1Changes for 2.3.23
2    Fixed underlinking of libqdmod and libqd_f_main (bug).
3    Don't set CC=$CXX (fixes a new build issue with autoconf-2.71, bug).
4    Fixed typo in the rule to build docs/qd.pdf (bug).
5    Fixed "make dist" rule (built headers cannot be dist'd, bug).
6    Fixed the build with --disable-inline (bug).
7    Don't add "-O2" to CXXFLAGS or FCFLAGS when they're empty (portability).
8    Don't manually search for compilers & flags in configure.ac (portability).
9    Removed the --enable-warnings and --enable-debug flags (portability).
10    Modernized all of configure.ac for use with autoconf-2.71 (housekeeping).
11
12Changes for 2.3.22
13    Made changes suggested by Vasiliy Sotnikov
14
15Changes for 2.3.21
16    Changed renorm in include/qd/qd_inline.h
17
18Changes for 2.3.20
19    added #include <string.h> to quadt_test.cpp
20    changed references to 2.3.20 from 2.3.18
21
22Changes for 2.3.19
23  - Updated qd_real.cpp and dd_real.cpp to fix a buffer overflow problem.
24
25Changes for 2.3.18
26  - Updated qd_real.cpp and dd_real.cpp to fix a problem in output.
27
28Changes for 2.3.17
29  - updated qd_real.cpp, to fix a problem with improper treatment of
30  negative arguments in nroot.
31
32Changes for 2.3.16
33  - Updated dd_real.cpp, to fix a problem with inaccurate values of
34  tanh for small arguments.
35
36Changes for 2.3.15
37  - Updated qd_real.cpp, to fix a problem with static definitions.
38
39Changes for 2.3.14
40  - Updated autoconfig (replaced config.sub and config.guess)
41
42Changes for 2.3.7
43  - Fixed bug in to_digits where digits larger than 10
44    where output occasionally.
45
46Changes for 2.3.6
47  - Added fmod (C++) and mod (Fortran) functions.
48
49Changes for 2.3.5
50  - Fixed bug in division of qd_real by dd_real.
51  - Fixed bug in ddoutc (Fortran ddmod.f).
52  - Now compiles with g++ 4.3.
53  - Distribute tests/coeff.dat.
54
55Changes for 2.3.4
56  - Fixed bug in Makefile for cygwin / mingw systems.
57
58Changes for 2.3.3
59  - Fixed bug in atan2.
60
61Changes for 2.3.2
62  - Fixed bug in sin / cos / sincos where too much accuracy was
63    lost for (moderately) large angles.
64  - Use fused-multiply add intrinsics on IA-64 platforms if
65    compiled by Intel compiler.
66  - Fixed bug in c_dd_write and c_qd_write.
67  - Fixed bug were qdext.mod was not being installed.
68
69Changes for 2.3.1
70  - Fixed bug in sincos and cos_taylor.  This affected the result
71    of trigonometric functions in some cases.
72
73Changes for 2.3.0
74  This is a fairly significant change, breaking API compatibility.
75  - Moved C++ main entry in libqdmod.a to libqd_f_main.a.
76    This allows to link Fortran code using QD with custom
77    C++ main function.  Pure Fortran code will need to be linked
78    with qd_f_main library in addition to qdmod and qd library.
79  - Constructors accepting pointers made explicit.
80  - Fortran routines labeled as elemental or pure, where appropriate.
81  - Write() is now to_string(), and now takes a single fmtflag.
82  - dd_real addition and multiplication made commutative.
83  - dd_real now represented as array of two doubles, instead of
84    two discrete scalars.
85  - New Fortran generic routines to read / write, operations with
86    complex and integers.
87  - Improved exp, sin, and cos functions.
88  - Removed unused constants and obscure constants only used internally
89    from public interface.
90
91Changes for 2.2.6
92  - Fixed bug in mixed precision multiplication: qd_real * dd_real.
93
94Changes for 2.2.5
95  - Bug fix in qd_real addition when --enable-ieee-add is specified.
96  - Debugging routines dump and dump_bits updated;
97    dump_components removed (just use dump).
98  - Fortran support for Fortran strings.  Use character arrays instead.
99  - Return NaN under error conditions.
100  - Added _inf constant; exp now returns Inf when argument is too large.
101  - Output formatting fixes for Inf and NaNs.
102  - Added more real-complex mixed arithmetic routines in Fortran
103    interface.
104
105Changes for 2.2.4
106  - Added random_number interface for Fortran modules.
107  - Use slightly more conservative values for eps.
108  - Avoid unnecessary overflow near overflow threshold.
109  - Added radix, digits, min/maxexponent, range, and precision
110    intrinsics to Fortran interface.
111  - Added safe_max (C++) and safe_huge (Fortran).
112
113Changes for 2.2.3
114  - Fix sign function bug in Fortran modules.
115
116Changes for 2.2.2
117  - Do not bother setting uninitialized dd_real and qd_reals to zero.
118  - Use clock_gettime if available for timing.
119  - Fortran I/O should be more consistent with C++ version.
120  - fpu.h is now included with dd_real.h.
121
122Changes for 2.2.1
123  - Minor fixes when printing in scientific format.
124  - Change search order of C++ compilers in Apple systems to avoid
125    case insensitive filesystems.
126
127Changes for 2.2.0
128  - Added F95 interface for complex types.
129  - Renamed dd.h and qd.h to dd_real.h and qd_real.h, respectively.
130    This will break older C++ code using 2.1.x library, but it was
131    conflicting with QuickDraw libraries on Macs.  (Hence the version
132    bump to 2.2).
133  - Removed overloaded typecast operators for int and double.  These
134    permitted *automatic* conversion of dd_real/qd_real to double or
135    int, which is somewhat dangerous.  Instead to_int and to_double
136    routines are added.
137
138Changes for 2.1.214
139  - Updated pslq_test.
140  - Implmented numeric_limits<>.
141  - Better polyroot.
142  - Added isnan, isfinite, isinf functions.
143  - Fix / improve input output functions.
144  - Drop Microsoft Visual C++ 6.0 support.
145  - More efficient dd_real::sin.
146
147Changes for 2.1.213
148  - Support for x86_64 platforms.
149  - Drop libtool support for now.
150
151Changes for 2.1.212
152  - Support for pathCC compiler.
153  - Added accurate and sloppy versions of add / sub / mul / div avaialble.
154  - Added autodetection of fma functions.
155
156Changes for 2.1  (2003-12-30)
157  - added automake scripts.
158  - use libtool to compile / link and build libraries.
159  - supports standard installation targets (make install).
160  - support for Intel C++ compilers (icc / ecc).
161  - Fortran programs are now linked by C++ compiler.
162  - support for building shared library.
163  - minor bug fixes.
164
165Changes for 2.0  (2003-12-08)
166  - all header files are in "include/qd" directory.
167  - added autoconf scripts.
168  - added config.h and qd_config.h to store configuration information.
169  - renamed x86_* routines to fpu_* routines.
170  - added separate Fortran interface (f_* routines).
171  - options for sloppy multiply and sloppy divison separated.
172  - fixed C interface to be actually in C syntax.
173  - updated / added README, AUTHORS, NEWS, and LICENSE files.
174  - minor bug fixes.
175
176Changes for 1.2  (2003-12-04)
177  - added "dist-clean" target in Makefile
178  - initialize dd and qd variables to zero
179  - increases tolerance for qd / dd tests
180  - changed .cc extension to .cpp
181  - updated README, COPYING, and NEWS files
182  - added ChangeLog file
183  - fixed bug in '-all' flag in qd_test
184  - minor bug fixes
185
186Changes for 1.1  (2002-10-22)
187  - added "Changes" file (this file)
188  - fixed <strings.h> to <string.h>
189  - fixed constant (3/4) * pi
190  - fixed exp(x) to return zero if x is a large negative number
191  - removed "docs" target in Makefile
192
193