xref: /netbsd-src/external/gpl3/gcc/dist/NEWS (revision 63aea4bd5b445e491ff0389fe27ec78b3099dba3)
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-4.8/index.html
9
10                             GCC 4.8 Release Series
11
12   June 23, 2015
13
14   The [1]GNU project and the GCC developers are pleased to announce the
15   release of GCC 4.8.5.
16
17   This release is a bug-fix release, containing fixes for regressions in
18   GCC 4.8.4 relative to previous releases of GCC.
19
20Release History
21
22   GCC 4.8.5
23          June 23, 2015 ([2]changes, [3]documentation)
24
25   GCC 4.8.4
26          December 19, 2014 ([4]changes, [5]documentation)
27
28   GCC 4.8.3
29          May 22, 2014 ([6]changes, [7]documentation)
30
31   GCC 4.8.2
32          October 16, 2013 ([8]changes, [9]documentation)
33
34   GCC 4.8.1
35          May 31, 2013 ([10]changes, [11]documentation)
36
37   GCC 4.8.0
38          March 22, 2013 ([12]changes, [13]documentation)
39
40References and Acknowledgements
41
42   GCC used to stand for the GNU C Compiler, but since the compiler
43   supports several other languages aside from C, it now stands for the
44   GNU Compiler Collection.
45
46   A list of [14]successful builds is updated as new information becomes
47   available.
48
49   The GCC developers would like to thank the numerous people that have
50   contributed new features, improvements, bug fixes, and other changes as
51   well as test results to GCC. This [15]amazing group of volunteers is
52   what makes GCC successful.
53
54   For additional information about GCC please refer to the [16]GCC
55   project web site or contact the [17]GCC development mailing list.
56
57   To obtain GCC please use [18]our mirror sites or [19]our SVN server.
58
59
60    For questions related to the use of GCC, please consult these web
61    pages and the [20]GCC manuals. If that fails, the
62    [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these
63    web pages and the development of GCC are welcome on our developer
64    list at [22]gcc@gcc.gnu.org. All of [23]our lists have public
65    archives.
66
67   Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and
68   distribution of this entire article is permitted in any medium,
69   provided this notice is preserved.
70
71   These pages are [25]maintained by the GCC team. Last modified
72   2015-06-23[26].
73
74References
75
76   1. http://www.gnu.org/
77   2. http://gcc.gnu.org/gcc-4.8/changes.html
78   3. https://gcc.gnu.org/onlinedocs/4.8.5/
79   4. http://gcc.gnu.org/gcc-4.8/changes.html
80   5. https://gcc.gnu.org/onlinedocs/4.8.4/
81   6. http://gcc.gnu.org/gcc-4.8/changes.html
82   7. https://gcc.gnu.org/onlinedocs/4.8.3/
83   8. http://gcc.gnu.org/gcc-4.8/changes.html
84   9. https://gcc.gnu.org/onlinedocs/4.8.2/
85  10. http://gcc.gnu.org/gcc-4.8/changes.html
86  11. https://gcc.gnu.org/onlinedocs/4.8.1/
87  12. http://gcc.gnu.org/gcc-4.8/changes.html
88  13. https://gcc.gnu.org/onlinedocs/4.8.0/
89  14. http://gcc.gnu.org/gcc-4.8/buildstat.html
90  15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
91  16. http://gcc.gnu.org/index.html
92  17. mailto:gcc@gcc.gnu.org
93  18. http://gcc.gnu.org/mirrors.html
94  19. http://gcc.gnu.org/svn.html
95  20. https://gcc.gnu.org/onlinedocs/
96  21. mailto:gcc-help@gcc.gnu.org
97  22. mailto:gcc@gcc.gnu.org
98  23. https://gcc.gnu.org/lists.html
99  24. http://www.fsf.org/
100  25. https://gcc.gnu.org/about.html
101  26. http://validator.w3.org/check/referer
102======================================================================
103http://gcc.gnu.org/gcc-4.8/changes.html
104
105                             GCC 4.8 Release Series
106                        Changes, New Features, and Fixes
107
108Caveats
109
110   GCC now uses C++ as its implementation language. This means that to
111   build GCC from sources, you will need a C++ compiler that understands
112   C++ 2003. For more details on the rationale and specific changes,
113   please refer to the [1]C++ conversion page.
114
115   To enable the Graphite framework for loop optimizations you now need
116   CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from
117   the [2]GCC infrastructure directory. The installation manual contains
118   more information about requirements to build GCC.
119
120   GCC now uses a more aggressive analysis to derive an upper bound for
121   the number of iterations of loops using constraints imposed by language
122   standards. This may cause non-conforming programs to no longer work as
123   expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new
124   option, -fno-aggressive-loop-optimizations, was added to disable this
125   aggressive analysis. In some loops that have known constant number of
126   iterations, but undefined behavior is known to occur in the loop before
127   reaching or during the last iteration, GCC will warn about the
128   undefined behavior in the loop instead of deriving lower upper bound of
129   the number of iterations for the loop. The warning can be disabled with
130   -Wno-aggressive-loop-optimizations.
131
132   On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules
133   for the layout of vectors that could lead to wrong code being
134   generated. Vectors larger than 8 bytes in size are now by default
135   aligned to an 8-byte boundary. This is an ABI change: code that makes
136   explicit use of vector types may be incompatible with binary objects
137   built with older versions of GCC. Auto-vectorized code is not affected
138   by this change.
139
140   On AVR, support has been removed for the command-line option
141   -mshort-calls deprecated in GCC 4.7.
142
143   On AVR, the configure option --with-avrlibc supported since GCC 4.7.2
144   is turned on per default for all non-RTEMS configurations. This option
145   arranges for a better integration of [3]AVR Libc with avr-gcc. For
146   technical details, see [4]PR54461. To turn off the option in non-RTEMS
147   configurations, use --with-avrlibc=no. If the compiler is configured
148   for RTEMS, the option is always turned off.
149
150   More information on porting to GCC 4.8 from previous versions of GCC
151   can be found in the [5]porting guide for this release.
152
153General Optimizer Improvements (and Changes)
154
155     * DWARF4 is now the default when generating DWARF debug information.
156       When -g is used on a platform that uses DWARF debugging
157       information, GCC will now default to -gdwarf-4
158       -fno-debug-types-section.
159       GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
160       consumers support DWARF4 by default. Before GCC 4.8 the default
161       version used was DWARF2. To make GCC 4.8 generate an older DWARF
162       version use -g together with -gdwarf-2 or -gdwarf-3. The default
163       for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf.
164     * A new general optimization level, -Og, has been introduced. It
165       addresses the need for fast compilation and a superior debugging
166       experience while providing a reasonable level of run-time
167       performance. Overall experience for development should be better
168       than the default optimization level -O0.
169     * A new option -ftree-partial-pre was added to control the partial
170       redundancy elimination (PRE) optimization. This option is enabled
171       by default at the -O3 optimization level, and it makes PRE more
172       aggressive.
173     * The option -fconserve-space has been removed; it was no longer
174       useful on most targets since GCC supports putting variables into
175       BSS without making them common.
176     * The struct reorg and matrix reorg optimizations (command-line
177       options -fipa-struct-reorg and -fipa-matrix-reorg) have been
178       removed. They did not always work correctly, nor did they work with
179       link-time optimization (LTO), hence were only applicable to
180       programs consisting of a single translation unit.
181     * Several scalability bottle-necks have been removed from GCC's
182       optimization passes. Compilation of extremely large functions, e.g.
183       due to the use of the flatten attribute in the "Eigen" C++ linear
184       algebra templates library, is significantly faster than previous
185       releases of GCC.
186     * Link-time optimization (LTO) improvements:
187          + LTO partitioning has been rewritten for better reliability and
188            maintanibility. Several important bugs leading to link
189            failures have been fixed.
190     * Interprocedural optimization improvements:
191          + A new symbol table has been implemented. It builds on existing
192            callgraph and varpool modules and provide a new API. Unusual
193            symbol visibilities and aliases are handled more consistently
194            leading to, for example, more aggressive unreachable code
195            removal with LTO.
196          + The inline heuristic can now bypass limits on the size of of
197            inlined functions when the inlining is particularly
198            profitable. This happens, for example, when loop bounds or
199            array strides get propagated.
200          + Values passed through aggregates (either by value or
201            reference) are now propagated at the inter-procedural level
202            leading to better inlining decisions (for example in the case
203            of Fortran array descriptors) and devirtualization.
204     * [6]AddressSanitizer , a fast memory error detector, has been added
205       and can be enabled via -fsanitize=address. Memory access
206       instructions will be instrumented to detect heap-, stack-, and
207       global-buffer overflow as well as use-after-free bugs. To get nicer
208       stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is
209       available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on
210       x86-64 Darwin.
211     * [7]ThreadSanitizer has been added and can be enabled via
212       -fsanitize=thread. Instructions will be instrumented to detect data
213       races. The ThreadSanitizer is available on x86-64 GNU/Linux.
214     * A new local register allocator (LRA) has been implemented, which
215       replaces the 26 year old reload pass and improves generated code
216       quality. For now it is active on the IA-32 and x86-64 targets.
217     * Support for transactional memory has been implemented on the
218       following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and
219       Alpha.
220
221New Languages and Language specific improvements
222
223  C family
224
225     * Each diagnostic emitted now includes the original source line and a
226       caret '^' indicating the column. The option
227       -fno-diagnostics-show-caret suppresses this information.
228     * The option -ftrack-macro-expansion=2 is now enabled by default.
229       This allows the compiler to display the macro expansion stack in
230       diagnostics. Combined with the caret information, an example
231       diagnostic showing these two features is:
232
233t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float
234')
235 #define MYMAX(A,B)    __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _
236_b = (B); __a < __b ? __b : __a; })
237
238              ^
239t.c:7:7: note: in expansion of macro 'MYMAX'
240   X = MYMAX(P, F);
241       ^
242
243     * A new -Wsizeof-pointer-memaccess warning has been added (also
244       enabled by -Wall) to warn about suspicious length parameters to
245       certain string and memory built-in functions if the argument uses
246       sizeof. This warning warns e.g. about memset (ptr, 0, sizeof
247       (ptr)); if ptr is not an array, but a pointer, and suggests a
248       possible fix, or about memcpy (&foo, ptr, sizeof (&foo));.
249     * The new option -Wpedantic is an alias for -pedantic, which is now
250       deprecated. The forms -Wno-pedantic, -Werror=pedantic, and
251       -Wno-error=pedantic work in the same way as for any other -W
252       option. One caveat is that -Werror=pedantic is not equivalent to
253       -pedantic-errors, since the latter makes into errors some warnings
254       that are not controlled by -Wpedantic, and the former only affects
255       diagnostics that are disabled when using -Wno-pedantic.
256     * The option -Wshadow no longer warns if a declaration shadows a
257       function declaration, unless the former declares a function or
258       pointer to function, because this is [8]a common and valid case in
259       real-world code.
260
261  C++
262
263     * G++ now implements the [9]C++11 thread_local keyword; this differs
264       from the GNU __thread keyword primarily in that it allows dynamic
265       initialization and destruction semantics. Unfortunately, this
266       support requires a run-time penalty for references to
267       non-function-local thread_local variables defined in a different
268       translation unit even if they don't need dynamic initialization, so
269       users may want to continue to use __thread for TLS variables with
270       static initialization semantics.
271       If the programmer can be sure that no use of the variable in a
272       non-defining TU needs to trigger dynamic initialization (either
273       because the variable is statically initialized, or a use of the
274       variable in the defining TU will be executed before any uses in
275       another TU), they can avoid this overhead with the
276       -fno-extern-tls-init option.
277       OpenMP threadprivate variables now also support dynamic
278       initialization and destruction by the same mechanism.
279     * G++ now implements the [10]C++11 attribute syntax, e.g.
280
281[[noreturn]] void f();
282
283       and also the alignment specifier, e.g.
284
285alignas(double) int i;
286
287     * G++ now implements [11]C++11 inheriting constructors, e.g.
288
289struct A { A(int); };
290struct B: A { using A::A; }; // defines B::B(int)
291B b(42); // OK
292
293     * As of GCC 4.8.1, G++ implements the change to decltype semantics
294       from [12]N3276.
295
296struct A f();
297decltype(f()) g();    // OK, return type of f() is not required to be complete.
298
299     * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g.
300
301struct A { int f() &; };
302int i = A().f();  // error, f() requires an lvalue object
303
304     * G++ now supports a -std=c++1y option for experimentation with
305       features proposed for the next revision of the standard, expected
306       around 2014. Currently the only difference from -std=c++11 is
307       support for return type deduction in normal functions, as proposed
308       in [14]N3386. Status of C++1y features in GCC 4.8 can be found
309       [15]here.
310     * The G++ namespace association extension, __attribute ((strong)),
311       has been deprecated. Inline namespaces should be used instead.
312     * G++ now supports a -fext-numeric-literal option to control whether
313       GNU numeric literal suffixes are accepted as extensions or
314       processed as C++11 user-defined numeric literal suffixes. The flag
315       is on (use suffixes for GNU literals) by default for -std=gnu++*,
316       and -std=c++98. The flag is off (use suffixes for user-defined
317       literals) by default for -std=c++11 and later.
318
319    Runtime Library (libstdc++)
320
321     * [16]Improved experimental support for the new ISO C++ standard,
322       C++11, including:
323          + forward_list meets the allocator-aware container requirements;
324          + this_thread::sleep_for(), this_thread::sleep_until() and
325            this_thread::yield() are defined without requiring the
326            configure option --enable-libstdcxx-time;
327     * Improvements to <random>:
328          + SSE optimized normal_distribution.
329          + Use of hardware RNG instruction for random_device on new x86
330            processors (requires the assembler to support the
331            instruction.)
332       and <ext/random>:
333          + New random number engine simd_fast_mersenne_twister_engine
334            with an optimized SSE implementation.
335          + New random number distributions beta_distribution,
336            normal_mv_distribution, rice_distribution,
337            nakagami_distribution, pareto_distribution, k_distribution,
338            arcsine_distribution, hoyt_distribution.
339     * Added --disable-libstdcxx-verbose configure option to disable
340       diagnostic messages issued when a process terminates abnormally.
341       This may be useful for embedded systems to reduce the size of
342       executables that link statically to the library.
343
344  Fortran
345
346     * Compatibility notice:
347          + Module files: The version of module files (.mod) has been
348            incremented. Fortran MODULEs compiled by earlier GCC versions
349            have to be recompiled, when they are USEd by files compiled
350            with GCC 4.8. GCC 4.8 is not able to read .mod files created
351            by earlier versions; attempting to do so gives an error
352            message.
353            Note: The ABI of the produced assembler data itself has not
354            changed; object files and libraries are fully compatible with
355            older versions except as noted below.
356          + ABI: Some internal names (used in the assembler/object file)
357            have changed for symbols declared in the specification part of
358            a module. If an affected module - or a file using it via use
359            association - is recompiled, the module and all files which
360            directly use such symbols have to be recompiled as well. This
361            change only affects the following kind of module symbols:
362               o Procedure pointers. Note: C-interoperable function
363                 pointers (type(c_funptr)) are not affected nor are
364                 procedure-pointer components.
365               o Deferred-length character strings.
366     * The [17]BACKTRACE intrinsic subroutine has been added. It shows a
367       backtrace at an arbitrary place in user code; program execution
368       continues normally afterwards.
369     * The [18]-Wc-binding-type warning option has been added (disabled by
370       default). It warns if the a variable might not be C interoperable;
371       in particular, if the variable has been declared using an intrinsic
372       type with default kind instead of using a kind parameter defined
373       for C interoperability in the intrinsic ISO_C_Binding module.
374       Before, this warning was always printed. The -Wc-binding-type
375       option is enabled by -Wall.
376     * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line
377       options have been added, which diagnose when code is inserted for
378       automatic (re)allocation of a variable during assignment. This
379       option can be used to decide whether it is safe to use
380       [20]-fno-realloc-lhs. Additionally, it can be used to find
381       automatic (re)allocation in hot loops. (For arrays, replacing
382       "var=" by "var(:)=" disables the automatic reallocation.)
383     * The [21]-Wcompare-reals command-line option has been added. When
384       this is set, warnings are issued when comparing REAL or COMPLEX
385       types for equality and inequality; consider replacing a == b by
386       abs(a -b) < eps with a suitable eps. -Wcompare-reals is enabled by
387       -Wextra.
388     * The [22]-Wtarget-lifetime command-line option has been added
389       (enabled with -Wall), which warns if the pointer in a pointer
390       assignment might outlive its target.
391     * Reading floating point numbers which use "q" for the exponential
392       (such as 4.0q0) is now supported as vendor extension for better
393       compatibility with old data files. It is strongly recommended to
394       use for I/O the equivalent but standard conforming "e" (such as
395       4.0e0).
396       (For Fortran source code, consider replacing the "q" in
397       floating-point literals by a kind parameter (e.g. 4.0e0_qp with a
398       suitable qp). Note that - in Fortran source code - replacing "q" by
399       a simple "e" is not equivalent.)
400     * The GFORTRAN_TMPDIR environment variable for specifying a
401       non-default directory for files opened with STATUS="SCRATCH", is
402       not used anymore. Instead gfortran checks the POSIX/GNU standard
403       TMPDIR environment variable. If TMPDIR is not defined, gfortran
404       falls back to other methods to determine the directory for
405       temporary files as documented in the [23]user manual.
406     * [24]Fortran 2003:
407          + Support for unlimited polymorphic variables (CLASS(*)) has
408            been added. Nonconstant character lengths are not yet
409            supported.
410     * [25]TS 29113:
411          + Assumed types (TYPE(*)) are now supported.
412          + Experimental support for assumed-rank arrays (dimension(..))
413            has been added. Note that currently gfortran's own array
414            descriptor is used, which is different from the one defined in
415            TS29113, see [26]gfortran's header file or use the [27]Chasm
416            Language Interoperability Tools.
417
418  Go
419
420     * GCC 4.8.2 provides a complete implementation of the Go 1.1.2
421       release.
422     * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1
423       release. The library support is not quite complete.
424     * Go has been tested on GNU/Linux and Solaris platforms for various
425       processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may
426       work on other platforms as well.
427
428New Targets and Target Specific Improvements
429
430  AArch64
431
432     * A new port has been added to support AArch64, the new 64-bit
433       architecture from ARM. Note that this is a separate port from the
434       existing 32-bit ARM port.
435     * The port provides initial support for the Cortex-A53 and the
436       Cortex-A57 processors with the command line options
437       -mcpu=cortex-a53 and -mcpu=cortex-a57.
438     * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769
439       has been added and can be enabled by giving the
440       -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
441       default by configuring GCC with the --enable-fix-cortex-a53-835769
442       option.
443
444  ARM
445
446     * Initial support has been added for the AArch32 extensions defined
447       in the ARMv8 architecture.
448     * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs.
449     * A new option, -mcpu=marvell-pj4, has been added to generate code
450       for the Marvell PJ4 processor.
451     * The compiler can now automatically generate the VFMA, VFMS, REVSH
452       and REV16 instructions.
453     * A new vectorizer cost model for Advanced SIMD configurations to
454       improve the auto-vectorization strategies used.
455     * The scheduler now takes into account the number of live registers
456       to reduce the amount of spilling that can occur. This should
457       improve code performance in large functions. The limit can be
458       removed by using the option -fno-sched-pressure.
459     * Improvements have been made to the Marvell iWMMX code generation
460       and support for the iWMMX2 SIMD unit has been added. The option
461       -mcpu=iwmmxt2 can be used to enable code generation for the latter.
462     * A number of code generation improvements for Thumb2 to reduce code
463       size when compiling for the M-profile processors.
464     * The RTEMS (arm-rtems) port has been updated to use the EABI.
465     * Code generation support for the old FPA and Maverick floating-point
466       architectures has been removed. Ports that previously relied on
467       these features have also been removed. This includes the targets:
468          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
469          + arm*-*-elf (use arm*-*-eabi)
470          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
471          + arm*-*-ecos-elf (no alternative)
472          + arm*-*-freebsd (no alternative)
473          + arm*-wince-pe* (no alternative).
474
475  AVR
476
477     * Support for the "Embedded C" fixed-point has been added. For
478       details, see the [28]GCC wiki and the [29]user manual. The support
479       is not complete.
480     * A new print modifier %r for register operands in inline assembler
481       is supported. It will print the raw register number without the
482       register prefix 'r':
483    /* Return the most significant byte of 'val', a 64-bit value.  */
484
485    unsigned char msb (long long val)
486    {
487      unsigned char c;
488      __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val));
489      return c;
490    }
491       The inline assembler in this example will generate code like
492    mov r24, 8+7
493       provided c is allocated to R24 and val is allocated to R8...R15.
494       This works because the GNU assembler accepts plain register numbers
495       without register prefix.
496     * Static initializers with 3-byte symbols are supported now:
497    extern const __memx char foo;
498    const __memx void *pfoo = &foo;
499       This requires at least Binutils 2.23.
500
501  IA-32/x86-64
502
503     * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with
504       SSE extensions disabled. Since the x86-64 ABI requires 16 byte
505       stack alignment, this is ABI incompatible and intended to be used
506       in controlled environments where stack space is an important
507       limitation. This option will lead to wrong code when functions
508       compiled with 16 byte stack alignment (such as functions from a
509       standard library) are called with misaligned stack. In this case,
510       SSE instructions may lead to misaligned memory access traps. In
511       addition, variable arguments will be handled incorrectly for 16
512       byte aligned objects (including x87 long double and __int128),
513       leading to wrong results. You must build all modules with
514       -mpreferred-stack-boundary=3, including any libraries. This
515       includes the system libraries and startup modules.
516     * Support for the new Intel processor codename Broadwell with RDSEED,
517       ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw,
518       -mrdseed command-line options.
519     * Support for the Intel RTM and HLE intrinsics, built-in functions
520       and code generation is available via -mrtm and -mhle.
521     * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets.
522       Intrinsics and built-in functions are available via -mfxsr, -mxsave
523       and -mxsaveopt respectively.
524     * New -maddress-mode=[short|long] options for x32.
525       -maddress-mode=short overrides default 64-bit addresses to 32-bit
526       by emitting the 0x67 address-size override prefix. This is the
527       default address mode for x32.
528     * New built-in functions to detect run-time CPU type and ISA:
529          + A built-in function __builtin_cpu_is has been added to detect
530            if the run-time CPU is of a particular type. It returns a
531            positive integer on a match and zero otherwise. It accepts one
532            string literal argument, the CPU name. For example,
533            __builtin_cpu_is("westmere") returns a positive integer if the
534            run-time CPU is an Intel Core i7 Westmere processor. Please
535            refer to the [30]user manual for the list of valid CPU names
536            recognized.
537          + A built-in function __builtin_cpu_supports has been added to
538            detect if the run-time CPU supports a particular ISA feature.
539            It returns a positive integer on a match and zero otherwise.
540            It accepts one string literal argument, the ISA feature. For
541            example, __builtin_cpu_supports("ssse3") returns a positive
542            integer if the run-time CPU supports SSSE3 instructions.
543            Please refer to the [31]user manual for the list of valid ISA
544            names recognized.
545       Caveat: If these built-in functions are called before any static
546       constructors are invoked, like during IFUNC initialization, then
547       the CPU detection initialization must be explicitly run using this
548       newly provided built-in function, __builtin_cpu_init. The
549       initialization needs to be done only once. For example, this is how
550       the invocation would look like inside an IFUNC initializer:
551    static void (*some_ifunc_resolver(void))(void)
552    {
553      __builtin_cpu_init();
554      if (__builtin_cpu_is("amdfam10h") ...
555      if (__builtin_cpu_supports("popcnt") ...
556    }
557
558     * Function Multiversioning Support with G++:
559       It is now possible to create multiple function versions each
560       targeting a specific processor and/or ISA. Function versions have
561       the same signature but different target attributes. For example,
562       here is a program with function versions:
563    __attribute__ ((target ("default")))
564    int foo(void)
565    {
566      return 1;
567    }
568
569    __attribute__ ((target ("sse4.2")))
570    int foo(void)
571    {
572      return 2;
573    }
574
575    int main (void)
576    {
577      int (*p) = &foo;
578      assert ((*p)() == foo());
579      return 0;
580    }
581
582       Please refer to this [32]wiki for more information.
583     * The x86 back end has been improved to allow option -fschedule-insns
584       to work reliably. This option can be used to schedule instructions
585       better and leads to improved performace in certain cases.
586     * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437
587       from the Mingw-w64 trunk.
588     * Support for new AMD family 15h processors (Steamroller core) is now
589       available through the -march=bdver3 and -mtune=bdver3 options.
590     * Support for new AMD family 16h processors (Jaguar core) is now
591       available through the -march=btver2 and -mtune=btver2 options.
592
593  FRV
594
595     * This target now supports the -fstack-usage command-line option.
596
597  MIPS
598
599     * GCC can now generate code specifically for the R4700, Broadcom XLP
600       and MIPS 34kn processors. The associated -march options are
601       -march=r4700, -march=xlp and -march=34kn respectively.
602     * GCC now generates better DSP code for MIPS 74k cores thanks to
603       further scheduling optimizations.
604     * The MIPS port now supports the -fstack-check option.
605     * GCC now passes the -mmcu and -mno-mcu options to the assembler.
606     * Previous versions of GCC would silently accept -fpic and -fPIC for
607       -mno-abicalls targets like mips*-elf. This combination was not
608       intended or supported, and did not generate position-independent
609       code. GCC 4.8 now reports an error when this combination is used.
610
611  PowerPC / PowerPC64 / RS6000
612
613     * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save,
614       restore or update the VRSAVE register by default. The respective
615       operating systems manage the VRSAVE register directly.
616     * Large TOC support has been added for AIX through the command line
617       option -mcmodel=large.
618     * Native Thread-Local Storage support has been added for AIX.
619     * VMX (Altivec) and VSX instruction sets now are enabled implicitly
620       when targetting processors that support those hardware features on
621       AIX 6.1 and above.
622
623  RX
624
625     * This target will now issue a warning message whenever multiple fast
626       interrupt handlers are found in the same compilation unit. This
627       feature can be turned off by the new
628       -mno-warn-multiple-fast-interrupts command-line option.
629
630  S/390, System z
631
632     * Support for the IBM zEnterprise zEC12 processor has been added.
633       When using the -march=zEC12 option, the compiler will generate code
634       making use of the following new instructions:
635          + load and trap instructions
636          + 2 new compare and trap instructions
637          + rotate and insert selected bits - without CC clobber
638       The -mtune=zEC12 option enables zEC12 specific instruction
639       scheduling without making use of new instructions.
640     * Register pressure sensitive instruction scheduling is enabled by
641       default.
642     * The ifunc function attribute is enabled by default.
643     * memcpy and memcmp invokations on big memory chunks or with run time
644       lengths are not generated inline anymore when tuning for z10 or
645       higher. The purpose is to make use of the IFUNC optimized versions
646       in Glibc.
647
648  SH
649
650     * The default alignment settings have been reduced to be less
651       aggressive. This results in more compact code for optimization
652       levels other than -Os.
653     * Improved support for the __atomic built-in functions:
654          + A new option -matomic-model=model selects the model for the
655            generated atomic sequences. The following models are
656            supported:
657
658              soft-gusa
659                      Software gUSA sequences (SH3* and SH4* only). On
660                      SH4A targets this will now also partially utilize
661                      the movco.l and movli.l instructions. This is the
662                      default when the target is sh3*-*-linux* or
663                      sh4*-*-linux*.
664
665              hard-llcs
666                      Hardware movco.l / movli.l sequences (SH4A only).
667
668              soft-tcb
669                      Software thread control block sequences.
670
671              soft-imask
672                      Software interrupt flipping sequences (privileged
673                      mode only). This is the default when the target is
674                      sh1*-*-linux* or sh2*-*-linux*.
675
676              none
677                      Generates function calls to the respective __atomic
678                      built-in functions. This is the default for SH64
679                      targets or when the target is not sh*-*-linux*.
680
681          + The option -msoft-atomic has been deprecated. It is now an
682            alias for -matomic-model=soft-gusa.
683          + A new option -mtas makes the compiler generate the tas.b
684            instruction for the __atomic_test_and_set built-in function
685            regardless of the selected atomic model.
686          + The __sync functions in libgcc now reflect the selected atomic
687            model when building the toolchain.
688     * Added support for the mov.b and mov.w instructions with
689       displacement addressing.
690     * Added support for the SH2A instructions movu.b and movu.w.
691     * Various improvements to code generated for integer arithmetic.
692     * Improvements to conditional branches and code that involves the T
693       bit. A new option -mzdcbranch tells the compiler to favor
694       zero-displacement branches. This is enabled by default for SH4*
695       targets.
696     * The pref instruction will now be emitted by the __builtin_prefetch
697       built-in function for SH3* targets.
698     * The fmac instruction will now be emitted by the fmaf standard
699       function and the __builtin_fmaf built-in function.
700     * The -mfused-madd option has been deprecated in favor of the
701       machine-independent -ffp-contract option. Notice that the fmac
702       instruction will now be generated by default for expressions like a
703       * b + c. This is due to the compiler default setting
704       -ffp-contract=fast.
705     * Added new options -mfsrra and -mfsca to allow the compiler using
706       the fsrra and fsca instructions on targets other than SH4A (where
707       they are already enabled by default).
708     * Added support for the __builtin_bswap32 built-in function. It is
709       now expanded as a sequence of swap.b and swap.w instructions
710       instead of a library function call.
711     * The behavior of the -mieee option has been fixed and the negative
712       form -mno-ieee has been added to control the IEEE conformance of
713       floating point comparisons. By default -mieee is now enabled and
714       the option -ffinite-math-only implicitly sets -mno-ieee.
715     * Added support for the built-in functions __builtin_thread_pointer
716       and __builtin_set_thread_pointer. This assumes that GBR is used to
717       hold the thread pointer of the current thread. Memory loads and
718       stores relative to the address returned by __builtin_thread_pointer
719       will now also utilize GBR based displacement address modes.
720     * The -mdiv= option for targets other than SHmedia has been fixed and
721       documented.
722
723  SPARC
724
725     * Added optimized instruction scheduling for Niagara4.
726
727  TILE-Gx
728
729     * Added support for the -mcmodel=MODEL command-line option. The
730       models supported are small and large.
731
732  V850
733
734     * This target now supports the E3V5 architecture via the use of the
735       new -mv850e3v5 command-line option. It also has experimental
736       support for the e3v5 LOOP instruction which can be enabled via the
737       new -mloop command-line option.
738
739  XStormy16
740
741     * This target now supports the -fstack-usage command-line option.
742
743Operating Systems
744
745  Windows (Cygwin)
746
747     * Executables are now linked against shared libgcc by default. The
748       previous default was to link statically, which can still be done by
749       explicitly specifying -static or static-libgcc on the command line.
750       However it is strongly advised against, as it will cause problems
751       for any application that makes use of DLLs compiled by GCC. It
752       should be alright for a monolithic stand-alone application that
753       only links against the Windows DLLs, but offers little or no
754       benefit.
755
756GCC 4.8.1
757
758   This is the [33]list of problem reports (PRs) from GCC's bug tracking
759   system that are known to be fixed in the 4.8.1 release. This list might
760   not be complete (that is, it is possible that some PRs that have been
761   fixed are not listed here).
762
763   The C++11 <chrono> std::chrono::system_clock and
764   std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they
765   both are now separate (never typedefs of each other), both use
766   std::chrono::nanoseconds resolution, on most GNU/Linux configurations
767   std::chrono::steady_clock is now finally monotonic, and both classes
768   are mangled differently than in the previous GCC releases.
769   std::chrono::system_clock::now() with std::chrono::microseconds resp.
770   std::chrono::seconds resolution is still exported for backwards
771   compatibility with default configured libstdc++. Note that libstdc++
772   configured with --enable-libstdcxx-time= used to be ABI incompatible
773   with default configured libstdc++ for those two classes and no ABI
774   compatibility can be offered for those configurations, so any C++11
775   code that uses those classes and has been compiled and linked against
776   libstdc++ configured with the non-default --enable-libstdcxx-time=
777   configuration option needs to be recompiled.
778
779GCC 4.8.2
780
781   This is the [34]list of problem reports (PRs) from GCC's bug tracking
782   system that are known to be fixed in the 4.8.2 release. This list might
783   not be complete (that is, it is possible that some PRs that have been
784   fixed are not listed here).
785
786GCC 4.8.3
787
788   This is the [35]list of problem reports (PRs) from GCC's bug tracking
789   system that are known to be fixed in the 4.8.3 release. This list might
790   not be complete (that is, it is possible that some PRs that have been
791   fixed are not listed here).
792
793   Support for the new powerpc64le-linux platform has been added. It
794   defaults to generating code that conforms to the ELFV2 ABI.
795
796GCC 4.8.4
797
798   This is the [36]list of problem reports (PRs) from GCC's bug tracking
799   system that are known to be fixed in the 4.8.4 release. This list might
800   not be complete (that is, it is possible that some PRs that have been
801   fixed are not listed here).
802
803GCC 4.8.5
804
805   This is the [37]list of problem reports (PRs) from GCC's bug tracking
806   system that are known to be fixed in the 4.8.5 release. This list might
807   not be complete (that is, it is possible that some PRs that have been
808   fixed are not listed here).
809
810
811    For questions related to the use of GCC, please consult these web
812    pages and the [38]GCC manuals. If that fails, the
813    [39]gcc-help@gcc.gnu.org mailing list might help. Comments on these
814    web pages and the development of GCC are welcome on our developer
815    list at [40]gcc@gcc.gnu.org. All of [41]our lists have public
816    archives.
817
818   Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and
819   distribution of this entire article is permitted in any medium,
820   provided this notice is preserved.
821
822   These pages are [43]maintained by the GCC team. Last modified
823   2015-06-23[44].
824
825References
826
827   1. https://gcc.gnu.org/wiki/cxx-conversion
828   2. ftp://gcc.gnu.org/pub/gcc/infrastructure/
829   3. http://www.nongnu.org/avr-libc/
830   4. https://gcc.gnu.org/PR54461
831   5. https://gcc.gnu.org/gcc-4.8/porting_to.html
832   6. https://code.google.com/p/address-sanitizer/
833   7. https://code.google.com/p/data-race-test/wiki/ThreadSanitizer
834   8. https://lkml.org/lkml/2006/11/28/239
835   9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
836  10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
837  11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
838  12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf
839  13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
840  14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html
841  15. http://gcc.gnu.org/projects/cxx1y.html
842  16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011
843  17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html
844  18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
845  19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
846  20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
847  21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
848  22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
849  23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html
850  24. https://gcc.gnu.org/wiki/Fortran2003Status
851  25. https://gcc.gnu.org/wiki/TS29113Status
852  26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text%2Fplain&view=co
853  27. http://chasm-interop.sourceforge.net/
854  28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support
855  29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html
856  30. https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
857  31. https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
858  32. https://gcc.gnu.org/wiki/FunctionMultiVersioning
859  33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1
860  34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2
861  35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3
862  36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4
863  37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5
864  38. https://gcc.gnu.org/onlinedocs/
865  39. mailto:gcc-help@gcc.gnu.org
866  40. mailto:gcc@gcc.gnu.org
867  41. https://gcc.gnu.org/lists.html
868  42. http://www.fsf.org/
869  43. https://gcc.gnu.org/about.html
870  44. http://validator.w3.org/check/referer
871======================================================================
872http://gcc.gnu.org/gcc-4.7/index.html
873
874                             GCC 4.7 Release Series
875
876   June 12, 2014
877
878   The [1]GNU project and the GCC developers are pleased to announce the
879   release of GCC 4.7.4.
880
881   This release is a bug-fix release, containing fixes for regressions in
882   GCC 4.7.3 relative to previous releases of GCC.
883
884Release History
885
886   GCC 4.7.4
887          June 12, 2014 ([2]changes, [3]documentation)
888
889   GCC 4.7.3
890          April 11, 2013 ([4]changes, [5]documentation)
891
892   GCC 4.7.2
893          September 20, 2012 ([6]changes, [7]documentation)
894
895   GCC 4.7.1
896          June 14, 2012 ([8]changes, [9]documentation)
897
898   GCC 4.7.0
899          March 22, 2012 ([10]changes, [11]documentation)
900
901References and Acknowledgements
902
903   GCC used to stand for the GNU C Compiler, but since the compiler
904   supports several other languages aside from C, it now stands for the
905   GNU Compiler Collection.
906
907   A list of [12]successful builds is updated as new information becomes
908   available.
909
910   The GCC developers would like to thank the numerous people that have
911   contributed new features, improvements, bug fixes, and other changes as
912   well as test results to GCC. This [13]amazing group of volunteers is
913   what makes GCC successful.
914
915   For additional information about GCC please refer to the [14]GCC
916   project web site or contact the [15]GCC development mailing list.
917
918   To obtain GCC please use [16]our mirror sites or [17]our SVN server.
919
920
921    For questions related to the use of GCC, please consult these web
922    pages and the [18]GCC manuals. If that fails, the
923    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
924    web pages and the development of GCC are welcome on our developer
925    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
926    archives.
927
928   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
929   distribution of this entire article is permitted in any medium,
930   provided this notice is preserved.
931
932   These pages are [23]maintained by the GCC team. Last modified
933   2014-06-12[24].
934
935References
936
937   1. http://www.gnu.org/
938   2. http://gcc.gnu.org/gcc-4.7/changes.html
939   3. https://gcc.gnu.org/onlinedocs/4.7.4/
940   4. http://gcc.gnu.org/gcc-4.7/changes.html
941   5. https://gcc.gnu.org/onlinedocs/4.7.3/
942   6. http://gcc.gnu.org/gcc-4.7/changes.html
943   7. https://gcc.gnu.org/onlinedocs/4.7.2/
944   8. http://gcc.gnu.org/gcc-4.7/changes.html
945   9. https://gcc.gnu.org/onlinedocs/4.7.1/
946  10. http://gcc.gnu.org/gcc-4.7/changes.html
947  11. https://gcc.gnu.org/onlinedocs/4.7.0/
948  12. http://gcc.gnu.org/gcc-4.7/buildstat.html
949  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
950  14. http://gcc.gnu.org/index.html
951  15. mailto:gcc@gcc.gnu.org
952  16. http://gcc.gnu.org/mirrors.html
953  17. http://gcc.gnu.org/svn.html
954  18. https://gcc.gnu.org/onlinedocs/
955  19. mailto:gcc-help@gcc.gnu.org
956  20. mailto:gcc@gcc.gnu.org
957  21. https://gcc.gnu.org/lists.html
958  22. http://www.fsf.org/
959  23. https://gcc.gnu.org/about.html
960  24. http://validator.w3.org/check/referer
961======================================================================
962http://gcc.gnu.org/gcc-4.7/changes.html
963
964                             GCC 4.7 Release Series
965                        Changes, New Features, and Fixes
966
967Caveats
968
969     * The -fconserve-space flag has been deprecated. The flag had no
970       effect for most targets: only targets without a global .bss section
971       and without support for switchable sections. Furthermore, the flag
972       only had an effect for G++, where it could result in wrong
973       semantics (please refer to the GCC manual for further details). The
974       flag will be removed in GCC 4.8
975     * Support for a number of older systems and recently unmaintained or
976       untested target ports of GCC has been declared obsolete in GCC 4.7.
977       Unless there is activity to revive them, the next release of GCC
978       will have their sources permanently removed.
979       All GCC ports for the following processor architectures have been
980       declared obsolete:
981          + picoChip (picochip-*)
982       The following ports for individual systems on particular
983       architectures have been obsoleted:
984          + IRIX 6.5 (mips-sgi-irix6.5)
985          + MIPS OpenBSD (mips*-*-openbsd*)
986          + Solaris 8 (*-*-solaris2.8). Details can be found in the
987            [1]announcement.
988          + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*)
989     * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A,
990       ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by
991       default, which for some sources generates code that accesses memory
992       on unaligned addresses. This requires the kernel of those systems
993       to enable such accesses (controlled by CP15 register c1, refer to
994       ARM documentation). Alternatively, or for compatibility with
995       kernels where unaligned accesses are not supported, all code has to
996       be compiled with -mno-unaligned-access. Upstream Linux kernel
997       releases have automatically and unconditionally supported unaligned
998       accesses as emitted by GCC due to this option being active since
999       version 2.6.28.
1000     * Support on ARM for the legacy floating-point accelerator (FPA) and
1001       the mixed-endian floating-point format that it used has been
1002       obsoleted. The ports that still use this format have been obsoleted
1003       as well. Many legacy ARM ports already provide an alternative that
1004       uses the VFP floating-point format. The obsolete ports will be
1005       deleted in the next release.
1006       The obsolete ports with alternatives are:
1007          + arm*-*-rtems (use arm*-*-rtemseabi)
1008          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
1009          + arm*-*-elf (use arm*-*-eabi)
1010          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
1011       Note, however, that these alternatives are not binary compatible
1012       with their legacy counterparts (although some can support running
1013       legacy applications).
1014       The obsolete ports that currently lack a modern alternative are:
1015          + arm*-*-ecos-elf
1016          + arm*-*-freebsd
1017          + arm*-wince-pe*
1018       New ports that support more recent versions of the architecture are
1019       welcome.
1020     * Support for the Maverick co-processor on ARM has been obsoleted.
1021       Code to support it will be deleted in the next release.
1022     * Support has been removed for Unix International threads on Solaris
1023       2, so the --enable-threads=solaris configure option and the
1024       -threads compiler option don't work any longer.
1025     * Support has been removed for the Solaris BSD Compatibility Package,
1026       which lives in /usr/ucbinclude and /usr/ucblib. It has been removed
1027       from Solaris 11, and was only intended as a migration aid from
1028       SunOS 4 to SunOS 5. The -compat-bsd compiler option is not
1029       recognized any longer.
1030     * The AVR port's libgcc has been improved and its multilib structure
1031       has been enhanced. As a result, all objects contributing to an
1032       application must either be compiled with GCC versions up to 4.6.x
1033       or with GCC versions 4.7.1 or later. If the compiler is used with
1034       AVR Libc, you need a version that supports the new layout, i.e.
1035       implements [2]#35407.
1036     * The AVR port's -mshort-calls command-line option has been
1037       deprecated. It will be removed in the GCC 4.8 release. See -mrelax
1038       for a replacement.
1039     * The AVR port only references startup code that clears .bss and the
1040       common section resp. initializes the .data and .rodata section
1041       provided respective sections (or subsections thereof) are not
1042       empty, see [3]PR18145. Applications that put all static storage
1043       objects into non-standard sections and / or define all static
1044       storage objects in assembler modules, must reference __do_clear_bss
1045       resp. __do_copy_data by hand or undefine the symbol(s) by means of
1046       -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data.
1047     * The ARM port's -mwords-little-endian option has been deprecated. It
1048       will be removed in a future release.
1049     * Support has been removed for the NetWare x86 configuration
1050       obsoleted in GCC 4.6.
1051     * It is no longer possible to use the "l" constraint in MIPS16 asm
1052       statements.
1053     * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard
1054       library which affected the ABI in C++11 mode: a data member was
1055       added to std::list changing its size and altering the definitions
1056       of some member functions, and std::pair's move constructor was
1057       non-trivial which altered the calling convention for functions with
1058       std::pair arguments or return types. The ABI incompatibilities have
1059       been fixed for GCC version 4.7.2 but as a result C++11 code
1060       compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11
1061       code compiled with different GCC versions and with C++98/C++03 code
1062       compiled with any version.
1063     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
1064       rules for the layout of vectors that could lead to wrong code being
1065       generated. Vectors larger than 8 bytes in size are now by default
1066       aligned to an 8-byte boundary. This is an ABI change: code that
1067       makes explicit use of vector types may be incompatible with binary
1068       objects built with older versions of GCC. Auto-vectorized code is
1069       not affected by this change. (This change affects GCC versions
1070       4.7.2 and later.)
1071     * More information on porting to GCC 4.7 from previous versions of
1072       GCC can be found in the [4]porting guide for this release.
1073
1074General Optimizer Improvements
1075
1076     * Support for a new parameter --param case-values-threshold=n was
1077       added to allow users to control the cutoff between doing switch
1078       statements as a series of if statements and using a jump table.
1079     * Link-time optimization (LTO) improvements:
1080          + Improved scalability and reduced memory usage. Link time
1081            optimization of Firefox now requires 3GB of RAM on a 64-bit
1082            system, while over 8GB was needed previously. Linking time has
1083            been improved, too. The serial stage of linking Firefox has
1084            been sped up by about a factor of 10.
1085          + Reduced size of object files and temporary storage used during
1086            linking.
1087          + Streaming performance (both outbound and inbound) has been
1088            improved.
1089          + ld -r is now supported with LTO.
1090          + Several bug fixes, especially in symbol table handling and
1091            merging.
1092     * Interprocedural optimization improvements:
1093          + Heuristics now take into account that after inlining code will
1094            be optimized out because of known values (or properties) of
1095            function parameters. For example:
1096void foo(int a)
1097{
1098  if (a > 10)
1099    ... huge code ...
1100}
1101void bar (void)
1102{
1103  foo (0);
1104}
1105
1106            The call of foo will be inlined into bar even when optimizing
1107            for code size. Constructs based on __builtin_constant_p are
1108            now understood by the inliner and code size estimates are
1109            evaluated a lot more realistically.
1110          + The representation of C++ virtual thunks and aliases (both
1111            implicit and defined via the alias attribute) has been
1112            re-engineered. Aliases no longer pose optimization barriers
1113            and calls to an alias can be inlined and otherwise optimized.
1114          + The inter-procedural constant propagation pass has been
1115            rewritten. It now performs generic function specialization.
1116            For example when compiling the following:
1117void foo(bool flag)
1118{
1119  if (flag)
1120    ... do something ...
1121  else
1122    ... do something else ...
1123}
1124void bar (void)
1125{
1126  foo (false);
1127  foo (true);
1128  foo (false);
1129  foo (true);
1130  foo (false);
1131  foo (true);
1132}
1133
1134            GCC will now produce two copies of foo. One with flag being
1135            true, while other with flag being false. This leads to
1136            performance improvements previously possible only by inlining
1137            all calls. Cloning causes a lot less code size growth.
1138     * A string length optimization pass has been added. It attempts to
1139       track string lengths and optimize various standard C string
1140       functions like strlen, strchr, strcpy, strcat, stpcpy and their
1141       _FORTIFY_SOURCE counterparts into faster alternatives. This pass is
1142       enabled by default at -O2 or above, unless optimizing for size, and
1143       can be disabled by the -fno-optimize-strlen option. The pass can
1144       e.g. optimize
1145char *bar (const char *a)
1146{
1147  size_t l = strlen (a) + 2;
1148  char *p = malloc (l); if (p == NULL) return p;
1149  strcpy (p, a); strcat (p, "/"); return p;
1150}
1151
1152       into:
1153char *bar (const char *a)
1154{
1155  size_t tmp = strlen (a);
1156  char *p = malloc (tmp + 2); if (p == NULL) return p;
1157  memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p;
1158}
1159
1160       or for hosted compilations where stpcpy is available in the runtime
1161       and headers provide its prototype, e.g.
1162void foo (char *a, const char *b, const char *c, const char *d)
1163{
1164  strcpy (a, b); strcat (a, c); strcat (a, d);
1165}
1166
1167       can be optimized into:
1168void foo (char *a, const char *b, const char *c, const char *d)
1169{
1170  strcpy (stpcpy (stpcpy (a, b), c), d);
1171}
1172
1173New Languages and Language specific improvements
1174
1175     * Version 3.1 of the [5]OpenMP specification is now supported for the
1176       C, C++, and Fortran compilers.
1177
1178  Ada
1179
1180     * The command-line option -feliminate-unused-debug-types has been
1181       re-enabled by default, as it is for the other languages, leading to
1182       a reduction in debug info size of 12.5% and more for relevant
1183       cases, as well as to a small compilation speedup.
1184
1185  C family
1186
1187     * A new built-in, __builtin_assume_aligned, has been added, through
1188       which the compiler can be hinted about pointer alignment and can
1189       use it to improve generated code.
1190     * A new warning option -Wunused-local-typedefs was added for C, C++,
1191       Objective-C and Objective-C++. This warning diagnoses typedefs
1192       locally defined in a function, and otherwise not used.
1193     * A new experimental command-line option -ftrack-macro-expansion was
1194       added for C, C++, Objective-C, Objective-C++ and Fortran. It allows
1195       the compiler to emit diagnostic about the current macro expansion
1196       stack when a compilation error occurs in a macro expansion.
1197     * Experimental support for transactional memory has been added. It
1198       includes support in the compiler, as well as a supporting runtime
1199       library called libitm. To compile code with transactional memory
1200       constructs, use the -fgnu-tm option.
1201       Support is currently available for Alpha, ARM, PowerPC, SH, SPARC,
1202       and 32-bit/64-bit x86 platforms.
1203       For more details on transactional memory see [6]the GCC WiKi.
1204     * Support for atomic operations specifying the C++11/C11 memory model
1205       has been added. These new __atomic routines replace the existing
1206       __sync built-in routines.
1207       Atomic support is also available for memory blocks. Lock-free
1208       instructions will be used if a memory block is the same size and
1209       alignment as a supported integer type. Atomic operations which do
1210       not have lock-free support are left as function calls. A set of
1211       library functions is available on the GCC atomic wiki in the
1212       "External Atomics Library" section.
1213       For more details on the memory models and features, see the
1214       [7]atomic wiki.
1215     * When a binary operation is performed on vector types and one of the
1216       operands is a uniform vector, it is possible to replace the vector
1217       with the generating element. For example:
1218typedef int v4si __attribute__ ((vector_size (16)));
1219v4si res, a = {1,2,3,4};
1220int x;
1221
1222res = 2 + a;  /* means {2,2,2,2} + a  */
1223res = a - x;  /* means a - {x,x,x,x}  */
1224
1225  C
1226
1227     * There is support for some more features from the C11 revision of
1228       the ISO C standard. GCC now accepts the options -std=c11 and
1229       -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x.
1230          + Unicode strings (previously supported only with options such
1231            as -std=gnu11, now supported with -std=c11), and the
1232            predefined macros __STDC_UTF_16__ and __STDC_UTF_32__.
1233          + Nonreturning functions (_Noreturn and <stdnoreturn.h>).
1234          + Alignment support (_Alignas, _Alignof, max_align_t,
1235            <stdalign.h>).
1236          + A built-in function __builtin_complex is provided to support C
1237            library implementation of the CMPLX family of macros.
1238
1239  C++
1240
1241     * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat
1242       options, which are equivalent to -std=c++0x, -std=gnu++0x, and
1243       -Wc++0x-compat, respectively.
1244     * G++ now implements [8]C++11 extended friend syntax:
1245
1246template<class W>
1247class Q
1248{
1249  static const int I = 2;
1250public:
1251  friend W;
1252};
1253
1254struct B
1255{
1256  int ar[Q<B>::I];
1257};
1258
1259     * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit
1260       override control.
1261
1262struct B {
1263  virtual void f() const final;
1264  virtual void f(int);
1265};
1266
1267struct D : B {
1268  void f() const;            // error: D::f attempts to override final B::f
1269  void f(long) override;     // error: doesn't override anything
1270  void f(int) override;      // ok
1271};
1272
1273struct E final { };
1274struct F: E { }; // error: deriving from final class
1275
1276     * G++ now implements [10]C++11 non-static data member initializers.
1277
1278struct A {
1279  int i = 42;
1280} a; // initializes a.i to 42
1281
1282     * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11
1283       user-defined literals.
1284
1285// Not actually a good approximation.  :)
1286constexpr long double operator"" _degrees (long double d) { return d * 0.0175; }
1287long double pi = 180.0_degrees;
1288
1289     * G++ now implements [12]C++11 alias-declarations.
1290
1291template <class T> using Ptr = T*;
1292Ptr<int> ip;  // decltype(ip) is int*
1293
1294     * Thanks to Ville Voutilainen and Pedro Lamarao, G++ now implements
1295       [13]C++11 delegating constructors.
1296
1297struct A {
1298  A(int);
1299  A(): A(42) { } // delegate to the A(int) constructor
1300};
1301
1302     * G++ now fully implements C++11 atomic classes rather than just
1303       integer derived classes.
1304
1305class POD {
1306  int a;
1307  int b;
1308};
1309std::atomic<POD> my_atomic_POD;
1310
1311     * G++ now sets the predefined macro __cplusplus to the correct value,
1312       199711L for C++98/03, and 201103L for C++11.
1313     * G++ now correctly implements the two-phase lookup rules such that
1314       an unqualified name used in a template must have an appropriate
1315       declaration found either in scope at the point of definition of the
1316       template or by argument-dependent lookup at the point of
1317       instantiation. As a result, code that relies on a second
1318       unqualified lookup at the point of instantiation to find functions
1319       declared after the template or in dependent bases will be rejected.
1320       The compiler will suggest ways to fix affected code, and using the
1321       -fpermissive compiler flag will allow the code to compile with a
1322       warning.
1323
1324template <class T>
1325void f() { g(T()); } // error, g(int) not found by argument-dependent lookup
1326void g(int) { } // fix by moving this declaration before the declaration of f
1327
1328template <class T>
1329struct A: T {
1330  // error, B::g(B) not found by argument-dependent lookup
1331  void f() { g(T()); } // fix by using this->g or A::g
1332};
1333
1334struct B { void g(B); };
1335
1336int main()
1337{
1338  f<int>();
1339  A<B>().f();
1340}
1341
1342     * G++ now properly re-uses stack space allocated for temporary
1343       objects when their lifetime ends, which can significantly lower
1344       stack consumption for some C++ functions. As a result of this, some
1345       code with undefined behavior will now break:
1346
1347const int &f(const int &i) { return i; }
1348....
1349const int &x = f(1);
1350const int &y = f(2);
1351
1352       Here, x refers to the temporary allocated to hold the 1 argument,
1353       which only lives until the end of the initialization; it
1354       immediately becomes a dangling reference. So the next statement
1355       re-uses the stack slot to hold the 2 argument, and users of x get
1356       that value instead.
1357       Note that this should not cause any change of behavior for
1358       temporaries of types with non-trivial destructors, as they are
1359       already destroyed at end of full-expression; the change is that now
1360       the storage is released as well.
1361     * A new command-line option -Wdelete-non-virtual-dtor has been added
1362       to warn when delete is used to destroy an instance of a class which
1363       has virtual functions and non-virtual destructor. It is unsafe to
1364       delete an instance of a derived class through a pointer to a base
1365       class if the base class does not have a virtual destructor. This
1366       warning is enabled by -Wall.
1367     * A new command-line option -Wzero-as-null-pointer-constant has been
1368       added to warn when a literal '0' is used as null pointer constant.
1369       It can be useful to facilitate the conversion to nullptr in C++11.
1370     * As per C++98, access-declarations are now deprecated by G++.
1371       Using-declarations are to be used instead. Furthermore, some
1372       efforts have been made to improve the support of class scope
1373       using-declarations. In particular, using-declarations referring to
1374       a dependent type now work as expected ([14]bug c++/14258).
1375     * The ELF symbol visibility of a template instantiation is now
1376       properly constrained by the visibility of its template arguments
1377       ([15]bug c++/35688).
1378
1379    Runtime Library (libstdc++)
1380
1381     * [16]Improved experimental support for the new ISO C++ standard,
1382       C++11, including:
1383          + using noexcept in most of the library;
1384          + implementations of pointer_traits, allocator_traits and
1385            scoped_allocator_adaptor;
1386          + uses-allocator construction for tuple;
1387          + vector meets the allocator-aware container requirements;
1388          + replacing monotonic_clock with steady_clock;
1389          + enabling the thread support library on most POSIX targets;
1390          + many small improvements to conform to the FDIS.
1391     * Added --enable-clocale=newlib configure option.
1392     * Debug Mode iterators for unordered associative containers.
1393     * Avoid polluting the global namespace and do not include <unistd.h>.
1394
1395  Fortran
1396
1397     * The compile flag [17]-fstack-arrays has been added, which causes
1398       all local arrays to be put on stack memory. For some programs this
1399       will improve the performance significantly. If your program uses
1400       very large local arrays, it is possible that you will have to
1401       extend your runtime limits for stack memory.
1402     * The [18]-Ofast flag now also implies [19]-fno-protect-parens and
1403       [20]-fstack-arrays.
1404     * Front-end optimizations can now be selected by the
1405       [21]-ffrontend-optimize option and deselected by the
1406       -fno-frontend-optimize option.
1407     * When front-end optimization removes a function call,
1408       [22]-Wfunction-elimination warns about that.
1409     * When performing front-end-optimization, the
1410       [23]-faggressive-function-elimination option allows the removal of
1411       duplicate function calls even for impure functions.
1412     * The flag [24]-Wreal-q-constant has been added, which warns if
1413       floating-point literals have been specified using q (such as
1414       1.0q0); the q marker is now supported as a vendor extension to
1415       denote quad precision (REAL(16) or, if not available, REAL(10)).
1416       Consider using a kind parameter (such as in 1.0_qp) instead, which
1417       can be obtained via [25]SELECTED_REAL_KIND.
1418     * The GFORTRAN_USE_STDERR environment variable has been removed. GNU
1419       Fortran now always prints error messages to standard error. If you
1420       wish to redirect standard error, please consult the manual for your
1421       OS, shell, batch environment etc. as appropriate.
1422     * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment
1423       variable have been removed. When encountering a serious error,
1424       gfortran will now always abort the program. Whether a core dump is
1425       generated depends on the user environment settings; see the ulimit
1426       -c setting for POSIX shells, limit coredumpsize for C shells, and
1427       the [26]WER user-mode dumps settings on Windows.
1428     * The [27]-fbacktrace option is now enabled by default. When
1429       encountering a fatal error, gfortran will attempt to print a
1430       backtrace to standard error before aborting. It can be disabled
1431       with -fno-backtrace. Note: On POSIX targets with the addr2line
1432       utility from GNU binutils, GNU Fortran can print a backtrace with
1433       function name, file name, line number information in addition to
1434       the addresses; otherwise only the addresses are printed.
1435     * [28]Fortran 2003:
1436          + Generic interface names which have the same name as derived
1437            types are now supported, which allows to write constructor
1438            functions. Note that Fortran does not support static
1439            constructor functions; only default initialization or an
1440            explicit structure-constructor initialization are available.
1441          + [29]Polymorphic (class) arrays are now supported.
1442     * [30]Fortran 2008:
1443          + Support for the DO CONCURRENT construct has been added, which
1444            allows the user to specify that individual loop iterations
1445            have no interdependencies.
1446          + [31]Coarrays: Full single-image support except for polymorphic
1447            coarrays. Additionally, preliminary support for multiple
1448            images via an MPI-based [32]coarray communication library has
1449            been added. Note: The library version is not yet usable as
1450            remote coarray access is not yet possible.
1451     * [33]TS 29113:
1452          + New flag [34]-std=f2008ts permits programs that are expected
1453            to conform to the Fortran 2008 standard and the draft
1454            Technical Specification (TS) 29113 on Further Interoperability
1455            of Fortran with C.
1456          + The OPTIONAL attribute is now allowed for dummy arguments of
1457            BIND(C) procedures.
1458          + The RANK intrinsic has been added.
1459          + The implementation of the ASYNCHRONOUS attribute in GCC is
1460            compatible with the candidate draft of TS 29113 (since GCC
1461            4.6).
1462
1463  Go
1464
1465     * GCC 4.7 implements the [35]Go 1 language standard. The library
1466       support in 4.7.0 is not quite complete, due to release timing.
1467       Release 4.7.1 includes complete support for Go 1. The Go library is
1468       from the Go 1.0.1 release.
1469     * Go has been tested on GNU/Linux and Solaris platforms. It may work
1470       on other platforms as well.
1471
1472New Targets and Target Specific Improvements
1473
1474  ARM
1475
1476     * GCC now supports the Cortex-A7 processor implementing the v7-a
1477       version of the architecture using the option -mcpu=cortex-a7.
1478     * The default vector size in auto-vectorization for NEON is now 128
1479       bits. If vectorization fails thusly, the vectorizer tries again
1480       with 64-bit vectors.
1481     * A new option -mvectorize-with-neon-double was added to allow users
1482       to change the vector size to 64 bits.
1483
1484  AVR
1485
1486     * GCC now supports the XMEGA architecture. This requires GNU binutils
1487       2.22 or later.
1488     * Support for the [36]named address spaces __flash, __flash1, ...,
1489       __flash5 and __memx has been added. These address spaces locate
1490       read-only data in flash memory and allow reading from flash memory
1491       by means of ordinary C code, i.e. without the need of (inline)
1492       assembler code:
1493
1494const __flash int values[] = { 42, 31 };
1495
1496int add_values (const __flash int *p, int i)
1497{
1498    return values[i] + *p;
1499}
1500
1501     * Support has been added for the AVR-specific configure option
1502       --with-avrlibc=yes in order to arrange for better integration of
1503       [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2
1504       and newer and will only take effect in non-RTEMS configurations. If
1505       avr-gcc is configured for RTEMS, the option will be ignored which
1506       is the same as specifying --with-avrlibc=no. See [38]PR54461 for
1507       more technical details.
1508     * Support for AVR-specific [39]built-in functions has been added.
1509     * Support has been added for the signed and unsigned 24-bit scalar
1510       integer types __int24 and __uint24.
1511     * New command-line options -maccumulate-args, -mbranch-cost=cost and
1512       -mstrict-X were added to allow better fine-tuning of code
1513       optimization.
1514     * The command option -fdata-sections now also takes affect on the
1515       section names of variables with the progmem attribute.
1516     * A new inline assembler print modifier %i to print a RAM address as
1517       I/O address has been added:
1518
1519#include <avr/io.h> /* Port Definitions from AVR-LibC */
1520
1521void set_portb (uint8_t value)
1522{
1523    asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory");
1524}
1525
1526       The offset between an I/O address and the RAM address for that I/O
1527       location is device-specific. This offset is taken into account when
1528       printing a RAM address with the %i modifier so that the address is
1529       suitable to be used as operand in an I/O command. The address must
1530       be a constant integer known at compile time.
1531     * The inline assembler constraint "R" to represent integers in the
1532       range -6 ... 5 has been removed without replacement.
1533     * Many optimizations to:
1534          + 64-bit integer arithmetic
1535          + Widening multiplication
1536          + Integer division by a constant
1537          + Avoid constant reloading in multi-byte instructions.
1538          + Micro-optimizations for special instruction sequences.
1539          + Generic built-in functions like __builtin_ffs*,
1540            __builtin_clz*, etc.
1541          + If-else decision trees generated by switch instructions
1542          + Merging of data located in flash memory
1543          + New libgcc variants for devices with 8-bit wide stack pointer
1544          + ...
1545     * Better documentation:
1546          + Handling of EIND and indirect jumps on devices with more than
1547            128 KiB of program memory.
1548          + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function
1549            registers.
1550          + Function attributes OS_main and OS_task.
1551          + AVR-specific built-in macros.
1552
1553  C6X
1554
1555     * Support has been added for the Texas Instruments C6X family of
1556       processors.
1557
1558  CR16
1559
1560     * Support has been added for National Semiconductor's CR16
1561       architecture.
1562
1563  Epiphany
1564
1565     * Support has been added for Adapteva's Epiphany architecture.
1566
1567  IA-32/x86-64
1568
1569     * Support for Intel AVX2 intrinsics, built-in functions and code
1570       generation is available via -mavx2.
1571     * Support for Intel BMI2 intrinsics, built-in functions and code
1572       generation is available via -mbmi2.
1573     * Implementation and automatic generation of __builtin_clz* using the
1574       lzcnt instruction is available via -mlzcnt.
1575     * Support for Intel FMA3 intrinsics and code generation is available
1576       via -mfma.
1577     * A new -mfsgsbase command-line option is available that makes GCC
1578       generate new segment register read/write instructions through
1579       dedicated built-ins.
1580     * Support for the new Intel rdrnd instruction is available via
1581       -mrdrnd.
1582     * Two additional AVX vector conversion instructions are available via
1583       -mf16c.
1584     * Support for new Intel processor codename IvyBridge with RDRND,
1585       FSGSBASE and F16C is available through -march=core-avx-i.
1586     * Support for the new Intel processor codename Haswell with AVX2,
1587       FMA, BMI, BMI2, LZCNT is available through -march=core-avx2.
1588     * Support for new AMD family 15h processors (Piledriver core) is now
1589       available through -march=bdver2 and -mtune=bdver2 options.
1590     * Support for [40]the x32 psABI is now available through the -mx32
1591       option.
1592     * Windows mingw targets are using the -mms-bitfields option by
1593       default.
1594     * Windows x86 targets are using the __thiscall calling convention for
1595       C++ class-member functions.
1596     * Support for the configure option --with-threads=posix for Windows
1597       mingw targets.
1598
1599  MIPS
1600
1601     * GCC now supports thread-local storage (TLS) for MIPS16. This
1602       requires GNU binutils 2.22 or later.
1603     * GCC can now generate code specifically for the Cavium Octeon+ and
1604       Octeon2 processors. The associated command-line options are
1605       -march=octeon+ and -march=octeon2 respectively. Both options
1606       require GNU binutils 2.22 or later.
1607     * GCC can now work around certain 24k errata, under the control of
1608       the command-line option -mfix-24k. These workarounds require GNU
1609       binutils 2.20 or later.
1610     * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build
1611       n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux
1612       toolchain that generates 32-bit code by default. Use the
1613       configure-time option --enable-targets=all to select these extra
1614       multilibs.
1615     * Passing -fno-delayed-branch now also stops the assembler from
1616       automatically filling delay slots.
1617
1618  PowerPC/PowerPC64
1619
1620     * Vectors of type vector long long or vector long are passed and
1621       returned using the same method as other vectors with the VSX
1622       instruction set. Previously GCC did not adhere to the ABI for
1623       128-bit vectors with 64-bit integer base types (PR 48857). This
1624       will also be fixed in the GCC 4.6.1 and 4.5.4 releases.
1625     * A new option -mno-pointers-to-nested-functions was added to allow
1626       AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify
1627       that the compiler should not load up the chain register (r11)
1628       before calling a function through a pointer. If you use this
1629       option, you cannot call nested functions through a pointer, or call
1630       other languages that might use the static chain.
1631     * A new option msave-toc-indirect was added to allow AIX
1632       32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we
1633       save the TOC in the prologue for indirect calls or generate the
1634       save inline. This can speed up some programs that call through a
1635       function pointer a lot, but it can slow down other functions that
1636       only call through a function pointer in exceptional cases.
1637     * The PowerPC port will now enable machine-specific built-in
1638       functions when the user switches the target machine using the
1639       #pragma GCC target or __attribute__ ((__target__ ("target"))) code
1640       sequences. In addition, the target macros are updated. However, due
1641       to the way the -save-temps switch is implemented, you won't see the
1642       effect of these additional macros being defined in preprocessor
1643       output.
1644
1645  SH
1646
1647     * A new option -msoft-atomic has been added. When it is specified,
1648       GCC will generate GNU/Linux-compatible gUSA atomic sequences for
1649       the new __atomic routines.
1650     * Since it is neither supported by GAS nor officially documented,
1651       code generation for little endian SH2A has been disabled.
1652       Specifying -ml with -m2a* will now result in a compiler error.
1653     * The defunct -mbranch-cost option has been fixed.
1654     * Some improvements to the generated code of:
1655          + Utilization of the tst #imm,R0 instruction.
1656          + Dynamic shift instructions on SH2A.
1657          + Integer absolute value calculations.
1658     * The -mdiv= option for targets other than SHmedia has been fixed and
1659       documented.
1660
1661  SPARC
1662
1663     * The option -mflat has been reinstated. When it is specified, the
1664       compiler will generate code for a single register window model.
1665       This is essentially a new implementation and the corresponding
1666       debugger support has been added to GDB 7.4.
1667     * Support for the options -mtune=native and -mcpu=native has been
1668       added on selected native platforms (GNU/Linux and Solaris).
1669     * Support for the SPARC T3 (Niagara 3) processor has been added.
1670     * VIS:
1671          + An intrinsics header visintrin.h has been added.
1672          + Builtin intrinsics for the VIS 1.0 edge handling and pixel
1673            compare instructions have been added.
1674          + The little-endian version of alignaddr is now supported.
1675          + When possible, VIS builtins are marked const, which should
1676            increase the compiler's ability to optimize VIS operations.
1677          + The compiler now properly tracks the %gsr register and how it
1678            behaves as an input for various VIS instructions.
1679          + Akin to fzero, the compiler can now generate fone instructions
1680            in order to set all of the bits of a floating-point register
1681            to 1.
1682          + The documentation for the VIS intrinsics in the GCC manual has
1683            been brought up to date and many inaccuracies were fixed.
1684          + Intrinsics for the VIS 2.0 bmask, bshuffle, and
1685            non-condition-code setting edge instructions have been added.
1686            Their availability is controlled by the new -mvis2 and
1687            -mno-vis2 options. They are enabled by default on
1688            UltraSPARC-III and later CPUs.
1689     * Support for UltraSPARC Fused Multiply-Add floating-point extensions
1690       has been added. These instructions are enabled by default on SPARC
1691       T3 (Niagara 3) and later CPUs.
1692
1693  TILE-Gx/TILEPro
1694
1695     * Support has been added for the Tilera TILE-Gx and TILEPro families
1696       of processors.
1697
1698Other significant improvements
1699
1700     * A new option (-grecord-gcc-switches) was added that appends
1701       compiler command-line options that might affect code generation to
1702       the DW_AT_producer attribute string in the DWARF debugging
1703       information.
1704     * GCC now supports various new GNU extensions to the DWARF debugging
1705       information format, like [41]entry value and [42]call site
1706       information, [43]typed DWARF stack or [44]a more compact macro
1707       representation. Support for these extensions has been added to GDB
1708       7.4. They can be disabled through the -gstrict-dwarf command-line
1709       option.
1710
1711GCC 4.7.1
1712
1713   This is the [45]list of problem reports (PRs) from GCC's bug tracking
1714   system that are known to be fixed in the 4.7.1 release. This list might
1715   not be complete (that is, it is possible that some PRs that have been
1716   fixed are not listed here).
1717
1718   The Go frontend in the 4.7.1 release fully supports the [46]Go 1
1719   language standard.
1720
1721GCC 4.7.2
1722
1723   This is the [47]list of problem reports (PRs) from GCC's bug tracking
1724   system that are known to be fixed in the 4.7.2 release. This list might
1725   not be complete (that is, it is possible that some PRs that have been
1726   fixed are not listed here).
1727
1728GCC 4.7.3
1729
1730   This is the [48]list of problem reports (PRs) from GCC's bug tracking
1731   system that are known to be fixed in the 4.7.3 release. This list might
1732   not be complete (that is, it is possible that some PRs that have been
1733   fixed are not listed here).
1734
1735GCC 4.7.4
1736
1737   This is the [49]list of problem reports (PRs) from GCC's bug tracking
1738   system that are known to be fixed in the 4.7.4 release. This list might
1739   not be complete (that is, it is possible that some PRs that have been
1740   fixed are not listed here).
1741
1742
1743    For questions related to the use of GCC, please consult these web
1744    pages and the [50]GCC manuals. If that fails, the
1745    [51]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1746    web pages and the development of GCC are welcome on our developer
1747    list at [52]gcc@gcc.gnu.org. All of [53]our lists have public
1748    archives.
1749
1750   Copyright (C) [54]Free Software Foundation, Inc. Verbatim copying and
1751   distribution of this entire article is permitted in any medium,
1752   provided this notice is preserved.
1753
1754   These pages are [55]maintained by the GCC team. Last modified
1755   2014-12-06[56].
1756
1757References
1758
1759   1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html
1760   2. http://savannah.nongnu.org/bugs/?35407
1761   3. https://gcc.gnu.org/PR18145
1762   4. https://gcc.gnu.org/gcc-4.7/porting_to.html
1763   5. http://openmp.org/wp/openmp-specifications/
1764   6. https://gcc.gnu.org/wiki/TransactionalMemory
1765   7. https://gcc.gnu.org/wiki/Atomic/GCCMM
1766   8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1767   9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1768  10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1769  11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1770  12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1771  13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1772  14. https://gcc.gnu.org/PR14258
1773  15. https://gcc.gnu.org/PR35688
1774  16. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011
1775  17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
1776  18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689
1777  19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270
1778  20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
1779  21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275
1780  22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170
1781  23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270
1782  24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149
1783  25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html
1784  26. http://msdn.microsoft.com/en-us/library/bb787181%28v=vs.85%29.aspx
1785  27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183
1786  28. https://gcc.gnu.org/wiki/Fortran2003Status
1787  29. https://gcc.gnu.org/wiki/OOP
1788  30. https://gcc.gnu.org/wiki/Fortran2008Status
1789  31. https://gcc.gnu.org/wiki/Coarray
1790  32. https://gcc.gnu.org/wiki/CoarrayLib
1791  33. https://gcc.gnu.org/wiki/TS29113Status
1792  34. 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
1793  35. http://weekly.golang.org/doc/go1.html
1794  36. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html
1795  37. http://nongnu.org/avr-libc/
1796  38. https://gcc.gnu.org/PR54461
1797  39. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html
1798  40. https://sites.google.com/site/x32abi/
1799  41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1
1800  42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2
1801  43. http://www.dwarfstd.org/doc/040408.1.html
1802  44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1
1803  45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1
1804  46. http://weekly.golang.org/doc/go1.html
1805  47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2
1806  48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3
1807  49. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4
1808  50. https://gcc.gnu.org/onlinedocs/
1809  51. mailto:gcc-help@gcc.gnu.org
1810  52. mailto:gcc@gcc.gnu.org
1811  53. https://gcc.gnu.org/lists.html
1812  54. http://www.fsf.org/
1813  55. https://gcc.gnu.org/about.html
1814  56. http://validator.w3.org/check/referer
1815======================================================================
1816http://gcc.gnu.org/gcc-4.6/index.html
1817
1818                             GCC 4.6 Release Series
1819
1820   April 12, 2013
1821
1822   The [1]GNU project and the GCC developers are pleased to announce the
1823   release of GCC 4.6.4.
1824
1825   This release is a bug-fix release, containing fixes for regressions in
1826   GCC 4.6.3 relative to previous releases of GCC.
1827
1828Release History
1829
1830   GCC 4.6.4
1831          April 12, 2013 ([2]changes, [3]documentation)
1832
1833   GCC 4.6.3
1834          March 1, 2012 ([4]changes, [5]documentation)
1835
1836   GCC 4.6.2
1837          October 26, 2011 ([6]changes, [7]documentation)
1838
1839   GCC 4.6.1
1840          June 27, 2011 ([8]changes, [9]documentation)
1841
1842   GCC 4.6.0
1843          March 25, 2011 ([10]changes, [11]documentation)
1844
1845References and Acknowledgements
1846
1847   GCC used to stand for the GNU C Compiler, but since the compiler
1848   supports several other languages aside from C, it now stands for the
1849   GNU Compiler Collection.
1850
1851   A list of [12]successful builds is updated as new information becomes
1852   available.
1853
1854   The GCC developers would like to thank the numerous people that have
1855   contributed new features, improvements, bug fixes, and other changes as
1856   well as test results to GCC. This [13]amazing group of volunteers is
1857   what makes GCC successful.
1858
1859   For additional information about GCC please refer to the [14]GCC
1860   project web site or contact the [15]GCC development mailing list.
1861
1862   To obtain GCC please use [16]our mirror sites or [17]our SVN server.
1863
1864
1865    For questions related to the use of GCC, please consult these web
1866    pages and the [18]GCC manuals. If that fails, the
1867    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1868    web pages and the development of GCC are welcome on our developer
1869    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
1870    archives.
1871
1872   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
1873   distribution of this entire article is permitted in any medium,
1874   provided this notice is preserved.
1875
1876   These pages are [23]maintained by the GCC team. Last modified
1877   2014-06-28[24].
1878
1879References
1880
1881   1. http://www.gnu.org/
1882   2. http://gcc.gnu.org/gcc-4.6/changes.html
1883   3. https://gcc.gnu.org/onlinedocs/4.6.4/
1884   4. http://gcc.gnu.org/gcc-4.6/changes.html
1885   5. https://gcc.gnu.org/onlinedocs/4.6.3/
1886   6. http://gcc.gnu.org/gcc-4.6/changes.html
1887   7. https://gcc.gnu.org/onlinedocs/4.6.2/
1888   8. http://gcc.gnu.org/gcc-4.6/changes.html
1889   9. https://gcc.gnu.org/onlinedocs/4.6.1/
1890  10. http://gcc.gnu.org/gcc-4.6/changes.html
1891  11. https://gcc.gnu.org/onlinedocs/4.6.0/
1892  12. http://gcc.gnu.org/gcc-4.6/buildstat.html
1893  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
1894  14. http://gcc.gnu.org/index.html
1895  15. mailto:gcc@gcc.gnu.org
1896  16. http://gcc.gnu.org/mirrors.html
1897  17. http://gcc.gnu.org/svn.html
1898  18. https://gcc.gnu.org/onlinedocs/
1899  19. mailto:gcc-help@gcc.gnu.org
1900  20. mailto:gcc@gcc.gnu.org
1901  21. https://gcc.gnu.org/lists.html
1902  22. http://www.fsf.org/
1903  23. https://gcc.gnu.org/about.html
1904  24. http://validator.w3.org/check/referer
1905======================================================================
1906http://gcc.gnu.org/gcc-4.6/changes.html
1907
1908                             GCC 4.6 Release Series
1909                        Changes, New Features, and Fixes
1910
1911Caveats
1912
1913     * The options -b <machine> and -V <version> have been removed because
1914       they were unreliable. Instead, users should directly run
1915       <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to
1916       run a different version of gcc.
1917     * GCC now has stricter checks for invalid command-line options. In
1918       particular, when gcc was called to link object files rather than
1919       compile source code, it would previously accept and ignore all
1920       options starting with --, including linker options such as
1921       --as-needed and --export-dynamic, although such options would
1922       result in errors if any source code was compiled. Such options, if
1923       unknown to the compiler, are now rejected in all cases; if the
1924       intent was to pass them to the linker, options such as
1925       -Wl,--as-needed should be used.
1926     * Versions of the GNU C library up to and including 2.11.1 included
1927       an [1]incorrect implementation of the cproj function. GCC optimizes
1928       its builtin cproj according to the behavior specified and allowed
1929       by the ISO C99 standard. If you want to avoid discrepancies between
1930       the C library and GCC's builtin transformations when using cproj in
1931       your code, use GLIBC 2.12 or later. If you are using an older GLIBC
1932       and actually rely on the incorrect behavior of cproj, then you can
1933       disable GCC's transformations using -fno-builtin-cproj.
1934     * The C-only intermodule optimization framework (IMA, enabled by
1935       -combine) has been removed in favor of the new generic link-time
1936       optimization framework (LTO) introduced in [2]GCC 4.5.0.
1937     * GCC now ships with the LGPL-licensed libquadmath library, which
1938       provides quad-precision mathematical functions for targets with a
1939       __float128 datatype. __float128 is available for targets on 32-bit
1940       x86, x86-64 and Itanium architectures. The libquadmath library is
1941       automatically built on such targets when building the Fortran
1942       compiler.
1943     * New -Wunused-but-set-variable and -Wunused-but-set-parameter
1944       warnings were added for C, C++, Objective-C and Objective-C++.
1945       These warnings diagnose variables respective parameters which are
1946       only set in the code and never otherwise used. Usually such
1947       variables are useless and often even the value assigned to them is
1948       computed needlessly, sometimes expensively. The
1949       -Wunused-but-set-variable warning is enabled by default by -Wall
1950       flag and -Wunused-but-set-parameter by -Wall -Wextra flags.
1951     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
1952       rules for the layout of vectors that could lead to wrong code being
1953       generated. Vectors larger than 8 bytes in size are now by default
1954       aligned to an 8-byte boundary. This is an ABI change: code that
1955       makes explicit use of vector types may be incompatible with binary
1956       objects built with older versions of GCC. Auto-vectorized code is
1957       not affected by this change. (This change affects GCC versions
1958       4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.)
1959     * On AVR, variables with the progmem attribute to locate data in
1960       flash memory must be qualified as const.
1961     * Support for a number of older systems and recently unmaintained or
1962       untested target ports of GCC has been declared obsolete in GCC 4.6.
1963       Unless there is activity to revive them, the next release of GCC
1964       will have their sources permanently removed.
1965       All GCC ports for the following processor architectures have been
1966       declared obsolete:
1967          + Argonaut ARC (arc-*)
1968          + National Semiconductor CRX (crx-*)
1969          + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*,
1970            m68hc12-*-*, m6812-*-*)
1971          + Sunplus S+core (score-*)
1972       The following ports for individual systems on particular
1973       architectures have been obsoleted:
1974          + Interix (i[34567]86-*-interix3*)
1975          + NetWare x86 (i[3456x]86-*-netware*)
1976          + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*)
1977          + MCore PE (mcore-*-pe*)
1978          + SH SymbianOS (sh*-*-symbianelf*)
1979          + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*)
1980          + M68K uClinux old ABI (m68k-*-uclinuxoldabi*)
1981          + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*,
1982            vax-*-netbsd*, but not *-*-netbsdelf*)
1983       The i[34567]86-*-pe alias for Cygwin targets has also been
1984       obsoleted; users should configure for i[34567]86-*-cygwin* instead.
1985       Certain configure options to control the set of libraries built
1986       with GCC on some targets have been obsoleted. On ARM targets, the
1987       options --disable-fpu, --disable-26bit, --disable-underscore,
1988       --disable-interwork, --disable-biendian and --disable-nofmult have
1989       been obsoleted. On MIPS targets, the options
1990       --disable-single-float, --disable-biendian and --disable-softfloat
1991       have been obsoleted.
1992     * Support has been removed for all the [3]configurations obsoleted in
1993       GCC 4.5.
1994     * More information on porting to GCC 4.6 from previous versions of
1995       GCC can be found in the [4]porting guide for this release.
1996
1997General Optimizer Improvements
1998
1999     * A new general optimization level, -Ofast, has been introduced. It
2000       combines the existing optimization level -O3 with options that can
2001       affect standards compliance but result in better optimized code.
2002       For example, -Ofast enables -ffast-math.
2003     * Link-time optimization improvements:
2004          + The [5]Scalable Whole Program Optimizer (WHOPR) project has
2005            stabilized to the point of being usable. It has become the
2006            default mode when using the LTO optimization model. Link time
2007            optimization can now split itself into multiple parallel
2008            compilations. Parallelism is controlled with -flto=n (where n
2009            specifies the number of compilations to execute in parallel).
2010            GCC can also cooperate with a GNU make job server by
2011            specifying the -flto=jobserver option and adding + to the
2012            beginning of the Makefile rule executing the linker.
2013            Classical LTO mode can be enforced by -flto-partition=none.
2014            This may result in small code quality improvements.
2015          + A large number of bugs were fixed. GCC itself, Mozilla Firefox
2016            and other large applications can be built with LTO enabled.
2017          + The linker plugin support improvements
2018               o Linker plugin is now enabled by default when the linker
2019                 is detected to have plugin support. This is the case for
2020                 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and
2021                 the Gold linker on ELF targets. Plugin support of the
2022                 Apple linker on Darwin is not compatible with GCC. The
2023                 linker plugin can also be controlled by the
2024                 -fuse-linker-plugin command line option.
2025               o Resolution information from the linker plugin is used to
2026                 drive whole program assumptions. Use of the linker plugin
2027                 results in more aggressive optimization on binaries and
2028                 on shared libraries that use the hidden visibility
2029                 attribute. Consequently the use of -fwhole-program is not
2030                 necessary in addition to LTO.
2031          + Hidden symbols used from non-LTO objects now have to be
2032            explicitly annotated with externally_visible when the linker
2033            plugin is not used.
2034          + C++ inline functions and virtual tables are now privatized
2035            more aggressively, leading to better inter-procedural
2036            optimization and faster dynamic linking.
2037          + Memory usage and intermediate language streaming performance
2038            have been improved.
2039          + Static constructors and destructors from individual units are
2040            inlined into a single function. This can significantly improve
2041            startup times of large C++ applications where static
2042            constructors are very common. For example, static constructors
2043            are used when including the iostream header.
2044          + Support for the Ada language has been added.
2045     * Interprocedural optimization improvements
2046          + The interprocedural framework was re-tuned for link time
2047            optimization. Several scalability issues were resolved.
2048          + Improved auto-detection of const and pure functions. Newly,
2049            noreturn functions are auto-detected.
2050            The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is
2051            available that informs users when adding attributes to headers
2052            might improve code generation.
2053          + A number of inlining heuristic improvements. In particular:
2054               o Partial inlining is now supported and enabled by default
2055                 at -O2 and greater. The feature can be controlled via
2056                 -fpartial-inlining.
2057                 Partial inlining splits functions with short hot path to
2058                 return. This allows more aggressive inlining of the hot
2059                 path leading to better performance and often to code size
2060                 reductions (because cold parts of functions are not
2061                 duplicated).
2062               o Scalability for large compilation units was improved
2063                 significantly.
2064               o Inlining of callbacks is now more aggressive.
2065               o Virtual methods are considered for inlining when the
2066                 caller is inlined and devirtualization is then possible.
2067               o Inlining when optimizing for size (either in cold regions
2068                 of a program or when compiling with -Os) was improved to
2069                 better handle C++ programs with larger abstraction
2070                 penalty, leading to smaller and faster code.
2071          + The IPA reference optimization pass detecting global variables
2072            used or modified by functions was strengthened and sped up.
2073          + Functions whose address was taken are now optimized out when
2074            all references to them are dead.
2075          + A new inter-procedural static profile estimation pass detects
2076            functions that are executed once or unlikely to be executed.
2077            Unlikely executed functions are optimized for size. Functions
2078            executed once are optimized for size except for the inner
2079            loops.
2080          + On most targets with named section support, functions used
2081            only at startup (static constructors and main), functions used
2082            only at exit and functions detected to be cold are placed into
2083            separate text segment subsections. This extends the
2084            -freorder-functions feature and is controlled by the same
2085            switch. The goal is to improve the startup time of large C++
2086            programs.
2087            Proper function placement requires linker support. GNU ld
2088            2.21.51 on ELF targets was updated to place those functions
2089            together within the text section leading to better code
2090            locality and faster startup times of large C++ programs. The
2091            feature is also supported in the Apple linker. Support in the
2092            gold linker is planned.
2093     * A new switch -fstack-usage has been added. It makes the compiler
2094       output stack usage information for the program, on a per-function
2095       basis, in an auxiliary file.
2096     * A new switch -fcombine-stack-adjustments has been added. It can be
2097       used to enable or disable the compiler's stack-slot combining pass
2098       which before was enabled automatically at -O1 and above, but could
2099       not be controlled on its own.
2100     * A new switch -fstrict-volatile-bitfields has been added. Using it
2101       indicates that accesses to volatile bitfields should use a single
2102       access of the width of the field's type. This option can be useful
2103       for precisely defining and accessing memory-mapped peripheral
2104       registers from C or C++.
2105
2106Compile time and memory usage improvements
2107
2108     * Datastructures used by the dataflow framework in GCC were
2109       reorganized for better memory usage and more cache locality.
2110       Compile time is improved especially on units with large functions
2111       (possibly resulting from a lot of inlining) not fitting into the
2112       processor cache. The compile time of the GCC C compiler binary with
2113       link-time optimization went down by over 10% (benchmarked on x86-64
2114       target).
2115
2116New Languages and Language specific improvements
2117
2118  Ada
2119
2120     * Stack checking has been improved on selected architectures (Alpha,
2121       IA-32/x86-64, RS/6000 and SPARC): it now will detect stack
2122       overflows in all cases on these architectures.
2123     * Initial support for Ada 2012 has been added.
2124
2125  C family
2126
2127     * A new warning, enabled by -Wdouble-promotion, has been added that
2128       warns about cases where a value of type float is implicitly
2129       promoted to double. This is especially helpful for CPUs that handle
2130       the former in hardware, but emulate the latter in software.
2131     * A new function attribute leaf was introduced. This attribute allows
2132       better inter-procedural optimization across calls to functions that
2133       return to the current unit only via returning or exception
2134       handling. This is the case for most library functions that have no
2135       callbacks.
2136     * Support for a new data type __int128 for targets having wide enough
2137       machine-mode support.
2138     * The new function attribute callee_pop_aggregate allows to specify
2139       if the caller or callee is responsible for popping the aggregate
2140       return pointer value from the stack.
2141     * Support for selectively enabling and disabling warnings via #pragma
2142       GCC diagnostic has been added. For instance:
2143#pragma GCC diagnostic error "-Wuninitialized"
2144  foo(a);                       /* error is given for this one */
2145#pragma GCC diagnostic push
2146#pragma GCC diagnostic ignored "-Wuninitialized"
2147  foo(b);                       /* no diagnostic for this one */
2148#pragma GCC diagnostic pop
2149  foo(c);                       /* error is given for this one */
2150#pragma GCC diagnostic pop
2151  foo(d);                       /* depends on command line options */
2152
2153     * The -fmax-errors=N option is now supported. Using this option
2154       causes the compiler to exit after N errors have been issued.
2155
2156  C
2157
2158     * There is now experimental support for some features from the
2159       upcoming C1X revision of the ISO C standard. This support may be
2160       selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions.
2161       Note that this support is experimental and may change incompatibly
2162       in future releases for consistency with changes to the C1X standard
2163       draft. The following features are newly supported as described in
2164       the N1539 draft of C1X (with changes agreed at the March 2011 WG14
2165       meeting); some other features were already supported with no
2166       compiler changes being needed, or have some support but not in full
2167       accord with N1539 (as amended).
2168          + Static assertions (_Static_assert keyword)
2169          + Typedef redefinition
2170          + New macros in <float.h>
2171          + Anonymous structures and unions
2172     * The new -fplan9-extensions option directs the compiler to support
2173       some extensions for anonymous struct fields which are implemented
2174       by the Plan 9 compiler. A pointer to a struct may be automatically
2175       converted to a pointer to an anonymous field when calling a
2176       function, in order to make the types match. An anonymous struct
2177       field whose type is a typedef name may be referred to using the
2178       typedef name.
2179
2180  C++
2181
2182     * Improved [7]experimental support for the upcoming C++0x ISO C++
2183       standard, including support for constexpr (thanks to Gabriel Dos
2184       Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide),
2185       noexcept, unrestricted unions, range-based for loops (thanks to
2186       Rodrigo Rivas Costa), opaque enum declarations (thanks also to
2187       Rodrigo), implicitly deleted functions and implicit move
2188       constructors.
2189     * When an extern declaration within a function does not match a
2190       declaration in the enclosing context, G++ now properly declares the
2191       name within the namespace of the function rather than the namespace
2192       which was open just before the function definition ([8]c++/43145).
2193     * GCC now warns by default when casting integers to larger pointer
2194       types. These warnings can be disabled with the option
2195       -Wno-int-to-pointer-cast, which is now also available in C++.
2196     * G++ no longer optimizes using the assumption that a value of
2197       enumeration type will fall within the range specified by the
2198       standard, since that assumption is easily violated with a
2199       conversion from integer type ([9]c++/43680). The old behavior can
2200       be restored with -fstrict-enums.
2201     * The new -fnothrow-opt flag changes the semantics of a throw()
2202       exception specification to match the proposed semantics of the
2203       noexcept specification: just call terminate if an exception tries
2204       to propagate out of a function with such an exception
2205       specification. This dramatically reduces or eliminates the code
2206       size overhead from adding the exception specification.
2207     * The new -Wnoexcept flag will suggest adding a noexcept qualifier to
2208       a function that the compiler can tell doesn't throw if it would
2209       change the value of a noexcept expression.
2210     * The -Wshadow option now warns if a local variable or type
2211       declaration shadows another type in C++. Note that the compiler
2212       will not warn if a local variable shadows a struct/class/enum, but
2213       will warn if it shadows an explicit typedef.
2214     * When an identifier is not found in the current scope, G++ now
2215       offers suggestions about which identifier might have been intended.
2216     * G++ now issues clearer diagnostics for missing semicolons after
2217       class, struct, and union definitions.
2218     * G++ now issues clearer diagnostics for missing semicolons after
2219       class member declarations.
2220     * G++ now issues clearer diagnostics when a colon is used in a place
2221       where a double-colon was intended.
2222     * G++ no longer accepts mutable on reference members ([10]c++/33558).
2223       Use -fpermissive to allow the old, non-conforming behaviour.
2224     * A few mangling fixes have been made, to attribute const/volatile on
2225       function pointer types, decltype of a plain decl, and use of a
2226       function parameter in the declaration of another parameter. By
2227       default the compiler still uses the old mangling, but emits aliases
2228       with the new mangling on targets that support strong aliases. Users
2229       can switch over entirely to the new mangling with -fabi-version=5
2230       or -fabi-version=0. -Wabi will now warn about code that uses the
2231       old mangling.
2232     * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified
2233       type to be default initialized unless the type has a user-declared
2234       default constructor. In 4.6.2 G++ implements the proposed
2235       resolution of [11]DR 253, so default initialization is allowed if
2236       it initializes all subobjects. Code that fails to compile can be
2237       fixed by providing an initializer e.g.
2238    struct A { A(); };
2239    struct B : A { int i; };
2240    const B b = B();
2241       Use -fpermissive to allow the old, non-conforming behaviour.
2242
2243    Runtime Library (libstdc++)
2244
2245     * [12]Improved experimental support for the upcoming ISO C++
2246       standard, C++0x, including using constexpr and nullptr.
2247     * Performance improvements to the [13]Debug Mode, thanks to Franc,ois
2248       Dumont.
2249     * Atomic operations used for reference-counting are annotated so that
2250       they can be understood by race detectors such as Helgrind, see
2251       [14]Data Race Hunting.
2252     * Most libstdc++ standard headers have been changed to no longer
2253       include the cstddef header as an implementation detail. Code that
2254       relied on that header being included as side-effect of including
2255       other standard headers will need to include cstddef explicitly.
2256
2257  Fortran
2258
2259     * On systems supporting the libquadmath library, GNU Fortran now also
2260       supports a quad-precision, kind=16 floating-point data type
2261       (REAL(16), COMPLEX(16)). As the data type is not fully supported in
2262       hardware, calculations might be one to two orders of magnitude
2263       slower than with the 4, 8 or 10 bytes floating-point data types.
2264       This change does not affect systems which support REAL(16) in
2265       hardware nor those which do not support libquadmath.
2266     * Much improved compile time for large array constructors.
2267     * In order to reduce execution time and memory consumption, use of
2268       temporary arrays in assignment expressions is avoided for many
2269       cases. The compiler now reverses loops in order to avoid generating
2270       a temporary array where possible.
2271     * Improved diagnostics, especially with -fwhole-file.
2272     * The -fwhole-file flag is now enabled by default. This improves code
2273       generation and diagnostics. It can be disabled using the deprecated
2274       -fno-whole-file flag.
2275     * Support the generation of Makefile dependencies via the [15]-M...
2276       flags of GCC; you may need to specify the -cpp option in addition.
2277       The dependencies take modules, Fortran's include, and CPP's
2278       #include into account. Note: Using -M for the module path is no
2279       longer supported, use -J instead.
2280     * The flag -Wconversion has been modified to only issue warnings
2281       where a conversion leads to information loss. This drastically
2282       reduces the number of warnings; -Wconversion is thus now enabled
2283       with -Wall. The flag -Wconversion-extra has been added and also
2284       warns about other conversions; -Wconversion-extra typically issues
2285       a huge number of warnings, most of which can be ignored.
2286     * A new command-line option -Wunused-dummy-argument warns about
2287       unused dummy arguments and is included in -Wall. Before,
2288       -Wunused-variable also warned about unused dummy arguments.
2289     * Fortran 2003 support has been extended:
2290          + Improved support for polymorphism between libraries and
2291            programs and for complicated inheritance patterns (cf.
2292            [16]object-oriented programming).
2293          + Experimental support of the ASSOCIATE construct.
2294          + In pointer assignments it is now possible to specify the lower
2295            bounds of the pointer and, for a rank-1 or a simply contiguous
2296            data-target, to remap the bounds.
2297          + Automatic (re)allocation: In intrinsic assignments to
2298            allocatable variables the left-hand side will be automatically
2299            allocated (if unallocated) or reallocated (if the shape or
2300            type parameter is different). To avoid the small performance
2301            penalty, you can use a(:) = ... instead of a = ... for arrays
2302            and character strings - or disable the feature using -std=f95
2303            or -fno-realloc-lhs.
2304          + Deferred type parameter: For scalar allocatable and pointer
2305            variables the character length can be deferred.
2306          + Namelist variables with allocatable and pointer attribute and
2307            nonconstant length type parameter are supported.
2308     * Fortran 2008 support has been extended:
2309          + Experimental [17]coarray support (for one image only, i.e.
2310            num_images() == 1); use the [18]-fcoarray=single flag to
2311            enable it.
2312          + The STOP and the new ERROR STOP statements now support all
2313            constant expressions.
2314          + Support for the CONTIGUOUS attribute.
2315          + Support for ALLOCATE with MOLD.
2316          + Support for the STORAGE_SIZE intrinsic inquiry function.
2317          + Support of the NORM2 and PARITY intrinsic functions.
2318          + The following bit intrinsics were added: POPCNT and POPPAR for
2319            counting the number of 1 bits and returning the parity; BGE,
2320            BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR
2321            for combined left and right shifts, MASKL and MASKR for simple
2322            left and right justified masks, MERGE_BITS for a bitwise merge
2323            using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations,
2324            and the transformational bit intrinsics IALL, IANY and
2325            IPARITY.
2326          + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine.
2327          + Support for the IMPURE attribute for procedures, which allows
2328            for ELEMENTAL procedures without the restrictions of PURE.
2329          + Null pointers (including NULL()) and not allocated variables
2330            can be used as actual argument to optional non-pointer,
2331            non-allocatable dummy arguments, denoting an absent argument.
2332          + Non-pointer variables with TARGET attribute can be used as
2333            actual argument to POINTER dummies with INTENT(IN)
2334          + Pointers including procedure pointers and those in a derived
2335            type (pointer components) can now be initialized by a target
2336            instead of only by NULL.
2337          + The EXIT statement (with construct-name) can now be used to
2338            leave not only the DO but also the ASSOCIATE, BLOCK, IF,
2339            SELECT CASE and SELECT TYPE constructs.
2340          + Internal procedures can now be used as actual argument.
2341          + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS
2342            and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV
2343            have been added; these arrays contain the supported kind
2344            values for the respective types.
2345          + The module procedures C_SIZEOF of the intrinsic module
2346            ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of
2347            ISO_FORTRAN_ENV have been implemented.
2348          + Minor changes: obsolescence diagnostics for ENTRY was added
2349            for -std=f2008; a line may start with a semicolon; for
2350            internal and module procedures END can be used instead of END
2351            SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes
2352            a RADIX argument; intrinsic types are supported for
2353            TYPE(intrinsic-type-spec); multiple type-bound procedures can
2354            be declared in a single PROCEDURE statement; implied-shape
2355            arrays are supported for named constants (PARAMETER). The
2356            transformational, three argument versions of BESSEL_JN and
2357            BESSEL_YN were added - the elemental, two-argument version had
2358            been added in GCC 4.4; note that the transformational
2359            functions use a recurrence algorithm.
2360
2361  Go
2362
2363   Support for the [19]Go programming language has been added to GCC. It
2364   is not enabled by default when you build GCC; use the
2365   --enable-languages configure option to build it. The driver program for
2366   compiling Go code is gccgo.
2367
2368   Go is currently known to work on GNU/Linux and RTEMS. Solaris support
2369   is in progress. It may or may not work on other platforms.
2370
2371  Objective-C and Objective-C++
2372
2373     * The -fobjc-exceptions flag is now required to enable Objective-C
2374       exception and synchronization syntax (introduced by the keywords
2375       @try, @catch, @finally and @synchronized).
2376     * A number of Objective-C 2.0 features and extensions are now
2377       supported by GCC. These features are enabled by default; you can
2378       disable them by using the new -fobjc-std=objc1 command-line option.
2379     * The Objective-C 2.0 dot-syntax is now supported. It is an
2380       alternative syntax for using getters and setters; object.count is
2381       automatically converted into [object count] or [object setCount:
2382       ...] depending on context; for example if (object.count > 0) is
2383       automatically compiled into the equivalent of if ([object count] >
2384       0) while object.count = 0; is automatically compiled into the
2385       equivalent ot [object setCount: 0];. The dot-syntax can be used
2386       with instance and class objects and with any setters or getters, no
2387       matter if they are part of a declared property or not.
2388     * Objective-C 2.0 declared properties are now supported. They are
2389       declared using the new @property keyword, and are most commonly
2390       used in conjunction with the new Objective-C 2.0 dot-syntax. The
2391       nonatomic, readonly, readwrite, assign, retain, copy, setter and
2392       getter attributes are all supported. Marking declared properties
2393       with __attribute__ ((deprecated)) is supported too.
2394     * The Objective-C 2.0 @synthesize and @dynamic keywords are
2395       supported. @synthesize causes the compiler to automatically
2396       synthesize a declared property, while @dynamic is used to disable
2397       all warnings for a declared property for which no implementation is
2398       provided at compile time. Synthesizing declared properties requires
2399       runtime support in most useful cases; to be able to use it with the
2400       GNU runtime, appropriate helper functions have been added to the
2401       GNU Objective-C runtime ABI, and are implemented by the GNU
2402       Objective-C runtime library shipped with GCC.
2403     * The Objective-C 2.0 fast enumeration syntax is supported in
2404       Objective-C. This is currently not yet available in Objective-C++.
2405       Fast enumeration requires support in the runtime, and such support
2406       has been added to the GNU Objective-C runtime library (shipped with
2407       GCC).
2408     * The Objective-C 2.0 @optional keyword is supported. It allows you
2409       to mark methods or properties in a protocol as optional as opposed
2410       to required.
2411     * The Objective-C 2.0 @package keyword is supported. It has currently
2412       the same effect as the @public keyword.
2413     * Objective-C 2.0 method attributes are supported. Currently the
2414       supported attributes are deprecated, sentinel, noreturn and format.
2415     * Objective-C 2.0 method argument attributes are supported. The most
2416       widely used attribute is unused, to mark an argument as unused in
2417       the implementation.
2418     * Objective-C 2.0 class and protocol attributes are supported.
2419       Currently the only supported attribute is deprecated.
2420     * Objective-C 2.0 class extensions are supported. A class extension
2421       has the same syntax as a category declaration with no category
2422       name, and the methods and properties declared in it are added
2423       directly to the main class. It is mostly used as an alternative to
2424       a category to add methods to a class without advertising them in
2425       the public headers, with the advantage that for class extensions
2426       the compiler checks that all the privately declared methods are
2427       actually implemented.
2428     * As a result of these enhancements, GCC can now be used to build
2429       Objective-C and Objective-C++ software that uses Foundation and
2430       other important system frameworks with the NeXT runtime on Darwin 9
2431       and Darwin 10 (OSX 10.5 and 10.6).
2432     * Many bugs in the compiler have been fixed in this release; in
2433       particular, LTO can now be used when compiling Objective-C and
2434       Objective-C++ and the parser is much more robust in dealing with
2435       invalid code.
2436
2437    Runtime Library (libobjc)
2438
2439     * The GNU Objective-C runtime library now defines the macro
2440       __GNU_LIBOBJC__ (with a value that is increased at every release
2441       where there is any change to the API) in objc/objc.h, making it
2442       easy to determine if the GNU Objective-C runtime library is being
2443       used, and if so, which version. Previous versions of the GNU
2444       Objective-C runtime library (and other Objective-C runtime
2445       libraries such as the Apple one) do not define this macro.
2446     * A new Objective-C 2.0 API, almost identical to the one implemented
2447       by the Apple Objective-C runtime, has been implemented in the GNU
2448       Objective-C runtime library. The new API hides the internals of
2449       most runtime structures but provides a more extensive set of
2450       functions to operate on them. It is much easier, for example, to
2451       create or modify classes at runtime. The new API also makes it
2452       easier to port software from Apple to GNU as almost no changes
2453       should be required. The old API is still supported for backwards
2454       compatibility; including the old objc/objc-api.h header file
2455       automatically selects the old API, while including the new
2456       objc/runtime.h header file automatically selects the new API.
2457       Support for the old API is being phased out and upgrading the
2458       software to use the new API is strongly recommended. To check for
2459       the availability of the new API, the __GNU_LIBOBJC__ macro can be
2460       used as older versions of the GNU Objective-C runtime library,
2461       which do not support the new API, do not define such a macro.
2462     * Runtime support for @synchronized has been added.
2463     * Runtime support for Objective-C 2.0 synthesized property accessors
2464       has been added.
2465     * Runtime support for Objective-C 2.0 fast enumeration has been
2466       added.
2467
2468New Targets and Target Specific Improvements
2469
2470  ARM
2471
2472     * GCC now supports the Cortex-M4 processor implementing the v7-em
2473       version of the architecture using the option -mcpu=cortex-m4.
2474     * Scheduling descriptions for the Cortex-M4, the Neon and the
2475       floating point units of the Cortex-A9 and a pipeline description
2476       for the Cortex-A5 have been added.
2477     * Synchronization primitives such as __sync_fetch_and_add and friends
2478       are now inlined for supported architectures rather than calling
2479       into a kernel helper function.
2480     * SSA loop prefetching is enabled by default for the Cortex-A9 at
2481       -O3.
2482     * Several improvements were committed to improve code generation for
2483       the ARM architecture including a rewritten implementation for load
2484       and store multiples.
2485     * Several enhancements were committed to improve SIMD code generation
2486       for NEON by adding support for widening instructions, misaligned
2487       loads and stores, vector conditionals and support for 64 bit
2488       arithmetic.
2489     * Support was added for the Faraday cores fa526, fa606te, fa626te,
2490       fmp626te, fmp626 and fa726te and can be used with the respective
2491       names as parameters to the -mcpu= option.
2492     * Basic support was added for Cortex-A15 and is available through
2493       -mcpu=cortex-a15.
2494     * GCC for AAPCS configurations now more closely adheres to the AAPCS
2495       specification by enabling -fstrict-volatile-bitfields by default.
2496
2497  IA-32/x86-64
2498
2499     * The new -fsplit-stack option permits programs to use a
2500       discontiguous stack. This is useful for threaded programs, in that
2501       it is no longer necessary to specify the maximum stack size when
2502       creating a thread. This feature is currently only implemented for
2503       32-bit and 64-bit x86 GNU/Linux targets.
2504     * Support for emitting profiler counter calls before function
2505       prologues. This is enabled via a new command-line option -mfentry.
2506     * Optimization for the Intel Core 2 processors is now available
2507       through the -march=core2 and -mtune=core2 options.
2508     * Support for Intel Core i3/i5/i7 processors is now available through
2509       the -march=corei7 and -mtune=corei7 options.
2510     * Support for Intel Core i3/i5/i7 processors with AVX is now
2511       available through the -march=corei7-avx and -mtune=corei7-avx
2512       options.
2513     * Support for AMD Bobcat (family 14) processors is now available
2514       through the -march=btver1 and -mtune=btver1 options.
2515     * Support for AMD Bulldozer (family 15) processors is now available
2516       through the -march=bdver1 and -mtune=bdver1 options.
2517     * The default setting (when not optimizing for size) for 32-bit
2518       GNU/Linux and Darwin x86 targets has been changed to
2519       -fomit-frame-pointer. The default can be reverted to
2520       -fno-omit-frame-pointer by configuring GCC with the
2521       --enable-frame-pointer configure option.
2522     * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support
2523       __float128 on 32-bit and 64-bit x86 targets.
2524     * AVX floating-point arithmetic can now be enabled by default at
2525       configure time with the new --with-fpmath=avx option.
2526     * The SSA loop prefetching pass is enabled when using -O3 when
2527       optimizing for CPUs where prefetching is beneficial (AMD CPUs newer
2528       than K6).
2529     * Support for TBM (Trailing Bit Manipulation) built-in functions and
2530       code generation is available via -mtbm.
2531     * Support for AMD's BMI (Bit Manipulation) built-in functions and
2532       code generation is available via -mbmi.
2533
2534  MicroBlaze
2535
2536     * Support has been added for the Xilinx MicroBlaze softcore processor
2537       (microblaze-elf) embedded target. This configurable processor is
2538       supported on several Xilinx Spartan and Virtex FPGAs.
2539
2540  MIPS
2541
2542     * GCC now supports the Loongson 3A processor. Its canonical -march=
2543       and -mtune= name is loongson3a.
2544
2545  MN10300 / AM33
2546
2547     * The inline assembly register constraint "A" has been renamed "c".
2548       This constraint is used to select a floating-point register that
2549       can be used as the destination of a multiply-accumulate
2550       instruction.
2551     * New inline assembly register constraints "A" and "D" have been
2552       added. These constraint letters resolve to all general registers
2553       when compiling for AM33, and resolve to address registers only or
2554       data registers only when compiling for MN10300.
2555     * The MDR register is represented in the compiler. One can access the
2556       register via the "z" constraint in inline assembly. It can be
2557       marked as clobbered or used as a local register variable via the
2558       "mdr" name. The compiler uses the RETF instruction if the function
2559       does not modify the MDR register, so it is important that inline
2560       assembly properly annotate any usage of the register.
2561
2562  PowerPC/PowerPC64
2563
2564     * GCC now supports the Applied Micro Titan processor with
2565       -mcpu=titan.
2566     * The -mrecip option has been added, which indicates whether the
2567       reciprocal and reciprocal square root instructions should be used.
2568     * The -mveclibabi=mass option can be used to enable the compiler to
2569       autovectorize mathematical functions using the Mathematical
2570       Acceleration Subsystem library.
2571     * The -msingle-pic-base option has been added, which instructs the
2572       compiler to avoid loading the PIC base register in function
2573       prologues. The PIC base register must be initialized by the runtime
2574       system.
2575     * The -mblock-move-inline-limit option has been added, which enables
2576       the user to control the maximum size of inlined memcpy calls and
2577       similar.
2578     * PowerPC64 GNU/Linux support for applications requiring a large TOC
2579       section has been improved. A new command-line option,
2580       -mcmodel=MODEL, controls this feature; valid values for MODEL are
2581       small, medium, or large.
2582     * The Altivec builtin functions vec_ld and vec_st have been modified
2583       to generate the Altivec memory instructions LVX and STVX, even if
2584       the -mvsx option is used. In the initial GCC 4.5 release, these
2585       builtin functions were changed to generate VSX memory reference
2586       instructions instead of Altivec memory instructions, but there are
2587       differences between the two instructions. If the VSX instruction
2588       set is available, you can now use the new builtin functions
2589       vec_vsx_ld and vec_vsx_st which always generates the VSX memory
2590       instructions.
2591     * The GCC compiler on AIX now defaults to a process layout with a
2592       larger data space allowing larger programs to be compiled.
2593     * The GCC long double type on AIX 6.1 and above has reverted to 64
2594       bit double precision, matching the AIX XL compiler default, because
2595       of missing C99 symbols required by the GCC runtime.
2596     * The default processor scheduling model and tuning for PowerPC64
2597       GNU/Linux and for AIX 6.1 and above now is POWER7.
2598     * Starting with GCC 4.6.1, vectors of type vector long long or vector
2599       long are passed and returned in the same method as other vectors
2600       with the VSX instruction set. Previously the GCC compiler did not
2601       adhere to the ABI for 128-bit vectors with 64-bit integer base
2602       types (PR 48857). This is also fixed in the GCC 4.5.4 release.
2603
2604  S/390, zSeries and System z9/z10, IBM zEnterprise z196
2605
2606     * Support for the zEnterprise z196 processor has been added. When
2607       using the -march=z196 option, the compiler will generate code
2608       making use of the following instruction facilities:
2609          + Conditional load/store
2610          + Distinct-operands
2611          + Floating-point-extension
2612          + Interlocked-access
2613          + Population-count
2614       The -mtune=z196 option avoids the compare and branch instructions
2615       as well as the load address instruction with an index register as
2616       much as possible and performs instruction scheduling appropriate
2617       for the new out-of-order pipeline architecture.
2618     * When using the -m31 -mzarch options the generated code still
2619       conforms to the 32-bit ABI but uses the general purpose registers
2620       as 64-bit registers internally. This requires a Linux kernel saving
2621       the whole 64-bit registers when doing a context switch. Kernels
2622       providing that feature indicate that by the 'highgprs' string in
2623       /proc/cpuinfo.
2624     * The SSA loop prefetching pass is enabled when using -O3.
2625
2626  SPARC
2627
2628     * GCC now supports the LEON series of SPARC V8 processors. The code
2629       generated by the compiler can either be tuned to it by means of the
2630       --with-tune=leon configure option and -mtune=leon compilation
2631       option, or the compiler can be built for the sparc-leon-{elf,linux}
2632       and sparc-leon3-{elf,linux} targets directly.
2633     * GCC has stopped sign/zero-extending parameter registers in the
2634       callee for functions taking parameters with sub-word size in 32-bit
2635       mode, since this is redundant with the specification of the ABI.
2636       GCC has never done so in 64-bit mode since this is also redundant.
2637     * The command line option -mfix-at697f has been added to enable the
2638       documented workaround for the single erratum of the Atmel AT697F
2639       processor.
2640
2641Operating Systems
2642
2643  Android
2644
2645     * GCC now supports the Bionic C library and provides a convenient way
2646       of building native libraries and applications for the Android
2647       platform. Refer to the documentation of the -mandroid and -mbionic
2648       options for details on building native code. At the moment, Android
2649       support is enabled only for ARM.
2650
2651  Darwin/Mac OS X
2652
2653     * General
2654          + Initial support for CFString types has been added.
2655            This allows GCC to build projects including the system Core
2656            Foundation frameworks. The GCC Objective-C family supports
2657            CFString "toll-free bridged" as per the Mac OS X system tools.
2658            CFString is also recognized in the context of format
2659            attributes and arguments (see the documentation for format
2660            attributes for limitations). At present, 8-bit character types
2661            are supported.
2662          + Object file size reduction.
2663            The Darwin zeroed memory allocators have been re-written to
2664            make more use of .zerofill sections. For non-debug code, this
2665            can reduce object file size significantly.
2666          + Objective-C family 64-bit support (NeXT ABI 2).
2667            Initial support has been added to support 64-bit Objective-C
2668            code using the Darwin/OS X native (NeXT) runtime. ABI version
2669            2 will be selected automatically when 64-bit code is built.
2670          + Objective-C family 32-bit ABI 1.
2671            For 32-bit code ABI 1 is also now also allowed. At present it
2672            must be selected manually using -fobjc-abi-version=1 where
2673            applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6).
2674     * x86 Architecture
2675          + The -mdynamic-no-pic option has been enabled.
2676            Code supporting -mdynamic-no-pic optimization has been added
2677            and is applicable to -m32 builds. The compiler bootstrap uses
2678            the option where appropriate.
2679          + The default value for -mtune= has been changed.
2680            Since Darwin systems are primarily Xeon, Core-2 or similar the
2681            default tuning has been changed to -mtune=core2.
2682          + Enable 128-bit long double (__float128) support on Darwin.
2683     * PPC Architecture
2684          + Darwin64 ABI.
2685            Several significant bugs have been fixed, such that GCC now
2686            produces code compatible with the Darwin64 PowerPC ABI.
2687          + libffi and boehm-gc.
2688            The Darwin ports of the libffi and boehm-gc libraries have
2689            been upgraded to include a Darwin64 implementation. This means
2690            that powerpc*-*-darwin9 platforms may now, for example, build
2691            Java applications with -m64 enabled.
2692          + Plug-in support has been enabled.
2693          + The -fsection-anchors option is now available although,
2694            presently, not heavily tested.
2695
2696  Solaris 2
2697
2698    New Features
2699
2700     * Support symbol versioning with the Sun linker.
2701     * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+.
2702     * Support thread-local storage (TLS) with the Sun assembler on
2703       Solaris 2/x86.
2704     * Support TLS on Solaris 8/9 if prerequisites are met.
2705     * Support COMDAT group with the GNU assembler and recent Sun linker.
2706     * Support the Sun assembler visibility syntax.
2707     * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp.
2708       -march=pentiumpro (Solaris 8/9).
2709     * Don't use SSE on Solaris 8/9 x86 by default.
2710     * Enable 128-bit long double (__float128) support on Solaris 2/x86.
2711
2712    ABI Change
2713
2714     * Change the ABI for returning 8-byte vectors like __m64 in MMX
2715       registers on Solaris 10+/x86 to match the Sun Studio 12.1+
2716       compilers. This is an incompatible change. If you use such types,
2717       you must either recompile all your code with the new compiler or
2718       use the new -mvect8-ret-in-mem option to remain compatible with
2719       previous versions of GCC and Sun Studio.
2720
2721  Windows x86/x86_64
2722
2723     * Initial support for decimal floating point.
2724     * Support for the __thiscall calling-convention.
2725     * Support for hot-patchable function prologues via the
2726       ms_hook_prologue attribute for x86_64 in addition to 32-bit x86.
2727     * Improvements of stack-probing and stack-allocation mechanisms.
2728     * Support of push/pop-macro pragma as preprocessor command.
2729       With #pragma push_macro("macro-name") the current definition of
2730       macro-name is saved and can be restored with #pragma
2731       pop_macro("macro-name") to its saved definition.
2732     * Enable 128-bit long double (__float128) support on MinGW and
2733       Cygwin.
2734
2735Other significant improvements
2736
2737  Installation changes
2738
2739     * An install-strip make target is provided that installs stripped
2740       executables, and may install libraries with unneeded or debugging
2741       sections stripped.
2742     * On Power7 systems, there is a potential problem if you build the
2743       GCC compiler with a host compiler using options that enable the VSX
2744       instruction set generation. If the host compiler has been patched
2745       so that the vec_ld and vec_st builtin functions generate Altivec
2746       memory instructions instead of VSX memory instructions, then you
2747       should be able to build the compiler with VSX instruction
2748       generation.
2749
2750Changes for GCC Developers
2751
2752   Note: these changes concern developers that develop GCC itself or
2753   software that integrates with GCC, such as plugins, and not the general
2754   GCC users.
2755     * The gengtype utility, which previously was internal to the GCC
2756       build process, has been enchanced to provide GC root information
2757       for plugins as necessary.
2758     * The old GC allocation interface of ggc_alloc and friends was
2759       replaced with a type-safe alternative.
2760
2761GCC 4.6.1
2762
2763   This is the [20]list of problem reports (PRs) from GCC's bug tracking
2764   system that are known to be fixed in the 4.6.1 release. This list might
2765   not be complete (that is, it is possible that some PRs that have been
2766   fixed are not listed here).
2767
2768GCC 4.6.2
2769
2770   This is the [21]list of problem reports (PRs) from GCC's bug tracking
2771   system that are known to be fixed in the 4.6.2 release. This list might
2772   not be complete (that is, it is possible that some PRs that have been
2773   fixed are not listed here).
2774
2775GCC 4.6.3
2776
2777   This is the [22]list of problem reports (PRs) from GCC's bug tracking
2778   system that are known to be fixed in the 4.6.3 release. This list might
2779   not be complete (that is, it is possible that some PRs that have been
2780   fixed are not listed here).
2781
2782GCC 4.6.4
2783
2784   This is the [23]list of problem reports (PRs) from GCC's bug tracking
2785   system that are known to be fixed in the 4.6.4 release. This list might
2786   not be complete (that is, it is possible that some PRs that have been
2787   fixed are not listed here).
2788
2789
2790    For questions related to the use of GCC, please consult these web
2791    pages and the [24]GCC manuals. If that fails, the
2792    [25]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2793    web pages and the development of GCC are welcome on our developer
2794    list at [26]gcc@gcc.gnu.org. All of [27]our lists have public
2795    archives.
2796
2797   Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and
2798   distribution of this entire article is permitted in any medium,
2799   provided this notice is preserved.
2800
2801   These pages are [29]maintained by the GCC team. Last modified
2802   2014-12-06[30].
2803
2804References
2805
2806   1. http://sourceware.org/bugzilla/show_bug.cgi?id=10401
2807   2. http://gcc.gnu.org/gcc-4.5/changes.html
2808   3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted
2809   4. http://gcc.gnu.org/gcc-4.6/porting_to.html
2810   5. http://gcc.gnu.org/projects/lto/whopr.pdf
2811   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
2812   7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html
2813   8. https://gcc.gnu.org/PR43145
2814   9. https://gcc.gnu.org/PR43680
2815  10. https://gcc.gnu.org/PR33558
2816  11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253
2817  12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x
2818  13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
2819  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races
2820  15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
2821  16. https://gcc.gnu.org/wiki/OOP
2822  17. https://gcc.gnu.org/wiki/Coarray
2823  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233
2824  19. http://golang.org/
2825  20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1
2826  21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2
2827  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3
2828  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4
2829  24. https://gcc.gnu.org/onlinedocs/
2830  25. mailto:gcc-help@gcc.gnu.org
2831  26. mailto:gcc@gcc.gnu.org
2832  27. https://gcc.gnu.org/lists.html
2833  28. http://www.fsf.org/
2834  29. https://gcc.gnu.org/about.html
2835  30. http://validator.w3.org/check/referer
2836======================================================================
2837http://gcc.gnu.org/gcc-4.5/index.html
2838
2839                             GCC 4.5 Release Series
2840
2841   Jul 2, 2012
2842
2843   The [1]GNU project and the GCC developers are pleased to announce the
2844   release of GCC 4.5.4.
2845
2846   This release is a bug-fix release, containing fixes for regressions in
2847   GCC 4.5.3 relative to previous releases of GCC.
2848
2849Release History
2850
2851   GCC 4.5.4
2852          Jul 2, 2012 ([2]changes)
2853
2854   GCC 4.5.3
2855          Apr 28, 2011 ([3]changes)
2856
2857   GCC 4.5.2
2858          Dec 16, 2010 ([4]changes)
2859
2860   GCC 4.5.1
2861          Jul 31, 2010 ([5]changes)
2862
2863   GCC 4.5.0
2864          April 14, 2010 ([6]changes)
2865
2866References and Acknowledgements
2867
2868   GCC used to stand for the GNU C Compiler, but since the compiler
2869   supports several other languages aside from C, it now stands for the
2870   GNU Compiler Collection.
2871
2872   A list of [7]successful builds is updated as new information becomes
2873   available.
2874
2875   The GCC developers would like to thank the numerous people that have
2876   contributed new features, improvements, bug fixes, and other changes as
2877   well as test results to GCC. This [8]amazing group of volunteers is
2878   what makes GCC successful.
2879
2880   For additional information about GCC please refer to the [9]GCC project
2881   web site or contact the [10]GCC development mailing list.
2882
2883   To obtain GCC please use [11]our mirror sites or [12]our SVN server.
2884
2885
2886    For questions related to the use of GCC, please consult these web
2887    pages and the [13]GCC manuals. If that fails, the
2888    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2889    web pages and the development of GCC are welcome on our developer
2890    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
2891    archives.
2892
2893   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
2894   distribution of this entire article is permitted in any medium,
2895   provided this notice is preserved.
2896
2897   These pages are [18]maintained by the GCC team. Last modified
2898   2014-06-28[19].
2899
2900References
2901
2902   1. http://www.gnu.org/
2903   2. http://gcc.gnu.org/gcc-4.5/changes.html
2904   3. http://gcc.gnu.org/gcc-4.5/changes.html
2905   4. http://gcc.gnu.org/gcc-4.5/changes.html
2906   5. http://gcc.gnu.org/gcc-4.5/changes.html
2907   6. http://gcc.gnu.org/gcc-4.5/changes.html
2908   7. http://gcc.gnu.org/gcc-4.5/buildstat.html
2909   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
2910   9. http://gcc.gnu.org/index.html
2911  10. mailto:gcc@gcc.gnu.org
2912  11. http://gcc.gnu.org/mirrors.html
2913  12. http://gcc.gnu.org/svn.html
2914  13. https://gcc.gnu.org/onlinedocs/
2915  14. mailto:gcc-help@gcc.gnu.org
2916  15. mailto:gcc@gcc.gnu.org
2917  16. https://gcc.gnu.org/lists.html
2918  17. http://www.fsf.org/
2919  18. https://gcc.gnu.org/about.html
2920  19. http://validator.w3.org/check/referer
2921======================================================================
2922http://gcc.gnu.org/gcc-4.5/changes.html
2923
2924                             GCC 4.5 Release Series
2925                        Changes, New Features, and Fixes
2926
2927Caveats
2928
2929     * GCC now requires the [1]MPC library in order to build. See the
2930       [2]prerequisites page for version requirements.
2931     * Support for a number of older systems and recently unmaintained or
2932       untested target ports of GCC has been declared obsolete in GCC 4.5.
2933       Unless there is activity to revive them, the next release of GCC
2934       will have their sources permanently removed.
2935       The following ports for individual systems on particular
2936       architectures have been obsoleted:
2937          + IRIX releases before 6.5 (mips-sgi-irix5*,
2938            mips-sgi-irix6.[0-4])
2939          + Solaris 7 (*-*-solaris2.7)
2940          + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*,
2941            alpha-dec-osf5.0*)
2942          + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions
2943            can be found in the [3]announcement.
2944       Support for the classic POWER architecture implemented in the
2945       original RIOS and RIOS2 processors of the old IBM RS/6000 product
2946       line has been obsoleted in the rs6000 port. This does not affect
2947       the new generation Power and PowerPC architectures.
2948     * Support has been removed for all the [4]configurations obsoleted in
2949       GCC 4.4.
2950     * Support has been removed for the protoize and unprotoize utilities,
2951       obsoleted in GCC 4.4.
2952     * Support has been removed for tuning for Itanium1 (Merced) variants.
2953       Note that code tuned for Itanium2 should also run correctly on
2954       Itanium1.
2955     * GCC now generates unwind info also for epilogues. DWARF debuginfo
2956       generated by GCC now uses more features of DWARF3 than before, and
2957       also some DWARF4 features. GDB older than 7.0 is not able to handle
2958       either of these, so to debug GCC 4.5 generated binaries or
2959       libraries GDB 7.0 or later is needed. You can disable use of DWARF4
2960       features with the -gdwarf-3 -gstrict-dwarf options, or use
2961       -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but
2962       epilogue unwind info is emitted unconditionally whenever unwind
2963       info is emitted.
2964     * On x86 targets, code containing floating-point calculations may run
2965       significantly slower when compiled with GCC 4.5 in strict C99
2966       conformance mode than they did with earlier GCC versions. This is
2967       due to stricter standard conformance of the compiler and can be
2968       avoided by using the option -fexcess-precision=fast; also see
2969       [5]below.
2970     * The function attribute noinline no longer prevents GCC from cloning
2971       the function. A new attribute noclone has been introduced for this
2972       purpose. Cloning a function means that it is duplicated and the new
2973       copy is specialized for certain contexts (for example when a
2974       parameter is a known constant).
2975
2976General Optimizer Improvements
2977
2978     * The -save-temps now takes an optional argument. The -save-temps and
2979       -save-temps=cwd switches write the temporary files in the current
2980       working directory based on the original source file. The
2981       -save-temps=obj switch will write files into the directory
2982       specified with the -o option, and the intermediate filenames are
2983       based on the output file. This will allow the user to get the
2984       compiler intermediate files when doing parallel builds without two
2985       builds of the same filename located in different directories from
2986       interfering with each other.
2987     * Debugging dumps are now created in the same directory as the object
2988       file rather than in the current working directory. This allows the
2989       user to get debugging dumps when doing parallel builds without two
2990       builds of the same filename interfering with each other.
2991     * GCC has been integrated with the [6]MPC library. This allows GCC to
2992       evaluate complex arithmetic at compile time [7]more accurately. It
2993       also allows GCC to evaluate calls to complex built-in math
2994       functions having constant arguments and replace them at compile
2995       time with their mathematically equivalent results. In doing so, GCC
2996       can generate correct results regardless of the math library
2997       implementation or floating point precision of the host platform.
2998       This also allows GCC to generate identical results regardless of
2999       whether one compiles in native or cross-compile configurations to a
3000       particular target. The following built-in functions take advantage
3001       of this new capability: cacos, cacosh, casin, casinh, catan,
3002       catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan,
3003       and ctanh. The float and long double variants of these functions
3004       (e.g. csinf and csinl) are also handled.
3005     * A new link-time optimizer has been added ([8]-flto). When this
3006       option is used, GCC generates a bytecode representation of each
3007       input file and writes it to specially-named sections in each object
3008       file. When the object files are linked together, all the function
3009       bodies are read from these named sections and instantiated as if
3010       they had been part of the same translation unit. This enables
3011       interprocedural optimizations to work across different files (and
3012       even different languages), potentially improving the performance of
3013       the generated code. To use the link-timer optimizer, -flto needs to
3014       be specified at compile time and during the final link. If the
3015       program does not require any symbols to be exported, it is possible
3016       to combine -flto and the experimental [9]-fwhopr with
3017       [10]-fwhole-program to allow the interprocedural optimizers to use
3018       more aggressive assumptions.
3019     * The automatic parallelization pass was enhanced to support
3020       parallelization of outer loops.
3021     * Automatic parallelization can be enabled as part of Graphite. In
3022       addition to -ftree-parallelize-loops=, specify
3023       -floop-parallelize-all to enable the Graphite-based optimization.
3024     * The infrastructure for optimizing based on [11]restrict qualified
3025       pointers has been rewritten and should result in code generation
3026       improvements. Optimizations based on restrict qualified pointers
3027       are now also available when using -fno-strict-aliasing.
3028     * There is a new optimization pass that attempts to change prototype
3029       of functions to avoid unused parameters, pass only relevant parts
3030       of structures and turn arguments passed by reference to arguments
3031       passed by value when possible. It is enabled by -O2 and above as
3032       well as -Os and can be manually invoked using the new command-line
3033       switch -fipa-sra.
3034     * GCC now optimize exception handling code. In particular cleanup
3035       regions that are proved to not have any effect are optimized out.
3036
3037New Languages and Language specific improvements
3038
3039  All languages
3040
3041     * The -fshow-column option is now on by default. This means error
3042       messages now have a column associated with them.
3043
3044  Ada
3045
3046     * Compilation of programs heavily using discriminated record types
3047       with variant parts has been sped up and generates more compact
3048       code.
3049     * Stack checking now works reasonably well on most plaforms. In some
3050       specific cases, stack overflows may still fail to be detected, but
3051       a compile-time warning will be issued for these cases.
3052
3053  C family
3054
3055     * If a header named in a #include directive is not found, the
3056       compiler exits immediately. This avoids a cascade of errors arising
3057       from declarations expected to be found in that header being
3058       missing.
3059     * A new built-in function __builtin_unreachable() has been added that
3060       tells the compiler that control will never reach that point. It may
3061       be used after asm statements that terminate by transferring control
3062       elsewhere, and in other places that are known to be unreachable.
3063     * The -Wlogical-op option now warns for logical expressions such as
3064       (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be
3065       mistakes. This option is disabled by default.
3066     * An asm goto feature has been added to allow asm statements that
3067       jump to C labels.
3068     * C++0x raw strings are supported for C++ and for C with -std=gnu99.
3069     * The deprecated attribute now takes an optional string argument, for
3070       example, __attribute__((deprecated("text string"))), that will be
3071       printed together with the deprecation warning.
3072
3073  C
3074
3075     * The -Wenum-compare option, which warns when comparing values of
3076       different enum types, now works for C. It formerly only worked for
3077       C++. This warning is enabled by -Wall. It may be avoided by using a
3078       type cast.
3079     * The -Wcast-qual option now warns about casts which are unsafe in
3080       that they permit const-correctness to be violated without further
3081       warnings. Specifically, it warns about cases where a qualifier is
3082       added when all the lower types are not const. For example, it warns
3083       about a cast from char ** to const char **.
3084     * The -Wc++-compat option is significantly improved. It issues new
3085       warnings for:
3086          + Using C++ reserved operator names as identifiers.
3087          + Conversions to enum types without explicit casts.
3088          + Using va_arg with an enum type.
3089          + Using different enum types in the two branches of ?:.
3090          + Using ++ or -- on a variable of enum type.
3091          + Using the same name as both a struct, union or enum tag and a
3092            typedef, unless the typedef refers to the tagged type itself.
3093          + Using a struct, union, or enum which is defined within another
3094            struct or union.
3095          + A struct field defined using a typedef if there is a field in
3096            the struct, or an enclosing struct, whose name is the typedef
3097            name.
3098          + Duplicate definitions at file scope.
3099          + Uninitialized const variables.
3100          + A global variable with an anonymous struct, union, or enum
3101            type.
3102          + Using a string constant to initialize a char array whose size
3103            is the length of the string.
3104     * The new -Wjump-misses-init option warns about cases where a goto or
3105       switch skips the initialization of a variable. This sort of branch
3106       is an error in C++ but not in C. This warning is enabled by
3107       -Wc++-compat.
3108     * GCC now ensures that a C99-conforming <stdint.h> is present on most
3109       targets, and uses information about the types in this header to
3110       implement the Fortran bindings to those types. GCC does not ensure
3111       the presence of such a header, and does not implement the Fortran
3112       bindings, on the following targets: NetBSD, VxWorks, VMS,
3113       SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF.
3114     * GCC now implements C90- and C99-conforming rules for constant
3115       expressions. This may cause warnings or errors for some code using
3116       expressions that can be folded to a constant but are not constant
3117       expressions as defined by ISO C.
3118     * All known target-independent C90 and C90 Amendment 1 conformance
3119       bugs, and all known target-independent C99 conformance bugs not
3120       related to floating point or extended identifiers, have been fixed.
3121     * The C decimal floating point support now includes support for the
3122       FLOAT_CONST_DECIMAL64 pragma.
3123     * The named address space feature from ISO/IEC TR 18037 is now
3124       supported. This is currently only implemented for the SPU
3125       processor.
3126
3127  C++
3128
3129     * Improved [12]experimental support for the upcoming C++0x ISO C++
3130       standard, including support for raw strings, lambda expressions and
3131       explicit type conversion operators.
3132     * When printing the name of a class template specialization, G++ will
3133       now omit any template arguments which come from default template
3134       arguments. This behavior (and the pretty-printing of function
3135       template specializations as template signature and arguments) can
3136       be disabled with the -fno-pretty-templates option.
3137     * Access control is now applied to typedef names used in a template,
3138       which may cause G++ to reject some ill-formed code that was
3139       accepted by earlier releases. The -fno-access-control option can be
3140       used as a temporary workaround until the code is corrected.
3141     * Compilation time for code that uses templates should now scale
3142       linearly with the number of instantiations rather than
3143       quadratically, as template instantiations are now looked up using
3144       hash tables.
3145     * Declarations of functions that look like builtin declarations of
3146       library functions are only considered to be redeclarations if they
3147       are declared with extern "C". This may cause problems with code
3148       that omits extern "C" on hand-written declarations of C library
3149       functions such as abort or memcpy. Such code is ill-formed, but was
3150       accepted by earlier releases.
3151     * Diagnostics that used to complain about passing non-POD types to
3152       ... or jumping past the declaration of a non-POD variable now check
3153       for triviality rather than PODness, as per C++0x.
3154     * In C++0x mode local and anonymous classes are now allowed as
3155       template arguments, and in declarations of variables and functions
3156       with linkage, so long as any such declaration that is used is also
3157       defined ([13]DR 757).
3158     * Labels may now have attributes, as has been permitted for a while
3159       in C. This is only permitted when the label definition and the
3160       attribute specifier is followed by a semicolon--i.e., the label
3161       applies to an empty statement. The only useful attribute for a
3162       label is unused.
3163     * G++ now implements [14]DR 176. Previously G++ did not support using
3164       the injected-class-name of a template base class as a type name,
3165       and lookup of the name found the declaration of the template in the
3166       enclosing scope. Now lookup of the name finds the
3167       injected-class-name, which can be used either as a type or as a
3168       template, depending on whether or not the name is followed by a
3169       template argument list. As a result of this change, some code that
3170       was previously accepted may be ill-formed because
3171         1. The injected-class-name is not accessible because it's from a
3172            private base, or
3173         2. The injected-class-name cannot be used as an argument for a
3174            template template parameter.
3175       In either of these cases, the code can be fixed by adding a
3176       nested-name-specifier to explicitly name the template. The first
3177       can be worked around with -fno-access-control; the second is only
3178       rejected with -pedantic.
3179     * A new standard mangling for SIMD vector types has been added, to
3180       avoid name clashes on systems with vectors of varying length. By
3181       default the compiler still uses the old mangling, but emits aliases
3182       with the new mangling on targets that support strong aliases. Users
3183       can switch over entirely to the new mangling with -fabi-version=4
3184       or -fabi-version=0. -Wabi will now warn about code that uses the
3185       old mangling.
3186     * The command-line option -ftemplate-depth-N is now written as
3187       -ftemplate-depth=N and the old form is deprecated.
3188     * Conversions between NULL and non-pointer types are now warned by
3189       default. The new option -Wno-conversion-null disables these
3190       warnings. Previously these warnings were only available when using
3191       -Wconversion explicitly.
3192
3193    Runtime Library (libstdc++)
3194
3195     * Improved experimental support for the upcoming ISO C++ standard,
3196       C++0x, including:
3197          + Support for <future>, <functional>, and <random>.
3198          + Existing facilities now exploit explicit operators and the
3199            newly implemented core C++0x features.
3200          + The header <cstdatomic> has been renamed to <atomic>.
3201     * An experimental [15]profile mode has been added. This is an
3202       implementation of many C++ standard library constructs with an
3203       additional analysis layer that gives performance improvement advice
3204       based on recognition of suboptimal usage patterns. For example,
3205#include <vector>
3206int main()
3207{
3208  std::vector<int> v;
3209  for (int k = 0; k < 1024; ++k)
3210    v.insert(v.begin(), k);
3211}
3212
3213       When instrumented via the profile mode, can return suggestions
3214       about the initial size and choice of the container used as follows:
3215vector-to-list: improvement = 5: call stack = 0x804842c ...
3216    : advice = change std::vector to std::list
3217vector-size: improvement = 3: call stack = 0x804842c ...
3218    : advice = change initial container size from 0 to 1024
3219
3220       These constructs can be substituted for the normal libstdc++
3221       constructs on a piecemeal basis, or all existing components can be
3222       transformed via the -D_GLIBCXX_PROFILE macro.
3223     * [16]Support for decimal floating-point arithmetic (aka ISO C++ TR
3224       24733) has been added. This support is in header file
3225       <decimal/decimal>, uses namespace std::decimal, and includes
3226       classes decimal32, decimal64, and decimal128.
3227     * Sources have been audited for application of function attributes
3228       nothrow, const, pure, and noreturn.
3229     * Python pretty-printers have been added for many standard library
3230       components that simplify the internal representation and present a
3231       more intuitive view of components when used with
3232       appropriately-advanced versions of GDB. For more information,
3233       please consult the more [17]detailed description.
3234     * The default behavior for comparing typeinfo names has changed, so
3235       in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero.
3236     * The new -static-libstdc++ option directs g++ to link the C++
3237       library statically, even if the default would normally be to link
3238       it dynamically.
3239
3240  Fortran
3241
3242     * The COMMON default padding has been changed - instead of adding the
3243       padding before a variable it is now added afterwards, which
3244       increases the compatibility with other vendors and helps to obtain
3245       the correct output in some cases. Cf. also the -falign-commons
3246       option ([18]added in 4.4).
3247     * The -finit-real= option now also supports the value snan for
3248       signalling not-a-number; to be effective, one additionally needs to
3249       enable trapping (e.g. via -ffpe-trap=). Note: Compile-time
3250       optimizations can turn a signalling NaN into a quiet one.
3251     * The new option -fcheck= has been added with the options bounds,
3252       array-temps, do, pointer, and recursive. The bounds and array-temps
3253       options are equivalent to -fbounds-check and
3254       -fcheck-array-temporaries. The do option checks for invalid
3255       modification of loop iteration variables, and the recursive option
3256       tests for recursive calls to subroutines/functions which are not
3257       marked as recursive. With pointer pointer association checks in
3258       calls are performed; however, neither undefined pointers nor
3259       pointers in expressions are handled. Using -fcheck=all enables all
3260       these run-time checks.
3261     * The run-time checking -fcheck=bounds now warns about invalid string
3262       lengths of character dummy arguments. Additionally, more
3263       compile-time checks have been added.
3264     * The new option [19]-fno-protect-parens has been added; if set, the
3265       compiler may reorder REAL and COMPLEX expressions without regard to
3266       parentheses.
3267     * GNU Fortran no longer links against libgfortranbegin. As before,
3268       MAIN__ (assembler symbol name) is the actual Fortran main program,
3269       which is invoked by the main function. However, main is now
3270       generated and put in the same object file as MAIN__. For the time
3271       being, libgfortranbegin still exists for backward compatibility.
3272       For details see the new [20]Mixed-Language Programming chapter in
3273       the manual.
3274     * The I/O library was restructured for performance and cleaner code.
3275     * Array assignments and WHERE are now run in parallel when OpenMP's
3276       WORKSHARE is used.
3277     * The experimental option -fwhole-file was added. The option allows
3278       whole-file checking of procedure arguments and allows for better
3279       optimizations. It can also be used with -fwhole-program, which is
3280       now also supported in gfortran.
3281     * More Fortran 2003 and Fortran 2008 mathematical functions can now
3282       be used as initialization expressions.
3283     * Some extended attributes such as STDCALL are now supported via the
3284       [21]GCC$ compiler directive.
3285     * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN
3286       intrinsic behaves now as if zero were always positive.
3287     * For legacy compatibiliy: On Cygwin and MinGW, the special files
3288       CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now
3289       supported.
3290     * Fortran 2003 support has been extended:
3291          + Procedure-pointer function results and procedure-pointer
3292            components (including PASS),
3293          + allocatable scalars (experimental),
3294          + DEFERRED type-bound procedures,
3295          + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements
3296            have been implemented.
3297          + The ALLOCATE statement supports type-specs and the SOURCE=
3298            argument.
3299          + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC
3300            type-bound procedure (i.e. as type-bound operators).
3301          + Rounding (ROUND=, RZ, ...) for output is now supported.
3302          + The INT_FAST{8,16,32,64,128}_T kind type parameters of the
3303            intrinsic module ISO_C_BINDING are now supported, except for
3304            the targets listed above as ones where GCC does not have
3305            <stdint.h> type information.
3306          + Extensible derived types with type-bound procedure or
3307            procedure pointer with PASS attribute now have to use CLASS in
3308            line with the Fortran 2003 standard; the workaround to use
3309            TYPE is no longer supported.
3310          + [22]Experimental, incomplete support for polymorphism,
3311            including CLASS, SELECT TYPE and dynamic dispatch of
3312            type-bound procedure calls. Some features do not work yet such
3313            as unlimited polymorphism (CLASS(*)).
3314     * Fortran 2008 support has been extended:
3315          + The OPEN statement now supports the NEWUNIT= option, which
3316            returns a unique file unit, thus preventing inadvertent use of
3317            the same unit in different parts of the program.
3318          + Support for unlimited format items has been added.
3319          + The INT{8,16,32} and REAL{32,64,128} kind type parameters of
3320            the intrinsic module ISO_FORTRAN_ENV are now supported.
3321          + Using complex arguments with TAN, SINH, COSH, TANH, ASIN,
3322            ACOS, and ATAN is now possible; the functions ASINH, ACOSH,
3323            and ATANH have been added (for real and complex arguments) and
3324            ATAN(Y,X) is now an alias for ATAN2(Y,X).
3325          + The BLOCK construct has been implemented.
3326
3327New Targets and Target Specific Improvements
3328
3329  AIX
3330
3331     * Full cross-toolchain support now available with GNU Binutils
3332
3333  ARM
3334
3335     * GCC now supports the Cortex-M0 and Cortex-A5 processors.
3336     * GCC now supports the ARM v7E-M architecture.
3337     * GCC now supports VFPv4-based FPUs and FPUs with
3338       single-precision-only VFP.
3339     * GCC has many improvements to optimization for other ARM processors,
3340       including scheduling support for the integer pipeline on Cortex-A9.
3341     * GCC now supports the IEEE 754-2008 half-precision floating-point
3342       type, and a variant ARM-specific half-precision type. This type is
3343       specified using __fp16, with the layout determined by
3344       -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and
3345       VFPv4 half-precision instructions will be used.
3346     * GCC now supports the variant of AAPCS that uses VFP registers for
3347       parameter passing and return values.
3348
3349  AVR
3350
3351     * The -mno-tablejump option has been removed because it has the same
3352       effect as the -fno-jump-tables option.
3353     * Added support for these new AVR devices:
3354          + ATmega8U2
3355          + ATmega16U2
3356          + ATmega32U2
3357
3358  IA-32/x86-64
3359
3360     * GCC now will set the default for -march= based on the configure
3361       target.
3362     * GCC now supports handling floating-point excess precision arising
3363       from use of the x87 floating-point unit in a way that conforms to
3364       ISO C99. This is enabled with -fexcess-precision=standard and with
3365       standards conformance options such as -std=c99, and may be disabled
3366       using -fexcess-precision=fast.
3367     * Support for the Intel Atom processor is now available through the
3368       -march=atom and -mtune=atom options.
3369     * A new -mcrc32 option is now available to enable crc32 intrinsics.
3370     * A new -mmovbe option is now available to enable GCC to use the
3371       movbe instruction to implement __builtin_bswap32 and
3372       __builtin_bswap64.
3373     * SSE math now can be enabled by default at configure time with the
3374       new --with-fpmath=sse option.
3375     * There is a new intrinsic header file, <x86intrin.h>. It should be
3376       included before using any IA-32/x86-64 intrinsics.
3377     * Support for the XOP, FMA4, and LWP instruction sets for the AMD
3378       Orochi processors are now available with the -mxop, -mfma4, and
3379       -mlwp options.
3380     * The -mabm option enables GCC to use the popcnt and lzcnt
3381       instructions on AMD processors.
3382     * The -mpopcnt option enables GCC to use the popcnt instructions on
3383       both AMD and Intel processors.
3384
3385  M68K/ColdFire
3386
3387     * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x
3388       and 5441x devices.
3389     * GCC now supports thread-local storage (TLS) on M68K and ColdFire
3390       processors.
3391
3392  MeP
3393
3394   Support has been added for the Toshiba Media embedded Processor (MeP,
3395   or mep-elf) embedded target.
3396
3397  MIPS
3398
3399     * GCC now supports MIPS 1004K processors.
3400     * GCC can now be configured with options --with-arch-32,
3401       --with-arch-64, --with-tune-32 and --with-tune-64 to control the
3402       default optimization separately for 32-bit and 64-bit modes.
3403     * MIPS targets now support an alternative _mcount interface, in which
3404       register $12 points to the function's save slot for register $31.
3405       This interface is selected by the -mcount-ra-address option; see
3406       the documentation for more details.
3407     * GNU/Linux targets can now generate read-only .eh_frame sections.
3408       This optimization requires GNU binutils 2.20 or above, and is only
3409       available if GCC is configured with a suitable version of binutils.
3410     * GNU/Linux targets can now attach special relocations to indirect
3411       calls, so that the linker can turn them into direct jumps or
3412       branches. This optimization requires GNU binutils 2.20 or later,
3413       and is automatically selected if GCC is configured with an
3414       appropriate version of binutils. It can be explicitly enabled or
3415       disabled using the -mrelax-pic-calls command-line option.
3416     * GCC now generates more heavily-optimized atomic operations on
3417       Octeon processors.
3418     * MIPS targets now support the -fstack-protector option.
3419     * GCC now supports an -msynci option, which specifies that synci is
3420       enough to flush the instruction cache, without help from the
3421       operating system. GCC uses this information to optimize
3422       automatically-generated cache flush operations, such as those used
3423       for nested functions in C. There is also a --with-synci
3424       configure-time option, which makes -msynci the default.
3425     * GCC supports four new function attributes for interrupt handlers:
3426       interrupt, use_shadow_register_set, keep_interrupts_masked and
3427       use_debug_exception_return. See the documentation for more details
3428       about these attributes.
3429
3430  RS/6000 (POWER/PowerPC)
3431
3432     * GCC now supports the Power ISA 2.06, which includes the VSX
3433       instructions that add vector 64-bit floating point support, new
3434       population count instructions, and conversions between floating
3435       point and unsigned types.
3436     * Support for the power7 processor is now available through the
3437       -mcpu=power7 and -mtune=power7.
3438     * GCC will now vectorize loops that contain simple math functions
3439       like copysign when generating code for altivec or VSX targets.
3440     * Support for the A2 processor is now available through the -mcpu=a2
3441       and -mtune=a2 options.
3442     * Support for the 476 processor is now available through the
3443       -mcpu={476,476fp} and -mtune={476,476fp} options.
3444     * Support for the e500mc64 processor is now available through the
3445       -mcpu=e500mc64 and -mtune=e500mc64 options.
3446     * GCC can now be configured with options --with-cpu-32,
3447       --with-cpu-64, --with-tune-32 and --with-tune-64 to control the
3448       default optimization separately for 32-bit and 64-bit modes.
3449     * Starting with GCC 4.5.4, vectors of type vector long long or vector
3450       long are passed and returned in the same method as other vectors
3451       with the VSX instruction set. Previously the GCC compiler did not
3452       adhere to the ABI for 128-bit vectors with 64-bit integer base
3453       types (PR 48857). This is also fixed in the GCC 4.6.1 release.
3454
3455  RX
3456
3457   Support has been added for the Renesas RX Processor (rx-elf) target.
3458
3459Operating Systems
3460
3461  Windows (Cygwin and MinGW)
3462
3463     * GCC now installs all the major language runtime libraries as DLLs
3464       when configured with the --enable-shared option.
3465     * GCC now makes use of the new support for aligned common variables
3466       in versions of binutils >= 2.20 to fix bugs in the support for SSE
3467       data types.
3468     * Improvements to the libffi support library increase the reliability
3469       of code generated by GCJ on all Windows platforms. Libgcj is
3470       enabled by default for the first time.
3471     * Libtool improvements simplify installation by placing the generated
3472       DLLs in the correct binaries directory.
3473     * Numerous other minor bugfixes and improvements, and substantial
3474       enhancements to the Fortran language support library.
3475
3476   >
3477
3478Other significant improvements
3479
3480  Plugins
3481
3482     * It is now possible to extend the compiler without having to modify
3483       its source code. A new option -fplugin=file.so tells GCC to load
3484       the shared object file.so and execute it as part of the compiler.
3485       The internal documentation describes the details on how plugins can
3486       interact with the compiler.
3487
3488  Installation changes
3489
3490     * The move to newer autotools changed default installation
3491       directories and switches to control them: The --with-datarootdir,
3492       --with-docdir, --with-pdfdir, and --with-htmldir switches are not
3493       used any more. Instead, you can now use --datarootdir, --docdir,
3494       --htmldir, and --pdfdir. The default installation directories have
3495       changed as follows according to the GNU Coding Standards:
3496
3497       datarootdir read-only architecture-independent data root [PREFIX/share]
3498       localedir   locale-specific message catalogs [DATAROOTDIR/locale]
3499       docdir      documentation root [DATAROOTDIR/doc/PACKAGE]
3500       htmldir     html documentation [DOCDIR]
3501       dvidir      dvi documentation [DOCDIR]
3502       pdfdir      pdf documentation [DOCDIR]
3503       psdir       ps documentation [DOCDIR]
3504       The following variables have new default values:
3505
3506       datadir read-only architecture-independent data [DATAROOTDIR]
3507       infodir info documentation [DATAROOTDIR/info]
3508       mandir  man documentation [DATAROOTDIR/man]
3509
3510GCC 4.5.1
3511
3512   This is the [23]list of problem reports (PRs) from GCC's bug tracking
3513   system that are known to be fixed in the 4.5.1 release. This list might
3514   not be complete (that is, it is possible that some PRs that have been
3515   fixed are not listed here).
3516
3517  All languages
3518
3519     * GCC's new link-time optimizer ([24]-flto) now also works on a few
3520       non-ELF targets:
3521          + Cygwin (*-cygwin*)
3522          + MinGW (*-mingw*)
3523          + Darwin on x86-64 (x86_64-apple-darwin*)
3524       LTO is not enabled by default for these targets. To enable LTO, you
3525       should configure with the --enable-lto option.
3526
3527GCC 4.5.2
3528
3529   This is the [25]list of problem reports (PRs) from GCC's bug tracking
3530   system that are known to be fixed in the 4.5.2 release. This list might
3531   not be complete (that is, it is possible that some PRs that have been
3532   fixed are not listed here).
3533
3534GCC 4.5.3
3535
3536   This is the [26]list of problem reports (PRs) from GCC's bug tracking
3537   system that are known to be fixed in the 4.5.3 release. This list might
3538   not be complete (that is, it is possible that some PRs that have been
3539   fixed are not listed here).
3540
3541   On the PowerPC compiler, the Altivec builtin functions vec_ld and
3542   vec_st have been modified to generate the Altivec memory instructions
3543   LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5
3544   release, these builtin functions were changed to generate VSX memory
3545   reference instructions instead of Altivec memory instructions, but
3546   there are differences between the two instructions. If the VSX
3547   instruction set is available, you can now use the new builtin functions
3548   vec_vsx_ld and vec_vsx_st which always generates the VSX memory
3549   instructions.
3550
3551GCC 4.5.4
3552
3553   This is the [27]list of problem reports (PRs) from GCC's bug tracking
3554   system that are known to be fixed in the 4.5.4 release. This list might
3555   not be complete (that is, it is possible that some PRs that have been
3556   fixed are not listed here).
3557
3558
3559    For questions related to the use of GCC, please consult these web
3560    pages and the [28]GCC manuals. If that fails, the
3561    [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3562    web pages and the development of GCC are welcome on our developer
3563    list at [30]gcc@gcc.gnu.org. All of [31]our lists have public
3564    archives.
3565
3566   Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and
3567   distribution of this entire article is permitted in any medium,
3568   provided this notice is preserved.
3569
3570   These pages are [33]maintained by the GCC team. Last modified
3571   2014-12-06[34].
3572
3573References
3574
3575   1. http://www.multiprecision.org/
3576   2. https://gcc.gnu.org/install/prerequisites.html
3577   3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html
3578   4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted
3579   5. http://gcc.gnu.org/gcc-4.5/changes.html#x86
3580   6. http://www.multiprecision.org/
3581   7. https://gcc.gnu.org/PR30789
3582   8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
3583   9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802
3584  10. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800
3585  11. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html
3586  12. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html
3587  13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757
3588  14. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176
3589  15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html
3590  16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733
3591  17. http://sourceware.org/gdb/wiki/STLSupport
3592  18. http://gcc.gnu.org/gcc-4.4/changes.html
3593  19. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
3594  20. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html
3595  21. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html
3596  22. https://gcc.gnu.org/wiki/OOP
3597  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1
3598  24. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
3599  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2
3600  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3
3601  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4
3602  28. https://gcc.gnu.org/onlinedocs/
3603  29. mailto:gcc-help@gcc.gnu.org
3604  30. mailto:gcc@gcc.gnu.org
3605  31. https://gcc.gnu.org/lists.html
3606  32. http://www.fsf.org/
3607  33. https://gcc.gnu.org/about.html
3608  34. http://validator.w3.org/check/referer
3609======================================================================
3610http://gcc.gnu.org/gcc-4.4/index.html
3611
3612                             GCC 4.4 Release Series
3613
3614   March 13, 2012
3615
3616   The [1]GNU project and the GCC developers are pleased to announce the
3617   release of GCC 4.4.7.
3618
3619   This release is a bug-fix release, containing fixes for regressions in
3620   GCC 4.4.6 relative to previous releases of GCC.
3621
3622Release History
3623
3624   GCC 4.4.7
3625          March 13, 2012 ([2]changes)
3626
3627   GCC 4.4.6
3628          April 16, 2011 ([3]changes)
3629
3630   GCC 4.4.5
3631          October 1, 2010 ([4]changes)
3632
3633   GCC 4.4.4
3634          April 29, 2010 ([5]changes)
3635
3636   GCC 4.4.3
3637          January 21, 2010 ([6]changes)
3638
3639   GCC 4.4.2
3640          October 15, 2009 ([7]changes)
3641
3642   GCC 4.4.1
3643          July 22, 2009 ([8]changes)
3644
3645   GCC 4.4.0
3646          April 21, 2009 ([9]changes)
3647
3648References and Acknowledgements
3649
3650   GCC used to stand for the GNU C Compiler, but since the compiler
3651   supports several other languages aside from C, it now stands for the
3652   GNU Compiler Collection.
3653
3654   A list of [10]successful builds is updated as new information becomes
3655   available.
3656
3657   The GCC developers would like to thank the numerous people that have
3658   contributed new features, improvements, bug fixes, and other changes as
3659   well as test results to GCC. This [11]amazing group of volunteers is
3660   what makes GCC successful.
3661
3662   For additional information about GCC please refer to the [12]GCC
3663   project web site or contact the [13]GCC development mailing list.
3664
3665   To obtain GCC please use [14]our mirror sites or [15]our SVN server.
3666
3667
3668    For questions related to the use of GCC, please consult these web
3669    pages and the [16]GCC manuals. If that fails, the
3670    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3671    web pages and the development of GCC are welcome on our developer
3672    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
3673    archives.
3674
3675   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
3676   distribution of this entire article is permitted in any medium,
3677   provided this notice is preserved.
3678
3679   These pages are [21]maintained by the GCC team. Last modified
3680   2014-06-28[22].
3681
3682References
3683
3684   1. http://www.gnu.org/
3685   2. http://gcc.gnu.org/gcc-4.4/changes.html
3686   3. http://gcc.gnu.org/gcc-4.4/changes.html
3687   4. http://gcc.gnu.org/gcc-4.4/changes.html
3688   5. http://gcc.gnu.org/gcc-4.4/changes.html
3689   6. http://gcc.gnu.org/gcc-4.4/changes.html
3690   7. http://gcc.gnu.org/gcc-4.4/changes.html
3691   8. http://gcc.gnu.org/gcc-4.4/changes.html
3692   9. http://gcc.gnu.org/gcc-4.4/changes.html
3693  10. http://gcc.gnu.org/gcc-4.4/buildstat.html
3694  11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
3695  12. http://gcc.gnu.org/index.html
3696  13. mailto:gcc@gcc.gnu.org
3697  14. http://gcc.gnu.org/mirrors.html
3698  15. http://gcc.gnu.org/svn.html
3699  16. https://gcc.gnu.org/onlinedocs/
3700  17. mailto:gcc-help@gcc.gnu.org
3701  18. mailto:gcc@gcc.gnu.org
3702  19. https://gcc.gnu.org/lists.html
3703  20. http://www.fsf.org/
3704  21. https://gcc.gnu.org/about.html
3705  22. http://validator.w3.org/check/referer
3706======================================================================
3707http://gcc.gnu.org/gcc-4.4/changes.html
3708
3709                             GCC 4.4 Release Series
3710                        Changes, New Features, and Fixes
3711
3712   The latest release in the 4.4 release series is [1]GCC 4.4.7.
3713
3714Caveats
3715
3716     * __builtin_stdarg_start has been completely removed from GCC.
3717       Support for <varargs.h> had been deprecated since GCC 4.0. Use
3718       __builtin_va_start as a replacement.
3719     * Some of the errors issued by the C++ front end that could be
3720       downgraded to warnings in previous releases by using -fpermissive
3721       are now warnings by default. They can be converted into errors by
3722       using -pedantic-errors.
3723     * Use of the cpp assertion extension will now emit a warning when
3724       -Wdeprecated or -pedantic is used. This extension has been
3725       deprecated for many years, but never warned about.
3726     * Packed bit-fields of type char were not properly bit-packed on many
3727       targets prior to GCC 4.4. On these targets, the fix in GCC 4.4
3728       causes an ABI change. For example there is no longer a 4-bit
3729       padding between field a and b in this structure:
3730    struct foo
3731    {
3732      char a:4;
3733      char b:8;
3734    } __attribute__ ((packed));
3735       There is a new warning to help identify fields that are affected:
3736    foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4
3737       The warning can be disabled with -Wno-packed-bitfield-compat.
3738     * On ARM EABI targets, the C++ mangling of the va_list type has been
3739       changed to conform to the current revision of the EABI. This does
3740       not affect the libstdc++ library included with GCC.
3741     * The SCOUNT and POS bits of the MIPS DSP control register are now
3742       treated as global. Previous versions of GCC treated these fields as
3743       call-clobbered instead.
3744     * The MIPS port no longer recognizes the h asm constraint. It was
3745       necessary to remove this constraint in order to avoid generating
3746       unpredictable code sequences.
3747       One of the main uses of the h constraint was to extract the high
3748       part of a multiplication on 64-bit targets. For example:
3749    asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y));
3750       You can now achieve the same effect using 128-bit types:
3751    typedef unsigned int uint128_t __attribute__((mode(TI)));
3752    result = ((uint128_t) x * y) >> 64;
3753       The second sequence is better in many ways. For example, if x and y
3754       are constants, the compiler can perform the multiplication at
3755       compile time. If x and y are not constants, the compiler can
3756       schedule the runtime multiplication better than it can schedule an
3757       asm statement.
3758     * Support for a number of older systems and recently unmaintained or
3759       untested target ports of GCC has been declared obsolete in GCC 4.4.
3760       Unless there is activity to revive them, the next release of GCC
3761       will have their sources permanently removed.
3762       The following ports for individual systems on particular
3763       architectures have been obsoleted:
3764          + Generic a.out on IA32 and m68k (i[34567]86-*-aout*,
3765            m68k-*-aout*)
3766          + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*,
3767            armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*,
3768            sh-*-*). This does not affect other more specific targets
3769            using the COFF object format on those architectures, or the
3770            more specific H8300 and SH targets (h8300-*-rtems*,
3771            h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*,
3772            sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks).
3773          + 2BSD on PDP-11 (pdp11-*-bsd)
3774          + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*,
3775            powerpc-ibm-aix4.[12]*)
3776          + Tuning support for Itanium1 (Merced) variants. Note that code
3777            tuned for Itanium2 should also run correctly on Itanium1.
3778     * The protoize and unprotoize utilities have been obsoleted and will
3779       be removed in GCC 4.5. These utilities have not been installed by
3780       default since GCC 3.0.
3781     * Support has been removed for all the [2]configurations obsoleted in
3782       GCC 4.3.
3783     * Unknown -Wno-* options are now silently ignored by GCC if no other
3784       diagnostics are issued. If other diagnostics are issued, then GCC
3785       warns about the unknown options.
3786     * More information on porting to GCC 4.4 from previous versions of
3787       GCC can be found in the [3]porting guide for this release.
3788
3789General Optimizer Improvements
3790
3791     * A new command-line switch -findirect-inlining has been added. When
3792       turned on it allows the inliner to also inline indirect calls that
3793       are discovered to have known targets at compile time thanks to
3794       previous inlining.
3795     * A new command-line switch -ftree-switch-conversion has been added.
3796       This new pass turns simple initializations of scalar variables in
3797       switch statements into initializations from a static array, given
3798       that all the values are known at compile time and the ratio between
3799       the new array size and the original switch branches does not exceed
3800       the parameter --param switch-conversion-max-branch-ratio (default
3801       is eight).
3802     * A new command-line switch -ftree-builtin-call-dce has been added.
3803       This optimization eliminates unnecessary calls to certain builtin
3804       functions when the return value is not used, in cases where the
3805       calls can not be eliminated entirely because the function may set
3806       errno. This optimization is on by default at -O2 and above.
3807     * A new command-line switch -fconserve-stack directs the compiler to
3808       minimize stack usage even if it makes the generated code slower.
3809       This affects inlining decisions.
3810     * When the assembler supports it, the compiler will now emit unwind
3811       information using assembler .cfi directives. This makes it possible
3812       to use such directives in inline assembler code. The new option
3813       -fno-dwarf2-cfi-asm directs the compiler to not use .cfi
3814       directives.
3815     * The [4]Graphite branch has been merged. This merge has brought in a
3816       new framework for loop optimizations based on a polyhedral
3817       intermediate representation. These optimizations apply to all the
3818       languages supported by GCC. The following new code transformations
3819       are available in GCC 4.4:
3820          + -floop-interchange performs loop interchange transformations
3821            on loops. Interchanging two nested loops switches the inner
3822            and outer loops. For example, given a loop like:
3823          DO J = 1, M
3824            DO I = 1, N
3825              A(J, I) = A(J, I) * C
3826            ENDDO
3827          ENDDO
3828
3829            loop interchange will transform the loop as if the user had
3830            written:
3831          DO I = 1, N
3832            DO J = 1, M
3833              A(J, I) = A(J, I) * C
3834            ENDDO
3835          ENDDO
3836
3837            which can be beneficial when N is larger than the caches,
3838            because in Fortran, the elements of an array are stored in
3839            memory contiguously by column, and the original loop iterates
3840            over rows, potentially creating at each access a cache miss.
3841          + -floop-strip-mine performs loop strip mining transformations
3842            on loops. Strip mining splits a loop into two nested loops.
3843            The outer loop has strides equal to the strip size and the
3844            inner loop has strides of the original loop within a strip.
3845            For example, given a loop like:
3846          DO I = 1, N
3847            A(I) = A(I) + C
3848          ENDDO
3849
3850            loop strip mining will transform the loop as if the user had
3851            written:
3852          DO II = 1, N, 4
3853            DO I = II, min (II + 3, N)
3854              A(I) = A(I) + C
3855            ENDDO
3856          ENDDO
3857
3858          + -floop-block performs loop blocking transformations on loops.
3859            Blocking strip mines each loop in the loop nest such that the
3860            memory accesses of the element loops fit inside caches. For
3861            example, given a loop like:
3862          DO I = 1, N
3863            DO J = 1, M
3864              A(J, I) = B(I) + C(J)
3865            ENDDO
3866          ENDDO
3867
3868            loop blocking will transform the loop as if the user had
3869            written:
3870          DO II = 1, N, 64
3871            DO JJ = 1, M, 64
3872              DO I = II, min (II + 63, N)
3873                DO J = JJ, min (JJ + 63, M)
3874                  A(J, I) = B(I) + C(J)
3875                ENDDO
3876              ENDDO
3877            ENDDO
3878          ENDDO
3879
3880            which can be beneficial when M is larger than the caches,
3881            because the innermost loop will iterate over a smaller amount
3882            of data that can be kept in the caches.
3883     * A new register allocator has replaced the old one. It is called
3884       integrated register allocator (IRA) because coalescing, register
3885       live range splitting, and hard register preferencing are done
3886       on-the-fly during coloring. It also has better integration with the
3887       reload pass. IRA is a regional register allocator which uses modern
3888       Chaitin-Briggs coloring instead of Chow's priority coloring used in
3889       the old register allocator. More info about IRA internals and
3890       options can be found in the GCC manuals.
3891     * A new instruction scheduler and software pipeliner, based on the
3892       selective scheduling approach, has been added. The new pass
3893       performs instruction unification, register renaming, substitution
3894       through register copies, and speculation during scheduling. The
3895       software pipeliner is able to pipeline non-countable loops. The new
3896       pass is targeted at scheduling-eager in-order platforms. In GCC 4.4
3897       it is available for the Intel Itanium platform working by default
3898       as the second scheduling pass (after register allocation) at the
3899       -O3 optimization level.
3900     * When using -fprofile-generate with a multi-threaded program, the
3901       profile counts may be slightly wrong due to race conditions. The
3902       new -fprofile-correction option directs the compiler to apply
3903       heuristics to smooth out the inconsistencies. By default the
3904       compiler will give an error message when it finds an inconsistent
3905       profile.
3906     * The new -fprofile-dir=PATH option permits setting the directory
3907       where profile data files are stored when using -fprofile-generate
3908       and friends, and the directory used when reading profile data files
3909       using -fprofile-use and friends.
3910
3911New warning options
3912
3913     * The new -Wframe-larger-than=NUMBER option directs GCC to emit a
3914       warning if any stack frame is larger than NUMBER bytes. This may be
3915       used to help ensure that code fits within a limited amount of stack
3916       space.
3917     * The command-line option -Wlarger-than-N is now written as
3918       -Wlarger-than=N and the old form is deprecated.
3919     * The new -Wno-mudflap option disables warnings about constructs
3920       which can not be instrumented when using -fmudflap.
3921
3922New Languages and Language specific improvements
3923
3924     * Version 3.0 of the [5]OpenMP specification is now supported for the
3925       C, C++, and Fortran compilers.
3926     * New character data types, per [6]TR 19769: New character types in
3927       C, are now supported for the C compiler in -std=gnu99 mode, as
3928       __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in
3929       -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too.
3930
3931  C family
3932
3933     * A new optimize attribute was added to allow programmers to change
3934       the optimization level and particular optimization options for an
3935       individual function. You can also change the optimization options
3936       via the GCC optimize pragma for functions defined after the pragma.
3937       The GCC push_options pragma and the GCC pop_options pragma allow
3938       you temporarily save and restore the options used. The GCC
3939       reset_options pragma restores the options to what was specified on
3940       the command line.
3941     * Uninitialized warnings do not require enabling optimization
3942       anymore, that is, -Wuninitialized can be used together with -O0.
3943       Nonetheless, the warnings given by -Wuninitialized will probably be
3944       more accurate if optimization is enabled.
3945     * -Wparentheses now warns about expressions such as (!x | y) and (!x
3946       & y). Using explicit parentheses, such as in ((!x) | y), silences
3947       this warning.
3948     * -Wsequence-point now warns within if, while,do while and for
3949       conditions, and within for begin/end expressions.
3950     * A new option -dU is available to dump definitions of preprocessor
3951       macros that are tested or expanded.
3952
3953  C++
3954
3955     * [7]Improved experimental support for the upcoming ISO C++ standard,
3956       C++0x. Including support for auto, inline namespaces, generalized
3957       initializer lists, defaulted and deleted functions, new character
3958       types, and scoped enums.
3959     * Those errors that may be downgraded to warnings to build legacy
3960       code now mention -fpermissive when -fdiagnostics-show-option is
3961       enabled.
3962     * -Wconversion now warns if the result of a static_cast to enumeral
3963       type is unspecified because the value is outside the range of the
3964       enumeral type.
3965     * -Wuninitialized now warns if a non-static reference or non-static
3966       const member appears in a class without constructors.
3967     * G++ now properly implements value-initialization, so objects with
3968       an initializer of () and an implicitly defined default constructor
3969       will be zero-initialized before the default constructor is called.
3970
3971    Runtime Library (libstdc++)
3972
3973     * Improved experimental support for the upcoming ISO C++ standard,
3974       C++0x, including:
3975          + Support for <chrono>, <condition_variable>, <cstdatomic>,
3976            <forward_list>, <initializer_list>, <mutex>, <ratio>,
3977            <system_error>, and <thread>.
3978          + unique_ptr, <algorithm> additions, exception propagation, and
3979            support for the new character types in <string> and <limits>.
3980          + Existing facilities now exploit initializer lists, defaulted
3981            and deleted functions, and the newly implemented core C++0x
3982            features.
3983          + Some standard containers are more efficient together with
3984            stateful allocators, i.e., no allocator is constructed on the
3985            fly at element construction time.
3986     * Experimental support for non-standard pointer types in containers.
3987     * The long standing libstdc++/30928 has been fixed for targets
3988       running glibc 2.10 or later.
3989     * As usual, many small and larger bug fixes, in particular quite a
3990       few corner cases in <locale>.
3991
3992  Fortran
3993
3994     * GNU Fortran now employs libcpp directly instead of using cc1 as an
3995       external preprocessor. The [8]-cpp option was added to allow manual
3996       invocation of the preprocessor without relying on filename
3997       extensions.
3998     * The [9]-Warray-temporaries option warns about array temporaries
3999       generated by the compiler, as an aid to optimization.
4000     * The [10]-fcheck-array-temporaries option has been added, printing a
4001       notification at run time, when an array temporary had to be created
4002       for an function argument. Contrary to -Warray-temporaries the
4003       warning is only printed if the array is noncontiguous.
4004     * Improved generation of DWARF debugging symbols
4005     * If using an intrinsic not part of the selected standard (via -std=
4006       and -fall-intrinsics) gfortran will now treat it as if this
4007       procedure were declared EXTERNAL and try to link to a user-supplied
4008       procedure. -Wintrinsics-std will warn whenever this happens. The
4009       now-useless option -Wnonstd-intrinsic was removed.
4010     * The flag -falign-commons has been added to control the alignment of
4011       variables in COMMON blocks, which is enabled by default in line
4012       with previous GCC version. Using -fno-align-commons one can force
4013       commons to be contiguous in memory as required by the Fortran
4014       standard, however, this slows down the memory access. The option
4015       -Walign-commons, which is enabled by default, warns when padding
4016       bytes were added for alignment. The proper solution is to sort the
4017       common objects by decreasing storage size, which avoids the
4018       alignment problems.
4019     * Fortran 2003 support has been extended:
4020          + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is
4021            now supported (except internal reads from/writes to wide
4022            strings). [11]-fbackslash now supports also \unnnn and
4023            \Unnnnnnnn to enter Unicode characters.
4024          + Asynchronous I/O (implemented as synchronous I/O) and the
4025            decimal=, size=, sign=, pad=, blank=, and delim= specifiers
4026            are now supported in I/O statements.
4027          + Support for Fortran 2003 structure constructors and for array
4028            constructor with typespec has been added.
4029          + Procedure Pointers (but not yet as component in derived types
4030            and as function results) are now supported.
4031          + Abstract types, type extension, and type-bound procedures
4032            (both PROCEDURE and GENERIC but not as operators). Note: As
4033            CLASS/polymorphyic types are not implemented, type-bound
4034            procedures with PASS accept as non-standard extension TYPE
4035            arguments.
4036     * Fortran 2008 support has been added:
4037          + The -std=f2008 option and support for the file extensions
4038            .f2008 and .F2008 has been added.
4039          + The g0 format descriptor is now supported.
4040          + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH,
4041            ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED
4042            are now available (some of them existed as GNU extension
4043            before). Note: The hyperbolic functions are not yet supporting
4044            complex arguments and the three- argument version of BESSEL_*N
4045            is not available.
4046          + The bit intrinsics LEADZ and TRAILZ have been added.
4047
4048  Java (GCJ)
4049
4050  Ada
4051
4052     * The Ada runtime now supports multilibs on many platforms including
4053       x86_64, SPARC and PowerPC. Their build is enabled by default.
4054
4055New Targets and Target Specific Improvements
4056
4057  ARM
4058
4059     * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and
4060       Cortex-R4F processors and has many other improvements to
4061       optimization for ARM processors.
4062     * GCC now supports the VFPv3 variant with 16 double-precision
4063       registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been
4064       renamed to -mfpu=vfpv3.
4065     * GCC now supports the -mfix-cortex-m3-ldrd option to work around an
4066       erratum on Cortex-M3 processors.
4067     * GCC now supports the __sync_* atomic operations for ARM EABI
4068       GNU/Linux.
4069     * The section anchors optimization is now enabled by default when
4070       optimizing for ARM.
4071     * GCC now uses a new EABI-compatible profiling interface for EABI
4072       targets. This requires a function __gnu_mcount_nc, which is
4073       provided by GNU libc versions 2.8 and later.
4074
4075  AVR
4076
4077     * The -mno-tablejump option has been deprecated because it has the
4078       same effect as the -fno-jump-tables option.
4079     * Added support for these new AVR devices:
4080          + ATA6289
4081          + ATtiny13A
4082          + ATtiny87
4083          + ATtiny167
4084          + ATtiny327
4085          + ATmega8C1
4086          + ATmega16C1
4087          + ATmega32C1
4088          + ATmega8M1
4089          + ATmega16M1
4090          + ATmega32M1
4091          + ATmega32U4
4092          + ATmega16HVB
4093          + ATmega4HVD
4094          + ATmega8HVD
4095          + ATmega64C1
4096          + ATmega64M1
4097          + ATmega16U4
4098          + ATmega32U6
4099          + ATmega128RFA1
4100          + AT90PWM81
4101          + AT90SCR100
4102          + M3000F
4103          + M3000S
4104          + M3001B
4105
4106  IA-32/x86-64
4107
4108     * Support for Intel AES built-in functions and code generation is
4109       available via -maes.
4110     * Support for Intel PCLMUL built-in function and code generation is
4111       available via -mpclmul.
4112     * Support for Intel AVX built-in functions and code generation is
4113       available via -mavx.
4114     * Automatically align the stack for local variables with alignment
4115       requirement.
4116     * GCC can now utilize the SVML library for vectorizing calls to a set
4117       of C99 functions if -mveclibabi=svml is specified and you link to
4118       an SVML ABI compatible library.
4119     * On x86-64, the ABI has been changed in the following cases to
4120       conform to the x86-64 ABI:
4121          + Passing/returning structures with flexible array member:
4122  struct foo
4123    {
4124      int i;
4125      int flex[];
4126    };
4127          + Passing/returning structures with complex float member:
4128  struct foo
4129    {
4130      int i;
4131      __complex__ float f;
4132    };
4133          + Passing/returning unions with long double member:
4134  union foo
4135    {
4136      int x;
4137      long double ld;
4138    };
4139       Code built with previous versions of GCC that uses any of these is
4140       not compatible with code built with GCC 4.4.0 or later.
4141     * A new target attribute was added to allow programmers to change the
4142       target options like -msse2 or -march=k8 for an individual function.
4143       You can also change the target options via the GCC target pragma
4144       for functions defined after the pragma.
4145     * GCC can now be configured with options --with-arch-32,
4146       --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and
4147       --with-tune-64 to control the default optimization separately for
4148       32-bit and 64-bit modes.
4149
4150  IA-32/IA64
4151
4152     * Support for __float128 (TFmode) IEEE quad type and corresponding
4153       TCmode IEEE complex quad type is available via the soft-fp library
4154       on IA-32/IA64 targets. This includes basic arithmetic operations
4155       (addition, subtraction, negation, multiplication and division) on
4156       __float128 real and TCmode complex values, the full set of IEEE
4157       comparisons between __float128 values, conversions to and from
4158       float, double and long double floating point types, as well as
4159       conversions to and from signed or unsigned integer, signed or
4160       unsigned long integer and signed or unsigned quad (TImode, IA64
4161       only) integer types. Additionally, all operations generate the full
4162       set of IEEE exceptions and support the full set of IEEE rounding
4163       modes.
4164
4165  M68K/ColdFire
4166
4167     * GCC now supports instruction scheduling for ColdFire V1, V3 and V4
4168       processors. (Scheduling support for ColdFire V2 processors was
4169       added in GCC 4.3.)
4170     * GCC now supports the -mxgot option to support programs requiring
4171       many GOT entries on ColdFire.
4172     * The m68k-*-linux-gnu target now builds multilibs by default.
4173
4174  MIPS
4175
4176     * MIPS Technologies have extended the original MIPS SVR4 ABI to
4177       include support for procedure linkage tables (PLTs) and copy
4178       relocations. These extensions allow GNU/Linux executables to use a
4179       significantly more efficient code model than the one defined by the
4180       original ABI.
4181       GCC support for this code model is available via a new command-line
4182       option, -mplt. There is also a new configure-time option,
4183       --with-mips-plt, to make -mplt the default.
4184       The new code model requires support from the assembler, the linker,
4185       and the runtime C library. This support is available in binutils
4186       2.19 and GLIBC 2.9.
4187     * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables
4188       and 32-bit GNU/Linux shared libraries. This feature requires GNU
4189       binutils 2.19 or above.
4190     * Support for RMI's XLR processor is now available through the
4191       -march=xlr and -mtune=xlr options.
4192     * 64-bit targets can now perform 128-bit multiplications inline,
4193       instead of relying on a libgcc function.
4194     * Native GNU/Linux toolchains now support -march=native and
4195       -mtune=native, which select the host processor.
4196     * GCC now supports the R10K, R12K, R14K and R16K processors. The
4197       canonical -march= and -mtune= names for these processors are
4198       r10000, r12000, r14000 and r16000 respectively.
4199     * GCC can now work around the side effects of speculative execution
4200       on R10K processors. Please see the documentation of the
4201       -mr10k-cache-barrier option for details.
4202     * Support for the MIPS64 Release 2 instruction set has been added.
4203       The option -march=mips64r2 enables generation of these
4204       instructions.
4205     * GCC now supports Cavium Networks' Octeon processor. This support is
4206       available through the -march=octeon and -mtune=octeon options.
4207     * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The
4208       canonical -march= and -mtune= names for these processors are
4209       loongson2e and loongson2f.
4210
4211  picochip
4212
4213   Picochip is a 16-bit processor. A typical picoChip contains over 250
4214   small cores, each with small amounts of memory. There are three
4215   processor variants (STAN, MEM and CTRL) with different instruction sets
4216   and memory configurations and they can be chosen using the -mae option.
4217
4218   This port is intended to be a "C" only port.
4219
4220  Power Architecture and PowerPC
4221
4222     * GCC now supports the e300c2, e300c3 and e500mc processors.
4223     * GCC now supports Xilinx processors with a single-precision FPU.
4224     * Decimal floating point is now supported for e500 processors.
4225
4226  S/390, zSeries and System z9/z10
4227
4228     * Support for the IBM System z10 EC/BC processor has been added. When
4229       using the -march=z10 option, the compiler will generate code making
4230       use of instructions provided by the General-Instruction-Extension
4231       Facility and the Execute-Extension Facility.
4232
4233  VxWorks
4234
4235     * GCC now supports the thread-local storage mechanism used on
4236       VxWorks.
4237
4238  Xtensa
4239
4240     * GCC now supports thread-local storage (TLS) for Xtensa processor
4241       configurations that include the Thread Pointer option. TLS also
4242       requires support from the assembler and linker; this support is
4243       provided in the GNU binutils beginning with version 2.19.
4244
4245Documentation improvements
4246
4247Other significant improvements
4248
4249GCC 4.4.1
4250
4251   This is the [12]list of problem reports (PRs) from GCC's bug tracking
4252   system that are known to be fixed in the 4.4.1 release. This list might
4253   not be complete (that is, it is possible that some PRs that have been
4254   fixed are not listed here).
4255
4256GCC 4.4.2
4257
4258   This is the [13]list of problem reports (PRs) from GCC's bug tracking
4259   system that are known to be fixed in the 4.4.2 release. This list might
4260   not be complete (that is, it is possible that some PRs that have been
4261   fixed are not listed here).
4262
4263GCC 4.4.3
4264
4265   This is the [14]list of problem reports (PRs) from GCC's bug tracking
4266   system that are known to be fixed in the 4.4.3 release. This list might
4267   not be complete (that is, it is possible that some PRs that have been
4268   fixed are not listed here).
4269
4270GCC 4.4.4
4271
4272   This is the [15]list of problem reports (PRs) from GCC's bug tracking
4273   system that are known to be fixed in the 4.4.4 release. This list might
4274   not be complete (that is, it is possible that some PRs that have been
4275   fixed are not listed here).
4276
4277GCC 4.4.5
4278
4279   This is the [16]list of problem reports (PRs) from GCC's bug tracking
4280   system that are known to be fixed in the 4.4.5 release. This list might
4281   not be complete (that is, it is possible that some PRs that have been
4282   fixed are not listed here).
4283
4284GCC 4.4.6
4285
4286   This is the [17]list of problem reports (PRs) from GCC's bug tracking
4287   system that are known to be fixed in the 4.4.6 release. This list might
4288   not be complete (that is, it is possible that some PRs that have been
4289   fixed are not listed here).
4290
4291GCC 4.4.7
4292
4293   This is the [18]list of problem reports (PRs) from GCC's bug tracking
4294   system that are known to be fixed in the 4.4.7 release. This list might
4295   not be complete (that is, it is possible that some PRs that have been
4296   fixed are not listed here).
4297
4298
4299    For questions related to the use of GCC, please consult these web
4300    pages and the [19]GCC manuals. If that fails, the
4301    [20]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4302    web pages and the development of GCC are welcome on our developer
4303    list at [21]gcc@gcc.gnu.org. All of [22]our lists have public
4304    archives.
4305
4306   Copyright (C) [23]Free Software Foundation, Inc. Verbatim copying and
4307   distribution of this entire article is permitted in any medium,
4308   provided this notice is preserved.
4309
4310   These pages are [24]maintained by the GCC team. Last modified
4311   2014-12-06[25].
4312
4313References
4314
4315   1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7
4316   2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted
4317   3. http://gcc.gnu.org/gcc-4.4/porting_to.html
4318   4. https://gcc.gnu.org/wiki/Graphite
4319   5. http://openmp.org/wp/openmp-specifications/
4320   6. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf
4321   7. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html
4322   8. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html
4323   9. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125
4324  10. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221
4325  11. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34
4326  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1
4327  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2
4328  14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3
4329  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4
4330  16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5
4331  17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6
4332  18. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7
4333  19. https://gcc.gnu.org/onlinedocs/
4334  20. mailto:gcc-help@gcc.gnu.org
4335  21. mailto:gcc@gcc.gnu.org
4336  22. https://gcc.gnu.org/lists.html
4337  23. http://www.fsf.org/
4338  24. https://gcc.gnu.org/about.html
4339  25. http://validator.w3.org/check/referer
4340======================================================================
4341http://gcc.gnu.org/gcc-4.3/index.html
4342
4343                             GCC 4.3 Release Series
4344
4345   Jun 27, 2011
4346
4347   The [1]GNU project and the GCC developers are pleased to announce the
4348   release of GCC 4.3.6.
4349
4350   This release is a bug-fix release, containing fixes for regressions in
4351   GCC 4.3.5 relative to previous releases of GCC.
4352
4353Release History
4354
4355   GCC 4.3.6
4356          Jun 27, 2011 ([2]changes)
4357
4358   GCC 4.3.5
4359          May 22, 2010 ([3]changes)
4360
4361   GCC 4.3.4
4362          August 4, 2009 ([4]changes)
4363
4364   GCC 4.3.3
4365          January 24, 2009 ([5]changes)
4366
4367   GCC 4.3.2
4368          August 27, 2008 ([6]changes)
4369
4370   GCC 4.3.1
4371          June 6, 2008 ([7]changes)
4372
4373   GCC 4.3.0
4374          March 5, 2008 ([8]changes)
4375
4376References and Acknowledgements
4377
4378   GCC used to stand for the GNU C Compiler, but since the compiler
4379   supports several other languages aside from C, it now stands for the
4380   GNU Compiler Collection.
4381
4382   A list of [9]successful builds is updated as new information becomes
4383   available.
4384
4385   The GCC developers would like to thank the numerous people that have
4386   contributed new features, improvements, bug fixes, and other changes as
4387   well as test results to GCC. This [10]amazing group of volunteers is
4388   what makes GCC successful.
4389
4390   For additional information about GCC please refer to the [11]GCC
4391   project web site or contact the [12]GCC development mailing list.
4392
4393   To obtain GCC please use [13]our mirror sites or [14]our SVN server.
4394
4395
4396    For questions related to the use of GCC, please consult these web
4397    pages and the [15]GCC manuals. If that fails, the
4398    [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4399    web pages and the development of GCC are welcome on our developer
4400    list at [17]gcc@gcc.gnu.org. All of [18]our lists have public
4401    archives.
4402
4403   Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and
4404   distribution of this entire article is permitted in any medium,
4405   provided this notice is preserved.
4406
4407   These pages are [20]maintained by the GCC team. Last modified
4408   2014-06-28[21].
4409
4410References
4411
4412   1. http://www.gnu.org/
4413   2. http://gcc.gnu.org/gcc-4.3/changes.html
4414   3. http://gcc.gnu.org/gcc-4.3/changes.html
4415   4. http://gcc.gnu.org/gcc-4.3/changes.html
4416   5. http://gcc.gnu.org/gcc-4.3/changes.html
4417   6. http://gcc.gnu.org/gcc-4.3/changes.html
4418   7. http://gcc.gnu.org/gcc-4.3/changes.html
4419   8. http://gcc.gnu.org/gcc-4.3/changes.html
4420   9. http://gcc.gnu.org/gcc-4.3/buildstat.html
4421  10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
4422  11. http://gcc.gnu.org/index.html
4423  12. mailto:gcc@gcc.gnu.org
4424  13. http://gcc.gnu.org/mirrors.html
4425  14. http://gcc.gnu.org/svn.html
4426  15. https://gcc.gnu.org/onlinedocs/
4427  16. mailto:gcc-help@gcc.gnu.org
4428  17. mailto:gcc@gcc.gnu.org
4429  18. https://gcc.gnu.org/lists.html
4430  19. http://www.fsf.org/
4431  20. https://gcc.gnu.org/about.html
4432  21. http://validator.w3.org/check/referer
4433======================================================================
4434http://gcc.gnu.org/gcc-4.3/changes.html
4435
4436                             GCC 4.3 Release Series
4437                        Changes, New Features, and Fixes
4438
4439   The latest release in the 4.3 release series is [1]GCC 4.3.5.
4440
4441Caveats
4442
4443     * GCC requires the [2]GMP and [3]MPFR libraries for building all the
4444       various front-end languages it supports. See the [4]prerequisites
4445       page for version requirements.
4446     * ColdFire targets now treat long double as having the same format as
4447       double. In earlier versions of GCC, they used the 68881 long double
4448       format instead.
4449     * The m68k-uclinux target now uses the same calling conventions as
4450       m68k-linux-gnu. You can select the original calling conventions by
4451       configuring for m68k-uclinuxoldabi instead. Note that
4452       m68k-uclinuxoldabi also retains the original 80-bit long double on
4453       ColdFire targets.
4454     * The -fforce-mem option has been removed because it has had no
4455       effect in the last few GCC releases.
4456     * The i386 -msvr3-shlib option has been removed since it is no longer
4457       used.
4458     * Fastcall for i386 has been changed not to pass aggregate arguments
4459       in registers, following Microsoft compilers.
4460     * Support for the AOF assembler has been removed from the ARM back
4461       end; this affects only the targets arm-semi-aof and armel-semi-aof,
4462       which are no longer recognized. We removed these targets without a
4463       deprecation period because we discovered that they have been
4464       unusable since GCC 4.0.0.
4465     * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*)
4466       has been removed. This support had been deprecated since GCC 4.0.0.
4467     * Support for a number of older systems and recently unmaintained or
4468       untested target ports of GCC has been declared obsolete in GCC 4.3.
4469       Unless there is activity to revive them, the next release of GCC
4470       will have their sources permanently removed.
4471       All GCC ports for the following processor architectures have been
4472       declared obsolete:
4473          + Morpho MT (mt-*)
4474       The following aliases for processor architectures have been
4475       declared obsolete. Users should use the indicated generic target
4476       names instead, with compile-time options such as -mcpu or
4477       configure-time options such as --with-cpu to control the
4478       configuration more precisely.
4479          + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-*
4480            instead).
4481          + parisc*-*-* (use hppa*-*-* instead).
4482          + m680[012]0-*-* (use m68k-*-* instead).
4483       All GCC ports for the following operating systems have been
4484       declared obsolete:
4485          + BeOS (*-*-beos*)
4486          + kaOS (*-*-kaos*)
4487          + GNU/Linux using the a.out object format (*-*-linux*aout*)
4488          + GNU/Linux using version 1 of the GNU C Library
4489            (*-*-linux*libc1*)
4490          + Solaris versions before Solaris 7 (*-*-solaris2.[0-6],
4491            *-*-solaris2.[0-6].*)
4492          + Miscellaneous System V (*-*-sysv*)
4493          + WindISS (*-*-windiss*)
4494       Also, those for some individual systems on particular architectures
4495       have been obsoleted:
4496          + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
4497          + CRIS with a.out object format (cris-*-aout)
4498          + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
4499          + OSF/1 on PA-RISC (hppa1.1-*-osf*)
4500          + PRO on PA-RISC (hppa1.1-*-pro*)
4501          + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*,
4502            i[34567]86-sequent-sysv4*)
4503          + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
4504          + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host
4505            was previously [5]removed in 2001, leaving only the support
4506            for UWIN as a target now being deprecated)
4507          + ChorusOS on PowerPC (powerpc-*-chorusos*)
4508          + All VAX configurations apart from NetBSD and OpenBSD
4509            (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
4510     * The [6]-Wconversion option has been modified. Its purpose now is to
4511       warn for implicit conversions that may alter a value. This new
4512       behavior is available for both C and C++. Warnings about
4513       conversions between signed and unsigned integers can be disabled by
4514       using -Wno-sign-conversion. In C++, they are disabled by default
4515       unless -Wsign-conversion is explicitly requested. The old behavior
4516       of -Wconversion, that is, warn for prototypes causing a type
4517       conversion that is different from what would happen to the same
4518       argument in the absence of a prototype, has been moved to a new
4519       option -Wtraditional-conversion, which is only available for C.
4520     * The -m386, -m486, -mpentium and -mpentiumpro tuning options have
4521       been removed because they were deprecated for more than 3 GCC major
4522       releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or
4523       -mtune=pentiumpro as a replacement.
4524     * The -funsafe-math-optimizations option now automatically turns on
4525       -fno-trapping-math in addition to -fno-signed-zeros, as it enables
4526       reassociation and thus may introduce or remove traps.
4527     * The -ftree-vectorize option is now on by default under -O3. In
4528       order to generate code for a SIMD extension, it has to be enabled
4529       as well: use -maltivec for PowerPC platforms and -msse/-msse2 for
4530       i?86 and x86_64.
4531     * More information on porting to GCC 4.3 from previous versions of
4532       GCC can be found in the [7]porting guide for this release.
4533
4534General Optimizer Improvements
4535
4536     * The GCC middle-end has been integrated with the [8]MPFR library.
4537       This allows GCC to evaluate and replace at compile-time calls to
4538       built-in math functions having constant arguments with their
4539       mathematically equivalent results. In making use of [9]MPFR, GCC
4540       can generate correct results regardless of the math library
4541       implementation or floating point precision of the host platform.
4542       This also allows GCC to generate identical results regardless of
4543       whether one compiles in native or cross-compile configurations to a
4544       particular target. The following built-in functions take advantage
4545       of this new capability: acos, acosh, asin, asinh, atan2, atan,
4546       atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1,
4547       fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10,
4548       log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh,
4549       tan, tanh, tgamma, y0, y1 and yn. The float and long double
4550       variants of these functions (e.g. sinf and sinl) are also handled.
4551       The sqrt and cabs functions with constant arguments were already
4552       optimized in prior GCC releases. Now they also use [10]MPFR.
4553     * A new forward propagation pass on RTL was added. The new pass
4554       replaces several slower transformations, resulting in compile-time
4555       improvements as well as better code generation in some cases.
4556     * A new command-line switch -frecord-gcc-switches has been added to
4557       GCC, although it is only enabled for some targets. The switch
4558       causes the command line that was used to invoke the compiler to be
4559       recorded into the object file that is being created. The exact
4560       format of this recording is target and binary file format
4561       dependent, but it usually takes the form of a note section
4562       containing ASCII text. The switch is related to the -fverbose-asm
4563       switch, but that one only records the information in the assembler
4564       output file as comments, so the information never reaches the
4565       object file.
4566     * The inliner heuristic is now aware of stack frame consumption. New
4567       command-line parameters --param large-stack-frame and --param
4568       large-stack-frame-growth can be used to limit stack frame size
4569       growth caused by inlining.
4570     * During feedback directed optimizations, the expected block size the
4571       memcpy, memset and bzero functions operate on is discovered and for
4572       cases of commonly used small sizes, specialized inline code is
4573       generated.
4574     * __builtin_expect no longer requires its argument to be a compile
4575       time constant.
4576     * Interprocedural optimization was reorganized to work on functions
4577       in SSA form. This enables more precise and cheaper dataflow
4578       analysis and makes writing interprocedural optimizations easier.
4579       The following improvements have been implemented on top of this
4580       framework:
4581          + Pre-inline optimization: Selected local optimization passes
4582            are run before the inliner (and other interprocedural passes)
4583            are executed. This significantly improves the accuracy of code
4584            growth estimates used by the inliner and reduces the overall
4585            memory footprint for large compilation units.
4586          + Early inlining (a simple bottom-up inliner pass inlining only
4587            functions whose body is smaller than the expected call
4588            overhead) is now executed with the early optimization passes,
4589            thus inlining already optimized function bodies into an
4590            unoptimized function that is subsequently optimized by early
4591            optimizers. This enables the compiler to quickly eliminate
4592            abstraction penalty in C++ programs.
4593          + Interprocedural constant propagation now operate on SSA form
4594            increasing accuracy of the analysis.
4595     * A new internal representation for GIMPLE statements has been
4596       contributed, resulting in compile-time memory savings.
4597     * The vectorizer was enhanced to support vectorization of outer
4598       loops, intra-iteration parallelism (loop-aware SLP), vectorization
4599       of strided accesses and loops with multiple data-types. Run-time
4600       dependency testing using loop versioning was added. The cost model,
4601       turned on by -fvect-cost-model, was developed.
4602
4603New Languages and Language specific improvements
4604
4605     * We have added new command-line options
4606       -finstrument-functions-exclude-function-list and
4607       -finstrument-functions-exclude-file-list. They provide more control
4608       over which functions are annotated by the -finstrument-functions
4609       option.
4610
4611  C family
4612
4613     * Implicit conversions between generic vector types are now only
4614       permitted when the two vectors in question have the same number of
4615       elements and compatible element types. (Note that the restriction
4616       involves compatible element types, not implicitly-convertible
4617       element types: thus, a vector type with element type int may not be
4618       implicitly converted to a vector type with element type unsigned
4619       int.) This restriction, which is in line with specifications for
4620       SIMD architectures such as AltiVec, may be relaxed using the flag
4621       -flax-vector-conversions. This flag is intended only as a
4622       compatibility measure and should not be used for new code.
4623     * -Warray-bounds has been added and is now enabled by default for
4624       -Wall . It produces warnings for array subscripts that can be
4625       determined at compile time to be always out of bounds.
4626       -Wno-array-bounds will disable the warning.
4627     * The constructor and destructor function attributes now accept
4628       optional priority arguments which control the order in which the
4629       constructor and destructor functions are run.
4630     * New [11]command-line options -Wtype-limits,
4631       -Wold-style-declaration, -Wmissing-parameter-type, -Wempty-body,
4632       -Wclobbered and -Wignored-qualifiers have been added for finer
4633       control of the diverse warnings enabled by -Wextra.
4634     * A new function attribute alloc_size has been added to mark up
4635       malloc style functions. For constant sized allocations this can be
4636       used to find out the size of the returned pointer using the
4637       __builtin_object_size() function for buffer overflow checking and
4638       similar. This supplements the already built-in malloc and calloc
4639       constant size handling.
4640     * Integer constants written in binary are now supported as a GCC
4641       extension. They consist of a prefix 0b or 0B, followed by a
4642       sequence of 0 and 1 digits.
4643     * A new predefined macro __COUNTER__ has been added. It expands to
4644       sequential integral values starting from 0. In conjunction with the
4645       ## operator, this provides a convenient means to generate unique
4646       identifiers.
4647     * A new command-line option -fdirectives-only has been added. It
4648       enables a special preprocessing mode which improves the performance
4649       of applications like distcc and ccache.
4650     * Fixed-point data types and operators have been added. They are
4651       based on Chapter 4 of the Embedded-C specification (n1169.pdf).
4652       Currently, only MIPS targets are supported.
4653     * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732,
4654       N1241, is now supported as a GCC extension to C for targets
4655       i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu,
4656       and x86_64-*-linux-gnu. The feature introduces new data types
4657       _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF,
4658       DD, and DL.
4659
4660  C++
4661
4662     * [12]Experimental support for the upcoming ISO C++ standard, C++0x.
4663     * -Wc++0x-compat has been added and is now enabled by default for
4664       -Wall. It produces warnings for constructs whose meaning differs
4665       between ISO C++ 1998 and C++0x.
4666     * The -Wparentheses option now works for C++ as it does for C. It
4667       warns if parentheses are omitted when operators with confusing
4668       precedence are nested. It also warns about ambiguous else
4669       statements. Since -Wparentheses is enabled by -Wall, this may cause
4670       additional warnings with existing C++ code which uses -Wall. These
4671       new warnings may be disabled by using -Wall -Wno-parentheses.
4672     * The -Wmissing-declarations now works for C++ as it does for C.
4673     * The -fvisibility-ms-compat flag was added, to make it easier to
4674       port larger projects using shared libraries from Microsoft's Visual
4675       Studio to ELF and Mach-O systems.
4676     * C++ attribute handling has been overhauled for template arguments
4677       (ie dependent types). In particular, __attribute__((aligned(T)));
4678       works for C++ types.
4679
4680    Runtime Library (libstdc++)
4681
4682     * [13]Experimental support for the upcoming ISO C++ standard, C++0x.
4683     * Support for TR1 mathematical special functions and regular
4684       expressions. ([14]Implementation status of TR1)
4685     * Default what implementations give more elaborate exception strings
4686       for bad_cast, bad_typeid, bad_exception, and bad_alloc.
4687     * Header dependencies have been streamlined, reducing unnecessary
4688       includes and pre-processed bloat.
4689     * Variadic template implementations of items in <tuple> and
4690       <functional>.
4691     * An experimental [15]parallel mode has been added. This is a
4692       parallel implementation of many C++ Standard library algorithms,
4693       like std::accumulate, std::for_each, std::transform, or std::sort,
4694       to give but four examples. These algorithms can be substituted for
4695       the normal (sequential) libstdc++ algorithms on a piecemeal basis,
4696       or all existing algorithms can be transformed via the
4697       -D_GLIBCXX_PARALLEL macro.
4698     * Debug mode versions of classes in <unordered_set> and
4699       <unordered_map>.
4700     * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are
4701       now <backward/hash_set> and <backward/hash_map>. This code:
4702    #include <ext/hash_set>
4703    __gnu_cxx::hash_set<int> s;
4704
4705       Can be transformed (in order of preference) to:
4706    #include <tr1/unordered_set>
4707    std::tr1::unordered_set<int> s;
4708
4709       or
4710    #include <backward/hash_set>
4711    __gnu_cxx::hash_set<int> s;
4712
4713       Similar transformations apply to __gnu_cxx::hash_map,
4714       __gnu_cxx::hash_multimap, __gnu_cxx::hash_set,
4715       __gnu_cxx::hash_multiset.
4716
4717  Fortran
4718
4719     * Due to the fact that the [16]GMP and [17]MPFR libraries are
4720       required for all languages, Fortran is no longer special in this
4721       regard and is available by default.
4722     * The [18]-fexternal-blas option has been added, which generates
4723       calls to BLAS routines for intrinsic matrix operations such as
4724       matmul rather than using the built-in algorithms.
4725     * Support to give a backtrace (compiler flag -fbacktrace or
4726       environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems
4727       only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a
4728       run-time error occured.
4729     * GNU Fortran now defines __GFORTRAN__ when it runs the C
4730       preprocessor (CPP).
4731     * The [19]-finit-local-zero, -finit-real, -finit-integer,
4732       -finit-character, and -finit-logical options have been added, which
4733       can be used to initialize local variables.
4734     * The intrinsic procedures [20]GAMMA and [21]LGAMMA have been added,
4735       which calculate the Gamma function and its logarithm. Use EXTERNAL
4736       gamma if you want to use your own gamma function.
4737     * GNU Fortran now regards the backslash character as literal (as
4738       required by the Fortran 2003 standard); using [22]-fbackslash GNU
4739       Fortran interprets backslashes as C-style escape characters.
4740     * The [23]interpretation of binary, octal and hexadecimal (BOZ)
4741       literal constants has been changed. Before they were always
4742       interpreted as integer; now they are bit-wise transferred as
4743       argument of INT, REAL, DBLE and CMPLX as required by the Fortran
4744       2003 standard, and for real and complex variables in DATA
4745       statements or when directly assigned to real and complex variables.
4746       Everywhere else and especially in expressions they are still
4747       regarded as integer constants.
4748     * Fortran 2003 support has been extended:
4749          + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
4750          + Pointer intent
4751          + Intrinsic module ISO_ENV_FORTRAN
4752          + Interoperability with C (ISO C Bindings)
4753          + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER
4754            attribute)
4755          + Fortran 2003 BOZ
4756
4757  Java (GCJ)
4758
4759     * GCJ now uses the Eclipse Java compiler for its Java parsing needs.
4760       This enables the use of all 1.5 language features, and fixes most
4761       existing front end bugs.
4762     * libgcj now supports all 1.5 language features which require runtime
4763       support: foreach, enum, annotations, generics, and auto-boxing.
4764     * We've made many changes to the tools shipped with gcj.
4765          + The old jv-scan tool has been removed. This tool never really
4766            worked properly. There is no replacement.
4767          + gcjh has been rewritten. Some of its more obscure options no
4768            longer work, but are still recognized in an attempt at
4769            compatibility. gjavah is a new program with similar
4770            functionality but different command-line options.
4771          + grmic and grmiregistry have been rewritten. grmid has been
4772            added.
4773          + gjar replaces the old fastjar.
4774          + gjarsigner (used for signing jars), gkeytool (used for key
4775            management), gorbd (for CORBA), gserialver (computes
4776            serialization UIDs), and gtnameserv (also for CORBA) are now
4777            installed.
4778     * The ability to dump the contents of the java run time heap to a
4779       file for off-line analysis has been added. The heap dumps may be
4780       analyzed with the new gc-analyze tool. They may be generated on
4781       out-of-memory conditions or on demand and are controlled by the new
4782       run time class gnu.gcj.util.GCInfo.
4783     * java.util.TimeZone can now read files from /usr/share/zoneinfo to
4784       provide correct, updated, timezone information. This means that
4785       packagers no longer have to update libgcj when a time zone change
4786       is published.
4787
4788New Targets and Target Specific Improvements
4789
4790  IA-32/x86-64
4791
4792     * Tuning for Intel Core 2 processors is available via -mtune=core2
4793       and -march=core2.
4794     * Tuning for AMD Geode processors is available via -mtune=geode and
4795       -march=geode.
4796     * Code generation of block move (memcpy) and block set (memset) was
4797       rewritten. GCC can now pick the best algorithm (loop, unrolled
4798       loop, instruction with rep prefix or a library call) based on the
4799       size of the block being copied and the CPU being optimized for. A
4800       new option -minline-stringops-dynamically has been added. With this
4801       option string operations of unknown size are expanded such that
4802       small blocks are copied by in-line code, while for large blocks a
4803       library call is used. This results in faster code than
4804       -minline-all-stringops when the library implementation is capable
4805       of using cache hierarchy hints. The heuristic choosing the
4806       particular algorithm can be overwritten via -mstringop-strategy.
4807       Newly also memset of values different from 0 is inlined.
4808     * GCC no longer places the cld instruction before string operations.
4809       Both i386 and x86-64 ABI documents mandate the direction flag to be
4810       clear at the entry of a function. It is now invalid to set the flag
4811       in asm statement without reseting it afterward.
4812     * Support for SSSE3 built-in functions and code generation are
4813       available via -mssse3.
4814     * Support for SSE4.1 built-in functions and code generation are
4815       available via -msse4.1.
4816     * Support for SSE4.2 built-in functions and code generation are
4817       available via -msse4.2.
4818     * Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
4819     * A new set of options -mpc32, -mpc64 and -mpc80 have been added to
4820       allow explicit control of x87 floating point precision.
4821     * Support for __float128 (TFmode) IEEE quad type and corresponding
4822       TCmode IEEE complex quad type is available via the soft-fp library
4823       on x86_64 targets. This includes basic arithmetic operations
4824       (addition, subtraction, negation, multiplication and division) on
4825       __float128 real and TCmode complex values, the full set of IEEE
4826       comparisons between __float128 values, conversions to and from
4827       float, double and long double floating point types, as well as
4828       conversions to and from signed or unsigned integer, signed or
4829       unsigned long integer and signed or unsigned quad (TImode) integer
4830       types. Additionally, all operations generate the full set of IEEE
4831       exceptions and support the full set of IEEE rounding modes.
4832     * GCC can now utilize the ACML library for vectorizing calls to a set
4833       of C99 functions on x86_64 if -mveclibabi=acml is specified and you
4834       link to an ACML ABI compatible library.
4835
4836  ARM
4837
4838     * Compiler and Library support for Thumb-2 and the ARMv7 architecture
4839       has been added.
4840
4841  CRIS
4842
4843    New features
4844
4845     * Compiler and Library support for the CRIS v32 architecture, as
4846       found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been
4847       added.
4848
4849    Configuration changes
4850
4851     * The cris-*-elf target now includes support for CRIS v32, including
4852       libraries, through the -march=v32 option.
4853     * A new crisv32-*-elf target defaults to generate code for CRIS v32.
4854     * A new crisv32-*-linux* target defaults to generate code for CRIS
4855       v32.
4856     * The cris-*-aout target has been obsoleted.
4857
4858    Improved support for built-in functions
4859
4860     * GCC can now use the lz and swapwbr instructions to implement the
4861       __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
4862     * __builtin_bswap32 is now implemented using the swapwb instruction,
4863       when available.
4864
4865  m68k and ColdFire
4866
4867    New features
4868
4869     * Support for several new ColdFire processors has been added. You can
4870       generate code for them using the new -mcpu option.
4871     * All targets now support ColdFire processors.
4872     * m68k-uclinux targets have improved support for C++ constructors and
4873       destructors, and for shared libraries.
4874     * It is now possible to set breakpoints on the first or last line of
4875       a function, even if there are no statements on that line.
4876
4877    Optimizations
4878
4879     * Support for sibling calls has been added.
4880     * More use is now made of the ColdFire mov3q instruction.
4881     * __builtin_clz is now implemented using the ff1 ColdFire
4882       instruction, when available.
4883     * GCC now honors the -m68010 option. 68010 code now uses clr rather
4884       than move to zero volatile memory.
4885     * 68020 targets and above can now use symbol(index.size*scale)
4886       addresses for indexed array accesses. Earlier compilers would
4887       always load the symbol into a base register first.
4888
4889    Configuration changes
4890
4891     * All m68k and ColdFire targets now allow the default processor to be
4892       set at configure time using --with-cpu.
4893     * A --with-arch configuration option has been added. This option
4894       allows you to restrict a target to ColdFire or non-ColdFire
4895       processors.
4896
4897    Preprocessor macros
4898
4899     * An __mcfv*__ macro is now defined for all ColdFire targets.
4900       (Earlier versions of GCC only defined __mcfv4e__.)
4901     * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
4902     * All targets now define __mc68010 and __mc68010__ when generating
4903       68010 code.
4904
4905    Command-line changes
4906
4907     * New command-line options -march, -mcpu, -mtune and -mhard-float
4908       have been added. These options apply to both m68k and ColdFire
4909       targets.
4910     * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative
4911       versions of -mshort, etc.
4912     * -fforce-addr has been removed. It is now ignored by the compiler.
4913
4914    Other improvements
4915
4916     * ColdFire targets now try to maintain a 4-byte-aligned stack where
4917       possible.
4918     * m68k-uclinux targets now try to avoid situations that lead to the
4919       load-time error: BINFMT_FLAT: reloc outside program.
4920
4921  MIPS
4922
4923    Changes to existing configurations
4924
4925     * libffi and libjava now support all three GNU/Linux ABIs: o32, n32
4926       and n64. Every GNU/Linux configuration now builds these libraries
4927       by default.
4928     * GNU/Linux configurations now generate -mno-shared code unless
4929       overridden by -fpic, -fPIC, -fpie or -fPIE.
4930     * mipsisa32*-linux-gnu configurations now generate hard-float code by
4931       default, just like other mipsisa32* and mips*-linux-gnu
4932       configurations. You can build a soft-float version of any
4933       mips*-linux-gnu configuration by passing --with-float=soft to
4934       configure.
4935     * mips-wrs-vxworks now supports run-time processes (RTPs).
4936
4937    Changes to existing command-line options
4938
4939     * The -march and -mtune options no longer accept 24k as a processor
4940       name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
4941     * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and
4942       34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The
4943       options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for
4944       24kx, 24kex and 34kx.
4945
4946    New configurations
4947
4948   GCC now supports the following configurations:
4949     * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by
4950       default. Earlier releases also recognized this configuration, but
4951       they treated it in the same way as mipsisa32*-linux-gnu*. Note that
4952       you can customize any mips*-linux-gnu* configuration to a
4953       particular ISA or processor by passing an appropriate --with-arch
4954       option to configure.
4955     * mipsisa*-sde-elf*, which provides compatibility with MIPS
4956       Technologies' SDE toolchains. The configuration uses the SDE
4957       libraries by default, but you can use it like other newlib-based
4958       ELF configurations by passing --with-newlib to configure. It is the
4959       only configuration besides mips64vr*-elf* to build MIPS16 as well
4960       as non-MIPS16 libraries.
4961     * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf*
4962       configuration, but uses the o32 and o64 ABIs instead of the 32-bit
4963       and 64-bit forms of the EABI.
4964
4965    New processors and application-specific extensions
4966
4967     * Support for the SmartMIPS ASE is available through the new
4968       -msmartmips option.
4969     * Support for revision 2 of the DSP ASE is available through the new
4970       -mdspr2 option. A new preprocessor macro called __mips_dsp_rev
4971       indicates the revision of the ASE in use.
4972     * Support for the 4KS and 74K families of processors is available
4973       through the -march and -mtune options.
4974
4975    Improved support for built-in functions
4976
4977     * GCC can now use load-linked, store-conditional and sync
4978       instructions to implement atomic built-in functions such as
4979       __sync_fetch_and_add. The memory reference must be 4 bytes wide for
4980       32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
4981     * GCC can now use the clz and dclz instructions to implement the
4982       __builtin_ctz and __builtin_ffs families of functions.
4983     * There is a new __builtin___clear_cache function for flushing the
4984       instruction cache. GCC expands this function inline on MIPS32
4985       revision 2 targets, otherwise it calls the function specified by
4986       -mcache-flush-func.
4987
4988    MIPS16 improvements
4989
4990     * GCC can now compile objects that contain a mixture of MIPS16 and
4991       non-MIPS16 code. There are two new attributes, mips16 and nomips16,
4992       for specifying which mode a function should use.
4993     * A new option called -minterlink-mips16 makes non-MIPS16 code
4994       link-compatible with MIPS16 code.
4995     * After many bug fixes, the long-standing MIPS16 -mhard-float support
4996       should now work fairly reliably.
4997     * GCC can now use the MIPS16e save and restore instructions.
4998     * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled
4999       with -G0 -fsection-anchors is often smaller than code compiled with
5000       -G8. However, please note that you must usually compile all objects
5001       in your application with the same -G option; see the documentation
5002       of -G for details.
5003     * A new option called-mcode-readable specifies which instructions are
5004       allowed to load from the code segment. -mcode-readable=yes is the
5005       default and says that any instruction may load from the code
5006       segment. The other alternatives are -mcode-readable=pcrel, which
5007       says that only PC-relative MIPS16 instructions may load from the
5008       code segment, and -mcode-readable=no, which says that no
5009       instruction may do so. Please see the documentation for more
5010       details, including example uses.
5011
5012    Small-data improvements
5013
5014   There are three new options for controlling small data:
5015     * -mno-extern-sdata, which disables small-data accesses for
5016       externally-defined variables. Code compiled with -Gn
5017       -mno-extern-sdata will be link-compatible with any -G setting
5018       between -G0 and -Gn inclusive.
5019     * -mno-local-sdata, which disables the use of small-data sections for
5020       data that is not externally visible. This option can be a useful
5021       way of reducing small-data usage in less performance-critical parts
5022       of an application.
5023     * -mno-gpopt, which disables the use of the $gp register while still
5024       honoring the -G limit when placing externally-visible data. This
5025       option implies -mno-extern-sdata and -mno-local-sdata and it can be
5026       useful in situations where $gp does not necessarily hold the
5027       expected value.
5028
5029    Miscellaneous improvements
5030
5031     * There is a new option called -mbranch-cost for tweaking the
5032       perceived cost of branches.
5033     * If GCC is configured to use a version of GAS that supports the
5034       .gnu_attribute directive, it will use that directive to record
5035       certain properties of the output code. .gnu_attribute is new to GAS
5036       2.18.
5037     * There are two new function attributes, near and far, for overriding
5038       the command-line setting of -mlong-calls on a function-by-function
5039       basis.
5040     * -mfp64, which previously required a 64-bit target, now works with
5041       MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and
5042       mipsisa*-sde-elf* configurations provide suitable library support.
5043     * GCC now recognizes the -mdmx and -mmt options and passes them down
5044       to the assembler. It does nothing else with the options at present.
5045
5046  SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture
5047  (BEA)
5048
5049     * Support has been added for this new architecture.
5050
5051  RS6000 (POWER/PowerPC)
5052
5053     * Support for the PowerPC 750CL paired-single instructions has been
5054       added with a new powerpc-*-linux*paired* target configuration. It
5055       is enabled by an associated -mpaired option and can be accessed
5056       using new built-in functions.
5057     * Support for auto-detecting architecture and system configuration to
5058       auto-select processor optimization tuning.
5059     * Support for VMX on AIX 5.3 has been added.
5060     * Support for AIX Version 6.1 has been added.
5061
5062  S/390, zSeries and System z9
5063
5064     * Support for the IBM System z9 EC/BC processor (z9 GA3) has been
5065       added. When using the -march=z9-ec option, the compiler will
5066       generate code making use of instructions provided by the decimal
5067       floating point facility and the floating point conversion facility
5068       (pfpo). Besides the instructions used to implement decimal floating
5069       point operations these facilities also contain instructions to move
5070       between general purpose and floating point registers and to modify
5071       and copy the sign-bit of floating point values.
5072     * When the -march=z9-ec option is used the new
5073       -mhard-dfp/-mno-hard-dfp options can be used to specify whether the
5074       decimal floating point hardware instructions will be used or not.
5075       If none of them is given the hardware support is enabled by
5076       default.
5077     * The -mstack-guard option can now be omitted when using stack
5078       checking via -mstack-size in order to let GCC choose a sensible
5079       stack guard value according to the frame size of each function.
5080     * Various changes to improve performance of generated code have been
5081       implemented, including:
5082          + The condition code set by an add logical with carry
5083            instruction is now available for overflow checks like: a + b +
5084            carry < b.
5085          + The test data class instruction is now used to implement
5086            sign-bit and infinity checks of binary and decimal floating
5087            point numbers.
5088
5089  SPARC
5090
5091     * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been
5092       added.
5093
5094  Xtensa
5095
5096     * Stack unwinding for exception handling now uses by default a
5097       specialized version of DWARF unwinding. This is not
5098       binary-compatible with the setjmp/longjmp (sjlj) unwinding used for
5099       Xtensa with previous versions of GCC.
5100     * For Xtensa processors that include the Conditional Store option,
5101       the built-in functions for atomic memory access are now implemented
5102       using S32C1I instructions.
5103     * If the Xtensa NSA option is available, GCC will use it to implement
5104       the __builtin_ctz and __builtin_clz functions.
5105
5106Documentation improvements
5107
5108     * Existing libstdc++ documentation has been edited and restructured
5109       into a single DocBook XML manual. The results can be viewed online
5110       [24]here.
5111
5112Other significant improvements
5113
5114     * The compiler's --help command-line option has been extended so that
5115       it now takes an optional set of arguments. These arguments restrict
5116       the information displayed to specific classes of command-line
5117       options, and possibly only a subset of those options. It is also
5118       now possible to replace the descriptive text associated with each
5119       displayed option with an indication of its current value, or for
5120       binary options, whether it has been enabled or disabled.
5121       Here are some examples. The following will display all the options
5122       controlling warning messages:
5123      --help=warnings
5124
5125       Whereas this will display all the undocumented, target specific
5126       options:
5127      --help=target,undocumented
5128
5129       This sequence of commands will display the binary optimizations
5130       that are enabled by -O3:
5131      gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
5132      gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
5133      diff /tmp/O2-opts /tmp/O3-opts | grep enabled
5134
5135     * The configure options --with-pkgversion and --with-bugurl have been
5136       added. These allow distributors of GCC to include a
5137       distributor-specific string in manuals and --version output and to
5138       specify the URL for reporting bugs in their versions of GCC.
5139
5140GCC 4.3.1
5141
5142   This is the [25]list of problem reports (PRs) from GCC's bug tracking
5143   system that are known to be fixed in the 4.3.1 release. This list might
5144   not be complete (that is, it is possible that some PRs that have been
5145   fixed are not listed here).
5146
5147Target Specific Changes
5148
5149  IA-32/x86-64
5150
5151    ABI changes
5152
5153     * Starting with GCC 4.3.1, decimal floating point variables are
5154       aligned to their natural boundaries when they are passed on the
5155       stack for i386.
5156
5157    Command-line changes
5158
5159     * Starting with GCC 4.3.1, the -mcld option has been added to
5160       automatically generate a cld instruction in the prologue of
5161       functions that use string instructions. This option is used for
5162       backward compatibility on some operating systems and can be enabled
5163       by default for 32-bit x86 targets by configuring GCC with the
5164       --enable-cld configure option.
5165
5166GCC 4.3.2
5167
5168   This is the [26]list of problem reports (PRs) from GCC's bug tracking
5169   system that are known to be fixed in the 4.3.2 release. This list might
5170   not be complete (that is, it is possible that some PRs that have been
5171   fixed are not listed here).
5172
5173GCC 4.3.3
5174
5175   This is the [27]list of problem reports (PRs) from GCC's bug tracking
5176   system that are known to be fixed in the 4.3.3 release. This list might
5177   not be complete (that is, it is possible that some PRs that have been
5178   fixed are not listed here).
5179
5180GCC 4.3.4
5181
5182   This is the [28]list of problem reports (PRs) from GCC's bug tracking
5183   system that are known to be fixed in the 4.3.4 release. This list might
5184   not be complete (that is, it is possible that some PRs that have been
5185   fixed are not listed here).
5186
5187GCC 4.3.5
5188
5189   This is the [29]list of problem reports (PRs) from GCC's bug tracking
5190   system that are known to be fixed in the 4.3.5 release. This list might
5191   not be complete (that is, it is possible that some PRs that have been
5192   fixed are not listed here).
5193
5194GCC 4.3.6
5195
5196   This is the [30]list of problem reports (PRs) from GCC's bug tracking
5197   system that are known to be fixed in the 4.3.6 release. This list might
5198   not be complete (that is, it is possible that some PRs that have been
5199   fixed are not listed here).
5200
5201
5202    For questions related to the use of GCC, please consult these web
5203    pages and the [31]GCC manuals. If that fails, the
5204    [32]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5205    web pages and the development of GCC are welcome on our developer
5206    list at [33]gcc@gcc.gnu.org. All of [34]our lists have public
5207    archives.
5208
5209   Copyright (C) [35]Free Software Foundation, Inc. Verbatim copying and
5210   distribution of this entire article is permitted in any medium,
5211   provided this notice is preserved.
5212
5213   These pages are [36]maintained by the GCC team. Last modified
5214   2014-06-28[37].
5215
5216References
5217
5218   1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5
5219   2. http://gmplib.org/
5220   3. http://www.mpfr.org/
5221   4. https://gcc.gnu.org/install/prerequisites.html
5222   5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html
5223   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
5224   7. http://gcc.gnu.org/gcc-4.3/porting_to.html
5225   8. http://www.mpfr.org/
5226   9. http://www.mpfr.org/
5227  10. http://www.mpfr.org/
5228  11. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
5229  12. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5230  13. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5231  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#m anual.intro.status.standard.tr1
5232  15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
5233  16. http://gmplib.org/
5234  17. http://www.mpfr.org/
5235  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options
5236  19. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167
5237  20. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html
5238  21. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html
5239  22. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html
5240  23. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html
5241  24. https://gcc.gnu.org/onlinedocs/libstdc++/
5242  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1
5243  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2
5244  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3
5245  28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4
5246  29. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5
5247  30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6
5248  31. https://gcc.gnu.org/onlinedocs/
5249  32. mailto:gcc-help@gcc.gnu.org
5250  33. mailto:gcc@gcc.gnu.org
5251  34. https://gcc.gnu.org/lists.html
5252  35. http://www.fsf.org/
5253  36. https://gcc.gnu.org/about.html
5254  37. http://validator.w3.org/check/referer
5255======================================================================
5256http://gcc.gnu.org/gcc-4.2/index.html
5257
5258                             GCC 4.2 Release Series
5259
5260   May 19, 2008
5261
5262   The [1]GNU project and the GCC developers are pleased to announce the
5263   release of GCC 4.2.4.
5264
5265   This release is a bug-fix release, containing fixes for regressions in
5266   GCC 4.2.3 relative to previous releases of GCC.
5267
5268Release History
5269
5270   GCC 4.2.4
5271          May 19, 2008 ([2]changes)
5272
5273   GCC 4.2.3
5274          February 1, 2008 ([3]changes)
5275
5276   GCC 4.2.2
5277          October 7, 2007 ([4]changes)
5278
5279   GCC 4.2.1
5280          July 18, 2007 ([5]changes)
5281
5282   GCC 4.2.0
5283          May 13, 2007 ([6]changes)
5284
5285References and Acknowledgements
5286
5287   GCC used to stand for the GNU C Compiler, but since the compiler
5288   supports several other languages aside from C, it now stands for the
5289   GNU Compiler Collection.
5290
5291   A list of [7]successful builds is updated as new information becomes
5292   available.
5293
5294   The GCC developers would like to thank the numerous people that have
5295   contributed new features, improvements, bug fixes, and other changes as
5296   well as test results to GCC. This [8]amazing group of volunteers is
5297   what makes GCC successful.
5298
5299   For additional information about GCC please refer to the [9]GCC project
5300   web site or contact the [10]GCC development mailing list.
5301
5302   To obtain GCC please use [11]our mirror sites or [12]our SVN server.
5303
5304
5305    For questions related to the use of GCC, please consult these web
5306    pages and the [13]GCC manuals. If that fails, the
5307    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5308    web pages and the development of GCC are welcome on our developer
5309    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
5310    archives.
5311
5312   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
5313   distribution of this entire article is permitted in any medium,
5314   provided this notice is preserved.
5315
5316   These pages are [18]maintained by the GCC team. Last modified
5317   2014-06-28[19].
5318
5319References
5320
5321   1. http://www.gnu.org/
5322   2. http://gcc.gnu.org/gcc-4.2/changes.html
5323   3. http://gcc.gnu.org/gcc-4.2/changes.html
5324   4. http://gcc.gnu.org/gcc-4.2/changes.html
5325   5. http://gcc.gnu.org/gcc-4.2/changes.html
5326   6. http://gcc.gnu.org/gcc-4.2/changes.html
5327   7. http://gcc.gnu.org/gcc-4.2/buildstat.html
5328   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5329   9. http://gcc.gnu.org/index.html
5330  10. mailto:gcc@gcc.gnu.org
5331  11. http://gcc.gnu.org/mirrors.html
5332  12. http://gcc.gnu.org/svn.html
5333  13. https://gcc.gnu.org/onlinedocs/
5334  14. mailto:gcc-help@gcc.gnu.org
5335  15. mailto:gcc@gcc.gnu.org
5336  16. https://gcc.gnu.org/lists.html
5337  17. http://www.fsf.org/
5338  18. https://gcc.gnu.org/about.html
5339  19. http://validator.w3.org/check/referer
5340======================================================================
5341http://gcc.gnu.org/gcc-4.2/changes.html
5342
5343                             GCC 4.2 Release Series
5344                        Changes, New Features, and Fixes
5345
5346Caveats
5347
5348     * GCC no longer accepts the -fshared-data option. This option has had
5349       no effect in any GCC 4 release; the targets to which the option
5350       used to apply had been removed before GCC 4.0.
5351
5352General Optimizer Improvements
5353
5354     * New command-line options specify the possible relationships among
5355       parameters and between parameters and global data. For example,
5356       -fargument-noalias-anything specifies that arguments do not alias
5357       any other storage.
5358       Each language will automatically use whatever option is required by
5359       the language standard. You should not need to use these options
5360       yourself.
5361
5362New Languages and Language specific improvements
5363
5364     * [1]OpenMP is now supported for the C, C++ and Fortran compilers.
5365     * New command-line options -fstrict-overflow and -Wstrict-overflow
5366       have been added. -fstrict-overflow tells the compiler that it may
5367       assume that the program follows the strict signed overflow
5368       semantics permitted for the language: for C and C++ this means that
5369       the compiler may assume that signed overflow does not occur. For
5370       example, a loop like
5371      for (i = 1; i > 0; i *= 2)
5372
5373       is presumably intended to continue looping until i overflows. With
5374       -fstrict-overflow, the compiler may assume that signed overflow
5375       will not occur, and transform this into an infinite loop.
5376       -fstrict-overflow is turned on by default at -O2, and may be
5377       disabled via -fno-strict-overflow. The -Wstrict-overflow option may
5378       be used to warn about cases where the compiler assumes that signed
5379       overflow will not occur. It takes five different levels:
5380       -Wstrict-overflow=1 to 5. See the [2]documentation for details.
5381       -Wstrict-overflow=1 is enabled by -Wall.
5382     * The new command-line option -fno-toplevel-reorder directs GCC to
5383       emit top-level functions, variables, and asm statements in the same
5384       order that they appear in the input file. This is intended to
5385       support existing code which relies on a particular ordering (for
5386       example, code which uses top-level asm statements to switch
5387       sections). For new code, it is generally better to use function and
5388       variable attributes. The -fno-toplevel-reorder option may be used
5389       for most cases which currently use -fno-unit-at-a-time. The
5390       -fno-unit-at-a-time option will be removed in some future version
5391       of GCC. If you know of a case which requires -fno-unit-at-a-time
5392       which is not fixed by -fno-toplevel-reorder, please open a bug
5393       report.
5394
5395  C family
5396
5397     * The pragma redefine_extname will now macro expand its tokens for
5398       compatibility with SunPRO.
5399     * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct
5400       GCC to handle inline functions as specified in the C99 standard. In
5401       preparation for this, GCC 4.2 will warn about any use of non-static
5402       inline functions in gnu99 or c99 mode. This new warning may be
5403       disabled with the new gnu_inline function attribute or the new
5404       -fgnu89-inline command-line option. Also, GCC 4.2 and later will
5405       define one of the preprocessor macros __GNUC_GNU_INLINE__ or
5406       __GNUC_STDC_INLINE__ to indicate the semantics of inline functions
5407       in the current compilation.
5408     * A new command-line option -Waddress has been added to warn about
5409       suspicious uses of memory addresses as, for example, using the
5410       address of a function in a conditional expression, and comparisons
5411       against the memory address of a string literal. This warning is
5412       enabled by -Wall.
5413
5414  C++
5415
5416     * C++ visibility handling has been overhauled.
5417       Restricted visiblity is propagated from classes to members, from
5418       functions to local statics, and from templates and template
5419       arguments to instantiations, unless the latter has explicitly
5420       declared visibility.
5421       The visibility attribute for a class must come between the
5422       class-key and the name, not after the closing brace.
5423       Attributes are now allowed for enums and elaborated-type-specifiers
5424       that only declare a type.
5425       Members of the anonymous namespace are now local to a particular
5426       translation unit, along with any other declarations which use them,
5427       though they are still treated as having external linkage for
5428       language semantics.
5429     * The (undocumented) extension which permitted templates with default
5430       arguments to be bound to template template parameters with fewer
5431       parameters has been removed. For example:
5432        template <template <typename> class C>
5433        void f(C<double>) {}
5434
5435        template <typename T, typename U = int>
5436        struct S {};
5437
5438        template void f(S<double>);
5439
5440       is no longer accepted by G++. The reason this code is not accepted
5441       is that S is a template with two parameters; therefore, it cannot
5442       be bound to C which has only one parameter.
5443     * The <?, >?, <?=, and >?= operators, deprecated in previous GCC
5444       releases, have been removed.
5445     * The command-line option -fconst-strings, deprecated in previous GCC
5446       releases, has been removed.
5447     * The configure variable enable-__cxa_atexit is now enabled by
5448       default for more targets. Enabling this variable is necessary in
5449       order for static destructors to be executed in the correct order,
5450       but it depends upon the presence of a non-standard C library in the
5451       target library in order to work. The variable is now enabled for
5452       more targets which are known to have suitable C libraries.
5453     * -Wextra will produce warnings for if statements with a semicolon as
5454       the only body, to catch code like:
5455         if (a);
5456            return 1;
5457         return 0;
5458
5459       To suppress the warning in valid cases, use { } instead.
5460     * The C++ frontend now also produces strict aliasing warnings when
5461       -fstrict-aliasing -Wstrict-aliasing is in effect.
5462
5463    Runtime Library (libstdc++)
5464
5465     * Added support for TR1 <random>, <complex>, and C compatibility
5466       headers. In addition, a lock-free version of shared_ptr was
5467       contributed as part of Phillip Jordan's Google Summer of Code
5468       project on lock-free containers. ([3]Implementation status of TR1)
5469     * In association with the Summer of Code work on lock-free
5470       containers, the interface for atomic builtins was adjusted,
5471       creating simpler alternatives for non-threaded code paths. Also,
5472       usage was consolidated and all elements were moved from namespace
5473       std to namespace__gnu_cxx. Affected interfaces are the functions
5474       __exchange_and_add, __atomic_add, and the objects __mutex,
5475       __recursive_mutex, and __scoped_lock.
5476     * Support for versioning weak symbol names via namespace association
5477       was added. However, as this changes the names of exported symbols,
5478       this is turned off by default in the current ABI. Intrepid users
5479       can enable this feature by using
5480       --enable-symvers=gnu-versioned-namespace during configuration.
5481     * Revised, simplified, and expanded policy-based associative
5482       containers, including data types for tree and trie forms
5483       (basic_tree, tree, trie), lists (list_update), and both
5484       collision-chaining and probing hash-based containers
5485       (basic_hash_table, cc_hash_table, gp_hash_table). More details per
5486       the [4]documentation.
5487     * The implementation of the debug mode was modified, whereby the
5488       debug namespaces were nested inside of namespace std and namespace
5489       __gnu_cxx in order to resolve some long standing corner cases
5490       involving name lookup. Debug functionality from the policy-based
5491       data structures was consolidated and enabled with the single macro,
5492       _GLIBCXX_DEBUG. See PR 26142 for more information.
5493     * Added extensions for type traits: __conditional_type,
5494       __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if.
5495     * Added a typelist implementation for compile-time meta-programming.
5496       Elements for typelist construction and operation can be found
5497       within namespace __gnu_cxx::typelist.
5498     * Added a new allocator, __gnu_cxx::throw_allocator, for testing
5499       exception-safety.
5500     * Enabled library-wide visibility control, allowing -fvisibility to
5501       be used.
5502     * Consolidated all nested namespaces and the conversion of
5503       __gnu_internal implementation-private details to anonymous
5504       namespaces whenever possible.
5505     * Implemented LWG resolutions DR 431 and DR 538.
5506
5507  Fortran
5508
5509     * Support for allocatable components has been added (TR 15581 and
5510       Fortran 2003).
5511     * Support for the Fortran 2003 streaming IO extension has been added.
5512     * The GNU Fortran compiler now uses 4-byte record markers by default
5513       for unformatted files to be compatible with g77 and most other
5514       compilers. The implementation allows for records greater than 2 GB
5515       and is compatible with several other compilers. Older versions of
5516       gfortran used 8-byte record markers by default (on most systems).
5517       In order to change the length of the record markers, e.g. to read
5518       unformatted files created by older gfortran versions, the
5519       [5]-frecord-marker=8 option can be used.
5520
5521  Java (GCJ)
5522
5523     * A new command-line option -static-libgcj has been added for targets
5524       that use a linker compatible with GNU Binutils. As its name
5525       implies, this causes libgcj to be linked statically. In some cases
5526       this causes the resulting executable to start faster and use less
5527       memory than if the shared version of libgcj were used. However
5528       caution should be used as it can also cause essential parts of the
5529       library to be omitted. Some of these issues are discussed in:
5530       [6]https://gcc.gnu.org/wiki/Statically_linking_libgcj
5531     * fastjar is no longer bundled with GCC. To build libgcj, you will
5532       need either InfoZIP (both zip and unzip) or an external jar
5533       program. In the former case, the GCC build will install a jar shell
5534       script that is based on InfoZIP and provides the same functionality
5535       as fastjar.
5536
5537New Targets and Target Specific Improvements
5538
5539  IA-32/x86-64
5540
5541     * -mtune=generic can now be used to generate code running well on
5542       common x86 chips. This includes AMD Athlon, AMD Opteron, Intel
5543       Pentium-M, Intel Pentium 4 and Intel Core 2.
5544     * -mtune=native and -march=native will produce code optimized for the
5545       host architecture as detected using the cpuid instruction.
5546     * Added a new command-line option -fstackrealign and and
5547       __attribute__ ((force_align_arg_pointer)) to realign the stack at
5548       runtime. This allows functions compiled with a vector-aligned stack
5549       to be invoked from legacy objects that keep only word-alignment.
5550
5551  SPARC
5552
5553     * The default CPU setting has been changed from V7 to V9 in 32-bit
5554       mode on Solaris 7 and above. This is already the case in 64-bit
5555       mode. It can be overridden by specifying --with-cpu at configure
5556       time.
5557     * Back-end support of built-in functions for atomic memory access has
5558       been implemented.
5559     * Support for the Sun UltraSPARC T1 (Niagara) processor has been
5560       added.
5561
5562  M32C
5563
5564     * Various bug fixes have made some functions (notably, functions
5565       returning structures) incompatible with previous releases.
5566       Recompiling all libraries is recommended. Note that code quality
5567       has considerably improved since 4.1, making a recompile even more
5568       beneficial.
5569
5570  MIPS
5571
5572     * Added support for the Broadcom SB-1A core.
5573
5574  IA-64
5575
5576     * Added support for IA-64 data and control speculation. By default
5577       speculation is enabled only during second scheduler pass. A number
5578       of machine flags was introduced to control the usage of speculation
5579       for both scheduler passes.
5580
5581  HPPA
5582
5583     * Added Java language support (libffi and libjava) for 32-bit HP-UX
5584       11 target.
5585
5586Obsolete Systems
5587
5588Documentation improvements
5589
5590  PDF Documentation
5591
5592     * A make pdf target has been added to the top-level makefile,
5593       enabling automated production of PDF documentation files.
5594       (Front-ends external to GCC should modify their Make-lang.in file
5595       to add a lang.pdf: target.)
5596
5597Other significant improvements
5598
5599  Build system improvements
5600
5601     * All the components of the compiler are now bootstrapped by default.
5602       This improves the resilience to bugs in the system compiler or
5603       binary compatibility problems, as well as providing better testing
5604       of GCC 4.2 itself. In addition, if you build the compiler from a
5605       combined tree, the assembler, linker, etc. will also be
5606       bootstrapped (i.e. built with themselves).
5607       You can disable this behavior, and go back to the pre-GCC 4.2 set
5608       up, by configuring GCC with --disable-bootstrap.
5609     * The rules that configure follows to find target tools resemble more
5610       closely the locations that the built compiler will search. In
5611       addition, you can use the new configure option --with-target-tools
5612       to specify where to find the target tools used during the build,
5613       without affecting what the built compiler will use.
5614       This can be especially useful when building packages of GCC. For
5615       example, you may want to build GCC with GNU as or ld, even if the
5616       resulting compiler to work with the native assembler and linker. To
5617       do so, you can use --with-target-tools to point to the native
5618       tools.
5619
5620  Incompatible changes to the build system
5621
5622     * Front-ends external to GCC should modify their Make-lang.in file to
5623       replace double-colon rules (e.g. dvi::) with normal rules (like
5624       lang.dvi:). Front-end makefile hooks do not use double-colon rules
5625       anymore.
5626     * Up to GCC 4.1, a popular way to specify the target tools used
5627       during the build was to create directories named gas, binutils,
5628       etc. in the build tree, and create links to the tools from there.
5629       This does not work any more when the compiler is bootstrapped. The
5630       new configure option --with-target-tools provides a better way to
5631       achieve the same effect, and works for all native and cross
5632       settings.
5633
5634
5635    For questions related to the use of GCC, please consult these web
5636    pages and the [7]GCC manuals. If that fails, the
5637    [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5638    web pages and the development of GCC are welcome on our developer
5639    list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
5640    archives.
5641
5642   Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
5643   distribution of this entire article is permitted in any medium,
5644   provided this notice is preserved.
5645
5646   These pages are [12]maintained by the GCC team. Last modified
5647   2014-06-28[13].
5648
5649References
5650
5651   1. http://gcc.gnu.org/projects/gomp/
5652   2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
5653   3. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#manual.intro.status.standard.tr1
5654   4. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html
5655   5. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html
5656   6. https://gcc.gnu.org/wiki/Statically_linking_libgcj
5657   7. https://gcc.gnu.org/onlinedocs/
5658   8. mailto:gcc-help@gcc.gnu.org
5659   9. mailto:gcc@gcc.gnu.org
5660  10. https://gcc.gnu.org/lists.html
5661  11. http://www.fsf.org/
5662  12. https://gcc.gnu.org/about.html
5663  13. http://validator.w3.org/check/referer
5664======================================================================
5665http://gcc.gnu.org/gcc-4.1/index.html
5666
5667                             GCC 4.1 Release Series
5668
5669   February 13, 2007
5670
5671   The [1]GNU project and the GCC developers are pleased to announce the
5672   release of GCC 4.1.2.
5673
5674   This release is a bug-fix release, containing fixes for regressions in
5675   GCC 4.1.1 relative to previous releases of GCC.
5676
5677Release History
5678
5679   GCC 4.1.2
5680          February 13, 2007 ([2]changes)
5681
5682   GCC 4.1.1
5683          May 24, 2006 ([3]changes)
5684
5685   GCC 4.1.0
5686          February 28, 2006 ([4]changes)
5687
5688References and Acknowledgements
5689
5690   GCC used to stand for the GNU C Compiler, but since the compiler
5691   supports several other languages aside from C, it now stands for the
5692   GNU Compiler Collection.
5693
5694   A list of [5]successful builds is updated as new information becomes
5695   available.
5696
5697   The GCC developers would like to thank the numerous people that have
5698   contributed new features, improvements, bug fixes, and other changes as
5699   well as test results to GCC. This [6]amazing group of volunteers is
5700   what makes GCC successful.
5701
5702   For additional information about GCC please refer to the [7]GCC project
5703   web site or contact the [8]GCC development mailing list.
5704
5705   To obtain GCC please use [9]our mirror sites or [10]our SVN server.
5706
5707
5708    For questions related to the use of GCC, please consult these web
5709    pages and the [11]GCC manuals. If that fails, the
5710    [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5711    web pages and the development of GCC are welcome on our developer
5712    list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
5713    archives.
5714
5715   Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
5716   distribution of this entire article is permitted in any medium,
5717   provided this notice is preserved.
5718
5719   These pages are [16]maintained by the GCC team. Last modified
5720   2014-06-28[17].
5721
5722References
5723
5724   1. http://www.gnu.org/
5725   2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
5726   3. http://gcc.gnu.org/gcc-4.1/changes.html
5727   4. http://gcc.gnu.org/gcc-4.1/changes.html
5728   5. http://gcc.gnu.org/gcc-4.1/buildstat.html
5729   6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5730   7. http://gcc.gnu.org/index.html
5731   8. mailto:gcc@gcc.gnu.org
5732   9. http://gcc.gnu.org/mirrors.html
5733  10. http://gcc.gnu.org/svn.html
5734  11. https://gcc.gnu.org/onlinedocs/
5735  12. mailto:gcc-help@gcc.gnu.org
5736  13. mailto:gcc@gcc.gnu.org
5737  14. https://gcc.gnu.org/lists.html
5738  15. http://www.fsf.org/
5739  16. https://gcc.gnu.org/about.html
5740  17. http://validator.w3.org/check/referer
5741======================================================================
5742http://gcc.gnu.org/gcc-4.1/changes.html
5743
5744                             GCC 4.1 Release Series
5745                        Changes, New Features, and Fixes
5746
5747   The latest release in the 4.1 release series is [1]GCC 4.1.2.
5748
5749Caveats
5750
5751General Optimizer Improvements
5752
5753     * GCC now has infrastructure for inter-procedural optimizations and
5754       the following inter-procedural optimizations are implemented:
5755          + Profile guided inlining. When doing profile feedback guided
5756            optimization, GCC can now use the profile to make better
5757            informed decisions on whether inlining of a function is
5758            profitable or not. This means that GCC will no longer inline
5759            functions at call sites that are not executed very often, and
5760            that functions at hot call sites are more likely to be
5761            inlined.
5762            A new parameter min-inline-recursive-probability is also now
5763            available to throttle recursive inlining of functions with
5764            small average recursive depths.
5765          + Discovery of pure and const functions, a form of side-effects
5766            analysis. While older GCC releases could also discover such
5767            special functions, the new IPA-based pass runs earlier so that
5768            the results are available to more optimizers. The pass is also
5769            simply more powerful than the old one.
5770          + Analysis of references to static variables and type escape
5771            analysis, also forms of side-effects analysis. The results of
5772            these passes allow the compiler to be less conservative about
5773            call-clobbered variables and references. This results in more
5774            redundant loads being eliminated and in making static
5775            variables candidates for register promotion.
5776          + Improvement of RTL-based alias analysis. The results of type
5777            escape analysis are fed to the RTL type-based alias analyzer,
5778            allowing it to disambiguate more memory references.
5779          + Interprocedural constant propagation and function versioning.
5780            This pass looks for functions that are always called with the
5781            same constant value for one or more of the function arguments,
5782            and propagates those constants into those functions.
5783          + GCC will now eliminate static variables whose usage was
5784            optimized out.
5785          + -fwhole-program --combine can now be used to make all
5786            functions in program static allowing whole program
5787            optimization. As an exception, the main function and all
5788            functions marked with the new externally_visible attribute are
5789            kept global so that programs can link with runtime libraries.
5790     * GCC can now do a form of partial dead code elimination (PDCE) that
5791       allows code motion of expressions to the paths where the result of
5792       the expression is actually needed. This is not always a win, so the
5793       pass has been limited to only consider profitable cases. Here is an
5794       example:
5795    int foo (int *, int *);
5796    int
5797    bar (int d)
5798    {
5799      int a, b, c;
5800      b = d + 1;
5801      c = d + 2;
5802      a = b + c;
5803      if (d)
5804        {
5805          foo (&b, &c);
5806          a = b + c;
5807        }
5808      printf ("%d\n", a);
5809    }
5810
5811       The a = b + c can be sunk to right before the printf. Normal code
5812       sinking will not do this, it will sink the first one above into the
5813       else-branch of the conditional jump, which still gives you two
5814       copies of the code.
5815     * GCC now has a value range propagation pass. This allows the
5816       compiler to eliminate bounds checks and branches. The results of
5817       the pass can also be used to accurately compute branch
5818       probabilities.
5819     * The pass to convert PHI nodes to straight-line code (a form of
5820       if-conversion for GIMPLE) has been improved significantly. The two
5821       most significant improvements are an improved algorithm to
5822       determine the order in which the PHI nodes are considered, and an
5823       improvement that allow the pass to consider if-conversions of basic
5824       blocks with more than two predecessors.
5825     * Alias analysis improvements. GCC can now differentiate between
5826       different fields of structures in Tree-SSA's virtual operands form.
5827       This lets stores/loads from non-overlapping structure fields not
5828       conflict. A new algorithm to compute points-to sets was contributed
5829       that can allows GCC to see now that p->a and p->b, where p is a
5830       pointer to a structure, can never point to the same field.
5831     * Various enhancements to auto-vectorization:
5832          + Incrementally preserve SSA form when vectorizing.
5833          + Incrementally preserve loop-closed form when vectorizing.
5834          + Improvements to peeling for alignment: generate better code
5835            when the misalignment of an access is known at compile time,
5836            or when different accesses are known to have the same
5837            misalignment, even if the misalignment amount itself is
5838            unknown.
5839          + Consider dependence distance in the vectorizer.
5840          + Externalize generic parts of data reference analysis to make
5841            this analysis available to other passes.
5842          + Vectorization of conditional code.
5843          + Reduction support.
5844     * GCC can now partition functions in sections of hot and cold code.
5845       This can significantly improve performance due to better
5846       instruction cache locality. This feature works best together with
5847       profile feedback driven optimization.
5848     * A new pass to avoid saving of unneeded arguments to the stack in
5849       vararg functions if the compiler can prove that they will not be
5850       needed.
5851     * Transition of basic block profiling to tree level implementation
5852       has been completed. The new implementation should be considerably
5853       more reliable (hopefully avoiding profile mismatch errors when
5854       using -fprofile-use or -fbranch-probabilities) and can be used to
5855       drive higher level optimizations, such as inlining.
5856       The -ftree-based-profiling command-line option was removed and
5857       -fprofile-use now implies disabling old RTL level loop optimizer
5858       (-fno-loop-optimize). Speculative prefetching optimization
5859       (originally enabled by -fspeculative-prefetching) was removed.
5860
5861New Languages and Language specific improvements
5862
5863  C and Objective-C
5864
5865     * The old Bison-based C and Objective-C parser has been replaced by a
5866       new, faster hand-written recursive-descent parser.
5867
5868  Ada
5869
5870     * The build infrastructure for the Ada runtime library and tools has
5871       been changed to be better integrated with the rest of the build
5872       infrastructure of GCC. This should make doing cross builds of Ada a
5873       bit easier.
5874
5875  C++
5876
5877     * ARM-style name-injection of friend declarations is no longer the
5878       default. For example:
5879          struct S {
5880            friend void f();
5881          };
5882
5883          void g() { f(); }
5884       will not be accepted; instead a declaration of f will need to be
5885       present outside of the scope of S. The new -ffriend-injection
5886       option will enable the old behavior.
5887     * The (undocumented) extension which permitted templates with default
5888       arguments to be bound to template template parameters with fewer
5889       parameters has been deprecated, and will be removed in the next
5890       major release of G++. For example:
5891       template <template <typename> class C>
5892       void f(C<double>) {}
5893
5894       template <typename T, typename U = int>
5895       struct S {};
5896
5897       template void f(S<double>);
5898
5899       makes use of the deprecated extension. The reason this code is not
5900       valid ISO C++ is that S is a template with two parameters;
5901       therefore, it cannot be bound to C which has only one parameter.
5902
5903    Runtime Library (libstdc++)
5904
5905     * Optimization work:
5906          + A new implementation of std::search_n is provided, better
5907            performing in case of random access iterators.
5908          + Added further efficient specializations of istream functions,
5909            i.e., character array and string extractors.
5910          + Other smaller improvements throughout.
5911     * Policy-based associative containers, designed for high-performance,
5912       flexibility and semantic safety are delivered in ext/pb_assoc.
5913     * A versatile string class, __gnu_cxx::__versa_string, providing
5914       facilities conforming to the standard requirements for
5915       basic_string, is delivered in <ext/vstring.h>. In particular:
5916          + Two base classes are provided: the default one avoids
5917            reference counting and is optimized for short strings; the
5918            alternate one, still uses it while improving in a few low
5919            level areas (e.g., alignment). See vstring_fwd.h for some
5920            useful typedefs.
5921          + Various algorithms have been rewritten (e.g., replace), the
5922            code streamlined and simple optimizations added.
5923          + Option 3 of DR 431 is implemented for both available bases,
5924            thus improving the support for stateful allocators.
5925     * As usual, many bugs have been fixed (e.g., libstdc++/13583,
5926       libstdc++/23953) and LWG resolutions put into effect for the first
5927       time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1
5928       Issue 6.19). The implementation status of TR1 is now tracked in the
5929       docs in tr1.html.
5930
5931  Objective-C++
5932
5933     * A new language front end for Objective-C++ has been added. This
5934       language allows users to mix the object oriented features of
5935       Objective-C with those of C++.
5936
5937  Java (GCJ)
5938
5939     * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19
5940       features (plus some 0.20 bug-fixes)
5941          + Networking
5942               o The java.net.HttpURLConnection implementation no longer
5943                 buffers the entire response body in memory. This means
5944                 that response bodies larger than available memory can now
5945                 be handled.
5946          + (N)IO
5947               o NIO FileChannel.map implementation, fast bulk put
5948                 implementation for DirectByteBuffer (speeds up this
5949                 method 10x).
5950               o FileChannel.lock() and FileChannel.force() implemented.
5951          + XML
5952               o gnu.xml fix for nodes created outside a namespace
5953                 context.
5954               o Add support for output indenting and
5955                 cdata-section-elements output instruction in
5956                 xml.transform.
5957               o xml.xpath corrections for cases where elements/attributes
5958                 might have been created in non-namespace-aware mode.
5959                 Corrections to handling of XSL variables and minor
5960                 conformance updates.
5961          + AWT
5962               o GNU JAWT implementation, the AWT Native Interface, which
5963                 allows direct access to native screen resources from
5964                 within a Canvas's paint method. GNU Classpath Examples
5965                 comes with a Demo, see libjava/classpath/examples/README.
5966               o awt.datatransfer updated to 1.5 with support for
5967                 FlavorEvents. The gtk+ awt peers now allow copy/paste of
5968                 text, images, URIs/files and serialized objects with
5969                 other applications and tracking clipboard change events
5970                 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized
5971                 objects are supported). A GNU Classpath Examples
5972                 datatransfer Demo was added to show the new
5973                 functionality.
5974               o Split gtk+ awt peers event handling in two threads and
5975                 improve gdk lock handling (solves several awt lock ups).
5976               o Speed up awt Image loading.
5977               o Better gtk+ scrollbar peer implementation when using gtk+
5978                 >= 2.6.
5979               o Handle image loading errors correctly for gdkpixbuf and
5980                 MediaTracker.
5981               o Better handle GDK lock. Properly prefix gtkpeer native
5982                 functions (cp_gtk).
5983               o GdkGraphics2D has been updated to use Cairo 0.5.x or
5984                 higher.
5985               o BufferedImage and GtkImage rewrites. All image drawing
5986                 operations should now work correctly (flipping requires
5987                 gtk+ >= 2.6)
5988               o Future Graphics2D, image and text work is documented at:
5989                 [2]http://developer.classpath.org/mediation/ClasspathGrap
5990                 hicsImagesText
5991               o When gtk+ 2.6 or higher is installed the default log
5992                 handler will produce stack traces whenever a WARNING,
5993                 CRITICAL or ERROR message is produced.
5994          + Free Swing
5995               o The RepaintManager has been reworked for more efficient
5996                 painting, especially for large GUIs.
5997               o The layout manager OverlayLayout has been implemented,
5998                 the BoxLayout has been rewritten to make use of the
5999                 SizeRequirements utility class and caching for more
6000                 efficient layout.
6001               o Improved accessibility support.
6002               o Significant progress has been made in the implementation
6003                 of the javax.swing.plaf.metal package, with most UI
6004                 delegates in a working state now. Please test this with
6005                 your own applications and provide feedback that will help
6006                 us to improve this package.
6007               o The GUI demo (gnu.classpath.examples.swing.Demo) has been
6008                 extended to highlight various features in our Free Swing
6009                 implementation. And it includes a look and feel switcher
6010                 for Metal (default), Ocean and GNU themes.
6011               o The javax.swing.plaf.multi package is now implemented.
6012               o Editing and several key actions for JTree and JTable were
6013                 implemented.
6014               o Lots of icons and look and feel improvements for Free
6015                 Swing basic and metal themes were added. Try running the
6016                 GNU Classpath Swing Demo in examples
6017                 (gnu.classpath.examples.swing.Demo) with:
6018                 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee
6019                 l or
6020                 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee
6021                 l
6022               o Start of styled text capabilites for java.swing.text.
6023               o DefaultMutableTreeNode pre-order, post-order, depth-first
6024                 and breadth-first traversal enumerations implemented.
6025               o JInternalFrame colors and titlebar draw properly.
6026               o JTree is working up to par (icons, selection and keyboard
6027                 traversal).
6028               o JMenus were made more compatible in visual and
6029                 programmatic behavior.
6030               o JTable changeSelection and multiple selections
6031                 implemented.
6032               o JButton and JToggleButton change states work properly
6033                 now.
6034               o JFileChooser fixes.
6035               o revalidate() and repaint() fixes which make Free Swing
6036                 much more responsive.
6037               o MetalIconFactory implemented.
6038               o Free Swing Top-Level Compatibility. JFrame, JDialog,
6039                 JApplet, JInternalFrame, and JWindow are now 1.5
6040                 compatible in the sense that you can call add() and
6041                 setLayout() directly on them, which will have the same
6042                 effect as calling getContentPane().add() and
6043                 getContentPane().setLayout().
6044               o The JTree interface has been completed. JTrees now
6045                 recognizes mouse clicks and selections work.
6046               o BoxLayout works properly now.
6047               o Fixed GrayFilter to actually work.
6048               o Metal SplitPane implemented.
6049               o Lots of Free Swing text and editor stuff work now.
6050          + Free RMI and Corba
6051               o Andrew Watson, Vice President and Technical Director of
6052                 the Object Management Group, has officially assigned us
6053                 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will
6054                 mark remote classpath-specific system exceptions.
6055                 Obtaining the VMCID means that GNU Classpath now is a
6056                 recogniseable type of node in a highly interoperable
6057                 CORBA world.
6058               o GNU Classpath now includes the first working draft to
6059                 support the RMI over IIOP protocol. The current
6060                 implementation is capable of remote invocations,
6061                 transferring various Serializables and Externalizables
6062                 via RMI-IIOP protocol. It can flatten graphs and, at
6063                 least for the simple cases, is interoperable with 1.5
6064                 JDKs.
6065               o org.omg.PortableInterceptor and related functionality in
6066                 other packages is now implemented:
6067                    # The sever and client interceptors work as required
6068                      since 1.4.
6069                    # The IOR interceptor works as needed for 1.5.
6070               o The org.omg.DynamicAny package is completed and passes
6071                 the prepared tests.
6072               o The Portable Object Adapter should now support the output
6073                 of the recent IDL to java compilers. These compilers now
6074                 generate servants and not CORBA objects as before, making
6075                 the output depend on the existing POA implementation.
6076                 Completing POA means that such code can already be tried
6077                 to run on Classpath. Our POA is tested for the following
6078                 usager scenarios:
6079                    # POA converts servant to the CORBA object.
6080                    # Servant provides to the CORBA object.
6081                    # POA activates new CORBA object with the given Object
6082                      Id (byte array) that is later accessible for the
6083                      servant.
6084                    # During the first call, the ServantActivator provides
6085                      servant for this and all subsequent calls on the
6086                      current object.
6087                    # During each call, the ServantLocator provides
6088                      servant for this call only.
6089                    # ServantLocator or ServantActivator forwards call to
6090                      another server.
6091                    # POA has a single servant, responsible for all
6092                      objects.
6093                    # POA has a default servant, but some objects are
6094                      explicitly connected to they specific servants.
6095                 The POA is verified using tests from the former
6096                 cost.omg.org.
6097               o The CORBA implementation is now a working prototype that
6098                 should support features up to 1.3 inclusive. We invite
6099                 groups writing CORBA dependent applications to try
6100                 Classpath implementation, reporting any possible bugs.
6101                 The CORBA prototype is interoperable with Sun's
6102                 implementation v 1.4, transferring object references,
6103                 primitive types, narrow and wide strings, arrays,
6104                 structures, trees, abstract interfaces and value types
6105                 (feature of CORBA 2.3) between these two platforms.
6106                 Remote exceptions are transferred and handled correctly.
6107                 The stringified object references (IORs) from various
6108                 sources are parsed as required. The transient (for
6109                 current session) and permanent (till jre restart)
6110                 redirections work. Both Little and Big Endian encoded
6111                 messages are accepted. The implementation is verified
6112                 using tests from the former cost.omg.org. The current
6113                 release includes working examples (see the examples
6114                 directory), demonstrating the client-server
6115                 communication, using either CORBA Request or IDL-based
6116                 stub (usually generated by a IDL to java compiler). These
6117                 examples also show how to use the Classpath CORBA naming
6118                 service. The IDL to java compiler is not yet written, but
6119                 as our library must be compatible, it naturally accepts
6120                 the output of other idlj implementations.
6121          + Misc
6122               o Updated TimeZone data against Olson tzdata2005l.
6123               o Make zip and jar packages UTF-8 clean.
6124               o "native" code builds and compiles (warning free) on
6125                 Darwin and Solaris.
6126               o java.util.logging.FileHandler now rotates files.
6127               o Start of a generic JDWP framework in gnu/classpath/jdwp.
6128                 This is unfinished, but feedback (at classpath@gnu.org)
6129                 from runtime hackers is greatly appreciated. Although
6130                 most of the work is currently being done around gcj/gij
6131                 we want this framework to be as VM neutral as possible.
6132                 Early design is described in:
6133                 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html
6134               o QT4 AWT peers, enable by giving configure
6135                 --enable-qt-peer. Included, but not ready for production
6136                 yet. They are explicitly disabled and not supported. But
6137                 if you want to help with the development of these new
6138                 features we are interested in feedback. You will have to
6139                 explicitly enable them to try them out (and they will
6140                 most likely contain bugs).
6141               o Documentation fixes all over the place. See
6142                 [4]http://developer.classpath.org/doc/
6143
6144New Targets and Target Specific Improvements
6145
6146  IA-32/x86-64
6147
6148     * The x86-64 medium model (that allows building applications whose
6149       data segment exceeds 4GB) was redesigned to match latest ABI draft.
6150       New implementation split large datastructures into separate segment
6151       improving performance of accesses to small datastructures and also
6152       allows linking of small model libraries into medium model programs
6153       as long as the libraries are not accessing the large datastructures
6154       directly. Medium model is also supported in position independent
6155       code now.
6156       The ABI change results in partial incompatibility among medium
6157       model objects. Linking medium model libraries (or objects) compiled
6158       with new compiler into medium model program compiled with older
6159       will likely result in exceeding ranges of relocations.
6160       Binutils 2.16.91 or newer are required for compiling medium model
6161       now.
6162
6163  RS6000 (POWER/PowerPC)
6164
6165     * The AltiVec vector primitives in <altivec.h> are now implemented in
6166       a way that puts a smaller burden on the preprocessor, instead
6167       processing the "overloading" in the front ends. This should benefit
6168       compilation speed on AltiVec vector code.
6169     * AltiVec initializers now are generated more efficiently.
6170     * The popcountb instruction available on POWER5 now is generated.
6171     * The floating point round to integer instructions available on
6172       POWER5+ now is generated.
6173     * Floating point divides can be synthesized using the floating point
6174       reciprocal estimate instructions.
6175     * Double precision floating point constants are initialized as single
6176       precision values if they can be represented exactly.
6177
6178  S/390, zSeries and System z9
6179
6180     * Support for the IBM System z9 109 processor has been added. When
6181       using the -march=z9-109 option, the compiler will generate code
6182       making use of instructions provided by the extended immediate
6183       facility.
6184     * Support for 128-bit IEEE floating point has been added. When using
6185       the -mlong-double-128 option, the compiler will map the long double
6186       data type to 128-bit IEEE floating point. Using this option
6187       constitutes an ABI change, and requires glibc support.
6188     * Various changes to improve performance of generated code have been
6189       implemented, including:
6190          + In functions that do not require a literal pool, register %r13
6191            (which is traditionally reserved as literal pool pointer), can
6192            now be freely used for other purposes by the compiler.
6193          + More precise tracking of register use allows the compiler to
6194            generate more efficient function prolog and epilog code in
6195            certain cases.
6196          + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING
6197            instructions are now used to implement C string functions.
6198          + The MOVE CHARACTER instruction with single byte overlap is now
6199            used to implement the memset function with non-zero fill byte.
6200          + The LOAD ZERO instructions are now used where appropriate.
6201          + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK,
6202            and INSERT IMMEDIATE instructions are now used more frequently
6203            to optimize bitfield operations.
6204          + The BRANCH ON COUNT instruction is now used more frequently.
6205            In particular, the fact that a loop contains a subroutine call
6206            no longer prevents the compiler from using this instruction.
6207          + The compiler is now aware that all shift and rotate
6208            instructions implicitly truncate the shift count to six bits.
6209     * Back-end support for the following generic features has been
6210       implemented:
6211          + The full set of [5]built-in functions for atomic memory
6212            access.
6213          + The -fstack-protector feature.
6214          + The optimization pass avoiding unnecessary stores of incoming
6215            argument registers in functions with variable argument list.
6216
6217  SPARC
6218
6219     * The default code model in 64-bit mode has been changed from
6220       Medium/Anywhere to Medium/Middle on Solaris.
6221     * TLS support is disabled by default on Solaris prior to release 10.
6222       It can be enabled on TLS-capable Solaris 9 versions (4/04 release
6223       and later) by specifying --enable-tls at configure time.
6224
6225  MorphoSys
6226
6227     * Support has been added for this new architecture.
6228
6229Obsolete Systems
6230
6231Documentation improvements
6232
6233Other significant improvements
6234
6235     * GCC can now emit code for protecting applications from
6236       stack-smashing attacks. The protection is realized by buffer
6237       overflow detection and reordering of stack variables to avoid
6238       pointer corruption.
6239     * Some built-in functions have been fortified to protect them against
6240       various buffer overflow (and format string) vulnerabilities.
6241       Compared to the mudflap bounds checking feature, the safe builtins
6242       have far smaller overhead. This means that programs built using
6243       safe builtins should not experience any measurable slowdown.
6244
6245GCC 4.1.2
6246
6247   This is the [6]list of problem reports (PRs) from GCC's bug tracking
6248   system that are known to be fixed in the 4.1.2 release. This list might
6249   not be complete (that is, it is possible that some PRs that have been
6250   fixed are not listed here).
6251
6252   When generating code for a shared library, GCC now recognizes that
6253   global functions may be replaced when the program runs. Therefore, it
6254   is now more conservative in deducing information from the bodies of
6255   functions. For example, in this example:
6256    void f() {}
6257    void g() {
6258     try { f(); }
6259     catch (...) {
6260       cout << "Exception";
6261     }
6262    }
6263
6264   G++ would previously have optimized away the catch clause, since it
6265   would have concluded that f cannot throw exceptions. Because users may
6266   replace f with another function in the main body of the program, this
6267   optimization is unsafe, and is no longer performed. If you wish G++ to
6268   continue to optimize as before, you must add a throw() clause to the
6269   declaration of f to make clear that it does not throw exceptions.
6270
6271
6272    For questions related to the use of GCC, please consult these web
6273    pages and the [7]GCC manuals. If that fails, the
6274    [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6275    web pages and the development of GCC are welcome on our developer
6276    list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
6277    archives.
6278
6279   Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
6280   distribution of this entire article is permitted in any medium,
6281   provided this notice is preserved.
6282
6283   These pages are [12]maintained by the GCC team. Last modified
6284   2014-06-28[13].
6285
6286References
6287
6288   1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
6289   2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText
6290   3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html
6291   4. http://developer.classpath.org/doc/
6292   5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
6293   6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2
6294   7. https://gcc.gnu.org/onlinedocs/
6295   8. mailto:gcc-help@gcc.gnu.org
6296   9. mailto:gcc@gcc.gnu.org
6297  10. https://gcc.gnu.org/lists.html
6298  11. http://www.fsf.org/
6299  12. https://gcc.gnu.org/about.html
6300  13. http://validator.w3.org/check/referer
6301======================================================================
6302http://gcc.gnu.org/gcc-4.0/index.html
6303
6304                             GCC 4.0 Release Series
6305
6306   January 31, 2007
6307
6308   The [1]GNU project and the GCC developers are pleased to announce the
6309   release of GCC 4.0.4.
6310
6311   This release is a bug-fix release, containing fixes for regressions in
6312   GCC 4.0.3 relative to previous releases of GCC.
6313
6314Release History
6315
6316   GCC 4.0.4
6317          January 31, 2007 ([2]changes)
6318
6319   GCC 4.0.3
6320          March 10, 2006 ([3]changes)
6321
6322   GCC 4.0.2
6323          September 28, 2005 ([4]changes)
6324
6325   GCC 4.0.1
6326          July 7, 2005 ([5]changes)
6327
6328   GCC 4.0.0
6329          April 20, 2005 ([6]changes)
6330
6331References and Acknowledgements
6332
6333   GCC used to stand for the GNU C Compiler, but since the compiler
6334   supports several other languages aside from C, it now stands for the
6335   GNU Compiler Collection.
6336
6337   A list of [7]successful builds is updated as new information becomes
6338   available.
6339
6340   The GCC developers would like to thank the numerous people that have
6341   contributed new features, improvements, bug fixes, and other changes as
6342   well as test results to GCC. This [8]amazing group of volunteers is
6343   what makes GCC successful.
6344
6345   For additional information about GCC please refer to the [9]GCC project
6346   web site or contact the [10]GCC development mailing list.
6347
6348   To obtain GCC please use [11]our mirror sites, or [12]our SVN server.
6349
6350
6351    For questions related to the use of GCC, please consult these web
6352    pages and the [13]GCC manuals. If that fails, the
6353    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6354    web pages and the development of GCC are welcome on our developer
6355    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
6356    archives.
6357
6358   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
6359   distribution of this entire article is permitted in any medium,
6360   provided this notice is preserved.
6361
6362   These pages are [18]maintained by the GCC team. Last modified
6363   2014-06-28[19].
6364
6365References
6366
6367   1. http://www.gnu.org/
6368   2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
6369   3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3
6370   4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2
6371   5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1
6372   6. http://gcc.gnu.org/gcc-4.0/changes.html
6373   7. http://gcc.gnu.org/gcc-4.0/buildstat.html
6374   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6375   9. http://gcc.gnu.org/index.html
6376  10. mailto:gcc@gcc.gnu.org
6377  11. http://gcc.gnu.org/mirrors.html
6378  12. http://gcc.gnu.org/svn.html
6379  13. https://gcc.gnu.org/onlinedocs/
6380  14. mailto:gcc-help@gcc.gnu.org
6381  15. mailto:gcc@gcc.gnu.org
6382  16. https://gcc.gnu.org/lists.html
6383  17. http://www.fsf.org/
6384  18. https://gcc.gnu.org/about.html
6385  19. http://validator.w3.org/check/referer
6386======================================================================
6387http://gcc.gnu.org/gcc-4.0/changes.html
6388
6389                             GCC 4.0 Release Series
6390                        Changes, New Features, and Fixes
6391
6392   The latest release in the 4.0 release series is [1]GCC 4.0.4.
6393
6394Caveats
6395
6396     * GCC now generates location lists by default when compiling with
6397       debug info and optimization.
6398          + GDB 6.0 and older crashes when it sees location lists. GDB 6.1
6399            or later is needed to debug binaries containing location
6400            lists.
6401          + When you are trying to view a value of a variable in a part of
6402            a function where it has no location (for example when the
6403            variable is no longer used and thus its location was used for
6404            something else) GDB will say that it is not available.
6405       You can disable generating location lists by -fno-var-tracking.
6406     * GCC no longer accepts the -fwritable-strings option. Use named
6407       character arrays when you need a writable string.
6408     * The options -freduce-all-givs and -fmove-all-movables have been
6409       discontinued. They were used to circumvent a shortcoming in the
6410       heuristics of the old loop optimization code with respect to common
6411       Fortran constructs. The new (tree) loop optimizer works differently
6412       and doesn't need those work-arounds.
6413     * The graph-coloring register allocator, formerly enabled by the
6414       option -fnew-ra, has been discontinued.
6415     * -I- has been deprecated. -iquote is meant to replace the need for
6416       this option.
6417     * The MIPS -membedded-pic and -mrnames options have been removed.
6418     * All MIPS targets now require the GNU assembler. In particular, IRIX
6419       configurations can no longer use the MIPSpro assemblers, although
6420       they do still support the MIPSpro linkers.
6421     * The SPARC option -mflat has been removed.
6422     * English-language diagnostic messages will now use Unicode quotation
6423       marks in UTF-8 locales. (Non-English messages already used the
6424       quotes appropriate for the language in previous releases.) If your
6425       terminal does not support UTF-8 but you are using a UTF-8 locale
6426       (such locales are the default on many GNU/Linux systems) then you
6427       should set LC_CTYPE=C in the environment to disable that locale.
6428       Programs that parse diagnostics and expect plain ASCII
6429       English-language messages should set LC_ALL=C. See [2]Markus Kuhn's
6430       explanation of Unicode quotation marks for more information.
6431     * The specs file is no longer installed on most platforms. Most users
6432       will be totally unaffected. However, if you are accustomed to
6433       editing the specs file yourself, you will now have to use the
6434       -dumpspecs option to generate the specs file, and then edit the
6435       resulting file.
6436
6437General Optimizer Improvements
6438
6439     * The [3]tree ssa branch has been merged. This merge has brought in a
6440       completely new optimization framework based on a higher level
6441       intermediate representation than the existing RTL representation.
6442       Numerous new code transformations based on the new framework are
6443       available in GCC 4.0, including:
6444          + Scalar replacement of aggregates
6445          + Constant propagation
6446          + Value range propagation
6447          + Partial redundancy elimination
6448          + Load and store motion
6449          + Strength reduction
6450          + Dead store elimination
6451          + Dead and unreachable code elimination
6452          + [4]Autovectorization
6453          + Loop interchange
6454          + Tail recursion by accumulation
6455       Many of these passes outperform their counterparts from previous
6456       GCC releases.
6457     * [5]Swing Modulo Scheduling (SMS). An RTL level instruction
6458       scheduling optimization intended for loops that perform heavy
6459       computations.
6460
6461New Languages and Language specific improvements
6462
6463  C family
6464
6465     * The sentinel attribute has been added to GCC. This function
6466       attribute allows GCC to warn when variadic functions such as execl
6467       are not NULL terminated. See the GCC manual for a complete
6468       description of its behavior.
6469     * Given __attribute__((alias("target"))) it is now an error if target
6470       is not a symbol, defined in the same translation unit. This also
6471       applies to aliases created by #pragma weak alias=target. This is
6472       because it's meaningless to define an alias to an undefined symbol.
6473       On Solaris, the native assembler would have caught this error, but
6474       GNU as does not.
6475
6476  C and Objective-C
6477
6478     * The -Wstrict-aliasing=2 option has been added. This warning catches
6479       all unsafe cases, but it may also give a warning for some cases
6480       that are safe.
6481     * The cast-as-lvalue, conditional-expression-as-lvalue and
6482       compound-expression-as-lvalue extensions, which were deprecated in
6483       3.3.4 and 3.4, have been removed.
6484     * The -fwritable-strings option, which was deprecated in 3.4, has
6485       been removed.
6486     * #pragma pack() semantics have been brought closer to those used by
6487       other compilers. This also applies to C++.
6488     * Taking the address of a variable with register storage is invalid
6489       in C. GCC now issues an error instead of a warning.
6490     * Arrays of incomplete element type are invalid in C. GCC now issues
6491       an error for such arrays. Declarations such as extern struct s x[];
6492       (where struct s has not been defined) can be moved after the
6493       definition of struct s. Function parameters declared as arrays of
6494       incomplete type can instead be declared as pointers.
6495
6496  C++
6497
6498     * When compiling without optimizations (-O0), the C++ frontend is
6499       much faster than in any previous versions of GCC. Independent
6500       testers have measured speed-ups up to 25% in real-world production
6501       code, compared to the 3.4 family (which was already the fastest
6502       version to date). Upgrading from older versions might show even
6503       bigger improvements.
6504     * ELF visibility attributes can now be applied to a class type, so
6505       that it affects every member function of a class at once, without
6506       having to specify each individually:
6507class __attribute__ ((visibility("hidden"))) Foo
6508{
6509   int foo1();
6510   void foo2();
6511};
6512       The syntax is deliberately similar to the __declspec() system used
6513       by Microsoft Windows based compilers, allowing cross-platform
6514       projects to easily reuse their existing macro system for denoting
6515       exports and imports. By explicitly marking internal classes never
6516       used outside a binary as hidden, one can completely avoid PLT
6517       indirection overheads during their usage by the compiler. You can
6518       find out more about the advantages of this at
6519       [6]http://www.akkadia.org/drepper/dsohowto.pdf
6520     * The -fvisibility-inlines-hidden option has been added which marks
6521       all inlineable functions as having hidden ELF visibility, thus
6522       removing their symbol and typeinfo from the exported symbol table
6523       of the output ELF binary. Using this option can reduce the exported
6524       symbol count of template-heavy code by up to 40% with no code
6525       change at all, thus notably improving link and load times for the
6526       binary as well as a reduction in size of up to 10%. Also, check the
6527       new [7]-fvisibility option.
6528     * The compiler now uses the library interface specified by the [8]C++
6529       ABI for thread-safe initialization of function-scope static
6530       variables. Most users should leave this alone, but embedded
6531       programmers may want to disable this by specifying
6532       -fno-threadsafe-statics for a small savings in code size.
6533     * Taking the address of an explicit register variable is no longer
6534       supported. Note that C++ allows taking the address of variables
6535       with register storage so this will continue to compile with a
6536       warning. For example, assuming that r0 is a machine register:
6537register int foo asm ("r0");
6538register int bar;
6539&foo; // error, no longer accepted
6540&bar; // OK, with a warning
6541     * G++ has an undocumented extension to virtual function covariancy
6542       rules that allowed the overrider to return a type that was
6543       implicitly convertable to the overridden function's return type.
6544       For instance a function returning void * could be overridden by a
6545       function returning T *. This is now deprecated and will be removed
6546       in a future release.
6547     * The G++ minimum and maximum operators (<? and >?) and their
6548       compound forms (<?=) and >?=) have been deprecated and will be
6549       removed in a future version. Code using these operators should be
6550       modified to use std::min and std::max instead.
6551     * Declaration of nested classes of class templates as friends are
6552       supported:
6553template <typename T> struct A {
6554  class B {};
6555};
6556class C {
6557  template <typename T> friend class A<T>::B;
6558};
6559       This complements the feature member functions of class templates as
6560       friends introduced in GCC 3.4.0.
6561     * When declaring a friend class using an unqualified name, classes
6562       outside the innermost non-class scope are not searched:
6563class A;
6564namespace N {
6565  class B {
6566    friend class A;   // Refer to N::A which has not been declared yet
6567                      // because name outside namespace N are not searched
6568    friend class ::A; // Refer to ::A
6569  };
6570}
6571       Hiding the friend name until declaration is still not implemented.
6572     * Friends of classes defined outside their namespace are correctly
6573       handled:
6574namespace N {
6575  class A;
6576}
6577class N::A {
6578  friend class B; // Refer to N::B in GCC 4.0.0
6579                  // but ::B in earlier versions of GCC
6580};
6581
6582    Runtime Library (libstdc++)
6583
6584     * Optimization work:
6585          + Added efficient specializations of istream functions for char
6586            and wchar_t.
6587          + Further performance tuning of strings, in particular wrt
6588            single-char append and getline.
6589          + iter_swap - and therefore most of the mutating algorithms -
6590            now makes an unqualified call to swap when the value_type of
6591            the two iterators is the same.
6592     * A large subset of the features in Technical Report 1 (TR1 for
6593       short) is experimentally delivered (i.e., no guarantees about the
6594       implementation are provided. In particular it is not promised that
6595       the library will remain link-compatible when code using TR1 is
6596       used):
6597          + General utilities such as reference_wrapper and shared_ptr.
6598          + Function objects, i.e., result_of, mem_fn, bind, function.
6599          + Support for metaprogramming.
6600          + New containers such as tuple, array, unordered_set,
6601            unordered_map, unordered_multiset, unordered_multimap.
6602     * As usual, many bugs have been fixed and LWG resolutions implemented
6603       for the first time (e.g., DR 409).
6604
6605  Java
6606
6607     * In order to prevent naming conflicts with other implementations of
6608       these tools, some GCJ binaries have been renamed:
6609          + rmic is now grmic,
6610          + rmiregistry is now grmiregistry, and
6611          + jar is now fastjar.
6612       In particular, these names were problematic for the jpackage.org
6613       packaging conventions which install symlinks in /usr/bin that point
6614       to the preferred versions of these tools.
6615     * The -findirect-dispatch argument to the compiler now works and
6616       generates code following a new "binary compatibility" ABI. Code
6617       compiled this way follows the binary compatibility rules of the
6618       Java Language Specification.
6619     * libgcj now has support for using GCJ as a JIT, using the
6620       gnu.gcj.jit family of system properties.
6621     * libgcj can now find a shared library corresponding to the bytecode
6622       representation of a class. See the documentation for the new
6623       gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system
6624       property.
6625     * There have been many improvements to the class library. Here are
6626       some highlights:
6627          + Much more of AWT and Swing exist.
6628          + Many new packages and classes were added, including
6629            java.util.regex, java.net.URI, javax.crypto,
6630            javax.crypto.interfaces, javax.crypto.spec, javax.net,
6631            javax.net.ssl, javax.security.auth,
6632            javax.security.auth.callback, javax.security.auth.login,
6633            javax.security.auth.x500, javax.security.sasl, org.ietf.jgss,
6634            javax.imageio, javax.imageio.event, javax.imageio.spi,
6635            javax.print, javax.print.attribute,
6636            javax.print.attribute.standard, javax.print.event, and
6637            javax.xml
6638          + Updated SAX and DOM, and imported GNU JAXP
6639
6640  Fortran
6641
6642     * A new [9]Fortran front end has replaced the aging GNU Fortran 77
6643       front end. The new front end supports Fortran 90 and Fortran 95. It
6644       may not yet be as stable as the old Fortran front end.
6645
6646  Ada
6647
6648     * Ada (with tasking and Zero Cost Exceptions) is now available on
6649       many more targets, including but not limited to: alpha-linux,
6650       hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux,
6651       s390x-linux, sparc-linux.
6652     * Some of the new Ada 2005 features are now implemented like
6653       Wide_Wide_Character and Ada.Containers.
6654     * Many bugs have been fixed, tools and documentation improved.
6655     * To compile Ada from the sources, install an older working Ada
6656       compiler and then use --enable-languages=ada at configuration time,
6657       since the Ada frontend is not currently activated by default. See
6658       the [10]Installing GCC for details.
6659
6660New Targets and Target Specific Improvements
6661
6662  H8/300
6663
6664     * The frame layout has changed. In the new layout, the prologue of a
6665       function first saves registers and then allocate space for locals,
6666       resulting in an 1% improvement on code size.
6667
6668  IA-32/x86-64 (AMD64)
6669
6670     * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10,
6671       log1p, log2, logb and tan mathematical builtins (and their float
6672       and long double variants) are now implemented as inline x87
6673       intrinsics when using -ffast-math.
6674     * The ceil, floor, nearbyint, rint and trunc mathematical builtins
6675       (and their float and long double variants) are now implemented as
6676       inline x87 intrinsics when using -ffast-math.
6677     * The x87's fsincos instruction is now used automatically with
6678       -ffast-math when calculating both the sin and cos of the same
6679       argument.
6680     * Instruction selection for multiplication and division by constants
6681       has been improved.
6682
6683  IA-64
6684
6685     * Floating point division, integer division and sqrt are now inlined,
6686       resulting in significant performance improvements on some codes.
6687
6688  MIPS
6689
6690     * Division by zero checks now use conditional traps if the target
6691       processor supports them. This decreases code size by one word per
6692       division operation. The old behavior (branch and break) can be
6693       obtained either at configure time by passing --with-divide=breaks
6694       to configure or at runtime by passing -mdivide-breaks to GCC.
6695     * Support for MIPS64 paired-single instructions has been added. It is
6696       enabled by -mpaired-single and can be accessed using both the
6697       target-independent vector extensions and new MIPS-specific built-in
6698       functions.
6699     * Support for the MIPS-3D ASE has been added. It is enabled by
6700       -mips3d and provides new MIPS-3D-specific built-in functions.
6701     * The -mexplicit-relocs option now supports static n64 code (as is
6702       used, for example, in 64-bit linux kernels). -mexplicit-relocs
6703       should now be feature-complete and is enabled by default when GCC
6704       is configured to use a compatible assembler.
6705     * Support for the NEC VR4130 series has been added. This support
6706       includes the use of VR-specific instructions and a new VR4130
6707       scheduler. Full VR4130 support can be selected with -march=vr4130
6708       while code for any ISA can be tuned for the VR4130 using
6709       -mtune=vr4130. There is also a new -mvr4130-align option that
6710       produces better schedules at the cost of increased code size.
6711     * Support for the Broadcom SB-1 has been extended. There is now an
6712       SB-1 scheduler as well as support for the SB-1-specific
6713       paired-single instructions. Full SB-1 support can be selected with
6714       -march=sb1 while code for any ISA can be optimized for the SB-1
6715       using -mtune=sb1.
6716     * The compiler can now work around errata in R4000, R4400, VR4120 and
6717       VR4130 processors. These workarounds are enabled by -mfix-r4000,
6718       -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120
6719       and VR4130 workarounds need binutils 2.16 or above.
6720     * IRIX shared libraries are now installed into the standard library
6721       directories: o32 libraries go into lib/, n32 libraries go into
6722       lib32/ and n64 libraries go into lib64/.
6723     * The compiler supports a new -msym32 option. It can be used to
6724       optimize n64 code in which all symbols are known to have 32-bit
6725       values.
6726
6727  S/390 and zSeries
6728
6729     * New command-line options help to generate code intended to run in
6730       an environment where stack space is restricted, e.g. Linux kernel
6731       code:
6732          + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time
6733            warnings for single functions that require large or dynamic
6734            stack frames.
6735          + -mstack-size and -mstack-guard generate code that checks for
6736            stack overflow at run time.
6737          + -mpacked-stack generates code that reduces the stack frame
6738            size of many functions by reusing unneeded parts of the stack
6739            bias area.
6740     * The -msoft-float option now ensures that generated code never
6741       accesses floating point registers.
6742     * The s390x-ibm-tpf target now fully supports C++, including
6743       exceptions and threads.
6744     * Various changes to improve performance of the generated code have
6745       been implemented, including:
6746          + GCC now uses sibling calls where possible.
6747          + Condition code handling has been optimized, allowing GCC to
6748            omit redundant comparisons in certain cases.
6749          + The cost function guiding many optimizations has been refined
6750            to more accurately represent the z900 and z990 processors.
6751          + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW
6752            instructions are now used to avoid conditional branches in
6753            certain cases.
6754          + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to
6755            optimize address arithmetic required to access large stack
6756            frames.
6757          + GCC now makes more efficient use of memory-to-memory type
6758            instructions (MVC, CLC, ...).
6759          + More precise tracking of special register use allows better
6760            instruction scheduling, in particular of the function prologue
6761            and epilogue sequences.
6762          + The Java front end now generates inline code to implement
6763            integer division, instead of calling library routines.
6764
6765  SPARC
6766
6767     * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and
6768       -mf934 have been removed. They have been replaced with -mcpu=xxx.
6769     * The internal model used to estimate the relative cost of each
6770       instruction has been updated. It is expected to give better results
6771       on recent UltraSPARC processors.
6772     * Code generation for function prologues and epilogues has been
6773       improved, resulting in better scheduling and allowing multiple exit
6774       points in functions.
6775     * Support for Sun's Visual Instruction Set (VIS) has been enhanced.
6776       It is enabled by -mvis and provides new built-in functions for VIS
6777       instructions on UltraSPARC processors.
6778     * The option -mapp-regs has been turned on by default on Solaris too.
6779
6780  NetWare
6781
6782     * Novell NetWare (on ix86, no other hardware platform was ever really
6783       supported by this OS) has been re-enabled and the ABI supported by
6784       GCC has been brought into sync with that of MetroWerks CodeWarrior
6785       (the ABI previously supported was that of some Unix systems, which
6786       NetWare never tried to support).
6787
6788Obsolete Systems
6789
6790   Support for a number of older systems has been declared obsolete in GCC
6791   4.0. Unless there is activity to revive them, the next release of GCC
6792   will have their sources permanently removed.
6793
6794   All GCC ports for the following processor architectures have been
6795   declared obsolete:
6796     * Intel i860
6797     * Ubicom IP2022
6798     * National Semiconductor NS32K
6799     * Texas Instruments TMS320C[34]x
6800
6801   Also, those for some individual systems have been obsoleted:
6802     * SPARC family
6803          + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf,
6804            sparc86x-*-elf)
6805          + OpenBSD 32-bit (sparc-*-openbsd*)
6806
6807Documentation improvements
6808
6809Other significant improvements
6810
6811     * Location lists are now generated by default when compiling with
6812       debug info and optimization. Location lists provide more accurate
6813       debug info about locations of variables and they allow debugging
6814       code compiled with -fomit-frame-pointer.
6815     * The -fvisibility option has been added which allows the default ELF
6816       visibility of all symbols to be set per compilation and the new
6817       #pragma GCC visibility preprocessor command allows the setting of
6818       default ELF visibility for a region of code. Using
6819       -fvisibility=hidden especially in combination with the new
6820       -fvisibility-inlines-hidden can yield substantial improvements in
6821       output binary quality including avoiding PLT indirection overheads,
6822       reduction of the exported symbol count by up to 60% (with resultant
6823       improvements to link and load times), better scope for the
6824       optimizer to improve code and up to a 20% reduction in binary size.
6825       Using these options correctly yields a binary with a similar symbol
6826       count to a Windows DLL.
6827       Perhaps more importantly, this new feature finally allows (with
6828       careful planning) complete avoidance of symbol clashes when
6829       manually loading shared objects with RTLD_GLOBAL, thus finally
6830       solving problems many projects such as python were forced to use
6831       RTLD_LOCAL for (with its resulting issues for C++ correctness). You
6832       can find more information about using these options at
6833       [11]https://gcc.gnu.org/wiki/Visibility.
6834     __________________________________________________________________
6835
6836GCC 4.0.1
6837
6838   This is the [12]list of problem reports (PRs) from GCC's bug tracking
6839   system that are known to be fixed in the 4.0.1 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
6843GCC 4.0.2
6844
6845   This is the [13]list of problem reports (PRs) from GCC's bug tracking
6846   system that are known to be fixed in the 4.0.2 release. This list might
6847   not be complete (that is, it is possible that some PRs that have been
6848   fixed are not listed here).
6849
6850   Unfortunately, due to a release engineering failure, this release has a
6851   regression on Solaris that will affect some C++ programs. We suggest
6852   that Solaris users apply a [14]patch that corrects the problem. Users
6853   who do not wish to apply the patch should explicitly link C++ programs
6854   with the -pthreads option, even if they do not use threads. This
6855   problem has been corrected in the current 4.0 branch sources and will
6856   not be present in GCC 4.0.3.
6857
6858GCC 4.0.3
6859
6860   Starting with this release, the function getcontext is recognized by
6861   the compiler as having the same semantics as the setjmp function. In
6862   particular, the compiler will ensure that all registers are dead before
6863   calling such a function and will emit a warning about the variables
6864   that may be clobbered after the second return from the function.
6865
6866GCC 4.0.4
6867
6868   This is the [15]list of problem reports (PRs) from GCC's bug tracking
6869   system that are known to be fixed in the 4.0.4 release. This list might
6870   not be complete (that is, it is possible that some PRs that have been
6871   fixed are not listed here).
6872
6873   The 4.0.4 release is provided for those that require a high degree of
6874   binary compatibility with previous 4.0.x releases. For most users, the
6875   GCC team recommends that version 4.1.1 or later be used instead."
6876
6877
6878    For questions related to the use of GCC, please consult these web
6879    pages and the [16]GCC manuals. If that fails, the
6880    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6881    web pages and the development of GCC are welcome on our developer
6882    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
6883    archives.
6884
6885   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
6886   distribution of this entire article is permitted in any medium,
6887   provided this notice is preserved.
6888
6889   These pages are [21]maintained by the GCC team. Last modified
6890   2014-06-28[22].
6891
6892References
6893
6894   1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
6895   2. http://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
6896   3. http://gcc.gnu.org/projects/tree-ssa/
6897   4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html
6898   5. http://gcc.gnu.org/news/sms.html
6899   6. http://www.akkadia.org/drepper/dsohowto.pdf
6900   7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility
6901   8. http://mentorembedded.github.com/cxx-abi/
6902   9. http://gcc.gnu.org/fortran/
6903  10. https://gcc.gnu.org/install/
6904  11. https://gcc.gnu.org/wiki/Visibility
6905  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1
6906  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2
6907  14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html
6908  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4
6909  16. https://gcc.gnu.org/onlinedocs/
6910  17. mailto:gcc-help@gcc.gnu.org
6911  18. mailto:gcc@gcc.gnu.org
6912  19. https://gcc.gnu.org/lists.html
6913  20. http://www.fsf.org/
6914  21. https://gcc.gnu.org/about.html
6915  22. http://validator.w3.org/check/referer
6916======================================================================
6917http://gcc.gnu.org/gcc-3.4/index.html
6918
6919                             GCC 3.4 Release Series
6920
6921   May 26, 2006
6922
6923   The [1]GNU project and the GCC developers are pleased to announce the
6924   release of GCC 3.4.6.
6925
6926   This release is a bug-fix release, containing fixes for regressions in
6927   GCC 3.4.4 relative to previous releases of GCC. This is the last of the
6928   3.4.x series.
6929
6930   The GCC 3.4 release series includes numerous [2]new features,
6931   improvements, bug fixes, and other changes, thanks to an [3]amazing
6932   group of volunteers.
6933
6934Release History
6935
6936   GCC 3.4.6
6937          March 6, 2006 ([4]changes)
6938
6939   GCC 3.4.5
6940          November 30, 2005 ([5]changes)
6941
6942   GCC 3.4.4
6943          May 18, 2005 ([6]changes)
6944
6945   GCC 3.4.3
6946          November 4, 2004 ([7]changes)
6947
6948   GCC 3.4.2
6949          September 6, 2004 ([8]changes)
6950
6951   GCC 3.4.1
6952          July 1, 2004 ([9]changes)
6953
6954   GCC 3.4.0
6955          April 18, 2004 ([10]changes)
6956
6957References and Acknowledgements
6958
6959   GCC used to stand for the GNU C Compiler, but since the compiler
6960   supports several other languages aside from C, it now stands for the
6961   GNU Compiler Collection.
6962
6963   A list of [11]successful builds is updated as new information becomes
6964   available.
6965
6966   The GCC developers would like to thank the numerous people that have
6967   contributed new features, improvements, bug fixes, and other changes as
6968   well as test results to GCC. This [12]amazing group of volunteers is
6969   what makes GCC successful.
6970
6971   For additional information about GCC please refer to the [13]GCC
6972   project web site or contact the [14]GCC development mailing list.
6973
6974   To obtain GCC please use [15]our mirror sites, or [16]our SVN server.
6975
6976
6977    For questions related to the use of GCC, please consult these web
6978    pages and the [17]GCC manuals. If that fails, the
6979    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6980    web pages and the development of GCC are welcome on our developer
6981    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
6982    archives.
6983
6984   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
6985   distribution of this entire article is permitted in any medium,
6986   provided this notice is preserved.
6987
6988   These pages are [22]maintained by the GCC team. Last modified
6989   2014-06-28[23].
6990
6991References
6992
6993   1. http://www.gnu.org/
6994   2. http://gcc.gnu.org/gcc-3.4/changes.html
6995   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6996   4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
6997   5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5
6998   6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4
6999   7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3
7000   8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2
7001   9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1
7002  10. http://gcc.gnu.org/gcc-3.4/changes.html
7003  11. http://gcc.gnu.org/gcc-3.4/buildstat.html
7004  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7005  13. http://gcc.gnu.org/index.html
7006  14. mailto:gcc@gcc.gnu.org
7007  15. http://gcc.gnu.org/mirrors.html
7008  16. http://gcc.gnu.org/svn.html
7009  17. https://gcc.gnu.org/onlinedocs/
7010  18. mailto:gcc-help@gcc.gnu.org
7011  19. mailto:gcc@gcc.gnu.org
7012  20. https://gcc.gnu.org/lists.html
7013  21. http://www.fsf.org/
7014  22. https://gcc.gnu.org/about.html
7015  23. http://validator.w3.org/check/referer
7016======================================================================
7017http://gcc.gnu.org/gcc-3.4/changes.html
7018
7019                             GCC 3.4 Release Series
7020                        Changes, New Features, and Fixes
7021
7022   The final release in the 3.4 release series is [1]GCC 3.4.6. The series
7023   is now closed.
7024
7025   GCC 3.4 has [2]many improvements in the C++ frontend. Before reporting
7026   a bug, please make sure it's really GCC, and not your code, that is
7027   broken.
7028
7029Caveats
7030
7031     * GNU Make is now required to build GCC.
7032     * With -nostdinc the preprocessor used to ignore both standard
7033       include paths and include paths contained in environment variables.
7034       It was neither documented nor intended that environment variable
7035       paths be ignored, so this has been corrected.
7036     * GCC no longer accepts the options -fvolatile, -fvolatile-global and
7037       -fvolatile-static. It is unlikely that they worked correctly in any
7038       3.x release.
7039     * GCC no longer ships <varargs.h>. Use <stdarg.h> instead.
7040     * Support for all the systems [3]obsoleted in GCC 3.3 has been
7041       removed from GCC 3.4. See below for a [4]list of systems which are
7042       obsoleted in this release.
7043     * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C
7044       compilers will not work.
7045     * The implementation of the [5]MIPS ABIs has changed. As a result,
7046       the code generated for certain MIPS targets will not be binary
7047       compatible with earlier releases.
7048     * In previous releases, the MIPS port had a fake "hilo" register with
7049       the user-visible name accum. This register has been removed.
7050     * The implementation of the [6]SPARC ABIs has changed. As a result,
7051       the code generated will not be binary compatible with earlier
7052       releases in certain cases.
7053     * The configure option --enable-threads=pthreads has been removed;
7054       use --enable-threads=posix instead, which should have the same
7055       effect.
7056     * Code size estimates used by inlining heuristics for C, Objective-C,
7057       C++ and Java have been redesigned significantly. As a result the
7058       parameters of -finline-insns, --param max-inline-insns-single and
7059       --param max-inline-insns-auto need to be reconsidered.
7060     * --param max-inline-slope and --param min-inline-insns have been
7061       removed; they are not needed for the new bottom-up inlining
7062       heuristics.
7063     * The new unit-at-a-time compilation scheme has several compatibility
7064       issues:
7065          + The order in which functions, variables, and top-level asm
7066            statements are emitted may have changed. Code relying on some
7067            particular ordering needs to be updated. The majority of such
7068            top-level asm statements can be replaced by section
7069            attributes.
7070          + Unreferenced static variables and functions are removed. This
7071            may result in undefined references when an asm statement
7072            refers to the variable/function directly. In that case either
7073            the variable/function shall be listed in asm statement operand
7074            or in the case of top-level asm statements the attribute used
7075            shall be used to force function/variable to be always output
7076            and considered as a possibly used by unknown code.
7077            For variables the attribute is accepted only by GCC 3.4 and
7078            newer, while for earlier versions it is sufficient to use
7079            unused to silence warnings about the variables not being
7080            referenced. To keep code portable across different GCC
7081            versions, you can use appropriate preprocessor conditionals.
7082          + Static functions now can use non-standard passing conventions
7083            that may break asm statements calling functions directly.
7084            Again the attribute used shall be used to prevent this
7085            behavior.
7086       As a temporary workaround, -fno-unit-at-a-time can be used, but
7087       this scheme may not be supported by future releases of GCC.
7088     * GCC 3.4 automatically places zero-initialized variables in the .bss
7089       section on some operating systems. Versions of GNU Emacs up to (and
7090       including) 21.3 will not work correctly when using this
7091       optimization; you can use -fno-zero-initialized-in-bss to disable
7092       it.
7093     * If GCC 3.4 is configured with --enable-threads=posix (the default
7094       on most targets that support pthreads) then _REENTRANT will be
7095       defined unconditionally by some libstdc++ headers. C++ code which
7096       relies on that macro to detect whether multi-threaded code is being
7097       compiled might change in meaning, possibly resulting in linker
7098       errors for single-threaded programs. Affected users of [7]Boost
7099       should compile single-threaded code with -DBOOST_DISABLE_THREADS.
7100       See Bugzilla for [8]more information.
7101
7102General Optimizer Improvements
7103
7104     * Usability of the profile feedback and coverage testing has been
7105       improved.
7106          + Performance of profiled programs has been improved by faster
7107            profile merging code.
7108          + Better use of the profile feedback for optimization (loop
7109            unrolling and loop peeling).
7110          + File locking support allowing fork() calls and parallel runs
7111            of profiled programs.
7112          + Coverage file format has been redesigned.
7113          + gcov coverage tool has been improved.
7114          + make profiledbootstrap available to build a faster compiler.
7115            Experiments made on i386 hardware showed an 11% speedup on -O0
7116            and a 7.5% speedup on -O2 compilation of a [9]large C++
7117            testcase.
7118          + New value profiling pass enabled via -fprofile-values
7119          + New value profile transformations pass enabled via -fvpt aims
7120            to optimize some code sequences by exploiting knowledge about
7121            value ranges or other properties of the operands. At the
7122            moment a conversion of expensive divisions into cheaper
7123            operations has been implemented.
7124          + New -fprofile-generate and -fprofile-use command-line options
7125            to simplify the use of profile feedback.
7126     * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and
7127       Java which is enabled via -funit-at-a-time (and implied by -O2). In
7128       this scheme a whole file is parsed first and optimized later. The
7129       following basic inter-procedural optimizations are implemented:
7130          + Removal of unreachable functions and variables
7131          + Discovery of local functions (functions with static linkage
7132            whose address is never taken)
7133          + On i386, these local functions use register parameter passing
7134            conventions.
7135          + Reordering of functions in topological order of the call graph
7136            to enable better propagation of optimizing hints (such as the
7137            stack alignments needed by functions) in the back end.
7138          + Call graph based out-of-order inlining heuristics which allows
7139            to limit overall compilation unit growth (--param
7140            inline-unit-growth).
7141       Overall, the unit-at-a-time scheme produces a 1.3% improvement for
7142       the SPECint2000 benchmark on the i386 architecture (AMD Athlon
7143       CPU).
7144     * More realistic code size estimates used by inlining for C,
7145       Objective-C, C++ and Java. The growth of large functions can now be
7146       limited via --param large-function-insns and --param
7147       large-function-growth.
7148     * A new cfg-level loop optimizer pass replaces the old loop unrolling
7149       pass and adds two other loop transformations -- loop peeling and
7150       loop unswitching -- and also uses the profile feedback to limit
7151       code growth. (The three optimizations are enabled by
7152       -funroll-loops, -fpeel-loops and -funswitch-loops flags,
7153       respectively).
7154       The old loop unroller still can be enabled by -fold-unroll-loops
7155       and may produce better code in some cases, especially when the
7156       webizer optimization pass is not run.
7157     * A new web construction pass enabled via -fweb (and implied by -O3)
7158       improves the quality of register allocation, CSE, first scheduling
7159       pass and some other optimization passes by avoiding re-use of
7160       pseudo registers with non-overlapping live ranges. The pass almost
7161       always improves code quality but does make debugging difficult and
7162       thus is not enabled by default by -O2
7163       The pass is especially effective as cleanup after code duplication
7164       passes, such as the loop unroller or the tracer.
7165     * Experimental implementations of superblock or trace scheduling in
7166       the second scheduling pass can be enabled via
7167       -fsched2-use-superblocks and -fsched2-use-traces, respectively.
7168
7169New Languages and Language specific improvements
7170
7171  Ada
7172
7173     * The Ada front end has been updated to include numerous bug fixes
7174       and enhancements. These include:
7175          + Improved project file support
7176          + Additional set of warnings about potential wrong code
7177          + Improved error messages
7178          + Improved code generation
7179          + Improved cross reference information
7180          + Improved inlining
7181          + Better run-time check elimination
7182          + Better error recovery
7183          + More efficient implementation of unbounded strings
7184          + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools,
7185            ...
7186          + New GNAT.xxxx packages (e.g. GNAT.Strings,
7187            GNAT.Exception_Action)
7188          + New pragmas
7189          + New -gnatS switch replacing gnatpsta
7190          + Implementation of new Ada features (in particular limited
7191            with, limited aggregates)
7192
7193  C/Objective-C/C++
7194
7195     * Precompiled headers are now supported. Precompiled headers can
7196       dramatically speed up compilation of some projects. There are some
7197       known defects in the current precompiled header implementation that
7198       will result in compiler crashes in relatively rare situations.
7199       Therefore, precompiled headers should be considered a "technology
7200       preview" in this release. Read the manual for details about how to
7201       use precompiled headers.
7202     * File handling in the preprocessor has been rewritten. GCC no longer
7203       gets confused by symlinks and hardlinks, and now has a correct
7204       implementation of #import and #pragma once. These two directives
7205       have therefore been un-deprecated.
7206     * The undocumented extension that allowed C programs to have a label
7207       at the end of a compound statement, which has been deprecated since
7208       GCC 3.0, has been removed.
7209     * The cast-as-lvalue extension has been removed for C++ and
7210       deprecated for C and Objective-C. In particular, code like this:
7211        int i;
7212        (char) i = 5;
7213
7214       or this:
7215        char *p;
7216        ((int *) p)++;
7217
7218       is no longer accepted for C++ and will not be accepted for C and
7219       Objective-C in a future version.
7220     * The conditional-expression-as-lvalue extension has been deprecated
7221       for C and Objective-C. In particular, code like this:
7222        int a, b, c;
7223        (a ? b : c) = 2;
7224
7225       will not be accepted for C and Objective-C in a future version.
7226     * The compound-expression-as-lvalue extension has been deprecated for
7227       C and Objective-C. In particular, code like this:
7228        int a, b;
7229        (a, b) = 2;
7230
7231       will not be accepted for C and Objective-C in a future version. A
7232       possible non-intrusive workaround is the following:
7233        (*(a, &b)) = 2;
7234
7235     * Several [10]built-in functions such as __builtin_popcount for
7236       counting bits, finding the highest and lowest bit in a word, and
7237       parity have been added.
7238     * The -fwritable-strings option has been deprecated and will be
7239       removed.
7240     * Many C math library functions are now recognized as built-ins and
7241       optimized.
7242     * The C, C++, and Objective-C compilers can now handle source files
7243       written in any character encoding supported by the host C library.
7244       The default input character set is taken from the current locale,
7245       and may be overridden with the -finput-charset command line option.
7246       In the future we will add support for inline encoding markers.
7247
7248  C++
7249
7250     * G++ is now much closer to full conformance to the ISO/ANSI C++
7251       standard. This means, among other things, that a lot of invalid
7252       constructs which used to be accepted in previous versions will now
7253       be rejected. It is very likely that existing C++ code will need to
7254       be fixed. This document lists some of the most common issues.
7255     * A hand-written recursive-descent C++ parser has replaced the
7256       YACC-derived C++ parser from previous GCC releases. The new parser
7257       contains much improved infrastructure needed for better parsing of
7258       C++ source codes, handling of extensions, and clean separation
7259       (where possible) between proper semantics analysis and parsing. The
7260       new parser fixes many bugs that were found in the old parser.
7261     * You must now use the typename and template keywords to disambiguate
7262       dependent names, as required by the C++ standard.
7263        struct K {
7264          typedef int mytype_t;
7265        };
7266
7267        template <class T1> struct A {
7268          template <class T2> struct B {
7269              void callme(void);
7270            };
7271
7272          template <int N> void bar(void)
7273          {
7274            // Use 'typename' to tell the parser that T1::mytype_t names
7275            //  a type. This is needed because the name is dependent (in
7276            //  this case, on template parameter T1).
7277            typename T1::mytype_t x;
7278            x = 0;
7279          }
7280        };
7281
7282        template <class T> void template_func(void)
7283        {
7284          // Use 'template' to prefix member templates within
7285          //  dependent types (a has type A<T>, which depends on
7286          //  the template parameter T).
7287          A<T> a;
7288          a.template bar<0>();
7289
7290          // Use 'template' to tell the parser that B is a nested
7291          //  template class (dependent on template parameter T), and
7292          //  'typename' because the whole A<T>::B<int> is
7293          //  the name of a type (again, dependent).
7294          typename A<T>::template B<int> b;
7295          b.callme();
7296        }
7297
7298        void non_template_func(void)
7299        {
7300          // Outside of any template class or function, no names can be
7301          //  dependent, so the use of the keyword 'typename' and 'template'
7302          //  is not needed (and actually forbidden).
7303          A<K> a;
7304          a.bar<0>();
7305          A<K>::B<float> b;
7306          b.callme();
7307        }
7308     * In a template definition, unqualified names will no longer find
7309       members of a dependent base (as specified by [temp.dep]/3 in the
7310       C++ standard). For example,
7311        template <typename T> struct B {
7312          int m;
7313          int n;
7314          int f ();
7315          int g ();
7316        };
7317        int n;
7318        int g ();
7319        template <typename T> struct C : B<T> {
7320          void h ()
7321          {
7322            m = 0; // error
7323            f ();  // error
7324            n = 0; // ::n is modified
7325            g ();  // ::g is called
7326          }
7327        };
7328       You must make the names dependent, e.g. by prefixing them with
7329       this->. Here is the corrected definition of C<T>::h,
7330        template <typename T> void C<T>::h ()
7331        {
7332          this->m = 0;
7333          this->f ();
7334          this->n = 0
7335          this->g ();
7336        }
7337       As an alternative solution (unfortunately not backwards compatible
7338       with GCC 3.3), you may use using declarations instead of this->:
7339        template <typename T> struct C : B<T> {
7340          using B<T>::m;
7341          using B<T>::f;
7342          using B<T>::n;
7343          using B<T>::g;
7344          void h ()
7345          {
7346            m = 0;
7347            f ();
7348            n = 0;
7349            g ();
7350          }
7351        };
7352     * In templates, all non-dependent names are now looked up and bound
7353       at definition time (while parsing the code), instead of later when
7354       the template is instantiated. For instance:
7355        void foo(int);
7356
7357        template <int> struct A {
7358          static void bar(void){
7359            foo('a');
7360          }
7361        };
7362
7363        void foo(char);
7364
7365        int main()
7366        {
7367          A<0>::bar();    // Calls foo(int), used to call foo(char).
7368        }
7369
7370     * In an explicit instantiation of a class template, you must use
7371       class or struct before the template-id:
7372        template <int N>
7373        class A {};
7374
7375        template A<0>;         // error, not accepted anymore
7376        template class A<0>;   // OK
7377     * The "named return value" and "implicit typename" extensions have
7378       been removed.
7379     * Default arguments in function types have been deprecated and will
7380       be removed.
7381     * ARM-style name-injection of friend declarations has been deprecated
7382       and will be removed. For example: struct S { friend void f(); };
7383       void g() { f(); } will not be accepted by future versions of G++;
7384       instead a declaration of "f" will need to be present outside of the
7385       scope of "S".
7386     * Covariant returns are implemented for all but varadic functions
7387       that require an adjustment.
7388     * When -pedantic is used, G++ now issues errors about spurious
7389       semicolons. For example,
7390        namespace N {}; // Invalid semicolon.
7391        void f() {}; // Invalid semicolon.
7392     * G++ no longer accepts attributes for a declarator after the
7393       initializer associated with that declarator. For example,
7394        X x(1) __attribute__((...));
7395       is no longer accepted. Instead, use:
7396        X x __attribute__((...)) (1);
7397     * Inside the scope of a template class, the name of the class itself
7398       can be treated as either a class or a template. So GCC used to
7399       accept the class name as argument of type template, and template
7400       template parameter. However this is not C++ standard compliant. Now
7401       the name is not treated as a valid template template argument
7402       unless you qualify the name by its scope. For example, the code
7403       below no longer compiles.
7404        template <template <class> class TT> class X {};
7405        template <class T> class Y {
7406          X<Y> x; // Invalid, Y is always a type template parameter.
7407        };
7408       The valid code for the above example is
7409          X< ::Y> x; // Valid.
7410       (Notice the space between < and : to prevent GCC to interpret this
7411       as a digraph for [.)
7412     * Friend declarations that refer to template specializations are
7413       rejected if the template has not already been declared. For
7414       example,
7415        template <typename T>
7416        class C {
7417          friend void f<> (C&);
7418        };
7419       is rejected. You must first declare f as a template,
7420        template <typename T>
7421        void f(T);
7422     * In case of friend declarations, every name used in the friend
7423       declaration must be accessible at the point of that declaration.
7424       Previous versions of G++ used to be less strict about this and
7425       allowed friend declarations for private class members, for example.
7426       See the ISO C++ Standard Committee's [11]defect report #209 for
7427       details.
7428     * Declaration of member functions of class templates as friends are
7429       supported. For example,
7430        template <typename T> struct A {
7431          void f();
7432        };
7433        class C {
7434          template <typename T> friend void A<T>::f();
7435        };
7436     * You must use template <> to introduce template specializations, as
7437       required by the standard. For example,
7438        template <typename T>
7439        struct S;
7440
7441        struct S<int> { };
7442       is rejected. You must write,
7443        template <> struct S<int> {};
7444     * G++ used to accept code like this,
7445        struct S {
7446          int h();
7447          void f(int i = g());
7448          int g(int i = h());
7449        };
7450       This behavior is not mandated by the standard. Now G++ issues an
7451       error about this code. To avoid the error, you must move the
7452       declaration of g before the declaration of f. The default arguments
7453       for g must be visible at the point where it is called.
7454     * The C++ ABI Section 3.3.3 specifications for the array construction
7455       routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return
7456       NULL when the allocator argument returns NULL. These changes are
7457       incorporated into the libstdc++ runtime library.
7458     * Using a name introduced by a typedef in a friend declaration or in
7459       an explicit instantiation is now rejected, as specified by the ISO
7460       C++ standard.
7461        class A;
7462        typedef A B;
7463        class C {
7464          friend class B;      // error, no typedef name here
7465          friend B;            // error, friend always needs class/struct/enum
7466          friend class A;      // OK
7467        };
7468
7469        template <int> class Q {};
7470        typedef Q<0> R;
7471        template class R;      // error, no typedef name here
7472        template class Q<0>;   // OK
7473     * When allocating an array with a new expression, GCC used to allow
7474       parentheses around the type name. This is actually ill-formed and
7475       it is now rejected:
7476        int* a = new (int)[10];    // error, not accepted anymore
7477        int* a = new int[10];      // OK
7478     * When binding an rvalue of class type to a reference, the copy
7479       constructor of the class must be accessible. For instance, consider
7480       the following code:
7481        class A
7482        {
7483        public:
7484          A();
7485
7486        private:
7487          A(const A&);   // private copy ctor
7488        };
7489
7490        A makeA(void);
7491        void foo(const A&);
7492
7493        void bar(void)
7494        {
7495          foo(A());       // error, copy ctor is not accessible
7496          foo(makeA());   // error, copy ctor is not accessible
7497
7498          A a1;
7499          foo(a1);        // OK, a1 is a lvalue
7500        }
7501       This might be surprising at first sight, especially since most
7502       popular compilers do not correctly implement this rule ([12]further
7503       details).
7504     * When forming a pointer to member or a pointer to member function,
7505       access checks for class visibility (public, protected, private) are
7506       now performed using the qualifying scope of the name itself. This
7507       is better explained with an example:
7508        class A
7509        {
7510        public:
7511          void pub_func();
7512        protected:
7513          void prot_func();
7514        private:
7515          void priv_func();
7516        };
7517
7518        class B : public A
7519        {
7520        public:
7521          void foo()
7522          {
7523            &A::pub_func;   // OK, pub_func is accessible through A
7524            &A::prot_func;  // error, cannot access prot_func through A
7525            &A::priv_func;  // error, cannot access priv_func through A
7526
7527            &B::pub_func;   // OK, pub_func is accessible through B
7528            &B::prot_func;  // OK, can access prot_func through B (within B)
7529            &B::priv_func;  // error, cannot access priv_func through B
7530          }
7531        };
7532
7533    Runtime Library (libstdc++)
7534
7535     * Optimization work:
7536          + Streamlined streambuf, filebuf, separate synched with C
7537            Standard I/O streambuf.
7538          + All formatted I/O now uses cached locale information.
7539          + STL optimizations (memory/speed for list, red-black trees as
7540            used by sets and maps).
7541          + More use of GCC builtins.
7542          + String optimizations (avoid contention on
7543            increment/decrement-and-test of the reference count in the
7544            empty-string object, constructor from input_iterators
7545            speedup).
7546     * Static linkage size reductions.
7547     * Large File Support (files larger than 2 GB on 32-bit systems).
7548     * Wide character and variable encoding filebuf work (UTF-8, Unicode).
7549     * Generic character traits.
7550     * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x,
7551       Solaris 2.7 and above, AIX 5.x, Irix 6.5.
7552     * The allocator class is now standard-conformant, and two additional
7553       extension allocators have been added, mt_alloc and
7554       bitmap_allocator.
7555     * PCH support: -include bits/stdc++.h (2x compile speedup).
7556     * Rewrote __cxa_demangle with support for C++ style allocators.
7557     * New debug modes for STL containers and iterators.
7558     * Testsuite rewrite: five times as many tests, plus increasingly
7559       sophisticated tests, including I/O, MT, multi-locale, wide and
7560       narrow characters.
7561     * Use current versions of GNU "autotools" for build/configuration.
7562
7563  Objective-C
7564
7565     * The Objective-C front end has been updated to include the numerous
7566       bug fixes and enhancements previously available only in Apple's
7567       version of GCC. These include:
7568          + Structured exception (@try... @catch... @finally, @throw) and
7569            synchronization (@synchronized) support. These are accessible
7570            via the -fobjc-exceptions switch; as of this writing, they may
7571            only be used in conjunction with -fnext-runtime on Mac OS X
7572            10.3 and later. See [13]Options Controlling Objective-C
7573            Dialect for more information.
7574          + An overhaul of @encode logic. The C99 _Bool and C++ bool type
7575            may now be encoded as 'B'. In addition, the back-end/codegen
7576            dependencies have been removed.
7577          + An overhaul of message dispatch construction, ensuring that
7578            the various receiver types (and casts thereof) are handled
7579            properly, and that correct diagnostics are issued.
7580          + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue"
7581            (-freplace-objc-classes) debugging modes, currently available
7582            on Mac OS X 10.3 and later. See [14]Options Controlling
7583            Objective-C Dialect for more information.
7584          + Access to optimized runtime entry points (-fno-nil-receivers )
7585            on the assumption that message receivers are never nil. This
7586            is currently available on Mac OS X 10.3 and later. See
7587            [15]Options Controlling Objective-C Dialect for more
7588            information.
7589
7590  Java
7591
7592     * Compiling a .jar file will now cause non-.class entries to be
7593       automatically compiled as resources.
7594     * libgcj has been ported to Darwin.
7595     * Jeff Sturm has adapted Jan Hubicka's call graph optimization code
7596       to gcj.
7597     * libgcj has a new gcjlib URL type; this lets URLClassLoader load
7598       code from shared libraries.
7599     * libgcj has been much more completely merged with [16]GNU Classpath.
7600     * Class loading is now much more correct; in particular the caller's
7601       class loader is now used when that is required.
7602     * [17]Eclipse 2.x will run out of the box using gij.
7603     * Parts of java.nio have been implemented. Direct and indirect
7604       buffers work, as do fundamental file and socket operations.
7605     * java.awt has been improved, though it is still not ready for
7606       general use.
7607     * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST
7608       method.
7609     * The MinGW port has matured. Enhancements include socket timeout
7610       support, thread interruption, improved Runtime.exec() handling and
7611       support for accented characters in filenames.
7612
7613  Fortran
7614
7615     * Fortran improvements are listed in the [18]Fortran documentation.
7616
7617New Targets and Target Specific Improvements
7618
7619  Alpha
7620
7621     * Several [19]built-in functions have been added such as
7622       __builtin_alpha_zap to allow utilizing the more obscure
7623       instructions of the CPU.
7624     * Parameter passing of complex arguments has changed to match the
7625       [20]ABI. This change is incompatible with previous GCC versions,
7626       but does fix compatibility with the Tru64 compiler and several
7627       corner cases where GCC was incompatible with itself.
7628
7629  ARM
7630
7631     * Nicolas Pitre has contributed his hand-coded floating-point support
7632       code for ARM. It is both significantly smaller and faster than the
7633       existing C-based implementation, even when building applications
7634       for Thumb. The arm-elf configuration has been converted to use the
7635       new code.
7636     * Support for the Intel's iWMMXt architecture, a second generation
7637       XScale processor, has been added. Enabled at run time with the
7638       -mcpu=iwmmxt command line switch.
7639     * A new ARM target has been added: arm-wince-pe. This is similar to
7640       the arm-pe target, but it defaults to using the APCS32 ABI.
7641     * The existing ARM pipeline description has been converted to the use
7642       the [21]DFA processor pipeline model. There is not much change in
7643       code performance, but the description is now [22]easier to
7644       understand.
7645     * Support for the Cirrus EP9312 Maverick floating point co-processor
7646       added. Enabled at run time with the -mcpu=ep9312 command line
7647       switch. Note however that the multilibs to support this chip are
7648       currently disabled in gcc/config/arm/t-arm-elf, so if you want to
7649       enable their production you will have to uncomment the entries in
7650       that file.
7651
7652  H8/300
7653
7654     * Support for long long has been added.
7655     * Support for saveall attribute has been added.
7656     * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code
7657       for H8/300H and H8S, which is much faster than the previous
7658       implementation.
7659     * A lot of small performance improvements.
7660
7661  IA-32/AMD64 (x86-64)
7662
7663     * Tuning for K8 (AMD Opteron/Athlon64) core is available via
7664       -march=k8 and -mcpu=k8.
7665     * Scalar SSE code generation carefully avoids reformatting penalties,
7666       hidden dependencies and minimizes the number of uops generated on
7667       both Intel and AMD CPUs.
7668     * Vector MMX and SSE operands are now passed in registers to improve
7669       performance and match the argument passing convention used by the
7670       Intel C++ Compiler. As a result it is not possible to call
7671       functions accepting vector arguments compiled by older GCC version.
7672     * Conditional jump elimination is now more aggressive on modern CPUs.
7673     * The Athlon ports has been converted to use the DFA processor
7674       pipeline description.
7675     * Optimization of indirect tail calls is now possible in a similar
7676       fashion as direct sibcall optimization.
7677     * Further small performance improvements.
7678     * -m128bit-long-double is now less buggy.
7679     * __float128 support in 64-bit compilation.
7680     * Support for data structures exceeding 2GB in 64-bit mode.
7681     * -mcpu has been renamed to -mtune.
7682
7683  IA-64
7684
7685     * Tuning code for the Itanium 2 processor has been added. The
7686       generation of code tuned for Itanium 2 (option -mtune=itanium2) is
7687       enabled by default now. To generate code tuned for Itanium 1 the
7688       option -mtune=itanium1 should be used.
7689     * [23]DFA processor pipeline descriptions for the IA-64 processors
7690       have been added. This resulted in about 3% improvement on the
7691       SPECInt2000 benchmark for Itanium 2.
7692     * Instruction bundling for the IA-64 processors has been rewritten
7693       using the DFA pipeline hazard recognizer. It resulted in about 60%
7694       compiler speedup on the SPECInt2000 C programs.
7695
7696  M32R
7697
7698     * Support for the M32R/2 processor has been added by Renesas.
7699     * Support for an M32R GNU/Linux target and PIC code generation has
7700       been added by Renesas.
7701
7702  M68000
7703
7704     * Bernardo Innocenti (Develer S.r.l.) has contributed the
7705       m68k-uclinux target, based on former work done by Paul Dale
7706       (SnapGear Inc.). Code generation for the ColdFire processors family
7707       has been enhanced and extended to support the MCF 53xx and MCF 54xx
7708       cores, integrating former work done by Peter Barada (Motorola).
7709
7710  MIPS
7711
7712    Processor-specific changes
7713
7714     * Support for the RM7000 and RM9000 processors has been added. It can
7715       be selected using the -march compiler option and should work with
7716       any MIPS I (mips-*) or MIPS III (mips64-*) configuration.
7717     * Support for revision 2 of the MIPS32 ISA has been added. It can be
7718       selected with the command-line option -march=mips32r2.
7719     * There is a new option, -mfix-sb1, to work around certain SB-1
7720       errata.
7721
7722    Configuration
7723
7724     * It is possible to customize GCC using the following configure-time
7725       options:
7726          + --with-arch, which specifies the default value of the -march
7727            option.
7728          + --with-tune, which specifies the default value of the -mtune
7729            option.
7730          + --with-abi, which specifies the default ABI.
7731          + --with-float=soft, which tells GCC to use software floating
7732            point by default.
7733          + --with-float=hard, which tells GCC to use hardware floating
7734            point by default.
7735     * A 64-bit GNU/Linux port has been added. The associated
7736       configurations are mips64-linux-gnu and mips64el-linux-gnu.
7737     * The 32-bit GNU/Linux port now supports Java.
7738     * The IRIX 6 configuration now supports the o32 ABI and will build
7739       o32 multilibs by default. This support is compatible with both
7740       binutils and the SGI tools, but note that several features,
7741       including debugging information and DWARF2 exception handling, are
7742       only available when using the GNU assembler. Use of the GNU
7743       assembler and linker (version 2.15 or above) is strongly
7744       recommended.
7745     * The IRIX 6 configuration now supports 128-bit long doubles.
7746     * There are two new RTEMS-specific configurations, mips-rtems and
7747       mipsel-rtems.
7748     * There are two new *-elf configurations, mipsisa32r2-elf and
7749       mipsisa32r2el-elf.
7750
7751    General
7752
7753     * Several [24]ABI bugs have been fixed. Unfortunately, these changes
7754       will break binary compatibility with earlier releases.
7755     * GCC can now use explicit relocation operators when generating
7756       -mabicalls code. This behavior is controlled by -mexplicit-relocs
7757       and can have several performance benefits. For example:
7758          + It allows for more optimization of GOT accesses, including
7759            better scheduling and redundancy elimination.
7760          + It allows sibling calls to be implemented as jumps.
7761          + n32 and n64 leaf functions can use a call-clobbered global
7762            pointer instead of $28.
7763          + The code to set up $gp can be removed from functions that
7764            don't need it.
7765     * A new option, -mxgot, allows the GOT to be bigger than 64k. This
7766       option is equivalent to the assembler's -xgot option and should be
7767       used instead of -Wa,-xgot.
7768     * Frame pointer elimination is now supported when generating 64-bit
7769       MIPS16 code.
7770     * Inline block moves have been optimized to take more account of
7771       alignment information.
7772     * Many internal changes have been made to the MIPS port, mostly aimed
7773       at reducing the reliance on assembler macros.
7774
7775  PowerPC
7776
7777     * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64
7778       [25]ABI incompatibilities regarding the way parameters are passed
7779       during functions calls. These changes may result in incompatibility
7780       between code compiled with GCC 3.3 and GCC 3.4.
7781
7782    PowerPC Darwin
7783
7784     * Support for shared/dylib gcc libraries has been added. It is
7785       enabled by default on powerpc-apple-darwin7.0.0 and up.
7786     * Libgcj is enabled by default. On systems older than
7787       powerpc-apple-darwin7.0.0 you need to install dlcompat.
7788     * 128-bit IBM extended precision format support added for long
7789       double.
7790
7791    PowerPC64 GNU/Linux
7792
7793     * By default, PowerPC64 GNU/Linux now uses natural alignment of
7794       structure elements. The old four byte alignment for double, with
7795       special rules for a struct starting with a double, can be chosen
7796       with -malign-power. This change may result in incompatibility
7797       between code compiled with GCC 3.3 and GCC 3.4.
7798     * -mabi=altivec is now the default rather than -mabi=no-altivec.
7799     * 128-bit IBM extended precision format support added for long
7800       double.
7801
7802  S/390 and zSeries
7803
7804     * New command-line options allow to specify the intended execution
7805       environment for generated code:
7806          + -mesa/-mzarch allows to specify whether to generate code
7807            running in ESA/390 mode or in z/Architecture mode (this is
7808            applicable to 31-bit code only).
7809          + -march allows to specify a minimum processor architecture
7810            level (g5, g6, z900, or z990).
7811          + -mtune allows to specify which processor to tune for.
7812     * It is possible to customize GCC using the following configure-time
7813       options:
7814          + --with-mode, which specifies whether to default to assuming
7815            ESA/390 or z/Architecture mode.
7816          + --with-arch, which specifies the default value of the -march
7817            option.
7818          + --with-tune, which specifies the default value of the -mtune
7819            option.
7820     * Support for the z990 processor has been added, and can be selected
7821       using -march=z990 or -mtune=z990. This includes instruction
7822       scheduling tuned for the superscalar instruction pipeline of the
7823       z990 processor as well as support for all new instructions provided
7824       by the long-displacement facility.
7825     * Support to generate 31-bit code optimized for zSeries processors
7826       (running in ESA/390 or in z/Architecture mode) has been added. This
7827       can be selected using -march=z900 and -mzarch respectively.
7828     * Instruction scheduling for the z900 and z990 processors now uses
7829       the DFA pipeline hazard recognizer.
7830     * GCC no longer generates code to maintain a stack backchain,
7831       previously used to generate stack backtraces for debugging
7832       purposes. As replacement that does not incur runtime overhead,
7833       DWARF-2 call frame information is provided by GCC; this is
7834       supported by GDB 6.1. The old behavior can be restored using the
7835       -mbackchain option.
7836     * The stack frame size of functions may now exceed 2 GB in 64-bit
7837       code.
7838     * A port for the 64-bit IBM TPF operating system has been added; the
7839       configuration is s390x-ibm-tpf. This configuration is supported as
7840       cross-compilation target only.
7841     * Various changes to improve the generated code have been
7842       implemented, including:
7843          + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT
7844            instructions to significantly speed up many floating-point
7845            applications.
7846          + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL
7847            WITH BORROW instructions to speed up long long arithmetic.
7848          + GCC now uses the SEARCH STRING instruction to implement
7849            strlen().
7850          + In many cases, function call overhead for 31-bit code has been
7851            reduced by placing the literal pool after the function code
7852            instead of after the function prolog.
7853          + Register 14 is no longer reserved in 64-bit code.
7854          + Handling of global register variables has been improved.
7855
7856  SPARC
7857
7858     * The option -mflat is deprecated.
7859     * Support for large (> 2GB) frames has been added to the 64-bit port.
7860     * Several [26]ABI bugs have been fixed. Unfortunately, these changes
7861       will break binary compatibility with earlier releases.
7862     * The default debugging format has been switched from STABS to
7863       DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already
7864       the default debugging format for 64-bit code on Solaris.
7865
7866  SuperH
7867
7868     * Support for the SH2E processor has been added. Enabled at run time
7869       with the -m2e command line switch, or at configure time by
7870       specifying sh2e as the machine part of the target triple.
7871
7872  V850
7873
7874     * Support for the Mitsubishi V850E1 processor has been added. This is
7875       a variant of the V850E processor with some additional debugging
7876       instructions.
7877
7878  Xtensa
7879
7880     * Several ABI bugs have been fixed. Unfortunately, these changes
7881       break binary compatibility with earlier releases.
7882          + For big-endian processors, the padding of aggregate return
7883            values larger than a word has changed. If the size of an
7884            aggregate return value is not a multiple of 32 bits, previous
7885            versions of GCC inserted padding in the most-significant bytes
7886            of the first return value register. Aggregates larger than a
7887            word are now padded in the least-significant bytes of the last
7888            return value register used. Aggregates smaller than a word are
7889            still padded in the most-significant bytes. The return value
7890            padding has not changed for little-endian processors.
7891          + Function arguments with 16-byte alignment are now properly
7892            aligned.
7893          + The implementation of the va_list type has changed. A va_list
7894            value created by va_start from a previous release cannot be
7895            used with va_arg from this release, or vice versa.
7896     * More processor configuration options for Xtensa processors are
7897       supported:
7898          + the ABS instruction is now optional;
7899          + the ADDX* and SUBX* instructions are now optional;
7900          + an experimental CONST16 instruction can be used to synthesize
7901            constants instead of loading them from constant pools.
7902       These and other Xtensa processor configuration options can no
7903       longer be enabled or disabled by command-line options; the
7904       processor configuration must be specified by the xtensa-config.h
7905       header file when building GCC. Additionally, the
7906       -mno-serialize-volatile option is no longer supported.
7907
7908Obsolete Systems
7909
7910   Support for a number of older systems has been declared obsolete in GCC
7911   3.4. Unless there is activity to revive them, the next release of GCC
7912   will have their sources permanently removed.
7913
7914   All configurations of the following processor architectures have been
7915   declared obsolete:
7916     * Mitsubishi D30V, d30v-*
7917     * AT&T DSP1600 and DSP1610, dsp16xx-*
7918     * Intel 80960, i960
7919
7920   Also, some individual systems have been obsoleted:
7921     * ARM Family
7922          + Support for generating code for operation in APCS/26 mode
7923            (-mapcs-26).
7924     * IBM ESA/390
7925          + "Bigfoot" port, i370-*. (The other port, s390-*, is actively
7926            maintained and supported.)
7927     * Intel 386 family
7928          + MOSS, i?86-moss-msdos and i?86-*-moss*
7929          + NCR 3000 running System V r.4, i?86-ncr-sysv4*
7930          + FreeBSD with a.out object format, i?86-*-freebsd*aout* and
7931            i?86-*-freebsd2*
7932          + GNU/Linux with a.out object format, i?86-linux*aout*
7933          + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1*
7934          + Interix versions before Interix 3, i?86-*-interix
7935          + Mach microkernel, i?86-mach*
7936          + SCO UnixWare with UDK, i?86-*-udk*
7937          + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]*
7938          + VSTa microkernel, i386-*-vsta
7939     * Motorola M68000 family
7940          + HPUX, m68k-hp-hpux* and m68000-hp-hpux*
7941          + NetBSD with a.out object format (before NetBSD 1.4),
7942            m68k-*-*-netbsd* except m68k-*-*-netbsdelf*
7943          + Generic System V r.4, m68k-*-sysv4*
7944     * VAX
7945          + Generic VAX, vax-*-* (This is generic VAX only; we have not
7946            obsoleted any VAX triples for specific operating systems.)
7947
7948Documentation improvements
7949
7950Other significant improvements
7951
7952     * The build system has undergone several significant cleanups.
7953       Subdirectories will only be configured if they are being built, and
7954       all subdirectory configures are run from the make command. The top
7955       level has been autoconfiscated.
7956     * Building GCC no longer writes to its source directory. This should
7957       help those wishing to share a read-only source directory over NFS
7958       or build from a CD. The exceptions to this feature are if you
7959       configure with either --enable-maintainer-mode or
7960       --enable-generated-files-in-srcdir.
7961     * The -W warning option has been renamed to -Wextra, which is more
7962       easily understood. The older spelling will be retained for
7963       backwards compatibility.
7964     * Substantial improvements in compile time have been made,
7965       particularly for non-optimizing compilations.
7966     __________________________________________________________________
7967
7968GCC 3.4.0
7969
7970  Bug Fixes
7971
7972   A vast number of bugs have been fixed in 3.4.0, too many to publish a
7973   complete list here. [27]Follow this link to query the Bugzilla database
7974   for the list of over 900 bugs fixed in 3.4.0. This is the list of all
7975   bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4
7976   regressions.
7977     __________________________________________________________________
7978
7979GCC 3.4.1
7980
7981  Bug Fixes
7982
7983   This section lists the problem reports (PRs) from GCC's bug tracking
7984   system that are known to be fixed in the 3.4.1 release. This list might
7985   not be complete (that is, it is possible that some PRs that have been
7986   fixed are not listed here).
7987
7988    Bootstrap failures
7989
7990     * [28]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler
7991       emitted - PIC related
7992     * [29]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf
7993     * [30]14760 A bug in configure.in prevents using both
7994       --program-suffix and --program-prefix
7995     * [31]14671 [hppa64] bootstrap fails: ICE in
7996       save_call_clobbered_regs, in caller_save.c
7997     * [32]15093 [alpha][Java] make bootstrap fails to configure libffi on
7998       Alpha
7999     * [33]15178 Solaris 9/x86 fails linking after stage 3
8000
8001    Multi-platform internal compiler errors (ICEs)
8002
8003     * [34]12753 (preprocessor) Memory corruption in preprocessor on bad
8004       input
8005     * [35]13985 ICE in gcc.c-torture/compile/930621-1.c
8006     * [36]14810 (c++) tree check failures with invalid code involving
8007       templates
8008     * [37]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in
8009       cp/parser.c
8010     * [38]15044 (c++) ICE on syntax error, template header
8011     * [39]15057 (c++) Compiling of conditional value throw constructs
8012       cause a segmentation violation
8013     * [40]15064 (c++) typeid of template parameter gives ICE
8014     * [41]15142 (c++) ICE when passing a string where a char* is expected
8015       in a throw statement
8016     * [42]15159 ICE in rtl_verify_flow_info_1
8017     * [43]15165 (c++) ICE in instantiate_template
8018     * [44]15193 Unary minus using pointer to V4SF vector causes
8019       -fforce-mem to exhaust all memory
8020     * [45]15209 (c++) Runs out of memory with packed structs
8021     * [46]15227 (c++) Trouble with invalid function definition
8022     * [47]15285 (c++) instantiate_type ICE when forming pointer to
8023       template function
8024     * [48]15299 (c++) ICE in resolve_overloaded_unification
8025     * [49]15329 (c++) ICE on constructor of member template
8026     * [50]15550 ICE in extract_insn, in recog.c
8027     * [51]15554 (c++) ICE in tsubst_copy, in cp/pt.c
8028     * [52]15640 (c++) ICE on invalid code in arg_assoc, in
8029       cp/name-lookup.c
8030     * [53]15666 [unit-at-a-time] Gcc abort on valid code
8031     * [54]15696 (c++) ICE with bad pointer-to-member code
8032     * [55]15701 (c++) ICE with friends and template template parameter
8033     * [56]15761 ICE in do_SUBST, in combine.c
8034     * [57]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops
8035
8036    Ada
8037
8038     * [58]14538 All RTEMS targets broken for gnat
8039
8040    C front end
8041
8042     * [59]12391 missing warning about assigning to an incomplete type
8043     * [60]14649 atan(1.0) should not be a constant expression
8044     * [61]15004 [unit-at-a-time] no warning for unused paramater in
8045       static function
8046     * [62]15749 --pedantic-errors behaves differently from --pedantic
8047       with C-compiler on GNU/Linux
8048
8049    C++ compiler and library
8050
8051     * [63]10646 non-const reference is incorrectly matched in a "const T"
8052       partial specialization
8053     * [64]12077 wcin.rdbuf()->in_avail() return value too high
8054     * [65]13598 enc_filebuf doesn't work
8055     * [66]14211 const_cast returns lvalue but should be rvalue
8056     * [67]14220 num_put::do_put() undesired float/double behavior
8057     * [68]14245 problem with user-defined allocators in std::basic_string
8058     * [69]14340 libstdc++ Debug mode: failure to convert iterator to
8059       const_iterator
8060     * [70]14600 __gnu_cxx::stdio_sync_filebuf should expose internal
8061       FILE*
8062     * [71]14668 no warning anymore for reevaluation of declaration
8063     * [72]14775 LFS (large file support) tests missing
8064     * [73]14821 Duplicate namespace alias declaration should not conflict
8065     * [74]14930 Friend declaration ignored
8066     * [75]14932 cannot use offsetof to get offsets of array elements in
8067       g++ 3.4.0
8068     * [76]14950 [non unit-at-a-time] always_inline does not mix with
8069       templates and -O0
8070     * [77]14962 g++ ignores #pragma redefine_extname
8071     * [78]14975 Segfault on low-level write error during imbue
8072     * [79]15002 Linewise stream input is unusably slow (std::string slow)
8073     * [80]15025 compiler accepts redeclaration of template as
8074       non-template
8075     * [81]15046 [arm] Math functions misdetected by cross configuration
8076     * [82]15069 a bit test on a variable of enum type is miscompiled
8077     * [83]15074 g++ -lsupc++ still links against libstdc++
8078     * [84]15083 spurious "statement has no effect" warning
8079     * [85]15096 parse error with templates and pointer to const member
8080     * [86]15287 combination of operator[] and operator .* fails in
8081       templates
8082     * [87]15317 __attribute__ unused in first parameter of constructor
8083       gives error
8084     * [88]15337 sizeof on incomplete type diagnostic
8085     * [89]15361 bitset<>::_Find_next fails
8086     * [90]15412 _GLIBCXX_ symbols symbols defined and used in different
8087       namespaces
8088     * [91]15427 valid code results in incomplete type error
8089     * [92]15471 Incorrect member pointer offsets in anonymous
8090       structs/unions
8091     * [93]15503 nested template problem
8092     * [94]15507 compiler hangs while laying out union
8093     * [95]15542 operator & and template definitions
8094     * [96]15565 SLES9: leading + sign for unsigned int with showpos
8095     * [97]15625 friend defined inside a template fails to find static
8096       function
8097     * [98]15629 Function templates, overloads, and friend name injection
8098     * [99]15742 'noreturn' attribute ignored in method of template
8099       functions.
8100     * [100]15775 Allocator::pointer consistently ignored
8101     * [101]15821 Duplicate namespace alias within namespace rejected
8102     * [102]15862 'enum yn' fails (confict with undeclared builtin)
8103     * [103]15875 rejects pointer to member in template
8104     * [104]15877 valid code using templates and anonymous enums is
8105       rejected
8106     * [105]15947 Puzzling error message for wrong destructor declaration
8107       in template class
8108     * [106]16020 cannot copy __gnu_debug::bitset
8109     * [107]16154 input iterator concept too restrictive
8110     * [108]16174 deducing top-level consts
8111
8112    Java
8113
8114     * [109]14315 Java compiler is not parallel make safe
8115
8116    Fortran
8117
8118     * [110]15151 [g77] incorrect logical i/o in 64-bit mode
8119
8120    Objective-C
8121
8122     * [111]7993 private variables cannot be shadowed in subclasses
8123
8124    Optimization bugs
8125
8126     * [112]15228 useless copies of floating point operands
8127     * [113]15345 [non-unit-at-a-time] unreferenced nested inline
8128       functions not optimized away
8129     * [114]15945 Incorrect floating point optimization
8130     * [115]15526 ftrapv aborts on 0 * (-1)
8131     * [116]14690 Miscompiled POOMA tests
8132     * [117]15112 GCC generates code to write to unchanging memory
8133
8134    Preprocessor
8135
8136     * [118]15067 Minor glitch in the source of cpp
8137
8138    Main driver program bugs
8139
8140     * [119]1963 collect2 interprets -oldstyle_liblookup as -o
8141       ldstyle_liblookup
8142
8143    x86-specific (Intel/AMD)
8144
8145     * [120]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND*
8146       section}
8147
8148    HPPA-specific
8149
8150     * [121]14782 GCC produces an unaligned data access at -O2
8151     * [122]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2
8152     * [123]15202 ICE in reload_cse_simplify_operands, in postreload.c
8153
8154    IA64-specific
8155
8156     * [124]14610 __float80 constants incorrectly emitted
8157     * [125]14813 init_array sections are initialized in the wrong order
8158     * [126]14857 GCC segfault on duplicated asm statement
8159     * [127]15598 Gcc 3.4 ICE on valid code
8160     * [128]15653 Gcc 3.4 ICE on valid code
8161
8162    MIPS-specific
8163
8164     * [129]15189 wrong filling of delay slot with -march=mips1 -G0
8165       -mno-split-addresses -mno-explicit-relocs
8166     * [130]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as
8167       2.14.91
8168     * [131]16144 Bogus reference to __divdf3 when -O1
8169     * [132]16176 Miscompilation of unaligned data in MIPS backend
8170
8171    PowerPC-specific
8172
8173     * [133]11591 ICE in gcc.dg/altivec-5.c
8174     * [134]12028 powerpc-eabispe produces bad sCOND operation
8175     * [135]14478 rs6000 geu/ltu patterns generate incorrect code
8176     * [136]14567 long double and va_arg complex args
8177     * [137]14715 Altivec stack layout may overlap gpr save with stack
8178       temps
8179     * [138]14902 (libstdc++) Stream checking functions fail when -pthread
8180       option is used.
8181     * [139]14924 Compiler ICE on valid code
8182     * [140]14960 -maltivec affects vector return with -mabi=no-altivec
8183     * [141]15106 vector varargs failure passing from altivec to
8184       non-altivec code for -m32
8185     * [142]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 &
8186       half-word operation
8187     * [143]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx
8188       and stvx
8189     * [144]15662 Segmentation fault when an exception is thrown - even if
8190       try and catch are specified
8191
8192    s390-specific
8193
8194     * [145]15054 Bad code due to overlapping stack temporaries
8195
8196    SPARC-specific
8197
8198     * [146]15783 ICE with union assignment in 64-bit mode
8199     * [147]15626 GCC 3.4 emits "ld: warning: relocation error:
8200       R_SPARC_UA32"
8201
8202    x86-64-specific
8203
8204     * [148]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64
8205     * [149]14723 Backported -march=nocona from mainline
8206     * [150]15290 __float128 failed to pass to function properly
8207
8208    Cygwin/Mingw32-specific
8209
8210     * [151]15250 Option -mms-bitfields support on GCC 3.4 is not
8211       conformant to MS layout
8212     * [152]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe
8213       worker on windows32 targets
8214
8215    Bugs specific to embedded processors
8216
8217     * [153]8309 [m68k] -m5200 produces erroneous SImode set of short
8218       varaible on stack
8219     * [154]13250 [SH] Gcc code for rotation clobbers the register, but
8220       gcc continues to use the register as if it was not clobbered
8221     * [155]13803 [coldfire] movqi operand constraints too restrictivefor
8222       TARGET_COLDFIRE
8223     * [156]14093 [SH] ICE for code when using -mhitachi option in SH
8224     * [157]14457 [m6811hc] ICE with simple c++ source
8225     * [158]14542 [m6811hc] ICE on simple source
8226     * [159]15100 [SH] cc1plus got hang-up on
8227       libstdc++-v3/testsuite/abi_check.cc
8228     * [160]15296 [CRIS] Delayed branch scheduling causing invalid code on
8229       cris-*
8230     * [161]15396 [SH] ICE with -O2 -fPIC
8231     * [162]15782 [coldfire] m68k_output_mi_thunk emits wrong code for
8232       ColdFire
8233
8234    Testsuite problems (compiler not affected)
8235
8236     * [163]11610 libstdc++ testcases 27_io/* don't work properly remotely
8237     * [164]15488 (libstdc++) possibly insufficient file permissions for
8238       executing test suite
8239     * [165]15489 (libstdc++) testsuite_files determined incorrectly
8240
8241    Documentation bugs
8242
8243     * [166]13928 (libstdc++) no whatis info in some man pages generated
8244       by doxygen
8245     * [167]14150 Ada documentation out of date
8246     * [168]14949 (c++) Need to document method visibility changes
8247     * [169]15123 libstdc++-doc: Allocators.3 manpage is empty
8248     __________________________________________________________________
8249
8250GCC 3.4.2
8251
8252  Bug Fixes
8253
8254   This section lists the problem reports (PRs) from GCC's bug tracking
8255   system that are known to be fixed in the 3.4.2 release. This list might
8256   not be complete (that is, it is possible that some PRs that have been
8257   fixed are not listed here).
8258
8259    Bootstrap failures and issues
8260
8261     * [170]16469 [mips-sgi-irix5.3] bootstrap fails in
8262       libstdc++-v3/testsuite
8263     * [171]16344 [hppa-linux-gnu] libstdc++'s PCH built by
8264       profiledbootstrap does not work with the built compiler
8265     * [172]16842 [Solaris/x86] mkheaders can not find mkheaders.conf
8266
8267    Multi-platform internal compiler errors (ICEs)
8268
8269     * [173]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in
8270       cp_parser_class_specifier, in cp/parser.c
8271     * [174]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
8272     * [175]15461 (c++) ICE due to NRV and inlining
8273     * [176]15890 (c++) ICE in c_expand_expr, in c-common.c
8274     * [177]16180 ICE: segmentation fault in RTL optimization
8275     * [178]16224 (c++) ICE in write_unscoped_name (template/namespace)
8276     * [179]16408 ICE: in delete_insn, in cfgrtl.c
8277     * [180]16529 (c++) ICE for: namespace-alias shall not be declared as
8278       the name of any other entity
8279     * [181]16698 (c++) ICE with exceptions and declaration of __cxa_throw
8280     * [182]16706 (c++) ICE in finish_member_declaration, in
8281       cp/semantics.c
8282     * [183]16810 (c++) Legal C++ program with cast gives ICE in
8283       build_ptrmemfunc
8284     * [184]16851 (c++) ICE when throwing a comma expression
8285     * [185]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c
8286     * [186]16904 (c++) ICE in finish_class_member_access_expr, in
8287       cp/typeck.c
8288     * [187]16905 (c++) ICE (segfault) with exceptions
8289     * [188]16964 (c++) ICE in cp_parser_class_specifier due to
8290       redefinition
8291     * [189]17068 (c++) ICE: tree check: expected class 'd', have 'x'
8292       (identifier_node) in dependent_template_p, in cp/pt.c
8293
8294    Preprocessor bugs
8295
8296     * [190]16366 Preprocessor option -remap causes memory corruption
8297
8298    Optimization
8299
8300     * [191]15345 unreferenced nested inline functions not optimized away
8301     * [192]16590 Incorrect execution when compiling with -O2
8302     * [193]16693 Bitwise AND is lost when used within a cast to an enum
8303       of the same precision
8304     * [194]17078 Jump into if(0) substatement fails
8305
8306    Problems in generated debug information
8307
8308     * [195]13956 incorrect stabs for nested local variables
8309
8310    C front end bugs
8311
8312     * [196]16684 GCC should not warn about redundant redeclarations of
8313       built-ins
8314
8315    C++ compiler and library
8316
8317     * [197]12658 Thread safety problems in locale::global() and
8318       locale::locale()
8319     * [198]13092 g++ accepts invalid pointer-to-member conversion
8320     * [199]15320 Excessive memory consumption
8321     * [200]16246 Incorrect template argument deduction
8322     * [201]16273 Memory exhausted when using nested classes and virtual
8323       functions
8324     * [202]16401 ostringstream in gcc 3.4.x very slow for big data
8325     * [203]16411 undefined reference to
8326       __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char>
8327       >::file()
8328     * [204]16489 G++ incorrectly rejects use of a null constant integral
8329       expression as a null constant pointer
8330     * [205]16618 offsetof fails with constant member
8331     * [206]16637 syntax error reported for valid input code
8332     * [207]16717 __attribute__((constructor)) broken in C++
8333     * [208]16813 compiler error in DEBUG version of range insertion
8334       std::map::insert
8335     * [209]16853 pointer-to-member initialization from incompatible one
8336       accepted
8337     * [210]16889 ambiguity is not detected
8338     * [211]16959 Segmentation fault in ios_base::sync_with_stdio
8339
8340    Java compiler and library
8341
8342     * [212]7587 direct threaded interpreter not thread-safe
8343     * [213]16473 ServerSocket accept() leaks file descriptors
8344     * [214]16478 Hash synchronization deadlock with finalizers
8345
8346    Alpha-specific
8347
8348     * [215]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c
8349     * [216]16974 could not split insn (ice in final_scan_insn, in
8350       final.c)
8351
8352    x86-specific
8353
8354     * [217]16298 ICE in output_operand
8355     * [218]17113 ICE with SSE2 intrinsics
8356
8357    x86-64 specific
8358
8359     * [219]14697 libstdc++ couldn't find 32bit libgcc_s
8360
8361    MIPS-specific
8362
8363     * [220]15869 [mips64] No NOP after LW (with -mips1 -O0)
8364     * [221]16325 [mips64] value profiling clobbers gp on mips
8365     * [222]16357 [mipsisa64-elf] ICE copying 7 bytes between extern
8366       char[]s
8367     * [223]16380 [mips64] Use of uninitialised register after dbra
8368       conversion
8369     * [224]16407 [mips64] Unaligned access to local variables
8370     * [225]16643 [mips64] verify_local_live_at_start ICE after
8371       crossjumping & cfgcleanup
8372
8373    ARM-specific
8374
8375     * [226]15927 THUMB -O2: strength-reduced iteration variable ends up
8376       off by 1
8377     * [227]15948 THUMB: ICE with non-commutative cbranch
8378     * [228]17019 THUMB: bad switch statement in md code for
8379       addsi3_cbranch_scratch
8380
8381    IA64-specific
8382
8383     * [229]16130 ICE on valid code: in bundling, in config/ia64/ia64.c
8384       (-mtune=merced)
8385     * [230]16142 ICE on valid code: in bundling, in config/ia64/ia64.c
8386       (-mtune=itanium)
8387     * [231]16278 Gcc failed to build Linux kernel with -mtune=merced
8388     * [232]16414 ICE on valid code: typo in comparison of asm_noperands
8389       result
8390     * [233]16445 ICE on valid code: don't count ignored insns
8391     * [234]16490 ICE (segfault) while compiling with -fprofile-use
8392     * [235]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS
8393
8394    PowerPC-specific
8395
8396     * [236]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x
8397     * [237]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions
8398       issue)
8399
8400    SPARC-specific
8401
8402     * [238]16199 ICE while compiling apache 2.0.49
8403     * [239]16416 -m64 doesn't imply -mcpu=v9 anymore
8404     * [240]16430 ICE when returning non-C aggregates larger than 16 bytes
8405
8406    Bugs specific to embedded processors
8407
8408     * [241]16379 [m32r] can't output large model function call of memcpy
8409     * [242]17093 [m32r] ICE with -msdata=use -O0
8410     * [243]17119 [m32r] ICE at switch case 0x8000
8411
8412    DJGPP-specific
8413
8414     * [244]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp
8415
8416    Alpha Tru64-specific
8417
8418     * [245]16210 libstdc++ gratuitously omits "long long" I/O
8419
8420    Testsuite, documentation issues (compiler is not affected):
8421
8422     * [246]15488 (libstdc++) possibly insufficient file permissions for
8423       executing test suite
8424     * [247]16250 ada/doctools runs makeinfo even in release tarball
8425     __________________________________________________________________
8426
8427GCC 3.4.3
8428
8429   This is the [248]list of problem reports (PRs) from GCC's bug tracking
8430   system that are known to be fixed in the 3.4.3 release. This list might
8431   not be complete (that is, it is possible that some PRs that have been
8432   fixed are not listed here).
8433
8434    Bootstrap failures
8435
8436     * [249]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1
8437     * [250]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold
8438       when undeclared
8439
8440    Internal compiler errors (ICEs) affecting multiple platforms
8441
8442     * [251]13948 (java) GCJ segmentation fault while compiling GL4Java
8443       .class files
8444     * [252]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
8445     * [253]16301 (c++) ICE when "strong" attribute is attached to a using
8446       directive
8447     * [254]16566 ICE with flexible arrays
8448     * [255]17023 ICE with nested functions in parameter declaration
8449     * [256]17027 ICE with noreturn function in loop at -O2
8450     * [257]17524 ICE in grokdeclarator, in cp/decl.c
8451     * [258]17826 (c++) ICE in cp_tree_equal
8452
8453    C and optimization bugs
8454
8455     * [259]15526 -ftrapv aborts on 0 * (-1)
8456     * [260]16999 #ident stopped working
8457     * [261]17503 quadratic behaviour in invalid_mode_change_p
8458     * [262]17581 Long long arithmetic fails inside a switch/case
8459       statement when compiled with -O2
8460     * [263]18129 -fwritable-strings doesn't work
8461
8462    C++ compiler and library bugs
8463
8464     * [264]10975 incorrect initial ostringstream::tellp()
8465     * [265]11722 Unbuffered filebuf::sgetn is slow
8466     * [266]14534 Unrecognizing static function as a template parameter
8467       when its return value is also templated
8468     * [267]15172 Copy constructor optimization in aggregate
8469       initialization
8470     * [268]15786 Bad error message for frequently occuring error.
8471     * [269]16162 Rejects valid member-template-definition
8472     * [270]16612 empty basic_strings can't live in shared memory
8473     * [271]16715 std::basic_iostream is instantiated when used, even
8474       though instantiations are already contained in libstdc++
8475     * [272]16848 code in /ext/demangle.h appears broken
8476     * [273]17132 GCC fails to eliminate function template specialization
8477       when argument deduction fails
8478     * [274]17259 One more _S_leaf incorrectly qualified with _RopeRep::
8479       in ropeimpl.h
8480     * [275]17327 use of `enumeral_type' in template type unification
8481     * [276]17393 "unused variable '._0'" warning with -Wall
8482     * [277]17501 Confusion with member templates
8483     * [278]17537 g++ not passing -lstdc++ to linker when all command line
8484       arguments are libraries
8485     * [279]17585 usage of unqualified name of static member from within
8486       class not allowed
8487     * [280]17821 Poor diagnostic for using "." instead of "->"
8488     * [281]17829 wrong error: call of overloaded function is ambiguous
8489     * [282]17851 Misleading diagnostic for invalid function declarations
8490       with undeclared types
8491     * [283]17976 Destructor is called twice
8492     * [284]18020 rejects valid definition of enum value in template
8493     * [285]18093 bogus conflict in namespace aliasing
8494     * [286]18140 C++ parser bug when using >> in templates
8495
8496    Fortran
8497
8498     * [287]17541 data statements with double precision constants fail
8499
8500    x86-specific
8501
8502     * [288]17853 -O2 ICE for MMX testcase
8503
8504    SPARC-specific
8505
8506     * [289]17245 ICE compiling gsl-1.5 statistics/lag1.c
8507
8508    Darwin-specific
8509
8510     * [290]17167 FATAL:Symbol L_foo$stub already defined.
8511
8512    AIX-specific
8513
8514     * [291]17277 could not catch an exception when specified -maix64
8515
8516    Solaris-specific
8517
8518     * [292]17505 <cmath> calls acosf(), ceilf(), and other functions
8519       missing from system libraries
8520
8521    HP/UX specific:
8522
8523     * [293]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl
8524
8525    ARM-specific
8526
8527     * [294]17384 ICE with mode attribute on structures
8528
8529    MIPS-specific
8530
8531     * [295]17770 No NOP after LWL with -mips1
8532
8533    Other embedded target specific
8534
8535     * [296]11476 [arc-elf] gcc ICE on newlib's vfprintf.c
8536     * [297]14064 [avr-elf] -fdata-sections triggers ICE
8537     * [298]14678 [m68hc11-elf] gcc ICE
8538     * [299]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__
8539     * [300]15790 [i686-coff] Alignment error building gcc with i686-coff
8540       target
8541     * [301]15886 [SH] Miscompilation with -O2 -fPIC
8542     * [302]16884 [avr-elf] [fweb related] bug while initializing
8543       variables
8544
8545    Bugs relating to debugger support
8546
8547     * [303]13841 missing debug info for _Complex function arguments
8548     * [304]15860 [big-endian targets] No DW_AT_location debug info is
8549       emitted for formal arguments to a function that uses "register"
8550       qualifiers
8551
8552    Testsuite issues (compiler not affected)
8553
8554     * [305]17465 Testsuite in libffi overrides LD_LIBRARY_PATH
8555     * [306]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH
8556     * [307]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit
8557       testsuite
8558
8559    Documentation
8560
8561     * [308]15498 typo in gcc manual: non-existing locale example en_UK,
8562       should be en_GB
8563     * [309]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap:
8564       document broken shell
8565     * [310]16406 USE_LD_AS_NEEDED undocumented
8566     __________________________________________________________________
8567
8568GCC 3.4.4
8569
8570   This is the [311]list of problem reports (PRs) from GCC's bug tracking
8571   system that are known to be fixed in the 3.4.4 release. This list might
8572   not be complete (that is, it is possible that some PRs that have been
8573   fixed are not listed here).
8574     __________________________________________________________________
8575
8576GCC 3.4.5
8577
8578   This is the [312]list of problem reports (PRs) from GCC's bug tracking
8579   system that are known to be fixed in the 3.4.5 release. This list might
8580   not be complete (that is, it is possible that some PRs that have been
8581   fixed are not listed here).
8582
8583    Bootstrap issues
8584
8585     * [313]24688 sco_math fixincl breaks math.h
8586
8587    C compiler bugs
8588
8589     * [314]17188 struct Foo { } redefinition
8590     * [315]20187 wrong code for ((unsigned char)(unsigned long
8591       long)((a?a:1)&(a*b)))?0:1)
8592     * [316]21873 infinite warning loop on bad array initializer
8593     * [317]21899 enum definition accepts values to be overriden
8594     * [318]22061 ICE in find_function_data, in function.c
8595     * [319]22308 Failure to diagnose violation of constraint 6.516p2
8596     * [320]22458 ICE on missing brace
8597     * [321]22589 ICE casting to long long
8598     * [322]24101 Segfault with preprocessed source
8599
8600    C++ compiler and library bugs
8601
8602     * [323]10611 operations on vector mode not recognized in C++
8603     * [324]13377 unexpected behavior of namespace usage directive
8604     * [325]16002 Strange error message with new parser
8605     * [326]17413 local classes as template argument
8606     * [327]17609 spurious error message after using keyword
8607     * [328]17618 ICE in cp_convert_to_pointer, in cp/cvt.c
8608     * [329]18124 ICE with invalid template template parameter
8609     * [330]18155 typedef in template declaration not rejected
8610     * [331]18177 ICE with const_cast for undeclared variable
8611     * [332]18368 C++ error message regression
8612     * [333]16378 ICE when returning a copy of a packed member
8613     * [334]18466 int ::i; accepted
8614     * [335]18512 ICE on invalid usage of template base class
8615     * [336]18454 ICE when returning undefined type
8616     * [337]18738 typename not allowed with non-dependent qualified name
8617     * [338]18803 rejects access to operator() in template
8618     * [339]19004 ICE in uses_template_parms, in cp/pt.c
8619     * [340]19208 Spurious error about variably modified type
8620     * [341]18253 bad error message / ICE for invalid template parameter
8621     * [342]19608 ICE after friend function definition in local class
8622     * [343]19884 ICE on explicit instantiation of a non-template
8623       constructor
8624     * [344]20153 ICE when C++ template function contains anonymous union
8625     * [345]20563 Infinite loop in diagnostic (and ice after error
8626       message)
8627     * [346]20789 ICE with incomplete type in template
8628     * [347]21336 Internal compiler error when using custom new operators
8629     * [348]21768 ICE in error message due to violation of coding
8630       conventions
8631     * [349]21853 constness of pointer to data member ignored
8632     * [350]21903 Default argument of template function causes a
8633       compile-time error
8634     * [351]21983 multiple diagnostics
8635     * [352]21987 New testsuite failure
8636       g++.dg/warn/conversion-function-1.C
8637     * [353]22153 ICE on invalid template specialization
8638     * [354]22172 Internal compiler error, seg fault.
8639     * [355]21286 filebuf::xsgetn vs pipes
8640     * [356]22233 ICE with wrong number of template parameters
8641     * [357]22508 ICE after invalid operator new
8642     * [358]22545 ICE with pointer to class member & user defined
8643       conversion operator
8644     * [359]23528 Wrong default allocator in ext/hash_map
8645     * [360]23550 char_traits requirements/1.cc test bad math
8646     * [361]23586 Bad diagnostic for invalid namespace-name
8647     * [362]23624 ICE in invert_truthvalue, in fold-const.c
8648     * [363]23639 Bad error message: not a member of '<declaration error>'
8649     * [364]23797 ICE on typename outside template
8650     * [365]23965 Bogus error message: no matching function for call to
8651       'foo(<type error>)'
8652     * [366]24052 &#`label_decl' not supported by dump_expr#<expression
8653       error>
8654     * [367]24580 virtual base class cause exception not to be caught
8655
8656    Problems in generated debug information
8657
8658     * [368]24267 Bad DWARF for altivec vectors
8659
8660    Optimizations issues
8661
8662     * [369]17810 ICE in verify_local_live_at_start
8663     * [370]17860 Wrong generated code for loop with varying bound
8664     * [371]21709 ICE on compile-time complex NaN
8665     * [372]21964 broken tail call at -O2 or more
8666     * [373]22167 Strange optimization bug when using -Os
8667     * [374]22619 Compilation failure for real_const_1.f and
8668       real_const_2.f90
8669     * [375]23241 Invalid code generated for comparison of uchar to 255
8670     * [376]23478 Miscompilation due to reloading of a var that is also
8671       used in EH pad
8672     * [377]24470 segmentation fault in cc1plus when compiling with -O
8673     * [378]24950 ICE in operand_subword_force
8674
8675    Precompiled headers problems
8676
8677     * [379]14400 Cannot compile qt-x11-free-3.3.0
8678     * [380]14940 PCH largefile test fails on various platforms
8679
8680    Preprocessor bugs
8681
8682     * [381]20239 ICE on empty preprocessed input
8683     * [382]15220 "gcc -E -MM -MG" reports missing system headers in
8684       source directory
8685
8686    Testsuite issues
8687
8688     * [383]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on
8689       i686-pc-linux-gnu
8690
8691    Alpha specific
8692
8693     * [384]21888 bootstrap failure with linker relaxation enabled
8694
8695    ARM specific
8696
8697     * [385]15342 [arm-linux]: ICE in verify_local_live_at_start
8698     * [386]23985 Memory aliasing information incorrect in inlined memcpy
8699
8700    ColdFile specific
8701
8702     * [387]16719 Illegal move of byte into address register causes
8703       compiler to ICE
8704
8705    HPPA specific
8706
8707     * [388]21723 ICE while building libgfortran
8708     * [389]21841 -mhp-ld/-mgnu-ld documentation
8709
8710    IA-64 specific
8711
8712     * [390]23644 IA-64 hardware models and configuration options
8713       documentation error
8714     * [391]24718 Shared libgcc not used for linking by default
8715
8716    M68000 specific
8717
8718     * [392]18421 ICE in reload_cse_simplify_operands, in postreload.c
8719
8720    MIPS specific
8721
8722     * [393]20621 ICE in change_address_1, in emit-rtl.c
8723
8724    PowerPC and PowerPC64 specific
8725
8726     * [394]18583 error on valid code: const
8727       __attribute__((altivec(vector__))) doesn't work in arrays
8728     * [395]20191 ICE in reload_cse_simplify_operands
8729     * [396]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined
8730     * [397]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set
8731     * [398]23404 gij trashes args of functions with more than 8 fp args
8732     * [399]23539 C & C++ compiler generating misaligned references
8733       regardless of compiler flags
8734     * [400]24102 floatdisf2_internal2 broken
8735     * [401]24465 -mminimal-toc miscompilation of __thread vars
8736
8737    Solaris specific
8738
8739     * [402]19933 Problem with define of HUGE_VAL in math_c99
8740     * [403]21889 Native Solaris assembler cannot grok DTP-relative debug
8741       symbols
8742
8743    SPARC specific
8744
8745     * [404]19300 PCH failures on sparc-linux
8746     * [405]20301 Assembler labels have a leading "-"
8747     * [406]20673 C PCH testsuite assembly comparison failure
8748
8749    x86 and x86_64 specific
8750
8751     * [407]18582 ICE with arrays of type V2DF
8752     * [408]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2
8753       -fsched2-use-traces
8754     * [409]21716 ICE in reg-stack.c's swap_rtx_condition
8755     * [410]24315 amd64 fails -fpeephole2
8756     __________________________________________________________________
8757
8758GCC 3.4.6
8759
8760   This is the [411]list of problem reports (PRs) from GCC's bug tracking
8761   system that are known to be fixed in the 3.4.6 release. This list might
8762   not be complete (that is, it is possible that some PRs that have been
8763   fixed are not listed here).
8764
8765
8766    For questions related to the use of GCC, please consult these web
8767    pages and the [412]GCC manuals. If that fails, the
8768    [413]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8769    web pages and the development of GCC are welcome on our developer
8770    list at [414]gcc@gcc.gnu.org. All of [415]our lists have public
8771    archives.
8772
8773   Copyright (C) [416]Free Software Foundation, Inc. Verbatim copying and
8774   distribution of this entire article is permitted in any medium,
8775   provided this notice is preserved.
8776
8777   These pages are [417]maintained by the GCC team. Last modified
8778   2014-06-28[418].
8779
8780References
8781
8782   1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
8783   2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus
8784   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
8785   4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems
8786   5. http://gcc.gnu.org/gcc-3.4/mips-abi.html
8787   6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
8788   7. http://www.boost.org/
8789   8. https://gcc.gnu.org/PR11953
8790   9. https://gcc.gnu.org/PR8361
8791  10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins
8792  11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209
8793  12. http://gcc.gnu.org/bugs/#cxx_rvalbind
8794  13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
8795  14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
8796  15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
8797  16. http://www.gnu.org/software/classpath/
8798  17. http://www.eclipse.org/
8799  18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html
8800  19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html
8801  20. http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51A_HTML/ARH9MBTE/DTMNPLTN.HTM#normal-argument-list-structure
8802  21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
8803  22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html
8804  23. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
8805  24. http://gcc.gnu.org/gcc-3.4/mips-abi.html
8806  25. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html
8807  26. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
8808  27. 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
8809  28. https://gcc.gnu.org/PR10129
8810  29. https://gcc.gnu.org/PR14576
8811  30. https://gcc.gnu.org/PR14760
8812  31. https://gcc.gnu.org/PR14671
8813  32. https://gcc.gnu.org/PR15093
8814  33. https://gcc.gnu.org/PR15178
8815  34. https://gcc.gnu.org/PR12753
8816  35. https://gcc.gnu.org/PR13985
8817  36. https://gcc.gnu.org/PR14810
8818  37. https://gcc.gnu.org/PR14883
8819  38. https://gcc.gnu.org/PR15044
8820  39. https://gcc.gnu.org/PR15057
8821  40. https://gcc.gnu.org/PR15064
8822  41. https://gcc.gnu.org/PR15142
8823  42. https://gcc.gnu.org/PR15159
8824  43. https://gcc.gnu.org/PR15165
8825  44. https://gcc.gnu.org/PR15193
8826  45. https://gcc.gnu.org/PR15209
8827  46. https://gcc.gnu.org/PR15227
8828  47. https://gcc.gnu.org/PR15285
8829  48. https://gcc.gnu.org/PR15299
8830  49. https://gcc.gnu.org/PR15329
8831  50. https://gcc.gnu.org/PR15550
8832  51. https://gcc.gnu.org/PR15554
8833  52. https://gcc.gnu.org/PR15640
8834  53. https://gcc.gnu.org/PR15666
8835  54. https://gcc.gnu.org/PR15696
8836  55. https://gcc.gnu.org/PR15701
8837  56. https://gcc.gnu.org/PR15761
8838  57. https://gcc.gnu.org/PR15829
8839  58. https://gcc.gnu.org/PR14538
8840  59. https://gcc.gnu.org/PR12391
8841  60. https://gcc.gnu.org/PR14649
8842  61. https://gcc.gnu.org/PR15004
8843  62. https://gcc.gnu.org/PR15749
8844  63. https://gcc.gnu.org/PR10646
8845  64. https://gcc.gnu.org/PR12077
8846  65. https://gcc.gnu.org/PR13598
8847  66. https://gcc.gnu.org/PR14211
8848  67. https://gcc.gnu.org/PR14220
8849  68. https://gcc.gnu.org/PR14245
8850  69. https://gcc.gnu.org/PR14340
8851  70. https://gcc.gnu.org/PR14600
8852  71. https://gcc.gnu.org/PR14668
8853  72. https://gcc.gnu.org/PR14775
8854  73. https://gcc.gnu.org/PR14821
8855  74. https://gcc.gnu.org/PR14930
8856  75. https://gcc.gnu.org/PR14932
8857  76. https://gcc.gnu.org/PR14950
8858  77. https://gcc.gnu.org/PR14962
8859  78. https://gcc.gnu.org/PR14975
8860  79. https://gcc.gnu.org/PR15002
8861  80. https://gcc.gnu.org/PR15025
8862  81. https://gcc.gnu.org/PR15046
8863  82. https://gcc.gnu.org/PR15069
8864  83. https://gcc.gnu.org/PR15074
8865  84. https://gcc.gnu.org/PR15083
8866  85. https://gcc.gnu.org/PR15096
8867  86. https://gcc.gnu.org/PR15287
8868  87. https://gcc.gnu.org/PR15317
8869  88. https://gcc.gnu.org/PR15337
8870  89. https://gcc.gnu.org/PR15361
8871  90. https://gcc.gnu.org/PR15412
8872  91. https://gcc.gnu.org/PR15427
8873  92. https://gcc.gnu.org/PR15471
8874  93. https://gcc.gnu.org/PR15503
8875  94. https://gcc.gnu.org/PR15507
8876  95. https://gcc.gnu.org/PR15542
8877  96. https://gcc.gnu.org/PR15565
8878  97. https://gcc.gnu.org/PR15625
8879  98. https://gcc.gnu.org/PR15629
8880  99. https://gcc.gnu.org/PR15742
8881 100. https://gcc.gnu.org/PR15775
8882 101. https://gcc.gnu.org/PR15821
8883 102. https://gcc.gnu.org/PR15862
8884 103. https://gcc.gnu.org/PR15875
8885 104. https://gcc.gnu.org/PR15877
8886 105. https://gcc.gnu.org/PR15947
8887 106. https://gcc.gnu.org/PR16020
8888 107. https://gcc.gnu.org/PR16154
8889 108. https://gcc.gnu.org/PR16174
8890 109. https://gcc.gnu.org/PR14315
8891 110. https://gcc.gnu.org/PR15151
8892 111. https://gcc.gnu.org/PR7993
8893 112. https://gcc.gnu.org/PR15228
8894 113. https://gcc.gnu.org/PR15345
8895 114. https://gcc.gnu.org/PR15945
8896 115. https://gcc.gnu.org/PR15526
8897 116. https://gcc.gnu.org/PR14690
8898 117. https://gcc.gnu.org/PR15112
8899 118. https://gcc.gnu.org/PR15067
8900 119. https://gcc.gnu.org/PR1963
8901 120. https://gcc.gnu.org/PR15717
8902 121. https://gcc.gnu.org/PR14782
8903 122. https://gcc.gnu.org/PR14828
8904 123. https://gcc.gnu.org/PR15202
8905 124. https://gcc.gnu.org/PR14610
8906 125. https://gcc.gnu.org/PR14813
8907 126. https://gcc.gnu.org/PR14857
8908 127. https://gcc.gnu.org/PR15598
8909 128. https://gcc.gnu.org/PR15653
8910 129. https://gcc.gnu.org/PR15189
8911 130. https://gcc.gnu.org/PR15331
8912 131. https://gcc.gnu.org/PR16144
8913 132. https://gcc.gnu.org/PR16176
8914 133. https://gcc.gnu.org/PR11591
8915 134. https://gcc.gnu.org/PR12028
8916 135. https://gcc.gnu.org/PR14478
8917 136. https://gcc.gnu.org/PR14567
8918 137. https://gcc.gnu.org/PR14715
8919 138. https://gcc.gnu.org/PR14902
8920 139. https://gcc.gnu.org/PR14924
8921 140. https://gcc.gnu.org/PR14960
8922 141. https://gcc.gnu.org/PR15106
8923 142. https://gcc.gnu.org/PR16026
8924 143. https://gcc.gnu.org/PR15191
8925 144. https://gcc.gnu.org/PR15662
8926 145. https://gcc.gnu.org/PR15054
8927 146. https://gcc.gnu.org/PR15783
8928 147. https://gcc.gnu.org/PR15626
8929 148. https://gcc.gnu.org/PR14326
8930 149. https://gcc.gnu.org/PR14723
8931 150. https://gcc.gnu.org/PR15290
8932 151. https://gcc.gnu.org/PR15250
8933 152. https://gcc.gnu.org/PR15551
8934 153. https://gcc.gnu.org/PR8309
8935 154. https://gcc.gnu.org/PR13250
8936 155. https://gcc.gnu.org/PR13803
8937 156. https://gcc.gnu.org/PR14093
8938 157. https://gcc.gnu.org/PR14457
8939 158. https://gcc.gnu.org/PR14542
8940 159. https://gcc.gnu.org/PR15100
8941 160. https://gcc.gnu.org/PR15296
8942 161. https://gcc.gnu.org/PR15396
8943 162. https://gcc.gnu.org/PR15782
8944 163. https://gcc.gnu.org/PR11610
8945 164. https://gcc.gnu.org/PR15488
8946 165. https://gcc.gnu.org/PR15489
8947 166. https://gcc.gnu.org/PR13928
8948 167. https://gcc.gnu.org/PR14150
8949 168. https://gcc.gnu.org/PR14949
8950 169. https://gcc.gnu.org/PR15123
8951 170. https://gcc.gnu.org/PR16469
8952 171. https://gcc.gnu.org/PR16344
8953 172. https://gcc.gnu.org/PR16842
8954 173. https://gcc.gnu.org/PR12608
8955 174. https://gcc.gnu.org/PR14492
8956 175. https://gcc.gnu.org/PR15461
8957 176. https://gcc.gnu.org/PR15890
8958 177. https://gcc.gnu.org/PR16180
8959 178. https://gcc.gnu.org/PR16224
8960 179. https://gcc.gnu.org/PR16408
8961 180. https://gcc.gnu.org/PR16529
8962 181. https://gcc.gnu.org/PR16698
8963 182. https://gcc.gnu.org/PR16706
8964 183. https://gcc.gnu.org/PR16810
8965 184. https://gcc.gnu.org/PR16851
8966 185. https://gcc.gnu.org/PR16870
8967 186. https://gcc.gnu.org/PR16904
8968 187. https://gcc.gnu.org/PR16905
8969 188. https://gcc.gnu.org/PR16964
8970 189. https://gcc.gnu.org/PR17068
8971 190. https://gcc.gnu.org/PR16366
8972 191. https://gcc.gnu.org/PR15345
8973 192. https://gcc.gnu.org/PR16590
8974 193. https://gcc.gnu.org/PR16693
8975 194. https://gcc.gnu.org/PR17078
8976 195. https://gcc.gnu.org/PR13956
8977 196. https://gcc.gnu.org/PR16684
8978 197. https://gcc.gnu.org/PR12658
8979 198. https://gcc.gnu.org/PR13092
8980 199. https://gcc.gnu.org/PR15320
8981 200. https://gcc.gnu.org/PR16246
8982 201. https://gcc.gnu.org/PR16273
8983 202. https://gcc.gnu.org/PR16401
8984 203. https://gcc.gnu.org/PR16411
8985 204. https://gcc.gnu.org/PR16489
8986 205. https://gcc.gnu.org/PR16618
8987 206. https://gcc.gnu.org/PR16637
8988 207. https://gcc.gnu.org/PR16717
8989 208. https://gcc.gnu.org/PR16813
8990 209. https://gcc.gnu.org/PR16853
8991 210. https://gcc.gnu.org/PR16889
8992 211. https://gcc.gnu.org/PR16959
8993 212. https://gcc.gnu.org/PR7587
8994 213. https://gcc.gnu.org/PR16473
8995 214. https://gcc.gnu.org/PR16478
8996 215. https://gcc.gnu.org/PR10695
8997 216. https://gcc.gnu.org/PR16974
8998 217. https://gcc.gnu.org/PR16298
8999 218. https://gcc.gnu.org/PR17113
9000 219. https://gcc.gnu.org/PR14697
9001 220. https://gcc.gnu.org/PR15869
9002 221. https://gcc.gnu.org/PR16325
9003 222. https://gcc.gnu.org/PR16357
9004 223. https://gcc.gnu.org/PR16380
9005 224. https://gcc.gnu.org/PR16407
9006 225. https://gcc.gnu.org/PR16643
9007 226. https://gcc.gnu.org/PR15927
9008 227. https://gcc.gnu.org/PR15948
9009 228. https://gcc.gnu.org/PR17019
9010 229. https://gcc.gnu.org/PR16130
9011 230. https://gcc.gnu.org/PR16142
9012 231. https://gcc.gnu.org/PR16278
9013 232. https://gcc.gnu.org/PR16414
9014 233. https://gcc.gnu.org/PR16445
9015 234. https://gcc.gnu.org/PR16490
9016 235. https://gcc.gnu.org/PR16683
9017 236. https://gcc.gnu.org/PR16195
9018 237. https://gcc.gnu.org/PR16239
9019 238. https://gcc.gnu.org/PR16199
9020 239. https://gcc.gnu.org/PR16416
9021 240. https://gcc.gnu.org/PR16430
9022 241. https://gcc.gnu.org/PR16379
9023 242. https://gcc.gnu.org/PR17093
9024 243. https://gcc.gnu.org/PR17119
9025 244. https://gcc.gnu.org/PR15928
9026 245. https://gcc.gnu.org/PR16210
9027 246. https://gcc.gnu.org/PR15488
9028 247. https://gcc.gnu.org/PR16250
9029 248. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3
9030 249. https://gcc.gnu.org/PR17369
9031 250. https://gcc.gnu.org/PR17850
9032 251. https://gcc.gnu.org/PR13948
9033 252. https://gcc.gnu.org/PR14492
9034 253. https://gcc.gnu.org/PR16301
9035 254. https://gcc.gnu.org/PR16566
9036 255. https://gcc.gnu.org/PR17023
9037 256. https://gcc.gnu.org/PR17027
9038 257. https://gcc.gnu.org/PR17524
9039 258. https://gcc.gnu.org/PR17826
9040 259. https://gcc.gnu.org/PR15526
9041 260. https://gcc.gnu.org/PR16999
9042 261. https://gcc.gnu.org/PR17503
9043 262. https://gcc.gnu.org/PR17581
9044 263. https://gcc.gnu.org/PR18129
9045 264. https://gcc.gnu.org/PR10975
9046 265. https://gcc.gnu.org/PR11722
9047 266. https://gcc.gnu.org/PR14534
9048 267. https://gcc.gnu.org/PR15172
9049 268. https://gcc.gnu.org/PR15786
9050 269. https://gcc.gnu.org/PR16162
9051 270. https://gcc.gnu.org/PR16612
9052 271. https://gcc.gnu.org/PR16715
9053 272. https://gcc.gnu.org/PR16848
9054 273. https://gcc.gnu.org/PR17132
9055 274. https://gcc.gnu.org/PR17259
9056 275. https://gcc.gnu.org/PR17327
9057 276. https://gcc.gnu.org/PR17393
9058 277. https://gcc.gnu.org/PR17501
9059 278. https://gcc.gnu.org/PR17537
9060 279. https://gcc.gnu.org/PR17585
9061 280. https://gcc.gnu.org/PR17821
9062 281. https://gcc.gnu.org/PR17829
9063 282. https://gcc.gnu.org/PR17851
9064 283. https://gcc.gnu.org/PR17976
9065 284. https://gcc.gnu.org/PR18020
9066 285. https://gcc.gnu.org/PR18093
9067 286. https://gcc.gnu.org/PR18140
9068 287. https://gcc.gnu.org/PR17541
9069 288. https://gcc.gnu.org/PR17853
9070 289. https://gcc.gnu.org/PR17245
9071 290. https://gcc.gnu.org/PR17167
9072 291. https://gcc.gnu.org/PR17277
9073 292. https://gcc.gnu.org/PR17505
9074 293. https://gcc.gnu.org/PR17684
9075 294. https://gcc.gnu.org/PR17384
9076 295. https://gcc.gnu.org/PR17770
9077 296. https://gcc.gnu.org/PR11476
9078 297. https://gcc.gnu.org/PR14064
9079 298. https://gcc.gnu.org/PR14678
9080 299. https://gcc.gnu.org/PR15583
9081 300. https://gcc.gnu.org/PR15790
9082 301. https://gcc.gnu.org/PR15886
9083 302. https://gcc.gnu.org/PR16884
9084 303. https://gcc.gnu.org/PR13841
9085 304. https://gcc.gnu.org/PR15860
9086 305. https://gcc.gnu.org/PR17465
9087 306. https://gcc.gnu.org/PR17469
9088 307. https://gcc.gnu.org/PR18138
9089 308. https://gcc.gnu.org/PR15498
9090 309. https://gcc.gnu.org/PR15747
9091 310. https://gcc.gnu.org/PR16406
9092 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4
9093 312. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5
9094 313. https://gcc.gnu.org/PR24688
9095 314. https://gcc.gnu.org/PR17188
9096 315. https://gcc.gnu.org/PR20187
9097 316. https://gcc.gnu.org/PR21873
9098 317. https://gcc.gnu.org/PR21899
9099 318. https://gcc.gnu.org/PR22061
9100 319. https://gcc.gnu.org/PR22208
9101 320. https://gcc.gnu.org/PR22458
9102 321. https://gcc.gnu.org/PR22589
9103 322. https://gcc.gnu.org/PR24101
9104 323. https://gcc.gnu.org/PR10611
9105 324. https://gcc.gnu.org/PR13377
9106 325. https://gcc.gnu.org/PR16002
9107 326. https://gcc.gnu.org/PR17413
9108 327. https://gcc.gnu.org/PR17609
9109 328. https://gcc.gnu.org/PR17618
9110 329. https://gcc.gnu.org/PR18124
9111 330. https://gcc.gnu.org/PR18155
9112 331. https://gcc.gnu.org/PR18177
9113 332. https://gcc.gnu.org/PR18368
9114 333. https://gcc.gnu.org/PR18378
9115 334. https://gcc.gnu.org/PR18466
9116 335. https://gcc.gnu.org/PR18512
9117 336. https://gcc.gnu.org/PR18545
9118 337. https://gcc.gnu.org/PR18738
9119 338. https://gcc.gnu.org/PR18803
9120 339. https://gcc.gnu.org/PR19004
9121 340. https://gcc.gnu.org/PR19208
9122 341. https://gcc.gnu.org/PR19253
9123 342. https://gcc.gnu.org/PR19608
9124 343. https://gcc.gnu.org/PR19884
9125 344. https://gcc.gnu.org/PR20153
9126 345. https://gcc.gnu.org/PR20563
9127 346. https://gcc.gnu.org/PR20789
9128 347. https://gcc.gnu.org/PR21336
9129 348. https://gcc.gnu.org/PR21768
9130 349. https://gcc.gnu.org/PR21853
9131 350. https://gcc.gnu.org/PR21903
9132 351. https://gcc.gnu.org/PR21983
9133 352. https://gcc.gnu.org/PR21987
9134 353. https://gcc.gnu.org/PR22153
9135 354. https://gcc.gnu.org/PR22172
9136 355. https://gcc.gnu.org/PR21286
9137 356. https://gcc.gnu.org/PR22233
9138 357. https://gcc.gnu.org/PR22508
9139 358. https://gcc.gnu.org/PR22545
9140 359. https://gcc.gnu.org/PR23528
9141 360. https://gcc.gnu.org/PR23550
9142 361. https://gcc.gnu.org/PR23586
9143 362. https://gcc.gnu.org/PR23624
9144 363. https://gcc.gnu.org/PR23639
9145 364. https://gcc.gnu.org/PR23797
9146 365. https://gcc.gnu.org/PR23965
9147 366. https://gcc.gnu.org/PR24052
9148 367. https://gcc.gnu.org/PR24580
9149 368. https://gcc.gnu.org/PR24267
9150 369. https://gcc.gnu.org/PR17810
9151 370. https://gcc.gnu.org/PR17860
9152 371. https://gcc.gnu.org/PR21709
9153 372. https://gcc.gnu.org/PR21964
9154 373. https://gcc.gnu.org/PR22167
9155 374. https://gcc.gnu.org/PR22619
9156 375. https://gcc.gnu.org/PR23241
9157 376. https://gcc.gnu.org/PR23478
9158 377. https://gcc.gnu.org/PR24470
9159 378. https://gcc.gnu.org/PR24950
9160 379. https://gcc.gnu.org/PR14400
9161 380. https://gcc.gnu.org/PR14940
9162 381. https://gcc.gnu.org/PR20239
9163 382. https://gcc.gnu.org/PR15220
9164 383. https://gcc.gnu.org/PR19275
9165 384. https://gcc.gnu.org/PR21888
9166 385. https://gcc.gnu.org/PR15342
9167 386. https://gcc.gnu.org/PR23985
9168 387. https://gcc.gnu.org/PR16719
9169 388. https://gcc.gnu.org/PR21723
9170 389. https://gcc.gnu.org/PR21841
9171 390. https://gcc.gnu.org/PR23644
9172 391. https://gcc.gnu.org/PR24718
9173 392. https://gcc.gnu.org/PR18421
9174 393. https://gcc.gnu.org/PR20621
9175 394. https://gcc.gnu.org/PR18583
9176 395. https://gcc.gnu.org/PR20191
9177 396. https://gcc.gnu.org/PR22083
9178 397. https://gcc.gnu.org/PR23070
9179 398. https://gcc.gnu.org/PR23404
9180 399. https://gcc.gnu.org/PR23539
9181 400. https://gcc.gnu.org/PR24102
9182 401. https://gcc.gnu.org/PR24465
9183 402. https://gcc.gnu.org/PR19933
9184 403. https://gcc.gnu.org/PR21889
9185 404. https://gcc.gnu.org/PR19300
9186 405. https://gcc.gnu.org/PR20301
9187 406. https://gcc.gnu.org/PR20673
9188 407. https://gcc.gnu.org/PR18582
9189 408. https://gcc.gnu.org/PR19340
9190 409. https://gcc.gnu.org/PR21716
9191 410. https://gcc.gnu.org/PR24315
9192 411. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6
9193 412. https://gcc.gnu.org/onlinedocs/
9194 413. mailto:gcc-help@gcc.gnu.org
9195 414. mailto:gcc@gcc.gnu.org
9196 415. https://gcc.gnu.org/lists.html
9197 416. http://www.fsf.org/
9198 417. https://gcc.gnu.org/about.html
9199 418. http://validator.w3.org/check/referer
9200======================================================================
9201http://gcc.gnu.org/gcc-3.3/index.html
9202
9203                             GCC 3.3 Release Series
9204
9205   May 03, 2005
9206
9207   The [1]GNU project and the GCC developers are pleased to announce the
9208   release of GCC 3.3.6.
9209
9210   This release is a bug-fix release, containing fixes for regressions in
9211   GCC 3.3.5 relative to previous releases of GCC.
9212
9213   This release is the last of the series 3.3.x.
9214
9215   The GCC 3.3 release series includes numerous [2]new features,
9216   improvements, bug fixes, and other changes, thanks to an [3]amazing
9217   group of volunteers.
9218
9219Release History
9220
9221   GCC 3.3.6
9222          May 3, 2005 ([4]changes)
9223
9224   GCC 3.3.5
9225          September 30, 2004 ([5]changes)
9226
9227   GCC 3.3.4
9228          May 31, 2004 ([6]changes)
9229
9230   GCC 3.3.3
9231          February 14, 2004 ([7]changes)
9232
9233   GCC 3.3.2
9234          October 16, 2003 ([8]changes)
9235
9236   GCC 3.3.1
9237          August 8, 2003 ([9]changes)
9238
9239   GCC 3.3
9240          May 14, 2003 ([10]changes)
9241
9242References and Acknowledgements
9243
9244   GCC used to stand for the GNU C Compiler, but since the compiler
9245   supports several other languages aside from C, it now stands for the
9246   GNU Compiler Collection.
9247
9248   A list of [11]successful builds is updated as new information becomes
9249   available.
9250
9251   The GCC developers would like to thank the numerous people that have
9252   contributed new features, improvements, bug fixes, and other changes as
9253   well as test results to GCC. This [12]amazing group of volunteers is
9254   what makes GCC successful.
9255
9256   For additional information about GCC please refer to the [13]GCC
9257   project web site or contact the [14]GCC development mailing list.
9258
9259   To obtain GCC please use [15]our mirror sites, or our CVS server.
9260
9261
9262    For questions related to the use of GCC, please consult these web
9263    pages and the [16]GCC manuals. If that fails, the
9264    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9265    web pages and the development of GCC are welcome on our developer
9266    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
9267    archives.
9268
9269   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
9270   distribution of this entire article is permitted in any medium,
9271   provided this notice is preserved.
9272
9273   These pages are [21]maintained by the GCC team. Last modified
9274   2014-06-28[22].
9275
9276References
9277
9278   1. http://www.gnu.org/
9279   2. http://gcc.gnu.org/gcc-3.3/changes.html
9280   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9281   4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
9282   5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5
9283   6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4
9284   7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3
9285   8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2
9286   9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1
9287  10. http://gcc.gnu.org/gcc-3.3/changes.html
9288  11. http://gcc.gnu.org/gcc-3.3/buildstat.html
9289  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9290  13. http://gcc.gnu.org/index.html
9291  14. mailto:gcc@gcc.gnu.org
9292  15. http://gcc.gnu.org/mirrors.html
9293  16. https://gcc.gnu.org/onlinedocs/
9294  17. mailto:gcc-help@gcc.gnu.org
9295  18. mailto:gcc@gcc.gnu.org
9296  19. https://gcc.gnu.org/lists.html
9297  20. http://www.fsf.org/
9298  21. https://gcc.gnu.org/about.html
9299  22. http://validator.w3.org/check/referer
9300======================================================================
9301http://gcc.gnu.org/gcc-3.3/changes.html
9302
9303                             GCC 3.3 Release Series
9304                        Changes, New Features, and Fixes
9305
9306   The latest release in the 3.3 release series is [1]GCC 3.3.6.
9307
9308Caveats
9309
9310     * The preprocessor no longer accepts multi-line string literals. They
9311       were deprecated in 3.0, 3.1, and 3.2.
9312     * The preprocessor no longer supports the -A- switch when appearing
9313       alone. -A- followed by an assertion is still supported.
9314     * Support for all the systems [2]obsoleted in GCC 3.1 has been
9315       removed from GCC 3.3. See below for a [3]list of systems which are
9316       obsoleted in this release.
9317     * Checking for null format arguments has been decoupled from the rest
9318       of the format checking mechanism. Programs which use the format
9319       attribute may regain this functionality by using the new [4]nonnull
9320       function attribute. Note that all functions for which GCC has a
9321       built-in format attribute, an appropriate built-in nonnull
9322       attribute is also applied.
9323     * The DWARF (version 1) debugging format has been deprecated and will
9324       be removed in a future version of GCC. Version 2 of the DWARF
9325       debugging format will continue to be supported for the foreseeable
9326       future.
9327     * The C and Objective-C compilers no longer accept the "Naming Types"
9328       extension (typedef foo = bar); it was already unavailable in C++.
9329       Code which uses it will need to be changed to use the "typeof"
9330       extension instead: typedef typeof(bar) foo. (We have removed this
9331       extension without a period of deprecation because it has caused the
9332       compiler to crash since version 3.0 and no one noticed until very
9333       recently. Thus we conclude it is not in widespread use.)
9334     * The -traditional C compiler option has been removed. It was
9335       deprecated in 3.1 and 3.2. (Traditional preprocessing remains
9336       available.) The <varargs.h> header, used for writing variadic
9337       functions in traditional C, still exists but will produce an error
9338       message if used.
9339     * GCC 3.3.1 automatically places zero-initialized variables in the
9340       .bss section on some operating systems. Versions of GNU Emacs up to
9341       (and including) 21.3 will not work correctly when using this
9342       optimization; you can use -fno-zero-initialized-in-bss to disable
9343       it.
9344
9345General Optimizer Improvements
9346
9347     * A new scheme for accurately describing processor pipelines, the
9348       [5]DFA scheduler, has been added.
9349     * Pavel Nejedly, Charles University Prague, has contributed new file
9350       format used by the edge coverage profiler (-fprofile-arcs).
9351       The new format is robust and diagnoses common mistakes where
9352       profiles from different versions (or compilations) of the program
9353       are combined resulting in nonsensical profiles and slow code to
9354       produced with profile feedback. Additionally this format allows
9355       extra data to be gathered. Currently, overall statistics are
9356       produced helping optimizers to identify hot spots of a program
9357       globally replacing the old intra-procedural scheme and resulting in
9358       better code. Note that the gcov tool from older GCC versions will
9359       not be able to parse the profiles generated by GCC 3.3 and vice
9360       versa.
9361     * Jan Hubicka, SuSE Labs, has contributed a new superblock formation
9362       pass enabled using -ftracer. This pass simplifies the control flow
9363       of functions allowing other optimizations to do better job.
9364       He also contributed the function reordering pass
9365       (-freorder-functions) to optimize function placement using profile
9366       feedback.
9367
9368New Languages and Language specific improvements
9369
9370  C/ObjC/C++
9371
9372     * The preprocessor now accepts directives within macro arguments. It
9373       processes them just as if they had not been within macro arguments.
9374     * The separate ISO and traditional preprocessors have been completely
9375       removed. The front end handles either type of preprocessed output
9376       if necessary.
9377     * In C99 mode preprocessor arithmetic is done in the precision of the
9378       target's intmax_t, as required by that standard.
9379     * The preprocessor can now copy comments inside macros to the output
9380       file when the macro is expanded. This feature, enabled using the
9381       -CC option, is intended for use by applications which place
9382       metadata or directives inside comments, such as lint.
9383     * The method of constructing the list of directories to be searched
9384       for header files has been revised. If a directory named by a -I
9385       option is a standard system include directory, the option is
9386       ignored to ensure that the default search order for system
9387       directories and the special treatment of system header files are
9388       not defeated.
9389     * A few more [6]ISO C99 features now work correctly.
9390     * A new function attribute, nonnull, has been added which allows
9391       pointer arguments to functions to be specified as requiring a
9392       non-null value. The compiler currently uses this information to
9393       issue a warning when it detects a null value passed in such an
9394       argument slot.
9395     * A new type attribute, may_alias, has been added. Accesses to
9396       objects with types with this attribute are not subjected to
9397       type-based alias analysis, but are instead assumed to be able to
9398       alias any other type of objects, just like the char type.
9399
9400  C++
9401
9402     * Type based alias analysis has been implemented for C++ aggregate
9403       types.
9404
9405  Objective-C
9406
9407     * Generate an error if Objective-C objects are passed by value in
9408       function and method calls.
9409     * When -Wselector is used, check the whole list of selectors at the
9410       end of compilation, and emit a warning if a @selector() is not
9411       known.
9412     * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
9413     * No longer need to include objc/objc-class.h to compile self calls
9414       in class methods (NeXT runtime only).
9415     * New -Wundeclared-selector option.
9416     * Removed selector bloating which was causing object files to be 10%
9417       bigger on average (GNU runtime only).
9418     * Using at run time @protocol() objects has been fixed in certain
9419       situations (GNU runtime only).
9420     * Type checking has been fixed and improved in many situations
9421       involving protocols.
9422
9423  Java
9424
9425     * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK
9426       1.4) API.
9427     * The JDK 1.4 assert facility has been implemented.
9428     * The bytecode interpreter is now direct threaded and thus faster.
9429
9430  Fortran
9431
9432     * Fortran improvements are listed in [7]the Fortran documentation.
9433
9434  Ada
9435
9436     * Ada tasking now works with glibc 2.3.x threading libraries.
9437
9438New Targets and Target Specific Improvements
9439
9440     * The following changes have been made to the HP-PA port:
9441          + The port now defaults to scheduling for the PA8000 series of
9442            processors.
9443          + Scheduling support for the PA7300 processor has been added.
9444          + The 32-bit port now supports weak symbols under HP-UX 11.
9445          + The handling of initializers and finalizers has been improved
9446            under HP-UX 11. The 64-bit port no longer uses collect2.
9447          + Dwarf2 EH support has been added to the 32-bit GNU/Linux port.
9448          + ABI fixes to correct the passing of small structures by value.
9449     * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to
9450       use the DFA processor pipeline description.
9451     * The following NetBSD configurations for the SuperH processor family
9452       have been added:
9453          + SH3, big-endian, sh-*-netbsdelf*
9454          + SH3, little-endian, shle-*-netbsdelf*
9455          + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
9456          + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
9457          + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
9458          + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
9459     * The following changes have been made to the IA-32/x86-64 port:
9460          + SSE2 and 3dNOW! intrinsics are now supported.
9461          + Support for thread local storage has been added to the IA-32
9462            and x86-64 ports.
9463          + The x86-64 port has been significantly improved.
9464     * The following changes have been made to the MIPS port:
9465          + All configurations now accept the -mabi switch. Note that you
9466            will need appropriate multilibs for this option to work
9467            properly.
9468          + ELF configurations will always pass an ABI flag to the
9469            assembler, except when the MIPS EABI is selected.
9470          + -mabi=64 no longer selects MIPS IV code.
9471          + The -mcpu option, which was deprecated in 3.1 and 3.2, has
9472            been removed from this release.
9473          + -march now changes the core ISA level. In previous releases,
9474            it would change the use of processor-specific extensions, but
9475            would leave the core ISA unchanged. For example, mips64-elf
9476            -march=r8000 will now generate MIPS IV code.
9477          + Under most configurations, -mipsN now acts as a synonym for
9478            -march.
9479          + There are some new preprocessor macros to describe the -march
9480            and -mtune settings. See the documentation of those options
9481            for details.
9482          + Support for the NEC VR-Series processors has been added. This
9483            includes the 54xx, 5500, and 41xx series.
9484          + Support for the Sandcraft sr71k processor has been added.
9485     * The following changes have been made to the S/390 port:
9486          + Support to build the Java runtime libraries has been added.
9487            Java is now enabled by default on s390-*-linux* and
9488            s390x-*-linux* targets.
9489          + Multilib support for the s390x-*-linux* target has been added;
9490            this allows to build 31-bit binaries using the -m31 option.
9491          + Support for thread local storage has been added.
9492          + Inline assembler code may now use the 'Q' constraint to
9493            specify memory operands without index register.
9494          + Various platform-specific performance improvements have been
9495            implemented; in particular, the compiler now uses the BRANCH
9496            ON COUNT family of instructions and makes more frequent use of
9497            the TEST UNDER MASK family of instructions.
9498     * The following changes have been made to the PowerPC port:
9499          + Support for IBM Power4 processor added.
9500          + Support for Motorola e500 SPE added.
9501          + Support for AIX 5.2 added.
9502          + Function and Data sections now supported on AIX.
9503          + Sibcall optimizations added.
9504     * The support for H8 Tiny is added to the H8/300 port with -mn.
9505
9506Obsolete Systems
9507
9508   Support for a number of older systems has been declared obsolete in GCC
9509   3.3. Unless there is activity to revive them, the next release of GCC
9510   will have their sources permanently removed.
9511
9512   All configurations of the following processor architectures have been
9513   declared obsolete:
9514     * Matsushita MN10200, mn10200-*-*
9515     * Motorola 88000, m88k-*-*
9516     * IBM ROMP, romp-*-*
9517
9518   Also, some individual systems have been obsoleted:
9519     * Alpha
9520          + Interix, alpha*-*-interix*
9521          + Linux libc1, alpha*-*-linux*libc1*
9522          + Linux ECOFF, alpha*-*-linux*ecoff*
9523     * ARM
9524          + Generic a.out, arm*-*-aout*
9525          + Conix, arm*-*-conix*
9526          + "Old ABI," arm*-*-oabi
9527          + StrongARM/COFF, strongarm-*-coff*
9528     * HPPA (PA-RISC)
9529          + Generic OSF, hppa1.0-*-osf*
9530          + Generic BSD, hppa1.0-*-bsd*
9531          + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
9532          + HiUX, hppa*-*-hiux*
9533          + Mach Lites, hppa*-*-lites*
9534     * Intel 386 family
9535          + Windows NT 3.x, i?86-*-win32
9536     * MC68000 family
9537          + HP systems, m68000-hp-bsd* and m68k-hp-bsd*
9538          + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and
9539            m68k-sun-mach*
9540          + AT&T systems, m68000-att-sysv*
9541          + Atari systems, m68k-atari-sysv*
9542          + Motorola systems, m68k-motorola-sysv*
9543          + NCR systems, m68k-ncr-sysv*
9544          + Plexus systems, m68k-plexus-sysv*
9545          + Commodore systems, m68k-cbm-sysv*
9546          + Citicorp TTI, m68k-tti-*
9547          + Unos, m68k-crds-unos*
9548          + Concurrent RTU, m68k-ccur-rtu*
9549          + Linux a.out, m68k-*-linux*aout*
9550          + Linux libc1, m68k-*-linux*libc1*
9551          + pSOS, m68k-*-psos*
9552     * MIPS
9553          + Generic ECOFF, mips*-*-ecoff*
9554          + SINIX, mips-sni-sysv4
9555          + Orion RTEMS, mips64orion-*-rtems*
9556     * National Semiconductor 32000
9557          + OpenBSD, ns32k-*-openbsd*
9558     * POWER (aka RS/6000) and PowerPC
9559          + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
9560          + Bull BOSX, rs6000-bull-bosx
9561          + Generic Mach, rs6000-*-mach*
9562          + Generic SysV, powerpc*-*-sysv*
9563          + Linux libc1, powerpc*-*-linux*libc1*
9564     * Sun SPARC
9565          + Generic a.out, sparc-*-aout*, sparclet-*-aout*,
9566            sparclite-*-aout*, and sparc86x-*-aout*
9567          + NetBSD a.out, sparc-*-netbsd*aout*
9568          + Generic BSD, sparc-*-bsd*
9569          + ChorusOS, sparc-*-chorusos*
9570          + Linux a.out, sparc-*-linux*aout*
9571          + Linux libc1, sparc-*-linux*libc1*
9572          + LynxOS, sparc-*-lynxos*
9573          + Solaris on HAL hardware, sparc-hal-solaris2*
9574          + SunOS versions 3 and 4, sparc-*-sunos[34]*
9575     * NEC V850
9576          + RTEMS, v850-*-rtems*
9577     * VAX
9578          + VMS, vax-*-vms*
9579
9580Documentation improvements
9581
9582Other significant improvements
9583
9584     * Almost all front-end dependencies in the compiler have been
9585       separated out into a set of language hooks. This should make adding
9586       a new front end clearer and easier.
9587     * One effect of removing the separate preprocessor is a small
9588       increase in the robustness of the compiler in general, and the
9589       maintainability of target descriptions. Previously target-specific
9590       built-in macros and others, such as __FAST_MATH__, had to be
9591       handled with so-called specs that were hard to maintain. Often they
9592       would fail to behave properly when conflicting options were
9593       supplied on the command line, and define macros in the user's
9594       namespace even when strict ISO compliance was requested.
9595       Integrating the preprocessor has cleanly solved these issues.
9596     * The Makefile suite now supports redirection of make install by
9597       means of the variable DESTDIR.
9598     __________________________________________________________________
9599
9600GCC 3.3
9601
9602   Detailed release notes for the GCC 3.3 release follow.
9603
9604  Bug Fixes
9605
9606    bootstrap failures
9607
9608     * [8]10140 cross compiler build failures: missing __mempcpy (DUP:
9609       [9]10198,[10]10338)
9610
9611    Internal compiler errors (multi-platform)
9612
9613     * [11]3581 large string causes segmentation fault in cc1
9614     * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler
9615     * [13]5533 (c++) ICE when processing std::accumulate(begin, end,
9616       init, invalid_op)
9617     * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
9618     * [15]6412 (c++) ICE in retrieve_specialization
9619     * [16]6620 (c++) partial template specialization causes an ICE
9620       (segmentation fault)
9621     * [17]6663 (c++) ICE with attribute aligned
9622     * [18]7068 ICE with incomplete types
9623     * [19]7083 (c++) ICE using -gstabs with dodgy class derivation
9624     * [20]7647 (c++) ICE when data member has the name of the enclosing
9625       class
9626     * [21]7675 ICE in fixup_var_refs_1
9627     * [22]7718 'complex' template instantiation causes ICE
9628     * [23]8116 (c++) ICE in member template function
9629     * [24]8358 (ada) Ada compiler accesses freed memory, crashes
9630     * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation
9631       fault
9632     * [26]8564 (c++) ICE in find_function_data, in function.c
9633     * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
9634     * [28]8766 (c++) ICE after failed initialization of static template
9635       variable
9636     * [29]8803 ICE in instantiate_virtual_regs_1, in function.c
9637     * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
9638     * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class
9639       definition
9640     * [32]9216 (c++) ICE on missing template parameter
9641     * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c
9642     * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO
9643       loop
9644     * [35]9429 (c++) ICE in template instantiation with a pointered new
9645       operator
9646     * [36]9516 Internal error when using a big array
9647     * [37]9600 (c++) ICE with typedefs in template class
9648     * [38]9629 (c++) virtual inheritance segfault
9649     * [39]9672 (c++) ICE: Error reporting routines re-entered
9650     * [40]9749 (c++) ICE in write_expression on invalid function
9651       prototype
9652     * [41]9794 (fortran) ICE: floating point exception during constant
9653       folding
9654     * [42]9829 (c++) Missing colon in nested namespace usage causes ICE
9655     * [43]9916 (c++) ICE with noreturn function in ?: statement
9656     * [44]9936 ICE with local function and variable-length 2d array
9657     * [45]10262 (c++) cc1plus crashes with large generated code
9658     * [46]10278 (c++) ICE in parser for invalid code
9659     * [47]10446 (c++) ICE on definition of nonexistent member function of
9660       nested class in a class template
9661     * [48]10451 (c++) ICE in grokdeclarator on spurious mutable
9662       declaration
9663     * [49]10506 (c++) ICE in build_new at cp/init.c with
9664       -fkeep-inline-functions and multiple inheritance
9665     * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the
9666       precision of the declared type
9667
9668    Optimization bugs
9669
9670     * [51]2001 Inordinately long compile times in reload CSE regs
9671     * [52]2391 Exponential compilation time explosion in combine
9672     * [53]2960 Duplicate loop conditions even with -Os
9673     * [54]4046 redundant conditional branch
9674     * [55]6405 Loop-unrolling related performance regressions
9675     * [56]6798 very long compile time with large case-statement
9676     * [57]6871 const objects shouldn't be moved to .bss
9677     * [58]6909 problem w/ -Os on modified loop-2c.c test case
9678     * [59]7189 gcc -O2 -Wall does not print ``control reaches end of
9679       non-void function'' warning
9680     * [60]7642 optimization problem with signbit()
9681     * [61]8634 incorrect code for inlining of memcpy under -O2
9682     * [62]8750 Cygwin prolog generation erroneously emitting __alloca as
9683       regular function call
9684
9685    C front end
9686
9687     * [63]2161 long if-else cascade overflows parser stack
9688     * [64]4319 short accepted on typedef'd char
9689     * [65]8602 incorrect line numbers in warning messages when using
9690       inline functions
9691     * [66]9177 -fdump-translation-unit: C front end deletes function_decl
9692       AST nodes and breaks debugging dumps
9693     * [67]9853 miscompilation of non-constant structure initializer
9694
9695    c++ compiler and library
9696
9697     * [68]45 legal template specialization code is rejected (DUP:
9698       [69]3784)
9699     * [70]764 lookup failure: friend operator and dereferencing a pointer
9700       and templates (DUP: [71]5116)
9701     * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP:
9702       2863)
9703     * [73]3663 G++ doesn't check access control during template
9704       instantiation
9705     * [74]3797 gcc fails to emit explicit specialization of a template
9706       member
9707     * [75]3948 Two destructors are called when no copy destructor is
9708       defined (ABI change)
9709     * [76]4137 Conversion operator within template is not accepted
9710     * [77]4361 bogus ambiguity taking the address of a member template
9711     * [78]4802 g++ accepts illegal template code (access to private
9712       member; DUP: [79]5837)
9713     * [80]4803 inline function is used but never defined, and g++ does
9714       not object
9715     * [81]5094 Partial specialization cannot be friend?
9716     * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
9717     * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run
9718       time
9719     * [84]7015 certain __asm__ constructs rejected
9720     * [85]7086 compile time regression (quadratic behavior in
9721       fixup_var_refs)
9722     * [86]7099 G++ doesn't set the noreturn attribute on std::exit and
9723       std::abort
9724     * [87]7247 copy constructor missing when inlining enabled (invalid
9725       optimization?)
9726     * [88]7441 string array initialization compilation time regression
9727       from seconds to minutes
9728     * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong
9729     * [90]7804 bad printing of floating point constant in warning message
9730     * [91]8099 Friend classes and template specializations
9731     * [92]8117 member function pointers and multiple inheritance
9732     * [93]8205 using declaration and multiple inheritance
9733     * [94]8645 unnecessary non-zero checks in stl_tree.h
9734     * [95]8724 explicit destructor call for incomplete class allowed
9735     * [96]8805 compile time regression with many member variables
9736     * [97]8691 -O3 and -fno-implicit-templates are incompatible
9737     * [98]8700 unhelpful error message for binding temp to reference
9738     * [99]8724 explicit destructor call for incomplete class allowed
9739     * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems
9740     * [101]9016 Failure to consistently constant fold "constant" C++
9741       objects
9742     * [102]9053 g++ confused about ambiguity of overloaded function
9743       templates
9744     * [103]9152 undefined virtual thunks
9745     * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out
9746     * [105]9297 data corruption due to codegen bug (when copying.)
9747     * [106]9318 i/ostream::operator>>/<<(streambuf*) broken
9748     * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf
9749     * [108]9400 bogus -Wshadow warning: shadowed declaration of this in
9750       local classes
9751     * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters
9752     * [110]9425 filebuf::pbackfail broken (DUP: [111]9439)
9753     * [112]9474 GCC freezes in compiling a weird code mixing <iostream>
9754       and <iostream.h>
9755     * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1)
9756       [114][DR 231]
9757     * [115]9555 ostream inserters fail to set badbit on exception
9758     * [116]9561 ostream inserters rethrow exception of wrong type
9759     * [117]9563 ostream::sentry returns true after a failed preparation
9760     * [118]9582 one-definition rule violation in std::allocator
9761     * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors
9762     * [120]9683 bug in initialization chains for static const variables
9763       from template classes
9764     * [121]9791 -Woverloaded-virtual reports hiding of destructor
9765     * [122]9817 collate::compare doesn't handle nul characters
9766     * [123]9825 filebuf::sputbackc breaks sbumpc
9767     * [124]9826 operator>>(basic_istream, basic_string) fails to compile
9768       with custom traits
9769     * [125]9924 Multiple using statements for builtin functions not
9770       allowed
9771     * [126]9946 destructor is not called for temporary object
9772     * [127]9964 filebuf::close() sometimes fails to close file
9773     * [128]9988 filebuf::overflow writes EOF to file
9774     * [129]10033 optimization breaks polymorphic references w/ typeid
9775       operator
9776     * [130]10097 filebuf::underflow drops characters
9777     * [131]10132 filebuf destructor can throw exceptions
9778     * [132]10180 gcc fails to warn about non-inlined function
9779     * [133]10199 method parametrized by template does not work everywhere
9780     * [134]10300 use of array-new (nothrow) in segfaults on NULL return
9781     * [135]10427 Stack corruption with variable-length automatic arrays
9782       and virtual destructors
9783     * [136]10503 Compilation never stops in fixed_type_or_null
9784
9785    Objective-C
9786
9787     * [137]5956 selectors aren't matched properly when added to the
9788       selector table
9789
9790    Fortran compiler and library
9791
9792     * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't
9793       detect
9794     * [139]3924 g77 generates code that is rejected by GAS if COFF debug
9795       info requested
9796     * [140]5634 doc: explain that configure --prefix=~/... does not work
9797     * [141]6367 multiple repeat counts confuse namelist read into array
9798     * [142]6491 Logical operations error on logicals when using
9799       -fugly-logint
9800     * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C"
9801     * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os
9802       on irix6.5
9803     * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should
9804       assume a direct access file
9805     * [146]7278 g77 "bug"; the executable misbehaves (with -O2
9806       -fno-automatic)
9807     * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows
9808     * [148]7388 Incorrect output with 0-based array of characters
9809     * [149]8587 Double complex zero ** double precision number -> NaN
9810       instead of zero
9811     * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning:
9812       unknown register name line-length-none
9813     * [151]10197 Direct access files not unformatted by default
9814
9815    Java compiler and library
9816
9817     * [152]6005 gcj fails to build rhug on alpha
9818     * [153]6389 System.getProperty("") should always throw an
9819       IllegalArgumentException
9820     * [154]6576 java.util.ResourceBundle.getResource ignores locale
9821     * [155]6652 new java.io.File("").getCanonicalFile() throws exception
9822     * [156]7060 getMethod() doesn't search super interface
9823     * [157]7073 bytecode interpreter gives wrong answer for interface
9824       getSuperclass()
9825     * [158]7180 possible bug in
9826       javax.naming.spi.NamingManager.getPlusPath()
9827     * [159]7416 java.security startup refs "GNU libgcj.security"
9828     * [160]7570 Runtime.exec with null envp: child doesn't inherit parent
9829       env (DUP: [161]7578)
9830     * [162]7611 Internal error while compiling libjava with -O
9831     * [163]7709 NullPointerException in _Jv_ResolvePoolEntry
9832     * [164]7766 ZipInputStream.available returns 0 immediately after
9833       construction
9834     * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public
9835     * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
9836     * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens'
9837     * [168]8234 ZipInputStream chokes when InputStream.read() returns
9838       small chunks
9839     * [169]8415 reflection bug: exception info for Method
9840     * [170]8481 java.Random.nextInt(int) may return negative
9841     * [171]8593 Error reading GZIPped files with BufferedReader
9842     * [172]8759 java.beans.Introspector has no flushCaches() or
9843       flushFromCaches() methods
9844     * [173]8997 spin() calls Thread.sleep
9845     * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd
9846       instead of the root content of C:
9847     * [175]9254 java::lang::Object::wait(), threads-win32.cc returns
9848       wrong return codes
9849     * [176]9271 Severe bias in java.security.SecureRandom
9850
9851    Ada compiler and library
9852
9853     * [177]6767 make gnatlib-shared fails on -laddr2line
9854     * [178]9911 gnatmake fails to link when GCC configured with
9855       --with-sjlj-exceptions=yes
9856     * [179]10020 Can't bootstrap gcc on AIX with Ada enabled
9857     * [180]10546 Ada tasking not working on Red Hat 9
9858
9859    preprocessor
9860
9861     * [181]7029 preprocessor should ignore #warning with -M
9862
9863    ARM-specific
9864
9865     * [182]2903 [arm] Optimization bug with long long arithmetic
9866     * [183]7873 arm-linux-gcc fails when assigning address to a bit field
9867
9868    FreeBSD-specific
9869
9870     * [184]7680 float functions undefined in math.h/cmath with #define
9871       _XOPEN_SOURCE
9872
9873    HP-UX or HP-PA-specific
9874
9875     * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
9876     * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to
9877       fputc_unlocked
9878     * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen
9879
9880    m68hc11-specific
9881
9882     * [188]6744 Bad assembler code generated: reference to pseudo
9883       register z
9884     * [189]7361 Internal compiler error in reload_cse_simplify_operands,
9885       in reload1.c
9886
9887    MIPS-specific
9888
9889     * [190]9496 [mips-linux] bug in optimizer?
9890
9891    PowerPC-specific
9892
9893     * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of
9894       space
9895     * [192]8480 reload ICEs for LAPACK code on powerpc64-linux
9896     * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg
9897     * [194]10315 [powerpc] ICE: in extract_insn, in recog.c
9898
9899    SPARC-specific
9900
9901     * [195]10267 (documentation) Wrong build instructions for
9902       *-*-solaris2*
9903
9904    x86-specific (Intel/AMD)
9905
9906     * [196]7916 ICE in instantiate_virtual_register_1
9907     * [197]7926 (c++) i486 instructions in header files make c++ programs
9908       crash on i386
9909     * [198]8555 ICE in gen_split_1231
9910     * [199]8994 ICE with -O -march=pentium4
9911     * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs
9912     * [201]9806 ICE in inline assembly with -fPIC flag
9913     * [202]10077 gcc -msse2 generates movd to move dwords between xmm
9914       regs
9915     * [203]10233 64-bit comparison only comparing bottom 32-bits
9916     * [204]10286 type-punning doesn't work with __m64 and -O
9917     * [205]10308 [x86] ICE with -O -fgcse or -O2
9918     __________________________________________________________________
9919
9920GCC 3.3.1
9921
9922  Bug Fixes
9923
9924   This section lists the problem reports (PRs) from GCC's bug tracking
9925   system that are known to be fixed in the 3.3.1 release. This list might
9926   not be complete (that is, it is possible that some PRs that have been
9927   fixed are not listed here).
9928
9929    Bootstrap failures
9930
9931     * [206]11272 [Solaris] make bootstrap fails while building libstdc++
9932
9933    Internal compiler errors (multi-platform)
9934
9935     * [207]5754 ICE on invalid nested template class
9936     * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64
9937       and --enable-checking
9938     * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c
9939     * [210]7053 (c++) ICE when declaring a function already defined as a
9940       friend method of a template class
9941     * [211]8164 (c++) ICE when using different const expressions as
9942       template parameter
9943     * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c
9944     * [213]9559 (c++) ICE with invalid initialization of a static const
9945     * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c
9946       when redeclaring a static member variable
9947     * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in
9948       dwarfout.c with -g -O -finline-functions
9949     * [216]10432 (c++) ICE in poplevel, in cp/decl.c
9950     * [217]10475 ICE in subreg_highpart_offset for code with long long
9951     * [218]10635 (c++) ICE when dereferencing an incomplete type casted
9952       from a void pointer
9953     * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while
9954       instantiating static member variables
9955     * [220]10700 ICE in copy_to_mode_reg on 64-bit targets
9956     * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c
9957     * [222]10796 (c++) ICE when defining an enum with two values: -1 and
9958       MAX_INT_64BIT
9959     * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x
9960       sched.c
9961     * [224]10939 (c++) ICE with template code
9962     * [225]10956 (c++) ICE when specializing a template member function
9963       of a template class, in tsubst, in cp/pt.c
9964     * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*()
9965       defined)
9966     * [227]11059 (c++) ICE with empty union
9967     * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with
9968       -O2 -fnon-call-exceptions
9969     * [229]11105 (c++) ICE in mangle_conv_op_name_for_type
9970     * [230]11149 (c++) ICE on error when instantiation with call function
9971       of a base type
9972     * [231]11228 (c++) ICE on new-expression using array operator new and
9973       default-initialization
9974     * [232]11282 (c++) Infinite memory usage after syntax error
9975     * [233]11301 (fortran) ICE with -fno-globals
9976     * [234]11308 (c++) ICE when using an enum type name as if it were a
9977       class or namespace
9978     * [235]11473 (c++) ICE with -gstabs when empty struct inherits from
9979       an empty struct
9980     * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR
9981     * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c:
9982       template member functions
9983
9984    Optimization bugs
9985
9986     * [238]11198 -O2 -frename-registers generates wrong code (aliasing
9987       problem)
9988     * [239]11304 Wrong code production with -fomit-frame-pointer
9989     * [240]11381 volatile memory access optimized away
9990     * [241]11536 [strength-reduce] -O2 optimization produces wrong code
9991     * [242]11557 constant folding bug generates wrong code
9992
9993    C front end
9994
9995     * [243]5897 No warning for statement after return
9996     * [244]11279 DWARF-2 output mishandles large enums
9997
9998    Preprocessor bugs
9999
10000     * [245]11022 no warning for non-compatible macro redefinition
10001
10002    C++ compiler and library
10003
10004     * [246]2330 static_cast<>() to a private base is allowed
10005     * [247]5388 Incorrect message "operands to ?: have different types"
10006     * [248]5390 Libiberty fails to demangle multi-digit template
10007       parameters
10008     * [249]7877 Incorrect parameter passing to specializations of member
10009       function templates
10010     * [250]9393 Anonymous namespaces and compiling the same file twice
10011     * [251]10032 -pedantic converts some errors to warnings
10012     * [252]10468 const typeof(x) is non-const, but only in templates
10013     * [253]10527 confused error message with "new int()" parameter
10014       initializer
10015     * [254]10679 parameter MIN_INLINE_INSNS is not honored
10016     * [255]10682 gcc chokes on a typedef for an enum inside a class
10017       template
10018     * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of
10019       0.
10020     * [257]10845 template member function (with nested template as
10021       parameter) cannot be called anymore if another unrelated template
10022       member function is defined
10023     * [258]10849 Cannot define an out-of-class specialization of a
10024       private nested template class
10025     * [259]10888 Suppress -Winline warnings for system headers
10026     * [260]10929 -Winline warns about functions for which no definition
10027       is visible
10028     * [261]10931 valid conversion static_cast<const unsigned
10029       int&>(lvalue-of-type-int) is rejected
10030     * [262]10940 Bad code with explicit specialization
10031     * [263]10968 If member function implicitly instantiated, explicit
10032       instantiation of class fails to instantiate it
10033     * [264]10990 Cannot convert with dynamic_cast<> to a private base
10034       class from within a member function
10035     * [265]11039 Bad interaction between implicit typename deprecation
10036       and friendship
10037     * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say
10038       "__unused__" instead
10039     * [267]11095 C++ iostream manipulator causes segfault when called
10040       with negative argument
10041     * [268]11098 g++ doesn't emit complete debugging information for
10042       local variables in destructors
10043     * [269]11137 GNU/Linux shared library constructors not called unless
10044       there's one global object
10045     * [270]11154 spurious ambiguity report for template class
10046       specialization
10047     * [271]11329 Compiler cannot find user defined implicit typecast
10048     * [272]11332 Spurious error with casts in ?: expression
10049     * [273]11431 static_cast behavior with subclasses when default
10050       constructor available
10051     * [274]11528 money_get facet does not accept "$.00" as valid
10052     * [275]11546 Type lookup problems in out-of-line definition of a
10053       class doubly nested from a template class
10054     * [276]11567 C++ code containing templated member function with same
10055       name as pure virtual member function results in linking failure
10056     * [277]11645 Failure to deal with using and private inheritance
10057
10058    Java compiler and library
10059
10060     * [278]5179 Qualified static field access doesn't initialize its
10061       class
10062     * [279]8204 gcj -O2 to native reorders certain instructions
10063       improperly
10064     * [280]10838 java.io.ObjectInputStream syntax error
10065     * [281]10886 The RMI registry that comes with GCJ does not work
10066       correctly
10067     * [282]11349 JNDI URL context factories not located correctly
10068
10069    x86-specific (Intel/AMD)
10070
10071     * [283]4823 ICE on inline assembly code
10072     * [284]8878 miscompilation with -O and SSE
10073     * [285]9815 (c++ library) atomicity.h - fails to compile with -O3
10074       -masm=intel
10075     * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads,
10076       in reload1.c
10077     * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2
10078     * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source
10079     * [289]11044 [x86] out of range loop instructions for FP code on K6
10080     * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE
10081       built-ins
10082     * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC"
10083       is used
10084
10085    SPARC- or Solaris- specific
10086
10087     * [292]9362 solaris 'as' dies when fed .s and "-gstabs"
10088     * [293]10142 [SPARC64] gcc produces wrong code when passing
10089       structures by value
10090     * [294]10663 New configure check aborts with Sun tools.
10091     * [295]10835 combinatorial explosion in scheduler on HyperSPARC
10092     * [296]10876 ICE in calculate_giv_inc when building KDE
10093     * [297]10955 wrong code at -O3 for structure argument in context of
10094       structure return
10095     * [298]11018 -mcpu=ultrasparc busts tar-1.13.25
10096     * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x
10097       Linux kernel
10098
10099    ia64 specific
10100
10101     * [300]10907 gcc violates the ia64 ABI (GP must be preserved)
10102     * [301]11320 scheduler bug (in machine depended reorganization pass)
10103     * [302]11599 bug with conditional and __builtin_prefetch
10104
10105    PowerPC specific
10106
10107     * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem
10108       during loop)
10109     * [304]10871 error in rs6000_stack_info save_size computation
10110     * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse
10111       cures it
10112
10113    m68k-specific
10114
10115     * [306]7594 [m68k] ICE on legal code associated with simplify-rtx
10116     * [307]10557 [m68k] ICE in subreg_offset_representable_p
10117     * [308]11054 [m68k] ICE in reg_overlap_mentioned_p
10118
10119    ARM-specific
10120
10121     * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for
10122       functions with __attribute__ ((interrupt ("IRQ")))
10123     * [310]10842 [arm] Clobbered link register is copied to pc under
10124       certain circumstances
10125     * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes
10126     * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno
10127       (3.4)
10128
10129    MIPS-specific
10130
10131     * [313]11084 ICE in propagate_one_insn, in flow.c
10132
10133    SH-specific
10134
10135     * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf
10136     * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c
10137     * [316]11096 i686-linux to sh-linux cross compiler fails to compile
10138       C++ files
10139
10140    GNU/Linux (or Hurd?) specific
10141
10142     * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3
10143
10144    UnixWare specific
10145
10146     * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare
10147       7.1.1
10148
10149    Cygwin (or mingw) specific
10150
10151     * [319]5287 ICE with dllimport attribute
10152     * [320]10148 [MingW/CygWin] Compiler dumps core
10153
10154    DJGPP specific
10155
10156     * [321]8787 GCC fails to emit .intel_syntax when invoked with
10157       -masm=intel on DJGPP
10158
10159    Darwin (and MacOS X) specific
10160
10161     * [322]10900 trampolines crash
10162
10163    Documentation
10164
10165     * [323]1607 (c++) Format attributes on methods undocumented
10166     * [324]4252 Invalid option `-fdump-translation-unit'
10167     * [325]4490 Clarify restrictions on -m96bit-long-double,
10168       -m128bit-long-double
10169     * [326]10355 document an issue with regparm attribute on some systems
10170       (e.g. Solaris)
10171     * [327]10726 (fortran) Documentation for function "IDate Intrinsic
10172       (Unix)" is wrong
10173     * [328]10805 document bug in old version of Sun assembler
10174     * [329]10815 warn against GNU binutils on AIX
10175     * [330]10877 document need for newer binutils on i?86-*-linux-gnu
10176     * [331]11280 Manual incorrect with respect to -freorder-blocks
10177     * [332]11466 Document -mlittle-endian and its restrictions for the
10178       sparc64 port
10179
10180    Testsuite bugs (compiler itself is not affected)
10181
10182     * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly
10183       report failure
10184     * [334]10810 gcc-3.3 fails make check: buffer overrun in
10185       test_demangle.c
10186     __________________________________________________________________
10187
10188GCC 3.3.2
10189
10190  Bug Fixes
10191
10192   This section lists the problem reports (PRs) from GCC's bug tracker
10193   that are known to be fixed in the 3.3.2 release. This list might not be
10194   complete (that is, it is possible that some PRs that have been fixed
10195   are not listed here).
10196
10197    Bootstrap failures and problems
10198
10199     * [335]8336 [SCO5] bootstrap config still tries to use COFF options
10200     * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with
10201       --enable-threads=posix
10202     * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap
10203     * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare
10204       7.1.1)
10205     * [339]11687 xstormy16-elf build fails in libf2c
10206     * [340]12263 [SGI IRIX] bootstrap fails during compile of
10207       libf2c/libI77/backspace.c
10208     * [341]12490 buffer overflow in scan-decls.c (during Solaris 9
10209       fix-header processing)
10210
10211    Internal compiler errors (multi-platform)
10212
10213     * [342]7277 Casting integers to vector types causes ICE
10214     * [343]7939 (c++) ICE on invalid function template specialization
10215     * [344]11063 (c++) ICE on parsing initialization list of const array
10216       member
10217     * [345]11207 ICE with negative index in array element designator
10218     * [346]11522 (fortran) g77 dwarf-2 ICE in
10219       add_abstract_origin_attribute
10220     * [347]11595 (c++) ICE on duplicate label definition
10221     * [348]11646 (c++) ICE in commit_one_edge_insertion with
10222       -fnon-call-exceptions -fgcse -O
10223     * [349]11665 ICE in struct initializer when taking address
10224     * [350]11852 (c++) ICE with bad struct initializer.
10225     * [351]11878 (c++) ICE in cp_expr_size
10226     * [352]11883 ICE with any -O on mercury-generated C code
10227     * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in
10228       cp/typeck2.c when applying typeid operator to template template
10229       parameter
10230     * [354]12146 ICE in lookup_template_function, in cp/pt.c
10231     * [355]12215 ICE in make_label_edge with -fnon-call-exceptions
10232       -fno-gcse -O2
10233     * [356]12369 (c++) ICE with templates and friends
10234     * [357]12446 ICE in emit_move_insn on complicated array reference
10235     * [358]12510 ICE in final_scan_insn
10236     * [359]12544 ICE with large parameters used in nested functions
10237
10238    C and optimization bugs
10239
10240     * [360]9862 spurious warnings with -W -finline-functions
10241     * [361]10962 lookup_field is a linear search on a linked list (can be
10242       slow if large struct)
10243     * [362]11370 -Wunreachable-code gives false complaints
10244     * [363]11637 invalid assembly with -fnon-call-exceptions
10245     * [364]11885 Problem with bitfields in packed structs
10246     * [365]12082 Inappropriate unreachable code warnings
10247     * [366]12180 Inline optimization fails for variadic function
10248     * [367]12340 loop unroller + gcse produces wrong code
10249
10250    C++ compiler and library
10251
10252     * [368]3907 nested template parameter collides with member name
10253     * [369]5293 confusing message when binding a temporary to a reference
10254     * [370]5296 [DR115] Pointers to functions and to template functions
10255       behave differently in deduction
10256     * [371]7939 ICE on function template specialization
10257     * [372]8656 Unable to assign function with __attribute__ and pointer
10258       return type to an appropriate variable
10259     * [373]10147 Confusing error message for invalid template function
10260       argument
10261     * [374]11400 std::search_n() makes assumptions about Size parameter
10262     * [375]11409 issues with using declarations, overloading, and
10263       built-in functions
10264     * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle
10265       multiple bits in mask
10266     * [377]11786 operator() call on variable in other namespace not
10267       recognized
10268     * [378]11867 static_cast ignores ambiguity
10269     * [379]11928 bug with conversion operators that are typedefs
10270     * [380]12114 Uninitialized memory accessed in dtor
10271     * [381]12163 static_cast + explicit constructor regression
10272     * [382]12181 Wrong code with comma operator and c++
10273     * [383]12236 regparm and fastcall messes up parameters
10274     * [384]12266 incorrect instantiation of unneeded template during
10275       overload resolution
10276     * [385]12296 istream::peek() doesn't set eofbit
10277     * [386]12298 [sjlj exceptions] Stack unwind destroys
10278       not-yet-constructed object
10279     * [387]12369 ICE with templates and friends
10280     * [388]12337 apparently infinite loop in g++
10281     * [389]12344 stdcall attribute ignored if function returns a pointer
10282     * [390]12451 missing(late) class forward declaration in cxxabi.h
10283     * [391]12486 g++ accepts invalid use of a qualified name
10284
10285    x86 specific (Intel/AMD)
10286
10287     * [392]8869 [x86 MMX] ICE with const variable optimization and MMX
10288       builtins
10289     * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions
10290       -O2
10291     * [394]11689 g++3.3 emits un-assembleable code for k6 architecture
10292     * [395]12116 [k6] Invalid assembly output values with X-MAME code
10293     * [396]12070 ICE converting between double and long double with
10294       -msoft-float
10295
10296    ia64-specific
10297
10298     * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc
10299     * [398]11535 __builtin_return_address may not work on ia64
10300     * [399]11693 [ia64] ICE in gen_nop_type
10301     * [400]12224 [ia64] Thread-local storage doesn't work
10302
10303    PowerPC-specific
10304
10305     * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux
10306       kernel
10307     * [402]11319 loop miscompiled on ppc32
10308     * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code
10309
10310    SPARC-specific
10311
10312     * [404]11662 wrong code for expr. with cast to long long and
10313       exclusive or
10314     * [405]11965 invalid assembler code for a shift < 32 operation
10315     * [406]12301 (c++) stack corruption when a returned expression throws
10316       an exception
10317
10318    Alpha-specific
10319
10320     * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of
10321       kernel 2.4.22-pre8
10322
10323    HPUX-specific
10324
10325     * [408]11313 problem with #pragma weak and static inline functions
10326     * [409]11712 __STDC_EXT__ not defined for C++ by default anymore?
10327
10328    Solaris specific
10329
10330     * [410]12166 Profiled programs crash if PROFDIR is set
10331
10332    Solaris-x86 specific
10333
10334     * [411]12101 i386 Solaris no longer works with GNU as?
10335
10336    Miscellaneous embedded target-specific bugs
10337
10338     * [412]10988 [m32r-elf] wrong blockmove code with -O3
10339     * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with
10340       -O2
10341     * [414]11902 [sh4] spec file improperly inserts rpath even when none
10342       needed
10343     * [415]11903 [sh4] -pthread fails to link due to error in spec file
10344       on sh4
10345     __________________________________________________________________
10346
10347GCC 3.3.3
10348
10349  Minor features
10350
10351   In addition to the bug fixes documented below, this release contains
10352   few minor features such as:
10353     * Support for --with-sysroot
10354     * Support for automatic detection of executable stacks
10355     * Support for SSE3 instructions
10356     * Support for thread local storage debugging under GDB on S390
10357
10358  Bug Fixes
10359
10360   This section lists the problem reports (PRs) from GCC's bug tracker
10361   that are known to be fixed in the 3.3.3 release. This list might not be
10362   complete (that is, it is possible that some PRs that have been fixed
10363   are not listed here).
10364
10365    Bootstrap failures and issues
10366
10367     * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails
10368     * [417]12399 boehm-gc fails (when building a cross compiler): libtool
10369       unable to infer tagged configuration
10370     * [418]13068 mklibgcc.in doesn't handle multi-level multilib
10371       subdirectories properly
10372
10373    Internal compiler errors (multi-platform)
10374
10375     * [419]10060 ICE (stack overflow) on huge file (300k lines) due to
10376       recursive behaviour of copy_rtx_if_shared, in emit_rtl.c
10377     * [420]10555 (c++) ICE on undefined template argument
10378     * [421]10706 (c++) ICE in mangle_class_name_for_template
10379     * [422]11496 (fortran) error in flow_loops_find when -funroll-loops
10380       active
10381     * [423]11741 ICE in pre_insert_copy_insn, in gcse.c
10382     * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0
10383     * [425]12632 (fortran) -fbounds-check ICE
10384     * [426]12712 (c++) ICE on short legit C++ code fragment with gcc
10385       3.3.2
10386     * [427]12726 (c++) ICE (segfault) on trivial code
10387     * [428]12890 (c++) ICE on compilation of class with throwing method
10388     * [429]12900 (c++) ICE in rtl_verify_flow_info_1
10389     * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on
10390       correct code with -O2 -fno-force-mem
10391     * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive
10392       template
10393     * [432]13318 ICE: floating point exception in the loop optimizer
10394     * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in
10395       except.c
10396     * [434]13574 (c++) invalid array default initializer in class lets
10397       gcc consume all memory and die
10398     * [435]13475 ICE on SIMD variables with partial value initialization
10399     * [436]13797 (c++) ICE on invalid template parameter
10400     * [437]13824 (java) gcj SEGV with simple .java program
10401
10402    C and optimization bugs
10403
10404     * [438]8776 loop invariants are not removed (most likely)
10405     * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing
10406       strncmp by memcmp
10407     * [440]11350 undefined labels with -Os -fPIC
10408     * [441]12826 Optimizer removes reference through volatile pointer
10409     * [442]12500 stabs debug info: void no longer a predefined / builtin
10410       type
10411     * [443]12941 builtin-bitops-1.c miscompilation (latent bug)
10412     * [444]12953 tree inliner bug (in inline_forbidden_p) and fix
10413     * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled
10414     * [446]13507 spurious printf format warning
10415     * [447]13382 Type information for const pointer disappears during
10416       optimization.
10417     * [448]13394 noreturn attribute ignored on recursive invokation
10418     * [449]13400 Compiled code crashes storing to read-only location
10419     * [450]13521 Endless loop in calculate_global_regs_live
10420
10421    C++ compiler and library
10422
10423   Some of the bug fixes in this list were made to implement decisions
10424   that the ISO C++ standards committee has made concerning several defect
10425   reports (DRs). Links in the list below point to detailed discussion of
10426   the relevant defect report.
10427     * [451]2094 unimplemented: use of `ptrmem_cst' in template type
10428       unification
10429     * [452]2294 using declaration confusion
10430     * [453]5050 template instantiation depth exceeds limit: recursion
10431       problem?
10432     * [454]9371 Bad exception handling in
10433       i/ostream::operator>>/<<(streambuf*)
10434     * [455]9546 bad exception handling in ostream members
10435     * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the
10436       face of unknown locales
10437     * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work
10438     * [459]10095 istream::operator>>(int&) sets ios::badbit when
10439       ios::failbit is set.
10440     * [460]11554 Warning about reordering of initializers doesn't mention
10441       location of constructor
10442     * [461]12297 istream::sentry::sentry() handles eof() incorrectly.
10443     * [462]12352 Exception safety problems in src/localename.cc
10444     * [463]12438 Memory leak in locale::combine()
10445     * [464]12540 Memory leak in locale::locale(const char*)
10446     * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented
10447     * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented
10448     * [470]12696 memory eating infinite loop in diagnostics (error
10449       recovery problem)
10450     * [471]12815 Code compiled with optimization behaves unexpectedly
10451     * [472]12862 Conflicts between typedefs/enums and namespace member
10452       declarations
10453     * [473]12926 Wrong value after assignment in initialize list using
10454       bit-fields
10455     * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented
10456     * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented
10457     * [478]13007 basic_streambuf::pubimbue, imbue wrong
10458     * [479]13009 Implicitly-defined assignment operator writes to wrong
10459       memory
10460     * [480]13057 regparm attribute not applied to destructor
10461     * [481]13070 -Wformat option ignored in g++
10462     * [482]13081 forward template declarations in <complex> let inlining
10463       fail
10464     * [483]13239 Assertion does not seem to work correctly anymore
10465     * [484]13262 "xxx is private within this context" when initializing a
10466       self-contained template class
10467     * [485]13290 simple typo in concept checking for std::generate_n
10468     * [486]13323 Template code does not compile in presence of typedef
10469     * [487]13369 __verify_grouping (and __add_grouping?) not correct
10470     * [488]13371 infinite loop with packed struct and inlining
10471     * [489]13445 Template argument replacement "dereferences" a typedef
10472     * [490]13461 Fails to access protected-ctor from public constant
10473     * [491]13462 Non-standard-conforming type set::pointer
10474     * [492]13478 gcc uses wrong constructor to initialize a const
10475       reference
10476     * [493]13544 "conflicting types" for enums in different scopes
10477     * [494]13650 string::compare should not (always) use
10478       traits_type::length()
10479     * [495]13683 bogus warning about passing non-PODs through ellipsis
10480     * [496]13688 Derived class is denied access to protected base class
10481       member class
10482     * [497]13774 Member variable cleared in virtual multiple inheritance
10483       class
10484     * [498]13884 Protect sstream.tcc from extern template use
10485
10486    Java compiler and library
10487
10488     * [499]10746 [win32] garbage collection crash in GCJ
10489
10490    Objective-C compiler and library
10491
10492     * [500]11433 Crash due to dereferencing null pointer when querying
10493       protocol
10494
10495    Fortran compiler and library
10496
10497     * [501]12633 logical expression gives incorrect result with
10498       -fugly-logint option
10499     * [502]13037 [gcse-lm] g77 generates incorrect code
10500     * [503]13213 Hex constant problem when compiling with -fugly-logint
10501       and -ftypeless-boz
10502
10503    x86-specific (Intel/AMD)
10504
10505     * [504]4490 ICE with -m128bit-long-double
10506     * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have
10507       `reg' in make_field_assignment, in combine.c
10508     * [506]12441 ICE: can't find a register to spill
10509     * [507]12943 array static-init failure under -fpic, -fPIC
10510     * [508]13608 Incorrect code with -O3 -ffast-math
10511
10512    PowerPC-specific
10513
10514     * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of
10515       __attribute__((aligned(16)))
10516     * [510]11793 ICE in extract_insn, in recog.c (const_vector's)
10517     * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in
10518       altivec.md)
10519     * [512]12537 g++ generates writeable text sections
10520
10521    SPARC-specific
10522
10523     * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0
10524       -m64
10525     * [514]12865 mprotect call to make trampoline executable may fail
10526     * [515]13354 ICE in sparc_emit_set_const32
10527
10528    ARM-specific
10529
10530     * [516]10467 [arm] ICE in pre_insert_copy_insn,
10531
10532    ia64-specific
10533
10534     * [517]11226 ICE passing struct arg with two floats
10535     * [518]11227 ICE for _Complex float, _Complex long double args
10536     * [519]12644 GCC 3.3.2 fails to compile glibc on ia64
10537     * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn
10538     * Various fixes for libunwind
10539
10540    Alpha-specific
10541
10542     * [521]12654 Incorrect comparison code generated for Alpha
10543     * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2
10544     * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2
10545
10546    HPPA-specific
10547
10548     * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c
10549     * [525]12158 [hppa] compilation does not terminate at -O1
10550
10551    S390-specific
10552
10553     * [526]11992 Wrong built-in code for memcmp with length 1<<24: only
10554       (1<<24)-1 possible for CLCL-Instruction
10555
10556    SH-specific
10557
10558     * [527]9365 segfault in gen_far_branch (config/sh/sh.c)
10559     * [528]10392 optimizer generates faulty array indexing
10560     * [529]11322 SH profiler outputs multiple definitions of symbol
10561     * [530]13069 gcc/config/sh/rtems.h broken
10562     * [531]13302 Putting a va_list in a struct causes seg fault
10563     * [532]13585 Incorrect optimization of call to sfunc
10564     * Fix inappropriately exported libgcc functions from the shared
10565       library
10566
10567    Other embedded target specific
10568
10569     * [533]8916 [mcore] unsigned char assign gets hosed.
10570     * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c
10571     * [535]13122 [h8300] local variable gets corrupted by function call
10572       when -fomit-frame-pointer is given
10573     * [536]13256 [cris] strict_low_part mistreated in delay slots
10574     * [537]13373 [mcore] optimization with -frerun-cse-after-loop
10575       -fexpensive-optimizations produces wrong code on mcore
10576
10577    GNU HURD-specific
10578
10579     * [538]12561 gcc/config/t-gnu needs updating to work with
10580       --with-sysroot
10581
10582    Tru64 Unix specific
10583
10584     * [539]6243 testsuite fails almost all tests due to no libintl in
10585       LD_LIBRARY_PATH during test.
10586     * [540]11397 weak aliases broken on Tru64 UNIX
10587
10588    AIX-specific
10589
10590     * [541]12505 build failure due to defines of uchar in cpphash.h and
10591       sys/types.h
10592     * [542]13150 WEAK symbols not exported by collect2
10593
10594    IRIX-specific
10595
10596     * [543]12666 fixincludes problem on IRIX 6.5.19m
10597
10598    Solaris-specific
10599
10600     * [544]12969 Including sys/byteorder.h breaks configure checks
10601
10602    Testsuite problems (compiler is not affected)
10603
10604     * [545]10819 testsuite creates CR+LF on compiler version lines in
10605       test summary files
10606     * [546]11612 abi_check not finding correct libgcc_s.so.1
10607
10608    Miscellaneous
10609
10610     * [547]13211 using -###, incorrect warnings about unused linker file
10611       are produced
10612     __________________________________________________________________
10613
10614GCC 3.3.4
10615
10616   This is the [548]list of problem reports (PRs) from GCC's bug tracking
10617   system that are known to be fixed in the 3.3.4 release. This list might
10618   not be complete (that is, it is possible that some PRs that have been
10619   fixed are not listed here).
10620     __________________________________________________________________
10621
10622GCC 3.3.5
10623
10624   This is the [549]list of problem reports (PRs) from GCC's bug tracking
10625   system that are known to be fixed in the 3.3.5 release. This list might
10626   not be complete (that is, it is possible that some PRs that have been
10627   fixed are not listed here).
10628     __________________________________________________________________
10629
10630GCC 3.3.6
10631
10632   This is the [550]list of problem reports (PRs) from GCC's bug tracking
10633   system that are known to be fixed in the 3.3.6 release. This list might
10634   not be complete (that is, it is possible that some PRs that have been
10635   fixed are not listed here).
10636
10637
10638    For questions related to the use of GCC, please consult these web
10639    pages and the [551]GCC manuals. If that fails, the
10640    [552]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10641    web pages and the development of GCC are welcome on our developer
10642    list at [553]gcc@gcc.gnu.org. All of [554]our lists have public
10643    archives.
10644
10645   Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and
10646   distribution of this entire article is permitted in any medium,
10647   provided this notice is preserved.
10648
10649   These pages are [556]maintained by the GCC team. Last modified
10650   2014-06-28[557].
10651
10652References
10653
10654   1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
10655   2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems
10656   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
10657   4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute
10658   5. http://gcc.gnu.org/news/dfa.html
10659   6. http://gcc.gnu.org/c99status.html
10660   7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html
10661   8. https://gcc.gnu.org/PR10140
10662   9. https://gcc.gnu.org/PR10198
10663  10. https://gcc.gnu.org/PR10338
10664  11. https://gcc.gnu.org/PR3581
10665  12. https://gcc.gnu.org/PR4382
10666  13. https://gcc.gnu.org/PR5533
10667  14. https://gcc.gnu.org/PR6387
10668  15. https://gcc.gnu.org/PR6412
10669  16. https://gcc.gnu.org/PR6620
10670  17. https://gcc.gnu.org/PR6663
10671  18. https://gcc.gnu.org/PR7068
10672  19. https://gcc.gnu.org/PR7083
10673  20. https://gcc.gnu.org/PR7647
10674  21. https://gcc.gnu.org/PR7675
10675  22. https://gcc.gnu.org/PR7718
10676  23. https://gcc.gnu.org/PR8116
10677  24. https://gcc.gnu.org/PR8358
10678  25. https://gcc.gnu.org/PR8511
10679  26. https://gcc.gnu.org/PR8564
10680  27. https://gcc.gnu.org/PR8660
10681  28. https://gcc.gnu.org/PR8766
10682  29. https://gcc.gnu.org/PR8803
10683  30. https://gcc.gnu.org/PR8846
10684  31. https://gcc.gnu.org/PR8906
10685  32. https://gcc.gnu.org/PR9216
10686  33. https://gcc.gnu.org/PR9261
10687  34. https://gcc.gnu.org/PR9263
10688  35. https://gcc.gnu.org/PR9429
10689  36. https://gcc.gnu.org/PR9516
10690  37. https://gcc.gnu.org/PR9600
10691  38. https://gcc.gnu.org/PR9629
10692  39. https://gcc.gnu.org/PR9672
10693  40. https://gcc.gnu.org/PR9749
10694  41. https://gcc.gnu.org/PR9794
10695  42. https://gcc.gnu.org/PR9829
10696  43. https://gcc.gnu.org/PR9916
10697  44. https://gcc.gnu.org/PR9936
10698  45. https://gcc.gnu.org/PR10262
10699  46. https://gcc.gnu.org/PR10278
10700  47. https://gcc.gnu.org/PR10446
10701  48. https://gcc.gnu.org/PR10451
10702  49. https://gcc.gnu.org/PR10506
10703  50. https://gcc.gnu.org/PR10549
10704  51. https://gcc.gnu.org/PR2001
10705  52. https://gcc.gnu.org/PR2391
10706  53. https://gcc.gnu.org/PR2960
10707  54. https://gcc.gnu.org/PR4046
10708  55. https://gcc.gnu.org/PR6405
10709  56. https://gcc.gnu.org/PR6798
10710  57. https://gcc.gnu.org/PR6871
10711  58. https://gcc.gnu.org/PR6909
10712  59. https://gcc.gnu.org/PR7189
10713  60. https://gcc.gnu.org/PR7642
10714  61. https://gcc.gnu.org/PR8634
10715  62. https://gcc.gnu.org/PR8750
10716  63. https://gcc.gnu.org/PR2161
10717  64. https://gcc.gnu.org/PR4319
10718  65. https://gcc.gnu.org/PR8602
10719  66. https://gcc.gnu.org/PR9177
10720  67. https://gcc.gnu.org/PR9853
10721  68. https://gcc.gnu.org/PR45
10722  69. https://gcc.gnu.org/PR3784
10723  70. https://gcc.gnu.org/PR764
10724  71. https://gcc.gnu.org/PR5116
10725  72. https://gcc.gnu.org/PR2862
10726  73. https://gcc.gnu.org/PR3663
10727  74. https://gcc.gnu.org/PR3797
10728  75. https://gcc.gnu.org/PR3948
10729  76. https://gcc.gnu.org/PR4137
10730  77. https://gcc.gnu.org/PR4361
10731  78. https://gcc.gnu.org/PR4802
10732  79. https://gcc.gnu.org/PR5837
10733  80. https://gcc.gnu.org/PR4803
10734  81. https://gcc.gnu.org/PR5094
10735  82. https://gcc.gnu.org/PR5730
10736  83. https://gcc.gnu.org/PR6713
10737  84. https://gcc.gnu.org/PR7015
10738  85. https://gcc.gnu.org/PR7086
10739  86. https://gcc.gnu.org/PR7099
10740  87. https://gcc.gnu.org/PR7247
10741  88. https://gcc.gnu.org/PR7441
10742  89. https://gcc.gnu.org/PR7768
10743  90. https://gcc.gnu.org/PR7804
10744  91. https://gcc.gnu.org/PR8099
10745  92. https://gcc.gnu.org/PR8117
10746  93. https://gcc.gnu.org/PR8205
10747  94. https://gcc.gnu.org/PR8645
10748  95. https://gcc.gnu.org/PR8724
10749  96. https://gcc.gnu.org/PR8805
10750  97. https://gcc.gnu.org/PR8691
10751  98. https://gcc.gnu.org/PR8700
10752  99. https://gcc.gnu.org/PR8724
10753 100. https://gcc.gnu.org/PR8949
10754 101. https://gcc.gnu.org/PR9016
10755 102. https://gcc.gnu.org/PR9053
10756 103. https://gcc.gnu.org/PR9152
10757 104. https://gcc.gnu.org/PR9182
10758 105. https://gcc.gnu.org/PR9297
10759 106. https://gcc.gnu.org/PR9318
10760 107. https://gcc.gnu.org/PR9320
10761 108. https://gcc.gnu.org/PR9400
10762 109. https://gcc.gnu.org/PR9424
10763 110. https://gcc.gnu.org/PR9425
10764 111. https://gcc.gnu.org/PR9439
10765 112. https://gcc.gnu.org/PR9474
10766 113. https://gcc.gnu.org/PR9548
10767 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231
10768 115. https://gcc.gnu.org/PR9555
10769 116. https://gcc.gnu.org/PR9561
10770 117. https://gcc.gnu.org/PR9563
10771 118. https://gcc.gnu.org/PR9582
10772 119. https://gcc.gnu.org/PR9622
10773 120. https://gcc.gnu.org/PR9683
10774 121. https://gcc.gnu.org/PR9791
10775 122. https://gcc.gnu.org/PR9817
10776 123. https://gcc.gnu.org/PR9825
10777 124. https://gcc.gnu.org/PR9826
10778 125. https://gcc.gnu.org/PR9924
10779 126. https://gcc.gnu.org/PR9946
10780 127. https://gcc.gnu.org/PR9964
10781 128. https://gcc.gnu.org/PR9988
10782 129. https://gcc.gnu.org/PR10033
10783 130. https://gcc.gnu.org/PR10097
10784 131. https://gcc.gnu.org/PR10132
10785 132. https://gcc.gnu.org/PR10180
10786 133. https://gcc.gnu.org/PR10199
10787 134. https://gcc.gnu.org/PR10300
10788 135. https://gcc.gnu.org/PR10427
10789 136. https://gcc.gnu.org/PR10503
10790 137. https://gcc.gnu.org/PR5956
10791 138. https://gcc.gnu.org/PR1832
10792 139. https://gcc.gnu.org/PR3924
10793 140. https://gcc.gnu.org/PR5634
10794 141. https://gcc.gnu.org/PR6367
10795 142. https://gcc.gnu.org/PR6491
10796 143. https://gcc.gnu.org/PR6742
10797 144. https://gcc.gnu.org/PR7113
10798 145. https://gcc.gnu.org/PR7236
10799 146. https://gcc.gnu.org/PR7278
10800 147. https://gcc.gnu.org/PR7384
10801 148. https://gcc.gnu.org/PR7388
10802 149. https://gcc.gnu.org/PR8587
10803 150. https://gcc.gnu.org/PR9038
10804 151. https://gcc.gnu.org/PR10197
10805 152. https://gcc.gnu.org/PR6005
10806 153. https://gcc.gnu.org/PR6389
10807 154. https://gcc.gnu.org/PR6576
10808 155. https://gcc.gnu.org/PR6652
10809 156. https://gcc.gnu.org/PR7060
10810 157. https://gcc.gnu.org/PR7073
10811 158. https://gcc.gnu.org/PR7180
10812 159. https://gcc.gnu.org/PR7416
10813 160. https://gcc.gnu.org/PR7570
10814 161. https://gcc.gnu.org/PR7578
10815 162. https://gcc.gnu.org/PR7611
10816 163. https://gcc.gnu.org/PR7709
10817 164. https://gcc.gnu.org/PR7766
10818 165. https://gcc.gnu.org/PR7785
10819 166. https://gcc.gnu.org/PR7786
10820 167. https://gcc.gnu.org/PR8142
10821 168. https://gcc.gnu.org/PR8234
10822 169. https://gcc.gnu.org/PR8415
10823 170. https://gcc.gnu.org/PR8481
10824 171. https://gcc.gnu.org/PR8593
10825 172. https://gcc.gnu.org/PR8759
10826 173. https://gcc.gnu.org/PR8997
10827 174. https://gcc.gnu.org/PR9253
10828 175. https://gcc.gnu.org/PR9254
10829 176. https://gcc.gnu.org/PR9271
10830 177. https://gcc.gnu.org/PR6767
10831 178. https://gcc.gnu.org/PR9911
10832 179. https://gcc.gnu.org/PR10020
10833 180. https://gcc.gnu.org/PR10546
10834 181. https://gcc.gnu.org/PR7029
10835 182. https://gcc.gnu.org/PR2903
10836 183. https://gcc.gnu.org/PR7873
10837 184. https://gcc.gnu.org/PR7680
10838 185. https://gcc.gnu.org/PR8705
10839 186. https://gcc.gnu.org/PR9986
10840 187. https://gcc.gnu.org/PR10056
10841 188. https://gcc.gnu.org/PR6744
10842 189. https://gcc.gnu.org/PR7361
10843 190. https://gcc.gnu.org/PR9496
10844 191. https://gcc.gnu.org/PR7067
10845 192. https://gcc.gnu.org/PR8480
10846 193. https://gcc.gnu.org/PR8784
10847 194. https://gcc.gnu.org/PR10315
10848 195. https://gcc.gnu.org/PR10267
10849 196. https://gcc.gnu.org/PR7916
10850 197. https://gcc.gnu.org/PR7926
10851 198. https://gcc.gnu.org/PR8555
10852 199. https://gcc.gnu.org/PR8994
10853 200. https://gcc.gnu.org/PR9426
10854 201. https://gcc.gnu.org/PR9806
10855 202. https://gcc.gnu.org/PR10077
10856 203. https://gcc.gnu.org/PR10233
10857 204. https://gcc.gnu.org/PR10286
10858 205. https://gcc.gnu.org/PR10308
10859 206. https://gcc.gnu.org/PR11272
10860 207. https://gcc.gnu.org/PR5754
10861 208. https://gcc.gnu.org/PR6597
10862 209. https://gcc.gnu.org/PR6949
10863 210. https://gcc.gnu.org/PR7053
10864 211. https://gcc.gnu.org/PR8164
10865 212. https://gcc.gnu.org/PR8384
10866 213. https://gcc.gnu.org/PR9559
10867 214. https://gcc.gnu.org/PR9649
10868 215. https://gcc.gnu.org/PR9864
10869 216. https://gcc.gnu.org/PR10432
10870 217. https://gcc.gnu.org/PR10475
10871 218. https://gcc.gnu.org/PR10635
10872 219. https://gcc.gnu.org/PR10661
10873 220. https://gcc.gnu.org/PR10700
10874 221. https://gcc.gnu.org/PR10712
10875 222. https://gcc.gnu.org/PR10796
10876 223. https://gcc.gnu.org/PR10890
10877 224. https://gcc.gnu.org/PR10939
10878 225. https://gcc.gnu.org/PR10956
10879 226. https://gcc.gnu.org/PR11041
10880 227. https://gcc.gnu.org/PR11059
10881 228. https://gcc.gnu.org/PR11083
10882 229. https://gcc.gnu.org/PR11105
10883 230. https://gcc.gnu.org/PR11149
10884 231. https://gcc.gnu.org/PR11228
10885 232. https://gcc.gnu.org/PR11282
10886 233. https://gcc.gnu.org/PR11301
10887 234. https://gcc.gnu.org/PR11308
10888 235. https://gcc.gnu.org/PR11473
10889 236. https://gcc.gnu.org/PR11503
10890 237. https://gcc.gnu.org/PR11513
10891 238. https://gcc.gnu.org/PR11198
10892 239. https://gcc.gnu.org/PR11304
10893 240. https://gcc.gnu.org/PR11381
10894 241. https://gcc.gnu.org/PR11536
10895 242. https://gcc.gnu.org/PR11557
10896 243. https://gcc.gnu.org/PR5897
10897 244. https://gcc.gnu.org/PR11279
10898 245. https://gcc.gnu.org/PR11022
10899 246. https://gcc.gnu.org/PR2330
10900 247. https://gcc.gnu.org/PR5388
10901 248. https://gcc.gnu.org/PR5390
10902 249. https://gcc.gnu.org/PR7877
10903 250. https://gcc.gnu.org/PR9393
10904 251. https://gcc.gnu.org/PR10032
10905 252. https://gcc.gnu.org/PR10468
10906 253. https://gcc.gnu.org/PR10527
10907 254. https://gcc.gnu.org/PR10679
10908 255. https://gcc.gnu.org/PR10682
10909 256. https://gcc.gnu.org/PR10689
10910 257. https://gcc.gnu.org/PR10845
10911 258. https://gcc.gnu.org/PR10849
10912 259. https://gcc.gnu.org/PR10888
10913 260. https://gcc.gnu.org/PR10929
10914 261. https://gcc.gnu.org/PR10931
10915 262. https://gcc.gnu.org/PR10940
10916 263. https://gcc.gnu.org/PR10968
10917 264. https://gcc.gnu.org/PR10990
10918 265. https://gcc.gnu.org/PR11039
10919 266. https://gcc.gnu.org/PR11062
10920 267. https://gcc.gnu.org/PR11095
10921 268. https://gcc.gnu.org/PR11098
10922 269. https://gcc.gnu.org/PR11137
10923 270. https://gcc.gnu.org/PR11154
10924 271. https://gcc.gnu.org/PR11329
10925 272. https://gcc.gnu.org/PR11332
10926 273. https://gcc.gnu.org/PR11431
10927 274. https://gcc.gnu.org/PR11528
10928 275. https://gcc.gnu.org/PR11546
10929 276. https://gcc.gnu.org/PR11567
10930 277. https://gcc.gnu.org/PR11645
10931 278. https://gcc.gnu.org/PR5179
10932 279. https://gcc.gnu.org/PR8204
10933 280. https://gcc.gnu.org/PR10838
10934 281. https://gcc.gnu.org/PR10886
10935 282. https://gcc.gnu.org/PR11349
10936 283. https://gcc.gnu.org/PR4823
10937 284. https://gcc.gnu.org/PR8878
10938 285. https://gcc.gnu.org/PR9815
10939 286. https://gcc.gnu.org/PR10402
10940 287. https://gcc.gnu.org/PR10504
10941 288. https://gcc.gnu.org/PR10673
10942 289. https://gcc.gnu.org/PR11044
10943 290. https://gcc.gnu.org/PR11089
10944 291. https://gcc.gnu.org/PR11420
10945 292. https://gcc.gnu.org/PR9362
10946 293. https://gcc.gnu.org/PR10142
10947 294. https://gcc.gnu.org/PR10663
10948 295. https://gcc.gnu.org/PR10835
10949 296. https://gcc.gnu.org/PR10876
10950 297. https://gcc.gnu.org/PR10955
10951 298. https://gcc.gnu.org/PR11018
10952 299. https://gcc.gnu.org/PR11556
10953 300. https://gcc.gnu.org/PR10907
10954 301. https://gcc.gnu.org/PR11320
10955 302. https://gcc.gnu.org/PR11599
10956 303. https://gcc.gnu.org/PR9745
10957 304. https://gcc.gnu.org/PR10871
10958 305. https://gcc.gnu.org/PR11440
10959 306. https://gcc.gnu.org/PR7594
10960 307. https://gcc.gnu.org/PR10557
10961 308. https://gcc.gnu.org/PR11054
10962 309. https://gcc.gnu.org/PR10834
10963 310. https://gcc.gnu.org/PR10842
10964 311. https://gcc.gnu.org/PR11052
10965 312. https://gcc.gnu.org/PR11183
10966 313. https://gcc.gnu.org/PR11084
10967 314. https://gcc.gnu.org/PR10331
10968 315. https://gcc.gnu.org/PR10413
10969 316. https://gcc.gnu.org/PR11096
10970 317. https://gcc.gnu.org/PR2873
10971 318. https://gcc.gnu.org/PR3163
10972 319. https://gcc.gnu.org/PR5287
10973 320. https://gcc.gnu.org/PR10148
10974 321. https://gcc.gnu.org/PR8787
10975 322. https://gcc.gnu.org/PR10900
10976 323. https://gcc.gnu.org/PR1607
10977 324. https://gcc.gnu.org/PR4252
10978 325. https://gcc.gnu.org/PR4490
10979 326. https://gcc.gnu.org/PR10355
10980 327. https://gcc.gnu.org/PR10726
10981 328. https://gcc.gnu.org/PR10805
10982 329. https://gcc.gnu.org/PR10815
10983 330. https://gcc.gnu.org/PR10877
10984 331. https://gcc.gnu.org/PR11280
10985 332. https://gcc.gnu.org/PR11466
10986 333. https://gcc.gnu.org/PR10737
10987 334. https://gcc.gnu.org/PR10810
10988 335. https://gcc.gnu.org/PR8336
10989 336. https://gcc.gnu.org/PR9330
10990 337. https://gcc.gnu.org/PR9631
10991 338. https://gcc.gnu.org/PR9877
10992 339. https://gcc.gnu.org/PR11687
10993 340. https://gcc.gnu.org/PR12263
10994 341. https://gcc.gnu.org/PR12490
10995 342. https://gcc.gnu.org/PR7277
10996 343. https://gcc.gnu.org/PR7939
10997 344. https://gcc.gnu.org/PR11063
10998 345. https://gcc.gnu.org/PR11207
10999 346. https://gcc.gnu.org/PR11522
11000 347. https://gcc.gnu.org/PR11595
11001 348. https://gcc.gnu.org/PR11646
11002 349. https://gcc.gnu.org/PR11665
11003 350. https://gcc.gnu.org/PR11852
11004 351. https://gcc.gnu.org/PR11878
11005 352. https://gcc.gnu.org/PR11883
11006 353. https://gcc.gnu.org/PR11991
11007 354. https://gcc.gnu.org/PR12146
11008 355. https://gcc.gnu.org/PR12215
11009 356. https://gcc.gnu.org/PR12369
11010 357. https://gcc.gnu.org/PR12446
11011 358. https://gcc.gnu.org/PR12510
11012 359. https://gcc.gnu.org/PR12544
11013 360. https://gcc.gnu.org/PR9862
11014 361. https://gcc.gnu.org/PR10962
11015 362. https://gcc.gnu.org/PR11370
11016 363. https://gcc.gnu.org/PR11637
11017 364. https://gcc.gnu.org/PR11885
11018 365. https://gcc.gnu.org/PR12082
11019 366. https://gcc.gnu.org/PR12180
11020 367. https://gcc.gnu.org/PR12340
11021 368. https://gcc.gnu.org/PR3907
11022 369. https://gcc.gnu.org/PR5293
11023 370. https://gcc.gnu.org/PR5296
11024 371. https://gcc.gnu.org/PR7939
11025 372. https://gcc.gnu.org/PR8656
11026 373. https://gcc.gnu.org/PR10147
11027 374. https://gcc.gnu.org/PR11400
11028 375. https://gcc.gnu.org/PR11409
11029 376. https://gcc.gnu.org/PR11740
11030 377. https://gcc.gnu.org/PR11786
11031 378. https://gcc.gnu.org/PR11867
11032 379. https://gcc.gnu.org/PR11928
11033 380. https://gcc.gnu.org/PR12114
11034 381. https://gcc.gnu.org/PR12163
11035 382. https://gcc.gnu.org/PR12181
11036 383. https://gcc.gnu.org/PR12236
11037 384. https://gcc.gnu.org/PR12266
11038 385. https://gcc.gnu.org/PR12296
11039 386. https://gcc.gnu.org/PR12298
11040 387. https://gcc.gnu.org/PR12369
11041 388. https://gcc.gnu.org/PR12337
11042 389. https://gcc.gnu.org/PR12344
11043 390. https://gcc.gnu.org/PR12451
11044 391. https://gcc.gnu.org/PR12486
11045 392. https://gcc.gnu.org/PR8869
11046 393. https://gcc.gnu.org/PR9786
11047 394. https://gcc.gnu.org/PR11689
11048 395. https://gcc.gnu.org/PR12116
11049 396. https://gcc.gnu.org/PR12070
11050 397. https://gcc.gnu.org/PR11184
11051 398. https://gcc.gnu.org/PR11535
11052 399. https://gcc.gnu.org/PR11693
11053 400. https://gcc.gnu.org/PR12224
11054 401. https://gcc.gnu.org/PR11087
11055 402. https://gcc.gnu.org/PR11319
11056 403. https://gcc.gnu.org/PR11949
11057 404. https://gcc.gnu.org/PR11662
11058 405. https://gcc.gnu.org/PR11965
11059 406. https://gcc.gnu.org/PR12301
11060 407. https://gcc.gnu.org/PR11717
11061 408. https://gcc.gnu.org/PR11313
11062 409. https://gcc.gnu.org/PR11712
11063 410. https://gcc.gnu.org/PR12166
11064 411. https://gcc.gnu.org/PR12101
11065 412. https://gcc.gnu.org/PR10988
11066 413. https://gcc.gnu.org/PR11805
11067 414. https://gcc.gnu.org/PR11902
11068 415. https://gcc.gnu.org/PR11903
11069 416. https://gcc.gnu.org/PR11890
11070 417. https://gcc.gnu.org/PR12399
11071 418. https://gcc.gnu.org/PR13068
11072 419. https://gcc.gnu.org/PR10060
11073 420. https://gcc.gnu.org/PR10555
11074 421. https://gcc.gnu.org/PR10706
11075 422. https://gcc.gnu.org/PR11496
11076 423. https://gcc.gnu.org/PR11741
11077 424. https://gcc.gnu.org/PR12440
11078 425. https://gcc.gnu.org/PR12632
11079 426. https://gcc.gnu.org/PR12712
11080 427. https://gcc.gnu.org/PR12726
11081 428. https://gcc.gnu.org/PR12890
11082 429. https://gcc.gnu.org/PR12900
11083 430. https://gcc.gnu.org/PR13060
11084 431. https://gcc.gnu.org/PR13289
11085 432. https://gcc.gnu.org/PR13318
11086 433. https://gcc.gnu.org/PR13392
11087 434. https://gcc.gnu.org/PR13574
11088 435. https://gcc.gnu.org/PR13475
11089 436. https://gcc.gnu.org/PR13797
11090 437. https://gcc.gnu.org/PR13824
11091 438. https://gcc.gnu.org/PR8776
11092 439. https://gcc.gnu.org/PR10339
11093 440. https://gcc.gnu.org/PR11350
11094 441. https://gcc.gnu.org/PR12826
11095 442. https://gcc.gnu.org/PR12500
11096 443. https://gcc.gnu.org/PR12941
11097 444. https://gcc.gnu.org/PR12953
11098 445. https://gcc.gnu.org/PR13041
11099 446. https://gcc.gnu.org/PR13507
11100 447. https://gcc.gnu.org/PR13382
11101 448. https://gcc.gnu.org/PR13394
11102 449. https://gcc.gnu.org/PR13400
11103 450. https://gcc.gnu.org/PR13521
11104 451. https://gcc.gnu.org/PR2094
11105 452. https://gcc.gnu.org/PR2294
11106 453. https://gcc.gnu.org/PR5050
11107 454. https://gcc.gnu.org/PR9371
11108 455. https://gcc.gnu.org/PR9546
11109 456. https://gcc.gnu.org/PR10081
11110 457. https://gcc.gnu.org/PR10093
11111 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61
11112 459. https://gcc.gnu.org/PR10095
11113 460. https://gcc.gnu.org/PR11554
11114 461. https://gcc.gnu.org/PR12297
11115 462. https://gcc.gnu.org/PR12352
11116 463. https://gcc.gnu.org/PR12438
11117 464. https://gcc.gnu.org/PR12540
11118 465. https://gcc.gnu.org/PR12594
11119 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60
11120 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63
11121 468. https://gcc.gnu.org/PR12657
11122 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292
11123 470. https://gcc.gnu.org/PR12696
11124 471. https://gcc.gnu.org/PR12815
11125 472. https://gcc.gnu.org/PR12862
11126 473. https://gcc.gnu.org/PR12926
11127 474. https://gcc.gnu.org/PR12967
11128 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html
11129 476. https://gcc.gnu.org/PR12971
11130 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328
11131 478. https://gcc.gnu.org/PR13007
11132 479. https://gcc.gnu.org/PR13009
11133 480. https://gcc.gnu.org/PR13057
11134 481. https://gcc.gnu.org/PR13070
11135 482. https://gcc.gnu.org/PR13081
11136 483. https://gcc.gnu.org/PR13239
11137 484. https://gcc.gnu.org/PR13262
11138 485. https://gcc.gnu.org/PR13290
11139 486. https://gcc.gnu.org/PR13323
11140 487. https://gcc.gnu.org/PR13369
11141 488. https://gcc.gnu.org/PR13371
11142 489. https://gcc.gnu.org/PR13445
11143 490. https://gcc.gnu.org/PR13461
11144 491. https://gcc.gnu.org/PR13462
11145 492. https://gcc.gnu.org/PR13478
11146 493. https://gcc.gnu.org/PR13544
11147 494. https://gcc.gnu.org/PR13650
11148 495. https://gcc.gnu.org/PR13683
11149 496. https://gcc.gnu.org/PR13688
11150 497. https://gcc.gnu.org/PR13774
11151 498. https://gcc.gnu.org/PR13884
11152 499. https://gcc.gnu.org/PR10746
11153 500. https://gcc.gnu.org/PR11433
11154 501. https://gcc.gnu.org/PR12633
11155 502. https://gcc.gnu.org/PR13037
11156 503. https://gcc.gnu.org/PR13213
11157 504. https://gcc.gnu.org/PR4490
11158 505. https://gcc.gnu.org/PR12292
11159 506. https://gcc.gnu.org/PR12441
11160 507. https://gcc.gnu.org/PR12943
11161 508. https://gcc.gnu.org/PR13608
11162 509. https://gcc.gnu.org/PR11598
11163 510. https://gcc.gnu.org/PR11793
11164 511. https://gcc.gnu.org/PR12467
11165 512. https://gcc.gnu.org/PR12537
11166 513. https://gcc.gnu.org/PR12496
11167 514. https://gcc.gnu.org/PR12865
11168 515. https://gcc.gnu.org/PR13354
11169 516. https://gcc.gnu.org/PR10467
11170 517. https://gcc.gnu.org/PR11226
11171 518. https://gcc.gnu.org/PR11227
11172 519. https://gcc.gnu.org/PR12644
11173 520. https://gcc.gnu.org/PR13149
11174 521. https://gcc.gnu.org/PR12654
11175 522. https://gcc.gnu.org/PR12965
11176 523. https://gcc.gnu.org/PR13031
11177 524. https://gcc.gnu.org/PR11634
11178 525. https://gcc.gnu.org/PR12158
11179 526. https://gcc.gnu.org/PR11992
11180 527. https://gcc.gnu.org/PR9365
11181 528. https://gcc.gnu.org/PR10392
11182 529. https://gcc.gnu.org/PR11322
11183 530. https://gcc.gnu.org/PR13069
11184 531. https://gcc.gnu.org/PR13302
11185 532. https://gcc.gnu.org/PR13585
11186 533. https://gcc.gnu.org/PR8916
11187 534. https://gcc.gnu.org/PR11576
11188 535. https://gcc.gnu.org/PR13122
11189 536. https://gcc.gnu.org/PR13256
11190 537. https://gcc.gnu.org/PR13373
11191 538. https://gcc.gnu.org/PR12561
11192 539. https://gcc.gnu.org/PR6243
11193 540. https://gcc.gnu.org/PR11397
11194 541. https://gcc.gnu.org/PR12505
11195 542. https://gcc.gnu.org/PR13150
11196 543. https://gcc.gnu.org/PR12666
11197 544. https://gcc.gnu.org/PR12969
11198 545. https://gcc.gnu.org/PR10819
11199 546. https://gcc.gnu.org/PR11612
11200 547. https://gcc.gnu.org/PR13211
11201 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4
11202 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5
11203 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6
11204 551. https://gcc.gnu.org/onlinedocs/
11205 552. mailto:gcc-help@gcc.gnu.org
11206 553. mailto:gcc@gcc.gnu.org
11207 554. https://gcc.gnu.org/lists.html
11208 555. http://www.fsf.org/
11209 556. https://gcc.gnu.org/about.html
11210 557. http://validator.w3.org/check/referer
11211======================================================================
11212http://gcc.gnu.org/gcc-3.2/index.html
11213
11214                             GCC 3.2 Release Series
11215
11216   April 25, 2003
11217
11218   The [1]GNU project and the GCC developers are pleased to announce the
11219   release of GCC 3.2.3.
11220
11221   The purpose of the GCC 3.2 release series is to provide a stable
11222   platform for OS distributors to use building their next releases. A
11223   primary objective was to stabilize the C++ ABI; we believe that the
11224   interface to the compiler and the C++ standard library are now
11225   relatively stable.
11226
11227   Be aware that C++ code compiled by GCC 3.2.x will (in general) not
11228   interoperate with code compiled by GCC 3.1.1 or earlier.
11229
11230   Please refer to our [2]detailed list of news, caveats, and bug-fixes
11231   for further information.
11232
11233Release History
11234
11235   GCC 3.2.3
11236          April 25, 2003 ([3]changes)
11237
11238   GCC 3.2.2
11239          February 5, 2003 ([4]changes)
11240
11241   GCC 3.2.1
11242          November 19, 2002 ([5]changes)
11243
11244   GCC 3.2
11245          August 14, 2002 ([6]changes)
11246
11247References and Acknowledgements
11248
11249   GCC used to stand for the GNU C Compiler, but since the compiler
11250   supports several other languages aside from C, it now stands for the
11251   GNU Compiler Collection.
11252
11253   A list of [7]successful builds is updated as new information becomes
11254   available.
11255
11256   The GCC developers would like to thank the numerous people that have
11257   contributed new features, improvements, bug fixes, and other changes as
11258   well as test results to GCC. This [8]amazing group of volunteers is
11259   what makes GCC successful.
11260
11261   For additional information about GCC please refer to the [9]GCC project
11262   web site or contact the [10]GCC development mailing list.
11263
11264   To obtain GCC please use [11]our mirror sites, or our CVS server.
11265
11266
11267    For questions related to the use of GCC, please consult these web
11268    pages and the [12]GCC manuals. If that fails, the
11269    [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11270    web pages and the development of GCC are welcome on our developer
11271    list at [14]gcc@gcc.gnu.org. All of [15]our lists have public
11272    archives.
11273
11274   Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and
11275   distribution of this entire article is permitted in any medium,
11276   provided this notice is preserved.
11277
11278   These pages are [17]maintained by the GCC team. Last modified
11279   2014-06-28[18].
11280
11281References
11282
11283   1. http://www.gnu.org/
11284   2. http://gcc.gnu.org/gcc-3.2/changes.html
11285   3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
11286   4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2
11287   5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1
11288   6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2
11289   7. http://gcc.gnu.org/gcc-3.2/buildstat.html
11290   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11291   9. http://gcc.gnu.org/index.html
11292  10. mailto:gcc@gcc.gnu.org
11293  11. http://gcc.gnu.org/mirrors.html
11294  12. https://gcc.gnu.org/onlinedocs/
11295  13. mailto:gcc-help@gcc.gnu.org
11296  14. mailto:gcc@gcc.gnu.org
11297  15. https://gcc.gnu.org/lists.html
11298  16. http://www.fsf.org/
11299  17. https://gcc.gnu.org/about.html
11300  18. http://validator.w3.org/check/referer
11301======================================================================
11302http://gcc.gnu.org/gcc-3.2/changes.html
11303
11304                             GCC 3.2 Release Series
11305                        Changes, New Features, and Fixes
11306
11307   The latest release in the 3.2 release series is [1]GCC 3.2.3.
11308
11309Caveats and New Features
11310
11311  Caveats
11312
11313     * The C++ compiler does not correctly zero-initialize
11314       pointers-to-data members. You must explicitly initialize them. For
11315       example: int S::*m(0); will work, but depending on
11316       default-initialization to zero will not work. This bug cannot be
11317       fixed in GCC 3.2 without inducing unacceptable risks. It will be
11318       fixed in GCC 3.3.
11319     * This GCC release is based on the GCC 3.1 sourcebase, and thus has
11320       all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has
11321       a number of C++ ABI fixes which make its C++ compiler generate
11322       binary code which is incompatible with the C++ compilers found in
11323       earlier GCC releases, including GCC 3.1 and GCC 3.1.1.
11324
11325  Frontend Enhancements
11326
11327    C/C++/Objective-C
11328
11329     * The method of constructing the list of directories to be searched
11330       for header files has been revised. If a directory named by a -I
11331       option is a standard system include directory, the option is
11332       ignored to ensure that the default search order for system
11333       directories and the special treatment of system header files are
11334       not defeated.
11335     * The C and Objective-C compilers no longer accept the "Naming Types"
11336       extension (typedef foo = bar); it was already unavailable in C++.
11337       Code which uses it will need to be changed to use the "typeof"
11338       extension instead: typedef typeof(bar) foo. (We have removed this
11339       extension without a period of deprecation because it has caused the
11340       compiler to crash since version 3.0 and no one noticed until very
11341       recently. Thus we conclude it is not in widespread use.)
11342
11343    C++
11344
11345     * GCC 3.2 fixed serveral differences between the C++ ABI implemented
11346       in GCC and the multi-vendor standard, but more have been found
11347       since the release. 3.2.1 adds a new warning, -Wabi, to warn about
11348       code which is affected by these bugs. We will fix these bugs in
11349       some future release, once we are confident that all have been
11350       found; until then, it is our intention to make changes to the ABI
11351       only if they are necessary for correct compilation of C++, as
11352       opposed to conformance to the ABI documents.
11353     * For details on how to build an ABI compliant compiler for GNU/Linux
11354       systems, check the [3]common C++ ABI page.
11355
11356  New Targets and Target Specific Improvements
11357
11358    IA-32
11359
11360     * Fixed a number of bugs in SSE and MMX intrinsics.
11361     * Fixed common compiler crashes with SSE instruction set enabled
11362       (implied by -march=pentium3, pentium4, athlon-xp)
11363     * __m128 and __m128i is not 128bit aligned when used in structures.
11364
11365    x86-64
11366
11367     * A bug whereby the compiler could generate bad code for bzero has
11368       been fixed.
11369     * ABI fixes (implying ABI incompatibilities with previous version in
11370       some corner cases)
11371     * Fixed prefetch code generation
11372     __________________________________________________________________
11373
11374GCC 3.2.3
11375
11376   3.2.3 is a bug fix release only; there are no new features that were
11377   not present in GCC 3.2.2.
11378
11379  Bug Fixes
11380
11381   This section lists the problem reports (PRs) from GCC's bug tracking
11382   system that are known to be fixed in the 3.2.3 release. This list might
11383   not be complete (that is, it is possible that some PRs that have been
11384   fixed are not listed here), and some of the titles have been changed to
11385   make them more clear.
11386
11387    Internal Compiler Errors (multi-platform)
11388
11389     * [4]3782: (c++) -quiet -fstats produces a segmentation fault in
11390       cc1plus
11391     * [5]6440: (c++) template specializations cause ICE
11392     * [6]7050: (c++) ICE on: (i ? get_string() : throw)
11393     * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c)
11394     * [8]7982: (c++) ICE due to infinite recursion (using STL set)
11395     * [9]8068: exceedingly high (infinite) memory usage
11396     * [10]8178: ICE with __builtin_ffs
11397     * [11]8396: ICE in copy_to_mode_reg, in explow.c
11398     * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c
11399     * [13]9768: ICE when optimizing inline code at -O2
11400     * [14]9798: (c++) Infinite recursion (segfault) in
11401       cp/decl.c:push_using_directive with recursive using directives
11402     * [15]9799: mismatching structure initializer with nested flexible
11403       array member: ICE
11404     * [16]9928: ICE on duplicate enum declaration
11405     * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects
11406       sparc, alpha)
11407     * [18]10352: ICE in find_reloads_toplev
11408     * [19]10336: ICE with -Wunreachable-code
11409
11410    C/optimizer bugs:
11411
11412     * [20]8224: Incorrect joining of signed and unsigned division
11413     * [21]8613: -O2 produces wrong code with builtin strlen and
11414       postincrements
11415     * [22]8828: gcc reports some code is unreachable when it is not
11416     * [23]9226: GCSE breaking argument passing
11417     * [24]9853: miscompilation of non-constant structure initializer
11418     * [25]9797: C99-style struct initializers are miscompiled
11419     * [26]9967: Some standard C function calls should not be replaced
11420       when optimizing for size
11421     * [27]10116: ce2: invalid merge of join_bb in the context of switch
11422       statements
11423     * [28]10171: wrong code for inlined function
11424     * [29]10175: -Wunreachable-code doesn't work for single lines
11425
11426    C++ compiler and library:
11427
11428     * [30]8316: Confusing diagnostic for code that misuses conversion
11429       operators
11430     * [31]9169: filebuf output fails if codecvt<>::out returns noconv
11431     * [32]9420: incomplete type incorrectly reported
11432     * [33]9459: typeof in return type specification of template not
11433       supported
11434     * [34]9507: filebuf::open handles ios_base::ate incorrectly
11435     * [35]9538: Out-of-bounds memory access in streambuf::sputbackc
11436     * [36]9602: Total confusion about template/friend/virtual/abstract
11437     * [37]9993: destructor not called for local object created within and
11438       returned from infinite loop
11439     * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2
11440       system
11441
11442    Java compiler and library:
11443
11444     * [39]9652: libgcj build fails on irix6.5.1[78]
11445     * [40]10144: gas on solaris complains about bad .stabs lines for
11446       java, native as unaffected
11447
11448    x86-specific (Intel/AMD):
11449
11450     * [41]8746: gcc miscompiles Linux kernel ppa driver on x86
11451     * [42]9888: -mcpu=k6 -Os produces out of range loop instructions
11452     * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu
11453       failed
11454     * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib)
11455       failed
11456
11457    SPARC-specific:
11458
11459     * [45]7784: [Sparc] ICE in extract_insn, in recog.c
11460     * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in
11461       unroll.c
11462     * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc
11463     * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in
11464       execute/loop-2d.c
11465     * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc
11466     * [50]9414: Scheduling bug on Ultrasparc
11467     * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64
11468
11469    m68k-specific:
11470
11471     * [52]7248: broken "inclusive or" code
11472     * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1
11473
11474    PowerPC-specific:
11475
11476     * [54]9732: Wrong code with -O2 -fPIC
11477     * [55]10073: ICE: powerpc cannot split insn
11478
11479    Alpha-specific:
11480
11481     * [56]7702: optimization problem on a DEC alpha under OSF1
11482     * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system
11483
11484    HP-specific:
11485
11486     * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275)
11487     * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10
11488       (missing symbol)
11489     * [60]10271: Floating point args don't get reloaded across function
11490       calls with -O2
11491
11492    MIPS specific:
11493
11494     * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in
11495       compile/920501-4.c
11496
11497    CRIS specific:
11498
11499     * [62]10377: gcc-3.2.2 creates bad assembler code for cris
11500
11501    Miscellaneous and minor bugs:
11502
11503     * [63]6955: collect2 says "core dumped" when there is no core
11504     __________________________________________________________________
11505
11506GCC 3.2.2
11507
11508   Beginning with 3.2.2, GCC's Makefile suite supports redirection of make
11509   install by means of the DESTDIR variable. Parts of the GCC tree have
11510   featured that support long before, but now it is available even from
11511   the top level.
11512
11513   Other than that, GCC 3.2.2 is a bug fix release only; there are no new
11514   features that were not present in GCC 3.2.1.
11515
11516  Bug Fixes
11517
11518   On the following i386-based systems GCC 3.2.1 broke the C ABI wrt.
11519   functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped
11520   with FreeBSD 5.0 does not have this problem), Interix, a.out-based
11521   GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI
11522   change, and thus restores ABI-compatibility with previous releases
11523   (except GCC 3.2.1) on these platforms.
11524
11525   This section lists the problem reports (PRs) from GCC's bug tracking
11526   system that are known to be fixed in the 3.2.2 release. This list might
11527   not be complete (that is, it is possible that some PRs that have been
11528   fixed are not listed here) and some of the titles have been changed to
11529   make them more clear.
11530
11531    Internal Compiler Errors (multi-platform)
11532
11533     * [64]5919: (c++) ICE when passing variable array to template
11534       function
11535     * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=)
11536     * [66]7507: ICE with -O2 when address of called function is a
11537       complicated expression
11538     * [67]7622: ICE with nested inline functions if function's address is
11539       taken
11540     * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR
11541       [69]9258)
11542     * [70]8031: (c++) ICE in code comparing typeids and casting from
11543       virtual base
11544     * [71]8275: ICE in simplify_subreg
11545     * [72]8332: (c++) builtin strlen/template interaction causes ICE
11546     * [73]8372: (c++) ICE on explicit call of destructor
11547     * [74]8439: (c, not c++) empty struct causes ICE
11548     * [75]8442: (c++) ICE with nested template classes
11549     * [76]8518: ICE when compiling mplayer ("extern inline" issue)
11550     * [77]8615: (c++) ICE with out-of-range character constant template
11551       argument
11552     * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307
11553     * [79]8799: (c++) ICE: error reporting routines re-entered
11554     * [80]9328: (c++) ICE with typeof(X) for overloaded X
11555     * [81]9465: (preprocessor) cpp -traditional ICE on null bytes
11556
11557    C++ (compiler and library) bugs
11558
11559     * [82]47: scoping in nested classes is broken
11560     * [83]6745: problems with iostream rdbuf() member function
11561     * [84]8214: conversion from const char* const to char* sometimes
11562       accepted illegally
11563     * [85]8493: builtin strlen and overload resolution (same bug as
11564       [86]8332)
11565     * [87]8503: strange behaviour of function types
11566     * [88]8727: compiler confused by inheritance from an anonymous struct
11567     * [89]7445: poor performance of std::locale::classic() in
11568       multi-threaded applications
11569     * [90]8230: mishandling of overflow in vector<T>::resize
11570     * [91]8399: sync_with_stdio(false) breaks unformatted input
11571     * [92]8662: illegal access of private member of unnamed class is
11572       accepted
11573     * [93]8707: "make distclean" fails in libstdc++-v3 directory
11574     * [94]8708: __USE_MALLOC doesn't work
11575     * [95]8790: Use of non-thread-safe strtok in src/localename.cc
11576     * [96]8887: Bug in date formats with --enable-clocale=generic
11577     * [97]9076: Call Frame Instructions are not handled correctly during
11578       unwind operation
11579     * [98]9151: std::setprecision limited to 16 digits when outputting a
11580       double to a stream
11581     * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers
11582     * [100]9269: libstdc++ headers: explicit specialization of function
11583       must precede its first use
11584     * [101]9322: return value of basic_streambuf<>::getloc affected by
11585       locale::global
11586     * [102]9433: segfault in runtime support for dynamic_cast
11587
11588    C and optimizer bugs
11589
11590     * [103]8032: GCC incorrectly initializes static structs that have
11591       flexible arrays
11592     * [104]8639: simple arithmetic expression broken
11593     * [105]8794: optimization improperly eliminates certain expressions
11594     * [106]8832: traditional "asm volatile" code is illegally optimized
11595     * [107]8988: loop optimizer bug: with -O2, code is generated that
11596       segfaults (found on i386, bug present for all platforms)
11597     * [108]9492: structure copy clobbers subsequent stores to structure
11598
11599    Objective-C bugs
11600
11601     * [109]9267: Objective-C parser won't build with newer bison versions
11602       (e.g. 1.875)
11603
11604    Ada bugs
11605
11606     * [110]8344: Ada build problem due to conflict between gcc/final.o,
11607       gcc/ada/final.o
11608
11609    Preprocessor bugs
11610
11611     * [111]8524: _Pragma within macros is improperly expanded
11612     * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with
11613       -fshort-wchar
11614
11615    ARM-specific
11616
11617     * [113]9090: arm ICE with >= -O2; regression from gcc-2.95
11618
11619    x86-specific (Intel/AMD)
11620
11621     * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction)
11622     * [115]8599: loop unroll bug with -march=k6-3
11623     * [116]9506: ABI breakage in structure return (affects BSD and
11624       Cygwin, but not GNU/Linux)
11625
11626    FreeBSD 5.0 specific
11627
11628     * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0
11629
11630    RTEMS-specific
11631
11632     * [118]9292: hppa1.1-rtems configurery problems
11633     * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug
11634     * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue
11635     * [121]9296: gthr-rtems regression
11636     * [122]9316: powerpc-rtems: extending multilibs
11637
11638    HP-PA specific
11639
11640     * [123]9493: ICE with -O2 when building a simple function
11641
11642    Documentation
11643
11644     * [124]7341: hyperlink to gcov in GCC documentation doesn't work
11645     * [125]8947: Please add a warning about "-malign-double" in docs
11646     * [126]7448, [127]8882: typo cleanups
11647     __________________________________________________________________
11648
11649GCC 3.2.1
11650
11651   3.2.1 adds a new warning, -Wabi. This option warns when GNU C++
11652   generates code that is known not to be binary-compatible with the
11653   vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included
11654   in the distribution, for details.
11655
11656   This release also removes an old GCC extension, "naming types", and the
11657   documentation now directs users to use a different GCC extension,
11658   __typeof__, instead. The feature had evidently been broken for a while.
11659
11660   Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and
11661   the new warning there are no new features that were not present in GCC
11662   3.2.
11663
11664   In addition, the previous fix for [128]PR 7445 (poor performance of
11665   std::locale::classic() in multi-threaded applications) was reverted
11666   ("unfixed"), because the "fix" was not thread-safe.
11667
11668  Bug Fixes
11669
11670   This section lists the problem reports (PRs) from GCC's bug tracking
11671   system that are known to be fixed in the 3.2.1 release. This list might
11672   not be complete (that is, it is possible that some PRs that have been
11673   fixed are not listed here). As you can see, the number of bug fixes is
11674   quite large, so it is strongly recommended that users of earlier GCC
11675   3.x releases upgrade to GCC 3.2.1.
11676
11677    Internal Compiler Errors (multi-platform)
11678
11679     * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c
11680     * [130]5661: (c++) ICE instantiating template on array of unknown
11681       size (bad code)
11682     * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on
11683       64-bit platforms
11684     * [132]6994: (c++) ICE in find_function_data
11685     * [133]7150: preprocessor: GCC -dM -E gives an ICE
11686     * [134]7160: ICE when optimizing branches without a return value
11687     * [135]7228: (c++) ICE when using member template and template
11688       function
11689     * [136]7266: (c++) ICE with -pedantic on missing typename
11690     * [137]7353: ICE from use of "Naming Types" extension, see above
11691     * [138]7411: ICE in instantiate_virtual_regs_1, in function.c
11692     * [139]7478: (c++) ICE on static_cast inside template
11693     * [140]7526: preprocessor core dump when _Pragma implies #pragma
11694       dependency
11695     * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803
11696       is a duplicate)
11697     * [143]7754: (c++) ICE on union with template parameter
11698     * [144]7788: (c++) redeclaring a definition as an incomplete class
11699       causes ICE
11700     * [145]8031: (c++) ICE in comptypes, in cp/typeck.c
11701     * [146]8055: preprocessor dies with SIG11 when building FreeBSD
11702       kernel
11703     * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related
11704       variables
11705     * [148]8134: (c++) ICE in force_store_init_value on legal code
11706     * [149]8149: (c++) ICE on incomplete type
11707     * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array
11708       initialization
11709
11710    C++ (compiler and library) bugs
11711
11712     * [151]5607: No pointer adjustment in covariant return types
11713     * [152]6579: Infinite loop with statement expressions in member
11714       initialization
11715     * [153]6803: Default copy constructor bug in GCC 3.1
11716     * [154]7176: g++ confused by friend and static member with same name
11717     * [155]7188: Segfault with template class and recursive (incorrect)
11718       initializer list
11719     * [156]7306: Regression: GCC 3.x fails to compile code with virtual
11720       inheritance if a method has a variable number of arguments
11721     * [157]7461: ctype<char>::classic_table() returns offset array on
11722       Cygwin
11723     * [158]7524: f(const float arg[3]) fails
11724     * [159]7584: Erroneous ambiguous base error on using declaration
11725     * [160]7676: Member template overloading problem
11726     * [161]7679: infinite loop when a right parenthesis is missing
11727     * [162]7811: default locale not taken from environment
11728     * [163]7961: compare( char *) implemented incorrectly in
11729       basic_string<>
11730     * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if
11731       streambuf::underflow() leaves gptr() NULL (dups: [165]8127,
11732       [166]6745)
11733     * [167]8096: deque::at() throws std::range_error instead of
11734       std::out_of_range
11735     * [168]8127: cout << cin.rdbuf() infinite loop
11736     * [169]8218: Excessively large memory consumed for classes with large
11737       array members
11738     * [170]8287: GCC 3.2: Destructor called for non-constructed local
11739       object
11740     * [171]8347: empty vector range used in string construction causes
11741       core dump
11742     * [172]8348: fail() flag is set in istringstream when eof() flag is
11743       set
11744     * [173]8391: regression: infinite loop in cp/decl2.c(finish_file)
11745
11746    C and optimizer bugs
11747
11748     * [174]6627: -fno-align-functions doesn't seem to disable function
11749       alignment
11750     * [175]6631: life_analysis misoptimizes code to initialize fields of
11751       a structure
11752     * [176]7102: unsigned char division results in floating exception
11753     * [177]7120: Run once loop should *always* be unrolled
11754       (pessimization)
11755     * [178]7209: Bug involving array referencing and ?: operator
11756     * [179]7515: invalid inlining of global function with -O3
11757     * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test
11758     * [181]8467: bug in sibling call optimization
11759
11760    Preprocessor bugs
11761
11762     * [182]4890: incorrect line markers from the traditional preprocessor
11763     * [183]7357: -M option omits system headers files (making it the same
11764       as -MM)
11765     * [184]7358: Changes to Sun's make Dependencies
11766     * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as
11767       C headers
11768     * [186]7862: Interrupting GCC -MD removes .d file but not .o
11769     * [187]8190: Failed compilation deletes -MD dependency file
11770     * [188]8524: _Pragma within macro is improperly expanded
11771
11772    x86 specific (Intel/AMD)
11773
11774     * [189]5351: (i686-only) function pass-by-value structure copy
11775       corrupts stack ([190]7591 is a duplicate)
11776     * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with
11777       -march=pentium3/pentium2/athlon (these are all the same underlying
11778       bug, in MMX register use)
11779     * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same
11780       as above?)
11781     * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken
11782     * [199]6981: wrong code in 64-bit manipulation on x86
11783     * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__
11784       macro
11785     * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE
11786       intrinsics are broken
11787     * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with
11788       -march=pentium4
11789     * [203]7693: Typo in i386 mmintrin.h header
11790     * [204]7723: ICE - Pentium3 sse - GCC 3.2
11791     * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse
11792     * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3
11793
11794    PowerPC specific
11795
11796     * [207]5967: GCC bug when profiling nested functions on powerpc
11797     * [208]6984: wrong code generated with -O2, -O3, -Os for do-while
11798       loop on PowerPC
11799     * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5
11800     * [210]7130: miscompiled code for GCC-3.1 on
11801       powerpc-unknown-linux-gnu with -funroll-all-loops
11802     * [211]7133: PowerPC ICE: unrecognizable insn
11803     * [212]7380: ICE in extract_insn, at recog.c:2148
11804     * [213]8252: ICE on Altivec code with optimization turned on
11805     * [214]8451: Altivec ICE in GCC 3.2
11806
11807    HP/PA specific
11808
11809     * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa
11810
11811    SPARC specific
11812
11813     * [216]6668: when using --disable-multilib, libgcc_s.so is installed
11814       in the wrong place on sparc-solaris
11815     * [217]7151: ICE when compiling for UltraSPARC
11816     * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long
11817       double and -O1
11818     * [219]7842: [REGRESSION] SPARC code gen bug
11819
11820    ARM specific
11821
11822     * [220]7856: [arm] invalid offset in constant pool reference
11823     * [221]7967: optimization produces wrong code (ARM)
11824
11825    Alpha specific
11826
11827     * [222]7374: __builtin_fabsl broken on alpha
11828
11829    IBM s390 specific
11830
11831     * [223]7370: ICE in fixup_var_refs_1 on s390x
11832     * [224]7409: loop optimization bug on s390x-linux-gnu
11833     * [225]8232: s390x: ICE when using bcmp with int length argument
11834
11835    SCO specific
11836
11837     * [226]7623: SCO OpenServer build fails with machmode.def: undefined
11838       symbol: BITS_PER_UNIT
11839
11840    m68k/Coldfire specific
11841
11842     * [227]8314: crtbegin, crtend need to be multilib'ed for this
11843       platform
11844
11845    Documentation
11846
11847     * [228]761: Document some undocumented options
11848     * [229]5610: Fix documentation about invoking SSE instructions
11849       (-mfpmath=sse)
11850     * [230]7484: List -Wmissing-declarations as C-only option
11851     * [231]7531: -mcmodel not documented for x86-64
11852     * [232]8120: Update documentation of bad use of ##
11853     __________________________________________________________________
11854
11855GCC 3.2
11856
11857   3.2 is a small bug fix release, but there is a change to the
11858   application binary interface (ABI), hence the change to the second part
11859   of the version number.
11860
11861   The main purpose of the 3.2 release is to correct a couple of problems
11862   in the C++ ABI, with the intention of providing a stable interface
11863   going forward.  Accordingly, 3.2 is only a small change to 3.1.1.
11864
11865  Bug Fixes
11866
11867    C++
11868
11869     * [233]7320: g++ 3.2 relocation problem
11870     * [234]7470: vtable: virtual function pointers not in declaration
11871       order
11872
11873    libstdc++
11874
11875     * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t
11876     * [236]6503, [237]6642, [238]7186: Problems with comparing or
11877       subtracting various types of const and non-const iterators
11878     * [239]7216: ambiguity with basic_iostream::traits_type
11879     * [240]7220: problem with basic_istream::ignore(0,delimiter)
11880     * [241]7222: locale::operator==() doesn't work on std::locale("")
11881     * [242]7286: placement operator delete issue
11882     * [243]7442: cxxabi.h does not match the C++ ABI
11883     * [244]7445: poor performance of std::locale::classic() in
11884       multi-threaded applications
11885
11886    x86-64 specific
11887
11888     * [245]7291: off-by-one in generated inline bzero code for x86-64
11889
11890
11891    For questions related to the use of GCC, please consult these web
11892    pages and the [246]GCC manuals. If that fails, the
11893    [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11894    web pages and the development of GCC are welcome on our developer
11895    list at [248]gcc@gcc.gnu.org. All of [249]our lists have public
11896    archives.
11897
11898   Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and
11899   distribution of this entire article is permitted in any medium,
11900   provided this notice is preserved.
11901
11902   These pages are [251]maintained by the GCC team. Last modified
11903   2014-06-27[252].
11904
11905References
11906
11907   1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
11908   2. http://gcc.gnu.org/gcc-3.1/changes.html
11909   3. http://gcc.gnu.org/gcc-3.2/c++-abi.html
11910   4. https://gcc.gnu.org/PR3782
11911   5. https://gcc.gnu.org/PR6440
11912   6. https://gcc.gnu.org/PR7050
11913   7. https://gcc.gnu.org/PR7741
11914   8. https://gcc.gnu.org/PR7982
11915   9. https://gcc.gnu.org/PR8068
11916  10. https://gcc.gnu.org/PR8178
11917  11. https://gcc.gnu.org/PR8396
11918  12. https://gcc.gnu.org/PR8674
11919  13. https://gcc.gnu.org/PR9768
11920  14. https://gcc.gnu.org/PR9798
11921  15. https://gcc.gnu.org/PR9799
11922  16. https://gcc.gnu.org/PR9928
11923  17. https://gcc.gnu.org/PR10114
11924  18. https://gcc.gnu.org/PR10352
11925  19. https://gcc.gnu.org/PR10336
11926  20. https://gcc.gnu.org/PR8224
11927  21. https://gcc.gnu.org/PR8613
11928  22. https://gcc.gnu.org/PR8828
11929  23. https://gcc.gnu.org/PR9226
11930  24. https://gcc.gnu.org/PR9853
11931  25. https://gcc.gnu.org/PR9797
11932  26. https://gcc.gnu.org/PR9967
11933  27. https://gcc.gnu.org/PR10116
11934  28. https://gcc.gnu.org/PR10171
11935  29. https://gcc.gnu.org/PR10175
11936  30. https://gcc.gnu.org/PR8316
11937  31. https://gcc.gnu.org/PR9169
11938  32. https://gcc.gnu.org/PR9420
11939  33. https://gcc.gnu.org/PR9459
11940  34. https://gcc.gnu.org/PR9507
11941  35. https://gcc.gnu.org/PR9538
11942  36. https://gcc.gnu.org/PR9602
11943  37. https://gcc.gnu.org/PR9993
11944  38. https://gcc.gnu.org/PR10167
11945  39. https://gcc.gnu.org/PR9652
11946  40. https://gcc.gnu.org/PR10144
11947  41. https://gcc.gnu.org/PR8746
11948  42. https://gcc.gnu.org/PR9888
11949  43. https://gcc.gnu.org/PR9638
11950  44. https://gcc.gnu.org/PR9954
11951  45. https://gcc.gnu.org/PR7784
11952  46. https://gcc.gnu.org/PR7796
11953  47. https://gcc.gnu.org/PR8281
11954  48. https://gcc.gnu.org/PR8366
11955  49. https://gcc.gnu.org/PR8726
11956  50. https://gcc.gnu.org/PR9414
11957  51. https://gcc.gnu.org/PR10067
11958  52. https://gcc.gnu.org/PR7248
11959  53. https://gcc.gnu.org/PR8343
11960  54. https://gcc.gnu.org/PR9732
11961  55. https://gcc.gnu.org/PR10073
11962  56. https://gcc.gnu.org/PR7702
11963  57. https://gcc.gnu.org/PR9671
11964  58. https://gcc.gnu.org/PR8694
11965  59. https://gcc.gnu.org/PR9953
11966  60. https://gcc.gnu.org/PR10271
11967  61. https://gcc.gnu.org/PR6362
11968  62. https://gcc.gnu.org/PR10377
11969  63. https://gcc.gnu.org/PR6955
11970  64. https://gcc.gnu.org/PR5919
11971  65. https://gcc.gnu.org/PR7129
11972  66. https://gcc.gnu.org/PR7507
11973  67. https://gcc.gnu.org/PR7622
11974  68. https://gcc.gnu.org/PR7681
11975  69. https://gcc.gnu.org/PR9528
11976  70. https://gcc.gnu.org/PR8031
11977  71. https://gcc.gnu.org/PR8275
11978  72. https://gcc.gnu.org/PR8332
11979  73. https://gcc.gnu.org/PR8372
11980  74. https://gcc.gnu.org/PR8439
11981  75. https://gcc.gnu.org/PR8442
11982  76. https://gcc.gnu.org/PR8518
11983  77. https://gcc.gnu.org/PR8615
11984  78. https://gcc.gnu.org/PR8663
11985  79. https://gcc.gnu.org/PR8799
11986  80. https://gcc.gnu.org/PR9328
11987  81. https://gcc.gnu.org/PR9465
11988  82. https://gcc.gnu.org/PR47
11989  83. https://gcc.gnu.org/PR6745
11990  84. https://gcc.gnu.org/PR8214
11991  85. https://gcc.gnu.org/PR8493
11992  86. https://gcc.gnu.org/PR8332
11993  87. https://gcc.gnu.org/PR8503
11994  88. https://gcc.gnu.org/PR8727
11995  89. https://gcc.gnu.org/PR7445
11996  90. https://gcc.gnu.org/PR8230
11997  91. https://gcc.gnu.org/PR8399
11998  92. https://gcc.gnu.org/PR8662
11999  93. https://gcc.gnu.org/PR8707
12000  94. https://gcc.gnu.org/PR8708
12001  95. https://gcc.gnu.org/PR8790
12002  96. https://gcc.gnu.org/PR8887
12003  97. https://gcc.gnu.org/PR9076
12004  98. https://gcc.gnu.org/PR9151
12005  99. https://gcc.gnu.org/PR9168
12006 100. https://gcc.gnu.org/PR9269
12007 101. https://gcc.gnu.org/PR9322
12008 102. https://gcc.gnu.org/PR9433
12009 103. https://gcc.gnu.org/PR8032
12010 104. https://gcc.gnu.org/PR8639
12011 105. https://gcc.gnu.org/PR8794
12012 106. https://gcc.gnu.org/PR8832
12013 107. https://gcc.gnu.org/PR8988
12014 108. https://gcc.gnu.org/PR9492
12015 109. https://gcc.gnu.org/PR9267
12016 110. https://gcc.gnu.org/PR8344
12017 111. https://gcc.gnu.org/PR8524
12018 112. https://gcc.gnu.org/PR8880
12019 113. https://gcc.gnu.org/PR9090
12020 114. https://gcc.gnu.org/PR8588
12021 115. https://gcc.gnu.org/PR8599
12022 116. https://gcc.gnu.org/PR9506
12023 117. https://gcc.gnu.org/PR9484
12024 118. https://gcc.gnu.org/PR9292
12025 119. https://gcc.gnu.org/PR9293
12026 120. https://gcc.gnu.org/PR9295
12027 121. https://gcc.gnu.org/PR9296
12028 122. https://gcc.gnu.org/PR9316
12029 123. https://gcc.gnu.org/PR9493
12030 124. https://gcc.gnu.org/PR7341
12031 125. https://gcc.gnu.org/PR8947
12032 126. https://gcc.gnu.org/PR7448
12033 127. https://gcc.gnu.org/PR8882
12034 128. https://gcc.gnu.org/PR7445
12035 129. https://gcc.gnu.org/PR2521
12036 130. https://gcc.gnu.org/PR5661
12037 131. https://gcc.gnu.org/PR6419
12038 132. https://gcc.gnu.org/PR6994
12039 133. https://gcc.gnu.org/PR7150
12040 134. https://gcc.gnu.org/PR7160
12041 135. https://gcc.gnu.org/PR7228
12042 136. https://gcc.gnu.org/PR7266
12043 137. https://gcc.gnu.org/PR7353
12044 138. https://gcc.gnu.org/PR7411
12045 139. https://gcc.gnu.org/PR7478
12046 140. https://gcc.gnu.org/PR7526
12047 141. https://gcc.gnu.org/PR7721
12048 142. https://gcc.gnu.org/PR7803
12049 143. https://gcc.gnu.org/PR7754
12050 144. https://gcc.gnu.org/PR7788
12051 145. https://gcc.gnu.org/PR8031
12052 146. https://gcc.gnu.org/PR8055
12053 147. https://gcc.gnu.org/PR8067
12054 148. https://gcc.gnu.org/PR8134
12055 149. https://gcc.gnu.org/PR8149
12056 150. https://gcc.gnu.org/PR8160
12057 151. https://gcc.gnu.org/PR5607
12058 152. https://gcc.gnu.org/PR6579
12059 153. https://gcc.gnu.org/PR6803
12060 154. https://gcc.gnu.org/PR7176
12061 155. https://gcc.gnu.org/PR7188
12062 156. https://gcc.gnu.org/PR7306
12063 157. https://gcc.gnu.org/PR7461
12064 158. https://gcc.gnu.org/PR7524
12065 159. https://gcc.gnu.org/PR7584
12066 160. https://gcc.gnu.org/PR7676
12067 161. https://gcc.gnu.org/PR7679
12068 162. https://gcc.gnu.org/PR7811
12069 163. https://gcc.gnu.org/PR7961
12070 164. https://gcc.gnu.org/PR8071
12071 165. https://gcc.gnu.org/PR8127
12072 166. https://gcc.gnu.org/PR6745
12073 167. https://gcc.gnu.org/PR8096
12074 168. https://gcc.gnu.org/PR8127
12075 169. https://gcc.gnu.org/PR8218
12076 170. https://gcc.gnu.org/PR8287
12077 171. https://gcc.gnu.org/PR8347
12078 172. https://gcc.gnu.org/PR8348
12079 173. https://gcc.gnu.org/PR8391
12080 174. https://gcc.gnu.org/PR6627
12081 175. https://gcc.gnu.org/PR6631
12082 176. https://gcc.gnu.org/PR7102
12083 177. https://gcc.gnu.org/PR7120
12084 178. https://gcc.gnu.org/PR7209
12085 179. https://gcc.gnu.org/PR7515
12086 180. https://gcc.gnu.org/PR7814
12087 181. https://gcc.gnu.org/PR8467
12088 182. https://gcc.gnu.org/PR4890
12089 183. https://gcc.gnu.org/PR7357
12090 184. https://gcc.gnu.org/PR7358
12091 185. https://gcc.gnu.org/PR7602
12092 186. https://gcc.gnu.org/PR7862
12093 187. https://gcc.gnu.org/PR8190
12094 188. https://gcc.gnu.org/PR8524
12095 189. https://gcc.gnu.org/PR5351
12096 190. https://gcc.gnu.org/PR7591
12097 191. https://gcc.gnu.org/PR6845
12098 192. https://gcc.gnu.org/PR7034
12099 193. https://gcc.gnu.org/PR7124
12100 194. https://gcc.gnu.org/PR7174
12101 195. https://gcc.gnu.org/PR7134
12102 196. https://gcc.gnu.org/PR7375
12103 197. https://gcc.gnu.org/PR7390
12104 198. https://gcc.gnu.org/PR6890
12105 199. https://gcc.gnu.org/PR6981
12106 200. https://gcc.gnu.org/PR7242
12107 201. https://gcc.gnu.org/PR7396
12108 202. https://gcc.gnu.org/PR7630
12109 203. https://gcc.gnu.org/PR7693
12110 204. https://gcc.gnu.org/PR7723
12111 205. https://gcc.gnu.org/PR7951
12112 206. https://gcc.gnu.org/PR8146
12113 207. https://gcc.gnu.org/PR5967
12114 208. https://gcc.gnu.org/PR6984
12115 209. https://gcc.gnu.org/PR7114
12116 210. https://gcc.gnu.org/PR7130
12117 211. https://gcc.gnu.org/PR7133
12118 212. https://gcc.gnu.org/PR7380
12119 213. https://gcc.gnu.org/PR8252
12120 214. https://gcc.gnu.org/PR8451
12121 215. https://gcc.gnu.org/PR7250
12122 216. https://gcc.gnu.org/PR6668
12123 217. https://gcc.gnu.org/PR7151
12124 218. https://gcc.gnu.org/PR7335
12125 219. https://gcc.gnu.org/PR7842
12126 220. https://gcc.gnu.org/PR7856
12127 221. https://gcc.gnu.org/PR7967
12128 222. https://gcc.gnu.org/PR7374
12129 223. https://gcc.gnu.org/PR7370
12130 224. https://gcc.gnu.org/PR7409
12131 225. https://gcc.gnu.org/PR8232
12132 226. https://gcc.gnu.org/PR7623
12133 227. https://gcc.gnu.org/PR8314
12134 228. https://gcc.gnu.org/PR761
12135 229. https://gcc.gnu.org/PR5610
12136 230. https://gcc.gnu.org/PR7484
12137 231. https://gcc.gnu.org/PR7531
12138 232. https://gcc.gnu.org/PR8120
12139 233. https://gcc.gnu.org/PR7320
12140 234. https://gcc.gnu.org/PR7470
12141 235. https://gcc.gnu.org/PR6410
12142 236. https://gcc.gnu.org/PR6503
12143 237. https://gcc.gnu.org/PR6642
12144 238. https://gcc.gnu.org/PR7186
12145 239. https://gcc.gnu.org/PR7216
12146 240. https://gcc.gnu.org/PR7220
12147 241. https://gcc.gnu.org/PR7222
12148 242. https://gcc.gnu.org/PR7286
12149 243. https://gcc.gnu.org/PR7442
12150 244. https://gcc.gnu.org/PR7445
12151 245. https://gcc.gnu.org/PR7291
12152 246. https://gcc.gnu.org/onlinedocs/
12153 247. mailto:gcc-help@gcc.gnu.org
12154 248. mailto:gcc@gcc.gnu.org
12155 249. https://gcc.gnu.org/lists.html
12156 250. http://www.fsf.org/
12157 251. https://gcc.gnu.org/about.html
12158 252. http://validator.w3.org/check/referer
12159======================================================================
12160http://gcc.gnu.org/gcc-3.1/index.html
12161
12162                                    GCC 3.1
12163
12164   July 27, 2002
12165
12166   The [1]GNU project and the GCC developers are pleased to announce the
12167   release of GCC 3.1.1.
12168
12169   The links below still apply to GCC 3.1.1.
12170
12171   May 15, 2002
12172
12173   The [2]GNU project and the GCC developers are pleased to announce the
12174   release of GCC 3.1.
12175
12176   GCC used to stand for the GNU C Compiler, but since the compiler
12177   supports several other languages aside from C, it now stands for the
12178   GNU Compiler Collection.
12179
12180   A list of [3]successful builds is updated as new information becomes
12181   available.
12182
12183   The GCC developers would like to thank the numerous people that have
12184   contributed [4]new features, improvements, bug fixes, and other changes
12185   as well as test results to GCC. This [5]amazing group of volunteers is
12186   what makes GCC successful.
12187
12188   For additional information about GCC please refer to the [6]GCC project
12189   web site or contact the [7]GCC development mailing list.
12190
12191   To obtain GCC please use [8]our mirror sites, or our CVS server.
12192     __________________________________________________________________
12193
12194
12195    For questions related to the use of GCC, please consult these web
12196    pages and the [9]GCC manuals. If that fails, the
12197    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12198    web pages and the development of GCC are welcome on our developer
12199    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
12200    archives.
12201
12202   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
12203   distribution of this entire article is permitted in any medium,
12204   provided this notice is preserved.
12205
12206   These pages are [14]maintained by the GCC team. Last modified
12207   2014-06-28[15].
12208
12209References
12210
12211   1. http://www.gnu.org/
12212   2. http://www.gnu.org/
12213   3. http://gcc.gnu.org/gcc-3.1/buildstat.html
12214   4. http://gcc.gnu.org/gcc-3.1/changes.html
12215   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12216   6. http://gcc.gnu.org/index.html
12217   7. mailto:gcc@gcc.gnu.org
12218   8. http://gcc.gnu.org/mirrors.html
12219   9. https://gcc.gnu.org/onlinedocs/
12220  10. mailto:gcc-help@gcc.gnu.org
12221  11. mailto:gcc@gcc.gnu.org
12222  12. https://gcc.gnu.org/lists.html
12223  13. http://www.fsf.org/
12224  14. https://gcc.gnu.org/about.html
12225  15. http://validator.w3.org/check/referer
12226======================================================================
12227http://gcc.gnu.org/gcc-3.1/changes.html
12228
12229                             GCC 3.1 Release Series
12230                        Changes, New Features, and Fixes
12231
12232Additional changes in GCC 3.1.1
12233
12234     * A bug related to how structures and unions are returned has been
12235       fixed for powerpc-*-netbsd*.
12236     * An important bug in the implementation of -fprefetch-loop-arrays
12237       has been fixed. Previously the optimization prefetched random
12238       blocks of memory for most targets except for i386.
12239     * The Java compiler now compiles Java programs much faster and also
12240       works with parallel make.
12241     * Nested functions have been fixed for mips*-*-netbsd*.
12242     * Some missing floating point support routines have beed added for
12243       mips*-*-netbsd*.
12244     * This [1]message gives additional information about the bugs fixed
12245       in this release.
12246
12247Caveats
12248
12249     * The -traditional C compiler option has been deprecated and will be
12250       removed in GCC 3.3. (It remains possible to preprocess non-C code
12251       with the traditional preprocessor.)
12252     * The default debugging format for most ELF platforms (including
12253       GNU/Linux and FreeBSD; notable exception is Solaris) has changed
12254       from stabs to DWARF2. This requires GDB 5.1.1 or later.
12255
12256General Optimizer Improvements
12257
12258     * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat,
12259       and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure
12260       for profile driven optimizations.
12261       Options -fprofile-arcs and -fbranch-probabilities can now be used
12262       to improve speed of the generated code by profiling the actual
12263       program behaviour on typical runs. In the absence of profile info
12264       the compiler attempts to guess the profile statically.
12265     * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to
12266       monitor performance of the generated code.
12267       According to the SPECInt2000 results on an AMD Athlon CPU, the code
12268       generated by GCC 3.1 is 6% faster on the average (8.2% faster with
12269       profile feedback) compared to GCC 3.0. The code produced by GCC 3.0
12270       is about 2.1% faster compared to 2.95.3. Tests were done using the
12271       -O2 -march=athlon command-line options.
12272     * Alexandre Oliva, of Red Hat, has generalized the tree inlining
12273       infrastructure developed by CodeSourcery, LLC for the C++ front
12274       end, so that it is now used in the C front end too. Inlining
12275       functions as trees exposes them earlier to the compiler, giving it
12276       more opportunities for optimization.
12277     * Support for data prefetching instructions has been added to the GCC
12278       back end and several targets. A new __builtin_prefetch intrinsic is
12279       available to explicitly insert prefetch instructions and
12280       experimental support for loop array prefetching has been added (see
12281       -fprefetch-loop-array documentation).
12282     * Support for emitting debugging information for macros has been
12283       added for DWARF2. It is activated using -g3.
12284
12285New Languages and Language specific improvements
12286
12287  C/C++
12288
12289     * A few more [4]ISO C99 features.
12290     * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0.
12291     * The preprocessor's symbol table has been merged with the symbol
12292       table of the C, C++ and Objective-C front ends.
12293     * The preprocessor consumes less memory than the preprocessor in GCC
12294       3.0, often significantly so. On normal input files, it typically
12295       consumes less memory than pre-3.0 cccp-based GCC, too.
12296
12297  C++
12298
12299     * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std
12300       was a workaround to allow std compliant code to work with the
12301       non-std compliant libstdc++-v2. libstdc++-v3 is std compliant.
12302     * The C++ ABI has been fixed so that void (A::*)() const is mangled
12303       as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only
12304       affects pointer to cv-qualified member function types.
12305     * The C++ ABI has been changed to correctly handle this code:
12306    struct A {
12307      void operator delete[] (void *, size_t);
12308    };
12309
12310    struct B : public A {
12311    };
12312
12313    new B[10];
12314
12315       The amount of storage allocated for the array will be greater than
12316       it was in 3.0, in order to store the number of elements in the
12317       array, so that the correct size can be passed to operator delete[]
12318       when the array is deleted. Previously, the value passed to operator
12319       delete[] was unpredictable.
12320       This change will only affect code that declares a two-argument
12321       operator delete[] with a second parameter of type size_t in a base
12322       class, and does not override that definition in a derived class.
12323     * The C++ ABI has been changed so that:
12324    struct A {
12325      void operator delete[] (void *, size_t);
12326      void operator delete[] (void *);
12327    };
12328
12329       does not cause unnecessary storage to be allocated when an array of
12330       A objects is allocated.
12331       This change will only affect code that declares both of these forms
12332       of operator delete[], and declared the two-argument form before the
12333       one-argument form.
12334     * The C++ ABI has been changed so that when a parameter is passed by
12335       value, any cleanup for that parameter is performed in the caller,
12336       as specified by the ia64 C++ ABI, rather than the called function
12337       as before. As a result, classes with a non-trivial destructor but a
12338       trivial copy constructor will be passed and returned by invisible
12339       reference, rather than by bitwise copy as before.
12340     * G++ now supports the "named return value optimization": for code
12341       like
12342    A f () {
12343      A a;
12344      ...
12345      return a;
12346    }
12347
12348       G++ will allocate a in the return value slot, so that the return
12349       becomes a no-op. For this to work, all return statements in the
12350       function must return the same variable.
12351     * Improvements to the C++ library are listed in [5]the libstdc++-v3
12352       FAQ.
12353
12354  Objective-C
12355
12356     * Annoying linker warnings (due to incorrect code being generated)
12357       have been fixed.
12358     * If a class method cannot be found, the compiler no longer issues a
12359       warning if a corresponding instance method exists in the root
12360       class.
12361     * Forward @protocol declarations have been fixed.
12362     * Loading of categories has been fixed in certain situations (GNU run
12363       time only).
12364     * The class lookup in the run-time library has been rewritten so that
12365       class method dispatch is more than twice as fast as it used to be
12366       (GNU run time only).
12367
12368  Java
12369
12370     * libgcj now includes RMI, java.lang.ref.*, javax.naming, and
12371       javax.transaction.
12372     * Property files and other system resources can be compiled into
12373       executables which use libgcj using the new gcj --resource feature.
12374     * libgcj has been ported to more platforms. In particular there is
12375       now a mostly-functional mingw32 (Windows) target port.
12376     * JNI and CNI invocation interfaces were implemented, so gcj-compiled
12377       Java code can now be called from a C/C++ application.
12378     * gcj can now use builtin functions for certain known methods, for
12379       instance Math.cos.
12380     * gcj can now automatically remove redundant array-store checks in
12381       some common cases.
12382     * The --no-store-checks optimization option was added. This can be
12383       used to omit runtime store checks for code which is known not to
12384       throw ArrayStoreException
12385     * The following third party interface standards were added to libgcj:
12386       org.w3c.dom and org.xml.sax.
12387     * java.security has been merged with GNU Classpath. The new package
12388       is now JDK 1.2 compliant, and much more complete.
12389     * A bytecode verifier was added to the libgcj interpreter.
12390     * java.lang.Character was rewritten to comply with the Unicode 3.0
12391       standard, and improve performance.
12392     * Partial support for many more locales was added to libgcj.
12393     * Socket timeouts have been implemented.
12394     * libgcj has been merged into a single shared library. There are no
12395       longer separate shared libraries for the garbage collector and
12396       zlib.
12397     * Several performance improvements were made to gcj and libgcj:
12398          + Hash synchronization (thin locks)
12399          + A special allocation path for finalizer-free objects
12400          + Thread-local allocation
12401          + Parallel GC, and other GC tweaks
12402
12403  Fortran
12404
12405   Fortran improvements are listed in [6]the Fortran documentation.
12406
12407  Ada
12408
12409   [7]Ada Core Technologies, Inc, has contributed its GNAT Ada 95 front
12410   end and associated tools. The GNAT compiler fully implements the Ada
12411   language as defined by the ISO/IEC 8652 standard.
12412
12413   Please note that the integration of the Ada front end is still work in
12414   progress.
12415
12416New Targets and Target Specific Improvements
12417
12418     * Hans-Peter Nilsson has contributed a port to [8]MMIX, the CPU
12419       architecture used in new editions of Donald E. Knuth's The Art of
12420       Computer Programming.
12421     * [9]Axis Communications has contributed its port to the CRIS CPU
12422       architecture, used in the ETRAX system-on-a-chip series. See
12423       [10]Axis' developer site for technical information.
12424     * Alexandre Oliva, of Red Hat, has contributed a port to the
12425       [11]SuperH SH5 64-bit RISC microprocessor architecture, extending
12426       the existing SH port.
12427     * UltraSPARC is fully supported in 64-bit mode. The option -m64
12428       enables it.
12429     * For compatibility with the Sun compiler #pragma redefine_extname
12430       has been implemented on Solaris.
12431     * The x86 back end has had some noticeable work done to it.
12432          + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas
12433            Jaeger have contributed a port to the AMD x86-64 architecture.
12434            For more information on x86-64 see [12]http://www.x86-64.org.
12435          + The compiler now supports MMX, 3DNow!, SSE, and SSE2
12436            instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will
12437            enable the respective instruction sets. Intel C++ compatible
12438            MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics
12439            will be added in next major release.
12440          + Following those improvements, targets for Pentium MMX, K6-2,
12441            K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were
12442            added. Refer to the documentation on -march= and -mcpu=
12443            options for details.
12444          + For those targets that support it, -mfpmath=sse will cause the
12445            compiler to generate SSE/SSE2 instructions for floating point
12446            math instead of x87 instructions. Usually, this will lead to
12447            quicker code -- especially on the Pentium 4. Note that only
12448            scalar floating point instructions are used and GCC does not
12449            exploit SIMD features yet.
12450          + Prefetch support has been added to the Pentium III, Pentium 4,
12451            K6-2, K6-3, and Athlon series.
12452          + Code generated for floating point to integer conversions has
12453            been improved leading to better performance of many 3D
12454            applications.
12455     * The PowerPC back end has added 64-bit PowerPC GNU/Linux support.
12456     * C++ support for AIX has been improved.
12457     * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the
12458       PowerPC port supporting the AltiVec programming model (SIMD). The
12459       support, though presently useful, is experimental and is expected
12460       to stabilize for 3.2. The support is written to conform to
12461       Motorola's AltiVec specs. See -maltivec.
12462
12463Obsolete Systems
12464
12465   Support for a number of older systems has been declared obsolete in GCC
12466   3.1. Unless there is activity to revive them, the next release of GCC
12467   will have their sources permanently removed.
12468
12469   All configurations of the following processor architectures have been
12470   declared obsolete:
12471     * MIL-STD-1750A, 1750a-*-*
12472     * AMD A29k, a29k-*-*
12473     * Convex, c*-convex-*
12474     * Clipper, clipper-*-*
12475     * Elxsi, elxsi-*-*
12476     * Intel i860, i860-*-*
12477     * Sun picoJava, pj-*-* and pjl-*-*
12478     * Western Electric 32000, we32k-*-*
12479
12480   Most configurations of the following processor architectures have been
12481   declared obsolete, but we are preserving a few systems which may have
12482   active developers. It is unlikely that the remaining systems will
12483   survive much longer unless we see definite signs of port activity.
12484     * Motorola 88000 except
12485          + Generic a.out, m88k-*-aout*
12486          + Generic SVR4, m88k-*-sysv4
12487          + OpenBSD, m88k-*-openbsd*
12488     * NS32k except
12489          + NetBSD, ns32k-*-netbsd*
12490          + OpenBSD, ns32k-*-openbsd*.
12491     * ROMP except
12492          + OpenBSD, romp-*-openbsd*.
12493
12494   Finally, only some configurations of these processor architectures are
12495   being obsoleted.
12496     * Alpha:
12497          + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka
12498            alpha*-*-osf[45], are still supported.)
12499     * ARM:
12500          + RISCiX, arm-*-riscix*.
12501     * i386:
12502          + 386BSD, i?86-*-bsd*
12503          + Chorus, i?86-*-chorusos*
12504          + DG/UX, i?86-*-dgux*
12505          + FreeBSD 1.x, i?86-*-freebsd1.*
12506          + IBM AIX, i?86-*-aix*
12507          + ISC UNIX, i?86-*-isc*
12508          + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld*
12509          + NEXTstep, i?86-next-*
12510          + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose*
12511          + RTEMS/coff, i?86-*-rtemscoff*
12512          + RTEMS/go32, i?86-go32-rtems*
12513          + Sequent/BSD, i?86-sequent-bsd*
12514          + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and
12515            i?86-sequent-sysv3*
12516          + SunOS, i?86-*-sunos*
12517     * Motorola 68000:
12518          + Altos, m68[k0]*-altos-*
12519          + Apollo, m68[k0]*-apollo-*
12520          + Apple A/UX, m68[k0]*-apple-*
12521          + Bull, m68[k0]*-bull-*
12522          + Convergent, m68[k0]*-convergent-*
12523          + Generic SVR3, m68[k0]*-*-sysv3*
12524          + ISI, m68[k0]*-isi-*
12525          + LynxOS, m68[k0]*-*-lynxos*
12526          + NEXT, m68[k0]*-next-*
12527          + RTEMS/coff, m68[k0]*-*-rtemscoff*
12528          + Sony, m68[k0]*-sony-*
12529     * MIPS:
12530          + DEC Ultrix, mips-*-ultrix* and mips-dec-*
12531          + Generic BSD, mips-*-bsd*
12532          + Generic System V, mips-*-sysv*
12533          + IRIX before version 5, mips-sgi-irix[1234]*
12534          + RiscOS, mips-*-riscos*
12535          + Sony, mips-sony-*
12536          + Tandem, mips-tandem-*
12537     * SPARC:
12538          + RTEMS/a.out, sparc-*-rtemsaout*.
12539
12540Documentation improvements
12541
12542     * The old manual ("Using and Porting the GNU Compiler Collection")
12543       has been replaced by a users manual ("Using the GNU Compiler
12544       Collection") and a separate internals reference manual ("GNU
12545       Compiler Collection Internals").
12546     * More complete and much improved documentation about GCC's internal
12547       representation used by the C and C++ front ends.
12548     * Many cleanups and improvements in general.
12549
12550
12551    For questions related to the use of GCC, please consult these web
12552    pages and the [13]GCC manuals. If that fails, the
12553    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12554    web pages and the development of GCC are welcome on our developer
12555    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
12556    archives.
12557
12558   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
12559   distribution of this entire article is permitted in any medium,
12560   provided this notice is preserved.
12561
12562   These pages are [18]maintained by the GCC team. Last modified
12563   2014-06-28[19].
12564
12565References
12566
12567   1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html
12568   2. http://gcc.gnu.org/news/profiledriven.html
12569   3. http://gcc.gnu.org/benchmarks/
12570   4. http://gcc.gnu.org/c99status.html
12571   5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html
12572   6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html
12573   7. http://www.adacore.com/
12574   8. http://www-cs-faculty.stanford.edu/~knuth/mmix.html
12575   9. http://www.axis.com/
12576  10. http://developer.axis.com/
12577  11. http://www.superh.com/
12578  12. http://www.x86-64.org/
12579  13. https://gcc.gnu.org/onlinedocs/
12580  14. mailto:gcc-help@gcc.gnu.org
12581  15. mailto:gcc@gcc.gnu.org
12582  16. https://gcc.gnu.org/lists.html
12583  17. http://www.fsf.org/
12584  18. https://gcc.gnu.org/about.html
12585  19. http://validator.w3.org/check/referer
12586======================================================================
12587http://gcc.gnu.org/gcc-3.0/index.html
12588
12589                                   GCC 3.0.4
12590
12591   February 20, 2002
12592
12593   The [1]GNU project and the GCC developers are pleased to announce the
12594   release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0
12595   series.
12596
12597   GCC used to stand for the GNU C Compiler, but since the compiler
12598   supports several other languages aside from C, it now stands for the
12599   GNU Compiler Collection.
12600
12601   GCC 3.0.x has several new optimizations, new targets, new languages and
12602   many other new features, relative to GCC 2.95.x. See the [2]new
12603   features page for a more complete list.
12604
12605   A list of [3]successful builds is updated as new information becomes
12606   available.
12607
12608   The GCC developers would like to thank the numerous people that have
12609   contributed new features, test results, bug fixes, etc to GCC. This
12610   [4]amazing group of volunteers is what makes GCC successful.
12611
12612   And finally, we can't in good conscience fail to mention some
12613   [5]caveats to using GCC 3.0.x.
12614
12615   For additional information about GCC please refer to the [6]GCC project
12616   web site or contact the [7]GCC development mailing list.
12617
12618   To obtain GCC please use [8]our mirror sites, or our CVS server.
12619     __________________________________________________________________
12620
12621Previous 3.0.x Releases
12622
12623   December 20, 2001: GCC 3.0.3 has been released.
12624   October 25, 2001: GCC 3.0.2 has been released.
12625   August 20, 2001: GCC 3.0.1 has been released.
12626   June 18, 2001: GCC 3.0 has been released.
12627
12628
12629    For questions related to the use of GCC, please consult these web
12630    pages and the [9]GCC manuals. If that fails, the
12631    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12632    web pages and the development of GCC are welcome on our developer
12633    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
12634    archives.
12635
12636   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
12637   distribution of this entire article is permitted in any medium,
12638   provided this notice is preserved.
12639
12640   These pages are [14]maintained by the GCC team. Last modified
12641   2014-06-28[15].
12642
12643References
12644
12645   1. http://www.gnu.org/
12646   2. http://gcc.gnu.org/gcc-3.0/features.html
12647   3. http://gcc.gnu.org/gcc-3.0/buildstat.html
12648   4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12649   5. http://gcc.gnu.org/gcc-3.0/caveats.html
12650   6. http://gcc.gnu.org/index.html
12651   7. mailto:gcc@gcc.gnu.org
12652   8. http://gcc.gnu.org/mirrors.html
12653   9. https://gcc.gnu.org/onlinedocs/
12654  10. mailto:gcc-help@gcc.gnu.org
12655  11. mailto:gcc@gcc.gnu.org
12656  12. https://gcc.gnu.org/lists.html
12657  13. http://www.fsf.org/
12658  14. https://gcc.gnu.org/about.html
12659  15. http://validator.w3.org/check/referer
12660======================================================================
12661http://gcc.gnu.org/gcc-3.0/features.html
12662
12663                              GCC 3.0 New Features
12664
12665Additional changes in GCC 3.0.4
12666
12667     * GCC 3.0 now supports newer versions of the [1]NetBSD operating
12668       system, which use the ELF object file format, on x86 processors.
12669     * Correct debugging information is generated from functions that have
12670       lines from multiple files (e.g. yacc output).
12671     * A fix for whitespace handling in the -traditional preprocessor,
12672       which can affect Fortran.
12673     * Fixes to the exception handling runtime.
12674     * More fixes for bad code generation in C++.
12675     * A fix for shared library generation under AIX 4.3.
12676     * Documentation updates.
12677     * Port of GCC to Tensilica's Xtensa processor contributed.
12678     * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link).
12679
12680Additional changes in GCC 3.0.3
12681
12682     * A fix to correct an accidental change to the PowerPC ABI.
12683     * Fixes for bad code generation on a variety of architectures.
12684     * Improvements to the debugging information generated for C++
12685       classes.
12686     * Fixes for bad code generation in C++.
12687     * A fix to avoid crashes in the C++ demangler.
12688     * A fix to the C++ standard library to avoid buffer overflows.
12689     * Miscellaneous improvements for a variety of architectures.
12690
12691Additional changes in GCC 3.0.2
12692
12693     * Fixes for bad code generation during loop unrolling.
12694     * Fixes for bad code generation by the sibling call optimization.
12695     * Minor improvements to x86 code generation.
12696     * Implementation of function descriptors in C++ vtables for IA64.
12697     * Numerous minor bug-fixes.
12698
12699Additional changes in GCC 3.0.1
12700
12701     * C++ fixes for incorrect code-generation.
12702     * Improved cross-compiling support for the C++ standard library.
12703     * Fixes for some embedded targets that worked in GCC 2.95.3, but not
12704       in GCC 3.0.
12705     * Fixes for various exception-handling bugs.
12706     * A port to the S/390 architecture.
12707
12708General Optimizer Improvements
12709
12710     * [2]Basic block reordering pass.
12711     * New if-conversion pass with support for conditional (predicated)
12712       execution.
12713     * New tail call and sibling call elimination optimizations.
12714     * New register renaming pass.
12715     * New (experimental) [3]static single assignment (SSA) representation
12716       support.
12717     * New dead-code elimination pass implemented using the SSA
12718       representation.
12719     * [4]Global null pointer test elimination.
12720     * [5]Global code hoisting/unification.
12721     * More builtins and optimizations for stdio.h, string.h and old BSD
12722       functions, as well as for ISO C99 functions.
12723     * New builtin __builtin_expect for giving hints to the branch
12724       predictor.
12725
12726New Languages and Language specific improvements
12727
12728     * The GNU Compiler for the Java(TM) language (GCJ) is now integrated
12729       and supported, including the run-time library containing most
12730       common non-GUI Java classes, a bytecode interpreter, and the Boehm
12731       conservative garbage collector. Many bugs have been fixed. GCJ can
12732       compile Java source or Java bytecodes to either native code or Java
12733       class files, and supports native methods written in either the
12734       standard JNI or the more efficient and convenient CNI.
12735     * Here is a [6]partial list of C++ improvements, both new features
12736       and those no longer supported.
12737     * New C++ ABI. On the IA-64 platform GCC is capable of
12738       inter-operating with other IA-64 compilers.
12739     * The new ABI also significantly reduces the size of symbol and debug
12740       information.
12741     * New [7]C++ support library and many C++ bug fixes, vastly improving
12742       our conformance to the ISO C++ standard.
12743     * New [8]inliner for C++.
12744     * Rewritten C preprocessor, integrated into the C, C++ and Objective
12745       C compilers, with very many improvements including ISO C99 support
12746       and [9]improvements to dependency generation.
12747     * Support for more [10]ISO C99 features.
12748     * Many improvements to support for checking calls to format functions
12749       such as printf and scanf, including support for ISO C99 format
12750       features, extensions from the Single Unix Specification and GNU
12751       libc 2.2, checking of strfmon formats and features to assist in
12752       auditing for format string security bugs.
12753     * New warnings for C code that may have undefined semantics because
12754       of violations of sequence point rules in the C standard (such as a
12755       = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall.
12756     * Additional warning option -Wfloat-equal.
12757     * Improvements to -Wtraditional.
12758     * Fortran improvements are listed in [11]the Fortran documentation.
12759
12760New Targets and Target Specific Improvements
12761
12762     * New x86 back end, generating much improved code.
12763     * Support for a generic i386-elf target contributed.
12764     * New option to emit x86 assembly code using Intel style syntax
12765       (-mintel-syntax).
12766     * HPUX 11 support contributed.
12767     * Improved PowerPC code generation, including scheduled prologue and
12768       epilogue.
12769     * Port of GCC to Intel's IA-64 processor contributed.
12770     * Port of GCC to Motorola's MCore 210 and 340 contributed.
12771     * New unified back-end for Arm, Thumb and StrongArm contributed.
12772     * Port of GCC to Intel's XScale processor contributed.
12773     * Port of GCC to Atmel's AVR microcontrollers contributed.
12774     * Port of GCC to Mitsubishi's D30V processor contributed.
12775     * Port of GCC to Matsushita's AM33 processor (a member of the MN10300
12776       processor family) contributed.
12777     * Port of GCC to Fujitsu's FR30 processor contributed.
12778     * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed.
12779     * Port of GCC to Sun's picoJava processor core contributed.
12780
12781Documentation improvements
12782
12783     * Substantially rewritten and improved C preprocessor manual.
12784     * Many improvements to other documentation.
12785     * Manpages for gcc, cpp and gcov are now generated automatically from
12786       the master Texinfo manual, eliminating the problem of manpages
12787       being out of date. (The generated manpages are only extracts from
12788       the full manual, which is provided in Texinfo form, from which
12789       info, HTML, other formats and a printed manual can be generated.)
12790     * Generated info files are included in the release tarballs alongside
12791       their Texinfo sources, avoiding problems on some platforms with
12792       building makeinfo as part of the GCC distribution.
12793
12794Other significant improvements
12795
12796     * Garbage collection used internally by the compiler for most memory
12797       allocation instead of obstacks.
12798     * Lengauer and Tarjan algorithm used for computing dominators in the
12799       CFG. This algorithm can be significantly faster and more space
12800       efficient than our older algorithm.
12801     * gccbug script provided to assist in submitting bug reports to our
12802       bug tracking system. (Bug reports previously submitted directly to
12803       our mailing lists, for which you received no bug tracking number,
12804       should be submitted again using gccbug if you can reproduce the
12805       problem with GCC 3.0.)
12806     * The internal libgcc library is [12]built as a shared library on
12807       systems that support it.
12808     * Extensive testsuite included with GCC, with many new tests. In
12809       addition to tests for GCC bugs that have been fixed, many tests
12810       have been added for language features, compiler warnings and
12811       builtin functions.
12812     * Additional language-independent warning options -Wpacked, -Wpadded,
12813       -Wunreachable-code and -Wdisabled-optimization.
12814     * Target-independent options -falign-functions, -falign-loops and
12815       -falign-jumps.
12816
12817   Plus a great many bug fixes and almost all the [13]features found in
12818   GCC 2.95.
12819
12820
12821    For questions related to the use of GCC, please consult these web
12822    pages and the [14]GCC manuals. If that fails, the
12823    [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12824    web pages and the development of GCC are welcome on our developer
12825    list at [16]gcc@gcc.gnu.org. All of [17]our lists have public
12826    archives.
12827
12828   Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and
12829   distribution of this entire article is permitted in any medium,
12830   provided this notice is preserved.
12831
12832   These pages are [19]maintained by the GCC team. Last modified
12833   2014-06-28[20].
12834
12835References
12836
12837   1. http://www.netbsd.org/
12838   2. http://gcc.gnu.org/news/reorder.html
12839   3. http://gcc.gnu.org/news/ssa.html
12840   4. http://gcc.gnu.org/news/null.html
12841   5. http://gcc.gnu.org/news/unify.html
12842   6. http://gcc.gnu.org/gcc-3.0/c++features.html
12843   7. http://gcc.gnu.org/libstdc++/
12844   8. http://gcc.gnu.org/news/inlining.html
12845   9. http://gcc.gnu.org/news/dependencies.html
12846  10. http://gcc.gnu.org/c99status.html
12847  11. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
12848  12. http://gcc.gnu.org/gcc-3.0/libgcc.html
12849  13. http://gcc.gnu.org/gcc-2.95/features.html
12850  14. https://gcc.gnu.org/onlinedocs/
12851  15. mailto:gcc-help@gcc.gnu.org
12852  16. mailto:gcc@gcc.gnu.org
12853  17. https://gcc.gnu.org/lists.html
12854  18. http://www.fsf.org/
12855  19. https://gcc.gnu.org/about.html
12856  20. http://validator.w3.org/check/referer
12857======================================================================
12858http://gcc.gnu.org/gcc-3.0/caveats.html
12859
12860                                GCC 3.0 Caveats
12861
12862     * -fstrict-aliasing is now part of -O2 and higher optimization
12863       levels. This allows the compiler to assume the strictest aliasing
12864       rules applicable to the language being compiled. For C and C++,
12865       this activates optimizations based on the type of expressions. This
12866       optimization may thus break old, non-compliant code.
12867     * Enumerations are now properly promoted to int in function
12868       parameters and function returns. Normally this change is not
12869       visible, but when using -fshort-enums this is an ABI change.
12870     * The undocumented extension that allowed C programs to have a label
12871       at the end of a compound statement has been deprecated and may be
12872       removed in a future version. Programs that now generate a warning
12873       about this may be fixed by adding a null statement (a single
12874       semicolon) after the label.
12875     * The poorly documented extension that allowed string constants in C,
12876       C++ and Objective C to contain unescaped newlines has been
12877       deprecated and may be removed in a future version. Programs using
12878       this extension may be fixed in several ways: the bare newline may
12879       be replaced by \n, or preceded by \n\, or string concatenation may
12880       be used with the bare newline preceded by \n" and " placed at the
12881       start of the next line.
12882     * The Chill compiler is not included in GCC 3.0, because of the lack
12883       of a volunteer to convert it to use garbage collection.
12884     * Certain non-standard iostream methods from earlier versions of
12885       libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach,
12886       ostream::form, and istream::gets.
12887     * The new C++ ABI is not yet fully supported by current (as of
12888       2001-07-01) releases and development versions of GDB, or any
12889       earlier versions. There is a problem setting breakpoints by line
12890       number, and other related issues that have been fixed in GCC 3.0
12891       but not yet handled in GDB:
12892       [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
12893
12894
12895    For questions related to the use of GCC, please consult these web
12896    pages and the [2]GCC manuals. If that fails, the
12897    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12898    web pages and the development of GCC are welcome on our developer
12899    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
12900
12901   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
12902   distribution of this entire article is permitted in any medium,
12903   provided this notice is preserved.
12904
12905   These pages are [7]maintained by the GCC team. Last modified
12906   2014-06-27[8].
12907
12908References
12909
12910   1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
12911   2. https://gcc.gnu.org/onlinedocs/
12912   3. mailto:gcc-help@gcc.gnu.org
12913   4. mailto:gcc@gcc.gnu.org
12914   5. https://gcc.gnu.org/lists.html
12915   6. http://www.fsf.org/
12916   7. https://gcc.gnu.org/about.html
12917   8. http://validator.w3.org/check/referer
12918======================================================================
12919http://gcc.gnu.org/gcc-2.95/index.html
12920
12921                                    GCC 2.95
12922
12923   March 16, 2001: The GNU project and the GCC developers are pleased to
12924   announce the release of GCC version 2.95.3.
12925
12926Release History
12927
12928   GCC 2.95.3
12929          March 16, 2001
12930
12931   GCC 2.95.2
12932          October 27, 1999
12933
12934   GCC 2.95.1
12935          August 19, 1999
12936
12937   GCC 2.95
12938          July 31, 1999. This is the first release of GCC since the April
12939          1999 GCC/EGCS reunification and includes nearly a year's worth
12940          of new development and bugfixes.
12941
12942References and Acknowledgements
12943
12944   GCC used to stand for the GNU C Compiler, but since the compiler
12945   supports several other languages aside from C, it now stands for the
12946   GNU Compiler Collection.
12947
12948   The whole suite has been extensively [1]regression tested and
12949   [2]package tested. It should be reliable and suitable for widespread
12950   use.
12951
12952   The compiler has several new optimizations, new targets, new languages
12953   and other new features. See the [3]new features page for a more
12954   complete list of new features found in the GCC 2.95 releases.
12955
12956   The sources include installation instructions in both HTML and
12957   plaintext forms in the install directory in the distribution. However,
12958   the most up to date installation instructions and [4]build/test status
12959   are on the web pages. We will update those pages as new information
12960   becomes available.
12961
12962   The GCC developers would like to thank the numerous people that have
12963   contributed new features, test results, bugfixes, etc to GCC. This
12964   [5]amazing group of volunteers is what makes GCC successful.
12965
12966   And finally, we can't in good conscience fail to mention some
12967   [6]caveats to using GCC 2.95.
12968
12969   Download GCC 2.95 from one of our many [7]mirror sites.
12970
12971   For additional information about GCC please see the [8]GCC project web
12972   server or contact the [9]GCC development mailing list.
12973
12974
12975    For questions related to the use of GCC, please consult these web
12976    pages and the [10]GCC manuals. If that fails, the
12977    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12978    web pages and the development of GCC are welcome on our developer
12979    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
12980    archives.
12981
12982   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
12983   distribution of this entire article is permitted in any medium,
12984   provided this notice is preserved.
12985
12986   These pages are [15]maintained by the GCC team. Last modified
12987   2014-06-28[16].
12988
12989References
12990
12991   1. http://gcc.gnu.org/gcc-2.95/regress.html
12992   2. http://gcc.gnu.org/gcc-2.95/othertest.html
12993   3. http://gcc.gnu.org/gcc-2.95/features.html
12994   4. http://gcc.gnu.org/gcc-2.95/buildstat.html
12995   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12996   6. http://gcc.gnu.org/gcc-2.95/caveats.html
12997   7. http://gcc.gnu.org/mirrors.html
12998   8. http://gcc.gnu.org/index.html
12999   9. mailto:gcc@gcc.gnu.org
13000  10. https://gcc.gnu.org/onlinedocs/
13001  11. mailto:gcc-help@gcc.gnu.org
13002  12. mailto:gcc@gcc.gnu.org
13003  13. https://gcc.gnu.org/lists.html
13004  14. http://www.fsf.org/
13005  15. https://gcc.gnu.org/about.html
13006  16. http://validator.w3.org/check/referer
13007======================================================================
13008http://gcc.gnu.org/gcc-2.95/features.html
13009
13010                             GCC 2.95 New Features
13011
13012     * General Optimizer Improvements:
13013          + [1]Localized register spilling to improve speed and code
13014            density especially on small register class machines.
13015          + [2]Global CSE using lazy code motion algorithms.
13016          + [3]Improved global constant/copy propagation.
13017          + [4]Improved control flow graph analysis and manipulation.
13018          + [5]Local dead store elimination.
13019          + [6]Memory Load hoisting/store sinking in loops.
13020          + [7]Type based alias analysis is enabled by default. Note this
13021            feature will expose bugs in the Linux kernel. Please refer to
13022            the FAQ (as shipped with GCC 2.95) for additional information
13023            on this issue.
13024          + Major revamp of GIV detection, combination and simplification
13025            to improve loop performance.
13026          + Major improvements to register allocation and reloading.
13027     * New Languages and Language specific improvements
13028          + [8]Many C++ improvements.
13029          + [9]Many Fortran improvements.
13030          + [10]Java front-end has been integrated. [11]runtime library is
13031            available separately.
13032          + [12]ISO C99 support
13033          + [13]Chill front-end and runtime has been integrated.
13034          + Boehm garbage collector support in libobjc.
13035          + More support for various pragmas which appear in vendor
13036            include files
13037     * New Targets and Target Specific Improvements
13038          + [14]SPARC backend rewrite.
13039          + -mschedule=8000 will optimize code for PA8000 class
13040            processors; -mpa-risc-2-0 will generate code for PA2.0
13041            processors
13042          + Various micro-optimizations for the ia32 port. K6
13043            optimizations
13044          + Compiler will attempt to align doubles in the stack on the
13045            ia32 port
13046          + Alpha EV6 support
13047          + PowerPC 750
13048          + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403.
13049            -mcpu=e603e was added to do -mcpu=603e and -msoft-float.
13050          + c3x, c4x
13051          + HyperSPARC
13052          + SparcLite86x
13053          + sh4
13054          + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix,
13055            arm-linux)
13056          + vxWorks targets include support for vxWorks threads
13057          + StrongARM 110 and ARM9 support added. ARM Scheduling
13058            parameters rewritten.
13059          + Various changes to the MIPS port to avoid assembler macros,
13060            which in turn improves performance
13061          + Various performance improvements to the i960 port.
13062          + Major rewrite of ns32k port
13063     * Other significant improvements
13064          + [15]Ability to dump cfg information and display it using vcg.
13065          + The new faster scheme for fixing vendor header files is
13066            enabled by default.
13067          + Experimental internationalization support.
13068          + multibyte character support
13069          + Some compile-time speedups for pathological problems
13070          + Better support for complex types
13071     * Plus the usual mountain of bugfixes
13072     * Core compiler is based on the gcc2 development tree from Sept 30,
13073       1998, so we have all of the [16]features found in GCC 2.8.
13074
13075Additional Changes in GCC 2.95.1
13076
13077     * Generic bugfixes and improvements
13078          + Various documentation fixes related to the GCC/EGCS merger.
13079          + Fix memory management bug which could lead to spurious aborts,
13080            core dumps or random parsing errors in the compiler.
13081          + Fix a couple bugs in the dwarf1 and dwarf2 debug record
13082            support.
13083          + Fix infinite loop in the CSE optimizer.
13084          + Avoid undefined behavior in compiler FP emulation code
13085          + Fix install problem when prefix is overridden on the make
13086            install command.
13087          + Fix problem with unwanted installation of assert.h on some
13088            systems.
13089          + Fix problem with finding the wrong assembler in a single tree
13090            build.
13091          + Avoid increasing the known alignment of a register that is
13092            already known to be a pointer.
13093     * Platform specific bugfixes and improvements
13094          + Codegen bugfix for prologue/epilogue for cpu32 target.
13095          + Fix long long code generation bug for the Coldfire target.
13096          + Fix various aborts in the SH compiler.
13097          + Fix bugs in libgcc support library for the SH.
13098          + Fix alpha ev6 code generation bug.
13099          + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on
13100            AIX platforms.
13101          + Fix -fpic code generation bug for rs6000/ppc svr4 targets.
13102          + Fix varargs/stdarg code generation bug for rs6000/ppc svr4
13103            targets.
13104          + Fix weak symbol handling for rs6000/ppc svr4 targets.
13105          + Fix various problems with 64bit code generation for the
13106            rs6000/ppc port.
13107          + Fix codegen bug which caused tetex to be mis-compiled on the
13108            x86.
13109          + Fix compiler abort in new cfg code exposed by x86 port.
13110          + Fix out of range array reference in code convert flat
13111            registers to the x87 stacked FP register file.
13112          + Fix minor vxworks configuration bug.
13113          + Fix return type of bsearch for SunOS 4.x.
13114     * Language & Runtime specific fixes.
13115          + The G++ signature extension has been deprecated. It will be
13116            removed in the next major release of G++. Use of signatures
13117            will result in a warning from the compiler.
13118          + Several bugs relating to templates and namespaces were fixed.
13119          + A bug that caused crashes when combining templates with -g on
13120            DWARF1 platforms was fixed.
13121          + Pointers-to-members, virtual functions, and multiple
13122            inheritance should now work together correctly.
13123          + Some code-generation bugs relating to function try blocks were
13124            fixed.
13125          + G++ is a little bit more lenient with certain archaic
13126            constructs than in GCC 2.95.
13127          + Fix to prevent shared library version #s from bring truncated
13128            to 1 digit
13129          + Fix missing std:: in the libstdc++ library.
13130          + Fix stream locking problems in libio.
13131          + Fix problem in java compiler driver.
13132
13133Additional Changes in GCC 2.95.2
13134
13135   The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While
13136   the optimizations performed by -fstrict-aliasing are valid according to
13137   the C and C++ standards, the optimization have caused some problems,
13138   particularly with old non-conforming code.
13139
13140   The GCC developers are experimenting with ways to warn users about code
13141   which violates the C/C++ standards, but those warnings are not ready
13142   for widespread use at this time. Rather than wait for those warnings
13143   the GCC developers have chosen to disable -fstrict-aliasing by default
13144   for the GCC 2.95.2 release.
13145
13146   We strongly encourage developers to find and fix code which violates
13147   the C/C++ standards as -fstrict-aliasing may be enabled by default in
13148   future releases. Use the option -fstrict-aliasing to re-enable these
13149   optimizations.
13150     * Generic bugfixes and improvements
13151          + Fix incorrectly optimized memory reference in global common
13152            subexpression elimination (GCSE) optimization pass.
13153          + Fix code generation bug in regmove.c in which it could
13154            incorrectly change a "const" value.
13155          + Fix bug in optimization of conditionals involving volatile
13156            memory references.
13157          + Avoid over-allocation of stack space for some procedures.
13158          + Fixed bug in the compiler which caused incorrect optimization
13159            of an obscure series of bit manipulations, shifts and
13160            arithmetic.
13161          + Fixed register allocator bug which caused teTeX to be
13162            mis-compiled on SPARC targets.
13163          + Avoid incorrect optimization of degenerate case statements for
13164            certain targets such as the ARM.
13165          + Fix out of range memory reference in the jump optimizer.
13166          + Avoid dereferencing null pointer in fix-header.
13167          + Fix test for GCC specific features so that it is possible to
13168            bootstrap with gcc-2.6.2 and older versions of GCC.
13169          + Fix typo in scheduler which could potentially cause out of
13170            range memory accesses.
13171          + Avoid incorrect loop reversal which caused incorrect code for
13172            certain loops on PowerPC targets.
13173          + Avoid incorrect optimization of switch statements on certain
13174            targets (for example the ARM).
13175     * Platform specific bugfixes and improvements
13176          + Work around bug in Sun V5.0 compilers which caused bootstrap
13177            comparison failures on SPARC targets.
13178          + Fix SPARC backend bug which caused aborts in final.c.
13179          + Fix sparc-hal-solaris2* configuration fragments.
13180          + Fix bug in sparc block profiling.
13181          + Fix obscure code generation bug for the PARISC targets.
13182          + Define __STDC_EXT__ for HPUX configurations.
13183          + Various POWERPC64 code generation bugfixes.
13184          + Fix abort for PPC targets using ELF (ex GNU/Linux).
13185          + Fix collect2 problems for AIX targets.
13186          + Correct handling of .file directive for PPC targets.
13187          + Fix bug in fix_trunc x86 patterns.
13188          + Fix x86 port to correctly pop the FP stack for functions that
13189            return structures in memory.
13190          + Fix minor bug in strlen x86 pattern.
13191          + Use stabs debugging instead of dwarf1 for x86-solaris targets.
13192          + Fix template repository code to handle leading underscore in
13193            mangled names.
13194          + Fix weak/weak alias support for OpenBSD.
13195          + GNU/Linux for the ARM has C++ compatible include files.
13196     * Language & Runtime specific fixes.
13197          + Fix handling of constructor attribute in the C front-end which
13198            caused problems building the Chill runtime library on some
13199            targets.
13200          + Fix minor problem merging type qualifiers in the C front-end.
13201          + Fix aliasing bug for pointers and references (C/C++).
13202          + Fix incorrect "non-constant initializer bug" when -traditional
13203            or -fwritable-strings is enabled.
13204          + Fix build error for Chill front-end on SunOS.
13205          + Do not complain about duplicate instantiations when using
13206            -frepo (C++).
13207          + Fix array bounds handling in C++ front-end which caused
13208            problems with dwarf debugging information in some
13209            circumstances.
13210          + Fix minor namespace problem.
13211          + Fix problem linking java programs.
13212
13213Additional Changes in GCC 2.95.3
13214
13215     * Generic bugfixes and improvements
13216          + Fix numerous problems that caused incorrect optimization in
13217            the register reloading code.
13218          + Fix numerous problems that caused incorrect optimization in
13219            the loop optimizer.
13220          + Fix aborts in the functions build_insn_chain and scan_loops
13221            under some circumstances.
13222          + Fix an alias analysis bug.
13223          + Fix an infinite compilation bug in the combiner.
13224          + A few problems with complex number support have been fixed.
13225          + It is no longer possible for gcc to act as a fork bomb when
13226            installed incorrectly.
13227          + The -fpack-struct option should be recognized now.
13228          + Fixed a bug that caused incorrect code to be generated due to
13229            a lost stack adjustment.
13230     * Platform specific bugfixes and improvements
13231          + Support building ARM toolchains hosted on Windows.
13232          + Fix attribute calculations in ARM toolchains.
13233          + arm-linux support has been improved.
13234          + Fix a PIC failure on sparc targets.
13235          + On ix86 targets, the regparm attribute should now work
13236            reliably.
13237          + Several updates for the h8300 port.
13238          + Fix problem building libio with glibc 2.2.
13239
13240
13241    For questions related to the use of GCC, please consult these web
13242    pages and the [17]GCC manuals. If that fails, the
13243    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13244    web pages and the development of GCC are welcome on our developer
13245    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
13246    archives.
13247
13248   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
13249   distribution of this entire article is permitted in any medium,
13250   provided this notice is preserved.
13251
13252   These pages are [22]maintained by the GCC team. Last modified
13253   2014-06-28[23].
13254
13255References
13256
13257   1. http://gcc.gnu.org/news/spill.html
13258   2. http://gcc.gnu.org/news/lcm.html
13259   3. http://gcc.gnu.org/news/cprop.html
13260   4. http://gcc.gnu.org/news/cfg.html
13261   5. http://gcc.gnu.org/news/dse.html
13262   6. http://gcc.gnu.org/news/hoist.html
13263   7. http://gcc.gnu.org/news/alias.html
13264   8. http://gcc.gnu.org/gcc-2.95/c++features.html
13265   9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
13266  10. http://gcc.gnu.org/java/gcj-announce.txt
13267  11. http://gcc.gnu.org/news/javaannounce.html
13268  12. http://gcc.gnu.org/c99status.html
13269  13. http://gcc.gnu.org/news/chill.html
13270  14. http://gcc.gnu.org/news/sparc.html
13271  15. http://gcc.gnu.org/news/egcs-vcg.html
13272  16. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13273  17. https://gcc.gnu.org/onlinedocs/
13274  18. mailto:gcc-help@gcc.gnu.org
13275  19. mailto:gcc@gcc.gnu.org
13276  20. https://gcc.gnu.org/lists.html
13277  21. http://www.fsf.org/
13278  22. https://gcc.gnu.org/about.html
13279  23. http://validator.w3.org/check/referer
13280======================================================================
13281http://gcc.gnu.org/gcc-2.95/caveats.html
13282
13283                                GCC 2.95 Caveats
13284
13285     * GCC 2.95 will issue an error for invalid asm statements that had
13286       been silently accepted by earlier versions of the compiler. This is
13287       particularly noticeable when compiling older versions of the Linux
13288       kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95)
13289       for more information on this issue.
13290     * GCC 2.95 implements type based alias analysis to disambiguate
13291       memory references. Some programs, particularly the Linux kernel
13292       violate ANSI/ISO aliasing rules and therefore may not operate
13293       correctly when compiled with GCC 2.95. Please refer to the FAQ (as
13294       shipped with GCC 2.95) for more information on this issue.
13295     * GCC 2.95 has a known bug in its handling of complex variables for
13296       64bit targets. Instead of silently generating incorrect code, GCC
13297       2.95 will issue a fatal error for situations it can not handle.
13298       This primarily affects the Fortran community as Fortran makes more
13299       use of complex variables than C or C++.
13300     * GCC 2.95 has an integrated libstdc++, but does not have an
13301       integrated libg++. Furthermore old libg++ releases will not work
13302       with GCC 2.95. You can retrieve a recent copy of libg++ from the
13303       [1]GCC ftp server.
13304       Note most C++ programs only need libstdc++.
13305     * Exception handling may not work with shared libraries, particularly
13306       on alphas, hppas, rs6000/powerpc and mips based platforms.
13307       Exception handling is known to work on x86 GNU/Linux platforms with
13308       shared libraries.
13309     * In general, GCC 2.95 is more rigorous about rejecting invalid C++
13310       code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0,
13311       or EGCS 1.1. As a result it may be necessary to fix C++ code before
13312       it will compile with GCC 2.95.
13313     * G++ is also converting toward the ISO C++ standard; as a result
13314       code which was previously valid (and thus accepted by other
13315       compilers and older versions of g++) may no longer be accepted. The
13316       flag -fpermissive may allow some non-conforming code to compile
13317       with GCC 2.95.
13318     * GCC 2.95 compiled C++ code is not binary compatible with EGCS
13319       1.1.x, EGCS 1.0.x or GCC 2.8.x.
13320     * GCC 2.95 does not have changes from the GCC 2.8 tree that were made
13321       between Sept 30, 1998 and April 30, 1999 (the official end of the
13322       GCC 2.8 project). Future GCC releases will include all the changes
13323       from the defunct GCC 2.8 sources.
13324
13325
13326    For questions related to the use of GCC, please consult these web
13327    pages and the [2]GCC manuals. If that fails, the
13328    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13329    web pages and the development of GCC are welcome on our developer
13330    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
13331
13332   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
13333   distribution of this entire article is permitted in any medium,
13334   provided this notice is preserved.
13335
13336   These pages are [7]maintained by the GCC team. Last modified
13337   2014-06-12[8].
13338
13339References
13340
13341   1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz
13342   2. https://gcc.gnu.org/onlinedocs/
13343   3. mailto:gcc-help@gcc.gnu.org
13344   4. mailto:gcc@gcc.gnu.org
13345   5. https://gcc.gnu.org/lists.html
13346   6. http://www.fsf.org/
13347   7. https://gcc.gnu.org/about.html
13348   8. http://validator.w3.org/check/referer
13349======================================================================
13350http://gcc.gnu.org/egcs-1.1/index.html
13351
13352                                    EGCS 1.1
13353
13354   September 3, 1998: We are pleased to announce the release of EGCS 1.1.
13355   December 1, 1998: We are pleased to announce the release of EGCS 1.1.1.
13356   March 15, 1999: We are pleased to announce the release of EGCS 1.1.2.
13357
13358   EGCS is a free software project to further the development of the GNU
13359   compilers using an open development environment.
13360
13361   EGCS 1.1 is a major new release of the EGCS compiler system. It has
13362   been [1]extensively tested and is believed to be stable and suitable
13363   for widespread use.
13364
13365   EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8
13366   development sources; it contains all of the new features found in GCC
13367   2.8.1 as well as all new development from GCC up to June 6, 1998.
13368
13369   EGCS 1.1 also contains many improvements and features not found in GCC
13370   or in older versions of EGCS:
13371     * Global common subexpression elimination and global constant/copy
13372       propagation (aka [2]gcse)
13373     * Ongoing improvements to the [3]alias analysis support to allow for
13374       better optimizations throughout the compiler.
13375     * Vastly improved [4]C++ compiler and integrated C++ runtime
13376       libraries.
13377     * Fixes for the /tmp symlink race security problems.
13378     * New targets including mips16, arm-thumb and 64 bit PowerPC.
13379     * Improvements to GNU Fortran (g77) compiler and runtime library made
13380       since g77 version 0.5.23.
13381
13382   See the [5]new features page for a more complete list of new features
13383   found in EGCS 1.1 releases.
13384
13385   EGCS 1.1.1 is a minor update to fix several serious problems in EGCS
13386   1.1:
13387     * General improvements and fixes
13388          + Avoid some stack overflows when compiling large functions.
13389          + Avoid incorrect loop invariant code motions.
13390          + Fix some core dumps on Linux kernel code.
13391          + Bring back the imake -Di386 and friends fix from EGCS 1.0.2.
13392          + Fix code generation problem in gcse.
13393          + Various documentation related fixes.
13394     * g++/libstdc++ improvements and fixes
13395          + MT safe EH fix for setjmp/longjmp based exception handling.
13396          + Fix a few bad interactions between optimization and exception
13397            handling.
13398          + Fixes for demangling of template names starting with "__".
13399          + Fix a bug that would fail to run destructors in some cases
13400            with -O2.
13401          + Fix 'new' of classes with virtual bases.
13402          + Fix crash building Qt on the Alpha.
13403          + Fix failure compiling WIFEXITED macro on GNU/Linux.
13404          + Fix some -frepo failures.
13405     * g77 and libf2c improvements and fixes
13406          + Various documentation fixes.
13407          + Avoid compiler crash on RAND intrinsic.
13408          + Fix minor bugs in makefiles exposed by BSD make programs.
13409          + Define _XOPEN_SOURCE for libI77 build to avoid potential
13410            problems on some 64-bit systems.
13411          + Fix problem with implicit endfile on rewind.
13412          + Fix spurious recursive I/O errors.
13413     * platform specific improvements and fixes
13414          + Match all versions of UnixWare7.
13415          + Do not assume x86 SVR4 or UnixWare targets can handle stabs.
13416          + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion
13417            from unsigned ints to double precision floats.
13418          + Fix ARM ABI issue with NetBSD.
13419          + Fix a few arm code generation bugs.
13420          + Fixincludes will fix additional broken SCO OpenServer header
13421            files.
13422          + Fix a m68k backend bug which caused invalid offsets in reg+d
13423            addresses.
13424          + Fix problems with 64bit AIX 4.3 support.
13425          + Fix handling of long longs for varargs/stdarg functions on the
13426            ppc.
13427          + Minor fixes to CPP predefines for Windows.
13428          + Fix code generation problems with gpr<->fpr copies for 64bit
13429            ppc.
13430          + Fix a few coldfire code generation bugs.
13431          + Fix some more header file problems on SunOS 4.x.
13432          + Fix assert.h handling for RTEMS.
13433          + Fix Windows handling of TREE_SYMBOL_REFERENCED.
13434          + Fix x86 compiler abort in reg-stack pass.
13435          + Fix cygwin/windows problem with section attributes.
13436          + Fix Alpha code generation problem exposed by SMP Linux
13437            kernels.
13438          + Fix typo in m68k 32->64bit integer conversion.
13439          + Make sure target libraries build with -fPIC for PPC & Alpha
13440            targets.
13441
13442   EGCS 1.1.2 is a minor update to fix several serious problems in EGCS
13443   1.1.1:
13444     * General improvements and fixes
13445          + Fix bug in loop optimizer which caused the SPARC (and
13446            potentially other) ports to segfault.
13447          + Fix infinite recursion in alias analysis and combiner code.
13448          + Fix bug in regclass preferencing.
13449          + Fix incorrect loop reversal which caused incorrect code to be
13450            generated for several targets.
13451          + Fix return value for builtin memcpy.
13452          + Reduce compile time for certain loops which exposed quadratic
13453            behavior in the loop optimizer.
13454          + Fix bug which caused volatile memory to be written multiple
13455            times when only one write was needed/desired.
13456          + Fix compiler abort in caller-save.c
13457          + Fix combiner bug which caused incorrect code generation for
13458            certain division by constant operations.
13459          + Fix incorrect code generation due to a bug in range check
13460            optimizations.
13461          + Fix incorrect code generation due to mis-handling of clobbered
13462            values in CSE.
13463          + Fix compiler abort/segfault due to incorrect register
13464            splitting when unrolling loops.
13465          + Fix code generation involving autoincremented addresses with
13466            ternary operators.
13467          + Work around bug in the scheduler which caused qt to be
13468            mis-compiled on some platforms.
13469          + Fix code generation problems with -fshort-enums.
13470          + Tighten security for temporary files.
13471          + Improve compile time for codes which make heavy use of
13472            overloaded functions.
13473          + Fix multiply defined constructor/destructor symbol problems.
13474          + Avoid setting bogus RPATH environment variable during
13475            bootstrap.
13476          + Avoid GNU-make dependencies in the texinfo subdir.
13477          + Install CPP wrapper script in $(prefix)/bin if --enable-cpp.
13478            --enable-cpp=<dirname> can be used to specify an additional
13479            install directory for the cpp wrapper script.
13480          + Fix CSE bug which caused incorrect label-label refs to appear
13481            on some platforms.
13482          + Avoid linking in EH routines from libgcc if they are not
13483            needed.
13484          + Avoid obscure bug in aliasing code.
13485          + Fix bug in weak symbol handling.
13486     * Platform-specific improvements and fixes
13487          + Fix detection of PPro/PII on Unixware 7.
13488          + Fix compiler segfault when building spec99 and other programs
13489            for SPARC targets.
13490          + Fix code-generation bugs for integer and floating point
13491            conditional move instructions on the PPro/PII.
13492          + Use fixincludes to fix byteorder problems on i?86-*-sysv.
13493          + Fix build failure for the arc port.
13494          + Fix floating point format configuration for i?86-gnu port.
13495          + Fix problems with hppa1.0-hp-hpux10.20 configuration when
13496            threads are enabled.
13497          + Fix coldfire code generation bugs.
13498          + Fix "unrecognized insn" problems for Alpha and PPC ports.
13499          + Fix h8/300 code generation problem with floating point values
13500            in memory.
13501          + Fix unrecognized insn problems for the m68k port.
13502          + Fix namespace-pollution problem for the x86 port.
13503          + Fix problems with old assembler on x86 NeXT systems.
13504          + Fix PIC code-generation problems for the SPARC port.
13505          + Fix minor bug with LONG_CALLS in PowerPC SVR4 support.
13506          + Fix minor ISO namespace violation in Alpha varargs/stdarg
13507            support.
13508          + Fix incorrect "braf" instruction usage for the SH port.
13509          + Fix minor bug in va-sh which prevented its use with -ansi.
13510          + Fix problems recognizing and supporting FreeBSD.
13511          + Handle OpenBSD systems correctly.
13512          + Minor fixincludes fix for Digital UNIX 4.0B.
13513          + Fix problems with ctors/dtors in SCO shared libraries.
13514          + Abort instead of generating incorrect code for PPro/PII
13515            floating point conditional moves.
13516          + Avoid multiply defined symbols on GNU/Linux systems using
13517            libc-5.4.xx.
13518          + Fix abort in alpha compiler.
13519     * Fortran-specific fixes
13520          + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year
13521            is in the documented, non-Y2K-compliant range of 0-99, instead
13522            of being returned as 100 in the year 2000.
13523          + Fix the `Date_and_Time' intrinsic (in libg2c) to return the
13524            milliseconds value properly in Values(8).
13525          + Fix the `LStat' intrinsic (in libg2c) to return device-ID
13526            information properly in SArray(7).
13527
13528   Each release includes installation instructions in both HTML and
13529   plaintext forms (see the INSTALL directory in the toplevel directory of
13530   the distribution). However, we also keep the most up to date
13531   installation instructions and [6]build/test status on our web page. We
13532   will update those pages as new information becomes available.
13533
13534   The EGCS project would like to thank the numerous people that have
13535   contributed new features, test results, bugfixes, etc. This [7]amazing
13536   group of volunteers is what makes EGCS successful.
13537
13538   And finally, we can't in good conscience fail to mention some
13539   [8]caveats to using EGCS 1.1.
13540
13541   Download EGCS from egcs.cygnus.com (USA California).
13542
13543   The EGCS 1.1 release is also available on many mirror sites.
13544   [9]Goto mirror list to find a closer site.
13545
13546
13547    For questions related to the use of GCC, please consult these web
13548    pages and the [10]GCC manuals. If that fails, the
13549    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13550    web pages and the development of GCC are welcome on our developer
13551    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
13552    archives.
13553
13554   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
13555   distribution of this entire article is permitted in any medium,
13556   provided this notice is preserved.
13557
13558   These pages are [15]maintained by the GCC team. Last modified
13559   2014-06-28[16].
13560
13561References
13562
13563   1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html
13564   2. http://gcc.gnu.org/news/gcse.html
13565   3. http://gcc.gnu.org/news/alias.html
13566   4. http://gcc.gnu.org/egcs-1.1/c++features.html
13567   5. http://gcc.gnu.org/egcs-1.1/features.html
13568   6. http://gcc.gnu.org/egcs-1.1/buildstat.html
13569   7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13570   8. http://gcc.gnu.org/egcs-1.1/caveats.html
13571   9. http://gcc.gnu.org/mirrors.html
13572  10. https://gcc.gnu.org/onlinedocs/
13573  11. mailto:gcc-help@gcc.gnu.org
13574  12. mailto:gcc@gcc.gnu.org
13575  13. https://gcc.gnu.org/lists.html
13576  14. http://www.fsf.org/
13577  15. https://gcc.gnu.org/about.html
13578  16. http://validator.w3.org/check/referer
13579======================================================================
13580http://gcc.gnu.org/egcs-1.1/features.html
13581
13582                             EGCS 1.1 new features
13583
13584     * Integrated GNU Fortran (g77) compiler and runtime library with
13585       improvements, based on g77 version 0.5.23.
13586     * Vast improvements in the C++ compiler; so many they have [1]page of
13587       their own!
13588     * Compiler implements [2]global common subexpression elimination and
13589       global copy/constant propagation.
13590     * More major improvements in the [3]alias analysis code.
13591     * More major improvements in the exception handling code to improve
13592       performance, lower static overhead and provide the infrastructure
13593       for future improvements.
13594     * The infamous /tmp symlink race security problems have been fixed.
13595     * The regmove optimization pass has been nearly completely rewritten
13596       to improve performance of generated code.
13597     * The compiler now recomputes register usage information before local
13598       register allocation. By providing more accurate information to the
13599       priority based allocator, we get better register allocation.
13600     * The register reloading phase of the compiler optimizes spill code
13601       much better than in previous releases.
13602     * Some bad interactions between the register allocator and
13603       instruction scheduler have been fixed, resulting in much better
13604       code for certain programs. Additionally, we have tuned the
13605       scheduler in various ways to improve performance of generated code
13606       for some architectures.
13607     * The compiler's branch shortening algorithms have been significantly
13608       improved to work better on targets which align jump targets.
13609     * The compiler now supports -Os to prefer optimizing for code space
13610       over optimizing for code speed.
13611     * The compiler will now totally eliminate library calls which compute
13612       constant values. This primarily helps targets with no integer
13613       div/mul support and targets without floating point support.
13614     * The compiler now supports an extensive "--help" option.
13615     * cpplib has been greatly improved and may be suitable for limited
13616       use.
13617     * Memory footprint for the compiler has been significantly reduced
13618       for some pathological cases.
13619     * The time to build EGCS has been improved for certain targets
13620       (particularly the alpha and mips platforms).
13621     * Many infrastructure improvements throughout the compiler, plus the
13622       usual mountain of bugfixes and minor improvements.
13623     * Target dependent improvements:
13624          + SPARC port now includes V8 plus and V9 support as well as
13625            performance tuning for Ultra class machines. The SPARC port
13626            now uses the Haifa scheduler.
13627          + Alpha port has been tuned for the EV6 processor and has an
13628            optimized expansion of memcpy/bzero. The Alpha port now uses
13629            the Haifa scheduler.
13630          + RS6000/PowerPC: support for the Power64 architecture and AIX
13631            4.3. The RS6000/PowerPC port now uses the Haifa scheduler.
13632          + x86: Alignment of static store data and jump targets is per
13633            Intel recommendations now. Various improvements throughout the
13634            x86 port to improve performance on Pentium processors
13635            (including improved epilogue sequences for Pentium chips and
13636            backend improvements which should help register allocation on
13637            all x86 variants. Conditional move support has been fixed and
13638            enabled for PPro processors. The x86 port also better supports
13639            64bit operations now. Unixware 7, a System V Release 5 target,
13640            is now supported and SCO OpenServer targets can support GAS.
13641          + MIPS has improved multiply/multiply-add support and now
13642            includes mips16 ISA support.
13643          + M68k has many micro-optimizations and Coldfire fixes.
13644     * Core compiler is based on the GCC development tree from June 9,
13645       1998, so we have all of the [4]features found in GCC 2.8.
13646
13647
13648    For questions related to the use of GCC, please consult these web
13649    pages and the [5]GCC manuals. If that fails, the
13650    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13651    web pages and the development of GCC are welcome on our developer
13652    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
13653
13654   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
13655   distribution of this entire article is permitted in any medium,
13656   provided this notice is preserved.
13657
13658   These pages are [10]maintained by the GCC team. Last modified
13659   2014-06-12[11].
13660
13661References
13662
13663   1. http://gcc.gnu.org/egcs-1.1/c++features.html
13664   2. http://gcc.gnu.org/news/gcse.html
13665   3. http://gcc.gnu.org/news/alias.html
13666   4. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13667   5. https://gcc.gnu.org/onlinedocs/
13668   6. mailto:gcc-help@gcc.gnu.org
13669   7. mailto:gcc@gcc.gnu.org
13670   8. https://gcc.gnu.org/lists.html
13671   9. http://www.fsf.org/
13672  10. https://gcc.gnu.org/about.html
13673  11. http://validator.w3.org/check/referer
13674======================================================================
13675http://gcc.gnu.org/egcs-1.1/caveats.html
13676
13677                                EGCS 1.1 Caveats
13678
13679     * EGCS has an integrated libstdc++, but does not have an integrated
13680       libg++. Furthermore old libg++ releases will not work with EGCS; HJ
13681       Lu has made a libg++-2.8.1.2 snapshot available which may work with
13682       EGCS.
13683       Note most C++ programs only need libstdc++.
13684     * Exception handling may not work with shared libraries, particularly
13685       on alphas, hppas, rs6000/powerpc and mips based platforms.
13686       Exception handling is known to work on x86-linux platforms with
13687       shared libraries.
13688     * Some versions of the Linux kernel have bugs which prevent them from
13689       being compiled or from running when compiled by EGCS. See the FAQ
13690       (as shipped with EGCS 1.1) for additional information.
13691     * In general, EGCS is more rigorous about rejecting invalid C++ code
13692       or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As
13693       a result it may be necessary to fix C++ code before it will compile
13694       with EGCS.
13695     * G++ is also converting toward the ISO C++ standard; as a result
13696       code which was previously valid (and thus accepted by other
13697       compilers and older versions of g++) may no longer be accepted.
13698     * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x
13699       or GCC 2.8.x due to changes necessary to support thread safe
13700       exception handling.
13701
13702
13703    For questions related to the use of GCC, please consult these web
13704    pages and the [1]GCC manuals. If that fails, the
13705    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13706    web pages and the development of GCC are welcome on our developer
13707    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
13708
13709   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
13710   distribution of this entire article is permitted in any medium,
13711   provided this notice is preserved.
13712
13713   These pages are [6]maintained by the GCC team. Last modified
13714   2014-06-12[7].
13715
13716References
13717
13718   1. https://gcc.gnu.org/onlinedocs/
13719   2. mailto:gcc-help@gcc.gnu.org
13720   3. mailto:gcc@gcc.gnu.org
13721   4. https://gcc.gnu.org/lists.html
13722   5. http://www.fsf.org/
13723   6. https://gcc.gnu.org/about.html
13724   7. http://validator.w3.org/check/referer
13725======================================================================
13726http://gcc.gnu.org/egcs-1.0/index.html
13727
13728                                    EGCS 1.0
13729
13730   December 3, 1997: We are pleased to announce the release of EGCS 1.0.
13731   January 6, 1998: We are pleased to announce the release of EGCS 1.0.1.
13732   March 16, 1998: We are pleased to announce the release of EGCS 1.0.2.
13733   May 15, 1998 We are pleased to announce the release of EGCS 1.0.3.
13734
13735   EGCS is a collaborative effort involving several groups of hackers
13736   using an open development model to accelerate development and testing
13737   of GNU compilers and runtime libraries.
13738
13739   An important goal of EGCS is to allow wide scale testing of
13740   experimental features and optimizations; therefore, EGCS contains some
13741   features and optimizations which are still under development. However,
13742   EGCS has been carefully tested and should be comparable in quality to
13743   most GCC releases.
13744
13745   EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8
13746   development sources; it contains nearly all of the new features found
13747   in GCC 2.8.
13748
13749   EGCS 1.0 also contains many improvements and features not found in GCC
13750   2.7 and even the GCC 2.8 series (which was released after the original
13751   EGCS 1.0 release).
13752     * Integrated C++ runtime libraries, including support for most major
13753       GNU/Linux systems!
13754     * The integrated libstdc++ library includes a verbatim copy of SGI's
13755       STL release.
13756     * Integrated GNU Fortran compiler.
13757     * New instruction scheduler.
13758     * New alias analysis code.
13759
13760   See the [1]new features page for a more complete list of new features.
13761
13762   EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few
13763   critical bugs and add support for Red Hat 5.0 Linux. Changes since the
13764   EGCS 1.0 release:
13765     * Add support for Red Hat 5.0 Linux and better support for Linux
13766       systems using glibc2.
13767       Many programs failed to link when compiled with EGCS 1.0 on Red Hat
13768       5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should
13769       fix these problems.
13770     * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception
13771       handling interfaces.
13772       To avoid future compatibility problems, we strongly urge anyone who
13773       is planning on distributing shared libraries that contain C++ code
13774       to upgrade to EGCS 1.0.1 first.
13775       Soon after EGCS 1.0 was released, the GCC developers made some
13776       incompatible changes in libgcc's exception handling interfaces.
13777       These changes were needed to solve problems on some platforms. This
13778       means that GCC 2.8.0, when released, will not be seamlessly
13779       compatible with shared libraries built by EGCS 1.0. The reason is
13780       that the libgcc.a in GCC 2.8.0 will not contain a function needed
13781       by the old interface.
13782       The result of this is that there may be compatibility problems with
13783       shared libraries built by EGCS 1.0 when used with GCC 2.8.0.
13784       With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface,
13785       and libgcc.a has the support routines for both the old and the new
13786       interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed,
13787       and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed).
13788       The maintainers of GCC 2.x have decided against including seamless
13789       support for the old interface in 2.8.0, since it was never
13790       "official", so to avoid future compatibility problems we recommend
13791       against distributing any shared libraries built by EGCS 1.0 that
13792       contain C++ code (upgrade to 1.0.1 and use that).
13793     * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends.
13794       The x86 changes fix code generation errors exposed when building
13795       glibc2 and the usual GNU/Linux dynamic linker (ld.so).
13796       The hppa change fixes a compiler abort when configured for use with
13797       RTEMS.
13798       The MIPS changes fix problems with the definition of LONG_MAX on
13799       newer systems, allow for command line selection of the target ABI,
13800       and fix one code generation problem.
13801       The rs6000/ppc change fixes some problems with passing structures
13802       to varargs/stdarg functions.
13803     * A few machine independent bugfixes, mostly to fix code generation
13804       errors when building Linux kernels or glibc.
13805     * Fix a few critical exception handling and template bugs in the C++
13806       compiler.
13807     * Fix Fortran namelist bug on alphas.
13808     * Fix build problems on x86-solaris systems.
13809
13810   EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several
13811   serious problems in EGCS 1.0.1.
13812     * General improvements and fixes
13813          + Memory consumption significantly reduced, especially for
13814            templates and inline functions.
13815          + Fix various problems with glibc2.1.
13816          + Fix loop optimization bug exposed by rs6000/ppc port.
13817          + Fix to avoid potential code generation problems in jump.c.
13818          + Fix some undefined symbol problems in dwarf1 debug support.
13819     * g++/libstdc++ improvements and fixes
13820          + libstdc++ in the EGCS release has been updated and should be
13821            link compatible with libstdc++-2.8.
13822          + Various fixes in libio/libstdc++ to work better on GNU/Linux
13823            systems.
13824          + Fix problems with duplicate symbols on systems that do not
13825            support weak symbols.
13826          + Memory corruption bug and undefined symbols in bastring have
13827            been fixed.
13828          + Various exception handling fixes.
13829          + Fix compiler abort for very long thunk names.
13830     * g77 improvements and fixes
13831          + Fix compiler crash for omitted bound in Fortran CASE
13832            statement.
13833          + Add missing entries to g77 lang-options.
13834          + Fix problem with -fpedantic in the g77 compiler.
13835          + Fix "backspace" problem with g77 on alphas.
13836          + Fix x86 backend problem with Fortran literals and -fpic.
13837          + Fix some of the problems with negative subscripts for g77 on
13838            alphas.
13839          + Fixes for Fortran builds on cygwin32/mingw32.
13840     * platform specific improvements and fixes
13841          + Fix long double problems on x86 (exposed by glibc).
13842          + x86 ports define i386 again to keep imake happy.
13843          + Fix exception handling support on NetBSD ports.
13844          + Several changes to collect2 to fix many problems with AIX.
13845          + Define __ELF__ for GNU/Linux on rs6000.
13846          + Fix -mcall-linux problem on GNU/Linux on rs6000.
13847          + Fix stdarg/vararg problem for GNU/Linux on rs6000.
13848          + Allow autoconf to select a proper install problem on AIX 3.1.
13849          + m68k port support includes -mcpu32 option as well as cpu32
13850            multilibs.
13851          + Fix stdarg bug for irix6.
13852          + Allow EGCS to build on irix5 without the gnu assembler.
13853          + Fix problem with static linking on sco5.
13854          + Fix bootstrap on sco5 with native compiler.
13855          + Fix for abort building newlib on H8 target.
13856          + Fix fixincludes handling of math.h on SunOS.
13857          + Minor fix for Motorola 3300 m68k systems.
13858
13859   EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few
13860   problems reported by Red Hat for builds of Red Hat 5.1.
13861     * Generic bugfixes:
13862          + Fix a typo in the libio library which resulted in incorrect
13863            behavior of istream::get.
13864          + Fix the Fortran negative array index problem.
13865          + Fix a major problem with the ObjC runtime thread support
13866            exposed by glibc2.
13867          + Reduce memory consumption of the Haifa scheduler.
13868     * Target specific bugfixes:
13869          + Fix one x86 floating point code generation bug exposed by
13870            glibc2 builds.
13871          + Fix one x86 internal compiler error exposed by glibc2 builds.
13872          + Fix profiling bugs on the Alpha.
13873          + Fix ImageMagick & emacs 20.2 build problems on the Alpha.
13874          + Fix rs6000/ppc bug when converting values from integer types
13875            to floating point types.
13876
13877   The EGCS 1.0 releases include installation instructions in both HTML
13878   and plaintext forms (see the INSTALL directory in the toplevel
13879   directory of the distribution). However, we also keep the most up to
13880   date installation instructions and [2]build/test status on our web
13881   page. We will update those pages as new information becomes available.
13882
13883   And, we can't in good conscience fail to mention some [3]caveats to
13884   using EGCS.
13885
13886   Update: Big thanks to Stanford for providing a high speed link for
13887   downloading EGCS (go.cygnus.com)!
13888
13889   Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com
13890   (USA California -- High speed link provided by Stanford).
13891
13892   The EGCS 1.0 release is also available many mirror sites.
13893   [4]Goto mirror list to find a closer site
13894
13895   We'd like to thank the numerous people that have contributed new
13896   features, test results, bugfixes, etc. Unfortunately, they're far too
13897   numerous to mention by name.
13898
13899
13900    For questions related to the use of GCC, please consult these web
13901    pages and the [5]GCC manuals. If that fails, the
13902    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13903    web pages and the development of GCC are welcome on our developer
13904    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
13905
13906   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
13907   distribution of this entire article is permitted in any medium,
13908   provided this notice is preserved.
13909
13910   These pages are [10]maintained by the GCC team. Last modified
13911   2014-06-28[11].
13912
13913References
13914
13915   1. http://gcc.gnu.org/egcs-1.0/features.html
13916   2. http://gcc.gnu.org/egcs-1.0/buildstat.html
13917   3. http://gcc.gnu.org/egcs-1.0/caveats.html
13918   4. http://gcc.gnu.org/mirrors.html
13919   5. https://gcc.gnu.org/onlinedocs/
13920   6. mailto:gcc-help@gcc.gnu.org
13921   7. mailto:gcc@gcc.gnu.org
13922   8. https://gcc.gnu.org/lists.html
13923   9. http://www.fsf.org/
13924  10. https://gcc.gnu.org/about.html
13925  11. http://validator.w3.org/check/referer
13926======================================================================
13927http://gcc.gnu.org/egcs-1.0/features.html
13928
13929                               EGCS 1.0 features
13930
13931     * Core compiler is based on the gcc2 development tree from Aug 2,
13932       1997, so we have most of the [1]features found in GCC 2.8.
13933     * Integrated GNU Fortran compiler based on g77-0.5.22-19970929.
13934     * Vast improvements in the C++ compiler; so many they have [2]page of
13935       their own!
13936     * Integrated C++ runtime libraries, including support for most major
13937       GNU/Linux systems!
13938     * New instruction scheduler from IBM Haifa which includes support for
13939       function wide instruction scheduling as well as superscalar
13940       scheduling.
13941     * Significantly improved alias analysis code.
13942     * Improved register allocation for two address machines.
13943     * Significant code generation improvements for Fortran code on
13944       Alphas.
13945     * Various optimizations from the g77 project as well as improved loop
13946       optimizations.
13947     * Dwarf2 debug format support for some targets.
13948     * egcs libstdc++ includes the SGI STL implementation without changes.
13949     * As a result of these and other changes, egcs libstc++ is not binary
13950       compatible with previous releases of libstdc++.
13951     * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO
13952       Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and
13953       1.1), Support for RTEMS on several embedded targets, Support for
13954       arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and
13955       MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc.
13956     * Integrated testsuites for gcc, g++, g77, libstdc++ and libio.
13957     * RS6000/PowerPC ports generate code which can run on all
13958       RS6000/PowerPC variants by default.
13959     * -mcpu= and -march= switches for the x86 port to allow better
13960       control over how the x86 port generates code.
13961     * Includes the template repository patch (aka repo patch); note the
13962       new template code makes repo obsolete for ELF systems using gnu-ld
13963       such as GNU/Linux.
13964     * Plus the usual assortment of bugfixes and improvements.
13965
13966
13967    For questions related to the use of GCC, please consult these web
13968    pages and the [3]GCC manuals. If that fails, the
13969    [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13970    web pages and the development of GCC are welcome on our developer
13971    list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives.
13972
13973   Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and
13974   distribution of this entire article is permitted in any medium,
13975   provided this notice is preserved.
13976
13977   These pages are [8]maintained by the GCC team. Last modified
13978   2014-06-12[9].
13979
13980References
13981
13982   1. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13983   2. http://gcc.gnu.org/egcs-1.0/c++features.html
13984   3. https://gcc.gnu.org/onlinedocs/
13985   4. mailto:gcc-help@gcc.gnu.org
13986   5. mailto:gcc@gcc.gnu.org
13987   6. https://gcc.gnu.org/lists.html
13988   7. http://www.fsf.org/
13989   8. https://gcc.gnu.org/about.html
13990   9. http://validator.w3.org/check/referer
13991======================================================================
13992http://gcc.gnu.org/egcs-1.0/caveats.html
13993
13994                                EGCS 1.0 Caveats
13995
13996     * EGCS has an integrated libstdc++, but does not have an integrated
13997       libg++. Furthermore old libg++ releases will not work with egc; HJ
13998       Lu has made a libg++-2.8.1.2 available which may work with EGCS.
13999       Note most C++ programs only need libstdc++.
14000     * Note that using -pedantic or -Wreturn-type can cause an explosion
14001       in the amount of memory needed for template-heavy C++ code, such as
14002       code that uses STL. Also note that -Wall includes -Wreturn-type, so
14003       if you use -Wall you will need to specify -Wno-return-type to turn
14004       it off.
14005     * Exception handling may not work with shared libraries, particularly
14006       on alphas, hppas, and mips based platforms. Exception handling is
14007       known to work on x86-linux platforms with shared libraries.
14008     * Some versions of the Linux kernel have bugs which prevent them from
14009       being compiled or from running when compiled by EGCS. See the FAQ
14010       (as shipped with EGCS 1.0) for additional information.
14011     * In general, EGCS is more rigorous about rejecting invalid C++ code
14012       or deprecated C++ constructs than G++ 2.7. As a result it may be
14013       necessary to fix C++ code before it will compile with EGCS.
14014     * G++ is also aggressively tracking the C++ standard; as a result
14015       code which was previously valid (and thus accepted by other
14016       compilers and older versions of G++) may no longer be accepted.
14017     * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS
14018       1.0.x and later releases should work with Red Hat Linux 5.0.
14019
14020
14021    For questions related to the use of GCC, please consult these web
14022    pages and the [1]GCC manuals. If that fails, the
14023    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14024    web pages and the development of GCC are welcome on our developer
14025    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
14026
14027   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
14028   distribution of this entire article is permitted in any medium,
14029   provided this notice is preserved.
14030
14031   These pages are [6]maintained by the GCC team. Last modified
14032   2014-06-12[7].
14033
14034References
14035
14036   1. https://gcc.gnu.org/onlinedocs/
14037   2. mailto:gcc-help@gcc.gnu.org
14038   3. mailto:gcc@gcc.gnu.org
14039   4. https://gcc.gnu.org/lists.html
14040   5. http://www.fsf.org/
14041   6. https://gcc.gnu.org/about.html
14042   7. http://validator.w3.org/check/referer
14043======================================================================
14044