xref: /netbsd-src/external/gpl3/gcc/dist/NEWS (revision b1bb3099bf4d47bbe8c7be5b78240a535263771f)
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   May 22, 2014
13
14   The [1]GNU project and the GCC developers are pleased to announce the
15   release of GCC 4.8.3.
16
17   This release is a bug-fix release, containing fixes for regressions in
18   GCC 4.8.2 relative to previous releases of GCC.
19
20Release History
21
22   GCC 4.8.3
23          May 22, 2014 ([2]changes, [3]documentation)
24
25   GCC 4.8.2
26          October 16, 2013 ([4]changes, [5]documentation)
27
28   GCC 4.8.1
29          May 31, 2013 ([6]changes, [7]documentation)
30
31   GCC 4.8.0
32          March 22, 2013 ([8]changes, [9]documentation)
33
34References and Acknowledgements
35
36   GCC used to stand for the GNU C Compiler, but since the compiler
37   supports several other languages aside from C, it now stands for the
38   GNU Compiler Collection.
39
40   A list of [10]successful builds is updated as new information becomes
41   available.
42
43   The GCC developers would like to thank the numerous people that have
44   contributed new features, improvements, bug fixes, and other changes as
45   well as test results to GCC. This [11]amazing group of volunteers is
46   what makes GCC successful.
47
48   For additional information about GCC please refer to the [12]GCC
49   project web site or contact the [13]GCC development mailing list.
50
51   To obtain GCC please use [14]our mirror sites or [15]our SVN server.
52
53
54    For questions related to the use of GCC, please consult these web
55    pages and the [16]GCC manuals. If that fails, the
56    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
57    web pages and the development of GCC are welcome on our developer
58    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
59    archives.
60
61   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
62   distribution of this entire article is permitted in any medium,
63   provided this notice is preserved.
64
65   These pages are [21]maintained by the GCC team. Last modified
66   2014-05-22[22].
67
68References
69
70   1. http://www.gnu.org/
71   2. http://gcc.gnu.org/gcc-4.8/changes.html
72   3. http://gcc.gnu.org/onlinedocs/4.8.3/
73   4. http://gcc.gnu.org/gcc-4.8/changes.html
74   5. http://gcc.gnu.org/onlinedocs/4.8.2/
75   6. http://gcc.gnu.org/gcc-4.8/changes.html
76   7. http://gcc.gnu.org/onlinedocs/4.8.1/
77   8. http://gcc.gnu.org/gcc-4.8/changes.html
78   9. http://gcc.gnu.org/onlinedocs/4.8.0/
79  10. http://gcc.gnu.org/gcc-4.8/buildstat.html
80  11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
81  12. http://gcc.gnu.org/index.html
82  13. mailto:gcc@gcc.gnu.org
83  14. http://gcc.gnu.org/mirrors.html
84  15. http://gcc.gnu.org/svn.html
85  16. http://gcc.gnu.org/onlinedocs/
86  17. mailto:gcc-help@gcc.gnu.org
87  18. mailto:gcc@gcc.gnu.org
88  19. http://gcc.gnu.org/lists.html
89  20. http://www.fsf.org/
90  21. http://gcc.gnu.org/about.html
91  22. http://validator.w3.org/check/referer
92======================================================================
93http://gcc.gnu.org/gcc-4.8/changes.html
94
95                             GCC 4.8 Release Series
96                        Changes, New Features, and Fixes
97
98Caveats
99
100   GCC now uses C++ as its implementation language. This means that to
101   build GCC from sources, you will need a C++ compiler that understands
102   C++ 2003. For more details on the rationale and specific changes,
103   please refer to the [1]C++ conversion page.
104
105   To enable the Graphite framework for loop optimizations you now need
106   CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from
107   the [2]GCC infrastructure directory. The installation manual contains
108   more information about requirements to build GCC.
109
110   GCC now uses a more aggressive analysis to derive an upper bound for
111   the number of iterations of loops using constraints imposed by language
112   standards. This may cause non-conforming programs to no longer work as
113   expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new
114   option, -fno-aggressive-loop-optimizations, was added to disable this
115   aggressive analysis. In some loops that have known constant number of
116   iterations, but undefined behavior is known to occur in the loop before
117   reaching or during the last iteration, GCC will warn about the
118   undefined behavior in the loop instead of deriving lower upper bound of
119   the number of iterations for the loop. The warning can be disabled with
120   -Wno-aggressive-loop-optimizations.
121
122   On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules
123   for the layout of vectors that could lead to wrong code being
124   generated. Vectors larger than 8 bytes in size are now by default
125   aligned to an 8-byte boundary. This is an ABI change: code that makes
126   explicit use of vector types may be incompatible with binary objects
127   built with older versions of GCC. Auto-vectorized code is not affected
128   by this change.
129
130   On AVR, support has been removed for the command-line option
131   -mshort-calls deprecated in GCC 4.7.
132
133   On AVR, the configure option --with-avrlibc supported since GCC 4.7.2
134   is turned on per default for all non-RTEMS configurations. This option
135   arranges for a better integration of [3]AVR Libc with avr-gcc. For
136   technical details, see [4]PR54461. To turn off the option in non-RTEMS
137   configurations, use --with-avrlibc=no. If the compiler is configured
138   for RTEMS, the option is always turned off.
139
140   More information on porting to GCC 4.8 from previous versions of GCC
141   can be found in the [5]porting guide for this release.
142
143General Optimizer Improvements (and Changes)
144
145     * DWARF4 is now the default when generating DWARF debug information.
146       When -g is used on a platform that uses DWARF debugging
147       information, GCC will now default to -gdwarf-4
148       -fno-debug-types-section.
149       GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
150       consumers support DWARF4 by default. Before GCC 4.8 the default
151       version used was DWARF2. To make GCC 4.8 generate an older DWARF
152       version use -g together with -gdwarf-2 or -gdwarf-3. The default
153       for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf.
154     * A new general optimization level, -Og, has been introduced. It
155       addresses the need for fast compilation and a superior debugging
156       experience while providing a reasonable level of runtime
157       performance. Overall experience for development should be better
158       than the default optimization level -O0.
159     * A new option -ftree-partial-pre was added to control the partial
160       redundancy elimination (PRE) optimization. This option is enabled
161       by default at the -O3 optimization level, and it makes PRE more
162       aggressive.
163     * The option -fconserve-space has been removed; it was no longer
164       useful on most targets since GCC supports putting variables into
165       BSS without making them common.
166     * The struct reorg and matrix reorg optimizations (command-line
167       options -fipa-struct-reorg and -fipa-matrix-reorg) have been
168       removed. They did not always work correctly, nor did they work with
169       link-time optimization (LTO), hence were only applicable to
170       programs consisting of a single translation unit.
171     * Several scalability bottle-necks have been removed from GCC's
172       optimization passes. Compilation of extremely large functions, e.g.
173       due to the use of the flatten attribute in the "Eigen" C++ linear
174       algebra templates library, is significantly faster than previous
175       releases of GCC.
176     * Link-time optimization (LTO) improvements:
177          + LTO partitioning has been rewritten for better reliability and
178            maintanibility. Several important bugs leading to link
179            failures have been fixed.
180     * Interprocedural optimization improvements:
181          + A new symbol table has been implemented. It builds on existing
182            callgraph and varpool modules and provide a new API. Unusual
183            symbol visibilities and aliases are handled more consistently
184            leading to, for example, more aggressive unreachable code
185            removal with LTO.
186          + The inline heuristic can now bypass limits on the size of of
187            inlined functions when the inlining is particularly
188            profitable. This happens, for example, when loop bounds or
189            array strides get propagated.
190          + Values passed through aggregates (either by value or
191            reference) are now propagated at the inter-procedural level
192            leading to better inlining decisions (for example in the case
193            of Fortran array descriptors) and devirtualization.
194     * [6]AddressSanitizer , a fast memory error detector, has been added
195       and can be enabled via -fsanitize=address. Memory access
196       instructions will be instrumented to detect heap-, stack-, and
197       global-buffer overflow as well as use-after-free bugs. To get nicer
198       stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is
199       available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on
200       x86-64 Darwin.
201     * [7]ThreadSanitizer has been added and can be enabled via
202       -fsanitize=thread. Instructions will be instrumented to detect data
203       races. The ThreadSanitizer is available on x86-64 GNU/Linux.
204     * A new local register allocator (LRA) has been implemented, which
205       replaces the 26 year old reload pass and improves generated code
206       quality. For now it is active on the IA-32 and x86-64 targets.
207     * Support for transactional memory has been implemented on the
208       following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and
209       Alpha.
210
211New Languages and Language specific improvements
212
213  C family
214
215     * Each diagnostic emitted now includes the original source line and a
216       caret '^' indicating the column. The option
217       -fno-diagnostics-show-caret suppresses this information.
218     * The option -ftrack-macro-expansion=2 is now enabled by default.
219       This allows the compiler to display the macro expansion stack in
220       diagnostics. Combined with the caret information, an example
221       diagnostic showing these two features is:
222
223t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float
224')
225 #define MYMAX(A,B)    __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _
226_b = (B); __a < __b ? __b : __a; })
227
228              ^
229t.c:7:7: note: in expansion of macro 'MYMAX'
230   X = MYMAX(P, F);
231       ^
232
233     * A new -Wsizeof-pointer-memaccess warning has been added (also
234       enabled by -Wall) to warn about suspicious length parameters to
235       certain string and memory built-in functions if the argument uses
236       sizeof. This warning warns e.g. about memset (ptr, 0, sizeof
237       (ptr)); if ptr is not an array, but a pointer, and suggests a
238       possible fix, or about memcpy (&foo, ptr, sizeof (&foo));.
239     * The new option -Wpedantic is an alias for -pedantic, which is now
240       deprecated. The forms -Wno-pedantic, -Werror=pedantic, and
241       -Wno-error=pedantic work in the same way as for any other -W
242       option. One caveat is that -Werror=pedantic is not equivalent to
243       -pedantic-errors, since the latter makes into errors some warnings
244       that are not controlled by -Wpedantic, and the former only affects
245       diagnostics that are disabled when using -Wno-pedantic.
246     * The option -Wshadow no longer warns if a declaration shadows a
247       function declaration, unless the former declares a function or
248       pointer to function, because this is [8]a common and valid case in
249       real-world code.
250
251  C++
252
253     * G++ now implements the [9]C++11 thread_local keyword; this differs
254       from the GNU __thread keyword primarily in that it allows dynamic
255       initialization and destruction semantics. Unfortunately, this
256       support requires a run-time penalty for references to
257       non-function-local thread_local variables defined in a different
258       translation unit even if they don't need dynamic initialization, so
259       users may want to continue to use __thread for TLS variables with
260       static initialization semantics.
261       If the programmer can be sure that no use of the variable in a
262       non-defining TU needs to trigger dynamic initialization (either
263       because the variable is statically initialized, or a use of the
264       variable in the defining TU will be executed before any uses in
265       another TU), they can avoid this overhead with the
266       -fno-extern-tls-init option.
267       OpenMP threadprivate variables now also support dynamic
268       initialization and destruction by the same mechanism.
269     * G++ now implements the [10]C++11 attribute syntax, e.g.
270
271[[noreturn]] void f();
272
273       and also the alignment specifier, e.g.
274
275alignas(double) int i;
276
277     * G++ now implements [11]C++11 inheriting constructors, e.g.
278
279struct A { A(int); };
280struct B: A { using A::A; }; // defines B::B(int)
281B b(42); // OK
282
283     * As of GCC 4.8.1, G++ implements the change to decltype semantics
284       from [12]N3276.
285
286struct A f();
287decltype(f()) g();    // OK, return type of f() is not required to be complete.
288
289     * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g.
290
291struct A { int f() &; };
292int i = A().f();  // error, f() requires an lvalue object
293
294     * G++ now supports a -std=c++1y option for experimentation with
295       features proposed for the next revision of the standard, expected
296       around 2014. Currently the only difference from -std=c++11 is
297       support for return type deduction in normal functions, as proposed
298       in [14]N3386. Status of C++1y features in GCC 4.8 can be found
299       [15]here.
300     * The G++ namespace association extension, __attribute ((strong)),
301       has been deprecated. Inline namespaces should be used instead.
302     * G++ now supports a -fext-numeric-literal option to control whether
303       GNU numeric literal suffixes are accepted as extensions or
304       processed as C++11 user-defined numeric literal suffixes. The flag
305       is on (use suffixes for GNU literals) by default for -std=gnu++*,
306       and -std=c++98. The flag is off (use suffixes for user-defined
307       literals) by default for -std=c++11 and later.
308
309    Runtime Library (libstdc++)
310
311     * [16]Improved experimental support for the new ISO C++ standard,
312       C++11, including:
313          + forward_list meets the allocator-aware container requirements;
314          + this_thread::sleep_for(), this_thread::sleep_until() and
315            this_thread::yield() are defined without requiring the
316            configure option --enable-libstdcxx-time;
317     * Improvements to <random>:
318          + SSE optimized normal_distribution.
319          + Use of hardware RNG instruction for random_device on new x86
320            processors (requires the assembler to support the
321            instruction.)
322       and <ext/random>:
323          + New random number engine simd_fast_mersenne_twister_engine
324            with an optimized SSE implementation.
325          + New random number distributions beta_distribution,
326            normal_mv_distribution, rice_distribution,
327            nakagami_distribution, pareto_distribution, k_distribution,
328            arcsine_distribution, hoyt_distribution.
329     * Added --disable-libstdcxx-verbose configure option to disable
330       diagnostic messages issued when a process terminates abnormally.
331       This may be useful for embedded systems to reduce the size of
332       executables that link statically to the library.
333
334  Fortran
335
336     * Compatibility notice:
337          + Module files: The version of module files (.mod) has been
338            incremented. Fortran MODULEs compiled by earlier GCC versions
339            have to be recompiled, when they are USEd by files compiled
340            with GCC 4.8. GCC 4.8 is not able to read .mod files created
341            by earlier versions; attempting to do so gives an error
342            message.
343            Note: The ABI of the produced assembler data itself has not
344            changed; object files and libraries are fully compatible with
345            older versions except as noted below.
346          + ABI: Some internal names (used in the assembler/object file)
347            have changed for symbols declared in the specification part of
348            a module. If an affected module - or a file using it via use
349            association - is recompiled, the module and all files which
350            directly use such symbols have to be recompiled as well. This
351            change only affects the following kind of module symbols:
352               o Procedure pointers. Note: C-interoperable function
353                 pointers (type(c_funptr)) are not affected nor are
354                 procedure-pointer components.
355               o Deferred-length character strings.
356     * The [17]BACKTRACE intrinsic subroutine has been added. It shows a
357       backtrace at an arbitrary place in user code; program execution
358       continues normally afterwards.
359     * The [18]-Wc-binding-type warning option has been added (disabled by
360       default). It warns if the a variable might not be C interoperable;
361       in particular, if the variable has been declared using an intrinsic
362       type with default kind instead of using a kind parameter defined
363       for C interoperability in the intrinsic ISO_C_Binding module.
364       Before, this warning was always printed. The -Wc-binding-type
365       option is enabled by -Wall.
366     * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line
367       options have been added, which diagnose when code to is inserted
368       for automatic (re)allocation of a variable during assignment. This
369       option can be used to decide whether it is safe to use
370       [20]-fno-realloc-lhs. Additionally, it can be used to find
371       automatic (re)allocation in hot loops. (For arrays, replacing
372       "var=" by "var(:)=" disables the automatic reallocation.)
373     * The [21]-Wcompare-reals command-line option has been added. When
374       this is set, warnings are issued when comparing REAL or COMPLEX
375       types for equality and inequality; consider replacing a == b by
376       abs(a -b) < eps with a suitable eps. -Wcompare-reals is enabled by
377       -Wextra.
378     * The [22]-Wtarget-lifetime command-line option has been added
379       (enabled with -Wall), which warns if the pointer in a pointer
380       assignment might outlive its target.
381     * Reading floating point numbers which use "q" for the exponential
382       (such as 4.0q0) is now supported as vendor extension for better
383       compatibility with old data files. It is strongly recommended to
384       use for I/O the equivalent but standard conforming "e" (such as
385       4.0e0).
386       (For Fortran source code, consider replacing the "q" in
387       floating-point literals by a kind parameter (e.g. 4.0e0_qp with a
388       suitable qp). Note that - in Fortran source code - replacing "q" by
389       a simple "e" is not equivalent.)
390     * The GFORTRAN_TMPDIR environment variable for specifying a
391       non-default directory for files opened with STATUS="SCRATCH", is
392       not used anymore. Instead gfortran checks the POSIX/GNU standard
393       TMPDIR environment variable. If TMPDIR is not defined, gfortran
394       falls back to other methods to determine the directory for
395       temporary files as documented in the [23]user manual.
396     * [24]Fortran 2003:
397          + Support for unlimited polymorphic variables (CLASS(*)) has
398            been added. Nonconstant character lengths are not yet
399            supported.
400     * [25]TS 29113:
401          + Assumed types (TYPE(*)) are now supported.
402          + Experimental support for assumed-rank arrays (dimension(..))
403            has been added. Note that currently gfortran's own array
404            descriptor is used, which is different from the one defined in
405            TS29113, see [26]gfortran's header file or use the [27]Chasm
406            Language Interoperability Tools.
407
408  Go
409
410     * GCC 4.8.2 provides a complete implementation of the Go 1.1.2
411       release.
412     * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1
413       release. The library support is not quite complete.
414     * Go has been tested on GNU/Linux and Solaris platforms for various
415       processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may
416       work on other platforms as well.
417
418New Targets and Target Specific Improvements
419
420  AArch64
421
422     * A new port has been added to support AArch64, the new 64-bit
423       architecture from ARM. Note that this is a separate port from the
424       existing 32-bit ARM port.
425     * The port provides initial support for the Cortex-A53 and the
426       Cortex-A57 processors with the command line options
427       -mcpu=cortex-a53 and -mcpu=cortex-a57.
428
429  ARM
430
431     * Initial support has been added for the AArch32 extensions defined
432       in the ARMv8 architecture.
433     * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs.
434     * A new option, -mcpu=marvell-pj4, has been added to generate code
435       for the Marvell PJ4 processor.
436     * The compiler can now automatically generate the VFMA, VFMS, REVSH
437       and REV16 instructions.
438     * A new vectorizer cost model for Advanced SIMD configurations to
439       improve the auto-vectorization strategies used.
440     * The scheduler now takes into account the number of live registers
441       to reduce the amount of spilling that can occur. This should
442       improve code performance in large functions. The limit can be
443       removed by using the option -fno-sched-pressure.
444     * Improvements have been made to the Marvell iWMMX code generation
445       and support for the iWMMX2 SIMD unit has been added. The option
446       -mcpu=iwmmxt2 can be used to enable code generation for the latter.
447     * A number of code generation improvements for Thumb2 to reduce code
448       size when compiling for the M-profile processors.
449     * The RTEMS (arm-rtems) port has been updated to use the EABI.
450     * Code generation support for the old FPA and Maverick floating-point
451       architectures has been removed. Ports that previously relied on
452       these features have also been removed. This includes the targets:
453          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
454          + arm*-*-elf (use arm*-*-eabi)
455          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
456          + arm*-*-ecos-elf (no alternative)
457          + arm*-*-freebsd (no alternative)
458          + arm*-wince-pe* (no alternative).
459
460  AVR
461
462     * Support for the "Embedded C" fixed-point has been added. For
463       details, see the [28]GCC wiki and the [29]user manual. The support
464       is not complete.
465     * A new print modifier %r for register operands in inline assembler
466       is supported. It will print the raw register number without the
467       register prefix 'r':
468    /* Return the most significant byte of 'val', a 64-bit value.  */
469
470    unsigned char msb (long long val)
471    {
472      unsigned char c;
473      __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val));
474      return c;
475    }
476       The inline assembler in this example will generate code like
477    mov r24, 8+7
478       provided c is allocated to R24 and val is allocated to R8...R15.
479       This works because the GNU assembler accepts plain register numbers
480       without register prefix.
481     * Static initializers with 3-byte symbols are supported now:
482    extern const __memx char foo;
483    const __memx void *pfoo = &foo;
484       This requires at least Binutils 2.23.
485
486  IA-32/x86-64
487
488     * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with
489       SSE extensions disabled. Since the x86-64 ABI requires 16 byte
490       stack alignment, this is ABI incompatible and intended to be used
491       in controlled environments where stack space is an important
492       limitation. This option will lead to wrong code when functions
493       compiled with 16 byte stack alignment (such as functions from a
494       standard library) are called with misaligned stack. In this case,
495       SSE instructions may lead to misaligned memory access traps. In
496       addition, variable arguments will be handled incorrectly for 16
497       byte aligned objects (including x87 long double and __int128),
498       leading to wrong results. You must build all modules with
499       -mpreferred-stack-boundary=3, including any libraries. This
500       includes the system libraries and startup modules.
501     * Support for the new Intel processor codename Broadwell with RDSEED,
502       ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw,
503       -mrdseed command-line options.
504     * Support for the Intel RTM and HLE intrinsics, built-in functions
505       and code generation is available via -mrtm and -mhle.
506     * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets.
507       Intrinsics and built-in functions are available via -mfxsr, -mxsave
508       and -mxsaveopt respectively.
509     * New -maddress-mode=[short|long] options for x32.
510       -maddress-mode=short overrides default 64-bit addresses to 32-bit
511       by emitting the 0x67 address-size override prefix. This is the
512       default address mode for x32.
513     * New built-in functions to detect run-time CPU type and ISA:
514          + A built-in function __builtin_cpu_is has been added to detect
515            if the run-time CPU is of a particular type. It returns a
516            positive integer on a match and zero otherwise. It accepts one
517            string literal argument, the CPU name. For example,
518            __builtin_cpu_is("westmere") returns a positive integer if the
519            run-time CPU is an Intel Core i7 Westmere processor. Please
520            refer to the [30]user manual for the list of valid CPU names
521            recognized.
522          + A built-in function __builtin_cpu_supports has been added to
523            detect if the run-time CPU supports a particular ISA feature.
524            It returns a positive integer on a match and zero otherwise.
525            It accepts one string literal argument, the ISA feature. For
526            example, __builtin_cpu_supports("ssse3") returns a positive
527            integer if the run-time CPU supports SSSE3 instructions.
528            Please refer to the [31]user manual for the list of valid ISA
529            names recognized.
530       Caveat: If these built-in functions are called before any static
531       constructors are invoked, like during IFUNC initialization, then
532       the CPU detection initialization must be explicitly run using this
533       newly provided built-in function, __builtin_cpu_init. The
534       initialization needs to be done only once. For example, this is how
535       the invocation would look like inside an IFUNC initializer:
536    static void (*some_ifunc_resolver(void))(void)
537    {
538      __builtin_cpu_init();
539      if (__builtin_cpu_is("amdfam10h") ...
540      if (__builtin_cpu_supports("popcnt") ...
541    }
542
543     * Function Multiversioning Support with G++:
544       It is now possible to create multiple function versions each
545       targeting a specific processor and/or ISA. Function versions have
546       the same signature but different target attributes. For example,
547       here is a program with function versions:
548    __attribute__ ((target ("default")))
549    int foo(void)
550    {
551      return 1;
552    }
553
554    __attribute__ ((target ("sse4.2")))
555    int foo(void)
556    {
557      return 2;
558    }
559
560    int main (void)
561    {
562      int (*p) = &foo;
563      assert ((*p)() == foo());
564      return 0;
565    }
566
567       Please refer to this [32]wiki for more information.
568     * The x86 back end has been improved to allow option -fschedule-insns
569       to work reliably. This option can be used to schedule instructions
570       better and leads to improved performace in certain cases.
571     * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437
572       from the Mingw-w64 trunk.
573     * Support for new AMD family 15h processors (Steamroller core) is now
574       available through the -march=bdver3 and -mtune=bdver3 options.
575     * Support for new AMD family 16h processors (Jaguar core) is now
576       available through the -march=btver2 and -mtune=btver2 options.
577
578  FRV
579
580     * This target now supports the -fstack-usage command-line option.
581
582  MIPS
583
584     * GCC can now generate code specifically for the R4700, Broadcom XLP
585       and MIPS 34kn processors. The associated -march options are
586       -march=r4700, -march=xlp and -march=34kn respectively.
587     * GCC now generates better DSP code for MIPS 74k cores thanks to
588       further scheduling optimizations.
589     * The MIPS port now supports the -fstack-check option.
590     * GCC now passes the -mmcu and -mno-mcu options to the assembler.
591     * Previous versions of GCC would silently accept -fpic and -fPIC for
592       -mno-abicalls targets like mips*-elf. This combination was not
593       intended or supported, and did not generate position-independent
594       code. GCC 4.8 now reports an error when this combination is used.
595
596  PowerPC / PowerPC64 / RS6000
597
598     * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save,
599       restore or update the VRSAVE register by default. The respective
600       operating systems manage the VRSAVE register directly.
601     * Large TOC support has been added for AIX through the command line
602       option -mcmodel=large.
603     * Native Thread-Local Storage support has been added for AIX.
604     * VMX (Altivec) and VSX instruction sets now are enabled implicitly
605       when targetting processors that support those hardware features on
606       AIX 6.1 and above.
607
608  RX
609
610     * This target will now issue a warning message whenever multiple fast
611       interrupt handlers are found in the same compilation unit. This
612       feature can be turned off by the new
613       -mno-warn-multiple-fast-interrupts command-line option.
614
615  S/390, System z
616
617     * Support for the IBM zEnterprise zEC12 processor has been added.
618       When using the -march=zEC12 option, the compiler will generate code
619       making use of the following new instructions:
620          + load and trap instructions
621          + 2 new compare and trap instructions
622          + rotate and insert selected bits - without CC clobber
623       The -mtune=zEC12 option enables zEC12 specific instruction
624       scheduling without making use of new instructions.
625     * Register pressure sensitive instruction scheduling is enabled by
626       default.
627     * The ifunc function attribute is enabled by default.
628     * memcpy and memcmp invokations on big memory chunks or with run time
629       lengths are not generated inline anymore when tuning for z10 or
630       higher. The purpose is to make use of the IFUNC optimized versions
631       in Glibc.
632
633  SH
634
635     * The default alignment settings have been reduced to be less
636       aggressive. This results in more compact code for optimization
637       levels other than -Os.
638     * Improved support for the __atomic built-in functions:
639          + A new option -matomic-model=model selects the model for the
640            generated atomic sequences. The following models are
641            supported:
642
643              soft-gusa
644                      Software gUSA sequences (SH3* and SH4* only). On
645                      SH4A targets this will now also partially utilize
646                      the movco.l and movli.l instructions. This is the
647                      default when the target is sh3*-*-linux* or
648                      sh4*-*-linux*.
649
650              hard-llcs
651                      Hardware movco.l / movli.l sequences (SH4A only).
652
653              soft-tcb
654                      Software thread control block sequences.
655
656              soft-imask
657                      Software interrupt flipping sequences (privileged
658                      mode only). This is the default when the target is
659                      sh1*-*-linux* or sh2*-*-linux*.
660
661              none
662                      Generates function calls to the respective __atomic
663                      built-in functions. This is the default for SH64
664                      targets or when the target is not sh*-*-linux*.
665
666          + The option -msoft-atomic has been deprecated. It is now an
667            alias for -matomic-model=soft-gusa.
668          + A new option -mtas makes the compiler generate the tas.b
669            instruction for the __atomic_test_and_set built-in function
670            regardless of the selected atomic model.
671          + The __sync functions in libgcc now reflect the selected atomic
672            model when building the toolchain.
673     * Added support for the mov.b and mov.w instructions with
674       displacement addressing.
675     * Added support for the SH2A instructions movu.b and movu.w.
676     * Various improvements to code generated for integer arithmetic.
677     * Improvements to conditional branches and code that involves the T
678       bit. A new option -mzdcbranch tells the compiler to favor
679       zero-displacement branches. This is enabled by default for SH4*
680       targets.
681     * The pref instruction will now be emitted by the __builtin_prefetch
682       built-in function for SH3* targets.
683     * The fmac instruction will now be emitted by the fmaf standard
684       function and the __builtin_fmaf built-in function.
685     * The -mfused-madd option has been deprecated in favor of the
686       machine-independent -ffp-contract option. Notice that the fmac
687       instruction will now be generated by default for expressions like a
688       * b + c. This is due to the compiler default setting
689       -ffp-contract=fast.
690     * Added new options -mfsrra and -mfsca to allow the compiler using
691       the fsrra and fsca instructions on targets other than SH4A (where
692       they are already enabled by default).
693     * Added support for the __builtin_bswap32 built-in function. It is
694       now expanded as a sequence of swap.b and swap.w instructions
695       instead of a library function call.
696     * The behavior of the -mieee option has been fixed and the negative
697       form -mno-ieee has been added to control the IEEE conformance of
698       floating point comparisons. By default -mieee is now enabled and
699       the option -ffinite-math-only implicitly sets -mno-ieee.
700     * Added support for the built-in functions __builtin_thread_pointer
701       and __builtin_set_thread_pointer. This assumes that GBR is used to
702       hold the thread pointer of the current thread. Memory loads and
703       stores relative to the address returned by __builtin_thread_pointer
704       will now also utilize GBR based displacement address modes.
705     * The -mdiv= option for targets other than SHmedia has been fixed and
706       documented.
707
708  SPARC
709
710     * Added optimized instruction scheduling for Niagara4.
711
712  TILE-Gx
713
714     * Added support for the -mcmodel=MODEL command-line option. The
715       models supported are small and large.
716
717  V850
718
719     * This target now supports the E3V5 architecture via the use of the
720       new -mv850e3v5 command-line option. It also has experimental
721       support for the e3v5 LOOP instruction which can be enabled via the
722       new -mloop command-line option.
723
724  XStormy16
725
726     * This target now supports the -fstack-usage command-line option.
727
728Operating Systems
729
730  Windows (Cygwin)
731
732     * Executables are now linked against shared libgcc by default. The
733       previous default was to link statically, which can still be done by
734       explicitly specifying -static or static-libgcc on the command line.
735       However it is strongly advised against, as it will cause problems
736       for any application that makes use of DLLs compiled by GCC. It
737       should be alright for a monolithic stand-alone application that
738       only links against the Windows DLLs, but offers little or no
739       benefit.
740
741GCC 4.8.1
742
743   This is the [33]list of problem reports (PRs) from GCC's bug tracking
744   system that are known to be fixed in the 4.8.1 release. This list might
745   not be complete (that is, it is possible that some PRs that have been
746   fixed are not listed here).
747
748   The C++11 <chrono> std::chrono::system_clock and
749   std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they
750   both are now separate (never typedefs of each other), both use
751   std::chrono::nanoseconds resolution, on most GNU/Linux configurations
752   std::chrono::steady_clock is now finally monotonic, and both classes
753   are mangled differently than in the previous GCC releases.
754   std::chrono::system_clock::now() with std::chrono::microseconds resp.
755   std::chrono::seconds resolution is still exported for backwards
756   compatibility with default configured libstdc++. Note that libstdc++
757   configured with --enable-libstdcxx-time= used to be ABI incompatible
758   with default configured libstdc++ for those two classes and no ABI
759   compatibility can be offered for those configurations, so any C++11
760   code that uses those classes and has been compiled and linked against
761   libstdc++ configured with the non-default --enable-libstdcxx-time=
762   configuration option needs to be recompiled.
763
764GCC 4.8.2
765
766   This is the [34]list of problem reports (PRs) from GCC's bug tracking
767   system that are known to be fixed in the 4.8.2 release. This list might
768   not be complete (that is, it is possible that some PRs that have been
769   fixed are not listed here).
770
771GCC 4.8.3
772
773   This is the [35]list of problem reports (PRs) from GCC's bug tracking
774   system that are known to be fixed in the 4.8.3 release. This list might
775   not be complete (that is, it is possible that some PRs that have been
776   fixed are not listed here).
777
778   Support for the new powerpc64le-linux platform has been added. It
779   defaults to generating code that conforms to the ELFV2 ABI.
780
781
782    For questions related to the use of GCC, please consult these web
783    pages and the [36]GCC manuals. If that fails, the
784    [37]gcc-help@gcc.gnu.org mailing list might help. Comments on these
785    web pages and the development of GCC are welcome on our developer
786    list at [38]gcc@gcc.gnu.org. All of [39]our lists have public
787    archives.
788
789   Copyright (C) [40]Free Software Foundation, Inc. Verbatim copying and
790   distribution of this entire article is permitted in any medium,
791   provided this notice is preserved.
792
793   These pages are [41]maintained by the GCC team. Last modified
794   2014-05-22[42].
795
796References
797
798   1. http://gcc.gnu.org/wiki/cxx-conversion
799   2. ftp://gcc.gnu.org/pub/gcc/infrastructure/
800   3. http://www.nongnu.org/avr-libc/
801   4. http://gcc.gnu.org/PR54461
802   5. http://gcc.gnu.org/gcc-4.8/porting_to.html
803   6. https://code.google.com/p/address-sanitizer/
804   7. https://code.google.com/p/data-race-test/wiki/ThreadSanitizer
805   8. https://lkml.org/lkml/2006/11/28/239
806   9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
807  10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
808  11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
809  12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf
810  13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
811  14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html
812  15. http://gcc.gnu.org/projects/cxx1y.html
813  16. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2011
814  17. http://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html
815  18. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
816  19. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
817  20. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
818  21. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
819  22. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
820  23. http://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html
821  24. http://gcc.gnu.org/wiki/Fortran2003Status
822  25. http://gcc.gnu.org/wiki/TS29113Status
823  26. http://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text%2Fplain&view=co
824  27. http://chasm-interop.sourceforge.net/
825  28. http://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support
826  29. http://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html
827  30. http://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
828  31. http://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
829  32. http://gcc.gnu.org/wiki/FunctionMultiVersioning
830  33. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1
831  34. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2
832  35. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3
833  36. http://gcc.gnu.org/onlinedocs/
834  37. mailto:gcc-help@gcc.gnu.org
835  38. mailto:gcc@gcc.gnu.org
836  39. http://gcc.gnu.org/lists.html
837  40. http://www.fsf.org/
838  41. http://gcc.gnu.org/about.html
839  42. http://validator.w3.org/check/referer
840======================================================================
841http://gcc.gnu.org/gcc-4.7/index.html
842
843                             GCC 4.7 Release Series
844
845   April 11, 2013
846
847   The [1]GNU project and the GCC developers are pleased to announce the
848   release of GCC 4.7.3.
849
850   This release is a bug-fix release, containing fixes for regressions in
851   GCC 4.7.2 relative to previous releases of GCC.
852
853Release History
854
855   GCC 4.7.3
856          April 11, 2013 ([2]changes, [3]documentation)
857
858   GCC 4.7.2
859          September 20, 2012 ([4]changes, [5]documentation)
860
861   GCC 4.7.1
862          June 14, 2012 ([6]changes, [7]documentation)
863
864   GCC 4.7.0
865          March 22, 2012 ([8]changes, [9]documentation)
866
867References and Acknowledgements
868
869   GCC used to stand for the GNU C Compiler, but since the compiler
870   supports several other languages aside from C, it now stands for the
871   GNU Compiler Collection.
872
873   A list of [10]successful builds is updated as new information becomes
874   available.
875
876   The GCC developers would like to thank the numerous people that have
877   contributed new features, improvements, bug fixes, and other changes as
878   well as test results to GCC. This [11]amazing group of volunteers is
879   what makes GCC successful.
880
881   For additional information about GCC please refer to the [12]GCC
882   project web site or contact the [13]GCC development mailing list.
883
884   To obtain GCC please use [14]our mirror sites or [15]our SVN server.
885
886
887    For questions related to the use of GCC, please consult these web
888    pages and the [16]GCC manuals. If that fails, the
889    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
890    web pages and the development of GCC are welcome on our developer
891    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
892    archives.
893
894   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
895   distribution of this entire article is permitted in any medium,
896   provided this notice is preserved.
897
898   These pages are [21]maintained by the GCC team. Last modified
899   2013-04-11[22].
900
901References
902
903   1. http://www.gnu.org/
904   2. http://gcc.gnu.org/gcc-4.7/changes.html
905   3. http://gcc.gnu.org/onlinedocs/4.7.3/
906   4. http://gcc.gnu.org/gcc-4.7/changes.html
907   5. http://gcc.gnu.org/onlinedocs/4.7.2/
908   6. http://gcc.gnu.org/gcc-4.7/changes.html
909   7. http://gcc.gnu.org/onlinedocs/4.7.1/
910   8. http://gcc.gnu.org/gcc-4.7/changes.html
911   9. http://gcc.gnu.org/onlinedocs/4.7.0/
912  10. http://gcc.gnu.org/gcc-4.7/buildstat.html
913  11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
914  12. http://gcc.gnu.org/index.html
915  13. mailto:gcc@gcc.gnu.org
916  14. http://gcc.gnu.org/mirrors.html
917  15. http://gcc.gnu.org/svn.html
918  16. http://gcc.gnu.org/onlinedocs/
919  17. mailto:gcc-help@gcc.gnu.org
920  18. mailto:gcc@gcc.gnu.org
921  19. http://gcc.gnu.org/lists.html
922  20. http://www.fsf.org/
923  21. http://gcc.gnu.org/about.html
924  22. http://validator.w3.org/check/referer
925======================================================================
926http://gcc.gnu.org/gcc-4.7/changes.html
927
928                             GCC 4.7 Release Series
929                        Changes, New Features, and Fixes
930
931Caveats
932
933     * The -fconserve-space flag has been deprecated. The flag had no
934       effect for most targets: only targets without a global .bss section
935       and without support for switchable sections. Furthermore, the flag
936       only had an effect for G++, where it could result in wrong
937       semantics (please refer to the GCC manual for further details). The
938       flag will be removed in GCC 4.8
939     * Support for a number of older systems and recently unmaintained or
940       untested target ports of GCC has been declared obsolete in GCC 4.7.
941       Unless there is activity to revive them, the next release of GCC
942       will have their sources permanently removed.
943       All GCC ports for the following processor architectures have been
944       declared obsolete:
945          + picoChip (picochip-*)
946       The following ports for individual systems on particular
947       architectures have been obsoleted:
948          + IRIX 6.5 (mips-sgi-irix6.5)
949          + MIPS OpenBSD (mips*-*-openbsd*)
950          + Solaris 8 (*-*-solaris2.8). Details can be found in the
951            [1]announcement.
952          + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*)
953     * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A,
954       ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by
955       default, which for some sources generates code that accesses memory
956       on unaligned addresses. This requires the kernel of those systems
957       to enable such accesses (controlled by CP15 register c1, refer to
958       ARM documentation). Alternatively, or for compatibility with
959       kernels where unaligned accesses are not supported, all code has to
960       be compiled with -mno-unaligned-access. Upstream Linux kernel
961       releases have automatically and unconditionally supported unaligned
962       accesses as emitted by GCC due to this option being active since
963       version 2.6.28.
964     * Support on ARM for the legacy floating-point accelerator (FPA) and
965       the mixed-endian floating-point format that it used has been
966       obsoleted. The ports that still use this format have been obsoleted
967       as well. Many legacy ARM ports already provide an alternative that
968       uses the VFP floating-point format. The obsolete ports will be
969       deleted in the next release.
970       The obsolete ports with alternatives are:
971          + arm*-*-rtems (use arm*-*-rtemseabi)
972          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
973          + arm*-*-elf (use arm*-*-eabi)
974          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
975       Note, however, that these alternatives are not binary compatible
976       with their legacy counterparts (although some can support running
977       legacy applications).
978       The obsolete ports that currently lack a modern alternative are:
979          + arm*-*-ecos-elf
980          + arm*-*-freebsd
981          + arm*-wince-pe*
982       New ports that support more recent versions of the architecture are
983       welcome.
984     * Support for the Maverick co-processor on ARM has been obsoleted.
985       Code to support it will be deleted in the next release.
986     * Support has been removed for Unix International threads on Solaris
987       2, so the --enable-threads=solaris configure option and the
988       -threads compiler option don't work any longer.
989     * Support has been removed for the Solaris BSD Compatibility Package,
990       which lives in /usr/ucbinclude and /usr/ucblib. It has been removed
991       from Solaris 11, and was only intended as a migration aid from
992       SunOS 4 to SunOS 5. The -compat-bsd compiler option is not
993       recognized any longer.
994     * The AVR port's libgcc has been improved and its multilib structure
995       has been enhanced. As a result, all objects contributing to an
996       application must either be compiled with GCC versions up to 4.6.x
997       or with GCC versions 4.7.1 or later. If the compiler is used with
998       AVR Libc, you need a version that supports the new layout, i.e.
999       implements [2]#35407.
1000     * The AVR port's -mshort-calls command-line option has been
1001       deprecated. It will be removed in the GCC 4.8 release. See -mrelax
1002       for a replacement.
1003     * The AVR port only references startup code that clears .bss and the
1004       common section resp. initializes the .data and .rodata section
1005       provided respective sections (or subsections thereof) are not
1006       empty, see [3]PR18145. Applications that put all static storage
1007       objects into non-standard sections and / or define all static
1008       storage objects in assembler modules, must reference __do_clear_bss
1009       resp. __do_copy_data by hand or undefine the symbol(s) by means of
1010       -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data.
1011     * The ARM port's -mwords-little-endian option has been deprecated. It
1012       will be removed in a future release.
1013     * Support has been removed for the NetWare x86 configuration
1014       obsoleted in GCC 4.6.
1015     * It is no longer possible to use the "l" constraint in MIPS16 asm
1016       statements.
1017     * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard
1018       library which affected the ABI in C++11 mode: a data member was
1019       added to std::list changing its size and altering the definitions
1020       of some member functions, and std::pair's move constructor was
1021       non-trivial which altered the calling convention for functions with
1022       std::pair arguments or return types. The ABI incompatibilities have
1023       been fixed for GCC version 4.7.2 but as a result C++11 code
1024       compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11
1025       code compiled with different GCC versions and with C++98/C++03 code
1026       compiled with any version.
1027     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
1028       rules for the layout of vectors that could lead to wrong code being
1029       generated. Vectors larger than 8 bytes in size are now by default
1030       aligned to an 8-byte boundary. This is an ABI change: code that
1031       makes explicit use of vector types may be incompatible with binary
1032       objects built with older versions of GCC. Auto-vectorized code is
1033       not affected by this change. (This change affects GCC versions
1034       4.7.2 and later.)
1035     * More information on porting to GCC 4.7 from previous versions of
1036       GCC can be found in the [4]porting guide for this release.
1037
1038General Optimizer Improvements
1039
1040     * Support for a new parameter --param case-values-threshold=n was
1041       added to allow users to control the cutoff between doing switch
1042       statements as a series of if statements and using a jump table.
1043     * Link-time optimization (LTO) improvements:
1044          + Improved scalability and reduced memory usage. Link time
1045            optimization of Firefox now requires 3GB of RAM on a 64-bit
1046            system, while over 8GB was needed previously. Linking time has
1047            been improved, too. The serial stage of linking Firefox has
1048            been sped up by about a factor of 10.
1049          + Reduced size of object files and temporary storage used during
1050            linking.
1051          + Streaming performance (both outbound and inbound) has been
1052            improved.
1053          + ld -r is now supported with LTO.
1054          + Several bug fixes, especially in symbol table handling and
1055            merging.
1056     * Interprocedural optimization improvements:
1057          + Heuristics now take into account that after inlining code will
1058            be optimized out because of known values (or properties) of
1059            function parameters. For example:
1060void foo(int a)
1061{
1062  if (a > 10)
1063    ... huge code ...
1064}
1065void bar (void)
1066{
1067  foo (0);
1068}
1069
1070            The call of foo will be inlined into bar even when optimizing
1071            for code size. Constructs based on __builtin_constant_p are
1072            now understood by the inliner and code size estimates are
1073            evaluated a lot more realistically.
1074          + The representation of C++ virtual thunks and aliases (both
1075            implicit and defined via the alias attribute) has been
1076            re-engineered. Aliases no longer pose optimization barriers
1077            and calls to an alias can be inlined and otherwise optimized.
1078          + The inter-procedural constant propagation pass has been
1079            rewritten. It now performs generic function specialization.
1080            For example when compiling the following:
1081void foo(bool flag)
1082{
1083  if (flag)
1084    ... do something ...
1085  else
1086    ... do something else ...
1087}
1088void bar (void)
1089{
1090  foo (false);
1091  foo (true);
1092  foo (false);
1093  foo (true);
1094  foo (false);
1095  foo (true);
1096}
1097
1098            GCC will now produce two copies of foo. One with flag being
1099            true, while other with flag being false. This leads to
1100            performance improvements previously possible only by inlining
1101            all calls. Cloning causes a lot less code size growth.
1102     * A string length optimization pass has been added. It attempts to
1103       track string lengths and optimize various standard C string
1104       functions like strlen, strchr, strcpy, strcat, stpcpy and their
1105       _FORTIFY_SOURCE counterparts into faster alternatives. This pass is
1106       enabled by default at -O2 or above, unless optimizing for size, and
1107       can be disabled by the -fno-optimize-strlen option. The pass can
1108       e.g. optimize
1109char *bar (const char *a)
1110{
1111  size_t l = strlen (a) + 2;
1112  char *p = malloc (l); if (p == NULL) return p;
1113  strcpy (p, a); strcat (p, "/"); return p;
1114}
1115
1116       into:
1117char *bar (const char *a)
1118{
1119  size_t tmp = strlen (a);
1120  char *p = malloc (tmp + 2); if (p == NULL) return p;
1121  memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p;
1122}
1123
1124       or for hosted compilations where stpcpy is available in the runtime
1125       and headers provide its prototype, e.g.
1126void foo (char *a, const char *b, const char *c, const char *d)
1127{
1128  strcpy (a, b); strcat (a, c); strcat (a, d);
1129}
1130
1131       can be optimized into:
1132void foo (char *a, const char *b, const char *c, const char *d)
1133{
1134  strcpy (stpcpy (stpcpy (a, b), c), d);
1135}
1136
1137New Languages and Language specific improvements
1138
1139     * Version 3.1 of the [5]OpenMP specification is now supported for the
1140       C, C++, and Fortran compilers.
1141
1142  Ada
1143
1144     * The command-line option -feliminate-unused-debug-types has been
1145       re-enabled by default, as it is for the other languages, leading to
1146       a reduction in debug info size of 12.5% and more for relevant
1147       cases, as well as to a small compilation speedup.
1148
1149  C family
1150
1151     * A new built-in, __builtin_assume_aligned, has been added, through
1152       which the compiler can be hinted about pointer alignment and can
1153       use it to improve generated code.
1154     * A new warning option -Wunused-local-typedefs was added for C, C++,
1155       Objective-C and Objective-C++. This warning diagnoses typedefs
1156       locally defined in a function, and otherwise not used.
1157     * A new experimental command-line option -ftrack-macro-expansion was
1158       added for C, C++, Objective-C, Objective-C++ and Fortran. It allows
1159       the compiler to emit diagnostic about the current macro expansion
1160       stack when a compilation error occurs in a macro expansion.
1161     * Experimental support for transactional memory has been added. It
1162       includes support in the compiler, as well as a supporting runtime
1163       library called libitm. To compile code with transactional memory
1164       constructs, use the -fgnu-tm option.
1165       Support is currently available for Alpha, ARM, PowerPC, SH, SPARC,
1166       and 32-bit/64-bit x86 platforms.
1167       For more details on transactional memory see [6]the GCC WiKi.
1168     * Support for atomic operations specifying the C++11/C11 memory model
1169       has been added. These new __atomic routines replace the existing
1170       __sync built-in routines.
1171       Atomic support is also available for memory blocks. Lock-free
1172       instructions will be used if a memory block is the same size and
1173       alignment as a supported integer type. Atomic operations which do
1174       not have lock-free support are left as function calls. A set of
1175       library functions is available on the GCC atomic wiki in the
1176       "External Atomics Library" section.
1177       For more details on the memory models and features, see the
1178       [7]atomic wiki.
1179     * When a binary operation is performed on vector types and one of the
1180       operands is a uniform vector, it is possible to replace the vector
1181       with the generating element. For example:
1182typedef int v4si __attribute__ ((vector_size (16)));
1183v4si res, a = {1,2,3,4};
1184int x;
1185
1186res = 2 + a;  /* means {2,2,2,2} + a  */
1187res = a - x;  /* means a - {x,x,x,x}  */
1188
1189  C
1190
1191     * There is support for some more features from the C11 revision of
1192       the ISO C standard. GCC now accepts the options -std=c11 and
1193       -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x.
1194          + Unicode strings (previously supported only with options such
1195            as -std=gnu11, now supported with -std=c11), and the
1196            predefined macros __STDC_UTF_16__ and __STDC_UTF_32__.
1197          + Nonreturning functions (_Noreturn and <stdnoreturn.h>).
1198          + Alignment support (_Alignas, _Alignof, max_align_t,
1199            <stdalign.h>).
1200          + A built-in function __builtin_complex is provided to support C
1201            library implementation of the CMPLX family of macros.
1202
1203  C++
1204
1205     * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat
1206       options, which are equivalent to -std=c++0x, -std=gnu++0x, and
1207       -Wc++0x-compat, respectively.
1208     * G++ now implements [8]C++11 extended friend syntax:
1209
1210template<class W>
1211class Q
1212{
1213  static const int I = 2;
1214public:
1215  friend W;
1216};
1217
1218struct B
1219{
1220  int ar[Q<B>::I];
1221};
1222
1223     * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit
1224       override control.
1225
1226struct B {
1227  virtual void f() const final;
1228  virtual void f(int);
1229};
1230
1231struct D : B {
1232  void f() const;            // error: D::f attempts to override final B::f
1233  void f(long) override;     // error: doesn't override anything
1234  void f(int) override;      // ok
1235};
1236
1237struct E final { };
1238struct F: E { }; // error: deriving from final class
1239
1240     * G++ now implements [10]C++11 non-static data member initializers.
1241
1242struct A {
1243  int i = 42;
1244} a; // initializes a.i to 42
1245
1246     * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11
1247       user-defined literals.
1248
1249// Not actually a good approximation.  :)
1250constexpr long double operator"" _degrees (long double d) { return d * 0.0175; }
1251long double pi = 180.0_degrees;
1252
1253     * G++ now implements [12]C++11 alias-declarations.
1254
1255template <class T> using Ptr = T*;
1256Ptr<int> ip;  // decltype(ip) is int*
1257
1258     * Thanks to Ville Voutilainen and Pedro Lamarao, G++ now implements
1259       [13]C++11 delegating constructors.
1260
1261struct A {
1262  A(int);
1263  A(): A(42) { } // delegate to the A(int) constructor
1264};
1265
1266     * G++ now fully implements C++11 atomic classes rather than just
1267       integer derived classes.
1268
1269class POD {
1270  int a;
1271  int b;
1272};
1273std::atomic<POD> my_atomic_POD;
1274
1275     * G++ now sets the predefined macro __cplusplus to the correct value,
1276       199711L for C++98/03, and 201103L for C++11.
1277     * G++ now correctly implements the two-phase lookup rules such that
1278       an unqualified name used in a template must have an appropriate
1279       declaration found either in scope at the point of definition of the
1280       template or by argument-dependent lookup at the point of
1281       instantiation. As a result, code that relies on a second
1282       unqualified lookup at the point of instantiation to find functions
1283       declared after the template or in dependent bases will be rejected.
1284       The compiler will suggest ways to fix affected code, and using the
1285       -fpermissive compiler flag will allow the code to compile with a
1286       warning.
1287
1288template <class T>
1289void f() { g(T()); } // error, g(int) not found by argument-dependent lookup
1290void g(int) { } // fix by moving this declaration before the declaration of f
1291
1292template <class T>
1293struct A: T {
1294  // error, B::g(B) not found by argument-dependent lookup
1295  void f() { g(T()); } // fix by using this->g or A::g
1296};
1297
1298struct B { void g(B); };
1299
1300int main()
1301{
1302  f<int>();
1303  A<B>().f();
1304}
1305
1306     * G++ now properly re-uses stack space allocated for temporary
1307       objects when their lifetime ends, which can significantly lower
1308       stack consumption for some C++ functions. As a result of this, some
1309       code with undefined behavior will now break:
1310
1311const int &f(const int &i) { return i; }
1312....
1313const int &x = f(1);
1314const int &y = f(2);
1315
1316       Here, x refers to the temporary allocated to hold the 1 argument,
1317       which only lives until the end of the initialization; it
1318       immediately becomes a dangling reference. So the next statement
1319       re-uses the stack slot to hold the 2 argument, and users of x get
1320       that value instead.
1321       Note that this should not cause any change of behavior for
1322       temporaries of types with non-trivial destructors, as they are
1323       already destroyed at end of full-expression; the change is that now
1324       the storage is released as well.
1325     * A new command-line option -Wdelete-non-virtual-dtor has been added
1326       to warn when delete is used to destroy an instance of a class which
1327       has virtual functions and non-virtual destructor. It is unsafe to
1328       delete an instance of a derived class through a pointer to a base
1329       class if the base class does not have a virtual destructor. This
1330       warning is enabled by -Wall.
1331     * A new command-line option -Wzero-as-null-pointer-constant has been
1332       added to warn when a literal '0' is used as null pointer constant.
1333       It can be useful to facilitate the conversion to nullptr in C++11.
1334     * As per C++98, access-declarations are now deprecated by G++.
1335       Using-declarations are to be used instead. Furthermore, some
1336       efforts have been made to improve the support of class scope
1337       using-declarations. In particular, using-declarations referring to
1338       a dependent type now work as expected ([14]bug c++/14258).
1339     * The ELF symbol visibility of a template instantiation is now
1340       properly constrained by the visibility of its template arguments
1341       ([15]bug c++/35688).
1342
1343    Runtime Library (libstdc++)
1344
1345     * [16]Improved experimental support for the new ISO C++ standard,
1346       C++11, including:
1347          + using noexcept in most of the library;
1348          + implementations of pointer_traits, allocator_traits and
1349            scoped_allocator_adaptor;
1350          + uses-allocator construction for tuple;
1351          + vector meets the allocator-aware container requirements;
1352          + replacing monotonic_clock with steady_clock;
1353          + enabling the thread support library on most POSIX targets;
1354          + many small improvements to conform to the FDIS.
1355     * Added --enable-clocale=newlib configure option.
1356     * Debug Mode iterators for unordered associative containers.
1357     * Avoid polluting the global namespace and do not include <unistd.h>.
1358
1359  Fortran
1360
1361     * The compile flag [17]-fstack-arrays has been added, which causes
1362       all local arrays to be put on stack memory. For some programs this
1363       will improve the performance significantly. If your program uses
1364       very large local arrays, it is possible that you will have to
1365       extend your runtime limits for stack memory.
1366     * The [18]-Ofast flag now also implies [19]-fno-protect-parens and
1367       [20]-fstack-arrays.
1368     * Front-end optimizations can now be selected by the
1369       [21]-ffrontend-optimize option and deselected by the
1370       -fno-frontend-optimize option.
1371     * When front-end optimization removes a function call,
1372       [22]-Wfunction-elimination warns about that.
1373     * When performing front-end-optimization, the
1374       [23]-faggressive-function-elimination option allows the removal of
1375       duplicate function calls even for impure functions.
1376     * The flag [24]-Wreal-q-constant has been added, which warns if
1377       floating-point literals have been specified using q (such as
1378       1.0q0); the q marker is now supported as a vendor extension to
1379       denote quad precision (REAL(16) or, if not available, REAL(10)).
1380       Consider using a kind parameter (such as in 1.0_qp) instead, which
1381       can be obtained via [25]SELECTED_REAL_KIND.
1382     * The GFORTRAN_USE_STDERR environment variable has been removed. GNU
1383       Fortran now always prints error messages to standard error. If you
1384       wish to redirect standard error, please consult the manual for your
1385       OS, shell, batch environment etc. as appropriate.
1386     * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment
1387       variable have been removed. When encountering a serious error,
1388       gfortran will now always abort the program. Whether a core dump is
1389       generated depends on the user environment settings; see the ulimit
1390       -c setting for POSIX shells, limit coredumpsize for C shells, and
1391       the [26]WER user-mode dumps settings on Windows.
1392     * The [27]-fbacktrace option is now enabled by default. When
1393       encountering a fatal error, gfortran will attempt to print a
1394       backtrace to standard error before aborting. It can be disabled
1395       with -fno-backtrace. Note: On POSIX targets with the addr2line
1396       utility from GNU binutils, GNU Fortran can print a backtrace with
1397       function name, file name, line number information in addition to
1398       the addresses; otherwise only the addresses are printed.
1399     * [28]Fortran 2003:
1400          + Generic interface names which have the same name as derived
1401            types are now supported, which allows to write constructor
1402            functions. Note that Fortran does not support static
1403            constructor functions; only default initialization or an
1404            explicit structure-constructor initialization are available.
1405          + [29]Polymorphic (class) arrays are now supported.
1406     * [30]Fortran 2008:
1407          + Support for the DO CONCURRENT construct has been added, which
1408            allows the user to specify that individual loop iterations
1409            have no interdependencies.
1410          + [31]Coarrays: Full single-image support except for polymorphic
1411            coarrays. Additionally, preliminary support for multiple
1412            images via an MPI-based [32]coarray communication library has
1413            been added. Note: The library version is not yet usable as
1414            remote coarray access is not yet possible.
1415     * [33]TS 29113:
1416          + New flag [34]-std=f2008ts permits programs that are expected
1417            to conform to the Fortran 2008 standard and the draft
1418            Technical Specification (TS) 29113 on Further Interoperability
1419            of Fortran with C.
1420          + The OPTIONAL attribute is now allowed for dummy arguments of
1421            BIND(C) procedures.
1422          + The RANK intrinsic has been added.
1423          + The implementation of the ASYNCHRONOUS attribute in GCC is
1424            compatible with the candidate draft of TS 29113 (since GCC
1425            4.6).
1426
1427  Go
1428
1429     * GCC 4.7 implements the [35]Go 1 language standard. The library
1430       support in 4.7.0 is not quite complete, due to release timing.
1431       Release 4.7.1 includes complete support for Go 1. The Go library is
1432       from the Go 1.0.1 release.
1433     * Go has been tested on GNU/Linux and Solaris platforms. It may work
1434       on other platforms as well.
1435
1436New Targets and Target Specific Improvements
1437
1438  ARM
1439
1440     * GCC now supports the Cortex-A7 processor implementing the v7-a
1441       version of the architecture using the option -mcpu=cortex-a7.
1442     * The default vector size in auto-vectorization for NEON is now 128
1443       bits. If vectorization fails thusly, the vectorizer tries again
1444       with 64-bit vectors.
1445     * A new option -mvectorize-with-neon-double was added to allow users
1446       to change the vector size to 64 bits.
1447
1448  AVR
1449
1450     * GCC now supports the XMEGA architecture. This requires GNU binutils
1451       2.22 or later.
1452     * Support for the [36]named address spaces __flash, __flash1, ...,
1453       __flash5 and __memx has been added. These address spaces locate
1454       read-only data in flash memory and allow reading from flash memory
1455       by means of ordinary C code, i.e. without the need of (inline)
1456       assembler code:
1457
1458const __flash int values[] = { 42, 31 };
1459
1460int add_values (const __flash int *p, int i)
1461{
1462    return values[i] + *p;
1463}
1464
1465     * Support has been added for the AVR-specific configure option
1466       --with-avrlibc=yes in order to arrange for better integration of
1467       [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2
1468       and newer and will only take effect in non-RTEMS configurations. If
1469       avr-gcc is configured for RTEMS, the option will be ignored which
1470       is the same as specifying --with-avrlibc=no. See [38]PR54461 for
1471       more technical details.
1472     * Support for AVR-specific [39]built-in functions has been added.
1473     * Support has been added for the signed and unsigned 24-bit scalar
1474       integer types __int24 and __uint24.
1475     * New command-line options -maccumulate-args, -mbranch-cost=cost and
1476       -mstrict-X were added to allow better fine-tuning of code
1477       optimization.
1478     * The command option -fdata-sections now also takes affect on the
1479       section names of variables with the progmem attribute.
1480     * A new inline assembler print modifier %i to print a RAM address as
1481       I/O address has been added:
1482
1483#include <avr/io.h> /* Port Definitions from AVR-LibC */
1484
1485void set_portb (uint8_t value)
1486{
1487    asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory");
1488}
1489
1490       The offset between an I/O address and the RAM address for that I/O
1491       location is device-specific. This offset is taken into account when
1492       printing a RAM address with the %i modifier so that the address is
1493       suitable to be used as operand in an I/O command. The address must
1494       be a constant integer known at compile time.
1495     * The inline assembler constraint "R" to represent integers in the
1496       range -6 ... 5 has been removed without replacement.
1497     * Many optimizations to:
1498          + 64-bit integer arithmetic
1499          + Widening multiplication
1500          + Integer division by a constant
1501          + Avoid constant reloading in multi-byte instructions.
1502          + Micro-optimizations for special instruction sequences.
1503          + Generic built-in functions like __builtin_ffs*,
1504            __builtin_clz*, etc.
1505          + If-else decision trees generated by switch instructions
1506          + Merging of data located in flash memory
1507          + New libgcc variants for devices with 8-bit wide stack pointer
1508          + ...
1509     * Better documentation:
1510          + Handling of EIND and indirect jumps on devices with more than
1511            128 KiB of program memory.
1512          + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function
1513            registers.
1514          + Function attributes OS_main and OS_task.
1515          + AVR-specific built-in macros.
1516
1517  C6X
1518
1519     * Support has been added for the Texas Instruments C6X family of
1520       processors.
1521
1522  CR16
1523
1524     * Support has been added for National Semiconductor's CR16
1525       architecture.
1526
1527  Epiphany
1528
1529     * Support has been added for Adapteva's Epiphany architecture.
1530
1531  IA-32/x86-64
1532
1533     * Support for Intel AVX2 intrinsics, built-in functions and code
1534       generation is available via -mavx2.
1535     * Support for Intel BMI2 intrinsics, built-in functions and code
1536       generation is available via -mbmi2.
1537     * Implementation and automatic generation of __builtin_clz* using the
1538       lzcnt instruction is available via -mlzcnt.
1539     * Support for Intel FMA3 intrinsics and code generation is available
1540       via -mfma.
1541     * A new -mfsgsbase command-line option is available that makes GCC
1542       generate new segment register read/write instructions through
1543       dedicated built-ins.
1544     * Support for the new Intel rdrnd instruction is available via
1545       -mrdrnd.
1546     * Two additional AVX vector conversion instructions are available via
1547       -mf16c.
1548     * Support for new Intel processor codename IvyBridge with RDRND,
1549       FSGSBASE and F16C is available through -march=core-avx-i.
1550     * Support for the new Intel processor codename Haswell with AVX2,
1551       FMA, BMI, BMI2, LZCNT is available through -march=core-avx2.
1552     * Support for new AMD family 15h processors (Piledriver core) is now
1553       available through -march=bdver2 and -mtune=bdver2 options.
1554     * Support for [40]the x32 psABI is now available through the -mx32
1555       option.
1556     * Windows mingw targets are using the -mms-bitfields option by
1557       default.
1558     * Windows x86 targets are using the __thiscall calling convention for
1559       C++ class-member functions.
1560     * Support for the configure option --with-threads=posix for Windows
1561       mingw targets.
1562
1563  MIPS
1564
1565     * GCC now supports thread-local storage (TLS) for MIPS16. This
1566       requires GNU binutils 2.22 or later.
1567     * GCC can now generate code specifically for the Cavium Octeon+ and
1568       Octeon2 processors. The associated command-line options are
1569       -march=octeon+ and -march=octeon2 respectively. Both options
1570       require GNU binutils 2.22 or later.
1571     * GCC can now work around certain 24k errata, under the control of
1572       the command-line option -mfix-24k. These workarounds require GNU
1573       binutils 2.20 or later.
1574     * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build
1575       n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux
1576       toolchain that generates 32-bit code by default. Use the
1577       configure-time option --enable-targets=all to select these extra
1578       multilibs.
1579     * Passing -fno-delayed-branch now also stops the assembler from
1580       automatically filling delay slots.
1581
1582  PowerPC/PowerPC64
1583
1584     * Vectors of type vector long long or vector long are passed and
1585       returned using the same method as other vectors with the VSX
1586       instruction set. Previously GCC did not adhere to the ABI for
1587       128-bit vectors with 64-bit integer base types (PR 48857). This
1588       will also be fixed in the GCC 4.6.1 and 4.5.4 releases.
1589     * A new option -mno-pointers-to-nested-functions was added to allow
1590       AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify
1591       that the compiler should not load up the chain register (r11)
1592       before calling a function through a pointer. If you use this
1593       option, you cannot call nested functions through a pointer, or call
1594       other languages that might use the static chain.
1595     * A new option msave-toc-indirect was added to allow AIX
1596       32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we
1597       save the TOC in the prologue for indirect calls or generate the
1598       save inline. This can speed up some programs that call through a
1599       function pointer a lot, but it can slow down other functions that
1600       only call through a function pointer in exceptional cases.
1601     * The PowerPC port will now enable machine-specific built-in
1602       functions when the user switches the target machine using the
1603       #pragma GCC target or __attribute__ ((__target__ ("target"))) code
1604       sequences. In addition, the target macros are updated. However, due
1605       to the way the -save-temps switch is implemented, you won't see the
1606       effect of these additional macros being defined in preprocessor
1607       output.
1608
1609  SH
1610
1611     * A new option -msoft-atomic has been added. When it is specified,
1612       GCC will generate GNU/Linux-compatible gUSA atomic sequences for
1613       the new __atomic routines.
1614     * Since it is neither supported by GAS nor officially documented,
1615       code generation for little endian SH2A has been disabled.
1616       Specifying -ml with -m2a* will now result in a compiler error.
1617     * The defunct -mbranch-cost option has been fixed.
1618     * Some improvements to the generated code of:
1619          + Utilization of the tst #imm,R0 instruction.
1620          + Dynamic shift instructions on SH2A.
1621          + Integer absolute value calculations.
1622     * The -mdiv= option for targets other than SHmedia has been fixed and
1623       documented.
1624
1625  SPARC
1626
1627     * The option -mflat has been reinstated. When it is specified, the
1628       compiler will generate code for a single register window model.
1629       This is essentially a new implementation and the corresponding
1630       debugger support has been added to GDB 7.4.
1631     * Support for the options -mtune=native and -mcpu=native has been
1632       added on selected native platforms (GNU/Linux and Solaris).
1633     * Support for the SPARC T3 (Niagara 3) processor has been added.
1634     * VIS:
1635          + An intrinsics header visintrin.h has been added.
1636          + Builtin intrinsics for the VIS 1.0 edge handling and pixel
1637            compare instructions have been added.
1638          + The little-endian version of alignaddr is now supported.
1639          + When possible, VIS builtins are marked const, which should
1640            increase the compiler's ability to optimize VIS operations.
1641          + The compiler now properly tracks the %gsr register and how it
1642            behaves as an input for various VIS instructions.
1643          + Akin to fzero, the compiler can now generate fone instructions
1644            in order to set all of the bits of a floating-point register
1645            to 1.
1646          + The documentation for the VIS intrinsics in the GCC manual has
1647            been brought up to date and many inaccuracies were fixed.
1648          + Intrinsics for the VIS 2.0 bmask, bshuffle, and
1649            non-condition-code setting edge instructions have been added.
1650            Their availability is controlled by the new -mvis2 and
1651            -mno-vis2 options. They are enabled by default on
1652            UltraSPARC-III and later CPUs.
1653     * Support for UltraSPARC Fused Multiply-Add floating-point extensions
1654       has been added. These instructions are enabled by default on SPARC
1655       T3 (Niagara 3) and later CPUs.
1656
1657  TILE-Gx/TILEPro
1658
1659     * Support has been added for the Tilera TILE-Gx and TILEPro families
1660       of processors.
1661
1662Other significant improvements
1663
1664     * A new option (-grecord-gcc-switches) was added that appends
1665       compiler command-line options that might affect code generation to
1666       the DW_AT_producer attribute string in the DWARF debugging
1667       information.
1668     * GCC now supports various new GNU extensions to the DWARF debugging
1669       information format, like [41]entry value and [42]call site
1670       information, [43]typed DWARF stack or [44]a more compact macro
1671       representation. Support for these extensions has been added to GDB
1672       7.4. They can be disabled through the -gstrict-dwarf command-line
1673       option.
1674
1675GCC 4.7.1
1676
1677   This is the [45]list of problem reports (PRs) from GCC's bug tracking
1678   system that are known to be fixed in the 4.7.1 release. This list might
1679   not be complete (that is, it is possible that some PRs that have been
1680   fixed are not listed here).
1681
1682   The Go frontend in the 4.7.1 release fully supports the [46]Go 1
1683   language standard.
1684
1685GCC 4.7.2
1686
1687   This is the [47]list of problem reports (PRs) from GCC's bug tracking
1688   system that are known to be fixed in the 4.7.2 release. This list might
1689   not be complete (that is, it is possible that some PRs that have been
1690   fixed are not listed here).
1691
1692GCC 4.7.3
1693
1694   This is the [48]list of problem reports (PRs) from GCC's bug tracking
1695   system that are known to be fixed in the 4.7.3 release. This list might
1696   not be complete (that is, it is possible that some PRs that have been
1697   fixed are not listed here).
1698
1699
1700    For questions related to the use of GCC, please consult these web
1701    pages and the [49]GCC manuals. If that fails, the
1702    [50]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1703    web pages and the development of GCC are welcome on our developer
1704    list at [51]gcc@gcc.gnu.org. All of [52]our lists have public
1705    archives.
1706
1707   Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and
1708   distribution of this entire article is permitted in any medium,
1709   provided this notice is preserved.
1710
1711   These pages are [54]maintained by the GCC team. Last modified
1712   2013-04-11[55].
1713
1714References
1715
1716   1. http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html
1717   2. http://savannah.nongnu.org/bugs/?35407
1718   3. http://gcc.gnu.org/PR18145
1719   4. http://gcc.gnu.org/gcc-4.7/porting_to.html
1720   5. http://openmp.org/wp/openmp-specifications/
1721   6. http://gcc.gnu.org/wiki/TransactionalMemory
1722   7. http://gcc.gnu.org/wiki/Atomic/GCCMM
1723   8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1724   9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1725  10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1726  11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1727  12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1728  13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1729  14. http://gcc.gnu.org/PR14258
1730  15. http://gcc.gnu.org/PR35688
1731  16. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/libstdc++/manual/manual/status.html#status.iso.2011
1732  17. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
1733  18. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689
1734  19. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270
1735  20. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
1736  21. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275
1737  22. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170
1738  23. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270
1739  24. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149
1740  25. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html
1741  26. http://msdn.microsoft.com/en-us/library/bb787181%28v=vs.85%29.aspx
1742  27. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183
1743  28. http://gcc.gnu.org/wiki/Fortran2003Status
1744  29. http://gcc.gnu.org/wiki/OOP
1745  30. http://gcc.gnu.org/wiki/Fortran2008Status
1746  31. http://gcc.gnu.org/wiki/Coarray
1747  32. http://gcc.gnu.org/wiki/CoarrayLib
1748  33. http://gcc.gnu.org/wiki/TS29113Status
1749  34. http://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
1750  35. http://weekly.golang.org/doc/go1.html
1751  36. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html
1752  37. http://nongnu.org/avr-libc/
1753  38. http://gcc.gnu.org/PR54461
1754  39. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html
1755  40. https://sites.google.com/site/x32abi/
1756  41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1
1757  42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2
1758  43. http://www.dwarfstd.org/doc/040408.1.html
1759  44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1
1760  45. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1
1761  46. http://weekly.golang.org/doc/go1.html
1762  47. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2
1763  48. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3
1764  49. http://gcc.gnu.org/onlinedocs/
1765  50. mailto:gcc-help@gcc.gnu.org
1766  51. mailto:gcc@gcc.gnu.org
1767  52. http://gcc.gnu.org/lists.html
1768  53. http://www.fsf.org/
1769  54. http://gcc.gnu.org/about.html
1770  55. http://validator.w3.org/check/referer
1771======================================================================
1772http://gcc.gnu.org/gcc-4.6/index.html
1773
1774                             GCC 4.6 Release Series
1775
1776   April 12, 2013
1777
1778   The [1]GNU project and the GCC developers are pleased to announce the
1779   release of GCC 4.6.4.
1780
1781   This release is a bug-fix release, containing fixes for regressions in
1782   GCC 4.6.3 relative to previous releases of GCC.
1783
1784Release History
1785
1786   GCC 4.6.4
1787          April 12, 2013 ([2]changes, [3]documentation)
1788
1789   GCC 4.6.3
1790          March 1, 2012 ([4]changes, [5]documentation)
1791
1792   GCC 4.6.2
1793          October 26, 2011 ([6]changes, [7]documentation)
1794
1795   GCC 4.6.1
1796          June 27, 2011 ([8]changes, [9]documentation)
1797
1798   GCC 4.6.0
1799          March 25, 2011 ([10]changes, [11]documentation)
1800
1801References and Acknowledgements
1802
1803   GCC used to stand for the GNU C Compiler, but since the compiler
1804   supports several other languages aside from C, it now stands for the
1805   GNU Compiler Collection.
1806
1807   A list of [12]successful builds is updated as new information becomes
1808   available.
1809
1810   The GCC developers would like to thank the numerous people that have
1811   contributed new features, improvements, bug fixes, and other changes as
1812   well as test results to GCC. This [13]amazing group of volunteers is
1813   what makes GCC successful.
1814
1815   For additional information about GCC please refer to the [14]GCC
1816   project web site or contact the [15]GCC development mailing list.
1817
1818   To obtain GCC please use [16]our mirror sites or [17]our SVN server.
1819
1820
1821    For questions related to the use of GCC, please consult these web
1822    pages and the [18]GCC manuals. If that fails, the
1823    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1824    web pages and the development of GCC are welcome on our developer
1825    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
1826    archives.
1827
1828   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
1829   distribution of this entire article is permitted in any medium,
1830   provided this notice is preserved.
1831
1832   These pages are [23]maintained by the GCC team. Last modified
1833   2013-04-12[24].
1834
1835References
1836
1837   1. http://www.gnu.org/
1838   2. http://gcc.gnu.org/gcc-4.6/changes.html
1839   3. http://gcc.gnu.org/onlinedocs/4.6.4/
1840   4. http://gcc.gnu.org/gcc-4.6/changes.html
1841   5. http://gcc.gnu.org/onlinedocs/4.6.3/
1842   6. http://gcc.gnu.org/gcc-4.6/changes.html
1843   7. http://gcc.gnu.org/onlinedocs/4.6.2/
1844   8. http://gcc.gnu.org/gcc-4.6/changes.html
1845   9. http://gcc.gnu.org/onlinedocs/4.6.1/
1846  10. http://gcc.gnu.org/gcc-4.6/changes.html
1847  11. http://gcc.gnu.org/onlinedocs/4.6.0/
1848  12. http://gcc.gnu.org/gcc-4.6/buildstat.html
1849  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
1850  14. http://gcc.gnu.org/index.html
1851  15. mailto:gcc@gcc.gnu.org
1852  16. http://gcc.gnu.org/mirrors.html
1853  17. http://gcc.gnu.org/svn.html
1854  18. http://gcc.gnu.org/onlinedocs/
1855  19. mailto:gcc-help@gcc.gnu.org
1856  20. mailto:gcc@gcc.gnu.org
1857  21. http://gcc.gnu.org/lists.html
1858  22. http://www.fsf.org/
1859  23. http://gcc.gnu.org/about.html
1860  24. http://validator.w3.org/check/referer
1861======================================================================
1862http://gcc.gnu.org/gcc-4.6/changes.html
1863
1864                             GCC 4.6 Release Series
1865                        Changes, New Features, and Fixes
1866
1867Caveats
1868
1869     * The options -b <machine> and -V <version> have been removed because
1870       they were unreliable. Instead, users should directly run
1871       <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to
1872       run a different version of gcc.
1873     * GCC now has stricter checks for invalid command-line options. In
1874       particular, when gcc was called to link object files rather than
1875       compile source code, it would previously accept and ignore all
1876       options starting with --, including linker options such as
1877       --as-needed and --export-dynamic, although such options would
1878       result in errors if any source code was compiled. Such options, if
1879       unknown to the compiler, are now rejected in all cases; if the
1880       intent was to pass them to the linker, options such as
1881       -Wl,--as-needed should be used.
1882     * Versions of the GNU C library up to and including 2.11.1 included
1883       an [1]incorrect implementation of the cproj function. GCC optimizes
1884       its builtin cproj according to the behavior specified and allowed
1885       by the ISO C99 standard. If you want to avoid discrepancies between
1886       the C library and GCC's builtin transformations when using cproj in
1887       your code, use GLIBC 2.12 or later. If you are using an older GLIBC
1888       and actually rely on the incorrect behavior of cproj, then you can
1889       disable GCC's transformations using -fno-builtin-cproj.
1890     * The C-only intermodule optimization framework (IMA, enabled by
1891       -combine) has been removed in favor of the new generic link-time
1892       optimization framework (LTO) introduced in [2]GCC 4.5.0.
1893     * GCC now ships with the LGPL-licensed libquadmath library, which
1894       provides quad-precision mathematical functions for targets with a
1895       __float128 datatype. __float128 is available for targets on 32-bit
1896       x86, x86-64 and Itanium architectures. The libquadmath library is
1897       automatically built on such targets when building the Fortran
1898       compiler.
1899     * New -Wunused-but-set-variable and -Wunused-but-set-parameter
1900       warnings were added for C, C++, Objective-C and Objective-C++.
1901       These warnings diagnose variables respective parameters which are
1902       only set in the code and never otherwise used. Usually such
1903       variables are useless and often even the value assigned to them is
1904       computed needlessly, sometimes expensively. The
1905       -Wunused-but-set-variable warning is enabled by default by -Wall
1906       flag and -Wunused-but-set-parameter by -Wall -Wextra flags.
1907     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
1908       rules for the layout of vectors that could lead to wrong code being
1909       generated. Vectors larger than 8 bytes in size are now by default
1910       aligned to an 8-byte boundary. This is an ABI change: code that
1911       makes explicit use of vector types may be incompatible with binary
1912       objects built with older versions of GCC. Auto-vectorized code is
1913       not affected by this change. (This change affects GCC versions
1914       4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.)
1915     * On AVR, variables with the progmem attribute to locate data in
1916       flash memory must be qualified as const.
1917     * Support for a number of older systems and recently unmaintained or
1918       untested target ports of GCC has been declared obsolete in GCC 4.6.
1919       Unless there is activity to revive them, the next release of GCC
1920       will have their sources permanently removed.
1921       All GCC ports for the following processor architectures have been
1922       declared obsolete:
1923          + Argonaut ARC (arc-*)
1924          + National Semiconductor CRX (crx-*)
1925          + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*,
1926            m68hc12-*-*, m6812-*-*)
1927          + Sunplus S+core (score-*)
1928       The following ports for individual systems on particular
1929       architectures have been obsoleted:
1930          + Interix (i[34567]86-*-interix3*)
1931          + NetWare x86 (i[3456x]86-*-netware*)
1932          + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*)
1933          + MCore PE (mcore-*-pe*)
1934          + SH SymbianOS (sh*-*-symbianelf*)
1935          + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*)
1936          + M68K uClinux old ABI (m68k-*-uclinuxoldabi*)
1937          + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*,
1938            vax-*-netbsd*, but not *-*-netbsdelf*)
1939       The i[34567]86-*-pe alias for Cygwin targets has also been
1940       obsoleted; users should configure for i[34567]86-*-cygwin* instead.
1941       Certain configure options to control the set of libraries built
1942       with GCC on some targets have been obsoleted. On ARM targets, the
1943       options --disable-fpu, --disable-26bit, --disable-underscore,
1944       --disable-interwork, --disable-biendian and --disable-nofmult have
1945       been obsoleted. On MIPS targets, the options
1946       --disable-single-float, --disable-biendian and --disable-softfloat
1947       have been obsoleted.
1948     * Support has been removed for all the [3]configurations obsoleted in
1949       GCC 4.5.
1950     * More information on porting to GCC 4.6 from previous versions of
1951       GCC can be found in the [4]porting guide for this release.
1952
1953General Optimizer Improvements
1954
1955     * A new general optimization level, -Ofast, has been introduced. It
1956       combines the existing optimization level -O3 with options that can
1957       affect standards compliance but result in better optimized code.
1958       For example, -Ofast enables -ffast-math.
1959     * Link-time optimization improvements:
1960          + The [5]Scalable Whole Program Optimizer (WHOPR) project has
1961            stabilized to the point of being usable. It has become the
1962            default mode when using the LTO optimization model. Link time
1963            optimization can now split itself into multiple parallel
1964            compilations. Parallelism is controlled with -flto=n (where n
1965            specifies the number of compilations to execute in parallel).
1966            GCC can also cooperate with a GNU make job server by
1967            specifying the -flto=jobserver option and adding + to the
1968            beginning of the Makefile rule executing the linker.
1969            Classical LTO mode can be enforced by -flto-partition=none.
1970            This may result in small code quality improvements.
1971          + A large number of bugs were fixed. GCC itself, Mozilla Firefox
1972            and other large applications can be built with LTO enabled.
1973          + The linker plugin support improvements
1974               o Linker plugin is now enabled by default when the linker
1975                 is detected to have plugin support. This is the case for
1976                 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and
1977                 the Gold linker on ELF targets. Plugin support of the
1978                 Apple linker on Darwin is not compatible with GCC. The
1979                 linker plugin can also be controlled by the
1980                 -fuse-linker-plugin command line option.
1981               o Resolution information from the linker plugin is used to
1982                 drive whole program assumptions. Use of the linker plugin
1983                 results in more aggressive optimization on binaries and
1984                 on shared libraries that use the hidden visibility
1985                 attribute. Consequently the use of -fwhole-program is not
1986                 necessary in addition to LTO.
1987          + Hidden symbols used from non-LTO objects now have to be
1988            explicitly annotated with externally_visible when the linker
1989            plugin is not used.
1990          + C++ inline functions and virtual tables are now privatized
1991            more aggressively, leading to better inter-procedural
1992            optimization and faster dynamic linking.
1993          + Memory usage and intermediate language streaming performance
1994            have been improved.
1995          + Static constructors and destructors from individual units are
1996            inlined into a single function. This can significantly improve
1997            startup times of large C++ applications where static
1998            constructors are very common. For example, static constructors
1999            are used when including the iostream header.
2000          + Support for the Ada language has been added.
2001     * Interprocedural optimization improvements
2002          + The interprocedural framework was re-tuned for link time
2003            optimization. Several scalability issues were resolved.
2004          + Improved auto-detection of const and pure functions. Newly,
2005            noreturn functions are auto-detected.
2006            The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is
2007            available that informs users when adding attributes to headers
2008            might improve code generation.
2009          + A number of inlining heuristic improvements. In particular:
2010               o Partial inlining is now supported and enabled by default
2011                 at -O2 and greater. The feature can be controlled via
2012                 -fpartial-inlining.
2013                 Partial inlining splits functions with short hot path to
2014                 return. This allows more aggressive inlining of the hot
2015                 path leading to better performance and often to code size
2016                 reductions (because cold parts of functions are not
2017                 duplicated).
2018               o Scalability for large compilation units was improved
2019                 significantly.
2020               o Inlining of callbacks is now more aggressive.
2021               o Virtual methods are considered for inlining when the
2022                 caller is inlined and devirtualization is then possible.
2023               o Inlining when optimizing for size (either in cold regions
2024                 of a program or when compiling with -Os) was improved to
2025                 better handle C++ programs with larger abstraction
2026                 penalty, leading to smaller and faster code.
2027          + The IPA reference optimization pass detecting global variables
2028            used or modified by functions was strengthened and sped up.
2029          + Functions whose address was taken are now optimized out when
2030            all references to them are dead.
2031          + A new inter-procedural static profile estimation pass detects
2032            functions that are executed once or unlikely to be executed.
2033            Unlikely executed functions are optimized for size. Functions
2034            executed once are optimized for size except for the inner
2035            loops.
2036          + On most targets with named section support, functions used
2037            only at startup (static constructors and main), functions used
2038            only at exit and functions detected to be cold are placed into
2039            separate text segment subsections. This extends the
2040            -freorder-functions feature and is controlled by the same
2041            switch. The goal is to improve the startup time of large C++
2042            programs.
2043            Proper function placement requires linker support. GNU ld
2044            2.21.51 on ELF targets was updated to place those functions
2045            together within the text section leading to better code
2046            locality and faster startup times of large C++ programs. The
2047            feature is also supported in the Apple linker. Support in the
2048            gold linker is planned.
2049     * A new switch -fstack-usage has been added. It makes the compiler
2050       output stack usage information for the program, on a per-function
2051       basis, in an auxiliary file.
2052     * A new switch -fcombine-stack-adjustments has been added. It can be
2053       used to enable or disable the compiler's stack-slot combining pass
2054       which before was enabled automatically at -O1 and above, but could
2055       not be controlled on its own.
2056     * A new switch -fstrict-volatile-bitfields has been added. Using it
2057       indicates that accesses to volatile bitfields should use a single
2058       access of the width of the field's type. This option can be useful
2059       for precisely defining and accessing memory-mapped peripheral
2060       registers from C or C++.
2061
2062Compile time and memory usage improvements
2063
2064     * Datastructures used by the dataflow framework in GCC were
2065       reorganized for better memory usage and more cache locality.
2066       Compile time is improved especially on units with large functions
2067       (possibly resulting from a lot of inlining) not fitting into the
2068       processor cache. The compile time of the GCC C compiler binary with
2069       link-time optimization went down by over 10% (benchmarked on x86-64
2070       target).
2071
2072New Languages and Language specific improvements
2073
2074  Ada
2075
2076     * Stack checking has been improved on selected architectures (Alpha,
2077       IA-32/x86-64, RS/6000 and SPARC): it now will detect stack
2078       overflows in all cases on these architectures.
2079     * Initial support for Ada 2012 has been added.
2080
2081  C family
2082
2083     * A new warning, enabled by -Wdouble-promotion, has been added that
2084       warns about cases where a value of type float is implicitly
2085       promoted to double. This is especially helpful for CPUs that handle
2086       the former in hardware, but emulate the latter in software.
2087     * A new function attribute leaf was introduced. This attribute allows
2088       better inter-procedural optimization across calls to functions that
2089       return to the current unit only via returning or exception
2090       handling. This is the case for most library functions that have no
2091       callbacks.
2092     * Support for a new data type __int128 for targets having wide enough
2093       machine-mode support.
2094     * The new function attribute callee_pop_aggregate allows to specify
2095       if the caller or callee is responsible for popping the aggregate
2096       return pointer value from the stack.
2097     * Support for selectively enabling and disabling warnings via #pragma
2098       GCC diagnostic has been added. For instance:
2099#pragma GCC diagnostic error "-Wuninitialized"
2100  foo(a);                       /* error is given for this one */
2101#pragma GCC diagnostic push
2102#pragma GCC diagnostic ignored "-Wuninitialized"
2103  foo(b);                       /* no diagnostic for this one */
2104#pragma GCC diagnostic pop
2105  foo(c);                       /* error is given for this one */
2106#pragma GCC diagnostic pop
2107  foo(d);                       /* depends on command line options */
2108
2109     * The -fmax-errors=N option is now supported. Using this option
2110       causes the compiler to exit after N errors have been issued.
2111
2112  C
2113
2114     * There is now experimental support for some features from the
2115       upcoming C1X revision of the ISO C standard. This support may be
2116       selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions.
2117       Note that this support is experimental and may change incompatibly
2118       in future releases for consistency with changes to the C1X standard
2119       draft. The following features are newly supported as described in
2120       the N1539 draft of C1X (with changes agreed at the March 2011 WG14
2121       meeting); some other features were already supported with no
2122       compiler changes being needed, or have some support but not in full
2123       accord with N1539 (as amended).
2124          + Static assertions (_Static_assert keyword)
2125          + Typedef redefinition
2126          + New macros in <float.h>
2127          + Anonymous structures and unions
2128     * The new -fplan9-extensions option directs the compiler to support
2129       some extensions for anonymous struct fields which are implemented
2130       by the Plan 9 compiler. A pointer to a struct may be automatically
2131       converted to a pointer to an anonymous field when calling a
2132       function, in order to make the types match. An anonymous struct
2133       field whose type is a typedef name may be referred to using the
2134       typedef name.
2135
2136  C++
2137
2138     * Improved [7]experimental support for the upcoming C++0x ISO C++
2139       standard, including support for constexpr (thanks to Gabriel Dos
2140       Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide),
2141       noexcept, unrestricted unions, range-based for loops (thanks to
2142       Rodrigo Rivas Costa), opaque enum declarations (thanks also to
2143       Rodrigo), implicitly deleted functions and implicit move
2144       constructors.
2145     * When an extern declaration within a function does not match a
2146       declaration in the enclosing context, G++ now properly declares the
2147       name within the namespace of the function rather than the namespace
2148       which was open just before the function definition ([8]c++/43145).
2149     * GCC now warns by default when casting integers to larger pointer
2150       types. These warnings can be disabled with the option
2151       -Wno-int-to-pointer-cast, which is now also available in C++.
2152     * G++ no longer optimizes using the assumption that a value of
2153       enumeration type will fall within the range specified by the
2154       standard, since that assumption is easily violated with a
2155       conversion from integer type ([9]c++/43680). The old behavior can
2156       be restored with -fstrict-enums.
2157     * The new -fnothrow-opt flag changes the semantics of a throw()
2158       exception specification to match the proposed semantics of the
2159       noexcept specification: just call terminate if an exception tries
2160       to propagate out of a function with such an exception
2161       specification. This dramatically reduces or eliminates the code
2162       size overhead from adding the exception specification.
2163     * The new -Wnoexcept flag will suggest adding a noexcept qualifier to
2164       a function that the compiler can tell doesn't throw if it would
2165       change the value of a noexcept expression.
2166     * The -Wshadow option now warns if a local variable or type
2167       declaration shadows another type in C++. Note that the compiler
2168       will not warn if a local variable shadows a struct/class/enum, but
2169       will warn if it shadows an explicit typedef.
2170     * When an identifier is not found in the current scope, G++ now
2171       offers suggestions about which identifier might have been intended.
2172     * G++ now issues clearer diagnostics for missing semicolons after
2173       class, struct, and union definitions.
2174     * G++ now issues clearer diagnostics for missing semicolons after
2175       class member declarations.
2176     * G++ now issues clearer diagnostics when a colon is used in a place
2177       where a double-colon was intended.
2178     * G++ no longer accepts mutable on reference members ([10]c++/33558).
2179       Use -fpermissive to allow the old, non-conforming behaviour.
2180     * A few mangling fixes have been made, to attribute const/volatile on
2181       function pointer types, decltype of a plain decl, and use of a
2182       function parameter in the declaration of another parameter. By
2183       default the compiler still uses the old mangling, but emits aliases
2184       with the new mangling on targets that support strong aliases. Users
2185       can switch over entirely to the new mangling with -fabi-version=5
2186       or -fabi-version=0. -Wabi will now warn about code that uses the
2187       old mangling.
2188     * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified
2189       type to be default initialized unless the type has a user-declared
2190       default constructor. In 4.6.2 G++ implements the proposed
2191       resolution of [11]DR 253, so default initialization is allowed if
2192       it initializes all subobjects. Code that fails to compile can be
2193       fixed by providing an initializer e.g.
2194    struct A { A(); };
2195    struct B : A { int i; };
2196    const B b = B();
2197       Use -fpermissive to allow the old, non-conforming behaviour.
2198
2199    Runtime Library (libstdc++)
2200
2201     * [12]Improved experimental support for the upcoming ISO C++
2202       standard, C++0x, including using constexpr and nullptr.
2203     * Performance improvements to the [13]Debug Mode, thanks to Franc,ois
2204       Dumont.
2205     * Atomic operations used for reference-counting are annotated so that
2206       they can be understood by race detectors such as Helgrind, see
2207       [14]Data Race Hunting.
2208     * Most libstdc++ standard headers have been changed to no longer
2209       include the cstddef header as an implementation detail. Code that
2210       relied on that header being included as side-effect of including
2211       other standard headers will need to include cstddef explicitly.
2212
2213  Fortran
2214
2215     * On systems supporting the libquadmath library, GNU Fortran now also
2216       supports a quad-precision, kind=16 floating-point data type
2217       (REAL(16), COMPLEX(16)). As the data type is not fully supported in
2218       hardware, calculations might be one to two orders of magnitude
2219       slower than with the 4, 8 or 10 bytes floating-point data types.
2220       This change does not affect systems which support REAL(16) in
2221       hardware nor those which do not support libquadmath.
2222     * Much improved compile time for large array constructors.
2223     * In order to reduce execution time and memory consumption, use of
2224       temporary arrays in assignment expressions is avoided for many
2225       cases. The compiler now reverses loops in order to avoid generating
2226       a temporary array where possible.
2227     * Improved diagnostics, especially with -fwhole-file.
2228     * The -fwhole-file flag is now enabled by default. This improves code
2229       generation and diagnostics. It can be disabled using the deprecated
2230       -fno-whole-file flag.
2231     * Support the generation of Makefile dependencies via the [15]-M...
2232       flags of GCC; you may need to specify the -cpp option in addition.
2233       The dependencies take modules, Fortran's include, and CPP's
2234       #include into account. Note: Using -M for the module path is no
2235       longer supported, use -J instead.
2236     * The flag -Wconversion has been modified to only issue warnings
2237       where a conversion leads to information loss. This drastically
2238       reduces the number of warnings; -Wconversion is thus now enabled
2239       with -Wall. The flag -Wconversion-extra has been added and also
2240       warns about other conversions; -Wconversion-extra typically issues
2241       a huge number of warnings, most of which can be ignored.
2242     * A new command-line option -Wunused-dummy-argument warns about
2243       unused dummy arguments and is included in -Wall. Before,
2244       -Wunused-variable also warned about unused dummy arguments.
2245     * Fortran 2003 support has been extended:
2246          + Improved support for polymorphism between libraries and
2247            programs and for complicated inheritance patterns (cf.
2248            [16]object-oriented programming).
2249          + Experimental support of the ASSOCIATE construct.
2250          + In pointer assignments it is now possible to specify the lower
2251            bounds of the pointer and, for a rank-1 or a simply contiguous
2252            data-target, to remap the bounds.
2253          + Automatic (re)allocation: In intrinsic assignments to
2254            allocatable variables the left-hand side will be automatically
2255            allocated (if unallocated) or reallocated (if the shape or
2256            type parameter is different). To avoid the small performance
2257            penalty, you can use a(:) = ... instead of a = ... for arrays
2258            and character strings - or disable the feature using -std=f95
2259            or -fno-realloc-lhs.
2260          + Deferred type parameter: For scalar allocatable and pointer
2261            variables the character length can be deferred.
2262          + Namelist variables with allocatable and pointer attribute and
2263            nonconstant length type parameter are supported.
2264     * Fortran 2008 support has been extended:
2265          + Experimental [17]coarray support (for one image only, i.e.
2266            num_images() == 1); use the [18]-fcoarray=single flag to
2267            enable it.
2268          + The STOP and the new ERROR STOP statements now support all
2269            constant expressions.
2270          + Support for the CONTIGUOUS attribute.
2271          + Support for ALLOCATE with MOLD.
2272          + Support for the STORAGE_SIZE intrinsic inquiry function.
2273          + Support of the NORM2 and PARITY intrinsic functions.
2274          + The following bit intrinsics were added: POPCNT and POPPAR for
2275            counting the number of 1 bits and returning the parity; BGE,
2276            BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR
2277            for combined left and right shifts, MASKL and MASKR for simple
2278            left and right justified masks, MERGE_BITS for a bitwise merge
2279            using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations,
2280            and the transformational bit intrinsics IALL, IANY and
2281            IPARITY.
2282          + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine.
2283          + Support for the IMPURE attribute for procedures, which allows
2284            for ELEMENTAL procedures without the restrictions of PURE.
2285          + Null pointers (including NULL()) and not allocated variables
2286            can be used as actual argument to optional non-pointer,
2287            non-allocatable dummy arguments, denoting an absent argument.
2288          + Non-pointer variables with TARGET attribute can be used as
2289            actual argument to POINTER dummies with INTENT(IN)
2290          + Pointers including procedure pointers and those in a derived
2291            type (pointer components) can now be initialized by a target
2292            instead of only by NULL.
2293          + The EXIT statement (with construct-name) can now be used to
2294            leave not only the DO but also the ASSOCIATE, BLOCK, IF,
2295            SELECT CASE and SELECT TYPE constructs.
2296          + Internal procedures can now be used as actual argument.
2297          + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS
2298            and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV
2299            have been added; these arrays contain the supported kind
2300            values for the respective types.
2301          + The module procedures C_SIZEOF of the intrinsic module
2302            ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of
2303            ISO_FORTRAN_ENV have been implemented.
2304          + Minor changes: obsolescence diagnostics for ENTRY was added
2305            for -std=f2008; a line may start with a semicolon; for
2306            internal and module procedures END can be used instead of END
2307            SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes
2308            a RADIX argument; intrinsic types are supported for
2309            TYPE(intrinsic-type-spec); multiple type-bound procedures can
2310            be declared in a single PROCEDURE statement; implied-shape
2311            arrays are supported for named constants (PARAMETER). The
2312            transformational, three argument versions of BESSEL_JN and
2313            BESSEL_YN were added - the elemental, two-argument version had
2314            been added in GCC 4.4; note that the transformational
2315            functions use a recurrence algorithm.
2316
2317  Go
2318
2319   Support for the [19]Go programming language has been added to GCC. It
2320   is not enabled by default when you build GCC; use the
2321   --enable-languages configure option to build it. The driver program for
2322   compiling Go code is gccgo.
2323
2324   Go is currently known to work on GNU/Linux and RTEMS. Solaris support
2325   is in progress. It may or may not work on other platforms.
2326
2327  Objective-C and Objective-C++
2328
2329     * The -fobjc-exceptions flag is now required to enable Objective-C
2330       exception and synchronization syntax (introduced by the keywords
2331       @try, @catch, @finally and @synchronized).
2332     * A number of Objective-C 2.0 features and extensions are now
2333       supported by GCC. These features are enabled by default; you can
2334       disable them by using the new -fobjc-std=objc1 command-line option.
2335     * The Objective-C 2.0 dot-syntax is now supported. It is an
2336       alternative syntax for using getters and setters; object.count is
2337       automatically converted into [object count] or [object setCount:
2338       ...] depending on context; for example if (object.count > 0) is
2339       automatically compiled into the equivalent of if ([object count] >
2340       0) while object.count = 0; is automatically compiled into the
2341       equivalent ot [object setCount: 0];. The dot-syntax can be used
2342       with instance and class objects and with any setters or getters, no
2343       matter if they are part of a declared property or not.
2344     * Objective-C 2.0 declared properties are now supported. They are
2345       declared using the new @property keyword, and are most commonly
2346       used in conjunction with the new Objective-C 2.0 dot-syntax. The
2347       nonatomic, readonly, readwrite, assign, retain, copy, setter and
2348       getter attributes are all supported. Marking declared properties
2349       with __attribute__ ((deprecated)) is supported too.
2350     * The Objective-C 2.0 @synthesize and @dynamic keywords are
2351       supported. @synthesize causes the compiler to automatically
2352       synthesize a declared property, while @dynamic is used to disable
2353       all warnings for a declared property for which no implementation is
2354       provided at compile time. Synthesizing declared properties requires
2355       runtime support in most useful cases; to be able to use it with the
2356       GNU runtime, appropriate helper functions have been added to the
2357       GNU Objective-C runtime ABI, and are implemented by the GNU
2358       Objective-C runtime library shipped with GCC.
2359     * The Objective-C 2.0 fast enumeration syntax is supported in
2360       Objective-C. This is currently not yet available in Objective-C++.
2361       Fast enumeration requires support in the runtime, and such support
2362       has been added to the GNU Objective-C runtime library (shipped with
2363       GCC).
2364     * The Objective-C 2.0 @optional keyword is supported. It allows you
2365       to mark methods or properties in a protocol as optional as opposed
2366       to required.
2367     * The Objective-C 2.0 @package keyword is supported. It has currently
2368       the same effect as the @public keyword.
2369     * Objective-C 2.0 method attributes are supported. Currently the
2370       supported attributes are deprecated, sentinel, noreturn and format.
2371     * Objective-C 2.0 method argument attributes are supported. The most
2372       widely used attribute is unused, to mark an argument as unused in
2373       the implementation.
2374     * Objective-C 2.0 class and protocol attributes are supported.
2375       Currently the only supported attribute is deprecated.
2376     * Objective-C 2.0 class extensions are supported. A class extension
2377       has the same syntax as a category declaration with no category
2378       name, and the methods and properties declared in it are added
2379       directly to the main class. It is mostly used as an alternative to
2380       a category to add methods to a class without advertising them in
2381       the public headers, with the advantage that for class extensions
2382       the compiler checks that all the privately declared methods are
2383       actually implemented.
2384     * As a result of these enhancements, GCC can now be used to build
2385       Objective-C and Objective-C++ software that uses Foundation and
2386       other important system frameworks with the NeXT runtime on Darwin 9
2387       and Darwin 10 (OSX 10.5 and 10.6).
2388     * Many bugs in the compiler have been fixed in this release; in
2389       particular, LTO can now be used when compiling Objective-C and
2390       Objective-C++ and the parser is much more robust in dealing with
2391       invalid code.
2392
2393    Runtime Library (libobjc)
2394
2395     * The GNU Objective-C runtime library now defines the macro
2396       __GNU_LIBOBJC__ (with a value that is increased at every release
2397       where there is any change to the API) in objc/objc.h, making it
2398       easy to determine if the GNU Objective-C runtime library is being
2399       used, and if so, which version. Previous versions of the GNU
2400       Objective-C runtime library (and other Objective-C runtime
2401       libraries such as the Apple one) do not define this macro.
2402     * A new Objective-C 2.0 API, almost identical to the one implemented
2403       by the Apple Objective-C runtime, has been implemented in the GNU
2404       Objective-C runtime library. The new API hides the internals of
2405       most runtime structures but provides a more extensive set of
2406       functions to operate on them. It is much easier, for example, to
2407       create or modify classes at runtime. The new API also makes it
2408       easier to port software from Apple to GNU as almost no changes
2409       should be required. The old API is still supported for backwards
2410       compatibility; including the old objc/objc-api.h header file
2411       automatically selects the old API, while including the new
2412       objc/runtime.h header file automatically selects the new API.
2413       Support for the old API is being phased out and upgrading the
2414       software to use the new API is strongly recommended. To check for
2415       the availability of the new API, the __GNU_LIBOBJC__ macro can be
2416       used as older versions of the GNU Objective-C runtime library,
2417       which do not support the new API, do not define such a macro.
2418     * Runtime support for @synchronized has been added.
2419     * Runtime support for Objective-C 2.0 synthesized property accessors
2420       has been added.
2421     * Runtime support for Objective-C 2.0 fast enumeration has been
2422       added.
2423
2424New Targets and Target Specific Improvements
2425
2426  ARM
2427
2428     * GCC now supports the Cortex-M4 processor implementing the v7-em
2429       version of the architecture using the option -mcpu=cortex-m4.
2430     * Scheduling descriptions for the Cortex-M4, the Neon and the
2431       floating point units of the Cortex-A9 and a pipeline description
2432       for the Cortex-A5 have been added.
2433     * Synchronization primitives such as __sync_fetch_and_add and friends
2434       are now inlined for supported architectures rather than calling
2435       into a kernel helper function.
2436     * SSA loop prefetching is enabled by default for the Cortex-A9 at
2437       -O3.
2438     * Several improvements were committed to improve code generation for
2439       the ARM architecture including a rewritten implementation for load
2440       and store multiples.
2441     * Several enhancements were committed to improve SIMD code generation
2442       for NEON by adding support for widening instructions, misaligned
2443       loads and stores, vector conditionals and support for 64 bit
2444       arithmetic.
2445     * Support was added for the Faraday cores fa526, fa606te, fa626te,
2446       fmp626te, fmp626 and fa726te and can be used with the respective
2447       names as parameters to the -mcpu= option.
2448     * Basic support was added for Cortex-A15 and is available through
2449       -mcpu=cortex-a15.
2450     * GCC for AAPCS configurations now more closely adheres to the AAPCS
2451       specification by enabling -fstrict-volatile-bitfields by default.
2452
2453  IA-32/x86-64
2454
2455     * The new -fsplit-stack option permits programs to use a
2456       discontiguous stack. This is useful for threaded programs, in that
2457       it is no longer necessary to specify the maximum stack size when
2458       creating a thread. This feature is currently only implemented for
2459       32-bit and 64-bit x86 GNU/Linux targets.
2460     * Support for emitting profiler counter calls before function
2461       prologues. This is enabled via a new command-line option -mfentry.
2462     * Optimization for the Intel Core 2 processors is now available
2463       through the -march=core2 and -mtune=core2 options.
2464     * Support for Intel Core i3/i5/i7 processors is now available through
2465       the -march=corei7 and -mtune=corei7 options.
2466     * Support for Intel Core i3/i5/i7 processors with AVX is now
2467       available through the -march=corei7-avx and -mtune=corei7-avx
2468       options.
2469     * Support for AMD Bobcat (family 14) processors is now available
2470       through the -march=btver1 and -mtune=btver1 options.
2471     * Support for AMD Bulldozer (family 15) processors is now available
2472       through the -march=bdver1 and -mtune=bdver1 options.
2473     * The default setting (when not optimizing for size) for 32-bit
2474       GNU/Linux and Darwin x86 targets has been changed to
2475       -fomit-frame-pointer. The default can be reverted to
2476       -fno-omit-frame-pointer by configuring GCC with the
2477       --enable-frame-pointer configure option.
2478     * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support
2479       __float128 on 32-bit and 64-bit x86 targets.
2480     * AVX floating-point arithmetic can now be enabled by default at
2481       configure time with the new --with-fpmath=avx option.
2482     * The SSA loop prefetching pass is enabled when using -O3 when
2483       optimizing for CPUs where prefetching is beneficial (AMD CPUs newer
2484       than K6).
2485     * Support for TBM (Trailing Bit Manipulation) built-in functions and
2486       code generation is available via -mtbm.
2487     * Support for AMD's BMI (Bit Manipulation) built-in functions and
2488       code generation is available via -mbmi.
2489
2490  MicroBlaze
2491
2492     * Support has been added for the Xilinx MicroBlaze softcore processor
2493       (microblaze-elf) embedded target. This configurable processor is
2494       supported on several Xilinx Spartan and Virtex FPGAs.
2495
2496  MIPS
2497
2498     * GCC now supports the Loongson 3A processor. Its canonical -march=
2499       and -mtune= name is loongson3a.
2500
2501  MN10300 / AM33
2502
2503     * The inline assembly register constraint "A" has been renamed "c".
2504       This constraint is used to select a floating-point register that
2505       can be used as the destination of a multiply-accumulate
2506       instruction.
2507     * New inline assembly register constraints "A" and "D" have been
2508       added. These constraint letters resolve to all general registers
2509       when compiling for AM33, and resolve to address registers only or
2510       data registers only when compiling for MN10300.
2511     * The MDR register is represented in the compiler. One can access the
2512       register via the "z" constraint in inline assembly. It can be
2513       marked as clobbered or used as a local register variable via the
2514       "mdr" name. The compiler uses the RETF instruction if the function
2515       does not modify the MDR register, so it is important that inline
2516       assembly properly annotate any usage of the register.
2517
2518  PowerPC/PowerPC64
2519
2520     * GCC now supports the Applied Micro Titan processor with
2521       -mcpu=titan.
2522     * The -mrecip option has been added, which indicates whether the
2523       reciprocal and reciprocal square root instructions should be used.
2524     * The -mveclibabi=mass option can be used to enable the compiler to
2525       autovectorize mathematical functions using the Mathematical
2526       Acceleration Subsystem library.
2527     * The -msingle-pic-base option has been added, which instructs the
2528       compiler to avoid loading the PIC base register in function
2529       prologues. The PIC base register must be initialized by the runtime
2530       system.
2531     * The -mblock-move-inline-limit option has been added, which enables
2532       the user to control the maximum size of inlined memcpy calls and
2533       similar.
2534     * PowerPC64 GNU/Linux support for applications requiring a large TOC
2535       section has been improved. A new command-line option,
2536       -mcmodel=MODEL, controls this feature; valid values for MODEL are
2537       small, medium, or large.
2538     * The Altivec builtin functions vec_ld and vec_st have been modified
2539       to generate the Altivec memory instructions LVX and STVX, even if
2540       the -mvsx option is used. In the initial GCC 4.5 release, these
2541       builtin functions were changed to generate VSX memory reference
2542       instructions instead of Altivec memory instructions, but there are
2543       differences between the two instructions. If the VSX instruction
2544       set is available, you can now use the new builtin functions
2545       vec_vsx_ld and vec_vsx_st which always generates the VSX memory
2546       instructions.
2547     * The GCC compiler on AIX now defaults to a process layout with a
2548       larger data space allowing larger programs to be compiled.
2549     * The GCC long double type on AIX 6.1 and above has reverted to 64
2550       bit double precision, matching the AIX XL compiler default, because
2551       of missing C99 symbols required by the GCC runtime.
2552     * The default processor scheduling model and tuning for PowerPC64
2553       GNU/Linux and for AIX 6.1 and above now is POWER7.
2554     * Starting with GCC 4.6.1, vectors of type vector long long or vector
2555       long are passed and returned in the same method as other vectors
2556       with the VSX instruction set. Previously the GCC compiler did not
2557       adhere to the ABI for 128-bit vectors with 64-bit integer base
2558       types (PR 48857). This is also fixed in the GCC 4.5.4 release.
2559
2560  S/390, zSeries and System z9/z10, IBM zEnterprise z196
2561
2562     * Support for the zEnterprise z196 processor has been added. When
2563       using the -march=z196 option, the compiler will generate code
2564       making use of the following instruction facilities:
2565          + Conditional load/store
2566          + Distinct-operands
2567          + Floating-point-extension
2568          + Interlocked-access
2569          + Population-count
2570       The -mtune=z196 option avoids the compare and branch instructions
2571       as well as the load address instruction with an index register as
2572       much as possible and performs instruction scheduling appropriate
2573       for the new out-of-order pipeline architecture.
2574     * When using the -m31 -mzarch options the generated code still
2575       conforms to the 32-bit ABI but uses the general purpose registers
2576       as 64-bit registers internally. This requires a Linux kernel saving
2577       the whole 64-bit registers when doing a context switch. Kernels
2578       providing that feature indicate that by the 'highgprs' string in
2579       /proc/cpuinfo.
2580     * The SSA loop prefetching pass is enabled when using -O3.
2581
2582  SPARC
2583
2584     * GCC now supports the LEON series of SPARC V8 processors. The code
2585       generated by the compiler can either be tuned to it by means of the
2586       --with-tune=leon configure option and -mtune=leon compilation
2587       option, or the compiler can be built for the sparc-leon-{elf,linux}
2588       and sparc-leon3-{elf,linux} targets directly.
2589     * GCC has stopped sign/zero-extending parameter registers in the
2590       callee for functions taking parameters with sub-word size in 32-bit
2591       mode, since this is redundant with the specification of the ABI.
2592       GCC has never done so in 64-bit mode since this is also redundant.
2593     * The command line option -mfix-at697f has been added to enable the
2594       documented workaround for the single erratum of the Atmel AT697F
2595       processor.
2596
2597Operating Systems
2598
2599  Android
2600
2601     * GCC now supports the Bionic C library and provides a convenient way
2602       of building native libraries and applications for the Android
2603       platform. Refer to the documentation of the -mandroid and -mbionic
2604       options for details on building native code. At the moment, Android
2605       support is enabled only for ARM.
2606
2607  Darwin/Mac OS X
2608
2609     * General
2610          + Initial support for CFString types has been added.
2611            This allows GCC to build projects including the system Core
2612            Foundation frameworks. The GCC Objective-C family supports
2613            CFString "toll-free bridged" as per the Mac OS X system tools.
2614            CFString is also recognized in the context of format
2615            attributes and arguments (see the documentation for format
2616            attributes for limitations). At present, 8-bit character types
2617            are supported.
2618          + Object file size reduction.
2619            The Darwin zeroed memory allocators have been re-written to
2620            make more use of .zerofill sections. For non-debug code, this
2621            can reduce object file size significantly.
2622          + Objective-C family 64-bit support (NeXT ABI 2).
2623            Initial support has been added to support 64-bit Objective-C
2624            code using the Darwin/OS X native (NeXT) runtime. ABI version
2625            2 will be selected automatically when 64-bit code is built.
2626          + Objective-C family 32-bit ABI 1.
2627            For 32-bit code ABI 1 is also now also allowed. At present it
2628            must be selected manually using -fobjc-abi-version=1 where
2629            applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6).
2630     * x86 Architecture
2631          + The -mdynamic-no-pic option has been enabled.
2632            Code supporting -mdynamic-no-pic optimization has been added
2633            and is applicable to -m32 builds. The compiler bootstrap uses
2634            the option where appropriate.
2635          + The default value for -mtune= has been changed.
2636            Since Darwin systems are primarily Xeon, Core-2 or similar the
2637            default tuning has been changed to -mtune=core2.
2638          + Enable 128-bit long double (__float128) support on Darwin.
2639     * PPC Architecture
2640          + Darwin64 ABI.
2641            Several significant bugs have been fixed, such that GCC now
2642            produces code compatible with the Darwin64 PowerPC ABI.
2643          + libffi and boehm-gc.
2644            The Darwin ports of the libffi and boehm-gc libraries have
2645            been upgraded to include a Darwin64 implementation. This means
2646            that powerpc*-*-darwin9 platforms may now, for example, build
2647            Java applications with -m64 enabled.
2648          + Plug-in support has been enabled.
2649          + The -fsection-anchors option is now available although,
2650            presently, not heavily tested.
2651
2652  Solaris 2
2653
2654    New Features
2655
2656     * Support symbol versioning with the Sun linker.
2657     * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+.
2658     * Support thread-local storage (TLS) with the Sun assembler on
2659       Solaris 2/x86.
2660     * Support TLS on Solaris 8/9 if prerequisites are met.
2661     * Support COMDAT group with the GNU assembler and recent Sun linker.
2662     * Support the Sun assembler visibility syntax.
2663     * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp.
2664       -march=pentiumpro (Solaris 8/9).
2665     * Don't use SSE on Solaris 8/9 x86 by default.
2666     * Enable 128-bit long double (__float128) support on Solaris 2/x86.
2667
2668    ABI Change
2669
2670     * Change the ABI for returning 8-byte vectors like __m64 in MMX
2671       registers on Solaris 10+/x86 to match the Sun Studio 12.1+
2672       compilers. This is an incompatible change. If you use such types,
2673       you must either recompile all your code with the new compiler or
2674       use the new -mvect8-ret-in-mem option to remain compatible with
2675       previous versions of GCC and Sun Studio.
2676
2677  Windows x86/x86_64
2678
2679     * Initial support for decimal floating point.
2680     * Support for the __thiscall calling-convention.
2681     * Support for hot-patchable function prologues via the
2682       ms_hook_prologue attribute for x86_64 in addition to 32-bit x86.
2683     * Improvements of stack-probing and stack-allocation mechanisms.
2684     * Support of push/pop-macro pragma as preprocessor command.
2685       With #pragma push_macro("macro-name") the current definition of
2686       macro-name is saved and can be restored with #pragma
2687       pop_macro("macro-name") to its saved definition.
2688     * Enable 128-bit long double (__float128) support on MinGW and
2689       Cygwin.
2690
2691Other significant improvements
2692
2693  Installation changes
2694
2695     * An install-strip make target is provided that installs stripped
2696       executables, and may install libraries with unneeded or debugging
2697       sections stripped.
2698     * On Power7 systems, there is a potential problem if you build the
2699       GCC compiler with a host compiler using options that enable the VSX
2700       instruction set generation. If the host compiler has been patched
2701       so that the vec_ld and vec_st builtin functions generate Altivec
2702       memory instructions instead of VSX memory instructions, then you
2703       should be able to build the compiler with VSX instruction
2704       generation.
2705
2706Changes for GCC Developers
2707
2708   Note: these changes concern developers that develop GCC itself or
2709   software that integrates with GCC, such as plugins, and not the general
2710   GCC users.
2711     * The gengtype utility, which previously was internal to the GCC
2712       build process, has been enchanced to provide GC root information
2713       for plugins as necessary.
2714     * The old GC allocation interface of ggc_alloc and friends was
2715       replaced with a type-safe alternative.
2716
2717GCC 4.6.1
2718
2719   This is the [20]list of problem reports (PRs) from GCC's bug tracking
2720   system that are known to be fixed in the 4.6.1 release. This list might
2721   not be complete (that is, it is possible that some PRs that have been
2722   fixed are not listed here).
2723
2724GCC 4.6.2
2725
2726   This is the [21]list of problem reports (PRs) from GCC's bug tracking
2727   system that are known to be fixed in the 4.6.2 release. This list might
2728   not be complete (that is, it is possible that some PRs that have been
2729   fixed are not listed here).
2730
2731GCC 4.6.3
2732
2733   This is the [22]list of problem reports (PRs) from GCC's bug tracking
2734   system that are known to be fixed in the 4.6.3 release. This list might
2735   not be complete (that is, it is possible that some PRs that have been
2736   fixed are not listed here).
2737
2738GCC 4.6.4
2739
2740   This is the [23]list of problem reports (PRs) from GCC's bug tracking
2741   system that are known to be fixed in the 4.6.4 release. This list might
2742   not be complete (that is, it is possible that some PRs that have been
2743   fixed are not listed here).
2744
2745
2746    For questions related to the use of GCC, please consult these web
2747    pages and the [24]GCC manuals. If that fails, the
2748    [25]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2749    web pages and the development of GCC are welcome on our developer
2750    list at [26]gcc@gcc.gnu.org. All of [27]our lists have public
2751    archives.
2752
2753   Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and
2754   distribution of this entire article is permitted in any medium,
2755   provided this notice is preserved.
2756
2757   These pages are [29]maintained by the GCC team. Last modified
2758   2013-04-12[30].
2759
2760References
2761
2762   1. http://sourceware.org/bugzilla/show_bug.cgi?id=10401
2763   2. http://gcc.gnu.org/gcc-4.5/changes.html
2764   3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted
2765   4. http://gcc.gnu.org/gcc-4.6/porting_to.html
2766   5. http://gcc.gnu.org/projects/lto/whopr.pdf
2767   6. http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
2768   7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html
2769   8. http://gcc.gnu.org/PR43145
2770   9. http://gcc.gnu.org/PR43680
2771  10. http://gcc.gnu.org/PR33558
2772  11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253
2773  12. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.200x
2774  13. http://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
2775  14. http://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races
2776  15. http://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
2777  16. http://gcc.gnu.org/wiki/OOP
2778  17. http://gcc.gnu.org/wiki/Coarray
2779  18. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233
2780  19. http://golang.org/
2781  20. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1
2782  21. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2
2783  22. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3
2784  23. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4
2785  24. http://gcc.gnu.org/onlinedocs/
2786  25. mailto:gcc-help@gcc.gnu.org
2787  26. mailto:gcc@gcc.gnu.org
2788  27. http://gcc.gnu.org/lists.html
2789  28. http://www.fsf.org/
2790  29. http://gcc.gnu.org/about.html
2791  30. http://validator.w3.org/check/referer
2792======================================================================
2793http://gcc.gnu.org/gcc-4.5/index.html
2794
2795                             GCC 4.5 Release Series
2796
2797   Jul 2, 2012
2798
2799   The [1]GNU project and the GCC developers are pleased to announce the
2800   release of GCC 4.5.4.
2801
2802   This release is a bug-fix release, containing fixes for regressions in
2803   GCC 4.5.3 relative to previous releases of GCC.
2804
2805Release History
2806
2807   GCC 4.5.4
2808          Jul 2, 2012 ([2]changes)
2809
2810   GCC 4.5.3
2811          Apr 28, 2011 ([3]changes)
2812
2813   GCC 4.5.2
2814          Dec 16, 2010 ([4]changes)
2815
2816   GCC 4.5.1
2817          Jul 31, 2010 ([5]changes)
2818
2819   GCC 4.5.0
2820          April 14, 2010 ([6]changes)
2821
2822References and Acknowledgements
2823
2824   GCC used to stand for the GNU C Compiler, but since the compiler
2825   supports several other languages aside from C, it now stands for the
2826   GNU Compiler Collection.
2827
2828   A list of [7]successful builds is updated as new information becomes
2829   available.
2830
2831   The GCC developers would like to thank the numerous people that have
2832   contributed new features, improvements, bug fixes, and other changes as
2833   well as test results to GCC. This [8]amazing group of volunteers is
2834   what makes GCC successful.
2835
2836   For additional information about GCC please refer to the [9]GCC project
2837   web site or contact the [10]GCC development mailing list.
2838
2839   To obtain GCC please use [11]our mirror sites or [12]our SVN server.
2840
2841
2842    For questions related to the use of GCC, please consult these web
2843    pages and the [13]GCC manuals. If that fails, the
2844    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2845    web pages and the development of GCC are welcome on our developer
2846    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
2847    archives.
2848
2849   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
2850   distribution of this entire article is permitted in any medium,
2851   provided this notice is preserved.
2852
2853   These pages are [18]maintained by the GCC team. Last modified
2854   2012-11-02[19].
2855
2856References
2857
2858   1. http://www.gnu.org/
2859   2. http://gcc.gnu.org/gcc-4.5/changes.html
2860   3. http://gcc.gnu.org/gcc-4.5/changes.html
2861   4. http://gcc.gnu.org/gcc-4.5/changes.html
2862   5. http://gcc.gnu.org/gcc-4.5/changes.html
2863   6. http://gcc.gnu.org/gcc-4.5/changes.html
2864   7. http://gcc.gnu.org/gcc-4.5/buildstat.html
2865   8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
2866   9. http://gcc.gnu.org/index.html
2867  10. mailto:gcc@gcc.gnu.org
2868  11. http://gcc.gnu.org/mirrors.html
2869  12. http://gcc.gnu.org/svn.html
2870  13. http://gcc.gnu.org/onlinedocs/
2871  14. mailto:gcc-help@gcc.gnu.org
2872  15. mailto:gcc@gcc.gnu.org
2873  16. http://gcc.gnu.org/lists.html
2874  17. http://www.fsf.org/
2875  18. http://gcc.gnu.org/about.html
2876  19. http://validator.w3.org/check/referer
2877======================================================================
2878http://gcc.gnu.org/gcc-4.5/changes.html
2879
2880                             GCC 4.5 Release Series
2881                        Changes, New Features, and Fixes
2882
2883Caveats
2884
2885     * GCC now requires the [1]MPC library in order to build. See the
2886       [2]prerequisites page for version requirements.
2887     * Support for a number of older systems and recently unmaintained or
2888       untested target ports of GCC has been declared obsolete in GCC 4.5.
2889       Unless there is activity to revive them, the next release of GCC
2890       will have their sources permanently removed.
2891       The following ports for individual systems on particular
2892       architectures have been obsoleted:
2893          + IRIX releases before 6.5 (mips-sgi-irix5*,
2894            mips-sgi-irix6.[0-4])
2895          + Solaris 7 (*-*-solaris2.7)
2896          + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*,
2897            alpha-dec-osf5.0*)
2898          + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions
2899            can be found in the [3]announcement.
2900       Support for the classic POWER architecture implemented in the
2901       original RIOS and RIOS2 processors of the old IBM RS/6000 product
2902       line has been obsoleted in the rs6000 port. This does not affect
2903       the new generation Power and PowerPC architectures.
2904     * Support has been removed for all the [4]configurations obsoleted in
2905       GCC 4.4.
2906     * Support has been removed for the protoize and unprotoize utilities,
2907       obsoleted in GCC 4.4.
2908     * Support has been removed for tuning for Itanium1 (Merced) variants.
2909       Note that code tuned for Itanium2 should also run correctly on
2910       Itanium1.
2911     * GCC now generates unwind info also for epilogues. DWARF debuginfo
2912       generated by GCC now uses more features of DWARF3 than before, and
2913       also some DWARF4 features. GDB older than 7.0 is not able to handle
2914       either of these, so to debug GCC 4.5 generated binaries or
2915       libraries GDB 7.0 or later is needed. You can disable use of DWARF4
2916       features with the -gdwarf-3 -gstrict-dwarf options, or use
2917       -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but
2918       epilogue unwind info is emitted unconditionally whenever unwind
2919       info is emitted.
2920     * On x86 targets, code containing floating-point calculations may run
2921       significantly slower when compiled with GCC 4.5 in strict C99
2922       conformance mode than they did with earlier GCC versions. This is
2923       due to stricter standard conformance of the compiler and can be
2924       avoided by using the option -fexcess-precision=fast; also see
2925       [5]below.
2926     * The function attribute noinline no longer prevents GCC from cloning
2927       the function. A new attribute noclone has been introduced for this
2928       purpose. Cloning a function means that it is duplicated and the new
2929       copy is specialized for certain contexts (for example when a
2930       parameter is a known constant).
2931
2932General Optimizer Improvements
2933
2934     * The -save-temps now takes an optional argument. The -save-temps and
2935       -save-temps=cwd switches write the temporary files in the current
2936       working directory based on the original source file. The
2937       -save-temps=obj switch will write files into the directory
2938       specified with the -o option, and the intermediate filenames are
2939       based on the output file. This will allow the user to get the
2940       compiler intermediate files when doing parallel builds without two
2941       builds of the same filename located in different directories from
2942       interfering with each other.
2943     * Debugging dumps are now created in the same directory as the object
2944       file rather than in the current working directory. This allows the
2945       user to get debugging dumps when doing parallel builds without two
2946       builds of the same filename interfering with each other.
2947     * GCC has been integrated with the [6]MPC library. This allows GCC to
2948       evaluate complex arithmetic at compile time [7]more accurately. It
2949       also allows GCC to evaluate calls to complex built-in math
2950       functions having constant arguments and replace them at compile
2951       time with their mathematically equivalent results. In doing so, GCC
2952       can generate correct results regardless of the math library
2953       implementation or floating point precision of the host platform.
2954       This also allows GCC to generate identical results regardless of
2955       whether one compiles in native or cross-compile configurations to a
2956       particular target. The following built-in functions take advantage
2957       of this new capability: cacos, cacosh, casin, casinh, catan,
2958       catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan,
2959       and ctanh. The float and long double variants of these functions
2960       (e.g. csinf and csinl) are also handled.
2961     * A new link-time optimizer has been added ([8]-flto). When this
2962       option is used, GCC generates a bytecode representation of each
2963       input file and writes it to specially-named sections in each object
2964       file. When the object files are linked together, all the function
2965       bodies are read from these named sections and instantiated as if
2966       they had been part of the same translation unit. This enables
2967       interprocedural optimizations to work across different files (and
2968       even different languages), potentially improving the performance of
2969       the generated code. To use the link-timer optimizer, -flto needs to
2970       be specified at compile time and during the final link. If the
2971       program does not require any symbols to be exported, it is possible
2972       to combine -flto and the experimental [9]-fwhopr with
2973       [10]-fwhole-program to allow the interprocedural optimizers to use
2974       more aggressive assumptions.
2975     * The automatic parallelization pass was enhanced to support
2976       parallelization of outer loops.
2977     * Automatic parallelization can be enabled as part of Graphite. In
2978       addition to -ftree-parallelize-loops=, specify
2979       -floop-parallelize-all to enable the Graphite-based optimization.
2980     * The infrastructure for optimizing based on [11]restrict qualified
2981       pointers has been rewritten and should result in code generation
2982       improvements. Optimizations based on restrict qualified pointers
2983       are now also available when using -fno-strict-aliasing.
2984     * There is a new optimization pass that attempts to change prototype
2985       of functions to avoid unused parameters, pass only relevant parts
2986       of structures and turn arguments passed by reference to arguments
2987       passed by value when possible. It is enabled by -O2 and above as
2988       well as -Os and can be manually invoked using the new command-line
2989       switch -fipa-sra.
2990     * GCC now optimize exception handling code. In particular cleanup
2991       regions that are proved to not have any effect are optimized out.
2992
2993New Languages and Language specific improvements
2994
2995  All languages
2996
2997     * The -fshow-column option is now on by default. This means error
2998       messages now have a column associated with them.
2999
3000  Ada
3001
3002     * Compilation of programs heavily using discriminated record types
3003       with variant parts has been sped up and generates more compact
3004       code.
3005     * Stack checking now works reasonably well on most plaforms. In some
3006       specific cases, stack overflows may still fail to be detected, but
3007       a compile-time warning will be issued for these cases.
3008
3009  C family
3010
3011     * If a header named in a #include directive is not found, the
3012       compiler exits immediately. This avoids a cascade of errors arising
3013       from declarations expected to be found in that header being
3014       missing.
3015     * A new built-in function __builtin_unreachable() has been added that
3016       tells the compiler that control will never reach that point. It may
3017       be used after asm statements that terminate by transferring control
3018       elsewhere, and in other places that are known to be unreachable.
3019     * The -Wlogical-op option now warns for logical expressions such as
3020       (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be
3021       mistakes. This option is disabled by default.
3022     * An asm goto feature has been added to allow asm statements that
3023       jump to C labels.
3024     * C++0x raw strings are supported for C++ and for C with -std=gnu99.
3025     * The deprecated attribute now takes an optional string argument, for
3026       example, __attribute__((deprecated("text string"))), that will be
3027       printed together with the deprecation warning.
3028
3029  C
3030
3031     * The -Wenum-compare option, which warns when comparing values of
3032       different enum types, now works for C. It formerly only worked for
3033       C++. This warning is enabled by -Wall. It may be avoided by using a
3034       type cast.
3035     * The -Wcast-qual option now warns about casts which are unsafe in
3036       that they permit const-correctness to be violated without further
3037       warnings. Specifically, it warns about cases where a qualifier is
3038       added when all the lower types are not const. For example, it warns
3039       about a cast from char ** to const char **.
3040     * The -Wc++-compat option is significantly improved. It issues new
3041       warnings for:
3042          + Using C++ reserved operator names as identifiers.
3043          + Conversions to enum types without explicit casts.
3044          + Using va_arg with an enum type.
3045          + Using different enum types in the two branches of ?:.
3046          + Using ++ or -- on a variable of enum type.
3047          + Using the same name as both a struct, union or enum tag and a
3048            typedef, unless the typedef refers to the tagged type itself.
3049          + Using a struct, union, or enum which is defined within another
3050            struct or union.
3051          + A struct field defined using a typedef if there is a field in
3052            the struct, or an enclosing struct, whose name is the typedef
3053            name.
3054          + Duplicate definitions at file scope.
3055          + Uninitialized const variables.
3056          + A global variable with an anonymous struct, union, or enum
3057            type.
3058          + Using a string constant to initialize a char array whose size
3059            is the length of the string.
3060     * The new -Wjump-misses-init option warns about cases where a goto or
3061       switch skips the initialization of a variable. This sort of branch
3062       is an error in C++ but not in C. This warning is enabled by
3063       -Wc++-compat.
3064     * GCC now ensures that a C99-conforming <stdint.h> is present on most
3065       targets, and uses information about the types in this header to
3066       implement the Fortran bindings to those types. GCC does not ensure
3067       the presence of such a header, and does not implement the Fortran
3068       bindings, on the following targets: NetBSD, VxWorks, VMS,
3069       SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF.
3070     * GCC now implements C90- and C99-conforming rules for constant
3071       expressions. This may cause warnings or errors for some code using
3072       expressions that can be folded to a constant but are not constant
3073       expressions as defined by ISO C.
3074     * All known target-independent C90 and C90 Amendment 1 conformance
3075       bugs, and all known target-independent C99 conformance bugs not
3076       related to floating point or extended identifiers, have been fixed.
3077     * The C decimal floating point support now includes support for the
3078       FLOAT_CONST_DECIMAL64 pragma.
3079     * The named address space feature from ISO/IEC TR 18037 is now
3080       supported. This is currently only implemented for the SPU
3081       processor.
3082
3083  C++
3084
3085     * Improved [12]experimental support for the upcoming C++0x ISO C++
3086       standard, including support for raw strings, lambda expressions and
3087       explicit type conversion operators.
3088     * When printing the name of a class template specialization, G++ will
3089       now omit any template arguments which come from default template
3090       arguments. This behavior (and the pretty-printing of function
3091       template specializations as template signature and arguments) can
3092       be disabled with the -fno-pretty-templates option.
3093     * Access control is now applied to typedef names used in a template,
3094       which may cause G++ to reject some ill-formed code that was
3095       accepted by earlier releases. The -fno-access-control option can be
3096       used as a temporary workaround until the code is corrected.
3097     * Compilation time for code that uses templates should now scale
3098       linearly with the number of instantiations rather than
3099       quadratically, as template instantiations are now looked up using
3100       hash tables.
3101     * Declarations of functions that look like builtin declarations of
3102       library functions are only considered to be redeclarations if they
3103       are declared with extern "C". This may cause problems with code
3104       that omits extern "C" on hand-written declarations of C library
3105       functions such as abort or memcpy. Such code is ill-formed, but was
3106       accepted by earlier releases.
3107     * Diagnostics that used to complain about passing non-POD types to
3108       ... or jumping past the declaration of a non-POD variable now check
3109       for triviality rather than PODness, as per C++0x.
3110     * In C++0x mode local and anonymous classes are now allowed as
3111       template arguments, and in declarations of variables and functions
3112       with linkage, so long as any such declaration that is used is also
3113       defined ([13]DR 757).
3114     * Labels may now have attributes, as has been permitted for a while
3115       in C. This is only permitted when the label definition and the
3116       attribute specifier is followed by a semicolon--i.e., the label
3117       applies to an empty statement. The only useful attribute for a
3118       label is unused.
3119     * G++ now implements [14]DR 176. Previously G++ did not support using
3120       the injected-class-name of a template base class as a type name,
3121       and lookup of the name found the declaration of the template in the
3122       enclosing scope. Now lookup of the name finds the
3123       injected-class-name, which can be used either as a type or as a
3124       template, depending on whether or not the name is followed by a
3125       template argument list. As a result of this change, some code that
3126       was previously accepted may be ill-formed because
3127         1. The injected-class-name is not accessible because it's from a
3128            private base, or
3129         2. The injected-class-name cannot be used as an argument for a
3130            template template parameter.
3131       In either of these cases, the code can be fixed by adding a
3132       nested-name-specifier to explicitly name the template. The first
3133       can be worked around with -fno-access-control; the second is only
3134       rejected with -pedantic.
3135     * A new standard mangling for SIMD vector types has been added, to
3136       avoid name clashes on systems with vectors of varying length. By
3137       default the compiler still uses the old mangling, but emits aliases
3138       with the new mangling on targets that support strong aliases. Users
3139       can switch over entirely to the new mangling with -fabi-version=4
3140       or -fabi-version=0. -Wabi will now warn about code that uses the
3141       old mangling.
3142     * The command-line option -ftemplate-depth-N is now written as
3143       -ftemplate-depth=N and the old form is deprecated.
3144     * Conversions between NULL and non-pointer types are now warned by
3145       default. The new option -Wno-conversion-null disables these
3146       warnings. Previously these warnings were only available when using
3147       -Wconversion explicitly.
3148
3149    Runtime Library (libstdc++)
3150
3151     * [15]Improved experimental support for the upcoming ISO C++
3152       standard, C++0x, including:
3153          + Support for <future>, <functional>, and <random>.
3154          + Existing facilities now exploit explicit operators and the
3155            newly implemented core C++0x features.
3156     * An experimental [16]profile mode has been added. This is an
3157       implementation of many C++ standard library constructs with an
3158       additional analysis layer that gives performance improvement advice
3159       based on recognition of suboptimal usage patterns. For example,
3160#include <vector>
3161int main()
3162{
3163  std::vector<int> v;
3164  for (int k = 0; k < 1024; ++k)
3165    v.insert(v.begin(), k);
3166}
3167
3168       When instrumented via the profile mode, can return suggestions
3169       about the initial size and choice of the container used as follows:
3170vector-to-list: improvement = 5: call stack = 0x804842c ...
3171    : advice = change std::vector to std::list
3172vector-size: improvement = 3: call stack = 0x804842c ...
3173    : advice = change initial container size from 0 to 1024
3174
3175       These constructs can be substituted for the normal libstdc++
3176       constructs on a piecemeal basis, or all existing components can be
3177       transformed via the -D_GLIBCXX_PROFILE macro.
3178     * [17]Support for decimal floating-point arithmetic (aka ISO C++ TR
3179       24733) has been added. This support is in header file
3180       <decimal/decimal>, uses namespace std::decimal, and includes
3181       classes decimal32, decimal64, and decimal128.
3182     * Sources have been audited for application of function attributes
3183       nothrow, const, pure, and noreturn.
3184     * Python pretty-printers have been added for many standard library
3185       components that simplify the internal representation and present a
3186       more intuitive view of components when used with
3187       appropriately-advanced versions of GDB. For more information,
3188       please consult the more [18]detailed description.
3189     * The default behavior for comparing typeinfo names has changed, so
3190       in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero.
3191     * The new -static-libstdc++ option directs g++ to link the C++
3192       library statically, even if the default would normally be to link
3193       it dynamically.
3194
3195  Fortran
3196
3197     * The COMMON default padding has been changed - instead of adding the
3198       padding before a variable it is now added afterwards, which
3199       increases the compatibility with other vendors and helps to obtain
3200       the correct output in some cases. Cf. also the -falign-commons
3201       option ([19]added in 4.4).
3202     * The -finit-real= option now also supports the value snan for
3203       signalling not-a-number; to be effective, one additionally needs to
3204       enable trapping (e.g. via -ffpe-trap=). Note: Compile-time
3205       optimizations can turn a signalling NaN into a quiet one.
3206     * The new option -fcheck= has been added with the options bounds,
3207       array-temps, do, pointer, and recursive. The bounds and array-temps
3208       options are equivalent to -fbounds-check and
3209       -fcheck-array-temporaries. The do option checks for invalid
3210       modification of loop iteration variables, and the recursive option
3211       tests for recursive calls to subroutines/functions which are not
3212       marked as recursive. With pointer pointer association checks in
3213       calls are performed; however, neither undefined pointers nor
3214       pointers in expressions are handled. Using -fcheck=all enables all
3215       these run-time checks.
3216     * The run-time checking -fcheck=bounds now warns about invalid string
3217       lengths of character dummy arguments. Additionally, more
3218       compile-time checks have been added.
3219     * The new option [20]-fno-protect-parens has been added; if set, the
3220       compiler may reorder REAL and COMPLEX expressions without regard to
3221       parentheses.
3222     * GNU Fortran no longer links against libgfortranbegin. As before,
3223       MAIN__ (assembler symbol name) is the actual Fortran main program,
3224       which is invoked by the main function. However, main is now
3225       generated and put in the same object file as MAIN__. For the time
3226       being, libgfortranbegin still exists for backward compatibility.
3227       For details see the new [21]Mixed-Language Programming chapter in
3228       the manual.
3229     * The I/O library was restructured for performance and cleaner code.
3230     * Array assignments and WHERE are now run in parallel when OpenMP's
3231       WORKSHARE is used.
3232     * The experimental option -fwhole-file was added. The option allows
3233       whole-file checking of procedure arguments and allows for better
3234       optimizations. It can also be used with -fwhole-program, which is
3235       now also supported in gfortran.
3236     * More Fortran 2003 and Fortran 2008 mathematical functions can now
3237       be used as initialization expressions.
3238     * Some extended attributes such as STDCALL are now supported via the
3239       [22]GCC$ compiler directive.
3240     * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN
3241       intrinsic behaves now as if zero were always positive.
3242     * For legacy compatibiliy: On Cygwin and MinGW, the special files
3243       CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now
3244       supported.
3245     * Fortran 2003 support has been extended:
3246          + Procedure-pointer function results and procedure-pointer
3247            components (including PASS),
3248          + allocatable scalars (experimental),
3249          + DEFERRED type-bound procedures,
3250          + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements
3251            have been implemented.
3252          + The ALLOCATE statement supports type-specs and the SOURCE=
3253            argument.
3254          + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC
3255            type-bound procedure (i.e. as type-bound operators).
3256          + Rounding (ROUND=, RZ, ...) for output is now supported.
3257          + The INT_FAST{8,16,32,64,128}_T kind type parameters of the
3258            intrinsic module ISO_C_BINDING are now supported, except for
3259            the targets listed above as ones where GCC does not have
3260            <stdint.h> type information.
3261          + Extensible derived types with type-bound procedure or
3262            procedure pointer with PASS attribute now have to use CLASS in
3263            line with the Fortran 2003 standard; the workaround to use
3264            TYPE is no longer supported.
3265          + [23]Experimental, incomplete support for polymorphism,
3266            including CLASS, SELECT TYPE and dynamic dispatch of
3267            type-bound procedure calls. Some features do not work yet such
3268            as unlimited polymorphism (CLASS(*)).
3269     * Fortran 2008 support has been extended:
3270          + The OPEN statement now supports the NEWUNIT= option, which
3271            returns a unique file unit, thus preventing inadvertent use of
3272            the same unit in different parts of the program.
3273          + Support for unlimited format items has been added.
3274          + The INT{8,16,32} and REAL{32,64,128} kind type parameters of
3275            the intrinsic module ISO_FORTRAN_ENV are now supported.
3276          + Using complex arguments with TAN, SINH, COSH, TANH, ASIN,
3277            ACOS, and ATAN is now possible; the functions ASINH, ACOSH,
3278            and ATANH have been added (for real and complex arguments) and
3279            ATAN(Y,X) is now an alias for ATAN2(Y,X).
3280          + The BLOCK construct has been implemented.
3281
3282New Targets and Target Specific Improvements
3283
3284  AIX
3285
3286     * Full cross-toolchain support now available with GNU Binutils
3287
3288  ARM
3289
3290     * GCC now supports the Cortex-M0 and Cortex-A5 processors.
3291     * GCC now supports the ARM v7E-M architecture.
3292     * GCC now supports VFPv4-based FPUs and FPUs with
3293       single-precision-only VFP.
3294     * GCC has many improvements to optimization for other ARM processors,
3295       including scheduling support for the integer pipeline on Cortex-A9.
3296     * GCC now supports the IEEE 754-2008 half-precision floating-point
3297       type, and a variant ARM-specific half-precision type. This type is
3298       specified using __fp16, with the layout determined by
3299       -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and
3300       VFPv4 half-precision instructions will be used.
3301     * GCC now supports the variant of AAPCS that uses VFP registers for
3302       parameter passing and return values.
3303
3304  AVR
3305
3306     * The -mno-tablejump option has been removed because it has the same
3307       effect as the -fno-jump-tables option.
3308     * Added support for these new AVR devices:
3309          + ATmega8U2
3310          + ATmega16U2
3311          + ATmega32U2
3312
3313  IA-32/x86-64
3314
3315     * GCC now will set the default for -march= based on the configure
3316       target.
3317     * GCC now supports handling floating-point excess precision arising
3318       from use of the x87 floating-point unit in a way that conforms to
3319       ISO C99. This is enabled with -fexcess-precision=standard and with
3320       standards conformance options such as -std=c99, and may be disabled
3321       using -fexcess-precision=fast.
3322     * Support for the Intel Atom processor is now available through the
3323       -march=atom and -mtune=atom options.
3324     * A new -mcrc32 option is now available to enable crc32 intrinsics.
3325     * A new -mmovbe option is now available to enable GCC to use the
3326       movbe instruction to implement __builtin_bswap32 and
3327       __builtin_bswap64.
3328     * SSE math now can be enabled by default at configure time with the
3329       new --with-fpmath=sse option.
3330     * There is a new intrinsic header file, <x86intrin.h>. It should be
3331       included before using any IA-32/x86-64 intrinsics.
3332     * Support for the XOP, FMA4, and LWP instruction sets for the AMD
3333       Orochi processors are now available with the -mxop, -mfma4, and
3334       -mlwp options.
3335     * The -mabm option enables GCC to use the popcnt and lzcnt
3336       instructions on AMD processors.
3337     * The -mpopcnt option enables GCC to use the popcnt instructions on
3338       both AMD and Intel processors.
3339
3340  M68K/ColdFire
3341
3342     * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x
3343       and 5441x devices.
3344     * GCC now supports thread-local storage (TLS) on M68K and ColdFire
3345       processors.
3346
3347  MeP
3348
3349   Support has been added for the Toshiba Media embedded Processor (MeP,
3350   or mep-elf) embedded target.
3351
3352  MIPS
3353
3354     * GCC now supports MIPS 1004K processors.
3355     * GCC can now be configured with options --with-arch-32,
3356       --with-arch-64, --with-tune-32 and --with-tune-64 to control the
3357       default optimization separately for 32-bit and 64-bit modes.
3358     * MIPS targets now support an alternative _mcount interface, in which
3359       register $12 points to the function's save slot for register $31.
3360       This interface is selected by the -mcount-ra-address option; see
3361       the documentation for more details.
3362     * GNU/Linux targets can now generate read-only .eh_frame sections.
3363       This optimization requires GNU binutils 2.20 or above, and is only
3364       available if GCC is configured with a suitable version of binutils.
3365     * GNU/Linux targets can now attach special relocations to indirect
3366       calls, so that the linker can turn them into direct jumps or
3367       branches. This optimization requires GNU binutils 2.20 or later,
3368       and is automatically selected if GCC is configured with an
3369       appropriate version of binutils. It can be explicitly enabled or
3370       disabled using the -mrelax-pic-calls command-line option.
3371     * GCC now generates more heavily-optimized atomic operations on
3372       Octeon processors.
3373     * MIPS targets now support the -fstack-protector option.
3374     * GCC now supports an -msynci option, which specifies that synci is
3375       enough to flush the instruction cache, without help from the
3376       operating system. GCC uses this information to optimize
3377       automatically-generated cache flush operations, such as those used
3378       for nested functions in C. There is also a --with-synci
3379       configure-time option, which makes -msynci the default.
3380     * GCC supports four new function attributes for interrupt handlers:
3381       interrupt, use_shadow_register_set, keep_interrupts_masked and
3382       use_debug_exception_return. See the documentation for more details
3383       about these attributes.
3384
3385  RS/6000 (POWER/PowerPC)
3386
3387     * GCC now supports the Power ISA 2.06, which includes the VSX
3388       instructions that add vector 64-bit floating point support, new
3389       population count instructions, and conversions between floating
3390       point and unsigned types.
3391     * Support for the power7 processor is now available through the
3392       -mcpu=power7 and -mtune=power7.
3393     * GCC will now vectorize loops that contain simple math functions
3394       like copysign when generating code for altivec or VSX targets.
3395     * Support for the A2 processor is now available through the -mcpu=a2
3396       and -mtune=a2 options.
3397     * Support for the 476 processor is now available through the
3398       -mcpu={476,476fp} and -mtune={476,476fp} options.
3399     * Support for the e500mc64 processor is now available through the
3400       -mcpu=e500mc64 and -mtune=e500mc64 options.
3401     * GCC can now be configured with options --with-cpu-32,
3402       --with-cpu-64, --with-tune-32 and --with-tune-64 to control the
3403       default optimization separately for 32-bit and 64-bit modes.
3404     * Starting with GCC 4.5.4, vectors of type vector long long or vector
3405       long are passed and returned in the same method as other vectors
3406       with the VSX instruction set. Previously the GCC compiler did not
3407       adhere to the ABI for 128-bit vectors with 64-bit integer base
3408       types (PR 48857). This is also fixed in the GCC 4.6.1 release.
3409
3410  RX
3411
3412   Support has been added for the Renesas RX Processor (rx-elf) target.
3413
3414Operating Systems
3415
3416  Windows (Cygwin and MinGW)
3417
3418     * GCC now installs all the major language runtime libraries as DLLs
3419       when configured with the --enable-shared option.
3420     * GCC now makes use of the new support for aligned common variables
3421       in versions of binutils >= 2.20 to fix bugs in the support for SSE
3422       data types.
3423     * Improvements to the libffi support library increase the reliability
3424       of code generated by GCJ on all Windows platforms. Libgcj is
3425       enabled by default for the first time.
3426     * Libtool improvements simplify installation by placing the generated
3427       DLLs in the correct binaries directory.
3428     * Numerous other minor bugfixes and improvements, and substantial
3429       enhancements to the Fortran language support library.
3430
3431   >
3432
3433Other significant improvements
3434
3435  Plugins
3436
3437     * It is now possible to extend the compiler without having to modify
3438       its source code. A new option -fplugin=file.so tells GCC to load
3439       the shared object file.so and execute it as part of the compiler.
3440       The internal documentation describes the details on how plugins can
3441       interact with the compiler.
3442
3443  Installation changes
3444
3445     * The move to newer autotools changed default installation
3446       directories and switches to control them: The --with-datarootdir,
3447       --with-docdir, --with-pdfdir, and --with-htmldir switches are not
3448       used any more. Instead, you can now use --datarootdir, --docdir,
3449       --htmldir, and --pdfdir. The default installation directories have
3450       changed as follows according to the GNU Coding Standards:
3451
3452       datarootdir read-only architecture-independent data root [PREFIX/share]
3453       localedir   locale-specific message catalogs [DATAROOTDIR/locale]
3454       docdir      documentation root [DATAROOTDIR/doc/PACKAGE]
3455       htmldir     html documentation [DOCDIR]
3456       dvidir      dvi documentation [DOCDIR]
3457       pdfdir      pdf documentation [DOCDIR]
3458       psdir       ps documentation [DOCDIR]
3459       The following variables have new default values:
3460
3461       datadir read-only architecture-independent data [DATAROOTDIR]
3462       infodir info documentation [DATAROOTDIR/info]
3463       mandir  man documentation [DATAROOTDIR/man]
3464
3465GCC 4.5.1
3466
3467   This is the [24]list of problem reports (PRs) from GCC's bug tracking
3468   system that are known to be fixed in the 4.5.1 release. This list might
3469   not be complete (that is, it is possible that some PRs that have been
3470   fixed are not listed here).
3471
3472  All languages
3473
3474     * GCC's new link-time optimizer ([25]-flto) now also works on a few
3475       non-ELF targets:
3476          + Cygwin (*-cygwin*)
3477          + MinGW (*-mingw*)
3478          + Darwin on x86-64 (x86_64-apple-darwin*)
3479       LTO is not enabled by default for these targets. To enable LTO, you
3480       should configure with the --enable-lto option.
3481
3482GCC 4.5.2
3483
3484   This is the [26]list of problem reports (PRs) from GCC's bug tracking
3485   system that are known to be fixed in the 4.5.2 release. This list might
3486   not be complete (that is, it is possible that some PRs that have been
3487   fixed are not listed here).
3488
3489GCC 4.5.3
3490
3491   This is the [27]list of problem reports (PRs) from GCC's bug tracking
3492   system that are known to be fixed in the 4.5.3 release. This list might
3493   not be complete (that is, it is possible that some PRs that have been
3494   fixed are not listed here).
3495
3496   On the PowerPC compiler, the Altivec builtin functions vec_ld and
3497   vec_st have been modified to generate the Altivec memory instructions
3498   LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5
3499   release, these builtin functions were changed to generate VSX memory
3500   reference instructions instead of Altivec memory instructions, but
3501   there are differences between the two instructions. If the VSX
3502   instruction set is available, you can now use the new builtin functions
3503   vec_vsx_ld and vec_vsx_st which always generates the VSX memory
3504   instructions.
3505
3506GCC 4.5.4
3507
3508   This is the [28]list of problem reports (PRs) from GCC's bug tracking
3509   system that are known to be fixed in the 4.5.4 release. This list might
3510   not be complete (that is, it is possible that some PRs that have been
3511   fixed are not listed here).
3512
3513
3514    For questions related to the use of GCC, please consult these web
3515    pages and the [29]GCC manuals. If that fails, the
3516    [30]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3517    web pages and the development of GCC are welcome on our developer
3518    list at [31]gcc@gcc.gnu.org. All of [32]our lists have public
3519    archives.
3520
3521   Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and
3522   distribution of this entire article is permitted in any medium,
3523   provided this notice is preserved.
3524
3525   These pages are [34]maintained by the GCC team. Last modified
3526   2012-11-02[35].
3527
3528References
3529
3530   1. http://www.multiprecision.org/
3531   2. http://gcc.gnu.org/install/prerequisites.html
3532   3. http://gcc.gnu.org/ml/gcc/2010-01/msg00510.html
3533   4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted
3534   5. http://gcc.gnu.org/gcc-4.5/changes.html#x86
3535   6. http://www.multiprecision.org/
3536   7. http://gcc.gnu.org/PR30789
3537   8. http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
3538   9. http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802
3539  10. http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800
3540  11. http://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html
3541  12. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html
3542  13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757
3543  14. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176
3544  15. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.200x
3545  16. http://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html
3546  17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733
3547  18. http://sourceware.org/gdb/wiki/STLSupport
3548  19. http://gcc.gnu.org/gcc-4.4/changes.html
3549  20. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
3550  21. http://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html
3551  22. http://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html
3552  23. http://gcc.gnu.org/wiki/OOP
3553  24. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1
3554  25. http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
3555  26. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2
3556  27. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3
3557  28. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4
3558  29. http://gcc.gnu.org/onlinedocs/
3559  30. mailto:gcc-help@gcc.gnu.org
3560  31. mailto:gcc@gcc.gnu.org
3561  32. http://gcc.gnu.org/lists.html
3562  33. http://www.fsf.org/
3563  34. http://gcc.gnu.org/about.html
3564  35. http://validator.w3.org/check/referer
3565======================================================================
3566http://gcc.gnu.org/gcc-4.4/index.html
3567
3568                             GCC 4.4 Release Series
3569
3570   March 13, 2012
3571
3572   The [1]GNU project and the GCC developers are pleased to announce the
3573   release of GCC 4.4.7.
3574
3575   This release is a bug-fix release, containing fixes for regressions in
3576   GCC 4.4.6 relative to previous releases of GCC.
3577
3578Release History
3579
3580   GCC 4.4.7
3581          March 13, 2012 ([2]changes)
3582
3583   GCC 4.4.6
3584          April 16, 2011 ([3]changes)
3585
3586   GCC 4.4.5
3587          October 1, 2010 ([4]changes)
3588
3589   GCC 4.4.4
3590          April 29, 2010 ([5]changes)
3591
3592   GCC 4.4.3
3593          January 21, 2010 ([6]changes)
3594
3595   GCC 4.4.2
3596          October 15, 2009 ([7]changes)
3597
3598   GCC 4.4.1
3599          July 22, 2009 ([8]changes)
3600
3601   GCC 4.4.0
3602          April 21, 2009 ([9]changes)
3603
3604References and Acknowledgements
3605
3606   GCC used to stand for the GNU C Compiler, but since the compiler
3607   supports several other languages aside from C, it now stands for the
3608   GNU Compiler Collection.
3609
3610   A list of [10]successful builds is updated as new information becomes
3611   available.
3612
3613   The GCC developers would like to thank the numerous people that have
3614   contributed new features, improvements, bug fixes, and other changes as
3615   well as test results to GCC. This [11]amazing group of volunteers is
3616   what makes GCC successful.
3617
3618   For additional information about GCC please refer to the [12]GCC
3619   project web site or contact the [13]GCC development mailing list.
3620
3621   To obtain GCC please use [14]our mirror sites or [15]our SVN server.
3622
3623
3624    For questions related to the use of GCC, please consult these web
3625    pages and the [16]GCC manuals. If that fails, the
3626    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3627    web pages and the development of GCC are welcome on our developer
3628    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
3629    archives.
3630
3631   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
3632   distribution of this entire article is permitted in any medium,
3633   provided this notice is preserved.
3634
3635   These pages are [21]maintained by the GCC team. Last modified
3636   2012-11-02[22].
3637
3638References
3639
3640   1. http://www.gnu.org/
3641   2. http://gcc.gnu.org/gcc-4.4/changes.html
3642   3. http://gcc.gnu.org/gcc-4.4/changes.html
3643   4. http://gcc.gnu.org/gcc-4.4/changes.html
3644   5. http://gcc.gnu.org/gcc-4.4/changes.html
3645   6. http://gcc.gnu.org/gcc-4.4/changes.html
3646   7. http://gcc.gnu.org/gcc-4.4/changes.html
3647   8. http://gcc.gnu.org/gcc-4.4/changes.html
3648   9. http://gcc.gnu.org/gcc-4.4/changes.html
3649  10. http://gcc.gnu.org/gcc-4.4/buildstat.html
3650  11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
3651  12. http://gcc.gnu.org/index.html
3652  13. mailto:gcc@gcc.gnu.org
3653  14. http://gcc.gnu.org/mirrors.html
3654  15. http://gcc.gnu.org/svn.html
3655  16. http://gcc.gnu.org/onlinedocs/
3656  17. mailto:gcc-help@gcc.gnu.org
3657  18. mailto:gcc@gcc.gnu.org
3658  19. http://gcc.gnu.org/lists.html
3659  20. http://www.fsf.org/
3660  21. http://gcc.gnu.org/about.html
3661  22. http://validator.w3.org/check/referer
3662======================================================================
3663http://gcc.gnu.org/gcc-4.4/changes.html
3664
3665                             GCC 4.4 Release Series
3666                        Changes, New Features, and Fixes
3667
3668   The latest release in the 4.4 release series is [1]GCC 4.4.7.
3669
3670Caveats
3671
3672     * __builtin_stdarg_start has been completely removed from GCC.
3673       Support for <varargs.h> had been deprecated since GCC 4.0. Use
3674       __builtin_va_start as a replacement.
3675     * Some of the errors issued by the C++ front end that could be
3676       downgraded to warnings in previous releases by using -fpermissive
3677       are now warnings by default. They can be converted into errors by
3678       using -pedantic-errors.
3679     * Use of the cpp assertion extension will now emit a warning when
3680       -Wdeprecated or -pedantic is used. This extension has been
3681       deprecated for many years, but never warned about.
3682     * Packed bit-fields of type char were not properly bit-packed on many
3683       targets prior to GCC 4.4. On these targets, the fix in GCC 4.4
3684       causes an ABI change. For example there is no longer a 4-bit
3685       padding between field a and b in this structure:
3686    struct foo
3687    {
3688      char a:4;
3689      char b:8;
3690    } __attribute__ ((packed));
3691       There is a new warning to help identify fields that are affected:
3692    foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4
3693       The warning can be disabled with -Wno-packed-bitfield-compat.
3694     * On ARM EABI targets, the C++ mangling of the va_list type has been
3695       changed to conform to the current revision of the EABI. This does
3696       not affect the libstdc++ library included with GCC.
3697     * The SCOUNT and POS bits of the MIPS DSP control register are now
3698       treated as global. Previous versions of GCC treated these fields as
3699       call-clobbered instead.
3700     * The MIPS port no longer recognizes the h asm constraint. It was
3701       necessary to remove this constraint in order to avoid generating
3702       unpredictable code sequences.
3703       One of the main uses of the h constraint was to extract the high
3704       part of a multiplication on 64-bit targets. For example:
3705    asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y));
3706       You can now achieve the same effect using 128-bit types:
3707    typedef unsigned int uint128_t __attribute__((mode(TI)));
3708    result = ((uint128_t) x * y) >> 64;
3709       The second sequence is better in many ways. For example, if x and y
3710       are constants, the compiler can perform the multiplication at
3711       compile time. If x and y are not constants, the compiler can
3712       schedule the runtime multiplication better than it can schedule an
3713       asm statement.
3714     * Support for a number of older systems and recently unmaintained or
3715       untested target ports of GCC has been declared obsolete in GCC 4.4.
3716       Unless there is activity to revive them, the next release of GCC
3717       will have their sources permanently removed.
3718       The following ports for individual systems on particular
3719       architectures have been obsoleted:
3720          + Generic a.out on IA32 and m68k (i[34567]86-*-aout*,
3721            m68k-*-aout*)
3722          + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*,
3723            armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*,
3724            sh-*-*). This does not affect other more specific targets
3725            using the COFF object format on those architectures, or the
3726            more specific H8300 and SH targets (h8300-*-rtems*,
3727            h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*,
3728            sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks).
3729          + 2BSD on PDP-11 (pdp11-*-bsd)
3730          + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*,
3731            powerpc-ibm-aix4.[12]*)
3732          + Tuning support for Itanium1 (Merced) variants. Note that code
3733            tuned for Itanium2 should also run correctly on Itanium1.
3734     * The protoize and unprotoize utilities have been obsoleted and will
3735       be removed in GCC 4.5. These utilities have not been installed by
3736       default since GCC 3.0.
3737     * Support has been removed for all the [2]configurations obsoleted in
3738       GCC 4.3.
3739     * Unknown -Wno-* options are now silently ignored by GCC if no other
3740       diagnostics are issued. If other diagnostics are issued, then GCC
3741       warns about the unknown options.
3742     * More information on porting to GCC 4.4 from previous versions of
3743       GCC can be found in the [3]porting guide for this release.
3744
3745General Optimizer Improvements
3746
3747     * A new command-line switch -findirect-inlining has been added. When
3748       turned on it allows the inliner to also inline indirect calls that
3749       are discovered to have known targets at compile time thanks to
3750       previous inlining.
3751     * A new command-line switch -ftree-switch-conversion has been added.
3752       This new pass turns simple initializations of scalar variables in
3753       switch statements into initializations from a static array, given
3754       that all the values are known at compile time and the ratio between
3755       the new array size and the original switch branches does not exceed
3756       the parameter --param switch-conversion-max-branch-ratio (default
3757       is eight).
3758     * A new command-line switch -ftree-builtin-call-dce has been added.
3759       This optimization eliminates unnecessary calls to certain builtin
3760       functions when the return value is not used, in cases where the
3761       calls can not be eliminated entirely because the function may set
3762       errno. This optimization is on by default at -O2 and above.
3763     * A new command-line switch -fconserve-stack directs the compiler to
3764       minimize stack usage even if it makes the generated code slower.
3765       This affects inlining decisions.
3766     * When the assembler supports it, the compiler will now emit unwind
3767       information using assembler .cfi directives. This makes it possible
3768       to use such directives in inline assembler code. The new option
3769       -fno-dwarf2-cfi-asm directs the compiler to not use .cfi
3770       directives.
3771     * The [4]Graphite branch has been merged. This merge has brought in a
3772       new framework for loop optimizations based on a polyhedral
3773       intermediate representation. These optimizations apply to all the
3774       languages supported by GCC. The following new code transformations
3775       are available in GCC 4.4:
3776          + -floop-interchange performs loop interchange transformations
3777            on loops. Interchanging two nested loops switches the inner
3778            and outer loops. For example, given a loop like:
3779          DO J = 1, M
3780            DO I = 1, N
3781              A(J, I) = A(J, I) * C
3782            ENDDO
3783          ENDDO
3784
3785            loop interchange will transform the loop as if the user had
3786            written:
3787          DO I = 1, N
3788            DO J = 1, M
3789              A(J, I) = A(J, I) * C
3790            ENDDO
3791          ENDDO
3792
3793            which can be beneficial when N is larger than the caches,
3794            because in Fortran, the elements of an array are stored in
3795            memory contiguously by column, and the original loop iterates
3796            over rows, potentially creating at each access a cache miss.
3797          + -floop-strip-mine performs loop strip mining transformations
3798            on loops. Strip mining splits a loop into two nested loops.
3799            The outer loop has strides equal to the strip size and the
3800            inner loop has strides of the original loop within a strip.
3801            For example, given a loop like:
3802          DO I = 1, N
3803            A(I) = A(I) + C
3804          ENDDO
3805
3806            loop strip mining will transform the loop as if the user had
3807            written:
3808          DO II = 1, N, 4
3809            DO I = II, min (II + 3, N)
3810              A(I) = A(I) + C
3811            ENDDO
3812          ENDDO
3813
3814          + -floop-block performs loop blocking transformations on loops.
3815            Blocking strip mines each loop in the loop nest such that the
3816            memory accesses of the element loops fit inside caches. For
3817            example, given a loop like:
3818          DO I = 1, N
3819            DO J = 1, M
3820              A(J, I) = B(I) + C(J)
3821            ENDDO
3822          ENDDO
3823
3824            loop blocking will transform the loop as if the user had
3825            written:
3826          DO II = 1, N, 64
3827            DO JJ = 1, M, 64
3828              DO I = II, min (II + 63, N)
3829                DO J = JJ, min (JJ + 63, M)
3830                  A(J, I) = B(I) + C(J)
3831                ENDDO
3832              ENDDO
3833            ENDDO
3834          ENDDO
3835
3836            which can be beneficial when M is larger than the caches,
3837            because the innermost loop will iterate over a smaller amount
3838            of data that can be kept in the caches.
3839     * A new register allocator has replaced the old one. It is called
3840       integrated register allocator (IRA) because coalescing, register
3841       live range splitting, and hard register preferencing are done
3842       on-the-fly during coloring. It also has better integration with the
3843       reload pass. IRA is a regional register allocator which uses modern
3844       Chaitin-Briggs coloring instead of Chow's priority coloring used in
3845       the old register allocator. More info about IRA internals and
3846       options can be found in the GCC manuals.
3847     * A new instruction scheduler and software pipeliner, based on the
3848       selective scheduling approach, has been added. The new pass
3849       performs instruction unification, register renaming, substitution
3850       through register copies, and speculation during scheduling. The
3851       software pipeliner is able to pipeline non-countable loops. The new
3852       pass is targeted at scheduling-eager in-order platforms. In GCC 4.4
3853       it is available for the Intel Itanium platform working by default
3854       as the second scheduling pass (after register allocation) at the
3855       -O3 optimization level.
3856     * When using -fprofile-generate with a multi-threaded program, the
3857       profile counts may be slightly wrong due to race conditions. The
3858       new -fprofile-correction option directs the compiler to apply
3859       heuristics to smooth out the inconsistencies. By default the
3860       compiler will give an error message when it finds an inconsistent
3861       profile.
3862     * The new -fprofile-dir=PATH option permits setting the directory
3863       where profile data files are stored when using -fprofile-generate
3864       and friends, and the directory used when reading profile data files
3865       using -fprofile-use and friends.
3866
3867New warning options
3868
3869     * The new -Wframe-larger-than=NUMBER option directs GCC to emit a
3870       warning if any stack frame is larger than NUMBER bytes. This may be
3871       used to help ensure that code fits within a limited amount of stack
3872       space.
3873     * The command-line option -Wlarger-than-N is now written as
3874       -Wlarger-than=N and the old form is deprecated.
3875     * The new -Wno-mudflap option disables warnings about constructs
3876       which can not be instrumented when using -fmudflap.
3877
3878New Languages and Language specific improvements
3879
3880     * Version 3.0 of the [5]OpenMP specification is now supported for the
3881       C, C++, and Fortran compilers.
3882     * New character data types, per [6]TR 19769: New character types in
3883       C, are now supported for the C compiler in -std=gnu99 mode, as
3884       __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in
3885       -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too.
3886
3887  C family
3888
3889     * A new optimize attribute was added to allow programmers to change
3890       the optimization level and particular optimization options for an
3891       individual function. You can also change the optimization options
3892       via the GCC optimize pragma for functions defined after the pragma.
3893       The GCC push_options pragma and the GCC pop_options pragma allow
3894       you temporarily save and restore the options used. The GCC
3895       reset_options pragma restores the options to what was specified on
3896       the command line.
3897     * Uninitialized warnings do not require enabling optimization
3898       anymore, that is, -Wuninitialized can be used together with -O0.
3899       Nonetheless, the warnings given by -Wuninitialized will probably be
3900       more accurate if optimization is enabled.
3901     * -Wparentheses now warns about expressions such as (!x | y) and (!x
3902       & y). Using explicit parentheses, such as in ((!x) | y), silences
3903       this warning.
3904     * -Wsequence-point now warns within if, while,do while and for
3905       conditions, and within for begin/end expressions.
3906     * A new option -dU is available to dump definitions of preprocessor
3907       macros that are tested or expanded.
3908
3909  C++
3910
3911     * [7]Improved experimental support for the upcoming ISO C++ standard,
3912       C++0x. Including support for auto, inline namespaces, generalized
3913       initializer lists, defaulted and deleted functions, new character
3914       types, and scoped enums.
3915     * Those errors that may be downgraded to warnings to build legacy
3916       code now mention -fpermissive when -fdiagnostics-show-option is
3917       enabled.
3918     * -Wconversion now warns if the result of a static_cast to enumeral
3919       type is unspecified because the value is outside the range of the
3920       enumeral type.
3921     * -Wuninitialized now warns if a non-static reference or non-static
3922       const member appears in a class without constructors.
3923     * G++ now properly implements value-initialization, so objects with
3924       an initializer of () and an implicitly defined default constructor
3925       will be zero-initialized before the default constructor is called.
3926
3927    Runtime Library (libstdc++)
3928
3929     * [8]Improved experimental support for the upcoming ISO C++ standard,
3930       C++0x, including:
3931          + Support for <chrono>, <condition_variable>, <cstdatomic>,
3932            <forward_list>, <initializer_list>, <mutex>, <ratio>,
3933            <system_error>, and <thread>.
3934          + unique_ptr, <algorithm> additions, exception propagation, and
3935            support for the new character types in <string> and <limits>.
3936          + Existing facilities now exploit initializer lists, defaulted
3937            and deleted functions, and the newly implemented core C++0x
3938            features.
3939          + Some standard containers are more efficient together with
3940            stateful allocators, i.e., no allocator is constructed on the
3941            fly at element construction time.
3942     * Experimental support for non-standard pointer types in containers.
3943     * The long standing libstdc++/30928 has been fixed for targets
3944       running glibc 2.10 or later.
3945     * As usual, many small and larger bug fixes, in particular quite a
3946       few corner cases in <locale>.
3947
3948  Fortran
3949
3950     * GNU Fortran now employs libcpp directly instead of using cc1 as an
3951       external preprocessor. The [9]-cpp option was added to allow manual
3952       invocation of the preprocessor without relying on filename
3953       extensions.
3954     * The [10]-Warray-temporaries option warns about array temporaries
3955       generated by the compiler, as an aid to optimization.
3956     * The [11]-fcheck-array-temporaries option has been added, printing a
3957       notification at run time, when an array temporary had to be created
3958       for an function argument. Contrary to -Warray-temporaries the
3959       warning is only printed if the array is noncontiguous.
3960     * Improved generation of DWARF debugging symbols
3961     * If using an intrinsic not part of the selected standard (via -std=
3962       and -fall-intrinsics) gfortran will now treat it as if this
3963       procedure were declared EXTERNAL and try to link to a user-supplied
3964       procedure. -Wintrinsics-std will warn whenever this happens. The
3965       now-useless option -Wnonstd-intrinsic was removed.
3966     * The flag -falign-commons has been added to control the alignment of
3967       variables in COMMON blocks, which is enabled by default in line
3968       with previous GCC version. Using -fno-align-commons one can force
3969       commons to be contiguous in memory as required by the Fortran
3970       standard, however, this slows down the memory access. The option
3971       -Walign-commons, which is enabled by default, warns when padding
3972       bytes were added for alignment. The proper solution is to sort the
3973       common objects by decreasing storage size, which avoids the
3974       alignment problems.
3975     * Fortran 2003 support has been extended:
3976          + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is
3977            now supported (except internal reads from/writes to wide
3978            strings). [12]-fbackslash now supports also \unnnn and
3979            \Unnnnnnnn to enter Unicode characters.
3980          + Asynchronous I/O (implemented as synchronous I/O) and the
3981            decimal=, size=, sign=, pad=, blank=, and delim= specifiers
3982            are now supported in I/O statements.
3983          + Support for Fortran 2003 structure constructors and for array
3984            constructor with typespec has been added.
3985          + Procedure Pointers (but not yet as component in derived types
3986            and as function results) are now supported.
3987          + Abstract types, type extension, and type-bound procedures
3988            (both PROCEDURE and GENERIC but not as operators). Note: As
3989            CLASS/polymorphyic types are not implemented, type-bound
3990            procedures with PASS accept as non-standard extension TYPE
3991            arguments.
3992     * Fortran 2008 support has been added:
3993          + The -std=f2008 option and support for the file extensions
3994            .f2008 and .F2008 has been added.
3995          + The g0 format descriptor is now supported.
3996          + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH,
3997            ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED
3998            are now available (some of them existed as GNU extension
3999            before). Note: The hyperbolic functions are not yet supporting
4000            complex arguments and the three- argument version of BESSEL_*N
4001            is not available.
4002          + The bit intrinsics LEADZ and TRAILZ have been added.
4003
4004  Java (GCJ)
4005
4006  Ada
4007
4008     * The Ada runtime now supports multilibs on many platforms including
4009       x86_64, SPARC and PowerPC. Their build is enabled by default.
4010
4011New Targets and Target Specific Improvements
4012
4013  ARM
4014
4015     * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and
4016       Cortex-R4F processors and has many other improvements to
4017       optimization for ARM processors.
4018     * GCC now supports the VFPv3 variant with 16 double-precision
4019       registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been
4020       renamed to -mfpu=vfpv3.
4021     * GCC now supports the -mfix-cortex-m3-ldrd option to work around an
4022       erratum on Cortex-M3 processors.
4023     * GCC now supports the __sync_* atomic operations for ARM EABI
4024       GNU/Linux.
4025     * The section anchors optimization is now enabled by default when
4026       optimizing for ARM.
4027     * GCC now uses a new EABI-compatible profiling interface for EABI
4028       targets. This requires a function __gnu_mcount_nc, which is
4029       provided by GNU libc versions 2.8 and later.
4030
4031  AVR
4032
4033     * The -mno-tablejump option has been deprecated because it has the
4034       same effect as the -fno-jump-tables option.
4035     * Added support for these new AVR devices:
4036          + ATA6289
4037          + ATtiny13A
4038          + ATtiny87
4039          + ATtiny167
4040          + ATtiny327
4041          + ATmega8C1
4042          + ATmega16C1
4043          + ATmega32C1
4044          + ATmega8M1
4045          + ATmega16M1
4046          + ATmega32M1
4047          + ATmega32U4
4048          + ATmega16HVB
4049          + ATmega4HVD
4050          + ATmega8HVD
4051          + ATmega64C1
4052          + ATmega64M1
4053          + ATmega16U4
4054          + ATmega32U6
4055          + ATmega128RFA1
4056          + AT90PWM81
4057          + AT90SCR100
4058          + M3000F
4059          + M3000S
4060          + M3001B
4061
4062  IA-32/x86-64
4063
4064     * Support for Intel AES built-in functions and code generation is
4065       available via -maes.
4066     * Support for Intel PCLMUL built-in function and code generation is
4067       available via -mpclmul.
4068     * Support for Intel AVX built-in functions and code generation is
4069       available via -mavx.
4070     * Automatically align the stack for local variables with alignment
4071       requirement.
4072     * GCC can now utilize the SVML library for vectorizing calls to a set
4073       of C99 functions if -mveclibabi=svml is specified and you link to
4074       an SVML ABI compatible library.
4075     * On x86-64, the ABI has been changed in the following cases to
4076       conform to the x86-64 ABI:
4077          + Passing/returning structures with flexible array member:
4078  struct foo
4079    {
4080      int i;
4081      int flex[];
4082    };
4083          + Passing/returning structures with complex float member:
4084  struct foo
4085    {
4086      int i;
4087      __complex__ float f;
4088    };
4089          + Passing/returning unions with long double member:
4090  union foo
4091    {
4092      int x;
4093      long double ld;
4094    };
4095       Code built with previous versions of GCC that uses any of these is
4096       not compatible with code built with GCC 4.4.0 or later.
4097     * A new target attribute was added to allow programmers to change the
4098       target options like -msse2 or -march=k8 for an individual function.
4099       You can also change the target options via the GCC target pragma
4100       for functions defined after the pragma.
4101     * GCC can now be configured with options --with-arch-32,
4102       --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and
4103       --with-tune-64 to control the default optimization separately for
4104       32-bit and 64-bit modes.
4105
4106  IA-32/IA64
4107
4108     * Support for __float128 (TFmode) IEEE quad type and corresponding
4109       TCmode IEEE complex quad type is available via the soft-fp library
4110       on IA-32/IA64 targets. This includes basic arithmetic operations
4111       (addition, subtraction, negation, multiplication and division) on
4112       __float128 real and TCmode complex values, the full set of IEEE
4113       comparisons between __float128 values, conversions to and from
4114       float, double and long double floating point types, as well as
4115       conversions to and from signed or unsigned integer, signed or
4116       unsigned long integer and signed or unsigned quad (TImode, IA64
4117       only) integer types. Additionally, all operations generate the full
4118       set of IEEE exceptions and support the full set of IEEE rounding
4119       modes.
4120
4121  M68K/ColdFire
4122
4123     * GCC now supports instruction scheduling for ColdFire V1, V3 and V4
4124       processors. (Scheduling support for ColdFire V2 processors was
4125       added in GCC 4.3.)
4126     * GCC now supports the -mxgot option to support programs requiring
4127       many GOT entries on ColdFire.
4128     * The m68k-*-linux-gnu target now builds multilibs by default.
4129
4130  MIPS
4131
4132     * MIPS Technologies have extended the original MIPS SVR4 ABI to
4133       include support for procedure linkage tables (PLTs) and copy
4134       relocations. These extensions allow GNU/Linux executables to use a
4135       significantly more efficient code model than the one defined by the
4136       original ABI.
4137       GCC support for this code model is available via a new command-line
4138       option, -mplt. There is also a new configure-time option,
4139       --with-mips-plt, to make -mplt the default.
4140       The new code model requires support from the assembler, the linker,
4141       and the runtime C library. This support is available in binutils
4142       2.19 and GLIBC 2.9.
4143     * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables
4144       and 32-bit GNU/Linux shared libraries. This feature requires GNU
4145       binutils 2.19 or above.
4146     * Support for RMI's XLR processor is now available through the
4147       -march=xlr and -mtune=xlr options.
4148     * 64-bit targets can now perform 128-bit multiplications inline,
4149       instead of relying on a libgcc function.
4150     * Native GNU/Linux toolchains now support -march=native and
4151       -mtune=native, which select the host processor.
4152     * GCC now supports the R10K, R12K, R14K and R16K processors. The
4153       canonical -march= and -mtune= names for these processors are
4154       r10000, r12000, r14000 and r16000 respectively.
4155     * GCC can now work around the side effects of speculative execution
4156       on R10K processors. Please see the documentation of the
4157       -mr10k-cache-barrier option for details.
4158     * Support for the MIPS64 Release 2 instruction set has been added.
4159       The option -march=mips64r2 enables generation of these
4160       instructions.
4161     * GCC now supports Cavium Networks' Octeon processor. This support is
4162       available through the -march=octeon and -mtune=octeon options.
4163     * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The
4164       canonical -march= and -mtune= names for these processors are
4165       loongson2e and loongson2f.
4166
4167  picochip
4168
4169   Picochip is a 16-bit processor. A typical picoChip contains over 250
4170   small cores, each with small amounts of memory. There are three
4171   processor variants (STAN, MEM and CTRL) with different instruction sets
4172   and memory configurations and they can be chosen using the -mae option.
4173
4174   This port is intended to be a "C" only port.
4175
4176  Power Architecture and PowerPC
4177
4178     * GCC now supports the e300c2, e300c3 and e500mc processors.
4179     * GCC now supports Xilinx processors with a single-precision FPU.
4180     * Decimal floating point is now supported for e500 processors.
4181
4182  S/390, zSeries and System z9/z10
4183
4184     * Support for the IBM System z10 EC/BC processor has been added. When
4185       using the -march=z10 option, the compiler will generate code making
4186       use of instructions provided by the General-Instruction-Extension
4187       Facility and the Execute-Extension Facility.
4188
4189  VxWorks
4190
4191     * GCC now supports the thread-local storage mechanism used on
4192       VxWorks.
4193
4194  Xtensa
4195
4196     * GCC now supports thread-local storage (TLS) for Xtensa processor
4197       configurations that include the Thread Pointer option. TLS also
4198       requires support from the assembler and linker; this support is
4199       provided in the GNU binutils beginning with version 2.19.
4200
4201Documentation improvements
4202
4203Other significant improvements
4204
4205GCC 4.4.1
4206
4207   This is the [13]list of problem reports (PRs) from GCC's bug tracking
4208   system that are known to be fixed in the 4.4.1 release. This list might
4209   not be complete (that is, it is possible that some PRs that have been
4210   fixed are not listed here).
4211
4212GCC 4.4.2
4213
4214   This is the [14]list of problem reports (PRs) from GCC's bug tracking
4215   system that are known to be fixed in the 4.4.2 release. This list might
4216   not be complete (that is, it is possible that some PRs that have been
4217   fixed are not listed here).
4218
4219GCC 4.4.3
4220
4221   This is the [15]list of problem reports (PRs) from GCC's bug tracking
4222   system that are known to be fixed in the 4.4.3 release. This list might
4223   not be complete (that is, it is possible that some PRs that have been
4224   fixed are not listed here).
4225
4226GCC 4.4.4
4227
4228   This is the [16]list of problem reports (PRs) from GCC's bug tracking
4229   system that are known to be fixed in the 4.4.4 release. This list might
4230   not be complete (that is, it is possible that some PRs that have been
4231   fixed are not listed here).
4232
4233GCC 4.4.5
4234
4235   This is the [17]list of problem reports (PRs) from GCC's bug tracking
4236   system that are known to be fixed in the 4.4.5 release. This list might
4237   not be complete (that is, it is possible that some PRs that have been
4238   fixed are not listed here).
4239
4240GCC 4.4.6
4241
4242   This is the [18]list of problem reports (PRs) from GCC's bug tracking
4243   system that are known to be fixed in the 4.4.6 release. This list might
4244   not be complete (that is, it is possible that some PRs that have been
4245   fixed are not listed here).
4246
4247GCC 4.4.7
4248
4249   This is the [19]list of problem reports (PRs) from GCC's bug tracking
4250   system that are known to be fixed in the 4.4.7 release. This list might
4251   not be complete (that is, it is possible that some PRs that have been
4252   fixed are not listed here).
4253
4254
4255    For questions related to the use of GCC, please consult these web
4256    pages and the [20]GCC manuals. If that fails, the
4257    [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4258    web pages and the development of GCC are welcome on our developer
4259    list at [22]gcc@gcc.gnu.org. All of [23]our lists have public
4260    archives.
4261
4262   Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and
4263   distribution of this entire article is permitted in any medium,
4264   provided this notice is preserved.
4265
4266   These pages are [25]maintained by the GCC team. Last modified
4267   2012-11-02[26].
4268
4269References
4270
4271   1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7
4272   2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted
4273   3. http://gcc.gnu.org/gcc-4.4/porting_to.html
4274   4. http://gcc.gnu.org/wiki/Graphite
4275   5. http://openmp.org/wp/openmp-specifications/
4276   6. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf
4277   7. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html
4278   8. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#id476343
4279   9. http://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html
4280  10. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125
4281  11. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221
4282  12. http://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34
4283  13. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1
4284  14. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2
4285  15. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3
4286  16. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4
4287  17. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5
4288  18. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6
4289  19. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7
4290  20. http://gcc.gnu.org/onlinedocs/
4291  21. mailto:gcc-help@gcc.gnu.org
4292  22. mailto:gcc@gcc.gnu.org
4293  23. http://gcc.gnu.org/lists.html
4294  24. http://www.fsf.org/
4295  25. http://gcc.gnu.org/about.html
4296  26. http://validator.w3.org/check/referer
4297======================================================================
4298http://gcc.gnu.org/gcc-4.3/index.html
4299
4300                             GCC 4.3 Release Series
4301
4302   Jun 27, 2011
4303
4304   The [1]GNU project and the GCC developers are pleased to announce the
4305   release of GCC 4.3.6.
4306
4307   This release is a bug-fix release, containing fixes for regressions in
4308   GCC 4.3.5 relative to previous releases of GCC.
4309
4310Release History
4311
4312   GCC 4.3.6
4313          Jun 27, 2011 ([2]changes)
4314
4315   GCC 4.3.5
4316          May 22, 2010 ([3]changes)
4317
4318   GCC 4.3.4
4319          August 4, 2009 ([4]changes)
4320
4321   GCC 4.3.3
4322          January 24, 2009 ([5]changes)
4323
4324   GCC 4.3.2
4325          August 27, 2008 ([6]changes)
4326
4327   GCC 4.3.1
4328          June 6, 2008 ([7]changes)
4329
4330   GCC 4.3.0
4331          March 5, 2008 ([8]changes)
4332
4333References and Acknowledgements
4334
4335   GCC used to stand for the GNU C Compiler, but since the compiler
4336   supports several other languages aside from C, it now stands for the
4337   GNU Compiler Collection.
4338
4339   A list of [9]successful builds is updated as new information becomes
4340   available.
4341
4342   The GCC developers would like to thank the numerous people that have
4343   contributed new features, improvements, bug fixes, and other changes as
4344   well as test results to GCC. This [10]amazing group of volunteers is
4345   what makes GCC successful.
4346
4347   For additional information about GCC please refer to the [11]GCC
4348   project web site or contact the [12]GCC development mailing list.
4349
4350   To obtain GCC please use [13]our mirror sites or [14]our SVN server.
4351
4352
4353    For questions related to the use of GCC, please consult these web
4354    pages and the [15]GCC manuals. If that fails, the
4355    [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4356    web pages and the development of GCC are welcome on our developer
4357    list at [17]gcc@gcc.gnu.org. All of [18]our lists have public
4358    archives.
4359
4360   Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and
4361   distribution of this entire article is permitted in any medium,
4362   provided this notice is preserved.
4363
4364   These pages are [20]maintained by the GCC team. Last modified
4365   2012-11-02[21].
4366
4367References
4368
4369   1. http://www.gnu.org/
4370   2. http://gcc.gnu.org/gcc-4.3/changes.html
4371   3. http://gcc.gnu.org/gcc-4.3/changes.html
4372   4. http://gcc.gnu.org/gcc-4.3/changes.html
4373   5. http://gcc.gnu.org/gcc-4.3/changes.html
4374   6. http://gcc.gnu.org/gcc-4.3/changes.html
4375   7. http://gcc.gnu.org/gcc-4.3/changes.html
4376   8. http://gcc.gnu.org/gcc-4.3/changes.html
4377   9. http://gcc.gnu.org/gcc-4.3/buildstat.html
4378  10. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
4379  11. http://gcc.gnu.org/index.html
4380  12. mailto:gcc@gcc.gnu.org
4381  13. http://gcc.gnu.org/mirrors.html
4382  14. http://gcc.gnu.org/svn.html
4383  15. http://gcc.gnu.org/onlinedocs/
4384  16. mailto:gcc-help@gcc.gnu.org
4385  17. mailto:gcc@gcc.gnu.org
4386  18. http://gcc.gnu.org/lists.html
4387  19. http://www.fsf.org/
4388  20. http://gcc.gnu.org/about.html
4389  21. http://validator.w3.org/check/referer
4390======================================================================
4391http://gcc.gnu.org/gcc-4.3/changes.html
4392
4393                             GCC 4.3 Release Series
4394                        Changes, New Features, and Fixes
4395
4396   The latest release in the 4.3 release series is [1]GCC 4.3.5.
4397
4398Caveats
4399
4400     * GCC requires the [2]GMP and [3]MPFR libraries for building all the
4401       various front-end languages it supports. See the [4]prerequisites
4402       page for version requirements.
4403     * ColdFire targets now treat long double as having the same format as
4404       double. In earlier versions of GCC, they used the 68881 long double
4405       format instead.
4406     * The m68k-uclinux target now uses the same calling conventions as
4407       m68k-linux-gnu. You can select the original calling conventions by
4408       configuring for m68k-uclinuxoldabi instead. Note that
4409       m68k-uclinuxoldabi also retains the original 80-bit long double on
4410       ColdFire targets.
4411     * The -fforce-mem option has been removed because it has had no
4412       effect in the last few GCC releases.
4413     * The i386 -msvr3-shlib option has been removed since it is no longer
4414       used.
4415     * Fastcall for i386 has been changed not to pass aggregate arguments
4416       in registers, following Microsoft compilers.
4417     * Support for the AOF assembler has been removed from the ARM back
4418       end; this affects only the targets arm-semi-aof and armel-semi-aof,
4419       which are no longer recognized. We removed these targets without a
4420       deprecation period because we discovered that they have been
4421       unusable since GCC 4.0.0.
4422     * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*)
4423       has been removed. This support had been deprecated since GCC 4.0.0.
4424     * Support for a number of older systems and recently unmaintained or
4425       untested target ports of GCC has been declared obsolete in GCC 4.3.
4426       Unless there is activity to revive them, the next release of GCC
4427       will have their sources permanently removed.
4428       All GCC ports for the following processor architectures have been
4429       declared obsolete:
4430          + Morpho MT (mt-*)
4431       The following aliases for processor architectures have been
4432       declared obsolete. Users should use the indicated generic target
4433       names instead, with compile-time options such as -mcpu or
4434       configure-time options such as --with-cpu to control the
4435       configuration more precisely.
4436          + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-*
4437            instead).
4438          + parisc*-*-* (use hppa*-*-* instead).
4439          + m680[012]0-*-* (use m68k-*-* instead).
4440       All GCC ports for the following operating systems have been
4441       declared obsolete:
4442          + BeOS (*-*-beos*)
4443          + kaOS (*-*-kaos*)
4444          + GNU/Linux using the a.out object format (*-*-linux*aout*)
4445          + GNU/Linux using version 1 of the GNU C Library
4446            (*-*-linux*libc1*)
4447          + Solaris versions before Solaris 7 (*-*-solaris2.[0-6],
4448            *-*-solaris2.[0-6].*)
4449          + Miscellaneous System V (*-*-sysv*)
4450          + WindISS (*-*-windiss*)
4451       Also, those for some individual systems on particular architectures
4452       have been obsoleted:
4453          + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
4454          + CRIS with a.out object format (cris-*-aout)
4455          + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
4456          + OSF/1 on PA-RISC (hppa1.1-*-osf*)
4457          + PRO on PA-RISC (hppa1.1-*-pro*)
4458          + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*,
4459            i[34567]86-sequent-sysv4*)
4460          + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
4461          + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host
4462            was previously [5]removed in 2001, leaving only the support
4463            for UWIN as a target now being deprecated)
4464          + ChorusOS on PowerPC (powerpc-*-chorusos*)
4465          + All VAX configurations apart from NetBSD and OpenBSD
4466            (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
4467     * The [6]-Wconversion option has been modified. Its purpose now is to
4468       warn for implicit conversions that may alter a value. This new
4469       behavior is available for both C and C++. Warnings about
4470       conversions between signed and unsigned integers can be disabled by
4471       using -Wno-sign-conversion. In C++, they are disabled by default
4472       unless -Wsign-conversion is explicitly requested. The old behavior
4473       of -Wconversion, that is, warn for prototypes causing a type
4474       conversion that is different from what would happen to the same
4475       argument in the absence of a prototype, has been moved to a new
4476       option -Wtraditional-conversion, which is only available for C.
4477     * The -m386, -m486, -mpentium and -mpentiumpro tuning options have
4478       been removed because they were deprecated for more than 3 GCC major
4479       releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or
4480       -mtune=pentiumpro as a replacement.
4481     * The -funsafe-math-optimizations option now automatically turns on
4482       -fno-trapping-math in addition to -fno-signed-zeros, as it enables
4483       reassociation and thus may introduce or remove traps.
4484     * The -ftree-vectorize option is now on by default under -O3. In
4485       order to generate code for a SIMD extension, it has to be enabled
4486       as well: use -maltivec for PowerPC platforms and -msse/-msse2 for
4487       i?86 and x86_64.
4488     * More information on porting to GCC 4.3 from previous versions of
4489       GCC can be found in the [7]porting guide for this release.
4490
4491General Optimizer Improvements
4492
4493     * The GCC middle-end has been integrated with the [8]MPFR library.
4494       This allows GCC to evaluate and replace at compile-time calls to
4495       built-in math functions having constant arguments with their
4496       mathematically equivalent results. In making use of [9]MPFR, GCC
4497       can generate correct results regardless of the math library
4498       implementation or floating point precision of the host platform.
4499       This also allows GCC to generate identical results regardless of
4500       whether one compiles in native or cross-compile configurations to a
4501       particular target. The following built-in functions take advantage
4502       of this new capability: acos, acosh, asin, asinh, atan2, atan,
4503       atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1,
4504       fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10,
4505       log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh,
4506       tan, tanh, tgamma, y0, y1 and yn. The float and long double
4507       variants of these functions (e.g. sinf and sinl) are also handled.
4508       The sqrt and cabs functions with constant arguments were already
4509       optimized in prior GCC releases. Now they also use [10]MPFR.
4510     * A new forward propagation pass on RTL was added. The new pass
4511       replaces several slower transformations, resulting in compile-time
4512       improvements as well as better code generation in some cases.
4513     * A new command-line switch -frecord-gcc-switches has been added to
4514       GCC, although it is only enabled for some targets. The switch
4515       causes the command line that was used to invoke the compiler to be
4516       recorded into the object file that is being created. The exact
4517       format of this recording is target and binary file format
4518       dependent, but it usually takes the form of a note section
4519       containing ASCII text. The switch is related to the -fverbose-asm
4520       switch, but that one only records the information in the assembler
4521       output file as comments, so the information never reaches the
4522       object file.
4523     * The inliner heuristic is now aware of stack frame consumption. New
4524       command-line parameters --param large-stack-frame and --param
4525       large-stack-frame-growth can be used to limit stack frame size
4526       growth caused by inlining.
4527     * During feedback directed optimizations, the expected block size the
4528       memcpy, memset and bzero functions operate on is discovered and for
4529       cases of commonly used small sizes, specialized inline code is
4530       generated.
4531     * __builtin_expect no longer requires its argument to be a compile
4532       time constant.
4533     * Interprocedural optimization was reorganized to work on functions
4534       in SSA form. This enables more precise and cheaper dataflow
4535       analysis and makes writing interprocedural optimizations easier.
4536       The following improvements have been implemented on top of this
4537       framework:
4538          + Pre-inline optimization: Selected local optimization passes
4539            are run before the inliner (and other interprocedural passes)
4540            are executed. This significantly improves the accuracy of code
4541            growth estimates used by the inliner and reduces the overall
4542            memory footprint for large compilation units.
4543          + Early inlining (a simple bottom-up inliner pass inlining only
4544            functions whose body is smaller than the expected call
4545            overhead) is now executed with the early optimization passes,
4546            thus inlining already optimized function bodies into an
4547            unoptimized function that is subsequently optimized by early
4548            optimizers. This enables the compiler to quickly eliminate
4549            abstraction penalty in C++ programs.
4550          + Interprocedural constant propagation now operate on SSA form
4551            increasing accuracy of the analysis.
4552     * A new internal representation for GIMPLE statements has been
4553       contributed, resulting in compile-time memory savings.
4554     * The vectorizer was enhanced to support vectorization of outer
4555       loops, intra-iteration parallelism (loop-aware SLP), vectorization
4556       of strided accesses and loops with multiple data-types. Run-time
4557       dependency testing using loop versioning was added. The cost model,
4558       turned on by -fvect-cost-model, was developed.
4559
4560New Languages and Language specific improvements
4561
4562     * We have added new command-line options
4563       -finstrument-functions-exclude-function-list and
4564       -finstrument-functions-exclude-file-list. They provide more control
4565       over which functions are annotated by the -finstrument-functions
4566       option.
4567
4568  C family
4569
4570     * Implicit conversions between generic vector types are now only
4571       permitted when the two vectors in question have the same number of
4572       elements and compatible element types. (Note that the restriction
4573       involves compatible element types, not implicitly-convertible
4574       element types: thus, a vector type with element type int may not be
4575       implicitly converted to a vector type with element type unsigned
4576       int.) This restriction, which is in line with specifications for
4577       SIMD architectures such as AltiVec, may be relaxed using the flag
4578       -flax-vector-conversions. This flag is intended only as a
4579       compatibility measure and should not be used for new code.
4580     * -Warray-bounds has been added and is now enabled by default for
4581       -Wall . It produces warnings for array subscripts that can be
4582       determined at compile time to be always out of bounds.
4583       -Wno-array-bounds will disable the warning.
4584     * The constructor and destructor function attributes now accept
4585       optional priority arguments which control the order in which the
4586       constructor and destructor functions are run.
4587     * New [11]command-line options -Wtype-limits,
4588       -Wold-style-declaration, -Wmissing-parameter-type, -Wempty-body,
4589       -Wclobbered and -Wignored-qualifiers have been added for finer
4590       control of the diverse warnings enabled by -Wextra.
4591     * A new function attribute alloc_size has been added to mark up
4592       malloc style functions. For constant sized allocations this can be
4593       used to find out the size of the returned pointer using the
4594       __builtin_object_size() function for buffer overflow checking and
4595       similar. This supplements the already built-in malloc and calloc
4596       constant size handling.
4597     * Integer constants written in binary are now supported as a GCC
4598       extension. They consist of a prefix 0b or 0B, followed by a
4599       sequence of 0 and 1 digits.
4600     * A new predefined macro __COUNTER__ has been added. It expands to
4601       sequential integral values starting from 0. In conjunction with the
4602       ## operator, this provides a convenient means to generate unique
4603       identifiers.
4604     * A new command-line option -fdirectives-only has been added. It
4605       enables a special preprocessing mode which improves the performance
4606       of applications like distcc and ccache.
4607     * Fixed-point data types and operators have been added. They are
4608       based on Chapter 4 of the Embedded-C specification (n1169.pdf).
4609       Currently, only MIPS targets are supported.
4610     * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732,
4611       N1241, is now supported as a GCC extension to C for targets
4612       i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu,
4613       and x86_64-*-linux-gnu. The feature introduces new data types
4614       _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF,
4615       DD, and DL.
4616
4617  C++
4618
4619     * [12]Experimental support for the upcoming ISO C++ standard, C++0x.
4620     * -Wc++0x-compat has been added and is now enabled by default for
4621       -Wall. It produces warnings for constructs whose meaning differs
4622       between ISO C++ 1998 and C++0x.
4623     * The -Wparentheses option now works for C++ as it does for C. It
4624       warns if parentheses are omitted when operators with confusing
4625       precedence are nested. It also warns about ambiguous else
4626       statements. Since -Wparentheses is enabled by -Wall, this may cause
4627       additional warnings with existing C++ code which uses -Wall. These
4628       new warnings may be disabled by using -Wall -Wno-parentheses.
4629     * The -Wmissing-declarations now works for C++ as it does for C.
4630     * The -fvisibility-ms-compat flag was added, to make it easier to
4631       port larger projects using shared libraries from Microsoft's Visual
4632       Studio to ELF and Mach-O systems.
4633     * C++ attribute handling has been overhauled for template arguments
4634       (ie dependent types). In particular, __attribute__((aligned(T)));
4635       works for C++ types.
4636
4637    Runtime Library (libstdc++)
4638
4639     * [13]Experimental support for the upcoming ISO C++ standard, C++0x.
4640     * Support for TR1 mathematical special functions and regular
4641       expressions. ([14]Implementation status of TR1)
4642     * Default what implementations give more elaborate exception strings
4643       for bad_cast, bad_typeid, bad_exception, and bad_alloc.
4644     * Header dependencies have been streamlined, reducing unnecessary
4645       includes and pre-processed bloat.
4646     * Variadic template implementations of items in <tuple> and
4647       <functional>.
4648     * An experimental [15]parallel mode has been added. This is a
4649       parallel implementation of many C++ Standard library algorithms,
4650       like std::accumulate, std::for_each, std::transform, or std::sort,
4651       to give but four examples. These algorithms can be substituted for
4652       the normal (sequential) libstdc++ algorithms on a piecemeal basis,
4653       or all existing algorithms can be transformed via the
4654       -D_GLIBCXX_PARALLEL macro.
4655     * Debug mode versions of classes in <unordered_set> and
4656       <unordered_map>.
4657     * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are
4658       now <backward/hash_set> and <backward/hash_map>. This code:
4659    #include <ext/hash_set>
4660    __gnu_cxx::hash_set<int> s;
4661
4662       Can be transformed (in order of preference) to:
4663    #include <tr1/unordered_set>
4664    std::tr1::unordered_set<int> s;
4665
4666       or
4667    #include <backward/hash_set>
4668    __gnu_cxx::hash_set<int> s;
4669
4670       Similar transformations apply to __gnu_cxx::hash_map,
4671       __gnu_cxx::hash_multimap, __gnu_cxx::hash_set,
4672       __gnu_cxx::hash_multiset.
4673
4674  Fortran
4675
4676     * Due to the fact that the [16]GMP and [17]MPFR libraries are
4677       required for all languages, Fortran is no longer special in this
4678       regard and is available by default.
4679     * The [18]-fexternal-blas option has been added, which generates
4680       calls to BLAS routines for intrinsic matrix operations such as
4681       matmul rather than using the built-in algorithms.
4682     * Support to give a backtrace (compiler flag -fbacktrace or
4683       environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems
4684       only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a
4685       run-time error occured.
4686     * GNU Fortran now defines __GFORTRAN__ when it runs the C
4687       preprocessor (CPP).
4688     * The [19]-finit-local-zero, -finit-real, -finit-integer,
4689       -finit-character, and -finit-logical options have been added, which
4690       can be used to initialize local variables.
4691     * The intrinsic procedures [20]GAMMA and [21]LGAMMA have been added,
4692       which calculate the Gamma function and its logarithm. Use EXTERNAL
4693       gamma if you want to use your own gamma function.
4694     * GNU Fortran now regards the backslash character as literal (as
4695       required by the Fortran 2003 standard); using [22]-fbackslash GNU
4696       Fortran interprets backslashes as C-style escape characters.
4697     * The [23]interpretation of binary, octal and hexadecimal (BOZ)
4698       literal constants has been changed. Before they were always
4699       interpreted as integer; now they are bit-wise transferred as
4700       argument of INT, REAL, DBLE and CMPLX as required by the Fortran
4701       2003 standard, and for real and complex variables in DATA
4702       statements or when directly assigned to real and complex variables.
4703       Everywhere else and especially in expressions they are still
4704       regarded as integer constants.
4705     * Fortran 2003 support has been extended:
4706          + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
4707          + Pointer intent
4708          + Intrinsic module ISO_ENV_FORTRAN
4709          + Interoperability with C (ISO C Bindings)
4710          + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER
4711            attribute)
4712          + Fortran 2003 BOZ
4713
4714  Java (GCJ)
4715
4716     * GCJ now uses the Eclipse Java compiler for its Java parsing needs.
4717       This enables the use of all 1.5 language features, and fixes most
4718       existing front end bugs.
4719     * libgcj now supports all 1.5 language features which require runtime
4720       support: foreach, enum, annotations, generics, and auto-boxing.
4721     * We've made many changes to the tools shipped with gcj.
4722          + The old jv-scan tool has been removed. This tool never really
4723            worked properly. There is no replacement.
4724          + gcjh has been rewritten. Some of its more obscure options no
4725            longer work, but are still recognized in an attempt at
4726            compatibility. gjavah is a new program with similar
4727            functionality but different command-line options.
4728          + grmic and grmiregistry have been rewritten. grmid has been
4729            added.
4730          + gjar replaces the old fastjar.
4731          + gjarsigner (used for signing jars), gkeytool (used for key
4732            management), gorbd (for CORBA), gserialver (computes
4733            serialization UIDs), and gtnameserv (also for CORBA) are now
4734            installed.
4735     * The ability to dump the contents of the java run time heap to a
4736       file for off-line analysis has been added. The heap dumps may be
4737       analyzed with the new gc-analyze tool. They may be generated on
4738       out-of-memory conditions or on demand and are controlled by the new
4739       run time class gnu.gcj.util.GCInfo.
4740     * java.util.TimeZone can now read files from /usr/share/zoneinfo to
4741       provide correct, updated, timezone information. This means that
4742       packagers no longer have to update libgcj when a time zone change
4743       is published.
4744
4745New Targets and Target Specific Improvements
4746
4747  IA-32/x86-64
4748
4749     * Tuning for Intel Core 2 processors is available via -mtune=core2
4750       and -march=core2.
4751     * Tuning for AMD Geode processors is available via -mtune=geode and
4752       -march=geode.
4753     * Code generation of block move (memcpy) and block set (memset) was
4754       rewritten. GCC can now pick the best algorithm (loop, unrolled
4755       loop, instruction with rep prefix or a library call) based on the
4756       size of the block being copied and the CPU being optimized for. A
4757       new option -minline-stringops-dynamically has been added. With this
4758       option string operations of unknown size are expanded such that
4759       small blocks are copied by in-line code, while for large blocks a
4760       library call is used. This results in faster code than
4761       -minline-all-stringops when the library implementation is capable
4762       of using cache hierarchy hints. The heuristic choosing the
4763       particular algorithm can be overwritten via -mstringop-strategy.
4764       Newly also memset of values different from 0 is inlined.
4765     * GCC no longer places the cld instruction before string operations.
4766       Both i386 and x86-64 ABI documents mandate the direction flag to be
4767       clear at the entry of a function. It is now invalid to set the flag
4768       in asm statement without reseting it afterward.
4769     * Support for SSSE3 built-in functions and code generation are
4770       available via -mssse3.
4771     * Support for SSE4.1 built-in functions and code generation are
4772       available via -msse4.1.
4773     * Support for SSE4.2 built-in functions and code generation are
4774       available via -msse4.2.
4775     * Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
4776     * A new set of options -mpc32, -mpc64 and -mpc80 have been added to
4777       allow explicit control of x87 floating point precision.
4778     * Support for __float128 (TFmode) IEEE quad type and corresponding
4779       TCmode IEEE complex quad type is available via the soft-fp library
4780       on x86_64 targets. This includes basic arithmetic operations
4781       (addition, subtraction, negation, multiplication and division) on
4782       __float128 real and TCmode complex values, the full set of IEEE
4783       comparisons between __float128 values, conversions to and from
4784       float, double and long double floating point types, as well as
4785       conversions to and from signed or unsigned integer, signed or
4786       unsigned long integer and signed or unsigned quad (TImode) integer
4787       types. Additionally, all operations generate the full set of IEEE
4788       exceptions and support the full set of IEEE rounding modes.
4789     * GCC can now utilize the ACML library for vectorizing calls to a set
4790       of C99 functions on x86_64 if -mveclibabi=acml is specified and you
4791       link to an ACML ABI compatible library.
4792
4793  ARM
4794
4795     * Compiler and Library support for Thumb-2 and the ARMv7 architecture
4796       has been added.
4797
4798  CRIS
4799
4800    New features
4801
4802     * Compiler and Library support for the CRIS v32 architecture, as
4803       found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been
4804       added.
4805
4806    Configuration changes
4807
4808     * The cris-*-elf target now includes support for CRIS v32, including
4809       libraries, through the -march=v32 option.
4810     * A new crisv32-*-elf target defaults to generate code for CRIS v32.
4811     * A new crisv32-*-linux* target defaults to generate code for CRIS
4812       v32.
4813     * The cris-*-aout target has been obsoleted.
4814
4815    Improved support for built-in functions
4816
4817     * GCC can now use the lz and swapwbr instructions to implement the
4818       __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
4819     * __builtin_bswap32 is now implemented using the swapwb instruction,
4820       when available.
4821
4822  m68k and ColdFire
4823
4824    New features
4825
4826     * Support for several new ColdFire processors has been added. You can
4827       generate code for them using the new -mcpu option.
4828     * All targets now support ColdFire processors.
4829     * m68k-uclinux targets have improved support for C++ constructors and
4830       destructors, and for shared libraries.
4831     * It is now possible to set breakpoints on the first or last line of
4832       a function, even if there are no statements on that line.
4833
4834    Optimizations
4835
4836     * Support for sibling calls has been added.
4837     * More use is now made of the ColdFire mov3q instruction.
4838     * __builtin_clz is now implemented using the ff1 ColdFire
4839       instruction, when available.
4840     * GCC now honors the -m68010 option. 68010 code now uses clr rather
4841       than move to zero volatile memory.
4842     * 68020 targets and above can now use symbol(index.size*scale)
4843       addresses for indexed array accesses. Earlier compilers would
4844       always load the symbol into a base register first.
4845
4846    Configuration changes
4847
4848     * All m68k and ColdFire targets now allow the default processor to be
4849       set at configure time using --with-cpu.
4850     * A --with-arch configuration option has been added. This option
4851       allows you to restrict a target to ColdFire or non-ColdFire
4852       processors.
4853
4854    Preprocessor macros
4855
4856     * An __mcfv*__ macro is now defined for all ColdFire targets.
4857       (Earlier versions of GCC only defined __mcfv4e__.)
4858     * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
4859     * All targets now define __mc68010 and __mc68010__ when generating
4860       68010 code.
4861
4862    Command-line changes
4863
4864     * New command-line options -march, -mcpu, -mtune and -mhard-float
4865       have been added. These options apply to both m68k and ColdFire
4866       targets.
4867     * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative
4868       versions of -mshort, etc.
4869     * -fforce-addr has been removed. It is now ignored by the compiler.
4870
4871    Other improvements
4872
4873     * ColdFire targets now try to maintain a 4-byte-aligned stack where
4874       possible.
4875     * m68k-uclinux targets now try to avoid situations that lead to the
4876       load-time error: BINFMT_FLAT: reloc outside program.
4877
4878  MIPS
4879
4880    Changes to existing configurations
4881
4882     * libffi and libjava now support all three GNU/Linux ABIs: o32, n32
4883       and n64. Every GNU/Linux configuration now builds these libraries
4884       by default.
4885     * GNU/Linux configurations now generate -mno-shared code unless
4886       overridden by -fpic, -fPIC, -fpie or -fPIE.
4887     * mipsisa32*-linux-gnu configurations now generate hard-float code by
4888       default, just like other mipsisa32* and mips*-linux-gnu
4889       configurations. You can build a soft-float version of any
4890       mips*-linux-gnu configuration by passing --with-float=soft to
4891       configure.
4892     * mips-wrs-vxworks now supports run-time processes (RTPs).
4893
4894    Changes to existing command-line options
4895
4896     * The -march and -mtune options no longer accept 24k as a processor
4897       name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
4898     * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and
4899       34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The
4900       options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for
4901       24kx, 24kex and 34kx.
4902
4903    New configurations
4904
4905   GCC now supports the following configurations:
4906     * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by
4907       default. Earlier releases also recognized this configuration, but
4908       they treated it in the same way as mipsisa32*-linux-gnu*. Note that
4909       you can customize any mips*-linux-gnu* configuration to a
4910       particular ISA or processor by passing an appropriate --with-arch
4911       option to configure.
4912     * mipsisa*-sde-elf*, which provides compatibility with MIPS
4913       Technologies' SDE toolchains. The configuration uses the SDE
4914       libraries by default, but you can use it like other newlib-based
4915       ELF configurations by passing --with-newlib to configure. It is the
4916       only configuration besides mips64vr*-elf* to build MIPS16 as well
4917       as non-MIPS16 libraries.
4918     * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf*
4919       configuration, but uses the o32 and o64 ABIs instead of the 32-bit
4920       and 64-bit forms of the EABI.
4921
4922    New processors and application-specific extensions
4923
4924     * Support for the SmartMIPS ASE is available through the new
4925       -msmartmips option.
4926     * Support for revision 2 of the DSP ASE is available through the new
4927       -mdspr2 option. A new preprocessor macro called __mips_dsp_rev
4928       indicates the revision of the ASE in use.
4929     * Support for the 4KS and 74K families of processors is available
4930       through the -march and -mtune options.
4931
4932    Improved support for built-in functions
4933
4934     * GCC can now use load-linked, store-conditional and sync
4935       instructions to implement atomic built-in functions such as
4936       __sync_fetch_and_add. The memory reference must be 4 bytes wide for
4937       32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
4938     * GCC can now use the clz and dclz instructions to implement the
4939       __builtin_ctz and __builtin_ffs families of functions.
4940     * There is a new __builtin___clear_cache function for flushing the
4941       instruction cache. GCC expands this function inline on MIPS32
4942       revision 2 targets, otherwise it calls the function specified by
4943       -mcache-flush-func.
4944
4945    MIPS16 improvements
4946
4947     * GCC can now compile objects that contain a mixture of MIPS16 and
4948       non-MIPS16 code. There are two new attributes, mips16 and nomips16,
4949       for specifying which mode a function should use.
4950     * A new option called -minterlink-mips16 makes non-MIPS16 code
4951       link-compatible with MIPS16 code.
4952     * After many bug fixes, the long-standing MIPS16 -mhard-float support
4953       should now work fairly reliably.
4954     * GCC can now use the MIPS16e save and restore instructions.
4955     * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled
4956       with -G0 -fsection-anchors is often smaller than code compiled with
4957       -G8. However, please note that you must usually compile all objects
4958       in your application with the same -G option; see the documentation
4959       of -G for details.
4960     * A new option called-mcode-readable specifies which instructions are
4961       allowed to load from the code segment. -mcode-readable=yes is the
4962       default and says that any instruction may load from the code
4963       segment. The other alternatives are -mcode-readable=pcrel, which
4964       says that only PC-relative MIPS16 instructions may load from the
4965       code segment, and -mcode-readable=no, which says that no
4966       instruction may do so. Please see the documentation for more
4967       details, including example uses.
4968
4969    Small-data improvements
4970
4971   There are three new options for controlling small data:
4972     * -mno-extern-sdata, which disables small-data accesses for
4973       externally-defined variables. Code compiled with -Gn
4974       -mno-extern-sdata will be link-compatible with any -G setting
4975       between -G0 and -Gn inclusive.
4976     * -mno-local-sdata, which disables the use of small-data sections for
4977       data that is not externally visible. This option can be a useful
4978       way of reducing small-data usage in less performance-critical parts
4979       of an application.
4980     * -mno-gpopt, which disables the use of the $gp register while still
4981       honoring the -G limit when placing externally-visible data. This
4982       option implies -mno-extern-sdata and -mno-local-sdata and it can be
4983       useful in situations where $gp does not necessarily hold the
4984       expected value.
4985
4986    Miscellaneous improvements
4987
4988     * There is a new option called -mbranch-cost for tweaking the
4989       perceived cost of branches.
4990     * If GCC is configured to use a version of GAS that supports the
4991       .gnu_attribute directive, it will use that directive to record
4992       certain properties of the output code. .gnu_attribute is new to GAS
4993       2.18.
4994     * There are two new function attributes, near and far, for overriding
4995       the command-line setting of -mlong-calls on a function-by-function
4996       basis.
4997     * -mfp64, which previously required a 64-bit target, now works with
4998       MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and
4999       mipsisa*-sde-elf* configurations provide suitable library support.
5000     * GCC now recognizes the -mdmx and -mmt options and passes them down
5001       to the assembler. It does nothing else with the options at present.
5002
5003  SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture
5004  (BEA)
5005
5006     * Support has been added for this new architecture.
5007
5008  RS6000 (POWER/PowerPC)
5009
5010     * Support for the PowerPC 750CL paired-single instructions has been
5011       added with a new powerpc-*-linux*paired* target configuration. It
5012       is enabled by an associated -mpaired option and can be accessed
5013       using new built-in functions.
5014     * Support for auto-detecting architecture and system configuration to
5015       auto-select processor optimization tuning.
5016     * Support for VMX on AIX 5.3 has been added.
5017     * Support for AIX Version 6.1 has been added.
5018
5019  S/390, zSeries and System z9
5020
5021     * Support for the IBM System z9 EC/BC processor (z9 GA3) has been
5022       added. When using the -march=z9-ec option, the compiler will
5023       generate code making use of instructions provided by the decimal
5024       floating point facility and the floating point conversion facility
5025       (pfpo). Besides the instructions used to implement decimal floating
5026       point operations these facilities also contain instructions to move
5027       between general purpose and floating point registers and to modify
5028       and copy the sign-bit of floating point values.
5029     * When the -march=z9-ec option is used the new
5030       -mhard-dfp/-mno-hard-dfp options can be used to specify whether the
5031       decimal floating point hardware instructions will be used or not.
5032       If none of them is given the hardware support is enabled by
5033       default.
5034     * The -mstack-guard option can now be omitted when using stack
5035       checking via -mstack-size in order to let GCC choose a sensible
5036       stack guard value according to the frame size of each function.
5037     * Various changes to improve performance of generated code have been
5038       implemented, including:
5039          + The condition code set by an add logical with carry
5040            instruction is now available for overflow checks like: a + b +
5041            carry < b.
5042          + The test data class instruction is now used to implement
5043            sign-bit and infinity checks of binary and decimal floating
5044            point numbers.
5045
5046  SPARC
5047
5048     * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been
5049       added.
5050
5051  Xtensa
5052
5053     * Stack unwinding for exception handling now uses by default a
5054       specialized version of DWARF unwinding. This is not
5055       binary-compatible with the setjmp/longjmp (sjlj) unwinding used for
5056       Xtensa with previous versions of GCC.
5057     * For Xtensa processors that include the Conditional Store option,
5058       the built-in functions for atomic memory access are now implemented
5059       using S32C1I instructions.
5060     * If the Xtensa NSA option is available, GCC will use it to implement
5061       the __builtin_ctz and __builtin_clz functions.
5062
5063Documentation improvements
5064
5065     * Existing libstdc++ documentation has been edited and restructured
5066       into a single DocBook XML manual. The results can be viewed online
5067       [24]here.
5068
5069Other significant improvements
5070
5071     * The compiler's --help command-line option has been extended so that
5072       it now takes an optional set of arguments. These arguments restrict
5073       the information displayed to specific classes of command-line
5074       options, and possibly only a subset of those options. It is also
5075       now possible to replace the descriptive text associated with each
5076       displayed option with an indication of its current value, or for
5077       binary options, whether it has been enabled or disabled.
5078       Here are some examples. The following will display all the options
5079       controlling warning messages:
5080      --help=warnings
5081
5082       Whereas this will display all the undocumented, target specific
5083       options:
5084      --help=target,undocumented
5085
5086       This sequence of commands will display the binary optimizations
5087       that are enabled by -O3:
5088      gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
5089      gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
5090      diff /tmp/O2-opts /tmp/O3-opts | grep enabled
5091
5092     * The configure options --with-pkgversion and --with-bugurl have been
5093       added. These allow distributors of GCC to include a
5094       distributor-specific string in manuals and --version output and to
5095       specify the URL for reporting bugs in their versions of GCC.
5096
5097GCC 4.3.1
5098
5099   This is the [25]list of problem reports (PRs) from GCC's bug tracking
5100   system that are known to be fixed in the 4.3.1 release. This list might
5101   not be complete (that is, it is possible that some PRs that have been
5102   fixed are not listed here).
5103
5104Target Specific Changes
5105
5106  IA-32/x86-64
5107
5108    ABI changes
5109
5110     * Starting with GCC 4.3.1, decimal floating point variables are
5111       aligned to their natural boundaries when they are passed on the
5112       stack for i386.
5113
5114    Command-line changes
5115
5116     * Starting with GCC 4.3.1, the -mcld option has been added to
5117       automatically generate a cld instruction in the prologue of
5118       functions that use string instructions. This option is used for
5119       backward compatibility on some operating systems and can be enabled
5120       by default for 32-bit x86 targets by configuring GCC with the
5121       --enable-cld configure option.
5122
5123GCC 4.3.2
5124
5125   This is the [26]list of problem reports (PRs) from GCC's bug tracking
5126   system that are known to be fixed in the 4.3.2 release. This list might
5127   not be complete (that is, it is possible that some PRs that have been
5128   fixed are not listed here).
5129
5130GCC 4.3.3
5131
5132   This is the [27]list of problem reports (PRs) from GCC's bug tracking
5133   system that are known to be fixed in the 4.3.3 release. This list might
5134   not be complete (that is, it is possible that some PRs that have been
5135   fixed are not listed here).
5136
5137GCC 4.3.4
5138
5139   This is the [28]list of problem reports (PRs) from GCC's bug tracking
5140   system that are known to be fixed in the 4.3.4 release. This list might
5141   not be complete (that is, it is possible that some PRs that have been
5142   fixed are not listed here).
5143
5144GCC 4.3.5
5145
5146   This is the [29]list of problem reports (PRs) from GCC's bug tracking
5147   system that are known to be fixed in the 4.3.5 release. This list might
5148   not be complete (that is, it is possible that some PRs that have been
5149   fixed are not listed here).
5150
5151GCC 4.3.6
5152
5153   This is the [30]list of problem reports (PRs) from GCC's bug tracking
5154   system that are known to be fixed in the 4.3.6 release. This list might
5155   not be complete (that is, it is possible that some PRs that have been
5156   fixed are not listed here).
5157
5158
5159    For questions related to the use of GCC, please consult these web
5160    pages and the [31]GCC manuals. If that fails, the
5161    [32]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5162    web pages and the development of GCC are welcome on our developer
5163    list at [33]gcc@gcc.gnu.org. All of [34]our lists have public
5164    archives.
5165
5166   Copyright (C) [35]Free Software Foundation, Inc. Verbatim copying and
5167   distribution of this entire article is permitted in any medium,
5168   provided this notice is preserved.
5169
5170   These pages are [36]maintained by the GCC team. Last modified
5171   2012-11-02[37].
5172
5173References
5174
5175   1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5
5176   2. http://gmplib.org/
5177   3. http://www.mpfr.org/
5178   4. http://gcc.gnu.org/install/prerequisites.html
5179   5. http://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html
5180   6. http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
5181   7. http://gcc.gnu.org/gcc-4.3/porting_to.html
5182   8. http://www.mpfr.org/
5183   9. http://www.mpfr.org/
5184  10. http://www.mpfr.org/
5185  11. http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
5186  12. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5187  13. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5188  14. http://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#m anual.intro.status.standard.tr1
5189  15. http://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
5190  16. http://gmplib.org/
5191  17. http://www.mpfr.org/
5192  18. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options
5193  19. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167
5194  20. http://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html
5195  21. http://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html
5196  22. http://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html
5197  23. http://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html
5198  24. http://gcc.gnu.org/onlinedocs/libstdc++/
5199  25. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1
5200  26. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2
5201  27. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3
5202  28. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4
5203  29. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5
5204  30. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6
5205  31. http://gcc.gnu.org/onlinedocs/
5206  32. mailto:gcc-help@gcc.gnu.org
5207  33. mailto:gcc@gcc.gnu.org
5208  34. http://gcc.gnu.org/lists.html
5209  35. http://www.fsf.org/
5210  36. http://gcc.gnu.org/about.html
5211  37. http://validator.w3.org/check/referer
5212======================================================================
5213http://gcc.gnu.org/gcc-4.2/index.html
5214
5215                             GCC 4.2 Release Series
5216
5217   May 19, 2008
5218
5219   The [1]GNU project and the GCC developers are pleased to announce the
5220   release of GCC 4.2.4.
5221
5222   This release is a bug-fix release, containing fixes for regressions in
5223   GCC 4.2.3 relative to previous releases of GCC.
5224
5225Release History
5226
5227   GCC 4.2.4
5228          May 19, 2008 ([2]changes)
5229
5230   GCC 4.2.3
5231          February 1, 2008 ([3]changes)
5232
5233   GCC 4.2.2
5234          October 7, 2007 ([4]changes)
5235
5236   GCC 4.2.1
5237          July 18, 2007 ([5]changes)
5238
5239   GCC 4.2.0
5240          May 13, 2007 ([6]changes)
5241
5242References and Acknowledgements
5243
5244   GCC used to stand for the GNU C Compiler, but since the compiler
5245   supports several other languages aside from C, it now stands for the
5246   GNU Compiler Collection.
5247
5248   A list of [7]successful builds is updated as new information becomes
5249   available.
5250
5251   The GCC developers would like to thank the numerous people that have
5252   contributed new features, improvements, bug fixes, and other changes as
5253   well as test results to GCC. This [8]amazing group of volunteers is
5254   what makes GCC successful.
5255
5256   For additional information about GCC please refer to the [9]GCC project
5257   web site or contact the [10]GCC development mailing list.
5258
5259   To obtain GCC please use [11]our mirror sites or [12]our SVN server.
5260
5261
5262    For questions related to the use of GCC, please consult these web
5263    pages and the [13]GCC manuals. If that fails, the
5264    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5265    web pages and the development of GCC are welcome on our developer
5266    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
5267    archives.
5268
5269   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
5270   distribution of this entire article is permitted in any medium,
5271   provided this notice is preserved.
5272
5273   These pages are [18]maintained by the GCC team. Last modified
5274   2012-11-02[19].
5275
5276References
5277
5278   1. http://www.gnu.org/
5279   2. http://gcc.gnu.org/gcc-4.2/changes.html
5280   3. http://gcc.gnu.org/gcc-4.2/changes.html
5281   4. http://gcc.gnu.org/gcc-4.2/changes.html
5282   5. http://gcc.gnu.org/gcc-4.2/changes.html
5283   6. http://gcc.gnu.org/gcc-4.2/changes.html
5284   7. http://gcc.gnu.org/gcc-4.2/buildstat.html
5285   8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5286   9. http://gcc.gnu.org/index.html
5287  10. mailto:gcc@gcc.gnu.org
5288  11. http://gcc.gnu.org/mirrors.html
5289  12. http://gcc.gnu.org/svn.html
5290  13. http://gcc.gnu.org/onlinedocs/
5291  14. mailto:gcc-help@gcc.gnu.org
5292  15. mailto:gcc@gcc.gnu.org
5293  16. http://gcc.gnu.org/lists.html
5294  17. http://www.fsf.org/
5295  18. http://gcc.gnu.org/about.html
5296  19. http://validator.w3.org/check/referer
5297======================================================================
5298http://gcc.gnu.org/gcc-4.2/changes.html
5299
5300                             GCC 4.2 Release Series
5301                        Changes, New Features, and Fixes
5302
5303Caveats
5304
5305     * GCC no longer accepts the -fshared-data option. This option has had
5306       no effect in any GCC 4 release; the targets to which the option
5307       used to apply had been removed before GCC 4.0.
5308
5309General Optimizer Improvements
5310
5311     * New command-line options specify the possible relationships among
5312       parameters and between parameters and global data. For example,
5313       -fargument-noalias-anything specifies that arguments do not alias
5314       any other storage.
5315       Each language will automatically use whatever option is required by
5316       the language standard. You should not need to use these options
5317       yourself.
5318
5319New Languages and Language specific improvements
5320
5321     * [1]OpenMP is now supported for the C, C++ and Fortran compilers.
5322     * New command-line options -fstrict-overflow and -Wstrict-overflow
5323       have been added. -fstrict-overflow tells the compiler that it may
5324       assume that the program follows the strict signed overflow
5325       semantics permitted for the language: for C and C++ this means that
5326       the compiler may assume that signed overflow does not occur. For
5327       example, a loop like
5328      for (i = 1; i > 0; i *= 2)
5329
5330       is presumably intended to continue looping until i overflows. With
5331       -fstrict-overflow, the compiler may assume that signed overflow
5332       will not occur, and transform this into an infinite loop.
5333       -fstrict-overflow is turned on by default at -O2, and may be
5334       disabled via -fno-strict-overflow. The -Wstrict-overflow option may
5335       be used to warn about cases where the compiler assumes that signed
5336       overflow will not occur. It takes five different levels:
5337       -Wstrict-overflow=1 to 5. See the [2]documentation for details.
5338       -Wstrict-overflow=1 is enabled by -Wall.
5339     * The new command-line option -fno-toplevel-reorder directs GCC to
5340       emit top-level functions, variables, and asm statements in the same
5341       order that they appear in the input file. This is intended to
5342       support existing code which relies on a particular ordering (for
5343       example, code which uses top-level asm statements to switch
5344       sections). For new code, it is generally better to use function and
5345       variable attributes. The -fno-toplevel-reorder option may be used
5346       for most cases which currently use -fno-unit-at-a-time. The
5347       -fno-unit-at-a-time option will be removed in some future version
5348       of GCC. If you know of a case which requires -fno-unit-at-a-time
5349       which is not fixed by -fno-toplevel-reorder, please open a bug
5350       report.
5351
5352  C family
5353
5354     * The pragma redefine_extname will now macro expand its tokens for
5355       compatibility with SunPRO.
5356     * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct
5357       GCC to handle inline functions as specified in the C99 standard. In
5358       preparation for this, GCC 4.2 will warn about any use of non-static
5359       inline functions in gnu99 or c99 mode. This new warning may be
5360       disabled with the new gnu_inline function attribute or the new
5361       -fgnu89-inline command-line option. Also, GCC 4.2 and later will
5362       define one of the preprocessor macros __GNUC_GNU_INLINE__ or
5363       __GNUC_STDC_INLINE__ to indicate the semantics of inline functions
5364       in the current compilation.
5365     * A new command-line option -Waddress has been added to warn about
5366       suspicious uses of memory addresses as, for example, using the
5367       address of a function in a conditional expression, and comparisons
5368       against the memory address of a string literal. This warning is
5369       enabled by -Wall.
5370
5371  C++
5372
5373     * C++ visibility handling has been overhauled.
5374       Restricted visiblity is propagated from classes to members, from
5375       functions to local statics, and from templates and template
5376       arguments to instantiations, unless the latter has explicitly
5377       declared visibility.
5378       The visibility attribute for a class must come between the
5379       class-key and the name, not after the closing brace.
5380       Attributes are now allowed for enums and elaborated-type-specifiers
5381       that only declare a type.
5382       Members of the anonymous namespace are now local to a particular
5383       translation unit, along with any other declarations which use them,
5384       though they are still treated as having external linkage for
5385       language semantics.
5386     * The (undocumented) extension which permitted templates with default
5387       arguments to be bound to template template parameters with fewer
5388       parameters has been removed. For example:
5389        template <template <typename> class C>
5390        void f(C<double>) {}
5391
5392        template <typename T, typename U = int>
5393        struct S {};
5394
5395        template void f(S<double>);
5396
5397       is no longer accepted by G++. The reason this code is not accepted
5398       is that S is a template with two parameters; therefore, it cannot
5399       be bound to C which has only one parameter.
5400     * The <?, >?, <?=, and >?= operators, deprecated in previous GCC
5401       releases, have been removed.
5402     * The command-line option -fconst-strings, deprecated in previous GCC
5403       releases, has been removed.
5404     * The configure variable enable-__cxa_atexit is now enabled by
5405       default for more targets. Enabling this variable is necessary in
5406       order for static destructors to be executed in the correct order,
5407       but it depends upon the presence of a non-standard C library in the
5408       target library in order to work. The variable is now enabled for
5409       more targets which are known to have suitable C libraries.
5410     * -Wextra will produce warnings for if statements with a semicolon as
5411       the only body, to catch code like:
5412         if (a);
5413            return 1;
5414         return 0;
5415
5416       To suppress the warning in valid cases, use { } instead.
5417     * The C++ frontend now also produces strict aliasing warnings when
5418       -fstrict-aliasing -Wstrict-aliasing is in effect.
5419
5420    Runtime Library (libstdc++)
5421
5422     * Added support for TR1 <random>, <complex>, and C compatibility
5423       headers. In addition, a lock-free version of shared_ptr was
5424       contributed as part of Phillip Jordan's Google Summer of Code
5425       project on lock-free containers. ([3]Implementation status of TR1)
5426     * In association with the Summer of Code work on lock-free
5427       containers, the interface for atomic builtins was adjusted,
5428       creating simpler alternatives for non-threaded code paths. Also,
5429       usage was consolidated and all elements were moved from namespace
5430       std to namespace__gnu_cxx. Affected interfaces are the functions
5431       __exchange_and_add, __atomic_add, and the objects __mutex,
5432       __recursive_mutex, and __scoped_lock.
5433     * Support for versioning weak symbol names via namespace association
5434       was added. However, as this changes the names of exported symbols,
5435       this is turned off by default in the current ABI. Intrepid users
5436       can enable this feature by using
5437       --enable-symvers=gnu-versioned-namespace during configuration.
5438     * Revised, simplified, and expanded policy-based associative
5439       containers, including data types for tree and trie forms
5440       (basic_tree, tree, trie), lists (list_update), and both
5441       collision-chaining and probing hash-based containers
5442       (basic_hash_table, cc_hash_table, gp_hash_table). More details per
5443       the [4]documentation.
5444     * The implementation of the debug mode was modified, whereby the
5445       debug namespaces were nested inside of namespace std and namespace
5446       __gnu_cxx in order to resolve some long standing corner cases
5447       involving name lookup. Debug functionality from the policy-based
5448       data structures was consolidated and enabled with the single macro,
5449       _GLIBCXX_DEBUG. See PR 26142 for more information.
5450     * Added extensions for type traits: __conditional_type,
5451       __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if.
5452     * Added a typelist implementation for compile-time meta-programming.
5453       Elements for typelist construction and operation can be found
5454       within namespace __gnu_cxx::typelist.
5455     * Added a new allocator, __gnu_cxx::throw_allocator, for testing
5456       exception-safety.
5457     * Enabled library-wide visibility control, allowing -fvisibility to
5458       be used.
5459     * Consolidated all nested namespaces and the conversion of
5460       __gnu_internal implementation-private details to anonymous
5461       namespaces whenever possible.
5462     * Implemented LWG resolutions DR 431 and DR 538.
5463
5464  Fortran
5465
5466     * Support for allocatable components has been added (TR 15581 and
5467       Fortran 2003).
5468     * Support for the Fortran 2003 streaming IO extension has been added.
5469     * The GNU Fortran compiler now uses 4-byte record markers by default
5470       for unformatted files to be compatible with g77 and most other
5471       compilers. The implementation allows for records greater than 2 GB
5472       and is compatible with several other compilers. Older versions of
5473       gfortran used 8-byte record markers by default (on most systems).
5474       In order to change the length of the record markers, e.g. to read
5475       unformatted files created by older gfortran versions, the
5476       [5]-frecord-marker=8 option can be used.
5477
5478  Java (GCJ)
5479
5480     * A new command-line option -static-libgcj has been added for targets
5481       that use a linker compatible with GNU Binutils. As its name
5482       implies, this causes libgcj to be linked statically. In some cases
5483       this causes the resulting executable to start faster and use less
5484       memory than if the shared version of libgcj were used. However
5485       caution should be used as it can also cause essential parts of the
5486       library to be omitted. Some of these issues are discussed in:
5487       [6]http://gcc.gnu.org/wiki/Statically_linking_libgcj
5488     * fastjar is no longer bundled with GCC. To build libgcj, you will
5489       need either InfoZIP (both zip and unzip) or an external jar
5490       program. In the former case, the GCC build will install a jar shell
5491       script that is based on InfoZIP and provides the same functionality
5492       as fastjar.
5493
5494New Targets and Target Specific Improvements
5495
5496  IA-32/x86-64
5497
5498     * -mtune=generic can now be used to generate code running well on
5499       common x86 chips. This includes AMD Athlon, AMD Opteron, Intel
5500       Pentium-M, Intel Pentium 4 and Intel Core 2.
5501     * -mtune=native and -march=native will produce code optimized for the
5502       host architecture as detected using the cpuid instruction.
5503     * Added a new command-line option -fstackrealign and and
5504       __attribute__ ((force_align_arg_pointer)) to realign the stack at
5505       runtime. This allows functions compiled with a vector-aligned stack
5506       to be invoked from legacy objects that keep only word-alignment.
5507
5508  SPARC
5509
5510     * The default CPU setting has been changed from V7 to V9 in 32-bit
5511       mode on Solaris 7 and above. This is already the case in 64-bit
5512       mode. It can be overridden by specifying --with-cpu at configure
5513       time.
5514     * Back-end support of built-in functions for atomic memory access has
5515       been implemented.
5516     * Support for the Sun UltraSPARC T1 (Niagara) processor has been
5517       added.
5518
5519  M32C
5520
5521     * Various bug fixes have made some functions (notably, functions
5522       returning structures) incompatible with previous releases.
5523       Recompiling all libraries is recommended. Note that code quality
5524       has considerably improved since 4.1, making a recompile even more
5525       beneficial.
5526
5527  MIPS
5528
5529     * Added support for the Broadcom SB-1A core.
5530
5531  IA-64
5532
5533     * Added support for IA-64 data and control speculation. By default
5534       speculation is enabled only during second scheduler pass. A number
5535       of machine flags was introduced to control the usage of speculation
5536       for both scheduler passes.
5537
5538  HPPA
5539
5540     * Added Java language support (libffi and libjava) for 32-bit HP-UX
5541       11 target.
5542
5543Obsolete Systems
5544
5545Documentation improvements
5546
5547  PDF Documentation
5548
5549     * A make pdf target has been added to the top-level makefile,
5550       enabling automated production of PDF documentation files.
5551       (Front-ends external to GCC should modify their Make-lang.in file
5552       to add a lang.pdf: target.)
5553
5554Other significant improvements
5555
5556  Build system improvements
5557
5558     * All the components of the compiler are now bootstrapped by default.
5559       This improves the resilience to bugs in the system compiler or
5560       binary compatibility problems, as well as providing better testing
5561       of GCC 4.2 itself. In addition, if you build the compiler from a
5562       combined tree, the assembler, linker, etc. will also be
5563       bootstrapped (i.e. built with themselves).
5564       You can disable this behavior, and go back to the pre-GCC 4.2 set
5565       up, by configuring GCC with --disable-bootstrap.
5566     * The rules that configure follows to find target tools resemble more
5567       closely the locations that the built compiler will search. In
5568       addition, you can use the new configure option --with-target-tools
5569       to specify where to find the target tools used during the build,
5570       without affecting what the built compiler will use.
5571       This can be especially useful when building packages of GCC. For
5572       example, you may want to build GCC with GNU as or ld, even if the
5573       resulting compiler to work with the native assembler and linker. To
5574       do so, you can use --with-target-tools to point to the native
5575       tools.
5576
5577  Incompatible changes to the build system
5578
5579     * Front-ends external to GCC should modify their Make-lang.in file to
5580       replace double-colon rules (e.g. dvi::) with normal rules (like
5581       lang.dvi:). Front-end makefile hooks do not use double-colon rules
5582       anymore.
5583     * Up to GCC 4.1, a popular way to specify the target tools used
5584       during the build was to create directories named gas, binutils,
5585       etc. in the build tree, and create links to the tools from there.
5586       This does not work any more when the compiler is bootstrapped. The
5587       new configure option --with-target-tools provides a better way to
5588       achieve the same effect, and works for all native and cross
5589       settings.
5590
5591
5592    For questions related to the use of GCC, please consult these web
5593    pages and the [7]GCC manuals. If that fails, the
5594    [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5595    web pages and the development of GCC are welcome on our developer
5596    list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
5597    archives.
5598
5599   Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
5600   distribution of this entire article is permitted in any medium,
5601   provided this notice is preserved.
5602
5603   These pages are [12]maintained by the GCC team. Last modified
5604   2012-11-02[13].
5605
5606References
5607
5608   1. http://gcc.gnu.org/projects/gomp/
5609   2. http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
5610   3. http://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#manual.intro.status.standard.tr1
5611   4. http://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html
5612   5. http://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html
5613   6. http://gcc.gnu.org/wiki/Statically_linking_libgcj
5614   7. http://gcc.gnu.org/onlinedocs/
5615   8. mailto:gcc-help@gcc.gnu.org
5616   9. mailto:gcc@gcc.gnu.org
5617  10. http://gcc.gnu.org/lists.html
5618  11. http://www.fsf.org/
5619  12. http://gcc.gnu.org/about.html
5620  13. http://validator.w3.org/check/referer
5621======================================================================
5622http://gcc.gnu.org/gcc-4.1/index.html
5623
5624                             GCC 4.1 Release Series
5625
5626   February 13, 2007
5627
5628   The [1]GNU project and the GCC developers are pleased to announce the
5629   release of GCC 4.1.2.
5630
5631   This release is a bug-fix release, containing fixes for regressions in
5632   GCC 4.1.1 relative to previous releases of GCC.
5633
5634Release History
5635
5636   GCC 4.1.2
5637          February 13, 2007 ([2]changes)
5638
5639   GCC 4.1.1
5640          May 24, 2006 ([3]changes)
5641
5642   GCC 4.1.0
5643          February 28, 2006 ([4]changes)
5644
5645References and Acknowledgements
5646
5647   GCC used to stand for the GNU C Compiler, but since the compiler
5648   supports several other languages aside from C, it now stands for the
5649   GNU Compiler Collection.
5650
5651   A list of [5]successful builds is updated as new information becomes
5652   available.
5653
5654   The GCC developers would like to thank the numerous people that have
5655   contributed new features, improvements, bug fixes, and other changes as
5656   well as test results to GCC. This [6]amazing group of volunteers is
5657   what makes GCC successful.
5658
5659   For additional information about GCC please refer to the [7]GCC project
5660   web site or contact the [8]GCC development mailing list.
5661
5662   To obtain GCC please use [9]our mirror sites or [10]our SVN server.
5663
5664
5665    For questions related to the use of GCC, please consult these web
5666    pages and the [11]GCC manuals. If that fails, the
5667    [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5668    web pages and the development of GCC are welcome on our developer
5669    list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
5670    archives.
5671
5672   Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
5673   distribution of this entire article is permitted in any medium,
5674   provided this notice is preserved.
5675
5676   These pages are [16]maintained by the GCC team. Last modified
5677   2012-11-02[17].
5678
5679References
5680
5681   1. http://www.gnu.org/
5682   2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
5683   3. http://gcc.gnu.org/gcc-4.1/changes.html
5684   4. http://gcc.gnu.org/gcc-4.1/changes.html
5685   5. http://gcc.gnu.org/gcc-4.1/buildstat.html
5686   6. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5687   7. http://gcc.gnu.org/index.html
5688   8. mailto:gcc@gcc.gnu.org
5689   9. http://gcc.gnu.org/mirrors.html
5690  10. http://gcc.gnu.org/svn.html
5691  11. http://gcc.gnu.org/onlinedocs/
5692  12. mailto:gcc-help@gcc.gnu.org
5693  13. mailto:gcc@gcc.gnu.org
5694  14. http://gcc.gnu.org/lists.html
5695  15. http://www.fsf.org/
5696  16. http://gcc.gnu.org/about.html
5697  17. http://validator.w3.org/check/referer
5698======================================================================
5699http://gcc.gnu.org/gcc-4.1/changes.html
5700
5701                             GCC 4.1 Release Series
5702                        Changes, New Features, and Fixes
5703
5704   The latest release in the 4.1 release series is [1]GCC 4.1.2.
5705
5706Caveats
5707
5708General Optimizer Improvements
5709
5710     * GCC now has infrastructure for inter-procedural optimizations and
5711       the following inter-procedural optimizations are implemented:
5712          + Profile guided inlining. When doing profile feedback guided
5713            optimization, GCC can now use the profile to make better
5714            informed decisions on whether inlining of a function is
5715            profitable or not. This means that GCC will no longer inline
5716            functions at call sites that are not executed very often, and
5717            that functions at hot call sites are more likely to be
5718            inlined.
5719            A new parameter min-inline-recursive-probability is also now
5720            available to throttle recursive inlining of functions with
5721            small average recursive depths.
5722          + Discovery of pure and const functions, a form of side-effects
5723            analysis. While older GCC releases could also discover such
5724            special functions, the new IPA-based pass runs earlier so that
5725            the results are available to more optimizers. The pass is also
5726            simply more powerful than the old one.
5727          + Analysis of references to static variables and type escape
5728            analysis, also forms of side-effects analysis. The results of
5729            these passes allow the compiler to be less conservative about
5730            call-clobbered variables and references. This results in more
5731            redundant loads being eliminated and in making static
5732            variables candidates for register promotion.
5733          + Improvement of RTL-based alias analysis. The results of type
5734            escape analysis are fed to the RTL type-based alias analyzer,
5735            allowing it to disambiguate more memory references.
5736          + Interprocedural constant propagation and function versioning.
5737            This pass looks for functions that are always called with the
5738            same constant value for one or more of the function arguments,
5739            and propagates those constants into those functions.
5740          + GCC will now eliminate static variables whose usage was
5741            optimized out.
5742          + -fwhole-program --combine can now be used to make all
5743            functions in program static allowing whole program
5744            optimization. As an exception, the main function and all
5745            functions marked with the new externally_visible attribute are
5746            kept global so that programs can link with runtime libraries.
5747     * GCC can now do a form of partial dead code elimination (PDCE) that
5748       allows code motion of expressions to the paths where the result of
5749       the expression is actually needed. This is not always a win, so the
5750       pass has been limited to only consider profitable cases. Here is an
5751       example:
5752    int foo (int *, int *);
5753    int
5754    bar (int d)
5755    {
5756      int a, b, c;
5757      b = d + 1;
5758      c = d + 2;
5759      a = b + c;
5760      if (d)
5761        {
5762          foo (&b, &c);
5763          a = b + c;
5764        }
5765      printf ("%d\n", a);
5766    }
5767
5768       The a = b + c can be sunk to right before the printf. Normal code
5769       sinking will not do this, it will sink the first one above into the
5770       else-branch of the conditional jump, which still gives you two
5771       copies of the code.
5772     * GCC now has a value range propagation pass. This allows the
5773       compiler to eliminate bounds checks and branches. The results of
5774       the pass can also be used to accurately compute branch
5775       probabilities.
5776     * The pass to convert PHI nodes to straight-line code (a form of
5777       if-conversion for GIMPLE) has been improved significantly. The two
5778       most significant improvements are an improved algorithm to
5779       determine the order in which the PHI nodes are considered, and an
5780       improvement that allow the pass to consider if-conversions of basic
5781       blocks with more than two predecessors.
5782     * Alias analysis improvements. GCC can now differentiate between
5783       different fields of structures in Tree-SSA's virtual operands form.
5784       This lets stores/loads from non-overlapping structure fields not
5785       conflict. A new algorithm to compute points-to sets was contributed
5786       that can allows GCC to see now that p->a and p->b, where p is a
5787       pointer to a structure, can never point to the same field.
5788     * Various enhancements to auto-vectorization:
5789          + Incrementally preserve SSA form when vectorizing.
5790          + Incrementally preserve loop-closed form when vectorizing.
5791          + Improvements to peeling for alignment: generate better code
5792            when the misalignment of an access is known at compile time,
5793            or when different accesses are known to have the same
5794            misalignment, even if the misalignment amount itself is
5795            unknown.
5796          + Consider dependence distance in the vectorizer.
5797          + Externalize generic parts of data reference analysis to make
5798            this analysis available to other passes.
5799          + Vectorization of conditional code.
5800          + Reduction support.
5801     * GCC can now partition functions in sections of hot and cold code.
5802       This can significantly improve performance due to better
5803       instruction cache locality. This feature works best together with
5804       profile feedback driven optimization.
5805     * A new pass to avoid saving of unneeded arguments to the stack in
5806       vararg functions if the compiler can prove that they will not be
5807       needed.
5808     * Transition of basic block profiling to tree level implementation
5809       has been completed. The new implementation should be considerably
5810       more reliable (hopefully avoiding profile mismatch errors when
5811       using -fprofile-use or -fbranch-probabilities) and can be used to
5812       drive higher level optimizations, such as inlining.
5813       The -ftree-based-profiling command-line option was removed and
5814       -fprofile-use now implies disabling old RTL level loop optimizer
5815       (-fno-loop-optimize). Speculative prefetching optimization
5816       (originally enabled by -fspeculative-prefetching) was removed.
5817
5818New Languages and Language specific improvements
5819
5820  C and Objective-C
5821
5822     * The old Bison-based C and Objective-C parser has been replaced by a
5823       new, faster hand-written recursive-descent parser.
5824
5825  Ada
5826
5827     * The build infrastructure for the Ada runtime library and tools has
5828       been changed to be better integrated with the rest of the build
5829       infrastructure of GCC. This should make doing cross builds of Ada a
5830       bit easier.
5831
5832  C++
5833
5834     * ARM-style name-injection of friend declarations is no longer the
5835       default. For example:
5836          struct S {
5837            friend void f();
5838          };
5839
5840          void g() { f(); }
5841       will not be accepted; instead a declaration of f will need to be
5842       present outside of the scope of S. The new -ffriend-injection
5843       option will enable the old behavior.
5844     * The (undocumented) extension which permitted templates with default
5845       arguments to be bound to template template parameters with fewer
5846       parameters has been deprecated, and will be removed in the next
5847       major release of G++. For example:
5848       template <template <typename> class C>
5849       void f(C<double>) {}
5850
5851       template <typename T, typename U = int>
5852       struct S {};
5853
5854       template void f(S<double>);
5855
5856       makes use of the deprecated extension. The reason this code is not
5857       valid ISO C++ is that S is a template with two parameters;
5858       therefore, it cannot be bound to C which has only one parameter.
5859
5860    Runtime Library (libstdc++)
5861
5862     * Optimization work:
5863          + A new implementation of std::search_n is provided, better
5864            performing in case of random access iterators.
5865          + Added further efficient specializations of istream functions,
5866            i.e., character array and string extractors.
5867          + Other smaller improvements throughout.
5868     * Policy-based associative containers, designed for high-performance,
5869       flexibility and semantic safety are delivered in ext/pb_assoc.
5870     * A versatile string class, __gnu_cxx::__versa_string, providing
5871       facilities conforming to the standard requirements for
5872       basic_string, is delivered in <ext/vstring.h>. In particular:
5873          + Two base classes are provided: the default one avoids
5874            reference counting and is optimized for short strings; the
5875            alternate one, still uses it while improving in a few low
5876            level areas (e.g., alignment). See vstring_fwd.h for some
5877            useful typedefs.
5878          + Various algorithms have been rewritten (e.g., replace), the
5879            code streamlined and simple optimizations added.
5880          + Option 3 of DR 431 is implemented for both available bases,
5881            thus improving the support for stateful allocators.
5882     * As usual, many bugs have been fixed (e.g., libstdc++/13583,
5883       libstdc++/23953) and LWG resolutions put into effect for the first
5884       time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1
5885       Issue 6.19). The implementation status of TR1 is now tracked in the
5886       docs in tr1.html.
5887
5888  Objective-C++
5889
5890     * A new language front end for Objective-C++ has been added. This
5891       language allows users to mix the object oriented features of
5892       Objective-C with those of C++.
5893
5894  Java (GCJ)
5895
5896     * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19
5897       features (plus some 0.20 bug-fixes)
5898          + Networking
5899               o The java.net.HttpURLConnection implementation no longer
5900                 buffers the entire response body in memory. This means
5901                 that response bodies larger than available memory can now
5902                 be handled.
5903          + (N)IO
5904               o NIO FileChannel.map implementation, fast bulk put
5905                 implementation for DirectByteBuffer (speeds up this
5906                 method 10x).
5907               o FileChannel.lock() and FileChannel.force() implemented.
5908          + XML
5909               o gnu.xml fix for nodes created outside a namespace
5910                 context.
5911               o Add support for output indenting and
5912                 cdata-section-elements output instruction in
5913                 xml.transform.
5914               o xml.xpath corrections for cases where elements/attributes
5915                 might have been created in non-namespace-aware mode.
5916                 Corrections to handling of XSL variables and minor
5917                 conformance updates.
5918          + AWT
5919               o GNU JAWT implementation, the AWT Native Interface, which
5920                 allows direct access to native screen resources from
5921                 within a Canvas's paint method. GNU Classpath Examples
5922                 comes with a Demo, see libjava/classpath/examples/README.
5923               o awt.datatransfer updated to 1.5 with support for
5924                 FlavorEvents. The gtk+ awt peers now allow copy/paste of
5925                 text, images, URIs/files and serialized objects with
5926                 other applications and tracking clipboard change events
5927                 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized
5928                 objects are supported). A GNU Classpath Examples
5929                 datatransfer Demo was added to show the new
5930                 functionality.
5931               o Split gtk+ awt peers event handling in two threads and
5932                 improve gdk lock handling (solves several awt lock ups).
5933               o Speed up awt Image loading.
5934               o Better gtk+ scrollbar peer implementation when using gtk+
5935                 >= 2.6.
5936               o Handle image loading errors correctly for gdkpixbuf and
5937                 MediaTracker.
5938               o Better handle GDK lock. Properly prefix gtkpeer native
5939                 functions (cp_gtk).
5940               o GdkGraphics2D has been updated to use Cairo 0.5.x or
5941                 higher.
5942               o BufferedImage and GtkImage rewrites. All image drawing
5943                 operations should now work correctly (flipping requires
5944                 gtk+ >= 2.6)
5945               o Future Graphics2D, image and text work is documented at:
5946                 [2]http://developer.classpath.org/mediation/ClasspathGrap
5947                 hicsImagesText
5948               o When gtk+ 2.6 or higher is installed the default log
5949                 handler will produce stack traces whenever a WARNING,
5950                 CRITICAL or ERROR message is produced.
5951          + Free Swing
5952               o The RepaintManager has been reworked for more efficient
5953                 painting, especially for large GUIs.
5954               o The layout manager OverlayLayout has been implemented,
5955                 the BoxLayout has been rewritten to make use of the
5956                 SizeRequirements utility class and caching for more
5957                 efficient layout.
5958               o Improved accessibility support.
5959               o Significant progress has been made in the implementation
5960                 of the javax.swing.plaf.metal package, with most UI
5961                 delegates in a working state now. Please test this with
5962                 your own applications and provide feedback that will help
5963                 us to improve this package.
5964               o The GUI demo (gnu.classpath.examples.swing.Demo) has been
5965                 extended to highlight various features in our Free Swing
5966                 implementation. And it includes a look and feel switcher
5967                 for Metal (default), Ocean and GNU themes.
5968               o The javax.swing.plaf.multi package is now implemented.
5969               o Editing and several key actions for JTree and JTable were
5970                 implemented.
5971               o Lots of icons and look and feel improvements for Free
5972                 Swing basic and metal themes were added. Try running the
5973                 GNU Classpath Swing Demo in examples
5974                 (gnu.classpath.examples.swing.Demo) with:
5975                 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee
5976                 l or
5977                 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee
5978                 l
5979               o Start of styled text capabilites for java.swing.text.
5980               o DefaultMutableTreeNode pre-order, post-order, depth-first
5981                 and breadth-first traversal enumerations implemented.
5982               o JInternalFrame colors and titlebar draw properly.
5983               o JTree is working up to par (icons, selection and keyboard
5984                 traversal).
5985               o JMenus were made more compatible in visual and
5986                 programmatic behavior.
5987               o JTable changeSelection and multiple selections
5988                 implemented.
5989               o JButton and JToggleButton change states work properly
5990                 now.
5991               o JFileChooser fixes.
5992               o revalidate() and repaint() fixes which make Free Swing
5993                 much more responsive.
5994               o MetalIconFactory implemented.
5995               o Free Swing Top-Level Compatibility. JFrame, JDialog,
5996                 JApplet, JInternalFrame, and JWindow are now 1.5
5997                 compatible in the sense that you can call add() and
5998                 setLayout() directly on them, which will have the same
5999                 effect as calling getContentPane().add() and
6000                 getContentPane().setLayout().
6001               o The JTree interface has been completed. JTrees now
6002                 recognizes mouse clicks and selections work.
6003               o BoxLayout works properly now.
6004               o Fixed GrayFilter to actually work.
6005               o Metal SplitPane implemented.
6006               o Lots of Free Swing text and editor stuff work now.
6007          + Free RMI and Corba
6008               o Andrew Watson, Vice President and Technical Director of
6009                 the Object Management Group, has officially assigned us
6010                 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will
6011                 mark remote classpath-specific system exceptions.
6012                 Obtaining the VMCID means that GNU Classpath now is a
6013                 recogniseable type of node in a highly interoperable
6014                 CORBA world.
6015               o GNU Classpath now includes the first working draft to
6016                 support the RMI over IIOP protocol. The current
6017                 implementation is capable of remote invocations,
6018                 transferring various Serializables and Externalizables
6019                 via RMI-IIOP protocol. It can flatten graphs and, at
6020                 least for the simple cases, is interoperable with 1.5
6021                 JDKs.
6022               o org.omg.PortableInterceptor and related functionality in
6023                 other packages is now implemented:
6024                    # The sever and client interceptors work as required
6025                      since 1.4.
6026                    # The IOR interceptor works as needed for 1.5.
6027               o The org.omg.DynamicAny package is completed and passes
6028                 the prepared tests.
6029               o The Portable Object Adapter should now support the output
6030                 of the recent IDL to java compilers. These compilers now
6031                 generate servants and not CORBA objects as before, making
6032                 the output depend on the existing POA implementation.
6033                 Completing POA means that such code can already be tried
6034                 to run on Classpath. Our POA is tested for the following
6035                 usager scenarios:
6036                    # POA converts servant to the CORBA object.
6037                    # Servant provides to the CORBA object.
6038                    # POA activates new CORBA object with the given Object
6039                      Id (byte array) that is later accessible for the
6040                      servant.
6041                    # During the first call, the ServantActivator provides
6042                      servant for this and all subsequent calls on the
6043                      current object.
6044                    # During each call, the ServantLocator provides
6045                      servant for this call only.
6046                    # ServantLocator or ServantActivator forwards call to
6047                      another server.
6048                    # POA has a single servant, responsible for all
6049                      objects.
6050                    # POA has a default servant, but some objects are
6051                      explicitly connected to they specific servants.
6052                 The POA is verified using tests from the former
6053                 cost.omg.org.
6054               o The CORBA implementation is now a working prototype that
6055                 should support features up to 1.3 inclusive. We invite
6056                 groups writing CORBA dependent applications to try
6057                 Classpath implementation, reporting any possible bugs.
6058                 The CORBA prototype is interoperable with Sun's
6059                 implementation v 1.4, transferring object references,
6060                 primitive types, narrow and wide strings, arrays,
6061                 structures, trees, abstract interfaces and value types
6062                 (feature of CORBA 2.3) between these two platforms.
6063                 Remote exceptions are transferred and handled correctly.
6064                 The stringified object references (IORs) from various
6065                 sources are parsed as required. The transient (for
6066                 current session) and permanent (till jre restart)
6067                 redirections work. Both Little and Big Endian encoded
6068                 messages are accepted. The implementation is verified
6069                 using tests from the former cost.omg.org. The current
6070                 release includes working examples (see the examples
6071                 directory), demonstrating the client-server
6072                 communication, using either CORBA Request or IDL-based
6073                 stub (usually generated by a IDL to java compiler). These
6074                 examples also show how to use the Classpath CORBA naming
6075                 service. The IDL to java compiler is not yet written, but
6076                 as our library must be compatible, it naturally accepts
6077                 the output of other idlj implementations.
6078          + Misc
6079               o Updated TimeZone data against Olson tzdata2005l.
6080               o Make zip and jar packages UTF-8 clean.
6081               o "native" code builds and compiles (warning free) on
6082                 Darwin and Solaris.
6083               o java.util.logging.FileHandler now rotates files.
6084               o Start of a generic JDWP framework in gnu/classpath/jdwp.
6085                 This is unfinished, but feedback (at classpath@gnu.org)
6086                 from runtime hackers is greatly appreciated. Although
6087                 most of the work is currently being done around gcj/gij
6088                 we want this framework to be as VM neutral as possible.
6089                 Early design is described in:
6090                 [3]http://gcc.gnu.org/ml/java/2005-05/msg00260.html
6091               o QT4 AWT peers, enable by giving configure
6092                 --enable-qt-peer. Included, but not ready for production
6093                 yet. They are explicitly disabled and not supported. But
6094                 if you want to help with the development of these new
6095                 features we are interested in feedback. You will have to
6096                 explicitly enable them to try them out (and they will
6097                 most likely contain bugs).
6098               o Documentation fixes all over the place. See
6099                 [4]http://developer.classpath.org/doc/
6100
6101New Targets and Target Specific Improvements
6102
6103  IA-32/x86-64
6104
6105     * The x86-64 medium model (that allows building applications whose
6106       data segment exceeds 4GB) was redesigned to match latest ABI draft.
6107       New implementation split large datastructures into separate segment
6108       improving performance of accesses to small datastructures and also
6109       allows linking of small model libraries into medium model programs
6110       as long as the libraries are not accessing the large datastructures
6111       directly. Medium model is also supported in position independent
6112       code now.
6113       The ABI change results in partial incompatibility among medium
6114       model objects. Linking medium model libraries (or objects) compiled
6115       with new compiler into medium model program compiled with older
6116       will likely result in exceeding ranges of relocations.
6117       Binutils 2.16.91 or newer are required for compiling medium model
6118       now.
6119
6120  RS6000 (POWER/PowerPC)
6121
6122     * The AltiVec vector primitives in <altivec.h> are now implemented in
6123       a way that puts a smaller burden on the preprocessor, instead
6124       processing the "overloading" in the front ends. This should benefit
6125       compilation speed on AltiVec vector code.
6126     * AltiVec initializers now are generated more efficiently.
6127     * The popcountb instruction available on POWER5 now is generated.
6128     * The floating point round to integer instructions available on
6129       POWER5+ now is generated.
6130     * Floating point divides can be synthesized using the floating point
6131       reciprocal estimate instructions.
6132     * Double precision floating point constants are initialized as single
6133       precision values if they can be represented exactly.
6134
6135  S/390, zSeries and System z9
6136
6137     * Support for the IBM System z9 109 processor has been added. When
6138       using the -march=z9-109 option, the compiler will generate code
6139       making use of instructions provided by the extended immediate
6140       facility.
6141     * Support for 128-bit IEEE floating point has been added. When using
6142       the -mlong-double-128 option, the compiler will map the long double
6143       data type to 128-bit IEEE floating point. Using this option
6144       constitutes an ABI change, and requires glibc support.
6145     * Various changes to improve performance of generated code have been
6146       implemented, including:
6147          + In functions that do not require a literal pool, register %r13
6148            (which is traditionally reserved as literal pool pointer), can
6149            now be freely used for other purposes by the compiler.
6150          + More precise tracking of register use allows the compiler to
6151            generate more efficient function prolog and epilog code in
6152            certain cases.
6153          + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING
6154            instructions are now used to implement C string functions.
6155          + The MOVE CHARACTER instruction with single byte overlap is now
6156            used to implement the memset function with non-zero fill byte.
6157          + The LOAD ZERO instructions are now used where appropriate.
6158          + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK,
6159            and INSERT IMMEDIATE instructions are now used more frequently
6160            to optimize bitfield operations.
6161          + The BRANCH ON COUNT instruction is now used more frequently.
6162            In particular, the fact that a loop contains a subroutine call
6163            no longer prevents the compiler from using this instruction.
6164          + The compiler is now aware that all shift and rotate
6165            instructions implicitly truncate the shift count to six bits.
6166     * Back-end support for the following generic features has been
6167       implemented:
6168          + The full set of [5]built-in functions for atomic memory
6169            access.
6170          + The -fstack-protector feature.
6171          + The optimization pass avoiding unnecessary stores of incoming
6172            argument registers in functions with variable argument list.
6173
6174  SPARC
6175
6176     * The default code model in 64-bit mode has been changed from
6177       Medium/Anywhere to Medium/Middle on Solaris.
6178     * TLS support is disabled by default on Solaris prior to release 10.
6179       It can be enabled on TLS-capable Solaris 9 versions (4/04 release
6180       and later) by specifying --enable-tls at configure time.
6181
6182  MorphoSys
6183
6184     * Support has been added for this new architecture.
6185
6186Obsolete Systems
6187
6188Documentation improvements
6189
6190Other significant improvements
6191
6192     * GCC can now emit code for protecting applications from
6193       stack-smashing attacks. The protection is realized by buffer
6194       overflow detection and reordering of stack variables to avoid
6195       pointer corruption.
6196     * Some built-in functions have been fortified to protect them against
6197       various buffer overflow (and format string) vulnerabilities.
6198       Compared to the mudflap bounds checking feature, the safe builtins
6199       have far smaller overhead. This means that programs built using
6200       safe builtins should not experience any measurable slowdown.
6201
6202GCC 4.1.2
6203
6204   This is the [6]list of problem reports (PRs) from GCC's bug tracking
6205   system that are known to be fixed in the 4.1.2 release. This list might
6206   not be complete (that is, it is possible that some PRs that have been
6207   fixed are not listed here).
6208
6209   When generating code for a shared library, GCC now recognizes that
6210   global functions may be replaced when the program runs. Therefore, it
6211   is now more conservative in deducing information from the bodies of
6212   functions. For example, in this example:
6213    void f() {}
6214    void g() {
6215     try { f(); }
6216     catch (...) {
6217       cout << "Exception";
6218     }
6219    }
6220
6221   G++ would previously have optimized away the catch clause, since it
6222   would have concluded that f cannot throw exceptions. Because users may
6223   replace f with another function in the main body of the program, this
6224   optimization is unsafe, and is no longer performed. If you wish G++ to
6225   continue to optimize as before, you must add a throw() clause to the
6226   declaration of f to make clear that it does not throw exceptions.
6227
6228
6229    For questions related to the use of GCC, please consult these web
6230    pages and the [7]GCC manuals. If that fails, the
6231    [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6232    web pages and the development of GCC are welcome on our developer
6233    list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
6234    archives.
6235
6236   Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
6237   distribution of this entire article is permitted in any medium,
6238   provided this notice is preserved.
6239
6240   These pages are [12]maintained by the GCC team. Last modified
6241   2012-11-02[13].
6242
6243References
6244
6245   1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
6246   2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText
6247   3. http://gcc.gnu.org/ml/java/2005-05/msg00260.html
6248   4. http://developer.classpath.org/doc/
6249   5. http://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
6250   6. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2
6251   7. http://gcc.gnu.org/onlinedocs/
6252   8. mailto:gcc-help@gcc.gnu.org
6253   9. mailto:gcc@gcc.gnu.org
6254  10. http://gcc.gnu.org/lists.html
6255  11. http://www.fsf.org/
6256  12. http://gcc.gnu.org/about.html
6257  13. http://validator.w3.org/check/referer
6258======================================================================
6259http://gcc.gnu.org/gcc-4.0/index.html
6260
6261                             GCC 4.0 Release Series
6262
6263   January 31, 2007
6264
6265   The [1]GNU project and the GCC developers are pleased to announce the
6266   release of GCC 4.0.4.
6267
6268   This release is a bug-fix release, containing fixes for regressions in
6269   GCC 4.0.3 relative to previous releases of GCC.
6270
6271Release History
6272
6273   GCC 4.0.4
6274          January 31, 2007 ([2]changes)
6275
6276   GCC 4.0.3
6277          March 10, 2006 ([3]changes)
6278
6279   GCC 4.0.2
6280          September 28, 2005 ([4]changes)
6281
6282   GCC 4.0.1
6283          July 7, 2005 ([5]changes)
6284
6285   GCC 4.0.0
6286          April 20, 2005 ([6]changes)
6287
6288References and Acknowledgements
6289
6290   GCC used to stand for the GNU C Compiler, but since the compiler
6291   supports several other languages aside from C, it now stands for the
6292   GNU Compiler Collection.
6293
6294   A list of [7]successful builds is updated as new information becomes
6295   available.
6296
6297   The GCC developers would like to thank the numerous people that have
6298   contributed new features, improvements, bug fixes, and other changes as
6299   well as test results to GCC. This [8]amazing group of volunteers is
6300   what makes GCC successful.
6301
6302   For additional information about GCC please refer to the [9]GCC project
6303   web site or contact the [10]GCC development mailing list.
6304
6305   To obtain GCC please use [11]our mirror sites, or [12]our SVN server.
6306
6307
6308    For questions related to the use of GCC, please consult these web
6309    pages and the [13]GCC manuals. If that fails, the
6310    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6311    web pages and the development of GCC are welcome on our developer
6312    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
6313    archives.
6314
6315   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
6316   distribution of this entire article is permitted in any medium,
6317   provided this notice is preserved.
6318
6319   These pages are [18]maintained by the GCC team. Last modified
6320   2012-11-02[19].
6321
6322References
6323
6324   1. http://www.gnu.org/
6325   2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
6326   3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3
6327   4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2
6328   5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1
6329   6. http://gcc.gnu.org/gcc-4.0/changes.html
6330   7. http://gcc.gnu.org/gcc-4.0/buildstat.html
6331   8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6332   9. http://gcc.gnu.org/index.html
6333  10. mailto:gcc@gcc.gnu.org
6334  11. http://gcc.gnu.org/mirrors.html
6335  12. http://gcc.gnu.org/svn.html
6336  13. http://gcc.gnu.org/onlinedocs/
6337  14. mailto:gcc-help@gcc.gnu.org
6338  15. mailto:gcc@gcc.gnu.org
6339  16. http://gcc.gnu.org/lists.html
6340  17. http://www.fsf.org/
6341  18. http://gcc.gnu.org/about.html
6342  19. http://validator.w3.org/check/referer
6343======================================================================
6344http://gcc.gnu.org/gcc-4.0/changes.html
6345
6346                             GCC 4.0 Release Series
6347                        Changes, New Features, and Fixes
6348
6349   The latest release in the 4.0 release series is [1]GCC 4.0.4.
6350
6351Caveats
6352
6353     * GCC now generates location lists by default when compiling with
6354       debug info and optimization.
6355          + GDB 6.0 and older crashes when it sees location lists. GDB 6.1
6356            or later is needed to debug binaries containing location
6357            lists.
6358          + When you are trying to view a value of a variable in a part of
6359            a function where it has no location (for example when the
6360            variable is no longer used and thus its location was used for
6361            something else) GDB will say that it is not available.
6362       You can disable generating location lists by -fno-var-tracking.
6363     * GCC no longer accepts the -fwritable-strings option. Use named
6364       character arrays when you need a writable string.
6365     * The options -freduce-all-givs and -fmove-all-movables have been
6366       discontinued. They were used to circumvent a shortcoming in the
6367       heuristics of the old loop optimization code with respect to common
6368       Fortran constructs. The new (tree) loop optimizer works differently
6369       and doesn't need those work-arounds.
6370     * The graph-coloring register allocator, formerly enabled by the
6371       option -fnew-ra, has been discontinued.
6372     * -I- has been deprecated. -iquote is meant to replace the need for
6373       this option.
6374     * The MIPS -membedded-pic and -mrnames options have been removed.
6375     * All MIPS targets now require the GNU assembler. In particular, IRIX
6376       configurations can no longer use the MIPSpro assemblers, although
6377       they do still support the MIPSpro linkers.
6378     * The SPARC option -mflat has been removed.
6379     * English-language diagnostic messages will now use Unicode quotation
6380       marks in UTF-8 locales. (Non-English messages already used the
6381       quotes appropriate for the language in previous releases.) If your
6382       terminal does not support UTF-8 but you are using a UTF-8 locale
6383       (such locales are the default on many GNU/Linux systems) then you
6384       should set LC_CTYPE=C in the environment to disable that locale.
6385       Programs that parse diagnostics and expect plain ASCII
6386       English-language messages should set LC_ALL=C. See [2]Markus Kuhn's
6387       explanation of Unicode quotation marks for more information.
6388     * The specs file is no longer installed on most platforms. Most users
6389       will be totally unaffected. However, if you are accustomed to
6390       editing the specs file yourself, you will now have to use the
6391       -dumpspecs option to generate the specs file, and then edit the
6392       resulting file.
6393
6394General Optimizer Improvements
6395
6396     * The [3]tree ssa branch has been merged. This merge has brought in a
6397       completely new optimization framework based on a higher level
6398       intermediate representation than the existing RTL representation.
6399       Numerous new code transformations based on the new framework are
6400       available in GCC 4.0, including:
6401          + Scalar replacement of aggregates
6402          + Constant propagation
6403          + Value range propagation
6404          + Partial redundancy elimination
6405          + Load and store motion
6406          + Strength reduction
6407          + Dead store elimination
6408          + Dead and unreachable code elimination
6409          + [4]Autovectorization
6410          + Loop interchange
6411          + Tail recursion by accumulation
6412       Many of these passes outperform their counterparts from previous
6413       GCC releases.
6414     * [5]Swing Modulo Scheduling (SMS). An RTL level instruction
6415       scheduling optimization intended for loops that perform heavy
6416       computations.
6417
6418New Languages and Language specific improvements
6419
6420  C family
6421
6422     * The sentinel attribute has been added to GCC. This function
6423       attribute allows GCC to warn when variadic functions such as execl
6424       are not NULL terminated. See the GCC manual for a complete
6425       description of its behavior.
6426     * Given __attribute__((alias("target"))) it is now an error if target
6427       is not a symbol, defined in the same translation unit. This also
6428       applies to aliases created by #pragma weak alias=target. This is
6429       because it's meaningless to define an alias to an undefined symbol.
6430       On Solaris, the native assembler would have caught this error, but
6431       GNU as does not.
6432
6433  C and Objective-C
6434
6435     * The -Wstrict-aliasing=2 option has been added. This warning catches
6436       all unsafe cases, but it may also give a warning for some cases
6437       that are safe.
6438     * The cast-as-lvalue, conditional-expression-as-lvalue and
6439       compound-expression-as-lvalue extensions, which were deprecated in
6440       3.3.4 and 3.4, have been removed.
6441     * The -fwritable-strings option, which was deprecated in 3.4, has
6442       been removed.
6443     * #pragma pack() semantics have been brought closer to those used by
6444       other compilers. This also applies to C++.
6445     * Taking the address of a variable with register storage is invalid
6446       in C. GCC now issues an error instead of a warning.
6447     * Arrays of incomplete element type are invalid in C. GCC now issues
6448       an error for such arrays. Declarations such as extern struct s x[];
6449       (where struct s has not been defined) can be moved after the
6450       definition of struct s. Function parameters declared as arrays of
6451       incomplete type can instead be declared as pointers.
6452
6453  C++
6454
6455     * When compiling without optimizations (-O0), the C++ frontend is
6456       much faster than in any previous versions of GCC. Independent
6457       testers have measured speed-ups up to 25% in real-world production
6458       code, compared to the 3.4 family (which was already the fastest
6459       version to date). Upgrading from older versions might show even
6460       bigger improvements.
6461     * ELF visibility attributes can now be applied to a class type, so
6462       that it affects every member function of a class at once, without
6463       having to specify each individually:
6464class __attribute__ ((visibility("hidden"))) Foo
6465{
6466   int foo1();
6467   void foo2();
6468};
6469       The syntax is deliberately similar to the __declspec() system used
6470       by Microsoft Windows based compilers, allowing cross-platform
6471       projects to easily reuse their existing macro system for denoting
6472       exports and imports. By explicitly marking internal classes never
6473       used outside a binary as hidden, one can completely avoid PLT
6474       indirection overheads during their usage by the compiler. You can
6475       find out more about the advantages of this at
6476       [6]http://www.akkadia.org/drepper/dsohowto.pdf
6477     * The -fvisibility-inlines-hidden option has been added which marks
6478       all inlineable functions as having hidden ELF visibility, thus
6479       removing their symbol and typeinfo from the exported symbol table
6480       of the output ELF binary. Using this option can reduce the exported
6481       symbol count of template-heavy code by up to 40% with no code
6482       change at all, thus notably improving link and load times for the
6483       binary as well as a reduction in size of up to 10%. Also, check the
6484       new [7]-fvisibility option.
6485     * The compiler now uses the library interface specified by the [8]C++
6486       ABI for thread-safe initialization of function-scope static
6487       variables. Most users should leave this alone, but embedded
6488       programmers may want to disable this by specifying
6489       -fno-threadsafe-statics for a small savings in code size.
6490     * Taking the address of an explicit register variable is no longer
6491       supported. Note that C++ allows taking the address of variables
6492       with register storage so this will continue to compile with a
6493       warning. For example, assuming that r0 is a machine register:
6494register int foo asm ("r0");
6495register int bar;
6496&foo; // error, no longer accepted
6497&bar; // OK, with a warning
6498     * G++ has an undocumented extension to virtual function covariancy
6499       rules that allowed the overrider to return a type that was
6500       implicitly convertable to the overridden function's return type.
6501       For instance a function returning void * could be overridden by a
6502       function returning T *. This is now deprecated and will be removed
6503       in a future release.
6504     * The G++ minimum and maximum operators (<? and >?) and their
6505       compound forms (<?=) and >?=) have been deprecated and will be
6506       removed in a future version. Code using these operators should be
6507       modified to use std::min and std::max instead.
6508     * Declaration of nested classes of class templates as friends are
6509       supported:
6510template <typename T> struct A {
6511  class B {};
6512};
6513class C {
6514  template <typename T> friend class A<T>::B;
6515};
6516       This complements the feature member functions of class templates as
6517       friends introduced in GCC 3.4.0.
6518     * When declaring a friend class using an unqualified name, classes
6519       outside the innermost non-class scope are not searched:
6520class A;
6521namespace N {
6522  class B {
6523    friend class A;   // Refer to N::A which has not been declared yet
6524                      // because name outside namespace N are not searched
6525    friend class ::A; // Refer to ::A
6526  };
6527}
6528       Hiding the friend name until declaration is still not implemented.
6529     * Friends of classes defined outside their namespace are correctly
6530       handled:
6531namespace N {
6532  class A;
6533}
6534class N::A {
6535  friend class B; // Refer to N::B in GCC 4.0.0
6536                  // but ::B in earlier versions of GCC
6537};
6538
6539    Runtime Library (libstdc++)
6540
6541     * Optimization work:
6542          + Added efficient specializations of istream functions for char
6543            and wchar_t.
6544          + Further performance tuning of strings, in particular wrt
6545            single-char append and getline.
6546          + iter_swap - and therefore most of the mutating algorithms -
6547            now makes an unqualified call to swap when the value_type of
6548            the two iterators is the same.
6549     * A large subset of the features in Technical Report 1 (TR1 for
6550       short) is experimentally delivered (i.e., no guarantees about the
6551       implementation are provided. In particular it is not promised that
6552       the library will remain link-compatible when code using TR1 is
6553       used):
6554          + General utilities such as reference_wrapper and shared_ptr.
6555          + Function objects, i.e., result_of, mem_fn, bind, function.
6556          + Support for metaprogramming.
6557          + New containers such as tuple, array, unordered_set,
6558            unordered_map, unordered_multiset, unordered_multimap.
6559     * As usual, many bugs have been fixed and LWG resolutions implemented
6560       for the first time (e.g., DR 409).
6561
6562  Java
6563
6564     * In order to prevent naming conflicts with other implementations of
6565       these tools, some GCJ binaries have been renamed:
6566          + rmic is now grmic,
6567          + rmiregistry is now grmiregistry, and
6568          + jar is now fastjar.
6569       In particular, these names were problematic for the jpackage.org
6570       packaging conventions which install symlinks in /usr/bin that point
6571       to the preferred versions of these tools.
6572     * The -findirect-dispatch argument to the compiler now works and
6573       generates code following a new "binary compatibility" ABI. Code
6574       compiled this way follows the binary compatibility rules of the
6575       Java Language Specification.
6576     * libgcj now has support for using GCJ as a JIT, using the
6577       gnu.gcj.jit family of system properties.
6578     * libgcj can now find a shared library corresponding to the bytecode
6579       representation of a class. See the documentation for the new
6580       gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system
6581       property.
6582     * There have been many improvements to the class library. Here are
6583       some highlights:
6584          + Much more of AWT and Swing exist.
6585          + Many new packages and classes were added, including
6586            java.util.regex, java.net.URI, javax.crypto,
6587            javax.crypto.interfaces, javax.crypto.spec, javax.net,
6588            javax.net.ssl, javax.security.auth,
6589            javax.security.auth.callback, javax.security.auth.login,
6590            javax.security.auth.x500, javax.security.sasl, org.ietf.jgss,
6591            javax.imageio, javax.imageio.event, javax.imageio.spi,
6592            javax.print, javax.print.attribute,
6593            javax.print.attribute.standard, javax.print.event, and
6594            javax.xml
6595          + Updated SAX and DOM, and imported GNU JAXP
6596
6597  Fortran
6598
6599     * A new [9]Fortran front end has replaced the aging GNU Fortran 77
6600       front end. The new front end supports Fortran 90 and Fortran 95. It
6601       may not yet be as stable as the old Fortran front end.
6602
6603  Ada
6604
6605     * Ada (with tasking and Zero Cost Exceptions) is now available on
6606       many more targets, including but not limited to: alpha-linux,
6607       hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux,
6608       s390x-linux, sparc-linux.
6609     * Some of the new Ada 2005 features are now implemented like
6610       Wide_Wide_Character and Ada.Containers.
6611     * Many bugs have been fixed, tools and documentation improved.
6612     * To compile Ada from the sources, install an older working Ada
6613       compiler and then use --enable-languages=ada at configuration time,
6614       since the Ada frontend is not currently activated by default. See
6615       the [10]Installing GCC for details.
6616
6617New Targets and Target Specific Improvements
6618
6619  H8/300
6620
6621     * The frame layout has changed. In the new layout, the prologue of a
6622       function first saves registers and then allocate space for locals,
6623       resulting in an 1% improvement on code size.
6624
6625  IA-32/x86-64 (AMD64)
6626
6627     * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10,
6628       log1p, log2, logb and tan mathematical builtins (and their float
6629       and long double variants) are now implemented as inline x87
6630       intrinsics when using -ffast-math.
6631     * The ceil, floor, nearbyint, rint and trunc mathematical builtins
6632       (and their float and long double variants) are now implemented as
6633       inline x87 intrinsics when using -ffast-math.
6634     * The x87's fsincos instruction is now used automatically with
6635       -ffast-math when calculating both the sin and cos of the same
6636       argument.
6637     * Instruction selection for multiplication and division by constants
6638       has been improved.
6639
6640  IA-64
6641
6642     * Floating point division, integer division and sqrt are now inlined,
6643       resulting in significant performance improvements on some codes.
6644
6645  MIPS
6646
6647     * Division by zero checks now use conditional traps if the target
6648       processor supports them. This decreases code size by one word per
6649       division operation. The old behavior (branch and break) can be
6650       obtained either at configure time by passing --with-divide=breaks
6651       to configure or at runtime by passing -mdivide-breaks to GCC.
6652     * Support for MIPS64 paired-single instructions has been added. It is
6653       enabled by -mpaired-single and can be accessed using both the
6654       target-independent vector extensions and new MIPS-specific built-in
6655       functions.
6656     * Support for the MIPS-3D ASE has been added. It is enabled by
6657       -mips3d and provides new MIPS-3D-specific built-in functions.
6658     * The -mexplicit-relocs option now supports static n64 code (as is
6659       used, for example, in 64-bit linux kernels). -mexplicit-relocs
6660       should now be feature-complete and is enabled by default when GCC
6661       is configured to use a compatible assembler.
6662     * Support for the NEC VR4130 series has been added. This support
6663       includes the use of VR-specific instructions and a new VR4130
6664       scheduler. Full VR4130 support can be selected with -march=vr4130
6665       while code for any ISA can be tuned for the VR4130 using
6666       -mtune=vr4130. There is also a new -mvr4130-align option that
6667       produces better schedules at the cost of increased code size.
6668     * Support for the Broadcom SB-1 has been extended. There is now an
6669       SB-1 scheduler as well as support for the SB-1-specific
6670       paired-single instructions. Full SB-1 support can be selected with
6671       -march=sb1 while code for any ISA can be optimized for the SB-1
6672       using -mtune=sb1.
6673     * The compiler can now work around errata in R4000, R4400, VR4120 and
6674       VR4130 processors. These workarounds are enabled by -mfix-r4000,
6675       -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120
6676       and VR4130 workarounds need binutils 2.16 or above.
6677     * IRIX shared libraries are now installed into the standard library
6678       directories: o32 libraries go into lib/, n32 libraries go into
6679       lib32/ and n64 libraries go into lib64/.
6680     * The compiler supports a new -msym32 option. It can be used to
6681       optimize n64 code in which all symbols are known to have 32-bit
6682       values.
6683
6684  S/390 and zSeries
6685
6686     * New command-line options help to generate code intended to run in
6687       an environment where stack space is restricted, e.g. Linux kernel
6688       code:
6689          + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time
6690            warnings for single functions that require large or dynamic
6691            stack frames.
6692          + -mstack-size and -mstack-guard generate code that checks for
6693            stack overflow at run time.
6694          + -mpacked-stack generates code that reduces the stack frame
6695            size of many functions by reusing unneeded parts of the stack
6696            bias area.
6697     * The -msoft-float option now ensures that generated code never
6698       accesses floating point registers.
6699     * The s390x-ibm-tpf target now fully supports C++, including
6700       exceptions and threads.
6701     * Various changes to improve performance of the generated code have
6702       been implemented, including:
6703          + GCC now uses sibling calls where possible.
6704          + Condition code handling has been optimized, allowing GCC to
6705            omit redundant comparisons in certain cases.
6706          + The cost function guiding many optimizations has been refined
6707            to more accurately represent the z900 and z990 processors.
6708          + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW
6709            instructions are now used to avoid conditional branches in
6710            certain cases.
6711          + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to
6712            optimize address arithmetic required to access large stack
6713            frames.
6714          + GCC now makes more efficient use of memory-to-memory type
6715            instructions (MVC, CLC, ...).
6716          + More precise tracking of special register use allows better
6717            instruction scheduling, in particular of the function prologue
6718            and epilogue sequences.
6719          + The Java front end now generates inline code to implement
6720            integer division, instead of calling library routines.
6721
6722  SPARC
6723
6724     * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and
6725       -mf934 have been removed. They have been replaced with -mcpu=xxx.
6726     * The internal model used to estimate the relative cost of each
6727       instruction has been updated. It is expected to give better results
6728       on recent UltraSPARC processors.
6729     * Code generation for function prologues and epilogues has been
6730       improved, resulting in better scheduling and allowing multiple exit
6731       points in functions.
6732     * Support for Sun's Visual Instruction Set (VIS) has been enhanced.
6733       It is enabled by -mvis and provides new built-in functions for VIS
6734       instructions on UltraSPARC processors.
6735     * The option -mapp-regs has been turned on by default on Solaris too.
6736
6737  NetWare
6738
6739     * Novell NetWare (on ix86, no other hardware platform was ever really
6740       supported by this OS) has been re-enabled and the ABI supported by
6741       GCC has been brought into sync with that of MetroWerks CodeWarrior
6742       (the ABI previously supported was that of some Unix systems, which
6743       NetWare never tried to support).
6744
6745Obsolete Systems
6746
6747   Support for a number of older systems has been declared obsolete in GCC
6748   4.0. Unless there is activity to revive them, the next release of GCC
6749   will have their sources permanently removed.
6750
6751   All GCC ports for the following processor architectures have been
6752   declared obsolete:
6753     * Intel i860
6754     * Ubicom IP2022
6755     * National Semiconductor NS32K
6756     * Texas Instruments TMS320C[34]x
6757
6758   Also, those for some individual systems have been obsoleted:
6759     * SPARC family
6760          + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf,
6761            sparc86x-*-elf)
6762          + OpenBSD 32-bit (sparc-*-openbsd*)
6763
6764Documentation improvements
6765
6766Other significant improvements
6767
6768     * Location lists are now generated by default when compiling with
6769       debug info and optimization. Location lists provide more accurate
6770       debug info about locations of variables and they allow debugging
6771       code compiled with -fomit-frame-pointer.
6772     * The -fvisibility option has been added which allows the default ELF
6773       visibility of all symbols to be set per compilation and the new
6774       #pragma GCC visibility preprocessor command allows the setting of
6775       default ELF visibility for a region of code. Using
6776       -fvisibility=hidden especially in combination with the new
6777       -fvisibility-inlines-hidden can yield substantial improvements in
6778       output binary quality including avoiding PLT indirection overheads,
6779       reduction of the exported symbol count by up to 60% (with resultant
6780       improvements to link and load times), better scope for the
6781       optimizer to improve code and up to a 20% reduction in binary size.
6782       Using these options correctly yields a binary with a similar symbol
6783       count to a Windows DLL.
6784       Perhaps more importantly, this new feature finally allows (with
6785       careful planning) complete avoidance of symbol clashes when
6786       manually loading shared objects with RTLD_GLOBAL, thus finally
6787       solving problems many projects such as python were forced to use
6788       RTLD_LOCAL for (with its resulting issues for C++ correctness). You
6789       can find more information about using these options at
6790       [11]http://gcc.gnu.org/wiki/Visibility.
6791     __________________________________________________________________
6792
6793GCC 4.0.1
6794
6795   This is the [12]list of problem reports (PRs) from GCC's bug tracking
6796   system that are known to be fixed in the 4.0.1 release. This list might
6797   not be complete (that is, it is possible that some PRs that have been
6798   fixed are not listed here).
6799
6800GCC 4.0.2
6801
6802   This is the [13]list of problem reports (PRs) from GCC's bug tracking
6803   system that are known to be fixed in the 4.0.2 release. This list might
6804   not be complete (that is, it is possible that some PRs that have been
6805   fixed are not listed here).
6806
6807   Unfortunately, due to a release engineering failure, this release has a
6808   regression on Solaris that will affect some C++ programs. We suggest
6809   that Solaris users apply a [14]patch that corrects the problem. Users
6810   who do not wish to apply the patch should explicitly link C++ programs
6811   with the -pthreads option, even if they do not use threads. This
6812   problem has been corrected in the current 4.0 branch sources and will
6813   not be present in GCC 4.0.3.
6814
6815GCC 4.0.3
6816
6817   Starting with this release, the function getcontext is recognized by
6818   the compiler as having the same semantics as the setjmp function. In
6819   particular, the compiler will ensure that all registers are dead before
6820   calling such a function and will emit a warning about the variables
6821   that may be clobbered after the second return from the function.
6822
6823GCC 4.0.4
6824
6825   This is the [15]list of problem reports (PRs) from GCC's bug tracking
6826   system that are known to be fixed in the 4.0.4 release. This list might
6827   not be complete (that is, it is possible that some PRs that have been
6828   fixed are not listed here).
6829
6830   The 4.0.4 release is provided for those that require a high degree of
6831   binary compatibility with previous 4.0.x releases. For most users, the
6832   GCC team recommends that version 4.1.1 or later be used instead."
6833
6834
6835    For questions related to the use of GCC, please consult these web
6836    pages and the [16]GCC manuals. If that fails, the
6837    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6838    web pages and the development of GCC are welcome on our developer
6839    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
6840    archives.
6841
6842   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
6843   distribution of this entire article is permitted in any medium,
6844   provided this notice is preserved.
6845
6846   These pages are [21]maintained by the GCC team. Last modified
6847   2012-11-02[22].
6848
6849References
6850
6851   1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
6852   2. http://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
6853   3. http://gcc.gnu.org/projects/tree-ssa/
6854   4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html
6855   5. http://gcc.gnu.org/news/sms.html
6856   6. http://www.akkadia.org/drepper/dsohowto.pdf
6857   7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility
6858   8. http://mentorembedded.github.com/cxx-abi/
6859   9. http://gcc.gnu.org/fortran/
6860  10. http://gcc.gnu.org/install/
6861  11. http://gcc.gnu.org/wiki/Visibility
6862  12. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1
6863  13. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2
6864  14. http://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html
6865  15. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4
6866  16. http://gcc.gnu.org/onlinedocs/
6867  17. mailto:gcc-help@gcc.gnu.org
6868  18. mailto:gcc@gcc.gnu.org
6869  19. http://gcc.gnu.org/lists.html
6870  20. http://www.fsf.org/
6871  21. http://gcc.gnu.org/about.html
6872  22. http://validator.w3.org/check/referer
6873======================================================================
6874http://gcc.gnu.org/gcc-3.4/index.html
6875
6876                             GCC 3.4 Release Series
6877
6878   May 26, 2006
6879
6880   The [1]GNU project and the GCC developers are pleased to announce the
6881   release of GCC 3.4.6.
6882
6883   This release is a bug-fix release, containing fixes for regressions in
6884   GCC 3.4.4 relative to previous releases of GCC. This is the last of the
6885   3.4.x series.
6886
6887   The GCC 3.4 release series includes numerous [2]new features,
6888   improvements, bug fixes, and other changes, thanks to an [3]amazing
6889   group of volunteers.
6890
6891Release History
6892
6893   GCC 3.4.6
6894          March 6, 2006 ([4]changes)
6895
6896   GCC 3.4.5
6897          November 30, 2005 ([5]changes)
6898
6899   GCC 3.4.4
6900          May 18, 2005 ([6]changes)
6901
6902   GCC 3.4.3
6903          November 4, 2004 ([7]changes)
6904
6905   GCC 3.4.2
6906          September 6, 2004 ([8]changes)
6907
6908   GCC 3.4.1
6909          July 1, 2004 ([9]changes)
6910
6911   GCC 3.4.0
6912          April 18, 2004 ([10]changes)
6913
6914References and Acknowledgements
6915
6916   GCC used to stand for the GNU C Compiler, but since the compiler
6917   supports several other languages aside from C, it now stands for the
6918   GNU Compiler Collection.
6919
6920   A list of [11]successful builds is updated as new information becomes
6921   available.
6922
6923   The GCC developers would like to thank the numerous people that have
6924   contributed new features, improvements, bug fixes, and other changes as
6925   well as test results to GCC. This [12]amazing group of volunteers is
6926   what makes GCC successful.
6927
6928   For additional information about GCC please refer to the [13]GCC
6929   project web site or contact the [14]GCC development mailing list.
6930
6931   To obtain GCC please use [15]our mirror sites, or [16]our SVN server.
6932
6933
6934    For questions related to the use of GCC, please consult these web
6935    pages and the [17]GCC manuals. If that fails, the
6936    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6937    web pages and the development of GCC are welcome on our developer
6938    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
6939    archives.
6940
6941   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
6942   distribution of this entire article is permitted in any medium,
6943   provided this notice is preserved.
6944
6945   These pages are [22]maintained by the GCC team. Last modified
6946   2012-11-02[23].
6947
6948References
6949
6950   1. http://www.gnu.org/
6951   2. http://gcc.gnu.org/gcc-3.4/changes.html
6952   3. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6953   4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
6954   5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5
6955   6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4
6956   7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3
6957   8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2
6958   9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1
6959  10. http://gcc.gnu.org/gcc-3.4/changes.html
6960  11. http://gcc.gnu.org/gcc-3.4/buildstat.html
6961  12. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6962  13. http://gcc.gnu.org/index.html
6963  14. mailto:gcc@gcc.gnu.org
6964  15. http://gcc.gnu.org/mirrors.html
6965  16. http://gcc.gnu.org/svn.html
6966  17. http://gcc.gnu.org/onlinedocs/
6967  18. mailto:gcc-help@gcc.gnu.org
6968  19. mailto:gcc@gcc.gnu.org
6969  20. http://gcc.gnu.org/lists.html
6970  21. http://www.fsf.org/
6971  22. http://gcc.gnu.org/about.html
6972  23. http://validator.w3.org/check/referer
6973======================================================================
6974http://gcc.gnu.org/gcc-3.4/changes.html
6975
6976                             GCC 3.4 Release Series
6977                        Changes, New Features, and Fixes
6978
6979   The final release in the 3.4 release series is [1]GCC 3.4.6. The series
6980   is now closed.
6981
6982   GCC 3.4 has [2]many improvements in the C++ frontend. Before reporting
6983   a bug, please make sure it's really GCC, and not your code, that is
6984   broken.
6985
6986Caveats
6987
6988     * GNU Make is now required to build GCC.
6989     * With -nostdinc the preprocessor used to ignore both standard
6990       include paths and include paths contained in environment variables.
6991       It was neither documented nor intended that environment variable
6992       paths be ignored, so this has been corrected.
6993     * GCC no longer accepts the options -fvolatile, -fvolatile-global and
6994       -fvolatile-static. It is unlikely that they worked correctly in any
6995       3.x release.
6996     * GCC no longer ships <varargs.h>. Use <stdarg.h> instead.
6997     * Support for all the systems [3]obsoleted in GCC 3.3 has been
6998       removed from GCC 3.4. See below for a [4]list of systems which are
6999       obsoleted in this release.
7000     * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C
7001       compilers will not work.
7002     * The implementation of the [5]MIPS ABIs has changed. As a result,
7003       the code generated for certain MIPS targets will not be binary
7004       compatible with earlier releases.
7005     * In previous releases, the MIPS port had a fake "hilo" register with
7006       the user-visible name accum. This register has been removed.
7007     * The implementation of the [6]SPARC ABIs has changed. As a result,
7008       the code generated will not be binary compatible with earlier
7009       releases in certain cases.
7010     * The configure option --enable-threads=pthreads has been removed;
7011       use --enable-threads=posix instead, which should have the same
7012       effect.
7013     * Code size estimates used by inlining heuristics for C, Objective-C,
7014       C++ and Java have been redesigned significantly. As a result the
7015       parameters of -finline-insns, --param max-inline-insns-single and
7016       --param max-inline-insns-auto need to be reconsidered.
7017     * --param max-inline-slope and --param min-inline-insns have been
7018       removed; they are not needed for the new bottom-up inlining
7019       heuristics.
7020     * The new unit-at-a-time compilation scheme has several compatibility
7021       issues:
7022          + The order in which functions, variables, and top-level asm
7023            statements are emitted may have changed. Code relying on some
7024            particular ordering needs to be updated. The majority of such
7025            top-level asm statements can be replaced by section
7026            attributes.
7027          + Unreferenced static variables and functions are removed. This
7028            may result in undefined references when an asm statement
7029            refers to the variable/function directly. In that case either
7030            the variable/function shall be listed in asm statement operand
7031            or in the case of top-level asm statements the attribute used
7032            shall be used to force function/variable to be always output
7033            and considered as a possibly used by unknown code.
7034            For variables the attribute is accepted only by GCC 3.4 and
7035            newer, while for earlier versions it is sufficient to use
7036            unused to silence warnings about the variables not being
7037            referenced. To keep code portable across different GCC
7038            versions, you can use appropriate preprocessor conditionals.
7039          + Static functions now can use non-standard passing conventions
7040            that may break asm statements calling functions directly.
7041            Again the attribute used shall be used to prevent this
7042            behavior.
7043       As a temporary workaround, -fno-unit-at-a-time can be used, but
7044       this scheme may not be supported by future releases of GCC.
7045     * GCC 3.4 automatically places zero-initialized variables in the .bss
7046       section on some operating systems. Versions of GNU Emacs up to (and
7047       including) 21.3 will not work correctly when using this
7048       optimization; you can use -fno-zero-initialized-in-bss to disable
7049       it.
7050     * If GCC 3.4 is configured with --enable-threads=posix (the default
7051       on most targets that support pthreads) then _REENTRANT will be
7052       defined unconditionally by some libstdc++ headers. C++ code which
7053       relies on that macro to detect whether multi-threaded code is being
7054       compiled might change in meaning, possibly resulting in linker
7055       errors for single-threaded programs. Affected users of [7]Boost
7056       should compile single-threaded code with -DBOOST_DISABLE_THREADS.
7057       See Bugzilla for [8]more information.
7058
7059General Optimizer Improvements
7060
7061     * Usability of the profile feedback and coverage testing has been
7062       improved.
7063          + Performance of profiled programs has been improved by faster
7064            profile merging code.
7065          + Better use of the profile feedback for optimization (loop
7066            unrolling and loop peeling).
7067          + File locking support allowing fork() calls and parallel runs
7068            of profiled programs.
7069          + Coverage file format has been redesigned.
7070          + gcov coverage tool has been improved.
7071          + make profiledbootstrap available to build a faster compiler.
7072            Experiments made on i386 hardware showed an 11% speedup on -O0
7073            and a 7.5% speedup on -O2 compilation of a [9]large C++
7074            testcase.
7075          + New value profiling pass enabled via -fprofile-values
7076          + New value profile transformations pass enabled via -fvpt aims
7077            to optimize some code sequences by exploiting knowledge about
7078            value ranges or other properties of the operands. At the
7079            moment a conversion of expensive divisions into cheaper
7080            operations has been implemented.
7081          + New -fprofile-generate and -fprofile-use command-line options
7082            to simplify the use of profile feedback.
7083     * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and
7084       Java which is enabled via -funit-at-a-time (and implied by -O2). In
7085       this scheme a whole file is parsed first and optimized later. The
7086       following basic inter-procedural optimizations are implemented:
7087          + Removal of unreachable functions and variables
7088          + Discovery of local functions (functions with static linkage
7089            whose address is never taken)
7090          + On i386, these local functions use register parameter passing
7091            conventions.
7092          + Reordering of functions in topological order of the call graph
7093            to enable better propagation of optimizing hints (such as the
7094            stack alignments needed by functions) in the back end.
7095          + Call graph based out-of-order inlining heuristics which allows
7096            to limit overall compilation unit growth (--param
7097            inline-unit-growth).
7098       Overall, the unit-at-a-time scheme produces a 1.3% improvement for
7099       the SPECint2000 benchmark on the i386 architecture (AMD Athlon
7100       CPU).
7101     * More realistic code size estimates used by inlining for C,
7102       Objective-C, C++ and Java. The growth of large functions can now be
7103       limited via --param large-function-insns and --param
7104       large-function-growth.
7105     * A new cfg-level loop optimizer pass replaces the old loop unrolling
7106       pass and adds two other loop transformations -- loop peeling and
7107       loop unswitching -- and also uses the profile feedback to limit
7108       code growth. (The three optimizations are enabled by
7109       -funroll-loops, -fpeel-loops and -funswitch-loops flags,
7110       respectively).
7111       The old loop unroller still can be enabled by -fold-unroll-loops
7112       and may produce better code in some cases, especially when the
7113       webizer optimization pass is not run.
7114     * A new web construction pass enabled via -fweb (and implied by -O3)
7115       improves the quality of register allocation, CSE, first scheduling
7116       pass and some other optimization passes by avoiding re-use of
7117       pseudo registers with non-overlapping live ranges. The pass almost
7118       always improves code quality but does make debugging difficult and
7119       thus is not enabled by default by -O2
7120       The pass is especially effective as cleanup after code duplication
7121       passes, such as the loop unroller or the tracer.
7122     * Experimental implementations of superblock or trace scheduling in
7123       the second scheduling pass can be enabled via
7124       -fsched2-use-superblocks and -fsched2-use-traces, respectively.
7125
7126New Languages and Language specific improvements
7127
7128  Ada
7129
7130     * The Ada front end has been updated to include numerous bug fixes
7131       and enhancements. These include:
7132          + Improved project file support
7133          + Additional set of warnings about potential wrong code
7134          + Improved error messages
7135          + Improved code generation
7136          + Improved cross reference information
7137          + Improved inlining
7138          + Better run-time check elimination
7139          + Better error recovery
7140          + More efficient implementation of unbounded strings
7141          + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools,
7142            ...
7143          + New GNAT.xxxx packages (e.g. GNAT.Strings,
7144            GNAT.Exception_Action)
7145          + New pragmas
7146          + New -gnatS switch replacing gnatpsta
7147          + Implementation of new Ada features (in particular limited
7148            with, limited aggregates)
7149
7150  C/Objective-C/C++
7151
7152     * Precompiled headers are now supported. Precompiled headers can
7153       dramatically speed up compilation of some projects. There are some
7154       known defects in the current precompiled header implementation that
7155       will result in compiler crashes in relatively rare situations.
7156       Therefore, precompiled headers should be considered a "technology
7157       preview" in this release. Read the manual for details about how to
7158       use precompiled headers.
7159     * File handling in the preprocessor has been rewritten. GCC no longer
7160       gets confused by symlinks and hardlinks, and now has a correct
7161       implementation of #import and #pragma once. These two directives
7162       have therefore been un-deprecated.
7163     * The undocumented extension that allowed C programs to have a label
7164       at the end of a compound statement, which has been deprecated since
7165       GCC 3.0, has been removed.
7166     * The cast-as-lvalue extension has been removed for C++ and
7167       deprecated for C and Objective-C. In particular, code like this:
7168        int i;
7169        (char) i = 5;
7170
7171       or this:
7172        char *p;
7173        ((int *) p)++;
7174
7175       is no longer accepted for C++ and will not be accepted for C and
7176       Objective-C in a future version.
7177     * The conditional-expression-as-lvalue extension has been deprecated
7178       for C and Objective-C. In particular, code like this:
7179        int a, b, c;
7180        (a ? b : c) = 2;
7181
7182       will not be accepted for C and Objective-C in a future version.
7183     * The compound-expression-as-lvalue extension has been deprecated for
7184       C and Objective-C. In particular, code like this:
7185        int a, b;
7186        (a, b) = 2;
7187
7188       will not be accepted for C and Objective-C in a future version. A
7189       possible non-intrusive workaround is the following:
7190        (*(a, &b)) = 2;
7191
7192     * Several [10]built-in functions such as __builtin_popcount for
7193       counting bits, finding the highest and lowest bit in a word, and
7194       parity have been added.
7195     * The -fwritable-strings option has been deprecated and will be
7196       removed.
7197     * Many C math library functions are now recognized as built-ins and
7198       optimized.
7199     * The C, C++, and Objective-C compilers can now handle source files
7200       written in any character encoding supported by the host C library.
7201       The default input character set is taken from the current locale,
7202       and may be overridden with the -finput-charset command line option.
7203       In the future we will add support for inline encoding markers.
7204
7205  C++
7206
7207     * G++ is now much closer to full conformance to the ISO/ANSI C++
7208       standard. This means, among other things, that a lot of invalid
7209       constructs which used to be accepted in previous versions will now
7210       be rejected. It is very likely that existing C++ code will need to
7211       be fixed. This document lists some of the most common issues.
7212     * A hand-written recursive-descent C++ parser has replaced the
7213       YACC-derived C++ parser from previous GCC releases. The new parser
7214       contains much improved infrastructure needed for better parsing of
7215       C++ source codes, handling of extensions, and clean separation
7216       (where possible) between proper semantics analysis and parsing. The
7217       new parser fixes many bugs that were found in the old parser.
7218     * You must now use the typename and template keywords to disambiguate
7219       dependent names, as required by the C++ standard.
7220        struct K {
7221          typedef int mytype_t;
7222        };
7223
7224        template <class T1> struct A {
7225          template <class T2> struct B {
7226              void callme(void);
7227            };
7228
7229          template <int N> void bar(void)
7230          {
7231            // Use 'typename' to tell the parser that T1::mytype_t names
7232            //  a type. This is needed because the name is dependent (in
7233            //  this case, on template parameter T1).
7234            typename T1::mytype_t x;
7235            x = 0;
7236          }
7237        };
7238
7239        template <class T> void template_func(void)
7240        {
7241          // Use 'template' to prefix member templates within
7242          //  dependent types (a has type A<T>, which depends on
7243          //  the template parameter T).
7244          A<T> a;
7245          a.template bar<0>();
7246
7247          // Use 'template' to tell the parser that B is a nested
7248          //  template class (dependent on template parameter T), and
7249          //  'typename' because the whole A<T>::B<int> is
7250          //  the name of a type (again, dependent).
7251          typename A<T>::template B<int> b;
7252          b.callme();
7253        }
7254
7255        void non_template_func(void)
7256        {
7257          // Outside of any template class or function, no names can be
7258          //  dependent, so the use of the keyword 'typename' and 'template'
7259          //  is not needed (and actually forbidden).
7260          A<K> a;
7261          a.bar<0>();
7262          A<K>::B<float> b;
7263          b.callme();
7264        }
7265     * In a template definition, unqualified names will no longer find
7266       members of a dependent base (as specified by [temp.dep]/3 in the
7267       C++ standard). For example,
7268        template <typename T> struct B {
7269          int m;
7270          int n;
7271          int f ();
7272          int g ();
7273        };
7274        int n;
7275        int g ();
7276        template <typename T> struct C : B<T> {
7277          void h ()
7278          {
7279            m = 0; // error
7280            f ();  // error
7281            n = 0; // ::n is modified
7282            g ();  // ::g is called
7283          }
7284        };
7285       You must make the names dependent, e.g. by prefixing them with
7286       this->. Here is the corrected definition of C<T>::h,
7287        template <typename T> void C<T>::h ()
7288        {
7289          this->m = 0;
7290          this->f ();
7291          this->n = 0
7292          this->g ();
7293        }
7294       As an alternative solution (unfortunately not backwards compatible
7295       with GCC 3.3), you may use using declarations instead of this->:
7296        template <typename T> struct C : B<T> {
7297          using B<T>::m;
7298          using B<T>::f;
7299          using B<T>::n;
7300          using B<T>::g;
7301          void h ()
7302          {
7303            m = 0;
7304            f ();
7305            n = 0;
7306            g ();
7307          }
7308        };
7309     * In templates, all non-dependent names are now looked up and bound
7310       at definition time (while parsing the code), instead of later when
7311       the template is instantiated. For instance:
7312        void foo(int);
7313
7314        template <int> struct A {
7315          static void bar(void){
7316            foo('a');
7317          }
7318        };
7319
7320        void foo(char);
7321
7322        int main()
7323        {
7324          A<0>::bar();    // Calls foo(int), used to call foo(char).
7325        }
7326
7327     * In an explicit instantiation of a class template, you must use
7328       class or struct before the template-id:
7329        template <int N>
7330        class A {};
7331
7332        template A<0>;         // error, not accepted anymore
7333        template class A<0>;   // OK
7334     * The "named return value" and "implicit typename" extensions have
7335       been removed.
7336     * Default arguments in function types have been deprecated and will
7337       be removed.
7338     * ARM-style name-injection of friend declarations has been deprecated
7339       and will be removed. For example: struct S { friend void f(); };
7340       void g() { f(); } will not be accepted by future versions of G++;
7341       instead a declaration of "f" will need to be present outside of the
7342       scope of "S".
7343     * Covariant returns are implemented for all but varadic functions
7344       that require an adjustment.
7345     * When -pedantic is used, G++ now issues errors about spurious
7346       semicolons. For example,
7347        namespace N {}; // Invalid semicolon.
7348        void f() {}; // Invalid semicolon.
7349     * G++ no longer accepts attributes for a declarator after the
7350       initializer associated with that declarator. For example,
7351        X x(1) __attribute__((...));
7352       is no longer accepted. Instead, use:
7353        X x __attribute__((...)) (1);
7354     * Inside the scope of a template class, the name of the class itself
7355       can be treated as either a class or a template. So GCC used to
7356       accept the class name as argument of type template, and template
7357       template parameter. However this is not C++ standard compliant. Now
7358       the name is not treated as a valid template template argument
7359       unless you qualify the name by its scope. For example, the code
7360       below no longer compiles.
7361        template <template <class> class TT> class X {};
7362        template <class T> class Y {
7363          X<Y> x; // Invalid, Y is always a type template parameter.
7364        };
7365       The valid code for the above example is
7366          X< ::Y> x; // Valid.
7367       (Notice the space between < and : to prevent GCC to interpret this
7368       as a digraph for [.)
7369     * Friend declarations that refer to template specializations are
7370       rejected if the template has not already been declared. For
7371       example,
7372        template <typename T>
7373        class C {
7374          friend void f<> (C&);
7375        };
7376       is rejected. You must first declare f as a template,
7377        template <typename T>
7378        void f(T);
7379     * In case of friend declarations, every name used in the friend
7380       declaration must be accessible at the point of that declaration.
7381       Previous versions of G++ used to be less strict about this and
7382       allowed friend declarations for private class members, for example.
7383       See the ISO C++ Standard Committee's [11]defect report #209 for
7384       details.
7385     * Declaration of member functions of class templates as friends are
7386       supported. For example,
7387        template <typename T> struct A {
7388          void f();
7389        };
7390        class C {
7391          template <typename T> friend void A<T>::f();
7392        };
7393     * You must use template <> to introduce template specializations, as
7394       required by the standard. For example,
7395        template <typename T>
7396        struct S;
7397
7398        struct S<int> { };
7399       is rejected. You must write,
7400        template <> struct S<int> {};
7401     * G++ used to accept code like this,
7402        struct S {
7403          int h();
7404          void f(int i = g());
7405          int g(int i = h());
7406        };
7407       This behavior is not mandated by the standard. Now G++ issues an
7408       error about this code. To avoid the error, you must move the
7409       declaration of g before the declaration of f. The default arguments
7410       for g must be visible at the point where it is called.
7411     * The C++ ABI Section 3.3.3 specifications for the array construction
7412       routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return
7413       NULL when the allocator argument returns NULL. These changes are
7414       incorporated into the libstdc++ runtime library.
7415     * Using a name introduced by a typedef in a friend declaration or in
7416       an explicit instantiation is now rejected, as specified by the ISO
7417       C++ standard.
7418        class A;
7419        typedef A B;
7420        class C {
7421          friend class B;      // error, no typedef name here
7422          friend B;            // error, friend always needs class/struct/enum
7423          friend class A;      // OK
7424        };
7425
7426        template <int> class Q {};
7427        typedef Q<0> R;
7428        template class R;      // error, no typedef name here
7429        template class Q<0>;   // OK
7430     * When allocating an array with a new expression, GCC used to allow
7431       parentheses around the type name. This is actually ill-formed and
7432       it is now rejected:
7433        int* a = new (int)[10];    // error, not accepted anymore
7434        int* a = new int[10];      // OK
7435     * When binding an rvalue of class type to a reference, the copy
7436       constructor of the class must be accessible. For instance, consider
7437       the following code:
7438        class A
7439        {
7440        public:
7441          A();
7442
7443        private:
7444          A(const A&);   // private copy ctor
7445        };
7446
7447        A makeA(void);
7448        void foo(const A&);
7449
7450        void bar(void)
7451        {
7452          foo(A());       // error, copy ctor is not accessible
7453          foo(makeA());   // error, copy ctor is not accessible
7454
7455          A a1;
7456          foo(a1);        // OK, a1 is a lvalue
7457        }
7458       This might be surprising at first sight, especially since most
7459       popular compilers do not correctly implement this rule ([12]further
7460       details).
7461     * When forming a pointer to member or a pointer to member function,
7462       access checks for class visibility (public, protected, private) are
7463       now performed using the qualifying scope of the name itself. This
7464       is better explained with an example:
7465        class A
7466        {
7467        public:
7468          void pub_func();
7469        protected:
7470          void prot_func();
7471        private:
7472          void priv_func();
7473        };
7474
7475        class B : public A
7476        {
7477        public:
7478          void foo()
7479          {
7480            &A::pub_func;   // OK, pub_func is accessible through A
7481            &A::prot_func;  // error, cannot access prot_func through A
7482            &A::priv_func;  // error, cannot access priv_func through A
7483
7484            &B::pub_func;   // OK, pub_func is accessible through B
7485            &B::prot_func;  // OK, can access prot_func through B (within B)
7486            &B::priv_func;  // error, cannot access priv_func through B
7487          }
7488        };
7489
7490    Runtime Library (libstdc++)
7491
7492     * Optimization work:
7493          + Streamlined streambuf, filebuf, separate synched with C
7494            Standard I/O streambuf.
7495          + All formatted I/O now uses cached locale information.
7496          + STL optimizations (memory/speed for list, red-black trees as
7497            used by sets and maps).
7498          + More use of GCC builtins.
7499          + String optimizations (avoid contention on
7500            increment/decrement-and-test of the reference count in the
7501            empty-string object, constructor from input_iterators
7502            speedup).
7503     * Static linkage size reductions.
7504     * Large File Support (files larger than 2 GB on 32-bit systems).
7505     * Wide character and variable encoding filebuf work (UTF-8, Unicode).
7506     * Generic character traits.
7507     * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x,
7508       Solaris 2.7 and above, AIX 5.x, Irix 6.5.
7509     * The allocator class is now standard-conformant, and two additional
7510       extension allocators have been added, mt_alloc and
7511       bitmap_allocator.
7512     * PCH support: -include bits/stdc++.h (2x compile speedup).
7513     * Rewrote __cxa_demangle with support for C++ style allocators.
7514     * New debug modes for STL containers and iterators.
7515     * Testsuite rewrite: five times as many tests, plus increasingly
7516       sophisticated tests, including I/O, MT, multi-locale, wide and
7517       narrow characters.
7518     * Use current versions of GNU "autotools" for build/configuration.
7519
7520  Objective-C
7521
7522     * The Objective-C front end has been updated to include the numerous
7523       bug fixes and enhancements previously available only in Apple's
7524       version of GCC. These include:
7525          + Structured exception (@try... @catch... @finally, @throw) and
7526            synchronization (@synchronized) support. These are accessible
7527            via the -fobjc-exceptions switch; as of this writing, they may
7528            only be used in conjunction with -fnext-runtime on Mac OS X
7529            10.3 and later. See [13]Options Controlling Objective-C
7530            Dialect for more information.
7531          + An overhaul of @encode logic. The C99 _Bool and C++ bool type
7532            may now be encoded as 'B'. In addition, the back-end/codegen
7533            dependencies have been removed.
7534          + An overhaul of message dispatch construction, ensuring that
7535            the various receiver types (and casts thereof) are handled
7536            properly, and that correct diagnostics are issued.
7537          + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue"
7538            (-freplace-objc-classes) debugging modes, currently available
7539            on Mac OS X 10.3 and later. See [14]Options Controlling
7540            Objective-C Dialect for more information.
7541          + Access to optimized runtime entry points (-fno-nil-receivers )
7542            on the assumption that message receivers are never nil. This
7543            is currently available on Mac OS X 10.3 and later. See
7544            [15]Options Controlling Objective-C Dialect for more
7545            information.
7546
7547  Java
7548
7549     * Compiling a .jar file will now cause non-.class entries to be
7550       automatically compiled as resources.
7551     * libgcj has been ported to Darwin.
7552     * Jeff Sturm has adapted Jan Hubicka's call graph optimization code
7553       to gcj.
7554     * libgcj has a new gcjlib URL type; this lets URLClassLoader load
7555       code from shared libraries.
7556     * libgcj has been much more completely merged with [16]GNU Classpath.
7557     * Class loading is now much more correct; in particular the caller's
7558       class loader is now used when that is required.
7559     * [17]Eclipse 2.x will run out of the box using gij.
7560     * Parts of java.nio have been implemented. Direct and indirect
7561       buffers work, as do fundamental file and socket operations.
7562     * java.awt has been improved, though it is still not ready for
7563       general use.
7564     * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST
7565       method.
7566     * The MinGW port has matured. Enhancements include socket timeout
7567       support, thread interruption, improved Runtime.exec() handling and
7568       support for accented characters in filenames.
7569
7570  Fortran
7571
7572     * Fortran improvements are listed in the [18]Fortran documentation.
7573
7574New Targets and Target Specific Improvements
7575
7576  Alpha
7577
7578     * Several [19]built-in functions have been added such as
7579       __builtin_alpha_zap to allow utilizing the more obscure
7580       instructions of the CPU.
7581     * Parameter passing of complex arguments has changed to match the
7582       [20]ABI. This change is incompatible with previous GCC versions,
7583       but does fix compatibility with the Tru64 compiler and several
7584       corner cases where GCC was incompatible with itself.
7585
7586  ARM
7587
7588     * Nicolas Pitre has contributed his hand-coded floating-point support
7589       code for ARM. It is both significantly smaller and faster than the
7590       existing C-based implementation, even when building applications
7591       for Thumb. The arm-elf configuration has been converted to use the
7592       new code.
7593     * Support for the Intel's iWMMXt architecture, a second generation
7594       XScale processor, has been added. Enabled at run time with the
7595       -mcpu=iwmmxt command line switch.
7596     * A new ARM target has been added: arm-wince-pe. This is similar to
7597       the arm-pe target, but it defaults to using the APCS32 ABI.
7598     * The existing ARM pipeline description has been converted to the use
7599       the [21]DFA processor pipeline model. There is not much change in
7600       code performance, but the description is now [22]easier to
7601       understand.
7602     * Support for the Cirrus EP9312 Maverick floating point co-processor
7603       added. Enabled at run time with the -mcpu=ep9312 command line
7604       switch. Note however that the multilibs to support this chip are
7605       currently disabled in gcc/config/arm/t-arm-elf, so if you want to
7606       enable their production you will have to uncomment the entries in
7607       that file.
7608
7609  H8/300
7610
7611     * Support for long long has been added.
7612     * Support for saveall attribute has been added.
7613     * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code
7614       for H8/300H and H8S, which is much faster than the previous
7615       implementation.
7616     * A lot of small performance improvements.
7617
7618  IA-32/AMD64 (x86-64)
7619
7620     * Tuning for K8 (AMD Opteron/Athlon64) core is available via
7621       -march=k8 and -mcpu=k8.
7622     * Scalar SSE code generation carefully avoids reformatting penalties,
7623       hidden dependencies and minimizes the number of uops generated on
7624       both Intel and AMD CPUs.
7625     * Vector MMX and SSE operands are now passed in registers to improve
7626       performance and match the argument passing convention used by the
7627       Intel C++ Compiler. As a result it is not possible to call
7628       functions accepting vector arguments compiled by older GCC version.
7629     * Conditional jump elimination is now more aggressive on modern CPUs.
7630     * The Athlon ports has been converted to use the DFA processor
7631       pipeline description.
7632     * Optimization of indirect tail calls is now possible in a similar
7633       fashion as direct sibcall optimization.
7634     * Further small performance improvements.
7635     * -m128bit-long-double is now less buggy.
7636     * __float128 support in 64-bit compilation.
7637     * Support for data structures exceeding 2GB in 64-bit mode.
7638     * -mcpu has been renamed to -mtune.
7639
7640  IA-64
7641
7642     * Tuning code for the Itanium 2 processor has been added. The
7643       generation of code tuned for Itanium 2 (option -mtune=itanium2) is
7644       enabled by default now. To generate code tuned for Itanium 1 the
7645       option -mtune=itanium1 should be used.
7646     * [23]DFA processor pipeline descriptions for the IA-64 processors
7647       have been added. This resulted in about 3% improvement on the
7648       SPECInt2000 benchmark for Itanium 2.
7649     * Instruction bundling for the IA-64 processors has been rewritten
7650       using the DFA pipeline hazard recognizer. It resulted in about 60%
7651       compiler speedup on the SPECInt2000 C programs.
7652
7653  M32R
7654
7655     * Support for the M32R/2 processor has been added by Renesas.
7656     * Support for an M32R GNU/Linux target and PIC code generation has
7657       been added by Renesas.
7658
7659  M68000
7660
7661     * Bernardo Innocenti (Develer S.r.l.) has contributed the
7662       m68k-uclinux target, based on former work done by Paul Dale
7663       (SnapGear Inc.). Code generation for the ColdFire processors family
7664       has been enhanced and extended to support the MCF 53xx and MCF 54xx
7665       cores, integrating former work done by Peter Barada (Motorola).
7666
7667  MIPS
7668
7669    Processor-specific changes
7670
7671     * Support for the RM7000 and RM9000 processors has been added. It can
7672       be selected using the -march compiler option and should work with
7673       any MIPS I (mips-*) or MIPS III (mips64-*) configuration.
7674     * Support for revision 2 of the MIPS32 ISA has been added. It can be
7675       selected with the command-line option -march=mips32r2.
7676     * There is a new option, -mfix-sb1, to work around certain SB-1
7677       errata.
7678
7679    Configuration
7680
7681     * It is possible to customize GCC using the following configure-time
7682       options:
7683          + --with-arch, which specifies the default value of the -march
7684            option.
7685          + --with-tune, which specifies the default value of the -mtune
7686            option.
7687          + --with-abi, which specifies the default ABI.
7688          + --with-float=soft, which tells GCC to use software floating
7689            point by default.
7690          + --with-float=hard, which tells GCC to use hardware floating
7691            point by default.
7692     * A 64-bit GNU/Linux port has been added. The associated
7693       configurations are mips64-linux-gnu and mips64el-linux-gnu.
7694     * The 32-bit GNU/Linux port now supports Java.
7695     * The IRIX 6 configuration now supports the o32 ABI and will build
7696       o32 multilibs by default. This support is compatible with both
7697       binutils and the SGI tools, but note that several features,
7698       including debugging information and DWARF2 exception handling, are
7699       only available when using the GNU assembler. Use of the GNU
7700       assembler and linker (version 2.15 or above) is strongly
7701       recommended.
7702     * The IRIX 6 configuration now supports 128-bit long doubles.
7703     * There are two new RTEMS-specific configurations, mips-rtems and
7704       mipsel-rtems.
7705     * There are two new *-elf configurations, mipsisa32r2-elf and
7706       mipsisa32r2el-elf.
7707
7708    General
7709
7710     * Several [24]ABI bugs have been fixed. Unfortunately, these changes
7711       will break binary compatibility with earlier releases.
7712     * GCC can now use explicit relocation operators when generating
7713       -mabicalls code. This behavior is controlled by -mexplicit-relocs
7714       and can have several performance benefits. For example:
7715          + It allows for more optimization of GOT accesses, including
7716            better scheduling and redundancy elimination.
7717          + It allows sibling calls to be implemented as jumps.
7718          + n32 and n64 leaf functions can use a call-clobbered global
7719            pointer instead of $28.
7720          + The code to set up $gp can be removed from functions that
7721            don't need it.
7722     * A new option, -mxgot, allows the GOT to be bigger than 64k. This
7723       option is equivalent to the assembler's -xgot option and should be
7724       used instead of -Wa,-xgot.
7725     * Frame pointer elimination is now supported when generating 64-bit
7726       MIPS16 code.
7727     * Inline block moves have been optimized to take more account of
7728       alignment information.
7729     * Many internal changes have been made to the MIPS port, mostly aimed
7730       at reducing the reliance on assembler macros.
7731
7732  PowerPC
7733
7734     * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64
7735       [25]ABI incompatibilities regarding the way parameters are passed
7736       during functions calls. These changes may result in incompatibility
7737       between code compiled with GCC 3.3 and GCC 3.4.
7738
7739    PowerPC Darwin
7740
7741     * Support for shared/dylib gcc libraries has been added. It is
7742       enabled by default on powerpc-apple-darwin7.0.0 and up.
7743     * Libgcj is enabled by default. On systems older than
7744       powerpc-apple-darwin7.0.0 you need to install dlcompat.
7745     * 128-bit IBM extended precision format support added for long
7746       double.
7747
7748    PowerPC64 GNU/Linux
7749
7750     * By default, PowerPC64 GNU/Linux now uses natural alignment of
7751       structure elements. The old four byte alignment for double, with
7752       special rules for a struct starting with a double, can be chosen
7753       with -malign-power. This change may result in incompatibility
7754       between code compiled with GCC 3.3 and GCC 3.4.
7755     * -mabi=altivec is now the default rather than -mabi=no-altivec.
7756     * 128-bit IBM extended precision format support added for long
7757       double.
7758
7759  S/390 and zSeries
7760
7761     * New command-line options allow to specify the intended execution
7762       environment for generated code:
7763          + -mesa/-mzarch allows to specify whether to generate code
7764            running in ESA/390 mode or in z/Architecture mode (this is
7765            applicable to 31-bit code only).
7766          + -march allows to specify a minimum processor architecture
7767            level (g5, g6, z900, or z990).
7768          + -mtune allows to specify which processor to tune for.
7769     * It is possible to customize GCC using the following configure-time
7770       options:
7771          + --with-mode, which specifies whether to default to assuming
7772            ESA/390 or z/Architecture mode.
7773          + --with-arch, which specifies the default value of the -march
7774            option.
7775          + --with-tune, which specifies the default value of the -mtune
7776            option.
7777     * Support for the z990 processor has been added, and can be selected
7778       using -march=z990 or -mtune=z990. This includes instruction
7779       scheduling tuned for the superscalar instruction pipeline of the
7780       z990 processor as well as support for all new instructions provided
7781       by the long-displacement facility.
7782     * Support to generate 31-bit code optimized for zSeries processors
7783       (running in ESA/390 or in z/Architecture mode) has been added. This
7784       can be selected using -march=z900 and -mzarch respectively.
7785     * Instruction scheduling for the z900 and z990 processors now uses
7786       the DFA pipeline hazard recognizer.
7787     * GCC no longer generates code to maintain a stack backchain,
7788       previously used to generate stack backtraces for debugging
7789       purposes. As replacement that does not incur runtime overhead,
7790       DWARF-2 call frame information is provided by GCC; this is
7791       supported by GDB 6.1. The old behavior can be restored using the
7792       -mbackchain option.
7793     * The stack frame size of functions may now exceed 2 GB in 64-bit
7794       code.
7795     * A port for the 64-bit IBM TPF operating system has been added; the
7796       configuration is s390x-ibm-tpf. This configuration is supported as
7797       cross-compilation target only.
7798     * Various changes to improve the generated code have been
7799       implemented, including:
7800          + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT
7801            instructions to significantly speed up many floating-point
7802            applications.
7803          + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL
7804            WITH BORROW instructions to speed up long long arithmetic.
7805          + GCC now uses the SEARCH STRING instruction to implement
7806            strlen().
7807          + In many cases, function call overhead for 31-bit code has been
7808            reduced by placing the literal pool after the function code
7809            instead of after the function prolog.
7810          + Register 14 is no longer reserved in 64-bit code.
7811          + Handling of global register variables has been improved.
7812
7813  SPARC
7814
7815     * The option -mflat is deprecated.
7816     * Support for large (> 2GB) frames has been added to the 64-bit port.
7817     * Several [26]ABI bugs have been fixed. Unfortunately, these changes
7818       will break binary compatibility with earlier releases.
7819     * The default debugging format has been switched from STABS to
7820       DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already
7821       the default debugging format for 64-bit code on Solaris.
7822
7823  SuperH
7824
7825     * Support for the SH2E processor has been added. Enabled at run time
7826       with the -m2e command line switch, or at configure time by
7827       specifying sh2e as the machine part of the target triple.
7828
7829  V850
7830
7831     * Support for the Mitsubishi V850E1 processor has been added. This is
7832       a variant of the V850E processor with some additional debugging
7833       instructions.
7834
7835  Xtensa
7836
7837     * Several ABI bugs have been fixed. Unfortunately, these changes
7838       break binary compatibility with earlier releases.
7839          + For big-endian processors, the padding of aggregate return
7840            values larger than a word has changed. If the size of an
7841            aggregate return value is not a multiple of 32 bits, previous
7842            versions of GCC inserted padding in the most-significant bytes
7843            of the first return value register. Aggregates larger than a
7844            word are now padded in the least-significant bytes of the last
7845            return value register used. Aggregates smaller than a word are
7846            still padded in the most-significant bytes. The return value
7847            padding has not changed for little-endian processors.
7848          + Function arguments with 16-byte alignment are now properly
7849            aligned.
7850          + The implementation of the va_list type has changed. A va_list
7851            value created by va_start from a previous release cannot be
7852            used with va_arg from this release, or vice versa.
7853     * More processor configuration options for Xtensa processors are
7854       supported:
7855          + the ABS instruction is now optional;
7856          + the ADDX* and SUBX* instructions are now optional;
7857          + an experimental CONST16 instruction can be used to synthesize
7858            constants instead of loading them from constant pools.
7859       These and other Xtensa processor configuration options can no
7860       longer be enabled or disabled by command-line options; the
7861       processor configuration must be specified by the xtensa-config.h
7862       header file when building GCC. Additionally, the
7863       -mno-serialize-volatile option is no longer supported.
7864
7865Obsolete Systems
7866
7867   Support for a number of older systems has been declared obsolete in GCC
7868   3.4. Unless there is activity to revive them, the next release of GCC
7869   will have their sources permanently removed.
7870
7871   All configurations of the following processor architectures have been
7872   declared obsolete:
7873     * Mitsubishi D30V, d30v-*
7874     * AT&T DSP1600 and DSP1610, dsp16xx-*
7875     * Intel 80960, i960
7876
7877   Also, some individual systems have been obsoleted:
7878     * ARM Family
7879          + Support for generating code for operation in APCS/26 mode
7880            (-mapcs-26).
7881     * IBM ESA/390
7882          + "Bigfoot" port, i370-*. (The other port, s390-*, is actively
7883            maintained and supported.)
7884     * Intel 386 family
7885          + MOSS, i?86-moss-msdos and i?86-*-moss*
7886          + NCR 3000 running System V r.4, i?86-ncr-sysv4*
7887          + FreeBSD with a.out object format, i?86-*-freebsd*aout* and
7888            i?86-*-freebsd2*
7889          + GNU/Linux with a.out object format, i?86-linux*aout*
7890          + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1*
7891          + Interix versions before Interix 3, i?86-*-interix
7892          + Mach microkernel, i?86-mach*
7893          + SCO UnixWare with UDK, i?86-*-udk*
7894          + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]*
7895          + VSTa microkernel, i386-*-vsta
7896     * Motorola M68000 family
7897          + HPUX, m68k-hp-hpux* and m68000-hp-hpux*
7898          + NetBSD with a.out object format (before NetBSD 1.4),
7899            m68k-*-*-netbsd* except m68k-*-*-netbsdelf*
7900          + Generic System V r.4, m68k-*-sysv4*
7901     * VAX
7902          + Generic VAX, vax-*-* (This is generic VAX only; we have not
7903            obsoleted any VAX triples for specific operating systems.)
7904
7905Documentation improvements
7906
7907Other significant improvements
7908
7909     * The build system has undergone several significant cleanups.
7910       Subdirectories will only be configured if they are being built, and
7911       all subdirectory configures are run from the make command. The top
7912       level has been autoconfiscated.
7913     * Building GCC no longer writes to its source directory. This should
7914       help those wishing to share a read-only source directory over NFS
7915       or build from a CD. The exceptions to this feature are if you
7916       configure with either --enable-maintainer-mode or
7917       --enable-generated-files-in-srcdir.
7918     * The -W warning option has been renamed to -Wextra, which is more
7919       easily understood. The older spelling will be retained for
7920       backwards compatibility.
7921     * Substantial improvements in compile time have been made,
7922       particularly for non-optimizing compilations.
7923     __________________________________________________________________
7924
7925GCC 3.4.0
7926
7927  Bug Fixes
7928
7929   A vast number of bugs have been fixed in 3.4.0, too many to publish a
7930   complete list here. [27]Follow this link to query the Bugzilla database
7931   for the list of over 900 bugs fixed in 3.4.0. This is the list of all
7932   bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4
7933   regressions.
7934     __________________________________________________________________
7935
7936GCC 3.4.1
7937
7938  Bug Fixes
7939
7940   This section lists the problem reports (PRs) from GCC's bug tracking
7941   system that are known to be fixed in the 3.4.1 release. This list might
7942   not be complete (that is, it is possible that some PRs that have been
7943   fixed are not listed here).
7944
7945    Bootstrap failures
7946
7947     * [28]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler
7948       emitted - PIC related
7949     * [29]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf
7950     * [30]14760 A bug in configure.in prevents using both
7951       --program-suffix and --program-prefix
7952     * [31]14671 [hppa64] bootstrap fails: ICE in
7953       save_call_clobbered_regs, in caller_save.c
7954     * [32]15093 [alpha][Java] make bootstrap fails to configure libffi on
7955       Alpha
7956     * [33]15178 Solaris 9/x86 fails linking after stage 3
7957
7958    Multi-platform internal compiler errors (ICEs)
7959
7960     * [34]12753 (preprocessor) Memory corruption in preprocessor on bad
7961       input
7962     * [35]13985 ICE in gcc.c-torture/compile/930621-1.c
7963     * [36]14810 (c++) tree check failures with invalid code involving
7964       templates
7965     * [37]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in
7966       cp/parser.c
7967     * [38]15044 (c++) ICE on syntax error, template header
7968     * [39]15057 (c++) Compiling of conditional value throw constructs
7969       cause a segmentation violation
7970     * [40]15064 (c++) typeid of template parameter gives ICE
7971     * [41]15142 (c++) ICE when passing a string where a char* is expected
7972       in a throw statement
7973     * [42]15159 ICE in rtl_verify_flow_info_1
7974     * [43]15165 (c++) ICE in instantiate_template
7975     * [44]15193 Unary minus using pointer to V4SF vector causes
7976       -fforce-mem to exhaust all memory
7977     * [45]15209 (c++) Runs out of memory with packed structs
7978     * [46]15227 (c++) Trouble with invalid function definition
7979     * [47]15285 (c++) instantiate_type ICE when forming pointer to
7980       template function
7981     * [48]15299 (c++) ICE in resolve_overloaded_unification
7982     * [49]15329 (c++) ICE on constructor of member template
7983     * [50]15550 ICE in extract_insn, in recog.c
7984     * [51]15554 (c++) ICE in tsubst_copy, in cp/pt.c
7985     * [52]15640 (c++) ICE on invalid code in arg_assoc, in
7986       cp/name-lookup.c
7987     * [53]15666 [unit-at-a-time] Gcc abort on valid code
7988     * [54]15696 (c++) ICE with bad pointer-to-member code
7989     * [55]15701 (c++) ICE with friends and template template parameter
7990     * [56]15761 ICE in do_SUBST, in combine.c
7991     * [57]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops
7992
7993    Ada
7994
7995     * [58]14538 All RTEMS targets broken for gnat
7996
7997    C front end
7998
7999     * [59]12391 missing warning about assigning to an incomplete type
8000     * [60]14649 atan(1.0) should not be a constant expression
8001     * [61]15004 [unit-at-a-time] no warning for unused paramater in
8002       static function
8003     * [62]15749 --pedantic-errors behaves differently from --pedantic
8004       with C-compiler on GNU/Linux
8005
8006    C++ compiler and library
8007
8008     * [63]10646 non-const reference is incorrectly matched in a "const T"
8009       partial specialization
8010     * [64]12077 wcin.rdbuf()->in_avail() return value too high
8011     * [65]13598 enc_filebuf doesn't work
8012     * [66]14211 const_cast returns lvalue but should be rvalue
8013     * [67]14220 num_put::do_put() undesired float/double behavior
8014     * [68]14245 problem with user-defined allocators in std::basic_string
8015     * [69]14340 libstdc++ Debug mode: failure to convert iterator to
8016       const_iterator
8017     * [70]14600 __gnu_cxx::stdio_sync_filebuf should expose internal
8018       FILE*
8019     * [71]14668 no warning anymore for reevaluation of declaration
8020     * [72]14775 LFS (large file support) tests missing
8021     * [73]14821 Duplicate namespace alias declaration should not conflict
8022     * [74]14930 Friend declaration ignored
8023     * [75]14932 cannot use offsetof to get offsets of array elements in
8024       g++ 3.4.0
8025     * [76]14950 [non unit-at-a-time] always_inline does not mix with
8026       templates and -O0
8027     * [77]14962 g++ ignores #pragma redefine_extname
8028     * [78]14975 Segfault on low-level write error during imbue
8029     * [79]15002 Linewise stream input is unusably slow (std::string slow)
8030     * [80]15025 compiler accepts redeclaration of template as
8031       non-template
8032     * [81]15046 [arm] Math functions misdetected by cross configuration
8033     * [82]15069 a bit test on a variable of enum type is miscompiled
8034     * [83]15074 g++ -lsupc++ still links against libstdc++
8035     * [84]15083 spurious "statement has no effect" warning
8036     * [85]15096 parse error with templates and pointer to const member
8037     * [86]15287 combination of operator[] and operator .* fails in
8038       templates
8039     * [87]15317 __attribute__ unused in first parameter of constructor
8040       gives error
8041     * [88]15337 sizeof on incomplete type diagnostic
8042     * [89]15361 bitset<>::_Find_next fails
8043     * [90]15412 _GLIBCXX_ symbols symbols defined and used in different
8044       namespaces
8045     * [91]15427 valid code results in incomplete type error
8046     * [92]15471 Incorrect member pointer offsets in anonymous
8047       structs/unions
8048     * [93]15503 nested template problem
8049     * [94]15507 compiler hangs while laying out union
8050     * [95]15542 operator & and template definitions
8051     * [96]15565 SLES9: leading + sign for unsigned int with showpos
8052     * [97]15625 friend defined inside a template fails to find static
8053       function
8054     * [98]15629 Function templates, overloads, and friend name injection
8055     * [99]15742 'noreturn' attribute ignored in method of template
8056       functions.
8057     * [100]15775 Allocator::pointer consistently ignored
8058     * [101]15821 Duplicate namespace alias within namespace rejected
8059     * [102]15862 'enum yn' fails (confict with undeclared builtin)
8060     * [103]15875 rejects pointer to member in template
8061     * [104]15877 valid code using templates and anonymous enums is
8062       rejected
8063     * [105]15947 Puzzling error message for wrong destructor declaration
8064       in template class
8065     * [106]16020 cannot copy __gnu_debug::bitset
8066     * [107]16154 input iterator concept too restrictive
8067     * [108]16174 deducing top-level consts
8068
8069    Java
8070
8071     * [109]14315 Java compiler is not parallel make safe
8072
8073    Fortran
8074
8075     * [110]15151 [g77] incorrect logical i/o in 64-bit mode
8076
8077    Objective-C
8078
8079     * [111]7993 private variables cannot be shadowed in subclasses
8080
8081    Optimization bugs
8082
8083     * [112]15228 useless copies of floating point operands
8084     * [113]15345 [non-unit-at-a-time] unreferenced nested inline
8085       functions not optimized away
8086     * [114]15945 Incorrect floating point optimization
8087     * [115]15526 ftrapv aborts on 0 * (-1)
8088     * [116]14690 Miscompiled POOMA tests
8089     * [117]15112 GCC generates code to write to unchanging memory
8090
8091    Preprocessor
8092
8093     * [118]15067 Minor glitch in the source of cpp
8094
8095    Main driver program bugs
8096
8097     * [119]1963 collect2 interprets -oldstyle_liblookup as -o
8098       ldstyle_liblookup
8099
8100    x86-specific (Intel/AMD)
8101
8102     * [120]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND*
8103       section}
8104
8105    HPPA-specific
8106
8107     * [121]14782 GCC produces an unaligned data access at -O2
8108     * [122]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2
8109     * [123]15202 ICE in reload_cse_simplify_operands, in postreload.c
8110
8111    IA64-specific
8112
8113     * [124]14610 __float80 constants incorrectly emitted
8114     * [125]14813 init_array sections are initialized in the wrong order
8115     * [126]14857 GCC segfault on duplicated asm statement
8116     * [127]15598 Gcc 3.4 ICE on valid code
8117     * [128]15653 Gcc 3.4 ICE on valid code
8118
8119    MIPS-specific
8120
8121     * [129]15189 wrong filling of delay slot with -march=mips1 -G0
8122       -mno-split-addresses -mno-explicit-relocs
8123     * [130]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as
8124       2.14.91
8125     * [131]16144 Bogus reference to __divdf3 when -O1
8126     * [132]16176 Miscompilation of unaligned data in MIPS backend
8127
8128    PowerPC-specific
8129
8130     * [133]11591 ICE in gcc.dg/altivec-5.c
8131     * [134]12028 powerpc-eabispe produces bad sCOND operation
8132     * [135]14478 rs6000 geu/ltu patterns generate incorrect code
8133     * [136]14567 long double and va_arg complex args
8134     * [137]14715 Altivec stack layout may overlap gpr save with stack
8135       temps
8136     * [138]14902 (libstdc++) Stream checking functions fail when -pthread
8137       option is used.
8138     * [139]14924 Compiler ICE on valid code
8139     * [140]14960 -maltivec affects vector return with -mabi=no-altivec
8140     * [141]15106 vector varargs failure passing from altivec to
8141       non-altivec code for -m32
8142     * [142]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 &
8143       half-word operation
8144     * [143]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx
8145       and stvx
8146     * [144]15662 Segmentation fault when an exception is thrown - even if
8147       try and catch are specified
8148
8149    s390-specific
8150
8151     * [145]15054 Bad code due to overlapping stack temporaries
8152
8153    SPARC-specific
8154
8155     * [146]15783 ICE with union assignment in 64-bit mode
8156     * [147]15626 GCC 3.4 emits "ld: warning: relocation error:
8157       R_SPARC_UA32"
8158
8159    x86-64-specific
8160
8161     * [148]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64
8162     * [149]14723 Backported -march=nocona from mainline
8163     * [150]15290 __float128 failed to pass to function properly
8164
8165    Cygwin/Mingw32-specific
8166
8167     * [151]15250 Option -mms-bitfields support on GCC 3.4 is not
8168       conformant to MS layout
8169     * [152]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe
8170       worker on windows32 targets
8171
8172    Bugs specific to embedded processors
8173
8174     * [153]8309 [m68k] -m5200 produces erroneous SImode set of short
8175       varaible on stack
8176     * [154]13250 [SH] Gcc code for rotation clobbers the register, but
8177       gcc continues to use the register as if it was not clobbered
8178     * [155]13803 [coldfire] movqi operand constraints too restrictivefor
8179       TARGET_COLDFIRE
8180     * [156]14093 [SH] ICE for code when using -mhitachi option in SH
8181     * [157]14457 [m6811hc] ICE with simple c++ source
8182     * [158]14542 [m6811hc] ICE on simple source
8183     * [159]15100 [SH] cc1plus got hang-up on
8184       libstdc++-v3/testsuite/abi_check.cc
8185     * [160]15296 [CRIS] Delayed branch scheduling causing invalid code on
8186       cris-*
8187     * [161]15396 [SH] ICE with -O2 -fPIC
8188     * [162]15782 [coldfire] m68k_output_mi_thunk emits wrong code for
8189       ColdFire
8190
8191    Testsuite problems (compiler not affected)
8192
8193     * [163]11610 libstdc++ testcases 27_io/* don't work properly remotely
8194     * [164]15488 (libstdc++) possibly insufficient file permissions for
8195       executing test suite
8196     * [165]15489 (libstdc++) testsuite_files determined incorrectly
8197
8198    Documentation bugs
8199
8200     * [166]13928 (libstdc++) no whatis info in some man pages generated
8201       by doxygen
8202     * [167]14150 Ada documentation out of date
8203     * [168]14949 (c++) Need to document method visibility changes
8204     * [169]15123 libstdc++-doc: Allocators.3 manpage is empty
8205     __________________________________________________________________
8206
8207GCC 3.4.2
8208
8209  Bug Fixes
8210
8211   This section lists the problem reports (PRs) from GCC's bug tracking
8212   system that are known to be fixed in the 3.4.2 release. This list might
8213   not be complete (that is, it is possible that some PRs that have been
8214   fixed are not listed here).
8215
8216    Bootstrap failures and issues
8217
8218     * [170]16469 [mips-sgi-irix5.3] bootstrap fails in
8219       libstdc++-v3/testsuite
8220     * [171]16344 [hppa-linux-gnu] libstdc++'s PCH built by
8221       profiledbootstrap does not work with the built compiler
8222     * [172]16842 [Solaris/x86] mkheaders can not find mkheaders.conf
8223
8224    Multi-platform internal compiler errors (ICEs)
8225
8226     * [173]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in
8227       cp_parser_class_specifier, in cp/parser.c
8228     * [174]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
8229     * [175]15461 (c++) ICE due to NRV and inlining
8230     * [176]15890 (c++) ICE in c_expand_expr, in c-common.c
8231     * [177]16180 ICE: segmentation fault in RTL optimization
8232     * [178]16224 (c++) ICE in write_unscoped_name (template/namespace)
8233     * [179]16408 ICE: in delete_insn, in cfgrtl.c
8234     * [180]16529 (c++) ICE for: namespace-alias shall not be declared as
8235       the name of any other entity
8236     * [181]16698 (c++) ICE with exceptions and declaration of __cxa_throw
8237     * [182]16706 (c++) ICE in finish_member_declaration, in
8238       cp/semantics.c
8239     * [183]16810 (c++) Legal C++ program with cast gives ICE in
8240       build_ptrmemfunc
8241     * [184]16851 (c++) ICE when throwing a comma expression
8242     * [185]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c
8243     * [186]16904 (c++) ICE in finish_class_member_access_expr, in
8244       cp/typeck.c
8245     * [187]16905 (c++) ICE (segfault) with exceptions
8246     * [188]16964 (c++) ICE in cp_parser_class_specifier due to
8247       redefinition
8248     * [189]17068 (c++) ICE: tree check: expected class 'd', have 'x'
8249       (identifier_node) in dependent_template_p, in cp/pt.c
8250
8251    Preprocessor bugs
8252
8253     * [190]16366 Preprocessor option -remap causes memory corruption
8254
8255    Optimization
8256
8257     * [191]15345 unreferenced nested inline functions not optimized away
8258     * [192]16590 Incorrect execution when compiling with -O2
8259     * [193]16693 Bitwise AND is lost when used within a cast to an enum
8260       of the same precision
8261     * [194]17078 Jump into if(0) substatement fails
8262
8263    Problems in generated debug information
8264
8265     * [195]13956 incorrect stabs for nested local variables
8266
8267    C front end bugs
8268
8269     * [196]16684 GCC should not warn about redundant redeclarations of
8270       built-ins
8271
8272    C++ compiler and library
8273
8274     * [197]12658 Thread safety problems in locale::global() and
8275       locale::locale()
8276     * [198]13092 g++ accepts invalid pointer-to-member conversion
8277     * [199]15320 Excessive memory consumption
8278     * [200]16246 Incorrect template argument deduction
8279     * [201]16273 Memory exhausted when using nested classes and virtual
8280       functions
8281     * [202]16401 ostringstream in gcc 3.4.x very slow for big data
8282     * [203]16411 undefined reference to
8283       __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char>
8284       >::file()
8285     * [204]16489 G++ incorrectly rejects use of a null constant integral
8286       expression as a null constant pointer
8287     * [205]16618 offsetof fails with constant member
8288     * [206]16637 syntax error reported for valid input code
8289     * [207]16717 __attribute__((constructor)) broken in C++
8290     * [208]16813 compiler error in DEBUG version of range insertion
8291       std::map::insert
8292     * [209]16853 pointer-to-member initialization from incompatible one
8293       accepted
8294     * [210]16889 ambiguity is not detected
8295     * [211]16959 Segmentation fault in ios_base::sync_with_stdio
8296
8297    Java compiler and library
8298
8299     * [212]7587 direct threaded interpreter not thread-safe
8300     * [213]16473 ServerSocket accept() leaks file descriptors
8301     * [214]16478 Hash synchronization deadlock with finalizers
8302
8303    Alpha-specific
8304
8305     * [215]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c
8306     * [216]16974 could not split insn (ice in final_scan_insn, in
8307       final.c)
8308
8309    x86-specific
8310
8311     * [217]16298 ICE in output_operand
8312     * [218]17113 ICE with SSE2 intrinsics
8313
8314    x86-64 specific
8315
8316     * [219]14697 libstdc++ couldn't find 32bit libgcc_s
8317
8318    MIPS-specific
8319
8320     * [220]15869 [mips64] No NOP after LW (with -mips1 -O0)
8321     * [221]16325 [mips64] value profiling clobbers gp on mips
8322     * [222]16357 [mipsisa64-elf] ICE copying 7 bytes between extern
8323       char[]s
8324     * [223]16380 [mips64] Use of uninitialised register after dbra
8325       conversion
8326     * [224]16407 [mips64] Unaligned access to local variables
8327     * [225]16643 [mips64] verify_local_live_at_start ICE after
8328       crossjumping & cfgcleanup
8329
8330    ARM-specific
8331
8332     * [226]15927 THUMB -O2: strength-reduced iteration variable ends up
8333       off by 1
8334     * [227]15948 THUMB: ICE with non-commutative cbranch
8335     * [228]17019 THUMB: bad switch statement in md code for
8336       addsi3_cbranch_scratch
8337
8338    IA64-specific
8339
8340     * [229]16130 ICE on valid code: in bundling, in config/ia64/ia64.c
8341       (-mtune=merced)
8342     * [230]16142 ICE on valid code: in bundling, in config/ia64/ia64.c
8343       (-mtune=itanium)
8344     * [231]16278 Gcc failed to build Linux kernel with -mtune=merced
8345     * [232]16414 ICE on valid code: typo in comparison of asm_noperands
8346       result
8347     * [233]16445 ICE on valid code: don't count ignored insns
8348     * [234]16490 ICE (segfault) while compiling with -fprofile-use
8349     * [235]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS
8350
8351    PowerPC-specific
8352
8353     * [236]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x
8354     * [237]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions
8355       issue)
8356
8357    SPARC-specific
8358
8359     * [238]16199 ICE while compiling apache 2.0.49
8360     * [239]16416 -m64 doesn't imply -mcpu=v9 anymore
8361     * [240]16430 ICE when returning non-C aggregates larger than 16 bytes
8362
8363    Bugs specific to embedded processors
8364
8365     * [241]16379 [m32r] can't output large model function call of memcpy
8366     * [242]17093 [m32r] ICE with -msdata=use -O0
8367     * [243]17119 [m32r] ICE at switch case 0x8000
8368
8369    DJGPP-specific
8370
8371     * [244]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp
8372
8373    Alpha Tru64-specific
8374
8375     * [245]16210 libstdc++ gratuitously omits "long long" I/O
8376
8377    Testsuite, documentation issues (compiler is not affected):
8378
8379     * [246]15488 (libstdc++) possibly insufficient file permissions for
8380       executing test suite
8381     * [247]16250 ada/doctools runs makeinfo even in release tarball
8382     __________________________________________________________________
8383
8384GCC 3.4.3
8385
8386   This is the [248]list of problem reports (PRs) from GCC's bug tracking
8387   system that are known to be fixed in the 3.4.3 release. This list might
8388   not be complete (that is, it is possible that some PRs that have been
8389   fixed are not listed here).
8390
8391    Bootstrap failures
8392
8393     * [249]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1
8394     * [250]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold
8395       when undeclared
8396
8397    Internal compiler errors (ICEs) affecting multiple platforms
8398
8399     * [251]13948 (java) GCJ segmentation fault while compiling GL4Java
8400       .class files
8401     * [252]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
8402     * [253]16301 (c++) ICE when "strong" attribute is attached to a using
8403       directive
8404     * [254]16566 ICE with flexible arrays
8405     * [255]17023 ICE with nested functions in parameter declaration
8406     * [256]17027 ICE with noreturn function in loop at -O2
8407     * [257]17524 ICE in grokdeclarator, in cp/decl.c
8408     * [258]17826 (c++) ICE in cp_tree_equal
8409
8410    C and optimization bugs
8411
8412     * [259]15526 -ftrapv aborts on 0 * (-1)
8413     * [260]16999 #ident stopped working
8414     * [261]17503 quadratic behaviour in invalid_mode_change_p
8415     * [262]17581 Long long arithmetic fails inside a switch/case
8416       statement when compiled with -O2
8417     * [263]18129 -fwritable-strings doesn't work
8418
8419    C++ compiler and library bugs
8420
8421     * [264]10975 incorrect initial ostringstream::tellp()
8422     * [265]11722 Unbuffered filebuf::sgetn is slow
8423     * [266]14534 Unrecognizing static function as a template parameter
8424       when its return value is also templated
8425     * [267]15172 Copy constructor optimization in aggregate
8426       initialization
8427     * [268]15786 Bad error message for frequently occuring error.
8428     * [269]16162 Rejects valid member-template-definition
8429     * [270]16612 empty basic_strings can't live in shared memory
8430     * [271]16715 std::basic_iostream is instantiated when used, even
8431       though instantiations are already contained in libstdc++
8432     * [272]16848 code in /ext/demangle.h appears broken
8433     * [273]17132 GCC fails to eliminate function template specialization
8434       when argument deduction fails
8435     * [274]17259 One more _S_leaf incorrectly qualified with _RopeRep::
8436       in ropeimpl.h
8437     * [275]17327 use of `enumeral_type' in template type unification
8438     * [276]17393 "unused variable '._0'" warning with -Wall
8439     * [277]17501 Confusion with member templates
8440     * [278]17537 g++ not passing -lstdc++ to linker when all command line
8441       arguments are libraries
8442     * [279]17585 usage of unqualified name of static member from within
8443       class not allowed
8444     * [280]17821 Poor diagnostic for using "." instead of "->"
8445     * [281]17829 wrong error: call of overloaded function is ambiguous
8446     * [282]17851 Misleading diagnostic for invalid function declarations
8447       with undeclared types
8448     * [283]17976 Destructor is called twice
8449     * [284]18020 rejects valid definition of enum value in template
8450     * [285]18093 bogus conflict in namespace aliasing
8451     * [286]18140 C++ parser bug when using >> in templates
8452
8453    Fortran
8454
8455     * [287]17541 data statements with double precision constants fail
8456
8457    x86-specific
8458
8459     * [288]17853 -O2 ICE for MMX testcase
8460
8461    SPARC-specific
8462
8463     * [289]17245 ICE compiling gsl-1.5 statistics/lag1.c
8464
8465    Darwin-specific
8466
8467     * [290]17167 FATAL:Symbol L_foo$stub already defined.
8468
8469    AIX-specific
8470
8471     * [291]17277 could not catch an exception when specified -maix64
8472
8473    Solaris-specific
8474
8475     * [292]17505 <cmath> calls acosf(), ceilf(), and other functions
8476       missing from system libraries
8477
8478    HP/UX specific:
8479
8480     * [293]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl
8481
8482    ARM-specific
8483
8484     * [294]17384 ICE with mode attribute on structures
8485
8486    MIPS-specific
8487
8488     * [295]17770 No NOP after LWL with -mips1
8489
8490    Other embedded target specific
8491
8492     * [296]11476 [arc-elf] gcc ICE on newlib's vfprintf.c
8493     * [297]14064 [avr-elf] -fdata-sections triggers ICE
8494     * [298]14678 [m68hc11-elf] gcc ICE
8495     * [299]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__
8496     * [300]15790 [i686-coff] Alignment error building gcc with i686-coff
8497       target
8498     * [301]15886 [SH] Miscompilation with -O2 -fPIC
8499     * [302]16884 [avr-elf] [fweb related] bug while initializing
8500       variables
8501
8502    Bugs relating to debugger support
8503
8504     * [303]13841 missing debug info for _Complex function arguments
8505     * [304]15860 [big-endian targets] No DW_AT_location debug info is
8506       emitted for formal arguments to a function that uses "register"
8507       qualifiers
8508
8509    Testsuite issues (compiler not affected)
8510
8511     * [305]17465 Testsuite in libffi overrides LD_LIBRARY_PATH
8512     * [306]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH
8513     * [307]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit
8514       testsuite
8515
8516    Documentation
8517
8518     * [308]15498 typo in gcc manual: non-existing locale example en_UK,
8519       should be en_GB
8520     * [309]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap:
8521       document broken shell
8522     * [310]16406 USE_LD_AS_NEEDED undocumented
8523     __________________________________________________________________
8524
8525GCC 3.4.4
8526
8527   This is the [311]list of problem reports (PRs) from GCC's bug tracking
8528   system that are known to be fixed in the 3.4.4 release. This list might
8529   not be complete (that is, it is possible that some PRs that have been
8530   fixed are not listed here).
8531     __________________________________________________________________
8532
8533GCC 3.4.5
8534
8535   This is the [312]list of problem reports (PRs) from GCC's bug tracking
8536   system that are known to be fixed in the 3.4.5 release. This list might
8537   not be complete (that is, it is possible that some PRs that have been
8538   fixed are not listed here).
8539
8540    Bootstrap issues
8541
8542     * [313]24688 sco_math fixincl breaks math.h
8543
8544    C compiler bugs
8545
8546     * [314]17188 struct Foo { } redefinition
8547     * [315]20187 wrong code for ((unsigned char)(unsigned long
8548       long)((a?a:1)&(a*b)))?0:1)
8549     * [316]21873 infinite warning loop on bad array initializer
8550     * [317]21899 enum definition accepts values to be overriden
8551     * [318]22061 ICE in find_function_data, in function.c
8552     * [319]22308 Failure to diagnose violation of constraint 6.516p2
8553     * [320]22458 ICE on missing brace
8554     * [321]22589 ICE casting to long long
8555     * [322]24101 Segfault with preprocessed source
8556
8557    C++ compiler and library bugs
8558
8559     * [323]10611 operations on vector mode not recognized in C++
8560     * [324]13377 unexpected behavior of namespace usage directive
8561     * [325]16002 Strange error message with new parser
8562     * [326]17413 local classes as template argument
8563     * [327]17609 spurious error message after using keyword
8564     * [328]17618 ICE in cp_convert_to_pointer, in cp/cvt.c
8565     * [329]18124 ICE with invalid template template parameter
8566     * [330]18155 typedef in template declaration not rejected
8567     * [331]18177 ICE with const_cast for undeclared variable
8568     * [332]18368 C++ error message regression
8569     * [333]16378 ICE when returning a copy of a packed member
8570     * [334]18466 int ::i; accepted
8571     * [335]18512 ICE on invalid usage of template base class
8572     * [336]18454 ICE when returning undefined type
8573     * [337]18738 typename not allowed with non-dependent qualified name
8574     * [338]18803 rejects access to operator() in template
8575     * [339]19004 ICE in uses_template_parms, in cp/pt.c
8576     * [340]19208 Spurious error about variably modified type
8577     * [341]18253 bad error message / ICE for invalid template parameter
8578     * [342]19608 ICE after friend function definition in local class
8579     * [343]19884 ICE on explicit instantiation of a non-template
8580       constructor
8581     * [344]20153 ICE when C++ template function contains anonymous union
8582     * [345]20563 Infinite loop in diagnostic (and ice after error
8583       message)
8584     * [346]20789 ICE with incomplete type in template
8585     * [347]21336 Internal compiler error when using custom new operators
8586     * [348]21768 ICE in error message due to violation of coding
8587       conventions
8588     * [349]21853 constness of pointer to data member ignored
8589     * [350]21903 Default argument of template function causes a
8590       compile-time error
8591     * [351]21983 multiple diagnostics
8592     * [352]21987 New testsuite failure
8593       g++.dg/warn/conversion-function-1.C
8594     * [353]22153 ICE on invalid template specialization
8595     * [354]22172 Internal compiler error, seg fault.
8596     * [355]21286 filebuf::xsgetn vs pipes
8597     * [356]22233 ICE with wrong number of template parameters
8598     * [357]22508 ICE after invalid operator new
8599     * [358]22545 ICE with pointer to class member & user defined
8600       conversion operator
8601     * [359]23528 Wrong default allocator in ext/hash_map
8602     * [360]23550 char_traits requirements/1.cc test bad math
8603     * [361]23586 Bad diagnostic for invalid namespace-name
8604     * [362]23624 ICE in invert_truthvalue, in fold-const.c
8605     * [363]23639 Bad error message: not a member of '<declaration error>'
8606     * [364]23797 ICE on typename outside template
8607     * [365]23965 Bogus error message: no matching function for call to
8608       'foo(<type error>)'
8609     * [366]24052 &#`label_decl' not supported by dump_expr#<expression
8610       error>
8611     * [367]24580 virtual base class cause exception not to be caught
8612
8613    Problems in generated debug information
8614
8615     * [368]24267 Bad DWARF for altivec vectors
8616
8617    Optimizations issues
8618
8619     * [369]17810 ICE in verify_local_live_at_start
8620     * [370]17860 Wrong generated code for loop with varying bound
8621     * [371]21709 ICE on compile-time complex NaN
8622     * [372]21964 broken tail call at -O2 or more
8623     * [373]22167 Strange optimization bug when using -Os
8624     * [374]22619 Compilation failure for real_const_1.f and
8625       real_const_2.f90
8626     * [375]23241 Invalid code generated for comparison of uchar to 255
8627     * [376]23478 Miscompilation due to reloading of a var that is also
8628       used in EH pad
8629     * [377]24470 segmentation fault in cc1plus when compiling with -O
8630     * [378]24950 ICE in operand_subword_force
8631
8632    Precompiled headers problems
8633
8634     * [379]14400 Cannot compile qt-x11-free-3.3.0
8635     * [380]14940 PCH largefile test fails on various platforms
8636
8637    Preprocessor bugs
8638
8639     * [381]20239 ICE on empty preprocessed input
8640     * [382]15220 "gcc -E -MM -MG" reports missing system headers in
8641       source directory
8642
8643    Testsuite issues
8644
8645     * [383]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on
8646       i686-pc-linux-gnu
8647
8648    Alpha specific
8649
8650     * [384]21888 bootstrap failure with linker relaxation enabled
8651
8652    ARM specific
8653
8654     * [385]15342 [arm-linux]: ICE in verify_local_live_at_start
8655     * [386]23985 Memory aliasing information incorrect in inlined memcpy
8656
8657    ColdFile specific
8658
8659     * [387]16719 Illegal move of byte into address register causes
8660       compiler to ICE
8661
8662    HPPA specific
8663
8664     * [388]21723 ICE while building libgfortran
8665     * [389]21841 -mhp-ld/-mgnu-ld documentation
8666
8667    IA-64 specific
8668
8669     * [390]23644 IA-64 hardware models and configuration options
8670       documentation error
8671     * [391]24718 Shared libgcc not used for linking by default
8672
8673    M68000 specific
8674
8675     * [392]18421 ICE in reload_cse_simplify_operands, in postreload.c
8676
8677    MIPS specific
8678
8679     * [393]20621 ICE in change_address_1, in emit-rtl.c
8680
8681    PowerPC and PowerPC64 specific
8682
8683     * [394]18583 error on valid code: const
8684       __attribute__((altivec(vector__))) doesn't work in arrays
8685     * [395]20191 ICE in reload_cse_simplify_operands
8686     * [396]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined
8687     * [397]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set
8688     * [398]23404 gij trashes args of functions with more than 8 fp args
8689     * [399]23539 C & C++ compiler generating misaligned references
8690       regardless of compiler flags
8691     * [400]24102 floatdisf2_internal2 broken
8692     * [401]24465 -mminimal-toc miscompilation of __thread vars
8693
8694    Solaris specific
8695
8696     * [402]19933 Problem with define of HUGE_VAL in math_c99
8697     * [403]21889 Native Solaris assembler cannot grok DTP-relative debug
8698       symbols
8699
8700    SPARC specific
8701
8702     * [404]19300 PCH failures on sparc-linux
8703     * [405]20301 Assembler labels have a leading "-"
8704     * [406]20673 C PCH testsuite assembly comparison failure
8705
8706    x86 and x86_64 specific
8707
8708     * [407]18582 ICE with arrays of type V2DF
8709     * [408]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2
8710       -fsched2-use-traces
8711     * [409]21716 ICE in reg-stack.c's swap_rtx_condition
8712     * [410]24315 amd64 fails -fpeephole2
8713     __________________________________________________________________
8714
8715GCC 3.4.6
8716
8717   This is the [411]list of problem reports (PRs) from GCC's bug tracking
8718   system that are known to be fixed in the 3.4.6 release. This list might
8719   not be complete (that is, it is possible that some PRs that have been
8720   fixed are not listed here).
8721
8722
8723    For questions related to the use of GCC, please consult these web
8724    pages and the [412]GCC manuals. If that fails, the
8725    [413]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8726    web pages and the development of GCC are welcome on our developer
8727    list at [414]gcc@gcc.gnu.org. All of [415]our lists have public
8728    archives.
8729
8730   Copyright (C) [416]Free Software Foundation, Inc. Verbatim copying and
8731   distribution of this entire article is permitted in any medium,
8732   provided this notice is preserved.
8733
8734   These pages are [417]maintained by the GCC team. Last modified
8735   2012-11-02[418].
8736
8737References
8738
8739   1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
8740   2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus
8741   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
8742   4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems
8743   5. http://gcc.gnu.org/gcc-3.4/mips-abi.html
8744   6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
8745   7. http://www.boost.org/
8746   8. http://gcc.gnu.org/PR11953
8747   9. http://gcc.gnu.org/PR8361
8748  10. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins
8749  11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209
8750  12. http://gcc.gnu.org/bugs/#cxx_rvalbind
8751  13. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
8752  14. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
8753  15. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
8754  16. http://www.gnu.org/software/classpath/
8755  17. http://www.eclipse.org/
8756  18. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html
8757  19. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html
8758  20. http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51A_HTML/ARH9MBTE/DTMNPLTN.HTM#normal-argument-list-structure
8759  21. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
8760  22. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html
8761  23. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
8762  24. http://gcc.gnu.org/gcc-3.4/mips-abi.html
8763  25. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html
8764  26. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
8765  27. http://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
8766  28. http://gcc.gnu.org/PR10129
8767  29. http://gcc.gnu.org/PR14576
8768  30. http://gcc.gnu.org/PR14760
8769  31. http://gcc.gnu.org/PR14671
8770  32. http://gcc.gnu.org/PR15093
8771  33. http://gcc.gnu.org/PR15178
8772  34. http://gcc.gnu.org/PR12753
8773  35. http://gcc.gnu.org/PR13985
8774  36. http://gcc.gnu.org/PR14810
8775  37. http://gcc.gnu.org/PR14883
8776  38. http://gcc.gnu.org/PR15044
8777  39. http://gcc.gnu.org/PR15057
8778  40. http://gcc.gnu.org/PR15064
8779  41. http://gcc.gnu.org/PR15142
8780  42. http://gcc.gnu.org/PR15159
8781  43. http://gcc.gnu.org/PR15165
8782  44. http://gcc.gnu.org/PR15193
8783  45. http://gcc.gnu.org/PR15209
8784  46. http://gcc.gnu.org/PR15227
8785  47. http://gcc.gnu.org/PR15285
8786  48. http://gcc.gnu.org/PR15299
8787  49. http://gcc.gnu.org/PR15329
8788  50. http://gcc.gnu.org/PR15550
8789  51. http://gcc.gnu.org/PR15554
8790  52. http://gcc.gnu.org/PR15640
8791  53. http://gcc.gnu.org/PR15666
8792  54. http://gcc.gnu.org/PR15696
8793  55. http://gcc.gnu.org/PR15701
8794  56. http://gcc.gnu.org/PR15761
8795  57. http://gcc.gnu.org/PR15829
8796  58. http://gcc.gnu.org/PR14538
8797  59. http://gcc.gnu.org/PR12391
8798  60. http://gcc.gnu.org/PR14649
8799  61. http://gcc.gnu.org/PR15004
8800  62. http://gcc.gnu.org/PR15749
8801  63. http://gcc.gnu.org/PR10646
8802  64. http://gcc.gnu.org/PR12077
8803  65. http://gcc.gnu.org/PR13598
8804  66. http://gcc.gnu.org/PR14211
8805  67. http://gcc.gnu.org/PR14220
8806  68. http://gcc.gnu.org/PR14245
8807  69. http://gcc.gnu.org/PR14340
8808  70. http://gcc.gnu.org/PR14600
8809  71. http://gcc.gnu.org/PR14668
8810  72. http://gcc.gnu.org/PR14775
8811  73. http://gcc.gnu.org/PR14821
8812  74. http://gcc.gnu.org/PR14930
8813  75. http://gcc.gnu.org/PR14932
8814  76. http://gcc.gnu.org/PR14950
8815  77. http://gcc.gnu.org/PR14962
8816  78. http://gcc.gnu.org/PR14975
8817  79. http://gcc.gnu.org/PR15002
8818  80. http://gcc.gnu.org/PR15025
8819  81. http://gcc.gnu.org/PR15046
8820  82. http://gcc.gnu.org/PR15069
8821  83. http://gcc.gnu.org/PR15074
8822  84. http://gcc.gnu.org/PR15083
8823  85. http://gcc.gnu.org/PR15096
8824  86. http://gcc.gnu.org/PR15287
8825  87. http://gcc.gnu.org/PR15317
8826  88. http://gcc.gnu.org/PR15337
8827  89. http://gcc.gnu.org/PR15361
8828  90. http://gcc.gnu.org/PR15412
8829  91. http://gcc.gnu.org/PR15427
8830  92. http://gcc.gnu.org/PR15471
8831  93. http://gcc.gnu.org/PR15503
8832  94. http://gcc.gnu.org/PR15507
8833  95. http://gcc.gnu.org/PR15542
8834  96. http://gcc.gnu.org/PR15565
8835  97. http://gcc.gnu.org/PR15625
8836  98. http://gcc.gnu.org/PR15629
8837  99. http://gcc.gnu.org/PR15742
8838 100. http://gcc.gnu.org/PR15775
8839 101. http://gcc.gnu.org/PR15821
8840 102. http://gcc.gnu.org/PR15862
8841 103. http://gcc.gnu.org/PR15875
8842 104. http://gcc.gnu.org/PR15877
8843 105. http://gcc.gnu.org/PR15947
8844 106. http://gcc.gnu.org/PR16020
8845 107. http://gcc.gnu.org/PR16154
8846 108. http://gcc.gnu.org/PR16174
8847 109. http://gcc.gnu.org/PR14315
8848 110. http://gcc.gnu.org/PR15151
8849 111. http://gcc.gnu.org/PR7993
8850 112. http://gcc.gnu.org/PR15228
8851 113. http://gcc.gnu.org/PR15345
8852 114. http://gcc.gnu.org/PR15945
8853 115. http://gcc.gnu.org/PR15526
8854 116. http://gcc.gnu.org/PR14690
8855 117. http://gcc.gnu.org/PR15112
8856 118. http://gcc.gnu.org/PR15067
8857 119. http://gcc.gnu.org/PR1963
8858 120. http://gcc.gnu.org/PR15717
8859 121. http://gcc.gnu.org/PR14782
8860 122. http://gcc.gnu.org/PR14828
8861 123. http://gcc.gnu.org/PR15202
8862 124. http://gcc.gnu.org/PR14610
8863 125. http://gcc.gnu.org/PR14813
8864 126. http://gcc.gnu.org/PR14857
8865 127. http://gcc.gnu.org/PR15598
8866 128. http://gcc.gnu.org/PR15653
8867 129. http://gcc.gnu.org/PR15189
8868 130. http://gcc.gnu.org/PR15331
8869 131. http://gcc.gnu.org/PR16144
8870 132. http://gcc.gnu.org/PR16176
8871 133. http://gcc.gnu.org/PR11591
8872 134. http://gcc.gnu.org/PR12028
8873 135. http://gcc.gnu.org/PR14478
8874 136. http://gcc.gnu.org/PR14567
8875 137. http://gcc.gnu.org/PR14715
8876 138. http://gcc.gnu.org/PR14902
8877 139. http://gcc.gnu.org/PR14924
8878 140. http://gcc.gnu.org/PR14960
8879 141. http://gcc.gnu.org/PR15106
8880 142. http://gcc.gnu.org/PR16026
8881 143. http://gcc.gnu.org/PR15191
8882 144. http://gcc.gnu.org/PR15662
8883 145. http://gcc.gnu.org/PR15054
8884 146. http://gcc.gnu.org/PR15783
8885 147. http://gcc.gnu.org/PR15626
8886 148. http://gcc.gnu.org/PR14326
8887 149. http://gcc.gnu.org/PR14723
8888 150. http://gcc.gnu.org/PR15290
8889 151. http://gcc.gnu.org/PR15250
8890 152. http://gcc.gnu.org/PR15551
8891 153. http://gcc.gnu.org/PR8309
8892 154. http://gcc.gnu.org/PR13250
8893 155. http://gcc.gnu.org/PR13803
8894 156. http://gcc.gnu.org/PR14093
8895 157. http://gcc.gnu.org/PR14457
8896 158. http://gcc.gnu.org/PR14542
8897 159. http://gcc.gnu.org/PR15100
8898 160. http://gcc.gnu.org/PR15296
8899 161. http://gcc.gnu.org/PR15396
8900 162. http://gcc.gnu.org/PR15782
8901 163. http://gcc.gnu.org/PR11610
8902 164. http://gcc.gnu.org/PR15488
8903 165. http://gcc.gnu.org/PR15489
8904 166. http://gcc.gnu.org/PR13928
8905 167. http://gcc.gnu.org/PR14150
8906 168. http://gcc.gnu.org/PR14949
8907 169. http://gcc.gnu.org/PR15123
8908 170. http://gcc.gnu.org/PR16469
8909 171. http://gcc.gnu.org/PR16344
8910 172. http://gcc.gnu.org/PR16842
8911 173. http://gcc.gnu.org/PR12608
8912 174. http://gcc.gnu.org/PR14492
8913 175. http://gcc.gnu.org/PR15461
8914 176. http://gcc.gnu.org/PR15890
8915 177. http://gcc.gnu.org/PR16180
8916 178. http://gcc.gnu.org/PR16224
8917 179. http://gcc.gnu.org/PR16408
8918 180. http://gcc.gnu.org/PR16529
8919 181. http://gcc.gnu.org/PR16698
8920 182. http://gcc.gnu.org/PR16706
8921 183. http://gcc.gnu.org/PR16810
8922 184. http://gcc.gnu.org/PR16851
8923 185. http://gcc.gnu.org/PR16870
8924 186. http://gcc.gnu.org/PR16904
8925 187. http://gcc.gnu.org/PR16905
8926 188. http://gcc.gnu.org/PR16964
8927 189. http://gcc.gnu.org/PR17068
8928 190. http://gcc.gnu.org/PR16366
8929 191. http://gcc.gnu.org/PR15345
8930 192. http://gcc.gnu.org/PR16590
8931 193. http://gcc.gnu.org/PR16693
8932 194. http://gcc.gnu.org/PR17078
8933 195. http://gcc.gnu.org/PR13956
8934 196. http://gcc.gnu.org/PR16684
8935 197. http://gcc.gnu.org/PR12658
8936 198. http://gcc.gnu.org/PR13092
8937 199. http://gcc.gnu.org/PR15320
8938 200. http://gcc.gnu.org/PR16246
8939 201. http://gcc.gnu.org/PR16273
8940 202. http://gcc.gnu.org/PR16401
8941 203. http://gcc.gnu.org/PR16411
8942 204. http://gcc.gnu.org/PR16489
8943 205. http://gcc.gnu.org/PR16618
8944 206. http://gcc.gnu.org/PR16637
8945 207. http://gcc.gnu.org/PR16717
8946 208. http://gcc.gnu.org/PR16813
8947 209. http://gcc.gnu.org/PR16853
8948 210. http://gcc.gnu.org/PR16889
8949 211. http://gcc.gnu.org/PR16959
8950 212. http://gcc.gnu.org/PR7587
8951 213. http://gcc.gnu.org/PR16473
8952 214. http://gcc.gnu.org/PR16478
8953 215. http://gcc.gnu.org/PR10695
8954 216. http://gcc.gnu.org/PR16974
8955 217. http://gcc.gnu.org/PR16298
8956 218. http://gcc.gnu.org/PR17113
8957 219. http://gcc.gnu.org/PR14697
8958 220. http://gcc.gnu.org/PR15869
8959 221. http://gcc.gnu.org/PR16325
8960 222. http://gcc.gnu.org/PR16357
8961 223. http://gcc.gnu.org/PR16380
8962 224. http://gcc.gnu.org/PR16407
8963 225. http://gcc.gnu.org/PR16643
8964 226. http://gcc.gnu.org/PR15927
8965 227. http://gcc.gnu.org/PR15948
8966 228. http://gcc.gnu.org/PR17019
8967 229. http://gcc.gnu.org/PR16130
8968 230. http://gcc.gnu.org/PR16142
8969 231. http://gcc.gnu.org/PR16278
8970 232. http://gcc.gnu.org/PR16414
8971 233. http://gcc.gnu.org/PR16445
8972 234. http://gcc.gnu.org/PR16490
8973 235. http://gcc.gnu.org/PR16683
8974 236. http://gcc.gnu.org/PR16195
8975 237. http://gcc.gnu.org/PR16239
8976 238. http://gcc.gnu.org/PR16199
8977 239. http://gcc.gnu.org/PR16416
8978 240. http://gcc.gnu.org/PR16430
8979 241. http://gcc.gnu.org/PR16379
8980 242. http://gcc.gnu.org/PR17093
8981 243. http://gcc.gnu.org/PR17119
8982 244. http://gcc.gnu.org/PR15928
8983 245. http://gcc.gnu.org/PR16210
8984 246. http://gcc.gnu.org/PR15488
8985 247. http://gcc.gnu.org/PR16250
8986 248. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3
8987 249. http://gcc.gnu.org/PR17369
8988 250. http://gcc.gnu.org/PR17850
8989 251. http://gcc.gnu.org/PR13948
8990 252. http://gcc.gnu.org/PR14492
8991 253. http://gcc.gnu.org/PR16301
8992 254. http://gcc.gnu.org/PR16566
8993 255. http://gcc.gnu.org/PR17023
8994 256. http://gcc.gnu.org/PR17027
8995 257. http://gcc.gnu.org/PR17524
8996 258. http://gcc.gnu.org/PR17826
8997 259. http://gcc.gnu.org/PR15526
8998 260. http://gcc.gnu.org/PR16999
8999 261. http://gcc.gnu.org/PR17503
9000 262. http://gcc.gnu.org/PR17581
9001 263. http://gcc.gnu.org/PR18129
9002 264. http://gcc.gnu.org/PR10975
9003 265. http://gcc.gnu.org/PR11722
9004 266. http://gcc.gnu.org/PR14534
9005 267. http://gcc.gnu.org/PR15172
9006 268. http://gcc.gnu.org/PR15786
9007 269. http://gcc.gnu.org/PR16162
9008 270. http://gcc.gnu.org/PR16612
9009 271. http://gcc.gnu.org/PR16715
9010 272. http://gcc.gnu.org/PR16848
9011 273. http://gcc.gnu.org/PR17132
9012 274. http://gcc.gnu.org/PR17259
9013 275. http://gcc.gnu.org/PR17327
9014 276. http://gcc.gnu.org/PR17393
9015 277. http://gcc.gnu.org/PR17501
9016 278. http://gcc.gnu.org/PR17537
9017 279. http://gcc.gnu.org/PR17585
9018 280. http://gcc.gnu.org/PR17821
9019 281. http://gcc.gnu.org/PR17829
9020 282. http://gcc.gnu.org/PR17851
9021 283. http://gcc.gnu.org/PR17976
9022 284. http://gcc.gnu.org/PR18020
9023 285. http://gcc.gnu.org/PR18093
9024 286. http://gcc.gnu.org/PR18140
9025 287. http://gcc.gnu.org/PR17541
9026 288. http://gcc.gnu.org/PR17853
9027 289. http://gcc.gnu.org/PR17245
9028 290. http://gcc.gnu.org/PR17167
9029 291. http://gcc.gnu.org/PR17277
9030 292. http://gcc.gnu.org/PR17505
9031 293. http://gcc.gnu.org/PR17684
9032 294. http://gcc.gnu.org/PR17384
9033 295. http://gcc.gnu.org/PR17770
9034 296. http://gcc.gnu.org/PR11476
9035 297. http://gcc.gnu.org/PR14064
9036 298. http://gcc.gnu.org/PR14678
9037 299. http://gcc.gnu.org/PR15583
9038 300. http://gcc.gnu.org/PR15790
9039 301. http://gcc.gnu.org/PR15886
9040 302. http://gcc.gnu.org/PR16884
9041 303. http://gcc.gnu.org/PR13841
9042 304. http://gcc.gnu.org/PR15860
9043 305. http://gcc.gnu.org/PR17465
9044 306. http://gcc.gnu.org/PR17469
9045 307. http://gcc.gnu.org/PR18138
9046 308. http://gcc.gnu.org/PR15498
9047 309. http://gcc.gnu.org/PR15747
9048 310. http://gcc.gnu.org/PR16406
9049 311. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4
9050 312. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5
9051 313. http://gcc.gnu.org/PR24688
9052 314. http://gcc.gnu.org/PR17188
9053 315. http://gcc.gnu.org/PR20187
9054 316. http://gcc.gnu.org/PR21873
9055 317. http://gcc.gnu.org/PR21899
9056 318. http://gcc.gnu.org/PR22061
9057 319. http://gcc.gnu.org/PR22208
9058 320. http://gcc.gnu.org/PR22458
9059 321. http://gcc.gnu.org/PR22589
9060 322. http://gcc.gnu.org/PR24101
9061 323. http://gcc.gnu.org/PR10611
9062 324. http://gcc.gnu.org/PR13377
9063 325. http://gcc.gnu.org/PR16002
9064 326. http://gcc.gnu.org/PR17413
9065 327. http://gcc.gnu.org/PR17609
9066 328. http://gcc.gnu.org/PR17618
9067 329. http://gcc.gnu.org/PR18124
9068 330. http://gcc.gnu.org/PR18155
9069 331. http://gcc.gnu.org/PR18177
9070 332. http://gcc.gnu.org/PR18368
9071 333. http://gcc.gnu.org/PR18378
9072 334. http://gcc.gnu.org/PR18466
9073 335. http://gcc.gnu.org/PR18512
9074 336. http://gcc.gnu.org/PR18545
9075 337. http://gcc.gnu.org/PR18738
9076 338. http://gcc.gnu.org/PR18803
9077 339. http://gcc.gnu.org/PR19004
9078 340. http://gcc.gnu.org/PR19208
9079 341. http://gcc.gnu.org/PR19253
9080 342. http://gcc.gnu.org/PR19608
9081 343. http://gcc.gnu.org/PR19884
9082 344. http://gcc.gnu.org/PR20153
9083 345. http://gcc.gnu.org/PR20563
9084 346. http://gcc.gnu.org/PR20789
9085 347. http://gcc.gnu.org/PR21336
9086 348. http://gcc.gnu.org/PR21768
9087 349. http://gcc.gnu.org/PR21853
9088 350. http://gcc.gnu.org/PR21903
9089 351. http://gcc.gnu.org/PR21983
9090 352. http://gcc.gnu.org/PR21987
9091 353. http://gcc.gnu.org/PR22153
9092 354. http://gcc.gnu.org/PR22172
9093 355. http://gcc.gnu.org/PR21286
9094 356. http://gcc.gnu.org/PR22233
9095 357. http://gcc.gnu.org/PR22508
9096 358. http://gcc.gnu.org/PR22545
9097 359. http://gcc.gnu.org/PR23528
9098 360. http://gcc.gnu.org/PR23550
9099 361. http://gcc.gnu.org/PR23586
9100 362. http://gcc.gnu.org/PR23624
9101 363. http://gcc.gnu.org/PR23639
9102 364. http://gcc.gnu.org/PR23797
9103 365. http://gcc.gnu.org/PR23965
9104 366. http://gcc.gnu.org/PR24052
9105 367. http://gcc.gnu.org/PR24580
9106 368. http://gcc.gnu.org/PR24267
9107 369. http://gcc.gnu.org/PR17810
9108 370. http://gcc.gnu.org/PR17860
9109 371. http://gcc/gnu.org/PR21709
9110 372. http://gcc.gnu.org/PR21964
9111 373. http://gcc.gnu.org/PR22167
9112 374. http://gcc.gnu.org/PR22619
9113 375. http://gcc.gnu.org/PR23241
9114 376. http://gcc.gnu.org/PR23478
9115 377. http://gcc.gnu.org/PR24470
9116 378. http://gcc.gnu.org/PR24950
9117 379. http://gcc.gnu.org/PR14400
9118 380. http://gcc.gnu.org/PR14940
9119 381. http://gcc.gnu.org/PR20239
9120 382. http://gcc.gnu.org/PR15220
9121 383. http://gcc.gnu.org/PR19275
9122 384. http://gcc.gnu.org/PR21888
9123 385. http://gcc.gnu.org/PR15342
9124 386. http://gcc.gnu.org/PR23985
9125 387. http://gcc.gnu.org/PR16719
9126 388. http://gcc.gnu.org/PR21723
9127 389. http://gcc.gnu.org/PR21841
9128 390. http://gcc.gnu.org/PR23644
9129 391. http://gcc.gnu.org/PR24718
9130 392. http://gcc.gnu.org/PR18421
9131 393. http://gcc.gnu.org/PR20621
9132 394. http://gcc.gnu.org/PR18583
9133 395. http://gcc.gnu.org/PR20191
9134 396. http://gcc.gnu.org/PR22083
9135 397. http://gcc.gnu.org/PR23070
9136 398. http://gcc.gnu.org/PR23404
9137 399. http://gcc.gnu.org/PR23539
9138 400. http://gcc.gnu.org/PR24102
9139 401. http://gcc.gnu.org/PR24465
9140 402. http://gcc.gnu.org/PR19933
9141 403. http://gcc.gnu.org/PR21889
9142 404. http://gcc.gnu.org/PR19300
9143 405. http://gcc.gnu.org/PR20301
9144 406. http://gcc.gnu.org/PR20673
9145 407. http://gcc.gnu.org/PR18582
9146 408. http://gcc.gnu.org/PR19340
9147 409. http://gcc.gnu.org/PR21716
9148 410. http://gcc.gnu.org/PR24315
9149 411. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6
9150 412. http://gcc.gnu.org/onlinedocs/
9151 413. mailto:gcc-help@gcc.gnu.org
9152 414. mailto:gcc@gcc.gnu.org
9153 415. http://gcc.gnu.org/lists.html
9154 416. http://www.fsf.org/
9155 417. http://gcc.gnu.org/about.html
9156 418. http://validator.w3.org/check/referer
9157======================================================================
9158http://gcc.gnu.org/gcc-3.3/index.html
9159
9160                             GCC 3.3 Release Series
9161
9162   May 03, 2005
9163
9164   The [1]GNU project and the GCC developers are pleased to announce the
9165   release of GCC 3.3.6.
9166
9167   This release is a bug-fix release, containing fixes for regressions in
9168   GCC 3.3.5 relative to previous releases of GCC.
9169
9170   This release is the last of the series 3.3.x.
9171
9172   The GCC 3.3 release series includes numerous [2]new features,
9173   improvements, bug fixes, and other changes, thanks to an [3]amazing
9174   group of volunteers.
9175
9176Release History
9177
9178   GCC 3.3.6
9179          May 3, 2005 ([4]changes)
9180
9181   GCC 3.3.5
9182          September 30, 2004 ([5]changes)
9183
9184   GCC 3.3.4
9185          May 31, 2004 ([6]changes)
9186
9187   GCC 3.3.3
9188          February 14, 2004 ([7]changes)
9189
9190   GCC 3.3.2
9191          October 16, 2003 ([8]changes)
9192
9193   GCC 3.3.1
9194          August 8, 2003 ([9]changes)
9195
9196   GCC 3.3
9197          May 14, 2003 ([10]changes)
9198
9199References and Acknowledgements
9200
9201   GCC used to stand for the GNU C Compiler, but since the compiler
9202   supports several other languages aside from C, it now stands for the
9203   GNU Compiler Collection.
9204
9205   A list of [11]successful builds is updated as new information becomes
9206   available.
9207
9208   The GCC developers would like to thank the numerous people that have
9209   contributed new features, improvements, bug fixes, and other changes as
9210   well as test results to GCC. This [12]amazing group of volunteers is
9211   what makes GCC successful.
9212
9213   For additional information about GCC please refer to the [13]GCC
9214   project web site or contact the [14]GCC development mailing list.
9215
9216   To obtain GCC please use [15]our mirror sites, or our CVS server.
9217
9218
9219    For questions related to the use of GCC, please consult these web
9220    pages and the [16]GCC manuals. If that fails, the
9221    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9222    web pages and the development of GCC are welcome on our developer
9223    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
9224    archives.
9225
9226   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
9227   distribution of this entire article is permitted in any medium,
9228   provided this notice is preserved.
9229
9230   These pages are [21]maintained by the GCC team. Last modified
9231   2012-11-02[22].
9232
9233References
9234
9235   1. http://www.gnu.org/
9236   2. http://gcc.gnu.org/gcc-3.3/changes.html
9237   3. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9238   4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
9239   5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5
9240   6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4
9241   7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3
9242   8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2
9243   9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1
9244  10. http://gcc.gnu.org/gcc-3.3/changes.html
9245  11. http://gcc.gnu.org/gcc-3.3/buildstat.html
9246  12. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9247  13. http://gcc.gnu.org/index.html
9248  14. mailto:gcc@gcc.gnu.org
9249  15. http://gcc.gnu.org/mirrors.html
9250  16. http://gcc.gnu.org/onlinedocs/
9251  17. mailto:gcc-help@gcc.gnu.org
9252  18. mailto:gcc@gcc.gnu.org
9253  19. http://gcc.gnu.org/lists.html
9254  20. http://www.fsf.org/
9255  21. http://gcc.gnu.org/about.html
9256  22. http://validator.w3.org/check/referer
9257======================================================================
9258http://gcc.gnu.org/gcc-3.3/changes.html
9259
9260                             GCC 3.3 Release Series
9261                        Changes, New Features, and Fixes
9262
9263   The latest release in the 3.3 release series is [1]GCC 3.3.6.
9264
9265Caveats
9266
9267     * The preprocessor no longer accepts multi-line string literals. They
9268       were deprecated in 3.0, 3.1, and 3.2.
9269     * The preprocessor no longer supports the -A- switch when appearing
9270       alone. -A- followed by an assertion is still supported.
9271     * Support for all the systems [2]obsoleted in GCC 3.1 has been
9272       removed from GCC 3.3. See below for a [3]list of systems which are
9273       obsoleted in this release.
9274     * Checking for null format arguments has been decoupled from the rest
9275       of the format checking mechanism. Programs which use the format
9276       attribute may regain this functionality by using the new [4]nonnull
9277       function attribute. Note that all functions for which GCC has a
9278       built-in format attribute, an appropriate built-in nonnull
9279       attribute is also applied.
9280     * The DWARF (version 1) debugging format has been deprecated and will
9281       be removed in a future version of GCC. Version 2 of the DWARF
9282       debugging format will continue to be supported for the foreseeable
9283       future.
9284     * The C and Objective-C compilers no longer accept the "Naming Types"
9285       extension (typedef foo = bar); it was already unavailable in C++.
9286       Code which uses it will need to be changed to use the "typeof"
9287       extension instead: typedef typeof(bar) foo. (We have removed this
9288       extension without a period of deprecation because it has caused the
9289       compiler to crash since version 3.0 and no one noticed until very
9290       recently. Thus we conclude it is not in widespread use.)
9291     * The -traditional C compiler option has been removed. It was
9292       deprecated in 3.1 and 3.2. (Traditional preprocessing remains
9293       available.) The <varargs.h> header, used for writing variadic
9294       functions in traditional C, still exists but will produce an error
9295       message if used.
9296     * GCC 3.3.1 automatically places zero-initialized variables in the
9297       .bss section on some operating systems. Versions of GNU Emacs up to
9298       (and including) 21.3 will not work correctly when using this
9299       optimization; you can use -fno-zero-initialized-in-bss to disable
9300       it.
9301
9302General Optimizer Improvements
9303
9304     * A new scheme for accurately describing processor pipelines, the
9305       [5]DFA scheduler, has been added.
9306     * Pavel Nejedly, Charles University Prague, has contributed new file
9307       format used by the edge coverage profiler (-fprofile-arcs).
9308       The new format is robust and diagnoses common mistakes where
9309       profiles from different versions (or compilations) of the program
9310       are combined resulting in nonsensical profiles and slow code to
9311       produced with profile feedback. Additionally this format allows
9312       extra data to be gathered. Currently, overall statistics are
9313       produced helping optimizers to identify hot spots of a program
9314       globally replacing the old intra-procedural scheme and resulting in
9315       better code. Note that the gcov tool from older GCC versions will
9316       not be able to parse the profiles generated by GCC 3.3 and vice
9317       versa.
9318     * Jan Hubicka, SuSE Labs, has contributed a new superblock formation
9319       pass enabled using -ftracer. This pass simplifies the control flow
9320       of functions allowing other optimizations to do better job.
9321       He also contributed the function reordering pass
9322       (-freorder-functions) to optimize function placement using profile
9323       feedback.
9324
9325New Languages and Language specific improvements
9326
9327  C/ObjC/C++
9328
9329     * The preprocessor now accepts directives within macro arguments. It
9330       processes them just as if they had not been within macro arguments.
9331     * The separate ISO and traditional preprocessors have been completely
9332       removed. The front end handles either type of preprocessed output
9333       if necessary.
9334     * In C99 mode preprocessor arithmetic is done in the precision of the
9335       target's intmax_t, as required by that standard.
9336     * The preprocessor can now copy comments inside macros to the output
9337       file when the macro is expanded. This feature, enabled using the
9338       -CC option, is intended for use by applications which place
9339       metadata or directives inside comments, such as lint.
9340     * The method of constructing the list of directories to be searched
9341       for header files has been revised. If a directory named by a -I
9342       option is a standard system include directory, the option is
9343       ignored to ensure that the default search order for system
9344       directories and the special treatment of system header files are
9345       not defeated.
9346     * A few more [6]ISO C99 features now work correctly.
9347     * A new function attribute, nonnull, has been added which allows
9348       pointer arguments to functions to be specified as requiring a
9349       non-null value. The compiler currently uses this information to
9350       issue a warning when it detects a null value passed in such an
9351       argument slot.
9352     * A new type attribute, may_alias, has been added. Accesses to
9353       objects with types with this attribute are not subjected to
9354       type-based alias analysis, but are instead assumed to be able to
9355       alias any other type of objects, just like the char type.
9356
9357  C++
9358
9359     * Type based alias analysis has been implemented for C++ aggregate
9360       types.
9361
9362  Objective-C
9363
9364     * Generate an error if Objective-C objects are passed by value in
9365       function and method calls.
9366     * When -Wselector is used, check the whole list of selectors at the
9367       end of compilation, and emit a warning if a @selector() is not
9368       known.
9369     * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
9370     * No longer need to include objc/objc-class.h to compile self calls
9371       in class methods (NeXT runtime only).
9372     * New -Wundeclared-selector option.
9373     * Removed selector bloating which was causing object files to be 10%
9374       bigger on average (GNU runtime only).
9375     * Using at run time @protocol() objects has been fixed in certain
9376       situations (GNU runtime only).
9377     * Type checking has been fixed and improved in many situations
9378       involving protocols.
9379
9380  Java
9381
9382     * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK
9383       1.4) API.
9384     * The JDK 1.4 assert facility has been implemented.
9385     * The bytecode interpreter is now direct threaded and thus faster.
9386
9387  Fortran
9388
9389     * Fortran improvements are listed in [7]the Fortran documentation.
9390
9391  Ada
9392
9393     * Ada tasking now works with glibc 2.3.x threading libraries.
9394
9395New Targets and Target Specific Improvements
9396
9397     * The following changes have been made to the HP-PA port:
9398          + The port now defaults to scheduling for the PA8000 series of
9399            processors.
9400          + Scheduling support for the PA7300 processor has been added.
9401          + The 32-bit port now supports weak symbols under HP-UX 11.
9402          + The handling of initializers and finalizers has been improved
9403            under HP-UX 11. The 64-bit port no longer uses collect2.
9404          + Dwarf2 EH support has been added to the 32-bit GNU/Linux port.
9405          + ABI fixes to correct the passing of small structures by value.
9406     * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to
9407       use the DFA processor pipeline description.
9408     * The following NetBSD configurations for the SuperH processor family
9409       have been added:
9410          + SH3, big-endian, sh-*-netbsdelf*
9411          + SH3, little-endian, shle-*-netbsdelf*
9412          + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
9413          + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
9414          + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
9415          + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
9416     * The following changes have been made to the IA-32/x86-64 port:
9417          + SSE2 and 3dNOW! intrinsics are now supported.
9418          + Support for thread local storage has been added to the IA-32
9419            and x86-64 ports.
9420          + The x86-64 port has been significantly improved.
9421     * The following changes have been made to the MIPS port:
9422          + All configurations now accept the -mabi switch. Note that you
9423            will need appropriate multilibs for this option to work
9424            properly.
9425          + ELF configurations will always pass an ABI flag to the
9426            assembler, except when the MIPS EABI is selected.
9427          + -mabi=64 no longer selects MIPS IV code.
9428          + The -mcpu option, which was deprecated in 3.1 and 3.2, has
9429            been removed from this release.
9430          + -march now changes the core ISA level. In previous releases,
9431            it would change the use of processor-specific extensions, but
9432            would leave the core ISA unchanged. For example, mips64-elf
9433            -march=r8000 will now generate MIPS IV code.
9434          + Under most configurations, -mipsN now acts as a synonym for
9435            -march.
9436          + There are some new preprocessor macros to describe the -march
9437            and -mtune settings. See the documentation of those options
9438            for details.
9439          + Support for the NEC VR-Series processors has been added. This
9440            includes the 54xx, 5500, and 41xx series.
9441          + Support for the Sandcraft sr71k processor has been added.
9442     * The following changes have been made to the S/390 port:
9443          + Support to build the Java runtime libraries has been added.
9444            Java is now enabled by default on s390-*-linux* and
9445            s390x-*-linux* targets.
9446          + Multilib support for the s390x-*-linux* target has been added;
9447            this allows to build 31-bit binaries using the -m31 option.
9448          + Support for thread local storage has been added.
9449          + Inline assembler code may now use the 'Q' constraint to
9450            specify memory operands without index register.
9451          + Various platform-specific performance improvements have been
9452            implemented; in particular, the compiler now uses the BRANCH
9453            ON COUNT family of instructions and makes more frequent use of
9454            the TEST UNDER MASK family of instructions.
9455     * The following changes have been made to the PowerPC port:
9456          + Support for IBM Power4 processor added.
9457          + Support for Motorola e500 SPE added.
9458          + Support for AIX 5.2 added.
9459          + Function and Data sections now supported on AIX.
9460          + Sibcall optimizations added.
9461     * The support for H8 Tiny is added to the H8/300 port with -mn.
9462
9463Obsolete Systems
9464
9465   Support for a number of older systems has been declared obsolete in GCC
9466   3.3. Unless there is activity to revive them, the next release of GCC
9467   will have their sources permanently removed.
9468
9469   All configurations of the following processor architectures have been
9470   declared obsolete:
9471     * Matsushita MN10200, mn10200-*-*
9472     * Motorola 88000, m88k-*-*
9473     * IBM ROMP, romp-*-*
9474
9475   Also, some individual systems have been obsoleted:
9476     * Alpha
9477          + Interix, alpha*-*-interix*
9478          + Linux libc1, alpha*-*-linux*libc1*
9479          + Linux ECOFF, alpha*-*-linux*ecoff*
9480     * ARM
9481          + Generic a.out, arm*-*-aout*
9482          + Conix, arm*-*-conix*
9483          + "Old ABI," arm*-*-oabi
9484          + StrongARM/COFF, strongarm-*-coff*
9485     * HPPA (PA-RISC)
9486          + Generic OSF, hppa1.0-*-osf*
9487          + Generic BSD, hppa1.0-*-bsd*
9488          + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
9489          + HiUX, hppa*-*-hiux*
9490          + Mach Lites, hppa*-*-lites*
9491     * Intel 386 family
9492          + Windows NT 3.x, i?86-*-win32
9493     * MC68000 family
9494          + HP systems, m68000-hp-bsd* and m68k-hp-bsd*
9495          + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and
9496            m68k-sun-mach*
9497          + AT&T systems, m68000-att-sysv*
9498          + Atari systems, m68k-atari-sysv*
9499          + Motorola systems, m68k-motorola-sysv*
9500          + NCR systems, m68k-ncr-sysv*
9501          + Plexus systems, m68k-plexus-sysv*
9502          + Commodore systems, m68k-cbm-sysv*
9503          + Citicorp TTI, m68k-tti-*
9504          + Unos, m68k-crds-unos*
9505          + Concurrent RTU, m68k-ccur-rtu*
9506          + Linux a.out, m68k-*-linux*aout*
9507          + Linux libc1, m68k-*-linux*libc1*
9508          + pSOS, m68k-*-psos*
9509     * MIPS
9510          + Generic ECOFF, mips*-*-ecoff*
9511          + SINIX, mips-sni-sysv4
9512          + Orion RTEMS, mips64orion-*-rtems*
9513     * National Semiconductor 32000
9514          + OpenBSD, ns32k-*-openbsd*
9515     * POWER (aka RS/6000) and PowerPC
9516          + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
9517          + Bull BOSX, rs6000-bull-bosx
9518          + Generic Mach, rs6000-*-mach*
9519          + Generic SysV, powerpc*-*-sysv*
9520          + Linux libc1, powerpc*-*-linux*libc1*
9521     * Sun SPARC
9522          + Generic a.out, sparc-*-aout*, sparclet-*-aout*,
9523            sparclite-*-aout*, and sparc86x-*-aout*
9524          + NetBSD a.out, sparc-*-netbsd*aout*
9525          + Generic BSD, sparc-*-bsd*
9526          + ChorusOS, sparc-*-chorusos*
9527          + Linux a.out, sparc-*-linux*aout*
9528          + Linux libc1, sparc-*-linux*libc1*
9529          + LynxOS, sparc-*-lynxos*
9530          + Solaris on HAL hardware, sparc-hal-solaris2*
9531          + SunOS versions 3 and 4, sparc-*-sunos[34]*
9532     * NEC V850
9533          + RTEMS, v850-*-rtems*
9534     * VAX
9535          + VMS, vax-*-vms*
9536
9537Documentation improvements
9538
9539Other significant improvements
9540
9541     * Almost all front-end dependencies in the compiler have been
9542       separated out into a set of language hooks. This should make adding
9543       a new front end clearer and easier.
9544     * One effect of removing the separate preprocessor is a small
9545       increase in the robustness of the compiler in general, and the
9546       maintainability of target descriptions. Previously target-specific
9547       built-in macros and others, such as __FAST_MATH__, had to be
9548       handled with so-called specs that were hard to maintain. Often they
9549       would fail to behave properly when conflicting options were
9550       supplied on the command line, and define macros in the user's
9551       namespace even when strict ISO compliance was requested.
9552       Integrating the preprocessor has cleanly solved these issues.
9553     * The Makefile suite now supports redirection of make install by
9554       means of the variable DESTDIR.
9555     __________________________________________________________________
9556
9557GCC 3.3
9558
9559   Detailed release notes for the GCC 3.3 release follow.
9560
9561  Bug Fixes
9562
9563    bootstrap failures
9564
9565     * [8]10140 cross compiler build failures: missing __mempcpy (DUP:
9566       [9]10198,[10]10338)
9567
9568    Internal compiler errors (multi-platform)
9569
9570     * [11]3581 large string causes segmentation fault in cc1
9571     * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler
9572     * [13]5533 (c++) ICE when processing std::accumulate(begin, end,
9573       init, invalid_op)
9574     * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
9575     * [15]6412 (c++) ICE in retrieve_specialization
9576     * [16]6620 (c++) partial template specialization causes an ICE
9577       (segmentation fault)
9578     * [17]6663 (c++) ICE with attribute aligned
9579     * [18]7068 ICE with incomplete types
9580     * [19]7083 (c++) ICE using -gstabs with dodgy class derivation
9581     * [20]7647 (c++) ICE when data member has the name of the enclosing
9582       class
9583     * [21]7675 ICE in fixup_var_refs_1
9584     * [22]7718 'complex' template instantiation causes ICE
9585     * [23]8116 (c++) ICE in member template function
9586     * [24]8358 (ada) Ada compiler accesses freed memory, crashes
9587     * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation
9588       fault
9589     * [26]8564 (c++) ICE in find_function_data, in function.c
9590     * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
9591     * [28]8766 (c++) ICE after failed initialization of static template
9592       variable
9593     * [29]8803 ICE in instantiate_virtual_regs_1, in function.c
9594     * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
9595     * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class
9596       definition
9597     * [32]9216 (c++) ICE on missing template parameter
9598     * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c
9599     * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO
9600       loop
9601     * [35]9429 (c++) ICE in template instantiation with a pointered new
9602       operator
9603     * [36]9516 Internal error when using a big array
9604     * [37]9600 (c++) ICE with typedefs in template class
9605     * [38]9629 (c++) virtual inheritance segfault
9606     * [39]9672 (c++) ICE: Error reporting routines re-entered
9607     * [40]9749 (c++) ICE in write_expression on invalid function
9608       prototype
9609     * [41]9794 (fortran) ICE: floating point exception during constant
9610       folding
9611     * [42]9829 (c++) Missing colon in nested namespace usage causes ICE
9612     * [43]9916 (c++) ICE with noreturn function in ?: statement
9613     * [44]9936 ICE with local function and variable-length 2d array
9614     * [45]10262 (c++) cc1plus crashes with large generated code
9615     * [46]10278 (c++) ICE in parser for invalid code
9616     * [47]10446 (c++) ICE on definition of nonexistent member function of
9617       nested class in a class template
9618     * [48]10451 (c++) ICE in grokdeclarator on spurious mutable
9619       declaration
9620     * [49]10506 (c++) ICE in build_new at cp/init.c with
9621       -fkeep-inline-functions and multiple inheritance
9622     * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the
9623       precision of the declared type
9624
9625    Optimization bugs
9626
9627     * [51]2001 Inordinately long compile times in reload CSE regs
9628     * [52]2391 Exponential compilation time explosion in combine
9629     * [53]2960 Duplicate loop conditions even with -Os
9630     * [54]4046 redundant conditional branch
9631     * [55]6405 Loop-unrolling related performance regressions
9632     * [56]6798 very long compile time with large case-statement
9633     * [57]6871 const objects shouldn't be moved to .bss
9634     * [58]6909 problem w/ -Os on modified loop-2c.c test case
9635     * [59]7189 gcc -O2 -Wall does not print ``control reaches end of
9636       non-void function'' warning
9637     * [60]7642 optimization problem with signbit()
9638     * [61]8634 incorrect code for inlining of memcpy under -O2
9639     * [62]8750 Cygwin prolog generation erroneously emitting __alloca as
9640       regular function call
9641
9642    C front end
9643
9644     * [63]2161 long if-else cascade overflows parser stack
9645     * [64]4319 short accepted on typedef'd char
9646     * [65]8602 incorrect line numbers in warning messages when using
9647       inline functions
9648     * [66]9177 -fdump-translation-unit: C front end deletes function_decl
9649       AST nodes and breaks debugging dumps
9650     * [67]9853 miscompilation of non-constant structure initializer
9651
9652    c++ compiler and library
9653
9654     * [68]45 legal template specialization code is rejected (DUP:
9655       [69]3784)
9656     * [70]764 lookup failure: friend operator and dereferencing a pointer
9657       and templates (DUP: [71]5116)
9658     * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP:
9659       2863)
9660     * [73]3663 G++ doesn't check access control during template
9661       instantiation
9662     * [74]3797 gcc fails to emit explicit specialization of a template
9663       member
9664     * [75]3948 Two destructors are called when no copy destructor is
9665       defined (ABI change)
9666     * [76]4137 Conversion operator within template is not accepted
9667     * [77]4361 bogus ambiguity taking the address of a member template
9668     * [78]4802 g++ accepts illegal template code (access to private
9669       member; DUP: [79]5837)
9670     * [80]4803 inline function is used but never defined, and g++ does
9671       not object
9672     * [81]5094 Partial specialization cannot be friend?
9673     * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
9674     * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run
9675       time
9676     * [84]7015 certain __asm__ constructs rejected
9677     * [85]7086 compile time regression (quadratic behavior in
9678       fixup_var_refs)
9679     * [86]7099 G++ doesn't set the noreturn attribute on std::exit and
9680       std::abort
9681     * [87]7247 copy constructor missing when inlining enabled (invalid
9682       optimization?)
9683     * [88]7441 string array initialization compilation time regression
9684       from seconds to minutes
9685     * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong
9686     * [90]7804 bad printing of floating point constant in warning message
9687     * [91]8099 Friend classes and template specializations
9688     * [92]8117 member function pointers and multiple inheritance
9689     * [93]8205 using declaration and multiple inheritance
9690     * [94]8645 unnecessary non-zero checks in stl_tree.h
9691     * [95]8724 explicit destructor call for incomplete class allowed
9692     * [96]8805 compile time regression with many member variables
9693     * [97]8691 -O3 and -fno-implicit-templates are incompatible
9694     * [98]8700 unhelpful error message for binding temp to reference
9695     * [99]8724 explicit destructor call for incomplete class allowed
9696     * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems
9697     * [101]9016 Failure to consistently constant fold "constant" C++
9698       objects
9699     * [102]9053 g++ confused about ambiguity of overloaded function
9700       templates
9701     * [103]9152 undefined virtual thunks
9702     * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out
9703     * [105]9297 data corruption due to codegen bug (when copying.)
9704     * [106]9318 i/ostream::operator>>/<<(streambuf*) broken
9705     * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf
9706     * [108]9400 bogus -Wshadow warning: shadowed declaration of this in
9707       local classes
9708     * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters
9709     * [110]9425 filebuf::pbackfail broken (DUP: [111]9439)
9710     * [112]9474 GCC freezes in compiling a weird code mixing <iostream>
9711       and <iostream.h>
9712     * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1)
9713       [114][DR 231]
9714     * [115]9555 ostream inserters fail to set badbit on exception
9715     * [116]9561 ostream inserters rethrow exception of wrong type
9716     * [117]9563 ostream::sentry returns true after a failed preparation
9717     * [118]9582 one-definition rule violation in std::allocator
9718     * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors
9719     * [120]9683 bug in initialization chains for static const variables
9720       from template classes
9721     * [121]9791 -Woverloaded-virtual reports hiding of destructor
9722     * [122]9817 collate::compare doesn't handle nul characters
9723     * [123]9825 filebuf::sputbackc breaks sbumpc
9724     * [124]9826 operator>>(basic_istream, basic_string) fails to compile
9725       with custom traits
9726     * [125]9924 Multiple using statements for builtin functions not
9727       allowed
9728     * [126]9946 destructor is not called for temporary object
9729     * [127]9964 filebuf::close() sometimes fails to close file
9730     * [128]9988 filebuf::overflow writes EOF to file
9731     * [129]10033 optimization breaks polymorphic references w/ typeid
9732       operator
9733     * [130]10097 filebuf::underflow drops characters
9734     * [131]10132 filebuf destructor can throw exceptions
9735     * [132]10180 gcc fails to warn about non-inlined function
9736     * [133]10199 method parametrized by template does not work everywhere
9737     * [134]10300 use of array-new (nothrow) in segfaults on NULL return
9738     * [135]10427 Stack corruption with variable-length automatic arrays
9739       and virtual destructors
9740     * [136]10503 Compilation never stops in fixed_type_or_null
9741
9742    Objective-C
9743
9744     * [137]5956 selectors aren't matched properly when added to the
9745       selector table
9746
9747    Fortran compiler and library
9748
9749     * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't
9750       detect
9751     * [139]3924 g77 generates code that is rejected by GAS if COFF debug
9752       info requested
9753     * [140]5634 doc: explain that configure --prefix=~/... does not work
9754     * [141]6367 multiple repeat counts confuse namelist read into array
9755     * [142]6491 Logical operations error on logicals when using
9756       -fugly-logint
9757     * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C"
9758     * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os
9759       on irix6.5
9760     * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should
9761       assume a direct access file
9762     * [146]7278 g77 "bug"; the executable misbehaves (with -O2
9763       -fno-automatic)
9764     * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows
9765     * [148]7388 Incorrect output with 0-based array of characters
9766     * [149]8587 Double complex zero ** double precision number -> NaN
9767       instead of zero
9768     * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning:
9769       unknown register name line-length-none
9770     * [151]10197 Direct access files not unformatted by default
9771
9772    Java compiler and library
9773
9774     * [152]6005 gcj fails to build rhug on alpha
9775     * [153]6389 System.getProperty("") should always throw an
9776       IllegalArgumentException
9777     * [154]6576 java.util.ResourceBundle.getResource ignores locale
9778     * [155]6652 new java.io.File("").getCanonicalFile() throws exception
9779     * [156]7060 getMethod() doesn't search super interface
9780     * [157]7073 bytecode interpreter gives wrong answer for interface
9781       getSuperclass()
9782     * [158]7180 possible bug in
9783       javax.naming.spi.NamingManager.getPlusPath()
9784     * [159]7416 java.security startup refs "GNU libgcj.security"
9785     * [160]7570 Runtime.exec with null envp: child doesn't inherit parent
9786       env (DUP: [161]7578)
9787     * [162]7611 Internal error while compiling libjava with -O
9788     * [163]7709 NullPointerException in _Jv_ResolvePoolEntry
9789     * [164]7766 ZipInputStream.available returns 0 immediately after
9790       construction
9791     * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public
9792     * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
9793     * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens'
9794     * [168]8234 ZipInputStream chokes when InputStream.read() returns
9795       small chunks
9796     * [169]8415 reflection bug: exception info for Method
9797     * [170]8481 java.Random.nextInt(int) may return negative
9798     * [171]8593 Error reading GZIPped files with BufferedReader
9799     * [172]8759 java.beans.Introspector has no flushCaches() or
9800       flushFromCaches() methods
9801     * [173]8997 spin() calls Thread.sleep
9802     * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd
9803       instead of the root content of C:
9804     * [175]9254 java::lang::Object::wait(), threads-win32.cc returns
9805       wrong return codes
9806     * [176]9271 Severe bias in java.security.SecureRandom
9807
9808    Ada compiler and library
9809
9810     * [177]6767 make gnatlib-shared fails on -laddr2line
9811     * [178]9911 gnatmake fails to link when GCC configured with
9812       --with-sjlj-exceptions=yes
9813     * [179]10020 Can't bootstrap gcc on AIX with Ada enabled
9814     * [180]10546 Ada tasking not working on Red Hat 9
9815
9816    preprocessor
9817
9818     * [181]7029 preprocessor should ignore #warning with -M
9819
9820    ARM-specific
9821
9822     * [182]2903 [arm] Optimization bug with long long arithmetic
9823     * [183]7873 arm-linux-gcc fails when assigning address to a bit field
9824
9825    FreeBSD-specific
9826
9827     * [184]7680 float functions undefined in math.h/cmath with #define
9828       _XOPEN_SOURCE
9829
9830    HP-UX or HP-PA-specific
9831
9832     * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
9833     * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to
9834       fputc_unlocked
9835     * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen
9836
9837    m68hc11-specific
9838
9839     * [188]6744 Bad assembler code generated: reference to pseudo
9840       register z
9841     * [189]7361 Internal compiler error in reload_cse_simplify_operands,
9842       in reload1.c
9843
9844    MIPS-specific
9845
9846     * [190]9496 [mips-linux] bug in optimizer?
9847
9848    PowerPC-specific
9849
9850     * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of
9851       space
9852     * [192]8480 reload ICEs for LAPACK code on powerpc64-linux
9853     * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg
9854     * [194]10315 [powerpc] ICE: in extract_insn, in recog.c
9855
9856    SPARC-specific
9857
9858     * [195]10267 (documentation) Wrong build instructions for
9859       *-*-solaris2*
9860
9861    x86-specific (Intel/AMD)
9862
9863     * [196]7916 ICE in instantiate_virtual_register_1
9864     * [197]7926 (c++) i486 instructions in header files make c++ programs
9865       crash on i386
9866     * [198]8555 ICE in gen_split_1231
9867     * [199]8994 ICE with -O -march=pentium4
9868     * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs
9869     * [201]9806 ICE in inline assembly with -fPIC flag
9870     * [202]10077 gcc -msse2 generates movd to move dwords between xmm
9871       regs
9872     * [203]10233 64-bit comparison only comparing bottom 32-bits
9873     * [204]10286 type-punning doesn't work with __m64 and -O
9874     * [205]10308 [x86] ICE with -O -fgcse or -O2
9875     __________________________________________________________________
9876
9877GCC 3.3.1
9878
9879  Bug Fixes
9880
9881   This section lists the problem reports (PRs) from GCC's bug tracking
9882   system that are known to be fixed in the 3.3.1 release. This list might
9883   not be complete (that is, it is possible that some PRs that have been
9884   fixed are not listed here).
9885
9886    Bootstrap failures
9887
9888     * [206]11272 [Solaris] make bootstrap fails while building libstdc++
9889
9890    Internal compiler errors (multi-platform)
9891
9892     * [207]5754 ICE on invalid nested template class
9893     * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64
9894       and --enable-checking
9895     * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c
9896     * [210]7053 (c++) ICE when declaring a function already defined as a
9897       friend method of a template class
9898     * [211]8164 (c++) ICE when using different const expressions as
9899       template parameter
9900     * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c
9901     * [213]9559 (c++) ICE with invalid initialization of a static const
9902     * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c
9903       when redeclaring a static member variable
9904     * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in
9905       dwarfout.c with -g -O -finline-functions
9906     * [216]10432 (c++) ICE in poplevel, in cp/decl.c
9907     * [217]10475 ICE in subreg_highpart_offset for code with long long
9908     * [218]10635 (c++) ICE when dereferencing an incomplete type casted
9909       from a void pointer
9910     * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while
9911       instantiating static member variables
9912     * [220]10700 ICE in copy_to_mode_reg on 64-bit targets
9913     * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c
9914     * [222]10796 (c++) ICE when defining an enum with two values: -1 and
9915       MAX_INT_64BIT
9916     * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x
9917       sched.c
9918     * [224]10939 (c++) ICE with template code
9919     * [225]10956 (c++) ICE when specializing a template member function
9920       of a template class, in tsubst, in cp/pt.c
9921     * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*()
9922       defined)
9923     * [227]11059 (c++) ICE with empty union
9924     * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with
9925       -O2 -fnon-call-exceptions
9926     * [229]11105 (c++) ICE in mangle_conv_op_name_for_type
9927     * [230]11149 (c++) ICE on error when instantiation with call function
9928       of a base type
9929     * [231]11228 (c++) ICE on new-expression using array operator new and
9930       default-initialization
9931     * [232]11282 (c++) Infinite memory usage after syntax error
9932     * [233]11301 (fortran) ICE with -fno-globals
9933     * [234]11308 (c++) ICE when using an enum type name as if it were a
9934       class or namespace
9935     * [235]11473 (c++) ICE with -gstabs when empty struct inherits from
9936       an empty struct
9937     * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR
9938     * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c:
9939       template member functions
9940
9941    Optimization bugs
9942
9943     * [238]11198 -O2 -frename-registers generates wrong code (aliasing
9944       problem)
9945     * [239]11304 Wrong code production with -fomit-frame-pointer
9946     * [240]11381 volatile memory access optimized away
9947     * [241]11536 [strength-reduce] -O2 optimization produces wrong code
9948     * [242]11557 constant folding bug generates wrong code
9949
9950    C front end
9951
9952     * [243]5897 No warning for statement after return
9953     * [244]11279 DWARF-2 output mishandles large enums
9954
9955    Preprocessor bugs
9956
9957     * [245]11022 no warning for non-compatible macro redefinition
9958
9959    C++ compiler and library
9960
9961     * [246]2330 static_cast<>() to a private base is allowed
9962     * [247]5388 Incorrect message "operands to ?: have different types"
9963     * [248]5390 Libiberty fails to demangle multi-digit template
9964       parameters
9965     * [249]7877 Incorrect parameter passing to specializations of member
9966       function templates
9967     * [250]9393 Anonymous namespaces and compiling the same file twice
9968     * [251]10032 -pedantic converts some errors to warnings
9969     * [252]10468 const typeof(x) is non-const, but only in templates
9970     * [253]10527 confused error message with "new int()" parameter
9971       initializer
9972     * [254]10679 parameter MIN_INLINE_INSNS is not honored
9973     * [255]10682 gcc chokes on a typedef for an enum inside a class
9974       template
9975     * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of
9976       0.
9977     * [257]10845 template member function (with nested template as
9978       parameter) cannot be called anymore if another unrelated template
9979       member function is defined
9980     * [258]10849 Cannot define an out-of-class specialization of a
9981       private nested template class
9982     * [259]10888 Suppress -Winline warnings for system headers
9983     * [260]10929 -Winline warns about functions for which no definition
9984       is visible
9985     * [261]10931 valid conversion static_cast<const unsigned
9986       int&>(lvalue-of-type-int) is rejected
9987     * [262]10940 Bad code with explicit specialization
9988     * [263]10968 If member function implicitly instantiated, explicit
9989       instantiation of class fails to instantiate it
9990     * [264]10990 Cannot convert with dynamic_cast<> to a private base
9991       class from within a member function
9992     * [265]11039 Bad interaction between implicit typename deprecation
9993       and friendship
9994     * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say
9995       "__unused__" instead
9996     * [267]11095 C++ iostream manipulator causes segfault when called
9997       with negative argument
9998     * [268]11098 g++ doesn't emit complete debugging information for
9999       local variables in destructors
10000     * [269]11137 GNU/Linux shared library constructors not called unless
10001       there's one global object
10002     * [270]11154 spurious ambiguity report for template class
10003       specialization
10004     * [271]11329 Compiler cannot find user defined implicit typecast
10005     * [272]11332 Spurious error with casts in ?: expression
10006     * [273]11431 static_cast behavior with subclasses when default
10007       constructor available
10008     * [274]11528 money_get facet does not accept "$.00" as valid
10009     * [275]11546 Type lookup problems in out-of-line definition of a
10010       class doubly nested from a template class
10011     * [276]11567 C++ code containing templated member function with same
10012       name as pure virtual member function results in linking failure
10013     * [277]11645 Failure to deal with using and private inheritance
10014
10015    Java compiler and library
10016
10017     * [278]5179 Qualified static field access doesn't initialize its
10018       class
10019     * [279]8204 gcj -O2 to native reorders certain instructions
10020       improperly
10021     * [280]10838 java.io.ObjectInputStream syntax error
10022     * [281]10886 The RMI registry that comes with GCJ does not work
10023       correctly
10024     * [282]11349 JNDI URL context factories not located correctly
10025
10026    x86-specific (Intel/AMD)
10027
10028     * [283]4823 ICE on inline assembly code
10029     * [284]8878 miscompilation with -O and SSE
10030     * [285]9815 (c++ library) atomicity.h - fails to compile with -O3
10031       -masm=intel
10032     * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads,
10033       in reload1.c
10034     * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2
10035     * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source
10036     * [289]11044 [x86] out of range loop instructions for FP code on K6
10037     * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE
10038       built-ins
10039     * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC"
10040       is used
10041
10042    SPARC- or Solaris- specific
10043
10044     * [292]9362 solaris 'as' dies when fed .s and "-gstabs"
10045     * [293]10142 [SPARC64] gcc produces wrong code when passing
10046       structures by value
10047     * [294]10663 New configure check aborts with Sun tools.
10048     * [295]10835 combinatorial explosion in scheduler on HyperSPARC
10049     * [296]10876 ICE in calculate_giv_inc when building KDE
10050     * [297]10955 wrong code at -O3 for structure argument in context of
10051       structure return
10052     * [298]11018 -mcpu=ultrasparc busts tar-1.13.25
10053     * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x
10054       Linux kernel
10055
10056    ia64 specific
10057
10058     * [300]10907 gcc violates the ia64 ABI (GP must be preserved)
10059     * [301]11320 scheduler bug (in machine depended reorganization pass)
10060     * [302]11599 bug with conditional and __builtin_prefetch
10061
10062    PowerPC specific
10063
10064     * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem
10065       during loop)
10066     * [304]10871 error in rs6000_stack_info save_size computation
10067     * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse
10068       cures it
10069
10070    m68k-specific
10071
10072     * [306]7594 [m68k] ICE on legal code associated with simplify-rtx
10073     * [307]10557 [m68k] ICE in subreg_offset_representable_p
10074     * [308]11054 [m68k] ICE in reg_overlap_mentioned_p
10075
10076    ARM-specific
10077
10078     * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for
10079       functions with __attribute__ ((interrupt ("IRQ")))
10080     * [310]10842 [arm] Clobbered link register is copied to pc under
10081       certain circumstances
10082     * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes
10083     * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno
10084       (3.4)
10085
10086    MIPS-specific
10087
10088     * [313]11084 ICE in propagate_one_insn, in flow.c
10089
10090    SH-specific
10091
10092     * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf
10093     * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c
10094     * [316]11096 i686-linux to sh-linux cross compiler fails to compile
10095       C++ files
10096
10097    GNU/Linux (or Hurd?) specific
10098
10099     * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3
10100
10101    UnixWare specific
10102
10103     * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare
10104       7.1.1
10105
10106    Cygwin (or mingw) specific
10107
10108     * [319]5287 ICE with dllimport attribute
10109     * [320]10148 [MingW/CygWin] Compiler dumps core
10110
10111    DJGPP specific
10112
10113     * [321]8787 GCC fails to emit .intel_syntax when invoked with
10114       -masm=intel on DJGPP
10115
10116    Darwin (and MacOS X) specific
10117
10118     * [322]10900 trampolines crash
10119
10120    Documentation
10121
10122     * [323]1607 (c++) Format attributes on methods undocumented
10123     * [324]4252 Invalid option `-fdump-translation-unit'
10124     * [325]4490 Clarify restrictions on -m96bit-long-double,
10125       -m128bit-long-double
10126     * [326]10355 document an issue with regparm attribute on some systems
10127       (e.g. Solaris)
10128     * [327]10726 (fortran) Documentation for function "IDate Intrinsic
10129       (Unix)" is wrong
10130     * [328]10805 document bug in old version of Sun assembler
10131     * [329]10815 warn against GNU binutils on AIX
10132     * [330]10877 document need for newer binutils on i?86-*-linux-gnu
10133     * [331]11280 Manual incorrect with respect to -freorder-blocks
10134     * [332]11466 Document -mlittle-endian and its restrictions for the
10135       sparc64 port
10136
10137    Testsuite bugs (compiler itself is not affected)
10138
10139     * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly
10140       report failure
10141     * [334]10810 gcc-3.3 fails make check: buffer overrun in
10142       test_demangle.c
10143     __________________________________________________________________
10144
10145GCC 3.3.2
10146
10147  Bug Fixes
10148
10149   This section lists the problem reports (PRs) from [335]GCC's bug
10150   tracking system that are known to be fixed in the 3.3.2 release. This
10151   list might not be complete (that is, it is possible that some PRs that
10152   have been fixed are not listed here).
10153
10154    Bootstrap failures and problems
10155
10156     * [336]8336 [SCO5] bootstrap config still tries to use COFF options
10157     * [337]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with
10158       --enable-threads=posix
10159     * [338]9631 [hppa64-linux] gcc-3.3 fails to bootstrap
10160     * [339]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare
10161       7.1.1)
10162     * [340]11687 xstormy16-elf build fails in libf2c
10163     * [341]12263 [SGI IRIX] bootstrap fails during compile of
10164       libf2c/libI77/backspace.c
10165     * [342]12490 buffer overflow in scan-decls.c (during Solaris 9
10166       fix-header processing)
10167
10168    Internal compiler errors (multi-platform)
10169
10170     * [343]7277 Casting integers to vector types causes ICE
10171     * [344]7939 (c++) ICE on invalid function template specialization
10172     * [345]11063 (c++) ICE on parsing initialization list of const array
10173       member
10174     * [346]11207 ICE with negative index in array element designator
10175     * [347]11522 (fortran) g77 dwarf-2 ICE in
10176       add_abstract_origin_attribute
10177     * [348]11595 (c++) ICE on duplicate label definition
10178     * [349]11646 (c++) ICE in commit_one_edge_insertion with
10179       -fnon-call-exceptions -fgcse -O
10180     * [350]11665 ICE in struct initializer when taking address
10181     * [351]11852 (c++) ICE with bad struct initializer.
10182     * [352]11878 (c++) ICE in cp_expr_size
10183     * [353]11883 ICE with any -O on mercury-generated C code
10184     * [354]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in
10185       cp/typeck2.c when applying typeid operator to template template
10186       parameter
10187     * [355]12146 ICE in lookup_template_function, in cp/pt.c
10188     * [356]12215 ICE in make_label_edge with -fnon-call-exceptions
10189       -fno-gcse -O2
10190     * [357]12369 (c++) ICE with templates and friends
10191     * [358]12446 ICE in emit_move_insn on complicated array reference
10192     * [359]12510 ICE in final_scan_insn
10193     * [360]12544 ICE with large parameters used in nested functions
10194
10195    C and optimization bugs
10196
10197     * [361]9862 spurious warnings with -W -finline-functions
10198     * [362]10962 lookup_field is a linear search on a linked list (can be
10199       slow if large struct)
10200     * [363]11370 -Wunreachable-code gives false complaints
10201     * [364]11637 invalid assembly with -fnon-call-exceptions
10202     * [365]11885 Problem with bitfields in packed structs
10203     * [366]12082 Inappropriate unreachable code warnings
10204     * [367]12180 Inline optimization fails for variadic function
10205     * [368]12340 loop unroller + gcse produces wrong code
10206
10207    C++ compiler and library
10208
10209     * [369]3907 nested template parameter collides with member name
10210     * [370]5293 confusing message when binding a temporary to a reference
10211     * [371]5296 [DR115] Pointers to functions and to template functions
10212       behave differently in deduction
10213     * [372]7939 ICE on function template specialization
10214     * [373]8656 Unable to assign function with __attribute__ and pointer
10215       return type to an appropriate variable
10216     * [374]10147 Confusing error message for invalid template function
10217       argument
10218     * [375]11400 std::search_n() makes assumptions about Size parameter
10219     * [376]11409 issues with using declarations, overloading, and
10220       built-in functions
10221     * [377]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle
10222       multiple bits in mask
10223     * [378]11786 operator() call on variable in other namespace not
10224       recognized
10225     * [379]11867 static_cast ignores ambiguity
10226     * [380]11928 bug with conversion operators that are typedefs
10227     * [381]12114 Uninitialized memory accessed in dtor
10228     * [382]12163 static_cast + explicit constructor regression
10229     * [383]12181 Wrong code with comma operator and c++
10230     * [384]12236 regparm and fastcall messes up parameters
10231     * [385]12266 incorrect instantiation of unneeded template during
10232       overload resolution
10233     * [386]12296 istream::peek() doesn't set eofbit
10234     * [387]12298 [sjlj exceptions] Stack unwind destroys
10235       not-yet-constructed object
10236     * [388]12369 ICE with templates and friends
10237     * [389]12337 apparently infinite loop in g++
10238     * [390]12344 stdcall attribute ignored if function returns a pointer
10239     * [391]12451 missing(late) class forward declaration in cxxabi.h
10240     * [392]12486 g++ accepts invalid use of a qualified name
10241
10242    x86 specific (Intel/AMD)
10243
10244     * [393]8869 [x86 MMX] ICE with const variable optimization and MMX
10245       builtins
10246     * [394]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions
10247       -O2
10248     * [395]11689 g++3.3 emits un-assembleable code for k6 architecture
10249     * [396]12116 [k6] Invalid assembly output values with X-MAME code
10250     * [397]12070 ICE converting between double and long double with
10251       -msoft-float
10252
10253    ia64-specific
10254
10255     * [398]11184 [ia64 hpux] ICE on __builtin_apply building libobjc
10256     * [399]11535 __builtin_return_address may not work on ia64
10257     * [400]11693 [ia64] ICE in gen_nop_type
10258     * [401]12224 [ia64] Thread-local storage doesn't work
10259
10260    PowerPC-specific
10261
10262     * [402]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux
10263       kernel
10264     * [403]11319 loop miscompiled on ppc32
10265     * [404]11949 ICE Compiler segfault with ffmpeg -maltivec code
10266
10267    SPARC-specific
10268
10269     * [405]11662 wrong code for expr. with cast to long long and
10270       exclusive or
10271     * [406]11965 invalid assembler code for a shift < 32 operation
10272     * [407]12301 (c++) stack corruption when a returned expression throws
10273       an exception
10274
10275    Alpha-specific
10276
10277     * [408]11717 [alpha-linux] unrecognizable insn compiling for.c of
10278       kernel 2.4.22-pre8
10279
10280    HPUX-specific
10281
10282     * [409]11313 problem with #pragma weak and static inline functions
10283     * [410]11712 __STDC_EXT__ not defined for C++ by default anymore?
10284
10285    Solaris specific
10286
10287     * [411]12166 Profiled programs crash if PROFDIR is set
10288
10289    Solaris-x86 specific
10290
10291     * [412]12101 i386 Solaris no longer works with GNU as?
10292
10293    Miscellaneous embedded target-specific bugs
10294
10295     * [413]10988 [m32r-elf] wrong blockmove code with -O3
10296     * [414]11805 [h8300-unknown-coff] [H8300] ICE for simple code with
10297       -O2
10298     * [415]11902 [sh4] spec file improperly inserts rpath even when none
10299       needed
10300     * [416]11903 [sh4] -pthread fails to link due to error in spec file
10301       on sh4
10302     __________________________________________________________________
10303
10304GCC 3.3.3
10305
10306  Minor features
10307
10308   In addition to the bug fixes documented below, this release contains
10309   few minor features such as:
10310     * Support for --with-sysroot
10311     * Support for automatic detection of executable stacks
10312     * Support for SSE3 instructions
10313     * Support for thread local storage debugging under GDB on S390
10314
10315  Bug Fixes
10316
10317   This section lists the problem reports (PRs) from [417]GCC's bug
10318   tracking system that are known to be fixed in the 3.3.3 release. This
10319   list might not be complete (that is, it is possible that some PRs that
10320   have been fixed are not listed here).
10321
10322    Bootstrap failures and issues
10323
10324     * [418]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails
10325     * [419]12399 boehm-gc fails (when building a cross compiler): libtool
10326       unable to infer tagged configuration
10327     * [420]13068 mklibgcc.in doesn't handle multi-level multilib
10328       subdirectories properly
10329
10330    Internal compiler errors (multi-platform)
10331
10332     * [421]10060 ICE (stack overflow) on huge file (300k lines) due to
10333       recursive behaviour of copy_rtx_if_shared, in emit_rtl.c
10334     * [422]10555 (c++) ICE on undefined template argument
10335     * [423]10706 (c++) ICE in mangle_class_name_for_template
10336     * [424]11496 (fortran) error in flow_loops_find when -funroll-loops
10337       active
10338     * [425]11741 ICE in pre_insert_copy_insn, in gcse.c
10339     * [426]12440 GCC crashes during compilation of quicktime4linux 2.0.0
10340     * [427]12632 (fortran) -fbounds-check ICE
10341     * [428]12712 (c++) ICE on short legit C++ code fragment with gcc
10342       3.3.2
10343     * [429]12726 (c++) ICE (segfault) on trivial code
10344     * [430]12890 (c++) ICE on compilation of class with throwing method
10345     * [431]12900 (c++) ICE in rtl_verify_flow_info_1
10346     * [432]13060 (fortran) ICE in fixup_var_refs_1, in function.c on
10347       correct code with -O2 -fno-force-mem
10348     * [433]13289 (c++) ICE in regenerate_decl_from_template on recursive
10349       template
10350     * [434]13318 ICE: floating point exception in the loop optimizer
10351     * [435]13392 (c++) ICE in convert_from_eh_region_ranges_1, in
10352       except.c
10353     * [436]13574 (c++) invalid array default initializer in class lets
10354       gcc consume all memory and die
10355     * [437]13475 ICE on SIMD variables with partial value initialization
10356     * [438]13797 (c++) ICE on invalid template parameter
10357     * [439]13824 (java) gcj SEGV with simple .java program
10358
10359    C and optimization bugs
10360
10361     * [440]8776 loop invariants are not removed (most likely)
10362     * [441]10339 [sparc,ppc,ppc64] Invalid optimization: replacing
10363       strncmp by memcmp
10364     * [442]11350 undefined labels with -Os -fPIC
10365     * [443]12826 Optimizer removes reference through volatile pointer
10366     * [444]12500 stabs debug info: void no longer a predefined / builtin
10367       type
10368     * [445]12941 builtin-bitops-1.c miscompilation (latent bug)
10369     * [446]12953 tree inliner bug (in inline_forbidden_p) and fix
10370     * [447]13041 linux-2.6/sound/core/oss/rate.c miscompiled
10371     * [448]13507 spurious printf format warning
10372     * [449]13382 Type information for const pointer disappears during
10373       optimization.
10374     * [450]13394 noreturn attribute ignored on recursive invokation
10375     * [451]13400 Compiled code crashes storing to read-only location
10376     * [452]13521 Endless loop in calculate_global_regs_live
10377
10378    C++ compiler and library
10379
10380   Some of the bug fixes in this list were made to implement decisions
10381   that the ISO C++ standards committee has made concerning several defect
10382   reports (DRs). Links in the list below point to detailed discussion of
10383   the relevant defect report.
10384     * [453]2094 unimplemented: use of `ptrmem_cst' in template type
10385       unification
10386     * [454]2294 using declaration confusion
10387     * [455]5050 template instantiation depth exceeds limit: recursion
10388       problem?
10389     * [456]9371 Bad exception handling in
10390       i/ostream::operator>>/<<(streambuf*)
10391     * [457]9546 bad exception handling in ostream members
10392     * [458]10081 basic_ios::_M_cache_locale leaves NULL members in the
10393       face of unknown locales
10394     * [459]10093 [460][DR 61] Setting failbit in exceptions doesn't work
10395     * [461]10095 istream::operator>>(int&) sets ios::badbit when
10396       ios::failbit is set.
10397     * [462]11554 Warning about reordering of initializers doesn't mention
10398       location of constructor
10399     * [463]12297 istream::sentry::sentry() handles eof() incorrectly.
10400     * [464]12352 Exception safety problems in src/localename.cc
10401     * [465]12438 Memory leak in locale::combine()
10402     * [466]12540 Memory leak in locale::locale(const char*)
10403     * [467]12594 DRs [468]60 [TC] and [469]63 [TC] not implemented
10404     * [470]12657 Resolution of [471]DR 292 (WP) still unimplemented
10405     * [472]12696 memory eating infinite loop in diagnostics (error
10406       recovery problem)
10407     * [473]12815 Code compiled with optimization behaves unexpectedly
10408     * [474]12862 Conflicts between typedefs/enums and namespace member
10409       declarations
10410     * [475]12926 Wrong value after assignment in initialize list using
10411       bit-fields
10412     * [476]12967 Resolution of [477]DR 300 [WP] still unimplemented
10413     * [478]12971 Resolution of [479]DR 328 [WP] still unimplemented
10414     * [480]13007 basic_streambuf::pubimbue, imbue wrong
10415     * [481]13009 Implicitly-defined assignment operator writes to wrong
10416       memory
10417     * [482]13057 regparm attribute not applied to destructor
10418     * [483]13070 -Wformat option ignored in g++
10419     * [484]13081 forward template declarations in <complex> let inlining
10420       fail
10421     * [485]13239 Assertion does not seem to work correctly anymore
10422     * [486]13262 "xxx is private within this context" when initializing a
10423       self-contained template class
10424     * [487]13290 simple typo in concept checking for std::generate_n
10425     * [488]13323 Template code does not compile in presence of typedef
10426     * [489]13369 __verify_grouping (and __add_grouping?) not correct
10427     * [490]13371 infinite loop with packed struct and inlining
10428     * [491]13445 Template argument replacement "dereferences" a typedef
10429     * [492]13461 Fails to access protected-ctor from public constant
10430     * [493]13462 Non-standard-conforming type set::pointer
10431     * [494]13478 gcc uses wrong constructor to initialize a const
10432       reference
10433     * [495]13544 "conflicting types" for enums in different scopes
10434     * [496]13650 string::compare should not (always) use
10435       traits_type::length()
10436     * [497]13683 bogus warning about passing non-PODs through ellipsis
10437     * [498]13688 Derived class is denied access to protected base class
10438       member class
10439     * [499]13774 Member variable cleared in virtual multiple inheritance
10440       class
10441     * [500]13884 Protect sstream.tcc from extern template use
10442
10443    Java compiler and library
10444
10445     * [501]10746 [win32] garbage collection crash in GCJ
10446
10447    Objective-C compiler and library
10448
10449     * [502]11433 Crash due to dereferencing null pointer when querying
10450       protocol
10451
10452    Fortran compiler and library
10453
10454     * [503]12633 logical expression gives incorrect result with
10455       -fugly-logint option
10456     * [504]13037 [gcse-lm] g77 generates incorrect code
10457     * [505]13213 Hex constant problem when compiling with -fugly-logint
10458       and -ftypeless-boz
10459
10460    x86-specific (Intel/AMD)
10461
10462     * [506]4490 ICE with -m128bit-long-double
10463     * [507]12292 [x86_64] ICE: RTL check: expected code `const_int', have
10464       `reg' in make_field_assignment, in combine.c
10465     * [508]12441 ICE: can't find a register to spill
10466     * [509]12943 array static-init failure under -fpic, -fPIC
10467     * [510]13608 Incorrect code with -O3 -ffast-math
10468
10469    PowerPC-specific
10470
10471     * [511]11598 testcase gcc.dg/20020118-1.c fails runtime check of
10472       __attribute__((aligned(16)))
10473     * [512]11793 ICE in extract_insn, in recog.c (const_vector's)
10474     * [513]12467 vmsumubm emitted when vmsummbm appropriate (typo in
10475       altivec.md)
10476     * [514]12537 g++ generates writeable text sections
10477
10478    SPARC-specific
10479
10480     * [515]12496 wrong result for __atomic_add(&value, -1) when using -O0
10481       -m64
10482     * [516]12865 mprotect call to make trampoline executable may fail
10483     * [517]13354 ICE in sparc_emit_set_const32
10484
10485    ARM-specific
10486
10487     * [518]10467 [arm] ICE in pre_insert_copy_insn,
10488
10489    ia64-specific
10490
10491     * [519]11226 ICE passing struct arg with two floats
10492     * [520]11227 ICE for _Complex float, _Complex long double args
10493     * [521]12644 GCC 3.3.2 fails to compile glibc on ia64
10494     * [522]13149 build gcc-3.3.2 1305 error:unrecognizable insn
10495     * Various fixes for libunwind
10496
10497    Alpha-specific
10498
10499     * [523]12654 Incorrect comparison code generated for Alpha
10500     * [524]12965 SEGV+ICE in cc1plus on alpha-linux with -O2
10501     * [525]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2
10502
10503    HPPA-specific
10504
10505     * [526]11634 [hppa] ICE in verify_local_live_at_start, in flow.c
10506     * [527]12158 [hppa] compilation does not terminate at -O1
10507
10508    S390-specific
10509
10510     * [528]11992 Wrong built-in code for memcmp with length 1<<24: only
10511       (1<<24)-1 possible for CLCL-Instruction
10512
10513    SH-specific
10514
10515     * [529]9365 segfault in gen_far_branch (config/sh/sh.c)
10516     * [530]10392 optimizer generates faulty array indexing
10517     * [531]11322 SH profiler outputs multiple definitions of symbol
10518     * [532]13069 gcc/config/sh/rtems.h broken
10519     * [533]13302 Putting a va_list in a struct causes seg fault
10520     * [534]13585 Incorrect optimization of call to sfunc
10521     * Fix inappropriately exported libgcc functions from the shared
10522       library
10523
10524    Other embedded target specific
10525
10526     * [535]8916 [mcore] unsigned char assign gets hosed.
10527     * [536]11576 [h8300] ICE in change_address_1, in emit-rtl.c
10528     * [537]13122 [h8300] local variable gets corrupted by function call
10529       when -fomit-frame-pointer is given
10530     * [538]13256 [cris] strict_low_part mistreated in delay slots
10531     * [539]13373 [mcore] optimization with -frerun-cse-after-loop
10532       -fexpensive-optimizations produces wrong code on mcore
10533
10534    GNU HURD-specific
10535
10536     * [540]12561 gcc/config/t-gnu needs updating to work with
10537       --with-sysroot
10538
10539    Tru64 Unix specific
10540
10541     * [541]6243 testsuite fails almost all tests due to no libintl in
10542       LD_LIBRARY_PATH during test.
10543     * [542]11397 weak aliases broken on Tru64 UNIX
10544
10545    AIX-specific
10546
10547     * [543]12505 build failure due to defines of uchar in cpphash.h and
10548       sys/types.h
10549     * [544]13150 WEAK symbols not exported by collect2
10550
10551    IRIX-specific
10552
10553     * [545]12666 fixincludes problem on IRIX 6.5.19m
10554
10555    Solaris-specific
10556
10557     * [546]12969 Including sys/byteorder.h breaks configure checks
10558
10559    Testsuite problems (compiler is not affected)
10560
10561     * [547]10819 testsuite creates CR+LF on compiler version lines in
10562       test summary files
10563     * [548]11612 abi_check not finding correct libgcc_s.so.1
10564
10565    Miscellaneous
10566
10567     * [549]13211 using -###, incorrect warnings about unused linker file
10568       are produced
10569     __________________________________________________________________
10570
10571GCC 3.3.4
10572
10573   This is the [550]list of problem reports (PRs) from GCC's bug tracking
10574   system that are known to be fixed in the 3.3.4 release. This list might
10575   not be complete (that is, it is possible that some PRs that have been
10576   fixed are not listed here).
10577     __________________________________________________________________
10578
10579GCC 3.3.5
10580
10581   This is the [551]list of problem reports (PRs) from GCC's bug tracking
10582   system that are known to be fixed in the 3.3.5 release. This list might
10583   not be complete (that is, it is possible that some PRs that have been
10584   fixed are not listed here).
10585     __________________________________________________________________
10586
10587GCC 3.3.6
10588
10589   This is the [552]list of problem reports (PRs) from GCC's bug tracking
10590   system that are known to be fixed in the 3.3.6 release. This list might
10591   not be complete (that is, it is possible that some PRs that have been
10592   fixed are not listed here).
10593
10594
10595    For questions related to the use of GCC, please consult these web
10596    pages and the [553]GCC manuals. If that fails, the
10597    [554]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10598    web pages and the development of GCC are welcome on our developer
10599    list at [555]gcc@gcc.gnu.org. All of [556]our lists have public
10600    archives.
10601
10602   Copyright (C) [557]Free Software Foundation, Inc. Verbatim copying and
10603   distribution of this entire article is permitted in any medium,
10604   provided this notice is preserved.
10605
10606   These pages are [558]maintained by the GCC team. Last modified
10607   2013-10-31[559].
10608
10609References
10610
10611   1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
10612   2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems
10613   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
10614   4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute
10615   5. http://gcc.gnu.org/news/dfa.html
10616   6. http://gcc.gnu.org/c99status.html
10617   7. http://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html
10618   8. http://gcc.gnu.org/PR10140
10619   9. http://gcc.gnu.org/PR10198
10620  10. http://gcc.gnu.org/PR10338
10621  11. http://gcc.gnu.org/PR3581
10622  12. http://gcc.gnu.org/PR4382
10623  13. http://gcc.gnu.org/PR5533
10624  14. http://gcc.gnu.org/PR6387
10625  15. http://gcc.gnu.org/PR6412
10626  16. http://gcc.gnu.org/PR6620
10627  17. http://gcc.gnu.org/PR6663
10628  18. http://gcc.gnu.org/PR7068
10629  19. http://gcc.gnu.org/PR7083
10630  20. http://gcc.gnu.org/PR7647
10631  21. http://gcc.gnu.org/PR7675
10632  22. http://gcc.gnu.org/PR7718
10633  23. http://gcc.gnu.org/PR8116
10634  24. http://gcc.gnu.org/PR8358
10635  25. http://gcc.gnu.org/PR8511
10636  26. http://gcc.gnu.org/PR8564
10637  27. http://gcc.gnu.org/PR8660
10638  28. http://gcc.gnu.org/PR8766
10639  29. http://gcc.gnu.org/PR8803
10640  30. http://gcc.gnu.org/PR8846
10641  31. http://gcc.gnu.org/PR8906
10642  32. http://gcc.gnu.org/PR9216
10643  33. http://gcc.gnu.org/PR9261
10644  34. http://gcc.gnu.org/PR9263
10645  35. http://gcc.gnu.org/PR9429
10646  36. http://gcc.gnu.org/PR9516
10647  37. http://gcc.gnu.org/PR9600
10648  38. http://gcc.gnu.org/PR9629
10649  39. http://gcc.gnu.org/PR9672
10650  40. http://gcc.gnu.org/PR9749
10651  41. http://gcc.gnu.org/PR9794
10652  42. http://gcc.gnu.org/PR9829
10653  43. http://gcc.gnu.org/PR9916
10654  44. http://gcc.gnu.org/PR9936
10655  45. http://gcc.gnu.org/PR10262
10656  46. http://gcc.gnu.org/PR10278
10657  47. http://gcc.gnu.org/PR10446
10658  48. http://gcc.gnu.org/PR10451
10659  49. http://gcc.gnu.org/PR10506
10660  50. http://gcc.gnu.org/PR10549
10661  51. http://gcc.gnu.org/PR2001
10662  52. http://gcc.gnu.org/PR2391
10663  53. http://gcc.gnu.org/PR2960
10664  54. http://gcc.gnu.org/PR4046
10665  55. http://gcc.gnu.org/PR6405
10666  56. http://gcc.gnu.org/PR6798
10667  57. http://gcc.gnu.org/PR6871
10668  58. http://gcc.gnu.org/PR6909
10669  59. http://gcc.gnu.org/PR7189
10670  60. http://gcc.gnu.org/PR7642
10671  61. http://gcc.gnu.org/PR8634
10672  62. http://gcc.gnu.org/PR8750
10673  63. http://gcc.gnu.org/PR2161
10674  64. http://gcc.gnu.org/PR4319
10675  65. http://gcc.gnu.org/PR8602
10676  66. http://gcc.gnu.org/PR9177
10677  67. http://gcc.gnu.org/PR9853
10678  68. http://gcc.gnu.org/PR45
10679  69. http://gcc.gnu.org/PR3784
10680  70. http://gcc.gnu.org/PR764
10681  71. http://gcc.gnu.org/PR5116
10682  72. http://gcc.gnu.org/PR2862
10683  73. http://gcc.gnu.org/PR3663
10684  74. http://gcc.gnu.org/PR3797
10685  75. http://gcc.gnu.org/PR3948
10686  76. http://gcc.gnu.org/PR4137
10687  77. http://gcc.gnu.org/PR4361
10688  78. http://gcc.gnu.org/PR4802
10689  79. http://gcc.gnu.org/PR5837
10690  80. http://gcc.gnu.org/PR4803
10691  81. http://gcc.gnu.org/PR5094
10692  82. http://gcc.gnu.org/PR5730
10693  83. http://gcc.gnu.org/PR6713
10694  84. http://gcc.gnu.org/PR7015
10695  85. http://gcc.gnu.org/PR7086
10696  86. http://gcc.gnu.org/PR7099
10697  87. http://gcc.gnu.org/PR7247
10698  88. http://gcc.gnu.org/PR7441
10699  89. http://gcc.gnu.org/PR7768
10700  90. http://gcc.gnu.org/PR7804
10701  91. http://gcc.gnu.org/PR8099
10702  92. http://gcc.gnu.org/PR8117
10703  93. http://gcc.gnu.org/PR8205
10704  94. http://gcc.gnu.org/PR8645
10705  95. http://gcc.gnu.org/PR8724
10706  96. http://gcc.gnu.org/PR8805
10707  97. http://gcc.gnu.org/PR8691
10708  98. http://gcc.gnu.org/PR8700
10709  99. http://gcc.gnu.org/PR8724
10710 100. http://gcc.gnu.org/PR8949
10711 101. http://gcc.gnu.org/PR9016
10712 102. http://gcc.gnu.org/PR9053
10713 103. http://gcc.gnu.org/PR9152
10714 104. http://gcc.gnu.org/PR9182
10715 105. http://gcc.gnu.org/PR9297
10716 106. http://gcc.gnu.org/PR9318
10717 107. http://gcc.gnu.org/PR9320
10718 108. http://gcc.gnu.org/PR9400
10719 109. http://gcc.gnu.org/PR9424
10720 110. http://gcc.gnu.org/PR9425
10721 111. http://gcc.gnu.org/PR9439
10722 112. http://gcc.gnu.org/PR9474
10723 113. http://gcc.gnu.org/PR9548
10724 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231
10725 115. http://gcc.gnu.org/PR9555
10726 116. http://gcc.gnu.org/PR9561
10727 117. http://gcc.gnu.org/PR9563
10728 118. http://gcc.gnu.org/PR9582
10729 119. http://gcc.gnu.org/PR9622
10730 120. http://gcc.gnu.org/PR9683
10731 121. http://gcc.gnu.org/PR9791
10732 122. http://gcc.gnu.org/PR9817
10733 123. http://gcc.gnu.org/PR9825
10734 124. http://gcc.gnu.org/PR9826
10735 125. http://gcc.gnu.org/PR9924
10736 126. http://gcc.gnu.org/PR9946
10737 127. http://gcc.gnu.org/PR9964
10738 128. http://gcc.gnu.org/PR9988
10739 129. http://gcc.gnu.org/PR10033
10740 130. http://gcc.gnu.org/PR10097
10741 131. http://gcc.gnu.org/PR10132
10742 132. http://gcc.gnu.org/PR10180
10743 133. http://gcc.gnu.org/PR10199
10744 134. http://gcc.gnu.org/PR10300
10745 135. http://gcc.gnu.org/PR10427
10746 136. http://gcc.gnu.org/PR10503
10747 137. http://gcc.gnu.org/PR5956
10748 138. http://gcc.gnu.org/PR1832
10749 139. http://gcc.gnu.org/PR3924
10750 140. http://gcc.gnu.org/PR5634
10751 141. http://gcc.gnu.org/PR6367
10752 142. http://gcc.gnu.org/PR6491
10753 143. http://gcc.gnu.org/PR6742
10754 144. http://gcc.gnu.org/PR7113
10755 145. http://gcc.gnu.org/PR7236
10756 146. http://gcc.gnu.org/PR7278
10757 147. http://gcc.gnu.org/PR7384
10758 148. http://gcc.gnu.org/PR7388
10759 149. http://gcc.gnu.org/PR8587
10760 150. http://gcc.gnu.org/PR9038
10761 151. http://gcc.gnu.org/PR10197
10762 152. http://gcc.gnu.org/PR6005
10763 153. http://gcc.gnu.org/PR6389
10764 154. http://gcc.gnu.org/PR6576
10765 155. http://gcc.gnu.org/PR6652
10766 156. http://gcc.gnu.org/PR7060
10767 157. http://gcc.gnu.org/PR7073
10768 158. http://gcc.gnu.org/PR7180
10769 159. http://gcc.gnu.org/PR7416
10770 160. http://gcc.gnu.org/PR7570
10771 161. http://gcc.gnu.org/PR7578
10772 162. http://gcc.gnu.org/PR7611
10773 163. http://gcc.gnu.org/PR7709
10774 164. http://gcc.gnu.org/PR7766
10775 165. http://gcc.gnu.org/PR7785
10776 166. http://gcc.gnu.org/PR7786
10777 167. http://gcc.gnu.org/PR8142
10778 168. http://gcc.gnu.org/PR8234
10779 169. http://gcc.gnu.org/PR8415
10780 170. http://gcc.gnu.org/PR8481
10781 171. http://gcc.gnu.org/PR8593
10782 172. http://gcc.gnu.org/PR8759
10783 173. http://gcc.gnu.org/PR8997
10784 174. http://gcc.gnu.org/PR9253
10785 175. http://gcc.gnu.org/PR9254
10786 176. http://gcc.gnu.org/PR9271
10787 177. http://gcc.gnu.org/PR6767
10788 178. http://gcc.gnu.org/PR9911
10789 179. http://gcc.gnu.org/PR10020
10790 180. http://gcc.gnu.org/PR10546
10791 181. http://gcc.gnu.org/PR7029
10792 182. http://gcc.gnu.org/PR2903
10793 183. http://gcc.gnu.org/PR7873
10794 184. http://gcc.gnu.org/PR7680
10795 185. http://gcc.gnu.org/PR8705
10796 186. http://gcc.gnu.org/PR9986
10797 187. http://gcc.gnu.org/PR10056
10798 188. http://gcc.gnu.org/PR6744
10799 189. http://gcc.gnu.org/PR7361
10800 190. http://gcc.gnu.org/PR9496
10801 191. http://gcc.gnu.org/PR7067
10802 192. http://gcc.gnu.org/PR8480
10803 193. http://gcc.gnu.org/PR8784
10804 194. http://gcc.gnu.org/PR10315
10805 195. http://gcc.gnu.org/PR10267
10806 196. http://gcc.gnu.org/PR7916
10807 197. http://gcc.gnu.org/PR7926
10808 198. http://gcc.gnu.org/PR8555
10809 199. http://gcc.gnu.org/PR8994
10810 200. http://gcc.gnu.org/PR9426
10811 201. http://gcc.gnu.org/PR9806
10812 202. http://gcc.gnu.org/PR10077
10813 203. http://gcc.gnu.org/PR10233
10814 204. http://gcc.gnu.org/PR10286
10815 205. http://gcc.gnu.org/PR10308
10816 206. http://gcc.gnu.org/PR11272
10817 207. http://gcc.gnu.org/PR5754
10818 208. http://gcc.gnu.org/PR6597
10819 209. http://gcc.gnu.org/PR6949
10820 210. http://gcc.gnu.org/PR7053
10821 211. http://gcc.gnu.org/PR8164
10822 212. http://gcc.gnu.org/PR8384
10823 213. http://gcc.gnu.org/PR9559
10824 214. http://gcc.gnu.org/PR9649
10825 215. http://gcc.gnu.org/PR9864
10826 216. http://gcc.gnu.org/PR10432
10827 217. http://gcc.gnu.org/PR10475
10828 218. http://gcc.gnu.org/PR10635
10829 219. http://gcc.gnu.org/PR10661
10830 220. http://gcc.gnu.org/PR10700
10831 221. http://gcc.gnu.org/PR10712
10832 222. http://gcc.gnu.org/PR10796
10833 223. http://gcc.gnu.org/PR10890
10834 224. http://gcc.gnu.org/PR10939
10835 225. http://gcc.gnu.org/PR10956
10836 226. http://gcc.gnu.org/PR11041
10837 227. http://gcc.gnu.org/PR11059
10838 228. http://gcc.gnu.org/PR11083
10839 229. http://gcc.gnu.org/PR11105
10840 230. http://gcc.gnu.org/PR11149
10841 231. http://gcc.gnu.org/PR11228
10842 232. http://gcc.gnu.org/PR11282
10843 233. http://gcc.gnu.org/PR11301
10844 234. http://gcc.gnu.org/PR11308
10845 235. http://gcc.gnu.org/PR11473
10846 236. http://gcc.gnu.org/PR11503
10847 237. http://gcc.gnu.org/PR11513
10848 238. http://gcc.gnu.org/PR11198
10849 239. http://gcc.gnu.org/PR11304
10850 240. http://gcc.gnu.org/PR11381
10851 241. http://gcc.gnu.org/PR11536
10852 242. http://gcc.gnu.org/PR11557
10853 243. http://gcc.gnu.org/PR5897
10854 244. http://gcc.gnu.org/PR11279
10855 245. http://gcc.gnu.org/PR11022
10856 246. http://gcc.gnu.org/PR2330
10857 247. http://gcc.gnu.org/PR5388
10858 248. http://gcc.gnu.org/PR5390
10859 249. http://gcc.gnu.org/PR7877
10860 250. http://gcc.gnu.org/PR9393
10861 251. http://gcc.gnu.org/PR10032
10862 252. http://gcc.gnu.org/PR10468
10863 253. http://gcc.gnu.org/PR10527
10864 254. http://gcc.gnu.org/PR10679
10865 255. http://gcc.gnu.org/PR10682
10866 256. http://gcc.gnu.org/PR10689
10867 257. http://gcc.gnu.org/PR10845
10868 258. http://gcc.gnu.org/PR10849
10869 259. http://gcc.gnu.org/PR10888
10870 260. http://gcc.gnu.org/PR10929
10871 261. http://gcc.gnu.org/PR10931
10872 262. http://gcc.gnu.org/PR10940
10873 263. http://gcc.gnu.org/PR10968
10874 264. http://gcc.gnu.org/PR10990
10875 265. http://gcc.gnu.org/PR11039
10876 266. http://gcc.gnu.org/PR11062
10877 267. http://gcc.gnu.org/PR11095
10878 268. http://gcc.gnu.org/PR11098
10879 269. http://gcc.gnu.org/PR11137
10880 270. http://gcc.gnu.org/PR11154
10881 271. http://gcc.gnu.org/PR11329
10882 272. http://gcc.gnu.org/PR11332
10883 273. http://gcc.gnu.org/PR11431
10884 274. http://gcc.gnu.org/PR11528
10885 275. http://gcc.gnu.org/PR11546
10886 276. http://gcc.gnu.org/PR11567
10887 277. http://gcc.gnu.org/PR11645
10888 278. http://gcc.gnu.org/PR5179
10889 279. http://gcc.gnu.org/PR8204
10890 280. http://gcc.gnu.org/PR10838
10891 281. http://gcc.gnu.org/PR10886
10892 282. http://gcc.gnu.org/PR11349
10893 283. http://gcc.gnu.org/PR4823
10894 284. http://gcc.gnu.org/PR8878
10895 285. http://gcc.gnu.org/PR9815
10896 286. http://gcc.gnu.org/PR10402
10897 287. http://gcc.gnu.org/PR10504
10898 288. http://gcc.gnu.org/PR10673
10899 289. http://gcc.gnu.org/PR11044
10900 290. http://gcc.gnu.org/PR11089
10901 291. http://gcc.gnu.org/PR11420
10902 292. http://gcc.gnu.org/PR9362
10903 293. http://gcc.gnu.org/PR10142
10904 294. http://gcc.gnu.org/PR10663
10905 295. http://gcc.gnu.org/PR10835
10906 296. http://gcc.gnu.org/PR10876
10907 297. http://gcc.gnu.org/PR10955
10908 298. http://gcc.gnu.org/PR11018
10909 299. http://gcc.gnu.org/PR11556
10910 300. http://gcc.gnu.org/PR10907
10911 301. http://gcc.gnu.org/PR11320
10912 302. http://gcc.gnu.org/PR11599
10913 303. http://gcc.gnu.org/PR9745
10914 304. http://gcc.gnu.org/PR10871
10915 305. http://gcc.gnu.org/PR11440
10916 306. http://gcc.gnu.org/PR7594
10917 307. http://gcc.gnu.org/PR10557
10918 308. http://gcc.gnu.org/PR11054
10919 309. http://gcc.gnu.org/PR10834
10920 310. http://gcc.gnu.org/PR10842
10921 311. http://gcc.gnu.org/PR11052
10922 312. http://gcc.gnu.org/PR11183
10923 313. http://gcc.gnu.org/PR11084
10924 314. http://gcc.gnu.org/PR10331
10925 315. http://gcc.gnu.org/PR10413
10926 316. http://gcc.gnu.org/PR11096
10927 317. http://gcc.gnu.org/PR2873
10928 318. http://gcc.gnu.org/PR3163
10929 319. http://gcc.gnu.org/PR5287
10930 320. http://gcc.gnu.org/PR10148
10931 321. http://gcc.gnu.org/PR8787
10932 322. http://gcc.gnu.org/PR10900
10933 323. http://gcc.gnu.org/PR1607
10934 324. http://gcc.gnu.org/PR4252
10935 325. http://gcc.gnu.org/PR4490
10936 326. http://gcc.gnu.org/PR10355
10937 327. http://gcc.gnu.org/PR10726
10938 328. http://gcc.gnu.org/PR10805
10939 329. http://gcc.gnu.org/PR10815
10940 330. http://gcc.gnu.org/PR10877
10941 331. http://gcc.gnu.org/PR11280
10942 332. http://gcc.gnu.org/PR11466
10943 333. http://gcc.gnu.org/PR10737
10944 334. http://gcc.gnu.org/PR10810
10945 335. http://gcc.gnu.org/bugzilla/
10946 336. http://gcc.gnu.org/PR8336
10947 337. http://gcc.gnu.org/PR9330
10948 338. http://gcc.gnu.org/PR9631
10949 339. http://gcc.gnu.org/PR9877
10950 340. http://gcc.gnu.org/PR11687
10951 341. http://gcc.gnu.org/PR12263
10952 342. http://gcc.gnu.org/PR12490
10953 343. http://gcc.gnu.org/PR7277
10954 344. http://gcc.gnu.org/PR7939
10955 345. http://gcc.gnu.org/PR11063
10956 346. http://gcc.gnu.org/PR11207
10957 347. http://gcc.gnu.org/PR11522
10958 348. http://gcc.gnu.org/PR11595
10959 349. http://gcc.gnu.org/PR11646
10960 350. http://gcc.gnu.org/PR11665
10961 351. http://gcc.gnu.org/PR11852
10962 352. http://gcc.gnu.org/PR11878
10963 353. http://gcc.gnu.org/PR11883
10964 354. http://gcc.gnu.org/PR11991
10965 355. http://gcc.gnu.org/PR12146
10966 356. http://gcc.gnu.org/PR12215
10967 357. http://gcc.gnu.org/PR12369
10968 358. http://gcc.gnu.org/PR12446
10969 359. http://gcc.gnu.org/PR12510
10970 360. http://gcc.gnu.org/PR12544
10971 361. http://gcc.gnu.org/PR9862
10972 362. http://gcc.gnu.org/PR10962
10973 363. http://gcc.gnu.org/PR11370
10974 364. http://gcc.gnu.org/PR11637
10975 365. http://gcc.gnu.org/PR11885
10976 366. http://gcc.gnu.org/PR12082
10977 367. http://gcc.gnu.org/PR12180
10978 368. http://gcc.gnu.org/PR12340
10979 369. http://gcc.gnu.org/PR3907
10980 370. http://gcc.gnu.org/PR5293
10981 371. http://gcc.gnu.org/PR5296
10982 372. http://gcc.gnu.org/PR7939
10983 373. http://gcc.gnu.org/PR8656
10984 374. http://gcc.gnu.org/PR10147
10985 375. http://gcc.gnu.org/PR11400
10986 376. http://gcc.gnu.org/PR11409
10987 377. http://gcc.gnu.org/PR11740
10988 378. http://gcc.gnu.org/PR11786
10989 379. http://gcc.gnu.org/PR11867
10990 380. http://gcc.gnu.org/PR11928
10991 381. http://gcc.gnu.org/PR12114
10992 382. http://gcc.gnu.org/PR12163
10993 383. http://gcc.gnu.org/PR12181
10994 384. http://gcc.gnu.org/PR12236
10995 385. http://gcc.gnu.org/PR12266
10996 386. http://gcc.gnu.org/PR12296
10997 387. http://gcc.gnu.org/PR12298
10998 388. http://gcc.gnu.org/PR12369
10999 389. http://gcc.gnu.org/PR12337
11000 390. http://gcc.gnu.org/PR12344
11001 391. http://gcc.gnu.org/PR12451
11002 392. http://gcc.gnu.org/PR12486
11003 393. http://gcc.gnu.org/PR8869
11004 394. http://gcc.gnu.org/PR9786
11005 395. http://gcc.gnu.org/PR11689
11006 396. http://gcc.gnu.org/PR12116
11007 397. http://gcc.gnu.org/PR12070
11008 398. http://gcc.gnu.org/PR11184
11009 399. http://gcc.gnu.org/PR11535
11010 400. http://gcc.gnu.org/PR11693
11011 401. http://gcc.gnu.org/PR12224
11012 402. http://gcc.gnu.org/PR11087
11013 403. http://gcc.gnu.org/PR11319
11014 404. http://gcc.gnu.org/PR11949
11015 405. http://gcc.gnu.org/PR11662
11016 406. http://gcc.gnu.org/PR11965
11017 407. http://gcc.gnu.org/PR12301
11018 408. http://gcc.gnu.org/PR11717
11019 409. http://gcc.gnu.org/PR11313
11020 410. http://gcc.gnu.org/PR11712
11021 411. http://gcc.gnu.org/PR12166
11022 412. http://gcc.gnu.org/PR12101
11023 413. http://gcc.gnu.org/PR10988
11024 414. http://gcc.gnu.org/PR11805
11025 415. http://gcc.gnu.org/PR11902
11026 416. http://gcc.gnu.org/PR11903
11027 417. http://gcc.gnu.org/bugzilla/
11028 418. http://gcc.gnu.org/PR11890
11029 419. http://gcc.gnu.org/PR12399
11030 420. http://gcc.gnu.org/PR13068
11031 421. http://gcc.gnu.org/PR10060
11032 422. http://gcc.gnu.org/PR10555
11033 423. http://gcc.gnu.org/PR10706
11034 424. http://gcc.gnu.org/PR11496
11035 425. http://gcc.gnu.org/PR11741
11036 426. http://gcc.gnu.org/PR12440
11037 427. http://gcc.gnu.org/PR12632
11038 428. http://gcc.gnu.org/PR12712
11039 429. http://gcc.gnu.org/PR12726
11040 430. http://gcc.gnu.org/PR12890
11041 431. http://gcc.gnu.org/PR12900
11042 432. http://gcc.gnu.org/PR13060
11043 433. http://gcc.gnu.org/PR13289
11044 434. http://gcc.gnu.org/PR13318
11045 435. http://gcc.gnu.org/PR13392
11046 436. http://gcc.gnu.org/PR13574
11047 437. http://gcc.gnu.org/PR13475
11048 438. http://gcc.gnu.org/PR13797
11049 439. http://gcc.gnu.org/PR13824
11050 440. http://gcc.gnu.org/PR8776
11051 441. http://gcc.gnu.org/PR10339
11052 442. http://gcc.gnu.org/PR11350
11053 443. http://gcc.gnu.org/PR12826
11054 444. http://gcc.gnu.org/PR12500
11055 445. http://gcc.gnu.org/PR12941
11056 446. http://gcc.gnu.org/PR12953
11057 447. http://gcc.gnu.org/PR13041
11058 448. http://gcc.gnu.org/PR13507
11059 449. http://gcc.gnu.org/PR13382
11060 450. http://gcc.gnu.org/PR13394
11061 451. http://gcc.gnu.org/PR13400
11062 452. http://gcc.gnu.org/PR13521
11063 453. http://gcc.gnu.org/PR2094
11064 454. http://gcc.gnu.org/PR2294
11065 455. http://gcc.gnu.org/PR5050
11066 456. http://gcc.gnu.org/PR9371
11067 457. http://gcc.gnu.org/PR9546
11068 458. http://gcc.gnu.org/PR10081
11069 459. http://gcc.gnu.org/PR10093
11070 460. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61
11071 461. http://gcc.gnu.org/PR10095
11072 462. http://gcc.gnu.org/PR11554
11073 463. http://gcc.gnu.org/PR12297
11074 464. http://gcc.gnu.org/PR12352
11075 465. http://gcc.gnu.org/PR12438
11076 466. http://gcc.gnu.org/PR12540
11077 467. http://gcc.gnu.org/PR12594
11078 468. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60
11079 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63
11080 470. http://gcc.gnu.org/PR12657
11081 471. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292
11082 472. http://gcc.gnu.org/PR12696
11083 473. http://gcc.gnu.org/PR12815
11084 474. http://gcc.gnu.org/PR12862
11085 475. http://gcc.gnu.org/PR12926
11086 476. http://gcc.gnu.org/PR12967
11087 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html
11088 478. http://gcc.gnu.org/PR12971
11089 479. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328
11090 480. http://gcc.gnu.org/PR13007
11091 481. http://gcc.gnu.org/PR13009
11092 482. http://gcc.gnu.org/PR13057
11093 483. http://gcc.gnu.org/PR13070
11094 484. http://gcc.gnu.org/PR13081
11095 485. http://gcc.gnu.org/PR13239
11096 486. http://gcc.gnu.org/PR13262
11097 487. http://gcc.gnu.org/PR13290
11098 488. http://gcc.gnu.org/PR13323
11099 489. http://gcc.gnu.org/PR13369
11100 490. http://gcc.gnu.org/PR13371
11101 491. http://gcc.gnu.org/PR13445
11102 492. http://gcc.gnu.org/PR13461
11103 493. http://gcc.gnu.org/PR13462
11104 494. http://gcc.gnu.org/PR13478
11105 495. http://gcc.gnu.org/PR13544
11106 496. http://gcc.gnu.org/PR13650
11107 497. http://gcc.gnu.org/PR13683
11108 498. http://gcc.gnu.org/PR13688
11109 499. http://gcc.gnu.org/PR13774
11110 500. http://gcc.gnu.org/PR13884
11111 501. http://gcc.gnu.org/PR10746
11112 502. http://gcc.gnu.org/PR11433
11113 503. http://gcc.gnu.org/PR12633
11114 504. http://gcc.gnu.org/PR13037
11115 505. http://gcc.gnu.org/PR13213
11116 506. http://gcc.gnu.org/PR4490
11117 507. http://gcc.gnu.org/PR12292
11118 508. http://gcc.gnu.org/PR12441
11119 509. http://gcc.gnu.org/PR12943
11120 510. http://gcc.gnu.org/PR13608
11121 511. http://gcc.gnu.org/PR11598
11122 512. http://gcc.gnu.org/PR11793
11123 513. http://gcc.gnu.org/PR12467
11124 514. http://gcc.gnu.org/PR12537
11125 515. http://gcc.gnu.org/PR12496
11126 516. http://gcc.gnu.org/PR12865
11127 517. http://gcc.gnu.org/PR13354
11128 518. http://gcc.gnu.org/PR10467
11129 519. http://gcc.gnu.org/PR11226
11130 520. http://gcc.gnu.org/PR11227
11131 521. http://gcc.gnu.org/PR12644
11132 522. http://gcc.gnu.org/PR13149
11133 523. http://gcc.gnu.org/PR12654
11134 524. http://gcc.gnu.org/PR12965
11135 525. http://gcc.gnu.org/PR13031
11136 526. http://gcc.gnu.org/PR11634
11137 527. http://gcc.gnu.org/PR12158
11138 528. http://gcc.gnu.org/PR11992
11139 529. http://gcc.gnu.org/PR9365
11140 530. http://gcc.gnu.org/PR10392
11141 531. http://gcc.gnu.org/PR11322
11142 532. http://gcc.gnu.org/PR13069
11143 533. http://gcc.gnu.org/PR13302
11144 534. http://gcc.gnu.org/PR13585
11145 535. http://gcc.gnu.org/PR8916
11146 536. http://gcc.gnu.org/PR11576
11147 537. http://gcc.gnu.org/PR13122
11148 538. http://gcc.gnu.org/PR13256
11149 539. http://gcc.gnu.org/PR13373
11150 540. http://gcc.gnu.org/PR12561
11151 541. http://gcc.gnu.org/PR6243
11152 542. http://gcc.gnu.org/PR11397
11153 543. http://gcc.gnu.org/PR12505
11154 544. http://gcc.gnu.org/PR13150
11155 545. http://gcc.gnu.org/PR12666
11156 546. http://gcc.gnu.org/PR12969
11157 547. http://gcc.gnu.org/PR10819
11158 548. http://gcc.gnu.org/PR11612
11159 549. http://gcc.gnu.org/PR13211
11160 550. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4
11161 551. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5
11162 552. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6
11163 553. http://gcc.gnu.org/onlinedocs/
11164 554. mailto:gcc-help@gcc.gnu.org
11165 555. mailto:gcc@gcc.gnu.org
11166 556. http://gcc.gnu.org/lists.html
11167 557. http://www.fsf.org/
11168 558. http://gcc.gnu.org/about.html
11169 559. http://validator.w3.org/check/referer
11170======================================================================
11171http://gcc.gnu.org/gcc-3.2/index.html
11172
11173                             GCC 3.2 Release Series
11174
11175   April 25, 2003
11176
11177   The [1]GNU project and the GCC developers are pleased to announce the
11178   release of GCC 3.2.3.
11179
11180   The purpose of the GCC 3.2 release series is to provide a stable
11181   platform for OS distributors to use building their next releases. A
11182   primary objective was to stabilize the C++ ABI; we believe that the
11183   interface to the compiler and the C++ standard library are now
11184   relatively stable.
11185
11186   Be aware that C++ code compiled by GCC 3.2.x will (in general) not
11187   interoperate with code compiled by GCC 3.1.1 or earlier.
11188
11189   Please refer to our [2]detailed list of news, caveats, and bug-fixes
11190   for further information.
11191
11192Release History
11193
11194   GCC 3.2.3
11195          April 25, 2003 ([3]changes)
11196
11197   GCC 3.2.2
11198          February 5, 2003 ([4]changes)
11199
11200   GCC 3.2.1
11201          November 19, 2002 ([5]changes)
11202
11203   GCC 3.2
11204          August 14, 2002 ([6]changes)
11205
11206References and Acknowledgements
11207
11208   GCC used to stand for the GNU C Compiler, but since the compiler
11209   supports several other languages aside from C, it now stands for the
11210   GNU Compiler Collection.
11211
11212   A list of [7]successful builds is updated as new information becomes
11213   available.
11214
11215   The GCC developers would like to thank the numerous people that have
11216   contributed new features, improvements, bug fixes, and other changes as
11217   well as test results to GCC. This [8]amazing group of volunteers is
11218   what makes GCC successful.
11219
11220   For additional information about GCC please refer to the [9]GCC project
11221   web site or contact the [10]GCC development mailing list.
11222
11223   To obtain GCC please use [11]our mirror sites, or our CVS server.
11224
11225
11226    For questions related to the use of GCC, please consult these web
11227    pages and the [12]GCC manuals. If that fails, the
11228    [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11229    web pages and the development of GCC are welcome on our developer
11230    list at [14]gcc@gcc.gnu.org. All of [15]our lists have public
11231    archives.
11232
11233   Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and
11234   distribution of this entire article is permitted in any medium,
11235   provided this notice is preserved.
11236
11237   These pages are [17]maintained by the GCC team. Last modified
11238   2012-11-02[18].
11239
11240References
11241
11242   1. http://www.gnu.org/
11243   2. http://gcc.gnu.org/gcc-3.2/changes.html
11244   3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
11245   4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2
11246   5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1
11247   6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2
11248   7. http://gcc.gnu.org/gcc-3.2/buildstat.html
11249   8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11250   9. http://gcc.gnu.org/index.html
11251  10. mailto:gcc@gcc.gnu.org
11252  11. http://gcc.gnu.org/mirrors.html
11253  12. http://gcc.gnu.org/onlinedocs/
11254  13. mailto:gcc-help@gcc.gnu.org
11255  14. mailto:gcc@gcc.gnu.org
11256  15. http://gcc.gnu.org/lists.html
11257  16. http://www.fsf.org/
11258  17. http://gcc.gnu.org/about.html
11259  18. http://validator.w3.org/check/referer
11260======================================================================
11261http://gcc.gnu.org/gcc-3.2/changes.html
11262
11263                             GCC 3.2 Release Series
11264                        Changes, New Features, and Fixes
11265
11266   The latest release in the 3.2 release series is [1]GCC 3.2.3.
11267
11268Caveats and New Features
11269
11270  Caveats
11271
11272     * The C++ compiler does not correctly zero-initialize
11273       pointers-to-data members. You must explicitly initialize them. For
11274       example: int S::*m(0); will work, but depending on
11275       default-initialization to zero will not work. This bug cannot be
11276       fixed in GCC 3.2 without inducing unacceptable risks. It will be
11277       fixed in GCC 3.3.
11278     * This GCC release is based on the GCC 3.1 sourcebase, and thus has
11279       all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has
11280       a number of C++ ABI fixes which make its C++ compiler generate
11281       binary code which is incompatible with the C++ compilers found in
11282       earlier GCC releases, including GCC 3.1 and GCC 3.1.1.
11283
11284  Frontend Enhancements
11285
11286    C/C++/Objective-C
11287
11288     * The method of constructing the list of directories to be searched
11289       for header files has been revised. If a directory named by a -I
11290       option is a standard system include directory, the option is
11291       ignored to ensure that the default search order for system
11292       directories and the special treatment of system header files are
11293       not defeated.
11294     * The C and Objective-C compilers no longer accept the "Naming Types"
11295       extension (typedef foo = bar); it was already unavailable in C++.
11296       Code which uses it will need to be changed to use the "typeof"
11297       extension instead: typedef typeof(bar) foo. (We have removed this
11298       extension without a period of deprecation because it has caused the
11299       compiler to crash since version 3.0 and no one noticed until very
11300       recently. Thus we conclude it is not in widespread use.)
11301
11302    C++
11303
11304     * GCC 3.2 fixed serveral differences between the C++ ABI implemented
11305       in GCC and the multi-vendor standard, but more have been found
11306       since the release. 3.2.1 adds a new warning, -Wabi, to warn about
11307       code which is affected by these bugs. We will fix these bugs in
11308       some future release, once we are confident that all have been
11309       found; until then, it is our intention to make changes to the ABI
11310       only if they are necessary for correct compilation of C++, as
11311       opposed to conformance to the ABI documents.
11312     * For details on how to build an ABI compliant compiler for GNU/Linux
11313       systems, check the [3]common C++ ABI page.
11314
11315  New Targets and Target Specific Improvements
11316
11317    IA-32
11318
11319     * Fixed a number of bugs in SSE and MMX intrinsics.
11320     * Fixed common compiler crashes with SSE instruction set enabled
11321       (implied by -march=pentium3, pentium4, athlon-xp)
11322     * __m128 and __m128i is not 128bit aligned when used in structures.
11323
11324    x86-64
11325
11326     * A bug whereby the compiler could generate bad code for bzero has
11327       been fixed.
11328     * ABI fixes (implying ABI incompatibilities with previous version in
11329       some corner cases)
11330     * Fixed prefetch code generation
11331     __________________________________________________________________
11332
11333GCC 3.2.3
11334
11335   3.2.3 is a bug fix release only; there are no new features that were
11336   not present in GCC 3.2.2.
11337
11338  Bug Fixes
11339
11340   This section lists the problem reports (PRs) from GCC's bug tracking
11341   system that are known to be fixed in the 3.2.3 release. This list might
11342   not be complete (that is, it is possible that some PRs that have been
11343   fixed are not listed here), and some of the titles have been changed to
11344   make them more clear.
11345
11346    Internal Compiler Errors (multi-platform)
11347
11348     * [4]3782: (c++) -quiet -fstats produces a segmentation fault in
11349       cc1plus
11350     * [5]6440: (c++) template specializations cause ICE
11351     * [6]7050: (c++) ICE on: (i ? get_string() : throw)
11352     * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c)
11353     * [8]7982: (c++) ICE due to infinite recursion (using STL set)
11354     * [9]8068: exceedingly high (infinite) memory usage
11355     * [10]8178: ICE with __builtin_ffs
11356     * [11]8396: ICE in copy_to_mode_reg, in explow.c
11357     * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c
11358     * [13]9768: ICE when optimizing inline code at -O2
11359     * [14]9798: (c++) Infinite recursion (segfault) in
11360       cp/decl.c:push_using_directive with recursive using directives
11361     * [15]9799: mismatching structure initializer with nested flexible
11362       array member: ICE
11363     * [16]9928: ICE on duplicate enum declaration
11364     * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects
11365       sparc, alpha)
11366     * [18]10352: ICE in find_reloads_toplev
11367     * [19]10336: ICE with -Wunreachable-code
11368
11369    C/optimizer bugs:
11370
11371     * [20]8224: Incorrect joining of signed and unsigned division
11372     * [21]8613: -O2 produces wrong code with builtin strlen and
11373       postincrements
11374     * [22]8828: gcc reports some code is unreachable when it is not
11375     * [23]9226: GCSE breaking argument passing
11376     * [24]9853: miscompilation of non-constant structure initializer
11377     * [25]9797: C99-style struct initializers are miscompiled
11378     * [26]9967: Some standard C function calls should not be replaced
11379       when optimizing for size
11380     * [27]10116: ce2: invalid merge of join_bb in the context of switch
11381       statements
11382     * [28]10171: wrong code for inlined function
11383     * [29]10175: -Wunreachable-code doesn't work for single lines
11384
11385    C++ compiler and library:
11386
11387     * [30]8316: Confusing diagnostic for code that misuses conversion
11388       operators
11389     * [31]9169: filebuf output fails if codecvt<>::out returns noconv
11390     * [32]9420: incomplete type incorrectly reported
11391     * [33]9459: typeof in return type specification of template not
11392       supported
11393     * [34]9507: filebuf::open handles ios_base::ate incorrectly
11394     * [35]9538: Out-of-bounds memory access in streambuf::sputbackc
11395     * [36]9602: Total confusion about template/friend/virtual/abstract
11396     * [37]9993: destructor not called for local object created within and
11397       returned from infinite loop
11398     * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2
11399       system
11400
11401    Java compiler and library:
11402
11403     * [39]9652: libgcj build fails on irix6.5.1[78]
11404     * [40]10144: gas on solaris complains about bad .stabs lines for
11405       java, native as unaffected
11406
11407    x86-specific (Intel/AMD):
11408
11409     * [41]8746: gcc miscompiles Linux kernel ppa driver on x86
11410     * [42]9888: -mcpu=k6 -Os produces out of range loop instructions
11411     * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu
11412       failed
11413     * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib)
11414       failed
11415
11416    SPARC-specific:
11417
11418     * [45]7784: [Sparc] ICE in extract_insn, in recog.c
11419     * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in
11420       unroll.c
11421     * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc
11422     * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in
11423       execute/loop-2d.c
11424     * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc
11425     * [50]9414: Scheduling bug on Ultrasparc
11426     * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64
11427
11428    m68k-specific:
11429
11430     * [52]7248: broken "inclusive or" code
11431     * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1
11432
11433    PowerPC-specific:
11434
11435     * [54]9732: Wrong code with -O2 -fPIC
11436     * [55]10073: ICE: powerpc cannot split insn
11437
11438    Alpha-specific:
11439
11440     * [56]7702: optimization problem on a DEC alpha under OSF1
11441     * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system
11442
11443    HP-specific:
11444
11445     * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275)
11446     * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10
11447       (missing symbol)
11448     * [60]10271: Floating point args don't get reloaded across function
11449       calls with -O2
11450
11451    MIPS specific:
11452
11453     * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in
11454       compile/920501-4.c
11455
11456    CRIS specific:
11457
11458     * [62]10377: gcc-3.2.2 creates bad assembler code for cris
11459
11460    Miscellaneous and minor bugs:
11461
11462     * [63]6955: collect2 says "core dumped" when there is no core
11463     __________________________________________________________________
11464
11465GCC 3.2.2
11466
11467   Beginning with 3.2.2, GCC's Makefile suite supports redirection of make
11468   install by means of the DESTDIR variable. Parts of the GCC tree have
11469   featured that support long before, but now it is available even from
11470   the top level.
11471
11472   Other than that, GCC 3.2.2 is a bug fix release only; there are no new
11473   features that were not present in GCC 3.2.1.
11474
11475  Bug Fixes
11476
11477   On the following i386-based systems GCC 3.2.1 broke the C ABI wrt.
11478   functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped
11479   with FreeBSD 5.0 does not have this problem), Interix, a.out-based
11480   GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI
11481   change, and thus restores ABI-compatibility with previous releases
11482   (except GCC 3.2.1) on these platforms.
11483
11484   This section lists the problem reports (PRs) from GCC's bug tracking
11485   system that are known to be fixed in the 3.2.2 release. This list might
11486   not be complete (that is, it is possible that some PRs that have been
11487   fixed are not listed here) and some of the titles have been changed to
11488   make them more clear.
11489
11490    Internal Compiler Errors (multi-platform)
11491
11492     * [64]5919: (c++) ICE when passing variable array to template
11493       function
11494     * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=)
11495     * [66]7507: ICE with -O2 when address of called function is a
11496       complicated expression
11497     * [67]7622: ICE with nested inline functions if function's address is
11498       taken
11499     * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR
11500       [69]9258)
11501     * [70]8031: (c++) ICE in code comparing typeids and casting from
11502       virtual base
11503     * [71]8275: ICE in simplify_subreg
11504     * [72]8332: (c++) builtin strlen/template interaction causes ICE
11505     * [73]8372: (c++) ICE on explicit call of destructor
11506     * [74]8439: (c, not c++) empty struct causes ICE
11507     * [75]8442: (c++) ICE with nested template classes
11508     * [76]8518: ICE when compiling mplayer ("extern inline" issue)
11509     * [77]8615: (c++) ICE with out-of-range character constant template
11510       argument
11511     * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307
11512     * [79]8799: (c++) ICE: error reporting routines re-entered
11513     * [80]9328: (c++) ICE with typeof(X) for overloaded X
11514     * [81]9465: (preprocessor) cpp -traditional ICE on null bytes
11515
11516    C++ (compiler and library) bugs
11517
11518     * [82]47: scoping in nested classes is broken
11519     * [83]6745: problems with iostream rdbuf() member function
11520     * [84]8214: conversion from const char* const to char* sometimes
11521       accepted illegally
11522     * [85]8493: builtin strlen and overload resolution (same bug as
11523       [86]8332)
11524     * [87]8503: strange behaviour of function types
11525     * [88]8727: compiler confused by inheritance from an anonymous struct
11526     * [89]7445: poor performance of std::locale::classic() in
11527       multi-threaded applications
11528     * [90]8230: mishandling of overflow in vector<T>::resize
11529     * [91]8399: sync_with_stdio(false) breaks unformatted input
11530     * [92]8662: illegal access of private member of unnamed class is
11531       accepted
11532     * [93]8707: "make distclean" fails in libstdc++-v3 directory
11533     * [94]8708: __USE_MALLOC doesn't work
11534     * [95]8790: Use of non-thread-safe strtok in src/localename.cc
11535     * [96]8887: Bug in date formats with --enable-clocale=generic
11536     * [97]9076: Call Frame Instructions are not handled correctly during
11537       unwind operation
11538     * [98]9151: std::setprecision limited to 16 digits when outputting a
11539       double to a stream
11540     * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers
11541     * [100]9269: libstdc++ headers: explicit specialization of function
11542       must precede its first use
11543     * [101]9322: return value of basic_streambuf<>::getloc affected by
11544       locale::global
11545     * [102]9433: segfault in runtime support for dynamic_cast
11546
11547    C and optimizer bugs
11548
11549     * [103]8032: GCC incorrectly initializes static structs that have
11550       flexible arrays
11551     * [104]8639: simple arithmetic expression broken
11552     * [105]8794: optimization improperly eliminates certain expressions
11553     * [106]8832: traditional "asm volatile" code is illegally optimized
11554     * [107]8988: loop optimizer bug: with -O2, code is generated that
11555       segfaults (found on i386, bug present for all platforms)
11556     * [108]9492: structure copy clobbers subsequent stores to structure
11557
11558    Objective-C bugs
11559
11560     * [109]9267: Objective-C parser won't build with newer bison versions
11561       (e.g. 1.875)
11562
11563    Ada bugs
11564
11565     * [110]8344: Ada build problem due to conflict between gcc/final.o,
11566       gcc/ada/final.o
11567
11568    Preprocessor bugs
11569
11570     * [111]8524: _Pragma within macros is improperly expanded
11571     * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with
11572       -fshort-wchar
11573
11574    ARM-specific
11575
11576     * [113]9090: arm ICE with >= -O2; regression from gcc-2.95
11577
11578    x86-specific (Intel/AMD)
11579
11580     * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction)
11581     * [115]8599: loop unroll bug with -march=k6-3
11582     * [116]9506: ABI breakage in structure return (affects BSD and
11583       Cygwin, but not GNU/Linux)
11584
11585    FreeBSD 5.0 specific
11586
11587     * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0
11588
11589    RTEMS-specific
11590
11591     * [118]9292: hppa1.1-rtems configurery problems
11592     * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug
11593     * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue
11594     * [121]9296: gthr-rtems regression
11595     * [122]9316: powerpc-rtems: extending multilibs
11596
11597    HP-PA specific
11598
11599     * [123]9493: ICE with -O2 when building a simple function
11600
11601    Documentation
11602
11603     * [124]7341: hyperlink to gcov in GCC documentation doesn't work
11604     * [125]8947: Please add a warning about "-malign-double" in docs
11605     * [126]7448, [127]8882: typo cleanups
11606     __________________________________________________________________
11607
11608GCC 3.2.1
11609
11610   3.2.1 adds a new warning, -Wabi. This option warns when GNU C++
11611   generates code that is known not to be binary-compatible with the
11612   vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included
11613   in the distribution, for details.
11614
11615   This release also removes an old GCC extension, "naming types", and the
11616   documentation now directs users to use a different GCC extension,
11617   __typeof__, instead. The feature had evidently been broken for a while.
11618
11619   Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and
11620   the new warning there are no new features that were not present in GCC
11621   3.2.
11622
11623   In addition, the previous fix for [128]PR 7445 (poor performance of
11624   std::locale::classic() in multi-threaded applications) was reverted
11625   ("unfixed"), because the "fix" was not thread-safe.
11626
11627  Bug Fixes
11628
11629   This section lists the problem reports (PRs) from GCC's bug tracking
11630   system that are known to be fixed in the 3.2.1 release. This list might
11631   not be complete (that is, it is possible that some PRs that have been
11632   fixed are not listed here). As you can see, the number of bug fixes is
11633   quite large, so it is strongly recommended that users of earlier GCC
11634   3.x releases upgrade to GCC 3.2.1.
11635
11636    Internal Compiler Errors (multi-platform)
11637
11638     * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c
11639     * [130]5661: (c++) ICE instantiating template on array of unknown
11640       size (bad code)
11641     * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on
11642       64-bit platforms
11643     * [132]6994: (c++) ICE in find_function_data
11644     * [133]7150: preprocessor: GCC -dM -E gives an ICE
11645     * [134]7160: ICE when optimizing branches without a return value
11646     * [135]7228: (c++) ICE when using member template and template
11647       function
11648     * [136]7266: (c++) ICE with -pedantic on missing typename
11649     * [137]7353: ICE from use of "Naming Types" extension, see above
11650     * [138]7411: ICE in instantiate_virtual_regs_1, in function.c
11651     * [139]7478: (c++) ICE on static_cast inside template
11652     * [140]7526: preprocessor core dump when _Pragma implies #pragma
11653       dependency
11654     * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803
11655       is a duplicate)
11656     * [143]7754: (c++) ICE on union with template parameter
11657     * [144]7788: (c++) redeclaring a definition as an incomplete class
11658       causes ICE
11659     * [145]8031: (c++) ICE in comptypes, in cp/typeck.c
11660     * [146]8055: preprocessor dies with SIG11 when building FreeBSD
11661       kernel
11662     * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related
11663       variables
11664     * [148]8134: (c++) ICE in force_store_init_value on legal code
11665     * [149]8149: (c++) ICE on incomplete type
11666     * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array
11667       initialization
11668
11669    C++ (compiler and library) bugs
11670
11671     * [151]5607: No pointer adjustment in covariant return types
11672     * [152]6579: Infinite loop with statement expressions in member
11673       initialization
11674     * [153]6803: Default copy constructor bug in GCC 3.1
11675     * [154]7176: g++ confused by friend and static member with same name
11676     * [155]7188: Segfault with template class and recursive (incorrect)
11677       initializer list
11678     * [156]7306: Regression: GCC 3.x fails to compile code with virtual
11679       inheritance if a method has a variable number of arguments
11680     * [157]7461: ctype<char>::classic_table() returns offset array on
11681       Cygwin
11682     * [158]7524: f(const float arg[3]) fails
11683     * [159]7584: Erroneous ambiguous base error on using declaration
11684     * [160]7676: Member template overloading problem
11685     * [161]7679: infinite loop when a right parenthesis is missing
11686     * [162]7811: default locale not taken from environment
11687     * [163]7961: compare( char *) implemented incorrectly in
11688       basic_string<>
11689     * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if
11690       streambuf::underflow() leaves gptr() NULL (dups: [165]8127,
11691       [166]6745)
11692     * [167]8096: deque::at() throws std::range_error instead of
11693       std::out_of_range
11694     * [168]8127: cout << cin.rdbuf() infinite loop
11695     * [169]8218: Excessively large memory consumed for classes with large
11696       array members
11697     * [170]8287: GCC 3.2: Destructor called for non-constructed local
11698       object
11699     * [171]8347: empty vector range used in string construction causes
11700       core dump
11701     * [172]8348: fail() flag is set in istringstream when eof() flag is
11702       set
11703     * [173]8391: regression: infinite loop in cp/decl2.c(finish_file)
11704
11705    C and optimizer bugs
11706
11707     * [174]6627: -fno-align-functions doesn't seem to disable function
11708       alignment
11709     * [175]6631: life_analysis misoptimizes code to initialize fields of
11710       a structure
11711     * [176]7102: unsigned char division results in floating exception
11712     * [177]7120: Run once loop should *always* be unrolled
11713       (pessimization)
11714     * [178]7209: Bug involving array referencing and ?: operator
11715     * [179]7515: invalid inlining of global function with -O3
11716     * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test
11717     * [181]8467: bug in sibling call optimization
11718
11719    Preprocessor bugs
11720
11721     * [182]4890: incorrect line markers from the traditional preprocessor
11722     * [183]7357: -M option omits system headers files (making it the same
11723       as -MM)
11724     * [184]7358: Changes to Sun's make Dependencies
11725     * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as
11726       C headers
11727     * [186]7862: Interrupting GCC -MD removes .d file but not .o
11728     * [187]8190: Failed compilation deletes -MD dependency file
11729     * [188]8524: _Pragma within macro is improperly expanded
11730
11731    x86 specific (Intel/AMD)
11732
11733     * [189]5351: (i686-only) function pass-by-value structure copy
11734       corrupts stack ([190]7591 is a duplicate)
11735     * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with
11736       -march=pentium3/pentium2/athlon (these are all the same underlying
11737       bug, in MMX register use)
11738     * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same
11739       as above?)
11740     * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken
11741     * [199]6981: wrong code in 64-bit manipulation on x86
11742     * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__
11743       macro
11744     * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE
11745       intrinsics are broken
11746     * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with
11747       -march=pentium4
11748     * [203]7693: Typo in i386 mmintrin.h header
11749     * [204]7723: ICE - Pentium3 sse - GCC 3.2
11750     * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse
11751     * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3
11752
11753    PowerPC specific
11754
11755     * [207]5967: GCC bug when profiling nested functions on powerpc
11756     * [208]6984: wrong code generated with -O2, -O3, -Os for do-while
11757       loop on PowerPC
11758     * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5
11759     * [210]7130: miscompiled code for GCC-3.1 on
11760       powerpc-unknown-linux-gnu with -funroll-all-loops
11761     * [211]7133: PowerPC ICE: unrecognizable insn
11762     * [212]7380: ICE in extract_insn, at recog.c:2148
11763     * [213]8252: ICE on Altivec code with optimization turned on
11764     * [214]8451: Altivec ICE in GCC 3.2
11765
11766    HP/PA specific
11767
11768     * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa
11769
11770    SPARC specific
11771
11772     * [216]6668: when using --disable-multilib, libgcc_s.so is installed
11773       in the wrong place on sparc-solaris
11774     * [217]7151: ICE when compiling for UltraSPARC
11775     * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long
11776       double and -O1
11777     * [219]7842: [REGRESSION] SPARC code gen bug
11778
11779    ARM specific
11780
11781     * [220]7856: [arm] invalid offset in constant pool reference
11782     * [221]7967: optimization produces wrong code (ARM)
11783
11784    Alpha specific
11785
11786     * [222]7374: __builtin_fabsl broken on alpha
11787
11788    IBM s390 specific
11789
11790     * [223]7370: ICE in fixup_var_refs_1 on s390x
11791     * [224]7409: loop optimization bug on s390x-linux-gnu
11792     * [225]8232: s390x: ICE when using bcmp with int length argument
11793
11794    SCO specific
11795
11796     * [226]7623: SCO OpenServer build fails with machmode.def: undefined
11797       symbol: BITS_PER_UNIT
11798
11799    m68k/Coldfire specific
11800
11801     * [227]8314: crtbegin, crtend need to be multilib'ed for this
11802       platform
11803
11804    Documentation
11805
11806     * [228]761: Document some undocumented options
11807     * [229]5610: Fix documentation about invoking SSE instructions
11808       (-mfpmath=sse)
11809     * [230]7484: List -Wmissing-declarations as C-only option
11810     * [231]7531: -mcmodel not documented for x86-64
11811     * [232]8120: Update documentation of bad use of ##
11812     __________________________________________________________________
11813
11814GCC 3.2
11815
11816   3.2 is a small bug fix release, but there is a change to the
11817   application binary interface (ABI), hence the change to the second part
11818   of the version number.
11819
11820   The main purpose of the 3.2 release is to correct a couple of problems
11821   in the C++ ABI, with the intention of providing a stable interface
11822   going forward.  Accordingly, 3.2 is only a small change to 3.1.1.
11823
11824  Bug Fixes
11825
11826    C++
11827
11828     * [233]7320: g++ 3.2 relocation problem
11829     * [234]7470: vtable: virtual function pointers not in declaration
11830       order
11831
11832    libstdc++
11833
11834     * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t
11835     * [236]6503, [237]6642, [238]7186: Problems with comparing or
11836       subtracting various types of const and non-const iterators
11837     * [239]7216: ambiguity with basic_iostream::traits_type
11838     * [240]7220: problem with basic_istream::ignore(0,delimiter)
11839     * [241]7222: locale::operator==() doesn't work on std::locale("")
11840     * [242]7286: placement operator delete issue
11841     * [243]7442: cxxabi.h does not match the C++ ABI
11842     * [244]7445: poor performance of std::locale::classic() in
11843       multi-threaded applications
11844
11845    x86-64 specific
11846
11847     * [245]7291: off-by-one in generated inline bzero code for x86-64
11848
11849
11850    For questions related to the use of GCC, please consult these web
11851    pages and the [246]GCC manuals. If that fails, the
11852    [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11853    web pages and the development of GCC are welcome on our developer
11854    list at [248]gcc@gcc.gnu.org. All of [249]our lists have public
11855    archives.
11856
11857   Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and
11858   distribution of this entire article is permitted in any medium,
11859   provided this notice is preserved.
11860
11861   These pages are [251]maintained by the GCC team. Last modified
11862   2012-11-02[252].
11863
11864References
11865
11866   1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
11867   2. http://gcc.gnu.org/gcc-3.1/changes.html
11868   3. http://gcc.gnu.org/gcc-3.2/c++-abi.html
11869   4. http://gcc.gnu.org/PR3782
11870   5. http://gcc.gnu.org/PR6440
11871   6. http://gcc.gnu.org/PR7050
11872   7. http://gcc.gnu.org/PR7741
11873   8. http://gcc.gnu.org/PR7982
11874   9. http://gcc.gnu.org/PR8068
11875  10. http://gcc.gnu.org/PR8178
11876  11. http://gcc.gnu.org/PR8396
11877  12. http://gcc.gnu.org/PR8674
11878  13. http://gcc.gnu.org/PR9768
11879  14. http://gcc.gnu.org/PR9798
11880  15. http://gcc.gnu.org/PR9799
11881  16. http://gcc.gnu.org/PR9928
11882  17. http://gcc.gnu.org/PR10114
11883  18. http://gcc.gnu.org/PR10352
11884  19. http://gcc.gnu.org/PR10336
11885  20. http://gcc.gnu.org/PR8224
11886  21. http://gcc.gnu.org/PR8613
11887  22. http://gcc.gnu.org/PR8828
11888  23. http://gcc.gnu.org/PR9226
11889  24. http://gcc.gnu.org/PR9853
11890  25. http://gcc.gnu.org/PR9797
11891  26. http://gcc.gnu.org/PR9967
11892  27. http://gcc.gnu.org/PR10116
11893  28. http://gcc.gnu.org/PR10171
11894  29. http://gcc.gnu.org/PR10175
11895  30. http://gcc.gnu.org/PR8316
11896  31. http://gcc.gnu.org/PR9169
11897  32. http://gcc.gnu.org/PR9420
11898  33. http://gcc.gnu.org/PR9459
11899  34. http://gcc.gnu.org/PR9507
11900  35. http://gcc.gnu.org/PR9538
11901  36. http://gcc.gnu.org/PR9602
11902  37. http://gcc.gnu.org/PR9993
11903  38. http://gcc.gnu.org/PR10167
11904  39. http://gcc.gnu.org/PR9652
11905  40. http://gcc.gnu.org/PR10144
11906  41. http://gcc.gnu.org/PR8746
11907  42. http://gcc.gnu.org/PR9888
11908  43. http://gcc.gnu.org/PR9638
11909  44. http://gcc.gnu.org/PR9954
11910  45. http://gcc.gnu.org/PR7784
11911  46. http://gcc.gnu.org/PR7796
11912  47. http://gcc.gnu.org/PR8281
11913  48. http://gcc.gnu.org/PR8366
11914  49. http://gcc.gnu.org/PR8726
11915  50. http://gcc.gnu.org/PR9414
11916  51. http://gcc.gnu.org/PR10067
11917  52. http://gcc.gnu.org/PR7248
11918  53. http://gcc.gnu.org/PR8343
11919  54. http://gcc.gnu.org/PR9732
11920  55. http://gcc.gnu.org/PR10073
11921  56. http://gcc.gnu.org/PR7702
11922  57. http://gcc.gnu.org/PR9671
11923  58. http://gcc.gnu.org/PR8694
11924  59. http://gcc.gnu.org/PR9953
11925  60. http://gcc.gnu.org/PR10271
11926  61. http://gcc.gnu.org/PR6362
11927  62. http://gcc.gnu.org/PR10377
11928  63. http://gcc.gnu.org/PR6955
11929  64. http://gcc.gnu.org/PR5919
11930  65. http://gcc.gnu.org/PR7129
11931  66. http://gcc.gnu.org/PR7507
11932  67. http://gcc.gnu.org/PR7622
11933  68. http://gcc.gnu.org/PR7681
11934  69. http://gcc.gnu.org/PR9528
11935  70. http://gcc.gnu.org/PR8031
11936  71. http://gcc.gnu.org/PR8275
11937  72. http://gcc.gnu.org/PR8332
11938  73. http://gcc.gnu.org/PR8372
11939  74. http://gcc.gnu.org/PR8439
11940  75. http://gcc.gnu.org/PR8442
11941  76. http://gcc.gnu.org/PR8518
11942  77. http://gcc.gnu.org/PR8615
11943  78. http://gcc.gnu.org/PR8663
11944  79. http://gcc.gnu.org/PR8799
11945  80. http://gcc.gnu.org/PR9328
11946  81. http://gcc.gnu.org/PR9465
11947  82. http://gcc.gnu.org/PR47
11948  83. http://gcc.gnu.org/PR6745
11949  84. http://gcc.gnu.org/PR8214
11950  85. http://gcc.gnu.org/PR8493
11951  86. http://gcc.gnu.org/PR8332
11952  87. http://gcc.gnu.org/PR8503
11953  88. http://gcc.gnu.org/PR8727
11954  89. http://gcc.gnu.org/PR7445
11955  90. http://gcc.gnu.org/PR8230
11956  91. http://gcc.gnu.org/PR8399
11957  92. http://gcc.gnu.org/PR8662
11958  93. http://gcc.gnu.org/PR8707
11959  94. http://gcc.gnu.org/PR8708
11960  95. http://gcc.gnu.org/PR8790
11961  96. http://gcc.gnu.org/PR8887
11962  97. http://gcc.gnu.org/PR9076
11963  98. http://gcc.gnu.org/PR9151
11964  99. http://gcc.gnu.org/PR9168
11965 100. http://gcc.gnu.org/PR9269
11966 101. http://gcc.gnu.org/PR9322
11967 102. http://gcc.gnu.org/PR9433
11968 103. http://gcc.gnu.org/PR8032
11969 104. http://gcc.gnu.org/PR8639
11970 105. http://gcc.gnu.org/PR8794
11971 106. http://gcc.gnu.org/PR8832
11972 107. http://gcc.gnu.org/PR8988
11973 108. http://gcc.gnu.org/PR9492
11974 109. http://gcc.gnu.org/PR9267
11975 110. http://gcc.gnu.org/PR8344
11976 111. http://gcc.gnu.org/PR8524
11977 112. http://gcc.gnu.org/PR8880
11978 113. http://gcc.gnu.org/PR9090
11979 114. http://gcc.gnu.org/PR8588
11980 115. http://gcc.gnu.org/PR8599
11981 116. http://gcc.gnu.org/PR9506
11982 117. http://gcc.gnu.org/PR9484
11983 118. http://gcc.gnu.org/PR9292
11984 119. http://gcc.gnu.org/PR9293
11985 120. http://gcc.gnu.org/PR9295
11986 121. http://gcc.gnu.org/PR9296
11987 122. http://gcc.gnu.org/PR9316
11988 123. http://gcc.gnu.org/PR9493
11989 124. http://gcc.gnu.org/PR7341
11990 125. http://gcc.gnu.org/PR8947
11991 126. http://gcc.gnu.org/PR7448
11992 127. http://gcc.gnu.org/PR8882
11993 128. http://gcc.gnu.org/PR7445
11994 129. http://gcc.gnu.org/PR2521
11995 130. http://gcc.gnu.org/PR5661
11996 131. http://gcc.gnu.org/PR6419
11997 132. http://gcc.gnu.org/PR6994
11998 133. http://gcc.gnu.org/PR7150
11999 134. http://gcc.gnu.org/PR7160
12000 135. http://gcc.gnu.org/PR7228
12001 136. http://gcc.gnu.org/PR7266
12002 137. http://gcc.gnu.org/PR7353
12003 138. http://gcc.gnu.org/PR7411
12004 139. http://gcc.gnu.org/PR7478
12005 140. http://gcc.gnu.org/PR7526
12006 141. http://gcc.gnu.org/PR7721
12007 142. http://gcc.gnu.org/PR7803
12008 143. http://gcc.gnu.org/PR7754
12009 144. http://gcc.gnu.org/PR7788
12010 145. http://gcc.gnu.org/PR8031
12011 146. http://gcc.gnu.org/PR8055
12012 147. http://gcc.gnu.org/PR8067
12013 148. http://gcc.gnu.org/PR8134
12014 149. http://gcc.gnu.org/PR8149
12015 150. http://gcc.gnu.org/PR8160
12016 151. http://gcc.gnu.org/PR5607
12017 152. http://gcc.gnu.org/PR6579
12018 153. http://gcc.gnu.org/PR6803
12019 154. http://gcc.gnu.org/PR7176
12020 155. http://gcc.gnu.org/PR7188
12021 156. http://gcc.gnu.org/PR7306
12022 157. http://gcc.gnu.org/PR7461
12023 158. http://gcc.gnu.org/PR7524
12024 159. http://gcc.gnu.org/PR7584
12025 160. http://gcc.gnu.org/PR7676
12026 161. http://gcc.gnu.org/PR7679
12027 162. http://gcc.gnu.org/PR7811
12028 163. http://gcc.gnu.org/PR7961
12029 164. http://gcc.gnu.org/PR8071
12030 165. http://gcc.gnu.org/PR8127
12031 166. http://gcc.gnu.org/PR6745
12032 167. http://gcc.gnu.org/PR8096
12033 168. http://gcc.gnu.org/PR8127
12034 169. http://gcc.gnu.org/PR8218
12035 170. http://gcc.gnu.org/PR8287
12036 171. http://gcc.gnu.org/PR8347
12037 172. http://gcc.gnu.org/PR8348
12038 173. http://gcc.gnu.org/PR8391
12039 174. http://gcc.gnu.org/PR6627
12040 175. http://gcc.gnu.org/PR6631
12041 176. http://gcc.gnu.org/PR7102
12042 177. http://gcc.gnu.org/PR7120
12043 178. http://gcc.gnu.org/PR7209
12044 179. http://gcc.gnu.org/PR7515
12045 180. http://gcc.gnu.org/PR7814
12046 181. http://gcc.gnu.org/PR8467
12047 182. http://gcc.gnu.org/PR4890
12048 183. http://gcc.gnu.org/PR7357
12049 184. http://gcc.gnu.org/PR7358
12050 185. http://gcc.gnu.org/PR7602
12051 186. http://gcc.gnu.org/PR7862
12052 187. http://gcc.gnu.org/PR8190
12053 188. http://gcc.gnu.org/PR8524
12054 189. http://gcc.gnu.org/PR5351
12055 190. http://gcc.gnu.org/PR7591
12056 191. http://gcc.gnu.org/PR6845
12057 192. http://gcc.gnu.org/PR7034
12058 193. http://gcc.gnu.org/PR7124
12059 194. http://gcc.gnu.org/PR7174
12060 195. http://gcc.gnu.org/PR7134
12061 196. http://gcc.gnu.org/PR7375
12062 197. http://gcc.gnu.org/PR7390
12063 198. http://gcc.gnu.org/PR6890
12064 199. http://gcc.gnu.org/PR6981
12065 200. http://gcc.gnu.org/PR7242
12066 201. http://gcc.gnu.org/PR7396
12067 202. http://gcc.gnu.org/PR7630
12068 203. http://gcc.gnu.org/PR7693
12069 204. http://gcc.gnu.org/PR7723
12070 205. http://gcc.gnu.org/PR7951
12071 206. http://gcc.gnu.org/PR8146
12072 207. http://gcc.gnu.org/PR5967
12073 208. http://gcc.gnu.org/PR6984
12074 209. http://gcc.gnu.org/PR7114
12075 210. http://gcc.gnu.org/PR7130
12076 211. http://gcc.gnu.org/PR7133
12077 212. http://gcc.gnu.org/PR7380
12078 213. http://gcc.gnu.org/PR8252
12079 214. http://gcc.gnu.org/PR8451
12080 215. http://gcc.gnu.org/PR7250
12081 216. http://gcc.gnu.org/PR6668
12082 217. http://gcc.gnu.org/PR7151
12083 218. http://gcc.gnu.org/PR7335
12084 219. http://gcc.gnu.org/PR7842
12085 220. http://gcc.gnu.org/PR7856
12086 221. http://gcc.gnu.org/PR7967
12087 222. http://gcc.gnu.org/PR7374
12088 223. http://gcc.gnu.org/PR7370
12089 224. http://gcc.gnu.org/PR7409
12090 225. http://gcc.gnu.org/PR8232
12091 226. http://gcc.gnu.org/PR7623
12092 227. http://gcc.gnu.org/PR8314
12093 228. http://gcc.gnu.org/PR761
12094 229. http://gcc.gnu.org/PR5610
12095 230. http://gcc.gnu.org/PR7484
12096 231. http://gcc.gnu.org/PR7531
12097 232. http://gcc.gnu.org/PR8120
12098 233. http://gcc.gnu.org/PR7320
12099 234. http://gcc.gnu.org/PR7470
12100 235. http://gcc.gnu.org/PR6410
12101 236. http://gcc.gnu.org/PR6503
12102 237. http://gcc.gnu.org/PR6642
12103 238. http://gcc.gnu.org/PR7186
12104 239. http://gcc.gnu.org/PR7216
12105 240. http://gcc.gnu.org/PR7220
12106 241. http://gcc.gnu.org/PR7222
12107 242. http://gcc.gnu.org/PR7286
12108 243. http://gcc.gnu.org/PR7442
12109 244. http://gcc.gnu.org/PR7445
12110 245. http://gcc.gnu.org/PR7291
12111 246. http://gcc.gnu.org/onlinedocs/
12112 247. mailto:gcc-help@gcc.gnu.org
12113 248. mailto:gcc@gcc.gnu.org
12114 249. http://gcc.gnu.org/lists.html
12115 250. http://www.fsf.org/
12116 251. http://gcc.gnu.org/about.html
12117 252. http://validator.w3.org/check/referer
12118======================================================================
12119http://gcc.gnu.org/gcc-3.1/index.html
12120
12121                                    GCC 3.1
12122
12123   July 27, 2002
12124
12125   The [1]GNU project and the GCC developers are pleased to announce the
12126   release of GCC 3.1.1.
12127
12128   The links below still apply to GCC 3.1.1.
12129
12130   May 15, 2002
12131
12132   The [2]GNU project and the GCC developers are pleased to announce the
12133   release of GCC 3.1.
12134
12135   GCC used to stand for the GNU C Compiler, but since the compiler
12136   supports several other languages aside from C, it now stands for the
12137   GNU Compiler Collection.
12138
12139   A list of [3]successful builds is updated as new information becomes
12140   available.
12141
12142   The GCC developers would like to thank the numerous people that have
12143   contributed [4]new features, improvements, bug fixes, and other changes
12144   as well as test results to GCC. This [5]amazing group of volunteers is
12145   what makes GCC successful.
12146
12147   For additional information about GCC please refer to the [6]GCC project
12148   web site or contact the [7]GCC development mailing list.
12149
12150   To obtain GCC please use [8]our mirror sites, or our CVS server.
12151     __________________________________________________________________
12152
12153
12154    For questions related to the use of GCC, please consult these web
12155    pages and the [9]GCC manuals. If that fails, the
12156    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12157    web pages and the development of GCC are welcome on our developer
12158    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
12159    archives.
12160
12161   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
12162   distribution of this entire article is permitted in any medium,
12163   provided this notice is preserved.
12164
12165   These pages are [14]maintained by the GCC team. Last modified
12166   2012-11-02[15].
12167
12168References
12169
12170   1. http://www.gnu.org/
12171   2. http://www.gnu.org/
12172   3. http://gcc.gnu.org/gcc-3.1/buildstat.html
12173   4. http://gcc.gnu.org/gcc-3.1/changes.html
12174   5. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12175   6. http://gcc.gnu.org/index.html
12176   7. mailto:gcc@gcc.gnu.org
12177   8. http://gcc.gnu.org/mirrors.html
12178   9. http://gcc.gnu.org/onlinedocs/
12179  10. mailto:gcc-help@gcc.gnu.org
12180  11. mailto:gcc@gcc.gnu.org
12181  12. http://gcc.gnu.org/lists.html
12182  13. http://www.fsf.org/
12183  14. http://gcc.gnu.org/about.html
12184  15. http://validator.w3.org/check/referer
12185======================================================================
12186http://gcc.gnu.org/gcc-3.1/changes.html
12187
12188                             GCC 3.1 Release Series
12189                        Changes, New Features, and Fixes
12190
12191Additional changes in GCC 3.1.1
12192
12193     * A bug related to how structures and unions are returned has been
12194       fixed for powerpc-*-netbsd*.
12195     * An important bug in the implementation of -fprefetch-loop-arrays
12196       has been fixed. Previously the optimization prefetched random
12197       blocks of memory for most targets except for i386.
12198     * The Java compiler now compiles Java programs much faster and also
12199       works with parallel make.
12200     * Nested functions have been fixed for mips*-*-netbsd*.
12201     * Some missing floating point support routines have beed added for
12202       mips*-*-netbsd*.
12203     * This [1]message gives additional information about the bugs fixed
12204       in this release.
12205
12206Caveats
12207
12208     * The -traditional C compiler option has been deprecated and will be
12209       removed in GCC 3.3. (It remains possible to preprocess non-C code
12210       with the traditional preprocessor.)
12211     * The default debugging format for most ELF platforms (including
12212       GNU/Linux and FreeBSD; notable exception is Solaris) has changed
12213       from stabs to DWARF2. This requires GDB 5.1.1 or later.
12214
12215General Optimizer Improvements
12216
12217     * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat,
12218       and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure
12219       for profile driven optimizations.
12220       Options -fprofile-arcs and -fbranch-probabilities can now be used
12221       to improve speed of the generated code by profiling the actual
12222       program behaviour on typical runs. In the absence of profile info
12223       the compiler attempts to guess the profile statically.
12224     * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to
12225       monitor performance of the generated code.
12226       According to the SPECInt2000 results on an AMD Athlon CPU, the code
12227       generated by GCC 3.1 is 6% faster on the average (8.2% faster with
12228       profile feedback) compared to GCC 3.0. The code produced by GCC 3.0
12229       is about 2.1% faster compared to 2.95.3. Tests were done using the
12230       -O2 -march=athlon command-line options.
12231     * Alexandre Oliva, of Red Hat, has generalized the tree inlining
12232       infrastructure developed by CodeSourcery, LLC for the C++ front
12233       end, so that it is now used in the C front end too. Inlining
12234       functions as trees exposes them earlier to the compiler, giving it
12235       more opportunities for optimization.
12236     * Support for data prefetching instructions has been added to the GCC
12237       back end and several targets. A new __builtin_prefetch intrinsic is
12238       available to explicitly insert prefetch instructions and
12239       experimental support for loop array prefetching has been added (see
12240       -fprefetch-loop-array documentation).
12241     * Support for emitting debugging information for macros has been
12242       added for DWARF2. It is activated using -g3.
12243
12244New Languages and Language specific improvements
12245
12246  C/C++
12247
12248     * A few more [4]ISO C99 features.
12249     * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0.
12250     * The preprocessor's symbol table has been merged with the symbol
12251       table of the C, C++ and Objective-C front ends.
12252     * The preprocessor consumes less memory than the preprocessor in GCC
12253       3.0, often significantly so. On normal input files, it typically
12254       consumes less memory than pre-3.0 cccp-based GCC, too.
12255
12256  C++
12257
12258     * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std
12259       was a workaround to allow std compliant code to work with the
12260       non-std compliant libstdc++-v2. libstdc++-v3 is std compliant.
12261     * The C++ ABI has been fixed so that void (A::*)() const is mangled
12262       as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only
12263       affects pointer to cv-qualified member function types.
12264     * The C++ ABI has been changed to correctly handle this code:
12265    struct A {
12266      void operator delete[] (void *, size_t);
12267    };
12268
12269    struct B : public A {
12270    };
12271
12272    new B[10];
12273
12274       The amount of storage allocated for the array will be greater than
12275       it was in 3.0, in order to store the number of elements in the
12276       array, so that the correct size can be passed to operator delete[]
12277       when the array is deleted. Previously, the value passed to operator
12278       delete[] was unpredictable.
12279       This change will only affect code that declares a two-argument
12280       operator delete[] with a second parameter of type size_t in a base
12281       class, and does not override that definition in a derived class.
12282     * The C++ ABI has been changed so that:
12283    struct A {
12284      void operator delete[] (void *, size_t);
12285      void operator delete[] (void *);
12286    };
12287
12288       does not cause unnecessary storage to be allocated when an array of
12289       A objects is allocated.
12290       This change will only affect code that declares both of these forms
12291       of operator delete[], and declared the two-argument form before the
12292       one-argument form.
12293     * The C++ ABI has been changed so that when a parameter is passed by
12294       value, any cleanup for that parameter is performed in the caller,
12295       as specified by the ia64 C++ ABI, rather than the called function
12296       as before. As a result, classes with a non-trivial destructor but a
12297       trivial copy constructor will be passed and returned by invisible
12298       reference, rather than by bitwise copy as before.
12299     * G++ now supports the "named return value optimization": for code
12300       like
12301    A f () {
12302      A a;
12303      ...
12304      return a;
12305    }
12306
12307       G++ will allocate a in the return value slot, so that the return
12308       becomes a no-op. For this to work, all return statements in the
12309       function must return the same variable.
12310     * Improvements to the C++ library are listed in [5]the libstdc++-v3
12311       FAQ.
12312
12313  Objective-C
12314
12315     * Annoying linker warnings (due to incorrect code being generated)
12316       have been fixed.
12317     * If a class method cannot be found, the compiler no longer issues a
12318       warning if a corresponding instance method exists in the root
12319       class.
12320     * Forward @protocol declarations have been fixed.
12321     * Loading of categories has been fixed in certain situations (GNU run
12322       time only).
12323     * The class lookup in the run-time library has been rewritten so that
12324       class method dispatch is more than twice as fast as it used to be
12325       (GNU run time only).
12326
12327  Java
12328
12329     * libgcj now includes RMI, java.lang.ref.*, javax.naming, and
12330       javax.transaction.
12331     * Property files and other system resources can be compiled into
12332       executables which use libgcj using the new gcj --resource feature.
12333     * libgcj has been ported to more platforms. In particular there is
12334       now a mostly-functional mingw32 (Windows) target port.
12335     * JNI and CNI invocation interfaces were implemented, so gcj-compiled
12336       Java code can now be called from a C/C++ application.
12337     * gcj can now use builtin functions for certain known methods, for
12338       instance Math.cos.
12339     * gcj can now automatically remove redundant array-store checks in
12340       some common cases.
12341     * The --no-store-checks optimization option was added. This can be
12342       used to omit runtime store checks for code which is known not to
12343       throw ArrayStoreException
12344     * The following third party interface standards were added to libgcj:
12345       org.w3c.dom and org.xml.sax.
12346     * java.security has been merged with GNU Classpath. The new package
12347       is now JDK 1.2 compliant, and much more complete.
12348     * A bytecode verifier was added to the libgcj interpreter.
12349     * java.lang.Character was rewritten to comply with the Unicode 3.0
12350       standard, and improve performance.
12351     * Partial support for many more locales was added to libgcj.
12352     * Socket timeouts have been implemented.
12353     * libgcj has been merged into a single shared library. There are no
12354       longer separate shared libraries for the garbage collector and
12355       zlib.
12356     * Several performance improvements were made to gcj and libgcj:
12357          + Hash synchronization (thin locks)
12358          + A special allocation path for finalizer-free objects
12359          + Thread-local allocation
12360          + Parallel GC, and other GC tweaks
12361
12362  Fortran
12363
12364   Fortran improvements are listed in [6]the Fortran documentation.
12365
12366  Ada
12367
12368   [7]Ada Core Technologies, Inc, has contributed its GNAT Ada 95 front
12369   end and associated tools. The GNAT compiler fully implements the Ada
12370   language as defined by the ISO/IEC 8652 standard.
12371
12372   Please note that the integration of the Ada front end is still work in
12373   progress.
12374
12375New Targets and Target Specific Improvements
12376
12377     * Hans-Peter Nilsson has contributed a port to [8]MMIX, the CPU
12378       architecture used in new editions of Donald E. Knuth's The Art of
12379       Computer Programming.
12380     * [9]Axis Communications has contributed its port to the CRIS CPU
12381       architecture, used in the ETRAX system-on-a-chip series. See
12382       [10]Axis' developer site for technical information.
12383     * Alexandre Oliva, of Red Hat, has contributed a port to the
12384       [11]SuperH SH5 64-bit RISC microprocessor architecture, extending
12385       the existing SH port.
12386     * UltraSPARC is fully supported in 64-bit mode. The option -m64
12387       enables it.
12388     * For compatibility with the Sun compiler #pragma redefine_extname
12389       has been implemented on Solaris.
12390     * The x86 back end has had some noticeable work done to it.
12391          + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas
12392            Jaeger have contributed a port to the AMD x86-64 architecture.
12393            For more information on x86-64 see [12]http://www.x86-64.org.
12394          + The compiler now supports MMX, 3DNow!, SSE, and SSE2
12395            instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will
12396            enable the respective instruction sets. Intel C++ compatible
12397            MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics
12398            will be added in next major release.
12399          + Following those improvements, targets for Pentium MMX, K6-2,
12400            K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were
12401            added. Refer to the documentation on -march= and -mcpu=
12402            options for details.
12403          + For those targets that support it, -mfpmath=sse will cause the
12404            compiler to generate SSE/SSE2 instructions for floating point
12405            math instead of x87 instructions. Usually, this will lead to
12406            quicker code -- especially on the Pentium 4. Note that only
12407            scalar floating point instructions are used and GCC does not
12408            exploit SIMD features yet.
12409          + Prefetch support has been added to the Pentium III, Pentium 4,
12410            K6-2, K6-3, and Athlon series.
12411          + Code generated for floating point to integer conversions has
12412            been improved leading to better performance of many 3D
12413            applications.
12414     * The PowerPC back end has added 64-bit PowerPC GNU/Linux support.
12415     * C++ support for AIX has been improved.
12416     * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the
12417       PowerPC port supporting the AltiVec programming model (SIMD). The
12418       support, though presently useful, is experimental and is expected
12419       to stabilize for 3.2. The support is written to conform to
12420       Motorola's AltiVec specs. See -maltivec.
12421
12422Obsolete Systems
12423
12424   Support for a number of older systems has been declared obsolete in GCC
12425   3.1. Unless there is activity to revive them, the next release of GCC
12426   will have their sources permanently removed.
12427
12428   All configurations of the following processor architectures have been
12429   declared obsolete:
12430     * MIL-STD-1750A, 1750a-*-*
12431     * AMD A29k, a29k-*-*
12432     * Convex, c*-convex-*
12433     * Clipper, clipper-*-*
12434     * Elxsi, elxsi-*-*
12435     * Intel i860, i860-*-*
12436     * Sun picoJava, pj-*-* and pjl-*-*
12437     * Western Electric 32000, we32k-*-*
12438
12439   Most configurations of the following processor architectures have been
12440   declared obsolete, but we are preserving a few systems which may have
12441   active developers. It is unlikely that the remaining systems will
12442   survive much longer unless we see definite signs of port activity.
12443     * Motorola 88000 except
12444          + Generic a.out, m88k-*-aout*
12445          + Generic SVR4, m88k-*-sysv4
12446          + OpenBSD, m88k-*-openbsd*
12447     * NS32k except
12448          + NetBSD, ns32k-*-netbsd*
12449          + OpenBSD, ns32k-*-openbsd*.
12450     * ROMP except
12451          + OpenBSD, romp-*-openbsd*.
12452
12453   Finally, only some configurations of these processor architectures are
12454   being obsoleted.
12455     * Alpha:
12456          + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka
12457            alpha*-*-osf[45], are still supported.)
12458     * ARM:
12459          + RISCiX, arm-*-riscix*.
12460     * i386:
12461          + 386BSD, i?86-*-bsd*
12462          + Chorus, i?86-*-chorusos*
12463          + DG/UX, i?86-*-dgux*
12464          + FreeBSD 1.x, i?86-*-freebsd1.*
12465          + IBM AIX, i?86-*-aix*
12466          + ISC UNIX, i?86-*-isc*
12467          + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld*
12468          + NEXTstep, i?86-next-*
12469          + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose*
12470          + RTEMS/coff, i?86-*-rtemscoff*
12471          + RTEMS/go32, i?86-go32-rtems*
12472          + Sequent/BSD, i?86-sequent-bsd*
12473          + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and
12474            i?86-sequent-sysv3*
12475          + SunOS, i?86-*-sunos*
12476     * Motorola 68000:
12477          + Altos, m68[k0]*-altos-*
12478          + Apollo, m68[k0]*-apollo-*
12479          + Apple A/UX, m68[k0]*-apple-*
12480          + Bull, m68[k0]*-bull-*
12481          + Convergent, m68[k0]*-convergent-*
12482          + Generic SVR3, m68[k0]*-*-sysv3*
12483          + ISI, m68[k0]*-isi-*
12484          + LynxOS, m68[k0]*-*-lynxos*
12485          + NEXT, m68[k0]*-next-*
12486          + RTEMS/coff, m68[k0]*-*-rtemscoff*
12487          + Sony, m68[k0]*-sony-*
12488     * MIPS:
12489          + DEC Ultrix, mips-*-ultrix* and mips-dec-*
12490          + Generic BSD, mips-*-bsd*
12491          + Generic System V, mips-*-sysv*
12492          + IRIX before version 5, mips-sgi-irix[1234]*
12493          + RiscOS, mips-*-riscos*
12494          + Sony, mips-sony-*
12495          + Tandem, mips-tandem-*
12496     * SPARC:
12497          + RTEMS/a.out, sparc-*-rtemsaout*.
12498
12499Documentation improvements
12500
12501     * The old manual ("Using and Porting the GNU Compiler Collection")
12502       has been replaced by a users manual ("Using the GNU Compiler
12503       Collection") and a separate internals reference manual ("GNU
12504       Compiler Collection Internals").
12505     * More complete and much improved documentation about GCC's internal
12506       representation used by the C and C++ front ends.
12507     * Many cleanups and improvements in general.
12508
12509
12510    For questions related to the use of GCC, please consult these web
12511    pages and the [13]GCC manuals. If that fails, the
12512    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12513    web pages and the development of GCC are welcome on our developer
12514    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
12515    archives.
12516
12517   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
12518   distribution of this entire article is permitted in any medium,
12519   provided this notice is preserved.
12520
12521   These pages are [18]maintained by the GCC team. Last modified
12522   2013-10-31[19].
12523
12524References
12525
12526   1. http://gcc.gnu.org/ml/gcc/2002-07/msg01208.html
12527   2. http://gcc.gnu.org/news/profiledriven.html
12528   3. http://gcc.gnu.org/benchmarks/
12529   4. http://gcc.gnu.org/c99status.html
12530   5. http://gcc.gnu.org/onlinedocs/libstdc++/faq.html
12531   6. http://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html
12532   7. http://www.adacore.com/
12533   8. http://www-cs-faculty.stanford.edu/~knuth/mmix.html
12534   9. http://www.axis.com/
12535  10. http://developer.axis.com/
12536  11. http://www.superh.com/
12537  12. http://www.x86-64.org/
12538  13. http://gcc.gnu.org/onlinedocs/
12539  14. mailto:gcc-help@gcc.gnu.org
12540  15. mailto:gcc@gcc.gnu.org
12541  16. http://gcc.gnu.org/lists.html
12542  17. http://www.fsf.org/
12543  18. http://gcc.gnu.org/about.html
12544  19. http://validator.w3.org/check/referer
12545======================================================================
12546http://gcc.gnu.org/gcc-3.0/gcc-3.0.html
12547
12548                                   GCC 3.0.4
12549
12550   February 20, 2002
12551
12552   The [1]GNU project and the GCC developers are pleased to announce the
12553   release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0
12554   series.
12555
12556   GCC used to stand for the GNU C Compiler, but since the compiler
12557   supports several other languages aside from C, it now stands for the
12558   GNU Compiler Collection.
12559
12560   GCC 3.0.x has several new optimizations, new targets, new languages and
12561   many other new features, relative to GCC 2.95.x. See the [2]new
12562   features page for a more complete list.
12563
12564   A list of [3]successful builds is updated as new information becomes
12565   available.
12566
12567   The GCC developers would like to thank the numerous people that have
12568   contributed new features, test results, bug fixes, etc to GCC. This
12569   [4]amazing group of volunteers is what makes GCC successful.
12570
12571   And finally, we can't in good conscience fail to mention some
12572   [5]caveats to using GCC 3.0.x.
12573
12574   For additional information about GCC please refer to the [6]GCC project
12575   web site or contact the [7]GCC development mailing list.
12576
12577   To obtain GCC please use [8]our mirror sites, or our CVS server.
12578     __________________________________________________________________
12579
12580Previous 3.0.x Releases
12581
12582   December 20, 2001: GCC 3.0.3 has been released.
12583   October 25, 2001: GCC 3.0.2 has been released.
12584   August 20, 2001: GCC 3.0.1 has been released.
12585   June 18, 2001: GCC 3.0 has been released.
12586
12587
12588    For questions related to the use of GCC, please consult these web
12589    pages and the [9]GCC manuals. If that fails, the
12590    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12591    web pages and the development of GCC are welcome on our developer
12592    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
12593    archives.
12594
12595   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
12596   distribution of this entire article is permitted in any medium,
12597   provided this notice is preserved.
12598
12599   These pages are [14]maintained by the GCC team. Last modified
12600   2012-11-02[15].
12601
12602References
12603
12604   1. http://www.gnu.org/
12605   2. http://gcc.gnu.org/gcc-3.0/features.html
12606   3. http://gcc.gnu.org/gcc-3.0/buildstat.html
12607   4. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12608   5. http://gcc.gnu.org/gcc-3.0/caveats.html
12609   6. http://gcc.gnu.org/index.html
12610   7. mailto:gcc@gcc.gnu.org
12611   8. http://gcc.gnu.org/mirrors.html
12612   9. http://gcc.gnu.org/onlinedocs/
12613  10. mailto:gcc-help@gcc.gnu.org
12614  11. mailto:gcc@gcc.gnu.org
12615  12. http://gcc.gnu.org/lists.html
12616  13. http://www.fsf.org/
12617  14. http://gcc.gnu.org/about.html
12618  15. http://validator.w3.org/check/referer
12619======================================================================
12620http://gcc.gnu.org/gcc-3.0/features.html
12621
12622                              GCC 3.0 New Features
12623
12624Additional changes in GCC 3.0.4
12625
12626     * GCC 3.0 now supports newer versions of the [1]NetBSD operating
12627       system, which use the ELF object file format, on x86 processors.
12628     * Correct debugging information is generated from functions that have
12629       lines from multiple files (e.g. yacc output).
12630     * A fix for whitespace handling in the -traditional preprocessor,
12631       which can affect Fortran.
12632     * Fixes to the exception handling runtime.
12633     * More fixes for bad code generation in C++.
12634     * A fix for shared library generation under AIX 4.3.
12635     * Documentation updates.
12636     * Port of GCC to Tensilica's Xtensa processor contributed.
12637     * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link).
12638
12639Additional changes in GCC 3.0.3
12640
12641     * A fix to correct an accidental change to the PowerPC ABI.
12642     * Fixes for bad code generation on a variety of architectures.
12643     * Improvements to the debugging information generated for C++
12644       classes.
12645     * Fixes for bad code generation in C++.
12646     * A fix to avoid crashes in the C++ demangler.
12647     * A fix to the C++ standard library to avoid buffer overflows.
12648     * Miscellaneous improvements for a variety of architectures.
12649
12650Additional changes in GCC 3.0.2
12651
12652     * Fixes for bad code generation during loop unrolling.
12653     * Fixes for bad code generation by the sibling call optimization.
12654     * Minor improvements to x86 code generation.
12655     * Implementation of function descriptors in C++ vtables for IA64.
12656     * Numerous minor bug-fixes.
12657
12658Additional changes in GCC 3.0.1
12659
12660     * C++ fixes for incorrect code-generation.
12661     * Improved cross-compiling support for the C++ standard library.
12662     * Fixes for some embedded targets that worked in GCC 2.95.3, but not
12663       in GCC 3.0.
12664     * Fixes for various exception-handling bugs.
12665     * A port to the S/390 architecture.
12666
12667General Optimizer Improvements
12668
12669     * [2]Basic block reordering pass.
12670     * New if-conversion pass with support for conditional (predicated)
12671       execution.
12672     * New tail call and sibling call elimination optimizations.
12673     * New register renaming pass.
12674     * New (experimental) [3]static single assignment (SSA) representation
12675       support.
12676     * New dead-code elimination pass implemented using the SSA
12677       representation.
12678     * [4]Global null pointer test elimination.
12679     * [5]Global code hoisting/unification.
12680     * More builtins and optimizations for stdio.h, string.h and old BSD
12681       functions, as well as for ISO C99 functions.
12682     * New builtin __builtin_expect for giving hints to the branch
12683       predictor.
12684
12685New Languages and Language specific improvements
12686
12687     * The GNU Compiler for the Java(TM) language (GCJ) is now integrated
12688       and supported, including the run-time library containing most
12689       common non-GUI Java classes, a bytecode interpreter, and the Boehm
12690       conservative garbage collector. Many bugs have been fixed. GCJ can
12691       compile Java source or Java bytecodes to either native code or Java
12692       class files, and supports native methods written in either the
12693       standard JNI or the more efficient and convenient CNI.
12694     * Here is a [6]partial list of C++ improvements, both new features
12695       and those no longer supported.
12696     * New C++ ABI. On the IA-64 platform GCC is capable of
12697       inter-operating with other IA-64 compilers.
12698     * The new ABI also significantly reduces the size of symbol and debug
12699       information.
12700     * New [7]C++ support library and many C++ bug fixes, vastly improving
12701       our conformance to the ISO C++ standard.
12702     * New [8]inliner for C++.
12703     * Rewritten C preprocessor, integrated into the C, C++ and Objective
12704       C compilers, with very many improvements including ISO C99 support
12705       and [9]improvements to dependency generation.
12706     * Support for more [10]ISO C99 features.
12707     * Many improvements to support for checking calls to format functions
12708       such as printf and scanf, including support for ISO C99 format
12709       features, extensions from the Single Unix Specification and GNU
12710       libc 2.2, checking of strfmon formats and features to assist in
12711       auditing for format string security bugs.
12712     * New warnings for C code that may have undefined semantics because
12713       of violations of sequence point rules in the C standard (such as a
12714       = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall.
12715     * Additional warning option -Wfloat-equal.
12716     * Improvements to -Wtraditional.
12717     * Fortran improvements are listed in [11]the Fortran documentation.
12718
12719New Targets and Target Specific Improvements
12720
12721     * New x86 back end, generating much improved code.
12722     * Support for a generic i386-elf target contributed.
12723     * New option to emit x86 assembly code using Intel style syntax
12724       (-mintel-syntax).
12725     * HPUX 11 support contributed.
12726     * Improved PowerPC code generation, including scheduled prologue and
12727       epilogue.
12728     * Port of GCC to Intel's IA-64 processor contributed.
12729     * Port of GCC to Motorola's MCore 210 and 340 contributed.
12730     * New unified back-end for Arm, Thumb and StrongArm contributed.
12731     * Port of GCC to Intel's XScale processor contributed.
12732     * Port of GCC to Atmel's AVR microcontrollers contributed.
12733     * Port of GCC to Mitsubishi's D30V processor contributed.
12734     * Port of GCC to Matsushita's AM33 processor (a member of the MN10300
12735       processor family) contributed.
12736     * Port of GCC to Fujitsu's FR30 processor contributed.
12737     * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed.
12738     * Port of GCC to Sun's picoJava processor core contributed.
12739
12740Documentation improvements
12741
12742     * Substantially rewritten and improved C preprocessor manual.
12743     * Many improvements to other documentation.
12744     * Manpages for gcc, cpp and gcov are now generated automatically from
12745       the master Texinfo manual, eliminating the problem of manpages
12746       being out of date. (The generated manpages are only extracts from
12747       the full manual, which is provided in Texinfo form, from which
12748       info, HTML, other formats and a printed manual can be generated.)
12749     * Generated info files are included in the release tarballs alongside
12750       their Texinfo sources, avoiding problems on some platforms with
12751       building makeinfo as part of the GCC distribution.
12752
12753Other significant improvements
12754
12755     * Garbage collection used internally by the compiler for most memory
12756       allocation instead of obstacks.
12757     * Lengauer and Tarjan algorithm used for computing dominators in the
12758       CFG. This algorithm can be significantly faster and more space
12759       efficient than our older algorithm.
12760     * gccbug script provided to assist in submitting bug reports to our
12761       bug tracking system. (Bug reports previously submitted directly to
12762       our mailing lists, for which you received no bug tracking number,
12763       should be submitted again using gccbug if you can reproduce the
12764       problem with GCC 3.0.)
12765     * The internal libgcc library is [12]built as a shared library on
12766       systems that support it.
12767     * Extensive testsuite included with GCC, with many new tests. In
12768       addition to tests for GCC bugs that have been fixed, many tests
12769       have been added for language features, compiler warnings and
12770       builtin functions.
12771     * Additional language-independent warning options -Wpacked, -Wpadded,
12772       -Wunreachable-code and -Wdisabled-optimization.
12773     * Target-independent options -falign-functions, -falign-loops and
12774       -falign-jumps.
12775
12776   Plus a great many bug fixes and almost all the [13]features found in
12777   GCC 2.95.
12778
12779
12780    For questions related to the use of GCC, please consult these web
12781    pages and the [14]GCC manuals. If that fails, the
12782    [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12783    web pages and the development of GCC are welcome on our developer
12784    list at [16]gcc@gcc.gnu.org. All of [17]our lists have public
12785    archives.
12786
12787   Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and
12788   distribution of this entire article is permitted in any medium,
12789   provided this notice is preserved.
12790
12791   These pages are [19]maintained by the GCC team. Last modified
12792   2013-12-03[20].
12793
12794References
12795
12796   1. http://www.netbsd.org/
12797   2. http://gcc.gnu.org/news/reorder.html
12798   3. http://gcc.gnu.org/news/ssa.html
12799   4. http://gcc.gnu.org/news/null.html
12800   5. http://gcc.gnu.org/news/unify.html
12801   6. http://gcc.gnu.org/gcc-3.0/c++features.html
12802   7. http://gcc.gnu.org/libstdc++/
12803   8. http://gcc.gnu.org/news/inlining.html
12804   9. http://gcc.gnu.org/news/dependencies.html
12805  10. http://gcc.gnu.org/c99status.html
12806  11. http://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
12807  12. http://gcc.gnu.org/gcc-3.0/libgcc.html
12808  13. http://gcc.gnu.org/gcc-2.95/features.html
12809  14. http://gcc.gnu.org/onlinedocs/
12810  15. mailto:gcc-help@gcc.gnu.org
12811  16. mailto:gcc@gcc.gnu.org
12812  17. http://gcc.gnu.org/lists.html
12813  18. http://www.fsf.org/
12814  19. http://gcc.gnu.org/about.html
12815  20. http://validator.w3.org/check/referer
12816======================================================================
12817http://gcc.gnu.org/gcc-3.0/caveats.html
12818
12819                                GCC 3.0 Caveats
12820
12821     * -fstrict-aliasing is now part of -O2 and higher optimization
12822       levels. This allows the compiler to assume the strictest aliasing
12823       rules applicable to the language being compiled. For C and C++,
12824       this activates optimizations based on the type of expressions. This
12825       optimization may thus break old, non-compliant code.
12826     * Enumerations are now properly promoted to int in function
12827       parameters and function returns. Normally this change is not
12828       visible, but when using -fshort-enums this is an ABI change.
12829     * The undocumented extension that allowed C programs to have a label
12830       at the end of a compound statement has been deprecated and may be
12831       removed in a future version. Programs that now generate a warning
12832       about this may be fixed by adding a null statement (a single
12833       semicolon) after the label.
12834     * The poorly documented extension that allowed string constants in C,
12835       C++ and Objective C to contain unescaped newlines has been
12836       deprecated and may be removed in a future version. Programs using
12837       this extension may be fixed in several ways: the bare newline may
12838       be replaced by \n, or preceded by \n\, or string concatenation may
12839       be used with the bare newline preceded by \n" and " placed at the
12840       start of the next line.
12841     * The Chill compiler is not included in GCC 3.0, because of the lack
12842       of a volunteer to convert it to use garbage collection.
12843     * Certain non-standard iostream methods from earlier versions of
12844       libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach,
12845       ostream::form, and istream::gets.
12846     * The new C++ ABI is not yet fully supported by current (as of
12847       2001-07-01) releases and development versions of GDB, or any
12848       earlier versions. There is a problem setting breakpoints by line
12849       number, and other related issues that have been fixed in GCC 3.0
12850       but not yet handled in GDB:
12851       [1]http://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
12852
12853
12854    For questions related to the use of GCC, please consult these web
12855    pages and the [2]GCC manuals. If that fails, the
12856    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12857    web pages and the development of GCC are welcome on our developer
12858    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
12859
12860   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
12861   distribution of this entire article is permitted in any medium,
12862   provided this notice is preserved.
12863
12864   These pages are [7]maintained by the GCC team. Last modified
12865   2012-11-02[8].
12866
12867References
12868
12869   1. http://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
12870   2. http://gcc.gnu.org/onlinedocs/
12871   3. mailto:gcc-help@gcc.gnu.org
12872   4. mailto:gcc@gcc.gnu.org
12873   5. http://gcc.gnu.org/lists.html
12874   6. http://www.fsf.org/
12875   7. http://gcc.gnu.org/about.html
12876   8. http://validator.w3.org/check/referer
12877======================================================================
12878http://gcc.gnu.org/gcc-2.95/index.html
12879
12880                                    GCC 2.95
12881
12882   March 16, 2001: The GNU project and the GCC developers are pleased to
12883   announce the release of GCC version 2.95.3.
12884
12885Release History
12886
12887   GCC 2.95.3
12888          March 16, 2001
12889
12890   GCC 2.95.2
12891          October 27, 1999
12892
12893   GCC 2.95.1
12894          August 19, 1999
12895
12896   GCC 2.95
12897          July 31, 1999. This is the first release of GCC since the April
12898          1999 GCC/EGCS reunification and includes nearly a year's worth
12899          of new development and bugfixes.
12900
12901References and Acknowledgements
12902
12903   GCC used to stand for the GNU C Compiler, but since the compiler
12904   supports several other languages aside from C, it now stands for the
12905   GNU Compiler Collection.
12906
12907   The whole suite has been extensively [1]regression tested and
12908   [2]package tested. It should be reliable and suitable for widespread
12909   use.
12910
12911   The compiler has several new optimizations, new targets, new languages
12912   and other new features. See the [3]new features page for a more
12913   complete list of new features found in the GCC 2.95 releases.
12914
12915   The sources include installation instructions in both HTML and
12916   plaintext forms in the install directory in the distribution. However,
12917   the most up to date [4]installation instructions and [5]build/test
12918   status are on the web pages. We will update those pages as new
12919   information becomes available.
12920
12921   The GCC developers would like to thank the numerous people that have
12922   contributed new features, test results, bugfixes, etc to GCC. This
12923   [6]amazing group of volunteers is what makes GCC successful.
12924
12925   And finally, we can't in good conscience fail to mention some
12926   [7]caveats to using GCC 2.95.
12927
12928   Download GCC 2.95 from one of our many [8]mirror sites.
12929
12930   For additional information about GCC please see the [9]GCC project web
12931   server or contact the [10]GCC development mailing list.
12932
12933
12934    For questions related to the use of GCC, please consult these web
12935    pages and the [11]GCC manuals. If that fails, the
12936    [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12937    web pages and the development of GCC are welcome on our developer
12938    list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
12939    archives.
12940
12941   Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
12942   distribution of this entire article is permitted in any medium,
12943   provided this notice is preserved.
12944
12945   These pages are [16]maintained by the GCC team. Last modified
12946   2012-11-02[17].
12947
12948References
12949
12950   1. http://gcc.gnu.org/gcc-2.95/regress.html
12951   2. http://gcc.gnu.org/gcc-2.95/othertest.html
12952   3. http://gcc.gnu.org/gcc-2.95/features.html
12953   4. http://gcc.gnu.org/install/
12954   5. http://gcc.gnu.org/gcc-2.95/buildstat.html
12955   6. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12956   7. http://gcc.gnu.org/gcc-2.95/caveats.html
12957   8. http://gcc.gnu.org/mirrors.html
12958   9. http://gcc.gnu.org/index.html
12959  10. mailto:gcc@gcc.gnu.org
12960  11. http://gcc.gnu.org/onlinedocs/
12961  12. mailto:gcc-help@gcc.gnu.org
12962  13. mailto:gcc@gcc.gnu.org
12963  14. http://gcc.gnu.org/lists.html
12964  15. http://www.fsf.org/
12965  16. http://gcc.gnu.org/about.html
12966  17. http://validator.w3.org/check/referer
12967======================================================================
12968http://gcc.gnu.org/gcc-2.95/features.html
12969
12970                             GCC 2.95 New Features
12971
12972     * General Optimizer Improvements:
12973          + [1]Localized register spilling to improve speed and code
12974            density especially on small register class machines.
12975          + [2]Global CSE using lazy code motion algorithms.
12976          + [3]Improved global constant/copy propagation.
12977          + [4]Improved control flow graph analysis and manipulation.
12978          + [5]Local dead store elimination.
12979          + [6]Memory Load hoisting/store sinking in loops.
12980          + [7]Type based alias analysis is enabled by default. Note this
12981            feature will expose bugs in the Linux kernel. Please refer to
12982            the FAQ (as shipped with GCC 2.95) for additional information
12983            on this issue.
12984          + Major revamp of GIV detection, combination and simplification
12985            to improve loop performance.
12986          + Major improvements to register allocation and reloading.
12987     * New Languages and Language specific improvements
12988          + [8]Many C++ improvements.
12989          + [9]Many Fortran improvements.
12990          + [10]Java front-end has been integrated. [11]runtime library is
12991            available separately.
12992          + [12]ISO C99 support
12993          + [13]Chill front-end and runtime has been integrated.
12994          + Boehm garbage collector support in libobjc.
12995          + More support for various pragmas which appear in vendor
12996            include files
12997     * New Targets and Target Specific Improvements
12998          + [14]SPARC backend rewrite.
12999          + -mschedule=8000 will optimize code for PA8000 class
13000            processors; -mpa-risc-2-0 will generate code for PA2.0
13001            processors
13002          + Various micro-optimizations for the ia32 port. K6
13003            optimizations
13004          + Compiler will attempt to align doubles in the stack on the
13005            ia32 port
13006          + Alpha EV6 support
13007          + PowerPC 750
13008          + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403.
13009            -mcpu=e603e was added to do -mcpu=603e and -msoft-float.
13010          + c3x, c4x
13011          + HyperSPARC
13012          + SparcLite86x
13013          + sh4
13014          + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix,
13015            arm-linux)
13016          + vxWorks targets include support for vxWorks threads
13017          + StrongARM 110 and ARM9 support added. ARM Scheduling
13018            parameters rewritten.
13019          + Various changes to the MIPS port to avoid assembler macros,
13020            which in turn improves performance
13021          + Various performance improvements to the i960 port.
13022          + Major rewrite of ns32k port
13023     * Other significant improvements
13024          + [15]Ability to dump cfg information and display it using vcg.
13025          + The new faster scheme for fixing vendor header files is
13026            enabled by default.
13027          + Experimental internationalization support.
13028          + multibyte character support
13029          + Some compile-time speedups for pathological problems
13030          + Better support for complex types
13031     * Plus the usual mountain of bugfixes
13032     * Core compiler is based on the gcc2 development tree from Sept 30,
13033       1998, so we have all of the [16]features found in GCC 2.8.
13034
13035Additional Changes in GCC 2.95.1
13036
13037     * Generic bugfixes and improvements
13038          + Various documentation fixes related to the GCC/EGCS merger.
13039          + Fix memory management bug which could lead to spurious aborts,
13040            core dumps or random parsing errors in the compiler.
13041          + Fix a couple bugs in the dwarf1 and dwarf2 debug record
13042            support.
13043          + Fix infinite loop in the CSE optimizer.
13044          + Avoid undefined behavior in compiler FP emulation code
13045          + Fix install problem when prefix is overridden on the make
13046            install command.
13047          + Fix problem with unwanted installation of assert.h on some
13048            systems.
13049          + Fix problem with finding the wrong assembler in a single tree
13050            build.
13051          + Avoid increasing the known alignment of a register that is
13052            already known to be a pointer.
13053     * Platform specific bugfixes and improvements
13054          + Codegen bugfix for prologue/epilogue for cpu32 target.
13055          + Fix long long code generation bug for the Coldfire target.
13056          + Fix various aborts in the SH compiler.
13057          + Fix bugs in libgcc support library for the SH.
13058          + Fix alpha ev6 code generation bug.
13059          + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on
13060            AIX platforms.
13061          + Fix -fpic code generation bug for rs6000/ppc svr4 targets.
13062          + Fix varargs/stdarg code generation bug for rs6000/ppc svr4
13063            targets.
13064          + Fix weak symbol handling for rs6000/ppc svr4 targets.
13065          + Fix various problems with 64bit code generation for the
13066            rs6000/ppc port.
13067          + Fix codegen bug which caused tetex to be mis-compiled on the
13068            x86.
13069          + Fix compiler abort in new cfg code exposed by x86 port.
13070          + Fix out of range array reference in code convert flat
13071            registers to the x87 stacked FP register file.
13072          + Fix minor vxworks configuration bug.
13073          + Fix return type of bsearch for SunOS 4.x.
13074     * Language & Runtime specific fixes.
13075          + The G++ signature extension has been deprecated. It will be
13076            removed in the next major release of G++. Use of signatures
13077            will result in a warning from the compiler.
13078          + Several bugs relating to templates and namespaces were fixed.
13079          + A bug that caused crashes when combining templates with -g on
13080            DWARF1 platforms was fixed.
13081          + Pointers-to-members, virtual functions, and multiple
13082            inheritance should now work together correctly.
13083          + Some code-generation bugs relating to function try blocks were
13084            fixed.
13085          + G++ is a little bit more lenient with certain archaic
13086            constructs than in GCC 2.95.
13087          + Fix to prevent shared library version #s from bring truncated
13088            to 1 digit
13089          + Fix missing std:: in the libstdc++ library.
13090          + Fix stream locking problems in libio.
13091          + Fix problem in java compiler driver.
13092
13093Additional Changes in GCC 2.95.2
13094
13095   The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While
13096   the optimizations performed by -fstrict-aliasing are valid according to
13097   the C and C++ standards, the optimization have caused some problems,
13098   particularly with old non-conforming code.
13099
13100   The GCC developers are experimenting with ways to warn users about code
13101   which violates the C/C++ standards, but those warnings are not ready
13102   for widespread use at this time. Rather than wait for those warnings
13103   the GCC developers have chosen to disable -fstrict-aliasing by default
13104   for the GCC 2.95.2 release.
13105
13106   We strongly encourage developers to find and fix code which violates
13107   the C/C++ standards as -fstrict-aliasing may be enabled by default in
13108   future releases. Use the option -fstrict-aliasing to re-enable these
13109   optimizations.
13110     * Generic bugfixes and improvements
13111          + Fix incorrectly optimized memory reference in global common
13112            subexpression elimination (GCSE) optimization pass.
13113          + Fix code generation bug in regmove.c in which it could
13114            incorrectly change a "const" value.
13115          + Fix bug in optimization of conditionals involving volatile
13116            memory references.
13117          + Avoid over-allocation of stack space for some procedures.
13118          + Fixed bug in the compiler which caused incorrect optimization
13119            of an obscure series of bit manipulations, shifts and
13120            arithmetic.
13121          + Fixed register allocator bug which caused teTeX to be
13122            mis-compiled on SPARC targets.
13123          + Avoid incorrect optimization of degenerate case statements for
13124            certain targets such as the ARM.
13125          + Fix out of range memory reference in the jump optimizer.
13126          + Avoid dereferencing null pointer in fix-header.
13127          + Fix test for GCC specific features so that it is possible to
13128            bootstrap with gcc-2.6.2 and older versions of GCC.
13129          + Fix typo in scheduler which could potentially cause out of
13130            range memory accesses.
13131          + Avoid incorrect loop reversal which caused incorrect code for
13132            certain loops on PowerPC targets.
13133          + Avoid incorrect optimization of switch statements on certain
13134            targets (for example the ARM).
13135     * Platform specific bugfixes and improvements
13136          + Work around bug in Sun V5.0 compilers which caused bootstrap
13137            comparison failures on SPARC targets.
13138          + Fix SPARC backend bug which caused aborts in final.c.
13139          + Fix sparc-hal-solaris2* configuration fragments.
13140          + Fix bug in sparc block profiling.
13141          + Fix obscure code generation bug for the PARISC targets.
13142          + Define __STDC_EXT__ for HPUX configurations.
13143          + Various POWERPC64 code generation bugfixes.
13144          + Fix abort for PPC targets using ELF (ex GNU/Linux).
13145          + Fix collect2 problems for AIX targets.
13146          + Correct handling of .file directive for PPC targets.
13147          + Fix bug in fix_trunc x86 patterns.
13148          + Fix x86 port to correctly pop the FP stack for functions that
13149            return structures in memory.
13150          + Fix minor bug in strlen x86 pattern.
13151          + Use stabs debugging instead of dwarf1 for x86-solaris targets.
13152          + Fix template repository code to handle leading underscore in
13153            mangled names.
13154          + Fix weak/weak alias support for OpenBSD.
13155          + GNU/Linux for the ARM has C++ compatible include files.
13156     * Language & Runtime specific fixes.
13157          + Fix handling of constructor attribute in the C front-end which
13158            caused problems building the Chill runtime library on some
13159            targets.
13160          + Fix minor problem merging type qualifiers in the C front-end.
13161          + Fix aliasing bug for pointers and references (C/C++).
13162          + Fix incorrect "non-constant initializer bug" when -traditional
13163            or -fwritable-strings is enabled.
13164          + Fix build error for Chill front-end on SunOS.
13165          + Do not complain about duplicate instantiations when using
13166            -frepo (C++).
13167          + Fix array bounds handling in C++ front-end which caused
13168            problems with dwarf debugging information in some
13169            circumstances.
13170          + Fix minor namespace problem.
13171          + Fix problem linking java programs.
13172
13173Additional Changes in GCC 2.95.3
13174
13175     * Generic bugfixes and improvements
13176          + Fix numerous problems that caused incorrect optimization in
13177            the register reloading code.
13178          + Fix numerous problems that caused incorrect optimization in
13179            the loop optimizer.
13180          + Fix aborts in the functions build_insn_chain and scan_loops
13181            under some circumstances.
13182          + Fix an alias analysis bug.
13183          + Fix an infinite compilation bug in the combiner.
13184          + A few problems with complex number support have been fixed.
13185          + It is no longer possible for gcc to act as a fork bomb when
13186            installed incorrectly.
13187          + The -fpack-struct option should be recognized now.
13188          + Fixed a bug that caused incorrect code to be generated due to
13189            a lost stack adjustment.
13190     * Platform specific bugfixes and improvements
13191          + Support building ARM toolchains hosted on Windows.
13192          + Fix attribute calculations in ARM toolchains.
13193          + arm-linux support has been improved.
13194          + Fix a PIC failure on sparc targets.
13195          + On ix86 targets, the regparm attribute should now work
13196            reliably.
13197          + Several updates for the h8300 port.
13198          + Fix problem building libio with glibc 2.2.
13199
13200
13201    For questions related to the use of GCC, please consult these web
13202    pages and the [17]GCC manuals. If that fails, the
13203    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13204    web pages and the development of GCC are welcome on our developer
13205    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
13206    archives.
13207
13208   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
13209   distribution of this entire article is permitted in any medium,
13210   provided this notice is preserved.
13211
13212   These pages are [22]maintained by the GCC team. Last modified
13213   2012-11-02[23].
13214
13215References
13216
13217   1. http://gcc.gnu.org/news/spill.html
13218   2. http://gcc.gnu.org/news/lcm.html
13219   3. http://gcc.gnu.org/news/cprop.html
13220   4. http://gcc.gnu.org/news/cfg.html
13221   5. http://gcc.gnu.org/news/dse.html
13222   6. http://gcc.gnu.org/news/hoist.html
13223   7. http://gcc.gnu.org/news/alias.html
13224   8. http://gcc.gnu.org/gcc-2.95/c++features.html
13225   9. http://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
13226  10. http://gcc.gnu.org/java/gcj-announce.txt
13227  11. http://gcc.gnu.org/news/javaannounce.html
13228  12. http://gcc.gnu.org/c99status.html
13229  13. http://gcc.gnu.org/news/chill.html
13230  14. http://gcc.gnu.org/news/sparc.html
13231  15. http://gcc.gnu.org/news/egcs-vcg.html
13232  16. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13233  17. http://gcc.gnu.org/onlinedocs/
13234  18. mailto:gcc-help@gcc.gnu.org
13235  19. mailto:gcc@gcc.gnu.org
13236  20. http://gcc.gnu.org/lists.html
13237  21. http://www.fsf.org/
13238  22. http://gcc.gnu.org/about.html
13239  23. http://validator.w3.org/check/referer
13240======================================================================
13241http://gcc.gnu.org/gcc-2.95/caveats.html
13242
13243                                GCC 2.95 Caveats
13244
13245     * GCC 2.95 will issue an error for invalid asm statements that had
13246       been silently accepted by earlier versions of the compiler. This is
13247       particularly noticeable when compiling older versions of the Linux
13248       kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95)
13249       for more information on this issue.
13250     * GCC 2.95 implements type based alias analysis to disambiguate
13251       memory references. Some programs, particularly the Linux kernel
13252       violate ANSI/ISO aliasing rules and therefore may not operate
13253       correctly when compiled with GCC 2.95. Please refer to the FAQ (as
13254       shipped with GCC 2.95) for more information on this issue.
13255     * GCC 2.95 has a known bug in its handling of complex variables for
13256       64bit targets. Instead of silently generating incorrect code, GCC
13257       2.95 will issue a fatal error for situations it can not handle.
13258       This primarily affects the Fortran community as Fortran makes more
13259       use of complex variables than C or C++.
13260     * GCC 2.95 has an integrated libstdc++, but does not have an
13261       integrated libg++. Furthermore old libg++ releases will not work
13262       with GCC 2.95. You can retrieve a recent copy of libg++ from the
13263       [1]GCC ftp server.
13264       Note most C++ programs only need libstdc++.
13265     * Exception handling may not work with shared libraries, particularly
13266       on alphas, hppas, rs6000/powerpc and mips based platforms.
13267       Exception handling is known to work on x86 GNU/Linux platforms with
13268       shared libraries.
13269     * In general, GCC 2.95 is more rigorous about rejecting invalid C++
13270       code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0,
13271       or EGCS 1.1. As a result it may be necessary to fix C++ code before
13272       it will compile with GCC 2.95.
13273     * G++ is also converting toward the ISO C++ standard; as a result
13274       code which was previously valid (and thus accepted by other
13275       compilers and older versions of g++) may no longer be accepted. The
13276       flag -fpermissive may allow some non-conforming code to compile
13277       with GCC 2.95.
13278     * GCC 2.95 compiled C++ code is not binary compatible with EGCS
13279       1.1.x, EGCS 1.0.x or GCC 2.8.x.
13280     * GCC 2.95 does not have changes from the GCC 2.8 tree that were made
13281       between Sept 30, 1998 and April 30, 1999 (the official end of the
13282       GCC 2.8 project). Future GCC releases will include all the changes
13283       from the defunct GCC 2.8 sources.
13284
13285
13286    For questions related to the use of GCC, please consult these web
13287    pages and the [2]GCC manuals. If that fails, the
13288    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13289    web pages and the development of GCC are welcome on our developer
13290    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
13291
13292   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
13293   distribution of this entire article is permitted in any medium,
13294   provided this notice is preserved.
13295
13296   These pages are [7]maintained by the GCC team. Last modified
13297   2012-11-02[8].
13298
13299References
13300
13301   1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz
13302   2. http://gcc.gnu.org/onlinedocs/
13303   3. mailto:gcc-help@gcc.gnu.org
13304   4. mailto:gcc@gcc.gnu.org
13305   5. http://gcc.gnu.org/lists.html
13306   6. http://www.fsf.org/
13307   7. http://gcc.gnu.org/about.html
13308   8. http://validator.w3.org/check/referer
13309======================================================================
13310http://gcc.gnu.org/egcs-1.1/index.html
13311
13312                                    EGCS 1.1
13313
13314   September 3, 1998: We are pleased to announce the release of EGCS 1.1.
13315   December 1, 1998: We are pleased to announce the release of EGCS 1.1.1.
13316   March 15, 1999: We are pleased to announce the release of EGCS 1.1.2.
13317
13318   EGCS is a free software project to further the development of the GNU
13319   compilers using an open development environment.
13320
13321   EGCS 1.1 is a major new release of the EGCS compiler system. It has
13322   been [1]extensively tested and is believed to be stable and suitable
13323   for widespread use.
13324
13325   EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8
13326   development sources; it contains all of the new features found in GCC
13327   2.8.1 as well as all new development from GCC up to June 6, 1998.
13328
13329   EGCS 1.1 also contains many improvements and features not found in GCC
13330   or in older versions of EGCS:
13331     * Global common subexpression elimination and global constant/copy
13332       propagation (aka [2]gcse)
13333     * Ongoing improvements to the [3]alias analysis support to allow for
13334       better optimizations throughout the compiler.
13335     * Vastly improved [4]C++ compiler and integrated C++ runtime
13336       libraries.
13337     * Fixes for the /tmp symlink race security problems.
13338     * New targets including mips16, arm-thumb and 64 bit PowerPC.
13339     * Improvements to GNU Fortran (g77) compiler and runtime library made
13340       since g77 version 0.5.23.
13341
13342   See the [5]new features page for a more complete list of new features
13343   found in EGCS 1.1 releases.
13344
13345   EGCS 1.1.1 is a minor update to fix several serious problems in EGCS
13346   1.1:
13347     * General improvements and fixes
13348          + Avoid some stack overflows when compiling large functions.
13349          + Avoid incorrect loop invariant code motions.
13350          + Fix some core dumps on Linux kernel code.
13351          + Bring back the imake -Di386 and friends fix from EGCS 1.0.2.
13352          + Fix code generation problem in gcse.
13353          + Various documentation related fixes.
13354     * g++/libstdc++ improvements and fixes
13355          + MT safe EH fix for setjmp/longjmp based exception handling.
13356          + Fix a few bad interactions between optimization and exception
13357            handling.
13358          + Fixes for demangling of template names starting with "__".
13359          + Fix a bug that would fail to run destructors in some cases
13360            with -O2.
13361          + Fix 'new' of classes with virtual bases.
13362          + Fix crash building Qt on the Alpha.
13363          + Fix failure compiling WIFEXITED macro on GNU/Linux.
13364          + Fix some -frepo failures.
13365     * g77 and libf2c improvements and fixes
13366          + Various documentation fixes.
13367          + Avoid compiler crash on RAND intrinsic.
13368          + Fix minor bugs in makefiles exposed by BSD make programs.
13369          + Define _XOPEN_SOURCE for libI77 build to avoid potential
13370            problems on some 64-bit systems.
13371          + Fix problem with implicit endfile on rewind.
13372          + Fix spurious recursive I/O errors.
13373     * platform specific improvements and fixes
13374          + Match all versions of UnixWare7.
13375          + Do not assume x86 SVR4 or UnixWare targets can handle stabs.
13376          + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion
13377            from unsigned ints to double precision floats.
13378          + Fix ARM ABI issue with NetBSD.
13379          + Fix a few arm code generation bugs.
13380          + Fixincludes will fix additional broken SCO OpenServer header
13381            files.
13382          + Fix a m68k backend bug which caused invalid offsets in reg+d
13383            addresses.
13384          + Fix problems with 64bit AIX 4.3 support.
13385          + Fix handling of long longs for varargs/stdarg functions on the
13386            ppc.
13387          + Minor fixes to CPP predefines for Windows.
13388          + Fix code generation problems with gpr<->fpr copies for 64bit
13389            ppc.
13390          + Fix a few coldfire code generation bugs.
13391          + Fix some more header file problems on SunOS 4.x.
13392          + Fix assert.h handling for RTEMS.
13393          + Fix Windows handling of TREE_SYMBOL_REFERENCED.
13394          + Fix x86 compiler abort in reg-stack pass.
13395          + Fix cygwin/windows problem with section attributes.
13396          + Fix Alpha code generation problem exposed by SMP Linux
13397            kernels.
13398          + Fix typo in m68k 32->64bit integer conversion.
13399          + Make sure target libraries build with -fPIC for PPC & Alpha
13400            targets.
13401
13402   EGCS 1.1.2 is a minor update to fix several serious problems in EGCS
13403   1.1.1:
13404     * General improvements and fixes
13405          + Fix bug in loop optimizer which caused the SPARC (and
13406            potentially other) ports to segfault.
13407          + Fix infinite recursion in alias analysis and combiner code.
13408          + Fix bug in regclass preferencing.
13409          + Fix incorrect loop reversal which caused incorrect code to be
13410            generated for several targets.
13411          + Fix return value for builtin memcpy.
13412          + Reduce compile time for certain loops which exposed quadratic
13413            behavior in the loop optimizer.
13414          + Fix bug which caused volatile memory to be written multiple
13415            times when only one write was needed/desired.
13416          + Fix compiler abort in caller-save.c
13417          + Fix combiner bug which caused incorrect code generation for
13418            certain division by constant operations.
13419          + Fix incorrect code generation due to a bug in range check
13420            optimizations.
13421          + Fix incorrect code generation due to mis-handling of clobbered
13422            values in CSE.
13423          + Fix compiler abort/segfault due to incorrect register
13424            splitting when unrolling loops.
13425          + Fix code generation involving autoincremented addresses with
13426            ternary operators.
13427          + Work around bug in the scheduler which caused qt to be
13428            mis-compiled on some platforms.
13429          + Fix code generation problems with -fshort-enums.
13430          + Tighten security for temporary files.
13431          + Improve compile time for codes which make heavy use of
13432            overloaded functions.
13433          + Fix multiply defined constructor/destructor symbol problems.
13434          + Avoid setting bogus RPATH environment variable during
13435            bootstrap.
13436          + Avoid GNU-make dependencies in the texinfo subdir.
13437          + Install CPP wrapper script in $(prefix)/bin if --enable-cpp.
13438            --enable-cpp=<dirname> can be used to specify an additional
13439            install directory for the cpp wrapper script.
13440          + Fix CSE bug which caused incorrect label-label refs to appear
13441            on some platforms.
13442          + Avoid linking in EH routines from libgcc if they are not
13443            needed.
13444          + Avoid obscure bug in aliasing code.
13445          + Fix bug in weak symbol handling.
13446     * Platform-specific improvements and fixes
13447          + Fix detection of PPro/PII on Unixware 7.
13448          + Fix compiler segfault when building spec99 and other programs
13449            for SPARC targets.
13450          + Fix code-generation bugs for integer and floating point
13451            conditional move instructions on the PPro/PII.
13452          + Use fixincludes to fix byteorder problems on i?86-*-sysv.
13453          + Fix build failure for the arc port.
13454          + Fix floating point format configuration for i?86-gnu port.
13455          + Fix problems with hppa1.0-hp-hpux10.20 configuration when
13456            threads are enabled.
13457          + Fix coldfire code generation bugs.
13458          + Fix "unrecognized insn" problems for Alpha and PPC ports.
13459          + Fix h8/300 code generation problem with floating point values
13460            in memory.
13461          + Fix unrecognized insn problems for the m68k port.
13462          + Fix namespace-pollution problem for the x86 port.
13463          + Fix problems with old assembler on x86 NeXT systems.
13464          + Fix PIC code-generation problems for the SPARC port.
13465          + Fix minor bug with LONG_CALLS in PowerPC SVR4 support.
13466          + Fix minor ISO namespace violation in Alpha varargs/stdarg
13467            support.
13468          + Fix incorrect "braf" instruction usage for the SH port.
13469          + Fix minor bug in va-sh which prevented its use with -ansi.
13470          + Fix problems recognizing and supporting FreeBSD.
13471          + Handle OpenBSD systems correctly.
13472          + Minor fixincludes fix for Digital UNIX 4.0B.
13473          + Fix problems with ctors/dtors in SCO shared libraries.
13474          + Abort instead of generating incorrect code for PPro/PII
13475            floating point conditional moves.
13476          + Avoid multiply defined symbols on GNU/Linux systems using
13477            libc-5.4.xx.
13478          + Fix abort in alpha compiler.
13479     * Fortran-specific fixes
13480          + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year
13481            is in the documented, non-Y2K-compliant range of 0-99, instead
13482            of being returned as 100 in the year 2000.
13483          + Fix the `Date_and_Time' intrinsic (in libg2c) to return the
13484            milliseconds value properly in Values(8).
13485          + Fix the `LStat' intrinsic (in libg2c) to return device-ID
13486            information properly in SArray(7).
13487
13488   Each release includes installation instructions in both HTML and
13489   plaintext forms (see the INSTALL directory in the toplevel directory of
13490   the distribution). However, we also keep the most up to date
13491   [6]installation instructions and [7]build/test status on our web page.
13492   We will update those pages as new information becomes available.
13493
13494   The EGCS project would like to thank the numerous people that have
13495   contributed new features, test results, bugfixes, etc. This [8]amazing
13496   group of volunteers is what makes EGCS successful.
13497
13498   And finally, we can't in good conscience fail to mention some
13499   [9]caveats to using EGCS 1.1.
13500
13501   Download EGCS from egcs.cygnus.com (USA California).
13502
13503   The EGCS 1.1 release is also available on many mirror sites.
13504   [10]Goto mirror list to find a closer site.
13505
13506
13507    For questions related to the use of GCC, please consult these web
13508    pages and the [11]GCC manuals. If that fails, the
13509    [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13510    web pages and the development of GCC are welcome on our developer
13511    list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
13512    archives.
13513
13514   Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
13515   distribution of this entire article is permitted in any medium,
13516   provided this notice is preserved.
13517
13518   These pages are [16]maintained by the GCC team. Last modified
13519   2012-11-02[17].
13520
13521References
13522
13523   1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html
13524   2. http://gcc.gnu.org/news/gcse.html
13525   3. http://gcc.gnu.org/news/alias.html
13526   4. http://gcc.gnu.org/egcs-1.1/c++features.html
13527   5. http://gcc.gnu.org/egcs-1.1/features.html
13528   6. http://gcc.gnu.org/install/
13529   7. http://gcc.gnu.org/egcs-1.1/buildstat.html
13530   8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13531   9. http://gcc.gnu.org/egcs-1.1/caveats.html
13532  10. http://gcc.gnu.org/mirrors.html
13533  11. http://gcc.gnu.org/onlinedocs/
13534  12. mailto:gcc-help@gcc.gnu.org
13535  13. mailto:gcc@gcc.gnu.org
13536  14. http://gcc.gnu.org/lists.html
13537  15. http://www.fsf.org/
13538  16. http://gcc.gnu.org/about.html
13539  17. http://validator.w3.org/check/referer
13540======================================================================
13541http://gcc.gnu.org/egcs-1.1/features.html
13542
13543                             EGCS 1.1 new features
13544
13545     * Integrated GNU Fortran (g77) compiler and runtime library with
13546       improvements, based on g77 version 0.5.23.
13547     * Vast improvements in the C++ compiler; so many they have [1]page of
13548       their own!
13549     * Compiler implements [2]global common subexpression elimination and
13550       global copy/constant propagation.
13551     * More major improvements in the [3]alias analysis code.
13552     * More major improvements in the exception handling code to improve
13553       performance, lower static overhead and provide the infrastructure
13554       for future improvements.
13555     * The infamous /tmp symlink race security problems have been fixed.
13556     * The regmove optimization pass has been nearly completely rewritten
13557       to improve performance of generated code.
13558     * The compiler now recomputes register usage information before local
13559       register allocation. By providing more accurate information to the
13560       priority based allocator, we get better register allocation.
13561     * The register reloading phase of the compiler optimizes spill code
13562       much better than in previous releases.
13563     * Some bad interactions between the register allocator and
13564       instruction scheduler have been fixed, resulting in much better
13565       code for certain programs. Additionally, we have tuned the
13566       scheduler in various ways to improve performance of generated code
13567       for some architectures.
13568     * The compiler's branch shortening algorithms have been significantly
13569       improved to work better on targets which align jump targets.
13570     * The compiler now supports -Os to prefer optimizing for code space
13571       over optimizing for code speed.
13572     * The compiler will now totally eliminate library calls which compute
13573       constant values. This primarily helps targets with no integer
13574       div/mul support and targets without floating point support.
13575     * The compiler now supports an extensive "--help" option.
13576     * cpplib has been greatly improved and may be suitable for limited
13577       use.
13578     * Memory footprint for the compiler has been significantly reduced
13579       for some pathological cases.
13580     * The time to build EGCS has been improved for certain targets
13581       (particularly the alpha and mips platforms).
13582     * Many infrastructure improvements throughout the compiler, plus the
13583       usual mountain of bugfixes and minor improvements.
13584     * Target dependent improvements:
13585          + SPARC port now includes V8 plus and V9 support as well as
13586            performance tuning for Ultra class machines. The SPARC port
13587            now uses the Haifa scheduler.
13588          + Alpha port has been tuned for the EV6 processor and has an
13589            optimized expansion of memcpy/bzero. The Alpha port now uses
13590            the Haifa scheduler.
13591          + RS6000/PowerPC: support for the Power64 architecture and AIX
13592            4.3. The RS6000/PowerPC port now uses the Haifa scheduler.
13593          + x86: Alignment of static store data and jump targets is per
13594            Intel recommendations now. Various improvements throughout the
13595            x86 port to improve performance on Pentium processors
13596            (including improved epilogue sequences for Pentium chips and
13597            backend improvements which should help register allocation on
13598            all x86 variants. Conditional move support has been fixed and
13599            enabled for PPro processors. The x86 port also better supports
13600            64bit operations now. Unixware 7, a System V Release 5 target,
13601            is now supported and SCO OpenServer targets can support GAS.
13602          + MIPS has improved multiply/multiply-add support and now
13603            includes mips16 ISA support.
13604          + M68k has many micro-optimizations and Coldfire fixes.
13605     * Core compiler is based on the GCC development tree from June 9,
13606       1998, so we have all of the [4]features found in GCC 2.8.
13607
13608
13609    For questions related to the use of GCC, please consult these web
13610    pages and the [5]GCC manuals. If that fails, the
13611    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13612    web pages and the development of GCC are welcome on our developer
13613    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
13614
13615   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
13616   distribution of this entire article is permitted in any medium,
13617   provided this notice is preserved.
13618
13619   These pages are [10]maintained by the GCC team. Last modified
13620   2012-11-02[11].
13621
13622References
13623
13624   1. http://gcc.gnu.org/egcs-1.1/c++features.html
13625   2. http://gcc.gnu.org/news/gcse.html
13626   3. http://gcc.gnu.org/news/alias.html
13627   4. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13628   5. http://gcc.gnu.org/onlinedocs/
13629   6. mailto:gcc-help@gcc.gnu.org
13630   7. mailto:gcc@gcc.gnu.org
13631   8. http://gcc.gnu.org/lists.html
13632   9. http://www.fsf.org/
13633  10. http://gcc.gnu.org/about.html
13634  11. http://validator.w3.org/check/referer
13635======================================================================
13636http://gcc.gnu.org/egcs-1.1/caveats.html
13637
13638                                EGCS 1.1 Caveats
13639
13640     * EGCS has an integrated libstdc++, but does not have an integrated
13641       libg++. Furthermore old libg++ releases will not work with EGCS; HJ
13642       Lu has made a libg++-2.8.1.2 snapshot available which may work with
13643       EGCS.
13644       Note most C++ programs only need libstdc++.
13645     * Exception handling may not work with shared libraries, particularly
13646       on alphas, hppas, rs6000/powerpc and mips based platforms.
13647       Exception handling is known to work on x86-linux platforms with
13648       shared libraries.
13649     * Some versions of the Linux kernel have bugs which prevent them from
13650       being compiled or from running when compiled by EGCS. See the FAQ
13651       (as shipped with EGCS 1.1) for additional information.
13652     * In general, EGCS is more rigorous about rejecting invalid C++ code
13653       or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As
13654       a result it may be necessary to fix C++ code before it will compile
13655       with EGCS.
13656     * G++ is also converting toward the ISO C++ standard; as a result
13657       code which was previously valid (and thus accepted by other
13658       compilers and older versions of g++) may no longer be accepted.
13659     * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x
13660       or GCC 2.8.x due to changes necessary to support thread safe
13661       exception handling.
13662
13663
13664    For questions related to the use of GCC, please consult these web
13665    pages and the [1]GCC manuals. If that fails, the
13666    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13667    web pages and the development of GCC are welcome on our developer
13668    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
13669
13670   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
13671   distribution of this entire article is permitted in any medium,
13672   provided this notice is preserved.
13673
13674   These pages are [6]maintained by the GCC team. Last modified
13675   2012-11-02[7].
13676
13677References
13678
13679   1. http://gcc.gnu.org/onlinedocs/
13680   2. mailto:gcc-help@gcc.gnu.org
13681   3. mailto:gcc@gcc.gnu.org
13682   4. http://gcc.gnu.org/lists.html
13683   5. http://www.fsf.org/
13684   6. http://gcc.gnu.org/about.html
13685   7. http://validator.w3.org/check/referer
13686======================================================================
13687http://gcc.gnu.org/egcs-1.0/index.html
13688
13689                                    EGCS 1.0
13690
13691   December 3, 1997: We are pleased to announce the release of EGCS 1.0.
13692   January 6, 1998: We are pleased to announce the release of EGCS 1.0.1.
13693   March 16, 1998: We are pleased to announce the release of EGCS 1.0.2.
13694   May 15, 1998 We are pleased to announce the release of EGCS 1.0.3.
13695
13696   EGCS is a collaborative effort involving several groups of hackers
13697   using an open development model to accelerate development and testing
13698   of GNU compilers and runtime libraries.
13699
13700   An important goal of EGCS is to allow wide scale testing of
13701   experimental features and optimizations; therefore, EGCS contains some
13702   features and optimizations which are still under development. However,
13703   EGCS has been carefully tested and should be comparable in quality to
13704   most GCC releases.
13705
13706   EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8
13707   development sources; it contains nearly all of the new features found
13708   in GCC 2.8.
13709
13710   EGCS 1.0 also contains many improvements and features not found in GCC
13711   2.7 and even the GCC 2.8 series (which was released after the original
13712   EGCS 1.0 release).
13713     * Integrated C++ runtime libraries, including support for most major
13714       GNU/Linux systems!
13715     * The integrated libstdc++ library includes a verbatim copy of SGI's
13716       STL release.
13717     * Integrated GNU Fortran compiler.
13718     * New instruction scheduler.
13719     * New alias analysis code.
13720
13721   See the [1]new features page for a more complete list of new features.
13722
13723   EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few
13724   critical bugs and add support for Red Hat 5.0 Linux. Changes since the
13725   EGCS 1.0 release:
13726     * Add support for Red Hat 5.0 Linux and better support for Linux
13727       systems using glibc2.
13728       Many programs failed to link when compiled with EGCS 1.0 on Red Hat
13729       5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should
13730       fix these problems.
13731     * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception
13732       handling interfaces.
13733       To avoid future compatibility problems, we strongly urge anyone who
13734       is planning on distributing shared libraries that contain C++ code
13735       to upgrade to EGCS 1.0.1 first.
13736       Soon after EGCS 1.0 was released, the GCC developers made some
13737       incompatible changes in libgcc's exception handling interfaces.
13738       These changes were needed to solve problems on some platforms. This
13739       means that GCC 2.8.0, when released, will not be seamlessly
13740       compatible with shared libraries built by EGCS 1.0. The reason is
13741       that the libgcc.a in GCC 2.8.0 will not contain a function needed
13742       by the old interface.
13743       The result of this is that there may be compatibility problems with
13744       shared libraries built by EGCS 1.0 when used with GCC 2.8.0.
13745       With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface,
13746       and libgcc.a has the support routines for both the old and the new
13747       interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed,
13748       and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed).
13749       The maintainers of GCC 2.x have decided against including seamless
13750       support for the old interface in 2.8.0, since it was never
13751       "official", so to avoid future compatibility problems we recommend
13752       against distributing any shared libraries built by EGCS 1.0 that
13753       contain C++ code (upgrade to 1.0.1 and use that).
13754     * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends.
13755       The x86 changes fix code generation errors exposed when building
13756       glibc2 and the usual GNU/Linux dynamic linker (ld.so).
13757       The hppa change fixes a compiler abort when configured for use with
13758       RTEMS.
13759       The MIPS changes fix problems with the definition of LONG_MAX on
13760       newer systems, allow for command line selection of the target ABI,
13761       and fix one code generation problem.
13762       The rs6000/ppc change fixes some problems with passing structures
13763       to varargs/stdarg functions.
13764     * A few machine independent bugfixes, mostly to fix code generation
13765       errors when building Linux kernels or glibc.
13766     * Fix a few critical exception handling and template bugs in the C++
13767       compiler.
13768     * Fix Fortran namelist bug on alphas.
13769     * Fix build problems on x86-solaris systems.
13770
13771   EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several
13772   serious problems in EGCS 1.0.1.
13773     * General improvements and fixes
13774          + Memory consumption significantly reduced, especially for
13775            templates and inline functions.
13776          + Fix various problems with glibc2.1.
13777          + Fix loop optimization bug exposed by rs6000/ppc port.
13778          + Fix to avoid potential code generation problems in jump.c.
13779          + Fix some undefined symbol problems in dwarf1 debug support.
13780     * g++/libstdc++ improvements and fixes
13781          + libstdc++ in the EGCS release has been updated and should be
13782            link compatible with libstdc++-2.8.
13783          + Various fixes in libio/libstdc++ to work better on GNU/Linux
13784            systems.
13785          + Fix problems with duplicate symbols on systems that do not
13786            support weak symbols.
13787          + Memory corruption bug and undefined symbols in bastring have
13788            been fixed.
13789          + Various exception handling fixes.
13790          + Fix compiler abort for very long thunk names.
13791     * g77 improvements and fixes
13792          + Fix compiler crash for omitted bound in Fortran CASE
13793            statement.
13794          + Add missing entries to g77 lang-options.
13795          + Fix problem with -fpedantic in the g77 compiler.
13796          + Fix "backspace" problem with g77 on alphas.
13797          + Fix x86 backend problem with Fortran literals and -fpic.
13798          + Fix some of the problems with negative subscripts for g77 on
13799            alphas.
13800          + Fixes for Fortran builds on cygwin32/mingw32.
13801     * platform specific improvements and fixes
13802          + Fix long double problems on x86 (exposed by glibc).
13803          + x86 ports define i386 again to keep imake happy.
13804          + Fix exception handling support on NetBSD ports.
13805          + Several changes to collect2 to fix many problems with AIX.
13806          + Define __ELF__ for GNU/Linux on rs6000.
13807          + Fix -mcall-linux problem on GNU/Linux on rs6000.
13808          + Fix stdarg/vararg problem for GNU/Linux on rs6000.
13809          + Allow autoconf to select a proper install problem on AIX 3.1.
13810          + m68k port support includes -mcpu32 option as well as cpu32
13811            multilibs.
13812          + Fix stdarg bug for irix6.
13813          + Allow EGCS to build on irix5 without the gnu assembler.
13814          + Fix problem with static linking on sco5.
13815          + Fix bootstrap on sco5 with native compiler.
13816          + Fix for abort building newlib on H8 target.
13817          + Fix fixincludes handling of math.h on SunOS.
13818          + Minor fix for Motorola 3300 m68k systems.
13819
13820   EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few
13821   problems reported by Red Hat for builds of Red Hat 5.1.
13822     * Generic bugfixes:
13823          + Fix a typo in the libio library which resulted in incorrect
13824            behavior of istream::get.
13825          + Fix the Fortran negative array index problem.
13826          + Fix a major problem with the ObjC runtime thread support
13827            exposed by glibc2.
13828          + Reduce memory consumption of the Haifa scheduler.
13829     * Target specific bugfixes:
13830          + Fix one x86 floating point code generation bug exposed by
13831            glibc2 builds.
13832          + Fix one x86 internal compiler error exposed by glibc2 builds.
13833          + Fix profiling bugs on the Alpha.
13834          + Fix ImageMagick & emacs 20.2 build problems on the Alpha.
13835          + Fix rs6000/ppc bug when converting values from integer types
13836            to floating point types.
13837
13838   The EGCS 1.0 releases include installation instructions in both HTML
13839   and plaintext forms (see the INSTALL directory in the toplevel
13840   directory of the distribution). However, we also keep the most up to
13841   date [2]installation instructions and [3]build/test status on our web
13842   page. We will update those pages as new information becomes available.
13843
13844   And, we can't in good conscience fail to mention some [4]caveats to
13845   using EGCS.
13846
13847   Update: Big thanks to Stanford for providing a high speed link for
13848   downloading EGCS (go.cygnus.com)!
13849
13850   Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com
13851   (USA California -- High speed link provided by Stanford).
13852
13853   The EGCS 1.0 release is also available many mirror sites.
13854   [5]Goto mirror list to find a closer site
13855
13856   We'd like to thank the numerous people that have contributed new
13857   features, test results, bugfixes, etc. Unfortunately, they're far too
13858   numerous to mention by name.
13859
13860
13861    For questions related to the use of GCC, please consult these web
13862    pages and the [6]GCC manuals. If that fails, the
13863    [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13864    web pages and the development of GCC are welcome on our developer
13865    list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives.
13866
13867   Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and
13868   distribution of this entire article is permitted in any medium,
13869   provided this notice is preserved.
13870
13871   These pages are [11]maintained by the GCC team. Last modified
13872   2013-12-03[12].
13873
13874References
13875
13876   1. http://gcc.gnu.org/egcs-1.0/features.html
13877   2. http://gcc.gnu.org/install/
13878   3. http://gcc.gnu.org/egcs-1.0/buildstat.html
13879   4. http://gcc.gnu.org/egcs-1.0/caveats.html
13880   5. http://gcc.gnu.org/mirrors.html
13881   6. http://gcc.gnu.org/onlinedocs/
13882   7. mailto:gcc-help@gcc.gnu.org
13883   8. mailto:gcc@gcc.gnu.org
13884   9. http://gcc.gnu.org/lists.html
13885  10. http://www.fsf.org/
13886  11. http://gcc.gnu.org/about.html
13887  12. http://validator.w3.org/check/referer
13888======================================================================
13889http://gcc.gnu.org/egcs-1.0/features.html
13890
13891                               EGCS 1.0 features
13892
13893     * Core compiler is based on the gcc2 development tree from Aug 2,
13894       1997, so we have most of the [1]features found in GCC 2.8.
13895     * Integrated GNU Fortran compiler based on g77-0.5.22-19970929.
13896     * Vast improvements in the C++ compiler; so many they have [2]page of
13897       their own!
13898     * Integrated C++ runtime libraries, including support for most major
13899       GNU/Linux systems!
13900     * New instruction scheduler from IBM Haifa which includes support for
13901       function wide instruction scheduling as well as superscalar
13902       scheduling.
13903     * Significantly improved alias analysis code.
13904     * Improved register allocation for two address machines.
13905     * Significant code generation improvements for Fortran code on
13906       Alphas.
13907     * Various optimizations from the g77 project as well as improved loop
13908       optimizations.
13909     * Dwarf2 debug format support for some targets.
13910     * egcs libstdc++ includes the SGI STL implementation without changes.
13911     * As a result of these and other changes, egcs libstc++ is not binary
13912       compatible with previous releases of libstdc++.
13913     * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO
13914       Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and
13915       1.1), Support for RTEMS on several embedded targets, Support for
13916       arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and
13917       MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc.
13918     * Integrated testsuites for gcc, g++, g77, libstdc++ and libio.
13919     * RS6000/PowerPC ports generate code which can run on all
13920       RS6000/PowerPC variants by default.
13921     * -mcpu= and -march= switches for the x86 port to allow better
13922       control over how the x86 port generates code.
13923     * Includes the template repository patch (aka repo patch); note the
13924       new template code makes repo obsolete for ELF systems using gnu-ld
13925       such as GNU/Linux.
13926     * Plus the usual assortment of bugfixes and improvements.
13927
13928
13929    For questions related to the use of GCC, please consult these web
13930    pages and the [3]GCC manuals. If that fails, the
13931    [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13932    web pages and the development of GCC are welcome on our developer
13933    list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives.
13934
13935   Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and
13936   distribution of this entire article is permitted in any medium,
13937   provided this notice is preserved.
13938
13939   These pages are [8]maintained by the GCC team. Last modified
13940   2012-11-02[9].
13941
13942References
13943
13944   1. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13945   2. http://gcc.gnu.org/egcs-1.0/c++features.html
13946   3. http://gcc.gnu.org/onlinedocs/
13947   4. mailto:gcc-help@gcc.gnu.org
13948   5. mailto:gcc@gcc.gnu.org
13949   6. http://gcc.gnu.org/lists.html
13950   7. http://www.fsf.org/
13951   8. http://gcc.gnu.org/about.html
13952   9. http://validator.w3.org/check/referer
13953======================================================================
13954http://gcc.gnu.org/egcs-1.0/caveats.html
13955
13956                                EGCS 1.0 Caveats
13957
13958     * EGCS has an integrated libstdc++, but does not have an integrated
13959       libg++. Furthermore old libg++ releases will not work with egc; HJ
13960       Lu has made a libg++-2.8.1.2 available which may work with EGCS.
13961       Note most C++ programs only need libstdc++.
13962     * Note that using -pedantic or -Wreturn-type can cause an explosion
13963       in the amount of memory needed for template-heavy C++ code, such as
13964       code that uses STL. Also note that -Wall includes -Wreturn-type, so
13965       if you use -Wall you will need to specify -Wno-return-type to turn
13966       it off.
13967     * Exception handling may not work with shared libraries, particularly
13968       on alphas, hppas, and mips based platforms. Exception handling is
13969       known to work on x86-linux platforms with shared libraries.
13970     * Some versions of the Linux kernel have bugs which prevent them from
13971       being compiled or from running when compiled by EGCS. See the FAQ
13972       (as shipped with EGCS 1.0) for additional information.
13973     * In general, EGCS is more rigorous about rejecting invalid C++ code
13974       or deprecated C++ constructs than G++ 2.7. As a result it may be
13975       necessary to fix C++ code before it will compile with EGCS.
13976     * G++ is also aggressively tracking the C++ standard; as a result
13977       code which was previously valid (and thus accepted by other
13978       compilers and older versions of G++) may no longer be accepted.
13979     * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS
13980       1.0.x and later releases should work with Red Hat Linux 5.0.
13981
13982
13983    For questions related to the use of GCC, please consult these web
13984    pages and the [1]GCC manuals. If that fails, the
13985    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13986    web pages and the development of GCC are welcome on our developer
13987    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
13988
13989   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
13990   distribution of this entire article is permitted in any medium,
13991   provided this notice is preserved.
13992
13993   These pages are [6]maintained by the GCC team. Last modified
13994   2012-11-02[7].
13995
13996References
13997
13998   1. http://gcc.gnu.org/onlinedocs/
13999   2. mailto:gcc-help@gcc.gnu.org
14000   3. mailto:gcc@gcc.gnu.org
14001   4. http://gcc.gnu.org/lists.html
14002   5. http://www.fsf.org/
14003   6. http://gcc.gnu.org/about.html
14004   7. http://validator.w3.org/check/referer
14005======================================================================
14006