xref: /netbsd/external/gpl3/gcc.old/dist/NEWS (revision 840b4d17)
1This file contains information about GCC releases which has been generated
2automatically from the online release notes.  It covers releases of GCC
3(and the former EGCS project) since EGCS 1.0, on the line of development
4that led to GCC 3. For information on GCC 2.8.1 and older releases of GCC 2,
5see ONEWS.
6
7======================================================================
8http://gcc.gnu.org/gcc-10/index.html
9
10                             GCC 10 Release Series
11
12   July, 7, 2023
13
14   The [1]GNU project and the GCC developers are pleased to announce the
15   release of GCC 10.5.
16
17   This release is a bug-fix release, containing fixes for regressions in
18   GCC 10.4 relative to previous releases of GCC.
19
20Release History
21
22   GCC 10.5
23          July 7, 2023 ([2]changes, [3]documentation)
24
25   GCC 10.4
26          June 28, 2022 ([4]changes, [5]documentation)
27
28   GCC 10.3
29          April 8, 2021 ([6]changes, [7]documentation)
30
31   GCC 10.2
32          July 23, 2020 ([8]changes, [9]documentation)
33
34   GCC 10.1
35          May 7, 2020 ([10]changes, [11]documentation)
36
37References and Acknowledgements
38
39   GCC used to stand for the GNU C Compiler, but since the compiler
40   supports several other languages aside from C, it now stands for the
41   GNU Compiler Collection.
42
43   A list of [12]successful builds is updated as new information becomes
44   available.
45
46   The GCC developers would like to thank the numerous people that have
47   contributed new features, improvements, bug fixes, and other changes as
48   well as test results to GCC. This [13]amazing group of volunteers is
49   what makes GCC successful.
50
51   For additional information about GCC please refer to the [14]GCC
52   project web site or contact the [15]GCC development mailing list.
53
54   To obtain GCC please use [16]our mirror sites or [17]our version
55   control system.
56
57
58    For questions related to the use of GCC, please consult these web
59    pages and the [18]GCC manuals. If that fails, the
60    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
61    web pages and the development of GCC are welcome on our developer
62    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
63    archives.
64
65   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
66   distribution of this entire article is permitted in any medium,
67   provided this notice is preserved.
68
69   These pages are [23]maintained by the GCC team. Last modified
70   2023-07-07.
71
72References
73
74   1. http://www.gnu.org/
75   2. http://gcc.gnu.org/gcc-10/changes.html
76   3. http://gcc.gnu.org/onlinedocs/10.5.0/
77   4. http://gcc.gnu.org/gcc-10/changes.html
78   5. http://gcc.gnu.org/onlinedocs/10.4.0/
79   6. http://gcc.gnu.org/gcc-10/changes.html
80   7. http://gcc.gnu.org/onlinedocs/10.3.0/
81   8. http://gcc.gnu.org/gcc-10/changes.html
82   9. http://gcc.gnu.org/onlinedocs/10.2.0/
83  10. http://gcc.gnu.org/gcc-10/changes.html
84  11. http://gcc.gnu.org/onlinedocs/10.1.0/
85  12. http://gcc.gnu.org/gcc-10/buildstat.html
86  13. http://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Contributors.html
87  14. http://gcc.gnu.org/index.html
88  15. mailto:gcc@gcc.gnu.org
89  16. http://gcc.gnu.org/mirrors.html
90  17. http://gcc.gnu.org/git.html
91  18. https://gcc.gnu.org/onlinedocs/
92  19. mailto:gcc-help@gcc.gnu.org
93  20. mailto:gcc@gcc.gnu.org
94  21. https://gcc.gnu.org/lists.html
95  22. https://www.fsf.org/
96  23. https://gcc.gnu.org/about.html
97======================================================================
98http://gcc.gnu.org/gcc-10/changes.html
99
100                             GCC 10 Release Series
101                        Changes, New Features, and Fixes
102
103   This page is a "brief" summary of some of the huge number of
104   improvements in GCC 10. You may also want to check out our [1]Porting
105   to GCC 10 page and the [2]full GCC documentation.
106
107Caveats
108
109     * An ABI incompatibility between C++14 and C++17 has been fixed. On
110       some targets a class with a zero-sized subobject would be passed
111       incorrectly when compiled as C++17 or C++20. See the [3]C++ notes
112       below for more details.
113     * The deprecated Profile Mode and array_allocator extensions have
114       been removed from libstdc++.
115     * The non-standard std::__is_nullptr_t type trait is deprecated and
116       will be removed from libstdc++ in a future release. The standard
117       trait std::is_null_pointer should be instead.
118     * The minimum version of the [4]MPFR library required for building
119       GCC has been increased to version 3.1.0 (released 2011-10-03).
120     * The automatic template instantiation at link time (-frepo) has been
121       removed.
122     * The --param allow-store-data-races internal parameter has been
123       removed in favor of a new official option -fallow-store-data-races.
124       While default behavior is unchanged and the new option allows to
125       correctly maintain a per compilation unit setting across link-time
126       optimization, alteration of the default via --param
127       allow-store-data-races will now be diagnosed and build systems have
128       to be adjusted accordingly.
129     * Offloading to Heterogeneous System Architecture Intermediate
130       Language (HSAIL) has been deprecated and will likely be removed in
131       a future release.
132     * The type of the std::iterator base class of
133       std::istreambuf_iterator was changed in C++98 mode to be consistent
134       with C++11 and later standards. See the [5]libstdc++ notes below
135       for more details.
136
137General Improvements
138
139     * New built-in functions:
140          + The [6]__has_builtin built-in preprocessor operator can be
141            used to query support for built-in functions provided by GCC
142            and other compilers that support it.
143          + __builtin_roundeven for the corresponding function from
144            ISO/IEC TS 18661.
145     * New command-line options:
146          + [7]-fallocation-dce removes unneeded pairs of new and delete
147            operators.
148          + [8]-fprofile-partial-training can now be used to inform the
149            compiler that code paths not covered by the training run
150            should not be optimized for size.
151          + [9]-fprofile-reproducible controls level of reproducibility of
152            profile gathered by [10]-fprofile-generate. This makes it
153            possible to rebuild program with same outcome which is useful,
154            for example, for distribution packages.
155          + [11]-fprofile-prefix-path can be used in combination with
156            -fprofile-generate=profile_dir and -fprofile-use=profile_dir
157            to inform GCC where the base directory of build source tree is
158            in case it differs between instrumentation and optimized
159            builds.
160          + [12]-fanalyzer enables a new static analysis pass and
161            associated warnings. This pass performs a time-consuming
162            exploration of paths through the code in the hope of detecting
163            various common errors, such as double-free bugs. This option
164            should be regarded as experimental in this release. In
165            particular, analysis of non-C code is unlikely to work.
166     * Inter-procedural optimization improvements:
167          + The inter-procedural scalar replacement of aggregates
168            (IPA-SRA) pass was re-implemented to work at link-time and can
169            now also remove computing and returning unused return values.
170          + [13]-finline-functions is now enabled at -O2 and was retuned
171            for better code size versus runtime performance trade-offs.
172            Inliner heuristics was also significantly sped up to avoid
173            negative impact to -flto -O2 compile times.
174          + Inliner heuristics and function cloning can now use
175            value-range information to predict effectivity of individual
176            transformations.
177          + During link-time optimization the C++ One Definition Rule is
178            used to increase precision of type based alias analysis.
179     * Link-time optimization improvements:
180          + A new binary [14]lto-dump has been added. It dumps various
181            information about LTO bytecode object files.
182          + The parallel phase of the LTO can automatically detect a
183            running make's jobserver or fall back to number of available
184            cores.
185          + The LTO bytecode can be compressed with the [15]zstd
186            algorithm. The configure script automatically detects zstd
187            support.
188          + Most --param values can now be specified at translation unit
189            granularity. This includes all parameters controlling the
190            inliner and other inter-procedural optimizations. Unlike
191            earlier releases, GCC 10 will ignore parameters controlling
192            optimizations specified at link-time and apply parameters
193            specified at compile-time in the same manner as done for
194            optimization flags.
195     * Profile driven optimization improvements:
196          + Profile maintenance during compilation and hot/cold code
197            partitioning have been improved.
198          + Using [16]-fprofile-values, an instrumented binary can track
199            multiple values (up to 4) for e.g. indirect calls and provide
200            more precise profile information.
201
202New Languages and Language-Specific Improvements
203
204     * Version 2.6 of the [17]OpenACC specification is now supported by
205       the C, C++ and Fortran compilers. See the [18]implementation status
206       section on the OpenACC wiki page and the [19]run-time library
207       documentation for further information.
208     * GCC 10 adds a number of newly implemented [20]OpenMP 5.0 features
209       such as conditional lastprivate clause, scan and loop directives,
210       order(concurrent) and use_device_addr clauses support, if clause on
211       simd construct, and partial support for the declare variant
212       directive, getting closer to full support of the OpenMP 5.0
213       standard.
214     * OpenMP and OpenACC now support [21]offloading to AMD Radeon (GCN)
215       GPUs; supported are the third-generation Fiji (fiji) and the
216       fifth-generation VEGA 10/VEGA 20 (gfx900 or gfx906).
217
218  C family
219
220     * New attributes:
221          + The access function and type attribute has been added to
222            describe how a function accesses objects passed to it by
223            pointer or reference, and to associate such arguments with
224            integer arguments denoting the objects' sizes. The attribute
225            is used to enable the detection of invalid accesses by
226            user-defined functions, such as those diagnosed by
227            -Wstringop-overflow.
228          + The symver attribute can be used to bind symbols to specific
229            version nodes on ELF platforms. This is preferred to using
230            inline assembly with GNU as symver directive because the
231            latter is not compatible with link-time optimizations.
232     * New warnings:
233          + [22]-Wstring-compare, enabled by -Wextra, warns about equality
234            and inequality expressions between zero and the result of a
235            call to either strcmp and strncmp that evaluate to a constant
236            as a result of the length of one argument being greater than
237            the size of the array pointed to by the other.
238          + [23]-Wzero-length-bounds, enabled by -Warray-bounds, warns
239            about accesses to elements of zero-length arrays that might
240            overlap other members of the same object.
241     * Enhancements to existing warnings:
242          + [24]-Warray-bounds detects more out-of-bounds accesses to
243            member arrays as well as accesses to elements of zero-length
244            arrays.
245          + [25]-Wformat-overflow makes full use of string length
246            information computed by the strlen optimization pass.
247          + [26]-Wrestrict detects overlapping accesses to dynamically
248            allocated objects.
249          + [27]-Wreturn-local-addr diagnoses more instances of return
250            statements returning addresses of automatic variables.
251          + [28]-Wstringop-overflow detects more out-of-bounds stores to
252            member arrays including zero-length arrays, dynamically
253            allocated objects and variable length arrays, as well as more
254            instances of reads of unterminated character arrays by string
255            built-in functions. The warning also detects out-of-bounds
256            accesses by calls to user-defined functions declared with the
257            new attribute access.
258          + [29]-Warith-conversion re-enables warnings from -Wconversion,
259            -Wfloat-conversion, and -Wsign-conversion that are now off by
260            default for an expression where the result of an arithmetic
261            operation will not fit in the target type due to promotion,
262            but the operands of the expression do fit in the target type.
263     * Extended characters in identifiers may now be specified directly in
264       the input encoding (UTF-8, by default), in addition to the UCN
265       syntax (\uNNNN or \UNNNNNNNN) that is already supported:
266
267static const int p = 3;
268int get_na�ve_pi() {
269  return p;
270}
271
272  C
273
274     * Several new features from the upcoming C2X revision of the ISO C
275       standard are supported with -std=c2x and -std=gnu2x. Some of these
276       features are also supported as extensions when compiling for older
277       language versions. In addition to the features listed, some
278       features previously supported as extensions and now added to the C
279       standard are enabled by default in C2X mode and not diagnosed with
280       -std=c2x -Wpedantic.
281          + The [[]] attribute syntax is supported, as in C++. Existing
282            attributes can be used with this syntax in forms such as
283            [[gnu::const]]. The standard attributes [[deprecated]],
284            [[fallthrough]] and [[maybe_unused]] are supported.
285          + UTF-8 character constants using the u8'' syntax are supported.
286          + <float.h> defines macros FLT_NORM_MAX, DBL_NORM_MAX and
287            LDBL_NORM_MAX.
288          + When decimal floating-point arithmetic is supported, <float.h>
289            defines macros DEC32_TRUE_MIN, DEC64_TRUE_MIN and
290            DEC128_TRUE_MIN, in addition to the macros that were
291            previously only defined if __STDC_WANT_DEC_FP__ was defined
292            before including <float.h>.
293          + In C2X mode, empty parentheses in a function definition give
294            that function a type with a prototype for subsequent calls;
295            other old-style function definitions are diagnosed by default
296            in C2X mode.
297          + The strftime format checking supports the %OB and %Ob formats.
298          + In C2X mode, -fno-fp-int-builtin-inexact is enabled by
299            default.
300     * GCC now defaults to -fno-common. As a result, global variable
301       accesses are more efficient on various targets. In C, global
302       variables with multiple tentative definitions now result in linker
303       errors. With -fcommon such definitions are silently merged during
304       linking.
305
306  C++
307
308     * Several C++20 features have been implemented:
309          + Concepts, including P0734R0, P0857R0, P1084R2, P1141R2,
310            P0848R3, P1616R1, P1452R2
311          + P1668R1, Permitting Unevaluated inline-assembly in constexpr
312            Functions
313          + P1161R3, Deprecate a[b,c]
314          + P0848R3, Conditionally Trivial Special Member Functions
315          + P1091R3, Extending structured bindings
316          + P1143R2, Adding the constinit keyword
317          + P1152R4, Deprecating volatile
318          + P0388R4, Permit conversions to arrays of unknown bound
319          + P0784R7, constexpr new
320          + P1301R4, [[nodiscard("with reason")]]
321          + P1814R0, class template argument deduction for alias templates
322          + P1816R0, class template argument deduction for aggregates
323          + P0960R3, Parenthesized initialization of aggregates
324          + P1331R2, Allow trivial default initialization in constexpr
325            contexts
326          + P1327R1, Allowing dynamic_cast and polymorphic typeid in
327            constexpr contexts
328          + P0912R5, Coroutines (requires -fcoroutines)
329     * Several C++ Defect Reports have been resolved, e.g.:
330          + DR 1560, lvalue-to-rvalue conversion in ?:
331          + DR 1813, __is_standard_layout for a class with repeated bases
332          + DR 2094, volatile scalars are trivially copyable,
333          + DR 2096, constraints on literal unions
334          + DR 2413, typename in conversion-function-ids
335          + DR 2352, Similar types and reference binding
336          + DR 1601, Promotion of enumeration with fixed underlying type
337          + DR 330, Qualification conversions and pointers to arrays of
338            pointers
339          + DR 1307, Overload resolution based on size of array
340            initializer-list
341          + DR 1710, Missing template keyword in class-or-decltype
342     * New warnings:
343          + [30]-Wmismatched-tags, disabled by default, warns about
344            declarations of structs, classes, and class templates and
345            their specializations with a class-key that does not match
346            either the definition or the first declaration if no
347            definition is provided. The option is provided to ease
348            portability to Windows-based compilers.
349          + [31]-Wredundant-tags, disabled by default, warns about
350            redundant class-key and enum-key in contexts where the key can
351            be eliminated without causing an syntactic ambiguity.
352     * G++ can now detect modifying constant objects in constexpr
353       evaluation (which is undefined behavior).
354     * G++ no longer emits bogus -Wsign-conversion warnings with explicit
355       casts.
356     * Narrowing is now detected in more contexts (e.g., case values).
357     * Memory consumption of the compiler has been reduced in constexpr
358       evaluation.
359     * The noexcept-specifier is now properly treated as a complete-class
360       context as per [class.mem].
361     * The attribute deprecated can now be used on namespaces too.
362     * The ABI of passing and returning certain C++ classes by value
363       changed on several targets in GCC 10, including [32]AArch64,
364       [33]ARM, [34]PowerPC ELFv2, [35]S/390 and [36]Itanium. These
365       changes affect classes with a zero-sized subobject (an empty base
366       class, or data member with the [[no_unique_address]] attribute)
367       where all other non-static data members have the same type (this is
368       called a "homogeneous aggregate" in some ABI specifications, or if
369       there is only one such member, a "single element"). In -std=c++17
370       and -std=c++20 modes, classes with an empty base class were not
371       considered to have a single element or to be a homogeneous
372       aggregate, and so could be passed differently (in the wrong
373       registers or at the wrong stack address). This could make code
374       compiled with -std=c++17 and -std=c++14 ABI incompatible. This has
375       been corrected and the empty bases are ignored in those ABI
376       decisions, so functions compiled with -std=c++14 and -std=c++17 are
377       now ABI compatible again. Example: struct empty {}; struct S :
378       empty { float f; }; void f(S);. Similarly, in classes containing
379       non-static data members with empty class types using the C++20
380       [[no_unique_address]] attribute, those members weren't ignored in
381       the ABI argument passing decisions as they should be. Both of these
382       ABI changes are now diagnosed with -Wpsabi.
383
384    Runtime Library (libstdc++)
385
386     * Improved experimental C++2a support, including:
387          + Library concepts in <concepts> and <iterator>.
388          + Constrained algorithms in <ranges>, <algorithm>, and <memory>
389            (thanks to Patrick Palka).
390          + New algorithms shift_left and shift_right (thanks to Patrick
391            Palka).
392          + std::span (thanks to JeanHeyd Meneide).
393          + Three-way comparisons in <compare> and throughout the library.
394          + Constexpr support in <algorithm> and elsewhere (thanks to
395            Edward Smith-Rowland).
396          + <stop_token> and std::jthread (thanks to Thomas Rodgers).
397          + std::atomic_ref and std::atomic<floating point>.
398          + Integer comparison functions (cmp_equal, cmp_less etc.).
399          + std::ssize, std::to_array.
400          + std::construct_at, std::destroy, constexpr std::allocator.
401          + Mathematical constants in <numbers>.
402     * Support for RDSEED in std::random_device.
403     * Reduced header dependencies, leading to faster compilation for some
404       code.
405     * The std::iterator base class of std::istreambuf_iterator was
406       changed in C++98 mode to be consistent with C++11 and later
407       standards. This is expected to have no noticeable effect except in
408       the unlikely case of a class which has potentially overlapping
409       subobjects of type std::istreambuf_iterator<C> and another iterator
410       type with a std::iterator<input_iterator_tag, C, ...> base class.
411       The layout of such a type might change when compiled as C++98.
412       [37]Bug 92285 has more details and concrete examples.
413
414  D
415
416     * Support for static foreach has been implemented.
417     * Aliases can now be created directly from any __traits that return
418       symbols or tuples. Previously, an AliasSeq was necessary in order
419       to alias their return.
420     * It is now possible to detect the language ABI specified for a
421       struct, class, or interface using __traits(getLinkage, ...).
422     * Support for core.math.toPrec intrinsics has been added. These
423       intrinsics guarantee the rounding to specific floating-point
424       precisions at specified points in the code.
425     * Support for pragma(inline) has been implemented. Previously the
426       pragma was recognized, but had no effect on the compilation.
427     * Optional parentheses in asm operands are deprecated and will be
428       removed in a future release.
429     * All content imported files are now included in the make dependency
430       list when compiling with -M.
431     * Compiler recognized attributes provided by the gcc.attribute module
432       will now take effect when applied to function prototypes as well as
433       when applied to full function declarations.
434     * Added a --enable-libphobos-checking configure option to control
435       whether run-time checks are compiled into the D runtime library.
436     * Added a --with-libphobos-druntime-only configure option to indicate
437       whether to build only the core D runtime library, or both the core
438       and standard libraries into libphobos.
439
440  Fortran
441
442     * use_device_addr of version 5.0 of the [38]OpenMP specification is
443       now supported. Note that otherwise OpenMP 4.5 is partially
444       supported in the Fortran compiler; the largest missing item is
445       structure element mapping.
446     * The default buffer size for I/O using unformatted files has been
447       increased to 1048576. The buffer size for can now be set at runtime
448       via the environment variables GFORTRAN_FORMATTED_BUFFER_SIZE and
449       GFORTRAN_UNFORMATTED_BUFFER_SIZE for formatted and unformatted
450       files, respectively.
451     * Mismatches between actual and dummy argument lists in a single file
452       are now rejected with an error. Use the new option
453       -fallow-argument-mismatch to turn these errors into warnings; this
454       option is implied with -std=legacy. -Wargument-mismatch has been
455       removed.
456     * The handling of a BOZ literal constant has been reworked to provide
457       better conformance to the Fortran 2008 and 2018 standards. In these
458       Fortran standards, a BOZ literal constant is a typeless and
459       kindless entity. As a part of the rework, documented and
460       undocumented extensions to the Fortran standard now emit errors
461       during compilation. Some of these extensions are permitted with the
462       -fallow-invalid-boz option, which degrades the error to a warning
463       and the code is compiled as with older gfortran.
464     * At any optimization level except-Os, gfortran now uses inline
465       packing for arguments instead of calling a library routine. If the
466       source contains a large number of arguments that need to be
467       repacked, code size or time for compilation can become excessive.
468       If that is the case, -fno-inline-arg-packing can be used to disable
469       inline argument packing.
470     * Legacy extensions:
471          + For formatted input/output, if the explicit widths after the
472            data-edit descriptors I, F and G have been omitted, default
473            widths are used.
474          + A blank format item at the end of a format specification, i.e.
475            nothing following the final comma, is allowed. Use the option
476            -fdec-blank-format-item; this option is implied with -fdec.
477          + The existing support for AUTOMATIC and STATIC attributes has
478            been extended to allow variables with the AUTOMATIC attribute
479            to be used in EQUIVALENCE statements. Use -fdec-static; this
480            option is implied by -fdec.
481          + Allow character literals in assignments and DATA statements
482            for numeric (INTEGER, REAL, or COMPLEX) or LOGICAL variables.
483            Use the option -fdec-char-conversions; this option is implied
484            with -fdec.
485          + DEC comparisons, i.e. allow Hollerith constants to be used in
486            comparisons with INTEGER, REAL, COMPLEX and CHARACTER
487            expressions. Use the option -fdec.
488     * Character type names in errors and warnings now include len in
489       addition to kind; * is used for assumed length. The kind is omitted
490       if it is the default kind. Examples: CHARACTER(12), CHARACTER(6,4).
491     * CO_BROADCAST now supports derived type variables including objects
492       with allocatable components. In this case, the optional arguments
493       STAT= and ERRMSG= are currently ignored.
494     * The handling of module and submodule names has been reworked to
495       allow the full 63-character length mandated by the standard.
496       Previously symbol names were truncated if the combined length of
497       module, submodule, and function name exceeded 126 characters. This
498       change therefore breaks the ABI, but only for cases where this 126
499       character limit was exceeded.
500
501  Go
502
503     * GCC 10 provides a complete implementation of the Go 1.14.6 user
504       packages.
505
506libgccjit
507
508     * The libgccjit API gained four new entry points:
509          + [39]gcc_jit_version_major, [40]gcc_jit_version_minor, and
510            [41]gcc_jit_version_patchlevel for programmatically checking
511            the libgccjit version from client code, and
512          + [42]gcc_jit_context_new_bitfield
513
514New Targets and Target Specific Improvements
515
516  AArch64 & arm
517
518     * The AArch64 and arm ports now support condition flag output
519       constraints in inline assembly, as indicated by the
520       __GCC_ASM_FLAG_OUTPUTS__. On arm this feature is only available for
521       A32 and T32 targets. Please refer to the documentation for more
522       details.
523
524  AArch64
525
526     * There have been several improvements related to the Scalable Vector
527       Extension (SVE):
528          + The SVE ACLE types and intrinsics are now supported. They can
529            be accessed using the header file arm_sve.h.
530          + It is now possible to create fixed-length SVE types using the
531            arm_sve_vector_bits attribute. For example:
532#if __ARM_FEATURE_SVE_BITS==512
533typedef svint32_t vec512 __attribute__((arm_sve_vector_bits(512)));
534typedef svbool_t pred512 __attribute__((arm_sve_vector_bits(512)));
535#endif
536          + -mlow-precision-div, -mlow-precision-sqrt and
537            -mlow-precision-recip-sqrt now work for SVE.
538          + -msve-vector-bits=128 now generates vector-length-specific
539            code for little-endian targets. It continues to generate
540            vector-length-agnostic code for big-endian targets, just as
541            previous releases did for all targets.
542          + The vectorizer is now able to use extending loads and
543            truncating stores, including gather loads and scatter stores.
544          + The vectorizer now compares the cost of vectorizing with SVE
545            and vectorizing with Advanced SIMD and tries to pick the best
546            one. Previously it would always use SVE if possible.
547          + If a vector loop uses Advanced SIMD rather than SVE, the
548            vectorizer now considers using SVE to vectorize the left-over
549            elements (the "scalar tail" or "epilog").
550          + Besides these specific points, there have been many general
551            improvements to the way that the vectorizer uses SVE.
552     * The -mbranch-protection=pac-ret option now accepts the optional
553       argument +b-key extension to perform return address signing with
554       the B-key instead of the A-key.
555     * The option -moutline-atomics has been added to aid deployment of
556       the Large System Extensions (LSE) on GNU/Linux systems built with a
557       baseline architecture targeting Armv8-A. When the option is
558       specified code is emitted to detect the presence of LSE
559       instructions at runtime and use them for standard atomic
560       operations. For more information please refer to the documentation.
561     * The Transactional Memory Extension is now supported through ACLE
562       intrinsics. It can be enabled through the +tme option extension
563       (for example, -march=armv8.5-a+tme).
564     * A number of features from Armv8.5-A are now supported through ACLE
565       intrinsics. These include:
566          + The random number instructions that can be enabled through the
567            (already present in GCC 9.1) +rng option extension.
568          + Floating-point intrinsics to round to integer instructions
569            from Armv8.5-A when targeting -march=armv8.5-a or later.
570          + Memory Tagging Extension intrinsics enabled through the
571            +memtag option extension.
572     * Similarly, the following Armv8.6-A features are now supported
573       through ACLE intrinsics:
574          + The bfloat16 extension. This extension is enabled
575            automatically when Armv8.6-A is selected (such as by
576            -march=armv8.6-a). It can also be enabled for Armv8.2-A and
577            later using the +bf16 option extension.
578          + The Matrix Multiply extension. This extension is split into
579            three parts, one for each supported data type:
580               o Support for 8-bit integer matrix multiply instructions.
581                 This extension is enabled automatically when Armv8.6-A is
582                 selected. It can also be enabled for Armv8.2-A and later
583                 using the +i8mm option extension.
584               o Support for 32-bit floating-point matrix multiply
585                 instructions. This extension can be enabled using the
586                 +f32mm option extension, which also has the effect of
587                 enabling SVE.
588               o Support for 64-bit floating-point matrix multiply
589                 instructions. This extension can be enabled using the
590                 +f64mm option extension, which likewise has the effect of
591                 enabling SVE.
592     * SVE2 is now supported through ACLE intrinsics and (to a limited
593       extent) through autovectorization. It can be enabled through the
594       +sve2 option extension (for example, -march=armv8.5-a+sve2).
595       Additional extensions can be enabled through +sve2-sm4, +sve2-aes,
596       +sve2-sha3 and +sve2-bitperm.
597     * Support has been added for the following processors (GCC
598       identifiers in parentheses):
599          + Arm Cortex-A77 (cortex-a77).
600          + Arm Cortex-A76AE (cortex-a76ae).
601          + Arm Cortex-A65 (cortex-a65).
602          + Arm Cortex-A65AE (cortex-a65ae).
603          + Arm Cortex-A34 (cortex-a34).
604          + Marvell ThunderX3 (thunderx3t110).
605       The GCC identifiers can be used as arguments to the -mcpu or -mtune
606       options, for example: -mcpu=cortex-a77 or -mtune=cortex-a65ae or as
607       arguments to the equivalent target attributes and pragmas.
608
609  arm
610
611     * Support for the FDPIC ABI has been added. It uses 64-bit function
612       descriptors to represent pointers to functions, and enables code
613       sharing on MMU-less systems. The corresponding target triple is
614       arm-uclinuxfdpiceabi, and the C library is uclibc-ng.
615     * Support has been added for the Arm EABI on NetBSD through the
616       arm*-*-netbsdelf-*eabi* triplet.
617     * The handling of 64-bit integer operations has been significantly
618       reworked and improved leading to improved performance and reduced
619       stack usage when using 64-bit integral data types. The option
620       -mneon-for-64bits is now deprecated and will be removed in a future
621       release.
622     * Support has been added for the following processors (GCC
623       identifiers in parentheses):
624          + Arm Cortex-A77 (cortex-a77).
625          + Arm Cortex-A76AE (cortex-a76ae).
626          + Arm Cortex-M35P (cortex-m35p).
627          + Arm Cortex-M55 (cortex-m55).
628       The GCC identifiers can be used as arguments to the -mcpu or -mtune
629       options, for example: -mcpu=cortex-a77 or -mtune=cortex-m35p.
630     * Support has been extended for the ACLE [43]data-processing
631       intrinsics to include 32-bit SIMD, saturating arithmetic, 16-bit
632       multiplication and other related intrinsics aimed at DSP algorithm
633       optimization.
634     * Support for -mpure-code in Thumb-1 (v6m) has been added: this
635       M-profile feature is no longer restricted to targets with MOVT. For
636       example, -mcpu=cortex-m0 now supports this option.
637     * Support for the [44]Armv8.1-M Mainline Architecture has been added.
638          + Armv8.1-M Mainline can be enabled by using the
639            -march=armv8.1-m.main command-line option.
640     * Support for the [45]MVE beta ACLE intrinsics has been added. These
641       intrinsics can be enabled by including the arm_mve.h header file
642       and passing the +mve or +mve.fp option extensions (for example:
643       -march=armv8.1-m.main+mve).
644     * Support for the Custom Datapath Extension beta ACLE [46]intrinsics
645       has been added.
646     * Support for Armv8.1-M Mainline Security Extensions architecture has
647       been added. The -mcmse option, when used in combination with an
648       Armv8.1-M Mainline architecture (for example: -march=armv8.1-m.main
649       -mcmse), now leads to the generation of improved code sequences
650       when changing security states.
651
652  AMD Radeon (GCN)
653
654     * Code generation and in particular vectorization support have been
655       much improved.
656
657  ARC
658
659     * The interrupt service routine functions save all used registers,
660       including extension registers and auxiliary registers used by Zero
661       Overhead Loops.
662     * Improve code size by using multiple short instructions instead of a
663       single long mov or ior instruction when its long immediate constant
664       is known.
665     * Fix usage of the accumulator register for ARC600.
666     * Fix issues with uncached attribute.
667     * Remove -mq-class option.
668     * Improve 64-bit integer addition and subtraction operations.
669
670  AVR
671
672     * Support for the XMEGA-like devices
673
674     ATtiny202, ATtiny204, ATtiny402, ATtiny404, ATtiny406, ATtiny804,
675     ATtiny806, ATtiny807, ATtiny1604, ATtiny1606, ATtiny1607, ATmega808,
676     ATmega809, ATmega1608, ATmega1609, ATmega3208, ATmega3209,
677     ATmega4808, ATmega4809
678       has been added.
679     * A new command-line option -nodevicespecs has been added. It allows
680       to provide a custom device-specs file by means of
681
682     avr-gcc -nodevicespecs -specs=my-spec-file <options>
683       and without the need to provide options -B and -mmcu=. See [47]AVR
684       command-line options for details. This feature is also available in
685       GCC 9.3+ and GCC 8.4+.
686     * New command-line options -mdouble=[32,64] and -mlong-double=[32,64]
687       have been added. They allow to choose the size (in bits) of the
688       double and long double types, respectively. Whether or not the
689       mentioned layouts are available, whether the options act as a
690       multilib option, and the default for either option are controlled
691       by the new [48]AVR configure options --with-double= and
692       --with-long-double=.
693     * A new configure option --with-libf7= has been added. It controls to
694       which level avr-libgcc provides 64-bit floating point support by
695       means of [49]Libf7.
696     * A new configure option --with-double-comparison= has been added.
697       It's unlikely you need to set this option by hand.
698
699  IA-32/x86-64
700
701     * Support to expand __builtin_roundeven into the appropriate SSE 4.1
702       instruction has been added.
703     * New ISA extension support for Intel ENQCMD was added to GCC. ENQCMD
704       intrinsics are available via the -menqcmd compiler switch.
705     * GCC now supports the Intel CPU named Cooperlake through
706       -march=cooperlake. The switch enables the AVX512BF16 ISA
707       extensions.
708     * GCC now supports the Intel CPU named Tigerlake through
709       -march=tigerlake. The switch enables the MOVDIRI MOVDIR64B
710       AVX512VP2INTERSECT ISA extensions.
711
712  MIPS
713
714     * The mips*-*-linux* targets now mark object files with appropriate
715       GNU-stack note, facilitating use of non-executable stack hardening
716       on GNU/Linux. The soft-float targets have this feature enabled by
717       default, while for hard-float targets it is required for GCC to be
718       configured with --with-glibc-version=2.31 against glibc 2.31 or
719       later.
720
721  PowerPC / PowerPC64 / RS6000
722
723     * Many vector builtins have been listed as deprecated in the
724       [50]64-Bit ELF V2 ABI Specification for quite a number of years.
725       The vector builtins listed in Tables A.8 through A.10 are now
726       deprecated for GCC 10, and will likely be removed from support in
727       GCC 11. Note that this does not result in any loss of function.
728       These deprecated builtins generally provide somewhat nonsensical
729       argument lists (for example, mixing signed, unsigned, and bool
730       vector arguments arbitrarily), or are duplicate builtins that are
731       inconsistent with the expected naming scheme. We expect that this
732       will be unlikely to affect much if any code, and any required code
733       changes will be trivial.
734
735  PRU
736
737     * A new back end targeting TI PRU I/O processors has been contributed
738       to GCC.
739
740  RISC-V
741
742     * The riscv*-*-* targets now require GNU binutils version 2.30 or
743       later, to support new assembly instructions produced by GCC.
744
745  V850
746
747     * The ABI for V850 nested functions has been changed. Previously the
748       V850 port used %r20 for the static chain pointer, now the port uses
749       %r19. This corrects a long standing latent bug in the v850 port
750       where a call to a nested function would unexpectedly change the
751       value in %r20.
752
753Operating Systems
754
755Improvements for plugin authors
756
757     * GCC diagnostics can now have a chain of events associated with
758       them, describing a path through the code that triggers the problem.
759       These can be printed by the diagnostics subsystem in various ways,
760       controlled by the [51]-fdiagnostics-path-format option, or captured
761       in JSON form via [52]-fdiagnostics-format=json.
762     * GCC diagnostics can now be associated with [53]CWE weakness
763       identifiers, which will appear on the standard error stream, and in
764       the JSON output from [54]-fdiagnostics-format=json.
765
766Other significant improvements
767
768     * To allow inline expansion of both memcpy and memmove, the existing
769       movmem instruction patterns used for non-overlapping memory copies
770       have been renamed to cpymem. The movmem name is now used for
771       overlapping memory moves, consistent with the library functions
772       memcpy and memmove.
773     * For many releases, when GCC emits a warning it prints the option
774       controlling that warning. As of GCC 10, that option text is now a
775       clickable hyperlink for the documentation of that option (assuming
776       a [55]sufficiently capable terminal). This behavior can be
777       controlled via a new [56]-fdiagnostics-urls option (along with
778       various environment variables and heuristics documented with that
779       option).
780
781GCC 10.1
782
783   This is the [57]list of problem reports (PRs) from GCC's bug tracking
784   system that are known to be fixed in the 10.1 release. This list might
785   not be complete (that is, it is possible that some PRs that have been
786   fixed are not listed here).
787
788GCC 10.2
789
790   This is the [58]list of problem reports (PRs) from GCC's bug tracking
791   system that are known to be fixed in the 10.2 release. This list might
792   not be complete (that is, it is possible that some PRs that have been
793   fixed are not listed here).
794
795GCC 10.3
796
797   This is the [59]list of problem reports (PRs) from GCC's bug tracking
798   system that are known to be fixed in the 10.3 release. This list might
799   not be complete (that is, it is possible that some PRs that have been
800   fixed are not listed here).
801
802  Target Specific Changes
803
804    AArch64
805
806     * A bug with the Random Number intrinsics in the arm_acle.h header
807       that resulted in an incorrect status result being returned has been
808       fixed.
809     * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune
810       options are -mcpu=a64fx and -mtune=a64fx respectively. In
811       particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and
812       tunes the code for the A64FX. This includes tuning the SVE code,
813       although by default the code is still length-agnostic and so works
814       for all SVE implementations. Adding -msve-vector-bits=512 makes the
815       code specific to 512-bit SVE.
816
817  x86-64
818
819     * GCC 10.3 supports AMD CPUs based on the znver3 core via
820       -march=znver3.
821
822GCC 10.4
823
824   This is the [60]list of problem reports (PRs) from GCC's bug tracking
825   system that are known to be fixed in the 10.4 release. This list might
826   not be complete (that is, it is possible that some PRs that have been
827   fixed are not listed here).
828
829  Target Specific Changes
830
831    x86-64
832
833     * The x86-64 ABI of passing and returning structures with a 64-bit
834       integer vector changed in GCC 10.1 when MMX is disabled. Disabling
835       MMX no longer changes how they are passed nor returned. This ABI
836       change is now diagnosed with -Wpsabi.
837
838GCC 10.5
839
840   This is the [61]list of problem reports (PRs) from GCC's bug tracking
841   system that are known to be fixed in the 10.5 release. This list might
842   not be complete (that is, it is possible that some PRs that have been
843   fixed are not listed here).
844
845
846    For questions related to the use of GCC, please consult these web
847    pages and the [62]GCC manuals. If that fails, the
848    [63]gcc-help@gcc.gnu.org mailing list might help. Comments on these
849    web pages and the development of GCC are welcome on our developer
850    list at [64]gcc@gcc.gnu.org. All of [65]our lists have public
851    archives.
852
853   Copyright (C) [66]Free Software Foundation, Inc. Verbatim copying and
854   distribution of this entire article is permitted in any medium,
855   provided this notice is preserved.
856
857   These pages are [67]maintained by the GCC team. Last modified
858   2023-07-07.
859
860References
861
862   1. http://gcc.gnu.org/gcc-10/porting_to.html
863   2. http://gcc.gnu.org/onlinedocs/index.html#current
864   3. http://gcc.gnu.org/gcc-10/changes.html#empty_base
865   4. https://www.mpfr.org/
866   5. http://gcc.gnu.org/gcc-10/changes.html#iterator_base
867   6. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/cpp/_005f_005fhas_005fbuiltin.html#g_t_005f_005fhas_005fbuiltin
868   7. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fno-allocation-dce
869   8. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-partial-training
870   9. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-reproducible
871  10. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-generate
872  11. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-prefix-path
873  12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Static-Analyzer-Options.html
874  13. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-finline-functions
875  14. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/lto-dump.html
876  15. https://facebook.github.io/zstd/
877  16. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-values
878  17. https://www.openacc.org/
879  18. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-10
880  19. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/libgomp/#toc-Enabling-OpenACC-1
881  20. https://www.openmp.org/specifications/
882  21. https://gcc.gnu.org/wiki/Offloading
883  22. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstring-compare
884  23. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wzero-length-bounds
885  24. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warray-bounds
886  25. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wformat-overflow
887  26. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wrestrict
888  27. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wreturn-local-addr
889  28. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstringop-overflow
890  29. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warith-conversion
891  30. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-tags
892  31. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-tags
893  32. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94383
894  33. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94711
895  34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94707
896  35. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94704
897  36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94706
898  37. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92285
899  38. https://www.openmp.org/specifications/
900  39. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_major
901  40. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_minor
902  41. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_patchlevel
903  42. https://gcc.gnu.org/onlinedocs/jit/topics/types.html#c.gcc_jit_context_new_bitfield
904  43. https://developer.arm.com/documentation/101028/0009/Data-processing-intrinsics
905  44. https://developer.arm.com/Architectures/M-Profile%20Architecture
906  45. https://developer.arm.com/architectures/instruction-sets/intrinsics/
907  46. https://developer.arm.com/documentation/101028/0010/Custom-Datapath-Extension
908  47. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/AVR-Options.html#index-nodevicespecs
909  48. https://gcc.gnu.org/install/configure.html#avr
910  49. https://gcc.gnu.org/wiki/avr-gcc#Libf7
911  50. https://openpowerfoundation.org/?resource_lib=64-bit-elf-v2-abi-specification-power-architecture
912  51. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-path-format
913  52. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
914  53. https://cwe.mitre.org/
915  54. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
916  55. https://gist.github.com/egmontkob/eb114294efbcd5adb1944c9f3cb5feda
917  56. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-urls
918  57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.0
919  58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.2
920  59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.3
921  60. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.4
922  61. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.5
923  62. https://gcc.gnu.org/onlinedocs/
924  63. mailto:gcc-help@gcc.gnu.org
925  64. mailto:gcc@gcc.gnu.org
926  65. https://gcc.gnu.org/lists.html
927  66. https://www.fsf.org/
928  67. https://gcc.gnu.org/about.html
929======================================================================
930http://gcc.gnu.org/gcc-9/index.html
931
932                              GCC 9 Release Series
933
934   (This release series is no longer supported.)
935
936   May 27, 2022
937
938   The [1]GNU project and the GCC developers are pleased to announce the
939   release of GCC 9.5.
940
941   This release is a bug-fix release, containing fixes for regressions in
942   GCC 9.4 relative to previous releases of GCC.
943
944Release History
945
946   GCC 9.5
947          May 27, 2022 ([2]changes, [3]documentation)
948
949   GCC 9.4
950          June 1, 2021 ([4]changes, [5]documentation)
951
952   GCC 9.3
953          Mar 12, 2020 ([6]changes, [7]documentation)
954
955   GCC 9.2
956          Aug 12, 2019 ([8]changes, [9]documentation)
957
958   GCC 9.1
959          May 3, 2019 ([10]changes, [11]documentation)
960
961References and Acknowledgements
962
963   GCC used to stand for the GNU C Compiler, but since the compiler
964   supports several other languages aside from C, it now stands for the
965   GNU Compiler Collection.
966
967   A list of [12]successful builds is updated as new information becomes
968   available.
969
970   The GCC developers would like to thank the numerous people that have
971   contributed new features, improvements, bug fixes, and other changes as
972   well as test results to GCC. This [13]amazing group of volunteers is
973   what makes GCC successful.
974
975   For additional information about GCC please refer to the [14]GCC
976   project web site or contact the [15]GCC development mailing list.
977
978   To obtain GCC please use [16]our mirror sites or [17]our version
979   control system.
980
981
982    For questions related to the use of GCC, please consult these web
983    pages and the [18]GCC manuals. If that fails, the
984    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
985    web pages and the development of GCC are welcome on our developer
986    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
987    archives.
988
989   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
990   distribution of this entire article is permitted in any medium,
991   provided this notice is preserved.
992
993   These pages are [23]maintained by the GCC team. Last modified
994   2022-10-26.
995
996References
997
998   1. http://www.gnu.org/
999   2. http://gcc.gnu.org/gcc-9/changes.html
1000   3. http://gcc.gnu.org/onlinedocs/9.5.0/
1001   4. http://gcc.gnu.org/gcc-9/changes.html
1002   5. http://gcc.gnu.org/onlinedocs/9.4.0/
1003   6. http://gcc.gnu.org/gcc-9/changes.html
1004   7. http://gcc.gnu.org/onlinedocs/9.3.0/
1005   8. http://gcc.gnu.org/gcc-9/changes.html
1006   9. http://gcc.gnu.org/onlinedocs/9.2.0/
1007  10. http://gcc.gnu.org/gcc-9/changes.html
1008  11. http://gcc.gnu.org/onlinedocs/9.1.0/
1009  12. http://gcc.gnu.org/gcc-9/buildstat.html
1010  13. http://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Contributors.html
1011  14. http://gcc.gnu.org/index.html
1012  15. mailto:gcc@gcc.gnu.org
1013  16. http://gcc.gnu.org/mirrors.html
1014  17. http://gcc.gnu.org/git.html
1015  18. https://gcc.gnu.org/onlinedocs/
1016  19. mailto:gcc-help@gcc.gnu.org
1017  20. mailto:gcc@gcc.gnu.org
1018  21. https://gcc.gnu.org/lists.html
1019  22. https://www.fsf.org/
1020  23. https://gcc.gnu.org/about.html
1021======================================================================
1022http://gcc.gnu.org/gcc-9/changes.html
1023
1024                              GCC 9 Release Series
1025                        Changes, New Features, and Fixes
1026
1027   This page is a "brief" summary of some of the huge number of
1028   improvements in GCC 9. You may also want to check out our [1]Porting to
1029   GCC 9 page and the [2]full GCC documentation.
1030
1031Caveats
1032
1033     * On Arm targets (arm*-*-*), [3]a bug in the implementation of the
1034       procedure call standard (AAPCS) in the GCC 6, 7 and 8 releases has
1035       been fixed: a structure containing a bit-field based on a 64-bit
1036       integral type and where no other element in a structure required
1037       64-bit alignment could be passed incorrectly to functions. This is
1038       an ABI change. If the option -Wpsabi is enabled (on by default) the
1039       compiler will emit a diagnostic note for code that might be
1040       affected.
1041     * Support for a number of older systems and recently unmaintained or
1042       untested target ports of GCC has been declared obsolete in GCC 9.
1043       Unless there is activity to revive them, the next release of GCC
1044       will have their sources permanently removed.
1045       The following ports for individual systems on particular
1046       architectures have been obsoleted:
1047          + Solaris 10 (*-*-solaris2.10). Details can be found in the
1048            [4]announcement.
1049          + Cell Broadband Engine SPU (spu*-*-*). Details can be found in
1050            the [5]announcement.
1051     * A change to the C++ std::rotate algorithm in GCC 9.1.0 can cause
1052       ABI incompatibilities with object files compiled with other
1053       versions of GCC. If the std::rotate algorithm is called with an
1054       empty range then it might cause a divide-by-zero error (as a SIGFPE
1055       signal) and crash. The change has been reverted for GCC 9.2.0 and
1056       future releases. For more details see [6]Bug 90920. The problem can
1057       be avoided by recompiling any objects that might call std::rotate
1058       with an empty range, so that the GCC 9.1.0 definition of
1059       std::rotate is not used.
1060     * The automatic template instantiation at link time ([7]-frepo) has
1061       been deprecated and will be removed in a future release.
1062     * The --with-default-libstdcxx-abi=gcc4-compatible configure option
1063       is broken in the 9.1 and 9.2 releases, producing a shared library
1064       with missing symbols (see [8]Bug 90361). As a workaround, configure
1065       without that option and build GCC as normal, then edit the
1066       installed <bits/c++config.h> headers to define the
1067       _GLIBCXX_USE_CXX11_ABI macro to 0.
1068
1069General Improvements
1070
1071   The following GCC command line options have been introduced or
1072   improved.
1073     * All command line options that take a byte-size argument accept
1074       64-bit integers as well as standard SI and IEC suffixes such as kb
1075       and KiB, MB and MiB, or GB and GiB denoting the corresponding
1076       multiples of bytes. See [9]Invoking GCC for more.
1077     * A new option [10]-flive-patching=[inline-only-static|inline-clone]
1078       generates code suitable for live patching. At the same time it
1079       provides multiple-level control over IPA optimizations. See the
1080       user guide for more details.
1081     * A new option, --completion, has been added to provide more fine
1082       option completion in a shell. It is intended to be used by
1083       Bash-completion.
1084     * GCC's diagnostics now print source code with a left margin showing
1085       line numbers, configurable with
1086       [11]-fno-diagnostics-show-line-numbers.
1087       GCC's diagnostics can also now label regions of the source code to
1088       show pertinent information, such as the types within an expression.
1089$ g++ t.cc
1090t.cc: In function 'int test(const shape&, const shape&)':
1091t.cc:15:4: error: no match for 'operator+' (operand types are 'boxed_value<doubl
1092e>' and 'boxed_value<double>')
1093   14 |   return (width(s1) * height(s1)
1094      |           ~~~~~~~~~~~~~~~~~~~~~~
1095      |                     |
1096      |                     boxed_value<[...]>
1097   15 |    + width(s2) * height(s2));
1098      |    ^ ~~~~~~~~~~~~~~~~~~~~~~
1099      |                |
1100      |                boxed_value<[...]>
1101
1102       These labels can be disabled via [12]-fno-diagnostics-show-labels.
1103     * A new option [13]-fdiagnostics-format=json has been introduced for
1104       emitting diagnostics in a machine-readable format.
1105     * The alignment-related options [14]-falign-functions,
1106       [15]-falign-labels, [16]-falign-loops, and [17]-falign-jumps
1107       received support for a secondary alignment (e.g.
1108       -falign-loops=n:m:n2:m2).
1109     * New pair of profiling options ([18]-fprofile-filter-files and
1110       [19]-fprofile-exclude-files) has been added. The options help to
1111       filter which source files are instrumented.
1112     * AddressSanitizer generates more compact redzones for automatic
1113       variables. That helps to reduce memory footprint of a sanitized
1114       binary.
1115     * Numerous improvements have been made to the output of
1116       [20]-fopt-info.
1117       Messages are now prefixed with optimized, missed, or note, rather
1118       than the old behavior of all being prefixed with note.
1119       The output from -fopt-info can now contain information on inlining
1120       decisions:
1121$ g++ -c inline.cc -O2 -fopt-info-inline-all
1122inline.cc:24:11: note: Considering inline candidate void foreach(T, T, void (*)(
1123E)) [with T = char**; E = char*]/2.
1124inline.cc:24:11: optimized:  Inlining void foreach(T, T, void (*)(E)) [with T =
1125char**; E = char*]/2 into int main(int, char**)/1.
1126inline.cc:19:12: missed:   not inlinable: void inline_me(char*)/0 -> int std::pu
1127ts(const char*)/3, function body not available
1128inline.cc:13:8: optimized:  Inlined void inline_me(char*)/4 into int main(int, c
1129har**)/1 which now has time 127.363637 and size 11, net change of +0.
1130Unit growth for small function inlining: 16->16 (0%)
1131
1132Inlined 2 calls, eliminated 1 functions
1133
1134
1135       The output from the vectorizer has been rationalized so that failed
1136       attempts to vectorize a loop are displayed in the form
1137    [LOOP-LOCATION]: couldn't vectorize this loop
1138    [PROBLEM-LOCATION]: because of [REASON]
1139
1140       rather than an exhaustive log of all decisions made by the
1141       vectorizer. For example:
1142$ gcc -c v.c -O3 -fopt-info-all-vec
1143v.c:7:3: missed: couldn't vectorize loop
1144v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" :  :  : "me
1145mory");
1146v.c:3:6: note: vectorized 0 loops in function.
1147v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" :  :  : "me
1148mory");
1149
1150       The old behavior can be obtained via a new -internals suboption of
1151       -fopt-info.
1152     * A new option, [21]-fsave-optimization-record has been added, which
1153       writes a SRCFILE.opt-record.json.gz file describing the
1154       optimization decisions made by GCC. This is similar to the output
1155       of -fopt-info, but with additional metadata such as the inlining
1156       chain, and profile information (if available).
1157     * Inter-procedural propagation of stack alignment can now be
1158       controlled by [22]-fipa-stack-alignment.
1159     * Propagation of addressability, readonly and writeonly flags on
1160       static variables can now be controlled by
1161       [23]-fipa-reference-addressable.
1162
1163   The following built-in functions have been introduced.
1164     * [24]__builtin_expect_with_probability to provide branch prediction
1165       probability hints to the optimizer.
1166     * [25]__builtin_has_attribute determines whether a function, type, or
1167       variable has been declared with some attribute.
1168     * [26]__builtin_speculation_safe_value can be used to help mitigate
1169       against unsafe speculative execution.
1170
1171   The following attributes have been introduced.
1172     * The [27]copy function attribute has been added. The attribute can
1173       also be applied to type definitions and to variable declarations.
1174
1175   A large number of improvements to code generation have been made,
1176   including but not limited to the following.
1177     * Switch expansion has been improved by using a different strategy
1178       (jump table, bit test, decision tree) for a subset of switch cases.
1179     * A linear function expression defined as a switch statement can be
1180       transformed by [28]-ftree-switch-conversion. For example:
1181
1182int
1183foo (int how)
1184{
1185  switch (how) {
1186    case 2: how = 205; break;
1187    case 3: how = 305; break;
1188    case 4: how = 405; break;
1189    case 5: how = 505; break;
1190    case 6: how = 605; break;
1191  }
1192  return how;
1193}
1194
1195       can be transformed into 100 * how + 5 (for values defined in the
1196       switch statement).
1197     * Inter-procedural optimization improvements:
1198          + Inliner defaults was tuned to better suits modern C++
1199            codebases especially when built with link time optimizations.
1200            New parameters max-inline-insns-small, max-inline-insns-size,
1201            uninlined-function-insns, uninlined-function-time,
1202            uninlined-thunk-insns, and uninlined-thunk-time were added.
1203          + Hot/cold partitioning is now more precise and aggressive.
1204          + Improved scalability for very large translation units
1205            (especially when link-time optimizing large programs).
1206     * Profile driven optimization improvements:
1207          + [29]-fprofile-use now enables [30]-fversion-loops-for-strides,
1208            [31]-floop-interchange, [32]-floop-unroll-and-jam,
1209            [33]-ftree-loop-distribution.
1210          + Streaming of counter histograms was removed. This reduces the
1211            size of profile files. Histogram is computed on the fly with
1212            link-time optimization. Parameter hot-bb-count-ws-permille was
1213            reduced from 999 to 990 to account for more precise
1214            histograms.
1215     * Link-time optimization improvements:
1216          + Types are now simplified prior streaming resulting in
1217            significant reductions of the LTO object files, link-time
1218            memory use, and improvements of link-time parallelism.
1219          + Default number of partitions (--param lto-partitions) was
1220            increased from 32 to 128 enabling effective use of CPUs with
1221            more than 32 hyperthreads. --param
1222            lto-max-streaming-parallelism can now be used to control
1223            number of streaming processes.
1224          + Warnings on C++ One Decl Rule violations (-Wodr) are now more
1225            informative and produce fewer redundant results.
1226       Overall compile time of Firefox 66 and LibreOffice 6.2.3 on an
1227       8-core machine was reduced by about 5% compared to GCC 8.3, and the
1228       size of LTO object files by 7%. LTO link-time improves by 11% on an
1229       8-core machine and scales significantly better for more parallel
1230       build environments. The serial stage of the link-time optimization
1231       is 28% faster consuming 20% less memory. The parallel stage now
1232       scales to up to 128 partitions rather than 32 and reduces memory
1233       use for every worker by 30%.
1234
1235   The following improvements to the gcov command-line utility have been
1236   made.
1237     * The gcov tool received a new option [34]--use-hotness-colors (-q)
1238       that can provide perf-like coloring of hot functions.
1239     * The gcov tool has changed its intermediate format to a new JSON
1240       format.
1241
1242New Languages and Language specific improvements
1243
1244   [35]OpenACC support in C, C++, and Fortran continues to be maintained
1245   and improved. Most of the OpenACC 2.5 specification is implemented. See
1246   the [36]implementation status section on the OpenACC wiki page for
1247   further information.
1248
1249  C family
1250
1251     * Version 5.0 of the [37]OpenMP specification is now partially
1252       supported in the C and C++ compilers. For details which features of
1253       OpenMP 5.0 are and which are not supported in the GCC 9 release see
1254       [38]this mail.
1255     * New extensions:
1256          + [39]__builtin_convertvector built-in for vector conversions
1257            has been added.
1258     * New warnings:
1259          + [40]-Waddress-of-packed-member, enabled by default, warns
1260            about an unaligned pointer value from the address of a packed
1261            member of a struct or union.
1262     * Enhancements to existing warnings:
1263          + [41]-Warray-bounds detects more instances of out-of-bounds
1264            indices.
1265          + [42]-Wattribute-alias also detects attribute mismatches
1266            between alias declarations and their targets, in addition to
1267            mismatches between their types.
1268          + [43]-Wformat-overflow and [44]-Wformat-truncation have been
1269            extended to all formatted input/output functions (where
1270            applicable) and enhanced to detect a subset of instances of
1271            reading past the end of unterminated constant character arrays
1272            in %s directives.
1273          + [45]-Wmissing-attributes detects instances of missing function
1274            attributes on declarations of aliases and weak references.
1275          + [46]-Wstringop-truncation also detects a subset of instances
1276            of reading past the end of unterminated constant character
1277            arrays,
1278     * If a macro is used with the wrong argument count, the C and C++
1279       front ends now show the definition of that macro via a note.
1280     * The spelling corrector now considers transposed letters, and the
1281       threshold for similarity has been tightened, to avoid nonsensical
1282       suggestions.
1283
1284  C
1285
1286     * There is now experimental support for -std=c2x, to select support
1287       for the upcoming C2X revision of the ISO C standard. This standard
1288       is in the early stages of development and the only feature
1289       supported in GCC 9 is _Static_assert with a single argument
1290       (support for _Static_assert with two arguments was added in C11 and
1291       GCC 4.6). There are also new options -std=gnu2x, for C2X with GNU
1292       extensions, and -Wc11-c2x-compat, to warn for uses of features
1293       added in C2X (such warnings are also enabled by use of -Wpedantic
1294       if not using -std=c2x or -std=gnu2x).
1295     * New warnings:
1296          + [47]-Wabsolute-value warns for calls to standard functions
1297            that compute the absolute value of an argument when a more
1298            appropriate standard function is available. For example,
1299            calling abs(3.14) triggers the warning because the appropriate
1300            function to call to compute the absolute value of a double
1301            argument is fabs. The option also triggers warnings when the
1302            argument in a call to such a function has an unsigned type.
1303            This warning can be suppressed with an explicit type cast and
1304            it is also enabled by -Wextra.
1305
1306  C++
1307
1308     * New warnings:
1309          + [48]-Wdeprecated-copy, implied by -Wextra, warns about the
1310            C++11 deprecation of implicitly declared copy constructor and
1311            assignment operator if one of them is user-provided.
1312            -Wdeprecated-copy-dtor also warns if the destructor is
1313            user-provided, as specified in C++11.
1314          + [49]-Winit-list-lifetime, on by default, warns about uses of
1315            std::initializer_list that are likely to result in a dangling
1316            pointer, such as returning or assigning from a temporary list.
1317          + [50]-Wredundant-move, implied by -Wextra, warns about
1318            redundant calls to std::move.
1319          + [51]-Wpessimizing-move, implied by -Wall, warns when a call to
1320            std::move prevents copy elision.
1321          + [52]-Wclass-conversion, on by default, warns when a conversion
1322            function will never be called due to the type it converts to.
1323     * The C++ front end has experimental support for some of the upcoming
1324       C++2a draft features with the -std=c++2a or -std=gnu++2a flags,
1325       including range-based for statements with initializer, default
1326       constructible and assignable stateless lambdas, lambdas in
1327       unevaluated contexts, language support for empty data members,
1328       allowing pack expansion in lambda init-capture, likely and unlikely
1329       attributes, class types in non-type template parameters, allowing
1330       virtual function calls in constant expressions, explicit(bool),
1331       std::is_constant_evaluated, nested inline namespaces, etc. For a
1332       full list of new features, see [53]the C++ status page.
1333     * The C++ front end now preserves source locations for literals,
1334       id-expression, and mem-initializer for longer. For example it is
1335       now able to pin-point the pertinent locations for bad
1336       initializations such as these
1337$ g++ -c bad-inits.cc
1338bad-inits.cc:10:14: error: cannot convert 'json' to 'int' in initialization
1339   10 |   { 3, json::object },
1340      |        ~~~~~~^~~~~~
1341      |              |
1342      |              json
1343bad-inits.cc:14:31: error: initializer-string for array of chars is too long [-f
1344permissive]
1345   14 | char buffers[3][5] = { "red", "green", "blue" };
1346      |                               ^~~~~~~
1347bad-inits.cc: In constructor 'X::X()':
1348bad-inits.cc:17:13: error: invalid conversion from 'int' to 'void*' [-fpermissiv
1349e]
1350   17 |   X() : one(42), two(42), three(42)
1351      |             ^~
1352      |             |
1353      |             int
1354
1355       rather than emitting the error at the final closing parenthesis or
1356       brace.
1357     * Error-reporting of overload resolution has been special-cased to
1358       make the case of a single failed candidate easier to read. For
1359       example:
1360$ g++ param-type-mismatch.cc
1361param-type-mismatch.cc: In function 'int test(int, const char*, float)':
1362param-type-mismatch.cc:8:32: error: cannot convert 'const char*' to 'const char*
1363*'
1364    8 |   return foo::member_1 (first, second, third);
1365      |                                ^~~~~~
1366      |                                |
1367      |                                const char*
1368param-type-mismatch.cc:3:46: note:   initializing argument 2 of 'static int foo:
1369:member_1(int, const char**, float)'
1370    3 |   static int member_1 (int one, const char **two, float three);
1371      |                                 ~~~~~~~~~~~~~^~~
1372
1373       highlights both the problematic argument, and the parameter that it
1374       can't be converted to.
1375     * Diagnostics involving binary operators now use color to distinguish
1376       the two operands, and label them separately (as per the example of
1377       source labelling above).
1378     * Diagnostics involving function calls now highlight the pertinent
1379       parameter of the declaration in more places.
1380$ g++ bad-conversion.cc
1381bad-conversion.cc: In function 'void caller()':
1382bad-conversion.cc:9:14: error: cannot convert 'bool' to 'void*'
1383    9 |   callee (0, false, 2);
1384      |              ^~~~~
1385      |              |
1386      |              bool
1387bad-conversion.cc:3:19: note:   initializing argument 2 of 'void callee(int, voi
1388d*, int)'
1389    3 | void callee (int, void *, int)
1390      |                   ^~~~~~
1391
1392     * The C++ front end's implementation of [54]-Wformat now shows
1393       precise locations within string literals, and underlines the
1394       pertinent arguments at bogus call sites (the C front end has been
1395       doing this since GCC 7). For example:
1396$ g++ -c bad-printf.cc -Wall
1397bad-printf.cc: In function 'void print_field(const char*, float, long int, long
1398int)':
1399bad-printf.cc:6:17: warning: field width specifier '*' expects argument of type
1400'int', but argument 3 has type 'long int' [-Wformat=]
1401    6 |   printf ("%s: %*ld ", fieldname, column - width, value);
1402      |                ~^~~               ~~~~~~~~~~~~~~
1403      |                 |                        |
1404      |                 int                      long int
1405bad-printf.cc:6:19: warning: format '%ld' expects argument of type 'long int', b
1406ut argument 4 has type 'double' [-Wformat=]
1407    6 |   printf ("%s: %*ld ", fieldname, column - width, value);
1408      |                ~~~^                               ~~~~~
1409      |                   |                               |
1410      |                   long int                        double
1411      |                %*f
1412
1413     * The C++ front end has gained new fix-it hints for forgetting the
1414       return *this; needed by various C++ operators:
1415$ g++ -c operator.cc
1416operator.cc: In member function 'boxed_ptr& boxed_ptr::operator=(const boxed_ptr
1417&)':
1418operator.cc:7:3: warning: no return statement in function returning non-void [-W
1419return-type]
1420    6 |     m_ptr = other.m_ptr;
1421  +++ |+    return *this;
1422    7 |   }
1423      |   ^
1424
1425       for when the compiler needs a typename:
1426$ g++ -c template.cc
1427template.cc:3:3: error: need 'typename' before 'Traits::type' because 'Traits' i
1428s a dependent scope
1429    3 |   Traits::type type;
1430      |   ^~~~~~
1431      |   typename
1432
1433       when trying to use an accessor member as if it were a data member:
1434$ g++ -c fncall.cc
1435fncall.cc: In function 'void hangman(const mystring&)':
1436fncall.cc:12:11: error: invalid use of member function 'int mystring::get_length
1437() const' (did you forget the '()' ?)
1438   12 |   if (str.get_length > 0)
1439      |       ~~~~^~~~~~~~~~
1440      |                     ()
1441
1442       for C++11's scoped enums:
1443$ g++ -c enums.cc
1444enums.cc: In function 'void json::test(const json::value&)':
1445enums.cc:12:26: error: 'STRING' was not declared in this scope; did you mean 'js
1446on::kind::STRING'?
1447   12 |     if (v.get_kind () == STRING)
1448      |                          ^~~~~~
1449      |                          json::kind::STRING
1450enums.cc:3:44: note: 'json::kind::STRING' declared here
1451    3 |   enum class kind { OBJECT, ARRAY, NUMBER, STRING, TRUE, FALSE, NULL_ };
1452      |                                            ^~~~~~
1453
1454       and a tweak to integrate the suggestions about misspelled members
1455       with that for accessors:
1456$ g++ -c accessor-fixit.cc
1457accessor-fixit.cc: In function 'int test(t*)':
1458accessor-fixit.cc:17:15: error: 'class t' has no member named 'ratio'; did you m
1459ean 'int t::m_ratio'? (accessible via 'int t::get_ratio() const')
1460   17 |   return ptr->ratio;
1461      |               ^~~~~
1462      |               get_ratio()
1463
1464       In addition, various diagnostics in the C++ front-end have been
1465       streamlined by consolidating the suggestion into the initial error,
1466       rather than emitting a follow-up note:
1467$ g++ typo.cc
1468typo.cc:5:13: error: 'BUFSIZE' was not declared in this scope; did you mean 'BUF
1469_SIZE'?
1470    5 | uint8_t buf[BUFSIZE];
1471      |             ^~~~~~~
1472      |             BUF_SIZE
1473
1474    Runtime Library (libstdc++)
1475
1476     * Improved support for C++17, including:
1477          + The C++17 implementation is no longer experimental.
1478          + Parallel algorithms and <execution> (requires [55]Thread
1479            Building Blocks 2018 or newer).
1480          + <memory_resource>.
1481          + Using the types and functions in <filesystem> does not require
1482            linking with -lstdc++fs now.
1483     * Improved experimental support for C++2a, including:
1484          + Type traits std::remove_cvref, std::unwrap_reference,
1485            std::unwrap_decay_ref, std::is_nothrow_convertible, and
1486            std::type_identity.
1487          + Headers <bit> and <version>.
1488          + Uniform container erasure (std::erase_if).
1489          + contains member of maps and sets.
1490          + String prefix and suffix checking (starts_with, ends_with).
1491          + Functions std::midpoint and std::lerp for interpolation.
1492          + std::bind_front.
1493          + std::visit<R>.
1494          + std::assume_aligned.
1495          + Uses-allocator construction utilities.
1496          + std::pmr::polymorphic_allocator<std::byte>.
1497          + Library support for char8_t type.
1498          + Destroying delete.
1499          + std::is_constant_evaluated() function.
1500     * Support for opening file streams with wide character paths on
1501       Windows
1502     * Incomplete support for the C++17 Filesystem library and the
1503       Filesystem TS on Windows.
1504     * Incomplete, experimental support for the Networking TS.
1505
1506  D
1507
1508     * Support for the D programming language has been added to GCC,
1509       implementing version 2.076 of the language and run-time library.
1510
1511  Fortran
1512
1513     * Asynchronous I/O is now fully supported. The program needs to be
1514       linked against the pthreads library to use it, otherwise the I/O is
1515       done synchronously. For systems which do not support POSIX
1516       condition variables, such as AIX, all I/O is still done
1517       synchronously.
1518     * The BACK argument for MINLOC and MAXLOC has been implemented.
1519     * The FINDLOC intrinsic function has been implemented.
1520     * The IS_CONTIGUOUS intrinsic function has been implemented.
1521     * Direct access to the real and imaginary parts of a complex variable
1522       via c%re and c%im has been implemented.
1523     * Type parameter inquiry via str%len and a%kind has been implemented.
1524     * C descriptors and the ISO_Fortran_binding.h source file have been
1525       implemented.
1526     * The MAX and MIN intrinsics are no longer guaranteed to return any
1527       particular value in case one of the arguments is a NaN. Note that
1528       this conforms to the Fortran standard and to what other Fortran
1529       compilers do. If there is a need to handle that case in some
1530       specific way, one needs to explicitly check for NaN's before
1531       calling MAX or MIN, e.g. by using the IEEE_IS_NAN function from the
1532       intrinsic module IEEE_ARITHMETIC.
1533     * A new command-line option [56]-fdec-include, set also by the
1534       [57]-fdec option, has been added to increase compatibility with
1535       legacy code. With this option, an INCLUDE directive is also parsed
1536       as a statement, which allows the directive to be spread across
1537       multiple source lines with line continuations.
1538     * A new [58]BUILTIN directive, has been added. The purpose of the
1539       directive is to provide an API between the GCC compiler and the GNU
1540       C Library which would define vector implementations of math
1541       routines.
1542
1543  Go
1544
1545     * GCC 9 provides a complete implementation of the Go 1.12.2 user
1546       packages.
1547
1548libgccjit
1549
1550     * The libgccjit API gained a new entry point:
1551       [59]gcc_jit_context_add_driver_option.
1552
1553New Targets and Target Specific Improvements
1554
1555  AArch64 & Arm
1556
1557     * Support has been added for the following processors (GCC
1558       identifiers in parentheses):
1559          + Arm Cortex-A76 (cortex-a76).
1560          + Arm Cortex-A55/Cortex-A76 DynamIQ big.LITTLE
1561            (cortex-a76.cortex-a55).
1562          + Arm Neoverse N1 (neoverse-n1).
1563       The GCC identifiers can be used as arguments to the -mcpu or -mtune
1564       options, for example: -mcpu=cortex-a76 or
1565       -mtune=cortex-a76.cortex-a55 or as arguments to the equivalent
1566       target attributes and pragmas.
1567     * The Armv8.3-A complex number instructions are now supported via
1568       intrinsics when the option -march=armv8.3-a or equivalent is
1569       specified. For the half-precision floating-point variants of these
1570       instructions use the architecture extension flag +fp16, e.g.
1571       -march=armv8.3-a+fp16.
1572       The intrinsics are defined by the ACLE specification.
1573     * The Armv8.5-A architecture is now supported through the
1574       -march=armv8.5-a option.
1575     * The Armv8.5-A architecture also adds some security features that
1576       are optional to all older architecture versions. These are now
1577       supported and only affect the assembler.
1578          + Speculation Barrier instruction through the -march=armv8-a+sb
1579            option.
1580          + Execution and Data Prediction Restriction instructions through
1581            the -march=armv8-a+predres option.
1582          + Speculative Store Bypass Safe instruction through the
1583            -march=armv8-a+ssbs option. This does not require a compiler
1584            option for Arm and thus -march=armv8-a+ssbs is an
1585            AArch64-specific option.
1586
1587      AArch64 specific
1588
1589     * Support has been added for the Arm Neoverse E1 processor
1590       (-mcpu=neoverse-e1).
1591     * The AArch64 port now has support for stack clash protection using
1592       the [60]-fstack-clash-protection option. The probing interval/guard
1593       size can be set by using --param
1594       stack-clash-protection-guard-size=12|16. The value of this
1595       parameter must be in bytes represented as a power of two. The two
1596       supported values for this parameter are 12 (for a 4KiB size, 2^12)
1597       and 16 (for a 64KiB size, 2^16). The default value is 16 (64Kb) and
1598       can be changed at configure time using the flag
1599       --with-stack-clash-protection-guard-size=12|16.
1600     * The option -msign-return-address= has been deprecated. This has
1601       been replaced by the new -mbranch-protection= option. This new
1602       option can now be used to enable the return address signing as well
1603       as the new Branch Target Identification feature of Armv8.5-A
1604       architecture. For more information on the arguments accepted by
1605       this option, please refer to [61]AArch64-Options.
1606     * The following optional extensions to Armv8.5-A architecture are now
1607       supported and only affect the assembler.
1608          + Random Number Generation instructions through the
1609            -march=armv8.5-a+rng option.
1610          + Memory Tagging Extension through the -march=armv8.5-a+memtag
1611            option.
1612
1613      Arm specific
1614
1615     * Support for the deprecated Armv2 and Armv3 architectures and their
1616       variants has been removed. Their corresponding -march values and
1617       the -mcpu options that used these architectures have been removed.
1618     * Support for the Armv5 and Armv5E architectures (which have no known
1619       implementations) has been removed. Note that Armv5T, Armv5TE and
1620       Armv5TEJ architectures remain supported.
1621     * Corrected FPU configurations for Cortex-R7 and Cortex-R8 when using
1622       their respective -mcpu options.
1623
1624  AMD GCN
1625
1626     * A new back end targeting AMD GCN GPUs has been contributed to GCC.
1627       The implementation is currently limited to compiling
1628       single-threaded, stand-alone programs. Future versions will add
1629       support for offloading multi-threaded kernels via OpenMP and
1630       OpenACC. The following devices are supported (GCC identifiers in
1631       parentheses):
1632          + Fiji (fiji).
1633          + Vega 10 (gfx900).
1634
1635  ARC
1636
1637     * LRA is now on by default for the ARC target. This can be controlled
1638       by -mlra.
1639     * Add support for frame code-density and branch-and-index
1640       instructions.
1641
1642  C-SKY
1643
1644     * A new back end targeting C-SKY V2 processors has been contributed
1645       to GCC.
1646
1647  IA-32/x86-64
1648
1649     * Support of Intel MPX (Memory Protection Extensions) has been
1650       removed.
1651     * New ISA extension support for Intel PTWRITE was added to GCC.
1652       PTWRITE intrinsics are available via the -mptwrite compiler switch.
1653     * GCC now supports the Intel CPU named Cascade Lake with AVX512
1654       extensions through -march=cascadelake. The switch enables the
1655       following ISA extensions: AVX512F, AVX512VL, AVX512CD, AVX512BW,
1656       AVX512DQ, AVX512VNNI.
1657
1658  MIPS
1659
1660     * The Loongson loongson-mmi and loongson-ext extensions have been
1661       split from loongson3a:
1662          + loongson-mmi contains the Loongson MMI (MultiMedia extensions
1663            Instructions).
1664          + loongson-ext contains the Loongson EXT (EXTensions
1665            instructions).
1666     * The Loongson EXT2 (EXTensions R2 instructions) are now supported.
1667          + loongson-ext2 contains the Loongson EXT2 instructions.
1668       Command-line options-m[no-]loongson-mmi, -m[no-]loongson-ext, and
1669       -m[no-]loongson-ext2 enable or disable those extensions.
1670     * Support has been added for the following processors (GCC
1671       identifiers in parentheses):
1672          + Loongson 3A1000 (gs464) which enables loongson-mmi,
1673            loongson-ext by default.
1674          + Loongson 3A2000/3A3000 (gs464e) which enables loongson-mmi,
1675            loongson-ext, loongson-ext2 by default.
1676          + Loongson 2K1000 (gs264e) which enables loongson-ext,
1677            loongson-ext2, msa by default.
1678       The GCC identifiers can be used as arguments to the -mcpu and
1679       -mtune options (as in -mcpu=gs464 or -mtune=gs464e) or as arguments
1680       to the equivalent target attributes and pragmas.
1681
1682  OpenRISC
1683
1684     * A new back end targeting OpenRISC processors has been contributed
1685       to GCC.
1686
1687  S/390, System z, IBM z Systems
1688
1689     * Support for the arch13 architecture has been added. When using the
1690       -march=arch13 option, the compiler will generate code making use of
1691       the new instructions introduced with the vector enhancement
1692       facility 2 and the miscellaneous instruction extension facility 2.
1693       The -mtune=arch13 option enables arch13 specific instruction
1694       scheduling without making use of new instructions.
1695     * Builtins for the new vector instructions have been added and can be
1696       enabled using the -mzvector option.
1697     * Support for ESA architecture machines g5 and g6 is deprecated since
1698       GCC 6.1.0 and has been removed now.
1699     * When compiling with -march=z14 or higher GCC emits alignments hints
1700       on the vector load/store instructions (8 or 16 byte).
1701     * Functions now have a default alignment of 16 bytes. This helps with
1702       branch prediction effects.
1703     * -mfentry is now supported. As well as the mcount mechanism the
1704       __fentry__ is called before the function prologue. However, since
1705       just a single instruction is required to call __fentry__ the call
1706       sequence imposes a smaller overhead than mcount (4 instructions).
1707       The produced code is compatible only with newer glibc versions,
1708       which provide the __fentry__ symbol and do not clobber r0 when
1709       resolving lazily bound functions. -mfentry is only supported when
1710       generating 64 bit code and does not work with nested C functions.
1711     * The -mnop-mcount option can be used to emit NOP instructions
1712       instead of an mcount or fentry call stub.
1713     * With the -mrecord-mcount option a __mcount_loc section is generated
1714       containing pointers to each profiling call stub. This is useful for
1715       automatically patching in and out calls.
1716
1717Operating Systems
1718
1719  Solaris
1720
1721     * g++ now unconditionally enables large file support when compiling
1722       32-bit code.
1723     * Support for the AddressSanitizer and UndefinedBehaviorSanitizer has
1724       been merged from LLVM. For the moment, this only works for 32-bit
1725       code on both SPARC and x86.
1726     * An initial port of the D runtime library has been completed on
1727       Solaris 11/x86. It requires the use of GNU as. Solaris 11/SPARC
1728       support is still work-in-progress.
1729
1730  Windows
1731
1732     * A C++ Microsoft ABI bitfield layout bug, [62]PR87137 has been
1733       fixed. A non-field declaration could cause the current bitfield
1734       allocation unit to be completed, incorrectly placing a following
1735       bitfield into a new allocation unit. The Microsoft ABI is selected
1736       for:
1737          + Mingw targets
1738          + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields
1739            option is specified, or __attribute__((ms_struct)) is used
1740          + SuperH targets when the -mhitachi option is specified, or
1741            __attribute__((renesas)) is used
1742
1743Improvements for plugin authors
1744
1745     * GCC's diagnostic subsystem now has a way to logically group
1746       together related diagnostics, auto_diagnostic_group. Such
1747       diagnostics will be nested by the output of
1748       [63]-fdiagnostics-format=json.
1749     * GCC now has a set of [64]user experience guidelines for GCC, with
1750       information and advice on implementing new diagnostics.
1751
1752Other significant improvements
1753
1754     * GCC's internal "selftest" suite now runs for C++ as well as C (in
1755       debug builds of the compiler).
1756
1757GCC 9.1
1758
1759   This is the [65]list of problem reports (PRs) from GCC's bug tracking
1760   system that are known to be fixed in the 9.1 release. This list might
1761   not be complete (that is, it is possible that some PRs that have been
1762   fixed are not listed here).
1763
1764GCC 9.2
1765
1766   This is the [66]list of problem reports (PRs) from GCC's bug tracking
1767   system that are known to be fixed in the 9.2 release. This list might
1768   not be complete (that is, it is possible that some PRs that have been
1769   fixed are not listed here).
1770
1771GCC 9.3
1772
1773   This is the [67]list of problem reports (PRs) from GCC's bug tracking
1774   system that are known to be fixed in the 9.3 release. This list might
1775   not be complete (that is, it is possible that some PRs that have been
1776   fixed are not listed here).
1777
1778GCC 9.4
1779
1780   This is the [68]list of problem reports (PRs) from GCC's bug tracking
1781   system that are known to be fixed in the 9.4 release. This list might
1782   not be complete (that is, it is possible that some PRs that have been
1783   fixed are not listed here).
1784
1785  Target Specific Changes
1786
1787    AArch64
1788
1789     * The option -moutline-atomics has been added to aid deployment of
1790       the Large System Extensions (LSE) on GNU/Linux systems built with a
1791       baseline architecture targeting Armv8-A. When the option is
1792       specified code is emitted to detect the presence of LSE
1793       instructions at run time and use them for standard atomic
1794       operations. For more information please refer to the documentation.
1795     * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune
1796       options are -mcpu=a64fx and -mtune=a64fx respectively. In
1797       particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and
1798       tunes the code for the A64FX. This includes tuning the SVE code,
1799       although by default the code is still length-agnostic and so works
1800       for all SVE implementations. Adding -msve-vector-bits=512 makes the
1801       code specific to 512-bit SVE.
1802
1803GCC 9.5
1804
1805   This is the [69]list of problem reports (PRs) from GCC's bug tracking
1806   system that are known to be fixed in the 9.5 release. This list might
1807   not be complete (that is, it is possible that some PRs that have been
1808   fixed are not listed here).
1809
1810
1811    For questions related to the use of GCC, please consult these web
1812    pages and the [70]GCC manuals. If that fails, the
1813    [71]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1814    web pages and the development of GCC are welcome on our developer
1815    list at [72]gcc@gcc.gnu.org. All of [73]our lists have public
1816    archives.
1817
1818   Copyright (C) [74]Free Software Foundation, Inc. Verbatim copying and
1819   distribution of this entire article is permitted in any medium,
1820   provided this notice is preserved.
1821
1822   These pages are [75]maintained by the GCC team. Last modified
1823   2023-02-22.
1824
1825References
1826
1827   1. http://gcc.gnu.org/gcc-9/porting_to.html
1828   2. http://gcc.gnu.org/onlinedocs/index.html#current
1829   3. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88469
1830   4. https://gcc.gnu.org/ml/gcc/2018-10/msg00139.html
1831   5. https://gcc.gnu.org/ml/gcc/2019-04/msg00023.html
1832   6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90920
1833   7. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-frepo
1834   8. https://gcc.gnu.org/PR90361
1835   9. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-GCC.html#Invoking-GCC
1836  10. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-flive-patching
1837  11. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-line-numbers
1838  12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-labels
1839  13. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
1840  14. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-functions
1841  15. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-labels
1842  16. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-loops
1843  17. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-jumps
1844  18. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-filter-files
1845  19. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-exclude-files
1846  20. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fopt-info
1847  21. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fsave-optimization-record
1848  22. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-stack-alignment
1849  23. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-reference-addressable
1850  24. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fexpect_005fwith_005fprobability
1851  25. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fhas_005fattribute-1
1852  26. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fspeculation_005fsafe_005fvalue-1
1853  27. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Common-Function-Attributes.html#index-copy-function-attribute
1854  28. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-switch-conversion
1855  29. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fprofile-use
1856  30. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fversion-loops-for-strides
1857  31. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-interchange
1858  32. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-unroll-and-jam
1859  33. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-loop-distribution
1860  34. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-Gcov.html#Invoking-Gcov
1861  35. https://www.openacc.org/
1862  36. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-9
1863  37. https://www.openmp.org/specifications/
1864  38. https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00628.html
1865  39. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Vector-Extensions.html#index-_005f_005fbuiltin_005fconvertvector
1866  40. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Waddress-of-packed-member
1867  41. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Warray-bounds
1868  42. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wattribute-alias
1869  43. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-overflow
1870  44. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-truncation
1871  45. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wmissing-attributes
1872  46. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wstringop-truncation
1873  47. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wabsolute-value
1874  48. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-copy
1875  49. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Winit-list-lifetime
1876  50. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move
1877  51. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move
1878  52. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-conversion
1879  53. http://gcc.gnu.org/projects/cxx-status.html#cxx2a
1880  54. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat
1881  55. https://github.com/oneapi-src/oneTBB
1882  56. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec-include
1883  57. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec
1884  58. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BUILTIN-directive.html#BUILTIN-directive
1885  59. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/jit/topics/contexts.html#gcc_jit_context_add_driver_option
1886  60. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fstack-protector
1887  61. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/AArch64-Options.html#AArch64-Options
1888  62. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137
1889  63. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
1890  64. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gccint/User-Experience-Guidelines.html
1891  65. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.0
1892  66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.2
1893  67. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.3
1894  68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.4
1895  69. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.5
1896  70. https://gcc.gnu.org/onlinedocs/
1897  71. mailto:gcc-help@gcc.gnu.org
1898  72. mailto:gcc@gcc.gnu.org
1899  73. https://gcc.gnu.org/lists.html
1900  74. https://www.fsf.org/
1901  75. https://gcc.gnu.org/about.html
1902======================================================================
1903http://gcc.gnu.org/gcc-8/index.html
1904
1905                              GCC 8 Release Series
1906
1907   (This release series is no longer supported.)
1908
1909   May 14, 2021
1910
1911   The [1]GNU project and the GCC developers are pleased to announce the
1912   release of GCC 8.5.
1913
1914   This release is a bug-fix release, containing fixes for regressions in
1915   GCC 8.4 relative to previous releases of GCC.
1916
1917Release History
1918
1919   GCC 8.5
1920          May 14, 2021 ([2]changes, [3]documentation)
1921
1922   GCC 8.4
1923          Mar 4, 2020 ([4]changes, [5]documentation)
1924
1925   GCC 8.3
1926          Feb 22, 2019 ([6]changes, [7]documentation)
1927
1928   GCC 8.2
1929          Jul 26, 2018 ([8]changes, [9]documentation)
1930
1931   GCC 8.1
1932          May 2, 2018 ([10]changes, [11]documentation)
1933
1934References and Acknowledgements
1935
1936   GCC used to stand for the GNU C Compiler, but since the compiler
1937   supports several other languages aside from C, it now stands for the
1938   GNU Compiler Collection.
1939
1940   A list of [12]successful builds is updated as new information becomes
1941   available.
1942
1943   The GCC developers would like to thank the numerous people that have
1944   contributed new features, improvements, bug fixes, and other changes as
1945   well as test results to GCC. This [13]amazing group of volunteers is
1946   what makes GCC successful.
1947
1948   For additional information about GCC please refer to the [14]GCC
1949   project web site or contact the [15]GCC development mailing list.
1950
1951   To obtain GCC please use [16]our mirror sites or [17]our version
1952   control system.
1953
1954
1955    For questions related to the use of GCC, please consult these web
1956    pages and the [18]GCC manuals. If that fails, the
1957    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1958    web pages and the development of GCC are welcome on our developer
1959    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
1960    archives.
1961
1962   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
1963   distribution of this entire article is permitted in any medium,
1964   provided this notice is preserved.
1965
1966   These pages are [23]maintained by the GCC team. Last modified
1967   2022-10-26.
1968
1969References
1970
1971   1. http://www.gnu.org/
1972   2. http://gcc.gnu.org/gcc-8/changes.html
1973   3. http://gcc.gnu.org/onlinedocs/8.5.0/
1974   4. http://gcc.gnu.org/gcc-8/changes.html
1975   5. http://gcc.gnu.org/onlinedocs/8.4.0/
1976   6. http://gcc.gnu.org/gcc-8/changes.html
1977   7. http://gcc.gnu.org/onlinedocs/8.3.0/
1978   8. http://gcc.gnu.org/gcc-8/changes.html
1979   9. http://gcc.gnu.org/onlinedocs/8.2.0/
1980  10. http://gcc.gnu.org/gcc-8/changes.html
1981  11. http://gcc.gnu.org/onlinedocs/8.1.0/
1982  12. http://gcc.gnu.org/gcc-8/buildstat.html
1983  13. http://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html
1984  14. http://gcc.gnu.org/index.html
1985  15. mailto:gcc@gcc.gnu.org
1986  16. http://gcc.gnu.org/mirrors.html
1987  17. http://gcc.gnu.org/git.html
1988  18. https://gcc.gnu.org/onlinedocs/
1989  19. mailto:gcc-help@gcc.gnu.org
1990  20. mailto:gcc@gcc.gnu.org
1991  21. https://gcc.gnu.org/lists.html
1992  22. https://www.fsf.org/
1993  23. https://gcc.gnu.org/about.html
1994======================================================================
1995http://gcc.gnu.org/gcc-8/changes.html
1996
1997                              GCC 8 Release Series
1998                        Changes, New Features, and Fixes
1999
2000   This page is a "brief" summary of some of the huge number of
2001   improvements in GCC 8. You may also want to check out our [1]Porting to
2002   GCC 8 page and the [2]full GCC documentation.
2003
2004Caveats
2005
2006     * Support for the obsolete SDB/coff debug info format has been
2007       removed. The option -gcoff no longer does anything.
2008     * The Cilk+ extensions to the C and C++ languages have been removed.
2009     * The MPX extensions to the C and C++ languages have been deprecated
2010       and will be removed in a future release.
2011     * The extension allowing arithmetic on std::atomic<void*> and types
2012       like std::atomic<R(*)()> has been deprecated.
2013     * The non-standard C++0x std::copy_exception function was removed.
2014       std::make_exception_ptr should be used instead.
2015     * Support for the powerpc*-*-*spe* target ports which have been
2016       recently unmaintained and untested in GCC has been declared
2017       obsolete in GCC 8 as [3]announced. Unless there is activity to
2018       revive them, the next release of GCC will have their sources
2019       permanently removed.
2020
2021General Improvements
2022
2023     * Inter-procedural optimization improvements:
2024          + Reworked run-time estimation metrics leading to more realistic
2025            guesses driving inliner and cloning heuristics.
2026          + The ipa-pure-const pass is extended to propagate the malloc
2027            attribute, and the corresponding warning option
2028            -Wsuggest-attribute=malloc emits a diagnostic for functions
2029            which can be annotated with the malloc attribute.
2030     * Profile driven optimization improvements:
2031          + New infrastructure for representing profiles (both statically
2032            guessed and profile feedback) which allows propagation of
2033            additional information about the reliability of the profile.
2034          + A number of improvements in the profile updating code solving
2035            problems found by new verification code.
2036          + Static detection of code which is not executed in a valid run
2037            of the program. This includes paths which trigger undefined
2038            behavior as well as calls to functions declared with the cold
2039            attribute. Newly the noreturn attribute does not imply all
2040            effects of cold to differentiate between exit (which is
2041            noreturn) and abort (which is in addition not executed in
2042            valid runs).
2043          + -freorder-blocks-and-partition, a pass splitting function
2044            bodies into hot and cold regions, is now enabled by default at
2045            -O2 and higher for x86 and x86-64.
2046     * Link-time optimization improvements:
2047          + We have significantly improved debug information on ELF
2048            targets using DWARF by properly preserving language-specific
2049            information. This allows for example the libstdc++
2050            pretty-printers to work with LTO optimized executables.
2051     * A new option -fcf-protection=[full|branch|return|none] is
2052       introduced to perform code instrumentation to increase program
2053       security by checking that target addresses of control-flow transfer
2054       instructions (such as indirect function call, function return,
2055       indirect jump) are valid. Currently the instrumentation is
2056       supported on x86 GNU/Linux targets only. See the user guide for
2057       further information about the option syntax and section "New
2058       Targets and Target Specific Improvements" for IA-32/x86-64 for more
2059       details.
2060     * The -gcolumn-info option is now enabled by default. It includes
2061       column information in addition to just filenames and line numbers
2062       in DWARF debugging information.
2063     * The polyhedral-based loop nest optimization pass
2064       -floop-nest-optimize has been overhauled. It's still considered
2065       experimental and may not result in any runtime improvements.
2066     * Two new classical loop nest optimization passes have been added.
2067       -floop-unroll-and-jam performs outer loop unrolling and fusing of
2068       the inner loop copies. -floop-interchange exchanges loops in a loop
2069       nest to improve data locality. Both passes are enabled by default
2070       at -O3 and above.
2071     * The classic loop nest optimization pass -ftree-loop-distribution
2072       has been improved and enabled by default at -O3 and above. It
2073       supports loop nest distribution in some restricted scenarios; it
2074       also supports cancellable innermost loop distribution with loop
2075       versioning under run-time alias checks.
2076     * The new option -fstack-clash-protection causes the compiler to
2077       insert probes whenever stack space is allocated statically or
2078       dynamically to reliably detect stack overflows and thus mitigate
2079       the attack vector that relies on jumping over a stack guard page as
2080       provided by the operating system.
2081     * A new pragma GCC unroll has been implemented in the C family of
2082       languages, as well as Fortran and Ada, so as to make it possible
2083       for the user to have a finer-grained control over the loop
2084       unrolling optimization.
2085     * GCC has been enhanced to detect more instances of meaningless or
2086       mutually exclusive attribute specifications and handle such
2087       conflicts more consistently. Mutually exclusive attribute
2088       specifications are ignored with a warning regardless of whether
2089       they appear on the same declaration or on distinct declarations of
2090       the same entity. For example, because the noreturn attribute on the
2091       second declaration below is mutually exclusive with the malloc
2092       attribute on the first, it is ignored and a warning is issued.
2093>
2094      void* __attribute__ ((malloc)) f (unsigned);
2095      void* __attribute__ ((noreturn)) f (unsigned);
2096
2097      warning: ignoring attribute 'noreturn' because it conflicts with attribute
2098 'malloc' [-Wattributes]
2099     * The gcov tool can distinguish functions that begin on a same line
2100       in a source file. This can be a different template instantiation or
2101       a class constructor:
2102
2103File 'ins.C'
2104Lines executed:100.00% of 8
2105Creating 'ins.C.gcov'
2106
2107        -:    0:Source:ins.C
2108        -:    0:Graph:ins.gcno
2109        -:    0:Data:ins.gcda
2110        -:    0:Runs:1
2111        -:    0:Programs:1
2112        -:    1:template<class T>
2113        -:    2:class Foo
2114        -:    3:{
2115        -:    4: public:
2116        2:    5:   Foo(): b (1000) {}
2117------------------
2118Foo<char>::Foo():
2119        1:    5:   Foo(): b (1000) {}
2120------------------
2121Foo<int>::Foo():
2122        1:    5:   Foo(): b (1000) {}
2123------------------
2124        2:    6:   void inc () { b++; }
2125------------------
2126Foo<char>::inc():
2127        1:    6:   void inc () { b++; }
2128------------------
2129Foo<int>::inc():
2130        1:    6:   void inc () { b++; }
2131------------------
2132        -:    7:
2133        -:    8:  private:
2134        -:    9:   int b;
2135        -:   10:};
2136        -:   11:
2137        1:   12:int main(int argc, char **argv)
2138        -:   13:{
2139        1:   14:  Foo<int> a;
2140        1:   15:  Foo<char> b;
2141        -:   16:
2142        1:   17:  a.inc ();
2143        1:   18:  b.inc ();
2144        1:   19:}
2145
2146     * The gcov tool has more accurate numbers for execution of lines in a
2147       source file.
2148     * The gcov tool can use TERM colors to provide more readable output.
2149     * AddressSanitizer gained a new pair of sanitization options,
2150       -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which
2151       warn about subtraction (or comparison) of pointers that point to a
2152       different memory object:
2153
2154int
2155main ()
2156{
2157  /* Heap allocated memory.  */
2158  char *heap1 = (char *)__builtin_malloc (42);
2159  char *heap2 = (char *)__builtin_malloc (42);
2160  if (heap1 > heap2)
2161      return 1;
2162
2163  return 0;
2164}
2165
2166==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000
216700050
2168    #0 0x40070f in main /tmp/pointer-compare.c:7
2169    #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
2170    #2 0x400629 in _start (/tmp/a.out+0x400629)
2171
21720x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604
217300000003a)
2174allocated by thread T0 here:
2175    #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan
2176_malloc_linux.cc:86
2177    #1 0x4006ea in main /tmp/pointer-compare.c:5
2178    #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
2179
21800x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604
218100000007a)
2182allocated by thread T0 here:
2183    #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan
2184_malloc_linux.cc:86
2185    #1 0x4006f8 in main /tmp/pointer-compare.c:6
2186    #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
2187
2188SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main
2189
2190     * The store merging pass has been enhanced to handle bit-fields and
2191       not just constant stores, but also data copying from adjacent
2192       memory locations into other adjacent memory locations, including
2193       bitwise logical operations on the data. The pass can also handle
2194       byte swapping into memory locations.
2195     * The undefined behavior sanitizer gained two new options included in
2196       -fsanitize=undefined: -fsanitize=builtin which diagnoses at run
2197       time invalid arguments to __builtin_clz or __builtin_ctz prefixed
2198       builtins, and -fsanitize=pointer-overflow which performs cheap run
2199       time tests for pointer wrapping.
2200     * A new attribute no_sanitize can be applied to functions to instruct
2201       the compiler not to do sanitization of the options provided as
2202       arguments to the attribute. Acceptable values for no_sanitize match
2203       those acceptable by the -fsanitize command-line option.
2204
2205void __attribute__ ((no_sanitize ("alignment", "object-size")))
2206f () { /* Do something. */; }
2207
2208New Languages and Language specific improvements
2209
2210  Ada
2211
2212     * For its internal exception handling used on the host for error
2213       recovery in the front-end, the compiler now relies on the native
2214       exception handling mechanism of the host platform, which should be
2215       more efficient than the former mechanism.
2216
2217  BRIG (HSAIL)
2218
2219   In this release cycle, the focus for the BRIGFE was on stabilization
2220   and performance improvements. Also a couple of completely new features
2221   were added.
2222     * Improved support for function and module scope group segment
2223       variables. PRM specs define function and module scope group segment
2224       variables as an experimental feature. However, PRM test suite uses
2225       them. Now group segment is handled by separate book keeping of
2226       module scope and function (kernel) offsets. Each function has a
2227       "frame" in the group segment offset to which is given as an
2228       argument, similar to traditional call stack frame handling.
2229     * Reduce the number of type conversions due to the untyped HSAIL
2230       registers. Instead of always representing the HSAIL's untyped
2231       registers as unsigned int, the gccbrig now pre-analyzes the BRIG
2232       code and builds the register variables as a type used the most when
2233       storing or reading data to/from each register. This reduces the
2234       number of total casts which cannot be always optimized away.
2235     * Support for BRIG_KIND_NONE directives.
2236     * Made -O3 the default optimization level for BRIGFE.
2237     * Fixed illegal addresses generated from address expressions which
2238       refer only to offset 0.
2239     * Fixed a bug with reg+offset addressing on 32b segments. In 'large'
2240       mode, the offset is treated as 32-bit unless it's in global,
2241       read-only or kernarg address space.
2242     * Fixed a crash caused sometimes by calls with more than 4 arguments.
2243     * Fixed a mis-execution issue with kernels that have both unexpanded
2244       ID functions and calls to subfunctions.
2245     * Treat HSAIL barrier builtins as setjmp/longjump style functions to
2246       avoid illegal optimizations.
2247     * Ensure per WI copies of private variables are aligned correctly.
2248     * libhsail-rt: Assume the host runtime allocates the work group
2249       memory.
2250
2251  C family
2252
2253     * New command-line options have been added for the C and C++
2254       compilers:
2255          + [4]-Wmultistatement-macros warns about unsafe macros expanding
2256            to multiple statements used as a body of a statement such as
2257            if, else, while, switch, or for.
2258          + [5]-Wstringop-truncation warns for calls to bounded string
2259            manipulation functions such as strncat, strncpy, and stpncpy
2260            that might either truncate the copied string or leave the
2261            destination unchanged. For example, the following call to
2262            strncat is diagnosed because it appends just three of the four
2263            characters from the source string.
2264void append (char *buf, size_t bufsize)
2265{
2266    strncat (buf, ".txt", 3);
2267}
2268warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [-
2269Wstringop-truncation]
2270            Similarly, in the following example, the call to strncpy
2271            specifies the size of the destination buffer as the bound. If
2272            the length of the source string is equal to or greater than
2273            this size the result of the copy will not be NUL-terminated.
2274            Therefore, the call is also diagnosed. To avoid the warning,
2275            specify sizeof buf - 1 as the bound and set the last element
2276            of the buffer to NUL.
2277void copy (const char *s)
2278{
2279    char buf[80];
2280    strncpy (buf, s, sizeof buf);
2281    ...
2282}
2283warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca
2284tion]
2285            The -Wstringop-truncation option is included in -Wall.
2286            Note that due to GCC bug [6]82944, defining strncat, strncpy,
2287            or stpncpy as a macro in a system header as some
2288            implementations do, suppresses the warning.
2289          + [7]-Wif-not-aligned controls warnings issued in response to
2290            invalid uses of objects declared with attribute
2291            [8]warn_if_not_aligned.
2292            The -Wif-not-aligned option is included in -Wall.
2293          + [9]-Wmissing-attributes warns when a declaration of a function
2294            is missing one or more attributes that a related function is
2295            declared with and whose absence may adversely affect the
2296            correctness or efficiency of generated code. For example, in
2297            C++, the warning is issued when an explicit specialization of
2298            a primary template declared with attribute alloc_align,
2299            alloc_size, assume_aligned, format, format_arg, malloc, or
2300            nonnull is declared without it. Attributes deprecated, error,
2301            and warning suppress the warning.
2302            The -Wmissing-attributes option is included in -Wall.
2303          + [10]-Wpacked-not-aligned warns when a struct or union declared
2304            with attribute packed defines a member with an explicitly
2305            specified alignment greater than 1. Such a member will wind up
2306            under-aligned. For example, a warning will be issued for the
2307            definition of struct A in the following:
2308struct __attribute__ ((aligned (8)))
2309S8 { char a[8]; };
2310
2311struct __attribute__ ((packed)) A
2312{
2313    struct S8 s8;
2314};
2315warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned]
2316            The -Wpacked-not-aligned option is included in -Wall.
2317          + -Wcast-function-type warns when a function pointer is cast to
2318            an incompatible function pointer. This warning is enabled by
2319            -Wextra.
2320          + -Wsizeof-pointer-div warns for suspicious divisions of the
2321            size of a pointer by the size of the elements it points to,
2322            which looks like the usual way to compute the array size but
2323            won't work out correctly with pointers. This warning is
2324            enabled by -Wall.
2325          + -Wcast-align=strict warns whenever a pointer is cast such that
2326            the required alignment of the target is increased. For
2327            example, warn if a char * is cast to an int * regardless of
2328            the target machine.
2329          + -fprofile-abs-path creates absolute path names in the .gcno
2330            files. This allows gcov to find the correct sources in
2331            projects where compilations occur with different working
2332            directories.
2333     * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and
2334       signed integer overflow is now undefined by default at all
2335       optimization levels. Using -fsanitize=signed-integer-overflow is
2336       now the preferred way to audit code, -Wstrict-overflow is
2337       deprecated.
2338     * The [11]-Warray-bounds option has been improved to detect more
2339       instances of out-of-bounds array indices and pointer offsets. For
2340       example, negative or excessive indices into flexible array members
2341       and string literals are detected.
2342     * The [12]-Wrestrict option introduced in GCC 7 has been enhanced to
2343       detect many more instances of overlapping accesses to objects via
2344       restrict-qualified arguments to standard memory and string
2345       manipulation functions such as memcpy and strcpy. For example, the
2346       strcpy call in the function below attempts to truncate the string
2347       by replacing its initial characters with the last four. However,
2348       because the function writes the terminating NUL into a[4], the
2349       copies overlap and the call is diagnosed.
2350void f (void)
2351{
2352    char a[] = "abcd1234";
2353    strcpy (a, a + 4);
2354    ...
2355}
2356warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset
2357 4 [-Wrestrict]
2358       The -Wrestrict option is included in -Wall.
2359     * Several optimizer enhancements have enabled improvements to the
2360       [13]-Wformat-overflow and [14]-Wformat-truncation options. The
2361       warnings detect more instances of buffer overflow and truncation
2362       than in GCC 7 and are better at avoiding certain kinds of false
2363       positives.
2364     * When reporting mismatching argument types at a function call, the C
2365       and C++ compilers now underline both the argument and the pertinent
2366       parameter in the declaration.
2367$ gcc arg-type-mismatch.cc
2368arg-type-mismatch.cc: In function 'int caller(int, int, float)':
2369arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*'
2370 [-fpermissive]
2371   return callee(first, second, third);
2372                        ^~~~~~
2373arg-type-mismatch.cc:1:40: note:   initializing argument 2 of 'int callee(int, c
2374onst char*, float)'
2375 extern int callee(int one, const char *two, float three);
2376                            ~~~~~~~~~~~~^~~
2377
2378     * When reporting on unrecognized identifiers, the C and C++ compilers
2379       will now emit fix-it hints suggesting #include directives for
2380       various headers in the C and C++ standard libraries.
2381$ gcc incomplete.c
2382incomplete.c: In function 'test':
2383incomplete.c:3:10: error: 'NULL' undeclared (first use in this function)
2384   return NULL;
2385          ^~~~
2386incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge
2387t to '#include <stddef.h>'?
2388incomplete.c:1:1:
2389+#include <stddef.h>
2390 const char *test(void)
2391incomplete.c:3:10:
2392   return NULL;
2393          ^~~~
2394incomplete.c:3:10: note: each undeclared identifier is reported only once for ea
2395ch function it appears in
2396
2397$ gcc incomplete.cc
2398incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type
2399 std::string s("hello world");
2400      ^~~~~~
2401incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you
2402forget to '#include <string>'?
2403+#include <string>
2404 std::string s("hello world");
2405 ^~~
2406
2407     * The C and C++ compilers now use more intuitive locations when
2408       reporting on missing semicolons, and offer fix-it hints:
2409$ gcc t.c
2410t.c: In function 'test':
2411t.c:3:12: error: expected ';' before '}' token
2412   return 42
2413            ^
2414            ;
2415 }
2416 ~
2417
2418     * When reporting on missing '}' and ')' tokens, the C and C++
2419       compilers will now highlight the corresponding '{' and '(' token,
2420       issuing a 'note' if it's on a separate line:
2421$ gcc unclosed.c
2422unclosed.c: In function 'log_when_out_of_range':
2423unclosed.c:12:50: error: expected ')' before '{' token
2424       && (temperature < MIN || temperature > MAX) {
2425                                                  ^~
2426                                                  )
2427unclosed.c:11:6: note: to match this '('
2428   if (logging_enabled && check_range ()
2429      ^
2430
2431       or highlighting it directly if it's on the same line:
2432$ gcc unclosed-2.c
2433unclosed-2.c: In function 'test':
2434unclosed-2.c:8:45: error: expected ')' before '{' token
2435   if (temperature < MIN || temperature > MAX {
2436      ~                                      ^~
2437                                             )
2438
2439       They will also emit fix-it hints.
2440
2441  C++
2442
2443     * GCC 8 (-fabi-version=12) has a couple of corrections to the calling
2444       convention, which changes the ABI for some uncommon code:
2445          + Passing an empty class as an argument now takes up no space on
2446            x86_64, as required by the psABI.
2447          + Passing or returning a class with only deleted copy and move
2448            constructors now uses the same calling convention as a class
2449            with a non-trivial copy or move constructor. This only affects
2450            C++17 mode, as in earlier standards passing or returning such
2451            a class was impossible.
2452          + WARNING: In GCC 8.1 the second change mistakenly also affects
2453            classes with a deleted copy constructor and defaulted trivial
2454            move constructor (bug [15]c++/86094). This issue is fixed in
2455            GCC 8.2 (-fabi-version=13).
2456       You can test whether these changes affect your code with -Wabi=11
2457       (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are
2458       problematic for your project, the GCC 7 ABI can be selected with
2459       -fabi-version=11.
2460     * The value of the C++11 alignof operator has been corrected to match
2461       C _Alignof (minimum alignment) rather than GNU __alignof__
2462       (preferred alignment); on ia32 targets this means that
2463       alignof(double) is now 4 rather than 8. Code that wants the
2464       preferred alignment should use __alignof__ instead.
2465     * New command-line options have been added for the C++ compiler to
2466       control warnings:
2467          + [16]-Wclass-memaccess warns when objects of non-trivial class
2468            types are manipulated in potentially unsafe ways by raw memory
2469            functions such as memcpy, or realloc. The warning helps detect
2470            calls that bypass user-defined constructors or copy-assignment
2471            operators, corrupt virtual table pointers, data members of
2472            const-qualified types or references, or member pointers. The
2473            warning also detects calls that would bypass access controls
2474            to data members. For example, a call such as:
2475        memcpy (&std::cout, &std::cerr, sizeof std::cout);
2476            results in
2477        warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t
2478o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n
2479o trivial copy-assignment [-Wclass-memaccess]
2480            The -Wclass-memaccess option is included in -Wall.
2481     * The C++ front end has experimental support for some of the upcoming
2482       C++2a draft features with the -std=c++2a or -std=gnu++2a flags,
2483       including designated initializers, default member initializers for
2484       bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported),
2485       lambda [=, this] captures, etc. For a full list of new features,
2486       see [17]the C++ status page.
2487     * When reporting on attempts to access private fields of a class or
2488       struct, the C++ compiler will now offer fix-it hints showing how to
2489       use an accessor function to get at the field in question, if one
2490       exists.
2491$ gcc accessor.cc
2492accessor.cc: In function 'void test(foo*)':
2493accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context
2494   if (ptr->m_ratio >= 0.5)
2495            ^~~~~~~
2496accessor.cc:7:10: note: declared private here
2497   double m_ratio;
2498          ^~~~~~~
2499accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double
2500 foo::get_ratio() const'
2501   if (ptr->m_ratio >= 0.5)
2502            ^~~~~~~
2503            get_ratio()
2504
2505     * The C++ compiler can now give you a hint if you use a macro before
2506       it was defined (e.g. if you mess up the order of your #include
2507       directives):
2508$ gcc ordering.cc
2509ordering.cc:2:24: error: expected ';' at end of member declaration
2510   virtual void clone() const OVERRIDE { }
2511                        ^~~~~
2512                             ;
2513ordering.cc:2:30: error: 'OVERRIDE' does not name a type
2514   virtual void clone() const OVERRIDE { }
2515                              ^~~~~~~~
2516ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined
2517In file included from ordering.cc:5:
2518c++11-compat.h:2: note: it was later defined here
2519 #define OVERRIDE override
2520
2521
2522     * The -Wold-style-cast diagnostic can now emit fix-it hints telling
2523       you when you can use a static_cast, const_cast, or
2524       reinterpret_cast.
2525$ gcc -c old-style-cast-fixits.cc -Wold-style-cast
2526old-style-cast-fixits.cc: In function 'void test(void*)':
2527old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [
2528-Wold-style-cast]
2529   foo *f = (foo *)ptr;
2530                   ^~~
2531            ----------
2532            static_cast<foo *> (ptr)
2533
2534     * When reporting on problems within extern "C" linkage
2535       specifications, the C++ compiler will now display the location of
2536       the start of the extern "C".
2537$ gcc -c extern-c.cc
2538extern-c.cc:3:1: error: template with C linkage
2539 template <typename T> void test (void);
2540 ^~~~~~~~
2541In file included from extern-c.cc:1:
2542unclosed.h:1:1: note: 'extern "C"' linkage started here
2543 extern "C" {
2544 ^~~~~~~~~~
2545extern-c.cc:3:39: error: expected '}' at end of input
2546 template <typename T> void test (void);
2547                                       ^
2548In file included from extern-c.cc:1:
2549unclosed.h:1:12: note: to match this '{'
2550 extern "C" {
2551            ^
2552
2553     * When reporting on mismatching template types, the C++ compiler will
2554       now use color to highlight the mismatching parts of the template,
2555       and will elide the parameters that are common between two
2556       mismatching templates, printing [...] instead:
2557$ gcc templates.cc
2558templates.cc: In function 'void test()':
2559templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl
2560e>' to 'vector<int>'
2561   fn_1(vector<double> ());
2562        ^~~~~~~~~~~~~~~~~
2563templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...]
2564,double>' to 'map<[...],int>'
2565   fn_2(map<int, double>());
2566        ^~~~~~~~~~~~~~~~~~
2567
2568       Those [...] elided parameters can be seen using -fno-elide-type:
2569$ gcc templates.cc -fno-elide-type
2570templates.cc: In function 'void test()':
2571templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl
2572e>' to 'vector<int>'
2573   fn_1(vector<double> ());
2574        ^~~~~~~~~~~~~~~~~
2575templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d
2576ouble>' to 'map<int,int>'
2577   fn_2(map<int, double>());
2578        ^~~~~~~~~~~~~~~~~~
2579
2580       The C++ compiler has also gained an option
2581       -fdiagnostics-show-template-tree which visualizes such mismatching
2582       templates in a hierarchical form:
2583$ gcc templates-2.cc -fdiagnostics-show-template-tree
2584templates-2.cc: In function 'void test()':
2585templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou
2586ble>' to 'vector<int>'
2587  vector<
2588    [double != int]>
2589   fn_1(vector<double> ());
2590        ^~~~~~~~~~~~~~~~~
2591templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve
2592ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m
2593ap<[...],vector<float>>,vector<float>>'
2594  map<
2595    map<
2596      [...],
2597      vector<
2598        [double != float]>>,
2599    vector<
2600      [double != float]>>
2601   fn_2(map<map<int, vector<double>>, vector<double>> ());
2602        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2603
2604       which again works with -fno-elide-type:
2605$ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type
2606templates-2.cc: In function 'void test()':
2607templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou
2608ble>' to 'vector<int>'
2609  vector<
2610    [double != int]>
2611   fn_1(vector<double> ());
2612        ^~~~~~~~~~~~~~~~~
2613templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve
2614ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map
2615<int,vector<float>>,vector<float>>'
2616  map<
2617    map<
2618      int,
2619      vector<
2620        [double != float]>>,
2621    vector<
2622      [double != float]>>
2623   fn_2(map<map<int, vector<double>>, vector<double>> ());
2624        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2625
2626     * Flowing off the end of a non-void function is considered
2627       unreachable and may be subject to optimization on that basis. As a
2628       result of this change, -Wreturn-type warnings are enabled by
2629       default for C++.
2630
2631    Runtime Library (libstdc++)
2632
2633     * Improved experimental support for C++17, including the following
2634       features:
2635          + Deduction guides to support class template argument deduction.
2636          + std::filesystem implementation.
2637          + std::char_traits<char> and std::char_traits<wchar_t> are
2638            usable in constant expressions.
2639          + std::to_chars and std::from_chars (for integers only, not for
2640            floating point types).
2641     * Experimental support for C++2a: std::to_address (thanks to Glen
2642       Fernandes) and std::endian.
2643     * On GNU/Linux, std::random_device::entropy() accesses the kernel's
2644       entropy count for the random device, if known (thanks to Xi
2645       Ruoyao).
2646     * Support for std::experimental::source_location.
2647     * AddressSanitizer integration for std::vector, detecting
2648       out-of-range accesses to the unused capacity of a vector.
2649     * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the
2650       Mathematical Special Functions.
2651
2652  Fortran
2653
2654     * The main version of libfortran has been changed to 5.
2655     * Parameterized derived types, a major feature of Fortran 2003, have
2656       been implemented.
2657     * Partial support is provided for Fortran 2018 teams, which are
2658       hierarchical subsets of images that execute independently of other
2659       image subsets.
2660     * The maximum rank for arrays has been increased to 15, conforming to
2661       the Fortran 2008 standard.
2662     * Transformational intrinsics are now fully supported in
2663       initialization expressions.
2664     * New flag -fc-prototypes to write C prototypes for BIND(C)
2665       procedures and variables.
2666     * If -fmax-stack-var-size is honored if given together with -Ofast,
2667       -fstack-arrays is no longer set in that case.
2668     * New options -fdefault-real-16 and -fdefault-real-10 to control the
2669       default kind of REAL variables.
2670     * A warning is now issued if an array subscript inside a DO loop
2671       could lead to an out-of-bounds-access. The new option
2672       -Wdo-subscript, enabled by -Wextra, warns about this even if the
2673       compiler can not prove that the code will be executed.
2674     * The Fortran front end now attempts to interchange loops if it is
2675       deemed profitable. So far, this is restricted to FORALL and DO
2676       CONCURRENT statements with multiple indices. This behavior be
2677       controlled with the new flag -ffrontend-loop-interchange, which is
2678       enabled with optimization by default. The
2679       -Wfrontend-loop-interchange option warns about such occurrences.
2680     * When an actual argument contains too few elements for a dummy
2681       argument, an error is now issued. The -std=legacy option can be
2682       used to still compile such code.
2683     * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit
2684       integers, making records larger than 2GiB possible.
2685     * The GFORTRAN_DEFAULT_RECL environment variable no longer has any
2686       effect. The record length for preconnected units is now larger than
2687       any practical limit, same as for sequential access units opened
2688       without an explicit RECL= specifier.
2689     * Character variables longer than HUGE(0) elements are now possible
2690       on 64-bit targets. Note that this changes the procedure call ABI
2691       for all procedures with character arguments on 64-bit targets, as
2692       the type of the hidden character length argument has changed. The
2693       hidden character length argument is now of type INTEGER(C_SIZE_T).
2694     * Partial support is provided for Fortran 2018 teams, which are
2695       hierarchical subsets of images that execute independently of other
2696       image subsets.
2697
2698  Go
2699
2700     * GCC 8 provides a complete implementation of the Go 1.10.1 user
2701       packages.
2702     * The garbage collector is now fully concurrent. As before, values
2703       stored on the stack are scanned conservatively, but value stored in
2704       the heap are scanned precisely.
2705     * Escape analysis is fully implemented and enabled by default in the
2706       Go front end. This significantly reduces the number of heap
2707       allocations by allocating values on the stack instead.
2708
2709libgccjit
2710
2711   The libgccjit API gained four new entry points:
2712     * [18]gcc_jit_type_get_vector and
2713     * [19]gcc_jit_context_new_rvalue_from_vector for working with
2714       vectors,
2715     * [20]gcc_jit_type_get_aligned
2716     * [21]gcc_jit_function_get_address
2717
2718   The C code generated by [22]gcc_jit_context_dump_reproducer_to_file is
2719   now easier-to-read.
2720
2721New Targets and Target Specific Improvements
2722
2723  AArch64
2724
2725     * The Armv8.4-A architecture is now supported. It can be used by
2726       specifying the -march=armv8.4-a option.
2727     * The Dot Product instructions are now supported as an optional
2728       extension to the Armv8.2-A architecture and newer and are mandatory
2729       on Armv8.4-A. The extension can be used by specifying the +dotprod
2730       architecture extension. E.g. -march=armv8.2-a+dotprod.
2731     * The Armv8-A +crypto extension has now been split into two
2732       extensions for finer grained control:
2733          + +aes which contains the Armv8-A AES crytographic instructions.
2734          + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic
2735            instructions.
2736       Using +crypto will now enable these two extensions.
2737     * New Armv8.4-A FP16 Floating Point Multiplication Variant
2738       instructions have been added. These instructions are mandatory in
2739       Armv8.4-A but available as an optional extension to Armv8.2-A and
2740       Armv8.3-A. The new extension can be used by specifying the +fp16fml
2741       architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A
2742       the instructions can be enabled by specifying +fp16.
2743     * New cryptographic instructions have been added as optional
2744       extensions to Armv8.2-A and newer. These instructions can be
2745       enabled with:
2746          + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This
2747            implies +sha2.
2748          + +sm4 New SM3 and SM4 instructions from Armv8.4-A.
2749     * The Scalable Vector Extension (SVE) is now supported as an optional
2750       extension to the Armv8.2-A architecture and newer. This support
2751       includes automatic vectorization with SVE instructions, but it does
2752       not yet include the SVE Arm C Language Extensions (ACLE). It can be
2753       enabled by specifying the +sve architecture extension (for example,
2754       -march=armv8.2-a+sve). By default, the generated code works with
2755       all vector lengths, but it can be made specific to N-bit vectors
2756       using -msve-vector-bits=N.
2757     * Support has been added for the following processors (GCC
2758       identifiers in parentheses):
2759          + Arm Cortex-A75 (cortex-a75).
2760          + Arm Cortex-A55 (cortex-a55).
2761          + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE
2762            (cortex-a75.cortex-a55).
2763       The GCC identifiers can be used as arguments to the -mcpu or -mtune
2764       options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as
2765       arguments to the equivalent target attributes and pragmas.
2766
2767  ARC
2768
2769     * Added support for:
2770          + Fast interrupts.
2771          + Naked functions.
2772          + aux variable attributes.
2773          + uncached type qualifier.
2774          + Secure functions via sjli instruction.
2775     * New exception handling implementation.
2776     * Revamped trampoline implementation.
2777     * Refactored small data feature implementation, controlled via the -G
2778       command-line option.
2779     * New support for reduced register set ARC architecture
2780       configurations, controlled via the -mrf16 command-line option.
2781     * Refurbished and improved support for zero overhead loops.
2782       Introduced -mlpc-width command-line option to control the width of
2783       the lp_count register.
2784
2785  ARM
2786
2787     * The -mfpu option now takes a new option setting of -mfpu=auto. When
2788       set to this the floating-point and SIMD settings are derived from
2789       the settings of the -mcpu or -march options. The internal CPU
2790       configurations have been updated with information about the
2791       permitted floating-point configurations supported. See the user
2792       guide for further information about the extended option syntax for
2793       controlling architectural extensions via the -march option.
2794       -mfpu=auto is now the default setting unless the compiler has been
2795       configured with an explicit --with-fpu option.
2796     * The -march and -mcpu options now accept optional extensions to the
2797       architecture or CPU option, allowing the user to enable or disable
2798       any such extensions supported by that architecture or CPU such as
2799       (but not limited to) floating-point and AdvancedSIMD. For example:
2800       the option -mcpu=cortex-a53+nofp will generate code for the
2801       Cortex-A53 processor with no floating-point support. This, in
2802       combination with the new -mfpu=auto option, provides a
2803       straightforward way of specifying a valid build target through a
2804       single -mcpu or -march option. The -mtune option accepts the same
2805       arguments as -mcpu but only the CPU name has an effect on tuning.
2806       The architecture extensions do not have any effect. For details of
2807       what extensions a particular architecture or CPU option supports
2808       please refer to the [23]documentation.
2809     * The -mstructure-size-boundary option has been deprecated and will
2810       be removed in a future release.
2811     * The default link behavior for Armv6 and Armv7-R targets has been
2812       changed to produce BE8 format when generating big-endian images. A
2813       new flag -mbe32 can be used to force the linker to produce legacy
2814       BE32 format images. There is no change of behavior for Armv6-M and
2815       other Armv7 or later targets: these already defaulted to BE8
2816       format. This change brings GCC into alignment with other compilers
2817       for the ARM architecture.
2818     * The Armv8-R architecture is now supported. It can be used by
2819       specifying the -march=armv8-r option.
2820     * The Armv8.3-A architecture is now supported. It can be used by
2821       specifying the -march=armv8.3-a option.
2822     * The Armv8.4-A architecture is now supported. It can be used by
2823       specifying the -march=armv8.4-a option.
2824     * The Dot Product instructions are now supported as an optional
2825       extension to the Armv8.2-A architecture and newer and are mandatory
2826       on Armv8.4-A. The extension can be used by specifying the +dotprod
2827       architecture extension. E.g. -march=armv8.2-a+dotprod.
2828     * Support for setting extensions and architectures using the GCC
2829       target pragma and attribute has been added. It can be used by
2830       specifying #pragma GCC target ("arch=..."), #pragma GCC target
2831       ("+extension"), __attribute__((target("arch=..."))) or
2832       __attribute__((target("+extension"))).
2833     * New Armv8.4-A FP16 Floating Point Multiplication Variant
2834       instructions have been added. These instructions are mandatory in
2835       Armv8.4-A but available as an optional extension to Armv8.2-A and
2836       Armv8.3-A. The new extension can be used by specifying the +fp16fml
2837       architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A
2838       the instructions can be enabled by specifying +fp16.
2839     * Support has been added for the following processors (GCC
2840       identifiers in parentheses):
2841          + Arm Cortex-A75 (cortex-a75).
2842          + Arm Cortex-A55 (cortex-a55).
2843          + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE
2844            (cortex-a75.cortex-a55).
2845          + Arm Cortex-R52 for Armv8-R (cortex-r52).
2846       The GCC identifiers can be used as arguments to the -mcpu or -mtune
2847       options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as
2848       arguments to the equivalent target attributes and pragmas.
2849
2850  AVR
2851
2852     * The AVR port now supports the following XMEGA-like devices:
2853
2854     ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417,
2855     ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617,
2856     ATtiny3214, ATtiny3216, ATtiny3217
2857       The new devices are listed under [24]-mmcu=avrxmega3.
2858          + These devices see flash memory in the RAM address space, so
2859            that features like PROGMEM and __flash are not needed any more
2860            (as opposed to other AVR families for which read-only data
2861            will be located in RAM except special, non-standard features
2862            are used to locate and access such data). This requires that
2863            the compiler is used with Binutils 2.29 or newer so that
2864            [25]read-only data will be located in flash memory.
2865          + A new command-line option -mshort-calls is supported. This
2866            option is used internally for multilib selection of the
2867            avrxmega3 variants. It is not an optimization option. Do not
2868            set it by hand.
2869     * The compiler now generates [26]efficient interrupt service routine
2870       (ISR) prologues and epilogues. This is achieved by using the new
2871       [27]AVR pseudo instruction __gcc_isr which is supported and
2872       resolved by the GNU assembler.
2873          + As the __gcc_isr pseudo-instruction will be resolved by the
2874            assembler, inline assembly is transparent to the process. This
2875            means that when inline assembly uses an instruction like INC
2876            that clobbers the condition code, then the assembler will
2877            detect this and generate an appropriate ISR prologue /
2878            epilogue chunk to save / restore SREG as needed.
2879          + A new command-line option -mno-gas-isr-prologues disables the
2880            generation of the __gcc_isr pseudo instruction. Any non-naked
2881            ISR will save and restore SREG, tmp_reg and zero_reg, no
2882            matter whether the respective register is clobbered or used.
2883          + The feature is turned on per default for all optimization
2884            levels except for -O0 and -Og. It is explicitly enabled by
2885            means of option -mgas-isr-prologues.
2886          + Support has been added for a new [28]AVR function attribute
2887            no_gccisr. It can be used to disable __gcc_isr pseudo
2888            instruction generation for individual ISRs.
2889          + This optimization is only available if GCC is configured with
2890            GNU Binutils 2.29 or newer; or at least with a version of
2891            Binutils that implements feature [29]PR21683.
2892     * The compiler no more saves / restores registers in main; the effect
2893       is the same as if attribute OS_task was specified for main. This
2894       optimization can be switched off by the new command-line option
2895       -mno-main-is-OS_task.
2896
2897  IA-32/x86-64
2898
2899     * The x86 port now supports the naked function attribute.
2900     * Better tuning for znver1 and Intel Core based CPUs.
2901     * Vectorization cost metrics has been reworked leading to significant
2902       improvements on some benchmarks.
2903     * GCC now supports the Intel CPU named Cannonlake through
2904       -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA
2905       and SHA ISA extensions.
2906     * GCC now supports the Intel CPU named Icelake through
2907       -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES,
2908       AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ
2909       ISA extensions.
2910     * GCC now supports the Intel Control-flow Enforcement Technology
2911       (CET) extension through -fcf-protection option.
2912
2913  NDS32
2914
2915     * New command-line options -mext-perf, -mext-perf2, and -mext-string
2916       have been added for performance extension instructions.
2917
2918  Nios II
2919
2920     * The Nios II back end has been improved to generate better-optimized
2921       code. Changes include switching to LRA, more accurate cost models,
2922       and more compact code for addressing static variables.
2923     * New command-line options -mgprel-sec= and -mr0rel-sec= have been
2924       added.
2925     * The stack-smashing protection options are now enabled on Nios II.
2926
2927  PA-RISC
2928
2929     * The default call ABI on 32-bit linux has been changed from callee
2930       copies to caller copies. This affects objects larger than eight
2931       bytes passed by value. The goal is to improve compatibility with
2932       x86 and resolve issues with OpenMP.
2933     * Other PA-RISC targets are unchanged.
2934
2935  PowerPC / PowerPC64 / RS6000
2936
2937     * The PowerPC SPE support is split off to a separate powerpcspe port.
2938       The separate port is deprecated and might be removed in a future
2939       release.
2940     * The Paired Single support (as used on some PPC750 CPUs, -mpaired,
2941       powerpc*-*-linux*paired*) is deprecated and will be removed in a
2942       future release.
2943     * The Xilinx floating point support (-mxilinx-fpu,
2944       powerpc-xilinx-eabi*) is deprecated and will be removed in a future
2945       release.
2946     * Support for using big-endian AltiVec intrinsics on a little-endian
2947       target (-maltivec=be) is deprecated and will be removed in a future
2948       release.
2949
2950  Tile
2951
2952     * The TILE-Gx port is deprecated and will be removed in a future
2953       release.
2954
2955Operating Systems
2956
2957  Windows
2958
2959     * GCC on Microsoft Windows can now be configured via
2960       --enable-mingw-wildcard or --disable-mingw-wildcard to force a
2961       specific behavior for GCC itself with regards to supporting the
2962       wildcard character. Prior versions of GCC would follow the
2963       configuration of the MinGW runtime. This behavior can still be
2964       obtained by not using the above options or by using
2965       --enable-mingw-wildcard=platform.
2966
2967Improvements for plugin authors
2968
2969     * Plugins can now register a callback hook for when comments are
2970       encountered by the C and C++ compilers, e.g. allowing for plugins
2971       to handle documentation markup in code comments.
2972     * The gdbinit support script for debugging GCC now has a
2973       break-on-diagnostic command, providing an easy way to trigger a
2974       breakpoint whenever a diagnostic is emitted.
2975     * The API for creating fix-it hints now supports newlines, and for
2976       emitting mutually incompatible fix-it hints for one diagnostic.
2977
2978GCC 8.1
2979
2980   This is the [30]list of problem reports (PRs) from GCC's bug tracking
2981   system that are known to be fixed in the 8.1 release. This list might
2982   not be complete (that is, it is possible that some PRs that have been
2983   fixed are not listed here).
2984
2985GCC 8.2
2986
2987   This is the [31]list of problem reports (PRs) from GCC's bug tracking
2988   system that are known to be fixed in the 8.2 release. This list might
2989   not be complete (that is, it is possible that some PRs that have been
2990   fixed are not listed here).
2991
2992  General Improvements
2993
2994     * Fixed LTO link-time performance problems caused by an overflow in
2995       the partitioning algorithm while building large binaries.
2996
2997  Language Specific Changes
2998
2999    C++
3000
3001   GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or
3002   returning of classes with a deleted copy constructor and defaulted
3003   trivial move constructor (bug [32]c++/86094). GCC 8.2 introduces
3004   -fabi-version=13 and makes it the default, ABI incompatibilities
3005   between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [33]C++
3006   changes for more details.
3007
3008  Target Specific Changes
3009
3010    IA-32/x86-64
3011
3012     * -mtune=native performance regression [34]PR84413 on Intel Skylake
3013       processors has been fixed.
3014
3015GCC 8.3
3016
3017   This is the [35]list of problem reports (PRs) from GCC's bug tracking
3018   system that are known to be fixed in the 8.3 release. This list might
3019   not be complete (that is, it is possible that some PRs that have been
3020   fixed are not listed here).
3021
3022  Windows
3023
3024     * A C++ Microsoft ABI bitfield layout bug, [36]PR87137 has been
3025       fixed. A non-field declaration could cause the current bitfield
3026       allocation unit to be completed, incorrectly placing a following
3027       bitfield into a new allocation unit. The Microsoft ABI is selected
3028       for:
3029          + Mingw targets
3030          + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields
3031            option is specified, or __attribute__((ms_struct)) is used
3032          + SuperH targets when the -mhitachi option is specified, or
3033            __attribute__((renesas)) is used
3034       GCC 8 introduced additional cases of this defect, but rather than
3035       resolve only those regressions, we decided to resolve all the cases
3036       of this defect in single change.
3037
3038GCC 8.4
3039
3040   This is the [37]list of problem reports (PRs) from GCC's bug tracking
3041   system that are known to be fixed in the 8.4 release. This list might
3042   not be complete (that is, it is possible that some PRs that have been
3043   fixed are not listed here).
3044
3045GCC 8.5
3046
3047   This is the [38]list of problem reports (PRs) from GCC's bug tracking
3048   system that are known to be fixed in the 8.5 release. This list might
3049   not be complete (that is, it is possible that some PRs that have been
3050   fixed are not listed here).
3051
3052  Target Specific Changes
3053
3054    AArch64
3055
3056     * The option -moutline-atomics has been added to aid deployment of
3057       the Large System Extensions (LSE) on GNU/Linux systems built with a
3058       baseline architecture targeting Armv8-A. When the option is
3059       specified code is emitted to detect the presence of LSE
3060       instructions at run time and use them for standard atomic
3061       operations. For more information please refer to the documentation.
3062
3063
3064    For questions related to the use of GCC, please consult these web
3065    pages and the [39]GCC manuals. If that fails, the
3066    [40]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3067    web pages and the development of GCC are welcome on our developer
3068    list at [41]gcc@gcc.gnu.org. All of [42]our lists have public
3069    archives.
3070
3071   Copyright (C) [43]Free Software Foundation, Inc. Verbatim copying and
3072   distribution of this entire article is permitted in any medium,
3073   provided this notice is preserved.
3074
3075   These pages are [44]maintained by the GCC team. Last modified
3076   2023-01-11.
3077
3078References
3079
3080   1. http://gcc.gnu.org/gcc-8/porting_to.html
3081   2. http://gcc.gnu.org/onlinedocs/index.html#current
3082   3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html
3083   4. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros
3084   5. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation
3085   6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944
3086   7. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned
3087   8. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute
3088   9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes
3089  10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned
3090  11. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds
3091  12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict
3092  13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow
3093  14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation
3094  15. https://gcc.gnu.org/PR86094
3095  16. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess
3096  17. http://gcc.gnu.org/projects/cxx-status.html#cxx2a
3097  18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector
3098  19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector
3099  20. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned
3100  21. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address
3101  22. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file
3102  23. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options
3103  24. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html
3104  25. https://sourceware.org/PR21472
3105  26. https://gcc.gnu.org/PR20296
3106  27. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html
3107  28. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html
3108  29. https://sourceware.org/PR21683
3109  30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0
3110  31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2
3111  32. https://gcc.gnu.org/PR86094
3112  33. http://gcc.gnu.org/gcc-8/changes.html#cxx
3113  34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413
3114  35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3
3115  36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137
3116  37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4
3117  38. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.5
3118  39. https://gcc.gnu.org/onlinedocs/
3119  40. mailto:gcc-help@gcc.gnu.org
3120  41. mailto:gcc@gcc.gnu.org
3121  42. https://gcc.gnu.org/lists.html
3122  43. https://www.fsf.org/
3123  44. https://gcc.gnu.org/about.html
3124======================================================================
3125http://gcc.gnu.org/gcc-7/index.html
3126
3127                              GCC 7 Release Series
3128
3129   (This release series is no longer supported.)
3130
3131   Nov 14, 2019
3132
3133   The [1]GNU project and the GCC developers are pleased to announce the
3134   release of GCC 7.5.
3135
3136   This release is a bug-fix release, containing fixes for regressions in
3137   GCC 7.4 relative to previous releases of GCC.
3138
3139Release History
3140
3141   GCC 7.5
3142          Nov 14, 2019 ([2]changes, [3]documentation)
3143
3144   GCC 7.4
3145          Dec 6, 2018 ([4]changes, [5]documentation)
3146
3147   GCC 7.3
3148          Jan 25, 2018 ([6]changes, [7]documentation)
3149
3150   GCC 7.2
3151          Aug 14, 2017 ([8]changes, [9]documentation)
3152
3153   GCC 7.1
3154          May 2, 2017 ([10]changes, [11]documentation)
3155
3156References and Acknowledgements
3157
3158   GCC used to stand for the GNU C Compiler, but since the compiler
3159   supports several other languages aside from C, it now stands for the
3160   GNU Compiler Collection.
3161
3162   A list of [12]successful builds is updated as new information becomes
3163   available.
3164
3165   The GCC developers would like to thank the numerous people that have
3166   contributed new features, improvements, bug fixes, and other changes as
3167   well as test results to GCC. This [13]amazing group of volunteers is
3168   what makes GCC successful.
3169
3170   For additional information about GCC please refer to the [14]GCC
3171   project web site or contact the [15]GCC development mailing list.
3172
3173   To obtain GCC please use [16]our mirror sites or [17]our version
3174   control system.
3175
3176
3177    For questions related to the use of GCC, please consult these web
3178    pages and the [18]GCC manuals. If that fails, the
3179    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3180    web pages and the development of GCC are welcome on our developer
3181    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
3182    archives.
3183
3184   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
3185   distribution of this entire article is permitted in any medium,
3186   provided this notice is preserved.
3187
3188   These pages are [23]maintained by the GCC team. Last modified
3189   2022-10-26.
3190
3191References
3192
3193   1. http://www.gnu.org/
3194   2. http://gcc.gnu.org/gcc-7/changes.html
3195   3. http://gcc.gnu.org/onlinedocs/7.5.0/
3196   4. http://gcc.gnu.org/gcc-7/changes.html
3197   5. http://gcc.gnu.org/onlinedocs/7.4.0/
3198   6. http://gcc.gnu.org/gcc-7/changes.html
3199   7. http://gcc.gnu.org/onlinedocs/7.3.0/
3200   8. http://gcc.gnu.org/gcc-7/changes.html
3201   9. http://gcc.gnu.org/onlinedocs/7.2.0/
3202  10. http://gcc.gnu.org/gcc-7/changes.html
3203  11. http://gcc.gnu.org/onlinedocs/7.1.0/
3204  12. http://gcc.gnu.org/gcc-7/buildstat.html
3205  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
3206  14. http://gcc.gnu.org/index.html
3207  15. mailto:gcc@gcc.gnu.org
3208  16. http://gcc.gnu.org/mirrors.html
3209  17. http://gcc.gnu.org/git.html
3210  18. https://gcc.gnu.org/onlinedocs/
3211  19. mailto:gcc-help@gcc.gnu.org
3212  20. mailto:gcc@gcc.gnu.org
3213  21. https://gcc.gnu.org/lists.html
3214  22. https://www.fsf.org/
3215  23. https://gcc.gnu.org/about.html
3216======================================================================
3217http://gcc.gnu.org/gcc-7/changes.html
3218
3219                              GCC 7 Release Series
3220                        Changes, New Features, and Fixes
3221
3222   This page is a brief summary of some of the huge number of improvements
3223   in GCC 7. For more information, see the [1]Porting to GCC 7 page and
3224   the [2]full GCC documentation.
3225
3226Caveats
3227
3228     * GCC now uses [3]LRA (a new local register allocator) by default for
3229       new targets.
3230     * The non-standard C++0x type traits has_trivial_default_constructor,
3231       has_trivial_copy_constructor and has_trivial_copy_assign have been
3232       removed.
3233     * The libstdc++ [4]Profile Mode has been deprecated and will be
3234       removed in a future version.
3235     * The Cilk+ extensions to the C and C++ languages have been
3236       deprecated.
3237     * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that
3238       affects conformance to the procedure call standard (AAPCS) has been
3239       fixed. The bug affects some C++ code where class objects are passed
3240       by value to functions and could result in incorrect or inconsistent
3241       code being generated. This is an ABI change. If the option -Wpsabi
3242       is enabled (on by default) the compiler will emit a diagnostic note
3243       for code that might be affected.
3244
3245General Optimizer Improvements
3246
3247     * GCC 7 can determine the return value or range of return values of
3248       some calls to the sprintf family of functions and make it available
3249       to other optimization passes. Some calls to the snprintf function
3250       with a zero size argument can be folded into constants. This
3251       optimization is included in -O1 and can be selectively controlled
3252       by the -fprintf-return-value option.
3253     * A new store merging pass has been added. It merges constant stores
3254       to adjacent memory locations into fewer, wider, stores. It is
3255       enabled by the -fstore-merging option and at the -O2 optimization
3256       level or higher (and -Os).
3257     * A new code hoisting optimization has been added to the partial
3258       redundancy elimination pass. It attempts to move evaluation of
3259       expressions executed on all paths to the function exit as early as
3260       possible. This primarily helps improve code size, but can improve
3261       the speed of the generated code as well. It is enabled by the
3262       -fcode-hoisting option and at the -O2 optimization level or higher
3263       (and -Os).
3264     * A new interprocedural bitwise constant propagation optimization has
3265       been added, which propagates knowledge about which bits of
3266       variables are known to be zero (including pointer alignment
3267       information) across the call graph. It is enabled by the
3268       -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled
3269       at the -O2 optimization level and higher (and -Os). This
3270       optimization supersedes interprocedural alignment propagation of
3271       GCC 6, and therefore the option -fipa-cp-alignment is now
3272       deprecated and ignored.
3273     * A new interprocedural value range propagation optimization has been
3274       added, which propagates integral range information across the call
3275       graph when variable values can be proven to be within those ranges.
3276       It is enabled by the -fipa-vrp option and at the -O2 optimization
3277       level and higher (and -Os).
3278     * A new loop splitting optimization pass has been added. Certain
3279       loops which contain a condition that is always true on one side of
3280       the iteration space and always false on the other are split into
3281       two loops, such that each of the two new loops iterates on just one
3282       side of the iteration space and the condition does not need to be
3283       checked inside of the loop. It is enabled by the -fsplit-loops
3284       option and at the -O3 optimization level or higher.
3285     * The shrink-wrapping optimization can now separate portions of
3286       prologues and epilogues to improve performance if some of the work
3287       done traditionally by prologues and epilogues is not needed on
3288       certain paths. This is controlled by the -fshrink-wrap-separate
3289       option, enabled by default. It requires target support, which is
3290       currently only implemented in the PowerPC and AArch64 ports.
3291     * AddressSanitizer gained a new sanitization option,
3292       -fsanitize-address-use-after-scope, which enables sanitization of
3293       variables whose address is taken and used after a scope where the
3294       variable is defined:
3295
3296int
3297main (int argc, char **argv)
3298{
3299  char *ptr;
3300    {
3301      char my_char;
3302      ptr = &my_char;
3303    }
3304
3305  *ptr = 123;
3306  return *ptr;
3307}
3308
3309==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99
33100 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958
3311WRITE of size 1 at 0x7fffb8dba990 thread T0
3312    #0 0x4006d4 in main /tmp/use-after-scope-1.c:10
3313    #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290)
3314    #2 0x400739 in _start (/tmp/a.out+0x400739)
3315
3316Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame
3317    #0 0x40067f in main /tmp/use-after-scope-1.c:3
3318
3319  This frame has 1 object(s):
3320    [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable
3321
3322       The option is enabled by default with -fsanitize=address and
3323       disabled by default with -fsanitize=kernel-address. Compared to the
3324       LLVM compiler, where the option already exists, the implementation
3325       in the GCC compiler has some improvements and advantages:
3326          + Complex uses of gotos and case labels are properly handled and
3327            should not report any false positive or false negatives.
3328          + C++ temporaries are sanitized.
3329          + Sanitization can handle invalid memory stores that are
3330            optimized out by the LLVM compiler when optimization is
3331            enabled.
3332     * The -fsanitize=signed-integer-overflow suboption of the
3333       UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even
3334       on arithmetic operations with generic vectors.
3335     * Version 5 of the DWARF debugging information standard is supported
3336       through the -gdwarf-5 option. The DWARF version 4 debugging
3337       information remains the default until consumers of debugging
3338       information are adjusted.
3339
3340New Languages and Language specific improvements
3341
3342   OpenACC support in C, C++, and Fortran continues to be maintained and
3343   improved. See the [6]OpenACC and [7]Offloading wiki pages for further
3344   information.
3345
3346  Ada
3347
3348     * On mainstream native platforms, Ada programs no longer require the
3349       stack to be made executable in order to run properly.
3350
3351  BRIG (HSAIL)
3352
3353   Support for processing BRIG 1.0 files was added in this release. BRIG
3354   is a binary format for HSAIL (Heterogeneous System Architecture
3355   Intermediate Language). The BRIG front end can be used for implementing
3356   HSAIL "finalizers" (compilation of HSAIL to a native ISA) for
3357   GCC-supported targets. An implementation of an HSAIL runtime library,
3358   libhsail-rt is also included.
3359
3360  C family
3361
3362     * New command-line options have been added for the C and C++
3363       compilers:
3364          + -Wimplicit-fallthrough warns when a switch case falls through.
3365            This warning has five different levels. The compiler is able
3366            to parse a wide range of fallthrough comments, depending on
3367            the level. It also handles control-flow statements, such as
3368            ifs. It's possible to suppress the warning by either adding a
3369            fallthrough comment, or by using a null statement:
3370            __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]];
3371            (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning
3372            is enabled by -Wextra.
3373          + -Wpointer-compare warns when a pointer is compared with a zero
3374            character constant. Such code is now invalid in C++11 and GCC
3375            rejects it. This warning is enabled by default.
3376          + -Wduplicated-branches warns when an if-else has identical
3377            branches.
3378          + -Wrestrict warns when an argument passed to a
3379            restrict-qualified parameter aliases with another argument.
3380          + -Wmemset-elt-size warns for memset calls, when the first
3381            argument references an array, and the third argument is a
3382            number equal to the number of elements of the array, but not
3383            the size of the array. This warning is enabled by -Wall.
3384          + -Wint-in-bool-context warns about suspicious uses of integer
3385            values where boolean values are expected. This warning is
3386            enabled by -Wall.
3387          + -Wswitch-unreachable warns when a switch statement has
3388            statements between the controlling expression and the first
3389            case label which will never be executed. This warning is
3390            enabled by default.
3391          + -Wexpansion-to-defined warns when defined is used outside #if.
3392            This warning is enabled by -Wextra or -Wpedantic.
3393          + -Wregister warns about uses of the register storage specifier.
3394            In C++17 this keyword has been removed and for C++17 this is a
3395            pedantic warning enabled by default. The warning is not
3396            emitted for the GNU Explicit Register Variables extension.
3397          + -Wvla-larger-than=N warns about unbounded uses of
3398            variable-length arrays, and about bounded uses of
3399            variable-length arrays whose bound can be larger than N bytes.
3400          + -Wduplicate-decl-specifier warns when a declaration has
3401            duplicate const, volatile, restrict or _Atomic specifier. This
3402            warning is enabled by -Wall.
3403     * GCC 6's C and C++ front ends were able to offer suggestions for
3404       misspelled field names:
3405
3406spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
3407you mean 'color'?
3408   return ptr->colour;
3409               ^~~~~~
3410
3411       GCC 7 greatly expands the scope of these suggestions. Firstly, it
3412       adds fix-it hints to such suggestions:
3413
3414spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
3415you mean 'color'?
3416   return ptr->colour;
3417               ^~~~~~
3418               color
3419
3420       The suggestions now cover many other things, such as misspelled
3421       function names:
3422
3423spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi
3424dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat
3425ion]
3426   gtk_widget_showall (w);
3427   ^~~~~~~~~~~~~~~~~~
3428   gtk_widget_show_all
3429
3430       misspelled macro names and enum values:
3431
3432spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun
3433ction); did you mean 'MAX_ITEMS'?
3434 int array[MAX_ITEM];
3435           ^~~~~~~~
3436           MAX_ITEMS
3437
3438       misspelled type names:
3439
3440spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si
3441gned'?
3442 void test (singed char e);
3443            ^~~~~~
3444            signed
3445
3446       and, in the C front end, named initializers:
3447
3448test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color
3449'?
3450 struct s test = { .colour = 3 };
3451                    ^~~~~~
3452                    color
3453
3454     * The preprocessor can now offer suggestions for misspelled
3455       directives, e.g.:
3456
3457test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif?
3458 #endfi
3459  ^~~~~
3460  endif
3461
3462     * Warnings about format strings now underline the pertinent part of
3463       the string, and can offer suggested fixes. In some cases, the
3464       pertinent argument is underlined.
3465
3466test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume
3467nt 3 has type 'int' [-Wformat=]
3468   printf ("foo: %d  bar: %s baz: %d", 100, i + j, 102);
3469                          ~^                ~~~~~
3470                          %d
3471
3472     * The new -Wdangling-else command-line option has been split out of
3473       -Wparentheses and warns about dangling else.
3474     * The -Wshadow warning has been split into three variants.
3475       -Wshadow=global warns for any shadowing. This is the default when
3476       using -Wshadow without any argument. -Wshadow=local only warns for
3477       a local variable shadowing another local variable or parameter.
3478       -Wshadow=compatible-local only warns for a local variable shadowing
3479       another local variable or parameter whose type is compatible (in
3480       C++ compatible means that the type of the shadowing variable can be
3481       converted to that of the shadowed variable).
3482       The following example shows the different kinds of shadow warnings:
3483
3484enum operation { add, count };
3485struct container { int nr; };
3486
3487int
3488container_count (struct container c, int count)
3489{
3490  int r = 0;
3491  for (int count = 0; count > 0; count--)
3492    {
3493      struct container count = c;
3494      r += count.nr;
3495    }
3496  return r;
3497}
3498
3499       -Wshadow=compatible-local will warn for the parameter being
3500       shadowed with the same type:
3501
3502warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow=
3503compatible-local]
3504   for (int count = 0; count > 0; count--)
3505            ^~~~~
3506warn-test.c:5:42: note: shadowed declaration is here
3507 container_count (struct container c, int count)
3508                                          ^~~~~
3509
3510       -Wshadow=local will warn for the above and for the shadowed
3511       declaration with incompatible type:
3512
3513warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws
3514hadow=local]
3515       struct container count = c;
3516                        ^~~~~
3517warn-test.c:8:12: note: shadowed declaration is here
3518   for (int count = 0; count > 0; count--)
3519            ^~~~~
3520
3521       -Wshadow=global will warn for all of the above and the shadowing of
3522       the global declaration:
3523
3524warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [
3525-Wshadow]
3526 container_count (struct container c, int count)
3527                                          ^~~~~
3528warn-test.c:1:23: note: shadowed declaration is here
3529 enum operation { add, count };
3530                       ^~~~~
3531
3532     * GCC 7 contains a number of enhancements that help detect buffer
3533       overflow and other forms of invalid memory accesses.
3534          + The -Walloc-size-larger-than=size option detects calls to
3535            standard and user-defined memory allocation functions
3536            decorated with attribute alloc_size whose argument exceeds the
3537            specified size (PTRDIFF_MAX by default). The option also
3538            detects arithmetic overflow in the computation of the size in
3539            two-argument allocation functions like calloc where the total
3540            size is the product of the two arguments. Since calls with an
3541            excessive size cannot succeed they are typically the result of
3542            programming errors. Such bugs have been known to be the source
3543            of security vulnerabilities and a target of exploits.
3544            -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall.
3545            For example, the following call to malloc incorrectly tries to
3546            avoid passing a negative argument to the function and instead
3547            ends up unconditionally invoking it with an argument less than
3548            or equal to zero. Since after conversion to the type of the
3549            argument of the function (size_t) a negative argument results
3550            in a value in excess of the maximum PTRDIFF_MAX the call is
3551            diagnosed.
3552
3553void* f (int n)
3554{
3555  return malloc (n > 0 ? 0 : n);
3556}
3557
3558warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2
3559147483647 [-Walloc-size-larger-than=]
3560
3561          + The -Walloc-zero option detects calls to standard and
3562            user-defined memory allocation functions decorated with
3563            attribute alloc_size with a zero argument. -Walloc-zero is not
3564            included in either -Wall or -Wextra and must be explicitly
3565            enabled.
3566          + The -Walloca option detects all calls to the alloca function
3567            in the program. -Walloca is not included in either -Wall or
3568            -Wextra and must be explicitly enabled.
3569          + The -Walloca-larger-than=size option detects calls to the
3570            alloca function whose argument either may exceed the specified
3571            size, or that is not known to be sufficiently constrained to
3572            avoid exceeding it. -Walloca-larger-than is not included in
3573            either -Wall or -Wextra and must be explicitly enabled.
3574            For example, compiling the following snippet with
3575            -Walloca-larger-than=1024 results in a warning because even
3576            though the code appears to call alloca only with sizes of 1kb
3577            and less, since n is signed, a negative value would result in
3578            a call to the function well in excess of the limit.
3579
3580void f (int n)
3581{
3582  char *d;
3583  if (n < 1025)
3584    d = alloca (n);
3585  else
3586    d = malloc (n);
3587  ...
3588}
3589
3590warning: argument to 'alloca may be too large due to conversion from 'int' to 'l
3591ong unsigned int' [-Walloca-larger-than=]
3592
3593            In contrast, a call to alloca that isn't bounded at all such
3594            as in the following function will elicit the warning below
3595            regardless of the size argument to the option.
3596
3597void f (size_t n)
3598{
3599  char *d = alloca (n);
3600  ...
3601}
3602
3603warning: unbounded use of 'alloca' [-Walloca-larger-than=]
3604
3605          + The -Wformat-overflow=level option detects certain and likely
3606            buffer overflow in calls to the sprintf family of formatted
3607            output functions. Although the option is enabled even without
3608            optimization it works best with -O2 and higher.
3609            For example, in the following snippet the call to sprintf is
3610            diagnosed because even though its output has been constrained
3611            using the modulo operation it could result in as many as three
3612            bytes if mday were negative. The solution is to either
3613            allocate a larger buffer or make sure the argument is not
3614            negative, for example by changing mday's type to unsigned or
3615            by making the type of the second operand of the modulo
3616            expression unsigned: 100U.
3617
3618void* f (int mday)
3619{
3620  char *buf = malloc (3);
3621  sprintf (buf, "%02i", mday % 100);
3622  return buf;
3623}
3624
3625warning: 'sprintf may write a terminating nul past the end of the destination [-
3626Wformat-overflow=]
3627note: 'sprintf' output between 3 and 4 bytes into a destination of size 3
3628
3629          + The -Wformat-truncation=level option detects certain and
3630            likely output truncation in calls to the snprintf family of
3631            formatted output functions. -Wformat-truncation=1 is included
3632            in -Wall and enabled without optimization but works best with
3633            -O2 and higher.
3634            For example, the following function attempts to format an
3635            integer between 0 and 255 in hexadecimal, including the 0x
3636            prefix, into a buffer of four characters. But since the
3637            function must always terminate output by the null character
3638            ('\0') such a buffer is only big enough to fit just one digit
3639            plus the prefix. Therefore the snprintf call is diagnosed. To
3640            avoid the warning either use a bigger buffer or handle the
3641            function's return value which indicates whether or not its
3642            output has been truncated.
3643
3644void f (unsigned x)
3645{
3646  char d[4];
3647  snprintf (d, sizeof d, "%#02x", x & 0xff);
3648  ...
3649}
3650
3651warning: 'snprintf' output may be truncated before the last format character [-W
3652format-truncation=]
3653note: 'snprintf' output between 3 and 5 bytes into a destination of size 4
3654
3655          + The -Wnonnull option has been enhanced to detect a broader set
3656            of cases of passing null pointers to functions that expect a
3657            non-null argument (those decorated with attribute nonnull). By
3658            taking advantage of optimizations the option can detect many
3659            more cases of the problem than in prior GCC versions.
3660          + The -Wstringop-overflow=type option detects buffer overflow in
3661            calls to string handling functions like memcpy and strcpy. The
3662            option relies on [8]Object Size Checking and has an effect
3663            similar to defining the _FORTIFY_SOURCE macro.
3664            -Wstringop-overflow=2 is enabled by default.
3665            For example, in the following snippet, because the call to
3666            strncat specifies a maximum that allows the function to write
3667            past the end of the destination, it is diagnosed. To correct
3668            the problem and avoid the overflow the function should be
3669            called with a size of at most sizeof d - strlen(d) - 1.
3670
3671void f (const char *fname)
3672{
3673  char d[8];
3674  strncpy (d, "/tmp/", sizeof d);
3675  strncat (d, fname, sizeof d);
3676  ...
3677}
3678
3679warning: specified bound 8 equals the size of the destination [-Wstringop-overfl
3680ow=]
3681
3682     * The <limits.h> header provided by GCC defines macros such as
3683       INT_WIDTH for the width in bits of integer types, if
3684       __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is
3685       included. The <stdint.h> header defines such macros as SIZE_WIDTH
3686       and INTMAX_WIDTH for the width of some standard typedef names for
3687       integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined
3688       before the header is included; note that GCC's implementation of
3689       this header is only used for freestanding compilations, not hosted
3690       compilations, on most systems. These macros come from ISO/IEC TS
3691       18661-1:2014.
3692     * The <float.h> header provided by GCC defines the macro
3693       CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if
3694       __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is
3695       included. This represents the number of decimal digits for which
3696       conversions between decimal character strings and binary formats,
3697       in both directions, are correctly rounded, and currently has the
3698       value of UINTMAX_MAX on all systems, reflecting that GCC's
3699       compile-time conversions are correctly rounded for any number of
3700       digits.
3701     * New __builtin_add_overflow_p, __builtin_sub_overflow_p,
3702       __builtin_mul_overflow_p built-in functions have been added. These
3703       work similarly to their siblings without the _p suffix, but do not
3704       actually store the result of the arithmetics anywhere, just return
3705       whether the operation would overflow. Calls to these built-ins with
3706       integer constant arguments evaluate to integer constants
3707       expressions.
3708       For example, in the following, c is assigned the result of a * b
3709       only if the multiplication does not overflow, otherwise it is
3710       assigned the value zero. The multiplication is performed at
3711       compile-time and without triggering a -Woverflow warning.
3712
3713enum {
3714  a = 12345678,
3715  b = 87654321,
3716  c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b
3717};
3718
3719  C
3720
3721     * The C front end now supports type names _FloatN for floating-point
3722       types with IEEE interchange formats and _FloatNx for floating-point
3723       types with IEEE extended formats. These type names come from
3724       ISO/IEC TS 18661-3:2015.
3725       The set of types supported depends on the target for which GCC is
3726       configured. Most targets support _Float32, _Float32x and _Float64.
3727       _Float128 is supported on targets where IEEE binary128 encoding was
3728       already supported as long double or __float128. _Float64x is
3729       supported on targets where a type with either binary128 or Intel
3730       extended precision format is available.
3731       Constants with these types are supported using suffixes fN, FN, fNx
3732       and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are
3733       defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is
3734       defined before it is included.
3735       These new types are always distinct from each other and from float,
3736       double and long double, even if they have the same encoding.
3737       Complex types such as _Complex _Float128 are also supported.
3738       Type-generic built-in functions such as __builtin_isinf support the
3739       new types, and the following type-specific built-in functions have
3740       versions (suffixed fN or fNx) for the new types:
3741       __builtin_copysign, __builtin_fabs, __builtin_huge_val,
3742       __builtin_inf, __builtin_nan, __builtin_nans.
3743     * Compilation with -fopenmp is now compatible with the C11 _Atomic
3744       keyword.
3745
3746  C++
3747
3748     * The C++ front end has experimental support for all of the current
3749       C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if
3750       constexpr, class template argument deduction, auto template
3751       parameters, and structured bindings. For a full list of new
3752       features, see [9]the C++ status page.
3753     * C++17 support for new of over-aligned types can be enabled in other
3754       modes with the -faligned-new flag.
3755     * The C++17 evaluation order requirements can be selected in other
3756       modes with the -fstrong-eval-order flag, or disabled in C++17 mode
3757       with -fno-strong-eval-order.
3758     * The default semantics of inherited constructors has changed in all
3759       modes, following [10]P0136. Essentially, overload resolution
3760       happens as if calling the inherited constructor directly, and the
3761       compiler fills in construction of the other bases and members as
3762       needed. Most uses should not need any changes. The old behavior can
3763       be restored with -fno-new-inheriting-ctors, or -fabi-version less
3764       than 11.
3765     * The resolution of DR 150 on matching of template template
3766       parameters, allowing default template arguments to make a template
3767       match a parameter, is currently enabled by default in C++17 mode
3768       only. The default can be overridden with -f{no-,}new-ttp-matching.
3769     * The C++ front end will now provide fix-it hints for some missing
3770       semicolons, allowing for automatic fixes by IDEs:
3771
3772test.cc:4:11: error: expected ';' after class definition
3773 class a {}
3774           ^
3775           ;
3776
3777     * -Waligned-new has been added to the C++ front end. It warns about
3778       new of type with extended alignment without -faligned-new.
3779
3780    Runtime Library (libstdc++)
3781
3782     * The type of exception thrown by iostreams, std::ios_base::failure,
3783       now uses the [11]cxx11 ABI.
3784     * Experimental support for C++17, including the following new
3785       features:
3786          + std::string_view;
3787          + std::any, std::optional, and std::variant;
3788          + std::invoke, std::is_invocable, std::is_nothrow_invocable, and
3789            invoke_result;
3790          + std::is_swappable, and std::is_nothrow_swappable;
3791          + std::apply, and std::make_from_tuple;
3792          + std::void_t, std::bool_constant, std::conjunction,
3793            std::disjunction, and std::negation;
3794          + Variable templates for type traits;
3795          + Mathematical Special Functions;
3796          + std::chrono::floor, std::chrono::ceil, std::chrono::round, and
3797            std::chrono::abs;
3798          + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot;
3799          + std::scoped_lock, std::shared_mutex,
3800            std::atomic<T>::is_always_lock_free;
3801          + std::sample, std::default_searcher, std::boyer_moore_searcher
3802            and std::boyer_moore_horspool_searcher;
3803          + Extraction and re-insertion of map and set nodes, try_emplace
3804            members for maps, and functions for accessing containers
3805            std::size, std::empty, and std::data;
3806          + std::shared_ptr support for arrays,
3807            std::shared_ptr<T>::weak_type,
3808            std::enable_shared_from_this<T>::weak_from_this(), and
3809            std::owner_less<void>;
3810          + std::byte;
3811          + std::as_const, std::not_fn,
3812            std::has_unique_object_representations, constexpr
3813            std::addressof.
3814       Thanks to Daniel Kr�gler, Tim Shen, Edward Smith-Rowland, and Ville
3815       Voutilainen for work on the C++17 support.
3816     * A new power-of-two rehashing policy for use with the _Hashtable
3817       internals, thanks to Fran�ois Dumont.
3818
3819  Fortran
3820
3821     * Support for a number of extensions for compatibility with legacy
3822       code with new flags:
3823          + -fdec-structure Support for DEC STRUCTURE and UNION
3824          + -fdec-intrinsic-ints Support for new integer intrinsics with
3825            B/I/J/K prefixes such as BABS, JIAND...
3826          + -fdec-math Support for additional math intrinsics, including
3827            COTAN and degree-valued trigonometric functions such as TAND,
3828            ASIND...
3829          + -fdec Enable the -fdec-* family of extensions.
3830     * New flag -finit-derived to allow default initialization of
3831       derived-type variables.
3832     * Improved DO loops with step equal to 1 or -1, generates faster code
3833       without a loop preheader. A new warning, -Wundefined-do-loop, warns
3834       when a loop iterates either to HUGE(i) (with step equal to 1), or
3835       to -HUGE(i) (with step equal to -1). Invalid behavior can be caught
3836       at run time with -fcheck=do enabled:
3837
3838program test
3839  implicit none
3840  integer(1) :: i
3841  do i = -HUGE(i)+10, -HUGE(i)-1, -1
3842    print *, i
3843  end do
3844end program test
3845
3846At line 8 of file do_check_12.f90
3847Fortran runtime error: Loop iterates infinitely
3848
3849     * Version 4.5 of the [12]OpenMP specification is now partially
3850       supported in the Fortran compiler; the largest missing item is
3851       structure element mapping.
3852     * User-defined derived-type input/output (UDTIO) is added.
3853     * Derived type coarrays with allocatable and pointer components are
3854       partially supported.
3855     * Non-constant stop codes and error stop codes (Fortran 2015
3856       feature).
3857     * Derived types with allocatable components of recursive type.
3858     * Intrinsic assignment to polymorphic variables.
3859     * Improved submodule support.
3860     * Improved diagnostics (polymorphic results in pure functions).
3861     * Coarray: Support for failed images (Fortan 2015 feature).
3862
3863  Go
3864
3865     * GCC 7 provides a complete implementation of the Go 1.8.1 user
3866       packages.
3867     * Compared to the Go 1.8.1 toolchain, the garbage collector is more
3868       conservative and less concurrent.
3869     * Escape analysis is available for experimental use via the
3870       -fgo-optimize-allocs option. The -fgo-debug-escape prints
3871       information useful for debugging escape analysis choices.
3872
3873  Java (GCJ)
3874
3875   The GCC Java front end and associated libjava runtime library have been
3876   removed from GCC.
3877
3878libgccjit
3879
3880   The libgccjit API gained support for marking calls as requiring
3881   tail-call optimization via a new entry point:
3882   [13]gcc_jit_rvalue_set_bool_require_tail_call.
3883
3884   libgccjit performs numerous checks at the API boundary, but if these
3885   succeed, it previously ignored errors and other diagnostics emitted
3886   within the core of GCC, and treated the compile of a gcc_jit_context as
3887   having succeeded. As of GCC 7 it now ensures that if any diagnostics
3888   are emitted, they are visible from the libgccjit API, and that the the
3889   context is flagged as having failed.
3890
3891New Targets and Target Specific Improvements
3892
3893  AArch64
3894
3895     * GCC has been updated to the latest revision of the procedure call
3896       standard (AAPCS64) to provide support for parameter passing when
3897       data types have been over-aligned.
3898     * The ARMv8.3-A architecture is now supported. It can be used by
3899       specifying the -march=armv8.3-a option.
3900     * The option -msign-return-address= is supported to enable return
3901       address protection using ARMv8.3-A Pointer Authentication
3902       Extensions. For more information on the arguments accepted by this
3903       option, please refer to [14]AArch64-Options.
3904     * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point
3905       Extensions are now supported. They can be used by specifying the
3906       -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit
3907       Floating-Point Extensions introduce new half-precision data
3908       processing floating-point instructions.
3909     * Support has been added for the following processors (GCC
3910       identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom
3911       Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium
3912       ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx
3913       (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1),
3914       Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor).
3915       The GCC identifiers can be used as arguments to the -mcpu or -mtune
3916       options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as
3917       arguments to the equivalent target attributes and pragmas.
3918
3919  ARC
3920
3921     * Added support for ARC HS and ARC EM processors.
3922     * Added support for ARC EM variation found in Intel QuarkSE SoCs.
3923     * Added support for NPS400 ARC700 based CPUs.
3924     * Thread Local Storage is now supported by ARC CPUs.
3925     * Fixed errors for ARC600 when using 32x16 multiplier option.
3926     * Fixed PIE for ARC CPUs.
3927     * New CPU templates are supported via multilib.
3928
3929  ARM
3930
3931     * Support for the ARMv5 and ARMv5E architectures has been deprecated
3932       (which have no known implementations) and will be removed in a
3933       future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ
3934       architectures remain supported. The values armv5 and armv5e of
3935       -march are thus deprecated.
3936     * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point
3937       Extensions are now supported. They can be used by specifying the
3938       -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit
3939       Floating-Point Extensions introduce new half-precision data
3940       processing floating-point instructions.
3941     * The ARMv8-M architecture is now supported in its two architecture
3942       profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and
3943       Floating-Point Extensions. They can be used by specifying the
3944       -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options.
3945     * Support has been added for the following processors (GCC
3946       identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM
3947       Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC
3948       identifiers can be used as arguments to the -mcpu or -mtune
3949       options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33.
3950     * A new command-line option -mpure-code has been added. It does not
3951       allow constant data to be placed in code sections. This option is
3952       only available when generating non-PIC code for ARMv7-M targets.
3953     * Support for the ACLE Coprocessor Intrinsics has been added. This
3954       enables the generation of coprocessor instructions through the use
3955       of intrinsics such as cdp, ldc, and others.
3956     * The configure option --with-multilib-list now accepts the value
3957       rmprofile to build multilib libraries for a range of embedded
3958       targets. See our [15]installation instructions for details.
3959
3960  AVR
3961
3962     * On the reduced Tiny cores, the progmem [16]variable attribute is
3963       now properly supported. Respective read-only variables are located
3964       in flash memory in section .progmem.data. No special code is needed
3965       to access such variables; the compiler automatically adds an offset
3966       of 0x4000 to all addresses, which is needed to access variables in
3967       flash memory. As opposed to ordinary cores where it is sufficient
3968       to specify the progmem attribute with definitions, on the reduced
3969       Tiny cores the attribute also has to be specified with (external)
3970       declarations:
3971
3972extern const int array[] __attribute__((__progmem__));
3973
3974int get_value2 (void)
3975{
3976  /* Access via addresses array + 0x4004 and array + 0x4005. */
3977  return array[2];
3978}
3979
3980const int* get_address (unsigned idx)
3981{
3982  /* Returns array + 0x4000 + 2 * idx. */
3983  return &array[idx];
3984}
3985
3986     * A new command-line option -Wmisspelled-isr has been added. It turns
3987       off -- or turns into errors -- warnings that are reported for
3988       interrupt service routines (ISRs) which don't follow AVR-LibC's
3989       naming convention of prefixing ISR names with __vector.
3990     * __builtin_avr_nops(n) is a new [17]built-in function that inserts n
3991       NOP instructions into the instruction stream. n must be a value
3992       known at compile time.
3993
3994  IA-32/x86-64
3995
3996     * Support for the AVX-512 Fused Multiply Accumulation Packed Single
3997       precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word
3998       variable precision (4VNNIW), AVX-512 Vector Population Count
3999       (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has
4000       been added.
4001
4002  NVPTX
4003
4004     * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs.
4005       See the [18]Offloading Wiki on how to configure it.
4006
4007  PowerPC / PowerPC64 / RS6000
4008
4009     * The PowerPC port now uses LRA by default.
4010     * GCC now diagnoses inline assembly that clobbers register r2. This
4011       has always been invalid code, and is no longer quietly tolerated.
4012     * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been
4013       enhanced to generate more of the new instructions by default, and
4014       to provide more built-in functions to generate code for other new
4015       instructions.
4016     * The configuration option --enable-gnu-indirect-function is now
4017       enabled by default on PowerPC GNU/Linux builds.
4018     * The PowerPC port will now allow 64-bit and 32-bit integer types to
4019       be allocated to the VSX vector registers (ISA 2.06 and above). In
4020       addition, on ISA 3.0, 16-bit and 8-bit integer types can be
4021       allocated in the vector registers. Previously, only 64-bit integer
4022       types were allowed in the traditional floating point registers.
4023     * New options -mstack-protector-guard=global,
4024       -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and
4025       -mstack-protector-guard-offset= change how the stack protector gets
4026       the value to use as canary.
4027
4028  S/390, System z, IBM z Systems, IBM Z
4029
4030     * Support for the IBM z14 processor has been added. When using the
4031       -march=z14 option, the compiler will generate code making use of
4032       the new instructions introduced with the vector enhancement
4033       facility and the miscellaneous instruction extension facility 2.
4034       The -mtune=z14 option enables z14 specific instruction scheduling
4035       without making use of new instructions.
4036     * Builtins for the new vector instructions have been added and can be
4037       enabled using the -mzvector option.
4038
4039  RISC-V
4040
4041     * Support for the RISC-V instruction set has been added.
4042
4043  RX
4044
4045   Basic support for atomic built-in function has been added. It is
4046   currently implemented by flipping interrupts off and on as needed.
4047
4048  SH
4049
4050     * Support for SH5/SH64 has been removed.
4051     * Improved utilization of delay slots on SH2A.
4052     * Improved utilization of zero-displacement conditional branches.
4053     * The following deprecated options have been removed
4054          + -mcbranchdi
4055          + -mcmpeqdi
4056          + -minvalid-symbols
4057          + -msoft-atomic
4058          + -mspace
4059          + -madjust-unroll
4060     * Support for the following SH2A instructions has been added
4061          + mov.b @-Rm,R0
4062          + mov.w @-Rm,R0
4063          + mov.l @-Rm,R0
4064          + mov.b R0,@Rn+
4065          + mov.w R0,@Rn+
4066          + mov.l R0,@Rn+
4067
4068  SPARC
4069
4070     * The SPARC port now uses LRA by default.
4071     * Support for the new Subtract-Extended-with-Carry instruction
4072       available in SPARC M7 (Niagara 7) has been added.
4073
4074Operating Systems
4075
4076  AIX
4077
4078     * Visibility support has been enabled for AIX 7.1 and above.
4079
4080  Fuchsia
4081
4082     * Support has been added for the [19]Fuchsia OS.
4083
4084  RTEMS
4085
4086     * The ABI changes on ARM so that no short enums are used by default.
4087
4088Other significant improvements
4089
4090     * -fverbose-asm previously emitted information on the meanings of
4091       assembly expressions. This has been extended so that it now also
4092       prints comments showing the source lines that correspond to the
4093       assembly, making it easier to read the generated assembly
4094       (especially with larger functions). For example, given this C
4095       source file:
4096
4097int test (int n)
4098{
4099  int i;
4100  int total = 0;
4101
4102  for (i = 0; i < n; i++)
4103    total += i * i;
4104  return total;
4105}
4106
4107       -fverbose-asm now gives output similar to this for the function
4108       body (when compiling for x86_64, with -Os):
4109
4110       .text
4111       .globl  test
4112       .type   test, @@function
4113test:
4114.LFB0:
4115       .cfi_startproc
4116# example.c:4:   int total = 0;
4117       xorl    %eax, %eax      # <retval>
4118# example.c:6:   for (i = 0; i < n; i++)
4119       xorl    %edx, %edx      # i
4120.L2:
4121# example.c:6:   for (i = 0; i < n; i++)
4122       cmpl    %edi, %edx      # n, i
4123       jge     .L5     #,
4124# example.c:7:     total += i * i;
4125       movl    %edx, %ecx      # i, tmp92
4126       imull   %edx, %ecx      # i, tmp92
4127# example.c:6:   for (i = 0; i < n; i++)
4128       incl    %edx    # i
4129# example.c:7:     total += i * i;
4130       addl    %ecx, %eax      # tmp92, <retval>
4131       jmp     .L2     #
4132.L5:
4133# example.c:10: }
4134       ret
4135       .cfi_endproc
4136
4137     * Two new options have been added for printing fix-it hints:
4138          + -fdiagnostics-parseable-fixits allows for fix-it hints to be
4139            emitted in a machine-readable form, suitable for consumption
4140            by IDEs. For example, given:
4141
4142spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
4143you mean 'color'?
4144   return ptr->colour;
4145               ^~~~~~
4146               color
4147
4148            it will emit:
4149
4150fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color"
4151
4152          + -fdiagnostics-generate-patch will print a patch in "unified"
4153            format after any diagnostics are printed, showing the result
4154            of applying all fix-it hints. For the above example it would
4155            emit:
4156
4157--- spellcheck-fields.cc
4158+++ spellcheck-fields.cc
4159@@ -49,5 +49,5 @@
4160
4161 color get_color(struct s *ptr)
4162 {
4163-  return ptr->colour;
4164+  return ptr->color;
4165 }
4166
4167     * The gcc and g++ driver programs will now provide suggestions for
4168       misspelled arguments to command-line options.
4169
4170$ gcc -c test.c -ftls-model=global-dinamic
4171gcc: error: unknown TLS model 'global-dinamic'
4172gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo
4173cal-dynamic local-exec; did you mean 'global-dynamic'?
4174
4175     * The compiler will now provide suggestions for misspelled
4176       parameters.
4177
4178$ gcc -c test.c --param max-early-inliner-iteration=3
4179cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma
4180x-early-inliner-iterations'?
4181
4182     * Profile-guided optimization (PGO) instrumentation, as well as test
4183       coverage (GCOV), can newly instrument constructors (functions marks
4184       with __attribute__((constructor))), destructors and C++
4185       constructors (and destructors) of classes that are used as the type
4186       of a global variable.
4187     * A new option -fprofile-update=atomic prevents creation of corrupted
4188       profiles created during an instrumentation run (-fprofile=generate)
4189       of an application. The downside of the option is a speed penalty.
4190       Providing -pthread on the command line selects atomic profile
4191       updating (when supported by the target).
4192     * GCC's already extensive testsuite has gained some new capabilities,
4193       to further improve the reliability of the compiler:
4194          + GCC now has an internal unit-testing API and a suite of tests
4195            for programmatic self-testing of subsystems.
4196          + GCC's C front end has been extended so that it can parse dumps
4197            of GCC's internal representations, allowing for DejaGnu tests
4198            that more directly exercise specific optimization passes. This
4199            covers both the [20]GIMPLE representation (for testing
4200            higher-level optimizations) and the [21]RTL representation,
4201            allowing for more direct testing of lower-level details, such
4202            as register allocation and instruction selection.
4203
4204GCC 7.1
4205
4206   This is the [22]list of problem reports (PRs) from GCC's bug tracking
4207   system that are known to be fixed in the 7.1 release. This list might
4208   not be complete (that is, it is possible that some PRs that have been
4209   fixed are not listed here).
4210
4211GCC 7.2
4212
4213   This is the [23]list of problem reports (PRs) from GCC's bug tracking
4214   system that are known to be fixed in the 7.2 release. This list might
4215   not be complete (that is, it is possible that some PRs that have been
4216   fixed are not listed here).
4217
4218  Target Specific Changes
4219
4220    SPARC
4221
4222     * Support for the SPARC M8 processor has been added.
4223     * The switches -mfix-ut700 and -mfix-gr712rc have been added to work
4224       around an erratum in LEON3FT processors.
4225     * Use of the Floating-point Multiply Single to Double (FsMULd)
4226       instruction can now be controlled by the -mfsmuld and -fno-fsmuld
4227       options.
4228
4229  Operating Systems
4230
4231    RTEMS
4232
4233     * The Ada run-time support uses now thread-local storage (TLS).
4234     * Support for RISC-V has been added.
4235     * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long
4236       double has been added.
4237
4238GCC 7.3
4239
4240   This is the [24]list of problem reports (PRs) from GCC's bug tracking
4241   system that are known to be fixed in the 7.3 release. This list might
4242   not be complete (that is, it is possible that some PRs that have been
4243   fixed are not listed here).
4244
4245  Target Specific Changes
4246
4247    SPARC
4248
4249     * Workarounds for the four [25]LEON3FT errata GRLIB-TN-0010..0013
4250       have been added. Relevant errata are activated by the target
4251       specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches.
4252
4253  Operating Systems
4254
4255    RTEMS
4256
4257     * Support has been added for Epiphany target.
4258
4259GCC 7.4
4260
4261   This is the [26]list of problem reports (PRs) from GCC's bug tracking
4262   system that are known to be fixed in the 7.4 release. This list might
4263   not be complete (that is, it is possible that some PRs that have been
4264   fixed are not listed here).
4265
4266GCC 7.5
4267
4268   This is the [27]list of problem reports (PRs) from GCC's bug tracking
4269   system that are known to be fixed in the 7.5 release. This list might
4270   not be complete (that is, it is possible that some PRs that have been
4271   fixed are not listed here).
4272
4273
4274    For questions related to the use of GCC, please consult these web
4275    pages and the [28]GCC manuals. If that fails, the
4276    [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4277    web pages and the development of GCC are welcome on our developer
4278    list at [30]gcc@gcc.gnu.org. All of [31]our lists have public
4279    archives.
4280
4281   Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and
4282   distribution of this entire article is permitted in any medium,
4283   provided this notice is preserved.
4284
4285   These pages are [33]maintained by the GCC team. Last modified
4286   2022-10-26.
4287
4288References
4289
4290   1. http://gcc.gnu.org/gcc-7/porting_to.html
4291   2. http://gcc.gnu.org/onlinedocs/index.html#current
4292   3. https://gcc.gnu.org/wiki/LRAIsDefault
4293   4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html
4294   5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728
4295   6. https://gcc.gnu.org/wiki/OpenACC
4296   7. https://gcc.gnu.org/wiki/Offloading
4297   8. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html
4298   9. https://gcc.gnu.org/projects/cxx-status.html#cxx1z
4299  10. https://wg21.link/p0136
4300  11. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html
4301  12. https://www.openmp.org/specifications/
4302  13. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call
4303  14. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options
4304  15. https://gcc.gnu.org/install/configure.html
4305  16. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html
4306  17. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html
4307  18. https://gcc.gnu.org/wiki/Offloading
4308  19. https://fuchsia.googlesource.com/
4309  20. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html
4310  21. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html
4311  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0
4312  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2
4313  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3
4314  25. https://www.gaisler.com/index.php/information/app-tech-notes
4315  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4
4316  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5
4317  28. https://gcc.gnu.org/onlinedocs/
4318  29. mailto:gcc-help@gcc.gnu.org
4319  30. mailto:gcc@gcc.gnu.org
4320  31. https://gcc.gnu.org/lists.html
4321  32. https://www.fsf.org/
4322  33. https://gcc.gnu.org/about.html
4323======================================================================
4324http://gcc.gnu.org/gcc-6/index.html
4325
4326                              GCC 6 Release Series
4327
4328   (This release series is no longer supported.)
4329
4330   October 26, 2018
4331
4332   The [1]GNU project and the GCC developers are pleased to announce the
4333   release of GCC 6.5.
4334
4335   This release is a bug-fix release, containing fixes for regressions in
4336   GCC 6.4 relative to previous releases of GCC.
4337
4338Release History
4339
4340   GCC 6.5
4341          October 26, 2018 ([2]changes, [3]documentation)
4342
4343   GCC 6.4
4344          July 4, 2017 ([4]changes, [5]documentation)
4345
4346   GCC 6.3
4347          December 21, 2016 ([6]changes, [7]documentation)
4348
4349   GCC 6.2
4350          August 22, 2016 ([8]changes, [9]documentation)
4351
4352   GCC 6.1
4353          April 27, 2016 ([10]changes, [11]documentation)
4354
4355References and Acknowledgements
4356
4357   GCC used to stand for the GNU C Compiler, but since the compiler
4358   supports several other languages aside from C, it now stands for the
4359   GNU Compiler Collection.
4360
4361   A list of [12]successful builds is updated as new information becomes
4362   available.
4363
4364   The GCC developers would like to thank the numerous people that have
4365   contributed new features, improvements, bug fixes, and other changes as
4366   well as test results to GCC. This [13]amazing group of volunteers is
4367   what makes GCC successful.
4368
4369   For additional information about GCC please refer to the [14]GCC
4370   project web site or contact the [15]GCC development mailing list.
4371
4372   To obtain GCC please use [16]our mirror sites or [17]our version
4373   control system.
4374
4375
4376    For questions related to the use of GCC, please consult these web
4377    pages and the [18]GCC manuals. If that fails, the
4378    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4379    web pages and the development of GCC are welcome on our developer
4380    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
4381    archives.
4382
4383   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
4384   distribution of this entire article is permitted in any medium,
4385   provided this notice is preserved.
4386
4387   These pages are [23]maintained by the GCC team. Last modified
4388   2022-10-26.
4389
4390References
4391
4392   1. http://www.gnu.org/
4393   2. http://gcc.gnu.org/gcc-6/changes.html
4394   3. http://gcc.gnu.org/onlinedocs/6.5.0/
4395   4. http://gcc.gnu.org/gcc-6/changes.html
4396   5. http://gcc.gnu.org/onlinedocs/6.4.0/
4397   6. http://gcc.gnu.org/gcc-6/changes.html
4398   7. http://gcc.gnu.org/onlinedocs/6.3.0/
4399   8. http://gcc.gnu.org/gcc-6/changes.html
4400   9. http://gcc.gnu.org/onlinedocs/6.2.0/
4401  10. http://gcc.gnu.org/gcc-6/changes.html
4402  11. http://gcc.gnu.org/onlinedocs/6.1.0/
4403  12. http://gcc.gnu.org/gcc-6/buildstat.html
4404  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
4405  14. http://gcc.gnu.org/index.html
4406  15. mailto:gcc@gcc.gnu.org
4407  16. http://gcc.gnu.org/mirrors.html
4408  17. http://gcc.gnu.org/git.html
4409  18. https://gcc.gnu.org/onlinedocs/
4410  19. mailto:gcc-help@gcc.gnu.org
4411  20. mailto:gcc@gcc.gnu.org
4412  21. https://gcc.gnu.org/lists.html
4413  22. https://www.fsf.org/
4414  23. https://gcc.gnu.org/about.html
4415======================================================================
4416http://gcc.gnu.org/gcc-6/changes.html
4417
4418                              GCC 6 Release Series
4419                        Changes, New Features, and Fixes
4420
4421   This page is a brief summary of some of the huge number of improvements
4422   in GCC 6. For more information, see the [1]Porting to GCC 6 page and
4423   the [2]full GCC documentation.
4424
4425Caveats
4426
4427     * The default mode for C++ is now -std=gnu++14 instead of
4428       -std=gnu++98.
4429     * Support for a number of older systems and recently unmaintained or
4430       untested target ports of GCC has been declared obsolete in GCC 6.
4431       Unless there is activity to revive them, the next release of GCC
4432       will have their sources permanently removed.
4433       The following ports for individual systems on particular
4434       architectures have been obsoleted:
4435          + SH5 / SH64 (sh64-*-*) as announced [3]here.
4436     * The AVR port requires binutils version 2.26.1 or later for the fix
4437       for [4]PR71151 to work.
4438     * The GCC 6.5 release has an accidental ABI incompatibility for
4439       nested std::pair objects, for more details see [5]PR 87822. The bug
4440       causes a layout change for pairs where the first member is also a
4441       pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series
4442       is closed so the bug in GCC 6.5 will not be fixed upstream, but
4443       there is a patch in the bug report to allow it to be fixed by
4444       anybody packaging GCC 6.5 or installing it themselves.
4445
4446General Optimizer Improvements
4447
4448     * UndefinedBehaviorSanitizer gained a new sanitization option,
4449       -fsanitize=bounds-strict, which enables strict checking of array
4450       bounds. In particular, it enables -fsanitize=bounds as well as
4451       instrumentation of flexible array member-like arrays.
4452     * Type-based alias analysis now disambiguates accesses to different
4453       pointers. This improves precision of the alias oracle by about
4454       20-30% on higher-level C++ programs. Programs doing invalid type
4455       punning of pointer types may now need -fno-strict-aliasing to work
4456       correctly.
4457     * Alias analysis now correctly supports the weakref and alias
4458       attributes. This allows accessing both a variable and its alias in
4459       one translation unit which is common with link-time optimization.
4460     * Value range propagation now assumes that the this pointer in C++
4461       member functions is non-null. This eliminates common null pointer
4462       checks but also breaks some non-conforming code-bases (such as
4463       Qt-5, Chromium, KDevelop). As a temporary work-around
4464       -fno-delete-null-pointer-checks can be used. Wrong code can be
4465       identified by using -fsanitize=undefined.
4466     * Link-time optimization improvements:
4467          + warning and error attributes are now correctly preserved by
4468            declaration linking and thus -D_FORTIFY_SOURCE=2 is now
4469            supported with -flto.
4470          + Type merging was fixed to handle C and Fortran
4471            interoperability rules as defined by the Fortran 2008 language
4472            standard.
4473            As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable
4474            with char in all cases because it is an array while char is
4475            scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In
4476            general, this inter-operability cannot be implemented, for
4477            example on targets where the argument passing convention for
4478            arrays differs from scalars.
4479          + More type information is now preserved at link time, reducing
4480            the loss of accuracy of the type-based alias analysis compared
4481            to builds without link-time optimization.
4482          + Invalid type punning on global variables and declarations is
4483            now reported with -Wodr-type-mismatch.
4484          + The size of LTO object files was reduced by about 11%
4485            (measured by compiling Firefox 46.0).
4486          + Link-time parallelization (enabled using -flto=n) was
4487            significantly improved by decreasing the size of streamed data
4488            when partitioning programs. The size of streamed IL while
4489            compiling Firefox 46.0 was reduced by 66%.
4490          + The linker plugin was extended to pass information about the
4491            type of binary produced to the GCC back end. (That can also be
4492            controlled manually by -flinker-output.) This makes it
4493            possible to properly configure the code generator and support
4494            incremental linking. Incremental linking of LTO objects by gcc
4495            -r is now supported for plugin-enabled setups.
4496            There are two ways to perform incremental linking:
4497              1. Linking by ld -r will result in an object file with all
4498                 sections from individual object files mechanically
4499                 merged. This delays the actual link-time optimization to
4500                 the final linking step and thus permits whole program
4501                 optimization. Linking the final binary with such object
4502                 files is however slower.
4503              2. Linking by gcc -r will lead to link-time optimization and
4504                 emit the final binary into the object file. Linking such
4505                 an object file is fast but avoids any benefits from whole
4506                 program optimization.
4507            GCC 7 will support incremental link-time optimization with gcc
4508            -r.
4509     * Inter-procedural optimization improvements:
4510          + Basic jump threading is now performed before profile
4511            construction and inline analysis, resulting in more realistic
4512            size and time estimates that drive the heuristics of the
4513            inliner and function cloning passes.
4514          + Function cloning now more aggressively eliminates unused
4515            function parameters.
4516
4517New Languages and Language specific improvements
4518
4519   Compared to GCC 5, the GCC 6 release series includes a much improved
4520   implementation of the [6]OpenACC 2.0a specification. Highlights are:
4521     * In addition to single-threaded host-fallback execution, offloading
4522       is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit
4523       little-endian GNU/Linux host systems. For nvptx offloading, with
4524       the OpenACC parallel construct, the execution model allows for an
4525       arbitrary number of gangs, up to 32 workers, and 32 vectors.
4526     * Initial support for parallelized execution of OpenACC kernels
4527       constructs:
4528          + Parallelization of a kernels region is switched on by
4529            -fopenacc combined with -O2 or higher.
4530          + Code is offloaded onto multiple gangs, but executes with just
4531            one worker, and a vector length of 1.
4532          + Directives inside a kernels region are not supported.
4533          + Loops with reductions can be parallelized.
4534          + Only kernels regions with one loop nest are parallelized.
4535          + Only the outer-most loop of a loop nest can be parallelized.
4536          + Loop nests containing sibling loops are not parallelized.
4537       Typically, using the OpenACC parallel construct gives much better
4538       performance, compared to the initial support of the OpenACC kernels
4539       construct.
4540     * The device_type clause is not supported. The bind and nohost
4541       clauses are not supported. The host_data directive is not supported
4542       in Fortran.
4543     * Nested parallelism (cf. CUDA dynamic parallelism) is not supported.
4544     * Usage of OpenACC constructs inside multithreaded contexts (such as
4545       created by OpenMP, or pthread programming) is not supported.
4546     * If a call to the acc_on_device function has a compile-time constant
4547       argument, the function call evaluates to a compile-time constant
4548       value only for C and C++ but not for Fortran.
4549
4550   See the [7]OpenACC and [8]Offloading wiki pages for further
4551   information.
4552
4553  C family
4554
4555     * Version 4.5 of the [9]OpenMP specification is now supported in the
4556       C and C++ compilers.
4557     * The C and C++ compilers now support attributes on enumerators. For
4558       instance, it is now possible to mark enumerators as deprecated:
4559
4560enum {
4561  newval,
4562  oldval __attribute__ ((deprecated ("too old")))
4563};
4564
4565     * Source locations for the C and C++ compilers are now tracked as
4566       ranges, rather than just points, making it easier to identify the
4567       subexpression of interest within a complicated expression. For
4568       example:
4569
4570test.cc: In function 'int test(int, int, foo, int, int)':
4571test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo'
4572)
4573   return p + q * r * s + t;
4574              ~~^~~
4575
4576       In addition, there is now initial support for precise diagnostic
4577       locations within strings:
4578
4579format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in
4580t' argument [-Wformat=]
4581   printf("%*d");
4582            ^
4583
4584     * Diagnostics can now contain "fix-it hints", which are displayed in
4585       context underneath the relevant source code. For example:
4586
4587fixits.c: In function 'bad_deref':
4588fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'?
4589   return ptr.x;
4590             ^
4591             ->
4592
4593     * The C and C++ compilers now offer suggestions for misspelled field
4594       names:
4595
4596spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
4597you mean 'color'?
4598   return ptr->colour;
4599               ^~~~~~
4600
4601     * New command-line options have been added for the C and C++
4602       compilers:
4603          + -Wshift-negative-value warns about left shifting a negative
4604            value.
4605          + -Wshift-overflow warns about left shift overflows. This
4606            warning is enabled by default. -Wshift-overflow=2 also warns
4607            about left-shifting 1 into the sign bit.
4608          + -Wtautological-compare warns if a self-comparison always
4609            evaluates to true or false. This warning is enabled by -Wall.
4610          + -Wnull-dereference warns if the compiler detects paths that
4611            trigger erroneous or undefined behavior due to dereferencing a
4612            null pointer. This option is only active when
4613            -fdelete-null-pointer-checks is active, which is enabled by
4614            optimizations in most targets. The precision of the warnings
4615            depends on the optimization options used.
4616          + -Wduplicated-cond warns about duplicated conditions in an
4617            if-else-if chain.
4618          + -Wmisleading-indentation warns about places where the
4619            indentation of the code gives a misleading idea of the block
4620            structure of the code to a human reader. For example, given
4621            [10]CVE-2014-1266:
4622
4623sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange':
4624sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin
4625g-indentation]
4626    if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0)
4627    ^~
4628sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly
4629indented as if it is guarded by the 'if'
4630        goto fail;
4631        ^~~~
4632
4633            This warning is enabled by -Wall.
4634     * The C and C++ compilers now emit saner error messages if
4635       merge-conflict markers are present in a source file.
4636
4637test.c:3:1: error: version control conflict marker in file
4638 <<<<<<< HEAD
4639 ^~~~~~~
4640
4641  C
4642
4643     * It is possible to disable warnings when an initialized field of a
4644       structure or a union with side effects is being overridden when
4645       using designated initializers via a new warning option
4646       -Woverride-init-side-effects.
4647     * A new type attribute scalar_storage_order applying to structures
4648       and unions has been introduced. It specifies the storage order (aka
4649       endianness) in memory of scalar fields in structures or unions.
4650
4651  C++
4652
4653     * The default mode has been changed to -std=gnu++14.
4654     * [11]C++ Concepts are now supported when compiling with -fconcepts.
4655     * -flifetime-dse is more aggressive in dead-store elimination in
4656       situations where a memory store to a location precedes a
4657       constructor to that memory location.
4658     * G++ now supports [12]C++17 fold expressions, u8 character literals,
4659       extended static_assert, and nested namespace definitions.
4660     * G++ now allows constant evaluation for all non-type template
4661       arguments.
4662     * G++ now supports C++ Transactional Memory when compiling with
4663       -fgnu-tm.
4664
4665    Runtime Library (libstdc++)
4666
4667     * Extensions to the C++ Library to support mathematical special
4668       functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland.
4669     * Experimental support for C++17, including the following new
4670       features:
4671          + std::uncaught_exceptions function (this is also available for
4672            -std=gnu++NN modes);
4673          + new member functions try_emplace and insert_or_assign for
4674            unique_key maps;
4675          + non-member functions std::size, std::empty, and std::data for
4676            accessing containers and arrays;
4677          + std::invoke;
4678          + std::shared_mutex;
4679          + std::void_t and std::bool_constant metaprogramming utilities.
4680       Thanks to Ville Voutilainen for contributing many of the C++17
4681       features.
4682     * An experimental implementation of the File System TS.
4683     * Experimental support for most features of the second version of the
4684       Library Fundamentals TS. This includes polymorphic memory resources
4685       and array support in shared_ptr, thanks to Fan You.
4686     * Some assertions checked by Debug Mode can now also be enabled by
4687       _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro
4688       have less run-time overhead than the full _GLIBCXX_DEBUG checks and
4689       don't affect the library ABI, so can be enabled per-translation
4690       unit.
4691     * Timed mutex types are supported on more targets, including Darwin.
4692     * Improved std::locale support for DragonFly and FreeBSD, thanks to
4693       John Marino and Andreas Tobler.
4694
4695  Fortran
4696
4697     * Fortran 2008 SUBMODULE support.
4698     * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY
4699       support.
4700     * Improved support for Fortran 2003 deferred-length character
4701       variables.
4702     * Improved support for OpenMP and OpenACC.
4703     * The MATMUL intrinsic is now inlined for straightforward cases if
4704       front-end optimization is active. The maximum size for inlining can
4705       be set to n with the -finline-matmul-limit=n option and turned off
4706       with -finline-matmul-limit=0.
4707     * The -Wconversion-extra option will warn about REAL constants which
4708       have excess precision for their kind.
4709     * The -Winteger-division option has been added, which warns about
4710       divisions of integer constants which are truncated. This option is
4711       included in -Wall by default.
4712
4713libgccjit
4714
4715     * The driver code is now run in-process within libgccjit, providing a
4716       small speed-up of the compilation process.
4717     * The API has gained entrypoints for
4718          + [13]timing how long was spent in different parts of code,
4719          + [14]creating switch statements,
4720          + [15]allowing unreachable basic blocks in a function, and
4721          + [16]adding arbitrary command-line options to a compilation.
4722
4723New Targets and Target Specific Improvements
4724
4725  AArch64
4726
4727     * A number of AArch64-specific options have been added. The most
4728       important ones are summarised in this section; for more detailed
4729       information please refer to the documentation.
4730     * The command-line options -march=native, -mcpu=native and
4731       -mtune=native are now available on native AArch64 GNU/Linux
4732       systems. Specifying these options causes GCC to auto-detect the
4733       host CPU and choose the optimal setting for that system.
4734     * -fpic is now supported when generating code for the small code
4735       model (-mcmodel=small). The size of the global offset table (GOT)
4736       is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the
4737       ILP32 SysV ABI.
4738     * The AArch64 port now supports target attributes and pragmas. Please
4739       refer to the [17]documentation for details of available attributes
4740       and pragmas as well as usage instructions.
4741     * Link-time optimization across translation units with different
4742       target-specific options is now supported.
4743     * The option -mtls-size= is now supported. It can be used to specify
4744       the bit size of TLS offsets, allowing GCC to generate better TLS
4745       instruction sequences.
4746     * The option -fno-plt is now fully functional.
4747     * The ARMv8.1-A architecture and the Large System Extensions are now
4748       supported. They can be used by specifying the -march=armv8.1-a
4749       option. Additionally, the +lse option extension can be used in a
4750       similar fashion to other option extensions. The Large System
4751       Extensions introduce new instructions that are used in the
4752       implementation of atomic operations.
4753     * The ACLE half-precision floating-point type __fp16 is now supported
4754       in the C and C++ languages.
4755     * The ARM Cortex-A35 processor is now supported via the
4756       -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the
4757       equivalent target attributes and pragmas.
4758     * The Qualcomm QDF24xx processor is now supported via the
4759       -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent
4760       target attributes and pragmas.
4761     * Code generation for the ARM Cortex-A57 processor is improved. Among
4762       general code generation improvements, a better algorithm is added
4763       for allocating registers to floating-point multiply-accumulate
4764       instructions offering increased performance when compiling with
4765       -mcpu=cortex-a57 or -mtune=cortex-a57.
4766     * Code generation for the ARM Cortex-A53 processor is improved. A
4767       more accurate instruction scheduling model for the processor is now
4768       used, and a number of compiler tuning parameters have been set to
4769       offer increased performance when compiling with -mcpu=cortex-a53 or
4770       -mtune=cortex-a53.
4771     * Code generation for the Samsung Exynos M1 processor is improved. A
4772       more accurate instruction scheduling model for the processor is now
4773       used, and a number of compiler tuning parameters have been set to
4774       offer increased performance when compiling with -mcpu=exynos-m1 or
4775       -mtune=exynos-m1.
4776     * Improvements in the generation of conditional branches and literal
4777       pools allow the compiler to compile functions of a large size.
4778       Constant pools are now placed into separate rodata sections. The
4779       new option -mpc-relative-literal-loads generates per-function
4780       literal pools, limiting the maximum size of functions to 1MiB.
4781     * Several correctness issues generating Advanced SIMD instructions
4782       for big-endian targets have been fixed resulting in improved code
4783       generation for ACLE intrinsics with -mbig-endian.
4784
4785  ARM
4786
4787     * Support for revisions of the ARM architecture prior to ARMv4t has
4788       been deprecated and will be removed in a future GCC release. The
4789       -mcpu and -mtune values that are deprecated are: arm2, arm250,
4790       arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di,
4791       arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500,
4792       arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm,
4793       strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value
4794       arm7tdmi is still supported. The values of -march that are
4795       deprecated are: armv2,armv2a,armv3,armv3m,armv4.
4796     * The ARM port now supports target attributes and pragmas. Please
4797       refer to the [18]documentation for details of available attributes
4798       and pragmas as well as usage instructions.
4799     * Support has been added for the following processors (GCC
4800       identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM
4801       Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC
4802       identifiers can be used as arguments to the -mcpu or -mtune
4803       options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35.
4804
4805  Heterogeneous Systems Architecture
4806
4807     * GCC can now generate HSAIL (Heterogeneous System Architecture
4808       Intermediate Language) for simple OpenMP device constructs if
4809       configured with --enable-offload-targets=hsa. A new libgomp plugin
4810       then runs the HSA GPU kernels implementing these constructs on HSA
4811       capable GPUs via a standard HSA run time.
4812       If the HSA compilation back end determines it cannot output HSAIL
4813       for a particular input, it gives a warning by default. These
4814       warnings can be suppressed with -Wno-hsa. To give a few examples,
4815       the HSA back end does not implement compilation of code using
4816       function pointers, automatic allocation of variable sized arrays,
4817       functions with variadic arguments as well as a number of other less
4818       common programming constructs.
4819       When compilation for HSA is enabled, the compiler attempts to
4820       compile composite OpenMP constructs
4821
4822#pragma omp target teams distribute parallel for
4823
4824       into parallel HSA GPU kernels.
4825
4826  IA-32/x86-64
4827
4828     * GCC now supports the Intel CPU named Skylake with AVX-512
4829       extensions through -march=skylake-avx512. The switch enables the
4830       following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW,
4831       AVX-512DQ.
4832     * Support for new AMD instructions monitorx and mwaitx has been
4833       added. This includes new intrinsic and built-in support. It is
4834       enabled through option -mmwaitx. The instructions monitorx and
4835       mwaitx implement the same functionality as the old monitor and
4836       mwait instructions. In addition, mwaitx adds a configurable timer.
4837       The timer value is received as third argument and stored in
4838       register %ebx.
4839     * x86-64 targets now allow stack realignment from a word-aligned
4840       stack pointer using the command-line option -mstackrealign or
4841       __attribute__ ((force_align_arg_pointer)). This allows functions
4842       compiled with a vector-aligned stack to be invoked from objects
4843       that keep only word-alignment.
4844     * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These
4845       can be used to access data via the %fs and %gs segments without
4846       having to resort to inline assembly. Please refer to the
4847       [19]documentation for usage instructions.
4848     * Support for AMD Zen (family 17h) processors is now available
4849       through the -march=znver1 and -mtune=znver1 options.
4850
4851  MeP
4852
4853     * Support for the MeP (mep-elf) architecture has been deprecated and
4854       will be removed in a future GCC release.
4855
4856  MSP430
4857
4858     * The MSP430 compiler now has the ability to automatically distribute
4859       code and data between low memory (addresses below 64K) and high
4860       memory. This only applies to parts that actually have both memory
4861       regions and only if the linker script for the part has been
4862       specifically set up to support this feature.
4863       A new attribute of either can be applied to both functions and
4864       data, and this tells the compiler to place the object into low
4865       memory if there is room and into high memory otherwise. Two other
4866       new attributes - lower and upper - can be used to explicitly state
4867       that an object should be placed in the specified memory region. If
4868       there is not enough left in that region the compilation will fail.
4869       Two new command-line options - -mcode-region=[lower|upper|either]
4870       and -mdata-region=[lower|upper|either] - can be used to tell the
4871       compiler what to do with objects that do not have one of these new
4872       attributes.
4873
4874  PowerPC / PowerPC64 / RS6000
4875
4876     * PowerPC64 now supports IEEE 128-bit floating-point using the
4877       __float128 data type. In GCC 6, this is not enabled by default, but
4878       you can enable it with -mfloat128. The IEEE 128-bit floating-point
4879       support requires the use of the VSX instruction set. IEEE 128-bit
4880       floating-point values are passed and returned as a single vector
4881       value. The software emulator for IEEE 128-bit floating-point
4882       support is only built on PowerPC GNU/Linux systems where the
4883       default CPU is at least power7. On future ISA 3.0 systems (POWER 9
4884       and later), you will be able to use the -mfloat128-hardware option
4885       to use the ISA 3.0 instructions that support IEEE 128-bit
4886       floating-point. An additional type (__ibm128) has been added to
4887       refer to the IBM extended double type that normally implements long
4888       double. This will allow for a future transition to implementing
4889       long double with IEEE 128-bit floating-point.
4890     * Basic support has been added for POWER9 hardware that will use the
4891       recently published OpenPOWER ISA 3.0 instructions. The following
4892       new switches are available:
4893          + -mcpu=power9: Implement all of the ISA 3.0 instructions
4894            supported by the compiler.
4895          + -mtune=power9: In the future, apply tuning for POWER9 systems.
4896            Currently, POWER8 tunings are used.
4897          + -mmodulo: Generate code using the ISA 3.0 integer instructions
4898            (modulus, count trailing zeros, array index support, integer
4899            multiply/add).
4900          + -mpower9-fusion: Generate code to suitably fuse instruction
4901            sequences for a POWER9 system.
4902          + -mpower9-dform: Generate code to use the new D-form
4903            (register+offset) memory instructions for the vector
4904            registers.
4905          + -mpower9-vector: Generate code using the new ISA 3.0 vector
4906            (VSX or Altivec) instructions.
4907          + -mpower9-minmax: Reserved for future development.
4908          + -mtoc-fusion: Keep TOC entries together to provide more fusion
4909            opportunities.
4910     * New constraints have been added to support IEEE 128-bit
4911       floating-point and ISA 3.0 instructions:
4912          + wb: Altivec register if -mpower9-dform is enabled.
4913          + we: VSX register if -mpower9-vector is enabled for 64-bit code
4914            generation.
4915          + wo: VSX register if -mpower9-vector is enabled.
4916          + wp: Reserved for future use if long double is implemented with
4917            IEEE 128-bit floating-point instead of IBM extended double.
4918          + wq: VSX register if -mfloat128 is enabled.
4919          + wF: Memory operand suitable for POWER9 fusion load/store.
4920          + wG: Memory operand suitable for TOC fusion memory references.
4921          + wL: Integer constant identifying the element number mfvsrld
4922            accesses within a vector.
4923     * Support has been added for __builtin_cpu_is() and
4924       __builtin_cpu_supports(), allowing for very fast access to
4925       AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of
4926       glibc 2.23 or later.
4927     * All hardware transactional memory builtins now correctly behave as
4928       memory barriers. Programmers can use #ifdef __TM_FENCE__ to
4929       determine whether their "old" compiler treats the builtins as
4930       barriers.
4931     * Split-stack support has been added for gccgo on PowerPC64 for both
4932       big- and little-endian (but not for 32-bit). The gold linker from
4933       at least binutils 2.25.1 must be available in the PATH when
4934       configuring and building gccgo to enable split stack. (The
4935       requirement for binutils 2.25.1 applies to PowerPC64 only.) The
4936       split-stack feature allows a small initial stack size to be
4937       allocated for each goroutine, which increases as needed.
4938     * GCC on PowerPC now supports the standard lround function.
4939     * A new configuration option ---with-advance-toolchain=at was added
4940       for PowerPC 64-bit GNU/Linux systems to use the header files,
4941       library files, and the dynamic linker from a specific Advance
4942       Toolchain release instead of the default versions that are provided
4943       by the GNU/Linux distribution. In general, this option is intended
4944       for the developers of GCC, and it is not intended for general use.
4945     * The "q", "S", "T", and "t" asm-constraints have been removed.
4946     * The "b", "B", "m", "M", and "W" format modifiers have been removed.
4947
4948  S/390, System z, IBM z Systems
4949
4950     * Support for the IBM z13 processor has been added. When using the
4951       -march=z13 option, the compiler will generate code making use of
4952       the new instructions and registers introduced with the vector
4953       extension facility. The -mtune=z13 option enables z13 specific
4954       instruction scheduling without making use of new instructions.
4955       Compiling code with -march=z13 reduces the default alignment of
4956       vector types bigger than 8 bytes to 8. This is an ABI change and
4957       care must be taken when linking modules compiled with different
4958       arch levels which interchange variables containing vector type
4959       values. For newly compiled code the GNU linker will emit a warning.
4960     * The -mzvector option enables a C/C++ language extension. This
4961       extension provides a new keyword vector which can be used to define
4962       vector type variables. (Note: This is not available when enforcing
4963       strict standard compliance e.g. with -std=c99. Either enable GNU
4964       extensions with e.g. -std=gnu99 or use __vector instead of vector.)
4965       Additionally a set of overloaded builtins is provided which is
4966       partially compatible to the PowerPC Altivec builtins. In order to
4967       make use of these builtins the vecintrin.h header file needs to be
4968       included.
4969     * The new command-line options -march=native, and -mtune=native are
4970       now available on native IBM z Systems. Specifying these options
4971       causes GCC to auto-detect the host CPU and choose the optimal
4972       setting for that system. If GCC is unable to detect the host CPU
4973       these options have no effect.
4974     * The IBM z Systems port now supports target attributes and pragmas.
4975       Please refer to the [20]documentation for details of available
4976       attributes and pragmas as well as usage instructions.
4977     * -fsplit-stack is now supported as part of the IBM z Systems port.
4978       This feature requires a recent gold linker to be used.
4979     * Support for the g5 and g6 -march=/-mtune= CPU level switches has
4980       been deprecated and will be removed in a future GCC release. -m31
4981       from now on defaults to -march=z900 if not specified otherwise.
4982       -march=native on a g5/g6 machine will default to -march=z900.
4983
4984  SH
4985
4986     * Support for SH5 / SH64 has been declared obsolete and will be
4987       removed in future releases.
4988     * Support for the FDPIC ABI has been added. It can be enabled using
4989       the new -mfdpic target option and --enable-fdpic configure option.
4990
4991  SPARC
4992
4993     * An ABI bug has been fixed in 64-bit mode. Unfortunately, this
4994       change will break binary compatibility with earlier releases for
4995       code it affects, but this should be pretty rare in practice. The
4996       conditions are: a 16-byte structure containing a double or a 8-byte
4997       vector in the second half is passed to a subprogram in slot #15,
4998       for example as 16th parameter if the first 15 ones have at most 8
4999       bytes. The double or vector was wrongly passed in floating-point
5000       register %d32 in lieu of on the stack as per the SPARC calling
5001       conventions.
5002
5003Operating Systems
5004
5005  AIX
5006
5007     * DWARF debugging support for AIX 7.1 has been enabled as an optional
5008       debugging format. A more recent Technology Level (TL) and GCC built
5009       with that level are required for full exploitation of DWARF
5010       debugging capabilities.
5011
5012  Linux
5013
5014     * Support for the [21]musl C library was added for the AArch64, ARM,
5015       MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and
5016       x86_64 targets. It can be selected using the new -mmusl option in
5017       case musl is not the default libc. GCC defaults to musl libc if it
5018       is built with a target triplet matching the *-linux-musl* pattern.
5019
5020  RTEMS
5021
5022     * The RTEMS thread model implementation changed. Mutexes now use
5023       self-contained objects defined in newlib <sys/lock.h> instead of
5024       Classic API semaphores. The keys for thread specific data and the
5025       once function are directly defined via <pthread.h>. Self-contained
5026       condition variables are provided via newlib <sys/lock.h>. The RTEMS
5027       thread model also supports C++11 threads.
5028     * OpenMP support now uses self-contained objects provided by newlib
5029       <sys/lock.h> and offers a significantly better performance compared
5030       to the POSIX configuration of libgomp. It is possible to configure
5031       thread pools for each scheduler instance via the environment
5032       variable GOMP_RTEMS_THREAD_POOLS.
5033
5034  Solaris
5035
5036     * Solaris 12 is now fully supported. Minimal support had already been
5037       present in GCC 5.3.
5038     * Solaris 12 provides a full set of startup files (crt1.o, crti.o,
5039       crtn.o), which GCC now prefers over its own ones.
5040     * Position independent executables (PIE) are now supported on Solaris
5041       12.
5042     * Constructor priority is now supported on Solaris 12 with the system
5043       linker.
5044     * libvtv has been ported to Solaris 11 and up.
5045
5046  Windows
5047
5048     * The option -mstackrealign is now automatically activated in 32-bit
5049       mode whenever the use of SSE instructions is requested.
5050
5051Other significant improvements
5052
5053     * The gcc and g++ driver programs will now provide suggestions for
5054       misspelled command-line options.
5055
5056$ gcc -static-libfortran test.f95
5057gcc: error: unrecognized command line option '-static-libfortran'; did you mean
5058'-static-libgfortran'?
5059
5060     * The --enable-default-pie configure option enables generation of PIE
5061       by default.
5062
5063                                    GCC 6.2
5064
5065   This is the [22]list of problem reports (PRs) from GCC's bug tracking
5066   system that are known to be fixed in the 6.2 release. This list might
5067   not be complete (that is, it is possible that some PRs that have been
5068   fixed are not listed here).
5069
5070Target Specific Changes
5071
5072  SPARC
5073
5074     * Support for --with-cpu-32 and --with-cpu-64 configure options has
5075       been added on bi-architecture platforms.
5076     * Support for the SPARC M7 (Niagara 7) processor has been added.
5077     * Support for the VIS 4.0 instruction set has been added.
5078
5079                                    GCC 6.3
5080
5081   This is the [23]list of problem reports (PRs) from GCC's bug tracking
5082   system that are known to be fixed in the 6.3 release. This list might
5083   not be complete (that is, it is possible that some PRs that have been
5084   fixed are not listed here).
5085
5086Target Specific Changes
5087
5088  IA-32/x86-64
5089
5090     * Support for the [24]deprecated pcommit instruction has been
5091       removed.
5092
5093                                    GCC 6.4
5094
5095   This is the [25]list of problem reports (PRs) from GCC's bug tracking
5096   system that are known to be fixed in the 6.4 release. This list might
5097   not be complete (that is, it is possible that some PRs that have been
5098   fixed are not listed here).
5099
5100Operating Systems
5101
5102  RTEMS
5103
5104     * The ABI changes on ARM so that no short enums are used by default.
5105
5106                                    GCC 6.5
5107
5108   This is the [26]list of problem reports (PRs) from GCC's bug tracking
5109   system that are known to be fixed in the 6.5 release. This list might
5110   not be complete (that is, it is possible that some PRs that have been
5111   fixed are not listed here).
5112
5113
5114    For questions related to the use of GCC, please consult these web
5115    pages and the [27]GCC manuals. If that fails, the
5116    [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5117    web pages and the development of GCC are welcome on our developer
5118    list at [29]gcc@gcc.gnu.org. All of [30]our lists have public
5119    archives.
5120
5121   Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and
5122   distribution of this entire article is permitted in any medium,
5123   provided this notice is preserved.
5124
5125   These pages are [32]maintained by the GCC team. Last modified
5126   2023-02-20.
5127
5128References
5129
5130   1. http://gcc.gnu.org/gcc-6/porting_to.html
5131   2. http://gcc.gnu.org/onlinedocs/index.html#current
5132   3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html
5133   4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151
5134   5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822
5135   6. https://www.openacc.org/
5136   7. https://gcc.gnu.org/wiki/OpenACC
5137   8. https://gcc.gnu.org/wiki/Offloading
5138   9. https://www.openmp.org/specifications/
5139  10. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266
5140  11. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf
5141  12. https://gcc.gnu.org/projects/cxx-status.html#cxx1z
5142  13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html
5143  14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch
5144  15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks
5145  16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option
5146  17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes
5147  18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes
5148  19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces
5149  20. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes
5150  21. http://www.musl-libc.org/
5151  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2
5152  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3
5153  24. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html
5154  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4
5155  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5
5156  27. https://gcc.gnu.org/onlinedocs/
5157  28. mailto:gcc-help@gcc.gnu.org
5158  29. mailto:gcc@gcc.gnu.org
5159  30. https://gcc.gnu.org/lists.html
5160  31. https://www.fsf.org/
5161  32. https://gcc.gnu.org/about.html
5162======================================================================
5163http://gcc.gnu.org/gcc-5/index.html
5164
5165                              GCC 5 Release Series
5166
5167   (This release series is no longer supported.)
5168
5169   October 10, 2017
5170
5171   The [1]GNU project and the GCC developers are pleased to announce the
5172   release of GCC 5.5.
5173
5174   This release is a bug-fix release, containing fixes for regressions in
5175   GCC 5.4 relative to previous releases of GCC.
5176
5177Release History
5178
5179   GCC 5.5
5180          October 10, 2017 ([2]changes, [3]documentation)
5181
5182   GCC 5.4
5183          June 3, 2016 ([4]changes, [5]documentation)
5184
5185   GCC 5.3
5186          December 4, 2015 ([6]changes, [7]documentation)
5187
5188   GCC 5.2
5189          July 16, 2015 ([8]changes, [9]documentation)
5190
5191   GCC 5.1
5192          April 22, 2015 ([10]changes, [11]documentation)
5193
5194References and Acknowledgements
5195
5196   GCC used to stand for the GNU C Compiler, but since the compiler
5197   supports several other languages aside from C, it now stands for the
5198   GNU Compiler Collection.
5199
5200   A list of [12]successful builds is updated as new information becomes
5201   available.
5202
5203   The GCC developers would like to thank the numerous people that have
5204   contributed new features, improvements, bug fixes, and other changes as
5205   well as test results to GCC. This [13]amazing group of volunteers is
5206   what makes GCC successful.
5207
5208   For additional information about GCC please refer to the [14]GCC
5209   project web site or contact the [15]GCC development mailing list.
5210
5211   To obtain GCC please use [16]our mirror sites or [17]our version
5212   control system.
5213
5214
5215    For questions related to the use of GCC, please consult these web
5216    pages and the [18]GCC manuals. If that fails, the
5217    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5218    web pages and the development of GCC are welcome on our developer
5219    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
5220    archives.
5221
5222   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
5223   distribution of this entire article is permitted in any medium,
5224   provided this notice is preserved.
5225
5226   These pages are [23]maintained by the GCC team. Last modified
5227   2022-10-26.
5228
5229References
5230
5231   1. http://www.gnu.org/
5232   2. http://gcc.gnu.org/gcc-5/changes.html
5233   3. http://gcc.gnu.org/onlinedocs/5.5.0/
5234   4. http://gcc.gnu.org/gcc-5/changes.html
5235   5. http://gcc.gnu.org/onlinedocs/5.4.0/
5236   6. http://gcc.gnu.org/gcc-5/changes.html
5237   7. http://gcc.gnu.org/onlinedocs/5.3.0/
5238   8. http://gcc.gnu.org/gcc-5/changes.html
5239   9. http://gcc.gnu.org/onlinedocs/5.2.0/
5240  10. http://gcc.gnu.org/gcc-5/changes.html
5241  11. http://gcc.gnu.org/onlinedocs/5.1.0/
5242  12. http://gcc.gnu.org/gcc-5/buildstat.html
5243  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5244  14. http://gcc.gnu.org/index.html
5245  15. mailto:gcc@gcc.gnu.org
5246  16. http://gcc.gnu.org/mirrors.html
5247  17. http://gcc.gnu.org/git.html
5248  18. https://gcc.gnu.org/onlinedocs/
5249  19. mailto:gcc-help@gcc.gnu.org
5250  20. mailto:gcc@gcc.gnu.org
5251  21. https://gcc.gnu.org/lists.html
5252  22. https://www.fsf.org/
5253  23. https://gcc.gnu.org/about.html
5254======================================================================
5255http://gcc.gnu.org/gcc-5/changes.html
5256
5257                              GCC 5 Release Series
5258                        Changes, New Features, and Fixes
5259
5260Caveats
5261
5262     * The default mode for C is now -std=gnu11 instead of -std=gnu89.
5263     * The C++ runtime library (libstdc++) uses a new ABI by default (see
5264       [1]below).
5265     * The Graphite framework for loop optimizations no longer requires
5266       the CLooG library, only ISL version 0.14 (recommended) or 0.12.2.
5267       The installation manual contains more information about
5268       requirements to build GCC.
5269     * The non-standard C++0x type traits has_trivial_default_constructor,
5270       has_trivial_copy_constructor and has_trivial_copy_assign have been
5271       deprecated and will be removed in a future version. The standard
5272       C++11 traits is_trivially_default_constructible,
5273       is_trivially_copy_constructible and is_trivially_copy_assignable
5274       should be used instead.
5275     * On AVR, support has been added for the devices
5276       ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer.
5277     * The AVR port uses a new scheme to describe supported devices: For
5278       each supported device the compiler provides a device-specific
5279       [2]spec file. If the compiler is used together with AVR-LibC, this
5280       requires at least GCC 5.2 and a version of AVR-LibC which
5281       implements [3]feature #44574.
5282
5283General Optimizer Improvements
5284
5285     * Inter-procedural optimization improvements:
5286          + An Identical Code Folding (ICF) pass (controlled via
5287            -fipa-icf) has been added. Compared to the identical code
5288            folding performed by the Gold linker this pass does not
5289            require function sections. It also performs merging before
5290            inlining, so inter-procedural optimizations are aware of the
5291            code re-use. On the other hand not all unifications performed
5292            by a linker are doable by GCC which must honor aliasing
5293            information. During link-time optimization of Firefox, this
5294            pass unifies about 31000 functions, that is 14% overall.
5295          + The devirtualization pass was significantly improved by adding
5296            better support for speculative devirtualization and dynamic
5297            type detection. About 50% of virtual calls in Firefox are now
5298            speculatively devirtualized during link-time optimization.
5299          + A new comdat localization pass allows the linker to eliminate
5300            more dead code in presence of C++ inline functions.
5301          + Virtual tables are now optimized. Local aliases are used to
5302            reduce dynamic linking time of C++ virtual tables on ELF
5303            targets and data alignment has been reduced to limit data
5304            segment bloat.
5305          + A new -fno-semantic-interposition option can be used to
5306            improve code quality of shared libraries where interposition
5307            of exported symbols is not allowed.
5308          + Write-only variables are now detected and optimized out.
5309          + With profile feedback the function inliner can now bypass
5310            --param inline-insns-auto and --param inline-insns-single
5311            limits for hot calls.
5312          + The IPA reference pass was significantly sped up making it
5313            feasible to enable -fipa-reference with -fprofile-generate.
5314            This also solves a bottleneck seen when building Chromium with
5315            link-time optimization.
5316          + The symbol table and call-graph API was reworked to C++ and
5317            simplified.
5318          + The interprocedural propagation of constants now also
5319            propagates alignments of pointer parameters. This for example
5320            means that the vectorizer often does not need to generate loop
5321            prologues and epilogues to make up for potential
5322            misalignments.
5323     * Link-time optimization improvements:
5324          + One Definition Rule based merging of C++ types has been
5325            implemented. Type merging enables better devirtualization and
5326            alias analysis. Streaming extra information needed to merge
5327            types adds about 2-6% of memory size and object size increase.
5328            This can be controlled by -flto-odr-type-merging.
5329          + Command-line optimization and target options are now streamed
5330            on a per-function basis and honored by the link-time
5331            optimizer. This change makes link-time optimization a more
5332            transparent replacement of per-file optimizations. It is now
5333            possible to build projects that require different optimization
5334            settings for different translation units (such as -ffast-math,
5335            -mavx, or -finline). Contrary to earlier GCC releases, the
5336            optimization and target options passed on the link command
5337            line are ignored.
5338            Note that this applies only to those command-line options that
5339            can be passed to optimize and target attributes. Command-line
5340            options affecting global code generation (such as -fpic),
5341            warnings (such as -Wodr), optimizations affecting the way
5342            static variables are optimized (such as -fcommon), debug
5343            output (such as -g), and --param parameters can be applied
5344            only to the whole link-time optimization unit. In these cases,
5345            it is recommended to consistently use the same options at both
5346            compile time and link time.
5347          + GCC bootstrap now uses slim LTO object files.
5348          + Memory usage and link times were improved. Tree merging was
5349            sped up, memory usage of GIMPLE declarations and types was
5350            reduced, and, support for on-demand streaming of variable
5351            constructors was added.
5352     * Feedback directed optimization improvements:
5353          + A new auto-FDO mode uses profiles collected by low overhead
5354            profiling tools (perf) instead of more expensive program
5355            instrumentation (via -fprofile-generate). SPEC2006 benchmarks
5356            on x86-64 improve by 4.7% with auto-FDO and by 7.3% with
5357            traditional feedback directed optimization.
5358          + Profile precision was improved in presence of C++ inline and
5359            extern inline functions.
5360          + The new gcov-tool utility allows manipulating profiles.
5361          + Profiles are now more tolerant to source file changes (this
5362            can be controlled by --param profile-func-internal-id).
5363     * Register allocation improvements:
5364          + A new local register allocator (LRA) sub-pass, controlled by
5365            -flra-remat, implements control-flow sensitive global register
5366            rematerialization. Instead of spilling and restoring a
5367            register value, it is recalculated if it is profitable. The
5368            sub-pass improved SPEC2000 generated code by 1% and 0.5%
5369            correspondingly on ARM and x86-64.
5370          + Reuse of the PIC hard register, instead of using a fixed
5371            register, was implemented on x86/x86-64 targets. This improves
5372            generated PIC code performance as more hard registers can be
5373            used. Shared libraries can significantly benefit from this
5374            optimization. Currently it is switched on only for x86/x86-64
5375            targets. As RA infrastructure is already implemented for PIC
5376            register reuse, other targets might follow this in the future.
5377          + A simple form of inter-procedural RA was implemented. When it
5378            is known that a called function does not use caller-saved
5379            registers, save/restore code is not generated around the call
5380            for such registers. This optimization can be controlled by
5381            -fipa-ra
5382          + LRA is now much more effective at generating spills of general
5383            registers into vector registers instead of memory on
5384            architectures (e.g., modern Intel processors) where this is
5385            profitable.
5386     * UndefinedBehaviorSanitizer gained a few new sanitization options:
5387          + -fsanitize=float-divide-by-zero: detect floating-point
5388            division by zero;
5389          + -fsanitize=float-cast-overflow: check that the result of
5390            floating-point type to integer conversions do not overflow;
5391          + -fsanitize=bounds: enable instrumentation of array bounds and
5392            detect out-of-bounds accesses;
5393          + -fsanitize=alignment: enable alignment checking, detect
5394            various misaligned objects;
5395          + -fsanitize=object-size: enable object size checking, detect
5396            various out-of-bounds accesses.
5397          + -fsanitize=vptr: enable checking of C++ member function calls,
5398            member accesses and some conversions between pointers to base
5399            and derived classes, detect if the referenced object does not
5400            have the correct dynamic type.
5401     * Pointer Bounds Checker, a bounds violation detector, has been added
5402       and can be enabled via -fcheck-pointer-bounds. Memory accesses are
5403       instrumented with run-time checks of used pointers against their
5404       bounds to detect pointer bounds violations (overflows). The Pointer
5405       Bounds Checker is available on x86/x86-64 GNU/Linux targets with a
5406       new ISA extension Intel MPX support. See the Pointer Bounds Checker
5407       [4]Wiki page for more details.
5408
5409New Languages and Language specific improvements
5410
5411     * [5]OpenMP 4.0 specification offloading features are now supported
5412       by the C, C++, and Fortran compilers. Generic changes:
5413          + Infrastructure (suitable for any vendor).
5414          + Testsuite which covers offloading from the [6]OpenMP 4.0
5415            Examples document.
5416       Specific for upcoming Intel Xeon Phi products:
5417          + Run-time library.
5418          + Card emulator.
5419     * GCC 5 includes a preliminary implementation of the OpenACC 2.0a
5420       specification. OpenACC is intended for programming accelerator
5421       devices such as GPUs. See [7]the OpenACC wiki page for more
5422       information.
5423
5424  C family
5425
5426     * The default setting of the -fdiagnostics-color= command-line option
5427       is now [8]configurable when building GCC using configuration option
5428       --with-diagnostics-color=. The possible values are: never, always,
5429       auto and auto-if-env. The new default auto uses color only when the
5430       standard error is a terminal. The default in GCC 4.9 was
5431       auto-if-env, which is equivalent to auto if there is a non-empty
5432       GCC_COLORS environment variable, and never otherwise. As in GCC
5433       4.9, an empty GCC_COLORS variable in the environment will always
5434       disable colors, no matter what the default is or what command-line
5435       options are used.
5436     * A new command-line option -Wswitch-bool has been added for the C
5437       and C++ compilers, which warns whenever a switch statement has an
5438       index of boolean type.
5439     * A new command-line option -Wlogical-not-parentheses has been added
5440       for the C and C++ compilers, which warns about "logical not" used
5441       on the left hand side operand of a comparison.
5442     * A new command-line option -Wsizeof-array-argument has been added
5443       for the C and C++ compilers, which warns when the sizeof operator
5444       is applied to a parameter that has been declared as an array in a
5445       function definition.
5446     * A new command-line option -Wbool-compare has been added for the C
5447       and C++ compilers, which warns about boolean expressions compared
5448       with an integer value different from true/false.
5449     * Full support for Cilk Plus has been added to the GCC compiler. Cilk
5450       Plus is an extension to the C and C++ languages to support data and
5451       task parallelism.
5452     * A new attribute no_reorder prevents reordering of selected symbols
5453       against other such symbols or inline assembler. This enables to
5454       link-time optimize the Linux kernel without having to resort to
5455       -fno-toplevel-reorder that disables several optimizations.
5456     * New preprocessor constructs, __has_include and __has_include_next,
5457       to test the availability of headers have been added.
5458       This demonstrates a way to include the header <optional> only if it
5459       is available:
5460
5461#ifdef __has_include
5462#  if __has_include(<optional>)
5463#    include <optional>
5464#    define have_optional 1
5465#  elif __has_include(<experimental/optional>)
5466#    include <experimental/optional>
5467#    define have_optional 1
5468#    define experimental_optional
5469#  else
5470#    define have_optional 0
5471#  endif
5472#endif
5473
5474       The header search paths for __has_include and __has_include_next
5475       are equivalent to those of the standard directive #include and the
5476       extension #include_next respectively.
5477     * A new built-in function-like macro to determine the existence of an
5478       attribute, __has_attribute, has been added. The equivalent built-in
5479       macro __has_cpp_attribute was added to C++ to support
5480       [9]Feature-testing recommendations for C++. The macro
5481       __has_attribute is added to all C-like languages as an extension:
5482
5483int
5484#ifdef __has_attribute
5485#  if __has_attribute(__noinline__)
5486  __attribute__((__noinline__))
5487#  endif
5488#endif
5489foo(int x);
5490
5491       If an attribute exists, a nonzero constant integer is returned. For
5492       standardized C++ attributes a date is returned, otherwise the
5493       constant returned is 1. Both __has_attribute and
5494       __has_cpp_attribute will add underscores to an attribute name if
5495       necessary to resolve the name. For C++11 and onwards the attribute
5496       may be scoped.
5497     * A new set of built-in functions for arithmetics with overflow
5498       checking has been added: __builtin_add_overflow,
5499       __builtin_sub_overflow and __builtin_mul_overflow and for
5500       compatibility with clang also other variants. These builtins have
5501       two integral arguments (which don't need to have the same type),
5502       the arguments are extended to infinite precision signed type, +, -
5503       or * is performed on those, and the result is stored in an integer
5504       variable pointed to by the last argument. If the stored value is
5505       equal to the infinite precision result, the built-in functions
5506       return false, otherwise true. The type of the integer variable that
5507       will hold the result can be different from the types of the first
5508       two arguments. The following snippet demonstrates how this can be
5509       used in computing the size for the calloc function:
5510
5511void *
5512calloc (size_t x, size_t y)
5513{
5514  size_t sz;
5515  if (__builtin_mul_overflow (x, y, &sz))
5516    return NULL;
5517  void *ret = malloc (sz);
5518  if (ret) memset (res, 0, sz);
5519  return ret;
5520}
5521
5522       On e.g. i?86 or x86-64 the above will result in a mul instruction
5523       followed by a jump on overflow.
5524     * The option -fextended-identifiers is now enabled by default for
5525       C++, and for C99 and later C versions. Various bugs in the
5526       implementation of extended identifiers have been fixed.
5527
5528  C
5529
5530     * The default mode has been changed to -std=gnu11.
5531     * A new command-line option -Wc90-c99-compat has been added to warn
5532       about features not present in ISO C90, but present in ISO C99.
5533     * A new command-line option -Wc99-c11-compat has been added to warn
5534       about features not present in ISO C99, but present in ISO C11.
5535     * It is possible to disable warnings about conversions between
5536       pointers that have incompatible types via a new warning option
5537       -Wno-incompatible-pointer-types; warnings about implicit
5538       incompatible integer to pointer and pointer to integer conversions
5539       via a new warning option -Wno-int-conversion; and warnings about
5540       qualifiers on pointers being discarded via a new warning option
5541       -Wno-discarded-qualifiers.
5542     * To allow proper use of const qualifiers with multidimensional
5543       arrays, GCC will not warn about incompatible pointer types anymore
5544       for conversions between pointers to arrays with and without const
5545       qualifier (except when using -pedantic). Instead, a new warning is
5546       emitted only if the const qualifier is lost. This can be controlled
5547       with a new warning option -Wno-discarded-array-qualifiers.
5548     * The C front end now generates more precise caret diagnostics.
5549     * The -pg command-line option now only affects the current file in an
5550       LTO build.
5551
5552  C++
5553
5554     * G++ now supports [10]C++14 variable templates.
5555     * -Wnon-virtual-dtor doesn't warn anymore for final classes.
5556     * Excessive template instantiation depth is now a fatal error. This
5557       prevents excessive diagnostics that usually do not help to identify
5558       the problem.
5559     * G++ and libstdc++ now implement the feature-testing macros from
5560       [11]Feature-testing recommendations for C++.
5561     * G++ now allows typename in a template template parameter.
5562
5563template<template<typename> typename X> struct D; // OK
5564
5565     * G++ now supports [12]C++14 aggregates with non-static data member
5566       initializers.
5567
5568struct A { int i, j = i; };
5569A a = { 42 }; // a.j is also 42
5570
5571     * G++ now supports [13]C++14 extended constexpr.
5572
5573constexpr int f (int i)
5574{
5575  int j = 0;
5576  for (; i > 0; --i)
5577    ++j;
5578  return j;
5579}
5580
5581constexpr int i = f(42); // i is 42
5582
5583     * G++ now supports the [14]C++14 sized deallocation functions.
5584
5585void operator delete (void *, std::size_t) noexcept;
5586void operator delete[] (void *, std::size_t) noexcept;
5587
5588     * A new One Definition Rule violation warning (controlled by -Wodr)
5589       detects mismatches in type definitions and virtual table contents
5590       during link-time optimization.
5591     * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help
5592       developers to annotate programs with final specifiers (or anonymous
5593       namespaces) to improve code generation. These warnings can be used
5594       at compile time, but they are more useful in combination with
5595       link-time optimization.
5596     * G++ no longer supports [15]N3639 variable length arrays, as they
5597       were removed from the C++14 working paper prior to ratification.
5598       GNU VLAs are still supported, so VLA support is now the same in
5599       C++14 mode as in C++98 and C++11 modes.
5600     * G++ now allows passing a non-trivially-copyable class via C
5601       varargs, which is conditionally-supported with
5602       implementation-defined semantics in the standard. This uses the
5603       same calling convention as a normal value parameter.
5604     * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So
5605       various mangling bugs are fixed, but G++ will still emit aliases
5606       with the old, wrong mangling where feasible. -Wabi=2 will warn
5607       about differences between ABI version 2 and the current setting.
5608     * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely
5609       to be unaffected, but -Wabi=8 will warn about a non-static data
5610       member with type std::nullptr_t which changes position due to this
5611       change.
5612
5613    Runtime Library (libstdc++)
5614
5615     * A [16]Dual ABI is provided by the library. A new ABI is enabled by
5616       default. The old ABI is still supported and can be used by defining
5617       the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++
5618       standard library headers.
5619     * A new implementation of std::string is enabled by default, using
5620       the small string optimization instead of copy-on-write reference
5621       counting.
5622     * A new implementation of std::list is enabled by default, with an
5623       O(1) size() function;
5624     * [17]Full support for C++11, including the following new features:
5625          + std::deque and std::vector<bool> meet the allocator-aware
5626            container requirements;
5627          + movable and swappable iostream classes;
5628          + support for std::align and std::aligned_union;
5629          + type traits std::is_trivially_copyable,
5630            std::is_trivially_constructible, std::is_trivially_assignable
5631            etc.;
5632          + I/O manipulators std::put_time, std::get_time, std::hexfloat
5633            and std::defaultfloat;
5634          + generic locale-aware std::isblank;
5635          + locale facets for Unicode conversion;
5636          + atomic operations for std::shared_ptr;
5637          + std::notify_all_at_thread_exit() and functions for making
5638            futures ready at thread exit.
5639     * Support for the C++11 hexfloat manipulator changes how the num_put
5640       facet formats floating point types when
5641       ios_base::fixed|ios_base::scientific is set in a stream's fmtflags.
5642       This change affects all language modes, even though the C++98
5643       standard gave no special meaning to that combination of flags. To
5644       prevent the use of hexadecimal notation for floating point types
5645       use str.unsetf(std::ios_base::floatfield) to clear the relevant
5646       bits in str.flags().
5647     * [18]Full experimental support for C++14, including the following
5648       new features:
5649          + std::is_final type trait;
5650          + heterogeneous comparison lookup in associative containers.
5651          + global functions cbegin, cend, rbegin, rend, crbegin, and
5652            crend for range access to containers, arrays and initializer
5653            lists.
5654     * [19]Improved experimental support for the Library Fundamentals TS,
5655       including:
5656          + class std::experimental::any;
5657          + function template std::experimental::apply;
5658          + function template std::experimental::sample;
5659          + function template std::experimental::search and related
5660            searcher types;
5661          + variable templates for type traits;
5662          + function template std::experimental::not_fn.
5663     * New random number distributions logistic_distribution and
5664       uniform_on_sphere_distribution as extensions.
5665     * [20]GDB Xmethods for containers and std::unique_ptr.
5666
5667  Fortran
5668
5669     * Compatibility notice:
5670          + The version of the module files (.mod) has been incremented.
5671          + For free-form source files [21]-Werror=line-truncation is now
5672            enabled by default. Note that comments exceeding the line
5673            length are not diagnosed. (For fixed-form source code, the
5674            same warning is available but turned off by default, such that
5675            excess characters are ignored. -ffree-line-length-n and
5676            -ffixed-line-length-n can be used to modify the default line
5677            lengths of 132 and 72 columns, respectively.)
5678          + The -Wtabs option is now more sensible: with -Wtabs the
5679            compiler warns if it encounters tabs and with -Wno-tabs this
5680            warning is turned off. Before, -Wno-tabs warned and -Wtabs
5681            disabled the warning. As before, this warning is also enabled
5682            by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts
5683            options of -std=.
5684     * Incomplete support for colorizing diagnostics emitted by gfortran
5685       has been added. The option [22]-fdiagnostics-color controls when
5686       color is used in diagnostics. The default value of this option can
5687       be [23]configured when building GCC. The GCC_COLORS environment
5688       variable can be used to customize the colors or disable coloring
5689       completely. Sample diagnostics output:
5690      $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90
5691      test.f90:6:1:
5692
5693       0 continue
5694       1
5695      Error: Zero is not a valid statement label at (1)
5696      test.f90:9:6:
5697
5698         USE foo
5699            1
5700      Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only]
5701
5702     * The -Wuse-without-only option has been added to warn when a USE
5703       statement has no ONLY qualifier and thus implicitly imports all
5704       public entities of the used module.
5705     * Formatted READ and WRITE statements now work correctly in
5706       locale-aware programs. For more information and potential caveats,
5707       see [24]Section 5.3 Thread-safety of the runtime library in the
5708       manual.
5709     * [25]Fortran 2003:
5710          + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and
5711            IEEE_ARITHMETIC) are now supported.
5712     * [26]Fortran 2008:
5713          + [27]Coarrays: Full experimental support of Fortran 2008's
5714            coarrays with -fcoarray=lib except for allocatable/pointer
5715            components of derived-type coarrays. GCC currently only ships
5716            with a single-image library (libcaf_single), but multi-image
5717            support based on MPI and GASNet is provided by the libraries
5718            of the [28]OpenCoarrays project.
5719     * TS18508 Additional Parallel Features in Fortran:
5720          + Support for the collective intrinsic subroutines CO_MAX,
5721            CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added,
5722            including -fcoarray=lib support.
5723          + Support for the new atomic intrinsics has been added,
5724            including -fcoarray=lib support.
5725     * Fortran 2015:
5726          + Support for IMPLICIT NONE (external, type).
5727          + ERROR STOP is now permitted in pure procedures.
5728
5729  Go
5730
5731     * GCC 5 provides a complete implementation of the Go 1.4.2 release.
5732     * Building GCC 5 with Go enabled will install two new programs:
5733       [29]go and [30]gofmt.
5734
5735libgccjit
5736
5737   New in GCC 5 is the ability to build GCC as a shared library for
5738   embedding in other processes (such as interpreters), suitable for
5739   Just-In-Time compilation to machine code.
5740
5741   The shared library has a [31]C API and a [32]C++ wrapper API providing
5742   some "syntactic sugar". There are also bindings available from 3rd
5743   parties for [33]Python and for [34]D.
5744
5745   For example, this library can be used by interpreters for [35]compiling
5746   functions from bytecode to machine code.
5747
5748   The library can also be used for ahead-of-time compilation, enabling
5749   GCC to be plugged into a pre-existing front end. An example of using
5750   this to build a compiler for an esoteric language we'll refer to as
5751   "brainf" can be seen [36]here.
5752
5753   libgccjit is licensed under the GPLv3 (or at your option, any later
5754   version)
5755
5756   It should be regarded as experimental at this time.
5757
5758New Targets and Target Specific Improvements
5759
5760  Reporting stack usage
5761
5762     * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the
5763       -fstack-usage option.
5764
5765  AArch64
5766
5767     * Code generation for the ARM Cortex-A57 processor has been improved.
5768       A more accurate instruction scheduling model for the processor is
5769       now used, and a number of compiler tuning parameters have been set
5770       to offer increased performance when compiling with -mcpu=cortex-a57
5771       or -mtune=cortex-a57.
5772     * A workaround for the ARM Cortex-A53 erratum 835769 has been added
5773       and can be enabled by giving the -mfix-cortex-a53-835769 option.
5774       Alternatively it can be enabled by default by configuring GCC with
5775       the --enable-fix-cortex-a53-835769 option.
5776     * The optional cryptographic extensions to the ARMv8-A architecture
5777       are no longer enabled by default when specifying the
5778       -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53
5779       options. To enable these extensions add +crypto to the value of
5780       -mcpu or -march e.g. -mcpu=cortex-a53+crypto.
5781     * Support has been added for the following processors (GCC
5782       identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and
5783       initial support for its big.LITTLE combination with the ARM
5784       Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx),
5785       Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1).
5786       The GCC identifiers can be used as arguments to the -mcpu or -mtune
5787       options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53.
5788       Using -mcpu=cortex-a72 requires a version of GNU binutils that has
5789       support for the Cortex-A72.
5790     * The transitional options -mlra and -mno-lra have been removed. The
5791       AArch64 backend now uses the local register allocator (LRA) only.
5792
5793  ARM
5794
5795     * Thumb-1 assembly code is now generated in unified syntax. The new
5796       option -masm-syntax-unified specifies whether inline assembly code
5797       is using unified syntax. By default the option is off which means
5798       non-unified syntax is used. However this is subject to change in
5799       future releases. Eventually the non-unified syntax will be
5800       deprecated.
5801     * It is now a configure-time error to use the --with-cpu configure
5802       option with either of --with-tune or --with-arch.
5803     * Code generation for the ARM Cortex-A57 processor has been improved.
5804       A more accurate instruction scheduling model for the processor is
5805       now used, and a number of compiler tuning parameters have been set
5806       to offer increased performance when compiling with -mcpu=cortex-a57
5807       or -mtune=cortex-a57.
5808     * Support has been added for the following processors (GCC
5809       identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and
5810       initial support for its big.LITTLE combination with the ARM
5811       Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and
5812       initial support for its big.LITTLE combination with the ARM
5813       Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7),
5814       Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1).
5815       The GCC identifiers can be used as arguments to the -mcpu or -mtune
5816       options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53.
5817       Using -mcpu=cortex-a72 requires a version of GNU binutils that has
5818       support for the Cortex-A72.
5819     * The deprecated option -mwords-little-endian has been removed.
5820     * The options -mapcs, -mapcs-frame, -mtpcs-frame and
5821       -mtpcs-leaf-frame which are only applicable to the old ABI have
5822       been deprecated.
5823     * The transitional options -mlra and -mno-lra have been removed. The
5824       ARM backend now uses the local register allocator (LRA) only.
5825
5826  AVR
5827
5828     * The compiler no more supports individual devices like ATmega8.
5829       Specifying, say, -mmcu=atmega8 triggers the usage of the
5830       device-specific [37]spec file specs-atmega8 which is part of the
5831       installation and describes options for the sub-processes like
5832       compiler proper, assembler and linker. You can add support for a
5833       new device -mmcu=mydevice as follows:
5834         1. In an empty directory /someplace, create a new directory
5835            device-specs.
5836         2. Copy a device spec file from the installed device-specs
5837            folder, follow the comments in that file and then save it as
5838            /someplace/device-specs/specs-mydevice.
5839         3. Add -B /someplace -mmcu=mydevice to the compiler's
5840            command-line options. Notice that /someplace must specify an
5841            absolute path and that mydevice must not start with "avr".
5842         4. Provided you have a device-specific library libmydevice.a
5843            available, you can put it at /someplace, dito for a
5844            device-specific startup file crtmydevice.o.
5845       The contents of the device spec files depend on the compiler's
5846       configuration, in particular on --with-avrlibc=no and whether or
5847       not it is configured for RTEMS.
5848     * A new command-line option -nodevicelib has been added. It prevents
5849       the compiler from linking against AVR-LibC's device-specific
5850       library libdevice.a.
5851     * The following three command-line options have been added:
5852
5853        -mrmw
5854                Set if the device supports the read-modify-write
5855                instructions LAC, LAS, LAT and XCH.
5856
5857        -mn-flash=size
5858                Specify the flash size of the device in units of 64 KiB,
5859                rounded up to the next integer as needed. This option
5860                affects the availability of the [38]AVR address-spaces.
5861
5862        -mskip-bug
5863                Set if the device is affected by the respective silicon
5864                bug.
5865
5866       In general, you don't need to set these options by hand. The new
5867       device-specific spec file will set them as needed.
5868
5869  IA-32/x86-64
5870
5871     * New ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of Intel's
5872       CPU codenamed Skylake Server was added to GCC. That includes inline
5873       assembly support, new intrinsics, and basic autovectorization.
5874       These new AVX-512 extensions are available via the following GCC
5875       switches: AVX-512 Vector Length EVEX feature: -mavx512vl, AVX-512
5876       Byte and Word instructions: -mavx512bw, AVX-512 Dword and Qword
5877       instructions: -mavx512dq, AVX-512 FMA-52 instructions: -mavx512ifma
5878       and for AVX-512 Vector Bit Manipulation Instructions: -mavx512vbmi.
5879     * New ISA extensions support Intel MPX was added to GCC. This new
5880       extension is available via the -mmpx compiler switch. Intel MPX is
5881       a set of processor features which, with compiler, run-time library
5882       and OS support, brings increased robustness to software by run-time
5883       checking pointer references against their bounds. In GCC Intel MPX
5884       is supported by Pointer Bounds Checker and libmpx run-time
5885       libraries.
5886     * The new -mrecord-mcount option for -pg generates a Linux kernel
5887       style table of pointers to mcount or __fentry__ calls at the
5888       beginning of functions. The new -mnop-mcount option in addition
5889       also generates nops in place of the __fentry__ or mcount call, so
5890       that a call per function can be later patched in. This can be used
5891       for low overhead tracing or hot code patching.
5892     * The new -malign-data option controls how GCC aligns variables.
5893       -malign-data=compat uses increased alignment compatible with GCC
5894       4.8 and earlier, -malign-data=abi uses alignment as specified by
5895       the psABI, and -malign-data=cacheline uses increased alignment to
5896       match the cache line size. -malign-data=compat is the default.
5897     * The new -mskip-rax-setup option skips setting up the RAX register
5898       when SSE is disabled and there are no variable arguments passed in
5899       vector registers. This can be used to optimize the Linux kernel.
5900
5901  MIPS
5902
5903     * MIPS Releases 3 and 5 are now directly supported. Use the
5904       command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5
5905       to enable code-generation for these processors.
5906     * The Imagination P5600 processor is now supported using the
5907       -march=p5600 command-line option.
5908     * The Cavium Octeon3 processor is now supported using the
5909       -march=octeon3 command-line option.
5910     * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6
5911       command-line options.
5912     * The o32 ABI has been modified and extended. The o32 64-bit
5913       floating-point register support is now obsolete and has been
5914       removed. It has been replaced by three ABI extensions FPXX, FP64A,
5915       and FP64. The meaning of the -mfp64 command-line option has
5916       changed. It is now used to enable the FP64A and FP64 ABI
5917       extensions.
5918          + The FPXX extension requires that code generated to access
5919            double-precision values use even-numbered registers. Code that
5920            adheres to this extension is link-compatible with all other
5921            o32 double-precision ABI variants and will execute correctly
5922            in all hardware FPU modes. The command-line options -mabi=32
5923            -mfpxx can be used to enable this extension. MIPS II is the
5924            minimum processor required.
5925          + The o32 FP64A extension requires that floating-point registers
5926            be 64-bit and odd-numbered single-precision registers are not
5927            allowed. Code that adheres to the o32 FP64A variant is
5928            link-compatible with all other o32 double-precision ABI
5929            variants. The command-line options -mabi=32 -mfp64
5930            -mno-odd-spreg can be used to enable this extension. MIPS32R2
5931            is the minimum processor required.
5932          + The o32 FP64 extension also requires that floating-point
5933            registers be 64-bit, but permits the use of single-precision
5934            registers. Code that adheres to the o32 FP64 variant is
5935            link-compatible with o32 FPXX and o32 FP64A variants only,
5936            i.e. it is not compatible with the original o32
5937            double-precision ABI. The command-line options -mabi=32 -mfp64
5938            -modd-spreg can be used to enable this extension. MIPS32R2 is
5939            the minimum processor required.
5940       The new ABI variants can be enabled by default using the configure
5941       time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32.
5942       It is strongly recommended that all vendors begin to set o32 FPXX
5943       as the default ABI. This will be required to run the generated code
5944       on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and
5945       MIPSR6 cores.
5946     * GCC will now pass all floating-point options to the assembler if
5947       GNU binutils 2.25 is used. As a result, any inline assembly code
5948       that uses hard-float instructions should be amended to include a
5949       .set directive to override the global assembler options when
5950       compiling for soft-float targets.
5951
5952  NDS32
5953
5954     * The variadic function ABI implementation is now compatible with
5955       past Andes toolchains where the caller uses registers to pass
5956       arguments and the callee is in charge of pushing them on stack.
5957     * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have
5958       been removed since they are not yet available in the nds32 port of
5959       GNU binutils.
5960     * A new option -mcmodel=[small|medium|large] supports varied code
5961       models on code generation. The -mgp-direct option became
5962       meaningless and can be discarded.
5963
5964  RX
5965
5966     * A new command line option -mno-allow-string-insns can be used to
5967       disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL,
5968       SWHILE and RMPA instructions. An erratum released by Renesas shows
5969       that it is unsafe to use these instructions on addresses within the
5970       I/O space of the processor. The new option can be used when the
5971       programmer is concerned that the I/O space might be accessed. The
5972       default is still to enable these instructions.
5973
5974  SH
5975
5976     * The compiler will now pass the appropriate --isa= option to the
5977       assembler.
5978     * The default handling for the GBR has been changed from call
5979       clobbered to call preserved. The old behavior can be reinstated by
5980       specifying the option -fcall-used-gbr.
5981     * Support for the SH4A fpchg instruction has been added which will be
5982       utilized when switching between single and double precision FPU
5983       modes.
5984     * The compiler no longer uses the __fpscr_values array for switching
5985       between single and double FPU precision modes on non-SH4A targets.
5986       Instead mode switching will now be performed by storing, modifying
5987       and reloading the FPSCR, so that other FPSCR bits are preserved
5988       across mode switches. The __fpscr_values array that is defined in
5989       libgcc is still present for backwards compatibility, but it will
5990       not be referenced by compiler generated code anymore.
5991     * New builtin functions __builtin_sh_get_fpscr and
5992       __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr
5993       function will mask the specified bits in such a way that the SZ, PR
5994       and FR mode bits will be preserved, while changing the other bits.
5995       These new functions do not reference the __fpscr_values array. The
5996       old functions __set_fpscr and __get_fpscr in libgcc which access
5997       the __fpscr_values array are still present for backwards
5998       compatibility, but their usage is highly discouraged.
5999     * Some improvements to code generated for __atomic built-in
6000       functions.
6001     * When compiling for SH2E the compiler will no longer force the usage
6002       of delay slots for conditional branch instructions bt and bf. The
6003       old behavior can be reinstated (e.g. to work around a hardware bug
6004       in the original SH7055) by specifying the new option
6005       -mcbranch-force-delay-slot.
6006
6007Operating Systems
6008
6009  AIX
6010
6011     * GCC now supports stabs debugging continuation lines to allow long
6012       stabs debug information without overflow that generates AIX linker
6013       errors.
6014
6015  DragonFly BSD
6016
6017     * GCC now supports the DragonFly BSD operating system.
6018
6019  FreeBSD
6020
6021     * GCC now supports the FreeBSD operating system for the arm port
6022       through the arm*-*-freebsd* target triplets.
6023
6024  VxWorks MILS
6025
6026     * GCC now supports the MILS (Multiple Independent Levels of Security)
6027       variant of WindRiver's VxWorks operating system for PowerPC
6028       targets.
6029
6030Other significant improvements
6031
6032     * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option
6033       to set the compiler to use.
6034
6035     * When the new command-line option -freport-bug is used, GCC
6036       automatically generates a developer-friendly reproducer whenever an
6037       internal compiler error is encountered.
6038
6039                                    GCC 5.2
6040
6041   This is the [39]list of problem reports (PRs) from GCC's bug tracking
6042   system that are known to be fixed in the 5.2 release. This list might
6043   not be complete (that is, it is possible that some PRs that have been
6044   fixed are not listed here).
6045
6046Target Specific Changes
6047
6048  IA-32/x86-64
6049
6050     * Support for new AMD instructions monitorx and mwaitx has been
6051       added. This includes new intrinsic and built-in support. It is
6052       enabled through option -mmwaitx. The instructions monitorx and
6053       mwaitx implement the same functionality as the old monitor and
6054       mwait instructions. In addition, mwaitx adds a configurable timer.
6055       The timer value is received as third argument and stored in
6056       register %ebx.
6057
6058  S/390, System z, IBM z Systems
6059
6060     * Support for the IBM z13 processor has been added. When using the
6061       -march=z13 option, the compiler will generate code making use of
6062       the new instructions and registers introduced with the vector
6063       extension facility. The -mtune=z13 option enables z13 specific
6064       instruction scheduling without making use of new instructions.
6065       Compiling code with -march=z13 reduces the default alignment of
6066       vector types bigger than 8 bytes to 8. This is an ABI change and
6067       care must be taken when linking modules compiled with different
6068       arch levels which interchange variables containing vector type
6069       values. For newly compiled code the GNU linker will emit a warning.
6070     * The -mzvector option enables a C/C++ language extension. This
6071       extension provides a new keyword vector which can be used to define
6072       vector type variables. (Note: This is not available when enforcing
6073       strict standard compliance e.g. with -std=c99. Either enable GNU
6074       extensions with e.g. -std=gnu99 or use __vector instead of vector.)
6075       Additionally a set of overloaded builtins is provided which is
6076       partially compatible to the PowerPC Altivec builtins. In order to
6077       make use of these builtins the vecintrin.h header file needs to be
6078       included.
6079
6080                                    GCC 5.3
6081
6082   This is the [40]list of problem reports (PRs) from GCC's bug tracking
6083   system that are known to be fixed in the 5.3 release. This list might
6084   not be complete (that is, it is possible that some PRs that have been
6085   fixed are not listed here).
6086
6087Target Specific Changes
6088
6089  IA-32/x86-64
6090
6091     * GCC now supports the Intel CPU named Skylake with AVX-512
6092       extensions through -march=skylake-avx512. The switch enables the
6093       following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW,
6094       AVX-512DQ.
6095
6096  S/390, System z, IBM z Systems
6097
6098     * With this version of GCC IBM z Systems support has been added to
6099       the GO runtime environment. GCC 5.3 has proven to be able to
6100       compile larger GO applications on IBM z Systems.
6101
6102                                    GCC 5.4
6103
6104   This is the [41]list of problem reports (PRs) from GCC's bug tracking
6105   system that are known to be fixed in the 5.4 release. This list might
6106   not be complete (that is, it is possible that some PRs that have been
6107   fixed are not listed here).
6108
6109                                    GCC 5.5
6110
6111   This is the [42]list of problem reports (PRs) from GCC's bug tracking
6112   system that are known to be fixed in the 5.5 release. This list might
6113   not be complete (that is, it is possible that some PRs that have been
6114   fixed are not listed here).
6115
6116Target Specific Changes
6117
6118  IA-32/x86-64
6119
6120     * Support for the [43]deprecated pcommit instruction has been
6121       removed.
6122
6123
6124    For questions related to the use of GCC, please consult these web
6125    pages and the [44]GCC manuals. If that fails, the
6126    [45]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6127    web pages and the development of GCC are welcome on our developer
6128    list at [46]gcc@gcc.gnu.org. All of [47]our lists have public
6129    archives.
6130
6131   Copyright (C) [48]Free Software Foundation, Inc. Verbatim copying and
6132   distribution of this entire article is permitted in any medium,
6133   provided this notice is preserved.
6134
6135   These pages are [49]maintained by the GCC team. Last modified
6136   2023-02-20.
6137
6138References
6139
6140   1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx
6141   2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html
6142   3. https://savannah.nongnu.org/bugs/?44574
6143   4. https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler
6144   5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf
6145   6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf
6146   7. https://gcc.gnu.org/wiki/OpenACC
6147   8. https://gcc.gnu.org/install/configure.html
6148   9. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations
6149  10. http://gcc.gnu.org/projects/cxx1y.html
6150  11. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations
6151  12. http://gcc.gnu.org/projects/cxx1y.html
6152  13. http://gcc.gnu.org/projects/cxx1y.html
6153  14. http://gcc.gnu.org/projects/cxx1y.html
6154  15. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html
6155  16. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html
6156  17. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011
6157  18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014
6158  19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014
6159  20. https://sourceware.org/gdb/current/onlinedocs/gdb#Xmethods-In-Python
6160  21. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html
6161  22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html
6162  23. https://gcc.gnu.org/install/configure.html
6163  24. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html
6164  25. https://gcc.gnu.org/wiki/Fortran2003Status
6165  26. https://gcc.gnu.org/wiki/Fortran2008Status
6166  27. https://gcc.gnu.org/wiki/Coarray
6167  28. http://www.opencoarrays.org/
6168  29. https://pkg.go.dev/cmd/go
6169  30. https://pkg.go.dev/cmd/gofmt
6170  31. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html
6171  32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html
6172  33. https://github.com/davidmalcolm/pygccjit
6173  34. https://github.com/ibuclaw/gccjitd
6174  35. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html
6175  36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html
6176  37. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html
6177  38. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html
6178  39. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2
6179  40. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3
6180  41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4
6181  42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5
6182  43. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html
6183  44. https://gcc.gnu.org/onlinedocs/
6184  45. mailto:gcc-help@gcc.gnu.org
6185  46. mailto:gcc@gcc.gnu.org
6186  47. https://gcc.gnu.org/lists.html
6187  48. https://www.fsf.org/
6188  49. https://gcc.gnu.org/about.html
6189======================================================================
6190http://gcc.gnu.org/gcc-4.9/index.html
6191
6192                             GCC 4.9 Release Series
6193
6194   (This release series is no longer supported.)
6195
6196   Aug 3, 2016
6197
6198   The [1]GNU project and the GCC developers are pleased to announce the
6199   release of GCC 4.9.4.
6200
6201   This release is a bug-fix release, containing fixes for regressions in
6202   GCC 4.9.3 relative to previous releases of GCC.
6203
6204Release History
6205
6206   GCC 4.9.4
6207          Aug 3, 2016 ([2]changes, [3]documentation)
6208
6209   GCC 4.9.3
6210          June 26, 2015 ([4]changes, [5]documentation)
6211
6212   GCC 4.9.2
6213          October 30, 2014 ([6]changes, [7]documentation)
6214
6215   GCC 4.9.1
6216          July 16, 2014 ([8]changes, [9]documentation)
6217
6218   GCC 4.9.0
6219          April 22, 2014 ([10]changes, [11]documentation)
6220
6221References and Acknowledgements
6222
6223   GCC used to stand for the GNU C Compiler, but since the compiler
6224   supports several other languages aside from C, it now stands for the
6225   GNU Compiler Collection.
6226
6227   A list of [12]successful builds is updated as new information becomes
6228   available.
6229
6230   The GCC developers would like to thank the numerous people that have
6231   contributed new features, improvements, bug fixes, and other changes as
6232   well as test results to GCC. This [13]amazing group of volunteers is
6233   what makes GCC successful.
6234
6235   For additional information about GCC please refer to the [14]GCC
6236   project web site or contact the [15]GCC development mailing list.
6237
6238   To obtain GCC please use [16]our mirror sites or [17]our version
6239   control system.
6240
6241
6242    For questions related to the use of GCC, please consult these web
6243    pages and the [18]GCC manuals. If that fails, the
6244    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6245    web pages and the development of GCC are welcome on our developer
6246    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
6247    archives.
6248
6249   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
6250   distribution of this entire article is permitted in any medium,
6251   provided this notice is preserved.
6252
6253   These pages are [23]maintained by the GCC team. Last modified
6254   2022-10-26.
6255
6256References
6257
6258   1. http://www.gnu.org/
6259   2. http://gcc.gnu.org/gcc-4.9/changes.html
6260   3. http://gcc.gnu.org/onlinedocs/4.9.4/
6261   4. http://gcc.gnu.org/gcc-4.9/changes.html
6262   5. http://gcc.gnu.org/onlinedocs/4.9.3/
6263   6. http://gcc.gnu.org/gcc-4.9/changes.html
6264   7. http://gcc.gnu.org/onlinedocs/4.9.2/
6265   8. http://gcc.gnu.org/gcc-4.9/changes.html
6266   9. http://gcc.gnu.org/onlinedocs/4.9.1/
6267  10. http://gcc.gnu.org/gcc-4.9/changes.html
6268  11. https://gcc.gnu.org/onlinedocs/4.9.0/
6269  12. http://gcc.gnu.org/gcc-4.9/buildstat.html
6270  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6271  14. http://gcc.gnu.org/index.html
6272  15. mailto:gcc@gcc.gnu.org
6273  16. http://gcc.gnu.org/mirrors.html
6274  17. http://gcc.gnu.org/git.html
6275  18. https://gcc.gnu.org/onlinedocs/
6276  19. mailto:gcc-help@gcc.gnu.org
6277  20. mailto:gcc@gcc.gnu.org
6278  21. https://gcc.gnu.org/lists.html
6279  22. https://www.fsf.org/
6280  23. https://gcc.gnu.org/about.html
6281======================================================================
6282http://gcc.gnu.org/gcc-4.9/changes.html
6283
6284                             GCC 4.9 Release Series
6285                        Changes, New Features, and Fixes
6286
6287Caveats
6288
6289     * The mudflap run time checker has been removed. The mudflap options
6290       remain, but do nothing.
6291     * Support for a number of older systems and recently unmaintained or
6292       untested target ports of GCC has been declared obsolete in GCC 4.9.
6293       Unless there is activity to revive them, the next release of GCC
6294       will have their sources permanently removed.
6295       The following ports for individual systems on particular
6296       architectures have been obsoleted:
6297          + Solaris 9 (*-*-solaris2.9). Details can be found in the
6298            [1]announcement.
6299     * On AArch64, the singleton vector types int64x1_t, uint64x1_t and
6300       float64x1_t exported by arm_neon.h are defined to be the same as
6301       their base types. This results in incorrect application of
6302       parameter passing rules to arguments of types int64x1_t and
6303       uint64x1_t, with respect to the AAPCS64 ABI specification. In
6304       addition, names of C++ functions with parameters of these types
6305       (including float64x1_t) are not mangled correctly. The current
6306       typedef declarations also unintentionally allow implicit casting
6307       between singleton vector types and their base types. These issues
6308       will be resolved in a near future release. See [2]PR60825 for more
6309       information.
6310
6311   More information on porting to GCC 4.9 from previous versions of GCC
6312   can be found in the [3]porting guide for this release.
6313
6314General Optimizer Improvements
6315
6316     * AddressSanitizer, a fast memory error detector, is now available on
6317       ARM.
6318     * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior
6319       detector, has been added and can be enabled via
6320       -fsanitize=undefined. Various computations will be instrumented to
6321       detect undefined behavior at runtime. UndefinedBehaviorSanitizer is
6322       currently available for the C and C++ languages.
6323     * Link-time optimization (LTO) improvements:
6324          + Type merging was rewritten. The new implementation is
6325            significantly faster and uses less memory.
6326          + Better partitioning algorithm resulting in less streaming
6327            during link time.
6328          + Early removal of virtual methods reduces the size of object
6329            files and improves link-time memory usage and compile time.
6330          + Function bodies are now loaded on-demand and released early
6331            improving overall memory usage at link time.
6332          + C++ hidden keyed methods can now be optimized out.
6333          + When using a linker plugin, compiling with the -flto option
6334            now generates slim object files (.o) which only contain
6335            intermediate language representation for LTO. Use
6336            -ffat-lto-objects to create files which contain additionally
6337            the object code. To generate static libraries suitable for LTO
6338            processing, use gcc-ar and gcc-ranlib; to list symbols from a
6339            slim object file use gcc-nm. (This requires that ar, ranlib
6340            and nm have been compiled with plugin support.)
6341       Memory usage building Firefox with debug enabled was reduced from
6342       15GB to 3.5GB; link time from 1700 seconds to 350 seconds.
6343     * Inter-procedural optimization improvements:
6344          + New type inheritance analysis module improving
6345            devirtualization. Devirtualization now takes into account
6346            anonymous name-spaces and the C++11 final keyword.
6347          + New speculative devirtualization pass (controlled by
6348            -fdevirtualize-speculatively.
6349          + Calls that were speculatively made direct are turned back to
6350            indirect where direct call is not cheaper.
6351          + Local aliases are introduced for symbols that are known to be
6352            semantically equivalent across shared libraries improving
6353            dynamic linking times.
6354     * Feedback directed optimization improvements:
6355          + Profiling of programs using C++ inline functions is now more
6356            reliable.
6357          + New time profiling determines typical order in which functions
6358            are executed.
6359          + A new function reordering pass (controlled by
6360            -freorder-functions) significantly reduces startup time of
6361            large applications. Until binutils support is completed, it is
6362            effective only with link-time optimization.
6363          + Feedback driven indirect call removal and devirtualization now
6364            handle cross-module calls when link-time optimization is
6365            enabled.
6366
6367New Languages and Language specific improvements
6368
6369     * Version 4.0 of the [4]OpenMP specification is now supported in the
6370       C and C++ compilers and starting with the 4.9.1 release also in the
6371       Fortran compiler. The new -fopenmp-simd option can be used to
6372       enable OpenMP's SIMD directives while ignoring other OpenMP
6373       directives. The new [5]-fsimd-cost-model= option permits to tune
6374       the vectorization cost model for loops annotated with OpenMP and
6375       Cilk Plus simd directives. -Wopenmp-simd warns when the current
6376       cost model overrides simd directives set by the user.
6377     * The -Wdate-time option has been added for the C, C++ and Fortran
6378       compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__
6379       macros are used. Those macros might prevent bit-wise-identical
6380       reproducible compilations.
6381
6382  Ada
6383
6384     * GNAT switched to Ada 2012 instead of Ada 2005 by default.
6385
6386  C family
6387
6388     * Support for colorizing diagnostics emitted by GCC has been added.
6389       The [6]-fdiagnostics-color=auto will enable it when outputting to
6390       terminals, -fdiagnostics-color=always unconditionally. The
6391       GCC_COLORS environment variable can be used to customize the colors
6392       or disable coloring. If GCC_COLORS variable is present in the
6393       environment, the default is -fdiagnostics-color=auto, otherwise
6394       -fdiagnostics-color=never.
6395       Sample diagnostics output:
6396    $ g++ -fdiagnostics-color=always -S -Wall test.C
6397    test.C: In function `int foo()':
6398    test.C:1:14: warning: no return statement in function returning non-void [-W
6399return-type]
6400     int foo () { }
6401                  ^
6402    test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use
6403 -ftemplate-depth= to increase the maximum) instantiating `struct X<100>'
6404     template <int N> struct X { static const int value = X<N-1>::value; }; temp
6405late struct X<1000>;
6406                                                  ^
6407    test.C:2:46:   recursively required from `const int X<999>::value'
6408    test.C:2:46:   required from `const int X<1000>::value'
6409    test.C:2:88:   required from here
6410
6411    test.C:2:46: error: incomplete type `X<100>' used in nested name specifier
6412
6413     * With the new [7]#pragma GCC ivdep, the user can assert that there
6414       are no loop-carried dependencies which would prevent concurrent
6415       execution of consecutive iterations using SIMD (single instruction
6416       multiple data) instructions.
6417     * Support for Cilk Plus has been added and can be enabled with the
6418       -fcilkplus option. Cilk Plus is an extension to the C and C++
6419       languages to support data and task parallelism. The present
6420       implementation follows ABI version 1.2; all features but _Cilk_for
6421       have been implemented.
6422
6423  C
6424
6425     * ISO C11 atomics (the _Atomic type specifier and qualifier and the
6426       <stdatomic.h> header) are now supported.
6427     * ISO C11 generic selections (_Generic keyword) are now supported.
6428     * ISO C11 thread-local storage (_Thread_local, similar to GNU C
6429       __thread) is now supported.
6430     * ISO C11 support is now at a similar level of completeness to ISO
6431       C99 support: substantially complete modulo bugs, extended
6432       identifiers (supported except for corner cases when
6433       -fextended-identifiers is used), floating-point issues (mainly but
6434       not entirely relating to optional C99 features from Annexes F and
6435       G) and the optional Annexes K (Bounds-checking interfaces) and L
6436       (Analyzability).
6437     * A new C extension __auto_type provides a subset of the
6438       functionality of C++11 auto in GNU C.
6439
6440  C++
6441
6442     * The G++ implementation of [8]C++1y return type deduction for normal
6443       functions has been updated to conform to [9]N3638, the proposal
6444       accepted into the working paper. Most notably, it adds
6445       decltype(auto) for getting decltype semantics rather than the
6446       template argument deduction semantics of plain auto:
6447
6448int& f();
6449         auto  i1 = f(); // int
6450decltype(auto) i2 = f(); // int&
6451
6452     * G++ supports [10]C++1y lambda capture initializers:
6453
6454[x = 42]{ ... };
6455
6456       Actually, they have been accepted since GCC 4.5, but now the
6457       compiler doesn't warn about them with -std=c++1y, and supports
6458       parenthesized and brace-enclosed initializers as well.
6459     * G++ supports [11]C++1y variable length arrays. G++ has supported
6460       GNU/C99-style VLAs for a long time, but now additionally supports
6461       initializers and lambda capture by reference. In C++1y mode G++
6462       will complain about VLA uses that are not permitted by the draft
6463       standard, such as forming a pointer to VLA type or applying sizeof
6464       to a VLA variable. Note that it now appears that VLAs will not be
6465       part of C++14, but will be part of a separate document and then
6466       perhaps C++17.
6467
6468void f(int n) {
6469  int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3
6470  [&a]{ for (int i : a) { cout << i << endl; } }();
6471  &a; // error, taking address of VLA
6472}
6473
6474     * G++ supports the [12]C++1y [[deprecated]] attribute modulo bugs in
6475       the underlying [[gnu::deprecated]] attribute. Classes and functions
6476       can be marked deprecated and a diagnostic message added:
6477
6478class A;
6479int bar(int n);
6480#if __cplusplus > 201103
6481class [[deprecated("A is deprecated in C++14; Use B instead")]] A;
6482[[deprecated("bar is unsafe; use foo() instead")]]
6483int bar(int n);
6484
6485int foo(int n);
6486class B;
6487#endif
6488A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead
6489int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo
6490o() instead
6491
6492     * G++ supports [13]C++1y digit separators. Long numeric literals can
6493       be subdivided with a single quote ' to enhance readability:
6494
6495int i = 1048576;
6496int j = 1'048'576;
6497int k = 0x10'0000;
6498int m = 0'004'000'000;
6499int n = 0b0001'0000'0000'0000'0000'0000;
6500
6501double x = 1.602'176'565e-19;
6502double y = 1.602'176'565e-1'9;
6503
6504     * G++ supports [14]C++1y generic (polymorphic) lambdas.
6505
6506// a functional object that will increment any type
6507auto incr = [](auto x) { return x++; };
6508
6509     * As a GNU extension, G++ supports explicit template parameter syntax
6510       for generic lambdas. This can be combined in the expected way with
6511       the standard auto syntax.
6512
6513// a functional object that will add two like-type objects
6514auto add = [] <typename T> (T a, T b) { return a + b; };
6515
6516     * G++ supports unconstrained generic functions as specified by �4.1.2
6517       and �5.1.1 of [15]N3889: Concepts Lite Specification. Briefly, auto
6518       may be used as a type-specifier in a parameter declaration of any
6519       function declarator in order to introduce an implicit function
6520       template parameter, akin to generic lambdas.
6521
6522// the following two function declarations are equivalent
6523auto incr(auto x) { return x++; }
6524template <typename T>
6525auto incr(T x) { return x++; }
6526
6527    Runtime Library (libstdc++)
6528
6529     * [16]Improved support for C++11, including:
6530          + support for <regex>;
6531          + The associative containers in <map> and <set> and the
6532            unordered associative containers in <unordered_map> and
6533            <unordered_set> meet the allocator-aware container
6534            requirements;
6535     * [17]Improved experimental support for the upcoming ISO C++
6536       standard, C++14, including:
6537          + fixing constexpr member functions without const;
6538          + implementation of the std::exchange() utility function;
6539          + addressing tuples by type;
6540          + implemention of std::make_unique;
6541          + implemention of std::shared_lock;
6542          + making std::result_of SFINAE-friendly;
6543          + adding operator() to std::integral_constant;
6544          + adding user-defined literals for standard library types
6545            std::basic_string, std::chrono::duration, and std::complex;
6546          + adding two range overloads to non-modifying sequence oprations
6547            std::equal and std::mismatch;
6548          + adding IO manipulators for quoted strings;
6549          + adding constexpr members to <utility>, <complex>, <chrono>,
6550            and some containers;
6551          + adding compile-time std::integer_sequence;
6552          + adding cleaner transformation traits;
6553          + making <functional>s operator functors easier to use and more
6554            generic;
6555     * An implementation of std::experimental::optional.
6556     * An implementation of std::experimental::string_view.
6557     * The non-standard function std::copy_exception has been deprecated
6558       and will be removed in a future version. std::make_exception_ptr
6559       should be used instead.
6560
6561  Fortran
6562
6563     * Compatibility notice:
6564          + Module files: The version of the module files (.mod) has been
6565            incremented; additionally, module files are now compressed.
6566            Fortran MODULEs compiled by earlier GCC versions have to be
6567            recompiled, when they are USEd by files compiled with GCC 4.9.
6568            GCC 4.9 is not able to read .mod files of earlier GCC
6569            versions; attempting to do so gives an error message. Note:
6570            The ABI of the produced assembler data itself has not changed:
6571            object files and libraries are fully compatible with older
6572            versions (except as stated below).
6573          + ABI changes:
6574               o The [18]argument passing ABI has changed for scalar dummy
6575                 arguments of type INTEGER, REAL, COMPLEX and LOGICAL,
6576                 which have both the VALUE and the OPTIONAL attributes.
6577               o To support finalization the virtual table associated with
6578                 polymorphic variables has changed. Code containing CLASS
6579                 should be recompiled, including all files which define
6580                 derived types involved in the type definition used by
6581                 polymorphic variables. (Note: Due to the incremented
6582                 module version, trying to mix old code with new code will
6583                 usually give an error message.)
6584          + GNU Fortran no longer deallocates allocatable variables or
6585            allocatable components of variables declared in the main
6586            program. Since Fortran 2008, the standard explicitly states
6587            that variables declared in the Fortran main program
6588            automatically have the SAVE attribute.
6589          + When opening files, the close-on-exec flag is set if the
6590            system supports such a feature. This is generally considered
6591            good practice these days, but if there is a need to pass file
6592            descriptors to child processes the parent process must now
6593            remember to clear the close-on-exec flag by calling fcntl(),
6594            e.g. via ISO_C_BINDING, before executing the child process.
6595     * The deprecated command-line option -fno-whole-file has been
6596       removed. (-fwhole-file is the default since GCC 4.6.)
6597       -fwhole-file/-fno-whole-file continue to be accepted but do not
6598       influence the code generation.
6599     * The compiler no longer unconditionally warns about DO loops with
6600       zero iterations. This warning is now controlled by the -Wzerotrip
6601       option, which is implied by -Wall.
6602     * The new NO_ARG_CHECK attribute of the [19]!GCC$ directive can be
6603       used to disable the type-kind-rank (TKR) argument check for a dummy
6604       argument. The feature is similar to ISO/IEC TS 29133:2012's
6605       TYPE(*), except that it additionally also disables the rank check.
6606       Variables with NO_ARG_CHECK have to be dummy arguments and may only
6607       be used as argument to ISO_C_BINDING's C_LOC and as actual argument
6608       to another NO_ARG_CHECK dummy argument; also the other constraints
6609       of TYPE(*) apply. The dummy arguments should be declared as scalar
6610       or assumed-size variable of type type(*) (recommended) - or of type
6611       integer, real, complex or logical. With NO_ARG_CHECK, a pointer to
6612       the data without further type or shape information is passed,
6613       similar to C's void*. Note that also TS 29113's
6614       type(*),dimension(..) accepts arguments of any type and rank;
6615       contrary to NO_ARG_CHECK assumed-rank arguments pass an array
6616       descriptor which contains the array shape and stride of the
6617       argument.
6618     * [20]Fortran 2003:
6619          + Finalization is now supported. It is currently only done for a
6620            subset of those situations in which it should occur.
6621          + Experimental support for scalar character components with
6622            deferred length (i.e. allocatable string length) in derived
6623            types has been added. (Deferred-length character variables are
6624            supported since GCC 4.6.)
6625     * [21]Fortran 2008:
6626          + When STOP or ERROR STOP are used to terminate the execution
6627            and any exception (but inexact) is signaling, a warning is
6628            printed to ERROR_UNIT, indicating which exceptions are
6629            signaling. The [22]-ffpe-summary= command-line option can be
6630            used to fine-tune for which exceptions the warning should be
6631            shown.
6632          + Rounding on input (READ) is now handled on systems where
6633            strtod honours the rounding mode. (For output, rounding is
6634            supported since GCC 4.5.) Note that for input, the compatible
6635            rounding mode is handled as nearest (i.e., rounding to an even
6636            least significant [cf. IEC 60559:1989] for a tie, while
6637            compatible rounds away from zero in that case).
6638
6639  Go
6640
6641     * GCC 4.9 provides a complete implementation of the Go 1.2.1 release.
6642
6643New Targets and Target Specific Improvements
6644
6645  AArch64
6646
6647     * The ARMv8-A crypto and CRC instructions are now supported through
6648       intrinsics. These are enabled when the architecture supports these
6649       and are available through the -march=armv8-a+crc and
6650       -march=armv8-a+crypto options.
6651     * Initial support for ILP32 has now been added to the compiler. This
6652       is now available through the command-line option -mabi=ilp32.
6653       Support for ILP32 is considered experimental as the ABI
6654       specification is still beta.
6655     * Coverage of more of the ISA including the SIMD extensions has been
6656       added. The Advanced SIMD intrinsics have also been improved.
6657     * The new local register allocator (LRA) is now on by default for the
6658       AArch64 backend.
6659     * The REE (Redundant extension elimination) pass has now been enabled
6660       by default for the AArch64 backend.
6661     * Tuning for the Cortex-A53 and Cortex-A57 has been improved.
6662     * Initial big.LITTLE tuning support for the combination of Cortex-A57
6663       and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
6664       option.
6665     * A number of structural changes have been made to both the ARM and
6666       AArch64 backends to facilitate improved code-generation.
6667     * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769
6668       has been added and can be enabled by giving the
6669       -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
6670       default by configuring GCC with the --enable-fix-cortex-a53-835769
6671       option.
6672
6673  ARC
6674
6675     * A port for Synopsys Designware ARC has been contributed by Embecosm
6676       and Synopsys Inc.
6677
6678  ARM
6679
6680     * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been
6681       disabled by default. This was found to generate better code in only
6682       a small number of cases. It can be turned back on with the
6683       -mneon-for-64bits option.
6684     * Further support for the ARMv8-A architecture, notably implementing
6685       the restriction around IT blocks in the Thumb32 instruction set has
6686       been added. The -mrestrict-it option can be used with
6687       -march=armv7-a or the -march=armv7ve options to make code
6688       generation fully compatible with the deprecated instructions in
6689       ARMv8-A.
6690     * Support has now been added for the ARMv7ve variant of the
6691       architecture. This can be used by the -march=armv7ve option.
6692     * The ARMv8-A crypto and CRC instructions are now supported through
6693       intrinsics and are available through the -march=armv8-a+crc and
6694       mfpu=crypto-neon-fp-armv8 options.
6695     * LRA is now on by default for the ARM target. This can be turned off
6696       using the -mno-lra option. This option is a purely transitionary
6697       command-line option and will be removed in a future release. We are
6698       interested in any bug reports regarding functional and performance
6699       regressions with LRA.
6700     * A new option -mslow-flash-data to improve performance of programs
6701       fetching data on slow flash memory has now been introduced for the
6702       ARMv7-M profile cores.
6703     * A new option -mpic-data-is-text-relative for targets that allows
6704       data segments to be relative to text segments has been added. This
6705       is on by default for all targets except VxWorks RTP.
6706     * A number of infrastructural changes have been made to both the ARM
6707       and AArch64 backends to facilitate improved code-generation.
6708     * GCC now supports Cortex-A12 and the Cortex-R7 through the
6709       -mcpu=cortex-a12 and -mcpu=cortex-r7 options.
6710     * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the
6711       -mcpu=cortex-a57 and -mcpu=cortex-a53 options.
6712     * Initial big.LITTLE tuning support for the combination of Cortex-A57
6713       and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
6714       option. Similar support was added for the combination of Cortex-A15
6715       and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option.
6716     * Further performance optimizations for the Cortex-A15 and the
6717       Cortex-M4 have been added.
6718     * A number of code generation improvements for Thumb2 to reduce code
6719       size when compiling for the M-profile processors.
6720
6721  AVR
6722
6723     * A new command-line option -mfract-convert-truncate has been added.
6724       It allows compiler to use truncation instead of rounding towards
6725       zero for fractional fixed-point types.
6726
6727  IA-32/x86-64
6728
6729     * -mfpmath=sse is now implied by -ffast-math on all targets where
6730       SSE2 is supported.
6731     * Intel AVX-512 support was added to GCC. That includes inline
6732       assembly support, new registers and extending existing ones, new
6733       intrinsics (covered by corresponding testsuite), and basic
6734       autovectorization. AVX-512 instructions are available via the
6735       following GCC switches: AVX-512 foundation instructions: -mavx512f,
6736       AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and
6737       reciprocal instructions: -mavx512er, AVX-512 conflict detection
6738       instructions: -mavx512cd.
6739     * It is now possible to call x86 intrinsics from select functions in
6740       a file that are tagged with the corresponding target attribute
6741       without having to compile the entire file with the -mxxx option.
6742       This improves the usability of x86 intrinsics and is particularly
6743       useful when doing [23]Function Multiversioning.
6744     * GCC now supports the new Intel microarchitecture named Silvermont
6745       through -march=silvermont.
6746     * GCC now supports the new Intel microarchitecture named Broadwell
6747       through -march=broadwell.
6748     * Optimizing for other Intel microarchitectures have been renamed to
6749       -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell.
6750     * -march=generic has been retuned for better support of Intel core
6751       and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel
6752       Pentium-M, and Pentium4 based CPUs is no longer considered
6753       important for generic.
6754     * -mtune=intel can now be used to generate code running well on the
6755       most current Intel processors, which are Haswell and Silvermont for
6756       GCC 4.9.
6757     * Support to encode 32-bit assembly instructions in 16-bit format is
6758       now available through the -m16 command-line option.
6759     * Better inlining of memcpy and memset that is aware of value ranges
6760       and produces shorter alignment prologues.
6761     * -mno-accumulate-outgoing-args is now honored when unwind
6762       information is output. Argument accumulation is also now turned off
6763       for portions of programs optimized for size.
6764     * Support for new AMD family 15h processors (Excavator core) is now
6765       available through the -march=bdver4 and -mtune=bdver4 options.
6766
6767  MSP430
6768
6769     * A new command-line option -mcpu= has been added to the MSP430
6770       backend. This option is used to specify the ISA to be used.
6771       Accepted values are msp430 (the default), msp430x and msp430xv2.
6772       The ISA is no longer deduced from the -mmcu= option as there are
6773       far too many different MCU names. The -mmcu= option is still
6774       supported, and this is still used to select linker scripts and
6775       generate a C preprocessor symbol that will be recognised by the
6776       msp430.h header file.
6777
6778  NDS32
6779
6780     * A new nds32 port supports the 32-bit architecture from Andes
6781       Technology Corporation.
6782     * The port provides initial support for the V2, V3, V3m instruction
6783       set architectures.
6784
6785  Nios II
6786
6787     * A port for the Altera Nios II has been contributed by Mentor
6788       Graphics.
6789
6790  PowerPC / PowerPC64 / RS6000
6791
6792     * GCC now supports Power ISA 2.07, which includes support for
6793       Hardware Transactional Memory (HTM), Quadword atomics and several
6794       VMX and VSX additions, including Crypto, 64-bit integer, 128-bit
6795       integer and decimal integer operations.
6796     * Support for the POWER8 processor is now available through the
6797       -mcpu=power8 and -mtune=power8 options.
6798     * The libitm library has been modified to add a HTM fastpath that
6799       automatically uses POWER's HTM hardware instructions when it is
6800       executing on a HTM enabled processor.
6801     * Support for the new powerpc64le-linux platform has been added. It
6802       defaults to generating code that conforms to the ELFV2 ABI.
6803
6804  S/390, System z
6805
6806     * Support for the Transactional Execution Facility included with the
6807       IBM zEnterprise zEC12 processor has been added. A set of GCC style
6808       builtins as well as XLC style builtins are provided. The builtins
6809       are enabled by default when using the -march=zEC12 option but can
6810       explicitly be disabled with -mno-htm. Using the GCC builtins also
6811       libitm supports hardware transactions on S/390.
6812     * The hotpatch features allows to prepare functions for hotpatching.
6813       A certain amount of bytes is reserved before the function entry
6814       label plus a NOP is inserted at its very beginning to implement a
6815       backward jump when applying a patch. The feature can either be
6816       enabled per compilation unit via the command-line option -mhotpatch
6817       or per function using the hotpatch attribute.
6818     * The shrink wrap optimization is now supported on S/390 and enabled
6819       by default.
6820     * A major rework of the routines to determine which registers need to
6821       be saved and restored in function prologue/epilogue now allow to
6822       use floating point registers as save slots. This will happen for
6823       certain leaf function with -march=z10 or higher.
6824     * The LRA rtl pass replaces reload by default on S/390.
6825
6826  RX
6827
6828     * The port now allows to specify the RX100, RX200, and RX600
6829       processors with the command-line options -mcpu=rx100, -mcpu=rx200
6830       and -mcpu=rx600.
6831
6832  SH
6833
6834     * Minor improvements to code generated for integer arithmetic and
6835       code that involves the T bit.
6836     * Added support for the SH2A clips and clipu instructions. The
6837       compiler will now try to utilize them for min/max expressions such
6838       as max (-128, min (127, x)).
6839     * Added support for the cmp/str instruction through built-in
6840       functions such as __builtin_strlen. When not optimizing for size,
6841       the compiler will now expand calls to e.g. strlen as an inlined
6842       sequences which utilize the cmp/str instruction.
6843     * Improved code generated around volatile memory loads and stores.
6844     * The option -mcbranchdi has been deprecated. Specifying it will
6845       result in a warning and will not influence code generation.
6846     * The option -mcmpeqdi has been deprecated. Specifying it will result
6847       in a warning and will not influence code generation.
6848
6849GCC 4.9.1
6850
6851   This is the [24]list of problem reports (PRs) from GCC's bug tracking
6852   system that are known to be fixed in the 4.9.1 release. This list might
6853   not be complete (that is, it is possible that some PRs that have been
6854   fixed are not listed here).
6855
6856   Version 4.0 of the OpenMP specification is supported even in Fortran,
6857   not just C and C++.
6858
6859GCC 4.9.2
6860
6861   This is the [25]list of problem reports (PRs) from GCC's bug tracking
6862   system that are known to be fixed in the 4.9.2 release. This list might
6863   not be complete (that is, it is possible that some PRs that have been
6864   fixed are not listed here).
6865
6866GCC 4.9.3
6867
6868   This is the [26]list of problem reports (PRs) from GCC's bug tracking
6869   system that are known to be fixed in the 4.9.3 release. This list might
6870   not be complete (that is, it is possible that some PRs that have been
6871   fixed are not listed here).
6872
6873GCC 4.9.4
6874
6875   This is the [27]list of problem reports (PRs) from GCC's bug tracking
6876   system that are known to be fixed in the 4.9.4 release. This list might
6877   not be complete (that is, it is possible that some PRs that have been
6878   fixed are not listed here).
6879
6880
6881    For questions related to the use of GCC, please consult these web
6882    pages and the [28]GCC manuals. If that fails, the
6883    [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6884    web pages and the development of GCC are welcome on our developer
6885    list at [30]gcc@gcc.gnu.org. All of [31]our lists have public
6886    archives.
6887
6888   Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and
6889   distribution of this entire article is permitted in any medium,
6890   provided this notice is preserved.
6891
6892   These pages are [33]maintained by the GCC team. Last modified
6893   2023-01-19.
6894
6895References
6896
6897   1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html
6898   2. https://gcc.gnu.org/PR60825
6899   3. https://gcc.gnu.org/gcc-4.9/porting_to.html
6900   4. https://www.openmp.org/specifications/
6901   5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908
6902   6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252
6903   7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html
6904   8. http://gcc.gnu.org/projects/cxx1y.html
6905   9. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html
6906  10. http://gcc.gnu.org/projects/cxx1y.html
6907  11. http://gcc.gnu.org/projects/cxx1y.html
6908  12. http://gcc.gnu.org/projects/cxx1y.html
6909  13. http://gcc.gnu.org/projects/cxx1y.html
6910  14. http://gcc.gnu.org/projects/cxx1y.html
6911  15. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf
6912  16. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011
6913  17. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014
6914  18. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html
6915  19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html
6916  20. https://gcc.gnu.org/wiki/Fortran2003Status
6917  21. https://gcc.gnu.org/wiki/Fortran2008Status
6918  22. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html
6919  23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html
6920  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1
6921  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2
6922  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3
6923  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4
6924  28. https://gcc.gnu.org/onlinedocs/
6925  29. mailto:gcc-help@gcc.gnu.org
6926  30. mailto:gcc@gcc.gnu.org
6927  31. https://gcc.gnu.org/lists.html
6928  32. https://www.fsf.org/
6929  33. https://gcc.gnu.org/about.html
6930======================================================================
6931http://gcc.gnu.org/gcc-4.8/index.html
6932
6933                             GCC 4.8 Release Series
6934
6935   (This release series is no longer supported.)
6936
6937   June 23, 2015
6938
6939   The [1]GNU project and the GCC developers are pleased to announce the
6940   release of GCC 4.8.5.
6941
6942   This release is a bug-fix release, containing fixes for regressions in
6943   GCC 4.8.4 relative to previous releases of GCC.
6944
6945Release History
6946
6947   GCC 4.8.5
6948          June 23, 2015 ([2]changes, [3]documentation)
6949
6950   GCC 4.8.4
6951          December 19, 2014 ([4]changes, [5]documentation)
6952
6953   GCC 4.8.3
6954          May 22, 2014 ([6]changes, [7]documentation)
6955
6956   GCC 4.8.2
6957          October 16, 2013 ([8]changes, [9]documentation)
6958
6959   GCC 4.8.1
6960          May 31, 2013 ([10]changes, [11]documentation)
6961
6962   GCC 4.8.0
6963          March 22, 2013 ([12]changes, [13]documentation)
6964
6965References and Acknowledgements
6966
6967   GCC used to stand for the GNU C Compiler, but since the compiler
6968   supports several other languages aside from C, it now stands for the
6969   GNU Compiler Collection.
6970
6971   A list of [14]successful builds is updated as new information becomes
6972   available.
6973
6974   The GCC developers would like to thank the numerous people that have
6975   contributed new features, improvements, bug fixes, and other changes as
6976   well as test results to GCC. This [15]amazing group of volunteers is
6977   what makes GCC successful.
6978
6979   For additional information about GCC please refer to the [16]GCC
6980   project web site or contact the [17]GCC development mailing list.
6981
6982   To obtain GCC please use [18]our mirror sites or [19]our version
6983   control system.
6984
6985
6986    For questions related to the use of GCC, please consult these web
6987    pages and the [20]GCC manuals. If that fails, the
6988    [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6989    web pages and the development of GCC are welcome on our developer
6990    list at [22]gcc@gcc.gnu.org. All of [23]our lists have public
6991    archives.
6992
6993   Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and
6994   distribution of this entire article is permitted in any medium,
6995   provided this notice is preserved.
6996
6997   These pages are [25]maintained by the GCC team. Last modified
6998   2022-10-26.
6999
7000References
7001
7002   1. http://www.gnu.org/
7003   2. http://gcc.gnu.org/gcc-4.8/changes.html
7004   3. https://gcc.gnu.org/onlinedocs/4.8.5/
7005   4. http://gcc.gnu.org/gcc-4.8/changes.html
7006   5. https://gcc.gnu.org/onlinedocs/4.8.4/
7007   6. http://gcc.gnu.org/gcc-4.8/changes.html
7008   7. https://gcc.gnu.org/onlinedocs/4.8.3/
7009   8. http://gcc.gnu.org/gcc-4.8/changes.html
7010   9. https://gcc.gnu.org/onlinedocs/4.8.2/
7011  10. http://gcc.gnu.org/gcc-4.8/changes.html
7012  11. https://gcc.gnu.org/onlinedocs/4.8.1/
7013  12. http://gcc.gnu.org/gcc-4.8/changes.html
7014  13. https://gcc.gnu.org/onlinedocs/4.8.0/
7015  14. http://gcc.gnu.org/gcc-4.8/buildstat.html
7016  15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7017  16. http://gcc.gnu.org/index.html
7018  17. mailto:gcc@gcc.gnu.org
7019  18. http://gcc.gnu.org/mirrors.html
7020  19. http://gcc.gnu.org/git.html
7021  20. https://gcc.gnu.org/onlinedocs/
7022  21. mailto:gcc-help@gcc.gnu.org
7023  22. mailto:gcc@gcc.gnu.org
7024  23. https://gcc.gnu.org/lists.html
7025  24. https://www.fsf.org/
7026  25. https://gcc.gnu.org/about.html
7027======================================================================
7028http://gcc.gnu.org/gcc-4.8/changes.html
7029
7030                             GCC 4.8 Release Series
7031                        Changes, New Features, and Fixes
7032
7033Caveats
7034
7035   GCC now uses C++ as its implementation language. This means that to
7036   build GCC from sources, you will need a C++ compiler that understands
7037   C++ 2003. For more details on the rationale and specific changes,
7038   please refer to the [1]C++ conversion page.
7039
7040   To enable the Graphite framework for loop optimizations you now need
7041   CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from
7042   the [2]GCC infrastructure directory. The installation manual contains
7043   more information about requirements to build GCC.
7044
7045   GCC now uses a more aggressive analysis to derive an upper bound for
7046   the number of iterations of loops using constraints imposed by language
7047   standards. This may cause non-conforming programs to no longer work as
7048   expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new
7049   option, -fno-aggressive-loop-optimizations, was added to disable this
7050   aggressive analysis. In some loops that have known constant number of
7051   iterations, but undefined behavior is known to occur in the loop before
7052   reaching or during the last iteration, GCC will warn about the
7053   undefined behavior in the loop instead of deriving lower upper bound of
7054   the number of iterations for the loop. The warning can be disabled with
7055   -Wno-aggressive-loop-optimizations.
7056
7057   On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules
7058   for the layout of vectors that could lead to wrong code being
7059   generated. Vectors larger than 8 bytes in size are now by default
7060   aligned to an 8-byte boundary. This is an ABI change: code that makes
7061   explicit use of vector types may be incompatible with binary objects
7062   built with older versions of GCC. Auto-vectorized code is not affected
7063   by this change.
7064
7065   On AVR, support has been removed for the command-line option
7066   -mshort-calls deprecated in GCC 4.7.
7067
7068   On AVR, the configure option --with-avrlibc supported since GCC 4.7.2
7069   is turned on per default for all non-RTEMS configurations. This option
7070   arranges for a better integration of [3]AVR Libc with avr-gcc. For
7071   technical details, see [4]PR54461. To turn off the option in non-RTEMS
7072   configurations, use --with-avrlibc=no. If the compiler is configured
7073   for RTEMS, the option is always turned off.
7074
7075   More information on porting to GCC 4.8 from previous versions of GCC
7076   can be found in the [5]porting guide for this release.
7077
7078General Optimizer Improvements (and Changes)
7079
7080     * DWARF4 is now the default when generating DWARF debug information.
7081       When -g is used on a platform that uses DWARF debugging
7082       information, GCC will now default to -gdwarf-4
7083       -fno-debug-types-section.
7084       GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
7085       consumers support DWARF4 by default. Before GCC 4.8 the default
7086       version used was DWARF2. To make GCC 4.8 generate an older DWARF
7087       version use -g together with -gdwarf-2 or -gdwarf-3. The default
7088       for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf.
7089     * A new general optimization level, -Og, has been introduced. It
7090       addresses the need for fast compilation and a superior debugging
7091       experience while providing a reasonable level of run-time
7092       performance. Overall experience for development should be better
7093       than the default optimization level -O0.
7094     * A new option -ftree-partial-pre was added to control the partial
7095       redundancy elimination (PRE) optimization. This option is enabled
7096       by default at the -O3 optimization level, and it makes PRE more
7097       aggressive.
7098     * The option -fconserve-space has been removed; it was no longer
7099       useful on most targets since GCC supports putting variables into
7100       BSS without making them common.
7101     * The struct reorg and matrix reorg optimizations (command-line
7102       options -fipa-struct-reorg and -fipa-matrix-reorg) have been
7103       removed. They did not always work correctly, nor did they work with
7104       link-time optimization (LTO), hence were only applicable to
7105       programs consisting of a single translation unit.
7106     * Several scalability bottle-necks have been removed from GCC's
7107       optimization passes. Compilation of extremely large functions, e.g.
7108       due to the use of the flatten attribute in the "Eigen" C++ linear
7109       algebra templates library, is significantly faster than previous
7110       releases of GCC.
7111     * Link-time optimization (LTO) improvements:
7112          + LTO partitioning has been rewritten for better reliability and
7113            maintanibility. Several important bugs leading to link
7114            failures have been fixed.
7115     * Interprocedural optimization improvements:
7116          + A new symbol table has been implemented. It builds on existing
7117            callgraph and varpool modules and provide a new API. Unusual
7118            symbol visibilities and aliases are handled more consistently
7119            leading to, for example, more aggressive unreachable code
7120            removal with LTO.
7121          + The inline heuristic can now bypass limits on the size of of
7122            inlined functions when the inlining is particularly
7123            profitable. This happens, for example, when loop bounds or
7124            array strides get propagated.
7125          + Values passed through aggregates (either by value or
7126            reference) are now propagated at the inter-procedural level
7127            leading to better inlining decisions (for example in the case
7128            of Fortran array descriptors) and devirtualization.
7129     * [6]AddressSanitizer , a fast memory error detector, has been added
7130       and can be enabled via -fsanitize=address. Memory access
7131       instructions will be instrumented to detect heap-, stack-, and
7132       global-buffer overflow as well as use-after-free bugs. To get nicer
7133       stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is
7134       available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on
7135       x86-64 Darwin.
7136     * [7]ThreadSanitizer has been added and can be enabled via
7137       -fsanitize=thread. Instructions will be instrumented to detect data
7138       races. The ThreadSanitizer is available on x86-64 GNU/Linux.
7139     * A new local register allocator (LRA) has been implemented, which
7140       replaces the 26 year old reload pass and improves generated code
7141       quality. For now it is active on the IA-32 and x86-64 targets.
7142     * Support for transactional memory has been implemented on the
7143       following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and
7144       Alpha.
7145
7146New Languages and Language specific improvements
7147
7148  C family
7149
7150     * Each diagnostic emitted now includes the original source line and a
7151       caret '^' indicating the column. The option
7152       -fno-diagnostics-show-caret suppresses this information.
7153     * The option -ftrack-macro-expansion=2 is now enabled by default.
7154       This allows the compiler to display the macro expansion stack in
7155       diagnostics. Combined with the caret information, an example
7156       diagnostic showing these two features is:
7157
7158t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float
7159')
7160 #define MYMAX(A,B)    __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _
7161_b = (B); __a < __b ? __b : __a; })
7162
7163              ^
7164t.c:7:7: note: in expansion of macro 'MYMAX'
7165   X = MYMAX(P, F);
7166       ^
7167
7168     * A new -Wsizeof-pointer-memaccess warning has been added (also
7169       enabled by -Wall) to warn about suspicious length parameters to
7170       certain string and memory built-in functions if the argument uses
7171       sizeof. This warning warns e.g. about memset (ptr, 0, sizeof
7172       (ptr)); if ptr is not an array, but a pointer, and suggests a
7173       possible fix, or about memcpy (&foo, ptr, sizeof (&foo));.
7174     * The new option -Wpedantic is an alias for -pedantic, which is now
7175       deprecated. The forms -Wno-pedantic, -Werror=pedantic, and
7176       -Wno-error=pedantic work in the same way as for any other -W
7177       option. One caveat is that -Werror=pedantic is not equivalent to
7178       -pedantic-errors, since the latter makes into errors some warnings
7179       that are not controlled by -Wpedantic, and the former only affects
7180       diagnostics that are disabled when using -Wno-pedantic.
7181     * The option -Wshadow no longer warns if a declaration shadows a
7182       function declaration, unless the former declares a function or
7183       pointer to function, because this is [8]a common and valid case in
7184       real-world code.
7185
7186  C++
7187
7188     * G++ now implements the [9]C++11 thread_local keyword; this differs
7189       from the GNU __thread keyword primarily in that it allows dynamic
7190       initialization and destruction semantics. Unfortunately, this
7191       support requires a run-time penalty for references to
7192       non-function-local thread_local variables defined in a different
7193       translation unit even if they don't need dynamic initialization, so
7194       users may want to continue to use __thread for TLS variables with
7195       static initialization semantics.
7196       If the programmer can be sure that no use of the variable in a
7197       non-defining TU needs to trigger dynamic initialization (either
7198       because the variable is statically initialized, or a use of the
7199       variable in the defining TU will be executed before any uses in
7200       another TU), they can avoid this overhead with the
7201       -fno-extern-tls-init option.
7202       OpenMP threadprivate variables now also support dynamic
7203       initialization and destruction by the same mechanism.
7204     * G++ now implements the [10]C++11 attribute syntax, e.g.
7205
7206[[noreturn]] void f();
7207
7208       and also the alignment specifier, e.g.
7209
7210alignas(double) int i;
7211
7212     * G++ now implements [11]C++11 inheriting constructors, e.g.
7213
7214struct A { A(int); };
7215struct B: A { using A::A; }; // defines B::B(int)
7216B b(42); // OK
7217
7218     * As of GCC 4.8.1, G++ implements the change to decltype semantics
7219       from [12]N3276.
7220
7221struct A f();
7222decltype(f()) g();    // OK, return type of f() is not required to be complete.
7223
7224     * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g.
7225
7226struct A { int f() &; };
7227int i = A().f();  // error, f() requires an lvalue object
7228
7229     * G++ now supports a -std=c++1y option for experimentation with
7230       features proposed for the next revision of the standard, expected
7231       around 2014. Currently the only difference from -std=c++11 is
7232       support for return type deduction in normal functions, as proposed
7233       in [14]N3386. Status of C++1y features in GCC 4.8 can be found
7234       [15]here.
7235     * The G++ namespace association extension, __attribute ((strong)),
7236       has been deprecated. Inline namespaces should be used instead.
7237     * G++ now supports a -fext-numeric-literal option to control whether
7238       GNU numeric literal suffixes are accepted as extensions or
7239       processed as C++11 user-defined numeric literal suffixes. The flag
7240       is on (use suffixes for GNU literals) by default for -std=gnu++*,
7241       and -std=c++98. The flag is off (use suffixes for user-defined
7242       literals) by default for -std=c++11 and later.
7243
7244    Runtime Library (libstdc++)
7245
7246     * [16]Improved experimental support for the new ISO C++ standard,
7247       C++11, including:
7248          + forward_list meets the allocator-aware container requirements;
7249          + this_thread::sleep_for(), this_thread::sleep_until() and
7250            this_thread::yield() are defined without requiring the
7251            configure option --enable-libstdcxx-time;
7252     * Improvements to <random>:
7253          + SSE optimized normal_distribution.
7254          + Use of hardware RNG instruction for random_device on new x86
7255            processors (requires the assembler to support the
7256            instruction.)
7257       and <ext/random>:
7258          + New random number engine simd_fast_mersenne_twister_engine
7259            with an optimized SSE implementation.
7260          + New random number distributions beta_distribution,
7261            normal_mv_distribution, rice_distribution,
7262            nakagami_distribution, pareto_distribution, k_distribution,
7263            arcsine_distribution, hoyt_distribution.
7264     * Added --disable-libstdcxx-verbose configure option to disable
7265       diagnostic messages issued when a process terminates abnormally.
7266       This may be useful for embedded systems to reduce the size of
7267       executables that link statically to the library.
7268
7269  Fortran
7270
7271     * Compatibility notice:
7272          + Module files: The version of module files (.mod) has been
7273            incremented. Fortran MODULEs compiled by earlier GCC versions
7274            have to be recompiled, when they are USEd by files compiled
7275            with GCC 4.8. GCC 4.8 is not able to read .mod files created
7276            by earlier versions; attempting to do so gives an error
7277            message.
7278            Note: The ABI of the produced assembler data itself has not
7279            changed; object files and libraries are fully compatible with
7280            older versions except as noted below.
7281          + ABI: Some internal names (used in the assembler/object file)
7282            have changed for symbols declared in the specification part of
7283            a module. If an affected module - or a file using it via use
7284            association - is recompiled, the module and all files which
7285            directly use such symbols have to be recompiled as well. This
7286            change only affects the following kind of module symbols:
7287               o Procedure pointers. Note: C-interoperable function
7288                 pointers (type(c_funptr)) are not affected nor are
7289                 procedure-pointer components.
7290               o Deferred-length character strings.
7291     * The [17]BACKTRACE intrinsic subroutine has been added. It shows a
7292       backtrace at an arbitrary place in user code; program execution
7293       continues normally afterwards.
7294     * The [18]-Wc-binding-type warning option has been added (disabled by
7295       default). It warns if the a variable might not be C interoperable;
7296       in particular, if the variable has been declared using an intrinsic
7297       type with default kind instead of using a kind parameter defined
7298       for C interoperability in the intrinsic ISO_C_Binding module.
7299       Before, this warning was always printed. The -Wc-binding-type
7300       option is enabled by -Wall.
7301     * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line
7302       options have been added, which diagnose when code is inserted for
7303       automatic (re)allocation of a variable during assignment. This
7304       option can be used to decide whether it is safe to use
7305       [20]-fno-realloc-lhs. Additionally, it can be used to find
7306       automatic (re)allocation in hot loops. (For arrays, replacing
7307       "var=" by "var(:)=" disables the automatic reallocation.)
7308     * The [21]-Wcompare-reals command-line option has been added. When
7309       this is set, warnings are issued when comparing REAL or COMPLEX
7310       types for equality and inequality; consider replacing a == b by
7311       abs(a-b) < eps with a suitable eps. -Wcompare-reals is enabled by
7312       -Wextra.
7313     * The [22]-Wtarget-lifetime command-line option has been added
7314       (enabled with -Wall), which warns if the pointer in a pointer
7315       assignment might outlive its target.
7316     * Reading floating point numbers which use "q" for the exponential
7317       (such as 4.0q0) is now supported as vendor extension for better
7318       compatibility with old data files. It is strongly recommended to
7319       use for I/O the equivalent but standard conforming "e" (such as
7320       4.0e0).
7321       (For Fortran source code, consider replacing the "q" in
7322       floating-point literals by a kind parameter (e.g. 4.0e0_qp with a
7323       suitable qp). Note that - in Fortran source code - replacing "q" by
7324       a simple "e" is not equivalent.)
7325     * The GFORTRAN_TMPDIR environment variable for specifying a
7326       non-default directory for files opened with STATUS="SCRATCH", is
7327       not used anymore. Instead gfortran checks the POSIX/GNU standard
7328       TMPDIR environment variable. If TMPDIR is not defined, gfortran
7329       falls back to other methods to determine the directory for
7330       temporary files as documented in the [23]user manual.
7331     * [24]Fortran 2003:
7332          + Support for unlimited polymorphic variables (CLASS(*)) has
7333            been added. Nonconstant character lengths are not yet
7334            supported.
7335     * [25]TS 29113:
7336          + Assumed types (TYPE(*)) are now supported.
7337          + Experimental support for assumed-rank arrays (dimension(..))
7338            has been added. Note that currently gfortran's own array
7339            descriptor is used, which is different from the one defined in
7340            TS29113, see [26]gfortran's header file or use the [27]Chasm
7341            Language Interoperability Tools.
7342
7343  Go
7344
7345     * GCC 4.8.2 provides a complete implementation of the Go 1.1.2
7346       release.
7347     * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1
7348       release. The library support is not quite complete.
7349     * Go has been tested on GNU/Linux and Solaris platforms for various
7350       processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may
7351       work on other platforms as well.
7352
7353New Targets and Target Specific Improvements
7354
7355  AArch64
7356
7357     * A new port has been added to support AArch64, the new 64-bit
7358       architecture from ARM. Note that this is a separate port from the
7359       existing 32-bit ARM port.
7360     * The port provides initial support for the Cortex-A53 and the
7361       Cortex-A57 processors with the command line options
7362       -mcpu=cortex-a53 and -mcpu=cortex-a57.
7363     * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769
7364       has been added and can be enabled by giving the
7365       -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
7366       default by configuring GCC with the --enable-fix-cortex-a53-835769
7367       option.
7368
7369  ARM
7370
7371     * Initial support has been added for the AArch32 extensions defined
7372       in the ARMv8 architecture.
7373     * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs.
7374     * A new option, -mcpu=marvell-pj4, has been added to generate code
7375       for the Marvell PJ4 processor.
7376     * The compiler can now automatically generate the VFMA, VFMS, REVSH
7377       and REV16 instructions.
7378     * A new vectorizer cost model for Advanced SIMD configurations to
7379       improve the auto-vectorization strategies used.
7380     * The scheduler now takes into account the number of live registers
7381       to reduce the amount of spilling that can occur. This should
7382       improve code performance in large functions. The limit can be
7383       removed by using the option -fno-sched-pressure.
7384     * Improvements have been made to the Marvell iWMMX code generation
7385       and support for the iWMMX2 SIMD unit has been added. The option
7386       -mcpu=iwmmxt2 can be used to enable code generation for the latter.
7387     * A number of code generation improvements for Thumb2 to reduce code
7388       size when compiling for the M-profile processors.
7389     * The RTEMS (arm-rtems) port has been updated to use the EABI.
7390     * Code generation support for the old FPA and Maverick floating-point
7391       architectures has been removed. Ports that previously relied on
7392       these features have also been removed. This includes the targets:
7393          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
7394          + arm*-*-elf (use arm*-*-eabi)
7395          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
7396          + arm*-*-ecos-elf (no alternative)
7397          + arm*-*-freebsd (no alternative)
7398          + arm*-wince-pe* (no alternative).
7399
7400  AVR
7401
7402     * Support for the "Embedded C" fixed-point has been added. For
7403       details, see the [28]GCC wiki and the [29]user manual. The support
7404       is not complete.
7405     * A new print modifier %r for register operands in inline assembler
7406       is supported. It will print the raw register number without the
7407       register prefix 'r':
7408    /* Return the most significant byte of 'val', a 64-bit value.  */
7409
7410    unsigned char msb (long long val)
7411    {
7412      unsigned char c;
7413      __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val));
7414      return c;
7415    }
7416       The inline assembler in this example will generate code like
7417    mov r24, 8+7
7418       provided c is allocated to R24 and val is allocated to R8...R15.
7419       This works because the GNU assembler accepts plain register numbers
7420       without register prefix.
7421     * Static initializers with 3-byte symbols are supported now:
7422    extern const __memx char foo;
7423    const __memx void *pfoo = &foo;
7424       This requires at least Binutils 2.23.
7425
7426  IA-32/x86-64
7427
7428     * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with
7429       SSE extensions disabled. Since the x86-64 ABI requires 16 byte
7430       stack alignment, this is ABI incompatible and intended to be used
7431       in controlled environments where stack space is an important
7432       limitation. This option will lead to wrong code when functions
7433       compiled with 16 byte stack alignment (such as functions from a
7434       standard library) are called with misaligned stack. In this case,
7435       SSE instructions may lead to misaligned memory access traps. In
7436       addition, variable arguments will be handled incorrectly for 16
7437       byte aligned objects (including x87 long double and __int128),
7438       leading to wrong results. You must build all modules with
7439       -mpreferred-stack-boundary=3, including any libraries. This
7440       includes the system libraries and startup modules.
7441     * Support for the new Intel processor codename Broadwell with RDSEED,
7442       ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw,
7443       -mrdseed command-line options.
7444     * Support for the Intel RTM and HLE intrinsics, built-in functions
7445       and code generation is available via -mrtm and -mhle.
7446     * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets.
7447       Intrinsics and built-in functions are available via -mfxsr, -mxsave
7448       and -mxsaveopt respectively.
7449     * New -maddress-mode=[short|long] options for x32.
7450       -maddress-mode=short overrides default 64-bit addresses to 32-bit
7451       by emitting the 0x67 address-size override prefix. This is the
7452       default address mode for x32.
7453     * New built-in functions to detect run-time CPU type and ISA:
7454          + A built-in function __builtin_cpu_is has been added to detect
7455            if the run-time CPU is of a particular type. It returns a
7456            positive integer on a match and zero otherwise. It accepts one
7457            string literal argument, the CPU name. For example,
7458            __builtin_cpu_is("westmere") returns a positive integer if the
7459            run-time CPU is an Intel Core i7 Westmere processor. Please
7460            refer to the [30]user manual for the list of valid CPU names
7461            recognized.
7462          + A built-in function __builtin_cpu_supports has been added to
7463            detect if the run-time CPU supports a particular ISA feature.
7464            It returns a positive integer on a match and zero otherwise.
7465            It accepts one string literal argument, the ISA feature. For
7466            example, __builtin_cpu_supports("ssse3") returns a positive
7467            integer if the run-time CPU supports SSSE3 instructions.
7468            Please refer to the [31]user manual for the list of valid ISA
7469            names recognized.
7470       Caveat: If these built-in functions are called before any static
7471       constructors are invoked, like during IFUNC initialization, then
7472       the CPU detection initialization must be explicitly run using this
7473       newly provided built-in function, __builtin_cpu_init. The
7474       initialization needs to be done only once. For example, this is how
7475       the invocation would look like inside an IFUNC initializer:
7476    static void (*some_ifunc_resolver(void))(void)
7477    {
7478      __builtin_cpu_init();
7479      if (__builtin_cpu_is("amdfam10h") ...
7480      if (__builtin_cpu_supports("popcnt") ...
7481    }
7482
7483     * Function Multiversioning Support with G++:
7484       It is now possible to create multiple function versions each
7485       targeting a specific processor and/or ISA. Function versions have
7486       the same signature but different target attributes. For example,
7487       here is a program with function versions:
7488    __attribute__ ((target ("default")))
7489    int foo(void)
7490    {
7491      return 1;
7492    }
7493
7494    __attribute__ ((target ("sse4.2")))
7495    int foo(void)
7496    {
7497      return 2;
7498    }
7499
7500    int main (void)
7501    {
7502      int (*p) = &foo;
7503      assert ((*p)() == foo());
7504      return 0;
7505    }
7506
7507       Please refer to this [32]wiki for more information.
7508     * The x86 back end has been improved to allow option -fschedule-insns
7509       to work reliably. This option can be used to schedule instructions
7510       better and leads to improved performace in certain cases.
7511     * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437
7512       from the Mingw-w64 trunk.
7513     * Support for new AMD family 15h processors (Steamroller core) is now
7514       available through the -march=bdver3 and -mtune=bdver3 options.
7515     * Support for new AMD family 16h processors (Jaguar core) is now
7516       available through the -march=btver2 and -mtune=btver2 options.
7517
7518  FRV
7519
7520     * This target now supports the -fstack-usage command-line option.
7521
7522  MIPS
7523
7524     * GCC can now generate code specifically for the R4700, Broadcom XLP
7525       and MIPS 34kn processors. The associated -march options are
7526       -march=r4700, -march=xlp and -march=34kn respectively.
7527     * GCC now generates better DSP code for MIPS 74k cores thanks to
7528       further scheduling optimizations.
7529     * The MIPS port now supports the -fstack-check option.
7530     * GCC now passes the -mmcu and -mno-mcu options to the assembler.
7531     * Previous versions of GCC would silently accept -fpic and -fPIC for
7532       -mno-abicalls targets like mips*-elf. This combination was not
7533       intended or supported, and did not generate position-independent
7534       code. GCC 4.8 now reports an error when this combination is used.
7535
7536  PowerPC / PowerPC64 / RS6000
7537
7538     * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save,
7539       restore or update the VRSAVE register by default. The respective
7540       operating systems manage the VRSAVE register directly.
7541     * Large TOC support has been added for AIX through the command line
7542       option -mcmodel=large.
7543     * Native Thread-Local Storage support has been added for AIX.
7544     * VMX (Altivec) and VSX instruction sets now are enabled implicitly
7545       when targetting processors that support those hardware features on
7546       AIX 6.1 and above.
7547
7548  RX
7549
7550     * This target will now issue a warning message whenever multiple fast
7551       interrupt handlers are found in the same compilation unit. This
7552       feature can be turned off by the new
7553       -mno-warn-multiple-fast-interrupts command-line option.
7554
7555  S/390, System z
7556
7557     * Support for the IBM zEnterprise zEC12 processor has been added.
7558       When using the -march=zEC12 option, the compiler will generate code
7559       making use of the following new instructions:
7560          + load and trap instructions
7561          + 2 new compare and trap instructions
7562          + rotate and insert selected bits - without CC clobber
7563       The -mtune=zEC12 option enables zEC12 specific instruction
7564       scheduling without making use of new instructions.
7565     * Register pressure sensitive instruction scheduling is enabled by
7566       default.
7567     * The ifunc function attribute is enabled by default.
7568     * memcpy and memcmp invokations on big memory chunks or with run time
7569       lengths are not generated inline anymore when tuning for z10 or
7570       higher. The purpose is to make use of the IFUNC optimized versions
7571       in Glibc.
7572
7573  SH
7574
7575     * The default alignment settings have been reduced to be less
7576       aggressive. This results in more compact code for optimization
7577       levels other than -Os.
7578     * Improved support for the __atomic built-in functions:
7579          + A new option -matomic-model=model selects the model for the
7580            generated atomic sequences. The following models are
7581            supported:
7582
7583              soft-gusa
7584                      Software gUSA sequences (SH3* and SH4* only). On
7585                      SH4A targets this will now also partially utilize
7586                      the movco.l and movli.l instructions. This is the
7587                      default when the target is sh3*-*-linux* or
7588                      sh4*-*-linux*.
7589
7590              hard-llcs
7591                      Hardware movco.l / movli.l sequences (SH4A only).
7592
7593              soft-tcb
7594                      Software thread control block sequences.
7595
7596              soft-imask
7597                      Software interrupt flipping sequences (privileged
7598                      mode only). This is the default when the target is
7599                      sh1*-*-linux* or sh2*-*-linux*.
7600
7601              none
7602                      Generates function calls to the respective __atomic
7603                      built-in functions. This is the default for SH64
7604                      targets or when the target is not sh*-*-linux*.
7605
7606          + The option -msoft-atomic has been deprecated. It is now an
7607            alias for -matomic-model=soft-gusa.
7608          + A new option -mtas makes the compiler generate the tas.b
7609            instruction for the __atomic_test_and_set built-in function
7610            regardless of the selected atomic model.
7611          + The __sync functions in libgcc now reflect the selected atomic
7612            model when building the toolchain.
7613     * Added support for the mov.b and mov.w instructions with
7614       displacement addressing.
7615     * Added support for the SH2A instructions movu.b and movu.w.
7616     * Various improvements to code generated for integer arithmetic.
7617     * Improvements to conditional branches and code that involves the T
7618       bit. A new option -mzdcbranch tells the compiler to favor
7619       zero-displacement branches. This is enabled by default for SH4*
7620       targets.
7621     * The pref instruction will now be emitted by the __builtin_prefetch
7622       built-in function for SH3* targets.
7623     * The fmac instruction will now be emitted by the fmaf standard
7624       function and the __builtin_fmaf built-in function.
7625     * The -mfused-madd option has been deprecated in favor of the
7626       machine-independent -ffp-contract option. Notice that the fmac
7627       instruction will now be generated by default for expressions like a
7628       * b + c. This is due to the compiler default setting
7629       -ffp-contract=fast.
7630     * Added new options -mfsrra and -mfsca to allow the compiler using
7631       the fsrra and fsca instructions on targets other than SH4A (where
7632       they are already enabled by default).
7633     * Added support for the __builtin_bswap32 built-in function. It is
7634       now expanded as a sequence of swap.b and swap.w instructions
7635       instead of a library function call.
7636     * The behavior of the -mieee option has been fixed and the negative
7637       form -mno-ieee has been added to control the IEEE conformance of
7638       floating point comparisons. By default -mieee is now enabled and
7639       the option -ffinite-math-only implicitly sets -mno-ieee.
7640     * Added support for the built-in functions __builtin_thread_pointer
7641       and __builtin_set_thread_pointer. This assumes that GBR is used to
7642       hold the thread pointer of the current thread. Memory loads and
7643       stores relative to the address returned by __builtin_thread_pointer
7644       will now also utilize GBR based displacement address modes.
7645     * The -mdiv= option for targets other than SHmedia has been fixed and
7646       documented.
7647
7648  SPARC
7649
7650     * Added optimized instruction scheduling for Niagara4.
7651
7652  TILE-Gx
7653
7654     * Added support for the -mcmodel=MODEL command-line option. The
7655       models supported are small and large.
7656
7657  V850
7658
7659     * This target now supports the E3V5 architecture via the use of the
7660       new -mv850e3v5 command-line option. It also has experimental
7661       support for the e3v5 LOOP instruction which can be enabled via the
7662       new -mloop command-line option.
7663
7664  XStormy16
7665
7666     * This target now supports the -fstack-usage command-line option.
7667
7668Operating Systems
7669
7670  OpenBSD
7671
7672     * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and
7673       support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated.
7674
7675  Windows (Cygwin)
7676
7677     * Executables are now linked against shared libgcc by default. The
7678       previous default was to link statically, which can still be done by
7679       explicitly specifying -static or static-libgcc on the command line.
7680       However it is strongly advised against, as it will cause problems
7681       for any application that makes use of DLLs compiled by GCC. It
7682       should be alright for a monolithic stand-alone application that
7683       only links against the Windows DLLs, but offers little or no
7684       benefit.
7685
7686GCC 4.8.1
7687
7688   This is the [33]list of problem reports (PRs) from GCC's bug tracking
7689   system that are known to be fixed in the 4.8.1 release. This list might
7690   not be complete (that is, it is possible that some PRs that have been
7691   fixed are not listed here).
7692
7693   The C++11 <chrono> std::chrono::system_clock and
7694   std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they
7695   both are now separate (never typedefs of each other), both use
7696   std::chrono::nanoseconds resolution, on most GNU/Linux configurations
7697   std::chrono::steady_clock is now finally monotonic, and both classes
7698   are mangled differently than in the previous GCC releases.
7699   std::chrono::system_clock::now() with std::chrono::microseconds resp.
7700   std::chrono::seconds resolution is still exported for backwards
7701   compatibility with default configured libstdc++. Note that libstdc++
7702   configured with --enable-libstdcxx-time= used to be ABI incompatible
7703   with default configured libstdc++ for those two classes and no ABI
7704   compatibility can be offered for those configurations, so any C++11
7705   code that uses those classes and has been compiled and linked against
7706   libstdc++ configured with the non-default --enable-libstdcxx-time=
7707   configuration option needs to be recompiled.
7708
7709GCC 4.8.2
7710
7711   This is the [34]list of problem reports (PRs) from GCC's bug tracking
7712   system that are known to be fixed in the 4.8.2 release. This list might
7713   not be complete (that is, it is possible that some PRs that have been
7714   fixed are not listed here).
7715
7716GCC 4.8.3
7717
7718   This is the [35]list of problem reports (PRs) from GCC's bug tracking
7719   system that are known to be fixed in the 4.8.3 release. This list might
7720   not be complete (that is, it is possible that some PRs that have been
7721   fixed are not listed here).
7722
7723   Support for the new powerpc64le-linux platform has been added. It
7724   defaults to generating code that conforms to the ELFV2 ABI.
7725
7726GCC 4.8.4
7727
7728   This is the [36]list of problem reports (PRs) from GCC's bug tracking
7729   system that are known to be fixed in the 4.8.4 release. This list might
7730   not be complete (that is, it is possible that some PRs that have been
7731   fixed are not listed here).
7732
7733GCC 4.8.5
7734
7735   This is the [37]list of problem reports (PRs) from GCC's bug tracking
7736   system that are known to be fixed in the 4.8.5 release. This list might
7737   not be complete (that is, it is possible that some PRs that have been
7738   fixed are not listed here).
7739
7740
7741    For questions related to the use of GCC, please consult these web
7742    pages and the [38]GCC manuals. If that fails, the
7743    [39]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7744    web pages and the development of GCC are welcome on our developer
7745    list at [40]gcc@gcc.gnu.org. All of [41]our lists have public
7746    archives.
7747
7748   Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and
7749   distribution of this entire article is permitted in any medium,
7750   provided this notice is preserved.
7751
7752   These pages are [43]maintained by the GCC team. Last modified
7753   2022-11-05.
7754
7755References
7756
7757   1. https://gcc.gnu.org/wiki/cxx-conversion
7758   2. ftp://gcc.gnu.org/pub/gcc/infrastructure/
7759   3. http://www.nongnu.org/avr-libc/
7760   4. https://gcc.gnu.org/PR54461
7761   5. https://gcc.gnu.org/gcc-4.8/porting_to.html
7762   6. https://github.com/google/sanitizers
7763   7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki
7764   8. https://lkml.org/lkml/2006/11/28/239
7765   9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
7766  10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
7767  11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
7768  12. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf
7769  13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
7770  14. https://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html
7771  15. http://gcc.gnu.org/projects/cxx1y.html
7772  16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011
7773  17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html
7774  18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
7775  19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
7776  20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
7777  21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
7778  22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
7779  23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html
7780  24. https://gcc.gnu.org/wiki/Fortran2003Status
7781  25. https://gcc.gnu.org/wiki/TS29113Status
7782  26. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libgfortran/libgfortran.h
7783  27. https://chasm-interop.sourceforge.net/
7784  28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support
7785  29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html
7786  30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html
7787  31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html
7788  32. https://gcc.gnu.org/wiki/FunctionMultiVersioning
7789  33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1
7790  34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2
7791  35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3
7792  36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4
7793  37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5
7794  38. https://gcc.gnu.org/onlinedocs/
7795  39. mailto:gcc-help@gcc.gnu.org
7796  40. mailto:gcc@gcc.gnu.org
7797  41. https://gcc.gnu.org/lists.html
7798  42. https://www.fsf.org/
7799  43. https://gcc.gnu.org/about.html
7800======================================================================
7801http://gcc.gnu.org/gcc-4.7/index.html
7802
7803                             GCC 4.7 Release Series
7804
7805   (This release series is no longer supported.)
7806
7807   June 12, 2014
7808
7809   The [1]GNU project and the GCC developers are pleased to announce the
7810   release of GCC 4.7.4.
7811
7812   This release is a bug-fix release, containing fixes for regressions in
7813   GCC 4.7.3 relative to previous releases of GCC.
7814
7815Release History
7816
7817   GCC 4.7.4
7818          June 12, 2014 ([2]changes, [3]documentation)
7819
7820   GCC 4.7.3
7821          April 11, 2013 ([4]changes, [5]documentation)
7822
7823   GCC 4.7.2
7824          September 20, 2012 ([6]changes, [7]documentation)
7825
7826   GCC 4.7.1
7827          June 14, 2012 ([8]changes, [9]documentation)
7828
7829   GCC 4.7.0
7830          March 22, 2012 ([10]changes, [11]documentation)
7831
7832References and Acknowledgements
7833
7834   GCC used to stand for the GNU C Compiler, but since the compiler
7835   supports several other languages aside from C, it now stands for the
7836   GNU Compiler Collection.
7837
7838   A list of [12]successful builds is updated as new information becomes
7839   available.
7840
7841   The GCC developers would like to thank the numerous people that have
7842   contributed new features, improvements, bug fixes, and other changes as
7843   well as test results to GCC. This [13]amazing group of volunteers is
7844   what makes GCC successful.
7845
7846   For additional information about GCC please refer to the [14]GCC
7847   project web site or contact the [15]GCC development mailing list.
7848
7849   To obtain GCC please use [16]our mirror sites or [17]our version
7850   control system.
7851
7852
7853    For questions related to the use of GCC, please consult these web
7854    pages and the [18]GCC manuals. If that fails, the
7855    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7856    web pages and the development of GCC are welcome on our developer
7857    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
7858    archives.
7859
7860   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
7861   distribution of this entire article is permitted in any medium,
7862   provided this notice is preserved.
7863
7864   These pages are [23]maintained by the GCC team. Last modified
7865   2022-10-26.
7866
7867References
7868
7869   1. http://www.gnu.org/
7870   2. http://gcc.gnu.org/gcc-4.7/changes.html
7871   3. https://gcc.gnu.org/onlinedocs/4.7.4/
7872   4. http://gcc.gnu.org/gcc-4.7/changes.html
7873   5. https://gcc.gnu.org/onlinedocs/4.7.3/
7874   6. http://gcc.gnu.org/gcc-4.7/changes.html
7875   7. https://gcc.gnu.org/onlinedocs/4.7.2/
7876   8. http://gcc.gnu.org/gcc-4.7/changes.html
7877   9. https://gcc.gnu.org/onlinedocs/4.7.1/
7878  10. http://gcc.gnu.org/gcc-4.7/changes.html
7879  11. https://gcc.gnu.org/onlinedocs/4.7.0/
7880  12. http://gcc.gnu.org/gcc-4.7/buildstat.html
7881  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7882  14. http://gcc.gnu.org/index.html
7883  15. mailto:gcc@gcc.gnu.org
7884  16. http://gcc.gnu.org/mirrors.html
7885  17. http://gcc.gnu.org/git.html
7886  18. https://gcc.gnu.org/onlinedocs/
7887  19. mailto:gcc-help@gcc.gnu.org
7888  20. mailto:gcc@gcc.gnu.org
7889  21. https://gcc.gnu.org/lists.html
7890  22. https://www.fsf.org/
7891  23. https://gcc.gnu.org/about.html
7892======================================================================
7893http://gcc.gnu.org/gcc-4.7/changes.html
7894
7895                             GCC 4.7 Release Series
7896                        Changes, New Features, and Fixes
7897
7898Caveats
7899
7900     * The -fconserve-space flag has been deprecated. The flag had no
7901       effect for most targets: only targets without a global .bss section
7902       and without support for switchable sections. Furthermore, the flag
7903       only had an effect for G++, where it could result in wrong
7904       semantics (please refer to the GCC manual for further details). The
7905       flag will be removed in GCC 4.8
7906     * Support for a number of older systems and recently unmaintained or
7907       untested target ports of GCC has been declared obsolete in GCC 4.7.
7908       Unless there is activity to revive them, the next release of GCC
7909       will have their sources permanently removed.
7910       All GCC ports for the following processor architectures have been
7911       declared obsolete:
7912          + picoChip (picochip-*)
7913       The following ports for individual systems on particular
7914       architectures have been obsoleted:
7915          + IRIX 6.5 (mips-sgi-irix6.5)
7916          + MIPS OpenBSD (mips*-*-openbsd*)
7917          + Solaris 8 (*-*-solaris2.8). Details can be found in the
7918            [1]announcement.
7919          + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*)
7920     * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A,
7921       ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by
7922       default, which for some sources generates code that accesses memory
7923       on unaligned addresses. This requires the kernel of those systems
7924       to enable such accesses (controlled by CP15 register c1, refer to
7925       ARM documentation). Alternatively, or for compatibility with
7926       kernels where unaligned accesses are not supported, all code has to
7927       be compiled with -mno-unaligned-access. Upstream Linux kernel
7928       releases have automatically and unconditionally supported unaligned
7929       accesses as emitted by GCC due to this option being active since
7930       version 2.6.28.
7931     * Support on ARM for the legacy floating-point accelerator (FPA) and
7932       the mixed-endian floating-point format that it used has been
7933       obsoleted. The ports that still use this format have been obsoleted
7934       as well. Many legacy ARM ports already provide an alternative that
7935       uses the VFP floating-point format. The obsolete ports will be
7936       deleted in the next release.
7937       The obsolete ports with alternatives are:
7938          + arm*-*-rtems (use arm*-*-rtemseabi)
7939          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
7940          + arm*-*-elf (use arm*-*-eabi)
7941          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
7942       Note, however, that these alternatives are not binary compatible
7943       with their legacy counterparts (although some can support running
7944       legacy applications).
7945       The obsolete ports that currently lack a modern alternative are:
7946          + arm*-*-ecos-elf
7947          + arm*-*-freebsd
7948          + arm*-wince-pe*
7949       New ports that support more recent versions of the architecture are
7950       welcome.
7951     * Support for the Maverick co-processor on ARM has been obsoleted.
7952       Code to support it will be deleted in the next release.
7953     * Support has been removed for Unix International threads on Solaris
7954       2, so the --enable-threads=solaris configure option and the
7955       -threads compiler option don't work any longer.
7956     * Support has been removed for the Solaris BSD Compatibility Package,
7957       which lives in /usr/ucbinclude and /usr/ucblib. It has been removed
7958       from Solaris 11, and was only intended as a migration aid from
7959       SunOS 4 to SunOS 5. The -compat-bsd compiler option is not
7960       recognized any longer.
7961     * The AVR port's libgcc has been improved and its multilib structure
7962       has been enhanced. As a result, all objects contributing to an
7963       application must either be compiled with GCC versions up to 4.6.x
7964       or with GCC versions 4.7.1 or later. If the compiler is used with
7965       AVR Libc, you need a version that supports the new layout, i.e.
7966       implements [2]#35407.
7967     * The AVR port's -mshort-calls command-line option has been
7968       deprecated. It will be removed in the GCC 4.8 release. See -mrelax
7969       for a replacement.
7970     * The AVR port only references startup code that clears .bss and the
7971       common section resp. initializes the .data and .rodata section
7972       provided respective sections (or subsections thereof) are not
7973       empty, see [3]PR18145. Applications that put all static storage
7974       objects into non-standard sections and / or define all static
7975       storage objects in assembler modules, must reference __do_clear_bss
7976       resp. __do_copy_data by hand or undefine the symbol(s) by means of
7977       -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data.
7978     * The ARM port's -mwords-little-endian option has been deprecated. It
7979       will be removed in a future release.
7980     * Support has been removed for the NetWare x86 configuration
7981       obsoleted in GCC 4.6.
7982     * It is no longer possible to use the "l" constraint in MIPS16 asm
7983       statements.
7984     * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard
7985       library which affected the ABI in C++11 mode: a data member was
7986       added to std::list changing its size and altering the definitions
7987       of some member functions, and std::pair's move constructor was
7988       non-trivial which altered the calling convention for functions with
7989       std::pair arguments or return types. The ABI incompatibilities have
7990       been fixed for GCC version 4.7.2 but as a result C++11 code
7991       compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11
7992       code compiled with different GCC versions and with C++98/C++03 code
7993       compiled with any version.
7994     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
7995       rules for the layout of vectors that could lead to wrong code being
7996       generated. Vectors larger than 8 bytes in size are now by default
7997       aligned to an 8-byte boundary. This is an ABI change: code that
7998       makes explicit use of vector types may be incompatible with binary
7999       objects built with older versions of GCC. Auto-vectorized code is
8000       not affected by this change. (This change affects GCC versions
8001       4.7.2 and later.)
8002     * More information on porting to GCC 4.7 from previous versions of
8003       GCC can be found in the [4]porting guide for this release.
8004
8005General Optimizer Improvements
8006
8007     * Support for a new parameter --param case-values-threshold=n was
8008       added to allow users to control the cutoff between doing switch
8009       statements as a series of if statements and using a jump table.
8010     * Link-time optimization (LTO) improvements:
8011          + Improved scalability and reduced memory usage. Link time
8012            optimization of Firefox now requires 3GB of RAM on a 64-bit
8013            system, while over 8GB was needed previously. Linking time has
8014            been improved, too. The serial stage of linking Firefox has
8015            been sped up by about a factor of 10.
8016          + Reduced size of object files and temporary storage used during
8017            linking.
8018          + Streaming performance (both outbound and inbound) has been
8019            improved.
8020          + ld -r is now supported with LTO.
8021          + Several bug fixes, especially in symbol table handling and
8022            merging.
8023     * Interprocedural optimization improvements:
8024          + Heuristics now take into account that after inlining code will
8025            be optimized out because of known values (or properties) of
8026            function parameters. For example:
8027void foo(int a)
8028{
8029  if (a > 10)
8030    ... huge code ...
8031}
8032void bar (void)
8033{
8034  foo (0);
8035}
8036
8037            The call of foo will be inlined into bar even when optimizing
8038            for code size. Constructs based on __builtin_constant_p are
8039            now understood by the inliner and code size estimates are
8040            evaluated a lot more realistically.
8041          + The representation of C++ virtual thunks and aliases (both
8042            implicit and defined via the alias attribute) has been
8043            re-engineered. Aliases no longer pose optimization barriers
8044            and calls to an alias can be inlined and otherwise optimized.
8045          + The inter-procedural constant propagation pass has been
8046            rewritten. It now performs generic function specialization.
8047            For example when compiling the following:
8048void foo(bool flag)
8049{
8050  if (flag)
8051    ... do something ...
8052  else
8053    ... do something else ...
8054}
8055void bar (void)
8056{
8057  foo (false);
8058  foo (true);
8059  foo (false);
8060  foo (true);
8061  foo (false);
8062  foo (true);
8063}
8064
8065            GCC will now produce two copies of foo. One with flag being
8066            true, while other with flag being false. This leads to
8067            performance improvements previously possible only by inlining
8068            all calls. Cloning causes a lot less code size growth.
8069     * A string length optimization pass has been added. It attempts to
8070       track string lengths and optimize various standard C string
8071       functions like strlen, strchr, strcpy, strcat, stpcpy and their
8072       _FORTIFY_SOURCE counterparts into faster alternatives. This pass is
8073       enabled by default at -O2 or above, unless optimizing for size, and
8074       can be disabled by the -fno-optimize-strlen option. The pass can
8075       e.g. optimize
8076char *bar (const char *a)
8077{
8078  size_t l = strlen (a) + 2;
8079  char *p = malloc (l); if (p == NULL) return p;
8080  strcpy (p, a); strcat (p, "/"); return p;
8081}
8082
8083       into:
8084char *bar (const char *a)
8085{
8086  size_t tmp = strlen (a);
8087  char *p = malloc (tmp + 2); if (p == NULL) return p;
8088  memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p;
8089}
8090
8091       or for hosted compilations where stpcpy is available in the runtime
8092       and headers provide its prototype, e.g.
8093void foo (char *a, const char *b, const char *c, const char *d)
8094{
8095  strcpy (a, b); strcat (a, c); strcat (a, d);
8096}
8097
8098       can be optimized into:
8099void foo (char *a, const char *b, const char *c, const char *d)
8100{
8101  strcpy (stpcpy (stpcpy (a, b), c), d);
8102}
8103
8104New Languages and Language specific improvements
8105
8106     * Version 3.1 of the OpenMP specification is now supported for the C,
8107       C++, and Fortran compilers.
8108
8109  Ada
8110
8111     * The command-line option -feliminate-unused-debug-types has been
8112       re-enabled by default, as it is for the other languages, leading to
8113       a reduction in debug info size of 12.5% and more for relevant
8114       cases, as well as to a small compilation speedup.
8115
8116  C family
8117
8118     * A new built-in, __builtin_assume_aligned, has been added, through
8119       which the compiler can be hinted about pointer alignment and can
8120       use it to improve generated code.
8121     * A new warning option -Wunused-local-typedefs was added for C, C++,
8122       Objective-C and Objective-C++. This warning diagnoses typedefs
8123       locally defined in a function, and otherwise not used.
8124     * A new experimental command-line option -ftrack-macro-expansion was
8125       added for C, C++, Objective-C, Objective-C++ and Fortran. It allows
8126       the compiler to emit diagnostic about the current macro expansion
8127       stack when a compilation error occurs in a macro expansion.
8128     * Experimental support for transactional memory has been added. It
8129       includes support in the compiler, as well as a supporting runtime
8130       library called libitm. To compile code with transactional memory
8131       constructs, use the -fgnu-tm option.
8132       Support is currently available for Alpha, ARM, PowerPC, SH, SPARC,
8133       and 32-bit/64-bit x86 platforms.
8134       For more details on transactional memory see [5]the GCC WiKi.
8135     * Support for atomic operations specifying the C++11/C11 memory model
8136       has been added. These new __atomic routines replace the existing
8137       __sync built-in routines.
8138       Atomic support is also available for memory blocks. Lock-free
8139       instructions will be used if a memory block is the same size and
8140       alignment as a supported integer type. Atomic operations which do
8141       not have lock-free support are left as function calls. A set of
8142       library functions is available on the GCC atomic wiki in the
8143       "External Atomics Library" section.
8144       For more details on the memory models and features, see the
8145       [6]atomic wiki.
8146     * When a binary operation is performed on vector types and one of the
8147       operands is a uniform vector, it is possible to replace the vector
8148       with the generating element. For example:
8149typedef int v4si __attribute__ ((vector_size (16)));
8150v4si res, a = {1,2,3,4};
8151int x;
8152
8153res = 2 + a;  /* means {2,2,2,2} + a  */
8154res = a - x;  /* means a - {x,x,x,x}  */
8155
8156  C
8157
8158     * There is support for some more features from the C11 revision of
8159       the ISO C standard. GCC now accepts the options -std=c11 and
8160       -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x.
8161          + Unicode strings (previously supported only with options such
8162            as -std=gnu11, now supported with -std=c11), and the
8163            predefined macros __STDC_UTF_16__ and __STDC_UTF_32__.
8164          + Nonreturning functions (_Noreturn and <stdnoreturn.h>).
8165          + Alignment support (_Alignas, _Alignof, max_align_t,
8166            <stdalign.h>).
8167          + A built-in function __builtin_complex is provided to support C
8168            library implementation of the CMPLX family of macros.
8169
8170  C++
8171
8172     * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat
8173       options, which are equivalent to -std=c++0x, -std=gnu++0x, and
8174       -Wc++0x-compat, respectively.
8175     * G++ now implements [7]C++11 extended friend syntax:
8176
8177template<class W>
8178class Q
8179{
8180  static const int I = 2;
8181public:
8182  friend W;
8183};
8184
8185struct B
8186{
8187  int ar[Q<B>::I];
8188};
8189
8190     * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit
8191       override control.
8192
8193struct B {
8194  virtual void f() const final;
8195  virtual void f(int);
8196};
8197
8198struct D : B {
8199  void f() const;            // error: D::f attempts to override final B::f
8200  void f(long) override;     // error: doesn't override anything
8201  void f(int) override;      // ok
8202};
8203
8204struct E final { };
8205struct F: E { }; // error: deriving from final class
8206
8207     * G++ now implements [9]C++11 non-static data member initializers.
8208
8209struct A {
8210  int i = 42;
8211} a; // initializes a.i to 42
8212
8213     * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11
8214       user-defined literals.
8215
8216// Not actually a good approximation.  :)
8217constexpr long double operator"" _degrees (long double d) { return d * 0.0175; }
8218long double pi = 180.0_degrees;
8219
8220     * G++ now implements [11]C++11 alias-declarations.
8221
8222template <class T> using Ptr = T*;
8223Ptr<int> ip;  // decltype(ip) is int*
8224
8225     * Thanks to Ville Voutilainen and Pedro Lamar�o, G++ now implements
8226       [12]C++11 delegating constructors.
8227
8228struct A {
8229  A(int);
8230  A(): A(42) { } // delegate to the A(int) constructor
8231};
8232
8233     * G++ now fully implements C++11 atomic classes rather than just
8234       integer derived classes.
8235
8236class POD {
8237  int a;
8238  int b;
8239};
8240std::atomic<POD> my_atomic_POD;
8241
8242     * G++ now sets the predefined macro __cplusplus to the correct value,
8243       199711L for C++98/03, and 201103L for C++11.
8244     * G++ now correctly implements the two-phase lookup rules such that
8245       an unqualified name used in a template must have an appropriate
8246       declaration found either in scope at the point of definition of the
8247       template or by argument-dependent lookup at the point of
8248       instantiation. As a result, code that relies on a second
8249       unqualified lookup at the point of instantiation to find functions
8250       declared after the template or in dependent bases will be rejected.
8251       The compiler will suggest ways to fix affected code, and using the
8252       -fpermissive compiler flag will allow the code to compile with a
8253       warning.
8254
8255template <class T>
8256void f() { g(T()); } // error, g(int) not found by argument-dependent lookup
8257void g(int) { } // fix by moving this declaration before the declaration of f
8258
8259template <class T>
8260struct A: T {
8261  // error, B::g(B) not found by argument-dependent lookup
8262  void f() { g(T()); } // fix by using this->g or A::g
8263};
8264
8265struct B { void g(B); };
8266
8267int main()
8268{
8269  f<int>();
8270  A<B>().f();
8271}
8272
8273     * G++ now properly re-uses stack space allocated for temporary
8274       objects when their lifetime ends, which can significantly lower
8275       stack consumption for some C++ functions. As a result of this, some
8276       code with undefined behavior will now break:
8277
8278const int &f(const int &i) { return i; }
8279....
8280const int &x = f(1);
8281const int &y = f(2);
8282
8283       Here, x refers to the temporary allocated to hold the 1 argument,
8284       which only lives until the end of the initialization; it
8285       immediately becomes a dangling reference. So the next statement
8286       re-uses the stack slot to hold the 2 argument, and users of x get
8287       that value instead.
8288       Note that this should not cause any change of behavior for
8289       temporaries of types with non-trivial destructors, as they are
8290       already destroyed at end of full-expression; the change is that now
8291       the storage is released as well.
8292     * A new command-line option -Wdelete-non-virtual-dtor has been added
8293       to warn when delete is used to destroy an instance of a class which
8294       has virtual functions and non-virtual destructor. It is unsafe to
8295       delete an instance of a derived class through a pointer to a base
8296       class if the base class does not have a virtual destructor. This
8297       warning is enabled by -Wall.
8298     * A new command-line option -Wzero-as-null-pointer-constant has been
8299       added to warn when a literal '0' is used as null pointer constant.
8300       It can be useful to facilitate the conversion to nullptr in C++11.
8301     * As per C++98, access-declarations are now deprecated by G++.
8302       Using-declarations are to be used instead. Furthermore, some
8303       efforts have been made to improve the support of class scope
8304       using-declarations. In particular, using-declarations referring to
8305       a dependent type now work as expected ([13]bug c++/14258).
8306     * The ELF symbol visibility of a template instantiation is now
8307       properly constrained by the visibility of its template arguments
8308       ([14]bug c++/35688).
8309
8310    Runtime Library (libstdc++)
8311
8312     * [15]Improved experimental support for the new ISO C++ standard,
8313       C++11, including:
8314          + using noexcept in most of the library;
8315          + implementations of pointer_traits, allocator_traits and
8316            scoped_allocator_adaptor;
8317          + uses-allocator construction for tuple;
8318          + vector meets the allocator-aware container requirements;
8319          + replacing monotonic_clock with steady_clock;
8320          + enabling the thread support library on most POSIX targets;
8321          + many small improvements to conform to the FDIS.
8322     * Added --enable-clocale=newlib configure option.
8323     * Debug Mode iterators for unordered associative containers.
8324     * Avoid polluting the global namespace and do not include <unistd.h>.
8325
8326  Fortran
8327
8328     * The compile flag [16]-fstack-arrays has been added, which causes
8329       all local arrays to be put on stack memory. For some programs this
8330       will improve the performance significantly. If your program uses
8331       very large local arrays, it is possible that you will have to
8332       extend your runtime limits for stack memory.
8333     * The [17]-Ofast flag now also implies [18]-fno-protect-parens and
8334       [19]-fstack-arrays.
8335     * Front-end optimizations can now be selected by the
8336       [20]-ffrontend-optimize option and deselected by the
8337       -fno-frontend-optimize option.
8338     * When front-end optimization removes a function call,
8339       [21]-Wfunction-elimination warns about that.
8340     * When performing front-end-optimization, the
8341       [22]-faggressive-function-elimination option allows the removal of
8342       duplicate function calls even for impure functions.
8343     * The flag [23]-Wreal-q-constant has been added, which warns if
8344       floating-point literals have been specified using q (such as
8345       1.0q0); the q marker is now supported as a vendor extension to
8346       denote quad precision (REAL(16) or, if not available, REAL(10)).
8347       Consider using a kind parameter (such as in 1.0_qp) instead, which
8348       can be obtained via [24]SELECTED_REAL_KIND.
8349     * The GFORTRAN_USE_STDERR environment variable has been removed. GNU
8350       Fortran now always prints error messages to standard error. If you
8351       wish to redirect standard error, please consult the manual for your
8352       OS, shell, batch environment etc. as appropriate.
8353     * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment
8354       variable have been removed. When encountering a serious error,
8355       gfortran will now always abort the program. Whether a core dump is
8356       generated depends on the user environment settings; see the ulimit
8357       -c setting for POSIX shells, limit coredumpsize for C shells, and
8358       the [25]WER user-mode dumps settings on Windows.
8359     * The [26]-fbacktrace option is now enabled by default. When
8360       encountering a fatal error, gfortran will attempt to print a
8361       backtrace to standard error before aborting. It can be disabled
8362       with -fno-backtrace. Note: On POSIX targets with the addr2line
8363       utility from GNU binutils, GNU Fortran can print a backtrace with
8364       function name, file name, line number information in addition to
8365       the addresses; otherwise only the addresses are printed.
8366     * [27]Fortran 2003:
8367          + Generic interface names which have the same name as derived
8368            types are now supported, which allows to write constructor
8369            functions. Note that Fortran does not support static
8370            constructor functions; only default initialization or an
8371            explicit structure-constructor initialization are available.
8372          + [28]Polymorphic (class) arrays are now supported.
8373     * [29]Fortran 2008:
8374          + Support for the DO CONCURRENT construct has been added, which
8375            allows the user to specify that individual loop iterations
8376            have no interdependencies.
8377          + [30]Coarrays: Full single-image support except for polymorphic
8378            coarrays. Additionally, preliminary support for multiple
8379            images via an MPI-based [31]coarray communication library has
8380            been added. Note: The library version is not yet usable as
8381            remote coarray access is not yet possible.
8382     * [32]TS 29113:
8383          + New flag [33]-std=f2008ts permits programs that are expected
8384            to conform to the Fortran 2008 standard and the draft
8385            Technical Specification (TS) 29113 on Further Interoperability
8386            of Fortran with C.
8387          + The OPTIONAL attribute is now allowed for dummy arguments of
8388            BIND(C) procedures.
8389          + The RANK intrinsic has been added.
8390          + The implementation of the ASYNCHRONOUS attribute in GCC is
8391            compatible with the candidate draft of TS 29113 (since GCC
8392            4.6).
8393
8394  Go
8395
8396     * GCC 4.7 implements the [34]Go 1 language standard. The library
8397       support in 4.7.0 is not quite complete, due to release timing.
8398       Release 4.7.1 includes complete support for Go 1. The Go library is
8399       from the Go 1.0.1 release.
8400     * Go has been tested on GNU/Linux and Solaris platforms. It may work
8401       on other platforms as well.
8402
8403New Targets and Target Specific Improvements
8404
8405  ARM
8406
8407     * GCC now supports the Cortex-A7 processor implementing the v7-a
8408       version of the architecture using the option -mcpu=cortex-a7.
8409     * The default vector size in auto-vectorization for NEON is now 128
8410       bits. If vectorization fails thusly, the vectorizer tries again
8411       with 64-bit vectors.
8412     * A new option -mvectorize-with-neon-double was added to allow users
8413       to change the vector size to 64 bits.
8414
8415  AVR
8416
8417     * GCC now supports the XMEGA architecture. This requires GNU binutils
8418       2.22 or later.
8419     * Support for the [35]named address spaces __flash, __flash1, ...,
8420       __flash5 and __memx has been added. These address spaces locate
8421       read-only data in flash memory and allow reading from flash memory
8422       by means of ordinary C code, i.e. without the need of (inline)
8423       assembler code:
8424
8425const __flash int values[] = { 42, 31 };
8426
8427int add_values (const __flash int *p, int i)
8428{
8429    return values[i] + *p;
8430}
8431
8432     * Support has been added for the AVR-specific configure option
8433       --with-avrlibc=yes in order to arrange for better integration of
8434       [36]AVR-Libc. This configure option is supported in avr-gcc 4.7.2
8435       and newer and will only take effect in non-RTEMS configurations. If
8436       avr-gcc is configured for RTEMS, the option will be ignored which
8437       is the same as specifying --with-avrlibc=no. See [37]PR54461 for
8438       more technical details.
8439     * Support for AVR-specific [38]built-in functions has been added.
8440     * Support has been added for the signed and unsigned 24-bit scalar
8441       integer types __int24 and __uint24.
8442     * New command-line options -maccumulate-args, -mbranch-cost=cost and
8443       -mstrict-X were added to allow better fine-tuning of code
8444       optimization.
8445     * The command-line option -fdata-sections now also takes affect on
8446       the section names of variables with the progmem attribute.
8447     * A new inline assembler print modifier %i to print a RAM address as
8448       I/O address has been added:
8449
8450#include <avr/io.h> /* Port Definitions from AVR-LibC */
8451
8452void set_portb (uint8_t value)
8453{
8454    asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory");
8455}
8456
8457       The offset between an I/O address and the RAM address for that I/O
8458       location is device-specific. This offset is taken into account when
8459       printing a RAM address with the %i modifier so that the address is
8460       suitable to be used as operand in an I/O command. The address must
8461       be a constant integer known at compile time.
8462     * The inline assembler constraint "R" to represent integers in the
8463       range -6 ... 5 has been removed without replacement.
8464     * Many optimizations to:
8465          + 64-bit integer arithmetic
8466          + Widening multiplication
8467          + Integer division by a constant
8468          + Avoid constant reloading in multi-byte instructions.
8469          + Micro-optimizations for special instruction sequences.
8470          + Generic built-in functions like __builtin_ffs*,
8471            __builtin_clz*, etc.
8472          + If-else decision trees generated by switch instructions
8473          + Merging of data located in flash memory
8474          + New libgcc variants for devices with 8-bit wide stack pointer
8475          + ...
8476     * Better documentation:
8477          + Handling of EIND and indirect jumps on devices with more than
8478            128 KiB of program memory.
8479          + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function
8480            registers.
8481          + Function attributes OS_main and OS_task.
8482          + AVR-specific built-in macros.
8483
8484  C6X
8485
8486     * Support has been added for the Texas Instruments C6X family of
8487       processors.
8488
8489  CR16
8490
8491     * Support has been added for National Semiconductor's CR16
8492       architecture.
8493
8494  Epiphany
8495
8496     * Support has been added for Adapteva's Epiphany architecture.
8497
8498  IA-32/x86-64
8499
8500     * Support for Intel AVX2 intrinsics, built-in functions and code
8501       generation is available via -mavx2.
8502     * Support for Intel BMI2 intrinsics, built-in functions and code
8503       generation is available via -mbmi2.
8504     * Implementation and automatic generation of __builtin_clz* using the
8505       lzcnt instruction is available via -mlzcnt.
8506     * Support for Intel FMA3 intrinsics and code generation is available
8507       via -mfma.
8508     * A new -mfsgsbase command-line option is available that makes GCC
8509       generate new segment register read/write instructions through
8510       dedicated built-ins.
8511     * Support for the new Intel rdrnd instruction is available via
8512       -mrdrnd.
8513     * Two additional AVX vector conversion instructions are available via
8514       -mf16c.
8515     * Support for new Intel processor codename IvyBridge with RDRND,
8516       FSGSBASE and F16C is available through -march=core-avx-i.
8517     * Support for the new Intel processor codename Haswell with AVX2,
8518       FMA, BMI, BMI2, LZCNT is available through -march=core-avx2.
8519     * Support for new AMD family 15h processors (Piledriver core) is now
8520       available through -march=bdver2 and -mtune=bdver2 options.
8521     * Support for [39]the x32 psABI is now available through the -mx32
8522       option.
8523     * Windows mingw targets are using the -mms-bitfields option by
8524       default.
8525     * Windows x86 targets are using the __thiscall calling convention for
8526       C++ class-member functions.
8527     * Support for the configure option --with-threads=posix for Windows
8528       mingw targets.
8529
8530  MIPS
8531
8532     * GCC now supports thread-local storage (TLS) for MIPS16. This
8533       requires GNU binutils 2.22 or later.
8534     * GCC can now generate code specifically for the Cavium Octeon+ and
8535       Octeon2 processors. The associated command-line options are
8536       -march=octeon+ and -march=octeon2 respectively. Both options
8537       require GNU binutils 2.22 or later.
8538     * GCC can now work around certain 24k errata, under the control of
8539       the command-line option -mfix-24k. These workarounds require GNU
8540       binutils 2.20 or later.
8541     * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build
8542       n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux
8543       toolchain that generates 32-bit code by default. Use the
8544       configure-time option --enable-targets=all to select these extra
8545       multilibs.
8546     * Passing -fno-delayed-branch now also stops the assembler from
8547       automatically filling delay slots.
8548
8549  PowerPC/PowerPC64
8550
8551     * Vectors of type vector long long or vector long are passed and
8552       returned using the same method as other vectors with the VSX
8553       instruction set. Previously GCC did not adhere to the ABI for
8554       128-bit vectors with 64-bit integer base types (PR 48857). This
8555       will also be fixed in the GCC 4.6.1 and 4.5.4 releases.
8556     * A new option -mno-pointers-to-nested-functions was added to allow
8557       AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify
8558       that the compiler should not load up the chain register (r11)
8559       before calling a function through a pointer. If you use this
8560       option, you cannot call nested functions through a pointer, or call
8561       other languages that might use the static chain.
8562     * A new option msave-toc-indirect was added to allow AIX
8563       32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we
8564       save the TOC in the prologue for indirect calls or generate the
8565       save inline. This can speed up some programs that call through a
8566       function pointer a lot, but it can slow down other functions that
8567       only call through a function pointer in exceptional cases.
8568     * The PowerPC port will now enable machine-specific built-in
8569       functions when the user switches the target machine using the
8570       #pragma GCC target or __attribute__ ((__target__ ("target"))) code
8571       sequences. In addition, the target macros are updated. However, due
8572       to the way the -save-temps switch is implemented, you won't see the
8573       effect of these additional macros being defined in preprocessor
8574       output.
8575
8576  SH
8577
8578     * A new option -msoft-atomic has been added. When it is specified,
8579       GCC will generate GNU/Linux-compatible gUSA atomic sequences for
8580       the new __atomic routines.
8581     * Since it is neither supported by GAS nor officially documented,
8582       code generation for little endian SH2A has been disabled.
8583       Specifying -ml with -m2a* will now result in a compiler error.
8584     * The defunct -mbranch-cost option has been fixed.
8585     * Some improvements to the generated code of:
8586          + Utilization of the tst #imm,R0 instruction.
8587          + Dynamic shift instructions on SH2A.
8588          + Integer absolute value calculations.
8589     * The -mdiv= option for targets other than SHmedia has been fixed and
8590       documented.
8591
8592  SPARC
8593
8594     * The option -mflat has been reinstated. When it is specified, the
8595       compiler will generate code for a single register window model.
8596       This is essentially a new implementation and the corresponding
8597       debugger support has been added to GDB 7.4.
8598     * Support for the options -mtune=native and -mcpu=native has been
8599       added on selected native platforms (GNU/Linux and Solaris).
8600     * Support for the SPARC T3 (Niagara 3) processor has been added.
8601     * VIS:
8602          + An intrinsics header visintrin.h has been added.
8603          + Builtin intrinsics for the VIS 1.0 edge handling and pixel
8604            compare instructions have been added.
8605          + The little-endian version of alignaddr is now supported.
8606          + When possible, VIS builtins are marked const, which should
8607            increase the compiler's ability to optimize VIS operations.
8608          + The compiler now properly tracks the %gsr register and how it
8609            behaves as an input for various VIS instructions.
8610          + Akin to fzero, the compiler can now generate fone instructions
8611            in order to set all of the bits of a floating-point register
8612            to 1.
8613          + The documentation for the VIS intrinsics in the GCC manual has
8614            been brought up to date and many inaccuracies were fixed.
8615          + Intrinsics for the VIS 2.0 bmask, bshuffle, and
8616            non-condition-code setting edge instructions have been added.
8617            Their availability is controlled by the new -mvis2 and
8618            -mno-vis2 options. They are enabled by default on
8619            UltraSPARC-III and later CPUs.
8620     * Support for UltraSPARC Fused Multiply-Add floating-point extensions
8621       has been added. These instructions are enabled by default on SPARC
8622       T3 (Niagara 3) and later CPUs.
8623
8624  TILE-Gx/TILEPro
8625
8626     * Support has been added for the Tilera TILE-Gx and TILEPro families
8627       of processors.
8628
8629Other significant improvements
8630
8631     * A new option (-grecord-gcc-switches) was added that appends
8632       compiler command-line options that might affect code generation to
8633       the DW_AT_producer attribute string in the DWARF debugging
8634       information.
8635     * GCC now supports various new GNU extensions to the DWARF debugging
8636       information format, like [40]entry value and [41]call site
8637       information, [42]typed DWARF stack or [43]a more compact macro
8638       representation. Support for these extensions has been added to GDB
8639       7.4. They can be disabled through the -gstrict-dwarf command-line
8640       option.
8641
8642GCC 4.7.1
8643
8644   This is the [44]list of problem reports (PRs) from GCC's bug tracking
8645   system that are known to be fixed in the 4.7.1 release. This list might
8646   not be complete (that is, it is possible that some PRs that have been
8647   fixed are not listed here).
8648
8649   The Go front end in the 4.7.1 release fully supports the [45]Go 1
8650   language standard.
8651
8652GCC 4.7.2
8653
8654   This is the [46]list of problem reports (PRs) from GCC's bug tracking
8655   system that are known to be fixed in the 4.7.2 release. This list might
8656   not be complete (that is, it is possible that some PRs that have been
8657   fixed are not listed here).
8658
8659GCC 4.7.3
8660
8661   This is the [47]list of problem reports (PRs) from GCC's bug tracking
8662   system that are known to be fixed in the 4.7.3 release. This list might
8663   not be complete (that is, it is possible that some PRs that have been
8664   fixed are not listed here).
8665
8666GCC 4.7.4
8667
8668   This is the [48]list of problem reports (PRs) from GCC's bug tracking
8669   system that are known to be fixed in the 4.7.4 release. This list might
8670   not be complete (that is, it is possible that some PRs that have been
8671   fixed are not listed here).
8672
8673
8674    For questions related to the use of GCC, please consult these web
8675    pages and the [49]GCC manuals. If that fails, the
8676    [50]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8677    web pages and the development of GCC are welcome on our developer
8678    list at [51]gcc@gcc.gnu.org. All of [52]our lists have public
8679    archives.
8680
8681   Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and
8682   distribution of this entire article is permitted in any medium,
8683   provided this notice is preserved.
8684
8685   These pages are [54]maintained by the GCC team. Last modified
8686   2023-03-29.
8687
8688References
8689
8690   1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html
8691   2. http://savannah.nongnu.org/bugs/?35407
8692   3. https://gcc.gnu.org/PR18145
8693   4. https://gcc.gnu.org/gcc-4.7/porting_to.html
8694   5. https://gcc.gnu.org/wiki/TransactionalMemory
8695   6. https://gcc.gnu.org/wiki/Atomic/GCCMM
8696   7. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8697   8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8698   9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8699  10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8700  11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8701  12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8702  13. https://gcc.gnu.org/PR14258
8703  14. https://gcc.gnu.org/PR35688
8704  15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011
8705  16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
8706  17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689
8707  18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270
8708  19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
8709  20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275
8710  21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170
8711  22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270
8712  23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149
8713  24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html
8714  25. https://learn.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps
8715  26. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183
8716  27. https://gcc.gnu.org/wiki/Fortran2003Status
8717  28. https://gcc.gnu.org/wiki/OOP
8718  29. https://gcc.gnu.org/wiki/Fortran2008Status
8719  30. https://gcc.gnu.org/wiki/Coarray
8720  31. https://gcc.gnu.org/wiki/CoarrayLib
8721  32. https://gcc.gnu.org/wiki/TS29113Status
8722  33. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53
8723  34. https://go.dev/doc/go1
8724  35. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html
8725  36. http://www.nongnu.org/avr-libc/
8726  37. https://gcc.gnu.org/PR54461
8727  38. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html
8728  39. https://sites.google.com/site/x32abi/
8729  40. https://dwarfstd.org/issues/100909.1.html
8730  41. https://dwarfstd.org/issues/100909.2.html
8731  42. https://dwarfstd.org/issues/140425.1.html
8732  43. https://dwarfstd.org/issues/110722.1.html
8733  44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1
8734  45. https://go.dev/doc/go1
8735  46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2
8736  47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3
8737  48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4
8738  49. https://gcc.gnu.org/onlinedocs/
8739  50. mailto:gcc-help@gcc.gnu.org
8740  51. mailto:gcc@gcc.gnu.org
8741  52. https://gcc.gnu.org/lists.html
8742  53. https://www.fsf.org/
8743  54. https://gcc.gnu.org/about.html
8744======================================================================
8745http://gcc.gnu.org/gcc-4.6/index.html
8746
8747                             GCC 4.6 Release Series
8748
8749   (This release series is no longer supported.)
8750
8751   April 12, 2013
8752
8753   The [1]GNU project and the GCC developers are pleased to announce the
8754   release of GCC 4.6.4.
8755
8756   This release is a bug-fix release, containing fixes for regressions in
8757   GCC 4.6.3 relative to previous releases of GCC.
8758
8759Release History
8760
8761   GCC 4.6.4
8762          April 12, 2013 ([2]changes, [3]documentation)
8763
8764   GCC 4.6.3
8765          March 1, 2012 ([4]changes, [5]documentation)
8766
8767   GCC 4.6.2
8768          October 26, 2011 ([6]changes, [7]documentation)
8769
8770   GCC 4.6.1
8771          June 27, 2011 ([8]changes, [9]documentation)
8772
8773   GCC 4.6.0
8774          March 25, 2011 ([10]changes, [11]documentation)
8775
8776References and Acknowledgements
8777
8778   GCC used to stand for the GNU C Compiler, but since the compiler
8779   supports several other languages aside from C, it now stands for the
8780   GNU Compiler Collection.
8781
8782   A list of [12]successful builds is updated as new information becomes
8783   available.
8784
8785   The GCC developers would like to thank the numerous people that have
8786   contributed new features, improvements, bug fixes, and other changes as
8787   well as test results to GCC. This [13]amazing group of volunteers is
8788   what makes GCC successful.
8789
8790   For additional information about GCC please refer to the [14]GCC
8791   project web site or contact the [15]GCC development mailing list.
8792
8793   To obtain GCC please use [16]our mirror sites or [17]our version
8794   control system.
8795
8796
8797    For questions related to the use of GCC, please consult these web
8798    pages and the [18]GCC manuals. If that fails, the
8799    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8800    web pages and the development of GCC are welcome on our developer
8801    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
8802    archives.
8803
8804   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
8805   distribution of this entire article is permitted in any medium,
8806   provided this notice is preserved.
8807
8808   These pages are [23]maintained by the GCC team. Last modified
8809   2022-10-26.
8810
8811References
8812
8813   1. http://www.gnu.org/
8814   2. http://gcc.gnu.org/gcc-4.6/changes.html
8815   3. https://gcc.gnu.org/onlinedocs/4.6.4/
8816   4. http://gcc.gnu.org/gcc-4.6/changes.html
8817   5. https://gcc.gnu.org/onlinedocs/4.6.3/
8818   6. http://gcc.gnu.org/gcc-4.6/changes.html
8819   7. https://gcc.gnu.org/onlinedocs/4.6.2/
8820   8. http://gcc.gnu.org/gcc-4.6/changes.html
8821   9. https://gcc.gnu.org/onlinedocs/4.6.1/
8822  10. http://gcc.gnu.org/gcc-4.6/changes.html
8823  11. https://gcc.gnu.org/onlinedocs/4.6.0/
8824  12. http://gcc.gnu.org/gcc-4.6/buildstat.html
8825  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
8826  14. http://gcc.gnu.org/index.html
8827  15. mailto:gcc@gcc.gnu.org
8828  16. http://gcc.gnu.org/mirrors.html
8829  17. http://gcc.gnu.org/git.html
8830  18. https://gcc.gnu.org/onlinedocs/
8831  19. mailto:gcc-help@gcc.gnu.org
8832  20. mailto:gcc@gcc.gnu.org
8833  21. https://gcc.gnu.org/lists.html
8834  22. https://www.fsf.org/
8835  23. https://gcc.gnu.org/about.html
8836======================================================================
8837http://gcc.gnu.org/gcc-4.6/changes.html
8838
8839                             GCC 4.6 Release Series
8840                        Changes, New Features, and Fixes
8841
8842Caveats
8843
8844     * The options -b <machine> and -V <version> have been removed because
8845       they were unreliable. Instead, users should directly run
8846       <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to
8847       run a different version of gcc.
8848     * GCC now has stricter checks for invalid command-line options. In
8849       particular, when gcc was called to link object files rather than
8850       compile source code, it would previously accept and ignore all
8851       options starting with --, including linker options such as
8852       --as-needed and --export-dynamic, although such options would
8853       result in errors if any source code was compiled. Such options, if
8854       unknown to the compiler, are now rejected in all cases; if the
8855       intent was to pass them to the linker, options such as
8856       -Wl,--as-needed should be used.
8857     * Versions of the GNU C library up to and including 2.11.1 included
8858       an [1]incorrect implementation of the cproj function. GCC optimizes
8859       its builtin cproj according to the behavior specified and allowed
8860       by the ISO C99 standard. If you want to avoid discrepancies between
8861       the C library and GCC's builtin transformations when using cproj in
8862       your code, use GLIBC 2.12 or later. If you are using an older GLIBC
8863       and actually rely on the incorrect behavior of cproj, then you can
8864       disable GCC's transformations using -fno-builtin-cproj.
8865     * The C-only intermodule optimization framework (IMA, enabled by
8866       -combine) has been removed in favor of the new generic link-time
8867       optimization framework (LTO) introduced in [2]GCC 4.5.0.
8868     * GCC now ships with the LGPL-licensed libquadmath library, which
8869       provides quad-precision mathematical functions for targets with a
8870       __float128 datatype. __float128 is available for targets on 32-bit
8871       x86, x86-64 and Itanium architectures. The libquadmath library is
8872       automatically built on such targets when building the Fortran
8873       compiler.
8874     * New -Wunused-but-set-variable and -Wunused-but-set-parameter
8875       warnings were added for C, C++, Objective-C and Objective-C++.
8876       These warnings diagnose variables respective parameters which are
8877       only set in the code and never otherwise used. Usually such
8878       variables are useless and often even the value assigned to them is
8879       computed needlessly, sometimes expensively. The
8880       -Wunused-but-set-variable warning is enabled by default by -Wall
8881       flag and -Wunused-but-set-parameter by -Wall -Wextra flags.
8882     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
8883       rules for the layout of vectors that could lead to wrong code being
8884       generated. Vectors larger than 8 bytes in size are now by default
8885       aligned to an 8-byte boundary. This is an ABI change: code that
8886       makes explicit use of vector types may be incompatible with binary
8887       objects built with older versions of GCC. Auto-vectorized code is
8888       not affected by this change. (This change affects GCC versions
8889       4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.)
8890     * On AVR, variables with the progmem attribute to locate data in
8891       flash memory must be qualified as const.
8892     * Support for a number of older systems and recently unmaintained or
8893       untested target ports of GCC has been declared obsolete in GCC 4.6.
8894       Unless there is activity to revive them, the next release of GCC
8895       will have their sources permanently removed.
8896       All GCC ports for the following processor architectures have been
8897       declared obsolete:
8898          + Argonaut ARC (arc-*)
8899          + National Semiconductor CRX (crx-*)
8900          + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*,
8901            m68hc12-*-*, m6812-*-*)
8902          + Sunplus S+core (score-*)
8903       The following ports for individual systems on particular
8904       architectures have been obsoleted:
8905          + Interix (i[34567]86-*-interix3*)
8906          + NetWare x86 (i[3456x]86-*-netware*)
8907          + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*)
8908          + MCore PE (mcore-*-pe*)
8909          + SH SymbianOS (sh*-*-symbianelf*)
8910          + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*)
8911          + M68K uClinux old ABI (m68k-*-uclinuxoldabi*)
8912          + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*,
8913            vax-*-netbsd*, but not *-*-netbsdelf*)
8914       The i[34567]86-*-pe alias for Cygwin targets has also been
8915       obsoleted; users should configure for i[34567]86-*-cygwin* instead.
8916       Certain configure options to control the set of libraries built
8917       with GCC on some targets have been obsoleted. On ARM targets, the
8918       options --disable-fpu, --disable-26bit, --disable-underscore,
8919       --disable-interwork, --disable-biendian and --disable-nofmult have
8920       been obsoleted. On MIPS targets, the options
8921       --disable-single-float, --disable-biendian and --disable-softfloat
8922       have been obsoleted.
8923     * Support has been removed for all the [3]configurations obsoleted in
8924       GCC 4.5.
8925     * More information on porting to GCC 4.6 from previous versions of
8926       GCC can be found in the [4]porting guide for this release.
8927
8928General Optimizer Improvements
8929
8930     * A new general optimization level, -Ofast, has been introduced. It
8931       combines the existing optimization level -O3 with options that can
8932       affect standards compliance but result in better optimized code.
8933       For example, -Ofast enables -ffast-math.
8934     * Link-time optimization improvements:
8935          + The [5]Scalable Whole Program Optimizer (WHOPR) project has
8936            stabilized to the point of being usable. It has become the
8937            default mode when using the LTO optimization model. Link time
8938            optimization can now split itself into multiple parallel
8939            compilations. Parallelism is controlled with -flto=n (where n
8940            specifies the number of compilations to execute in parallel).
8941            GCC can also cooperate with a GNU make job server by
8942            specifying the -flto=jobserver option and adding + to the
8943            beginning of the Makefile rule executing the linker.
8944            Classical LTO mode can be enforced by -flto-partition=none.
8945            This may result in small code quality improvements.
8946          + A large number of bugs were fixed. GCC itself, Mozilla Firefox
8947            and other large applications can be built with LTO enabled.
8948          + The linker plugin support improvements
8949               o Linker plugin is now enabled by default when the linker
8950                 is detected to have plugin support. This is the case for
8951                 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and
8952                 the Gold linker on ELF targets. Plugin support of the
8953                 Apple linker on Darwin is not compatible with GCC. The
8954                 linker plugin can also be controlled by the
8955                 -fuse-linker-plugin command-line option.
8956               o Resolution information from the linker plugin is used to
8957                 drive whole program assumptions. Use of the linker plugin
8958                 results in more aggressive optimization on binaries and
8959                 on shared libraries that use the hidden visibility
8960                 attribute. Consequently the use of -fwhole-program is not
8961                 necessary in addition to LTO.
8962          + Hidden symbols used from non-LTO objects now have to be
8963            explicitly annotated with externally_visible when the linker
8964            plugin is not used.
8965          + C++ inline functions and virtual tables are now privatized
8966            more aggressively, leading to better inter-procedural
8967            optimization and faster dynamic linking.
8968          + Memory usage and intermediate language streaming performance
8969            have been improved.
8970          + Static constructors and destructors from individual units are
8971            inlined into a single function. This can significantly improve
8972            startup times of large C++ applications where static
8973            constructors are very common. For example, static constructors
8974            are used when including the iostream header.
8975          + Support for the Ada language has been added.
8976     * Interprocedural optimization improvements
8977          + The interprocedural framework was re-tuned for link time
8978            optimization. Several scalability issues were resolved.
8979          + Improved auto-detection of const and pure functions. Newly,
8980            noreturn functions are auto-detected.
8981            The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is
8982            available that informs users when adding attributes to headers
8983            might improve code generation.
8984          + A number of inlining heuristic improvements. In particular:
8985               o Partial inlining is now supported and enabled by default
8986                 at -O2 and greater. The feature can be controlled via
8987                 -fpartial-inlining.
8988                 Partial inlining splits functions with short hot path to
8989                 return. This allows more aggressive inlining of the hot
8990                 path leading to better performance and often to code size
8991                 reductions (because cold parts of functions are not
8992                 duplicated).
8993               o Scalability for large compilation units was improved
8994                 significantly.
8995               o Inlining of callbacks is now more aggressive.
8996               o Virtual methods are considered for inlining when the
8997                 caller is inlined and devirtualization is then possible.
8998               o Inlining when optimizing for size (either in cold regions
8999                 of a program or when compiling with -Os) was improved to
9000                 better handle C++ programs with larger abstraction
9001                 penalty, leading to smaller and faster code.
9002          + The IPA reference optimization pass detecting global variables
9003            used or modified by functions was strengthened and sped up.
9004          + Functions whose address was taken are now optimized out when
9005            all references to them are dead.
9006          + A new inter-procedural static profile estimation pass detects
9007            functions that are executed once or unlikely to be executed.
9008            Unlikely executed functions are optimized for size. Functions
9009            executed once are optimized for size except for the inner
9010            loops.
9011          + On most targets with named section support, functions used
9012            only at startup (static constructors and main), functions used
9013            only at exit and functions detected to be cold are placed into
9014            separate text segment subsections. This extends the
9015            -freorder-functions feature and is controlled by the same
9016            switch. The goal is to improve the startup time of large C++
9017            programs.
9018            Proper function placement requires linker support. GNU ld
9019            2.21.51 on ELF targets was updated to place those functions
9020            together within the text section leading to better code
9021            locality and faster startup times of large C++ programs. The
9022            feature is also supported in the Apple linker. Support in the
9023            gold linker is planned.
9024     * A new switch -fstack-usage has been added. It makes the compiler
9025       output stack usage information for the program, on a per-function
9026       basis, in an auxiliary file.
9027     * A new switch -fcombine-stack-adjustments has been added. It can be
9028       used to enable or disable the compiler's stack-slot combining pass
9029       which before was enabled automatically at -O1 and above, but could
9030       not be controlled on its own.
9031     * A new switch -fstrict-volatile-bitfields has been added. Using it
9032       indicates that accesses to volatile bitfields should use a single
9033       access of the width of the field's type. This option can be useful
9034       for precisely defining and accessing memory-mapped peripheral
9035       registers from C or C++.
9036
9037Compile time and memory usage improvements
9038
9039     * Datastructures used by the dataflow framework in GCC were
9040       reorganized for better memory usage and more cache locality.
9041       Compile time is improved especially on units with large functions
9042       (possibly resulting from a lot of inlining) not fitting into the
9043       processor cache. The compile time of the GCC C compiler binary with
9044       link-time optimization went down by over 10% (benchmarked on x86-64
9045       target).
9046
9047New Languages and Language specific improvements
9048
9049  Ada
9050
9051     * Stack checking has been improved on selected architectures (Alpha,
9052       IA-32/x86-64, RS/6000 and SPARC): it now will detect stack
9053       overflows in all cases on these architectures.
9054     * Initial support for Ada 2012 has been added.
9055
9056  C family
9057
9058     * A new warning, enabled by -Wdouble-promotion, has been added that
9059       warns about cases where a value of type float is implicitly
9060       promoted to double. This is especially helpful for CPUs that handle
9061       the former in hardware, but emulate the latter in software.
9062     * A new function attribute leaf was introduced. This attribute allows
9063       better inter-procedural optimization across calls to functions that
9064       return to the current unit only via returning or exception
9065       handling. This is the case for most library functions that have no
9066       callbacks.
9067     * Support for a new data type __int128 for targets having wide enough
9068       machine-mode support.
9069     * The new function attribute callee_pop_aggregate allows to specify
9070       if the caller or callee is responsible for popping the aggregate
9071       return pointer value from the stack.
9072     * Support for selectively enabling and disabling warnings via #pragma
9073       GCC diagnostic has been added. For instance:
9074#pragma GCC diagnostic error "-Wuninitialized"
9075  foo(a);                       /* error is given for this one */
9076#pragma GCC diagnostic push
9077#pragma GCC diagnostic ignored "-Wuninitialized"
9078  foo(b);                       /* no diagnostic for this one */
9079#pragma GCC diagnostic pop
9080  foo(c);                       /* error is given for this one */
9081#pragma GCC diagnostic pop
9082  foo(d);                       /* depends on command-line options */
9083
9084     * The -fmax-errors=N option is now supported. Using this option
9085       causes the compiler to exit after N errors have been issued.
9086
9087  C
9088
9089     * There is now experimental support for some features from the
9090       upcoming C1X revision of the ISO C standard. This support may be
9091       selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions.
9092       Note that this support is experimental and may change incompatibly
9093       in future releases for consistency with changes to the C1X standard
9094       draft. The following features are newly supported as described in
9095       the N1539 draft of C1X (with changes agreed at the March 2011 WG14
9096       meeting); some other features were already supported with no
9097       compiler changes being needed, or have some support but not in full
9098       accord with N1539 (as amended).
9099          + Static assertions (_Static_assert keyword)
9100          + Typedef redefinition
9101          + New macros in <float.h>
9102          + Anonymous structures and unions
9103     * The new -fplan9-extensions option directs the compiler to support
9104       some extensions for anonymous struct fields which are implemented
9105       by the Plan 9 compiler. A pointer to a struct may be automatically
9106       converted to a pointer to an anonymous field when calling a
9107       function, in order to make the types match. An anonymous struct
9108       field whose type is a typedef name may be referred to using the
9109       typedef name.
9110
9111  C++
9112
9113     * Improved [7]experimental support for the upcoming C++0x ISO C++
9114       standard, including support for constexpr (thanks to Gabriel Dos
9115       Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide),
9116       noexcept, unrestricted unions, range-based for loops (thanks to
9117       Rodrigo Rivas Costa), opaque enum declarations (thanks also to
9118       Rodrigo), implicitly deleted functions and implicit move
9119       constructors.
9120     * When an extern declaration within a function does not match a
9121       declaration in the enclosing context, G++ now properly declares the
9122       name within the namespace of the function rather than the namespace
9123       which was open just before the function definition ([8]c++/43145).
9124     * GCC now warns by default when casting integers to larger pointer
9125       types. These warnings can be disabled with the option
9126       -Wno-int-to-pointer-cast, which is now also available in C++.
9127     * G++ no longer optimizes using the assumption that a value of
9128       enumeration type will fall within the range specified by the
9129       standard, since that assumption is easily violated with a
9130       conversion from integer type ([9]c++/43680). The old behavior can
9131       be restored with -fstrict-enums.
9132     * The new -fnothrow-opt flag changes the semantics of a throw()
9133       exception specification to match the proposed semantics of the
9134       noexcept specification: just call terminate if an exception tries
9135       to propagate out of a function with such an exception
9136       specification. This dramatically reduces or eliminates the code
9137       size overhead from adding the exception specification.
9138     * The new -Wnoexcept flag will suggest adding a noexcept qualifier to
9139       a function that the compiler can tell doesn't throw if it would
9140       change the value of a noexcept expression.
9141     * The -Wshadow option now warns if a local variable or type
9142       declaration shadows another type in C++. Note that the compiler
9143       will not warn if a local variable shadows a struct/class/enum, but
9144       will warn if it shadows an explicit typedef.
9145     * When an identifier is not found in the current scope, G++ now
9146       offers suggestions about which identifier might have been intended.
9147     * G++ now issues clearer diagnostics for missing semicolons after
9148       class, struct, and union definitions.
9149     * G++ now issues clearer diagnostics for missing semicolons after
9150       class member declarations.
9151     * G++ now issues clearer diagnostics when a colon is used in a place
9152       where a double-colon was intended.
9153     * G++ no longer accepts mutable on reference members ([10]c++/33558).
9154       Use -fpermissive to allow the old, non-conforming behaviour.
9155     * A few mangling fixes have been made, to attribute const/volatile on
9156       function pointer types, decltype of a plain decl, and use of a
9157       function parameter in the declaration of another parameter. By
9158       default the compiler still uses the old mangling, but emits aliases
9159       with the new mangling on targets that support strong aliases. Users
9160       can switch over entirely to the new mangling with -fabi-version=5
9161       or -fabi-version=0. -Wabi will now warn about code that uses the
9162       old mangling.
9163     * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified
9164       type to be default initialized unless the type has a user-declared
9165       default constructor. In 4.6.2 G++ implements the proposed
9166       resolution of [11]DR 253, so default initialization is allowed if
9167       it initializes all subobjects. Code that fails to compile can be
9168       fixed by providing an initializer e.g.
9169    struct A { A(); };
9170    struct B : A { int i; };
9171    const B b = B();
9172       Use -fpermissive to allow the old, non-conforming behaviour.
9173
9174    Runtime Library (libstdc++)
9175
9176     * [12]Improved experimental support for the upcoming ISO C++
9177       standard, C++0x, including using constexpr and nullptr.
9178     * Performance improvements to the [13]Debug Mode, thanks to Fran�ois
9179       Dumont.
9180     * Atomic operations used for reference-counting are annotated so that
9181       they can be understood by race detectors such as Helgrind, see
9182       [14]Data Race Hunting.
9183     * Most libstdc++ standard headers have been changed to no longer
9184       include the cstddef header as an implementation detail. Code that
9185       relied on that header being included as side-effect of including
9186       other standard headers will need to include cstddef explicitly.
9187
9188  Fortran
9189
9190     * On systems supporting the libquadmath library, GNU Fortran now also
9191       supports a quad-precision, kind=16 floating-point data type
9192       (REAL(16), COMPLEX(16)). As the data type is not fully supported in
9193       hardware, calculations might be one to two orders of magnitude
9194       slower than with the 4, 8 or 10 bytes floating-point data types.
9195       This change does not affect systems which support REAL(16) in
9196       hardware nor those which do not support libquadmath.
9197     * Much improved compile time for large array constructors.
9198     * In order to reduce execution time and memory consumption, use of
9199       temporary arrays in assignment expressions is avoided for many
9200       cases. The compiler now reverses loops in order to avoid generating
9201       a temporary array where possible.
9202     * Improved diagnostics, especially with -fwhole-file.
9203     * The -fwhole-file flag is now enabled by default. This improves code
9204       generation and diagnostics. It can be disabled using the deprecated
9205       -fno-whole-file flag.
9206     * Support the generation of Makefile dependencies via the [15]-M...
9207       flags of GCC; you may need to specify the -cpp option in addition.
9208       The dependencies take modules, Fortran's include, and CPP's
9209       #include into account. Note: Using -M for the module path is no
9210       longer supported, use -J instead.
9211     * The flag -Wconversion has been modified to only issue warnings
9212       where a conversion leads to information loss. This drastically
9213       reduces the number of warnings; -Wconversion is thus now enabled
9214       with -Wall. The flag -Wconversion-extra has been added and also
9215       warns about other conversions; -Wconversion-extra typically issues
9216       a huge number of warnings, most of which can be ignored.
9217     * A new command-line option -Wunused-dummy-argument warns about
9218       unused dummy arguments and is included in -Wall. Before,
9219       -Wunused-variable also warned about unused dummy arguments.
9220     * Fortran 2003 support has been extended:
9221          + Improved support for polymorphism between libraries and
9222            programs and for complicated inheritance patterns (cf.
9223            [16]object-oriented programming).
9224          + Experimental support of the ASSOCIATE construct.
9225          + In pointer assignments it is now possible to specify the lower
9226            bounds of the pointer and, for a rank-1 or a simply contiguous
9227            data-target, to remap the bounds.
9228          + Automatic (re)allocation: In intrinsic assignments to
9229            allocatable variables the left-hand side will be automatically
9230            allocated (if unallocated) or reallocated (if the shape or
9231            type parameter is different). To avoid the small performance
9232            penalty, you can use a(:) = ... instead of a = ... for arrays
9233            and character strings - or disable the feature using -std=f95
9234            or -fno-realloc-lhs.
9235          + Deferred type parameter: For scalar allocatable and pointer
9236            variables the character length can be deferred.
9237          + Namelist variables with allocatable and pointer attribute and
9238            nonconstant length type parameter are supported.
9239     * Fortran 2008 support has been extended:
9240          + Experimental [17]coarray support (for one image only, i.e.
9241            num_images() == 1); use the [18]-fcoarray=single flag to
9242            enable it.
9243          + The STOP and the new ERROR STOP statements now support all
9244            constant expressions.
9245          + Support for the CONTIGUOUS attribute.
9246          + Support for ALLOCATE with MOLD.
9247          + Support for the STORAGE_SIZE intrinsic inquiry function.
9248          + Support of the NORM2 and PARITY intrinsic functions.
9249          + The following bit intrinsics were added: POPCNT and POPPAR for
9250            counting the number of 1 bits and returning the parity; BGE,
9251            BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR
9252            for combined left and right shifts, MASKL and MASKR for simple
9253            left and right justified masks, MERGE_BITS for a bitwise merge
9254            using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations,
9255            and the transformational bit intrinsics IALL, IANY and
9256            IPARITY.
9257          + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine.
9258          + Support for the IMPURE attribute for procedures, which allows
9259            for ELEMENTAL procedures without the restrictions of PURE.
9260          + Null pointers (including NULL()) and not allocated variables
9261            can be used as actual argument to optional non-pointer,
9262            non-allocatable dummy arguments, denoting an absent argument.
9263          + Non-pointer variables with TARGET attribute can be used as
9264            actual argument to POINTER dummies with INTENT(IN)
9265          + Pointers including procedure pointers and those in a derived
9266            type (pointer components) can now be initialized by a target
9267            instead of only by NULL.
9268          + The EXIT statement (with construct-name) can now be used to
9269            leave not only the DO but also the ASSOCIATE, BLOCK, IF,
9270            SELECT CASE and SELECT TYPE constructs.
9271          + Internal procedures can now be used as actual argument.
9272          + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS
9273            and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV
9274            have been added; these arrays contain the supported kind
9275            values for the respective types.
9276          + The module procedures C_SIZEOF of the intrinsic module
9277            ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of
9278            ISO_FORTRAN_ENV have been implemented.
9279          + Minor changes: obsolescence diagnostics for ENTRY was added
9280            for -std=f2008; a line may start with a semicolon; for
9281            internal and module procedures END can be used instead of END
9282            SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes
9283            a RADIX argument; intrinsic types are supported for
9284            TYPE(intrinsic-type-spec); multiple type-bound procedures can
9285            be declared in a single PROCEDURE statement; implied-shape
9286            arrays are supported for named constants (PARAMETER). The
9287            transformational, three argument versions of BESSEL_JN and
9288            BESSEL_YN were added - the elemental, two-argument version had
9289            been added in GCC 4.4; note that the transformational
9290            functions use a recurrence algorithm.
9291
9292  Go
9293
9294   Support for the Go programming language has been added to GCC. It is
9295   not enabled by default when you build GCC; use the --enable-languages
9296   configure option to build it. The driver program for compiling Go code
9297   is gccgo.
9298
9299   Go is currently known to work on GNU/Linux and RTEMS. Solaris support
9300   is in progress. It may or may not work on other platforms.
9301
9302  Objective-C and Objective-C++
9303
9304     * The -fobjc-exceptions flag is now required to enable Objective-C
9305       exception and synchronization syntax (introduced by the keywords
9306       @try, @catch, @finally and @synchronized).
9307     * A number of Objective-C 2.0 features and extensions are now
9308       supported by GCC. These features are enabled by default; you can
9309       disable them by using the new -fobjc-std=objc1 command-line option.
9310     * The Objective-C 2.0 dot-syntax is now supported. It is an
9311       alternative syntax for using getters and setters; object.count is
9312       automatically converted into [object count] or [object setCount:
9313       ...] depending on context; for example if (object.count > 0) is
9314       automatically compiled into the equivalent of if ([object count] >
9315       0) while object.count = 0; is automatically compiled into the
9316       equivalent ot [object setCount: 0];. The dot-syntax can be used
9317       with instance and class objects and with any setters or getters, no
9318       matter if they are part of a declared property or not.
9319     * Objective-C 2.0 declared properties are now supported. They are
9320       declared using the new @property keyword, and are most commonly
9321       used in conjunction with the new Objective-C 2.0 dot-syntax. The
9322       nonatomic, readonly, readwrite, assign, retain, copy, setter and
9323       getter attributes are all supported. Marking declared properties
9324       with __attribute__ ((deprecated)) is supported too.
9325     * The Objective-C 2.0 @synthesize and @dynamic keywords are
9326       supported. @synthesize causes the compiler to automatically
9327       synthesize a declared property, while @dynamic is used to disable
9328       all warnings for a declared property for which no implementation is
9329       provided at compile time. Synthesizing declared properties requires
9330       runtime support in most useful cases; to be able to use it with the
9331       GNU runtime, appropriate helper functions have been added to the
9332       GNU Objective-C runtime ABI, and are implemented by the GNU
9333       Objective-C runtime library shipped with GCC.
9334     * The Objective-C 2.0 fast enumeration syntax is supported in
9335       Objective-C. This is currently not yet available in Objective-C++.
9336       Fast enumeration requires support in the runtime, and such support
9337       has been added to the GNU Objective-C runtime library (shipped with
9338       GCC).
9339     * The Objective-C 2.0 @optional keyword is supported. It allows you
9340       to mark methods or properties in a protocol as optional as opposed
9341       to required.
9342     * The Objective-C 2.0 @package keyword is supported. It has currently
9343       the same effect as the @public keyword.
9344     * Objective-C 2.0 method attributes are supported. Currently the
9345       supported attributes are deprecated, sentinel, noreturn and format.
9346     * Objective-C 2.0 method argument attributes are supported. The most
9347       widely used attribute is unused, to mark an argument as unused in
9348       the implementation.
9349     * Objective-C 2.0 class and protocol attributes are supported.
9350       Currently the only supported attribute is deprecated.
9351     * Objective-C 2.0 class extensions are supported. A class extension
9352       has the same syntax as a category declaration with no category
9353       name, and the methods and properties declared in it are added
9354       directly to the main class. It is mostly used as an alternative to
9355       a category to add methods to a class without advertising them in
9356       the public headers, with the advantage that for class extensions
9357       the compiler checks that all the privately declared methods are
9358       actually implemented.
9359     * As a result of these enhancements, GCC can now be used to build
9360       Objective-C and Objective-C++ software that uses Foundation and
9361       other important system frameworks with the NeXT runtime on Darwin 9
9362       and Darwin 10 (OSX 10.5 and 10.6).
9363     * Many bugs in the compiler have been fixed in this release; in
9364       particular, LTO can now be used when compiling Objective-C and
9365       Objective-C++ and the parser is much more robust in dealing with
9366       invalid code.
9367
9368    Runtime Library (libobjc)
9369
9370     * The GNU Objective-C runtime library now defines the macro
9371       __GNU_LIBOBJC__ (with a value that is increased at every release
9372       where there is any change to the API) in objc/objc.h, making it
9373       easy to determine if the GNU Objective-C runtime library is being
9374       used, and if so, which version. Previous versions of the GNU
9375       Objective-C runtime library (and other Objective-C runtime
9376       libraries such as the Apple one) do not define this macro.
9377     * A new Objective-C 2.0 API, almost identical to the one implemented
9378       by the Apple Objective-C runtime, has been implemented in the GNU
9379       Objective-C runtime library. The new API hides the internals of
9380       most runtime structures but provides a more extensive set of
9381       functions to operate on them. It is much easier, for example, to
9382       create or modify classes at runtime. The new API also makes it
9383       easier to port software from Apple to GNU as almost no changes
9384       should be required. The old API is still supported for backwards
9385       compatibility; including the old objc/objc-api.h header file
9386       automatically selects the old API, while including the new
9387       objc/runtime.h header file automatically selects the new API.
9388       Support for the old API is being phased out and upgrading the
9389       software to use the new API is strongly recommended. To check for
9390       the availability of the new API, the __GNU_LIBOBJC__ macro can be
9391       used as older versions of the GNU Objective-C runtime library,
9392       which do not support the new API, do not define such a macro.
9393     * Runtime support for @synchronized has been added.
9394     * Runtime support for Objective-C 2.0 synthesized property accessors
9395       has been added.
9396     * Runtime support for Objective-C 2.0 fast enumeration has been
9397       added.
9398
9399New Targets and Target Specific Improvements
9400
9401  ARM
9402
9403     * GCC now supports the Cortex-M4 processor implementing the v7-em
9404       version of the architecture using the option -mcpu=cortex-m4.
9405     * Scheduling descriptions for the Cortex-M4, the Neon and the
9406       floating point units of the Cortex-A9 and a pipeline description
9407       for the Cortex-A5 have been added.
9408     * Synchronization primitives such as __sync_fetch_and_add and friends
9409       are now inlined for supported architectures rather than calling
9410       into a kernel helper function.
9411     * SSA loop prefetching is enabled by default for the Cortex-A9 at
9412       -O3.
9413     * Several improvements were committed to improve code generation for
9414       the ARM architecture including a rewritten implementation for load
9415       and store multiples.
9416     * Several enhancements were committed to improve SIMD code generation
9417       for NEON by adding support for widening instructions, misaligned
9418       loads and stores, vector conditionals and support for 64 bit
9419       arithmetic.
9420     * Support was added for the Faraday cores fa526, fa606te, fa626te,
9421       fmp626te, fmp626 and fa726te and can be used with the respective
9422       names as parameters to the -mcpu= option.
9423     * Basic support was added for Cortex-A15 and is available through
9424       -mcpu=cortex-a15.
9425     * GCC for AAPCS configurations now more closely adheres to the AAPCS
9426       specification by enabling -fstrict-volatile-bitfields by default.
9427
9428  IA-32/x86-64
9429
9430     * The new -fsplit-stack option permits programs to use a
9431       discontiguous stack. This is useful for threaded programs, in that
9432       it is no longer necessary to specify the maximum stack size when
9433       creating a thread. This feature is currently only implemented for
9434       32-bit and 64-bit x86 GNU/Linux targets.
9435     * Support for emitting profiler counter calls before function
9436       prologues. This is enabled via a new command-line option -mfentry.
9437     * Optimization for the Intel Core 2 processors is now available
9438       through the -march=core2 and -mtune=core2 options.
9439     * Support for Intel Core i3/i5/i7 processors is now available through
9440       the -march=corei7 and -mtune=corei7 options.
9441     * Support for Intel Core i3/i5/i7 processors with AVX is now
9442       available through the -march=corei7-avx and -mtune=corei7-avx
9443       options.
9444     * Support for AMD Bobcat (family 14) processors is now available
9445       through the -march=btver1 and -mtune=btver1 options.
9446     * Support for AMD Bulldozer (family 15) processors is now available
9447       through the -march=bdver1 and -mtune=bdver1 options.
9448     * The default setting (when not optimizing for size) for 32-bit
9449       GNU/Linux and Darwin x86 targets has been changed to
9450       -fomit-frame-pointer. The default can be reverted to
9451       -fno-omit-frame-pointer by configuring GCC with the
9452       --enable-frame-pointer configure option.
9453     * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support
9454       __float128 on 32-bit and 64-bit x86 targets.
9455     * AVX floating-point arithmetic can now be enabled by default at
9456       configure time with the new --with-fpmath=avx option.
9457     * The SSA loop prefetching pass is enabled when using -O3 when
9458       optimizing for CPUs where prefetching is beneficial (AMD CPUs newer
9459       than K6).
9460     * Support for TBM (Trailing Bit Manipulation) built-in functions and
9461       code generation is available via -mtbm.
9462     * Support for AMD's BMI (Bit Manipulation) built-in functions and
9463       code generation is available via -mbmi.
9464
9465  MicroBlaze
9466
9467     * Support has been added for the Xilinx MicroBlaze softcore processor
9468       (microblaze-elf) embedded target. This configurable processor is
9469       supported on several Xilinx Spartan and Virtex FPGAs.
9470
9471  MIPS
9472
9473     * GCC now supports the Loongson 3A processor. Its canonical -march=
9474       and -mtune= name is loongson3a.
9475
9476  MN10300 / AM33
9477
9478     * The inline assembly register constraint "A" has been renamed "c".
9479       This constraint is used to select a floating-point register that
9480       can be used as the destination of a multiply-accumulate
9481       instruction.
9482     * New inline assembly register constraints "A" and "D" have been
9483       added. These constraint letters resolve to all general registers
9484       when compiling for AM33, and resolve to address registers only or
9485       data registers only when compiling for MN10300.
9486     * The MDR register is represented in the compiler. One can access the
9487       register via the "z" constraint in inline assembly. It can be
9488       marked as clobbered or used as a local register variable via the
9489       "mdr" name. The compiler uses the RETF instruction if the function
9490       does not modify the MDR register, so it is important that inline
9491       assembly properly annotate any usage of the register.
9492
9493  PowerPC/PowerPC64
9494
9495     * GCC now supports the Applied Micro Titan processor with
9496       -mcpu=titan.
9497     * The -mrecip option has been added, which indicates whether the
9498       reciprocal and reciprocal square root instructions should be used.
9499     * The -mveclibabi=mass option can be used to enable the compiler to
9500       autovectorize mathematical functions using the Mathematical
9501       Acceleration Subsystem library.
9502     * The -msingle-pic-base option has been added, which instructs the
9503       compiler to avoid loading the PIC base register in function
9504       prologues. The PIC base register must be initialized by the runtime
9505       system.
9506     * The -mblock-move-inline-limit option has been added, which enables
9507       the user to control the maximum size of inlined memcpy calls and
9508       similar.
9509     * PowerPC64 GNU/Linux support for applications requiring a large TOC
9510       section has been improved. A new command-line option,
9511       -mcmodel=MODEL, controls this feature; valid values for MODEL are
9512       small, medium, or large.
9513     * The Altivec builtin functions vec_ld and vec_st have been modified
9514       to generate the Altivec memory instructions LVX and STVX, even if
9515       the -mvsx option is used. In the initial GCC 4.5 release, these
9516       builtin functions were changed to generate VSX memory reference
9517       instructions instead of Altivec memory instructions, but there are
9518       differences between the two instructions. If the VSX instruction
9519       set is available, you can now use the new builtin functions
9520       vec_vsx_ld and vec_vsx_st which always generates the VSX memory
9521       instructions.
9522     * The GCC compiler on AIX now defaults to a process layout with a
9523       larger data space allowing larger programs to be compiled.
9524     * The GCC long double type on AIX 6.1 and above has reverted to 64
9525       bit double precision, matching the AIX XL compiler default, because
9526       of missing C99 symbols required by the GCC runtime.
9527     * The default processor scheduling model and tuning for PowerPC64
9528       GNU/Linux and for AIX 6.1 and above now is POWER7.
9529     * Starting with GCC 4.6.1, vectors of type vector long long or vector
9530       long are passed and returned in the same method as other vectors
9531       with the VSX instruction set. Previously the GCC compiler did not
9532       adhere to the ABI for 128-bit vectors with 64-bit integer base
9533       types (PR 48857). This is also fixed in the GCC 4.5.4 release.
9534
9535  S/390, zSeries and System z9/z10, IBM zEnterprise z196
9536
9537     * Support for the zEnterprise z196 processor has been added. When
9538       using the -march=z196 option, the compiler will generate code
9539       making use of the following instruction facilities:
9540          + Conditional load/store
9541          + Distinct-operands
9542          + Floating-point-extension
9543          + Interlocked-access
9544          + Population-count
9545       The -mtune=z196 option avoids the compare and branch instructions
9546       as well as the load address instruction with an index register as
9547       much as possible and performs instruction scheduling appropriate
9548       for the new out-of-order pipeline architecture.
9549     * When using the -m31 -mzarch options the generated code still
9550       conforms to the 32-bit ABI but uses the general purpose registers
9551       as 64-bit registers internally. This requires a Linux kernel saving
9552       the whole 64-bit registers when doing a context switch. Kernels
9553       providing that feature indicate that by the 'highgprs' string in
9554       /proc/cpuinfo.
9555     * The SSA loop prefetching pass is enabled when using -O3.
9556
9557  SPARC
9558
9559     * GCC now supports the LEON series of SPARC V8 processors. The code
9560       generated by the compiler can either be tuned to it by means of the
9561       --with-tune=leon configure option and -mtune=leon compilation
9562       option, or the compiler can be built for the sparc-leon-{elf,linux}
9563       and sparc-leon3-{elf,linux} targets directly.
9564     * GCC has stopped sign/zero-extending parameter registers in the
9565       callee for functions taking parameters with sub-word size in 32-bit
9566       mode, since this is redundant with the specification of the ABI.
9567       GCC has never done so in 64-bit mode since this is also redundant.
9568     * The command-line option -mfix-at697f has been added to enable the
9569       documented workaround for the single erratum of the Atmel AT697F
9570       processor.
9571
9572Operating Systems
9573
9574  Android
9575
9576     * GCC now supports the Bionic C library and provides a convenient way
9577       of building native libraries and applications for the Android
9578       platform. Refer to the documentation of the -mandroid and -mbionic
9579       options for details on building native code. At the moment, Android
9580       support is enabled only for ARM.
9581
9582  Darwin/Mac OS X
9583
9584     * General
9585          + Initial support for CFString types has been added.
9586            This allows GCC to build projects including the system Core
9587            Foundation frameworks. The GCC Objective-C family supports
9588            CFString "toll-free bridged" as per the Mac OS X system tools.
9589            CFString is also recognized in the context of format
9590            attributes and arguments (see the documentation for format
9591            attributes for limitations). At present, 8-bit character types
9592            are supported.
9593          + Object file size reduction.
9594            The Darwin zeroed memory allocators have been re-written to
9595            make more use of .zerofill sections. For non-debug code, this
9596            can reduce object file size significantly.
9597          + Objective-C family 64-bit support (NeXT ABI 2).
9598            Initial support has been added to support 64-bit Objective-C
9599            code using the Darwin/OS X native (NeXT) runtime. ABI version
9600            2 will be selected automatically when 64-bit code is built.
9601          + Objective-C family 32-bit ABI 1.
9602            For 32-bit code ABI 1 is also now also allowed. At present it
9603            must be selected manually using -fobjc-abi-version=1 where
9604            applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6).
9605     * x86 Architecture
9606          + The -mdynamic-no-pic option has been enabled.
9607            Code supporting -mdynamic-no-pic optimization has been added
9608            and is applicable to -m32 builds. The compiler bootstrap uses
9609            the option where appropriate.
9610          + The default value for -mtune= has been changed.
9611            Since Darwin systems are primarily Xeon, Core-2 or similar the
9612            default tuning has been changed to -mtune=core2.
9613          + Enable 128-bit long double (__float128) support on Darwin.
9614     * PPC Architecture
9615          + Darwin64 ABI.
9616            Several significant bugs have been fixed, such that GCC now
9617            produces code compatible with the Darwin64 PowerPC ABI.
9618          + libffi and boehm-gc.
9619            The Darwin ports of the libffi and boehm-gc libraries have
9620            been upgraded to include a Darwin64 implementation. This means
9621            that powerpc*-*-darwin9 platforms may now, for example, build
9622            Java applications with -m64 enabled.
9623          + Plug-in support has been enabled.
9624          + The -fsection-anchors option is now available although,
9625            presently, not heavily tested.
9626
9627  Solaris 2
9628
9629    New Features
9630
9631     * Support symbol versioning with the Sun linker.
9632     * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+.
9633     * Support thread-local storage (TLS) with the Sun assembler on
9634       Solaris 2/x86.
9635     * Support TLS on Solaris 8/9 if prerequisites are met.
9636     * Support COMDAT group with the GNU assembler and recent Sun linker.
9637     * Support the Sun assembler visibility syntax.
9638     * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp.
9639       -march=pentiumpro (Solaris 8/9).
9640     * Don't use SSE on Solaris 8/9 x86 by default.
9641     * Enable 128-bit long double (__float128) support on Solaris 2/x86.
9642
9643    ABI Change
9644
9645     * Change the ABI for returning 8-byte vectors like __m64 in MMX
9646       registers on Solaris 10+/x86 to match the Sun Studio 12.1+
9647       compilers. This is an incompatible change. If you use such types,
9648       you must either recompile all your code with the new compiler or
9649       use the new -mvect8-ret-in-mem option to remain compatible with
9650       previous versions of GCC and Sun Studio.
9651
9652  Windows x86/x86_64
9653
9654     * Initial support for decimal floating point.
9655     * Support for the __thiscall calling-convention.
9656     * Support for hot-patchable function prologues via the
9657       ms_hook_prologue attribute for x86_64 in addition to 32-bit x86.
9658     * Improvements of stack-probing and stack-allocation mechanisms.
9659     * Support of push/pop-macro pragma as preprocessor command.
9660       With #pragma push_macro("macro-name") the current definition of
9661       macro-name is saved and can be restored with #pragma
9662       pop_macro("macro-name") to its saved definition.
9663     * Enable 128-bit long double (__float128) support on MinGW and
9664       Cygwin.
9665
9666Other significant improvements
9667
9668  Installation changes
9669
9670     * An install-strip make target is provided that installs stripped
9671       executables, and may install libraries with unneeded or debugging
9672       sections stripped.
9673     * On Power7 systems, there is a potential problem if you build the
9674       GCC compiler with a host compiler using options that enable the VSX
9675       instruction set generation. If the host compiler has been patched
9676       so that the vec_ld and vec_st builtin functions generate Altivec
9677       memory instructions instead of VSX memory instructions, then you
9678       should be able to build the compiler with VSX instruction
9679       generation.
9680
9681Changes for GCC Developers
9682
9683   Note: these changes concern developers that develop GCC itself or
9684   software that integrates with GCC, such as plugins, and not the general
9685   GCC users.
9686     * The gengtype utility, which previously was internal to the GCC
9687       build process, has been enchanced to provide GC root information
9688       for plugins as necessary.
9689     * The old GC allocation interface of ggc_alloc and friends was
9690       replaced with a type-safe alternative.
9691
9692GCC 4.6.1
9693
9694   This is the [19]list of problem reports (PRs) from GCC's bug tracking
9695   system that are known to be fixed in the 4.6.1 release. This list might
9696   not be complete (that is, it is possible that some PRs that have been
9697   fixed are not listed here).
9698
9699GCC 4.6.2
9700
9701   This is the [20]list of problem reports (PRs) from GCC's bug tracking
9702   system that are known to be fixed in the 4.6.2 release. This list might
9703   not be complete (that is, it is possible that some PRs that have been
9704   fixed are not listed here).
9705
9706GCC 4.6.3
9707
9708   This is the [21]list of problem reports (PRs) from GCC's bug tracking
9709   system that are known to be fixed in the 4.6.3 release. This list might
9710   not be complete (that is, it is possible that some PRs that have been
9711   fixed are not listed here).
9712
9713GCC 4.6.4
9714
9715   This is the [22]list of problem reports (PRs) from GCC's bug tracking
9716   system that are known to be fixed in the 4.6.4 release. This list might
9717   not be complete (that is, it is possible that some PRs that have been
9718   fixed are not listed here).
9719
9720
9721    For questions related to the use of GCC, please consult these web
9722    pages and the [23]GCC manuals. If that fails, the
9723    [24]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9724    web pages and the development of GCC are welcome on our developer
9725    list at [25]gcc@gcc.gnu.org. All of [26]our lists have public
9726    archives.
9727
9728   Copyright (C) [27]Free Software Foundation, Inc. Verbatim copying and
9729   distribution of this entire article is permitted in any medium,
9730   provided this notice is preserved.
9731
9732   These pages are [28]maintained by the GCC team. Last modified
9733   2023-01-18.
9734
9735References
9736
9737   1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401
9738   2. http://gcc.gnu.org/gcc-4.5/changes.html
9739   3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted
9740   4. http://gcc.gnu.org/gcc-4.6/porting_to.html
9741   5. http://gcc.gnu.org/projects/lto/whopr.pdf
9742   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
9743   7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html
9744   8. https://gcc.gnu.org/PR43145
9745   9. https://gcc.gnu.org/PR43680
9746  10. https://gcc.gnu.org/PR33558
9747  11. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253
9748  12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x
9749  13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
9750  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races
9751  15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
9752  16. https://gcc.gnu.org/wiki/OOP
9753  17. https://gcc.gnu.org/wiki/Coarray
9754  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233
9755  19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1
9756  20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2
9757  21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3
9758  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4
9759  23. https://gcc.gnu.org/onlinedocs/
9760  24. mailto:gcc-help@gcc.gnu.org
9761  25. mailto:gcc@gcc.gnu.org
9762  26. https://gcc.gnu.org/lists.html
9763  27. https://www.fsf.org/
9764  28. https://gcc.gnu.org/about.html
9765======================================================================
9766http://gcc.gnu.org/gcc-4.5/index.html
9767
9768                             GCC 4.5 Release Series
9769
9770   (This release series is no longer supported.)
9771
9772   Jul 2, 2012
9773
9774   The [1]GNU project and the GCC developers are pleased to announce the
9775   release of GCC 4.5.4.
9776
9777   This release is a bug-fix release, containing fixes for regressions in
9778   GCC 4.5.3 relative to previous releases of GCC.
9779
9780Release History
9781
9782   GCC 4.5.4
9783          Jul 2, 2012 ([2]changes)
9784
9785   GCC 4.5.3
9786          Apr 28, 2011 ([3]changes)
9787
9788   GCC 4.5.2
9789          Dec 16, 2010 ([4]changes)
9790
9791   GCC 4.5.1
9792          Jul 31, 2010 ([5]changes)
9793
9794   GCC 4.5.0
9795          April 14, 2010 ([6]changes)
9796
9797References and Acknowledgements
9798
9799   GCC used to stand for the GNU C Compiler, but since the compiler
9800   supports several other languages aside from C, it now stands for the
9801   GNU Compiler Collection.
9802
9803   A list of [7]successful builds is updated as new information becomes
9804   available.
9805
9806   The GCC developers would like to thank the numerous people that have
9807   contributed new features, improvements, bug fixes, and other changes as
9808   well as test results to GCC. This [8]amazing group of volunteers is
9809   what makes GCC successful.
9810
9811   For additional information about GCC please refer to the [9]GCC project
9812   web site or contact the [10]GCC development mailing list.
9813
9814   To obtain GCC please use [11]our mirror sites or [12]our version
9815   control system.
9816
9817
9818    For questions related to the use of GCC, please consult these web
9819    pages and the [13]GCC manuals. If that fails, the
9820    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9821    web pages and the development of GCC are welcome on our developer
9822    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
9823    archives.
9824
9825   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
9826   distribution of this entire article is permitted in any medium,
9827   provided this notice is preserved.
9828
9829   These pages are [18]maintained by the GCC team. Last modified
9830   2022-10-26.
9831
9832References
9833
9834   1. http://www.gnu.org/
9835   2. http://gcc.gnu.org/gcc-4.5/changes.html
9836   3. http://gcc.gnu.org/gcc-4.5/changes.html
9837   4. http://gcc.gnu.org/gcc-4.5/changes.html
9838   5. http://gcc.gnu.org/gcc-4.5/changes.html
9839   6. http://gcc.gnu.org/gcc-4.5/changes.html
9840   7. http://gcc.gnu.org/gcc-4.5/buildstat.html
9841   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9842   9. http://gcc.gnu.org/index.html
9843  10. mailto:gcc@gcc.gnu.org
9844  11. http://gcc.gnu.org/mirrors.html
9845  12. http://gcc.gnu.org/git.html
9846  13. https://gcc.gnu.org/onlinedocs/
9847  14. mailto:gcc-help@gcc.gnu.org
9848  15. mailto:gcc@gcc.gnu.org
9849  16. https://gcc.gnu.org/lists.html
9850  17. https://www.fsf.org/
9851  18. https://gcc.gnu.org/about.html
9852======================================================================
9853http://gcc.gnu.org/gcc-4.5/changes.html
9854
9855                             GCC 4.5 Release Series
9856                        Changes, New Features, and Fixes
9857
9858Caveats
9859
9860     * GCC now requires the [1]MPC library in order to build. See the
9861       [2]prerequisites page for version requirements.
9862     * Support for a number of older systems and recently unmaintained or
9863       untested target ports of GCC has been declared obsolete in GCC 4.5.
9864       Unless there is activity to revive them, the next release of GCC
9865       will have their sources permanently removed.
9866       The following ports for individual systems on particular
9867       architectures have been obsoleted:
9868          + IRIX releases before 6.5 (mips-sgi-irix5*,
9869            mips-sgi-irix6.[0-4])
9870          + Solaris 7 (*-*-solaris2.7)
9871          + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*,
9872            alpha-dec-osf5.0*)
9873          + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions
9874            can be found in the [3]announcement.
9875       Support for the classic POWER architecture implemented in the
9876       original RIOS and RIOS2 processors of the old IBM RS/6000 product
9877       line has been obsoleted in the rs6000 port. This does not affect
9878       the new generation Power and PowerPC architectures.
9879     * Support has been removed for all the [4]configurations obsoleted in
9880       GCC 4.4.
9881     * Support has been removed for the protoize and unprotoize utilities,
9882       obsoleted in GCC 4.4.
9883     * Support has been removed for tuning for Itanium1 (Merced) variants.
9884       Note that code tuned for Itanium2 should also run correctly on
9885       Itanium1.
9886     * GCC now generates unwind info also for epilogues. DWARF debuginfo
9887       generated by GCC now uses more features of DWARF3 than before, and
9888       also some DWARF4 features. GDB older than 7.0 is not able to handle
9889       either of these, so to debug GCC 4.5 generated binaries or
9890       libraries GDB 7.0 or later is needed. You can disable use of DWARF4
9891       features with the -gdwarf-3 -gstrict-dwarf options, or use
9892       -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but
9893       epilogue unwind info is emitted unconditionally whenever unwind
9894       info is emitted.
9895     * On x86 targets, code containing floating-point calculations may run
9896       significantly more slowly when compiled with GCC 4.5 in strict C99
9897       conformance mode than they did with earlier GCC versions. This is
9898       due to stricter standard conformance of the compiler and can be
9899       avoided by using the option -fexcess-precision=fast; also see
9900       [5]below.
9901     * The function attribute noinline no longer prevents GCC from cloning
9902       the function. A new attribute noclone has been introduced for this
9903       purpose. Cloning a function means that it is duplicated and the new
9904       copy is specialized for certain contexts (for example when a
9905       parameter is a known constant).
9906
9907General Optimizer Improvements
9908
9909     * The -save-temps now takes an optional argument. The -save-temps and
9910       -save-temps=cwd switches write the temporary files in the current
9911       working directory based on the original source file. The
9912       -save-temps=obj switch will write files into the directory
9913       specified with the -o option, and the intermediate filenames are
9914       based on the output file. This will allow the user to get the
9915       compiler intermediate files when doing parallel builds without two
9916       builds of the same filename located in different directories from
9917       interfering with each other.
9918     * Debugging dumps are now created in the same directory as the object
9919       file rather than in the current working directory. This allows the
9920       user to get debugging dumps when doing parallel builds without two
9921       builds of the same filename interfering with each other.
9922     * GCC has been integrated with the MPC library. This allows GCC to
9923       evaluate complex arithmetic at compile time [6]more accurately. It
9924       also allows GCC to evaluate calls to complex built-in math
9925       functions having constant arguments and replace them at compile
9926       time with their mathematically equivalent results. In doing so, GCC
9927       can generate correct results regardless of the math library
9928       implementation or floating point precision of the host platform.
9929       This also allows GCC to generate identical results regardless of
9930       whether one compiles in native or cross-compile configurations to a
9931       particular target. The following built-in functions take advantage
9932       of this new capability: cacos, cacosh, casin, casinh, catan,
9933       catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan,
9934       and ctanh. The float and long double variants of these functions
9935       (e.g. csinf and csinl) are also handled.
9936     * A new link-time optimizer has been added ([7]-flto). When this
9937       option is used, GCC generates a bytecode representation of each
9938       input file and writes it to specially-named sections in each object
9939       file. When the object files are linked together, all the function
9940       bodies are read from these named sections and instantiated as if
9941       they had been part of the same translation unit. This enables
9942       interprocedural optimizations to work across different files (and
9943       even different languages), potentially improving the performance of
9944       the generated code. To use the link-timer optimizer, -flto needs to
9945       be specified at compile time and during the final link. If the
9946       program does not require any symbols to be exported, it is possible
9947       to combine -flto and the experimental [8]-fwhopr with
9948       [9]-fwhole-program to allow the interprocedural optimizers to use
9949       more aggressive assumptions.
9950     * The automatic parallelization pass was enhanced to support
9951       parallelization of outer loops.
9952     * Automatic parallelization can be enabled as part of Graphite. In
9953       addition to -ftree-parallelize-loops=, specify
9954       -floop-parallelize-all to enable the Graphite-based optimization.
9955     * The infrastructure for optimizing based on [10]restrict qualified
9956       pointers has been rewritten and should result in code generation
9957       improvements. Optimizations based on restrict qualified pointers
9958       are now also available when using -fno-strict-aliasing.
9959     * There is a new optimization pass that attempts to change prototype
9960       of functions to avoid unused parameters, pass only relevant parts
9961       of structures and turn arguments passed by reference to arguments
9962       passed by value when possible. It is enabled by -O2 and above as
9963       well as -Os and can be manually invoked using the new command-line
9964       switch -fipa-sra.
9965     * GCC now optimize exception handling code. In particular cleanup
9966       regions that are proved to not have any effect are optimized out.
9967
9968New Languages and Language specific improvements
9969
9970  All languages
9971
9972     * The -fshow-column option is now on by default. This means error
9973       messages now have a column associated with them.
9974
9975  Ada
9976
9977     * Compilation of programs heavily using discriminated record types
9978       with variant parts has been sped up and generates more compact
9979       code.
9980     * Stack checking now works reasonably well on most plaforms. In some
9981       specific cases, stack overflows may still fail to be detected, but
9982       a compile-time warning will be issued for these cases.
9983
9984  C family
9985
9986     * If a header named in a #include directive is not found, the
9987       compiler exits immediately. This avoids a cascade of errors arising
9988       from declarations expected to be found in that header being
9989       missing.
9990     * A new built-in function __builtin_unreachable() has been added that
9991       tells the compiler that control will never reach that point. It may
9992       be used after asm statements that terminate by transferring control
9993       elsewhere, and in other places that are known to be unreachable.
9994     * The -Wlogical-op option now warns for logical expressions such as
9995       (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be
9996       mistakes. This option is disabled by default.
9997     * An asm goto feature has been added to allow asm statements that
9998       jump to C labels.
9999     * C++0x raw strings are supported for C++ and for C with -std=gnu99.
10000     * The deprecated attribute now takes an optional string argument, for
10001       example, __attribute__((deprecated("text string"))), that will be
10002       printed together with the deprecation warning.
10003
10004  C
10005
10006     * The -Wenum-compare option, which warns when comparing values of
10007       different enum types, now works for C. It formerly only worked for
10008       C++. This warning is enabled by -Wall. It may be avoided by using a
10009       type cast.
10010     * The -Wcast-qual option now warns about casts which are unsafe in
10011       that they permit const-correctness to be violated without further
10012       warnings. Specifically, it warns about cases where a qualifier is
10013       added when all the lower types are not const. For example, it warns
10014       about a cast from char ** to const char **.
10015     * The -Wc++-compat option is significantly improved. It issues new
10016       warnings for:
10017          + Using C++ reserved operator names as identifiers.
10018          + Conversions to enum types without explicit casts.
10019          + Using va_arg with an enum type.
10020          + Using different enum types in the two branches of ?:.
10021          + Using ++ or -- on a variable of enum type.
10022          + Using the same name as both a struct, union or enum tag and a
10023            typedef, unless the typedef refers to the tagged type itself.
10024          + Using a struct, union, or enum which is defined within another
10025            struct or union.
10026          + A struct field defined using a typedef if there is a field in
10027            the struct, or an enclosing struct, whose name is the typedef
10028            name.
10029          + Duplicate definitions at file scope.
10030          + Uninitialized const variables.
10031          + A global variable with an anonymous struct, union, or enum
10032            type.
10033          + Using a string constant to initialize a char array whose size
10034            is the length of the string.
10035     * The new -Wjump-misses-init option warns about cases where a goto or
10036       switch skips the initialization of a variable. This sort of branch
10037       is an error in C++ but not in C. This warning is enabled by
10038       -Wc++-compat.
10039     * GCC now ensures that a C99-conforming <stdint.h> is present on most
10040       targets, and uses information about the types in this header to
10041       implement the Fortran bindings to those types. GCC does not ensure
10042       the presence of such a header, and does not implement the Fortran
10043       bindings, on the following targets: NetBSD, VxWorks, VMS,
10044       SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF.
10045     * GCC now implements C90- and C99-conforming rules for constant
10046       expressions. This may cause warnings or errors for some code using
10047       expressions that can be folded to a constant but are not constant
10048       expressions as defined by ISO C.
10049     * All known target-independent C90 and C90 Amendment 1 conformance
10050       bugs, and all known target-independent C99 conformance bugs not
10051       related to floating point or extended identifiers, have been fixed.
10052     * The C decimal floating point support now includes support for the
10053       FLOAT_CONST_DECIMAL64 pragma.
10054     * The named address space feature from ISO/IEC TR 18037 is now
10055       supported. This is currently only implemented for the SPU
10056       processor.
10057
10058  C++
10059
10060     * Improved [11]experimental support for the upcoming C++0x ISO C++
10061       standard, including support for raw strings, lambda expressions and
10062       explicit type conversion operators.
10063     * When printing the name of a class template specialization, G++ will
10064       now omit any template arguments which come from default template
10065       arguments. This behavior (and the pretty-printing of function
10066       template specializations as template signature and arguments) can
10067       be disabled with the -fno-pretty-templates option.
10068     * Access control is now applied to typedef names used in a template,
10069       which may cause G++ to reject some ill-formed code that was
10070       accepted by earlier releases. The -fno-access-control option can be
10071       used as a temporary workaround until the code is corrected.
10072     * Compilation time for code that uses templates should now scale
10073       linearly with the number of instantiations rather than
10074       quadratically, as template instantiations are now looked up using
10075       hash tables.
10076     * Declarations of functions that look like builtin declarations of
10077       library functions are only considered to be redeclarations if they
10078       are declared with extern "C". This may cause problems with code
10079       that omits extern "C" on hand-written declarations of C library
10080       functions such as abort or memcpy. Such code is ill-formed, but was
10081       accepted by earlier releases.
10082     * Diagnostics that used to complain about passing non-POD types to
10083       ... or jumping past the declaration of a non-POD variable now check
10084       for triviality rather than PODness, as per C++0x.
10085     * In C++0x mode local and anonymous classes are now allowed as
10086       template arguments, and in declarations of variables and functions
10087       with linkage, so long as any such declaration that is used is also
10088       defined ([12]DR 757).
10089     * Labels may now have attributes, as has been permitted for a while
10090       in C. This is only permitted when the label definition and the
10091       attribute specifier is followed by a semicolon--i.e., the label
10092       applies to an empty statement. The only useful attribute for a
10093       label is unused.
10094     * G++ now implements [13]DR 176. Previously G++ did not support using
10095       the injected-class-name of a template base class as a type name,
10096       and lookup of the name found the declaration of the template in the
10097       enclosing scope. Now lookup of the name finds the
10098       injected-class-name, which can be used either as a type or as a
10099       template, depending on whether or not the name is followed by a
10100       template argument list. As a result of this change, some code that
10101       was previously accepted may be ill-formed because
10102         1. The injected-class-name is not accessible because it's from a
10103            private base, or
10104         2. The injected-class-name cannot be used as an argument for a
10105            template template parameter.
10106       In either of these cases, the code can be fixed by adding a
10107       nested-name-specifier to explicitly name the template. The first
10108       can be worked around with -fno-access-control; the second is only
10109       rejected with -pedantic.
10110     * A new standard mangling for SIMD vector types has been added, to
10111       avoid name clashes on systems with vectors of varying length. By
10112       default the compiler still uses the old mangling, but emits aliases
10113       with the new mangling on targets that support strong aliases. Users
10114       can switch over entirely to the new mangling with -fabi-version=4
10115       or -fabi-version=0. -Wabi will now warn about code that uses the
10116       old mangling.
10117     * The command-line option -ftemplate-depth-N is now written as
10118       -ftemplate-depth=N and the old form is deprecated.
10119     * Conversions between NULL and non-pointer types are now warned by
10120       default. The new option -Wno-conversion-null disables these
10121       warnings. Previously these warnings were only available when using
10122       -Wconversion explicitly.
10123
10124    Runtime Library (libstdc++)
10125
10126     * Improved experimental support for the upcoming ISO C++ standard,
10127       C++0x, including:
10128          + Support for <future>, <functional>, and <random>.
10129          + Existing facilities now exploit explicit operators and the
10130            newly implemented core C++0x features.
10131          + The header <cstdatomic> has been renamed to <atomic>.
10132     * An experimental [14]profile mode has been added. This is an
10133       implementation of many C++ standard library constructs with an
10134       additional analysis layer that gives performance improvement advice
10135       based on recognition of suboptimal usage patterns. For example,
10136#include <vector>
10137int main()
10138{
10139  std::vector<int> v;
10140  for (int k = 0; k < 1024; ++k)
10141    v.insert(v.begin(), k);
10142}
10143
10144       When instrumented via the profile mode, can return suggestions
10145       about the initial size and choice of the container used as follows:
10146vector-to-list: improvement = 5: call stack = 0x804842c ...
10147    : advice = change std::vector to std::list
10148vector-size: improvement = 3: call stack = 0x804842c ...
10149    : advice = change initial container size from 0 to 1024
10150
10151       These constructs can be substituted for the normal libstdc++
10152       constructs on a piecemeal basis, or all existing components can be
10153       transformed via the -D_GLIBCXX_PROFILE macro.
10154     * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR
10155       24733) has been added. This support is in header file
10156       <decimal/decimal>, uses namespace std::decimal, and includes
10157       classes decimal32, decimal64, and decimal128.
10158     * Sources have been audited for application of function attributes
10159       nothrow, const, pure, and noreturn.
10160     * Python pretty-printers have been added for many standard library
10161       components that simplify the internal representation and present a
10162       more intuitive view of components when used with
10163       appropriately-advanced versions of GDB. For more information,
10164       please consult the more [16]detailed description.
10165     * The default behavior for comparing typeinfo names has changed, so
10166       in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero.
10167     * The new -static-libstdc++ option directs g++ to link the C++
10168       library statically, even if the default would normally be to link
10169       it dynamically.
10170
10171  Fortran
10172
10173     * The COMMON default padding has been changed - instead of adding the
10174       padding before a variable it is now added afterwards, which
10175       increases the compatibility with other vendors and helps to obtain
10176       the correct output in some cases. Cf. also the -falign-commons
10177       option ([17]added in 4.4).
10178     * The -finit-real= option now also supports the value snan for
10179       signaling not-a-number; to be effective, one additionally needs to
10180       enable trapping (e.g. via -ffpe-trap=). Note: Compile-time
10181       optimizations can turn a signaling NaN into a quiet one.
10182     * The new option -fcheck= has been added with the options bounds,
10183       array-temps, do, pointer, and recursive. The bounds and array-temps
10184       options are equivalent to -fbounds-check and
10185       -fcheck-array-temporaries. The do option checks for invalid
10186       modification of loop iteration variables, and the recursive option
10187       tests for recursive calls to subroutines/functions which are not
10188       marked as recursive. With pointer pointer association checks in
10189       calls are performed; however, neither undefined pointers nor
10190       pointers in expressions are handled. Using -fcheck=all enables all
10191       these run-time checks.
10192     * The run-time checking -fcheck=bounds now warns about invalid string
10193       lengths of character dummy arguments. Additionally, more
10194       compile-time checks have been added.
10195     * The new option [18]-fno-protect-parens has been added; if set, the
10196       compiler may reorder REAL and COMPLEX expressions without regard to
10197       parentheses.
10198     * GNU Fortran no longer links against libgfortranbegin. As before,
10199       MAIN__ (assembler symbol name) is the actual Fortran main program,
10200       which is invoked by the main function. However, main is now
10201       generated and put in the same object file as MAIN__. For the time
10202       being, libgfortranbegin still exists for backward compatibility.
10203       For details see the new [19]Mixed-Language Programming chapter in
10204       the manual.
10205     * The I/O library was restructured for performance and cleaner code.
10206     * Array assignments and WHERE are now run in parallel when OpenMP's
10207       WORKSHARE is used.
10208     * The experimental option -fwhole-file was added. The option allows
10209       whole-file checking of procedure arguments and allows for better
10210       optimizations. It can also be used with -fwhole-program, which is
10211       now also supported in gfortran.
10212     * More Fortran 2003 and Fortran 2008 mathematical functions can now
10213       be used as initialization expressions.
10214     * Some extended attributes such as STDCALL are now supported via the
10215       [20]GCC$ compiler directive.
10216     * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN
10217       intrinsic behaves now as if zero were always positive.
10218     * For legacy compatibiliy: On Cygwin and MinGW, the special files
10219       CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now
10220       supported.
10221     * Fortran 2003 support has been extended:
10222          + Procedure-pointer function results and procedure-pointer
10223            components (including PASS),
10224          + allocatable scalars (experimental),
10225          + DEFERRED type-bound procedures,
10226          + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements
10227            have been implemented.
10228          + The ALLOCATE statement supports type-specs and the SOURCE=
10229            argument.
10230          + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC
10231            type-bound procedure (i.e. as type-bound operators).
10232          + Rounding (ROUND=, RZ, ...) for output is now supported.
10233          + The INT_FAST{8,16,32,64,128}_T kind type parameters of the
10234            intrinsic module ISO_C_BINDING are now supported, except for
10235            the targets listed above as ones where GCC does not have
10236            <stdint.h> type information.
10237          + Extensible derived types with type-bound procedure or
10238            procedure pointer with PASS attribute now have to use CLASS in
10239            line with the Fortran 2003 standard; the workaround to use
10240            TYPE is no longer supported.
10241          + [21]Experimental, incomplete support for polymorphism,
10242            including CLASS, SELECT TYPE and dynamic dispatch of
10243            type-bound procedure calls. Some features do not work yet such
10244            as unlimited polymorphism (CLASS(*)).
10245     * Fortran 2008 support has been extended:
10246          + The OPEN statement now supports the NEWUNIT= option, which
10247            returns a unique file unit, thus preventing inadvertent use of
10248            the same unit in different parts of the program.
10249          + Support for unlimited format items has been added.
10250          + The INT{8,16,32} and REAL{32,64,128} kind type parameters of
10251            the intrinsic module ISO_FORTRAN_ENV are now supported.
10252          + Using complex arguments with TAN, SINH, COSH, TANH, ASIN,
10253            ACOS, and ATAN is now possible; the functions ASINH, ACOSH,
10254            and ATANH have been added (for real and complex arguments) and
10255            ATAN(Y,X) is now an alias for ATAN2(Y,X).
10256          + The BLOCK construct has been implemented.
10257
10258New Targets and Target Specific Improvements
10259
10260  AIX
10261
10262     * Full cross-toolchain support now available with GNU Binutils
10263
10264  ARM
10265
10266     * GCC now supports the Cortex-M0 and Cortex-A5 processors.
10267     * GCC now supports the ARM v7E-M architecture.
10268     * GCC now supports VFPv4-based FPUs and FPUs with
10269       single-precision-only VFP.
10270     * GCC has many improvements to optimization for other ARM processors,
10271       including scheduling support for the integer pipeline on Cortex-A9.
10272     * GCC now supports the IEEE 754-2008 half-precision floating-point
10273       type, and a variant ARM-specific half-precision type. This type is
10274       specified using __fp16, with the layout determined by
10275       -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and
10276       VFPv4 half-precision instructions will be used.
10277     * GCC now supports the variant of AAPCS that uses VFP registers for
10278       parameter passing and return values.
10279
10280  AVR
10281
10282     * The -mno-tablejump option has been removed because it has the same
10283       effect as the -fno-jump-tables option.
10284     * Added support for these new AVR devices:
10285          + ATmega8U2
10286          + ATmega16U2
10287          + ATmega32U2
10288
10289  IA-32/x86-64
10290
10291     * GCC now will set the default for -march= based on the configure
10292       target.
10293     * GCC now supports handling floating-point excess precision arising
10294       from use of the x87 floating-point unit in a way that conforms to
10295       ISO C99. This is enabled with -fexcess-precision=standard and with
10296       standards conformance options such as -std=c99, and may be disabled
10297       using -fexcess-precision=fast.
10298     * Support for the Intel Atom processor is now available through the
10299       -march=atom and -mtune=atom options.
10300     * A new -mcrc32 option is now available to enable crc32 intrinsics.
10301     * A new -mmovbe option is now available to enable GCC to use the
10302       movbe instruction to implement __builtin_bswap32 and
10303       __builtin_bswap64.
10304     * SSE math now can be enabled by default at configure time with the
10305       new --with-fpmath=sse option.
10306     * There is a new intrinsic header file, <x86intrin.h>. It should be
10307       included before using any IA-32/x86-64 intrinsics.
10308     * Support for the XOP, FMA4, and LWP instruction sets for the AMD
10309       Orochi processors are now available with the -mxop, -mfma4, and
10310       -mlwp options.
10311     * The -mabm option enables GCC to use the popcnt and lzcnt
10312       instructions on AMD processors.
10313     * The -mpopcnt option enables GCC to use the popcnt instructions on
10314       both AMD and Intel processors.
10315
10316  M68K/ColdFire
10317
10318     * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x
10319       and 5441x devices.
10320     * GCC now supports thread-local storage (TLS) on M68K and ColdFire
10321       processors.
10322
10323  MeP
10324
10325   Support has been added for the Toshiba Media embedded Processor (MeP,
10326   or mep-elf) embedded target.
10327
10328  MIPS
10329
10330     * GCC now supports MIPS 1004K processors.
10331     * GCC can now be configured with options --with-arch-32,
10332       --with-arch-64, --with-tune-32 and --with-tune-64 to control the
10333       default optimization separately for 32-bit and 64-bit modes.
10334     * MIPS targets now support an alternative _mcount interface, in which
10335       register $12 points to the function's save slot for register $31.
10336       This interface is selected by the -mcount-ra-address option; see
10337       the documentation for more details.
10338     * GNU/Linux targets can now generate read-only .eh_frame sections.
10339       This optimization requires GNU binutils 2.20 or above, and is only
10340       available if GCC is configured with a suitable version of binutils.
10341     * GNU/Linux targets can now attach special relocations to indirect
10342       calls, so that the linker can turn them into direct jumps or
10343       branches. This optimization requires GNU binutils 2.20 or later,
10344       and is automatically selected if GCC is configured with an
10345       appropriate version of binutils. It can be explicitly enabled or
10346       disabled using the -mrelax-pic-calls command-line option.
10347     * GCC now generates more heavily-optimized atomic operations on
10348       Octeon processors.
10349     * MIPS targets now support the -fstack-protector option.
10350     * GCC now supports an -msynci option, which specifies that synci is
10351       enough to flush the instruction cache, without help from the
10352       operating system. GCC uses this information to optimize
10353       automatically-generated cache flush operations, such as those used
10354       for nested functions in C. There is also a --with-synci
10355       configure-time option, which makes -msynci the default.
10356     * GCC supports four new function attributes for interrupt handlers:
10357       interrupt, use_shadow_register_set, keep_interrupts_masked and
10358       use_debug_exception_return. See the documentation for more details
10359       about these attributes.
10360
10361  RS/6000 (POWER/PowerPC)
10362
10363     * GCC now supports the Power ISA 2.06, which includes the VSX
10364       instructions that add vector 64-bit floating point support, new
10365       population count instructions, and conversions between floating
10366       point and unsigned types.
10367     * Support for the power7 processor is now available through the
10368       -mcpu=power7 and -mtune=power7.
10369     * GCC will now vectorize loops that contain simple math functions
10370       like copysign when generating code for altivec or VSX targets.
10371     * Support for the A2 processor is now available through the -mcpu=a2
10372       and -mtune=a2 options.
10373     * Support for the 476 processor is now available through the
10374       -mcpu={476,476fp} and -mtune={476,476fp} options.
10375     * Support for the e500mc64 processor is now available through the
10376       -mcpu=e500mc64 and -mtune=e500mc64 options.
10377     * GCC can now be configured with options --with-cpu-32,
10378       --with-cpu-64, --with-tune-32 and --with-tune-64 to control the
10379       default optimization separately for 32-bit and 64-bit modes.
10380     * Starting with GCC 4.5.4, vectors of type vector long long or vector
10381       long are passed and returned in the same method as other vectors
10382       with the VSX instruction set. Previously the GCC compiler did not
10383       adhere to the ABI for 128-bit vectors with 64-bit integer base
10384       types (PR 48857). This is also fixed in the GCC 4.6.1 release.
10385
10386  RX
10387
10388   Support has been added for the Renesas RX Processor (rx-elf) target.
10389
10390Operating Systems
10391
10392  Windows (Cygwin and MinGW)
10393
10394     * GCC now installs all the major language runtime libraries as DLLs
10395       when configured with the --enable-shared option.
10396     * GCC now makes use of the new support for aligned common variables
10397       in versions of binutils >= 2.20 to fix bugs in the support for SSE
10398       data types.
10399     * Improvements to the libffi support library increase the reliability
10400       of code generated by GCJ on all Windows platforms. Libgcj is
10401       enabled by default for the first time.
10402     * Libtool improvements simplify installation by placing the generated
10403       DLLs in the correct binaries directory.
10404     * Numerous other minor bugfixes and improvements, and substantial
10405       enhancements to the Fortran language support library.
10406
10407   >
10408
10409Other significant improvements
10410
10411  Plugins
10412
10413     * It is now possible to extend the compiler without having to modify
10414       its source code. A new option -fplugin=file.so tells GCC to load
10415       the shared object file.so and execute it as part of the compiler.
10416       The internal documentation describes the details on how plugins can
10417       interact with the compiler.
10418
10419  Installation changes
10420
10421     * The move to newer autotools changed default installation
10422       directories and switches to control them: The --with-datarootdir,
10423       --with-docdir, --with-pdfdir, and --with-htmldir switches are not
10424       used any more. Instead, you can now use --datarootdir, --docdir,
10425       --htmldir, and --pdfdir. The default installation directories have
10426       changed as follows according to the GNU Coding Standards:
10427
10428       datarootdir read-only architecture-independent data root [PREFIX/share]
10429       localedir   locale-specific message catalogs [DATAROOTDIR/locale]
10430       docdir      documentation root [DATAROOTDIR/doc/PACKAGE]
10431       htmldir     html documentation [DOCDIR]
10432       dvidir      dvi documentation [DOCDIR]
10433       pdfdir      pdf documentation [DOCDIR]
10434       psdir       ps documentation [DOCDIR]
10435       The following variables have new default values:
10436
10437       datadir read-only architecture-independent data [DATAROOTDIR]
10438       infodir info documentation [DATAROOTDIR/info]
10439       mandir  man documentation [DATAROOTDIR/man]
10440
10441GCC 4.5.1
10442
10443   This is the [22]list of problem reports (PRs) from GCC's bug tracking
10444   system that are known to be fixed in the 4.5.1 release. This list might
10445   not be complete (that is, it is possible that some PRs that have been
10446   fixed are not listed here).
10447
10448  All languages
10449
10450     * GCC's new link-time optimizer ([23]-flto) now also works on a few
10451       non-ELF targets:
10452          + Cygwin (*-cygwin*)
10453          + MinGW (*-mingw*)
10454          + Darwin on x86-64 (x86_64-apple-darwin*)
10455       LTO is not enabled by default for these targets. To enable LTO, you
10456       should configure with the --enable-lto option.
10457
10458GCC 4.5.2
10459
10460   This is the [24]list of problem reports (PRs) from GCC's bug tracking
10461   system that are known to be fixed in the 4.5.2 release. This list might
10462   not be complete (that is, it is possible that some PRs that have been
10463   fixed are not listed here).
10464
10465GCC 4.5.3
10466
10467   This is the [25]list of problem reports (PRs) from GCC's bug tracking
10468   system that are known to be fixed in the 4.5.3 release. This list might
10469   not be complete (that is, it is possible that some PRs that have been
10470   fixed are not listed here).
10471
10472   On the PowerPC compiler, the Altivec builtin functions vec_ld and
10473   vec_st have been modified to generate the Altivec memory instructions
10474   LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5
10475   release, these builtin functions were changed to generate VSX memory
10476   reference instructions instead of Altivec memory instructions, but
10477   there are differences between the two instructions. If the VSX
10478   instruction set is available, you can now use the new builtin functions
10479   vec_vsx_ld and vec_vsx_st which always generates the VSX memory
10480   instructions.
10481
10482GCC 4.5.4
10483
10484   This is the [26]list of problem reports (PRs) from GCC's bug tracking
10485   system that are known to be fixed in the 4.5.4 release. This list might
10486   not be complete (that is, it is possible that some PRs that have been
10487   fixed are not listed here).
10488
10489
10490    For questions related to the use of GCC, please consult these web
10491    pages and the [27]GCC manuals. If that fails, the
10492    [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10493    web pages and the development of GCC are welcome on our developer
10494    list at [29]gcc@gcc.gnu.org. All of [30]our lists have public
10495    archives.
10496
10497   Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and
10498   distribution of this entire article is permitted in any medium,
10499   provided this notice is preserved.
10500
10501   These pages are [32]maintained by the GCC team. Last modified
10502   2023-01-15.
10503
10504References
10505
10506   1. https://www.multiprecision.org/mpc/
10507   2. https://gcc.gnu.org/install/prerequisites.html
10508   3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html
10509   4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted
10510   5. http://gcc.gnu.org/gcc-4.5/changes.html#x86
10511   6. https://gcc.gnu.org/PR30789
10512   7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
10513   8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802
10514   9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800
10515  10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html
10516  11. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html
10517  12. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757
10518  13. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176
10519  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html
10520  15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733
10521  16. https://sourceware.org/gdb/wiki/STLSupport
10522  17. http://gcc.gnu.org/gcc-4.4/changes.html
10523  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
10524  19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html
10525  20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html
10526  21. https://gcc.gnu.org/wiki/OOP
10527  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1
10528  23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
10529  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2
10530  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3
10531  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4
10532  27. https://gcc.gnu.org/onlinedocs/
10533  28. mailto:gcc-help@gcc.gnu.org
10534  29. mailto:gcc@gcc.gnu.org
10535  30. https://gcc.gnu.org/lists.html
10536  31. https://www.fsf.org/
10537  32. https://gcc.gnu.org/about.html
10538======================================================================
10539http://gcc.gnu.org/gcc-4.4/index.html
10540
10541                             GCC 4.4 Release Series
10542
10543   This release series is no longer maintained.
10544
10545   March 13, 2012
10546
10547   The [1]GNU project and the GCC developers are pleased to announce the
10548   release of GCC 4.4.7.
10549
10550   This release is a bug-fix release, containing fixes for regressions in
10551   GCC 4.4.6 relative to previous releases of GCC.
10552
10553Release History
10554
10555   GCC 4.4.7
10556          March 13, 2012 ([2]changes)
10557
10558   GCC 4.4.6
10559          April 16, 2011 ([3]changes)
10560
10561   GCC 4.4.5
10562          October 1, 2010 ([4]changes)
10563
10564   GCC 4.4.4
10565          April 29, 2010 ([5]changes)
10566
10567   GCC 4.4.3
10568          January 21, 2010 ([6]changes)
10569
10570   GCC 4.4.2
10571          October 15, 2009 ([7]changes)
10572
10573   GCC 4.4.1
10574          July 22, 2009 ([8]changes)
10575
10576   GCC 4.4.0
10577          April 21, 2009 ([9]changes)
10578
10579References and Acknowledgements
10580
10581   GCC used to stand for the GNU C Compiler, but since the compiler
10582   supports several other languages aside from C, it now stands for the
10583   GNU Compiler Collection.
10584
10585   A list of [10]successful builds is updated as new information becomes
10586   available.
10587
10588   The GCC developers would like to thank the numerous people that have
10589   contributed new features, improvements, bug fixes, and other changes as
10590   well as test results to GCC. This [11]amazing group of volunteers is
10591   what makes GCC successful.
10592
10593   For additional information about GCC please refer to the [12]GCC
10594   project web site or contact the [13]GCC development mailing list.
10595
10596   To obtain GCC please use [14]our mirror sites or [15]our version
10597   control system.
10598
10599
10600    For questions related to the use of GCC, please consult these web
10601    pages and the [16]GCC manuals. If that fails, the
10602    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10603    web pages and the development of GCC are welcome on our developer
10604    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
10605    archives.
10606
10607   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
10608   distribution of this entire article is permitted in any medium,
10609   provided this notice is preserved.
10610
10611   These pages are [21]maintained by the GCC team. Last modified
10612   2022-10-26.
10613
10614References
10615
10616   1. http://www.gnu.org/
10617   2. http://gcc.gnu.org/gcc-4.4/changes.html
10618   3. http://gcc.gnu.org/gcc-4.4/changes.html
10619   4. http://gcc.gnu.org/gcc-4.4/changes.html
10620   5. http://gcc.gnu.org/gcc-4.4/changes.html
10621   6. http://gcc.gnu.org/gcc-4.4/changes.html
10622   7. http://gcc.gnu.org/gcc-4.4/changes.html
10623   8. http://gcc.gnu.org/gcc-4.4/changes.html
10624   9. http://gcc.gnu.org/gcc-4.4/changes.html
10625  10. http://gcc.gnu.org/gcc-4.4/buildstat.html
10626  11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
10627  12. http://gcc.gnu.org/index.html
10628  13. mailto:gcc@gcc.gnu.org
10629  14. http://gcc.gnu.org/mirrors.html
10630  15. http://gcc.gnu.org/git.html
10631  16. https://gcc.gnu.org/onlinedocs/
10632  17. mailto:gcc-help@gcc.gnu.org
10633  18. mailto:gcc@gcc.gnu.org
10634  19. https://gcc.gnu.org/lists.html
10635  20. https://www.fsf.org/
10636  21. https://gcc.gnu.org/about.html
10637======================================================================
10638http://gcc.gnu.org/gcc-4.4/changes.html
10639
10640                             GCC 4.4 Release Series
10641                        Changes, New Features, and Fixes
10642
10643   The latest release in the 4.4 release series is [1]GCC 4.4.7.
10644
10645Caveats
10646
10647     * __builtin_stdarg_start has been completely removed from GCC.
10648       Support for <varargs.h> had been deprecated since GCC 4.0. Use
10649       __builtin_va_start as a replacement.
10650     * Some of the errors issued by the C++ front end that could be
10651       downgraded to warnings in previous releases by using -fpermissive
10652       are now warnings by default. They can be converted into errors by
10653       using -pedantic-errors.
10654     * Use of the cpp assertion extension will now emit a warning when
10655       -Wdeprecated or -pedantic is used. This extension has been
10656       deprecated for many years, but never warned about.
10657     * Packed bit-fields of type char were not properly bit-packed on many
10658       targets prior to GCC 4.4. On these targets, the fix in GCC 4.4
10659       causes an ABI change. For example there is no longer a 4-bit
10660       padding between field a and b in this structure:
10661    struct foo
10662    {
10663      char a:4;
10664      char b:8;
10665    } __attribute__ ((packed));
10666       There is a new warning to help identify fields that are affected:
10667    foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4
10668       The warning can be disabled with -Wno-packed-bitfield-compat.
10669     * On ARM EABI targets, the C++ mangling of the va_list type has been
10670       changed to conform to the current revision of the EABI. This does
10671       not affect the libstdc++ library included with GCC.
10672     * The SCOUNT and POS bits of the MIPS DSP control register are now
10673       treated as global. Previous versions of GCC treated these fields as
10674       call-clobbered instead.
10675     * The MIPS port no longer recognizes the h asm constraint. It was
10676       necessary to remove this constraint in order to avoid generating
10677       unpredictable code sequences.
10678       One of the main uses of the h constraint was to extract the high
10679       part of a multiplication on 64-bit targets. For example:
10680    asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y));
10681       You can now achieve the same effect using 128-bit types:
10682    typedef unsigned int uint128_t __attribute__((mode(TI)));
10683    result = ((uint128_t) x * y) >> 64;
10684       The second sequence is better in many ways. For example, if x and y
10685       are constants, the compiler can perform the multiplication at
10686       compile time. If x and y are not constants, the compiler can
10687       schedule the runtime multiplication better than it can schedule an
10688       asm statement.
10689     * Support for a number of older systems and recently unmaintained or
10690       untested target ports of GCC has been declared obsolete in GCC 4.4.
10691       Unless there is activity to revive them, the next release of GCC
10692       will have their sources permanently removed.
10693       The following ports for individual systems on particular
10694       architectures have been obsoleted:
10695          + Generic a.out on IA32 and m68k (i[34567]86-*-aout*,
10696            m68k-*-aout*)
10697          + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*,
10698            armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*,
10699            sh-*-*). This does not affect other more specific targets
10700            using the COFF object format on those architectures, or the
10701            more specific H8300 and SH targets (h8300-*-rtems*,
10702            h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*,
10703            sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks).
10704          + 2BSD on PDP-11 (pdp11-*-bsd)
10705          + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*,
10706            powerpc-ibm-aix4.[12]*)
10707          + Tuning support for Itanium1 (Merced) variants. Note that code
10708            tuned for Itanium2 should also run correctly on Itanium1.
10709     * The protoize and unprotoize utilities have been obsoleted and will
10710       be removed in GCC 4.5. These utilities have not been installed by
10711       default since GCC 3.0.
10712     * Support has been removed for all the [2]configurations obsoleted in
10713       GCC 4.3.
10714     * Unknown -Wno-* options are now silently ignored by GCC if no other
10715       diagnostics are issued. If other diagnostics are issued, then GCC
10716       warns about the unknown options.
10717     * More information on porting to GCC 4.4 from previous versions of
10718       GCC can be found in the [3]porting guide for this release.
10719
10720General Optimizer Improvements
10721
10722     * A new command-line switch -findirect-inlining has been added. When
10723       turned on it allows the inliner to also inline indirect calls that
10724       are discovered to have known targets at compile time thanks to
10725       previous inlining.
10726     * A new command-line switch -ftree-switch-conversion has been added.
10727       This new pass turns simple initializations of scalar variables in
10728       switch statements into initializations from a static array, given
10729       that all the values are known at compile time and the ratio between
10730       the new array size and the original switch branches does not exceed
10731       the parameter --param switch-conversion-max-branch-ratio (default
10732       is eight).
10733     * A new command-line switch -ftree-builtin-call-dce has been added.
10734       This optimization eliminates unnecessary calls to certain builtin
10735       functions when the return value is not used, in cases where the
10736       calls can not be eliminated entirely because the function may set
10737       errno. This optimization is on by default at -O2 and above.
10738     * A new command-line switch -fconserve-stack directs the compiler to
10739       minimize stack usage even if it makes the generated code slower.
10740       This affects inlining decisions.
10741     * When the assembler supports it, the compiler will now emit unwind
10742       information using assembler .cfi directives. This makes it possible
10743       to use such directives in inline assembler code. The new option
10744       -fno-dwarf2-cfi-asm directs the compiler to not use .cfi
10745       directives.
10746     * The [4]Graphite branch has been merged. This merge has brought in a
10747       new framework for loop optimizations based on a polyhedral
10748       intermediate representation. These optimizations apply to all the
10749       languages supported by GCC. The following new code transformations
10750       are available in GCC 4.4:
10751          + -floop-interchange performs loop interchange transformations
10752            on loops. Interchanging two nested loops switches the inner
10753            and outer loops. For example, given a loop like:
10754          DO J = 1, M
10755            DO I = 1, N
10756              A(J, I) = A(J, I) * C
10757            ENDDO
10758          ENDDO
10759
10760            loop interchange will transform the loop as if the user had
10761            written:
10762          DO I = 1, N
10763            DO J = 1, M
10764              A(J, I) = A(J, I) * C
10765            ENDDO
10766          ENDDO
10767
10768            which can be beneficial when N is larger than the caches,
10769            because in Fortran, the elements of an array are stored in
10770            memory contiguously by column, and the original loop iterates
10771            over rows, potentially creating at each access a cache miss.
10772          + -floop-strip-mine performs loop strip mining transformations
10773            on loops. Strip mining splits a loop into two nested loops.
10774            The outer loop has strides equal to the strip size and the
10775            inner loop has strides of the original loop within a strip.
10776            For example, given a loop like:
10777          DO I = 1, N
10778            A(I) = A(I) + C
10779          ENDDO
10780
10781            loop strip mining will transform the loop as if the user had
10782            written:
10783          DO II = 1, N, 4
10784            DO I = II, min (II + 3, N)
10785              A(I) = A(I) + C
10786            ENDDO
10787          ENDDO
10788
10789          + -floop-block performs loop blocking transformations on loops.
10790            Blocking strip mines each loop in the loop nest such that the
10791            memory accesses of the element loops fit inside caches. For
10792            example, given a loop like:
10793          DO I = 1, N
10794            DO J = 1, M
10795              A(J, I) = B(I) + C(J)
10796            ENDDO
10797          ENDDO
10798
10799            loop blocking will transform the loop as if the user had
10800            written:
10801          DO II = 1, N, 64
10802            DO JJ = 1, M, 64
10803              DO I = II, min (II + 63, N)
10804                DO J = JJ, min (JJ + 63, M)
10805                  A(J, I) = B(I) + C(J)
10806                ENDDO
10807              ENDDO
10808            ENDDO
10809          ENDDO
10810
10811            which can be beneficial when M is larger than the caches,
10812            because the innermost loop will iterate over a smaller amount
10813            of data that can be kept in the caches.
10814     * A new register allocator has replaced the old one. It is called
10815       integrated register allocator (IRA) because coalescing, register
10816       live range splitting, and hard register preferencing are done
10817       on-the-fly during coloring. It also has better integration with the
10818       reload pass. IRA is a regional register allocator which uses modern
10819       Chaitin-Briggs coloring instead of Chow's priority coloring used in
10820       the old register allocator. More info about IRA internals and
10821       options can be found in the GCC manuals.
10822     * A new instruction scheduler and software pipeliner, based on the
10823       selective scheduling approach, has been added. The new pass
10824       performs instruction unification, register renaming, substitution
10825       through register copies, and speculation during scheduling. The
10826       software pipeliner is able to pipeline non-countable loops. The new
10827       pass is targeted at scheduling-eager in-order platforms. In GCC 4.4
10828       it is available for the Intel Itanium platform working by default
10829       as the second scheduling pass (after register allocation) at the
10830       -O3 optimization level.
10831     * When using -fprofile-generate with a multi-threaded program, the
10832       profile counts may be slightly wrong due to race conditions. The
10833       new -fprofile-correction option directs the compiler to apply
10834       heuristics to smooth out the inconsistencies. By default the
10835       compiler will give an error message when it finds an inconsistent
10836       profile.
10837     * The new -fprofile-dir=PATH option permits setting the directory
10838       where profile data files are stored when using -fprofile-generate
10839       and friends, and the directory used when reading profile data files
10840       using -fprofile-use and friends.
10841
10842New warning options
10843
10844     * The new -Wframe-larger-than=NUMBER option directs GCC to emit a
10845       warning if any stack frame is larger than NUMBER bytes. This may be
10846       used to help ensure that code fits within a limited amount of stack
10847       space.
10848     * The command-line option -Wlarger-than-N is now written as
10849       -Wlarger-than=N and the old form is deprecated.
10850     * The new -Wno-mudflap option disables warnings about constructs
10851       which can not be instrumented when using -fmudflap.
10852
10853New Languages and Language specific improvements
10854
10855     * Version 3.0 of the OpenMP specification is now supported for the C,
10856       C++, and Fortran compilers.
10857     * New character data types, per [5]TR 19769: New character types in
10858       C, are now supported for the C compiler in -std=gnu99 mode, as
10859       __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in
10860       -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too.
10861
10862  C family
10863
10864     * A new optimize attribute was added to allow programmers to change
10865       the optimization level and particular optimization options for an
10866       individual function. You can also change the optimization options
10867       via the GCC optimize pragma for functions defined after the pragma.
10868       The GCC push_options pragma and the GCC pop_options pragma allow
10869       you temporarily save and restore the options used. The GCC
10870       reset_options pragma restores the options to what was specified on
10871       the command line.
10872     * Uninitialized warnings do not require enabling optimization
10873       anymore, that is, -Wuninitialized can be used together with -O0.
10874       Nonetheless, the warnings given by -Wuninitialized will probably be
10875       more accurate if optimization is enabled.
10876     * -Wparentheses now warns about expressions such as (!x | y) and (!x
10877       & y). Using explicit parentheses, such as in ((!x) | y), silences
10878       this warning.
10879     * -Wsequence-point now warns within if, while,do while and for
10880       conditions, and within for begin/end expressions.
10881     * A new option -dU is available to dump definitions of preprocessor
10882       macros that are tested or expanded.
10883
10884  C++
10885
10886     * [6]Improved experimental support for the upcoming ISO C++ standard,
10887       C++0x. Including support for auto, inline namespaces, generalized
10888       initializer lists, defaulted and deleted functions, new character
10889       types, and scoped enums.
10890     * Those errors that may be downgraded to warnings to build legacy
10891       code now mention -fpermissive when -fdiagnostics-show-option is
10892       enabled.
10893     * -Wconversion now warns if the result of a static_cast to enumeral
10894       type is unspecified because the value is outside the range of the
10895       enumeral type.
10896     * -Wuninitialized now warns if a non-static reference or non-static
10897       const member appears in a class without constructors.
10898     * G++ now properly implements value-initialization, so objects with
10899       an initializer of () and an implicitly defined default constructor
10900       will be zero-initialized before the default constructor is called.
10901
10902    Runtime Library (libstdc++)
10903
10904     * Improved experimental support for the upcoming ISO C++ standard,
10905       C++0x, including:
10906          + Support for <chrono>, <condition_variable>, <cstdatomic>,
10907            <forward_list>, <initializer_list>, <mutex>, <ratio>,
10908            <system_error>, and <thread>.
10909          + unique_ptr, <algorithm> additions, exception propagation, and
10910            support for the new character types in <string> and <limits>.
10911          + Existing facilities now exploit initializer lists, defaulted
10912            and deleted functions, and the newly implemented core C++0x
10913            features.
10914          + Some standard containers are more efficient together with
10915            stateful allocators, i.e., no allocator is constructed on the
10916            fly at element construction time.
10917     * Experimental support for non-standard pointer types in containers.
10918     * The long standing libstdc++/30928 has been fixed for targets
10919       running glibc 2.10 or later.
10920     * As usual, many small and larger bug fixes, in particular quite a
10921       few corner cases in <locale>.
10922
10923  Fortran
10924
10925     * GNU Fortran now employs libcpp directly instead of using cc1 as an
10926       external preprocessor. The [7]-cpp option was added to allow manual
10927       invocation of the preprocessor without relying on filename
10928       extensions.
10929     * The [8]-Warray-temporaries option warns about array temporaries
10930       generated by the compiler, as an aid to optimization.
10931     * The [9]-fcheck-array-temporaries option has been added, printing a
10932       notification at run time, when an array temporary had to be created
10933       for an function argument. Contrary to -Warray-temporaries the
10934       warning is only printed if the array is noncontiguous.
10935     * Improved generation of DWARF debugging symbols
10936     * If using an intrinsic not part of the selected standard (via -std=
10937       and -fall-intrinsics) gfortran will now treat it as if this
10938       procedure were declared EXTERNAL and try to link to a user-supplied
10939       procedure. -Wintrinsics-std will warn whenever this happens. The
10940       now-useless option -Wnonstd-intrinsic was removed.
10941     * The flag -falign-commons has been added to control the alignment of
10942       variables in COMMON blocks, which is enabled by default in line
10943       with previous GCC version. Using -fno-align-commons one can force
10944       commons to be contiguous in memory as required by the Fortran
10945       standard, however, this slows down the memory access. The option
10946       -Walign-commons, which is enabled by default, warns when padding
10947       bytes were added for alignment. The proper solution is to sort the
10948       common objects by decreasing storage size, which avoids the
10949       alignment problems.
10950     * Fortran 2003 support has been extended:
10951          + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is
10952            now supported (except internal reads from/writes to wide
10953            strings). [10]-fbackslash now supports also \unnnn and
10954            \Unnnnnnnn to enter Unicode characters.
10955          + Asynchronous I/O (implemented as synchronous I/O) and the
10956            decimal=, size=, sign=, pad=, blank=, and delim= specifiers
10957            are now supported in I/O statements.
10958          + Support for Fortran 2003 structure constructors and for array
10959            constructor with typespec has been added.
10960          + Procedure Pointers (but not yet as component in derived types
10961            and as function results) are now supported.
10962          + Abstract types, type extension, and type-bound procedures
10963            (both PROCEDURE and GENERIC but not as operators). Note: As
10964            CLASS/polymorphyic types are not implemented, type-bound
10965            procedures with PASS accept as non-standard extension TYPE
10966            arguments.
10967     * Fortran 2008 support has been added:
10968          + The -std=f2008 option and support for the file extensions
10969            .f2008 and .F2008 has been added.
10970          + The g0 format descriptor is now supported.
10971          + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH,
10972            ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED
10973            are now available (some of them existed as GNU extension
10974            before). Note: The hyperbolic functions are not yet supporting
10975            complex arguments and the three- argument version of BESSEL_*N
10976            is not available.
10977          + The bit intrinsics LEADZ and TRAILZ have been added.
10978
10979  Java (GCJ)
10980
10981  Ada
10982
10983     * The Ada runtime now supports multilibs on many platforms including
10984       x86_64, SPARC and PowerPC. Their build is enabled by default.
10985
10986New Targets and Target Specific Improvements
10987
10988  ARM
10989
10990     * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and
10991       Cortex-R4F processors and has many other improvements to
10992       optimization for ARM processors.
10993     * GCC now supports the VFPv3 variant with 16 double-precision
10994       registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been
10995       renamed to -mfpu=vfpv3.
10996     * GCC now supports the -mfix-cortex-m3-ldrd option to work around an
10997       erratum on Cortex-M3 processors.
10998     * GCC now supports the __sync_* atomic operations for ARM EABI
10999       GNU/Linux.
11000     * The section anchors optimization is now enabled by default when
11001       optimizing for ARM.
11002     * GCC now uses a new EABI-compatible profiling interface for EABI
11003       targets. This requires a function __gnu_mcount_nc, which is
11004       provided by GNU libc versions 2.8 and later.
11005
11006  AVR
11007
11008     * The -mno-tablejump option has been deprecated because it has the
11009       same effect as the -fno-jump-tables option.
11010     * Added support for these new AVR devices:
11011          + ATA6289
11012          + ATtiny13A
11013          + ATtiny87
11014          + ATtiny167
11015          + ATtiny327
11016          + ATmega8C1
11017          + ATmega16C1
11018          + ATmega32C1
11019          + ATmega8M1
11020          + ATmega16M1
11021          + ATmega32M1
11022          + ATmega32U4
11023          + ATmega16HVB
11024          + ATmega4HVD
11025          + ATmega8HVD
11026          + ATmega64C1
11027          + ATmega64M1
11028          + ATmega16U4
11029          + ATmega32U6
11030          + ATmega128RFA1
11031          + AT90PWM81
11032          + AT90SCR100
11033          + M3000F
11034          + M3000S
11035          + M3001B
11036
11037  IA-32/x86-64
11038
11039     * Support for Intel AES built-in functions and code generation is
11040       available via -maes.
11041     * Support for Intel PCLMUL built-in function and code generation is
11042       available via -mpclmul.
11043     * Support for Intel AVX built-in functions and code generation is
11044       available via -mavx.
11045     * Automatically align the stack for local variables with alignment
11046       requirement.
11047     * GCC can now utilize the SVML library for vectorizing calls to a set
11048       of C99 functions if -mveclibabi=svml is specified and you link to
11049       an SVML ABI compatible library.
11050     * On x86-64, the ABI has been changed in the following cases to
11051       conform to the x86-64 ABI:
11052          + Passing/returning structures with flexible array member:
11053  struct foo
11054    {
11055      int i;
11056      int flex[];
11057    };
11058          + Passing/returning structures with complex float member:
11059  struct foo
11060    {
11061      int i;
11062      __complex__ float f;
11063    };
11064          + Passing/returning unions with long double member:
11065  union foo
11066    {
11067      int x;
11068      long double ld;
11069    };
11070       Code built with previous versions of GCC that uses any of these is
11071       not compatible with code built with GCC 4.4.0 or later.
11072     * A new target attribute was added to allow programmers to change the
11073       target options like -msse2 or -march=k8 for an individual function.
11074       You can also change the target options via the GCC target pragma
11075       for functions defined after the pragma.
11076     * GCC can now be configured with options --with-arch-32,
11077       --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and
11078       --with-tune-64 to control the default optimization separately for
11079       32-bit and 64-bit modes.
11080
11081  IA-32/IA64
11082
11083     * Support for __float128 (TFmode) IEEE quad type and corresponding
11084       TCmode IEEE complex quad type is available via the soft-fp library
11085       on IA-32/IA64 targets. This includes basic arithmetic operations
11086       (addition, subtraction, negation, multiplication and division) on
11087       __float128 real and TCmode complex values, the full set of IEEE
11088       comparisons between __float128 values, conversions to and from
11089       float, double and long double floating point types, as well as
11090       conversions to and from signed or unsigned integer, signed or
11091       unsigned long integer and signed or unsigned quad (TImode, IA64
11092       only) integer types. Additionally, all operations generate the full
11093       set of IEEE exceptions and support the full set of IEEE rounding
11094       modes.
11095
11096  M68K/ColdFire
11097
11098     * GCC now supports instruction scheduling for ColdFire V1, V3 and V4
11099       processors. (Scheduling support for ColdFire V2 processors was
11100       added in GCC 4.3.)
11101     * GCC now supports the -mxgot option to support programs requiring
11102       many GOT entries on ColdFire.
11103     * The m68k-*-linux-gnu target now builds multilibs by default.
11104
11105  MIPS
11106
11107     * MIPS Technologies have extended the original MIPS SVR4 ABI to
11108       include support for procedure linkage tables (PLTs) and copy
11109       relocations. These extensions allow GNU/Linux executables to use a
11110       significantly more efficient code model than the one defined by the
11111       original ABI.
11112       GCC support for this code model is available via a new command-line
11113       option, -mplt. There is also a new configure-time option,
11114       --with-mips-plt, to make -mplt the default.
11115       The new code model requires support from the assembler, the linker,
11116       and the runtime C library. This support is available in binutils
11117       2.19 and GLIBC 2.9.
11118     * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables
11119       and 32-bit GNU/Linux shared libraries. This feature requires GNU
11120       binutils 2.19 or above.
11121     * Support for RMI's XLR processor is now available through the
11122       -march=xlr and -mtune=xlr options.
11123     * 64-bit targets can now perform 128-bit multiplications inline,
11124       instead of relying on a libgcc function.
11125     * Native GNU/Linux toolchains now support -march=native and
11126       -mtune=native, which select the host processor.
11127     * GCC now supports the R10K, R12K, R14K and R16K processors. The
11128       canonical -march= and -mtune= names for these processors are
11129       r10000, r12000, r14000 and r16000 respectively.
11130     * GCC can now work around the side effects of speculative execution
11131       on R10K processors. Please see the documentation of the
11132       -mr10k-cache-barrier option for details.
11133     * Support for the MIPS64 Release 2 instruction set has been added.
11134       The option -march=mips64r2 enables generation of these
11135       instructions.
11136     * GCC now supports Cavium Networks' Octeon processor. This support is
11137       available through the -march=octeon and -mtune=octeon options.
11138     * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The
11139       canonical -march= and -mtune= names for these processors are
11140       loongson2e and loongson2f.
11141
11142  picochip
11143
11144   Picochip is a 16-bit processor. A typical picoChip contains over 250
11145   small cores, each with small amounts of memory. There are three
11146   processor variants (STAN, MEM and CTRL) with different instruction sets
11147   and memory configurations and they can be chosen using the -mae option.
11148
11149   This port is intended to be a "C" only port.
11150
11151  Power Architecture and PowerPC
11152
11153     * GCC now supports the e300c2, e300c3 and e500mc processors.
11154     * GCC now supports Xilinx processors with a single-precision FPU.
11155     * Decimal floating point is now supported for e500 processors.
11156
11157  S/390, zSeries and System z9/z10
11158
11159     * Support for the IBM System z10 EC/BC processor has been added. When
11160       using the -march=z10 option, the compiler will generate code making
11161       use of instructions provided by the General-Instruction-Extension
11162       Facility and the Execute-Extension Facility.
11163
11164  VxWorks
11165
11166     * GCC now supports the thread-local storage mechanism used on
11167       VxWorks.
11168
11169  Xtensa
11170
11171     * GCC now supports thread-local storage (TLS) for Xtensa processor
11172       configurations that include the Thread Pointer option. TLS also
11173       requires support from the assembler and linker; this support is
11174       provided in the GNU binutils beginning with version 2.19.
11175
11176Documentation improvements
11177
11178Other significant improvements
11179
11180GCC 4.4.1
11181
11182   This is the [11]list of problem reports (PRs) from GCC's bug tracking
11183   system that are known to be fixed in the 4.4.1 release. This list might
11184   not be complete (that is, it is possible that some PRs that have been
11185   fixed are not listed here).
11186
11187GCC 4.4.2
11188
11189   This is the [12]list of problem reports (PRs) from GCC's bug tracking
11190   system that are known to be fixed in the 4.4.2 release. This list might
11191   not be complete (that is, it is possible that some PRs that have been
11192   fixed are not listed here).
11193
11194GCC 4.4.3
11195
11196   This is the [13]list of problem reports (PRs) from GCC's bug tracking
11197   system that are known to be fixed in the 4.4.3 release. This list might
11198   not be complete (that is, it is possible that some PRs that have been
11199   fixed are not listed here).
11200
11201GCC 4.4.4
11202
11203   This is the [14]list of problem reports (PRs) from GCC's bug tracking
11204   system that are known to be fixed in the 4.4.4 release. This list might
11205   not be complete (that is, it is possible that some PRs that have been
11206   fixed are not listed here).
11207
11208GCC 4.4.5
11209
11210   This is the [15]list of problem reports (PRs) from GCC's bug tracking
11211   system that are known to be fixed in the 4.4.5 release. This list might
11212   not be complete (that is, it is possible that some PRs that have been
11213   fixed are not listed here).
11214
11215GCC 4.4.6
11216
11217   This is the [16]list of problem reports (PRs) from GCC's bug tracking
11218   system that are known to be fixed in the 4.4.6 release. This list might
11219   not be complete (that is, it is possible that some PRs that have been
11220   fixed are not listed here).
11221
11222GCC 4.4.7
11223
11224   This is the [17]list of problem reports (PRs) from GCC's bug tracking
11225   system that are known to be fixed in the 4.4.7 release. This list might
11226   not be complete (that is, it is possible that some PRs that have been
11227   fixed are not listed here).
11228
11229
11230    For questions related to the use of GCC, please consult these web
11231    pages and the [18]GCC manuals. If that fails, the
11232    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11233    web pages and the development of GCC are welcome on our developer
11234    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
11235    archives.
11236
11237   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
11238   distribution of this entire article is permitted in any medium,
11239   provided this notice is preserved.
11240
11241   These pages are [23]maintained by the GCC team. Last modified
11242   2022-11-01.
11243
11244References
11245
11246   1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7
11247   2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted
11248   3. http://gcc.gnu.org/gcc-4.4/porting_to.html
11249   4. https://gcc.gnu.org/wiki/Graphite
11250   5. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf
11251   6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html
11252   7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html
11253   8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125
11254   9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221
11255  10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34
11256  11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1
11257  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2
11258  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3
11259  14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4
11260  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5
11261  16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6
11262  17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7
11263  18. https://gcc.gnu.org/onlinedocs/
11264  19. mailto:gcc-help@gcc.gnu.org
11265  20. mailto:gcc@gcc.gnu.org
11266  21. https://gcc.gnu.org/lists.html
11267  22. https://www.fsf.org/
11268  23. https://gcc.gnu.org/about.html
11269======================================================================
11270http://gcc.gnu.org/gcc-4.3/index.html
11271
11272                             GCC 4.3 Release Series
11273
11274   (This release series is no longer supported.)
11275
11276   Jun 27, 2011
11277
11278   The [1]GNU project and the GCC developers are pleased to announce the
11279   release of GCC 4.3.6.
11280
11281   This release is a bug-fix release, containing fixes for regressions in
11282   GCC 4.3.5 relative to previous releases of GCC.
11283
11284Release History
11285
11286   GCC 4.3.6
11287          Jun 27, 2011 ([2]changes)
11288
11289   GCC 4.3.5
11290          May 22, 2010 ([3]changes)
11291
11292   GCC 4.3.4
11293          August 4, 2009 ([4]changes)
11294
11295   GCC 4.3.3
11296          January 24, 2009 ([5]changes)
11297
11298   GCC 4.3.2
11299          August 27, 2008 ([6]changes)
11300
11301   GCC 4.3.1
11302          June 6, 2008 ([7]changes)
11303
11304   GCC 4.3.0
11305          March 5, 2008 ([8]changes)
11306
11307References and Acknowledgements
11308
11309   GCC used to stand for the GNU C Compiler, but since the compiler
11310   supports several other languages aside from C, it now stands for the
11311   GNU Compiler Collection.
11312
11313   A list of [9]successful builds is updated as new information becomes
11314   available.
11315
11316   The GCC developers would like to thank the numerous people that have
11317   contributed new features, improvements, bug fixes, and other changes as
11318   well as test results to GCC. This [10]amazing group of volunteers is
11319   what makes GCC successful.
11320
11321   For additional information about GCC please refer to the [11]GCC
11322   project web site or contact the [12]GCC development mailing list.
11323
11324   To obtain GCC please use [13]our mirror sites or [14]our version
11325   control system.
11326
11327
11328    For questions related to the use of GCC, please consult these web
11329    pages and the [15]GCC manuals. If that fails, the
11330    [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11331    web pages and the development of GCC are welcome on our developer
11332    list at [17]gcc@gcc.gnu.org. All of [18]our lists have public
11333    archives.
11334
11335   Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and
11336   distribution of this entire article is permitted in any medium,
11337   provided this notice is preserved.
11338
11339   These pages are [20]maintained by the GCC team. Last modified
11340   2022-10-26.
11341
11342References
11343
11344   1. http://www.gnu.org/
11345   2. http://gcc.gnu.org/gcc-4.3/changes.html
11346   3. http://gcc.gnu.org/gcc-4.3/changes.html
11347   4. http://gcc.gnu.org/gcc-4.3/changes.html
11348   5. http://gcc.gnu.org/gcc-4.3/changes.html
11349   6. http://gcc.gnu.org/gcc-4.3/changes.html
11350   7. http://gcc.gnu.org/gcc-4.3/changes.html
11351   8. http://gcc.gnu.org/gcc-4.3/changes.html
11352   9. http://gcc.gnu.org/gcc-4.3/buildstat.html
11353  10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11354  11. http://gcc.gnu.org/index.html
11355  12. mailto:gcc@gcc.gnu.org
11356  13. http://gcc.gnu.org/mirrors.html
11357  14. http://gcc.gnu.org/git.html
11358  15. https://gcc.gnu.org/onlinedocs/
11359  16. mailto:gcc-help@gcc.gnu.org
11360  17. mailto:gcc@gcc.gnu.org
11361  18. https://gcc.gnu.org/lists.html
11362  19. https://www.fsf.org/
11363  20. https://gcc.gnu.org/about.html
11364======================================================================
11365http://gcc.gnu.org/gcc-4.3/changes.html
11366
11367                             GCC 4.3 Release Series
11368                        Changes, New Features, and Fixes
11369
11370   The latest release in the 4.3 release series is [1]GCC 4.3.5.
11371
11372Caveats
11373
11374     * GCC requires the [2]GMP and [3]MPFR libraries for building all the
11375       various front-end languages it supports. See the [4]prerequisites
11376       page for version requirements.
11377     * ColdFire targets now treat long double as having the same format as
11378       double. In earlier versions of GCC, they used the 68881 long double
11379       format instead.
11380     * The m68k-uclinux target now uses the same calling conventions as
11381       m68k-linux-gnu. You can select the original calling conventions by
11382       configuring for m68k-uclinuxoldabi instead. Note that
11383       m68k-uclinuxoldabi also retains the original 80-bit long double on
11384       ColdFire targets.
11385     * The -fforce-mem option has been removed because it has had no
11386       effect in the last few GCC releases.
11387     * The i386 -msvr3-shlib option has been removed since it is no longer
11388       used.
11389     * Fastcall for i386 has been changed not to pass aggregate arguments
11390       in registers, following Microsoft compilers.
11391     * Support for the AOF assembler has been removed from the ARM back
11392       end; this affects only the targets arm-semi-aof and armel-semi-aof,
11393       which are no longer recognized. We removed these targets without a
11394       deprecation period because we discovered that they have been
11395       unusable since GCC 4.0.0.
11396     * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*)
11397       has been removed. This support had been deprecated since GCC 4.0.0.
11398     * Support for a number of older systems and recently unmaintained or
11399       untested target ports of GCC has been declared obsolete in GCC 4.3.
11400       Unless there is activity to revive them, the next release of GCC
11401       will have their sources permanently removed.
11402       All GCC ports for the following processor architectures have been
11403       declared obsolete:
11404          + Morpho MT (mt-*)
11405       The following aliases for processor architectures have been
11406       declared obsolete. Users should use the indicated generic target
11407       names instead, with compile-time options such as -mcpu or
11408       configure-time options such as --with-cpu to control the
11409       configuration more precisely.
11410          + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-*
11411            instead).
11412          + parisc*-*-* (use hppa*-*-* instead).
11413          + m680[012]0-*-* (use m68k-*-* instead).
11414       All GCC ports for the following operating systems have been
11415       declared obsolete:
11416          + BeOS (*-*-beos*)
11417          + kaOS (*-*-kaos*)
11418          + GNU/Linux using the a.out object format (*-*-linux*aout*)
11419          + GNU/Linux using version 1 of the GNU C Library
11420            (*-*-linux*libc1*)
11421          + Solaris versions before Solaris 7 (*-*-solaris2.[0-6],
11422            *-*-solaris2.[0-6].*)
11423          + Miscellaneous System V (*-*-sysv*)
11424          + WindISS (*-*-windiss*)
11425       Also, those for some individual systems on particular architectures
11426       have been obsoleted:
11427          + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
11428          + CRIS with a.out object format (cris-*-aout)
11429          + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
11430          + OSF/1 on PA-RISC (hppa1.1-*-osf*)
11431          + PRO on PA-RISC (hppa1.1-*-pro*)
11432          + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*,
11433            i[34567]86-sequent-sysv4*)
11434          + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
11435          + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host
11436            was previously [5]removed in 2001, leaving only the support
11437            for UWIN as a target now being deprecated)
11438          + ChorusOS on PowerPC (powerpc-*-chorusos*)
11439          + All VAX configurations apart from NetBSD and OpenBSD
11440            (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
11441     * The [6]-Wconversion option has been modified. Its purpose now is to
11442       warn for implicit conversions that may alter a value. This new
11443       behavior is available for both C and C++. Warnings about
11444       conversions between signed and unsigned integers can be disabled by
11445       using -Wno-sign-conversion. In C++, they are disabled by default
11446       unless -Wsign-conversion is explicitly requested. The old behavior
11447       of -Wconversion, that is, warn for prototypes causing a type
11448       conversion that is different from what would happen to the same
11449       argument in the absence of a prototype, has been moved to a new
11450       option -Wtraditional-conversion, which is only available for C.
11451     * The -m386, -m486, -mpentium and -mpentiumpro tuning options have
11452       been removed because they were deprecated for more than 3 GCC major
11453       releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or
11454       -mtune=pentiumpro as a replacement.
11455     * The -funsafe-math-optimizations option now automatically turns on
11456       -fno-trapping-math in addition to -fno-signed-zeros, as it enables
11457       reassociation and thus may introduce or remove traps.
11458     * The -ftree-vectorize option is now on by default under -O3. In
11459       order to generate code for a SIMD extension, it has to be enabled
11460       as well: use -maltivec for PowerPC platforms and -msse/-msse2 for
11461       i?86 and x86_64.
11462     * More information on porting to GCC 4.3 from previous versions of
11463       GCC can be found in the [7]porting guide for this release.
11464
11465General Optimizer Improvements
11466
11467     * The GCC middle-end has been integrated with the MPFR library. This
11468       allows GCC to evaluate and replace at compile-time calls to
11469       built-in math functions having constant arguments with their
11470       mathematically equivalent results. In making use of MPFR, GCC can
11471       generate correct results regardless of the math library
11472       implementation or floating point precision of the host platform.
11473       This also allows GCC to generate identical results regardless of
11474       whether one compiles in native or cross-compile configurations to a
11475       particular target. The following built-in functions take advantage
11476       of this new capability: acos, acosh, asin, asinh, atan2, atan,
11477       atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1,
11478       fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10,
11479       log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh,
11480       tan, tanh, tgamma, y0, y1 and yn. The float and long double
11481       variants of these functions (e.g. sinf and sinl) are also handled.
11482       The sqrt and cabs functions with constant arguments were already
11483       optimized in prior GCC releases. Now they also use MPFR.
11484     * A new forward propagation pass on RTL was added. The new pass
11485       replaces several slower transformations, resulting in compile-time
11486       improvements as well as better code generation in some cases.
11487     * A new command-line switch -frecord-gcc-switches has been added to
11488       GCC, although it is only enabled for some targets. The switch
11489       causes the command line that was used to invoke the compiler to be
11490       recorded into the object file that is being created. The exact
11491       format of this recording is target and binary file format
11492       dependent, but it usually takes the form of a note section
11493       containing ASCII text. The switch is related to the -fverbose-asm
11494       switch, but that one only records the information in the assembler
11495       output file as comments, so the information never reaches the
11496       object file.
11497     * The inliner heuristic is now aware of stack frame consumption. New
11498       command-line parameters --param large-stack-frame and --param
11499       large-stack-frame-growth can be used to limit stack frame size
11500       growth caused by inlining.
11501     * During feedback directed optimizations, the expected block size the
11502       memcpy, memset and bzero functions operate on is discovered and for
11503       cases of commonly used small sizes, specialized inline code is
11504       generated.
11505     * __builtin_expect no longer requires its argument to be a compile
11506       time constant.
11507     * Interprocedural optimization was reorganized to work on functions
11508       in SSA form. This enables more precise and cheaper dataflow
11509       analysis and makes writing interprocedural optimizations easier.
11510       The following improvements have been implemented on top of this
11511       framework:
11512          + Pre-inline optimization: Selected local optimization passes
11513            are run before the inliner (and other interprocedural passes)
11514            are executed. This significantly improves the accuracy of code
11515            growth estimates used by the inliner and reduces the overall
11516            memory footprint for large compilation units.
11517          + Early inlining (a simple bottom-up inliner pass inlining only
11518            functions whose body is smaller than the expected call
11519            overhead) is now executed with the early optimization passes,
11520            thus inlining already optimized function bodies into an
11521            unoptimized function that is subsequently optimized by early
11522            optimizers. This enables the compiler to quickly eliminate
11523            abstraction penalty in C++ programs.
11524          + Interprocedural constant propagation now operate on SSA form
11525            increasing accuracy of the analysis.
11526     * A new internal representation for GIMPLE statements has been
11527       contributed, resulting in compile-time memory savings.
11528     * The vectorizer was enhanced to support vectorization of outer
11529       loops, intra-iteration parallelism (loop-aware SLP), vectorization
11530       of strided accesses and loops with multiple data-types. Run-time
11531       dependency testing using loop versioning was added. The cost model,
11532       turned on by -fvect-cost-model, was developed.
11533
11534New Languages and Language specific improvements
11535
11536     * We have added new command-line options
11537       -finstrument-functions-exclude-function-list and
11538       -finstrument-functions-exclude-file-list. They provide more control
11539       over which functions are annotated by the -finstrument-functions
11540       option.
11541
11542  C family
11543
11544     * Implicit conversions between generic vector types are now only
11545       permitted when the two vectors in question have the same number of
11546       elements and compatible element types. (Note that the restriction
11547       involves compatible element types, not implicitly-convertible
11548       element types: thus, a vector type with element type int may not be
11549       implicitly converted to a vector type with element type unsigned
11550       int.) This restriction, which is in line with specifications for
11551       SIMD architectures such as AltiVec, may be relaxed using the flag
11552       -flax-vector-conversions. This flag is intended only as a
11553       compatibility measure and should not be used for new code.
11554     * -Warray-bounds has been added and is now enabled by default for
11555       -Wall . It produces warnings for array subscripts that can be
11556       determined at compile time to be always out of bounds.
11557       -Wno-array-bounds will disable the warning.
11558     * The constructor and destructor function attributes now accept
11559       optional priority arguments which control the order in which the
11560       constructor and destructor functions are run.
11561     * New [8]command-line options -Wtype-limits, -Wold-style-declaration,
11562       -Wmissing-parameter-type, -Wempty-body, -Wclobbered and
11563       -Wignored-qualifiers have been added for finer control of the
11564       diverse warnings enabled by -Wextra.
11565     * A new function attribute alloc_size has been added to mark up
11566       malloc style functions. For constant sized allocations this can be
11567       used to find out the size of the returned pointer using the
11568       __builtin_object_size() function for buffer overflow checking and
11569       similar. This supplements the already built-in malloc and calloc
11570       constant size handling.
11571     * Integer constants written in binary are now supported as a GCC
11572       extension. They consist of a prefix 0b or 0B, followed by a
11573       sequence of 0 and 1 digits.
11574     * A new predefined macro __COUNTER__ has been added. It expands to
11575       sequential integral values starting from 0. In conjunction with the
11576       ## operator, this provides a convenient means to generate unique
11577       identifiers.
11578     * A new command-line option -fdirectives-only has been added. It
11579       enables a special preprocessing mode which improves the performance
11580       of applications like distcc and ccache.
11581     * Fixed-point data types and operators have been added. They are
11582       based on Chapter 4 of the Embedded-C specification (n1169.pdf).
11583       Currently, only MIPS targets are supported.
11584     * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732,
11585       N1241, is now supported as a GCC extension to C for targets
11586       i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu,
11587       and x86_64-*-linux-gnu. The feature introduces new data types
11588       _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF,
11589       DD, and DL.
11590
11591  C++
11592
11593     * [9]Experimental support for the upcoming ISO C++ standard, C++0x.
11594     * -Wc++0x-compat has been added and is now enabled by default for
11595       -Wall. It produces warnings for constructs whose meaning differs
11596       between ISO C++ 1998 and C++0x.
11597     * The -Wparentheses option now works for C++ as it does for C. It
11598       warns if parentheses are omitted when operators with confusing
11599       precedence are nested. It also warns about ambiguous else
11600       statements. Since -Wparentheses is enabled by -Wall, this may cause
11601       additional warnings with existing C++ code which uses -Wall. These
11602       new warnings may be disabled by using -Wall -Wno-parentheses.
11603     * The -Wmissing-declarations now works for C++ as it does for C.
11604     * The -fvisibility-ms-compat flag was added, to make it easier to
11605       port larger projects using shared libraries from Microsoft's Visual
11606       Studio to ELF and Mach-O systems.
11607     * C++ attribute handling has been overhauled for template arguments
11608       (ie dependent types). In particular, __attribute__((aligned(T)));
11609       works for C++ types.
11610
11611    Runtime Library (libstdc++)
11612
11613     * [10]Experimental support for the upcoming ISO C++ standard, C++0x.
11614     * Support for TR1 mathematical special functions and regular
11615       expressions.
11616     * Default what implementations give more elaborate exception strings
11617       for bad_cast, bad_typeid, bad_exception, and bad_alloc.
11618     * Header dependencies have been streamlined, reducing unnecessary
11619       includes and pre-processed bloat.
11620     * Variadic template implementations of items in <tuple> and
11621       <functional>.
11622     * An experimental [11]parallel mode has been added. This is a
11623       parallel implementation of many C++ Standard library algorithms,
11624       like std::accumulate, std::for_each, std::transform, or std::sort,
11625       to give but four examples. These algorithms can be substituted for
11626       the normal (sequential) libstdc++ algorithms on a piecemeal basis,
11627       or all existing algorithms can be transformed via the
11628       -D_GLIBCXX_PARALLEL macro.
11629     * Debug mode versions of classes in <unordered_set> and
11630       <unordered_map>.
11631     * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are
11632       now <backward/hash_set> and <backward/hash_map>. This code:
11633    #include <ext/hash_set>
11634    __gnu_cxx::hash_set<int> s;
11635
11636       Can be transformed (in order of preference) to:
11637    #include <tr1/unordered_set>
11638    std::tr1::unordered_set<int> s;
11639
11640       or
11641    #include <backward/hash_set>
11642    __gnu_cxx::hash_set<int> s;
11643
11644       Similar transformations apply to __gnu_cxx::hash_map,
11645       __gnu_cxx::hash_multimap, __gnu_cxx::hash_set,
11646       __gnu_cxx::hash_multiset.
11647
11648  Fortran
11649
11650     * Due to the fact that the GMP and MPFR libraries are required for
11651       all languages, Fortran is no longer special in this regard and is
11652       available by default.
11653     * The [12]-fexternal-blas option has been added, which generates
11654       calls to BLAS routines for intrinsic matrix operations such as
11655       matmul rather than using the built-in algorithms.
11656     * Support to give a backtrace (compiler flag -fbacktrace or
11657       environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems
11658       only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a
11659       run-time error occured.
11660     * GNU Fortran now defines __GFORTRAN__ when it runs the C
11661       preprocessor (CPP).
11662     * The [13]-finit-local-zero, -finit-real, -finit-integer,
11663       -finit-character, and -finit-logical options have been added, which
11664       can be used to initialize local variables.
11665     * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added,
11666       which calculate the Gamma function and its logarithm. Use EXTERNAL
11667       gamma if you want to use your own gamma function.
11668     * GNU Fortran now regards the backslash character as literal (as
11669       required by the Fortran 2003 standard); using [16]-fbackslash GNU
11670       Fortran interprets backslashes as C-style escape characters.
11671     * The [17]interpretation of binary, octal and hexadecimal (BOZ)
11672       literal constants has been changed. Before they were always
11673       interpreted as integer; now they are bit-wise transferred as
11674       argument of INT, REAL, DBLE and CMPLX as required by the Fortran
11675       2003 standard, and for real and complex variables in DATA
11676       statements or when directly assigned to real and complex variables.
11677       Everywhere else and especially in expressions they are still
11678       regarded as integer constants.
11679     * Fortran 2003 support has been extended:
11680          + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
11681          + Pointer intent
11682          + Intrinsic module ISO_ENV_FORTRAN
11683          + Interoperability with C (ISO C Bindings)
11684          + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER
11685            attribute)
11686          + Fortran 2003 BOZ
11687
11688  Java (GCJ)
11689
11690     * GCJ now uses the Eclipse Java compiler for its Java parsing needs.
11691       This enables the use of all 1.5 language features, and fixes most
11692       existing front end bugs.
11693     * libgcj now supports all 1.5 language features which require runtime
11694       support: foreach, enum, annotations, generics, and auto-boxing.
11695     * We've made many changes to the tools shipped with gcj.
11696          + The old jv-scan tool has been removed. This tool never really
11697            worked properly. There is no replacement.
11698          + gcjh has been rewritten. Some of its more obscure options no
11699            longer work, but are still recognized in an attempt at
11700            compatibility. gjavah is a new program with similar
11701            functionality but different command-line options.
11702          + grmic and grmiregistry have been rewritten. grmid has been
11703            added.
11704          + gjar replaces the old fastjar.
11705          + gjarsigner (used for signing jars), gkeytool (used for key
11706            management), gorbd (for CORBA), gserialver (computes
11707            serialization UIDs), and gtnameserv (also for CORBA) are now
11708            installed.
11709     * The ability to dump the contents of the java run time heap to a
11710       file for off-line analysis has been added. The heap dumps may be
11711       analyzed with the new gc-analyze tool. They may be generated on
11712       out-of-memory conditions or on demand and are controlled by the new
11713       run time class gnu.gcj.util.GCInfo.
11714     * java.util.TimeZone can now read files from /usr/share/zoneinfo to
11715       provide correct, updated, timezone information. This means that
11716       packagers no longer have to update libgcj when a time zone change
11717       is published.
11718
11719New Targets and Target Specific Improvements
11720
11721  IA-32/x86-64
11722
11723     * Tuning for Intel Core 2 processors is available via -mtune=core2
11724       and -march=core2.
11725     * Tuning for AMD Geode processors is available via -mtune=geode and
11726       -march=geode.
11727     * Code generation of block move (memcpy) and block set (memset) was
11728       rewritten. GCC can now pick the best algorithm (loop, unrolled
11729       loop, instruction with rep prefix or a library call) based on the
11730       size of the block being copied and the CPU being optimized for. A
11731       new option -minline-stringops-dynamically has been added. With this
11732       option string operations of unknown size are expanded such that
11733       small blocks are copied by in-line code, while for large blocks a
11734       library call is used. This results in faster code than
11735       -minline-all-stringops when the library implementation is capable
11736       of using cache hierarchy hints. The heuristic choosing the
11737       particular algorithm can be overwritten via -mstringop-strategy.
11738       Newly also memset of values different from 0 is inlined.
11739     * GCC no longer places the cld instruction before string operations.
11740       Both i386 and x86-64 ABI documents mandate the direction flag to be
11741       clear at the entry of a function. It is now invalid to set the flag
11742       in asm statement without reseting it afterward.
11743     * Support for SSSE3 built-in functions and code generation are
11744       available via -mssse3.
11745     * Support for SSE4.1 built-in functions and code generation are
11746       available via -msse4.1.
11747     * Support for SSE4.2 built-in functions and code generation are
11748       available via -msse4.2.
11749     * Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
11750     * A new set of options -mpc32, -mpc64 and -mpc80 have been added to
11751       allow explicit control of x87 floating point precision.
11752     * Support for __float128 (TFmode) IEEE quad type and corresponding
11753       TCmode IEEE complex quad type is available via the soft-fp library
11754       on x86_64 targets. This includes basic arithmetic operations
11755       (addition, subtraction, negation, multiplication and division) on
11756       __float128 real and TCmode complex values, the full set of IEEE
11757       comparisons between __float128 values, conversions to and from
11758       float, double and long double floating point types, as well as
11759       conversions to and from signed or unsigned integer, signed or
11760       unsigned long integer and signed or unsigned quad (TImode) integer
11761       types. Additionally, all operations generate the full set of IEEE
11762       exceptions and support the full set of IEEE rounding modes.
11763     * GCC can now utilize the ACML library for vectorizing calls to a set
11764       of C99 functions on x86_64 if -mveclibabi=acml is specified and you
11765       link to an ACML ABI compatible library.
11766
11767  ARM
11768
11769     * Compiler and Library support for Thumb-2 and the ARMv7 architecture
11770       has been added.
11771
11772  CRIS
11773
11774    New features
11775
11776     * Compiler and Library support for the CRIS v32 architecture, as
11777       found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been
11778       added.
11779
11780    Configuration changes
11781
11782     * The cris-*-elf target now includes support for CRIS v32, including
11783       libraries, through the -march=v32 option.
11784     * A new crisv32-*-elf target defaults to generate code for CRIS v32.
11785     * A new crisv32-*-linux* target defaults to generate code for CRIS
11786       v32.
11787     * The cris-*-aout target has been obsoleted.
11788
11789    Improved support for built-in functions
11790
11791     * GCC can now use the lz and swapwbr instructions to implement the
11792       __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
11793     * __builtin_bswap32 is now implemented using the swapwb instruction,
11794       when available.
11795
11796  m68k and ColdFire
11797
11798    New features
11799
11800     * Support for several new ColdFire processors has been added. You can
11801       generate code for them using the new -mcpu option.
11802     * All targets now support ColdFire processors.
11803     * m68k-uclinux targets have improved support for C++ constructors and
11804       destructors, and for shared libraries.
11805     * It is now possible to set breakpoints on the first or last line of
11806       a function, even if there are no statements on that line.
11807
11808    Optimizations
11809
11810     * Support for sibling calls has been added.
11811     * More use is now made of the ColdFire mov3q instruction.
11812     * __builtin_clz is now implemented using the ff1 ColdFire
11813       instruction, when available.
11814     * GCC now honors the -m68010 option. 68010 code now uses clr rather
11815       than move to zero volatile memory.
11816     * 68020 targets and above can now use symbol(index.size*scale)
11817       addresses for indexed array accesses. Earlier compilers would
11818       always load the symbol into a base register first.
11819
11820    Configuration changes
11821
11822     * All m68k and ColdFire targets now allow the default processor to be
11823       set at configure time using --with-cpu.
11824     * A --with-arch configuration option has been added. This option
11825       allows you to restrict a target to ColdFire or non-ColdFire
11826       processors.
11827
11828    Preprocessor macros
11829
11830     * An __mcfv*__ macro is now defined for all ColdFire targets.
11831       (Earlier versions of GCC only defined __mcfv4e__.)
11832     * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
11833     * All targets now define __mc68010 and __mc68010__ when generating
11834       68010 code.
11835
11836    Command-line changes
11837
11838     * New command-line options -march, -mcpu, -mtune and -mhard-float
11839       have been added. These options apply to both m68k and ColdFire
11840       targets.
11841     * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative
11842       versions of -mshort, etc.
11843     * -fforce-addr has been removed. It is now ignored by the compiler.
11844
11845    Other improvements
11846
11847     * ColdFire targets now try to maintain a 4-byte-aligned stack where
11848       possible.
11849     * m68k-uclinux targets now try to avoid situations that lead to the
11850       load-time error: BINFMT_FLAT: reloc outside program.
11851
11852  MIPS
11853
11854    Changes to existing configurations
11855
11856     * libffi and libjava now support all three GNU/Linux ABIs: o32, n32
11857       and n64. Every GNU/Linux configuration now builds these libraries
11858       by default.
11859     * GNU/Linux configurations now generate -mno-shared code unless
11860       overridden by -fpic, -fPIC, -fpie or -fPIE.
11861     * mipsisa32*-linux-gnu configurations now generate hard-float code by
11862       default, just like other mipsisa32* and mips*-linux-gnu
11863       configurations. You can build a soft-float version of any
11864       mips*-linux-gnu configuration by passing --with-float=soft to
11865       configure.
11866     * mips-wrs-vxworks now supports run-time processes (RTPs).
11867
11868    Changes to existing command-line options
11869
11870     * The -march and -mtune options no longer accept 24k as a processor
11871       name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
11872     * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and
11873       34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The
11874       options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for
11875       24kx, 24kex and 34kx.
11876
11877    New configurations
11878
11879   GCC now supports the following configurations:
11880     * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by
11881       default. Earlier releases also recognized this configuration, but
11882       they treated it in the same way as mipsisa32*-linux-gnu*. Note that
11883       you can customize any mips*-linux-gnu* configuration to a
11884       particular ISA or processor by passing an appropriate --with-arch
11885       option to configure.
11886     * mipsisa*-sde-elf*, which provides compatibility with MIPS
11887       Technologies' SDE toolchains. The configuration uses the SDE
11888       libraries by default, but you can use it like other newlib-based
11889       ELF configurations by passing --with-newlib to configure. It is the
11890       only configuration besides mips64vr*-elf* to build MIPS16 as well
11891       as non-MIPS16 libraries.
11892     * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf*
11893       configuration, but uses the o32 and o64 ABIs instead of the 32-bit
11894       and 64-bit forms of the EABI.
11895
11896    New processors and application-specific extensions
11897
11898     * Support for the SmartMIPS ASE is available through the new
11899       -msmartmips option.
11900     * Support for revision 2 of the DSP ASE is available through the new
11901       -mdspr2 option. A new preprocessor macro called __mips_dsp_rev
11902       indicates the revision of the ASE in use.
11903     * Support for the 4KS and 74K families of processors is available
11904       through the -march and -mtune options.
11905
11906    Improved support for built-in functions
11907
11908     * GCC can now use load-linked, store-conditional and sync
11909       instructions to implement atomic built-in functions such as
11910       __sync_fetch_and_add. The memory reference must be 4 bytes wide for
11911       32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
11912     * GCC can now use the clz and dclz instructions to implement the
11913       __builtin_ctz and __builtin_ffs families of functions.
11914     * There is a new __builtin___clear_cache function for flushing the
11915       instruction cache. GCC expands this function inline on MIPS32
11916       revision 2 targets, otherwise it calls the function specified by
11917       -mcache-flush-func.
11918
11919    MIPS16 improvements
11920
11921     * GCC can now compile objects that contain a mixture of MIPS16 and
11922       non-MIPS16 code. There are two new attributes, mips16 and nomips16,
11923       for specifying which mode a function should use.
11924     * A new option called -minterlink-mips16 makes non-MIPS16 code
11925       link-compatible with MIPS16 code.
11926     * After many bug fixes, the long-standing MIPS16 -mhard-float support
11927       should now work fairly reliably.
11928     * GCC can now use the MIPS16e save and restore instructions.
11929     * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled
11930       with -G0 -fsection-anchors is often smaller than code compiled with
11931       -G8. However, please note that you must usually compile all objects
11932       in your application with the same -G option; see the documentation
11933       of -G for details.
11934     * A new option called-mcode-readable specifies which instructions are
11935       allowed to load from the code segment. -mcode-readable=yes is the
11936       default and says that any instruction may load from the code
11937       segment. The other alternatives are -mcode-readable=pcrel, which
11938       says that only PC-relative MIPS16 instructions may load from the
11939       code segment, and -mcode-readable=no, which says that no
11940       instruction may do so. Please see the documentation for more
11941       details, including example uses.
11942
11943    Small-data improvements
11944
11945   There are three new options for controlling small data:
11946     * -mno-extern-sdata, which disables small-data accesses for
11947       externally-defined variables. Code compiled with -Gn
11948       -mno-extern-sdata will be link-compatible with any -G setting
11949       between -G0 and -Gn inclusive.
11950     * -mno-local-sdata, which disables the use of small-data sections for
11951       data that is not externally visible. This option can be a useful
11952       way of reducing small-data usage in less performance-critical parts
11953       of an application.
11954     * -mno-gpopt, which disables the use of the $gp register while still
11955       honoring the -G limit when placing externally-visible data. This
11956       option implies -mno-extern-sdata and -mno-local-sdata and it can be
11957       useful in situations where $gp does not necessarily hold the
11958       expected value.
11959
11960    Miscellaneous improvements
11961
11962     * There is a new option called -mbranch-cost for tweaking the
11963       perceived cost of branches.
11964     * If GCC is configured to use a version of GAS that supports the
11965       .gnu_attribute directive, it will use that directive to record
11966       certain properties of the output code. .gnu_attribute is new to GAS
11967       2.18.
11968     * There are two new function attributes, near and far, for overriding
11969       the command-line setting of -mlong-calls on a function-by-function
11970       basis.
11971     * -mfp64, which previously required a 64-bit target, now works with
11972       MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and
11973       mipsisa*-sde-elf* configurations provide suitable library support.
11974     * GCC now recognizes the -mdmx and -mmt options and passes them down
11975       to the assembler. It does nothing else with the options at present.
11976
11977  SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture
11978  (BEA)
11979
11980     * Support has been added for this new architecture.
11981
11982  RS6000 (POWER/PowerPC)
11983
11984     * Support for the PowerPC 750CL paired-single instructions has been
11985       added with a new powerpc-*-linux*paired* target configuration. It
11986       is enabled by an associated -mpaired option and can be accessed
11987       using new built-in functions.
11988     * Support for auto-detecting architecture and system configuration to
11989       auto-select processor optimization tuning.
11990     * Support for VMX on AIX 5.3 has been added.
11991     * Support for AIX Version 6.1 has been added.
11992
11993  S/390, zSeries and System z9
11994
11995     * Support for the IBM System z9 EC/BC processor (z9 GA3) has been
11996       added. When using the -march=z9-ec option, the compiler will
11997       generate code making use of instructions provided by the decimal
11998       floating point facility and the floating point conversion facility
11999       (pfpo). Besides the instructions used to implement decimal floating
12000       point operations these facilities also contain instructions to move
12001       between general purpose and floating point registers and to modify
12002       and copy the sign-bit of floating point values.
12003     * When the -march=z9-ec option is used the new
12004       -mhard-dfp/-mno-hard-dfp options can be used to specify whether the
12005       decimal floating point hardware instructions will be used or not.
12006       If none of them is given the hardware support is enabled by
12007       default.
12008     * The -mstack-guard option can now be omitted when using stack
12009       checking via -mstack-size in order to let GCC choose a sensible
12010       stack guard value according to the frame size of each function.
12011     * Various changes to improve performance of generated code have been
12012       implemented, including:
12013          + The condition code set by an add logical with carry
12014            instruction is now available for overflow checks like: a + b +
12015            carry < b.
12016          + The test data class instruction is now used to implement
12017            sign-bit and infinity checks of binary and decimal floating
12018            point numbers.
12019
12020  SPARC
12021
12022     * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been
12023       added.
12024
12025  Xtensa
12026
12027     * Stack unwinding for exception handling now uses by default a
12028       specialized version of DWARF unwinding. This is not
12029       binary-compatible with the setjmp/longjmp (sjlj) unwinding used for
12030       Xtensa with previous versions of GCC.
12031     * For Xtensa processors that include the Conditional Store option,
12032       the built-in functions for atomic memory access are now implemented
12033       using S32C1I instructions.
12034     * If the Xtensa NSA option is available, GCC will use it to implement
12035       the __builtin_ctz and __builtin_clz functions.
12036
12037Documentation improvements
12038
12039     * Existing libstdc++ documentation has been edited and restructured
12040       into a single DocBook XML manual. The results can be viewed online
12041       [18]here.
12042
12043Other significant improvements
12044
12045     * The compiler's --help command-line option has been extended so that
12046       it now takes an optional set of arguments. These arguments restrict
12047       the information displayed to specific classes of command-line
12048       options, and possibly only a subset of those options. It is also
12049       now possible to replace the descriptive text associated with each
12050       displayed option with an indication of its current value, or for
12051       binary options, whether it has been enabled or disabled.
12052       Here are some examples. The following will display all the options
12053       controlling warning messages:
12054      --help=warnings
12055
12056       Whereas this will display all the undocumented, target specific
12057       options:
12058      --help=target,undocumented
12059
12060       This sequence of commands will display the binary optimizations
12061       that are enabled by -O3:
12062      gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
12063      gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
12064      diff /tmp/O2-opts /tmp/O3-opts | grep enabled
12065
12066     * The configure options --with-pkgversion and --with-bugurl have been
12067       added. These allow distributors of GCC to include a
12068       distributor-specific string in manuals and --version output and to
12069       specify the URL for reporting bugs in their versions of GCC.
12070
12071GCC 4.3.1
12072
12073   This is the [19]list of problem reports (PRs) from GCC's bug tracking
12074   system that are known to be fixed in the 4.3.1 release. This list might
12075   not be complete (that is, it is possible that some PRs that have been
12076   fixed are not listed here).
12077
12078Target Specific Changes
12079
12080  IA-32/x86-64
12081
12082    ABI changes
12083
12084     * Starting with GCC 4.3.1, decimal floating point variables are
12085       aligned to their natural boundaries when they are passed on the
12086       stack for i386.
12087
12088    Command-line changes
12089
12090     * Starting with GCC 4.3.1, the -mcld option has been added to
12091       automatically generate a cld instruction in the prologue of
12092       functions that use string instructions. This option is used for
12093       backward compatibility on some operating systems and can be enabled
12094       by default for 32-bit x86 targets by configuring GCC with the
12095       --enable-cld configure option.
12096
12097GCC 4.3.2
12098
12099   This is the [20]list of problem reports (PRs) from GCC's bug tracking
12100   system that are known to be fixed in the 4.3.2 release. This list might
12101   not be complete (that is, it is possible that some PRs that have been
12102   fixed are not listed here).
12103
12104GCC 4.3.3
12105
12106   This is the [21]list of problem reports (PRs) from GCC's bug tracking
12107   system that are known to be fixed in the 4.3.3 release. This list might
12108   not be complete (that is, it is possible that some PRs that have been
12109   fixed are not listed here).
12110
12111GCC 4.3.4
12112
12113   This is the [22]list of problem reports (PRs) from GCC's bug tracking
12114   system that are known to be fixed in the 4.3.4 release. This list might
12115   not be complete (that is, it is possible that some PRs that have been
12116   fixed are not listed here).
12117
12118GCC 4.3.5
12119
12120   This is the [23]list of problem reports (PRs) from GCC's bug tracking
12121   system that are known to be fixed in the 4.3.5 release. This list might
12122   not be complete (that is, it is possible that some PRs that have been
12123   fixed are not listed here).
12124
12125GCC 4.3.6
12126
12127   This is the [24]list of problem reports (PRs) from GCC's bug tracking
12128   system that are known to be fixed in the 4.3.6 release. This list might
12129   not be complete (that is, it is possible that some PRs that have been
12130   fixed are not listed here).
12131
12132
12133    For questions related to the use of GCC, please consult these web
12134    pages and the [25]GCC manuals. If that fails, the
12135    [26]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12136    web pages and the development of GCC are welcome on our developer
12137    list at [27]gcc@gcc.gnu.org. All of [28]our lists have public
12138    archives.
12139
12140   Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and
12141   distribution of this entire article is permitted in any medium,
12142   provided this notice is preserved.
12143
12144   These pages are [30]maintained by the GCC team. Last modified
12145   2022-10-26.
12146
12147References
12148
12149   1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5
12150   2. https://gmplib.org/
12151   3. https://www.mpfr.org/
12152   4. https://gcc.gnu.org/install/prerequisites.html
12153   5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html
12154   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
12155   7. http://gcc.gnu.org/gcc-4.3/porting_to.html
12156   8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
12157   9. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
12158  10. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
12159  11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
12160  12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options
12161  13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167
12162  14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html
12163  15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html
12164  16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html
12165  17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html
12166  18. https://gcc.gnu.org/onlinedocs/libstdc++/
12167  19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1
12168  20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2
12169  21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3
12170  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4
12171  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5
12172  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6
12173  25. https://gcc.gnu.org/onlinedocs/
12174  26. mailto:gcc-help@gcc.gnu.org
12175  27. mailto:gcc@gcc.gnu.org
12176  28. https://gcc.gnu.org/lists.html
12177  29. https://www.fsf.org/
12178  30. https://gcc.gnu.org/about.html
12179======================================================================
12180http://gcc.gnu.org/gcc-4.2/index.html
12181
12182                             GCC 4.2 Release Series
12183
12184   (This release series is no longer supported.)
12185
12186   May 19, 2008
12187
12188   The [1]GNU project and the GCC developers are pleased to announce the
12189   release of GCC 4.2.4.
12190
12191   This release is a bug-fix release, containing fixes for regressions in
12192   GCC 4.2.3 relative to previous releases of GCC.
12193
12194Release History
12195
12196   GCC 4.2.4
12197          May 19, 2008 ([2]changes)
12198
12199   GCC 4.2.3
12200          February 1, 2008 ([3]changes)
12201
12202   GCC 4.2.2
12203          October 7, 2007 ([4]changes)
12204
12205   GCC 4.2.1
12206          July 18, 2007 ([5]changes)
12207
12208   GCC 4.2.0
12209          May 13, 2007 ([6]changes)
12210
12211References and Acknowledgements
12212
12213   GCC used to stand for the GNU C Compiler, but since the compiler
12214   supports several other languages aside from C, it now stands for the
12215   GNU Compiler Collection.
12216
12217   A list of [7]successful builds is updated as new information becomes
12218   available.
12219
12220   The GCC developers would like to thank the numerous people that have
12221   contributed new features, improvements, bug fixes, and other changes as
12222   well as test results to GCC. This [8]amazing group of volunteers is
12223   what makes GCC successful.
12224
12225   For additional information about GCC please refer to the [9]GCC project
12226   web site or contact the [10]GCC development mailing list.
12227
12228   To obtain GCC please use [11]our mirror sites or [12]our version
12229   control system.
12230
12231
12232    For questions related to the use of GCC, please consult these web
12233    pages and the [13]GCC manuals. If that fails, the
12234    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12235    web pages and the development of GCC are welcome on our developer
12236    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
12237    archives.
12238
12239   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
12240   distribution of this entire article is permitted in any medium,
12241   provided this notice is preserved.
12242
12243   These pages are [18]maintained by the GCC team. Last modified
12244   2022-10-26.
12245
12246References
12247
12248   1. http://www.gnu.org/
12249   2. http://gcc.gnu.org/gcc-4.2/changes.html
12250   3. http://gcc.gnu.org/gcc-4.2/changes.html
12251   4. http://gcc.gnu.org/gcc-4.2/changes.html
12252   5. http://gcc.gnu.org/gcc-4.2/changes.html
12253   6. http://gcc.gnu.org/gcc-4.2/changes.html
12254   7. http://gcc.gnu.org/gcc-4.2/buildstat.html
12255   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12256   9. http://gcc.gnu.org/index.html
12257  10. mailto:gcc@gcc.gnu.org
12258  11. http://gcc.gnu.org/mirrors.html
12259  12. http://gcc.gnu.org/git.html
12260  13. https://gcc.gnu.org/onlinedocs/
12261  14. mailto:gcc-help@gcc.gnu.org
12262  15. mailto:gcc@gcc.gnu.org
12263  16. https://gcc.gnu.org/lists.html
12264  17. https://www.fsf.org/
12265  18. https://gcc.gnu.org/about.html
12266======================================================================
12267http://gcc.gnu.org/gcc-4.2/changes.html
12268
12269                             GCC 4.2 Release Series
12270                        Changes, New Features, and Fixes
12271
12272Caveats
12273
12274     * GCC no longer accepts the -fshared-data option. This option has had
12275       no effect in any GCC 4 release; the targets to which the option
12276       used to apply had been removed before GCC 4.0.
12277
12278General Optimizer Improvements
12279
12280     * New command-line options specify the possible relationships among
12281       parameters and between parameters and global data. For example,
12282       -fargument-noalias-anything specifies that arguments do not alias
12283       any other storage.
12284       Each language will automatically use whatever option is required by
12285       the language standard. You should not need to use these options
12286       yourself.
12287
12288New Languages and Language specific improvements
12289
12290     * [1]OpenMP is now supported for the C, C++ and Fortran compilers.
12291     * New command-line options -fstrict-overflow and -Wstrict-overflow
12292       have been added. -fstrict-overflow tells the compiler that it may
12293       assume that the program follows the strict signed overflow
12294       semantics permitted for the language: for C and C++ this means that
12295       the compiler may assume that signed overflow does not occur. For
12296       example, a loop like
12297      for (i = 1; i > 0; i *= 2)
12298
12299       is presumably intended to continue looping until i overflows. With
12300       -fstrict-overflow, the compiler may assume that signed overflow
12301       will not occur, and transform this into an infinite loop.
12302       -fstrict-overflow is turned on by default at -O2, and may be
12303       disabled via -fno-strict-overflow. The -Wstrict-overflow option may
12304       be used to warn about cases where the compiler assumes that signed
12305       overflow will not occur. It takes five different levels:
12306       -Wstrict-overflow=1 to 5. See the [2]documentation for details.
12307       -Wstrict-overflow=1 is enabled by -Wall.
12308     * The new command-line option -fno-toplevel-reorder directs GCC to
12309       emit top-level functions, variables, and asm statements in the same
12310       order that they appear in the input file. This is intended to
12311       support existing code which relies on a particular ordering (for
12312       example, code which uses top-level asm statements to switch
12313       sections). For new code, it is generally better to use function and
12314       variable attributes. The -fno-toplevel-reorder option may be used
12315       for most cases which currently use -fno-unit-at-a-time. The
12316       -fno-unit-at-a-time option will be removed in some future version
12317       of GCC. If you know of a case which requires -fno-unit-at-a-time
12318       which is not fixed by -fno-toplevel-reorder, please open a bug
12319       report.
12320
12321  C family
12322
12323     * The pragma redefine_extname will now macro expand its tokens for
12324       compatibility with SunPRO.
12325     * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct
12326       GCC to handle inline functions as specified in the C99 standard. In
12327       preparation for this, GCC 4.2 will warn about any use of non-static
12328       inline functions in gnu99 or c99 mode. This new warning may be
12329       disabled with the new gnu_inline function attribute or the new
12330       -fgnu89-inline command-line option. Also, GCC 4.2 and later will
12331       define one of the preprocessor macros __GNUC_GNU_INLINE__ or
12332       __GNUC_STDC_INLINE__ to indicate the semantics of inline functions
12333       in the current compilation.
12334     * A new command-line option -Waddress has been added to warn about
12335       suspicious uses of memory addresses as, for example, using the
12336       address of a function in a conditional expression, and comparisons
12337       against the memory address of a string literal. This warning is
12338       enabled by -Wall.
12339
12340  C++
12341
12342     * C++ visibility handling has been overhauled.
12343       Restricted visiblity is propagated from classes to members, from
12344       functions to local statics, and from templates and template
12345       arguments to instantiations, unless the latter has explicitly
12346       declared visibility.
12347       The visibility attribute for a class must come between the
12348       class-key and the name, not after the closing brace.
12349       Attributes are now allowed for enums and elaborated-type-specifiers
12350       that only declare a type.
12351       Members of the anonymous namespace are now local to a particular
12352       translation unit, along with any other declarations which use them,
12353       though they are still treated as having external linkage for
12354       language semantics.
12355     * The (undocumented) extension which permitted templates with default
12356       arguments to be bound to template template parameters with fewer
12357       parameters has been removed. For example:
12358        template <template <typename> class C>
12359        void f(C<double>) {}
12360
12361        template <typename T, typename U = int>
12362        struct S {};
12363
12364        template void f(S<double>);
12365
12366       is no longer accepted by G++. The reason this code is not accepted
12367       is that S is a template with two parameters; therefore, it cannot
12368       be bound to C which has only one parameter.
12369     * The <?, >?, <?=, and >?= operators, deprecated in previous GCC
12370       releases, have been removed.
12371     * The command-line option -fconst-strings, deprecated in previous GCC
12372       releases, has been removed.
12373     * The configure variable enable-__cxa_atexit is now enabled by
12374       default for more targets. Enabling this variable is necessary in
12375       order for static destructors to be executed in the correct order,
12376       but it depends upon the presence of a non-standard C library in the
12377       target library in order to work. The variable is now enabled for
12378       more targets which are known to have suitable C libraries.
12379     * -Wextra will produce warnings for if statements with a semicolon as
12380       the only body, to catch code like:
12381         if (a);
12382            return 1;
12383         return 0;
12384
12385       To suppress the warning in valid cases, use { } instead.
12386     * The C++ front end now also produces strict aliasing warnings when
12387       -fstrict-aliasing -Wstrict-aliasing is in effect.
12388
12389    Runtime Library (libstdc++)
12390
12391     * Added support for TR1 <random>, <complex>, and C compatibility
12392       headers. In addition, a lock-free version of shared_ptr was
12393       contributed as part of Phillip Jordan's Google Summer of Code
12394       project on lock-free containers.
12395     * In association with the Summer of Code work on lock-free
12396       containers, the interface for atomic builtins was adjusted,
12397       creating simpler alternatives for non-threaded code paths. Also,
12398       usage was consolidated and all elements were moved from namespace
12399       std to namespace__gnu_cxx. Affected interfaces are the functions
12400       __exchange_and_add, __atomic_add, and the objects __mutex,
12401       __recursive_mutex, and __scoped_lock.
12402     * Support for versioning weak symbol names via namespace association
12403       was added. However, as this changes the names of exported symbols,
12404       this is turned off by default in the current ABI. Intrepid users
12405       can enable this feature by using
12406       --enable-symvers=gnu-versioned-namespace during configuration.
12407     * Revised, simplified, and expanded policy-based associative
12408       containers, including data types for tree and trie forms
12409       (basic_tree, tree, trie), lists (list_update), and both
12410       collision-chaining and probing hash-based containers
12411       (basic_hash_table, cc_hash_table, gp_hash_table). More details per
12412       the [3]documentation.
12413     * The implementation of the debug mode was modified, whereby the
12414       debug namespaces were nested inside of namespace std and namespace
12415       __gnu_cxx in order to resolve some long standing corner cases
12416       involving name lookup. Debug functionality from the policy-based
12417       data structures was consolidated and enabled with the single macro,
12418       _GLIBCXX_DEBUG. See PR 26142 for more information.
12419     * Added extensions for type traits: __conditional_type,
12420       __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if.
12421     * Added a typelist implementation for compile-time meta-programming.
12422       Elements for typelist construction and operation can be found
12423       within namespace __gnu_cxx::typelist.
12424     * Added a new allocator, __gnu_cxx::throw_allocator, for testing
12425       exception-safety.
12426     * Enabled library-wide visibility control, allowing -fvisibility to
12427       be used.
12428     * Consolidated all nested namespaces and the conversion of
12429       __gnu_internal implementation-private details to anonymous
12430       namespaces whenever possible.
12431     * Implemented LWG resolutions DR 431 and DR 538.
12432
12433  Fortran
12434
12435     * Support for allocatable components has been added (TR 15581 and
12436       Fortran 2003).
12437     * Support for the Fortran 2003 streaming IO extension has been added.
12438     * The GNU Fortran compiler now uses 4-byte record markers by default
12439       for unformatted files to be compatible with g77 and most other
12440       compilers. The implementation allows for records greater than 2 GB
12441       and is compatible with several other compilers. Older versions of
12442       gfortran used 8-byte record markers by default (on most systems).
12443       In order to change the length of the record markers, e.g. to read
12444       unformatted files created by older gfortran versions, the
12445       [4]-frecord-marker=8 option can be used.
12446
12447  Java (GCJ)
12448
12449     * A new command-line option -static-libgcj has been added for targets
12450       that use a linker compatible with GNU Binutils. As its name
12451       implies, this causes libgcj to be linked statically. In some cases
12452       this causes the resulting executable to start faster and use less
12453       memory than if the shared version of libgcj were used. However
12454       caution should be used as it can also cause essential parts of the
12455       library to be omitted. Some of these issues are discussed in:
12456       [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj
12457     * fastjar is no longer bundled with GCC. To build libgcj, you will
12458       need either InfoZIP (both zip and unzip) or an external jar
12459       program. In the former case, the GCC build will install a jar shell
12460       script that is based on InfoZIP and provides the same functionality
12461       as fastjar.
12462
12463New Targets and Target Specific Improvements
12464
12465  IA-32/x86-64
12466
12467     * -mtune=generic can now be used to generate code running well on
12468       common x86 chips. This includes AMD Athlon, AMD Opteron, Intel
12469       Pentium-M, Intel Pentium 4 and Intel Core 2.
12470     * -mtune=native and -march=native will produce code optimized for the
12471       host architecture as detected using the cpuid instruction.
12472     * Added a new command-line option -fstackrealign and and
12473       __attribute__ ((force_align_arg_pointer)) to realign the stack at
12474       runtime. This allows functions compiled with a vector-aligned stack
12475       to be invoked from legacy objects that keep only word-alignment.
12476
12477  SPARC
12478
12479     * The default CPU setting has been changed from V7 to V9 in 32-bit
12480       mode on Solaris 7 and above. This is already the case in 64-bit
12481       mode. It can be overridden by specifying --with-cpu at configure
12482       time.
12483     * Back-end support of built-in functions for atomic memory access has
12484       been implemented.
12485     * Support for the Sun UltraSPARC T1 (Niagara) processor has been
12486       added.
12487
12488  M32C
12489
12490     * Various bug fixes have made some functions (notably, functions
12491       returning structures) incompatible with previous releases.
12492       Recompiling all libraries is recommended. Note that code quality
12493       has considerably improved since 4.1, making a recompile even more
12494       beneficial.
12495
12496  MIPS
12497
12498     * Added support for the Broadcom SB-1A core.
12499
12500  IA-64
12501
12502     * Added support for IA-64 data and control speculation. By default
12503       speculation is enabled only during second scheduler pass. A number
12504       of machine flags was introduced to control the usage of speculation
12505       for both scheduler passes.
12506
12507  HPPA
12508
12509     * Added Java language support (libffi and libjava) for 32-bit HP-UX
12510       11 target.
12511
12512Obsolete Systems
12513
12514Documentation improvements
12515
12516  PDF Documentation
12517
12518     * A make pdf target has been added to the top-level makefile,
12519       enabling automated production of PDF documentation files.
12520       (Front-ends external to GCC should modify their Make-lang.in file
12521       to add a lang.pdf: target.)
12522
12523Other significant improvements
12524
12525  Build system improvements
12526
12527     * All the components of the compiler are now bootstrapped by default.
12528       This improves the resilience to bugs in the system compiler or
12529       binary compatibility problems, as well as providing better testing
12530       of GCC 4.2 itself. In addition, if you build the compiler from a
12531       combined tree, the assembler, linker, etc. will also be
12532       bootstrapped (i.e. built with themselves).
12533       You can disable this behavior, and go back to the pre-GCC 4.2 set
12534       up, by configuring GCC with --disable-bootstrap.
12535     * The rules that configure follows to find target tools resemble more
12536       closely the locations that the built compiler will search. In
12537       addition, you can use the new configure option --with-target-tools
12538       to specify where to find the target tools used during the build,
12539       without affecting what the built compiler will use.
12540       This can be especially useful when building packages of GCC. For
12541       example, you may want to build GCC with GNU as or ld, even if the
12542       resulting compiler to work with the native assembler and linker. To
12543       do so, you can use --with-target-tools to point to the native
12544       tools.
12545
12546  Incompatible changes to the build system
12547
12548     * Front-ends external to GCC should modify their Make-lang.in file to
12549       replace double-colon rules (e.g. dvi::) with normal rules (like
12550       lang.dvi:). Front-end makefile hooks do not use double-colon rules
12551       anymore.
12552     * Up to GCC 4.1, a popular way to specify the target tools used
12553       during the build was to create directories named gas, binutils,
12554       etc. in the build tree, and create links to the tools from there.
12555       This does not work any more when the compiler is bootstrapped. The
12556       new configure option --with-target-tools provides a better way to
12557       achieve the same effect, and works for all native and cross
12558       settings.
12559
12560
12561    For questions related to the use of GCC, please consult these web
12562    pages and the [6]GCC manuals. If that fails, the
12563    [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12564    web pages and the development of GCC are welcome on our developer
12565    list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives.
12566
12567   Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and
12568   distribution of this entire article is permitted in any medium,
12569   provided this notice is preserved.
12570
12571   These pages are [11]maintained by the GCC team. Last modified
12572   2022-10-26.
12573
12574References
12575
12576   1. http://gcc.gnu.org/projects/gomp/
12577   2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
12578   3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html
12579   4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html
12580   5. https://gcc.gnu.org/wiki/Statically_linking_libgcj
12581   6. https://gcc.gnu.org/onlinedocs/
12582   7. mailto:gcc-help@gcc.gnu.org
12583   8. mailto:gcc@gcc.gnu.org
12584   9. https://gcc.gnu.org/lists.html
12585  10. https://www.fsf.org/
12586  11. https://gcc.gnu.org/about.html
12587======================================================================
12588http://gcc.gnu.org/gcc-4.1/index.html
12589
12590                             GCC 4.1 Release Series
12591
12592   (This release series is no longer supported.)
12593
12594   February 13, 2007
12595
12596   The [1]GNU project and the GCC developers are pleased to announce the
12597   release of GCC 4.1.2.
12598
12599   This release is a bug-fix release, containing fixes for regressions in
12600   GCC 4.1.1 relative to previous releases of GCC.
12601
12602Release History
12603
12604   GCC 4.1.2
12605          February 13, 2007 ([2]changes)
12606
12607   GCC 4.1.1
12608          May 24, 2006 ([3]changes)
12609
12610   GCC 4.1.0
12611          February 28, 2006 ([4]changes)
12612
12613References and Acknowledgements
12614
12615   GCC used to stand for the GNU C Compiler, but since the compiler
12616   supports several other languages aside from C, it now stands for the
12617   GNU Compiler Collection.
12618
12619   A list of [5]successful builds is updated as new information becomes
12620   available.
12621
12622   The GCC developers would like to thank the numerous people that have
12623   contributed new features, improvements, bug fixes, and other changes as
12624   well as test results to GCC. This [6]amazing group of volunteers is
12625   what makes GCC successful.
12626
12627   For additional information about GCC please refer to the [7]GCC project
12628   web site or contact the [8]GCC development mailing list.
12629
12630   To obtain GCC please use [9]our mirror sites or [10]our version control
12631   system.
12632
12633
12634    For questions related to the use of GCC, please consult these web
12635    pages and the [11]GCC manuals. If that fails, the
12636    [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12637    web pages and the development of GCC are welcome on our developer
12638    list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
12639    archives.
12640
12641   Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
12642   distribution of this entire article is permitted in any medium,
12643   provided this notice is preserved.
12644
12645   These pages are [16]maintained by the GCC team. Last modified
12646   2022-10-26.
12647
12648References
12649
12650   1. http://www.gnu.org/
12651   2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
12652   3. http://gcc.gnu.org/gcc-4.1/changes.html
12653   4. http://gcc.gnu.org/gcc-4.1/changes.html
12654   5. http://gcc.gnu.org/gcc-4.1/buildstat.html
12655   6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12656   7. http://gcc.gnu.org/index.html
12657   8. mailto:gcc@gcc.gnu.org
12658   9. http://gcc.gnu.org/mirrors.html
12659  10. http://gcc.gnu.org/git.html
12660  11. https://gcc.gnu.org/onlinedocs/
12661  12. mailto:gcc-help@gcc.gnu.org
12662  13. mailto:gcc@gcc.gnu.org
12663  14. https://gcc.gnu.org/lists.html
12664  15. https://www.fsf.org/
12665  16. https://gcc.gnu.org/about.html
12666======================================================================
12667http://gcc.gnu.org/gcc-4.1/changes.html
12668
12669                             GCC 4.1 Release Series
12670                        Changes, New Features, and Fixes
12671
12672   The latest release in the 4.1 release series is [1]GCC 4.1.2.
12673
12674Caveats
12675
12676General Optimizer Improvements
12677
12678     * GCC now has infrastructure for inter-procedural optimizations and
12679       the following inter-procedural optimizations are implemented:
12680          + Profile guided inlining. When doing profile feedback guided
12681            optimization, GCC can now use the profile to make better
12682            informed decisions on whether inlining of a function is
12683            profitable or not. This means that GCC will no longer inline
12684            functions at call sites that are not executed very often, and
12685            that functions at hot call sites are more likely to be
12686            inlined.
12687            A new parameter min-inline-recursive-probability is also now
12688            available to throttle recursive inlining of functions with
12689            small average recursive depths.
12690          + Discovery of pure and const functions, a form of side-effects
12691            analysis. While older GCC releases could also discover such
12692            special functions, the new IPA-based pass runs earlier so that
12693            the results are available to more optimizers. The pass is also
12694            simply more powerful than the old one.
12695          + Analysis of references to static variables and type escape
12696            analysis, also forms of side-effects analysis. The results of
12697            these passes allow the compiler to be less conservative about
12698            call-clobbered variables and references. This results in more
12699            redundant loads being eliminated and in making static
12700            variables candidates for register promotion.
12701          + Improvement of RTL-based alias analysis. The results of type
12702            escape analysis are fed to the RTL type-based alias analyzer,
12703            allowing it to disambiguate more memory references.
12704          + Interprocedural constant propagation and function versioning.
12705            This pass looks for functions that are always called with the
12706            same constant value for one or more of the function arguments,
12707            and propagates those constants into those functions.
12708          + GCC will now eliminate static variables whose usage was
12709            optimized out.
12710          + -fwhole-program --combine can now be used to make all
12711            functions in program static allowing whole program
12712            optimization. As an exception, the main function and all
12713            functions marked with the new externally_visible attribute are
12714            kept global so that programs can link with runtime libraries.
12715     * GCC can now do a form of partial dead code elimination (PDCE) that
12716       allows code motion of expressions to the paths where the result of
12717       the expression is actually needed. This is not always a win, so the
12718       pass has been limited to only consider profitable cases. Here is an
12719       example:
12720    int foo (int *, int *);
12721    int
12722    bar (int d)
12723    {
12724      int a, b, c;
12725      b = d + 1;
12726      c = d + 2;
12727      a = b + c;
12728      if (d)
12729        {
12730          foo (&b, &c);
12731          a = b + c;
12732        }
12733      printf ("%d\n", a);
12734    }
12735
12736       The a = b + c can be sunk to right before the printf. Normal code
12737       sinking will not do this, it will sink the first one above into the
12738       else-branch of the conditional jump, which still gives you two
12739       copies of the code.
12740     * GCC now has a value range propagation pass. This allows the
12741       compiler to eliminate bounds checks and branches. The results of
12742       the pass can also be used to accurately compute branch
12743       probabilities.
12744     * The pass to convert PHI nodes to straight-line code (a form of
12745       if-conversion for GIMPLE) has been improved significantly. The two
12746       most significant improvements are an improved algorithm to
12747       determine the order in which the PHI nodes are considered, and an
12748       improvement that allow the pass to consider if-conversions of basic
12749       blocks with more than two predecessors.
12750     * Alias analysis improvements. GCC can now differentiate between
12751       different fields of structures in Tree-SSA's virtual operands form.
12752       This lets stores/loads from non-overlapping structure fields not
12753       conflict. A new algorithm to compute points-to sets was contributed
12754       that can allows GCC to see now that p->a and p->b, where p is a
12755       pointer to a structure, can never point to the same field.
12756     * Various enhancements to auto-vectorization:
12757          + Incrementally preserve SSA form when vectorizing.
12758          + Incrementally preserve loop-closed form when vectorizing.
12759          + Improvements to peeling for alignment: generate better code
12760            when the misalignment of an access is known at compile time,
12761            or when different accesses are known to have the same
12762            misalignment, even if the misalignment amount itself is
12763            unknown.
12764          + Consider dependence distance in the vectorizer.
12765          + Externalize generic parts of data reference analysis to make
12766            this analysis available to other passes.
12767          + Vectorization of conditional code.
12768          + Reduction support.
12769     * GCC can now partition functions in sections of hot and cold code.
12770       This can significantly improve performance due to better
12771       instruction cache locality. This feature works best together with
12772       profile feedback driven optimization.
12773     * A new pass to avoid saving of unneeded arguments to the stack in
12774       vararg functions if the compiler can prove that they will not be
12775       needed.
12776     * Transition of basic block profiling to tree level implementation
12777       has been completed. The new implementation should be considerably
12778       more reliable (hopefully avoiding profile mismatch errors when
12779       using -fprofile-use or -fbranch-probabilities) and can be used to
12780       drive higher level optimizations, such as inlining.
12781       The -ftree-based-profiling command-line option was removed and
12782       -fprofile-use now implies disabling old RTL level loop optimizer
12783       (-fno-loop-optimize). Speculative prefetching optimization
12784       (originally enabled by -fspeculative-prefetching) was removed.
12785
12786New Languages and Language specific improvements
12787
12788  C and Objective-C
12789
12790     * The old Bison-based C and Objective-C parser has been replaced by a
12791       new, faster hand-written recursive-descent parser.
12792
12793  Ada
12794
12795     * The build infrastructure for the Ada runtime library and tools has
12796       been changed to be better integrated with the rest of the build
12797       infrastructure of GCC. This should make doing cross builds of Ada a
12798       bit easier.
12799
12800  C++
12801
12802     * ARM-style name-injection of friend declarations is no longer the
12803       default. For example:
12804          struct S {
12805            friend void f();
12806          };
12807
12808          void g() { f(); }
12809       will not be accepted; instead a declaration of f will need to be
12810       present outside of the scope of S. The new -ffriend-injection
12811       option will enable the old behavior.
12812     * The (undocumented) extension which permitted templates with default
12813       arguments to be bound to template template parameters with fewer
12814       parameters has been deprecated, and will be removed in the next
12815       major release of G++. For example:
12816       template <template <typename> class C>
12817       void f(C<double>) {}
12818
12819       template <typename T, typename U = int>
12820       struct S {};
12821
12822       template void f(S<double>);
12823
12824       makes use of the deprecated extension. The reason this code is not
12825       valid ISO C++ is that S is a template with two parameters;
12826       therefore, it cannot be bound to C which has only one parameter.
12827
12828    Runtime Library (libstdc++)
12829
12830     * Optimization work:
12831          + A new implementation of std::search_n is provided, better
12832            performing in case of random access iterators.
12833          + Added further efficient specializations of istream functions,
12834            i.e., character array and string extractors.
12835          + Other smaller improvements throughout.
12836     * Policy-based associative containers, designed for high-performance,
12837       flexibility and semantic safety are delivered in ext/pb_assoc.
12838     * A versatile string class, __gnu_cxx::__versa_string, providing
12839       facilities conforming to the standard requirements for
12840       basic_string, is delivered in <ext/vstring.h>. In particular:
12841          + Two base classes are provided: the default one avoids
12842            reference counting and is optimized for short strings; the
12843            alternate one, still uses it while improving in a few low
12844            level areas (e.g., alignment). See vstring_fwd.h for some
12845            useful typedefs.
12846          + Various algorithms have been rewritten (e.g., replace), the
12847            code streamlined and simple optimizations added.
12848          + Option 3 of DR 431 is implemented for both available bases,
12849            thus improving the support for stateful allocators.
12850     * As usual, many bugs have been fixed (e.g., libstdc++/13583,
12851       libstdc++/23953) and LWG resolutions put into effect for the first
12852       time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1
12853       Issue 6.19). The implementation status of TR1 is now tracked in the
12854       docs in tr1.html.
12855
12856  Objective-C++
12857
12858     * A new language front end for Objective-C++ has been added. This
12859       language allows users to mix the object oriented features of
12860       Objective-C with those of C++.
12861
12862  Java (GCJ)
12863
12864     * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19
12865       features (plus some 0.20 bug-fixes)
12866          + Networking
12867               o The java.net.HttpURLConnection implementation no longer
12868                 buffers the entire response body in memory. This means
12869                 that response bodies larger than available memory can now
12870                 be handled.
12871          + (N)IO
12872               o NIO FileChannel.map implementation, fast bulk put
12873                 implementation for DirectByteBuffer (speeds up this
12874                 method 10x).
12875               o FileChannel.lock() and FileChannel.force() implemented.
12876          + XML
12877               o gnu.xml fix for nodes created outside a namespace
12878                 context.
12879               o Add support for output indenting and
12880                 cdata-section-elements output instruction in
12881                 xml.transform.
12882               o xml.xpath corrections for cases where elements/attributes
12883                 might have been created in non-namespace-aware mode.
12884                 Corrections to handling of XSL variables and minor
12885                 conformance updates.
12886          + AWT
12887               o GNU JAWT implementation, the AWT Native Interface, which
12888                 allows direct access to native screen resources from
12889                 within a Canvas's paint method. GNU Classpath Examples
12890                 comes with a Demo, see libjava/classpath/examples/README.
12891               o awt.datatransfer updated to 1.5 with support for
12892                 FlavorEvents. The gtk+ awt peers now allow copy/paste of
12893                 text, images, URIs/files and serialized objects with
12894                 other applications and tracking clipboard change events
12895                 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized
12896                 objects are supported). A GNU Classpath Examples
12897                 datatransfer Demo was added to show the new
12898                 functionality.
12899               o Split gtk+ awt peers event handling in two threads and
12900                 improve gdk lock handling (solves several awt lock ups).
12901               o Speed up awt Image loading.
12902               o Better gtk+ scrollbar peer implementation when using gtk+
12903                 >= 2.6.
12904               o Handle image loading errors correctly for gdkpixbuf and
12905                 MediaTracker.
12906               o Better handle GDK lock. Properly prefix gtkpeer native
12907                 functions (cp_gtk).
12908               o GdkGraphics2D has been updated to use Cairo 0.5.x or
12909                 higher.
12910               o BufferedImage and GtkImage rewrites. All image drawing
12911                 operations should now work correctly (flipping requires
12912                 gtk+ >= 2.6)
12913               o When gtk+ 2.6 or higher is installed the default log
12914                 handler will produce stack traces whenever a WARNING,
12915                 CRITICAL or ERROR message is produced.
12916          + Free Swing
12917               o The RepaintManager has been reworked for more efficient
12918                 painting, especially for large GUIs.
12919               o The layout manager OverlayLayout has been implemented,
12920                 the BoxLayout has been rewritten to make use of the
12921                 SizeRequirements utility class and caching for more
12922                 efficient layout.
12923               o Improved accessibility support.
12924               o Significant progress has been made in the implementation
12925                 of the javax.swing.plaf.metal package, with most UI
12926                 delegates in a working state now. Please test this with
12927                 your own applications and provide feedback that will help
12928                 us to improve this package.
12929               o The GUI demo (gnu.classpath.examples.swing.Demo) has been
12930                 extended to highlight various features in our Free Swing
12931                 implementation. And it includes a look and feel switcher
12932                 for Metal (default), Ocean and GNU themes.
12933               o The javax.swing.plaf.multi package is now implemented.
12934               o Editing and several key actions for JTree and JTable were
12935                 implemented.
12936               o Lots of icons and look and feel improvements for Free
12937                 Swing basic and metal themes were added. Try running the
12938                 GNU Classpath Swing Demo in examples
12939                 (gnu.classpath.examples.swing.Demo) with:
12940                 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee
12941                 l or
12942                 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee
12943                 l
12944               o Start of styled text capabilites for java.swing.text.
12945               o DefaultMutableTreeNode pre-order, post-order, depth-first
12946                 and breadth-first traversal enumerations implemented.
12947               o JInternalFrame colors and titlebar draw properly.
12948               o JTree is working up to par (icons, selection and keyboard
12949                 traversal).
12950               o JMenus were made more compatible in visual and
12951                 programmatic behavior.
12952               o JTable changeSelection and multiple selections
12953                 implemented.
12954               o JButton and JToggleButton change states work properly
12955                 now.
12956               o JFileChooser fixes.
12957               o revalidate() and repaint() fixes which make Free Swing
12958                 much more responsive.
12959               o MetalIconFactory implemented.
12960               o Free Swing Top-Level Compatibility. JFrame, JDialog,
12961                 JApplet, JInternalFrame, and JWindow are now 1.5
12962                 compatible in the sense that you can call add() and
12963                 setLayout() directly on them, which will have the same
12964                 effect as calling getContentPane().add() and
12965                 getContentPane().setLayout().
12966               o The JTree interface has been completed. JTrees now
12967                 recognizes mouse clicks and selections work.
12968               o BoxLayout works properly now.
12969               o Fixed GrayFilter to actually work.
12970               o Metal SplitPane implemented.
12971               o Lots of Free Swing text and editor stuff work now.
12972          + Free RMI and Corba
12973               o Andrew Watson, Vice President and Technical Director of
12974                 the Object Management Group, has officially assigned us
12975                 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will
12976                 mark remote classpath-specific system exceptions.
12977                 Obtaining the VMCID means that GNU Classpath now is a
12978                 recogniseable type of node in a highly interoperable
12979                 CORBA world.
12980               o GNU Classpath now includes the first working draft to
12981                 support the RMI over IIOP protocol. The current
12982                 implementation is capable of remote invocations,
12983                 transferring various Serializables and Externalizables
12984                 via RMI-IIOP protocol. It can flatten graphs and, at
12985                 least for the simple cases, is interoperable with 1.5
12986                 JDKs.
12987               o org.omg.PortableInterceptor and related functionality in
12988                 other packages is now implemented:
12989                    # The sever and client interceptors work as required
12990                      since 1.4.
12991                    # The IOR interceptor works as needed for 1.5.
12992               o The org.omg.DynamicAny package is completed and passes
12993                 the prepared tests.
12994               o The Portable Object Adapter should now support the output
12995                 of the recent IDL to java compilers. These compilers now
12996                 generate servants and not CORBA objects as before, making
12997                 the output depend on the existing POA implementation.
12998                 Completing POA means that such code can already be tried
12999                 to run on Classpath. Our POA is tested for the following
13000                 usager scenarios:
13001                    # POA converts servant to the CORBA object.
13002                    # Servant provides to the CORBA object.
13003                    # POA activates new CORBA object with the given Object
13004                      Id (byte array) that is later accessible for the
13005                      servant.
13006                    # During the first call, the ServantActivator provides
13007                      servant for this and all subsequent calls on the
13008                      current object.
13009                    # During each call, the ServantLocator provides
13010                      servant for this call only.
13011                    # ServantLocator or ServantActivator forwards call to
13012                      another server.
13013                    # POA has a single servant, responsible for all
13014                      objects.
13015                    # POA has a default servant, but some objects are
13016                      explicitly connected to they specific servants.
13017                 The POA is verified using tests from the former
13018                 cost.omg.org.
13019               o The CORBA implementation is now a working prototype that
13020                 should support features up to 1.3 inclusive. We invite
13021                 groups writing CORBA dependent applications to try
13022                 Classpath implementation, reporting any possible bugs.
13023                 The CORBA prototype is interoperable with Sun's
13024                 implementation v 1.4, transferring object references,
13025                 primitive types, narrow and wide strings, arrays,
13026                 structures, trees, abstract interfaces and value types
13027                 (feature of CORBA 2.3) between these two platforms.
13028                 Remote exceptions are transferred and handled correctly.
13029                 The stringified object references (IORs) from various
13030                 sources are parsed as required. The transient (for
13031                 current session) and permanent (till jre restart)
13032                 redirections work. Both Little and Big Endian encoded
13033                 messages are accepted. The implementation is verified
13034                 using tests from the former cost.omg.org. The current
13035                 release includes working examples (see the examples
13036                 directory), demonstrating the client-server
13037                 communication, using either CORBA Request or IDL-based
13038                 stub (usually generated by a IDL to java compiler). These
13039                 examples also show how to use the Classpath CORBA naming
13040                 service. The IDL to java compiler is not yet written, but
13041                 as our library must be compatible, it naturally accepts
13042                 the output of other idlj implementations.
13043          + Misc
13044               o Updated TimeZone data against Olson tzdata2005l.
13045               o Make zip and jar packages UTF-8 clean.
13046               o "native" code builds and compiles (warning free) on
13047                 Darwin and Solaris.
13048               o java.util.logging.FileHandler now rotates files.
13049               o Start of a generic JDWP framework in gnu/classpath/jdwp.
13050                 This is unfinished, but feedback (at classpath@gnu.org)
13051                 from runtime hackers is greatly appreciated. Although
13052                 most of the work is currently being done around gcj/gij
13053                 we want this framework to be as VM neutral as possible.
13054                 Early design is described in:
13055                 [2]https://gcc.gnu.org/ml/java/2005-05/msg00260.html
13056               o QT4 AWT peers, enable by giving configure
13057                 --enable-qt-peer. Included, but not ready for production
13058                 yet. They are explicitly disabled and not supported. But
13059                 if you want to help with the development of these new
13060                 features we are interested in feedback. You will have to
13061                 explicitly enable them to try them out (and they will
13062                 most likely contain bugs).
13063               o Documentation fixes all over the place. See
13064                 [3]https://developer.classpath.org/doc/
13065
13066New Targets and Target Specific Improvements
13067
13068  IA-32/x86-64
13069
13070     * The x86-64 medium model (that allows building applications whose
13071       data segment exceeds 4GB) was redesigned to match latest ABI draft.
13072       New implementation split large datastructures into separate segment
13073       improving performance of accesses to small datastructures and also
13074       allows linking of small model libraries into medium model programs
13075       as long as the libraries are not accessing the large datastructures
13076       directly. Medium model is also supported in position independent
13077       code now.
13078       The ABI change results in partial incompatibility among medium
13079       model objects. Linking medium model libraries (or objects) compiled
13080       with new compiler into medium model program compiled with older
13081       will likely result in exceeding ranges of relocations.
13082       Binutils 2.16.91 or newer are required for compiling medium model
13083       now.
13084
13085  RS6000 (POWER/PowerPC)
13086
13087     * The AltiVec vector primitives in <altivec.h> are now implemented in
13088       a way that puts a smaller burden on the preprocessor, instead
13089       processing the "overloading" in the front ends. This should benefit
13090       compilation speed on AltiVec vector code.
13091     * AltiVec initializers now are generated more efficiently.
13092     * The popcountb instruction available on POWER5 now is generated.
13093     * The floating point round to integer instructions available on
13094       POWER5+ now is generated.
13095     * Floating point divides can be synthesized using the floating point
13096       reciprocal estimate instructions.
13097     * Double precision floating point constants are initialized as single
13098       precision values if they can be represented exactly.
13099
13100  S/390, zSeries and System z9
13101
13102     * Support for the IBM System z9 109 processor has been added. When
13103       using the -march=z9-109 option, the compiler will generate code
13104       making use of instructions provided by the extended immediate
13105       facility.
13106     * Support for 128-bit IEEE floating point has been added. When using
13107       the -mlong-double-128 option, the compiler will map the long double
13108       data type to 128-bit IEEE floating point. Using this option
13109       constitutes an ABI change, and requires glibc support.
13110     * Various changes to improve performance of generated code have been
13111       implemented, including:
13112          + In functions that do not require a literal pool, register %r13
13113            (which is traditionally reserved as literal pool pointer), can
13114            now be freely used for other purposes by the compiler.
13115          + More precise tracking of register use allows the compiler to
13116            generate more efficient function prolog and epilog code in
13117            certain cases.
13118          + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING
13119            instructions are now used to implement C string functions.
13120          + The MOVE CHARACTER instruction with single byte overlap is now
13121            used to implement the memset function with non-zero fill byte.
13122          + The LOAD ZERO instructions are now used where appropriate.
13123          + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK,
13124            and INSERT IMMEDIATE instructions are now used more frequently
13125            to optimize bitfield operations.
13126          + The BRANCH ON COUNT instruction is now used more frequently.
13127            In particular, the fact that a loop contains a subroutine call
13128            no longer prevents the compiler from using this instruction.
13129          + The compiler is now aware that all shift and rotate
13130            instructions implicitly truncate the shift count to six bits.
13131     * Back-end support for the following generic features has been
13132       implemented:
13133          + The full set of [4]built-in functions for atomic memory
13134            access.
13135          + The -fstack-protector feature.
13136          + The optimization pass avoiding unnecessary stores of incoming
13137            argument registers in functions with variable argument list.
13138
13139  SPARC
13140
13141     * The default code model in 64-bit mode has been changed from
13142       Medium/Anywhere to Medium/Middle on Solaris.
13143     * TLS support is disabled by default on Solaris prior to release 10.
13144       It can be enabled on TLS-capable Solaris 9 versions (4/04 release
13145       and later) by specifying --enable-tls at configure time.
13146
13147  MorphoSys
13148
13149     * Support has been added for this new architecture.
13150
13151Obsolete Systems
13152
13153Documentation improvements
13154
13155Other significant improvements
13156
13157     * GCC can now emit code for protecting applications from
13158       stack-smashing attacks. The protection is realized by buffer
13159       overflow detection and reordering of stack variables to avoid
13160       pointer corruption.
13161     * Some built-in functions have been fortified to protect them against
13162       various buffer overflow (and format string) vulnerabilities.
13163       Compared to the mudflap bounds checking feature, the safe builtins
13164       have far smaller overhead. This means that programs built using
13165       safe builtins should not experience any measurable slowdown.
13166
13167GCC 4.1.2
13168
13169   This is the [5]list of problem reports (PRs) from GCC's bug tracking
13170   system that are known to be fixed in the 4.1.2 release. This list might
13171   not be complete (that is, it is possible that some PRs that have been
13172   fixed are not listed here).
13173
13174   When generating code for a shared library, GCC now recognizes that
13175   global functions may be replaced when the program runs. Therefore, it
13176   is now more conservative in deducing information from the bodies of
13177   functions. For example, in this example:
13178    void f() {}
13179    void g() {
13180     try { f(); }
13181     catch (...) {
13182       cout << "Exception";
13183     }
13184    }
13185
13186   G++ would previously have optimized away the catch clause, since it
13187   would have concluded that f cannot throw exceptions. Because users may
13188   replace f with another function in the main body of the program, this
13189   optimization is unsafe, and is no longer performed. If you wish G++ to
13190   continue to optimize as before, you must add a throw() clause to the
13191   declaration of f to make clear that it does not throw exceptions.
13192
13193
13194    For questions related to the use of GCC, please consult these web
13195    pages and the [6]GCC manuals. If that fails, the
13196    [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13197    web pages and the development of GCC are welcome on our developer
13198    list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives.
13199
13200   Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and
13201   distribution of this entire article is permitted in any medium,
13202   provided this notice is preserved.
13203
13204   These pages are [11]maintained by the GCC team. Last modified
13205   2022-10-26.
13206
13207References
13208
13209   1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
13210   2. https://gcc.gnu.org/ml/java/2005-05/msg00260.html
13211   3. https://developer.classpath.org/doc/
13212   4. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
13213   5. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2
13214   6. https://gcc.gnu.org/onlinedocs/
13215   7. mailto:gcc-help@gcc.gnu.org
13216   8. mailto:gcc@gcc.gnu.org
13217   9. https://gcc.gnu.org/lists.html
13218  10. https://www.fsf.org/
13219  11. https://gcc.gnu.org/about.html
13220======================================================================
13221http://gcc.gnu.org/gcc-4.0/index.html
13222
13223                             GCC 4.0 Release Series
13224
13225   (This release series is no longer supported.)
13226
13227   January 31, 2007
13228
13229   The [1]GNU project and the GCC developers are pleased to announce the
13230   release of GCC 4.0.4.
13231
13232   This release is a bug-fix release, containing fixes for regressions in
13233   GCC 4.0.3 relative to previous releases of GCC.
13234
13235Release History
13236
13237   GCC 4.0.4
13238          January 31, 2007 ([2]changes)
13239
13240   GCC 4.0.3
13241          March 10, 2006 ([3]changes)
13242
13243   GCC 4.0.2
13244          September 28, 2005 ([4]changes)
13245
13246   GCC 4.0.1
13247          July 7, 2005 ([5]changes)
13248
13249   GCC 4.0.0
13250          April 20, 2005 ([6]changes)
13251
13252References and Acknowledgements
13253
13254   GCC used to stand for the GNU C Compiler, but since the compiler
13255   supports several other languages aside from C, it now stands for the
13256   GNU Compiler Collection.
13257
13258   A list of [7]successful builds is updated as new information becomes
13259   available.
13260
13261   The GCC developers would like to thank the numerous people that have
13262   contributed new features, improvements, bug fixes, and other changes as
13263   well as test results to GCC. This [8]amazing group of volunteers is
13264   what makes GCC successful.
13265
13266   For additional information about GCC please refer to the [9]GCC project
13267   web site or contact the [10]GCC development mailing list.
13268
13269   To obtain GCC please use [11]our mirror sites, or [12]our version
13270   control system.
13271
13272
13273    For questions related to the use of GCC, please consult these web
13274    pages and the [13]GCC manuals. If that fails, the
13275    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13276    web pages and the development of GCC are welcome on our developer
13277    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
13278    archives.
13279
13280   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
13281   distribution of this entire article is permitted in any medium,
13282   provided this notice is preserved.
13283
13284   These pages are [18]maintained by the GCC team. Last modified
13285   2022-10-26.
13286
13287References
13288
13289   1. http://www.gnu.org/
13290   2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
13291   3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3
13292   4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2
13293   5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1
13294   6. http://gcc.gnu.org/gcc-4.0/changes.html
13295   7. http://gcc.gnu.org/gcc-4.0/buildstat.html
13296   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13297   9. http://gcc.gnu.org/index.html
13298  10. mailto:gcc@gcc.gnu.org
13299  11. http://gcc.gnu.org/mirrors.html
13300  12. http://gcc.gnu.org/git.html
13301  13. https://gcc.gnu.org/onlinedocs/
13302  14. mailto:gcc-help@gcc.gnu.org
13303  15. mailto:gcc@gcc.gnu.org
13304  16. https://gcc.gnu.org/lists.html
13305  17. https://www.fsf.org/
13306  18. https://gcc.gnu.org/about.html
13307======================================================================
13308http://gcc.gnu.org/gcc-4.0/changes.html
13309
13310                             GCC 4.0 Release Series
13311                        Changes, New Features, and Fixes
13312
13313   The latest release in the 4.0 release series is [1]GCC 4.0.4.
13314
13315Caveats
13316
13317     * GCC now generates location lists by default when compiling with
13318       debug info and optimization.
13319          + GDB 6.0 and older crashes when it sees location lists. GDB 6.1
13320            or later is needed to debug binaries containing location
13321            lists.
13322          + When you are trying to view a value of a variable in a part of
13323            a function where it has no location (for example when the
13324            variable is no longer used and thus its location was used for
13325            something else) GDB will say that it is not available.
13326       You can disable generating location lists by -fno-var-tracking.
13327     * GCC no longer accepts the -fwritable-strings option. Use named
13328       character arrays when you need a writable string.
13329     * The options -freduce-all-givs and -fmove-all-movables have been
13330       discontinued. They were used to circumvent a shortcoming in the
13331       heuristics of the old loop optimization code with respect to common
13332       Fortran constructs. The new (tree) loop optimizer works differently
13333       and doesn't need those work-arounds.
13334     * The graph-coloring register allocator, formerly enabled by the
13335       option -fnew-ra, has been discontinued.
13336     * -I- has been deprecated. -iquote is meant to replace the need for
13337       this option.
13338     * The MIPS -membedded-pic and -mrnames options have been removed.
13339     * All MIPS targets now require the GNU assembler. In particular, IRIX
13340       configurations can no longer use the MIPSpro assemblers, although
13341       they do still support the MIPSpro linkers.
13342     * The SPARC option -mflat has been removed.
13343     * English-language diagnostic messages will now use Unicode quotation
13344       marks in UTF-8 locales. (Non-English messages already used the
13345       quotes appropriate for the language in previous releases.) If your
13346       terminal does not support UTF-8 but you are using a UTF-8 locale
13347       (such locales are the default on many GNU/Linux systems) then you
13348       should set LC_CTYPE=C in the environment to disable that locale.
13349       Programs that parse diagnostics and expect plain ASCII
13350       English-language messages should set LC_ALL=C. See [2]Markus Kuhn's
13351       explanation of Unicode quotation marks for more information.
13352     * The specs file is no longer installed on most platforms. Most users
13353       will be totally unaffected. However, if you are accustomed to
13354       editing the specs file yourself, you will now have to use the
13355       -dumpspecs option to generate the specs file, and then edit the
13356       resulting file.
13357
13358General Optimizer Improvements
13359
13360     * The [3]tree ssa branch has been merged. This merge has brought in a
13361       completely new optimization framework based on a higher level
13362       intermediate representation than the existing RTL representation.
13363       Numerous new code transformations based on the new framework are
13364       available in GCC 4.0, including:
13365          + Scalar replacement of aggregates
13366          + Constant propagation
13367          + Value range propagation
13368          + Partial redundancy elimination
13369          + Load and store motion
13370          + Strength reduction
13371          + Dead store elimination
13372          + Dead and unreachable code elimination
13373          + [4]Autovectorization
13374          + Loop interchange
13375          + Tail recursion by accumulation
13376       Many of these passes outperform their counterparts from previous
13377       GCC releases.
13378     * [5]Swing Modulo Scheduling (SMS). An RTL level instruction
13379       scheduling optimization intended for loops that perform heavy
13380       computations.
13381
13382New Languages and Language specific improvements
13383
13384  C family
13385
13386     * The sentinel attribute has been added to GCC. This function
13387       attribute allows GCC to warn when variadic functions such as execl
13388       are not NULL terminated. See the GCC manual for a complete
13389       description of its behavior.
13390     * Given __attribute__((alias("target"))) it is now an error if target
13391       is not a symbol, defined in the same translation unit. This also
13392       applies to aliases created by #pragma weak alias=target. This is
13393       because it's meaningless to define an alias to an undefined symbol.
13394       On Solaris, the native assembler would have caught this error, but
13395       GNU as does not.
13396
13397  C and Objective-C
13398
13399     * The -Wstrict-aliasing=2 option has been added. This warning catches
13400       all unsafe cases, but it may also give a warning for some cases
13401       that are safe.
13402     * The cast-as-lvalue, conditional-expression-as-lvalue and
13403       compound-expression-as-lvalue extensions, which were deprecated in
13404       3.3.4 and 3.4, have been removed.
13405     * The -fwritable-strings option, which was deprecated in 3.4, has
13406       been removed.
13407     * #pragma pack() semantics have been brought closer to those used by
13408       other compilers. This also applies to C++.
13409     * Taking the address of a variable with register storage is invalid
13410       in C. GCC now issues an error instead of a warning.
13411     * Arrays of incomplete element type are invalid in C. GCC now issues
13412       an error for such arrays. Declarations such as extern struct s x[];
13413       (where struct s has not been defined) can be moved after the
13414       definition of struct s. Function parameters declared as arrays of
13415       incomplete type can instead be declared as pointers.
13416
13417  C++
13418
13419     * When compiling without optimizations (-O0), the C++ front end is
13420       much faster than in any previous versions of GCC. Independent
13421       testers have measured speed-ups up to 25% in real-world production
13422       code, compared to the 3.4 family (which was already the fastest
13423       version to date). Upgrading from older versions might show even
13424       bigger improvements.
13425     * ELF visibility attributes can now be applied to a class type, so
13426       that it affects every member function of a class at once, without
13427       having to specify each individually:
13428class __attribute__ ((visibility("hidden"))) Foo
13429{
13430   int foo1();
13431   void foo2();
13432};
13433       The syntax is deliberately similar to the __declspec() system used
13434       by Microsoft Windows based compilers, allowing cross-platform
13435       projects to easily reuse their existing macro system for denoting
13436       exports and imports. By explicitly marking internal classes never
13437       used outside a binary as hidden, one can completely avoid PLT
13438       indirection overheads during their usage by the compiler. You can
13439       find out more about the advantages of this at
13440       [6]https://www.akkadia.org/drepper/dsohowto.pdf
13441     * The -fvisibility-inlines-hidden option has been added which marks
13442       all inlineable functions as having hidden ELF visibility, thus
13443       removing their symbol and typeinfo from the exported symbol table
13444       of the output ELF binary. Using this option can reduce the exported
13445       symbol count of template-heavy code by up to 40% with no code
13446       change at all, thus notably improving link and load times for the
13447       binary as well as a reduction in size of up to 10%. Also, check the
13448       new [7]-fvisibility option.
13449     * The compiler now uses the library interface specified by the [8]C++
13450       ABI for thread-safe initialization of function-scope static
13451       variables. Most users should leave this alone, but embedded
13452       programmers may want to disable this by specifying
13453       -fno-threadsafe-statics for a small savings in code size.
13454     * Taking the address of an explicit register variable is no longer
13455       supported. Note that C++ allows taking the address of variables
13456       with register storage so this will continue to compile with a
13457       warning. For example, assuming that r0 is a machine register:
13458register int foo asm ("r0");
13459register int bar;
13460&foo; // error, no longer accepted
13461&bar; // OK, with a warning
13462     * G++ has an undocumented extension to virtual function covariancy
13463       rules that allowed the overrider to return a type that was
13464       implicitly convertable to the overridden function's return type.
13465       For instance a function returning void * could be overridden by a
13466       function returning T *. This is now deprecated and will be removed
13467       in a future release.
13468     * The G++ minimum and maximum operators (<? and >?) and their
13469       compound forms (<?=) and >?=) have been deprecated and will be
13470       removed in a future version. Code using these operators should be
13471       modified to use std::min and std::max instead.
13472     * Declaration of nested classes of class templates as friends are
13473       supported:
13474template <typename T> struct A {
13475  class B {};
13476};
13477class C {
13478  template <typename T> friend class A<T>::B;
13479};
13480       This complements the feature member functions of class templates as
13481       friends introduced in GCC 3.4.0.
13482     * When declaring a friend class using an unqualified name, classes
13483       outside the innermost non-class scope are not searched:
13484class A;
13485namespace N {
13486  class B {
13487    friend class A;   // Refer to N::A which has not been declared yet
13488                      // because name outside namespace N are not searched
13489    friend class ::A; // Refer to ::A
13490  };
13491}
13492       Hiding the friend name until declaration is still not implemented.
13493     * Friends of classes defined outside their namespace are correctly
13494       handled:
13495namespace N {
13496  class A;
13497}
13498class N::A {
13499  friend class B; // Refer to N::B in GCC 4.0.0
13500                  // but ::B in earlier versions of GCC
13501};
13502
13503    Runtime Library (libstdc++)
13504
13505     * Optimization work:
13506          + Added efficient specializations of istream functions for char
13507            and wchar_t.
13508          + Further performance tuning of strings, in particular wrt
13509            single-char append and getline.
13510          + iter_swap - and therefore most of the mutating algorithms -
13511            now makes an unqualified call to swap when the value_type of
13512            the two iterators is the same.
13513     * A large subset of the features in Technical Report 1 (TR1 for
13514       short) is experimentally delivered (i.e., no guarantees about the
13515       implementation are provided. In particular it is not promised that
13516       the library will remain link-compatible when code using TR1 is
13517       used):
13518          + General utilities such as reference_wrapper and shared_ptr.
13519          + Function objects, i.e., result_of, mem_fn, bind, function.
13520          + Support for metaprogramming.
13521          + New containers such as tuple, array, unordered_set,
13522            unordered_map, unordered_multiset, unordered_multimap.
13523     * As usual, many bugs have been fixed and LWG resolutions implemented
13524       for the first time (e.g., DR 409).
13525
13526  Java
13527
13528     * In order to prevent naming conflicts with other implementations of
13529       these tools, some GCJ binaries have been renamed:
13530          + rmic is now grmic,
13531          + rmiregistry is now grmiregistry, and
13532          + jar is now fastjar.
13533       In particular, these names were problematic for the jpackage.org
13534       packaging conventions which install symlinks in /usr/bin that point
13535       to the preferred versions of these tools.
13536     * The -findirect-dispatch argument to the compiler now works and
13537       generates code following a new "binary compatibility" ABI. Code
13538       compiled this way follows the binary compatibility rules of the
13539       Java Language Specification.
13540     * libgcj now has support for using GCJ as a JIT, using the
13541       gnu.gcj.jit family of system properties.
13542     * libgcj can now find a shared library corresponding to the bytecode
13543       representation of a class. See the documentation for the new
13544       gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system
13545       property.
13546     * There have been many improvements to the class library. Here are
13547       some highlights:
13548          + Much more of AWT and Swing exist.
13549          + Many new packages and classes were added, including
13550            java.util.regex, java.net.URI, javax.crypto,
13551            javax.crypto.interfaces, javax.crypto.spec, javax.net,
13552            javax.net.ssl, javax.security.auth,
13553            javax.security.auth.callback, javax.security.auth.login,
13554            javax.security.auth.x500, javax.security.sasl, org.ietf.jgss,
13555            javax.imageio, javax.imageio.event, javax.imageio.spi,
13556            javax.print, javax.print.attribute,
13557            javax.print.attribute.standard, javax.print.event, and
13558            javax.xml
13559          + Updated SAX and DOM, and imported GNU JAXP
13560
13561  Fortran
13562
13563     * A new [9]Fortran front end has replaced the aging GNU Fortran 77
13564       front end. The new front end supports Fortran 90 and Fortran 95. It
13565       may not yet be as stable as the old Fortran front end.
13566
13567  Ada
13568
13569     * Ada (with tasking and Zero Cost Exceptions) is now available on
13570       many more targets, including but not limited to: alpha-linux,
13571       hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux,
13572       s390x-linux, sparc-linux.
13573     * Some of the new Ada 2005 features are now implemented like
13574       Wide_Wide_Character and Ada.Containers.
13575     * Many bugs have been fixed, tools and documentation improved.
13576     * To compile Ada from the sources, install an older working Ada
13577       compiler and then use --enable-languages=ada at configuration time,
13578       since the Ada front end is not currently activated by default. See
13579       the [10]Installing GCC for details.
13580
13581New Targets and Target Specific Improvements
13582
13583  H8/300
13584
13585     * The frame layout has changed. In the new layout, the prologue of a
13586       function first saves registers and then allocate space for locals,
13587       resulting in an 1% improvement on code size.
13588
13589  IA-32/x86-64 (AMD64)
13590
13591     * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10,
13592       log1p, log2, logb and tan mathematical builtins (and their float
13593       and long double variants) are now implemented as inline x87
13594       intrinsics when using -ffast-math.
13595     * The ceil, floor, nearbyint, rint and trunc mathematical builtins
13596       (and their float and long double variants) are now implemented as
13597       inline x87 intrinsics when using -ffast-math.
13598     * The x87's fsincos instruction is now used automatically with
13599       -ffast-math when calculating both the sin and cos of the same
13600       argument.
13601     * Instruction selection for multiplication and division by constants
13602       has been improved.
13603
13604  IA-64
13605
13606     * Floating point division, integer division and sqrt are now inlined,
13607       resulting in significant performance improvements on some codes.
13608
13609  MIPS
13610
13611     * Division by zero checks now use conditional traps if the target
13612       processor supports them. This decreases code size by one word per
13613       division operation. The old behavior (branch and break) can be
13614       obtained either at configure time by passing --with-divide=breaks
13615       to configure or at runtime by passing -mdivide-breaks to GCC.
13616     * Support for MIPS64 paired-single instructions has been added. It is
13617       enabled by -mpaired-single and can be accessed using both the
13618       target-independent vector extensions and new MIPS-specific built-in
13619       functions.
13620     * Support for the MIPS-3D ASE has been added. It is enabled by
13621       -mips3d and provides new MIPS-3D-specific built-in functions.
13622     * The -mexplicit-relocs option now supports static n64 code (as is
13623       used, for example, in 64-bit linux kernels). -mexplicit-relocs
13624       should now be feature-complete and is enabled by default when GCC
13625       is configured to use a compatible assembler.
13626     * Support for the NEC VR4130 series has been added. This support
13627       includes the use of VR-specific instructions and a new VR4130
13628       scheduler. Full VR4130 support can be selected with -march=vr4130
13629       while code for any ISA can be tuned for the VR4130 using
13630       -mtune=vr4130. There is also a new -mvr4130-align option that
13631       produces better schedules at the cost of increased code size.
13632     * Support for the Broadcom SB-1 has been extended. There is now an
13633       SB-1 scheduler as well as support for the SB-1-specific
13634       paired-single instructions. Full SB-1 support can be selected with
13635       -march=sb1 while code for any ISA can be optimized for the SB-1
13636       using -mtune=sb1.
13637     * The compiler can now work around errata in R4000, R4400, VR4120 and
13638       VR4130 processors. These workarounds are enabled by -mfix-r4000,
13639       -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120
13640       and VR4130 workarounds need binutils 2.16 or above.
13641     * IRIX shared libraries are now installed into the standard library
13642       directories: o32 libraries go into lib/, n32 libraries go into
13643       lib32/ and n64 libraries go into lib64/.
13644     * The compiler supports a new -msym32 option. It can be used to
13645       optimize n64 code in which all symbols are known to have 32-bit
13646       values.
13647
13648  S/390 and zSeries
13649
13650     * New command-line options help to generate code intended to run in
13651       an environment where stack space is restricted, e.g. Linux kernel
13652       code:
13653          + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time
13654            warnings for single functions that require large or dynamic
13655            stack frames.
13656          + -mstack-size and -mstack-guard generate code that checks for
13657            stack overflow at run time.
13658          + -mpacked-stack generates code that reduces the stack frame
13659            size of many functions by reusing unneeded parts of the stack
13660            bias area.
13661     * The -msoft-float option now ensures that generated code never
13662       accesses floating point registers.
13663     * The s390x-ibm-tpf target now fully supports C++, including
13664       exceptions and threads.
13665     * Various changes to improve performance of the generated code have
13666       been implemented, including:
13667          + GCC now uses sibling calls where possible.
13668          + Condition code handling has been optimized, allowing GCC to
13669            omit redundant comparisons in certain cases.
13670          + The cost function guiding many optimizations has been refined
13671            to more accurately represent the z900 and z990 processors.
13672          + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW
13673            instructions are now used to avoid conditional branches in
13674            certain cases.
13675          + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to
13676            optimize address arithmetic required to access large stack
13677            frames.
13678          + GCC now makes more efficient use of memory-to-memory type
13679            instructions (MVC, CLC, ...).
13680          + More precise tracking of special register use allows better
13681            instruction scheduling, in particular of the function prologue
13682            and epilogue sequences.
13683          + The Java front end now generates inline code to implement
13684            integer division, instead of calling library routines.
13685
13686  SPARC
13687
13688     * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and
13689       -mf934 have been removed. They have been replaced with -mcpu=xxx.
13690     * The internal model used to estimate the relative cost of each
13691       instruction has been updated. It is expected to give better results
13692       on recent UltraSPARC processors.
13693     * Code generation for function prologues and epilogues has been
13694       improved, resulting in better scheduling and allowing multiple exit
13695       points in functions.
13696     * Support for Sun's Visual Instruction Set (VIS) has been enhanced.
13697       It is enabled by -mvis and provides new built-in functions for VIS
13698       instructions on UltraSPARC processors.
13699     * The option -mapp-regs has been turned on by default on Solaris too.
13700
13701  NetWare
13702
13703     * Novell NetWare (on ix86, no other hardware platform was ever really
13704       supported by this OS) has been re-enabled and the ABI supported by
13705       GCC has been brought into sync with that of MetroWerks CodeWarrior
13706       (the ABI previously supported was that of some Unix systems, which
13707       NetWare never tried to support).
13708
13709Obsolete Systems
13710
13711   Support for a number of older systems has been declared obsolete in GCC
13712   4.0. Unless there is activity to revive them, the next release of GCC
13713   will have their sources permanently removed.
13714
13715   All GCC ports for the following processor architectures have been
13716   declared obsolete:
13717     * Intel i860
13718     * Ubicom IP2022
13719     * National Semiconductor NS32K (ns32k)
13720     * Texas Instruments TMS320C[34]x
13721
13722   Also, those for some individual systems have been obsoleted:
13723     * SPARC family
13724          + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf,
13725            sparc86x-*-elf)
13726          + OpenBSD 32-bit (sparc-*-openbsd*)
13727
13728Documentation improvements
13729
13730Other significant improvements
13731
13732     * Location lists are now generated by default when compiling with
13733       debug info and optimization. Location lists provide more accurate
13734       debug info about locations of variables and they allow debugging
13735       code compiled with -fomit-frame-pointer.
13736     * The -fvisibility option has been added which allows the default ELF
13737       visibility of all symbols to be set per compilation and the new
13738       #pragma GCC visibility preprocessor command allows the setting of
13739       default ELF visibility for a region of code. Using
13740       -fvisibility=hidden especially in combination with the new
13741       -fvisibility-inlines-hidden can yield substantial improvements in
13742       output binary quality including avoiding PLT indirection overheads,
13743       reduction of the exported symbol count by up to 60% (with resultant
13744       improvements to link and load times), better scope for the
13745       optimizer to improve code and up to a 20% reduction in binary size.
13746       Using these options correctly yields a binary with a similar symbol
13747       count to a Windows DLL.
13748       Perhaps more importantly, this new feature finally allows (with
13749       careful planning) complete avoidance of symbol clashes when
13750       manually loading shared objects with RTLD_GLOBAL, thus finally
13751       solving problems many projects such as python were forced to use
13752       RTLD_LOCAL for (with its resulting issues for C++ correctness). You
13753       can find more information about using these options at
13754       [11]https://gcc.gnu.org/wiki/Visibility.
13755     __________________________________________________________________
13756
13757GCC 4.0.1
13758
13759   This is the [12]list of problem reports (PRs) from GCC's bug tracking
13760   system that are known to be fixed in the 4.0.1 release. This list might
13761   not be complete (that is, it is possible that some PRs that have been
13762   fixed are not listed here).
13763
13764GCC 4.0.2
13765
13766   This is the [13]list of problem reports (PRs) from GCC's bug tracking
13767   system that are known to be fixed in the 4.0.2 release. This list might
13768   not be complete (that is, it is possible that some PRs that have been
13769   fixed are not listed here).
13770
13771   Unfortunately, due to a release engineering failure, this release has a
13772   regression on Solaris that will affect some C++ programs. We suggest
13773   that Solaris users apply a [14]patch that corrects the problem. Users
13774   who do not wish to apply the patch should explicitly link C++ programs
13775   with the -pthreads option, even if they do not use threads. This
13776   problem has been corrected in the current 4.0 branch sources and will
13777   not be present in GCC 4.0.3.
13778
13779GCC 4.0.3
13780
13781   Starting with this release, the function getcontext is recognized by
13782   the compiler as having the same semantics as the setjmp function. In
13783   particular, the compiler will ensure that all registers are dead before
13784   calling such a function and will emit a warning about the variables
13785   that may be clobbered after the second return from the function.
13786
13787GCC 4.0.4
13788
13789   This is the [15]list of problem reports (PRs) from GCC's bug tracking
13790   system that are known to be fixed in the 4.0.4 release. This list might
13791   not be complete (that is, it is possible that some PRs that have been
13792   fixed are not listed here).
13793
13794   The 4.0.4 release is provided for those that require a high degree of
13795   binary compatibility with previous 4.0.x releases. For most users, the
13796   GCC team recommends that version 4.1.1 or later be used instead."
13797
13798
13799    For questions related to the use of GCC, please consult these web
13800    pages and the [16]GCC manuals. If that fails, the
13801    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13802    web pages and the development of GCC are welcome on our developer
13803    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
13804    archives.
13805
13806   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
13807   distribution of this entire article is permitted in any medium,
13808   provided this notice is preserved.
13809
13810   These pages are [21]maintained by the GCC team. Last modified
13811   2022-10-26.
13812
13813References
13814
13815   1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
13816   2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
13817   3. http://gcc.gnu.org/projects/tree-ssa/
13818   4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html
13819   5. http://gcc.gnu.org/news/sms.html
13820   6. https://www.akkadia.org/drepper/dsohowto.pdf
13821   7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility
13822   8. https://itanium-cxx-abi.github.io/cxx-abi/
13823   9. http://gcc.gnu.org/fortran/
13824  10. https://gcc.gnu.org/install/
13825  11. https://gcc.gnu.org/wiki/Visibility
13826  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1
13827  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2
13828  14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html
13829  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4
13830  16. https://gcc.gnu.org/onlinedocs/
13831  17. mailto:gcc-help@gcc.gnu.org
13832  18. mailto:gcc@gcc.gnu.org
13833  19. https://gcc.gnu.org/lists.html
13834  20. https://www.fsf.org/
13835  21. https://gcc.gnu.org/about.html
13836======================================================================
13837http://gcc.gnu.org/gcc-3.4/index.html
13838
13839                             GCC 3.4 Release Series
13840
13841   (This release series is no longer supported.)
13842
13843   May 26, 2006
13844
13845   The [1]GNU project and the GCC developers are pleased to announce the
13846   release of GCC 3.4.6.
13847
13848   This release is a bug-fix release, containing fixes for regressions in
13849   GCC 3.4.4 relative to previous releases of GCC. This is the last of the
13850   3.4.x series.
13851
13852   The GCC 3.4 release series includes numerous [2]new features,
13853   improvements, bug fixes, and other changes, thanks to an [3]amazing
13854   group of volunteers.
13855
13856Release History
13857
13858   GCC 3.4.6
13859          March 6, 2006 ([4]changes)
13860
13861   GCC 3.4.5
13862          November 30, 2005 ([5]changes)
13863
13864   GCC 3.4.4
13865          May 18, 2005 ([6]changes)
13866
13867   GCC 3.4.3
13868          November 4, 2004 ([7]changes)
13869
13870   GCC 3.4.2
13871          September 6, 2004 ([8]changes)
13872
13873   GCC 3.4.1
13874          July 1, 2004 ([9]changes)
13875
13876   GCC 3.4.0
13877          April 18, 2004 ([10]changes)
13878
13879References and Acknowledgements
13880
13881   GCC used to stand for the GNU C Compiler, but since the compiler
13882   supports several other languages aside from C, it now stands for the
13883   GNU Compiler Collection.
13884
13885   A list of [11]successful builds is updated as new information becomes
13886   available.
13887
13888   The GCC developers would like to thank the numerous people that have
13889   contributed new features, improvements, bug fixes, and other changes as
13890   well as test results to GCC. This [12]amazing group of volunteers is
13891   what makes GCC successful.
13892
13893   For additional information about GCC please refer to the [13]GCC
13894   project web site or contact the [14]GCC development mailing list.
13895
13896   To obtain GCC please use [15]our mirror sites, or [16]our version
13897   control system.
13898
13899
13900    For questions related to the use of GCC, please consult these web
13901    pages and the [17]GCC manuals. If that fails, the
13902    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13903    web pages and the development of GCC are welcome on our developer
13904    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
13905    archives.
13906
13907   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
13908   distribution of this entire article is permitted in any medium,
13909   provided this notice is preserved.
13910
13911   These pages are [22]maintained by the GCC team. Last modified
13912   2022-10-26.
13913
13914References
13915
13916   1. http://www.gnu.org/
13917   2. http://gcc.gnu.org/gcc-3.4/changes.html
13918   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13919   4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
13920   5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5
13921   6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4
13922   7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3
13923   8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2
13924   9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1
13925  10. http://gcc.gnu.org/gcc-3.4/changes.html
13926  11. http://gcc.gnu.org/gcc-3.4/buildstat.html
13927  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13928  13. http://gcc.gnu.org/index.html
13929  14. mailto:gcc@gcc.gnu.org
13930  15. http://gcc.gnu.org/mirrors.html
13931  16. http://gcc.gnu.org/git.html
13932  17. https://gcc.gnu.org/onlinedocs/
13933  18. mailto:gcc-help@gcc.gnu.org
13934  19. mailto:gcc@gcc.gnu.org
13935  20. https://gcc.gnu.org/lists.html
13936  21. https://www.fsf.org/
13937  22. https://gcc.gnu.org/about.html
13938======================================================================
13939http://gcc.gnu.org/gcc-3.4/changes.html
13940
13941                             GCC 3.4 Release Series
13942                        Changes, New Features, and Fixes
13943
13944   The final release in the 3.4 release series is [1]GCC 3.4.6. The series
13945   is now closed.
13946
13947   GCC 3.4 has [2]many improvements in the C++ front end. Before reporting
13948   a bug, please make sure it's really GCC, and not your code, that is
13949   broken.
13950
13951Caveats
13952
13953     * GNU Make is now required to build GCC.
13954     * With -nostdinc the preprocessor used to ignore both standard
13955       include paths and include paths contained in environment variables.
13956       It was neither documented nor intended that environment variable
13957       paths be ignored, so this has been corrected.
13958     * GCC no longer accepts the options -fvolatile, -fvolatile-global and
13959       -fvolatile-static. It is unlikely that they worked correctly in any
13960       3.x release.
13961     * GCC no longer ships <varargs.h>. Use <stdarg.h> instead.
13962     * Support for all the systems [3]obsoleted in GCC 3.3 has been
13963       removed from GCC 3.4. See below for a [4]list of systems which are
13964       obsoleted in this release.
13965     * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C
13966       compilers will not work.
13967     * The implementation of the [5]MIPS ABIs has changed. As a result,
13968       the code generated for certain MIPS targets will not be binary
13969       compatible with earlier releases.
13970     * In previous releases, the MIPS port had a fake "hilo" register with
13971       the user-visible name accum. This register has been removed.
13972     * The implementation of the [6]SPARC ABIs has changed. As a result,
13973       the code generated will not be binary compatible with earlier
13974       releases in certain cases.
13975     * The configure option --enable-threads=pthreads has been removed;
13976       use --enable-threads=posix instead, which should have the same
13977       effect.
13978     * Code size estimates used by inlining heuristics for C, Objective-C,
13979       C++ and Java have been redesigned significantly. As a result the
13980       parameters of -finline-insns, --param max-inline-insns-single and
13981       --param max-inline-insns-auto need to be reconsidered.
13982     * --param max-inline-slope and --param min-inline-insns have been
13983       removed; they are not needed for the new bottom-up inlining
13984       heuristics.
13985     * The new unit-at-a-time compilation scheme has several compatibility
13986       issues:
13987          + The order in which functions, variables, and top-level asm
13988            statements are emitted may have changed. Code relying on some
13989            particular ordering needs to be updated. The majority of such
13990            top-level asm statements can be replaced by section
13991            attributes.
13992          + Unreferenced static variables and functions are removed. This
13993            may result in undefined references when an asm statement
13994            refers to the variable/function directly. In that case either
13995            the variable/function shall be listed in asm statement operand
13996            or in the case of top-level asm statements the attribute used
13997            shall be used to force function/variable to be always output
13998            and considered as a possibly used by unknown code.
13999            For variables the attribute is accepted only by GCC 3.4 and
14000            newer, while for earlier versions it is sufficient to use
14001            unused to silence warnings about the variables not being
14002            referenced. To keep code portable across different GCC
14003            versions, you can use appropriate preprocessor conditionals.
14004          + Static functions now can use non-standard passing conventions
14005            that may break asm statements calling functions directly.
14006            Again the attribute used shall be used to prevent this
14007            behavior.
14008       As a temporary workaround, -fno-unit-at-a-time can be used, but
14009       this scheme may not be supported by future releases of GCC.
14010     * GCC 3.4 automatically places zero-initialized variables in the .bss
14011       section on some operating systems. Versions of GNU Emacs up to (and
14012       including) 21.3 will not work correctly when using this
14013       optimization; you can use -fno-zero-initialized-in-bss to disable
14014       it.
14015     * If GCC 3.4 is configured with --enable-threads=posix (the default
14016       on most targets that support pthreads) then _REENTRANT will be
14017       defined unconditionally by some libstdc++ headers. C++ code which
14018       relies on that macro to detect whether multi-threaded code is being
14019       compiled might change in meaning, possibly resulting in linker
14020       errors for single-threaded programs. Affected users of [7]Boost
14021       should compile single-threaded code with -DBOOST_DISABLE_THREADS.
14022       See Bugzilla for [8]more information.
14023
14024General Optimizer Improvements
14025
14026     * Usability of the profile feedback and coverage testing has been
14027       improved.
14028          + Performance of profiled programs has been improved by faster
14029            profile merging code.
14030          + Better use of the profile feedback for optimization (loop
14031            unrolling and loop peeling).
14032          + File locking support allowing fork() calls and parallel runs
14033            of profiled programs.
14034          + Coverage file format has been redesigned.
14035          + gcov coverage tool has been improved.
14036          + make profiledbootstrap available to build a faster compiler.
14037            Experiments made on i386 hardware showed an 11% speedup on -O0
14038            and a 7.5% speedup on -O2 compilation of a [9]large C++
14039            testcase.
14040          + New value profiling pass enabled via -fprofile-values
14041          + New value profile transformations pass enabled via -fvpt aims
14042            to optimize some code sequences by exploiting knowledge about
14043            value ranges or other properties of the operands. At the
14044            moment a conversion of expensive divisions into cheaper
14045            operations has been implemented.
14046          + New -fprofile-generate and -fprofile-use command-line options
14047            to simplify the use of profile feedback.
14048     * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and
14049       Java which is enabled via -funit-at-a-time (and implied by -O2). In
14050       this scheme a whole file is parsed first and optimized later. The
14051       following basic inter-procedural optimizations are implemented:
14052          + Removal of unreachable functions and variables
14053          + Discovery of local functions (functions with static linkage
14054            whose address is never taken)
14055          + On i386, these local functions use register parameter passing
14056            conventions.
14057          + Reordering of functions in topological order of the call graph
14058            to enable better propagation of optimizing hints (such as the
14059            stack alignments needed by functions) in the back end.
14060          + Call graph based out-of-order inlining heuristics which allows
14061            to limit overall compilation unit growth (--param
14062            inline-unit-growth).
14063       Overall, the unit-at-a-time scheme produces a 1.3% improvement for
14064       the SPECint2000 benchmark on the i386 architecture (AMD Athlon
14065       CPU).
14066     * More realistic code size estimates used by inlining for C,
14067       Objective-C, C++ and Java. The growth of large functions can now be
14068       limited via --param large-function-insns and --param
14069       large-function-growth.
14070     * A new cfg-level loop optimizer pass replaces the old loop unrolling
14071       pass and adds two other loop transformations -- loop peeling and
14072       loop unswitching -- and also uses the profile feedback to limit
14073       code growth. (The three optimizations are enabled by
14074       -funroll-loops, -fpeel-loops and -funswitch-loops flags,
14075       respectively).
14076       The old loop unroller still can be enabled by -fold-unroll-loops
14077       and may produce better code in some cases, especially when the
14078       webizer optimization pass is not run.
14079     * A new web construction pass enabled via -fweb (and implied by -O3)
14080       improves the quality of register allocation, CSE, first scheduling
14081       pass and some other optimization passes by avoiding re-use of
14082       pseudo registers with non-overlapping live ranges. The pass almost
14083       always improves code quality but does make debugging difficult and
14084       thus is not enabled by default by -O2
14085       The pass is especially effective as cleanup after code duplication
14086       passes, such as the loop unroller or the tracer.
14087     * Experimental implementations of superblock or trace scheduling in
14088       the second scheduling pass can be enabled via
14089       -fsched2-use-superblocks and -fsched2-use-traces, respectively.
14090
14091New Languages and Language specific improvements
14092
14093  Ada
14094
14095     * The Ada front end has been updated to include numerous bug fixes
14096       and enhancements. These include:
14097          + Improved project file support
14098          + Additional set of warnings about potential wrong code
14099          + Improved error messages
14100          + Improved code generation
14101          + Improved cross reference information
14102          + Improved inlining
14103          + Better run-time check elimination
14104          + Better error recovery
14105          + More efficient implementation of unbounded strings
14106          + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools,
14107            ...
14108          + New GNAT.xxxx packages (e.g. GNAT.Strings,
14109            GNAT.Exception_Action)
14110          + New pragmas
14111          + New -gnatS switch replacing gnatpsta
14112          + Implementation of new Ada features (in particular limited
14113            with, limited aggregates)
14114
14115  C/Objective-C/C++
14116
14117     * Precompiled headers are now supported. Precompiled headers can
14118       dramatically speed up compilation of some projects. There are some
14119       known defects in the current precompiled header implementation that
14120       will result in compiler crashes in relatively rare situations.
14121       Therefore, precompiled headers should be considered a "technology
14122       preview" in this release. Read the manual for details about how to
14123       use precompiled headers.
14124     * File handling in the preprocessor has been rewritten. GCC no longer
14125       gets confused by symlinks and hardlinks, and now has a correct
14126       implementation of #import and #pragma once. These two directives
14127       have therefore been un-deprecated.
14128     * The undocumented extension that allowed C programs to have a label
14129       at the end of a compound statement, which has been deprecated since
14130       GCC 3.0, has been removed.
14131     * The cast-as-lvalue extension has been removed for C++ and
14132       deprecated for C and Objective-C. In particular, code like this:
14133        int i;
14134        (char) i = 5;
14135
14136       or this:
14137        char *p;
14138        ((int *) p)++;
14139
14140       is no longer accepted for C++ and will not be accepted for C and
14141       Objective-C in a future version.
14142     * The conditional-expression-as-lvalue extension has been deprecated
14143       for C and Objective-C. In particular, code like this:
14144        int a, b, c;
14145        (a ? b : c) = 2;
14146
14147       will not be accepted for C and Objective-C in a future version.
14148     * The compound-expression-as-lvalue extension has been deprecated for
14149       C and Objective-C. In particular, code like this:
14150        int a, b;
14151        (a, b) = 2;
14152
14153       will not be accepted for C and Objective-C in a future version. A
14154       possible non-intrusive workaround is the following:
14155        (*(a, &b)) = 2;
14156
14157     * Several [10]built-in functions such as __builtin_popcount for
14158       counting bits, finding the highest and lowest bit in a word, and
14159       parity have been added.
14160     * The -fwritable-strings option has been deprecated and will be
14161       removed.
14162     * Many C math library functions are now recognized as built-ins and
14163       optimized.
14164     * The C, C++, and Objective-C compilers can now handle source files
14165       written in any character encoding supported by the host C library.
14166       The default input character set is taken from the current locale,
14167       and may be overridden with the -finput-charset command line option.
14168       In the future we will add support for inline encoding markers.
14169
14170  C++
14171
14172     * G++ is now much closer to full conformance to the ISO/ANSI C++
14173       standard. This means, among other things, that a lot of invalid
14174       constructs which used to be accepted in previous versions will now
14175       be rejected. It is very likely that existing C++ code will need to
14176       be fixed. This document lists some of the most common issues.
14177     * A hand-written recursive-descent C++ parser has replaced the
14178       YACC-derived C++ parser from previous GCC releases. The new parser
14179       contains much improved infrastructure needed for better parsing of
14180       C++ source codes, handling of extensions, and clean separation
14181       (where possible) between proper semantics analysis and parsing. The
14182       new parser fixes many bugs that were found in the old parser.
14183     * You must now use the typename and template keywords to disambiguate
14184       dependent names, as required by the C++ standard.
14185        struct K {
14186          typedef int mytype_t;
14187        };
14188
14189        template <class T1> struct A {
14190          template <class T2> struct B {
14191              void callme(void);
14192            };
14193
14194          template <int N> void bar(void)
14195          {
14196            // Use 'typename' to tell the parser that T1::mytype_t names
14197            //  a type. This is needed because the name is dependent (in
14198            //  this case, on template parameter T1).
14199            typename T1::mytype_t x;
14200            x = 0;
14201          }
14202        };
14203
14204        template <class T> void template_func(void)
14205        {
14206          // Use 'template' to prefix member templates within
14207          //  dependent types (a has type A<T>, which depends on
14208          //  the template parameter T).
14209          A<T> a;
14210          a.template bar<0>();
14211
14212          // Use 'template' to tell the parser that B is a nested
14213          //  template class (dependent on template parameter T), and
14214          //  'typename' because the whole A<T>::B<int> is
14215          //  the name of a type (again, dependent).
14216          typename A<T>::template B<int> b;
14217          b.callme();
14218        }
14219
14220        void non_template_func(void)
14221        {
14222          // Outside of any template class or function, no names can be
14223          //  dependent, so the use of the keyword 'typename' and 'template'
14224          //  is not needed (and actually forbidden).
14225          A<K> a;
14226          a.bar<0>();
14227          A<K>::B<float> b;
14228          b.callme();
14229        }
14230     * In a template definition, unqualified names will no longer find
14231       members of a dependent base (as specified by [temp.dep]/3 in the
14232       C++ standard). For example,
14233        template <typename T> struct B {
14234          int m;
14235          int n;
14236          int f ();
14237          int g ();
14238        };
14239        int n;
14240        int g ();
14241        template <typename T> struct C : B<T> {
14242          void h ()
14243          {
14244            m = 0; // error
14245            f ();  // error
14246            n = 0; // ::n is modified
14247            g ();  // ::g is called
14248          }
14249        };
14250       You must make the names dependent, e.g. by prefixing them with
14251       this->. Here is the corrected definition of C<T>::h,
14252        template <typename T> void C<T>::h ()
14253        {
14254          this->m = 0;
14255          this->f ();
14256          this->n = 0
14257          this->g ();
14258        }
14259       As an alternative solution (unfortunately not backwards compatible
14260       with GCC 3.3), you may use using declarations instead of this->:
14261        template <typename T> struct C : B<T> {
14262          using B<T>::m;
14263          using B<T>::f;
14264          using B<T>::n;
14265          using B<T>::g;
14266          void h ()
14267          {
14268            m = 0;
14269            f ();
14270            n = 0;
14271            g ();
14272          }
14273        };
14274     * In templates, all non-dependent names are now looked up and bound
14275       at definition time (while parsing the code), instead of later when
14276       the template is instantiated. For instance:
14277        void foo(int);
14278
14279        template <int> struct A {
14280          static void bar(void){
14281            foo('a');
14282          }
14283        };
14284
14285        void foo(char);
14286
14287        int main()
14288        {
14289          A<0>::bar();    // Calls foo(int), used to call foo(char).
14290        }
14291
14292     * In an explicit instantiation of a class template, you must use
14293       class or struct before the template-id:
14294        template <int N>
14295        class A {};
14296
14297        template A<0>;         // error, not accepted anymore
14298        template class A<0>;   // OK
14299     * The "named return value" and "implicit typename" extensions have
14300       been removed.
14301     * Default arguments in function types have been deprecated and will
14302       be removed.
14303     * ARM-style name-injection of friend declarations has been deprecated
14304       and will be removed. For example: struct S { friend void f(); };
14305       void g() { f(); } will not be accepted by future versions of G++;
14306       instead a declaration of "f" will need to be present outside of the
14307       scope of "S".
14308     * Covariant returns are implemented for all but varadic functions
14309       that require an adjustment.
14310     * When -pedantic is used, G++ now issues errors about spurious
14311       semicolons. For example,
14312        namespace N {}; // Invalid semicolon.
14313        void f() {}; // Invalid semicolon.
14314     * G++ no longer accepts attributes for a declarator after the
14315       initializer associated with that declarator. For example,
14316        X x(1) __attribute__((...));
14317       is no longer accepted. Instead, use:
14318        X x __attribute__((...)) (1);
14319     * Inside the scope of a template class, the name of the class itself
14320       can be treated as either a class or a template. So GCC used to
14321       accept the class name as argument of type template, and template
14322       template parameter. However this is not C++ standard compliant. Now
14323       the name is not treated as a valid template template argument
14324       unless you qualify the name by its scope. For example, the code
14325       below no longer compiles.
14326        template <template <class> class TT> class X {};
14327        template <class T> class Y {
14328          X<Y> x; // Invalid, Y is always a type template parameter.
14329        };
14330       The valid code for the above example is
14331          X< ::Y> x; // Valid.
14332       (Notice the space between < and : to prevent GCC to interpret this
14333       as a digraph for [.)
14334     * Friend declarations that refer to template specializations are
14335       rejected if the template has not already been declared. For
14336       example,
14337        template <typename T>
14338        class C {
14339          friend void f<> (C&);
14340        };
14341       is rejected. You must first declare f as a template,
14342        template <typename T>
14343        void f(T);
14344     * In case of friend declarations, every name used in the friend
14345       declaration must be accessible at the point of that declaration.
14346       Previous versions of G++ used to be less strict about this and
14347       allowed friend declarations for private class members, for example.
14348       See the ISO C++ Standard Committee's [11]defect report #209 for
14349       details.
14350     * Declaration of member functions of class templates as friends are
14351       supported. For example,
14352        template <typename T> struct A {
14353          void f();
14354        };
14355        class C {
14356          template <typename T> friend void A<T>::f();
14357        };
14358     * You must use template <> to introduce template specializations, as
14359       required by the standard. For example,
14360        template <typename T>
14361        struct S;
14362
14363        struct S<int> { };
14364       is rejected. You must write,
14365        template <> struct S<int> {};
14366     * G++ used to accept code like this,
14367        struct S {
14368          int h();
14369          void f(int i = g());
14370          int g(int i = h());
14371        };
14372       This behavior is not mandated by the standard. Now G++ issues an
14373       error about this code. To avoid the error, you must move the
14374       declaration of g before the declaration of f. The default arguments
14375       for g must be visible at the point where it is called.
14376     * The C++ ABI Section 3.3.3 specifications for the array construction
14377       routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return
14378       NULL when the allocator argument returns NULL. These changes are
14379       incorporated into the libstdc++ runtime library.
14380     * Using a name introduced by a typedef in a friend declaration or in
14381       an explicit instantiation is now rejected, as specified by the ISO
14382       C++ standard.
14383        class A;
14384        typedef A B;
14385        class C {
14386          friend class B;      // error, no typedef name here
14387          friend B;            // error, friend always needs class/struct/enum
14388          friend class A;      // OK
14389        };
14390
14391        template <int> class Q {};
14392        typedef Q<0> R;
14393        template class R;      // error, no typedef name here
14394        template class Q<0>;   // OK
14395     * When allocating an array with a new expression, GCC used to allow
14396       parentheses around the type name. This is actually ill-formed and
14397       it is now rejected:
14398        int* a = new (int)[10];    // error, not accepted anymore
14399        int* a = new int[10];      // OK
14400     * When binding an rvalue of class type to a reference, the copy
14401       constructor of the class must be accessible. For instance, consider
14402       the following code:
14403        class A
14404        {
14405        public:
14406          A();
14407
14408        private:
14409          A(const A&);   // private copy ctor
14410        };
14411
14412        A makeA(void);
14413        void foo(const A&);
14414
14415        void bar(void)
14416        {
14417          foo(A());       // error, copy ctor is not accessible
14418          foo(makeA());   // error, copy ctor is not accessible
14419
14420          A a1;
14421          foo(a1);        // OK, a1 is a lvalue
14422        }
14423       This might be surprising at first sight, especially since most
14424       popular compilers do not correctly implement this rule ([12]further
14425       details).
14426     * When forming a pointer to member or a pointer to member function,
14427       access checks for class visibility (public, protected, private) are
14428       now performed using the qualifying scope of the name itself. This
14429       is better explained with an example:
14430        class A
14431        {
14432        public:
14433          void pub_func();
14434        protected:
14435          void prot_func();
14436        private:
14437          void priv_func();
14438        };
14439
14440        class B : public A
14441        {
14442        public:
14443          void foo()
14444          {
14445            &A::pub_func;   // OK, pub_func is accessible through A
14446            &A::prot_func;  // error, cannot access prot_func through A
14447            &A::priv_func;  // error, cannot access priv_func through A
14448
14449            &B::pub_func;   // OK, pub_func is accessible through B
14450            &B::prot_func;  // OK, can access prot_func through B (within B)
14451            &B::priv_func;  // error, cannot access priv_func through B
14452          }
14453        };
14454
14455    Runtime Library (libstdc++)
14456
14457     * Optimization work:
14458          + Streamlined streambuf, filebuf, separate synched with C
14459            Standard I/O streambuf.
14460          + All formatted I/O now uses cached locale information.
14461          + STL optimizations (memory/speed for list, red-black trees as
14462            used by sets and maps).
14463          + More use of GCC builtins.
14464          + String optimizations (avoid contention on
14465            increment/decrement-and-test of the reference count in the
14466            empty-string object, constructor from input_iterators
14467            speedup).
14468     * Static linkage size reductions.
14469     * Large File Support (files larger than 2 GB on 32-bit systems).
14470     * Wide character and variable encoding filebuf work (UTF-8, Unicode).
14471     * Generic character traits.
14472     * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x,
14473       Solaris 2.7 and above, AIX 5.x, Irix 6.5.
14474     * The allocator class is now standard-conformant, and two additional
14475       extension allocators have been added, mt_alloc and
14476       bitmap_allocator.
14477     * PCH support: -include bits/stdc++.h (2x compile speedup).
14478     * Rewrote __cxa_demangle with support for C++ style allocators.
14479     * New debug modes for STL containers and iterators.
14480     * Testsuite rewrite: five times as many tests, plus increasingly
14481       sophisticated tests, including I/O, MT, multi-locale, wide and
14482       narrow characters.
14483     * Use current versions of GNU "autotools" for build/configuration.
14484
14485  Objective-C
14486
14487     * The Objective-C front end has been updated to include the numerous
14488       bug fixes and enhancements previously available only in Apple's
14489       version of GCC. These include:
14490          + Structured exception (@try... @catch... @finally, @throw) and
14491            synchronization (@synchronized) support. These are accessible
14492            via the -fobjc-exceptions switch; as of this writing, they may
14493            only be used in conjunction with -fnext-runtime on Mac OS X
14494            10.3 and later. See [13]Options Controlling Objective-C
14495            Dialect for more information.
14496          + An overhaul of @encode logic. The C99 _Bool and C++ bool type
14497            may now be encoded as 'B'. In addition, the back-end/codegen
14498            dependencies have been removed.
14499          + An overhaul of message dispatch construction, ensuring that
14500            the various receiver types (and casts thereof) are handled
14501            properly, and that correct diagnostics are issued.
14502          + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue"
14503            (-freplace-objc-classes) debugging modes, currently available
14504            on Mac OS X 10.3 and later. See [14]Options Controlling
14505            Objective-C Dialect for more information.
14506          + Access to optimized runtime entry points (-fno-nil-receivers )
14507            on the assumption that message receivers are never nil. This
14508            is currently available on Mac OS X 10.3 and later. See
14509            [15]Options Controlling Objective-C Dialect for more
14510            information.
14511
14512  Java
14513
14514     * Compiling a .jar file will now cause non-.class entries to be
14515       automatically compiled as resources.
14516     * libgcj has been ported to Darwin.
14517     * Jeff Sturm has adapted Jan Hubicka's call graph optimization code
14518       to gcj.
14519     * libgcj has a new gcjlib URL type; this lets URLClassLoader load
14520       code from shared libraries.
14521     * libgcj has been much more completely merged with [16]GNU Classpath.
14522     * Class loading is now much more correct; in particular the caller's
14523       class loader is now used when that is required.
14524     * [17]Eclipse 2.x will run out of the box using gij.
14525     * Parts of java.nio have been implemented. Direct and indirect
14526       buffers work, as do fundamental file and socket operations.
14527     * java.awt has been improved, though it is still not ready for
14528       general use.
14529     * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST
14530       method.
14531     * The MinGW port has matured. Enhancements include socket timeout
14532       support, thread interruption, improved Runtime.exec() handling and
14533       support for accented characters in filenames.
14534
14535  Fortran
14536
14537     * Fortran improvements are listed in the [18]Fortran documentation.
14538
14539New Targets and Target Specific Improvements
14540
14541  Alpha
14542
14543     * Several [19]built-in functions have been added such as
14544       __builtin_alpha_zap to allow utilizing the more obscure
14545       instructions of the CPU.
14546     * Parameter passing of complex arguments has changed to match the
14547       ABI. This change is incompatible with previous GCC versions, but
14548       does fix compatibility with the Tru64 compiler and several corner
14549       cases where GCC was incompatible with itself.
14550
14551  ARM
14552
14553     * Nicolas Pitre has contributed his hand-coded floating-point support
14554       code for ARM. It is both significantly smaller and faster than the
14555       existing C-based implementation, even when building applications
14556       for Thumb. The arm-elf configuration has been converted to use the
14557       new code.
14558     * Support for the Intel's iWMMXt architecture, a second generation
14559       XScale processor, has been added. Enabled at run time with the
14560       -mcpu=iwmmxt command line switch.
14561     * A new ARM target has been added: arm-wince-pe. This is similar to
14562       the arm-pe target, but it defaults to using the APCS32 ABI.
14563     * The existing ARM pipeline description has been converted to the use
14564       the [20]DFA processor pipeline model. There is not much change in
14565       code performance, but the description is now [21]easier to
14566       understand.
14567     * Support for the Cirrus EP9312 Maverick floating point co-processor
14568       added. Enabled at run time with the -mcpu=ep9312 command line
14569       switch. Note however that the multilibs to support this chip are
14570       currently disabled in gcc/config/arm/t-arm-elf, so if you want to
14571       enable their production you will have to uncomment the entries in
14572       that file.
14573
14574  H8/300
14575
14576     * Support for long long has been added.
14577     * Support for saveall attribute has been added.
14578     * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code
14579       for H8/300H and H8S, which is much faster than the previous
14580       implementation.
14581     * A lot of small performance improvements.
14582
14583  IA-32/AMD64 (x86-64)
14584
14585     * Tuning for K8 (AMD Opteron/Athlon64) core is available via
14586       -march=k8 and -mcpu=k8.
14587     * Scalar SSE code generation carefully avoids reformatting penalties,
14588       hidden dependencies and minimizes the number of uops generated on
14589       both Intel and AMD CPUs.
14590     * Vector MMX and SSE operands are now passed in registers to improve
14591       performance and match the argument passing convention used by the
14592       Intel C++ Compiler. As a result it is not possible to call
14593       functions accepting vector arguments compiled by older GCC version.
14594     * Conditional jump elimination is now more aggressive on modern CPUs.
14595     * The Athlon ports has been converted to use the DFA processor
14596       pipeline description.
14597     * Optimization of indirect tail calls is now possible in a similar
14598       fashion as direct sibcall optimization.
14599     * Further small performance improvements.
14600     * -m128bit-long-double is now less buggy.
14601     * __float128 support in 64-bit compilation.
14602     * Support for data structures exceeding 2GB in 64-bit mode.
14603     * -mcpu has been renamed to -mtune.
14604
14605  IA-64
14606
14607     * Tuning code for the Itanium 2 processor has been added. The
14608       generation of code tuned for Itanium 2 (option -mtune=itanium2) is
14609       enabled by default now. To generate code tuned for Itanium 1 the
14610       option -mtune=itanium1 should be used.
14611     * [22]DFA processor pipeline descriptions for the IA-64 processors
14612       have been added. This resulted in about 3% improvement on the
14613       SPECInt2000 benchmark for Itanium 2.
14614     * Instruction bundling for the IA-64 processors has been rewritten
14615       using the DFA pipeline hazard recognizer. It resulted in about 60%
14616       compiler speedup on the SPECInt2000 C programs.
14617
14618  M32R
14619
14620     * Support for the M32R/2 processor has been added by Renesas.
14621     * Support for an M32R GNU/Linux target and PIC code generation has
14622       been added by Renesas.
14623
14624  M68000
14625
14626     * Bernardo Innocenti (Develer S.r.l.) has contributed the
14627       m68k-uclinux target, based on former work done by Paul Dale
14628       (SnapGear Inc.). Code generation for the ColdFire processors family
14629       has been enhanced and extended to support the MCF 53xx and MCF 54xx
14630       cores, integrating former work done by Peter Barada (Motorola).
14631
14632  MIPS
14633
14634    Processor-specific changes
14635
14636     * Support for the RM7000 and RM9000 processors has been added. It can
14637       be selected using the -march compiler option and should work with
14638       any MIPS I (mips-*) or MIPS III (mips64-*) configuration.
14639     * Support for revision 2 of the MIPS32 ISA has been added. It can be
14640       selected with the command-line option -march=mips32r2.
14641     * There is a new option, -mfix-sb1, to work around certain SB-1
14642       errata.
14643
14644    Configuration
14645
14646     * It is possible to customize GCC using the following configure-time
14647       options:
14648          + --with-arch, which specifies the default value of the -march
14649            option.
14650          + --with-tune, which specifies the default value of the -mtune
14651            option.
14652          + --with-abi, which specifies the default ABI.
14653          + --with-float=soft, which tells GCC to use software floating
14654            point by default.
14655          + --with-float=hard, which tells GCC to use hardware floating
14656            point by default.
14657     * A 64-bit GNU/Linux port has been added. The associated
14658       configurations are mips64-linux-gnu and mips64el-linux-gnu.
14659     * The 32-bit GNU/Linux port now supports Java.
14660     * The IRIX 6 configuration now supports the o32 ABI and will build
14661       o32 multilibs by default. This support is compatible with both
14662       binutils and the SGI tools, but note that several features,
14663       including debugging information and DWARF2 exception handling, are
14664       only available when using the GNU assembler. Use of the GNU
14665       assembler and linker (version 2.15 or above) is strongly
14666       recommended.
14667     * The IRIX 6 configuration now supports 128-bit long doubles.
14668     * There are two new RTEMS-specific configurations, mips-rtems and
14669       mipsel-rtems.
14670     * There are two new *-elf configurations, mipsisa32r2-elf and
14671       mipsisa32r2el-elf.
14672
14673    General
14674
14675     * Several [23]ABI bugs have been fixed. Unfortunately, these changes
14676       will break binary compatibility with earlier releases.
14677     * GCC can now use explicit relocation operators when generating
14678       -mabicalls code. This behavior is controlled by -mexplicit-relocs
14679       and can have several performance benefits. For example:
14680          + It allows for more optimization of GOT accesses, including
14681            better scheduling and redundancy elimination.
14682          + It allows sibling calls to be implemented as jumps.
14683          + n32 and n64 leaf functions can use a call-clobbered global
14684            pointer instead of $28.
14685          + The code to set up $gp can be removed from functions that
14686            don't need it.
14687     * A new option, -mxgot, allows the GOT to be bigger than 64k. This
14688       option is equivalent to the assembler's -xgot option and should be
14689       used instead of -Wa,-xgot.
14690     * Frame pointer elimination is now supported when generating 64-bit
14691       MIPS16 code.
14692     * Inline block moves have been optimized to take more account of
14693       alignment information.
14694     * Many internal changes have been made to the MIPS port, mostly aimed
14695       at reducing the reliance on assembler macros.
14696
14697  PowerPC
14698
14699     * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64
14700       [24]ABI incompatibilities regarding the way parameters are passed
14701       during functions calls. These changes may result in incompatibility
14702       between code compiled with GCC 3.3 and GCC 3.4.
14703
14704    PowerPC Darwin
14705
14706     * Support for shared/dylib gcc libraries has been added. It is
14707       enabled by default on powerpc-apple-darwin7.0.0 and up.
14708     * Libgcj is enabled by default. On systems older than
14709       powerpc-apple-darwin7.0.0 you need to install dlcompat.
14710     * 128-bit IBM extended precision format support added for long
14711       double.
14712
14713    PowerPC64 GNU/Linux
14714
14715     * By default, PowerPC64 GNU/Linux now uses natural alignment of
14716       structure elements. The old four byte alignment for double, with
14717       special rules for a struct starting with a double, can be chosen
14718       with -malign-power. This change may result in incompatibility
14719       between code compiled with GCC 3.3 and GCC 3.4.
14720     * -mabi=altivec is now the default rather than -mabi=no-altivec.
14721     * 128-bit IBM extended precision format support added for long
14722       double.
14723
14724  S/390 and zSeries
14725
14726     * New command-line options allow to specify the intended execution
14727       environment for generated code:
14728          + -mesa/-mzarch allows to specify whether to generate code
14729            running in ESA/390 mode or in z/Architecture mode (this is
14730            applicable to 31-bit code only).
14731          + -march allows to specify a minimum processor architecture
14732            level (g5, g6, z900, or z990).
14733          + -mtune allows to specify which processor to tune for.
14734     * It is possible to customize GCC using the following configure-time
14735       options:
14736          + --with-mode, which specifies whether to default to assuming
14737            ESA/390 or z/Architecture mode.
14738          + --with-arch, which specifies the default value of the -march
14739            option.
14740          + --with-tune, which specifies the default value of the -mtune
14741            option.
14742     * Support for the z990 processor has been added, and can be selected
14743       using -march=z990 or -mtune=z990. This includes instruction
14744       scheduling tuned for the superscalar instruction pipeline of the
14745       z990 processor as well as support for all new instructions provided
14746       by the long-displacement facility.
14747     * Support to generate 31-bit code optimized for zSeries processors
14748       (running in ESA/390 or in z/Architecture mode) has been added. This
14749       can be selected using -march=z900 and -mzarch respectively.
14750     * Instruction scheduling for the z900 and z990 processors now uses
14751       the DFA pipeline hazard recognizer.
14752     * GCC no longer generates code to maintain a stack backchain,
14753       previously used to generate stack backtraces for debugging
14754       purposes. As replacement that does not incur runtime overhead,
14755       DWARF-2 call frame information is provided by GCC; this is
14756       supported by GDB 6.1. The old behavior can be restored using the
14757       -mbackchain option.
14758     * The stack frame size of functions may now exceed 2 GB in 64-bit
14759       code.
14760     * A port for the 64-bit IBM TPF operating system has been added; the
14761       configuration is s390x-ibm-tpf. This configuration is supported as
14762       cross-compilation target only.
14763     * Various changes to improve the generated code have been
14764       implemented, including:
14765          + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT
14766            instructions to significantly speed up many floating-point
14767            applications.
14768          + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL
14769            WITH BORROW instructions to speed up long long arithmetic.
14770          + GCC now uses the SEARCH STRING instruction to implement
14771            strlen().
14772          + In many cases, function call overhead for 31-bit code has been
14773            reduced by placing the literal pool after the function code
14774            instead of after the function prolog.
14775          + Register 14 is no longer reserved in 64-bit code.
14776          + Handling of global register variables has been improved.
14777
14778  SPARC
14779
14780     * The option -mflat is deprecated.
14781     * Support for large (> 2GB) frames has been added to the 64-bit port.
14782     * Several [25]ABI bugs have been fixed. Unfortunately, these changes
14783       will break binary compatibility with earlier releases.
14784     * The default debugging format has been switched from STABS to
14785       DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already
14786       the default debugging format for 64-bit code on Solaris.
14787
14788  SuperH
14789
14790     * Support for the SH2E processor has been added. Enabled at run time
14791       with the -m2e command line switch, or at configure time by
14792       specifying sh2e as the machine part of the target triple.
14793
14794  V850
14795
14796     * Support for the Mitsubishi V850E1 processor has been added. This is
14797       a variant of the V850E processor with some additional debugging
14798       instructions.
14799
14800  Xtensa
14801
14802     * Several ABI bugs have been fixed. Unfortunately, these changes
14803       break binary compatibility with earlier releases.
14804          + For big-endian processors, the padding of aggregate return
14805            values larger than a word has changed. If the size of an
14806            aggregate return value is not a multiple of 32 bits, previous
14807            versions of GCC inserted padding in the most-significant bytes
14808            of the first return value register. Aggregates larger than a
14809            word are now padded in the least-significant bytes of the last
14810            return value register used. Aggregates smaller than a word are
14811            still padded in the most-significant bytes. The return value
14812            padding has not changed for little-endian processors.
14813          + Function arguments with 16-byte alignment are now properly
14814            aligned.
14815          + The implementation of the va_list type has changed. A va_list
14816            value created by va_start from a previous release cannot be
14817            used with va_arg from this release, or vice versa.
14818     * More processor configuration options for Xtensa processors are
14819       supported:
14820          + the ABS instruction is now optional;
14821          + the ADDX* and SUBX* instructions are now optional;
14822          + an experimental CONST16 instruction can be used to synthesize
14823            constants instead of loading them from constant pools.
14824       These and other Xtensa processor configuration options can no
14825       longer be enabled or disabled by command-line options; the
14826       processor configuration must be specified by the xtensa-config.h
14827       header file when building GCC. Additionally, the
14828       -mno-serialize-volatile option is no longer supported.
14829
14830Obsolete Systems
14831
14832   Support for a number of older systems has been declared obsolete in GCC
14833   3.4. Unless there is activity to revive them, the next release of GCC
14834   will have their sources permanently removed.
14835
14836   All configurations of the following processor architectures have been
14837   declared obsolete:
14838     * Mitsubishi D30V, d30v-*
14839     * AT&T DSP1600 and DSP1610, dsp16xx-*
14840     * Intel 80960, i960
14841
14842   Also, some individual systems have been obsoleted:
14843     * ARM Family
14844          + Support for generating code for operation in APCS/26 mode
14845            (-mapcs-26).
14846     * IBM ESA/390
14847          + "Bigfoot" port, i370-*. (The other port, s390-*, is actively
14848            maintained and supported.)
14849     * Intel 386 family
14850          + MOSS, i?86-moss-msdos and i?86-*-moss*
14851          + NCR 3000 running System V r.4, i?86-ncr-sysv4*
14852          + FreeBSD with a.out object format, i?86-*-freebsd*aout* and
14853            i?86-*-freebsd2*
14854          + GNU/Linux with a.out object format, i?86-linux*aout*
14855          + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1*
14856          + Interix versions before Interix 3, i?86-*-interix
14857          + Mach microkernel, i?86-mach*
14858          + SCO UnixWare with UDK, i?86-*-udk*
14859          + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]*
14860          + VSTa microkernel, i386-*-vsta
14861     * Motorola M68000 family
14862          + HPUX, m68k-hp-hpux* and m68000-hp-hpux*
14863          + NetBSD with a.out object format (before NetBSD 1.4),
14864            m68k-*-*-netbsd* except m68k-*-*-netbsdelf*
14865          + Generic System V r.4, m68k-*-sysv4*
14866     * VAX
14867          + Generic VAX, vax-*-* (This is generic VAX only; we have not
14868            obsoleted any VAX triples for specific operating systems.)
14869
14870Documentation improvements
14871
14872Other significant improvements
14873
14874     * The build system has undergone several significant cleanups.
14875       Subdirectories will only be configured if they are being built, and
14876       all subdirectory configures are run from the make command. The top
14877       level has been autoconfiscated.
14878     * Building GCC no longer writes to its source directory. This should
14879       help those wishing to share a read-only source directory over NFS
14880       or build from a CD. The exceptions to this feature are if you
14881       configure with either --enable-maintainer-mode or
14882       --enable-generated-files-in-srcdir.
14883     * The -W warning option has been renamed to -Wextra, which is more
14884       easily understood. The older spelling will be retained for
14885       backwards compatibility.
14886     * Substantial improvements in compile time have been made,
14887       particularly for non-optimizing compilations.
14888     __________________________________________________________________
14889
14890GCC 3.4.0
14891
14892  Bug Fixes
14893
14894   A vast number of bugs have been fixed in 3.4.0, too many to publish a
14895   complete list here. [26]Follow this link to query the Bugzilla database
14896   for the list of over 900 bugs fixed in 3.4.0. This is the list of all
14897   bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4
14898   regressions.
14899     __________________________________________________________________
14900
14901GCC 3.4.1
14902
14903  Bug Fixes
14904
14905   This section lists the problem reports (PRs) from GCC's bug tracking
14906   system that are known to be fixed in the 3.4.1 release. This list might
14907   not be complete (that is, it is possible that some PRs that have been
14908   fixed are not listed here).
14909
14910    Bootstrap failures
14911
14912     * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler
14913       emitted - PIC related
14914     * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf
14915     * [29]14760 A bug in configure.in prevents using both
14916       --program-suffix and --program-prefix
14917     * [30]14671 [hppa64] bootstrap fails: ICE in
14918       save_call_clobbered_regs, in caller_save.c
14919     * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on
14920       Alpha
14921     * [32]15178 Solaris 9/x86 fails linking after stage 3
14922
14923    Multi-platform internal compiler errors (ICEs)
14924
14925     * [33]12753 (preprocessor) Memory corruption in preprocessor on bad
14926       input
14927     * [34]13985 ICE in gcc.c-torture/compile/930621-1.c
14928     * [35]14810 (c++) tree check failures with invalid code involving
14929       templates
14930     * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in
14931       cp/parser.c
14932     * [37]15044 (c++) ICE on syntax error, template header
14933     * [38]15057 (c++) Compiling of conditional value throw constructs
14934       cause a segmentation violation
14935     * [39]15064 (c++) typeid of template parameter gives ICE
14936     * [40]15142 (c++) ICE when passing a string where a char* is expected
14937       in a throw statement
14938     * [41]15159 ICE in rtl_verify_flow_info_1
14939     * [42]15165 (c++) ICE in instantiate_template
14940     * [43]15193 Unary minus using pointer to V4SF vector causes
14941       -fforce-mem to exhaust all memory
14942     * [44]15209 (c++) Runs out of memory with packed structs
14943     * [45]15227 (c++) Trouble with invalid function definition
14944     * [46]15285 (c++) instantiate_type ICE when forming pointer to
14945       template function
14946     * [47]15299 (c++) ICE in resolve_overloaded_unification
14947     * [48]15329 (c++) ICE on constructor of member template
14948     * [49]15550 ICE in extract_insn, in recog.c
14949     * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c
14950     * [51]15640 (c++) ICE on invalid code in arg_assoc, in
14951       cp/name-lookup.c
14952     * [52]15666 [unit-at-a-time] Gcc abort on valid code
14953     * [53]15696 (c++) ICE with bad pointer-to-member code
14954     * [54]15701 (c++) ICE with friends and template template parameter
14955     * [55]15761 ICE in do_SUBST, in combine.c
14956     * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops
14957
14958    Ada
14959
14960     * [57]14538 All RTEMS targets broken for gnat
14961
14962    C front end
14963
14964     * [58]12391 missing warning about assigning to an incomplete type
14965     * [59]14649 atan(1.0) should not be a constant expression
14966     * [60]15004 [unit-at-a-time] no warning for unused paramater in
14967       static function
14968     * [61]15749 --pedantic-errors behaves differently from --pedantic
14969       with C-compiler on GNU/Linux
14970
14971    C++ compiler and library
14972
14973     * [62]10646 non-const reference is incorrectly matched in a "const T"
14974       partial specialization
14975     * [63]12077 wcin.rdbuf()->in_avail() return value too high
14976     * [64]13598 enc_filebuf doesn't work
14977     * [65]14211 const_cast returns lvalue but should be rvalue
14978     * [66]14220 num_put::do_put() undesired float/double behavior
14979     * [67]14245 problem with user-defined allocators in std::basic_string
14980     * [68]14340 libstdc++ Debug mode: failure to convert iterator to
14981       const_iterator
14982     * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal
14983       FILE*
14984     * [70]14668 no warning anymore for reevaluation of declaration
14985     * [71]14775 LFS (large file support) tests missing
14986     * [72]14821 Duplicate namespace alias declaration should not conflict
14987     * [73]14930 Friend declaration ignored
14988     * [74]14932 cannot use offsetof to get offsets of array elements in
14989       g++ 3.4.0
14990     * [75]14950 [non unit-at-a-time] always_inline does not mix with
14991       templates and -O0
14992     * [76]14962 g++ ignores #pragma redefine_extname
14993     * [77]14975 Segfault on low-level write error during imbue
14994     * [78]15002 Linewise stream input is unusably slow (std::string slow)
14995     * [79]15025 compiler accepts redeclaration of template as
14996       non-template
14997     * [80]15046 [arm] Math functions misdetected by cross configuration
14998     * [81]15069 a bit test on a variable of enum type is miscompiled
14999     * [82]15074 g++ -lsupc++ still links against libstdc++
15000     * [83]15083 spurious "statement has no effect" warning
15001     * [84]15096 parse error with templates and pointer to const member
15002     * [85]15287 combination of operator[] and operator .* fails in
15003       templates
15004     * [86]15317 __attribute__ unused in first parameter of constructor
15005       gives error
15006     * [87]15337 sizeof on incomplete type diagnostic
15007     * [88]15361 bitset<>::_Find_next fails
15008     * [89]15412 _GLIBCXX_ symbols symbols defined and used in different
15009       namespaces
15010     * [90]15427 valid code results in incomplete type error
15011     * [91]15471 Incorrect member pointer offsets in anonymous
15012       structs/unions
15013     * [92]15503 nested template problem
15014     * [93]15507 compiler hangs while laying out union
15015     * [94]15542 operator & and template definitions
15016     * [95]15565 SLES9: leading + sign for unsigned int with showpos
15017     * [96]15625 friend defined inside a template fails to find static
15018       function
15019     * [97]15629 Function templates, overloads, and friend name injection
15020     * [98]15742 'noreturn' attribute ignored in method of template
15021       functions.
15022     * [99]15775 Allocator::pointer consistently ignored
15023     * [100]15821 Duplicate namespace alias within namespace rejected
15024     * [101]15862 'enum yn' fails (confict with undeclared builtin)
15025     * [102]15875 rejects pointer to member in template
15026     * [103]15877 valid code using templates and anonymous enums is
15027       rejected
15028     * [104]15947 Puzzling error message for wrong destructor declaration
15029       in template class
15030     * [105]16020 cannot copy __gnu_debug::bitset
15031     * [106]16154 input iterator concept too restrictive
15032     * [107]16174 deducing top-level consts
15033
15034    Java
15035
15036     * [108]14315 Java compiler is not parallel make safe
15037
15038    Fortran
15039
15040     * [109]15151 [g77] incorrect logical i/o in 64-bit mode
15041
15042    Objective-C
15043
15044     * [110]7993 private variables cannot be shadowed in subclasses
15045
15046    Optimization bugs
15047
15048     * [111]15228 useless copies of floating point operands
15049     * [112]15345 [non-unit-at-a-time] unreferenced nested inline
15050       functions not optimized away
15051     * [113]15945 Incorrect floating point optimization
15052     * [114]15526 ftrapv aborts on 0 * (-1)
15053     * [115]14690 Miscompiled POOMA tests
15054     * [116]15112 GCC generates code to write to unchanging memory
15055
15056    Preprocessor
15057
15058     * [117]15067 Minor glitch in the source of cpp
15059
15060    Main driver program bugs
15061
15062     * [118]1963 collect2 interprets -oldstyle_liblookup as -o
15063       ldstyle_liblookup
15064
15065    x86-specific (Intel/AMD)
15066
15067     * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND*
15068       section}
15069
15070    HPPA-specific
15071
15072     * [120]14782 GCC produces an unaligned data access at -O2
15073     * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2
15074     * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c
15075
15076    IA64-specific
15077
15078     * [123]14610 __float80 constants incorrectly emitted
15079     * [124]14813 init_array sections are initialized in the wrong order
15080     * [125]14857 GCC segfault on duplicated asm statement
15081     * [126]15598 Gcc 3.4 ICE on valid code
15082     * [127]15653 Gcc 3.4 ICE on valid code
15083
15084    MIPS-specific
15085
15086     * [128]15189 wrong filling of delay slot with -march=mips1 -G0
15087       -mno-split-addresses -mno-explicit-relocs
15088     * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as
15089       2.14.91
15090     * [130]16144 Bogus reference to __divdf3 when -O1
15091     * [131]16176 Miscompilation of unaligned data in MIPS backend
15092
15093    PowerPC-specific
15094
15095     * [132]11591 ICE in gcc.dg/altivec-5.c
15096     * [133]12028 powerpc-eabispe produces bad sCOND operation
15097     * [134]14478 rs6000 geu/ltu patterns generate incorrect code
15098     * [135]14567 long double and va_arg complex args
15099     * [136]14715 Altivec stack layout may overlap gpr save with stack
15100       temps
15101     * [137]14902 (libstdc++) Stream checking functions fail when -pthread
15102       option is used.
15103     * [138]14924 Compiler ICE on valid code
15104     * [139]14960 -maltivec affects vector return with -mabi=no-altivec
15105     * [140]15106 vector varargs failure passing from altivec to
15106       non-altivec code for -m32
15107     * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 &
15108       half-word operation
15109     * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx
15110       and stvx
15111     * [143]15662 Segmentation fault when an exception is thrown - even if
15112       try and catch are specified
15113
15114    s390-specific
15115
15116     * [144]15054 Bad code due to overlapping stack temporaries
15117
15118    SPARC-specific
15119
15120     * [145]15783 ICE with union assignment in 64-bit mode
15121     * [146]15626 GCC 3.4 emits "ld: warning: relocation error:
15122       R_SPARC_UA32"
15123
15124    x86-64-specific
15125
15126     * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64
15127     * [148]14723 Backported -march=nocona from mainline
15128     * [149]15290 __float128 failed to pass to function properly
15129
15130    Cygwin/Mingw32-specific
15131
15132     * [150]15250 Option -mms-bitfields support on GCC 3.4 is not
15133       conformant to MS layout
15134     * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe
15135       worker on windows32 targets
15136
15137    Bugs specific to embedded processors
15138
15139     * [152]8309 [m68k] -m5200 produces erroneous SImode set of short
15140       varaible on stack
15141     * [153]13250 [SH] Gcc code for rotation clobbers the register, but
15142       gcc continues to use the register as if it was not clobbered
15143     * [154]13803 [coldfire] movqi operand constraints too restrictivefor
15144       TARGET_COLDFIRE
15145     * [155]14093 [SH] ICE for code when using -mhitachi option in SH
15146     * [156]14457 [m6811hc] ICE with simple c++ source
15147     * [157]14542 [m6811hc] ICE on simple source
15148     * [158]15100 [SH] cc1plus got hang-up on
15149       libstdc++-v3/testsuite/abi_check.cc
15150     * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on
15151       cris-*
15152     * [160]15396 [SH] ICE with -O2 -fPIC
15153     * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for
15154       ColdFire
15155
15156    Testsuite problems (compiler not affected)
15157
15158     * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely
15159     * [163]15488 (libstdc++) possibly insufficient file permissions for
15160       executing test suite
15161     * [164]15489 (libstdc++) testsuite_files determined incorrectly
15162
15163    Documentation bugs
15164
15165     * [165]13928 (libstdc++) no whatis info in some man pages generated
15166       by doxygen
15167     * [166]14150 Ada documentation out of date
15168     * [167]14949 (c++) Need to document method visibility changes
15169     * [168]15123 libstdc++-doc: Allocators.3 manpage is empty
15170     __________________________________________________________________
15171
15172GCC 3.4.2
15173
15174  Bug Fixes
15175
15176   This section lists the problem reports (PRs) from GCC's bug tracking
15177   system that are known to be fixed in the 3.4.2 release. This list might
15178   not be complete (that is, it is possible that some PRs that have been
15179   fixed are not listed here).
15180
15181    Bootstrap failures and issues
15182
15183     * [169]16469 [mips-sgi-irix5.3] bootstrap fails in
15184       libstdc++-v3/testsuite
15185     * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by
15186       profiledbootstrap does not work with the built compiler
15187     * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf
15188
15189    Multi-platform internal compiler errors (ICEs)
15190
15191     * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in
15192       cp_parser_class_specifier, in cp/parser.c
15193     * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
15194     * [174]15461 (c++) ICE due to NRV and inlining
15195     * [175]15890 (c++) ICE in c_expand_expr, in c-common.c
15196     * [176]16180 ICE: segmentation fault in RTL optimization
15197     * [177]16224 (c++) ICE in write_unscoped_name (template/namespace)
15198     * [178]16408 ICE: in delete_insn, in cfgrtl.c
15199     * [179]16529 (c++) ICE for: namespace-alias shall not be declared as
15200       the name of any other entity
15201     * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw
15202     * [181]16706 (c++) ICE in finish_member_declaration, in
15203       cp/semantics.c
15204     * [182]16810 (c++) Legal C++ program with cast gives ICE in
15205       build_ptrmemfunc
15206     * [183]16851 (c++) ICE when throwing a comma expression
15207     * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c
15208     * [185]16904 (c++) ICE in finish_class_member_access_expr, in
15209       cp/typeck.c
15210     * [186]16905 (c++) ICE (segfault) with exceptions
15211     * [187]16964 (c++) ICE in cp_parser_class_specifier due to
15212       redefinition
15213     * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x'
15214       (identifier_node) in dependent_template_p, in cp/pt.c
15215
15216    Preprocessor bugs
15217
15218     * [189]16366 Preprocessor option -remap causes memory corruption
15219
15220    Optimization
15221
15222     * [190]15345 unreferenced nested inline functions not optimized away
15223     * [191]16590 Incorrect execution when compiling with -O2
15224     * [192]16693 Bitwise AND is lost when used within a cast to an enum
15225       of the same precision
15226     * [193]17078 Jump into if(0) substatement fails
15227
15228    Problems in generated debug information
15229
15230     * [194]13956 incorrect stabs for nested local variables
15231
15232    C front end bugs
15233
15234     * [195]16684 GCC should not warn about redundant redeclarations of
15235       built-ins
15236
15237    C++ compiler and library
15238
15239     * [196]12658 Thread safety problems in locale::global() and
15240       locale::locale()
15241     * [197]13092 g++ accepts invalid pointer-to-member conversion
15242     * [198]15320 Excessive memory consumption
15243     * [199]16246 Incorrect template argument deduction
15244     * [200]16273 Memory exhausted when using nested classes and virtual
15245       functions
15246     * [201]16401 ostringstream in gcc 3.4.x very slow for big data
15247     * [202]16411 undefined reference to
15248       __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char>
15249       >::file()
15250     * [203]16489 G++ incorrectly rejects use of a null constant integral
15251       expression as a null constant pointer
15252     * [204]16618 offsetof fails with constant member
15253     * [205]16637 syntax error reported for valid input code
15254     * [206]16717 __attribute__((constructor)) broken in C++
15255     * [207]16813 compiler error in DEBUG version of range insertion
15256       std::map::insert
15257     * [208]16853 pointer-to-member initialization from incompatible one
15258       accepted
15259     * [209]16889 ambiguity is not detected
15260     * [210]16959 Segmentation fault in ios_base::sync_with_stdio
15261
15262    Java compiler and library
15263
15264     * [211]7587 direct threaded interpreter not thread-safe
15265     * [212]16473 ServerSocket accept() leaks file descriptors
15266     * [213]16478 Hash synchronization deadlock with finalizers
15267
15268    Alpha-specific
15269
15270     * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c
15271     * [215]16974 could not split insn (ice in final_scan_insn, in
15272       final.c)
15273
15274    x86-specific
15275
15276     * [216]16298 ICE in output_operand
15277     * [217]17113 ICE with SSE2 intrinsics
15278
15279    x86-64 specific
15280
15281     * [218]14697 libstdc++ couldn't find 32bit libgcc_s
15282
15283    MIPS-specific
15284
15285     * [219]15869 [mips64] No NOP after LW (with -mips1 -O0)
15286     * [220]16325 [mips64] value profiling clobbers gp on mips
15287     * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern
15288       char[]s
15289     * [222]16380 [mips64] Use of uninitialised register after dbra
15290       conversion
15291     * [223]16407 [mips64] Unaligned access to local variables
15292     * [224]16643 [mips64] verify_local_live_at_start ICE after
15293       crossjumping & cfgcleanup
15294
15295    ARM-specific
15296
15297     * [225]15927 THUMB -O2: strength-reduced iteration variable ends up
15298       off by 1
15299     * [226]15948 THUMB: ICE with non-commutative cbranch
15300     * [227]17019 THUMB: bad switch statement in md code for
15301       addsi3_cbranch_scratch
15302
15303    IA64-specific
15304
15305     * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c
15306       (-mtune=merced)
15307     * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c
15308       (-mtune=itanium)
15309     * [230]16278 Gcc failed to build Linux kernel with -mtune=merced
15310     * [231]16414 ICE on valid code: typo in comparison of asm_noperands
15311       result
15312     * [232]16445 ICE on valid code: don't count ignored insns
15313     * [233]16490 ICE (segfault) while compiling with -fprofile-use
15314     * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS
15315
15316    PowerPC-specific
15317
15318     * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x
15319     * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions
15320       issue)
15321
15322    SPARC-specific
15323
15324     * [237]16199 ICE while compiling apache 2.0.49
15325     * [238]16416 -m64 doesn't imply -mcpu=v9 anymore
15326     * [239]16430 ICE when returning non-C aggregates larger than 16 bytes
15327
15328    Bugs specific to embedded processors
15329
15330     * [240]16379 [m32r] can't output large model function call of memcpy
15331     * [241]17093 [m32r] ICE with -msdata=use -O0
15332     * [242]17119 [m32r] ICE at switch case 0x8000
15333
15334    DJGPP-specific
15335
15336     * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp
15337
15338    Alpha Tru64-specific
15339
15340     * [244]16210 libstdc++ gratuitously omits "long long" I/O
15341
15342    Testsuite, documentation issues (compiler is not affected):
15343
15344     * [245]15488 (libstdc++) possibly insufficient file permissions for
15345       executing test suite
15346     * [246]16250 ada/doctools runs makeinfo even in release tarball
15347     __________________________________________________________________
15348
15349GCC 3.4.3
15350
15351   This is the [247]list of problem reports (PRs) from GCC's bug tracking
15352   system that are known to be fixed in the 3.4.3 release. This list might
15353   not be complete (that is, it is possible that some PRs that have been
15354   fixed are not listed here).
15355
15356    Bootstrap failures
15357
15358     * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1
15359     * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold
15360       when undeclared
15361
15362    Internal compiler errors (ICEs) affecting multiple platforms
15363
15364     * [250]13948 (java) GCJ segmentation fault while compiling GL4Java
15365       .class files
15366     * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
15367     * [252]16301 (c++) ICE when "strong" attribute is attached to a using
15368       directive
15369     * [253]16566 ICE with flexible arrays
15370     * [254]17023 ICE with nested functions in parameter declaration
15371     * [255]17027 ICE with noreturn function in loop at -O2
15372     * [256]17524 ICE in grokdeclarator, in cp/decl.c
15373     * [257]17826 (c++) ICE in cp_tree_equal
15374
15375    C and optimization bugs
15376
15377     * [258]15526 -ftrapv aborts on 0 * (-1)
15378     * [259]16999 #ident stopped working
15379     * [260]17503 quadratic behaviour in invalid_mode_change_p
15380     * [261]17581 Long long arithmetic fails inside a switch/case
15381       statement when compiled with -O2
15382     * [262]18129 -fwritable-strings doesn't work
15383
15384    C++ compiler and library bugs
15385
15386     * [263]10975 incorrect initial ostringstream::tellp()
15387     * [264]11722 Unbuffered filebuf::sgetn is slow
15388     * [265]14534 Unrecognizing static function as a template parameter
15389       when its return value is also templated
15390     * [266]15172 Copy constructor optimization in aggregate
15391       initialization
15392     * [267]15786 Bad error message for frequently occuring error.
15393     * [268]16162 Rejects valid member-template-definition
15394     * [269]16612 empty basic_strings can't live in shared memory
15395     * [270]16715 std::basic_iostream is instantiated when used, even
15396       though instantiations are already contained in libstdc++
15397     * [271]16848 code in /ext/demangle.h appears broken
15398     * [272]17132 GCC fails to eliminate function template specialization
15399       when argument deduction fails
15400     * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep::
15401       in ropeimpl.h
15402     * [274]17327 use of `enumeral_type' in template type unification
15403     * [275]17393 "unused variable '._0'" warning with -Wall
15404     * [276]17501 Confusion with member templates
15405     * [277]17537 g++ not passing -lstdc++ to linker when all command line
15406       arguments are libraries
15407     * [278]17585 usage of unqualified name of static member from within
15408       class not allowed
15409     * [279]17821 Poor diagnostic for using "." instead of "->"
15410     * [280]17829 wrong error: call of overloaded function is ambiguous
15411     * [281]17851 Misleading diagnostic for invalid function declarations
15412       with undeclared types
15413     * [282]17976 Destructor is called twice
15414     * [283]18020 rejects valid definition of enum value in template
15415     * [284]18093 bogus conflict in namespace aliasing
15416     * [285]18140 C++ parser bug when using >> in templates
15417
15418    Fortran
15419
15420     * [286]17541 data statements with double precision constants fail
15421
15422    x86-specific
15423
15424     * [287]17853 -O2 ICE for MMX testcase
15425
15426    SPARC-specific
15427
15428     * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c
15429
15430    Darwin-specific
15431
15432     * [289]17167 FATAL:Symbol L_foo$stub already defined.
15433
15434    AIX-specific
15435
15436     * [290]17277 could not catch an exception when specified -maix64
15437
15438    Solaris-specific
15439
15440     * [291]17505 <cmath> calls acosf(), ceilf(), and other functions
15441       missing from system libraries
15442
15443    HP/UX specific:
15444
15445     * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl
15446
15447    ARM-specific
15448
15449     * [293]17384 ICE with mode attribute on structures
15450
15451    MIPS-specific
15452
15453     * [294]17770 No NOP after LWL with -mips1
15454
15455    Other embedded target specific
15456
15457     * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c
15458     * [296]14064 [avr-elf] -fdata-sections triggers ICE
15459     * [297]14678 [m68hc11-elf] gcc ICE
15460     * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__
15461     * [299]15790 [i686-coff] Alignment error building gcc with i686-coff
15462       target
15463     * [300]15886 [SH] Miscompilation with -O2 -fPIC
15464     * [301]16884 [avr-elf] [fweb related] bug while initializing
15465       variables
15466
15467    Bugs relating to debugger support
15468
15469     * [302]13841 missing debug info for _Complex function arguments
15470     * [303]15860 [big-endian targets] No DW_AT_location debug info is
15471       emitted for formal arguments to a function that uses "register"
15472       qualifiers
15473
15474    Testsuite issues (compiler not affected)
15475
15476     * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH
15477     * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH
15478     * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit
15479       testsuite
15480
15481    Documentation
15482
15483     * [307]15498 typo in gcc manual: non-existing locale example en_UK,
15484       should be en_GB
15485     * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap:
15486       document broken shell
15487     * [309]16406 USE_LD_AS_NEEDED undocumented
15488     __________________________________________________________________
15489
15490GCC 3.4.4
15491
15492   This is the [310]list of problem reports (PRs) from GCC's bug tracking
15493   system that are known to be fixed in the 3.4.4 release. This list might
15494   not be complete (that is, it is possible that some PRs that have been
15495   fixed are not listed here).
15496     __________________________________________________________________
15497
15498GCC 3.4.5
15499
15500   This is the [311]list of problem reports (PRs) from GCC's bug tracking
15501   system that are known to be fixed in the 3.4.5 release. This list might
15502   not be complete (that is, it is possible that some PRs that have been
15503   fixed are not listed here).
15504
15505    Bootstrap issues
15506
15507     * [312]24688 sco_math fixincl breaks math.h
15508
15509    C compiler bugs
15510
15511     * [313]17188 struct Foo { } redefinition
15512     * [314]20187 wrong code for ((unsigned char)(unsigned long
15513       long)((a?a:1)&(a*b)))?0:1)
15514     * [315]21873 infinite warning loop on bad array initializer
15515     * [316]21899 enum definition accepts values to be overriden
15516     * [317]22061 ICE in find_function_data, in function.c
15517     * [318]22308 Failure to diagnose violation of constraint 6.516p2
15518     * [319]22458 ICE on missing brace
15519     * [320]22589 ICE casting to long long
15520     * [321]24101 Segfault with preprocessed source
15521
15522    C++ compiler and library bugs
15523
15524     * [322]10611 operations on vector mode not recognized in C++
15525     * [323]13377 unexpected behavior of namespace usage directive
15526     * [324]16002 Strange error message with new parser
15527     * [325]17413 local classes as template argument
15528     * [326]17609 spurious error message after using keyword
15529     * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c
15530     * [328]18124 ICE with invalid template template parameter
15531     * [329]18155 typedef in template declaration not rejected
15532     * [330]18177 ICE with const_cast for undeclared variable
15533     * [331]18368 C++ error message regression
15534     * [332]16378 ICE when returning a copy of a packed member
15535     * [333]18466 int ::i; accepted
15536     * [334]18512 ICE on invalid usage of template base class
15537     * [335]18454 ICE when returning undefined type
15538     * [336]18738 typename not allowed with non-dependent qualified name
15539     * [337]18803 rejects access to operator() in template
15540     * [338]19004 ICE in uses_template_parms, in cp/pt.c
15541     * [339]19208 Spurious error about variably modified type
15542     * [340]18253 bad error message / ICE for invalid template parameter
15543     * [341]19608 ICE after friend function definition in local class
15544     * [342]19884 ICE on explicit instantiation of a non-template
15545       constructor
15546     * [343]20153 ICE when C++ template function contains anonymous union
15547     * [344]20563 Infinite loop in diagnostic (and ice after error
15548       message)
15549     * [345]20789 ICE with incomplete type in template
15550     * [346]21336 Internal compiler error when using custom new operators
15551     * [347]21768 ICE in error message due to violation of coding
15552       conventions
15553     * [348]21853 constness of pointer to data member ignored
15554     * [349]21903 Default argument of template function causes a
15555       compile-time error
15556     * [350]21983 multiple diagnostics
15557     * [351]21987 New testsuite failure
15558       g++.dg/warn/conversion-function-1.C
15559     * [352]22153 ICE on invalid template specialization
15560     * [353]22172 Internal compiler error, seg fault.
15561     * [354]21286 filebuf::xsgetn vs pipes
15562     * [355]22233 ICE with wrong number of template parameters
15563     * [356]22508 ICE after invalid operator new
15564     * [357]22545 ICE with pointer to class member & user defined
15565       conversion operator
15566     * [358]23528 Wrong default allocator in ext/hash_map
15567     * [359]23550 char_traits requirements/1.cc test bad math
15568     * [360]23586 Bad diagnostic for invalid namespace-name
15569     * [361]23624 ICE in invert_truthvalue, in fold-const.c
15570     * [362]23639 Bad error message: not a member of '<declaration error>'
15571     * [363]23797 ICE on typename outside template
15572     * [364]23965 Bogus error message: no matching function for call to
15573       'foo(<type error>)'
15574     * [365]24052 &#`label_decl' not supported by dump_expr#<expression
15575       error>
15576     * [366]24580 virtual base class cause exception not to be caught
15577
15578    Problems in generated debug information
15579
15580     * [367]24267 Bad DWARF for altivec vectors
15581
15582    Optimizations issues
15583
15584     * [368]17810 ICE in verify_local_live_at_start
15585     * [369]17860 Wrong generated code for loop with varying bound
15586     * [370]21709 ICE on compile-time complex NaN
15587     * [371]21964 broken tail call at -O2 or more
15588     * [372]22167 Strange optimization bug when using -Os
15589     * [373]22619 Compilation failure for real_const_1.f and
15590       real_const_2.f90
15591     * [374]23241 Invalid code generated for comparison of uchar to 255
15592     * [375]23478 Miscompilation due to reloading of a var that is also
15593       used in EH pad
15594     * [376]24470 segmentation fault in cc1plus when compiling with -O
15595     * [377]24950 ICE in operand_subword_force
15596
15597    Precompiled headers problems
15598
15599     * [378]14400 Cannot compile qt-x11-free-3.3.0
15600     * [379]14940 PCH largefile test fails on various platforms
15601
15602    Preprocessor bugs
15603
15604     * [380]20239 ICE on empty preprocessed input
15605     * [381]15220 "gcc -E -MM -MG" reports missing system headers in
15606       source directory
15607
15608    Testsuite issues
15609
15610     * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on
15611       i686-pc-linux-gnu
15612
15613    Alpha specific
15614
15615     * [383]21888 bootstrap failure with linker relaxation enabled
15616
15617    ARM specific
15618
15619     * [384]15342 [arm-linux]: ICE in verify_local_live_at_start
15620     * [385]23985 Memory aliasing information incorrect in inlined memcpy
15621
15622    ColdFile specific
15623
15624     * [386]16719 Illegal move of byte into address register causes
15625       compiler to ICE
15626
15627    HPPA specific
15628
15629     * [387]21723 ICE while building libgfortran
15630     * [388]21841 -mhp-ld/-mgnu-ld documentation
15631
15632    IA-64 specific
15633
15634     * [389]23644 IA-64 hardware models and configuration options
15635       documentation error
15636     * [390]24718 Shared libgcc not used for linking by default
15637
15638    M68000 specific
15639
15640     * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c
15641
15642    MIPS specific
15643
15644     * [392]20621 ICE in change_address_1, in emit-rtl.c
15645
15646    PowerPC and PowerPC64 specific
15647
15648     * [393]18583 error on valid code: const
15649       __attribute__((altivec(vector__))) doesn't work in arrays
15650     * [394]20191 ICE in reload_cse_simplify_operands
15651     * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined
15652     * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set
15653     * [397]23404 gij trashes args of functions with more than 8 fp args
15654     * [398]23539 C & C++ compiler generating misaligned references
15655       regardless of compiler flags
15656     * [399]24102 floatdisf2_internal2 broken
15657     * [400]24465 -mminimal-toc miscompilation of __thread vars
15658
15659    Solaris specific
15660
15661     * [401]19933 Problem with define of HUGE_VAL in math_c99
15662     * [402]21889 Native Solaris assembler cannot grok DTP-relative debug
15663       symbols
15664
15665    SPARC specific
15666
15667     * [403]19300 PCH failures on sparc-linux
15668     * [404]20301 Assembler labels have a leading "-"
15669     * [405]20673 C PCH testsuite assembly comparison failure
15670
15671    x86 and x86_64 specific
15672
15673     * [406]18582 ICE with arrays of type V2DF
15674     * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2
15675       -fsched2-use-traces
15676     * [408]21716 ICE in reg-stack.c's swap_rtx_condition
15677     * [409]24315 amd64 fails -fpeephole2
15678     __________________________________________________________________
15679
15680GCC 3.4.6
15681
15682   This is the [410]list of problem reports (PRs) from GCC's bug tracking
15683   system that are known to be fixed in the 3.4.6 release. This list might
15684   not be complete (that is, it is possible that some PRs that have been
15685   fixed are not listed here).
15686
15687
15688    For questions related to the use of GCC, please consult these web
15689    pages and the [411]GCC manuals. If that fails, the
15690    [412]gcc-help@gcc.gnu.org mailing list might help. Comments on these
15691    web pages and the development of GCC are welcome on our developer
15692    list at [413]gcc@gcc.gnu.org. All of [414]our lists have public
15693    archives.
15694
15695   Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and
15696   distribution of this entire article is permitted in any medium,
15697   provided this notice is preserved.
15698
15699   These pages are [416]maintained by the GCC team. Last modified
15700   2023-01-27.
15701
15702References
15703
15704   1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
15705   2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus
15706   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
15707   4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems
15708   5. http://gcc.gnu.org/gcc-3.4/mips-abi.html
15709   6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
15710   7. https://www.boost.org/
15711   8. https://gcc.gnu.org/PR11953
15712   9. https://gcc.gnu.org/PR8361
15713  10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins
15714  11. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209
15715  12. http://gcc.gnu.org/bugs/#cxx_rvalbind
15716  13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
15717  14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
15718  15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
15719  16. http://www.gnu.org/software/classpath/
15720  17. https://www.eclipse.org/
15721  18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html
15722  19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html
15723  20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
15724  21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html
15725  22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
15726  23. http://gcc.gnu.org/gcc-3.4/mips-abi.html
15727  24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html
15728  25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
15729  26. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=%5C%5B3%5C.4.*%5BRr%5Degression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED
15730  27. https://gcc.gnu.org/PR10129
15731  28. https://gcc.gnu.org/PR14576
15732  29. https://gcc.gnu.org/PR14760
15733  30. https://gcc.gnu.org/PR14671
15734  31. https://gcc.gnu.org/PR15093
15735  32. https://gcc.gnu.org/PR15178
15736  33. https://gcc.gnu.org/PR12753
15737  34. https://gcc.gnu.org/PR13985
15738  35. https://gcc.gnu.org/PR14810
15739  36. https://gcc.gnu.org/PR14883
15740  37. https://gcc.gnu.org/PR15044
15741  38. https://gcc.gnu.org/PR15057
15742  39. https://gcc.gnu.org/PR15064
15743  40. https://gcc.gnu.org/PR15142
15744  41. https://gcc.gnu.org/PR15159
15745  42. https://gcc.gnu.org/PR15165
15746  43. https://gcc.gnu.org/PR15193
15747  44. https://gcc.gnu.org/PR15209
15748  45. https://gcc.gnu.org/PR15227
15749  46. https://gcc.gnu.org/PR15285
15750  47. https://gcc.gnu.org/PR15299
15751  48. https://gcc.gnu.org/PR15329
15752  49. https://gcc.gnu.org/PR15550
15753  50. https://gcc.gnu.org/PR15554
15754  51. https://gcc.gnu.org/PR15640
15755  52. https://gcc.gnu.org/PR15666
15756  53. https://gcc.gnu.org/PR15696
15757  54. https://gcc.gnu.org/PR15701
15758  55. https://gcc.gnu.org/PR15761
15759  56. https://gcc.gnu.org/PR15829
15760  57. https://gcc.gnu.org/PR14538
15761  58. https://gcc.gnu.org/PR12391
15762  59. https://gcc.gnu.org/PR14649
15763  60. https://gcc.gnu.org/PR15004
15764  61. https://gcc.gnu.org/PR15749
15765  62. https://gcc.gnu.org/PR10646
15766  63. https://gcc.gnu.org/PR12077
15767  64. https://gcc.gnu.org/PR13598
15768  65. https://gcc.gnu.org/PR14211
15769  66. https://gcc.gnu.org/PR14220
15770  67. https://gcc.gnu.org/PR14245
15771  68. https://gcc.gnu.org/PR14340
15772  69. https://gcc.gnu.org/PR14600
15773  70. https://gcc.gnu.org/PR14668
15774  71. https://gcc.gnu.org/PR14775
15775  72. https://gcc.gnu.org/PR14821
15776  73. https://gcc.gnu.org/PR14930
15777  74. https://gcc.gnu.org/PR14932
15778  75. https://gcc.gnu.org/PR14950
15779  76. https://gcc.gnu.org/PR14962
15780  77. https://gcc.gnu.org/PR14975
15781  78. https://gcc.gnu.org/PR15002
15782  79. https://gcc.gnu.org/PR15025
15783  80. https://gcc.gnu.org/PR15046
15784  81. https://gcc.gnu.org/PR15069
15785  82. https://gcc.gnu.org/PR15074
15786  83. https://gcc.gnu.org/PR15083
15787  84. https://gcc.gnu.org/PR15096
15788  85. https://gcc.gnu.org/PR15287
15789  86. https://gcc.gnu.org/PR15317
15790  87. https://gcc.gnu.org/PR15337
15791  88. https://gcc.gnu.org/PR15361
15792  89. https://gcc.gnu.org/PR15412
15793  90. https://gcc.gnu.org/PR15427
15794  91. https://gcc.gnu.org/PR15471
15795  92. https://gcc.gnu.org/PR15503
15796  93. https://gcc.gnu.org/PR15507
15797  94. https://gcc.gnu.org/PR15542
15798  95. https://gcc.gnu.org/PR15565
15799  96. https://gcc.gnu.org/PR15625
15800  97. https://gcc.gnu.org/PR15629
15801  98. https://gcc.gnu.org/PR15742
15802  99. https://gcc.gnu.org/PR15775
15803 100. https://gcc.gnu.org/PR15821
15804 101. https://gcc.gnu.org/PR15862
15805 102. https://gcc.gnu.org/PR15875
15806 103. https://gcc.gnu.org/PR15877
15807 104. https://gcc.gnu.org/PR15947
15808 105. https://gcc.gnu.org/PR16020
15809 106. https://gcc.gnu.org/PR16154
15810 107. https://gcc.gnu.org/PR16174
15811 108. https://gcc.gnu.org/PR14315
15812 109. https://gcc.gnu.org/PR15151
15813 110. https://gcc.gnu.org/PR7993
15814 111. https://gcc.gnu.org/PR15228
15815 112. https://gcc.gnu.org/PR15345
15816 113. https://gcc.gnu.org/PR15945
15817 114. https://gcc.gnu.org/PR15526
15818 115. https://gcc.gnu.org/PR14690
15819 116. https://gcc.gnu.org/PR15112
15820 117. https://gcc.gnu.org/PR15067
15821 118. https://gcc.gnu.org/PR1963
15822 119. https://gcc.gnu.org/PR15717
15823 120. https://gcc.gnu.org/PR14782
15824 121. https://gcc.gnu.org/PR14828
15825 122. https://gcc.gnu.org/PR15202
15826 123. https://gcc.gnu.org/PR14610
15827 124. https://gcc.gnu.org/PR14813
15828 125. https://gcc.gnu.org/PR14857
15829 126. https://gcc.gnu.org/PR15598
15830 127. https://gcc.gnu.org/PR15653
15831 128. https://gcc.gnu.org/PR15189
15832 129. https://gcc.gnu.org/PR15331
15833 130. https://gcc.gnu.org/PR16144
15834 131. https://gcc.gnu.org/PR16176
15835 132. https://gcc.gnu.org/PR11591
15836 133. https://gcc.gnu.org/PR12028
15837 134. https://gcc.gnu.org/PR14478
15838 135. https://gcc.gnu.org/PR14567
15839 136. https://gcc.gnu.org/PR14715
15840 137. https://gcc.gnu.org/PR14902
15841 138. https://gcc.gnu.org/PR14924
15842 139. https://gcc.gnu.org/PR14960
15843 140. https://gcc.gnu.org/PR15106
15844 141. https://gcc.gnu.org/PR16026
15845 142. https://gcc.gnu.org/PR15191
15846 143. https://gcc.gnu.org/PR15662
15847 144. https://gcc.gnu.org/PR15054
15848 145. https://gcc.gnu.org/PR15783
15849 146. https://gcc.gnu.org/PR15626
15850 147. https://gcc.gnu.org/PR14326
15851 148. https://gcc.gnu.org/PR14723
15852 149. https://gcc.gnu.org/PR15290
15853 150. https://gcc.gnu.org/PR15250
15854 151. https://gcc.gnu.org/PR15551
15855 152. https://gcc.gnu.org/PR8309
15856 153. https://gcc.gnu.org/PR13250
15857 154. https://gcc.gnu.org/PR13803
15858 155. https://gcc.gnu.org/PR14093
15859 156. https://gcc.gnu.org/PR14457
15860 157. https://gcc.gnu.org/PR14542
15861 158. https://gcc.gnu.org/PR15100
15862 159. https://gcc.gnu.org/PR15296
15863 160. https://gcc.gnu.org/PR15396
15864 161. https://gcc.gnu.org/PR15782
15865 162. https://gcc.gnu.org/PR11610
15866 163. https://gcc.gnu.org/PR15488
15867 164. https://gcc.gnu.org/PR15489
15868 165. https://gcc.gnu.org/PR13928
15869 166. https://gcc.gnu.org/PR14150
15870 167. https://gcc.gnu.org/PR14949
15871 168. https://gcc.gnu.org/PR15123
15872 169. https://gcc.gnu.org/PR16469
15873 170. https://gcc.gnu.org/PR16344
15874 171. https://gcc.gnu.org/PR16842
15875 172. https://gcc.gnu.org/PR12608
15876 173. https://gcc.gnu.org/PR14492
15877 174. https://gcc.gnu.org/PR15461
15878 175. https://gcc.gnu.org/PR15890
15879 176. https://gcc.gnu.org/PR16180
15880 177. https://gcc.gnu.org/PR16224
15881 178. https://gcc.gnu.org/PR16408
15882 179. https://gcc.gnu.org/PR16529
15883 180. https://gcc.gnu.org/PR16698
15884 181. https://gcc.gnu.org/PR16706
15885 182. https://gcc.gnu.org/PR16810
15886 183. https://gcc.gnu.org/PR16851
15887 184. https://gcc.gnu.org/PR16870
15888 185. https://gcc.gnu.org/PR16904
15889 186. https://gcc.gnu.org/PR16905
15890 187. https://gcc.gnu.org/PR16964
15891 188. https://gcc.gnu.org/PR17068
15892 189. https://gcc.gnu.org/PR16366
15893 190. https://gcc.gnu.org/PR15345
15894 191. https://gcc.gnu.org/PR16590
15895 192. https://gcc.gnu.org/PR16693
15896 193. https://gcc.gnu.org/PR17078
15897 194. https://gcc.gnu.org/PR13956
15898 195. https://gcc.gnu.org/PR16684
15899 196. https://gcc.gnu.org/PR12658
15900 197. https://gcc.gnu.org/PR13092
15901 198. https://gcc.gnu.org/PR15320
15902 199. https://gcc.gnu.org/PR16246
15903 200. https://gcc.gnu.org/PR16273
15904 201. https://gcc.gnu.org/PR16401
15905 202. https://gcc.gnu.org/PR16411
15906 203. https://gcc.gnu.org/PR16489
15907 204. https://gcc.gnu.org/PR16618
15908 205. https://gcc.gnu.org/PR16637
15909 206. https://gcc.gnu.org/PR16717
15910 207. https://gcc.gnu.org/PR16813
15911 208. https://gcc.gnu.org/PR16853
15912 209. https://gcc.gnu.org/PR16889
15913 210. https://gcc.gnu.org/PR16959
15914 211. https://gcc.gnu.org/PR7587
15915 212. https://gcc.gnu.org/PR16473
15916 213. https://gcc.gnu.org/PR16478
15917 214. https://gcc.gnu.org/PR10695
15918 215. https://gcc.gnu.org/PR16974
15919 216. https://gcc.gnu.org/PR16298
15920 217. https://gcc.gnu.org/PR17113
15921 218. https://gcc.gnu.org/PR14697
15922 219. https://gcc.gnu.org/PR15869
15923 220. https://gcc.gnu.org/PR16325
15924 221. https://gcc.gnu.org/PR16357
15925 222. https://gcc.gnu.org/PR16380
15926 223. https://gcc.gnu.org/PR16407
15927 224. https://gcc.gnu.org/PR16643
15928 225. https://gcc.gnu.org/PR15927
15929 226. https://gcc.gnu.org/PR15948
15930 227. https://gcc.gnu.org/PR17019
15931 228. https://gcc.gnu.org/PR16130
15932 229. https://gcc.gnu.org/PR16142
15933 230. https://gcc.gnu.org/PR16278
15934 231. https://gcc.gnu.org/PR16414
15935 232. https://gcc.gnu.org/PR16445
15936 233. https://gcc.gnu.org/PR16490
15937 234. https://gcc.gnu.org/PR16683
15938 235. https://gcc.gnu.org/PR16195
15939 236. https://gcc.gnu.org/PR16239
15940 237. https://gcc.gnu.org/PR16199
15941 238. https://gcc.gnu.org/PR16416
15942 239. https://gcc.gnu.org/PR16430
15943 240. https://gcc.gnu.org/PR16379
15944 241. https://gcc.gnu.org/PR17093
15945 242. https://gcc.gnu.org/PR17119
15946 243. https://gcc.gnu.org/PR15928
15947 244. https://gcc.gnu.org/PR16210
15948 245. https://gcc.gnu.org/PR15488
15949 246. https://gcc.gnu.org/PR16250
15950 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3
15951 248. https://gcc.gnu.org/PR17369
15952 249. https://gcc.gnu.org/PR17850
15953 250. https://gcc.gnu.org/PR13948
15954 251. https://gcc.gnu.org/PR14492
15955 252. https://gcc.gnu.org/PR16301
15956 253. https://gcc.gnu.org/PR16566
15957 254. https://gcc.gnu.org/PR17023
15958 255. https://gcc.gnu.org/PR17027
15959 256. https://gcc.gnu.org/PR17524
15960 257. https://gcc.gnu.org/PR17826
15961 258. https://gcc.gnu.org/PR15526
15962 259. https://gcc.gnu.org/PR16999
15963 260. https://gcc.gnu.org/PR17503
15964 261. https://gcc.gnu.org/PR17581
15965 262. https://gcc.gnu.org/PR18129
15966 263. https://gcc.gnu.org/PR10975
15967 264. https://gcc.gnu.org/PR11722
15968 265. https://gcc.gnu.org/PR14534
15969 266. https://gcc.gnu.org/PR15172
15970 267. https://gcc.gnu.org/PR15786
15971 268. https://gcc.gnu.org/PR16162
15972 269. https://gcc.gnu.org/PR16612
15973 270. https://gcc.gnu.org/PR16715
15974 271. https://gcc.gnu.org/PR16848
15975 272. https://gcc.gnu.org/PR17132
15976 273. https://gcc.gnu.org/PR17259
15977 274. https://gcc.gnu.org/PR17327
15978 275. https://gcc.gnu.org/PR17393
15979 276. https://gcc.gnu.org/PR17501
15980 277. https://gcc.gnu.org/PR17537
15981 278. https://gcc.gnu.org/PR17585
15982 279. https://gcc.gnu.org/PR17821
15983 280. https://gcc.gnu.org/PR17829
15984 281. https://gcc.gnu.org/PR17851
15985 282. https://gcc.gnu.org/PR17976
15986 283. https://gcc.gnu.org/PR18020
15987 284. https://gcc.gnu.org/PR18093
15988 285. https://gcc.gnu.org/PR18140
15989 286. https://gcc.gnu.org/PR17541
15990 287. https://gcc.gnu.org/PR17853
15991 288. https://gcc.gnu.org/PR17245
15992 289. https://gcc.gnu.org/PR17167
15993 290. https://gcc.gnu.org/PR17277
15994 291. https://gcc.gnu.org/PR17505
15995 292. https://gcc.gnu.org/PR17684
15996 293. https://gcc.gnu.org/PR17384
15997 294. https://gcc.gnu.org/PR17770
15998 295. https://gcc.gnu.org/PR11476
15999 296. https://gcc.gnu.org/PR14064
16000 297. https://gcc.gnu.org/PR14678
16001 298. https://gcc.gnu.org/PR15583
16002 299. https://gcc.gnu.org/PR15790
16003 300. https://gcc.gnu.org/PR15886
16004 301. https://gcc.gnu.org/PR16884
16005 302. https://gcc.gnu.org/PR13841
16006 303. https://gcc.gnu.org/PR15860
16007 304. https://gcc.gnu.org/PR17465
16008 305. https://gcc.gnu.org/PR17469
16009 306. https://gcc.gnu.org/PR18138
16010 307. https://gcc.gnu.org/PR15498
16011 308. https://gcc.gnu.org/PR15747
16012 309. https://gcc.gnu.org/PR16406
16013 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4
16014 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5
16015 312. https://gcc.gnu.org/PR24688
16016 313. https://gcc.gnu.org/PR17188
16017 314. https://gcc.gnu.org/PR20187
16018 315. https://gcc.gnu.org/PR21873
16019 316. https://gcc.gnu.org/PR21899
16020 317. https://gcc.gnu.org/PR22061
16021 318. https://gcc.gnu.org/PR22208
16022 319. https://gcc.gnu.org/PR22458
16023 320. https://gcc.gnu.org/PR22589
16024 321. https://gcc.gnu.org/PR24101
16025 322. https://gcc.gnu.org/PR10611
16026 323. https://gcc.gnu.org/PR13377
16027 324. https://gcc.gnu.org/PR16002
16028 325. https://gcc.gnu.org/PR17413
16029 326. https://gcc.gnu.org/PR17609
16030 327. https://gcc.gnu.org/PR17618
16031 328. https://gcc.gnu.org/PR18124
16032 329. https://gcc.gnu.org/PR18155
16033 330. https://gcc.gnu.org/PR18177
16034 331. https://gcc.gnu.org/PR18368
16035 332. https://gcc.gnu.org/PR18378
16036 333. https://gcc.gnu.org/PR18466
16037 334. https://gcc.gnu.org/PR18512
16038 335. https://gcc.gnu.org/PR18545
16039 336. https://gcc.gnu.org/PR18738
16040 337. https://gcc.gnu.org/PR18803
16041 338. https://gcc.gnu.org/PR19004
16042 339. https://gcc.gnu.org/PR19208
16043 340. https://gcc.gnu.org/PR19253
16044 341. https://gcc.gnu.org/PR19608
16045 342. https://gcc.gnu.org/PR19884
16046 343. https://gcc.gnu.org/PR20153
16047 344. https://gcc.gnu.org/PR20563
16048 345. https://gcc.gnu.org/PR20789
16049 346. https://gcc.gnu.org/PR21336
16050 347. https://gcc.gnu.org/PR21768
16051 348. https://gcc.gnu.org/PR21853
16052 349. https://gcc.gnu.org/PR21903
16053 350. https://gcc.gnu.org/PR21983
16054 351. https://gcc.gnu.org/PR21987
16055 352. https://gcc.gnu.org/PR22153
16056 353. https://gcc.gnu.org/PR22172
16057 354. https://gcc.gnu.org/PR21286
16058 355. https://gcc.gnu.org/PR22233
16059 356. https://gcc.gnu.org/PR22508
16060 357. https://gcc.gnu.org/PR22545
16061 358. https://gcc.gnu.org/PR23528
16062 359. https://gcc.gnu.org/PR23550
16063 360. https://gcc.gnu.org/PR23586
16064 361. https://gcc.gnu.org/PR23624
16065 362. https://gcc.gnu.org/PR23639
16066 363. https://gcc.gnu.org/PR23797
16067 364. https://gcc.gnu.org/PR23965
16068 365. https://gcc.gnu.org/PR24052
16069 366. https://gcc.gnu.org/PR24580
16070 367. https://gcc.gnu.org/PR24267
16071 368. https://gcc.gnu.org/PR17810
16072 369. https://gcc.gnu.org/PR17860
16073 370. https://gcc.gnu.org/PR21709
16074 371. https://gcc.gnu.org/PR21964
16075 372. https://gcc.gnu.org/PR22167
16076 373. https://gcc.gnu.org/PR22619
16077 374. https://gcc.gnu.org/PR23241
16078 375. https://gcc.gnu.org/PR23478
16079 376. https://gcc.gnu.org/PR24470
16080 377. https://gcc.gnu.org/PR24950
16081 378. https://gcc.gnu.org/PR14400
16082 379. https://gcc.gnu.org/PR14940
16083 380. https://gcc.gnu.org/PR20239
16084 381. https://gcc.gnu.org/PR15220
16085 382. https://gcc.gnu.org/PR19275
16086 383. https://gcc.gnu.org/PR21888
16087 384. https://gcc.gnu.org/PR15342
16088 385. https://gcc.gnu.org/PR23985
16089 386. https://gcc.gnu.org/PR16719
16090 387. https://gcc.gnu.org/PR21723
16091 388. https://gcc.gnu.org/PR21841
16092 389. https://gcc.gnu.org/PR23644
16093 390. https://gcc.gnu.org/PR24718
16094 391. https://gcc.gnu.org/PR18421
16095 392. https://gcc.gnu.org/PR20621
16096 393. https://gcc.gnu.org/PR18583
16097 394. https://gcc.gnu.org/PR20191
16098 395. https://gcc.gnu.org/PR22083
16099 396. https://gcc.gnu.org/PR23070
16100 397. https://gcc.gnu.org/PR23404
16101 398. https://gcc.gnu.org/PR23539
16102 399. https://gcc.gnu.org/PR24102
16103 400. https://gcc.gnu.org/PR24465
16104 401. https://gcc.gnu.org/PR19933
16105 402. https://gcc.gnu.org/PR21889
16106 403. https://gcc.gnu.org/PR19300
16107 404. https://gcc.gnu.org/PR20301
16108 405. https://gcc.gnu.org/PR20673
16109 406. https://gcc.gnu.org/PR18582
16110 407. https://gcc.gnu.org/PR19340
16111 408. https://gcc.gnu.org/PR21716
16112 409. https://gcc.gnu.org/PR24315
16113 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6
16114 411. https://gcc.gnu.org/onlinedocs/
16115 412. mailto:gcc-help@gcc.gnu.org
16116 413. mailto:gcc@gcc.gnu.org
16117 414. https://gcc.gnu.org/lists.html
16118 415. https://www.fsf.org/
16119 416. https://gcc.gnu.org/about.html
16120======================================================================
16121http://gcc.gnu.org/gcc-3.3/index.html
16122
16123                             GCC 3.3 Release Series
16124
16125   (This release series is no longer supported.)
16126
16127   May 03, 2005
16128
16129   The [1]GNU project and the GCC developers are pleased to announce the
16130   release of GCC 3.3.6.
16131
16132   This release is a bug-fix release, containing fixes for regressions in
16133   GCC 3.3.5 relative to previous releases of GCC.
16134
16135   This release is the last of the series 3.3.x.
16136
16137   The GCC 3.3 release series includes numerous [2]new features,
16138   improvements, bug fixes, and other changes, thanks to an [3]amazing
16139   group of volunteers.
16140
16141Release History
16142
16143   GCC 3.3.6
16144          May 3, 2005 ([4]changes)
16145
16146   GCC 3.3.5
16147          September 30, 2004 ([5]changes)
16148
16149   GCC 3.3.4
16150          May 31, 2004 ([6]changes)
16151
16152   GCC 3.3.3
16153          February 14, 2004 ([7]changes)
16154
16155   GCC 3.3.2
16156          October 16, 2003 ([8]changes)
16157
16158   GCC 3.3.1
16159          August 8, 2003 ([9]changes)
16160
16161   GCC 3.3
16162          May 14, 2003 ([10]changes)
16163
16164References and Acknowledgements
16165
16166   GCC used to stand for the GNU C Compiler, but since the compiler
16167   supports several other languages aside from C, it now stands for the
16168   GNU Compiler Collection.
16169
16170   A list of [11]successful builds is updated as new information becomes
16171   available.
16172
16173   The GCC developers would like to thank the numerous people that have
16174   contributed new features, improvements, bug fixes, and other changes as
16175   well as test results to GCC. This [12]amazing group of volunteers is
16176   what makes GCC successful.
16177
16178   For additional information about GCC please refer to the [13]GCC
16179   project web site or contact the [14]GCC development mailing list.
16180
16181   To obtain GCC please use [15]our mirror sites, or our CVS server.
16182
16183
16184    For questions related to the use of GCC, please consult these web
16185    pages and the [16]GCC manuals. If that fails, the
16186    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
16187    web pages and the development of GCC are welcome on our developer
16188    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
16189    archives.
16190
16191   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
16192   distribution of this entire article is permitted in any medium,
16193   provided this notice is preserved.
16194
16195   These pages are [21]maintained by the GCC team. Last modified
16196   2022-10-26.
16197
16198References
16199
16200   1. http://www.gnu.org/
16201   2. http://gcc.gnu.org/gcc-3.3/changes.html
16202   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
16203   4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
16204   5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5
16205   6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4
16206   7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3
16207   8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2
16208   9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1
16209  10. http://gcc.gnu.org/gcc-3.3/changes.html
16210  11. http://gcc.gnu.org/gcc-3.3/buildstat.html
16211  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
16212  13. http://gcc.gnu.org/index.html
16213  14. mailto:gcc@gcc.gnu.org
16214  15. http://gcc.gnu.org/mirrors.html
16215  16. https://gcc.gnu.org/onlinedocs/
16216  17. mailto:gcc-help@gcc.gnu.org
16217  18. mailto:gcc@gcc.gnu.org
16218  19. https://gcc.gnu.org/lists.html
16219  20. https://www.fsf.org/
16220  21. https://gcc.gnu.org/about.html
16221======================================================================
16222http://gcc.gnu.org/gcc-3.3/changes.html
16223
16224                             GCC 3.3 Release Series
16225                        Changes, New Features, and Fixes
16226
16227   The latest release in the 3.3 release series is [1]GCC 3.3.6.
16228
16229Caveats
16230
16231     * The preprocessor no longer accepts multi-line string literals. They
16232       were deprecated in 3.0, 3.1, and 3.2.
16233     * The preprocessor no longer supports the -A- switch when appearing
16234       alone. -A- followed by an assertion is still supported.
16235     * Support for all the systems [2]obsoleted in GCC 3.1 has been
16236       removed from GCC 3.3. See below for a [3]list of systems which are
16237       obsoleted in this release.
16238     * Checking for null format arguments has been decoupled from the rest
16239       of the format checking mechanism. Programs which use the format
16240       attribute may regain this functionality by using the new [4]nonnull
16241       function attribute. Note that all functions for which GCC has a
16242       built-in format attribute, an appropriate built-in nonnull
16243       attribute is also applied.
16244     * The DWARF (version 1) debugging format has been deprecated and will
16245       be removed in a future version of GCC. Version 2 of the DWARF
16246       debugging format will continue to be supported for the foreseeable
16247       future.
16248     * The C and Objective-C compilers no longer accept the "Naming Types"
16249       extension (typedef foo = bar); it was already unavailable in C++.
16250       Code which uses it will need to be changed to use the "typeof"
16251       extension instead: typedef typeof(bar) foo. (We have removed this
16252       extension without a period of deprecation because it has caused the
16253       compiler to crash since version 3.0 and no one noticed until very
16254       recently. Thus we conclude it is not in widespread use.)
16255     * The -traditional C compiler option has been removed. It was
16256       deprecated in 3.1 and 3.2. (Traditional preprocessing remains
16257       available.) The <varargs.h> header, used for writing variadic
16258       functions in traditional C, still exists but will produce an error
16259       message if used.
16260     * GCC 3.3.1 automatically places zero-initialized variables in the
16261       .bss section on some operating systems. Versions of GNU Emacs up to
16262       (and including) 21.3 will not work correctly when using this
16263       optimization; you can use -fno-zero-initialized-in-bss to disable
16264       it.
16265
16266General Optimizer Improvements
16267
16268     * A new scheme for accurately describing processor pipelines, the
16269       [5]DFA scheduler, has been added.
16270     * Pavel Nejedly, Charles University Prague, has contributed new file
16271       format used by the edge coverage profiler (-fprofile-arcs).
16272       The new format is robust and diagnoses common mistakes where
16273       profiles from different versions (or compilations) of the program
16274       are combined resulting in nonsensical profiles and slow code to
16275       produced with profile feedback. Additionally this format allows
16276       extra data to be gathered. Currently, overall statistics are
16277       produced helping optimizers to identify hot spots of a program
16278       globally replacing the old intra-procedural scheme and resulting in
16279       better code. Note that the gcov tool from older GCC versions will
16280       not be able to parse the profiles generated by GCC 3.3 and vice
16281       versa.
16282     * Jan Hubicka, SuSE Labs, has contributed a new superblock formation
16283       pass enabled using -ftracer. This pass simplifies the control flow
16284       of functions allowing other optimizations to do better job.
16285       He also contributed the function reordering pass
16286       (-freorder-functions) to optimize function placement using profile
16287       feedback.
16288
16289New Languages and Language specific improvements
16290
16291  C/ObjC/C++
16292
16293     * The preprocessor now accepts directives within macro arguments. It
16294       processes them just as if they had not been within macro arguments.
16295     * The separate ISO and traditional preprocessors have been completely
16296       removed. The front end handles either type of preprocessed output
16297       if necessary.
16298     * In C99 mode preprocessor arithmetic is done in the precision of the
16299       target's intmax_t, as required by that standard.
16300     * The preprocessor can now copy comments inside macros to the output
16301       file when the macro is expanded. This feature, enabled using the
16302       -CC option, is intended for use by applications which place
16303       metadata or directives inside comments, such as lint.
16304     * The method of constructing the list of directories to be searched
16305       for header files has been revised. If a directory named by a -I
16306       option is a standard system include directory, the option is
16307       ignored to ensure that the default search order for system
16308       directories and the special treatment of system header files are
16309       not defeated.
16310     * A few more [6]ISO C99 features now work correctly.
16311     * A new function attribute, nonnull, has been added which allows
16312       pointer arguments to functions to be specified as requiring a
16313       non-null value. The compiler currently uses this information to
16314       issue a warning when it detects a null value passed in such an
16315       argument slot.
16316     * A new type attribute, may_alias, has been added. Accesses to
16317       objects with types with this attribute are not subjected to
16318       type-based alias analysis, but are instead assumed to be able to
16319       alias any other type of objects, just like the char type.
16320
16321  C++
16322
16323     * Type based alias analysis has been implemented for C++ aggregate
16324       types.
16325
16326  Objective-C
16327
16328     * Generate an error if Objective-C objects are passed by value in
16329       function and method calls.
16330     * When -Wselector is used, check the whole list of selectors at the
16331       end of compilation, and emit a warning if a @selector() is not
16332       known.
16333     * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
16334     * No longer need to include objc/objc-class.h to compile self calls
16335       in class methods (NeXT runtime only).
16336     * New -Wundeclared-selector option.
16337     * Removed selector bloating which was causing object files to be 10%
16338       bigger on average (GNU runtime only).
16339     * Using at run time @protocol() objects has been fixed in certain
16340       situations (GNU runtime only).
16341     * Type checking has been fixed and improved in many situations
16342       involving protocols.
16343
16344  Java
16345
16346     * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK
16347       1.4) API.
16348     * The JDK 1.4 assert facility has been implemented.
16349     * The bytecode interpreter is now direct threaded and thus faster.
16350
16351  Fortran
16352
16353     * Fortran improvements are listed in [7]the Fortran documentation.
16354
16355  Ada
16356
16357     * Ada tasking now works with glibc 2.3.x threading libraries.
16358
16359New Targets and Target Specific Improvements
16360
16361     * The following changes have been made to the HP-PA port:
16362          + The port now defaults to scheduling for the PA8000 series of
16363            processors.
16364          + Scheduling support for the PA7300 processor has been added.
16365          + The 32-bit port now supports weak symbols under HP-UX 11.
16366          + The handling of initializers and finalizers has been improved
16367            under HP-UX 11. The 64-bit port no longer uses collect2.
16368          + Dwarf2 EH support has been added to the 32-bit GNU/Linux port.
16369          + ABI fixes to correct the passing of small structures by value.
16370     * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to
16371       use the DFA processor pipeline description.
16372     * The following NetBSD configurations for the SuperH processor family
16373       have been added:
16374          + SH3, big-endian, sh-*-netbsdelf*
16375          + SH3, little-endian, shle-*-netbsdelf*
16376          + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
16377          + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
16378          + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
16379          + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
16380     * The following changes have been made to the IA-32/x86-64 port:
16381          + SSE2 and 3dNOW! intrinsics are now supported.
16382          + Support for thread local storage has been added to the IA-32
16383            and x86-64 ports.
16384          + The x86-64 port has been significantly improved.
16385     * The following changes have been made to the MIPS port:
16386          + All configurations now accept the -mabi switch. Note that you
16387            will need appropriate multilibs for this option to work
16388            properly.
16389          + ELF configurations will always pass an ABI flag to the
16390            assembler, except when the MIPS EABI is selected.
16391          + -mabi=64 no longer selects MIPS IV code.
16392          + The -mcpu option, which was deprecated in 3.1 and 3.2, has
16393            been removed from this release.
16394          + -march now changes the core ISA level. In previous releases,
16395            it would change the use of processor-specific extensions, but
16396            would leave the core ISA unchanged. For example, mips64-elf
16397            -march=r8000 will now generate MIPS IV code.
16398          + Under most configurations, -mipsN now acts as a synonym for
16399            -march.
16400          + There are some new preprocessor macros to describe the -march
16401            and -mtune settings. See the documentation of those options
16402            for details.
16403          + Support for the NEC VR-Series processors has been added. This
16404            includes the 54xx, 5500, and 41xx series.
16405          + Support for the Sandcraft sr71k processor has been added.
16406     * The following changes have been made to the S/390 port:
16407          + Support to build the Java runtime libraries has been added.
16408            Java is now enabled by default on s390-*-linux* and
16409            s390x-*-linux* targets.
16410          + Multilib support for the s390x-*-linux* target has been added;
16411            this allows to build 31-bit binaries using the -m31 option.
16412          + Support for thread local storage has been added.
16413          + Inline assembler code may now use the 'Q' constraint to
16414            specify memory operands without index register.
16415          + Various platform-specific performance improvements have been
16416            implemented; in particular, the compiler now uses the BRANCH
16417            ON COUNT family of instructions and makes more frequent use of
16418            the TEST UNDER MASK family of instructions.
16419     * The following changes have been made to the PowerPC port:
16420          + Support for IBM Power4 processor added.
16421          + Support for Motorola e500 SPE added.
16422          + Support for AIX 5.2 added.
16423          + Function and Data sections now supported on AIX.
16424          + Sibcall optimizations added.
16425     * The support for H8 Tiny is added to the H8/300 port with -mn.
16426
16427Obsolete Systems
16428
16429   Support for a number of older systems has been declared obsolete in GCC
16430   3.3. Unless there is activity to revive them, the next release of GCC
16431   will have their sources permanently removed.
16432
16433   All configurations of the following processor architectures have been
16434   declared obsolete:
16435     * Matsushita MN10200, mn10200-*-*
16436     * Motorola 88000, m88k-*-*
16437     * IBM ROMP, romp-*-*
16438
16439   Also, some individual systems have been obsoleted:
16440     * Alpha
16441          + Interix, alpha*-*-interix*
16442          + Linux libc1, alpha*-*-linux*libc1*
16443          + Linux ECOFF, alpha*-*-linux*ecoff*
16444     * ARM
16445          + Generic a.out, arm*-*-aout*
16446          + Conix, arm*-*-conix*
16447          + "Old ABI," arm*-*-oabi
16448          + StrongARM/COFF, strongarm-*-coff*
16449     * HPPA (PA-RISC)
16450          + Generic OSF, hppa1.0-*-osf*
16451          + Generic BSD, hppa1.0-*-bsd*
16452          + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
16453          + HiUX, hppa*-*-hiux*
16454          + Mach Lites, hppa*-*-lites*
16455     * Intel 386 family
16456          + Windows NT 3.x, i?86-*-win32
16457     * MC68000 family
16458          + HP systems, m68000-hp-bsd* and m68k-hp-bsd*
16459          + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and
16460            m68k-sun-mach*
16461          + AT&T systems, m68000-att-sysv*
16462          + Atari systems, m68k-atari-sysv*
16463          + Motorola systems, m68k-motorola-sysv*
16464          + NCR systems, m68k-ncr-sysv*
16465          + Plexus systems, m68k-plexus-sysv*
16466          + Commodore systems, m68k-cbm-sysv*
16467          + Citicorp TTI, m68k-tti-*
16468          + Unos, m68k-crds-unos*
16469          + Concurrent RTU, m68k-ccur-rtu*
16470          + Linux a.out, m68k-*-linux*aout*
16471          + Linux libc1, m68k-*-linux*libc1*
16472          + pSOS, m68k-*-psos*
16473     * MIPS
16474          + Generic ECOFF, mips*-*-ecoff*
16475          + SINIX, mips-sni-sysv4
16476          + Orion RTEMS, mips64orion-*-rtems*
16477     * National Semiconductor 32000
16478          + OpenBSD, ns32k-*-openbsd*
16479     * POWER (aka RS/6000) and PowerPC
16480          + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
16481          + Bull BOSX, rs6000-bull-bosx
16482          + Generic Mach, rs6000-*-mach*
16483          + Generic SysV, powerpc*-*-sysv*
16484          + Linux libc1, powerpc*-*-linux*libc1*
16485     * Sun SPARC
16486          + Generic a.out, sparc-*-aout*, sparclet-*-aout*,
16487            sparclite-*-aout*, and sparc86x-*-aout*
16488          + NetBSD a.out, sparc-*-netbsd*aout*
16489          + Generic BSD, sparc-*-bsd*
16490          + ChorusOS, sparc-*-chorusos*
16491          + Linux a.out, sparc-*-linux*aout*
16492          + Linux libc1, sparc-*-linux*libc1*
16493          + LynxOS, sparc-*-lynxos*
16494          + Solaris on HAL hardware, sparc-hal-solaris2*
16495          + SunOS versions 3 and 4, sparc-*-sunos[34]*
16496     * NEC V850
16497          + RTEMS, v850-*-rtems*
16498     * VAX
16499          + VMS, vax-*-vms*
16500
16501Documentation improvements
16502
16503Other significant improvements
16504
16505     * Almost all front-end dependencies in the compiler have been
16506       separated out into a set of language hooks. This should make adding
16507       a new front end clearer and easier.
16508     * One effect of removing the separate preprocessor is a small
16509       increase in the robustness of the compiler in general, and the
16510       maintainability of target descriptions. Previously target-specific
16511       built-in macros and others, such as __FAST_MATH__, had to be
16512       handled with so-called specs that were hard to maintain. Often they
16513       would fail to behave properly when conflicting options were
16514       supplied on the command line, and define macros in the user's
16515       namespace even when strict ISO compliance was requested.
16516       Integrating the preprocessor has cleanly solved these issues.
16517     * The Makefile suite now supports redirection of make install by
16518       means of the variable DESTDIR.
16519     __________________________________________________________________
16520
16521GCC 3.3
16522
16523   Detailed release notes for the GCC 3.3 release follow.
16524
16525  Bug Fixes
16526
16527    bootstrap failures
16528
16529     * [8]10140 cross compiler build failures: missing __mempcpy (DUP:
16530       [9]10198,[10]10338)
16531
16532    Internal compiler errors (multi-platform)
16533
16534     * [11]3581 large string causes segmentation fault in cc1
16535     * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler
16536     * [13]5533 (c++) ICE when processing std::accumulate(begin, end,
16537       init, invalid_op)
16538     * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
16539     * [15]6412 (c++) ICE in retrieve_specialization
16540     * [16]6620 (c++) partial template specialization causes an ICE
16541       (segmentation fault)
16542     * [17]6663 (c++) ICE with attribute aligned
16543     * [18]7068 ICE with incomplete types
16544     * [19]7083 (c++) ICE using -gstabs with dodgy class derivation
16545     * [20]7647 (c++) ICE when data member has the name of the enclosing
16546       class
16547     * [21]7675 ICE in fixup_var_refs_1
16548     * [22]7718 'complex' template instantiation causes ICE
16549     * [23]8116 (c++) ICE in member template function
16550     * [24]8358 (ada) Ada compiler accesses freed memory, crashes
16551     * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation
16552       fault
16553     * [26]8564 (c++) ICE in find_function_data, in function.c
16554     * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
16555     * [28]8766 (c++) ICE after failed initialization of static template
16556       variable
16557     * [29]8803 ICE in instantiate_virtual_regs_1, in function.c
16558     * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
16559     * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class
16560       definition
16561     * [32]9216 (c++) ICE on missing template parameter
16562     * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c
16563     * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO
16564       loop
16565     * [35]9429 (c++) ICE in template instantiation with a pointered new
16566       operator
16567     * [36]9516 Internal error when using a big array
16568     * [37]9600 (c++) ICE with typedefs in template class
16569     * [38]9629 (c++) virtual inheritance segfault
16570     * [39]9672 (c++) ICE: Error reporting routines re-entered
16571     * [40]9749 (c++) ICE in write_expression on invalid function
16572       prototype
16573     * [41]9794 (fortran) ICE: floating point exception during constant
16574       folding
16575     * [42]9829 (c++) Missing colon in nested namespace usage causes ICE
16576     * [43]9916 (c++) ICE with noreturn function in ?: statement
16577     * [44]9936 ICE with local function and variable-length 2d array
16578     * [45]10262 (c++) cc1plus crashes with large generated code
16579     * [46]10278 (c++) ICE in parser for invalid code
16580     * [47]10446 (c++) ICE on definition of nonexistent member function of
16581       nested class in a class template
16582     * [48]10451 (c++) ICE in grokdeclarator on spurious mutable
16583       declaration
16584     * [49]10506 (c++) ICE in build_new at cp/init.c with
16585       -fkeep-inline-functions and multiple inheritance
16586     * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the
16587       precision of the declared type
16588
16589    Optimization bugs
16590
16591     * [51]2001 Inordinately long compile times in reload CSE regs
16592     * [52]2391 Exponential compilation time explosion in combine
16593     * [53]2960 Duplicate loop conditions even with -Os
16594     * [54]4046 redundant conditional branch
16595     * [55]6405 Loop-unrolling related performance regressions
16596     * [56]6798 very long compile time with large case-statement
16597     * [57]6871 const objects shouldn't be moved to .bss
16598     * [58]6909 problem w/ -Os on modified loop-2c.c test case
16599     * [59]7189 gcc -O2 -Wall does not print ``control reaches end of
16600       non-void function'' warning
16601     * [60]7642 optimization problem with signbit()
16602     * [61]8634 incorrect code for inlining of memcpy under -O2
16603     * [62]8750 Cygwin prolog generation erroneously emitting __alloca as
16604       regular function call
16605
16606    C front end
16607
16608     * [63]2161 long if-else cascade overflows parser stack
16609     * [64]4319 short accepted on typedef'd char
16610     * [65]8602 incorrect line numbers in warning messages when using
16611       inline functions
16612     * [66]9177 -fdump-translation-unit: C front end deletes function_decl
16613       AST nodes and breaks debugging dumps
16614     * [67]9853 miscompilation of non-constant structure initializer
16615
16616    c++ compiler and library
16617
16618     * [68]45 legal template specialization code is rejected (DUP:
16619       [69]3784)
16620     * [70]764 lookup failure: friend operator and dereferencing a pointer
16621       and templates (DUP: [71]5116)
16622     * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP:
16623       2863)
16624     * [73]3663 G++ doesn't check access control during template
16625       instantiation
16626     * [74]3797 gcc fails to emit explicit specialization of a template
16627       member
16628     * [75]3948 Two destructors are called when no copy destructor is
16629       defined (ABI change)
16630     * [76]4137 Conversion operator within template is not accepted
16631     * [77]4361 bogus ambiguity taking the address of a member template
16632     * [78]4802 g++ accepts illegal template code (access to private
16633       member; DUP: [79]5837)
16634     * [80]4803 inline function is used but never defined, and g++ does
16635       not object
16636     * [81]5094 Partial specialization cannot be friend?
16637     * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
16638     * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run
16639       time
16640     * [84]7015 certain __asm__ constructs rejected
16641     * [85]7086 compile time regression (quadratic behavior in
16642       fixup_var_refs)
16643     * [86]7099 G++ doesn't set the noreturn attribute on std::exit and
16644       std::abort
16645     * [87]7247 copy constructor missing when inlining enabled (invalid
16646       optimization?)
16647     * [88]7441 string array initialization compilation time regression
16648       from seconds to minutes
16649     * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong
16650     * [90]7804 bad printing of floating point constant in warning message
16651     * [91]8099 Friend classes and template specializations
16652     * [92]8117 member function pointers and multiple inheritance
16653     * [93]8205 using declaration and multiple inheritance
16654     * [94]8645 unnecessary non-zero checks in stl_tree.h
16655     * [95]8724 explicit destructor call for incomplete class allowed
16656     * [96]8805 compile time regression with many member variables
16657     * [97]8691 -O3 and -fno-implicit-templates are incompatible
16658     * [98]8700 unhelpful error message for binding temp to reference
16659     * [99]8724 explicit destructor call for incomplete class allowed
16660     * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems
16661     * [101]9016 Failure to consistently constant fold "constant" C++
16662       objects
16663     * [102]9053 g++ confused about ambiguity of overloaded function
16664       templates
16665     * [103]9152 undefined virtual thunks
16666     * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out
16667     * [105]9297 data corruption due to codegen bug (when copying.)
16668     * [106]9318 i/ostream::operator>>/<<(streambuf*) broken
16669     * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf
16670     * [108]9400 bogus -Wshadow warning: shadowed declaration of this in
16671       local classes
16672     * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters
16673     * [110]9425 filebuf::pbackfail broken (DUP: [111]9439)
16674     * [112]9474 GCC freezes in compiling a weird code mixing <iostream>
16675       and <iostream.h>
16676     * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1)
16677       [114][DR 231]
16678     * [115]9555 ostream inserters fail to set badbit on exception
16679     * [116]9561 ostream inserters rethrow exception of wrong type
16680     * [117]9563 ostream::sentry returns true after a failed preparation
16681     * [118]9582 one-definition rule violation in std::allocator
16682     * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors
16683     * [120]9683 bug in initialization chains for static const variables
16684       from template classes
16685     * [121]9791 -Woverloaded-virtual reports hiding of destructor
16686     * [122]9817 collate::compare doesn't handle nul characters
16687     * [123]9825 filebuf::sputbackc breaks sbumpc
16688     * [124]9826 operator>>(basic_istream, basic_string) fails to compile
16689       with custom traits
16690     * [125]9924 Multiple using statements for builtin functions not
16691       allowed
16692     * [126]9946 destructor is not called for temporary object
16693     * [127]9964 filebuf::close() sometimes fails to close file
16694     * [128]9988 filebuf::overflow writes EOF to file
16695     * [129]10033 optimization breaks polymorphic references w/ typeid
16696       operator
16697     * [130]10097 filebuf::underflow drops characters
16698     * [131]10132 filebuf destructor can throw exceptions
16699     * [132]10180 gcc fails to warn about non-inlined function
16700     * [133]10199 method parametrized by template does not work everywhere
16701     * [134]10300 use of array-new (nothrow) in segfaults on NULL return
16702     * [135]10427 Stack corruption with variable-length automatic arrays
16703       and virtual destructors
16704     * [136]10503 Compilation never stops in fixed_type_or_null
16705
16706    Objective-C
16707
16708     * [137]5956 selectors aren't matched properly when added to the
16709       selector table
16710
16711    Fortran compiler and library
16712
16713     * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't
16714       detect
16715     * [139]3924 g77 generates code that is rejected by GAS if COFF debug
16716       info requested
16717     * [140]5634 doc: explain that configure --prefix=~/... does not work
16718     * [141]6367 multiple repeat counts confuse namelist read into array
16719     * [142]6491 Logical operations error on logicals when using
16720       -fugly-logint
16721     * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C"
16722     * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os
16723       on irix6.5
16724     * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should
16725       assume a direct access file
16726     * [146]7278 g77 "bug"; the executable misbehaves (with -O2
16727       -fno-automatic)
16728     * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows
16729     * [148]7388 Incorrect output with 0-based array of characters
16730     * [149]8587 Double complex zero ** double precision number -> NaN
16731       instead of zero
16732     * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning:
16733       unknown register name line-length-none
16734     * [151]10197 Direct access files not unformatted by default
16735
16736    Java compiler and library
16737
16738     * [152]6005 gcj fails to build rhug on alpha
16739     * [153]6389 System.getProperty("") should always throw an
16740       IllegalArgumentException
16741     * [154]6576 java.util.ResourceBundle.getResource ignores locale
16742     * [155]6652 new java.io.File("").getCanonicalFile() throws exception
16743     * [156]7060 getMethod() doesn't search super interface
16744     * [157]7073 bytecode interpreter gives wrong answer for interface
16745       getSuperclass()
16746     * [158]7180 possible bug in
16747       javax.naming.spi.NamingManager.getPlusPath()
16748     * [159]7416 java.security startup refs "GNU libgcj.security"
16749     * [160]7570 Runtime.exec with null envp: child doesn't inherit parent
16750       env (DUP: [161]7578)
16751     * [162]7611 Internal error while compiling libjava with -O
16752     * [163]7709 NullPointerException in _Jv_ResolvePoolEntry
16753     * [164]7766 ZipInputStream.available returns 0 immediately after
16754       construction
16755     * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public
16756     * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
16757     * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens'
16758     * [168]8234 ZipInputStream chokes when InputStream.read() returns
16759       small chunks
16760     * [169]8415 reflection bug: exception info for Method
16761     * [170]8481 java.Random.nextInt(int) may return negative
16762     * [171]8593 Error reading GZIPped files with BufferedReader
16763     * [172]8759 java.beans.Introspector has no flushCaches() or
16764       flushFromCaches() methods
16765     * [173]8997 spin() calls Thread.sleep
16766     * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd
16767       instead of the root content of C:
16768     * [175]9254 java::lang::Object::wait(), threads-win32.cc returns
16769       wrong return codes
16770     * [176]9271 Severe bias in java.security.SecureRandom
16771
16772    Ada compiler and library
16773
16774     * [177]6767 make gnatlib-shared fails on -laddr2line
16775     * [178]9911 gnatmake fails to link when GCC configured with
16776       --with-sjlj-exceptions=yes
16777     * [179]10020 Can't bootstrap gcc on AIX with Ada enabled
16778     * [180]10546 Ada tasking not working on Red Hat 9
16779
16780    preprocessor
16781
16782     * [181]7029 preprocessor should ignore #warning with -M
16783
16784    ARM-specific
16785
16786     * [182]2903 [arm] Optimization bug with long long arithmetic
16787     * [183]7873 arm-linux-gcc fails when assigning address to a bit field
16788
16789    FreeBSD-specific
16790
16791     * [184]7680 float functions undefined in math.h/cmath with #define
16792       _XOPEN_SOURCE
16793
16794    HP-UX or HP-PA-specific
16795
16796     * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
16797     * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to
16798       fputc_unlocked
16799     * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen
16800
16801    m68hc11-specific
16802
16803     * [188]6744 Bad assembler code generated: reference to pseudo
16804       register z
16805     * [189]7361 Internal compiler error in reload_cse_simplify_operands,
16806       in reload1.c
16807
16808    MIPS-specific
16809
16810     * [190]9496 [mips-linux] bug in optimizer?
16811
16812    PowerPC-specific
16813
16814     * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of
16815       space
16816     * [192]8480 reload ICEs for LAPACK code on powerpc64-linux
16817     * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg
16818     * [194]10315 [powerpc] ICE: in extract_insn, in recog.c
16819
16820    SPARC-specific
16821
16822     * [195]10267 (documentation) Wrong build instructions for
16823       *-*-solaris2*
16824
16825    x86-specific (Intel/AMD)
16826
16827     * [196]7916 ICE in instantiate_virtual_register_1
16828     * [197]7926 (c++) i486 instructions in header files make c++ programs
16829       crash on i386
16830     * [198]8555 ICE in gen_split_1231
16831     * [199]8994 ICE with -O -march=pentium4
16832     * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs
16833     * [201]9806 ICE in inline assembly with -fPIC flag
16834     * [202]10077 gcc -msse2 generates movd to move dwords between xmm
16835       regs
16836     * [203]10233 64-bit comparison only comparing bottom 32-bits
16837     * [204]10286 type-punning doesn't work with __m64 and -O
16838     * [205]10308 [x86] ICE with -O -fgcse or -O2
16839     __________________________________________________________________
16840
16841GCC 3.3.1
16842
16843  Bug Fixes
16844
16845   This section lists the problem reports (PRs) from GCC's bug tracking
16846   system that are known to be fixed in the 3.3.1 release. This list might
16847   not be complete (that is, it is possible that some PRs that have been
16848   fixed are not listed here).
16849
16850    Bootstrap failures
16851
16852     * [206]11272 [Solaris] make bootstrap fails while building libstdc++
16853
16854    Internal compiler errors (multi-platform)
16855
16856     * [207]5754 ICE on invalid nested template class
16857     * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64
16858       and --enable-checking
16859     * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c
16860     * [210]7053 (c++) ICE when declaring a function already defined as a
16861       friend method of a template class
16862     * [211]8164 (c++) ICE when using different const expressions as
16863       template parameter
16864     * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c
16865     * [213]9559 (c++) ICE with invalid initialization of a static const
16866     * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c
16867       when redeclaring a static member variable
16868     * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in
16869       dwarfout.c with -g -O -finline-functions
16870     * [216]10432 (c++) ICE in poplevel, in cp/decl.c
16871     * [217]10475 ICE in subreg_highpart_offset for code with long long
16872     * [218]10635 (c++) ICE when dereferencing an incomplete type casted
16873       from a void pointer
16874     * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while
16875       instantiating static member variables
16876     * [220]10700 ICE in copy_to_mode_reg on 64-bit targets
16877     * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c
16878     * [222]10796 (c++) ICE when defining an enum with two values: -1 and
16879       MAX_INT_64BIT
16880     * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x
16881       sched.c
16882     * [224]10939 (c++) ICE with template code
16883     * [225]10956 (c++) ICE when specializing a template member function
16884       of a template class, in tsubst, in cp/pt.c
16885     * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*()
16886       defined)
16887     * [227]11059 (c++) ICE with empty union
16888     * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with
16889       -O2 -fnon-call-exceptions
16890     * [229]11105 (c++) ICE in mangle_conv_op_name_for_type
16891     * [230]11149 (c++) ICE on error when instantiation with call function
16892       of a base type
16893     * [231]11228 (c++) ICE on new-expression using array operator new and
16894       default-initialization
16895     * [232]11282 (c++) Infinite memory usage after syntax error
16896     * [233]11301 (fortran) ICE with -fno-globals
16897     * [234]11308 (c++) ICE when using an enum type name as if it were a
16898       class or namespace
16899     * [235]11473 (c++) ICE with -gstabs when empty struct inherits from
16900       an empty struct
16901     * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR
16902     * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c:
16903       template member functions
16904
16905    Optimization bugs
16906
16907     * [238]11198 -O2 -frename-registers generates wrong code (aliasing
16908       problem)
16909     * [239]11304 Wrong code production with -fomit-frame-pointer
16910     * [240]11381 volatile memory access optimized away
16911     * [241]11536 [strength-reduce] -O2 optimization produces wrong code
16912     * [242]11557 constant folding bug generates wrong code
16913
16914    C front end
16915
16916     * [243]5897 No warning for statement after return
16917     * [244]11279 DWARF-2 output mishandles large enums
16918
16919    Preprocessor bugs
16920
16921     * [245]11022 no warning for non-compatible macro redefinition
16922
16923    C++ compiler and library
16924
16925     * [246]2330 static_cast<>() to a private base is allowed
16926     * [247]5388 Incorrect message "operands to ?: have different types"
16927     * [248]5390 Libiberty fails to demangle multi-digit template
16928       parameters
16929     * [249]7877 Incorrect parameter passing to specializations of member
16930       function templates
16931     * [250]9393 Anonymous namespaces and compiling the same file twice
16932     * [251]10032 -pedantic converts some errors to warnings
16933     * [252]10468 const typeof(x) is non-const, but only in templates
16934     * [253]10527 confused error message with "new int()" parameter
16935       initializer
16936     * [254]10679 parameter MIN_INLINE_INSNS is not honored
16937     * [255]10682 gcc chokes on a typedef for an enum inside a class
16938       template
16939     * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of
16940       0.
16941     * [257]10845 template member function (with nested template as
16942       parameter) cannot be called anymore if another unrelated template
16943       member function is defined
16944     * [258]10849 Cannot define an out-of-class specialization of a
16945       private nested template class
16946     * [259]10888 Suppress -Winline warnings for system headers
16947     * [260]10929 -Winline warns about functions for which no definition
16948       is visible
16949     * [261]10931 valid conversion static_cast<const unsigned
16950       int&>(lvalue-of-type-int) is rejected
16951     * [262]10940 Bad code with explicit specialization
16952     * [263]10968 If member function implicitly instantiated, explicit
16953       instantiation of class fails to instantiate it
16954     * [264]10990 Cannot convert with dynamic_cast<> to a private base
16955       class from within a member function
16956     * [265]11039 Bad interaction between implicit typename deprecation
16957       and friendship
16958     * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say
16959       "__unused__" instead
16960     * [267]11095 C++ iostream manipulator causes segfault when called
16961       with negative argument
16962     * [268]11098 g++ doesn't emit complete debugging information for
16963       local variables in destructors
16964     * [269]11137 GNU/Linux shared library constructors not called unless
16965       there's one global object
16966     * [270]11154 spurious ambiguity report for template class
16967       specialization
16968     * [271]11329 Compiler cannot find user defined implicit typecast
16969     * [272]11332 Spurious error with casts in ?: expression
16970     * [273]11431 static_cast behavior with subclasses when default
16971       constructor available
16972     * [274]11528 money_get facet does not accept "$.00" as valid
16973     * [275]11546 Type lookup problems in out-of-line definition of a
16974       class doubly nested from a template class
16975     * [276]11567 C++ code containing templated member function with same
16976       name as pure virtual member function results in linking failure
16977     * [277]11645 Failure to deal with using and private inheritance
16978
16979    Java compiler and library
16980
16981     * [278]5179 Qualified static field access doesn't initialize its
16982       class
16983     * [279]8204 gcj -O2 to native reorders certain instructions
16984       improperly
16985     * [280]10838 java.io.ObjectInputStream syntax error
16986     * [281]10886 The RMI registry that comes with GCJ does not work
16987       correctly
16988     * [282]11349 JNDI URL context factories not located correctly
16989
16990    x86-specific (Intel/AMD)
16991
16992     * [283]4823 ICE on inline assembly code
16993     * [284]8878 miscompilation with -O and SSE
16994     * [285]9815 (c++ library) atomicity.h - fails to compile with -O3
16995       -masm=intel
16996     * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads,
16997       in reload1.c
16998     * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2
16999     * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source
17000     * [289]11044 [x86] out of range loop instructions for FP code on K6
17001     * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE
17002       built-ins
17003     * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC"
17004       is used
17005
17006    SPARC- or Solaris- specific
17007
17008     * [292]9362 solaris 'as' dies when fed .s and "-gstabs"
17009     * [293]10142 [SPARC64] gcc produces wrong code when passing
17010       structures by value
17011     * [294]10663 New configure check aborts with Sun tools.
17012     * [295]10835 combinatorial explosion in scheduler on HyperSPARC
17013     * [296]10876 ICE in calculate_giv_inc when building KDE
17014     * [297]10955 wrong code at -O3 for structure argument in context of
17015       structure return
17016     * [298]11018 -mcpu=ultrasparc busts tar-1.13.25
17017     * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x
17018       Linux kernel
17019
17020    ia64 specific
17021
17022     * [300]10907 gcc violates the ia64 ABI (GP must be preserved)
17023     * [301]11320 scheduler bug (in machine depended reorganization pass)
17024     * [302]11599 bug with conditional and __builtin_prefetch
17025
17026    PowerPC specific
17027
17028     * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem
17029       during loop)
17030     * [304]10871 error in rs6000_stack_info save_size computation
17031     * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse
17032       cures it
17033
17034    m68k-specific
17035
17036     * [306]7594 [m68k] ICE on legal code associated with simplify-rtx
17037     * [307]10557 [m68k] ICE in subreg_offset_representable_p
17038     * [308]11054 [m68k] ICE in reg_overlap_mentioned_p
17039
17040    ARM-specific
17041
17042     * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for
17043       functions with __attribute__ ((interrupt ("IRQ")))
17044     * [310]10842 [arm] Clobbered link register is copied to pc under
17045       certain circumstances
17046     * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes
17047     * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno
17048       (3.4)
17049
17050    MIPS-specific
17051
17052     * [313]11084 ICE in propagate_one_insn, in flow.c
17053
17054    SH-specific
17055
17056     * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf
17057     * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c
17058     * [316]11096 i686-linux to sh-linux cross compiler fails to compile
17059       C++ files
17060
17061    GNU/Linux (or Hurd?) specific
17062
17063     * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3
17064
17065    UnixWare specific
17066
17067     * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare
17068       7.1.1
17069
17070    Cygwin (or mingw) specific
17071
17072     * [319]5287 ICE with dllimport attribute
17073     * [320]10148 [MingW/CygWin] Compiler dumps core
17074
17075    DJGPP specific
17076
17077     * [321]8787 GCC fails to emit .intel_syntax when invoked with
17078       -masm=intel on DJGPP
17079
17080    Darwin (and MacOS X) specific
17081
17082     * [322]10900 trampolines crash
17083
17084    Documentation
17085
17086     * [323]1607 (c++) Format attributes on methods undocumented
17087     * [324]4252 Invalid option `-fdump-translation-unit'
17088     * [325]4490 Clarify restrictions on -m96bit-long-double,
17089       -m128bit-long-double
17090     * [326]10355 document an issue with regparm attribute on some systems
17091       (e.g. Solaris)
17092     * [327]10726 (fortran) Documentation for function "IDate Intrinsic
17093       (Unix)" is wrong
17094     * [328]10805 document bug in old version of Sun assembler
17095     * [329]10815 warn against GNU binutils on AIX
17096     * [330]10877 document need for newer binutils on i?86-*-linux-gnu
17097     * [331]11280 Manual incorrect with respect to -freorder-blocks
17098     * [332]11466 Document -mlittle-endian and its restrictions for the
17099       sparc64 port
17100
17101    Testsuite bugs (compiler itself is not affected)
17102
17103     * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly
17104       report failure
17105     * [334]10810 gcc-3.3 fails make check: buffer overrun in
17106       test_demangle.c
17107     __________________________________________________________________
17108
17109GCC 3.3.2
17110
17111  Bug Fixes
17112
17113   This section lists the problem reports (PRs) from GCC's bug tracker
17114   that are known to be fixed in the 3.3.2 release. This list might not be
17115   complete (that is, it is possible that some PRs that have been fixed
17116   are not listed here).
17117
17118    Bootstrap failures and problems
17119
17120     * [335]8336 [SCO5] bootstrap config still tries to use COFF options
17121     * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with
17122       --enable-threads=posix
17123     * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap
17124     * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare
17125       7.1.1)
17126     * [339]11687 xstormy16-elf build fails in libf2c
17127     * [340]12263 [SGI IRIX] bootstrap fails during compile of
17128       libf2c/libI77/backspace.c
17129     * [341]12490 buffer overflow in scan-decls.c (during Solaris 9
17130       fix-header processing)
17131
17132    Internal compiler errors (multi-platform)
17133
17134     * [342]7277 Casting integers to vector types causes ICE
17135     * [343]7939 (c++) ICE on invalid function template specialization
17136     * [344]11063 (c++) ICE on parsing initialization list of const array
17137       member
17138     * [345]11207 ICE with negative index in array element designator
17139     * [346]11522 (fortran) g77 dwarf-2 ICE in
17140       add_abstract_origin_attribute
17141     * [347]11595 (c++) ICE on duplicate label definition
17142     * [348]11646 (c++) ICE in commit_one_edge_insertion with
17143       -fnon-call-exceptions -fgcse -O
17144     * [349]11665 ICE in struct initializer when taking address
17145     * [350]11852 (c++) ICE with bad struct initializer.
17146     * [351]11878 (c++) ICE in cp_expr_size
17147     * [352]11883 ICE with any -O on mercury-generated C code
17148     * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in
17149       cp/typeck2.c when applying typeid operator to template template
17150       parameter
17151     * [354]12146 ICE in lookup_template_function, in cp/pt.c
17152     * [355]12215 ICE in make_label_edge with -fnon-call-exceptions
17153       -fno-gcse -O2
17154     * [356]12369 (c++) ICE with templates and friends
17155     * [357]12446 ICE in emit_move_insn on complicated array reference
17156     * [358]12510 ICE in final_scan_insn
17157     * [359]12544 ICE with large parameters used in nested functions
17158
17159    C and optimization bugs
17160
17161     * [360]9862 spurious warnings with -W -finline-functions
17162     * [361]10962 lookup_field is a linear search on a linked list (can be
17163       slow if large struct)
17164     * [362]11370 -Wunreachable-code gives false complaints
17165     * [363]11637 invalid assembly with -fnon-call-exceptions
17166     * [364]11885 Problem with bitfields in packed structs
17167     * [365]12082 Inappropriate unreachable code warnings
17168     * [366]12180 Inline optimization fails for variadic function
17169     * [367]12340 loop unroller + gcse produces wrong code
17170
17171    C++ compiler and library
17172
17173     * [368]3907 nested template parameter collides with member name
17174     * [369]5293 confusing message when binding a temporary to a reference
17175     * [370]5296 [DR115] Pointers to functions and to template functions
17176       behave differently in deduction
17177     * [371]7939 ICE on function template specialization
17178     * [372]8656 Unable to assign function with __attribute__ and pointer
17179       return type to an appropriate variable
17180     * [373]10147 Confusing error message for invalid template function
17181       argument
17182     * [374]11400 std::search_n() makes assumptions about Size parameter
17183     * [375]11409 issues with using declarations, overloading, and
17184       built-in functions
17185     * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle
17186       multiple bits in mask
17187     * [377]11786 operator() call on variable in other namespace not
17188       recognized
17189     * [378]11867 static_cast ignores ambiguity
17190     * [379]11928 bug with conversion operators that are typedefs
17191     * [380]12114 Uninitialized memory accessed in dtor
17192     * [381]12163 static_cast + explicit constructor regression
17193     * [382]12181 Wrong code with comma operator and c++
17194     * [383]12236 regparm and fastcall messes up parameters
17195     * [384]12266 incorrect instantiation of unneeded template during
17196       overload resolution
17197     * [385]12296 istream::peek() doesn't set eofbit
17198     * [386]12298 [sjlj exceptions] Stack unwind destroys
17199       not-yet-constructed object
17200     * [387]12369 ICE with templates and friends
17201     * [388]12337 apparently infinite loop in g++
17202     * [389]12344 stdcall attribute ignored if function returns a pointer
17203     * [390]12451 missing(late) class forward declaration in cxxabi.h
17204     * [391]12486 g++ accepts invalid use of a qualified name
17205
17206    x86 specific (Intel/AMD)
17207
17208     * [392]8869 [x86 MMX] ICE with const variable optimization and MMX
17209       builtins
17210     * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions
17211       -O2
17212     * [394]11689 g++3.3 emits un-assembleable code for k6 architecture
17213     * [395]12116 [k6] Invalid assembly output values with X-MAME code
17214     * [396]12070 ICE converting between double and long double with
17215       -msoft-float
17216
17217    ia64-specific
17218
17219     * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc
17220     * [398]11535 __builtin_return_address may not work on ia64
17221     * [399]11693 [ia64] ICE in gen_nop_type
17222     * [400]12224 [ia64] Thread-local storage doesn't work
17223
17224    PowerPC-specific
17225
17226     * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux
17227       kernel
17228     * [402]11319 loop miscompiled on ppc32
17229     * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code
17230
17231    SPARC-specific
17232
17233     * [404]11662 wrong code for expr. with cast to long long and
17234       exclusive or
17235     * [405]11965 invalid assembler code for a shift < 32 operation
17236     * [406]12301 (c++) stack corruption when a returned expression throws
17237       an exception
17238
17239    Alpha-specific
17240
17241     * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of
17242       kernel 2.4.22-pre8
17243
17244    HPUX-specific
17245
17246     * [408]11313 problem with #pragma weak and static inline functions
17247     * [409]11712 __STDC_EXT__ not defined for C++ by default anymore?
17248
17249    Solaris specific
17250
17251     * [410]12166 Profiled programs crash if PROFDIR is set
17252
17253    Solaris-x86 specific
17254
17255     * [411]12101 i386 Solaris no longer works with GNU as?
17256
17257    Miscellaneous embedded target-specific bugs
17258
17259     * [412]10988 [m32r-elf] wrong blockmove code with -O3
17260     * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with
17261       -O2
17262     * [414]11902 [sh4] spec file improperly inserts rpath even when none
17263       needed
17264     * [415]11903 [sh4] -pthread fails to link due to error in spec file
17265       on sh4
17266     __________________________________________________________________
17267
17268GCC 3.3.3
17269
17270  Minor features
17271
17272   In addition to the bug fixes documented below, this release contains
17273   few minor features such as:
17274     * Support for --with-sysroot
17275     * Support for automatic detection of executable stacks
17276     * Support for SSE3 instructions
17277     * Support for thread local storage debugging under GDB on S390
17278
17279  Bug Fixes
17280
17281   This section lists the problem reports (PRs) from GCC's bug tracker
17282   that are known to be fixed in the 3.3.3 release. This list might not be
17283   complete (that is, it is possible that some PRs that have been fixed
17284   are not listed here).
17285
17286    Bootstrap failures and issues
17287
17288     * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails
17289     * [417]12399 boehm-gc fails (when building a cross compiler): libtool
17290       unable to infer tagged configuration
17291     * [418]13068 mklibgcc.in doesn't handle multi-level multilib
17292       subdirectories properly
17293
17294    Internal compiler errors (multi-platform)
17295
17296     * [419]10060 ICE (stack overflow) on huge file (300k lines) due to
17297       recursive behaviour of copy_rtx_if_shared, in emit_rtl.c
17298     * [420]10555 (c++) ICE on undefined template argument
17299     * [421]10706 (c++) ICE in mangle_class_name_for_template
17300     * [422]11496 (fortran) error in flow_loops_find when -funroll-loops
17301       active
17302     * [423]11741 ICE in pre_insert_copy_insn, in gcse.c
17303     * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0
17304     * [425]12632 (fortran) -fbounds-check ICE
17305     * [426]12712 (c++) ICE on short legit C++ code fragment with gcc
17306       3.3.2
17307     * [427]12726 (c++) ICE (segfault) on trivial code
17308     * [428]12890 (c++) ICE on compilation of class with throwing method
17309     * [429]12900 (c++) ICE in rtl_verify_flow_info_1
17310     * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on
17311       correct code with -O2 -fno-force-mem
17312     * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive
17313       template
17314     * [432]13318 ICE: floating point exception in the loop optimizer
17315     * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in
17316       except.c
17317     * [434]13574 (c++) invalid array default initializer in class lets
17318       gcc consume all memory and die
17319     * [435]13475 ICE on SIMD variables with partial value initialization
17320     * [436]13797 (c++) ICE on invalid template parameter
17321     * [437]13824 (java) gcj SEGV with simple .java program
17322
17323    C and optimization bugs
17324
17325     * [438]8776 loop invariants are not removed (most likely)
17326     * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing
17327       strncmp by memcmp
17328     * [440]11350 undefined labels with -Os -fPIC
17329     * [441]12826 Optimizer removes reference through volatile pointer
17330     * [442]12500 stabs debug info: void no longer a predefined / builtin
17331       type
17332     * [443]12941 builtin-bitops-1.c miscompilation (latent bug)
17333     * [444]12953 tree inliner bug (in inline_forbidden_p) and fix
17334     * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled
17335     * [446]13507 spurious printf format warning
17336     * [447]13382 Type information for const pointer disappears during
17337       optimization.
17338     * [448]13394 noreturn attribute ignored on recursive invokation
17339     * [449]13400 Compiled code crashes storing to read-only location
17340     * [450]13521 Endless loop in calculate_global_regs_live
17341
17342    C++ compiler and library
17343
17344   Some of the bug fixes in this list were made to implement decisions
17345   that the ISO C++ standards committee has made concerning several defect
17346   reports (DRs). Links in the list below point to detailed discussion of
17347   the relevant defect report.
17348     * [451]2094 unimplemented: use of `ptrmem_cst' in template type
17349       unification
17350     * [452]2294 using declaration confusion
17351     * [453]5050 template instantiation depth exceeds limit: recursion
17352       problem?
17353     * [454]9371 Bad exception handling in
17354       i/ostream::operator>>/<<(streambuf*)
17355     * [455]9546 bad exception handling in ostream members
17356     * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the
17357       face of unknown locales
17358     * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work
17359     * [459]10095 istream::operator>>(int&) sets ios::badbit when
17360       ios::failbit is set.
17361     * [460]11554 Warning about reordering of initializers doesn't mention
17362       location of constructor
17363     * [461]12297 istream::sentry::sentry() handles eof() incorrectly.
17364     * [462]12352 Exception safety problems in src/localename.cc
17365     * [463]12438 Memory leak in locale::combine()
17366     * [464]12540 Memory leak in locale::locale(const char*)
17367     * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented
17368     * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented
17369     * [470]12696 memory eating infinite loop in diagnostics (error
17370       recovery problem)
17371     * [471]12815 Code compiled with optimization behaves unexpectedly
17372     * [472]12862 Conflicts between typedefs/enums and namespace member
17373       declarations
17374     * [473]12926 Wrong value after assignment in initialize list using
17375       bit-fields
17376     * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented
17377     * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented
17378     * [478]13007 basic_streambuf::pubimbue, imbue wrong
17379     * [479]13009 Implicitly-defined assignment operator writes to wrong
17380       memory
17381     * [480]13057 regparm attribute not applied to destructor
17382     * [481]13070 -Wformat option ignored in g++
17383     * [482]13081 forward template declarations in <complex> let inlining
17384       fail
17385     * [483]13239 Assertion does not seem to work correctly anymore
17386     * [484]13262 "xxx is private within this context" when initializing a
17387       self-contained template class
17388     * [485]13290 simple typo in concept checking for std::generate_n
17389     * [486]13323 Template code does not compile in presence of typedef
17390     * [487]13369 __verify_grouping (and __add_grouping?) not correct
17391     * [488]13371 infinite loop with packed struct and inlining
17392     * [489]13445 Template argument replacement "dereferences" a typedef
17393     * [490]13461 Fails to access protected-ctor from public constant
17394     * [491]13462 Non-standard-conforming type set::pointer
17395     * [492]13478 gcc uses wrong constructor to initialize a const
17396       reference
17397     * [493]13544 "conflicting types" for enums in different scopes
17398     * [494]13650 string::compare should not (always) use
17399       traits_type::length()
17400     * [495]13683 bogus warning about passing non-PODs through ellipsis
17401     * [496]13688 Derived class is denied access to protected base class
17402       member class
17403     * [497]13774 Member variable cleared in virtual multiple inheritance
17404       class
17405     * [498]13884 Protect sstream.tcc from extern template use
17406
17407    Java compiler and library
17408
17409     * [499]10746 [win32] garbage collection crash in GCJ
17410
17411    Objective-C compiler and library
17412
17413     * [500]11433 Crash due to dereferencing null pointer when querying
17414       protocol
17415
17416    Fortran compiler and library
17417
17418     * [501]12633 logical expression gives incorrect result with
17419       -fugly-logint option
17420     * [502]13037 [gcse-lm] g77 generates incorrect code
17421     * [503]13213 Hex constant problem when compiling with -fugly-logint
17422       and -ftypeless-boz
17423
17424    x86-specific (Intel/AMD)
17425
17426     * [504]4490 ICE with -m128bit-long-double
17427     * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have
17428       `reg' in make_field_assignment, in combine.c
17429     * [506]12441 ICE: can't find a register to spill
17430     * [507]12943 array static-init failure under -fpic, -fPIC
17431     * [508]13608 Incorrect code with -O3 -ffast-math
17432
17433    PowerPC-specific
17434
17435     * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of
17436       __attribute__((aligned(16)))
17437     * [510]11793 ICE in extract_insn, in recog.c (const_vector's)
17438     * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in
17439       altivec.md)
17440     * [512]12537 g++ generates writeable text sections
17441
17442    SPARC-specific
17443
17444     * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0
17445       -m64
17446     * [514]12865 mprotect call to make trampoline executable may fail
17447     * [515]13354 ICE in sparc_emit_set_const32
17448
17449    ARM-specific
17450
17451     * [516]10467 [arm] ICE in pre_insert_copy_insn,
17452
17453    ia64-specific
17454
17455     * [517]11226 ICE passing struct arg with two floats
17456     * [518]11227 ICE for _Complex float, _Complex long double args
17457     * [519]12644 GCC 3.3.2 fails to compile glibc on ia64
17458     * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn
17459     * Various fixes for libunwind
17460
17461    Alpha-specific
17462
17463     * [521]12654 Incorrect comparison code generated for Alpha
17464     * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2
17465     * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2
17466
17467    HPPA-specific
17468
17469     * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c
17470     * [525]12158 [hppa] compilation does not terminate at -O1
17471
17472    S390-specific
17473
17474     * [526]11992 Wrong built-in code for memcmp with length 1<<24: only
17475       (1<<24)-1 possible for CLCL-Instruction
17476
17477    SH-specific
17478
17479     * [527]9365 segfault in gen_far_branch (config/sh/sh.c)
17480     * [528]10392 optimizer generates faulty array indexing
17481     * [529]11322 SH profiler outputs multiple definitions of symbol
17482     * [530]13069 gcc/config/sh/rtems.h broken
17483     * [531]13302 Putting a va_list in a struct causes seg fault
17484     * [532]13585 Incorrect optimization of call to sfunc
17485     * Fix inappropriately exported libgcc functions from the shared
17486       library
17487
17488    Other embedded target specific
17489
17490     * [533]8916 [mcore] unsigned char assign gets hosed.
17491     * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c
17492     * [535]13122 [h8300] local variable gets corrupted by function call
17493       when -fomit-frame-pointer is given
17494     * [536]13256 [cris] strict_low_part mistreated in delay slots
17495     * [537]13373 [mcore] optimization with -frerun-cse-after-loop
17496       -fexpensive-optimizations produces wrong code on mcore
17497
17498    GNU HURD-specific
17499
17500     * [538]12561 gcc/config/t-gnu needs updating to work with
17501       --with-sysroot
17502
17503    Tru64 Unix specific
17504
17505     * [539]6243 testsuite fails almost all tests due to no libintl in
17506       LD_LIBRARY_PATH during test.
17507     * [540]11397 weak aliases broken on Tru64 UNIX
17508
17509    AIX-specific
17510
17511     * [541]12505 build failure due to defines of uchar in cpphash.h and
17512       sys/types.h
17513     * [542]13150 WEAK symbols not exported by collect2
17514
17515    IRIX-specific
17516
17517     * [543]12666 fixincludes problem on IRIX 6.5.19m
17518
17519    Solaris-specific
17520
17521     * [544]12969 Including sys/byteorder.h breaks configure checks
17522
17523    Testsuite problems (compiler is not affected)
17524
17525     * [545]10819 testsuite creates CR+LF on compiler version lines in
17526       test summary files
17527     * [546]11612 abi_check not finding correct libgcc_s.so.1
17528
17529    Miscellaneous
17530
17531     * [547]13211 using -###, incorrect warnings about unused linker file
17532       are produced
17533     __________________________________________________________________
17534
17535GCC 3.3.4
17536
17537   This is the [548]list of problem reports (PRs) from GCC's bug tracking
17538   system that are known to be fixed in the 3.3.4 release. This list might
17539   not be complete (that is, it is possible that some PRs that have been
17540   fixed are not listed here).
17541     __________________________________________________________________
17542
17543GCC 3.3.5
17544
17545   This is the [549]list of problem reports (PRs) from GCC's bug tracking
17546   system that are known to be fixed in the 3.3.5 release. This list might
17547   not be complete (that is, it is possible that some PRs that have been
17548   fixed are not listed here).
17549     __________________________________________________________________
17550
17551GCC 3.3.6
17552
17553   This is the [550]list of problem reports (PRs) from GCC's bug tracking
17554   system that are known to be fixed in the 3.3.6 release. This list might
17555   not be complete (that is, it is possible that some PRs that have been
17556   fixed are not listed here).
17557
17558
17559    For questions related to the use of GCC, please consult these web
17560    pages and the [551]GCC manuals. If that fails, the
17561    [552]gcc-help@gcc.gnu.org mailing list might help. Comments on these
17562    web pages and the development of GCC are welcome on our developer
17563    list at [553]gcc@gcc.gnu.org. All of [554]our lists have public
17564    archives.
17565
17566   Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and
17567   distribution of this entire article is permitted in any medium,
17568   provided this notice is preserved.
17569
17570   These pages are [556]maintained by the GCC team. Last modified
17571   2023-01-19.
17572
17573References
17574
17575   1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
17576   2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems
17577   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
17578   4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute
17579   5. http://gcc.gnu.org/news/dfa.html
17580   6. http://gcc.gnu.org/c99status.html
17581   7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html
17582   8. https://gcc.gnu.org/PR10140
17583   9. https://gcc.gnu.org/PR10198
17584  10. https://gcc.gnu.org/PR10338
17585  11. https://gcc.gnu.org/PR3581
17586  12. https://gcc.gnu.org/PR4382
17587  13. https://gcc.gnu.org/PR5533
17588  14. https://gcc.gnu.org/PR6387
17589  15. https://gcc.gnu.org/PR6412
17590  16. https://gcc.gnu.org/PR6620
17591  17. https://gcc.gnu.org/PR6663
17592  18. https://gcc.gnu.org/PR7068
17593  19. https://gcc.gnu.org/PR7083
17594  20. https://gcc.gnu.org/PR7647
17595  21. https://gcc.gnu.org/PR7675
17596  22. https://gcc.gnu.org/PR7718
17597  23. https://gcc.gnu.org/PR8116
17598  24. https://gcc.gnu.org/PR8358
17599  25. https://gcc.gnu.org/PR8511
17600  26. https://gcc.gnu.org/PR8564
17601  27. https://gcc.gnu.org/PR8660
17602  28. https://gcc.gnu.org/PR8766
17603  29. https://gcc.gnu.org/PR8803
17604  30. https://gcc.gnu.org/PR8846
17605  31. https://gcc.gnu.org/PR8906
17606  32. https://gcc.gnu.org/PR9216
17607  33. https://gcc.gnu.org/PR9261
17608  34. https://gcc.gnu.org/PR9263
17609  35. https://gcc.gnu.org/PR9429
17610  36. https://gcc.gnu.org/PR9516
17611  37. https://gcc.gnu.org/PR9600
17612  38. https://gcc.gnu.org/PR9629
17613  39. https://gcc.gnu.org/PR9672
17614  40. https://gcc.gnu.org/PR9749
17615  41. https://gcc.gnu.org/PR9794
17616  42. https://gcc.gnu.org/PR9829
17617  43. https://gcc.gnu.org/PR9916
17618  44. https://gcc.gnu.org/PR9936
17619  45. https://gcc.gnu.org/PR10262
17620  46. https://gcc.gnu.org/PR10278
17621  47. https://gcc.gnu.org/PR10446
17622  48. https://gcc.gnu.org/PR10451
17623  49. https://gcc.gnu.org/PR10506
17624  50. https://gcc.gnu.org/PR10549
17625  51. https://gcc.gnu.org/PR2001
17626  52. https://gcc.gnu.org/PR2391
17627  53. https://gcc.gnu.org/PR2960
17628  54. https://gcc.gnu.org/PR4046
17629  55. https://gcc.gnu.org/PR6405
17630  56. https://gcc.gnu.org/PR6798
17631  57. https://gcc.gnu.org/PR6871
17632  58. https://gcc.gnu.org/PR6909
17633  59. https://gcc.gnu.org/PR7189
17634  60. https://gcc.gnu.org/PR7642
17635  61. https://gcc.gnu.org/PR8634
17636  62. https://gcc.gnu.org/PR8750
17637  63. https://gcc.gnu.org/PR2161
17638  64. https://gcc.gnu.org/PR4319
17639  65. https://gcc.gnu.org/PR8602
17640  66. https://gcc.gnu.org/PR9177
17641  67. https://gcc.gnu.org/PR9853
17642  68. https://gcc.gnu.org/PR45
17643  69. https://gcc.gnu.org/PR3784
17644  70. https://gcc.gnu.org/PR764
17645  71. https://gcc.gnu.org/PR5116
17646  72. https://gcc.gnu.org/PR2862
17647  73. https://gcc.gnu.org/PR3663
17648  74. https://gcc.gnu.org/PR3797
17649  75. https://gcc.gnu.org/PR3948
17650  76. https://gcc.gnu.org/PR4137
17651  77. https://gcc.gnu.org/PR4361
17652  78. https://gcc.gnu.org/PR4802
17653  79. https://gcc.gnu.org/PR5837
17654  80. https://gcc.gnu.org/PR4803
17655  81. https://gcc.gnu.org/PR5094
17656  82. https://gcc.gnu.org/PR5730
17657  83. https://gcc.gnu.org/PR6713
17658  84. https://gcc.gnu.org/PR7015
17659  85. https://gcc.gnu.org/PR7086
17660  86. https://gcc.gnu.org/PR7099
17661  87. https://gcc.gnu.org/PR7247
17662  88. https://gcc.gnu.org/PR7441
17663  89. https://gcc.gnu.org/PR7768
17664  90. https://gcc.gnu.org/PR7804
17665  91. https://gcc.gnu.org/PR8099
17666  92. https://gcc.gnu.org/PR8117
17667  93. https://gcc.gnu.org/PR8205
17668  94. https://gcc.gnu.org/PR8645
17669  95. https://gcc.gnu.org/PR8724
17670  96. https://gcc.gnu.org/PR8805
17671  97. https://gcc.gnu.org/PR8691
17672  98. https://gcc.gnu.org/PR8700
17673  99. https://gcc.gnu.org/PR8724
17674 100. https://gcc.gnu.org/PR8949
17675 101. https://gcc.gnu.org/PR9016
17676 102. https://gcc.gnu.org/PR9053
17677 103. https://gcc.gnu.org/PR9152
17678 104. https://gcc.gnu.org/PR9182
17679 105. https://gcc.gnu.org/PR9297
17680 106. https://gcc.gnu.org/PR9318
17681 107. https://gcc.gnu.org/PR9320
17682 108. https://gcc.gnu.org/PR9400
17683 109. https://gcc.gnu.org/PR9424
17684 110. https://gcc.gnu.org/PR9425
17685 111. https://gcc.gnu.org/PR9439
17686 112. https://gcc.gnu.org/PR9474
17687 113. https://gcc.gnu.org/PR9548
17688 114. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231
17689 115. https://gcc.gnu.org/PR9555
17690 116. https://gcc.gnu.org/PR9561
17691 117. https://gcc.gnu.org/PR9563
17692 118. https://gcc.gnu.org/PR9582
17693 119. https://gcc.gnu.org/PR9622
17694 120. https://gcc.gnu.org/PR9683
17695 121. https://gcc.gnu.org/PR9791
17696 122. https://gcc.gnu.org/PR9817
17697 123. https://gcc.gnu.org/PR9825
17698 124. https://gcc.gnu.org/PR9826
17699 125. https://gcc.gnu.org/PR9924
17700 126. https://gcc.gnu.org/PR9946
17701 127. https://gcc.gnu.org/PR9964
17702 128. https://gcc.gnu.org/PR9988
17703 129. https://gcc.gnu.org/PR10033
17704 130. https://gcc.gnu.org/PR10097
17705 131. https://gcc.gnu.org/PR10132
17706 132. https://gcc.gnu.org/PR10180
17707 133. https://gcc.gnu.org/PR10199
17708 134. https://gcc.gnu.org/PR10300
17709 135. https://gcc.gnu.org/PR10427
17710 136. https://gcc.gnu.org/PR10503
17711 137. https://gcc.gnu.org/PR5956
17712 138. https://gcc.gnu.org/PR1832
17713 139. https://gcc.gnu.org/PR3924
17714 140. https://gcc.gnu.org/PR5634
17715 141. https://gcc.gnu.org/PR6367
17716 142. https://gcc.gnu.org/PR6491
17717 143. https://gcc.gnu.org/PR6742
17718 144. https://gcc.gnu.org/PR7113
17719 145. https://gcc.gnu.org/PR7236
17720 146. https://gcc.gnu.org/PR7278
17721 147. https://gcc.gnu.org/PR7384
17722 148. https://gcc.gnu.org/PR7388
17723 149. https://gcc.gnu.org/PR8587
17724 150. https://gcc.gnu.org/PR9038
17725 151. https://gcc.gnu.org/PR10197
17726 152. https://gcc.gnu.org/PR6005
17727 153. https://gcc.gnu.org/PR6389
17728 154. https://gcc.gnu.org/PR6576
17729 155. https://gcc.gnu.org/PR6652
17730 156. https://gcc.gnu.org/PR7060
17731 157. https://gcc.gnu.org/PR7073
17732 158. https://gcc.gnu.org/PR7180
17733 159. https://gcc.gnu.org/PR7416
17734 160. https://gcc.gnu.org/PR7570
17735 161. https://gcc.gnu.org/PR7578
17736 162. https://gcc.gnu.org/PR7611
17737 163. https://gcc.gnu.org/PR7709
17738 164. https://gcc.gnu.org/PR7766
17739 165. https://gcc.gnu.org/PR7785
17740 166. https://gcc.gnu.org/PR7786
17741 167. https://gcc.gnu.org/PR8142
17742 168. https://gcc.gnu.org/PR8234
17743 169. https://gcc.gnu.org/PR8415
17744 170. https://gcc.gnu.org/PR8481
17745 171. https://gcc.gnu.org/PR8593
17746 172. https://gcc.gnu.org/PR8759
17747 173. https://gcc.gnu.org/PR8997
17748 174. https://gcc.gnu.org/PR9253
17749 175. https://gcc.gnu.org/PR9254
17750 176. https://gcc.gnu.org/PR9271
17751 177. https://gcc.gnu.org/PR6767
17752 178. https://gcc.gnu.org/PR9911
17753 179. https://gcc.gnu.org/PR10020
17754 180. https://gcc.gnu.org/PR10546
17755 181. https://gcc.gnu.org/PR7029
17756 182. https://gcc.gnu.org/PR2903
17757 183. https://gcc.gnu.org/PR7873
17758 184. https://gcc.gnu.org/PR7680
17759 185. https://gcc.gnu.org/PR8705
17760 186. https://gcc.gnu.org/PR9986
17761 187. https://gcc.gnu.org/PR10056
17762 188. https://gcc.gnu.org/PR6744
17763 189. https://gcc.gnu.org/PR7361
17764 190. https://gcc.gnu.org/PR9496
17765 191. https://gcc.gnu.org/PR7067
17766 192. https://gcc.gnu.org/PR8480
17767 193. https://gcc.gnu.org/PR8784
17768 194. https://gcc.gnu.org/PR10315
17769 195. https://gcc.gnu.org/PR10267
17770 196. https://gcc.gnu.org/PR7916
17771 197. https://gcc.gnu.org/PR7926
17772 198. https://gcc.gnu.org/PR8555
17773 199. https://gcc.gnu.org/PR8994
17774 200. https://gcc.gnu.org/PR9426
17775 201. https://gcc.gnu.org/PR9806
17776 202. https://gcc.gnu.org/PR10077
17777 203. https://gcc.gnu.org/PR10233
17778 204. https://gcc.gnu.org/PR10286
17779 205. https://gcc.gnu.org/PR10308
17780 206. https://gcc.gnu.org/PR11272
17781 207. https://gcc.gnu.org/PR5754
17782 208. https://gcc.gnu.org/PR6597
17783 209. https://gcc.gnu.org/PR6949
17784 210. https://gcc.gnu.org/PR7053
17785 211. https://gcc.gnu.org/PR8164
17786 212. https://gcc.gnu.org/PR8384
17787 213. https://gcc.gnu.org/PR9559
17788 214. https://gcc.gnu.org/PR9649
17789 215. https://gcc.gnu.org/PR9864
17790 216. https://gcc.gnu.org/PR10432
17791 217. https://gcc.gnu.org/PR10475
17792 218. https://gcc.gnu.org/PR10635
17793 219. https://gcc.gnu.org/PR10661
17794 220. https://gcc.gnu.org/PR10700
17795 221. https://gcc.gnu.org/PR10712
17796 222. https://gcc.gnu.org/PR10796
17797 223. https://gcc.gnu.org/PR10890
17798 224. https://gcc.gnu.org/PR10939
17799 225. https://gcc.gnu.org/PR10956
17800 226. https://gcc.gnu.org/PR11041
17801 227. https://gcc.gnu.org/PR11059
17802 228. https://gcc.gnu.org/PR11083
17803 229. https://gcc.gnu.org/PR11105
17804 230. https://gcc.gnu.org/PR11149
17805 231. https://gcc.gnu.org/PR11228
17806 232. https://gcc.gnu.org/PR11282
17807 233. https://gcc.gnu.org/PR11301
17808 234. https://gcc.gnu.org/PR11308
17809 235. https://gcc.gnu.org/PR11473
17810 236. https://gcc.gnu.org/PR11503
17811 237. https://gcc.gnu.org/PR11513
17812 238. https://gcc.gnu.org/PR11198
17813 239. https://gcc.gnu.org/PR11304
17814 240. https://gcc.gnu.org/PR11381
17815 241. https://gcc.gnu.org/PR11536
17816 242. https://gcc.gnu.org/PR11557
17817 243. https://gcc.gnu.org/PR5897
17818 244. https://gcc.gnu.org/PR11279
17819 245. https://gcc.gnu.org/PR11022
17820 246. https://gcc.gnu.org/PR2330
17821 247. https://gcc.gnu.org/PR5388
17822 248. https://gcc.gnu.org/PR5390
17823 249. https://gcc.gnu.org/PR7877
17824 250. https://gcc.gnu.org/PR9393
17825 251. https://gcc.gnu.org/PR10032
17826 252. https://gcc.gnu.org/PR10468
17827 253. https://gcc.gnu.org/PR10527
17828 254. https://gcc.gnu.org/PR10679
17829 255. https://gcc.gnu.org/PR10682
17830 256. https://gcc.gnu.org/PR10689
17831 257. https://gcc.gnu.org/PR10845
17832 258. https://gcc.gnu.org/PR10849
17833 259. https://gcc.gnu.org/PR10888
17834 260. https://gcc.gnu.org/PR10929
17835 261. https://gcc.gnu.org/PR10931
17836 262. https://gcc.gnu.org/PR10940
17837 263. https://gcc.gnu.org/PR10968
17838 264. https://gcc.gnu.org/PR10990
17839 265. https://gcc.gnu.org/PR11039
17840 266. https://gcc.gnu.org/PR11062
17841 267. https://gcc.gnu.org/PR11095
17842 268. https://gcc.gnu.org/PR11098
17843 269. https://gcc.gnu.org/PR11137
17844 270. https://gcc.gnu.org/PR11154
17845 271. https://gcc.gnu.org/PR11329
17846 272. https://gcc.gnu.org/PR11332
17847 273. https://gcc.gnu.org/PR11431
17848 274. https://gcc.gnu.org/PR11528
17849 275. https://gcc.gnu.org/PR11546
17850 276. https://gcc.gnu.org/PR11567
17851 277. https://gcc.gnu.org/PR11645
17852 278. https://gcc.gnu.org/PR5179
17853 279. https://gcc.gnu.org/PR8204
17854 280. https://gcc.gnu.org/PR10838
17855 281. https://gcc.gnu.org/PR10886
17856 282. https://gcc.gnu.org/PR11349
17857 283. https://gcc.gnu.org/PR4823
17858 284. https://gcc.gnu.org/PR8878
17859 285. https://gcc.gnu.org/PR9815
17860 286. https://gcc.gnu.org/PR10402
17861 287. https://gcc.gnu.org/PR10504
17862 288. https://gcc.gnu.org/PR10673
17863 289. https://gcc.gnu.org/PR11044
17864 290. https://gcc.gnu.org/PR11089
17865 291. https://gcc.gnu.org/PR11420
17866 292. https://gcc.gnu.org/PR9362
17867 293. https://gcc.gnu.org/PR10142
17868 294. https://gcc.gnu.org/PR10663
17869 295. https://gcc.gnu.org/PR10835
17870 296. https://gcc.gnu.org/PR10876
17871 297. https://gcc.gnu.org/PR10955
17872 298. https://gcc.gnu.org/PR11018
17873 299. https://gcc.gnu.org/PR11556
17874 300. https://gcc.gnu.org/PR10907
17875 301. https://gcc.gnu.org/PR11320
17876 302. https://gcc.gnu.org/PR11599
17877 303. https://gcc.gnu.org/PR9745
17878 304. https://gcc.gnu.org/PR10871
17879 305. https://gcc.gnu.org/PR11440
17880 306. https://gcc.gnu.org/PR7594
17881 307. https://gcc.gnu.org/PR10557
17882 308. https://gcc.gnu.org/PR11054
17883 309. https://gcc.gnu.org/PR10834
17884 310. https://gcc.gnu.org/PR10842
17885 311. https://gcc.gnu.org/PR11052
17886 312. https://gcc.gnu.org/PR11183
17887 313. https://gcc.gnu.org/PR11084
17888 314. https://gcc.gnu.org/PR10331
17889 315. https://gcc.gnu.org/PR10413
17890 316. https://gcc.gnu.org/PR11096
17891 317. https://gcc.gnu.org/PR2873
17892 318. https://gcc.gnu.org/PR3163
17893 319. https://gcc.gnu.org/PR5287
17894 320. https://gcc.gnu.org/PR10148
17895 321. https://gcc.gnu.org/PR8787
17896 322. https://gcc.gnu.org/PR10900
17897 323. https://gcc.gnu.org/PR1607
17898 324. https://gcc.gnu.org/PR4252
17899 325. https://gcc.gnu.org/PR4490
17900 326. https://gcc.gnu.org/PR10355
17901 327. https://gcc.gnu.org/PR10726
17902 328. https://gcc.gnu.org/PR10805
17903 329. https://gcc.gnu.org/PR10815
17904 330. https://gcc.gnu.org/PR10877
17905 331. https://gcc.gnu.org/PR11280
17906 332. https://gcc.gnu.org/PR11466
17907 333. https://gcc.gnu.org/PR10737
17908 334. https://gcc.gnu.org/PR10810
17909 335. https://gcc.gnu.org/PR8336
17910 336. https://gcc.gnu.org/PR9330
17911 337. https://gcc.gnu.org/PR9631
17912 338. https://gcc.gnu.org/PR9877
17913 339. https://gcc.gnu.org/PR11687
17914 340. https://gcc.gnu.org/PR12263
17915 341. https://gcc.gnu.org/PR12490
17916 342. https://gcc.gnu.org/PR7277
17917 343. https://gcc.gnu.org/PR7939
17918 344. https://gcc.gnu.org/PR11063
17919 345. https://gcc.gnu.org/PR11207
17920 346. https://gcc.gnu.org/PR11522
17921 347. https://gcc.gnu.org/PR11595
17922 348. https://gcc.gnu.org/PR11646
17923 349. https://gcc.gnu.org/PR11665
17924 350. https://gcc.gnu.org/PR11852
17925 351. https://gcc.gnu.org/PR11878
17926 352. https://gcc.gnu.org/PR11883
17927 353. https://gcc.gnu.org/PR11991
17928 354. https://gcc.gnu.org/PR12146
17929 355. https://gcc.gnu.org/PR12215
17930 356. https://gcc.gnu.org/PR12369
17931 357. https://gcc.gnu.org/PR12446
17932 358. https://gcc.gnu.org/PR12510
17933 359. https://gcc.gnu.org/PR12544
17934 360. https://gcc.gnu.org/PR9862
17935 361. https://gcc.gnu.org/PR10962
17936 362. https://gcc.gnu.org/PR11370
17937 363. https://gcc.gnu.org/PR11637
17938 364. https://gcc.gnu.org/PR11885
17939 365. https://gcc.gnu.org/PR12082
17940 366. https://gcc.gnu.org/PR12180
17941 367. https://gcc.gnu.org/PR12340
17942 368. https://gcc.gnu.org/PR3907
17943 369. https://gcc.gnu.org/PR5293
17944 370. https://gcc.gnu.org/PR5296
17945 371. https://gcc.gnu.org/PR7939
17946 372. https://gcc.gnu.org/PR8656
17947 373. https://gcc.gnu.org/PR10147
17948 374. https://gcc.gnu.org/PR11400
17949 375. https://gcc.gnu.org/PR11409
17950 376. https://gcc.gnu.org/PR11740
17951 377. https://gcc.gnu.org/PR11786
17952 378. https://gcc.gnu.org/PR11867
17953 379. https://gcc.gnu.org/PR11928
17954 380. https://gcc.gnu.org/PR12114
17955 381. https://gcc.gnu.org/PR12163
17956 382. https://gcc.gnu.org/PR12181
17957 383. https://gcc.gnu.org/PR12236
17958 384. https://gcc.gnu.org/PR12266
17959 385. https://gcc.gnu.org/PR12296
17960 386. https://gcc.gnu.org/PR12298
17961 387. https://gcc.gnu.org/PR12369
17962 388. https://gcc.gnu.org/PR12337
17963 389. https://gcc.gnu.org/PR12344
17964 390. https://gcc.gnu.org/PR12451
17965 391. https://gcc.gnu.org/PR12486
17966 392. https://gcc.gnu.org/PR8869
17967 393. https://gcc.gnu.org/PR9786
17968 394. https://gcc.gnu.org/PR11689
17969 395. https://gcc.gnu.org/PR12116
17970 396. https://gcc.gnu.org/PR12070
17971 397. https://gcc.gnu.org/PR11184
17972 398. https://gcc.gnu.org/PR11535
17973 399. https://gcc.gnu.org/PR11693
17974 400. https://gcc.gnu.org/PR12224
17975 401. https://gcc.gnu.org/PR11087
17976 402. https://gcc.gnu.org/PR11319
17977 403. https://gcc.gnu.org/PR11949
17978 404. https://gcc.gnu.org/PR11662
17979 405. https://gcc.gnu.org/PR11965
17980 406. https://gcc.gnu.org/PR12301
17981 407. https://gcc.gnu.org/PR11717
17982 408. https://gcc.gnu.org/PR11313
17983 409. https://gcc.gnu.org/PR11712
17984 410. https://gcc.gnu.org/PR12166
17985 411. https://gcc.gnu.org/PR12101
17986 412. https://gcc.gnu.org/PR10988
17987 413. https://gcc.gnu.org/PR11805
17988 414. https://gcc.gnu.org/PR11902
17989 415. https://gcc.gnu.org/PR11903
17990 416. https://gcc.gnu.org/PR11890
17991 417. https://gcc.gnu.org/PR12399
17992 418. https://gcc.gnu.org/PR13068
17993 419. https://gcc.gnu.org/PR10060
17994 420. https://gcc.gnu.org/PR10555
17995 421. https://gcc.gnu.org/PR10706
17996 422. https://gcc.gnu.org/PR11496
17997 423. https://gcc.gnu.org/PR11741
17998 424. https://gcc.gnu.org/PR12440
17999 425. https://gcc.gnu.org/PR12632
18000 426. https://gcc.gnu.org/PR12712
18001 427. https://gcc.gnu.org/PR12726
18002 428. https://gcc.gnu.org/PR12890
18003 429. https://gcc.gnu.org/PR12900
18004 430. https://gcc.gnu.org/PR13060
18005 431. https://gcc.gnu.org/PR13289
18006 432. https://gcc.gnu.org/PR13318
18007 433. https://gcc.gnu.org/PR13392
18008 434. https://gcc.gnu.org/PR13574
18009 435. https://gcc.gnu.org/PR13475
18010 436. https://gcc.gnu.org/PR13797
18011 437. https://gcc.gnu.org/PR13824
18012 438. https://gcc.gnu.org/PR8776
18013 439. https://gcc.gnu.org/PR10339
18014 440. https://gcc.gnu.org/PR11350
18015 441. https://gcc.gnu.org/PR12826
18016 442. https://gcc.gnu.org/PR12500
18017 443. https://gcc.gnu.org/PR12941
18018 444. https://gcc.gnu.org/PR12953
18019 445. https://gcc.gnu.org/PR13041
18020 446. https://gcc.gnu.org/PR13507
18021 447. https://gcc.gnu.org/PR13382
18022 448. https://gcc.gnu.org/PR13394
18023 449. https://gcc.gnu.org/PR13400
18024 450. https://gcc.gnu.org/PR13521
18025 451. https://gcc.gnu.org/PR2094
18026 452. https://gcc.gnu.org/PR2294
18027 453. https://gcc.gnu.org/PR5050
18028 454. https://gcc.gnu.org/PR9371
18029 455. https://gcc.gnu.org/PR9546
18030 456. https://gcc.gnu.org/PR10081
18031 457. https://gcc.gnu.org/PR10093
18032 458. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61
18033 459. https://gcc.gnu.org/PR10095
18034 460. https://gcc.gnu.org/PR11554
18035 461. https://gcc.gnu.org/PR12297
18036 462. https://gcc.gnu.org/PR12352
18037 463. https://gcc.gnu.org/PR12438
18038 464. https://gcc.gnu.org/PR12540
18039 465. https://gcc.gnu.org/PR12594
18040 466. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60
18041 467. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63
18042 468. https://gcc.gnu.org/PR12657
18043 469. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292
18044 470. https://gcc.gnu.org/PR12696
18045 471. https://gcc.gnu.org/PR12815
18046 472. https://gcc.gnu.org/PR12862
18047 473. https://gcc.gnu.org/PR12926
18048 474. https://gcc.gnu.org/PR12967
18049 475. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html
18050 476. https://gcc.gnu.org/PR12971
18051 477. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328
18052 478. https://gcc.gnu.org/PR13007
18053 479. https://gcc.gnu.org/PR13009
18054 480. https://gcc.gnu.org/PR13057
18055 481. https://gcc.gnu.org/PR13070
18056 482. https://gcc.gnu.org/PR13081
18057 483. https://gcc.gnu.org/PR13239
18058 484. https://gcc.gnu.org/PR13262
18059 485. https://gcc.gnu.org/PR13290
18060 486. https://gcc.gnu.org/PR13323
18061 487. https://gcc.gnu.org/PR13369
18062 488. https://gcc.gnu.org/PR13371
18063 489. https://gcc.gnu.org/PR13445
18064 490. https://gcc.gnu.org/PR13461
18065 491. https://gcc.gnu.org/PR13462
18066 492. https://gcc.gnu.org/PR13478
18067 493. https://gcc.gnu.org/PR13544
18068 494. https://gcc.gnu.org/PR13650
18069 495. https://gcc.gnu.org/PR13683
18070 496. https://gcc.gnu.org/PR13688
18071 497. https://gcc.gnu.org/PR13774
18072 498. https://gcc.gnu.org/PR13884
18073 499. https://gcc.gnu.org/PR10746
18074 500. https://gcc.gnu.org/PR11433
18075 501. https://gcc.gnu.org/PR12633
18076 502. https://gcc.gnu.org/PR13037
18077 503. https://gcc.gnu.org/PR13213
18078 504. https://gcc.gnu.org/PR4490
18079 505. https://gcc.gnu.org/PR12292
18080 506. https://gcc.gnu.org/PR12441
18081 507. https://gcc.gnu.org/PR12943
18082 508. https://gcc.gnu.org/PR13608
18083 509. https://gcc.gnu.org/PR11598
18084 510. https://gcc.gnu.org/PR11793
18085 511. https://gcc.gnu.org/PR12467
18086 512. https://gcc.gnu.org/PR12537
18087 513. https://gcc.gnu.org/PR12496
18088 514. https://gcc.gnu.org/PR12865
18089 515. https://gcc.gnu.org/PR13354
18090 516. https://gcc.gnu.org/PR10467
18091 517. https://gcc.gnu.org/PR11226
18092 518. https://gcc.gnu.org/PR11227
18093 519. https://gcc.gnu.org/PR12644
18094 520. https://gcc.gnu.org/PR13149
18095 521. https://gcc.gnu.org/PR12654
18096 522. https://gcc.gnu.org/PR12965
18097 523. https://gcc.gnu.org/PR13031
18098 524. https://gcc.gnu.org/PR11634
18099 525. https://gcc.gnu.org/PR12158
18100 526. https://gcc.gnu.org/PR11992
18101 527. https://gcc.gnu.org/PR9365
18102 528. https://gcc.gnu.org/PR10392
18103 529. https://gcc.gnu.org/PR11322
18104 530. https://gcc.gnu.org/PR13069
18105 531. https://gcc.gnu.org/PR13302
18106 532. https://gcc.gnu.org/PR13585
18107 533. https://gcc.gnu.org/PR8916
18108 534. https://gcc.gnu.org/PR11576
18109 535. https://gcc.gnu.org/PR13122
18110 536. https://gcc.gnu.org/PR13256
18111 537. https://gcc.gnu.org/PR13373
18112 538. https://gcc.gnu.org/PR12561
18113 539. https://gcc.gnu.org/PR6243
18114 540. https://gcc.gnu.org/PR11397
18115 541. https://gcc.gnu.org/PR12505
18116 542. https://gcc.gnu.org/PR13150
18117 543. https://gcc.gnu.org/PR12666
18118 544. https://gcc.gnu.org/PR12969
18119 545. https://gcc.gnu.org/PR10819
18120 546. https://gcc.gnu.org/PR11612
18121 547. https://gcc.gnu.org/PR13211
18122 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4
18123 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5
18124 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6
18125 551. https://gcc.gnu.org/onlinedocs/
18126 552. mailto:gcc-help@gcc.gnu.org
18127 553. mailto:gcc@gcc.gnu.org
18128 554. https://gcc.gnu.org/lists.html
18129 555. https://www.fsf.org/
18130 556. https://gcc.gnu.org/about.html
18131======================================================================
18132http://gcc.gnu.org/gcc-3.2/index.html
18133
18134                             GCC 3.2 Release Series
18135
18136   (This release series is no longer supported.)
18137
18138   April 25, 2003
18139
18140   The [1]GNU project and the GCC developers are pleased to announce the
18141   release of GCC 3.2.3.
18142
18143   The purpose of the GCC 3.2 release series is to provide a stable
18144   platform for OS distributors to use building their next releases. A
18145   primary objective was to stabilize the C++ ABI; we believe that the
18146   interface to the compiler and the C++ standard library are now
18147   relatively stable.
18148
18149   Be aware that C++ code compiled by GCC 3.2.x will (in general) not
18150   interoperate with code compiled by GCC 3.1.1 or earlier.
18151
18152   Please refer to our [2]detailed list of news, caveats, and bug-fixes
18153   for further information.
18154
18155Release History
18156
18157   GCC 3.2.3
18158          April 25, 2003 ([3]changes)
18159
18160   GCC 3.2.2
18161          February 5, 2003 ([4]changes)
18162
18163   GCC 3.2.1
18164          November 19, 2002 ([5]changes)
18165
18166   GCC 3.2
18167          August 14, 2002 ([6]changes)
18168
18169References and Acknowledgements
18170
18171   GCC used to stand for the GNU C Compiler, but since the compiler
18172   supports several other languages aside from C, it now stands for the
18173   GNU Compiler Collection.
18174
18175   A list of [7]successful builds is updated as new information becomes
18176   available.
18177
18178   The GCC developers would like to thank the numerous people that have
18179   contributed new features, improvements, bug fixes, and other changes as
18180   well as test results to GCC. This [8]amazing group of volunteers is
18181   what makes GCC successful.
18182
18183   For additional information about GCC please refer to the [9]GCC project
18184   web site or contact the [10]GCC development mailing list.
18185
18186   To obtain GCC please use [11]our mirror sites, or our CVS server.
18187
18188
18189    For questions related to the use of GCC, please consult these web
18190    pages and the [12]GCC manuals. If that fails, the
18191    [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18192    web pages and the development of GCC are welcome on our developer
18193    list at [14]gcc@gcc.gnu.org. All of [15]our lists have public
18194    archives.
18195
18196   Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and
18197   distribution of this entire article is permitted in any medium,
18198   provided this notice is preserved.
18199
18200   These pages are [17]maintained by the GCC team. Last modified
18201   2022-10-26.
18202
18203References
18204
18205   1. http://www.gnu.org/
18206   2. http://gcc.gnu.org/gcc-3.2/changes.html
18207   3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
18208   4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2
18209   5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1
18210   6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2
18211   7. http://gcc.gnu.org/gcc-3.2/buildstat.html
18212   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
18213   9. http://gcc.gnu.org/index.html
18214  10. mailto:gcc@gcc.gnu.org
18215  11. http://gcc.gnu.org/mirrors.html
18216  12. https://gcc.gnu.org/onlinedocs/
18217  13. mailto:gcc-help@gcc.gnu.org
18218  14. mailto:gcc@gcc.gnu.org
18219  15. https://gcc.gnu.org/lists.html
18220  16. https://www.fsf.org/
18221  17. https://gcc.gnu.org/about.html
18222======================================================================
18223http://gcc.gnu.org/gcc-3.2/changes.html
18224
18225                             GCC 3.2 Release Series
18226                        Changes, New Features, and Fixes
18227
18228   The latest release in the 3.2 release series is [1]GCC 3.2.3.
18229
18230Caveats and New Features
18231
18232  Caveats
18233
18234     * The C++ compiler does not correctly zero-initialize
18235       pointers-to-data members. You must explicitly initialize them. For
18236       example: int S::*m(0); will work, but depending on
18237       default-initialization to zero will not work. This bug cannot be
18238       fixed in GCC 3.2 without inducing unacceptable risks. It will be
18239       fixed in GCC 3.3.
18240     * This GCC release is based on the GCC 3.1 sourcebase, and thus has
18241       all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has
18242       a number of C++ ABI fixes which make its C++ compiler generate
18243       binary code which is incompatible with the C++ compilers found in
18244       earlier GCC releases, including GCC 3.1 and GCC 3.1.1.
18245
18246  Frontend Enhancements
18247
18248    C/C++/Objective-C
18249
18250     * The method of constructing the list of directories to be searched
18251       for header files has been revised. If a directory named by a -I
18252       option is a standard system include directory, the option is
18253       ignored to ensure that the default search order for system
18254       directories and the special treatment of system header files are
18255       not defeated.
18256     * The C and Objective-C compilers no longer accept the "Naming Types"
18257       extension (typedef foo = bar); it was already unavailable in C++.
18258       Code which uses it will need to be changed to use the "typeof"
18259       extension instead: typedef typeof(bar) foo. (We have removed this
18260       extension without a period of deprecation because it has caused the
18261       compiler to crash since version 3.0 and no one noticed until very
18262       recently. Thus we conclude it is not in widespread use.)
18263
18264    C++
18265
18266     * GCC 3.2 fixed serveral differences between the C++ ABI implemented
18267       in GCC and the multi-vendor standard, but more have been found
18268       since the release. 3.2.1 adds a new warning, -Wabi, to warn about
18269       code which is affected by these bugs. We will fix these bugs in
18270       some future release, once we are confident that all have been
18271       found; until then, it is our intention to make changes to the ABI
18272       only if they are necessary for correct compilation of C++, as
18273       opposed to conformance to the ABI documents.
18274     * For details on how to build an ABI compliant compiler for GNU/Linux
18275       systems, check the [3]common C++ ABI page.
18276
18277  New Targets and Target Specific Improvements
18278
18279    IA-32
18280
18281     * Fixed a number of bugs in SSE and MMX intrinsics.
18282     * Fixed common compiler crashes with SSE instruction set enabled
18283       (implied by -march=pentium3, pentium4, athlon-xp)
18284     * __m128 and __m128i is not 128bit aligned when used in structures.
18285
18286    x86-64
18287
18288     * A bug whereby the compiler could generate bad code for bzero has
18289       been fixed.
18290     * ABI fixes (implying ABI incompatibilities with previous version in
18291       some corner cases)
18292     * Fixed prefetch code generation
18293     __________________________________________________________________
18294
18295GCC 3.2.3
18296
18297   3.2.3 is a bug fix release only; there are no new features that were
18298   not present in GCC 3.2.2.
18299
18300  Bug Fixes
18301
18302   This section lists the problem reports (PRs) from GCC's bug tracking
18303   system that are known to be fixed in the 3.2.3 release. This list might
18304   not be complete (that is, it is possible that some PRs that have been
18305   fixed are not listed here), and some of the titles have been changed to
18306   make them more clear.
18307
18308    Internal Compiler Errors (multi-platform)
18309
18310     * [4]3782: (c++) -quiet -fstats produces a segmentation fault in
18311       cc1plus
18312     * [5]6440: (c++) template specializations cause ICE
18313     * [6]7050: (c++) ICE on: (i ? get_string() : throw)
18314     * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c)
18315     * [8]7982: (c++) ICE due to infinite recursion (using STL set)
18316     * [9]8068: exceedingly high (infinite) memory usage
18317     * [10]8178: ICE with __builtin_ffs
18318     * [11]8396: ICE in copy_to_mode_reg, in explow.c
18319     * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c
18320     * [13]9768: ICE when optimizing inline code at -O2
18321     * [14]9798: (c++) Infinite recursion (segfault) in
18322       cp/decl.c:push_using_directive with recursive using directives
18323     * [15]9799: mismatching structure initializer with nested flexible
18324       array member: ICE
18325     * [16]9928: ICE on duplicate enum declaration
18326     * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects
18327       sparc, alpha)
18328     * [18]10352: ICE in find_reloads_toplev
18329     * [19]10336: ICE with -Wunreachable-code
18330
18331    C/optimizer bugs:
18332
18333     * [20]8224: Incorrect joining of signed and unsigned division
18334     * [21]8613: -O2 produces wrong code with builtin strlen and
18335       postincrements
18336     * [22]8828: gcc reports some code is unreachable when it is not
18337     * [23]9226: GCSE breaking argument passing
18338     * [24]9853: miscompilation of non-constant structure initializer
18339     * [25]9797: C99-style struct initializers are miscompiled
18340     * [26]9967: Some standard C function calls should not be replaced
18341       when optimizing for size
18342     * [27]10116: ce2: invalid merge of join_bb in the context of switch
18343       statements
18344     * [28]10171: wrong code for inlined function
18345     * [29]10175: -Wunreachable-code doesn't work for single lines
18346
18347    C++ compiler and library:
18348
18349     * [30]8316: Confusing diagnostic for code that misuses conversion
18350       operators
18351     * [31]9169: filebuf output fails if codecvt<>::out returns noconv
18352     * [32]9420: incomplete type incorrectly reported
18353     * [33]9459: typeof in return type specification of template not
18354       supported
18355     * [34]9507: filebuf::open handles ios_base::ate incorrectly
18356     * [35]9538: Out-of-bounds memory access in streambuf::sputbackc
18357     * [36]9602: Total confusion about template/friend/virtual/abstract
18358     * [37]9993: destructor not called for local object created within and
18359       returned from infinite loop
18360     * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2
18361       system
18362
18363    Java compiler and library:
18364
18365     * [39]9652: libgcj build fails on irix6.5.1[78]
18366     * [40]10144: gas on solaris complains about bad .stabs lines for
18367       java, native as unaffected
18368
18369    x86-specific (Intel/AMD):
18370
18371     * [41]8746: gcc miscompiles Linux kernel ppa driver on x86
18372     * [42]9888: -mcpu=k6 -Os produces out of range loop instructions
18373     * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu
18374       failed
18375     * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib)
18376       failed
18377
18378    SPARC-specific:
18379
18380     * [45]7784: [Sparc] ICE in extract_insn, in recog.c
18381     * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in
18382       unroll.c
18383     * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc
18384     * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in
18385       execute/loop-2d.c
18386     * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc
18387     * [50]9414: Scheduling bug on Ultrasparc
18388     * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64
18389
18390    m68k-specific:
18391
18392     * [52]7248: broken "inclusive or" code
18393     * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1
18394
18395    PowerPC-specific:
18396
18397     * [54]9732: Wrong code with -O2 -fPIC
18398     * [55]10073: ICE: powerpc cannot split insn
18399
18400    Alpha-specific:
18401
18402     * [56]7702: optimization problem on a DEC alpha under OSF1
18403     * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system
18404
18405    HP-specific:
18406
18407     * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275)
18408     * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10
18409       (missing symbol)
18410     * [60]10271: Floating point args don't get reloaded across function
18411       calls with -O2
18412
18413    MIPS specific:
18414
18415     * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in
18416       compile/920501-4.c
18417
18418    CRIS specific:
18419
18420     * [62]10377: gcc-3.2.2 creates bad assembler code for cris
18421
18422    Miscellaneous and minor bugs:
18423
18424     * [63]6955: collect2 says "core dumped" when there is no core
18425     __________________________________________________________________
18426
18427GCC 3.2.2
18428
18429   Beginning with 3.2.2, GCC's Makefile suite supports redirection of make
18430   install by means of the DESTDIR variable. Parts of the GCC tree have
18431   featured that support long before, but now it is available even from
18432   the top level.
18433
18434   Other than that, GCC 3.2.2 is a bug fix release only; there are no new
18435   features that were not present in GCC 3.2.1.
18436
18437  Bug Fixes
18438
18439   On the following i386-based systems GCC 3.2.1 broke the C ABI wrt.
18440   functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped
18441   with FreeBSD 5.0 does not have this problem), Interix, a.out-based
18442   GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI
18443   change, and thus restores ABI-compatibility with previous releases
18444   (except GCC 3.2.1) on these platforms.
18445
18446   This section lists the problem reports (PRs) from GCC's bug tracking
18447   system that are known to be fixed in the 3.2.2 release. This list might
18448   not be complete (that is, it is possible that some PRs that have been
18449   fixed are not listed here) and some of the titles have been changed to
18450   make them more clear.
18451
18452    Internal Compiler Errors (multi-platform)
18453
18454     * [64]5919: (c++) ICE when passing variable array to template
18455       function
18456     * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=)
18457     * [66]7507: ICE with -O2 when address of called function is a
18458       complicated expression
18459     * [67]7622: ICE with nested inline functions if function's address is
18460       taken
18461     * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR
18462       [69]9258)
18463     * [70]8031: (c++) ICE in code comparing typeids and casting from
18464       virtual base
18465     * [71]8275: ICE in simplify_subreg
18466     * [72]8332: (c++) builtin strlen/template interaction causes ICE
18467     * [73]8372: (c++) ICE on explicit call of destructor
18468     * [74]8439: (c, not c++) empty struct causes ICE
18469     * [75]8442: (c++) ICE with nested template classes
18470     * [76]8518: ICE when compiling mplayer ("extern inline" issue)
18471     * [77]8615: (c++) ICE with out-of-range character constant template
18472       argument
18473     * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307
18474     * [79]8799: (c++) ICE: error reporting routines re-entered
18475     * [80]9328: (c++) ICE with typeof(X) for overloaded X
18476     * [81]9465: (preprocessor) cpp -traditional ICE on null bytes
18477
18478    C++ (compiler and library) bugs
18479
18480     * [82]47: scoping in nested classes is broken
18481     * [83]6745: problems with iostream rdbuf() member function
18482     * [84]8214: conversion from const char* const to char* sometimes
18483       accepted illegally
18484     * [85]8493: builtin strlen and overload resolution (same bug as
18485       [86]8332)
18486     * [87]8503: strange behaviour of function types
18487     * [88]8727: compiler confused by inheritance from an anonymous struct
18488     * [89]7445: poor performance of std::locale::classic() in
18489       multi-threaded applications
18490     * [90]8230: mishandling of overflow in vector<T>::resize
18491     * [91]8399: sync_with_stdio(false) breaks unformatted input
18492     * [92]8662: illegal access of private member of unnamed class is
18493       accepted
18494     * [93]8707: "make distclean" fails in libstdc++-v3 directory
18495     * [94]8708: __USE_MALLOC doesn't work
18496     * [95]8790: Use of non-thread-safe strtok in src/localename.cc
18497     * [96]8887: Bug in date formats with --enable-clocale=generic
18498     * [97]9076: Call Frame Instructions are not handled correctly during
18499       unwind operation
18500     * [98]9151: std::setprecision limited to 16 digits when outputting a
18501       double to a stream
18502     * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers
18503     * [100]9269: libstdc++ headers: explicit specialization of function
18504       must precede its first use
18505     * [101]9322: return value of basic_streambuf<>::getloc affected by
18506       locale::global
18507     * [102]9433: segfault in runtime support for dynamic_cast
18508
18509    C and optimizer bugs
18510
18511     * [103]8032: GCC incorrectly initializes static structs that have
18512       flexible arrays
18513     * [104]8639: simple arithmetic expression broken
18514     * [105]8794: optimization improperly eliminates certain expressions
18515     * [106]8832: traditional "asm volatile" code is illegally optimized
18516     * [107]8988: loop optimizer bug: with -O2, code is generated that
18517       segfaults (found on i386, bug present for all platforms)
18518     * [108]9492: structure copy clobbers subsequent stores to structure
18519
18520    Objective-C bugs
18521
18522     * [109]9267: Objective-C parser won't build with newer bison versions
18523       (e.g. 1.875)
18524
18525    Ada bugs
18526
18527     * [110]8344: Ada build problem due to conflict between gcc/final.o,
18528       gcc/ada/final.o
18529
18530    Preprocessor bugs
18531
18532     * [111]8524: _Pragma within macros is improperly expanded
18533     * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with
18534       -fshort-wchar
18535
18536    ARM-specific
18537
18538     * [113]9090: arm ICE with >= -O2; regression from gcc-2.95
18539
18540    x86-specific (Intel/AMD)
18541
18542     * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction)
18543     * [115]8599: loop unroll bug with -march=k6-3
18544     * [116]9506: ABI breakage in structure return (affects BSD and
18545       Cygwin, but not GNU/Linux)
18546
18547    FreeBSD 5.0 specific
18548
18549     * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0
18550
18551    RTEMS-specific
18552
18553     * [118]9292: hppa1.1-rtems configurery problems
18554     * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug
18555     * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue
18556     * [121]9296: gthr-rtems regression
18557     * [122]9316: powerpc-rtems: extending multilibs
18558
18559    HP-PA specific
18560
18561     * [123]9493: ICE with -O2 when building a simple function
18562
18563    Documentation
18564
18565     * [124]7341: hyperlink to gcov in GCC documentation doesn't work
18566     * [125]8947: Please add a warning about "-malign-double" in docs
18567     * [126]7448, [127]8882: typo cleanups
18568     __________________________________________________________________
18569
18570GCC 3.2.1
18571
18572   3.2.1 adds a new warning, -Wabi. This option warns when GNU C++
18573   generates code that is known not to be binary-compatible with the
18574   vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included
18575   in the distribution, for details.
18576
18577   This release also removes an old GCC extension, "naming types", and the
18578   documentation now directs users to use a different GCC extension,
18579   __typeof__, instead. The feature had evidently been broken for a while.
18580
18581   Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and
18582   the new warning there are no new features that were not present in GCC
18583   3.2.
18584
18585   In addition, the previous fix for [128]PR 7445 (poor performance of
18586   std::locale::classic() in multi-threaded applications) was reverted
18587   ("unfixed"), because the "fix" was not thread-safe.
18588
18589  Bug Fixes
18590
18591   This section lists the problem reports (PRs) from GCC's bug tracking
18592   system that are known to be fixed in the 3.2.1 release. This list might
18593   not be complete (that is, it is possible that some PRs that have been
18594   fixed are not listed here). As you can see, the number of bug fixes is
18595   quite large, so it is strongly recommended that users of earlier GCC
18596   3.x releases upgrade to GCC 3.2.1.
18597
18598    Internal Compiler Errors (multi-platform)
18599
18600     * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c
18601     * [130]5661: (c++) ICE instantiating template on array of unknown
18602       size (bad code)
18603     * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on
18604       64-bit platforms
18605     * [132]6994: (c++) ICE in find_function_data
18606     * [133]7150: preprocessor: GCC -dM -E gives an ICE
18607     * [134]7160: ICE when optimizing branches without a return value
18608     * [135]7228: (c++) ICE when using member template and template
18609       function
18610     * [136]7266: (c++) ICE with -pedantic on missing typename
18611     * [137]7353: ICE from use of "Naming Types" extension, see above
18612     * [138]7411: ICE in instantiate_virtual_regs_1, in function.c
18613     * [139]7478: (c++) ICE on static_cast inside template
18614     * [140]7526: preprocessor core dump when _Pragma implies #pragma
18615       dependency
18616     * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803
18617       is a duplicate)
18618     * [143]7754: (c++) ICE on union with template parameter
18619     * [144]7788: (c++) redeclaring a definition as an incomplete class
18620       causes ICE
18621     * [145]8031: (c++) ICE in comptypes, in cp/typeck.c
18622     * [146]8055: preprocessor dies with SIG11 when building FreeBSD
18623       kernel
18624     * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related
18625       variables
18626     * [148]8134: (c++) ICE in force_store_init_value on legal code
18627     * [149]8149: (c++) ICE on incomplete type
18628     * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array
18629       initialization
18630
18631    C++ (compiler and library) bugs
18632
18633     * [151]5607: No pointer adjustment in covariant return types
18634     * [152]6579: Infinite loop with statement expressions in member
18635       initialization
18636     * [153]6803: Default copy constructor bug in GCC 3.1
18637     * [154]7176: g++ confused by friend and static member with same name
18638     * [155]7188: Segfault with template class and recursive (incorrect)
18639       initializer list
18640     * [156]7306: Regression: GCC 3.x fails to compile code with virtual
18641       inheritance if a method has a variable number of arguments
18642     * [157]7461: ctype<char>::classic_table() returns offset array on
18643       Cygwin
18644     * [158]7524: f(const float arg[3]) fails
18645     * [159]7584: Erroneous ambiguous base error on using declaration
18646     * [160]7676: Member template overloading problem
18647     * [161]7679: infinite loop when a right parenthesis is missing
18648     * [162]7811: default locale not taken from environment
18649     * [163]7961: compare( char *) implemented incorrectly in
18650       basic_string<>
18651     * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if
18652       streambuf::underflow() leaves gptr() NULL (dups: [165]8127,
18653       [166]6745)
18654     * [167]8096: deque::at() throws std::range_error instead of
18655       std::out_of_range
18656     * [168]8127: cout << cin.rdbuf() infinite loop
18657     * [169]8218: Excessively large memory consumed for classes with large
18658       array members
18659     * [170]8287: GCC 3.2: Destructor called for non-constructed local
18660       object
18661     * [171]8347: empty vector range used in string construction causes
18662       core dump
18663     * [172]8348: fail() flag is set in istringstream when eof() flag is
18664       set
18665     * [173]8391: regression: infinite loop in cp/decl2.c(finish_file)
18666
18667    C and optimizer bugs
18668
18669     * [174]6627: -fno-align-functions doesn't seem to disable function
18670       alignment
18671     * [175]6631: life_analysis misoptimizes code to initialize fields of
18672       a structure
18673     * [176]7102: unsigned char division results in floating exception
18674     * [177]7120: Run once loop should *always* be unrolled
18675       (pessimization)
18676     * [178]7209: Bug involving array referencing and ?: operator
18677     * [179]7515: invalid inlining of global function with -O3
18678     * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test
18679     * [181]8467: bug in sibling call optimization
18680
18681    Preprocessor bugs
18682
18683     * [182]4890: incorrect line markers from the traditional preprocessor
18684     * [183]7357: -M option omits system headers files (making it the same
18685       as -MM)
18686     * [184]7358: Changes to Sun's make Dependencies
18687     * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as
18688       C headers
18689     * [186]7862: Interrupting GCC -MD removes .d file but not .o
18690     * [187]8190: Failed compilation deletes -MD dependency file
18691     * [188]8524: _Pragma within macro is improperly expanded
18692
18693    x86 specific (Intel/AMD)
18694
18695     * [189]5351: (i686-only) function pass-by-value structure copy
18696       corrupts stack ([190]7591 is a duplicate)
18697     * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with
18698       -march=pentium3/pentium2/athlon (these are all the same underlying
18699       bug, in MMX register use)
18700     * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same
18701       as above?)
18702     * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken
18703     * [199]6981: wrong code in 64-bit manipulation on x86
18704     * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__
18705       macro
18706     * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE
18707       intrinsics are broken
18708     * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with
18709       -march=pentium4
18710     * [203]7693: Typo in i386 mmintrin.h header
18711     * [204]7723: ICE - Pentium3 sse - GCC 3.2
18712     * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse
18713     * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3
18714
18715    PowerPC specific
18716
18717     * [207]5967: GCC bug when profiling nested functions on powerpc
18718     * [208]6984: wrong code generated with -O2, -O3, -Os for do-while
18719       loop on PowerPC
18720     * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5
18721     * [210]7130: miscompiled code for GCC-3.1 on
18722       powerpc-unknown-linux-gnu with -funroll-all-loops
18723     * [211]7133: PowerPC ICE: unrecognizable insn
18724     * [212]7380: ICE in extract_insn, at recog.c:2148
18725     * [213]8252: ICE on Altivec code with optimization turned on
18726     * [214]8451: Altivec ICE in GCC 3.2
18727
18728    HP/PA specific
18729
18730     * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa
18731
18732    SPARC specific
18733
18734     * [216]6668: when using --disable-multilib, libgcc_s.so is installed
18735       in the wrong place on sparc-solaris
18736     * [217]7151: ICE when compiling for UltraSPARC
18737     * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long
18738       double and -O1
18739     * [219]7842: [REGRESSION] SPARC code gen bug
18740
18741    ARM specific
18742
18743     * [220]7856: [arm] invalid offset in constant pool reference
18744     * [221]7967: optimization produces wrong code (ARM)
18745
18746    Alpha specific
18747
18748     * [222]7374: __builtin_fabsl broken on alpha
18749
18750    IBM s390 specific
18751
18752     * [223]7370: ICE in fixup_var_refs_1 on s390x
18753     * [224]7409: loop optimization bug on s390x-linux-gnu
18754     * [225]8232: s390x: ICE when using bcmp with int length argument
18755
18756    SCO specific
18757
18758     * [226]7623: SCO OpenServer build fails with machmode.def: undefined
18759       symbol: BITS_PER_UNIT
18760
18761    m68k/Coldfire specific
18762
18763     * [227]8314: crtbegin, crtend need to be multilib'ed for this
18764       platform
18765
18766    Documentation
18767
18768     * [228]761: Document some undocumented options
18769     * [229]5610: Fix documentation about invoking SSE instructions
18770       (-mfpmath=sse)
18771     * [230]7484: List -Wmissing-declarations as C-only option
18772     * [231]7531: -mcmodel not documented for x86-64
18773     * [232]8120: Update documentation of bad use of ##
18774     __________________________________________________________________
18775
18776GCC 3.2
18777
18778   3.2 is a small bug fix release, but there is a change to the
18779   application binary interface (ABI), hence the change to the second part
18780   of the version number.
18781
18782   The main purpose of the 3.2 release is to correct a couple of problems
18783   in the C++ ABI, with the intention of providing a stable interface
18784   going forward.  Accordingly, 3.2 is only a small change to 3.1.1.
18785
18786  Bug Fixes
18787
18788    C++
18789
18790     * [233]7320: g++ 3.2 relocation problem
18791     * [234]7470: vtable: virtual function pointers not in declaration
18792       order
18793
18794    libstdc++
18795
18796     * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t
18797     * [236]6503, [237]6642, [238]7186: Problems with comparing or
18798       subtracting various types of const and non-const iterators
18799     * [239]7216: ambiguity with basic_iostream::traits_type
18800     * [240]7220: problem with basic_istream::ignore(0,delimiter)
18801     * [241]7222: locale::operator==() doesn't work on std::locale("")
18802     * [242]7286: placement operator delete issue
18803     * [243]7442: cxxabi.h does not match the C++ ABI
18804     * [244]7445: poor performance of std::locale::classic() in
18805       multi-threaded applications
18806
18807    x86-64 specific
18808
18809     * [245]7291: off-by-one in generated inline bzero code for x86-64
18810
18811
18812    For questions related to the use of GCC, please consult these web
18813    pages and the [246]GCC manuals. If that fails, the
18814    [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18815    web pages and the development of GCC are welcome on our developer
18816    list at [248]gcc@gcc.gnu.org. All of [249]our lists have public
18817    archives.
18818
18819   Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and
18820   distribution of this entire article is permitted in any medium,
18821   provided this notice is preserved.
18822
18823   These pages are [251]maintained by the GCC team. Last modified
18824   2022-10-26.
18825
18826References
18827
18828   1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
18829   2. http://gcc.gnu.org/gcc-3.1/changes.html
18830   3. http://gcc.gnu.org/gcc-3.2/c++-abi.html
18831   4. https://gcc.gnu.org/PR3782
18832   5. https://gcc.gnu.org/PR6440
18833   6. https://gcc.gnu.org/PR7050
18834   7. https://gcc.gnu.org/PR7741
18835   8. https://gcc.gnu.org/PR7982
18836   9. https://gcc.gnu.org/PR8068
18837  10. https://gcc.gnu.org/PR8178
18838  11. https://gcc.gnu.org/PR8396
18839  12. https://gcc.gnu.org/PR8674
18840  13. https://gcc.gnu.org/PR9768
18841  14. https://gcc.gnu.org/PR9798
18842  15. https://gcc.gnu.org/PR9799
18843  16. https://gcc.gnu.org/PR9928
18844  17. https://gcc.gnu.org/PR10114
18845  18. https://gcc.gnu.org/PR10352
18846  19. https://gcc.gnu.org/PR10336
18847  20. https://gcc.gnu.org/PR8224
18848  21. https://gcc.gnu.org/PR8613
18849  22. https://gcc.gnu.org/PR8828
18850  23. https://gcc.gnu.org/PR9226
18851  24. https://gcc.gnu.org/PR9853
18852  25. https://gcc.gnu.org/PR9797
18853  26. https://gcc.gnu.org/PR9967
18854  27. https://gcc.gnu.org/PR10116
18855  28. https://gcc.gnu.org/PR10171
18856  29. https://gcc.gnu.org/PR10175
18857  30. https://gcc.gnu.org/PR8316
18858  31. https://gcc.gnu.org/PR9169
18859  32. https://gcc.gnu.org/PR9420
18860  33. https://gcc.gnu.org/PR9459
18861  34. https://gcc.gnu.org/PR9507
18862  35. https://gcc.gnu.org/PR9538
18863  36. https://gcc.gnu.org/PR9602
18864  37. https://gcc.gnu.org/PR9993
18865  38. https://gcc.gnu.org/PR10167
18866  39. https://gcc.gnu.org/PR9652
18867  40. https://gcc.gnu.org/PR10144
18868  41. https://gcc.gnu.org/PR8746
18869  42. https://gcc.gnu.org/PR9888
18870  43. https://gcc.gnu.org/PR9638
18871  44. https://gcc.gnu.org/PR9954
18872  45. https://gcc.gnu.org/PR7784
18873  46. https://gcc.gnu.org/PR7796
18874  47. https://gcc.gnu.org/PR8281
18875  48. https://gcc.gnu.org/PR8366
18876  49. https://gcc.gnu.org/PR8726
18877  50. https://gcc.gnu.org/PR9414
18878  51. https://gcc.gnu.org/PR10067
18879  52. https://gcc.gnu.org/PR7248
18880  53. https://gcc.gnu.org/PR8343
18881  54. https://gcc.gnu.org/PR9732
18882  55. https://gcc.gnu.org/PR10073
18883  56. https://gcc.gnu.org/PR7702
18884  57. https://gcc.gnu.org/PR9671
18885  58. https://gcc.gnu.org/PR8694
18886  59. https://gcc.gnu.org/PR9953
18887  60. https://gcc.gnu.org/PR10271
18888  61. https://gcc.gnu.org/PR6362
18889  62. https://gcc.gnu.org/PR10377
18890  63. https://gcc.gnu.org/PR6955
18891  64. https://gcc.gnu.org/PR5919
18892  65. https://gcc.gnu.org/PR7129
18893  66. https://gcc.gnu.org/PR7507
18894  67. https://gcc.gnu.org/PR7622
18895  68. https://gcc.gnu.org/PR7681
18896  69. https://gcc.gnu.org/PR9528
18897  70. https://gcc.gnu.org/PR8031
18898  71. https://gcc.gnu.org/PR8275
18899  72. https://gcc.gnu.org/PR8332
18900  73. https://gcc.gnu.org/PR8372
18901  74. https://gcc.gnu.org/PR8439
18902  75. https://gcc.gnu.org/PR8442
18903  76. https://gcc.gnu.org/PR8518
18904  77. https://gcc.gnu.org/PR8615
18905  78. https://gcc.gnu.org/PR8663
18906  79. https://gcc.gnu.org/PR8799
18907  80. https://gcc.gnu.org/PR9328
18908  81. https://gcc.gnu.org/PR9465
18909  82. https://gcc.gnu.org/PR47
18910  83. https://gcc.gnu.org/PR6745
18911  84. https://gcc.gnu.org/PR8214
18912  85. https://gcc.gnu.org/PR8493
18913  86. https://gcc.gnu.org/PR8332
18914  87. https://gcc.gnu.org/PR8503
18915  88. https://gcc.gnu.org/PR8727
18916  89. https://gcc.gnu.org/PR7445
18917  90. https://gcc.gnu.org/PR8230
18918  91. https://gcc.gnu.org/PR8399
18919  92. https://gcc.gnu.org/PR8662
18920  93. https://gcc.gnu.org/PR8707
18921  94. https://gcc.gnu.org/PR8708
18922  95. https://gcc.gnu.org/PR8790
18923  96. https://gcc.gnu.org/PR8887
18924  97. https://gcc.gnu.org/PR9076
18925  98. https://gcc.gnu.org/PR9151
18926  99. https://gcc.gnu.org/PR9168
18927 100. https://gcc.gnu.org/PR9269
18928 101. https://gcc.gnu.org/PR9322
18929 102. https://gcc.gnu.org/PR9433
18930 103. https://gcc.gnu.org/PR8032
18931 104. https://gcc.gnu.org/PR8639
18932 105. https://gcc.gnu.org/PR8794
18933 106. https://gcc.gnu.org/PR8832
18934 107. https://gcc.gnu.org/PR8988
18935 108. https://gcc.gnu.org/PR9492
18936 109. https://gcc.gnu.org/PR9267
18937 110. https://gcc.gnu.org/PR8344
18938 111. https://gcc.gnu.org/PR8524
18939 112. https://gcc.gnu.org/PR8880
18940 113. https://gcc.gnu.org/PR9090
18941 114. https://gcc.gnu.org/PR8588
18942 115. https://gcc.gnu.org/PR8599
18943 116. https://gcc.gnu.org/PR9506
18944 117. https://gcc.gnu.org/PR9484
18945 118. https://gcc.gnu.org/PR9292
18946 119. https://gcc.gnu.org/PR9293
18947 120. https://gcc.gnu.org/PR9295
18948 121. https://gcc.gnu.org/PR9296
18949 122. https://gcc.gnu.org/PR9316
18950 123. https://gcc.gnu.org/PR9493
18951 124. https://gcc.gnu.org/PR7341
18952 125. https://gcc.gnu.org/PR8947
18953 126. https://gcc.gnu.org/PR7448
18954 127. https://gcc.gnu.org/PR8882
18955 128. https://gcc.gnu.org/PR7445
18956 129. https://gcc.gnu.org/PR2521
18957 130. https://gcc.gnu.org/PR5661
18958 131. https://gcc.gnu.org/PR6419
18959 132. https://gcc.gnu.org/PR6994
18960 133. https://gcc.gnu.org/PR7150
18961 134. https://gcc.gnu.org/PR7160
18962 135. https://gcc.gnu.org/PR7228
18963 136. https://gcc.gnu.org/PR7266
18964 137. https://gcc.gnu.org/PR7353
18965 138. https://gcc.gnu.org/PR7411
18966 139. https://gcc.gnu.org/PR7478
18967 140. https://gcc.gnu.org/PR7526
18968 141. https://gcc.gnu.org/PR7721
18969 142. https://gcc.gnu.org/PR7803
18970 143. https://gcc.gnu.org/PR7754
18971 144. https://gcc.gnu.org/PR7788
18972 145. https://gcc.gnu.org/PR8031
18973 146. https://gcc.gnu.org/PR8055
18974 147. https://gcc.gnu.org/PR8067
18975 148. https://gcc.gnu.org/PR8134
18976 149. https://gcc.gnu.org/PR8149
18977 150. https://gcc.gnu.org/PR8160
18978 151. https://gcc.gnu.org/PR5607
18979 152. https://gcc.gnu.org/PR6579
18980 153. https://gcc.gnu.org/PR6803
18981 154. https://gcc.gnu.org/PR7176
18982 155. https://gcc.gnu.org/PR7188
18983 156. https://gcc.gnu.org/PR7306
18984 157. https://gcc.gnu.org/PR7461
18985 158. https://gcc.gnu.org/PR7524
18986 159. https://gcc.gnu.org/PR7584
18987 160. https://gcc.gnu.org/PR7676
18988 161. https://gcc.gnu.org/PR7679
18989 162. https://gcc.gnu.org/PR7811
18990 163. https://gcc.gnu.org/PR7961
18991 164. https://gcc.gnu.org/PR8071
18992 165. https://gcc.gnu.org/PR8127
18993 166. https://gcc.gnu.org/PR6745
18994 167. https://gcc.gnu.org/PR8096
18995 168. https://gcc.gnu.org/PR8127
18996 169. https://gcc.gnu.org/PR8218
18997 170. https://gcc.gnu.org/PR8287
18998 171. https://gcc.gnu.org/PR8347
18999 172. https://gcc.gnu.org/PR8348
19000 173. https://gcc.gnu.org/PR8391
19001 174. https://gcc.gnu.org/PR6627
19002 175. https://gcc.gnu.org/PR6631
19003 176. https://gcc.gnu.org/PR7102
19004 177. https://gcc.gnu.org/PR7120
19005 178. https://gcc.gnu.org/PR7209
19006 179. https://gcc.gnu.org/PR7515
19007 180. https://gcc.gnu.org/PR7814
19008 181. https://gcc.gnu.org/PR8467
19009 182. https://gcc.gnu.org/PR4890
19010 183. https://gcc.gnu.org/PR7357
19011 184. https://gcc.gnu.org/PR7358
19012 185. https://gcc.gnu.org/PR7602
19013 186. https://gcc.gnu.org/PR7862
19014 187. https://gcc.gnu.org/PR8190
19015 188. https://gcc.gnu.org/PR8524
19016 189. https://gcc.gnu.org/PR5351
19017 190. https://gcc.gnu.org/PR7591
19018 191. https://gcc.gnu.org/PR6845
19019 192. https://gcc.gnu.org/PR7034
19020 193. https://gcc.gnu.org/PR7124
19021 194. https://gcc.gnu.org/PR7174
19022 195. https://gcc.gnu.org/PR7134
19023 196. https://gcc.gnu.org/PR7375
19024 197. https://gcc.gnu.org/PR7390
19025 198. https://gcc.gnu.org/PR6890
19026 199. https://gcc.gnu.org/PR6981
19027 200. https://gcc.gnu.org/PR7242
19028 201. https://gcc.gnu.org/PR7396
19029 202. https://gcc.gnu.org/PR7630
19030 203. https://gcc.gnu.org/PR7693
19031 204. https://gcc.gnu.org/PR7723
19032 205. https://gcc.gnu.org/PR7951
19033 206. https://gcc.gnu.org/PR8146
19034 207. https://gcc.gnu.org/PR5967
19035 208. https://gcc.gnu.org/PR6984
19036 209. https://gcc.gnu.org/PR7114
19037 210. https://gcc.gnu.org/PR7130
19038 211. https://gcc.gnu.org/PR7133
19039 212. https://gcc.gnu.org/PR7380
19040 213. https://gcc.gnu.org/PR8252
19041 214. https://gcc.gnu.org/PR8451
19042 215. https://gcc.gnu.org/PR7250
19043 216. https://gcc.gnu.org/PR6668
19044 217. https://gcc.gnu.org/PR7151
19045 218. https://gcc.gnu.org/PR7335
19046 219. https://gcc.gnu.org/PR7842
19047 220. https://gcc.gnu.org/PR7856
19048 221. https://gcc.gnu.org/PR7967
19049 222. https://gcc.gnu.org/PR7374
19050 223. https://gcc.gnu.org/PR7370
19051 224. https://gcc.gnu.org/PR7409
19052 225. https://gcc.gnu.org/PR8232
19053 226. https://gcc.gnu.org/PR7623
19054 227. https://gcc.gnu.org/PR8314
19055 228. https://gcc.gnu.org/PR761
19056 229. https://gcc.gnu.org/PR5610
19057 230. https://gcc.gnu.org/PR7484
19058 231. https://gcc.gnu.org/PR7531
19059 232. https://gcc.gnu.org/PR8120
19060 233. https://gcc.gnu.org/PR7320
19061 234. https://gcc.gnu.org/PR7470
19062 235. https://gcc.gnu.org/PR6410
19063 236. https://gcc.gnu.org/PR6503
19064 237. https://gcc.gnu.org/PR6642
19065 238. https://gcc.gnu.org/PR7186
19066 239. https://gcc.gnu.org/PR7216
19067 240. https://gcc.gnu.org/PR7220
19068 241. https://gcc.gnu.org/PR7222
19069 242. https://gcc.gnu.org/PR7286
19070 243. https://gcc.gnu.org/PR7442
19071 244. https://gcc.gnu.org/PR7445
19072 245. https://gcc.gnu.org/PR7291
19073 246. https://gcc.gnu.org/onlinedocs/
19074 247. mailto:gcc-help@gcc.gnu.org
19075 248. mailto:gcc@gcc.gnu.org
19076 249. https://gcc.gnu.org/lists.html
19077 250. https://www.fsf.org/
19078 251. https://gcc.gnu.org/about.html
19079======================================================================
19080http://gcc.gnu.org/gcc-3.1/index.html
19081
19082                                    GCC 3.1
19083
19084   (This release series is no longer supported.)
19085
19086   July 27, 2002
19087
19088   The [1]GNU project and the GCC developers are pleased to announce the
19089   release of GCC 3.1.1.
19090
19091   The links below still apply to GCC 3.1.1.
19092
19093   May 15, 2002
19094
19095   The [2]GNU project and the GCC developers are pleased to announce the
19096   release of GCC 3.1.
19097
19098   GCC used to stand for the GNU C Compiler, but since the compiler
19099   supports several other languages aside from C, it now stands for the
19100   GNU Compiler Collection.
19101
19102   A list of [3]successful builds is updated as new information becomes
19103   available.
19104
19105   The GCC developers would like to thank the numerous people that have
19106   contributed [4]new features, improvements, bug fixes, and other changes
19107   as well as test results to GCC. This [5]amazing group of volunteers is
19108   what makes GCC successful.
19109
19110   For additional information about GCC please refer to the [6]GCC project
19111   web site or contact the [7]GCC development mailing list.
19112
19113   To obtain GCC please use [8]our mirror sites, or our CVS server.
19114     __________________________________________________________________
19115
19116
19117    For questions related to the use of GCC, please consult these web
19118    pages and the [9]GCC manuals. If that fails, the
19119    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19120    web pages and the development of GCC are welcome on our developer
19121    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
19122    archives.
19123
19124   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
19125   distribution of this entire article is permitted in any medium,
19126   provided this notice is preserved.
19127
19128   These pages are [14]maintained by the GCC team. Last modified
19129   2022-10-26.
19130
19131References
19132
19133   1. http://www.gnu.org/
19134   2. http://www.gnu.org/
19135   3. http://gcc.gnu.org/gcc-3.1/buildstat.html
19136   4. http://gcc.gnu.org/gcc-3.1/changes.html
19137   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
19138   6. http://gcc.gnu.org/index.html
19139   7. mailto:gcc@gcc.gnu.org
19140   8. http://gcc.gnu.org/mirrors.html
19141   9. https://gcc.gnu.org/onlinedocs/
19142  10. mailto:gcc-help@gcc.gnu.org
19143  11. mailto:gcc@gcc.gnu.org
19144  12. https://gcc.gnu.org/lists.html
19145  13. https://www.fsf.org/
19146  14. https://gcc.gnu.org/about.html
19147======================================================================
19148http://gcc.gnu.org/gcc-3.1/changes.html
19149
19150                             GCC 3.1 Release Series
19151                        Changes, New Features, and Fixes
19152
19153Additional changes in GCC 3.1.1
19154
19155     * A bug related to how structures and unions are returned has been
19156       fixed for powerpc-*-netbsd*.
19157     * An important bug in the implementation of -fprefetch-loop-arrays
19158       has been fixed. Previously the optimization prefetched random
19159       blocks of memory for most targets except for i386.
19160     * The Java compiler now compiles Java programs much faster and also
19161       works with parallel make.
19162     * Nested functions have been fixed for mips*-*-netbsd*.
19163     * Some missing floating point support routines have beed added for
19164       mips*-*-netbsd*.
19165     * This [1]message gives additional information about the bugs fixed
19166       in this release.
19167
19168Caveats
19169
19170     * The -traditional C compiler option has been deprecated and will be
19171       removed in GCC 3.3. (It remains possible to preprocess non-C code
19172       with the traditional preprocessor.)
19173     * The default debugging format for most ELF platforms (including
19174       GNU/Linux and FreeBSD; notable exception is Solaris) has changed
19175       from stabs to DWARF2. This requires GDB 5.1.1 or later.
19176
19177General Optimizer Improvements
19178
19179     * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat,
19180       and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure
19181       for profile driven optimizations.
19182       Options -fprofile-arcs and -fbranch-probabilities can now be used
19183       to improve speed of the generated code by profiling the actual
19184       program behaviour on typical runs. In the absence of profile info
19185       the compiler attempts to guess the profile statically.
19186     * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to
19187       monitor performance of the generated code.
19188       According to the SPECInt2000 results on an AMD Athlon CPU, the code
19189       generated by GCC 3.1 is 6% faster on the average (8.2% faster with
19190       profile feedback) compared to GCC 3.0. The code produced by GCC 3.0
19191       is about 2.1% faster compared to 2.95.3. Tests were done using the
19192       -O2 -march=athlon command-line options.
19193     * Alexandre Oliva, of Red Hat, has generalized the tree inlining
19194       infrastructure developed by CodeSourcery, LLC for the C++ front
19195       end, so that it is now used in the C front end too. Inlining
19196       functions as trees exposes them earlier to the compiler, giving it
19197       more opportunities for optimization.
19198     * Support for data prefetching instructions has been added to the GCC
19199       back end and several targets. A new __builtin_prefetch intrinsic is
19200       available to explicitly insert prefetch instructions and
19201       experimental support for loop array prefetching has been added (see
19202       -fprefetch-loop-array documentation).
19203     * Support for emitting debugging information for macros has been
19204       added for DWARF2. It is activated using -g3.
19205
19206New Languages and Language specific improvements
19207
19208  C/C++
19209
19210     * A few more [4]ISO C99 features.
19211     * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0.
19212     * The preprocessor's symbol table has been merged with the symbol
19213       table of the C, C++ and Objective-C front ends.
19214     * The preprocessor consumes less memory than the preprocessor in GCC
19215       3.0, often significantly so. On normal input files, it typically
19216       consumes less memory than pre-3.0 cccp-based GCC, too.
19217
19218  C++
19219
19220     * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std
19221       was a workaround to allow std compliant code to work with the
19222       non-std compliant libstdc++-v2. libstdc++-v3 is std compliant.
19223     * The C++ ABI has been fixed so that void (A::*)() const is mangled
19224       as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only
19225       affects pointer to cv-qualified member function types.
19226     * The C++ ABI has been changed to correctly handle this code:
19227    struct A {
19228      void operator delete[] (void *, size_t);
19229    };
19230
19231    struct B : public A {
19232    };
19233
19234    new B[10];
19235
19236       The amount of storage allocated for the array will be greater than
19237       it was in 3.0, in order to store the number of elements in the
19238       array, so that the correct size can be passed to operator delete[]
19239       when the array is deleted. Previously, the value passed to operator
19240       delete[] was unpredictable.
19241       This change will only affect code that declares a two-argument
19242       operator delete[] with a second parameter of type size_t in a base
19243       class, and does not override that definition in a derived class.
19244     * The C++ ABI has been changed so that:
19245    struct A {
19246      void operator delete[] (void *, size_t);
19247      void operator delete[] (void *);
19248    };
19249
19250       does not cause unnecessary storage to be allocated when an array of
19251       A objects is allocated.
19252       This change will only affect code that declares both of these forms
19253       of operator delete[], and declared the two-argument form before the
19254       one-argument form.
19255     * The C++ ABI has been changed so that when a parameter is passed by
19256       value, any cleanup for that parameter is performed in the caller,
19257       as specified by the ia64 C++ ABI, rather than the called function
19258       as before. As a result, classes with a non-trivial destructor but a
19259       trivial copy constructor will be passed and returned by invisible
19260       reference, rather than by bitwise copy as before.
19261     * G++ now supports the "named return value optimization": for code
19262       like
19263    A f () {
19264      A a;
19265      ...
19266      return a;
19267    }
19268
19269       G++ will allocate a in the return value slot, so that the return
19270       becomes a no-op. For this to work, all return statements in the
19271       function must return the same variable.
19272     * Improvements to the C++ library are listed in [5]the libstdc++-v3
19273       FAQ.
19274
19275  Objective-C
19276
19277     * Annoying linker warnings (due to incorrect code being generated)
19278       have been fixed.
19279     * If a class method cannot be found, the compiler no longer issues a
19280       warning if a corresponding instance method exists in the root
19281       class.
19282     * Forward @protocol declarations have been fixed.
19283     * Loading of categories has been fixed in certain situations (GNU run
19284       time only).
19285     * The class lookup in the run-time library has been rewritten so that
19286       class method dispatch is more than twice as fast as it used to be
19287       (GNU run time only).
19288
19289  Java
19290
19291     * libgcj now includes RMI, java.lang.ref.*, javax.naming, and
19292       javax.transaction.
19293     * Property files and other system resources can be compiled into
19294       executables which use libgcj using the new gcj --resource feature.
19295     * libgcj has been ported to more platforms. In particular there is
19296       now a mostly-functional mingw32 (Windows) target port.
19297     * JNI and CNI invocation interfaces were implemented, so gcj-compiled
19298       Java code can now be called from a C/C++ application.
19299     * gcj can now use builtin functions for certain known methods, for
19300       instance Math.cos.
19301     * gcj can now automatically remove redundant array-store checks in
19302       some common cases.
19303     * The --no-store-checks optimization option was added. This can be
19304       used to omit runtime store checks for code which is known not to
19305       throw ArrayStoreException
19306     * The following third party interface standards were added to libgcj:
19307       org.w3c.dom and org.xml.sax.
19308     * java.security has been merged with GNU Classpath. The new package
19309       is now JDK 1.2 compliant, and much more complete.
19310     * A bytecode verifier was added to the libgcj interpreter.
19311     * java.lang.Character was rewritten to comply with the Unicode 3.0
19312       standard, and improve performance.
19313     * Partial support for many more locales was added to libgcj.
19314     * Socket timeouts have been implemented.
19315     * libgcj has been merged into a single shared library. There are no
19316       longer separate shared libraries for the garbage collector and
19317       zlib.
19318     * Several performance improvements were made to gcj and libgcj:
19319          + Hash synchronization (thin locks)
19320          + A special allocation path for finalizer-free objects
19321          + Thread-local allocation
19322          + Parallel GC, and other GC tweaks
19323
19324  Fortran
19325
19326   Fortran improvements are listed in [6]the Fortran documentation.
19327
19328  Ada
19329
19330   [7]AdaCore, has contributed its GNAT Ada 95 front end and associated
19331   tools. The GNAT compiler fully implements the Ada language as defined
19332   by the ISO/IEC 8652 standard.
19333
19334   Please note that the integration of the Ada front end is still work in
19335   progress.
19336
19337New Targets and Target Specific Improvements
19338
19339     * Hans-Peter Nilsson has contributed a port to MMIX, the CPU
19340       architecture used in new editions of Donald E. Knuth's The Art of
19341       Computer Programming.
19342     * Axis Communications has contributed its port to the CRIS CPU
19343       architecture, used in the ETRAX system-on-a-chip series.
19344     * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH
19345       SH5 64-bit RISC microprocessor architecture, extending the existing
19346       SH port.
19347     * UltraSPARC is fully supported in 64-bit mode. The option -m64
19348       enables it.
19349     * For compatibility with the Sun compiler #pragma redefine_extname
19350       has been implemented on Solaris.
19351     * The x86 back end has had some noticeable work done to it.
19352          + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas
19353            Jaeger have contributed a port to the AMD x86-64 architecture.
19354            For more information on x86-64 see http://www.x86-64.org.
19355          + The compiler now supports MMX, 3DNow!, SSE, and SSE2
19356            instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will
19357            enable the respective instruction sets. Intel C++ compatible
19358            MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics
19359            will be added in next major release.
19360          + Following those improvements, targets for Pentium MMX, K6-2,
19361            K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were
19362            added. Refer to the documentation on -march= and -mcpu=
19363            options for details.
19364          + For those targets that support it, -mfpmath=sse will cause the
19365            compiler to generate SSE/SSE2 instructions for floating point
19366            math instead of x87 instructions. Usually, this will lead to
19367            quicker code -- especially on the Pentium 4. Note that only
19368            scalar floating point instructions are used and GCC does not
19369            exploit SIMD features yet.
19370          + Prefetch support has been added to the Pentium III, Pentium 4,
19371            K6-2, K6-3, and Athlon series.
19372          + Code generated for floating point to integer conversions has
19373            been improved leading to better performance of many 3D
19374            applications.
19375     * The PowerPC back end has added 64-bit PowerPC GNU/Linux support.
19376     * C++ support for AIX has been improved.
19377     * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the
19378       PowerPC port supporting the AltiVec programming model (SIMD). The
19379       support, though presently useful, is experimental and is expected
19380       to stabilize for 3.2. The support is written to conform to
19381       Motorola's AltiVec specs. See -maltivec.
19382
19383Obsolete Systems
19384
19385   Support for a number of older systems has been declared obsolete in GCC
19386   3.1. Unless there is activity to revive them, the next release of GCC
19387   will have their sources permanently removed.
19388
19389   All configurations of the following processor architectures have been
19390   declared obsolete:
19391     * MIL-STD-1750A, 1750a-*-*
19392     * AMD A29k, a29k-*-*
19393     * Convex, c*-convex-*
19394     * Clipper, clipper-*-*
19395     * Elxsi, elxsi-*-*
19396     * Intel i860, i860-*-*
19397     * Sun picoJava, pj-*-* and pjl-*-*
19398     * Western Electric 32000, we32k-*-*
19399
19400   Most configurations of the following processor architectures have been
19401   declared obsolete, but we are preserving a few systems which may have
19402   active developers. It is unlikely that the remaining systems will
19403   survive much longer unless we see definite signs of port activity.
19404     * Motorola 88000 except
19405          + Generic a.out, m88k-*-aout*
19406          + Generic SVR4, m88k-*-sysv4
19407          + OpenBSD, m88k-*-openbsd*
19408     * NS32k except
19409          + NetBSD, ns32k-*-netbsd*
19410          + OpenBSD, ns32k-*-openbsd*.
19411     * ROMP except
19412          + OpenBSD, romp-*-openbsd*.
19413
19414   Finally, only some configurations of these processor architectures are
19415   being obsoleted.
19416     * Alpha:
19417          + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka
19418            alpha*-*-osf[45], are still supported.)
19419     * ARM:
19420          + RISCiX, arm-*-riscix*.
19421     * i386:
19422          + 386BSD, i?86-*-bsd*
19423          + Chorus, i?86-*-chorusos*
19424          + DG/UX, i?86-*-dgux*
19425          + FreeBSD 1.x, i?86-*-freebsd1.*
19426          + IBM AIX, i?86-*-aix*
19427          + ISC UNIX, i?86-*-isc*
19428          + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld*
19429          + NEXTstep, i?86-next-*
19430          + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose*
19431          + RTEMS/coff, i?86-*-rtemscoff*
19432          + RTEMS/go32, i?86-go32-rtems*
19433          + Sequent/BSD, i?86-sequent-bsd*
19434          + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and
19435            i?86-sequent-sysv3*
19436          + SunOS, i?86-*-sunos*
19437     * Motorola 68000:
19438          + Altos, m68[k0]*-altos-*
19439          + Apollo, m68[k0]*-apollo-*
19440          + Apple A/UX, m68[k0]*-apple-*
19441          + Bull, m68[k0]*-bull-*
19442          + Convergent, m68[k0]*-convergent-*
19443          + Generic SVR3, m68[k0]*-*-sysv3*
19444          + ISI, m68[k0]*-isi-*
19445          + LynxOS, m68[k0]*-*-lynxos*
19446          + NEXT, m68[k0]*-next-*
19447          + RTEMS/coff, m68[k0]*-*-rtemscoff*
19448          + Sony, m68[k0]*-sony-*
19449     * MIPS:
19450          + DEC Ultrix, mips-*-ultrix* and mips-dec-*
19451          + Generic BSD, mips-*-bsd*
19452          + Generic System V, mips-*-sysv*
19453          + IRIX before version 5, mips-sgi-irix[1234]*
19454          + RiscOS, mips-*-riscos*
19455          + Sony, mips-sony-*
19456          + Tandem, mips-tandem-*
19457     * SPARC:
19458          + RTEMS/a.out, sparc-*-rtemsaout*.
19459
19460Documentation improvements
19461
19462     * The old manual ("Using and Porting the GNU Compiler Collection")
19463       has been replaced by a users manual ("Using the GNU Compiler
19464       Collection") and a separate internals reference manual ("GNU
19465       Compiler Collection Internals").
19466     * More complete and much improved documentation about GCC's internal
19467       representation used by the C and C++ front ends.
19468     * Many cleanups and improvements in general.
19469
19470
19471    For questions related to the use of GCC, please consult these web
19472    pages and the [8]GCC manuals. If that fails, the
19473    [9]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19474    web pages and the development of GCC are welcome on our developer
19475    list at [10]gcc@gcc.gnu.org. All of [11]our lists have public
19476    archives.
19477
19478   Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and
19479   distribution of this entire article is permitted in any medium,
19480   provided this notice is preserved.
19481
19482   These pages are [13]maintained by the GCC team. Last modified
19483   2022-10-26.
19484
19485References
19486
19487   1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html
19488   2. http://gcc.gnu.org/news/profiledriven.html
19489   3. http://gcc.gnu.org/benchmarks/
19490   4. http://gcc.gnu.org/c99status.html
19491   5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html
19492   6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html
19493   7. https://www.adacore.com/
19494   8. https://gcc.gnu.org/onlinedocs/
19495   9. mailto:gcc-help@gcc.gnu.org
19496  10. mailto:gcc@gcc.gnu.org
19497  11. https://gcc.gnu.org/lists.html
19498  12. https://www.fsf.org/
19499  13. https://gcc.gnu.org/about.html
19500======================================================================
19501http://gcc.gnu.org/gcc-3.0/index.html
19502
19503                                   GCC 3.0.4
19504
19505   (This release series is no longer supported.)
19506
19507   February 20, 2002
19508
19509   The [1]GNU project and the GCC developers are pleased to announce the
19510   release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0
19511   series.
19512
19513   GCC used to stand for the GNU C Compiler, but since the compiler
19514   supports several other languages aside from C, it now stands for the
19515   GNU Compiler Collection.
19516
19517   GCC 3.0.x has several new optimizations, new targets, new languages and
19518   many other new features, relative to GCC 2.95.x. See the [2]new
19519   features page for a more complete list.
19520
19521   A list of [3]successful builds is updated as new information becomes
19522   available.
19523
19524   The GCC developers would like to thank the numerous people that have
19525   contributed new features, test results, bug fixes, etc to GCC. This
19526   [4]amazing group of volunteers is what makes GCC successful.
19527
19528   And finally, we can't in good conscience fail to mention some
19529   [5]caveats to using GCC 3.0.x.
19530
19531   For additional information about GCC please refer to the [6]GCC project
19532   web site or contact the [7]GCC development mailing list.
19533
19534   To obtain GCC please use [8]our mirror sites, or our CVS server.
19535     __________________________________________________________________
19536
19537Previous 3.0.x Releases
19538
19539   December 20, 2001: GCC 3.0.3 has been released.
19540   October 25, 2001: GCC 3.0.2 has been released.
19541   August 20, 2001: GCC 3.0.1 has been released.
19542   June 18, 2001: GCC 3.0 has been released.
19543
19544
19545    For questions related to the use of GCC, please consult these web
19546    pages and the [9]GCC manuals. If that fails, the
19547    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19548    web pages and the development of GCC are welcome on our developer
19549    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
19550    archives.
19551
19552   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
19553   distribution of this entire article is permitted in any medium,
19554   provided this notice is preserved.
19555
19556   These pages are [14]maintained by the GCC team. Last modified
19557   2022-10-26.
19558
19559References
19560
19561   1. http://www.gnu.org/
19562   2. http://gcc.gnu.org/gcc-3.0/features.html
19563   3. http://gcc.gnu.org/gcc-3.0/buildstat.html
19564   4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
19565   5. http://gcc.gnu.org/gcc-3.0/caveats.html
19566   6. http://gcc.gnu.org/index.html
19567   7. mailto:gcc@gcc.gnu.org
19568   8. http://gcc.gnu.org/mirrors.html
19569   9. https://gcc.gnu.org/onlinedocs/
19570  10. mailto:gcc-help@gcc.gnu.org
19571  11. mailto:gcc@gcc.gnu.org
19572  12. https://gcc.gnu.org/lists.html
19573  13. https://www.fsf.org/
19574  14. https://gcc.gnu.org/about.html
19575======================================================================
19576http://gcc.gnu.org/gcc-3.0/features.html
19577
19578                              GCC 3.0 New Features
19579
19580Additional changes in GCC 3.0.4
19581
19582     * GCC 3.0 now supports newer versions of the [1]NetBSD operating
19583       system, which use the ELF object file format, on x86 processors.
19584     * Correct debugging information is generated from functions that have
19585       lines from multiple files (e.g. yacc output).
19586     * A fix for whitespace handling in the -traditional preprocessor,
19587       which can affect Fortran.
19588     * Fixes to the exception handling runtime.
19589     * More fixes for bad code generation in C++.
19590     * A fix for shared library generation under AIX 4.3.
19591     * Documentation updates.
19592     * Port of GCC to Tensilica's Xtensa processor contributed.
19593     * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link).
19594
19595Additional changes in GCC 3.0.3
19596
19597     * A fix to correct an accidental change to the PowerPC ABI.
19598     * Fixes for bad code generation on a variety of architectures.
19599     * Improvements to the debugging information generated for C++
19600       classes.
19601     * Fixes for bad code generation in C++.
19602     * A fix to avoid crashes in the C++ demangler.
19603     * A fix to the C++ standard library to avoid buffer overflows.
19604     * Miscellaneous improvements for a variety of architectures.
19605
19606Additional changes in GCC 3.0.2
19607
19608     * Fixes for bad code generation during loop unrolling.
19609     * Fixes for bad code generation by the sibling call optimization.
19610     * Minor improvements to x86 code generation.
19611     * Implementation of function descriptors in C++ vtables for IA64.
19612     * Numerous minor bug-fixes.
19613
19614Additional changes in GCC 3.0.1
19615
19616     * C++ fixes for incorrect code-generation.
19617     * Improved cross-compiling support for the C++ standard library.
19618     * Fixes for some embedded targets that worked in GCC 2.95.3, but not
19619       in GCC 3.0.
19620     * Fixes for various exception-handling bugs.
19621     * A port to the S/390 architecture.
19622
19623General Optimizer Improvements
19624
19625     * [2]Basic block reordering pass.
19626     * New if-conversion pass with support for conditional (predicated)
19627       execution.
19628     * New tail call and sibling call elimination optimizations.
19629     * New register renaming pass.
19630     * New (experimental) [3]static single assignment (SSA) representation
19631       support.
19632     * New dead-code elimination pass implemented using the SSA
19633       representation.
19634     * [4]Global null pointer test elimination.
19635     * [5]Global code hoisting/unification.
19636     * More builtins and optimizations for stdio.h, string.h and old BSD
19637       functions, as well as for ISO C99 functions.
19638     * New builtin __builtin_expect for giving hints to the branch
19639       predictor.
19640
19641New Languages and Language specific improvements
19642
19643     * The GNU Compiler for the Java(TM) language (GCJ) is now integrated
19644       and supported, including the run-time library containing most
19645       common non-GUI Java classes, a bytecode interpreter, and the Boehm
19646       conservative garbage collector. Many bugs have been fixed. GCJ can
19647       compile Java source or Java bytecodes to either native code or Java
19648       class files, and supports native methods written in either the
19649       standard JNI or the more efficient and convenient CNI.
19650     * Here is a [6]partial list of C++ improvements, both new features
19651       and those no longer supported.
19652     * New C++ ABI. On the IA-64 platform GCC is capable of
19653       inter-operating with other IA-64 compilers.
19654     * The new ABI also significantly reduces the size of symbol and debug
19655       information.
19656     * New C++ support library and many C++ bug fixes, vastly improving
19657       our conformance to the ISO C++ standard.
19658     * New [7]inliner for C++.
19659     * Rewritten C preprocessor, integrated into the C, C++ and Objective
19660       C compilers, with very many improvements including ISO C99 support
19661       and [8]improvements to dependency generation.
19662     * Support for more [9]ISO C99 features.
19663     * Many improvements to support for checking calls to format functions
19664       such as printf and scanf, including support for ISO C99 format
19665       features, extensions from the Single Unix Specification and GNU
19666       libc 2.2, checking of strfmon formats and features to assist in
19667       auditing for format string security bugs.
19668     * New warnings for C code that may have undefined semantics because
19669       of violations of sequence point rules in the C standard (such as a
19670       = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall.
19671     * Additional warning option -Wfloat-equal.
19672     * Improvements to -Wtraditional.
19673     * Fortran improvements are listed in [10]the Fortran documentation.
19674
19675New Targets and Target Specific Improvements
19676
19677     * New x86 back end, generating much improved code.
19678     * Support for a generic i386-elf target contributed.
19679     * New option to emit x86 assembly code using Intel style syntax
19680       (-mintel-syntax).
19681     * HPUX 11 support contributed.
19682     * Improved PowerPC code generation, including scheduled prologue and
19683       epilogue.
19684     * Port of GCC to Intel's IA-64 processor contributed.
19685     * Port of GCC to Motorola's MCore 210 and 340 contributed.
19686     * New unified back-end for Arm, Thumb and StrongArm contributed.
19687     * Port of GCC to Intel's XScale processor contributed.
19688     * Port of GCC to Atmel's AVR microcontrollers contributed.
19689     * Port of GCC to Mitsubishi's D30V processor contributed.
19690     * Port of GCC to Matsushita's AM33 processor (a member of the MN10300
19691       processor family) contributed.
19692     * Port of GCC to Fujitsu's FR30 processor contributed.
19693     * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed.
19694     * Port of GCC to Sun's picoJava processor core contributed.
19695
19696Documentation improvements
19697
19698     * Substantially rewritten and improved C preprocessor manual.
19699     * Many improvements to other documentation.
19700     * Manpages for gcc, cpp and gcov are now generated automatically from
19701       the master Texinfo manual, eliminating the problem of manpages
19702       being out of date. (The generated manpages are only extracts from
19703       the full manual, which is provided in Texinfo form, from which
19704       info, HTML, other formats and a printed manual can be generated.)
19705     * Generated info files are included in the release tarballs alongside
19706       their Texinfo sources, avoiding problems on some platforms with
19707       building makeinfo as part of the GCC distribution.
19708
19709Other significant improvements
19710
19711     * Garbage collection used internally by the compiler for most memory
19712       allocation instead of obstacks.
19713     * Lengauer and Tarjan algorithm used for computing dominators in the
19714       CFG. This algorithm can be significantly faster and more space
19715       efficient than our older algorithm.
19716     * gccbug script provided to assist in submitting bug reports to our
19717       bug tracking system. (Bug reports previously submitted directly to
19718       our mailing lists, for which you received no bug tracking number,
19719       should be submitted again using gccbug if you can reproduce the
19720       problem with GCC 3.0.)
19721     * The internal libgcc library is [11]built as a shared library on
19722       systems that support it.
19723     * Extensive testsuite included with GCC, with many new tests. In
19724       addition to tests for GCC bugs that have been fixed, many tests
19725       have been added for language features, compiler warnings and
19726       builtin functions.
19727     * Additional language-independent warning options -Wpacked, -Wpadded,
19728       -Wunreachable-code and -Wdisabled-optimization.
19729     * Target-independent options -falign-functions, -falign-loops and
19730       -falign-jumps.
19731
19732   Plus a great many bug fixes and almost all the [12]features found in
19733   GCC 2.95.
19734
19735
19736    For questions related to the use of GCC, please consult these web
19737    pages and the [13]GCC manuals. If that fails, the
19738    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19739    web pages and the development of GCC are welcome on our developer
19740    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
19741    archives.
19742
19743   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
19744   distribution of this entire article is permitted in any medium,
19745   provided this notice is preserved.
19746
19747   These pages are [18]maintained by the GCC team. Last modified
19748   2022-10-26.
19749
19750References
19751
19752   1. http://www.netbsd.org/
19753   2. http://gcc.gnu.org/news/reorder.html
19754   3. http://gcc.gnu.org/news/ssa.html
19755   4. http://gcc.gnu.org/news/null.html
19756   5. http://gcc.gnu.org/news/unify.html
19757   6. http://gcc.gnu.org/gcc-3.0/c++features.html
19758   7. http://gcc.gnu.org/news/inlining.html
19759   8. http://gcc.gnu.org/news/dependencies.html
19760   9. http://gcc.gnu.org/c99status.html
19761  10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
19762  11. http://gcc.gnu.org/gcc-3.0/libgcc.html
19763  12. http://gcc.gnu.org/gcc-2.95/features.html
19764  13. https://gcc.gnu.org/onlinedocs/
19765  14. mailto:gcc-help@gcc.gnu.org
19766  15. mailto:gcc@gcc.gnu.org
19767  16. https://gcc.gnu.org/lists.html
19768  17. https://www.fsf.org/
19769  18. https://gcc.gnu.org/about.html
19770======================================================================
19771http://gcc.gnu.org/gcc-3.0/caveats.html
19772
19773                                GCC 3.0 Caveats
19774
19775     * -fstrict-aliasing is now part of -O2 and higher optimization
19776       levels. This allows the compiler to assume the strictest aliasing
19777       rules applicable to the language being compiled. For C and C++,
19778       this activates optimizations based on the type of expressions. This
19779       optimization may thus break old, non-compliant code.
19780     * Enumerations are now properly promoted to int in function
19781       parameters and function returns. Normally this change is not
19782       visible, but when using -fshort-enums this is an ABI change.
19783     * The undocumented extension that allowed C programs to have a label
19784       at the end of a compound statement has been deprecated and may be
19785       removed in a future version. Programs that now generate a warning
19786       about this may be fixed by adding a null statement (a single
19787       semicolon) after the label.
19788     * The poorly documented extension that allowed string constants in C,
19789       C++ and Objective C to contain unescaped newlines has been
19790       deprecated and may be removed in a future version. Programs using
19791       this extension may be fixed in several ways: the bare newline may
19792       be replaced by \n, or preceded by \n\, or string concatenation may
19793       be used with the bare newline preceded by \n" and " placed at the
19794       start of the next line.
19795     * The Chill compiler is not included in GCC 3.0, because of the lack
19796       of a volunteer to convert it to use garbage collection.
19797     * Certain non-standard iostream methods from earlier versions of
19798       libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach,
19799       ostream::form, and istream::gets.
19800     * The new C++ ABI is not yet fully supported by current (as of
19801       2001-07-01) releases and development versions of GDB, or any
19802       earlier versions. There is a problem setting breakpoints by line
19803       number, and other related issues that have been fixed in GCC 3.0
19804       but not yet handled in GDB:
19805       [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
19806
19807
19808    For questions related to the use of GCC, please consult these web
19809    pages and the [2]GCC manuals. If that fails, the
19810    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19811    web pages and the development of GCC are welcome on our developer
19812    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
19813
19814   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
19815   distribution of this entire article is permitted in any medium,
19816   provided this notice is preserved.
19817
19818   These pages are [7]maintained by the GCC team. Last modified
19819   2022-10-26.
19820
19821References
19822
19823   1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
19824   2. https://gcc.gnu.org/onlinedocs/
19825   3. mailto:gcc-help@gcc.gnu.org
19826   4. mailto:gcc@gcc.gnu.org
19827   5. https://gcc.gnu.org/lists.html
19828   6. https://www.fsf.org/
19829   7. https://gcc.gnu.org/about.html
19830======================================================================
19831http://gcc.gnu.org/gcc-2.95/index.html
19832
19833                                    GCC 2.95
19834
19835   (This release series is no longer supported.)
19836
19837   March 16, 2001: The GNU project and the GCC developers are pleased to
19838   announce the release of GCC version 2.95.3.
19839
19840Release History
19841
19842   GCC 2.95.3
19843          March 16, 2001
19844
19845   GCC 2.95.2
19846          October 27, 1999
19847
19848   GCC 2.95.1
19849          August 19, 1999
19850
19851   GCC 2.95
19852          July 31, 1999. This is the first release of GCC since the April
19853          1999 GCC/EGCS reunification and includes nearly a year's worth
19854          of new development and bugfixes.
19855
19856References and Acknowledgements
19857
19858   GCC used to stand for the GNU C Compiler, but since the compiler
19859   supports several other languages aside from C, it now stands for the
19860   GNU Compiler Collection.
19861
19862   The whole suite has been extensively [1]regression tested and
19863   [2]package tested. It should be reliable and suitable for widespread
19864   use.
19865
19866   The compiler has several new optimizations, new targets, new languages
19867   and other new features. See the [3]new features page for a more
19868   complete list of new features found in the GCC 2.95 releases.
19869
19870   The sources include installation instructions in both HTML and
19871   plaintext forms in the install directory in the distribution. However,
19872   the most up to date installation instructions and [4]build/test status
19873   are on the web pages. We will update those pages as new information
19874   becomes available.
19875
19876   The GCC developers would like to thank the numerous people that have
19877   contributed new features, test results, bugfixes, etc to GCC. This
19878   [5]amazing group of volunteers is what makes GCC successful.
19879
19880   And finally, we can't in good conscience fail to mention some
19881   [6]caveats to using GCC 2.95.
19882
19883   Download GCC 2.95 from one of our many [7]mirror sites.
19884
19885   For additional information about GCC please see the [8]GCC project web
19886   server or contact the [9]GCC development mailing list.
19887
19888
19889    For questions related to the use of GCC, please consult these web
19890    pages and the [10]GCC manuals. If that fails, the
19891    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19892    web pages and the development of GCC are welcome on our developer
19893    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
19894    archives.
19895
19896   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
19897   distribution of this entire article is permitted in any medium,
19898   provided this notice is preserved.
19899
19900   These pages are [15]maintained by the GCC team. Last modified
19901   2022-10-26.
19902
19903References
19904
19905   1. http://gcc.gnu.org/gcc-2.95/regress.html
19906   2. http://gcc.gnu.org/gcc-2.95/othertest.html
19907   3. http://gcc.gnu.org/gcc-2.95/features.html
19908   4. http://gcc.gnu.org/gcc-2.95/buildstat.html
19909   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
19910   6. http://gcc.gnu.org/gcc-2.95/caveats.html
19911   7. http://gcc.gnu.org/mirrors.html
19912   8. http://gcc.gnu.org/index.html
19913   9. mailto:gcc@gcc.gnu.org
19914  10. https://gcc.gnu.org/onlinedocs/
19915  11. mailto:gcc-help@gcc.gnu.org
19916  12. mailto:gcc@gcc.gnu.org
19917  13. https://gcc.gnu.org/lists.html
19918  14. https://www.fsf.org/
19919  15. https://gcc.gnu.org/about.html
19920======================================================================
19921http://gcc.gnu.org/gcc-2.95/features.html
19922
19923                             GCC 2.95 New Features
19924
19925     * General Optimizer Improvements:
19926          + [1]Localized register spilling to improve speed and code
19927            density especially on small register class machines.
19928          + [2]Global CSE using lazy code motion algorithms.
19929          + [3]Improved global constant/copy propagation.
19930          + [4]Improved control flow graph analysis and manipulation.
19931          + [5]Local dead store elimination.
19932          + [6]Memory Load hoisting/store sinking in loops.
19933          + [7]Type based alias analysis is enabled by default. Note this
19934            feature will expose bugs in the Linux kernel. Please refer to
19935            the FAQ (as shipped with GCC 2.95) for additional information
19936            on this issue.
19937          + Major revamp of GIV detection, combination and simplification
19938            to improve loop performance.
19939          + Major improvements to register allocation and reloading.
19940     * New Languages and Language specific improvements
19941          + [8]Many C++ improvements.
19942          + [9]Many Fortran improvements.
19943          + [10]Java front-end has been integrated. A [11]runtime library
19944            is available separately.
19945          + [12]ISO C99 support
19946          + [13]Chill front-end and runtime has been integrated.
19947          + Boehm garbage collector support in libobjc.
19948          + More support for various pragmas which appear in vendor
19949            include files
19950     * New Targets and Target Specific Improvements
19951          + [14]SPARC backend rewrite.
19952          + -mschedule=8000 will optimize code for PA8000 class
19953            processors; -mpa-risc-2-0 will generate code for PA2.0
19954            processors
19955          + Various micro-optimizations for the ia32 port. K6
19956            optimizations
19957          + Compiler will attempt to align doubles in the stack on the
19958            ia32 port
19959          + Alpha EV6 support
19960          + PowerPC 750
19961          + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403.
19962            -mcpu=e603e was added to do -mcpu=603e and -msoft-float.
19963          + c3x, c4x
19964          + HyperSPARC
19965          + SparcLite86x
19966          + sh4
19967          + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix,
19968            arm-linux)
19969          + vxWorks targets include support for vxWorks threads
19970          + StrongARM 110 and ARM9 support added. ARM Scheduling
19971            parameters rewritten.
19972          + Various changes to the MIPS port to avoid assembler macros,
19973            which in turn improves performance
19974          + Various performance improvements to the i960 port.
19975          + Major rewrite of ns32k port
19976     * Other significant improvements
19977          + [15]Ability to dump cfg information and display it using vcg.
19978          + The new faster scheme for fixing vendor header files is
19979            enabled by default.
19980          + Experimental internationalization support.
19981          + multibyte character support
19982          + Some compile-time speedups for pathological problems
19983          + Better support for complex types
19984     * Plus the usual mountain of bugfixes
19985     * Core compiler is based on the gcc2 development tree from Sept 30,
19986       1998, so we have all of the [16]features found in GCC 2.8.
19987
19988Additional Changes in GCC 2.95.1
19989
19990     * Generic bugfixes and improvements
19991          + Various documentation fixes related to the GCC/EGCS merger.
19992          + Fix memory management bug which could lead to spurious aborts,
19993            core dumps or random parsing errors in the compiler.
19994          + Fix a couple bugs in the dwarf1 and dwarf2 debug record
19995            support.
19996          + Fix infinite loop in the CSE optimizer.
19997          + Avoid undefined behavior in compiler FP emulation code
19998          + Fix install problem when prefix is overridden on the make
19999            install command.
20000          + Fix problem with unwanted installation of assert.h on some
20001            systems.
20002          + Fix problem with finding the wrong assembler in a single tree
20003            build.
20004          + Avoid increasing the known alignment of a register that is
20005            already known to be a pointer.
20006     * Platform specific bugfixes and improvements
20007          + Codegen bugfix for prologue/epilogue for cpu32 target.
20008          + Fix long long code generation bug for the Coldfire target.
20009          + Fix various aborts in the SH compiler.
20010          + Fix bugs in libgcc support library for the SH.
20011          + Fix alpha ev6 code generation bug.
20012          + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on
20013            AIX platforms.
20014          + Fix -fpic code generation bug for rs6000/ppc svr4 targets.
20015          + Fix varargs/stdarg code generation bug for rs6000/ppc svr4
20016            targets.
20017          + Fix weak symbol handling for rs6000/ppc svr4 targets.
20018          + Fix various problems with 64bit code generation for the
20019            rs6000/ppc port.
20020          + Fix codegen bug which caused tetex to be mis-compiled on the
20021            x86.
20022          + Fix compiler abort in new cfg code exposed by x86 port.
20023          + Fix out of range array reference in code convert flat
20024            registers to the x87 stacked FP register file.
20025          + Fix minor vxworks configuration bug.
20026          + Fix return type of bsearch for SunOS 4.x.
20027     * Language & Runtime specific fixes.
20028          + The G++ signature extension has been deprecated. It will be
20029            removed in the next major release of G++. Use of signatures
20030            will result in a warning from the compiler.
20031          + Several bugs relating to templates and namespaces were fixed.
20032          + A bug that caused crashes when combining templates with -g on
20033            DWARF1 platforms was fixed.
20034          + Pointers-to-members, virtual functions, and multiple
20035            inheritance should now work together correctly.
20036          + Some code-generation bugs relating to function try blocks were
20037            fixed.
20038          + G++ is a little bit more lenient with certain archaic
20039            constructs than in GCC 2.95.
20040          + Fix to prevent shared library version #s from bring truncated
20041            to 1 digit
20042          + Fix missing std:: in the libstdc++ library.
20043          + Fix stream locking problems in libio.
20044          + Fix problem in java compiler driver.
20045
20046Additional Changes in GCC 2.95.2
20047
20048   The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While
20049   the optimizations performed by -fstrict-aliasing are valid according to
20050   the C and C++ standards, the optimization have caused some problems,
20051   particularly with old non-conforming code.
20052
20053   The GCC developers are experimenting with ways to warn users about code
20054   which violates the C/C++ standards, but those warnings are not ready
20055   for widespread use at this time. Rather than wait for those warnings
20056   the GCC developers have chosen to disable -fstrict-aliasing by default
20057   for the GCC 2.95.2 release.
20058
20059   We strongly encourage developers to find and fix code which violates
20060   the C/C++ standards as -fstrict-aliasing may be enabled by default in
20061   future releases. Use the option -fstrict-aliasing to re-enable these
20062   optimizations.
20063     * Generic bugfixes and improvements
20064          + Fix incorrectly optimized memory reference in global common
20065            subexpression elimination (GCSE) optimization pass.
20066          + Fix code generation bug in regmove.c in which it could
20067            incorrectly change a "const" value.
20068          + Fix bug in optimization of conditionals involving volatile
20069            memory references.
20070          + Avoid over-allocation of stack space for some procedures.
20071          + Fixed bug in the compiler which caused incorrect optimization
20072            of an obscure series of bit manipulations, shifts and
20073            arithmetic.
20074          + Fixed register allocator bug which caused teTeX to be
20075            mis-compiled on SPARC targets.
20076          + Avoid incorrect optimization of degenerate case statements for
20077            certain targets such as the ARM.
20078          + Fix out of range memory reference in the jump optimizer.
20079          + Avoid dereferencing null pointer in fix-header.
20080          + Fix test for GCC specific features so that it is possible to
20081            bootstrap with gcc-2.6.2 and older versions of GCC.
20082          + Fix typo in scheduler which could potentially cause out of
20083            range memory accesses.
20084          + Avoid incorrect loop reversal which caused incorrect code for
20085            certain loops on PowerPC targets.
20086          + Avoid incorrect optimization of switch statements on certain
20087            targets (for example the ARM).
20088     * Platform specific bugfixes and improvements
20089          + Work around bug in Sun V5.0 compilers which caused bootstrap
20090            comparison failures on SPARC targets.
20091          + Fix SPARC backend bug which caused aborts in final.c.
20092          + Fix sparc-hal-solaris2* configuration fragments.
20093          + Fix bug in sparc block profiling.
20094          + Fix obscure code generation bug for the PARISC targets.
20095          + Define __STDC_EXT__ for HPUX configurations.
20096          + Various POWERPC64 code generation bugfixes.
20097          + Fix abort for PPC targets using ELF (ex GNU/Linux).
20098          + Fix collect2 problems for AIX targets.
20099          + Correct handling of .file directive for PPC targets.
20100          + Fix bug in fix_trunc x86 patterns.
20101          + Fix x86 port to correctly pop the FP stack for functions that
20102            return structures in memory.
20103          + Fix minor bug in strlen x86 pattern.
20104          + Use stabs debugging instead of dwarf1 for x86-solaris targets.
20105          + Fix template repository code to handle leading underscore in
20106            mangled names.
20107          + Fix weak/weak alias support for OpenBSD.
20108          + GNU/Linux for the ARM has C++ compatible include files.
20109     * Language & Runtime specific fixes.
20110          + Fix handling of constructor attribute in the C front-end which
20111            caused problems building the Chill runtime library on some
20112            targets.
20113          + Fix minor problem merging type qualifiers in the C front-end.
20114          + Fix aliasing bug for pointers and references (C/C++).
20115          + Fix incorrect "non-constant initializer bug" when -traditional
20116            or -fwritable-strings is enabled.
20117          + Fix build error for Chill front-end on SunOS.
20118          + Do not complain about duplicate instantiations when using
20119            -frepo (C++).
20120          + Fix array bounds handling in C++ front-end which caused
20121            problems with dwarf debugging information in some
20122            circumstances.
20123          + Fix minor namespace problem.
20124          + Fix problem linking java programs.
20125
20126Additional Changes in GCC 2.95.3
20127
20128     * Generic bugfixes and improvements
20129          + Fix numerous problems that caused incorrect optimization in
20130            the register reloading code.
20131          + Fix numerous problems that caused incorrect optimization in
20132            the loop optimizer.
20133          + Fix aborts in the functions build_insn_chain and scan_loops
20134            under some circumstances.
20135          + Fix an alias analysis bug.
20136          + Fix an infinite compilation bug in the combiner.
20137          + A few problems with complex number support have been fixed.
20138          + It is no longer possible for gcc to act as a fork bomb when
20139            installed incorrectly.
20140          + The -fpack-struct option should be recognized now.
20141          + Fixed a bug that caused incorrect code to be generated due to
20142            a lost stack adjustment.
20143     * Platform specific bugfixes and improvements
20144          + Support building ARM toolchains hosted on Windows.
20145          + Fix attribute calculations in ARM toolchains.
20146          + arm-linux support has been improved.
20147          + Fix a PIC failure on sparc targets.
20148          + On ix86 targets, the regparm attribute should now work
20149            reliably.
20150          + Several updates for the h8300 port.
20151          + Fix problem building libio with glibc 2.2.
20152
20153
20154    For questions related to the use of GCC, please consult these web
20155    pages and the [17]GCC manuals. If that fails, the
20156    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20157    web pages and the development of GCC are welcome on our developer
20158    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
20159    archives.
20160
20161   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
20162   distribution of this entire article is permitted in any medium,
20163   provided this notice is preserved.
20164
20165   These pages are [22]maintained by the GCC team. Last modified
20166   2022-10-26.
20167
20168References
20169
20170   1. http://gcc.gnu.org/news/spill.html
20171   2. http://gcc.gnu.org/news/lcm.html
20172   3. http://gcc.gnu.org/news/cprop.html
20173   4. http://gcc.gnu.org/news/cfg.html
20174   5. http://gcc.gnu.org/news/dse.html
20175   6. http://gcc.gnu.org/news/hoist.html
20176   7. http://gcc.gnu.org/news/alias.html
20177   8. http://gcc.gnu.org/gcc-2.95/c++features.html
20178   9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
20179  10. http://gcc.gnu.org/news/gcj-announce.txt
20180  11. http://gcc.gnu.org/news/javaannounce.html
20181  12. http://gcc.gnu.org/c99status.html
20182  13. http://gcc.gnu.org/news/chill.html
20183  14. http://gcc.gnu.org/news/sparc.html
20184  15. http://gcc.gnu.org/news/egcs-vcg.html
20185  16. http://gcc.gnu.org/egcs-1.0/features-2.8.html
20186  17. https://gcc.gnu.org/onlinedocs/
20187  18. mailto:gcc-help@gcc.gnu.org
20188  19. mailto:gcc@gcc.gnu.org
20189  20. https://gcc.gnu.org/lists.html
20190  21. https://www.fsf.org/
20191  22. https://gcc.gnu.org/about.html
20192======================================================================
20193http://gcc.gnu.org/gcc-2.95/caveats.html
20194
20195                                GCC 2.95 Caveats
20196
20197     * GCC 2.95 will issue an error for invalid asm statements that had
20198       been silently accepted by earlier versions of the compiler. This is
20199       particularly noticeable when compiling older versions of the Linux
20200       kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95)
20201       for more information on this issue.
20202     * GCC 2.95 implements type based alias analysis to disambiguate
20203       memory references. Some programs, particularly the Linux kernel
20204       violate ANSI/ISO aliasing rules and therefore may not operate
20205       correctly when compiled with GCC 2.95. Please refer to the FAQ (as
20206       shipped with GCC 2.95) for more information on this issue.
20207     * GCC 2.95 has a known bug in its handling of complex variables for
20208       64bit targets. Instead of silently generating incorrect code, GCC
20209       2.95 will issue a fatal error for situations it can not handle.
20210       This primarily affects the Fortran community as Fortran makes more
20211       use of complex variables than C or C++.
20212     * GCC 2.95 has an integrated libstdc++, but does not have an
20213       integrated libg++. Furthermore old libg++ releases will not work
20214       with GCC 2.95. You can retrieve a recent copy of libg++ from the
20215       [1]GCC ftp server.
20216       Note most C++ programs only need libstdc++.
20217     * Exception handling may not work with shared libraries, particularly
20218       on alphas, hppas, rs6000/powerpc and mips based platforms.
20219       Exception handling is known to work on x86 GNU/Linux platforms with
20220       shared libraries.
20221     * In general, GCC 2.95 is more rigorous about rejecting invalid C++
20222       code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0,
20223       or EGCS 1.1. As a result it may be necessary to fix C++ code before
20224       it will compile with GCC 2.95.
20225     * G++ is also converting toward the ISO C++ standard; as a result
20226       code which was previously valid (and thus accepted by other
20227       compilers and older versions of g++) may no longer be accepted. The
20228       flag -fpermissive may allow some non-conforming code to compile
20229       with GCC 2.95.
20230     * GCC 2.95 compiled C++ code is not binary compatible with EGCS
20231       1.1.x, EGCS 1.0.x or GCC 2.8.x.
20232     * GCC 2.95 does not have changes from the GCC 2.8 tree that were made
20233       between Sept 30, 1998 and April 30, 1999 (the official end of the
20234       GCC 2.8 project). Future GCC releases will include all the changes
20235       from the defunct GCC 2.8 sources.
20236
20237
20238    For questions related to the use of GCC, please consult these web
20239    pages and the [2]GCC manuals. If that fails, the
20240    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20241    web pages and the development of GCC are welcome on our developer
20242    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
20243
20244   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
20245   distribution of this entire article is permitted in any medium,
20246   provided this notice is preserved.
20247
20248   These pages are [7]maintained by the GCC team. Last modified
20249   2022-10-26.
20250
20251References
20252
20253   1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz
20254   2. https://gcc.gnu.org/onlinedocs/
20255   3. mailto:gcc-help@gcc.gnu.org
20256   4. mailto:gcc@gcc.gnu.org
20257   5. https://gcc.gnu.org/lists.html
20258   6. https://www.fsf.org/
20259   7. https://gcc.gnu.org/about.html
20260======================================================================
20261http://gcc.gnu.org/egcs-1.1/index.html
20262
20263                                    EGCS 1.1
20264
20265   September 3, 1998: We are pleased to announce the release of EGCS 1.1.
20266   December 1, 1998: We are pleased to announce the release of EGCS 1.1.1.
20267   March 15, 1999: We are pleased to announce the release of EGCS 1.1.2.
20268
20269   EGCS is a free software project to further the development of the GNU
20270   compilers using an open development environment.
20271
20272   EGCS 1.1 is a major new release of the EGCS compiler system. It has
20273   been [1]extensively tested and is believed to be stable and suitable
20274   for widespread use.
20275
20276   EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8
20277   development sources; it contains all of the new features found in GCC
20278   2.8.1 as well as all new development from GCC up to June 6, 1998.
20279
20280   EGCS 1.1 also contains many improvements and features not found in GCC
20281   or in older versions of EGCS:
20282     * Global common subexpression elimination and global constant/copy
20283       propagation (aka [2]gcse)
20284     * Ongoing improvements to the [3]alias analysis support to allow for
20285       better optimizations throughout the compiler.
20286     * Vastly improved [4]C++ compiler and integrated C++ runtime
20287       libraries.
20288     * Fixes for the /tmp symlink race security problems.
20289     * New targets including mips16, arm-thumb and 64 bit PowerPC.
20290     * Improvements to GNU Fortran (g77) compiler and runtime library made
20291       since g77 version 0.5.23.
20292
20293   See the [5]new features page for a more complete list of new features
20294   found in EGCS 1.1 releases.
20295
20296   EGCS 1.1.1 is a minor update to fix several serious problems in EGCS
20297   1.1:
20298     * General improvements and fixes
20299          + Avoid some stack overflows when compiling large functions.
20300          + Avoid incorrect loop invariant code motions.
20301          + Fix some core dumps on Linux kernel code.
20302          + Bring back the imake -Di386 and friends fix from EGCS 1.0.2.
20303          + Fix code generation problem in gcse.
20304          + Various documentation related fixes.
20305     * g++/libstdc++ improvements and fixes
20306          + MT safe EH fix for setjmp/longjmp based exception handling.
20307          + Fix a few bad interactions between optimization and exception
20308            handling.
20309          + Fixes for demangling of template names starting with "__".
20310          + Fix a bug that would fail to run destructors in some cases
20311            with -O2.
20312          + Fix 'new' of classes with virtual bases.
20313          + Fix crash building Qt on the Alpha.
20314          + Fix failure compiling WIFEXITED macro on GNU/Linux.
20315          + Fix some -frepo failures.
20316     * g77 and libf2c improvements and fixes
20317          + Various documentation fixes.
20318          + Avoid compiler crash on RAND intrinsic.
20319          + Fix minor bugs in makefiles exposed by BSD make programs.
20320          + Define _XOPEN_SOURCE for libI77 build to avoid potential
20321            problems on some 64-bit systems.
20322          + Fix problem with implicit endfile on rewind.
20323          + Fix spurious recursive I/O errors.
20324     * platform specific improvements and fixes
20325          + Match all versions of UnixWare7.
20326          + Do not assume x86 SVR4 or UnixWare targets can handle stabs.
20327          + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion
20328            from unsigned ints to double precision floats.
20329          + Fix ARM ABI issue with NetBSD.
20330          + Fix a few arm code generation bugs.
20331          + Fixincludes will fix additional broken SCO OpenServer header
20332            files.
20333          + Fix a m68k backend bug which caused invalid offsets in reg+d
20334            addresses.
20335          + Fix problems with 64bit AIX 4.3 support.
20336          + Fix handling of long longs for varargs/stdarg functions on the
20337            ppc.
20338          + Minor fixes to CPP predefines for Windows.
20339          + Fix code generation problems with gpr<->fpr copies for 64bit
20340            ppc.
20341          + Fix a few coldfire code generation bugs.
20342          + Fix some more header file problems on SunOS 4.x.
20343          + Fix assert.h handling for RTEMS.
20344          + Fix Windows handling of TREE_SYMBOL_REFERENCED.
20345          + Fix x86 compiler abort in reg-stack pass.
20346          + Fix cygwin/windows problem with section attributes.
20347          + Fix Alpha code generation problem exposed by SMP Linux
20348            kernels.
20349          + Fix typo in m68k 32->64bit integer conversion.
20350          + Make sure target libraries build with -fPIC for PPC & Alpha
20351            targets.
20352
20353   EGCS 1.1.2 is a minor update to fix several serious problems in EGCS
20354   1.1.1:
20355     * General improvements and fixes
20356          + Fix bug in loop optimizer which caused the SPARC (and
20357            potentially other) ports to segfault.
20358          + Fix infinite recursion in alias analysis and combiner code.
20359          + Fix bug in regclass preferencing.
20360          + Fix incorrect loop reversal which caused incorrect code to be
20361            generated for several targets.
20362          + Fix return value for builtin memcpy.
20363          + Reduce compile time for certain loops which exposed quadratic
20364            behavior in the loop optimizer.
20365          + Fix bug which caused volatile memory to be written multiple
20366            times when only one write was needed/desired.
20367          + Fix compiler abort in caller-save.c
20368          + Fix combiner bug which caused incorrect code generation for
20369            certain division by constant operations.
20370          + Fix incorrect code generation due to a bug in range check
20371            optimizations.
20372          + Fix incorrect code generation due to mis-handling of clobbered
20373            values in CSE.
20374          + Fix compiler abort/segfault due to incorrect register
20375            splitting when unrolling loops.
20376          + Fix code generation involving autoincremented addresses with
20377            ternary operators.
20378          + Work around bug in the scheduler which caused qt to be
20379            mis-compiled on some platforms.
20380          + Fix code generation problems with -fshort-enums.
20381          + Tighten security for temporary files.
20382          + Improve compile time for codes which make heavy use of
20383            overloaded functions.
20384          + Fix multiply defined constructor/destructor symbol problems.
20385          + Avoid setting bogus RPATH environment variable during
20386            bootstrap.
20387          + Avoid GNU-make dependencies in the texinfo subdir.
20388          + Install CPP wrapper script in $(prefix)/bin if --enable-cpp.
20389            --enable-cpp=<dirname> can be used to specify an additional
20390            install directory for the cpp wrapper script.
20391          + Fix CSE bug which caused incorrect label-label refs to appear
20392            on some platforms.
20393          + Avoid linking in EH routines from libgcc if they are not
20394            needed.
20395          + Avoid obscure bug in aliasing code.
20396          + Fix bug in weak symbol handling.
20397     * Platform-specific improvements and fixes
20398          + Fix detection of PPro/PII on Unixware 7.
20399          + Fix compiler segfault when building spec99 and other programs
20400            for SPARC targets.
20401          + Fix code-generation bugs for integer and floating point
20402            conditional move instructions on the PPro/PII.
20403          + Use fixincludes to fix byteorder problems on i?86-*-sysv.
20404          + Fix build failure for the arc port.
20405          + Fix floating point format configuration for i?86-gnu port.
20406          + Fix problems with hppa1.0-hp-hpux10.20 configuration when
20407            threads are enabled.
20408          + Fix coldfire code generation bugs.
20409          + Fix "unrecognized insn" problems for Alpha and PPC ports.
20410          + Fix h8/300 code generation problem with floating point values
20411            in memory.
20412          + Fix unrecognized insn problems for the m68k port.
20413          + Fix namespace-pollution problem for the x86 port.
20414          + Fix problems with old assembler on x86 NeXT systems.
20415          + Fix PIC code-generation problems for the SPARC port.
20416          + Fix minor bug with LONG_CALLS in PowerPC SVR4 support.
20417          + Fix minor ISO namespace violation in Alpha varargs/stdarg
20418            support.
20419          + Fix incorrect "braf" instruction usage for the SH port.
20420          + Fix minor bug in va-sh which prevented its use with -ansi.
20421          + Fix problems recognizing and supporting FreeBSD.
20422          + Handle OpenBSD systems correctly.
20423          + Minor fixincludes fix for Digital UNIX 4.0B.
20424          + Fix problems with ctors/dtors in SCO shared libraries.
20425          + Abort instead of generating incorrect code for PPro/PII
20426            floating point conditional moves.
20427          + Avoid multiply defined symbols on GNU/Linux systems using
20428            libc-5.4.xx.
20429          + Fix abort in alpha compiler.
20430     * Fortran-specific fixes
20431          + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year
20432            is in the documented, non-Y2K-compliant range of 0-99, instead
20433            of being returned as 100 in the year 2000.
20434          + Fix the `Date_and_Time' intrinsic (in libg2c) to return the
20435            milliseconds value properly in Values(8).
20436          + Fix the `LStat' intrinsic (in libg2c) to return device-ID
20437            information properly in SArray(7).
20438
20439   Each release includes installation instructions in both HTML and
20440   plaintext forms (see the INSTALL directory in the toplevel directory of
20441   the distribution). However, we also keep the most up to date
20442   installation instructions and [6]build/test status on our web page. We
20443   will update those pages as new information becomes available.
20444
20445   The EGCS project would like to thank the numerous people that have
20446   contributed new features, test results, bugfixes, etc. This [7]amazing
20447   group of volunteers is what makes EGCS successful.
20448
20449   And finally, we can't in good conscience fail to mention some
20450   [8]caveats to using EGCS 1.1.
20451
20452   Download EGCS from egcs.cygnus.com (USA California).
20453
20454   The EGCS 1.1 release is also available on many mirror sites.
20455   [9]Goto mirror list to find a closer site.
20456
20457
20458    For questions related to the use of GCC, please consult these web
20459    pages and the [10]GCC manuals. If that fails, the
20460    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20461    web pages and the development of GCC are welcome on our developer
20462    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
20463    archives.
20464
20465   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
20466   distribution of this entire article is permitted in any medium,
20467   provided this notice is preserved.
20468
20469   These pages are [15]maintained by the GCC team. Last modified
20470   2023-01-21.
20471
20472References
20473
20474   1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html
20475   2. http://gcc.gnu.org/news/gcse.html
20476   3. http://gcc.gnu.org/news/alias.html
20477   4. http://gcc.gnu.org/egcs-1.1/c++features.html
20478   5. http://gcc.gnu.org/egcs-1.1/features.html
20479   6. http://gcc.gnu.org/egcs-1.1/buildstat.html
20480   7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
20481   8. http://gcc.gnu.org/egcs-1.1/caveats.html
20482   9. http://gcc.gnu.org/mirrors.html
20483  10. https://gcc.gnu.org/onlinedocs/
20484  11. mailto:gcc-help@gcc.gnu.org
20485  12. mailto:gcc@gcc.gnu.org
20486  13. https://gcc.gnu.org/lists.html
20487  14. https://www.fsf.org/
20488  15. https://gcc.gnu.org/about.html
20489======================================================================
20490http://gcc.gnu.org/egcs-1.1/features.html
20491
20492                             EGCS 1.1 new features
20493
20494     * Integrated GNU Fortran (g77) compiler and runtime library with
20495       improvements, based on g77 version 0.5.23.
20496     * Vast improvements in the C++ compiler; so many they have [1]page of
20497       their own!
20498     * Compiler implements [2]global common subexpression elimination and
20499       global copy/constant propagation.
20500     * More major improvements in the [3]alias analysis code.
20501     * More major improvements in the exception handling code to improve
20502       performance, lower static overhead and provide the infrastructure
20503       for future improvements.
20504     * The infamous /tmp symlink race security problems have been fixed.
20505     * The regmove optimization pass has been nearly completely rewritten
20506       to improve performance of generated code.
20507     * The compiler now recomputes register usage information before local
20508       register allocation. By providing more accurate information to the
20509       priority based allocator, we get better register allocation.
20510     * The register reloading phase of the compiler optimizes spill code
20511       much better than in previous releases.
20512     * Some bad interactions between the register allocator and
20513       instruction scheduler have been fixed, resulting in much better
20514       code for certain programs. Additionally, we have tuned the
20515       scheduler in various ways to improve performance of generated code
20516       for some architectures.
20517     * The compiler's branch shortening algorithms have been significantly
20518       improved to work better on targets which align jump targets.
20519     * The compiler now supports -Os to prefer optimizing for code space
20520       over optimizing for code speed.
20521     * The compiler will now totally eliminate library calls which compute
20522       constant values. This primarily helps targets with no integer
20523       div/mul support and targets without floating point support.
20524     * The compiler now supports an extensive "--help" option.
20525     * cpplib has been greatly improved and may be suitable for limited
20526       use.
20527     * Memory footprint for the compiler has been significantly reduced
20528       for some pathological cases.
20529     * The time to build EGCS has been improved for certain targets
20530       (particularly the alpha and mips platforms).
20531     * Many infrastructure improvements throughout the compiler, plus the
20532       usual mountain of bugfixes and minor improvements.
20533     * Target dependent improvements:
20534          + SPARC port now includes V8 plus and V9 support as well as
20535            performance tuning for Ultra class machines. The SPARC port
20536            now uses the Haifa scheduler.
20537          + Alpha port has been tuned for the EV6 processor and has an
20538            optimized expansion of memcpy/bzero. The Alpha port now uses
20539            the Haifa scheduler.
20540          + RS6000/PowerPC: support for the Power64 architecture and AIX
20541            4.3. The RS6000/PowerPC port now uses the Haifa scheduler.
20542          + x86: Alignment of static store data and jump targets is per
20543            Intel recommendations now. Various improvements throughout the
20544            x86 port to improve performance on Pentium processors
20545            (including improved epilogue sequences for Pentium chips and
20546            backend improvements which should help register allocation on
20547            all x86 variants. Conditional move support has been fixed and
20548            enabled for PPro processors. The x86 port also better supports
20549            64bit operations now. Unixware 7, a System V Release 5 target,
20550            is now supported and SCO OpenServer targets can support GAS.
20551          + MIPS has improved multiply/multiply-add support and now
20552            includes mips16 ISA support.
20553          + M68k has many micro-optimizations and Coldfire fixes.
20554     * Core compiler is based on the GCC development tree from June 9,
20555       1998, so we have all of the [4]features found in GCC 2.8.
20556
20557
20558    For questions related to the use of GCC, please consult these web
20559    pages and the [5]GCC manuals. If that fails, the
20560    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20561    web pages and the development of GCC are welcome on our developer
20562    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
20563
20564   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
20565   distribution of this entire article is permitted in any medium,
20566   provided this notice is preserved.
20567
20568   These pages are [10]maintained by the GCC team. Last modified
20569   2022-10-26.
20570
20571References
20572
20573   1. http://gcc.gnu.org/egcs-1.1/c++features.html
20574   2. http://gcc.gnu.org/news/gcse.html
20575   3. http://gcc.gnu.org/news/alias.html
20576   4. http://gcc.gnu.org/egcs-1.0/features-2.8.html
20577   5. https://gcc.gnu.org/onlinedocs/
20578   6. mailto:gcc-help@gcc.gnu.org
20579   7. mailto:gcc@gcc.gnu.org
20580   8. https://gcc.gnu.org/lists.html
20581   9. https://www.fsf.org/
20582  10. https://gcc.gnu.org/about.html
20583======================================================================
20584http://gcc.gnu.org/egcs-1.1/caveats.html
20585
20586                                EGCS 1.1 Caveats
20587
20588     * EGCS has an integrated libstdc++, but does not have an integrated
20589       libg++. Furthermore old libg++ releases will not work with EGCS; HJ
20590       Lu has made a libg++-2.8.1.2 snapshot available which may work with
20591       EGCS.
20592       Note most C++ programs only need libstdc++.
20593     * Exception handling may not work with shared libraries, particularly
20594       on alphas, hppas, rs6000/powerpc and mips based platforms.
20595       Exception handling is known to work on x86-linux platforms with
20596       shared libraries.
20597     * Some versions of the Linux kernel have bugs which prevent them from
20598       being compiled or from running when compiled by EGCS. See the FAQ
20599       (as shipped with EGCS 1.1) for additional information.
20600     * In general, EGCS is more rigorous about rejecting invalid C++ code
20601       or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As
20602       a result it may be necessary to fix C++ code before it will compile
20603       with EGCS.
20604     * G++ is also converting toward the ISO C++ standard; as a result
20605       code which was previously valid (and thus accepted by other
20606       compilers and older versions of g++) may no longer be accepted.
20607     * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x
20608       or GCC 2.8.x due to changes necessary to support thread safe
20609       exception handling.
20610
20611
20612    For questions related to the use of GCC, please consult these web
20613    pages and the [1]GCC manuals. If that fails, the
20614    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20615    web pages and the development of GCC are welcome on our developer
20616    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
20617
20618   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
20619   distribution of this entire article is permitted in any medium,
20620   provided this notice is preserved.
20621
20622   These pages are [6]maintained by the GCC team. Last modified
20623   2022-10-26.
20624
20625References
20626
20627   1. https://gcc.gnu.org/onlinedocs/
20628   2. mailto:gcc-help@gcc.gnu.org
20629   3. mailto:gcc@gcc.gnu.org
20630   4. https://gcc.gnu.org/lists.html
20631   5. https://www.fsf.org/
20632   6. https://gcc.gnu.org/about.html
20633======================================================================
20634http://gcc.gnu.org/egcs-1.0/index.html
20635
20636                                    EGCS 1.0
20637
20638   December 3, 1997: We are pleased to announce the release of EGCS 1.0.
20639   January 6, 1998: We are pleased to announce the release of EGCS 1.0.1.
20640   March 16, 1998: We are pleased to announce the release of EGCS 1.0.2.
20641   May 15, 1998 We are pleased to announce the release of EGCS 1.0.3.
20642
20643   EGCS is a collaborative effort involving several groups of hackers
20644   using an open development model to accelerate development and testing
20645   of GNU compilers and runtime libraries.
20646
20647   An important goal of EGCS is to allow wide scale testing of
20648   experimental features and optimizations; therefore, EGCS contains some
20649   features and optimizations which are still under development. However,
20650   EGCS has been carefully tested and should be comparable in quality to
20651   most GCC releases.
20652
20653   EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8
20654   development sources; it contains nearly all of the new features found
20655   in GCC 2.8.
20656
20657   EGCS 1.0 also contains many improvements and features not found in GCC
20658   2.7 and even the GCC 2.8 series (which was released after the original
20659   EGCS 1.0 release).
20660     * Integrated C++ runtime libraries, including support for most major
20661       GNU/Linux systems!
20662     * The integrated libstdc++ library includes a verbatim copy of SGI's
20663       STL release.
20664     * Integrated GNU Fortran compiler.
20665     * New instruction scheduler.
20666     * New alias analysis code.
20667
20668   See the [1]new features page for a more complete list of new features.
20669
20670   EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few
20671   critical bugs and add support for Red Hat 5.0 Linux. Changes since the
20672   EGCS 1.0 release:
20673     * Add support for Red Hat 5.0 Linux and better support for Linux
20674       systems using glibc2.
20675       Many programs failed to link when compiled with EGCS 1.0 on Red Hat
20676       5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should
20677       fix these problems.
20678     * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception
20679       handling interfaces.
20680       To avoid future compatibility problems, we strongly urge anyone who
20681       is planning on distributing shared libraries that contain C++ code
20682       to upgrade to EGCS 1.0.1 first.
20683       Soon after EGCS 1.0 was released, the GCC developers made some
20684       incompatible changes in libgcc's exception handling interfaces.
20685       These changes were needed to solve problems on some platforms. This
20686       means that GCC 2.8.0, when released, will not be seamlessly
20687       compatible with shared libraries built by EGCS 1.0. The reason is
20688       that the libgcc.a in GCC 2.8.0 will not contain a function needed
20689       by the old interface.
20690       The result of this is that there may be compatibility problems with
20691       shared libraries built by EGCS 1.0 when used with GCC 2.8.0.
20692       With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface,
20693       and libgcc.a has the support routines for both the old and the new
20694       interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed,
20695       and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed).
20696       The maintainers of GCC 2.x have decided against including seamless
20697       support for the old interface in 2.8.0, since it was never
20698       "official", so to avoid future compatibility problems we recommend
20699       against distributing any shared libraries built by EGCS 1.0 that
20700       contain C++ code (upgrade to 1.0.1 and use that).
20701     * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends.
20702       The x86 changes fix code generation errors exposed when building
20703       glibc2 and the usual GNU/Linux dynamic linker (ld.so).
20704       The hppa change fixes a compiler abort when configured for use with
20705       RTEMS.
20706       The MIPS changes fix problems with the definition of LONG_MAX on
20707       newer systems, allow for command line selection of the target ABI,
20708       and fix one code generation problem.
20709       The rs6000/ppc change fixes some problems with passing structures
20710       to varargs/stdarg functions.
20711     * A few machine independent bugfixes, mostly to fix code generation
20712       errors when building Linux kernels or glibc.
20713     * Fix a few critical exception handling and template bugs in the C++
20714       compiler.
20715     * Fix Fortran namelist bug on alphas.
20716     * Fix build problems on x86-solaris systems.
20717
20718   EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several
20719   serious problems in EGCS 1.0.1.
20720     * General improvements and fixes
20721          + Memory consumption significantly reduced, especially for
20722            templates and inline functions.
20723          + Fix various problems with glibc2.1.
20724          + Fix loop optimization bug exposed by rs6000/ppc port.
20725          + Fix to avoid potential code generation problems in jump.c.
20726          + Fix some undefined symbol problems in dwarf1 debug support.
20727     * g++/libstdc++ improvements and fixes
20728          + libstdc++ in the EGCS release has been updated and should be
20729            link compatible with libstdc++-2.8.
20730          + Various fixes in libio/libstdc++ to work better on GNU/Linux
20731            systems.
20732          + Fix problems with duplicate symbols on systems that do not
20733            support weak symbols.
20734          + Memory corruption bug and undefined symbols in bastring have
20735            been fixed.
20736          + Various exception handling fixes.
20737          + Fix compiler abort for very long thunk names.
20738     * g77 improvements and fixes
20739          + Fix compiler crash for omitted bound in Fortran CASE
20740            statement.
20741          + Add missing entries to g77 lang-options.
20742          + Fix problem with -fpedantic in the g77 compiler.
20743          + Fix "backspace" problem with g77 on alphas.
20744          + Fix x86 backend problem with Fortran literals and -fpic.
20745          + Fix some of the problems with negative subscripts for g77 on
20746            alphas.
20747          + Fixes for Fortran builds on cygwin32/mingw32.
20748     * platform specific improvements and fixes
20749          + Fix long double problems on x86 (exposed by glibc).
20750          + x86 ports define i386 again to keep imake happy.
20751          + Fix exception handling support on NetBSD ports.
20752          + Several changes to collect2 to fix many problems with AIX.
20753          + Define __ELF__ for GNU/Linux on rs6000.
20754          + Fix -mcall-linux problem on GNU/Linux on rs6000.
20755          + Fix stdarg/vararg problem for GNU/Linux on rs6000.
20756          + Allow autoconf to select a proper install problem on AIX 3.1.
20757          + m68k port support includes -mcpu32 option as well as cpu32
20758            multilibs.
20759          + Fix stdarg bug for irix6.
20760          + Allow EGCS to build on irix5 without the gnu assembler.
20761          + Fix problem with static linking on sco5.
20762          + Fix bootstrap on sco5 with native compiler.
20763          + Fix for abort building newlib on H8 target.
20764          + Fix fixincludes handling of math.h on SunOS.
20765          + Minor fix for Motorola 3300 m68k systems.
20766
20767   EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few
20768   problems reported by Red Hat for builds of Red Hat 5.1.
20769     * Generic bugfixes:
20770          + Fix a typo in the libio library which resulted in incorrect
20771            behavior of istream::get.
20772          + Fix the Fortran negative array index problem.
20773          + Fix a major problem with the ObjC runtime thread support
20774            exposed by glibc2.
20775          + Reduce memory consumption of the Haifa scheduler.
20776     * Target specific bugfixes:
20777          + Fix one x86 floating point code generation bug exposed by
20778            glibc2 builds.
20779          + Fix one x86 internal compiler error exposed by glibc2 builds.
20780          + Fix profiling bugs on the Alpha.
20781          + Fix ImageMagick & emacs 20.2 build problems on the Alpha.
20782          + Fix rs6000/ppc bug when converting values from integer types
20783            to floating point types.
20784
20785   The EGCS 1.0 releases include installation instructions in both HTML
20786   and plaintext forms (see the INSTALL directory in the toplevel
20787   directory of the distribution). However, we also keep the most up to
20788   date installation instructions and [2]build/test status on our web
20789   page. We will update those pages as new information becomes available.
20790
20791   And, we can't in good conscience fail to mention some [3]caveats to
20792   using EGCS.
20793
20794   Update: Big thanks to Stanford for providing a high speed link for
20795   downloading EGCS (go.cygnus.com)!
20796
20797   Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com
20798   (USA California -- High speed link provided by Stanford).
20799
20800   The EGCS 1.0 release is also available many mirror sites.
20801   [4]Goto mirror list to find a closer site
20802
20803   We'd like to thank the numerous people that have contributed new
20804   features, test results, bugfixes, etc. Unfortunately, they're far too
20805   numerous to mention by name.
20806
20807
20808    For questions related to the use of GCC, please consult these web
20809    pages and the [5]GCC manuals. If that fails, the
20810    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20811    web pages and the development of GCC are welcome on our developer
20812    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
20813
20814   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
20815   distribution of this entire article is permitted in any medium,
20816   provided this notice is preserved.
20817
20818   These pages are [10]maintained by the GCC team. Last modified
20819   2023-01-21.
20820
20821References
20822
20823   1. http://gcc.gnu.org/egcs-1.0/features.html
20824   2. http://gcc.gnu.org/egcs-1.0/buildstat.html
20825   3. http://gcc.gnu.org/egcs-1.0/caveats.html
20826   4. http://gcc.gnu.org/mirrors.html
20827   5. https://gcc.gnu.org/onlinedocs/
20828   6. mailto:gcc-help@gcc.gnu.org
20829   7. mailto:gcc@gcc.gnu.org
20830   8. https://gcc.gnu.org/lists.html
20831   9. https://www.fsf.org/
20832  10. https://gcc.gnu.org/about.html
20833======================================================================
20834http://gcc.gnu.org/egcs-1.0/features.html
20835
20836                               EGCS 1.0 features
20837
20838     * Core compiler is based on the gcc2 development tree from Aug 2,
20839       1997, so we have most of the [1]features found in GCC 2.8.
20840     * Integrated GNU Fortran compiler based on g77-0.5.22-19970929.
20841     * Vast improvements in the C++ compiler; so many they have [2]page of
20842       their own!
20843     * Integrated C++ runtime libraries, including support for most major
20844       GNU/Linux systems!
20845     * New instruction scheduler from IBM Haifa which includes support for
20846       function wide instruction scheduling as well as superscalar
20847       scheduling.
20848     * Significantly improved alias analysis code.
20849     * Improved register allocation for two address machines.
20850     * Significant code generation improvements for Fortran code on
20851       Alphas.
20852     * Various optimizations from the g77 project as well as improved loop
20853       optimizations.
20854     * Dwarf2 debug format support for some targets.
20855     * egcs libstdc++ includes the SGI STL implementation without changes.
20856     * As a result of these and other changes, egcs libstc++ is not binary
20857       compatible with previous releases of libstdc++.
20858     * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO
20859       Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and
20860       1.1), Support for RTEMS on several embedded targets, Support for
20861       arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and
20862       MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc.
20863     * Integrated testsuites for gcc, g++, g77, libstdc++ and libio.
20864     * RS6000/PowerPC ports generate code which can run on all
20865       RS6000/PowerPC variants by default.
20866     * -mcpu= and -march= switches for the x86 port to allow better
20867       control over how the x86 port generates code.
20868     * Includes the template repository patch (aka repo patch); note the
20869       new template code makes repo obsolete for ELF systems using gnu-ld
20870       such as GNU/Linux.
20871     * Plus the usual assortment of bugfixes and improvements.
20872
20873
20874    For questions related to the use of GCC, please consult these web
20875    pages and the [3]GCC manuals. If that fails, the
20876    [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20877    web pages and the development of GCC are welcome on our developer
20878    list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives.
20879
20880   Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and
20881   distribution of this entire article is permitted in any medium,
20882   provided this notice is preserved.
20883
20884   These pages are [8]maintained by the GCC team. Last modified
20885   2022-10-26.
20886
20887References
20888
20889   1. http://gcc.gnu.org/egcs-1.0/features-2.8.html
20890   2. http://gcc.gnu.org/egcs-1.0/c++features.html
20891   3. https://gcc.gnu.org/onlinedocs/
20892   4. mailto:gcc-help@gcc.gnu.org
20893   5. mailto:gcc@gcc.gnu.org
20894   6. https://gcc.gnu.org/lists.html
20895   7. https://www.fsf.org/
20896   8. https://gcc.gnu.org/about.html
20897======================================================================
20898http://gcc.gnu.org/egcs-1.0/caveats.html
20899
20900                                EGCS 1.0 Caveats
20901
20902     * EGCS has an integrated libstdc++, but does not have an integrated
20903       libg++. Furthermore old libg++ releases will not work with egc; HJ
20904       Lu has made a libg++-2.8.1.2 available which may work with EGCS.
20905       Note most C++ programs only need libstdc++.
20906     * Note that using -pedantic or -Wreturn-type can cause an explosion
20907       in the amount of memory needed for template-heavy C++ code, such as
20908       code that uses STL. Also note that -Wall includes -Wreturn-type, so
20909       if you use -Wall you will need to specify -Wno-return-type to turn
20910       it off.
20911     * Exception handling may not work with shared libraries, particularly
20912       on alphas, hppas, and mips based platforms. Exception handling is
20913       known to work on x86-linux platforms with shared libraries.
20914     * Some versions of the Linux kernel have bugs which prevent them from
20915       being compiled or from running when compiled by EGCS. See the FAQ
20916       (as shipped with EGCS 1.0) for additional information.
20917     * In general, EGCS is more rigorous about rejecting invalid C++ code
20918       or deprecated C++ constructs than G++ 2.7. As a result it may be
20919       necessary to fix C++ code before it will compile with EGCS.
20920     * G++ is also aggressively tracking the C++ standard; as a result
20921       code which was previously valid (and thus accepted by other
20922       compilers and older versions of G++) may no longer be accepted.
20923     * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS
20924       1.0.x and later releases should work with Red Hat Linux 5.0.
20925
20926
20927    For questions related to the use of GCC, please consult these web
20928    pages and the [1]GCC manuals. If that fails, the
20929    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20930    web pages and the development of GCC are welcome on our developer
20931    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
20932
20933   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
20934   distribution of this entire article is permitted in any medium,
20935   provided this notice is preserved.
20936
20937   These pages are [6]maintained by the GCC team. Last modified
20938   2022-10-26.
20939
20940References
20941
20942   1. https://gcc.gnu.org/onlinedocs/
20943   2. mailto:gcc-help@gcc.gnu.org
20944   3. mailto:gcc@gcc.gnu.org
20945   4. https://gcc.gnu.org/lists.html
20946   5. https://www.fsf.org/
20947   6. https://gcc.gnu.org/about.html
20948======================================================================
20949