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-11/index.html 9 10 GCC 11 Release Series 11 12 July 28, 2021 13 14 The GCC developers are pleased to announce the release of GCC 11.2. 15 16 This release is a bug-fix release, containing fixes for regressions in 17 GCC 11.1 relative to previous releases of GCC. 18 19Release History 20 21 GCC 11.2 22 July 28, 2021 ([1]changes, [2]documentation) 23 24 GCC 11.1 25 April 27, 2021 ([3]changes, [4]documentation) 26 27References and Acknowledgements 28 29 GCC used to stand for the GNU C Compiler, but since the compiler 30 supports several other languages aside from C, it now stands for the 31 GNU Compiler Collection. 32 33 A list of [5]successful builds is updated as new information becomes 34 available. 35 36 The GCC developers would like to thank the numerous people that have 37 contributed new features, improvements, bug fixes, and other changes as 38 well as test results to GCC. This [6]amazing group of volunteers is 39 what makes GCC successful. 40 41 For additional information about GCC please refer to the [7]GCC project 42 web site or contact the [8]GCC development mailing list. 43 44 To obtain GCC please use [9]our mirror sites or [10]our version control 45 system. 46 47 48 For questions related to the use of GCC, please consult these web 49 pages and the [11]GCC manuals. If that fails, the 50 [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these 51 web pages and the development of GCC are welcome on our developer 52 list at [13]gcc@gcc.gnu.org. All of [14]our lists have public 53 archives. 54 55 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 56 distribution of this entire article is permitted in any medium, 57 provided this notice is preserved. 58 59 These pages are [16]maintained by the GCC team. Last modified 60 2021-07-28[17]. 61 62References 63 64 1. http://gcc.gnu.org/gcc-11/changes.html 65 2. http://gcc.gnu.org/onlinedocs/11.2.0/ 66 3. http://gcc.gnu.org/gcc-11/changes.html 67 4. http://gcc.gnu.org/onlinedocs/11.1.0/ 68 5. http://gcc.gnu.org/gcc-11/buildstat.html 69 6. http://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Contributors.html 70 7. http://gcc.gnu.org/index.html 71 8. mailto:gcc@gcc.gnu.org 72 9. http://gcc.gnu.org/mirrors.html 73 10. http://gcc.gnu.org/git.html 74 11. https://gcc.gnu.org/onlinedocs/ 75 12. mailto:gcc-help@gcc.gnu.org 76 13. mailto:gcc@gcc.gnu.org 77 14. https://gcc.gnu.org/lists.html 78 15. https://www.fsf.org/ 79 16. https://gcc.gnu.org/about.html 80 17. http://validator.w3.org/check/referer 81====================================================================== 82http://gcc.gnu.org/gcc-11/changes.html 83 84 GCC 11 Release Series 85 Changes, New Features, and Fixes 86 87 This page is a "brief" summary of some of the huge number of 88 improvements in GCC 11. You may also want to check out our [1]Porting 89 to GCC 11 page and the [2]full GCC documentation. 90 91Caveats 92 93 * The default mode for C++ is now -std=gnu++17 instead of 94 -std=gnu++14. Note that [3]C++17 changes to template template 95 parameter matching can be disabled independently of other features 96 with -fno-new-ttp-matching. 97 * When building GCC itself, the host compiler must now support C++11, 98 rather than C++98. In particular bootstrapping GCC 11 using an 99 older version of GCC requires a binary of GCC 4.8 or later, rather 100 than of GCC 3.4 or later as was the case for bootstrapping GCC 10. 101 * Naming and location of auxiliary and dump output files changed. If 102 you compile multiple input files in a single command, if you enable 103 Link Time Optimization, or if you use -dumpbase, -dumpdir, 104 -save-temps=*, and you expect any file other than the primary 105 output file(s) to be created as a side effect, watch out for 106 improvements and a few surprises. See [4]the patch, particularly 107 its textual description, for more details about the changes. 108 * -gsplit-dwarf no longer enables debug info generation on its own 109 but requires a separate -g for this. 110 * The libstdc++ configure option --enable-cheaders=c_std is 111 deprecated and will be removed in a future release. It should be 112 possible to use --enable-cheaders=c_global (the default) with no 113 change in behaviour. 114 * The front end for compiling BRIG format of Heterogeneous System 115 Architecture Intermediate Language (HSAIL) has been deprecated and 116 will likely be removed in a future release. 117 * Some short options of the gcov tool have been renamed: -i to -j and 118 -j to -H. 119 120General Improvements 121 122 * [5]ThreadSanitizer improvements to support alternative runtimes and 123 environments. The [6]Linux Kernel Concurrency Sanitizer (KCSAN) is 124 now supported. 125 + Add --param tsan-distinguish-volatile to optionally emit 126 instrumentation distinguishing volatile accesses. 127 + Add --param tsan-instrument-func-entry-exit to optionally 128 control if function entries and exits should be instrumented. 129 * In previous releases of GCC, the "column numbers" emitted in 130 diagnostics were actually a count of bytes from the start of the 131 source line. This could be problematic, both because of: 132 + multibyte characters (requiring more than one byte to encode), 133 and 134 + multicolumn characters (requiring more than one column to 135 display in a monospace font) 136 For example, the character p ("GREEK SMALL LETTER PI (U+03C0)") 137 occupies one column, and its UTF-8 encoding requires two bytes; the 138 character 🙂 ("SLIGHTLY SMILING FACE (U+1F642)") occupies 139 two columns, and its UTF-8 encoding requires four bytes. 140 In GCC 11 the column numbers default to being column numbers, 141 respecting multi-column characters. The old behavior can be 142 restored using a new option [7]-fdiagnostics-column-unit=byte. 143 There is also a new option [8]-fdiagnostics-column-origin=, 144 allowing the pre-existing default of the left-hand column being 145 column 1 to be overridden if desired (e.g. for 0-based columns). 146 The output of [9]-fdiagnostics-format=json has been extended to 147 supply both byte counts and column numbers for all source 148 locations. 149 Additionally, in previous releases of GCC, tab characters in the 150 source would be emitted verbatim when quoting source code, but be 151 prefixed with whitespace or line number information, leading to 152 misalignments in the resulting output when compared with the actual 153 source. Tab characters are now printed as an appropriate number of 154 spaces, using the [10]-ftabstop option (which defaults to 8 spaces 155 per tab stop). 156 * Introduce [11]Hardware-assisted AddressSanitizer support. This 157 sanitizer currently only works for the AArch64 target. It helps 158 debug address problems similarly to [12]AddressSanitizer but is 159 based on partial hardware assistance and provides probabilistic 160 protection to use less RAM at run time. [13]Hardware-assisted 161 AddressSanitizer is not production-ready for user space, and is 162 provided mainly for use compiling the Linux Kernel. 163 To use this sanitizer the command line arguments are: 164 + -fsanitize=hwaddress to instrument userspace code. 165 + -fsanitize=kernel-hwaddress to instrument kernel code. 166 * For targets that produce DWARF debugging information GCC now 167 defaults to [14]DWARF version 5 (with the exception of VxWorks and 168 Darwin/Mac OS X which default to version 2 and AIX which defaults 169 to version 4). This can produce up to 25% more compact debug 170 information compared to earlier versions. 171 To take full advantage of DWARF version 5 GCC needs to be build 172 against binutils version 2.35.2 or higher. When GCC is build 173 against earlier versions of binutils GCC will still emit DWARF 174 version 5 for most debuginfo data, but will generate version 4 175 debug line tables (even when explicitly given -gdwarf-5). 176 The following debug information consumers can process DWARF version 177 5: 178 + GDB 8.0, or higher 179 + valgrind 3.17.0 180 + elfutils 0.172, or higher (for use with systemtap, 181 dwarves/pahole, perf and libabigail) 182 + dwz 0.14 183 Programs embedding libbacktrace are urged to upgrade to the version 184 shipping with GCC 11. 185 To make GCC 11 generate an older DWARF version use -g together with 186 -gdwarf-2, -gdwarf-3 or -gdwarf-4. 187 * Vectorizer improvements: 188 + The straight-line code vectorizer now considers the whole 189 function when vectorizing and can handle opportunities 190 crossing CFG merges and backedges. 191 * A series of conditional expressions that compare the same variable 192 can be transformed into a switch statement if each of them contains 193 a comparison expression. Example: 194 int IsHTMLWhitespace(int aChar) { 195 return aChar == 0x0009 || aChar == 0x000A || 196 aChar == 0x000C || aChar == 0x000D || 197 aChar == 0x0020; 198 } 199 200 This statement can be transformed into a switch statement and then 201 expanded into a bit-test. 202 * New command-line options: 203 + [15]-fbit-tests, enabled by default, can be used to enable or 204 disable switch expansion using bit-tests. 205 * Inter-procedural optimization improvements: 206 + A new IPA-modref pass was added to track side effects of 207 function calls and improve precision of points-to-analysis. 208 The pass can be controlled by the [16]-fipa-modref option. 209 + The identical code folding pass (controlled by [17]-fipa-icf) 210 was significantly improved to increase the number of unified 211 functions and to reduce compile-time memory use. 212 + IPA-CP (Interprocedural constant propagation) heuristics 213 improved its estimation of potential usefulness of known loop 214 bounds and strides by taking the estimated frequency of these 215 loops into account. 216 * Link-time optimization improvements: 217 + The LTO bytecode format was optimized for smaller object files 218 and faster streaming. 219 + Memory allocation of the linking stage was improved to reduce 220 peak memory use. 221 * Profile driven optimization improvements: 222 + Using [18]-fprofile-values, was improved by tracking more 223 target values for e.g. indirect calls. 224 + GCOV data file format outputs smaller files by representing 225 zero counters in a more compact way. 226 227New Languages and Language specific improvements 228 229 * GCC 11 adds support for non-rectangular loop nests in OpenMP 230 constructs and the allocator routines of [19]OpenMP 5.0, including 231 initial allocate clause support in C/C++. The OMP_TARGET_OFFLOAD 232 environment variable and the active-levels routines are now 233 supported. For C/C++, the declare variant and map support has been 234 extended. For Fortran, OpenMP 4.5 is now fully supported and OpenMP 235 5.0 support has been extended, including the following features 236 which were before only available in C and C++: order(concurrent), 237 device_type, memorder-clauses for flush, lastprivate with 238 conditional modifier, atomic construct and reduction clause 239 extensions of OpenMP 5.0, if clause with simd and cancel modifiers, 240 target data without map clause, and limited support for the 241 requires construct. 242 * Version 2.6 of the [20]OpenACC specification continues to be 243 maintained and improved in the C, C++ and Fortran compilers. See 244 the [21]implementation status section on the OpenACC wiki page and 245 the [22]run-time library documentation for further information. 246 247 C family 248 249 * New attributes: 250 + The [23]no_stack_protector attribute has been added to mark 251 functions which should not be instrumented with stack 252 protection (-fstack-protector). 253 + The existing [24]malloc attribute has been extended so that it 254 can be used to identify allocator/deallocator API pairs. A 255 pair of new [25]-Wmismatched-dealloc and 256 [26]-Wmismatched-new-delete warnings will complain about 257 mismatched calls, and [27]-Wfree-nonheap-object about 258 deallocation calls with pointers not obtained from allocation 259 functions. Additionally, the static analyzer will use these 260 attributes when checking for leaks, double-frees, 261 use-after-frees, and similar issues. 262 * New warnings: 263 + [28]-Wmismatched-dealloc, enabled by default, warns about 264 calls to deallocation functions with pointers returned from 265 mismatched allocation functions. 266 + [29]-Wsizeof-array-div, enabled by -Wall, warns about 267 divisions of two sizeof operators when the first one is 268 applied to an array and the divisor does not equal the size of 269 the array element. 270 + [30]-Wstringop-overread, enabled by default, warns about calls 271 to string functions reading past the end of the arrays passed 272 to them as arguments. In prior GCC releases most instances of 273 his warning are diagnosed by -Wstringop-overflow. 274 + [31]-Wtsan, enabled by default, warns about unsupported 275 features in ThreadSanitizer (currently 276 std::atomic_thread_fence). 277 * Enhancements to existing warnings: 278 + [32]-Wfree-nonheap-object detects many more instances of calls 279 to deallocation functions with pointers not returned from a 280 dynamic memory allocation function. 281 + [33]-Wmaybe-uninitialized diagnoses passing pointers or 282 references to uninitialized memory to functions taking 283 const-qualified arguments. 284 + [34]-Wuninitialized detects reads from uninitialized 285 dynamically allocated memory. 286 * For ELF targets that support the GNU or FreeBSD OSABIs, the used 287 attribute will now save the symbol declaration it is applied to 288 from linker garbage collection. 289 To support this behavior, used symbols that have not been placed in 290 specific sections (e.g. with the section attribute, or the 291 -f{function,data}-sections options) will be placed in new, unique 292 sections. 293 This functionality requires Binutils version 2.36 or later. 294 295 C 296 297 * Several new features from the upcoming C2X revision of the ISO C 298 standard are supported with -std=c2x and -std=gnu2x. Some of these 299 features are also supported as extensions when compiling for older 300 language versions. In addition to the features listed, some 301 features previously supported as extensions and now added to the C 302 standard are enabled by default in C2X mode and not diagnosed with 303 -std=c2x -Wpedantic. 304 + The BOOL_MAX and BOOL_WIDTH macros are provided in <limits.h>. 305 + As in C++, function definitions no longer need to give names 306 for unused function parameters. 307 + The expansions of the true and false macros in <stdbool.h> 308 have changed so that they have type bool. 309 + The [[nodiscard]] standard attribute is now supported. 310 + The __has_c_attribute preprocessor operator is now supported. 311 + Macros INFINITY, NAN, FLT_SNAN, DBL_SNAN, LDBL_SNAN, 312 DEC_INFINITY, DEC_NAN, and corresponding signaling NaN macros 313 for _FloatN, _FloatNx and _DecimalN types, are provided in 314 <float.h>. There are also corresponding built-in functions 315 __builtin_nansdN for decimal signaling NaNs. 316 + Macros FLT_IS_IEC_60559, DBL_IS_IEC_60559 and 317 LDBL_IS_IEC_60559 are provided in <float.h>. 318 + The feature test macro __STDC_WANT_IEC_60559_EXT__ is 319 supported by <float.h>. 320 + Labels may appear before declarations and at the end of a 321 compound statement. 322 * New warnings: 323 + [35]-Warray-parameter, enabled by -Wall, warns about 324 redeclarations of functions with ordinary array arguments 325 declared using inconsistent forms. The warning also enables 326 the detection of the likely out of bounds accesses in calls to 327 such functions with smaller arrays. 328 + [36]-Wvla-parameter, enabled by -Wall, warns redeclarations of 329 functions with variable length array arguments declared using 330 inconsistent forms or with mismatched bounds. The warning also 331 enables the detection of the likely out of bounds accesses in 332 calls to such functions with smaller arrays. 333 334 C++ 335 336 * The default mode has been changed to -std=gnu++17. 337 * Several C++20 features have been implemented: 338 + the compiler now supports consteval virtual functions 339 + P2082R1, Fixing CTAD for aggregates 340 + P0593R6, Pseudo-destructors end object lifetimes 341 + P1907R1, Inconsistencies with non-type template parameters 342 (complete implementation) 343 + P1975R0, Fixing the wording of parenthesized 344 aggregate-initialization 345 + P1009R2, Array size deduction in new-expressions 346 + P1099R5, using enum 347 + Modules, Requires -fmodules-ts and some aspects are 348 incomplete. Refer to [37]C++ 20 Status 349 * The C++ front end has experimental support for some of the upcoming 350 C++23 draft features with the -std=c++23, -std=gnu++23, -std=c++2b 351 or -std=gnu++2b flags, including 352 + P0330R8, Literal Suffix for (signed) size_t. 353 For a full list of new features, see [38]the C++ status page. 354 * Several C++ Defect Reports have been resolved, e.g.: 355 + DR 625, Use of auto as a template-argument 356 + DR 1512, Pointer comparison vs qualification conversions 357 + DR 1722, Should lambda to function pointer conversion function 358 be noexcept? 359 + DR 1914, Duplicate standard attributes 360 + DR 2032, Default template-arguments of variable templates 361 + DR 2289, Uniqueness of decomposition declaration names 362 + DR 2237, Can a template-id name a constructor? 363 + DR 2303, Partial ordering and recursive variadic inheritance 364 + DR 2369, Ordering between constraints and substitution 365 + DR 2450, braced-init-list as a template-argument 366 * G++ now performs better access checking in templates ([39]PR41437). 367 * reinterpret_casts in constexpr evaluation are now checked more 368 completely ([40]PR95307). 369 * The diagnostic for static_assert has been improved: the compiler 370 now shows the expression including its template arguments (if there 371 were any), and can point to the failing clause if the condition 372 comprised of any logical AND operators ([41]PR97518). 373 * New warnings: 374 + [42]-Wctad-maybe-unsupported, disabled by default, warns about 375 performing class template argument deduction on a type with no 376 deduction guides. 377 + [43]-Wrange-loop-construct, enabled by -Wall, warns when a 378 range-based for-loop is creating unnecessary and expensive 379 copies. 380 + [44]-Wdeprecated-enum-enum-conversion, enabled by default in 381 C++20, warns about deprecated arithmetic conversions on 382 operands of enumeration types, as outlined in 383 [depr.arith.conv.enum]. 384 + [45]-Wdeprecated-enum-float-conversion, enabled by default in 385 C++20, warns about deprecated arithmetic conversions on 386 operands where one is of enumeration type and the other is of 387 a floating-point type, as outlined in [depr.arith.conv.enum]. 388 + [46]-Wmismatched-new-delete, enabled by -Wall, warns about 389 calls to C++ operator delete with pointers returned from 390 mismatched forms of operator new or from other mismatched 391 allocation functions. 392 + [47]-Wvexing-parse, enabled by default, warns about the most 393 vexing parse rule: the cases when a declaration looks like a 394 variable definition, but the C++ language requires it to be 395 interpreted as a function declaration. 396 * Enhancements to existing warnings: 397 + [48]-Wnonnull considers the implicit this argument of every 398 C++ nonstatic member function to have been implicitly declared 399 with attribute nonnull and triggers warnings for calls where 400 the pointer is null. 401 402 Runtime Library (libstdc++) 403 404 * Improved C++17 support, including: 405 + std::from_chars and std::to_chars for floating-point types. 406 * Improved experimental C++20 support, including: 407 + Calendar additions to <chrono>. Thanks to Cassio Neri for 408 optimizations. 409 + std::bit_cast 410 + std::source_location 411 + Atomic wait and notify operations. 412 + <barrier>, <latch>, and <semaphore> 413 + <syncstream> 414 + Efficient access to basic_stringbuf's buffer. 415 + Heterogeneous lookup in unordered containers. 416 * Experimental C++23 support, including: 417 + contains member functions for strings, thanks to Paul Fee. 418 + std::to_underlying, std::is_scoped_enum 419 * Experimental support for Data-Parallel Types (simd) from the 420 Parallelism 2 TS, thanks to Matthias Kretz. 421 * Faster std::uniform_int_distribution, thanks to Daniel Lemire. 422 423 D 424 425 * New features: 426 + A new bottom type typeof(*null) has been added to represent 427 run-time errors and non-terminating functions. This also 428 introduces a new standard alias for the type named noreturn, 429 and is implicitly imported into every module. 430 + Printf-like and scanf-like functions are now detected by 431 prefixing them with pragma(printf) for printf-like functions 432 or pragma(scanf) for scanf-like functions. 433 + The __traits() expression now supports the extensions 434 isDeprecated, isDisabled, isFuture, isModule, isPackage, 435 child, isReturnOnStack, isZeroInit, getTargetInfo, 436 getLocation, hasPostblit, isCopyable, getVisibility, and 437 totype. 438 + An expression-based contract syntax has been added to the 439 language. 440 + Function literals can now return a value by reference with the 441 ref keyword. 442 + A new syntax is available to declare aliases to function types 443 using the alias syntax based on the assignment operator. 444 + New types __c_complex_float, __c_complex_double, 445 __c_complex_real, and __c_wchar_t have been added for 446 interfacing with C and C++ code, and are available from the 447 core.stdc.config module. 448 + User-defined attributes can now be used to annotate enum 449 members, alias declarations, and function parameters. 450 + Templates alias parameters can now be instantiated with basic 451 types such as int or void function(). 452 + The mixin construct can now be used as types in the form 453 mixin(string) var. 454 + The mixin construct can now take an argument list, same as 455 pragma(msg). 456 * New intrinsics: 457 + Bitwise rotate intrinsics core.bitop.rol and core.bitop.ror 458 have been added. 459 + Byte swap intrinsic core.bitop.byteswap for swapping bytes in 460 a 2-byte ushort has been added. 461 + Math intrinsics available from core.math now have overloads 462 for float and double types. 463 + Volatile intrinsics core.volatile.volatileLoad and 464 core.volatile.volatileStore have been moved from the 465 core.bitop module. 466 * New attributes: 467 + The following GCC attributes are now recognized and available 468 from the gcc.attributes module with short-hand aliases for 469 convenience: 470 o @attribute("alloc_size", arguments) or 471 @alloc_size(arguments). 472 o @attribute("always_inline") or @always_inline. 473 o @attribute("used") or @used. 474 o @attribute("optimize", arguments) or 475 @optimize(arguments). 476 o @attribute("cold") or @cold. 477 o @attribute("noplt") or @noplt. 478 o @attribute("target_clones", arguments) or 479 @target_clones(arguments). 480 o @attribute("no_icf") or @no_icf. 481 o @attribute("noipa") or @noipa. 482 o @attribute("symver", arguments) or @symver(arguments). 483 + New aliases have been added to gcc.attributes for 484 compatibility with ldc.attributes. 485 o The @allocSize(arguments) attribute is the same as 486 @alloc_size(arguments), but uses a 0-based index for 487 function arguments. 488 o The @assumeUsed attribute is an alias for 489 @attribute("used"). 490 o The @fastmath attribute is an alias for 491 @optimize("Ofast"). 492 o The @naked attribute is an alias for @attribute("naked"). 493 This attribute may not be available on all targets. 494 o The @restrict attribute has been added to specify that a 495 function parameter is to be restrict-qualified in the C99 496 sense of the term. 497 o The @optStrategy(strategy) attribute is an alias for 498 @optimize("O0") when the strategy is "none", otherwise 499 @optimize("Os") for the "optsize" and "minsize" 500 strategies. 501 o The @polly attribute is an alias for 502 @optimize("loop-parallelize-all"). 503 o The @section(name) attribute is an alias for 504 @attribute("section", name). 505 o The @target(arguments) attribute is an alias for 506 attribute("target", arguments). 507 o The @weak attribute is an alias for @attribute("weak"). 508 * New language options: 509 + -fweak-templates, added to control whether declarations that 510 can be defined in multiple objects should be emitted as weak 511 symbols. The default is to emit all symbols with extern 512 linkage as weak, unless the target lacks support for weak 513 symbols. 514 + -Wdeprecated, this option is now enabled by default. 515 + -Wextra, this option now turns on all warnings that are not 516 part of the core D language front-end - -Waddress, 517 -Wcast-result, -Wunknown-pragmas. 518 + -Wvarargs, added to turn on warnings about questionable usage 519 of the va_start intrinsic. 520 * Deprecated and removed features: 521 + Compiler-recognized attributes are now made available from the 522 gcc.attributes module, the former module gcc.attribute has 523 been deprecated and will be removed in a future release. 524 + The @attribute("alias") attribute has been removed, as it had 525 been superseded by pragma(mangle). 526 + The @attribute("forceinline") attribute has been removed and 527 renamed to @always_inline. 528 + __vector types that are not supported in hardware are now 529 rejected at compile-time. Previously all vector types were 530 accepted by the compiler and emulated when target support was 531 absent. 532 + The extern(Pascal) linkage attribute has been removed. 533 + The deprecation phase for -ftransition=import and 534 -ftransition=checkimports is finished. These switches no 535 longer have an effect and are now removed. Symbols that are 536 not visible in a particular scope will no longer be found by 537 the compiler. 538 + It is now an error to use private variables selectively 539 imported from other modules. Due to a bug, some imported 540 private members were visible from other modules, violating the 541 specification. 542 + The -fweak compiler switch has been removed, as it existed 543 only for testing. 544 545 Fortran 546 547 * Added DEPRECATED to !GCC$'s attributes directive. 548 549 Go 550 551 * GCC 11 provides a complete implementation of the Go 1.16.3 user 552 packages. 553 554libgccjit 555 556 * libgccjit was marked as merely "Alpha" quality when [49]originally 557 added in GCC 5. Given that we have maintained [50]API and ABI 558 compatibility since then and it is in use by various projects, we 559 have removed that caveat. 560 * libgccjit can now be built for MinGW 561 * The libgccjit API gained 10 new entry points: 562 + [51]gcc_jit_global_set_initializer 563 + 9 entrypoints for [52]directly embedding asm statements into a 564 compile, analogous to inline asm in the C front end 565 566New Targets and Target Specific Improvements 567 568 AArch64 & arm 569 570 * A number of new CPUs are supported through arguments to the -mcpu 571 and -mtune options in both the arm and aarch64 backends (GCC 572 identifiers in parentheses): 573 + Arm Cortex-A78 (cortex-a78). 574 + Arm Cortex-A78AE (cortex-a78ae). 575 + Arm Cortex-A78C (cortex-a78c). 576 + Arm Cortex-X1 (cortex-x1). 577 + Arm Neoverse V1 (neoverse-v1). 578 + Arm Neoverse N2 (neoverse-n2). 579 * GCC can now auto-vectorize operations performing addition, 580 subtraction, multiplication and the accumulate/subtract variants on 581 complex numbers, taking advantage of the Advanced SIMD instructions 582 in the Armv8.3-a (AArch64/AArch32), SVE (AArch64), SVE2 (AArch64) 583 and MVE (AArch32 M-profile) instruction sets. 584 585 AArch64 586 587 * In addition to the above, the following AArch64-only CPUs are now 588 supported: 589 + Fujitsu A64FX (a64fx). 590 + Arm Cortex-R82 (cortex-r82). 591 * The AArch64 Armv8-R architecture is now supported through the 592 -march=armv8-r option. 593 * Mitigation against the [53]Straight-line Speculation vulnerability 594 is supported with the -mharden-sls= option. Please refer to the 595 documentation for usage instructions. 596 * The availability of Advanced SIMD intrinsics available through the 597 arm_neon.h header is improved and GCC 11 supports the full set of 598 intrinsics defined by ACLE Q3 2020. 599 600 AMD Radeon (GCN) 601 602 * Initial support for gfx908 GPUs has been added. 603 604 arm 605 606 * Initial auto-vectorization is now available when targeting the MVE 607 instruction set. 608 * GCC can now make use of the Low Overhead Branch instruction in 609 Armv8.1-M to optimize loop counters and checks. 610 * The -mcpu=cortex-m55 option now supports the extensions +nomve and 611 +nomve.fp to control generation of MVE and MVE floating-point 612 instructions. 613 614 IA-32/x86-64 615 616 * New ISA extension support for Intel TSXLDTRK was added to GCC. 617 TSXLDTRK intrinsics are available via the -mtsxldtrk compiler 618 switch. 619 * New ISA extension support for Intel SERIALIZE was added to GCC. 620 SERIALIZE intrinsics are available via the -mserialize compiler 621 switch. 622 * New ISA extension support for Intel HRESET was added to GCC. HRESET 623 intrinsics are available via the -mhreset compiler switch. 624 * New ISA extension support for Intel UINTR was added to GCC. UINTR 625 intrinsics are available via the -muintr compiler switch. 626 * New ISA extension support for Intel KEYLOCKER was added to GCC. 627 KEYLOCKER intrinsics are available via the -mkeylocker compiler 628 switch. 629 * New ISA extension support for Intel AMX-TILE, AMX-INT8, AMX-BF16 630 was added to GCC. AMX-TILE, AMX-INT8, AMX-BF16 intrinsics are 631 available via the -mamx-tile, -mamx-int8, -mamx-bf16 compiler 632 switches. 633 * New ISA extension support for Intel AVX-VNNI was added to GCC. 634 AVX-VNNI intrinsics are available via the -mavxvnni compiler 635 switch. 636 * GCC now supports the Intel CPU named Sapphire Rapids through 637 -march=sapphirerapids. The switch enables the MOVDIRI, MOVDIR64B, 638 AVX512VP2INTERSECT, ENQCMD, CLDEMOTE, SERIALIZE, PTWRITE, WAITPKG, 639 TSXLDTRK, AMT-TILE, AMX-INT8, AMX-BF16, and AVX-VNNI ISA 640 extensions. 641 * GCC now supports the Intel CPU named Alderlake through 642 -march=alderlake. The switch enables the CLDEMOTE, PTWRITE, 643 WAITPKG, SERIALIZE, KEYLOCKER, AVX-VNNI, and HRESET ISA extensions. 644 * GCC now supports the Intel CPU named Rocketlake through 645 -march=rocketlake. Rocket Lake is based on Icelake client and minus 646 SGX. 647 * GCC now supports AMD CPUs based on the znver3 core via 648 -march=znver3. 649 * GCC now supports micro-architecture levels defined in the x86-64 650 psABI via -march=x86-64-v2, -march=x86-64-v3 and -march=x86-64-v4. 651 652 Nios II 653 654 * The options -mcustom-insn=N no longer produce compiler warnings if 655 the custom instruction is not generated due to missing optimization 656 options such as -fno-math-errno, -ffinite-math-only, or 657 -funsafe-math-optimizations. These warnings were not consistently 658 emitted for all custom instructions. 659 * The -mcustom-fpu-cfg=fph2 option has been added to enable the 660 custom instructions supported by the Nios II Floating Point 661 Hardware 2 Component. 662 663 NVPTX 664 665 * The -misa default has changed from sm_30 to sm_35. 666 * The -m32 compiler switch has been removed. 667 * The -msoft-stack-reserve-local format has been fixed. Previously, 668 it accepted -msoft-stack-reserve-local<n>. It now accepts 669 -msoft-stack-reserve-local=<n>. 670 671 S/390, System z, IBM Z Systems 672 673 * The behavior when compiling with -fexcess-precision=standard (e.g., 674 implied by -std=c99) on s390(x) targets can now be controlled at 675 configure time with the flag --enable-s390-excess-float-precision. 676 When enabled, GCC will maintain previous behavior and evaluate 677 float expressions in double precision, which aligns with the 678 definition of float_t as double. With the flag disabled, GCC will 679 always evaluate float expressions in single precision. In native 680 builds and cross compiles that have target libc headers, GCC will 681 by default match the definition of float_t in the installed glibc. 682 683 RISC-V 684 685 * Support address sanitizer for RISC-V. 686 * Support big-endian for RISC-V, thanks to Marcus Comstedt. 687 * Implement new style of architecture extension test macros: each 688 architecture extension has a corresponding feature test macro, 689 which can be used to test its existence and version information. 690 * Legacy architecture extension test macros like __riscv_atomic are 691 deprecated, but will still be supported for at least 2 release 692 cycles. 693 * Support IFUNC for riscv*-*-linux*. 694 * Add new option -misa-spec=* to control ISA spec version. This 695 controls the default version of each extensions. It defaults to 696 2.2. 697 * Introduce the --with-multilib-generator configure time option. This 698 allows for flexible config multi-lib settings. Its syntax is the 699 same as RISC-V's multilib-generator. 700 * Extend the sytax for multilib-generator, support expansion operator 701 * to reduce the complexity of complicated multi-lib re-use rules. 702 * Support -mcpu=* option aligned with RISC-V clang/LLVM. It sets the 703 pipeline model and architecture extensions, like -mtune=* plus 704 -march=*. 705 * Support for TLS stack protector canary access, thanks to Cooper Qu. 706 * Support __builtin_thread_pointer for RISC-V. 707 * Introduce shorten_memrefs optimization, which reduces the code size 708 for memory access, thanks to Craig Blackmore. 709 710Operating Systems 711 712 AIX 713 714 * GCC for AIX can be built as a 64 bit application and the runtime is 715 built as FAT libraries containing both 32 bit and 64 bit objects. 716 * Support AIX Vector Extended ABI with -mabi=vec-extabi. 717 * Thread-Local uninitiated data placed in local common section. 718 * Use thread-safe access in ctype. 719 * Link with libc128.a when long-double-128 enabled. 720 721Improvements to Static Analyzer 722 723 * The implementation of how program state is tracked within 724 [54]-fanalyzer has been completely rewritten for GCC 11, fixing 725 numerous bugs, and allowing for the analyzer to scale to larger C 726 source files. 727 * The analysis of allocations and deallocations has been generalized 728 beyond malloc and free. 729 + As preliminary work towards eventually supporting C++, the 730 malloc/free checking will also check new/delete and 731 new[]/delete[]. However, C++ is not yet properly supported by 732 [55]-fanalyzer (for example, exception-handling is 733 unimplemented). 734 + As noted above, the existing [56]malloc attribute has been 735 extended so that it can be used to identify 736 allocator/deallocator API pairs. The analyzer will use these 737 attributes when checking for leaks, double-frees, 738 use-after-frees, and similar issues. 739 + A new [57]-Wanalyzer-mismatching-deallocation warning has been 740 added, covering such mismatches as using scalar delete rather 741 vector delete[]. 742 * The analyzer has gained warnings 743 [58]-Wanalyzer-shift-count-negative, 744 [59]-Wanalyzer-shift-count-overflow, [60]-Wanalyzer-write-to-const, 745 and [61]-Wanalyzer-write-to-string-literal, all enabled by default 746 when [62]-fanalyzer is enabled. 747 * The analyzer can now be extended by GCC plugins, allowing for 748 domain-specific path-sensitive warnings. An example of using a 749 [63]GCC plugin to check for misuses of CPython's global interpreter 750 lock can be seen in the test suite 751 * The analyzer has gained new debugging options 752 [64]-fdump-analyzer-json and [65]-fno-analyzer-feasibility. 753 754Other significant improvements 755 756 * GCC has gained a new environment variable 757 [66]GCC_EXTRA_DIAGNOSTIC_OUTPUT which can be used by IDEs to 758 request machine-readable fix-it hints without needing to adjust 759 build flags. 760 761GCC 11.1 762 763 This is the [67]list of problem reports (PRs) from GCC's bug tracking 764 system that are known to be fixed in the 11.1 release. This list might 765 not be complete (that is, it is possible that some PRs that have been 766 fixed are not listed here). 767 768GCC 11.2 769 770 This is the [68]list of problem reports (PRs) from GCC's bug tracking 771 system that are known to be fixed in the 11.2 release. This list might 772 not be complete (that is, it is possible that some PRs that have been 773 fixed are not listed here). 774 775 776 For questions related to the use of GCC, please consult these web 777 pages and the [69]GCC manuals. If that fails, the 778 [70]gcc-help@gcc.gnu.org mailing list might help. Comments on these 779 web pages and the development of GCC are welcome on our developer 780 list at [71]gcc@gcc.gnu.org. All of [72]our lists have public 781 archives. 782 783 Copyright (C) [73]Free Software Foundation, Inc. Verbatim copying and 784 distribution of this entire article is permitted in any medium, 785 provided this notice is preserved. 786 787 These pages are [74]maintained by the GCC team. Last modified 788 2021-07-28[75]. 789 790References 791 792 1. http://gcc.gnu.org/gcc-11/porting_to.html 793 2. http://gcc.gnu.org/onlinedocs/index.html#current 794 3. https://wg21.link/p0522r0 795 4. https://gcc.gnu.org/pipermail/gcc-patches/2020-May/546494.html 796 5. https://github.com/google/sanitizers/wiki/ThreadSanitizerCppManual 797 6. https://www.kernel.org/doc/html/latest/dev-tools/kcsan.html 798 7. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-unit 799 8. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-origin 800 9. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 801 10. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Preprocessor-Options.html#index-ftabstop 802 11. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 803 12. https://github.com/google/sanitizers/wiki/AddressSanitizer 804 13. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html 805 14. http://dwarfstd.org/doc/DWARF5.pdf 806 15. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Code-Gen-Options.html#index-fno-bit-tests 807 16. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-modref 808 17. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-ocf 809 18. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fprofile-values 810 19. https://www.openmp.org/specifications/ 811 20. https://www.openacc.org/ 812 21. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-11 813 22. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/libgomp/Enabling-OpenACC.html 814 23. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-no_005fstack_005fprotector-function-attribute 815 24. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 816 25. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 817 26. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 818 27. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 819 28. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc 820 29. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wsizeof-array-div 821 30. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wstringop-overread 822 31. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wtsan 823 32. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object 824 33. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized 825 34. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wuninitialized 826 35. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Warray-parameter 827 36. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wvla-parameter 828 37. http://gcc.gnu.org/projects/cxx-status.html#cxx20 829 38. http://gcc.gnu.org/projects/cxx-status.html#cxx23 830 39. https://gcc.gnu.org/PR41437 831 40. https://gcc.gnu.org/PR95307 832 41. https://gcc.gnu.org/PR97518 833 42. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wctad-maybe-unsupported 834 43. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wrange-loop-construct 835 44. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-enum-conversion 836 45. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-float-conversion 837 46. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete 838 47. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wvexing-parse 839 48. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wnonnull 840 49. http://gcc.gnu.org/gcc-5/changes.html#jit 841 50. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/compatibility.html 842 51. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/expressions.html#c.gcc_jit_global_set_initializer 843 52. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/asm.html 844 53. https://developer.arm.com/support/arm-security-updates/speculative-processor-vulnerability/downloads/straight-line-speculation 845 54. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 846 55. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 847 56. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute 848 57. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-mismatching-deallocation 849 58. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-negative 850 59. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-overflow 851 60. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const 852 61. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal 853 62. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html 854 63. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=66dde7bc64b75d4a338266333c9c490b12d49825 855 64. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fdump-analyzer-json 856 65. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fno-analyzer-feasibility 857 66. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Environment-Variables.html#index-GCC_005fEXTRA_005fDIAGNOSTIC_005fOUTPUT 858 67. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.0 859 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.2 860 69. https://gcc.gnu.org/onlinedocs/ 861 70. mailto:gcc-help@gcc.gnu.org 862 71. mailto:gcc@gcc.gnu.org 863 72. https://gcc.gnu.org/lists.html 864 73. https://www.fsf.org/ 865 74. https://gcc.gnu.org/about.html 866 75. http://validator.w3.org/check/referer 867====================================================================== 868http://gcc.gnu.org/gcc-10/index.html 869 870 GCC 10 Release Series 871 872 April 8, 2021 873 874 The [1]GNU project and the GCC developers are pleased to announce the 875 release of GCC 10.3. 876 877 This release is a bug-fix release, containing fixes for regressions in 878 GCC 10.2 relative to previous releases of GCC. 879 880Release History 881 882 GCC 10.3 883 April 8, 2021 ([2]changes, [3]documentation) 884 885 GCC 10.2 886 July 23, 2020 ([4]changes, [5]documentation) 887 888 GCC 10.1 889 May 7, 2020 ([6]changes, [7]documentation) 890 891References and Acknowledgements 892 893 GCC used to stand for the GNU C Compiler, but since the compiler 894 supports several other languages aside from C, it now stands for the 895 GNU Compiler Collection. 896 897 A list of [8]successful builds is updated as new information becomes 898 available. 899 900 The GCC developers would like to thank the numerous people that have 901 contributed new features, improvements, bug fixes, and other changes as 902 well as test results to GCC. This [9]amazing group of volunteers is 903 what makes GCC successful. 904 905 For additional information about GCC please refer to the [10]GCC 906 project web site or contact the [11]GCC development mailing list. 907 908 To obtain GCC please use [12]our mirror sites or [13]our version 909 control system. 910 911 912 For questions related to the use of GCC, please consult these web 913 pages and the [14]GCC manuals. If that fails, the 914 [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these 915 web pages and the development of GCC are welcome on our developer 916 list at [16]gcc@gcc.gnu.org. All of [17]our lists have public 917 archives. 918 919 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 920 distribution of this entire article is permitted in any medium, 921 provided this notice is preserved. 922 923 These pages are [19]maintained by the GCC team. Last modified 924 2021-07-28[20]. 925 926References 927 928 1. http://www.gnu.org/ 929 2. http://gcc.gnu.org/gcc-10/changes.html 930 3. http://gcc.gnu.org/onlinedocs/10.3.0/ 931 4. http://gcc.gnu.org/gcc-10/changes.html 932 5. http://gcc.gnu.org/onlinedocs/10.2.0/ 933 6. http://gcc.gnu.org/gcc-10/changes.html 934 7. http://gcc.gnu.org/onlinedocs/10.1.0/ 935 8. http://gcc.gnu.org/gcc-10/buildstat.html 936 9. http://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Contributors.html 937 10. http://gcc.gnu.org/index.html 938 11. mailto:gcc@gcc.gnu.org 939 12. http://gcc.gnu.org/mirrors.html 940 13. http://gcc.gnu.org/git.html 941 14. https://gcc.gnu.org/onlinedocs/ 942 15. mailto:gcc-help@gcc.gnu.org 943 16. mailto:gcc@gcc.gnu.org 944 17. https://gcc.gnu.org/lists.html 945 18. https://www.fsf.org/ 946 19. https://gcc.gnu.org/about.html 947 20. http://validator.w3.org/check/referer 948====================================================================== 949http://gcc.gnu.org/gcc-10/changes.html 950 951 GCC 10 Release Series 952 Changes, New Features, and Fixes 953 954 This page is a "brief" summary of some of the huge number of 955 improvements in GCC 10. You may also want to check out our [1]Porting 956 to GCC 10 page and the [2]full GCC documentation. 957 958Caveats 959 960 * An ABI incompatibility between C++14 and C++17 has been fixed. On 961 some targets a class with a zero-sized subobject would be passed 962 incorrectly when compiled as C++17 or C++20. See the [3]C++ notes 963 below for more details. 964 * The deprecated Profile Mode and array_allocator extensions have 965 been removed from libstdc++. 966 * The non-standard std::__is_nullptr_t type trait is deprecated and 967 will be removed from libstdc++ in a future release. The standard 968 trait std::is_null_pointer should be instead. 969 * The minimum version of the [4]MPFR library required for building 970 GCC has been increased to version 3.1.0 (released 2011-10-03). 971 * The automatic template instantiation at link time (-frepo) has been 972 removed. 973 * The --param allow-store-data-races internal parameter has been 974 removed in favor of a new official option -fallow-store-data-races. 975 While default behavior is unchanged and the new option allows to 976 correctly maintain a per compilation unit setting across link-time 977 optimization, alteration of the default via --param 978 allow-store-data-races will now be diagnosed and build systems have 979 to be adjusted accordingly. 980 * Offloading to Heterogeneous System Architecture Intermediate 981 Language (HSAIL) has been deprecated and will likely be removed in 982 a future release. 983 * The type of the std::iterator base class of 984 std::istreambuf_iterator was changed in C++98 mode to be consistent 985 with C++11 and later standards. See the [5]libstdc++ notes below 986 for more details. 987 988General Improvements 989 990 * New built-in functions: 991 + The [6]__has_builtin built-in preprocessor operator can be 992 used to query support for built-in functions provided by GCC 993 and other compilers that support it. 994 + __builtin_roundeven for the corresponding function from 995 ISO/IEC TS 18661. 996 * New command-line options: 997 + [7]-fallocation-dce removes unneeded pairs of new and delete 998 operators. 999 + [8]-fprofile-partial-training can now be used to inform the 1000 compiler that code paths not covered by the training run 1001 should not be optimized for size. 1002 + [9]-fprofile-reproducible controls level of reproducibility of 1003 profile gathered by [10]-fprofile-generate. This makes it 1004 possible to rebuild program with same outcome which is useful, 1005 for example, for distribution packages. 1006 + [11]-fprofile-prefix-path can be used in combination with 1007 -fprofile-generate=profile_dir and -fprofile-use=profile_dir 1008 to inform GCC where the base directory of build source tree is 1009 in case it differs between instrumentation and optimized 1010 builds. 1011 + [12]-fanalyzer enables a new static analysis pass and 1012 associated warnings. This pass performs a time-consuming 1013 exploration of paths through the code in the hope of detecting 1014 various common errors, such as double-free bugs. This option 1015 should be regarded as experimental in this release. In 1016 particular, analysis of non-C code is unlikely to work. 1017 * Inter-procedural optimization improvements: 1018 + The inter-procedural scalar replacement of aggregates 1019 (IPA-SRA) pass was re-implemented to work at link-time and can 1020 now also remove computing and returning unused return values. 1021 + [13]-finline-functions is now enabled at -O2 and was retuned 1022 for better code size versus runtime performance trade-offs. 1023 Inliner heuristics was also significantly sped up to avoid 1024 negative impact to -flto -O2 compile times. 1025 + Inliner heuristics and function cloning can now use 1026 value-range information to predict effectivity of individual 1027 transformations. 1028 + During link-time optimization the C++ One Definition Rule is 1029 used to increase precision of type based alias analysis. 1030 * Link-time optimization improvements: 1031 + A new binary [14]lto-dump has been added. It dumps various 1032 information about LTO bytecode object files. 1033 + The parallel phase of the LTO can automatically detect a 1034 running make's jobserver or fall back to number of available 1035 cores. 1036 + The LTO bytecode can be compressed with the [15]zstd 1037 algorithm. The configure script automatically detects zstd 1038 support. 1039 + Most --param values can now be specified at translation unit 1040 granularity. This includes all parameters controlling the 1041 inliner and other inter-procedural optimizations. Unlike 1042 earlier releases, GCC 10 will ignore parameters controlling 1043 optimizations specified at link-time and apply parameters 1044 specified at compile-time in the same manner as done for 1045 optimization flags. 1046 * Profile driven optimization improvements: 1047 + Profile maintenance during compilation and hot/cold code 1048 partitioning have been improved. 1049 + Using [16]-fprofile-values, an instrumented binary can track 1050 multiple values (up to 4) for e.g. indirect calls and provide 1051 more precise profile information. 1052 1053New Languages and Language-Specific Improvements 1054 1055 * Version 2.6 of the [17]OpenACC specification is now supported in 1056 the C, C++ and Fortran compilers. See the [18]implementation status 1057 section on the OpenACC wiki page and the [19]run-time library 1058 documentation for further information. 1059 * GCC 10 adds a number of newly implemented [20]OpenMP 5.0 features 1060 on top of the GCC 9 release such as conditional lastprivate clause, 1061 scan and loop directives, order(concurrent) and use_device_addr 1062 clauses support, if clause on simd construct or partial support for 1063 the declare variant directive, getting closer to full support of 1064 the OpenMP 5.0 standard. 1065 * OpenMP and OpenACC now support [21]offloading to AMD Radeon (GCN) 1066 GPUs; supported are the third-generation Fiji (fiji) and the 1067 fifth-generation VEGA 10/VEGA 20 (gfx900 or gfx906). 1068 1069 C family 1070 1071 * New attributes: 1072 + The access function and type attribute has been added to 1073 describe how a function accesses objects passed to it by 1074 pointer or reference, and to associate such arguments with 1075 integer arguments denoting the objects' sizes. The attribute 1076 is used to enable the detection of invalid accesses by 1077 user-defined functions, such as those diagnosed by 1078 -Wstringop-overflow. 1079 + The symver attribute can be used to bind symbols to specific 1080 version nodes on ELF platforms. This is preferred to using 1081 inline assembly with GNU as symver directive because the 1082 latter is not compatible with link-time optimizations. 1083 * New warnings: 1084 + [22]-Wstring-compare, enabled by -Wextra, warns about equality 1085 and inequality expressions between zero and the result of a 1086 call to either strcmp and strncmp that evaluate to a constant 1087 as a result of the length of one argument being greater than 1088 the size of the array pointed to by the other. 1089 + [23]-Wzero-length-bounds, enabled by -Warray-bounds, warns 1090 about accesses to elements of zero-length arrays that might 1091 overlap other members of the same object. 1092 * Enhancements to existing warnings: 1093 + [24]-Warray-bounds detects more out-of-bounds accesses to 1094 member arrays as well as accesses to elements of zero-length 1095 arrays. 1096 + [25]-Wformat-overflow makes full use of string length 1097 information computed by the strlen optimization pass. 1098 + [26]-Wrestrict detects overlapping accesses to dynamically 1099 allocated objects. 1100 + [27]-Wreturn-local-addr diagnoses more instances of return 1101 statements returning addresses of automatic variables. 1102 + [28]-Wstringop-overflow detects more out-of-bounds stores to 1103 member arrays including zero-length arrays, dynamically 1104 allocated objects and variable length arrays, as well as more 1105 instances of reads of unterminated character arrays by string 1106 built-in functions. The warning also detects out-of-bounds 1107 accesses by calls to user-defined functions declared with the 1108 new attribute access. 1109 + [29]-Warith-conversion re-enables warnings from -Wconversion, 1110 -Wfloat-conversion, and -Wsign-conversion that are now off by 1111 default for an expression where the result of an arithmetic 1112 operation will not fit in the target type due to promotion, 1113 but the operands of the expression do fit in the target type. 1114 * Extended characters in identifiers may now be specified directly in 1115 the input encoding (UTF-8, by default), in addition to the UCN 1116 syntax (\uNNNN or \UNNNNNNNN) that is already supported: 1117 1118static const int p = 3; 1119int get_na�ve_pi() { 1120 return p; 1121} 1122 1123 C 1124 1125 * Several new features from the upcoming C2X revision of the ISO C 1126 standard are supported with -std=c2x and -std=gnu2x. Some of these 1127 features are also supported as extensions when compiling for older 1128 language versions. In addition to the features listed, some 1129 features previously supported as extensions and now added to the C 1130 standard are enabled by default in C2X mode and not diagnosed with 1131 -std=c2x -Wpedantic. 1132 + The [[]] attribute syntax is supported, as in C++. Existing 1133 attributes can be used with this syntax in forms such as 1134 [[gnu::const]]. The standard attributes [[deprecated]], 1135 [[fallthrough]] and [[maybe_unused]] are supported. 1136 + UTF-8 character constants using the u8'' syntax are supported. 1137 + <float.h> defines macros FLT_NORM_MAX, DBL_NORM_MAX and 1138 LDBL_NORM_MAX. 1139 + When decimal floating-point arithmetic is supported, <float.h> 1140 defines macros DEC32_TRUE_MIN, DEC64_TRUE_MIN and 1141 DEC128_TRUE_MIN, in addition to the macros that were 1142 previously only defined if __STDC_WANT_DEC_FP__ was defined 1143 before including <float.h>. 1144 + In C2X mode, empty parentheses in a function definition give 1145 that function a type with a prototype for subsequent calls; 1146 other old-style function definitions are diagnosed by default 1147 in C2X mode. 1148 + The strftime format checking supports the %OB and %Ob formats. 1149 + In C2X mode, -fno-fp-int-builtin-inexact is enabled by 1150 default. 1151 * GCC now defaults to -fno-common. As a result, global variable 1152 accesses are more efficient on various targets. In C, global 1153 variables with multiple tentative definitions now result in linker 1154 errors. With -fcommon such definitions are silently merged during 1155 linking. 1156 1157 C++ 1158 1159 * Several C++20 features have been implemented: 1160 + Concepts, including P0734R0, P0857R0, P1084R2, P1141R2, 1161 P0848R3, P1616R1, P1452R2 1162 + P1668R1, Permitting Unevaluated inline-assembly in constexpr 1163 Functions 1164 + P1161R3, Deprecate a[b,c] 1165 + P0848R3, Conditionally Trivial Special Member Functions 1166 + P1091R3, Extending structured bindings 1167 + P1143R2, Adding the constinit keyword 1168 + P1152R4, Deprecating volatile 1169 + P0388R4, Permit conversions to arrays of unknown bound 1170 + P0784R7, constexpr new 1171 + P1301R4, [[nodiscard("with reason")]] 1172 + P1814R0, class template argument deduction for alias templates 1173 + P1816R0, class template argument deduction for aggregates 1174 + P0960R3, Parenthesized initialization of aggregates 1175 + P1331R2, Allow trivial default initialization in constexpr 1176 contexts 1177 + P1327R1, Allowing dynamic_cast and polymorphic typeid in 1178 constexpr contexts 1179 + P0912R5, Coroutines (requires -fcoroutines) 1180 * Several C++ Defect Reports have been resolved, e.g.: 1181 + DR 1560, lvalue-to-rvalue conversion in ?: 1182 + DR 1813, __is_standard_layout for a class with repeated bases 1183 + DR 2094, volatile scalars are trivially copyable, 1184 + DR 2096, constraints on literal unions 1185 + DR 2413, typename in conversion-function-ids 1186 + DR 2352, Similar types and reference binding 1187 + DR 1601, Promotion of enumeration with fixed underlying type 1188 + DR 330, Qualification conversions and pointers to arrays of 1189 pointers 1190 + DR 1307, Overload resolution based on size of array 1191 initializer-list 1192 + DR 1710, Missing template keyword in class-or-decltype 1193 * New warnings: 1194 + [30]-Wmismatched-tags, disabled by default, warns about 1195 declarations of structs, classes, and class templates and 1196 their specializations with a class-key that does not match 1197 either the definition or the first declaration if no 1198 definition is provided. The option is provided to ease 1199 portability to Windows-based compilers. 1200 + [31]-Wredundant-tags, disabled by default, warns about 1201 redundant class-key and enum-key in contexts where the key can 1202 be eliminated without causing an syntactic ambiguity. 1203 * G++ can now detect modifying constant objects in constexpr 1204 evaluation (which is undefined behavior). 1205 * G++ no longer emits bogus -Wsign-conversion warnings with explicit 1206 casts. 1207 * Narrowing is now detected in more contexts (e.g., case values). 1208 * Memory consumption of the compiler has been reduced in constexpr 1209 evaluation. 1210 * The noexcept-specifier is now properly treated as a complete-class 1211 context as per [class.mem]. 1212 * The attribute deprecated can now be used on namespaces too. 1213 * The ABI of passing and returning certain C++ classes by value 1214 changed on several targets in GCC 10, including [32]AArch64, 1215 [33]ARM, [34]PowerPC ELFv2, [35]S/390 and [36]Itanium. These 1216 changes affect classes with a zero-sized subobject (an empty base 1217 class, or data member with the [[no_unique_address]] attribute) 1218 where all other non-static data members have the same type (this is 1219 called a "homogeneous aggregate" in some ABI specifications, or if 1220 there is only one such member, a "single element"). In -std=c++17 1221 and -std=c++20 modes, classes with an empty base class were not 1222 considered to have a single element or to be a homogeneous 1223 aggregate, and so could be passed differently (in the wrong 1224 registers or at the wrong stack address). This could make code 1225 compiled with -std=c++17 and -std=c++14 ABI incompatible. This has 1226 been corrected and the empty bases are ignored in those ABI 1227 decisions, so functions compiled with -std=c++14 and -std=c++17 are 1228 now ABI compatible again. Example: struct empty {}; struct S : 1229 empty { float f; }; void f(S);. Similarly, in classes containing 1230 non-static data members with empty class types using the C++20 1231 [[no_unique_address]] attribute, those members weren't ignored in 1232 the ABI argument passing decisions as they should be. Both of these 1233 ABI changes are now diagnosed with -Wpsabi. 1234 1235 Runtime Library (libstdc++) 1236 1237 * Improved experimental C++2a support, including: 1238 + Library concepts in <concepts> and <iterator>. 1239 + Constrained algorithms in <ranges>, <algorithm>, and <memory> 1240 (thanks to Patrick Palka). 1241 + New algorithms shift_left and shift_right (thanks to Patrick 1242 Palka). 1243 + std::span (thanks to JeanHeyd Meneide). 1244 + Three-way comparisons in <compare> and throughout the library. 1245 + Constexpr support in <algorithm> and elsewhere (thanks to 1246 Edward Smith-Rowland). 1247 + <stop_token> and std::jthread (thanks to Thomas Rodgers). 1248 + std::atomic_ref and std::atomic<floating point>. 1249 + Integer comparison functions (cmp_equal, cmp_less etc.). 1250 + std::ssize, std::to_array. 1251 + std::construct_at, std::destroy, constexpr std::allocator. 1252 + Mathematical constants in <numbers>. 1253 * Support for RDSEED in std::random_device. 1254 * Reduced header dependencies, leading to faster compilation for some 1255 code. 1256 * The std::iterator base class of std::istreambuf_iterator was 1257 changed in C++98 mode to be consistent with C++11 and later 1258 standards. This is expected to have no noticeable effect except in 1259 the unlikely case of a class which has potentially overlapping 1260 subobjects of type std::istreambuf_iterator<C> and another iterator 1261 type with a std::iterator<input_iterator_tag, C, ...> base class. 1262 The layout of such a type might change when compiled as C++98. 1263 [37]Bug 92285 has more details and concrete examples. 1264 1265 D 1266 1267 * Support for static foreach has been implemented. 1268 * Aliases can now be created directly from any __traits that return 1269 symbols or tuples. Previously, an AliasSeq was necessary in order 1270 to alias their return. 1271 * It is now possible to detect the language ABI specified for a 1272 struct, class, or interface using __traits(getLinkage, ...). 1273 * Support for core.math.toPrec intrinsics has been added. These 1274 intrinsics guarantee the rounding to specific floating-point 1275 precisions at specified points in the code. 1276 * Support for pragma(inline) has been implemented. Previously the 1277 pragma was recognized, but had no effect on the compilation. 1278 * Optional parentheses in asm operands are deprecated and will be 1279 removed in a future release. 1280 * All content imported files are now included in the make dependency 1281 list when compiling with -M. 1282 * Compiler recognized attributes provided by the gcc.attribute module 1283 will now take effect when applied to function prototypes as well as 1284 when applied to full function declarations. 1285 * Added a --enable-libphobos-checking configure option to control 1286 whether run-time checks are compiled into the D runtime library. 1287 * Added a --with-libphobos-druntime-only configure option to indicate 1288 whether to build only the core D runtime library, or both the core 1289 and standard libraries into libphobos. 1290 1291 Fortran 1292 1293 * use_device_addr of version 5.0 of the [38]OpenMP specification is 1294 now supported. Note that otherwise OpenMP 4.5 is partially 1295 supported in the Fortran compiler; the largest missing item is 1296 structure element mapping. 1297 * The default buffer size for I/O using unformatted files has been 1298 increased to 1048576. The buffer size for can now be set at runtime 1299 via the environment variables GFORTRAN_FORMATTED_BUFFER_SIZE and 1300 GFORTRAN_UNFORMATTED_BUFFER_SIZE for formatted and unformatted 1301 files, respectively. 1302 * Mismatches between actual and dummy argument lists in a single file 1303 are now rejected with an error. Use the new option 1304 -fallow-argument-mismatch to turn these errors into warnings; this 1305 option is implied with -std=legacy. -Wargument-mismatch has been 1306 removed. 1307 * The handling of a BOZ literal constant has been reworked to provide 1308 better conformance to the Fortran 2008 and 2018 standards. In these 1309 Fortran standards, a BOZ literal constant is a typeless and 1310 kindless entity. As a part of the rework, documented and 1311 undocumented extensions to the Fortran standard now emit errors 1312 during compilation. Some of these extensions are permitted with the 1313 -fallow-invalid-boz option, which degrades the error to a warning 1314 and the code is compiled as with older gfortran. 1315 * At any optimization level except-Os, gfortran now uses inline 1316 packing for arguments instead of calling a library routine. If the 1317 source contains a large number of arguments that need to be 1318 repacked, code size or time for compilation can become excessive. 1319 If that is the case, -fno-inline-arg-packing can be used to disable 1320 inline argument packing. 1321 * Legacy extensions: 1322 + For formatted input/output, if the explicit widths after the 1323 data-edit descriptors I, F and G have been omitted, default 1324 widths are used. 1325 + A blank format item at the end of a format specification, i.e. 1326 nothing following the final comma, is allowed. Use the option 1327 -fdec-blank-format-item; this option is implied with -fdec. 1328 + The existing support for AUTOMATIC and STATIC attributes has 1329 been extended to allow variables with the AUTOMATIC attribute 1330 to be used in EQUIVALENCE statements. Use -fdec-static; this 1331 option is implied by -fdec. 1332 + Allow character literals in assignments and DATA statements 1333 for numeric (INTEGER, REAL, or COMPLEX) or LOGICAL variables. 1334 Use the option -fdec-char-conversions; this option is implied 1335 with -fdec. 1336 + DEC comparisons, i.e. allow Hollerith constants to be used in 1337 comparisons with INTEGER, REAL, COMPLEX and CHARACTER 1338 expressions. Use the option -fdec. 1339 * Character type names in errors and warnings now include len in 1340 addition to kind; * is used for assumed length. The kind is omitted 1341 if it is the default kind. Examples: CHARACTER(12), CHARACTER(6,4). 1342 * CO_BROADCAST now supports derived type variables including objects 1343 with allocatable components. In this case, the optional arguments 1344 STAT= and ERRMSG= are currently ignored. 1345 * The handling of module and submodule names has been reworked to 1346 allow the full 63-character length mandated by the standard. 1347 Previously symbol names were truncated if the combined length of 1348 module, submodule, and function name exceeded 126 characters. This 1349 change therefore breaks the ABI, but only for cases where this 126 1350 character limit was exceeded. 1351 1352 Go 1353 1354 * GCC 10 provides a complete implementation of the Go 1.14.6 user 1355 packages. 1356 1357libgccjit 1358 1359 * The libgccjit API gained four new entry points: 1360 + [39]gcc_jit_version_major, [40]gcc_jit_version_minor, and 1361 [41]gcc_jit_version_patchlevel for programmatically checking 1362 the libgccjit version from client code, and 1363 + [42]gcc_jit_context_new_bitfield 1364 1365New Targets and Target Specific Improvements 1366 1367 AArch64 & arm 1368 1369 * The AArch64 and arm ports now support condition flag output 1370 constraints in inline assembly, as indicated by the 1371 __GCC_ASM_FLAG_OUTPUTS__. On arm this feature is only available for 1372 A32 and T32 targets. Please refer to the documentation for more 1373 details. 1374 1375 AArch64 1376 1377 * There have been several improvements related to the Scalable Vector 1378 Extension (SVE): 1379 + The SVE ACLE types and intrinsics are now supported. They can 1380 be accessed using the header file arm_sve.h. 1381 + It is now possible to create fixed-length SVE types using the 1382 arm_sve_vector_bits attribute. For example: 1383#if __ARM_FEATURE_SVE_BITS==512 1384typedef svint32_t vec512 __attribute__((arm_sve_vector_bits(512))); 1385typedef svbool_t pred512 __attribute__((arm_sve_vector_bits(512))); 1386#endif 1387 + -mlow-precision-div, -mlow-precision-sqrt and 1388 -mlow-precision-recip-sqrt now work for SVE. 1389 + -msve-vector-bits=128 now generates vector-length-specific 1390 code for little-endian targets. It continues to generate 1391 vector-length-agnostic code for big-endian targets, just as 1392 previous releases did for all targets. 1393 + The vectorizer is now able to use extending loads and 1394 truncating stores, including gather loads and scatter stores. 1395 + The vectorizer now compares the cost of vectorizing with SVE 1396 and vectorizing with Advanced SIMD and tries to pick the best 1397 one. Previously it would always use SVE if possible. 1398 + If a vector loop uses Advanced SIMD rather than SVE, the 1399 vectorizer now considers using SVE to vectorize the left-over 1400 elements (the "scalar tail" or "epilog"). 1401 + Besides these specific points, there have been many general 1402 improvements to the way that the vectorizer uses SVE. 1403 * The -mbranch-protection=pac-ret option now accepts the optional 1404 argument +b-key extension to perform return address signing with 1405 the B-key instead of the A-key. 1406 * The option -moutline-atomics has been added to aid deployment of 1407 the Large System Extensions (LSE) on GNU/Linux systems built with a 1408 baseline architecture targeting Armv8-A. When the option is 1409 specified code is emitted to detect the presence of LSE 1410 instructions at runtime and use them for standard atomic 1411 operations. For more information please refer to the documentation. 1412 * The Transactional Memory Extension is now supported through ACLE 1413 intrinsics. It can be enabled through the +tme option extension 1414 (for example, -march=armv8.5-a+tme). 1415 * A number of features from Armv8.5-A are now supported through ACLE 1416 intrinsics. These include: 1417 + The random number instructions that can be enabled through the 1418 (already present in GCC 9.1) +rng option extension. 1419 + Floating-point intrinsics to round to integer instructions 1420 from Armv8.5-A when targeting -march=armv8.5-a or later. 1421 + Memory Tagging Extension intrinsics enabled through the 1422 +memtag option extension. 1423 * Similarly, the following Armv8.6-A features are now supported 1424 through ACLE intrinsics: 1425 + The bfloat16 extension. This extension is enabled 1426 automatically when Armv8.6-A is selected (such as by 1427 -march=armv8.6-a). It can also be enabled for Armv8.2-A and 1428 later using the +bf16 option extension. 1429 + The Matrix Multiply extension. This extension is split into 1430 three parts, one for each supported data type: 1431 o Support for 8-bit integer matrix multiply instructions. 1432 This extension is enabled automatically when Armv8.6-A is 1433 selected. It can also be enabled for Armv8.2-A and later 1434 using the +i8mm option extension. 1435 o Support for 32-bit floating-point matrix multiply 1436 instructions. This extension can be enabled using the 1437 +f32mm option extension, which also has the effect of 1438 enabling SVE. 1439 o Support for 64-bit floating-point matrix multiply 1440 instructions. This extension can be enabled using the 1441 +f64mm option extension, which likewise has the effect of 1442 enabling SVE. 1443 * SVE2 is now supported through ACLE intrinsics and (to a limited 1444 extent) through autovectorization. It can be enabled through the 1445 +sve2 option extension (for example, -march=armv8.5-a+sve2). 1446 Additional extensions can be enabled through +sve2-sm4, +sve2-aes, 1447 +sve2-sha3 and +sve2-bitperm. 1448 * Support has been added for the following processors (GCC 1449 identifiers in parentheses): 1450 + Arm Cortex-A77 (cortex-a77). 1451 + Arm Cortex-A76AE (cortex-a76ae). 1452 + Arm Cortex-A65 (cortex-a65). 1453 + Arm Cortex-A65AE (cortex-a65ae). 1454 + Arm Cortex-A34 (cortex-a34). 1455 + Marvell ThunderX3 (thunderx3t110). 1456 The GCC identifiers can be used as arguments to the -mcpu or -mtune 1457 options, for example: -mcpu=cortex-a77 or -mtune=cortex-a65ae or as 1458 arguments to the equivalent target attributes and pragmas. 1459 1460 arm 1461 1462 * Support for the FDPIC ABI has been added. It uses 64-bit function 1463 descriptors to represent pointers to functions, and enables code 1464 sharing on MMU-less systems. The corresponding target triple is 1465 arm-uclinuxfdpiceabi, and the C library is uclibc-ng. 1466 * Support has been added for the Arm EABI on NetBSD through the 1467 arm*-*-netbsdelf-*eabi* triplet. 1468 * The handling of 64-bit integer operations has been significantly 1469 reworked and improved leading to improved performance and reduced 1470 stack usage when using 64-bit integral data types. The option 1471 -mneon-for-64bits is now deprecated and will be removed in a future 1472 release. 1473 * Support has been added for the following processors (GCC 1474 identifiers in parentheses): 1475 + Arm Cortex-A77 (cortex-a77). 1476 + Arm Cortex-A76AE (cortex-a76ae). 1477 + Arm Cortex-M35P (cortex-m35p). 1478 + Arm Cortex-M55 (cortex-m55). 1479 The GCC identifiers can be used as arguments to the -mcpu or -mtune 1480 options, for example: -mcpu=cortex-a77 or -mtune=cortex-m35p. 1481 * Support has been extended for the ACLE [43]data-processing 1482 intrinsics to include 32-bit SIMD, saturating arithmetic, 16-bit 1483 multiplication and other related intrinsics aimed at DSP algorithm 1484 optimization. 1485 * Support for -mpure-code in Thumb-1 (v6m) has been added: this 1486 M-profile feature is no longer restricted to targets with MOVT. For 1487 example, -mcpu=cortex-m0 now supports this option. 1488 * Support for the [44]Armv8.1-M Mainline Architecture has been added. 1489 + Armv8.1-M Mainline can be enabled by using the 1490 -march=armv8.1-m.main command-line option. 1491 * Support for the [45]MVE beta ACLE intrinsics has been added. These 1492 intrinsics can be enabled by including the arm_mve.h header file 1493 and passing the +mve or +mve.fp option extensions (for example: 1494 -march=armv8.1-m.main+mve). 1495 * Support for the Custom Datapath Extension beta ACLE [46]intrinsics 1496 has been added. 1497 * Support for Armv8.1-M Mainline Security Extensions architecture has 1498 been added. The -mcmse option, when used in combination with an 1499 Armv8.1-M Mainline architecture (for example: -march=armv8.1-m.main 1500 -mcmse), now leads to the generation of improved code sequences 1501 when changing security states. 1502 1503 AMD Radeon (GCN) 1504 1505 * The code generation and in particular the vectorization support has 1506 been much improved. 1507 1508 ARC 1509 1510 * The interrupt service routine functions save all used registers, 1511 including extension registers and auxiliary registers used by Zero 1512 Overhead Loops. 1513 * Improve code size by using multiple short instructions instead of a 1514 single long mov or ior instruction when its long immediate constant 1515 is known. 1516 * Fix usage of the accumulator register for ARC600. 1517 * Fix issues with uncached attribute. 1518 * Remove -mq-class option. 1519 * Improve 64-bit integer addition and subtraction operations. 1520 1521 AVR 1522 1523 * Support for the XMEGA-like devices 1524 1525 ATtiny202, ATtiny204, ATtiny402, ATtiny404, ATtiny406, ATtiny804, 1526 ATtiny806, ATtiny807, ATtiny1604, ATtiny1606, ATtiny1607, ATmega808, 1527 ATmega809, ATmega1608, ATmega1609, ATmega3208, ATmega3209, 1528 ATmega4808, ATmega4809 1529 has been added. 1530 * A new command-line option -nodevicespecs has been added. It allows 1531 to provide a custom device-specs file by means of 1532 1533 avr-gcc -nodevicespecs -specs=my-spec-file <options> 1534 and without the need to provide options -B and -mmcu=. See [47]AVR 1535 command-line options for details. This feature is also available in 1536 GCC 9.3+ and GCC 8.4+. 1537 * New command-line options -mdouble=[32,64] and -mlong-double=[32,64] 1538 have been added. They allow to choose the size (in bits) of the 1539 double and long double types, respectively. Whether or not the 1540 mentioned layouts are available, whether the options act as a 1541 multilib option, and the default for either option are controlled 1542 by the new [48]AVR configure options --with-double= and 1543 --with-long-double=. 1544 * A new configure option --with-libf7= has been added. It controls to 1545 which level avr-libgcc provides 64-bit floating point support by 1546 means of [49]Libf7. 1547 * A new configure option --with-double-comparison= has been added. 1548 It's unlikely you need to set this option by hand. 1549 1550 IA-32/x86-64 1551 1552 * Support to expand __builtin_roundeven into the appropriate SSE 4.1 1553 instruction has been added. 1554 * New ISA extension support for Intel ENQCMD was added to GCC. ENQCMD 1555 intrinsics are available via the -menqcmd compiler switch. 1556 * GCC now supports the Intel CPU named Cooperlake through 1557 -march=cooperlake. The switch enables the AVX512BF16 ISA 1558 extensions. 1559 * GCC now supports the Intel CPU named Tigerlake through 1560 -march=tigerlake. The switch enables the MOVDIRI MOVDIR64B 1561 AVX512VP2INTERSECT ISA extensions. 1562 1563 MIPS 1564 1565 * The mips*-*-linux* targets now mark object files with appropriate 1566 GNU-stack note, facilitating use of non-executable stack hardening 1567 on GNU/Linux. The soft-float targets have this feature enabled by 1568 default, while for hard-float targets it is required for GCC to be 1569 configured with --with-glibc-version=2.31 against glibc 2.31 or 1570 later. 1571 1572 PowerPC / PowerPC64 / RS6000 1573 1574 * Many vector builtins have been listed as deprecated in the 1575 [50]64-Bit ELF V2 ABI Specification for quite a number of years. 1576 The vector builtins listed in Tables A.8 through A.10 are now 1577 deprecated for GCC 10, and will likely be removed from support in 1578 GCC 11. Note that this does not result in any loss of function. 1579 These deprecated builtins generally provide somewhat nonsensical 1580 argument lists (for example, mixing signed, unsigned, and bool 1581 vector arguments arbitrarily), or are duplicate builtins that are 1582 inconsistent with the expected naming scheme. We expect that this 1583 will be unlikely to affect much if any code, and any required code 1584 changes will be trivial. 1585 1586 PRU 1587 1588 * A new back end targeting TI PRU I/O processors has been contributed 1589 to GCC. 1590 1591 RISC-V 1592 1593 * The riscv*-*-* targets now require GNU binutils version 2.30 or 1594 later, to support new assembly instructions produced by GCC. 1595 1596 V850 1597 1598 * The ABI for V850 nested functions has been changed. Previously the 1599 V850 port used %r20 for the static chain pointer, now the port uses 1600 %r19. This corrects a long standing latent bug in the v850 port 1601 where a call to a nested function would unexpectedly change the 1602 value in %r20. 1603 1604Operating Systems 1605 1606Improvements for plugin authors 1607 1608 * GCC diagnostics can now have a chain of events associated with 1609 them, describing a path through the code that triggers the problem. 1610 These can be printed by the diagnostics subsystem in various ways, 1611 controlled by the [51]-fdiagnostics-path-format option, or captured 1612 in JSON form via [52]-fdiagnostics-format=json. 1613 * GCC diagnostics can now be associated with [53]CWE weakness 1614 identifiers, which will appear on the standard error stream, and in 1615 the JSON output from [54]-fdiagnostics-format=json. 1616 1617Other significant improvements 1618 1619 * To allow inline expansion of both memcpy and memmove, the existing 1620 movmem instruction patterns used for non-overlapping memory copies 1621 have been renamed to cpymem. The movmem name is now used for 1622 overlapping memory moves, consistent with the library functions 1623 memcpy and memmove. 1624 * For many releases, when GCC emits a warning it prints the option 1625 controlling that warning. As of GCC 10, that option text is now a 1626 clickable hyperlink for the documentation of that option (assuming 1627 a [55]sufficiently capable terminal). This behavior can be 1628 controlled via a new [56]-fdiagnostics-urls option (along with 1629 various environment variables and heuristics documented with that 1630 option). 1631 1632GCC 10.1 1633 1634 This is the [57]list of problem reports (PRs) from GCC's bug tracking 1635 system that are known to be fixed in the 10.1 release. This list might 1636 not be complete (that is, it is possible that some PRs that have been 1637 fixed are not listed here). 1638 1639GCC 10.2 1640 1641 This is the [58]list of problem reports (PRs) from GCC's bug tracking 1642 system that are known to be fixed in the 10.2 release. This list might 1643 not be complete (that is, it is possible that some PRs that have been 1644 fixed are not listed here). 1645 1646GCC 10.3 1647 1648 This is the [59]list of problem reports (PRs) from GCC's bug tracking 1649 system that are known to be fixed in the 10.3 release. This list might 1650 not be complete (that is, it is possible that some PRs that have been 1651 fixed are not listed here). 1652 1653 Target Specific Changes 1654 1655 AArch64 1656 1657 * A bug with the Random Number intrinsics in the arm_acle.h header 1658 that resulted in an incorrect status result being returned has been 1659 fixed. 1660 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 1661 options are -mcpu=a64fx and -mtune=a64fx respectively. In 1662 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 1663 tunes the code for the A64FX. This includes tuning the SVE code, 1664 although by default the code is still length-agnostic and so works 1665 for all SVE implementations. Adding -msve-vector-bits=512 makes the 1666 code specific to 512-bit SVE. 1667 1668 x86-64 1669 1670 * GCC 10.3 supports AMD CPUs based on the znver3 core via 1671 -march=znver3. 1672 1673 1674 For questions related to the use of GCC, please consult these web 1675 pages and the [60]GCC manuals. If that fails, the 1676 [61]gcc-help@gcc.gnu.org mailing list might help. Comments on these 1677 web pages and the development of GCC are welcome on our developer 1678 list at [62]gcc@gcc.gnu.org. All of [63]our lists have public 1679 archives. 1680 1681 Copyright (C) [64]Free Software Foundation, Inc. Verbatim copying and 1682 distribution of this entire article is permitted in any medium, 1683 provided this notice is preserved. 1684 1685 These pages are [65]maintained by the GCC team. Last modified 1686 2021-07-28[66]. 1687 1688References 1689 1690 1. http://gcc.gnu.org/gcc-10/porting_to.html 1691 2. http://gcc.gnu.org/onlinedocs/index.html#current 1692 3. http://gcc.gnu.org/gcc-10/changes.html#empty_base 1693 4. https://www.mpfr.org/ 1694 5. http://gcc.gnu.org/gcc-10/changes.html#iterator_base 1695 6. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/cpp/_005f_005fhas_005fbuiltin.html#g_t_005f_005fhas_005fbuiltin 1696 7. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fno-allocation-dce 1697 8. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-partial-training 1698 9. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-reproducible 1699 10. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-generate 1700 11. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-prefix-path 1701 12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Static-Analyzer-Options.html 1702 13. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-finline-functions 1703 14. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/lto-dump.html 1704 15. https://facebook.github.io/zstd/ 1705 16. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-values 1706 17. https://www.openacc.org/ 1707 18. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-10 1708 19. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/libgomp/#toc-Enabling-OpenACC-1 1709 20. https://www.openmp.org/specifications/ 1710 21. https://gcc.gnu.org/wiki/Offloading 1711 22. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstring-compare 1712 23. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wzero-length-bounds 1713 24. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warray-bounds 1714 25. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wformat-overflow 1715 26. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wrestrict 1716 27. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wreturn-local-addr 1717 28. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstringop-overflow 1718 29. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warith-conversion 1719 30. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-tags 1720 31. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-tags 1721 32. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94383 1722 33. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94711 1723 34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94707 1724 35. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94704 1725 36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94706 1726 37. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92285 1727 38. https://www.openmp.org/specifications/ 1728 39. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_major 1729 40. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_minor 1730 41. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_patchlevel 1731 42. https://gcc.gnu.org/onlinedocs/jit/topics/types.html#c.gcc_jit_context_new_bitfield 1732 43. https://developer.arm.com/documentation/101028/0009/Data-processing-intrinsics 1733 44. https://developer.arm.com/architectures/cpu-architecture/m-profile 1734 45. https://developer.arm.com/architectures/instruction-sets/simd-isas/helium/helium-intrinsics 1735 46. https://developer.arm.com/documentation/101028/0010/Custom-Datapath-Extension 1736 47. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/AVR-Options.html#index-nodevicespecs 1737 48. https://gcc.gnu.org/install/configure.html#avr 1738 49. https://gcc.gnu.org/wiki/avr-gcc#Libf7 1739 50. https://openpowerfoundation.org/?resource_lib=64-bit-elf-v2-abi-specification-power-architecture 1740 51. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-path-format 1741 52. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 1742 53. https://cwe.mitre.org/ 1743 54. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 1744 55. https://gist.github.com/egmontkob/eb114294efbcd5adb1944c9f3cb5feda 1745 56. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-urls 1746 57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.0 1747 58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.2 1748 59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.3 1749 60. https://gcc.gnu.org/onlinedocs/ 1750 61. mailto:gcc-help@gcc.gnu.org 1751 62. mailto:gcc@gcc.gnu.org 1752 63. https://gcc.gnu.org/lists.html 1753 64. https://www.fsf.org/ 1754 65. https://gcc.gnu.org/about.html 1755 66. http://validator.w3.org/check/referer 1756====================================================================== 1757http://gcc.gnu.org/gcc-9/index.html 1758 1759 GCC 9 Release Series 1760 1761 June 1, 2021 1762 1763 The [1]GNU project and the GCC developers are pleased to announce the 1764 release of GCC 9.4. 1765 1766 This release is a bug-fix release, containing fixes for regressions in 1767 GCC 9.3 relative to previous releases of GCC. 1768 1769Release History 1770 1771 GCC 9.4 1772 June 1, 2021 ([2]changes, [3]documentation) 1773 1774 GCC 9.3 1775 Mar 12, 2020 ([4]changes, [5]documentation) 1776 1777 GCC 9.2 1778 Aug 12, 2019 ([6]changes, [7]documentation) 1779 1780 GCC 9.1 1781 May 3, 2019 ([8]changes, [9]documentation) 1782 1783References and Acknowledgements 1784 1785 GCC used to stand for the GNU C Compiler, but since the compiler 1786 supports several other languages aside from C, it now stands for the 1787 GNU Compiler Collection. 1788 1789 A list of [10]successful builds is updated as new information becomes 1790 available. 1791 1792 The GCC developers would like to thank the numerous people that have 1793 contributed new features, improvements, bug fixes, and other changes as 1794 well as test results to GCC. This [11]amazing group of volunteers is 1795 what makes GCC successful. 1796 1797 For additional information about GCC please refer to the [12]GCC 1798 project web site or contact the [13]GCC development mailing list. 1799 1800 To obtain GCC please use [14]our mirror sites or [15]our version 1801 control system. 1802 1803 1804 For questions related to the use of GCC, please consult these web 1805 pages and the [16]GCC manuals. If that fails, the 1806 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these 1807 web pages and the development of GCC are welcome on our developer 1808 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public 1809 archives. 1810 1811 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 1812 distribution of this entire article is permitted in any medium, 1813 provided this notice is preserved. 1814 1815 These pages are [21]maintained by the GCC team. Last modified 1816 2021-07-28[22]. 1817 1818References 1819 1820 1. http://www.gnu.org/ 1821 2. http://gcc.gnu.org/gcc-9/changes.html 1822 3. http://gcc.gnu.org/onlinedocs/9.4.0/ 1823 4. http://gcc.gnu.org/gcc-9/changes.html 1824 5. http://gcc.gnu.org/onlinedocs/9.3.0/ 1825 6. http://gcc.gnu.org/gcc-9/changes.html 1826 7. http://gcc.gnu.org/onlinedocs/9.2.0/ 1827 8. http://gcc.gnu.org/gcc-9/changes.html 1828 9. http://gcc.gnu.org/onlinedocs/9.1.0/ 1829 10. http://gcc.gnu.org/gcc-9/buildstat.html 1830 11. http://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Contributors.html 1831 12. http://gcc.gnu.org/index.html 1832 13. mailto:gcc@gcc.gnu.org 1833 14. http://gcc.gnu.org/mirrors.html 1834 15. http://gcc.gnu.org/git.html 1835 16. https://gcc.gnu.org/onlinedocs/ 1836 17. mailto:gcc-help@gcc.gnu.org 1837 18. mailto:gcc@gcc.gnu.org 1838 19. https://gcc.gnu.org/lists.html 1839 20. https://www.fsf.org/ 1840 21. https://gcc.gnu.org/about.html 1841 22. http://validator.w3.org/check/referer 1842====================================================================== 1843http://gcc.gnu.org/gcc-9/changes.html 1844 1845 GCC 9 Release Series 1846 Changes, New Features, and Fixes 1847 1848 This page is a "brief" summary of some of the huge number of 1849 improvements in GCC 9. You may also want to check out our [1]Porting to 1850 GCC 9 page and the [2]full GCC documentation. 1851 1852Caveats 1853 1854 * On Arm targets (arm*-*-*), [3]a bug in the implementation of the 1855 procedure call standard (AAPCS) in the GCC 6, 7 and 8 releases has 1856 been fixed: a structure containing a bit-field based on a 64-bit 1857 integral type and where no other element in a structure required 1858 64-bit alignment could be passed incorrectly to functions. This is 1859 an ABI change. If the option -Wpsabi is enabled (on by default) the 1860 compiler will emit a diagnostic note for code that might be 1861 affected. 1862 * Support for a number of older systems and recently unmaintained or 1863 untested target ports of GCC has been declared obsolete in GCC 9. 1864 Unless there is activity to revive them, the next release of GCC 1865 will have their sources permanently removed. 1866 The following ports for individual systems on particular 1867 architectures have been obsoleted: 1868 + Solaris 10 (*-*-solaris2.10). Details can be found in the 1869 [4]announcement. 1870 + Cell Broadband Engine SPU (spu*-*-*). Details can be found in 1871 the [5]announcement. 1872 * A change to the C++ std::rotate algorithm in GCC 9.1.0 can cause 1873 ABI incompatibilities with object files compiled with other 1874 versions of GCC. If the std::rotate algorithm is called with an 1875 empty range then it might cause a divide-by-zero error (as a SIGFPE 1876 signal) and crash. The change has been reverted for GCC 9.2.0 and 1877 future releases. For more details see [6]Bug 90920. The problem can 1878 be avoided by recompiling any objects that might call std::rotate 1879 with an empty range, so that the GCC 9.1.0 definition of 1880 std::rotate is not used. 1881 * The automatic template instantiation at link time ([7]-frepo) has 1882 been deprecated and will be removed in a future release. 1883 * The --with-default-libstdcxx-abi=gcc4-compatible configure option 1884 is broken in the 9.1 and 9.2 releases, producing a shared library 1885 with missing symbols (see [8]Bug 90361). As a workaround, configure 1886 without that option and build GCC as normal, then edit the 1887 installed <bits/c++config.h> headers to define the 1888 _GLIBCXX_USE_CXX11_ABI macro to 0. 1889 1890General Improvements 1891 1892 The following GCC command line options have been introduced or 1893 improved. 1894 * All command line options that take a byte-size argument accept 1895 64-bit integers as well as standard SI and IEC suffixes such as kb 1896 and KiB, MB and MiB, or GB and GiB denoting the corresponding 1897 multiples of bytes. See [9]Invoking GCC for more. 1898 * A new option, 1899 [10]-flive-patching=[inline-only-static|inline-clone], has been 1900 introduced to provide a safe compilation for live-patching. At the 1901 same time, provides multiple-level control on the enabled IPA 1902 optimizations. See the user guide for more details about the 1903 option. 1904 * A new option, --completion, has been added to provide more fine 1905 option completion in a shell. It is intended to be used by 1906 Bash-completion. 1907 * GCC's diagnostics now print source code with a left margin showing 1908 line numbers, configurable with 1909 [11]-fno-diagnostics-show-line-numbers. 1910 GCC's diagnostics can also now label regions of the source code to 1911 show pertinent information, such as the types within an expression. 1912$ g++ t.cc 1913t.cc: In function 'int test(const shape&, const shape&)': 1914t.cc:15:4: error: no match for 'operator+' (operand types are 'boxed_value<doubl 1915e>' and 'boxed_value<double>') 1916 14 | return (width(s1) * height(s1) 1917 | ~~~~~~~~~~~~~~~~~~~~~~ 1918 | | 1919 | boxed_value<[...]> 1920 15 | + width(s2) * height(s2)); 1921 | ^ ~~~~~~~~~~~~~~~~~~~~~~ 1922 | | 1923 | boxed_value<[...]> 1924 1925 These labels can be disabled via [12]-fno-diagnostics-show-labels. 1926 * A new option [13]-fdiagnostics-format=json has been introduced for 1927 emitting diagnostics in a machine-readable format. 1928 * The alignment-related options [14]-falign-functions, 1929 [15]-falign-labels, [16]-falign-loops, and [17]-falign-jumps 1930 received support for a secondary alignment (e.g. 1931 -falign-loops=n:m:n2:m2). 1932 * New pair of profiling options ([18]-fprofile-filter-files and 1933 [19]-fprofile-exclude-files) has been added. The options help to 1934 filter which source files are instrumented. 1935 * AddressSanitizer generates more compact redzones for automatic 1936 variables. That helps to reduce memory footprint of a sanitized 1937 binary. 1938 * Numerous improvements have been made to the output of 1939 [20]-fopt-info. 1940 Messages are now prefixed with optimized, missed, or note, rather 1941 than the old behavior of all being prefixed with note. 1942 The output from -fopt-info can now contain information on inlining 1943 decisions: 1944$ g++ -c inline.cc -O2 -fopt-info-inline-all 1945inline.cc:24:11: note: Considering inline candidate void foreach(T, T, void (*)( 1946E)) [with T = char**; E = char*]/2. 1947inline.cc:24:11: optimized: Inlining void foreach(T, T, void (*)(E)) [with T = 1948char**; E = char*]/2 into int main(int, char**)/1. 1949inline.cc:19:12: missed: not inlinable: void inline_me(char*)/0 -> int std::pu 1950ts(const char*)/3, function body not available 1951inline.cc:13:8: optimized: Inlined void inline_me(char*)/4 into int main(int, c 1952har**)/1 which now has time 127.363637 and size 11, net change of +0. 1953Unit growth for small function inlining: 16->16 (0%) 1954 1955Inlined 2 calls, eliminated 1 functions 1956 1957 1958 The output from the vectorizer has been rationalized so that failed 1959 attempts to vectorize a loop are displayed in the form 1960 [LOOP-LOCATION]: couldn't vectorize this loop 1961 [PROBLEM-LOCATION]: because of [REASON] 1962 1963 rather than an exhaustive log of all decisions made by the 1964 vectorizer. For example: 1965$ gcc -c v.c -O3 -fopt-info-all-vec 1966v.c:7:3: missed: couldn't vectorize loop 1967v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 1968mory"); 1969v.c:3:6: note: vectorized 0 loops in function. 1970v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" : : : "me 1971mory"); 1972 1973 The old behavior can be obtained via a new -internals suboption of 1974 -fopt-info. 1975 * A new option, [21]-fsave-optimization-record has been added, which 1976 writes a SRCFILE.opt-record.json.gz file describing the 1977 optimization decisions made by GCC. This is similar to the output 1978 of -fopt-info, but with additional metadata such as the inlining 1979 chain, and profile information (if available). 1980 * Inter-procedural propagation of stack alignment can now be 1981 controlled by [22]-fipa-stack-alignment. 1982 * Propagation of addressability, readonly and writeonly flags on 1983 static variables can now be controlled by 1984 [23]-fipa-reference-addressable. 1985 1986 The following built-in functions have been introduced. 1987 * [24]__builtin_expect_with_probability to provide branch prediction 1988 probability hints to the optimizer. 1989 * [25]__builtin_has_attribute determines whether a function, type, or 1990 variable has been declared with some attribute. 1991 * [26]__builtin_speculation_safe_value can be used to help mitigate 1992 against unsafe speculative execution. 1993 1994 The following attributes have been introduced. 1995 * The [27]copy function attribute has been added. The attribute can 1996 also be applied to type definitions and to variable declarations. 1997 1998 A large number of improvements to code generation have been made, 1999 including but not limited to the following. 2000 * Switch expansion has been improved by using a different strategy 2001 (jump table, bit test, decision tree) for a subset of switch cases. 2002 * A linear function expression defined as a switch statement can be 2003 transformed by [28]-ftree-switch-conversion. For example: 2004 2005int 2006foo (int how) 2007{ 2008 switch (how) { 2009 case 2: how = 205; break; 2010 case 3: how = 305; break; 2011 case 4: how = 405; break; 2012 case 5: how = 505; break; 2013 case 6: how = 605; break; 2014 } 2015 return how; 2016} 2017 2018 can be transformed into 100 * how + 5 (for values defined in the 2019 switch statement). 2020 * Inter-procedural optimization improvements: 2021 + Inliner defaults was tuned to better suits modern C++ 2022 codebases especially when built with link time optimizations. 2023 New parameters max-inline-insns-small, max-inline-insns-size, 2024 uninlined-function-insns, uninlined-function-time, 2025 uninlined-thunk-insns, and uninlined-thunk-time were added. 2026 + Hot/cold partitioning is now more precise and aggressive. 2027 + Improved scalability for very large translation units 2028 (especially when link-time optimizing large programs). 2029 * Profile driven optimization improvements: 2030 + [29]-fprofile-use now enables [30]-fversion-loops-for-strides, 2031 [31]-floop-interchange, [32]-floop-unroll-and-jam, 2032 [33]-ftree-loop-distribution. 2033 + Streaming of counter histograms was removed. This reduces the 2034 size of profile files. Histogram is computed on the fly with 2035 link-time optimization. Parameter hot-bb-count-ws-permille was 2036 reduced from 999 to 990 to account for more precise 2037 histograms. 2038 * Link-time optimization improvements: 2039 + Types are now simplified prior streaming resulting in 2040 significant reductions of the LTO object files, link-time 2041 memory use, and improvements of link-time parallelism. 2042 + Default number of partitions (--param lto-partitions) was 2043 increased from 32 to 128 enabling effective use of CPUs with 2044 more than 32 hyperthreads. --param 2045 lto-max-streaming-parallelism can now be used to control 2046 number of streaming processes. 2047 + Warnings on C++ One Decl Rule violations (-Wodr) are now more 2048 informative and produce fewer redundant results. 2049 Overall compile time of Firefox 66 and LibreOffice 6.2.3 on an 2050 8-core machine was reduced by about 5% compared to GCC 8.3, and the 2051 size of LTO object files by 7%. LTO link-time improves by 11% on an 2052 8-core machine and scales significantly better for more parallel 2053 build environments. The serial stage of the link-time optimization 2054 is 28% faster consuming 20% less memory. The parallel stage now 2055 scales to up to 128 partitions rather than 32 and reduces memory 2056 use for every worker by 30%. 2057 2058 The following improvements to the gcov command-line utility have been 2059 made. 2060 * The gcov tool received a new option [34]--use-hotness-colors (-q) 2061 that can provide perf-like coloring of hot functions. 2062 * The gcov tool has changed its intermediate format to a new JSON 2063 format. 2064 2065New Languages and Language specific improvements 2066 2067 [35]OpenACC support in C, C++, and Fortran continues to be maintained 2068 and improved. Most of the OpenACC 2.5 specification is implemented. See 2069 the [36]implementation status section on the OpenACC wiki page for 2070 further information. 2071 2072 C family 2073 2074 * Version 5.0 of the [37]OpenMP specification is now partially 2075 supported in the C and C++ compilers. For details which features of 2076 OpenMP 5.0 are and which are not supported in the GCC 9 release see 2077 [38]this mail. 2078 * New extensions: 2079 + [39]__builtin_convertvector built-in for vector conversions 2080 has been added. 2081 * New warnings: 2082 + [40]-Waddress-of-packed-member, enabled by default, warns 2083 about an unaligned pointer value from the address of a packed 2084 member of a struct or union. 2085 * Enhancements to existing warnings: 2086 + [41]-Warray-bounds detects more instances of out-of-bounds 2087 indices. 2088 + [42]-Wattribute-alias also detects attribute mismatches 2089 between alias declarations and their targets, in addition to 2090 mismatches between their types. 2091 + [43]-Wformat-overflow and [44]-Wformat-truncation have been 2092 extended to all formatted input/output functions (where 2093 applicable) and enhanced to detect a subset of instances of 2094 reading past the end of unterminated constant character arrays 2095 in %s directives. 2096 + [45]-Wmissing-attributes detects instances of missing function 2097 attributes on declarations of aliases and weak references. 2098 + [46]-Wstringop-truncation also detects a subset of instances 2099 of reading past the end of unterminated constant character 2100 arrays, 2101 * If a macro is used with the wrong argument count, the C and C++ 2102 front ends now show the definition of that macro via a note. 2103 * The spelling corrector now considers transposed letters, and the 2104 threshold for similarity has been tightened, to avoid nonsensical 2105 suggestions. 2106 2107 C 2108 2109 * There is now experimental support for -std=c2x, to select support 2110 for the upcoming C2X revision of the ISO C standard. This standard 2111 is in the early stages of development and the only feature 2112 supported in GCC 9 is _Static_assert with a single argument 2113 (support for _Static_assert with two arguments was added in C11 and 2114 GCC 4.6). There are also new options -std=gnu2x, for C2X with GNU 2115 extensions, and -Wc11-c2x-compat, to warn for uses of features 2116 added in C2X (such warnings are also enabled by use of -Wpedantic 2117 if not using -std=c2x or -std=gnu2x). 2118 * New warnings: 2119 + [47]-Wabsolute-value warns for calls to standard functions 2120 that compute the absolute value of an argument when a more 2121 appropriate standard function is available. For example, 2122 calling abs(3.14) triggers the warning because the appropriate 2123 function to call to compute the absolute value of a double 2124 argument is fabs. The option also triggers warnings when the 2125 argument in a call to such a function has an unsigned type. 2126 This warning can be suppressed with an explicit type cast and 2127 it is also enabled by -Wextra. 2128 2129 C++ 2130 2131 * New warnings: 2132 + [48]-Wdeprecated-copy, implied by -Wextra, warns about the 2133 C++11 deprecation of implicitly declared copy constructor and 2134 assignment operator if one of them is user-provided. 2135 -Wdeprecated-copy-dtor also warns if the destructor is 2136 user-provided, as specified in C++11. 2137 + [49]-Winit-list-lifetime, on by default, warns about uses of 2138 std::initializer_list that are likely to result in a dangling 2139 pointer, such as returning or assigning from a temporary list. 2140 + [50]-Wredundant-move, implied by -Wextra, warns about 2141 redundant calls to std::move. 2142 + [51]-Wpessimizing-move, implied by -Wall, warns when a call to 2143 std::move prevents copy elision. 2144 + [52]-Wclass-conversion, on by default, warns when a conversion 2145 function will never be called due to the type it converts to. 2146 * The C++ front end has experimental support for some of the upcoming 2147 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 2148 including range-based for statements with initializer, default 2149 constructible and assignable stateless lambdas, lambdas in 2150 unevaluated contexts, language support for empty data members, 2151 allowing pack expansion in lambda init-capture, likely and unlikely 2152 attributes, class types in non-type template parameters, allowing 2153 virtual function calls in constant expressions, explicit(bool), 2154 std::is_constant_evaluated, nested inline namespaces, etc. For a 2155 full list of new features, see [53]the C++ status page. 2156 * The C++ front end now preserves source locations for literals, 2157 id-expression, and mem-initializer for longer. For example it is 2158 now able to pin-point the pertinent locations for bad 2159 initializations such as these 2160$ g++ -c bad-inits.cc 2161bad-inits.cc:10:14: error: cannot convert 'json' to 'int' in initialization 2162 10 | { 3, json::object }, 2163 | ~~~~~~^~~~~~ 2164 | | 2165 | json 2166bad-inits.cc:14:31: error: initializer-string for array of chars is too long [-f 2167permissive] 2168 14 | char buffers[3][5] = { "red", "green", "blue" }; 2169 | ^~~~~~~ 2170bad-inits.cc: In constructor 'X::X()': 2171bad-inits.cc:17:13: error: invalid conversion from 'int' to 'void*' [-fpermissiv 2172e] 2173 17 | X() : one(42), two(42), three(42) 2174 | ^~ 2175 | | 2176 | int 2177 2178 rather than emitting the error at the final closing parenthesis or 2179 brace. 2180 * Error-reporting of overload resolution has been special-cased to 2181 make the case of a single failed candidate easier to read. For 2182 example: 2183$ g++ param-type-mismatch.cc 2184param-type-mismatch.cc: In function 'int test(int, const char*, float)': 2185param-type-mismatch.cc:8:32: error: cannot convert 'const char*' to 'const char* 2186*' 2187 8 | return foo::member_1 (first, second, third); 2188 | ^~~~~~ 2189 | | 2190 | const char* 2191param-type-mismatch.cc:3:46: note: initializing argument 2 of 'static int foo: 2192:member_1(int, const char**, float)' 2193 3 | static int member_1 (int one, const char **two, float three); 2194 | ~~~~~~~~~~~~~^~~ 2195 2196 highlights both the problematic argument, and the parameter that it 2197 can't be converted to. 2198 * Diagnostics involving binary operators now use color to distinguish 2199 the two operands, and label them separately (as per the example of 2200 source labelling above). 2201 * Diagnostics involving function calls now highlight the pertinent 2202 parameter of the declaration in more places. 2203$ g++ bad-conversion.cc 2204bad-conversion.cc: In function 'void caller()': 2205bad-conversion.cc:9:14: error: cannot convert 'bool' to 'void*' 2206 9 | callee (0, false, 2); 2207 | ^~~~~ 2208 | | 2209 | bool 2210bad-conversion.cc:3:19: note: initializing argument 2 of 'void callee(int, voi 2211d*, int)' 2212 3 | void callee (int, void *, int) 2213 | ^~~~~~ 2214 2215 * The C++ front end's implementation of [54]-Wformat now shows 2216 precise locations within string literals, and underlines the 2217 pertinent arguments at bogus call sites (the C front end has been 2218 doing this since GCC 7). For example: 2219$ g++ -c bad-printf.cc -Wall 2220bad-printf.cc: In function 'void print_field(const char*, float, long int, long 2221int)': 2222bad-printf.cc:6:17: warning: field width specifier '*' expects argument of type 2223'int', but argument 3 has type 'long int' [-Wformat=] 2224 6 | printf ("%s: %*ld ", fieldname, column - width, value); 2225 | ~^~~ ~~~~~~~~~~~~~~ 2226 | | | 2227 | int long int 2228bad-printf.cc:6:19: warning: format '%ld' expects argument of type 'long int', b 2229ut argument 4 has type 'double' [-Wformat=] 2230 6 | printf ("%s: %*ld ", fieldname, column - width, value); 2231 | ~~~^ ~~~~~ 2232 | | | 2233 | long int double 2234 | %*f 2235 2236 * The C++ front end has gained new fix-it hints for forgetting the 2237 return *this; needed by various C++ operators: 2238$ g++ -c operator.cc 2239operator.cc: In member function 'boxed_ptr& boxed_ptr::operator=(const boxed_ptr 2240&)': 2241operator.cc:7:3: warning: no return statement in function returning non-void [-W 2242return-type] 2243 6 | m_ptr = other.m_ptr; 2244 +++ |+ return *this; 2245 7 | } 2246 | ^ 2247 2248 for when the compiler needs a typename: 2249$ g++ -c template.cc 2250template.cc:3:3: error: need 'typename' before 'Traits::type' because 'Traits' i 2251s a dependent scope 2252 3 | Traits::type type; 2253 | ^~~~~~ 2254 | typename 2255 2256 when trying to use an accessor member as if it were a data member: 2257$ g++ -c fncall.cc 2258fncall.cc: In function 'void hangman(const mystring&)': 2259fncall.cc:12:11: error: invalid use of member function 'int mystring::get_length 2260() const' (did you forget the '()' ?) 2261 12 | if (str.get_length > 0) 2262 | ~~~~^~~~~~~~~~ 2263 | () 2264 2265 for C++11's scoped enums: 2266$ g++ -c enums.cc 2267enums.cc: In function 'void json::test(const json::value&)': 2268enums.cc:12:26: error: 'STRING' was not declared in this scope; did you mean 'js 2269on::kind::STRING'? 2270 12 | if (v.get_kind () == STRING) 2271 | ^~~~~~ 2272 | json::kind::STRING 2273enums.cc:3:44: note: 'json::kind::STRING' declared here 2274 3 | enum class kind { OBJECT, ARRAY, NUMBER, STRING, TRUE, FALSE, NULL_ }; 2275 | ^~~~~~ 2276 2277 and a tweak to integrate the suggestions about misspelled members 2278 with that for accessors: 2279$ g++ -c accessor-fixit.cc 2280accessor-fixit.cc: In function 'int test(t*)': 2281accessor-fixit.cc:17:15: error: 'class t' has no member named 'ratio'; did you m 2282ean 'int t::m_ratio'? (accessible via 'int t::get_ratio() const') 2283 17 | return ptr->ratio; 2284 | ^~~~~ 2285 | get_ratio() 2286 2287 In addition, various diagnostics in the C++ front-end have been 2288 streamlined by consolidating the suggestion into the initial error, 2289 rather than emitting a follow-up note: 2290$ g++ typo.cc 2291typo.cc:5:13: error: 'BUFSIZE' was not declared in this scope; did you mean 'BUF 2292_SIZE'? 2293 5 | uint8_t buf[BUFSIZE]; 2294 | ^~~~~~~ 2295 | BUF_SIZE 2296 2297 Runtime Library (libstdc++) 2298 2299 * Improved support for C++17, including: 2300 + The C++17 implementation is no longer experimental. 2301 + Parallel algorithms and <execution> (requires [55]Thread 2302 Building Blocks 2018 or newer). 2303 + <memory_resource>. 2304 + Using the types and functions in <filesystem> does not require 2305 linking with -lstdc++fs now. 2306 * Improved experimental support for C++2a, including: 2307 + Type traits std::remove_cvref, std::unwrap_reference, 2308 std::unwrap_decay_ref, std::is_nothrow_convertible, and 2309 std::type_identity. 2310 + Headers <bit> and <version>. 2311 + Uniform container erasure (std::erase_if). 2312 + contains member of maps and sets. 2313 + String prefix and suffix checking (starts_with, ends_with). 2314 + Functions std::midpoint and std::lerp for interpolation. 2315 + std::bind_front. 2316 + std::visit<R>. 2317 + std::assume_aligned. 2318 + Uses-allocator construction utilities. 2319 + std::pmr::polymorphic_allocator<std::byte>. 2320 + Library support for char8_t type. 2321 + Destroying delete. 2322 + std::is_constant_evaluated() function. 2323 * Support for opening file streams with wide character paths on 2324 Windows 2325 * Incomplete support for the C++17 Filesystem library and the 2326 Filesystem TS on Windows. 2327 * Incomplete, experimental support for the Networking TS. 2328 2329 D 2330 2331 * Support for the D programming language has been added to GCC, 2332 implementing version 2.076 of the language and run-time library. 2333 2334 Fortran 2335 2336 * Asynchronous I/O is now fully supported. The program needs to be 2337 linked against the pthreads library to use it, otherwise the I/O is 2338 done synchronously. For systems which do not support POSIX 2339 condition variables, such as AIX, all I/O is still done 2340 synchronously. 2341 * The BACK argument for MINLOC and MAXLOC has been implemented. 2342 * The FINDLOC intrinsic function has been implemented. 2343 * The IS_CONTIGUOUS intrinsic function has been implemented. 2344 * Direct access to the real and imaginary parts of a complex variable 2345 via c%re and c%im has been implemented. 2346 * Type parameter inquiry via str%len and a%kind has been implemented. 2347 * C descriptors and the ISO_Fortran_binding.h source file have been 2348 implemented. 2349 * The MAX and MIN intrinsics are no longer guaranteed to return any 2350 particular value in case one of the arguments is a NaN. Note that 2351 this conforms to the Fortran standard and to what other Fortran 2352 compilers do. If there is a need to handle that case in some 2353 specific way, one needs to explicitly check for NaN's before 2354 calling MAX or MIN, e.g. by using the IEEE_IS_NAN function from the 2355 intrinsic module IEEE_ARITHMETIC. 2356 * A new command-line option [56]-fdec-include, set also by the 2357 [57]-fdec option, has been added to increase compatibility with 2358 legacy code. With this option, an INCLUDE directive is also parsed 2359 as a statement, which allows the directive to be spread across 2360 multiple source lines with line continuations. 2361 * A new [58]BUILTIN directive, has been added. The purpose of the 2362 directive is to provide an API between the GCC compiler and the GNU 2363 C Library which would define vector implementations of math 2364 routines. 2365 2366 Go 2367 2368 * GCC 9 provides a complete implementation of the Go 1.12.2 user 2369 packages. 2370 2371libgccjit 2372 2373 * The libgccjit API gained a new entry point: 2374 [59]gcc_jit_context_add_driver_option. 2375 2376New Targets and Target Specific Improvements 2377 2378 AArch64 & Arm 2379 2380 * Support has been added for the following processors (GCC 2381 identifiers in parentheses): 2382 + Arm Cortex-A76 (cortex-a76). 2383 + Arm Cortex-A55/Cortex-A76 DynamIQ big.LITTLE 2384 (cortex-a76.cortex-a55). 2385 + Arm Neoverse N1 (neoverse-n1). 2386 The GCC identifiers can be used as arguments to the -mcpu or -mtune 2387 options, for example: -mcpu=cortex-a76 or 2388 -mtune=cortex-a76.cortex-a55 or as arguments to the equivalent 2389 target attributes and pragmas. 2390 * The Armv8.3-A complex number instructions are now supported via 2391 intrinsics when the option -march=armv8.3-a or equivalent is 2392 specified. For the half-precision floating-point variants of these 2393 instructions use the architecture extension flag +fp16, e.g. 2394 -march=armv8.3-a+fp16. 2395 The intrinsics are defined by the ACLE specification. 2396 * The Armv8.5-A architecture is now supported through the 2397 -march=armv8.5-a option. 2398 * The Armv8.5-A architecture also adds some security features that 2399 are optional to all older architecture versions. These are now 2400 supported and only affect the assembler. 2401 + Speculation Barrier instruction through the -march=armv8-a+sb 2402 option. 2403 + Execution and Data Prediction Restriction instructions through 2404 the -march=armv8-a+predres option. 2405 + Speculative Store Bypass Safe instruction through the 2406 -march=armv8-a+ssbs option. This does not require a compiler 2407 option for Arm and thus -march=armv8-a+ssbs is an 2408 AArch64-specific option. 2409 2410 AArch64 specific 2411 2412 * Support has been added for the Arm Neoverse E1 processor 2413 (-mcpu=neoverse-e1). 2414 * The AArch64 port now has support for stack clash protection using 2415 the [60]-fstack-clash-protection option. The probing interval/guard 2416 size can be set by using --param 2417 stack-clash-protection-guard-size=12|16. The value of this 2418 parameter must be in bytes represented as a power of two. The two 2419 supported values for this parameter are 12 (for a 4KiB size, 2^12) 2420 and 16 (for a 64KiB size, 2^16). The default value is 16 (64Kb) and 2421 can be changed at configure time using the flag 2422 --with-stack-clash-protection-guard-size=12|16. 2423 * The option -msign-return-address= has been deprecated. This has 2424 been replaced by the new -mbranch-protection= option. This new 2425 option can now be used to enable the return address signing as well 2426 as the new Branch Target Identification feature of Armv8.5-A 2427 architecture. For more information on the arguments accepted by 2428 this option, please refer to [61]AArch64-Options. 2429 * The following optional extensions to Armv8.5-A architecture are now 2430 supported and only affect the assembler. 2431 + Random Number Generation instructions through the 2432 -march=armv8.5-a+rng option. 2433 + Memory Tagging Extension through the -march=armv8.5-a+memtag 2434 option. 2435 2436 Arm specific 2437 2438 * Support for the deprecated Armv2 and Armv3 architectures and their 2439 variants has been removed. Their corresponding -march values and 2440 the -mcpu options that used these architectures have been removed. 2441 * Support for the Armv5 and Armv5E architectures (which have no known 2442 implementations) has been removed. Note that Armv5T, Armv5TE and 2443 Armv5TEJ architectures remain supported. 2444 * Corrected FPU configurations for Cortex-R7 and Cortex-R8 when using 2445 their respective -mcpu options. 2446 2447 AMD GCN 2448 2449 * A new back end targeting AMD GCN GPUs has been contributed to GCC. 2450 The implementation is currently limited to compiling 2451 single-threaded, stand-alone programs. Future versions will add 2452 support for offloading multi-threaded kernels via OpenMP and 2453 OpenACC. The following devices are supported (GCC identifiers in 2454 parentheses): 2455 + Fiji (fiji). 2456 + Vega 10 (gfx900). 2457 2458 ARC 2459 2460 * LRA is now on by default for the ARC target. This can be controlled 2461 by -mlra. 2462 * Add support for frame code-density and branch-and-index 2463 instructions. 2464 2465 C-SKY 2466 2467 * A new back end targeting C-SKY V2 processors has been contributed 2468 to GCC. 2469 2470 IA-32/x86-64 2471 2472 * Support of Intel MPX (Memory Protection Extensions) has been 2473 removed. 2474 * New ISA extension support for Intel PTWRITE was added to GCC. 2475 PTWRITE intrinsics are available via the -mptwrite compiler switch. 2476 * GCC now supports the Intel CPU named Cascade Lake with AVX512 2477 extensions through -march=cascadelake. The switch enables the 2478 following ISA extensions: AVX512F, AVX512VL, AVX512CD, AVX512BW, 2479 AVX512DQ, AVX512VNNI. 2480 2481 OpenRISC 2482 2483 * A new back end targeting OpenRISC processors has been contributed 2484 to GCC. 2485 2486 S/390, System z, IBM z Systems 2487 2488 * Support for the arch13 architecture has been added. When using the 2489 -march=arch13 option, the compiler will generate code making use of 2490 the new instructions introduced with the vector enhancement 2491 facility 2 and the miscellaneous instruction extension facility 2. 2492 The -mtune=arch13 option enables arch13 specific instruction 2493 scheduling without making use of new instructions. 2494 * Builtins for the new vector instructions have been added and can be 2495 enabled using the -mzvector option. 2496 * Support for ESA architecture machines g5 and g6 is deprecated since 2497 GCC 6.1.0 and has been removed now. 2498 * When compiling with -march=z14 or higher GCC emits alignments hints 2499 on the vector load/store instructions (8 or 16 byte). 2500 * Functions now have a default alignment of 16 bytes. This helps with 2501 branch prediction effects. 2502 * -mfentry is now supported. As well as the mcount mechanism the 2503 __fentry__ is called before the function prologue. However, since 2504 just a single instruction is required to call __fentry__ the call 2505 sequence imposes a smaller overhead than mcount (4 instructions). 2506 The produced code is compatible only with newer glibc versions, 2507 which provide the __fentry__ symbol and do not clobber r0 when 2508 resolving lazily bound functions. -mfentry is only supported when 2509 generating 64 bit code and does not work with nested C functions. 2510 * The -mnop-mcount option can be used to emit NOP instructions 2511 instead of an mcount or fentry call stub. 2512 * With the -mrecord-mcount option a __mcount_loc section is generated 2513 containing pointers to each profiling call stub. This is useful for 2514 automatically patching in and out calls. 2515 2516Operating Systems 2517 2518 Solaris 2519 2520 * g++ now unconditionally enables large file support when compiling 2521 32-bit code. 2522 * Support for the AddressSanitizer and UndefinedBehaviorSanitizer has 2523 been merged from LLVM. For the moment, this only works for 32-bit 2524 code on both SPARC and x86. 2525 * An initial port of the D runtime library has been completed on 2526 Solaris 11/x86. It requires the use of GNU as. Solaris 11/SPARC 2527 support is still work-in-progress. 2528 2529 Windows 2530 2531 * A C++ Microsoft ABI bitfield layout bug, [62]PR87137 has been 2532 fixed. A non-field declaration could cause the current bitfield 2533 allocation unit to be completed, incorrectly placing a following 2534 bitfield into a new allocation unit. The Microsoft ABI is selected 2535 for: 2536 + Mingw targets 2537 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 2538 option is specified, or __attribute__((ms_struct)) is used 2539 + SuperH targets when the -mhitachi option is specified, or 2540 __attribute__((renesas)) is used 2541 2542Improvements for plugin authors 2543 2544 * GCC's diagnostic subsystem now has a way to logically group 2545 together related diagnostics, auto_diagnostic_group. Such 2546 diagnostics will be nested by the output of 2547 [63]-fdiagnostics-format=json. 2548 * GCC now has a set of [64]user experience guidelines for GCC, with 2549 information and advice on implementing new diagnostics. 2550 2551Other significant improvements 2552 2553 * GCC's internal "selftest" suite now runs for C++ as well as C (in 2554 debug builds of the compiler). 2555 2556GCC 9.1 2557 2558 This is the [65]list of problem reports (PRs) from GCC's bug tracking 2559 system that are known to be fixed in the 9.1 release. This list might 2560 not be complete (that is, it is possible that some PRs that have been 2561 fixed are not listed here). 2562 2563GCC 9.2 2564 2565 This is the [66]list of problem reports (PRs) from GCC's bug tracking 2566 system that are known to be fixed in the 9.2 release. This list might 2567 not be complete (that is, it is possible that some PRs that have been 2568 fixed are not listed here). 2569 2570GCC 9.3 2571 2572 This is the [67]list of problem reports (PRs) from GCC's bug tracking 2573 system that are known to be fixed in the 9.3 release. This list might 2574 not be complete (that is, it is possible that some PRs that have been 2575 fixed are not listed here). 2576 2577GCC 9.4 2578 2579 This is the [68]list of problem reports (PRs) from GCC's bug tracking 2580 system that are known to be fixed in the 9.4 release. This list might 2581 not be complete (that is, it is possible that some PRs that have been 2582 fixed are not listed here). 2583 2584 Target Specific Changes 2585 2586 AArch64 2587 2588 * The option -moutline-atomics has been added to aid deployment of 2589 the Large System Extensions (LSE) on GNU/Linux systems built with a 2590 baseline architecture targeting Armv8-A. When the option is 2591 specified code is emitted to detect the presence of LSE 2592 instructions at run time and use them for standard atomic 2593 operations. For more information please refer to the documentation. 2594 * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune 2595 options are -mcpu=a64fx and -mtune=a64fx respectively. In 2596 particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and 2597 tunes the code for the A64FX. This includes tuning the SVE code, 2598 although by default the code is still length-agnostic and so works 2599 for all SVE implementations. Adding -msve-vector-bits=512 makes the 2600 code specific to 512-bit SVE. 2601 2602 2603 For questions related to the use of GCC, please consult these web 2604 pages and the [69]GCC manuals. If that fails, the 2605 [70]gcc-help@gcc.gnu.org mailing list might help. Comments on these 2606 web pages and the development of GCC are welcome on our developer 2607 list at [71]gcc@gcc.gnu.org. All of [72]our lists have public 2608 archives. 2609 2610 Copyright (C) [73]Free Software Foundation, Inc. Verbatim copying and 2611 distribution of this entire article is permitted in any medium, 2612 provided this notice is preserved. 2613 2614 These pages are [74]maintained by the GCC team. Last modified 2615 2021-07-28[75]. 2616 2617References 2618 2619 1. http://gcc.gnu.org/gcc-9/porting_to.html 2620 2. http://gcc.gnu.org/onlinedocs/index.html#current 2621 3. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88469 2622 4. https://gcc.gnu.org/ml/gcc/2018-10/msg00139.html 2623 5. https://gcc.gnu.org/ml/gcc/2019-04/msg00023.html 2624 6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90920 2625 7. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-frepo 2626 8. https://gcc.gnu.org/PR90361 2627 9. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-GCC.html#Invoking-GCC 2628 10. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-flive-patching 2629 11. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-line-numbers 2630 12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-labels 2631 13. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2632 14. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-functions 2633 15. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-labels 2634 16. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-loops 2635 17. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-jumps 2636 18. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-filter-files 2637 19. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-exclude-files 2638 20. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fopt-info 2639 21. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fsave-optimization-record 2640 22. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-stack-alignment 2641 23. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-reference-addressable 2642 24. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fexpect_005fwith_005fprobability 2643 25. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fhas_005fattribute-1 2644 26. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fspeculation_005fsafe_005fvalue-1 2645 27. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Common-Function-Attributes.html#index-copy-function-attribute 2646 28. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-switch-conversion 2647 29. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fprofile-use 2648 30. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fversion-loops-for-strides 2649 31. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-interchange 2650 32. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-unroll-and-jam 2651 33. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-loop-distribution 2652 34. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-Gcov.html#Invoking-Gcov 2653 35. https://www.openacc.org/ 2654 36. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-9 2655 37. https://www.openmp.org/specifications/ 2656 38. https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00628.html 2657 39. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Vector-Extensions.html#index-_005f_005fbuiltin_005fconvertvector 2658 40. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Waddress-of-packed-member 2659 41. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Warray-bounds 2660 42. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wattribute-alias 2661 43. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-overflow 2662 44. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-truncation 2663 45. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 2664 46. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 2665 47. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wabsolute-value 2666 48. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-copy 2667 49. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Winit-list-lifetime 2668 50. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move 2669 51. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move 2670 52. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-conversion 2671 53. http://gcc.gnu.org/projects/cxx-status.html#cxx2a 2672 54. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat 2673 55. https://github.com/oneapi-src/oneTBB 2674 56. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec-include 2675 57. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec 2676 58. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BUILTIN-directive.html#BUILTIN-directive 2677 59. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/jit/topics/contexts.html#gcc_jit_context_add_driver_option 2678 60. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fstack-protector 2679 61. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/AArch64-Options.html#AArch64-Options 2680 62. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 2681 63. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format 2682 64. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gccint/User-Experience-Guidelines.html 2683 65. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.0 2684 66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.2 2685 67. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.3 2686 68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.4 2687 69. https://gcc.gnu.org/onlinedocs/ 2688 70. mailto:gcc-help@gcc.gnu.org 2689 71. mailto:gcc@gcc.gnu.org 2690 72. https://gcc.gnu.org/lists.html 2691 73. https://www.fsf.org/ 2692 74. https://gcc.gnu.org/about.html 2693 75. http://validator.w3.org/check/referer 2694====================================================================== 2695http://gcc.gnu.org/gcc-8/index.html 2696 2697 GCC 8 Release Series 2698 2699 May 14, 2021 2700 2701 The [1]GNU project and the GCC developers are pleased to announce the 2702 release of GCC 8.5. 2703 2704 This release is a bug-fix release, containing fixes for regressions in 2705 GCC 8.4 relative to previous releases of GCC. 2706 2707Release History 2708 2709 GCC 8.5 2710 May 14, 2021 ([2]changes, [3]documentation) 2711 2712 GCC 8.4 2713 Mar 4, 2020 ([4]changes, [5]documentation) 2714 2715 GCC 8.3 2716 Feb 22, 2019 ([6]changes, [7]documentation) 2717 2718 GCC 8.2 2719 Jul 26, 2018 ([8]changes, [9]documentation) 2720 2721 GCC 8.1 2722 May 2, 2018 ([10]changes, [11]documentation) 2723 2724References and Acknowledgements 2725 2726 GCC used to stand for the GNU C Compiler, but since the compiler 2727 supports several other languages aside from C, it now stands for the 2728 GNU Compiler Collection. 2729 2730 A list of [12]successful builds is updated as new information becomes 2731 available. 2732 2733 The GCC developers would like to thank the numerous people that have 2734 contributed new features, improvements, bug fixes, and other changes as 2735 well as test results to GCC. This [13]amazing group of volunteers is 2736 what makes GCC successful. 2737 2738 For additional information about GCC please refer to the [14]GCC 2739 project web site or contact the [15]GCC development mailing list. 2740 2741 To obtain GCC please use [16]our mirror sites or [17]our version 2742 control system. 2743 2744 2745 For questions related to the use of GCC, please consult these web 2746 pages and the [18]GCC manuals. If that fails, the 2747 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 2748 web pages and the development of GCC are welcome on our developer 2749 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 2750 archives. 2751 2752 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 2753 distribution of this entire article is permitted in any medium, 2754 provided this notice is preserved. 2755 2756 These pages are [23]maintained by the GCC team. Last modified 2757 2021-07-28[24]. 2758 2759References 2760 2761 1. http://www.gnu.org/ 2762 2. http://gcc.gnu.org/gcc-8/changes.html 2763 3. http://gcc.gnu.org/onlinedocs/8.5.0/ 2764 4. http://gcc.gnu.org/gcc-8/changes.html 2765 5. http://gcc.gnu.org/onlinedocs/8.4.0/ 2766 6. http://gcc.gnu.org/gcc-8/changes.html 2767 7. http://gcc.gnu.org/onlinedocs/8.3.0/ 2768 8. http://gcc.gnu.org/gcc-8/changes.html 2769 9. http://gcc.gnu.org/onlinedocs/8.2.0/ 2770 10. http://gcc.gnu.org/gcc-8/changes.html 2771 11. http://gcc.gnu.org/onlinedocs/8.1.0/ 2772 12. http://gcc.gnu.org/gcc-8/buildstat.html 2773 13. http://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html 2774 14. http://gcc.gnu.org/index.html 2775 15. mailto:gcc@gcc.gnu.org 2776 16. http://gcc.gnu.org/mirrors.html 2777 17. http://gcc.gnu.org/git.html 2778 18. https://gcc.gnu.org/onlinedocs/ 2779 19. mailto:gcc-help@gcc.gnu.org 2780 20. mailto:gcc@gcc.gnu.org 2781 21. https://gcc.gnu.org/lists.html 2782 22. https://www.fsf.org/ 2783 23. https://gcc.gnu.org/about.html 2784 24. http://validator.w3.org/check/referer 2785====================================================================== 2786http://gcc.gnu.org/gcc-8/changes.html 2787 2788 GCC 8 Release Series 2789 Changes, New Features, and Fixes 2790 2791 This page is a "brief" summary of some of the huge number of 2792 improvements in GCC 8. You may also want to check out our [1]Porting to 2793 GCC 8 page and the [2]full GCC documentation. 2794 2795Caveats 2796 2797 * Support for the obsolete SDB/coff debug info format has been 2798 removed. The option -gcoff no longer does anything. 2799 * The Cilk+ extensions to the C and C++ languages have been removed. 2800 * The MPX extensions to the C and C++ languages have been deprecated 2801 and will be removed in a future release. 2802 * The extension allowing arithmetic on std::atomic<void*> and types 2803 like std::atomic<R(*)()> has been deprecated. 2804 * The non-standard C++0x std::copy_exception function was removed. 2805 std::make_exception_ptr should be used instead. 2806 * Support for the powerpc*-*-*spe* target ports which have been 2807 recently unmaintained and untested in GCC has been declared 2808 obsolete in GCC 8 as [3]announced. Unless there is activity to 2809 revive them, the next release of GCC will have their sources 2810 permanently removed. 2811 2812General Improvements 2813 2814 * Inter-procedural optimization improvements: 2815 + Reworked run-time estimation metrics leading to more realistic 2816 guesses driving inliner and cloning heuristics. 2817 + The ipa-pure-const pass is extended to propagate the malloc 2818 attribute, and the corresponding warning option 2819 -Wsuggest-attribute=malloc emits a diagnostic for functions 2820 which can be annotated with the malloc attribute. 2821 * Profile driven optimization improvements: 2822 + New infrastructure for representing profiles (both statically 2823 guessed and profile feedback) which allows propagation of 2824 additional information about the reliability of the profile. 2825 + A number of improvements in the profile updating code solving 2826 problems found by new verification code. 2827 + Static detection of code which is not executed in a valid run 2828 of the program. This includes paths which trigger undefined 2829 behavior as well as calls to functions declared with the cold 2830 attribute. Newly the noreturn attribute does not imply all 2831 effects of cold to differentiate between exit (which is 2832 noreturn) and abort (which is in addition not executed in 2833 valid runs). 2834 + -freorder-blocks-and-partition, a pass splitting function 2835 bodies into hot and cold regions, is now enabled by default at 2836 -O2 and higher for x86 and x86-64. 2837 * Link-time optimization improvements: 2838 + We have significantly improved debug information on ELF 2839 targets using DWARF by properly preserving language-specific 2840 information. This allows for example the libstdc++ 2841 pretty-printers to work with LTO optimized executables. 2842 * A new option -fcf-protection=[full|branch|return|none] is 2843 introduced to perform code instrumentation to increase program 2844 security by checking that target addresses of control-flow transfer 2845 instructions (such as indirect function call, function return, 2846 indirect jump) are valid. Currently the instrumentation is 2847 supported on x86 GNU/Linux targets only. See the user guide for 2848 further information about the option syntax and section "New 2849 Targets and Target Specific Improvements" for IA-32/x86-64 for more 2850 details. 2851 * The -gcolumn-info option is now enabled by default. It includes 2852 column information in addition to just filenames and line numbers 2853 in DWARF debugging information. 2854 * The polyhedral-based loop nest optimization pass 2855 -floop-nest-optimize has been overhauled. It's still considered 2856 experimental and may not result in any runtime improvements. 2857 * Two new classical loop nest optimization passes have been added. 2858 -floop-unroll-and-jam performs outer loop unrolling and fusing of 2859 the inner loop copies. -floop-interchange exchanges loops in a loop 2860 nest to improve data locality. Both passes are enabled by default 2861 at -O3 and above. 2862 * The classic loop nest optimization pass -ftree-loop-distribution 2863 has been improved and enabled by default at -O3 and above. It 2864 supports loop nest distribution in some restricted scenarios; it 2865 also supports cancellable innermost loop distribution with loop 2866 versioning under run-time alias checks. 2867 * The new option -fstack-clash-protection causes the compiler to 2868 insert probes whenever stack space is allocated statically or 2869 dynamically to reliably detect stack overflows and thus mitigate 2870 the attack vector that relies on jumping over a stack guard page as 2871 provided by the operating system. 2872 * A new pragma GCC unroll has been implemented in the C family of 2873 languages, as well as Fortran and Ada, so as to make it possible 2874 for the user to have a finer-grained control over the loop 2875 unrolling optimization. 2876 * GCC has been enhanced to detect more instances of meaningless or 2877 mutually exclusive attribute specifications and handle such 2878 conflicts more consistently. Mutually exclusive attribute 2879 specifications are ignored with a warning regardless of whether 2880 they appear on the same declaration or on distinct declarations of 2881 the same entity. For example, because the noreturn attribute on the 2882 second declaration below is mutually exclusive with the malloc 2883 attribute on the first, it is ignored and a warning is issued. 2884> 2885 void* __attribute__ ((malloc)) f (unsigned); 2886 void* __attribute__ ((noreturn)) f (unsigned); 2887 2888 warning: ignoring attribute 'noreturn' because it conflicts with attribute 2889 'malloc' [-Wattributes] 2890 * The gcov tool can distinguish functions that begin on a same line 2891 in a source file. This can be a different template instantiation or 2892 a class constructor: 2893 2894File 'ins.C' 2895Lines executed:100.00% of 8 2896Creating 'ins.C.gcov' 2897 2898 -: 0:Source:ins.C 2899 -: 0:Graph:ins.gcno 2900 -: 0:Data:ins.gcda 2901 -: 0:Runs:1 2902 -: 0:Programs:1 2903 -: 1:template<class T> 2904 -: 2:class Foo 2905 -: 3:{ 2906 -: 4: public: 2907 2: 5: Foo(): b (1000) {} 2908------------------ 2909Foo<char>::Foo(): 2910 1: 5: Foo(): b (1000) {} 2911------------------ 2912Foo<int>::Foo(): 2913 1: 5: Foo(): b (1000) {} 2914------------------ 2915 2: 6: void inc () { b++; } 2916------------------ 2917Foo<char>::inc(): 2918 1: 6: void inc () { b++; } 2919------------------ 2920Foo<int>::inc(): 2921 1: 6: void inc () { b++; } 2922------------------ 2923 -: 7: 2924 -: 8: private: 2925 -: 9: int b; 2926 -: 10:}; 2927 -: 11: 2928 1: 12:int main(int argc, char **argv) 2929 -: 13:{ 2930 1: 14: Foo<int> a; 2931 1: 15: Foo<char> b; 2932 -: 16: 2933 1: 17: a.inc (); 2934 1: 18: b.inc (); 2935 1: 19:} 2936 2937 * The gcov tool has more accurate numbers for execution of lines in a 2938 source file. 2939 * The gcov tool can use TERM colors to provide more readable output. 2940 * AddressSanitizer gained a new pair of sanitization options, 2941 -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which 2942 warn about subtraction (or comparison) of pointers that point to a 2943 different memory object: 2944 2945int 2946main () 2947{ 2948 /* Heap allocated memory. */ 2949 char *heap1 = (char *)__builtin_malloc (42); 2950 char *heap2 = (char *)__builtin_malloc (42); 2951 if (heap1 > heap2) 2952 return 1; 2953 2954 return 0; 2955} 2956 2957==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000 295800050 2959 #0 0x40070f in main /tmp/pointer-compare.c:7 2960 #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 2961 #2 0x400629 in _start (/tmp/a.out+0x400629) 2962 29630x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604 296400000003a) 2965allocated by thread T0 here: 2966 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 2967_malloc_linux.cc:86 2968 #1 0x4006ea in main /tmp/pointer-compare.c:5 2969 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 2970 29710x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604 297200000007a) 2973allocated by thread T0 here: 2974 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 2975_malloc_linux.cc:86 2976 #1 0x4006f8 in main /tmp/pointer-compare.c:6 2977 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 2978 2979SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main 2980 2981 * The store merging pass has been enhanced to handle bit-fields and 2982 not just constant stores, but also data copying from adjacent 2983 memory locations into other adjacent memory locations, including 2984 bitwise logical operations on the data. The pass can also handle 2985 byte swapping into memory locations. 2986 * The undefined behavior sanitizer gained two new options included in 2987 -fsanitize=undefined: -fsanitize=builtin which diagnoses at run 2988 time invalid arguments to __builtin_clz or __builtin_ctz prefixed 2989 builtins, and -fsanitize=pointer-overflow which performs cheap run 2990 time tests for pointer wrapping. 2991 * A new attribute no_sanitize can be applied to functions to instruct 2992 the compiler not to do sanitization of the options provided as 2993 arguments to the attribute. Acceptable values for no_sanitize match 2994 those acceptable by the -fsanitize command-line option. 2995 2996void __attribute__ ((no_sanitize ("alignment", "object-size"))) 2997f () { /* Do something. */; } 2998 2999New Languages and Language specific improvements 3000 3001 Ada 3002 3003 * For its internal exception handling used on the host for error 3004 recovery in the front-end, the compiler now relies on the native 3005 exception handling mechanism of the host platform, which should be 3006 more efficient than the former mechanism. 3007 3008 BRIG (HSAIL) 3009 3010 In this release cycle, the focus for the BRIGFE was on stabilization 3011 and performance improvements. Also a couple of completely new features 3012 were added. 3013 * Improved support for function and module scope group segment 3014 variables. PRM specs define function and module scope group segment 3015 variables as an experimental feature. However, PRM test suite uses 3016 them. Now group segment is handled by separate book keeping of 3017 module scope and function (kernel) offsets. Each function has a 3018 "frame" in the group segment offset to which is given as an 3019 argument, similar to traditional call stack frame handling. 3020 * Reduce the number of type conversions due to the untyped HSAIL 3021 registers. Instead of always representing the HSAIL's untyped 3022 registers as unsigned int, the gccbrig now pre-analyzes the BRIG 3023 code and builds the register variables as a type used the most when 3024 storing or reading data to/from each register. This reduces the 3025 number of total casts which cannot be always optimized away. 3026 * Support for BRIG_KIND_NONE directives. 3027 * Made -O3 the default optimization level for BRIGFE. 3028 * Fixed illegal addresses generated from address expressions which 3029 refer only to offset 0. 3030 * Fixed a bug with reg+offset addressing on 32b segments. In 'large' 3031 mode, the offset is treated as 32bits unless it's in global, 3032 read-only or kernarg address space. 3033 * Fixed a crash caused sometimes by calls with more than 4 arguments. 3034 * Fixed a mis-execution issue with kernels that have both unexpanded 3035 ID functions and calls to subfunctions. 3036 * Treat HSAIL barrier builtins as setjmp/longjump style functions to 3037 avoid illegal optimizations. 3038 * Ensure per WI copies of private variables are aligned correctly. 3039 * libhsail-rt: Assume the host runtime allocates the work group 3040 memory. 3041 3042 C family 3043 3044 * New command-line options have been added for the C and C++ 3045 compilers: 3046 + [4]-Wmultistatement-macros warns about unsafe macros expanding 3047 to multiple statements used as a body of a statement such as 3048 if, else, while, switch, or for. 3049 + [5]-Wstringop-truncation warns for calls to bounded string 3050 manipulation functions such as strncat, strncpy, and stpncpy 3051 that might either truncate the copied string or leave the 3052 destination unchanged. For example, the following call to 3053 strncat is diagnosed because it appends just three of the four 3054 characters from the source string. 3055void append (char *buf, size_t bufsize) 3056{ 3057 strncat (buf, ".txt", 3); 3058} 3059warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [- 3060Wstringop-truncation] 3061 Similarly, in the following example, the call to strncpy 3062 specifies the size of the destination buffer as the bound. If 3063 the length of the source string is equal to or greater than 3064 this size the result of the copy will not be NUL-terminated. 3065 Therefore, the call is also diagnosed. To avoid the warning, 3066 specify sizeof buf - 1 as the bound and set the last element 3067 of the buffer to NUL. 3068void copy (const char *s) 3069{ 3070 char buf[80]; 3071 strncpy (buf, s, sizeof buf); 3072 ... 3073} 3074warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca 3075tion] 3076 The -Wstringop-truncation option is included in -Wall. 3077 Note that due to GCC bug [6]82944, defining strncat, strncpy, 3078 or stpncpy as a macro in a system header as some 3079 implementations do, suppresses the warning. 3080 + [7]-Wif-not-aligned controls warnings issued in response to 3081 invalid uses of objects declared with attribute 3082 [8]warn_if_not_aligned. 3083 The -Wif-not-aligned option is included in -Wall. 3084 + [9]-Wmissing-attributes warns when a declaration of a function 3085 is missing one or more attributes that a related function is 3086 declared with and whose absence may adversely affect the 3087 correctness or efficiency of generated code. For example, in 3088 C++, the warning is issued when an explicit specialization of 3089 a primary template declared with attribute alloc_align, 3090 alloc_size, assume_aligned, format, format_arg, malloc, or 3091 nonnull is declared without it. Attributes deprecated, error, 3092 and warning suppress the warning. 3093 The -Wmissing-attributes option is included in -Wall. 3094 + [10]-Wpacked-not-aligned warns when a struct or union declared 3095 with attribute packed defines a member with an explicitly 3096 specified alignment greater than 1. Such a member will wind up 3097 under-aligned. For example, a warning will be issued for the 3098 definition of struct A in the following: 3099struct __attribute__ ((aligned (8))) 3100S8 { char a[8]; }; 3101 3102struct __attribute__ ((packed)) A 3103{ 3104 struct S8 s8; 3105}; 3106warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned] 3107 The -Wpacked-not-aligned option is included in -Wall. 3108 + -Wcast-function-type warns when a function pointer is cast to 3109 an incompatible function pointer. This warning is enabled by 3110 -Wextra. 3111 + -Wsizeof-pointer-div warns for suspicious divisions of the 3112 size of a pointer by the size of the elements it points to, 3113 which looks like the usual way to compute the array size but 3114 won't work out correctly with pointers. This warning is 3115 enabled by -Wall. 3116 + -Wcast-align=strict warns whenever a pointer is cast such that 3117 the required alignment of the target is increased. For 3118 example, warn if a char * is cast to an int * regardless of 3119 the target machine. 3120 + -fprofile-abs-path creates absolute path names in the .gcno 3121 files. This allows gcov to find the correct sources in 3122 projects where compilations occur with different working 3123 directories. 3124 * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and 3125 signed integer overflow is now undefined by default at all 3126 optimization levels. Using -fsanitize=signed-integer-overflow is 3127 now the preferred way to audit code, -Wstrict-overflow is 3128 deprecated. 3129 * The [11]-Warray-bounds option has been improved to detect more 3130 instances of out-of-bounds array indices and pointer offsets. For 3131 example, negative or excessive indices into flexible array members 3132 and string literals are detected. 3133 * The [12]-Wrestrict option introduced in GCC 7 has been enhanced to 3134 detect many more instances of overlapping accesses to objects via 3135 restrict-qualified arguments to standard memory and string 3136 manipulation functions such as memcpy and strcpy. For example, the 3137 strcpy call in the function below attempts to truncate the string 3138 by replacing its initial characters with the last four. However, 3139 because the function writes the terminating NUL into a[4], the 3140 copies overlap and the call is diagnosed. 3141void f (void) 3142{ 3143 char a[] = "abcd1234"; 3144 strcpy (a, a + 4); 3145 ... 3146} 3147warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset 3148 4 [-Wrestrict] 3149 The -Wrestrict option is included in -Wall. 3150 * Several optimizer enhancements have enabled improvements to the 3151 [13]-Wformat-overflow and [14]-Wformat-truncation options. The 3152 warnings detect more instances of buffer overflow and truncation 3153 than in GCC 7 and are better at avoiding certain kinds of false 3154 positives. 3155 * When reporting mismatching argument types at a function call, the C 3156 and C++ compilers now underline both the argument and the pertinent 3157 parameter in the declaration. 3158$ gcc arg-type-mismatch.cc 3159arg-type-mismatch.cc: In function 'int caller(int, int, float)': 3160arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*' 3161 [-fpermissive] 3162 return callee(first, second, third); 3163 ^~~~~~ 3164arg-type-mismatch.cc:1:40: note: initializing argument 2 of 'int callee(int, c 3165onst char*, float)' 3166 extern int callee(int one, const char *two, float three); 3167 ~~~~~~~~~~~~^~~ 3168 3169 * When reporting on unrecognized identifiers, the C and C++ compilers 3170 will now emit fix-it hints suggesting #include directives for 3171 various headers in the C and C++ standard libraries. 3172$ gcc incomplete.c 3173incomplete.c: In function 'test': 3174incomplete.c:3:10: error: 'NULL' undeclared (first use in this function) 3175 return NULL; 3176 ^~~~ 3177incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge 3178t to '#include <stddef.h>'? 3179incomplete.c:1:1: 3180+#include <stddef.h> 3181 const char *test(void) 3182incomplete.c:3:10: 3183 return NULL; 3184 ^~~~ 3185incomplete.c:3:10: note: each undeclared identifier is reported only once for ea 3186ch function it appears in 3187 3188$ gcc incomplete.cc 3189incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type 3190 std::string s("hello world"); 3191 ^~~~~~ 3192incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you 3193forget to '#include <string>'? 3194+#include <string> 3195 std::string s("hello world"); 3196 ^~~ 3197 3198 * The C and C++ compilers now use more intuitive locations when 3199 reporting on missing semicolons, and offer fix-it hints: 3200$ gcc t.c 3201t.c: In function 'test': 3202t.c:3:12: error: expected ';' before '}' token 3203 return 42 3204 ^ 3205 ; 3206 } 3207 ~ 3208 3209 * When reporting on missing '}' and ')' tokens, the C and C++ 3210 compilers will now highlight the corresponding '{' and '(' token, 3211 issuing a 'note' if it's on a separate line: 3212$ gcc unclosed.c 3213unclosed.c: In function 'log_when_out_of_range': 3214unclosed.c:12:50: error: expected ')' before '{' token 3215 && (temperature < MIN || temperature > MAX) { 3216 ^~ 3217 ) 3218unclosed.c:11:6: note: to match this '(' 3219 if (logging_enabled && check_range () 3220 ^ 3221 3222 or highlighting it directly if it's on the same line: 3223$ gcc unclosed-2.c 3224unclosed-2.c: In function 'test': 3225unclosed-2.c:8:45: error: expected ')' before '{' token 3226 if (temperature < MIN || temperature > MAX { 3227 ~ ^~ 3228 ) 3229 3230 They will also emit fix-it hints. 3231 3232 C++ 3233 3234 * GCC 8 (-fabi-version=12) has a couple of corrections to the calling 3235 convention, which changes the ABI for some uncommon code: 3236 + Passing an empty class as an argument now takes up no space on 3237 x86_64, as required by the psABI. 3238 + Passing or returning a class with only deleted copy and move 3239 constructors now uses the same calling convention as a class 3240 with a non-trivial copy or move constructor. This only affects 3241 C++17 mode, as in earlier standards passing or returning such 3242 a class was impossible. 3243 + WARNING: In GCC 8.1 the second change mistakenly also affects 3244 classes with a deleted copy constructor and defaulted trivial 3245 move constructor (bug [15]c++/86094). This issue is fixed in 3246 GCC 8.2 (-fabi-version=13). 3247 You can test whether these changes affect your code with -Wabi=11 3248 (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are 3249 problematic for your project, the GCC 7 ABI can be selected with 3250 -fabi-version=11. 3251 * The value of the C++11 alignof operator has been corrected to match 3252 C _Alignof (minimum alignment) rather than GNU __alignof__ 3253 (preferred alignment); on ia32 targets this means that 3254 alignof(double) is now 4 rather than 8. Code that wants the 3255 preferred alignment should use __alignof__ instead. 3256 * New command-line options have been added for the C++ compiler to 3257 control warnings: 3258 + [16]-Wclass-memaccess warns when objects of non-trivial class 3259 types are manipulated in potentially unsafe ways by raw memory 3260 functions such as memcpy, or realloc. The warning helps detect 3261 calls that bypass user-defined constructors or copy-assignment 3262 operators, corrupt virtual table pointers, data members of 3263 const-qualified types or references, or member pointers. The 3264 warning also detects calls that would bypass access controls 3265 to data members. For example, a call such as: 3266 memcpy (&std::cout, &std::cerr, sizeof std::cout); 3267 results in 3268 warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t 3269o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n 3270o trivial copy-assignment [-Wclass-memaccess] 3271 The -Wclass-memaccess option is included in -Wall. 3272 * The C++ front end has experimental support for some of the upcoming 3273 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 3274 including designated initializers, default member initializers for 3275 bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported), 3276 lambda [=, this] captures, etc. For a full list of new features, 3277 see [17]the C++ status page. 3278 * When reporting on attempts to access private fields of a class or 3279 struct, the C++ compiler will now offer fix-it hints showing how to 3280 use an accessor function to get at the field in question, if one 3281 exists. 3282$ gcc accessor.cc 3283accessor.cc: In function 'void test(foo*)': 3284accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context 3285 if (ptr->m_ratio >= 0.5) 3286 ^~~~~~~ 3287accessor.cc:7:10: note: declared private here 3288 double m_ratio; 3289 ^~~~~~~ 3290accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double 3291 foo::get_ratio() const' 3292 if (ptr->m_ratio >= 0.5) 3293 ^~~~~~~ 3294 get_ratio() 3295 3296 * The C++ compiler can now give you a hint if you use a macro before 3297 it was defined (e.g. if you mess up the order of your #include 3298 directives): 3299$ gcc ordering.cc 3300ordering.cc:2:24: error: expected ';' at end of member declaration 3301 virtual void clone() const OVERRIDE { } 3302 ^~~~~ 3303 ; 3304ordering.cc:2:30: error: 'OVERRIDE' does not name a type 3305 virtual void clone() const OVERRIDE { } 3306 ^~~~~~~~ 3307ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined 3308In file included from ordering.cc:5: 3309c++11-compat.h:2: note: it was later defined here 3310 #define OVERRIDE override 3311 3312 3313 * The -Wold-style-cast diagnostic can now emit fix-it hints telling 3314 you when you can use a static_cast, const_cast, or 3315 reinterpret_cast. 3316$ gcc -c old-style-cast-fixits.cc -Wold-style-cast 3317old-style-cast-fixits.cc: In function 'void test(void*)': 3318old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [ 3319-Wold-style-cast] 3320 foo *f = (foo *)ptr; 3321 ^~~ 3322 ---------- 3323 static_cast<foo *> (ptr) 3324 3325 * When reporting on problems within extern "C" linkage 3326 specifications, the C++ compiler will now display the location of 3327 the start of the extern "C". 3328$ gcc -c extern-c.cc 3329extern-c.cc:3:1: error: template with C linkage 3330 template <typename T> void test (void); 3331 ^~~~~~~~ 3332In file included from extern-c.cc:1: 3333unclosed.h:1:1: note: 'extern "C"' linkage started here 3334 extern "C" { 3335 ^~~~~~~~~~ 3336extern-c.cc:3:39: error: expected '}' at end of input 3337 template <typename T> void test (void); 3338 ^ 3339In file included from extern-c.cc:1: 3340unclosed.h:1:12: note: to match this '{' 3341 extern "C" { 3342 ^ 3343 3344 * When reporting on mismatching template types, the C++ compiler will 3345 now use color to highlight the mismatching parts of the template, 3346 and will elide the parameters that are common between two 3347 mismatching templates, printing [...] instead: 3348$ gcc templates.cc 3349templates.cc: In function 'void test()': 3350templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 3351e>' to 'vector<int>' 3352 fn_1(vector<double> ()); 3353 ^~~~~~~~~~~~~~~~~ 3354templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...] 3355,double>' to 'map<[...],int>' 3356 fn_2(map<int, double>()); 3357 ^~~~~~~~~~~~~~~~~~ 3358 3359 Those [...] elided parameters can be seen using -fno-elide-type: 3360$ gcc templates.cc -fno-elide-type 3361templates.cc: In function 'void test()': 3362templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 3363e>' to 'vector<int>' 3364 fn_1(vector<double> ()); 3365 ^~~~~~~~~~~~~~~~~ 3366templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d 3367ouble>' to 'map<int,int>' 3368 fn_2(map<int, double>()); 3369 ^~~~~~~~~~~~~~~~~~ 3370 3371 The C++ compiler has also gained an option 3372 -fdiagnostics-show-template-tree which visualizes such mismatching 3373 templates in a hierarchical form: 3374$ gcc templates-2.cc -fdiagnostics-show-template-tree 3375templates-2.cc: In function 'void test()': 3376templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 3377ble>' to 'vector<int>' 3378 vector< 3379 [double != int]> 3380 fn_1(vector<double> ()); 3381 ^~~~~~~~~~~~~~~~~ 3382templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 3383ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m 3384ap<[...],vector<float>>,vector<float>>' 3385 map< 3386 map< 3387 [...], 3388 vector< 3389 [double != float]>>, 3390 vector< 3391 [double != float]>> 3392 fn_2(map<map<int, vector<double>>, vector<double>> ()); 3393 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 3394 3395 which again works with -fno-elide-type: 3396$ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type 3397templates-2.cc: In function 'void test()': 3398templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 3399ble>' to 'vector<int>' 3400 vector< 3401 [double != int]> 3402 fn_1(vector<double> ()); 3403 ^~~~~~~~~~~~~~~~~ 3404templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 3405ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map 3406<int,vector<float>>,vector<float>>' 3407 map< 3408 map< 3409 int, 3410 vector< 3411 [double != float]>>, 3412 vector< 3413 [double != float]>> 3414 fn_2(map<map<int, vector<double>>, vector<double>> ()); 3415 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 3416 3417 * Flowing off the end of a non-void function is considered 3418 unreachable and may be subject to optimization on that basis. As a 3419 result of this change, -Wreturn-type warnings are enabled by 3420 default for C++. 3421 3422 Runtime Library (libstdc++) 3423 3424 * Improved experimental support for C++17, including the following 3425 features: 3426 + Deduction guides to support class template argument deduction. 3427 + std::filesystem implementation. 3428 + std::char_traits<char> and std::char_traits<wchar_t> are 3429 usable in constant expressions. 3430 + std::to_chars and std::from_chars (for integers only, not for 3431 floating point types). 3432 * Experimental support for C++2a: std::to_address (thanks to Glen 3433 Fernandes) and std::endian. 3434 * On GNU/Linux, std::random_device::entropy() accesses the kernel's 3435 entropy count for the random device, if known (thanks to Xi 3436 Ruoyao). 3437 * Support for std::experimental::source_location. 3438 * AddressSanitizer integration for std::vector, detecting 3439 out-of-range accesses to the unused capacity of a vector. 3440 * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the 3441 Mathematical Special Functions. 3442 3443 Fortran 3444 3445 * The main version of libfortran has been changed to 5. 3446 * Parameterized derived types, a major feature of Fortran 2003, have 3447 been implemented. 3448 * Partial support is provided for Fortran 2018 teams, which are 3449 hierarchical subsets of images that execute independently of other 3450 image subsets. 3451 * The maximum rank for arrays has been increased to 15, conforming to 3452 the Fortran 2008 standard. 3453 * Transformational intrinsics are now fully supported in 3454 initialization expressions. 3455 * New flag -fc-prototypes to write C prototypes for BIND(C) 3456 procedures and variables. 3457 * If -fmax-stack-var-size is honored if given together with -Ofast, 3458 -fstack-arrays is no longer set in that case. 3459 * New options -fdefault-real-16 and -fdefault-real-10 to control the 3460 default kind of REAL variables. 3461 * A warning is now issued if an array subscript inside a DO loop 3462 could lead to an out-of-bounds-access. The new option 3463 -Wdo-subscript, enabled by -Wextra, warns about this even if the 3464 compiler can not prove that the code will be executed. 3465 * The Fortran front end now attempts to interchange loops if it is 3466 deemed profitable. So far, this is restricted to FORALL and DO 3467 CONCURRENT statements with multiple indices. This behavior be 3468 controlled with the new flag -ffrontend-loop-interchange, which is 3469 enabled with optimization by default. The 3470 -Wfrontend-loop-interchange option warns about such occurrences. 3471 * When an actual argument contains too few elements for a dummy 3472 argument, an error is now issued. The -std=legacy option can be 3473 used to still compile such code. 3474 * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit 3475 integers, making records larger than 2GiB possible. 3476 * The GFORTRAN_DEFAULT_RECL environment variable no longer has any 3477 effect. The record length for preconnected units is now larger than 3478 any practical limit, same as for sequential access units opened 3479 without an explicit RECL= specifier. 3480 * Character variables longer than HUGE(0) elements are now possible 3481 on 64-bit targets. Note that this changes the procedure call ABI 3482 for all procedures with character arguments on 64-bit targets, as 3483 the type of the hidden character length argument has changed. The 3484 hidden character length argument is now of type INTEGER(C_SIZE_T). 3485 * Partial support is provided for Fortran 2018 teams, which are 3486 hierarchical subsets of images that execute independently of other 3487 image subsets. 3488 3489 Go 3490 3491 * GCC 8 provides a complete implementation of the Go 1.10.1 user 3492 packages. 3493 * The garbage collector is now fully concurrent. As before, values 3494 stored on the stack are scanned conservatively, but value stored in 3495 the heap are scanned precisely. 3496 * Escape analysis is fully implemented and enabled by default in the 3497 Go front end. This significantly reduces the number of heap 3498 allocations by allocating values on the stack instead. 3499 3500libgccjit 3501 3502 The libgccjit API gained four new entry points: 3503 * [18]gcc_jit_type_get_vector and 3504 * [19]gcc_jit_context_new_rvalue_from_vector for working with 3505 vectors, 3506 * [20]gcc_jit_type_get_aligned 3507 * [21]gcc_jit_function_get_address 3508 3509 The C code generated by [22]gcc_jit_context_dump_reproducer_to_file is 3510 now easier-to-read. 3511 3512New Targets and Target Specific Improvements 3513 3514 AArch64 3515 3516 * The Armv8.4-A architecture is now supported. It can be used by 3517 specifying the -march=armv8.4-a option. 3518 * The Dot Product instructions are now supported as an optional 3519 extension to the Armv8.2-A architecture and newer and are mandatory 3520 on Armv8.4-A. The extension can be used by specifying the +dotprod 3521 architecture extension. E.g. -march=armv8.2-a+dotprod. 3522 * The Armv8-A +crypto extension has now been split into two 3523 extensions for finer grained control: 3524 + +aes which contains the Armv8-A AES crytographic instructions. 3525 + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic 3526 instructions. 3527 Using +crypto will now enable these two extensions. 3528 * New Armv8.4-A FP16 Floating Point Multiplication Variant 3529 instructions have been added. These instructions are mandatory in 3530 Armv8.4-A but available as an optional extension to Armv8.2-A and 3531 Armv8.3-A. The new extension can be used by specifying the +fp16fml 3532 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 3533 the instructions can be enabled by specifying +fp16. 3534 * New cryptographic instructions have been added as optional 3535 extensions to Armv8.2-A and newer. These instructions can be 3536 enabled with: 3537 + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This 3538 implies +sha2. 3539 + +sm4 New SM3 and SM4 instructions from Armv8.4-A. 3540 * The Scalable Vector Extension (SVE) is now supported as an optional 3541 extension to the Armv8.2-A architecture and newer. This support 3542 includes automatic vectorization with SVE instructions, but it does 3543 not yet include the SVE Arm C Language Extensions (ACLE). It can be 3544 enabled by specifying the +sve architecture extension (for example, 3545 -march=armv8.2-a+sve). By default, the generated code works with 3546 all vector lengths, but it can be made specific to N-bit vectors 3547 using -msve-vector-bits=N. 3548 * Support has been added for the following processors (GCC 3549 identifiers in parentheses): 3550 + Arm Cortex-A75 (cortex-a75). 3551 + Arm Cortex-A55 (cortex-a55). 3552 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 3553 (cortex-a75.cortex-a55). 3554 The GCC identifiers can be used as arguments to the -mcpu or -mtune 3555 options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as 3556 arguments to the equivalent target attributes and pragmas. 3557 3558 ARC 3559 3560 * Added support for: 3561 + Fast interrupts. 3562 + Naked functions. 3563 + aux variable attributes. 3564 + uncached type qualifier. 3565 + Secure functions via sjli instruction. 3566 * New exception handling implementation. 3567 * Revamped trampoline implementation. 3568 * Refactored small data feature implementation, controlled via -G 3569 command line option. 3570 * New support for reduced register set ARC architecture 3571 configurations, controlled via -mrf16 command line option. 3572 * Refurbished and improved support for zero overhead loops. 3573 Introduced -mlpc-width command line option to control the width of 3574 lp_count register. 3575 3576 ARM 3577 3578 * The -mfpu option now takes a new option setting of -mfpu=auto. When 3579 set to this the floating-point and SIMD settings are derived from 3580 the settings of the -mcpu or -march options. The internal CPU 3581 configurations have been updated with information about the 3582 permitted floating-point configurations supported. See the user 3583 guide for further information about the extended option syntax for 3584 controlling architectural extensions via the -march option. 3585 -mfpu=auto is now the default setting unless the compiler has been 3586 configured with an explicit --with-fpu option. 3587 * The -march and -mcpu options now accept optional extensions to the 3588 architecture or CPU option, allowing the user to enable or disable 3589 any such extensions supported by that architecture or CPU such as 3590 (but not limited to) floating-point and AdvancedSIMD. For example: 3591 the option -mcpu=cortex-a53+nofp will generate code for the 3592 Cortex-A53 processor with no floating-point support. This, in 3593 combination with the new -mfpu=auto option, provides a 3594 straightforward way of specifying a valid build target through a 3595 single -mcpu or -march option. The -mtune option accepts the same 3596 arguments as -mcpu but only the CPU name has an effect on tuning. 3597 The architecture extensions do not have any effect. For details of 3598 what extensions a particular architecture or CPU option supports 3599 please refer to the [23]documentation. 3600 * The -mstructure-size-boundary option has been deprecated and will 3601 be removed in a future release. 3602 * The default link behavior for Armv6 and Armv7-R targets has been 3603 changed to produce BE8 format when generating big-endian images. A 3604 new flag -mbe32 can be used to force the linker to produce legacy 3605 BE32 format images. There is no change of behavior for Armv6-M and 3606 other Armv7 or later targets: these already defaulted to BE8 3607 format. This change brings GCC into alignment with other compilers 3608 for the ARM architecture. 3609 * The Armv8-R architecture is now supported. It can be used by 3610 specifying the -march=armv8-r option. 3611 * The Armv8.3-A architecture is now supported. It can be used by 3612 specifying the -march=armv8.3-a option. 3613 * The Armv8.4-A architecture is now supported. It can be used by 3614 specifying the -march=armv8.4-a option. 3615 * The Dot Product instructions are now supported as an optional 3616 extension to the Armv8.2-A architecture and newer and are mandatory 3617 on Armv8.4-A. The extension can be used by specifying the +dotprod 3618 architecture extension. E.g. -march=armv8.2-a+dotprod. 3619 * Support for setting extensions and architectures using the GCC 3620 target pragma and attribute has been added. It can be used by 3621 specifying #pragma GCC target ("arch=..."), #pragma GCC target 3622 ("+extension"), __attribute__((target("arch=..."))) or 3623 __attribute__((target("+extension"))). 3624 * New Armv8.4-A FP16 Floating Point Multiplication Variant 3625 instructions have been added. These instructions are mandatory in 3626 Armv8.4-A but available as an optional extension to Armv8.2-A and 3627 Armv8.3-A. The new extension can be used by specifying the +fp16fml 3628 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 3629 the instructions can be enabled by specifying +fp16. 3630 * Support has been added for the following processors (GCC 3631 identifiers in parentheses): 3632 + Arm Cortex-A75 (cortex-a75). 3633 + Arm Cortex-A55 (cortex-a55). 3634 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 3635 (cortex-a75.cortex-a55). 3636 + Arm Cortex-R52 for Armv8-R (cortex-r52). 3637 The GCC identifiers can be used as arguments to the -mcpu or -mtune 3638 options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as 3639 arguments to the equivalent target attributes and pragmas. 3640 3641 AVR 3642 3643 * The AVR port now supports the following XMEGA-like devices: 3644 3645 ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417, 3646 ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617, 3647 ATtiny3214, ATtiny3216, ATtiny3217 3648 The new devices are listed under [24]-mmcu=avrxmega3. 3649 + These devices see flash memory in the RAM address space, so 3650 that features like PROGMEM and __flash are not needed any more 3651 (as opposed to other AVR families for which read-only data 3652 will be located in RAM except special, non-standard features 3653 are used to locate and access such data). This requires that 3654 the compiler is used with Binutils 2.29 or newer so that 3655 [25]read-only data will be located in flash memory. 3656 + A new command-line option -mshort-calls is supported. This 3657 option is used internally for multilib selection of the 3658 avrxmega3 variants. It is not an optimization option. Do not 3659 set it by hand. 3660 * The compiler now generates [26]efficient interrupt service routine 3661 (ISR) prologues and epilogues. This is achieved by using the new 3662 [27]AVR pseudo instruction __gcc_isr which is supported and 3663 resolved by the GNU assembler. 3664 + As the __gcc_isr pseudo-instruction will be resolved by the 3665 assembler, inline assembly is transparent to the process. This 3666 means that when inline assembly uses an instruction like INC 3667 that clobbers the condition code, then the assembler will 3668 detect this and generate an appropriate ISR prologue / 3669 epilogue chunk to save / restore SREG as needed. 3670 + A new command-line option -mno-gas-isr-prologues disables the 3671 generation of the __gcc_isr pseudo instruction. Any non-naked 3672 ISR will save and restore SREG, tmp_reg and zero_reg, no 3673 matter whether the respective register is clobbered or used. 3674 + The feature is turned on per default for all optimization 3675 levels except for -O0 and -Og. It is explicitly enabled by 3676 means of option -mgas-isr-prologues. 3677 + Support has been added for a new [28]AVR function attribute 3678 no_gccisr. It can be used to disable __gcc_isr pseudo 3679 instruction generation for individual ISRs. 3680 + This optimization is only available if GCC is configured with 3681 GNU Binutils 2.29 or newer; or at least with a version of 3682 Binutils that implements feature [29]PR21683. 3683 * The compiler no more saves / restores registers in main; the effect 3684 is the same as if attribute OS_task was specified for main. This 3685 optimization can be switched off by the new command-line option 3686 -mno-main-is-OS_task. 3687 3688 IA-32/x86-64 3689 3690 * The x86 port now supports the naked function attribute. 3691 * Better tuning for znver1 and Intel Core based CPUs. 3692 * Vectorization cost metrics has been reworked leading to significant 3693 improvements on some benchmarks. 3694 * GCC now supports the Intel CPU named Cannonlake through 3695 -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA 3696 and SHA ISA extensions. 3697 * GCC now supports the Intel CPU named Icelake through 3698 -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES, 3699 AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ 3700 ISA extensions. 3701 * GCC now supports the Intel Control-flow Enforcement Technology 3702 (CET) extension through -fcf-protection option. 3703 3704 NDS32 3705 3706 * New command-line options -mext-perf, -mext-perf2, and -mext-string 3707 have been added for performance extension instructions. 3708 3709 Nios II 3710 3711 * The Nios II back end has been improved to generate better-optimized 3712 code. Changes include switching to LRA, more accurate cost models, 3713 and more compact code for addressing static variables. 3714 * New command-line options -mgprel-sec= and -mr0rel-sec= have been 3715 added. 3716 * The stack-smashing protection options are now enabled on Nios II. 3717 3718 PA-RISC 3719 3720 * The default call ABI on 32-bit linux has been changed from callee 3721 copies to caller copies. This affects objects larger than eight 3722 bytes passed by value. The goal is to improve compatibility with 3723 x86 and resolve issues with OpenMP. 3724 * Other PA-RISC targets are unchanged. 3725 3726 PowerPC / PowerPC64 / RS6000 3727 3728 * The PowerPC SPE support is split off to a separate powerpcspe port. 3729 The separate port is deprecated and might be removed in a future 3730 release. 3731 * The Paired Single support (as used on some PPC750 CPUs, -mpaired, 3732 powerpc*-*-linux*paired*) is deprecated and will be removed in a 3733 future release. 3734 * The Xilinx floating point support (-mxilinx-fpu, 3735 powerpc-xilinx-eabi*) is deprecated and will be removed in a future 3736 release. 3737 * Support for using big-endian AltiVec intrinsics on a little-endian 3738 target (-maltivec=be) is deprecated and will be removed in a future 3739 release. 3740 3741 Tile 3742 3743 * The TILE-Gx port is deprecated and will be removed in a future 3744 release. 3745 3746Operating Systems 3747 3748 Windows 3749 3750 * GCC on Microsoft Windows can now be configured via 3751 --enable-mingw-wildcard or --disable-mingw-wildcard to force a 3752 specific behavior for GCC itself with regards to supporting the 3753 wildcard character. Prior versions of GCC would follow the 3754 configuration of the MinGW runtime. This behavior can still be 3755 obtained by not using the above options or by using 3756 --enable-mingw-wildcard=platform. 3757 3758Improvements for plugin authors 3759 3760 * Plugins can now register a callback hook for when comments are 3761 encountered by the C and C++ compilers, e.g. allowing for plugins 3762 to handle documentation markup in code comments. 3763 * The gdbinit support script for debugging GCC now has a 3764 break-on-diagnostic command, providing an easy way to trigger a 3765 breakpoint whenever a diagnostic is emitted. 3766 * The API for creating fix-it hints now supports newlines, and for 3767 emitting mutually incompatible fix-it hints for one diagnostic. 3768 3769GCC 8.1 3770 3771 This is the [30]list of problem reports (PRs) from GCC's bug tracking 3772 system that are known to be fixed in the 8.1 release. This list might 3773 not be complete (that is, it is possible that some PRs that have been 3774 fixed are not listed here). 3775 3776GCC 8.2 3777 3778 This is the [31]list of problem reports (PRs) from GCC's bug tracking 3779 system that are known to be fixed in the 8.2 release. This list might 3780 not be complete (that is, it is possible that some PRs that have been 3781 fixed are not listed here). 3782 3783 General Improvements 3784 3785 * Fixed LTO link-time performance problems caused by an overflow in 3786 the partitioning algorithm while building large binaries. 3787 3788 Language Specific Changes 3789 3790 C++ 3791 3792 GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or 3793 returning of classes with a deleted copy constructor and defaulted 3794 trivial move constructor (bug [32]c++/86094). GCC 8.2 introduces 3795 -fabi-version=13 and makes it the default, ABI incompatibilities 3796 between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [33]C++ 3797 changes for more details. 3798 3799 Target Specific Changes 3800 3801 IA-32/x86-64 3802 3803 * -mtune=native performance regression [34]PR84413 on Intel Skylake 3804 processors has been fixed. 3805 3806GCC 8.3 3807 3808 This is the [35]list of problem reports (PRs) from GCC's bug tracking 3809 system that are known to be fixed in the 8.3 release. This list might 3810 not be complete (that is, it is possible that some PRs that have been 3811 fixed are not listed here). 3812 3813 Windows 3814 3815 * A C++ Microsoft ABI bitfield layout bug, [36]PR87137 has been 3816 fixed. A non-field declaration could cause the current bitfield 3817 allocation unit to be completed, incorrectly placing a following 3818 bitfield into a new allocation unit. The Microsoft ABI is selected 3819 for: 3820 + Mingw targets 3821 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 3822 option is specified, or __attribute__((ms_struct)) is used 3823 + SuperH targets when the -mhitachi option is specified, or 3824 __attribute__((renesas)) is used 3825 GCC 8 introduced additional cases of this defect, but rather than 3826 resolve only those regressions, we decided to resolve all the cases 3827 of this defect in single change. 3828 3829GCC 8.4 3830 3831 This is the [37]list of problem reports (PRs) from GCC's bug tracking 3832 system that are known to be fixed in the 8.4 release. This list might 3833 not be complete (that is, it is possible that some PRs that have been 3834 fixed are not listed here). 3835 3836GCC 8.5 3837 3838 This is the [38]list of problem reports (PRs) from GCC's bug tracking 3839 system that are known to be fixed in the 8.5 release. This list might 3840 not be complete (that is, it is possible that some PRs that have been 3841 fixed are not listed here). 3842 3843 Target Specific Changes 3844 3845 AArch64 3846 3847 * The option -moutline-atomics has been added to aid deployment of 3848 the Large System Extensions (LSE) on GNU/Linux systems built with a 3849 baseline architecture targeting Armv8-A. When the option is 3850 specified code is emitted to detect the presence of LSE 3851 instructions at run time and use them for standard atomic 3852 operations. For more information please refer to the documentation. 3853 3854 3855 For questions related to the use of GCC, please consult these web 3856 pages and the [39]GCC manuals. If that fails, the 3857 [40]gcc-help@gcc.gnu.org mailing list might help. Comments on these 3858 web pages and the development of GCC are welcome on our developer 3859 list at [41]gcc@gcc.gnu.org. All of [42]our lists have public 3860 archives. 3861 3862 Copyright (C) [43]Free Software Foundation, Inc. Verbatim copying and 3863 distribution of this entire article is permitted in any medium, 3864 provided this notice is preserved. 3865 3866 These pages are [44]maintained by the GCC team. Last modified 3867 2021-07-28[45]. 3868 3869References 3870 3871 1. http://gcc.gnu.org/gcc-8/porting_to.html 3872 2. http://gcc.gnu.org/onlinedocs/index.html#current 3873 3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html 3874 4. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros 3875 5. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 3876 6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944 3877 7. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned 3878 8. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute 3879 9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 3880 10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned 3881 11. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds 3882 12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict 3883 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow 3884 14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation 3885 15. https://gcc.gnu.org/PR86094 3886 16. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess 3887 17. http://gcc.gnu.org/projects/cxx-status.html#cxx2a 3888 18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector 3889 19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector 3890 20. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned 3891 21. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address 3892 22. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file 3893 23. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options 3894 24. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html 3895 25. https://sourceware.org/PR21472 3896 26. https://gcc.gnu.org/PR20296 3897 27. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html 3898 28. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html 3899 29. https://sourceware.org/PR21683 3900 30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0 3901 31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2 3902 32. https://gcc.gnu.org/PR86094 3903 33. http://gcc.gnu.org/gcc-8/changes.html#cxx 3904 34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413 3905 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3 3906 36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 3907 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4 3908 38. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.5 3909 39. https://gcc.gnu.org/onlinedocs/ 3910 40. mailto:gcc-help@gcc.gnu.org 3911 41. mailto:gcc@gcc.gnu.org 3912 42. https://gcc.gnu.org/lists.html 3913 43. https://www.fsf.org/ 3914 44. https://gcc.gnu.org/about.html 3915 45. http://validator.w3.org/check/referer 3916====================================================================== 3917http://gcc.gnu.org/gcc-7/index.html 3918 3919 GCC 7 Release Series 3920 3921 Nov 14, 2019 3922 3923 The [1]GNU project and the GCC developers are pleased to announce the 3924 release of GCC 7.5. 3925 3926 This release is a bug-fix release, containing fixes for regressions in 3927 GCC 7.4 relative to previous releases of GCC. 3928 3929Release History 3930 3931 GCC 7.5 3932 Nov 14, 2019 ([2]changes, [3]documentation) 3933 3934 GCC 7.4 3935 Dec 6, 2018 ([4]changes, [5]documentation) 3936 3937 GCC 7.3 3938 Jan 25, 2018 ([6]changes, [7]documentation) 3939 3940 GCC 7.2 3941 Aug 14, 2017 ([8]changes, [9]documentation) 3942 3943 GCC 7.1 3944 May 2, 2017 ([10]changes, [11]documentation) 3945 3946References and Acknowledgements 3947 3948 GCC used to stand for the GNU C Compiler, but since the compiler 3949 supports several other languages aside from C, it now stands for the 3950 GNU Compiler Collection. 3951 3952 A list of [12]successful builds is updated as new information becomes 3953 available. 3954 3955 The GCC developers would like to thank the numerous people that have 3956 contributed new features, improvements, bug fixes, and other changes as 3957 well as test results to GCC. This [13]amazing group of volunteers is 3958 what makes GCC successful. 3959 3960 For additional information about GCC please refer to the [14]GCC 3961 project web site or contact the [15]GCC development mailing list. 3962 3963 To obtain GCC please use [16]our mirror sites or [17]our version 3964 control system. 3965 3966 3967 For questions related to the use of GCC, please consult these web 3968 pages and the [18]GCC manuals. If that fails, the 3969 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 3970 web pages and the development of GCC are welcome on our developer 3971 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 3972 archives. 3973 3974 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 3975 distribution of this entire article is permitted in any medium, 3976 provided this notice is preserved. 3977 3978 These pages are [23]maintained by the GCC team. Last modified 3979 2021-07-28[24]. 3980 3981References 3982 3983 1. http://www.gnu.org/ 3984 2. http://gcc.gnu.org/gcc-7/changes.html 3985 3. http://gcc.gnu.org/onlinedocs/7.5.0/ 3986 4. http://gcc.gnu.org/gcc-7/changes.html 3987 5. http://gcc.gnu.org/onlinedocs/7.4.0/ 3988 6. http://gcc.gnu.org/gcc-7/changes.html 3989 7. http://gcc.gnu.org/onlinedocs/7.3.0/ 3990 8. http://gcc.gnu.org/gcc-7/changes.html 3991 9. http://gcc.gnu.org/onlinedocs/7.2.0/ 3992 10. http://gcc.gnu.org/gcc-7/changes.html 3993 11. http://gcc.gnu.org/onlinedocs/7.1.0/ 3994 12. http://gcc.gnu.org/gcc-7/buildstat.html 3995 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 3996 14. http://gcc.gnu.org/index.html 3997 15. mailto:gcc@gcc.gnu.org 3998 16. http://gcc.gnu.org/mirrors.html 3999 17. http://gcc.gnu.org/git.html 4000 18. https://gcc.gnu.org/onlinedocs/ 4001 19. mailto:gcc-help@gcc.gnu.org 4002 20. mailto:gcc@gcc.gnu.org 4003 21. https://gcc.gnu.org/lists.html 4004 22. https://www.fsf.org/ 4005 23. https://gcc.gnu.org/about.html 4006 24. http://validator.w3.org/check/referer 4007====================================================================== 4008http://gcc.gnu.org/gcc-7/changes.html 4009 4010 GCC 7 Release Series 4011 Changes, New Features, and Fixes 4012 4013 This page is a brief summary of some of the huge number of improvements 4014 in GCC 7. For more information, see the [1]Porting to GCC 7 page and 4015 the [2]full GCC documentation. 4016 4017Caveats 4018 4019 * GCC now uses [3]LRA (a new local register allocator) by default for 4020 new targets. 4021 * The non-standard C++0x type traits has_trivial_default_constructor, 4022 has_trivial_copy_constructor and has_trivial_copy_assign have been 4023 removed. 4024 * The libstdc++ [4]Profile Mode has been deprecated and will be 4025 removed in a future version. 4026 * The Cilk+ extensions to the C and C++ languages have been 4027 deprecated. 4028 * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that 4029 affects conformance to the procedure call standard (AAPCS) has been 4030 fixed. The bug affects some C++ code where class objects are passed 4031 by value to functions and could result in incorrect or inconsistent 4032 code being generated. This is an ABI change. If the option -Wpsabi 4033 is enabled (on by default) the compiler will emit a diagnostic note 4034 for code that might be affected. 4035 4036General Optimizer Improvements 4037 4038 * GCC 7 can determine the return value or range of return values of 4039 some calls to the sprintf family of functions and make it available 4040 to other optimization passes. Some calls to the snprintf function 4041 with a zero size argument can be folded into constants. This 4042 optimization is included in -O1 and can be selectively controlled 4043 by the -fprintf-return-value option. 4044 * A new store merging pass has been added. It merges constant stores 4045 to adjacent memory locations into fewer, wider, stores. It is 4046 enabled by the -fstore-merging option and at the -O2 optimization 4047 level or higher (and -Os). 4048 * A new code hoisting optimization has been added to the partial 4049 redundancy elimination pass. It attempts to move evaluation of 4050 expressions executed on all paths to the function exit as early as 4051 possible. This primarily helps improve code size, but can improve 4052 the speed of the generated code as well. It is enabled by the 4053 -fcode-hoisting option and at the -O2 optimization level or higher 4054 (and -Os). 4055 * A new interprocedural bitwise constant propagation optimization has 4056 been added, which propagates knowledge about which bits of 4057 variables are known to be zero (including pointer alignment 4058 information) across the call graph. It is enabled by the 4059 -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled 4060 at the -O2 optimization level and higher (and -Os). This 4061 optimization supersedes interprocedural alignment propagation of 4062 GCC 6, and therefore the option -fipa-cp-alignment is now 4063 deprecated and ignored. 4064 * A new interprocedural value range propagation optimization has been 4065 added, which propagates integral range information across the call 4066 graph when variable values can be proven to be within those ranges. 4067 It is enabled by the -fipa-vrp option and at the -O2 optimization 4068 level and higher (and -Os). 4069 * A new loop splitting optimization pass has been added. Certain 4070 loops which contain a condition that is always true on one side of 4071 the iteration space and always false on the other are split into 4072 two loops, such that each of the two new loops iterates on just one 4073 side of the iteration space and the condition does not need to be 4074 checked inside of the loop. It is enabled by the -fsplit-loops 4075 option and at the -O3 optimization level or higher. 4076 * The shrink-wrapping optimization can now separate portions of 4077 prologues and epilogues to improve performance if some of the work 4078 done traditionally by prologues and epilogues is not needed on 4079 certain paths. This is controlled by the -fshrink-wrap-separate 4080 option, enabled by default. It requires target support, which is 4081 currently only implemented in the PowerPC and AArch64 ports. 4082 * AddressSanitizer gained a new sanitization option, 4083 -fsanitize-address-use-after-scope, which enables sanitization of 4084 variables whose address is taken and used after a scope where the 4085 variable is defined: 4086 4087int 4088main (int argc, char **argv) 4089{ 4090 char *ptr; 4091 { 4092 char my_char; 4093 ptr = &my_char; 4094 } 4095 4096 *ptr = 123; 4097 return *ptr; 4098} 4099 4100==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99 41010 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958 4102WRITE of size 1 at 0x7fffb8dba990 thread T0 4103 #0 0x4006d4 in main /tmp/use-after-scope-1.c:10 4104 #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290) 4105 #2 0x400739 in _start (/tmp/a.out+0x400739) 4106 4107Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame 4108 #0 0x40067f in main /tmp/use-after-scope-1.c:3 4109 4110 This frame has 1 object(s): 4111 [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable 4112 4113 The option is enabled by default with -fsanitize=address and 4114 disabled by default with -fsanitize=kernel-address. Compared to the 4115 LLVM compiler, where the option already exists, the implementation 4116 in the GCC compiler has some improvements and advantages: 4117 + Complex uses of gotos and case labels are properly handled and 4118 should not report any false positive or false negatives. 4119 + C++ temporaries are sanitized. 4120 + Sanitization can handle invalid memory stores that are 4121 optimized out by the LLVM compiler when optimization is 4122 enabled. 4123 * The -fsanitize=signed-integer-overflow suboption of the 4124 UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even 4125 on arithmetic operations with generic vectors. 4126 * Version 5 of the [6]DWARF debugging information standard is 4127 supported through the -gdwarf-5 option. The DWARF version 4 4128 debugging information remains the default until consumers of 4129 debugging information are adjusted. 4130 4131New Languages and Language specific improvements 4132 4133 OpenACC support in C, C++, and Fortran continues to be maintained and 4134 improved. See the [7]OpenACC and [8]Offloading wiki pages for further 4135 information. 4136 4137 Ada 4138 4139 * On mainstream native platforms, Ada programs no longer require the 4140 stack to be made executable in order to run properly. 4141 4142 BRIG (HSAIL) 4143 4144 Support for processing BRIG 1.0 files was added in this release. BRIG 4145 is a binary format for HSAIL (Heterogeneous System Architecture 4146 Intermediate Language). The BRIG front end can be used for implementing 4147 HSAIL "finalizers" (compilation of HSAIL to a native ISA) for 4148 GCC-supported targets. An implementation of an HSAIL runtime library, 4149 libhsail-rt is also included. 4150 4151 C family 4152 4153 * New command-line options have been added for the C and C++ 4154 compilers: 4155 + -Wimplicit-fallthrough warns when a switch case falls through. 4156 This warning has five different levels. The compiler is able 4157 to parse a wide range of fallthrough comments, depending on 4158 the level. It also handles control-flow statements, such as 4159 ifs. It's possible to suppress the warning by either adding a 4160 fallthrough comment, or by using a null statement: 4161 __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]]; 4162 (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning 4163 is enabled by -Wextra. 4164 + -Wpointer-compare warns when a pointer is compared with a zero 4165 character constant. Such code is now invalid in C++11 and GCC 4166 rejects it. This warning is enabled by default. 4167 + -Wduplicated-branches warns when an if-else has identical 4168 branches. 4169 + -Wrestrict warns when an argument passed to a 4170 restrict-qualified parameter aliases with another argument. 4171 + -Wmemset-elt-size warns for memset calls, when the first 4172 argument references an array, and the third argument is a 4173 number equal to the number of elements of the array, but not 4174 the size of the array. This warning is enabled by -Wall. 4175 + -Wint-in-bool-context warns about suspicious uses of integer 4176 values where boolean values are expected. This warning is 4177 enabled by -Wall. 4178 + -Wswitch-unreachable warns when a switch statement has 4179 statements between the controlling expression and the first 4180 case label which will never be executed. This warning is 4181 enabled by default. 4182 + -Wexpansion-to-defined warns when defined is used outside #if. 4183 This warning is enabled by -Wextra or -Wpedantic. 4184 + -Wregister warns about uses of the register storage specifier. 4185 In C++17 this keyword has been removed and for C++17 this is a 4186 pedantic warning enabled by default. The warning is not 4187 emitted for the GNU Explicit Register Variables extension. 4188 + -Wvla-larger-than=N warns about unbounded uses of 4189 variable-length arrays, and about bounded uses of 4190 variable-length arrays whose bound can be larger than N bytes. 4191 + -Wduplicate-decl-specifier warns when a declaration has 4192 duplicate const, volatile, restrict or _Atomic specifier. This 4193 warning is enabled by -Wall. 4194 * GCC 6's C and C++ front ends were able to offer suggestions for 4195 misspelled field names: 4196 4197spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 4198you mean 'color'? 4199 return ptr->colour; 4200 ^~~~~~ 4201 4202 GCC 7 greatly expands the scope of these suggestions. Firstly, it 4203 adds fix-it hints to such suggestions: 4204 4205spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 4206you mean 'color'? 4207 return ptr->colour; 4208 ^~~~~~ 4209 color 4210 4211 The suggestions now cover many other things, such as misspelled 4212 function names: 4213 4214spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi 4215dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat 4216ion] 4217 gtk_widget_showall (w); 4218 ^~~~~~~~~~~~~~~~~~ 4219 gtk_widget_show_all 4220 4221 misspelled macro names and enum values: 4222 4223spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun 4224ction); did you mean 'MAX_ITEMS'? 4225 int array[MAX_ITEM]; 4226 ^~~~~~~~ 4227 MAX_ITEMS 4228 4229 misspelled type names: 4230 4231spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si 4232gned'? 4233 void test (singed char e); 4234 ^~~~~~ 4235 signed 4236 4237 and, in the C front end, named initializers: 4238 4239test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color 4240'? 4241 struct s test = { .colour = 3 }; 4242 ^~~~~~ 4243 color 4244 4245 * The preprocessor can now offer suggestions for misspelled 4246 directives, e.g.: 4247 4248test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif? 4249 #endfi 4250 ^~~~~ 4251 endif 4252 4253 * Warnings about format strings now underline the pertinent part of 4254 the string, and can offer suggested fixes. In some cases, the 4255 pertinent argument is underlined. 4256 4257test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume 4258nt 3 has type 'int' [-Wformat=] 4259 printf ("foo: %d bar: %s baz: %d", 100, i + j, 102); 4260 ~^ ~~~~~ 4261 %d 4262 4263 * The new -Wdangling-else command-line option has been split out of 4264 -Wparentheses and warns about dangling else. 4265 * The -Wshadow warning has been split into three variants. 4266 -Wshadow=global warns for any shadowing. This is the default when 4267 using -Wshadow without any argument. -Wshadow=local only warns for 4268 a local variable shadowing another local variable or parameter. 4269 -Wshadow=compatible-local only warns for a local variable shadowing 4270 another local variable or parameter whose type is compatible (in 4271 C++ compatible means that the type of the shadowing variable can be 4272 converted to that of the shadowed variable). 4273 The following example shows the different kinds of shadow warnings: 4274 4275enum operation { add, count }; 4276struct container { int nr; }; 4277 4278int 4279container_count (struct container c, int count) 4280{ 4281 int r = 0; 4282 for (int count = 0; count > 0; count--) 4283 { 4284 struct container count = c; 4285 r += count.nr; 4286 } 4287 return r; 4288} 4289 4290 -Wshadow=compatible-local will warn for the parameter being 4291 shadowed with the same type: 4292 4293warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow= 4294compatible-local] 4295 for (int count = 0; count > 0; count--) 4296 ^~~~~ 4297warn-test.c:5:42: note: shadowed declaration is here 4298 container_count (struct container c, int count) 4299 ^~~~~ 4300 4301 -Wshadow=local will warn for the above and for the shadowed 4302 declaration with incompatible type: 4303 4304warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws 4305hadow=local] 4306 struct container count = c; 4307 ^~~~~ 4308warn-test.c:8:12: note: shadowed declaration is here 4309 for (int count = 0; count > 0; count--) 4310 ^~~~~ 4311 4312 -Wshadow=global will warn for all of the above and the shadowing of 4313 the global declaration: 4314 4315warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [ 4316-Wshadow] 4317 container_count (struct container c, int count) 4318 ^~~~~ 4319warn-test.c:1:23: note: shadowed declaration is here 4320 enum operation { add, count }; 4321 ^~~~~ 4322 4323 * GCC 7 contains a number of enhancements that help detect buffer 4324 overflow and other forms of invalid memory accesses. 4325 + The -Walloc-size-larger-than=size option detects calls to 4326 standard and user-defined memory allocation functions 4327 decorated with attribute alloc_size whose argument exceeds the 4328 specified size (PTRDIFF_MAX by default). The option also 4329 detects arithmetic overflow in the computation of the size in 4330 two-argument allocation functions like calloc where the total 4331 size is the product of the two arguments. Since calls with an 4332 excessive size cannot succeed they are typically the result of 4333 programming errors. Such bugs have been known to be the source 4334 of security vulnerabilities and a target of exploits. 4335 -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall. 4336 For example, the following call to malloc incorrectly tries to 4337 avoid passing a negative argument to the function and instead 4338 ends up unconditionally invoking it with an argument less than 4339 or equal to zero. Since after conversion to the type of the 4340 argument of the function (size_t) a negative argument results 4341 in a value in excess of the maximum PTRDIFF_MAX the call is 4342 diagnosed. 4343 4344void* f (int n) 4345{ 4346 return malloc (n > 0 ? 0 : n); 4347} 4348 4349warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2 4350147483647 [-Walloc-size-larger-than=] 4351 4352 + The -Walloc-zero option detects calls to standard and 4353 user-defined memory allocation functions decorated with 4354 attribute alloc_size with a zero argument. -Walloc-zero is not 4355 included in either -Wall or -Wextra and must be explicitly 4356 enabled. 4357 + The -Walloca option detects all calls to the alloca function 4358 in the program. -Walloca is not included in either -Wall or 4359 -Wextra and must be explicitly enabled. 4360 + The -Walloca-larger-than=size option detects calls to the 4361 alloca function whose argument either may exceed the specified 4362 size, or that is not known to be sufficiently constrained to 4363 avoid exceeding it. -Walloca-larger-than is not included in 4364 either -Wall or -Wextra and must be explicitly enabled. 4365 For example, compiling the following snippet with 4366 -Walloca-larger-than=1024 results in a warning because even 4367 though the code appears to call alloca only with sizes of 1kb 4368 and less, since n is signed, a negative value would result in 4369 a call to the function well in excess of the limit. 4370 4371void f (int n) 4372{ 4373 char *d; 4374 if (n < 1025) 4375 d = alloca (n); 4376 else 4377 d = malloc (n); 4378 ... 4379} 4380 4381warning: argument to 'alloca may be too large due to conversion from 'int' to 'l 4382ong unsigned int' [-Walloca-larger-than=] 4383 4384 In contrast, a call to alloca that isn't bounded at all such 4385 as in the following function will elicit the warning below 4386 regardless of the size argument to the option. 4387 4388void f (size_t n) 4389{ 4390 char *d = alloca (n); 4391 ... 4392} 4393 4394warning: unbounded use of 'alloca' [-Walloca-larger-than=] 4395 4396 + The -Wformat-overflow=level option detects certain and likely 4397 buffer overflow in calls to the sprintf family of formatted 4398 output functions. Although the option is enabled even without 4399 optimization it works best with -O2 and higher. 4400 For example, in the following snippet the call to sprintf is 4401 diagnosed because even though its output has been constrained 4402 using the modulo operation it could result in as many as three 4403 bytes if mday were negative. The solution is to either 4404 allocate a larger buffer or make sure the argument is not 4405 negative, for example by changing mday's type to unsigned or 4406 by making the type of the second operand of the modulo 4407 expression unsigned: 100U. 4408 4409void* f (int mday) 4410{ 4411 char *buf = malloc (3); 4412 sprintf (buf, "%02i", mday % 100); 4413 return buf; 4414} 4415 4416warning: 'sprintf may write a terminating nul past the end of the destination [- 4417Wformat-overflow=] 4418note: 'sprintf' output between 3 and 4 bytes into a destination of size 3 4419 4420 + The -Wformat-truncation=level option detects certain and 4421 likely output truncation in calls to the snprintf family of 4422 formatted output functions. -Wformat-truncation=1 is included 4423 in -Wall and enabled without optimization but works best with 4424 -O2 and higher. 4425 For example, the following function attempts to format an 4426 integer between 0 and 255 in hexadecimal, including the 0x 4427 prefix, into a buffer of four characters. But since the 4428 function must always terminate output by the null character 4429 ('\0') such a buffer is only big enough to fit just one digit 4430 plus the prefix. Therefore the snprintf call is diagnosed. To 4431 avoid the warning either use a bigger buffer or handle the 4432 function's return value which indicates whether or not its 4433 output has been truncated. 4434 4435void f (unsigned x) 4436{ 4437 char d[4]; 4438 snprintf (d, sizeof d, "%#02x", x & 0xff); 4439 ... 4440} 4441 4442warning: 'snprintf' output may be truncated before the last format character [-W 4443format-truncation=] 4444note: 'snprintf' output between 3 and 5 bytes into a destination of size 4 4445 4446 + The -Wnonnull option has been enhanced to detect a broader set 4447 of cases of passing null pointers to functions that expect a 4448 non-null argument (those decorated with attribute nonnull). By 4449 taking advantage of optimizations the option can detect many 4450 more cases of the problem than in prior GCC versions. 4451 + The -Wstringop-overflow=type option detects buffer overflow in 4452 calls to string handling functions like memcpy and strcpy. The 4453 option relies on [9]Object Size Checking and has an effect 4454 similar to defining the _FORTIFY_SOURCE macro. 4455 -Wstringop-overflow=2 is enabled by default. 4456 For example, in the following snippet, because the call to 4457 strncat specifies a maximum that allows the function to write 4458 past the end of the destination, it is diagnosed. To correct 4459 the problem and avoid the overflow the function should be 4460 called with a size of at most sizeof d - strlen(d) - 1. 4461 4462void f (const char *fname) 4463{ 4464 char d[8]; 4465 strncpy (d, "/tmp/", sizeof d); 4466 strncat (d, fname, sizeof d); 4467 ... 4468} 4469 4470warning: specified bound 8 equals the size of the destination [-Wstringop-overfl 4471ow=] 4472 4473 * The <limits.h> header provided by GCC defines macros such as 4474 INT_WIDTH for the width in bits of integer types, if 4475 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 4476 included. The <stdint.h> header defines such macros as SIZE_WIDTH 4477 and INTMAX_WIDTH for the width of some standard typedef names for 4478 integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined 4479 before the header is included; note that GCC's implementation of 4480 this header is only used for freestanding compilations, not hosted 4481 compilations, on most systems. These macros come from ISO/IEC TS 4482 18661-1:2014. 4483 * The <float.h> header provided by GCC defines the macro 4484 CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if 4485 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 4486 included. This represents the number of decimal digits for which 4487 conversions between decimal character strings and binary formats, 4488 in both directions, are correctly rounded, and currently has the 4489 value of UINTMAX_MAX on all systems, reflecting that GCC's 4490 compile-time conversions are correctly rounded for any number of 4491 digits. 4492 * New __builtin_add_overflow_p, __builtin_sub_overflow_p, 4493 __builtin_mul_overflow_p built-in functions have been added. These 4494 work similarly to their siblings without the _p suffix, but do not 4495 actually store the result of the arithmetics anywhere, just return 4496 whether the operation would overflow. Calls to these built-ins with 4497 integer constant arguments evaluate to integer constants 4498 expressions. 4499 For example, in the following, c is assigned the result of a * b 4500 only if the multiplication does not overflow, otherwise it is 4501 assigned the value zero. The multiplication is performed at 4502 compile-time and without triggering a -Woverflow warning. 4503 4504enum { 4505 a = 12345678, 4506 b = 87654321, 4507 c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b 4508}; 4509 4510 C 4511 4512 * The C front end now supports type names _FloatN for floating-point 4513 types with IEEE interchange formats and _FloatNx for floating-point 4514 types with IEEE extended formats. These type names come from 4515 ISO/IEC TS 18661-3:2015. 4516 The set of types supported depends on the target for which GCC is 4517 configured. Most targets support _Float32, _Float32x and _Float64. 4518 _Float128 is supported on targets where IEEE binary128 encoding was 4519 already supported as long double or __float128. _Float64x is 4520 supported on targets where a type with either binary128 or Intel 4521 extended precision format is available. 4522 Constants with these types are supported using suffixes fN, FN, fNx 4523 and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are 4524 defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is 4525 defined before it is included. 4526 These new types are always distinct from each other and from float, 4527 double and long double, even if they have the same encoding. 4528 Complex types such as _Complex _Float128 are also supported. 4529 Type-generic built-in functions such as __builtin_isinf support the 4530 new types, and the following type-specific built-in functions have 4531 versions (suffixed fN or fNx) for the new types: 4532 __builtin_copysign, __builtin_fabs, __builtin_huge_val, 4533 __builtin_inf, __builtin_nan, __builtin_nans. 4534 * Compilation with -fopenmp is now compatible with the C11 _Atomic 4535 keyword. 4536 4537 C++ 4538 4539 * The C++ front end has experimental support for all of the current 4540 C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if 4541 constexpr, class template argument deduction, auto template 4542 parameters, and structured bindings. For a full list of new 4543 features, see [10]the C++ status page. 4544 * C++17 support for new of over-aligned types can be enabled in other 4545 modes with the -faligned-new flag. 4546 * The C++17 evaluation order requirements can be selected in other 4547 modes with the -fstrong-eval-order flag, or disabled in C++17 mode 4548 with -fno-strong-eval-order. 4549 * The default semantics of inherited constructors has changed in all 4550 modes, following [11]P0136. Essentially, overload resolution 4551 happens as if calling the inherited constructor directly, and the 4552 compiler fills in construction of the other bases and members as 4553 needed. Most uses should not need any changes. The old behavior can 4554 be restored with -fno-new-inheriting-ctors, or -fabi-version less 4555 than 11. 4556 * The resolution of DR 150 on matching of template template 4557 parameters, allowing default template arguments to make a template 4558 match a parameter, is currently enabled by default in C++17 mode 4559 only. The default can be overridden with -f{no-,}new-ttp-matching. 4560 * The C++ front end will now provide fix-it hints for some missing 4561 semicolons, allowing for automatic fixes by IDEs: 4562 4563test.cc:4:11: error: expected ';' after class definition 4564 class a {} 4565 ^ 4566 ; 4567 4568 * -Waligned-new has been added to the C++ front end. It warns about 4569 new of type with extended alignment without -faligned-new. 4570 4571 Runtime Library (libstdc++) 4572 4573 * The type of exception thrown by iostreams, std::ios_base::failure, 4574 now uses the [12]cxx11 ABI. 4575 * Experimental support for C++17, including the following new 4576 features: 4577 + std::string_view; 4578 + std::any, std::optional, and std::variant; 4579 + std::invoke, std::is_invocable, std::is_nothrow_invocable, and 4580 invoke_result; 4581 + std::is_swappable, and std::is_nothrow_swappable; 4582 + std::apply, and std::make_from_tuple; 4583 + std::void_t, std::bool_constant, std::conjunction, 4584 std::disjunction, and std::negation; 4585 + Variable templates for type traits; 4586 + Mathematical Special Functions; 4587 + std::chrono::floor, std::chrono::ceil, std::chrono::round, and 4588 std::chrono::abs; 4589 + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot; 4590 + std::scoped_lock, std::shared_mutex, 4591 std::atomic<T>::is_always_lock_free; 4592 + std::sample, std::default_searcher, std::boyer_moore_searcher 4593 and std::boyer_moore_horspool_searcher; 4594 + Extraction and re-insertion of map and set nodes, try_emplace 4595 members for maps, and functions for accessing containers 4596 std::size, std::empty, and std::data; 4597 + std::shared_ptr support for arrays, 4598 std::shared_ptr<T>::weak_type, 4599 std::enable_shared_from_this<T>::weak_from_this(), and 4600 std::owner_less<void>; 4601 + std::byte; 4602 + std::as_const, std::not_fn, 4603 std::has_unique_object_representations, constexpr 4604 std::addressof. 4605 Thanks to Daniel Kr�gler, Tim Shen, Edward Smith-Rowland, and Ville 4606 Voutilainen for work on the C++17 support. 4607 * A new power-of-two rehashing policy for use with the _Hashtable 4608 internals, thanks to Fran�ois Dumont. 4609 4610 Fortran 4611 4612 * Support for a number of extensions for compatibility with legacy 4613 code with new flags: 4614 + -fdec-structure Support for DEC STRUCTURE and UNION 4615 + -fdec-intrinsic-ints Support for new integer intrinsics with 4616 B/I/J/K prefixes such as BABS, JIAND... 4617 + -fdec-math Support for additional math intrinsics, including 4618 COTAN and degree-valued trigonometric functions such as TAND, 4619 ASIND... 4620 + -fdec Enable the -fdec-* family of extensions. 4621 * New flag -finit-derived to allow default initialization of 4622 derived-type variables. 4623 * Improved DO loops with step equal to 1 or -1, generates faster code 4624 without a loop preheader. A new warning, -Wundefined-do-loop, warns 4625 when a loop iterates either to HUGE(i) (with step equal to 1), or 4626 to -HUGE(i) (with step equal to -1). Invalid behavior can be caught 4627 at run time with -fcheck=do enabled: 4628 4629program test 4630 implicit none 4631 integer(1) :: i 4632 do i = -HUGE(i)+10, -HUGE(i)-1, -1 4633 print *, i 4634 end do 4635end program test 4636 4637At line 8 of file do_check_12.f90 4638Fortran runtime error: Loop iterates infinitely 4639 4640 * Version 4.5 of the [13]OpenMP specification is now partially 4641 supported in the Fortran compiler; the largest missing item is 4642 structure element mapping. 4643 * User-defined derived-type input/output (UDTIO) is added. 4644 * Derived type coarrays with allocatable and pointer components are 4645 partially supported. 4646 * Non-constant stop codes and error stop codes (Fortran 2015 4647 feature). 4648 * Derived types with allocatable components of recursive type. 4649 * Intrinsic assignment to polymorphic variables. 4650 * Improved submodule support. 4651 * Improved diagnostics (polymorphic results in pure functions). 4652 * Coarray: Support for failed images (Fortan 2015 feature). 4653 4654 Go 4655 4656 * GCC 7 provides a complete implementation of the Go 1.8.1 user 4657 packages. 4658 * Compared to the Go 1.8.1 toolchain, the garbage collector is more 4659 conservative and less concurrent. 4660 * Escape analysis is available for experimental use via the 4661 -fgo-optimize-allocs option. The -fgo-debug-escape prints 4662 information useful for debugging escape analysis choices. 4663 4664 Java (GCJ) 4665 4666 The GCC Java front end and associated libjava runtime library have been 4667 removed from GCC. 4668 4669libgccjit 4670 4671 The libgccjit API gained support for marking calls as requiring 4672 tail-call optimization via a new entry point: 4673 [14]gcc_jit_rvalue_set_bool_require_tail_call. 4674 4675 libgccjit performs numerous checks at the API boundary, but if these 4676 succeed, it previously ignored errors and other diagnostics emitted 4677 within the core of GCC, and treated the compile of a gcc_jit_context as 4678 having succeeded. As of GCC 7 it now ensures that if any diagnostics 4679 are emitted, they are visible from the libgccjit API, and that the the 4680 context is flagged as having failed. 4681 4682New Targets and Target Specific Improvements 4683 4684 AArch64 4685 4686 * GCC has been updated to the latest revision of the procedure call 4687 standard (AAPCS64) to provide support for parameter passing when 4688 data types have been over-aligned. 4689 * The ARMv8.3-A architecture is now supported. It can be used by 4690 specifying the -march=armv8.3-a option. 4691 * The option -msign-return-address= is supported to enable return 4692 address protection using ARMv8.3-A Pointer Authentication 4693 Extensions. For more information on the arguments accepted by this 4694 option, please refer to [15]AArch64-Options. 4695 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 4696 Extensions are now supported. They can be used by specifying the 4697 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 4698 Floating-Point Extensions introduce new half-precision data 4699 processing floating-point instructions. 4700 * Support has been added for the following processors (GCC 4701 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom 4702 Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium 4703 ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx 4704 (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1), 4705 Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor). 4706 The GCC identifiers can be used as arguments to the -mcpu or -mtune 4707 options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as 4708 arguments to the equivalent target attributes and pragmas. 4709 4710 ARC 4711 4712 * Added support for ARC HS and ARC EM processors. 4713 * Added support for ARC EM variation found in Intel QuarkSE SoCs. 4714 * Added support for NPS400 ARC700 based CPUs. 4715 * Thread Local Storage is now supported by ARC CPUs. 4716 * Fixed errors for ARC600 when using 32x16 multiplier option. 4717 * Fixed PIE for ARC CPUs. 4718 * New CPU templates are supported via multilib. 4719 4720 ARM 4721 4722 * Support for the ARMv5 and ARMv5E architectures has been deprecated 4723 (which have no known implementations) and will be removed in a 4724 future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ 4725 architectures remain supported. The values armv5 and armv5e of 4726 -march are thus deprecated. 4727 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 4728 Extensions are now supported. They can be used by specifying the 4729 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 4730 Floating-Point Extensions introduce new half-precision data 4731 processing floating-point instructions. 4732 * The ARMv8-M architecture is now supported in its two architecture 4733 profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and 4734 Floating-Point Extensions. They can be used by specifying the 4735 -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options. 4736 * Support has been added for the following processors (GCC 4737 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM 4738 Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC 4739 identifiers can be used as arguments to the -mcpu or -mtune 4740 options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33. 4741 * A new command-line option -mpure-code has been added. It does not 4742 allow constant data to be placed in code sections. This option is 4743 only available when generating non-PIC code for ARMv7-M targets. 4744 * Support for the ACLE Coprocessor Intrinsics has been added. This 4745 enables the generation of coprocessor instructions through the use 4746 of intrinsics such as cdp, ldc, and others. 4747 * The configure option --with-multilib-list now accepts the value 4748 rmprofile to build multilib libraries for a range of embedded 4749 targets. See our [16]installation instructions for details. 4750 4751 AVR 4752 4753 * On the reduced Tiny cores, the progmem [17]variable attribute is 4754 now properly supported. Respective read-only variables are located 4755 in flash memory in section .progmem.data. No special code is needed 4756 to access such variables; the compiler automatically adds an offset 4757 of 0x4000 to all addresses, which is needed to access variables in 4758 flash memory. As opposed to ordinary cores where it is sufficient 4759 to specify the progmem attribute with definitions, on the reduced 4760 Tiny cores the attribute also has to be specified with (external) 4761 declarations: 4762 4763extern const int array[] __attribute__((__progmem__)); 4764 4765int get_value2 (void) 4766{ 4767 /* Access via addresses array + 0x4004 and array + 0x4005. */ 4768 return array[2]; 4769} 4770 4771const int* get_address (unsigned idx) 4772{ 4773 /* Returns array + 0x4000 + 2 * idx. */ 4774 return &array[idx]; 4775} 4776 4777 * A new command-line option -Wmisspelled-isr has been added. It turns 4778 off -- or turns into errors -- warnings that are reported for 4779 interrupt service routines (ISRs) which don't follow AVR-LibC's 4780 naming convention of prefixing ISR names with __vector. 4781 * __builtin_avr_nops(n) is a new [18]built-in function that inserts n 4782 NOP instructions into the instruction stream. n must be a value 4783 known at compile time. 4784 4785 IA-32/x86-64 4786 4787 * Support for the AVX-512 Fused Multiply Accumulation Packed Single 4788 precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word 4789 variable precision (4VNNIW), AVX-512 Vector Population Count 4790 (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has 4791 been added. 4792 4793 NVPTX 4794 4795 * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs. 4796 See the [19]Offloading Wiki on how to configure it. 4797 4798 PowerPC / PowerPC64 / RS6000 4799 4800 * The PowerPC port now uses LRA by default. 4801 * GCC now diagnoses inline assembly that clobbers register r2. This 4802 has always been invalid code, and is no longer quietly tolerated. 4803 * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been 4804 enhanced to generate more of the new instructions by default, and 4805 to provide more built-in functions to generate code for other new 4806 instructions. 4807 * The configuration option --enable-gnu-indirect-function is now 4808 enabled by default on PowerPC GNU/Linux builds. 4809 * The PowerPC port will now allow 64-bit and 32-bit integer types to 4810 be allocated to the VSX vector registers (ISA 2.06 and above). In 4811 addition, on ISA 3.0, 16-bit and 8-bit integer types can be 4812 allocated in the vector registers. Previously, only 64-bit integer 4813 types were allowed in the traditional floating point registers. 4814 * New options -mstack-protector-guard=global, 4815 -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and 4816 -mstack-protector-guard-offset= change how the stack protector gets 4817 the value to use as canary. 4818 4819 S/390, System z, IBM z Systems, IBM Z 4820 4821 * Support for the IBM z14 processor has been added. When using the 4822 -march=z14 option, the compiler will generate code making use of 4823 the new instructions introduced with the vector enhancement 4824 facility and the miscellaneous instruction extension facility 2. 4825 The -mtune=z14 option enables z14 specific instruction scheduling 4826 without making use of new instructions. 4827 * Builtins for the new vector instructions have been added and can be 4828 enabled using the -mzvector option. 4829 4830 RISC-V 4831 4832 * Support for the RISC-V instruction set has been added. 4833 4834 RX 4835 4836 Basic support for atomic built-in function has been added. It is 4837 currently implemented by flipping interrupts off and on as needed. 4838 4839 SH 4840 4841 * Support for SH5/SH64 has been removed. 4842 * Improved utilization of delay slots on SH2A. 4843 * Improved utilization of zero-displacement conditional branches. 4844 * The following deprecated options have been removed 4845 + -mcbranchdi 4846 + -mcmpeqdi 4847 + -minvalid-symbols 4848 + -msoft-atomic 4849 + -mspace 4850 + -madjust-unroll 4851 * Support for the following SH2A instructions has been added 4852 + mov.b @-Rm,R0 4853 + mov.w @-Rm,R0 4854 + mov.l @-Rm,R0 4855 + mov.b R0,@Rn+ 4856 + mov.w R0,@Rn+ 4857 + mov.l R0,@Rn+ 4858 4859 SPARC 4860 4861 * The SPARC port now uses LRA by default. 4862 * Support for the new Subtract-Extended-with-Carry instruction 4863 available in SPARC M7 (Niagara 7) has been added. 4864 4865Operating Systems 4866 4867 AIX 4868 4869 * Visibility support has been enabled for AIX 7.1 and above. 4870 4871 Fuchsia 4872 4873 * Support has been added for the [20]Fuchsia OS. 4874 4875 RTEMS 4876 4877 * The ABI changes on ARM so that no short enums are used by default. 4878 4879Other significant improvements 4880 4881 * -fverbose-asm previously emitted information on the meanings of 4882 assembly expressions. This has been extended so that it now also 4883 prints comments showing the source lines that correspond to the 4884 assembly, making it easier to read the generated assembly 4885 (especially with larger functions). For example, given this C 4886 source file: 4887 4888int test (int n) 4889{ 4890 int i; 4891 int total = 0; 4892 4893 for (i = 0; i < n; i++) 4894 total += i * i; 4895 return total; 4896} 4897 4898 -fverbose-asm now gives output similar to this for the function 4899 body (when compiling for x86_64, with -Os): 4900 4901 .text 4902 .globl test 4903 .type test, @@function 4904test: 4905.LFB0: 4906 .cfi_startproc 4907# example.c:4: int total = 0; 4908 xorl %eax, %eax # <retval> 4909# example.c:6: for (i = 0; i < n; i++) 4910 xorl %edx, %edx # i 4911.L2: 4912# example.c:6: for (i = 0; i < n; i++) 4913 cmpl %edi, %edx # n, i 4914 jge .L5 #, 4915# example.c:7: total += i * i; 4916 movl %edx, %ecx # i, tmp92 4917 imull %edx, %ecx # i, tmp92 4918# example.c:6: for (i = 0; i < n; i++) 4919 incl %edx # i 4920# example.c:7: total += i * i; 4921 addl %ecx, %eax # tmp92, <retval> 4922 jmp .L2 # 4923.L5: 4924# example.c:10: } 4925 ret 4926 .cfi_endproc 4927 4928 * Two new options have been added for printing fix-it hints: 4929 + -fdiagnostics-parseable-fixits allows for fix-it hints to be 4930 emitted in a machine-readable form, suitable for consumption 4931 by IDEs. For example, given: 4932 4933spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 4934you mean 'color'? 4935 return ptr->colour; 4936 ^~~~~~ 4937 color 4938 4939 it will emit: 4940 4941fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color" 4942 4943 + -fdiagnostics-generate-patch will print a patch in "unified" 4944 format after any diagnostics are printed, showing the result 4945 of applying all fix-it hints. For the above example it would 4946 emit: 4947 4948--- spellcheck-fields.cc 4949+++ spellcheck-fields.cc 4950@@ -49,5 +49,5 @@ 4951 4952 color get_color(struct s *ptr) 4953 { 4954- return ptr->colour; 4955+ return ptr->color; 4956 } 4957 4958 * The gcc and g++ driver programs will now provide suggestions for 4959 misspelled arguments to command-line options. 4960 4961$ gcc -c test.c -ftls-model=global-dinamic 4962gcc: error: unknown TLS model 'global-dinamic' 4963gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo 4964cal-dynamic local-exec; did you mean 'global-dynamic'? 4965 4966 * The compiler will now provide suggestions for misspelled 4967 parameters. 4968 4969$ gcc -c test.c --param max-early-inliner-iteration=3 4970cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma 4971x-early-inliner-iterations'? 4972 4973 * Profile-guided optimization (PGO) instrumentation, as well as test 4974 coverage (GCOV), can newly instrument constructors (functions marks 4975 with __attribute__((constructor))), destructors and C++ 4976 constructors (and destructors) of classes that are used as the type 4977 of a global variable. 4978 * A new option -fprofile-update=atomic prevents creation of corrupted 4979 profiles created during an instrumentation run (-fprofile=generate) 4980 of an application. The downside of the option is a speed penalty. 4981 Providing -pthread on the command line selects atomic profile 4982 updating (when supported by the target). 4983 * GCC's already extensive testsuite has gained some new capabilities, 4984 to further improve the reliability of the compiler: 4985 + GCC now has an internal unit-testing API and a suite of tests 4986 for programmatic self-testing of subsystems. 4987 + GCC's C front end has been extended so that it can parse dumps 4988 of GCC's internal representations, allowing for DejaGnu tests 4989 that more directly exercise specific optimization passes. This 4990 covers both the [21]GIMPLE representation (for testing 4991 higher-level optimizations) and the [22]RTL representation, 4992 allowing for more direct testing of lower-level details, such 4993 as register allocation and instruction selection. 4994 4995GCC 7.1 4996 4997 This is the [23]list of problem reports (PRs) from GCC's bug tracking 4998 system that are known to be fixed in the 7.1 release. This list might 4999 not be complete (that is, it is possible that some PRs that have been 5000 fixed are not listed here). 5001 5002GCC 7.2 5003 5004 This is the [24]list of problem reports (PRs) from GCC's bug tracking 5005 system that are known to be fixed in the 7.2 release. This list might 5006 not be complete (that is, it is possible that some PRs that have been 5007 fixed are not listed here). 5008 5009 Target Specific Changes 5010 5011 SPARC 5012 5013 * Support for the SPARC M8 processor has been added. 5014 * The switches -mfix-ut700 and -mfix-gr712rc have been added to work 5015 around an erratum in LEON3FT processors. 5016 * Use of the Floating-point Multiply Single to Double (FsMULd) 5017 instruction can now be controlled by the -mfsmuld and -fno-fsmuld 5018 options. 5019 5020 Operating Systems 5021 5022 RTEMS 5023 5024 * The Ada run-time support uses now thread-local storage (TLS). 5025 * Support for RISC-V has been added. 5026 * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long 5027 double has been added. 5028 5029GCC 7.3 5030 5031 This is the [25]list of problem reports (PRs) from GCC's bug tracking 5032 system that are known to be fixed in the 7.3 release. This list might 5033 not be complete (that is, it is possible that some PRs that have been 5034 fixed are not listed here). 5035 5036 Target Specific Changes 5037 5038 SPARC 5039 5040 * Workarounds for the four [26]LEON3FT errata GRLIB-TN-0010..0013 5041 have been added. Relevant errata are activated by the target 5042 specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches. 5043 5044 Operating Systems 5045 5046 RTEMS 5047 5048 * Support has been added for Epiphany target. 5049 5050GCC 7.4 5051 5052 This is the [27]list of problem reports (PRs) from GCC's bug tracking 5053 system that are known to be fixed in the 7.4 release. This list might 5054 not be complete (that is, it is possible that some PRs that have been 5055 fixed are not listed here). 5056 5057GCC 7.5 5058 5059 This is the [28]list of problem reports (PRs) from GCC's bug tracking 5060 system that are known to be fixed in the 7.5 release. This list might 5061 not be complete (that is, it is possible that some PRs that have been 5062 fixed are not listed here). 5063 5064 5065 For questions related to the use of GCC, please consult these web 5066 pages and the [29]GCC manuals. If that fails, the 5067 [30]gcc-help@gcc.gnu.org mailing list might help. Comments on these 5068 web pages and the development of GCC are welcome on our developer 5069 list at [31]gcc@gcc.gnu.org. All of [32]our lists have public 5070 archives. 5071 5072 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 5073 distribution of this entire article is permitted in any medium, 5074 provided this notice is preserved. 5075 5076 These pages are [34]maintained by the GCC team. Last modified 5077 2021-07-28[35]. 5078 5079References 5080 5081 1. http://gcc.gnu.org/gcc-7/porting_to.html 5082 2. http://gcc.gnu.org/onlinedocs/index.html#current 5083 3. https://gcc.gnu.org/wiki/LRAIsDefault 5084 4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html 5085 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728 5086 6. http://www.dwarfstd.org/Download.php 5087 7. https://gcc.gnu.org/wiki/OpenACC 5088 8. https://gcc.gnu.org/wiki/Offloading 5089 9. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html 5090 10. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 5091 11. https://wg21.link/p0136 5092 12. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html 5093 13. https://www.openmp.org/specifications/ 5094 14. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call 5095 15. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options 5096 16. https://gcc.gnu.org/install/configure.html 5097 17. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html 5098 18. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html 5099 19. https://gcc.gnu.org/wiki/Offloading 5100 20. https://fuchsia.googlesource.com/ 5101 21. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html 5102 22. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html 5103 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0 5104 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2 5105 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3 5106 26. https://www.gaisler.com/index.php/information/app-tech-notes 5107 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4 5108 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5 5109 29. https://gcc.gnu.org/onlinedocs/ 5110 30. mailto:gcc-help@gcc.gnu.org 5111 31. mailto:gcc@gcc.gnu.org 5112 32. https://gcc.gnu.org/lists.html 5113 33. https://www.fsf.org/ 5114 34. https://gcc.gnu.org/about.html 5115 35. http://validator.w3.org/check/referer 5116====================================================================== 5117http://gcc.gnu.org/gcc-6/index.html 5118 5119 GCC 6 Release Series 5120 5121 (This release series is no longer supported.) 5122 5123 October 26, 2018 5124 5125 The [1]GNU project and the GCC developers are pleased to announce the 5126 release of GCC 6.5. 5127 5128 This release is a bug-fix release, containing fixes for regressions in 5129 GCC 6.4 relative to previous releases of GCC. 5130 5131Release History 5132 5133 GCC 6.5 5134 October 26, 2018 ([2]changes, [3]documentation) 5135 5136 GCC 6.4 5137 July 4, 2017 ([4]changes, [5]documentation) 5138 5139 GCC 6.3 5140 December 21, 2016 ([6]changes, [7]documentation) 5141 5142 GCC 6.2 5143 August 22, 2016 ([8]changes, [9]documentation) 5144 5145 GCC 6.1 5146 April 27, 2016 ([10]changes, [11]documentation) 5147 5148References and Acknowledgements 5149 5150 GCC used to stand for the GNU C Compiler, but since the compiler 5151 supports several other languages aside from C, it now stands for the 5152 GNU Compiler Collection. 5153 5154 A list of [12]successful builds is updated as new information becomes 5155 available. 5156 5157 The GCC developers would like to thank the numerous people that have 5158 contributed new features, improvements, bug fixes, and other changes as 5159 well as test results to GCC. This [13]amazing group of volunteers is 5160 what makes GCC successful. 5161 5162 For additional information about GCC please refer to the [14]GCC 5163 project web site or contact the [15]GCC development mailing list. 5164 5165 To obtain GCC please use [16]our mirror sites or [17]our version 5166 control system. 5167 5168 5169 For questions related to the use of GCC, please consult these web 5170 pages and the [18]GCC manuals. If that fails, the 5171 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 5172 web pages and the development of GCC are welcome on our developer 5173 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 5174 archives. 5175 5176 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 5177 distribution of this entire article is permitted in any medium, 5178 provided this notice is preserved. 5179 5180 These pages are [23]maintained by the GCC team. Last modified 5181 2021-07-28[24]. 5182 5183References 5184 5185 1. http://www.gnu.org/ 5186 2. http://gcc.gnu.org/gcc-6/changes.html 5187 3. http://gcc.gnu.org/onlinedocs/6.5.0/ 5188 4. http://gcc.gnu.org/gcc-6/changes.html 5189 5. http://gcc.gnu.org/onlinedocs/6.4.0/ 5190 6. http://gcc.gnu.org/gcc-6/changes.html 5191 7. http://gcc.gnu.org/onlinedocs/6.3.0/ 5192 8. http://gcc.gnu.org/gcc-6/changes.html 5193 9. http://gcc.gnu.org/onlinedocs/6.2.0/ 5194 10. http://gcc.gnu.org/gcc-6/changes.html 5195 11. http://gcc.gnu.org/onlinedocs/6.1.0/ 5196 12. http://gcc.gnu.org/gcc-6/buildstat.html 5197 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5198 14. http://gcc.gnu.org/index.html 5199 15. mailto:gcc@gcc.gnu.org 5200 16. http://gcc.gnu.org/mirrors.html 5201 17. http://gcc.gnu.org/git.html 5202 18. https://gcc.gnu.org/onlinedocs/ 5203 19. mailto:gcc-help@gcc.gnu.org 5204 20. mailto:gcc@gcc.gnu.org 5205 21. https://gcc.gnu.org/lists.html 5206 22. https://www.fsf.org/ 5207 23. https://gcc.gnu.org/about.html 5208 24. http://validator.w3.org/check/referer 5209====================================================================== 5210http://gcc.gnu.org/gcc-6/changes.html 5211 5212 GCC 6 Release Series 5213 Changes, New Features, and Fixes 5214 5215 This page is a brief summary of some of the huge number of improvements 5216 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 5217 the [2]full GCC documentation. 5218 5219Caveats 5220 5221 * The default mode for C++ is now -std=gnu++14 instead of 5222 -std=gnu++98. 5223 * Support for a number of older systems and recently unmaintained or 5224 untested target ports of GCC has been declared obsolete in GCC 6. 5225 Unless there is activity to revive them, the next release of GCC 5226 will have their sources permanently removed. 5227 The following ports for individual systems on particular 5228 architectures have been obsoleted: 5229 + SH5 / SH64 (sh64-*-*) as announced [3]here. 5230 * The AVR port requires binutils version 2.26.1 or later for the fix 5231 for [4]PR71151 to work. 5232 * The GCC 6.5 release has an accidental ABI incompatibility for 5233 nested std::pair objects, for more details see [5]PR 87822. The bug 5234 causes a layout change for pairs where the first member is also a 5235 pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series 5236 is closed so the bug in GCC 6.5 will not be fixed upstream, but 5237 there is a patch in the bug report to allow it to be fixed by 5238 anybody packaging GCC 6.5 or installing it themselves. 5239 5240General Optimizer Improvements 5241 5242 * UndefinedBehaviorSanitizer gained a new sanitization option, 5243 -fsanitize=bounds-strict, which enables strict checking of array 5244 bounds. In particular, it enables -fsanitize=bounds as well as 5245 instrumentation of flexible array member-like arrays. 5246 * Type-based alias analysis now disambiguates accesses to different 5247 pointers. This improves precision of the alias oracle by about 5248 20-30% on higher-level C++ programs. Programs doing invalid type 5249 punning of pointer types may now need -fno-strict-aliasing to work 5250 correctly. 5251 * Alias analysis now correctly supports the weakref and alias 5252 attributes. This allows accessing both a variable and its alias in 5253 one translation unit which is common with link-time optimization. 5254 * Value range propagation now assumes that the this pointer in C++ 5255 member functions is non-null. This eliminates common null pointer 5256 checks but also breaks some non-conforming code-bases (such as 5257 Qt-5, Chromium, KDevelop). As a temporary work-around 5258 -fno-delete-null-pointer-checks can be used. Wrong code can be 5259 identified by using -fsanitize=undefined. 5260 * Link-time optimization improvements: 5261 + warning and error attributes are now correctly preserved by 5262 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 5263 supported with -flto. 5264 + Type merging was fixed to handle C and Fortran 5265 interoperability rules as defined by the Fortran 2008 language 5266 standard. 5267 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 5268 with char in all cases because it is an array while char is 5269 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 5270 general, this inter-operability cannot be implemented, for 5271 example on targets where the argument passing convention for 5272 arrays differs from scalars. 5273 + More type information is now preserved at link time, reducing 5274 the loss of accuracy of the type-based alias analysis compared 5275 to builds without link-time optimization. 5276 + Invalid type punning on global variables and declarations is 5277 now reported with -Wodr-type-mismatch. 5278 + The size of LTO object files was reduced by about 11% 5279 (measured by compiling Firefox 46.0). 5280 + Link-time parallelization (enabled using -flto=n) was 5281 significantly improved by decreasing the size of streamed data 5282 when partitioning programs. The size of streamed IL while 5283 compiling Firefox 46.0 was reduced by 66%. 5284 + The linker plugin was extended to pass information about the 5285 type of binary produced to the GCC back end. (That can also be 5286 controlled manually by -flinker-output.) This makes it 5287 possible to properly configure the code generator and support 5288 incremental linking. Incremental linking of LTO objects by gcc 5289 -r is now supported for plugin-enabled setups. 5290 There are two ways to perform incremental linking: 5291 1. Linking by ld -r will result in an object file with all 5292 sections from individual object files mechanically 5293 merged. This delays the actual link-time optimization to 5294 the final linking step and thus permits whole program 5295 optimization. Linking the final binary with such object 5296 files is however slower. 5297 2. Linking by gcc -r will lead to link-time optimization and 5298 emit the final binary into the object file. Linking such 5299 an object file is fast but avoids any benefits from whole 5300 program optimization. 5301 GCC 7 will support incremental link-time optimization with gcc 5302 -r. 5303 * Inter-procedural optimization improvements: 5304 + Basic jump threading is now performed before profile 5305 construction and inline analysis, resulting in more realistic 5306 size and time estimates that drive the heuristics of the 5307 inliner and function cloning passes. 5308 + Function cloning now more aggressively eliminates unused 5309 function parameters. 5310 5311New Languages and Language specific improvements 5312 5313 Compared to GCC 5, the GCC 6 release series includes a much improved 5314 implementation of the [6]OpenACC 2.0a specification. Highlights are: 5315 * In addition to single-threaded host-fallback execution, offloading 5316 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 5317 little-endian GNU/Linux host systems. For nvptx offloading, with 5318 the OpenACC parallel construct, the execution model allows for an 5319 arbitrary number of gangs, up to 32 workers, and 32 vectors. 5320 * Initial support for parallelized execution of OpenACC kernels 5321 constructs: 5322 + Parallelization of a kernels region is switched on by 5323 -fopenacc combined with -O2 or higher. 5324 + Code is offloaded onto multiple gangs, but executes with just 5325 one worker, and a vector length of 1. 5326 + Directives inside a kernels region are not supported. 5327 + Loops with reductions can be parallelized. 5328 + Only kernels regions with one loop nest are parallelized. 5329 + Only the outer-most loop of a loop nest can be parallelized. 5330 + Loop nests containing sibling loops are not parallelized. 5331 Typically, using the OpenACC parallel construct gives much better 5332 performance, compared to the initial support of the OpenACC kernels 5333 construct. 5334 * The device_type clause is not supported. The bind and nohost 5335 clauses are not supported. The host_data directive is not supported 5336 in Fortran. 5337 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 5338 * Usage of OpenACC constructs inside multithreaded contexts (such as 5339 created by OpenMP, or pthread programming) is not supported. 5340 * If a call to the acc_on_device function has a compile-time constant 5341 argument, the function call evaluates to a compile-time constant 5342 value only for C and C++ but not for Fortran. 5343 5344 See the [7]OpenACC and [8]Offloading wiki pages for further 5345 information. 5346 5347 C family 5348 5349 * Version 4.5 of the [9]OpenMP specification is now supported in the 5350 C and C++ compilers. 5351 * The C and C++ compilers now support attributes on enumerators. For 5352 instance, it is now possible to mark enumerators as deprecated: 5353 5354enum { 5355 newval, 5356 oldval __attribute__ ((deprecated ("too old"))) 5357}; 5358 5359 * Source locations for the C and C++ compilers are now tracked as 5360 ranges, rather than just points, making it easier to identify the 5361 subexpression of interest within a complicated expression. For 5362 example: 5363 5364test.cc: In function 'int test(int, int, foo, int, int)': 5365test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 5366) 5367 return p + q * r * s + t; 5368 ~~^~~ 5369 5370 In addition, there is now initial support for precise diagnostic 5371 locations within strings: 5372 5373format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 5374t' argument [-Wformat=] 5375 printf("%*d"); 5376 ^ 5377 5378 * Diagnostics can now contain "fix-it hints", which are displayed in 5379 context underneath the relevant source code. For example: 5380 5381fixits.c: In function 'bad_deref': 5382fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 5383 return ptr.x; 5384 ^ 5385 -> 5386 5387 * The C and C++ compilers now offer suggestions for misspelled field 5388 names: 5389 5390spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 5391you mean 'color'? 5392 return ptr->colour; 5393 ^~~~~~ 5394 5395 * New command-line options have been added for the C and C++ 5396 compilers: 5397 + -Wshift-negative-value warns about left shifting a negative 5398 value. 5399 + -Wshift-overflow warns about left shift overflows. This 5400 warning is enabled by default. -Wshift-overflow=2 also warns 5401 about left-shifting 1 into the sign bit. 5402 + -Wtautological-compare warns if a self-comparison always 5403 evaluates to true or false. This warning is enabled by -Wall. 5404 + -Wnull-dereference warns if the compiler detects paths that 5405 trigger erroneous or undefined behavior due to dereferencing a 5406 null pointer. This option is only active when 5407 -fdelete-null-pointer-checks is active, which is enabled by 5408 optimizations in most targets. The precision of the warnings 5409 depends on the optimization options used. 5410 + -Wduplicated-cond warns about duplicated conditions in an 5411 if-else-if chain. 5412 + -Wmisleading-indentation warns about places where the 5413 indentation of the code gives a misleading idea of the block 5414 structure of the code to a human reader. For example, given 5415 [10]CVE-2014-1266: 5416 5417sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 5418sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 5419g-indentation] 5420 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 5421 ^~ 5422sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 5423indented as if it is guarded by the 'if' 5424 goto fail; 5425 ^~~~ 5426 5427 This warning is enabled by -Wall. 5428 * The C and C++ compilers now emit saner error messages if 5429 merge-conflict markers are present in a source file. 5430 5431test.c:3:1: error: version control conflict marker in file 5432 <<<<<<< HEAD 5433 ^~~~~~~ 5434 5435 C 5436 5437 * It is possible to disable warnings when an initialized field of a 5438 structure or a union with side effects is being overridden when 5439 using designated initializers via a new warning option 5440 -Woverride-init-side-effects. 5441 * A new type attribute scalar_storage_order applying to structures 5442 and unions has been introduced. It specifies the storage order (aka 5443 endianness) in memory of scalar fields in structures or unions. 5444 5445 C++ 5446 5447 * The default mode has been changed to -std=gnu++14. 5448 * [11]C++ Concepts are now supported when compiling with -fconcepts. 5449 * -flifetime-dse is more aggressive in dead-store elimination in 5450 situations where a memory store to a location precedes a 5451 constructor to that memory location. 5452 * G++ now supports [12]C++17 fold expressions, u8 character literals, 5453 extended static_assert, and nested namespace definitions. 5454 * G++ now allows constant evaluation for all non-type template 5455 arguments. 5456 * G++ now supports C++ Transactional Memory when compiling with 5457 -fgnu-tm. 5458 5459 Runtime Library (libstdc++) 5460 5461 * Extensions to the C++ Library to support mathematical special 5462 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 5463 * Experimental support for C++17, including the following new 5464 features: 5465 + std::uncaught_exceptions function (this is also available for 5466 -std=gnu++NN modes); 5467 + new member functions try_emplace and insert_or_assign for 5468 unique_key maps; 5469 + non-member functions std::size, std::empty, and std::data for 5470 accessing containers and arrays; 5471 + std::invoke; 5472 + std::shared_mutex; 5473 + std::void_t and std::bool_constant metaprogramming utilities. 5474 Thanks to Ville Voutilainen for contributing many of the C++17 5475 features. 5476 * An experimental implementation of the File System TS. 5477 * Experimental support for most features of the second version of the 5478 Library Fundamentals TS. This includes polymorphic memory resources 5479 and array support in shared_ptr, thanks to Fan You. 5480 * Some assertions checked by Debug Mode can now also be enabled by 5481 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 5482 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 5483 don't affect the library ABI, so can be enabled per-translation 5484 unit. 5485 * Timed mutex types are supported on more targets, including Darwin. 5486 * Improved std::locale support for DragonFly and FreeBSD, thanks to 5487 John Marino and Andreas Tobler. 5488 5489 Fortran 5490 5491 * Fortran 2008 SUBMODULE support. 5492 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 5493 support. 5494 * Improved support for Fortran 2003 deferred-length character 5495 variables. 5496 * Improved support for OpenMP and OpenACC. 5497 * The MATMUL intrinsic is now inlined for straightforward cases if 5498 front-end optimization is active. The maximum size for inlining can 5499 be set to n with the -finline-matmul-limit=n option and turned off 5500 with -finline-matmul-limit=0. 5501 * The -Wconversion-extra option will warn about REAL constants which 5502 have excess precision for their kind. 5503 * The -Winteger-division option has been added, which warns about 5504 divisions of integer constants which are truncated. This option is 5505 included in -Wall by default. 5506 5507libgccjit 5508 5509 * The driver code is now run in-process within libgccjit, providing a 5510 small speed-up of the compilation process. 5511 * The API has gained entrypoints for 5512 + [13]timing how long was spent in different parts of code, 5513 + [14]creating switch statements, 5514 + [15]allowing unreachable basic blocks in a function, and 5515 + [16]adding arbitrary command-line options to a compilation. 5516 5517New Targets and Target Specific Improvements 5518 5519 AArch64 5520 5521 * A number of AArch64-specific options have been added. The most 5522 important ones are summarised in this section; for more detailed 5523 information please refer to the documentation. 5524 * The command-line options -march=native, -mcpu=native and 5525 -mtune=native are now available on native AArch64 GNU/Linux 5526 systems. Specifying these options causes GCC to auto-detect the 5527 host CPU and choose the optimal setting for that system. 5528 * -fpic is now supported when generating code for the small code 5529 model (-mcmodel=small). The size of the global offset table (GOT) 5530 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 5531 ILP32 SysV ABI. 5532 * The AArch64 port now supports target attributes and pragmas. Please 5533 refer to the [17]documentation for details of available attributes 5534 and pragmas as well as usage instructions. 5535 * Link-time optimization across translation units with different 5536 target-specific options is now supported. 5537 * The option -mtls-size= is now supported. It can be used to specify 5538 the bit size of TLS offsets, allowing GCC to generate better TLS 5539 instruction sequences. 5540 * The option -fno-plt is now fully functional. 5541 * The ARMv8.1-A architecture and the Large System Extensions are now 5542 supported. They can be used by specifying the -march=armv8.1-a 5543 option. Additionally, the +lse option extension can be used in a 5544 similar fashion to other option extensions. The Large System 5545 Extensions introduce new instructions that are used in the 5546 implementation of atomic operations. 5547 * The ACLE half-precision floating-point type __fp16 is now supported 5548 in the C and C++ languages. 5549 * The ARM Cortex-A35 processor is now supported via the 5550 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 5551 equivalent target attributes and pragmas. 5552 * The Qualcomm QDF24xx processor is now supported via the 5553 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 5554 target attributes and pragmas. 5555 * Code generation for the ARM Cortex-A57 processor is improved. Among 5556 general code generation improvements, a better algorithm is added 5557 for allocating registers to floating-point multiply-accumulate 5558 instructions offering increased performance when compiling with 5559 -mcpu=cortex-a57 or -mtune=cortex-a57. 5560 * Code generation for the ARM Cortex-A53 processor is improved. A 5561 more accurate instruction scheduling model for the processor is now 5562 used, and a number of compiler tuning parameters have been set to 5563 offer increased performance when compiling with -mcpu=cortex-a53 or 5564 -mtune=cortex-a53. 5565 * Code generation for the Samsung Exynos M1 processor is improved. A 5566 more accurate instruction scheduling model for the processor is now 5567 used, and a number of compiler tuning parameters have been set to 5568 offer increased performance when compiling with -mcpu=exynos-m1 or 5569 -mtune=exynos-m1. 5570 * Improvements in the generation of conditional branches and literal 5571 pools allow the compiler to compile functions of a large size. 5572 Constant pools are now placed into separate rodata sections. The 5573 new option -mpc-relative-literal-loads generates per-function 5574 literal pools, limiting the maximum size of functions to 1MiB. 5575 * Several correctness issues generating Advanced SIMD instructions 5576 for big-endian targets have been fixed resulting in improved code 5577 generation for ACLE intrinsics with -mbig-endian. 5578 5579 ARM 5580 5581 * Support for revisions of the ARM architecture prior to ARMv4t has 5582 been deprecated and will be removed in a future GCC release. The 5583 -mcpu and -mtune values that are deprecated are: arm2, arm250, 5584 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 5585 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 5586 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 5587 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 5588 arm7tdmi is still supported. The values of -march that are 5589 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 5590 * The ARM port now supports target attributes and pragmas. Please 5591 refer to the [18]documentation for details of available attributes 5592 and pragmas as well as usage instructions. 5593 * Support has been added for the following processors (GCC 5594 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 5595 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 5596 identifiers can be used as arguments to the -mcpu or -mtune 5597 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 5598 5599 Heterogeneous Systems Architecture 5600 5601 * GCC can now generate HSAIL (Heterogeneous System Architecture 5602 Intermediate Language) for simple OpenMP device constructs if 5603 configured with --enable-offload-targets=hsa. A new libgomp plugin 5604 then runs the HSA GPU kernels implementing these constructs on HSA 5605 capable GPUs via a standard HSA run time. 5606 If the HSA compilation back end determines it cannot output HSAIL 5607 for a particular input, it gives a warning by default. These 5608 warnings can be suppressed with -Wno-hsa. To give a few examples, 5609 the HSA back end does not implement compilation of code using 5610 function pointers, automatic allocation of variable sized arrays, 5611 functions with variadic arguments as well as a number of other less 5612 common programming constructs. 5613 When compilation for HSA is enabled, the compiler attempts to 5614 compile composite OpenMP constructs 5615 5616#pragma omp target teams distribute parallel for 5617 5618 into parallel HSA GPU kernels. 5619 5620 IA-32/x86-64 5621 5622 * GCC now supports the Intel CPU named Skylake with AVX-512 5623 extensions through -march=skylake-avx512. The switch enables the 5624 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 5625 AVX-512DQ. 5626 * Support for new AMD instructions monitorx and mwaitx has been 5627 added. This includes new intrinsic and built-in support. It is 5628 enabled through option -mmwaitx. The instructions monitorx and 5629 mwaitx implement the same functionality as the old monitor and 5630 mwait instructions. In addition mwaitx adds a configurable timer. 5631 The timer value is received as third argument and stored in 5632 register %ebx. 5633 * x86-64 targets now allow stack realignment from a word-aligned 5634 stack pointer using the command-line option -mstackrealign or 5635 __attribute__ ((force_align_arg_pointer)). This allows functions 5636 compiled with a vector-aligned stack to be invoked from objects 5637 that keep only word-alignment. 5638 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 5639 can be used to access data via the %fs and %gs segments without 5640 having to resort to inline assembly. Please refer to the 5641 [19]documentation for usage instructions. 5642 * Support for AMD Zen (family 17h) processors is now available 5643 through the -march=znver1 and -mtune=znver1 options. 5644 5645 MeP 5646 5647 * Support for the MeP (mep-elf) architecture has been deprecated and 5648 will be removed in a future GCC release. 5649 5650 MSP430 5651 5652 * The MSP430 compiler now has the ability to automatically distribute 5653 code and data between low memory (addresses below 64K) and high 5654 memory. This only applies to parts that actually have both memory 5655 regions and only if the linker script for the part has been 5656 specifically set up to support this feature. 5657 A new attribute of either can be applied to both functions and 5658 data, and this tells the compiler to place the object into low 5659 memory if there is room and into high memory otherwise. Two other 5660 new attributes - lower and upper - can be used to explicitly state 5661 that an object should be placed in the specified memory region. If 5662 there is not enough left in that region the compilation will fail. 5663 Two new command-line options - -mcode-region=[lower|upper|either] 5664 and -mdata-region=[lower|upper|either] - can be used to tell the 5665 compiler what to do with objects that do not have one of these new 5666 attributes. 5667 5668 PowerPC / PowerPC64 / RS6000 5669 5670 * PowerPC64 now supports IEEE 128-bit floating-point using the 5671 __float128 data type. In GCC 6, this is not enabled by default, but 5672 you can enable it with -mfloat128. The IEEE 128-bit floating-point 5673 support requires the use of the VSX instruction set. IEEE 128-bit 5674 floating-point values are passed and returned as a single vector 5675 value. The software emulator for IEEE 128-bit floating-point 5676 support is only built on PowerPC GNU/Linux systems where the 5677 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 5678 and later), you will be able to use the -mfloat128-hardware option 5679 to use the ISA 3.0 instructions that support IEEE 128-bit 5680 floating-point. An additional type (__ibm128) has been added to 5681 refer to the IBM extended double type that normally implements long 5682 double. This will allow for a future transition to implementing 5683 long double with IEEE 128-bit floating-point. 5684 * Basic support has been added for POWER9 hardware that will use the 5685 recently published OpenPOWER ISA 3.0 instructions. The following 5686 new switches are available: 5687 + -mcpu=power9: Implement all of the ISA 3.0 instructions 5688 supported by the compiler. 5689 + -mtune=power9: In the future, apply tuning for POWER9 systems. 5690 Currently, POWER8 tunings are used. 5691 + -mmodulo: Generate code using the ISA 3.0 integer instructions 5692 (modulus, count trailing zeros, array index support, integer 5693 multiply/add). 5694 + -mpower9-fusion: Generate code to suitably fuse instruction 5695 sequences for a POWER9 system. 5696 + -mpower9-dform: Generate code to use the new D-form 5697 (register+offset) memory instructions for the vector 5698 registers. 5699 + -mpower9-vector: Generate code using the new ISA 3.0 vector 5700 (VSX or Altivec) instructions. 5701 + -mpower9-minmax: Reserved for future development. 5702 + -mtoc-fusion: Keep TOC entries together to provide more fusion 5703 opportunities. 5704 * New constraints have been added to support IEEE 128-bit 5705 floating-point and ISA 3.0 instructions: 5706 + wb: Altivec register if -mpower9-dform is enabled. 5707 + we: VSX register if -mpower9-vector is enabled for 64-bit code 5708 generation. 5709 + wo: VSX register if -mpower9-vector is enabled. 5710 + wp: Reserved for future use if long double is implemented with 5711 IEEE 128-bit floating-point instead of IBM extended double. 5712 + wq: VSX register if -mfloat128 is enabled. 5713 + wF: Memory operand suitable for POWER9 fusion load/store. 5714 + wG: Memory operand suitable for TOC fusion memory references. 5715 + wL: Integer constant identifying the element number mfvsrld 5716 accesses within a vector. 5717 * Support has been added for __builtin_cpu_is() and 5718 __builtin_cpu_supports(), allowing for very fast access to 5719 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 5720 glibc 2.23 or later. 5721 * All hardware transactional memory builtins now correctly behave as 5722 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 5723 determine whether their "old" compiler treats the builtins as 5724 barriers. 5725 * Split-stack support has been added for gccgo on PowerPC64 for both 5726 big- and little-endian (but not for 32-bit). The gold linker from 5727 at least binutils 2.25.1 must be available in the PATH when 5728 configuring and building gccgo to enable split stack. (The 5729 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 5730 split-stack feature allows a small initial stack size to be 5731 allocated for each goroutine, which increases as needed. 5732 * GCC on PowerPC now supports the standard lround function. 5733 * A new configuration option ---with-advance-toolchain=at was added 5734 for PowerPC 64-bit GNU/Linux systems to use the header files, 5735 library files, and the dynamic linker from a specific Advance 5736 Toolchain release instead of the default versions that are provided 5737 by the GNU/Linux distribution. In general, this option is intended 5738 for the developers of GCC, and it is not intended for general use. 5739 * The "q", "S", "T", and "t" asm-constraints have been removed. 5740 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 5741 5742 S/390, System z, IBM z Systems 5743 5744 * Support for the IBM z13 processor has been added. When using the 5745 -march=z13 option, the compiler will generate code making use of 5746 the new instructions and registers introduced with the vector 5747 extension facility. The -mtune=z13 option enables z13 specific 5748 instruction scheduling without making use of new instructions. 5749 Compiling code with -march=z13 reduces the default alignment of 5750 vector types bigger than 8 bytes to 8. This is an ABI change and 5751 care must be taken when linking modules compiled with different 5752 arch levels which interchange variables containing vector type 5753 values. For newly compiled code the GNU linker will emit a warning. 5754 * The -mzvector option enables a C/C++ language extension. This 5755 extension provides a new keyword vector which can be used to define 5756 vector type variables. (Note: This is not available when enforcing 5757 strict standard compliance e.g. with -std=c99. Either enable GNU 5758 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 5759 Additionally a set of overloaded builtins is provided which is 5760 partially compatible to the PowerPC Altivec builtins. In order to 5761 make use of these builtins the vecintrin.h header file needs to be 5762 included. 5763 * The new command-line options -march=native, and -mtune=native are 5764 now available on native IBM z Systems. Specifying these options 5765 causes GCC to auto-detect the host CPU and choose the optimal 5766 setting for that system. If GCC is unable to detect the host CPU 5767 these options have no effect. 5768 * The IBM z Systems port now supports target attributes and pragmas. 5769 Please refer to the [20]documentation for details of available 5770 attributes and pragmas as well as usage instructions. 5771 * -fsplit-stack is now supported as part of the IBM z Systems port. 5772 This feature requires a recent gold linker to be used. 5773 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 5774 been deprecated and will be removed in a future GCC release. -m31 5775 from now on defaults to -march=z900 if not specified otherwise. 5776 -march=native on a g5/g6 machine will default to -march=z900. 5777 5778 SH 5779 5780 * Support for SH5 / SH64 has been declared obsolete and will be 5781 removed in future releases. 5782 * Support for the FDPIC ABI has been added. It can be enabled using 5783 the new -mfdpic target option and --enable-fdpic configure option. 5784 5785 SPARC 5786 5787 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 5788 change will break binary compatibility with earlier releases for 5789 code it affects, but this should be pretty rare in practice. The 5790 conditions are: a 16-byte structure containing a double or a 8-byte 5791 vector in the second half is passed to a subprogram in slot #15, 5792 for example as 16th parameter if the first 15 ones have at most 8 5793 bytes. The double or vector was wrongly passed in floating-point 5794 register %d32 in lieu of on the stack as per the SPARC calling 5795 conventions. 5796 5797Operating Systems 5798 5799 AIX 5800 5801 * DWARF debugging support for AIX 7.1 has been enabled as an optional 5802 debugging format. A more recent Technology Level (TL) and GCC built 5803 with that level are required for full exploitation of DWARF 5804 debugging capabilities. 5805 5806 Linux 5807 5808 * Support for the [21]musl C library was added for the AArch64, ARM, 5809 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 5810 x86_64 targets. It can be selected using the new -mmusl option in 5811 case musl is not the default libc. GCC defaults to musl libc if it 5812 is built with a target triplet matching the *-linux-musl* pattern. 5813 5814 RTEMS 5815 5816 * The RTEMS thread model implementation changed. Mutexes now use 5817 self-contained objects defined in Newlib <sys/lock.h> instead of 5818 Classic API semaphores. The keys for thread specific data and the 5819 once function are directly defined via <pthread.h>. Self-contained 5820 condition variables are provided via Newlib <sys/lock.h>. The RTEMS 5821 thread model also supports C++11 threads. 5822 * OpenMP support now uses self-contained objects provided by Newlib 5823 <sys/lock.h> and offers a significantly better performance compared 5824 to the POSIX configuration of libgomp. It is possible to configure 5825 thread pools for each scheduler instance via the environment 5826 variable GOMP_RTEMS_THREAD_POOLS. 5827 5828 Solaris 5829 5830 * Solaris 12 is now fully supported. Minimal support had already been 5831 present in GCC 5.3. 5832 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 5833 crtn.o), which GCC now prefers over its own ones. 5834 * Position independent executables (PIE) are now supported on Solaris 5835 12. 5836 * Constructor priority is now supported on Solaris 12 with the system 5837 linker. 5838 * libvtv has been ported to Solaris 11 and up. 5839 5840 Windows 5841 5842 * The option -mstackrealign is now automatically activated in 32-bit 5843 mode whenever the use of SSE instructions is requested. 5844 5845Other significant improvements 5846 5847 * The gcc and g++ driver programs will now provide suggestions for 5848 misspelled command-line options. 5849 5850$ gcc -static-libfortran test.f95 5851gcc: error: unrecognized command line option '-static-libfortran'; did you mean 5852'-static-libgfortran'? 5853 5854 * The --enable-default-pie configure option enables generation of PIE 5855 by default. 5856 5857 GCC 6.2 5858 5859 This is the [22]list of problem reports (PRs) from GCC's bug tracking 5860 system that are known to be fixed in the 6.2 release. This list might 5861 not be complete (that is, it is possible that some PRs that have been 5862 fixed are not listed here). 5863 5864Target Specific Changes 5865 5866 SPARC 5867 5868 * Support for --with-cpu-32 and --with-cpu-64 configure options has 5869 been added on bi-architecture platforms. 5870 * Support for the SPARC M7 (Niagara 7) processor has been added. 5871 * Support for the VIS 4.0 instruction set has been added. 5872 5873 GCC 6.3 5874 5875 This is the [23]list of problem reports (PRs) from GCC's bug tracking 5876 system that are known to be fixed in the 6.3 release. This list might 5877 not be complete (that is, it is possible that some PRs that have been 5878 fixed are not listed here). 5879 5880Target Specific Changes 5881 5882 IA-32/x86-64 5883 5884 * Support for the [24]deprecated pcommit instruction has been 5885 removed. 5886 5887 GCC 6.4 5888 5889 This is the [25]list of problem reports (PRs) from GCC's bug tracking 5890 system that are known to be fixed in the 6.4 release. This list might 5891 not be complete (that is, it is possible that some PRs that have been 5892 fixed are not listed here). 5893 5894Operating Systems 5895 5896 RTEMS 5897 5898 * The ABI changes on ARM so that no short enums are used by default. 5899 5900 GCC 6.5 5901 5902 This is the [26]list of problem reports (PRs) from GCC's bug tracking 5903 system that are known to be fixed in the 6.5 release. This list might 5904 not be complete (that is, it is possible that some PRs that have been 5905 fixed are not listed here). 5906 5907 5908 For questions related to the use of GCC, please consult these web 5909 pages and the [27]GCC manuals. If that fails, the 5910 [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these 5911 web pages and the development of GCC are welcome on our developer 5912 list at [29]gcc@gcc.gnu.org. All of [30]our lists have public 5913 archives. 5914 5915 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 5916 distribution of this entire article is permitted in any medium, 5917 provided this notice is preserved. 5918 5919 These pages are [32]maintained by the GCC team. Last modified 5920 2021-07-28[33]. 5921 5922References 5923 5924 1. http://gcc.gnu.org/gcc-6/porting_to.html 5925 2. http://gcc.gnu.org/onlinedocs/index.html#current 5926 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 5927 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 5928 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822 5929 6. https://www.openacc.org/ 5930 7. https://gcc.gnu.org/wiki/OpenACC 5931 8. https://gcc.gnu.org/wiki/Offloading 5932 9. https://www.openmp.org/specifications/ 5933 10. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266 5934 11. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 5935 12. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 5936 13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 5937 14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 5938 15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 5939 16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 5940 17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 5941 18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 5942 19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 5943 20. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 5944 21. http://www.musl-libc.org/ 5945 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 5946 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 5947 24. https://software.intel.com/content/www/us/en/develop/blogs/deprecate-pcommit-instruction.html 5948 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 5949 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5 5950 27. https://gcc.gnu.org/onlinedocs/ 5951 28. mailto:gcc-help@gcc.gnu.org 5952 29. mailto:gcc@gcc.gnu.org 5953 30. https://gcc.gnu.org/lists.html 5954 31. https://www.fsf.org/ 5955 32. https://gcc.gnu.org/about.html 5956 33. http://validator.w3.org/check/referer 5957====================================================================== 5958http://gcc.gnu.org/gcc-5/index.html 5959 5960 GCC 5 Release Series 5961 5962 (This release series is no longer supported.) 5963 5964 October 10, 2017 5965 5966 The [1]GNU project and the GCC developers are pleased to announce the 5967 release of GCC 5.5. 5968 5969 This release is a bug-fix release, containing fixes for regressions in 5970 GCC 5.4 relative to previous releases of GCC. 5971 5972Release History 5973 5974 GCC 5.5 5975 October 10, 2017 ([2]changes, [3]documentation) 5976 5977 GCC 5.4 5978 June 3, 2016 ([4]changes, [5]documentation) 5979 5980 GCC 5.3 5981 December 4, 2015 ([6]changes, [7]documentation) 5982 5983 GCC 5.2 5984 July 16, 2015 ([8]changes, [9]documentation) 5985 5986 GCC 5.1 5987 April 22, 2015 ([10]changes, [11]documentation) 5988 5989References and Acknowledgements 5990 5991 GCC used to stand for the GNU C Compiler, but since the compiler 5992 supports several other languages aside from C, it now stands for the 5993 GNU Compiler Collection. 5994 5995 A list of [12]successful builds is updated as new information becomes 5996 available. 5997 5998 The GCC developers would like to thank the numerous people that have 5999 contributed new features, improvements, bug fixes, and other changes as 6000 well as test results to GCC. This [13]amazing group of volunteers is 6001 what makes GCC successful. 6002 6003 For additional information about GCC please refer to the [14]GCC 6004 project web site or contact the [15]GCC development mailing list. 6005 6006 To obtain GCC please use [16]our mirror sites or [17]our version 6007 control system. 6008 6009 6010 For questions related to the use of GCC, please consult these web 6011 pages and the [18]GCC manuals. If that fails, the 6012 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 6013 web pages and the development of GCC are welcome on our developer 6014 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 6015 archives. 6016 6017 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 6018 distribution of this entire article is permitted in any medium, 6019 provided this notice is preserved. 6020 6021 These pages are [23]maintained by the GCC team. Last modified 6022 2021-07-28[24]. 6023 6024References 6025 6026 1. http://www.gnu.org/ 6027 2. http://gcc.gnu.org/gcc-5/changes.html 6028 3. http://gcc.gnu.org/onlinedocs/5.5.0/ 6029 4. http://gcc.gnu.org/gcc-5/changes.html 6030 5. http://gcc.gnu.org/onlinedocs/5.4.0/ 6031 6. http://gcc.gnu.org/gcc-5/changes.html 6032 7. http://gcc.gnu.org/onlinedocs/5.3.0/ 6033 8. http://gcc.gnu.org/gcc-5/changes.html 6034 9. http://gcc.gnu.org/onlinedocs/5.2.0/ 6035 10. http://gcc.gnu.org/gcc-5/changes.html 6036 11. http://gcc.gnu.org/onlinedocs/5.1.0/ 6037 12. http://gcc.gnu.org/gcc-5/buildstat.html 6038 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6039 14. http://gcc.gnu.org/index.html 6040 15. mailto:gcc@gcc.gnu.org 6041 16. http://gcc.gnu.org/mirrors.html 6042 17. http://gcc.gnu.org/git.html 6043 18. https://gcc.gnu.org/onlinedocs/ 6044 19. mailto:gcc-help@gcc.gnu.org 6045 20. mailto:gcc@gcc.gnu.org 6046 21. https://gcc.gnu.org/lists.html 6047 22. https://www.fsf.org/ 6048 23. https://gcc.gnu.org/about.html 6049 24. http://validator.w3.org/check/referer 6050====================================================================== 6051http://gcc.gnu.org/gcc-5/changes.html 6052 6053 GCC 5 Release Series 6054 Changes, New Features, and Fixes 6055 6056Caveats 6057 6058 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 6059 * The C++ runtime library (libstdc++) uses a new ABI by default (see 6060 [1]below). 6061 * The Graphite framework for loop optimizations no longer requires 6062 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 6063 The installation manual contains more information about 6064 requirements to build GCC. 6065 * The non-standard C++0x type traits has_trivial_default_constructor, 6066 has_trivial_copy_constructor and has_trivial_copy_assign have been 6067 deprecated and will be removed in a future version. The standard 6068 C++11 traits is_trivially_default_constructible, 6069 is_trivially_copy_constructible and is_trivially_copy_assignable 6070 should be used instead. 6071 * On AVR, support has been added for the devices 6072 ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer. 6073 * The AVR port uses a new scheme to describe supported devices: For 6074 each supported device the compiler provides a device-specific 6075 [2]spec file. If the compiler is used together with AVR-LibC, this 6076 requires at least GCC 5.2 and a version of AVR-LibC which 6077 implements [3]feature #44574. 6078 6079General Optimizer Improvements 6080 6081 * Inter-procedural optimization improvements: 6082 + An Identical Code Folding (ICF) pass (controlled via 6083 -fipa-icf) has been added. Compared to the identical code 6084 folding performed by the Gold linker this pass does not 6085 require function sections. It also performs merging before 6086 inlining, so inter-procedural optimizations are aware of the 6087 code re-use. On the other hand not all unifications performed 6088 by a linker are doable by GCC which must honor aliasing 6089 information. During link-time optimization of Firefox, this 6090 pass unifies about 31000 functions, that is 14% overall. 6091 + The devirtualization pass was significantly improved by adding 6092 better support for speculative devirtualization and dynamic 6093 type detection. About 50% of virtual calls in Firefox are now 6094 speculatively devirtualized during link-time optimization. 6095 + A new comdat localization pass allows the linker to eliminate 6096 more dead code in presence of C++ inline functions. 6097 + Virtual tables are now optimized. Local aliases are used to 6098 reduce dynamic linking time of C++ virtual tables on ELF 6099 targets and data alignment has been reduced to limit data 6100 segment bloat. 6101 + A new -fno-semantic-interposition option can be used to 6102 improve code quality of shared libraries where interposition 6103 of exported symbols is not allowed. 6104 + Write-only variables are now detected and optimized out. 6105 + With profile feedback the function inliner can now bypass 6106 --param inline-insns-auto and --param inline-insns-single 6107 limits for hot calls. 6108 + The IPA reference pass was significantly sped up making it 6109 feasible to enable -fipa-reference with -fprofile-generate. 6110 This also solves a bottleneck seen when building Chromium with 6111 link-time optimization. 6112 + The symbol table and call-graph API was reworked to C++ and 6113 simplified. 6114 + The interprocedural propagation of constants now also 6115 propagates alignments of pointer parameters. This for example 6116 means that the vectorizer often does not need to generate loop 6117 prologues and epilogues to make up for potential 6118 misalignments. 6119 * Link-time optimization improvements: 6120 + One Definition Rule based merging of C++ types has been 6121 implemented. Type merging enables better devirtualization and 6122 alias analysis. Streaming extra information needed to merge 6123 types adds about 2-6% of memory size and object size increase. 6124 This can be controlled by -flto-odr-type-merging. 6125 + Command-line optimization and target options are now streamed 6126 on a per-function basis and honored by the link-time 6127 optimizer. This change makes link-time optimization a more 6128 transparent replacement of per-file optimizations. It is now 6129 possible to build projects that require different optimization 6130 settings for different translation units (such as -ffast-math, 6131 -mavx, or -finline). Contrary to earlier GCC releases, the 6132 optimization and target options passed on the link command 6133 line are ignored. 6134 Note that this applies only to those command-line options that 6135 can be passed to optimize and target attributes. Command-line 6136 options affecting global code generation (such as -fpic), 6137 warnings (such as -Wodr), optimizations affecting the way 6138 static variables are optimized (such as -fcommon), debug 6139 output (such as -g), and --param parameters can be applied 6140 only to the whole link-time optimization unit. In these cases, 6141 it is recommended to consistently use the same options at both 6142 compile time and link time. 6143 + GCC bootstrap now uses slim LTO object files. 6144 + Memory usage and link times were improved. Tree merging was 6145 sped up, memory usage of GIMPLE declarations and types was 6146 reduced, and, support for on-demand streaming of variable 6147 constructors was added. 6148 * Feedback directed optimization improvements: 6149 + A new auto-FDO mode uses profiles collected by low overhead 6150 profiling tools (perf) instead of more expensive program 6151 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 6152 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 6153 traditional feedback directed optimization. 6154 + Profile precision was improved in presence of C++ inline and 6155 extern inline functions. 6156 + The new gcov-tool utility allows manipulating profiles. 6157 + Profiles are now more tolerant to source file changes (this 6158 can be controlled by --param profile-func-internal-id). 6159 * Register allocation improvements: 6160 + A new local register allocator (LRA) sub-pass, controlled by 6161 -flra-remat, implements control-flow sensitive global register 6162 rematerialization. Instead of spilling and restoring a 6163 register value, it is recalculated if it is profitable. The 6164 sub-pass improved SPEC2000 generated code by 1% and 0.5% 6165 correspondingly on ARM and x86-64. 6166 + Reuse of the PIC hard register, instead of using a fixed 6167 register, was implemented on x86/x86-64 targets. This improves 6168 generated PIC code performance as more hard registers can be 6169 used. Shared libraries can significantly benefit from this 6170 optimization. Currently it is switched on only for x86/x86-64 6171 targets. As RA infrastructure is already implemented for PIC 6172 register reuse, other targets might follow this in the future. 6173 + A simple form of inter-procedural RA was implemented. When it 6174 is known that a called function does not use caller-saved 6175 registers, save/restore code is not generated around the call 6176 for such registers. This optimization can be controlled by 6177 -fipa-ra 6178 + LRA is now much more effective at generating spills of general 6179 registers into vector registers instead of memory on 6180 architectures (e.g., modern Intel processors) where this is 6181 profitable. 6182 * UndefinedBehaviorSanitizer gained a few new sanitization options: 6183 + -fsanitize=float-divide-by-zero: detect floating-point 6184 division by zero; 6185 + -fsanitize=float-cast-overflow: check that the result of 6186 floating-point type to integer conversions do not overflow; 6187 + -fsanitize=bounds: enable instrumentation of array bounds and 6188 detect out-of-bounds accesses; 6189 + -fsanitize=alignment: enable alignment checking, detect 6190 various misaligned objects; 6191 + -fsanitize=object-size: enable object size checking, detect 6192 various out-of-bounds accesses. 6193 + -fsanitize=vptr: enable checking of C++ member function calls, 6194 member accesses and some conversions between pointers to base 6195 and derived classes, detect if the referenced object does not 6196 have the correct dynamic type. 6197 * Pointer Bounds Checker, a bounds violation detector, has been added 6198 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 6199 instrumented with run-time checks of used pointers against their 6200 bounds to detect pointer bounds violations (overflows). The Pointer 6201 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 6202 new ISA extension Intel MPX support. See the Pointer Bounds Checker 6203 [4]Wiki page for more details. 6204 6205New Languages and Language specific improvements 6206 6207 * [5]OpenMP 4.0 specification offloading features are now supported 6208 by the C, C++, and Fortran compilers. Generic changes: 6209 + Infrastructure (suitable for any vendor). 6210 + Testsuite which covers offloading from the [6]OpenMP 4.0 6211 Examples document. 6212 Specific for upcoming Intel Xeon Phi products: 6213 + Run-time library. 6214 + Card emulator. 6215 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 6216 specification. OpenACC is intended for programming accelerator 6217 devices such as GPUs. See [7]the OpenACC wiki page for more 6218 information. 6219 6220 C family 6221 6222 * The default setting of the -fdiagnostics-color= command-line option 6223 is now [8]configurable when building GCC using configuration option 6224 --with-diagnostics-color=. The possible values are: never, always, 6225 auto and auto-if-env. The new default auto uses color only when the 6226 standard error is a terminal. The default in GCC 4.9 was 6227 auto-if-env, which is equivalent to auto if there is a non-empty 6228 GCC_COLORS environment variable, and never otherwise. As in GCC 6229 4.9, an empty GCC_COLORS variable in the environment will always 6230 disable colors, no matter what the default is or what command-line 6231 options are used. 6232 * A new command-line option -Wswitch-bool has been added for the C 6233 and C++ compilers, which warns whenever a switch statement has an 6234 index of boolean type. 6235 * A new command-line option -Wlogical-not-parentheses has been added 6236 for the C and C++ compilers, which warns about "logical not" used 6237 on the left hand side operand of a comparison. 6238 * A new command-line option -Wsizeof-array-argument has been added 6239 for the C and C++ compilers, which warns when the sizeof operator 6240 is applied to a parameter that has been declared as an array in a 6241 function definition. 6242 * A new command-line option -Wbool-compare has been added for the C 6243 and C++ compilers, which warns about boolean expressions compared 6244 with an integer value different from true/false. 6245 * Full support for Cilk Plus has been added to the GCC compiler. Cilk 6246 Plus is an extension to the C and C++ languages to support data and 6247 task parallelism. 6248 * A new attribute no_reorder prevents reordering of selected symbols 6249 against other such symbols or inline assembler. This enables to 6250 link-time optimize the Linux kernel without having to resort to 6251 -fno-toplevel-reorder that disables several optimizations. 6252 * New preprocessor constructs, __has_include and __has_include_next, 6253 to test the availability of headers have been added. 6254 This demonstrates a way to include the header <optional> only if it 6255 is available: 6256 6257#ifdef __has_include 6258# if __has_include(<optional>) 6259# include <optional> 6260# define have_optional 1 6261# elif __has_include(<experimental/optional>) 6262# include <experimental/optional> 6263# define have_optional 1 6264# define experimental_optional 6265# else 6266# define have_optional 0 6267# endif 6268#endif 6269 6270 The header search paths for __has_include and __has_include_next 6271 are equivalent to those of the standard directive #include and the 6272 extension #include_next respectively. 6273 * A new built-in function-like macro to determine the existence of an 6274 attribute, __has_attribute, has been added. The equivalent built-in 6275 macro __has_cpp_attribute was added to C++ to support 6276 [9]Feature-testing recommendations for C++. The macro 6277 __has_attribute is added to all C-like languages as an extension: 6278 6279int 6280#ifdef __has_attribute 6281# if __has_attribute(__noinline__) 6282 __attribute__((__noinline__)) 6283# endif 6284#endif 6285foo(int x); 6286 6287 If an attribute exists, a nonzero constant integer is returned. For 6288 standardized C++ attributes a date is returned, otherwise the 6289 constant returned is 1. Both __has_attribute and 6290 __has_cpp_attribute will add underscores to an attribute name if 6291 necessary to resolve the name. For C++11 and onwards the attribute 6292 may be scoped. 6293 * A new set of built-in functions for arithmetics with overflow 6294 checking has been added: __builtin_add_overflow, 6295 __builtin_sub_overflow and __builtin_mul_overflow and for 6296 compatibility with clang also other variants. These builtins have 6297 two integral arguments (which don't need to have the same type), 6298 the arguments are extended to infinite precision signed type, +, - 6299 or * is performed on those, and the result is stored in an integer 6300 variable pointed to by the last argument. If the stored value is 6301 equal to the infinite precision result, the built-in functions 6302 return false, otherwise true. The type of the integer variable that 6303 will hold the result can be different from the types of the first 6304 two arguments. The following snippet demonstrates how this can be 6305 used in computing the size for the calloc function: 6306 6307void * 6308calloc (size_t x, size_t y) 6309{ 6310 size_t sz; 6311 if (__builtin_mul_overflow (x, y, &sz)) 6312 return NULL; 6313 void *ret = malloc (sz); 6314 if (ret) memset (res, 0, sz); 6315 return ret; 6316} 6317 6318 On e.g. i?86 or x86-64 the above will result in a mul instruction 6319 followed by a jump on overflow. 6320 * The option -fextended-identifiers is now enabled by default for 6321 C++, and for C99 and later C versions. Various bugs in the 6322 implementation of extended identifiers have been fixed. 6323 6324 C 6325 6326 * The default mode has been changed to -std=gnu11. 6327 * A new command-line option -Wc90-c99-compat has been added to warn 6328 about features not present in ISO C90, but present in ISO C99. 6329 * A new command-line option -Wc99-c11-compat has been added to warn 6330 about features not present in ISO C99, but present in ISO C11. 6331 * It is possible to disable warnings about conversions between 6332 pointers that have incompatible types via a new warning option 6333 -Wno-incompatible-pointer-types; warnings about implicit 6334 incompatible integer to pointer and pointer to integer conversions 6335 via a new warning option -Wno-int-conversion; and warnings about 6336 qualifiers on pointers being discarded via a new warning option 6337 -Wno-discarded-qualifiers. 6338 * To allow proper use of const qualifiers with multidimensional 6339 arrays, GCC will not warn about incompatible pointer types anymore 6340 for conversions between pointers to arrays with and without const 6341 qualifier (except when using -pedantic). Instead, a new warning is 6342 emitted only if the const qualifier is lost. This can be controlled 6343 with a new warning option -Wno-discarded-array-qualifiers. 6344 * The C front end now generates more precise caret diagnostics. 6345 * The -pg command-line option now only affects the current file in an 6346 LTO build. 6347 6348 C++ 6349 6350 * G++ now supports [10]C++14 variable templates. 6351 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 6352 * Excessive template instantiation depth is now a fatal error. This 6353 prevents excessive diagnostics that usually do not help to identify 6354 the problem. 6355 * G++ and libstdc++ now implement the feature-testing macros from 6356 [11]Feature-testing recommendations for C++. 6357 * G++ now allows typename in a template template parameter. 6358 6359template<template<typename> typename X> struct D; // OK 6360 6361 * G++ now supports [12]C++14 aggregates with non-static data member 6362 initializers. 6363 6364struct A { int i, j = i; }; 6365A a = { 42 }; // a.j is also 42 6366 6367 * G++ now supports [13]C++14 extended constexpr. 6368 6369constexpr int f (int i) 6370{ 6371 int j = 0; 6372 for (; i > 0; --i) 6373 ++j; 6374 return j; 6375} 6376 6377constexpr int i = f(42); // i is 42 6378 6379 * G++ now supports the [14]C++14 sized deallocation functions. 6380 6381void operator delete (void *, std::size_t) noexcept; 6382void operator delete[] (void *, std::size_t) noexcept; 6383 6384 * A new One Definition Rule violation warning (controlled by -Wodr) 6385 detects mismatches in type definitions and virtual table contents 6386 during link-time optimization. 6387 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 6388 developers to annotate programs with final specifiers (or anonymous 6389 namespaces) to improve code generation. These warnings can be used 6390 at compile time, but they are more useful in combination with 6391 link-time optimization. 6392 * G++ no longer supports [15]N3639 variable length arrays, as they 6393 were removed from the C++14 working paper prior to ratification. 6394 GNU VLAs are still supported, so VLA support is now the same in 6395 C++14 mode as in C++98 and C++11 modes. 6396 * G++ now allows passing a non-trivially-copyable class via C 6397 varargs, which is conditionally-supported with 6398 implementation-defined semantics in the standard. This uses the 6399 same calling convention as a normal value parameter. 6400 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 6401 various mangling bugs are fixed, but G++ will still emit aliases 6402 with the old, wrong mangling where feasible. -Wabi=2 will warn 6403 about differences between ABI version 2 and the current setting. 6404 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 6405 to be unaffected, but -Wabi=8 will warn about a non-static data 6406 member with type std::nullptr_t which changes position due to this 6407 change. 6408 6409 Runtime Library (libstdc++) 6410 6411 * A [16]Dual ABI is provided by the library. A new ABI is enabled by 6412 default. The old ABI is still supported and can be used by defining 6413 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 6414 standard library headers. 6415 * A new implementation of std::string is enabled by default, using 6416 the small string optimization instead of copy-on-write reference 6417 counting. 6418 * A new implementation of std::list is enabled by default, with an 6419 O(1) size() function; 6420 * [17]Full support for C++11, including the following new features: 6421 + std::deque and std::vector<bool> meet the allocator-aware 6422 container requirements; 6423 + movable and swappable iostream classes; 6424 + support for std::align and std::aligned_union; 6425 + type traits std::is_trivially_copyable, 6426 std::is_trivially_constructible, std::is_trivially_assignable 6427 etc.; 6428 + I/O manipulators std::put_time, std::get_time, std::hexfloat 6429 and std::defaultfloat; 6430 + generic locale-aware std::isblank; 6431 + locale facets for Unicode conversion; 6432 + atomic operations for std::shared_ptr; 6433 + std::notify_all_at_thread_exit() and functions for making 6434 futures ready at thread exit. 6435 * Support for the C++11 hexfloat manipulator changes how the num_put 6436 facet formats floating point types when 6437 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 6438 This change affects all language modes, even though the C++98 6439 standard gave no special meaning to that combination of flags. To 6440 prevent the use of hexadecimal notation for floating point types 6441 use str.unsetf(std::ios_base::floatfield) to clear the relevant 6442 bits in str.flags(). 6443 * [18]Full experimental support for C++14, including the following 6444 new features: 6445 + std::is_final type trait; 6446 + heterogeneous comparison lookup in associative containers. 6447 + global functions cbegin, cend, rbegin, rend, crbegin, and 6448 crend for range access to containers, arrays and initializer 6449 lists. 6450 * [19]Improved experimental support for the Library Fundamentals TS, 6451 including: 6452 + class std::experimental::any; 6453 + function template std::experimental::apply; 6454 + function template std::experimental::sample; 6455 + function template std::experimental::search and related 6456 searcher types; 6457 + variable templates for type traits; 6458 + function template std::experimental::not_fn. 6459 * New random number distributions logistic_distribution and 6460 uniform_on_sphere_distribution as extensions. 6461 * [20]GDB Xmethods for containers and std::unique_ptr. 6462 6463 Fortran 6464 6465 * Compatibility notice: 6466 + The version of the module files (.mod) has been incremented. 6467 + For free-form source files [21]-Werror=line-truncation is now 6468 enabled by default. Note that comments exceeding the line 6469 length are not diagnosed. (For fixed-form source code, the 6470 same warning is available but turned off by default, such that 6471 excess characters are ignored. -ffree-line-length-n and 6472 -ffixed-line-length-n can be used to modify the default line 6473 lengths of 132 and 72 columns, respectively.) 6474 + The -Wtabs option is now more sensible: with -Wtabs the 6475 compiler warns if it encounters tabs and with -Wno-tabs this 6476 warning is turned off. Before, -Wno-tabs warned and -Wtabs 6477 disabled the warning. As before, this warning is also enabled 6478 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 6479 options of -std=. 6480 * Incomplete support for colorizing diagnostics emitted by gfortran 6481 has been added. The option [22]-fdiagnostics-color controls when 6482 color is used in diagnostics. The default value of this option can 6483 be [23]configured when building GCC. The GCC_COLORS environment 6484 variable can be used to customize the colors or disable coloring 6485 completely. Sample diagnostics output: 6486 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 6487 test.f90:6:1: 6488 6489 0 continue 6490 1 6491 Error: Zero is not a valid statement label at (1) 6492 test.f90:9:6: 6493 6494 USE foo 6495 1 6496 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 6497 6498 * The -Wuse-without-only option has been added to warn when a USE 6499 statement has no ONLY qualifier and thus implicitly imports all 6500 public entities of the used module. 6501 * Formatted READ and WRITE statements now work correctly in 6502 locale-aware programs. For more information and potential caveats, 6503 see [24]Section 5.3 Thread-safety of the runtime library in the 6504 manual. 6505 * [25]Fortran 2003: 6506 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 6507 IEEE_ARITHMETIC) are now supported. 6508 * [26]Fortran 2008: 6509 + [27]Coarrays: Full experimental support of Fortran 2008's 6510 coarrays with -fcoarray=lib except for allocatable/pointer 6511 components of derived-type coarrays. GCC currently only ships 6512 with a single-image library (libcaf_single), but multi-image 6513 support based on MPI and GASNet is provided by the libraries 6514 of the [28]OpenCoarrays project. 6515 * TS18508 Additional Parallel Features in Fortran: 6516 + Support for the collective intrinsic subroutines CO_MAX, 6517 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 6518 including -fcoarray=lib support. 6519 + Support for the new atomic intrinsics has been added, 6520 including -fcoarray=lib support. 6521 * Fortran 2015: 6522 + Support for IMPLICIT NONE (external, type). 6523 + ERROR STOP is now permitted in pure procedures. 6524 6525 Go 6526 6527 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 6528 * Building GCC 5 with Go enabled will install two new programs: 6529 [29]go and [30]gofmt. 6530 6531libgccjit 6532 6533 New in GCC 5 is the ability to build GCC as a shared library for 6534 embedding in other processes (such as interpreters), suitable for 6535 Just-In-Time compilation to machine code. 6536 6537 The shared library has a [31]C API and a [32]C++ wrapper API providing 6538 some "syntactic sugar". There are also bindings available from 3rd 6539 parties for [33]Python and for [34]D. 6540 6541 For example, this library can be used by interpreters for [35]compiling 6542 functions from bytecode to machine code. 6543 6544 The library can also be used for ahead-of-time compilation, enabling 6545 GCC to be plugged into a pre-existing front end. An example of using 6546 this to build a compiler for an esoteric language we'll refer to as 6547 "brainf" can be seen [36]here. 6548 6549 libgccjit is licensed under the GPLv3 (or at your option, any later 6550 version) 6551 6552 It should be regarded as experimental at this time. 6553 6554New Targets and Target Specific Improvements 6555 6556 Reporting stack usage 6557 6558 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 6559 -fstack-usage option. 6560 6561 AArch64 6562 6563 * Code generation for the ARM Cortex-A57 processor has been improved. 6564 A more accurate instruction scheduling model for the processor is 6565 now used, and a number of compiler tuning parameters have been set 6566 to offer increased performance when compiling with -mcpu=cortex-a57 6567 or -mtune=cortex-a57. 6568 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 6569 and can be enabled by giving the -mfix-cortex-a53-835769 option. 6570 Alternatively it can be enabled by default by configuring GCC with 6571 the --enable-fix-cortex-a53-835769 option. 6572 * The optional cryptographic extensions to the ARMv8-A architecture 6573 are no longer enabled by default when specifying the 6574 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 6575 options. To enable these extensions add +crypto to the value of 6576 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 6577 * Support has been added for the following processors (GCC 6578 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 6579 initial support for its big.LITTLE combination with the ARM 6580 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 6581 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 6582 The GCC identifiers can be used as arguments to the -mcpu or -mtune 6583 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 6584 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 6585 support for the Cortex-A72. 6586 * The transitional options -mlra and -mno-lra have been removed. The 6587 AArch64 backend now uses the local register allocator (LRA) only. 6588 6589 ARM 6590 6591 * Thumb-1 assembly code is now generated in unified syntax. The new 6592 option -masm-syntax-unified specifies whether inline assembly code 6593 is using unified syntax. By default the option is off which means 6594 non-unified syntax is used. However this is subject to change in 6595 future releases. Eventually the non-unified syntax will be 6596 deprecated. 6597 * It is now a configure-time error to use the --with-cpu configure 6598 option with either of --with-tune or --with-arch. 6599 * Code generation for the ARM Cortex-A57 processor has been improved. 6600 A more accurate instruction scheduling model for the processor is 6601 now used, and a number of compiler tuning parameters have been set 6602 to offer increased performance when compiling with -mcpu=cortex-a57 6603 or -mtune=cortex-a57. 6604 * Support has been added for the following processors (GCC 6605 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 6606 initial support for its big.LITTLE combination with the ARM 6607 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 6608 initial support for its big.LITTLE combination with the ARM 6609 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 6610 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 6611 The GCC identifiers can be used as arguments to the -mcpu or -mtune 6612 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 6613 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 6614 support for the Cortex-A72. 6615 * The deprecated option -mwords-little-endian has been removed. 6616 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 6617 -mtpcs-leaf-frame which are only applicable to the old ABI have 6618 been deprecated. 6619 * The transitional options -mlra and -mno-lra have been removed. The 6620 ARM backend now uses the local register allocator (LRA) only. 6621 6622 AVR 6623 6624 * The compiler no more supports individual devices like ATmega8. 6625 Specifying, say, -mmcu=atmega8 triggers the usage of the 6626 device-specific [37]spec file specs-atmega8 which is part of the 6627 installation and describes options for the sub-processes like 6628 compiler proper, assembler and linker. You can add support for a 6629 new device -mmcu=mydevice as follows: 6630 1. In an empty directory /someplace, create a new directory 6631 device-specs. 6632 2. Copy a device spec file from the installed device-specs 6633 folder, follow the comments in that file and then save it as 6634 /someplace/device-specs/specs-mydevice. 6635 3. Add -B /someplace -mmcu=mydevice to the compiler's 6636 command-line options. Notice that /someplace must specify an 6637 absolute path and that mydevice must not start with "avr". 6638 4. Provided you have a device-specific library libmydevice.a 6639 available, you can put it at /someplace, dito for a 6640 device-specific startup file crtmydevice.o. 6641 The contents of the device spec files depend on the compiler's 6642 configuration, in particular on --with-avrlibc=no and whether or 6643 not it is configured for RTEMS. 6644 * A new command-line option -nodevicelib has been added. It prevents 6645 the compiler from linking against AVR-LibC's device-specific 6646 library libdevice.a. 6647 * The following three command-line options have been added: 6648 6649 -mrmw 6650 Set if the device supports the read-modify-write 6651 instructions LAC, LAS, LAT and XCH. 6652 6653 -mn-flash=size 6654 Specify the flash size of the device in units of 64 KiB, 6655 rounded up to the next integer as needed. This option 6656 affects the availability of the [38]AVR address-spaces. 6657 6658 -mskip-bug 6659 Set if the device is affected by the respective silicon 6660 bug. 6661 6662 In general, you don't need to set these options by hand. The new 6663 device-specific spec file will set them as needed. 6664 6665 IA-32/x86-64 6666 6667 * New [39]ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of 6668 Intel's CPU codenamed Skylake Server was added to GCC. That 6669 includes inline assembly support, new intrinsics, and basic 6670 autovectorization. These new AVX-512 extensions are available via 6671 the following GCC switches: AVX-512 Vector Length EVEX feature: 6672 -mavx512vl, AVX-512 Byte and Word instructions: -mavx512bw, AVX-512 6673 Dword and Qword instructions: -mavx512dq, AVX-512 FMA-52 6674 instructions: -mavx512ifma and for AVX-512 Vector Bit Manipulation 6675 Instructions: -mavx512vbmi. 6676 * New ISA extensions support Intel MPX was added to GCC. This new 6677 extension is available via the -mmpx compiler switch. Intel MPX is 6678 a set of processor features which, with compiler, run-time library 6679 and OS support, brings increased robustness to software by run-time 6680 checking pointer references against their bounds. In GCC Intel MPX 6681 is supported by Pointer Bounds Checker and libmpx run-time 6682 libraries. 6683 * The new -mrecord-mcount option for -pg generates a Linux kernel 6684 style table of pointers to mcount or __fentry__ calls at the 6685 beginning of functions. The new -mnop-mcount option in addition 6686 also generates nops in place of the __fentry__ or mcount call, so 6687 that a call per function can be later patched in. This can be used 6688 for low overhead tracing or hot code patching. 6689 * The new -malign-data option controls how GCC aligns variables. 6690 -malign-data=compat uses increased alignment compatible with GCC 6691 4.8 and earlier, -malign-data=abi uses alignment as specified by 6692 the psABI, and -malign-data=cacheline uses increased alignment to 6693 match the cache line size. -malign-data=compat is the default. 6694 * The new -mskip-rax-setup option skips setting up the RAX register 6695 when SSE is disabled and there are no variable arguments passed in 6696 vector registers. This can be used to optimize the Linux kernel. 6697 6698 MIPS 6699 6700 * MIPS Releases 3 and 5 are now directly supported. Use the 6701 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 6702 to enable code-generation for these processors. 6703 * The Imagination P5600 processor is now supported using the 6704 -march=p5600 command-line option. 6705 * The Cavium Octeon3 processor is now supported using the 6706 -march=octeon3 command-line option. 6707 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 6708 command-line options. 6709 * The o32 ABI has been modified and extended. The o32 64-bit 6710 floating-point register support is now obsolete and has been 6711 removed. It has been replaced by three ABI extensions FPXX, FP64A, 6712 and FP64. The meaning of the -mfp64 command-line option has 6713 changed. It is now used to enable the FP64A and FP64 ABI 6714 extensions. 6715 + The FPXX extension requires that code generated to access 6716 double-precision values use even-numbered registers. Code that 6717 adheres to this extension is link-compatible with all other 6718 o32 double-precision ABI variants and will execute correctly 6719 in all hardware FPU modes. The command-line options -mabi=32 6720 -mfpxx can be used to enable this extension. MIPS II is the 6721 minimum processor required. 6722 + The o32 FP64A extension requires that floating-point registers 6723 be 64-bit and odd-numbered single-precision registers are not 6724 allowed. Code that adheres to the o32 FP64A variant is 6725 link-compatible with all other o32 double-precision ABI 6726 variants. The command-line options -mabi=32 -mfp64 6727 -mno-odd-spreg can be used to enable this extension. MIPS32R2 6728 is the minimum processor required. 6729 + The o32 FP64 extension also requires that floating-point 6730 registers be 64-bit, but permits the use of single-precision 6731 registers. Code that adheres to the o32 FP64 variant is 6732 link-compatible with o32 FPXX and o32 FP64A variants only, 6733 i.e. it is not compatible with the original o32 6734 double-precision ABI. The command-line options -mabi=32 -mfp64 6735 -modd-spreg can be used to enable this extension. MIPS32R2 is 6736 the minimum processor required. 6737 The new ABI variants can be enabled by default using the configure 6738 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 6739 It is strongly recommended that all vendors begin to set o32 FPXX 6740 as the default ABI. This will be required to run the generated code 6741 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 6742 MIPSR6 cores. 6743 * GCC will now pass all floating-point options to the assembler if 6744 GNU binutils 2.25 is used. As a result, any inline assembly code 6745 that uses hard-float instructions should be amended to include a 6746 .set directive to override the global assembler options when 6747 compiling for soft-float targets. 6748 6749 NDS32 6750 6751 * The variadic function ABI implementation is now compatible with 6752 past Andes toolchains where the caller uses registers to pass 6753 arguments and the callee is in charge of pushing them on stack. 6754 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 6755 been removed since they are not yet available in the nds32 port of 6756 GNU binutils. 6757 * A new option -mcmodel=[small|medium|large] supports varied code 6758 models on code generation. The -mgp-direct option became 6759 meaningless and can be discarded. 6760 6761 RX 6762 6763 * A new command line option -mno-allow-string-insns can be used to 6764 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 6765 SWHILE and RMPA instructions. An erratum released by Renesas shows 6766 that it is unsafe to use these instructions on addresses within the 6767 I/O space of the processor. The new option can be used when the 6768 programmer is concerned that the I/O space might be accessed. The 6769 default is still to enable these instructions. 6770 6771 SH 6772 6773 * The compiler will now pass the appropriate --isa= option to the 6774 assembler. 6775 * The default handling for the GBR has been changed from call 6776 clobbered to call preserved. The old behavior can be reinstated by 6777 specifying the option -fcall-used-gbr. 6778 * Support for the SH4A fpchg instruction has been added which will be 6779 utilized when switching between single and double precision FPU 6780 modes. 6781 * The compiler no longer uses the __fpscr_values array for switching 6782 between single and double FPU precision modes on non-SH4A targets. 6783 Instead mode switching will now be performed by storing, modifying 6784 and reloading the FPSCR, so that other FPSCR bits are preserved 6785 across mode switches. The __fpscr_values array that is defined in 6786 libgcc is still present for backwards compatibility, but it will 6787 not be referenced by compiler generated code anymore. 6788 * New builtin functions __builtin_sh_get_fpscr and 6789 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 6790 function will mask the specified bits in such a way that the SZ, PR 6791 and FR mode bits will be preserved, while changing the other bits. 6792 These new functions do not reference the __fpscr_values array. The 6793 old functions __set_fpscr and __get_fpscr in libgcc which access 6794 the __fpscr_values array are still present for backwards 6795 compatibility, but their usage is highly discouraged. 6796 * Some improvements to code generated for __atomic built-in 6797 functions. 6798 * When compiling for SH2E the compiler will no longer force the usage 6799 of delay slots for conditional branch instructions bt and bf. The 6800 old behavior can be reinstated (e.g. to work around a hardware bug 6801 in the original SH7055) by specifying the new option 6802 -mcbranch-force-delay-slot. 6803 6804Operating Systems 6805 6806 AIX 6807 6808 * GCC now supports stabs debugging continuation lines to allow long 6809 stabs debug information without overflow that generates AIX linker 6810 errors. 6811 6812 DragonFly BSD 6813 6814 * GCC now supports the DragonFly BSD operating system. 6815 6816 FreeBSD 6817 6818 * GCC now supports the FreeBSD operating system for the arm port 6819 through the arm*-*-freebsd* target triplets. 6820 6821 VxWorks MILS 6822 6823 * GCC now supports the MILS (Multiple Independent Levels of Security) 6824 variant of WindRiver's VxWorks operating system for PowerPC 6825 targets. 6826 6827Other significant improvements 6828 6829 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 6830 to set the compiler to use. 6831 6832 * When the new command-line option -freport-bug is used, GCC 6833 automatically generates a developer-friendly reproducer whenever an 6834 internal compiler error is encountered. 6835 6836 GCC 5.2 6837 6838 This is the [40]list of problem reports (PRs) from GCC's bug tracking 6839 system that are known to be fixed in the 5.2 release. This list might 6840 not be complete (that is, it is possible that some PRs that have been 6841 fixed are not listed here). 6842 6843Target Specific Changes 6844 6845 IA-32/x86-64 6846 6847 * Support for new AMD instructions monitorx and mwaitx has been 6848 added. This includes new intrinsic and built-in support. It is 6849 enabled through option -mmwaitx. The instructions monitorx and 6850 mwaitx implement the same functionality as the old monitor and 6851 mwait instructions. In addition mwaitx adds a configurable timer. 6852 The timer value is received as third argument and stored in 6853 register %ebx. 6854 6855 S/390, System z, IBM z Systems 6856 6857 * Support for the IBM z13 processor has been added. When using the 6858 -march=z13 option, the compiler will generate code making use of 6859 the new instructions and registers introduced with the vector 6860 extension facility. The -mtune=z13 option enables z13 specific 6861 instruction scheduling without making use of new instructions. 6862 Compiling code with -march=z13 reduces the default alignment of 6863 vector types bigger than 8 bytes to 8. This is an ABI change and 6864 care must be taken when linking modules compiled with different 6865 arch levels which interchange variables containing vector type 6866 values. For newly compiled code the GNU linker will emit a warning. 6867 * The -mzvector option enables a C/C++ language extension. This 6868 extension provides a new keyword vector which can be used to define 6869 vector type variables. (Note: This is not available when enforcing 6870 strict standard compliance e.g. with -std=c99. Either enable GNU 6871 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 6872 Additionally a set of overloaded builtins is provided which is 6873 partially compatible to the PowerPC Altivec builtins. In order to 6874 make use of these builtins the vecintrin.h header file needs to be 6875 included. 6876 6877 GCC 5.3 6878 6879 This is the [41]list of problem reports (PRs) from GCC's bug tracking 6880 system that are known to be fixed in the 5.3 release. This list might 6881 not be complete (that is, it is possible that some PRs that have been 6882 fixed are not listed here). 6883 6884Target Specific Changes 6885 6886 IA-32/x86-64 6887 6888 * GCC now supports the Intel CPU named Skylake with AVX-512 6889 extensions through -march=skylake-avx512. The switch enables the 6890 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 6891 AVX-512DQ. 6892 6893 S/390, System z, IBM z Systems 6894 6895 * With this version of GCC IBM z Systems support has been added to 6896 the GO runtime environment. GCC 5.3 has proven to be able to 6897 compile larger GO applications on IBM z Systems. 6898 6899 GCC 5.4 6900 6901 This is the [42]list of problem reports (PRs) from GCC's bug tracking 6902 system that are known to be fixed in the 5.4 release. This list might 6903 not be complete (that is, it is possible that some PRs that have been 6904 fixed are not listed here). 6905 6906 GCC 5.5 6907 6908 This is the [43]list of problem reports (PRs) from GCC's bug tracking 6909 system that are known to be fixed in the 5.5 release. This list might 6910 not be complete (that is, it is possible that some PRs that have been 6911 fixed are not listed here). 6912 6913Target Specific Changes 6914 6915 IA-32/x86-64 6916 6917 * Support for the [44]deprecated pcommit instruction has been 6918 removed. 6919 6920 6921 For questions related to the use of GCC, please consult these web 6922 pages and the [45]GCC manuals. If that fails, the 6923 [46]gcc-help@gcc.gnu.org mailing list might help. Comments on these 6924 web pages and the development of GCC are welcome on our developer 6925 list at [47]gcc@gcc.gnu.org. All of [48]our lists have public 6926 archives. 6927 6928 Copyright (C) [49]Free Software Foundation, Inc. Verbatim copying and 6929 distribution of this entire article is permitted in any medium, 6930 provided this notice is preserved. 6931 6932 These pages are [50]maintained by the GCC team. Last modified 6933 2021-07-28[51]. 6934 6935References 6936 6937 1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx 6938 2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 6939 3. https://savannah.nongnu.org/bugs/?44574 6940 4. https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler 6941 5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 6942 6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 6943 7. https://gcc.gnu.org/wiki/OpenACC 6944 8. https://gcc.gnu.org/install/configure.html 6945 9. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 6946 10. http://gcc.gnu.org/projects/cxx1y.html 6947 11. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 6948 12. http://gcc.gnu.org/projects/cxx1y.html 6949 13. http://gcc.gnu.org/projects/cxx1y.html 6950 14. http://gcc.gnu.org/projects/cxx1y.html 6951 15. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 6952 16. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 6953 17. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 6954 18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 6955 19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 6956 20. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html 6957 21. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 6958 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 6959 23. https://gcc.gnu.org/install/configure.html 6960 24. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 6961 25. https://gcc.gnu.org/wiki/Fortran2003Status 6962 26. https://gcc.gnu.org/wiki/Fortran2008Status 6963 27. https://gcc.gnu.org/wiki/Coarray 6964 28. http://www.opencoarrays.org/ 6965 29. https://golang.org/cmd/go/ 6966 30. https://golang.org/cmd/gofmt/ 6967 31. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 6968 32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 6969 33. https://github.com/davidmalcolm/pygccjit 6970 34. https://github.com/ibuclaw/gccjitd 6971 35. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 6972 36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 6973 37. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 6974 38. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 6975 39. https://software.intel.com/sites/default/files/managed/39/c5/325462-sdm-vol-1-2abcd-3abcd.pdf 6976 40. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 6977 41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 6978 42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 6979 43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 6980 44. https://software.intel.com/content/www/us/en/develop/blogs/deprecate-pcommit-instruction.html 6981 45. https://gcc.gnu.org/onlinedocs/ 6982 46. mailto:gcc-help@gcc.gnu.org 6983 47. mailto:gcc@gcc.gnu.org 6984 48. https://gcc.gnu.org/lists.html 6985 49. https://www.fsf.org/ 6986 50. https://gcc.gnu.org/about.html 6987 51. http://validator.w3.org/check/referer 6988====================================================================== 6989http://gcc.gnu.org/gcc-4.9/index.html 6990 6991 GCC 4.9 Release Series 6992 6993 (This release series is no longer supported.) 6994 6995 Aug 3, 2016 6996 6997 The [1]GNU project and the GCC developers are pleased to announce the 6998 release of GCC 4.9.4. 6999 7000 This release is a bug-fix release, containing fixes for regressions in 7001 GCC 4.9.3 relative to previous releases of GCC. 7002 7003Release History 7004 7005 GCC 4.9.4 7006 Aug 3, 2016 ([2]changes, [3]documentation) 7007 7008 GCC 4.9.3 7009 June 26, 2015 ([4]changes, [5]documentation) 7010 7011 GCC 4.9.2 7012 October 30, 2014 ([6]changes, [7]documentation) 7013 7014 GCC 4.9.1 7015 July 16, 2014 ([8]changes, [9]documentation) 7016 7017 GCC 4.9.0 7018 April 22, 2014 ([10]changes, [11]documentation) 7019 7020References and Acknowledgements 7021 7022 GCC used to stand for the GNU C Compiler, but since the compiler 7023 supports several other languages aside from C, it now stands for the 7024 GNU Compiler Collection. 7025 7026 A list of [12]successful builds is updated as new information becomes 7027 available. 7028 7029 The GCC developers would like to thank the numerous people that have 7030 contributed new features, improvements, bug fixes, and other changes as 7031 well as test results to GCC. This [13]amazing group of volunteers is 7032 what makes GCC successful. 7033 7034 For additional information about GCC please refer to the [14]GCC 7035 project web site or contact the [15]GCC development mailing list. 7036 7037 To obtain GCC please use [16]our mirror sites or [17]our version 7038 control system. 7039 7040 7041 For questions related to the use of GCC, please consult these web 7042 pages and the [18]GCC manuals. If that fails, the 7043 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 7044 web pages and the development of GCC are welcome on our developer 7045 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 7046 archives. 7047 7048 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 7049 distribution of this entire article is permitted in any medium, 7050 provided this notice is preserved. 7051 7052 These pages are [23]maintained by the GCC team. Last modified 7053 2021-07-28[24]. 7054 7055References 7056 7057 1. http://www.gnu.org/ 7058 2. http://gcc.gnu.org/gcc-4.9/changes.html 7059 3. http://gcc.gnu.org/onlinedocs/4.9.4/ 7060 4. http://gcc.gnu.org/gcc-4.9/changes.html 7061 5. http://gcc.gnu.org/onlinedocs/4.9.3/ 7062 6. http://gcc.gnu.org/gcc-4.9/changes.html 7063 7. http://gcc.gnu.org/onlinedocs/4.9.2/ 7064 8. http://gcc.gnu.org/gcc-4.9/changes.html 7065 9. http://gcc.gnu.org/onlinedocs/4.9.1/ 7066 10. http://gcc.gnu.org/gcc-4.9/changes.html 7067 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 7068 12. http://gcc.gnu.org/gcc-4.9/buildstat.html 7069 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7070 14. http://gcc.gnu.org/index.html 7071 15. mailto:gcc@gcc.gnu.org 7072 16. http://gcc.gnu.org/mirrors.html 7073 17. http://gcc.gnu.org/git.html 7074 18. https://gcc.gnu.org/onlinedocs/ 7075 19. mailto:gcc-help@gcc.gnu.org 7076 20. mailto:gcc@gcc.gnu.org 7077 21. https://gcc.gnu.org/lists.html 7078 22. https://www.fsf.org/ 7079 23. https://gcc.gnu.org/about.html 7080 24. http://validator.w3.org/check/referer 7081====================================================================== 7082http://gcc.gnu.org/gcc-4.9/changes.html 7083 7084 GCC 4.9 Release Series 7085 Changes, New Features, and Fixes 7086 7087Caveats 7088 7089 * The mudflap run time checker has been removed. The mudflap options 7090 remain, but do nothing. 7091 * Support for a number of older systems and recently unmaintained or 7092 untested target ports of GCC has been declared obsolete in GCC 4.9. 7093 Unless there is activity to revive them, the next release of GCC 7094 will have their sources permanently removed. 7095 The following ports for individual systems on particular 7096 architectures have been obsoleted: 7097 + Solaris 9 (*-*-solaris2.9). Details can be found in the 7098 [1]announcement. 7099 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 7100 float64x1_t exported by arm_neon.h are defined to be the same as 7101 their base types. This results in incorrect application of 7102 parameter passing rules to arguments of types int64x1_t and 7103 uint64x1_t, with respect to the AAPCS64 ABI specification. In 7104 addition, names of C++ functions with parameters of these types 7105 (including float64x1_t) are not mangled correctly. The current 7106 typedef declarations also unintentionally allow implicit casting 7107 between singleton vector types and their base types. These issues 7108 will be resolved in a near future release. See [2]PR60825 for more 7109 information. 7110 7111 More information on porting to GCC 4.9 from previous versions of GCC 7112 can be found in the [3]porting guide for this release. 7113 7114General Optimizer Improvements 7115 7116 * AddressSanitizer, a fast memory error detector, is now available on 7117 ARM. 7118 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 7119 detector, has been added and can be enabled via 7120 -fsanitize=undefined. Various computations will be instrumented to 7121 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 7122 currently available for the C and C++ languages. 7123 * Link-time optimization (LTO) improvements: 7124 + Type merging was rewritten. The new implementation is 7125 significantly faster and uses less memory. 7126 + Better partitioning algorithm resulting in less streaming 7127 during link time. 7128 + Early removal of virtual methods reduces the size of object 7129 files and improves link-time memory usage and compile time. 7130 + Function bodies are now loaded on-demand and released early 7131 improving overall memory usage at link time. 7132 + C++ hidden keyed methods can now be optimized out. 7133 + When using a linker plugin, compiling with the -flto option 7134 now generates slim object files (.o) which only contain 7135 intermediate language representation for LTO. Use 7136 -ffat-lto-objects to create files which contain additionally 7137 the object code. To generate static libraries suitable for LTO 7138 processing, use gcc-ar and gcc-ranlib; to list symbols from a 7139 slim object file use gcc-nm. (This requires that ar, ranlib 7140 and nm have been compiled with plugin support.) 7141 Memory usage building Firefox with debug enabled was reduced from 7142 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 7143 * Inter-procedural optimization improvements: 7144 + New type inheritance analysis module improving 7145 devirtualization. Devirtualization now takes into account 7146 anonymous name-spaces and the C++11 final keyword. 7147 + New speculative devirtualization pass (controlled by 7148 -fdevirtualize-speculatively. 7149 + Calls that were speculatively made direct are turned back to 7150 indirect where direct call is not cheaper. 7151 + Local aliases are introduced for symbols that are known to be 7152 semantically equivalent across shared libraries improving 7153 dynamic linking times. 7154 * Feedback directed optimization improvements: 7155 + Profiling of programs using C++ inline functions is now more 7156 reliable. 7157 + New time profiling determines typical order in which functions 7158 are executed. 7159 + A new function reordering pass (controlled by 7160 -freorder-functions) significantly reduces startup time of 7161 large applications. Until binutils support is completed, it is 7162 effective only with link-time optimization. 7163 + Feedback driven indirect call removal and devirtualization now 7164 handle cross-module calls when link-time optimization is 7165 enabled. 7166 7167New Languages and Language specific improvements 7168 7169 * Version 4.0 of the [4]OpenMP specification is now supported in the 7170 C and C++ compilers and starting with the 4.9.1 release also in the 7171 Fortran compiler. The new -fopenmp-simd option can be used to 7172 enable OpenMP's SIMD directives while ignoring other OpenMP 7173 directives. The new [5]-fsimd-cost-model= option permits to tune 7174 the vectorization cost model for loops annotated with OpenMP and 7175 Cilk Plus simd directives. -Wopenmp-simd warns when the current 7176 cost model overrides simd directives set by the user. 7177 * The -Wdate-time option has been added for the C, C++ and Fortran 7178 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 7179 macros are used. Those macros might prevent bit-wise-identical 7180 reproducible compilations. 7181 7182 Ada 7183 7184 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 7185 7186 C family 7187 7188 * Support for colorizing diagnostics emitted by GCC has been added. 7189 The [6]-fdiagnostics-color=auto will enable it when outputting to 7190 terminals, -fdiagnostics-color=always unconditionally. The 7191 GCC_COLORS environment variable can be used to customize the colors 7192 or disable coloring. If GCC_COLORS variable is present in the 7193 environment, the default is -fdiagnostics-color=auto, otherwise 7194 -fdiagnostics-color=never. 7195 Sample diagnostics output: 7196 $ g++ -fdiagnostics-color=always -S -Wall test.C 7197 test.C: In function `int foo()': 7198 test.C:1:14: warning: no return statement in function returning non-void [-W 7199return-type] 7200 int foo () { } 7201 ^ 7202 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 7203 -ftemplate-depth= to increase the maximum) instantiating `struct X<100>' 7204 template <int N> struct X { static const int value = X<N-1>::value; }; temp 7205late struct X<1000>; 7206 ^ 7207 test.C:2:46: recursively required from `const int X<999>::value' 7208 test.C:2:46: required from `const int X<1000>::value' 7209 test.C:2:88: required from here 7210 7211 test.C:2:46: error: incomplete type `X<100>' used in nested name specifier 7212 7213 * With the new [7]#pragma GCC ivdep, the user can assert that there 7214 are no loop-carried dependencies which would prevent concurrent 7215 execution of consecutive iterations using SIMD (single instruction 7216 multiple data) instructions. 7217 * Support for Cilk Plus has been added and can be enabled with the 7218 -fcilkplus option. Cilk Plus is an extension to the C and C++ 7219 languages to support data and task parallelism. The present 7220 implementation follows ABI version 1.2; all features but _Cilk_for 7221 have been implemented. 7222 7223 C 7224 7225 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 7226 <stdatomic.h> header) are now supported. 7227 * ISO C11 generic selections (_Generic keyword) are now supported. 7228 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 7229 __thread) is now supported. 7230 * ISO C11 support is now at a similar level of completeness to ISO 7231 C99 support: substantially complete modulo bugs, extended 7232 identifiers (supported except for corner cases when 7233 -fextended-identifiers is used), floating-point issues (mainly but 7234 not entirely relating to optional C99 features from Annexes F and 7235 G) and the optional Annexes K (Bounds-checking interfaces) and L 7236 (Analyzability). 7237 * A new C extension __auto_type provides a subset of the 7238 functionality of C++11 auto in GNU C. 7239 7240 C++ 7241 7242 * The G++ implementation of [8]C++1y return type deduction for normal 7243 functions has been updated to conform to [9]N3638, the proposal 7244 accepted into the working paper. Most notably, it adds 7245 decltype(auto) for getting decltype semantics rather than the 7246 template argument deduction semantics of plain auto: 7247 7248int& f(); 7249 auto i1 = f(); // int 7250decltype(auto) i2 = f(); // int& 7251 7252 * G++ supports [10]C++1y lambda capture initializers: 7253 7254[x = 42]{ ... }; 7255 7256 Actually, they have been accepted since GCC 4.5, but now the 7257 compiler doesn't warn about them with -std=c++1y, and supports 7258 parenthesized and brace-enclosed initializers as well. 7259 * G++ supports [11]C++1y variable length arrays. G++ has supported 7260 GNU/C99-style VLAs for a long time, but now additionally supports 7261 initializers and lambda capture by reference. In C++1y mode G++ 7262 will complain about VLA uses that are not permitted by the draft 7263 standard, such as forming a pointer to VLA type or applying sizeof 7264 to a VLA variable. Note that it now appears that VLAs will not be 7265 part of C++14, but will be part of a separate document and then 7266 perhaps C++17. 7267 7268void f(int n) { 7269 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 7270 [&a]{ for (int i : a) { cout << i << endl; } }(); 7271 &a; // error, taking address of VLA 7272} 7273 7274 * G++ supports the [12]C++1y [[deprecated]] attribute modulo bugs in 7275 the underlying [[gnu::deprecated]] attribute. Classes and functions 7276 can be marked deprecated and a diagnostic message added: 7277 7278class A; 7279int bar(int n); 7280#if __cplusplus > 201103 7281class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 7282[[deprecated("bar is unsafe; use foo() instead")]] 7283int bar(int n); 7284 7285int foo(int n); 7286class B; 7287#endif 7288A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 7289int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 7290o() instead 7291 7292 * G++ supports [13]C++1y digit separators. Long numeric literals can 7293 be subdivided with a single quote ' to enhance readability: 7294 7295int i = 1048576; 7296int j = 1'048'576; 7297int k = 0x10'0000; 7298int m = 0'004'000'000; 7299int n = 0b0001'0000'0000'0000'0000'0000; 7300 7301double x = 1.602'176'565e-19; 7302double y = 1.602'176'565e-1'9; 7303 7304 * G++ supports [14]C++1y generic (polymorphic) lambdas. 7305 7306// a functional object that will increment any type 7307auto incr = [](auto x) { return x++; }; 7308 7309 * As a GNU extension, G++ supports explicit template parameter syntax 7310 for generic lambdas. This can be combined in the expected way with 7311 the standard auto syntax. 7312 7313// a functional object that will add two like-type objects 7314auto add = [] <typename T> (T a, T b) { return a + b; }; 7315 7316 * G++ supports unconstrained generic functions as specified by �4.1.2 7317 and �5.1.1 of [15]N3889: Concepts Lite Specification. Briefly, auto 7318 may be used as a type-specifier in a parameter declaration of any 7319 function declarator in order to introduce an implicit function 7320 template parameter, akin to generic lambdas. 7321 7322// the following two function declarations are equivalent 7323auto incr(auto x) { return x++; } 7324template <typename T> 7325auto incr(T x) { return x++; } 7326 7327 Runtime Library (libstdc++) 7328 7329 * [16]Improved support for C++11, including: 7330 + support for <regex>; 7331 + The associative containers in <map> and <set> and the 7332 unordered associative containers in <unordered_map> and 7333 <unordered_set> meet the allocator-aware container 7334 requirements; 7335 * [17]Improved experimental support for the upcoming ISO C++ 7336 standard, C++14, including: 7337 + fixing constexpr member functions without const; 7338 + implementation of the std::exchange() utility function; 7339 + addressing tuples by type; 7340 + implemention of std::make_unique; 7341 + implemention of std::shared_lock; 7342 + making std::result_of SFINAE-friendly; 7343 + adding operator() to std::integral_constant; 7344 + adding user-defined literals for standard library types 7345 std::basic_string, std::chrono::duration, and std::complex; 7346 + adding two range overloads to non-modifying sequence oprations 7347 std::equal and std::mismatch; 7348 + adding IO manipulators for quoted strings; 7349 + adding constexpr members to <utility>, <complex>, <chrono>, 7350 and some containers; 7351 + adding compile-time std::integer_sequence; 7352 + adding cleaner transformation traits; 7353 + making <functional>s operator functors easier to use and more 7354 generic; 7355 * An implementation of std::experimental::optional. 7356 * An implementation of std::experimental::string_view. 7357 * The non-standard function std::copy_exception has been deprecated 7358 and will be removed in a future version. std::make_exception_ptr 7359 should be used instead. 7360 7361 Fortran 7362 7363 * Compatibility notice: 7364 + Module files: The version of the module files (.mod) has been 7365 incremented; additionally, module files are now compressed. 7366 Fortran MODULEs compiled by earlier GCC versions have to be 7367 recompiled, when they are USEd by files compiled with GCC 4.9. 7368 GCC 4.9 is not able to read .mod files of earlier GCC 7369 versions; attempting to do so gives an error message. Note: 7370 The ABI of the produced assembler data itself has not changed: 7371 object files and libraries are fully compatible with older 7372 versions (except as stated below). 7373 + ABI changes: 7374 o The [18]argument passing ABI has changed for scalar dummy 7375 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 7376 which have both the VALUE and the OPTIONAL attributes. 7377 o To support finalization the virtual table associated with 7378 polymorphic variables has changed. Code containing CLASS 7379 should be recompiled, including all files which define 7380 derived types involved in the type definition used by 7381 polymorphic variables. (Note: Due to the incremented 7382 module version, trying to mix old code with new code will 7383 usually give an error message.) 7384 + GNU Fortran no longer deallocates allocatable variables or 7385 allocatable components of variables declared in the main 7386 program. Since Fortran 2008, the standard explicitly states 7387 that variables declared in the Fortran main program 7388 automatically have the SAVE attribute. 7389 + When opening files, the close-on-exec flag is set if the 7390 system supports such a feature. This is generally considered 7391 good practice these days, but if there is a need to pass file 7392 descriptors to child processes the parent process must now 7393 remember to clear the close-on-exec flag by calling fcntl(), 7394 e.g. via ISO_C_BINDING, before executing the child process. 7395 * The deprecated command-line option -fno-whole-file has been 7396 removed. (-fwhole-file is the default since GCC 4.6.) 7397 -fwhole-file/-fno-whole-file continue to be accepted but do not 7398 influence the code generation. 7399 * The compiler no longer unconditionally warns about DO loops with 7400 zero iterations. This warning is now controlled by the -Wzerotrip 7401 option, which is implied by -Wall. 7402 * The new NO_ARG_CHECK attribute of the [19]!GCC$ directive can be 7403 used to disable the type-kind-rank (TKR) argument check for a dummy 7404 argument. The feature is similar to ISO/IEC TS 29133:2012's 7405 TYPE(*), except that it additionally also disables the rank check. 7406 Variables with NO_ARG_CHECK have to be dummy arguments and may only 7407 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 7408 to another NO_ARG_CHECK dummy argument; also the other constraints 7409 of TYPE(*) apply. The dummy arguments should be declared as scalar 7410 or assumed-size variable of type type(*) (recommended) - or of type 7411 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 7412 the data without further type or shape information is passed, 7413 similar to C's void*. Note that also TS 29113's 7414 type(*),dimension(..) accepts arguments of any type and rank; 7415 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 7416 descriptor which contains the array shape and stride of the 7417 argument. 7418 * [20]Fortran 2003: 7419 + Finalization is now supported. It is currently only done for a 7420 subset of those situations in which it should occur. 7421 + Experimental support for scalar character components with 7422 deferred length (i.e. allocatable string length) in derived 7423 types has been added. (Deferred-length character variables are 7424 supported since GCC 4.6.) 7425 * [21]Fortran 2008: 7426 + When STOP or ERROR STOP are used to terminate the execution 7427 and any exception (but inexact) is signaling, a warning is 7428 printed to ERROR_UNIT, indicating which exceptions are 7429 signaling. The [22]-ffpe-summary= command-line option can be 7430 used to fine-tune for which exceptions the warning should be 7431 shown. 7432 + Rounding on input (READ) is now handled on systems where 7433 strtod honours the rounding mode. (For output, rounding is 7434 supported since GCC 4.5.) Note that for input, the compatible 7435 rounding mode is handled as nearest (i.e., rounding to an even 7436 least significant [cf. IEC 60559:1989] for a tie, while 7437 compatible rounds away from zero in that case). 7438 7439 Go 7440 7441 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 7442 7443New Targets and Target Specific Improvements 7444 7445 AArch64 7446 7447 * The ARMv8-A crypto and CRC instructions are now supported through 7448 intrinsics. These are enabled when the architecture supports these 7449 and are available through the -march=armv8-a+crc and 7450 -march=armv8-a+crypto options. 7451 * Initial support for ILP32 has now been added to the compiler. This 7452 is now available through the command-line option -mabi=ilp32. 7453 Support for ILP32 is considered experimental as the ABI 7454 specification is still beta. 7455 * Coverage of more of the ISA including the SIMD extensions has been 7456 added. The Advanced SIMD intrinsics have also been improved. 7457 * The new local register allocator (LRA) is now on by default for the 7458 AArch64 backend. 7459 * The REE (Redundant extension elimination) pass has now been enabled 7460 by default for the AArch64 backend. 7461 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 7462 * Initial big.LITTLE tuning support for the combination of Cortex-A57 7463 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 7464 option. 7465 * A number of structural changes have been made to both the ARM and 7466 AArch64 backends to facilitate improved code-generation. 7467 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 7468 has been added and can be enabled by giving the 7469 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 7470 default by configuring GCC with the --enable-fix-cortex-a53-835769 7471 option. 7472 7473 ARC 7474 7475 * A port for Synopsys Designware ARC has been contributed by Embecosm 7476 and Synopsys Inc. 7477 7478 ARM 7479 7480 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 7481 disabled by default. This was found to generate better code in only 7482 a small number of cases. It can be turned back on with the 7483 -mneon-for-64bits option. 7484 * Further support for the ARMv8-A architecture, notably implementing 7485 the restriction around IT blocks in the Thumb32 instruction set has 7486 been added. The -mrestrict-it option can be used with 7487 -march=armv7-a or the -march=armv7ve options to make code 7488 generation fully compatible with the deprecated instructions in 7489 ARMv8-A. 7490 * Support has now been added for the ARMv7ve variant of the 7491 architecture. This can be used by the -march=armv7ve option. 7492 * The ARMv8-A crypto and CRC instructions are now supported through 7493 intrinsics and are available through the -march=armv8-a+crc and 7494 mfpu=crypto-neon-fp-armv8 options. 7495 * LRA is now on by default for the ARM target. This can be turned off 7496 using the -mno-lra option. This option is a purely transitionary 7497 command-line option and will be removed in a future release. We are 7498 interested in any bug reports regarding functional and performance 7499 regressions with LRA. 7500 * A new option -mslow-flash-data to improve performance of programs 7501 fetching data on slow flash memory has now been introduced for the 7502 ARMv7-M profile cores. 7503 * A new option -mpic-data-is-text-relative for targets that allows 7504 data segments to be relative to text segments has been added. This 7505 is on by default for all targets except VxWorks RTP. 7506 * A number of infrastructural changes have been made to both the ARM 7507 and AArch64 backends to facilitate improved code-generation. 7508 * GCC now supports Cortex-A12 and the Cortex-R7 through the 7509 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 7510 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 7511 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 7512 * Initial big.LITTLE tuning support for the combination of Cortex-A57 7513 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 7514 option. Similar support was added for the combination of Cortex-A15 7515 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 7516 * Further performance optimizations for the Cortex-A15 and the 7517 Cortex-M4 have been added. 7518 * A number of code generation improvements for Thumb2 to reduce code 7519 size when compiling for the M-profile processors. 7520 7521 AVR 7522 7523 * A new command-line option -mfract-convert-truncate has been added. 7524 It allows compiler to use truncation instead of rounding towards 7525 zero for fractional fixed-point types. 7526 7527 IA-32/x86-64 7528 7529 * -mfpmath=sse is now implied by -ffast-math on all targets where 7530 SSE2 is supported. 7531 * Intel AVX-512 support was added to GCC. That includes inline 7532 assembly support, new registers and extending existing ones, new 7533 intrinsics (covered by corresponding testsuite), and basic 7534 autovectorization. AVX-512 instructions are available via the 7535 following GCC switches: AVX-512 foundation instructions: -mavx512f, 7536 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 7537 reciprocal instructions: -mavx512er, AVX-512 conflict detection 7538 instructions: -mavx512cd. 7539 * It is now possible to call x86 intrinsics from select functions in 7540 a file that are tagged with the corresponding target attribute 7541 without having to compile the entire file with the -mxxx option. 7542 This improves the usability of x86 intrinsics and is particularly 7543 useful when doing [23]Function Multiversioning. 7544 * GCC now supports the new Intel microarchitecture named Silvermont 7545 through -march=silvermont. 7546 * GCC now supports the new Intel microarchitecture named Broadwell 7547 through -march=broadwell. 7548 * Optimizing for other Intel microarchitectures have been renamed to 7549 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 7550 * -march=generic has been retuned for better support of Intel core 7551 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 7552 Pentium-M, and Pentium4 based CPUs is no longer considered 7553 important for generic. 7554 * -mtune=intel can now be used to generate code running well on the 7555 most current Intel processors, which are Haswell and Silvermont for 7556 GCC 4.9. 7557 * Support to encode 32-bit assembly instructions in 16-bit format is 7558 now available through the -m16 command-line option. 7559 * Better inlining of memcpy and memset that is aware of value ranges 7560 and produces shorter alignment prologues. 7561 * -mno-accumulate-outgoing-args is now honored when unwind 7562 information is output. Argument accumulation is also now turned off 7563 for portions of programs optimized for size. 7564 * Support for new AMD family 15h processors (Excavator core) is now 7565 available through the -march=bdver4 and -mtune=bdver4 options. 7566 7567 MSP430 7568 7569 * A new command-line option -mcpu= has been added to the MSP430 7570 backend. This option is used to specify the ISA to be used. 7571 Accepted values are msp430 (the default), msp430x and msp430xv2. 7572 The ISA is no longer deduced from the -mmcu= option as there are 7573 far too many different MCU names. The -mmcu= option is still 7574 supported, and this is still used to select linker scripts and 7575 generate a C preprocessor symbol that will be recognised by the 7576 msp430.h header file. 7577 7578 NDS32 7579 7580 * A new nds32 port supports the 32-bit architecture from Andes 7581 Technology Corporation. 7582 * The port provides initial support for the V2, V3, V3m instruction 7583 set architectures. 7584 7585 Nios II 7586 7587 * A port for the Altera Nios II has been contributed by Mentor 7588 Graphics. 7589 7590 PowerPC / PowerPC64 / RS6000 7591 7592 * GCC now supports Power ISA 2.07, which includes support for 7593 Hardware Transactional Memory (HTM), Quadword atomics and several 7594 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 7595 integer and decimal integer operations. 7596 * Support for the POWER8 processor is now available through the 7597 -mcpu=power8 and -mtune=power8 options. 7598 * The libitm library has been modified to add a HTM fastpath that 7599 automatically uses POWER's HTM hardware instructions when it is 7600 executing on a HTM enabled processor. 7601 * Support for the new powerpc64le-linux platform has been added. It 7602 defaults to generating code that conforms to the ELFV2 ABI. 7603 7604 S/390, System z 7605 7606 * Support for the Transactional Execution Facility included with the 7607 IBM zEnterprise zEC12 processor has been added. A set of GCC style 7608 builtins as well as XLC style builtins are provided. The builtins 7609 are enabled by default when using the -march=zEC12 option but can 7610 explicitly be disabled with -mno-htm. Using the GCC builtins also 7611 libitm supports hardware transactions on S/390. 7612 * The hotpatch features allows to prepare functions for hotpatching. 7613 A certain amount of bytes is reserved before the function entry 7614 label plus a NOP is inserted at its very beginning to implement a 7615 backward jump when applying a patch. The feature can either be 7616 enabled per compilation unit via the command-line option -mhotpatch 7617 or per function using the hotpatch attribute. 7618 * The shrink wrap optimization is now supported on S/390 and enabled 7619 by default. 7620 * A major rework of the routines to determine which registers need to 7621 be saved and restored in function prologue/epilogue now allow to 7622 use floating point registers as save slots. This will happen for 7623 certain leaf function with -march=z10 or higher. 7624 * The LRA rtl pass replaces reload by default on S/390. 7625 7626 RX 7627 7628 * The port now allows to specify the RX100, RX200, and RX600 7629 processors with the command-line options -mcpu=rx100, -mcpu=rx200 7630 and -mcpu=rx600. 7631 7632 SH 7633 7634 * Minor improvements to code generated for integer arithmetic and 7635 code that involves the T bit. 7636 * Added support for the SH2A clips and clipu instructions. The 7637 compiler will now try to utilize them for min/max expressions such 7638 as max (-128, min (127, x)). 7639 * Added support for the cmp/str instruction through built-in 7640 functions such as __builtin_strlen. When not optimizing for size, 7641 the compiler will now expand calls to e.g. strlen as an inlined 7642 sequences which utilize the cmp/str instruction. 7643 * Improved code generated around volatile memory loads and stores. 7644 * The option -mcbranchdi has been deprecated. Specifying it will 7645 result in a warning and will not influence code generation. 7646 * The option -mcmpeqdi has been deprecated. Specifying it will result 7647 in a warning and will not influence code generation. 7648 7649GCC 4.9.1 7650 7651 This is the [24]list of problem reports (PRs) from GCC's bug tracking 7652 system that are known to be fixed in the 4.9.1 release. This list might 7653 not be complete (that is, it is possible that some PRs that have been 7654 fixed are not listed here). 7655 7656 Version 4.0 of the OpenMP specification is supported even in Fortran, 7657 not just C and C++. 7658 7659GCC 4.9.2 7660 7661 This is the [25]list of problem reports (PRs) from GCC's bug tracking 7662 system that are known to be fixed in the 4.9.2 release. This list might 7663 not be complete (that is, it is possible that some PRs that have been 7664 fixed are not listed here). 7665 7666GCC 4.9.3 7667 7668 This is the [26]list of problem reports (PRs) from GCC's bug tracking 7669 system that are known to be fixed in the 4.9.3 release. This list might 7670 not be complete (that is, it is possible that some PRs that have been 7671 fixed are not listed here). 7672 7673GCC 4.9.4 7674 7675 This is the [27]list of problem reports (PRs) from GCC's bug tracking 7676 system that are known to be fixed in the 4.9.4 release. This list might 7677 not be complete (that is, it is possible that some PRs that have been 7678 fixed are not listed here). 7679 7680 7681 For questions related to the use of GCC, please consult these web 7682 pages and the [28]GCC manuals. If that fails, the 7683 [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these 7684 web pages and the development of GCC are welcome on our developer 7685 list at [30]gcc@gcc.gnu.org. All of [31]our lists have public 7686 archives. 7687 7688 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and 7689 distribution of this entire article is permitted in any medium, 7690 provided this notice is preserved. 7691 7692 These pages are [33]maintained by the GCC team. Last modified 7693 2021-07-28[34]. 7694 7695References 7696 7697 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 7698 2. https://gcc.gnu.org/PR60825 7699 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 7700 4. https://www.openmp.org/specifications/ 7701 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 7702 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 7703 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 7704 8. http://gcc.gnu.org/projects/cxx1y.html 7705 9. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 7706 10. http://gcc.gnu.org/projects/cxx1y.html 7707 11. http://gcc.gnu.org/projects/cxx1y.html 7708 12. http://gcc.gnu.org/projects/cxx1y.html 7709 13. http://gcc.gnu.org/projects/cxx1y.html 7710 14. http://gcc.gnu.org/projects/cxx1y.html 7711 15. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 7712 16. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 7713 17. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 7714 18. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 7715 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 7716 20. https://gcc.gnu.org/wiki/Fortran2003Status 7717 21. https://gcc.gnu.org/wiki/Fortran2008Status 7718 22. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 7719 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 7720 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 7721 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 7722 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 7723 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 7724 28. https://gcc.gnu.org/onlinedocs/ 7725 29. mailto:gcc-help@gcc.gnu.org 7726 30. mailto:gcc@gcc.gnu.org 7727 31. https://gcc.gnu.org/lists.html 7728 32. https://www.fsf.org/ 7729 33. https://gcc.gnu.org/about.html 7730 34. http://validator.w3.org/check/referer 7731====================================================================== 7732http://gcc.gnu.org/gcc-4.8/index.html 7733 7734 GCC 4.8 Release Series 7735 7736 (This release series is no longer supported.) 7737 7738 June 23, 2015 7739 7740 The [1]GNU project and the GCC developers are pleased to announce the 7741 release of GCC 4.8.5. 7742 7743 This release is a bug-fix release, containing fixes for regressions in 7744 GCC 4.8.4 relative to previous releases of GCC. 7745 7746Release History 7747 7748 GCC 4.8.5 7749 June 23, 2015 ([2]changes, [3]documentation) 7750 7751 GCC 4.8.4 7752 December 19, 2014 ([4]changes, [5]documentation) 7753 7754 GCC 4.8.3 7755 May 22, 2014 ([6]changes, [7]documentation) 7756 7757 GCC 4.8.2 7758 October 16, 2013 ([8]changes, [9]documentation) 7759 7760 GCC 4.8.1 7761 May 31, 2013 ([10]changes, [11]documentation) 7762 7763 GCC 4.8.0 7764 March 22, 2013 ([12]changes, [13]documentation) 7765 7766References and Acknowledgements 7767 7768 GCC used to stand for the GNU C Compiler, but since the compiler 7769 supports several other languages aside from C, it now stands for the 7770 GNU Compiler Collection. 7771 7772 A list of [14]successful builds is updated as new information becomes 7773 available. 7774 7775 The GCC developers would like to thank the numerous people that have 7776 contributed new features, improvements, bug fixes, and other changes as 7777 well as test results to GCC. This [15]amazing group of volunteers is 7778 what makes GCC successful. 7779 7780 For additional information about GCC please refer to the [16]GCC 7781 project web site or contact the [17]GCC development mailing list. 7782 7783 To obtain GCC please use [18]our mirror sites or [19]our version 7784 control system. 7785 7786 7787 For questions related to the use of GCC, please consult these web 7788 pages and the [20]GCC manuals. If that fails, the 7789 [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these 7790 web pages and the development of GCC are welcome on our developer 7791 list at [22]gcc@gcc.gnu.org. All of [23]our lists have public 7792 archives. 7793 7794 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 7795 distribution of this entire article is permitted in any medium, 7796 provided this notice is preserved. 7797 7798 These pages are [25]maintained by the GCC team. Last modified 7799 2021-07-28[26]. 7800 7801References 7802 7803 1. http://www.gnu.org/ 7804 2. http://gcc.gnu.org/gcc-4.8/changes.html 7805 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 7806 4. http://gcc.gnu.org/gcc-4.8/changes.html 7807 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 7808 6. http://gcc.gnu.org/gcc-4.8/changes.html 7809 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 7810 8. http://gcc.gnu.org/gcc-4.8/changes.html 7811 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 7812 10. http://gcc.gnu.org/gcc-4.8/changes.html 7813 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 7814 12. http://gcc.gnu.org/gcc-4.8/changes.html 7815 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 7816 14. http://gcc.gnu.org/gcc-4.8/buildstat.html 7817 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7818 16. http://gcc.gnu.org/index.html 7819 17. mailto:gcc@gcc.gnu.org 7820 18. http://gcc.gnu.org/mirrors.html 7821 19. http://gcc.gnu.org/git.html 7822 20. https://gcc.gnu.org/onlinedocs/ 7823 21. mailto:gcc-help@gcc.gnu.org 7824 22. mailto:gcc@gcc.gnu.org 7825 23. https://gcc.gnu.org/lists.html 7826 24. https://www.fsf.org/ 7827 25. https://gcc.gnu.org/about.html 7828 26. http://validator.w3.org/check/referer 7829====================================================================== 7830http://gcc.gnu.org/gcc-4.8/changes.html 7831 7832 GCC 4.8 Release Series 7833 Changes, New Features, and Fixes 7834 7835Caveats 7836 7837 GCC now uses C++ as its implementation language. This means that to 7838 build GCC from sources, you will need a C++ compiler that understands 7839 C++ 2003. For more details on the rationale and specific changes, 7840 please refer to the [1]C++ conversion page. 7841 7842 To enable the Graphite framework for loop optimizations you now need 7843 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 7844 the [2]GCC infrastructure directory. The installation manual contains 7845 more information about requirements to build GCC. 7846 7847 GCC now uses a more aggressive analysis to derive an upper bound for 7848 the number of iterations of loops using constraints imposed by language 7849 standards. This may cause non-conforming programs to no longer work as 7850 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 7851 option, -fno-aggressive-loop-optimizations, was added to disable this 7852 aggressive analysis. In some loops that have known constant number of 7853 iterations, but undefined behavior is known to occur in the loop before 7854 reaching or during the last iteration, GCC will warn about the 7855 undefined behavior in the loop instead of deriving lower upper bound of 7856 the number of iterations for the loop. The warning can be disabled with 7857 -Wno-aggressive-loop-optimizations. 7858 7859 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 7860 for the layout of vectors that could lead to wrong code being 7861 generated. Vectors larger than 8 bytes in size are now by default 7862 aligned to an 8-byte boundary. This is an ABI change: code that makes 7863 explicit use of vector types may be incompatible with binary objects 7864 built with older versions of GCC. Auto-vectorized code is not affected 7865 by this change. 7866 7867 On AVR, support has been removed for the command-line option 7868 -mshort-calls deprecated in GCC 4.7. 7869 7870 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2 7871 is turned on per default for all non-RTEMS configurations. This option 7872 arranges for a better integration of [3]AVR Libc with avr-gcc. For 7873 technical details, see [4]PR54461. To turn off the option in non-RTEMS 7874 configurations, use --with-avrlibc=no. If the compiler is configured 7875 for RTEMS, the option is always turned off. 7876 7877 More information on porting to GCC 4.8 from previous versions of GCC 7878 can be found in the [5]porting guide for this release. 7879 7880General Optimizer Improvements (and Changes) 7881 7882 * DWARF4 is now the default when generating DWARF debug information. 7883 When -g is used on a platform that uses DWARF debugging 7884 information, GCC will now default to -gdwarf-4 7885 -fno-debug-types-section. 7886 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 7887 consumers support DWARF4 by default. Before GCC 4.8 the default 7888 version used was DWARF2. To make GCC 4.8 generate an older DWARF 7889 version use -g together with -gdwarf-2 or -gdwarf-3. The default 7890 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 7891 * A new general optimization level, -Og, has been introduced. It 7892 addresses the need for fast compilation and a superior debugging 7893 experience while providing a reasonable level of run-time 7894 performance. Overall experience for development should be better 7895 than the default optimization level -O0. 7896 * A new option -ftree-partial-pre was added to control the partial 7897 redundancy elimination (PRE) optimization. This option is enabled 7898 by default at the -O3 optimization level, and it makes PRE more 7899 aggressive. 7900 * The option -fconserve-space has been removed; it was no longer 7901 useful on most targets since GCC supports putting variables into 7902 BSS without making them common. 7903 * The struct reorg and matrix reorg optimizations (command-line 7904 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 7905 removed. They did not always work correctly, nor did they work with 7906 link-time optimization (LTO), hence were only applicable to 7907 programs consisting of a single translation unit. 7908 * Several scalability bottle-necks have been removed from GCC's 7909 optimization passes. Compilation of extremely large functions, e.g. 7910 due to the use of the flatten attribute in the "Eigen" C++ linear 7911 algebra templates library, is significantly faster than previous 7912 releases of GCC. 7913 * Link-time optimization (LTO) improvements: 7914 + LTO partitioning has been rewritten for better reliability and 7915 maintanibility. Several important bugs leading to link 7916 failures have been fixed. 7917 * Interprocedural optimization improvements: 7918 + A new symbol table has been implemented. It builds on existing 7919 callgraph and varpool modules and provide a new API. Unusual 7920 symbol visibilities and aliases are handled more consistently 7921 leading to, for example, more aggressive unreachable code 7922 removal with LTO. 7923 + The inline heuristic can now bypass limits on the size of of 7924 inlined functions when the inlining is particularly 7925 profitable. This happens, for example, when loop bounds or 7926 array strides get propagated. 7927 + Values passed through aggregates (either by value or 7928 reference) are now propagated at the inter-procedural level 7929 leading to better inlining decisions (for example in the case 7930 of Fortran array descriptors) and devirtualization. 7931 * [6]AddressSanitizer , a fast memory error detector, has been added 7932 and can be enabled via -fsanitize=address. Memory access 7933 instructions will be instrumented to detect heap-, stack-, and 7934 global-buffer overflow as well as use-after-free bugs. To get nicer 7935 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 7936 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 7937 x86-64 Darwin. 7938 * [7]ThreadSanitizer has been added and can be enabled via 7939 -fsanitize=thread. Instructions will be instrumented to detect data 7940 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 7941 * A new local register allocator (LRA) has been implemented, which 7942 replaces the 26 year old reload pass and improves generated code 7943 quality. For now it is active on the IA-32 and x86-64 targets. 7944 * Support for transactional memory has been implemented on the 7945 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 7946 Alpha. 7947 7948New Languages and Language specific improvements 7949 7950 C family 7951 7952 * Each diagnostic emitted now includes the original source line and a 7953 caret '^' indicating the column. The option 7954 -fno-diagnostics-show-caret suppresses this information. 7955 * The option -ftrack-macro-expansion=2 is now enabled by default. 7956 This allows the compiler to display the macro expansion stack in 7957 diagnostics. Combined with the caret information, an example 7958 diagnostic showing these two features is: 7959 7960t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float 7961') 7962 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 7963_b = (B); __a < __b ? __b : __a; }) 7964 7965 ^ 7966t.c:7:7: note: in expansion of macro 'MYMAX' 7967 X = MYMAX(P, F); 7968 ^ 7969 7970 * A new -Wsizeof-pointer-memaccess warning has been added (also 7971 enabled by -Wall) to warn about suspicious length parameters to 7972 certain string and memory built-in functions if the argument uses 7973 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 7974 (ptr)); if ptr is not an array, but a pointer, and suggests a 7975 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 7976 * The new option -Wpedantic is an alias for -pedantic, which is now 7977 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 7978 -Wno-error=pedantic work in the same way as for any other -W 7979 option. One caveat is that -Werror=pedantic is not equivalent to 7980 -pedantic-errors, since the latter makes into errors some warnings 7981 that are not controlled by -Wpedantic, and the former only affects 7982 diagnostics that are disabled when using -Wno-pedantic. 7983 * The option -Wshadow no longer warns if a declaration shadows a 7984 function declaration, unless the former declares a function or 7985 pointer to function, because this is [8]a common and valid case in 7986 real-world code. 7987 7988 C++ 7989 7990 * G++ now implements the [9]C++11 thread_local keyword; this differs 7991 from the GNU __thread keyword primarily in that it allows dynamic 7992 initialization and destruction semantics. Unfortunately, this 7993 support requires a run-time penalty for references to 7994 non-function-local thread_local variables defined in a different 7995 translation unit even if they don't need dynamic initialization, so 7996 users may want to continue to use __thread for TLS variables with 7997 static initialization semantics. 7998 If the programmer can be sure that no use of the variable in a 7999 non-defining TU needs to trigger dynamic initialization (either 8000 because the variable is statically initialized, or a use of the 8001 variable in the defining TU will be executed before any uses in 8002 another TU), they can avoid this overhead with the 8003 -fno-extern-tls-init option. 8004 OpenMP threadprivate variables now also support dynamic 8005 initialization and destruction by the same mechanism. 8006 * G++ now implements the [10]C++11 attribute syntax, e.g. 8007 8008[[noreturn]] void f(); 8009 8010 and also the alignment specifier, e.g. 8011 8012alignas(double) int i; 8013 8014 * G++ now implements [11]C++11 inheriting constructors, e.g. 8015 8016struct A { A(int); }; 8017struct B: A { using A::A; }; // defines B::B(int) 8018B b(42); // OK 8019 8020 * As of GCC 4.8.1, G++ implements the change to decltype semantics 8021 from [12]N3276. 8022 8023struct A f(); 8024decltype(f()) g(); // OK, return type of f() is not required to be complete. 8025 8026 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g. 8027 8028struct A { int f() &; }; 8029int i = A().f(); // error, f() requires an lvalue object 8030 8031 * G++ now supports a -std=c++1y option for experimentation with 8032 features proposed for the next revision of the standard, expected 8033 around 2014. Currently the only difference from -std=c++11 is 8034 support for return type deduction in normal functions, as proposed 8035 in [14]N3386. Status of C++1y features in GCC 4.8 can be found 8036 [15]here. 8037 * The G++ namespace association extension, __attribute ((strong)), 8038 has been deprecated. Inline namespaces should be used instead. 8039 * G++ now supports a -fext-numeric-literal option to control whether 8040 GNU numeric literal suffixes are accepted as extensions or 8041 processed as C++11 user-defined numeric literal suffixes. The flag 8042 is on (use suffixes for GNU literals) by default for -std=gnu++*, 8043 and -std=c++98. The flag is off (use suffixes for user-defined 8044 literals) by default for -std=c++11 and later. 8045 8046 Runtime Library (libstdc++) 8047 8048 * [16]Improved experimental support for the new ISO C++ standard, 8049 C++11, including: 8050 + forward_list meets the allocator-aware container requirements; 8051 + this_thread::sleep_for(), this_thread::sleep_until() and 8052 this_thread::yield() are defined without requiring the 8053 configure option --enable-libstdcxx-time; 8054 * Improvements to <random>: 8055 + SSE optimized normal_distribution. 8056 + Use of hardware RNG instruction for random_device on new x86 8057 processors (requires the assembler to support the 8058 instruction.) 8059 and <ext/random>: 8060 + New random number engine simd_fast_mersenne_twister_engine 8061 with an optimized SSE implementation. 8062 + New random number distributions beta_distribution, 8063 normal_mv_distribution, rice_distribution, 8064 nakagami_distribution, pareto_distribution, k_distribution, 8065 arcsine_distribution, hoyt_distribution. 8066 * Added --disable-libstdcxx-verbose configure option to disable 8067 diagnostic messages issued when a process terminates abnormally. 8068 This may be useful for embedded systems to reduce the size of 8069 executables that link statically to the library. 8070 8071 Fortran 8072 8073 * Compatibility notice: 8074 + Module files: The version of module files (.mod) has been 8075 incremented. Fortran MODULEs compiled by earlier GCC versions 8076 have to be recompiled, when they are USEd by files compiled 8077 with GCC 4.8. GCC 4.8 is not able to read .mod files created 8078 by earlier versions; attempting to do so gives an error 8079 message. 8080 Note: The ABI of the produced assembler data itself has not 8081 changed; object files and libraries are fully compatible with 8082 older versions except as noted below. 8083 + ABI: Some internal names (used in the assembler/object file) 8084 have changed for symbols declared in the specification part of 8085 a module. If an affected module - or a file using it via use 8086 association - is recompiled, the module and all files which 8087 directly use such symbols have to be recompiled as well. This 8088 change only affects the following kind of module symbols: 8089 o Procedure pointers. Note: C-interoperable function 8090 pointers (type(c_funptr)) are not affected nor are 8091 procedure-pointer components. 8092 o Deferred-length character strings. 8093 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 8094 backtrace at an arbitrary place in user code; program execution 8095 continues normally afterwards. 8096 * The [18]-Wc-binding-type warning option has been added (disabled by 8097 default). It warns if the a variable might not be C interoperable; 8098 in particular, if the variable has been declared using an intrinsic 8099 type with default kind instead of using a kind parameter defined 8100 for C interoperability in the intrinsic ISO_C_Binding module. 8101 Before, this warning was always printed. The -Wc-binding-type 8102 option is enabled by -Wall. 8103 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 8104 options have been added, which diagnose when code is inserted for 8105 automatic (re)allocation of a variable during assignment. This 8106 option can be used to decide whether it is safe to use 8107 [20]-fno-realloc-lhs. Additionally, it can be used to find 8108 automatic (re)allocation in hot loops. (For arrays, replacing 8109 "var=" by "var(:)=" disables the automatic reallocation.) 8110 * The [21]-Wcompare-reals command-line option has been added. When 8111 this is set, warnings are issued when comparing REAL or COMPLEX 8112 types for equality and inequality; consider replacing a == b by 8113 abs(a-b) < eps with a suitable eps. -Wcompare-reals is enabled by 8114 -Wextra. 8115 * The [22]-Wtarget-lifetime command-line option has been added 8116 (enabled with -Wall), which warns if the pointer in a pointer 8117 assignment might outlive its target. 8118 * Reading floating point numbers which use "q" for the exponential 8119 (such as 4.0q0) is now supported as vendor extension for better 8120 compatibility with old data files. It is strongly recommended to 8121 use for I/O the equivalent but standard conforming "e" (such as 8122 4.0e0). 8123 (For Fortran source code, consider replacing the "q" in 8124 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 8125 suitable qp). Note that - in Fortran source code - replacing "q" by 8126 a simple "e" is not equivalent.) 8127 * The GFORTRAN_TMPDIR environment variable for specifying a 8128 non-default directory for files opened with STATUS="SCRATCH", is 8129 not used anymore. Instead gfortran checks the POSIX/GNU standard 8130 TMPDIR environment variable. If TMPDIR is not defined, gfortran 8131 falls back to other methods to determine the directory for 8132 temporary files as documented in the [23]user manual. 8133 * [24]Fortran 2003: 8134 + Support for unlimited polymorphic variables (CLASS(*)) has 8135 been added. Nonconstant character lengths are not yet 8136 supported. 8137 * [25]TS 29113: 8138 + Assumed types (TYPE(*)) are now supported. 8139 + Experimental support for assumed-rank arrays (dimension(..)) 8140 has been added. Note that currently gfortran's own array 8141 descriptor is used, which is different from the one defined in 8142 TS29113, see [26]gfortran's header file or use the [27]Chasm 8143 Language Interoperability Tools. 8144 8145 Go 8146 8147 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 8148 release. 8149 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 8150 release. The library support is not quite complete. 8151 * Go has been tested on GNU/Linux and Solaris platforms for various 8152 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 8153 work on other platforms as well. 8154 8155New Targets and Target Specific Improvements 8156 8157 AArch64 8158 8159 * A new port has been added to support AArch64, the new 64-bit 8160 architecture from ARM. Note that this is a separate port from the 8161 existing 32-bit ARM port. 8162 * The port provides initial support for the Cortex-A53 and the 8163 Cortex-A57 processors with the command line options 8164 -mcpu=cortex-a53 and -mcpu=cortex-a57. 8165 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 8166 has been added and can be enabled by giving the 8167 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 8168 default by configuring GCC with the --enable-fix-cortex-a53-835769 8169 option. 8170 8171 ARM 8172 8173 * Initial support has been added for the AArch32 extensions defined 8174 in the ARMv8 architecture. 8175 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 8176 * A new option, -mcpu=marvell-pj4, has been added to generate code 8177 for the Marvell PJ4 processor. 8178 * The compiler can now automatically generate the VFMA, VFMS, REVSH 8179 and REV16 instructions. 8180 * A new vectorizer cost model for Advanced SIMD configurations to 8181 improve the auto-vectorization strategies used. 8182 * The scheduler now takes into account the number of live registers 8183 to reduce the amount of spilling that can occur. This should 8184 improve code performance in large functions. The limit can be 8185 removed by using the option -fno-sched-pressure. 8186 * Improvements have been made to the Marvell iWMMX code generation 8187 and support for the iWMMX2 SIMD unit has been added. The option 8188 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 8189 * A number of code generation improvements for Thumb2 to reduce code 8190 size when compiling for the M-profile processors. 8191 * The RTEMS (arm-rtems) port has been updated to use the EABI. 8192 * Code generation support for the old FPA and Maverick floating-point 8193 architectures has been removed. Ports that previously relied on 8194 these features have also been removed. This includes the targets: 8195 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 8196 + arm*-*-elf (use arm*-*-eabi) 8197 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 8198 + arm*-*-ecos-elf (no alternative) 8199 + arm*-*-freebsd (no alternative) 8200 + arm*-wince-pe* (no alternative). 8201 8202 AVR 8203 8204 * Support for the "Embedded C" fixed-point has been added. For 8205 details, see the [28]GCC wiki and the [29]user manual. The support 8206 is not complete. 8207 * A new print modifier %r for register operands in inline assembler 8208 is supported. It will print the raw register number without the 8209 register prefix 'r': 8210 /* Return the most significant byte of 'val', a 64-bit value. */ 8211 8212 unsigned char msb (long long val) 8213 { 8214 unsigned char c; 8215 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 8216 return c; 8217 } 8218 The inline assembler in this example will generate code like 8219 mov r24, 8+7 8220 provided c is allocated to R24 and val is allocated to R8...R15. 8221 This works because the GNU assembler accepts plain register numbers 8222 without register prefix. 8223 * Static initializers with 3-byte symbols are supported now: 8224 extern const __memx char foo; 8225 const __memx void *pfoo = &foo; 8226 This requires at least Binutils 2.23. 8227 8228 IA-32/x86-64 8229 8230 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 8231 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 8232 stack alignment, this is ABI incompatible and intended to be used 8233 in controlled environments where stack space is an important 8234 limitation. This option will lead to wrong code when functions 8235 compiled with 16 byte stack alignment (such as functions from a 8236 standard library) are called with misaligned stack. In this case, 8237 SSE instructions may lead to misaligned memory access traps. In 8238 addition, variable arguments will be handled incorrectly for 16 8239 byte aligned objects (including x87 long double and __int128), 8240 leading to wrong results. You must build all modules with 8241 -mpreferred-stack-boundary=3, including any libraries. This 8242 includes the system libraries and startup modules. 8243 * Support for the new Intel processor codename Broadwell with RDSEED, 8244 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 8245 -mrdseed command-line options. 8246 * Support for the Intel RTM and HLE intrinsics, built-in functions 8247 and code generation is available via -mrtm and -mhle. 8248 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 8249 Intrinsics and built-in functions are available via -mfxsr, -mxsave 8250 and -mxsaveopt respectively. 8251 * New -maddress-mode=[short|long] options for x32. 8252 -maddress-mode=short overrides default 64-bit addresses to 32-bit 8253 by emitting the 0x67 address-size override prefix. This is the 8254 default address mode for x32. 8255 * New built-in functions to detect run-time CPU type and ISA: 8256 + A built-in function __builtin_cpu_is has been added to detect 8257 if the run-time CPU is of a particular type. It returns a 8258 positive integer on a match and zero otherwise. It accepts one 8259 string literal argument, the CPU name. For example, 8260 __builtin_cpu_is("westmere") returns a positive integer if the 8261 run-time CPU is an Intel Core i7 Westmere processor. Please 8262 refer to the [30]user manual for the list of valid CPU names 8263 recognized. 8264 + A built-in function __builtin_cpu_supports has been added to 8265 detect if the run-time CPU supports a particular ISA feature. 8266 It returns a positive integer on a match and zero otherwise. 8267 It accepts one string literal argument, the ISA feature. For 8268 example, __builtin_cpu_supports("ssse3") returns a positive 8269 integer if the run-time CPU supports SSSE3 instructions. 8270 Please refer to the [31]user manual for the list of valid ISA 8271 names recognized. 8272 Caveat: If these built-in functions are called before any static 8273 constructors are invoked, like during IFUNC initialization, then 8274 the CPU detection initialization must be explicitly run using this 8275 newly provided built-in function, __builtin_cpu_init. The 8276 initialization needs to be done only once. For example, this is how 8277 the invocation would look like inside an IFUNC initializer: 8278 static void (*some_ifunc_resolver(void))(void) 8279 { 8280 __builtin_cpu_init(); 8281 if (__builtin_cpu_is("amdfam10h") ... 8282 if (__builtin_cpu_supports("popcnt") ... 8283 } 8284 8285 * Function Multiversioning Support with G++: 8286 It is now possible to create multiple function versions each 8287 targeting a specific processor and/or ISA. Function versions have 8288 the same signature but different target attributes. For example, 8289 here is a program with function versions: 8290 __attribute__ ((target ("default"))) 8291 int foo(void) 8292 { 8293 return 1; 8294 } 8295 8296 __attribute__ ((target ("sse4.2"))) 8297 int foo(void) 8298 { 8299 return 2; 8300 } 8301 8302 int main (void) 8303 { 8304 int (*p) = &foo; 8305 assert ((*p)() == foo()); 8306 return 0; 8307 } 8308 8309 Please refer to this [32]wiki for more information. 8310 * The x86 back end has been improved to allow option -fschedule-insns 8311 to work reliably. This option can be used to schedule instructions 8312 better and leads to improved performace in certain cases. 8313 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 8314 from the Mingw-w64 trunk. 8315 * Support for new AMD family 15h processors (Steamroller core) is now 8316 available through the -march=bdver3 and -mtune=bdver3 options. 8317 * Support for new AMD family 16h processors (Jaguar core) is now 8318 available through the -march=btver2 and -mtune=btver2 options. 8319 8320 FRV 8321 8322 * This target now supports the -fstack-usage command-line option. 8323 8324 MIPS 8325 8326 * GCC can now generate code specifically for the R4700, Broadcom XLP 8327 and MIPS 34kn processors. The associated -march options are 8328 -march=r4700, -march=xlp and -march=34kn respectively. 8329 * GCC now generates better DSP code for MIPS 74k cores thanks to 8330 further scheduling optimizations. 8331 * The MIPS port now supports the -fstack-check option. 8332 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 8333 * Previous versions of GCC would silently accept -fpic and -fPIC for 8334 -mno-abicalls targets like mips*-elf. This combination was not 8335 intended or supported, and did not generate position-independent 8336 code. GCC 4.8 now reports an error when this combination is used. 8337 8338 PowerPC / PowerPC64 / RS6000 8339 8340 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 8341 restore or update the VRSAVE register by default. The respective 8342 operating systems manage the VRSAVE register directly. 8343 * Large TOC support has been added for AIX through the command line 8344 option -mcmodel=large. 8345 * Native Thread-Local Storage support has been added for AIX. 8346 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 8347 when targetting processors that support those hardware features on 8348 AIX 6.1 and above. 8349 8350 RX 8351 8352 * This target will now issue a warning message whenever multiple fast 8353 interrupt handlers are found in the same compilation unit. This 8354 feature can be turned off by the new 8355 -mno-warn-multiple-fast-interrupts command-line option. 8356 8357 S/390, System z 8358 8359 * Support for the IBM zEnterprise zEC12 processor has been added. 8360 When using the -march=zEC12 option, the compiler will generate code 8361 making use of the following new instructions: 8362 + load and trap instructions 8363 + 2 new compare and trap instructions 8364 + rotate and insert selected bits - without CC clobber 8365 The -mtune=zEC12 option enables zEC12 specific instruction 8366 scheduling without making use of new instructions. 8367 * Register pressure sensitive instruction scheduling is enabled by 8368 default. 8369 * The ifunc function attribute is enabled by default. 8370 * memcpy and memcmp invokations on big memory chunks or with run time 8371 lengths are not generated inline anymore when tuning for z10 or 8372 higher. The purpose is to make use of the IFUNC optimized versions 8373 in Glibc. 8374 8375 SH 8376 8377 * The default alignment settings have been reduced to be less 8378 aggressive. This results in more compact code for optimization 8379 levels other than -Os. 8380 * Improved support for the __atomic built-in functions: 8381 + A new option -matomic-model=model selects the model for the 8382 generated atomic sequences. The following models are 8383 supported: 8384 8385 soft-gusa 8386 Software gUSA sequences (SH3* and SH4* only). On 8387 SH4A targets this will now also partially utilize 8388 the movco.l and movli.l instructions. This is the 8389 default when the target is sh3*-*-linux* or 8390 sh4*-*-linux*. 8391 8392 hard-llcs 8393 Hardware movco.l / movli.l sequences (SH4A only). 8394 8395 soft-tcb 8396 Software thread control block sequences. 8397 8398 soft-imask 8399 Software interrupt flipping sequences (privileged 8400 mode only). This is the default when the target is 8401 sh1*-*-linux* or sh2*-*-linux*. 8402 8403 none 8404 Generates function calls to the respective __atomic 8405 built-in functions. This is the default for SH64 8406 targets or when the target is not sh*-*-linux*. 8407 8408 + The option -msoft-atomic has been deprecated. It is now an 8409 alias for -matomic-model=soft-gusa. 8410 + A new option -mtas makes the compiler generate the tas.b 8411 instruction for the __atomic_test_and_set built-in function 8412 regardless of the selected atomic model. 8413 + The __sync functions in libgcc now reflect the selected atomic 8414 model when building the toolchain. 8415 * Added support for the mov.b and mov.w instructions with 8416 displacement addressing. 8417 * Added support for the SH2A instructions movu.b and movu.w. 8418 * Various improvements to code generated for integer arithmetic. 8419 * Improvements to conditional branches and code that involves the T 8420 bit. A new option -mzdcbranch tells the compiler to favor 8421 zero-displacement branches. This is enabled by default for SH4* 8422 targets. 8423 * The pref instruction will now be emitted by the __builtin_prefetch 8424 built-in function for SH3* targets. 8425 * The fmac instruction will now be emitted by the fmaf standard 8426 function and the __builtin_fmaf built-in function. 8427 * The -mfused-madd option has been deprecated in favor of the 8428 machine-independent -ffp-contract option. Notice that the fmac 8429 instruction will now be generated by default for expressions like a 8430 * b + c. This is due to the compiler default setting 8431 -ffp-contract=fast. 8432 * Added new options -mfsrra and -mfsca to allow the compiler using 8433 the fsrra and fsca instructions on targets other than SH4A (where 8434 they are already enabled by default). 8435 * Added support for the __builtin_bswap32 built-in function. It is 8436 now expanded as a sequence of swap.b and swap.w instructions 8437 instead of a library function call. 8438 * The behavior of the -mieee option has been fixed and the negative 8439 form -mno-ieee has been added to control the IEEE conformance of 8440 floating point comparisons. By default -mieee is now enabled and 8441 the option -ffinite-math-only implicitly sets -mno-ieee. 8442 * Added support for the built-in functions __builtin_thread_pointer 8443 and __builtin_set_thread_pointer. This assumes that GBR is used to 8444 hold the thread pointer of the current thread. Memory loads and 8445 stores relative to the address returned by __builtin_thread_pointer 8446 will now also utilize GBR based displacement address modes. 8447 * The -mdiv= option for targets other than SHmedia has been fixed and 8448 documented. 8449 8450 SPARC 8451 8452 * Added optimized instruction scheduling for Niagara4. 8453 8454 TILE-Gx 8455 8456 * Added support for the -mcmodel=MODEL command-line option. The 8457 models supported are small and large. 8458 8459 V850 8460 8461 * This target now supports the E3V5 architecture via the use of the 8462 new -mv850e3v5 command-line option. It also has experimental 8463 support for the e3v5 LOOP instruction which can be enabled via the 8464 new -mloop command-line option. 8465 8466 XStormy16 8467 8468 * This target now supports the -fstack-usage command-line option. 8469 8470Operating Systems 8471 8472 OpenBSD 8473 8474 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 8475 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 8476 8477 Windows (Cygwin) 8478 8479 * Executables are now linked against shared libgcc by default. The 8480 previous default was to link statically, which can still be done by 8481 explicitly specifying -static or static-libgcc on the command line. 8482 However it is strongly advised against, as it will cause problems 8483 for any application that makes use of DLLs compiled by GCC. It 8484 should be alright for a monolithic stand-alone application that 8485 only links against the Windows DLLs, but offers little or no 8486 benefit. 8487 8488GCC 4.8.1 8489 8490 This is the [33]list of problem reports (PRs) from GCC's bug tracking 8491 system that are known to be fixed in the 4.8.1 release. This list might 8492 not be complete (that is, it is possible that some PRs that have been 8493 fixed are not listed here). 8494 8495 The C++11 <chrono> std::chrono::system_clock and 8496 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 8497 both are now separate (never typedefs of each other), both use 8498 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 8499 std::chrono::steady_clock is now finally monotonic, and both classes 8500 are mangled differently than in the previous GCC releases. 8501 std::chrono::system_clock::now() with std::chrono::microseconds resp. 8502 std::chrono::seconds resolution is still exported for backwards 8503 compatibility with default configured libstdc++. Note that libstdc++ 8504 configured with --enable-libstdcxx-time= used to be ABI incompatible 8505 with default configured libstdc++ for those two classes and no ABI 8506 compatibility can be offered for those configurations, so any C++11 8507 code that uses those classes and has been compiled and linked against 8508 libstdc++ configured with the non-default --enable-libstdcxx-time= 8509 configuration option needs to be recompiled. 8510 8511GCC 4.8.2 8512 8513 This is the [34]list of problem reports (PRs) from GCC's bug tracking 8514 system that are known to be fixed in the 4.8.2 release. This list might 8515 not be complete (that is, it is possible that some PRs that have been 8516 fixed are not listed here). 8517 8518GCC 4.8.3 8519 8520 This is the [35]list of problem reports (PRs) from GCC's bug tracking 8521 system that are known to be fixed in the 4.8.3 release. This list might 8522 not be complete (that is, it is possible that some PRs that have been 8523 fixed are not listed here). 8524 8525 Support for the new powerpc64le-linux platform has been added. It 8526 defaults to generating code that conforms to the ELFV2 ABI. 8527 8528GCC 4.8.4 8529 8530 This is the [36]list of problem reports (PRs) from GCC's bug tracking 8531 system that are known to be fixed in the 4.8.4 release. This list might 8532 not be complete (that is, it is possible that some PRs that have been 8533 fixed are not listed here). 8534 8535GCC 4.8.5 8536 8537 This is the [37]list of problem reports (PRs) from GCC's bug tracking 8538 system that are known to be fixed in the 4.8.5 release. This list might 8539 not be complete (that is, it is possible that some PRs that have been 8540 fixed are not listed here). 8541 8542 8543 For questions related to the use of GCC, please consult these web 8544 pages and the [38]GCC manuals. If that fails, the 8545 [39]gcc-help@gcc.gnu.org mailing list might help. Comments on these 8546 web pages and the development of GCC are welcome on our developer 8547 list at [40]gcc@gcc.gnu.org. All of [41]our lists have public 8548 archives. 8549 8550 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 8551 distribution of this entire article is permitted in any medium, 8552 provided this notice is preserved. 8553 8554 These pages are [43]maintained by the GCC team. Last modified 8555 2021-07-28[44]. 8556 8557References 8558 8559 1. https://gcc.gnu.org/wiki/cxx-conversion 8560 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 8561 3. http://www.nongnu.org/avr-libc/ 8562 4. https://gcc.gnu.org/PR54461 8563 5. https://gcc.gnu.org/gcc-4.8/porting_to.html 8564 6. https://github.com/google/sanitizers 8565 7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 8566 8. https://lkml.org/lkml/2006/11/28/239 8567 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 8568 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 8569 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 8570 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 8571 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 8572 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 8573 15. http://gcc.gnu.org/projects/cxx1y.html 8574 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 8575 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 8576 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 8577 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 8578 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 8579 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 8580 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 8581 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 8582 24. https://gcc.gnu.org/wiki/Fortran2003Status 8583 25. https://gcc.gnu.org/wiki/TS29113Status 8584 26. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libgfortran/libgfortran.h 8585 27. http://chasm-interop.sourceforge.net/ 8586 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 8587 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 8588 30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 8589 31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 8590 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning 8591 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 8592 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 8593 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 8594 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 8595 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 8596 38. https://gcc.gnu.org/onlinedocs/ 8597 39. mailto:gcc-help@gcc.gnu.org 8598 40. mailto:gcc@gcc.gnu.org 8599 41. https://gcc.gnu.org/lists.html 8600 42. https://www.fsf.org/ 8601 43. https://gcc.gnu.org/about.html 8602 44. http://validator.w3.org/check/referer 8603====================================================================== 8604http://gcc.gnu.org/gcc-4.7/index.html 8605 8606 GCC 4.7 Release Series 8607 8608 (This release series is no longer supported.) 8609 8610 June 12, 2014 8611 8612 The [1]GNU project and the GCC developers are pleased to announce the 8613 release of GCC 4.7.4. 8614 8615 This release is a bug-fix release, containing fixes for regressions in 8616 GCC 4.7.3 relative to previous releases of GCC. 8617 8618Release History 8619 8620 GCC 4.7.4 8621 June 12, 2014 ([2]changes, [3]documentation) 8622 8623 GCC 4.7.3 8624 April 11, 2013 ([4]changes, [5]documentation) 8625 8626 GCC 4.7.2 8627 September 20, 2012 ([6]changes, [7]documentation) 8628 8629 GCC 4.7.1 8630 June 14, 2012 ([8]changes, [9]documentation) 8631 8632 GCC 4.7.0 8633 March 22, 2012 ([10]changes, [11]documentation) 8634 8635References and Acknowledgements 8636 8637 GCC used to stand for the GNU C Compiler, but since the compiler 8638 supports several other languages aside from C, it now stands for the 8639 GNU Compiler Collection. 8640 8641 A list of [12]successful builds is updated as new information becomes 8642 available. 8643 8644 The GCC developers would like to thank the numerous people that have 8645 contributed new features, improvements, bug fixes, and other changes as 8646 well as test results to GCC. This [13]amazing group of volunteers is 8647 what makes GCC successful. 8648 8649 For additional information about GCC please refer to the [14]GCC 8650 project web site or contact the [15]GCC development mailing list. 8651 8652 To obtain GCC please use [16]our mirror sites or [17]our version 8653 control system. 8654 8655 8656 For questions related to the use of GCC, please consult these web 8657 pages and the [18]GCC manuals. If that fails, the 8658 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 8659 web pages and the development of GCC are welcome on our developer 8660 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 8661 archives. 8662 8663 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 8664 distribution of this entire article is permitted in any medium, 8665 provided this notice is preserved. 8666 8667 These pages are [23]maintained by the GCC team. Last modified 8668 2021-07-28[24]. 8669 8670References 8671 8672 1. http://www.gnu.org/ 8673 2. http://gcc.gnu.org/gcc-4.7/changes.html 8674 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 8675 4. http://gcc.gnu.org/gcc-4.7/changes.html 8676 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 8677 6. http://gcc.gnu.org/gcc-4.7/changes.html 8678 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 8679 8. http://gcc.gnu.org/gcc-4.7/changes.html 8680 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 8681 10. http://gcc.gnu.org/gcc-4.7/changes.html 8682 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 8683 12. http://gcc.gnu.org/gcc-4.7/buildstat.html 8684 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8685 14. http://gcc.gnu.org/index.html 8686 15. mailto:gcc@gcc.gnu.org 8687 16. http://gcc.gnu.org/mirrors.html 8688 17. http://gcc.gnu.org/git.html 8689 18. https://gcc.gnu.org/onlinedocs/ 8690 19. mailto:gcc-help@gcc.gnu.org 8691 20. mailto:gcc@gcc.gnu.org 8692 21. https://gcc.gnu.org/lists.html 8693 22. https://www.fsf.org/ 8694 23. https://gcc.gnu.org/about.html 8695 24. http://validator.w3.org/check/referer 8696====================================================================== 8697http://gcc.gnu.org/gcc-4.7/changes.html 8698 8699 GCC 4.7 Release Series 8700 Changes, New Features, and Fixes 8701 8702Caveats 8703 8704 * The -fconserve-space flag has been deprecated. The flag had no 8705 effect for most targets: only targets without a global .bss section 8706 and without support for switchable sections. Furthermore, the flag 8707 only had an effect for G++, where it could result in wrong 8708 semantics (please refer to the GCC manual for further details). The 8709 flag will be removed in GCC 4.8 8710 * Support for a number of older systems and recently unmaintained or 8711 untested target ports of GCC has been declared obsolete in GCC 4.7. 8712 Unless there is activity to revive them, the next release of GCC 8713 will have their sources permanently removed. 8714 All GCC ports for the following processor architectures have been 8715 declared obsolete: 8716 + picoChip (picochip-*) 8717 The following ports for individual systems on particular 8718 architectures have been obsoleted: 8719 + IRIX 6.5 (mips-sgi-irix6.5) 8720 + MIPS OpenBSD (mips*-*-openbsd*) 8721 + Solaris 8 (*-*-solaris2.8). Details can be found in the 8722 [1]announcement. 8723 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 8724 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 8725 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 8726 default, which for some sources generates code that accesses memory 8727 on unaligned addresses. This requires the kernel of those systems 8728 to enable such accesses (controlled by CP15 register c1, refer to 8729 ARM documentation). Alternatively, or for compatibility with 8730 kernels where unaligned accesses are not supported, all code has to 8731 be compiled with -mno-unaligned-access. Upstream Linux kernel 8732 releases have automatically and unconditionally supported unaligned 8733 accesses as emitted by GCC due to this option being active since 8734 version 2.6.28. 8735 * Support on ARM for the legacy floating-point accelerator (FPA) and 8736 the mixed-endian floating-point format that it used has been 8737 obsoleted. The ports that still use this format have been obsoleted 8738 as well. Many legacy ARM ports already provide an alternative that 8739 uses the VFP floating-point format. The obsolete ports will be 8740 deleted in the next release. 8741 The obsolete ports with alternatives are: 8742 + arm*-*-rtems (use arm*-*-rtemseabi) 8743 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 8744 + arm*-*-elf (use arm*-*-eabi) 8745 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 8746 Note, however, that these alternatives are not binary compatible 8747 with their legacy counterparts (although some can support running 8748 legacy applications). 8749 The obsolete ports that currently lack a modern alternative are: 8750 + arm*-*-ecos-elf 8751 + arm*-*-freebsd 8752 + arm*-wince-pe* 8753 New ports that support more recent versions of the architecture are 8754 welcome. 8755 * Support for the Maverick co-processor on ARM has been obsoleted. 8756 Code to support it will be deleted in the next release. 8757 * Support has been removed for Unix International threads on Solaris 8758 2, so the --enable-threads=solaris configure option and the 8759 -threads compiler option don't work any longer. 8760 * Support has been removed for the Solaris BSD Compatibility Package, 8761 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 8762 from Solaris 11, and was only intended as a migration aid from 8763 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 8764 recognized any longer. 8765 * The AVR port's libgcc has been improved and its multilib structure 8766 has been enhanced. As a result, all objects contributing to an 8767 application must either be compiled with GCC versions up to 4.6.x 8768 or with GCC versions 4.7.1 or later. If the compiler is used with 8769 AVR Libc, you need a version that supports the new layout, i.e. 8770 implements [2]#35407. 8771 * The AVR port's -mshort-calls command-line option has been 8772 deprecated. It will be removed in the GCC 4.8 release. See -mrelax 8773 for a replacement. 8774 * The AVR port only references startup code that clears .bss and the 8775 common section resp. initializes the .data and .rodata section 8776 provided respective sections (or subsections thereof) are not 8777 empty, see [3]PR18145. Applications that put all static storage 8778 objects into non-standard sections and / or define all static 8779 storage objects in assembler modules, must reference __do_clear_bss 8780 resp. __do_copy_data by hand or undefine the symbol(s) by means of 8781 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 8782 * The ARM port's -mwords-little-endian option has been deprecated. It 8783 will be removed in a future release. 8784 * Support has been removed for the NetWare x86 configuration 8785 obsoleted in GCC 4.6. 8786 * It is no longer possible to use the "l" constraint in MIPS16 asm 8787 statements. 8788 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 8789 library which affected the ABI in C++11 mode: a data member was 8790 added to std::list changing its size and altering the definitions 8791 of some member functions, and std::pair's move constructor was 8792 non-trivial which altered the calling convention for functions with 8793 std::pair arguments or return types. The ABI incompatibilities have 8794 been fixed for GCC version 4.7.2 but as a result C++11 code 8795 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 8796 code compiled with different GCC versions and with C++98/C++03 code 8797 compiled with any version. 8798 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 8799 rules for the layout of vectors that could lead to wrong code being 8800 generated. Vectors larger than 8 bytes in size are now by default 8801 aligned to an 8-byte boundary. This is an ABI change: code that 8802 makes explicit use of vector types may be incompatible with binary 8803 objects built with older versions of GCC. Auto-vectorized code is 8804 not affected by this change. (This change affects GCC versions 8805 4.7.2 and later.) 8806 * More information on porting to GCC 4.7 from previous versions of 8807 GCC can be found in the [4]porting guide for this release. 8808 8809General Optimizer Improvements 8810 8811 * Support for a new parameter --param case-values-threshold=n was 8812 added to allow users to control the cutoff between doing switch 8813 statements as a series of if statements and using a jump table. 8814 * Link-time optimization (LTO) improvements: 8815 + Improved scalability and reduced memory usage. Link time 8816 optimization of Firefox now requires 3GB of RAM on a 64-bit 8817 system, while over 8GB was needed previously. Linking time has 8818 been improved, too. The serial stage of linking Firefox has 8819 been sped up by about a factor of 10. 8820 + Reduced size of object files and temporary storage used during 8821 linking. 8822 + Streaming performance (both outbound and inbound) has been 8823 improved. 8824 + ld -r is now supported with LTO. 8825 + Several bug fixes, especially in symbol table handling and 8826 merging. 8827 * Interprocedural optimization improvements: 8828 + Heuristics now take into account that after inlining code will 8829 be optimized out because of known values (or properties) of 8830 function parameters. For example: 8831void foo(int a) 8832{ 8833 if (a > 10) 8834 ... huge code ... 8835} 8836void bar (void) 8837{ 8838 foo (0); 8839} 8840 8841 The call of foo will be inlined into bar even when optimizing 8842 for code size. Constructs based on __builtin_constant_p are 8843 now understood by the inliner and code size estimates are 8844 evaluated a lot more realistically. 8845 + The representation of C++ virtual thunks and aliases (both 8846 implicit and defined via the alias attribute) has been 8847 re-engineered. Aliases no longer pose optimization barriers 8848 and calls to an alias can be inlined and otherwise optimized. 8849 + The inter-procedural constant propagation pass has been 8850 rewritten. It now performs generic function specialization. 8851 For example when compiling the following: 8852void foo(bool flag) 8853{ 8854 if (flag) 8855 ... do something ... 8856 else 8857 ... do something else ... 8858} 8859void bar (void) 8860{ 8861 foo (false); 8862 foo (true); 8863 foo (false); 8864 foo (true); 8865 foo (false); 8866 foo (true); 8867} 8868 8869 GCC will now produce two copies of foo. One with flag being 8870 true, while other with flag being false. This leads to 8871 performance improvements previously possible only by inlining 8872 all calls. Cloning causes a lot less code size growth. 8873 * A string length optimization pass has been added. It attempts to 8874 track string lengths and optimize various standard C string 8875 functions like strlen, strchr, strcpy, strcat, stpcpy and their 8876 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 8877 enabled by default at -O2 or above, unless optimizing for size, and 8878 can be disabled by the -fno-optimize-strlen option. The pass can 8879 e.g. optimize 8880char *bar (const char *a) 8881{ 8882 size_t l = strlen (a) + 2; 8883 char *p = malloc (l); if (p == NULL) return p; 8884 strcpy (p, a); strcat (p, "/"); return p; 8885} 8886 8887 into: 8888char *bar (const char *a) 8889{ 8890 size_t tmp = strlen (a); 8891 char *p = malloc (tmp + 2); if (p == NULL) return p; 8892 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 8893} 8894 8895 or for hosted compilations where stpcpy is available in the runtime 8896 and headers provide its prototype, e.g. 8897void foo (char *a, const char *b, const char *c, const char *d) 8898{ 8899 strcpy (a, b); strcat (a, c); strcat (a, d); 8900} 8901 8902 can be optimized into: 8903void foo (char *a, const char *b, const char *c, const char *d) 8904{ 8905 strcpy (stpcpy (stpcpy (a, b), c), d); 8906} 8907 8908New Languages and Language specific improvements 8909 8910 * Version 3.1 of the OpenMP specification is now supported for the C, 8911 C++, and Fortran compilers. 8912 8913 Ada 8914 8915 * The command-line option -feliminate-unused-debug-types has been 8916 re-enabled by default, as it is for the other languages, leading to 8917 a reduction in debug info size of 12.5% and more for relevant 8918 cases, as well as to a small compilation speedup. 8919 8920 C family 8921 8922 * A new built-in, __builtin_assume_aligned, has been added, through 8923 which the compiler can be hinted about pointer alignment and can 8924 use it to improve generated code. 8925 * A new warning option -Wunused-local-typedefs was added for C, C++, 8926 Objective-C and Objective-C++. This warning diagnoses typedefs 8927 locally defined in a function, and otherwise not used. 8928 * A new experimental command-line option -ftrack-macro-expansion was 8929 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 8930 the compiler to emit diagnostic about the current macro expansion 8931 stack when a compilation error occurs in a macro expansion. 8932 * Experimental support for transactional memory has been added. It 8933 includes support in the compiler, as well as a supporting runtime 8934 library called libitm. To compile code with transactional memory 8935 constructs, use the -fgnu-tm option. 8936 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 8937 and 32-bit/64-bit x86 platforms. 8938 For more details on transactional memory see [5]the GCC WiKi. 8939 * Support for atomic operations specifying the C++11/C11 memory model 8940 has been added. These new __atomic routines replace the existing 8941 __sync built-in routines. 8942 Atomic support is also available for memory blocks. Lock-free 8943 instructions will be used if a memory block is the same size and 8944 alignment as a supported integer type. Atomic operations which do 8945 not have lock-free support are left as function calls. A set of 8946 library functions is available on the GCC atomic wiki in the 8947 "External Atomics Library" section. 8948 For more details on the memory models and features, see the 8949 [6]atomic wiki. 8950 * When a binary operation is performed on vector types and one of the 8951 operands is a uniform vector, it is possible to replace the vector 8952 with the generating element. For example: 8953typedef int v4si __attribute__ ((vector_size (16))); 8954v4si res, a = {1,2,3,4}; 8955int x; 8956 8957res = 2 + a; /* means {2,2,2,2} + a */ 8958res = a - x; /* means a - {x,x,x,x} */ 8959 8960 C 8961 8962 * There is support for some more features from the C11 revision of 8963 the ISO C standard. GCC now accepts the options -std=c11 and 8964 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 8965 + Unicode strings (previously supported only with options such 8966 as -std=gnu11, now supported with -std=c11), and the 8967 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 8968 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 8969 + Alignment support (_Alignas, _Alignof, max_align_t, 8970 <stdalign.h>). 8971 + A built-in function __builtin_complex is provided to support C 8972 library implementation of the CMPLX family of macros. 8973 8974 C++ 8975 8976 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 8977 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 8978 -Wc++0x-compat, respectively. 8979 * G++ now implements [7]C++11 extended friend syntax: 8980 8981template<class W> 8982class Q 8983{ 8984 static const int I = 2; 8985public: 8986 friend W; 8987}; 8988 8989struct B 8990{ 8991 int ar[Q<B>::I]; 8992}; 8993 8994 * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit 8995 override control. 8996 8997struct B { 8998 virtual void f() const final; 8999 virtual void f(int); 9000}; 9001 9002struct D : B { 9003 void f() const; // error: D::f attempts to override final B::f 9004 void f(long) override; // error: doesn't override anything 9005 void f(int) override; // ok 9006}; 9007 9008struct E final { }; 9009struct F: E { }; // error: deriving from final class 9010 9011 * G++ now implements [9]C++11 non-static data member initializers. 9012 9013struct A { 9014 int i = 42; 9015} a; // initializes a.i to 42 9016 9017 * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11 9018 user-defined literals. 9019 9020// Not actually a good approximation. :) 9021constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 9022long double pi = 180.0_degrees; 9023 9024 * G++ now implements [11]C++11 alias-declarations. 9025 9026template <class T> using Ptr = T*; 9027Ptr<int> ip; // decltype(ip) is int* 9028 9029 * Thanks to Ville Voutilainen and Pedro Lamar�o, G++ now implements 9030 [12]C++11 delegating constructors. 9031 9032struct A { 9033 A(int); 9034 A(): A(42) { } // delegate to the A(int) constructor 9035}; 9036 9037 * G++ now fully implements C++11 atomic classes rather than just 9038 integer derived classes. 9039 9040class POD { 9041 int a; 9042 int b; 9043}; 9044std::atomic<POD> my_atomic_POD; 9045 9046 * G++ now sets the predefined macro __cplusplus to the correct value, 9047 199711L for C++98/03, and 201103L for C++11. 9048 * G++ now correctly implements the two-phase lookup rules such that 9049 an unqualified name used in a template must have an appropriate 9050 declaration found either in scope at the point of definition of the 9051 template or by argument-dependent lookup at the point of 9052 instantiation. As a result, code that relies on a second 9053 unqualified lookup at the point of instantiation to find functions 9054 declared after the template or in dependent bases will be rejected. 9055 The compiler will suggest ways to fix affected code, and using the 9056 -fpermissive compiler flag will allow the code to compile with a 9057 warning. 9058 9059template <class T> 9060void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 9061void g(int) { } // fix by moving this declaration before the declaration of f 9062 9063template <class T> 9064struct A: T { 9065 // error, B::g(B) not found by argument-dependent lookup 9066 void f() { g(T()); } // fix by using this->g or A::g 9067}; 9068 9069struct B { void g(B); }; 9070 9071int main() 9072{ 9073 f<int>(); 9074 A<B>().f(); 9075} 9076 9077 * G++ now properly re-uses stack space allocated for temporary 9078 objects when their lifetime ends, which can significantly lower 9079 stack consumption for some C++ functions. As a result of this, some 9080 code with undefined behavior will now break: 9081 9082const int &f(const int &i) { return i; } 9083.... 9084const int &x = f(1); 9085const int &y = f(2); 9086 9087 Here, x refers to the temporary allocated to hold the 1 argument, 9088 which only lives until the end of the initialization; it 9089 immediately becomes a dangling reference. So the next statement 9090 re-uses the stack slot to hold the 2 argument, and users of x get 9091 that value instead. 9092 Note that this should not cause any change of behavior for 9093 temporaries of types with non-trivial destructors, as they are 9094 already destroyed at end of full-expression; the change is that now 9095 the storage is released as well. 9096 * A new command-line option -Wdelete-non-virtual-dtor has been added 9097 to warn when delete is used to destroy an instance of a class which 9098 has virtual functions and non-virtual destructor. It is unsafe to 9099 delete an instance of a derived class through a pointer to a base 9100 class if the base class does not have a virtual destructor. This 9101 warning is enabled by -Wall. 9102 * A new command-line option -Wzero-as-null-pointer-constant has been 9103 added to warn when a literal '0' is used as null pointer constant. 9104 It can be useful to facilitate the conversion to nullptr in C++11. 9105 * As per C++98, access-declarations are now deprecated by G++. 9106 Using-declarations are to be used instead. Furthermore, some 9107 efforts have been made to improve the support of class scope 9108 using-declarations. In particular, using-declarations referring to 9109 a dependent type now work as expected ([13]bug c++/14258). 9110 * The ELF symbol visibility of a template instantiation is now 9111 properly constrained by the visibility of its template arguments 9112 ([14]bug c++/35688). 9113 9114 Runtime Library (libstdc++) 9115 9116 * [15]Improved experimental support for the new ISO C++ standard, 9117 C++11, including: 9118 + using noexcept in most of the library; 9119 + implementations of pointer_traits, allocator_traits and 9120 scoped_allocator_adaptor; 9121 + uses-allocator construction for tuple; 9122 + vector meets the allocator-aware container requirements; 9123 + replacing monotonic_clock with steady_clock; 9124 + enabling the thread support library on most POSIX targets; 9125 + many small improvements to conform to the FDIS. 9126 * Added --enable-clocale=newlib configure option. 9127 * Debug Mode iterators for unordered associative containers. 9128 * Avoid polluting the global namespace and do not include <unistd.h>. 9129 9130 Fortran 9131 9132 * The compile flag [16]-fstack-arrays has been added, which causes 9133 all local arrays to be put on stack memory. For some programs this 9134 will improve the performance significantly. If your program uses 9135 very large local arrays, it is possible that you will have to 9136 extend your runtime limits for stack memory. 9137 * The [17]-Ofast flag now also implies [18]-fno-protect-parens and 9138 [19]-fstack-arrays. 9139 * Front-end optimizations can now be selected by the 9140 [20]-ffrontend-optimize option and deselected by the 9141 -fno-frontend-optimize option. 9142 * When front-end optimization removes a function call, 9143 [21]-Wfunction-elimination warns about that. 9144 * When performing front-end-optimization, the 9145 [22]-faggressive-function-elimination option allows the removal of 9146 duplicate function calls even for impure functions. 9147 * The flag [23]-Wreal-q-constant has been added, which warns if 9148 floating-point literals have been specified using q (such as 9149 1.0q0); the q marker is now supported as a vendor extension to 9150 denote quad precision (REAL(16) or, if not available, REAL(10)). 9151 Consider using a kind parameter (such as in 1.0_qp) instead, which 9152 can be obtained via [24]SELECTED_REAL_KIND. 9153 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 9154 Fortran now always prints error messages to standard error. If you 9155 wish to redirect standard error, please consult the manual for your 9156 OS, shell, batch environment etc. as appropriate. 9157 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 9158 variable have been removed. When encountering a serious error, 9159 gfortran will now always abort the program. Whether a core dump is 9160 generated depends on the user environment settings; see the ulimit 9161 -c setting for POSIX shells, limit coredumpsize for C shells, and 9162 the [25]WER user-mode dumps settings on Windows. 9163 * The [26]-fbacktrace option is now enabled by default. When 9164 encountering a fatal error, gfortran will attempt to print a 9165 backtrace to standard error before aborting. It can be disabled 9166 with -fno-backtrace. Note: On POSIX targets with the addr2line 9167 utility from GNU binutils, GNU Fortran can print a backtrace with 9168 function name, file name, line number information in addition to 9169 the addresses; otherwise only the addresses are printed. 9170 * [27]Fortran 2003: 9171 + Generic interface names which have the same name as derived 9172 types are now supported, which allows to write constructor 9173 functions. Note that Fortran does not support static 9174 constructor functions; only default initialization or an 9175 explicit structure-constructor initialization are available. 9176 + [28]Polymorphic (class) arrays are now supported. 9177 * [29]Fortran 2008: 9178 + Support for the DO CONCURRENT construct has been added, which 9179 allows the user to specify that individual loop iterations 9180 have no interdependencies. 9181 + [30]Coarrays: Full single-image support except for polymorphic 9182 coarrays. Additionally, preliminary support for multiple 9183 images via an MPI-based [31]coarray communication library has 9184 been added. Note: The library version is not yet usable as 9185 remote coarray access is not yet possible. 9186 * [32]TS 29113: 9187 + New flag [33]-std=f2008ts permits programs that are expected 9188 to conform to the Fortran 2008 standard and the draft 9189 Technical Specification (TS) 29113 on Further Interoperability 9190 of Fortran with C. 9191 + The OPTIONAL attribute is now allowed for dummy arguments of 9192 BIND(C) procedures. 9193 + The RANK intrinsic has been added. 9194 + The implementation of the ASYNCHRONOUS attribute in GCC is 9195 compatible with the candidate draft of TS 29113 (since GCC 9196 4.6). 9197 9198 Go 9199 9200 * GCC 4.7 implements the [34]Go 1 language standard. The library 9201 support in 4.7.0 is not quite complete, due to release timing. 9202 Release 4.7.1 includes complete support for Go 1. The Go library is 9203 from the Go 1.0.1 release. 9204 * Go has been tested on GNU/Linux and Solaris platforms. It may work 9205 on other platforms as well. 9206 9207New Targets and Target Specific Improvements 9208 9209 ARM 9210 9211 * GCC now supports the Cortex-A7 processor implementing the v7-a 9212 version of the architecture using the option -mcpu=cortex-a7. 9213 * The default vector size in auto-vectorization for NEON is now 128 9214 bits. If vectorization fails thusly, the vectorizer tries again 9215 with 64-bit vectors. 9216 * A new option -mvectorize-with-neon-double was added to allow users 9217 to change the vector size to 64 bits. 9218 9219 AVR 9220 9221 * GCC now supports the XMEGA architecture. This requires GNU binutils 9222 2.22 or later. 9223 * Support for the [35]named address spaces __flash, __flash1, ..., 9224 __flash5 and __memx has been added. These address spaces locate 9225 read-only data in flash memory and allow reading from flash memory 9226 by means of ordinary C code, i.e. without the need of (inline) 9227 assembler code: 9228 9229const __flash int values[] = { 42, 31 }; 9230 9231int add_values (const __flash int *p, int i) 9232{ 9233 return values[i] + *p; 9234} 9235 9236 * Support has been added for the AVR-specific configure option 9237 --with-avrlibc=yes in order to arrange for better integration of 9238 [36]AVR-Libc. This configure option is supported in avr-gcc 4.7.2 9239 and newer and will only take effect in non-RTEMS configurations. If 9240 avr-gcc is configured for RTEMS, the option will be ignored which 9241 is the same as specifying --with-avrlibc=no. See [37]PR54461 for 9242 more technical details. 9243 * Support for AVR-specific [38]built-in functions has been added. 9244 * Support has been added for the signed and unsigned 24-bit scalar 9245 integer types __int24 and __uint24. 9246 * New command-line options -maccumulate-args, -mbranch-cost=cost and 9247 -mstrict-X were added to allow better fine-tuning of code 9248 optimization. 9249 * The command-line option -fdata-sections now also takes affect on 9250 the section names of variables with the progmem attribute. 9251 * A new inline assembler print modifier %i to print a RAM address as 9252 I/O address has been added: 9253 9254#include <avr/io.h> /* Port Definitions from AVR-LibC */ 9255 9256void set_portb (uint8_t value) 9257{ 9258 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 9259} 9260 9261 The offset between an I/O address and the RAM address for that I/O 9262 location is device-specific. This offset is taken into account when 9263 printing a RAM address with the %i modifier so that the address is 9264 suitable to be used as operand in an I/O command. The address must 9265 be a constant integer known at compile time. 9266 * The inline assembler constraint "R" to represent integers in the 9267 range -6 ... 5 has been removed without replacement. 9268 * Many optimizations to: 9269 + 64-bit integer arithmetic 9270 + Widening multiplication 9271 + Integer division by a constant 9272 + Avoid constant reloading in multi-byte instructions. 9273 + Micro-optimizations for special instruction sequences. 9274 + Generic built-in functions like __builtin_ffs*, 9275 __builtin_clz*, etc. 9276 + If-else decision trees generated by switch instructions 9277 + Merging of data located in flash memory 9278 + New libgcc variants for devices with 8-bit wide stack pointer 9279 + ... 9280 * Better documentation: 9281 + Handling of EIND and indirect jumps on devices with more than 9282 128 KiB of program memory. 9283 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 9284 registers. 9285 + Function attributes OS_main and OS_task. 9286 + AVR-specific built-in macros. 9287 9288 C6X 9289 9290 * Support has been added for the Texas Instruments C6X family of 9291 processors. 9292 9293 CR16 9294 9295 * Support has been added for National Semiconductor's CR16 9296 architecture. 9297 9298 Epiphany 9299 9300 * Support has been added for Adapteva's Epiphany architecture. 9301 9302 IA-32/x86-64 9303 9304 * Support for Intel AVX2 intrinsics, built-in functions and code 9305 generation is available via -mavx2. 9306 * Support for Intel BMI2 intrinsics, built-in functions and code 9307 generation is available via -mbmi2. 9308 * Implementation and automatic generation of __builtin_clz* using the 9309 lzcnt instruction is available via -mlzcnt. 9310 * Support for Intel FMA3 intrinsics and code generation is available 9311 via -mfma. 9312 * A new -mfsgsbase command-line option is available that makes GCC 9313 generate new segment register read/write instructions through 9314 dedicated built-ins. 9315 * Support for the new Intel rdrnd instruction is available via 9316 -mrdrnd. 9317 * Two additional AVX vector conversion instructions are available via 9318 -mf16c. 9319 * Support for new Intel processor codename IvyBridge with RDRND, 9320 FSGSBASE and F16C is available through -march=core-avx-i. 9321 * Support for the new Intel processor codename Haswell with AVX2, 9322 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 9323 * Support for new AMD family 15h processors (Piledriver core) is now 9324 available through -march=bdver2 and -mtune=bdver2 options. 9325 * Support for [39]the x32 psABI is now available through the -mx32 9326 option. 9327 * Windows mingw targets are using the -mms-bitfields option by 9328 default. 9329 * Windows x86 targets are using the __thiscall calling convention for 9330 C++ class-member functions. 9331 * Support for the configure option --with-threads=posix for Windows 9332 mingw targets. 9333 9334 MIPS 9335 9336 * GCC now supports thread-local storage (TLS) for MIPS16. This 9337 requires GNU binutils 2.22 or later. 9338 * GCC can now generate code specifically for the Cavium Octeon+ and 9339 Octeon2 processors. The associated command-line options are 9340 -march=octeon+ and -march=octeon2 respectively. Both options 9341 require GNU binutils 2.22 or later. 9342 * GCC can now work around certain 24k errata, under the control of 9343 the command-line option -mfix-24k. These workarounds require GNU 9344 binutils 2.20 or later. 9345 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 9346 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 9347 toolchain that generates 32-bit code by default. Use the 9348 configure-time option --enable-targets=all to select these extra 9349 multilibs. 9350 * Passing -fno-delayed-branch now also stops the assembler from 9351 automatically filling delay slots. 9352 9353 PowerPC/PowerPC64 9354 9355 * Vectors of type vector long long or vector long are passed and 9356 returned using the same method as other vectors with the VSX 9357 instruction set. Previously GCC did not adhere to the ABI for 9358 128-bit vectors with 64-bit integer base types (PR 48857). This 9359 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 9360 * A new option -mno-pointers-to-nested-functions was added to allow 9361 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 9362 that the compiler should not load up the chain register (r11) 9363 before calling a function through a pointer. If you use this 9364 option, you cannot call nested functions through a pointer, or call 9365 other languages that might use the static chain. 9366 * A new option msave-toc-indirect was added to allow AIX 9367 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 9368 save the TOC in the prologue for indirect calls or generate the 9369 save inline. This can speed up some programs that call through a 9370 function pointer a lot, but it can slow down other functions that 9371 only call through a function pointer in exceptional cases. 9372 * The PowerPC port will now enable machine-specific built-in 9373 functions when the user switches the target machine using the 9374 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 9375 sequences. In addition, the target macros are updated. However, due 9376 to the way the -save-temps switch is implemented, you won't see the 9377 effect of these additional macros being defined in preprocessor 9378 output. 9379 9380 SH 9381 9382 * A new option -msoft-atomic has been added. When it is specified, 9383 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 9384 the new __atomic routines. 9385 * Since it is neither supported by GAS nor officially documented, 9386 code generation for little endian SH2A has been disabled. 9387 Specifying -ml with -m2a* will now result in a compiler error. 9388 * The defunct -mbranch-cost option has been fixed. 9389 * Some improvements to the generated code of: 9390 + Utilization of the tst #imm,R0 instruction. 9391 + Dynamic shift instructions on SH2A. 9392 + Integer absolute value calculations. 9393 * The -mdiv= option for targets other than SHmedia has been fixed and 9394 documented. 9395 9396 SPARC 9397 9398 * The option -mflat has been reinstated. When it is specified, the 9399 compiler will generate code for a single register window model. 9400 This is essentially a new implementation and the corresponding 9401 debugger support has been added to GDB 7.4. 9402 * Support for the options -mtune=native and -mcpu=native has been 9403 added on selected native platforms (GNU/Linux and Solaris). 9404 * Support for the SPARC T3 (Niagara 3) processor has been added. 9405 * VIS: 9406 + An intrinsics header visintrin.h has been added. 9407 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 9408 compare instructions have been added. 9409 + The little-endian version of alignaddr is now supported. 9410 + When possible, VIS builtins are marked const, which should 9411 increase the compiler's ability to optimize VIS operations. 9412 + The compiler now properly tracks the %gsr register and how it 9413 behaves as an input for various VIS instructions. 9414 + Akin to fzero, the compiler can now generate fone instructions 9415 in order to set all of the bits of a floating-point register 9416 to 1. 9417 + The documentation for the VIS intrinsics in the GCC manual has 9418 been brought up to date and many inaccuracies were fixed. 9419 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 9420 non-condition-code setting edge instructions have been added. 9421 Their availability is controlled by the new -mvis2 and 9422 -mno-vis2 options. They are enabled by default on 9423 UltraSPARC-III and later CPUs. 9424 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 9425 has been added. These instructions are enabled by default on SPARC 9426 T3 (Niagara 3) and later CPUs. 9427 9428 TILE-Gx/TILEPro 9429 9430 * Support has been added for the Tilera TILE-Gx and TILEPro families 9431 of processors. 9432 9433Other significant improvements 9434 9435 * A new option (-grecord-gcc-switches) was added that appends 9436 compiler command-line options that might affect code generation to 9437 the DW_AT_producer attribute string in the DWARF debugging 9438 information. 9439 * GCC now supports various new GNU extensions to the DWARF debugging 9440 information format, like [40]entry value and [41]call site 9441 information, [42]typed DWARF stack or [43]a more compact macro 9442 representation. Support for these extensions has been added to GDB 9443 7.4. They can be disabled through the -gstrict-dwarf command-line 9444 option. 9445 9446GCC 4.7.1 9447 9448 This is the [44]list of problem reports (PRs) from GCC's bug tracking 9449 system that are known to be fixed in the 4.7.1 release. This list might 9450 not be complete (that is, it is possible that some PRs that have been 9451 fixed are not listed here). 9452 9453 The Go front end in the 4.7.1 release fully supports the [45]Go 1 9454 language standard. 9455 9456GCC 4.7.2 9457 9458 This is the [46]list of problem reports (PRs) from GCC's bug tracking 9459 system that are known to be fixed in the 4.7.2 release. This list might 9460 not be complete (that is, it is possible that some PRs that have been 9461 fixed are not listed here). 9462 9463GCC 4.7.3 9464 9465 This is the [47]list of problem reports (PRs) from GCC's bug tracking 9466 system that are known to be fixed in the 4.7.3 release. This list might 9467 not be complete (that is, it is possible that some PRs that have been 9468 fixed are not listed here). 9469 9470GCC 4.7.4 9471 9472 This is the [48]list of problem reports (PRs) from GCC's bug tracking 9473 system that are known to be fixed in the 4.7.4 release. This list might 9474 not be complete (that is, it is possible that some PRs that have been 9475 fixed are not listed here). 9476 9477 9478 For questions related to the use of GCC, please consult these web 9479 pages and the [49]GCC manuals. If that fails, the 9480 [50]gcc-help@gcc.gnu.org mailing list might help. Comments on these 9481 web pages and the development of GCC are welcome on our developer 9482 list at [51]gcc@gcc.gnu.org. All of [52]our lists have public 9483 archives. 9484 9485 Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and 9486 distribution of this entire article is permitted in any medium, 9487 provided this notice is preserved. 9488 9489 These pages are [54]maintained by the GCC team. Last modified 9490 2021-07-28[55]. 9491 9492References 9493 9494 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 9495 2. http://savannah.nongnu.org/bugs/?35407 9496 3. https://gcc.gnu.org/PR18145 9497 4. https://gcc.gnu.org/gcc-4.7/porting_to.html 9498 5. https://gcc.gnu.org/wiki/TransactionalMemory 9499 6. https://gcc.gnu.org/wiki/Atomic/GCCMM 9500 7. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 9501 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 9502 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 9503 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 9504 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 9505 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 9506 13. https://gcc.gnu.org/PR14258 9507 14. https://gcc.gnu.org/PR35688 9508 15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 9509 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 9510 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 9511 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 9512 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 9513 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 9514 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 9515 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 9516 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 9517 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 9518 25. https://docs.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps 9519 26. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 9520 27. https://gcc.gnu.org/wiki/Fortran2003Status 9521 28. https://gcc.gnu.org/wiki/OOP 9522 29. https://gcc.gnu.org/wiki/Fortran2008Status 9523 30. https://gcc.gnu.org/wiki/Coarray 9524 31. https://gcc.gnu.org/wiki/CoarrayLib 9525 32. https://gcc.gnu.org/wiki/TS29113Status 9526 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 9527 34. https://golang.org/doc/go1 9528 35. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 9529 36. http://nongnu.org/avr-libc/ 9530 37. https://gcc.gnu.org/PR54461 9531 38. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html 9532 39. https://sites.google.com/site/x32abi/ 9533 40. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1 9534 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2 9535 42. http://www.dwarfstd.org/ShowIssue.php?issue=140425.1 9536 43. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1 9537 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 9538 45. https://golang.org/doc/go1 9539 46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 9540 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 9541 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 9542 49. https://gcc.gnu.org/onlinedocs/ 9543 50. mailto:gcc-help@gcc.gnu.org 9544 51. mailto:gcc@gcc.gnu.org 9545 52. https://gcc.gnu.org/lists.html 9546 53. https://www.fsf.org/ 9547 54. https://gcc.gnu.org/about.html 9548 55. http://validator.w3.org/check/referer 9549====================================================================== 9550http://gcc.gnu.org/gcc-4.6/index.html 9551 9552 GCC 4.6 Release Series 9553 9554 (This release series is no longer supported.) 9555 9556 April 12, 2013 9557 9558 The [1]GNU project and the GCC developers are pleased to announce the 9559 release of GCC 4.6.4. 9560 9561 This release is a bug-fix release, containing fixes for regressions in 9562 GCC 4.6.3 relative to previous releases of GCC. 9563 9564Release History 9565 9566 GCC 4.6.4 9567 April 12, 2013 ([2]changes, [3]documentation) 9568 9569 GCC 4.6.3 9570 March 1, 2012 ([4]changes, [5]documentation) 9571 9572 GCC 4.6.2 9573 October 26, 2011 ([6]changes, [7]documentation) 9574 9575 GCC 4.6.1 9576 June 27, 2011 ([8]changes, [9]documentation) 9577 9578 GCC 4.6.0 9579 March 25, 2011 ([10]changes, [11]documentation) 9580 9581References and Acknowledgements 9582 9583 GCC used to stand for the GNU C Compiler, but since the compiler 9584 supports several other languages aside from C, it now stands for the 9585 GNU Compiler Collection. 9586 9587 A list of [12]successful builds is updated as new information becomes 9588 available. 9589 9590 The GCC developers would like to thank the numerous people that have 9591 contributed new features, improvements, bug fixes, and other changes as 9592 well as test results to GCC. This [13]amazing group of volunteers is 9593 what makes GCC successful. 9594 9595 For additional information about GCC please refer to the [14]GCC 9596 project web site or contact the [15]GCC development mailing list. 9597 9598 To obtain GCC please use [16]our mirror sites or [17]our version 9599 control system. 9600 9601 9602 For questions related to the use of GCC, please consult these web 9603 pages and the [18]GCC manuals. If that fails, the 9604 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 9605 web pages and the development of GCC are welcome on our developer 9606 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 9607 archives. 9608 9609 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 9610 distribution of this entire article is permitted in any medium, 9611 provided this notice is preserved. 9612 9613 These pages are [23]maintained by the GCC team. Last modified 9614 2021-07-28[24]. 9615 9616References 9617 9618 1. http://www.gnu.org/ 9619 2. http://gcc.gnu.org/gcc-4.6/changes.html 9620 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 9621 4. http://gcc.gnu.org/gcc-4.6/changes.html 9622 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 9623 6. http://gcc.gnu.org/gcc-4.6/changes.html 9624 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 9625 8. http://gcc.gnu.org/gcc-4.6/changes.html 9626 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 9627 10. http://gcc.gnu.org/gcc-4.6/changes.html 9628 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 9629 12. http://gcc.gnu.org/gcc-4.6/buildstat.html 9630 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9631 14. http://gcc.gnu.org/index.html 9632 15. mailto:gcc@gcc.gnu.org 9633 16. http://gcc.gnu.org/mirrors.html 9634 17. http://gcc.gnu.org/git.html 9635 18. https://gcc.gnu.org/onlinedocs/ 9636 19. mailto:gcc-help@gcc.gnu.org 9637 20. mailto:gcc@gcc.gnu.org 9638 21. https://gcc.gnu.org/lists.html 9639 22. https://www.fsf.org/ 9640 23. https://gcc.gnu.org/about.html 9641 24. http://validator.w3.org/check/referer 9642====================================================================== 9643http://gcc.gnu.org/gcc-4.6/changes.html 9644 9645 GCC 4.6 Release Series 9646 Changes, New Features, and Fixes 9647 9648Caveats 9649 9650 * The options -b <machine> and -V <version> have been removed because 9651 they were unreliable. Instead, users should directly run 9652 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 9653 run a different version of gcc. 9654 * GCC now has stricter checks for invalid command-line options. In 9655 particular, when gcc was called to link object files rather than 9656 compile source code, it would previously accept and ignore all 9657 options starting with --, including linker options such as 9658 --as-needed and --export-dynamic, although such options would 9659 result in errors if any source code was compiled. Such options, if 9660 unknown to the compiler, are now rejected in all cases; if the 9661 intent was to pass them to the linker, options such as 9662 -Wl,--as-needed should be used. 9663 * Versions of the GNU C library up to and including 2.11.1 included 9664 an [1]incorrect implementation of the cproj function. GCC optimizes 9665 its builtin cproj according to the behavior specified and allowed 9666 by the ISO C99 standard. If you want to avoid discrepancies between 9667 the C library and GCC's builtin transformations when using cproj in 9668 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 9669 and actually rely on the incorrect behavior of cproj, then you can 9670 disable GCC's transformations using -fno-builtin-cproj. 9671 * The C-only intermodule optimization framework (IMA, enabled by 9672 -combine) has been removed in favor of the new generic link-time 9673 optimization framework (LTO) introduced in [2]GCC 4.5.0. 9674 * GCC now ships with the LGPL-licensed libquadmath library, which 9675 provides quad-precision mathematical functions for targets with a 9676 __float128 datatype. __float128 is available for targets on 32-bit 9677 x86, x86-64 and Itanium architectures. The libquadmath library is 9678 automatically built on such targets when building the Fortran 9679 compiler. 9680 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 9681 warnings were added for C, C++, Objective-C and Objective-C++. 9682 These warnings diagnose variables respective parameters which are 9683 only set in the code and never otherwise used. Usually such 9684 variables are useless and often even the value assigned to them is 9685 computed needlessly, sometimes expensively. The 9686 -Wunused-but-set-variable warning is enabled by default by -Wall 9687 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 9688 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 9689 rules for the layout of vectors that could lead to wrong code being 9690 generated. Vectors larger than 8 bytes in size are now by default 9691 aligned to an 8-byte boundary. This is an ABI change: code that 9692 makes explicit use of vector types may be incompatible with binary 9693 objects built with older versions of GCC. Auto-vectorized code is 9694 not affected by this change. (This change affects GCC versions 9695 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 9696 * On AVR, variables with the progmem attribute to locate data in 9697 flash memory must be qualified as const. 9698 * Support for a number of older systems and recently unmaintained or 9699 untested target ports of GCC has been declared obsolete in GCC 4.6. 9700 Unless there is activity to revive them, the next release of GCC 9701 will have their sources permanently removed. 9702 All GCC ports for the following processor architectures have been 9703 declared obsolete: 9704 + Argonaut ARC (arc-*) 9705 + National Semiconductor CRX (crx-*) 9706 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 9707 m68hc12-*-*, m6812-*-*) 9708 + Sunplus S+core (score-*) 9709 The following ports for individual systems on particular 9710 architectures have been obsoleted: 9711 + Interix (i[34567]86-*-interix3*) 9712 + NetWare x86 (i[3456x]86-*-netware*) 9713 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 9714 + MCore PE (mcore-*-pe*) 9715 + SH SymbianOS (sh*-*-symbianelf*) 9716 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 9717 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 9718 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 9719 vax-*-netbsd*, but not *-*-netbsdelf*) 9720 The i[34567]86-*-pe alias for Cygwin targets has also been 9721 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 9722 Certain configure options to control the set of libraries built 9723 with GCC on some targets have been obsoleted. On ARM targets, the 9724 options --disable-fpu, --disable-26bit, --disable-underscore, 9725 --disable-interwork, --disable-biendian and --disable-nofmult have 9726 been obsoleted. On MIPS targets, the options 9727 --disable-single-float, --disable-biendian and --disable-softfloat 9728 have been obsoleted. 9729 * Support has been removed for all the [3]configurations obsoleted in 9730 GCC 4.5. 9731 * More information on porting to GCC 4.6 from previous versions of 9732 GCC can be found in the [4]porting guide for this release. 9733 9734General Optimizer Improvements 9735 9736 * A new general optimization level, -Ofast, has been introduced. It 9737 combines the existing optimization level -O3 with options that can 9738 affect standards compliance but result in better optimized code. 9739 For example, -Ofast enables -ffast-math. 9740 * Link-time optimization improvements: 9741 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 9742 stabilized to the point of being usable. It has become the 9743 default mode when using the LTO optimization model. Link time 9744 optimization can now split itself into multiple parallel 9745 compilations. Parallelism is controlled with -flto=n (where n 9746 specifies the number of compilations to execute in parallel). 9747 GCC can also cooperate with a GNU make job server by 9748 specifying the -flto=jobserver option and adding + to the 9749 beginning of the Makefile rule executing the linker. 9750 Classical LTO mode can be enforced by -flto-partition=none. 9751 This may result in small code quality improvements. 9752 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 9753 and other large applications can be built with LTO enabled. 9754 + The linker plugin support improvements 9755 o Linker plugin is now enabled by default when the linker 9756 is detected to have plugin support. This is the case for 9757 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 9758 the Gold linker on ELF targets. Plugin support of the 9759 Apple linker on Darwin is not compatible with GCC. The 9760 linker plugin can also be controlled by the 9761 -fuse-linker-plugin command-line option. 9762 o Resolution information from the linker plugin is used to 9763 drive whole program assumptions. Use of the linker plugin 9764 results in more aggressive optimization on binaries and 9765 on shared libraries that use the hidden visibility 9766 attribute. Consequently the use of -fwhole-program is not 9767 necessary in addition to LTO. 9768 + Hidden symbols used from non-LTO objects now have to be 9769 explicitly annotated with externally_visible when the linker 9770 plugin is not used. 9771 + C++ inline functions and virtual tables are now privatized 9772 more aggressively, leading to better inter-procedural 9773 optimization and faster dynamic linking. 9774 + Memory usage and intermediate language streaming performance 9775 have been improved. 9776 + Static constructors and destructors from individual units are 9777 inlined into a single function. This can significantly improve 9778 startup times of large C++ applications where static 9779 constructors are very common. For example, static constructors 9780 are used when including the iostream header. 9781 + Support for the Ada language has been added. 9782 * Interprocedural optimization improvements 9783 + The interprocedural framework was re-tuned for link time 9784 optimization. Several scalability issues were resolved. 9785 + Improved auto-detection of const and pure functions. Newly, 9786 noreturn functions are auto-detected. 9787 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 9788 available that informs users when adding attributes to headers 9789 might improve code generation. 9790 + A number of inlining heuristic improvements. In particular: 9791 o Partial inlining is now supported and enabled by default 9792 at -O2 and greater. The feature can be controlled via 9793 -fpartial-inlining. 9794 Partial inlining splits functions with short hot path to 9795 return. This allows more aggressive inlining of the hot 9796 path leading to better performance and often to code size 9797 reductions (because cold parts of functions are not 9798 duplicated). 9799 o Scalability for large compilation units was improved 9800 significantly. 9801 o Inlining of callbacks is now more aggressive. 9802 o Virtual methods are considered for inlining when the 9803 caller is inlined and devirtualization is then possible. 9804 o Inlining when optimizing for size (either in cold regions 9805 of a program or when compiling with -Os) was improved to 9806 better handle C++ programs with larger abstraction 9807 penalty, leading to smaller and faster code. 9808 + The IPA reference optimization pass detecting global variables 9809 used or modified by functions was strengthened and sped up. 9810 + Functions whose address was taken are now optimized out when 9811 all references to them are dead. 9812 + A new inter-procedural static profile estimation pass detects 9813 functions that are executed once or unlikely to be executed. 9814 Unlikely executed functions are optimized for size. Functions 9815 executed once are optimized for size except for the inner 9816 loops. 9817 + On most targets with named section support, functions used 9818 only at startup (static constructors and main), functions used 9819 only at exit and functions detected to be cold are placed into 9820 separate text segment subsections. This extends the 9821 -freorder-functions feature and is controlled by the same 9822 switch. The goal is to improve the startup time of large C++ 9823 programs. 9824 Proper function placement requires linker support. GNU ld 9825 2.21.51 on ELF targets was updated to place those functions 9826 together within the text section leading to better code 9827 locality and faster startup times of large C++ programs. The 9828 feature is also supported in the Apple linker. Support in the 9829 gold linker is planned. 9830 * A new switch -fstack-usage has been added. It makes the compiler 9831 output stack usage information for the program, on a per-function 9832 basis, in an auxiliary file. 9833 * A new switch -fcombine-stack-adjustments has been added. It can be 9834 used to enable or disable the compiler's stack-slot combining pass 9835 which before was enabled automatically at -O1 and above, but could 9836 not be controlled on its own. 9837 * A new switch -fstrict-volatile-bitfields has been added. Using it 9838 indicates that accesses to volatile bitfields should use a single 9839 access of the width of the field's type. This option can be useful 9840 for precisely defining and accessing memory-mapped peripheral 9841 registers from C or C++. 9842 9843Compile time and memory usage improvements 9844 9845 * Datastructures used by the dataflow framework in GCC were 9846 reorganized for better memory usage and more cache locality. 9847 Compile time is improved especially on units with large functions 9848 (possibly resulting from a lot of inlining) not fitting into the 9849 processor cache. The compile time of the GCC C compiler binary with 9850 link-time optimization went down by over 10% (benchmarked on x86-64 9851 target). 9852 9853New Languages and Language specific improvements 9854 9855 Ada 9856 9857 * Stack checking has been improved on selected architectures (Alpha, 9858 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 9859 overflows in all cases on these architectures. 9860 * Initial support for Ada 2012 has been added. 9861 9862 C family 9863 9864 * A new warning, enabled by -Wdouble-promotion, has been added that 9865 warns about cases where a value of type float is implicitly 9866 promoted to double. This is especially helpful for CPUs that handle 9867 the former in hardware, but emulate the latter in software. 9868 * A new function attribute leaf was introduced. This attribute allows 9869 better inter-procedural optimization across calls to functions that 9870 return to the current unit only via returning or exception 9871 handling. This is the case for most library functions that have no 9872 callbacks. 9873 * Support for a new data type __int128 for targets having wide enough 9874 machine-mode support. 9875 * The new function attribute callee_pop_aggregate allows to specify 9876 if the caller or callee is responsible for popping the aggregate 9877 return pointer value from the stack. 9878 * Support for selectively enabling and disabling warnings via #pragma 9879 GCC diagnostic has been added. For instance: 9880#pragma GCC diagnostic error "-Wuninitialized" 9881 foo(a); /* error is given for this one */ 9882#pragma GCC diagnostic push 9883#pragma GCC diagnostic ignored "-Wuninitialized" 9884 foo(b); /* no diagnostic for this one */ 9885#pragma GCC diagnostic pop 9886 foo(c); /* error is given for this one */ 9887#pragma GCC diagnostic pop 9888 foo(d); /* depends on command-line options */ 9889 9890 * The -fmax-errors=N option is now supported. Using this option 9891 causes the compiler to exit after N errors have been issued. 9892 9893 C 9894 9895 * There is now experimental support for some features from the 9896 upcoming C1X revision of the ISO C standard. This support may be 9897 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 9898 Note that this support is experimental and may change incompatibly 9899 in future releases for consistency with changes to the C1X standard 9900 draft. The following features are newly supported as described in 9901 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 9902 meeting); some other features were already supported with no 9903 compiler changes being needed, or have some support but not in full 9904 accord with N1539 (as amended). 9905 + Static assertions (_Static_assert keyword) 9906 + Typedef redefinition 9907 + New macros in <float.h> 9908 + Anonymous structures and unions 9909 * The new -fplan9-extensions option directs the compiler to support 9910 some extensions for anonymous struct fields which are implemented 9911 by the Plan 9 compiler. A pointer to a struct may be automatically 9912 converted to a pointer to an anonymous field when calling a 9913 function, in order to make the types match. An anonymous struct 9914 field whose type is a typedef name may be referred to using the 9915 typedef name. 9916 9917 C++ 9918 9919 * Improved [7]experimental support for the upcoming C++0x ISO C++ 9920 standard, including support for constexpr (thanks to Gabriel Dos 9921 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 9922 noexcept, unrestricted unions, range-based for loops (thanks to 9923 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 9924 Rodrigo), implicitly deleted functions and implicit move 9925 constructors. 9926 * When an extern declaration within a function does not match a 9927 declaration in the enclosing context, G++ now properly declares the 9928 name within the namespace of the function rather than the namespace 9929 which was open just before the function definition ([8]c++/43145). 9930 * GCC now warns by default when casting integers to larger pointer 9931 types. These warnings can be disabled with the option 9932 -Wno-int-to-pointer-cast, which is now also available in C++. 9933 * G++ no longer optimizes using the assumption that a value of 9934 enumeration type will fall within the range specified by the 9935 standard, since that assumption is easily violated with a 9936 conversion from integer type ([9]c++/43680). The old behavior can 9937 be restored with -fstrict-enums. 9938 * The new -fnothrow-opt flag changes the semantics of a throw() 9939 exception specification to match the proposed semantics of the 9940 noexcept specification: just call terminate if an exception tries 9941 to propagate out of a function with such an exception 9942 specification. This dramatically reduces or eliminates the code 9943 size overhead from adding the exception specification. 9944 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 9945 a function that the compiler can tell doesn't throw if it would 9946 change the value of a noexcept expression. 9947 * The -Wshadow option now warns if a local variable or type 9948 declaration shadows another type in C++. Note that the compiler 9949 will not warn if a local variable shadows a struct/class/enum, but 9950 will warn if it shadows an explicit typedef. 9951 * When an identifier is not found in the current scope, G++ now 9952 offers suggestions about which identifier might have been intended. 9953 * G++ now issues clearer diagnostics for missing semicolons after 9954 class, struct, and union definitions. 9955 * G++ now issues clearer diagnostics for missing semicolons after 9956 class member declarations. 9957 * G++ now issues clearer diagnostics when a colon is used in a place 9958 where a double-colon was intended. 9959 * G++ no longer accepts mutable on reference members ([10]c++/33558). 9960 Use -fpermissive to allow the old, non-conforming behaviour. 9961 * A few mangling fixes have been made, to attribute const/volatile on 9962 function pointer types, decltype of a plain decl, and use of a 9963 function parameter in the declaration of another parameter. By 9964 default the compiler still uses the old mangling, but emits aliases 9965 with the new mangling on targets that support strong aliases. Users 9966 can switch over entirely to the new mangling with -fabi-version=5 9967 or -fabi-version=0. -Wabi will now warn about code that uses the 9968 old mangling. 9969 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 9970 type to be default initialized unless the type has a user-declared 9971 default constructor. In 4.6.2 G++ implements the proposed 9972 resolution of [11]DR 253, so default initialization is allowed if 9973 it initializes all subobjects. Code that fails to compile can be 9974 fixed by providing an initializer e.g. 9975 struct A { A(); }; 9976 struct B : A { int i; }; 9977 const B b = B(); 9978 Use -fpermissive to allow the old, non-conforming behaviour. 9979 9980 Runtime Library (libstdc++) 9981 9982 * [12]Improved experimental support for the upcoming ISO C++ 9983 standard, C++0x, including using constexpr and nullptr. 9984 * Performance improvements to the [13]Debug Mode, thanks to Fran�ois 9985 Dumont. 9986 * Atomic operations used for reference-counting are annotated so that 9987 they can be understood by race detectors such as Helgrind, see 9988 [14]Data Race Hunting. 9989 * Most libstdc++ standard headers have been changed to no longer 9990 include the cstddef header as an implementation detail. Code that 9991 relied on that header being included as side-effect of including 9992 other standard headers will need to include cstddef explicitly. 9993 9994 Fortran 9995 9996 * On systems supporting the libquadmath library, GNU Fortran now also 9997 supports a quad-precision, kind=16 floating-point data type 9998 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 9999 hardware, calculations might be one to two orders of magnitude 10000 slower than with the 4, 8 or 10 bytes floating-point data types. 10001 This change does not affect systems which support REAL(16) in 10002 hardware nor those which do not support libquadmath. 10003 * Much improved compile time for large array constructors. 10004 * In order to reduce execution time and memory consumption, use of 10005 temporary arrays in assignment expressions is avoided for many 10006 cases. The compiler now reverses loops in order to avoid generating 10007 a temporary array where possible. 10008 * Improved diagnostics, especially with -fwhole-file. 10009 * The -fwhole-file flag is now enabled by default. This improves code 10010 generation and diagnostics. It can be disabled using the deprecated 10011 -fno-whole-file flag. 10012 * Support the generation of Makefile dependencies via the [15]-M... 10013 flags of GCC; you may need to specify the -cpp option in addition. 10014 The dependencies take modules, Fortran's include, and CPP's 10015 #include into account. Note: Using -M for the module path is no 10016 longer supported, use -J instead. 10017 * The flag -Wconversion has been modified to only issue warnings 10018 where a conversion leads to information loss. This drastically 10019 reduces the number of warnings; -Wconversion is thus now enabled 10020 with -Wall. The flag -Wconversion-extra has been added and also 10021 warns about other conversions; -Wconversion-extra typically issues 10022 a huge number of warnings, most of which can be ignored. 10023 * A new command-line option -Wunused-dummy-argument warns about 10024 unused dummy arguments and is included in -Wall. Before, 10025 -Wunused-variable also warned about unused dummy arguments. 10026 * Fortran 2003 support has been extended: 10027 + Improved support for polymorphism between libraries and 10028 programs and for complicated inheritance patterns (cf. 10029 [16]object-oriented programming). 10030 + Experimental support of the ASSOCIATE construct. 10031 + In pointer assignments it is now possible to specify the lower 10032 bounds of the pointer and, for a rank-1 or a simply contiguous 10033 data-target, to remap the bounds. 10034 + Automatic (re)allocation: In intrinsic assignments to 10035 allocatable variables the left-hand side will be automatically 10036 allocated (if unallocated) or reallocated (if the shape or 10037 type parameter is different). To avoid the small performance 10038 penalty, you can use a(:) = ... instead of a = ... for arrays 10039 and character strings - or disable the feature using -std=f95 10040 or -fno-realloc-lhs. 10041 + Deferred type parameter: For scalar allocatable and pointer 10042 variables the character length can be deferred. 10043 + Namelist variables with allocatable and pointer attribute and 10044 nonconstant length type parameter are supported. 10045 * Fortran 2008 support has been extended: 10046 + Experimental [17]coarray support (for one image only, i.e. 10047 num_images() == 1); use the [18]-fcoarray=single flag to 10048 enable it. 10049 + The STOP and the new ERROR STOP statements now support all 10050 constant expressions. 10051 + Support for the CONTIGUOUS attribute. 10052 + Support for ALLOCATE with MOLD. 10053 + Support for the STORAGE_SIZE intrinsic inquiry function. 10054 + Support of the NORM2 and PARITY intrinsic functions. 10055 + The following bit intrinsics were added: POPCNT and POPPAR for 10056 counting the number of 1 bits and returning the parity; BGE, 10057 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 10058 for combined left and right shifts, MASKL and MASKR for simple 10059 left and right justified masks, MERGE_BITS for a bitwise merge 10060 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 10061 and the transformational bit intrinsics IALL, IANY and 10062 IPARITY. 10063 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 10064 + Support for the IMPURE attribute for procedures, which allows 10065 for ELEMENTAL procedures without the restrictions of PURE. 10066 + Null pointers (including NULL()) and not allocated variables 10067 can be used as actual argument to optional non-pointer, 10068 non-allocatable dummy arguments, denoting an absent argument. 10069 + Non-pointer variables with TARGET attribute can be used as 10070 actual argument to POINTER dummies with INTENT(IN) 10071 + Pointers including procedure pointers and those in a derived 10072 type (pointer components) can now be initialized by a target 10073 instead of only by NULL. 10074 + The EXIT statement (with construct-name) can now be used to 10075 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 10076 SELECT CASE and SELECT TYPE constructs. 10077 + Internal procedures can now be used as actual argument. 10078 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 10079 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 10080 have been added; these arrays contain the supported kind 10081 values for the respective types. 10082 + The module procedures C_SIZEOF of the intrinsic module 10083 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 10084 ISO_FORTRAN_ENV have been implemented. 10085 + Minor changes: obsolescence diagnostics for ENTRY was added 10086 for -std=f2008; a line may start with a semicolon; for 10087 internal and module procedures END can be used instead of END 10088 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 10089 a RADIX argument; intrinsic types are supported for 10090 TYPE(intrinsic-type-spec); multiple type-bound procedures can 10091 be declared in a single PROCEDURE statement; implied-shape 10092 arrays are supported for named constants (PARAMETER). The 10093 transformational, three argument versions of BESSEL_JN and 10094 BESSEL_YN were added - the elemental, two-argument version had 10095 been added in GCC 4.4; note that the transformational 10096 functions use a recurrence algorithm. 10097 10098 Go 10099 10100 Support for the [19]Go programming language has been added to GCC. It 10101 is not enabled by default when you build GCC; use the 10102 --enable-languages configure option to build it. The driver program for 10103 compiling Go code is gccgo. 10104 10105 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 10106 is in progress. It may or may not work on other platforms. 10107 10108 Objective-C and Objective-C++ 10109 10110 * The -fobjc-exceptions flag is now required to enable Objective-C 10111 exception and synchronization syntax (introduced by the keywords 10112 @try, @catch, @finally and @synchronized). 10113 * A number of Objective-C 2.0 features and extensions are now 10114 supported by GCC. These features are enabled by default; you can 10115 disable them by using the new -fobjc-std=objc1 command-line option. 10116 * The Objective-C 2.0 dot-syntax is now supported. It is an 10117 alternative syntax for using getters and setters; object.count is 10118 automatically converted into [object count] or [object setCount: 10119 ...] depending on context; for example if (object.count > 0) is 10120 automatically compiled into the equivalent of if ([object count] > 10121 0) while object.count = 0; is automatically compiled into the 10122 equivalent ot [object setCount: 0];. The dot-syntax can be used 10123 with instance and class objects and with any setters or getters, no 10124 matter if they are part of a declared property or not. 10125 * Objective-C 2.0 declared properties are now supported. They are 10126 declared using the new @property keyword, and are most commonly 10127 used in conjunction with the new Objective-C 2.0 dot-syntax. The 10128 nonatomic, readonly, readwrite, assign, retain, copy, setter and 10129 getter attributes are all supported. Marking declared properties 10130 with __attribute__ ((deprecated)) is supported too. 10131 * The Objective-C 2.0 @synthesize and @dynamic keywords are 10132 supported. @synthesize causes the compiler to automatically 10133 synthesize a declared property, while @dynamic is used to disable 10134 all warnings for a declared property for which no implementation is 10135 provided at compile time. Synthesizing declared properties requires 10136 runtime support in most useful cases; to be able to use it with the 10137 GNU runtime, appropriate helper functions have been added to the 10138 GNU Objective-C runtime ABI, and are implemented by the GNU 10139 Objective-C runtime library shipped with GCC. 10140 * The Objective-C 2.0 fast enumeration syntax is supported in 10141 Objective-C. This is currently not yet available in Objective-C++. 10142 Fast enumeration requires support in the runtime, and such support 10143 has been added to the GNU Objective-C runtime library (shipped with 10144 GCC). 10145 * The Objective-C 2.0 @optional keyword is supported. It allows you 10146 to mark methods or properties in a protocol as optional as opposed 10147 to required. 10148 * The Objective-C 2.0 @package keyword is supported. It has currently 10149 the same effect as the @public keyword. 10150 * Objective-C 2.0 method attributes are supported. Currently the 10151 supported attributes are deprecated, sentinel, noreturn and format. 10152 * Objective-C 2.0 method argument attributes are supported. The most 10153 widely used attribute is unused, to mark an argument as unused in 10154 the implementation. 10155 * Objective-C 2.0 class and protocol attributes are supported. 10156 Currently the only supported attribute is deprecated. 10157 * Objective-C 2.0 class extensions are supported. A class extension 10158 has the same syntax as a category declaration with no category 10159 name, and the methods and properties declared in it are added 10160 directly to the main class. It is mostly used as an alternative to 10161 a category to add methods to a class without advertising them in 10162 the public headers, with the advantage that for class extensions 10163 the compiler checks that all the privately declared methods are 10164 actually implemented. 10165 * As a result of these enhancements, GCC can now be used to build 10166 Objective-C and Objective-C++ software that uses Foundation and 10167 other important system frameworks with the NeXT runtime on Darwin 9 10168 and Darwin 10 (OSX 10.5 and 10.6). 10169 * Many bugs in the compiler have been fixed in this release; in 10170 particular, LTO can now be used when compiling Objective-C and 10171 Objective-C++ and the parser is much more robust in dealing with 10172 invalid code. 10173 10174 Runtime Library (libobjc) 10175 10176 * The GNU Objective-C runtime library now defines the macro 10177 __GNU_LIBOBJC__ (with a value that is increased at every release 10178 where there is any change to the API) in objc/objc.h, making it 10179 easy to determine if the GNU Objective-C runtime library is being 10180 used, and if so, which version. Previous versions of the GNU 10181 Objective-C runtime library (and other Objective-C runtime 10182 libraries such as the Apple one) do not define this macro. 10183 * A new Objective-C 2.0 API, almost identical to the one implemented 10184 by the Apple Objective-C runtime, has been implemented in the GNU 10185 Objective-C runtime library. The new API hides the internals of 10186 most runtime structures but provides a more extensive set of 10187 functions to operate on them. It is much easier, for example, to 10188 create or modify classes at runtime. The new API also makes it 10189 easier to port software from Apple to GNU as almost no changes 10190 should be required. The old API is still supported for backwards 10191 compatibility; including the old objc/objc-api.h header file 10192 automatically selects the old API, while including the new 10193 objc/runtime.h header file automatically selects the new API. 10194 Support for the old API is being phased out and upgrading the 10195 software to use the new API is strongly recommended. To check for 10196 the availability of the new API, the __GNU_LIBOBJC__ macro can be 10197 used as older versions of the GNU Objective-C runtime library, 10198 which do not support the new API, do not define such a macro. 10199 * Runtime support for @synchronized has been added. 10200 * Runtime support for Objective-C 2.0 synthesized property accessors 10201 has been added. 10202 * Runtime support for Objective-C 2.0 fast enumeration has been 10203 added. 10204 10205New Targets and Target Specific Improvements 10206 10207 ARM 10208 10209 * GCC now supports the Cortex-M4 processor implementing the v7-em 10210 version of the architecture using the option -mcpu=cortex-m4. 10211 * Scheduling descriptions for the Cortex-M4, the Neon and the 10212 floating point units of the Cortex-A9 and a pipeline description 10213 for the Cortex-A5 have been added. 10214 * Synchronization primitives such as __sync_fetch_and_add and friends 10215 are now inlined for supported architectures rather than calling 10216 into a kernel helper function. 10217 * SSA loop prefetching is enabled by default for the Cortex-A9 at 10218 -O3. 10219 * Several improvements were committed to improve code generation for 10220 the ARM architecture including a rewritten implementation for load 10221 and store multiples. 10222 * Several enhancements were committed to improve SIMD code generation 10223 for NEON by adding support for widening instructions, misaligned 10224 loads and stores, vector conditionals and support for 64 bit 10225 arithmetic. 10226 * Support was added for the Faraday cores fa526, fa606te, fa626te, 10227 fmp626te, fmp626 and fa726te and can be used with the respective 10228 names as parameters to the -mcpu= option. 10229 * Basic support was added for Cortex-A15 and is available through 10230 -mcpu=cortex-a15. 10231 * GCC for AAPCS configurations now more closely adheres to the AAPCS 10232 specification by enabling -fstrict-volatile-bitfields by default. 10233 10234 IA-32/x86-64 10235 10236 * The new -fsplit-stack option permits programs to use a 10237 discontiguous stack. This is useful for threaded programs, in that 10238 it is no longer necessary to specify the maximum stack size when 10239 creating a thread. This feature is currently only implemented for 10240 32-bit and 64-bit x86 GNU/Linux targets. 10241 * Support for emitting profiler counter calls before function 10242 prologues. This is enabled via a new command-line option -mfentry. 10243 * Optimization for the Intel Core 2 processors is now available 10244 through the -march=core2 and -mtune=core2 options. 10245 * Support for Intel Core i3/i5/i7 processors is now available through 10246 the -march=corei7 and -mtune=corei7 options. 10247 * Support for Intel Core i3/i5/i7 processors with AVX is now 10248 available through the -march=corei7-avx and -mtune=corei7-avx 10249 options. 10250 * Support for AMD Bobcat (family 14) processors is now available 10251 through the -march=btver1 and -mtune=btver1 options. 10252 * Support for AMD Bulldozer (family 15) processors is now available 10253 through the -march=bdver1 and -mtune=bdver1 options. 10254 * The default setting (when not optimizing for size) for 32-bit 10255 GNU/Linux and Darwin x86 targets has been changed to 10256 -fomit-frame-pointer. The default can be reverted to 10257 -fno-omit-frame-pointer by configuring GCC with the 10258 --enable-frame-pointer configure option. 10259 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 10260 __float128 on 32-bit and 64-bit x86 targets. 10261 * AVX floating-point arithmetic can now be enabled by default at 10262 configure time with the new --with-fpmath=avx option. 10263 * The SSA loop prefetching pass is enabled when using -O3 when 10264 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 10265 than K6). 10266 * Support for TBM (Trailing Bit Manipulation) built-in functions and 10267 code generation is available via -mtbm. 10268 * Support for AMD's BMI (Bit Manipulation) built-in functions and 10269 code generation is available via -mbmi. 10270 10271 MicroBlaze 10272 10273 * Support has been added for the Xilinx MicroBlaze softcore processor 10274 (microblaze-elf) embedded target. This configurable processor is 10275 supported on several Xilinx Spartan and Virtex FPGAs. 10276 10277 MIPS 10278 10279 * GCC now supports the Loongson 3A processor. Its canonical -march= 10280 and -mtune= name is loongson3a. 10281 10282 MN10300 / AM33 10283 10284 * The inline assembly register constraint "A" has been renamed "c". 10285 This constraint is used to select a floating-point register that 10286 can be used as the destination of a multiply-accumulate 10287 instruction. 10288 * New inline assembly register constraints "A" and "D" have been 10289 added. These constraint letters resolve to all general registers 10290 when compiling for AM33, and resolve to address registers only or 10291 data registers only when compiling for MN10300. 10292 * The MDR register is represented in the compiler. One can access the 10293 register via the "z" constraint in inline assembly. It can be 10294 marked as clobbered or used as a local register variable via the 10295 "mdr" name. The compiler uses the RETF instruction if the function 10296 does not modify the MDR register, so it is important that inline 10297 assembly properly annotate any usage of the register. 10298 10299 PowerPC/PowerPC64 10300 10301 * GCC now supports the Applied Micro Titan processor with 10302 -mcpu=titan. 10303 * The -mrecip option has been added, which indicates whether the 10304 reciprocal and reciprocal square root instructions should be used. 10305 * The -mveclibabi=mass option can be used to enable the compiler to 10306 autovectorize mathematical functions using the Mathematical 10307 Acceleration Subsystem library. 10308 * The -msingle-pic-base option has been added, which instructs the 10309 compiler to avoid loading the PIC base register in function 10310 prologues. The PIC base register must be initialized by the runtime 10311 system. 10312 * The -mblock-move-inline-limit option has been added, which enables 10313 the user to control the maximum size of inlined memcpy calls and 10314 similar. 10315 * PowerPC64 GNU/Linux support for applications requiring a large TOC 10316 section has been improved. A new command-line option, 10317 -mcmodel=MODEL, controls this feature; valid values for MODEL are 10318 small, medium, or large. 10319 * The Altivec builtin functions vec_ld and vec_st have been modified 10320 to generate the Altivec memory instructions LVX and STVX, even if 10321 the -mvsx option is used. In the initial GCC 4.5 release, these 10322 builtin functions were changed to generate VSX memory reference 10323 instructions instead of Altivec memory instructions, but there are 10324 differences between the two instructions. If the VSX instruction 10325 set is available, you can now use the new builtin functions 10326 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 10327 instructions. 10328 * The GCC compiler on AIX now defaults to a process layout with a 10329 larger data space allowing larger programs to be compiled. 10330 * The GCC long double type on AIX 6.1 and above has reverted to 64 10331 bit double precision, matching the AIX XL compiler default, because 10332 of missing C99 symbols required by the GCC runtime. 10333 * The default processor scheduling model and tuning for PowerPC64 10334 GNU/Linux and for AIX 6.1 and above now is POWER7. 10335 * Starting with GCC 4.6.1, vectors of type vector long long or vector 10336 long are passed and returned in the same method as other vectors 10337 with the VSX instruction set. Previously the GCC compiler did not 10338 adhere to the ABI for 128-bit vectors with 64-bit integer base 10339 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 10340 10341 S/390, zSeries and System z9/z10, IBM zEnterprise z196 10342 10343 * Support for the zEnterprise z196 processor has been added. When 10344 using the -march=z196 option, the compiler will generate code 10345 making use of the following instruction facilities: 10346 + Conditional load/store 10347 + Distinct-operands 10348 + Floating-point-extension 10349 + Interlocked-access 10350 + Population-count 10351 The -mtune=z196 option avoids the compare and branch instructions 10352 as well as the load address instruction with an index register as 10353 much as possible and performs instruction scheduling appropriate 10354 for the new out-of-order pipeline architecture. 10355 * When using the -m31 -mzarch options the generated code still 10356 conforms to the 32-bit ABI but uses the general purpose registers 10357 as 64-bit registers internally. This requires a Linux kernel saving 10358 the whole 64-bit registers when doing a context switch. Kernels 10359 providing that feature indicate that by the 'highgprs' string in 10360 /proc/cpuinfo. 10361 * The SSA loop prefetching pass is enabled when using -O3. 10362 10363 SPARC 10364 10365 * GCC now supports the LEON series of SPARC V8 processors. The code 10366 generated by the compiler can either be tuned to it by means of the 10367 --with-tune=leon configure option and -mtune=leon compilation 10368 option, or the compiler can be built for the sparc-leon-{elf,linux} 10369 and sparc-leon3-{elf,linux} targets directly. 10370 * GCC has stopped sign/zero-extending parameter registers in the 10371 callee for functions taking parameters with sub-word size in 32-bit 10372 mode, since this is redundant with the specification of the ABI. 10373 GCC has never done so in 64-bit mode since this is also redundant. 10374 * The command-line option -mfix-at697f has been added to enable the 10375 documented workaround for the single erratum of the Atmel AT697F 10376 processor. 10377 10378Operating Systems 10379 10380 Android 10381 10382 * GCC now supports the Bionic C library and provides a convenient way 10383 of building native libraries and applications for the Android 10384 platform. Refer to the documentation of the -mandroid and -mbionic 10385 options for details on building native code. At the moment, Android 10386 support is enabled only for ARM. 10387 10388 Darwin/Mac OS X 10389 10390 * General 10391 + Initial support for CFString types has been added. 10392 This allows GCC to build projects including the system Core 10393 Foundation frameworks. The GCC Objective-C family supports 10394 CFString "toll-free bridged" as per the Mac OS X system tools. 10395 CFString is also recognized in the context of format 10396 attributes and arguments (see the documentation for format 10397 attributes for limitations). At present, 8-bit character types 10398 are supported. 10399 + Object file size reduction. 10400 The Darwin zeroed memory allocators have been re-written to 10401 make more use of .zerofill sections. For non-debug code, this 10402 can reduce object file size significantly. 10403 + Objective-C family 64-bit support (NeXT ABI 2). 10404 Initial support has been added to support 64-bit Objective-C 10405 code using the Darwin/OS X native (NeXT) runtime. ABI version 10406 2 will be selected automatically when 64-bit code is built. 10407 + Objective-C family 32-bit ABI 1. 10408 For 32-bit code ABI 1 is also now also allowed. At present it 10409 must be selected manually using -fobjc-abi-version=1 where 10410 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 10411 * x86 Architecture 10412 + The -mdynamic-no-pic option has been enabled. 10413 Code supporting -mdynamic-no-pic optimization has been added 10414 and is applicable to -m32 builds. The compiler bootstrap uses 10415 the option where appropriate. 10416 + The default value for -mtune= has been changed. 10417 Since Darwin systems are primarily Xeon, Core-2 or similar the 10418 default tuning has been changed to -mtune=core2. 10419 + Enable 128-bit long double (__float128) support on Darwin. 10420 * PPC Architecture 10421 + Darwin64 ABI. 10422 Several significant bugs have been fixed, such that GCC now 10423 produces code compatible with the Darwin64 PowerPC ABI. 10424 + libffi and boehm-gc. 10425 The Darwin ports of the libffi and boehm-gc libraries have 10426 been upgraded to include a Darwin64 implementation. This means 10427 that powerpc*-*-darwin9 platforms may now, for example, build 10428 Java applications with -m64 enabled. 10429 + Plug-in support has been enabled. 10430 + The -fsection-anchors option is now available although, 10431 presently, not heavily tested. 10432 10433 Solaris 2 10434 10435 New Features 10436 10437 * Support symbol versioning with the Sun linker. 10438 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 10439 * Support thread-local storage (TLS) with the Sun assembler on 10440 Solaris 2/x86. 10441 * Support TLS on Solaris 8/9 if prerequisites are met. 10442 * Support COMDAT group with the GNU assembler and recent Sun linker. 10443 * Support the Sun assembler visibility syntax. 10444 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 10445 -march=pentiumpro (Solaris 8/9). 10446 * Don't use SSE on Solaris 8/9 x86 by default. 10447 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 10448 10449 ABI Change 10450 10451 * Change the ABI for returning 8-byte vectors like __m64 in MMX 10452 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 10453 compilers. This is an incompatible change. If you use such types, 10454 you must either recompile all your code with the new compiler or 10455 use the new -mvect8-ret-in-mem option to remain compatible with 10456 previous versions of GCC and Sun Studio. 10457 10458 Windows x86/x86_64 10459 10460 * Initial support for decimal floating point. 10461 * Support for the __thiscall calling-convention. 10462 * Support for hot-patchable function prologues via the 10463 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 10464 * Improvements of stack-probing and stack-allocation mechanisms. 10465 * Support of push/pop-macro pragma as preprocessor command. 10466 With #pragma push_macro("macro-name") the current definition of 10467 macro-name is saved and can be restored with #pragma 10468 pop_macro("macro-name") to its saved definition. 10469 * Enable 128-bit long double (__float128) support on MinGW and 10470 Cygwin. 10471 10472Other significant improvements 10473 10474 Installation changes 10475 10476 * An install-strip make target is provided that installs stripped 10477 executables, and may install libraries with unneeded or debugging 10478 sections stripped. 10479 * On Power7 systems, there is a potential problem if you build the 10480 GCC compiler with a host compiler using options that enable the VSX 10481 instruction set generation. If the host compiler has been patched 10482 so that the vec_ld and vec_st builtin functions generate Altivec 10483 memory instructions instead of VSX memory instructions, then you 10484 should be able to build the compiler with VSX instruction 10485 generation. 10486 10487Changes for GCC Developers 10488 10489 Note: these changes concern developers that develop GCC itself or 10490 software that integrates with GCC, such as plugins, and not the general 10491 GCC users. 10492 * The gengtype utility, which previously was internal to the GCC 10493 build process, has been enchanced to provide GC root information 10494 for plugins as necessary. 10495 * The old GC allocation interface of ggc_alloc and friends was 10496 replaced with a type-safe alternative. 10497 10498GCC 4.6.1 10499 10500 This is the [20]list of problem reports (PRs) from GCC's bug tracking 10501 system that are known to be fixed in the 4.6.1 release. This list might 10502 not be complete (that is, it is possible that some PRs that have been 10503 fixed are not listed here). 10504 10505GCC 4.6.2 10506 10507 This is the [21]list of problem reports (PRs) from GCC's bug tracking 10508 system that are known to be fixed in the 4.6.2 release. This list might 10509 not be complete (that is, it is possible that some PRs that have been 10510 fixed are not listed here). 10511 10512GCC 4.6.3 10513 10514 This is the [22]list of problem reports (PRs) from GCC's bug tracking 10515 system that are known to be fixed in the 4.6.3 release. This list might 10516 not be complete (that is, it is possible that some PRs that have been 10517 fixed are not listed here). 10518 10519GCC 4.6.4 10520 10521 This is the [23]list of problem reports (PRs) from GCC's bug tracking 10522 system that are known to be fixed in the 4.6.4 release. This list might 10523 not be complete (that is, it is possible that some PRs that have been 10524 fixed are not listed here). 10525 10526 10527 For questions related to the use of GCC, please consult these web 10528 pages and the [24]GCC manuals. If that fails, the 10529 [25]gcc-help@gcc.gnu.org mailing list might help. Comments on these 10530 web pages and the development of GCC are welcome on our developer 10531 list at [26]gcc@gcc.gnu.org. All of [27]our lists have public 10532 archives. 10533 10534 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and 10535 distribution of this entire article is permitted in any medium, 10536 provided this notice is preserved. 10537 10538 These pages are [29]maintained by the GCC team. Last modified 10539 2021-07-28[30]. 10540 10541References 10542 10543 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 10544 2. http://gcc.gnu.org/gcc-4.5/changes.html 10545 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 10546 4. http://gcc.gnu.org/gcc-4.6/porting_to.html 10547 5. http://gcc.gnu.org/projects/lto/whopr.pdf 10548 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 10549 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html 10550 8. https://gcc.gnu.org/PR43145 10551 9. https://gcc.gnu.org/PR43680 10552 10. https://gcc.gnu.org/PR33558 10553 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 10554 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 10555 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 10556 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 10557 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 10558 16. https://gcc.gnu.org/wiki/OOP 10559 17. https://gcc.gnu.org/wiki/Coarray 10560 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 10561 19. https://golang.org/ 10562 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 10563 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 10564 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 10565 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 10566 24. https://gcc.gnu.org/onlinedocs/ 10567 25. mailto:gcc-help@gcc.gnu.org 10568 26. mailto:gcc@gcc.gnu.org 10569 27. https://gcc.gnu.org/lists.html 10570 28. https://www.fsf.org/ 10571 29. https://gcc.gnu.org/about.html 10572 30. http://validator.w3.org/check/referer 10573====================================================================== 10574http://gcc.gnu.org/gcc-4.5/index.html 10575 10576 GCC 4.5 Release Series 10577 10578 (This release series is no longer supported.) 10579 10580 Jul 2, 2012 10581 10582 The [1]GNU project and the GCC developers are pleased to announce the 10583 release of GCC 4.5.4. 10584 10585 This release is a bug-fix release, containing fixes for regressions in 10586 GCC 4.5.3 relative to previous releases of GCC. 10587 10588Release History 10589 10590 GCC 4.5.4 10591 Jul 2, 2012 ([2]changes) 10592 10593 GCC 4.5.3 10594 Apr 28, 2011 ([3]changes) 10595 10596 GCC 4.5.2 10597 Dec 16, 2010 ([4]changes) 10598 10599 GCC 4.5.1 10600 Jul 31, 2010 ([5]changes) 10601 10602 GCC 4.5.0 10603 April 14, 2010 ([6]changes) 10604 10605References and Acknowledgements 10606 10607 GCC used to stand for the GNU C Compiler, but since the compiler 10608 supports several other languages aside from C, it now stands for the 10609 GNU Compiler Collection. 10610 10611 A list of [7]successful builds is updated as new information becomes 10612 available. 10613 10614 The GCC developers would like to thank the numerous people that have 10615 contributed new features, improvements, bug fixes, and other changes as 10616 well as test results to GCC. This [8]amazing group of volunteers is 10617 what makes GCC successful. 10618 10619 For additional information about GCC please refer to the [9]GCC project 10620 web site or contact the [10]GCC development mailing list. 10621 10622 To obtain GCC please use [11]our mirror sites or [12]our version 10623 control system. 10624 10625 10626 For questions related to the use of GCC, please consult these web 10627 pages and the [13]GCC manuals. If that fails, the 10628 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these 10629 web pages and the development of GCC are welcome on our developer 10630 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public 10631 archives. 10632 10633 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 10634 distribution of this entire article is permitted in any medium, 10635 provided this notice is preserved. 10636 10637 These pages are [18]maintained by the GCC team. Last modified 10638 2021-07-28[19]. 10639 10640References 10641 10642 1. http://www.gnu.org/ 10643 2. http://gcc.gnu.org/gcc-4.5/changes.html 10644 3. http://gcc.gnu.org/gcc-4.5/changes.html 10645 4. http://gcc.gnu.org/gcc-4.5/changes.html 10646 5. http://gcc.gnu.org/gcc-4.5/changes.html 10647 6. http://gcc.gnu.org/gcc-4.5/changes.html 10648 7. http://gcc.gnu.org/gcc-4.5/buildstat.html 10649 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10650 9. http://gcc.gnu.org/index.html 10651 10. mailto:gcc@gcc.gnu.org 10652 11. http://gcc.gnu.org/mirrors.html 10653 12. http://gcc.gnu.org/git.html 10654 13. https://gcc.gnu.org/onlinedocs/ 10655 14. mailto:gcc-help@gcc.gnu.org 10656 15. mailto:gcc@gcc.gnu.org 10657 16. https://gcc.gnu.org/lists.html 10658 17. https://www.fsf.org/ 10659 18. https://gcc.gnu.org/about.html 10660 19. http://validator.w3.org/check/referer 10661====================================================================== 10662http://gcc.gnu.org/gcc-4.5/changes.html 10663 10664 GCC 4.5 Release Series 10665 Changes, New Features, and Fixes 10666 10667Caveats 10668 10669 * GCC now requires the [1]MPC library in order to build. See the 10670 [2]prerequisites page for version requirements. 10671 * Support for a number of older systems and recently unmaintained or 10672 untested target ports of GCC has been declared obsolete in GCC 4.5. 10673 Unless there is activity to revive them, the next release of GCC 10674 will have their sources permanently removed. 10675 The following ports for individual systems on particular 10676 architectures have been obsoleted: 10677 + IRIX releases before 6.5 (mips-sgi-irix5*, 10678 mips-sgi-irix6.[0-4]) 10679 + Solaris 7 (*-*-solaris2.7) 10680 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 10681 alpha-dec-osf5.0*) 10682 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 10683 can be found in the [3]announcement. 10684 Support for the classic POWER architecture implemented in the 10685 original RIOS and RIOS2 processors of the old IBM RS/6000 product 10686 line has been obsoleted in the rs6000 port. This does not affect 10687 the new generation Power and PowerPC architectures. 10688 * Support has been removed for all the [4]configurations obsoleted in 10689 GCC 4.4. 10690 * Support has been removed for the protoize and unprotoize utilities, 10691 obsoleted in GCC 4.4. 10692 * Support has been removed for tuning for Itanium1 (Merced) variants. 10693 Note that code tuned for Itanium2 should also run correctly on 10694 Itanium1. 10695 * GCC now generates unwind info also for epilogues. DWARF debuginfo 10696 generated by GCC now uses more features of DWARF3 than before, and 10697 also some DWARF4 features. GDB older than 7.0 is not able to handle 10698 either of these, so to debug GCC 4.5 generated binaries or 10699 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 10700 features with the -gdwarf-3 -gstrict-dwarf options, or use 10701 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 10702 epilogue unwind info is emitted unconditionally whenever unwind 10703 info is emitted. 10704 * On x86 targets, code containing floating-point calculations may run 10705 significantly more slowly when compiled with GCC 4.5 in strict C99 10706 conformance mode than they did with earlier GCC versions. This is 10707 due to stricter standard conformance of the compiler and can be 10708 avoided by using the option -fexcess-precision=fast; also see 10709 [5]below. 10710 * The function attribute noinline no longer prevents GCC from cloning 10711 the function. A new attribute noclone has been introduced for this 10712 purpose. Cloning a function means that it is duplicated and the new 10713 copy is specialized for certain contexts (for example when a 10714 parameter is a known constant). 10715 10716General Optimizer Improvements 10717 10718 * The -save-temps now takes an optional argument. The -save-temps and 10719 -save-temps=cwd switches write the temporary files in the current 10720 working directory based on the original source file. The 10721 -save-temps=obj switch will write files into the directory 10722 specified with the -o option, and the intermediate filenames are 10723 based on the output file. This will allow the user to get the 10724 compiler intermediate files when doing parallel builds without two 10725 builds of the same filename located in different directories from 10726 interfering with each other. 10727 * Debugging dumps are now created in the same directory as the object 10728 file rather than in the current working directory. This allows the 10729 user to get debugging dumps when doing parallel builds without two 10730 builds of the same filename interfering with each other. 10731 * GCC has been integrated with the MPC library. This allows GCC to 10732 evaluate complex arithmetic at compile time [6]more accurately. It 10733 also allows GCC to evaluate calls to complex built-in math 10734 functions having constant arguments and replace them at compile 10735 time with their mathematically equivalent results. In doing so, GCC 10736 can generate correct results regardless of the math library 10737 implementation or floating point precision of the host platform. 10738 This also allows GCC to generate identical results regardless of 10739 whether one compiles in native or cross-compile configurations to a 10740 particular target. The following built-in functions take advantage 10741 of this new capability: cacos, cacosh, casin, casinh, catan, 10742 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 10743 and ctanh. The float and long double variants of these functions 10744 (e.g. csinf and csinl) are also handled. 10745 * A new link-time optimizer has been added ([7]-flto). When this 10746 option is used, GCC generates a bytecode representation of each 10747 input file and writes it to specially-named sections in each object 10748 file. When the object files are linked together, all the function 10749 bodies are read from these named sections and instantiated as if 10750 they had been part of the same translation unit. This enables 10751 interprocedural optimizations to work across different files (and 10752 even different languages), potentially improving the performance of 10753 the generated code. To use the link-timer optimizer, -flto needs to 10754 be specified at compile time and during the final link. If the 10755 program does not require any symbols to be exported, it is possible 10756 to combine -flto and the experimental [8]-fwhopr with 10757 [9]-fwhole-program to allow the interprocedural optimizers to use 10758 more aggressive assumptions. 10759 * The automatic parallelization pass was enhanced to support 10760 parallelization of outer loops. 10761 * Automatic parallelization can be enabled as part of Graphite. In 10762 addition to -ftree-parallelize-loops=, specify 10763 -floop-parallelize-all to enable the Graphite-based optimization. 10764 * The infrastructure for optimizing based on [10]restrict qualified 10765 pointers has been rewritten and should result in code generation 10766 improvements. Optimizations based on restrict qualified pointers 10767 are now also available when using -fno-strict-aliasing. 10768 * There is a new optimization pass that attempts to change prototype 10769 of functions to avoid unused parameters, pass only relevant parts 10770 of structures and turn arguments passed by reference to arguments 10771 passed by value when possible. It is enabled by -O2 and above as 10772 well as -Os and can be manually invoked using the new command-line 10773 switch -fipa-sra. 10774 * GCC now optimize exception handling code. In particular cleanup 10775 regions that are proved to not have any effect are optimized out. 10776 10777New Languages and Language specific improvements 10778 10779 All languages 10780 10781 * The -fshow-column option is now on by default. This means error 10782 messages now have a column associated with them. 10783 10784 Ada 10785 10786 * Compilation of programs heavily using discriminated record types 10787 with variant parts has been sped up and generates more compact 10788 code. 10789 * Stack checking now works reasonably well on most plaforms. In some 10790 specific cases, stack overflows may still fail to be detected, but 10791 a compile-time warning will be issued for these cases. 10792 10793 C family 10794 10795 * If a header named in a #include directive is not found, the 10796 compiler exits immediately. This avoids a cascade of errors arising 10797 from declarations expected to be found in that header being 10798 missing. 10799 * A new built-in function __builtin_unreachable() has been added that 10800 tells the compiler that control will never reach that point. It may 10801 be used after asm statements that terminate by transferring control 10802 elsewhere, and in other places that are known to be unreachable. 10803 * The -Wlogical-op option now warns for logical expressions such as 10804 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 10805 mistakes. This option is disabled by default. 10806 * An asm goto feature has been added to allow asm statements that 10807 jump to C labels. 10808 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 10809 * The deprecated attribute now takes an optional string argument, for 10810 example, __attribute__((deprecated("text string"))), that will be 10811 printed together with the deprecation warning. 10812 10813 C 10814 10815 * The -Wenum-compare option, which warns when comparing values of 10816 different enum types, now works for C. It formerly only worked for 10817 C++. This warning is enabled by -Wall. It may be avoided by using a 10818 type cast. 10819 * The -Wcast-qual option now warns about casts which are unsafe in 10820 that they permit const-correctness to be violated without further 10821 warnings. Specifically, it warns about cases where a qualifier is 10822 added when all the lower types are not const. For example, it warns 10823 about a cast from char ** to const char **. 10824 * The -Wc++-compat option is significantly improved. It issues new 10825 warnings for: 10826 + Using C++ reserved operator names as identifiers. 10827 + Conversions to enum types without explicit casts. 10828 + Using va_arg with an enum type. 10829 + Using different enum types in the two branches of ?:. 10830 + Using ++ or -- on a variable of enum type. 10831 + Using the same name as both a struct, union or enum tag and a 10832 typedef, unless the typedef refers to the tagged type itself. 10833 + Using a struct, union, or enum which is defined within another 10834 struct or union. 10835 + A struct field defined using a typedef if there is a field in 10836 the struct, or an enclosing struct, whose name is the typedef 10837 name. 10838 + Duplicate definitions at file scope. 10839 + Uninitialized const variables. 10840 + A global variable with an anonymous struct, union, or enum 10841 type. 10842 + Using a string constant to initialize a char array whose size 10843 is the length of the string. 10844 * The new -Wjump-misses-init option warns about cases where a goto or 10845 switch skips the initialization of a variable. This sort of branch 10846 is an error in C++ but not in C. This warning is enabled by 10847 -Wc++-compat. 10848 * GCC now ensures that a C99-conforming <stdint.h> is present on most 10849 targets, and uses information about the types in this header to 10850 implement the Fortran bindings to those types. GCC does not ensure 10851 the presence of such a header, and does not implement the Fortran 10852 bindings, on the following targets: NetBSD, VxWorks, VMS, 10853 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 10854 * GCC now implements C90- and C99-conforming rules for constant 10855 expressions. This may cause warnings or errors for some code using 10856 expressions that can be folded to a constant but are not constant 10857 expressions as defined by ISO C. 10858 * All known target-independent C90 and C90 Amendment 1 conformance 10859 bugs, and all known target-independent C99 conformance bugs not 10860 related to floating point or extended identifiers, have been fixed. 10861 * The C decimal floating point support now includes support for the 10862 FLOAT_CONST_DECIMAL64 pragma. 10863 * The named address space feature from ISO/IEC TR 18037 is now 10864 supported. This is currently only implemented for the SPU 10865 processor. 10866 10867 C++ 10868 10869 * Improved [11]experimental support for the upcoming C++0x ISO C++ 10870 standard, including support for raw strings, lambda expressions and 10871 explicit type conversion operators. 10872 * When printing the name of a class template specialization, G++ will 10873 now omit any template arguments which come from default template 10874 arguments. This behavior (and the pretty-printing of function 10875 template specializations as template signature and arguments) can 10876 be disabled with the -fno-pretty-templates option. 10877 * Access control is now applied to typedef names used in a template, 10878 which may cause G++ to reject some ill-formed code that was 10879 accepted by earlier releases. The -fno-access-control option can be 10880 used as a temporary workaround until the code is corrected. 10881 * Compilation time for code that uses templates should now scale 10882 linearly with the number of instantiations rather than 10883 quadratically, as template instantiations are now looked up using 10884 hash tables. 10885 * Declarations of functions that look like builtin declarations of 10886 library functions are only considered to be redeclarations if they 10887 are declared with extern "C". This may cause problems with code 10888 that omits extern "C" on hand-written declarations of C library 10889 functions such as abort or memcpy. Such code is ill-formed, but was 10890 accepted by earlier releases. 10891 * Diagnostics that used to complain about passing non-POD types to 10892 ... or jumping past the declaration of a non-POD variable now check 10893 for triviality rather than PODness, as per C++0x. 10894 * In C++0x mode local and anonymous classes are now allowed as 10895 template arguments, and in declarations of variables and functions 10896 with linkage, so long as any such declaration that is used is also 10897 defined ([12]DR 757). 10898 * Labels may now have attributes, as has been permitted for a while 10899 in C. This is only permitted when the label definition and the 10900 attribute specifier is followed by a semicolon--i.e., the label 10901 applies to an empty statement. The only useful attribute for a 10902 label is unused. 10903 * G++ now implements [13]DR 176. Previously G++ did not support using 10904 the injected-class-name of a template base class as a type name, 10905 and lookup of the name found the declaration of the template in the 10906 enclosing scope. Now lookup of the name finds the 10907 injected-class-name, which can be used either as a type or as a 10908 template, depending on whether or not the name is followed by a 10909 template argument list. As a result of this change, some code that 10910 was previously accepted may be ill-formed because 10911 1. The injected-class-name is not accessible because it's from a 10912 private base, or 10913 2. The injected-class-name cannot be used as an argument for a 10914 template template parameter. 10915 In either of these cases, the code can be fixed by adding a 10916 nested-name-specifier to explicitly name the template. The first 10917 can be worked around with -fno-access-control; the second is only 10918 rejected with -pedantic. 10919 * A new standard mangling for SIMD vector types has been added, to 10920 avoid name clashes on systems with vectors of varying length. By 10921 default the compiler still uses the old mangling, but emits aliases 10922 with the new mangling on targets that support strong aliases. Users 10923 can switch over entirely to the new mangling with -fabi-version=4 10924 or -fabi-version=0. -Wabi will now warn about code that uses the 10925 old mangling. 10926 * The command-line option -ftemplate-depth-N is now written as 10927 -ftemplate-depth=N and the old form is deprecated. 10928 * Conversions between NULL and non-pointer types are now warned by 10929 default. The new option -Wno-conversion-null disables these 10930 warnings. Previously these warnings were only available when using 10931 -Wconversion explicitly. 10932 10933 Runtime Library (libstdc++) 10934 10935 * Improved experimental support for the upcoming ISO C++ standard, 10936 C++0x, including: 10937 + Support for <future>, <functional>, and <random>. 10938 + Existing facilities now exploit explicit operators and the 10939 newly implemented core C++0x features. 10940 + The header <cstdatomic> has been renamed to <atomic>. 10941 * An experimental [14]profile mode has been added. This is an 10942 implementation of many C++ standard library constructs with an 10943 additional analysis layer that gives performance improvement advice 10944 based on recognition of suboptimal usage patterns. For example, 10945#include <vector> 10946int main() 10947{ 10948 std::vector<int> v; 10949 for (int k = 0; k < 1024; ++k) 10950 v.insert(v.begin(), k); 10951} 10952 10953 When instrumented via the profile mode, can return suggestions 10954 about the initial size and choice of the container used as follows: 10955vector-to-list: improvement = 5: call stack = 0x804842c ... 10956 : advice = change std::vector to std::list 10957vector-size: improvement = 3: call stack = 0x804842c ... 10958 : advice = change initial container size from 0 to 1024 10959 10960 These constructs can be substituted for the normal libstdc++ 10961 constructs on a piecemeal basis, or all existing components can be 10962 transformed via the -D_GLIBCXX_PROFILE macro. 10963 * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR 10964 24733) has been added. This support is in header file 10965 <decimal/decimal>, uses namespace std::decimal, and includes 10966 classes decimal32, decimal64, and decimal128. 10967 * Sources have been audited for application of function attributes 10968 nothrow, const, pure, and noreturn. 10969 * Python pretty-printers have been added for many standard library 10970 components that simplify the internal representation and present a 10971 more intuitive view of components when used with 10972 appropriately-advanced versions of GDB. For more information, 10973 please consult the more [16]detailed description. 10974 * The default behavior for comparing typeinfo names has changed, so 10975 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 10976 * The new -static-libstdc++ option directs g++ to link the C++ 10977 library statically, even if the default would normally be to link 10978 it dynamically. 10979 10980 Fortran 10981 10982 * The COMMON default padding has been changed - instead of adding the 10983 padding before a variable it is now added afterwards, which 10984 increases the compatibility with other vendors and helps to obtain 10985 the correct output in some cases. Cf. also the -falign-commons 10986 option ([17]added in 4.4). 10987 * The -finit-real= option now also supports the value snan for 10988 signaling not-a-number; to be effective, one additionally needs to 10989 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 10990 optimizations can turn a signaling NaN into a quiet one. 10991 * The new option -fcheck= has been added with the options bounds, 10992 array-temps, do, pointer, and recursive. The bounds and array-temps 10993 options are equivalent to -fbounds-check and 10994 -fcheck-array-temporaries. The do option checks for invalid 10995 modification of loop iteration variables, and the recursive option 10996 tests for recursive calls to subroutines/functions which are not 10997 marked as recursive. With pointer pointer association checks in 10998 calls are performed; however, neither undefined pointers nor 10999 pointers in expressions are handled. Using -fcheck=all enables all 11000 these run-time checks. 11001 * The run-time checking -fcheck=bounds now warns about invalid string 11002 lengths of character dummy arguments. Additionally, more 11003 compile-time checks have been added. 11004 * The new option [18]-fno-protect-parens has been added; if set, the 11005 compiler may reorder REAL and COMPLEX expressions without regard to 11006 parentheses. 11007 * GNU Fortran no longer links against libgfortranbegin. As before, 11008 MAIN__ (assembler symbol name) is the actual Fortran main program, 11009 which is invoked by the main function. However, main is now 11010 generated and put in the same object file as MAIN__. For the time 11011 being, libgfortranbegin still exists for backward compatibility. 11012 For details see the new [19]Mixed-Language Programming chapter in 11013 the manual. 11014 * The I/O library was restructured for performance and cleaner code. 11015 * Array assignments and WHERE are now run in parallel when OpenMP's 11016 WORKSHARE is used. 11017 * The experimental option -fwhole-file was added. The option allows 11018 whole-file checking of procedure arguments and allows for better 11019 optimizations. It can also be used with -fwhole-program, which is 11020 now also supported in gfortran. 11021 * More Fortran 2003 and Fortran 2008 mathematical functions can now 11022 be used as initialization expressions. 11023 * Some extended attributes such as STDCALL are now supported via the 11024 [20]GCC$ compiler directive. 11025 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 11026 intrinsic behaves now as if zero were always positive. 11027 * For legacy compatibiliy: On Cygwin and MinGW, the special files 11028 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 11029 supported. 11030 * Fortran 2003 support has been extended: 11031 + Procedure-pointer function results and procedure-pointer 11032 components (including PASS), 11033 + allocatable scalars (experimental), 11034 + DEFERRED type-bound procedures, 11035 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 11036 have been implemented. 11037 + The ALLOCATE statement supports type-specs and the SOURCE= 11038 argument. 11039 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 11040 type-bound procedure (i.e. as type-bound operators). 11041 + Rounding (ROUND=, RZ, ...) for output is now supported. 11042 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 11043 intrinsic module ISO_C_BINDING are now supported, except for 11044 the targets listed above as ones where GCC does not have 11045 <stdint.h> type information. 11046 + Extensible derived types with type-bound procedure or 11047 procedure pointer with PASS attribute now have to use CLASS in 11048 line with the Fortran 2003 standard; the workaround to use 11049 TYPE is no longer supported. 11050 + [21]Experimental, incomplete support for polymorphism, 11051 including CLASS, SELECT TYPE and dynamic dispatch of 11052 type-bound procedure calls. Some features do not work yet such 11053 as unlimited polymorphism (CLASS(*)). 11054 * Fortran 2008 support has been extended: 11055 + The OPEN statement now supports the NEWUNIT= option, which 11056 returns a unique file unit, thus preventing inadvertent use of 11057 the same unit in different parts of the program. 11058 + Support for unlimited format items has been added. 11059 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 11060 the intrinsic module ISO_FORTRAN_ENV are now supported. 11061 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 11062 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 11063 and ATANH have been added (for real and complex arguments) and 11064 ATAN(Y,X) is now an alias for ATAN2(Y,X). 11065 + The BLOCK construct has been implemented. 11066 11067New Targets and Target Specific Improvements 11068 11069 AIX 11070 11071 * Full cross-toolchain support now available with GNU Binutils 11072 11073 ARM 11074 11075 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 11076 * GCC now supports the ARM v7E-M architecture. 11077 * GCC now supports VFPv4-based FPUs and FPUs with 11078 single-precision-only VFP. 11079 * GCC has many improvements to optimization for other ARM processors, 11080 including scheduling support for the integer pipeline on Cortex-A9. 11081 * GCC now supports the IEEE 754-2008 half-precision floating-point 11082 type, and a variant ARM-specific half-precision type. This type is 11083 specified using __fp16, with the layout determined by 11084 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 11085 VFPv4 half-precision instructions will be used. 11086 * GCC now supports the variant of AAPCS that uses VFP registers for 11087 parameter passing and return values. 11088 11089 AVR 11090 11091 * The -mno-tablejump option has been removed because it has the same 11092 effect as the -fno-jump-tables option. 11093 * Added support for these new AVR devices: 11094 + ATmega8U2 11095 + ATmega16U2 11096 + ATmega32U2 11097 11098 IA-32/x86-64 11099 11100 * GCC now will set the default for -march= based on the configure 11101 target. 11102 * GCC now supports handling floating-point excess precision arising 11103 from use of the x87 floating-point unit in a way that conforms to 11104 ISO C99. This is enabled with -fexcess-precision=standard and with 11105 standards conformance options such as -std=c99, and may be disabled 11106 using -fexcess-precision=fast. 11107 * Support for the Intel Atom processor is now available through the 11108 -march=atom and -mtune=atom options. 11109 * A new -mcrc32 option is now available to enable crc32 intrinsics. 11110 * A new -mmovbe option is now available to enable GCC to use the 11111 movbe instruction to implement __builtin_bswap32 and 11112 __builtin_bswap64. 11113 * SSE math now can be enabled by default at configure time with the 11114 new --with-fpmath=sse option. 11115 * There is a new intrinsic header file, <x86intrin.h>. It should be 11116 included before using any IA-32/x86-64 intrinsics. 11117 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 11118 Orochi processors are now available with the -mxop, -mfma4, and 11119 -mlwp options. 11120 * The -mabm option enables GCC to use the popcnt and lzcnt 11121 instructions on AMD processors. 11122 * The -mpopcnt option enables GCC to use the popcnt instructions on 11123 both AMD and Intel processors. 11124 11125 M68K/ColdFire 11126 11127 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 11128 and 5441x devices. 11129 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 11130 processors. 11131 11132 MeP 11133 11134 Support has been added for the Toshiba Media embedded Processor (MeP, 11135 or mep-elf) embedded target. 11136 11137 MIPS 11138 11139 * GCC now supports MIPS 1004K processors. 11140 * GCC can now be configured with options --with-arch-32, 11141 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 11142 default optimization separately for 32-bit and 64-bit modes. 11143 * MIPS targets now support an alternative _mcount interface, in which 11144 register $12 points to the function's save slot for register $31. 11145 This interface is selected by the -mcount-ra-address option; see 11146 the documentation for more details. 11147 * GNU/Linux targets can now generate read-only .eh_frame sections. 11148 This optimization requires GNU binutils 2.20 or above, and is only 11149 available if GCC is configured with a suitable version of binutils. 11150 * GNU/Linux targets can now attach special relocations to indirect 11151 calls, so that the linker can turn them into direct jumps or 11152 branches. This optimization requires GNU binutils 2.20 or later, 11153 and is automatically selected if GCC is configured with an 11154 appropriate version of binutils. It can be explicitly enabled or 11155 disabled using the -mrelax-pic-calls command-line option. 11156 * GCC now generates more heavily-optimized atomic operations on 11157 Octeon processors. 11158 * MIPS targets now support the -fstack-protector option. 11159 * GCC now supports an -msynci option, which specifies that synci is 11160 enough to flush the instruction cache, without help from the 11161 operating system. GCC uses this information to optimize 11162 automatically-generated cache flush operations, such as those used 11163 for nested functions in C. There is also a --with-synci 11164 configure-time option, which makes -msynci the default. 11165 * GCC supports four new function attributes for interrupt handlers: 11166 interrupt, use_shadow_register_set, keep_interrupts_masked and 11167 use_debug_exception_return. See the documentation for more details 11168 about these attributes. 11169 11170 RS/6000 (POWER/PowerPC) 11171 11172 * GCC now supports the Power ISA 2.06, which includes the VSX 11173 instructions that add vector 64-bit floating point support, new 11174 population count instructions, and conversions between floating 11175 point and unsigned types. 11176 * Support for the power7 processor is now available through the 11177 -mcpu=power7 and -mtune=power7. 11178 * GCC will now vectorize loops that contain simple math functions 11179 like copysign when generating code for altivec or VSX targets. 11180 * Support for the A2 processor is now available through the -mcpu=a2 11181 and -mtune=a2 options. 11182 * Support for the 476 processor is now available through the 11183 -mcpu={476,476fp} and -mtune={476,476fp} options. 11184 * Support for the e500mc64 processor is now available through the 11185 -mcpu=e500mc64 and -mtune=e500mc64 options. 11186 * GCC can now be configured with options --with-cpu-32, 11187 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 11188 default optimization separately for 32-bit and 64-bit modes. 11189 * Starting with GCC 4.5.4, vectors of type vector long long or vector 11190 long are passed and returned in the same method as other vectors 11191 with the VSX instruction set. Previously the GCC compiler did not 11192 adhere to the ABI for 128-bit vectors with 64-bit integer base 11193 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 11194 11195 RX 11196 11197 Support has been added for the Renesas RX Processor (rx-elf) target. 11198 11199Operating Systems 11200 11201 Windows (Cygwin and MinGW) 11202 11203 * GCC now installs all the major language runtime libraries as DLLs 11204 when configured with the --enable-shared option. 11205 * GCC now makes use of the new support for aligned common variables 11206 in versions of binutils >= 2.20 to fix bugs in the support for SSE 11207 data types. 11208 * Improvements to the libffi support library increase the reliability 11209 of code generated by GCJ on all Windows platforms. Libgcj is 11210 enabled by default for the first time. 11211 * Libtool improvements simplify installation by placing the generated 11212 DLLs in the correct binaries directory. 11213 * Numerous other minor bugfixes and improvements, and substantial 11214 enhancements to the Fortran language support library. 11215 11216 > 11217 11218Other significant improvements 11219 11220 Plugins 11221 11222 * It is now possible to extend the compiler without having to modify 11223 its source code. A new option -fplugin=file.so tells GCC to load 11224 the shared object file.so and execute it as part of the compiler. 11225 The internal documentation describes the details on how plugins can 11226 interact with the compiler. 11227 11228 Installation changes 11229 11230 * The move to newer autotools changed default installation 11231 directories and switches to control them: The --with-datarootdir, 11232 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 11233 used any more. Instead, you can now use --datarootdir, --docdir, 11234 --htmldir, and --pdfdir. The default installation directories have 11235 changed as follows according to the GNU Coding Standards: 11236 11237 datarootdir read-only architecture-independent data root [PREFIX/share] 11238 localedir locale-specific message catalogs [DATAROOTDIR/locale] 11239 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 11240 htmldir html documentation [DOCDIR] 11241 dvidir dvi documentation [DOCDIR] 11242 pdfdir pdf documentation [DOCDIR] 11243 psdir ps documentation [DOCDIR] 11244 The following variables have new default values: 11245 11246 datadir read-only architecture-independent data [DATAROOTDIR] 11247 infodir info documentation [DATAROOTDIR/info] 11248 mandir man documentation [DATAROOTDIR/man] 11249 11250GCC 4.5.1 11251 11252 This is the [22]list of problem reports (PRs) from GCC's bug tracking 11253 system that are known to be fixed in the 4.5.1 release. This list might 11254 not be complete (that is, it is possible that some PRs that have been 11255 fixed are not listed here). 11256 11257 All languages 11258 11259 * GCC's new link-time optimizer ([23]-flto) now also works on a few 11260 non-ELF targets: 11261 + Cygwin (*-cygwin*) 11262 + MinGW (*-mingw*) 11263 + Darwin on x86-64 (x86_64-apple-darwin*) 11264 LTO is not enabled by default for these targets. To enable LTO, you 11265 should configure with the --enable-lto option. 11266 11267GCC 4.5.2 11268 11269 This is the [24]list of problem reports (PRs) from GCC's bug tracking 11270 system that are known to be fixed in the 4.5.2 release. This list might 11271 not be complete (that is, it is possible that some PRs that have been 11272 fixed are not listed here). 11273 11274GCC 4.5.3 11275 11276 This is the [25]list of problem reports (PRs) from GCC's bug tracking 11277 system that are known to be fixed in the 4.5.3 release. This list might 11278 not be complete (that is, it is possible that some PRs that have been 11279 fixed are not listed here). 11280 11281 On the PowerPC compiler, the Altivec builtin functions vec_ld and 11282 vec_st have been modified to generate the Altivec memory instructions 11283 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 11284 release, these builtin functions were changed to generate VSX memory 11285 reference instructions instead of Altivec memory instructions, but 11286 there are differences between the two instructions. If the VSX 11287 instruction set is available, you can now use the new builtin functions 11288 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 11289 instructions. 11290 11291GCC 4.5.4 11292 11293 This is the [26]list of problem reports (PRs) from GCC's bug tracking 11294 system that are known to be fixed in the 4.5.4 release. This list might 11295 not be complete (that is, it is possible that some PRs that have been 11296 fixed are not listed here). 11297 11298 11299 For questions related to the use of GCC, please consult these web 11300 pages and the [27]GCC manuals. If that fails, the 11301 [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these 11302 web pages and the development of GCC are welcome on our developer 11303 list at [29]gcc@gcc.gnu.org. All of [30]our lists have public 11304 archives. 11305 11306 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 11307 distribution of this entire article is permitted in any medium, 11308 provided this notice is preserved. 11309 11310 These pages are [32]maintained by the GCC team. Last modified 11311 2021-07-28[33]. 11312 11313References 11314 11315 1. http://www.multiprecision.org/mpc/ 11316 2. https://gcc.gnu.org/install/prerequisites.html 11317 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 11318 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 11319 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86 11320 6. https://gcc.gnu.org/PR30789 11321 7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 11322 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 11323 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 11324 10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 11325 11. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html 11326 12. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 11327 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 11328 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 11329 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 11330 16. https://sourceware.org/gdb/wiki/STLSupport 11331 17. http://gcc.gnu.org/gcc-4.4/changes.html 11332 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 11333 19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 11334 20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 11335 21. https://gcc.gnu.org/wiki/OOP 11336 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 11337 23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 11338 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 11339 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 11340 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 11341 27. https://gcc.gnu.org/onlinedocs/ 11342 28. mailto:gcc-help@gcc.gnu.org 11343 29. mailto:gcc@gcc.gnu.org 11344 30. https://gcc.gnu.org/lists.html 11345 31. https://www.fsf.org/ 11346 32. https://gcc.gnu.org/about.html 11347 33. http://validator.w3.org/check/referer 11348====================================================================== 11349http://gcc.gnu.org/gcc-4.4/index.html 11350 11351 GCC 4.4 Release Series 11352 11353 This release series is no longer maintained. 11354 11355 March 13, 2012 11356 11357 The [1]GNU project and the GCC developers are pleased to announce the 11358 release of GCC 4.4.7. 11359 11360 This release is a bug-fix release, containing fixes for regressions in 11361 GCC 4.4.6 relative to previous releases of GCC. 11362 11363Release History 11364 11365 GCC 4.4.7 11366 March 13, 2012 ([2]changes) 11367 11368 GCC 4.4.6 11369 April 16, 2011 ([3]changes) 11370 11371 GCC 4.4.5 11372 October 1, 2010 ([4]changes) 11373 11374 GCC 4.4.4 11375 April 29, 2010 ([5]changes) 11376 11377 GCC 4.4.3 11378 January 21, 2010 ([6]changes) 11379 11380 GCC 4.4.2 11381 October 15, 2009 ([7]changes) 11382 11383 GCC 4.4.1 11384 July 22, 2009 ([8]changes) 11385 11386 GCC 4.4.0 11387 April 21, 2009 ([9]changes) 11388 11389References and Acknowledgements 11390 11391 GCC used to stand for the GNU C Compiler, but since the compiler 11392 supports several other languages aside from C, it now stands for the 11393 GNU Compiler Collection. 11394 11395 A list of [10]successful builds is updated as new information becomes 11396 available. 11397 11398 The GCC developers would like to thank the numerous people that have 11399 contributed new features, improvements, bug fixes, and other changes as 11400 well as test results to GCC. This [11]amazing group of volunteers is 11401 what makes GCC successful. 11402 11403 For additional information about GCC please refer to the [12]GCC 11404 project web site or contact the [13]GCC development mailing list. 11405 11406 To obtain GCC please use [14]our mirror sites or [15]our version 11407 control system. 11408 11409 11410 For questions related to the use of GCC, please consult these web 11411 pages and the [16]GCC manuals. If that fails, the 11412 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these 11413 web pages and the development of GCC are welcome on our developer 11414 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public 11415 archives. 11416 11417 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 11418 distribution of this entire article is permitted in any medium, 11419 provided this notice is preserved. 11420 11421 These pages are [21]maintained by the GCC team. Last modified 11422 2021-07-28[22]. 11423 11424References 11425 11426 1. http://www.gnu.org/ 11427 2. http://gcc.gnu.org/gcc-4.4/changes.html 11428 3. http://gcc.gnu.org/gcc-4.4/changes.html 11429 4. http://gcc.gnu.org/gcc-4.4/changes.html 11430 5. http://gcc.gnu.org/gcc-4.4/changes.html 11431 6. http://gcc.gnu.org/gcc-4.4/changes.html 11432 7. http://gcc.gnu.org/gcc-4.4/changes.html 11433 8. http://gcc.gnu.org/gcc-4.4/changes.html 11434 9. http://gcc.gnu.org/gcc-4.4/changes.html 11435 10. http://gcc.gnu.org/gcc-4.4/buildstat.html 11436 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11437 12. http://gcc.gnu.org/index.html 11438 13. mailto:gcc@gcc.gnu.org 11439 14. http://gcc.gnu.org/mirrors.html 11440 15. http://gcc.gnu.org/git.html 11441 16. https://gcc.gnu.org/onlinedocs/ 11442 17. mailto:gcc-help@gcc.gnu.org 11443 18. mailto:gcc@gcc.gnu.org 11444 19. https://gcc.gnu.org/lists.html 11445 20. https://www.fsf.org/ 11446 21. https://gcc.gnu.org/about.html 11447 22. http://validator.w3.org/check/referer 11448====================================================================== 11449http://gcc.gnu.org/gcc-4.4/changes.html 11450 11451 GCC 4.4 Release Series 11452 Changes, New Features, and Fixes 11453 11454 The latest release in the 4.4 release series is [1]GCC 4.4.7. 11455 11456Caveats 11457 11458 * __builtin_stdarg_start has been completely removed from GCC. 11459 Support for <varargs.h> had been deprecated since GCC 4.0. Use 11460 __builtin_va_start as a replacement. 11461 * Some of the errors issued by the C++ front end that could be 11462 downgraded to warnings in previous releases by using -fpermissive 11463 are now warnings by default. They can be converted into errors by 11464 using -pedantic-errors. 11465 * Use of the cpp assertion extension will now emit a warning when 11466 -Wdeprecated or -pedantic is used. This extension has been 11467 deprecated for many years, but never warned about. 11468 * Packed bit-fields of type char were not properly bit-packed on many 11469 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 11470 causes an ABI change. For example there is no longer a 4-bit 11471 padding between field a and b in this structure: 11472 struct foo 11473 { 11474 char a:4; 11475 char b:8; 11476 } __attribute__ ((packed)); 11477 There is a new warning to help identify fields that are affected: 11478 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 11479 The warning can be disabled with -Wno-packed-bitfield-compat. 11480 * On ARM EABI targets, the C++ mangling of the va_list type has been 11481 changed to conform to the current revision of the EABI. This does 11482 not affect the libstdc++ library included with GCC. 11483 * The SCOUNT and POS bits of the MIPS DSP control register are now 11484 treated as global. Previous versions of GCC treated these fields as 11485 call-clobbered instead. 11486 * The MIPS port no longer recognizes the h asm constraint. It was 11487 necessary to remove this constraint in order to avoid generating 11488 unpredictable code sequences. 11489 One of the main uses of the h constraint was to extract the high 11490 part of a multiplication on 64-bit targets. For example: 11491 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 11492 You can now achieve the same effect using 128-bit types: 11493 typedef unsigned int uint128_t __attribute__((mode(TI))); 11494 result = ((uint128_t) x * y) >> 64; 11495 The second sequence is better in many ways. For example, if x and y 11496 are constants, the compiler can perform the multiplication at 11497 compile time. If x and y are not constants, the compiler can 11498 schedule the runtime multiplication better than it can schedule an 11499 asm statement. 11500 * Support for a number of older systems and recently unmaintained or 11501 untested target ports of GCC has been declared obsolete in GCC 4.4. 11502 Unless there is activity to revive them, the next release of GCC 11503 will have their sources permanently removed. 11504 The following ports for individual systems on particular 11505 architectures have been obsoleted: 11506 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 11507 m68k-*-aout*) 11508 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 11509 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 11510 sh-*-*). This does not affect other more specific targets 11511 using the COFF object format on those architectures, or the 11512 more specific H8300 and SH targets (h8300-*-rtems*, 11513 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 11514 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 11515 + 2BSD on PDP-11 (pdp11-*-bsd) 11516 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 11517 powerpc-ibm-aix4.[12]*) 11518 + Tuning support for Itanium1 (Merced) variants. Note that code 11519 tuned for Itanium2 should also run correctly on Itanium1. 11520 * The protoize and unprotoize utilities have been obsoleted and will 11521 be removed in GCC 4.5. These utilities have not been installed by 11522 default since GCC 3.0. 11523 * Support has been removed for all the [2]configurations obsoleted in 11524 GCC 4.3. 11525 * Unknown -Wno-* options are now silently ignored by GCC if no other 11526 diagnostics are issued. If other diagnostics are issued, then GCC 11527 warns about the unknown options. 11528 * More information on porting to GCC 4.4 from previous versions of 11529 GCC can be found in the [3]porting guide for this release. 11530 11531General Optimizer Improvements 11532 11533 * A new command-line switch -findirect-inlining has been added. When 11534 turned on it allows the inliner to also inline indirect calls that 11535 are discovered to have known targets at compile time thanks to 11536 previous inlining. 11537 * A new command-line switch -ftree-switch-conversion has been added. 11538 This new pass turns simple initializations of scalar variables in 11539 switch statements into initializations from a static array, given 11540 that all the values are known at compile time and the ratio between 11541 the new array size and the original switch branches does not exceed 11542 the parameter --param switch-conversion-max-branch-ratio (default 11543 is eight). 11544 * A new command-line switch -ftree-builtin-call-dce has been added. 11545 This optimization eliminates unnecessary calls to certain builtin 11546 functions when the return value is not used, in cases where the 11547 calls can not be eliminated entirely because the function may set 11548 errno. This optimization is on by default at -O2 and above. 11549 * A new command-line switch -fconserve-stack directs the compiler to 11550 minimize stack usage even if it makes the generated code slower. 11551 This affects inlining decisions. 11552 * When the assembler supports it, the compiler will now emit unwind 11553 information using assembler .cfi directives. This makes it possible 11554 to use such directives in inline assembler code. The new option 11555 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 11556 directives. 11557 * The [4]Graphite branch has been merged. This merge has brought in a 11558 new framework for loop optimizations based on a polyhedral 11559 intermediate representation. These optimizations apply to all the 11560 languages supported by GCC. The following new code transformations 11561 are available in GCC 4.4: 11562 + -floop-interchange performs loop interchange transformations 11563 on loops. Interchanging two nested loops switches the inner 11564 and outer loops. For example, given a loop like: 11565 DO J = 1, M 11566 DO I = 1, N 11567 A(J, I) = A(J, I) * C 11568 ENDDO 11569 ENDDO 11570 11571 loop interchange will transform the loop as if the user had 11572 written: 11573 DO I = 1, N 11574 DO J = 1, M 11575 A(J, I) = A(J, I) * C 11576 ENDDO 11577 ENDDO 11578 11579 which can be beneficial when N is larger than the caches, 11580 because in Fortran, the elements of an array are stored in 11581 memory contiguously by column, and the original loop iterates 11582 over rows, potentially creating at each access a cache miss. 11583 + -floop-strip-mine performs loop strip mining transformations 11584 on loops. Strip mining splits a loop into two nested loops. 11585 The outer loop has strides equal to the strip size and the 11586 inner loop has strides of the original loop within a strip. 11587 For example, given a loop like: 11588 DO I = 1, N 11589 A(I) = A(I) + C 11590 ENDDO 11591 11592 loop strip mining will transform the loop as if the user had 11593 written: 11594 DO II = 1, N, 4 11595 DO I = II, min (II + 3, N) 11596 A(I) = A(I) + C 11597 ENDDO 11598 ENDDO 11599 11600 + -floop-block performs loop blocking transformations on loops. 11601 Blocking strip mines each loop in the loop nest such that the 11602 memory accesses of the element loops fit inside caches. For 11603 example, given a loop like: 11604 DO I = 1, N 11605 DO J = 1, M 11606 A(J, I) = B(I) + C(J) 11607 ENDDO 11608 ENDDO 11609 11610 loop blocking will transform the loop as if the user had 11611 written: 11612 DO II = 1, N, 64 11613 DO JJ = 1, M, 64 11614 DO I = II, min (II + 63, N) 11615 DO J = JJ, min (JJ + 63, M) 11616 A(J, I) = B(I) + C(J) 11617 ENDDO 11618 ENDDO 11619 ENDDO 11620 ENDDO 11621 11622 which can be beneficial when M is larger than the caches, 11623 because the innermost loop will iterate over a smaller amount 11624 of data that can be kept in the caches. 11625 * A new register allocator has replaced the old one. It is called 11626 integrated register allocator (IRA) because coalescing, register 11627 live range splitting, and hard register preferencing are done 11628 on-the-fly during coloring. It also has better integration with the 11629 reload pass. IRA is a regional register allocator which uses modern 11630 Chaitin-Briggs coloring instead of Chow's priority coloring used in 11631 the old register allocator. More info about IRA internals and 11632 options can be found in the GCC manuals. 11633 * A new instruction scheduler and software pipeliner, based on the 11634 selective scheduling approach, has been added. The new pass 11635 performs instruction unification, register renaming, substitution 11636 through register copies, and speculation during scheduling. The 11637 software pipeliner is able to pipeline non-countable loops. The new 11638 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 11639 it is available for the Intel Itanium platform working by default 11640 as the second scheduling pass (after register allocation) at the 11641 -O3 optimization level. 11642 * When using -fprofile-generate with a multi-threaded program, the 11643 profile counts may be slightly wrong due to race conditions. The 11644 new -fprofile-correction option directs the compiler to apply 11645 heuristics to smooth out the inconsistencies. By default the 11646 compiler will give an error message when it finds an inconsistent 11647 profile. 11648 * The new -fprofile-dir=PATH option permits setting the directory 11649 where profile data files are stored when using -fprofile-generate 11650 and friends, and the directory used when reading profile data files 11651 using -fprofile-use and friends. 11652 11653New warning options 11654 11655 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 11656 warning if any stack frame is larger than NUMBER bytes. This may be 11657 used to help ensure that code fits within a limited amount of stack 11658 space. 11659 * The command-line option -Wlarger-than-N is now written as 11660 -Wlarger-than=N and the old form is deprecated. 11661 * The new -Wno-mudflap option disables warnings about constructs 11662 which can not be instrumented when using -fmudflap. 11663 11664New Languages and Language specific improvements 11665 11666 * Version 3.0 of the OpenMP specification is now supported for the C, 11667 C++, and Fortran compilers. 11668 * New character data types, per [5]TR 19769: New character types in 11669 C, are now supported for the C compiler in -std=gnu99 mode, as 11670 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 11671 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 11672 11673 C family 11674 11675 * A new optimize attribute was added to allow programmers to change 11676 the optimization level and particular optimization options for an 11677 individual function. You can also change the optimization options 11678 via the GCC optimize pragma for functions defined after the pragma. 11679 The GCC push_options pragma and the GCC pop_options pragma allow 11680 you temporarily save and restore the options used. The GCC 11681 reset_options pragma restores the options to what was specified on 11682 the command line. 11683 * Uninitialized warnings do not require enabling optimization 11684 anymore, that is, -Wuninitialized can be used together with -O0. 11685 Nonetheless, the warnings given by -Wuninitialized will probably be 11686 more accurate if optimization is enabled. 11687 * -Wparentheses now warns about expressions such as (!x | y) and (!x 11688 & y). Using explicit parentheses, such as in ((!x) | y), silences 11689 this warning. 11690 * -Wsequence-point now warns within if, while,do while and for 11691 conditions, and within for begin/end expressions. 11692 * A new option -dU is available to dump definitions of preprocessor 11693 macros that are tested or expanded. 11694 11695 C++ 11696 11697 * [6]Improved experimental support for the upcoming ISO C++ standard, 11698 C++0x. Including support for auto, inline namespaces, generalized 11699 initializer lists, defaulted and deleted functions, new character 11700 types, and scoped enums. 11701 * Those errors that may be downgraded to warnings to build legacy 11702 code now mention -fpermissive when -fdiagnostics-show-option is 11703 enabled. 11704 * -Wconversion now warns if the result of a static_cast to enumeral 11705 type is unspecified because the value is outside the range of the 11706 enumeral type. 11707 * -Wuninitialized now warns if a non-static reference or non-static 11708 const member appears in a class without constructors. 11709 * G++ now properly implements value-initialization, so objects with 11710 an initializer of () and an implicitly defined default constructor 11711 will be zero-initialized before the default constructor is called. 11712 11713 Runtime Library (libstdc++) 11714 11715 * Improved experimental support for the upcoming ISO C++ standard, 11716 C++0x, including: 11717 + Support for <chrono>, <condition_variable>, <cstdatomic>, 11718 <forward_list>, <initializer_list>, <mutex>, <ratio>, 11719 <system_error>, and <thread>. 11720 + unique_ptr, <algorithm> additions, exception propagation, and 11721 support for the new character types in <string> and <limits>. 11722 + Existing facilities now exploit initializer lists, defaulted 11723 and deleted functions, and the newly implemented core C++0x 11724 features. 11725 + Some standard containers are more efficient together with 11726 stateful allocators, i.e., no allocator is constructed on the 11727 fly at element construction time. 11728 * Experimental support for non-standard pointer types in containers. 11729 * The long standing libstdc++/30928 has been fixed for targets 11730 running glibc 2.10 or later. 11731 * As usual, many small and larger bug fixes, in particular quite a 11732 few corner cases in <locale>. 11733 11734 Fortran 11735 11736 * GNU Fortran now employs libcpp directly instead of using cc1 as an 11737 external preprocessor. The [7]-cpp option was added to allow manual 11738 invocation of the preprocessor without relying on filename 11739 extensions. 11740 * The [8]-Warray-temporaries option warns about array temporaries 11741 generated by the compiler, as an aid to optimization. 11742 * The [9]-fcheck-array-temporaries option has been added, printing a 11743 notification at run time, when an array temporary had to be created 11744 for an function argument. Contrary to -Warray-temporaries the 11745 warning is only printed if the array is noncontiguous. 11746 * Improved generation of DWARF debugging symbols 11747 * If using an intrinsic not part of the selected standard (via -std= 11748 and -fall-intrinsics) gfortran will now treat it as if this 11749 procedure were declared EXTERNAL and try to link to a user-supplied 11750 procedure. -Wintrinsics-std will warn whenever this happens. The 11751 now-useless option -Wnonstd-intrinsic was removed. 11752 * The flag -falign-commons has been added to control the alignment of 11753 variables in COMMON blocks, which is enabled by default in line 11754 with previous GCC version. Using -fno-align-commons one can force 11755 commons to be contiguous in memory as required by the Fortran 11756 standard, however, this slows down the memory access. The option 11757 -Walign-commons, which is enabled by default, warns when padding 11758 bytes were added for alignment. The proper solution is to sort the 11759 common objects by decreasing storage size, which avoids the 11760 alignment problems. 11761 * Fortran 2003 support has been extended: 11762 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 11763 now supported (except internal reads from/writes to wide 11764 strings). [10]-fbackslash now supports also \unnnn and 11765 \Unnnnnnnn to enter Unicode characters. 11766 + Asynchronous I/O (implemented as synchronous I/O) and the 11767 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 11768 are now supported in I/O statements. 11769 + Support for Fortran 2003 structure constructors and for array 11770 constructor with typespec has been added. 11771 + Procedure Pointers (but not yet as component in derived types 11772 and as function results) are now supported. 11773 + Abstract types, type extension, and type-bound procedures 11774 (both PROCEDURE and GENERIC but not as operators). Note: As 11775 CLASS/polymorphyic types are not implemented, type-bound 11776 procedures with PASS accept as non-standard extension TYPE 11777 arguments. 11778 * Fortran 2008 support has been added: 11779 + The -std=f2008 option and support for the file extensions 11780 .f2008 and .F2008 has been added. 11781 + The g0 format descriptor is now supported. 11782 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 11783 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 11784 are now available (some of them existed as GNU extension 11785 before). Note: The hyperbolic functions are not yet supporting 11786 complex arguments and the three- argument version of BESSEL_*N 11787 is not available. 11788 + The bit intrinsics LEADZ and TRAILZ have been added. 11789 11790 Java (GCJ) 11791 11792 Ada 11793 11794 * The Ada runtime now supports multilibs on many platforms including 11795 x86_64, SPARC and PowerPC. Their build is enabled by default. 11796 11797New Targets and Target Specific Improvements 11798 11799 ARM 11800 11801 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 11802 Cortex-R4F processors and has many other improvements to 11803 optimization for ARM processors. 11804 * GCC now supports the VFPv3 variant with 16 double-precision 11805 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 11806 renamed to -mfpu=vfpv3. 11807 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 11808 erratum on Cortex-M3 processors. 11809 * GCC now supports the __sync_* atomic operations for ARM EABI 11810 GNU/Linux. 11811 * The section anchors optimization is now enabled by default when 11812 optimizing for ARM. 11813 * GCC now uses a new EABI-compatible profiling interface for EABI 11814 targets. This requires a function __gnu_mcount_nc, which is 11815 provided by GNU libc versions 2.8 and later. 11816 11817 AVR 11818 11819 * The -mno-tablejump option has been deprecated because it has the 11820 same effect as the -fno-jump-tables option. 11821 * Added support for these new AVR devices: 11822 + ATA6289 11823 + ATtiny13A 11824 + ATtiny87 11825 + ATtiny167 11826 + ATtiny327 11827 + ATmega8C1 11828 + ATmega16C1 11829 + ATmega32C1 11830 + ATmega8M1 11831 + ATmega16M1 11832 + ATmega32M1 11833 + ATmega32U4 11834 + ATmega16HVB 11835 + ATmega4HVD 11836 + ATmega8HVD 11837 + ATmega64C1 11838 + ATmega64M1 11839 + ATmega16U4 11840 + ATmega32U6 11841 + ATmega128RFA1 11842 + AT90PWM81 11843 + AT90SCR100 11844 + M3000F 11845 + M3000S 11846 + M3001B 11847 11848 IA-32/x86-64 11849 11850 * Support for Intel AES built-in functions and code generation is 11851 available via -maes. 11852 * Support for Intel PCLMUL built-in function and code generation is 11853 available via -mpclmul. 11854 * Support for Intel AVX built-in functions and code generation is 11855 available via -mavx. 11856 * Automatically align the stack for local variables with alignment 11857 requirement. 11858 * GCC can now utilize the SVML library for vectorizing calls to a set 11859 of C99 functions if -mveclibabi=svml is specified and you link to 11860 an SVML ABI compatible library. 11861 * On x86-64, the ABI has been changed in the following cases to 11862 conform to the x86-64 ABI: 11863 + Passing/returning structures with flexible array member: 11864 struct foo 11865 { 11866 int i; 11867 int flex[]; 11868 }; 11869 + Passing/returning structures with complex float member: 11870 struct foo 11871 { 11872 int i; 11873 __complex__ float f; 11874 }; 11875 + Passing/returning unions with long double member: 11876 union foo 11877 { 11878 int x; 11879 long double ld; 11880 }; 11881 Code built with previous versions of GCC that uses any of these is 11882 not compatible with code built with GCC 4.4.0 or later. 11883 * A new target attribute was added to allow programmers to change the 11884 target options like -msse2 or -march=k8 for an individual function. 11885 You can also change the target options via the GCC target pragma 11886 for functions defined after the pragma. 11887 * GCC can now be configured with options --with-arch-32, 11888 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 11889 --with-tune-64 to control the default optimization separately for 11890 32-bit and 64-bit modes. 11891 11892 IA-32/IA64 11893 11894 * Support for __float128 (TFmode) IEEE quad type and corresponding 11895 TCmode IEEE complex quad type is available via the soft-fp library 11896 on IA-32/IA64 targets. This includes basic arithmetic operations 11897 (addition, subtraction, negation, multiplication and division) on 11898 __float128 real and TCmode complex values, the full set of IEEE 11899 comparisons between __float128 values, conversions to and from 11900 float, double and long double floating point types, as well as 11901 conversions to and from signed or unsigned integer, signed or 11902 unsigned long integer and signed or unsigned quad (TImode, IA64 11903 only) integer types. Additionally, all operations generate the full 11904 set of IEEE exceptions and support the full set of IEEE rounding 11905 modes. 11906 11907 M68K/ColdFire 11908 11909 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 11910 processors. (Scheduling support for ColdFire V2 processors was 11911 added in GCC 4.3.) 11912 * GCC now supports the -mxgot option to support programs requiring 11913 many GOT entries on ColdFire. 11914 * The m68k-*-linux-gnu target now builds multilibs by default. 11915 11916 MIPS 11917 11918 * MIPS Technologies have extended the original MIPS SVR4 ABI to 11919 include support for procedure linkage tables (PLTs) and copy 11920 relocations. These extensions allow GNU/Linux executables to use a 11921 significantly more efficient code model than the one defined by the 11922 original ABI. 11923 GCC support for this code model is available via a new command-line 11924 option, -mplt. There is also a new configure-time option, 11925 --with-mips-plt, to make -mplt the default. 11926 The new code model requires support from the assembler, the linker, 11927 and the runtime C library. This support is available in binutils 11928 2.19 and GLIBC 2.9. 11929 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 11930 and 32-bit GNU/Linux shared libraries. This feature requires GNU 11931 binutils 2.19 or above. 11932 * Support for RMI's XLR processor is now available through the 11933 -march=xlr and -mtune=xlr options. 11934 * 64-bit targets can now perform 128-bit multiplications inline, 11935 instead of relying on a libgcc function. 11936 * Native GNU/Linux toolchains now support -march=native and 11937 -mtune=native, which select the host processor. 11938 * GCC now supports the R10K, R12K, R14K and R16K processors. The 11939 canonical -march= and -mtune= names for these processors are 11940 r10000, r12000, r14000 and r16000 respectively. 11941 * GCC can now work around the side effects of speculative execution 11942 on R10K processors. Please see the documentation of the 11943 -mr10k-cache-barrier option for details. 11944 * Support for the MIPS64 Release 2 instruction set has been added. 11945 The option -march=mips64r2 enables generation of these 11946 instructions. 11947 * GCC now supports Cavium Networks' Octeon processor. This support is 11948 available through the -march=octeon and -mtune=octeon options. 11949 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 11950 canonical -march= and -mtune= names for these processors are 11951 loongson2e and loongson2f. 11952 11953 picochip 11954 11955 Picochip is a 16-bit processor. A typical picoChip contains over 250 11956 small cores, each with small amounts of memory. There are three 11957 processor variants (STAN, MEM and CTRL) with different instruction sets 11958 and memory configurations and they can be chosen using the -mae option. 11959 11960 This port is intended to be a "C" only port. 11961 11962 Power Architecture and PowerPC 11963 11964 * GCC now supports the e300c2, e300c3 and e500mc processors. 11965 * GCC now supports Xilinx processors with a single-precision FPU. 11966 * Decimal floating point is now supported for e500 processors. 11967 11968 S/390, zSeries and System z9/z10 11969 11970 * Support for the IBM System z10 EC/BC processor has been added. When 11971 using the -march=z10 option, the compiler will generate code making 11972 use of instructions provided by the General-Instruction-Extension 11973 Facility and the Execute-Extension Facility. 11974 11975 VxWorks 11976 11977 * GCC now supports the thread-local storage mechanism used on 11978 VxWorks. 11979 11980 Xtensa 11981 11982 * GCC now supports thread-local storage (TLS) for Xtensa processor 11983 configurations that include the Thread Pointer option. TLS also 11984 requires support from the assembler and linker; this support is 11985 provided in the GNU binutils beginning with version 2.19. 11986 11987Documentation improvements 11988 11989Other significant improvements 11990 11991GCC 4.4.1 11992 11993 This is the [11]list of problem reports (PRs) from GCC's bug tracking 11994 system that are known to be fixed in the 4.4.1 release. This list might 11995 not be complete (that is, it is possible that some PRs that have been 11996 fixed are not listed here). 11997 11998GCC 4.4.2 11999 12000 This is the [12]list of problem reports (PRs) from GCC's bug tracking 12001 system that are known to be fixed in the 4.4.2 release. This list might 12002 not be complete (that is, it is possible that some PRs that have been 12003 fixed are not listed here). 12004 12005GCC 4.4.3 12006 12007 This is the [13]list of problem reports (PRs) from GCC's bug tracking 12008 system that are known to be fixed in the 4.4.3 release. This list might 12009 not be complete (that is, it is possible that some PRs that have been 12010 fixed are not listed here). 12011 12012GCC 4.4.4 12013 12014 This is the [14]list of problem reports (PRs) from GCC's bug tracking 12015 system that are known to be fixed in the 4.4.4 release. This list might 12016 not be complete (that is, it is possible that some PRs that have been 12017 fixed are not listed here). 12018 12019GCC 4.4.5 12020 12021 This is the [15]list of problem reports (PRs) from GCC's bug tracking 12022 system that are known to be fixed in the 4.4.5 release. This list might 12023 not be complete (that is, it is possible that some PRs that have been 12024 fixed are not listed here). 12025 12026GCC 4.4.6 12027 12028 This is the [16]list of problem reports (PRs) from GCC's bug tracking 12029 system that are known to be fixed in the 4.4.6 release. This list might 12030 not be complete (that is, it is possible that some PRs that have been 12031 fixed are not listed here). 12032 12033GCC 4.4.7 12034 12035 This is the [17]list of problem reports (PRs) from GCC's bug tracking 12036 system that are known to be fixed in the 4.4.7 release. This list might 12037 not be complete (that is, it is possible that some PRs that have been 12038 fixed are not listed here). 12039 12040 12041 For questions related to the use of GCC, please consult these web 12042 pages and the [18]GCC manuals. If that fails, the 12043 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these 12044 web pages and the development of GCC are welcome on our developer 12045 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public 12046 archives. 12047 12048 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 12049 distribution of this entire article is permitted in any medium, 12050 provided this notice is preserved. 12051 12052 These pages are [23]maintained by the GCC team. Last modified 12053 2021-07-28[24]. 12054 12055References 12056 12057 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 12058 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 12059 3. http://gcc.gnu.org/gcc-4.4/porting_to.html 12060 4. https://gcc.gnu.org/wiki/Graphite 12061 5. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 12062 6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html 12063 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 12064 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 12065 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 12066 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 12067 11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 12068 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 12069 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 12070 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 12071 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 12072 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 12073 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 12074 18. https://gcc.gnu.org/onlinedocs/ 12075 19. mailto:gcc-help@gcc.gnu.org 12076 20. mailto:gcc@gcc.gnu.org 12077 21. https://gcc.gnu.org/lists.html 12078 22. https://www.fsf.org/ 12079 23. https://gcc.gnu.org/about.html 12080 24. http://validator.w3.org/check/referer 12081====================================================================== 12082http://gcc.gnu.org/gcc-4.3/index.html 12083 12084 GCC 4.3 Release Series 12085 12086 (This release series is no longer supported.) 12087 12088 Jun 27, 2011 12089 12090 The [1]GNU project and the GCC developers are pleased to announce the 12091 release of GCC 4.3.6. 12092 12093 This release is a bug-fix release, containing fixes for regressions in 12094 GCC 4.3.5 relative to previous releases of GCC. 12095 12096Release History 12097 12098 GCC 4.3.6 12099 Jun 27, 2011 ([2]changes) 12100 12101 GCC 4.3.5 12102 May 22, 2010 ([3]changes) 12103 12104 GCC 4.3.4 12105 August 4, 2009 ([4]changes) 12106 12107 GCC 4.3.3 12108 January 24, 2009 ([5]changes) 12109 12110 GCC 4.3.2 12111 August 27, 2008 ([6]changes) 12112 12113 GCC 4.3.1 12114 June 6, 2008 ([7]changes) 12115 12116 GCC 4.3.0 12117 March 5, 2008 ([8]changes) 12118 12119References and Acknowledgements 12120 12121 GCC used to stand for the GNU C Compiler, but since the compiler 12122 supports several other languages aside from C, it now stands for the 12123 GNU Compiler Collection. 12124 12125 A list of [9]successful builds is updated as new information becomes 12126 available. 12127 12128 The GCC developers would like to thank the numerous people that have 12129 contributed new features, improvements, bug fixes, and other changes as 12130 well as test results to GCC. This [10]amazing group of volunteers is 12131 what makes GCC successful. 12132 12133 For additional information about GCC please refer to the [11]GCC 12134 project web site or contact the [12]GCC development mailing list. 12135 12136 To obtain GCC please use [13]our mirror sites or [14]our version 12137 control system. 12138 12139 12140 For questions related to the use of GCC, please consult these web 12141 pages and the [15]GCC manuals. If that fails, the 12142 [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these 12143 web pages and the development of GCC are welcome on our developer 12144 list at [17]gcc@gcc.gnu.org. All of [18]our lists have public 12145 archives. 12146 12147 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 12148 distribution of this entire article is permitted in any medium, 12149 provided this notice is preserved. 12150 12151 These pages are [20]maintained by the GCC team. Last modified 12152 2021-07-28[21]. 12153 12154References 12155 12156 1. http://www.gnu.org/ 12157 2. http://gcc.gnu.org/gcc-4.3/changes.html 12158 3. http://gcc.gnu.org/gcc-4.3/changes.html 12159 4. http://gcc.gnu.org/gcc-4.3/changes.html 12160 5. http://gcc.gnu.org/gcc-4.3/changes.html 12161 6. http://gcc.gnu.org/gcc-4.3/changes.html 12162 7. http://gcc.gnu.org/gcc-4.3/changes.html 12163 8. http://gcc.gnu.org/gcc-4.3/changes.html 12164 9. http://gcc.gnu.org/gcc-4.3/buildstat.html 12165 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 12166 11. http://gcc.gnu.org/index.html 12167 12. mailto:gcc@gcc.gnu.org 12168 13. http://gcc.gnu.org/mirrors.html 12169 14. http://gcc.gnu.org/git.html 12170 15. https://gcc.gnu.org/onlinedocs/ 12171 16. mailto:gcc-help@gcc.gnu.org 12172 17. mailto:gcc@gcc.gnu.org 12173 18. https://gcc.gnu.org/lists.html 12174 19. https://www.fsf.org/ 12175 20. https://gcc.gnu.org/about.html 12176 21. http://validator.w3.org/check/referer 12177====================================================================== 12178http://gcc.gnu.org/gcc-4.3/changes.html 12179 12180 GCC 4.3 Release Series 12181 Changes, New Features, and Fixes 12182 12183 The latest release in the 4.3 release series is [1]GCC 4.3.5. 12184 12185Caveats 12186 12187 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 12188 various front-end languages it supports. See the [4]prerequisites 12189 page for version requirements. 12190 * ColdFire targets now treat long double as having the same format as 12191 double. In earlier versions of GCC, they used the 68881 long double 12192 format instead. 12193 * The m68k-uclinux target now uses the same calling conventions as 12194 m68k-linux-gnu. You can select the original calling conventions by 12195 configuring for m68k-uclinuxoldabi instead. Note that 12196 m68k-uclinuxoldabi also retains the original 80-bit long double on 12197 ColdFire targets. 12198 * The -fforce-mem option has been removed because it has had no 12199 effect in the last few GCC releases. 12200 * The i386 -msvr3-shlib option has been removed since it is no longer 12201 used. 12202 * Fastcall for i386 has been changed not to pass aggregate arguments 12203 in registers, following Microsoft compilers. 12204 * Support for the AOF assembler has been removed from the ARM back 12205 end; this affects only the targets arm-semi-aof and armel-semi-aof, 12206 which are no longer recognized. We removed these targets without a 12207 deprecation period because we discovered that they have been 12208 unusable since GCC 4.0.0. 12209 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 12210 has been removed. This support had been deprecated since GCC 4.0.0. 12211 * Support for a number of older systems and recently unmaintained or 12212 untested target ports of GCC has been declared obsolete in GCC 4.3. 12213 Unless there is activity to revive them, the next release of GCC 12214 will have their sources permanently removed. 12215 All GCC ports for the following processor architectures have been 12216 declared obsolete: 12217 + Morpho MT (mt-*) 12218 The following aliases for processor architectures have been 12219 declared obsolete. Users should use the indicated generic target 12220 names instead, with compile-time options such as -mcpu or 12221 configure-time options such as --with-cpu to control the 12222 configuration more precisely. 12223 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 12224 instead). 12225 + parisc*-*-* (use hppa*-*-* instead). 12226 + m680[012]0-*-* (use m68k-*-* instead). 12227 All GCC ports for the following operating systems have been 12228 declared obsolete: 12229 + BeOS (*-*-beos*) 12230 + kaOS (*-*-kaos*) 12231 + GNU/Linux using the a.out object format (*-*-linux*aout*) 12232 + GNU/Linux using version 1 of the GNU C Library 12233 (*-*-linux*libc1*) 12234 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 12235 *-*-solaris2.[0-6].*) 12236 + Miscellaneous System V (*-*-sysv*) 12237 + WindISS (*-*-windiss*) 12238 Also, those for some individual systems on particular architectures 12239 have been obsoleted: 12240 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 12241 + CRIS with a.out object format (cris-*-aout) 12242 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 12243 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 12244 + PRO on PA-RISC (hppa1.1-*-pro*) 12245 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 12246 i[34567]86-sequent-sysv4*) 12247 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 12248 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 12249 was previously [5]removed in 2001, leaving only the support 12250 for UWIN as a target now being deprecated) 12251 + ChorusOS on PowerPC (powerpc-*-chorusos*) 12252 + All VAX configurations apart from NetBSD and OpenBSD 12253 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 12254 * The [6]-Wconversion option has been modified. Its purpose now is to 12255 warn for implicit conversions that may alter a value. This new 12256 behavior is available for both C and C++. Warnings about 12257 conversions between signed and unsigned integers can be disabled by 12258 using -Wno-sign-conversion. In C++, they are disabled by default 12259 unless -Wsign-conversion is explicitly requested. The old behavior 12260 of -Wconversion, that is, warn for prototypes causing a type 12261 conversion that is different from what would happen to the same 12262 argument in the absence of a prototype, has been moved to a new 12263 option -Wtraditional-conversion, which is only available for C. 12264 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 12265 been removed because they were deprecated for more than 3 GCC major 12266 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 12267 -mtune=pentiumpro as a replacement. 12268 * The -funsafe-math-optimizations option now automatically turns on 12269 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 12270 reassociation and thus may introduce or remove traps. 12271 * The -ftree-vectorize option is now on by default under -O3. In 12272 order to generate code for a SIMD extension, it has to be enabled 12273 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 12274 i?86 and x86_64. 12275 * More information on porting to GCC 4.3 from previous versions of 12276 GCC can be found in the [7]porting guide for this release. 12277 12278General Optimizer Improvements 12279 12280 * The GCC middle-end has been integrated with the MPFR library. This 12281 allows GCC to evaluate and replace at compile-time calls to 12282 built-in math functions having constant arguments with their 12283 mathematically equivalent results. In making use of MPFR, GCC can 12284 generate correct results regardless of the math library 12285 implementation or floating point precision of the host platform. 12286 This also allows GCC to generate identical results regardless of 12287 whether one compiles in native or cross-compile configurations to a 12288 particular target. The following built-in functions take advantage 12289 of this new capability: acos, acosh, asin, asinh, atan2, atan, 12290 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 12291 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 12292 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 12293 tan, tanh, tgamma, y0, y1 and yn. The float and long double 12294 variants of these functions (e.g. sinf and sinl) are also handled. 12295 The sqrt and cabs functions with constant arguments were already 12296 optimized in prior GCC releases. Now they also use MPFR. 12297 * A new forward propagation pass on RTL was added. The new pass 12298 replaces several slower transformations, resulting in compile-time 12299 improvements as well as better code generation in some cases. 12300 * A new command-line switch -frecord-gcc-switches has been added to 12301 GCC, although it is only enabled for some targets. The switch 12302 causes the command line that was used to invoke the compiler to be 12303 recorded into the object file that is being created. The exact 12304 format of this recording is target and binary file format 12305 dependent, but it usually takes the form of a note section 12306 containing ASCII text. The switch is related to the -fverbose-asm 12307 switch, but that one only records the information in the assembler 12308 output file as comments, so the information never reaches the 12309 object file. 12310 * The inliner heuristic is now aware of stack frame consumption. New 12311 command-line parameters --param large-stack-frame and --param 12312 large-stack-frame-growth can be used to limit stack frame size 12313 growth caused by inlining. 12314 * During feedback directed optimizations, the expected block size the 12315 memcpy, memset and bzero functions operate on is discovered and for 12316 cases of commonly used small sizes, specialized inline code is 12317 generated. 12318 * __builtin_expect no longer requires its argument to be a compile 12319 time constant. 12320 * Interprocedural optimization was reorganized to work on functions 12321 in SSA form. This enables more precise and cheaper dataflow 12322 analysis and makes writing interprocedural optimizations easier. 12323 The following improvements have been implemented on top of this 12324 framework: 12325 + Pre-inline optimization: Selected local optimization passes 12326 are run before the inliner (and other interprocedural passes) 12327 are executed. This significantly improves the accuracy of code 12328 growth estimates used by the inliner and reduces the overall 12329 memory footprint for large compilation units. 12330 + Early inlining (a simple bottom-up inliner pass inlining only 12331 functions whose body is smaller than the expected call 12332 overhead) is now executed with the early optimization passes, 12333 thus inlining already optimized function bodies into an 12334 unoptimized function that is subsequently optimized by early 12335 optimizers. This enables the compiler to quickly eliminate 12336 abstraction penalty in C++ programs. 12337 + Interprocedural constant propagation now operate on SSA form 12338 increasing accuracy of the analysis. 12339 * A new internal representation for GIMPLE statements has been 12340 contributed, resulting in compile-time memory savings. 12341 * The vectorizer was enhanced to support vectorization of outer 12342 loops, intra-iteration parallelism (loop-aware SLP), vectorization 12343 of strided accesses and loops with multiple data-types. Run-time 12344 dependency testing using loop versioning was added. The cost model, 12345 turned on by -fvect-cost-model, was developed. 12346 12347New Languages and Language specific improvements 12348 12349 * We have added new command-line options 12350 -finstrument-functions-exclude-function-list and 12351 -finstrument-functions-exclude-file-list. They provide more control 12352 over which functions are annotated by the -finstrument-functions 12353 option. 12354 12355 C family 12356 12357 * Implicit conversions between generic vector types are now only 12358 permitted when the two vectors in question have the same number of 12359 elements and compatible element types. (Note that the restriction 12360 involves compatible element types, not implicitly-convertible 12361 element types: thus, a vector type with element type int may not be 12362 implicitly converted to a vector type with element type unsigned 12363 int.) This restriction, which is in line with specifications for 12364 SIMD architectures such as AltiVec, may be relaxed using the flag 12365 -flax-vector-conversions. This flag is intended only as a 12366 compatibility measure and should not be used for new code. 12367 * -Warray-bounds has been added and is now enabled by default for 12368 -Wall . It produces warnings for array subscripts that can be 12369 determined at compile time to be always out of bounds. 12370 -Wno-array-bounds will disable the warning. 12371 * The constructor and destructor function attributes now accept 12372 optional priority arguments which control the order in which the 12373 constructor and destructor functions are run. 12374 * New [8]command-line options -Wtype-limits, -Wold-style-declaration, 12375 -Wmissing-parameter-type, -Wempty-body, -Wclobbered and 12376 -Wignored-qualifiers have been added for finer control of the 12377 diverse warnings enabled by -Wextra. 12378 * A new function attribute alloc_size has been added to mark up 12379 malloc style functions. For constant sized allocations this can be 12380 used to find out the size of the returned pointer using the 12381 __builtin_object_size() function for buffer overflow checking and 12382 similar. This supplements the already built-in malloc and calloc 12383 constant size handling. 12384 * Integer constants written in binary are now supported as a GCC 12385 extension. They consist of a prefix 0b or 0B, followed by a 12386 sequence of 0 and 1 digits. 12387 * A new predefined macro __COUNTER__ has been added. It expands to 12388 sequential integral values starting from 0. In conjunction with the 12389 ## operator, this provides a convenient means to generate unique 12390 identifiers. 12391 * A new command-line option -fdirectives-only has been added. It 12392 enables a special preprocessing mode which improves the performance 12393 of applications like distcc and ccache. 12394 * Fixed-point data types and operators have been added. They are 12395 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 12396 Currently, only MIPS targets are supported. 12397 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 12398 N1241, is now supported as a GCC extension to C for targets 12399 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 12400 and x86_64-*-linux-gnu. The feature introduces new data types 12401 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 12402 DD, and DL. 12403 12404 C++ 12405 12406 * [9]Experimental support for the upcoming ISO C++ standard, C++0x. 12407 * -Wc++0x-compat has been added and is now enabled by default for 12408 -Wall. It produces warnings for constructs whose meaning differs 12409 between ISO C++ 1998 and C++0x. 12410 * The -Wparentheses option now works for C++ as it does for C. It 12411 warns if parentheses are omitted when operators with confusing 12412 precedence are nested. It also warns about ambiguous else 12413 statements. Since -Wparentheses is enabled by -Wall, this may cause 12414 additional warnings with existing C++ code which uses -Wall. These 12415 new warnings may be disabled by using -Wall -Wno-parentheses. 12416 * The -Wmissing-declarations now works for C++ as it does for C. 12417 * The -fvisibility-ms-compat flag was added, to make it easier to 12418 port larger projects using shared libraries from Microsoft's Visual 12419 Studio to ELF and Mach-O systems. 12420 * C++ attribute handling has been overhauled for template arguments 12421 (ie dependent types). In particular, __attribute__((aligned(T))); 12422 works for C++ types. 12423 12424 Runtime Library (libstdc++) 12425 12426 * [10]Experimental support for the upcoming ISO C++ standard, C++0x. 12427 * Support for TR1 mathematical special functions and regular 12428 expressions. 12429 * Default what implementations give more elaborate exception strings 12430 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 12431 * Header dependencies have been streamlined, reducing unnecessary 12432 includes and pre-processed bloat. 12433 * Variadic template implementations of items in <tuple> and 12434 <functional>. 12435 * An experimental [11]parallel mode has been added. This is a 12436 parallel implementation of many C++ Standard library algorithms, 12437 like std::accumulate, std::for_each, std::transform, or std::sort, 12438 to give but four examples. These algorithms can be substituted for 12439 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 12440 or all existing algorithms can be transformed via the 12441 -D_GLIBCXX_PARALLEL macro. 12442 * Debug mode versions of classes in <unordered_set> and 12443 <unordered_map>. 12444 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 12445 now <backward/hash_set> and <backward/hash_map>. This code: 12446 #include <ext/hash_set> 12447 __gnu_cxx::hash_set<int> s; 12448 12449 Can be transformed (in order of preference) to: 12450 #include <tr1/unordered_set> 12451 std::tr1::unordered_set<int> s; 12452 12453 or 12454 #include <backward/hash_set> 12455 __gnu_cxx::hash_set<int> s; 12456 12457 Similar transformations apply to __gnu_cxx::hash_map, 12458 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 12459 __gnu_cxx::hash_multiset. 12460 12461 Fortran 12462 12463 * Due to the fact that the GMP and MPFR libraries are required for 12464 all languages, Fortran is no longer special in this regard and is 12465 available by default. 12466 * The [12]-fexternal-blas option has been added, which generates 12467 calls to BLAS routines for intrinsic matrix operations such as 12468 matmul rather than using the built-in algorithms. 12469 * Support to give a backtrace (compiler flag -fbacktrace or 12470 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 12471 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 12472 run-time error occured. 12473 * GNU Fortran now defines __GFORTRAN__ when it runs the C 12474 preprocessor (CPP). 12475 * The [13]-finit-local-zero, -finit-real, -finit-integer, 12476 -finit-character, and -finit-logical options have been added, which 12477 can be used to initialize local variables. 12478 * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added, 12479 which calculate the Gamma function and its logarithm. Use EXTERNAL 12480 gamma if you want to use your own gamma function. 12481 * GNU Fortran now regards the backslash character as literal (as 12482 required by the Fortran 2003 standard); using [16]-fbackslash GNU 12483 Fortran interprets backslashes as C-style escape characters. 12484 * The [17]interpretation of binary, octal and hexadecimal (BOZ) 12485 literal constants has been changed. Before they were always 12486 interpreted as integer; now they are bit-wise transferred as 12487 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 12488 2003 standard, and for real and complex variables in DATA 12489 statements or when directly assigned to real and complex variables. 12490 Everywhere else and especially in expressions they are still 12491 regarded as integer constants. 12492 * Fortran 2003 support has been extended: 12493 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 12494 + Pointer intent 12495 + Intrinsic module ISO_ENV_FORTRAN 12496 + Interoperability with C (ISO C Bindings) 12497 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 12498 attribute) 12499 + Fortran 2003 BOZ 12500 12501 Java (GCJ) 12502 12503 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 12504 This enables the use of all 1.5 language features, and fixes most 12505 existing front end bugs. 12506 * libgcj now supports all 1.5 language features which require runtime 12507 support: foreach, enum, annotations, generics, and auto-boxing. 12508 * We've made many changes to the tools shipped with gcj. 12509 + The old jv-scan tool has been removed. This tool never really 12510 worked properly. There is no replacement. 12511 + gcjh has been rewritten. Some of its more obscure options no 12512 longer work, but are still recognized in an attempt at 12513 compatibility. gjavah is a new program with similar 12514 functionality but different command-line options. 12515 + grmic and grmiregistry have been rewritten. grmid has been 12516 added. 12517 + gjar replaces the old fastjar. 12518 + gjarsigner (used for signing jars), gkeytool (used for key 12519 management), gorbd (for CORBA), gserialver (computes 12520 serialization UIDs), and gtnameserv (also for CORBA) are now 12521 installed. 12522 * The ability to dump the contents of the java run time heap to a 12523 file for off-line analysis has been added. The heap dumps may be 12524 analyzed with the new gc-analyze tool. They may be generated on 12525 out-of-memory conditions or on demand and are controlled by the new 12526 run time class gnu.gcj.util.GCInfo. 12527 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 12528 provide correct, updated, timezone information. This means that 12529 packagers no longer have to update libgcj when a time zone change 12530 is published. 12531 12532New Targets and Target Specific Improvements 12533 12534 IA-32/x86-64 12535 12536 * Tuning for Intel Core 2 processors is available via -mtune=core2 12537 and -march=core2. 12538 * Tuning for AMD Geode processors is available via -mtune=geode and 12539 -march=geode. 12540 * Code generation of block move (memcpy) and block set (memset) was 12541 rewritten. GCC can now pick the best algorithm (loop, unrolled 12542 loop, instruction with rep prefix or a library call) based on the 12543 size of the block being copied and the CPU being optimized for. A 12544 new option -minline-stringops-dynamically has been added. With this 12545 option string operations of unknown size are expanded such that 12546 small blocks are copied by in-line code, while for large blocks a 12547 library call is used. This results in faster code than 12548 -minline-all-stringops when the library implementation is capable 12549 of using cache hierarchy hints. The heuristic choosing the 12550 particular algorithm can be overwritten via -mstringop-strategy. 12551 Newly also memset of values different from 0 is inlined. 12552 * GCC no longer places the cld instruction before string operations. 12553 Both i386 and x86-64 ABI documents mandate the direction flag to be 12554 clear at the entry of a function. It is now invalid to set the flag 12555 in asm statement without reseting it afterward. 12556 * Support for SSSE3 built-in functions and code generation are 12557 available via -mssse3. 12558 * Support for SSE4.1 built-in functions and code generation are 12559 available via -msse4.1. 12560 * Support for SSE4.2 built-in functions and code generation are 12561 available via -msse4.2. 12562 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 12563 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 12564 allow explicit control of x87 floating point precision. 12565 * Support for __float128 (TFmode) IEEE quad type and corresponding 12566 TCmode IEEE complex quad type is available via the soft-fp library 12567 on x86_64 targets. This includes basic arithmetic operations 12568 (addition, subtraction, negation, multiplication and division) on 12569 __float128 real and TCmode complex values, the full set of IEEE 12570 comparisons between __float128 values, conversions to and from 12571 float, double and long double floating point types, as well as 12572 conversions to and from signed or unsigned integer, signed or 12573 unsigned long integer and signed or unsigned quad (TImode) integer 12574 types. Additionally, all operations generate the full set of IEEE 12575 exceptions and support the full set of IEEE rounding modes. 12576 * GCC can now utilize the ACML library for vectorizing calls to a set 12577 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 12578 link to an ACML ABI compatible library. 12579 12580 ARM 12581 12582 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 12583 has been added. 12584 12585 CRIS 12586 12587 New features 12588 12589 * Compiler and Library support for the CRIS v32 architecture, as 12590 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 12591 added. 12592 12593 Configuration changes 12594 12595 * The cris-*-elf target now includes support for CRIS v32, including 12596 libraries, through the -march=v32 option. 12597 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 12598 * A new crisv32-*-linux* target defaults to generate code for CRIS 12599 v32. 12600 * The cris-*-aout target has been obsoleted. 12601 12602 Improved support for built-in functions 12603 12604 * GCC can now use the lz and swapwbr instructions to implement the 12605 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 12606 * __builtin_bswap32 is now implemented using the swapwb instruction, 12607 when available. 12608 12609 m68k and ColdFire 12610 12611 New features 12612 12613 * Support for several new ColdFire processors has been added. You can 12614 generate code for them using the new -mcpu option. 12615 * All targets now support ColdFire processors. 12616 * m68k-uclinux targets have improved support for C++ constructors and 12617 destructors, and for shared libraries. 12618 * It is now possible to set breakpoints on the first or last line of 12619 a function, even if there are no statements on that line. 12620 12621 Optimizations 12622 12623 * Support for sibling calls has been added. 12624 * More use is now made of the ColdFire mov3q instruction. 12625 * __builtin_clz is now implemented using the ff1 ColdFire 12626 instruction, when available. 12627 * GCC now honors the -m68010 option. 68010 code now uses clr rather 12628 than move to zero volatile memory. 12629 * 68020 targets and above can now use symbol(index.size*scale) 12630 addresses for indexed array accesses. Earlier compilers would 12631 always load the symbol into a base register first. 12632 12633 Configuration changes 12634 12635 * All m68k and ColdFire targets now allow the default processor to be 12636 set at configure time using --with-cpu. 12637 * A --with-arch configuration option has been added. This option 12638 allows you to restrict a target to ColdFire or non-ColdFire 12639 processors. 12640 12641 Preprocessor macros 12642 12643 * An __mcfv*__ macro is now defined for all ColdFire targets. 12644 (Earlier versions of GCC only defined __mcfv4e__.) 12645 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 12646 * All targets now define __mc68010 and __mc68010__ when generating 12647 68010 code. 12648 12649 Command-line changes 12650 12651 * New command-line options -march, -mcpu, -mtune and -mhard-float 12652 have been added. These options apply to both m68k and ColdFire 12653 targets. 12654 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 12655 versions of -mshort, etc. 12656 * -fforce-addr has been removed. It is now ignored by the compiler. 12657 12658 Other improvements 12659 12660 * ColdFire targets now try to maintain a 4-byte-aligned stack where 12661 possible. 12662 * m68k-uclinux targets now try to avoid situations that lead to the 12663 load-time error: BINFMT_FLAT: reloc outside program. 12664 12665 MIPS 12666 12667 Changes to existing configurations 12668 12669 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 12670 and n64. Every GNU/Linux configuration now builds these libraries 12671 by default. 12672 * GNU/Linux configurations now generate -mno-shared code unless 12673 overridden by -fpic, -fPIC, -fpie or -fPIE. 12674 * mipsisa32*-linux-gnu configurations now generate hard-float code by 12675 default, just like other mipsisa32* and mips*-linux-gnu 12676 configurations. You can build a soft-float version of any 12677 mips*-linux-gnu configuration by passing --with-float=soft to 12678 configure. 12679 * mips-wrs-vxworks now supports run-time processes (RTPs). 12680 12681 Changes to existing command-line options 12682 12683 * The -march and -mtune options no longer accept 24k as a processor 12684 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 12685 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 12686 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 12687 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 12688 24kx, 24kex and 34kx. 12689 12690 New configurations 12691 12692 GCC now supports the following configurations: 12693 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 12694 default. Earlier releases also recognized this configuration, but 12695 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 12696 you can customize any mips*-linux-gnu* configuration to a 12697 particular ISA or processor by passing an appropriate --with-arch 12698 option to configure. 12699 * mipsisa*-sde-elf*, which provides compatibility with MIPS 12700 Technologies' SDE toolchains. The configuration uses the SDE 12701 libraries by default, but you can use it like other newlib-based 12702 ELF configurations by passing --with-newlib to configure. It is the 12703 only configuration besides mips64vr*-elf* to build MIPS16 as well 12704 as non-MIPS16 libraries. 12705 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 12706 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 12707 and 64-bit forms of the EABI. 12708 12709 New processors and application-specific extensions 12710 12711 * Support for the SmartMIPS ASE is available through the new 12712 -msmartmips option. 12713 * Support for revision 2 of the DSP ASE is available through the new 12714 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 12715 indicates the revision of the ASE in use. 12716 * Support for the 4KS and 74K families of processors is available 12717 through the -march and -mtune options. 12718 12719 Improved support for built-in functions 12720 12721 * GCC can now use load-linked, store-conditional and sync 12722 instructions to implement atomic built-in functions such as 12723 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 12724 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 12725 * GCC can now use the clz and dclz instructions to implement the 12726 __builtin_ctz and __builtin_ffs families of functions. 12727 * There is a new __builtin___clear_cache function for flushing the 12728 instruction cache. GCC expands this function inline on MIPS32 12729 revision 2 targets, otherwise it calls the function specified by 12730 -mcache-flush-func. 12731 12732 MIPS16 improvements 12733 12734 * GCC can now compile objects that contain a mixture of MIPS16 and 12735 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 12736 for specifying which mode a function should use. 12737 * A new option called -minterlink-mips16 makes non-MIPS16 code 12738 link-compatible with MIPS16 code. 12739 * After many bug fixes, the long-standing MIPS16 -mhard-float support 12740 should now work fairly reliably. 12741 * GCC can now use the MIPS16e save and restore instructions. 12742 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 12743 with -G0 -fsection-anchors is often smaller than code compiled with 12744 -G8. However, please note that you must usually compile all objects 12745 in your application with the same -G option; see the documentation 12746 of -G for details. 12747 * A new option called-mcode-readable specifies which instructions are 12748 allowed to load from the code segment. -mcode-readable=yes is the 12749 default and says that any instruction may load from the code 12750 segment. The other alternatives are -mcode-readable=pcrel, which 12751 says that only PC-relative MIPS16 instructions may load from the 12752 code segment, and -mcode-readable=no, which says that no 12753 instruction may do so. Please see the documentation for more 12754 details, including example uses. 12755 12756 Small-data improvements 12757 12758 There are three new options for controlling small data: 12759 * -mno-extern-sdata, which disables small-data accesses for 12760 externally-defined variables. Code compiled with -Gn 12761 -mno-extern-sdata will be link-compatible with any -G setting 12762 between -G0 and -Gn inclusive. 12763 * -mno-local-sdata, which disables the use of small-data sections for 12764 data that is not externally visible. This option can be a useful 12765 way of reducing small-data usage in less performance-critical parts 12766 of an application. 12767 * -mno-gpopt, which disables the use of the $gp register while still 12768 honoring the -G limit when placing externally-visible data. This 12769 option implies -mno-extern-sdata and -mno-local-sdata and it can be 12770 useful in situations where $gp does not necessarily hold the 12771 expected value. 12772 12773 Miscellaneous improvements 12774 12775 * There is a new option called -mbranch-cost for tweaking the 12776 perceived cost of branches. 12777 * If GCC is configured to use a version of GAS that supports the 12778 .gnu_attribute directive, it will use that directive to record 12779 certain properties of the output code. .gnu_attribute is new to GAS 12780 2.18. 12781 * There are two new function attributes, near and far, for overriding 12782 the command-line setting of -mlong-calls on a function-by-function 12783 basis. 12784 * -mfp64, which previously required a 64-bit target, now works with 12785 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 12786 mipsisa*-sde-elf* configurations provide suitable library support. 12787 * GCC now recognizes the -mdmx and -mmt options and passes them down 12788 to the assembler. It does nothing else with the options at present. 12789 12790 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 12791 (BEA) 12792 12793 * Support has been added for this new architecture. 12794 12795 RS6000 (POWER/PowerPC) 12796 12797 * Support for the PowerPC 750CL paired-single instructions has been 12798 added with a new powerpc-*-linux*paired* target configuration. It 12799 is enabled by an associated -mpaired option and can be accessed 12800 using new built-in functions. 12801 * Support for auto-detecting architecture and system configuration to 12802 auto-select processor optimization tuning. 12803 * Support for VMX on AIX 5.3 has been added. 12804 * Support for AIX Version 6.1 has been added. 12805 12806 S/390, zSeries and System z9 12807 12808 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 12809 added. When using the -march=z9-ec option, the compiler will 12810 generate code making use of instructions provided by the decimal 12811 floating point facility and the floating point conversion facility 12812 (pfpo). Besides the instructions used to implement decimal floating 12813 point operations these facilities also contain instructions to move 12814 between general purpose and floating point registers and to modify 12815 and copy the sign-bit of floating point values. 12816 * When the -march=z9-ec option is used the new 12817 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 12818 decimal floating point hardware instructions will be used or not. 12819 If none of them is given the hardware support is enabled by 12820 default. 12821 * The -mstack-guard option can now be omitted when using stack 12822 checking via -mstack-size in order to let GCC choose a sensible 12823 stack guard value according to the frame size of each function. 12824 * Various changes to improve performance of generated code have been 12825 implemented, including: 12826 + The condition code set by an add logical with carry 12827 instruction is now available for overflow checks like: a + b + 12828 carry < b. 12829 + The test data class instruction is now used to implement 12830 sign-bit and infinity checks of binary and decimal floating 12831 point numbers. 12832 12833 SPARC 12834 12835 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 12836 added. 12837 12838 Xtensa 12839 12840 * Stack unwinding for exception handling now uses by default a 12841 specialized version of DWARF unwinding. This is not 12842 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 12843 Xtensa with previous versions of GCC. 12844 * For Xtensa processors that include the Conditional Store option, 12845 the built-in functions for atomic memory access are now implemented 12846 using S32C1I instructions. 12847 * If the Xtensa NSA option is available, GCC will use it to implement 12848 the __builtin_ctz and __builtin_clz functions. 12849 12850Documentation improvements 12851 12852 * Existing libstdc++ documentation has been edited and restructured 12853 into a single DocBook XML manual. The results can be viewed online 12854 [18]here. 12855 12856Other significant improvements 12857 12858 * The compiler's --help command-line option has been extended so that 12859 it now takes an optional set of arguments. These arguments restrict 12860 the information displayed to specific classes of command-line 12861 options, and possibly only a subset of those options. It is also 12862 now possible to replace the descriptive text associated with each 12863 displayed option with an indication of its current value, or for 12864 binary options, whether it has been enabled or disabled. 12865 Here are some examples. The following will display all the options 12866 controlling warning messages: 12867 --help=warnings 12868 12869 Whereas this will display all the undocumented, target specific 12870 options: 12871 --help=target,undocumented 12872 12873 This sequence of commands will display the binary optimizations 12874 that are enabled by -O3: 12875 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 12876 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 12877 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 12878 12879 * The configure options --with-pkgversion and --with-bugurl have been 12880 added. These allow distributors of GCC to include a 12881 distributor-specific string in manuals and --version output and to 12882 specify the URL for reporting bugs in their versions of GCC. 12883 12884GCC 4.3.1 12885 12886 This is the [19]list of problem reports (PRs) from GCC's bug tracking 12887 system that are known to be fixed in the 4.3.1 release. This list might 12888 not be complete (that is, it is possible that some PRs that have been 12889 fixed are not listed here). 12890 12891Target Specific Changes 12892 12893 IA-32/x86-64 12894 12895 ABI changes 12896 12897 * Starting with GCC 4.3.1, decimal floating point variables are 12898 aligned to their natural boundaries when they are passed on the 12899 stack for i386. 12900 12901 Command-line changes 12902 12903 * Starting with GCC 4.3.1, the -mcld option has been added to 12904 automatically generate a cld instruction in the prologue of 12905 functions that use string instructions. This option is used for 12906 backward compatibility on some operating systems and can be enabled 12907 by default for 32-bit x86 targets by configuring GCC with the 12908 --enable-cld configure option. 12909 12910GCC 4.3.2 12911 12912 This is the [20]list of problem reports (PRs) from GCC's bug tracking 12913 system that are known to be fixed in the 4.3.2 release. This list might 12914 not be complete (that is, it is possible that some PRs that have been 12915 fixed are not listed here). 12916 12917GCC 4.3.3 12918 12919 This is the [21]list of problem reports (PRs) from GCC's bug tracking 12920 system that are known to be fixed in the 4.3.3 release. This list might 12921 not be complete (that is, it is possible that some PRs that have been 12922 fixed are not listed here). 12923 12924GCC 4.3.4 12925 12926 This is the [22]list of problem reports (PRs) from GCC's bug tracking 12927 system that are known to be fixed in the 4.3.4 release. This list might 12928 not be complete (that is, it is possible that some PRs that have been 12929 fixed are not listed here). 12930 12931GCC 4.3.5 12932 12933 This is the [23]list of problem reports (PRs) from GCC's bug tracking 12934 system that are known to be fixed in the 4.3.5 release. This list might 12935 not be complete (that is, it is possible that some PRs that have been 12936 fixed are not listed here). 12937 12938GCC 4.3.6 12939 12940 This is the [24]list of problem reports (PRs) from GCC's bug tracking 12941 system that are known to be fixed in the 4.3.6 release. This list might 12942 not be complete (that is, it is possible that some PRs that have been 12943 fixed are not listed here). 12944 12945 12946 For questions related to the use of GCC, please consult these web 12947 pages and the [25]GCC manuals. If that fails, the 12948 [26]gcc-help@gcc.gnu.org mailing list might help. Comments on these 12949 web pages and the development of GCC are welcome on our developer 12950 list at [27]gcc@gcc.gnu.org. All of [28]our lists have public 12951 archives. 12952 12953 Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and 12954 distribution of this entire article is permitted in any medium, 12955 provided this notice is preserved. 12956 12957 These pages are [30]maintained by the GCC team. Last modified 12958 2021-07-28[31]. 12959 12960References 12961 12962 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 12963 2. https://gmplib.org/ 12964 3. https://www.mpfr.org/ 12965 4. https://gcc.gnu.org/install/prerequisites.html 12966 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 12967 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 12968 7. http://gcc.gnu.org/gcc-4.3/porting_to.html 12969 8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 12970 9. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 12971 10. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 12972 11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 12973 12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 12974 13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 12975 14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 12976 15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 12977 16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 12978 17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 12979 18. https://gcc.gnu.org/onlinedocs/libstdc++/ 12980 19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 12981 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 12982 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 12983 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 12984 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 12985 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 12986 25. https://gcc.gnu.org/onlinedocs/ 12987 26. mailto:gcc-help@gcc.gnu.org 12988 27. mailto:gcc@gcc.gnu.org 12989 28. https://gcc.gnu.org/lists.html 12990 29. https://www.fsf.org/ 12991 30. https://gcc.gnu.org/about.html 12992 31. http://validator.w3.org/check/referer 12993====================================================================== 12994http://gcc.gnu.org/gcc-4.2/index.html 12995 12996 GCC 4.2 Release Series 12997 12998 (This release series is no longer supported.) 12999 13000 May 19, 2008 13001 13002 The [1]GNU project and the GCC developers are pleased to announce the 13003 release of GCC 4.2.4. 13004 13005 This release is a bug-fix release, containing fixes for regressions in 13006 GCC 4.2.3 relative to previous releases of GCC. 13007 13008Release History 13009 13010 GCC 4.2.4 13011 May 19, 2008 ([2]changes) 13012 13013 GCC 4.2.3 13014 February 1, 2008 ([3]changes) 13015 13016 GCC 4.2.2 13017 October 7, 2007 ([4]changes) 13018 13019 GCC 4.2.1 13020 July 18, 2007 ([5]changes) 13021 13022 GCC 4.2.0 13023 May 13, 2007 ([6]changes) 13024 13025References and Acknowledgements 13026 13027 GCC used to stand for the GNU C Compiler, but since the compiler 13028 supports several other languages aside from C, it now stands for the 13029 GNU Compiler Collection. 13030 13031 A list of [7]successful builds is updated as new information becomes 13032 available. 13033 13034 The GCC developers would like to thank the numerous people that have 13035 contributed new features, improvements, bug fixes, and other changes as 13036 well as test results to GCC. This [8]amazing group of volunteers is 13037 what makes GCC successful. 13038 13039 For additional information about GCC please refer to the [9]GCC project 13040 web site or contact the [10]GCC development mailing list. 13041 13042 To obtain GCC please use [11]our mirror sites or [12]our version 13043 control system. 13044 13045 13046 For questions related to the use of GCC, please consult these web 13047 pages and the [13]GCC manuals. If that fails, the 13048 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these 13049 web pages and the development of GCC are welcome on our developer 13050 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public 13051 archives. 13052 13053 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 13054 distribution of this entire article is permitted in any medium, 13055 provided this notice is preserved. 13056 13057 These pages are [18]maintained by the GCC team. Last modified 13058 2021-07-28[19]. 13059 13060References 13061 13062 1. http://www.gnu.org/ 13063 2. http://gcc.gnu.org/gcc-4.2/changes.html 13064 3. http://gcc.gnu.org/gcc-4.2/changes.html 13065 4. http://gcc.gnu.org/gcc-4.2/changes.html 13066 5. http://gcc.gnu.org/gcc-4.2/changes.html 13067 6. http://gcc.gnu.org/gcc-4.2/changes.html 13068 7. http://gcc.gnu.org/gcc-4.2/buildstat.html 13069 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13070 9. http://gcc.gnu.org/index.html 13071 10. mailto:gcc@gcc.gnu.org 13072 11. http://gcc.gnu.org/mirrors.html 13073 12. http://gcc.gnu.org/git.html 13074 13. https://gcc.gnu.org/onlinedocs/ 13075 14. mailto:gcc-help@gcc.gnu.org 13076 15. mailto:gcc@gcc.gnu.org 13077 16. https://gcc.gnu.org/lists.html 13078 17. https://www.fsf.org/ 13079 18. https://gcc.gnu.org/about.html 13080 19. http://validator.w3.org/check/referer 13081====================================================================== 13082http://gcc.gnu.org/gcc-4.2/changes.html 13083 13084 GCC 4.2 Release Series 13085 Changes, New Features, and Fixes 13086 13087Caveats 13088 13089 * GCC no longer accepts the -fshared-data option. This option has had 13090 no effect in any GCC 4 release; the targets to which the option 13091 used to apply had been removed before GCC 4.0. 13092 13093General Optimizer Improvements 13094 13095 * New command-line options specify the possible relationships among 13096 parameters and between parameters and global data. For example, 13097 -fargument-noalias-anything specifies that arguments do not alias 13098 any other storage. 13099 Each language will automatically use whatever option is required by 13100 the language standard. You should not need to use these options 13101 yourself. 13102 13103New Languages and Language specific improvements 13104 13105 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 13106 * New command-line options -fstrict-overflow and -Wstrict-overflow 13107 have been added. -fstrict-overflow tells the compiler that it may 13108 assume that the program follows the strict signed overflow 13109 semantics permitted for the language: for C and C++ this means that 13110 the compiler may assume that signed overflow does not occur. For 13111 example, a loop like 13112 for (i = 1; i > 0; i *= 2) 13113 13114 is presumably intended to continue looping until i overflows. With 13115 -fstrict-overflow, the compiler may assume that signed overflow 13116 will not occur, and transform this into an infinite loop. 13117 -fstrict-overflow is turned on by default at -O2, and may be 13118 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 13119 be used to warn about cases where the compiler assumes that signed 13120 overflow will not occur. It takes five different levels: 13121 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 13122 -Wstrict-overflow=1 is enabled by -Wall. 13123 * The new command-line option -fno-toplevel-reorder directs GCC to 13124 emit top-level functions, variables, and asm statements in the same 13125 order that they appear in the input file. This is intended to 13126 support existing code which relies on a particular ordering (for 13127 example, code which uses top-level asm statements to switch 13128 sections). For new code, it is generally better to use function and 13129 variable attributes. The -fno-toplevel-reorder option may be used 13130 for most cases which currently use -fno-unit-at-a-time. The 13131 -fno-unit-at-a-time option will be removed in some future version 13132 of GCC. If you know of a case which requires -fno-unit-at-a-time 13133 which is not fixed by -fno-toplevel-reorder, please open a bug 13134 report. 13135 13136 C family 13137 13138 * The pragma redefine_extname will now macro expand its tokens for 13139 compatibility with SunPRO. 13140 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 13141 GCC to handle inline functions as specified in the C99 standard. In 13142 preparation for this, GCC 4.2 will warn about any use of non-static 13143 inline functions in gnu99 or c99 mode. This new warning may be 13144 disabled with the new gnu_inline function attribute or the new 13145 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 13146 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 13147 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 13148 in the current compilation. 13149 * A new command-line option -Waddress has been added to warn about 13150 suspicious uses of memory addresses as, for example, using the 13151 address of a function in a conditional expression, and comparisons 13152 against the memory address of a string literal. This warning is 13153 enabled by -Wall. 13154 13155 C++ 13156 13157 * C++ visibility handling has been overhauled. 13158 Restricted visiblity is propagated from classes to members, from 13159 functions to local statics, and from templates and template 13160 arguments to instantiations, unless the latter has explicitly 13161 declared visibility. 13162 The visibility attribute for a class must come between the 13163 class-key and the name, not after the closing brace. 13164 Attributes are now allowed for enums and elaborated-type-specifiers 13165 that only declare a type. 13166 Members of the anonymous namespace are now local to a particular 13167 translation unit, along with any other declarations which use them, 13168 though they are still treated as having external linkage for 13169 language semantics. 13170 * The (undocumented) extension which permitted templates with default 13171 arguments to be bound to template template parameters with fewer 13172 parameters has been removed. For example: 13173 template <template <typename> class C> 13174 void f(C<double>) {} 13175 13176 template <typename T, typename U = int> 13177 struct S {}; 13178 13179 template void f(S<double>); 13180 13181 is no longer accepted by G++. The reason this code is not accepted 13182 is that S is a template with two parameters; therefore, it cannot 13183 be bound to C which has only one parameter. 13184 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 13185 releases, have been removed. 13186 * The command-line option -fconst-strings, deprecated in previous GCC 13187 releases, has been removed. 13188 * The configure variable enable-__cxa_atexit is now enabled by 13189 default for more targets. Enabling this variable is necessary in 13190 order for static destructors to be executed in the correct order, 13191 but it depends upon the presence of a non-standard C library in the 13192 target library in order to work. The variable is now enabled for 13193 more targets which are known to have suitable C libraries. 13194 * -Wextra will produce warnings for if statements with a semicolon as 13195 the only body, to catch code like: 13196 if (a); 13197 return 1; 13198 return 0; 13199 13200 To suppress the warning in valid cases, use { } instead. 13201 * The C++ front end now also produces strict aliasing warnings when 13202 -fstrict-aliasing -Wstrict-aliasing is in effect. 13203 13204 Runtime Library (libstdc++) 13205 13206 * Added support for TR1 <random>, <complex>, and C compatibility 13207 headers. In addition, a lock-free version of shared_ptr was 13208 contributed as part of Phillip Jordan's Google Summer of Code 13209 project on lock-free containers. 13210 * In association with the Summer of Code work on lock-free 13211 containers, the interface for atomic builtins was adjusted, 13212 creating simpler alternatives for non-threaded code paths. Also, 13213 usage was consolidated and all elements were moved from namespace 13214 std to namespace__gnu_cxx. Affected interfaces are the functions 13215 __exchange_and_add, __atomic_add, and the objects __mutex, 13216 __recursive_mutex, and __scoped_lock. 13217 * Support for versioning weak symbol names via namespace association 13218 was added. However, as this changes the names of exported symbols, 13219 this is turned off by default in the current ABI. Intrepid users 13220 can enable this feature by using 13221 --enable-symvers=gnu-versioned-namespace during configuration. 13222 * Revised, simplified, and expanded policy-based associative 13223 containers, including data types for tree and trie forms 13224 (basic_tree, tree, trie), lists (list_update), and both 13225 collision-chaining and probing hash-based containers 13226 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 13227 the [3]documentation. 13228 * The implementation of the debug mode was modified, whereby the 13229 debug namespaces were nested inside of namespace std and namespace 13230 __gnu_cxx in order to resolve some long standing corner cases 13231 involving name lookup. Debug functionality from the policy-based 13232 data structures was consolidated and enabled with the single macro, 13233 _GLIBCXX_DEBUG. See PR 26142 for more information. 13234 * Added extensions for type traits: __conditional_type, 13235 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 13236 * Added a typelist implementation for compile-time meta-programming. 13237 Elements for typelist construction and operation can be found 13238 within namespace __gnu_cxx::typelist. 13239 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 13240 exception-safety. 13241 * Enabled library-wide visibility control, allowing -fvisibility to 13242 be used. 13243 * Consolidated all nested namespaces and the conversion of 13244 __gnu_internal implementation-private details to anonymous 13245 namespaces whenever possible. 13246 * Implemented LWG resolutions DR 431 and DR 538. 13247 13248 Fortran 13249 13250 * Support for allocatable components has been added (TR 15581 and 13251 Fortran 2003). 13252 * Support for the Fortran 2003 streaming IO extension has been added. 13253 * The GNU Fortran compiler now uses 4-byte record markers by default 13254 for unformatted files to be compatible with g77 and most other 13255 compilers. The implementation allows for records greater than 2 GB 13256 and is compatible with several other compilers. Older versions of 13257 gfortran used 8-byte record markers by default (on most systems). 13258 In order to change the length of the record markers, e.g. to read 13259 unformatted files created by older gfortran versions, the 13260 [4]-frecord-marker=8 option can be used. 13261 13262 Java (GCJ) 13263 13264 * A new command-line option -static-libgcj has been added for targets 13265 that use a linker compatible with GNU Binutils. As its name 13266 implies, this causes libgcj to be linked statically. In some cases 13267 this causes the resulting executable to start faster and use less 13268 memory than if the shared version of libgcj were used. However 13269 caution should be used as it can also cause essential parts of the 13270 library to be omitted. Some of these issues are discussed in: 13271 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 13272 * fastjar is no longer bundled with GCC. To build libgcj, you will 13273 need either InfoZIP (both zip and unzip) or an external jar 13274 program. In the former case, the GCC build will install a jar shell 13275 script that is based on InfoZIP and provides the same functionality 13276 as fastjar. 13277 13278New Targets and Target Specific Improvements 13279 13280 IA-32/x86-64 13281 13282 * -mtune=generic can now be used to generate code running well on 13283 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 13284 Pentium-M, Intel Pentium 4 and Intel Core 2. 13285 * -mtune=native and -march=native will produce code optimized for the 13286 host architecture as detected using the cpuid instruction. 13287 * Added a new command-line option -fstackrealign and and 13288 __attribute__ ((force_align_arg_pointer)) to realign the stack at 13289 runtime. This allows functions compiled with a vector-aligned stack 13290 to be invoked from legacy objects that keep only word-alignment. 13291 13292 SPARC 13293 13294 * The default CPU setting has been changed from V7 to V9 in 32-bit 13295 mode on Solaris 7 and above. This is already the case in 64-bit 13296 mode. It can be overridden by specifying --with-cpu at configure 13297 time. 13298 * Back-end support of built-in functions for atomic memory access has 13299 been implemented. 13300 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 13301 added. 13302 13303 M32C 13304 13305 * Various bug fixes have made some functions (notably, functions 13306 returning structures) incompatible with previous releases. 13307 Recompiling all libraries is recommended. Note that code quality 13308 has considerably improved since 4.1, making a recompile even more 13309 beneficial. 13310 13311 MIPS 13312 13313 * Added support for the Broadcom SB-1A core. 13314 13315 IA-64 13316 13317 * Added support for IA-64 data and control speculation. By default 13318 speculation is enabled only during second scheduler pass. A number 13319 of machine flags was introduced to control the usage of speculation 13320 for both scheduler passes. 13321 13322 HPPA 13323 13324 * Added Java language support (libffi and libjava) for 32-bit HP-UX 13325 11 target. 13326 13327Obsolete Systems 13328 13329Documentation improvements 13330 13331 PDF Documentation 13332 13333 * A make pdf target has been added to the top-level makefile, 13334 enabling automated production of PDF documentation files. 13335 (Front-ends external to GCC should modify their Make-lang.in file 13336 to add a lang.pdf: target.) 13337 13338Other significant improvements 13339 13340 Build system improvements 13341 13342 * All the components of the compiler are now bootstrapped by default. 13343 This improves the resilience to bugs in the system compiler or 13344 binary compatibility problems, as well as providing better testing 13345 of GCC 4.2 itself. In addition, if you build the compiler from a 13346 combined tree, the assembler, linker, etc. will also be 13347 bootstrapped (i.e. built with themselves). 13348 You can disable this behavior, and go back to the pre-GCC 4.2 set 13349 up, by configuring GCC with --disable-bootstrap. 13350 * The rules that configure follows to find target tools resemble more 13351 closely the locations that the built compiler will search. In 13352 addition, you can use the new configure option --with-target-tools 13353 to specify where to find the target tools used during the build, 13354 without affecting what the built compiler will use. 13355 This can be especially useful when building packages of GCC. For 13356 example, you may want to build GCC with GNU as or ld, even if the 13357 resulting compiler to work with the native assembler and linker. To 13358 do so, you can use --with-target-tools to point to the native 13359 tools. 13360 13361 Incompatible changes to the build system 13362 13363 * Front-ends external to GCC should modify their Make-lang.in file to 13364 replace double-colon rules (e.g. dvi::) with normal rules (like 13365 lang.dvi:). Front-end makefile hooks do not use double-colon rules 13366 anymore. 13367 * Up to GCC 4.1, a popular way to specify the target tools used 13368 during the build was to create directories named gas, binutils, 13369 etc. in the build tree, and create links to the tools from there. 13370 This does not work any more when the compiler is bootstrapped. The 13371 new configure option --with-target-tools provides a better way to 13372 achieve the same effect, and works for all native and cross 13373 settings. 13374 13375 13376 For questions related to the use of GCC, please consult these web 13377 pages and the [6]GCC manuals. If that fails, the 13378 [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these 13379 web pages and the development of GCC are welcome on our developer 13380 list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives. 13381 13382 Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and 13383 distribution of this entire article is permitted in any medium, 13384 provided this notice is preserved. 13385 13386 These pages are [11]maintained by the GCC team. Last modified 13387 2021-07-28[12]. 13388 13389References 13390 13391 1. http://gcc.gnu.org/projects/gomp/ 13392 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 13393 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 13394 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 13395 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 13396 6. https://gcc.gnu.org/onlinedocs/ 13397 7. mailto:gcc-help@gcc.gnu.org 13398 8. mailto:gcc@gcc.gnu.org 13399 9. https://gcc.gnu.org/lists.html 13400 10. https://www.fsf.org/ 13401 11. https://gcc.gnu.org/about.html 13402 12. http://validator.w3.org/check/referer 13403====================================================================== 13404http://gcc.gnu.org/gcc-4.1/index.html 13405 13406 GCC 4.1 Release Series 13407 13408 (This release series is no longer supported.) 13409 13410 February 13, 2007 13411 13412 The [1]GNU project and the GCC developers are pleased to announce the 13413 release of GCC 4.1.2. 13414 13415 This release is a bug-fix release, containing fixes for regressions in 13416 GCC 4.1.1 relative to previous releases of GCC. 13417 13418Release History 13419 13420 GCC 4.1.2 13421 February 13, 2007 ([2]changes) 13422 13423 GCC 4.1.1 13424 May 24, 2006 ([3]changes) 13425 13426 GCC 4.1.0 13427 February 28, 2006 ([4]changes) 13428 13429References and Acknowledgements 13430 13431 GCC used to stand for the GNU C Compiler, but since the compiler 13432 supports several other languages aside from C, it now stands for the 13433 GNU Compiler Collection. 13434 13435 A list of [5]successful builds is updated as new information becomes 13436 available. 13437 13438 The GCC developers would like to thank the numerous people that have 13439 contributed new features, improvements, bug fixes, and other changes as 13440 well as test results to GCC. This [6]amazing group of volunteers is 13441 what makes GCC successful. 13442 13443 For additional information about GCC please refer to the [7]GCC project 13444 web site or contact the [8]GCC development mailing list. 13445 13446 To obtain GCC please use [9]our mirror sites or [10]our version control 13447 system. 13448 13449 13450 For questions related to the use of GCC, please consult these web 13451 pages and the [11]GCC manuals. If that fails, the 13452 [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these 13453 web pages and the development of GCC are welcome on our developer 13454 list at [13]gcc@gcc.gnu.org. All of [14]our lists have public 13455 archives. 13456 13457 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 13458 distribution of this entire article is permitted in any medium, 13459 provided this notice is preserved. 13460 13461 These pages are [16]maintained by the GCC team. Last modified 13462 2021-07-28[17]. 13463 13464References 13465 13466 1. http://www.gnu.org/ 13467 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 13468 3. http://gcc.gnu.org/gcc-4.1/changes.html 13469 4. http://gcc.gnu.org/gcc-4.1/changes.html 13470 5. http://gcc.gnu.org/gcc-4.1/buildstat.html 13471 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13472 7. http://gcc.gnu.org/index.html 13473 8. mailto:gcc@gcc.gnu.org 13474 9. http://gcc.gnu.org/mirrors.html 13475 10. http://gcc.gnu.org/git.html 13476 11. https://gcc.gnu.org/onlinedocs/ 13477 12. mailto:gcc-help@gcc.gnu.org 13478 13. mailto:gcc@gcc.gnu.org 13479 14. https://gcc.gnu.org/lists.html 13480 15. https://www.fsf.org/ 13481 16. https://gcc.gnu.org/about.html 13482 17. http://validator.w3.org/check/referer 13483====================================================================== 13484http://gcc.gnu.org/gcc-4.1/changes.html 13485 13486 GCC 4.1 Release Series 13487 Changes, New Features, and Fixes 13488 13489 The latest release in the 4.1 release series is [1]GCC 4.1.2. 13490 13491Caveats 13492 13493General Optimizer Improvements 13494 13495 * GCC now has infrastructure for inter-procedural optimizations and 13496 the following inter-procedural optimizations are implemented: 13497 + Profile guided inlining. When doing profile feedback guided 13498 optimization, GCC can now use the profile to make better 13499 informed decisions on whether inlining of a function is 13500 profitable or not. This means that GCC will no longer inline 13501 functions at call sites that are not executed very often, and 13502 that functions at hot call sites are more likely to be 13503 inlined. 13504 A new parameter min-inline-recursive-probability is also now 13505 available to throttle recursive inlining of functions with 13506 small average recursive depths. 13507 + Discovery of pure and const functions, a form of side-effects 13508 analysis. While older GCC releases could also discover such 13509 special functions, the new IPA-based pass runs earlier so that 13510 the results are available to more optimizers. The pass is also 13511 simply more powerful than the old one. 13512 + Analysis of references to static variables and type escape 13513 analysis, also forms of side-effects analysis. The results of 13514 these passes allow the compiler to be less conservative about 13515 call-clobbered variables and references. This results in more 13516 redundant loads being eliminated and in making static 13517 variables candidates for register promotion. 13518 + Improvement of RTL-based alias analysis. The results of type 13519 escape analysis are fed to the RTL type-based alias analyzer, 13520 allowing it to disambiguate more memory references. 13521 + Interprocedural constant propagation and function versioning. 13522 This pass looks for functions that are always called with the 13523 same constant value for one or more of the function arguments, 13524 and propagates those constants into those functions. 13525 + GCC will now eliminate static variables whose usage was 13526 optimized out. 13527 + -fwhole-program --combine can now be used to make all 13528 functions in program static allowing whole program 13529 optimization. As an exception, the main function and all 13530 functions marked with the new externally_visible attribute are 13531 kept global so that programs can link with runtime libraries. 13532 * GCC can now do a form of partial dead code elimination (PDCE) that 13533 allows code motion of expressions to the paths where the result of 13534 the expression is actually needed. This is not always a win, so the 13535 pass has been limited to only consider profitable cases. Here is an 13536 example: 13537 int foo (int *, int *); 13538 int 13539 bar (int d) 13540 { 13541 int a, b, c; 13542 b = d + 1; 13543 c = d + 2; 13544 a = b + c; 13545 if (d) 13546 { 13547 foo (&b, &c); 13548 a = b + c; 13549 } 13550 printf ("%d\n", a); 13551 } 13552 13553 The a = b + c can be sunk to right before the printf. Normal code 13554 sinking will not do this, it will sink the first one above into the 13555 else-branch of the conditional jump, which still gives you two 13556 copies of the code. 13557 * GCC now has a value range propagation pass. This allows the 13558 compiler to eliminate bounds checks and branches. The results of 13559 the pass can also be used to accurately compute branch 13560 probabilities. 13561 * The pass to convert PHI nodes to straight-line code (a form of 13562 if-conversion for GIMPLE) has been improved significantly. The two 13563 most significant improvements are an improved algorithm to 13564 determine the order in which the PHI nodes are considered, and an 13565 improvement that allow the pass to consider if-conversions of basic 13566 blocks with more than two predecessors. 13567 * Alias analysis improvements. GCC can now differentiate between 13568 different fields of structures in Tree-SSA's virtual operands form. 13569 This lets stores/loads from non-overlapping structure fields not 13570 conflict. A new algorithm to compute points-to sets was contributed 13571 that can allows GCC to see now that p->a and p->b, where p is a 13572 pointer to a structure, can never point to the same field. 13573 * Various enhancements to auto-vectorization: 13574 + Incrementally preserve SSA form when vectorizing. 13575 + Incrementally preserve loop-closed form when vectorizing. 13576 + Improvements to peeling for alignment: generate better code 13577 when the misalignment of an access is known at compile time, 13578 or when different accesses are known to have the same 13579 misalignment, even if the misalignment amount itself is 13580 unknown. 13581 + Consider dependence distance in the vectorizer. 13582 + Externalize generic parts of data reference analysis to make 13583 this analysis available to other passes. 13584 + Vectorization of conditional code. 13585 + Reduction support. 13586 * GCC can now partition functions in sections of hot and cold code. 13587 This can significantly improve performance due to better 13588 instruction cache locality. This feature works best together with 13589 profile feedback driven optimization. 13590 * A new pass to avoid saving of unneeded arguments to the stack in 13591 vararg functions if the compiler can prove that they will not be 13592 needed. 13593 * Transition of basic block profiling to tree level implementation 13594 has been completed. The new implementation should be considerably 13595 more reliable (hopefully avoiding profile mismatch errors when 13596 using -fprofile-use or -fbranch-probabilities) and can be used to 13597 drive higher level optimizations, such as inlining. 13598 The -ftree-based-profiling command-line option was removed and 13599 -fprofile-use now implies disabling old RTL level loop optimizer 13600 (-fno-loop-optimize). Speculative prefetching optimization 13601 (originally enabled by -fspeculative-prefetching) was removed. 13602 13603New Languages and Language specific improvements 13604 13605 C and Objective-C 13606 13607 * The old Bison-based C and Objective-C parser has been replaced by a 13608 new, faster hand-written recursive-descent parser. 13609 13610 Ada 13611 13612 * The build infrastructure for the Ada runtime library and tools has 13613 been changed to be better integrated with the rest of the build 13614 infrastructure of GCC. This should make doing cross builds of Ada a 13615 bit easier. 13616 13617 C++ 13618 13619 * ARM-style name-injection of friend declarations is no longer the 13620 default. For example: 13621 struct S { 13622 friend void f(); 13623 }; 13624 13625 void g() { f(); } 13626 will not be accepted; instead a declaration of f will need to be 13627 present outside of the scope of S. The new -ffriend-injection 13628 option will enable the old behavior. 13629 * The (undocumented) extension which permitted templates with default 13630 arguments to be bound to template template parameters with fewer 13631 parameters has been deprecated, and will be removed in the next 13632 major release of G++. For example: 13633 template <template <typename> class C> 13634 void f(C<double>) {} 13635 13636 template <typename T, typename U = int> 13637 struct S {}; 13638 13639 template void f(S<double>); 13640 13641 makes use of the deprecated extension. The reason this code is not 13642 valid ISO C++ is that S is a template with two parameters; 13643 therefore, it cannot be bound to C which has only one parameter. 13644 13645 Runtime Library (libstdc++) 13646 13647 * Optimization work: 13648 + A new implementation of std::search_n is provided, better 13649 performing in case of random access iterators. 13650 + Added further efficient specializations of istream functions, 13651 i.e., character array and string extractors. 13652 + Other smaller improvements throughout. 13653 * Policy-based associative containers, designed for high-performance, 13654 flexibility and semantic safety are delivered in ext/pb_assoc. 13655 * A versatile string class, __gnu_cxx::__versa_string, providing 13656 facilities conforming to the standard requirements for 13657 basic_string, is delivered in <ext/vstring.h>. In particular: 13658 + Two base classes are provided: the default one avoids 13659 reference counting and is optimized for short strings; the 13660 alternate one, still uses it while improving in a few low 13661 level areas (e.g., alignment). See vstring_fwd.h for some 13662 useful typedefs. 13663 + Various algorithms have been rewritten (e.g., replace), the 13664 code streamlined and simple optimizations added. 13665 + Option 3 of DR 431 is implemented for both available bases, 13666 thus improving the support for stateful allocators. 13667 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 13668 libstdc++/23953) and LWG resolutions put into effect for the first 13669 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 13670 Issue 6.19). The implementation status of TR1 is now tracked in the 13671 docs in tr1.html. 13672 13673 Objective-C++ 13674 13675 * A new language front end for Objective-C++ has been added. This 13676 language allows users to mix the object oriented features of 13677 Objective-C with those of C++. 13678 13679 Java (GCJ) 13680 13681 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 13682 features (plus some 0.20 bug-fixes) 13683 + Networking 13684 o The java.net.HttpURLConnection implementation no longer 13685 buffers the entire response body in memory. This means 13686 that response bodies larger than available memory can now 13687 be handled. 13688 + (N)IO 13689 o NIO FileChannel.map implementation, fast bulk put 13690 implementation for DirectByteBuffer (speeds up this 13691 method 10x). 13692 o FileChannel.lock() and FileChannel.force() implemented. 13693 + XML 13694 o gnu.xml fix for nodes created outside a namespace 13695 context. 13696 o Add support for output indenting and 13697 cdata-section-elements output instruction in 13698 xml.transform. 13699 o xml.xpath corrections for cases where elements/attributes 13700 might have been created in non-namespace-aware mode. 13701 Corrections to handling of XSL variables and minor 13702 conformance updates. 13703 + AWT 13704 o GNU JAWT implementation, the AWT Native Interface, which 13705 allows direct access to native screen resources from 13706 within a Canvas's paint method. GNU Classpath Examples 13707 comes with a Demo, see libjava/classpath/examples/README. 13708 o awt.datatransfer updated to 1.5 with support for 13709 FlavorEvents. The gtk+ awt peers now allow copy/paste of 13710 text, images, URIs/files and serialized objects with 13711 other applications and tracking clipboard change events 13712 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 13713 objects are supported). A GNU Classpath Examples 13714 datatransfer Demo was added to show the new 13715 functionality. 13716 o Split gtk+ awt peers event handling in two threads and 13717 improve gdk lock handling (solves several awt lock ups). 13718 o Speed up awt Image loading. 13719 o Better gtk+ scrollbar peer implementation when using gtk+ 13720 >= 2.6. 13721 o Handle image loading errors correctly for gdkpixbuf and 13722 MediaTracker. 13723 o Better handle GDK lock. Properly prefix gtkpeer native 13724 functions (cp_gtk). 13725 o GdkGraphics2D has been updated to use Cairo 0.5.x or 13726 higher. 13727 o BufferedImage and GtkImage rewrites. All image drawing 13728 operations should now work correctly (flipping requires 13729 gtk+ >= 2.6) 13730 o Future Graphics2D, image and text work is documented at: 13731 [2]http://developer.classpath.org/mediation/ClasspathGrap 13732 hicsImagesText 13733 o When gtk+ 2.6 or higher is installed the default log 13734 handler will produce stack traces whenever a WARNING, 13735 CRITICAL or ERROR message is produced. 13736 + Free Swing 13737 o The RepaintManager has been reworked for more efficient 13738 painting, especially for large GUIs. 13739 o The layout manager OverlayLayout has been implemented, 13740 the BoxLayout has been rewritten to make use of the 13741 SizeRequirements utility class and caching for more 13742 efficient layout. 13743 o Improved accessibility support. 13744 o Significant progress has been made in the implementation 13745 of the javax.swing.plaf.metal package, with most UI 13746 delegates in a working state now. Please test this with 13747 your own applications and provide feedback that will help 13748 us to improve this package. 13749 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 13750 extended to highlight various features in our Free Swing 13751 implementation. And it includes a look and feel switcher 13752 for Metal (default), Ocean and GNU themes. 13753 o The javax.swing.plaf.multi package is now implemented. 13754 o Editing and several key actions for JTree and JTable were 13755 implemented. 13756 o Lots of icons and look and feel improvements for Free 13757 Swing basic and metal themes were added. Try running the 13758 GNU Classpath Swing Demo in examples 13759 (gnu.classpath.examples.swing.Demo) with: 13760 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 13761 l or 13762 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 13763 l 13764 o Start of styled text capabilites for java.swing.text. 13765 o DefaultMutableTreeNode pre-order, post-order, depth-first 13766 and breadth-first traversal enumerations implemented. 13767 o JInternalFrame colors and titlebar draw properly. 13768 o JTree is working up to par (icons, selection and keyboard 13769 traversal). 13770 o JMenus were made more compatible in visual and 13771 programmatic behavior. 13772 o JTable changeSelection and multiple selections 13773 implemented. 13774 o JButton and JToggleButton change states work properly 13775 now. 13776 o JFileChooser fixes. 13777 o revalidate() and repaint() fixes which make Free Swing 13778 much more responsive. 13779 o MetalIconFactory implemented. 13780 o Free Swing Top-Level Compatibility. JFrame, JDialog, 13781 JApplet, JInternalFrame, and JWindow are now 1.5 13782 compatible in the sense that you can call add() and 13783 setLayout() directly on them, which will have the same 13784 effect as calling getContentPane().add() and 13785 getContentPane().setLayout(). 13786 o The JTree interface has been completed. JTrees now 13787 recognizes mouse clicks and selections work. 13788 o BoxLayout works properly now. 13789 o Fixed GrayFilter to actually work. 13790 o Metal SplitPane implemented. 13791 o Lots of Free Swing text and editor stuff work now. 13792 + Free RMI and Corba 13793 o Andrew Watson, Vice President and Technical Director of 13794 the Object Management Group, has officially assigned us 13795 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 13796 mark remote classpath-specific system exceptions. 13797 Obtaining the VMCID means that GNU Classpath now is a 13798 recogniseable type of node in a highly interoperable 13799 CORBA world. 13800 o GNU Classpath now includes the first working draft to 13801 support the RMI over IIOP protocol. The current 13802 implementation is capable of remote invocations, 13803 transferring various Serializables and Externalizables 13804 via RMI-IIOP protocol. It can flatten graphs and, at 13805 least for the simple cases, is interoperable with 1.5 13806 JDKs. 13807 o org.omg.PortableInterceptor and related functionality in 13808 other packages is now implemented: 13809 # The sever and client interceptors work as required 13810 since 1.4. 13811 # The IOR interceptor works as needed for 1.5. 13812 o The org.omg.DynamicAny package is completed and passes 13813 the prepared tests. 13814 o The Portable Object Adapter should now support the output 13815 of the recent IDL to java compilers. These compilers now 13816 generate servants and not CORBA objects as before, making 13817 the output depend on the existing POA implementation. 13818 Completing POA means that such code can already be tried 13819 to run on Classpath. Our POA is tested for the following 13820 usager scenarios: 13821 # POA converts servant to the CORBA object. 13822 # Servant provides to the CORBA object. 13823 # POA activates new CORBA object with the given Object 13824 Id (byte array) that is later accessible for the 13825 servant. 13826 # During the first call, the ServantActivator provides 13827 servant for this and all subsequent calls on the 13828 current object. 13829 # During each call, the ServantLocator provides 13830 servant for this call only. 13831 # ServantLocator or ServantActivator forwards call to 13832 another server. 13833 # POA has a single servant, responsible for all 13834 objects. 13835 # POA has a default servant, but some objects are 13836 explicitly connected to they specific servants. 13837 The POA is verified using tests from the former 13838 cost.omg.org. 13839 o The CORBA implementation is now a working prototype that 13840 should support features up to 1.3 inclusive. We invite 13841 groups writing CORBA dependent applications to try 13842 Classpath implementation, reporting any possible bugs. 13843 The CORBA prototype is interoperable with Sun's 13844 implementation v 1.4, transferring object references, 13845 primitive types, narrow and wide strings, arrays, 13846 structures, trees, abstract interfaces and value types 13847 (feature of CORBA 2.3) between these two platforms. 13848 Remote exceptions are transferred and handled correctly. 13849 The stringified object references (IORs) from various 13850 sources are parsed as required. The transient (for 13851 current session) and permanent (till jre restart) 13852 redirections work. Both Little and Big Endian encoded 13853 messages are accepted. The implementation is verified 13854 using tests from the former cost.omg.org. The current 13855 release includes working examples (see the examples 13856 directory), demonstrating the client-server 13857 communication, using either CORBA Request or IDL-based 13858 stub (usually generated by a IDL to java compiler). These 13859 examples also show how to use the Classpath CORBA naming 13860 service. The IDL to java compiler is not yet written, but 13861 as our library must be compatible, it naturally accepts 13862 the output of other idlj implementations. 13863 + Misc 13864 o Updated TimeZone data against Olson tzdata2005l. 13865 o Make zip and jar packages UTF-8 clean. 13866 o "native" code builds and compiles (warning free) on 13867 Darwin and Solaris. 13868 o java.util.logging.FileHandler now rotates files. 13869 o Start of a generic JDWP framework in gnu/classpath/jdwp. 13870 This is unfinished, but feedback (at classpath@gnu.org) 13871 from runtime hackers is greatly appreciated. Although 13872 most of the work is currently being done around gcj/gij 13873 we want this framework to be as VM neutral as possible. 13874 Early design is described in: 13875 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 13876 o QT4 AWT peers, enable by giving configure 13877 --enable-qt-peer. Included, but not ready for production 13878 yet. They are explicitly disabled and not supported. But 13879 if you want to help with the development of these new 13880 features we are interested in feedback. You will have to 13881 explicitly enable them to try them out (and they will 13882 most likely contain bugs). 13883 o Documentation fixes all over the place. See 13884 [4]http://developer.classpath.org/doc/ 13885 13886New Targets and Target Specific Improvements 13887 13888 IA-32/x86-64 13889 13890 * The x86-64 medium model (that allows building applications whose 13891 data segment exceeds 4GB) was redesigned to match latest ABI draft. 13892 New implementation split large datastructures into separate segment 13893 improving performance of accesses to small datastructures and also 13894 allows linking of small model libraries into medium model programs 13895 as long as the libraries are not accessing the large datastructures 13896 directly. Medium model is also supported in position independent 13897 code now. 13898 The ABI change results in partial incompatibility among medium 13899 model objects. Linking medium model libraries (or objects) compiled 13900 with new compiler into medium model program compiled with older 13901 will likely result in exceeding ranges of relocations. 13902 Binutils 2.16.91 or newer are required for compiling medium model 13903 now. 13904 13905 RS6000 (POWER/PowerPC) 13906 13907 * The AltiVec vector primitives in <altivec.h> are now implemented in 13908 a way that puts a smaller burden on the preprocessor, instead 13909 processing the "overloading" in the front ends. This should benefit 13910 compilation speed on AltiVec vector code. 13911 * AltiVec initializers now are generated more efficiently. 13912 * The popcountb instruction available on POWER5 now is generated. 13913 * The floating point round to integer instructions available on 13914 POWER5+ now is generated. 13915 * Floating point divides can be synthesized using the floating point 13916 reciprocal estimate instructions. 13917 * Double precision floating point constants are initialized as single 13918 precision values if they can be represented exactly. 13919 13920 S/390, zSeries and System z9 13921 13922 * Support for the IBM System z9 109 processor has been added. When 13923 using the -march=z9-109 option, the compiler will generate code 13924 making use of instructions provided by the extended immediate 13925 facility. 13926 * Support for 128-bit IEEE floating point has been added. When using 13927 the -mlong-double-128 option, the compiler will map the long double 13928 data type to 128-bit IEEE floating point. Using this option 13929 constitutes an ABI change, and requires glibc support. 13930 * Various changes to improve performance of generated code have been 13931 implemented, including: 13932 + In functions that do not require a literal pool, register %r13 13933 (which is traditionally reserved as literal pool pointer), can 13934 now be freely used for other purposes by the compiler. 13935 + More precise tracking of register use allows the compiler to 13936 generate more efficient function prolog and epilog code in 13937 certain cases. 13938 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 13939 instructions are now used to implement C string functions. 13940 + The MOVE CHARACTER instruction with single byte overlap is now 13941 used to implement the memset function with non-zero fill byte. 13942 + The LOAD ZERO instructions are now used where appropriate. 13943 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 13944 and INSERT IMMEDIATE instructions are now used more frequently 13945 to optimize bitfield operations. 13946 + The BRANCH ON COUNT instruction is now used more frequently. 13947 In particular, the fact that a loop contains a subroutine call 13948 no longer prevents the compiler from using this instruction. 13949 + The compiler is now aware that all shift and rotate 13950 instructions implicitly truncate the shift count to six bits. 13951 * Back-end support for the following generic features has been 13952 implemented: 13953 + The full set of [5]built-in functions for atomic memory 13954 access. 13955 + The -fstack-protector feature. 13956 + The optimization pass avoiding unnecessary stores of incoming 13957 argument registers in functions with variable argument list. 13958 13959 SPARC 13960 13961 * The default code model in 64-bit mode has been changed from 13962 Medium/Anywhere to Medium/Middle on Solaris. 13963 * TLS support is disabled by default on Solaris prior to release 10. 13964 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 13965 and later) by specifying --enable-tls at configure time. 13966 13967 MorphoSys 13968 13969 * Support has been added for this new architecture. 13970 13971Obsolete Systems 13972 13973Documentation improvements 13974 13975Other significant improvements 13976 13977 * GCC can now emit code for protecting applications from 13978 stack-smashing attacks. The protection is realized by buffer 13979 overflow detection and reordering of stack variables to avoid 13980 pointer corruption. 13981 * Some built-in functions have been fortified to protect them against 13982 various buffer overflow (and format string) vulnerabilities. 13983 Compared to the mudflap bounds checking feature, the safe builtins 13984 have far smaller overhead. This means that programs built using 13985 safe builtins should not experience any measurable slowdown. 13986 13987GCC 4.1.2 13988 13989 This is the [6]list of problem reports (PRs) from GCC's bug tracking 13990 system that are known to be fixed in the 4.1.2 release. This list might 13991 not be complete (that is, it is possible that some PRs that have been 13992 fixed are not listed here). 13993 13994 When generating code for a shared library, GCC now recognizes that 13995 global functions may be replaced when the program runs. Therefore, it 13996 is now more conservative in deducing information from the bodies of 13997 functions. For example, in this example: 13998 void f() {} 13999 void g() { 14000 try { f(); } 14001 catch (...) { 14002 cout << "Exception"; 14003 } 14004 } 14005 14006 G++ would previously have optimized away the catch clause, since it 14007 would have concluded that f cannot throw exceptions. Because users may 14008 replace f with another function in the main body of the program, this 14009 optimization is unsafe, and is no longer performed. If you wish G++ to 14010 continue to optimize as before, you must add a throw() clause to the 14011 declaration of f to make clear that it does not throw exceptions. 14012 14013 14014 For questions related to the use of GCC, please consult these web 14015 pages and the [7]GCC manuals. If that fails, the 14016 [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14017 web pages and the development of GCC are welcome on our developer 14018 list at [9]gcc@gcc.gnu.org. All of [10]our lists have public 14019 archives. 14020 14021 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 14022 distribution of this entire article is permitted in any medium, 14023 provided this notice is preserved. 14024 14025 These pages are [12]maintained by the GCC team. Last modified 14026 2021-07-28[13]. 14027 14028References 14029 14030 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 14031 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText 14032 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 14033 4. http://developer.classpath.org/doc/ 14034 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 14035 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 14036 7. https://gcc.gnu.org/onlinedocs/ 14037 8. mailto:gcc-help@gcc.gnu.org 14038 9. mailto:gcc@gcc.gnu.org 14039 10. https://gcc.gnu.org/lists.html 14040 11. https://www.fsf.org/ 14041 12. https://gcc.gnu.org/about.html 14042 13. http://validator.w3.org/check/referer 14043====================================================================== 14044http://gcc.gnu.org/gcc-4.0/index.html 14045 14046 GCC 4.0 Release Series 14047 14048 (This release series is no longer supported.) 14049 14050 January 31, 2007 14051 14052 The [1]GNU project and the GCC developers are pleased to announce the 14053 release of GCC 4.0.4. 14054 14055 This release is a bug-fix release, containing fixes for regressions in 14056 GCC 4.0.3 relative to previous releases of GCC. 14057 14058Release History 14059 14060 GCC 4.0.4 14061 January 31, 2007 ([2]changes) 14062 14063 GCC 4.0.3 14064 March 10, 2006 ([3]changes) 14065 14066 GCC 4.0.2 14067 September 28, 2005 ([4]changes) 14068 14069 GCC 4.0.1 14070 July 7, 2005 ([5]changes) 14071 14072 GCC 4.0.0 14073 April 20, 2005 ([6]changes) 14074 14075References and Acknowledgements 14076 14077 GCC used to stand for the GNU C Compiler, but since the compiler 14078 supports several other languages aside from C, it now stands for the 14079 GNU Compiler Collection. 14080 14081 A list of [7]successful builds is updated as new information becomes 14082 available. 14083 14084 The GCC developers would like to thank the numerous people that have 14085 contributed new features, improvements, bug fixes, and other changes as 14086 well as test results to GCC. This [8]amazing group of volunteers is 14087 what makes GCC successful. 14088 14089 For additional information about GCC please refer to the [9]GCC project 14090 web site or contact the [10]GCC development mailing list. 14091 14092 To obtain GCC please use [11]our mirror sites, or [12]our version 14093 control system. 14094 14095 14096 For questions related to the use of GCC, please consult these web 14097 pages and the [13]GCC manuals. If that fails, the 14098 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14099 web pages and the development of GCC are welcome on our developer 14100 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public 14101 archives. 14102 14103 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 14104 distribution of this entire article is permitted in any medium, 14105 provided this notice is preserved. 14106 14107 These pages are [18]maintained by the GCC team. Last modified 14108 2021-07-28[19]. 14109 14110References 14111 14112 1. http://www.gnu.org/ 14113 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 14114 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 14115 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 14116 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 14117 6. http://gcc.gnu.org/gcc-4.0/changes.html 14118 7. http://gcc.gnu.org/gcc-4.0/buildstat.html 14119 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14120 9. http://gcc.gnu.org/index.html 14121 10. mailto:gcc@gcc.gnu.org 14122 11. http://gcc.gnu.org/mirrors.html 14123 12. http://gcc.gnu.org/git.html 14124 13. https://gcc.gnu.org/onlinedocs/ 14125 14. mailto:gcc-help@gcc.gnu.org 14126 15. mailto:gcc@gcc.gnu.org 14127 16. https://gcc.gnu.org/lists.html 14128 17. https://www.fsf.org/ 14129 18. https://gcc.gnu.org/about.html 14130 19. http://validator.w3.org/check/referer 14131====================================================================== 14132http://gcc.gnu.org/gcc-4.0/changes.html 14133 14134 GCC 4.0 Release Series 14135 Changes, New Features, and Fixes 14136 14137 The latest release in the 4.0 release series is [1]GCC 4.0.4. 14138 14139Caveats 14140 14141 * GCC now generates location lists by default when compiling with 14142 debug info and optimization. 14143 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 14144 or later is needed to debug binaries containing location 14145 lists. 14146 + When you are trying to view a value of a variable in a part of 14147 a function where it has no location (for example when the 14148 variable is no longer used and thus its location was used for 14149 something else) GDB will say that it is not available. 14150 You can disable generating location lists by -fno-var-tracking. 14151 * GCC no longer accepts the -fwritable-strings option. Use named 14152 character arrays when you need a writable string. 14153 * The options -freduce-all-givs and -fmove-all-movables have been 14154 discontinued. They were used to circumvent a shortcoming in the 14155 heuristics of the old loop optimization code with respect to common 14156 Fortran constructs. The new (tree) loop optimizer works differently 14157 and doesn't need those work-arounds. 14158 * The graph-coloring register allocator, formerly enabled by the 14159 option -fnew-ra, has been discontinued. 14160 * -I- has been deprecated. -iquote is meant to replace the need for 14161 this option. 14162 * The MIPS -membedded-pic and -mrnames options have been removed. 14163 * All MIPS targets now require the GNU assembler. In particular, IRIX 14164 configurations can no longer use the MIPSpro assemblers, although 14165 they do still support the MIPSpro linkers. 14166 * The SPARC option -mflat has been removed. 14167 * English-language diagnostic messages will now use Unicode quotation 14168 marks in UTF-8 locales. (Non-English messages already used the 14169 quotes appropriate for the language in previous releases.) If your 14170 terminal does not support UTF-8 but you are using a UTF-8 locale 14171 (such locales are the default on many GNU/Linux systems) then you 14172 should set LC_CTYPE=C in the environment to disable that locale. 14173 Programs that parse diagnostics and expect plain ASCII 14174 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 14175 explanation of Unicode quotation marks for more information. 14176 * The specs file is no longer installed on most platforms. Most users 14177 will be totally unaffected. However, if you are accustomed to 14178 editing the specs file yourself, you will now have to use the 14179 -dumpspecs option to generate the specs file, and then edit the 14180 resulting file. 14181 14182General Optimizer Improvements 14183 14184 * The [3]tree ssa branch has been merged. This merge has brought in a 14185 completely new optimization framework based on a higher level 14186 intermediate representation than the existing RTL representation. 14187 Numerous new code transformations based on the new framework are 14188 available in GCC 4.0, including: 14189 + Scalar replacement of aggregates 14190 + Constant propagation 14191 + Value range propagation 14192 + Partial redundancy elimination 14193 + Load and store motion 14194 + Strength reduction 14195 + Dead store elimination 14196 + Dead and unreachable code elimination 14197 + [4]Autovectorization 14198 + Loop interchange 14199 + Tail recursion by accumulation 14200 Many of these passes outperform their counterparts from previous 14201 GCC releases. 14202 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 14203 scheduling optimization intended for loops that perform heavy 14204 computations. 14205 14206New Languages and Language specific improvements 14207 14208 C family 14209 14210 * The sentinel attribute has been added to GCC. This function 14211 attribute allows GCC to warn when variadic functions such as execl 14212 are not NULL terminated. See the GCC manual for a complete 14213 description of its behavior. 14214 * Given __attribute__((alias("target"))) it is now an error if target 14215 is not a symbol, defined in the same translation unit. This also 14216 applies to aliases created by #pragma weak alias=target. This is 14217 because it's meaningless to define an alias to an undefined symbol. 14218 On Solaris, the native assembler would have caught this error, but 14219 GNU as does not. 14220 14221 C and Objective-C 14222 14223 * The -Wstrict-aliasing=2 option has been added. This warning catches 14224 all unsafe cases, but it may also give a warning for some cases 14225 that are safe. 14226 * The cast-as-lvalue, conditional-expression-as-lvalue and 14227 compound-expression-as-lvalue extensions, which were deprecated in 14228 3.3.4 and 3.4, have been removed. 14229 * The -fwritable-strings option, which was deprecated in 3.4, has 14230 been removed. 14231 * #pragma pack() semantics have been brought closer to those used by 14232 other compilers. This also applies to C++. 14233 * Taking the address of a variable with register storage is invalid 14234 in C. GCC now issues an error instead of a warning. 14235 * Arrays of incomplete element type are invalid in C. GCC now issues 14236 an error for such arrays. Declarations such as extern struct s x[]; 14237 (where struct s has not been defined) can be moved after the 14238 definition of struct s. Function parameters declared as arrays of 14239 incomplete type can instead be declared as pointers. 14240 14241 C++ 14242 14243 * When compiling without optimizations (-O0), the C++ front end is 14244 much faster than in any previous versions of GCC. Independent 14245 testers have measured speed-ups up to 25% in real-world production 14246 code, compared to the 3.4 family (which was already the fastest 14247 version to date). Upgrading from older versions might show even 14248 bigger improvements. 14249 * ELF visibility attributes can now be applied to a class type, so 14250 that it affects every member function of a class at once, without 14251 having to specify each individually: 14252class __attribute__ ((visibility("hidden"))) Foo 14253{ 14254 int foo1(); 14255 void foo2(); 14256}; 14257 The syntax is deliberately similar to the __declspec() system used 14258 by Microsoft Windows based compilers, allowing cross-platform 14259 projects to easily reuse their existing macro system for denoting 14260 exports and imports. By explicitly marking internal classes never 14261 used outside a binary as hidden, one can completely avoid PLT 14262 indirection overheads during their usage by the compiler. You can 14263 find out more about the advantages of this at 14264 [6]https://www.akkadia.org/drepper/dsohowto.pdf 14265 * The -fvisibility-inlines-hidden option has been added which marks 14266 all inlineable functions as having hidden ELF visibility, thus 14267 removing their symbol and typeinfo from the exported symbol table 14268 of the output ELF binary. Using this option can reduce the exported 14269 symbol count of template-heavy code by up to 40% with no code 14270 change at all, thus notably improving link and load times for the 14271 binary as well as a reduction in size of up to 10%. Also, check the 14272 new [7]-fvisibility option. 14273 * The compiler now uses the library interface specified by the [8]C++ 14274 ABI for thread-safe initialization of function-scope static 14275 variables. Most users should leave this alone, but embedded 14276 programmers may want to disable this by specifying 14277 -fno-threadsafe-statics for a small savings in code size. 14278 * Taking the address of an explicit register variable is no longer 14279 supported. Note that C++ allows taking the address of variables 14280 with register storage so this will continue to compile with a 14281 warning. For example, assuming that r0 is a machine register: 14282register int foo asm ("r0"); 14283register int bar; 14284&foo; // error, no longer accepted 14285&bar; // OK, with a warning 14286 * G++ has an undocumented extension to virtual function covariancy 14287 rules that allowed the overrider to return a type that was 14288 implicitly convertable to the overridden function's return type. 14289 For instance a function returning void * could be overridden by a 14290 function returning T *. This is now deprecated and will be removed 14291 in a future release. 14292 * The G++ minimum and maximum operators (<? and >?) and their 14293 compound forms (<?=) and >?=) have been deprecated and will be 14294 removed in a future version. Code using these operators should be 14295 modified to use std::min and std::max instead. 14296 * Declaration of nested classes of class templates as friends are 14297 supported: 14298template <typename T> struct A { 14299 class B {}; 14300}; 14301class C { 14302 template <typename T> friend class A<T>::B; 14303}; 14304 This complements the feature member functions of class templates as 14305 friends introduced in GCC 3.4.0. 14306 * When declaring a friend class using an unqualified name, classes 14307 outside the innermost non-class scope are not searched: 14308class A; 14309namespace N { 14310 class B { 14311 friend class A; // Refer to N::A which has not been declared yet 14312 // because name outside namespace N are not searched 14313 friend class ::A; // Refer to ::A 14314 }; 14315} 14316 Hiding the friend name until declaration is still not implemented. 14317 * Friends of classes defined outside their namespace are correctly 14318 handled: 14319namespace N { 14320 class A; 14321} 14322class N::A { 14323 friend class B; // Refer to N::B in GCC 4.0.0 14324 // but ::B in earlier versions of GCC 14325}; 14326 14327 Runtime Library (libstdc++) 14328 14329 * Optimization work: 14330 + Added efficient specializations of istream functions for char 14331 and wchar_t. 14332 + Further performance tuning of strings, in particular wrt 14333 single-char append and getline. 14334 + iter_swap - and therefore most of the mutating algorithms - 14335 now makes an unqualified call to swap when the value_type of 14336 the two iterators is the same. 14337 * A large subset of the features in Technical Report 1 (TR1 for 14338 short) is experimentally delivered (i.e., no guarantees about the 14339 implementation are provided. In particular it is not promised that 14340 the library will remain link-compatible when code using TR1 is 14341 used): 14342 + General utilities such as reference_wrapper and shared_ptr. 14343 + Function objects, i.e., result_of, mem_fn, bind, function. 14344 + Support for metaprogramming. 14345 + New containers such as tuple, array, unordered_set, 14346 unordered_map, unordered_multiset, unordered_multimap. 14347 * As usual, many bugs have been fixed and LWG resolutions implemented 14348 for the first time (e.g., DR 409). 14349 14350 Java 14351 14352 * In order to prevent naming conflicts with other implementations of 14353 these tools, some GCJ binaries have been renamed: 14354 + rmic is now grmic, 14355 + rmiregistry is now grmiregistry, and 14356 + jar is now fastjar. 14357 In particular, these names were problematic for the jpackage.org 14358 packaging conventions which install symlinks in /usr/bin that point 14359 to the preferred versions of these tools. 14360 * The -findirect-dispatch argument to the compiler now works and 14361 generates code following a new "binary compatibility" ABI. Code 14362 compiled this way follows the binary compatibility rules of the 14363 Java Language Specification. 14364 * libgcj now has support for using GCJ as a JIT, using the 14365 gnu.gcj.jit family of system properties. 14366 * libgcj can now find a shared library corresponding to the bytecode 14367 representation of a class. See the documentation for the new 14368 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 14369 property. 14370 * There have been many improvements to the class library. Here are 14371 some highlights: 14372 + Much more of AWT and Swing exist. 14373 + Many new packages and classes were added, including 14374 java.util.regex, java.net.URI, javax.crypto, 14375 javax.crypto.interfaces, javax.crypto.spec, javax.net, 14376 javax.net.ssl, javax.security.auth, 14377 javax.security.auth.callback, javax.security.auth.login, 14378 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 14379 javax.imageio, javax.imageio.event, javax.imageio.spi, 14380 javax.print, javax.print.attribute, 14381 javax.print.attribute.standard, javax.print.event, and 14382 javax.xml 14383 + Updated SAX and DOM, and imported GNU JAXP 14384 14385 Fortran 14386 14387 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 14388 front end. The new front end supports Fortran 90 and Fortran 95. It 14389 may not yet be as stable as the old Fortran front end. 14390 14391 Ada 14392 14393 * Ada (with tasking and Zero Cost Exceptions) is now available on 14394 many more targets, including but not limited to: alpha-linux, 14395 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 14396 s390x-linux, sparc-linux. 14397 * Some of the new Ada 2005 features are now implemented like 14398 Wide_Wide_Character and Ada.Containers. 14399 * Many bugs have been fixed, tools and documentation improved. 14400 * To compile Ada from the sources, install an older working Ada 14401 compiler and then use --enable-languages=ada at configuration time, 14402 since the Ada front end is not currently activated by default. See 14403 the [10]Installing GCC for details. 14404 14405New Targets and Target Specific Improvements 14406 14407 H8/300 14408 14409 * The frame layout has changed. In the new layout, the prologue of a 14410 function first saves registers and then allocate space for locals, 14411 resulting in an 1% improvement on code size. 14412 14413 IA-32/x86-64 (AMD64) 14414 14415 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 14416 log1p, log2, logb and tan mathematical builtins (and their float 14417 and long double variants) are now implemented as inline x87 14418 intrinsics when using -ffast-math. 14419 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 14420 (and their float and long double variants) are now implemented as 14421 inline x87 intrinsics when using -ffast-math. 14422 * The x87's fsincos instruction is now used automatically with 14423 -ffast-math when calculating both the sin and cos of the same 14424 argument. 14425 * Instruction selection for multiplication and division by constants 14426 has been improved. 14427 14428 IA-64 14429 14430 * Floating point division, integer division and sqrt are now inlined, 14431 resulting in significant performance improvements on some codes. 14432 14433 MIPS 14434 14435 * Division by zero checks now use conditional traps if the target 14436 processor supports them. This decreases code size by one word per 14437 division operation. The old behavior (branch and break) can be 14438 obtained either at configure time by passing --with-divide=breaks 14439 to configure or at runtime by passing -mdivide-breaks to GCC. 14440 * Support for MIPS64 paired-single instructions has been added. It is 14441 enabled by -mpaired-single and can be accessed using both the 14442 target-independent vector extensions and new MIPS-specific built-in 14443 functions. 14444 * Support for the MIPS-3D ASE has been added. It is enabled by 14445 -mips3d and provides new MIPS-3D-specific built-in functions. 14446 * The -mexplicit-relocs option now supports static n64 code (as is 14447 used, for example, in 64-bit linux kernels). -mexplicit-relocs 14448 should now be feature-complete and is enabled by default when GCC 14449 is configured to use a compatible assembler. 14450 * Support for the NEC VR4130 series has been added. This support 14451 includes the use of VR-specific instructions and a new VR4130 14452 scheduler. Full VR4130 support can be selected with -march=vr4130 14453 while code for any ISA can be tuned for the VR4130 using 14454 -mtune=vr4130. There is also a new -mvr4130-align option that 14455 produces better schedules at the cost of increased code size. 14456 * Support for the Broadcom SB-1 has been extended. There is now an 14457 SB-1 scheduler as well as support for the SB-1-specific 14458 paired-single instructions. Full SB-1 support can be selected with 14459 -march=sb1 while code for any ISA can be optimized for the SB-1 14460 using -mtune=sb1. 14461 * The compiler can now work around errata in R4000, R4400, VR4120 and 14462 VR4130 processors. These workarounds are enabled by -mfix-r4000, 14463 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 14464 and VR4130 workarounds need binutils 2.16 or above. 14465 * IRIX shared libraries are now installed into the standard library 14466 directories: o32 libraries go into lib/, n32 libraries go into 14467 lib32/ and n64 libraries go into lib64/. 14468 * The compiler supports a new -msym32 option. It can be used to 14469 optimize n64 code in which all symbols are known to have 32-bit 14470 values. 14471 14472 S/390 and zSeries 14473 14474 * New command-line options help to generate code intended to run in 14475 an environment where stack space is restricted, e.g. Linux kernel 14476 code: 14477 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 14478 warnings for single functions that require large or dynamic 14479 stack frames. 14480 + -mstack-size and -mstack-guard generate code that checks for 14481 stack overflow at run time. 14482 + -mpacked-stack generates code that reduces the stack frame 14483 size of many functions by reusing unneeded parts of the stack 14484 bias area. 14485 * The -msoft-float option now ensures that generated code never 14486 accesses floating point registers. 14487 * The s390x-ibm-tpf target now fully supports C++, including 14488 exceptions and threads. 14489 * Various changes to improve performance of the generated code have 14490 been implemented, including: 14491 + GCC now uses sibling calls where possible. 14492 + Condition code handling has been optimized, allowing GCC to 14493 omit redundant comparisons in certain cases. 14494 + The cost function guiding many optimizations has been refined 14495 to more accurately represent the z900 and z990 processors. 14496 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 14497 instructions are now used to avoid conditional branches in 14498 certain cases. 14499 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 14500 optimize address arithmetic required to access large stack 14501 frames. 14502 + GCC now makes more efficient use of memory-to-memory type 14503 instructions (MVC, CLC, ...). 14504 + More precise tracking of special register use allows better 14505 instruction scheduling, in particular of the function prologue 14506 and epilogue sequences. 14507 + The Java front end now generates inline code to implement 14508 integer division, instead of calling library routines. 14509 14510 SPARC 14511 14512 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 14513 -mf934 have been removed. They have been replaced with -mcpu=xxx. 14514 * The internal model used to estimate the relative cost of each 14515 instruction has been updated. It is expected to give better results 14516 on recent UltraSPARC processors. 14517 * Code generation for function prologues and epilogues has been 14518 improved, resulting in better scheduling and allowing multiple exit 14519 points in functions. 14520 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 14521 It is enabled by -mvis and provides new built-in functions for VIS 14522 instructions on UltraSPARC processors. 14523 * The option -mapp-regs has been turned on by default on Solaris too. 14524 14525 NetWare 14526 14527 * Novell NetWare (on ix86, no other hardware platform was ever really 14528 supported by this OS) has been re-enabled and the ABI supported by 14529 GCC has been brought into sync with that of MetroWerks CodeWarrior 14530 (the ABI previously supported was that of some Unix systems, which 14531 NetWare never tried to support). 14532 14533Obsolete Systems 14534 14535 Support for a number of older systems has been declared obsolete in GCC 14536 4.0. Unless there is activity to revive them, the next release of GCC 14537 will have their sources permanently removed. 14538 14539 All GCC ports for the following processor architectures have been 14540 declared obsolete: 14541 * Intel i860 14542 * Ubicom IP2022 14543 * National Semiconductor NS32K (ns32k) 14544 * Texas Instruments TMS320C[34]x 14545 14546 Also, those for some individual systems have been obsoleted: 14547 * SPARC family 14548 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 14549 sparc86x-*-elf) 14550 + OpenBSD 32-bit (sparc-*-openbsd*) 14551 14552Documentation improvements 14553 14554Other significant improvements 14555 14556 * Location lists are now generated by default when compiling with 14557 debug info and optimization. Location lists provide more accurate 14558 debug info about locations of variables and they allow debugging 14559 code compiled with -fomit-frame-pointer. 14560 * The -fvisibility option has been added which allows the default ELF 14561 visibility of all symbols to be set per compilation and the new 14562 #pragma GCC visibility preprocessor command allows the setting of 14563 default ELF visibility for a region of code. Using 14564 -fvisibility=hidden especially in combination with the new 14565 -fvisibility-inlines-hidden can yield substantial improvements in 14566 output binary quality including avoiding PLT indirection overheads, 14567 reduction of the exported symbol count by up to 60% (with resultant 14568 improvements to link and load times), better scope for the 14569 optimizer to improve code and up to a 20% reduction in binary size. 14570 Using these options correctly yields a binary with a similar symbol 14571 count to a Windows DLL. 14572 Perhaps more importantly, this new feature finally allows (with 14573 careful planning) complete avoidance of symbol clashes when 14574 manually loading shared objects with RTLD_GLOBAL, thus finally 14575 solving problems many projects such as python were forced to use 14576 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 14577 can find more information about using these options at 14578 [11]https://gcc.gnu.org/wiki/Visibility. 14579 __________________________________________________________________ 14580 14581GCC 4.0.1 14582 14583 This is the [12]list of problem reports (PRs) from GCC's bug tracking 14584 system that are known to be fixed in the 4.0.1 release. This list might 14585 not be complete (that is, it is possible that some PRs that have been 14586 fixed are not listed here). 14587 14588GCC 4.0.2 14589 14590 This is the [13]list of problem reports (PRs) from GCC's bug tracking 14591 system that are known to be fixed in the 4.0.2 release. This list might 14592 not be complete (that is, it is possible that some PRs that have been 14593 fixed are not listed here). 14594 14595 Unfortunately, due to a release engineering failure, this release has a 14596 regression on Solaris that will affect some C++ programs. We suggest 14597 that Solaris users apply a [14]patch that corrects the problem. Users 14598 who do not wish to apply the patch should explicitly link C++ programs 14599 with the -pthreads option, even if they do not use threads. This 14600 problem has been corrected in the current 4.0 branch sources and will 14601 not be present in GCC 4.0.3. 14602 14603GCC 4.0.3 14604 14605 Starting with this release, the function getcontext is recognized by 14606 the compiler as having the same semantics as the setjmp function. In 14607 particular, the compiler will ensure that all registers are dead before 14608 calling such a function and will emit a warning about the variables 14609 that may be clobbered after the second return from the function. 14610 14611GCC 4.0.4 14612 14613 This is the [15]list of problem reports (PRs) from GCC's bug tracking 14614 system that are known to be fixed in the 4.0.4 release. This list might 14615 not be complete (that is, it is possible that some PRs that have been 14616 fixed are not listed here). 14617 14618 The 4.0.4 release is provided for those that require a high degree of 14619 binary compatibility with previous 4.0.x releases. For most users, the 14620 GCC team recommends that version 4.1.1 or later be used instead." 14621 14622 14623 For questions related to the use of GCC, please consult these web 14624 pages and the [16]GCC manuals. If that fails, the 14625 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14626 web pages and the development of GCC are welcome on our developer 14627 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public 14628 archives. 14629 14630 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 14631 distribution of this entire article is permitted in any medium, 14632 provided this notice is preserved. 14633 14634 These pages are [21]maintained by the GCC team. Last modified 14635 2021-07-28[22]. 14636 14637References 14638 14639 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 14640 2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 14641 3. http://gcc.gnu.org/projects/tree-ssa/ 14642 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html 14643 5. http://gcc.gnu.org/news/sms.html 14644 6. https://www.akkadia.org/drepper/dsohowto.pdf 14645 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility 14646 8. https://itanium-cxx-abi.github.io/cxx-abi/ 14647 9. http://gcc.gnu.org/fortran/ 14648 10. https://gcc.gnu.org/install/ 14649 11. https://gcc.gnu.org/wiki/Visibility 14650 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 14651 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 14652 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 14653 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 14654 16. https://gcc.gnu.org/onlinedocs/ 14655 17. mailto:gcc-help@gcc.gnu.org 14656 18. mailto:gcc@gcc.gnu.org 14657 19. https://gcc.gnu.org/lists.html 14658 20. https://www.fsf.org/ 14659 21. https://gcc.gnu.org/about.html 14660 22. http://validator.w3.org/check/referer 14661====================================================================== 14662http://gcc.gnu.org/gcc-3.4/index.html 14663 14664 GCC 3.4 Release Series 14665 14666 (This release series is no longer supported.) 14667 14668 May 26, 2006 14669 14670 The [1]GNU project and the GCC developers are pleased to announce the 14671 release of GCC 3.4.6. 14672 14673 This release is a bug-fix release, containing fixes for regressions in 14674 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 14675 3.4.x series. 14676 14677 The GCC 3.4 release series includes numerous [2]new features, 14678 improvements, bug fixes, and other changes, thanks to an [3]amazing 14679 group of volunteers. 14680 14681Release History 14682 14683 GCC 3.4.6 14684 March 6, 2006 ([4]changes) 14685 14686 GCC 3.4.5 14687 November 30, 2005 ([5]changes) 14688 14689 GCC 3.4.4 14690 May 18, 2005 ([6]changes) 14691 14692 GCC 3.4.3 14693 November 4, 2004 ([7]changes) 14694 14695 GCC 3.4.2 14696 September 6, 2004 ([8]changes) 14697 14698 GCC 3.4.1 14699 July 1, 2004 ([9]changes) 14700 14701 GCC 3.4.0 14702 April 18, 2004 ([10]changes) 14703 14704References and Acknowledgements 14705 14706 GCC used to stand for the GNU C Compiler, but since the compiler 14707 supports several other languages aside from C, it now stands for the 14708 GNU Compiler Collection. 14709 14710 A list of [11]successful builds is updated as new information becomes 14711 available. 14712 14713 The GCC developers would like to thank the numerous people that have 14714 contributed new features, improvements, bug fixes, and other changes as 14715 well as test results to GCC. This [12]amazing group of volunteers is 14716 what makes GCC successful. 14717 14718 For additional information about GCC please refer to the [13]GCC 14719 project web site or contact the [14]GCC development mailing list. 14720 14721 To obtain GCC please use [15]our mirror sites, or [16]our version 14722 control system. 14723 14724 14725 For questions related to the use of GCC, please consult these web 14726 pages and the [17]GCC manuals. If that fails, the 14727 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these 14728 web pages and the development of GCC are welcome on our developer 14729 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public 14730 archives. 14731 14732 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 14733 distribution of this entire article is permitted in any medium, 14734 provided this notice is preserved. 14735 14736 These pages are [22]maintained by the GCC team. Last modified 14737 2021-07-28[23]. 14738 14739References 14740 14741 1. http://www.gnu.org/ 14742 2. http://gcc.gnu.org/gcc-3.4/changes.html 14743 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14744 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 14745 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 14746 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 14747 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 14748 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 14749 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 14750 10. http://gcc.gnu.org/gcc-3.4/changes.html 14751 11. http://gcc.gnu.org/gcc-3.4/buildstat.html 14752 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14753 13. http://gcc.gnu.org/index.html 14754 14. mailto:gcc@gcc.gnu.org 14755 15. http://gcc.gnu.org/mirrors.html 14756 16. http://gcc.gnu.org/git.html 14757 17. https://gcc.gnu.org/onlinedocs/ 14758 18. mailto:gcc-help@gcc.gnu.org 14759 19. mailto:gcc@gcc.gnu.org 14760 20. https://gcc.gnu.org/lists.html 14761 21. https://www.fsf.org/ 14762 22. https://gcc.gnu.org/about.html 14763 23. http://validator.w3.org/check/referer 14764====================================================================== 14765http://gcc.gnu.org/gcc-3.4/changes.html 14766 14767 GCC 3.4 Release Series 14768 Changes, New Features, and Fixes 14769 14770 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 14771 is now closed. 14772 14773 GCC 3.4 has [2]many improvements in the C++ front end. Before reporting 14774 a bug, please make sure it's really GCC, and not your code, that is 14775 broken. 14776 14777Caveats 14778 14779 * GNU Make is now required to build GCC. 14780 * With -nostdinc the preprocessor used to ignore both standard 14781 include paths and include paths contained in environment variables. 14782 It was neither documented nor intended that environment variable 14783 paths be ignored, so this has been corrected. 14784 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 14785 -fvolatile-static. It is unlikely that they worked correctly in any 14786 3.x release. 14787 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 14788 * Support for all the systems [3]obsoleted in GCC 3.3 has been 14789 removed from GCC 3.4. See below for a [4]list of systems which are 14790 obsoleted in this release. 14791 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 14792 compilers will not work. 14793 * The implementation of the [5]MIPS ABIs has changed. As a result, 14794 the code generated for certain MIPS targets will not be binary 14795 compatible with earlier releases. 14796 * In previous releases, the MIPS port had a fake "hilo" register with 14797 the user-visible name accum. This register has been removed. 14798 * The implementation of the [6]SPARC ABIs has changed. As a result, 14799 the code generated will not be binary compatible with earlier 14800 releases in certain cases. 14801 * The configure option --enable-threads=pthreads has been removed; 14802 use --enable-threads=posix instead, which should have the same 14803 effect. 14804 * Code size estimates used by inlining heuristics for C, Objective-C, 14805 C++ and Java have been redesigned significantly. As a result the 14806 parameters of -finline-insns, --param max-inline-insns-single and 14807 --param max-inline-insns-auto need to be reconsidered. 14808 * --param max-inline-slope and --param min-inline-insns have been 14809 removed; they are not needed for the new bottom-up inlining 14810 heuristics. 14811 * The new unit-at-a-time compilation scheme has several compatibility 14812 issues: 14813 + The order in which functions, variables, and top-level asm 14814 statements are emitted may have changed. Code relying on some 14815 particular ordering needs to be updated. The majority of such 14816 top-level asm statements can be replaced by section 14817 attributes. 14818 + Unreferenced static variables and functions are removed. This 14819 may result in undefined references when an asm statement 14820 refers to the variable/function directly. In that case either 14821 the variable/function shall be listed in asm statement operand 14822 or in the case of top-level asm statements the attribute used 14823 shall be used to force function/variable to be always output 14824 and considered as a possibly used by unknown code. 14825 For variables the attribute is accepted only by GCC 3.4 and 14826 newer, while for earlier versions it is sufficient to use 14827 unused to silence warnings about the variables not being 14828 referenced. To keep code portable across different GCC 14829 versions, you can use appropriate preprocessor conditionals. 14830 + Static functions now can use non-standard passing conventions 14831 that may break asm statements calling functions directly. 14832 Again the attribute used shall be used to prevent this 14833 behavior. 14834 As a temporary workaround, -fno-unit-at-a-time can be used, but 14835 this scheme may not be supported by future releases of GCC. 14836 * GCC 3.4 automatically places zero-initialized variables in the .bss 14837 section on some operating systems. Versions of GNU Emacs up to (and 14838 including) 21.3 will not work correctly when using this 14839 optimization; you can use -fno-zero-initialized-in-bss to disable 14840 it. 14841 * If GCC 3.4 is configured with --enable-threads=posix (the default 14842 on most targets that support pthreads) then _REENTRANT will be 14843 defined unconditionally by some libstdc++ headers. C++ code which 14844 relies on that macro to detect whether multi-threaded code is being 14845 compiled might change in meaning, possibly resulting in linker 14846 errors for single-threaded programs. Affected users of [7]Boost 14847 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 14848 See Bugzilla for [8]more information. 14849 14850General Optimizer Improvements 14851 14852 * Usability of the profile feedback and coverage testing has been 14853 improved. 14854 + Performance of profiled programs has been improved by faster 14855 profile merging code. 14856 + Better use of the profile feedback for optimization (loop 14857 unrolling and loop peeling). 14858 + File locking support allowing fork() calls and parallel runs 14859 of profiled programs. 14860 + Coverage file format has been redesigned. 14861 + gcov coverage tool has been improved. 14862 + make profiledbootstrap available to build a faster compiler. 14863 Experiments made on i386 hardware showed an 11% speedup on -O0 14864 and a 7.5% speedup on -O2 compilation of a [9]large C++ 14865 testcase. 14866 + New value profiling pass enabled via -fprofile-values 14867 + New value profile transformations pass enabled via -fvpt aims 14868 to optimize some code sequences by exploiting knowledge about 14869 value ranges or other properties of the operands. At the 14870 moment a conversion of expensive divisions into cheaper 14871 operations has been implemented. 14872 + New -fprofile-generate and -fprofile-use command-line options 14873 to simplify the use of profile feedback. 14874 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 14875 Java which is enabled via -funit-at-a-time (and implied by -O2). In 14876 this scheme a whole file is parsed first and optimized later. The 14877 following basic inter-procedural optimizations are implemented: 14878 + Removal of unreachable functions and variables 14879 + Discovery of local functions (functions with static linkage 14880 whose address is never taken) 14881 + On i386, these local functions use register parameter passing 14882 conventions. 14883 + Reordering of functions in topological order of the call graph 14884 to enable better propagation of optimizing hints (such as the 14885 stack alignments needed by functions) in the back end. 14886 + Call graph based out-of-order inlining heuristics which allows 14887 to limit overall compilation unit growth (--param 14888 inline-unit-growth). 14889 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 14890 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 14891 CPU). 14892 * More realistic code size estimates used by inlining for C, 14893 Objective-C, C++ and Java. The growth of large functions can now be 14894 limited via --param large-function-insns and --param 14895 large-function-growth. 14896 * A new cfg-level loop optimizer pass replaces the old loop unrolling 14897 pass and adds two other loop transformations -- loop peeling and 14898 loop unswitching -- and also uses the profile feedback to limit 14899 code growth. (The three optimizations are enabled by 14900 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 14901 respectively). 14902 The old loop unroller still can be enabled by -fold-unroll-loops 14903 and may produce better code in some cases, especially when the 14904 webizer optimization pass is not run. 14905 * A new web construction pass enabled via -fweb (and implied by -O3) 14906 improves the quality of register allocation, CSE, first scheduling 14907 pass and some other optimization passes by avoiding re-use of 14908 pseudo registers with non-overlapping live ranges. The pass almost 14909 always improves code quality but does make debugging difficult and 14910 thus is not enabled by default by -O2 14911 The pass is especially effective as cleanup after code duplication 14912 passes, such as the loop unroller or the tracer. 14913 * Experimental implementations of superblock or trace scheduling in 14914 the second scheduling pass can be enabled via 14915 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 14916 14917New Languages and Language specific improvements 14918 14919 Ada 14920 14921 * The Ada front end has been updated to include numerous bug fixes 14922 and enhancements. These include: 14923 + Improved project file support 14924 + Additional set of warnings about potential wrong code 14925 + Improved error messages 14926 + Improved code generation 14927 + Improved cross reference information 14928 + Improved inlining 14929 + Better run-time check elimination 14930 + Better error recovery 14931 + More efficient implementation of unbounded strings 14932 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 14933 ... 14934 + New GNAT.xxxx packages (e.g. GNAT.Strings, 14935 GNAT.Exception_Action) 14936 + New pragmas 14937 + New -gnatS switch replacing gnatpsta 14938 + Implementation of new Ada features (in particular limited 14939 with, limited aggregates) 14940 14941 C/Objective-C/C++ 14942 14943 * Precompiled headers are now supported. Precompiled headers can 14944 dramatically speed up compilation of some projects. There are some 14945 known defects in the current precompiled header implementation that 14946 will result in compiler crashes in relatively rare situations. 14947 Therefore, precompiled headers should be considered a "technology 14948 preview" in this release. Read the manual for details about how to 14949 use precompiled headers. 14950 * File handling in the preprocessor has been rewritten. GCC no longer 14951 gets confused by symlinks and hardlinks, and now has a correct 14952 implementation of #import and #pragma once. These two directives 14953 have therefore been un-deprecated. 14954 * The undocumented extension that allowed C programs to have a label 14955 at the end of a compound statement, which has been deprecated since 14956 GCC 3.0, has been removed. 14957 * The cast-as-lvalue extension has been removed for C++ and 14958 deprecated for C and Objective-C. In particular, code like this: 14959 int i; 14960 (char) i = 5; 14961 14962 or this: 14963 char *p; 14964 ((int *) p)++; 14965 14966 is no longer accepted for C++ and will not be accepted for C and 14967 Objective-C in a future version. 14968 * The conditional-expression-as-lvalue extension has been deprecated 14969 for C and Objective-C. In particular, code like this: 14970 int a, b, c; 14971 (a ? b : c) = 2; 14972 14973 will not be accepted for C and Objective-C in a future version. 14974 * The compound-expression-as-lvalue extension has been deprecated for 14975 C and Objective-C. In particular, code like this: 14976 int a, b; 14977 (a, b) = 2; 14978 14979 will not be accepted for C and Objective-C in a future version. A 14980 possible non-intrusive workaround is the following: 14981 (*(a, &b)) = 2; 14982 14983 * Several [10]built-in functions such as __builtin_popcount for 14984 counting bits, finding the highest and lowest bit in a word, and 14985 parity have been added. 14986 * The -fwritable-strings option has been deprecated and will be 14987 removed. 14988 * Many C math library functions are now recognized as built-ins and 14989 optimized. 14990 * The C, C++, and Objective-C compilers can now handle source files 14991 written in any character encoding supported by the host C library. 14992 The default input character set is taken from the current locale, 14993 and may be overridden with the -finput-charset command line option. 14994 In the future we will add support for inline encoding markers. 14995 14996 C++ 14997 14998 * G++ is now much closer to full conformance to the ISO/ANSI C++ 14999 standard. This means, among other things, that a lot of invalid 15000 constructs which used to be accepted in previous versions will now 15001 be rejected. It is very likely that existing C++ code will need to 15002 be fixed. This document lists some of the most common issues. 15003 * A hand-written recursive-descent C++ parser has replaced the 15004 YACC-derived C++ parser from previous GCC releases. The new parser 15005 contains much improved infrastructure needed for better parsing of 15006 C++ source codes, handling of extensions, and clean separation 15007 (where possible) between proper semantics analysis and parsing. The 15008 new parser fixes many bugs that were found in the old parser. 15009 * You must now use the typename and template keywords to disambiguate 15010 dependent names, as required by the C++ standard. 15011 struct K { 15012 typedef int mytype_t; 15013 }; 15014 15015 template <class T1> struct A { 15016 template <class T2> struct B { 15017 void callme(void); 15018 }; 15019 15020 template <int N> void bar(void) 15021 { 15022 // Use 'typename' to tell the parser that T1::mytype_t names 15023 // a type. This is needed because the name is dependent (in 15024 // this case, on template parameter T1). 15025 typename T1::mytype_t x; 15026 x = 0; 15027 } 15028 }; 15029 15030 template <class T> void template_func(void) 15031 { 15032 // Use 'template' to prefix member templates within 15033 // dependent types (a has type A<T>, which depends on 15034 // the template parameter T). 15035 A<T> a; 15036 a.template bar<0>(); 15037 15038 // Use 'template' to tell the parser that B is a nested 15039 // template class (dependent on template parameter T), and 15040 // 'typename' because the whole A<T>::B<int> is 15041 // the name of a type (again, dependent). 15042 typename A<T>::template B<int> b; 15043 b.callme(); 15044 } 15045 15046 void non_template_func(void) 15047 { 15048 // Outside of any template class or function, no names can be 15049 // dependent, so the use of the keyword 'typename' and 'template' 15050 // is not needed (and actually forbidden). 15051 A<K> a; 15052 a.bar<0>(); 15053 A<K>::B<float> b; 15054 b.callme(); 15055 } 15056 * In a template definition, unqualified names will no longer find 15057 members of a dependent base (as specified by [temp.dep]/3 in the 15058 C++ standard). For example, 15059 template <typename T> struct B { 15060 int m; 15061 int n; 15062 int f (); 15063 int g (); 15064 }; 15065 int n; 15066 int g (); 15067 template <typename T> struct C : B<T> { 15068 void h () 15069 { 15070 m = 0; // error 15071 f (); // error 15072 n = 0; // ::n is modified 15073 g (); // ::g is called 15074 } 15075 }; 15076 You must make the names dependent, e.g. by prefixing them with 15077 this->. Here is the corrected definition of C<T>::h, 15078 template <typename T> void C<T>::h () 15079 { 15080 this->m = 0; 15081 this->f (); 15082 this->n = 0 15083 this->g (); 15084 } 15085 As an alternative solution (unfortunately not backwards compatible 15086 with GCC 3.3), you may use using declarations instead of this->: 15087 template <typename T> struct C : B<T> { 15088 using B<T>::m; 15089 using B<T>::f; 15090 using B<T>::n; 15091 using B<T>::g; 15092 void h () 15093 { 15094 m = 0; 15095 f (); 15096 n = 0; 15097 g (); 15098 } 15099 }; 15100 * In templates, all non-dependent names are now looked up and bound 15101 at definition time (while parsing the code), instead of later when 15102 the template is instantiated. For instance: 15103 void foo(int); 15104 15105 template <int> struct A { 15106 static void bar(void){ 15107 foo('a'); 15108 } 15109 }; 15110 15111 void foo(char); 15112 15113 int main() 15114 { 15115 A<0>::bar(); // Calls foo(int), used to call foo(char). 15116 } 15117 15118 * In an explicit instantiation of a class template, you must use 15119 class or struct before the template-id: 15120 template <int N> 15121 class A {}; 15122 15123 template A<0>; // error, not accepted anymore 15124 template class A<0>; // OK 15125 * The "named return value" and "implicit typename" extensions have 15126 been removed. 15127 * Default arguments in function types have been deprecated and will 15128 be removed. 15129 * ARM-style name-injection of friend declarations has been deprecated 15130 and will be removed. For example: struct S { friend void f(); }; 15131 void g() { f(); } will not be accepted by future versions of G++; 15132 instead a declaration of "f" will need to be present outside of the 15133 scope of "S". 15134 * Covariant returns are implemented for all but varadic functions 15135 that require an adjustment. 15136 * When -pedantic is used, G++ now issues errors about spurious 15137 semicolons. For example, 15138 namespace N {}; // Invalid semicolon. 15139 void f() {}; // Invalid semicolon. 15140 * G++ no longer accepts attributes for a declarator after the 15141 initializer associated with that declarator. For example, 15142 X x(1) __attribute__((...)); 15143 is no longer accepted. Instead, use: 15144 X x __attribute__((...)) (1); 15145 * Inside the scope of a template class, the name of the class itself 15146 can be treated as either a class or a template. So GCC used to 15147 accept the class name as argument of type template, and template 15148 template parameter. However this is not C++ standard compliant. Now 15149 the name is not treated as a valid template template argument 15150 unless you qualify the name by its scope. For example, the code 15151 below no longer compiles. 15152 template <template <class> class TT> class X {}; 15153 template <class T> class Y { 15154 X<Y> x; // Invalid, Y is always a type template parameter. 15155 }; 15156 The valid code for the above example is 15157 X< ::Y> x; // Valid. 15158 (Notice the space between < and : to prevent GCC to interpret this 15159 as a digraph for [.) 15160 * Friend declarations that refer to template specializations are 15161 rejected if the template has not already been declared. For 15162 example, 15163 template <typename T> 15164 class C { 15165 friend void f<> (C&); 15166 }; 15167 is rejected. You must first declare f as a template, 15168 template <typename T> 15169 void f(T); 15170 * In case of friend declarations, every name used in the friend 15171 declaration must be accessible at the point of that declaration. 15172 Previous versions of G++ used to be less strict about this and 15173 allowed friend declarations for private class members, for example. 15174 See the ISO C++ Standard Committee's [11]defect report #209 for 15175 details. 15176 * Declaration of member functions of class templates as friends are 15177 supported. For example, 15178 template <typename T> struct A { 15179 void f(); 15180 }; 15181 class C { 15182 template <typename T> friend void A<T>::f(); 15183 }; 15184 * You must use template <> to introduce template specializations, as 15185 required by the standard. For example, 15186 template <typename T> 15187 struct S; 15188 15189 struct S<int> { }; 15190 is rejected. You must write, 15191 template <> struct S<int> {}; 15192 * G++ used to accept code like this, 15193 struct S { 15194 int h(); 15195 void f(int i = g()); 15196 int g(int i = h()); 15197 }; 15198 This behavior is not mandated by the standard. Now G++ issues an 15199 error about this code. To avoid the error, you must move the 15200 declaration of g before the declaration of f. The default arguments 15201 for g must be visible at the point where it is called. 15202 * The C++ ABI Section 3.3.3 specifications for the array construction 15203 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 15204 NULL when the allocator argument returns NULL. These changes are 15205 incorporated into the libstdc++ runtime library. 15206 * Using a name introduced by a typedef in a friend declaration or in 15207 an explicit instantiation is now rejected, as specified by the ISO 15208 C++ standard. 15209 class A; 15210 typedef A B; 15211 class C { 15212 friend class B; // error, no typedef name here 15213 friend B; // error, friend always needs class/struct/enum 15214 friend class A; // OK 15215 }; 15216 15217 template <int> class Q {}; 15218 typedef Q<0> R; 15219 template class R; // error, no typedef name here 15220 template class Q<0>; // OK 15221 * When allocating an array with a new expression, GCC used to allow 15222 parentheses around the type name. This is actually ill-formed and 15223 it is now rejected: 15224 int* a = new (int)[10]; // error, not accepted anymore 15225 int* a = new int[10]; // OK 15226 * When binding an rvalue of class type to a reference, the copy 15227 constructor of the class must be accessible. For instance, consider 15228 the following code: 15229 class A 15230 { 15231 public: 15232 A(); 15233 15234 private: 15235 A(const A&); // private copy ctor 15236 }; 15237 15238 A makeA(void); 15239 void foo(const A&); 15240 15241 void bar(void) 15242 { 15243 foo(A()); // error, copy ctor is not accessible 15244 foo(makeA()); // error, copy ctor is not accessible 15245 15246 A a1; 15247 foo(a1); // OK, a1 is a lvalue 15248 } 15249 This might be surprising at first sight, especially since most 15250 popular compilers do not correctly implement this rule ([12]further 15251 details). 15252 * When forming a pointer to member or a pointer to member function, 15253 access checks for class visibility (public, protected, private) are 15254 now performed using the qualifying scope of the name itself. This 15255 is better explained with an example: 15256 class A 15257 { 15258 public: 15259 void pub_func(); 15260 protected: 15261 void prot_func(); 15262 private: 15263 void priv_func(); 15264 }; 15265 15266 class B : public A 15267 { 15268 public: 15269 void foo() 15270 { 15271 &A::pub_func; // OK, pub_func is accessible through A 15272 &A::prot_func; // error, cannot access prot_func through A 15273 &A::priv_func; // error, cannot access priv_func through A 15274 15275 &B::pub_func; // OK, pub_func is accessible through B 15276 &B::prot_func; // OK, can access prot_func through B (within B) 15277 &B::priv_func; // error, cannot access priv_func through B 15278 } 15279 }; 15280 15281 Runtime Library (libstdc++) 15282 15283 * Optimization work: 15284 + Streamlined streambuf, filebuf, separate synched with C 15285 Standard I/O streambuf. 15286 + All formatted I/O now uses cached locale information. 15287 + STL optimizations (memory/speed for list, red-black trees as 15288 used by sets and maps). 15289 + More use of GCC builtins. 15290 + String optimizations (avoid contention on 15291 increment/decrement-and-test of the reference count in the 15292 empty-string object, constructor from input_iterators 15293 speedup). 15294 * Static linkage size reductions. 15295 * Large File Support (files larger than 2 GB on 32-bit systems). 15296 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 15297 * Generic character traits. 15298 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 15299 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 15300 * The allocator class is now standard-conformant, and two additional 15301 extension allocators have been added, mt_alloc and 15302 bitmap_allocator. 15303 * PCH support: -include bits/stdc++.h (2x compile speedup). 15304 * Rewrote __cxa_demangle with support for C++ style allocators. 15305 * New debug modes for STL containers and iterators. 15306 * Testsuite rewrite: five times as many tests, plus increasingly 15307 sophisticated tests, including I/O, MT, multi-locale, wide and 15308 narrow characters. 15309 * Use current versions of GNU "autotools" for build/configuration. 15310 15311 Objective-C 15312 15313 * The Objective-C front end has been updated to include the numerous 15314 bug fixes and enhancements previously available only in Apple's 15315 version of GCC. These include: 15316 + Structured exception (@try... @catch... @finally, @throw) and 15317 synchronization (@synchronized) support. These are accessible 15318 via the -fobjc-exceptions switch; as of this writing, they may 15319 only be used in conjunction with -fnext-runtime on Mac OS X 15320 10.3 and later. See [13]Options Controlling Objective-C 15321 Dialect for more information. 15322 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 15323 may now be encoded as 'B'. In addition, the back-end/codegen 15324 dependencies have been removed. 15325 + An overhaul of message dispatch construction, ensuring that 15326 the various receiver types (and casts thereof) are handled 15327 properly, and that correct diagnostics are issued. 15328 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 15329 (-freplace-objc-classes) debugging modes, currently available 15330 on Mac OS X 10.3 and later. See [14]Options Controlling 15331 Objective-C Dialect for more information. 15332 + Access to optimized runtime entry points (-fno-nil-receivers ) 15333 on the assumption that message receivers are never nil. This 15334 is currently available on Mac OS X 10.3 and later. See 15335 [15]Options Controlling Objective-C Dialect for more 15336 information. 15337 15338 Java 15339 15340 * Compiling a .jar file will now cause non-.class entries to be 15341 automatically compiled as resources. 15342 * libgcj has been ported to Darwin. 15343 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 15344 to gcj. 15345 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 15346 code from shared libraries. 15347 * libgcj has been much more completely merged with [16]GNU Classpath. 15348 * Class loading is now much more correct; in particular the caller's 15349 class loader is now used when that is required. 15350 * [17]Eclipse 2.x will run out of the box using gij. 15351 * Parts of java.nio have been implemented. Direct and indirect 15352 buffers work, as do fundamental file and socket operations. 15353 * java.awt has been improved, though it is still not ready for 15354 general use. 15355 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 15356 method. 15357 * The MinGW port has matured. Enhancements include socket timeout 15358 support, thread interruption, improved Runtime.exec() handling and 15359 support for accented characters in filenames. 15360 15361 Fortran 15362 15363 * Fortran improvements are listed in the [18]Fortran documentation. 15364 15365New Targets and Target Specific Improvements 15366 15367 Alpha 15368 15369 * Several [19]built-in functions have been added such as 15370 __builtin_alpha_zap to allow utilizing the more obscure 15371 instructions of the CPU. 15372 * Parameter passing of complex arguments has changed to match the 15373 ABI. This change is incompatible with previous GCC versions, but 15374 does fix compatibility with the Tru64 compiler and several corner 15375 cases where GCC was incompatible with itself. 15376 15377 ARM 15378 15379 * Nicolas Pitre has contributed his hand-coded floating-point support 15380 code for ARM. It is both significantly smaller and faster than the 15381 existing C-based implementation, even when building applications 15382 for Thumb. The arm-elf configuration has been converted to use the 15383 new code. 15384 * Support for the Intel's iWMMXt architecture, a second generation 15385 XScale processor, has been added. Enabled at run time with the 15386 -mcpu=iwmmxt command line switch. 15387 * A new ARM target has been added: arm-wince-pe. This is similar to 15388 the arm-pe target, but it defaults to using the APCS32 ABI. 15389 * The existing ARM pipeline description has been converted to the use 15390 the [20]DFA processor pipeline model. There is not much change in 15391 code performance, but the description is now [21]easier to 15392 understand. 15393 * Support for the Cirrus EP9312 Maverick floating point co-processor 15394 added. Enabled at run time with the -mcpu=ep9312 command line 15395 switch. Note however that the multilibs to support this chip are 15396 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 15397 enable their production you will have to uncomment the entries in 15398 that file. 15399 15400 H8/300 15401 15402 * Support for long long has been added. 15403 * Support for saveall attribute has been added. 15404 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 15405 for H8/300H and H8S, which is much faster than the previous 15406 implementation. 15407 * A lot of small performance improvements. 15408 15409 IA-32/AMD64 (x86-64) 15410 15411 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 15412 -march=k8 and -mcpu=k8. 15413 * Scalar SSE code generation carefully avoids reformatting penalties, 15414 hidden dependencies and minimizes the number of uops generated on 15415 both Intel and AMD CPUs. 15416 * Vector MMX and SSE operands are now passed in registers to improve 15417 performance and match the argument passing convention used by the 15418 Intel C++ Compiler. As a result it is not possible to call 15419 functions accepting vector arguments compiled by older GCC version. 15420 * Conditional jump elimination is now more aggressive on modern CPUs. 15421 * The Athlon ports has been converted to use the DFA processor 15422 pipeline description. 15423 * Optimization of indirect tail calls is now possible in a similar 15424 fashion as direct sibcall optimization. 15425 * Further small performance improvements. 15426 * -m128bit-long-double is now less buggy. 15427 * __float128 support in 64-bit compilation. 15428 * Support for data structures exceeding 2GB in 64-bit mode. 15429 * -mcpu has been renamed to -mtune. 15430 15431 IA-64 15432 15433 * Tuning code for the Itanium 2 processor has been added. The 15434 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 15435 enabled by default now. To generate code tuned for Itanium 1 the 15436 option -mtune=itanium1 should be used. 15437 * [22]DFA processor pipeline descriptions for the IA-64 processors 15438 have been added. This resulted in about 3% improvement on the 15439 SPECInt2000 benchmark for Itanium 2. 15440 * Instruction bundling for the IA-64 processors has been rewritten 15441 using the DFA pipeline hazard recognizer. It resulted in about 60% 15442 compiler speedup on the SPECInt2000 C programs. 15443 15444 M32R 15445 15446 * Support for the M32R/2 processor has been added by Renesas. 15447 * Support for an M32R GNU/Linux target and PIC code generation has 15448 been added by Renesas. 15449 15450 M68000 15451 15452 * Bernardo Innocenti (Develer S.r.l.) has contributed the 15453 m68k-uclinux target, based on former work done by Paul Dale 15454 (SnapGear Inc.). Code generation for the ColdFire processors family 15455 has been enhanced and extended to support the MCF 53xx and MCF 54xx 15456 cores, integrating former work done by Peter Barada (Motorola). 15457 15458 MIPS 15459 15460 Processor-specific changes 15461 15462 * Support for the RM7000 and RM9000 processors has been added. It can 15463 be selected using the -march compiler option and should work with 15464 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 15465 * Support for revision 2 of the MIPS32 ISA has been added. It can be 15466 selected with the command-line option -march=mips32r2. 15467 * There is a new option, -mfix-sb1, to work around certain SB-1 15468 errata. 15469 15470 Configuration 15471 15472 * It is possible to customize GCC using the following configure-time 15473 options: 15474 + --with-arch, which specifies the default value of the -march 15475 option. 15476 + --with-tune, which specifies the default value of the -mtune 15477 option. 15478 + --with-abi, which specifies the default ABI. 15479 + --with-float=soft, which tells GCC to use software floating 15480 point by default. 15481 + --with-float=hard, which tells GCC to use hardware floating 15482 point by default. 15483 * A 64-bit GNU/Linux port has been added. The associated 15484 configurations are mips64-linux-gnu and mips64el-linux-gnu. 15485 * The 32-bit GNU/Linux port now supports Java. 15486 * The IRIX 6 configuration now supports the o32 ABI and will build 15487 o32 multilibs by default. This support is compatible with both 15488 binutils and the SGI tools, but note that several features, 15489 including debugging information and DWARF2 exception handling, are 15490 only available when using the GNU assembler. Use of the GNU 15491 assembler and linker (version 2.15 or above) is strongly 15492 recommended. 15493 * The IRIX 6 configuration now supports 128-bit long doubles. 15494 * There are two new RTEMS-specific configurations, mips-rtems and 15495 mipsel-rtems. 15496 * There are two new *-elf configurations, mipsisa32r2-elf and 15497 mipsisa32r2el-elf. 15498 15499 General 15500 15501 * Several [23]ABI bugs have been fixed. Unfortunately, these changes 15502 will break binary compatibility with earlier releases. 15503 * GCC can now use explicit relocation operators when generating 15504 -mabicalls code. This behavior is controlled by -mexplicit-relocs 15505 and can have several performance benefits. For example: 15506 + It allows for more optimization of GOT accesses, including 15507 better scheduling and redundancy elimination. 15508 + It allows sibling calls to be implemented as jumps. 15509 + n32 and n64 leaf functions can use a call-clobbered global 15510 pointer instead of $28. 15511 + The code to set up $gp can be removed from functions that 15512 don't need it. 15513 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 15514 option is equivalent to the assembler's -xgot option and should be 15515 used instead of -Wa,-xgot. 15516 * Frame pointer elimination is now supported when generating 64-bit 15517 MIPS16 code. 15518 * Inline block moves have been optimized to take more account of 15519 alignment information. 15520 * Many internal changes have been made to the MIPS port, mostly aimed 15521 at reducing the reliance on assembler macros. 15522 15523 PowerPC 15524 15525 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 15526 [24]ABI incompatibilities regarding the way parameters are passed 15527 during functions calls. These changes may result in incompatibility 15528 between code compiled with GCC 3.3 and GCC 3.4. 15529 15530 PowerPC Darwin 15531 15532 * Support for shared/dylib gcc libraries has been added. It is 15533 enabled by default on powerpc-apple-darwin7.0.0 and up. 15534 * Libgcj is enabled by default. On systems older than 15535 powerpc-apple-darwin7.0.0 you need to install dlcompat. 15536 * 128-bit IBM extended precision format support added for long 15537 double. 15538 15539 PowerPC64 GNU/Linux 15540 15541 * By default, PowerPC64 GNU/Linux now uses natural alignment of 15542 structure elements. The old four byte alignment for double, with 15543 special rules for a struct starting with a double, can be chosen 15544 with -malign-power. This change may result in incompatibility 15545 between code compiled with GCC 3.3 and GCC 3.4. 15546 * -mabi=altivec is now the default rather than -mabi=no-altivec. 15547 * 128-bit IBM extended precision format support added for long 15548 double. 15549 15550 S/390 and zSeries 15551 15552 * New command-line options allow to specify the intended execution 15553 environment for generated code: 15554 + -mesa/-mzarch allows to specify whether to generate code 15555 running in ESA/390 mode or in z/Architecture mode (this is 15556 applicable to 31-bit code only). 15557 + -march allows to specify a minimum processor architecture 15558 level (g5, g6, z900, or z990). 15559 + -mtune allows to specify which processor to tune for. 15560 * It is possible to customize GCC using the following configure-time 15561 options: 15562 + --with-mode, which specifies whether to default to assuming 15563 ESA/390 or z/Architecture mode. 15564 + --with-arch, which specifies the default value of the -march 15565 option. 15566 + --with-tune, which specifies the default value of the -mtune 15567 option. 15568 * Support for the z990 processor has been added, and can be selected 15569 using -march=z990 or -mtune=z990. This includes instruction 15570 scheduling tuned for the superscalar instruction pipeline of the 15571 z990 processor as well as support for all new instructions provided 15572 by the long-displacement facility. 15573 * Support to generate 31-bit code optimized for zSeries processors 15574 (running in ESA/390 or in z/Architecture mode) has been added. This 15575 can be selected using -march=z900 and -mzarch respectively. 15576 * Instruction scheduling for the z900 and z990 processors now uses 15577 the DFA pipeline hazard recognizer. 15578 * GCC no longer generates code to maintain a stack backchain, 15579 previously used to generate stack backtraces for debugging 15580 purposes. As replacement that does not incur runtime overhead, 15581 DWARF-2 call frame information is provided by GCC; this is 15582 supported by GDB 6.1. The old behavior can be restored using the 15583 -mbackchain option. 15584 * The stack frame size of functions may now exceed 2 GB in 64-bit 15585 code. 15586 * A port for the 64-bit IBM TPF operating system has been added; the 15587 configuration is s390x-ibm-tpf. This configuration is supported as 15588 cross-compilation target only. 15589 * Various changes to improve the generated code have been 15590 implemented, including: 15591 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 15592 instructions to significantly speed up many floating-point 15593 applications. 15594 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 15595 WITH BORROW instructions to speed up long long arithmetic. 15596 + GCC now uses the SEARCH STRING instruction to implement 15597 strlen(). 15598 + In many cases, function call overhead for 31-bit code has been 15599 reduced by placing the literal pool after the function code 15600 instead of after the function prolog. 15601 + Register 14 is no longer reserved in 64-bit code. 15602 + Handling of global register variables has been improved. 15603 15604 SPARC 15605 15606 * The option -mflat is deprecated. 15607 * Support for large (> 2GB) frames has been added to the 64-bit port. 15608 * Several [25]ABI bugs have been fixed. Unfortunately, these changes 15609 will break binary compatibility with earlier releases. 15610 * The default debugging format has been switched from STABS to 15611 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 15612 the default debugging format for 64-bit code on Solaris. 15613 15614 SuperH 15615 15616 * Support for the SH2E processor has been added. Enabled at run time 15617 with the -m2e command line switch, or at configure time by 15618 specifying sh2e as the machine part of the target triple. 15619 15620 V850 15621 15622 * Support for the Mitsubishi V850E1 processor has been added. This is 15623 a variant of the V850E processor with some additional debugging 15624 instructions. 15625 15626 Xtensa 15627 15628 * Several ABI bugs have been fixed. Unfortunately, these changes 15629 break binary compatibility with earlier releases. 15630 + For big-endian processors, the padding of aggregate return 15631 values larger than a word has changed. If the size of an 15632 aggregate return value is not a multiple of 32 bits, previous 15633 versions of GCC inserted padding in the most-significant bytes 15634 of the first return value register. Aggregates larger than a 15635 word are now padded in the least-significant bytes of the last 15636 return value register used. Aggregates smaller than a word are 15637 still padded in the most-significant bytes. The return value 15638 padding has not changed for little-endian processors. 15639 + Function arguments with 16-byte alignment are now properly 15640 aligned. 15641 + The implementation of the va_list type has changed. A va_list 15642 value created by va_start from a previous release cannot be 15643 used with va_arg from this release, or vice versa. 15644 * More processor configuration options for Xtensa processors are 15645 supported: 15646 + the ABS instruction is now optional; 15647 + the ADDX* and SUBX* instructions are now optional; 15648 + an experimental CONST16 instruction can be used to synthesize 15649 constants instead of loading them from constant pools. 15650 These and other Xtensa processor configuration options can no 15651 longer be enabled or disabled by command-line options; the 15652 processor configuration must be specified by the xtensa-config.h 15653 header file when building GCC. Additionally, the 15654 -mno-serialize-volatile option is no longer supported. 15655 15656Obsolete Systems 15657 15658 Support for a number of older systems has been declared obsolete in GCC 15659 3.4. Unless there is activity to revive them, the next release of GCC 15660 will have their sources permanently removed. 15661 15662 All configurations of the following processor architectures have been 15663 declared obsolete: 15664 * Mitsubishi D30V, d30v-* 15665 * AT&T DSP1600 and DSP1610, dsp16xx-* 15666 * Intel 80960, i960 15667 15668 Also, some individual systems have been obsoleted: 15669 * ARM Family 15670 + Support for generating code for operation in APCS/26 mode 15671 (-mapcs-26). 15672 * IBM ESA/390 15673 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 15674 maintained and supported.) 15675 * Intel 386 family 15676 + MOSS, i?86-moss-msdos and i?86-*-moss* 15677 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 15678 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 15679 i?86-*-freebsd2* 15680 + GNU/Linux with a.out object format, i?86-linux*aout* 15681 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 15682 + Interix versions before Interix 3, i?86-*-interix 15683 + Mach microkernel, i?86-mach* 15684 + SCO UnixWare with UDK, i?86-*-udk* 15685 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 15686 + VSTa microkernel, i386-*-vsta 15687 * Motorola M68000 family 15688 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 15689 + NetBSD with a.out object format (before NetBSD 1.4), 15690 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 15691 + Generic System V r.4, m68k-*-sysv4* 15692 * VAX 15693 + Generic VAX, vax-*-* (This is generic VAX only; we have not 15694 obsoleted any VAX triples for specific operating systems.) 15695 15696Documentation improvements 15697 15698Other significant improvements 15699 15700 * The build system has undergone several significant cleanups. 15701 Subdirectories will only be configured if they are being built, and 15702 all subdirectory configures are run from the make command. The top 15703 level has been autoconfiscated. 15704 * Building GCC no longer writes to its source directory. This should 15705 help those wishing to share a read-only source directory over NFS 15706 or build from a CD. The exceptions to this feature are if you 15707 configure with either --enable-maintainer-mode or 15708 --enable-generated-files-in-srcdir. 15709 * The -W warning option has been renamed to -Wextra, which is more 15710 easily understood. The older spelling will be retained for 15711 backwards compatibility. 15712 * Substantial improvements in compile time have been made, 15713 particularly for non-optimizing compilations. 15714 __________________________________________________________________ 15715 15716GCC 3.4.0 15717 15718 Bug Fixes 15719 15720 A vast number of bugs have been fixed in 3.4.0, too many to publish a 15721 complete list here. [26]Follow this link to query the Bugzilla database 15722 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 15723 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 15724 regressions. 15725 __________________________________________________________________ 15726 15727GCC 3.4.1 15728 15729 Bug Fixes 15730 15731 This section lists the problem reports (PRs) from GCC's bug tracking 15732 system that are known to be fixed in the 3.4.1 release. This list might 15733 not be complete (that is, it is possible that some PRs that have been 15734 fixed are not listed here). 15735 15736 Bootstrap failures 15737 15738 * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 15739 emitted - PIC related 15740 * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 15741 * [29]14760 A bug in configure.in prevents using both 15742 --program-suffix and --program-prefix 15743 * [30]14671 [hppa64] bootstrap fails: ICE in 15744 save_call_clobbered_regs, in caller_save.c 15745 * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on 15746 Alpha 15747 * [32]15178 Solaris 9/x86 fails linking after stage 3 15748 15749 Multi-platform internal compiler errors (ICEs) 15750 15751 * [33]12753 (preprocessor) Memory corruption in preprocessor on bad 15752 input 15753 * [34]13985 ICE in gcc.c-torture/compile/930621-1.c 15754 * [35]14810 (c++) tree check failures with invalid code involving 15755 templates 15756 * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 15757 cp/parser.c 15758 * [37]15044 (c++) ICE on syntax error, template header 15759 * [38]15057 (c++) Compiling of conditional value throw constructs 15760 cause a segmentation violation 15761 * [39]15064 (c++) typeid of template parameter gives ICE 15762 * [40]15142 (c++) ICE when passing a string where a char* is expected 15763 in a throw statement 15764 * [41]15159 ICE in rtl_verify_flow_info_1 15765 * [42]15165 (c++) ICE in instantiate_template 15766 * [43]15193 Unary minus using pointer to V4SF vector causes 15767 -fforce-mem to exhaust all memory 15768 * [44]15209 (c++) Runs out of memory with packed structs 15769 * [45]15227 (c++) Trouble with invalid function definition 15770 * [46]15285 (c++) instantiate_type ICE when forming pointer to 15771 template function 15772 * [47]15299 (c++) ICE in resolve_overloaded_unification 15773 * [48]15329 (c++) ICE on constructor of member template 15774 * [49]15550 ICE in extract_insn, in recog.c 15775 * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c 15776 * [51]15640 (c++) ICE on invalid code in arg_assoc, in 15777 cp/name-lookup.c 15778 * [52]15666 [unit-at-a-time] Gcc abort on valid code 15779 * [53]15696 (c++) ICE with bad pointer-to-member code 15780 * [54]15701 (c++) ICE with friends and template template parameter 15781 * [55]15761 ICE in do_SUBST, in combine.c 15782 * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 15783 15784 Ada 15785 15786 * [57]14538 All RTEMS targets broken for gnat 15787 15788 C front end 15789 15790 * [58]12391 missing warning about assigning to an incomplete type 15791 * [59]14649 atan(1.0) should not be a constant expression 15792 * [60]15004 [unit-at-a-time] no warning for unused paramater in 15793 static function 15794 * [61]15749 --pedantic-errors behaves differently from --pedantic 15795 with C-compiler on GNU/Linux 15796 15797 C++ compiler and library 15798 15799 * [62]10646 non-const reference is incorrectly matched in a "const T" 15800 partial specialization 15801 * [63]12077 wcin.rdbuf()->in_avail() return value too high 15802 * [64]13598 enc_filebuf doesn't work 15803 * [65]14211 const_cast returns lvalue but should be rvalue 15804 * [66]14220 num_put::do_put() undesired float/double behavior 15805 * [67]14245 problem with user-defined allocators in std::basic_string 15806 * [68]14340 libstdc++ Debug mode: failure to convert iterator to 15807 const_iterator 15808 * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 15809 FILE* 15810 * [70]14668 no warning anymore for reevaluation of declaration 15811 * [71]14775 LFS (large file support) tests missing 15812 * [72]14821 Duplicate namespace alias declaration should not conflict 15813 * [73]14930 Friend declaration ignored 15814 * [74]14932 cannot use offsetof to get offsets of array elements in 15815 g++ 3.4.0 15816 * [75]14950 [non unit-at-a-time] always_inline does not mix with 15817 templates and -O0 15818 * [76]14962 g++ ignores #pragma redefine_extname 15819 * [77]14975 Segfault on low-level write error during imbue 15820 * [78]15002 Linewise stream input is unusably slow (std::string slow) 15821 * [79]15025 compiler accepts redeclaration of template as 15822 non-template 15823 * [80]15046 [arm] Math functions misdetected by cross configuration 15824 * [81]15069 a bit test on a variable of enum type is miscompiled 15825 * [82]15074 g++ -lsupc++ still links against libstdc++ 15826 * [83]15083 spurious "statement has no effect" warning 15827 * [84]15096 parse error with templates and pointer to const member 15828 * [85]15287 combination of operator[] and operator .* fails in 15829 templates 15830 * [86]15317 __attribute__ unused in first parameter of constructor 15831 gives error 15832 * [87]15337 sizeof on incomplete type diagnostic 15833 * [88]15361 bitset<>::_Find_next fails 15834 * [89]15412 _GLIBCXX_ symbols symbols defined and used in different 15835 namespaces 15836 * [90]15427 valid code results in incomplete type error 15837 * [91]15471 Incorrect member pointer offsets in anonymous 15838 structs/unions 15839 * [92]15503 nested template problem 15840 * [93]15507 compiler hangs while laying out union 15841 * [94]15542 operator & and template definitions 15842 * [95]15565 SLES9: leading + sign for unsigned int with showpos 15843 * [96]15625 friend defined inside a template fails to find static 15844 function 15845 * [97]15629 Function templates, overloads, and friend name injection 15846 * [98]15742 'noreturn' attribute ignored in method of template 15847 functions. 15848 * [99]15775 Allocator::pointer consistently ignored 15849 * [100]15821 Duplicate namespace alias within namespace rejected 15850 * [101]15862 'enum yn' fails (confict with undeclared builtin) 15851 * [102]15875 rejects pointer to member in template 15852 * [103]15877 valid code using templates and anonymous enums is 15853 rejected 15854 * [104]15947 Puzzling error message for wrong destructor declaration 15855 in template class 15856 * [105]16020 cannot copy __gnu_debug::bitset 15857 * [106]16154 input iterator concept too restrictive 15858 * [107]16174 deducing top-level consts 15859 15860 Java 15861 15862 * [108]14315 Java compiler is not parallel make safe 15863 15864 Fortran 15865 15866 * [109]15151 [g77] incorrect logical i/o in 64-bit mode 15867 15868 Objective-C 15869 15870 * [110]7993 private variables cannot be shadowed in subclasses 15871 15872 Optimization bugs 15873 15874 * [111]15228 useless copies of floating point operands 15875 * [112]15345 [non-unit-at-a-time] unreferenced nested inline 15876 functions not optimized away 15877 * [113]15945 Incorrect floating point optimization 15878 * [114]15526 ftrapv aborts on 0 * (-1) 15879 * [115]14690 Miscompiled POOMA tests 15880 * [116]15112 GCC generates code to write to unchanging memory 15881 15882 Preprocessor 15883 15884 * [117]15067 Minor glitch in the source of cpp 15885 15886 Main driver program bugs 15887 15888 * [118]1963 collect2 interprets -oldstyle_liblookup as -o 15889 ldstyle_liblookup 15890 15891 x86-specific (Intel/AMD) 15892 15893 * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 15894 section} 15895 15896 HPPA-specific 15897 15898 * [120]14782 GCC produces an unaligned data access at -O2 15899 * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 15900 * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c 15901 15902 IA64-specific 15903 15904 * [123]14610 __float80 constants incorrectly emitted 15905 * [124]14813 init_array sections are initialized in the wrong order 15906 * [125]14857 GCC segfault on duplicated asm statement 15907 * [126]15598 Gcc 3.4 ICE on valid code 15908 * [127]15653 Gcc 3.4 ICE on valid code 15909 15910 MIPS-specific 15911 15912 * [128]15189 wrong filling of delay slot with -march=mips1 -G0 15913 -mno-split-addresses -mno-explicit-relocs 15914 * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 15915 2.14.91 15916 * [130]16144 Bogus reference to __divdf3 when -O1 15917 * [131]16176 Miscompilation of unaligned data in MIPS backend 15918 15919 PowerPC-specific 15920 15921 * [132]11591 ICE in gcc.dg/altivec-5.c 15922 * [133]12028 powerpc-eabispe produces bad sCOND operation 15923 * [134]14478 rs6000 geu/ltu patterns generate incorrect code 15924 * [135]14567 long double and va_arg complex args 15925 * [136]14715 Altivec stack layout may overlap gpr save with stack 15926 temps 15927 * [137]14902 (libstdc++) Stream checking functions fail when -pthread 15928 option is used. 15929 * [138]14924 Compiler ICE on valid code 15930 * [139]14960 -maltivec affects vector return with -mabi=no-altivec 15931 * [140]15106 vector varargs failure passing from altivec to 15932 non-altivec code for -m32 15933 * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 15934 half-word operation 15935 * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 15936 and stvx 15937 * [143]15662 Segmentation fault when an exception is thrown - even if 15938 try and catch are specified 15939 15940 s390-specific 15941 15942 * [144]15054 Bad code due to overlapping stack temporaries 15943 15944 SPARC-specific 15945 15946 * [145]15783 ICE with union assignment in 64-bit mode 15947 * [146]15626 GCC 3.4 emits "ld: warning: relocation error: 15948 R_SPARC_UA32" 15949 15950 x86-64-specific 15951 15952 * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 15953 * [148]14723 Backported -march=nocona from mainline 15954 * [149]15290 __float128 failed to pass to function properly 15955 15956 Cygwin/Mingw32-specific 15957 15958 * [150]15250 Option -mms-bitfields support on GCC 3.4 is not 15959 conformant to MS layout 15960 * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 15961 worker on windows32 targets 15962 15963 Bugs specific to embedded processors 15964 15965 * [152]8309 [m68k] -m5200 produces erroneous SImode set of short 15966 varaible on stack 15967 * [153]13250 [SH] Gcc code for rotation clobbers the register, but 15968 gcc continues to use the register as if it was not clobbered 15969 * [154]13803 [coldfire] movqi operand constraints too restrictivefor 15970 TARGET_COLDFIRE 15971 * [155]14093 [SH] ICE for code when using -mhitachi option in SH 15972 * [156]14457 [m6811hc] ICE with simple c++ source 15973 * [157]14542 [m6811hc] ICE on simple source 15974 * [158]15100 [SH] cc1plus got hang-up on 15975 libstdc++-v3/testsuite/abi_check.cc 15976 * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on 15977 cris-* 15978 * [160]15396 [SH] ICE with -O2 -fPIC 15979 * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 15980 ColdFire 15981 15982 Testsuite problems (compiler not affected) 15983 15984 * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely 15985 * [163]15488 (libstdc++) possibly insufficient file permissions for 15986 executing test suite 15987 * [164]15489 (libstdc++) testsuite_files determined incorrectly 15988 15989 Documentation bugs 15990 15991 * [165]13928 (libstdc++) no whatis info in some man pages generated 15992 by doxygen 15993 * [166]14150 Ada documentation out of date 15994 * [167]14949 (c++) Need to document method visibility changes 15995 * [168]15123 libstdc++-doc: Allocators.3 manpage is empty 15996 __________________________________________________________________ 15997 15998GCC 3.4.2 15999 16000 Bug Fixes 16001 16002 This section lists the problem reports (PRs) from GCC's bug tracking 16003 system that are known to be fixed in the 3.4.2 release. This list might 16004 not be complete (that is, it is possible that some PRs that have been 16005 fixed are not listed here). 16006 16007 Bootstrap failures and issues 16008 16009 * [169]16469 [mips-sgi-irix5.3] bootstrap fails in 16010 libstdc++-v3/testsuite 16011 * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by 16012 profiledbootstrap does not work with the built compiler 16013 * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 16014 16015 Multi-platform internal compiler errors (ICEs) 16016 16017 * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 16018 cp_parser_class_specifier, in cp/parser.c 16019 * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 16020 * [174]15461 (c++) ICE due to NRV and inlining 16021 * [175]15890 (c++) ICE in c_expand_expr, in c-common.c 16022 * [176]16180 ICE: segmentation fault in RTL optimization 16023 * [177]16224 (c++) ICE in write_unscoped_name (template/namespace) 16024 * [178]16408 ICE: in delete_insn, in cfgrtl.c 16025 * [179]16529 (c++) ICE for: namespace-alias shall not be declared as 16026 the name of any other entity 16027 * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw 16028 * [181]16706 (c++) ICE in finish_member_declaration, in 16029 cp/semantics.c 16030 * [182]16810 (c++) Legal C++ program with cast gives ICE in 16031 build_ptrmemfunc 16032 * [183]16851 (c++) ICE when throwing a comma expression 16033 * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 16034 * [185]16904 (c++) ICE in finish_class_member_access_expr, in 16035 cp/typeck.c 16036 * [186]16905 (c++) ICE (segfault) with exceptions 16037 * [187]16964 (c++) ICE in cp_parser_class_specifier due to 16038 redefinition 16039 * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x' 16040 (identifier_node) in dependent_template_p, in cp/pt.c 16041 16042 Preprocessor bugs 16043 16044 * [189]16366 Preprocessor option -remap causes memory corruption 16045 16046 Optimization 16047 16048 * [190]15345 unreferenced nested inline functions not optimized away 16049 * [191]16590 Incorrect execution when compiling with -O2 16050 * [192]16693 Bitwise AND is lost when used within a cast to an enum 16051 of the same precision 16052 * [193]17078 Jump into if(0) substatement fails 16053 16054 Problems in generated debug information 16055 16056 * [194]13956 incorrect stabs for nested local variables 16057 16058 C front end bugs 16059 16060 * [195]16684 GCC should not warn about redundant redeclarations of 16061 built-ins 16062 16063 C++ compiler and library 16064 16065 * [196]12658 Thread safety problems in locale::global() and 16066 locale::locale() 16067 * [197]13092 g++ accepts invalid pointer-to-member conversion 16068 * [198]15320 Excessive memory consumption 16069 * [199]16246 Incorrect template argument deduction 16070 * [200]16273 Memory exhausted when using nested classes and virtual 16071 functions 16072 * [201]16401 ostringstream in gcc 3.4.x very slow for big data 16073 * [202]16411 undefined reference to 16074 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 16075 >::file() 16076 * [203]16489 G++ incorrectly rejects use of a null constant integral 16077 expression as a null constant pointer 16078 * [204]16618 offsetof fails with constant member 16079 * [205]16637 syntax error reported for valid input code 16080 * [206]16717 __attribute__((constructor)) broken in C++ 16081 * [207]16813 compiler error in DEBUG version of range insertion 16082 std::map::insert 16083 * [208]16853 pointer-to-member initialization from incompatible one 16084 accepted 16085 * [209]16889 ambiguity is not detected 16086 * [210]16959 Segmentation fault in ios_base::sync_with_stdio 16087 16088 Java compiler and library 16089 16090 * [211]7587 direct threaded interpreter not thread-safe 16091 * [212]16473 ServerSocket accept() leaks file descriptors 16092 * [213]16478 Hash synchronization deadlock with finalizers 16093 16094 Alpha-specific 16095 16096 * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 16097 * [215]16974 could not split insn (ice in final_scan_insn, in 16098 final.c) 16099 16100 x86-specific 16101 16102 * [216]16298 ICE in output_operand 16103 * [217]17113 ICE with SSE2 intrinsics 16104 16105 x86-64 specific 16106 16107 * [218]14697 libstdc++ couldn't find 32bit libgcc_s 16108 16109 MIPS-specific 16110 16111 * [219]15869 [mips64] No NOP after LW (with -mips1 -O0) 16112 * [220]16325 [mips64] value profiling clobbers gp on mips 16113 * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 16114 char[]s 16115 * [222]16380 [mips64] Use of uninitialised register after dbra 16116 conversion 16117 * [223]16407 [mips64] Unaligned access to local variables 16118 * [224]16643 [mips64] verify_local_live_at_start ICE after 16119 crossjumping & cfgcleanup 16120 16121 ARM-specific 16122 16123 * [225]15927 THUMB -O2: strength-reduced iteration variable ends up 16124 off by 1 16125 * [226]15948 THUMB: ICE with non-commutative cbranch 16126 * [227]17019 THUMB: bad switch statement in md code for 16127 addsi3_cbranch_scratch 16128 16129 IA64-specific 16130 16131 * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 16132 (-mtune=merced) 16133 * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 16134 (-mtune=itanium) 16135 * [230]16278 Gcc failed to build Linux kernel with -mtune=merced 16136 * [231]16414 ICE on valid code: typo in comparison of asm_noperands 16137 result 16138 * [232]16445 ICE on valid code: don't count ignored insns 16139 * [233]16490 ICE (segfault) while compiling with -fprofile-use 16140 * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 16141 16142 PowerPC-specific 16143 16144 * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 16145 * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 16146 issue) 16147 16148 SPARC-specific 16149 16150 * [237]16199 ICE while compiling apache 2.0.49 16151 * [238]16416 -m64 doesn't imply -mcpu=v9 anymore 16152 * [239]16430 ICE when returning non-C aggregates larger than 16 bytes 16153 16154 Bugs specific to embedded processors 16155 16156 * [240]16379 [m32r] can't output large model function call of memcpy 16157 * [241]17093 [m32r] ICE with -msdata=use -O0 16158 * [242]17119 [m32r] ICE at switch case 0x8000 16159 16160 DJGPP-specific 16161 16162 * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 16163 16164 Alpha Tru64-specific 16165 16166 * [244]16210 libstdc++ gratuitously omits "long long" I/O 16167 16168 Testsuite, documentation issues (compiler is not affected): 16169 16170 * [245]15488 (libstdc++) possibly insufficient file permissions for 16171 executing test suite 16172 * [246]16250 ada/doctools runs makeinfo even in release tarball 16173 __________________________________________________________________ 16174 16175GCC 3.4.3 16176 16177 This is the [247]list of problem reports (PRs) from GCC's bug tracking 16178 system that are known to be fixed in the 3.4.3 release. This list might 16179 not be complete (that is, it is possible that some PRs that have been 16180 fixed are not listed here). 16181 16182 Bootstrap failures 16183 16184 * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 16185 * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 16186 when undeclared 16187 16188 Internal compiler errors (ICEs) affecting multiple platforms 16189 16190 * [250]13948 (java) GCJ segmentation fault while compiling GL4Java 16191 .class files 16192 * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 16193 * [252]16301 (c++) ICE when "strong" attribute is attached to a using 16194 directive 16195 * [253]16566 ICE with flexible arrays 16196 * [254]17023 ICE with nested functions in parameter declaration 16197 * [255]17027 ICE with noreturn function in loop at -O2 16198 * [256]17524 ICE in grokdeclarator, in cp/decl.c 16199 * [257]17826 (c++) ICE in cp_tree_equal 16200 16201 C and optimization bugs 16202 16203 * [258]15526 -ftrapv aborts on 0 * (-1) 16204 * [259]16999 #ident stopped working 16205 * [260]17503 quadratic behaviour in invalid_mode_change_p 16206 * [261]17581 Long long arithmetic fails inside a switch/case 16207 statement when compiled with -O2 16208 * [262]18129 -fwritable-strings doesn't work 16209 16210 C++ compiler and library bugs 16211 16212 * [263]10975 incorrect initial ostringstream::tellp() 16213 * [264]11722 Unbuffered filebuf::sgetn is slow 16214 * [265]14534 Unrecognizing static function as a template parameter 16215 when its return value is also templated 16216 * [266]15172 Copy constructor optimization in aggregate 16217 initialization 16218 * [267]15786 Bad error message for frequently occuring error. 16219 * [268]16162 Rejects valid member-template-definition 16220 * [269]16612 empty basic_strings can't live in shared memory 16221 * [270]16715 std::basic_iostream is instantiated when used, even 16222 though instantiations are already contained in libstdc++ 16223 * [271]16848 code in /ext/demangle.h appears broken 16224 * [272]17132 GCC fails to eliminate function template specialization 16225 when argument deduction fails 16226 * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 16227 in ropeimpl.h 16228 * [274]17327 use of `enumeral_type' in template type unification 16229 * [275]17393 "unused variable '._0'" warning with -Wall 16230 * [276]17501 Confusion with member templates 16231 * [277]17537 g++ not passing -lstdc++ to linker when all command line 16232 arguments are libraries 16233 * [278]17585 usage of unqualified name of static member from within 16234 class not allowed 16235 * [279]17821 Poor diagnostic for using "." instead of "->" 16236 * [280]17829 wrong error: call of overloaded function is ambiguous 16237 * [281]17851 Misleading diagnostic for invalid function declarations 16238 with undeclared types 16239 * [282]17976 Destructor is called twice 16240 * [283]18020 rejects valid definition of enum value in template 16241 * [284]18093 bogus conflict in namespace aliasing 16242 * [285]18140 C++ parser bug when using >> in templates 16243 16244 Fortran 16245 16246 * [286]17541 data statements with double precision constants fail 16247 16248 x86-specific 16249 16250 * [287]17853 -O2 ICE for MMX testcase 16251 16252 SPARC-specific 16253 16254 * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c 16255 16256 Darwin-specific 16257 16258 * [289]17167 FATAL:Symbol L_foo$stub already defined. 16259 16260 AIX-specific 16261 16262 * [290]17277 could not catch an exception when specified -maix64 16263 16264 Solaris-specific 16265 16266 * [291]17505 <cmath> calls acosf(), ceilf(), and other functions 16267 missing from system libraries 16268 16269 HP/UX specific: 16270 16271 * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 16272 16273 ARM-specific 16274 16275 * [293]17384 ICE with mode attribute on structures 16276 16277 MIPS-specific 16278 16279 * [294]17770 No NOP after LWL with -mips1 16280 16281 Other embedded target specific 16282 16283 * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 16284 * [296]14064 [avr-elf] -fdata-sections triggers ICE 16285 * [297]14678 [m68hc11-elf] gcc ICE 16286 * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 16287 * [299]15790 [i686-coff] Alignment error building gcc with i686-coff 16288 target 16289 * [300]15886 [SH] Miscompilation with -O2 -fPIC 16290 * [301]16884 [avr-elf] [fweb related] bug while initializing 16291 variables 16292 16293 Bugs relating to debugger support 16294 16295 * [302]13841 missing debug info for _Complex function arguments 16296 * [303]15860 [big-endian targets] No DW_AT_location debug info is 16297 emitted for formal arguments to a function that uses "register" 16298 qualifiers 16299 16300 Testsuite issues (compiler not affected) 16301 16302 * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 16303 * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 16304 * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 16305 testsuite 16306 16307 Documentation 16308 16309 * [307]15498 typo in gcc manual: non-existing locale example en_UK, 16310 should be en_GB 16311 * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 16312 document broken shell 16313 * [309]16406 USE_LD_AS_NEEDED undocumented 16314 __________________________________________________________________ 16315 16316GCC 3.4.4 16317 16318 This is the [310]list of problem reports (PRs) from GCC's bug tracking 16319 system that are known to be fixed in the 3.4.4 release. This list might 16320 not be complete (that is, it is possible that some PRs that have been 16321 fixed are not listed here). 16322 __________________________________________________________________ 16323 16324GCC 3.4.5 16325 16326 This is the [311]list of problem reports (PRs) from GCC's bug tracking 16327 system that are known to be fixed in the 3.4.5 release. This list might 16328 not be complete (that is, it is possible that some PRs that have been 16329 fixed are not listed here). 16330 16331 Bootstrap issues 16332 16333 * [312]24688 sco_math fixincl breaks math.h 16334 16335 C compiler bugs 16336 16337 * [313]17188 struct Foo { } redefinition 16338 * [314]20187 wrong code for ((unsigned char)(unsigned long 16339 long)((a?a:1)&(a*b)))?0:1) 16340 * [315]21873 infinite warning loop on bad array initializer 16341 * [316]21899 enum definition accepts values to be overriden 16342 * [317]22061 ICE in find_function_data, in function.c 16343 * [318]22308 Failure to diagnose violation of constraint 6.516p2 16344 * [319]22458 ICE on missing brace 16345 * [320]22589 ICE casting to long long 16346 * [321]24101 Segfault with preprocessed source 16347 16348 C++ compiler and library bugs 16349 16350 * [322]10611 operations on vector mode not recognized in C++ 16351 * [323]13377 unexpected behavior of namespace usage directive 16352 * [324]16002 Strange error message with new parser 16353 * [325]17413 local classes as template argument 16354 * [326]17609 spurious error message after using keyword 16355 * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 16356 * [328]18124 ICE with invalid template template parameter 16357 * [329]18155 typedef in template declaration not rejected 16358 * [330]18177 ICE with const_cast for undeclared variable 16359 * [331]18368 C++ error message regression 16360 * [332]16378 ICE when returning a copy of a packed member 16361 * [333]18466 int ::i; accepted 16362 * [334]18512 ICE on invalid usage of template base class 16363 * [335]18454 ICE when returning undefined type 16364 * [336]18738 typename not allowed with non-dependent qualified name 16365 * [337]18803 rejects access to operator() in template 16366 * [338]19004 ICE in uses_template_parms, in cp/pt.c 16367 * [339]19208 Spurious error about variably modified type 16368 * [340]18253 bad error message / ICE for invalid template parameter 16369 * [341]19608 ICE after friend function definition in local class 16370 * [342]19884 ICE on explicit instantiation of a non-template 16371 constructor 16372 * [343]20153 ICE when C++ template function contains anonymous union 16373 * [344]20563 Infinite loop in diagnostic (and ice after error 16374 message) 16375 * [345]20789 ICE with incomplete type in template 16376 * [346]21336 Internal compiler error when using custom new operators 16377 * [347]21768 ICE in error message due to violation of coding 16378 conventions 16379 * [348]21853 constness of pointer to data member ignored 16380 * [349]21903 Default argument of template function causes a 16381 compile-time error 16382 * [350]21983 multiple diagnostics 16383 * [351]21987 New testsuite failure 16384 g++.dg/warn/conversion-function-1.C 16385 * [352]22153 ICE on invalid template specialization 16386 * [353]22172 Internal compiler error, seg fault. 16387 * [354]21286 filebuf::xsgetn vs pipes 16388 * [355]22233 ICE with wrong number of template parameters 16389 * [356]22508 ICE after invalid operator new 16390 * [357]22545 ICE with pointer to class member & user defined 16391 conversion operator 16392 * [358]23528 Wrong default allocator in ext/hash_map 16393 * [359]23550 char_traits requirements/1.cc test bad math 16394 * [360]23586 Bad diagnostic for invalid namespace-name 16395 * [361]23624 ICE in invert_truthvalue, in fold-const.c 16396 * [362]23639 Bad error message: not a member of '<declaration error>' 16397 * [363]23797 ICE on typename outside template 16398 * [364]23965 Bogus error message: no matching function for call to 16399 'foo(<type error>)' 16400 * [365]24052 &#`label_decl' not supported by dump_expr#<expression 16401 error> 16402 * [366]24580 virtual base class cause exception not to be caught 16403 16404 Problems in generated debug information 16405 16406 * [367]24267 Bad DWARF for altivec vectors 16407 16408 Optimizations issues 16409 16410 * [368]17810 ICE in verify_local_live_at_start 16411 * [369]17860 Wrong generated code for loop with varying bound 16412 * [370]21709 ICE on compile-time complex NaN 16413 * [371]21964 broken tail call at -O2 or more 16414 * [372]22167 Strange optimization bug when using -Os 16415 * [373]22619 Compilation failure for real_const_1.f and 16416 real_const_2.f90 16417 * [374]23241 Invalid code generated for comparison of uchar to 255 16418 * [375]23478 Miscompilation due to reloading of a var that is also 16419 used in EH pad 16420 * [376]24470 segmentation fault in cc1plus when compiling with -O 16421 * [377]24950 ICE in operand_subword_force 16422 16423 Precompiled headers problems 16424 16425 * [378]14400 Cannot compile qt-x11-free-3.3.0 16426 * [379]14940 PCH largefile test fails on various platforms 16427 16428 Preprocessor bugs 16429 16430 * [380]20239 ICE on empty preprocessed input 16431 * [381]15220 "gcc -E -MM -MG" reports missing system headers in 16432 source directory 16433 16434 Testsuite issues 16435 16436 * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 16437 i686-pc-linux-gnu 16438 16439 Alpha specific 16440 16441 * [383]21888 bootstrap failure with linker relaxation enabled 16442 16443 ARM specific 16444 16445 * [384]15342 [arm-linux]: ICE in verify_local_live_at_start 16446 * [385]23985 Memory aliasing information incorrect in inlined memcpy 16447 16448 ColdFile specific 16449 16450 * [386]16719 Illegal move of byte into address register causes 16451 compiler to ICE 16452 16453 HPPA specific 16454 16455 * [387]21723 ICE while building libgfortran 16456 * [388]21841 -mhp-ld/-mgnu-ld documentation 16457 16458 IA-64 specific 16459 16460 * [389]23644 IA-64 hardware models and configuration options 16461 documentation error 16462 * [390]24718 Shared libgcc not used for linking by default 16463 16464 M68000 specific 16465 16466 * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c 16467 16468 MIPS specific 16469 16470 * [392]20621 ICE in change_address_1, in emit-rtl.c 16471 16472 PowerPC and PowerPC64 specific 16473 16474 * [393]18583 error on valid code: const 16475 __attribute__((altivec(vector__))) doesn't work in arrays 16476 * [394]20191 ICE in reload_cse_simplify_operands 16477 * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 16478 * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 16479 * [397]23404 gij trashes args of functions with more than 8 fp args 16480 * [398]23539 C & C++ compiler generating misaligned references 16481 regardless of compiler flags 16482 * [399]24102 floatdisf2_internal2 broken 16483 * [400]24465 -mminimal-toc miscompilation of __thread vars 16484 16485 Solaris specific 16486 16487 * [401]19933 Problem with define of HUGE_VAL in math_c99 16488 * [402]21889 Native Solaris assembler cannot grok DTP-relative debug 16489 symbols 16490 16491 SPARC specific 16492 16493 * [403]19300 PCH failures on sparc-linux 16494 * [404]20301 Assembler labels have a leading "-" 16495 * [405]20673 C PCH testsuite assembly comparison failure 16496 16497 x86 and x86_64 specific 16498 16499 * [406]18582 ICE with arrays of type V2DF 16500 * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 16501 -fsched2-use-traces 16502 * [408]21716 ICE in reg-stack.c's swap_rtx_condition 16503 * [409]24315 amd64 fails -fpeephole2 16504 __________________________________________________________________ 16505 16506GCC 3.4.6 16507 16508 This is the [410]list of problem reports (PRs) from GCC's bug tracking 16509 system that are known to be fixed in the 3.4.6 release. This list might 16510 not be complete (that is, it is possible that some PRs that have been 16511 fixed are not listed here). 16512 16513 16514 For questions related to the use of GCC, please consult these web 16515 pages and the [411]GCC manuals. If that fails, the 16516 [412]gcc-help@gcc.gnu.org mailing list might help. Comments on these 16517 web pages and the development of GCC are welcome on our developer 16518 list at [413]gcc@gcc.gnu.org. All of [414]our lists have public 16519 archives. 16520 16521 Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and 16522 distribution of this entire article is permitted in any medium, 16523 provided this notice is preserved. 16524 16525 These pages are [416]maintained by the GCC team. Last modified 16526 2021-07-28[417]. 16527 16528References 16529 16530 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 16531 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 16532 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 16533 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 16534 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html 16535 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 16536 7. https://www.boost.org/ 16537 8. https://gcc.gnu.org/PR11953 16538 9. https://gcc.gnu.org/PR8361 16539 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins 16540 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 16541 12. http://gcc.gnu.org/bugs/#cxx_rvalbind 16542 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 16543 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 16544 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 16545 16. http://www.gnu.org/software/classpath/ 16546 17. http://www.eclipse.org/ 16547 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 16548 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 16549 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 16550 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 16551 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 16552 23. http://gcc.gnu.org/gcc-3.4/mips-abi.html 16553 24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html 16554 25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 16555 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 16556 27. https://gcc.gnu.org/PR10129 16557 28. https://gcc.gnu.org/PR14576 16558 29. https://gcc.gnu.org/PR14760 16559 30. https://gcc.gnu.org/PR14671 16560 31. https://gcc.gnu.org/PR15093 16561 32. https://gcc.gnu.org/PR15178 16562 33. https://gcc.gnu.org/PR12753 16563 34. https://gcc.gnu.org/PR13985 16564 35. https://gcc.gnu.org/PR14810 16565 36. https://gcc.gnu.org/PR14883 16566 37. https://gcc.gnu.org/PR15044 16567 38. https://gcc.gnu.org/PR15057 16568 39. https://gcc.gnu.org/PR15064 16569 40. https://gcc.gnu.org/PR15142 16570 41. https://gcc.gnu.org/PR15159 16571 42. https://gcc.gnu.org/PR15165 16572 43. https://gcc.gnu.org/PR15193 16573 44. https://gcc.gnu.org/PR15209 16574 45. https://gcc.gnu.org/PR15227 16575 46. https://gcc.gnu.org/PR15285 16576 47. https://gcc.gnu.org/PR15299 16577 48. https://gcc.gnu.org/PR15329 16578 49. https://gcc.gnu.org/PR15550 16579 50. https://gcc.gnu.org/PR15554 16580 51. https://gcc.gnu.org/PR15640 16581 52. https://gcc.gnu.org/PR15666 16582 53. https://gcc.gnu.org/PR15696 16583 54. https://gcc.gnu.org/PR15701 16584 55. https://gcc.gnu.org/PR15761 16585 56. https://gcc.gnu.org/PR15829 16586 57. https://gcc.gnu.org/PR14538 16587 58. https://gcc.gnu.org/PR12391 16588 59. https://gcc.gnu.org/PR14649 16589 60. https://gcc.gnu.org/PR15004 16590 61. https://gcc.gnu.org/PR15749 16591 62. https://gcc.gnu.org/PR10646 16592 63. https://gcc.gnu.org/PR12077 16593 64. https://gcc.gnu.org/PR13598 16594 65. https://gcc.gnu.org/PR14211 16595 66. https://gcc.gnu.org/PR14220 16596 67. https://gcc.gnu.org/PR14245 16597 68. https://gcc.gnu.org/PR14340 16598 69. https://gcc.gnu.org/PR14600 16599 70. https://gcc.gnu.org/PR14668 16600 71. https://gcc.gnu.org/PR14775 16601 72. https://gcc.gnu.org/PR14821 16602 73. https://gcc.gnu.org/PR14930 16603 74. https://gcc.gnu.org/PR14932 16604 75. https://gcc.gnu.org/PR14950 16605 76. https://gcc.gnu.org/PR14962 16606 77. https://gcc.gnu.org/PR14975 16607 78. https://gcc.gnu.org/PR15002 16608 79. https://gcc.gnu.org/PR15025 16609 80. https://gcc.gnu.org/PR15046 16610 81. https://gcc.gnu.org/PR15069 16611 82. https://gcc.gnu.org/PR15074 16612 83. https://gcc.gnu.org/PR15083 16613 84. https://gcc.gnu.org/PR15096 16614 85. https://gcc.gnu.org/PR15287 16615 86. https://gcc.gnu.org/PR15317 16616 87. https://gcc.gnu.org/PR15337 16617 88. https://gcc.gnu.org/PR15361 16618 89. https://gcc.gnu.org/PR15412 16619 90. https://gcc.gnu.org/PR15427 16620 91. https://gcc.gnu.org/PR15471 16621 92. https://gcc.gnu.org/PR15503 16622 93. https://gcc.gnu.org/PR15507 16623 94. https://gcc.gnu.org/PR15542 16624 95. https://gcc.gnu.org/PR15565 16625 96. https://gcc.gnu.org/PR15625 16626 97. https://gcc.gnu.org/PR15629 16627 98. https://gcc.gnu.org/PR15742 16628 99. https://gcc.gnu.org/PR15775 16629 100. https://gcc.gnu.org/PR15821 16630 101. https://gcc.gnu.org/PR15862 16631 102. https://gcc.gnu.org/PR15875 16632 103. https://gcc.gnu.org/PR15877 16633 104. https://gcc.gnu.org/PR15947 16634 105. https://gcc.gnu.org/PR16020 16635 106. https://gcc.gnu.org/PR16154 16636 107. https://gcc.gnu.org/PR16174 16637 108. https://gcc.gnu.org/PR14315 16638 109. https://gcc.gnu.org/PR15151 16639 110. https://gcc.gnu.org/PR7993 16640 111. https://gcc.gnu.org/PR15228 16641 112. https://gcc.gnu.org/PR15345 16642 113. https://gcc.gnu.org/PR15945 16643 114. https://gcc.gnu.org/PR15526 16644 115. https://gcc.gnu.org/PR14690 16645 116. https://gcc.gnu.org/PR15112 16646 117. https://gcc.gnu.org/PR15067 16647 118. https://gcc.gnu.org/PR1963 16648 119. https://gcc.gnu.org/PR15717 16649 120. https://gcc.gnu.org/PR14782 16650 121. https://gcc.gnu.org/PR14828 16651 122. https://gcc.gnu.org/PR15202 16652 123. https://gcc.gnu.org/PR14610 16653 124. https://gcc.gnu.org/PR14813 16654 125. https://gcc.gnu.org/PR14857 16655 126. https://gcc.gnu.org/PR15598 16656 127. https://gcc.gnu.org/PR15653 16657 128. https://gcc.gnu.org/PR15189 16658 129. https://gcc.gnu.org/PR15331 16659 130. https://gcc.gnu.org/PR16144 16660 131. https://gcc.gnu.org/PR16176 16661 132. https://gcc.gnu.org/PR11591 16662 133. https://gcc.gnu.org/PR12028 16663 134. https://gcc.gnu.org/PR14478 16664 135. https://gcc.gnu.org/PR14567 16665 136. https://gcc.gnu.org/PR14715 16666 137. https://gcc.gnu.org/PR14902 16667 138. https://gcc.gnu.org/PR14924 16668 139. https://gcc.gnu.org/PR14960 16669 140. https://gcc.gnu.org/PR15106 16670 141. https://gcc.gnu.org/PR16026 16671 142. https://gcc.gnu.org/PR15191 16672 143. https://gcc.gnu.org/PR15662 16673 144. https://gcc.gnu.org/PR15054 16674 145. https://gcc.gnu.org/PR15783 16675 146. https://gcc.gnu.org/PR15626 16676 147. https://gcc.gnu.org/PR14326 16677 148. https://gcc.gnu.org/PR14723 16678 149. https://gcc.gnu.org/PR15290 16679 150. https://gcc.gnu.org/PR15250 16680 151. https://gcc.gnu.org/PR15551 16681 152. https://gcc.gnu.org/PR8309 16682 153. https://gcc.gnu.org/PR13250 16683 154. https://gcc.gnu.org/PR13803 16684 155. https://gcc.gnu.org/PR14093 16685 156. https://gcc.gnu.org/PR14457 16686 157. https://gcc.gnu.org/PR14542 16687 158. https://gcc.gnu.org/PR15100 16688 159. https://gcc.gnu.org/PR15296 16689 160. https://gcc.gnu.org/PR15396 16690 161. https://gcc.gnu.org/PR15782 16691 162. https://gcc.gnu.org/PR11610 16692 163. https://gcc.gnu.org/PR15488 16693 164. https://gcc.gnu.org/PR15489 16694 165. https://gcc.gnu.org/PR13928 16695 166. https://gcc.gnu.org/PR14150 16696 167. https://gcc.gnu.org/PR14949 16697 168. https://gcc.gnu.org/PR15123 16698 169. https://gcc.gnu.org/PR16469 16699 170. https://gcc.gnu.org/PR16344 16700 171. https://gcc.gnu.org/PR16842 16701 172. https://gcc.gnu.org/PR12608 16702 173. https://gcc.gnu.org/PR14492 16703 174. https://gcc.gnu.org/PR15461 16704 175. https://gcc.gnu.org/PR15890 16705 176. https://gcc.gnu.org/PR16180 16706 177. https://gcc.gnu.org/PR16224 16707 178. https://gcc.gnu.org/PR16408 16708 179. https://gcc.gnu.org/PR16529 16709 180. https://gcc.gnu.org/PR16698 16710 181. https://gcc.gnu.org/PR16706 16711 182. https://gcc.gnu.org/PR16810 16712 183. https://gcc.gnu.org/PR16851 16713 184. https://gcc.gnu.org/PR16870 16714 185. https://gcc.gnu.org/PR16904 16715 186. https://gcc.gnu.org/PR16905 16716 187. https://gcc.gnu.org/PR16964 16717 188. https://gcc.gnu.org/PR17068 16718 189. https://gcc.gnu.org/PR16366 16719 190. https://gcc.gnu.org/PR15345 16720 191. https://gcc.gnu.org/PR16590 16721 192. https://gcc.gnu.org/PR16693 16722 193. https://gcc.gnu.org/PR17078 16723 194. https://gcc.gnu.org/PR13956 16724 195. https://gcc.gnu.org/PR16684 16725 196. https://gcc.gnu.org/PR12658 16726 197. https://gcc.gnu.org/PR13092 16727 198. https://gcc.gnu.org/PR15320 16728 199. https://gcc.gnu.org/PR16246 16729 200. https://gcc.gnu.org/PR16273 16730 201. https://gcc.gnu.org/PR16401 16731 202. https://gcc.gnu.org/PR16411 16732 203. https://gcc.gnu.org/PR16489 16733 204. https://gcc.gnu.org/PR16618 16734 205. https://gcc.gnu.org/PR16637 16735 206. https://gcc.gnu.org/PR16717 16736 207. https://gcc.gnu.org/PR16813 16737 208. https://gcc.gnu.org/PR16853 16738 209. https://gcc.gnu.org/PR16889 16739 210. https://gcc.gnu.org/PR16959 16740 211. https://gcc.gnu.org/PR7587 16741 212. https://gcc.gnu.org/PR16473 16742 213. https://gcc.gnu.org/PR16478 16743 214. https://gcc.gnu.org/PR10695 16744 215. https://gcc.gnu.org/PR16974 16745 216. https://gcc.gnu.org/PR16298 16746 217. https://gcc.gnu.org/PR17113 16747 218. https://gcc.gnu.org/PR14697 16748 219. https://gcc.gnu.org/PR15869 16749 220. https://gcc.gnu.org/PR16325 16750 221. https://gcc.gnu.org/PR16357 16751 222. https://gcc.gnu.org/PR16380 16752 223. https://gcc.gnu.org/PR16407 16753 224. https://gcc.gnu.org/PR16643 16754 225. https://gcc.gnu.org/PR15927 16755 226. https://gcc.gnu.org/PR15948 16756 227. https://gcc.gnu.org/PR17019 16757 228. https://gcc.gnu.org/PR16130 16758 229. https://gcc.gnu.org/PR16142 16759 230. https://gcc.gnu.org/PR16278 16760 231. https://gcc.gnu.org/PR16414 16761 232. https://gcc.gnu.org/PR16445 16762 233. https://gcc.gnu.org/PR16490 16763 234. https://gcc.gnu.org/PR16683 16764 235. https://gcc.gnu.org/PR16195 16765 236. https://gcc.gnu.org/PR16239 16766 237. https://gcc.gnu.org/PR16199 16767 238. https://gcc.gnu.org/PR16416 16768 239. https://gcc.gnu.org/PR16430 16769 240. https://gcc.gnu.org/PR16379 16770 241. https://gcc.gnu.org/PR17093 16771 242. https://gcc.gnu.org/PR17119 16772 243. https://gcc.gnu.org/PR15928 16773 244. https://gcc.gnu.org/PR16210 16774 245. https://gcc.gnu.org/PR15488 16775 246. https://gcc.gnu.org/PR16250 16776 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 16777 248. https://gcc.gnu.org/PR17369 16778 249. https://gcc.gnu.org/PR17850 16779 250. https://gcc.gnu.org/PR13948 16780 251. https://gcc.gnu.org/PR14492 16781 252. https://gcc.gnu.org/PR16301 16782 253. https://gcc.gnu.org/PR16566 16783 254. https://gcc.gnu.org/PR17023 16784 255. https://gcc.gnu.org/PR17027 16785 256. https://gcc.gnu.org/PR17524 16786 257. https://gcc.gnu.org/PR17826 16787 258. https://gcc.gnu.org/PR15526 16788 259. https://gcc.gnu.org/PR16999 16789 260. https://gcc.gnu.org/PR17503 16790 261. https://gcc.gnu.org/PR17581 16791 262. https://gcc.gnu.org/PR18129 16792 263. https://gcc.gnu.org/PR10975 16793 264. https://gcc.gnu.org/PR11722 16794 265. https://gcc.gnu.org/PR14534 16795 266. https://gcc.gnu.org/PR15172 16796 267. https://gcc.gnu.org/PR15786 16797 268. https://gcc.gnu.org/PR16162 16798 269. https://gcc.gnu.org/PR16612 16799 270. https://gcc.gnu.org/PR16715 16800 271. https://gcc.gnu.org/PR16848 16801 272. https://gcc.gnu.org/PR17132 16802 273. https://gcc.gnu.org/PR17259 16803 274. https://gcc.gnu.org/PR17327 16804 275. https://gcc.gnu.org/PR17393 16805 276. https://gcc.gnu.org/PR17501 16806 277. https://gcc.gnu.org/PR17537 16807 278. https://gcc.gnu.org/PR17585 16808 279. https://gcc.gnu.org/PR17821 16809 280. https://gcc.gnu.org/PR17829 16810 281. https://gcc.gnu.org/PR17851 16811 282. https://gcc.gnu.org/PR17976 16812 283. https://gcc.gnu.org/PR18020 16813 284. https://gcc.gnu.org/PR18093 16814 285. https://gcc.gnu.org/PR18140 16815 286. https://gcc.gnu.org/PR17541 16816 287. https://gcc.gnu.org/PR17853 16817 288. https://gcc.gnu.org/PR17245 16818 289. https://gcc.gnu.org/PR17167 16819 290. https://gcc.gnu.org/PR17277 16820 291. https://gcc.gnu.org/PR17505 16821 292. https://gcc.gnu.org/PR17684 16822 293. https://gcc.gnu.org/PR17384 16823 294. https://gcc.gnu.org/PR17770 16824 295. https://gcc.gnu.org/PR11476 16825 296. https://gcc.gnu.org/PR14064 16826 297. https://gcc.gnu.org/PR14678 16827 298. https://gcc.gnu.org/PR15583 16828 299. https://gcc.gnu.org/PR15790 16829 300. https://gcc.gnu.org/PR15886 16830 301. https://gcc.gnu.org/PR16884 16831 302. https://gcc.gnu.org/PR13841 16832 303. https://gcc.gnu.org/PR15860 16833 304. https://gcc.gnu.org/PR17465 16834 305. https://gcc.gnu.org/PR17469 16835 306. https://gcc.gnu.org/PR18138 16836 307. https://gcc.gnu.org/PR15498 16837 308. https://gcc.gnu.org/PR15747 16838 309. https://gcc.gnu.org/PR16406 16839 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 16840 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 16841 312. https://gcc.gnu.org/PR24688 16842 313. https://gcc.gnu.org/PR17188 16843 314. https://gcc.gnu.org/PR20187 16844 315. https://gcc.gnu.org/PR21873 16845 316. https://gcc.gnu.org/PR21899 16846 317. https://gcc.gnu.org/PR22061 16847 318. https://gcc.gnu.org/PR22208 16848 319. https://gcc.gnu.org/PR22458 16849 320. https://gcc.gnu.org/PR22589 16850 321. https://gcc.gnu.org/PR24101 16851 322. https://gcc.gnu.org/PR10611 16852 323. https://gcc.gnu.org/PR13377 16853 324. https://gcc.gnu.org/PR16002 16854 325. https://gcc.gnu.org/PR17413 16855 326. https://gcc.gnu.org/PR17609 16856 327. https://gcc.gnu.org/PR17618 16857 328. https://gcc.gnu.org/PR18124 16858 329. https://gcc.gnu.org/PR18155 16859 330. https://gcc.gnu.org/PR18177 16860 331. https://gcc.gnu.org/PR18368 16861 332. https://gcc.gnu.org/PR18378 16862 333. https://gcc.gnu.org/PR18466 16863 334. https://gcc.gnu.org/PR18512 16864 335. https://gcc.gnu.org/PR18545 16865 336. https://gcc.gnu.org/PR18738 16866 337. https://gcc.gnu.org/PR18803 16867 338. https://gcc.gnu.org/PR19004 16868 339. https://gcc.gnu.org/PR19208 16869 340. https://gcc.gnu.org/PR19253 16870 341. https://gcc.gnu.org/PR19608 16871 342. https://gcc.gnu.org/PR19884 16872 343. https://gcc.gnu.org/PR20153 16873 344. https://gcc.gnu.org/PR20563 16874 345. https://gcc.gnu.org/PR20789 16875 346. https://gcc.gnu.org/PR21336 16876 347. https://gcc.gnu.org/PR21768 16877 348. https://gcc.gnu.org/PR21853 16878 349. https://gcc.gnu.org/PR21903 16879 350. https://gcc.gnu.org/PR21983 16880 351. https://gcc.gnu.org/PR21987 16881 352. https://gcc.gnu.org/PR22153 16882 353. https://gcc.gnu.org/PR22172 16883 354. https://gcc.gnu.org/PR21286 16884 355. https://gcc.gnu.org/PR22233 16885 356. https://gcc.gnu.org/PR22508 16886 357. https://gcc.gnu.org/PR22545 16887 358. https://gcc.gnu.org/PR23528 16888 359. https://gcc.gnu.org/PR23550 16889 360. https://gcc.gnu.org/PR23586 16890 361. https://gcc.gnu.org/PR23624 16891 362. https://gcc.gnu.org/PR23639 16892 363. https://gcc.gnu.org/PR23797 16893 364. https://gcc.gnu.org/PR23965 16894 365. https://gcc.gnu.org/PR24052 16895 366. https://gcc.gnu.org/PR24580 16896 367. https://gcc.gnu.org/PR24267 16897 368. https://gcc.gnu.org/PR17810 16898 369. https://gcc.gnu.org/PR17860 16899 370. https://gcc.gnu.org/PR21709 16900 371. https://gcc.gnu.org/PR21964 16901 372. https://gcc.gnu.org/PR22167 16902 373. https://gcc.gnu.org/PR22619 16903 374. https://gcc.gnu.org/PR23241 16904 375. https://gcc.gnu.org/PR23478 16905 376. https://gcc.gnu.org/PR24470 16906 377. https://gcc.gnu.org/PR24950 16907 378. https://gcc.gnu.org/PR14400 16908 379. https://gcc.gnu.org/PR14940 16909 380. https://gcc.gnu.org/PR20239 16910 381. https://gcc.gnu.org/PR15220 16911 382. https://gcc.gnu.org/PR19275 16912 383. https://gcc.gnu.org/PR21888 16913 384. https://gcc.gnu.org/PR15342 16914 385. https://gcc.gnu.org/PR23985 16915 386. https://gcc.gnu.org/PR16719 16916 387. https://gcc.gnu.org/PR21723 16917 388. https://gcc.gnu.org/PR21841 16918 389. https://gcc.gnu.org/PR23644 16919 390. https://gcc.gnu.org/PR24718 16920 391. https://gcc.gnu.org/PR18421 16921 392. https://gcc.gnu.org/PR20621 16922 393. https://gcc.gnu.org/PR18583 16923 394. https://gcc.gnu.org/PR20191 16924 395. https://gcc.gnu.org/PR22083 16925 396. https://gcc.gnu.org/PR23070 16926 397. https://gcc.gnu.org/PR23404 16927 398. https://gcc.gnu.org/PR23539 16928 399. https://gcc.gnu.org/PR24102 16929 400. https://gcc.gnu.org/PR24465 16930 401. https://gcc.gnu.org/PR19933 16931 402. https://gcc.gnu.org/PR21889 16932 403. https://gcc.gnu.org/PR19300 16933 404. https://gcc.gnu.org/PR20301 16934 405. https://gcc.gnu.org/PR20673 16935 406. https://gcc.gnu.org/PR18582 16936 407. https://gcc.gnu.org/PR19340 16937 408. https://gcc.gnu.org/PR21716 16938 409. https://gcc.gnu.org/PR24315 16939 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 16940 411. https://gcc.gnu.org/onlinedocs/ 16941 412. mailto:gcc-help@gcc.gnu.org 16942 413. mailto:gcc@gcc.gnu.org 16943 414. https://gcc.gnu.org/lists.html 16944 415. https://www.fsf.org/ 16945 416. https://gcc.gnu.org/about.html 16946 417. http://validator.w3.org/check/referer 16947====================================================================== 16948http://gcc.gnu.org/gcc-3.3/index.html 16949 16950 GCC 3.3 Release Series 16951 16952 (This release series is no longer supported.) 16953 16954 May 03, 2005 16955 16956 The [1]GNU project and the GCC developers are pleased to announce the 16957 release of GCC 3.3.6. 16958 16959 This release is a bug-fix release, containing fixes for regressions in 16960 GCC 3.3.5 relative to previous releases of GCC. 16961 16962 This release is the last of the series 3.3.x. 16963 16964 The GCC 3.3 release series includes numerous [2]new features, 16965 improvements, bug fixes, and other changes, thanks to an [3]amazing 16966 group of volunteers. 16967 16968Release History 16969 16970 GCC 3.3.6 16971 May 3, 2005 ([4]changes) 16972 16973 GCC 3.3.5 16974 September 30, 2004 ([5]changes) 16975 16976 GCC 3.3.4 16977 May 31, 2004 ([6]changes) 16978 16979 GCC 3.3.3 16980 February 14, 2004 ([7]changes) 16981 16982 GCC 3.3.2 16983 October 16, 2003 ([8]changes) 16984 16985 GCC 3.3.1 16986 August 8, 2003 ([9]changes) 16987 16988 GCC 3.3 16989 May 14, 2003 ([10]changes) 16990 16991References and Acknowledgements 16992 16993 GCC used to stand for the GNU C Compiler, but since the compiler 16994 supports several other languages aside from C, it now stands for the 16995 GNU Compiler Collection. 16996 16997 A list of [11]successful builds is updated as new information becomes 16998 available. 16999 17000 The GCC developers would like to thank the numerous people that have 17001 contributed new features, improvements, bug fixes, and other changes as 17002 well as test results to GCC. This [12]amazing group of volunteers is 17003 what makes GCC successful. 17004 17005 For additional information about GCC please refer to the [13]GCC 17006 project web site or contact the [14]GCC development mailing list. 17007 17008 To obtain GCC please use [15]our mirror sites, or our CVS server. 17009 17010 17011 For questions related to the use of GCC, please consult these web 17012 pages and the [16]GCC manuals. If that fails, the 17013 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these 17014 web pages and the development of GCC are welcome on our developer 17015 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public 17016 archives. 17017 17018 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 17019 distribution of this entire article is permitted in any medium, 17020 provided this notice is preserved. 17021 17022 These pages are [21]maintained by the GCC team. Last modified 17023 2021-07-28[22]. 17024 17025References 17026 17027 1. http://www.gnu.org/ 17028 2. http://gcc.gnu.org/gcc-3.3/changes.html 17029 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17030 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 17031 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 17032 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 17033 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 17034 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 17035 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 17036 10. http://gcc.gnu.org/gcc-3.3/changes.html 17037 11. http://gcc.gnu.org/gcc-3.3/buildstat.html 17038 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17039 13. http://gcc.gnu.org/index.html 17040 14. mailto:gcc@gcc.gnu.org 17041 15. http://gcc.gnu.org/mirrors.html 17042 16. https://gcc.gnu.org/onlinedocs/ 17043 17. mailto:gcc-help@gcc.gnu.org 17044 18. mailto:gcc@gcc.gnu.org 17045 19. https://gcc.gnu.org/lists.html 17046 20. https://www.fsf.org/ 17047 21. https://gcc.gnu.org/about.html 17048 22. http://validator.w3.org/check/referer 17049====================================================================== 17050http://gcc.gnu.org/gcc-3.3/changes.html 17051 17052 GCC 3.3 Release Series 17053 Changes, New Features, and Fixes 17054 17055 The latest release in the 3.3 release series is [1]GCC 3.3.6. 17056 17057Caveats 17058 17059 * The preprocessor no longer accepts multi-line string literals. They 17060 were deprecated in 3.0, 3.1, and 3.2. 17061 * The preprocessor no longer supports the -A- switch when appearing 17062 alone. -A- followed by an assertion is still supported. 17063 * Support for all the systems [2]obsoleted in GCC 3.1 has been 17064 removed from GCC 3.3. See below for a [3]list of systems which are 17065 obsoleted in this release. 17066 * Checking for null format arguments has been decoupled from the rest 17067 of the format checking mechanism. Programs which use the format 17068 attribute may regain this functionality by using the new [4]nonnull 17069 function attribute. Note that all functions for which GCC has a 17070 built-in format attribute, an appropriate built-in nonnull 17071 attribute is also applied. 17072 * The DWARF (version 1) debugging format has been deprecated and will 17073 be removed in a future version of GCC. Version 2 of the DWARF 17074 debugging format will continue to be supported for the foreseeable 17075 future. 17076 * The C and Objective-C compilers no longer accept the "Naming Types" 17077 extension (typedef foo = bar); it was already unavailable in C++. 17078 Code which uses it will need to be changed to use the "typeof" 17079 extension instead: typedef typeof(bar) foo. (We have removed this 17080 extension without a period of deprecation because it has caused the 17081 compiler to crash since version 3.0 and no one noticed until very 17082 recently. Thus we conclude it is not in widespread use.) 17083 * The -traditional C compiler option has been removed. It was 17084 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 17085 available.) The <varargs.h> header, used for writing variadic 17086 functions in traditional C, still exists but will produce an error 17087 message if used. 17088 * GCC 3.3.1 automatically places zero-initialized variables in the 17089 .bss section on some operating systems. Versions of GNU Emacs up to 17090 (and including) 21.3 will not work correctly when using this 17091 optimization; you can use -fno-zero-initialized-in-bss to disable 17092 it. 17093 17094General Optimizer Improvements 17095 17096 * A new scheme for accurately describing processor pipelines, the 17097 [5]DFA scheduler, has been added. 17098 * Pavel Nejedly, Charles University Prague, has contributed new file 17099 format used by the edge coverage profiler (-fprofile-arcs). 17100 The new format is robust and diagnoses common mistakes where 17101 profiles from different versions (or compilations) of the program 17102 are combined resulting in nonsensical profiles and slow code to 17103 produced with profile feedback. Additionally this format allows 17104 extra data to be gathered. Currently, overall statistics are 17105 produced helping optimizers to identify hot spots of a program 17106 globally replacing the old intra-procedural scheme and resulting in 17107 better code. Note that the gcov tool from older GCC versions will 17108 not be able to parse the profiles generated by GCC 3.3 and vice 17109 versa. 17110 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 17111 pass enabled using -ftracer. This pass simplifies the control flow 17112 of functions allowing other optimizations to do better job. 17113 He also contributed the function reordering pass 17114 (-freorder-functions) to optimize function placement using profile 17115 feedback. 17116 17117New Languages and Language specific improvements 17118 17119 C/ObjC/C++ 17120 17121 * The preprocessor now accepts directives within macro arguments. It 17122 processes them just as if they had not been within macro arguments. 17123 * The separate ISO and traditional preprocessors have been completely 17124 removed. The front end handles either type of preprocessed output 17125 if necessary. 17126 * In C99 mode preprocessor arithmetic is done in the precision of the 17127 target's intmax_t, as required by that standard. 17128 * The preprocessor can now copy comments inside macros to the output 17129 file when the macro is expanded. This feature, enabled using the 17130 -CC option, is intended for use by applications which place 17131 metadata or directives inside comments, such as lint. 17132 * The method of constructing the list of directories to be searched 17133 for header files has been revised. If a directory named by a -I 17134 option is a standard system include directory, the option is 17135 ignored to ensure that the default search order for system 17136 directories and the special treatment of system header files are 17137 not defeated. 17138 * A few more [6]ISO C99 features now work correctly. 17139 * A new function attribute, nonnull, has been added which allows 17140 pointer arguments to functions to be specified as requiring a 17141 non-null value. The compiler currently uses this information to 17142 issue a warning when it detects a null value passed in such an 17143 argument slot. 17144 * A new type attribute, may_alias, has been added. Accesses to 17145 objects with types with this attribute are not subjected to 17146 type-based alias analysis, but are instead assumed to be able to 17147 alias any other type of objects, just like the char type. 17148 17149 C++ 17150 17151 * Type based alias analysis has been implemented for C++ aggregate 17152 types. 17153 17154 Objective-C 17155 17156 * Generate an error if Objective-C objects are passed by value in 17157 function and method calls. 17158 * When -Wselector is used, check the whole list of selectors at the 17159 end of compilation, and emit a warning if a @selector() is not 17160 known. 17161 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 17162 * No longer need to include objc/objc-class.h to compile self calls 17163 in class methods (NeXT runtime only). 17164 * New -Wundeclared-selector option. 17165 * Removed selector bloating which was causing object files to be 10% 17166 bigger on average (GNU runtime only). 17167 * Using at run time @protocol() objects has been fixed in certain 17168 situations (GNU runtime only). 17169 * Type checking has been fixed and improved in many situations 17170 involving protocols. 17171 17172 Java 17173 17174 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 17175 1.4) API. 17176 * The JDK 1.4 assert facility has been implemented. 17177 * The bytecode interpreter is now direct threaded and thus faster. 17178 17179 Fortran 17180 17181 * Fortran improvements are listed in [7]the Fortran documentation. 17182 17183 Ada 17184 17185 * Ada tasking now works with glibc 2.3.x threading libraries. 17186 17187New Targets and Target Specific Improvements 17188 17189 * The following changes have been made to the HP-PA port: 17190 + The port now defaults to scheduling for the PA8000 series of 17191 processors. 17192 + Scheduling support for the PA7300 processor has been added. 17193 + The 32-bit port now supports weak symbols under HP-UX 11. 17194 + The handling of initializers and finalizers has been improved 17195 under HP-UX 11. The 64-bit port no longer uses collect2. 17196 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 17197 + ABI fixes to correct the passing of small structures by value. 17198 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 17199 use the DFA processor pipeline description. 17200 * The following NetBSD configurations for the SuperH processor family 17201 have been added: 17202 + SH3, big-endian, sh-*-netbsdelf* 17203 + SH3, little-endian, shle-*-netbsdelf* 17204 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 17205 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 17206 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 17207 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 17208 * The following changes have been made to the IA-32/x86-64 port: 17209 + SSE2 and 3dNOW! intrinsics are now supported. 17210 + Support for thread local storage has been added to the IA-32 17211 and x86-64 ports. 17212 + The x86-64 port has been significantly improved. 17213 * The following changes have been made to the MIPS port: 17214 + All configurations now accept the -mabi switch. Note that you 17215 will need appropriate multilibs for this option to work 17216 properly. 17217 + ELF configurations will always pass an ABI flag to the 17218 assembler, except when the MIPS EABI is selected. 17219 + -mabi=64 no longer selects MIPS IV code. 17220 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 17221 been removed from this release. 17222 + -march now changes the core ISA level. In previous releases, 17223 it would change the use of processor-specific extensions, but 17224 would leave the core ISA unchanged. For example, mips64-elf 17225 -march=r8000 will now generate MIPS IV code. 17226 + Under most configurations, -mipsN now acts as a synonym for 17227 -march. 17228 + There are some new preprocessor macros to describe the -march 17229 and -mtune settings. See the documentation of those options 17230 for details. 17231 + Support for the NEC VR-Series processors has been added. This 17232 includes the 54xx, 5500, and 41xx series. 17233 + Support for the Sandcraft sr71k processor has been added. 17234 * The following changes have been made to the S/390 port: 17235 + Support to build the Java runtime libraries has been added. 17236 Java is now enabled by default on s390-*-linux* and 17237 s390x-*-linux* targets. 17238 + Multilib support for the s390x-*-linux* target has been added; 17239 this allows to build 31-bit binaries using the -m31 option. 17240 + Support for thread local storage has been added. 17241 + Inline assembler code may now use the 'Q' constraint to 17242 specify memory operands without index register. 17243 + Various platform-specific performance improvements have been 17244 implemented; in particular, the compiler now uses the BRANCH 17245 ON COUNT family of instructions and makes more frequent use of 17246 the TEST UNDER MASK family of instructions. 17247 * The following changes have been made to the PowerPC port: 17248 + Support for IBM Power4 processor added. 17249 + Support for Motorola e500 SPE added. 17250 + Support for AIX 5.2 added. 17251 + Function and Data sections now supported on AIX. 17252 + Sibcall optimizations added. 17253 * The support for H8 Tiny is added to the H8/300 port with -mn. 17254 17255Obsolete Systems 17256 17257 Support for a number of older systems has been declared obsolete in GCC 17258 3.3. Unless there is activity to revive them, the next release of GCC 17259 will have their sources permanently removed. 17260 17261 All configurations of the following processor architectures have been 17262 declared obsolete: 17263 * Matsushita MN10200, mn10200-*-* 17264 * Motorola 88000, m88k-*-* 17265 * IBM ROMP, romp-*-* 17266 17267 Also, some individual systems have been obsoleted: 17268 * Alpha 17269 + Interix, alpha*-*-interix* 17270 + Linux libc1, alpha*-*-linux*libc1* 17271 + Linux ECOFF, alpha*-*-linux*ecoff* 17272 * ARM 17273 + Generic a.out, arm*-*-aout* 17274 + Conix, arm*-*-conix* 17275 + "Old ABI," arm*-*-oabi 17276 + StrongARM/COFF, strongarm-*-coff* 17277 * HPPA (PA-RISC) 17278 + Generic OSF, hppa1.0-*-osf* 17279 + Generic BSD, hppa1.0-*-bsd* 17280 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 17281 + HiUX, hppa*-*-hiux* 17282 + Mach Lites, hppa*-*-lites* 17283 * Intel 386 family 17284 + Windows NT 3.x, i?86-*-win32 17285 * MC68000 family 17286 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 17287 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 17288 m68k-sun-mach* 17289 + AT&T systems, m68000-att-sysv* 17290 + Atari systems, m68k-atari-sysv* 17291 + Motorola systems, m68k-motorola-sysv* 17292 + NCR systems, m68k-ncr-sysv* 17293 + Plexus systems, m68k-plexus-sysv* 17294 + Commodore systems, m68k-cbm-sysv* 17295 + Citicorp TTI, m68k-tti-* 17296 + Unos, m68k-crds-unos* 17297 + Concurrent RTU, m68k-ccur-rtu* 17298 + Linux a.out, m68k-*-linux*aout* 17299 + Linux libc1, m68k-*-linux*libc1* 17300 + pSOS, m68k-*-psos* 17301 * MIPS 17302 + Generic ECOFF, mips*-*-ecoff* 17303 + SINIX, mips-sni-sysv4 17304 + Orion RTEMS, mips64orion-*-rtems* 17305 * National Semiconductor 32000 17306 + OpenBSD, ns32k-*-openbsd* 17307 * POWER (aka RS/6000) and PowerPC 17308 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 17309 + Bull BOSX, rs6000-bull-bosx 17310 + Generic Mach, rs6000-*-mach* 17311 + Generic SysV, powerpc*-*-sysv* 17312 + Linux libc1, powerpc*-*-linux*libc1* 17313 * Sun SPARC 17314 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 17315 sparclite-*-aout*, and sparc86x-*-aout* 17316 + NetBSD a.out, sparc-*-netbsd*aout* 17317 + Generic BSD, sparc-*-bsd* 17318 + ChorusOS, sparc-*-chorusos* 17319 + Linux a.out, sparc-*-linux*aout* 17320 + Linux libc1, sparc-*-linux*libc1* 17321 + LynxOS, sparc-*-lynxos* 17322 + Solaris on HAL hardware, sparc-hal-solaris2* 17323 + SunOS versions 3 and 4, sparc-*-sunos[34]* 17324 * NEC V850 17325 + RTEMS, v850-*-rtems* 17326 * VAX 17327 + VMS, vax-*-vms* 17328 17329Documentation improvements 17330 17331Other significant improvements 17332 17333 * Almost all front-end dependencies in the compiler have been 17334 separated out into a set of language hooks. This should make adding 17335 a new front end clearer and easier. 17336 * One effect of removing the separate preprocessor is a small 17337 increase in the robustness of the compiler in general, and the 17338 maintainability of target descriptions. Previously target-specific 17339 built-in macros and others, such as __FAST_MATH__, had to be 17340 handled with so-called specs that were hard to maintain. Often they 17341 would fail to behave properly when conflicting options were 17342 supplied on the command line, and define macros in the user's 17343 namespace even when strict ISO compliance was requested. 17344 Integrating the preprocessor has cleanly solved these issues. 17345 * The Makefile suite now supports redirection of make install by 17346 means of the variable DESTDIR. 17347 __________________________________________________________________ 17348 17349GCC 3.3 17350 17351 Detailed release notes for the GCC 3.3 release follow. 17352 17353 Bug Fixes 17354 17355 bootstrap failures 17356 17357 * [8]10140 cross compiler build failures: missing __mempcpy (DUP: 17358 [9]10198,[10]10338) 17359 17360 Internal compiler errors (multi-platform) 17361 17362 * [11]3581 large string causes segmentation fault in cc1 17363 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 17364 * [13]5533 (c++) ICE when processing std::accumulate(begin, end, 17365 init, invalid_op) 17366 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 17367 * [15]6412 (c++) ICE in retrieve_specialization 17368 * [16]6620 (c++) partial template specialization causes an ICE 17369 (segmentation fault) 17370 * [17]6663 (c++) ICE with attribute aligned 17371 * [18]7068 ICE with incomplete types 17372 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation 17373 * [20]7647 (c++) ICE when data member has the name of the enclosing 17374 class 17375 * [21]7675 ICE in fixup_var_refs_1 17376 * [22]7718 'complex' template instantiation causes ICE 17377 * [23]8116 (c++) ICE in member template function 17378 * [24]8358 (ada) Ada compiler accesses freed memory, crashes 17379 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 17380 fault 17381 * [26]8564 (c++) ICE in find_function_data, in function.c 17382 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 17383 * [28]8766 (c++) ICE after failed initialization of static template 17384 variable 17385 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c 17386 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 17387 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class 17388 definition 17389 * [32]9216 (c++) ICE on missing template parameter 17390 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c 17391 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 17392 loop 17393 * [35]9429 (c++) ICE in template instantiation with a pointered new 17394 operator 17395 * [36]9516 Internal error when using a big array 17396 * [37]9600 (c++) ICE with typedefs in template class 17397 * [38]9629 (c++) virtual inheritance segfault 17398 * [39]9672 (c++) ICE: Error reporting routines re-entered 17399 * [40]9749 (c++) ICE in write_expression on invalid function 17400 prototype 17401 * [41]9794 (fortran) ICE: floating point exception during constant 17402 folding 17403 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE 17404 * [43]9916 (c++) ICE with noreturn function in ?: statement 17405 * [44]9936 ICE with local function and variable-length 2d array 17406 * [45]10262 (c++) cc1plus crashes with large generated code 17407 * [46]10278 (c++) ICE in parser for invalid code 17408 * [47]10446 (c++) ICE on definition of nonexistent member function of 17409 nested class in a class template 17410 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable 17411 declaration 17412 * [49]10506 (c++) ICE in build_new at cp/init.c with 17413 -fkeep-inline-functions and multiple inheritance 17414 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the 17415 precision of the declared type 17416 17417 Optimization bugs 17418 17419 * [51]2001 Inordinately long compile times in reload CSE regs 17420 * [52]2391 Exponential compilation time explosion in combine 17421 * [53]2960 Duplicate loop conditions even with -Os 17422 * [54]4046 redundant conditional branch 17423 * [55]6405 Loop-unrolling related performance regressions 17424 * [56]6798 very long compile time with large case-statement 17425 * [57]6871 const objects shouldn't be moved to .bss 17426 * [58]6909 problem w/ -Os on modified loop-2c.c test case 17427 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of 17428 non-void function'' warning 17429 * [60]7642 optimization problem with signbit() 17430 * [61]8634 incorrect code for inlining of memcpy under -O2 17431 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as 17432 regular function call 17433 17434 C front end 17435 17436 * [63]2161 long if-else cascade overflows parser stack 17437 * [64]4319 short accepted on typedef'd char 17438 * [65]8602 incorrect line numbers in warning messages when using 17439 inline functions 17440 * [66]9177 -fdump-translation-unit: C front end deletes function_decl 17441 AST nodes and breaks debugging dumps 17442 * [67]9853 miscompilation of non-constant structure initializer 17443 17444 c++ compiler and library 17445 17446 * [68]45 legal template specialization code is rejected (DUP: 17447 [69]3784) 17448 * [70]764 lookup failure: friend operator and dereferencing a pointer 17449 and templates (DUP: [71]5116) 17450 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP: 17451 2863) 17452 * [73]3663 G++ doesn't check access control during template 17453 instantiation 17454 * [74]3797 gcc fails to emit explicit specialization of a template 17455 member 17456 * [75]3948 Two destructors are called when no copy destructor is 17457 defined (ABI change) 17458 * [76]4137 Conversion operator within template is not accepted 17459 * [77]4361 bogus ambiguity taking the address of a member template 17460 * [78]4802 g++ accepts illegal template code (access to private 17461 member; DUP: [79]5837) 17462 * [80]4803 inline function is used but never defined, and g++ does 17463 not object 17464 * [81]5094 Partial specialization cannot be friend? 17465 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 17466 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 17467 time 17468 * [84]7015 certain __asm__ constructs rejected 17469 * [85]7086 compile time regression (quadratic behavior in 17470 fixup_var_refs) 17471 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and 17472 std::abort 17473 * [87]7247 copy constructor missing when inlining enabled (invalid 17474 optimization?) 17475 * [88]7441 string array initialization compilation time regression 17476 from seconds to minutes 17477 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong 17478 * [90]7804 bad printing of floating point constant in warning message 17479 * [91]8099 Friend classes and template specializations 17480 * [92]8117 member function pointers and multiple inheritance 17481 * [93]8205 using declaration and multiple inheritance 17482 * [94]8645 unnecessary non-zero checks in stl_tree.h 17483 * [95]8724 explicit destructor call for incomplete class allowed 17484 * [96]8805 compile time regression with many member variables 17485 * [97]8691 -O3 and -fno-implicit-templates are incompatible 17486 * [98]8700 unhelpful error message for binding temp to reference 17487 * [99]8724 explicit destructor call for incomplete class allowed 17488 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems 17489 * [101]9016 Failure to consistently constant fold "constant" C++ 17490 objects 17491 * [102]9053 g++ confused about ambiguity of overloaded function 17492 templates 17493 * [103]9152 undefined virtual thunks 17494 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out 17495 * [105]9297 data corruption due to codegen bug (when copying.) 17496 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken 17497 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 17498 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in 17499 local classes 17500 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters 17501 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439) 17502 * [112]9474 GCC freezes in compiling a weird code mixing <iostream> 17503 and <iostream.h> 17504 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1) 17505 [114][DR 231] 17506 * [115]9555 ostream inserters fail to set badbit on exception 17507 * [116]9561 ostream inserters rethrow exception of wrong type 17508 * [117]9563 ostream::sentry returns true after a failed preparation 17509 * [118]9582 one-definition rule violation in std::allocator 17510 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors 17511 * [120]9683 bug in initialization chains for static const variables 17512 from template classes 17513 * [121]9791 -Woverloaded-virtual reports hiding of destructor 17514 * [122]9817 collate::compare doesn't handle nul characters 17515 * [123]9825 filebuf::sputbackc breaks sbumpc 17516 * [124]9826 operator>>(basic_istream, basic_string) fails to compile 17517 with custom traits 17518 * [125]9924 Multiple using statements for builtin functions not 17519 allowed 17520 * [126]9946 destructor is not called for temporary object 17521 * [127]9964 filebuf::close() sometimes fails to close file 17522 * [128]9988 filebuf::overflow writes EOF to file 17523 * [129]10033 optimization breaks polymorphic references w/ typeid 17524 operator 17525 * [130]10097 filebuf::underflow drops characters 17526 * [131]10132 filebuf destructor can throw exceptions 17527 * [132]10180 gcc fails to warn about non-inlined function 17528 * [133]10199 method parametrized by template does not work everywhere 17529 * [134]10300 use of array-new (nothrow) in segfaults on NULL return 17530 * [135]10427 Stack corruption with variable-length automatic arrays 17531 and virtual destructors 17532 * [136]10503 Compilation never stops in fixed_type_or_null 17533 17534 Objective-C 17535 17536 * [137]5956 selectors aren't matched properly when added to the 17537 selector table 17538 17539 Fortran compiler and library 17540 17541 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't 17542 detect 17543 * [139]3924 g77 generates code that is rejected by GAS if COFF debug 17544 info requested 17545 * [140]5634 doc: explain that configure --prefix=~/... does not work 17546 * [141]6367 multiple repeat counts confuse namelist read into array 17547 * [142]6491 Logical operations error on logicals when using 17548 -fugly-logint 17549 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C" 17550 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 17551 on irix6.5 17552 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 17553 assume a direct access file 17554 * [146]7278 g77 "bug"; the executable misbehaves (with -O2 17555 -fno-automatic) 17556 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows 17557 * [148]7388 Incorrect output with 0-based array of characters 17558 * [149]8587 Double complex zero ** double precision number -> NaN 17559 instead of zero 17560 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 17561 unknown register name line-length-none 17562 * [151]10197 Direct access files not unformatted by default 17563 17564 Java compiler and library 17565 17566 * [152]6005 gcj fails to build rhug on alpha 17567 * [153]6389 System.getProperty("") should always throw an 17568 IllegalArgumentException 17569 * [154]6576 java.util.ResourceBundle.getResource ignores locale 17570 * [155]6652 new java.io.File("").getCanonicalFile() throws exception 17571 * [156]7060 getMethod() doesn't search super interface 17572 * [157]7073 bytecode interpreter gives wrong answer for interface 17573 getSuperclass() 17574 * [158]7180 possible bug in 17575 javax.naming.spi.NamingManager.getPlusPath() 17576 * [159]7416 java.security startup refs "GNU libgcj.security" 17577 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent 17578 env (DUP: [161]7578) 17579 * [162]7611 Internal error while compiling libjava with -O 17580 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry 17581 * [164]7766 ZipInputStream.available returns 0 immediately after 17582 construction 17583 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 17584 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 17585 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens' 17586 * [168]8234 ZipInputStream chokes when InputStream.read() returns 17587 small chunks 17588 * [169]8415 reflection bug: exception info for Method 17589 * [170]8481 java.Random.nextInt(int) may return negative 17590 * [171]8593 Error reading GZIPped files with BufferedReader 17591 * [172]8759 java.beans.Introspector has no flushCaches() or 17592 flushFromCaches() methods 17593 * [173]8997 spin() calls Thread.sleep 17594 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 17595 instead of the root content of C: 17596 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns 17597 wrong return codes 17598 * [176]9271 Severe bias in java.security.SecureRandom 17599 17600 Ada compiler and library 17601 17602 * [177]6767 make gnatlib-shared fails on -laddr2line 17603 * [178]9911 gnatmake fails to link when GCC configured with 17604 --with-sjlj-exceptions=yes 17605 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled 17606 * [180]10546 Ada tasking not working on Red Hat 9 17607 17608 preprocessor 17609 17610 * [181]7029 preprocessor should ignore #warning with -M 17611 17612 ARM-specific 17613 17614 * [182]2903 [arm] Optimization bug with long long arithmetic 17615 * [183]7873 arm-linux-gcc fails when assigning address to a bit field 17616 17617 FreeBSD-specific 17618 17619 * [184]7680 float functions undefined in math.h/cmath with #define 17620 _XOPEN_SOURCE 17621 17622 HP-UX or HP-PA-specific 17623 17624 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 17625 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 17626 fputc_unlocked 17627 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 17628 17629 m68hc11-specific 17630 17631 * [188]6744 Bad assembler code generated: reference to pseudo 17632 register z 17633 * [189]7361 Internal compiler error in reload_cse_simplify_operands, 17634 in reload1.c 17635 17636 MIPS-specific 17637 17638 * [190]9496 [mips-linux] bug in optimizer? 17639 17640 PowerPC-specific 17641 17642 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 17643 space 17644 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux 17645 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg 17646 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c 17647 17648 SPARC-specific 17649 17650 * [195]10267 (documentation) Wrong build instructions for 17651 *-*-solaris2* 17652 17653 x86-specific (Intel/AMD) 17654 17655 * [196]7916 ICE in instantiate_virtual_register_1 17656 * [197]7926 (c++) i486 instructions in header files make c++ programs 17657 crash on i386 17658 * [198]8555 ICE in gen_split_1231 17659 * [199]8994 ICE with -O -march=pentium4 17660 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs 17661 * [201]9806 ICE in inline assembly with -fPIC flag 17662 * [202]10077 gcc -msse2 generates movd to move dwords between xmm 17663 regs 17664 * [203]10233 64-bit comparison only comparing bottom 32-bits 17665 * [204]10286 type-punning doesn't work with __m64 and -O 17666 * [205]10308 [x86] ICE with -O -fgcse or -O2 17667 __________________________________________________________________ 17668 17669GCC 3.3.1 17670 17671 Bug Fixes 17672 17673 This section lists the problem reports (PRs) from GCC's bug tracking 17674 system that are known to be fixed in the 3.3.1 release. This list might 17675 not be complete (that is, it is possible that some PRs that have been 17676 fixed are not listed here). 17677 17678 Bootstrap failures 17679 17680 * [206]11272 [Solaris] make bootstrap fails while building libstdc++ 17681 17682 Internal compiler errors (multi-platform) 17683 17684 * [207]5754 ICE on invalid nested template class 17685 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 17686 and --enable-checking 17687 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c 17688 * [210]7053 (c++) ICE when declaring a function already defined as a 17689 friend method of a template class 17690 * [211]8164 (c++) ICE when using different const expressions as 17691 template parameter 17692 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c 17693 * [213]9559 (c++) ICE with invalid initialization of a static const 17694 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 17695 when redeclaring a static member variable 17696 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in 17697 dwarfout.c with -g -O -finline-functions 17698 * [216]10432 (c++) ICE in poplevel, in cp/decl.c 17699 * [217]10475 ICE in subreg_highpart_offset for code with long long 17700 * [218]10635 (c++) ICE when dereferencing an incomplete type casted 17701 from a void pointer 17702 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 17703 instantiating static member variables 17704 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets 17705 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 17706 * [222]10796 (c++) ICE when defining an enum with two values: -1 and 17707 MAX_INT_64BIT 17708 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 17709 sched.c 17710 * [224]10939 (c++) ICE with template code 17711 * [225]10956 (c++) ICE when specializing a template member function 17712 of a template class, in tsubst, in cp/pt.c 17713 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*() 17714 defined) 17715 * [227]11059 (c++) ICE with empty union 17716 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 17717 -O2 -fnon-call-exceptions 17718 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type 17719 * [230]11149 (c++) ICE on error when instantiation with call function 17720 of a base type 17721 * [231]11228 (c++) ICE on new-expression using array operator new and 17722 default-initialization 17723 * [232]11282 (c++) Infinite memory usage after syntax error 17724 * [233]11301 (fortran) ICE with -fno-globals 17725 * [234]11308 (c++) ICE when using an enum type name as if it were a 17726 class or namespace 17727 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from 17728 an empty struct 17729 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR 17730 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 17731 template member functions 17732 17733 Optimization bugs 17734 17735 * [238]11198 -O2 -frename-registers generates wrong code (aliasing 17736 problem) 17737 * [239]11304 Wrong code production with -fomit-frame-pointer 17738 * [240]11381 volatile memory access optimized away 17739 * [241]11536 [strength-reduce] -O2 optimization produces wrong code 17740 * [242]11557 constant folding bug generates wrong code 17741 17742 C front end 17743 17744 * [243]5897 No warning for statement after return 17745 * [244]11279 DWARF-2 output mishandles large enums 17746 17747 Preprocessor bugs 17748 17749 * [245]11022 no warning for non-compatible macro redefinition 17750 17751 C++ compiler and library 17752 17753 * [246]2330 static_cast<>() to a private base is allowed 17754 * [247]5388 Incorrect message "operands to ?: have different types" 17755 * [248]5390 Libiberty fails to demangle multi-digit template 17756 parameters 17757 * [249]7877 Incorrect parameter passing to specializations of member 17758 function templates 17759 * [250]9393 Anonymous namespaces and compiling the same file twice 17760 * [251]10032 -pedantic converts some errors to warnings 17761 * [252]10468 const typeof(x) is non-const, but only in templates 17762 * [253]10527 confused error message with "new int()" parameter 17763 initializer 17764 * [254]10679 parameter MIN_INLINE_INSNS is not honored 17765 * [255]10682 gcc chokes on a typedef for an enum inside a class 17766 template 17767 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 17768 0. 17769 * [257]10845 template member function (with nested template as 17770 parameter) cannot be called anymore if another unrelated template 17771 member function is defined 17772 * [258]10849 Cannot define an out-of-class specialization of a 17773 private nested template class 17774 * [259]10888 Suppress -Winline warnings for system headers 17775 * [260]10929 -Winline warns about functions for which no definition 17776 is visible 17777 * [261]10931 valid conversion static_cast<const unsigned 17778 int&>(lvalue-of-type-int) is rejected 17779 * [262]10940 Bad code with explicit specialization 17780 * [263]10968 If member function implicitly instantiated, explicit 17781 instantiation of class fails to instantiate it 17782 * [264]10990 Cannot convert with dynamic_cast<> to a private base 17783 class from within a member function 17784 * [265]11039 Bad interaction between implicit typename deprecation 17785 and friendship 17786 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say 17787 "__unused__" instead 17788 * [267]11095 C++ iostream manipulator causes segfault when called 17789 with negative argument 17790 * [268]11098 g++ doesn't emit complete debugging information for 17791 local variables in destructors 17792 * [269]11137 GNU/Linux shared library constructors not called unless 17793 there's one global object 17794 * [270]11154 spurious ambiguity report for template class 17795 specialization 17796 * [271]11329 Compiler cannot find user defined implicit typecast 17797 * [272]11332 Spurious error with casts in ?: expression 17798 * [273]11431 static_cast behavior with subclasses when default 17799 constructor available 17800 * [274]11528 money_get facet does not accept "$.00" as valid 17801 * [275]11546 Type lookup problems in out-of-line definition of a 17802 class doubly nested from a template class 17803 * [276]11567 C++ code containing templated member function with same 17804 name as pure virtual member function results in linking failure 17805 * [277]11645 Failure to deal with using and private inheritance 17806 17807 Java compiler and library 17808 17809 * [278]5179 Qualified static field access doesn't initialize its 17810 class 17811 * [279]8204 gcj -O2 to native reorders certain instructions 17812 improperly 17813 * [280]10838 java.io.ObjectInputStream syntax error 17814 * [281]10886 The RMI registry that comes with GCJ does not work 17815 correctly 17816 * [282]11349 JNDI URL context factories not located correctly 17817 17818 x86-specific (Intel/AMD) 17819 17820 * [283]4823 ICE on inline assembly code 17821 * [284]8878 miscompilation with -O and SSE 17822 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3 17823 -masm=intel 17824 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 17825 in reload1.c 17826 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 17827 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source 17828 * [289]11044 [x86] out of range loop instructions for FP code on K6 17829 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE 17830 built-ins 17831 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 17832 is used 17833 17834 SPARC- or Solaris- specific 17835 17836 * [292]9362 solaris 'as' dies when fed .s and "-gstabs" 17837 * [293]10142 [SPARC64] gcc produces wrong code when passing 17838 structures by value 17839 * [294]10663 New configure check aborts with Sun tools. 17840 * [295]10835 combinatorial explosion in scheduler on HyperSPARC 17841 * [296]10876 ICE in calculate_giv_inc when building KDE 17842 * [297]10955 wrong code at -O3 for structure argument in context of 17843 structure return 17844 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25 17845 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 17846 Linux kernel 17847 17848 ia64 specific 17849 17850 * [300]10907 gcc violates the ia64 ABI (GP must be preserved) 17851 * [301]11320 scheduler bug (in machine depended reorganization pass) 17852 * [302]11599 bug with conditional and __builtin_prefetch 17853 17854 PowerPC specific 17855 17856 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 17857 during loop) 17858 * [304]10871 error in rs6000_stack_info save_size computation 17859 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 17860 cures it 17861 17862 m68k-specific 17863 17864 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx 17865 * [307]10557 [m68k] ICE in subreg_offset_representable_p 17866 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p 17867 17868 ARM-specific 17869 17870 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for 17871 functions with __attribute__ ((interrupt ("IRQ"))) 17872 * [310]10842 [arm] Clobbered link register is copied to pc under 17873 certain circumstances 17874 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes 17875 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 17876 (3.4) 17877 17878 MIPS-specific 17879 17880 * [313]11084 ICE in propagate_one_insn, in flow.c 17881 17882 SH-specific 17883 17884 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf 17885 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 17886 * [316]11096 i686-linux to sh-linux cross compiler fails to compile 17887 C++ files 17888 17889 GNU/Linux (or Hurd?) specific 17890 17891 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 17892 17893 UnixWare specific 17894 17895 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 17896 7.1.1 17897 17898 Cygwin (or mingw) specific 17899 17900 * [319]5287 ICE with dllimport attribute 17901 * [320]10148 [MingW/CygWin] Compiler dumps core 17902 17903 DJGPP specific 17904 17905 * [321]8787 GCC fails to emit .intel_syntax when invoked with 17906 -masm=intel on DJGPP 17907 17908 Darwin (and MacOS X) specific 17909 17910 * [322]10900 trampolines crash 17911 17912 Documentation 17913 17914 * [323]1607 (c++) Format attributes on methods undocumented 17915 * [324]4252 Invalid option `-fdump-translation-unit' 17916 * [325]4490 Clarify restrictions on -m96bit-long-double, 17917 -m128bit-long-double 17918 * [326]10355 document an issue with regparm attribute on some systems 17919 (e.g. Solaris) 17920 * [327]10726 (fortran) Documentation for function "IDate Intrinsic 17921 (Unix)" is wrong 17922 * [328]10805 document bug in old version of Sun assembler 17923 * [329]10815 warn against GNU binutils on AIX 17924 * [330]10877 document need for newer binutils on i?86-*-linux-gnu 17925 * [331]11280 Manual incorrect with respect to -freorder-blocks 17926 * [332]11466 Document -mlittle-endian and its restrictions for the 17927 sparc64 port 17928 17929 Testsuite bugs (compiler itself is not affected) 17930 17931 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 17932 report failure 17933 * [334]10810 gcc-3.3 fails make check: buffer overrun in 17934 test_demangle.c 17935 __________________________________________________________________ 17936 17937GCC 3.3.2 17938 17939 Bug Fixes 17940 17941 This section lists the problem reports (PRs) from GCC's bug tracker 17942 that are known to be fixed in the 3.3.2 release. This list might not be 17943 complete (that is, it is possible that some PRs that have been fixed 17944 are not listed here). 17945 17946 Bootstrap failures and problems 17947 17948 * [335]8336 [SCO5] bootstrap config still tries to use COFF options 17949 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 17950 --enable-threads=posix 17951 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 17952 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 17953 7.1.1) 17954 * [339]11687 xstormy16-elf build fails in libf2c 17955 * [340]12263 [SGI IRIX] bootstrap fails during compile of 17956 libf2c/libI77/backspace.c 17957 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9 17958 fix-header processing) 17959 17960 Internal compiler errors (multi-platform) 17961 17962 * [342]7277 Casting integers to vector types causes ICE 17963 * [343]7939 (c++) ICE on invalid function template specialization 17964 * [344]11063 (c++) ICE on parsing initialization list of const array 17965 member 17966 * [345]11207 ICE with negative index in array element designator 17967 * [346]11522 (fortran) g77 dwarf-2 ICE in 17968 add_abstract_origin_attribute 17969 * [347]11595 (c++) ICE on duplicate label definition 17970 * [348]11646 (c++) ICE in commit_one_edge_insertion with 17971 -fnon-call-exceptions -fgcse -O 17972 * [349]11665 ICE in struct initializer when taking address 17973 * [350]11852 (c++) ICE with bad struct initializer. 17974 * [351]11878 (c++) ICE in cp_expr_size 17975 * [352]11883 ICE with any -O on mercury-generated C code 17976 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 17977 cp/typeck2.c when applying typeid operator to template template 17978 parameter 17979 * [354]12146 ICE in lookup_template_function, in cp/pt.c 17980 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions 17981 -fno-gcse -O2 17982 * [356]12369 (c++) ICE with templates and friends 17983 * [357]12446 ICE in emit_move_insn on complicated array reference 17984 * [358]12510 ICE in final_scan_insn 17985 * [359]12544 ICE with large parameters used in nested functions 17986 17987 C and optimization bugs 17988 17989 * [360]9862 spurious warnings with -W -finline-functions 17990 * [361]10962 lookup_field is a linear search on a linked list (can be 17991 slow if large struct) 17992 * [362]11370 -Wunreachable-code gives false complaints 17993 * [363]11637 invalid assembly with -fnon-call-exceptions 17994 * [364]11885 Problem with bitfields in packed structs 17995 * [365]12082 Inappropriate unreachable code warnings 17996 * [366]12180 Inline optimization fails for variadic function 17997 * [367]12340 loop unroller + gcse produces wrong code 17998 17999 C++ compiler and library 18000 18001 * [368]3907 nested template parameter collides with member name 18002 * [369]5293 confusing message when binding a temporary to a reference 18003 * [370]5296 [DR115] Pointers to functions and to template functions 18004 behave differently in deduction 18005 * [371]7939 ICE on function template specialization 18006 * [372]8656 Unable to assign function with __attribute__ and pointer 18007 return type to an appropriate variable 18008 * [373]10147 Confusing error message for invalid template function 18009 argument 18010 * [374]11400 std::search_n() makes assumptions about Size parameter 18011 * [375]11409 issues with using declarations, overloading, and 18012 built-in functions 18013 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 18014 multiple bits in mask 18015 * [377]11786 operator() call on variable in other namespace not 18016 recognized 18017 * [378]11867 static_cast ignores ambiguity 18018 * [379]11928 bug with conversion operators that are typedefs 18019 * [380]12114 Uninitialized memory accessed in dtor 18020 * [381]12163 static_cast + explicit constructor regression 18021 * [382]12181 Wrong code with comma operator and c++ 18022 * [383]12236 regparm and fastcall messes up parameters 18023 * [384]12266 incorrect instantiation of unneeded template during 18024 overload resolution 18025 * [385]12296 istream::peek() doesn't set eofbit 18026 * [386]12298 [sjlj exceptions] Stack unwind destroys 18027 not-yet-constructed object 18028 * [387]12369 ICE with templates and friends 18029 * [388]12337 apparently infinite loop in g++ 18030 * [389]12344 stdcall attribute ignored if function returns a pointer 18031 * [390]12451 missing(late) class forward declaration in cxxabi.h 18032 * [391]12486 g++ accepts invalid use of a qualified name 18033 18034 x86 specific (Intel/AMD) 18035 18036 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX 18037 builtins 18038 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 18039 -O2 18040 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture 18041 * [395]12116 [k6] Invalid assembly output values with X-MAME code 18042 * [396]12070 ICE converting between double and long double with 18043 -msoft-float 18044 18045 ia64-specific 18046 18047 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 18048 * [398]11535 __builtin_return_address may not work on ia64 18049 * [399]11693 [ia64] ICE in gen_nop_type 18050 * [400]12224 [ia64] Thread-local storage doesn't work 18051 18052 PowerPC-specific 18053 18054 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 18055 kernel 18056 * [402]11319 loop miscompiled on ppc32 18057 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code 18058 18059 SPARC-specific 18060 18061 * [404]11662 wrong code for expr. with cast to long long and 18062 exclusive or 18063 * [405]11965 invalid assembler code for a shift < 32 operation 18064 * [406]12301 (c++) stack corruption when a returned expression throws 18065 an exception 18066 18067 Alpha-specific 18068 18069 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of 18070 kernel 2.4.22-pre8 18071 18072 HPUX-specific 18073 18074 * [408]11313 problem with #pragma weak and static inline functions 18075 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore? 18076 18077 Solaris specific 18078 18079 * [410]12166 Profiled programs crash if PROFDIR is set 18080 18081 Solaris-x86 specific 18082 18083 * [411]12101 i386 Solaris no longer works with GNU as? 18084 18085 Miscellaneous embedded target-specific bugs 18086 18087 * [412]10988 [m32r-elf] wrong blockmove code with -O3 18088 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 18089 -O2 18090 * [414]11902 [sh4] spec file improperly inserts rpath even when none 18091 needed 18092 * [415]11903 [sh4] -pthread fails to link due to error in spec file 18093 on sh4 18094 __________________________________________________________________ 18095 18096GCC 3.3.3 18097 18098 Minor features 18099 18100 In addition to the bug fixes documented below, this release contains 18101 few minor features such as: 18102 * Support for --with-sysroot 18103 * Support for automatic detection of executable stacks 18104 * Support for SSE3 instructions 18105 * Support for thread local storage debugging under GDB on S390 18106 18107 Bug Fixes 18108 18109 This section lists the problem reports (PRs) from GCC's bug tracker 18110 that are known to be fixed in the 3.3.3 release. This list might not be 18111 complete (that is, it is possible that some PRs that have been fixed 18112 are not listed here). 18113 18114 Bootstrap failures and issues 18115 18116 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 18117 * [417]12399 boehm-gc fails (when building a cross compiler): libtool 18118 unable to infer tagged configuration 18119 * [418]13068 mklibgcc.in doesn't handle multi-level multilib 18120 subdirectories properly 18121 18122 Internal compiler errors (multi-platform) 18123 18124 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to 18125 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 18126 * [420]10555 (c++) ICE on undefined template argument 18127 * [421]10706 (c++) ICE in mangle_class_name_for_template 18128 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops 18129 active 18130 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c 18131 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0 18132 * [425]12632 (fortran) -fbounds-check ICE 18133 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc 18134 3.3.2 18135 * [427]12726 (c++) ICE (segfault) on trivial code 18136 * [428]12890 (c++) ICE on compilation of class with throwing method 18137 * [429]12900 (c++) ICE in rtl_verify_flow_info_1 18138 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 18139 correct code with -O2 -fno-force-mem 18140 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive 18141 template 18142 * [432]13318 ICE: floating point exception in the loop optimizer 18143 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 18144 except.c 18145 * [434]13574 (c++) invalid array default initializer in class lets 18146 gcc consume all memory and die 18147 * [435]13475 ICE on SIMD variables with partial value initialization 18148 * [436]13797 (c++) ICE on invalid template parameter 18149 * [437]13824 (java) gcj SEGV with simple .java program 18150 18151 C and optimization bugs 18152 18153 * [438]8776 loop invariants are not removed (most likely) 18154 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 18155 strncmp by memcmp 18156 * [440]11350 undefined labels with -Os -fPIC 18157 * [441]12826 Optimizer removes reference through volatile pointer 18158 * [442]12500 stabs debug info: void no longer a predefined / builtin 18159 type 18160 * [443]12941 builtin-bitops-1.c miscompilation (latent bug) 18161 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix 18162 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled 18163 * [446]13507 spurious printf format warning 18164 * [447]13382 Type information for const pointer disappears during 18165 optimization. 18166 * [448]13394 noreturn attribute ignored on recursive invokation 18167 * [449]13400 Compiled code crashes storing to read-only location 18168 * [450]13521 Endless loop in calculate_global_regs_live 18169 18170 C++ compiler and library 18171 18172 Some of the bug fixes in this list were made to implement decisions 18173 that the ISO C++ standards committee has made concerning several defect 18174 reports (DRs). Links in the list below point to detailed discussion of 18175 the relevant defect report. 18176 * [451]2094 unimplemented: use of `ptrmem_cst' in template type 18177 unification 18178 * [452]2294 using declaration confusion 18179 * [453]5050 template instantiation depth exceeds limit: recursion 18180 problem? 18181 * [454]9371 Bad exception handling in 18182 i/ostream::operator>>/<<(streambuf*) 18183 * [455]9546 bad exception handling in ostream members 18184 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the 18185 face of unknown locales 18186 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work 18187 * [459]10095 istream::operator>>(int&) sets ios::badbit when 18188 ios::failbit is set. 18189 * [460]11554 Warning about reordering of initializers doesn't mention 18190 location of constructor 18191 * [461]12297 istream::sentry::sentry() handles eof() incorrectly. 18192 * [462]12352 Exception safety problems in src/localename.cc 18193 * [463]12438 Memory leak in locale::combine() 18194 * [464]12540 Memory leak in locale::locale(const char*) 18195 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented 18196 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented 18197 * [470]12696 memory eating infinite loop in diagnostics (error 18198 recovery problem) 18199 * [471]12815 Code compiled with optimization behaves unexpectedly 18200 * [472]12862 Conflicts between typedefs/enums and namespace member 18201 declarations 18202 * [473]12926 Wrong value after assignment in initialize list using 18203 bit-fields 18204 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented 18205 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented 18206 * [478]13007 basic_streambuf::pubimbue, imbue wrong 18207 * [479]13009 Implicitly-defined assignment operator writes to wrong 18208 memory 18209 * [480]13057 regparm attribute not applied to destructor 18210 * [481]13070 -Wformat option ignored in g++ 18211 * [482]13081 forward template declarations in <complex> let inlining 18212 fail 18213 * [483]13239 Assertion does not seem to work correctly anymore 18214 * [484]13262 "xxx is private within this context" when initializing a 18215 self-contained template class 18216 * [485]13290 simple typo in concept checking for std::generate_n 18217 * [486]13323 Template code does not compile in presence of typedef 18218 * [487]13369 __verify_grouping (and __add_grouping?) not correct 18219 * [488]13371 infinite loop with packed struct and inlining 18220 * [489]13445 Template argument replacement "dereferences" a typedef 18221 * [490]13461 Fails to access protected-ctor from public constant 18222 * [491]13462 Non-standard-conforming type set::pointer 18223 * [492]13478 gcc uses wrong constructor to initialize a const 18224 reference 18225 * [493]13544 "conflicting types" for enums in different scopes 18226 * [494]13650 string::compare should not (always) use 18227 traits_type::length() 18228 * [495]13683 bogus warning about passing non-PODs through ellipsis 18229 * [496]13688 Derived class is denied access to protected base class 18230 member class 18231 * [497]13774 Member variable cleared in virtual multiple inheritance 18232 class 18233 * [498]13884 Protect sstream.tcc from extern template use 18234 18235 Java compiler and library 18236 18237 * [499]10746 [win32] garbage collection crash in GCJ 18238 18239 Objective-C compiler and library 18240 18241 * [500]11433 Crash due to dereferencing null pointer when querying 18242 protocol 18243 18244 Fortran compiler and library 18245 18246 * [501]12633 logical expression gives incorrect result with 18247 -fugly-logint option 18248 * [502]13037 [gcse-lm] g77 generates incorrect code 18249 * [503]13213 Hex constant problem when compiling with -fugly-logint 18250 and -ftypeless-boz 18251 18252 x86-specific (Intel/AMD) 18253 18254 * [504]4490 ICE with -m128bit-long-double 18255 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have 18256 `reg' in make_field_assignment, in combine.c 18257 * [506]12441 ICE: can't find a register to spill 18258 * [507]12943 array static-init failure under -fpic, -fPIC 18259 * [508]13608 Incorrect code with -O3 -ffast-math 18260 18261 PowerPC-specific 18262 18263 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of 18264 __attribute__((aligned(16))) 18265 * [510]11793 ICE in extract_insn, in recog.c (const_vector's) 18266 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in 18267 altivec.md) 18268 * [512]12537 g++ generates writeable text sections 18269 18270 SPARC-specific 18271 18272 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0 18273 -m64 18274 * [514]12865 mprotect call to make trampoline executable may fail 18275 * [515]13354 ICE in sparc_emit_set_const32 18276 18277 ARM-specific 18278 18279 * [516]10467 [arm] ICE in pre_insert_copy_insn, 18280 18281 ia64-specific 18282 18283 * [517]11226 ICE passing struct arg with two floats 18284 * [518]11227 ICE for _Complex float, _Complex long double args 18285 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64 18286 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn 18287 * Various fixes for libunwind 18288 18289 Alpha-specific 18290 18291 * [521]12654 Incorrect comparison code generated for Alpha 18292 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 18293 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 18294 18295 HPPA-specific 18296 18297 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 18298 * [525]12158 [hppa] compilation does not terminate at -O1 18299 18300 S390-specific 18301 18302 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only 18303 (1<<24)-1 possible for CLCL-Instruction 18304 18305 SH-specific 18306 18307 * [527]9365 segfault in gen_far_branch (config/sh/sh.c) 18308 * [528]10392 optimizer generates faulty array indexing 18309 * [529]11322 SH profiler outputs multiple definitions of symbol 18310 * [530]13069 gcc/config/sh/rtems.h broken 18311 * [531]13302 Putting a va_list in a struct causes seg fault 18312 * [532]13585 Incorrect optimization of call to sfunc 18313 * Fix inappropriately exported libgcc functions from the shared 18314 library 18315 18316 Other embedded target specific 18317 18318 * [533]8916 [mcore] unsigned char assign gets hosed. 18319 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c 18320 * [535]13122 [h8300] local variable gets corrupted by function call 18321 when -fomit-frame-pointer is given 18322 * [536]13256 [cris] strict_low_part mistreated in delay slots 18323 * [537]13373 [mcore] optimization with -frerun-cse-after-loop 18324 -fexpensive-optimizations produces wrong code on mcore 18325 18326 GNU HURD-specific 18327 18328 * [538]12561 gcc/config/t-gnu needs updating to work with 18329 --with-sysroot 18330 18331 Tru64 Unix specific 18332 18333 * [539]6243 testsuite fails almost all tests due to no libintl in 18334 LD_LIBRARY_PATH during test. 18335 * [540]11397 weak aliases broken on Tru64 UNIX 18336 18337 AIX-specific 18338 18339 * [541]12505 build failure due to defines of uchar in cpphash.h and 18340 sys/types.h 18341 * [542]13150 WEAK symbols not exported by collect2 18342 18343 IRIX-specific 18344 18345 * [543]12666 fixincludes problem on IRIX 6.5.19m 18346 18347 Solaris-specific 18348 18349 * [544]12969 Including sys/byteorder.h breaks configure checks 18350 18351 Testsuite problems (compiler is not affected) 18352 18353 * [545]10819 testsuite creates CR+LF on compiler version lines in 18354 test summary files 18355 * [546]11612 abi_check not finding correct libgcc_s.so.1 18356 18357 Miscellaneous 18358 18359 * [547]13211 using -###, incorrect warnings about unused linker file 18360 are produced 18361 __________________________________________________________________ 18362 18363GCC 3.3.4 18364 18365 This is the [548]list of problem reports (PRs) from GCC's bug tracking 18366 system that are known to be fixed in the 3.3.4 release. This list might 18367 not be complete (that is, it is possible that some PRs that have been 18368 fixed are not listed here). 18369 __________________________________________________________________ 18370 18371GCC 3.3.5 18372 18373 This is the [549]list of problem reports (PRs) from GCC's bug tracking 18374 system that are known to be fixed in the 3.3.5 release. This list might 18375 not be complete (that is, it is possible that some PRs that have been 18376 fixed are not listed here). 18377 __________________________________________________________________ 18378 18379GCC 3.3.6 18380 18381 This is the [550]list of problem reports (PRs) from GCC's bug tracking 18382 system that are known to be fixed in the 3.3.6 release. This list might 18383 not be complete (that is, it is possible that some PRs that have been 18384 fixed are not listed here). 18385 18386 18387 For questions related to the use of GCC, please consult these web 18388 pages and the [551]GCC manuals. If that fails, the 18389 [552]gcc-help@gcc.gnu.org mailing list might help. Comments on these 18390 web pages and the development of GCC are welcome on our developer 18391 list at [553]gcc@gcc.gnu.org. All of [554]our lists have public 18392 archives. 18393 18394 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and 18395 distribution of this entire article is permitted in any medium, 18396 provided this notice is preserved. 18397 18398 These pages are [556]maintained by the GCC team. Last modified 18399 2021-07-28[557]. 18400 18401References 18402 18403 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 18404 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 18405 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 18406 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 18407 5. http://gcc.gnu.org/news/dfa.html 18408 6. http://gcc.gnu.org/c99status.html 18409 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 18410 8. https://gcc.gnu.org/PR10140 18411 9. https://gcc.gnu.org/PR10198 18412 10. https://gcc.gnu.org/PR10338 18413 11. https://gcc.gnu.org/PR3581 18414 12. https://gcc.gnu.org/PR4382 18415 13. https://gcc.gnu.org/PR5533 18416 14. https://gcc.gnu.org/PR6387 18417 15. https://gcc.gnu.org/PR6412 18418 16. https://gcc.gnu.org/PR6620 18419 17. https://gcc.gnu.org/PR6663 18420 18. https://gcc.gnu.org/PR7068 18421 19. https://gcc.gnu.org/PR7083 18422 20. https://gcc.gnu.org/PR7647 18423 21. https://gcc.gnu.org/PR7675 18424 22. https://gcc.gnu.org/PR7718 18425 23. https://gcc.gnu.org/PR8116 18426 24. https://gcc.gnu.org/PR8358 18427 25. https://gcc.gnu.org/PR8511 18428 26. https://gcc.gnu.org/PR8564 18429 27. https://gcc.gnu.org/PR8660 18430 28. https://gcc.gnu.org/PR8766 18431 29. https://gcc.gnu.org/PR8803 18432 30. https://gcc.gnu.org/PR8846 18433 31. https://gcc.gnu.org/PR8906 18434 32. https://gcc.gnu.org/PR9216 18435 33. https://gcc.gnu.org/PR9261 18436 34. https://gcc.gnu.org/PR9263 18437 35. https://gcc.gnu.org/PR9429 18438 36. https://gcc.gnu.org/PR9516 18439 37. https://gcc.gnu.org/PR9600 18440 38. https://gcc.gnu.org/PR9629 18441 39. https://gcc.gnu.org/PR9672 18442 40. https://gcc.gnu.org/PR9749 18443 41. https://gcc.gnu.org/PR9794 18444 42. https://gcc.gnu.org/PR9829 18445 43. https://gcc.gnu.org/PR9916 18446 44. https://gcc.gnu.org/PR9936 18447 45. https://gcc.gnu.org/PR10262 18448 46. https://gcc.gnu.org/PR10278 18449 47. https://gcc.gnu.org/PR10446 18450 48. https://gcc.gnu.org/PR10451 18451 49. https://gcc.gnu.org/PR10506 18452 50. https://gcc.gnu.org/PR10549 18453 51. https://gcc.gnu.org/PR2001 18454 52. https://gcc.gnu.org/PR2391 18455 53. https://gcc.gnu.org/PR2960 18456 54. https://gcc.gnu.org/PR4046 18457 55. https://gcc.gnu.org/PR6405 18458 56. https://gcc.gnu.org/PR6798 18459 57. https://gcc.gnu.org/PR6871 18460 58. https://gcc.gnu.org/PR6909 18461 59. https://gcc.gnu.org/PR7189 18462 60. https://gcc.gnu.org/PR7642 18463 61. https://gcc.gnu.org/PR8634 18464 62. https://gcc.gnu.org/PR8750 18465 63. https://gcc.gnu.org/PR2161 18466 64. https://gcc.gnu.org/PR4319 18467 65. https://gcc.gnu.org/PR8602 18468 66. https://gcc.gnu.org/PR9177 18469 67. https://gcc.gnu.org/PR9853 18470 68. https://gcc.gnu.org/PR45 18471 69. https://gcc.gnu.org/PR3784 18472 70. https://gcc.gnu.org/PR764 18473 71. https://gcc.gnu.org/PR5116 18474 72. https://gcc.gnu.org/PR2862 18475 73. https://gcc.gnu.org/PR3663 18476 74. https://gcc.gnu.org/PR3797 18477 75. https://gcc.gnu.org/PR3948 18478 76. https://gcc.gnu.org/PR4137 18479 77. https://gcc.gnu.org/PR4361 18480 78. https://gcc.gnu.org/PR4802 18481 79. https://gcc.gnu.org/PR5837 18482 80. https://gcc.gnu.org/PR4803 18483 81. https://gcc.gnu.org/PR5094 18484 82. https://gcc.gnu.org/PR5730 18485 83. https://gcc.gnu.org/PR6713 18486 84. https://gcc.gnu.org/PR7015 18487 85. https://gcc.gnu.org/PR7086 18488 86. https://gcc.gnu.org/PR7099 18489 87. https://gcc.gnu.org/PR7247 18490 88. https://gcc.gnu.org/PR7441 18491 89. https://gcc.gnu.org/PR7768 18492 90. https://gcc.gnu.org/PR7804 18493 91. https://gcc.gnu.org/PR8099 18494 92. https://gcc.gnu.org/PR8117 18495 93. https://gcc.gnu.org/PR8205 18496 94. https://gcc.gnu.org/PR8645 18497 95. https://gcc.gnu.org/PR8724 18498 96. https://gcc.gnu.org/PR8805 18499 97. https://gcc.gnu.org/PR8691 18500 98. https://gcc.gnu.org/PR8700 18501 99. https://gcc.gnu.org/PR8724 18502 100. https://gcc.gnu.org/PR8949 18503 101. https://gcc.gnu.org/PR9016 18504 102. https://gcc.gnu.org/PR9053 18505 103. https://gcc.gnu.org/PR9152 18506 104. https://gcc.gnu.org/PR9182 18507 105. https://gcc.gnu.org/PR9297 18508 106. https://gcc.gnu.org/PR9318 18509 107. https://gcc.gnu.org/PR9320 18510 108. https://gcc.gnu.org/PR9400 18511 109. https://gcc.gnu.org/PR9424 18512 110. https://gcc.gnu.org/PR9425 18513 111. https://gcc.gnu.org/PR9439 18514 112. https://gcc.gnu.org/PR9474 18515 113. https://gcc.gnu.org/PR9548 18516 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 18517 115. https://gcc.gnu.org/PR9555 18518 116. https://gcc.gnu.org/PR9561 18519 117. https://gcc.gnu.org/PR9563 18520 118. https://gcc.gnu.org/PR9582 18521 119. https://gcc.gnu.org/PR9622 18522 120. https://gcc.gnu.org/PR9683 18523 121. https://gcc.gnu.org/PR9791 18524 122. https://gcc.gnu.org/PR9817 18525 123. https://gcc.gnu.org/PR9825 18526 124. https://gcc.gnu.org/PR9826 18527 125. https://gcc.gnu.org/PR9924 18528 126. https://gcc.gnu.org/PR9946 18529 127. https://gcc.gnu.org/PR9964 18530 128. https://gcc.gnu.org/PR9988 18531 129. https://gcc.gnu.org/PR10033 18532 130. https://gcc.gnu.org/PR10097 18533 131. https://gcc.gnu.org/PR10132 18534 132. https://gcc.gnu.org/PR10180 18535 133. https://gcc.gnu.org/PR10199 18536 134. https://gcc.gnu.org/PR10300 18537 135. https://gcc.gnu.org/PR10427 18538 136. https://gcc.gnu.org/PR10503 18539 137. https://gcc.gnu.org/PR5956 18540 138. https://gcc.gnu.org/PR1832 18541 139. https://gcc.gnu.org/PR3924 18542 140. https://gcc.gnu.org/PR5634 18543 141. https://gcc.gnu.org/PR6367 18544 142. https://gcc.gnu.org/PR6491 18545 143. https://gcc.gnu.org/PR6742 18546 144. https://gcc.gnu.org/PR7113 18547 145. https://gcc.gnu.org/PR7236 18548 146. https://gcc.gnu.org/PR7278 18549 147. https://gcc.gnu.org/PR7384 18550 148. https://gcc.gnu.org/PR7388 18551 149. https://gcc.gnu.org/PR8587 18552 150. https://gcc.gnu.org/PR9038 18553 151. https://gcc.gnu.org/PR10197 18554 152. https://gcc.gnu.org/PR6005 18555 153. https://gcc.gnu.org/PR6389 18556 154. https://gcc.gnu.org/PR6576 18557 155. https://gcc.gnu.org/PR6652 18558 156. https://gcc.gnu.org/PR7060 18559 157. https://gcc.gnu.org/PR7073 18560 158. https://gcc.gnu.org/PR7180 18561 159. https://gcc.gnu.org/PR7416 18562 160. https://gcc.gnu.org/PR7570 18563 161. https://gcc.gnu.org/PR7578 18564 162. https://gcc.gnu.org/PR7611 18565 163. https://gcc.gnu.org/PR7709 18566 164. https://gcc.gnu.org/PR7766 18567 165. https://gcc.gnu.org/PR7785 18568 166. https://gcc.gnu.org/PR7786 18569 167. https://gcc.gnu.org/PR8142 18570 168. https://gcc.gnu.org/PR8234 18571 169. https://gcc.gnu.org/PR8415 18572 170. https://gcc.gnu.org/PR8481 18573 171. https://gcc.gnu.org/PR8593 18574 172. https://gcc.gnu.org/PR8759 18575 173. https://gcc.gnu.org/PR8997 18576 174. https://gcc.gnu.org/PR9253 18577 175. https://gcc.gnu.org/PR9254 18578 176. https://gcc.gnu.org/PR9271 18579 177. https://gcc.gnu.org/PR6767 18580 178. https://gcc.gnu.org/PR9911 18581 179. https://gcc.gnu.org/PR10020 18582 180. https://gcc.gnu.org/PR10546 18583 181. https://gcc.gnu.org/PR7029 18584 182. https://gcc.gnu.org/PR2903 18585 183. https://gcc.gnu.org/PR7873 18586 184. https://gcc.gnu.org/PR7680 18587 185. https://gcc.gnu.org/PR8705 18588 186. https://gcc.gnu.org/PR9986 18589 187. https://gcc.gnu.org/PR10056 18590 188. https://gcc.gnu.org/PR6744 18591 189. https://gcc.gnu.org/PR7361 18592 190. https://gcc.gnu.org/PR9496 18593 191. https://gcc.gnu.org/PR7067 18594 192. https://gcc.gnu.org/PR8480 18595 193. https://gcc.gnu.org/PR8784 18596 194. https://gcc.gnu.org/PR10315 18597 195. https://gcc.gnu.org/PR10267 18598 196. https://gcc.gnu.org/PR7916 18599 197. https://gcc.gnu.org/PR7926 18600 198. https://gcc.gnu.org/PR8555 18601 199. https://gcc.gnu.org/PR8994 18602 200. https://gcc.gnu.org/PR9426 18603 201. https://gcc.gnu.org/PR9806 18604 202. https://gcc.gnu.org/PR10077 18605 203. https://gcc.gnu.org/PR10233 18606 204. https://gcc.gnu.org/PR10286 18607 205. https://gcc.gnu.org/PR10308 18608 206. https://gcc.gnu.org/PR11272 18609 207. https://gcc.gnu.org/PR5754 18610 208. https://gcc.gnu.org/PR6597 18611 209. https://gcc.gnu.org/PR6949 18612 210. https://gcc.gnu.org/PR7053 18613 211. https://gcc.gnu.org/PR8164 18614 212. https://gcc.gnu.org/PR8384 18615 213. https://gcc.gnu.org/PR9559 18616 214. https://gcc.gnu.org/PR9649 18617 215. https://gcc.gnu.org/PR9864 18618 216. https://gcc.gnu.org/PR10432 18619 217. https://gcc.gnu.org/PR10475 18620 218. https://gcc.gnu.org/PR10635 18621 219. https://gcc.gnu.org/PR10661 18622 220. https://gcc.gnu.org/PR10700 18623 221. https://gcc.gnu.org/PR10712 18624 222. https://gcc.gnu.org/PR10796 18625 223. https://gcc.gnu.org/PR10890 18626 224. https://gcc.gnu.org/PR10939 18627 225. https://gcc.gnu.org/PR10956 18628 226. https://gcc.gnu.org/PR11041 18629 227. https://gcc.gnu.org/PR11059 18630 228. https://gcc.gnu.org/PR11083 18631 229. https://gcc.gnu.org/PR11105 18632 230. https://gcc.gnu.org/PR11149 18633 231. https://gcc.gnu.org/PR11228 18634 232. https://gcc.gnu.org/PR11282 18635 233. https://gcc.gnu.org/PR11301 18636 234. https://gcc.gnu.org/PR11308 18637 235. https://gcc.gnu.org/PR11473 18638 236. https://gcc.gnu.org/PR11503 18639 237. https://gcc.gnu.org/PR11513 18640 238. https://gcc.gnu.org/PR11198 18641 239. https://gcc.gnu.org/PR11304 18642 240. https://gcc.gnu.org/PR11381 18643 241. https://gcc.gnu.org/PR11536 18644 242. https://gcc.gnu.org/PR11557 18645 243. https://gcc.gnu.org/PR5897 18646 244. https://gcc.gnu.org/PR11279 18647 245. https://gcc.gnu.org/PR11022 18648 246. https://gcc.gnu.org/PR2330 18649 247. https://gcc.gnu.org/PR5388 18650 248. https://gcc.gnu.org/PR5390 18651 249. https://gcc.gnu.org/PR7877 18652 250. https://gcc.gnu.org/PR9393 18653 251. https://gcc.gnu.org/PR10032 18654 252. https://gcc.gnu.org/PR10468 18655 253. https://gcc.gnu.org/PR10527 18656 254. https://gcc.gnu.org/PR10679 18657 255. https://gcc.gnu.org/PR10682 18658 256. https://gcc.gnu.org/PR10689 18659 257. https://gcc.gnu.org/PR10845 18660 258. https://gcc.gnu.org/PR10849 18661 259. https://gcc.gnu.org/PR10888 18662 260. https://gcc.gnu.org/PR10929 18663 261. https://gcc.gnu.org/PR10931 18664 262. https://gcc.gnu.org/PR10940 18665 263. https://gcc.gnu.org/PR10968 18666 264. https://gcc.gnu.org/PR10990 18667 265. https://gcc.gnu.org/PR11039 18668 266. https://gcc.gnu.org/PR11062 18669 267. https://gcc.gnu.org/PR11095 18670 268. https://gcc.gnu.org/PR11098 18671 269. https://gcc.gnu.org/PR11137 18672 270. https://gcc.gnu.org/PR11154 18673 271. https://gcc.gnu.org/PR11329 18674 272. https://gcc.gnu.org/PR11332 18675 273. https://gcc.gnu.org/PR11431 18676 274. https://gcc.gnu.org/PR11528 18677 275. https://gcc.gnu.org/PR11546 18678 276. https://gcc.gnu.org/PR11567 18679 277. https://gcc.gnu.org/PR11645 18680 278. https://gcc.gnu.org/PR5179 18681 279. https://gcc.gnu.org/PR8204 18682 280. https://gcc.gnu.org/PR10838 18683 281. https://gcc.gnu.org/PR10886 18684 282. https://gcc.gnu.org/PR11349 18685 283. https://gcc.gnu.org/PR4823 18686 284. https://gcc.gnu.org/PR8878 18687 285. https://gcc.gnu.org/PR9815 18688 286. https://gcc.gnu.org/PR10402 18689 287. https://gcc.gnu.org/PR10504 18690 288. https://gcc.gnu.org/PR10673 18691 289. https://gcc.gnu.org/PR11044 18692 290. https://gcc.gnu.org/PR11089 18693 291. https://gcc.gnu.org/PR11420 18694 292. https://gcc.gnu.org/PR9362 18695 293. https://gcc.gnu.org/PR10142 18696 294. https://gcc.gnu.org/PR10663 18697 295. https://gcc.gnu.org/PR10835 18698 296. https://gcc.gnu.org/PR10876 18699 297. https://gcc.gnu.org/PR10955 18700 298. https://gcc.gnu.org/PR11018 18701 299. https://gcc.gnu.org/PR11556 18702 300. https://gcc.gnu.org/PR10907 18703 301. https://gcc.gnu.org/PR11320 18704 302. https://gcc.gnu.org/PR11599 18705 303. https://gcc.gnu.org/PR9745 18706 304. https://gcc.gnu.org/PR10871 18707 305. https://gcc.gnu.org/PR11440 18708 306. https://gcc.gnu.org/PR7594 18709 307. https://gcc.gnu.org/PR10557 18710 308. https://gcc.gnu.org/PR11054 18711 309. https://gcc.gnu.org/PR10834 18712 310. https://gcc.gnu.org/PR10842 18713 311. https://gcc.gnu.org/PR11052 18714 312. https://gcc.gnu.org/PR11183 18715 313. https://gcc.gnu.org/PR11084 18716 314. https://gcc.gnu.org/PR10331 18717 315. https://gcc.gnu.org/PR10413 18718 316. https://gcc.gnu.org/PR11096 18719 317. https://gcc.gnu.org/PR2873 18720 318. https://gcc.gnu.org/PR3163 18721 319. https://gcc.gnu.org/PR5287 18722 320. https://gcc.gnu.org/PR10148 18723 321. https://gcc.gnu.org/PR8787 18724 322. https://gcc.gnu.org/PR10900 18725 323. https://gcc.gnu.org/PR1607 18726 324. https://gcc.gnu.org/PR4252 18727 325. https://gcc.gnu.org/PR4490 18728 326. https://gcc.gnu.org/PR10355 18729 327. https://gcc.gnu.org/PR10726 18730 328. https://gcc.gnu.org/PR10805 18731 329. https://gcc.gnu.org/PR10815 18732 330. https://gcc.gnu.org/PR10877 18733 331. https://gcc.gnu.org/PR11280 18734 332. https://gcc.gnu.org/PR11466 18735 333. https://gcc.gnu.org/PR10737 18736 334. https://gcc.gnu.org/PR10810 18737 335. https://gcc.gnu.org/PR8336 18738 336. https://gcc.gnu.org/PR9330 18739 337. https://gcc.gnu.org/PR9631 18740 338. https://gcc.gnu.org/PR9877 18741 339. https://gcc.gnu.org/PR11687 18742 340. https://gcc.gnu.org/PR12263 18743 341. https://gcc.gnu.org/PR12490 18744 342. https://gcc.gnu.org/PR7277 18745 343. https://gcc.gnu.org/PR7939 18746 344. https://gcc.gnu.org/PR11063 18747 345. https://gcc.gnu.org/PR11207 18748 346. https://gcc.gnu.org/PR11522 18749 347. https://gcc.gnu.org/PR11595 18750 348. https://gcc.gnu.org/PR11646 18751 349. https://gcc.gnu.org/PR11665 18752 350. https://gcc.gnu.org/PR11852 18753 351. https://gcc.gnu.org/PR11878 18754 352. https://gcc.gnu.org/PR11883 18755 353. https://gcc.gnu.org/PR11991 18756 354. https://gcc.gnu.org/PR12146 18757 355. https://gcc.gnu.org/PR12215 18758 356. https://gcc.gnu.org/PR12369 18759 357. https://gcc.gnu.org/PR12446 18760 358. https://gcc.gnu.org/PR12510 18761 359. https://gcc.gnu.org/PR12544 18762 360. https://gcc.gnu.org/PR9862 18763 361. https://gcc.gnu.org/PR10962 18764 362. https://gcc.gnu.org/PR11370 18765 363. https://gcc.gnu.org/PR11637 18766 364. https://gcc.gnu.org/PR11885 18767 365. https://gcc.gnu.org/PR12082 18768 366. https://gcc.gnu.org/PR12180 18769 367. https://gcc.gnu.org/PR12340 18770 368. https://gcc.gnu.org/PR3907 18771 369. https://gcc.gnu.org/PR5293 18772 370. https://gcc.gnu.org/PR5296 18773 371. https://gcc.gnu.org/PR7939 18774 372. https://gcc.gnu.org/PR8656 18775 373. https://gcc.gnu.org/PR10147 18776 374. https://gcc.gnu.org/PR11400 18777 375. https://gcc.gnu.org/PR11409 18778 376. https://gcc.gnu.org/PR11740 18779 377. https://gcc.gnu.org/PR11786 18780 378. https://gcc.gnu.org/PR11867 18781 379. https://gcc.gnu.org/PR11928 18782 380. https://gcc.gnu.org/PR12114 18783 381. https://gcc.gnu.org/PR12163 18784 382. https://gcc.gnu.org/PR12181 18785 383. https://gcc.gnu.org/PR12236 18786 384. https://gcc.gnu.org/PR12266 18787 385. https://gcc.gnu.org/PR12296 18788 386. https://gcc.gnu.org/PR12298 18789 387. https://gcc.gnu.org/PR12369 18790 388. https://gcc.gnu.org/PR12337 18791 389. https://gcc.gnu.org/PR12344 18792 390. https://gcc.gnu.org/PR12451 18793 391. https://gcc.gnu.org/PR12486 18794 392. https://gcc.gnu.org/PR8869 18795 393. https://gcc.gnu.org/PR9786 18796 394. https://gcc.gnu.org/PR11689 18797 395. https://gcc.gnu.org/PR12116 18798 396. https://gcc.gnu.org/PR12070 18799 397. https://gcc.gnu.org/PR11184 18800 398. https://gcc.gnu.org/PR11535 18801 399. https://gcc.gnu.org/PR11693 18802 400. https://gcc.gnu.org/PR12224 18803 401. https://gcc.gnu.org/PR11087 18804 402. https://gcc.gnu.org/PR11319 18805 403. https://gcc.gnu.org/PR11949 18806 404. https://gcc.gnu.org/PR11662 18807 405. https://gcc.gnu.org/PR11965 18808 406. https://gcc.gnu.org/PR12301 18809 407. https://gcc.gnu.org/PR11717 18810 408. https://gcc.gnu.org/PR11313 18811 409. https://gcc.gnu.org/PR11712 18812 410. https://gcc.gnu.org/PR12166 18813 411. https://gcc.gnu.org/PR12101 18814 412. https://gcc.gnu.org/PR10988 18815 413. https://gcc.gnu.org/PR11805 18816 414. https://gcc.gnu.org/PR11902 18817 415. https://gcc.gnu.org/PR11903 18818 416. https://gcc.gnu.org/PR11890 18819 417. https://gcc.gnu.org/PR12399 18820 418. https://gcc.gnu.org/PR13068 18821 419. https://gcc.gnu.org/PR10060 18822 420. https://gcc.gnu.org/PR10555 18823 421. https://gcc.gnu.org/PR10706 18824 422. https://gcc.gnu.org/PR11496 18825 423. https://gcc.gnu.org/PR11741 18826 424. https://gcc.gnu.org/PR12440 18827 425. https://gcc.gnu.org/PR12632 18828 426. https://gcc.gnu.org/PR12712 18829 427. https://gcc.gnu.org/PR12726 18830 428. https://gcc.gnu.org/PR12890 18831 429. https://gcc.gnu.org/PR12900 18832 430. https://gcc.gnu.org/PR13060 18833 431. https://gcc.gnu.org/PR13289 18834 432. https://gcc.gnu.org/PR13318 18835 433. https://gcc.gnu.org/PR13392 18836 434. https://gcc.gnu.org/PR13574 18837 435. https://gcc.gnu.org/PR13475 18838 436. https://gcc.gnu.org/PR13797 18839 437. https://gcc.gnu.org/PR13824 18840 438. https://gcc.gnu.org/PR8776 18841 439. https://gcc.gnu.org/PR10339 18842 440. https://gcc.gnu.org/PR11350 18843 441. https://gcc.gnu.org/PR12826 18844 442. https://gcc.gnu.org/PR12500 18845 443. https://gcc.gnu.org/PR12941 18846 444. https://gcc.gnu.org/PR12953 18847 445. https://gcc.gnu.org/PR13041 18848 446. https://gcc.gnu.org/PR13507 18849 447. https://gcc.gnu.org/PR13382 18850 448. https://gcc.gnu.org/PR13394 18851 449. https://gcc.gnu.org/PR13400 18852 450. https://gcc.gnu.org/PR13521 18853 451. https://gcc.gnu.org/PR2094 18854 452. https://gcc.gnu.org/PR2294 18855 453. https://gcc.gnu.org/PR5050 18856 454. https://gcc.gnu.org/PR9371 18857 455. https://gcc.gnu.org/PR9546 18858 456. https://gcc.gnu.org/PR10081 18859 457. https://gcc.gnu.org/PR10093 18860 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 18861 459. https://gcc.gnu.org/PR10095 18862 460. https://gcc.gnu.org/PR11554 18863 461. https://gcc.gnu.org/PR12297 18864 462. https://gcc.gnu.org/PR12352 18865 463. https://gcc.gnu.org/PR12438 18866 464. https://gcc.gnu.org/PR12540 18867 465. https://gcc.gnu.org/PR12594 18868 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 18869 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 18870 468. https://gcc.gnu.org/PR12657 18871 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 18872 470. https://gcc.gnu.org/PR12696 18873 471. https://gcc.gnu.org/PR12815 18874 472. https://gcc.gnu.org/PR12862 18875 473. https://gcc.gnu.org/PR12926 18876 474. https://gcc.gnu.org/PR12967 18877 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 18878 476. https://gcc.gnu.org/PR12971 18879 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 18880 478. https://gcc.gnu.org/PR13007 18881 479. https://gcc.gnu.org/PR13009 18882 480. https://gcc.gnu.org/PR13057 18883 481. https://gcc.gnu.org/PR13070 18884 482. https://gcc.gnu.org/PR13081 18885 483. https://gcc.gnu.org/PR13239 18886 484. https://gcc.gnu.org/PR13262 18887 485. https://gcc.gnu.org/PR13290 18888 486. https://gcc.gnu.org/PR13323 18889 487. https://gcc.gnu.org/PR13369 18890 488. https://gcc.gnu.org/PR13371 18891 489. https://gcc.gnu.org/PR13445 18892 490. https://gcc.gnu.org/PR13461 18893 491. https://gcc.gnu.org/PR13462 18894 492. https://gcc.gnu.org/PR13478 18895 493. https://gcc.gnu.org/PR13544 18896 494. https://gcc.gnu.org/PR13650 18897 495. https://gcc.gnu.org/PR13683 18898 496. https://gcc.gnu.org/PR13688 18899 497. https://gcc.gnu.org/PR13774 18900 498. https://gcc.gnu.org/PR13884 18901 499. https://gcc.gnu.org/PR10746 18902 500. https://gcc.gnu.org/PR11433 18903 501. https://gcc.gnu.org/PR12633 18904 502. https://gcc.gnu.org/PR13037 18905 503. https://gcc.gnu.org/PR13213 18906 504. https://gcc.gnu.org/PR4490 18907 505. https://gcc.gnu.org/PR12292 18908 506. https://gcc.gnu.org/PR12441 18909 507. https://gcc.gnu.org/PR12943 18910 508. https://gcc.gnu.org/PR13608 18911 509. https://gcc.gnu.org/PR11598 18912 510. https://gcc.gnu.org/PR11793 18913 511. https://gcc.gnu.org/PR12467 18914 512. https://gcc.gnu.org/PR12537 18915 513. https://gcc.gnu.org/PR12496 18916 514. https://gcc.gnu.org/PR12865 18917 515. https://gcc.gnu.org/PR13354 18918 516. https://gcc.gnu.org/PR10467 18919 517. https://gcc.gnu.org/PR11226 18920 518. https://gcc.gnu.org/PR11227 18921 519. https://gcc.gnu.org/PR12644 18922 520. https://gcc.gnu.org/PR13149 18923 521. https://gcc.gnu.org/PR12654 18924 522. https://gcc.gnu.org/PR12965 18925 523. https://gcc.gnu.org/PR13031 18926 524. https://gcc.gnu.org/PR11634 18927 525. https://gcc.gnu.org/PR12158 18928 526. https://gcc.gnu.org/PR11992 18929 527. https://gcc.gnu.org/PR9365 18930 528. https://gcc.gnu.org/PR10392 18931 529. https://gcc.gnu.org/PR11322 18932 530. https://gcc.gnu.org/PR13069 18933 531. https://gcc.gnu.org/PR13302 18934 532. https://gcc.gnu.org/PR13585 18935 533. https://gcc.gnu.org/PR8916 18936 534. https://gcc.gnu.org/PR11576 18937 535. https://gcc.gnu.org/PR13122 18938 536. https://gcc.gnu.org/PR13256 18939 537. https://gcc.gnu.org/PR13373 18940 538. https://gcc.gnu.org/PR12561 18941 539. https://gcc.gnu.org/PR6243 18942 540. https://gcc.gnu.org/PR11397 18943 541. https://gcc.gnu.org/PR12505 18944 542. https://gcc.gnu.org/PR13150 18945 543. https://gcc.gnu.org/PR12666 18946 544. https://gcc.gnu.org/PR12969 18947 545. https://gcc.gnu.org/PR10819 18948 546. https://gcc.gnu.org/PR11612 18949 547. https://gcc.gnu.org/PR13211 18950 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 18951 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 18952 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 18953 551. https://gcc.gnu.org/onlinedocs/ 18954 552. mailto:gcc-help@gcc.gnu.org 18955 553. mailto:gcc@gcc.gnu.org 18956 554. https://gcc.gnu.org/lists.html 18957 555. https://www.fsf.org/ 18958 556. https://gcc.gnu.org/about.html 18959 557. http://validator.w3.org/check/referer 18960====================================================================== 18961http://gcc.gnu.org/gcc-3.2/index.html 18962 18963 GCC 3.2 Release Series 18964 18965 (This release series is no longer supported.) 18966 18967 April 25, 2003 18968 18969 The [1]GNU project and the GCC developers are pleased to announce the 18970 release of GCC 3.2.3. 18971 18972 The purpose of the GCC 3.2 release series is to provide a stable 18973 platform for OS distributors to use building their next releases. A 18974 primary objective was to stabilize the C++ ABI; we believe that the 18975 interface to the compiler and the C++ standard library are now 18976 relatively stable. 18977 18978 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 18979 interoperate with code compiled by GCC 3.1.1 or earlier. 18980 18981 Please refer to our [2]detailed list of news, caveats, and bug-fixes 18982 for further information. 18983 18984Release History 18985 18986 GCC 3.2.3 18987 April 25, 2003 ([3]changes) 18988 18989 GCC 3.2.2 18990 February 5, 2003 ([4]changes) 18991 18992 GCC 3.2.1 18993 November 19, 2002 ([5]changes) 18994 18995 GCC 3.2 18996 August 14, 2002 ([6]changes) 18997 18998References and Acknowledgements 18999 19000 GCC used to stand for the GNU C Compiler, but since the compiler 19001 supports several other languages aside from C, it now stands for the 19002 GNU Compiler Collection. 19003 19004 A list of [7]successful builds is updated as new information becomes 19005 available. 19006 19007 The GCC developers would like to thank the numerous people that have 19008 contributed new features, improvements, bug fixes, and other changes as 19009 well as test results to GCC. This [8]amazing group of volunteers is 19010 what makes GCC successful. 19011 19012 For additional information about GCC please refer to the [9]GCC project 19013 web site or contact the [10]GCC development mailing list. 19014 19015 To obtain GCC please use [11]our mirror sites, or our CVS server. 19016 19017 19018 For questions related to the use of GCC, please consult these web 19019 pages and the [12]GCC manuals. If that fails, the 19020 [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these 19021 web pages and the development of GCC are welcome on our developer 19022 list at [14]gcc@gcc.gnu.org. All of [15]our lists have public 19023 archives. 19024 19025 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 19026 distribution of this entire article is permitted in any medium, 19027 provided this notice is preserved. 19028 19029 These pages are [17]maintained by the GCC team. Last modified 19030 2021-07-28[18]. 19031 19032References 19033 19034 1. http://www.gnu.org/ 19035 2. http://gcc.gnu.org/gcc-3.2/changes.html 19036 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 19037 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 19038 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 19039 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2 19040 7. http://gcc.gnu.org/gcc-3.2/buildstat.html 19041 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 19042 9. http://gcc.gnu.org/index.html 19043 10. mailto:gcc@gcc.gnu.org 19044 11. http://gcc.gnu.org/mirrors.html 19045 12. https://gcc.gnu.org/onlinedocs/ 19046 13. mailto:gcc-help@gcc.gnu.org 19047 14. mailto:gcc@gcc.gnu.org 19048 15. https://gcc.gnu.org/lists.html 19049 16. https://www.fsf.org/ 19050 17. https://gcc.gnu.org/about.html 19051 18. http://validator.w3.org/check/referer 19052====================================================================== 19053http://gcc.gnu.org/gcc-3.2/changes.html 19054 19055 GCC 3.2 Release Series 19056 Changes, New Features, and Fixes 19057 19058 The latest release in the 3.2 release series is [1]GCC 3.2.3. 19059 19060Caveats and New Features 19061 19062 Caveats 19063 19064 * The C++ compiler does not correctly zero-initialize 19065 pointers-to-data members. You must explicitly initialize them. For 19066 example: int S::*m(0); will work, but depending on 19067 default-initialization to zero will not work. This bug cannot be 19068 fixed in GCC 3.2 without inducing unacceptable risks. It will be 19069 fixed in GCC 3.3. 19070 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 19071 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 19072 a number of C++ ABI fixes which make its C++ compiler generate 19073 binary code which is incompatible with the C++ compilers found in 19074 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 19075 19076 Frontend Enhancements 19077 19078 C/C++/Objective-C 19079 19080 * The method of constructing the list of directories to be searched 19081 for header files has been revised. If a directory named by a -I 19082 option is a standard system include directory, the option is 19083 ignored to ensure that the default search order for system 19084 directories and the special treatment of system header files are 19085 not defeated. 19086 * The C and Objective-C compilers no longer accept the "Naming Types" 19087 extension (typedef foo = bar); it was already unavailable in C++. 19088 Code which uses it will need to be changed to use the "typeof" 19089 extension instead: typedef typeof(bar) foo. (We have removed this 19090 extension without a period of deprecation because it has caused the 19091 compiler to crash since version 3.0 and no one noticed until very 19092 recently. Thus we conclude it is not in widespread use.) 19093 19094 C++ 19095 19096 * GCC 3.2 fixed serveral differences between the C++ ABI implemented 19097 in GCC and the multi-vendor standard, but more have been found 19098 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 19099 code which is affected by these bugs. We will fix these bugs in 19100 some future release, once we are confident that all have been 19101 found; until then, it is our intention to make changes to the ABI 19102 only if they are necessary for correct compilation of C++, as 19103 opposed to conformance to the ABI documents. 19104 * For details on how to build an ABI compliant compiler for GNU/Linux 19105 systems, check the [3]common C++ ABI page. 19106 19107 New Targets and Target Specific Improvements 19108 19109 IA-32 19110 19111 * Fixed a number of bugs in SSE and MMX intrinsics. 19112 * Fixed common compiler crashes with SSE instruction set enabled 19113 (implied by -march=pentium3, pentium4, athlon-xp) 19114 * __m128 and __m128i is not 128bit aligned when used in structures. 19115 19116 x86-64 19117 19118 * A bug whereby the compiler could generate bad code for bzero has 19119 been fixed. 19120 * ABI fixes (implying ABI incompatibilities with previous version in 19121 some corner cases) 19122 * Fixed prefetch code generation 19123 __________________________________________________________________ 19124 19125GCC 3.2.3 19126 19127 3.2.3 is a bug fix release only; there are no new features that were 19128 not present in GCC 3.2.2. 19129 19130 Bug Fixes 19131 19132 This section lists the problem reports (PRs) from GCC's bug tracking 19133 system that are known to be fixed in the 3.2.3 release. This list might 19134 not be complete (that is, it is possible that some PRs that have been 19135 fixed are not listed here), and some of the titles have been changed to 19136 make them more clear. 19137 19138 Internal Compiler Errors (multi-platform) 19139 19140 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in 19141 cc1plus 19142 * [5]6440: (c++) template specializations cause ICE 19143 * [6]7050: (c++) ICE on: (i ? get_string() : throw) 19144 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 19145 * [8]7982: (c++) ICE due to infinite recursion (using STL set) 19146 * [9]8068: exceedingly high (infinite) memory usage 19147 * [10]8178: ICE with __builtin_ffs 19148 * [11]8396: ICE in copy_to_mode_reg, in explow.c 19149 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 19150 * [13]9768: ICE when optimizing inline code at -O2 19151 * [14]9798: (c++) Infinite recursion (segfault) in 19152 cp/decl.c:push_using_directive with recursive using directives 19153 * [15]9799: mismatching structure initializer with nested flexible 19154 array member: ICE 19155 * [16]9928: ICE on duplicate enum declaration 19156 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 19157 sparc, alpha) 19158 * [18]10352: ICE in find_reloads_toplev 19159 * [19]10336: ICE with -Wunreachable-code 19160 19161 C/optimizer bugs: 19162 19163 * [20]8224: Incorrect joining of signed and unsigned division 19164 * [21]8613: -O2 produces wrong code with builtin strlen and 19165 postincrements 19166 * [22]8828: gcc reports some code is unreachable when it is not 19167 * [23]9226: GCSE breaking argument passing 19168 * [24]9853: miscompilation of non-constant structure initializer 19169 * [25]9797: C99-style struct initializers are miscompiled 19170 * [26]9967: Some standard C function calls should not be replaced 19171 when optimizing for size 19172 * [27]10116: ce2: invalid merge of join_bb in the context of switch 19173 statements 19174 * [28]10171: wrong code for inlined function 19175 * [29]10175: -Wunreachable-code doesn't work for single lines 19176 19177 C++ compiler and library: 19178 19179 * [30]8316: Confusing diagnostic for code that misuses conversion 19180 operators 19181 * [31]9169: filebuf output fails if codecvt<>::out returns noconv 19182 * [32]9420: incomplete type incorrectly reported 19183 * [33]9459: typeof in return type specification of template not 19184 supported 19185 * [34]9507: filebuf::open handles ios_base::ate incorrectly 19186 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc 19187 * [36]9602: Total confusion about template/friend/virtual/abstract 19188 * [37]9993: destructor not called for local object created within and 19189 returned from infinite loop 19190 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 19191 system 19192 19193 Java compiler and library: 19194 19195 * [39]9652: libgcj build fails on irix6.5.1[78] 19196 * [40]10144: gas on solaris complains about bad .stabs lines for 19197 java, native as unaffected 19198 19199 x86-specific (Intel/AMD): 19200 19201 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86 19202 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions 19203 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 19204 failed 19205 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 19206 failed 19207 19208 SPARC-specific: 19209 19210 * [45]7784: [Sparc] ICE in extract_insn, in recog.c 19211 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in 19212 unroll.c 19213 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 19214 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 19215 execute/loop-2d.c 19216 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 19217 * [50]9414: Scheduling bug on Ultrasparc 19218 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64 19219 19220 m68k-specific: 19221 19222 * [52]7248: broken "inclusive or" code 19223 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 19224 19225 PowerPC-specific: 19226 19227 * [54]9732: Wrong code with -O2 -fPIC 19228 * [55]10073: ICE: powerpc cannot split insn 19229 19230 Alpha-specific: 19231 19232 * [56]7702: optimization problem on a DEC alpha under OSF1 19233 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 19234 19235 HP-specific: 19236 19237 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 19238 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 19239 (missing symbol) 19240 * [60]10271: Floating point args don't get reloaded across function 19241 calls with -O2 19242 19243 MIPS specific: 19244 19245 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 19246 compile/920501-4.c 19247 19248 CRIS specific: 19249 19250 * [62]10377: gcc-3.2.2 creates bad assembler code for cris 19251 19252 Miscellaneous and minor bugs: 19253 19254 * [63]6955: collect2 says "core dumped" when there is no core 19255 __________________________________________________________________ 19256 19257GCC 3.2.2 19258 19259 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 19260 install by means of the DESTDIR variable. Parts of the GCC tree have 19261 featured that support long before, but now it is available even from 19262 the top level. 19263 19264 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 19265 features that were not present in GCC 3.2.1. 19266 19267 Bug Fixes 19268 19269 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 19270 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 19271 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 19272 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 19273 change, and thus restores ABI-compatibility with previous releases 19274 (except GCC 3.2.1) on these platforms. 19275 19276 This section lists the problem reports (PRs) from GCC's bug tracking 19277 system that are known to be fixed in the 3.2.2 release. This list might 19278 not be complete (that is, it is possible that some PRs that have been 19279 fixed are not listed here) and some of the titles have been changed to 19280 make them more clear. 19281 19282 Internal Compiler Errors (multi-platform) 19283 19284 * [64]5919: (c++) ICE when passing variable array to template 19285 function 19286 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 19287 * [66]7507: ICE with -O2 when address of called function is a 19288 complicated expression 19289 * [67]7622: ICE with nested inline functions if function's address is 19290 taken 19291 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 19292 [69]9258) 19293 * [70]8031: (c++) ICE in code comparing typeids and casting from 19294 virtual base 19295 * [71]8275: ICE in simplify_subreg 19296 * [72]8332: (c++) builtin strlen/template interaction causes ICE 19297 * [73]8372: (c++) ICE on explicit call of destructor 19298 * [74]8439: (c, not c++) empty struct causes ICE 19299 * [75]8442: (c++) ICE with nested template classes 19300 * [76]8518: ICE when compiling mplayer ("extern inline" issue) 19301 * [77]8615: (c++) ICE with out-of-range character constant template 19302 argument 19303 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 19304 * [79]8799: (c++) ICE: error reporting routines re-entered 19305 * [80]9328: (c++) ICE with typeof(X) for overloaded X 19306 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes 19307 19308 C++ (compiler and library) bugs 19309 19310 * [82]47: scoping in nested classes is broken 19311 * [83]6745: problems with iostream rdbuf() member function 19312 * [84]8214: conversion from const char* const to char* sometimes 19313 accepted illegally 19314 * [85]8493: builtin strlen and overload resolution (same bug as 19315 [86]8332) 19316 * [87]8503: strange behaviour of function types 19317 * [88]8727: compiler confused by inheritance from an anonymous struct 19318 * [89]7445: poor performance of std::locale::classic() in 19319 multi-threaded applications 19320 * [90]8230: mishandling of overflow in vector<T>::resize 19321 * [91]8399: sync_with_stdio(false) breaks unformatted input 19322 * [92]8662: illegal access of private member of unnamed class is 19323 accepted 19324 * [93]8707: "make distclean" fails in libstdc++-v3 directory 19325 * [94]8708: __USE_MALLOC doesn't work 19326 * [95]8790: Use of non-thread-safe strtok in src/localename.cc 19327 * [96]8887: Bug in date formats with --enable-clocale=generic 19328 * [97]9076: Call Frame Instructions are not handled correctly during 19329 unwind operation 19330 * [98]9151: std::setprecision limited to 16 digits when outputting a 19331 double to a stream 19332 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers 19333 * [100]9269: libstdc++ headers: explicit specialization of function 19334 must precede its first use 19335 * [101]9322: return value of basic_streambuf<>::getloc affected by 19336 locale::global 19337 * [102]9433: segfault in runtime support for dynamic_cast 19338 19339 C and optimizer bugs 19340 19341 * [103]8032: GCC incorrectly initializes static structs that have 19342 flexible arrays 19343 * [104]8639: simple arithmetic expression broken 19344 * [105]8794: optimization improperly eliminates certain expressions 19345 * [106]8832: traditional "asm volatile" code is illegally optimized 19346 * [107]8988: loop optimizer bug: with -O2, code is generated that 19347 segfaults (found on i386, bug present for all platforms) 19348 * [108]9492: structure copy clobbers subsequent stores to structure 19349 19350 Objective-C bugs 19351 19352 * [109]9267: Objective-C parser won't build with newer bison versions 19353 (e.g. 1.875) 19354 19355 Ada bugs 19356 19357 * [110]8344: Ada build problem due to conflict between gcc/final.o, 19358 gcc/ada/final.o 19359 19360 Preprocessor bugs 19361 19362 * [111]8524: _Pragma within macros is improperly expanded 19363 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 19364 -fshort-wchar 19365 19366 ARM-specific 19367 19368 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95 19369 19370 x86-specific (Intel/AMD) 19371 19372 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 19373 * [115]8599: loop unroll bug with -march=k6-3 19374 * [116]9506: ABI breakage in structure return (affects BSD and 19375 Cygwin, but not GNU/Linux) 19376 19377 FreeBSD 5.0 specific 19378 19379 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 19380 19381 RTEMS-specific 19382 19383 * [118]9292: hppa1.1-rtems configurery problems 19384 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 19385 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue 19386 * [121]9296: gthr-rtems regression 19387 * [122]9316: powerpc-rtems: extending multilibs 19388 19389 HP-PA specific 19390 19391 * [123]9493: ICE with -O2 when building a simple function 19392 19393 Documentation 19394 19395 * [124]7341: hyperlink to gcov in GCC documentation doesn't work 19396 * [125]8947: Please add a warning about "-malign-double" in docs 19397 * [126]7448, [127]8882: typo cleanups 19398 __________________________________________________________________ 19399 19400GCC 3.2.1 19401 19402 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 19403 generates code that is known not to be binary-compatible with the 19404 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 19405 in the distribution, for details. 19406 19407 This release also removes an old GCC extension, "naming types", and the 19408 documentation now directs users to use a different GCC extension, 19409 __typeof__, instead. The feature had evidently been broken for a while. 19410 19411 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 19412 the new warning there are no new features that were not present in GCC 19413 3.2. 19414 19415 In addition, the previous fix for [128]PR 7445 (poor performance of 19416 std::locale::classic() in multi-threaded applications) was reverted 19417 ("unfixed"), because the "fix" was not thread-safe. 19418 19419 Bug Fixes 19420 19421 This section lists the problem reports (PRs) from GCC's bug tracking 19422 system that are known to be fixed in the 3.2.1 release. This list might 19423 not be complete (that is, it is possible that some PRs that have been 19424 fixed are not listed here). As you can see, the number of bug fixes is 19425 quite large, so it is strongly recommended that users of earlier GCC 19426 3.x releases upgrade to GCC 3.2.1. 19427 19428 Internal Compiler Errors (multi-platform) 19429 19430 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 19431 * [130]5661: (c++) ICE instantiating template on array of unknown 19432 size (bad code) 19433 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 19434 64-bit platforms 19435 * [132]6994: (c++) ICE in find_function_data 19436 * [133]7150: preprocessor: GCC -dM -E gives an ICE 19437 * [134]7160: ICE when optimizing branches without a return value 19438 * [135]7228: (c++) ICE when using member template and template 19439 function 19440 * [136]7266: (c++) ICE with -pedantic on missing typename 19441 * [137]7353: ICE from use of "Naming Types" extension, see above 19442 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c 19443 * [139]7478: (c++) ICE on static_cast inside template 19444 * [140]7526: preprocessor core dump when _Pragma implies #pragma 19445 dependency 19446 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803 19447 is a duplicate) 19448 * [143]7754: (c++) ICE on union with template parameter 19449 * [144]7788: (c++) redeclaring a definition as an incomplete class 19450 causes ICE 19451 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c 19452 * [146]8055: preprocessor dies with SIG11 when building FreeBSD 19453 kernel 19454 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 19455 variables 19456 * [148]8134: (c++) ICE in force_store_init_value on legal code 19457 * [149]8149: (c++) ICE on incomplete type 19458 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 19459 initialization 19460 19461 C++ (compiler and library) bugs 19462 19463 * [151]5607: No pointer adjustment in covariant return types 19464 * [152]6579: Infinite loop with statement expressions in member 19465 initialization 19466 * [153]6803: Default copy constructor bug in GCC 3.1 19467 * [154]7176: g++ confused by friend and static member with same name 19468 * [155]7188: Segfault with template class and recursive (incorrect) 19469 initializer list 19470 * [156]7306: Regression: GCC 3.x fails to compile code with virtual 19471 inheritance if a method has a variable number of arguments 19472 * [157]7461: ctype<char>::classic_table() returns offset array on 19473 Cygwin 19474 * [158]7524: f(const float arg[3]) fails 19475 * [159]7584: Erroneous ambiguous base error on using declaration 19476 * [160]7676: Member template overloading problem 19477 * [161]7679: infinite loop when a right parenthesis is missing 19478 * [162]7811: default locale not taken from environment 19479 * [163]7961: compare( char *) implemented incorrectly in 19480 basic_string<> 19481 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if 19482 streambuf::underflow() leaves gptr() NULL (dups: [165]8127, 19483 [166]6745) 19484 * [167]8096: deque::at() throws std::range_error instead of 19485 std::out_of_range 19486 * [168]8127: cout << cin.rdbuf() infinite loop 19487 * [169]8218: Excessively large memory consumed for classes with large 19488 array members 19489 * [170]8287: GCC 3.2: Destructor called for non-constructed local 19490 object 19491 * [171]8347: empty vector range used in string construction causes 19492 core dump 19493 * [172]8348: fail() flag is set in istringstream when eof() flag is 19494 set 19495 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file) 19496 19497 C and optimizer bugs 19498 19499 * [174]6627: -fno-align-functions doesn't seem to disable function 19500 alignment 19501 * [175]6631: life_analysis misoptimizes code to initialize fields of 19502 a structure 19503 * [176]7102: unsigned char division results in floating exception 19504 * [177]7120: Run once loop should *always* be unrolled 19505 (pessimization) 19506 * [178]7209: Bug involving array referencing and ?: operator 19507 * [179]7515: invalid inlining of global function with -O3 19508 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test 19509 * [181]8467: bug in sibling call optimization 19510 19511 Preprocessor bugs 19512 19513 * [182]4890: incorrect line markers from the traditional preprocessor 19514 * [183]7357: -M option omits system headers files (making it the same 19515 as -MM) 19516 * [184]7358: Changes to Sun's make Dependencies 19517 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 19518 C headers 19519 * [186]7862: Interrupting GCC -MD removes .d file but not .o 19520 * [187]8190: Failed compilation deletes -MD dependency file 19521 * [188]8524: _Pragma within macro is improperly expanded 19522 19523 x86 specific (Intel/AMD) 19524 19525 * [189]5351: (i686-only) function pass-by-value structure copy 19526 corrupts stack ([190]7591 is a duplicate) 19527 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with 19528 -march=pentium3/pentium2/athlon (these are all the same underlying 19529 bug, in MMX register use) 19530 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same 19531 as above?) 19532 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 19533 * [199]6981: wrong code in 64-bit manipulation on x86 19534 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 19535 macro 19536 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 19537 intrinsics are broken 19538 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 19539 -march=pentium4 19540 * [203]7693: Typo in i386 mmintrin.h header 19541 * [204]7723: ICE - Pentium3 sse - GCC 3.2 19542 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 19543 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 19544 19545 PowerPC specific 19546 19547 * [207]5967: GCC bug when profiling nested functions on powerpc 19548 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while 19549 loop on PowerPC 19550 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 19551 * [210]7130: miscompiled code for GCC-3.1 on 19552 powerpc-unknown-linux-gnu with -funroll-all-loops 19553 * [211]7133: PowerPC ICE: unrecognizable insn 19554 * [212]7380: ICE in extract_insn, at recog.c:2148 19555 * [213]8252: ICE on Altivec code with optimization turned on 19556 * [214]8451: Altivec ICE in GCC 3.2 19557 19558 HP/PA specific 19559 19560 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa 19561 19562 SPARC specific 19563 19564 * [216]6668: when using --disable-multilib, libgcc_s.so is installed 19565 in the wrong place on sparc-solaris 19566 * [217]7151: ICE when compiling for UltraSPARC 19567 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 19568 double and -O1 19569 * [219]7842: [REGRESSION] SPARC code gen bug 19570 19571 ARM specific 19572 19573 * [220]7856: [arm] invalid offset in constant pool reference 19574 * [221]7967: optimization produces wrong code (ARM) 19575 19576 Alpha specific 19577 19578 * [222]7374: __builtin_fabsl broken on alpha 19579 19580 IBM s390 specific 19581 19582 * [223]7370: ICE in fixup_var_refs_1 on s390x 19583 * [224]7409: loop optimization bug on s390x-linux-gnu 19584 * [225]8232: s390x: ICE when using bcmp with int length argument 19585 19586 SCO specific 19587 19588 * [226]7623: SCO OpenServer build fails with machmode.def: undefined 19589 symbol: BITS_PER_UNIT 19590 19591 m68k/Coldfire specific 19592 19593 * [227]8314: crtbegin, crtend need to be multilib'ed for this 19594 platform 19595 19596 Documentation 19597 19598 * [228]761: Document some undocumented options 19599 * [229]5610: Fix documentation about invoking SSE instructions 19600 (-mfpmath=sse) 19601 * [230]7484: List -Wmissing-declarations as C-only option 19602 * [231]7531: -mcmodel not documented for x86-64 19603 * [232]8120: Update documentation of bad use of ## 19604 __________________________________________________________________ 19605 19606GCC 3.2 19607 19608 3.2 is a small bug fix release, but there is a change to the 19609 application binary interface (ABI), hence the change to the second part 19610 of the version number. 19611 19612 The main purpose of the 3.2 release is to correct a couple of problems 19613 in the C++ ABI, with the intention of providing a stable interface 19614 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 19615 19616 Bug Fixes 19617 19618 C++ 19619 19620 * [233]7320: g++ 3.2 relocation problem 19621 * [234]7470: vtable: virtual function pointers not in declaration 19622 order 19623 19624 libstdc++ 19625 19626 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t 19627 * [236]6503, [237]6642, [238]7186: Problems with comparing or 19628 subtracting various types of const and non-const iterators 19629 * [239]7216: ambiguity with basic_iostream::traits_type 19630 * [240]7220: problem with basic_istream::ignore(0,delimiter) 19631 * [241]7222: locale::operator==() doesn't work on std::locale("") 19632 * [242]7286: placement operator delete issue 19633 * [243]7442: cxxabi.h does not match the C++ ABI 19634 * [244]7445: poor performance of std::locale::classic() in 19635 multi-threaded applications 19636 19637 x86-64 specific 19638 19639 * [245]7291: off-by-one in generated inline bzero code for x86-64 19640 19641 19642 For questions related to the use of GCC, please consult these web 19643 pages and the [246]GCC manuals. If that fails, the 19644 [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these 19645 web pages and the development of GCC are welcome on our developer 19646 list at [248]gcc@gcc.gnu.org. All of [249]our lists have public 19647 archives. 19648 19649 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and 19650 distribution of this entire article is permitted in any medium, 19651 provided this notice is preserved. 19652 19653 These pages are [251]maintained by the GCC team. Last modified 19654 2021-07-28[252]. 19655 19656References 19657 19658 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 19659 2. http://gcc.gnu.org/gcc-3.1/changes.html 19660 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html 19661 4. https://gcc.gnu.org/PR3782 19662 5. https://gcc.gnu.org/PR6440 19663 6. https://gcc.gnu.org/PR7050 19664 7. https://gcc.gnu.org/PR7741 19665 8. https://gcc.gnu.org/PR7982 19666 9. https://gcc.gnu.org/PR8068 19667 10. https://gcc.gnu.org/PR8178 19668 11. https://gcc.gnu.org/PR8396 19669 12. https://gcc.gnu.org/PR8674 19670 13. https://gcc.gnu.org/PR9768 19671 14. https://gcc.gnu.org/PR9798 19672 15. https://gcc.gnu.org/PR9799 19673 16. https://gcc.gnu.org/PR9928 19674 17. https://gcc.gnu.org/PR10114 19675 18. https://gcc.gnu.org/PR10352 19676 19. https://gcc.gnu.org/PR10336 19677 20. https://gcc.gnu.org/PR8224 19678 21. https://gcc.gnu.org/PR8613 19679 22. https://gcc.gnu.org/PR8828 19680 23. https://gcc.gnu.org/PR9226 19681 24. https://gcc.gnu.org/PR9853 19682 25. https://gcc.gnu.org/PR9797 19683 26. https://gcc.gnu.org/PR9967 19684 27. https://gcc.gnu.org/PR10116 19685 28. https://gcc.gnu.org/PR10171 19686 29. https://gcc.gnu.org/PR10175 19687 30. https://gcc.gnu.org/PR8316 19688 31. https://gcc.gnu.org/PR9169 19689 32. https://gcc.gnu.org/PR9420 19690 33. https://gcc.gnu.org/PR9459 19691 34. https://gcc.gnu.org/PR9507 19692 35. https://gcc.gnu.org/PR9538 19693 36. https://gcc.gnu.org/PR9602 19694 37. https://gcc.gnu.org/PR9993 19695 38. https://gcc.gnu.org/PR10167 19696 39. https://gcc.gnu.org/PR9652 19697 40. https://gcc.gnu.org/PR10144 19698 41. https://gcc.gnu.org/PR8746 19699 42. https://gcc.gnu.org/PR9888 19700 43. https://gcc.gnu.org/PR9638 19701 44. https://gcc.gnu.org/PR9954 19702 45. https://gcc.gnu.org/PR7784 19703 46. https://gcc.gnu.org/PR7796 19704 47. https://gcc.gnu.org/PR8281 19705 48. https://gcc.gnu.org/PR8366 19706 49. https://gcc.gnu.org/PR8726 19707 50. https://gcc.gnu.org/PR9414 19708 51. https://gcc.gnu.org/PR10067 19709 52. https://gcc.gnu.org/PR7248 19710 53. https://gcc.gnu.org/PR8343 19711 54. https://gcc.gnu.org/PR9732 19712 55. https://gcc.gnu.org/PR10073 19713 56. https://gcc.gnu.org/PR7702 19714 57. https://gcc.gnu.org/PR9671 19715 58. https://gcc.gnu.org/PR8694 19716 59. https://gcc.gnu.org/PR9953 19717 60. https://gcc.gnu.org/PR10271 19718 61. https://gcc.gnu.org/PR6362 19719 62. https://gcc.gnu.org/PR10377 19720 63. https://gcc.gnu.org/PR6955 19721 64. https://gcc.gnu.org/PR5919 19722 65. https://gcc.gnu.org/PR7129 19723 66. https://gcc.gnu.org/PR7507 19724 67. https://gcc.gnu.org/PR7622 19725 68. https://gcc.gnu.org/PR7681 19726 69. https://gcc.gnu.org/PR9528 19727 70. https://gcc.gnu.org/PR8031 19728 71. https://gcc.gnu.org/PR8275 19729 72. https://gcc.gnu.org/PR8332 19730 73. https://gcc.gnu.org/PR8372 19731 74. https://gcc.gnu.org/PR8439 19732 75. https://gcc.gnu.org/PR8442 19733 76. https://gcc.gnu.org/PR8518 19734 77. https://gcc.gnu.org/PR8615 19735 78. https://gcc.gnu.org/PR8663 19736 79. https://gcc.gnu.org/PR8799 19737 80. https://gcc.gnu.org/PR9328 19738 81. https://gcc.gnu.org/PR9465 19739 82. https://gcc.gnu.org/PR47 19740 83. https://gcc.gnu.org/PR6745 19741 84. https://gcc.gnu.org/PR8214 19742 85. https://gcc.gnu.org/PR8493 19743 86. https://gcc.gnu.org/PR8332 19744 87. https://gcc.gnu.org/PR8503 19745 88. https://gcc.gnu.org/PR8727 19746 89. https://gcc.gnu.org/PR7445 19747 90. https://gcc.gnu.org/PR8230 19748 91. https://gcc.gnu.org/PR8399 19749 92. https://gcc.gnu.org/PR8662 19750 93. https://gcc.gnu.org/PR8707 19751 94. https://gcc.gnu.org/PR8708 19752 95. https://gcc.gnu.org/PR8790 19753 96. https://gcc.gnu.org/PR8887 19754 97. https://gcc.gnu.org/PR9076 19755 98. https://gcc.gnu.org/PR9151 19756 99. https://gcc.gnu.org/PR9168 19757 100. https://gcc.gnu.org/PR9269 19758 101. https://gcc.gnu.org/PR9322 19759 102. https://gcc.gnu.org/PR9433 19760 103. https://gcc.gnu.org/PR8032 19761 104. https://gcc.gnu.org/PR8639 19762 105. https://gcc.gnu.org/PR8794 19763 106. https://gcc.gnu.org/PR8832 19764 107. https://gcc.gnu.org/PR8988 19765 108. https://gcc.gnu.org/PR9492 19766 109. https://gcc.gnu.org/PR9267 19767 110. https://gcc.gnu.org/PR8344 19768 111. https://gcc.gnu.org/PR8524 19769 112. https://gcc.gnu.org/PR8880 19770 113. https://gcc.gnu.org/PR9090 19771 114. https://gcc.gnu.org/PR8588 19772 115. https://gcc.gnu.org/PR8599 19773 116. https://gcc.gnu.org/PR9506 19774 117. https://gcc.gnu.org/PR9484 19775 118. https://gcc.gnu.org/PR9292 19776 119. https://gcc.gnu.org/PR9293 19777 120. https://gcc.gnu.org/PR9295 19778 121. https://gcc.gnu.org/PR9296 19779 122. https://gcc.gnu.org/PR9316 19780 123. https://gcc.gnu.org/PR9493 19781 124. https://gcc.gnu.org/PR7341 19782 125. https://gcc.gnu.org/PR8947 19783 126. https://gcc.gnu.org/PR7448 19784 127. https://gcc.gnu.org/PR8882 19785 128. https://gcc.gnu.org/PR7445 19786 129. https://gcc.gnu.org/PR2521 19787 130. https://gcc.gnu.org/PR5661 19788 131. https://gcc.gnu.org/PR6419 19789 132. https://gcc.gnu.org/PR6994 19790 133. https://gcc.gnu.org/PR7150 19791 134. https://gcc.gnu.org/PR7160 19792 135. https://gcc.gnu.org/PR7228 19793 136. https://gcc.gnu.org/PR7266 19794 137. https://gcc.gnu.org/PR7353 19795 138. https://gcc.gnu.org/PR7411 19796 139. https://gcc.gnu.org/PR7478 19797 140. https://gcc.gnu.org/PR7526 19798 141. https://gcc.gnu.org/PR7721 19799 142. https://gcc.gnu.org/PR7803 19800 143. https://gcc.gnu.org/PR7754 19801 144. https://gcc.gnu.org/PR7788 19802 145. https://gcc.gnu.org/PR8031 19803 146. https://gcc.gnu.org/PR8055 19804 147. https://gcc.gnu.org/PR8067 19805 148. https://gcc.gnu.org/PR8134 19806 149. https://gcc.gnu.org/PR8149 19807 150. https://gcc.gnu.org/PR8160 19808 151. https://gcc.gnu.org/PR5607 19809 152. https://gcc.gnu.org/PR6579 19810 153. https://gcc.gnu.org/PR6803 19811 154. https://gcc.gnu.org/PR7176 19812 155. https://gcc.gnu.org/PR7188 19813 156. https://gcc.gnu.org/PR7306 19814 157. https://gcc.gnu.org/PR7461 19815 158. https://gcc.gnu.org/PR7524 19816 159. https://gcc.gnu.org/PR7584 19817 160. https://gcc.gnu.org/PR7676 19818 161. https://gcc.gnu.org/PR7679 19819 162. https://gcc.gnu.org/PR7811 19820 163. https://gcc.gnu.org/PR7961 19821 164. https://gcc.gnu.org/PR8071 19822 165. https://gcc.gnu.org/PR8127 19823 166. https://gcc.gnu.org/PR6745 19824 167. https://gcc.gnu.org/PR8096 19825 168. https://gcc.gnu.org/PR8127 19826 169. https://gcc.gnu.org/PR8218 19827 170. https://gcc.gnu.org/PR8287 19828 171. https://gcc.gnu.org/PR8347 19829 172. https://gcc.gnu.org/PR8348 19830 173. https://gcc.gnu.org/PR8391 19831 174. https://gcc.gnu.org/PR6627 19832 175. https://gcc.gnu.org/PR6631 19833 176. https://gcc.gnu.org/PR7102 19834 177. https://gcc.gnu.org/PR7120 19835 178. https://gcc.gnu.org/PR7209 19836 179. https://gcc.gnu.org/PR7515 19837 180. https://gcc.gnu.org/PR7814 19838 181. https://gcc.gnu.org/PR8467 19839 182. https://gcc.gnu.org/PR4890 19840 183. https://gcc.gnu.org/PR7357 19841 184. https://gcc.gnu.org/PR7358 19842 185. https://gcc.gnu.org/PR7602 19843 186. https://gcc.gnu.org/PR7862 19844 187. https://gcc.gnu.org/PR8190 19845 188. https://gcc.gnu.org/PR8524 19846 189. https://gcc.gnu.org/PR5351 19847 190. https://gcc.gnu.org/PR7591 19848 191. https://gcc.gnu.org/PR6845 19849 192. https://gcc.gnu.org/PR7034 19850 193. https://gcc.gnu.org/PR7124 19851 194. https://gcc.gnu.org/PR7174 19852 195. https://gcc.gnu.org/PR7134 19853 196. https://gcc.gnu.org/PR7375 19854 197. https://gcc.gnu.org/PR7390 19855 198. https://gcc.gnu.org/PR6890 19856 199. https://gcc.gnu.org/PR6981 19857 200. https://gcc.gnu.org/PR7242 19858 201. https://gcc.gnu.org/PR7396 19859 202. https://gcc.gnu.org/PR7630 19860 203. https://gcc.gnu.org/PR7693 19861 204. https://gcc.gnu.org/PR7723 19862 205. https://gcc.gnu.org/PR7951 19863 206. https://gcc.gnu.org/PR8146 19864 207. https://gcc.gnu.org/PR5967 19865 208. https://gcc.gnu.org/PR6984 19866 209. https://gcc.gnu.org/PR7114 19867 210. https://gcc.gnu.org/PR7130 19868 211. https://gcc.gnu.org/PR7133 19869 212. https://gcc.gnu.org/PR7380 19870 213. https://gcc.gnu.org/PR8252 19871 214. https://gcc.gnu.org/PR8451 19872 215. https://gcc.gnu.org/PR7250 19873 216. https://gcc.gnu.org/PR6668 19874 217. https://gcc.gnu.org/PR7151 19875 218. https://gcc.gnu.org/PR7335 19876 219. https://gcc.gnu.org/PR7842 19877 220. https://gcc.gnu.org/PR7856 19878 221. https://gcc.gnu.org/PR7967 19879 222. https://gcc.gnu.org/PR7374 19880 223. https://gcc.gnu.org/PR7370 19881 224. https://gcc.gnu.org/PR7409 19882 225. https://gcc.gnu.org/PR8232 19883 226. https://gcc.gnu.org/PR7623 19884 227. https://gcc.gnu.org/PR8314 19885 228. https://gcc.gnu.org/PR761 19886 229. https://gcc.gnu.org/PR5610 19887 230. https://gcc.gnu.org/PR7484 19888 231. https://gcc.gnu.org/PR7531 19889 232. https://gcc.gnu.org/PR8120 19890 233. https://gcc.gnu.org/PR7320 19891 234. https://gcc.gnu.org/PR7470 19892 235. https://gcc.gnu.org/PR6410 19893 236. https://gcc.gnu.org/PR6503 19894 237. https://gcc.gnu.org/PR6642 19895 238. https://gcc.gnu.org/PR7186 19896 239. https://gcc.gnu.org/PR7216 19897 240. https://gcc.gnu.org/PR7220 19898 241. https://gcc.gnu.org/PR7222 19899 242. https://gcc.gnu.org/PR7286 19900 243. https://gcc.gnu.org/PR7442 19901 244. https://gcc.gnu.org/PR7445 19902 245. https://gcc.gnu.org/PR7291 19903 246. https://gcc.gnu.org/onlinedocs/ 19904 247. mailto:gcc-help@gcc.gnu.org 19905 248. mailto:gcc@gcc.gnu.org 19906 249. https://gcc.gnu.org/lists.html 19907 250. https://www.fsf.org/ 19908 251. https://gcc.gnu.org/about.html 19909 252. http://validator.w3.org/check/referer 19910====================================================================== 19911http://gcc.gnu.org/gcc-3.1/index.html 19912 19913 GCC 3.1 19914 19915 (This release series is no longer supported.) 19916 19917 July 27, 2002 19918 19919 The [1]GNU project and the GCC developers are pleased to announce the 19920 release of GCC 3.1.1. 19921 19922 The links below still apply to GCC 3.1.1. 19923 19924 May 15, 2002 19925 19926 The [2]GNU project and the GCC developers are pleased to announce the 19927 release of GCC 3.1. 19928 19929 GCC used to stand for the GNU C Compiler, but since the compiler 19930 supports several other languages aside from C, it now stands for the 19931 GNU Compiler Collection. 19932 19933 A list of [3]successful builds is updated as new information becomes 19934 available. 19935 19936 The GCC developers would like to thank the numerous people that have 19937 contributed [4]new features, improvements, bug fixes, and other changes 19938 as well as test results to GCC. This [5]amazing group of volunteers is 19939 what makes GCC successful. 19940 19941 For additional information about GCC please refer to the [6]GCC project 19942 web site or contact the [7]GCC development mailing list. 19943 19944 To obtain GCC please use [8]our mirror sites, or our CVS server. 19945 __________________________________________________________________ 19946 19947 19948 For questions related to the use of GCC, please consult these web 19949 pages and the [9]GCC manuals. If that fails, the 19950 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these 19951 web pages and the development of GCC are welcome on our developer 19952 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public 19953 archives. 19954 19955 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 19956 distribution of this entire article is permitted in any medium, 19957 provided this notice is preserved. 19958 19959 These pages are [14]maintained by the GCC team. Last modified 19960 2021-07-28[15]. 19961 19962References 19963 19964 1. http://www.gnu.org/ 19965 2. http://www.gnu.org/ 19966 3. http://gcc.gnu.org/gcc-3.1/buildstat.html 19967 4. http://gcc.gnu.org/gcc-3.1/changes.html 19968 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 19969 6. http://gcc.gnu.org/index.html 19970 7. mailto:gcc@gcc.gnu.org 19971 8. http://gcc.gnu.org/mirrors.html 19972 9. https://gcc.gnu.org/onlinedocs/ 19973 10. mailto:gcc-help@gcc.gnu.org 19974 11. mailto:gcc@gcc.gnu.org 19975 12. https://gcc.gnu.org/lists.html 19976 13. https://www.fsf.org/ 19977 14. https://gcc.gnu.org/about.html 19978 15. http://validator.w3.org/check/referer 19979====================================================================== 19980http://gcc.gnu.org/gcc-3.1/changes.html 19981 19982 GCC 3.1 Release Series 19983 Changes, New Features, and Fixes 19984 19985Additional changes in GCC 3.1.1 19986 19987 * A bug related to how structures and unions are returned has been 19988 fixed for powerpc-*-netbsd*. 19989 * An important bug in the implementation of -fprefetch-loop-arrays 19990 has been fixed. Previously the optimization prefetched random 19991 blocks of memory for most targets except for i386. 19992 * The Java compiler now compiles Java programs much faster and also 19993 works with parallel make. 19994 * Nested functions have been fixed for mips*-*-netbsd*. 19995 * Some missing floating point support routines have beed added for 19996 mips*-*-netbsd*. 19997 * This [1]message gives additional information about the bugs fixed 19998 in this release. 19999 20000Caveats 20001 20002 * The -traditional C compiler option has been deprecated and will be 20003 removed in GCC 3.3. (It remains possible to preprocess non-C code 20004 with the traditional preprocessor.) 20005 * The default debugging format for most ELF platforms (including 20006 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 20007 from stabs to DWARF2. This requires GDB 5.1.1 or later. 20008 20009General Optimizer Improvements 20010 20011 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 20012 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 20013 for profile driven optimizations. 20014 Options -fprofile-arcs and -fbranch-probabilities can now be used 20015 to improve speed of the generated code by profiling the actual 20016 program behaviour on typical runs. In the absence of profile info 20017 the compiler attempts to guess the profile statically. 20018 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 20019 monitor performance of the generated code. 20020 According to the SPECInt2000 results on an AMD Athlon CPU, the code 20021 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 20022 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 20023 is about 2.1% faster compared to 2.95.3. Tests were done using the 20024 -O2 -march=athlon command-line options. 20025 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 20026 infrastructure developed by CodeSourcery, LLC for the C++ front 20027 end, so that it is now used in the C front end too. Inlining 20028 functions as trees exposes them earlier to the compiler, giving it 20029 more opportunities for optimization. 20030 * Support for data prefetching instructions has been added to the GCC 20031 back end and several targets. A new __builtin_prefetch intrinsic is 20032 available to explicitly insert prefetch instructions and 20033 experimental support for loop array prefetching has been added (see 20034 -fprefetch-loop-array documentation). 20035 * Support for emitting debugging information for macros has been 20036 added for DWARF2. It is activated using -g3. 20037 20038New Languages and Language specific improvements 20039 20040 C/C++ 20041 20042 * A few more [4]ISO C99 features. 20043 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 20044 * The preprocessor's symbol table has been merged with the symbol 20045 table of the C, C++ and Objective-C front ends. 20046 * The preprocessor consumes less memory than the preprocessor in GCC 20047 3.0, often significantly so. On normal input files, it typically 20048 consumes less memory than pre-3.0 cccp-based GCC, too. 20049 20050 C++ 20051 20052 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 20053 was a workaround to allow std compliant code to work with the 20054 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 20055 * The C++ ABI has been fixed so that void (A::*)() const is mangled 20056 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 20057 affects pointer to cv-qualified member function types. 20058 * The C++ ABI has been changed to correctly handle this code: 20059 struct A { 20060 void operator delete[] (void *, size_t); 20061 }; 20062 20063 struct B : public A { 20064 }; 20065 20066 new B[10]; 20067 20068 The amount of storage allocated for the array will be greater than 20069 it was in 3.0, in order to store the number of elements in the 20070 array, so that the correct size can be passed to operator delete[] 20071 when the array is deleted. Previously, the value passed to operator 20072 delete[] was unpredictable. 20073 This change will only affect code that declares a two-argument 20074 operator delete[] with a second parameter of type size_t in a base 20075 class, and does not override that definition in a derived class. 20076 * The C++ ABI has been changed so that: 20077 struct A { 20078 void operator delete[] (void *, size_t); 20079 void operator delete[] (void *); 20080 }; 20081 20082 does not cause unnecessary storage to be allocated when an array of 20083 A objects is allocated. 20084 This change will only affect code that declares both of these forms 20085 of operator delete[], and declared the two-argument form before the 20086 one-argument form. 20087 * The C++ ABI has been changed so that when a parameter is passed by 20088 value, any cleanup for that parameter is performed in the caller, 20089 as specified by the ia64 C++ ABI, rather than the called function 20090 as before. As a result, classes with a non-trivial destructor but a 20091 trivial copy constructor will be passed and returned by invisible 20092 reference, rather than by bitwise copy as before. 20093 * G++ now supports the "named return value optimization": for code 20094 like 20095 A f () { 20096 A a; 20097 ... 20098 return a; 20099 } 20100 20101 G++ will allocate a in the return value slot, so that the return 20102 becomes a no-op. For this to work, all return statements in the 20103 function must return the same variable. 20104 * Improvements to the C++ library are listed in [5]the libstdc++-v3 20105 FAQ. 20106 20107 Objective-C 20108 20109 * Annoying linker warnings (due to incorrect code being generated) 20110 have been fixed. 20111 * If a class method cannot be found, the compiler no longer issues a 20112 warning if a corresponding instance method exists in the root 20113 class. 20114 * Forward @protocol declarations have been fixed. 20115 * Loading of categories has been fixed in certain situations (GNU run 20116 time only). 20117 * The class lookup in the run-time library has been rewritten so that 20118 class method dispatch is more than twice as fast as it used to be 20119 (GNU run time only). 20120 20121 Java 20122 20123 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 20124 javax.transaction. 20125 * Property files and other system resources can be compiled into 20126 executables which use libgcj using the new gcj --resource feature. 20127 * libgcj has been ported to more platforms. In particular there is 20128 now a mostly-functional mingw32 (Windows) target port. 20129 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 20130 Java code can now be called from a C/C++ application. 20131 * gcj can now use builtin functions for certain known methods, for 20132 instance Math.cos. 20133 * gcj can now automatically remove redundant array-store checks in 20134 some common cases. 20135 * The --no-store-checks optimization option was added. This can be 20136 used to omit runtime store checks for code which is known not to 20137 throw ArrayStoreException 20138 * The following third party interface standards were added to libgcj: 20139 org.w3c.dom and org.xml.sax. 20140 * java.security has been merged with GNU Classpath. The new package 20141 is now JDK 1.2 compliant, and much more complete. 20142 * A bytecode verifier was added to the libgcj interpreter. 20143 * java.lang.Character was rewritten to comply with the Unicode 3.0 20144 standard, and improve performance. 20145 * Partial support for many more locales was added to libgcj. 20146 * Socket timeouts have been implemented. 20147 * libgcj has been merged into a single shared library. There are no 20148 longer separate shared libraries for the garbage collector and 20149 zlib. 20150 * Several performance improvements were made to gcj and libgcj: 20151 + Hash synchronization (thin locks) 20152 + A special allocation path for finalizer-free objects 20153 + Thread-local allocation 20154 + Parallel GC, and other GC tweaks 20155 20156 Fortran 20157 20158 Fortran improvements are listed in [6]the Fortran documentation. 20159 20160 Ada 20161 20162 [7]AdaCore, has contributed its GNAT Ada 95 front end and associated 20163 tools. The GNAT compiler fully implements the Ada language as defined 20164 by the ISO/IEC 8652 standard. 20165 20166 Please note that the integration of the Ada front end is still work in 20167 progress. 20168 20169New Targets and Target Specific Improvements 20170 20171 * Hans-Peter Nilsson has contributed a port to MMIX, the CPU 20172 architecture used in new editions of Donald E. Knuth's The Art of 20173 Computer Programming. 20174 * Axis Communications has contributed its port to the CRIS CPU 20175 architecture, used in the ETRAX system-on-a-chip series. 20176 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 20177 SH5 64-bit RISC microprocessor architecture, extending the existing 20178 SH port. 20179 * UltraSPARC is fully supported in 64-bit mode. The option -m64 20180 enables it. 20181 * For compatibility with the Sun compiler #pragma redefine_extname 20182 has been implemented on Solaris. 20183 * The x86 back end has had some noticeable work done to it. 20184 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 20185 Jaeger have contributed a port to the AMD x86-64 architecture. 20186 For more information on x86-64 see http://www.x86-64.org. 20187 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 20188 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 20189 enable the respective instruction sets. Intel C++ compatible 20190 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 20191 will be added in next major release. 20192 + Following those improvements, targets for Pentium MMX, K6-2, 20193 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 20194 added. Refer to the documentation on -march= and -mcpu= 20195 options for details. 20196 + For those targets that support it, -mfpmath=sse will cause the 20197 compiler to generate SSE/SSE2 instructions for floating point 20198 math instead of x87 instructions. Usually, this will lead to 20199 quicker code -- especially on the Pentium 4. Note that only 20200 scalar floating point instructions are used and GCC does not 20201 exploit SIMD features yet. 20202 + Prefetch support has been added to the Pentium III, Pentium 4, 20203 K6-2, K6-3, and Athlon series. 20204 + Code generated for floating point to integer conversions has 20205 been improved leading to better performance of many 3D 20206 applications. 20207 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 20208 * C++ support for AIX has been improved. 20209 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 20210 PowerPC port supporting the AltiVec programming model (SIMD). The 20211 support, though presently useful, is experimental and is expected 20212 to stabilize for 3.2. The support is written to conform to 20213 Motorola's AltiVec specs. See -maltivec. 20214 20215Obsolete Systems 20216 20217 Support for a number of older systems has been declared obsolete in GCC 20218 3.1. Unless there is activity to revive them, the next release of GCC 20219 will have their sources permanently removed. 20220 20221 All configurations of the following processor architectures have been 20222 declared obsolete: 20223 * MIL-STD-1750A, 1750a-*-* 20224 * AMD A29k, a29k-*-* 20225 * Convex, c*-convex-* 20226 * Clipper, clipper-*-* 20227 * Elxsi, elxsi-*-* 20228 * Intel i860, i860-*-* 20229 * Sun picoJava, pj-*-* and pjl-*-* 20230 * Western Electric 32000, we32k-*-* 20231 20232 Most configurations of the following processor architectures have been 20233 declared obsolete, but we are preserving a few systems which may have 20234 active developers. It is unlikely that the remaining systems will 20235 survive much longer unless we see definite signs of port activity. 20236 * Motorola 88000 except 20237 + Generic a.out, m88k-*-aout* 20238 + Generic SVR4, m88k-*-sysv4 20239 + OpenBSD, m88k-*-openbsd* 20240 * NS32k except 20241 + NetBSD, ns32k-*-netbsd* 20242 + OpenBSD, ns32k-*-openbsd*. 20243 * ROMP except 20244 + OpenBSD, romp-*-openbsd*. 20245 20246 Finally, only some configurations of these processor architectures are 20247 being obsoleted. 20248 * Alpha: 20249 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 20250 alpha*-*-osf[45], are still supported.) 20251 * ARM: 20252 + RISCiX, arm-*-riscix*. 20253 * i386: 20254 + 386BSD, i?86-*-bsd* 20255 + Chorus, i?86-*-chorusos* 20256 + DG/UX, i?86-*-dgux* 20257 + FreeBSD 1.x, i?86-*-freebsd1.* 20258 + IBM AIX, i?86-*-aix* 20259 + ISC UNIX, i?86-*-isc* 20260 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 20261 + NEXTstep, i?86-next-* 20262 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 20263 + RTEMS/coff, i?86-*-rtemscoff* 20264 + RTEMS/go32, i?86-go32-rtems* 20265 + Sequent/BSD, i?86-sequent-bsd* 20266 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 20267 i?86-sequent-sysv3* 20268 + SunOS, i?86-*-sunos* 20269 * Motorola 68000: 20270 + Altos, m68[k0]*-altos-* 20271 + Apollo, m68[k0]*-apollo-* 20272 + Apple A/UX, m68[k0]*-apple-* 20273 + Bull, m68[k0]*-bull-* 20274 + Convergent, m68[k0]*-convergent-* 20275 + Generic SVR3, m68[k0]*-*-sysv3* 20276 + ISI, m68[k0]*-isi-* 20277 + LynxOS, m68[k0]*-*-lynxos* 20278 + NEXT, m68[k0]*-next-* 20279 + RTEMS/coff, m68[k0]*-*-rtemscoff* 20280 + Sony, m68[k0]*-sony-* 20281 * MIPS: 20282 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 20283 + Generic BSD, mips-*-bsd* 20284 + Generic System V, mips-*-sysv* 20285 + IRIX before version 5, mips-sgi-irix[1234]* 20286 + RiscOS, mips-*-riscos* 20287 + Sony, mips-sony-* 20288 + Tandem, mips-tandem-* 20289 * SPARC: 20290 + RTEMS/a.out, sparc-*-rtemsaout*. 20291 20292Documentation improvements 20293 20294 * The old manual ("Using and Porting the GNU Compiler Collection") 20295 has been replaced by a users manual ("Using the GNU Compiler 20296 Collection") and a separate internals reference manual ("GNU 20297 Compiler Collection Internals"). 20298 * More complete and much improved documentation about GCC's internal 20299 representation used by the C and C++ front ends. 20300 * Many cleanups and improvements in general. 20301 20302 20303 For questions related to the use of GCC, please consult these web 20304 pages and the [8]GCC manuals. If that fails, the 20305 [9]gcc-help@gcc.gnu.org mailing list might help. Comments on these 20306 web pages and the development of GCC are welcome on our developer 20307 list at [10]gcc@gcc.gnu.org. All of [11]our lists have public 20308 archives. 20309 20310 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 20311 distribution of this entire article is permitted in any medium, 20312 provided this notice is preserved. 20313 20314 These pages are [13]maintained by the GCC team. Last modified 20315 2021-07-28[14]. 20316 20317References 20318 20319 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 20320 2. http://gcc.gnu.org/news/profiledriven.html 20321 3. http://gcc.gnu.org/benchmarks/ 20322 4. http://gcc.gnu.org/c99status.html 20323 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 20324 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 20325 7. https://www.adacore.com/ 20326 8. https://gcc.gnu.org/onlinedocs/ 20327 9. mailto:gcc-help@gcc.gnu.org 20328 10. mailto:gcc@gcc.gnu.org 20329 11. https://gcc.gnu.org/lists.html 20330 12. https://www.fsf.org/ 20331 13. https://gcc.gnu.org/about.html 20332 14. http://validator.w3.org/check/referer 20333====================================================================== 20334http://gcc.gnu.org/gcc-3.0/index.html 20335 20336 GCC 3.0.4 20337 20338 (This release series is no longer supported.) 20339 20340 February 20, 2002 20341 20342 The [1]GNU project and the GCC developers are pleased to announce the 20343 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 20344 series. 20345 20346 GCC used to stand for the GNU C Compiler, but since the compiler 20347 supports several other languages aside from C, it now stands for the 20348 GNU Compiler Collection. 20349 20350 GCC 3.0.x has several new optimizations, new targets, new languages and 20351 many other new features, relative to GCC 2.95.x. See the [2]new 20352 features page for a more complete list. 20353 20354 A list of [3]successful builds is updated as new information becomes 20355 available. 20356 20357 The GCC developers would like to thank the numerous people that have 20358 contributed new features, test results, bug fixes, etc to GCC. This 20359 [4]amazing group of volunteers is what makes GCC successful. 20360 20361 And finally, we can't in good conscience fail to mention some 20362 [5]caveats to using GCC 3.0.x. 20363 20364 For additional information about GCC please refer to the [6]GCC project 20365 web site or contact the [7]GCC development mailing list. 20366 20367 To obtain GCC please use [8]our mirror sites, or our CVS server. 20368 __________________________________________________________________ 20369 20370Previous 3.0.x Releases 20371 20372 December 20, 2001: GCC 3.0.3 has been released. 20373 October 25, 2001: GCC 3.0.2 has been released. 20374 August 20, 2001: GCC 3.0.1 has been released. 20375 June 18, 2001: GCC 3.0 has been released. 20376 20377 20378 For questions related to the use of GCC, please consult these web 20379 pages and the [9]GCC manuals. If that fails, the 20380 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these 20381 web pages and the development of GCC are welcome on our developer 20382 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public 20383 archives. 20384 20385 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 20386 distribution of this entire article is permitted in any medium, 20387 provided this notice is preserved. 20388 20389 These pages are [14]maintained by the GCC team. Last modified 20390 2021-07-28[15]. 20391 20392References 20393 20394 1. http://www.gnu.org/ 20395 2. http://gcc.gnu.org/gcc-3.0/features.html 20396 3. http://gcc.gnu.org/gcc-3.0/buildstat.html 20397 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 20398 5. http://gcc.gnu.org/gcc-3.0/caveats.html 20399 6. http://gcc.gnu.org/index.html 20400 7. mailto:gcc@gcc.gnu.org 20401 8. http://gcc.gnu.org/mirrors.html 20402 9. https://gcc.gnu.org/onlinedocs/ 20403 10. mailto:gcc-help@gcc.gnu.org 20404 11. mailto:gcc@gcc.gnu.org 20405 12. https://gcc.gnu.org/lists.html 20406 13. https://www.fsf.org/ 20407 14. https://gcc.gnu.org/about.html 20408 15. http://validator.w3.org/check/referer 20409====================================================================== 20410http://gcc.gnu.org/gcc-3.0/features.html 20411 20412 GCC 3.0 New Features 20413 20414Additional changes in GCC 3.0.4 20415 20416 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 20417 system, which use the ELF object file format, on x86 processors. 20418 * Correct debugging information is generated from functions that have 20419 lines from multiple files (e.g. yacc output). 20420 * A fix for whitespace handling in the -traditional preprocessor, 20421 which can affect Fortran. 20422 * Fixes to the exception handling runtime. 20423 * More fixes for bad code generation in C++. 20424 * A fix for shared library generation under AIX 4.3. 20425 * Documentation updates. 20426 * Port of GCC to Tensilica's Xtensa processor contributed. 20427 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 20428 20429Additional changes in GCC 3.0.3 20430 20431 * A fix to correct an accidental change to the PowerPC ABI. 20432 * Fixes for bad code generation on a variety of architectures. 20433 * Improvements to the debugging information generated for C++ 20434 classes. 20435 * Fixes for bad code generation in C++. 20436 * A fix to avoid crashes in the C++ demangler. 20437 * A fix to the C++ standard library to avoid buffer overflows. 20438 * Miscellaneous improvements for a variety of architectures. 20439 20440Additional changes in GCC 3.0.2 20441 20442 * Fixes for bad code generation during loop unrolling. 20443 * Fixes for bad code generation by the sibling call optimization. 20444 * Minor improvements to x86 code generation. 20445 * Implementation of function descriptors in C++ vtables for IA64. 20446 * Numerous minor bug-fixes. 20447 20448Additional changes in GCC 3.0.1 20449 20450 * C++ fixes for incorrect code-generation. 20451 * Improved cross-compiling support for the C++ standard library. 20452 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 20453 in GCC 3.0. 20454 * Fixes for various exception-handling bugs. 20455 * A port to the S/390 architecture. 20456 20457General Optimizer Improvements 20458 20459 * [2]Basic block reordering pass. 20460 * New if-conversion pass with support for conditional (predicated) 20461 execution. 20462 * New tail call and sibling call elimination optimizations. 20463 * New register renaming pass. 20464 * New (experimental) [3]static single assignment (SSA) representation 20465 support. 20466 * New dead-code elimination pass implemented using the SSA 20467 representation. 20468 * [4]Global null pointer test elimination. 20469 * [5]Global code hoisting/unification. 20470 * More builtins and optimizations for stdio.h, string.h and old BSD 20471 functions, as well as for ISO C99 functions. 20472 * New builtin __builtin_expect for giving hints to the branch 20473 predictor. 20474 20475New Languages and Language specific improvements 20476 20477 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 20478 and supported, including the run-time library containing most 20479 common non-GUI Java classes, a bytecode interpreter, and the Boehm 20480 conservative garbage collector. Many bugs have been fixed. GCJ can 20481 compile Java source or Java bytecodes to either native code or Java 20482 class files, and supports native methods written in either the 20483 standard JNI or the more efficient and convenient CNI. 20484 * Here is a [6]partial list of C++ improvements, both new features 20485 and those no longer supported. 20486 * New C++ ABI. On the IA-64 platform GCC is capable of 20487 inter-operating with other IA-64 compilers. 20488 * The new ABI also significantly reduces the size of symbol and debug 20489 information. 20490 * New C++ support library and many C++ bug fixes, vastly improving 20491 our conformance to the ISO C++ standard. 20492 * New [7]inliner for C++. 20493 * Rewritten C preprocessor, integrated into the C, C++ and Objective 20494 C compilers, with very many improvements including ISO C99 support 20495 and [8]improvements to dependency generation. 20496 * Support for more [9]ISO C99 features. 20497 * Many improvements to support for checking calls to format functions 20498 such as printf and scanf, including support for ISO C99 format 20499 features, extensions from the Single Unix Specification and GNU 20500 libc 2.2, checking of strfmon formats and features to assist in 20501 auditing for format string security bugs. 20502 * New warnings for C code that may have undefined semantics because 20503 of violations of sequence point rules in the C standard (such as a 20504 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 20505 * Additional warning option -Wfloat-equal. 20506 * Improvements to -Wtraditional. 20507 * Fortran improvements are listed in [10]the Fortran documentation. 20508 20509New Targets and Target Specific Improvements 20510 20511 * New x86 back end, generating much improved code. 20512 * Support for a generic i386-elf target contributed. 20513 * New option to emit x86 assembly code using Intel style syntax 20514 (-mintel-syntax). 20515 * HPUX 11 support contributed. 20516 * Improved PowerPC code generation, including scheduled prologue and 20517 epilogue. 20518 * Port of GCC to Intel's IA-64 processor contributed. 20519 * Port of GCC to Motorola's MCore 210 and 340 contributed. 20520 * New unified back-end for Arm, Thumb and StrongArm contributed. 20521 * Port of GCC to Intel's XScale processor contributed. 20522 * Port of GCC to Atmel's AVR microcontrollers contributed. 20523 * Port of GCC to Mitsubishi's D30V processor contributed. 20524 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 20525 processor family) contributed. 20526 * Port of GCC to Fujitsu's FR30 processor contributed. 20527 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 20528 * Port of GCC to Sun's picoJava processor core contributed. 20529 20530Documentation improvements 20531 20532 * Substantially rewritten and improved C preprocessor manual. 20533 * Many improvements to other documentation. 20534 * Manpages for gcc, cpp and gcov are now generated automatically from 20535 the master Texinfo manual, eliminating the problem of manpages 20536 being out of date. (The generated manpages are only extracts from 20537 the full manual, which is provided in Texinfo form, from which 20538 info, HTML, other formats and a printed manual can be generated.) 20539 * Generated info files are included in the release tarballs alongside 20540 their Texinfo sources, avoiding problems on some platforms with 20541 building makeinfo as part of the GCC distribution. 20542 20543Other significant improvements 20544 20545 * Garbage collection used internally by the compiler for most memory 20546 allocation instead of obstacks. 20547 * Lengauer and Tarjan algorithm used for computing dominators in the 20548 CFG. This algorithm can be significantly faster and more space 20549 efficient than our older algorithm. 20550 * gccbug script provided to assist in submitting bug reports to our 20551 bug tracking system. (Bug reports previously submitted directly to 20552 our mailing lists, for which you received no bug tracking number, 20553 should be submitted again using gccbug if you can reproduce the 20554 problem with GCC 3.0.) 20555 * The internal libgcc library is [11]built as a shared library on 20556 systems that support it. 20557 * Extensive testsuite included with GCC, with many new tests. In 20558 addition to tests for GCC bugs that have been fixed, many tests 20559 have been added for language features, compiler warnings and 20560 builtin functions. 20561 * Additional language-independent warning options -Wpacked, -Wpadded, 20562 -Wunreachable-code and -Wdisabled-optimization. 20563 * Target-independent options -falign-functions, -falign-loops and 20564 -falign-jumps. 20565 20566 Plus a great many bug fixes and almost all the [12]features found in 20567 GCC 2.95. 20568 20569 20570 For questions related to the use of GCC, please consult these web 20571 pages and the [13]GCC manuals. If that fails, the 20572 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these 20573 web pages and the development of GCC are welcome on our developer 20574 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public 20575 archives. 20576 20577 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 20578 distribution of this entire article is permitted in any medium, 20579 provided this notice is preserved. 20580 20581 These pages are [18]maintained by the GCC team. Last modified 20582 2021-07-28[19]. 20583 20584References 20585 20586 1. http://www.netbsd.org/ 20587 2. http://gcc.gnu.org/news/reorder.html 20588 3. http://gcc.gnu.org/news/ssa.html 20589 4. http://gcc.gnu.org/news/null.html 20590 5. http://gcc.gnu.org/news/unify.html 20591 6. http://gcc.gnu.org/gcc-3.0/c++features.html 20592 7. http://gcc.gnu.org/news/inlining.html 20593 8. http://gcc.gnu.org/news/dependencies.html 20594 9. http://gcc.gnu.org/c99status.html 20595 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 20596 11. http://gcc.gnu.org/gcc-3.0/libgcc.html 20597 12. http://gcc.gnu.org/gcc-2.95/features.html 20598 13. https://gcc.gnu.org/onlinedocs/ 20599 14. mailto:gcc-help@gcc.gnu.org 20600 15. mailto:gcc@gcc.gnu.org 20601 16. https://gcc.gnu.org/lists.html 20602 17. https://www.fsf.org/ 20603 18. https://gcc.gnu.org/about.html 20604 19. http://validator.w3.org/check/referer 20605====================================================================== 20606http://gcc.gnu.org/gcc-3.0/caveats.html 20607 20608 GCC 3.0 Caveats 20609 20610 * -fstrict-aliasing is now part of -O2 and higher optimization 20611 levels. This allows the compiler to assume the strictest aliasing 20612 rules applicable to the language being compiled. For C and C++, 20613 this activates optimizations based on the type of expressions. This 20614 optimization may thus break old, non-compliant code. 20615 * Enumerations are now properly promoted to int in function 20616 parameters and function returns. Normally this change is not 20617 visible, but when using -fshort-enums this is an ABI change. 20618 * The undocumented extension that allowed C programs to have a label 20619 at the end of a compound statement has been deprecated and may be 20620 removed in a future version. Programs that now generate a warning 20621 about this may be fixed by adding a null statement (a single 20622 semicolon) after the label. 20623 * The poorly documented extension that allowed string constants in C, 20624 C++ and Objective C to contain unescaped newlines has been 20625 deprecated and may be removed in a future version. Programs using 20626 this extension may be fixed in several ways: the bare newline may 20627 be replaced by \n, or preceded by \n\, or string concatenation may 20628 be used with the bare newline preceded by \n" and " placed at the 20629 start of the next line. 20630 * The Chill compiler is not included in GCC 3.0, because of the lack 20631 of a volunteer to convert it to use garbage collection. 20632 * Certain non-standard iostream methods from earlier versions of 20633 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 20634 ostream::form, and istream::gets. 20635 * The new C++ ABI is not yet fully supported by current (as of 20636 2001-07-01) releases and development versions of GDB, or any 20637 earlier versions. There is a problem setting breakpoints by line 20638 number, and other related issues that have been fixed in GCC 3.0 20639 but not yet handled in GDB: 20640 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 20641 20642 20643 For questions related to the use of GCC, please consult these web 20644 pages and the [2]GCC manuals. If that fails, the 20645 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these 20646 web pages and the development of GCC are welcome on our developer 20647 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives. 20648 20649 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 20650 distribution of this entire article is permitted in any medium, 20651 provided this notice is preserved. 20652 20653 These pages are [7]maintained by the GCC team. Last modified 20654 2021-07-28[8]. 20655 20656References 20657 20658 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 20659 2. https://gcc.gnu.org/onlinedocs/ 20660 3. mailto:gcc-help@gcc.gnu.org 20661 4. mailto:gcc@gcc.gnu.org 20662 5. https://gcc.gnu.org/lists.html 20663 6. https://www.fsf.org/ 20664 7. https://gcc.gnu.org/about.html 20665 8. http://validator.w3.org/check/referer 20666====================================================================== 20667http://gcc.gnu.org/gcc-2.95/index.html 20668 20669 GCC 2.95 20670 20671 (This release series is no longer supported.) 20672 20673 March 16, 2001: The GNU project and the GCC developers are pleased to 20674 announce the release of GCC version 2.95.3. 20675 20676Release History 20677 20678 GCC 2.95.3 20679 March 16, 2001 20680 20681 GCC 2.95.2 20682 October 27, 1999 20683 20684 GCC 2.95.1 20685 August 19, 1999 20686 20687 GCC 2.95 20688 July 31, 1999. This is the first release of GCC since the April 20689 1999 GCC/EGCS reunification and includes nearly a year's worth 20690 of new development and bugfixes. 20691 20692References and Acknowledgements 20693 20694 GCC used to stand for the GNU C Compiler, but since the compiler 20695 supports several other languages aside from C, it now stands for the 20696 GNU Compiler Collection. 20697 20698 The whole suite has been extensively [1]regression tested and 20699 [2]package tested. It should be reliable and suitable for widespread 20700 use. 20701 20702 The compiler has several new optimizations, new targets, new languages 20703 and other new features. See the [3]new features page for a more 20704 complete list of new features found in the GCC 2.95 releases. 20705 20706 The sources include installation instructions in both HTML and 20707 plaintext forms in the install directory in the distribution. However, 20708 the most up to date installation instructions and [4]build/test status 20709 are on the web pages. We will update those pages as new information 20710 becomes available. 20711 20712 The GCC developers would like to thank the numerous people that have 20713 contributed new features, test results, bugfixes, etc to GCC. This 20714 [5]amazing group of volunteers is what makes GCC successful. 20715 20716 And finally, we can't in good conscience fail to mention some 20717 [6]caveats to using GCC 2.95. 20718 20719 Download GCC 2.95 from one of our many [7]mirror sites. 20720 20721 For additional information about GCC please see the [8]GCC project web 20722 server or contact the [9]GCC development mailing list. 20723 20724 20725 For questions related to the use of GCC, please consult these web 20726 pages and the [10]GCC manuals. If that fails, the 20727 [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these 20728 web pages and the development of GCC are welcome on our developer 20729 list at [12]gcc@gcc.gnu.org. All of [13]our lists have public 20730 archives. 20731 20732 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 20733 distribution of this entire article is permitted in any medium, 20734 provided this notice is preserved. 20735 20736 These pages are [15]maintained by the GCC team. Last modified 20737 2021-07-28[16]. 20738 20739References 20740 20741 1. http://gcc.gnu.org/gcc-2.95/regress.html 20742 2. http://gcc.gnu.org/gcc-2.95/othertest.html 20743 3. http://gcc.gnu.org/gcc-2.95/features.html 20744 4. http://gcc.gnu.org/gcc-2.95/buildstat.html 20745 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 20746 6. http://gcc.gnu.org/gcc-2.95/caveats.html 20747 7. http://gcc.gnu.org/mirrors.html 20748 8. http://gcc.gnu.org/index.html 20749 9. mailto:gcc@gcc.gnu.org 20750 10. https://gcc.gnu.org/onlinedocs/ 20751 11. mailto:gcc-help@gcc.gnu.org 20752 12. mailto:gcc@gcc.gnu.org 20753 13. https://gcc.gnu.org/lists.html 20754 14. https://www.fsf.org/ 20755 15. https://gcc.gnu.org/about.html 20756 16. http://validator.w3.org/check/referer 20757====================================================================== 20758http://gcc.gnu.org/gcc-2.95/features.html 20759 20760 GCC 2.95 New Features 20761 20762 * General Optimizer Improvements: 20763 + [1]Localized register spilling to improve speed and code 20764 density especially on small register class machines. 20765 + [2]Global CSE using lazy code motion algorithms. 20766 + [3]Improved global constant/copy propagation. 20767 + [4]Improved control flow graph analysis and manipulation. 20768 + [5]Local dead store elimination. 20769 + [6]Memory Load hoisting/store sinking in loops. 20770 + [7]Type based alias analysis is enabled by default. Note this 20771 feature will expose bugs in the Linux kernel. Please refer to 20772 the FAQ (as shipped with GCC 2.95) for additional information 20773 on this issue. 20774 + Major revamp of GIV detection, combination and simplification 20775 to improve loop performance. 20776 + Major improvements to register allocation and reloading. 20777 * New Languages and Language specific improvements 20778 + [8]Many C++ improvements. 20779 + [9]Many Fortran improvements. 20780 + [10]Java front-end has been integrated. A [11]runtime library 20781 is available separately. 20782 + [12]ISO C99 support 20783 + [13]Chill front-end and runtime has been integrated. 20784 + Boehm garbage collector support in libobjc. 20785 + More support for various pragmas which appear in vendor 20786 include files 20787 * New Targets and Target Specific Improvements 20788 + [14]SPARC backend rewrite. 20789 + -mschedule=8000 will optimize code for PA8000 class 20790 processors; -mpa-risc-2-0 will generate code for PA2.0 20791 processors 20792 + Various micro-optimizations for the ia32 port. K6 20793 optimizations 20794 + Compiler will attempt to align doubles in the stack on the 20795 ia32 port 20796 + Alpha EV6 support 20797 + PowerPC 750 20798 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 20799 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 20800 + c3x, c4x 20801 + HyperSPARC 20802 + SparcLite86x 20803 + sh4 20804 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 20805 arm-linux) 20806 + vxWorks targets include support for vxWorks threads 20807 + StrongARM 110 and ARM9 support added. ARM Scheduling 20808 parameters rewritten. 20809 + Various changes to the MIPS port to avoid assembler macros, 20810 which in turn improves performance 20811 + Various performance improvements to the i960 port. 20812 + Major rewrite of ns32k port 20813 * Other significant improvements 20814 + [15]Ability to dump cfg information and display it using vcg. 20815 + The new faster scheme for fixing vendor header files is 20816 enabled by default. 20817 + Experimental internationalization support. 20818 + multibyte character support 20819 + Some compile-time speedups for pathological problems 20820 + Better support for complex types 20821 * Plus the usual mountain of bugfixes 20822 * Core compiler is based on the gcc2 development tree from Sept 30, 20823 1998, so we have all of the [16]features found in GCC 2.8. 20824 20825Additional Changes in GCC 2.95.1 20826 20827 * Generic bugfixes and improvements 20828 + Various documentation fixes related to the GCC/EGCS merger. 20829 + Fix memory management bug which could lead to spurious aborts, 20830 core dumps or random parsing errors in the compiler. 20831 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 20832 support. 20833 + Fix infinite loop in the CSE optimizer. 20834 + Avoid undefined behavior in compiler FP emulation code 20835 + Fix install problem when prefix is overridden on the make 20836 install command. 20837 + Fix problem with unwanted installation of assert.h on some 20838 systems. 20839 + Fix problem with finding the wrong assembler in a single tree 20840 build. 20841 + Avoid increasing the known alignment of a register that is 20842 already known to be a pointer. 20843 * Platform specific bugfixes and improvements 20844 + Codegen bugfix for prologue/epilogue for cpu32 target. 20845 + Fix long long code generation bug for the Coldfire target. 20846 + Fix various aborts in the SH compiler. 20847 + Fix bugs in libgcc support library for the SH. 20848 + Fix alpha ev6 code generation bug. 20849 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 20850 AIX platforms. 20851 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 20852 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 20853 targets. 20854 + Fix weak symbol handling for rs6000/ppc svr4 targets. 20855 + Fix various problems with 64bit code generation for the 20856 rs6000/ppc port. 20857 + Fix codegen bug which caused tetex to be mis-compiled on the 20858 x86. 20859 + Fix compiler abort in new cfg code exposed by x86 port. 20860 + Fix out of range array reference in code convert flat 20861 registers to the x87 stacked FP register file. 20862 + Fix minor vxworks configuration bug. 20863 + Fix return type of bsearch for SunOS 4.x. 20864 * Language & Runtime specific fixes. 20865 + The G++ signature extension has been deprecated. It will be 20866 removed in the next major release of G++. Use of signatures 20867 will result in a warning from the compiler. 20868 + Several bugs relating to templates and namespaces were fixed. 20869 + A bug that caused crashes when combining templates with -g on 20870 DWARF1 platforms was fixed. 20871 + Pointers-to-members, virtual functions, and multiple 20872 inheritance should now work together correctly. 20873 + Some code-generation bugs relating to function try blocks were 20874 fixed. 20875 + G++ is a little bit more lenient with certain archaic 20876 constructs than in GCC 2.95. 20877 + Fix to prevent shared library version #s from bring truncated 20878 to 1 digit 20879 + Fix missing std:: in the libstdc++ library. 20880 + Fix stream locking problems in libio. 20881 + Fix problem in java compiler driver. 20882 20883Additional Changes in GCC 2.95.2 20884 20885 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 20886 the optimizations performed by -fstrict-aliasing are valid according to 20887 the C and C++ standards, the optimization have caused some problems, 20888 particularly with old non-conforming code. 20889 20890 The GCC developers are experimenting with ways to warn users about code 20891 which violates the C/C++ standards, but those warnings are not ready 20892 for widespread use at this time. Rather than wait for those warnings 20893 the GCC developers have chosen to disable -fstrict-aliasing by default 20894 for the GCC 2.95.2 release. 20895 20896 We strongly encourage developers to find and fix code which violates 20897 the C/C++ standards as -fstrict-aliasing may be enabled by default in 20898 future releases. Use the option -fstrict-aliasing to re-enable these 20899 optimizations. 20900 * Generic bugfixes and improvements 20901 + Fix incorrectly optimized memory reference in global common 20902 subexpression elimination (GCSE) optimization pass. 20903 + Fix code generation bug in regmove.c in which it could 20904 incorrectly change a "const" value. 20905 + Fix bug in optimization of conditionals involving volatile 20906 memory references. 20907 + Avoid over-allocation of stack space for some procedures. 20908 + Fixed bug in the compiler which caused incorrect optimization 20909 of an obscure series of bit manipulations, shifts and 20910 arithmetic. 20911 + Fixed register allocator bug which caused teTeX to be 20912 mis-compiled on SPARC targets. 20913 + Avoid incorrect optimization of degenerate case statements for 20914 certain targets such as the ARM. 20915 + Fix out of range memory reference in the jump optimizer. 20916 + Avoid dereferencing null pointer in fix-header. 20917 + Fix test for GCC specific features so that it is possible to 20918 bootstrap with gcc-2.6.2 and older versions of GCC. 20919 + Fix typo in scheduler which could potentially cause out of 20920 range memory accesses. 20921 + Avoid incorrect loop reversal which caused incorrect code for 20922 certain loops on PowerPC targets. 20923 + Avoid incorrect optimization of switch statements on certain 20924 targets (for example the ARM). 20925 * Platform specific bugfixes and improvements 20926 + Work around bug in Sun V5.0 compilers which caused bootstrap 20927 comparison failures on SPARC targets. 20928 + Fix SPARC backend bug which caused aborts in final.c. 20929 + Fix sparc-hal-solaris2* configuration fragments. 20930 + Fix bug in sparc block profiling. 20931 + Fix obscure code generation bug for the PARISC targets. 20932 + Define __STDC_EXT__ for HPUX configurations. 20933 + Various POWERPC64 code generation bugfixes. 20934 + Fix abort for PPC targets using ELF (ex GNU/Linux). 20935 + Fix collect2 problems for AIX targets. 20936 + Correct handling of .file directive for PPC targets. 20937 + Fix bug in fix_trunc x86 patterns. 20938 + Fix x86 port to correctly pop the FP stack for functions that 20939 return structures in memory. 20940 + Fix minor bug in strlen x86 pattern. 20941 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 20942 + Fix template repository code to handle leading underscore in 20943 mangled names. 20944 + Fix weak/weak alias support for OpenBSD. 20945 + GNU/Linux for the ARM has C++ compatible include files. 20946 * Language & Runtime specific fixes. 20947 + Fix handling of constructor attribute in the C front-end which 20948 caused problems building the Chill runtime library on some 20949 targets. 20950 + Fix minor problem merging type qualifiers in the C front-end. 20951 + Fix aliasing bug for pointers and references (C/C++). 20952 + Fix incorrect "non-constant initializer bug" when -traditional 20953 or -fwritable-strings is enabled. 20954 + Fix build error for Chill front-end on SunOS. 20955 + Do not complain about duplicate instantiations when using 20956 -frepo (C++). 20957 + Fix array bounds handling in C++ front-end which caused 20958 problems with dwarf debugging information in some 20959 circumstances. 20960 + Fix minor namespace problem. 20961 + Fix problem linking java programs. 20962 20963Additional Changes in GCC 2.95.3 20964 20965 * Generic bugfixes and improvements 20966 + Fix numerous problems that caused incorrect optimization in 20967 the register reloading code. 20968 + Fix numerous problems that caused incorrect optimization in 20969 the loop optimizer. 20970 + Fix aborts in the functions build_insn_chain and scan_loops 20971 under some circumstances. 20972 + Fix an alias analysis bug. 20973 + Fix an infinite compilation bug in the combiner. 20974 + A few problems with complex number support have been fixed. 20975 + It is no longer possible for gcc to act as a fork bomb when 20976 installed incorrectly. 20977 + The -fpack-struct option should be recognized now. 20978 + Fixed a bug that caused incorrect code to be generated due to 20979 a lost stack adjustment. 20980 * Platform specific bugfixes and improvements 20981 + Support building ARM toolchains hosted on Windows. 20982 + Fix attribute calculations in ARM toolchains. 20983 + arm-linux support has been improved. 20984 + Fix a PIC failure on sparc targets. 20985 + On ix86 targets, the regparm attribute should now work 20986 reliably. 20987 + Several updates for the h8300 port. 20988 + Fix problem building libio with glibc 2.2. 20989 20990 20991 For questions related to the use of GCC, please consult these web 20992 pages and the [17]GCC manuals. If that fails, the 20993 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these 20994 web pages and the development of GCC are welcome on our developer 20995 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public 20996 archives. 20997 20998 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 20999 distribution of this entire article is permitted in any medium, 21000 provided this notice is preserved. 21001 21002 These pages are [22]maintained by the GCC team. Last modified 21003 2021-07-28[23]. 21004 21005References 21006 21007 1. http://gcc.gnu.org/news/spill.html 21008 2. http://gcc.gnu.org/news/lcm.html 21009 3. http://gcc.gnu.org/news/cprop.html 21010 4. http://gcc.gnu.org/news/cfg.html 21011 5. http://gcc.gnu.org/news/dse.html 21012 6. http://gcc.gnu.org/news/hoist.html 21013 7. http://gcc.gnu.org/news/alias.html 21014 8. http://gcc.gnu.org/gcc-2.95/c++features.html 21015 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 21016 10. http://gcc.gnu.org/news/gcj-announce.txt 21017 11. http://gcc.gnu.org/news/javaannounce.html 21018 12. http://gcc.gnu.org/c99status.html 21019 13. http://gcc.gnu.org/news/chill.html 21020 14. http://gcc.gnu.org/news/sparc.html 21021 15. http://gcc.gnu.org/news/egcs-vcg.html 21022 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html 21023 17. https://gcc.gnu.org/onlinedocs/ 21024 18. mailto:gcc-help@gcc.gnu.org 21025 19. mailto:gcc@gcc.gnu.org 21026 20. https://gcc.gnu.org/lists.html 21027 21. https://www.fsf.org/ 21028 22. https://gcc.gnu.org/about.html 21029 23. http://validator.w3.org/check/referer 21030====================================================================== 21031http://gcc.gnu.org/gcc-2.95/caveats.html 21032 21033 GCC 2.95 Caveats 21034 21035 * GCC 2.95 will issue an error for invalid asm statements that had 21036 been silently accepted by earlier versions of the compiler. This is 21037 particularly noticeable when compiling older versions of the Linux 21038 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 21039 for more information on this issue. 21040 * GCC 2.95 implements type based alias analysis to disambiguate 21041 memory references. Some programs, particularly the Linux kernel 21042 violate ANSI/ISO aliasing rules and therefore may not operate 21043 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 21044 shipped with GCC 2.95) for more information on this issue. 21045 * GCC 2.95 has a known bug in its handling of complex variables for 21046 64bit targets. Instead of silently generating incorrect code, GCC 21047 2.95 will issue a fatal error for situations it can not handle. 21048 This primarily affects the Fortran community as Fortran makes more 21049 use of complex variables than C or C++. 21050 * GCC 2.95 has an integrated libstdc++, but does not have an 21051 integrated libg++. Furthermore old libg++ releases will not work 21052 with GCC 2.95. You can retrieve a recent copy of libg++ from the 21053 [1]GCC ftp server. 21054 Note most C++ programs only need libstdc++. 21055 * Exception handling may not work with shared libraries, particularly 21056 on alphas, hppas, rs6000/powerpc and mips based platforms. 21057 Exception handling is known to work on x86 GNU/Linux platforms with 21058 shared libraries. 21059 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 21060 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 21061 or EGCS 1.1. As a result it may be necessary to fix C++ code before 21062 it will compile with GCC 2.95. 21063 * G++ is also converting toward the ISO C++ standard; as a result 21064 code which was previously valid (and thus accepted by other 21065 compilers and older versions of g++) may no longer be accepted. The 21066 flag -fpermissive may allow some non-conforming code to compile 21067 with GCC 2.95. 21068 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 21069 1.1.x, EGCS 1.0.x or GCC 2.8.x. 21070 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 21071 between Sept 30, 1998 and April 30, 1999 (the official end of the 21072 GCC 2.8 project). Future GCC releases will include all the changes 21073 from the defunct GCC 2.8 sources. 21074 21075 21076 For questions related to the use of GCC, please consult these web 21077 pages and the [2]GCC manuals. If that fails, the 21078 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these 21079 web pages and the development of GCC are welcome on our developer 21080 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives. 21081 21082 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 21083 distribution of this entire article is permitted in any medium, 21084 provided this notice is preserved. 21085 21086 These pages are [7]maintained by the GCC team. Last modified 21087 2021-07-28[8]. 21088 21089References 21090 21091 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 21092 2. https://gcc.gnu.org/onlinedocs/ 21093 3. mailto:gcc-help@gcc.gnu.org 21094 4. mailto:gcc@gcc.gnu.org 21095 5. https://gcc.gnu.org/lists.html 21096 6. https://www.fsf.org/ 21097 7. https://gcc.gnu.org/about.html 21098 8. http://validator.w3.org/check/referer 21099====================================================================== 21100http://gcc.gnu.org/egcs-1.1/index.html 21101 21102 EGCS 1.1 21103 21104 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 21105 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 21106 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 21107 21108 EGCS is a free software project to further the development of the GNU 21109 compilers using an open development environment. 21110 21111 EGCS 1.1 is a major new release of the EGCS compiler system. It has 21112 been [1]extensively tested and is believed to be stable and suitable 21113 for widespread use. 21114 21115 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 21116 development sources; it contains all of the new features found in GCC 21117 2.8.1 as well as all new development from GCC up to June 6, 1998. 21118 21119 EGCS 1.1 also contains many improvements and features not found in GCC 21120 or in older versions of EGCS: 21121 * Global common subexpression elimination and global constant/copy 21122 propagation (aka [2]gcse) 21123 * Ongoing improvements to the [3]alias analysis support to allow for 21124 better optimizations throughout the compiler. 21125 * Vastly improved [4]C++ compiler and integrated C++ runtime 21126 libraries. 21127 * Fixes for the /tmp symlink race security problems. 21128 * New targets including mips16, arm-thumb and 64 bit PowerPC. 21129 * Improvements to GNU Fortran (g77) compiler and runtime library made 21130 since g77 version 0.5.23. 21131 21132 See the [5]new features page for a more complete list of new features 21133 found in EGCS 1.1 releases. 21134 21135 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 21136 1.1: 21137 * General improvements and fixes 21138 + Avoid some stack overflows when compiling large functions. 21139 + Avoid incorrect loop invariant code motions. 21140 + Fix some core dumps on Linux kernel code. 21141 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 21142 + Fix code generation problem in gcse. 21143 + Various documentation related fixes. 21144 * g++/libstdc++ improvements and fixes 21145 + MT safe EH fix for setjmp/longjmp based exception handling. 21146 + Fix a few bad interactions between optimization and exception 21147 handling. 21148 + Fixes for demangling of template names starting with "__". 21149 + Fix a bug that would fail to run destructors in some cases 21150 with -O2. 21151 + Fix 'new' of classes with virtual bases. 21152 + Fix crash building Qt on the Alpha. 21153 + Fix failure compiling WIFEXITED macro on GNU/Linux. 21154 + Fix some -frepo failures. 21155 * g77 and libf2c improvements and fixes 21156 + Various documentation fixes. 21157 + Avoid compiler crash on RAND intrinsic. 21158 + Fix minor bugs in makefiles exposed by BSD make programs. 21159 + Define _XOPEN_SOURCE for libI77 build to avoid potential 21160 problems on some 64-bit systems. 21161 + Fix problem with implicit endfile on rewind. 21162 + Fix spurious recursive I/O errors. 21163 * platform specific improvements and fixes 21164 + Match all versions of UnixWare7. 21165 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 21166 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 21167 from unsigned ints to double precision floats. 21168 + Fix ARM ABI issue with NetBSD. 21169 + Fix a few arm code generation bugs. 21170 + Fixincludes will fix additional broken SCO OpenServer header 21171 files. 21172 + Fix a m68k backend bug which caused invalid offsets in reg+d 21173 addresses. 21174 + Fix problems with 64bit AIX 4.3 support. 21175 + Fix handling of long longs for varargs/stdarg functions on the 21176 ppc. 21177 + Minor fixes to CPP predefines for Windows. 21178 + Fix code generation problems with gpr<->fpr copies for 64bit 21179 ppc. 21180 + Fix a few coldfire code generation bugs. 21181 + Fix some more header file problems on SunOS 4.x. 21182 + Fix assert.h handling for RTEMS. 21183 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 21184 + Fix x86 compiler abort in reg-stack pass. 21185 + Fix cygwin/windows problem with section attributes. 21186 + Fix Alpha code generation problem exposed by SMP Linux 21187 kernels. 21188 + Fix typo in m68k 32->64bit integer conversion. 21189 + Make sure target libraries build with -fPIC for PPC & Alpha 21190 targets. 21191 21192 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 21193 1.1.1: 21194 * General improvements and fixes 21195 + Fix bug in loop optimizer which caused the SPARC (and 21196 potentially other) ports to segfault. 21197 + Fix infinite recursion in alias analysis and combiner code. 21198 + Fix bug in regclass preferencing. 21199 + Fix incorrect loop reversal which caused incorrect code to be 21200 generated for several targets. 21201 + Fix return value for builtin memcpy. 21202 + Reduce compile time for certain loops which exposed quadratic 21203 behavior in the loop optimizer. 21204 + Fix bug which caused volatile memory to be written multiple 21205 times when only one write was needed/desired. 21206 + Fix compiler abort in caller-save.c 21207 + Fix combiner bug which caused incorrect code generation for 21208 certain division by constant operations. 21209 + Fix incorrect code generation due to a bug in range check 21210 optimizations. 21211 + Fix incorrect code generation due to mis-handling of clobbered 21212 values in CSE. 21213 + Fix compiler abort/segfault due to incorrect register 21214 splitting when unrolling loops. 21215 + Fix code generation involving autoincremented addresses with 21216 ternary operators. 21217 + Work around bug in the scheduler which caused qt to be 21218 mis-compiled on some platforms. 21219 + Fix code generation problems with -fshort-enums. 21220 + Tighten security for temporary files. 21221 + Improve compile time for codes which make heavy use of 21222 overloaded functions. 21223 + Fix multiply defined constructor/destructor symbol problems. 21224 + Avoid setting bogus RPATH environment variable during 21225 bootstrap. 21226 + Avoid GNU-make dependencies in the texinfo subdir. 21227 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 21228 --enable-cpp=<dirname> can be used to specify an additional 21229 install directory for the cpp wrapper script. 21230 + Fix CSE bug which caused incorrect label-label refs to appear 21231 on some platforms. 21232 + Avoid linking in EH routines from libgcc if they are not 21233 needed. 21234 + Avoid obscure bug in aliasing code. 21235 + Fix bug in weak symbol handling. 21236 * Platform-specific improvements and fixes 21237 + Fix detection of PPro/PII on Unixware 7. 21238 + Fix compiler segfault when building spec99 and other programs 21239 for SPARC targets. 21240 + Fix code-generation bugs for integer and floating point 21241 conditional move instructions on the PPro/PII. 21242 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 21243 + Fix build failure for the arc port. 21244 + Fix floating point format configuration for i?86-gnu port. 21245 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 21246 threads are enabled. 21247 + Fix coldfire code generation bugs. 21248 + Fix "unrecognized insn" problems for Alpha and PPC ports. 21249 + Fix h8/300 code generation problem with floating point values 21250 in memory. 21251 + Fix unrecognized insn problems for the m68k port. 21252 + Fix namespace-pollution problem for the x86 port. 21253 + Fix problems with old assembler on x86 NeXT systems. 21254 + Fix PIC code-generation problems for the SPARC port. 21255 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 21256 + Fix minor ISO namespace violation in Alpha varargs/stdarg 21257 support. 21258 + Fix incorrect "braf" instruction usage for the SH port. 21259 + Fix minor bug in va-sh which prevented its use with -ansi. 21260 + Fix problems recognizing and supporting FreeBSD. 21261 + Handle OpenBSD systems correctly. 21262 + Minor fixincludes fix for Digital UNIX 4.0B. 21263 + Fix problems with ctors/dtors in SCO shared libraries. 21264 + Abort instead of generating incorrect code for PPro/PII 21265 floating point conditional moves. 21266 + Avoid multiply defined symbols on GNU/Linux systems using 21267 libc-5.4.xx. 21268 + Fix abort in alpha compiler. 21269 * Fortran-specific fixes 21270 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 21271 is in the documented, non-Y2K-compliant range of 0-99, instead 21272 of being returned as 100 in the year 2000. 21273 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 21274 milliseconds value properly in Values(8). 21275 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 21276 information properly in SArray(7). 21277 21278 Each release includes installation instructions in both HTML and 21279 plaintext forms (see the INSTALL directory in the toplevel directory of 21280 the distribution). However, we also keep the most up to date 21281 installation instructions and [6]build/test status on our web page. We 21282 will update those pages as new information becomes available. 21283 21284 The EGCS project would like to thank the numerous people that have 21285 contributed new features, test results, bugfixes, etc. This [7]amazing 21286 group of volunteers is what makes EGCS successful. 21287 21288 And finally, we can't in good conscience fail to mention some 21289 [8]caveats to using EGCS 1.1. 21290 21291 Download EGCS from egcs.cygnus.com (USA California). 21292 21293 The EGCS 1.1 release is also available on many mirror sites. 21294 [9]Goto mirror list to find a closer site. 21295 21296 21297 For questions related to the use of GCC, please consult these web 21298 pages and the [10]GCC manuals. If that fails, the 21299 [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these 21300 web pages and the development of GCC are welcome on our developer 21301 list at [12]gcc@gcc.gnu.org. All of [13]our lists have public 21302 archives. 21303 21304 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 21305 distribution of this entire article is permitted in any medium, 21306 provided this notice is preserved. 21307 21308 These pages are [15]maintained by the GCC team. Last modified 21309 2021-07-28[16]. 21310 21311References 21312 21313 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 21314 2. http://gcc.gnu.org/news/gcse.html 21315 3. http://gcc.gnu.org/news/alias.html 21316 4. http://gcc.gnu.org/egcs-1.1/c++features.html 21317 5. http://gcc.gnu.org/egcs-1.1/features.html 21318 6. http://gcc.gnu.org/egcs-1.1/buildstat.html 21319 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 21320 8. http://gcc.gnu.org/egcs-1.1/caveats.html 21321 9. http://gcc.gnu.org/mirrors.html 21322 10. https://gcc.gnu.org/onlinedocs/ 21323 11. mailto:gcc-help@gcc.gnu.org 21324 12. mailto:gcc@gcc.gnu.org 21325 13. https://gcc.gnu.org/lists.html 21326 14. https://www.fsf.org/ 21327 15. https://gcc.gnu.org/about.html 21328 16. http://validator.w3.org/check/referer 21329====================================================================== 21330http://gcc.gnu.org/egcs-1.1/features.html 21331 21332 EGCS 1.1 new features 21333 21334 * Integrated GNU Fortran (g77) compiler and runtime library with 21335 improvements, based on g77 version 0.5.23. 21336 * Vast improvements in the C++ compiler; so many they have [1]page of 21337 their own! 21338 * Compiler implements [2]global common subexpression elimination and 21339 global copy/constant propagation. 21340 * More major improvements in the [3]alias analysis code. 21341 * More major improvements in the exception handling code to improve 21342 performance, lower static overhead and provide the infrastructure 21343 for future improvements. 21344 * The infamous /tmp symlink race security problems have been fixed. 21345 * The regmove optimization pass has been nearly completely rewritten 21346 to improve performance of generated code. 21347 * The compiler now recomputes register usage information before local 21348 register allocation. By providing more accurate information to the 21349 priority based allocator, we get better register allocation. 21350 * The register reloading phase of the compiler optimizes spill code 21351 much better than in previous releases. 21352 * Some bad interactions between the register allocator and 21353 instruction scheduler have been fixed, resulting in much better 21354 code for certain programs. Additionally, we have tuned the 21355 scheduler in various ways to improve performance of generated code 21356 for some architectures. 21357 * The compiler's branch shortening algorithms have been significantly 21358 improved to work better on targets which align jump targets. 21359 * The compiler now supports -Os to prefer optimizing for code space 21360 over optimizing for code speed. 21361 * The compiler will now totally eliminate library calls which compute 21362 constant values. This primarily helps targets with no integer 21363 div/mul support and targets without floating point support. 21364 * The compiler now supports an extensive "--help" option. 21365 * cpplib has been greatly improved and may be suitable for limited 21366 use. 21367 * Memory footprint for the compiler has been significantly reduced 21368 for some pathological cases. 21369 * The time to build EGCS has been improved for certain targets 21370 (particularly the alpha and mips platforms). 21371 * Many infrastructure improvements throughout the compiler, plus the 21372 usual mountain of bugfixes and minor improvements. 21373 * Target dependent improvements: 21374 + SPARC port now includes V8 plus and V9 support as well as 21375 performance tuning for Ultra class machines. The SPARC port 21376 now uses the Haifa scheduler. 21377 + Alpha port has been tuned for the EV6 processor and has an 21378 optimized expansion of memcpy/bzero. The Alpha port now uses 21379 the Haifa scheduler. 21380 + RS6000/PowerPC: support for the Power64 architecture and AIX 21381 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 21382 + x86: Alignment of static store data and jump targets is per 21383 Intel recommendations now. Various improvements throughout the 21384 x86 port to improve performance on Pentium processors 21385 (including improved epilogue sequences for Pentium chips and 21386 backend improvements which should help register allocation on 21387 all x86 variants. Conditional move support has been fixed and 21388 enabled for PPro processors. The x86 port also better supports 21389 64bit operations now. Unixware 7, a System V Release 5 target, 21390 is now supported and SCO OpenServer targets can support GAS. 21391 + MIPS has improved multiply/multiply-add support and now 21392 includes mips16 ISA support. 21393 + M68k has many micro-optimizations and Coldfire fixes. 21394 * Core compiler is based on the GCC development tree from June 9, 21395 1998, so we have all of the [4]features found in GCC 2.8. 21396 21397 21398 For questions related to the use of GCC, please consult these web 21399 pages and the [5]GCC manuals. If that fails, the 21400 [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these 21401 web pages and the development of GCC are welcome on our developer 21402 list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives. 21403 21404 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 21405 distribution of this entire article is permitted in any medium, 21406 provided this notice is preserved. 21407 21408 These pages are [10]maintained by the GCC team. Last modified 21409 2021-07-28[11]. 21410 21411References 21412 21413 1. http://gcc.gnu.org/egcs-1.1/c++features.html 21414 2. http://gcc.gnu.org/news/gcse.html 21415 3. http://gcc.gnu.org/news/alias.html 21416 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html 21417 5. https://gcc.gnu.org/onlinedocs/ 21418 6. mailto:gcc-help@gcc.gnu.org 21419 7. mailto:gcc@gcc.gnu.org 21420 8. https://gcc.gnu.org/lists.html 21421 9. https://www.fsf.org/ 21422 10. https://gcc.gnu.org/about.html 21423 11. http://validator.w3.org/check/referer 21424====================================================================== 21425http://gcc.gnu.org/egcs-1.1/caveats.html 21426 21427 EGCS 1.1 Caveats 21428 21429 * EGCS has an integrated libstdc++, but does not have an integrated 21430 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 21431 Lu has made a libg++-2.8.1.2 snapshot available which may work with 21432 EGCS. 21433 Note most C++ programs only need libstdc++. 21434 * Exception handling may not work with shared libraries, particularly 21435 on alphas, hppas, rs6000/powerpc and mips based platforms. 21436 Exception handling is known to work on x86-linux platforms with 21437 shared libraries. 21438 * Some versions of the Linux kernel have bugs which prevent them from 21439 being compiled or from running when compiled by EGCS. See the FAQ 21440 (as shipped with EGCS 1.1) for additional information. 21441 * In general, EGCS is more rigorous about rejecting invalid C++ code 21442 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 21443 a result it may be necessary to fix C++ code before it will compile 21444 with EGCS. 21445 * G++ is also converting toward the ISO C++ standard; as a result 21446 code which was previously valid (and thus accepted by other 21447 compilers and older versions of g++) may no longer be accepted. 21448 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 21449 or GCC 2.8.x due to changes necessary to support thread safe 21450 exception handling. 21451 21452 21453 For questions related to the use of GCC, please consult these web 21454 pages and the [1]GCC manuals. If that fails, the 21455 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these 21456 web pages and the development of GCC are welcome on our developer 21457 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives. 21458 21459 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 21460 distribution of this entire article is permitted in any medium, 21461 provided this notice is preserved. 21462 21463 These pages are [6]maintained by the GCC team. Last modified 21464 2021-07-28[7]. 21465 21466References 21467 21468 1. https://gcc.gnu.org/onlinedocs/ 21469 2. mailto:gcc-help@gcc.gnu.org 21470 3. mailto:gcc@gcc.gnu.org 21471 4. https://gcc.gnu.org/lists.html 21472 5. https://www.fsf.org/ 21473 6. https://gcc.gnu.org/about.html 21474 7. http://validator.w3.org/check/referer 21475====================================================================== 21476http://gcc.gnu.org/egcs-1.0/index.html 21477 21478 EGCS 1.0 21479 21480 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 21481 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 21482 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 21483 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 21484 21485 EGCS is a collaborative effort involving several groups of hackers 21486 using an open development model to accelerate development and testing 21487 of GNU compilers and runtime libraries. 21488 21489 An important goal of EGCS is to allow wide scale testing of 21490 experimental features and optimizations; therefore, EGCS contains some 21491 features and optimizations which are still under development. However, 21492 EGCS has been carefully tested and should be comparable in quality to 21493 most GCC releases. 21494 21495 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 21496 development sources; it contains nearly all of the new features found 21497 in GCC 2.8. 21498 21499 EGCS 1.0 also contains many improvements and features not found in GCC 21500 2.7 and even the GCC 2.8 series (which was released after the original 21501 EGCS 1.0 release). 21502 * Integrated C++ runtime libraries, including support for most major 21503 GNU/Linux systems! 21504 * The integrated libstdc++ library includes a verbatim copy of SGI's 21505 STL release. 21506 * Integrated GNU Fortran compiler. 21507 * New instruction scheduler. 21508 * New alias analysis code. 21509 21510 See the [1]new features page for a more complete list of new features. 21511 21512 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 21513 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 21514 EGCS 1.0 release: 21515 * Add support for Red Hat 5.0 Linux and better support for Linux 21516 systems using glibc2. 21517 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 21518 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 21519 fix these problems. 21520 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 21521 handling interfaces. 21522 To avoid future compatibility problems, we strongly urge anyone who 21523 is planning on distributing shared libraries that contain C++ code 21524 to upgrade to EGCS 1.0.1 first. 21525 Soon after EGCS 1.0 was released, the GCC developers made some 21526 incompatible changes in libgcc's exception handling interfaces. 21527 These changes were needed to solve problems on some platforms. This 21528 means that GCC 2.8.0, when released, will not be seamlessly 21529 compatible with shared libraries built by EGCS 1.0. The reason is 21530 that the libgcc.a in GCC 2.8.0 will not contain a function needed 21531 by the old interface. 21532 The result of this is that there may be compatibility problems with 21533 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 21534 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 21535 and libgcc.a has the support routines for both the old and the new 21536 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 21537 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 21538 The maintainers of GCC 2.x have decided against including seamless 21539 support for the old interface in 2.8.0, since it was never 21540 "official", so to avoid future compatibility problems we recommend 21541 against distributing any shared libraries built by EGCS 1.0 that 21542 contain C++ code (upgrade to 1.0.1 and use that). 21543 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 21544 The x86 changes fix code generation errors exposed when building 21545 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 21546 The hppa change fixes a compiler abort when configured for use with 21547 RTEMS. 21548 The MIPS changes fix problems with the definition of LONG_MAX on 21549 newer systems, allow for command line selection of the target ABI, 21550 and fix one code generation problem. 21551 The rs6000/ppc change fixes some problems with passing structures 21552 to varargs/stdarg functions. 21553 * A few machine independent bugfixes, mostly to fix code generation 21554 errors when building Linux kernels or glibc. 21555 * Fix a few critical exception handling and template bugs in the C++ 21556 compiler. 21557 * Fix Fortran namelist bug on alphas. 21558 * Fix build problems on x86-solaris systems. 21559 21560 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 21561 serious problems in EGCS 1.0.1. 21562 * General improvements and fixes 21563 + Memory consumption significantly reduced, especially for 21564 templates and inline functions. 21565 + Fix various problems with glibc2.1. 21566 + Fix loop optimization bug exposed by rs6000/ppc port. 21567 + Fix to avoid potential code generation problems in jump.c. 21568 + Fix some undefined symbol problems in dwarf1 debug support. 21569 * g++/libstdc++ improvements and fixes 21570 + libstdc++ in the EGCS release has been updated and should be 21571 link compatible with libstdc++-2.8. 21572 + Various fixes in libio/libstdc++ to work better on GNU/Linux 21573 systems. 21574 + Fix problems with duplicate symbols on systems that do not 21575 support weak symbols. 21576 + Memory corruption bug and undefined symbols in bastring have 21577 been fixed. 21578 + Various exception handling fixes. 21579 + Fix compiler abort for very long thunk names. 21580 * g77 improvements and fixes 21581 + Fix compiler crash for omitted bound in Fortran CASE 21582 statement. 21583 + Add missing entries to g77 lang-options. 21584 + Fix problem with -fpedantic in the g77 compiler. 21585 + Fix "backspace" problem with g77 on alphas. 21586 + Fix x86 backend problem with Fortran literals and -fpic. 21587 + Fix some of the problems with negative subscripts for g77 on 21588 alphas. 21589 + Fixes for Fortran builds on cygwin32/mingw32. 21590 * platform specific improvements and fixes 21591 + Fix long double problems on x86 (exposed by glibc). 21592 + x86 ports define i386 again to keep imake happy. 21593 + Fix exception handling support on NetBSD ports. 21594 + Several changes to collect2 to fix many problems with AIX. 21595 + Define __ELF__ for GNU/Linux on rs6000. 21596 + Fix -mcall-linux problem on GNU/Linux on rs6000. 21597 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 21598 + Allow autoconf to select a proper install problem on AIX 3.1. 21599 + m68k port support includes -mcpu32 option as well as cpu32 21600 multilibs. 21601 + Fix stdarg bug for irix6. 21602 + Allow EGCS to build on irix5 without the gnu assembler. 21603 + Fix problem with static linking on sco5. 21604 + Fix bootstrap on sco5 with native compiler. 21605 + Fix for abort building newlib on H8 target. 21606 + Fix fixincludes handling of math.h on SunOS. 21607 + Minor fix for Motorola 3300 m68k systems. 21608 21609 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 21610 problems reported by Red Hat for builds of Red Hat 5.1. 21611 * Generic bugfixes: 21612 + Fix a typo in the libio library which resulted in incorrect 21613 behavior of istream::get. 21614 + Fix the Fortran negative array index problem. 21615 + Fix a major problem with the ObjC runtime thread support 21616 exposed by glibc2. 21617 + Reduce memory consumption of the Haifa scheduler. 21618 * Target specific bugfixes: 21619 + Fix one x86 floating point code generation bug exposed by 21620 glibc2 builds. 21621 + Fix one x86 internal compiler error exposed by glibc2 builds. 21622 + Fix profiling bugs on the Alpha. 21623 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 21624 + Fix rs6000/ppc bug when converting values from integer types 21625 to floating point types. 21626 21627 The EGCS 1.0 releases include installation instructions in both HTML 21628 and plaintext forms (see the INSTALL directory in the toplevel 21629 directory of the distribution). However, we also keep the most up to 21630 date installation instructions and [2]build/test status on our web 21631 page. We will update those pages as new information becomes available. 21632 21633 And, we can't in good conscience fail to mention some [3]caveats to 21634 using EGCS. 21635 21636 Update: Big thanks to Stanford for providing a high speed link for 21637 downloading EGCS (go.cygnus.com)! 21638 21639 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 21640 (USA California -- High speed link provided by Stanford). 21641 21642 The EGCS 1.0 release is also available many mirror sites. 21643 [4]Goto mirror list to find a closer site 21644 21645 We'd like to thank the numerous people that have contributed new 21646 features, test results, bugfixes, etc. Unfortunately, they're far too 21647 numerous to mention by name. 21648 21649 21650 For questions related to the use of GCC, please consult these web 21651 pages and the [5]GCC manuals. If that fails, the 21652 [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these 21653 web pages and the development of GCC are welcome on our developer 21654 list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives. 21655 21656 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 21657 distribution of this entire article is permitted in any medium, 21658 provided this notice is preserved. 21659 21660 These pages are [10]maintained by the GCC team. Last modified 21661 2021-07-28[11]. 21662 21663References 21664 21665 1. http://gcc.gnu.org/egcs-1.0/features.html 21666 2. http://gcc.gnu.org/egcs-1.0/buildstat.html 21667 3. http://gcc.gnu.org/egcs-1.0/caveats.html 21668 4. http://gcc.gnu.org/mirrors.html 21669 5. https://gcc.gnu.org/onlinedocs/ 21670 6. mailto:gcc-help@gcc.gnu.org 21671 7. mailto:gcc@gcc.gnu.org 21672 8. https://gcc.gnu.org/lists.html 21673 9. https://www.fsf.org/ 21674 10. https://gcc.gnu.org/about.html 21675 11. http://validator.w3.org/check/referer 21676====================================================================== 21677http://gcc.gnu.org/egcs-1.0/features.html 21678 21679 EGCS 1.0 features 21680 21681 * Core compiler is based on the gcc2 development tree from Aug 2, 21682 1997, so we have most of the [1]features found in GCC 2.8. 21683 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 21684 * Vast improvements in the C++ compiler; so many they have [2]page of 21685 their own! 21686 * Integrated C++ runtime libraries, including support for most major 21687 GNU/Linux systems! 21688 * New instruction scheduler from IBM Haifa which includes support for 21689 function wide instruction scheduling as well as superscalar 21690 scheduling. 21691 * Significantly improved alias analysis code. 21692 * Improved register allocation for two address machines. 21693 * Significant code generation improvements for Fortran code on 21694 Alphas. 21695 * Various optimizations from the g77 project as well as improved loop 21696 optimizations. 21697 * Dwarf2 debug format support for some targets. 21698 * egcs libstdc++ includes the SGI STL implementation without changes. 21699 * As a result of these and other changes, egcs libstc++ is not binary 21700 compatible with previous releases of libstdc++. 21701 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 21702 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 21703 1.1), Support for RTEMS on several embedded targets, Support for 21704 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 21705 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 21706 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 21707 * RS6000/PowerPC ports generate code which can run on all 21708 RS6000/PowerPC variants by default. 21709 * -mcpu= and -march= switches for the x86 port to allow better 21710 control over how the x86 port generates code. 21711 * Includes the template repository patch (aka repo patch); note the 21712 new template code makes repo obsolete for ELF systems using gnu-ld 21713 such as GNU/Linux. 21714 * Plus the usual assortment of bugfixes and improvements. 21715 21716 21717 For questions related to the use of GCC, please consult these web 21718 pages and the [3]GCC manuals. If that fails, the 21719 [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these 21720 web pages and the development of GCC are welcome on our developer 21721 list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives. 21722 21723 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 21724 distribution of this entire article is permitted in any medium, 21725 provided this notice is preserved. 21726 21727 These pages are [8]maintained by the GCC team. Last modified 21728 2021-07-28[9]. 21729 21730References 21731 21732 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html 21733 2. http://gcc.gnu.org/egcs-1.0/c++features.html 21734 3. https://gcc.gnu.org/onlinedocs/ 21735 4. mailto:gcc-help@gcc.gnu.org 21736 5. mailto:gcc@gcc.gnu.org 21737 6. https://gcc.gnu.org/lists.html 21738 7. https://www.fsf.org/ 21739 8. https://gcc.gnu.org/about.html 21740 9. http://validator.w3.org/check/referer 21741====================================================================== 21742http://gcc.gnu.org/egcs-1.0/caveats.html 21743 21744 EGCS 1.0 Caveats 21745 21746 * EGCS has an integrated libstdc++, but does not have an integrated 21747 libg++. Furthermore old libg++ releases will not work with egc; HJ 21748 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 21749 Note most C++ programs only need libstdc++. 21750 * Note that using -pedantic or -Wreturn-type can cause an explosion 21751 in the amount of memory needed for template-heavy C++ code, such as 21752 code that uses STL. Also note that -Wall includes -Wreturn-type, so 21753 if you use -Wall you will need to specify -Wno-return-type to turn 21754 it off. 21755 * Exception handling may not work with shared libraries, particularly 21756 on alphas, hppas, and mips based platforms. Exception handling is 21757 known to work on x86-linux platforms with shared libraries. 21758 * Some versions of the Linux kernel have bugs which prevent them from 21759 being compiled or from running when compiled by EGCS. See the FAQ 21760 (as shipped with EGCS 1.0) for additional information. 21761 * In general, EGCS is more rigorous about rejecting invalid C++ code 21762 or deprecated C++ constructs than G++ 2.7. As a result it may be 21763 necessary to fix C++ code before it will compile with EGCS. 21764 * G++ is also aggressively tracking the C++ standard; as a result 21765 code which was previously valid (and thus accepted by other 21766 compilers and older versions of G++) may no longer be accepted. 21767 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 21768 1.0.x and later releases should work with Red Hat Linux 5.0. 21769 21770 21771 For questions related to the use of GCC, please consult these web 21772 pages and the [1]GCC manuals. If that fails, the 21773 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these 21774 web pages and the development of GCC are welcome on our developer 21775 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives. 21776 21777 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 21778 distribution of this entire article is permitted in any medium, 21779 provided this notice is preserved. 21780 21781 These pages are [6]maintained by the GCC team. Last modified 21782 2021-07-28[7]. 21783 21784References 21785 21786 1. https://gcc.gnu.org/onlinedocs/ 21787 2. mailto:gcc-help@gcc.gnu.org 21788 3. mailto:gcc@gcc.gnu.org 21789 4. https://gcc.gnu.org/lists.html 21790 5. https://www.fsf.org/ 21791 6. https://gcc.gnu.org/about.html 21792 7. http://validator.w3.org/check/referer 21793====================================================================== 21794