1 This file contains information about GCC releases which has been generated
2 automatically 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
4 that led to GCC 3. For information on GCC 2.8.1 and older releases of GCC 2,
7 ======================================================================
8 http://gcc.gnu.org/gcc-4.8/index.html
13 The [1]GNU project and the GCC developers are pleased to announce the
16 This release is a bug-fix release, containing fixes for regressions in
17 GCC 4.8.1 relative to previous releases of GCC.
22 October 16, 2013 ([2]changes, [3]documentation)
25 May 31, 2013 ([4]changes, [5]documentation)
28 March 22, 2013 ([6]changes, [7]documentation)
30 References and Acknowledgements
32 GCC used to stand for the GNU C Compiler, but since the compiler
33 supports several other languages aside from C, it now stands for the
34 GNU Compiler Collection.
36 A list of [8]successful builds is updated as new information becomes
39 The GCC developers would like to thank the numerous people that have
40 contributed new features, improvements, bug fixes, and other changes as
41 well as test results to GCC. This [9]amazing group of volunteers is
42 what makes GCC successful.
44 For additional information about GCC please refer to the [10]GCC
45 project web site or contact the [11]GCC development mailing list.
47 To obtain GCC please use [12]our mirror sites or [13]our SVN server.
50 For questions related to the use of GCC, please consult these web
51 pages and the [14]GCC manuals. If that fails, the
52 [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these
53 web pages and the development of GCC are welcome on our developer
54 list at [16]gcc@gcc.gnu.org. All of [17]our lists have public
57 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and
58 distribution of this entire article is permitted in any medium,
59 provided this notice is preserved.
61 These pages are [19]maintained by the GCC team. Last modified
66 1. http://www.gnu.org/
67 2. http://gcc.gnu.org/gcc-4.8/changes.html
68 3. http://gcc.gnu.org/onlinedocs/4.8.2/
69 4. http://gcc.gnu.org/gcc-4.8/changes.html
70 5. http://gcc.gnu.org/onlinedocs/4.8.1/
71 6. http://gcc.gnu.org/gcc-4.8/changes.html
72 7. http://gcc.gnu.org/onlinedocs/4.8.0/
73 8. http://gcc.gnu.org/gcc-4.8/buildstat.html
74 9. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
75 10. http://gcc.gnu.org/index.html
76 11. mailto:gcc@gcc.gnu.org
77 12. http://gcc.gnu.org/mirrors.html
78 13. http://gcc.gnu.org/svn.html
79 14. http://gcc.gnu.org/onlinedocs/
80 15. mailto:gcc-help@gcc.gnu.org
81 16. mailto:gcc@gcc.gnu.org
82 17. http://gcc.gnu.org/lists.html
83 18. http://www.fsf.org/
84 19. http://gcc.gnu.org/about.html
85 20. http://validator.w3.org/check/referer
86 ======================================================================
87 http://gcc.gnu.org/gcc-4.8/changes.html
88 GCC 4.8 Release Series
89 Changes, New Features, and Fixes
93 GCC now uses C++ as its implementation language. This means that to
94 build GCC from sources, you will need a C++ compiler that understands
95 C++ 2003. For more details on the rationale and specific changes,
96 please refer to the [1]C++ conversion page.
98 To enable the Graphite framework for loop optimizations you now need
99 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from
100 the [2]GCC infrastructure directory. The installation manual contains
101 more information about requirements to build GCC.
103 GCC now uses a more aggressive analysis to derive an upper bound for
104 the number of iterations of loops using constraints imposed by language
105 standards. This may cause non-conforming programs to no longer work as
106 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new
107 option, -fno-aggressive-loop-optimizations, was added to disable this
108 aggressive analysis. In some loops that have known constant number of
109 iterations, but undefined behavior is known to occur in the loop before
110 reaching or during the last iteration, GCC will warn about the
111 undefined behavior in the loop instead of deriving lower upper bound of
112 the number of iterations for the loop. The warning can be disabled with
113 -Wno-aggressive-loop-optimizations.
115 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules
116 for the layout of vectors that could lead to wrong code being
117 generated. Vectors larger than 8 bytes in size are now by default
118 aligned to an 8-byte boundary. This is an ABI change: code that makes
119 explicit use of vector types may be incompatible with binary objects
120 built with older versions of GCC. Auto-vectorized code is not affected
123 On AVR, support has been removed for the command-line option
124 -mshort-calls deprecated in GCC 4.7.
126 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2
127 is turned on per default for all non-RTEMS configurations. This option
128 arranges for a better integration of [3]AVR Libc with avr-gcc. For
129 technical details, see [4]PR54461. To turn off the option in non-RTEMS
130 configurations, use --with-avrlibc=no. If the compiler is configured
131 for RTEMS, the option is always turned off.
133 More information on porting to GCC 4.8 from previous versions of GCC
134 can be found in the [5]porting guide for this release.
136 General Optimizer Improvements (and Changes)
138 * DWARF4 is now the default when generating DWARF debug information.
139 When -g is used on a platform that uses DWARF debugging
140 information, GCC will now default to -gdwarf-4
141 -fno-debug-types-section.
142 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
143 consumers support DWARF4 by default. Before GCC 4.8 the default
144 version used was DWARF2. To make GCC 4.8 generate an older DWARF
145 version use -g together with -gdwarf-2 or -gdwarf-3. The default
146 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf.
147 * A new general optimization level, -Og, has been introduced. It
148 addresses the need for fast compilation and a superior debugging
149 experience while providing a reasonable level of runtime
150 performance. Overall experience for development should be better
151 than the default optimization level -O0.
152 * A new option -ftree-partial-pre was added to control the partial
153 redundancy elimination (PRE) optimization. This option is enabled
154 by default at the -O3 optimization level, and it makes PRE more
156 * The option -fconserve-space has been removed; it was no longer
157 useful on most targets since GCC supports putting variables into
158 BSS without making them common.
159 * The struct reorg and matrix reorg optimizations (command-line
160 options -fipa-struct-reorg and -fipa-matrix-reorg) have been
161 removed. They did not always work correctly, nor did they work with
162 link-time optimization (LTO), hence were only applicable to
163 programs consisting of a single translation unit.
164 * Several scalability bottle-necks have been removed from GCC's
165 optimization passes. Compilation of extremely large functions, e.g.
166 due to the use of the flatten attribute in the "Eigen" C++ linear
167 algebra templates library, is significantly faster than previous
169 * Link-time optimization (LTO) improvements:
170 + LTO partitioning has been rewritten for better reliability and
171 maintanibility. Several important bugs leading to link
172 failures have been fixed.
173 * Interprocedural optimization improvements:
174 + A new symbol table has been implemented. It builds on existing
175 callgraph and varpool modules and provide a new API. Unusual
176 symbol visibilities and aliases are handled more consistently
177 leading to, for example, more aggressive unreachable code
179 + The inline heuristic can now bypass limits on the size of of
180 inlined functions when the inlining is particularly
181 profitable. This happens, for example, when loop bounds or
182 array strides get propagated.
183 + Values passed through aggregates (either by value or
184 reference) are now propagated at the inter-procedural level
185 leading to better inlining decisions (for example in the case
186 of Fortran array descriptors) and devirtualization.
187 * [6]AddressSanitizer , a fast memory error detector, has been added
188 and can be enabled via -fsanitize=address. Memory access
189 instructions will be instrumented to detect heap-, stack-, and
190 global-buffer overflow as well as use-after-free bugs. To get nicer
191 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is
192 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on
194 * [7]ThreadSanitizer has been added and can be enabled via
195 -fsanitize=thread. Instructions will be instrumented to detect data
196 races. The ThreadSanitizer is available on x86-64 GNU/Linux.
197 * A new local register allocator (LRA) has been implemented, which
198 replaces the 26 year old reload pass and improves generated code
199 quality. For now it is active on the ia32 and x86-64 targets.
200 * Support for transactional memory has been implemented on the
201 following architectures: IA-32/x86-64, ARM, PowerPC, SH, Sparc, and
204 New Languages and Language specific improvements
208 * Each diagnostic emitted now includes the original source line and a
209 caret '^' indicating the column. The option
210 -fno-diagnostics-show-caret suppresses this information.
211 * The option -ftrack-macro-expansion=2 is now enabled by default.
212 This allows the compiler to display the macro expansion stack in
213 diagnostics. Combined with the caret information, an example
214 diagnostic showing these two features is:
216 t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float
218 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _
219 _b = (B); __a < __b ? __b : __a; })
222 t.c:7:7: note: in expansion of macro 'MYMAX'
226 * A new -Wsizeof-pointer-memaccess warning has been added (also
227 enabled by -Wall) to warn about suspicious length parameters to
228 certain string and memory built-in functions if the argument uses
229 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof
230 (ptr)); if ptr is not an array, but a pointer, and suggests a
231 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));.
232 * The new option -Wpedantic is an alias for -pedantic, which is now
233 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and
234 -Wno-error=pedantic work in the same way as for any other -W
235 option. One caveat is that -Werror=pedantic is not equivalent to
236 -pedantic-errors, since the latter makes into errors some warnings
237 that are not controlled by -Wpedantic, and the former only affects
238 diagnostics that are disabled when using -Wno-pedantic.
239 * The option -Wshadow no longer warns if a declaration shadows a
240 function declaration, unless the former declares a function or
241 pointer to function, because this is [8]a common and valid case in
246 * G++ now implements the [9]C++11 thread_local keyword; this differs
247 from the GNU __thread keyword primarily in that it allows dynamic
248 initialization and destruction semantics. Unfortunately, this
249 support requires a run-time penalty for references to
250 non-function-local thread_local variables defined in a different
251 translation unit even if they don't need dynamic initialization, so
252 users may want to continue to use __thread for TLS variables with
253 static initialization semantics.
254 If the programmer can be sure that no use of the variable in a
255 non-defining TU needs to trigger dynamic initialization (either
256 because the variable is statically initialized, or a use of the
257 variable in the defining TU will be executed before any uses in
258 another TU), they can avoid this overhead with the
259 -fno-extern-tls-init option.
260 OpenMP threadprivate variables now also support dynamic
261 initialization and destruction by the same mechanism.
262 * G++ now implements the [10]C++11 attribute syntax, e.g.
264 [[noreturn]] void f();
266 and also the alignment specifier, e.g.
268 alignas(double) int i;
270 * G++ now implements [11]C++11 inheriting constructors, e.g.
272 struct A { A(int); };
273 struct B: A { using A::A; }; // defines B::B(int)
276 * As of GCC 4.8.1, G++ implements the change to decltype semantics
280 decltype(f()) g(); // OK, return type of f() is not required to be complete.
282 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g.
284 struct A { int f() &; };
285 int i = A().f(); // error, f() requires an lvalue object
287 * G++ now supports a -std=c++1y option for experimentation with
288 features proposed for the next revision of the standard, expected
289 around 2014. Currently the only difference from -std=c++11 is
290 support for return type deduction in normal functions, as proposed
291 in [14]N3386. Status of C++1y features in GCC 4.8 can be found
293 * The G++ namespace association extension, __attribute ((strong)),
294 has been deprecated. Inline namespaces should be used instead.
295 * G++ now supports a -fext-numeric-literal option to control whether
296 GNU numeric literal suffixes are accepted as extensions or
297 processed as C++11 user-defined numeric literal suffixes. The flag
298 is on (use suffixes for GNU literals) by default for -std=gnu++*,
299 and -std=c++98. The flag is off (use suffixes for user-defined
300 literals) by default for -std=c++11 and later.
302 Runtime Library (libstdc++)
304 * [16]Improved experimental support for the new ISO C++ standard,
306 + forward_list meets the allocator-aware container requirements;
307 + this_thread::sleep_for(), this_thread::sleep_until() and
308 this_thread::yield() are defined without requiring the
309 configure option --enable-libstdcxx-time;
310 * Improvements to <random>:
311 + SSE optimized normal_distribution.
312 + Use of hardware RNG instruction for random_device on new x86
313 processors (requires the assembler to support the
316 + New random number engine simd_fast_mersenne_twister_engine
317 with an optimized SSE implementation.
318 + New random number distributions beta_distribution,
319 normal_mv_distribution, rice_distribution,
320 nakagami_distribution, pareto_distribution, k_distribution,
321 arcsine_distribution, hoyt_distribution.
322 * Added --disable-libstdcxx-verbose configure option to disable
323 diagnostic messages issued when a process terminates abnormally.
324 This may be useful for embedded systems to reduce the size of
325 executables that link statically to the library.
329 * Compatibility notice:
330 + Module files: The version of module files (.mod) has been
331 incremented. Fortran MODULEs compiled by earlier GCC versions
332 have to be recompiled, when they are USEd by files compiled
333 with GCC 4.8. GCC 4.8 is not able to read .mod files created
334 by earlier versions; attempting to do so gives an error
336 Note: The ABI of the produced assembler data itself has not
337 changed; object files and libraries are fully compatible with
338 older versions except as noted below.
339 + ABI: Some internal names (used in the assembler/object file)
340 have changed for symbols declared in the specification part of
341 a module. If an affected module - or a file using it via use
342 association - is recompiled, the module and all files which
343 directly use such symbols have to be recompiled as well. This
344 change only affects the following kind of module symbols:
345 o Procedure pointers. Note: C-interoperable function
346 pointers (type(c_funptr)) are not affected nor are
347 procedure-pointer components.
348 o Deferred-length character strings.
349 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a
350 backtrace at an arbitrary place in user code; program execution
351 continues normally afterwards.
352 * The [18]-Wc-binding-type warning option has been added (disabled by
353 default). It warns if the a variable might not be C interoperable;
354 in particular, if the variable has been declared using an intrinsic
355 type with default kind instead of using a kind parameter defined
356 for C interoperability in the intrinsic ISO_C_Binding module.
357 Before, this warning was always printed. The -Wc-binding-type
358 option is enabled by -Wall.
359 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line
360 options have been added, which diagnose when code to is inserted
361 for automatic (re)allocation of a variable during assignment. This
362 option can be used to decide whether it is safe to use
363 [20]-fno-realloc-lhs. Additionally, it can be used to find
364 automatic (re)allocation in hot loops. (For arrays, replacing
365 "var=" by "var(:)=" disables the automatic reallocation.)
366 * The [21]-Wcompare-reals command-line option has been added. When
367 this is set, warnings are issued when comparing REAL or COMPLEX
368 types for equality and inequality; consider replacing a == b by
369 abs(a-b) < eps with a suitable eps. -Wcompare-reals is enabled by
371 * The [22]-Wtarget-lifetime command-line option has been added
372 (enabled with -Wall), which warns if the pointer in a pointer
373 assignment might outlive its target.
374 * Reading floating point numbers which use "q" for the exponential
375 (such as 4.0q0) is now supported as vendor extension for better
376 compatibility with old data files. It is strongly recommended to
377 use for I/O the equivalent but standard conforming "e" (such as
379 (For Fortran source code, consider replacing the "q" in
380 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a
381 suitable qp). Note that - in Fortran source code - replacing "q" by
382 a simple "e" is not equivalent.)
383 * The GFORTRAN_TMPDIR environment variable for specifying a
384 non-default directory for files opened with STATUS="SCRATCH", is
385 not used anymore. Instead gfortran checks the POSIX/GNU standard
386 TMPDIR environment variable. If TMPDIR is not defined, gfortran
387 falls back to other methods to determine the directory for
388 temporary files as documented in the [23]user manual.
390 + Support for unlimited polymorphic variables (CLASS(*)) has
391 been added. Nonconstant character lengths are not yet
394 + Assumed types (TYPE(*)) are now supported.
395 + Experimental support for assumed-rank arrays (dimension(..))
396 has been added. Note that currently gfortran's own array
397 descriptor is used, which is different from the one defined in
398 TS29113, see [26]gfortran's header file or use the [27]Chasm
399 Language Interoperability Tools.
403 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1
404 release. The library support is not quite complete.
405 * Go has been tested on GNU/Linux and Solaris platforms for various
406 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may
407 work on other platforms as well.
409 New Targets and Target Specific Improvements
413 * A new port has been added to support AArch64, the new 64-bit
414 architecture from ARM. Note that this is a separate port from the
415 existing 32-bit ARM port.
416 * The port provides initial support for the Cortex-A53 and the
417 Cortex-A57 processors with the command line options
418 -mcpu=cortex-a53 and -mcpu=cortex-a57.
422 * Initial support has been added for the AArch32 extensions defined
423 in the ARMv8 architecture.
424 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs.
425 * A new option, -mcpu=marvell-pj4, has been added to generate code
426 for the Marvell PJ4 processor.
427 * The compiler can now automatically generate the VFMA, VFMS, REVSH
428 and REV16 instructions.
429 * A new vectorizer cost model for Advanced SIMD configurations to
430 improve the auto-vectorization strategies used.
431 * The scheduler now takes into account the number of live registers
432 to reduce the amount of spilling that can occur. This should
433 improve code performance in large functions. The limit can be
434 removed by using the option -fno-sched-pressure.
435 * Improvements have been made to the Marvell iWMMX code generation
436 and support for the iWMMX2 SIMD unit has been added. The option
437 -mcpu=iwmmxt2 can be used to enable code generation for the latter.
438 * A number of code generation improvements for Thumb2 to reduce code
439 size when compiling for the M-profile processors.
440 * The RTEMS (arm-rtems) port has been updated to use the EABI.
441 * Code generation support for the old FPA and Maverick floating-point
442 architectures has been removed. Ports that previously relied on
443 these features have also been removed. This includes the targets:
444 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
445 + arm*-*-elf (use arm*-*-eabi)
446 + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
447 + arm*-*-ecos-elf (no alternative)
448 + arm*-*-freebsd (no alternative)
449 + arm*-wince-pe* (no alternative).
453 * Support for the "Embedded C" fixed-point has been added. For
454 details, see the [28]GCC wiki and the [29]user manual. The support
456 * A new print modifier %r for register operands in inline assembler
457 is supported. It will print the raw register number without the
459 /* Return the most significant byte of 'val', a 64-bit value. */
461 unsigned char msb (long long val)
464 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val));
467 The inline assembler in this example will generate code like
469 provided c is allocated to R24 and val is allocated to R8...R15.
470 This works because the GNU assembler accepts plain register numbers
471 without register prefix.
472 * Static initializers with 3-byte symbols are supported now:
473 extern const __memx char foo;
474 const __memx void *pfoo = &foo;
475 This requires at least Binutils 2.23.
479 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with
480 SSE extensions disabled. Since the x86-64 ABI requires 16 byte
481 stack alignment, this is ABI incompatible and intended to be used
482 in controlled environments where stack space is an important
483 limitation. This option will lead to wrong code when functions
484 compiled with 16 byte stack alignment (such as functions from a
485 standard library) are called with misaligned stack. In this case,
486 SSE instructions may lead to misaligned memory access traps. In
487 addition, variable arguments will be handled incorrectly for 16
488 byte aligned objects (including x87 long double and __int128),
489 leading to wrong results. You must build all modules with
490 -mpreferred-stack-boundary=3, including any libraries. This
491 includes the system libraries and startup modules.
492 * Support for the new Intel processor codename Broadwell with RDSEED,
493 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw,
494 -mrdseed command-line options.
495 * Support for the Intel RTM and HLE intrinsics, built-in functions
496 and code generation is available via -mrtm and -mhle.
497 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets.
498 Intrinsics and built-in functions are available via -mfxsr, -mxsave
499 and -mxsaveopt respectively.
500 * New -maddress-mode=[short|long] options for x32.
501 -maddress-mode=short overrides default 64-bit addresses to 32-bit
502 by emitting the 0x67 address-size override prefix. This is the
503 default address mode for x32.
504 * New built-in functions to detect run-time CPU type and ISA:
505 + A built-in function __builtin_cpu_is has been added to detect
506 if the run-time CPU is of a particular type. It returns a
507 positive integer on a match and zero otherwise. It accepts one
508 string literal argument, the CPU name. For example,
509 __builtin_cpu_is("westmere") returns a positive integer if the
510 run-time CPU is an Intel Core i7 Westmere processor. Please
511 refer to the [30]user manual for the list of valid CPU names
513 + A built-in function __builtin_cpu_supports has been added to
514 detect if the run-time CPU supports a particular ISA feature.
515 It returns a positive integer on a match and zero otherwise.
516 It accepts one string literal argument, the ISA feature. For
517 example, __builtin_cpu_supports("ssse3") returns a positive
518 integer if the run-time CPU supports SSSE3 instructions.
519 Please refer to the [31]user manual for the list of valid ISA
521 Caveat: If these built-in functions are called before any static
522 constructors are invoked, like during IFUNC initialization, then
523 the CPU detection initialization must be explicitly run using this
524 newly provided built-in function, __builtin_cpu_init. The
525 initialization needs to be done only once. For example, this is how
526 the invocation would look like inside an IFUNC initializer:
527 static void (*some_ifunc_resolver(void))(void)
529 __builtin_cpu_init();
530 if (__builtin_cpu_is("amdfam10h") ...
531 if (__builtin_cpu_supports("popcnt") ...
534 * Function Multiversioning Support with G++:
535 It is now possible to create multiple function versions each
536 targeting a specific processor and/or ISA. Function versions have
537 the same signature but different target attributes. For example,
538 here is a program with function versions:
539 __attribute__ ((target ("default")))
545 __attribute__ ((target ("sse4.2")))
554 assert ((*p)() == foo());
558 Please refer to this [32]wiki for more information.
559 * The x86 backend has been improved to allow option -fschedule-insns
560 to work reliably. This option can be used to schedule instructions
561 better and leads to improved performace in certain cases.
562 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437
563 from the Mingw-w64 trunk.
564 * Support for new AMD family 15h processors (Steamroller core) is now
565 available through the -march=bdver3 and -mtune=bdver3 options.
566 * Support for new AMD family 16h processors (Jaguar core) is now
567 available through the -march=btver2 and -mtune=btver2 options.
571 * This target now supports the -fstack-usage command-line option.
575 * GCC can now generate code specifically for the R4700, Broadcom XLP
576 and MIPS 34kn processors. The associated -march options are
577 -march=r4700, -march=xlp and -march=34kn respectively.
578 * GCC now generates better DSP code for MIPS 74k cores thanks to
579 further scheduling optimizations.
580 * The MIPS port now supports the -fstack-check option.
581 * GCC now passes the -mmcu and -mno-mcu options to the assembler.
582 * Previous versions of GCC would silently accept -fpic and -fPIC for
583 -mno-abicalls targets like mips*-elf. This combination was not
584 intended or supported, and did not generate position-independent
585 code. GCC 4.8 now reports an error when this combination is used.
587 PowerPC / PowerPC64 / RS6000
589 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save,
590 restore or update the VRSAVE register by default. The respective
591 operating systems manage the VRSAVE register directly.
592 * Large TOC support has been added for AIX through the command line
593 option -mcmodel=large.
594 * Native Thread-Local Storage support has been added for AIX.
595 * VMX (Altivec) and VSX instruction sets now are enabled implicitly
596 when targetting processors that support those hardware features on
601 * This target will now issue a warning message whenever multiple fast
602 interrupt handlers are found in the same compilation unit. This
603 feature can be turned off by the new
604 -mno-warn-multiple-fast-interrupts command-line option.
608 * Support for the IBM zEnterprise zEC12 processor has been added.
609 When using the -march=zEC12 option, the compiler will generate code
610 making use of the following new instructions:
611 + load and trap instructions
612 + 2 new compare and trap instructions
613 + rotate and insert selected bits - without CC clobber
614 The -mtune=zEC12 option enables zEC12 specific instruction
615 scheduling without making use of new instructions.
616 * Register pressure sensitive instruction scheduling is enabled by
618 * The ifunc function attribute is enabled by default.
619 * memcpy and memcmp invokations on big memory chunks or with run time
620 lengths are not generated inline anymore when tuning for z10 or
621 higher. The purpose is to make use of the IFUNC optimized versions
626 * The default alignment settings have been reduced to be less
627 aggressive. This results in more compact code for optimization
628 levels other than -Os.
629 * Improved support for the __atomic built-in functions:
630 + A new option -matomic-model=model selects the model for the
631 generated atomic sequences. The following models are
635 Software gUSA sequences (SH3* and SH4* only). On
636 SH4A targets this will now also partially utilize
637 the movco.l and movli.l instructions. This is the
638 default when the target is sh3*-*-linux* or
642 Hardware movco.l / movli.l sequences (SH4A only).
645 Software thread control block sequences.
648 Software interrupt flipping sequences (privileged
649 mode only). This is the default when the target is
650 sh1*-*-linux* or sh2*-*-linux*.
653 Generates function calls to the respective __atomic
654 built-in functions. This is the default for SH64
655 targets or when the target is not sh*-*-linux*.
657 + The option -msoft-atomic has been deprecated. It is now an
658 alias for -matomic-model=soft-gusa.
659 + A new option -mtas makes the compiler generate the tas.b
660 instruction for the __atomic_test_and_set built-in function
661 regardless of the selected atomic model.
662 + The __sync functions in libgcc now reflect the selected atomic
663 model when building the toolchain.
664 * Added support for the mov.b and mov.w instructions with
665 displacement addressing.
666 * Added support for the SH2A instructions movu.b and movu.w.
667 * Various improvements to code generated for integer arithmetic.
668 * Improvements to conditional branches and code that involves the T
669 bit. A new option -mzdcbranch tells the compiler to favor
670 zero-displacement branches. This is enabled by default for SH4*
672 * The pref instruction will now be emitted by the __builtin_prefetch
673 built-in function for SH3* targets.
674 * The fmac instruction will now be emitted by the fmaf standard
675 function and the __builtin_fmaf built-in function.
676 * The -mfused-madd option has been deprecated in favor of the
677 machine-independent -ffp-contract option. Notice that the fmac
678 instruction will now be generated by default for expressions like a
679 * b + c. This is due to the compiler default setting
681 * Added new options -mfsrra and -mfsca to allow the compiler using
682 the fsrra and fsca instructions on targets other than SH4A (where
683 they are already enabled by default).
684 * Added support for the __builtin_bswap32 built-in function. It is
685 now expanded as a sequence of swap.b and swap.w instructions
686 instead of a library function call.
687 * The behavior of the -mieee option has been fixed and the negative
688 form -mno-ieee has been added to control the IEEE conformance of
689 floating point comparisons. By default -mieee is now enabled and
690 the option -ffinite-math-only implicitly sets -mno-ieee.
691 * Added support for the built-in functions __builtin_thread_pointer
692 and __builtin_set_thread_pointer. This assumes that GBR is used to
693 hold the thread pointer of the current thread. Memory loads and
694 stores relative to the address returned by __builtin_thread_pointer
695 will now also utilize GBR based displacement address modes.
696 * The -mdiv= option for targets other than SHmedia has been fixed and
701 * Added optimized instruction scheduling for Niagara4.
705 * Added support for the -mcmodel=MODEL command-line option. The
706 models supported are small and large.
710 * This target now supports the E3V5 architecture via the use of the
711 new -mv850e3v5 command-line option. It also has experimental
712 support for the e3v5 LOOP instruction which can be enabled via the
713 new -mloop command-line option.
717 * This target now supports the -fstack-usage command-line option.
723 * Executables are now linked against shared libgcc by default. The
724 previous default was to link statically, which can still be done by
725 explicitly specifying -static or static-libgcc on the command line.
726 However it is strongly advised against, as it will cause problems
727 for any application that makes use of DLLs compiled by GCC. It
728 should be alright for a monolithic stand-alone application that
729 only links against the Windows DLLs, but offers little or no
734 This is the [33]list of problem reports (PRs) from GCC's bug tracking
735 system that are known to be fixed in the 4.8.1 release. This list might
736 not be complete (that is, it is possible that some PRs that have been
737 fixed are not listed here).
739 The C++11 <chrono> std::chrono::system_clock and
740 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they
741 both are now separate (never typedefs of each other), both use
742 std::chrono::nanoseconds resolution, on most GNU/Linux configurations
743 std::chrono::steady_clock is now finally monotonic, and both classes
744 are mangled differently than in the previous GCC releases.
745 std::chrono::system_clock::now() with std::chrono::microseconds resp.
746 std::chrono::seconds resolution is still exported for backwards
747 compatibility with default configured libstdc++. Note that libstdc++
748 configured with --enable-libstdcxx-time= used to be ABI incompatible
749 with default configured libstdc++ for those two classes and no ABI
750 compatibility can be offered for those configurations, so any C++11
751 code that uses those classes and has been compiled and linked against
752 libstdc++ configured with the non-default --enable-libstdcxx-time=
753 configuration option needs to be recompiled.
757 This is the [34]list of problem reports (PRs) from GCC's bug tracking
758 system that are known to be fixed in the 4.8.2 release. This list might
759 not be complete (that is, it is possible that some PRs that have been
760 fixed are not listed here).
763 For questions related to the use of GCC, please consult these web
764 pages and the [35]GCC manuals. If that fails, the
765 [36]gcc-help@gcc.gnu.org mailing list might help. Comments on these
766 web pages and the development of GCC are welcome on our developer
767 list at [37]gcc@gcc.gnu.org. All of [38]our lists have public
770 Copyright (C) [39]Free Software Foundation, Inc. Verbatim copying and
771 distribution of this entire article is permitted in any medium,
772 provided this notice is preserved.
774 These pages are [40]maintained by the GCC team. Last modified
779 1. http://gcc.gnu.org/wiki/cxx-conversion
780 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/
781 3. http://www.nongnu.org/avr-libc/
782 4. http://gcc.gnu.org/PR54461
783 5. http://gcc.gnu.org/gcc-4.8/porting_to.html
784 6. https://code.google.com/p/address-sanitizer/
785 7. https://code.google.com/p/data-race-test/wiki/ThreadSanitizer
786 8. https://lkml.org/lkml/2006/11/28/239
787 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
788 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
789 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
790 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf
791 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
792 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html
793 15. http://gcc.gnu.org/projects/cxx1y.html
794 16. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2011
795 17. http://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html
796 18. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
797 19. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
798 20. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
799 21. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
800 22. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
801 23. http://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html
802 24. http://gcc.gnu.org/wiki/Fortran2003Status
803 25. http://gcc.gnu.org/wiki/TS29113Status
804 26. http://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text%2Fplain&view=co
805 27. http://chasm-interop.sourceforge.net/
806 28. http://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support
807 29. http://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html
808 30. http://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
809 31. http://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
810 32. http://gcc.gnu.org/wiki/FunctionMultiVersioning
811 33. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1
812 34. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2
813 35. http://gcc.gnu.org/onlinedocs/
814 36. mailto:gcc-help@gcc.gnu.org
815 37. mailto:gcc@gcc.gnu.org
816 38. http://gcc.gnu.org/lists.html
817 39. http://www.fsf.org/
818 40. http://gcc.gnu.org/about.html
819 41. http://validator.w3.org/check/referer
820 ======================================================================
821 http://gcc.gnu.org/gcc-4.7/index.html
822 GCC 4.7 Release Series
826 The [1]GNU project and the GCC developers are pleased to announce the
827 release of GCC 4.7.3.
829 This release is a bug-fix release, containing fixes for regressions in
830 GCC 4.7.2 relative to previous releases of GCC.
835 April 11, 2013 ([2]changes, [3]documentation)
838 September 20, 2012 ([4]changes, [5]documentation)
841 June 14, 2012 ([6]changes, [7]documentation)
844 March 22, 2012 ([8]changes, [9]documentation)
846 References and Acknowledgements
848 GCC used to stand for the GNU C Compiler, but since the compiler
849 supports several other languages aside from C, it now stands for the
850 GNU Compiler Collection.
852 A list of [10]successful builds is updated as new information becomes
855 The GCC developers would like to thank the numerous people that have
856 contributed new features, improvements, bug fixes, and other changes as
857 well as test results to GCC. This [11]amazing group of volunteers is
858 what makes GCC successful.
860 For additional information about GCC please refer to the [12]GCC
861 project web site or contact the [13]GCC development mailing list.
863 To obtain GCC please use [14]our mirror sites or [15]our SVN server.
866 For questions related to the use of GCC, please consult these web
867 pages and the [16]GCC manuals. If that fails, the
868 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
869 web pages and the development of GCC are welcome on our developer
870 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
873 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
874 distribution of this entire article is permitted in any medium,
875 provided this notice is preserved.
877 These pages are [21]maintained by the GCC team. Last modified
882 1. http://www.gnu.org/
883 2. http://gcc.gnu.org/gcc-4.7/changes.html
884 3. http://gcc.gnu.org/onlinedocs/4.7.3/
885 4. http://gcc.gnu.org/gcc-4.7/changes.html
886 5. http://gcc.gnu.org/onlinedocs/4.7.2/
887 6. http://gcc.gnu.org/gcc-4.7/changes.html
888 7. http://gcc.gnu.org/onlinedocs/4.7.1/
889 8. http://gcc.gnu.org/gcc-4.7/changes.html
890 9. http://gcc.gnu.org/onlinedocs/4.7.0/
891 10. http://gcc.gnu.org/gcc-4.7/buildstat.html
892 11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
893 12. http://gcc.gnu.org/index.html
894 13. mailto:gcc@gcc.gnu.org
895 14. http://gcc.gnu.org/mirrors.html
896 15. http://gcc.gnu.org/svn.html
897 16. http://gcc.gnu.org/onlinedocs/
898 17. mailto:gcc-help@gcc.gnu.org
899 18. mailto:gcc@gcc.gnu.org
900 19. http://gcc.gnu.org/lists.html
901 20. http://www.fsf.org/
902 21. http://gcc.gnu.org/about.html
903 22. http://validator.w3.org/check/referer
904 ======================================================================
905 http://gcc.gnu.org/gcc-4.7/changes.html
906 GCC 4.7 Release Series
907 Changes, New Features, and Fixes
911 * The -fconserve-space flag has been deprecated. The flag had no
912 effect for most targets: only targets without a global .bss section
913 and without support for switchable sections. Furthermore, the flag
914 only had an effect for G++, where it could result in wrong
915 semantics (please refer to the GCC manual for further details). The
916 flag will be removed in GCC 4.8
917 * Support for a number of older systems and recently unmaintained or
918 untested target ports of GCC has been declared obsolete in GCC 4.7.
919 Unless there is activity to revive them, the next release of GCC
920 will have their sources permanently removed.
921 All GCC ports for the following processor architectures have been
923 + picoChip (picochip-*)
924 The following ports for individual systems on particular
925 architectures have been obsoleted:
926 + IRIX 6.5 (mips-sgi-irix6.5)
927 + MIPS OpenBSD (mips*-*-openbsd*)
928 + Solaris 8 (*-*-solaris2.8). Details can be found in the
930 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*)
931 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A,
932 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by
933 default, which for some sources generates code that accesses memory
934 on unaligned addresses. This requires the kernel of those systems
935 to enable such accesses (controlled by CP15 register c1, refer to
936 ARM documentation). Alternatively, or for compatibility with
937 kernels where unaligned accesses are not supported, all code has to
938 be compiled with -mno-unaligned-access. Upstream Linux kernel
939 releases have automatically and unconditionally supported unaligned
940 accesses as emitted by GCC due to this option being active since
942 * Support on ARM for the legacy floating-point accelerator (FPA) and
943 the mixed-endian floating-point format that it used has been
944 obsoleted. The ports that still use this format have been obsoleted
945 as well. Many legacy ARM ports already provide an alternative that
946 uses the VFP floating-point format. The obsolete ports will be
947 deleted in the next release.
948 The obsolete ports with alternatives are:
949 + arm*-*-rtems (use arm*-*-rtemseabi)
950 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
951 + arm*-*-elf (use arm*-*-eabi)
952 + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
953 Note, however, that these alternatives are not binary compatible
954 with their legacy counterparts (although some can support running
955 legacy applications).
956 The obsolete ports that currently lack a modern alternative are:
960 New ports that support more recent versions of the architecture are
962 * Support for the Maverick co-processor on ARM has been obsoleted.
963 Code to support it will be deleted in the next release.
964 * Support has been removed for Unix International threads on Solaris
965 2, so the --enable-threads=solaris configure option and the
966 -threads compiler option don't work any longer.
967 * Support has been removed for the Solaris BSD Compatibility Package,
968 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed
969 from Solaris 11, and was only intended as a migration aid from
970 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not
971 recognized any longer.
972 * The AVR port's libgcc has been improved and its multilib structure
973 has been enhanced. As a result, all objects contributing to an
974 application must either be compiled with GCC versions up to 4.6.x
975 or with GCC versions 4.7.1 or later. If the compiler is used with
976 AVR Libc, you need a version that supports the new layout, i.e.
977 implements [2]#35407.
978 * The AVR port's -mshort-calls command-line option has been
979 deprecated. It will be removed in the GCC 4.8 release. See -mrelax
981 * The AVR port only references startup code that clears .bss and the
982 common section resp. initializes the .data and .rodata section
983 provided respective sections (or subsections thereof) are not
984 empty, see [3]PR18145. Applications that put all static storage
985 objects into non-standard sections and / or define all static
986 storage objects in assembler modules, must reference __do_clear_bss
987 resp. __do_copy_data by hand or undefine the symbol(s) by means of
988 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data.
989 * The ARM port's -mwords-little-endian option has been deprecated. It
990 will be removed in a future release.
991 * Support has been removed for the NetWare x86 configuration
992 obsoleted in GCC 4.6.
993 * It is no longer possible to use the "l" constraint in MIPS16 asm
995 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard
996 library which affected the ABI in C++11 mode: a data member was
997 added to std::list changing its size and altering the definitions
998 of some member functions, and std::pair's move constructor was
999 non-trivial which altered the calling convention for functions with
1000 std::pair arguments or return types. The ABI incompatibilities have
1001 been fixed for GCC version 4.7.2 but as a result C++11 code
1002 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11
1003 code compiled with different GCC versions and with C++98/C++03 code
1004 compiled with any version.
1005 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
1006 rules for the layout of vectors that could lead to wrong code being
1007 generated. Vectors larger than 8 bytes in size are now by default
1008 aligned to an 8-byte boundary. This is an ABI change: code that
1009 makes explicit use of vector types may be incompatible with binary
1010 objects built with older versions of GCC. Auto-vectorized code is
1011 not affected by this change. (This change affects GCC versions
1013 * More information on porting to GCC 4.7 from previous versions of
1014 GCC can be found in the [4]porting guide for this release.
1016 General Optimizer Improvements
1018 * Support for a new parameter --param case-values-threshold=n was
1019 added to allow users to control the cutoff between doing switch
1020 statements as a series of if statements and using a jump table.
1021 * Link-time optimization (LTO) improvements:
1022 + Improved scalability and reduced memory usage. Link time
1023 optimization of Firefox now requires 3GB of RAM on a 64-bit
1024 system, while over 8GB was needed previously. Linking time has
1025 been improved, too. The serial stage of linking Firefox has
1026 been sped up by about a factor of 10.
1027 + Reduced size of object files and temporary storage used during
1029 + Streaming performance (both outbound and inbound) has been
1031 + ld -r is now supported with LTO.
1032 + Several bug fixes, especially in symbol table handling and
1034 * Interprocedural optimization improvements:
1035 + Heuristics now take into account that after inlining code will
1036 be optimized out because of known values (or properties) of
1037 function parameters. For example:
1048 The call of foo will be inlined into bar even when optimizing
1049 for code size. Constructs based on __builtin_constant_p are
1050 now understood by the inliner and code size estimates are
1051 evaluated a lot more realistically.
1052 + The representation of C++ virtual thunks and aliases (both
1053 implicit and defined via the alias attribute) has been
1054 re-engineered. Aliases no longer pose optimization barriers
1055 and calls to an alias can be inlined and otherwise optimized.
1056 + The inter-procedural constant propagation pass has been
1057 rewritten. It now performs generic function specialization.
1058 For example when compiling the following:
1062 ... do something ...
1064 ... do something else ...
1076 GCC will now produce two copies of foo. One with flag being
1077 true, while other with flag being false. This leads to
1078 performance improvements previously possible only by inlining
1079 all calls. Cloning causes a lot less code size growth.
1080 * A string length optimization pass has been added. It attempts to
1081 track string lengths and optimize various standard C string
1082 functions like strlen, strchr, strcpy, strcat, stpcpy and their
1083 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is
1084 enabled by default at -O2 or above, unless optimizing for size, and
1085 can be disabled by the -fno-optimize-strlen option. The pass can
1087 char *bar (const char *a)
1089 size_t l = strlen (a) + 2;
1090 char *p = malloc (l); if (p == NULL) return p;
1091 strcpy (p, a); strcat (p, "/"); return p;
1095 char *bar (const char *a)
1097 size_t tmp = strlen (a);
1098 char *p = malloc (tmp + 2); if (p == NULL) return p;
1099 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p;
1102 or for hosted compilations where stpcpy is available in the runtime
1103 and headers provide its prototype, e.g.
1104 void foo (char *a, const char *b, const char *c, const char *d)
1106 strcpy (a, b); strcat (a, c); strcat (a, d);
1109 can be optimized into:
1110 void foo (char *a, const char *b, const char *c, const char *d)
1112 strcpy (stpcpy (stpcpy (a, b), c), d);
1115 New Languages and Language specific improvements
1117 * Version 3.1 of the [5]OpenMP specification is now supported for the
1118 C, C++, and Fortran compilers.
1122 * The command-line option -feliminate-unused-debug-types has been
1123 re-enabled by default, as it is for the other languages, leading to
1124 a reduction in debug info size of 12.5% and more for relevant
1125 cases, as well as to a small compilation speedup.
1129 * A new built-in, __builtin_assume_aligned, has been added, through
1130 which the compiler can be hinted about pointer alignment and can
1131 use it to improve generated code.
1132 * A new warning option -Wunused-local-typedefs was added for C, C++,
1133 Objective-C and Objective-C++. This warning diagnoses typedefs
1134 locally defined in a function, and otherwise not used.
1135 * A new experimental command-line option -ftrack-macro-expansion was
1136 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows
1137 the compiler to emit diagnostic about the current macro expansion
1138 stack when a compilation error occurs in a macro expansion.
1139 * Experimental support for transactional memory has been added. It
1140 includes support in the compiler, as well as a supporting runtime
1141 library called libitm. To compile code with transactional memory
1142 constructs, use the -fgnu-tm option.
1143 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC,
1144 and 32-bit/64-bit x86 platforms.
1145 For more details on transactional memory see [6]the GCC WiKi.
1146 * Support for atomic operations specifying the C++11/C11 memory model
1147 has been added. These new __atomic routines replace the existing
1148 __sync built-in routines.
1149 Atomic support is also available for memory blocks. Lock-free
1150 instructions will be used if a memory block is the same size and
1151 alignment as a supported integer type. Atomic operations which do
1152 not have lock-free support are left as function calls. A set of
1153 library functions is available on the GCC atomic wiki in the
1154 "External Atomics Library" section.
1155 For more details on the memory models and features, see the
1157 * When a binary operation is performed on vector types and one of the
1158 operands is a uniform vector, it is possible to replace the vector
1159 with the generating element. For example:
1160 typedef int v4si __attribute__ ((vector_size (16)));
1161 v4si res, a = {1,2,3,4};
1164 res = 2 + a; /* means {2,2,2,2} + a */
1165 res = a - x; /* means a - {x,x,x,x} */
1169 * There is support for some more features from the C11 revision of
1170 the ISO C standard. GCC now accepts the options -std=c11 and
1171 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x.
1172 + Unicode strings (previously supported only with options such
1173 as -std=gnu11, now supported with -std=c11), and the
1174 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__.
1175 + Nonreturning functions (_Noreturn and <stdnoreturn.h>).
1176 + Alignment support (_Alignas, _Alignof, max_align_t,
1178 + A built-in function __builtin_complex is provided to support C
1179 library implementation of the CMPLX family of macros.
1183 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat
1184 options, which are equivalent to -std=c++0x, -std=gnu++0x, and
1185 -Wc++0x-compat, respectively.
1186 * G++ now implements [8]C++11 extended friend syntax:
1191 static const int I = 2;
1201 * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit
1205 virtual void f() const final;
1206 virtual void f(int);
1210 void f() const; // error: D::f attempts to override final B::f
1211 void f(long) override; // error: doesn't override anything
1212 void f(int) override; // ok
1216 struct F: E { }; // error: deriving from final class
1218 * G++ now implements [10]C++11 non-static data member initializers.
1222 } a; // initializes a.i to 42
1224 * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11
1225 user-defined literals.
1227 // Not actually a good approximation. :)
1228 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; }
1229 long double pi = 180.0_degrees;
1231 * G++ now implements [12]C++11 alias-declarations.
1233 template <class T> using Ptr = T*;
1234 Ptr<int> ip; // decltype(ip) is int*
1236 * Thanks to Ville Voutilainen and Pedro Lamarao, G++ now implements
1237 [13]C++11 delegating constructors.
1241 A(): A(42) { } // delegate to the A(int) constructor
1244 * G++ now fully implements C++11 atomic classes rather than just
1245 integer derived classes.
1251 std::atomic<POD> my_atomic_POD;
1253 * G++ now sets the predefined macro __cplusplus to the correct value,
1254 199711L for C++98/03, and 201103L for C++11.
1255 * G++ now correctly implements the two-phase lookup rules such that
1256 an unqualified name used in a template must have an appropriate
1257 declaration found either in scope at the point of definition of the
1258 template or by argument-dependent lookup at the point of
1259 instantiation. As a result, code that relies on a second
1260 unqualified lookup at the point of instantiation to find functions
1261 declared after the template or in dependent bases will be rejected.
1262 The compiler will suggest ways to fix affected code, and using the
1263 -fpermissive compiler flag will allow the code to compile with a
1267 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup
1268 void g(int) { } // fix by moving this declaration before the declaration of f
1272 // error, B::g(B) not found by argument-dependent lookup
1273 void f() { g(T()); } // fix by using this->g or A::g
1276 struct B { void g(B); };
1284 * G++ now properly re-uses stack space allocated for temporary
1285 objects when their lifetime ends, which can significantly lower
1286 stack consumption for some C++ functions. As a result of this, some
1287 code with undefined behavior will now break:
1289 const int &f(const int &i) { return i; }
1291 const int &x = f(1);
1292 const int &y = f(2);
1294 Here, x refers to the temporary allocated to hold the 1 argument,
1295 which only lives until the end of the initialization; it
1296 immediately becomes a dangling reference. So the next statement
1297 re-uses the stack slot to hold the 2 argument, and users of x get
1299 Note that this should not cause any change of behavior for
1300 temporaries of types with non-trivial destructors, as they are
1301 already destroyed at end of full-expression; the change is that now
1302 the storage is released as well.
1303 * A new command-line option -Wdelete-non-virtual-dtor has been added
1304 to warn when delete is used to destroy an instance of a class which
1305 has virtual functions and non-virtual destructor. It is unsafe to
1306 delete an instance of a derived class through a pointer to a base
1307 class if the base class does not have a virtual destructor. This
1308 warning is enabled by -Wall.
1309 * A new command-line option -Wzero-as-null-pointer-constant has been
1310 added to warn when a literal '0' is used as null pointer constant.
1311 It can be useful to facilitate the conversion to nullptr in C++11.
1312 * As per C++98, access-declarations are now deprecated by G++.
1313 Using-declarations are to be used instead. Furthermore, some
1314 efforts have been made to improve the support of class scope
1315 using-declarations. In particular, using-declarations referring to
1316 a dependent type now work as expected ([14]bug c++/14258).
1317 * The ELF symbol visibility of a template instantiation is now
1318 properly constrained by the visibility of its template arguments
1319 ([15]bug c++/35688).
1321 Runtime Library (libstdc++)
1323 * [16]Improved experimental support for the new ISO C++ standard,
1325 + using noexcept in most of the library;
1326 + implementations of pointer_traits, allocator_traits and
1327 scoped_allocator_adaptor;
1328 + uses-allocator construction for tuple;
1329 + vector meets the allocator-aware container requirements;
1330 + replacing monotonic_clock with steady_clock;
1331 + enabling the thread support library on most POSIX targets;
1332 + many small improvements to conform to the FDIS.
1333 * Added --enable-clocale=newlib configure option.
1334 * Debug Mode iterators for unordered associative containers.
1335 * Avoid polluting the global namespace and do not include <unistd.h>.
1339 * The compile flag [17]-fstack-arrays has been added, which causes
1340 all local arrays to be put on stack memory. For some programs this
1341 will improve the performance significantly. If your program uses
1342 very large local arrays, it is possible that you will have to
1343 extend your runtime limits for stack memory.
1344 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and
1346 * Front-end optimizations can now be selected by the
1347 [21]-ffrontend-optimize option and deselected by the
1348 -fno-frontend-optimize option.
1349 * When front-end optimization removes a function call,
1350 [22]-Wfunction-elimination warns about that.
1351 * When performing front-end-optimization, the
1352 [23]-faggressive-function-elimination option allows the removal of
1353 duplicate function calls even for impure functions.
1354 * The flag [24]-Wreal-q-constant has been added, which warns if
1355 floating-point literals have been specified using q (such as
1356 1.0q0); the q marker is now supported as a vendor extension to
1357 denote quad precision (REAL(16) or, if not available, REAL(10)).
1358 Consider using a kind parameter (such as in 1.0_qp) instead, which
1359 can be obtained via [25]SELECTED_REAL_KIND.
1360 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU
1361 Fortran now always prints error messages to standard error. If you
1362 wish to redirect standard error, please consult the manual for your
1363 OS, shell, batch environment etc. as appropriate.
1364 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment
1365 variable have been removed. When encountering a serious error,
1366 gfortran will now always abort the program. Whether a core dump is
1367 generated depends on the user environment settings; see the ulimit
1368 -c setting for POSIX shells, limit coredumpsize for C shells, and
1369 the [26]WER user-mode dumps settings on Windows.
1370 * The [27]-fbacktrace option is now enabled by default. When
1371 encountering a fatal error, gfortran will attempt to print a
1372 backtrace to standard error before aborting. It can be disabled
1373 with -fno-backtrace. Note: On POSIX targets with the addr2line
1374 utility from GNU binutils, GNU Fortran can print a backtrace with
1375 function name, file name, line number information in addition to
1376 the addresses; otherwise only the addresses are printed.
1378 + Generic interface names which have the same name as derived
1379 types are now supported, which allows to write constructor
1380 functions. Note that Fortran does not support static
1381 constructor functions; only default initialization or an
1382 explicit structure-constructor initialization are available.
1383 + [29]Polymorphic (class) arrays are now supported.
1385 + Support for the DO CONCURRENT construct has been added, which
1386 allows the user to specify that individual loop iterations
1387 have no interdependencies.
1388 + [31]Coarrays: Full single-image support except for polymorphic
1389 coarrays. Additionally, preliminary support for multiple
1390 images via an MPI-based [32]coarray communication library has
1391 been added. Note: The library version is not yet usable as
1392 remote coarray access is not yet possible.
1394 + New flag [34]-std=f2008ts permits programs that are expected
1395 to conform to the Fortran 2008 standard and the draft
1396 Technical Specification (TS) 29113 on Further Interoperability
1398 + The OPTIONAL attribute is now allowed for dummy arguments of
1400 + The RANK intrinsic has been added.
1401 + The implementation of the ASYNCHRONOUS attribute in GCC is
1402 compatible with the candidate draft of TS 29113 (since GCC
1407 * GCC 4.7 implements the [35]Go 1 language standard. The library
1408 support in 4.7.0 is not quite complete, due to release timing.
1409 Release 4.7.1 includes complete support for Go 1. The Go library is
1410 from the Go 1.0.1 release.
1411 * Go has been tested on GNU/Linux and Solaris platforms. It may work
1412 on other platforms as well.
1414 New Targets and Target Specific Improvements
1418 * GCC now supports the Cortex-A7 processor implementing the v7-a
1419 version of the architecture using the option -mcpu=cortex-a7.
1420 * The default vector size in auto-vectorization for NEON is now 128
1421 bits. If vectorization fails thusly, the vectorizer tries again
1422 with 64-bit vectors.
1423 * A new option -mvectorize-with-neon-double was added to allow users
1424 to change the vector size to 64 bits.
1428 * GCC now supports the XMEGA architecture. This requires GNU binutils
1430 * Support for the [36]named address spaces __flash, __flash1, ...,
1431 __flash5 and __memx has been added. These address spaces locate
1432 read-only data in flash memory and allow reading from flash memory
1433 by means of ordinary C code, i.e. without the need of (inline)
1436 const __flash int values[] = { 42, 31 };
1438 int add_values (const __flash int *p, int i)
1440 return values[i] + *p;
1443 * Support has been added for the AVR-specific configure option
1444 --with-avrlibc=yes in order to arrange for better integration of
1445 [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2
1446 and newer and will only take effect in non-RTEMS configurations. If
1447 avr-gcc is configured for RTEMS, the option will be ignored which
1448 is the same as specifying --with-avrlibc=no. See [38]PR54461 for
1449 more technical details.
1450 * Support for AVR-specific [39]built-in functions has been added.
1451 * Support has been added for the signed and unsigned 24-bit scalar
1452 integer types __int24 and __uint24.
1453 * New command-line options -maccumulate-args, -mbranch-cost=cost and
1454 -mstrict-X were added to allow better fine-tuning of code
1456 * The command option -fdata-sections now also takes affect on the
1457 section names of variables with the progmem attribute.
1458 * A new inline assembler print modifier %i to print a RAM address as
1459 I/O address has been added:
1461 #include <avr/io.h> /* Port Definitions from AVR-LibC */
1463 void set_portb (uint8_t value)
1465 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory");
1468 The offset between an I/O address and the RAM address for that I/O
1469 location is device-specific. This offset is taken into account when
1470 printing a RAM address with the %i modifier so that the address is
1471 suitable to be used as operand in an I/O command. The address must
1472 be a constant integer known at compile time.
1473 * The inline assembler constraint "R" to represent integers in the
1474 range -6 ... 5 has been removed without replacement.
1475 * Many optimizations to:
1476 + 64-bit integer arithmetic
1477 + Widening multiplication
1478 + Integer division by a constant
1479 + Avoid constant reloading in multi-byte instructions.
1480 + Micro-optimizations for special instruction sequences.
1481 + Generic built-in functions like __builtin_ffs*,
1482 __builtin_clz*, etc.
1483 + If-else decision trees generated by switch instructions
1484 + Merging of data located in flash memory
1485 + New libgcc variants for devices with 8-bit wide stack pointer
1487 * Better documentation:
1488 + Handling of EIND and indirect jumps on devices with more than
1489 128 KiB of program memory.
1490 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function
1492 + Function attributes OS_main and OS_task.
1493 + AVR-specific built-in macros.
1497 * Support has been added for the Texas Instruments C6X family of
1502 * Support has been added for National Semiconductor's CR16
1507 * Support has been added for Adapteva's Epiphany architecture.
1511 * Support for Intel AVX2 intrinsics, built-in functions and code
1512 generation is available via -mavx2.
1513 * Support for Intel BMI2 intrinsics, built-in functions and code
1514 generation is available via -mbmi2.
1515 * Implementation and automatic generation of __builtin_clz* using the
1516 lzcnt instruction is available via -mlzcnt.
1517 * Support for Intel FMA3 intrinsics and code generation is available
1519 * A new -mfsgsbase command-line option is available that makes GCC
1520 generate new segment register read/write instructions through
1521 dedicated built-ins.
1522 * Support for the new Intel rdrnd instruction is available via
1524 * Two additional AVX vector conversion instructions are available via
1526 * Support for new Intel processor codename IvyBridge with RDRND,
1527 FSGSBASE and F16C is available through -march=core-avx-i.
1528 * Support for the new Intel processor codename Haswell with AVX2,
1529 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2.
1530 * Support for new AMD family 15h processors (Piledriver core) is now
1531 available through -march=bdver2 and -mtune=bdver2 options.
1532 * Support for [40]the x32 psABI is now available through the -mx32
1534 * Windows mingw targets are using the -mms-bitfields option by
1536 * Windows x86 targets are using the __thiscall calling convention for
1537 C++ class-member functions.
1538 * Support for the configure option --with-threads=posix for Windows
1543 * GCC now supports thread-local storage (TLS) for MIPS16. This
1544 requires GNU binutils 2.22 or later.
1545 * GCC can now generate code specifically for the Cavium Octeon+ and
1546 Octeon2 processors. The associated command-line options are
1547 -march=octeon+ and -march=octeon2 respectively. Both options
1548 require GNU binutils 2.22 or later.
1549 * GCC can now work around certain 24k errata, under the control of
1550 the command-line option -mfix-24k. These workarounds require GNU
1551 binutils 2.20 or later.
1552 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build
1553 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux
1554 toolchain that generates 32-bit code by default. Use the
1555 configure-time option --enable-targets=all to select these extra
1557 * Passing -fno-delayed-branch now also stops the assembler from
1558 automatically filling delay slots.
1562 * Vectors of type vector long long or vector long are passed and
1563 returned using the same method as other vectors with the VSX
1564 instruction set. Previously GCC did not adhere to the ABI for
1565 128-bit vectors with 64-bit integer base types (PR 48857). This
1566 will also be fixed in the GCC 4.6.1 and 4.5.4 releases.
1567 * A new option -mno-pointers-to-nested-functions was added to allow
1568 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify
1569 that the compiler should not load up the chain register (r11)
1570 before calling a function through a pointer. If you use this
1571 option, you cannot call nested functions through a pointer, or call
1572 other languages that might use the static chain.
1573 * A new option msave-toc-indirect was added to allow AIX
1574 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we
1575 save the TOC in the prologue for indirect calls or generate the
1576 save inline. This can speed up some programs that call through a
1577 function pointer a lot, but it can slow down other functions that
1578 only call through a function pointer in exceptional cases.
1579 * The PowerPC port will now enable machine-specific built-in
1580 functions when the user switches the target machine using the
1581 #pragma GCC target or __attribute__ ((__target__ ("target"))) code
1582 sequences. In addition, the target macros are updated. However, due
1583 to the way the -save-temps switch is implemented, you won't see the
1584 effect of these additional macros being defined in preprocessor
1589 * A new option -msoft-atomic has been added. When it is specified,
1590 GCC will generate GNU/Linux-compatible gUSA atomic sequences for
1591 the new __atomic routines.
1592 * Since it is neither supported by GAS nor officially documented,
1593 code generation for little endian SH2A has been disabled.
1594 Specifying -ml with -m2a* will now result in a compiler error.
1595 * The defunct -mbranch-cost option has been fixed.
1596 * Some improvements to the generated code of:
1597 + Utilization of the tst #imm,R0 instruction.
1598 + Dynamic shift instructions on SH2A.
1599 + Integer absolute value calculations.
1600 * The -mdiv= option for targets other than SHmedia has been fixed and
1605 * The option -mflat has been reinstated. When it is specified, the
1606 compiler will generate code for a single register window model.
1607 This is essentially a new implementation and the corresponding
1608 debugger support has been added to GDB 7.4.
1609 * Support for the options -mtune=native and -mcpu=native has been
1610 added on selected native platforms (GNU/Linux and Solaris).
1611 * Support for the SPARC T3 (Niagara 3) processor has been added.
1613 + An intrinsics header visintrin.h has been added.
1614 + Builtin intrinsics for the VIS 1.0 edge handling and pixel
1615 compare instructions have been added.
1616 + The little-endian version of alignaddr is now supported.
1617 + When possible, VIS builtins are marked const, which should
1618 increase the compiler's ability to optimize VIS operations.
1619 + The compiler now properly tracks the %gsr register and how it
1620 behaves as an input for various VIS instructions.
1621 + Akin to fzero, the compiler can now generate fone instructions
1622 in order to set all of the bits of a floating-point register
1624 + The documentation for the VIS intrinsics in the GCC manual has
1625 been brought up to date and many inaccuracies were fixed.
1626 + Intrinsics for the VIS 2.0 bmask, bshuffle, and
1627 non-condition-code setting edge instructions have been added.
1628 Their availability is controlled by the new -mvis2 and
1629 -mno-vis2 options. They are enabled by default on
1630 UltraSPARC-III and later CPUs.
1631 * Support for UltraSPARC Fused Multiply-Add floating-point extensions
1632 has been added. These instructions are enabled by default on SPARC
1633 T3 (Niagara 3) and later CPUs.
1637 * Support has been added for the Tilera TILE-Gx and TILEPro families
1640 Other significant improvements
1642 * A new option (-grecord-gcc-switches) was added that appends
1643 compiler command-line options that might affect code generation to
1644 the DW_AT_producer attribute string in the DWARF debugging
1646 * GCC now supports various new GNU extensions to the DWARF debugging
1647 information format, like [41]entry value and [42]call site
1648 information, [43]typed DWARF stack or [44]a more compact macro
1649 representation. Support for these extensions has been added to GDB
1650 7.4. They can be disabled through the -gstrict-dwarf command-line
1655 This is the [45]list of problem reports (PRs) from GCC's bug tracking
1656 system that are known to be fixed in the 4.7.1 release. This list might
1657 not be complete (that is, it is possible that some PRs that have been
1658 fixed are not listed here).
1660 The Go frontend in the 4.7.1 release fully supports the [46]Go 1
1665 This is the [47]list of problem reports (PRs) from GCC's bug tracking
1666 system that are known to be fixed in the 4.7.2 release. This list might
1667 not be complete (that is, it is possible that some PRs that have been
1668 fixed are not listed here).
1672 This is the [48]list of problem reports (PRs) from GCC's bug tracking
1673 system that are known to be fixed in the 4.7.3 release. This list might
1674 not be complete (that is, it is possible that some PRs that have been
1675 fixed are not listed here).
1678 For questions related to the use of GCC, please consult these web
1679 pages and the [49]GCC manuals. If that fails, the
1680 [50]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1681 web pages and the development of GCC are welcome on our developer
1682 list at [51]gcc@gcc.gnu.org. All of [52]our lists have public
1685 Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and
1686 distribution of this entire article is permitted in any medium,
1687 provided this notice is preserved.
1689 These pages are [54]maintained by the GCC team. Last modified
1694 1. http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html
1695 2. http://savannah.nongnu.org/bugs/?35407
1696 3. http://gcc.gnu.org/PR18145
1697 4. http://gcc.gnu.org/gcc-4.7/porting_to.html
1698 5. http://openmp.org/wp/openmp-specifications/
1699 6. http://gcc.gnu.org/wiki/TransactionalMemory
1700 7. http://gcc.gnu.org/wiki/Atomic/GCCMM
1701 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1702 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1703 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1704 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1705 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1706 13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
1707 14. http://gcc.gnu.org/PR14258
1708 15. http://gcc.gnu.org/PR35688
1709 16. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/libstdc++/manual/manual/status.html#status.iso.2011
1710 17. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
1711 18. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689
1712 19. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270
1713 20. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
1714 21. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275
1715 22. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170
1716 23. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270
1717 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
1718 25. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html
1719 26. http://msdn.microsoft.com/en-us/library/bb787181%28v=vs.85%29.aspx
1720 27. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183
1721 28. http://gcc.gnu.org/wiki/Fortran2003Status
1722 29. http://gcc.gnu.org/wiki/OOP
1723 30. http://gcc.gnu.org/wiki/Fortran2008Status
1724 31. http://gcc.gnu.org/wiki/Coarray
1725 32. http://gcc.gnu.org/wiki/CoarrayLib
1726 33. http://gcc.gnu.org/wiki/TS29113Status
1727 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
1728 35. http://weekly.golang.org/doc/go1.html
1729 36. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html
1730 37. http://nongnu.org/avr-libc/
1731 38. http://gcc.gnu.org/PR54461
1732 39. http://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html
1733 40. https://sites.google.com/site/x32abi/
1734 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1
1735 42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2
1736 43. http://www.dwarfstd.org/doc/040408.1.html
1737 44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1
1738 45. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1
1739 46. http://weekly.golang.org/doc/go1.html
1740 47. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2
1741 48. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3
1742 49. http://gcc.gnu.org/onlinedocs/
1743 50. mailto:gcc-help@gcc.gnu.org
1744 51. mailto:gcc@gcc.gnu.org
1745 52. http://gcc.gnu.org/lists.html
1746 53. http://www.fsf.org/
1747 54. http://gcc.gnu.org/about.html
1748 55. http://validator.w3.org/check/referer
1749 ======================================================================
1750 http://gcc.gnu.org/gcc-4.6/index.html
1751 GCC 4.6 Release Series
1755 The [1]GNU project and the GCC developers are pleased to announce the
1756 release of GCC 4.6.4.
1758 This release is a bug-fix release, containing fixes for regressions in
1759 GCC 4.6.3 relative to previous releases of GCC.
1764 April 12, 2013 ([2]changes, [3]documentation)
1767 March 1, 2012 ([4]changes, [5]documentation)
1770 October 26, 2011 ([6]changes, [7]documentation)
1773 June 27, 2011 ([8]changes, [9]documentation)
1776 March 25, 2011 ([10]changes, [11]documentation)
1778 References and Acknowledgements
1780 GCC used to stand for the GNU C Compiler, but since the compiler
1781 supports several other languages aside from C, it now stands for the
1782 GNU Compiler Collection.
1784 A list of [12]successful builds is updated as new information becomes
1787 The GCC developers would like to thank the numerous people that have
1788 contributed new features, improvements, bug fixes, and other changes as
1789 well as test results to GCC. This [13]amazing group of volunteers is
1790 what makes GCC successful.
1792 For additional information about GCC please refer to the [14]GCC
1793 project web site or contact the [15]GCC development mailing list.
1795 To obtain GCC please use [16]our mirror sites or [17]our SVN server.
1798 For questions related to the use of GCC, please consult these web
1799 pages and the [18]GCC manuals. If that fails, the
1800 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1801 web pages and the development of GCC are welcome on our developer
1802 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
1805 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
1806 distribution of this entire article is permitted in any medium,
1807 provided this notice is preserved.
1809 These pages are [23]maintained by the GCC team. Last modified
1814 1. http://www.gnu.org/
1815 2. http://gcc.gnu.org/gcc-4.6/changes.html
1816 3. http://gcc.gnu.org/onlinedocs/4.6.4/
1817 4. http://gcc.gnu.org/gcc-4.6/changes.html
1818 5. http://gcc.gnu.org/onlinedocs/4.6.3/
1819 6. http://gcc.gnu.org/gcc-4.6/changes.html
1820 7. http://gcc.gnu.org/onlinedocs/4.6.2/
1821 8. http://gcc.gnu.org/gcc-4.6/changes.html
1822 9. http://gcc.gnu.org/onlinedocs/4.6.1/
1823 10. http://gcc.gnu.org/gcc-4.6/changes.html
1824 11. http://gcc.gnu.org/onlinedocs/4.6.0/
1825 12. http://gcc.gnu.org/gcc-4.6/buildstat.html
1826 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
1827 14. http://gcc.gnu.org/index.html
1828 15. mailto:gcc@gcc.gnu.org
1829 16. http://gcc.gnu.org/mirrors.html
1830 17. http://gcc.gnu.org/svn.html
1831 18. http://gcc.gnu.org/onlinedocs/
1832 19. mailto:gcc-help@gcc.gnu.org
1833 20. mailto:gcc@gcc.gnu.org
1834 21. http://gcc.gnu.org/lists.html
1835 22. http://www.fsf.org/
1836 23. http://gcc.gnu.org/about.html
1837 24. http://validator.w3.org/check/referer
1838 ======================================================================
1839 http://gcc.gnu.org/gcc-4.6/changes.html
1840 GCC 4.6 Release Series
1841 Changes, New Features, and Fixes
1845 * The options -b <machine> and -V <version> have been removed because
1846 they were unreliable. Instead, users should directly run
1847 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to
1848 run a different version of gcc.
1849 * GCC now has stricter checks for invalid command-line options. In
1850 particular, when gcc was called to link object files rather than
1851 compile source code, it would previously accept and ignore all
1852 options starting with --, including linker options such as
1853 --as-needed and --export-dynamic, although such options would
1854 result in errors if any source code was compiled. Such options, if
1855 unknown to the compiler, are now rejected in all cases; if the
1856 intent was to pass them to the linker, options such as
1857 -Wl,--as-needed should be used.
1858 * Versions of the GNU C library up to and including 2.11.1 included
1859 an [1]incorrect implementation of the cproj function. GCC optimizes
1860 its builtin cproj according to the behavior specified and allowed
1861 by the ISO C99 standard. If you want to avoid discrepancies between
1862 the C library and GCC's builtin transformations when using cproj in
1863 your code, use GLIBC 2.12 or later. If you are using an older GLIBC
1864 and actually rely on the incorrect behavior of cproj, then you can
1865 disable GCC's transformations using -fno-builtin-cproj.
1866 * The C-only intermodule optimization framework (IMA, enabled by
1867 -combine) has been removed in favor of the new generic link-time
1868 optimization framework (LTO) introduced in [2]GCC 4.5.0.
1869 * GCC now ships with the LGPL-licensed libquadmath library, which
1870 provides quad-precision mathematical functions for targets with a
1871 __float128 datatype. __float128 is available for targets on 32-bit
1872 x86, x86-64 and Itanium architectures. The libquadmath library is
1873 automatically built on such targets when building the Fortran
1875 * New -Wunused-but-set-variable and -Wunused-but-set-parameter
1876 warnings were added for C, C++, Objective-C and Objective-C++.
1877 These warnings diagnose variables respective parameters which are
1878 only set in the code and never otherwise used. Usually such
1879 variables are useless and often even the value assigned to them is
1880 computed needlessly, sometimes expensively. The
1881 -Wunused-but-set-variable warning is enabled by default by -Wall
1882 flag and -Wunused-but-set-parameter by -Wall -Wextra flags.
1883 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
1884 rules for the layout of vectors that could lead to wrong code being
1885 generated. Vectors larger than 8 bytes in size are now by default
1886 aligned to an 8-byte boundary. This is an ABI change: code that
1887 makes explicit use of vector types may be incompatible with binary
1888 objects built with older versions of GCC. Auto-vectorized code is
1889 not affected by this change. (This change affects GCC versions
1890 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.)
1891 * On AVR, variables with the progmem attribute to locate data in
1892 flash memory must be qualified as const.
1893 * Support for a number of older systems and recently unmaintained or
1894 untested target ports of GCC has been declared obsolete in GCC 4.6.
1895 Unless there is activity to revive them, the next release of GCC
1896 will have their sources permanently removed.
1897 All GCC ports for the following processor architectures have been
1899 + Argonaut ARC (arc-*)
1900 + National Semiconductor CRX (crx-*)
1901 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*,
1902 m68hc12-*-*, m6812-*-*)
1903 + Sunplus S+core (score-*)
1904 The following ports for individual systems on particular
1905 architectures have been obsoleted:
1906 + Interix (i[34567]86-*-interix3*)
1907 + NetWare x86 (i[3456x]86-*-netware*)
1908 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*)
1909 + MCore PE (mcore-*-pe*)
1910 + SH SymbianOS (sh*-*-symbianelf*)
1911 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*)
1912 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*)
1913 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*,
1914 vax-*-netbsd*, but not *-*-netbsdelf*)
1915 The i[34567]86-*-pe alias for Cygwin targets has also been
1916 obsoleted; users should configure for i[34567]86-*-cygwin* instead.
1917 Certain configure options to control the set of libraries built
1918 with GCC on some targets have been obsoleted. On ARM targets, the
1919 options --disable-fpu, --disable-26bit, --disable-underscore,
1920 --disable-interwork, --disable-biendian and --disable-nofmult have
1921 been obsoleted. On MIPS targets, the options
1922 --disable-single-float, --disable-biendian and --disable-softfloat
1923 have been obsoleted.
1924 * Support has been removed for all the [3]configurations obsoleted in
1926 * More information on porting to GCC 4.6 from previous versions of
1927 GCC can be found in the [4]porting guide for this release.
1929 General Optimizer Improvements
1931 * A new general optimization level, -Ofast, has been introduced. It
1932 combines the existing optimization level -O3 with options that can
1933 affect standards compliance but result in better optimized code.
1934 For example, -Ofast enables -ffast-math.
1935 * Link-time optimization improvements:
1936 + The [5]Scalable Whole Program Optimizer (WHOPR) project has
1937 stabilized to the point of being usable. It has become the
1938 default mode when using the LTO optimization model. Link time
1939 optimization can now split itself into multiple parallel
1940 compilations. Parallelism is controlled with -flto=n (where n
1941 specifies the number of compilations to execute in parallel).
1942 GCC can also cooperate with a GNU make job server by
1943 specifying the -flto=jobserver option and adding + to the
1944 beginning of the Makefile rule executing the linker.
1945 Classical LTO mode can be enforced by -flto-partition=none.
1946 This may result in small code quality improvements.
1947 + A large number of bugs were fixed. GCC itself, Mozilla Firefox
1948 and other large applications can be built with LTO enabled.
1949 + The linker plugin support improvements
1950 o Linker plugin is now enabled by default when the linker
1951 is detected to have plugin support. This is the case for
1952 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and
1953 the Gold linker on ELF targets. Plugin support of the
1954 Apple linker on Darwin is not compatible with GCC. The
1955 linker plugin can also be controlled by the
1956 -fuse-linker-plugin command line option.
1957 o Resolution information from the linker plugin is used to
1958 drive whole program assumptions. Use of the linker plugin
1959 results in more aggressive optimization on binaries and
1960 on shared libraries that use the hidden visibility
1961 attribute. Consequently the use of -fwhole-program is not
1962 necessary in addition to LTO.
1963 + Hidden symbols used from non-LTO objects now have to be
1964 explicitly annotated with externally_visible when the linker
1966 + C++ inline functions and virtual tables are now privatized
1967 more aggressively, leading to better inter-procedural
1968 optimization and faster dynamic linking.
1969 + Memory usage and intermediate language streaming performance
1971 + Static constructors and destructors from individual units are
1972 inlined into a single function. This can significantly improve
1973 startup times of large C++ applications where static
1974 constructors are very common. For example, static constructors
1975 are used when including the iostream header.
1976 + Support for the Ada language has been added.
1977 * Interprocedural optimization improvements
1978 + The interprocedural framework was re-tuned for link time
1979 optimization. Several scalability issues were resolved.
1980 + Improved auto-detection of const and pure functions. Newly,
1981 noreturn functions are auto-detected.
1982 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is
1983 available that informs users when adding attributes to headers
1984 might improve code generation.
1985 + A number of inlining heuristic improvements. In particular:
1986 o Partial inlining is now supported and enabled by default
1987 at -O2 and greater. The feature can be controlled via
1989 Partial inlining splits functions with short hot path to
1990 return. This allows more aggressive inlining of the hot
1991 path leading to better performance and often to code size
1992 reductions (because cold parts of functions are not
1994 o Scalability for large compilation units was improved
1996 o Inlining of callbacks is now more aggressive.
1997 o Virtual methods are considered for inlining when the
1998 caller is inlined and devirtualization is then possible.
1999 o Inlining when optimizing for size (either in cold regions
2000 of a program or when compiling with -Os) was improved to
2001 better handle C++ programs with larger abstraction
2002 penalty, leading to smaller and faster code.
2003 + The IPA reference optimization pass detecting global variables
2004 used or modified by functions was strengthened and sped up.
2005 + Functions whose address was taken are now optimized out when
2006 all references to them are dead.
2007 + A new inter-procedural static profile estimation pass detects
2008 functions that are executed once or unlikely to be executed.
2009 Unlikely executed functions are optimized for size. Functions
2010 executed once are optimized for size except for the inner
2012 + On most targets with named section support, functions used
2013 only at startup (static constructors and main), functions used
2014 only at exit and functions detected to be cold are placed into
2015 separate text segment subsections. This extends the
2016 -freorder-functions feature and is controlled by the same
2017 switch. The goal is to improve the startup time of large C++
2019 Proper function placement requires linker support. GNU ld
2020 2.21.51 on ELF targets was updated to place those functions
2021 together within the text section leading to better code
2022 locality and faster startup times of large C++ programs. The
2023 feature is also supported in the Apple linker. Support in the
2024 gold linker is planned.
2025 * A new switch -fstack-usage has been added. It makes the compiler
2026 output stack usage information for the program, on a per-function
2027 basis, in an auxiliary file.
2028 * A new switch -fcombine-stack-adjustments has been added. It can be
2029 used to enable or disable the compiler's stack-slot combining pass
2030 which before was enabled automatically at -O1 and above, but could
2031 not be controlled on its own.
2032 * A new switch -fstrict-volatile-bitfields has been added. Using it
2033 indicates that accesses to volatile bitfields should use a single
2034 access of the width of the field's type. This option can be useful
2035 for precisely defining and accessing memory-mapped peripheral
2036 registers from C or C++.
2038 Compile time and memory usage improvements
2040 * Datastructures used by the dataflow framework in GCC were
2041 reorganized for better memory usage and more cache locality.
2042 Compile time is improved especially on units with large functions
2043 (possibly resulting from a lot of inlining) not fitting into the
2044 processor cache. The compile time of the GCC C compiler binary with
2045 link-time optimization went down by over 10% (benchmarked on x86-64
2048 New Languages and Language specific improvements
2052 * Stack checking has been improved on selected architectures (Alpha,
2053 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack
2054 overflows in all cases on these architectures.
2055 * Initial support for Ada 2012 has been added.
2059 * A new warning, enabled by -Wdouble-promotion, has been added that
2060 warns about cases where a value of type float is implicitly
2061 promoted to double. This is especially helpful for CPUs that handle
2062 the former in hardware, but emulate the latter in software.
2063 * A new function attribute leaf was introduced. This attribute allows
2064 better inter-procedural optimization across calls to functions that
2065 return to the current unit only via returning or exception
2066 handling. This is the case for most library functions that have no
2068 * Support for a new data type __int128 for targets having wide enough
2069 machine-mode support.
2070 * The new function attribute callee_pop_aggregate allows to specify
2071 if the caller or callee is responsible for popping the aggregate
2072 return pointer value from the stack.
2073 * Support for selectively enabling and disabling warnings via #pragma
2074 GCC diagnostic has been added. For instance:
2075 #pragma GCC diagnostic error "-Wuninitialized"
2076 foo(a); /* error is given for this one */
2077 #pragma GCC diagnostic push
2078 #pragma GCC diagnostic ignored "-Wuninitialized"
2079 foo(b); /* no diagnostic for this one */
2080 #pragma GCC diagnostic pop
2081 foo(c); /* error is given for this one */
2082 #pragma GCC diagnostic pop
2083 foo(d); /* depends on command line options */
2085 * The -fmax-errors=N option is now supported. Using this option
2086 causes the compiler to exit after N errors have been issued.
2090 * There is now experimental support for some features from the
2091 upcoming C1X revision of the ISO C standard. This support may be
2092 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions.
2093 Note that this support is experimental and may change incompatibly
2094 in future releases for consistency with changes to the C1X standard
2095 draft. The following features are newly supported as described in
2096 the N1539 draft of C1X (with changes agreed at the March 2011 WG14
2097 meeting); some other features were already supported with no
2098 compiler changes being needed, or have some support but not in full
2099 accord with N1539 (as amended).
2100 + Static assertions (_Static_assert keyword)
2101 + Typedef redefinition
2102 + New macros in <float.h>
2103 + Anonymous structures and unions
2104 * The new -fplan9-extensions option directs the compiler to support
2105 some extensions for anonymous struct fields which are implemented
2106 by the Plan 9 compiler. A pointer to a struct may be automatically
2107 converted to a pointer to an anonymous field when calling a
2108 function, in order to make the types match. An anonymous struct
2109 field whose type is a typedef name may be referred to using the
2114 * Improved [7]experimental support for the upcoming C++0x ISO C++
2115 standard, including support for constexpr (thanks to Gabriel Dos
2116 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide),
2117 noexcept, unrestricted unions, range-based for loops (thanks to
2118 Rodrigo Rivas Costa), opaque enum declarations (thanks also to
2119 Rodrigo), implicitly deleted functions and implicit move
2121 * When an extern declaration within a function does not match a
2122 declaration in the enclosing context, G++ now properly declares the
2123 name within the namespace of the function rather than the namespace
2124 which was open just before the function definition ([8]c++/43145).
2125 * GCC now warns by default when casting integers to larger pointer
2126 types. These warnings can be disabled with the option
2127 -Wno-int-to-pointer-cast, which is now also available in C++.
2128 * G++ no longer optimizes using the assumption that a value of
2129 enumeration type will fall within the range specified by the
2130 standard, since that assumption is easily violated with a
2131 conversion from integer type ([9]c++/43680). The old behavior can
2132 be restored with -fstrict-enums.
2133 * The new -fnothrow-opt flag changes the semantics of a throw()
2134 exception specification to match the proposed semantics of the
2135 noexcept specification: just call terminate if an exception tries
2136 to propagate out of a function with such an exception
2137 specification. This dramatically reduces or eliminates the code
2138 size overhead from adding the exception specification.
2139 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to
2140 a function that the compiler can tell doesn't throw if it would
2141 change the value of a noexcept expression.
2142 * The -Wshadow option now warns if a local variable or type
2143 declaration shadows another type in C++. Note that the compiler
2144 will not warn if a local variable shadows a struct/class/enum, but
2145 will warn if it shadows an explicit typedef.
2146 * When an identifier is not found in the current scope, G++ now
2147 offers suggestions about which identifier might have been intended.
2148 * G++ now issues clearer diagnostics for missing semicolons after
2149 class, struct, and union definitions.
2150 * G++ now issues clearer diagnostics for missing semicolons after
2151 class member declarations.
2152 * G++ now issues clearer diagnostics when a colon is used in a place
2153 where a double-colon was intended.
2154 * G++ no longer accepts mutable on reference members ([10]c++/33558).
2155 Use -fpermissive to allow the old, non-conforming behaviour.
2156 * A few mangling fixes have been made, to attribute const/volatile on
2157 function pointer types, decltype of a plain decl, and use of a
2158 function parameter in the declaration of another parameter. By
2159 default the compiler still uses the old mangling, but emits aliases
2160 with the new mangling on targets that support strong aliases. Users
2161 can switch over entirely to the new mangling with -fabi-version=5
2162 or -fabi-version=0. -Wabi will now warn about code that uses the
2164 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified
2165 type to be default initialized unless the type has a user-declared
2166 default constructor. In 4.6.2 G++ implements the proposed
2167 resolution of [11]DR 253, so default initialization is allowed if
2168 it initializes all subobjects. Code that fails to compile can be
2169 fixed by providing an initializer e.g.
2171 struct B : A { int i; };
2173 Use -fpermissive to allow the old, non-conforming behaviour.
2175 Runtime Library (libstdc++)
2177 * [12]Improved experimental support for the upcoming ISO C++
2178 standard, C++0x, including using constexpr and nullptr.
2179 * Performance improvements to the [13]Debug Mode, thanks to Franc,ois
2181 * Atomic operations used for reference-counting are annotated so that
2182 they can be understood by race detectors such as Helgrind, see
2183 [14]Data Race Hunting.
2184 * Most libstdc++ standard headers have been changed to no longer
2185 include the cstddef header as an implementation detail. Code that
2186 relied on that header being included as side-effect of including
2187 other standard headers will need to include cstddef explicitly.
2191 * On systems supporting the libquadmath library, GNU Fortran now also
2192 supports a quad-precision, kind=16 floating-point data type
2193 (REAL(16), COMPLEX(16)). As the data type is not fully supported in
2194 hardware, calculations might be one to two orders of magnitude
2195 slower than with the 4, 8 or 10 bytes floating-point data types.
2196 This change does not affect systems which support REAL(16) in
2197 hardware nor those which do not support libquadmath.
2198 * Much improved compile time for large array constructors.
2199 * In order to reduce execution time and memory consumption, use of
2200 temporary arrays in assignment expressions is avoided for many
2201 cases. The compiler now reverses loops in order to avoid generating
2202 a temporary array where possible.
2203 * Improved diagnostics, especially with -fwhole-file.
2204 * The -fwhole-file flag is now enabled by default. This improves code
2205 generation and diagnostics. It can be disabled using the deprecated
2206 -fno-whole-file flag.
2207 * Support the generation of Makefile dependencies via the [15]-M...
2208 flags of GCC; you may need to specify the -cpp option in addition.
2209 The dependencies take modules, Fortran's include, and CPP's
2210 #include into account. Note: Using -M for the module path is no
2211 longer supported, use -J instead.
2212 * The flag -Wconversion has been modified to only issue warnings
2213 where a conversion leads to information loss. This drastically
2214 reduces the number of warnings; -Wconversion is thus now enabled
2215 with -Wall. The flag -Wconversion-extra has been added and also
2216 warns about other conversions; -Wconversion-extra typically issues
2217 a huge number of warnings, most of which can be ignored.
2218 * A new command-line option -Wunused-dummy-argument warns about
2219 unused dummy arguments and is included in -Wall. Before,
2220 -Wunused-variable also warned about unused dummy arguments.
2221 * Fortran 2003 support has been extended:
2222 + Improved support for polymorphism between libraries and
2223 programs and for complicated inheritance patterns (cf.
2224 [16]object-oriented programming).
2225 + Experimental support of the ASSOCIATE construct.
2226 + In pointer assignments it is now possible to specify the lower
2227 bounds of the pointer and, for a rank-1 or a simply contiguous
2228 data-target, to remap the bounds.
2229 + Automatic (re)allocation: In intrinsic assignments to
2230 allocatable variables the left-hand side will be automatically
2231 allocated (if unallocated) or reallocated (if the shape or
2232 type parameter is different). To avoid the small performance
2233 penalty, you can use a(:) = ... instead of a = ... for arrays
2234 and character strings - or disable the feature using -std=f95
2235 or -fno-realloc-lhs.
2236 + Deferred type parameter: For scalar allocatable and pointer
2237 variables the character length can be deferred.
2238 + Namelist variables with allocatable and pointer attribute and
2239 nonconstant length type parameter are supported.
2240 * Fortran 2008 support has been extended:
2241 + Experimental [17]coarray support (for one image only, i.e.
2242 num_images() == 1); use the [18]-fcoarray=single flag to
2244 + The STOP and the new ERROR STOP statements now support all
2245 constant expressions.
2246 + Support for the CONTIGUOUS attribute.
2247 + Support for ALLOCATE with MOLD.
2248 + Support for the STORAGE_SIZE intrinsic inquiry function.
2249 + Support of the NORM2 and PARITY intrinsic functions.
2250 + The following bit intrinsics were added: POPCNT and POPPAR for
2251 counting the number of 1 bits and returning the parity; BGE,
2252 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR
2253 for combined left and right shifts, MASKL and MASKR for simple
2254 left and right justified masks, MERGE_BITS for a bitwise merge
2255 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations,
2256 and the transformational bit intrinsics IALL, IANY and
2258 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine.
2259 + Support for the IMPURE attribute for procedures, which allows
2260 for ELEMENTAL procedures without the restrictions of PURE.
2261 + Null pointers (including NULL()) and not allocated variables
2262 can be used as actual argument to optional non-pointer,
2263 non-allocatable dummy arguments, denoting an absent argument.
2264 + Non-pointer variables with TARGET attribute can be used as
2265 actual argument to POINTER dummies with INTENT(IN)
2266 + Pointers including procedure pointers and those in a derived
2267 type (pointer components) can now be initialized by a target
2268 instead of only by NULL.
2269 + The EXIT statement (with construct-name) can now be used to
2270 leave not only the DO but also the ASSOCIATE, BLOCK, IF,
2271 SELECT CASE and SELECT TYPE constructs.
2272 + Internal procedures can now be used as actual argument.
2273 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS
2274 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV
2275 have been added; these arrays contain the supported kind
2276 values for the respective types.
2277 + The module procedures C_SIZEOF of the intrinsic module
2278 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of
2279 ISO_FORTRAN_ENV have been implemented.
2280 + Minor changes: obsolescence diagnostics for ENTRY was added
2281 for -std=f2008; a line may start with a semicolon; for
2282 internal and module procedures END can be used instead of END
2283 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes
2284 a RADIX argument; intrinsic types are supported for
2285 TYPE(intrinsic-type-spec); multiple type-bound procedures can
2286 be declared in a single PROCEDURE statement; implied-shape
2287 arrays are supported for named constants (PARAMETER). The
2288 transformational, three argument versions of BESSEL_JN and
2289 BESSEL_YN were added - the elemental, two-argument version had
2290 been added in GCC 4.4; note that the transformational
2291 functions use a recurrence algorithm.
2295 Support for the [19]Go programming language has been added to GCC. It
2296 is not enabled by default when you build GCC; use the
2297 --enable-languages configure option to build it. The driver program for
2298 compiling Go code is gccgo.
2300 Go is currently known to work on GNU/Linux and RTEMS. Solaris support
2301 is in progress. It may or may not work on other platforms.
2303 Objective-C and Objective-C++
2305 * The -fobjc-exceptions flag is now required to enable Objective-C
2306 exception and synchronization syntax (introduced by the keywords
2307 @try, @catch, @finally and @synchronized).
2308 * A number of Objective-C 2.0 features and extensions are now
2309 supported by GCC. These features are enabled by default; you can
2310 disable them by using the new -fobjc-std=objc1 command-line option.
2311 * The Objective-C 2.0 dot-syntax is now supported. It is an
2312 alternative syntax for using getters and setters; object.count is
2313 automatically converted into [object count] or [object setCount:
2314 ...] depending on context; for example if (object.count > 0) is
2315 automatically compiled into the equivalent of if ([object count] >
2316 0) while object.count = 0; is automatically compiled into the
2317 equivalent ot [object setCount: 0];. The dot-syntax can be used
2318 with instance and class objects and with any setters or getters, no
2319 matter if they are part of a declared property or not.
2320 * Objective-C 2.0 declared properties are now supported. They are
2321 declared using the new @property keyword, and are most commonly
2322 used in conjunction with the new Objective-C 2.0 dot-syntax. The
2323 nonatomic, readonly, readwrite, assign, retain, copy, setter and
2324 getter attributes are all supported. Marking declared properties
2325 with __attribute__ ((deprecated)) is supported too.
2326 * The Objective-C 2.0 @synthesize and @dynamic keywords are
2327 supported. @synthesize causes the compiler to automatically
2328 synthesize a declared property, while @dynamic is used to disable
2329 all warnings for a declared property for which no implementation is
2330 provided at compile time. Synthesizing declared properties requires
2331 runtime support in most useful cases; to be able to use it with the
2332 GNU runtime, appropriate helper functions have been added to the
2333 GNU Objective-C runtime ABI, and are implemented by the GNU
2334 Objective-C runtime library shipped with GCC.
2335 * The Objective-C 2.0 fast enumeration syntax is supported in
2336 Objective-C. This is currently not yet available in Objective-C++.
2337 Fast enumeration requires support in the runtime, and such support
2338 has been added to the GNU Objective-C runtime library (shipped with
2340 * The Objective-C 2.0 @optional keyword is supported. It allows you
2341 to mark methods or properties in a protocol as optional as opposed
2343 * The Objective-C 2.0 @package keyword is supported. It has currently
2344 the same effect as the @public keyword.
2345 * Objective-C 2.0 method attributes are supported. Currently the
2346 supported attributes are deprecated, sentinel, noreturn and format.
2347 * Objective-C 2.0 method argument attributes are supported. The most
2348 widely used attribute is unused, to mark an argument as unused in
2350 * Objective-C 2.0 class and protocol attributes are supported.
2351 Currently the only supported attribute is deprecated.
2352 * Objective-C 2.0 class extensions are supported. A class extension
2353 has the same syntax as a category declaration with no category
2354 name, and the methods and properties declared in it are added
2355 directly to the main class. It is mostly used as an alternative to
2356 a category to add methods to a class without advertising them in
2357 the public headers, with the advantage that for class extensions
2358 the compiler checks that all the privately declared methods are
2359 actually implemented.
2360 * As a result of these enhancements, GCC can now be used to build
2361 Objective-C and Objective-C++ software that uses Foundation and
2362 other important system frameworks with the NeXT runtime on Darwin 9
2363 and Darwin 10 (OSX 10.5 and 10.6).
2364 * Many bugs in the compiler have been fixed in this release; in
2365 particular, LTO can now be used when compiling Objective-C and
2366 Objective-C++ and the parser is much more robust in dealing with
2369 Runtime Library (libobjc)
2371 * The GNU Objective-C runtime library now defines the macro
2372 __GNU_LIBOBJC__ (with a value that is increased at every release
2373 where there is any change to the API) in objc/objc.h, making it
2374 easy to determine if the GNU Objective-C runtime library is being
2375 used, and if so, which version. Previous versions of the GNU
2376 Objective-C runtime library (and other Objective-C runtime
2377 libraries such as the Apple one) do not define this macro.
2378 * A new Objective-C 2.0 API, almost identical to the one implemented
2379 by the Apple Objective-C runtime, has been implemented in the GNU
2380 Objective-C runtime library. The new API hides the internals of
2381 most runtime structures but provides a more extensive set of
2382 functions to operate on them. It is much easier, for example, to
2383 create or modify classes at runtime. The new API also makes it
2384 easier to port software from Apple to GNU as almost no changes
2385 should be required. The old API is still supported for backwards
2386 compatibility; including the old objc/objc-api.h header file
2387 automatically selects the old API, while including the new
2388 objc/runtime.h header file automatically selects the new API.
2389 Support for the old API is being phased out and upgrading the
2390 software to use the new API is strongly recommended. To check for
2391 the availability of the new API, the __GNU_LIBOBJC__ macro can be
2392 used as older versions of the GNU Objective-C runtime library,
2393 which do not support the new API, do not define such a macro.
2394 * Runtime support for @synchronized has been added.
2395 * Runtime support for Objective-C 2.0 synthesized property accessors
2397 * Runtime support for Objective-C 2.0 fast enumeration has been
2400 New Targets and Target Specific Improvements
2404 * GCC now supports the Cortex-M4 processor implementing the v7-em
2405 version of the architecture using the option -mcpu=cortex-m4.
2406 * Scheduling descriptions for the Cortex-M4, the Neon and the
2407 floating point units of the Cortex-A9 and a pipeline description
2408 for the Cortex-A5 have been added.
2409 * Synchronization primitives such as __sync_fetch_and_add and friends
2410 are now inlined for supported architectures rather than calling
2411 into a kernel helper function.
2412 * SSA loop prefetching is enabled by default for the Cortex-A9 at
2414 * Several improvements were committed to improve code generation for
2415 the ARM architecture including a rewritten implementation for load
2416 and store multiples.
2417 * Several enhancements were committed to improve SIMD code generation
2418 for NEON by adding support for widening instructions, misaligned
2419 loads and stores, vector conditionals and support for 64 bit
2421 * Support was added for the Faraday cores fa526, fa606te, fa626te,
2422 fmp626te, fmp626 and fa726te and can be used with the respective
2423 names as parameters to the -mcpu= option.
2424 * Basic support was added for Cortex-A15 and is available through
2426 * GCC for AAPCS configurations now more closely adheres to the AAPCS
2427 specification by enabling -fstrict-volatile-bitfields by default.
2431 * The new -fsplit-stack option permits programs to use a
2432 discontiguous stack. This is useful for threaded programs, in that
2433 it is no longer necessary to specify the maximum stack size when
2434 creating a thread. This feature is currently only implemented for
2435 32-bit and 64-bit x86 GNU/Linux targets.
2436 * Support for emitting profiler counter calls before function
2437 prologues. This is enabled via a new command-line option -mfentry.
2438 * Optimization for the Intel Core 2 processors is now available
2439 through the -march=core2 and -mtune=core2 options.
2440 * Support for Intel Core i3/i5/i7 processors is now available through
2441 the -march=corei7 and -mtune=corei7 options.
2442 * Support for Intel Core i3/i5/i7 processors with AVX is now
2443 available through the -march=corei7-avx and -mtune=corei7-avx
2445 * Support for AMD Bobcat (family 14) processors is now available
2446 through the -march=btver1 and -mtune=btver1 options.
2447 * Support for AMD Bulldozer (family 15) processors is now available
2448 through the -march=bdver1 and -mtune=bdver1 options.
2449 * The default setting (when not optimizing for size) for 32-bit
2450 GNU/Linux and Darwin x86 targets has been changed to
2451 -fomit-frame-pointer. The default can be reverted to
2452 -fno-omit-frame-pointer by configuring GCC with the
2453 --enable-frame-pointer configure option.
2454 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support
2455 __float128 on 32-bit and 64-bit x86 targets.
2456 * AVX floating-point arithmetic can now be enabled by default at
2457 configure time with the new --with-fpmath=avx option.
2458 * The SSA loop prefetching pass is enabled when using -O3 when
2459 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer
2461 * Support for TBM (Trailing Bit Manipulation) built-in functions and
2462 code generation is available via -mtbm.
2463 * Support for AMD's BMI (Bit Manipulation) built-in functions and
2464 code generation is available via -mbmi.
2468 * Support has been added for the Xilinx MicroBlaze softcore processor
2469 (microblaze-elf) embedded target. This configurable processor is
2470 supported on several Xilinx Spartan and Virtex FPGAs.
2474 * GCC now supports the Loongson 3A processor. Its canonical -march=
2475 and -mtune= name is loongson3a.
2479 * The inline assembly register constraint "A" has been renamed "c".
2480 This constraint is used to select a floating-point register that
2481 can be used as the destination of a multiply-accumulate
2483 * New inline assembly register constraints "A" and "D" have been
2484 added. These constraint letters resolve to all general registers
2485 when compiling for AM33, and resolve to address registers only or
2486 data registers only when compiling for MN10300.
2487 * The MDR register is represented in the compiler. One can access the
2488 register via the "z" constraint in inline assembly. It can be
2489 marked as clobbered or used as a local register variable via the
2490 "mdr" name. The compiler uses the RETF instruction if the function
2491 does not modify the MDR register, so it is important that inline
2492 assembly properly annotate any usage of the register.
2496 * GCC now supports the Applied Micro Titan processor with
2498 * The -mrecip option has been added, which indicates whether the
2499 reciprocal and reciprocal square root instructions should be used.
2500 * The -mveclibabi=mass option can be used to enable the compiler to
2501 autovectorize mathematical functions using the Mathematical
2502 Acceleration Subsystem library.
2503 * The -msingle-pic-base option has been added, which instructs the
2504 compiler to avoid loading the PIC base register in function
2505 prologues. The PIC base register must be initialized by the runtime
2507 * The -mblock-move-inline-limit option has been added, which enables
2508 the user to control the maximum size of inlined memcpy calls and
2510 * PowerPC64 GNU/Linux support for applications requiring a large TOC
2511 section has been improved. A new command-line option,
2512 -mcmodel=MODEL, controls this feature; valid values for MODEL are
2513 small, medium, or large.
2514 * The Altivec builtin functions vec_ld and vec_st have been modified
2515 to generate the Altivec memory instructions LVX and STVX, even if
2516 the -mvsx option is used. In the initial GCC 4.5 release, these
2517 builtin functions were changed to generate VSX memory reference
2518 instructions instead of Altivec memory instructions, but there are
2519 differences between the two instructions. If the VSX instruction
2520 set is available, you can now use the new builtin functions
2521 vec_vsx_ld and vec_vsx_st which always generates the VSX memory
2523 * The GCC compiler on AIX now defaults to a process layout with a
2524 larger data space allowing larger programs to be compiled.
2525 * The GCC long double type on AIX 6.1 and above has reverted to 64
2526 bit double precision, matching the AIX XL compiler default, because
2527 of missing C99 symbols required by the GCC runtime.
2528 * The default processor scheduling model and tuning for PowerPC64
2529 GNU/Linux and for AIX 6.1 and above now is POWER7.
2530 * Starting with GCC 4.6.1, vectors of type vector long long or vector
2531 long are passed and returned in the same method as other vectors
2532 with the VSX instruction set. Previously the GCC compiler did not
2533 adhere to the ABI for 128-bit vectors with 64-bit integer base
2534 types (PR 48857). This is also fixed in the GCC 4.5.4 release.
2536 S/390, zSeries and System z9/z10, IBM zEnterprise z196
2538 * Support for the zEnterprise z196 processor has been added. When
2539 using the -march=z196 option, the compiler will generate code
2540 making use of the following instruction facilities:
2541 + Conditional load/store
2543 + Floating-point-extension
2544 + Interlocked-access
2546 The -mtune=z196 option avoids the compare and branch instructions
2547 as well as the load address instruction with an index register as
2548 much as possible and performs instruction scheduling appropriate
2549 for the new out-of-order pipeline architecture.
2550 * When using the -m31 -mzarch options the generated code still
2551 conforms to the 32-bit ABI but uses the general purpose registers
2552 as 64-bit registers internally. This requires a Linux kernel saving
2553 the whole 64-bit registers when doing a context switch. Kernels
2554 providing that feature indicate that by the 'highgprs' string in
2556 * The SSA loop prefetching pass is enabled when using -O3.
2560 * GCC now supports the LEON series of SPARC V8 processors. The code
2561 generated by the compiler can either be tuned to it by means of the
2562 --with-tune=leon configure option and -mtune=leon compilation
2563 option, or the compiler can be built for the sparc-leon-{elf,linux}
2564 and sparc-leon3-{elf,linux} targets directly.
2565 * GCC has stopped sign/zero-extending parameter registers in the
2566 callee for functions taking parameters with sub-word size in 32-bit
2567 mode, since this is redundant with the specification of the ABI.
2568 GCC has never done so in 64-bit mode since this is also redundant.
2569 * The command line option -mfix-at697f has been added to enable the
2570 documented workaround for the single erratum of the Atmel AT697F
2577 * GCC now supports the Bionic C library and provides a convenient way
2578 of building native libraries and applications for the Android
2579 platform. Refer to the documentation of the -mandroid and -mbionic
2580 options for details on building native code. At the moment, Android
2581 support is enabled only for ARM.
2586 + Initial support for CFString types has been added.
2587 This allows GCC to build projects including the system Core
2588 Foundation frameworks. The GCC Objective-C family supports
2589 CFString "toll-free bridged" as per the Mac OS X system tools.
2590 CFString is also recognized in the context of format
2591 attributes and arguments (see the documentation for format
2592 attributes for limitations). At present, 8-bit character types
2594 + Object file size reduction.
2595 The Darwin zeroed memory allocators have been re-written to
2596 make more use of .zerofill sections. For non-debug code, this
2597 can reduce object file size significantly.
2598 + Objective-C family 64-bit support (NeXT ABI 2).
2599 Initial support has been added to support 64-bit Objective-C
2600 code using the Darwin/OS X native (NeXT) runtime. ABI version
2601 2 will be selected automatically when 64-bit code is built.
2602 + Objective-C family 32-bit ABI 1.
2603 For 32-bit code ABI 1 is also now also allowed. At present it
2604 must be selected manually using -fobjc-abi-version=1 where
2605 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6).
2607 + The -mdynamic-no-pic option has been enabled.
2608 Code supporting -mdynamic-no-pic optimization has been added
2609 and is applicable to -m32 builds. The compiler bootstrap uses
2610 the option where appropriate.
2611 + The default value for -mtune= has been changed.
2612 Since Darwin systems are primarily Xeon, Core-2 or similar the
2613 default tuning has been changed to -mtune=core2.
2614 + Enable 128-bit long double (__float128) support on Darwin.
2617 Several significant bugs have been fixed, such that GCC now
2618 produces code compatible with the Darwin64 PowerPC ABI.
2619 + libffi and boehm-gc.
2620 The Darwin ports of the libffi and boehm-gc libraries have
2621 been upgraded to include a Darwin64 implementation. This means
2622 that powerpc*-*-darwin9 platforms may now, for example, build
2623 Java applications with -m64 enabled.
2624 + Plug-in support has been enabled.
2625 + The -fsection-anchors option is now available although,
2626 presently, not heavily tested.
2632 * Support symbol versioning with the Sun linker.
2633 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+.
2634 * Support thread-local storage (TLS) with the Sun assembler on
2636 * Support TLS on Solaris 8/9 if prerequisites are met.
2637 * Support COMDAT group with the GNU assembler and recent Sun linker.
2638 * Support the Sun assembler visibility syntax.
2639 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp.
2640 -march=pentiumpro (Solaris 8/9).
2641 * Don't use SSE on Solaris 8/9 x86 by default.
2642 * Enable 128-bit long double (__float128) support on Solaris 2/x86.
2646 * Change the ABI for returning 8-byte vectors like __m64 in MMX
2647 registers on Solaris 10+/x86 to match the Sun Studio 12.1+
2648 compilers. This is an incompatible change. If you use such types,
2649 you must either recompile all your code with the new compiler or
2650 use the new -mvect8-ret-in-mem option to remain compatible with
2651 previous versions of GCC and Sun Studio.
2655 * Initial support for decimal floating point.
2656 * Support for the __thiscall calling-convention.
2657 * Support for hot-patchable function prologues via the
2658 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86.
2659 * Improvements of stack-probing and stack-allocation mechanisms.
2660 * Support of push/pop-macro pragma as preprocessor command.
2661 With #pragma push_macro("macro-name") the current definition of
2662 macro-name is saved and can be restored with #pragma
2663 pop_macro("macro-name") to its saved definition.
2664 * Enable 128-bit long double (__float128) support on MinGW and
2667 Other significant improvements
2669 Installation changes
2671 * An install-strip make target is provided that installs stripped
2672 executables, and may install libraries with unneeded or debugging
2674 * On Power7 systems, there is a potential problem if you build the
2675 GCC compiler with a host compiler using options that enable the VSX
2676 instruction set generation. If the host compiler has been patched
2677 so that the vec_ld and vec_st builtin functions generate Altivec
2678 memory instructions instead of VSX memory instructions, then you
2679 should be able to build the compiler with VSX instruction
2682 Changes for GCC Developers
2684 Note: these changes concern developers that develop GCC itself or
2685 software that integrates with GCC, such as plugins, and not the general
2687 * The gengtype utility, which previously was internal to the GCC
2688 build process, has been enchanced to provide GC root information
2689 for plugins as necessary.
2690 * The old GC allocation interface of ggc_alloc and friends was
2691 replaced with a type-safe alternative.
2695 This is the [20]list of problem reports (PRs) from GCC's bug tracking
2696 system that are known to be fixed in the 4.6.1 release. This list might
2697 not be complete (that is, it is possible that some PRs that have been
2698 fixed are not listed here).
2702 This is the [21]list of problem reports (PRs) from GCC's bug tracking
2703 system that are known to be fixed in the 4.6.2 release. This list might
2704 not be complete (that is, it is possible that some PRs that have been
2705 fixed are not listed here).
2709 This is the [22]list of problem reports (PRs) from GCC's bug tracking
2710 system that are known to be fixed in the 4.6.3 release. This list might
2711 not be complete (that is, it is possible that some PRs that have been
2712 fixed are not listed here).
2716 This is the [23]list of problem reports (PRs) from GCC's bug tracking
2717 system that are known to be fixed in the 4.6.4 release. This list might
2718 not be complete (that is, it is possible that some PRs that have been
2719 fixed are not listed here).
2722 For questions related to the use of GCC, please consult these web
2723 pages and the [24]GCC manuals. If that fails, the
2724 [25]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2725 web pages and the development of GCC are welcome on our developer
2726 list at [26]gcc@gcc.gnu.org. All of [27]our lists have public
2729 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and
2730 distribution of this entire article is permitted in any medium,
2731 provided this notice is preserved.
2733 These pages are [29]maintained by the GCC team. Last modified
2738 1. http://sourceware.org/bugzilla/show_bug.cgi?id=10401
2739 2. http://gcc.gnu.org/gcc-4.5/changes.html
2740 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted
2741 4. http://gcc.gnu.org/gcc-4.6/porting_to.html
2742 5. http://gcc.gnu.org/projects/lto/whopr.pdf
2743 6. http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
2744 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html
2745 8. http://gcc.gnu.org/PR43145
2746 9. http://gcc.gnu.org/PR43680
2747 10. http://gcc.gnu.org/PR33558
2748 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253
2749 12. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.200x
2750 13. http://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
2751 14. http://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races
2752 15. http://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
2753 16. http://gcc.gnu.org/wiki/OOP
2754 17. http://gcc.gnu.org/wiki/Coarray
2755 18. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233
2756 19. http://golang.org/
2757 20. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1
2758 21. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2
2759 22. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3
2760 23. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4
2761 24. http://gcc.gnu.org/onlinedocs/
2762 25. mailto:gcc-help@gcc.gnu.org
2763 26. mailto:gcc@gcc.gnu.org
2764 27. http://gcc.gnu.org/lists.html
2765 28. http://www.fsf.org/
2766 29. http://gcc.gnu.org/about.html
2767 30. http://validator.w3.org/check/referer
2768 ======================================================================
2769 http://gcc.gnu.org/gcc-4.5/index.html
2770 GCC 4.5 Release Series
2774 The [1]GNU project and the GCC developers are pleased to announce the
2775 release of GCC 4.5.4.
2777 This release is a bug-fix release, containing fixes for regressions in
2778 GCC 4.5.3 relative to previous releases of GCC.
2783 Jul 2, 2012 ([2]changes)
2786 Apr 28, 2011 ([3]changes)
2789 Dec 16, 2010 ([4]changes)
2792 Jul 31, 2010 ([5]changes)
2795 April 14, 2010 ([6]changes)
2797 References and Acknowledgements
2799 GCC used to stand for the GNU C Compiler, but since the compiler
2800 supports several other languages aside from C, it now stands for the
2801 GNU Compiler Collection.
2803 A list of [7]successful builds is updated as new information becomes
2806 The GCC developers would like to thank the numerous people that have
2807 contributed new features, improvements, bug fixes, and other changes as
2808 well as test results to GCC. This [8]amazing group of volunteers is
2809 what makes GCC successful.
2811 For additional information about GCC please refer to the [9]GCC project
2812 web site or contact the [10]GCC development mailing list.
2814 To obtain GCC please use [11]our mirror sites or [12]our SVN server.
2817 For questions related to the use of GCC, please consult these web
2818 pages and the [13]GCC manuals. If that fails, the
2819 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2820 web pages and the development of GCC are welcome on our developer
2821 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
2824 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
2825 distribution of this entire article is permitted in any medium,
2826 provided this notice is preserved.
2828 These pages are [18]maintained by the GCC team. Last modified
2833 1. http://www.gnu.org/
2834 2. http://gcc.gnu.org/gcc-4.5/changes.html
2835 3. http://gcc.gnu.org/gcc-4.5/changes.html
2836 4. http://gcc.gnu.org/gcc-4.5/changes.html
2837 5. http://gcc.gnu.org/gcc-4.5/changes.html
2838 6. http://gcc.gnu.org/gcc-4.5/changes.html
2839 7. http://gcc.gnu.org/gcc-4.5/buildstat.html
2840 8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
2841 9. http://gcc.gnu.org/index.html
2842 10. mailto:gcc@gcc.gnu.org
2843 11. http://gcc.gnu.org/mirrors.html
2844 12. http://gcc.gnu.org/svn.html
2845 13. http://gcc.gnu.org/onlinedocs/
2846 14. mailto:gcc-help@gcc.gnu.org
2847 15. mailto:gcc@gcc.gnu.org
2848 16. http://gcc.gnu.org/lists.html
2849 17. http://www.fsf.org/
2850 18. http://gcc.gnu.org/about.html
2851 19. http://validator.w3.org/check/referer
2852 ======================================================================
2853 http://gcc.gnu.org/gcc-4.5/changes.html
2854 GCC 4.5 Release Series
2855 Changes, New Features, and Fixes
2859 * GCC now requires the [1]MPC library in order to build. See the
2860 [2]prerequisites page for version requirements.
2861 * Support for a number of older systems and recently unmaintained or
2862 untested target ports of GCC has been declared obsolete in GCC 4.5.
2863 Unless there is activity to revive them, the next release of GCC
2864 will have their sources permanently removed.
2865 The following ports for individual systems on particular
2866 architectures have been obsoleted:
2867 + IRIX releases before 6.5 (mips-sgi-irix5*,
2868 mips-sgi-irix6.[0-4])
2869 + Solaris 7 (*-*-solaris2.7)
2870 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*,
2872 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions
2873 can be found in the [3]announcement.
2874 Support for the classic POWER architecture implemented in the
2875 original RIOS and RIOS2 processors of the old IBM RS/6000 product
2876 line has been obsoleted in the rs6000 port. This does not affect
2877 the new generation Power and PowerPC architectures.
2878 * Support has been removed for all the [4]configurations obsoleted in
2880 * Support has been removed for the protoize and unprotoize utilities,
2881 obsoleted in GCC 4.4.
2882 * Support has been removed for tuning for Itanium1 (Merced) variants.
2883 Note that code tuned for Itanium2 should also run correctly on
2885 * GCC now generates unwind info also for epilogues. DWARF debuginfo
2886 generated by GCC now uses more features of DWARF3 than before, and
2887 also some DWARF4 features. GDB older than 7.0 is not able to handle
2888 either of these, so to debug GCC 4.5 generated binaries or
2889 libraries GDB 7.0 or later is needed. You can disable use of DWARF4
2890 features with the -gdwarf-3 -gstrict-dwarf options, or use
2891 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but
2892 epilogue unwind info is emitted unconditionally whenever unwind
2894 * On x86 targets, code containing floating-point calculations may run
2895 significantly slower when compiled with GCC 4.5 in strict C99
2896 conformance mode than they did with earlier GCC versions. This is
2897 due to stricter standard conformance of the compiler and can be
2898 avoided by using the option -fexcess-precision=fast; also see
2900 * The function attribute noinline no longer prevents GCC from cloning
2901 the function. A new attribute noclone has been introduced for this
2902 purpose. Cloning a function means that it is duplicated and the new
2903 copy is specialized for certain contexts (for example when a
2904 parameter is a known constant).
2906 General Optimizer Improvements
2908 * The -save-temps now takes an optional argument. The -save-temps and
2909 -save-temps=cwd switches write the temporary files in the current
2910 working directory based on the original source file. The
2911 -save-temps=obj switch will write files into the directory
2912 specified with the -o option, and the intermediate filenames are
2913 based on the output file. This will allow the user to get the
2914 compiler intermediate files when doing parallel builds without two
2915 builds of the same filename located in different directories from
2916 interfering with each other.
2917 * Debugging dumps are now created in the same directory as the object
2918 file rather than in the current working directory. This allows the
2919 user to get debugging dumps when doing parallel builds without two
2920 builds of the same filename interfering with each other.
2921 * GCC has been integrated with the [6]MPC library. This allows GCC to
2922 evaluate complex arithmetic at compile time [7]more accurately. It
2923 also allows GCC to evaluate calls to complex built-in math
2924 functions having constant arguments and replace them at compile
2925 time with their mathematically equivalent results. In doing so, GCC
2926 can generate correct results regardless of the math library
2927 implementation or floating point precision of the host platform.
2928 This also allows GCC to generate identical results regardless of
2929 whether one compiles in native or cross-compile configurations to a
2930 particular target. The following built-in functions take advantage
2931 of this new capability: cacos, cacosh, casin, casinh, catan,
2932 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan,
2933 and ctanh. The float and long double variants of these functions
2934 (e.g. csinf and csinl) are also handled.
2935 * A new link-time optimizer has been added ([8]-flto). When this
2936 option is used, GCC generates a bytecode representation of each
2937 input file and writes it to specially-named sections in each object
2938 file. When the object files are linked together, all the function
2939 bodies are read from these named sections and instantiated as if
2940 they had been part of the same translation unit. This enables
2941 interprocedural optimizations to work across different files (and
2942 even different languages), potentially improving the performance of
2943 the generated code. To use the link-timer optimizer, -flto needs to
2944 be specified at compile time and during the final link. If the
2945 program does not require any symbols to be exported, it is possible
2946 to combine -flto and the experimental [9]-fwhopr with
2947 [10]-fwhole-program to allow the interprocedural optimizers to use
2948 more aggressive assumptions.
2949 * The automatic parallelization pass was enhanced to support
2950 parallelization of outer loops.
2951 * Automatic parallelization can be enabled as part of Graphite. In
2952 addition to -ftree-parallelize-loops=, specify
2953 -floop-parallelize-all to enable the Graphite-based optimization.
2954 * The infrastructure for optimizing based on [11]restrict qualified
2955 pointers has been rewritten and should result in code generation
2956 improvements. Optimizations based on restrict qualified pointers
2957 are now also available when using -fno-strict-aliasing.
2958 * There is a new optimization pass that attempts to change prototype
2959 of functions to avoid unused parameters, pass only relevant parts
2960 of structures and turn arguments passed by reference to arguments
2961 passed by value when possible. It is enabled by -O2 and above as
2962 well as -Os and can be manually invoked using the new command-line
2964 * GCC now optimize exception handling code. In particular cleanup
2965 regions that are proved to not have any effect are optimized out.
2967 New Languages and Language specific improvements
2971 * The -fshow-column option is now on by default. This means error
2972 messages now have a column associated with them.
2976 * Compilation of programs heavily using discriminated record types
2977 with variant parts has been sped up and generates more compact
2979 * Stack checking now works reasonably well on most plaforms. In some
2980 specific cases, stack overflows may still fail to be detected, but
2981 a compile-time warning will be issued for these cases.
2985 * If a header named in a #include directive is not found, the
2986 compiler exits immediately. This avoids a cascade of errors arising
2987 from declarations expected to be found in that header being
2989 * A new built-in function __builtin_unreachable() has been added that
2990 tells the compiler that control will never reach that point. It may
2991 be used after asm statements that terminate by transferring control
2992 elsewhere, and in other places that are known to be unreachable.
2993 * The -Wlogical-op option now warns for logical expressions such as
2994 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be
2995 mistakes. This option is disabled by default.
2996 * An asm goto feature has been added to allow asm statements that
2998 * C++0x raw strings are supported for C++ and for C with -std=gnu99.
2999 * The deprecated attribute now takes an optional string argument, for
3000 example, __attribute__((deprecated("text string"))), that will be
3001 printed together with the deprecation warning.
3005 * The -Wenum-compare option, which warns when comparing values of
3006 different enum types, now works for C. It formerly only worked for
3007 C++. This warning is enabled by -Wall. It may be avoided by using a
3009 * The -Wcast-qual option now warns about casts which are unsafe in
3010 that they permit const-correctness to be violated without further
3011 warnings. Specifically, it warns about cases where a qualifier is
3012 added when all the lower types are not const. For example, it warns
3013 about a cast from char ** to const char **.
3014 * The -Wc++-compat option is significantly improved. It issues new
3016 + Using C++ reserved operator names as identifiers.
3017 + Conversions to enum types without explicit casts.
3018 + Using va_arg with an enum type.
3019 + Using different enum types in the two branches of ?:.
3020 + Using ++ or -- on a variable of enum type.
3021 + Using the same name as both a struct, union or enum tag and a
3022 typedef, unless the typedef refers to the tagged type itself.
3023 + Using a struct, union, or enum which is defined within another
3025 + A struct field defined using a typedef if there is a field in
3026 the struct, or an enclosing struct, whose name is the typedef
3028 + Duplicate definitions at file scope.
3029 + Uninitialized const variables.
3030 + A global variable with an anonymous struct, union, or enum
3032 + Using a string constant to initialize a char array whose size
3033 is the length of the string.
3034 * The new -Wjump-misses-init option warns about cases where a goto or
3035 switch skips the initialization of a variable. This sort of branch
3036 is an error in C++ but not in C. This warning is enabled by
3038 * GCC now ensures that a C99-conforming <stdint.h> is present on most
3039 targets, and uses information about the types in this header to
3040 implement the Fortran bindings to those types. GCC does not ensure
3041 the presence of such a header, and does not implement the Fortran
3042 bindings, on the following targets: NetBSD, VxWorks, VMS,
3043 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF.
3044 * GCC now implements C90- and C99-conforming rules for constant
3045 expressions. This may cause warnings or errors for some code using
3046 expressions that can be folded to a constant but are not constant
3047 expressions as defined by ISO C.
3048 * All known target-independent C90 and C90 Amendment 1 conformance
3049 bugs, and all known target-independent C99 conformance bugs not
3050 related to floating point or extended identifiers, have been fixed.
3051 * The C decimal floating point support now includes support for the
3052 FLOAT_CONST_DECIMAL64 pragma.
3053 * The named address space feature from ISO/IEC TR 18037 is now
3054 supported. This is currently only implemented for the SPU
3059 * Improved [12]experimental support for the upcoming C++0x ISO C++
3060 standard, including support for raw strings, lambda expressions and
3061 explicit type conversion operators.
3062 * When printing the name of a class template specialization, G++ will
3063 now omit any template arguments which come from default template
3064 arguments. This behavior (and the pretty-printing of function
3065 template specializations as template signature and arguments) can
3066 be disabled with the -fno-pretty-templates option.
3067 * Access control is now applied to typedef names used in a template,
3068 which may cause G++ to reject some ill-formed code that was
3069 accepted by earlier releases. The -fno-access-control option can be
3070 used as a temporary workaround until the code is corrected.
3071 * Compilation time for code that uses templates should now scale
3072 linearly with the number of instantiations rather than
3073 quadratically, as template instantiations are now looked up using
3075 * Declarations of functions that look like builtin declarations of
3076 library functions are only considered to be redeclarations if they
3077 are declared with extern "C". This may cause problems with code
3078 that omits extern "C" on hand-written declarations of C library
3079 functions such as abort or memcpy. Such code is ill-formed, but was
3080 accepted by earlier releases.
3081 * Diagnostics that used to complain about passing non-POD types to
3082 ... or jumping past the declaration of a non-POD variable now check
3083 for triviality rather than PODness, as per C++0x.
3084 * In C++0x mode local and anonymous classes are now allowed as
3085 template arguments, and in declarations of variables and functions
3086 with linkage, so long as any such declaration that is used is also
3087 defined ([13]DR 757).
3088 * Labels may now have attributes, as has been permitted for a while
3089 in C. This is only permitted when the label definition and the
3090 attribute specifier is followed by a semicolon--i.e., the label
3091 applies to an empty statement. The only useful attribute for a
3093 * G++ now implements [14]DR 176. Previously G++ did not support using
3094 the injected-class-name of a template base class as a type name,
3095 and lookup of the name found the declaration of the template in the
3096 enclosing scope. Now lookup of the name finds the
3097 injected-class-name, which can be used either as a type or as a
3098 template, depending on whether or not the name is followed by a
3099 template argument list. As a result of this change, some code that
3100 was previously accepted may be ill-formed because
3101 1. The injected-class-name is not accessible because it's from a
3103 2. The injected-class-name cannot be used as an argument for a
3104 template template parameter.
3105 In either of these cases, the code can be fixed by adding a
3106 nested-name-specifier to explicitly name the template. The first
3107 can be worked around with -fno-access-control; the second is only
3108 rejected with -pedantic.
3109 * A new standard mangling for SIMD vector types has been added, to
3110 avoid name clashes on systems with vectors of varying length. By
3111 default the compiler still uses the old mangling, but emits aliases
3112 with the new mangling on targets that support strong aliases. Users
3113 can switch over entirely to the new mangling with -fabi-version=4
3114 or -fabi-version=0. -Wabi will now warn about code that uses the
3116 * The command-line option -ftemplate-depth-N is now written as
3117 -ftemplate-depth=N and the old form is deprecated.
3118 * Conversions between NULL and non-pointer types are now warned by
3119 default. The new option -Wno-conversion-null disables these
3120 warnings. Previously these warnings were only available when using
3121 -Wconversion explicitly.
3123 Runtime Library (libstdc++)
3125 * [15]Improved experimental support for the upcoming ISO C++
3126 standard, C++0x, including:
3127 + Support for <future>, <functional>, and <random>.
3128 + Existing facilities now exploit explicit operators and the
3129 newly implemented core C++0x features.
3130 * An experimental [16]profile mode has been added. This is an
3131 implementation of many C++ standard library constructs with an
3132 additional analysis layer that gives performance improvement advice
3133 based on recognition of suboptimal usage patterns. For example,
3138 for (int k = 0; k < 1024; ++k)
3139 v.insert(v.begin(), k);
3142 When instrumented via the profile mode, can return suggestions
3143 about the initial size and choice of the container used as follows:
3144 vector-to-list: improvement = 5: call stack = 0x804842c ...
3145 : advice = change std::vector to std::list
3146 vector-size: improvement = 3: call stack = 0x804842c ...
3147 : advice = change initial container size from 0 to 1024
3149 These constructs can be substituted for the normal libstdc++
3150 constructs on a piecemeal basis, or all existing components can be
3151 transformed via the -D_GLIBCXX_PROFILE macro.
3152 * [17]Support for decimal floating-point arithmetic (aka ISO C++ TR
3153 24733) has been added. This support is in header file
3154 <decimal/decimal>, uses namespace std::decimal, and includes
3155 classes decimal32, decimal64, and decimal128.
3156 * Sources have been audited for application of function attributes
3157 nothrow, const, pure, and noreturn.
3158 * Python pretty-printers have been added for many standard library
3159 components that simplify the internal representation and present a
3160 more intuitive view of components when used with
3161 appropriately-advanced versions of GDB. For more information,
3162 please consult the more [18]detailed description.
3163 * The default behavior for comparing typeinfo names has changed, so
3164 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero.
3165 * The new -static-libstdc++ option directs g++ to link the C++
3166 library statically, even if the default would normally be to link
3171 * The COMMON default padding has been changed - instead of adding the
3172 padding before a variable it is now added afterwards, which
3173 increases the compatibility with other vendors and helps to obtain
3174 the correct output in some cases. Cf. also the -falign-commons
3175 option ([19]added in 4.4).
3176 * The -finit-real= option now also supports the value snan for
3177 signalling not-a-number; to be effective, one additionally needs to
3178 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time
3179 optimizations can turn a signalling NaN into a quiet one.
3180 * The new option -fcheck= has been added with the options bounds,
3181 array-temps, do, pointer, and recursive. The bounds and array-temps
3182 options are equivalent to -fbounds-check and
3183 -fcheck-array-temporaries. The do option checks for invalid
3184 modification of loop iteration variables, and the recursive option
3185 tests for recursive calls to subroutines/functions which are not
3186 marked as recursive. With pointer pointer association checks in
3187 calls are performed; however, neither undefined pointers nor
3188 pointers in expressions are handled. Using -fcheck=all enables all
3189 these run-time checks.
3190 * The run-time checking -fcheck=bounds now warns about invalid string
3191 lengths of character dummy arguments. Additionally, more
3192 compile-time checks have been added.
3193 * The new option [20]-fno-protect-parens has been added; if set, the
3194 compiler may reorder REAL and COMPLEX expressions without regard to
3196 * GNU Fortran no longer links against libgfortranbegin. As before,
3197 MAIN__ (assembler symbol name) is the actual Fortran main program,
3198 which is invoked by the main function. However, main is now
3199 generated and put in the same object file as MAIN__. For the time
3200 being, libgfortranbegin still exists for backward compatibility.
3201 For details see the new [21]Mixed-Language Programming chapter in
3203 * The I/O library was restructured for performance and cleaner code.
3204 * Array assignments and WHERE are now run in parallel when OpenMP's
3206 * The experimental option -fwhole-file was added. The option allows
3207 whole-file checking of procedure arguments and allows for better
3208 optimizations. It can also be used with -fwhole-program, which is
3209 now also supported in gfortran.
3210 * More Fortran 2003 and Fortran 2008 mathematical functions can now
3211 be used as initialization expressions.
3212 * Some extended attributes such as STDCALL are now supported via the
3213 [22]GCC$ compiler directive.
3214 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN
3215 intrinsic behaves now as if zero were always positive.
3216 * For legacy compatibiliy: On Cygwin and MinGW, the special files
3217 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now
3219 * Fortran 2003 support has been extended:
3220 + Procedure-pointer function results and procedure-pointer
3221 components (including PASS),
3222 + allocatable scalars (experimental),
3223 + DEFERRED type-bound procedures,
3224 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements
3225 have been implemented.
3226 + The ALLOCATE statement supports type-specs and the SOURCE=
3228 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC
3229 type-bound procedure (i.e. as type-bound operators).
3230 + Rounding (ROUND=, RZ, ...) for output is now supported.
3231 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the
3232 intrinsic module ISO_C_BINDING are now supported, except for
3233 the targets listed above as ones where GCC does not have
3234 <stdint.h> type information.
3235 + Extensible derived types with type-bound procedure or
3236 procedure pointer with PASS attribute now have to use CLASS in
3237 line with the Fortran 2003 standard; the workaround to use
3238 TYPE is no longer supported.
3239 + [23]Experimental, incomplete support for polymorphism,
3240 including CLASS, SELECT TYPE and dynamic dispatch of
3241 type-bound procedure calls. Some features do not work yet such
3242 as unlimited polymorphism (CLASS(*)).
3243 * Fortran 2008 support has been extended:
3244 + The OPEN statement now supports the NEWUNIT= option, which
3245 returns a unique file unit, thus preventing inadvertent use of
3246 the same unit in different parts of the program.
3247 + Support for unlimited format items has been added.
3248 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of
3249 the intrinsic module ISO_FORTRAN_ENV are now supported.
3250 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN,
3251 ACOS, and ATAN is now possible; the functions ASINH, ACOSH,
3252 and ATANH have been added (for real and complex arguments) and
3253 ATAN(Y,X) is now an alias for ATAN2(Y,X).
3254 + The BLOCK construct has been implemented.
3256 New Targets and Target Specific Improvements
3260 * Full cross-toolchain support now available with GNU Binutils
3264 * GCC now supports the Cortex-M0 and Cortex-A5 processors.
3265 * GCC now supports the ARM v7E-M architecture.
3266 * GCC now supports VFPv4-based FPUs and FPUs with
3267 single-precision-only VFP.
3268 * GCC has many improvements to optimization for other ARM processors,
3269 including scheduling support for the integer pipeline on Cortex-A9.
3270 * GCC now supports the IEEE 754-2008 half-precision floating-point
3271 type, and a variant ARM-specific half-precision type. This type is
3272 specified using __fp16, with the layout determined by
3273 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and
3274 VFPv4 half-precision instructions will be used.
3275 * GCC now supports the variant of AAPCS that uses VFP registers for
3276 parameter passing and return values.
3280 * The -mno-tablejump option has been removed because it has the same
3281 effect as the -fno-jump-tables option.
3282 * Added support for these new AVR devices:
3289 * GCC now will set the default for -march= based on the configure
3291 * GCC now supports handling floating-point excess precision arising
3292 from use of the x87 floating-point unit in a way that conforms to
3293 ISO C99. This is enabled with -fexcess-precision=standard and with
3294 standards conformance options such as -std=c99, and may be disabled
3295 using -fexcess-precision=fast.
3296 * Support for the Intel Atom processor is now available through the
3297 -march=atom and -mtune=atom options.
3298 * A new -mcrc32 option is now available to enable crc32 intrinsics.
3299 * A new -mmovbe option is now available to enable GCC to use the
3300 movbe instruction to implement __builtin_bswap32 and
3302 * SSE math now can be enabled by default at configure time with the
3303 new --with-fpmath=sse option.
3304 * There is a new intrinsic header file, <x86intrin.h>. It should be
3305 included before using any IA-32/x86-64 intrinsics.
3306 * Support for the XOP, FMA4, and LWP instruction sets for the AMD
3307 Orochi processors are now available with the -mxop, -mfma4, and
3309 * The -mabm option enables GCC to use the popcnt and lzcnt
3310 instructions on AMD processors.
3311 * The -mpopcnt option enables GCC to use the popcnt instructions on
3312 both AMD and Intel processors.
3316 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x
3318 * GCC now supports thread-local storage (TLS) on M68K and ColdFire
3323 Support has been added for the Toshiba Media embedded Processor (MeP,
3324 or mep-elf) embedded target.
3328 * GCC now supports MIPS 1004K processors.
3329 * GCC can now be configured with options --with-arch-32,
3330 --with-arch-64, --with-tune-32 and --with-tune-64 to control the
3331 default optimization separately for 32-bit and 64-bit modes.
3332 * MIPS targets now support an alternative _mcount interface, in which
3333 register $12 points to the function's save slot for register $31.
3334 This interface is selected by the -mcount-ra-address option; see
3335 the documentation for more details.
3336 * GNU/Linux targets can now generate read-only .eh_frame sections.
3337 This optimization requires GNU binutils 2.20 or above, and is only
3338 available if GCC is configured with a suitable version of binutils.
3339 * GNU/Linux targets can now attach special relocations to indirect
3340 calls, so that the linker can turn them into direct jumps or
3341 branches. This optimization requires GNU binutils 2.20 or later,
3342 and is automatically selected if GCC is configured with an
3343 appropriate version of binutils. It can be explicitly enabled or
3344 disabled using the -mrelax-pic-calls command-line option.
3345 * GCC now generates more heavily-optimized atomic operations on
3347 * MIPS targets now support the -fstack-protector option.
3348 * GCC now supports an -msynci option, which specifies that synci is
3349 enough to flush the instruction cache, without help from the
3350 operating system. GCC uses this information to optimize
3351 automatically-generated cache flush operations, such as those used
3352 for nested functions in C. There is also a --with-synci
3353 configure-time option, which makes -msynci the default.
3354 * GCC supports four new function attributes for interrupt handlers:
3355 interrupt, use_shadow_register_set, keep_interrupts_masked and
3356 use_debug_exception_return. See the documentation for more details
3357 about these attributes.
3359 RS/6000 (POWER/PowerPC)
3361 * GCC now supports the Power ISA 2.06, which includes the VSX
3362 instructions that add vector 64-bit floating point support, new
3363 population count instructions, and conversions between floating
3364 point and unsigned types.
3365 * Support for the power7 processor is now available through the
3366 -mcpu=power7 and -mtune=power7.
3367 * GCC will now vectorize loops that contain simple math functions
3368 like copysign when generating code for altivec or VSX targets.
3369 * Support for the A2 processor is now available through the -mcpu=a2
3370 and -mtune=a2 options.
3371 * Support for the 476 processor is now available through the
3372 -mcpu={476,476fp} and -mtune={476,476fp} options.
3373 * Support for the e500mc64 processor is now available through the
3374 -mcpu=e500mc64 and -mtune=e500mc64 options.
3375 * GCC can now be configured with options --with-cpu-32,
3376 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the
3377 default optimization separately for 32-bit and 64-bit modes.
3378 * Starting with GCC 4.5.4, vectors of type vector long long or vector
3379 long are passed and returned in the same method as other vectors
3380 with the VSX instruction set. Previously the GCC compiler did not
3381 adhere to the ABI for 128-bit vectors with 64-bit integer base
3382 types (PR 48857). This is also fixed in the GCC 4.6.1 release.
3386 Support has been added for the Renesas RX Processor (rx-elf) target.
3390 Windows (Cygwin and MinGW)
3392 * GCC now installs all the major language runtime libraries as DLLs
3393 when configured with the --enable-shared option.
3394 * GCC now makes use of the new support for aligned common variables
3395 in versions of binutils >= 2.20 to fix bugs in the support for SSE
3397 * Improvements to the libffi support library increase the reliability
3398 of code generated by GCJ on all Windows platforms. Libgcj is
3399 enabled by default for the first time.
3400 * Libtool improvements simplify installation by placing the generated
3401 DLLs in the correct binaries directory.
3402 * Numerous other minor bugfixes and improvements, and substantial
3403 enhancements to the Fortran language support library.
3407 Other significant improvements
3411 * It is now possible to extend the compiler without having to modify
3412 its source code. A new option -fplugin=file.so tells GCC to load
3413 the shared object file.so and execute it as part of the compiler.
3414 The internal documentation describes the details on how plugins can
3415 interact with the compiler.
3417 Installation changes
3419 * The move to newer autotools changed default installation
3420 directories and switches to control them: The --with-datarootdir,
3421 --with-docdir, --with-pdfdir, and --with-htmldir switches are not
3422 used any more. Instead, you can now use --datarootdir, --docdir,
3423 --htmldir, and --pdfdir. The default installation directories have
3424 changed as follows according to the GNU Coding Standards:
3426 datarootdir read-only architecture-independent data root [PREFIX/share]
3427 localedir locale-specific message catalogs [DATAROOTDIR/locale]
3428 docdir documentation root [DATAROOTDIR/doc/PACKAGE]
3429 htmldir html documentation [DOCDIR]
3430 dvidir dvi documentation [DOCDIR]
3431 pdfdir pdf documentation [DOCDIR]
3432 psdir ps documentation [DOCDIR]
3433 The following variables have new default values:
3435 datadir read-only architecture-independent data [DATAROOTDIR]
3436 infodir info documentation [DATAROOTDIR/info]
3437 mandir man documentation [DATAROOTDIR/man]
3441 This is the [24]list of problem reports (PRs) from GCC's bug tracking
3442 system that are known to be fixed in the 4.5.1 release. This list might
3443 not be complete (that is, it is possible that some PRs that have been
3444 fixed are not listed here).
3448 * GCC's new link-time optimizer ([25]-flto) now also works on a few
3450 + Cygwin (*-cygwin*)
3452 + Darwin on x86-64 (x86_64-apple-darwin*)
3453 LTO is not enabled by default for these targets. To enable LTO, you
3454 should configure with the --enable-lto option.
3458 This is the [26]list of problem reports (PRs) from GCC's bug tracking
3459 system that are known to be fixed in the 4.5.2 release. This list might
3460 not be complete (that is, it is possible that some PRs that have been
3461 fixed are not listed here).
3465 This is the [27]list of problem reports (PRs) from GCC's bug tracking
3466 system that are known to be fixed in the 4.5.3 release. This list might
3467 not be complete (that is, it is possible that some PRs that have been
3468 fixed are not listed here).
3470 On the PowerPC compiler, the Altivec builtin functions vec_ld and
3471 vec_st have been modified to generate the Altivec memory instructions
3472 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5
3473 release, these builtin functions were changed to generate VSX memory
3474 reference instructions instead of Altivec memory instructions, but
3475 there are differences between the two instructions. If the VSX
3476 instruction set is available, you can now use the new builtin functions
3477 vec_vsx_ld and vec_vsx_st which always generates the VSX memory
3482 This is the [28]list of problem reports (PRs) from GCC's bug tracking
3483 system that are known to be fixed in the 4.5.4 release. This list might
3484 not be complete (that is, it is possible that some PRs that have been
3485 fixed are not listed here).
3488 For questions related to the use of GCC, please consult these web
3489 pages and the [29]GCC manuals. If that fails, the
3490 [30]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3491 web pages and the development of GCC are welcome on our developer
3492 list at [31]gcc@gcc.gnu.org. All of [32]our lists have public
3495 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and
3496 distribution of this entire article is permitted in any medium,
3497 provided this notice is preserved.
3499 These pages are [34]maintained by the GCC team. Last modified
3504 1. http://www.multiprecision.org/
3505 2. http://gcc.gnu.org/install/prerequisites.html
3506 3. http://gcc.gnu.org/ml/gcc/2010-01/msg00510.html
3507 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted
3508 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86
3509 6. http://www.multiprecision.org/
3510 7. http://gcc.gnu.org/PR30789
3511 8. http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
3512 9. http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802
3513 10. http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800
3514 11. http://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html
3515 12. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html
3516 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757
3517 14. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176
3518 15. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.200x
3519 16. http://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html
3520 17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733
3521 18. http://sourceware.org/gdb/wiki/STLSupport
3522 19. http://gcc.gnu.org/gcc-4.4/changes.html
3523 20. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
3524 21. http://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html
3525 22. http://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html
3526 23. http://gcc.gnu.org/wiki/OOP
3527 24. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1
3528 25. http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
3529 26. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2
3530 27. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3
3531 28. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4
3532 29. http://gcc.gnu.org/onlinedocs/
3533 30. mailto:gcc-help@gcc.gnu.org
3534 31. mailto:gcc@gcc.gnu.org
3535 32. http://gcc.gnu.org/lists.html
3536 33. http://www.fsf.org/
3537 34. http://gcc.gnu.org/about.html
3538 35. http://validator.w3.org/check/referer
3539 ======================================================================
3540 http://gcc.gnu.org/gcc-4.4/index.html
3541 GCC 4.4 Release Series
3545 The [1]GNU project and the GCC developers are pleased to announce the
3546 release of GCC 4.4.7.
3548 This release is a bug-fix release, containing fixes for regressions in
3549 GCC 4.4.6 relative to previous releases of GCC.
3554 March 13, 2012 ([2]changes)
3557 April 16, 2011 ([3]changes)
3560 October 1, 2010 ([4]changes)
3563 April 29, 2010 ([5]changes)
3566 January 21, 2010 ([6]changes)
3569 October 15, 2009 ([7]changes)
3572 July 22, 2009 ([8]changes)
3575 April 21, 2009 ([9]changes)
3577 References and Acknowledgements
3579 GCC used to stand for the GNU C Compiler, but since the compiler
3580 supports several other languages aside from C, it now stands for the
3581 GNU Compiler Collection.
3583 A list of [10]successful builds is updated as new information becomes
3586 The GCC developers would like to thank the numerous people that have
3587 contributed new features, improvements, bug fixes, and other changes as
3588 well as test results to GCC. This [11]amazing group of volunteers is
3589 what makes GCC successful.
3591 For additional information about GCC please refer to the [12]GCC
3592 project web site or contact the [13]GCC development mailing list.
3594 To obtain GCC please use [14]our mirror sites or [15]our SVN server.
3597 For questions related to the use of GCC, please consult these web
3598 pages and the [16]GCC manuals. If that fails, the
3599 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3600 web pages and the development of GCC are welcome on our developer
3601 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
3604 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
3605 distribution of this entire article is permitted in any medium,
3606 provided this notice is preserved.
3608 These pages are [21]maintained by the GCC team. Last modified
3613 1. http://www.gnu.org/
3614 2. http://gcc.gnu.org/gcc-4.4/changes.html
3615 3. http://gcc.gnu.org/gcc-4.4/changes.html
3616 4. http://gcc.gnu.org/gcc-4.4/changes.html
3617 5. http://gcc.gnu.org/gcc-4.4/changes.html
3618 6. http://gcc.gnu.org/gcc-4.4/changes.html
3619 7. http://gcc.gnu.org/gcc-4.4/changes.html
3620 8. http://gcc.gnu.org/gcc-4.4/changes.html
3621 9. http://gcc.gnu.org/gcc-4.4/changes.html
3622 10. http://gcc.gnu.org/gcc-4.4/buildstat.html
3623 11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
3624 12. http://gcc.gnu.org/index.html
3625 13. mailto:gcc@gcc.gnu.org
3626 14. http://gcc.gnu.org/mirrors.html
3627 15. http://gcc.gnu.org/svn.html
3628 16. http://gcc.gnu.org/onlinedocs/
3629 17. mailto:gcc-help@gcc.gnu.org
3630 18. mailto:gcc@gcc.gnu.org
3631 19. http://gcc.gnu.org/lists.html
3632 20. http://www.fsf.org/
3633 21. http://gcc.gnu.org/about.html
3634 22. http://validator.w3.org/check/referer
3635 ======================================================================
3636 http://gcc.gnu.org/gcc-4.4/changes.html
3637 GCC 4.4 Release Series
3638 Changes, New Features, and Fixes
3640 The latest release in the 4.4 release series is [1]GCC 4.4.7.
3644 * __builtin_stdarg_start has been completely removed from GCC.
3645 Support for <varargs.h> had been deprecated since GCC 4.0. Use
3646 __builtin_va_start as a replacement.
3647 * Some of the errors issued by the C++ front end that could be
3648 downgraded to warnings in previous releases by using -fpermissive
3649 are now warnings by default. They can be converted into errors by
3650 using -pedantic-errors.
3651 * Use of the cpp assertion extension will now emit a warning when
3652 -Wdeprecated or -pedantic is used. This extension has been
3653 deprecated for many years, but never warned about.
3654 * Packed bit-fields of type char were not properly bit-packed on many
3655 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4
3656 causes an ABI change. For example there is no longer a 4-bit
3657 padding between field a and b in this structure:
3662 } __attribute__ ((packed));
3663 There is a new warning to help identify fields that are affected:
3664 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4
3665 The warning can be disabled with -Wno-packed-bitfield-compat.
3666 * On ARM EABI targets, the C++ mangling of the va_list type has been
3667 changed to conform to the current revision of the EABI. This does
3668 not affect the libstdc++ library included with GCC.
3669 * The SCOUNT and POS bits of the MIPS DSP control register are now
3670 treated as global. Previous versions of GCC treated these fields as
3671 call-clobbered instead.
3672 * The MIPS port no longer recognizes the h asm constraint. It was
3673 necessary to remove this constraint in order to avoid generating
3674 unpredictable code sequences.
3675 One of the main uses of the h constraint was to extract the high
3676 part of a multiplication on 64-bit targets. For example:
3677 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y));
3678 You can now achieve the same effect using 128-bit types:
3679 typedef unsigned int uint128_t __attribute__((mode(TI)));
3680 result = ((uint128_t) x * y) >> 64;
3681 The second sequence is better in many ways. For example, if x and y
3682 are constants, the compiler can perform the multiplication at
3683 compile time. If x and y are not constants, the compiler can
3684 schedule the runtime multiplication better than it can schedule an
3686 * Support for a number of older systems and recently unmaintained or
3687 untested target ports of GCC has been declared obsolete in GCC 4.4.
3688 Unless there is activity to revive them, the next release of GCC
3689 will have their sources permanently removed.
3690 The following ports for individual systems on particular
3691 architectures have been obsoleted:
3692 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*,
3694 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*,
3695 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*,
3696 sh-*-*). This does not affect other more specific targets
3697 using the COFF object format on those architectures, or the
3698 more specific H8300 and SH targets (h8300-*-rtems*,
3699 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*,
3700 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks).
3701 + 2BSD on PDP-11 (pdp11-*-bsd)
3702 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*,
3703 powerpc-ibm-aix4.[12]*)
3704 + Tuning support for Itanium1 (Merced) variants. Note that code
3705 tuned for Itanium2 should also run correctly on Itanium1.
3706 * The protoize and unprotoize utilities have been obsoleted and will
3707 be removed in GCC 4.5. These utilities have not been installed by
3708 default since GCC 3.0.
3709 * Support has been removed for all the [2]configurations obsoleted in
3711 * Unknown -Wno-* options are now silently ignored by GCC if no other
3712 diagnostics are issued. If other diagnostics are issued, then GCC
3713 warns about the unknown options.
3714 * More information on porting to GCC 4.4 from previous versions of
3715 GCC can be found in the [3]porting guide for this release.
3717 General Optimizer Improvements
3719 * A new command-line switch -findirect-inlining has been added. When
3720 turned on it allows the inliner to also inline indirect calls that
3721 are discovered to have known targets at compile time thanks to
3723 * A new command-line switch -ftree-switch-conversion has been added.
3724 This new pass turns simple initializations of scalar variables in
3725 switch statements into initializations from a static array, given
3726 that all the values are known at compile time and the ratio between
3727 the new array size and the original switch branches does not exceed
3728 the parameter --param switch-conversion-max-branch-ratio (default
3730 * A new command-line switch -ftree-builtin-call-dce has been added.
3731 This optimization eliminates unnecessary calls to certain builtin
3732 functions when the return value is not used, in cases where the
3733 calls can not be eliminated entirely because the function may set
3734 errno. This optimization is on by default at -O2 and above.
3735 * A new command-line switch -fconserve-stack directs the compiler to
3736 minimize stack usage even if it makes the generated code slower.
3737 This affects inlining decisions.
3738 * When the assembler supports it, the compiler will now emit unwind
3739 information using assembler .cfi directives. This makes it possible
3740 to use such directives in inline assembler code. The new option
3741 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi
3743 * The [4]Graphite branch has been merged. This merge has brought in a
3744 new framework for loop optimizations based on a polyhedral
3745 intermediate representation. These optimizations apply to all the
3746 languages supported by GCC. The following new code transformations
3747 are available in GCC 4.4:
3748 + -floop-interchange performs loop interchange transformations
3749 on loops. Interchanging two nested loops switches the inner
3750 and outer loops. For example, given a loop like:
3753 A(J, I) = A(J, I) * C
3757 loop interchange will transform the loop as if the user had
3761 A(J, I) = A(J, I) * C
3765 which can be beneficial when N is larger than the caches,
3766 because in Fortran, the elements of an array are stored in
3767 memory contiguously by column, and the original loop iterates
3768 over rows, potentially creating at each access a cache miss.
3769 + -floop-strip-mine performs loop strip mining transformations
3770 on loops. Strip mining splits a loop into two nested loops.
3771 The outer loop has strides equal to the strip size and the
3772 inner loop has strides of the original loop within a strip.
3773 For example, given a loop like:
3778 loop strip mining will transform the loop as if the user had
3781 DO I = II, min (II + 3, N)
3786 + -floop-block performs loop blocking transformations on loops.
3787 Blocking strip mines each loop in the loop nest such that the
3788 memory accesses of the element loops fit inside caches. For
3789 example, given a loop like:
3792 A(J, I) = B(I) + C(J)
3796 loop blocking will transform the loop as if the user had
3800 DO I = II, min (II + 63, N)
3801 DO J = JJ, min (JJ + 63, M)
3802 A(J, I) = B(I) + C(J)
3808 which can be beneficial when M is larger than the caches,
3809 because the innermost loop will iterate over a smaller amount
3810 of data that can be kept in the caches.
3811 * A new register allocator has replaced the old one. It is called
3812 integrated register allocator (IRA) because coalescing, register
3813 live range splitting, and hard register preferencing are done
3814 on-the-fly during coloring. It also has better integration with the
3815 reload pass. IRA is a regional register allocator which uses modern
3816 Chaitin-Briggs coloring instead of Chow's priority coloring used in
3817 the old register allocator. More info about IRA internals and
3818 options can be found in the GCC manuals.
3819 * A new instruction scheduler and software pipeliner, based on the
3820 selective scheduling approach, has been added. The new pass
3821 performs instruction unification, register renaming, substitution
3822 through register copies, and speculation during scheduling. The
3823 software pipeliner is able to pipeline non-countable loops. The new
3824 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4
3825 it is available for the Intel Itanium platform working by default
3826 as the second scheduling pass (after register allocation) at the
3827 -O3 optimization level.
3828 * When using -fprofile-generate with a multi-threaded program, the
3829 profile counts may be slightly wrong due to race conditions. The
3830 new -fprofile-correction option directs the compiler to apply
3831 heuristics to smooth out the inconsistencies. By default the
3832 compiler will give an error message when it finds an inconsistent
3834 * The new -fprofile-dir=PATH option permits setting the directory
3835 where profile data files are stored when using -fprofile-generate
3836 and friends, and the directory used when reading profile data files
3837 using -fprofile-use and friends.
3841 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a
3842 warning if any stack frame is larger than NUMBER bytes. This may be
3843 used to help ensure that code fits within a limited amount of stack
3845 * The command-line option -Wlarger-than-N is now written as
3846 -Wlarger-than=N and the old form is deprecated.
3847 * The new -Wno-mudflap option disables warnings about constructs
3848 which can not be instrumented when using -fmudflap.
3850 New Languages and Language specific improvements
3852 * Version 3.0 of the [5]OpenMP specification is now supported for the
3853 C, C++, and Fortran compilers.
3854 * New character data types, per [6]TR 19769: New character types in
3855 C, are now supported for the C compiler in -std=gnu99 mode, as
3856 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in
3857 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too.
3861 * A new optimize attribute was added to allow programmers to change
3862 the optimization level and particular optimization options for an
3863 individual function. You can also change the optimization options
3864 via the GCC optimize pragma for functions defined after the pragma.
3865 The GCC push_options pragma and the GCC pop_options pragma allow
3866 you temporarily save and restore the options used. The GCC
3867 reset_options pragma restores the options to what was specified on
3869 * Uninitialized warnings do not require enabling optimization
3870 anymore, that is, -Wuninitialized can be used together with -O0.
3871 Nonetheless, the warnings given by -Wuninitialized will probably be
3872 more accurate if optimization is enabled.
3873 * -Wparentheses now warns about expressions such as (!x | y) and (!x
3874 & y). Using explicit parentheses, such as in ((!x) | y), silences
3876 * -Wsequence-point now warns within if, while,do while and for
3877 conditions, and within for begin/end expressions.
3878 * A new option -dU is available to dump definitions of preprocessor
3879 macros that are tested or expanded.
3883 * [7]Improved experimental support for the upcoming ISO C++ standard,
3884 C++0x. Including support for auto, inline namespaces, generalized
3885 initializer lists, defaulted and deleted functions, new character
3886 types, and scoped enums.
3887 * Those errors that may be downgraded to warnings to build legacy
3888 code now mention -fpermissive when -fdiagnostics-show-option is
3890 * -Wconversion now warns if the result of a static_cast to enumeral
3891 type is unspecified because the value is outside the range of the
3893 * -Wuninitialized now warns if a non-static reference or non-static
3894 const member appears in a class without constructors.
3895 * G++ now properly implements value-initialization, so objects with
3896 an initializer of () and an implicitly defined default constructor
3897 will be zero-initialized before the default constructor is called.
3899 Runtime Library (libstdc++)
3901 * [8]Improved experimental support for the upcoming ISO C++ standard,
3903 + Support for <chrono>, <condition_variable>, <cstdatomic>,
3904 <forward_list>, <initializer_list>, <mutex>, <ratio>,
3905 <system_error>, and <thread>.
3906 + unique_ptr, <algorithm> additions, exception propagation, and
3907 support for the new character types in <string> and <limits>.
3908 + Existing facilities now exploit initializer lists, defaulted
3909 and deleted functions, and the newly implemented core C++0x
3911 + Some standard containers are more efficient together with
3912 stateful allocators, i.e., no allocator is constructed on the
3913 fly at element construction time.
3914 * Experimental support for non-standard pointer types in containers.
3915 * The long standing libstdc++/30928 has been fixed for targets
3916 running glibc 2.10 or later.
3917 * As usual, many small and larger bug fixes, in particular quite a
3918 few corner cases in <locale>.
3922 * GNU Fortran now employs libcpp directly instead of using cc1 as an
3923 external preprocessor. The [9]-cpp option was added to allow manual
3924 invocation of the preprocessor without relying on filename
3926 * The [10]-Warray-temporaries option warns about array temporaries
3927 generated by the compiler, as an aid to optimization.
3928 * The [11]-fcheck-array-temporaries option has been added, printing a
3929 notification at run time, when an array temporary had to be created
3930 for an function argument. Contrary to -Warray-temporaries the
3931 warning is only printed if the array is noncontiguous.
3932 * Improved generation of DWARF debugging symbols
3933 * If using an intrinsic not part of the selected standard (via -std=
3934 and -fall-intrinsics) gfortran will now treat it as if this
3935 procedure were declared EXTERNAL and try to link to a user-supplied
3936 procedure. -Wintrinsics-std will warn whenever this happens. The
3937 now-useless option -Wnonstd-intrinsic was removed.
3938 * The flag -falign-commons has been added to control the alignment of
3939 variables in COMMON blocks, which is enabled by default in line
3940 with previous GCC version. Using -fno-align-commons one can force
3941 commons to be contiguous in memory as required by the Fortran
3942 standard, however, this slows down the memory access. The option
3943 -Walign-commons, which is enabled by default, warns when padding
3944 bytes were added for alignment. The proper solution is to sort the
3945 common objects by decreasing storage size, which avoids the
3947 * Fortran 2003 support has been extended:
3948 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is
3949 now supported (except internal reads from/writes to wide
3950 strings). [12]-fbackslash now supports also \unnnn and
3951 \Unnnnnnnn to enter Unicode characters.
3952 + Asynchronous I/O (implemented as synchronous I/O) and the
3953 decimal=, size=, sign=, pad=, blank=, and delim= specifiers
3954 are now supported in I/O statements.
3955 + Support for Fortran 2003 structure constructors and for array
3956 constructor with typespec has been added.
3957 + Procedure Pointers (but not yet as component in derived types
3958 and as function results) are now supported.
3959 + Abstract types, type extension, and type-bound procedures
3960 (both PROCEDURE and GENERIC but not as operators). Note: As
3961 CLASS/polymorphyic types are not implemented, type-bound
3962 procedures with PASS accept as non-standard extension TYPE
3964 * Fortran 2008 support has been added:
3965 + The -std=f2008 option and support for the file extensions
3966 .f2008 and .F2008 has been added.
3967 + The g0 format descriptor is now supported.
3968 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH,
3969 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED
3970 are now available (some of them existed as GNU extension
3971 before). Note: The hyperbolic functions are not yet supporting
3972 complex arguments and the three- argument version of BESSEL_*N
3974 + The bit intrinsics LEADZ and TRAILZ have been added.
3980 * The Ada runtime now supports multilibs on many platforms including
3981 x86_64, SPARC and PowerPC. Their build is enabled by default.
3983 New Targets and Target Specific Improvements
3987 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and
3988 Cortex-R4F processors and has many other improvements to
3989 optimization for ARM processors.
3990 * GCC now supports the VFPv3 variant with 16 double-precision
3991 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been
3992 renamed to -mfpu=vfpv3.
3993 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an
3994 erratum on Cortex-M3 processors.
3995 * GCC now supports the __sync_* atomic operations for ARM EABI
3997 * The section anchors optimization is now enabled by default when
3999 * GCC now uses a new EABI-compatible profiling interface for EABI
4000 targets. This requires a function __gnu_mcount_nc, which is
4001 provided by GNU libc versions 2.8 and later.
4005 * The -mno-tablejump option has been deprecated because it has the
4006 same effect as the -fno-jump-tables option.
4007 * Added support for these new AVR devices:
4036 * Support for Intel AES built-in functions and code generation is
4037 available via -maes.
4038 * Support for Intel PCLMUL built-in function and code generation is
4039 available via -mpclmul.
4040 * Support for Intel AVX built-in functions and code generation is
4041 available via -mavx.
4042 * Automatically align the stack for local variables with alignment
4044 * GCC can now utilize the SVML library for vectorizing calls to a set
4045 of C99 functions if -mveclibabi=svml is specified and you link to
4046 an SVML ABI compatible library.
4047 * On x86-64, the ABI has been changed in the following cases to
4048 conform to the x86-64 ABI:
4049 + Passing/returning structures with flexible array member:
4055 + Passing/returning structures with complex float member:
4059 __complex__ float f;
4061 + Passing/returning unions with long double member:
4067 Code built with previous versions of GCC that uses any of these is
4068 not compatible with code built with GCC 4.4.0 or later.
4069 * A new target attribute was added to allow programmers to change the
4070 target options like -msse2 or -march=k8 for an individual function.
4071 You can also change the target options via the GCC target pragma
4072 for functions defined after the pragma.
4073 * GCC can now be configured with options --with-arch-32,
4074 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and
4075 --with-tune-64 to control the default optimization separately for
4076 32-bit and 64-bit modes.
4080 * Support for __float128 (TFmode) IEEE quad type and corresponding
4081 TCmode IEEE complex quad type is available via the soft-fp library
4082 on IA-32/IA64 targets. This includes basic arithmetic operations
4083 (addition, subtraction, negation, multiplication and division) on
4084 __float128 real and TCmode complex values, the full set of IEEE
4085 comparisons between __float128 values, conversions to and from
4086 float, double and long double floating point types, as well as
4087 conversions to and from signed or unsigned integer, signed or
4088 unsigned long integer and signed or unsigned quad (TImode, IA64
4089 only) integer types. Additionally, all operations generate the full
4090 set of IEEE exceptions and support the full set of IEEE rounding
4095 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4
4096 processors. (Scheduling support for ColdFire V2 processors was
4098 * GCC now supports the -mxgot option to support programs requiring
4099 many GOT entries on ColdFire.
4100 * The m68k-*-linux-gnu target now builds multilibs by default.
4104 * MIPS Technologies have extended the original MIPS SVR4 ABI to
4105 include support for procedure linkage tables (PLTs) and copy
4106 relocations. These extensions allow GNU/Linux executables to use a
4107 significantly more efficient code model than the one defined by the
4109 GCC support for this code model is available via a new command-line
4110 option, -mplt. There is also a new configure-time option,
4111 --with-mips-plt, to make -mplt the default.
4112 The new code model requires support from the assembler, the linker,
4113 and the runtime C library. This support is available in binutils
4115 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables
4116 and 32-bit GNU/Linux shared libraries. This feature requires GNU
4117 binutils 2.19 or above.
4118 * Support for RMI's XLR processor is now available through the
4119 -march=xlr and -mtune=xlr options.
4120 * 64-bit targets can now perform 128-bit multiplications inline,
4121 instead of relying on a libgcc function.
4122 * Native GNU/Linux toolchains now support -march=native and
4123 -mtune=native, which select the host processor.
4124 * GCC now supports the R10K, R12K, R14K and R16K processors. The
4125 canonical -march= and -mtune= names for these processors are
4126 r10000, r12000, r14000 and r16000 respectively.
4127 * GCC can now work around the side effects of speculative execution
4128 on R10K processors. Please see the documentation of the
4129 -mr10k-cache-barrier option for details.
4130 * Support for the MIPS64 Release 2 instruction set has been added.
4131 The option -march=mips64r2 enables generation of these
4133 * GCC now supports Cavium Networks' Octeon processor. This support is
4134 available through the -march=octeon and -mtune=octeon options.
4135 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The
4136 canonical -march= and -mtune= names for these processors are
4137 loongson2e and loongson2f.
4141 Picochip is a 16-bit processor. A typical picoChip contains over 250
4142 small cores, each with small amounts of memory. There are three
4143 processor variants (STAN, MEM and CTRL) with different instruction sets
4144 and memory configurations and they can be chosen using the -mae option.
4146 This port is intended to be a "C" only port.
4148 Power Architecture and PowerPC
4150 * GCC now supports the e300c2, e300c3 and e500mc processors.
4151 * GCC now supports Xilinx processors with a single-precision FPU.
4152 * Decimal floating point is now supported for e500 processors.
4154 S/390, zSeries and System z9/z10
4156 * Support for the IBM System z10 EC/BC processor has been added. When
4157 using the -march=z10 option, the compiler will generate code making
4158 use of instructions provided by the General-Instruction-Extension
4159 Facility and the Execute-Extension Facility.
4163 * GCC now supports the thread-local storage mechanism used on
4168 * GCC now supports thread-local storage (TLS) for Xtensa processor
4169 configurations that include the Thread Pointer option. TLS also
4170 requires support from the assembler and linker; this support is
4171 provided in the GNU binutils beginning with version 2.19.
4173 Documentation improvements
4175 Other significant improvements
4179 This is the [13]list of problem reports (PRs) from GCC's bug tracking
4180 system that are known to be fixed in the 4.4.1 release. This list might
4181 not be complete (that is, it is possible that some PRs that have been
4182 fixed are not listed here).
4186 This is the [14]list of problem reports (PRs) from GCC's bug tracking
4187 system that are known to be fixed in the 4.4.2 release. This list might
4188 not be complete (that is, it is possible that some PRs that have been
4189 fixed are not listed here).
4193 This is the [15]list of problem reports (PRs) from GCC's bug tracking
4194 system that are known to be fixed in the 4.4.3 release. This list might
4195 not be complete (that is, it is possible that some PRs that have been
4196 fixed are not listed here).
4200 This is the [16]list of problem reports (PRs) from GCC's bug tracking
4201 system that are known to be fixed in the 4.4.4 release. This list might
4202 not be complete (that is, it is possible that some PRs that have been
4203 fixed are not listed here).
4207 This is the [17]list of problem reports (PRs) from GCC's bug tracking
4208 system that are known to be fixed in the 4.4.5 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).
4214 This is the [18]list of problem reports (PRs) from GCC's bug tracking
4215 system that are known to be fixed in the 4.4.6 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).
4221 This is the [19]list of problem reports (PRs) from GCC's bug tracking
4222 system that are known to be fixed in the 4.4.7 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).
4227 For questions related to the use of GCC, please consult these web
4228 pages and the [20]GCC manuals. If that fails, the
4229 [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4230 web pages and the development of GCC are welcome on our developer
4231 list at [22]gcc@gcc.gnu.org. All of [23]our lists have public
4234 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and
4235 distribution of this entire article is permitted in any medium,
4236 provided this notice is preserved.
4238 These pages are [25]maintained by the GCC team. Last modified
4243 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7
4244 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted
4245 3. http://gcc.gnu.org/gcc-4.4/porting_to.html
4246 4. http://gcc.gnu.org/wiki/Graphite
4247 5. http://openmp.org/wp/openmp-specifications/
4248 6. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf
4249 7. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html
4250 8. http://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#id476343
4251 9. http://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html
4252 10. http://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125
4253 11. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221
4254 12. http://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34
4255 13. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1
4256 14. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2
4257 15. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3
4258 16. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4
4259 17. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5
4260 18. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6
4261 19. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7
4262 20. http://gcc.gnu.org/onlinedocs/
4263 21. mailto:gcc-help@gcc.gnu.org
4264 22. mailto:gcc@gcc.gnu.org
4265 23. http://gcc.gnu.org/lists.html
4266 24. http://www.fsf.org/
4267 25. http://gcc.gnu.org/about.html
4268 26. http://validator.w3.org/check/referer
4269 ======================================================================
4270 http://gcc.gnu.org/gcc-4.3/index.html
4271 GCC 4.3 Release Series
4275 The [1]GNU project and the GCC developers are pleased to announce the
4276 release of GCC 4.3.6.
4278 This release is a bug-fix release, containing fixes for regressions in
4279 GCC 4.3.5 relative to previous releases of GCC.
4284 Jun 27, 2011 ([2]changes)
4287 May 22, 2010 ([3]changes)
4290 August 4, 2009 ([4]changes)
4293 January 24, 2009 ([5]changes)
4296 August 27, 2008 ([6]changes)
4299 June 6, 2008 ([7]changes)
4302 March 5, 2008 ([8]changes)
4304 References and Acknowledgements
4306 GCC used to stand for the GNU C Compiler, but since the compiler
4307 supports several other languages aside from C, it now stands for the
4308 GNU Compiler Collection.
4310 A list of [9]successful builds is updated as new information becomes
4313 The GCC developers would like to thank the numerous people that have
4314 contributed new features, improvements, bug fixes, and other changes as
4315 well as test results to GCC. This [10]amazing group of volunteers is
4316 what makes GCC successful.
4318 For additional information about GCC please refer to the [11]GCC
4319 project web site or contact the [12]GCC development mailing list.
4321 To obtain GCC please use [13]our mirror sites or [14]our SVN server.
4324 For questions related to the use of GCC, please consult these web
4325 pages and the [15]GCC manuals. If that fails, the
4326 [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4327 web pages and the development of GCC are welcome on our developer
4328 list at [17]gcc@gcc.gnu.org. All of [18]our lists have public
4331 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and
4332 distribution of this entire article is permitted in any medium,
4333 provided this notice is preserved.
4335 These pages are [20]maintained by the GCC team. Last modified
4340 1. http://www.gnu.org/
4341 2. http://gcc.gnu.org/gcc-4.3/changes.html
4342 3. http://gcc.gnu.org/gcc-4.3/changes.html
4343 4. http://gcc.gnu.org/gcc-4.3/changes.html
4344 5. http://gcc.gnu.org/gcc-4.3/changes.html
4345 6. http://gcc.gnu.org/gcc-4.3/changes.html
4346 7. http://gcc.gnu.org/gcc-4.3/changes.html
4347 8. http://gcc.gnu.org/gcc-4.3/changes.html
4348 9. http://gcc.gnu.org/gcc-4.3/buildstat.html
4349 10. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
4350 11. http://gcc.gnu.org/index.html
4351 12. mailto:gcc@gcc.gnu.org
4352 13. http://gcc.gnu.org/mirrors.html
4353 14. http://gcc.gnu.org/svn.html
4354 15. http://gcc.gnu.org/onlinedocs/
4355 16. mailto:gcc-help@gcc.gnu.org
4356 17. mailto:gcc@gcc.gnu.org
4357 18. http://gcc.gnu.org/lists.html
4358 19. http://www.fsf.org/
4359 20. http://gcc.gnu.org/about.html
4360 21. http://validator.w3.org/check/referer
4361 ======================================================================
4362 http://gcc.gnu.org/gcc-4.3/changes.html
4363 GCC 4.3 Release Series
4364 Changes, New Features, and Fixes
4366 The latest release in the 4.3 release series is [1]GCC 4.3.5.
4370 * GCC requires the [2]GMP and [3]MPFR libraries for building all the
4371 various front-end languages it supports. See the [4]prerequisites
4372 page for version requirements.
4373 * ColdFire targets now treat long double as having the same format as
4374 double. In earlier versions of GCC, they used the 68881 long double
4376 * The m68k-uclinux target now uses the same calling conventions as
4377 m68k-linux-gnu. You can select the original calling conventions by
4378 configuring for m68k-uclinuxoldabi instead. Note that
4379 m68k-uclinuxoldabi also retains the original 80-bit long double on
4381 * The -fforce-mem option has been removed because it has had no
4382 effect in the last few GCC releases.
4383 * The i386 -msvr3-shlib option has been removed since it is no longer
4385 * Fastcall for i386 has been changed not to pass aggregate arguments
4386 in registers, following Microsoft compilers.
4387 * Support for the AOF assembler has been removed from the ARM back
4388 end; this affects only the targets arm-semi-aof and armel-semi-aof,
4389 which are no longer recognized. We removed these targets without a
4390 deprecation period because we discovered that they have been
4391 unusable since GCC 4.0.0.
4392 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*)
4393 has been removed. This support had been deprecated since GCC 4.0.0.
4394 * Support for a number of older systems and recently unmaintained or
4395 untested target ports of GCC has been declared obsolete in GCC 4.3.
4396 Unless there is activity to revive them, the next release of GCC
4397 will have their sources permanently removed.
4398 All GCC ports for the following processor architectures have been
4401 The following aliases for processor architectures have been
4402 declared obsolete. Users should use the indicated generic target
4403 names instead, with compile-time options such as -mcpu or
4404 configure-time options such as --with-cpu to control the
4405 configuration more precisely.
4406 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-*
4408 + parisc*-*-* (use hppa*-*-* instead).
4409 + m680[012]0-*-* (use m68k-*-* instead).
4410 All GCC ports for the following operating systems have been
4414 + GNU/Linux using the a.out object format (*-*-linux*aout*)
4415 + GNU/Linux using version 1 of the GNU C Library
4417 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6],
4418 *-*-solaris2.[0-6].*)
4419 + Miscellaneous System V (*-*-sysv*)
4420 + WindISS (*-*-windiss*)
4421 Also, those for some individual systems on particular architectures
4422 have been obsoleted:
4423 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
4424 + CRIS with a.out object format (cris-*-aout)
4425 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
4426 + OSF/1 on PA-RISC (hppa1.1-*-osf*)
4427 + PRO on PA-RISC (hppa1.1-*-pro*)
4428 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*,
4429 i[34567]86-sequent-sysv4*)
4430 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
4431 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host
4432 was previously [5]removed in 2001, leaving only the support
4433 for UWIN as a target now being deprecated)
4434 + ChorusOS on PowerPC (powerpc-*-chorusos*)
4435 + All VAX configurations apart from NetBSD and OpenBSD
4436 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
4437 * The [6]-Wconversion option has been modified. Its purpose now is to
4438 warn for implicit conversions that may alter a value. This new
4439 behavior is available for both C and C++. Warnings about
4440 conversions between signed and unsigned integers can be disabled by
4441 using -Wno-sign-conversion. In C++, they are disabled by default
4442 unless -Wsign-conversion is explicitly requested. The old behavior
4443 of -Wconversion, that is, warn for prototypes causing a type
4444 conversion that is different from what would happen to the same
4445 argument in the absence of a prototype, has been moved to a new
4446 option -Wtraditional-conversion, which is only available for C.
4447 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have
4448 been removed because they were deprecated for more than 3 GCC major
4449 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or
4450 -mtune=pentiumpro as a replacement.
4451 * The -funsafe-math-optimizations option now automatically turns on
4452 -fno-trapping-math in addition to -fno-signed-zeros, as it enables
4453 reassociation and thus may introduce or remove traps.
4454 * The -ftree-vectorize option is now on by default under -O3. In
4455 order to generate code for a SIMD extension, it has to be enabled
4456 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for
4458 * More information on porting to GCC 4.3 from previous versions of
4459 GCC can be found in the [7]porting guide for this release.
4461 General Optimizer Improvements
4463 * The GCC middle-end has been integrated with the [8]MPFR library.
4464 This allows GCC to evaluate and replace at compile-time calls to
4465 built-in math functions having constant arguments with their
4466 mathematically equivalent results. In making use of [9]MPFR, GCC
4467 can generate correct results regardless of the math library
4468 implementation or floating point precision of the host platform.
4469 This also allows GCC to generate identical results regardless of
4470 whether one compiles in native or cross-compile configurations to a
4471 particular target. The following built-in functions take advantage
4472 of this new capability: acos, acosh, asin, asinh, atan2, atan,
4473 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1,
4474 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10,
4475 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh,
4476 tan, tanh, tgamma, y0, y1 and yn. The float and long double
4477 variants of these functions (e.g. sinf and sinl) are also handled.
4478 The sqrt and cabs functions with constant arguments were already
4479 optimized in prior GCC releases. Now they also use [10]MPFR.
4480 * A new forward propagation pass on RTL was added. The new pass
4481 replaces several slower transformations, resulting in compile-time
4482 improvements as well as better code generation in some cases.
4483 * A new command-line switch -frecord-gcc-switches has been added to
4484 GCC, although it is only enabled for some targets. The switch
4485 causes the command line that was used to invoke the compiler to be
4486 recorded into the object file that is being created. The exact
4487 format of this recording is target and binary file format
4488 dependent, but it usually takes the form of a note section
4489 containing ASCII text. The switch is related to the -fverbose-asm
4490 switch, but that one only records the information in the assembler
4491 output file as comments, so the information never reaches the
4493 * The inliner heuristic is now aware of stack frame consumption. New
4494 command-line parameters --param large-stack-frame and --param
4495 large-stack-frame-growth can be used to limit stack frame size
4496 growth caused by inlining.
4497 * During feedback directed optimizations, the expected block size the
4498 memcpy, memset and bzero functions operate on is discovered and for
4499 cases of commonly used small sizes, specialized inline code is
4501 * __builtin_expect no longer requires its argument to be a compile
4503 * Interprocedural optimization was reorganized to work on functions
4504 in SSA form. This enables more precise and cheaper dataflow
4505 analysis and makes writing interprocedural optimizations easier.
4506 The following improvements have been implemented on top of this
4508 + Pre-inline optimization: Selected local optimization passes
4509 are run before the inliner (and other interprocedural passes)
4510 are executed. This significantly improves the accuracy of code
4511 growth estimates used by the inliner and reduces the overall
4512 memory footprint for large compilation units.
4513 + Early inlining (a simple bottom-up inliner pass inlining only
4514 functions whose body is smaller than the expected call
4515 overhead) is now executed with the early optimization passes,
4516 thus inlining already optimized function bodies into an
4517 unoptimized function that is subsequently optimized by early
4518 optimizers. This enables the compiler to quickly eliminate
4519 abstraction penalty in C++ programs.
4520 + Interprocedural constant propagation now operate on SSA form
4521 increasing accuracy of the analysis.
4522 * A new internal representation for GIMPLE statements has been
4523 contributed, resulting in compile-time memory savings.
4524 * The vectorizer was enhanced to support vectorization of outer
4525 loops, intra-iteration parallelism (loop-aware SLP), vectorization
4526 of strided accesses and loops with multiple data-types. Run-time
4527 dependency testing using loop versioning was added. The cost model,
4528 turned on by -fvect-cost-model, was developed.
4530 New Languages and Language specific improvements
4532 * We have added new command-line options
4533 -finstrument-functions-exclude-function-list and
4534 -finstrument-functions-exclude-file-list. They provide more control
4535 over which functions are annotated by the -finstrument-functions
4540 * Implicit conversions between generic vector types are now only
4541 permitted when the two vectors in question have the same number of
4542 elements and compatible element types. (Note that the restriction
4543 involves compatible element types, not implicitly-convertible
4544 element types: thus, a vector type with element type int may not be
4545 implicitly converted to a vector type with element type unsigned
4546 int.) This restriction, which is in line with specifications for
4547 SIMD architectures such as AltiVec, may be relaxed using the flag
4548 -flax-vector-conversions. This flag is intended only as a
4549 compatibility measure and should not be used for new code.
4550 * -Warray-bounds has been added and is now enabled by default for
4551 -Wall . It produces warnings for array subscripts that can be
4552 determined at compile time to be always out of bounds.
4553 -Wno-array-bounds will disable the warning.
4554 * The constructor and destructor function attributes now accept
4555 optional priority arguments which control the order in which the
4556 constructor and destructor functions are run.
4557 * New [11]command-line options -Wtype-limits,
4558 -Wold-style-declaration, -Wmissing-parameter-type, -Wempty-body,
4559 -Wclobbered and -Wignored-qualifiers have been added for finer
4560 control of the diverse warnings enabled by -Wextra.
4561 * A new function attribute alloc_size has been added to mark up
4562 malloc style functions. For constant sized allocations this can be
4563 used to find out the size of the returned pointer using the
4564 __builtin_object_size() function for buffer overflow checking and
4565 similar. This supplements the already built-in malloc and calloc
4566 constant size handling.
4567 * Integer constants written in binary are now supported as a GCC
4568 extension. They consist of a prefix 0b or 0B, followed by a
4569 sequence of 0 and 1 digits.
4570 * A new predefined macro __COUNTER__ has been added. It expands to
4571 sequential integral values starting from 0. In conjunction with the
4572 ## operator, this provides a convenient means to generate unique
4574 * A new command-line option -fdirectives-only has been added. It
4575 enables a special preprocessing mode which improves the performance
4576 of applications like distcc and ccache.
4577 * Fixed-point data types and operators have been added. They are
4578 based on Chapter 4 of the Embedded-C specification (n1169.pdf).
4579 Currently, only MIPS targets are supported.
4580 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732,
4581 N1241, is now supported as a GCC extension to C for targets
4582 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu,
4583 and x86_64-*-linux-gnu. The feature introduces new data types
4584 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF,
4589 * [12]Experimental support for the upcoming ISO C++ standard, C++0x.
4590 * -Wc++0x-compat has been added and is now enabled by default for
4591 -Wall. It produces warnings for constructs whose meaning differs
4592 between ISO C++ 1998 and C++0x.
4593 * The -Wparentheses option now works for C++ as it does for C. It
4594 warns if parentheses are omitted when operators with confusing
4595 precedence are nested. It also warns about ambiguous else
4596 statements. Since -Wparentheses is enabled by -Wall, this may cause
4597 additional warnings with existing C++ code which uses -Wall. These
4598 new warnings may be disabled by using -Wall -Wno-parentheses.
4599 * The -Wmissing-declarations now works for C++ as it does for C.
4600 * The -fvisibility-ms-compat flag was added, to make it easier to
4601 port larger projects using shared libraries from Microsoft's Visual
4602 Studio to ELF and Mach-O systems.
4603 * C++ attribute handling has been overhauled for template arguments
4604 (ie dependent types). In particular, __attribute__((aligned(T)));
4605 works for C++ types.
4607 Runtime Library (libstdc++)
4609 * [13]Experimental support for the upcoming ISO C++ standard, C++0x.
4610 * Support for TR1 mathematical special functions and regular
4611 expressions. ([14]Implementation status of TR1)
4612 * Default what implementations give more elaborate exception strings
4613 for bad_cast, bad_typeid, bad_exception, and bad_alloc.
4614 * Header dependencies have been streamlined, reducing unnecessary
4615 includes and pre-processed bloat.
4616 * Variadic template implementations of items in <tuple> and
4618 * An experimental [15]parallel mode has been added. This is a
4619 parallel implementation of many C++ Standard library algorithms,
4620 like std::accumulate, std::for_each, std::transform, or std::sort,
4621 to give but four examples. These algorithms can be substituted for
4622 the normal (sequential) libstdc++ algorithms on a piecemeal basis,
4623 or all existing algorithms can be transformed via the
4624 -D_GLIBCXX_PARALLEL macro.
4625 * Debug mode versions of classes in <unordered_set> and
4627 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are
4628 now <backward/hash_set> and <backward/hash_map>. This code:
4629 #include <ext/hash_set>
4630 __gnu_cxx::hash_set<int> s;
4632 Can be transformed (in order of preference) to:
4633 #include <tr1/unordered_set>
4634 std::tr1::unordered_set<int> s;
4637 #include <backward/hash_set>
4638 __gnu_cxx::hash_set<int> s;
4640 Similar transformations apply to __gnu_cxx::hash_map,
4641 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set,
4642 __gnu_cxx::hash_multiset.
4646 * Due to the fact that the [16]GMP and [17]MPFR libraries are
4647 required for all languages, Fortran is no longer special in this
4648 regard and is available by default.
4649 * The [18]-fexternal-blas option has been added, which generates
4650 calls to BLAS routines for intrinsic matrix operations such as
4651 matmul rather than using the built-in algorithms.
4652 * Support to give a backtrace (compiler flag -fbacktrace or
4653 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems
4654 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a
4655 run-time error occured.
4656 * GNU Fortran now defines __GFORTRAN__ when it runs the C
4658 * The [19]-finit-local-zero, -finit-real, -finit-integer,
4659 -finit-character, and -finit-logical options have been added, which
4660 can be used to initialize local variables.
4661 * The intrinsic procedures [20]GAMMA and [21]LGAMMA have been added,
4662 which calculate the Gamma function and its logarithm. Use EXTERNAL
4663 gamma if you want to use your own gamma function.
4664 * GNU Fortran now regards the backslash character as literal (as
4665 required by the Fortran 2003 standard); using [22]-fbackslash GNU
4666 Fortran interprets backslashes as C-style escape characters.
4667 * The [23]interpretation of binary, octal and hexadecimal (BOZ)
4668 literal constants has been changed. Before they were always
4669 interpreted as integer; now they are bit-wise transferred as
4670 argument of INT, REAL, DBLE and CMPLX as required by the Fortran
4671 2003 standard, and for real and complex variables in DATA
4672 statements or when directly assigned to real and complex variables.
4673 Everywhere else and especially in expressions they are still
4674 regarded as integer constants.
4675 * Fortran 2003 support has been extended:
4676 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
4678 + Intrinsic module ISO_ENV_FORTRAN
4679 + Interoperability with C (ISO C Bindings)
4680 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER
4686 * GCJ now uses the Eclipse Java compiler for its Java parsing needs.
4687 This enables the use of all 1.5 language features, and fixes most
4688 existing front end bugs.
4689 * libgcj now supports all 1.5 language features which require runtime
4690 support: foreach, enum, annotations, generics, and auto-boxing.
4691 * We've made many changes to the tools shipped with gcj.
4692 + The old jv-scan tool has been removed. This tool never really
4693 worked properly. There is no replacement.
4694 + gcjh has been rewritten. Some of its more obscure options no
4695 longer work, but are still recognized in an attempt at
4696 compatibility. gjavah is a new program with similar
4697 functionality but different command-line options.
4698 + grmic and grmiregistry have been rewritten. grmid has been
4700 + gjar replaces the old fastjar.
4701 + gjarsigner (used for signing jars), gkeytool (used for key
4702 management), gorbd (for CORBA), gserialver (computes
4703 serialization UIDs), and gtnameserv (also for CORBA) are now
4705 * The ability to dump the contents of the java run time heap to a
4706 file for off-line analysis has been added. The heap dumps may be
4707 analyzed with the new gc-analyze tool. They may be generated on
4708 out-of-memory conditions or on demand and are controlled by the new
4709 run time class gnu.gcj.util.GCInfo.
4710 * java.util.TimeZone can now read files from /usr/share/zoneinfo to
4711 provide correct, updated, timezone information. This means that
4712 packagers no longer have to update libgcj when a time zone change
4715 New Targets and Target Specific Improvements
4719 * Tuning for Intel Core 2 processors is available via -mtune=core2
4721 * Tuning for AMD Geode processors is available via -mtune=geode and
4723 * Code generation of block move (memcpy) and block set (memset) was
4724 rewritten. GCC can now pick the best algorithm (loop, unrolled
4725 loop, instruction with rep prefix or a library call) based on the
4726 size of the block being copied and the CPU being optimized for. A
4727 new option -minline-stringops-dynamically has been added. With this
4728 option string operations of unknown size are expanded such that
4729 small blocks are copied by in-line code, while for large blocks a
4730 library call is used. This results in faster code than
4731 -minline-all-stringops when the library implementation is capable
4732 of using cache hierarchy hints. The heuristic choosing the
4733 particular algorithm can be overwritten via -mstringop-strategy.
4734 Newly also memset of values different from 0 is inlined.
4735 * GCC no longer places the cld instruction before string operations.
4736 Both i386 and x86-64 ABI documents mandate the direction flag to be
4737 clear at the entry of a function. It is now invalid to set the flag
4738 in asm statement without reseting it afterward.
4739 * Support for SSSE3 built-in functions and code generation are
4740 available via -mssse3.
4741 * Support for SSE4.1 built-in functions and code generation are
4742 available via -msse4.1.
4743 * Support for SSE4.2 built-in functions and code generation are
4744 available via -msse4.2.
4745 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
4746 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to
4747 allow explicit control of x87 floating point precision.
4748 * Support for __float128 (TFmode) IEEE quad type and corresponding
4749 TCmode IEEE complex quad type is available via the soft-fp library
4750 on x86_64 targets. This includes basic arithmetic operations
4751 (addition, subtraction, negation, multiplication and division) on
4752 __float128 real and TCmode complex values, the full set of IEEE
4753 comparisons between __float128 values, conversions to and from
4754 float, double and long double floating point types, as well as
4755 conversions to and from signed or unsigned integer, signed or
4756 unsigned long integer and signed or unsigned quad (TImode) integer
4757 types. Additionally, all operations generate the full set of IEEE
4758 exceptions and support the full set of IEEE rounding modes.
4759 * GCC can now utilize the ACML library for vectorizing calls to a set
4760 of C99 functions on x86_64 if -mveclibabi=acml is specified and you
4761 link to an ACML ABI compatible library.
4765 * Compiler and Library support for Thumb-2 and the ARMv7 architecture
4772 * Compiler and Library support for the CRIS v32 architecture, as
4773 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been
4776 Configuration changes
4778 * The cris-*-elf target now includes support for CRIS v32, including
4779 libraries, through the -march=v32 option.
4780 * A new crisv32-*-elf target defaults to generate code for CRIS v32.
4781 * A new crisv32-*-linux* target defaults to generate code for CRIS
4783 * The cris-*-aout target has been obsoleted.
4785 Improved support for built-in functions
4787 * GCC can now use the lz and swapwbr instructions to implement the
4788 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
4789 * __builtin_bswap32 is now implemented using the swapwb instruction,
4796 * Support for several new ColdFire processors has been added. You can
4797 generate code for them using the new -mcpu option.
4798 * All targets now support ColdFire processors.
4799 * m68k-uclinux targets have improved support for C++ constructors and
4800 destructors, and for shared libraries.
4801 * It is now possible to set breakpoints on the first or last line of
4802 a function, even if there are no statements on that line.
4806 * Support for sibling calls has been added.
4807 * More use is now made of the ColdFire mov3q instruction.
4808 * __builtin_clz is now implemented using the ff1 ColdFire
4809 instruction, when available.
4810 * GCC now honors the -m68010 option. 68010 code now uses clr rather
4811 than move to zero volatile memory.
4812 * 68020 targets and above can now use symbol(index.size*scale)
4813 addresses for indexed array accesses. Earlier compilers would
4814 always load the symbol into a base register first.
4816 Configuration changes
4818 * All m68k and ColdFire targets now allow the default processor to be
4819 set at configure time using --with-cpu.
4820 * A --with-arch configuration option has been added. This option
4821 allows you to restrict a target to ColdFire or non-ColdFire
4826 * An __mcfv*__ macro is now defined for all ColdFire targets.
4827 (Earlier versions of GCC only defined __mcfv4e__.)
4828 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
4829 * All targets now define __mc68010 and __mc68010__ when generating
4832 Command-line changes
4834 * New command-line options -march, -mcpu, -mtune and -mhard-float
4835 have been added. These options apply to both m68k and ColdFire
4837 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative
4838 versions of -mshort, etc.
4839 * -fforce-addr has been removed. It is now ignored by the compiler.
4843 * ColdFire targets now try to maintain a 4-byte-aligned stack where
4845 * m68k-uclinux targets now try to avoid situations that lead to the
4846 load-time error: BINFMT_FLAT: reloc outside program.
4850 Changes to existing configurations
4852 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32
4853 and n64. Every GNU/Linux configuration now builds these libraries
4855 * GNU/Linux configurations now generate -mno-shared code unless
4856 overridden by -fpic, -fPIC, -fpie or -fPIE.
4857 * mipsisa32*-linux-gnu configurations now generate hard-float code by
4858 default, just like other mipsisa32* and mips*-linux-gnu
4859 configurations. You can build a soft-float version of any
4860 mips*-linux-gnu configuration by passing --with-float=soft to
4862 * mips-wrs-vxworks now supports run-time processes (RTPs).
4864 Changes to existing command-line options
4866 * The -march and -mtune options no longer accept 24k as a processor
4867 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
4868 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and
4869 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The
4870 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for
4871 24kx, 24kex and 34kx.
4875 GCC now supports the following configurations:
4876 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by
4877 default. Earlier releases also recognized this configuration, but
4878 they treated it in the same way as mipsisa32*-linux-gnu*. Note that
4879 you can customize any mips*-linux-gnu* configuration to a
4880 particular ISA or processor by passing an appropriate --with-arch
4881 option to configure.
4882 * mipsisa*-sde-elf*, which provides compatibility with MIPS
4883 Technologies' SDE toolchains. The configuration uses the SDE
4884 libraries by default, but you can use it like other newlib-based
4885 ELF configurations by passing --with-newlib to configure. It is the
4886 only configuration besides mips64vr*-elf* to build MIPS16 as well
4887 as non-MIPS16 libraries.
4888 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf*
4889 configuration, but uses the o32 and o64 ABIs instead of the 32-bit
4890 and 64-bit forms of the EABI.
4892 New processors and application-specific extensions
4894 * Support for the SmartMIPS ASE is available through the new
4896 * Support for revision 2 of the DSP ASE is available through the new
4897 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev
4898 indicates the revision of the ASE in use.
4899 * Support for the 4KS and 74K families of processors is available
4900 through the -march and -mtune options.
4902 Improved support for built-in functions
4904 * GCC can now use load-linked, store-conditional and sync
4905 instructions to implement atomic built-in functions such as
4906 __sync_fetch_and_add. The memory reference must be 4 bytes wide for
4907 32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
4908 * GCC can now use the clz and dclz instructions to implement the
4909 __builtin_ctz and __builtin_ffs families of functions.
4910 * There is a new __builtin___clear_cache function for flushing the
4911 instruction cache. GCC expands this function inline on MIPS32
4912 revision 2 targets, otherwise it calls the function specified by
4917 * GCC can now compile objects that contain a mixture of MIPS16 and
4918 non-MIPS16 code. There are two new attributes, mips16 and nomips16,
4919 for specifying which mode a function should use.
4920 * A new option called -minterlink-mips16 makes non-MIPS16 code
4921 link-compatible with MIPS16 code.
4922 * After many bug fixes, the long-standing MIPS16 -mhard-float support
4923 should now work fairly reliably.
4924 * GCC can now use the MIPS16e save and restore instructions.
4925 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled
4926 with -G0 -fsection-anchors is often smaller than code compiled with
4927 -G8. However, please note that you must usually compile all objects
4928 in your application with the same -G option; see the documentation
4930 * A new option called-mcode-readable specifies which instructions are
4931 allowed to load from the code segment. -mcode-readable=yes is the
4932 default and says that any instruction may load from the code
4933 segment. The other alternatives are -mcode-readable=pcrel, which
4934 says that only PC-relative MIPS16 instructions may load from the
4935 code segment, and -mcode-readable=no, which says that no
4936 instruction may do so. Please see the documentation for more
4937 details, including example uses.
4939 Small-data improvements
4941 There are three new options for controlling small data:
4942 * -mno-extern-sdata, which disables small-data accesses for
4943 externally-defined variables. Code compiled with -Gn
4944 -mno-extern-sdata will be link-compatible with any -G setting
4945 between -G0 and -Gn inclusive.
4946 * -mno-local-sdata, which disables the use of small-data sections for
4947 data that is not externally visible. This option can be a useful
4948 way of reducing small-data usage in less performance-critical parts
4950 * -mno-gpopt, which disables the use of the $gp register while still
4951 honoring the -G limit when placing externally-visible data. This
4952 option implies -mno-extern-sdata and -mno-local-sdata and it can be
4953 useful in situations where $gp does not necessarily hold the
4956 Miscellaneous improvements
4958 * There is a new option called -mbranch-cost for tweaking the
4959 perceived cost of branches.
4960 * If GCC is configured to use a version of GAS that supports the
4961 .gnu_attribute directive, it will use that directive to record
4962 certain properties of the output code. .gnu_attribute is new to GAS
4964 * There are two new function attributes, near and far, for overriding
4965 the command-line setting of -mlong-calls on a function-by-function
4967 * -mfp64, which previously required a 64-bit target, now works with
4968 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and
4969 mipsisa*-sde-elf* configurations provide suitable library support.
4970 * GCC now recognizes the -mdmx and -mmt options and passes them down
4971 to the assembler. It does nothing else with the options at present.
4973 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture
4976 * Support has been added for this new architecture.
4978 RS6000 (POWER/PowerPC)
4980 * Support for the PowerPC 750CL paired-single instructions has been
4981 added with a new powerpc-*-linux*paired* target configuration. It
4982 is enabled by an associated -mpaired option and can be accessed
4983 using new built-in functions.
4984 * Support for auto-detecting architecture and system configuration to
4985 auto-select processor optimization tuning.
4986 * Support for VMX on AIX 5.3 has been added.
4987 * Support for AIX Version 6.1 has been added.
4989 S/390, zSeries and System z9
4991 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been
4992 added. When using the -march=z9-ec option, the compiler will
4993 generate code making use of instructions provided by the decimal
4994 floating point facility and the floating point conversion facility
4995 (pfpo). Besides the instructions used to implement decimal floating
4996 point operations these facilities also contain instructions to move
4997 between general purpose and floating point registers and to modify
4998 and copy the sign-bit of floating point values.
4999 * When the -march=z9-ec option is used the new
5000 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the
5001 decimal floating point hardware instructions will be used or not.
5002 If none of them is given the hardware support is enabled by
5004 * The -mstack-guard option can now be omitted when using stack
5005 checking via -mstack-size in order to let GCC choose a sensible
5006 stack guard value according to the frame size of each function.
5007 * Various changes to improve performance of generated code have been
5008 implemented, including:
5009 + The condition code set by an add logical with carry
5010 instruction is now available for overflow checks like: a + b +
5012 + The test data class instruction is now used to implement
5013 sign-bit and infinity checks of binary and decimal floating
5018 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been
5023 * Stack unwinding for exception handling now uses by default a
5024 specialized version of DWARF unwinding. This is not
5025 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for
5026 Xtensa with previous versions of GCC.
5027 * For Xtensa processors that include the Conditional Store option,
5028 the built-in functions for atomic memory access are now implemented
5029 using S32C1I instructions.
5030 * If the Xtensa NSA option is available, GCC will use it to implement
5031 the __builtin_ctz and __builtin_clz functions.
5033 Documentation improvements
5035 * Existing libstdc++ documentation has been edited and restructured
5036 into a single DocBook XML manual. The results can be viewed online
5039 Other significant improvements
5041 * The compiler's --help command-line option has been extended so that
5042 it now takes an optional set of arguments. These arguments restrict
5043 the information displayed to specific classes of command-line
5044 options, and possibly only a subset of those options. It is also
5045 now possible to replace the descriptive text associated with each
5046 displayed option with an indication of its current value, or for
5047 binary options, whether it has been enabled or disabled.
5048 Here are some examples. The following will display all the options
5049 controlling warning messages:
5052 Whereas this will display all the undocumented, target specific
5054 --help=target,undocumented
5056 This sequence of commands will display the binary optimizations
5057 that are enabled by -O3:
5058 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
5059 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
5060 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
5062 * The configure options --with-pkgversion and --with-bugurl have been
5063 added. These allow distributors of GCC to include a
5064 distributor-specific string in manuals and --version output and to
5065 specify the URL for reporting bugs in their versions of GCC.
5069 This is the [25]list of problem reports (PRs) from GCC's bug tracking
5070 system that are known to be fixed in the 4.3.1 release. This list might
5071 not be complete (that is, it is possible that some PRs that have been
5072 fixed are not listed here).
5074 Target Specific Changes
5080 * Starting with GCC 4.3.1, decimal floating point variables are
5081 aligned to their natural boundaries when they are passed on the
5084 Command-line changes
5086 * Starting with GCC 4.3.1, the -mcld option has been added to
5087 automatically generate a cld instruction in the prologue of
5088 functions that use string instructions. This option is used for
5089 backward compatibility on some operating systems and can be enabled
5090 by default for 32-bit x86 targets by configuring GCC with the
5091 --enable-cld configure option.
5095 This is the [26]list of problem reports (PRs) from GCC's bug tracking
5096 system that are known to be fixed in the 4.3.2 release. This list might
5097 not be complete (that is, it is possible that some PRs that have been
5098 fixed are not listed here).
5102 This is the [27]list of problem reports (PRs) from GCC's bug tracking
5103 system that are known to be fixed in the 4.3.3 release. This list might
5104 not be complete (that is, it is possible that some PRs that have been
5105 fixed are not listed here).
5109 This is the [28]list of problem reports (PRs) from GCC's bug tracking
5110 system that are known to be fixed in the 4.3.4 release. This list might
5111 not be complete (that is, it is possible that some PRs that have been
5112 fixed are not listed here).
5116 This is the [29]list of problem reports (PRs) from GCC's bug tracking
5117 system that are known to be fixed in the 4.3.5 release. This list might
5118 not be complete (that is, it is possible that some PRs that have been
5119 fixed are not listed here).
5123 This is the [30]list of problem reports (PRs) from GCC's bug tracking
5124 system that are known to be fixed in the 4.3.6 release. This list might
5125 not be complete (that is, it is possible that some PRs that have been
5126 fixed are not listed here).
5129 For questions related to the use of GCC, please consult these web
5130 pages and the [31]GCC manuals. If that fails, the
5131 [32]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5132 web pages and the development of GCC are welcome on our developer
5133 list at [33]gcc@gcc.gnu.org. All of [34]our lists have public
5136 Copyright (C) [35]Free Software Foundation, Inc. Verbatim copying and
5137 distribution of this entire article is permitted in any medium,
5138 provided this notice is preserved.
5140 These pages are [36]maintained by the GCC team. Last modified
5145 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5
5146 2. http://gmplib.org/
5147 3. http://www.mpfr.org/
5148 4. http://gcc.gnu.org/install/prerequisites.html
5149 5. http://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html
5150 6. http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
5151 7. http://gcc.gnu.org/gcc-4.3/porting_to.html
5152 8. http://www.mpfr.org/
5153 9. http://www.mpfr.org/
5154 10. http://www.mpfr.org/
5155 11. http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
5156 12. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5157 13. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5158 14. http://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#m anual.intro.status.standard.tr1
5159 15. http://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
5160 16. http://gmplib.org/
5161 17. http://www.mpfr.org/
5162 18. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options
5163 19. http://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167
5164 20. http://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html
5165 21. http://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html
5166 22. http://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html
5167 23. http://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html
5168 24. http://gcc.gnu.org/onlinedocs/libstdc++/
5169 25. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1
5170 26. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2
5171 27. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3
5172 28. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4
5173 29. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5
5174 30. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6
5175 31. http://gcc.gnu.org/onlinedocs/
5176 32. mailto:gcc-help@gcc.gnu.org
5177 33. mailto:gcc@gcc.gnu.org
5178 34. http://gcc.gnu.org/lists.html
5179 35. http://www.fsf.org/
5180 36. http://gcc.gnu.org/about.html
5181 37. http://validator.w3.org/check/referer
5182 ======================================================================
5183 http://gcc.gnu.org/gcc-4.2/index.html
5184 GCC 4.2 Release Series
5188 The [1]GNU project and the GCC developers are pleased to announce the
5189 release of GCC 4.2.4.
5191 This release is a bug-fix release, containing fixes for regressions in
5192 GCC 4.2.3 relative to previous releases of GCC.
5197 May 19, 2008 ([2]changes)
5200 February 1, 2008 ([3]changes)
5203 October 7, 2007 ([4]changes)
5206 July 18, 2007 ([5]changes)
5209 May 13, 2007 ([6]changes)
5211 References and Acknowledgements
5213 GCC used to stand for the GNU C Compiler, but since the compiler
5214 supports several other languages aside from C, it now stands for the
5215 GNU Compiler Collection.
5217 A list of [7]successful builds is updated as new information becomes
5220 The GCC developers would like to thank the numerous people that have
5221 contributed new features, improvements, bug fixes, and other changes as
5222 well as test results to GCC. This [8]amazing group of volunteers is
5223 what makes GCC successful.
5225 For additional information about GCC please refer to the [9]GCC project
5226 web site or contact the [10]GCC development mailing list.
5228 To obtain GCC please use [11]our mirror sites or [12]our SVN server.
5231 For questions related to the use of GCC, please consult these web
5232 pages and the [13]GCC manuals. If that fails, the
5233 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5234 web pages and the development of GCC are welcome on our developer
5235 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
5238 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
5239 distribution of this entire article is permitted in any medium,
5240 provided this notice is preserved.
5242 These pages are [18]maintained by the GCC team. Last modified
5247 1. http://www.gnu.org/
5248 2. http://gcc.gnu.org/gcc-4.2/changes.html
5249 3. http://gcc.gnu.org/gcc-4.2/changes.html
5250 4. http://gcc.gnu.org/gcc-4.2/changes.html
5251 5. http://gcc.gnu.org/gcc-4.2/changes.html
5252 6. http://gcc.gnu.org/gcc-4.2/changes.html
5253 7. http://gcc.gnu.org/gcc-4.2/buildstat.html
5254 8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5255 9. http://gcc.gnu.org/index.html
5256 10. mailto:gcc@gcc.gnu.org
5257 11. http://gcc.gnu.org/mirrors.html
5258 12. http://gcc.gnu.org/svn.html
5259 13. http://gcc.gnu.org/onlinedocs/
5260 14. mailto:gcc-help@gcc.gnu.org
5261 15. mailto:gcc@gcc.gnu.org
5262 16. http://gcc.gnu.org/lists.html
5263 17. http://www.fsf.org/
5264 18. http://gcc.gnu.org/about.html
5265 19. http://validator.w3.org/check/referer
5266 ======================================================================
5267 http://gcc.gnu.org/gcc-4.2/changes.html
5268 GCC 4.2 Release Series
5269 Changes, New Features, and Fixes
5273 * GCC no longer accepts the -fshared-data option. This option has had
5274 no effect in any GCC 4 release; the targets to which the option
5275 used to apply had been removed before GCC 4.0.
5277 General Optimizer Improvements
5279 * New command-line options specify the possible relationships among
5280 parameters and between parameters and global data. For example,
5281 -fargument-noalias-anything specifies that arguments do not alias
5283 Each language will automatically use whatever option is required by
5284 the language standard. You should not need to use these options
5287 New Languages and Language specific improvements
5289 * [1]OpenMP is now supported for the C, C++ and Fortran compilers.
5290 * New command-line options -fstrict-overflow and -Wstrict-overflow
5291 have been added. -fstrict-overflow tells the compiler that it may
5292 assume that the program follows the strict signed overflow
5293 semantics permitted for the language: for C and C++ this means that
5294 the compiler may assume that signed overflow does not occur. For
5295 example, a loop like
5296 for (i = 1; i > 0; i *= 2)
5298 is presumably intended to continue looping until i overflows. With
5299 -fstrict-overflow, the compiler may assume that signed overflow
5300 will not occur, and transform this into an infinite loop.
5301 -fstrict-overflow is turned on by default at -O2, and may be
5302 disabled via -fno-strict-overflow. The -Wstrict-overflow option may
5303 be used to warn about cases where the compiler assumes that signed
5304 overflow will not occur. It takes five different levels:
5305 -Wstrict-overflow=1 to 5. See the [2]documentation for details.
5306 -Wstrict-overflow=1 is enabled by -Wall.
5307 * The new command-line option -fno-toplevel-reorder directs GCC to
5308 emit top-level functions, variables, and asm statements in the same
5309 order that they appear in the input file. This is intended to
5310 support existing code which relies on a particular ordering (for
5311 example, code which uses top-level asm statements to switch
5312 sections). For new code, it is generally better to use function and
5313 variable attributes. The -fno-toplevel-reorder option may be used
5314 for most cases which currently use -fno-unit-at-a-time. The
5315 -fno-unit-at-a-time option will be removed in some future version
5316 of GCC. If you know of a case which requires -fno-unit-at-a-time
5317 which is not fixed by -fno-toplevel-reorder, please open a bug
5322 * The pragma redefine_extname will now macro expand its tokens for
5323 compatibility with SunPRO.
5324 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct
5325 GCC to handle inline functions as specified in the C99 standard. In
5326 preparation for this, GCC 4.2 will warn about any use of non-static
5327 inline functions in gnu99 or c99 mode. This new warning may be
5328 disabled with the new gnu_inline function attribute or the new
5329 -fgnu89-inline command-line option. Also, GCC 4.2 and later will
5330 define one of the preprocessor macros __GNUC_GNU_INLINE__ or
5331 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions
5332 in the current compilation.
5333 * A new command-line option -Waddress has been added to warn about
5334 suspicious uses of memory addresses as, for example, using the
5335 address of a function in a conditional expression, and comparisons
5336 against the memory address of a string literal. This warning is
5341 * C++ visibility handling has been overhauled.
5342 Restricted visiblity is propagated from classes to members, from
5343 functions to local statics, and from templates and template
5344 arguments to instantiations, unless the latter has explicitly
5345 declared visibility.
5346 The visibility attribute for a class must come between the
5347 class-key and the name, not after the closing brace.
5348 Attributes are now allowed for enums and elaborated-type-specifiers
5349 that only declare a type.
5350 Members of the anonymous namespace are now local to a particular
5351 translation unit, along with any other declarations which use them,
5352 though they are still treated as having external linkage for
5354 * The (undocumented) extension which permitted templates with default
5355 arguments to be bound to template template parameters with fewer
5356 parameters has been removed. For example:
5357 template <template <typename> class C>
5358 void f(C<double>) {}
5360 template <typename T, typename U = int>
5363 template void f(S<double>);
5365 is no longer accepted by G++. The reason this code is not accepted
5366 is that S is a template with two parameters; therefore, it cannot
5367 be bound to C which has only one parameter.
5368 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC
5369 releases, have been removed.
5370 * The command-line option -fconst-strings, deprecated in previous GCC
5371 releases, has been removed.
5372 * The configure variable enable-__cxa_atexit is now enabled by
5373 default for more targets. Enabling this variable is necessary in
5374 order for static destructors to be executed in the correct order,
5375 but it depends upon the presence of a non-standard C library in the
5376 target library in order to work. The variable is now enabled for
5377 more targets which are known to have suitable C libraries.
5378 * -Wextra will produce warnings for if statements with a semicolon as
5379 the only body, to catch code like:
5384 To suppress the warning in valid cases, use { } instead.
5385 * The C++ frontend now also produces strict aliasing warnings when
5386 -fstrict-aliasing -Wstrict-aliasing is in effect.
5388 Runtime Library (libstdc++)
5390 * Added support for TR1 <random>, <complex>, and C compatibility
5391 headers. In addition, a lock-free version of shared_ptr was
5392 contributed as part of Phillip Jordan's Google Summer of Code
5393 project on lock-free containers. ([3]Implementation status of TR1)
5394 * In association with the Summer of Code work on lock-free
5395 containers, the interface for atomic builtins was adjusted,
5396 creating simpler alternatives for non-threaded code paths. Also,
5397 usage was consolidated and all elements were moved from namespace
5398 std to namespace__gnu_cxx. Affected interfaces are the functions
5399 __exchange_and_add, __atomic_add, and the objects __mutex,
5400 __recursive_mutex, and __scoped_lock.
5401 * Support for versioning weak symbol names via namespace association
5402 was added. However, as this changes the names of exported symbols,
5403 this is turned off by default in the current ABI. Intrepid users
5404 can enable this feature by using
5405 --enable-symvers=gnu-versioned-namespace during configuration.
5406 * Revised, simplified, and expanded policy-based associative
5407 containers, including data types for tree and trie forms
5408 (basic_tree, tree, trie), lists (list_update), and both
5409 collision-chaining and probing hash-based containers
5410 (basic_hash_table, cc_hash_table, gp_hash_table). More details per
5411 the [4]documentation.
5412 * The implementation of the debug mode was modified, whereby the
5413 debug namespaces were nested inside of namespace std and namespace
5414 __gnu_cxx in order to resolve some long standing corner cases
5415 involving name lookup. Debug functionality from the policy-based
5416 data structures was consolidated and enabled with the single macro,
5417 _GLIBCXX_DEBUG. See PR 26142 for more information.
5418 * Added extensions for type traits: __conditional_type,
5419 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if.
5420 * Added a typelist implementation for compile-time meta-programming.
5421 Elements for typelist construction and operation can be found
5422 within namespace __gnu_cxx::typelist.
5423 * Added a new allocator, __gnu_cxx::throw_allocator, for testing
5425 * Enabled library-wide visibility control, allowing -fvisibility to
5427 * Consolidated all nested namespaces and the conversion of
5428 __gnu_internal implementation-private details to anonymous
5429 namespaces whenever possible.
5430 * Implemented LWG resolutions DR 431 and DR 538.
5434 * Support for allocatable components has been added (TR 15581 and
5436 * Support for the Fortran 2003 streaming IO extension has been added.
5437 * The GNU Fortran compiler now uses 4-byte record markers by default
5438 for unformatted files to be compatible with g77 and most other
5439 compilers. The implementation allows for records greater than 2 GB
5440 and is compatible with several other compilers. Older versions of
5441 gfortran used 8-byte record markers by default (on most systems).
5442 In order to change the length of the record markers, e.g. to read
5443 unformatted files created by older gfortran versions, the
5444 [5]-frecord-marker=8 option can be used.
5448 * A new command-line option -static-libgcj has been added for targets
5449 that use a linker compatible with GNU Binutils. As its name
5450 implies, this causes libgcj to be linked statically. In some cases
5451 this causes the resulting executable to start faster and use less
5452 memory than if the shared version of libgcj were used. However
5453 caution should be used as it can also cause essential parts of the
5454 library to be omitted. Some of these issues are discussed in:
5455 [6]http://gcc.gnu.org/wiki/Statically_linking_libgcj
5456 * fastjar is no longer bundled with GCC. To build libgcj, you will
5457 need either InfoZIP (both zip and unzip) or an external jar
5458 program. In the former case, the GCC build will install a jar shell
5459 script that is based on InfoZIP and provides the same functionality
5462 New Targets and Target Specific Improvements
5466 * -mtune=generic can now be used to generate code running well on
5467 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel
5468 Pentium-M, Intel Pentium 4 and Intel Core 2.
5469 * -mtune=native and -march=native will produce code optimized for the
5470 host architecture as detected using the cpuid instruction.
5471 * Added a new command-line option -fstackrealign and and
5472 __attribute__ ((force_align_arg_pointer)) to realign the stack at
5473 runtime. This allows functions compiled with a vector-aligned stack
5474 to be invoked from legacy objects that keep only word-alignment.
5478 * The default CPU setting has been changed from V7 to V9 in 32-bit
5479 mode on Solaris 7 and above. This is already the case in 64-bit
5480 mode. It can be overridden by specifying --with-cpu at configure
5482 * Back-end support of built-in functions for atomic memory access has
5484 * Support for the Sun UltraSPARC T1 (Niagara) processor has been
5489 * Various bug fixes have made some functions (notably, functions
5490 returning structures) incompatible with previous releases.
5491 Recompiling all libraries is recommended. Note that code quality
5492 has considerably improved since 4.1, making a recompile even more
5497 * Added support for the Broadcom SB-1A core.
5501 * Added support for IA-64 data and control speculation. By default
5502 speculation is enabled only during second scheduler pass. A number
5503 of machine flags was introduced to control the usage of speculation
5504 for both scheduler passes.
5508 * Added Java language support (libffi and libjava) for 32-bit HP-UX
5513 Documentation improvements
5517 * A make pdf target has been added to the top-level makefile,
5518 enabling automated production of PDF documentation files.
5519 (Front-ends external to GCC should modify their Make-lang.in file
5520 to add a lang.pdf: target.)
5522 Other significant improvements
5524 Build system improvements
5526 * All the components of the compiler are now bootstrapped by default.
5527 This improves the resilience to bugs in the system compiler or
5528 binary compatibility problems, as well as providing better testing
5529 of GCC 4.2 itself. In addition, if you build the compiler from a
5530 combined tree, the assembler, linker, etc. will also be
5531 bootstrapped (i.e. built with themselves).
5532 You can disable this behavior, and go back to the pre-GCC 4.2 set
5533 up, by configuring GCC with --disable-bootstrap.
5534 * The rules that configure follows to find target tools resemble more
5535 closely the locations that the built compiler will search. In
5536 addition, you can use the new configure option --with-target-tools
5537 to specify where to find the target tools used during the build,
5538 without affecting what the built compiler will use.
5539 This can be especially useful when building packages of GCC. For
5540 example, you may want to build GCC with GNU as or ld, even if the
5541 resulting compiler to work with the native assembler and linker. To
5542 do so, you can use --with-target-tools to point to the native
5545 Incompatible changes to the build system
5547 * Front-ends external to GCC should modify their Make-lang.in file to
5548 replace double-colon rules (e.g. dvi::) with normal rules (like
5549 lang.dvi:). Front-end makefile hooks do not use double-colon rules
5551 * Up to GCC 4.1, a popular way to specify the target tools used
5552 during the build was to create directories named gas, binutils,
5553 etc. in the build tree, and create links to the tools from there.
5554 This does not work any more when the compiler is bootstrapped. The
5555 new configure option --with-target-tools provides a better way to
5556 achieve the same effect, and works for all native and cross
5560 For questions related to the use of GCC, please consult these web
5561 pages and the [7]GCC manuals. If that fails, the
5562 [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5563 web pages and the development of GCC are welcome on our developer
5564 list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
5567 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
5568 distribution of this entire article is permitted in any medium,
5569 provided this notice is preserved.
5571 These pages are [12]maintained by the GCC team. Last modified
5576 1. http://gcc.gnu.org/projects/gomp/
5577 2. http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
5578 3. http://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#manual.intro.status.standard.tr1
5579 4. http://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html
5580 5. http://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html
5581 6. http://gcc.gnu.org/wiki/Statically_linking_libgcj
5582 7. http://gcc.gnu.org/onlinedocs/
5583 8. mailto:gcc-help@gcc.gnu.org
5584 9. mailto:gcc@gcc.gnu.org
5585 10. http://gcc.gnu.org/lists.html
5586 11. http://www.fsf.org/
5587 12. http://gcc.gnu.org/about.html
5588 13. http://validator.w3.org/check/referer
5589 ======================================================================
5590 http://gcc.gnu.org/gcc-4.1/index.html
5591 GCC 4.1 Release Series
5595 The [1]GNU project and the GCC developers are pleased to announce the
5596 release of GCC 4.1.2.
5598 This release is a bug-fix release, containing fixes for regressions in
5599 GCC 4.1.1 relative to previous releases of GCC.
5604 February 13, 2007 ([2]changes)
5607 May 24, 2006 ([3]changes)
5610 February 28, 2006 ([4]changes)
5612 References and Acknowledgements
5614 GCC used to stand for the GNU C Compiler, but since the compiler
5615 supports several other languages aside from C, it now stands for the
5616 GNU Compiler Collection.
5618 A list of [5]successful builds is updated as new information becomes
5621 The GCC developers would like to thank the numerous people that have
5622 contributed new features, improvements, bug fixes, and other changes as
5623 well as test results to GCC. This [6]amazing group of volunteers is
5624 what makes GCC successful.
5626 For additional information about GCC please refer to the [7]GCC project
5627 web site or contact the [8]GCC development mailing list.
5629 To obtain GCC please use [9]our mirror sites or [10]our SVN server.
5632 For questions related to the use of GCC, please consult these web
5633 pages and the [11]GCC manuals. If that fails, the
5634 [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5635 web pages and the development of GCC are welcome on our developer
5636 list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
5639 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
5640 distribution of this entire article is permitted in any medium,
5641 provided this notice is preserved.
5643 These pages are [16]maintained by the GCC team. Last modified
5648 1. http://www.gnu.org/
5649 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
5650 3. http://gcc.gnu.org/gcc-4.1/changes.html
5651 4. http://gcc.gnu.org/gcc-4.1/changes.html
5652 5. http://gcc.gnu.org/gcc-4.1/buildstat.html
5653 6. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5654 7. http://gcc.gnu.org/index.html
5655 8. mailto:gcc@gcc.gnu.org
5656 9. http://gcc.gnu.org/mirrors.html
5657 10. http://gcc.gnu.org/svn.html
5658 11. http://gcc.gnu.org/onlinedocs/
5659 12. mailto:gcc-help@gcc.gnu.org
5660 13. mailto:gcc@gcc.gnu.org
5661 14. http://gcc.gnu.org/lists.html
5662 15. http://www.fsf.org/
5663 16. http://gcc.gnu.org/about.html
5664 17. http://validator.w3.org/check/referer
5665 ======================================================================
5666 http://gcc.gnu.org/gcc-4.1/changes.html
5667 GCC 4.1 Release Series
5668 Changes, New Features, and Fixes
5670 The latest release in the 4.1 release series is [1]GCC 4.1.2.
5674 General Optimizer Improvements
5676 * GCC now has infrastructure for inter-procedural optimizations and
5677 the following inter-procedural optimizations are implemented:
5678 + Profile guided inlining. When doing profile feedback guided
5679 optimization, GCC can now use the profile to make better
5680 informed decisions on whether inlining of a function is
5681 profitable or not. This means that GCC will no longer inline
5682 functions at call sites that are not executed very often, and
5683 that functions at hot call sites are more likely to be
5685 A new parameter min-inline-recursive-probability is also now
5686 available to throttle recursive inlining of functions with
5687 small average recursive depths.
5688 + Discovery of pure and const functions, a form of side-effects
5689 analysis. While older GCC releases could also discover such
5690 special functions, the new IPA-based pass runs earlier so that
5691 the results are available to more optimizers. The pass is also
5692 simply more powerful than the old one.
5693 + Analysis of references to static variables and type escape
5694 analysis, also forms of side-effects analysis. The results of
5695 these passes allow the compiler to be less conservative about
5696 call-clobbered variables and references. This results in more
5697 redundant loads being eliminated and in making static
5698 variables candidates for register promotion.
5699 + Improvement of RTL-based alias analysis. The results of type
5700 escape analysis are fed to the RTL type-based alias analyzer,
5701 allowing it to disambiguate more memory references.
5702 + Interprocedural constant propagation and function versioning.
5703 This pass looks for functions that are always called with the
5704 same constant value for one or more of the function arguments,
5705 and propagates those constants into those functions.
5706 + GCC will now eliminate static variables whose usage was
5708 + -fwhole-program --combine can now be used to make all
5709 functions in program static allowing whole program
5710 optimization. As an exception, the main function and all
5711 functions marked with the new externally_visible attribute are
5712 kept global so that programs can link with runtime libraries.
5713 * GCC can now do a form of partial dead code elimination (PDCE) that
5714 allows code motion of expressions to the paths where the result of
5715 the expression is actually needed. This is not always a win, so the
5716 pass has been limited to only consider profitable cases. Here is an
5718 int foo (int *, int *);
5734 The a = b + c can be sunk to right before the printf. Normal code
5735 sinking will not do this, it will sink the first one above into the
5736 else-branch of the conditional jump, which still gives you two
5738 * GCC now has a value range propagation pass. This allows the
5739 compiler to eliminate bounds checks and branches. The results of
5740 the pass can also be used to accurately compute branch
5742 * The pass to convert PHI nodes to straight-line code (a form of
5743 if-conversion for GIMPLE) has been improved significantly. The two
5744 most significant improvements are an improved algorithm to
5745 determine the order in which the PHI nodes are considered, and an
5746 improvement that allow the pass to consider if-conversions of basic
5747 blocks with more than two predecessors.
5748 * Alias analysis improvements. GCC can now differentiate between
5749 different fields of structures in Tree-SSA's virtual operands form.
5750 This lets stores/loads from non-overlapping structure fields not
5751 conflict. A new algorithm to compute points-to sets was contributed
5752 that can allows GCC to see now that p->a and p->b, where p is a
5753 pointer to a structure, can never point to the same field.
5754 * Various enhancements to auto-vectorization:
5755 + Incrementally preserve SSA form when vectorizing.
5756 + Incrementally preserve loop-closed form when vectorizing.
5757 + Improvements to peeling for alignment: generate better code
5758 when the misalignment of an access is known at compile time,
5759 or when different accesses are known to have the same
5760 misalignment, even if the misalignment amount itself is
5762 + Consider dependence distance in the vectorizer.
5763 + Externalize generic parts of data reference analysis to make
5764 this analysis available to other passes.
5765 + Vectorization of conditional code.
5766 + Reduction support.
5767 * GCC can now partition functions in sections of hot and cold code.
5768 This can significantly improve performance due to better
5769 instruction cache locality. This feature works best together with
5770 profile feedback driven optimization.
5771 * A new pass to avoid saving of unneeded arguments to the stack in
5772 vararg functions if the compiler can prove that they will not be
5774 * Transition of basic block profiling to tree level implementation
5775 has been completed. The new implementation should be considerably
5776 more reliable (hopefully avoiding profile mismatch errors when
5777 using -fprofile-use or -fbranch-probabilities) and can be used to
5778 drive higher level optimizations, such as inlining.
5779 The -ftree-based-profiling command-line option was removed and
5780 -fprofile-use now implies disabling old RTL level loop optimizer
5781 (-fno-loop-optimize). Speculative prefetching optimization
5782 (originally enabled by -fspeculative-prefetching) was removed.
5784 New Languages and Language specific improvements
5788 * The old Bison-based C and Objective-C parser has been replaced by a
5789 new, faster hand-written recursive-descent parser.
5793 * The build infrastructure for the Ada runtime library and tools has
5794 been changed to be better integrated with the rest of the build
5795 infrastructure of GCC. This should make doing cross builds of Ada a
5800 * ARM-style name-injection of friend declarations is no longer the
5801 default. For example:
5807 will not be accepted; instead a declaration of f will need to be
5808 present outside of the scope of S. The new -ffriend-injection
5809 option will enable the old behavior.
5810 * The (undocumented) extension which permitted templates with default
5811 arguments to be bound to template template parameters with fewer
5812 parameters has been deprecated, and will be removed in the next
5813 major release of G++. For example:
5814 template <template <typename> class C>
5815 void f(C<double>) {}
5817 template <typename T, typename U = int>
5820 template void f(S<double>);
5822 makes use of the deprecated extension. The reason this code is not
5823 valid ISO C++ is that S is a template with two parameters;
5824 therefore, it cannot be bound to C which has only one parameter.
5826 Runtime Library (libstdc++)
5828 * Optimization work:
5829 + A new implementation of std::search_n is provided, better
5830 performing in case of random access iterators.
5831 + Added further efficient specializations of istream functions,
5832 i.e., character array and string extractors.
5833 + Other smaller improvements throughout.
5834 * Policy-based associative containers, designed for high-performance,
5835 flexibility and semantic safety are delivered in ext/pb_assoc.
5836 * A versatile string class, __gnu_cxx::__versa_string, providing
5837 facilities conforming to the standard requirements for
5838 basic_string, is delivered in <ext/vstring.h>. In particular:
5839 + Two base classes are provided: the default one avoids
5840 reference counting and is optimized for short strings; the
5841 alternate one, still uses it while improving in a few low
5842 level areas (e.g., alignment). See vstring_fwd.h for some
5844 + Various algorithms have been rewritten (e.g., replace), the
5845 code streamlined and simple optimizations added.
5846 + Option 3 of DR 431 is implemented for both available bases,
5847 thus improving the support for stateful allocators.
5848 * As usual, many bugs have been fixed (e.g., libstdc++/13583,
5849 libstdc++/23953) and LWG resolutions put into effect for the first
5850 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1
5851 Issue 6.19). The implementation status of TR1 is now tracked in the
5856 * A new language front end for Objective-C++ has been added. This
5857 language allows users to mix the object oriented features of
5858 Objective-C with those of C++.
5862 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19
5863 features (plus some 0.20 bug-fixes)
5865 o The java.net.HttpURLConnection implementation no longer
5866 buffers the entire response body in memory. This means
5867 that response bodies larger than available memory can now
5870 o NIO FileChannel.map implementation, fast bulk put
5871 implementation for DirectByteBuffer (speeds up this
5873 o FileChannel.lock() and FileChannel.force() implemented.
5875 o gnu.xml fix for nodes created outside a namespace
5877 o Add support for output indenting and
5878 cdata-section-elements output instruction in
5880 o xml.xpath corrections for cases where elements/attributes
5881 might have been created in non-namespace-aware mode.
5882 Corrections to handling of XSL variables and minor
5883 conformance updates.
5885 o GNU JAWT implementation, the AWT Native Interface, which
5886 allows direct access to native screen resources from
5887 within a Canvas's paint method. GNU Classpath Examples
5888 comes with a Demo, see libjava/classpath/examples/README.
5889 o awt.datatransfer updated to 1.5 with support for
5890 FlavorEvents. The gtk+ awt peers now allow copy/paste of
5891 text, images, URIs/files and serialized objects with
5892 other applications and tracking clipboard change events
5893 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized
5894 objects are supported). A GNU Classpath Examples
5895 datatransfer Demo was added to show the new
5897 o Split gtk+ awt peers event handling in two threads and
5898 improve gdk lock handling (solves several awt lock ups).
5899 o Speed up awt Image loading.
5900 o Better gtk+ scrollbar peer implementation when using gtk+
5902 o Handle image loading errors correctly for gdkpixbuf and
5904 o Better handle GDK lock. Properly prefix gtkpeer native
5906 o GdkGraphics2D has been updated to use Cairo 0.5.x or
5908 o BufferedImage and GtkImage rewrites. All image drawing
5909 operations should now work correctly (flipping requires
5911 o Future Graphics2D, image and text work is documented at:
5912 [2]http://developer.classpath.org/mediation/ClasspathGrap
5914 o When gtk+ 2.6 or higher is installed the default log
5915 handler will produce stack traces whenever a WARNING,
5916 CRITICAL or ERROR message is produced.
5918 o The RepaintManager has been reworked for more efficient
5919 painting, especially for large GUIs.
5920 o The layout manager OverlayLayout has been implemented,
5921 the BoxLayout has been rewritten to make use of the
5922 SizeRequirements utility class and caching for more
5924 o Improved accessibility support.
5925 o Significant progress has been made in the implementation
5926 of the javax.swing.plaf.metal package, with most UI
5927 delegates in a working state now. Please test this with
5928 your own applications and provide feedback that will help
5929 us to improve this package.
5930 o The GUI demo (gnu.classpath.examples.swing.Demo) has been
5931 extended to highlight various features in our Free Swing
5932 implementation. And it includes a look and feel switcher
5933 for Metal (default), Ocean and GNU themes.
5934 o The javax.swing.plaf.multi package is now implemented.
5935 o Editing and several key actions for JTree and JTable were
5937 o Lots of icons and look and feel improvements for Free
5938 Swing basic and metal themes were added. Try running the
5939 GNU Classpath Swing Demo in examples
5940 (gnu.classpath.examples.swing.Demo) with:
5941 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee
5943 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee
5945 o Start of styled text capabilites for java.swing.text.
5946 o DefaultMutableTreeNode pre-order, post-order, depth-first
5947 and breadth-first traversal enumerations implemented.
5948 o JInternalFrame colors and titlebar draw properly.
5949 o JTree is working up to par (icons, selection and keyboard
5951 o JMenus were made more compatible in visual and
5952 programmatic behavior.
5953 o JTable changeSelection and multiple selections
5955 o JButton and JToggleButton change states work properly
5957 o JFileChooser fixes.
5958 o revalidate() and repaint() fixes which make Free Swing
5959 much more responsive.
5960 o MetalIconFactory implemented.
5961 o Free Swing Top-Level Compatibility. JFrame, JDialog,
5962 JApplet, JInternalFrame, and JWindow are now 1.5
5963 compatible in the sense that you can call add() and
5964 setLayout() directly on them, which will have the same
5965 effect as calling getContentPane().add() and
5966 getContentPane().setLayout().
5967 o The JTree interface has been completed. JTrees now
5968 recognizes mouse clicks and selections work.
5969 o BoxLayout works properly now.
5970 o Fixed GrayFilter to actually work.
5971 o Metal SplitPane implemented.
5972 o Lots of Free Swing text and editor stuff work now.
5973 + Free RMI and Corba
5974 o Andrew Watson, Vice President and Technical Director of
5975 the Object Management Group, has officially assigned us
5976 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will
5977 mark remote classpath-specific system exceptions.
5978 Obtaining the VMCID means that GNU Classpath now is a
5979 recogniseable type of node in a highly interoperable
5981 o GNU Classpath now includes the first working draft to
5982 support the RMI over IIOP protocol. The current
5983 implementation is capable of remote invocations,
5984 transferring various Serializables and Externalizables
5985 via RMI-IIOP protocol. It can flatten graphs and, at
5986 least for the simple cases, is interoperable with 1.5
5988 o org.omg.PortableInterceptor and related functionality in
5989 other packages is now implemented:
5990 # The sever and client interceptors work as required
5992 # The IOR interceptor works as needed for 1.5.
5993 o The org.omg.DynamicAny package is completed and passes
5995 o The Portable Object Adapter should now support the output
5996 of the recent IDL to java compilers. These compilers now
5997 generate servants and not CORBA objects as before, making
5998 the output depend on the existing POA implementation.
5999 Completing POA means that such code can already be tried
6000 to run on Classpath. Our POA is tested for the following
6002 # POA converts servant to the CORBA object.
6003 # Servant provides to the CORBA object.
6004 # POA activates new CORBA object with the given Object
6005 Id (byte array) that is later accessible for the
6007 # During the first call, the ServantActivator provides
6008 servant for this and all subsequent calls on the
6010 # During each call, the ServantLocator provides
6011 servant for this call only.
6012 # ServantLocator or ServantActivator forwards call to
6014 # POA has a single servant, responsible for all
6016 # POA has a default servant, but some objects are
6017 explicitly connected to they specific servants.
6018 The POA is verified using tests from the former
6020 o The CORBA implementation is now a working prototype that
6021 should support features up to 1.3 inclusive. We invite
6022 groups writing CORBA dependent applications to try
6023 Classpath implementation, reporting any possible bugs.
6024 The CORBA prototype is interoperable with Sun's
6025 implementation v 1.4, transferring object references,
6026 primitive types, narrow and wide strings, arrays,
6027 structures, trees, abstract interfaces and value types
6028 (feature of CORBA 2.3) between these two platforms.
6029 Remote exceptions are transferred and handled correctly.
6030 The stringified object references (IORs) from various
6031 sources are parsed as required. The transient (for
6032 current session) and permanent (till jre restart)
6033 redirections work. Both Little and Big Endian encoded
6034 messages are accepted. The implementation is verified
6035 using tests from the former cost.omg.org. The current
6036 release includes working examples (see the examples
6037 directory), demonstrating the client-server
6038 communication, using either CORBA Request or IDL-based
6039 stub (usually generated by a IDL to java compiler). These
6040 examples also show how to use the Classpath CORBA naming
6041 service. The IDL to java compiler is not yet written, but
6042 as our library must be compatible, it naturally accepts
6043 the output of other idlj implementations.
6045 o Updated TimeZone data against Olson tzdata2005l.
6046 o Make zip and jar packages UTF-8 clean.
6047 o "native" code builds and compiles (warning free) on
6049 o java.util.logging.FileHandler now rotates files.
6050 o Start of a generic JDWP framework in gnu/classpath/jdwp.
6051 This is unfinished, but feedback (at classpath@gnu.org)
6052 from runtime hackers is greatly appreciated. Although
6053 most of the work is currently being done around gcj/gij
6054 we want this framework to be as VM neutral as possible.
6055 Early design is described in:
6056 [3]http://gcc.gnu.org/ml/java/2005-05/msg00260.html
6057 o QT4 AWT peers, enable by giving configure
6058 --enable-qt-peer. Included, but not ready for production
6059 yet. They are explicitly disabled and not supported. But
6060 if you want to help with the development of these new
6061 features we are interested in feedback. You will have to
6062 explicitly enable them to try them out (and they will
6063 most likely contain bugs).
6064 o Documentation fixes all over the place. See
6065 [4]http://developer.classpath.org/doc/
6067 New Targets and Target Specific Improvements
6071 * The x86-64 medium model (that allows building applications whose
6072 data segment exceeds 4GB) was redesigned to match latest ABI draft.
6073 New implementation split large datastructures into separate segment
6074 improving performance of accesses to small datastructures and also
6075 allows linking of small model libraries into medium model programs
6076 as long as the libraries are not accessing the large datastructures
6077 directly. Medium model is also supported in position independent
6079 The ABI change results in partial incompatibility among medium
6080 model objects. Linking medium model libraries (or objects) compiled
6081 with new compiler into medium model program compiled with older
6082 will likely result in exceeding ranges of relocations.
6083 Binutils 2.16.91 or newer are required for compiling medium model
6086 RS6000 (POWER/PowerPC)
6088 * The AltiVec vector primitives in <altivec.h> are now implemented in
6089 a way that puts a smaller burden on the preprocessor, instead
6090 processing the "overloading" in the front ends. This should benefit
6091 compilation speed on AltiVec vector code.
6092 * AltiVec initializers now are generated more efficiently.
6093 * The popcountb instruction available on POWER5 now is generated.
6094 * The floating point round to integer instructions available on
6095 POWER5+ now is generated.
6096 * Floating point divides can be synthesized using the floating point
6097 reciprocal estimate instructions.
6098 * Double precision floating point constants are initialized as single
6099 precision values if they can be represented exactly.
6101 S/390, zSeries and System z9
6103 * Support for the IBM System z9 109 processor has been added. When
6104 using the -march=z9-109 option, the compiler will generate code
6105 making use of instructions provided by the extended immediate
6107 * Support for 128-bit IEEE floating point has been added. When using
6108 the -mlong-double-128 option, the compiler will map the long double
6109 data type to 128-bit IEEE floating point. Using this option
6110 constitutes an ABI change, and requires glibc support.
6111 * Various changes to improve performance of generated code have been
6112 implemented, including:
6113 + In functions that do not require a literal pool, register %r13
6114 (which is traditionally reserved as literal pool pointer), can
6115 now be freely used for other purposes by the compiler.
6116 + More precise tracking of register use allows the compiler to
6117 generate more efficient function prolog and epilog code in
6119 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING
6120 instructions are now used to implement C string functions.
6121 + The MOVE CHARACTER instruction with single byte overlap is now
6122 used to implement the memset function with non-zero fill byte.
6123 + The LOAD ZERO instructions are now used where appropriate.
6124 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK,
6125 and INSERT IMMEDIATE instructions are now used more frequently
6126 to optimize bitfield operations.
6127 + The BRANCH ON COUNT instruction is now used more frequently.
6128 In particular, the fact that a loop contains a subroutine call
6129 no longer prevents the compiler from using this instruction.
6130 + The compiler is now aware that all shift and rotate
6131 instructions implicitly truncate the shift count to six bits.
6132 * Back-end support for the following generic features has been
6134 + The full set of [5]built-in functions for atomic memory
6136 + The -fstack-protector feature.
6137 + The optimization pass avoiding unnecessary stores of incoming
6138 argument registers in functions with variable argument list.
6142 * The default code model in 64-bit mode has been changed from
6143 Medium/Anywhere to Medium/Middle on Solaris.
6144 * TLS support is disabled by default on Solaris prior to release 10.
6145 It can be enabled on TLS-capable Solaris 9 versions (4/04 release
6146 and later) by specifying --enable-tls at configure time.
6150 * Support has been added for this new architecture.
6154 Documentation improvements
6156 Other significant improvements
6158 * GCC can now emit code for protecting applications from
6159 stack-smashing attacks. The protection is realized by buffer
6160 overflow detection and reordering of stack variables to avoid
6162 * Some built-in functions have been fortified to protect them against
6163 various buffer overflow (and format string) vulnerabilities.
6164 Compared to the mudflap bounds checking feature, the safe builtins
6165 have far smaller overhead. This means that programs built using
6166 safe builtins should not experience any measurable slowdown.
6170 This is the [6]list of problem reports (PRs) from GCC's bug tracking
6171 system that are known to be fixed in the 4.1.2 release. This list might
6172 not be complete (that is, it is possible that some PRs that have been
6173 fixed are not listed here).
6175 When generating code for a shared library, GCC now recognizes that
6176 global functions may be replaced when the program runs. Therefore, it
6177 is now more conservative in deducing information from the bodies of
6178 functions. For example, in this example:
6183 cout << "Exception";
6187 G++ would previously have optimized away the catch clause, since it
6188 would have concluded that f cannot throw exceptions. Because users may
6189 replace f with another function in the main body of the program, this
6190 optimization is unsafe, and is no longer performed. If you wish G++ to
6191 continue to optimize as before, you must add a throw() clause to the
6192 declaration of f to make clear that it does not throw exceptions.
6195 For questions related to the use of GCC, please consult these web
6196 pages and the [7]GCC manuals. If that fails, the
6197 [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6198 web pages and the development of GCC are welcome on our developer
6199 list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
6202 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
6203 distribution of this entire article is permitted in any medium,
6204 provided this notice is preserved.
6206 These pages are [12]maintained by the GCC team. Last modified
6211 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
6212 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText
6213 3. http://gcc.gnu.org/ml/java/2005-05/msg00260.html
6214 4. http://developer.classpath.org/doc/
6215 5. http://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
6216 6. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2
6217 7. http://gcc.gnu.org/onlinedocs/
6218 8. mailto:gcc-help@gcc.gnu.org
6219 9. mailto:gcc@gcc.gnu.org
6220 10. http://gcc.gnu.org/lists.html
6221 11. http://www.fsf.org/
6222 12. http://gcc.gnu.org/about.html
6223 13. http://validator.w3.org/check/referer
6224 ======================================================================
6225 http://gcc.gnu.org/gcc-4.0/index.html
6226 GCC 4.0 Release Series
6230 The [1]GNU project and the GCC developers are pleased to announce the
6231 release of GCC 4.0.4.
6233 This release is a bug-fix release, containing fixes for regressions in
6234 GCC 4.0.3 relative to previous releases of GCC.
6239 January 31, 2007 ([2]changes)
6242 March 10, 2006 ([3]changes)
6245 September 28, 2005 ([4]changes)
6248 July 7, 2005 ([5]changes)
6251 April 20, 2005 ([6]changes)
6253 References and Acknowledgements
6255 GCC used to stand for the GNU C Compiler, but since the compiler
6256 supports several other languages aside from C, it now stands for the
6257 GNU Compiler Collection.
6259 A list of [7]successful builds is updated as new information becomes
6262 The GCC developers would like to thank the numerous people that have
6263 contributed new features, improvements, bug fixes, and other changes as
6264 well as test results to GCC. This [8]amazing group of volunteers is
6265 what makes GCC successful.
6267 For additional information about GCC please refer to the [9]GCC project
6268 web site or contact the [10]GCC development mailing list.
6270 To obtain GCC please use [11]our mirror sites, or [12]our SVN server.
6273 For questions related to the use of GCC, please consult these web
6274 pages and the [13]GCC manuals. If that fails, the
6275 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6276 web pages and the development of GCC are welcome on our developer
6277 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
6280 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
6281 distribution of this entire article is permitted in any medium,
6282 provided this notice is preserved.
6284 These pages are [18]maintained by the GCC team. Last modified
6289 1. http://www.gnu.org/
6290 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
6291 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3
6292 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2
6293 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1
6294 6. http://gcc.gnu.org/gcc-4.0/changes.html
6295 7. http://gcc.gnu.org/gcc-4.0/buildstat.html
6296 8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6297 9. http://gcc.gnu.org/index.html
6298 10. mailto:gcc@gcc.gnu.org
6299 11. http://gcc.gnu.org/mirrors.html
6300 12. http://gcc.gnu.org/svn.html
6301 13. http://gcc.gnu.org/onlinedocs/
6302 14. mailto:gcc-help@gcc.gnu.org
6303 15. mailto:gcc@gcc.gnu.org
6304 16. http://gcc.gnu.org/lists.html
6305 17. http://www.fsf.org/
6306 18. http://gcc.gnu.org/about.html
6307 19. http://validator.w3.org/check/referer
6308 ======================================================================
6309 http://gcc.gnu.org/gcc-4.0/changes.html
6310 GCC 4.0 Release Series
6311 Changes, New Features, and Fixes
6313 The latest release in the 4.0 release series is [1]GCC 4.0.4.
6317 * GCC now generates location lists by default when compiling with
6318 debug info and optimization.
6319 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1
6320 or later is needed to debug binaries containing location
6322 + When you are trying to view a value of a variable in a part of
6323 a function where it has no location (for example when the
6324 variable is no longer used and thus its location was used for
6325 something else) GDB will say that it is not available.
6326 You can disable generating location lists by -fno-var-tracking.
6327 * GCC no longer accepts the -fwritable-strings option. Use named
6328 character arrays when you need a writable string.
6329 * The options -freduce-all-givs and -fmove-all-movables have been
6330 discontinued. They were used to circumvent a shortcoming in the
6331 heuristics of the old loop optimization code with respect to common
6332 Fortran constructs. The new (tree) loop optimizer works differently
6333 and doesn't need those work-arounds.
6334 * The graph-coloring register allocator, formerly enabled by the
6335 option -fnew-ra, has been discontinued.
6336 * -I- has been deprecated. -iquote is meant to replace the need for
6338 * The MIPS -membedded-pic and -mrnames options have been removed.
6339 * All MIPS targets now require the GNU assembler. In particular, IRIX
6340 configurations can no longer use the MIPSpro assemblers, although
6341 they do still support the MIPSpro linkers.
6342 * The SPARC option -mflat has been removed.
6343 * English-language diagnostic messages will now use Unicode quotation
6344 marks in UTF-8 locales. (Non-English messages already used the
6345 quotes appropriate for the language in previous releases.) If your
6346 terminal does not support UTF-8 but you are using a UTF-8 locale
6347 (such locales are the default on many GNU/Linux systems) then you
6348 should set LC_CTYPE=C in the environment to disable that locale.
6349 Programs that parse diagnostics and expect plain ASCII
6350 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's
6351 explanation of Unicode quotation marks for more information.
6352 * The specs file is no longer installed on most platforms. Most users
6353 will be totally unaffected. However, if you are accustomed to
6354 editing the specs file yourself, you will now have to use the
6355 -dumpspecs option to generate the specs file, and then edit the
6358 General Optimizer Improvements
6360 * The [3]tree ssa branch has been merged. This merge has brought in a
6361 completely new optimization framework based on a higher level
6362 intermediate representation than the existing RTL representation.
6363 Numerous new code transformations based on the new framework are
6364 available in GCC 4.0, including:
6365 + Scalar replacement of aggregates
6366 + Constant propagation
6367 + Value range propagation
6368 + Partial redundancy elimination
6369 + Load and store motion
6370 + Strength reduction
6371 + Dead store elimination
6372 + Dead and unreachable code elimination
6373 + [4]Autovectorization
6375 + Tail recursion by accumulation
6376 Many of these passes outperform their counterparts from previous
6378 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction
6379 scheduling optimization intended for loops that perform heavy
6382 New Languages and Language specific improvements
6386 * The sentinel attribute has been added to GCC. This function
6387 attribute allows GCC to warn when variadic functions such as execl
6388 are not NULL terminated. See the GCC manual for a complete
6389 description of its behavior.
6390 * Given __attribute__((alias("target"))) it is now an error if target
6391 is not a symbol, defined in the same translation unit. This also
6392 applies to aliases created by #pragma weak alias=target. This is
6393 because it's meaningless to define an alias to an undefined symbol.
6394 On Solaris, the native assembler would have caught this error, but
6399 * The -Wstrict-aliasing=2 option has been added. This warning catches
6400 all unsafe cases, but it may also give a warning for some cases
6402 * The cast-as-lvalue, conditional-expression-as-lvalue and
6403 compound-expression-as-lvalue extensions, which were deprecated in
6404 3.3.4 and 3.4, have been removed.
6405 * The -fwritable-strings option, which was deprecated in 3.4, has
6407 * #pragma pack() semantics have been brought closer to those used by
6408 other compilers. This also applies to C++.
6409 * Taking the address of a variable with register storage is invalid
6410 in C. GCC now issues an error instead of a warning.
6411 * Arrays of incomplete element type are invalid in C. GCC now issues
6412 an error for such arrays. Declarations such as extern struct s x[];
6413 (where struct s has not been defined) can be moved after the
6414 definition of struct s. Function parameters declared as arrays of
6415 incomplete type can instead be declared as pointers.
6419 * When compiling without optimizations (-O0), the C++ frontend is
6420 much faster than in any previous versions of GCC. Independent
6421 testers have measured speed-ups up to 25% in real-world production
6422 code, compared to the 3.4 family (which was already the fastest
6423 version to date). Upgrading from older versions might show even
6424 bigger improvements.
6425 * ELF visibility attributes can now be applied to a class type, so
6426 that it affects every member function of a class at once, without
6427 having to specify each individually:
6428 class __attribute__ ((visibility("hidden"))) Foo
6433 The syntax is deliberately similar to the __declspec() system used
6434 by Microsoft Windows based compilers, allowing cross-platform
6435 projects to easily reuse their existing macro system for denoting
6436 exports and imports. By explicitly marking internal classes never
6437 used outside a binary as hidden, one can completely avoid PLT
6438 indirection overheads during their usage by the compiler. You can
6439 find out more about the advantages of this at
6440 [6]http://www.akkadia.org/drepper/dsohowto.pdf
6441 * The -fvisibility-inlines-hidden option has been added which marks
6442 all inlineable functions as having hidden ELF visibility, thus
6443 removing their symbol and typeinfo from the exported symbol table
6444 of the output ELF binary. Using this option can reduce the exported
6445 symbol count of template-heavy code by up to 40% with no code
6446 change at all, thus notably improving link and load times for the
6447 binary as well as a reduction in size of up to 10%. Also, check the
6448 new [7]-fvisibility option.
6449 * The compiler now uses the library interface specified by the [8]C++
6450 ABI for thread-safe initialization of function-scope static
6451 variables. Most users should leave this alone, but embedded
6452 programmers may want to disable this by specifying
6453 -fno-threadsafe-statics for a small savings in code size.
6454 * Taking the address of an explicit register variable is no longer
6455 supported. Note that C++ allows taking the address of variables
6456 with register storage so this will continue to compile with a
6457 warning. For example, assuming that r0 is a machine register:
6458 register int foo asm ("r0");
6460 &foo; // error, no longer accepted
6461 &bar; // OK, with a warning
6462 * G++ has an undocumented extension to virtual function covariancy
6463 rules that allowed the overrider to return a type that was
6464 implicitly convertable to the overridden function's return type.
6465 For instance a function returning void * could be overridden by a
6466 function returning T *. This is now deprecated and will be removed
6467 in a future release.
6468 * The G++ minimum and maximum operators (<? and >?) and their
6469 compound forms (<?=) and >?=) have been deprecated and will be
6470 removed in a future version. Code using these operators should be
6471 modified to use std::min and std::max instead.
6472 * Declaration of nested classes of class templates as friends are
6474 template <typename T> struct A {
6478 template <typename T> friend class A<T>::B;
6480 This complements the feature member functions of class templates as
6481 friends introduced in GCC 3.4.0.
6482 * When declaring a friend class using an unqualified name, classes
6483 outside the innermost non-class scope are not searched:
6487 friend class A; // Refer to N::A which has not been declared yet
6488 // because name outside namespace N are not searched
6489 friend class ::A; // Refer to ::A
6492 Hiding the friend name until declaration is still not implemented.
6493 * Friends of classes defined outside their namespace are correctly
6499 friend class B; // Refer to N::B in GCC 4.0.0
6500 // but ::B in earlier versions of GCC
6503 Runtime Library (libstdc++)
6505 * Optimization work:
6506 + Added efficient specializations of istream functions for char
6508 + Further performance tuning of strings, in particular wrt
6509 single-char append and getline.
6510 + iter_swap - and therefore most of the mutating algorithms -
6511 now makes an unqualified call to swap when the value_type of
6512 the two iterators is the same.
6513 * A large subset of the features in Technical Report 1 (TR1 for
6514 short) is experimentally delivered (i.e., no guarantees about the
6515 implementation are provided. In particular it is not promised that
6516 the library will remain link-compatible when code using TR1 is
6518 + General utilities such as reference_wrapper and shared_ptr.
6519 + Function objects, i.e., result_of, mem_fn, bind, function.
6520 + Support for metaprogramming.
6521 + New containers such as tuple, array, unordered_set,
6522 unordered_map, unordered_multiset, unordered_multimap.
6523 * As usual, many bugs have been fixed and LWG resolutions implemented
6524 for the first time (e.g., DR 409).
6528 * In order to prevent naming conflicts with other implementations of
6529 these tools, some GCJ binaries have been renamed:
6530 + rmic is now grmic,
6531 + rmiregistry is now grmiregistry, and
6532 + jar is now fastjar.
6533 In particular, these names were problematic for the jpackage.org
6534 packaging conventions which install symlinks in /usr/bin that point
6535 to the preferred versions of these tools.
6536 * The -findirect-dispatch argument to the compiler now works and
6537 generates code following a new "binary compatibility" ABI. Code
6538 compiled this way follows the binary compatibility rules of the
6539 Java Language Specification.
6540 * libgcj now has support for using GCJ as a JIT, using the
6541 gnu.gcj.jit family of system properties.
6542 * libgcj can now find a shared library corresponding to the bytecode
6543 representation of a class. See the documentation for the new
6544 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system
6546 * There have been many improvements to the class library. Here are
6548 + Much more of AWT and Swing exist.
6549 + Many new packages and classes were added, including
6550 java.util.regex, java.net.URI, javax.crypto,
6551 javax.crypto.interfaces, javax.crypto.spec, javax.net,
6552 javax.net.ssl, javax.security.auth,
6553 javax.security.auth.callback, javax.security.auth.login,
6554 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss,
6555 javax.imageio, javax.imageio.event, javax.imageio.spi,
6556 javax.print, javax.print.attribute,
6557 javax.print.attribute.standard, javax.print.event, and
6559 + Updated SAX and DOM, and imported GNU JAXP
6563 * A new [9]Fortran front end has replaced the aging GNU Fortran 77
6564 front end. The new front end supports Fortran 90 and Fortran 95. It
6565 may not yet be as stable as the old Fortran front end.
6569 * Ada (with tasking and Zero Cost Exceptions) is now available on
6570 many more targets, including but not limited to: alpha-linux,
6571 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux,
6572 s390x-linux, sparc-linux.
6573 * Some of the new Ada 2005 features are now implemented like
6574 Wide_Wide_Character and Ada.Containers.
6575 * Many bugs have been fixed, tools and documentation improved.
6576 * To compile Ada from the sources, install an older working Ada
6577 compiler and then use --enable-languages=ada at configuration time,
6578 since the Ada frontend is not currently activated by default. See
6579 the [10]Installing GCC for details.
6581 New Targets and Target Specific Improvements
6585 * The frame layout has changed. In the new layout, the prologue of a
6586 function first saves registers and then allocate space for locals,
6587 resulting in an 1% improvement on code size.
6589 IA-32/x86-64 (AMD64)
6591 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10,
6592 log1p, log2, logb and tan mathematical builtins (and their float
6593 and long double variants) are now implemented as inline x87
6594 intrinsics when using -ffast-math.
6595 * The ceil, floor, nearbyint, rint and trunc mathematical builtins
6596 (and their float and long double variants) are now implemented as
6597 inline x87 intrinsics when using -ffast-math.
6598 * The x87's fsincos instruction is now used automatically with
6599 -ffast-math when calculating both the sin and cos of the same
6601 * Instruction selection for multiplication and division by constants
6606 * Floating point division, integer division and sqrt are now inlined,
6607 resulting in significant performance improvements on some codes.
6611 * Division by zero checks now use conditional traps if the target
6612 processor supports them. This decreases code size by one word per
6613 division operation. The old behavior (branch and break) can be
6614 obtained either at configure time by passing --with-divide=breaks
6615 to configure or at runtime by passing -mdivide-breaks to GCC.
6616 * Support for MIPS64 paired-single instructions has been added. It is
6617 enabled by -mpaired-single and can be accessed using both the
6618 target-independent vector extensions and new MIPS-specific built-in
6620 * Support for the MIPS-3D ASE has been added. It is enabled by
6621 -mips3d and provides new MIPS-3D-specific built-in functions.
6622 * The -mexplicit-relocs option now supports static n64 code (as is
6623 used, for example, in 64-bit linux kernels). -mexplicit-relocs
6624 should now be feature-complete and is enabled by default when GCC
6625 is configured to use a compatible assembler.
6626 * Support for the NEC VR4130 series has been added. This support
6627 includes the use of VR-specific instructions and a new VR4130
6628 scheduler. Full VR4130 support can be selected with -march=vr4130
6629 while code for any ISA can be tuned for the VR4130 using
6630 -mtune=vr4130. There is also a new -mvr4130-align option that
6631 produces better schedules at the cost of increased code size.
6632 * Support for the Broadcom SB-1 has been extended. There is now an
6633 SB-1 scheduler as well as support for the SB-1-specific
6634 paired-single instructions. Full SB-1 support can be selected with
6635 -march=sb1 while code for any ISA can be optimized for the SB-1
6637 * The compiler can now work around errata in R4000, R4400, VR4120 and
6638 VR4130 processors. These workarounds are enabled by -mfix-r4000,
6639 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120
6640 and VR4130 workarounds need binutils 2.16 or above.
6641 * IRIX shared libraries are now installed into the standard library
6642 directories: o32 libraries go into lib/, n32 libraries go into
6643 lib32/ and n64 libraries go into lib64/.
6644 * The compiler supports a new -msym32 option. It can be used to
6645 optimize n64 code in which all symbols are known to have 32-bit
6650 * New command-line options help to generate code intended to run in
6651 an environment where stack space is restricted, e.g. Linux kernel
6653 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time
6654 warnings for single functions that require large or dynamic
6656 + -mstack-size and -mstack-guard generate code that checks for
6657 stack overflow at run time.
6658 + -mpacked-stack generates code that reduces the stack frame
6659 size of many functions by reusing unneeded parts of the stack
6661 * The -msoft-float option now ensures that generated code never
6662 accesses floating point registers.
6663 * The s390x-ibm-tpf target now fully supports C++, including
6664 exceptions and threads.
6665 * Various changes to improve performance of the generated code have
6666 been implemented, including:
6667 + GCC now uses sibling calls where possible.
6668 + Condition code handling has been optimized, allowing GCC to
6669 omit redundant comparisons in certain cases.
6670 + The cost function guiding many optimizations has been refined
6671 to more accurately represent the z900 and z990 processors.
6672 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW
6673 instructions are now used to avoid conditional branches in
6675 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to
6676 optimize address arithmetic required to access large stack
6678 + GCC now makes more efficient use of memory-to-memory type
6679 instructions (MVC, CLC, ...).
6680 + More precise tracking of special register use allows better
6681 instruction scheduling, in particular of the function prologue
6682 and epilogue sequences.
6683 + The Java front end now generates inline code to implement
6684 integer division, instead of calling library routines.
6688 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and
6689 -mf934 have been removed. They have been replaced with -mcpu=xxx.
6690 * The internal model used to estimate the relative cost of each
6691 instruction has been updated. It is expected to give better results
6692 on recent UltraSPARC processors.
6693 * Code generation for function prologues and epilogues has been
6694 improved, resulting in better scheduling and allowing multiple exit
6695 points in functions.
6696 * Support for Sun's Visual Instruction Set (VIS) has been enhanced.
6697 It is enabled by -mvis and provides new built-in functions for VIS
6698 instructions on UltraSPARC processors.
6699 * The option -mapp-regs has been turned on by default on Solaris too.
6703 * Novell NetWare (on ix86, no other hardware platform was ever really
6704 supported by this OS) has been re-enabled and the ABI supported by
6705 GCC has been brought into sync with that of MetroWerks CodeWarrior
6706 (the ABI previously supported was that of some Unix systems, which
6707 NetWare never tried to support).
6711 Support for a number of older systems has been declared obsolete in GCC
6712 4.0. Unless there is activity to revive them, the next release of GCC
6713 will have their sources permanently removed.
6715 All GCC ports for the following processor architectures have been
6719 * National Semiconductor NS32K
6720 * Texas Instruments TMS320C[34]x
6722 Also, those for some individual systems have been obsoleted:
6724 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf,
6726 + OpenBSD 32-bit (sparc-*-openbsd*)
6728 Documentation improvements
6730 Other significant improvements
6732 * Location lists are now generated by default when compiling with
6733 debug info and optimization. Location lists provide more accurate
6734 debug info about locations of variables and they allow debugging
6735 code compiled with -fomit-frame-pointer.
6736 * The -fvisibility option has been added which allows the default ELF
6737 visibility of all symbols to be set per compilation and the new
6738 #pragma GCC visibility preprocessor command allows the setting of
6739 default ELF visibility for a region of code. Using
6740 -fvisibility=hidden especially in combination with the new
6741 -fvisibility-inlines-hidden can yield substantial improvements in
6742 output binary quality including avoiding PLT indirection overheads,
6743 reduction of the exported symbol count by up to 60% (with resultant
6744 improvements to link and load times), better scope for the
6745 optimizer to improve code and up to a 20% reduction in binary size.
6746 Using these options correctly yields a binary with a similar symbol
6747 count to a Windows DLL.
6748 Perhaps more importantly, this new feature finally allows (with
6749 careful planning) complete avoidance of symbol clashes when
6750 manually loading shared objects with RTLD_GLOBAL, thus finally
6751 solving problems many projects such as python were forced to use
6752 RTLD_LOCAL for (with its resulting issues for C++ correctness). You
6753 can find more information about using these options at
6754 [11]http://gcc.gnu.org/wiki/Visibility.
6755 __________________________________________________________________
6759 This is the [12]list of problem reports (PRs) from GCC's bug tracking
6760 system that are known to be fixed in the 4.0.1 release. This list might
6761 not be complete (that is, it is possible that some PRs that have been
6762 fixed are not listed here).
6766 This is the [13]list of problem reports (PRs) from GCC's bug tracking
6767 system that are known to be fixed in the 4.0.2 release. This list might
6768 not be complete (that is, it is possible that some PRs that have been
6769 fixed are not listed here).
6771 Unfortunately, due to a release engineering failure, this release has a
6772 regression on Solaris that will affect some C++ programs. We suggest
6773 that Solaris users apply a [14]patch that corrects the problem. Users
6774 who do not wish to apply the patch should explicitly link C++ programs
6775 with the -pthreads option, even if they do not use threads. This
6776 problem has been corrected in the current 4.0 branch sources and will
6777 not be present in GCC 4.0.3.
6781 Starting with this release, the function getcontext is recognized by
6782 the compiler as having the same semantics as the setjmp function. In
6783 particular, the compiler will ensure that all registers are dead before
6784 calling such a function and will emit a warning about the variables
6785 that may be clobbered after the second return from the function.
6789 This is the [15]list of problem reports (PRs) from GCC's bug tracking
6790 system that are known to be fixed in the 4.0.4 release. This list might
6791 not be complete (that is, it is possible that some PRs that have been
6792 fixed are not listed here).
6794 The 4.0.4 release is provided for those that require a high degree of
6795 binary compatibility with previous 4.0.x releases. For most users, the
6796 GCC team recommends that version 4.1.1 or later be used instead."
6799 For questions related to the use of GCC, please consult these web
6800 pages and the [16]GCC manuals. If that fails, the
6801 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6802 web pages and the development of GCC are welcome on our developer
6803 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
6806 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
6807 distribution of this entire article is permitted in any medium,
6808 provided this notice is preserved.
6810 These pages are [21]maintained by the GCC team. Last modified
6815 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
6816 2. http://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
6817 3. http://gcc.gnu.org/projects/tree-ssa/
6818 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html
6819 5. http://gcc.gnu.org/news/sms.html
6820 6. http://www.akkadia.org/drepper/dsohowto.pdf
6821 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility
6822 8. http://mentorembedded.github.com/cxx-abi/
6823 9. http://gcc.gnu.org/fortran/
6824 10. http://gcc.gnu.org/install/
6825 11. http://gcc.gnu.org/wiki/Visibility
6826 12. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1
6827 13. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2
6828 14. http://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html
6829 15. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4
6830 16. http://gcc.gnu.org/onlinedocs/
6831 17. mailto:gcc-help@gcc.gnu.org
6832 18. mailto:gcc@gcc.gnu.org
6833 19. http://gcc.gnu.org/lists.html
6834 20. http://www.fsf.org/
6835 21. http://gcc.gnu.org/about.html
6836 22. http://validator.w3.org/check/referer
6837 ======================================================================
6838 http://gcc.gnu.org/gcc-3.4/index.html
6839 GCC 3.4 Release Series
6843 The [1]GNU project and the GCC developers are pleased to announce the
6844 release of GCC 3.4.6.
6846 This release is a bug-fix release, containing fixes for regressions in
6847 GCC 3.4.4 relative to previous releases of GCC. This is the last of the
6850 The GCC 3.4 release series includes numerous [2]new features,
6851 improvements, bug fixes, and other changes, thanks to an [3]amazing
6852 group of volunteers.
6857 March 6, 2006 ([4]changes)
6860 November 30, 2005 ([5]changes)
6863 May 18, 2005 ([6]changes)
6866 November 4, 2004 ([7]changes)
6869 September 6, 2004 ([8]changes)
6872 July 1, 2004 ([9]changes)
6875 April 18, 2004 ([10]changes)
6877 References and Acknowledgements
6879 GCC used to stand for the GNU C Compiler, but since the compiler
6880 supports several other languages aside from C, it now stands for the
6881 GNU Compiler Collection.
6883 A list of [11]successful builds is updated as new information becomes
6886 The GCC developers would like to thank the numerous people that have
6887 contributed new features, improvements, bug fixes, and other changes as
6888 well as test results to GCC. This [12]amazing group of volunteers is
6889 what makes GCC successful.
6891 For additional information about GCC please refer to the [13]GCC
6892 project web site or contact the [14]GCC development mailing list.
6894 To obtain GCC please use [15]our mirror sites, or [16]our SVN server.
6897 For questions related to the use of GCC, please consult these web
6898 pages and the [17]GCC manuals. If that fails, the
6899 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6900 web pages and the development of GCC are welcome on our developer
6901 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
6904 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
6905 distribution of this entire article is permitted in any medium,
6906 provided this notice is preserved.
6908 These pages are [22]maintained by the GCC team. Last modified
6913 1. http://www.gnu.org/
6914 2. http://gcc.gnu.org/gcc-3.4/changes.html
6915 3. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6916 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
6917 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5
6918 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4
6919 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3
6920 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2
6921 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1
6922 10. http://gcc.gnu.org/gcc-3.4/changes.html
6923 11. http://gcc.gnu.org/gcc-3.4/buildstat.html
6924 12. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6925 13. http://gcc.gnu.org/index.html
6926 14. mailto:gcc@gcc.gnu.org
6927 15. http://gcc.gnu.org/mirrors.html
6928 16. http://gcc.gnu.org/svn.html
6929 17. http://gcc.gnu.org/onlinedocs/
6930 18. mailto:gcc-help@gcc.gnu.org
6931 19. mailto:gcc@gcc.gnu.org
6932 20. http://gcc.gnu.org/lists.html
6933 21. http://www.fsf.org/
6934 22. http://gcc.gnu.org/about.html
6935 23. http://validator.w3.org/check/referer
6936 ======================================================================
6937 http://gcc.gnu.org/gcc-3.4/changes.html
6938 GCC 3.4 Release Series
6939 Changes, New Features, and Fixes
6941 The final release in the 3.4 release series is [1]GCC 3.4.6. The series
6944 GCC 3.4 has [2]many improvements in the C++ frontend. Before reporting
6945 a bug, please make sure it's really GCC, and not your code, that is
6950 * GNU Make is now required to build GCC.
6951 * With -nostdinc the preprocessor used to ignore both standard
6952 include paths and include paths contained in environment variables.
6953 It was neither documented nor intended that environment variable
6954 paths be ignored, so this has been corrected.
6955 * GCC no longer accepts the options -fvolatile, -fvolatile-global and
6956 -fvolatile-static. It is unlikely that they worked correctly in any
6958 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead.
6959 * Support for all the systems [3]obsoleted in GCC 3.3 has been
6960 removed from GCC 3.4. See below for a [4]list of systems which are
6961 obsoleted in this release.
6962 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C
6963 compilers will not work.
6964 * The implementation of the [5]MIPS ABIs has changed. As a result,
6965 the code generated for certain MIPS targets will not be binary
6966 compatible with earlier releases.
6967 * In previous releases, the MIPS port had a fake "hilo" register with
6968 the user-visible name accum. This register has been removed.
6969 * The implementation of the [6]SPARC ABIs has changed. As a result,
6970 the code generated will not be binary compatible with earlier
6971 releases in certain cases.
6972 * The configure option --enable-threads=pthreads has been removed;
6973 use --enable-threads=posix instead, which should have the same
6975 * Code size estimates used by inlining heuristics for C, Objective-C,
6976 C++ and Java have been redesigned significantly. As a result the
6977 parameters of -finline-insns, --param max-inline-insns-single and
6978 --param max-inline-insns-auto need to be reconsidered.
6979 * --param max-inline-slope and --param min-inline-insns have been
6980 removed; they are not needed for the new bottom-up inlining
6982 * The new unit-at-a-time compilation scheme has several compatibility
6984 + The order in which functions, variables, and top-level asm
6985 statements are emitted may have changed. Code relying on some
6986 particular ordering needs to be updated. The majority of such
6987 top-level asm statements can be replaced by section
6989 + Unreferenced static variables and functions are removed. This
6990 may result in undefined references when an asm statement
6991 refers to the variable/function directly. In that case either
6992 the variable/function shall be listed in asm statement operand
6993 or in the case of top-level asm statements the attribute used
6994 shall be used to force function/variable to be always output
6995 and considered as a possibly used by unknown code.
6996 For variables the attribute is accepted only by GCC 3.4 and
6997 newer, while for earlier versions it is sufficient to use
6998 unused to silence warnings about the variables not being
6999 referenced. To keep code portable across different GCC
7000 versions, you can use appropriate preprocessor conditionals.
7001 + Static functions now can use non-standard passing conventions
7002 that may break asm statements calling functions directly.
7003 Again the attribute used shall be used to prevent this
7005 As a temporary workaround, -fno-unit-at-a-time can be used, but
7006 this scheme may not be supported by future releases of GCC.
7007 * GCC 3.4 automatically places zero-initialized variables in the .bss
7008 section on some operating systems. Versions of GNU Emacs up to (and
7009 including) 21.3 will not work correctly when using this
7010 optimization; you can use -fno-zero-initialized-in-bss to disable
7012 * If GCC 3.4 is configured with --enable-threads=posix (the default
7013 on most targets that support pthreads) then _REENTRANT will be
7014 defined unconditionally by some libstdc++ headers. C++ code which
7015 relies on that macro to detect whether multi-threaded code is being
7016 compiled might change in meaning, possibly resulting in linker
7017 errors for single-threaded programs. Affected users of [7]Boost
7018 should compile single-threaded code with -DBOOST_DISABLE_THREADS.
7019 See Bugzilla for [8]more information.
7021 General Optimizer Improvements
7023 * Usability of the profile feedback and coverage testing has been
7025 + Performance of profiled programs has been improved by faster
7026 profile merging code.
7027 + Better use of the profile feedback for optimization (loop
7028 unrolling and loop peeling).
7029 + File locking support allowing fork() calls and parallel runs
7030 of profiled programs.
7031 + Coverage file format has been redesigned.
7032 + gcov coverage tool has been improved.
7033 + make profiledbootstrap available to build a faster compiler.
7034 Experiments made on i386 hardware showed an 11% speedup on -O0
7035 and a 7.5% speedup on -O2 compilation of a [9]large C++
7037 + New value profiling pass enabled via -fprofile-values
7038 + New value profile transformations pass enabled via -fvpt aims
7039 to optimize some code sequences by exploiting knowledge about
7040 value ranges or other properties of the operands. At the
7041 moment a conversion of expensive divisions into cheaper
7042 operations has been implemented.
7043 + New -fprofile-generate and -fprofile-use command-line options
7044 to simplify the use of profile feedback.
7045 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and
7046 Java which is enabled via -funit-at-a-time (and implied by -O2). In
7047 this scheme a whole file is parsed first and optimized later. The
7048 following basic inter-procedural optimizations are implemented:
7049 + Removal of unreachable functions and variables
7050 + Discovery of local functions (functions with static linkage
7051 whose address is never taken)
7052 + On i386, these local functions use register parameter passing
7054 + Reordering of functions in topological order of the call graph
7055 to enable better propagation of optimizing hints (such as the
7056 stack alignments needed by functions) in the back end.
7057 + Call graph based out-of-order inlining heuristics which allows
7058 to limit overall compilation unit growth (--param
7059 inline-unit-growth).
7060 Overall, the unit-at-a-time scheme produces a 1.3% improvement for
7061 the SPECint2000 benchmark on the i386 architecture (AMD Athlon
7063 * More realistic code size estimates used by inlining for C,
7064 Objective-C, C++ and Java. The growth of large functions can now be
7065 limited via --param large-function-insns and --param
7066 large-function-growth.
7067 * A new cfg-level loop optimizer pass replaces the old loop unrolling
7068 pass and adds two other loop transformations -- loop peeling and
7069 loop unswitching -- and also uses the profile feedback to limit
7070 code growth. (The three optimizations are enabled by
7071 -funroll-loops, -fpeel-loops and -funswitch-loops flags,
7073 The old loop unroller still can be enabled by -fold-unroll-loops
7074 and may produce better code in some cases, especially when the
7075 webizer optimization pass is not run.
7076 * A new web construction pass enabled via -fweb (and implied by -O3)
7077 improves the quality of register allocation, CSE, first scheduling
7078 pass and some other optimization passes by avoiding re-use of
7079 pseudo registers with non-overlapping live ranges. The pass almost
7080 always improves code quality but does make debugging difficult and
7081 thus is not enabled by default by -O2
7082 The pass is especially effective as cleanup after code duplication
7083 passes, such as the loop unroller or the tracer.
7084 * Experimental implementations of superblock or trace scheduling in
7085 the second scheduling pass can be enabled via
7086 -fsched2-use-superblocks and -fsched2-use-traces, respectively.
7088 New Languages and Language specific improvements
7092 * The Ada front end has been updated to include numerous bug fixes
7093 and enhancements. These include:
7094 + Improved project file support
7095 + Additional set of warnings about potential wrong code
7096 + Improved error messages
7097 + Improved code generation
7098 + Improved cross reference information
7100 + Better run-time check elimination
7101 + Better error recovery
7102 + More efficient implementation of unbounded strings
7103 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools,
7105 + New GNAT.xxxx packages (e.g. GNAT.Strings,
7106 GNAT.Exception_Action)
7108 + New -gnatS switch replacing gnatpsta
7109 + Implementation of new Ada features (in particular limited
7110 with, limited aggregates)
7114 * Precompiled headers are now supported. Precompiled headers can
7115 dramatically speed up compilation of some projects. There are some
7116 known defects in the current precompiled header implementation that
7117 will result in compiler crashes in relatively rare situations.
7118 Therefore, precompiled headers should be considered a "technology
7119 preview" in this release. Read the manual for details about how to
7120 use precompiled headers.
7121 * File handling in the preprocessor has been rewritten. GCC no longer
7122 gets confused by symlinks and hardlinks, and now has a correct
7123 implementation of #import and #pragma once. These two directives
7124 have therefore been un-deprecated.
7125 * The undocumented extension that allowed C programs to have a label
7126 at the end of a compound statement, which has been deprecated since
7127 GCC 3.0, has been removed.
7128 * The cast-as-lvalue extension has been removed for C++ and
7129 deprecated for C and Objective-C. In particular, code like this:
7137 is no longer accepted for C++ and will not be accepted for C and
7138 Objective-C in a future version.
7139 * The conditional-expression-as-lvalue extension has been deprecated
7140 for C and Objective-C. In particular, code like this:
7144 will not be accepted for C and Objective-C in a future version.
7145 * The compound-expression-as-lvalue extension has been deprecated for
7146 C and Objective-C. In particular, code like this:
7150 will not be accepted for C and Objective-C in a future version. A
7151 possible non-intrusive workaround is the following:
7154 * Several [10]built-in functions such as __builtin_popcount for
7155 counting bits, finding the highest and lowest bit in a word, and
7156 parity have been added.
7157 * The -fwritable-strings option has been deprecated and will be
7159 * Many C math library functions are now recognized as built-ins and
7161 * The C, C++, and Objective-C compilers can now handle source files
7162 written in any character encoding supported by the host C library.
7163 The default input character set is taken from the current locale,
7164 and may be overridden with the -finput-charset command line option.
7165 In the future we will add support for inline encoding markers.
7169 * G++ is now much closer to full conformance to the ISO/ANSI C++
7170 standard. This means, among other things, that a lot of invalid
7171 constructs which used to be accepted in previous versions will now
7172 be rejected. It is very likely that existing C++ code will need to
7173 be fixed. This document lists some of the most common issues.
7174 * A hand-written recursive-descent C++ parser has replaced the
7175 YACC-derived C++ parser from previous GCC releases. The new parser
7176 contains much improved infrastructure needed for better parsing of
7177 C++ source codes, handling of extensions, and clean separation
7178 (where possible) between proper semantics analysis and parsing. The
7179 new parser fixes many bugs that were found in the old parser.
7180 * You must now use the typename and template keywords to disambiguate
7181 dependent names, as required by the C++ standard.
7183 typedef int mytype_t;
7186 template <class T1> struct A {
7187 template <class T2> struct B {
7191 template <int N> void bar(void)
7193 // Use 'typename' to tell the parser that T1::mytype_t names
7194 // a type. This is needed because the name is dependent (in
7195 // this case, on template parameter T1).
7196 typename T1::mytype_t x;
7201 template <class T> void template_func(void)
7203 // Use 'template' to prefix member templates within
7204 // dependent types (a has type A<T>, which depends on
7205 // the template parameter T).
7207 a.template bar<0>();
7209 // Use 'template' to tell the parser that B is a nested
7210 // template class (dependent on template parameter T), and
7211 // 'typename' because the whole A<T>::B<int> is
7212 // the name of a type (again, dependent).
7213 typename A<T>::template B<int> b;
7217 void non_template_func(void)
7219 // Outside of any template class or function, no names can be
7220 // dependent, so the use of the keyword 'typename' and 'template'
7221 // is not needed (and actually forbidden).
7227 * In a template definition, unqualified names will no longer find
7228 members of a dependent base (as specified by [temp.dep]/3 in the
7229 C++ standard). For example,
7230 template <typename T> struct B {
7238 template <typename T> struct C : B<T> {
7243 n = 0; // ::n is modified
7244 g (); // ::g is called
7247 You must make the names dependent, e.g. by prefixing them with
7248 this->. Here is the corrected definition of C<T>::h,
7249 template <typename T> void C<T>::h ()
7256 As an alternative solution (unfortunately not backwards compatible
7257 with GCC 3.3), you may use using declarations instead of this->:
7258 template <typename T> struct C : B<T> {
7271 * In templates, all non-dependent names are now looked up and bound
7272 at definition time (while parsing the code), instead of later when
7273 the template is instantiated. For instance:
7276 template <int> struct A {
7277 static void bar(void){
7286 A<0>::bar(); // Calls foo(int), used to call foo(char).
7289 * In an explicit instantiation of a class template, you must use
7290 class or struct before the template-id:
7294 template A<0>; // error, not accepted anymore
7295 template class A<0>; // OK
7296 * The "named return value" and "implicit typename" extensions have
7298 * Default arguments in function types have been deprecated and will
7300 * ARM-style name-injection of friend declarations has been deprecated
7301 and will be removed. For example: struct S { friend void f(); };
7302 void g() { f(); } will not be accepted by future versions of G++;
7303 instead a declaration of "f" will need to be present outside of the
7305 * Covariant returns are implemented for all but varadic functions
7306 that require an adjustment.
7307 * When -pedantic is used, G++ now issues errors about spurious
7308 semicolons. For example,
7309 namespace N {}; // Invalid semicolon.
7310 void f() {}; // Invalid semicolon.
7311 * G++ no longer accepts attributes for a declarator after the
7312 initializer associated with that declarator. For example,
7313 X x(1) __attribute__((...));
7314 is no longer accepted. Instead, use:
7315 X x __attribute__((...)) (1);
7316 * Inside the scope of a template class, the name of the class itself
7317 can be treated as either a class or a template. So GCC used to
7318 accept the class name as argument of type template, and template
7319 template parameter. However this is not C++ standard compliant. Now
7320 the name is not treated as a valid template template argument
7321 unless you qualify the name by its scope. For example, the code
7322 below no longer compiles.
7323 template <template <class> class TT> class X {};
7324 template <class T> class Y {
7325 X<Y> x; // Invalid, Y is always a type template parameter.
7327 The valid code for the above example is
7328 X< ::Y> x; // Valid.
7329 (Notice the space between < and : to prevent GCC to interpret this
7330 as a digraph for [.)
7331 * Friend declarations that refer to template specializations are
7332 rejected if the template has not already been declared. For
7334 template <typename T>
7336 friend void f<> (C&);
7338 is rejected. You must first declare f as a template,
7339 template <typename T>
7341 * In case of friend declarations, every name used in the friend
7342 declaration must be accessible at the point of that declaration.
7343 Previous versions of G++ used to be less strict about this and
7344 allowed friend declarations for private class members, for example.
7345 See the ISO C++ Standard Committee's [11]defect report #209 for
7347 * Declaration of member functions of class templates as friends are
7348 supported. For example,
7349 template <typename T> struct A {
7353 template <typename T> friend void A<T>::f();
7355 * You must use template <> to introduce template specializations, as
7356 required by the standard. For example,
7357 template <typename T>
7361 is rejected. You must write,
7362 template <> struct S<int> {};
7363 * G++ used to accept code like this,
7366 void f(int i = g());
7369 This behavior is not mandated by the standard. Now G++ issues an
7370 error about this code. To avoid the error, you must move the
7371 declaration of g before the declaration of f. The default arguments
7372 for g must be visible at the point where it is called.
7373 * The C++ ABI Section 3.3.3 specifications for the array construction
7374 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return
7375 NULL when the allocator argument returns NULL. These changes are
7376 incorporated into the libstdc++ runtime library.
7377 * Using a name introduced by a typedef in a friend declaration or in
7378 an explicit instantiation is now rejected, as specified by the ISO
7383 friend class B; // error, no typedef name here
7384 friend B; // error, friend always needs class/struct/enum
7385 friend class A; // OK
7388 template <int> class Q {};
7390 template class R; // error, no typedef name here
7391 template class Q<0>; // OK
7392 * When allocating an array with a new expression, GCC used to allow
7393 parentheses around the type name. This is actually ill-formed and
7395 int* a = new (int)[10]; // error, not accepted anymore
7396 int* a = new int[10]; // OK
7397 * When binding an rvalue of class type to a reference, the copy
7398 constructor of the class must be accessible. For instance, consider
7406 A(const A&); // private copy ctor
7414 foo(A()); // error, copy ctor is not accessible
7415 foo(makeA()); // error, copy ctor is not accessible
7418 foo(a1); // OK, a1 is a lvalue
7420 This might be surprising at first sight, especially since most
7421 popular compilers do not correctly implement this rule ([12]further
7423 * When forming a pointer to member or a pointer to member function,
7424 access checks for class visibility (public, protected, private) are
7425 now performed using the qualifying scope of the name itself. This
7426 is better explained with an example:
7442 &A::pub_func; // OK, pub_func is accessible through A
7443 &A::prot_func; // error, cannot access prot_func through A
7444 &A::priv_func; // error, cannot access priv_func through A
7446 &B::pub_func; // OK, pub_func is accessible through B
7447 &B::prot_func; // OK, can access prot_func through B (within B)
7448 &B::priv_func; // error, cannot access priv_func through B
7452 Runtime Library (libstdc++)
7454 * Optimization work:
7455 + Streamlined streambuf, filebuf, separate synched with C
7456 Standard I/O streambuf.
7457 + All formatted I/O now uses cached locale information.
7458 + STL optimizations (memory/speed for list, red-black trees as
7459 used by sets and maps).
7460 + More use of GCC builtins.
7461 + String optimizations (avoid contention on
7462 increment/decrement-and-test of the reference count in the
7463 empty-string object, constructor from input_iterators
7465 * Static linkage size reductions.
7466 * Large File Support (files larger than 2 GB on 32-bit systems).
7467 * Wide character and variable encoding filebuf work (UTF-8, Unicode).
7468 * Generic character traits.
7469 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x,
7470 Solaris 2.7 and above, AIX 5.x, Irix 6.5.
7471 * The allocator class is now standard-conformant, and two additional
7472 extension allocators have been added, mt_alloc and
7474 * PCH support: -include bits/stdc++.h (2x compile speedup).
7475 * Rewrote __cxa_demangle with support for C++ style allocators.
7476 * New debug modes for STL containers and iterators.
7477 * Testsuite rewrite: five times as many tests, plus increasingly
7478 sophisticated tests, including I/O, MT, multi-locale, wide and
7480 * Use current versions of GNU "autotools" for build/configuration.
7484 * The Objective-C front end has been updated to include the numerous
7485 bug fixes and enhancements previously available only in Apple's
7486 version of GCC. These include:
7487 + Structured exception (@try... @catch... @finally, @throw) and
7488 synchronization (@synchronized) support. These are accessible
7489 via the -fobjc-exceptions switch; as of this writing, they may
7490 only be used in conjunction with -fnext-runtime on Mac OS X
7491 10.3 and later. See [13]Options Controlling Objective-C
7492 Dialect for more information.
7493 + An overhaul of @encode logic. The C99 _Bool and C++ bool type
7494 may now be encoded as 'B'. In addition, the back-end/codegen
7495 dependencies have been removed.
7496 + An overhaul of message dispatch construction, ensuring that
7497 the various receiver types (and casts thereof) are handled
7498 properly, and that correct diagnostics are issued.
7499 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue"
7500 (-freplace-objc-classes) debugging modes, currently available
7501 on Mac OS X 10.3 and later. See [14]Options Controlling
7502 Objective-C Dialect for more information.
7503 + Access to optimized runtime entry points (-fno-nil-receivers )
7504 on the assumption that message receivers are never nil. This
7505 is currently available on Mac OS X 10.3 and later. See
7506 [15]Options Controlling Objective-C Dialect for more
7511 * Compiling a .jar file will now cause non-.class entries to be
7512 automatically compiled as resources.
7513 * libgcj has been ported to Darwin.
7514 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code
7516 * libgcj has a new gcjlib URL type; this lets URLClassLoader load
7517 code from shared libraries.
7518 * libgcj has been much more completely merged with [16]GNU Classpath.
7519 * Class loading is now much more correct; in particular the caller's
7520 class loader is now used when that is required.
7521 * [17]Eclipse 2.x will run out of the box using gij.
7522 * Parts of java.nio have been implemented. Direct and indirect
7523 buffers work, as do fundamental file and socket operations.
7524 * java.awt has been improved, though it is still not ready for
7526 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST
7528 * The MinGW port has matured. Enhancements include socket timeout
7529 support, thread interruption, improved Runtime.exec() handling and
7530 support for accented characters in filenames.
7534 * Fortran improvements are listed in the [18]Fortran documentation.
7536 New Targets and Target Specific Improvements
7540 * Several [19]built-in functions have been added such as
7541 __builtin_alpha_zap to allow utilizing the more obscure
7542 instructions of the CPU.
7543 * Parameter passing of complex arguments has changed to match the
7544 [20]ABI. This change is incompatible with previous GCC versions,
7545 but does fix compatibility with the Tru64 compiler and several
7546 corner cases where GCC was incompatible with itself.
7550 * Nicolas Pitre has contributed his hand-coded floating-point support
7551 code for ARM. It is both significantly smaller and faster than the
7552 existing C-based implementation, even when building applications
7553 for Thumb. The arm-elf configuration has been converted to use the
7555 * Support for the Intel's iWMMXt architecture, a second generation
7556 XScale processor, has been added. Enabled at run time with the
7557 -mcpu=iwmmxt command line switch.
7558 * A new ARM target has been added: arm-wince-pe. This is similar to
7559 the arm-pe target, but it defaults to using the APCS32 ABI.
7560 * The existing ARM pipeline description has been converted to the use
7561 the [21]DFA processor pipeline model. There is not much change in
7562 code performance, but the description is now [22]easier to
7564 * Support for the Cirrus EP9312 Maverick floating point co-processor
7565 added. Enabled at run time with the -mcpu=ep9312 command line
7566 switch. Note however that the multilibs to support this chip are
7567 currently disabled in gcc/config/arm/t-arm-elf, so if you want to
7568 enable their production you will have to uncomment the entries in
7573 * Support for long long has been added.
7574 * Support for saveall attribute has been added.
7575 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code
7576 for H8/300H and H8S, which is much faster than the previous
7578 * A lot of small performance improvements.
7580 IA-32/AMD64 (x86-64)
7582 * Tuning for K8 (AMD Opteron/Athlon64) core is available via
7583 -march=k8 and -mcpu=k8.
7584 * Scalar SSE code generation carefully avoids reformatting penalties,
7585 hidden dependencies and minimizes the number of uops generated on
7586 both Intel and AMD CPUs.
7587 * Vector MMX and SSE operands are now passed in registers to improve
7588 performance and match the argument passing convention used by the
7589 Intel C++ Compiler. As a result it is not possible to call
7590 functions accepting vector arguments compiled by older GCC version.
7591 * Conditional jump elimination is now more aggressive on modern CPUs.
7592 * The Athlon ports has been converted to use the DFA processor
7593 pipeline description.
7594 * Optimization of indirect tail calls is now possible in a similar
7595 fashion as direct sibcall optimization.
7596 * Further small performance improvements.
7597 * -m128bit-long-double is now less buggy.
7598 * __float128 support in 64-bit compilation.
7599 * Support for data structures exceeding 2GB in 64-bit mode.
7600 * -mcpu has been renamed to -mtune.
7604 * Tuning code for the Itanium 2 processor has been added. The
7605 generation of code tuned for Itanium 2 (option -mtune=itanium2) is
7606 enabled by default now. To generate code tuned for Itanium 1 the
7607 option -mtune=itanium1 should be used.
7608 * [23]DFA processor pipeline descriptions for the IA-64 processors
7609 have been added. This resulted in about 3% improvement on the
7610 SPECInt2000 benchmark for Itanium 2.
7611 * Instruction bundling for the IA-64 processors has been rewritten
7612 using the DFA pipeline hazard recognizer. It resulted in about 60%
7613 compiler speedup on the SPECInt2000 C programs.
7617 * Support for the M32R/2 processor has been added by Renesas.
7618 * Support for an M32R GNU/Linux target and PIC code generation has
7619 been added by Renesas.
7623 * Bernardo Innocenti (Develer S.r.l.) has contributed the
7624 m68k-uclinux target, based on former work done by Paul Dale
7625 (SnapGear Inc.). Code generation for the ColdFire processors family
7626 has been enhanced and extended to support the MCF 53xx and MCF 54xx
7627 cores, integrating former work done by Peter Barada (Motorola).
7631 Processor-specific changes
7633 * Support for the RM7000 and RM9000 processors has been added. It can
7634 be selected using the -march compiler option and should work with
7635 any MIPS I (mips-*) or MIPS III (mips64-*) configuration.
7636 * Support for revision 2 of the MIPS32 ISA has been added. It can be
7637 selected with the command-line option -march=mips32r2.
7638 * There is a new option, -mfix-sb1, to work around certain SB-1
7643 * It is possible to customize GCC using the following configure-time
7645 + --with-arch, which specifies the default value of the -march
7647 + --with-tune, which specifies the default value of the -mtune
7649 + --with-abi, which specifies the default ABI.
7650 + --with-float=soft, which tells GCC to use software floating
7652 + --with-float=hard, which tells GCC to use hardware floating
7654 * A 64-bit GNU/Linux port has been added. The associated
7655 configurations are mips64-linux-gnu and mips64el-linux-gnu.
7656 * The 32-bit GNU/Linux port now supports Java.
7657 * The IRIX 6 configuration now supports the o32 ABI and will build
7658 o32 multilibs by default. This support is compatible with both
7659 binutils and the SGI tools, but note that several features,
7660 including debugging information and DWARF2 exception handling, are
7661 only available when using the GNU assembler. Use of the GNU
7662 assembler and linker (version 2.15 or above) is strongly
7664 * The IRIX 6 configuration now supports 128-bit long doubles.
7665 * There are two new RTEMS-specific configurations, mips-rtems and
7667 * There are two new *-elf configurations, mipsisa32r2-elf and
7672 * Several [24]ABI bugs have been fixed. Unfortunately, these changes
7673 will break binary compatibility with earlier releases.
7674 * GCC can now use explicit relocation operators when generating
7675 -mabicalls code. This behavior is controlled by -mexplicit-relocs
7676 and can have several performance benefits. For example:
7677 + It allows for more optimization of GOT accesses, including
7678 better scheduling and redundancy elimination.
7679 + It allows sibling calls to be implemented as jumps.
7680 + n32 and n64 leaf functions can use a call-clobbered global
7681 pointer instead of $28.
7682 + The code to set up $gp can be removed from functions that
7684 * A new option, -mxgot, allows the GOT to be bigger than 64k. This
7685 option is equivalent to the assembler's -xgot option and should be
7686 used instead of -Wa,-xgot.
7687 * Frame pointer elimination is now supported when generating 64-bit
7689 * Inline block moves have been optimized to take more account of
7690 alignment information.
7691 * Many internal changes have been made to the MIPS port, mostly aimed
7692 at reducing the reliance on assembler macros.
7696 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64
7697 [25]ABI incompatibilities regarding the way parameters are passed
7698 during functions calls. These changes may result in incompatibility
7699 between code compiled with GCC 3.3 and GCC 3.4.
7703 * Support for shared/dylib gcc libraries has been added. It is
7704 enabled by default on powerpc-apple-darwin7.0.0 and up.
7705 * Libgcj is enabled by default. On systems older than
7706 powerpc-apple-darwin7.0.0 you need to install dlcompat.
7707 * 128-bit IBM extended precision format support added for long
7712 * By default, PowerPC64 GNU/Linux now uses natural alignment of
7713 structure elements. The old four byte alignment for double, with
7714 special rules for a struct starting with a double, can be chosen
7715 with -malign-power. This change may result in incompatibility
7716 between code compiled with GCC 3.3 and GCC 3.4.
7717 * -mabi=altivec is now the default rather than -mabi=no-altivec.
7718 * 128-bit IBM extended precision format support added for long
7723 * New command-line options allow to specify the intended execution
7724 environment for generated code:
7725 + -mesa/-mzarch allows to specify whether to generate code
7726 running in ESA/390 mode or in z/Architecture mode (this is
7727 applicable to 31-bit code only).
7728 + -march allows to specify a minimum processor architecture
7729 level (g5, g6, z900, or z990).
7730 + -mtune allows to specify which processor to tune for.
7731 * It is possible to customize GCC using the following configure-time
7733 + --with-mode, which specifies whether to default to assuming
7734 ESA/390 or z/Architecture mode.
7735 + --with-arch, which specifies the default value of the -march
7737 + --with-tune, which specifies the default value of the -mtune
7739 * Support for the z990 processor has been added, and can be selected
7740 using -march=z990 or -mtune=z990. This includes instruction
7741 scheduling tuned for the superscalar instruction pipeline of the
7742 z990 processor as well as support for all new instructions provided
7743 by the long-displacement facility.
7744 * Support to generate 31-bit code optimized for zSeries processors
7745 (running in ESA/390 or in z/Architecture mode) has been added. This
7746 can be selected using -march=z900 and -mzarch respectively.
7747 * Instruction scheduling for the z900 and z990 processors now uses
7748 the DFA pipeline hazard recognizer.
7749 * GCC no longer generates code to maintain a stack backchain,
7750 previously used to generate stack backtraces for debugging
7751 purposes. As replacement that does not incur runtime overhead,
7752 DWARF-2 call frame information is provided by GCC; this is
7753 supported by GDB 6.1. The old behavior can be restored using the
7755 * The stack frame size of functions may now exceed 2 GB in 64-bit
7757 * A port for the 64-bit IBM TPF operating system has been added; the
7758 configuration is s390x-ibm-tpf. This configuration is supported as
7759 cross-compilation target only.
7760 * Various changes to improve the generated code have been
7761 implemented, including:
7762 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT
7763 instructions to significantly speed up many floating-point
7765 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL
7766 WITH BORROW instructions to speed up long long arithmetic.
7767 + GCC now uses the SEARCH STRING instruction to implement
7769 + In many cases, function call overhead for 31-bit code has been
7770 reduced by placing the literal pool after the function code
7771 instead of after the function prolog.
7772 + Register 14 is no longer reserved in 64-bit code.
7773 + Handling of global register variables has been improved.
7777 * The option -mflat is deprecated.
7778 * Support for large (> 2GB) frames has been added to the 64-bit port.
7779 * Several [26]ABI bugs have been fixed. Unfortunately, these changes
7780 will break binary compatibility with earlier releases.
7781 * The default debugging format has been switched from STABS to
7782 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already
7783 the default debugging format for 64-bit code on Solaris.
7787 * Support for the SH2E processor has been added. Enabled at run time
7788 with the -m2e command line switch, or at configure time by
7789 specifying sh2e as the machine part of the target triple.
7793 * Support for the Mitsubishi V850E1 processor has been added. This is
7794 a variant of the V850E processor with some additional debugging
7799 * Several ABI bugs have been fixed. Unfortunately, these changes
7800 break binary compatibility with earlier releases.
7801 + For big-endian processors, the padding of aggregate return
7802 values larger than a word has changed. If the size of an
7803 aggregate return value is not a multiple of 32 bits, previous
7804 versions of GCC inserted padding in the most-significant bytes
7805 of the first return value register. Aggregates larger than a
7806 word are now padded in the least-significant bytes of the last
7807 return value register used. Aggregates smaller than a word are
7808 still padded in the most-significant bytes. The return value
7809 padding has not changed for little-endian processors.
7810 + Function arguments with 16-byte alignment are now properly
7812 + The implementation of the va_list type has changed. A va_list
7813 value created by va_start from a previous release cannot be
7814 used with va_arg from this release, or vice versa.
7815 * More processor configuration options for Xtensa processors are
7817 + the ABS instruction is now optional;
7818 + the ADDX* and SUBX* instructions are now optional;
7819 + an experimental CONST16 instruction can be used to synthesize
7820 constants instead of loading them from constant pools.
7821 These and other Xtensa processor configuration options can no
7822 longer be enabled or disabled by command-line options; the
7823 processor configuration must be specified by the xtensa-config.h
7824 header file when building GCC. Additionally, the
7825 -mno-serialize-volatile option is no longer supported.
7829 Support for a number of older systems has been declared obsolete in GCC
7830 3.4. Unless there is activity to revive them, the next release of GCC
7831 will have their sources permanently removed.
7833 All configurations of the following processor architectures have been
7835 * Mitsubishi D30V, d30v-*
7836 * AT&T DSP1600 and DSP1610, dsp16xx-*
7839 Also, some individual systems have been obsoleted:
7841 + Support for generating code for operation in APCS/26 mode
7844 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively
7845 maintained and supported.)
7847 + MOSS, i?86-moss-msdos and i?86-*-moss*
7848 + NCR 3000 running System V r.4, i?86-ncr-sysv4*
7849 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and
7851 + GNU/Linux with a.out object format, i?86-linux*aout*
7852 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1*
7853 + Interix versions before Interix 3, i?86-*-interix
7854 + Mach microkernel, i?86-mach*
7855 + SCO UnixWare with UDK, i?86-*-udk*
7856 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]*
7857 + VSTa microkernel, i386-*-vsta
7858 * Motorola M68000 family
7859 + HPUX, m68k-hp-hpux* and m68000-hp-hpux*
7860 + NetBSD with a.out object format (before NetBSD 1.4),
7861 m68k-*-*-netbsd* except m68k-*-*-netbsdelf*
7862 + Generic System V r.4, m68k-*-sysv4*
7864 + Generic VAX, vax-*-* (This is generic VAX only; we have not
7865 obsoleted any VAX triples for specific operating systems.)
7867 Documentation improvements
7869 Other significant improvements
7871 * The build system has undergone several significant cleanups.
7872 Subdirectories will only be configured if they are being built, and
7873 all subdirectory configures are run from the make command. The top
7874 level has been autoconfiscated.
7875 * Building GCC no longer writes to its source directory. This should
7876 help those wishing to share a read-only source directory over NFS
7877 or build from a CD. The exceptions to this feature are if you
7878 configure with either --enable-maintainer-mode or
7879 --enable-generated-files-in-srcdir.
7880 * The -W warning option has been renamed to -Wextra, which is more
7881 easily understood. The older spelling will be retained for
7882 backwards compatibility.
7883 * Substantial improvements in compile time have been made,
7884 particularly for non-optimizing compilations.
7885 __________________________________________________________________
7891 A vast number of bugs have been fixed in 3.4.0, too many to publish a
7892 complete list here. [27]Follow this link to query the Bugzilla database
7893 for the list of over 900 bugs fixed in 3.4.0. This is the list of all
7894 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4
7896 __________________________________________________________________
7902 This section lists the problem reports (PRs) from GCC's bug tracking
7903 system that are known to be fixed in the 3.4.1 release. This list might
7904 not be complete (that is, it is possible that some PRs that have been
7905 fixed are not listed here).
7909 * [28]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler
7910 emitted - PIC related
7911 * [29]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf
7912 * [30]14760 A bug in configure.in prevents using both
7913 --program-suffix and --program-prefix
7914 * [31]14671 [hppa64] bootstrap fails: ICE in
7915 save_call_clobbered_regs, in caller_save.c
7916 * [32]15093 [alpha][Java] make bootstrap fails to configure libffi on
7918 * [33]15178 Solaris 9/x86 fails linking after stage 3
7920 Multi-platform internal compiler errors (ICEs)
7922 * [34]12753 (preprocessor) Memory corruption in preprocessor on bad
7924 * [35]13985 ICE in gcc.c-torture/compile/930621-1.c
7925 * [36]14810 (c++) tree check failures with invalid code involving
7927 * [37]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in
7929 * [38]15044 (c++) ICE on syntax error, template header
7930 * [39]15057 (c++) Compiling of conditional value throw constructs
7931 cause a segmentation violation
7932 * [40]15064 (c++) typeid of template parameter gives ICE
7933 * [41]15142 (c++) ICE when passing a string where a char* is expected
7934 in a throw statement
7935 * [42]15159 ICE in rtl_verify_flow_info_1
7936 * [43]15165 (c++) ICE in instantiate_template
7937 * [44]15193 Unary minus using pointer to V4SF vector causes
7938 -fforce-mem to exhaust all memory
7939 * [45]15209 (c++) Runs out of memory with packed structs
7940 * [46]15227 (c++) Trouble with invalid function definition
7941 * [47]15285 (c++) instantiate_type ICE when forming pointer to
7943 * [48]15299 (c++) ICE in resolve_overloaded_unification
7944 * [49]15329 (c++) ICE on constructor of member template
7945 * [50]15550 ICE in extract_insn, in recog.c
7946 * [51]15554 (c++) ICE in tsubst_copy, in cp/pt.c
7947 * [52]15640 (c++) ICE on invalid code in arg_assoc, in
7949 * [53]15666 [unit-at-a-time] Gcc abort on valid code
7950 * [54]15696 (c++) ICE with bad pointer-to-member code
7951 * [55]15701 (c++) ICE with friends and template template parameter
7952 * [56]15761 ICE in do_SUBST, in combine.c
7953 * [57]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops
7957 * [58]14538 All RTEMS targets broken for gnat
7961 * [59]12391 missing warning about assigning to an incomplete type
7962 * [60]14649 atan(1.0) should not be a constant expression
7963 * [61]15004 [unit-at-a-time] no warning for unused paramater in
7965 * [62]15749 --pedantic-errors behaves differently from --pedantic
7966 with C-compiler on GNU/Linux
7968 C++ compiler and library
7970 * [63]10646 non-const reference is incorrectly matched in a "const T"
7971 partial specialization
7972 * [64]12077 wcin.rdbuf()->in_avail() return value too high
7973 * [65]13598 enc_filebuf doesn't work
7974 * [66]14211 const_cast returns lvalue but should be rvalue
7975 * [67]14220 num_put::do_put() undesired float/double behavior
7976 * [68]14245 problem with user-defined allocators in std::basic_string
7977 * [69]14340 libstdc++ Debug mode: failure to convert iterator to
7979 * [70]14600 __gnu_cxx::stdio_sync_filebuf should expose internal
7981 * [71]14668 no warning anymore for reevaluation of declaration
7982 * [72]14775 LFS (large file support) tests missing
7983 * [73]14821 Duplicate namespace alias declaration should not conflict
7984 * [74]14930 Friend declaration ignored
7985 * [75]14932 cannot use offsetof to get offsets of array elements in
7987 * [76]14950 [non unit-at-a-time] always_inline does not mix with
7989 * [77]14962 g++ ignores #pragma redefine_extname
7990 * [78]14975 Segfault on low-level write error during imbue
7991 * [79]15002 Linewise stream input is unusably slow (std::string slow)
7992 * [80]15025 compiler accepts redeclaration of template as
7994 * [81]15046 [arm] Math functions misdetected by cross configuration
7995 * [82]15069 a bit test on a variable of enum type is miscompiled
7996 * [83]15074 g++ -lsupc++ still links against libstdc++
7997 * [84]15083 spurious "statement has no effect" warning
7998 * [85]15096 parse error with templates and pointer to const member
7999 * [86]15287 combination of operator[] and operator .* fails in
8001 * [87]15317 __attribute__ unused in first parameter of constructor
8003 * [88]15337 sizeof on incomplete type diagnostic
8004 * [89]15361 bitset<>::_Find_next fails
8005 * [90]15412 _GLIBCXX_ symbols symbols defined and used in different
8007 * [91]15427 valid code results in incomplete type error
8008 * [92]15471 Incorrect member pointer offsets in anonymous
8010 * [93]15503 nested template problem
8011 * [94]15507 compiler hangs while laying out union
8012 * [95]15542 operator & and template definitions
8013 * [96]15565 SLES9: leading + sign for unsigned int with showpos
8014 * [97]15625 friend defined inside a template fails to find static
8016 * [98]15629 Function templates, overloads, and friend name injection
8017 * [99]15742 'noreturn' attribute ignored in method of template
8019 * [100]15775 Allocator::pointer consistently ignored
8020 * [101]15821 Duplicate namespace alias within namespace rejected
8021 * [102]15862 'enum yn' fails (confict with undeclared builtin)
8022 * [103]15875 rejects pointer to member in template
8023 * [104]15877 valid code using templates and anonymous enums is
8025 * [105]15947 Puzzling error message for wrong destructor declaration
8027 * [106]16020 cannot copy __gnu_debug::bitset
8028 * [107]16154 input iterator concept too restrictive
8029 * [108]16174 deducing top-level consts
8033 * [109]14315 Java compiler is not parallel make safe
8037 * [110]15151 [g77] incorrect logical i/o in 64-bit mode
8041 * [111]7993 private variables cannot be shadowed in subclasses
8045 * [112]15228 useless copies of floating point operands
8046 * [113]15345 [non-unit-at-a-time] unreferenced nested inline
8047 functions not optimized away
8048 * [114]15945 Incorrect floating point optimization
8049 * [115]15526 ftrapv aborts on 0 * (-1)
8050 * [116]14690 Miscompiled POOMA tests
8051 * [117]15112 GCC generates code to write to unchanging memory
8055 * [118]15067 Minor glitch in the source of cpp
8057 Main driver program bugs
8059 * [119]1963 collect2 interprets -oldstyle_liblookup as -o
8062 x86-specific (Intel/AMD)
8064 * [120]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND*
8069 * [121]14782 GCC produces an unaligned data access at -O2
8070 * [122]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2
8071 * [123]15202 ICE in reload_cse_simplify_operands, in postreload.c
8075 * [124]14610 __float80 constants incorrectly emitted
8076 * [125]14813 init_array sections are initialized in the wrong order
8077 * [126]14857 GCC segfault on duplicated asm statement
8078 * [127]15598 Gcc 3.4 ICE on valid code
8079 * [128]15653 Gcc 3.4 ICE on valid code
8083 * [129]15189 wrong filling of delay slot with -march=mips1 -G0
8084 -mno-split-addresses -mno-explicit-relocs
8085 * [130]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as
8087 * [131]16144 Bogus reference to __divdf3 when -O1
8088 * [132]16176 Miscompilation of unaligned data in MIPS backend
8092 * [133]11591 ICE in gcc.dg/altivec-5.c
8093 * [134]12028 powerpc-eabispe produces bad sCOND operation
8094 * [135]14478 rs6000 geu/ltu patterns generate incorrect code
8095 * [136]14567 long double and va_arg complex args
8096 * [137]14715 Altivec stack layout may overlap gpr save with stack
8098 * [138]14902 (libstdc++) Stream checking functions fail when -pthread
8100 * [139]14924 Compiler ICE on valid code
8101 * [140]14960 -maltivec affects vector return with -mabi=no-altivec
8102 * [141]15106 vector varargs failure passing from altivec to
8103 non-altivec code for -m32
8104 * [142]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 &
8106 * [143]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx
8108 * [144]15662 Segmentation fault when an exception is thrown - even if
8109 try and catch are specified
8113 * [145]15054 Bad code due to overlapping stack temporaries
8117 * [146]15783 ICE with union assignment in 64-bit mode
8118 * [147]15626 GCC 3.4 emits "ld: warning: relocation error:
8123 * [148]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64
8124 * [149]14723 Backported -march=nocona from mainline
8125 * [150]15290 __float128 failed to pass to function properly
8127 Cygwin/Mingw32-specific
8129 * [151]15250 Option -mms-bitfields support on GCC 3.4 is not
8130 conformant to MS layout
8131 * [152]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe
8132 worker on windows32 targets
8134 Bugs specific to embedded processors
8136 * [153]8309 [m68k] -m5200 produces erroneous SImode set of short
8138 * [154]13250 [SH] Gcc code for rotation clobbers the register, but
8139 gcc continues to use the register as if it was not clobbered
8140 * [155]13803 [coldfire] movqi operand constraints too restrictivefor
8142 * [156]14093 [SH] ICE for code when using -mhitachi option in SH
8143 * [157]14457 [m6811hc] ICE with simple c++ source
8144 * [158]14542 [m6811hc] ICE on simple source
8145 * [159]15100 [SH] cc1plus got hang-up on
8146 libstdc++-v3/testsuite/abi_check.cc
8147 * [160]15296 [CRIS] Delayed branch scheduling causing invalid code on
8149 * [161]15396 [SH] ICE with -O2 -fPIC
8150 * [162]15782 [coldfire] m68k_output_mi_thunk emits wrong code for
8153 Testsuite problems (compiler not affected)
8155 * [163]11610 libstdc++ testcases 27_io/* don't work properly remotely
8156 * [164]15488 (libstdc++) possibly insufficient file permissions for
8157 executing test suite
8158 * [165]15489 (libstdc++) testsuite_files determined incorrectly
8162 * [166]13928 (libstdc++) no whatis info in some man pages generated
8164 * [167]14150 Ada documentation out of date
8165 * [168]14949 (c++) Need to document method visibility changes
8166 * [169]15123 libstdc++-doc: Allocators.3 manpage is empty
8167 __________________________________________________________________
8173 This section lists the problem reports (PRs) from GCC's bug tracking
8174 system that are known to be fixed in the 3.4.2 release. This list might
8175 not be complete (that is, it is possible that some PRs that have been
8176 fixed are not listed here).
8178 Bootstrap failures and issues
8180 * [170]16469 [mips-sgi-irix5.3] bootstrap fails in
8181 libstdc++-v3/testsuite
8182 * [171]16344 [hppa-linux-gnu] libstdc++'s PCH built by
8183 profiledbootstrap does not work with the built compiler
8184 * [172]16842 [Solaris/x86] mkheaders can not find mkheaders.conf
8186 Multi-platform internal compiler errors (ICEs)
8188 * [173]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in
8189 cp_parser_class_specifier, in cp/parser.c
8190 * [174]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
8191 * [175]15461 (c++) ICE due to NRV and inlining
8192 * [176]15890 (c++) ICE in c_expand_expr, in c-common.c
8193 * [177]16180 ICE: segmentation fault in RTL optimization
8194 * [178]16224 (c++) ICE in write_unscoped_name (template/namespace)
8195 * [179]16408 ICE: in delete_insn, in cfgrtl.c
8196 * [180]16529 (c++) ICE for: namespace-alias shall not be declared as
8197 the name of any other entity
8198 * [181]16698 (c++) ICE with exceptions and declaration of __cxa_throw
8199 * [182]16706 (c++) ICE in finish_member_declaration, in
8201 * [183]16810 (c++) Legal C++ program with cast gives ICE in
8203 * [184]16851 (c++) ICE when throwing a comma expression
8204 * [185]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c
8205 * [186]16904 (c++) ICE in finish_class_member_access_expr, in
8207 * [187]16905 (c++) ICE (segfault) with exceptions
8208 * [188]16964 (c++) ICE in cp_parser_class_specifier due to
8210 * [189]17068 (c++) ICE: tree check: expected class 'd', have 'x'
8211 (identifier_node) in dependent_template_p, in cp/pt.c
8215 * [190]16366 Preprocessor option -remap causes memory corruption
8219 * [191]15345 unreferenced nested inline functions not optimized away
8220 * [192]16590 Incorrect execution when compiling with -O2
8221 * [193]16693 Bitwise AND is lost when used within a cast to an enum
8222 of the same precision
8223 * [194]17078 Jump into if(0) substatement fails
8225 Problems in generated debug information
8227 * [195]13956 incorrect stabs for nested local variables
8231 * [196]16684 GCC should not warn about redundant redeclarations of
8234 C++ compiler and library
8236 * [197]12658 Thread safety problems in locale::global() and
8238 * [198]13092 g++ accepts invalid pointer-to-member conversion
8239 * [199]15320 Excessive memory consumption
8240 * [200]16246 Incorrect template argument deduction
8241 * [201]16273 Memory exhausted when using nested classes and virtual
8243 * [202]16401 ostringstream in gcc 3.4.x very slow for big data
8244 * [203]16411 undefined reference to
8245 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char>
8247 * [204]16489 G++ incorrectly rejects use of a null constant integral
8248 expression as a null constant pointer
8249 * [205]16618 offsetof fails with constant member
8250 * [206]16637 syntax error reported for valid input code
8251 * [207]16717 __attribute__((constructor)) broken in C++
8252 * [208]16813 compiler error in DEBUG version of range insertion
8254 * [209]16853 pointer-to-member initialization from incompatible one
8256 * [210]16889 ambiguity is not detected
8257 * [211]16959 Segmentation fault in ios_base::sync_with_stdio
8259 Java compiler and library
8261 * [212]7587 direct threaded interpreter not thread-safe
8262 * [213]16473 ServerSocket accept() leaks file descriptors
8263 * [214]16478 Hash synchronization deadlock with finalizers
8267 * [215]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c
8268 * [216]16974 could not split insn (ice in final_scan_insn, in
8273 * [217]16298 ICE in output_operand
8274 * [218]17113 ICE with SSE2 intrinsics
8278 * [219]14697 libstdc++ couldn't find 32bit libgcc_s
8282 * [220]15869 [mips64] No NOP after LW (with -mips1 -O0)
8283 * [221]16325 [mips64] value profiling clobbers gp on mips
8284 * [222]16357 [mipsisa64-elf] ICE copying 7 bytes between extern
8286 * [223]16380 [mips64] Use of uninitialised register after dbra
8288 * [224]16407 [mips64] Unaligned access to local variables
8289 * [225]16643 [mips64] verify_local_live_at_start ICE after
8290 crossjumping & cfgcleanup
8294 * [226]15927 THUMB -O2: strength-reduced iteration variable ends up
8296 * [227]15948 THUMB: ICE with non-commutative cbranch
8297 * [228]17019 THUMB: bad switch statement in md code for
8298 addsi3_cbranch_scratch
8302 * [229]16130 ICE on valid code: in bundling, in config/ia64/ia64.c
8304 * [230]16142 ICE on valid code: in bundling, in config/ia64/ia64.c
8306 * [231]16278 Gcc failed to build Linux kernel with -mtune=merced
8307 * [232]16414 ICE on valid code: typo in comparison of asm_noperands
8309 * [233]16445 ICE on valid code: don't count ignored insns
8310 * [234]16490 ICE (segfault) while compiling with -fprofile-use
8311 * [235]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS
8315 * [236]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x
8316 * [237]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions
8321 * [238]16199 ICE while compiling apache 2.0.49
8322 * [239]16416 -m64 doesn't imply -mcpu=v9 anymore
8323 * [240]16430 ICE when returning non-C aggregates larger than 16 bytes
8325 Bugs specific to embedded processors
8327 * [241]16379 [m32r] can't output large model function call of memcpy
8328 * [242]17093 [m32r] ICE with -msdata=use -O0
8329 * [243]17119 [m32r] ICE at switch case 0x8000
8333 * [244]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp
8335 Alpha Tru64-specific
8337 * [245]16210 libstdc++ gratuitously omits "long long" I/O
8339 Testsuite, documentation issues (compiler is not affected):
8341 * [246]15488 (libstdc++) possibly insufficient file permissions for
8342 executing test suite
8343 * [247]16250 ada/doctools runs makeinfo even in release tarball
8344 __________________________________________________________________
8348 This is the [248]list of problem reports (PRs) from GCC's bug tracking
8349 system that are known to be fixed in the 3.4.3 release. This list might
8350 not be complete (that is, it is possible that some PRs that have been
8351 fixed are not listed here).
8355 * [249]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1
8356 * [250]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold
8359 Internal compiler errors (ICEs) affecting multiple platforms
8361 * [251]13948 (java) GCJ segmentation fault while compiling GL4Java
8363 * [252]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
8364 * [253]16301 (c++) ICE when "strong" attribute is attached to a using
8366 * [254]16566 ICE with flexible arrays
8367 * [255]17023 ICE with nested functions in parameter declaration
8368 * [256]17027 ICE with noreturn function in loop at -O2
8369 * [257]17524 ICE in grokdeclarator, in cp/decl.c
8370 * [258]17826 (c++) ICE in cp_tree_equal
8372 C and optimization bugs
8374 * [259]15526 -ftrapv aborts on 0 * (-1)
8375 * [260]16999 #ident stopped working
8376 * [261]17503 quadratic behaviour in invalid_mode_change_p
8377 * [262]17581 Long long arithmetic fails inside a switch/case
8378 statement when compiled with -O2
8379 * [263]18129 -fwritable-strings doesn't work
8381 C++ compiler and library bugs
8383 * [264]10975 incorrect initial ostringstream::tellp()
8384 * [265]11722 Unbuffered filebuf::sgetn is slow
8385 * [266]14534 Unrecognizing static function as a template parameter
8386 when its return value is also templated
8387 * [267]15172 Copy constructor optimization in aggregate
8389 * [268]15786 Bad error message for frequently occuring error.
8390 * [269]16162 Rejects valid member-template-definition
8391 * [270]16612 empty basic_strings can't live in shared memory
8392 * [271]16715 std::basic_iostream is instantiated when used, even
8393 though instantiations are already contained in libstdc++
8394 * [272]16848 code in /ext/demangle.h appears broken
8395 * [273]17132 GCC fails to eliminate function template specialization
8396 when argument deduction fails
8397 * [274]17259 One more _S_leaf incorrectly qualified with _RopeRep::
8399 * [275]17327 use of `enumeral_type' in template type unification
8400 * [276]17393 "unused variable '._0'" warning with -Wall
8401 * [277]17501 Confusion with member templates
8402 * [278]17537 g++ not passing -lstdc++ to linker when all command line
8403 arguments are libraries
8404 * [279]17585 usage of unqualified name of static member from within
8406 * [280]17821 Poor diagnostic for using "." instead of "->"
8407 * [281]17829 wrong error: call of overloaded function is ambiguous
8408 * [282]17851 Misleading diagnostic for invalid function declarations
8409 with undeclared types
8410 * [283]17976 Destructor is called twice
8411 * [284]18020 rejects valid definition of enum value in template
8412 * [285]18093 bogus conflict in namespace aliasing
8413 * [286]18140 C++ parser bug when using >> in templates
8417 * [287]17541 data statements with double precision constants fail
8421 * [288]17853 -O2 ICE for MMX testcase
8425 * [289]17245 ICE compiling gsl-1.5 statistics/lag1.c
8429 * [290]17167 FATAL:Symbol L_foo$stub already defined.
8433 * [291]17277 could not catch an exception when specified -maix64
8437 * [292]17505 <cmath> calls acosf(), ceilf(), and other functions
8438 missing from system libraries
8442 * [293]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl
8446 * [294]17384 ICE with mode attribute on structures
8450 * [295]17770 No NOP after LWL with -mips1
8452 Other embedded target specific
8454 * [296]11476 [arc-elf] gcc ICE on newlib's vfprintf.c
8455 * [297]14064 [avr-elf] -fdata-sections triggers ICE
8456 * [298]14678 [m68hc11-elf] gcc ICE
8457 * [299]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__
8458 * [300]15790 [i686-coff] Alignment error building gcc with i686-coff
8460 * [301]15886 [SH] Miscompilation with -O2 -fPIC
8461 * [302]16884 [avr-elf] [fweb related] bug while initializing
8464 Bugs relating to debugger support
8466 * [303]13841 missing debug info for _Complex function arguments
8467 * [304]15860 [big-endian targets] No DW_AT_location debug info is
8468 emitted for formal arguments to a function that uses "register"
8471 Testsuite issues (compiler not affected)
8473 * [305]17465 Testsuite in libffi overrides LD_LIBRARY_PATH
8474 * [306]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH
8475 * [307]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit
8480 * [308]15498 typo in gcc manual: non-existing locale example en_UK,
8482 * [309]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap:
8483 document broken shell
8484 * [310]16406 USE_LD_AS_NEEDED undocumented
8485 __________________________________________________________________
8489 This is the [311]list of problem reports (PRs) from GCC's bug tracking
8490 system that are known to be fixed in the 3.4.4 release. This list might
8491 not be complete (that is, it is possible that some PRs that have been
8492 fixed are not listed here).
8493 __________________________________________________________________
8497 This is the [312]list of problem reports (PRs) from GCC's bug tracking
8498 system that are known to be fixed in the 3.4.5 release. This list might
8499 not be complete (that is, it is possible that some PRs that have been
8500 fixed are not listed here).
8504 * [313]24688 sco_math fixincl breaks math.h
8508 * [314]17188 struct Foo { } redefinition
8509 * [315]20187 wrong code for ((unsigned char)(unsigned long
8510 long)((a?a:1)&(a*b)))?0:1)
8511 * [316]21873 infinite warning loop on bad array initializer
8512 * [317]21899 enum definition accepts values to be overriden
8513 * [318]22061 ICE in find_function_data, in function.c
8514 * [319]22308 Failure to diagnose violation of constraint 6.516p2
8515 * [320]22458 ICE on missing brace
8516 * [321]22589 ICE casting to long long
8517 * [322]24101 Segfault with preprocessed source
8519 C++ compiler and library bugs
8521 * [323]10611 operations on vector mode not recognized in C++
8522 * [324]13377 unexpected behavior of namespace usage directive
8523 * [325]16002 Strange error message with new parser
8524 * [326]17413 local classes as template argument
8525 * [327]17609 spurious error message after using keyword
8526 * [328]17618 ICE in cp_convert_to_pointer, in cp/cvt.c
8527 * [329]18124 ICE with invalid template template parameter
8528 * [330]18155 typedef in template declaration not rejected
8529 * [331]18177 ICE with const_cast for undeclared variable
8530 * [332]18368 C++ error message regression
8531 * [333]16378 ICE when returning a copy of a packed member
8532 * [334]18466 int ::i; accepted
8533 * [335]18512 ICE on invalid usage of template base class
8534 * [336]18454 ICE when returning undefined type
8535 * [337]18738 typename not allowed with non-dependent qualified name
8536 * [338]18803 rejects access to operator() in template
8537 * [339]19004 ICE in uses_template_parms, in cp/pt.c
8538 * [340]19208 Spurious error about variably modified type
8539 * [341]18253 bad error message / ICE for invalid template parameter
8540 * [342]19608 ICE after friend function definition in local class
8541 * [343]19884 ICE on explicit instantiation of a non-template
8543 * [344]20153 ICE when C++ template function contains anonymous union
8544 * [345]20563 Infinite loop in diagnostic (and ice after error
8546 * [346]20789 ICE with incomplete type in template
8547 * [347]21336 Internal compiler error when using custom new operators
8548 * [348]21768 ICE in error message due to violation of coding
8550 * [349]21853 constness of pointer to data member ignored
8551 * [350]21903 Default argument of template function causes a
8553 * [351]21983 multiple diagnostics
8554 * [352]21987 New testsuite failure
8555 g++.dg/warn/conversion-function-1.C
8556 * [353]22153 ICE on invalid template specialization
8557 * [354]22172 Internal compiler error, seg fault.
8558 * [355]21286 filebuf::xsgetn vs pipes
8559 * [356]22233 ICE with wrong number of template parameters
8560 * [357]22508 ICE after invalid operator new
8561 * [358]22545 ICE with pointer to class member & user defined
8563 * [359]23528 Wrong default allocator in ext/hash_map
8564 * [360]23550 char_traits requirements/1.cc test bad math
8565 * [361]23586 Bad diagnostic for invalid namespace-name
8566 * [362]23624 ICE in invert_truthvalue, in fold-const.c
8567 * [363]23639 Bad error message: not a member of '<declaration error>'
8568 * [364]23797 ICE on typename outside template
8569 * [365]23965 Bogus error message: no matching function for call to
8571 * [366]24052 &#`label_decl' not supported by dump_expr#<expression
8573 * [367]24580 virtual base class cause exception not to be caught
8575 Problems in generated debug information
8577 * [368]24267 Bad DWARF for altivec vectors
8579 Optimizations issues
8581 * [369]17810 ICE in verify_local_live_at_start
8582 * [370]17860 Wrong generated code for loop with varying bound
8583 * [371]21709 ICE on compile-time complex NaN
8584 * [372]21964 broken tail call at -O2 or more
8585 * [373]22167 Strange optimization bug when using -Os
8586 * [374]22619 Compilation failure for real_const_1.f and
8588 * [375]23241 Invalid code generated for comparison of uchar to 255
8589 * [376]23478 Miscompilation due to reloading of a var that is also
8591 * [377]24470 segmentation fault in cc1plus when compiling with -O
8592 * [378]24950 ICE in operand_subword_force
8594 Precompiled headers problems
8596 * [379]14400 Cannot compile qt-x11-free-3.3.0
8597 * [380]14940 PCH largefile test fails on various platforms
8601 * [381]20239 ICE on empty preprocessed input
8602 * [382]15220 "gcc -E -MM -MG" reports missing system headers in
8607 * [383]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on
8612 * [384]21888 bootstrap failure with linker relaxation enabled
8616 * [385]15342 [arm-linux]: ICE in verify_local_live_at_start
8617 * [386]23985 Memory aliasing information incorrect in inlined memcpy
8621 * [387]16719 Illegal move of byte into address register causes
8626 * [388]21723 ICE while building libgfortran
8627 * [389]21841 -mhp-ld/-mgnu-ld documentation
8631 * [390]23644 IA-64 hardware models and configuration options
8633 * [391]24718 Shared libgcc not used for linking by default
8637 * [392]18421 ICE in reload_cse_simplify_operands, in postreload.c
8641 * [393]20621 ICE in change_address_1, in emit-rtl.c
8643 PowerPC and PowerPC64 specific
8645 * [394]18583 error on valid code: const
8646 __attribute__((altivec(vector__))) doesn't work in arrays
8647 * [395]20191 ICE in reload_cse_simplify_operands
8648 * [396]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined
8649 * [397]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set
8650 * [398]23404 gij trashes args of functions with more than 8 fp args
8651 * [399]23539 C & C++ compiler generating misaligned references
8652 regardless of compiler flags
8653 * [400]24102 floatdisf2_internal2 broken
8654 * [401]24465 -mminimal-toc miscompilation of __thread vars
8658 * [402]19933 Problem with define of HUGE_VAL in math_c99
8659 * [403]21889 Native Solaris assembler cannot grok DTP-relative debug
8664 * [404]19300 PCH failures on sparc-linux
8665 * [405]20301 Assembler labels have a leading "-"
8666 * [406]20673 C PCH testsuite assembly comparison failure
8668 x86 and x86_64 specific
8670 * [407]18582 ICE with arrays of type V2DF
8671 * [408]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2
8673 * [409]21716 ICE in reg-stack.c's swap_rtx_condition
8674 * [410]24315 amd64 fails -fpeephole2
8675 __________________________________________________________________
8679 This is the [411]list of problem reports (PRs) from GCC's bug tracking
8680 system that are known to be fixed in the 3.4.6 release. This list might
8681 not be complete (that is, it is possible that some PRs that have been
8682 fixed are not listed here).
8685 For questions related to the use of GCC, please consult these web
8686 pages and the [412]GCC manuals. If that fails, the
8687 [413]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8688 web pages and the development of GCC are welcome on our developer
8689 list at [414]gcc@gcc.gnu.org. All of [415]our lists have public
8692 Copyright (C) [416]Free Software Foundation, Inc. Verbatim copying and
8693 distribution of this entire article is permitted in any medium,
8694 provided this notice is preserved.
8696 These pages are [417]maintained by the GCC team. Last modified
8701 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
8702 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus
8703 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
8704 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems
8705 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html
8706 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
8707 7. http://www.boost.org/
8708 8. http://gcc.gnu.org/PR11953
8709 9. http://gcc.gnu.org/PR8361
8710 10. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins
8711 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209
8712 12. http://gcc.gnu.org/bugs/#cxx_rvalbind
8713 13. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
8714 14. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
8715 15. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
8716 16. http://www.gnu.org/software/classpath/
8717 17. http://www.eclipse.org/
8718 18. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html
8719 19. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html
8720 20. http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51A_HTML/ARH9MBTE/DTMNPLTN.HTM#normal-argument-list-structure
8721 21. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
8722 22. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html
8723 23. http://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
8724 24. http://gcc.gnu.org/gcc-3.4/mips-abi.html
8725 25. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html
8726 26. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
8727 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
8728 28. http://gcc.gnu.org/PR10129
8729 29. http://gcc.gnu.org/PR14576
8730 30. http://gcc.gnu.org/PR14760
8731 31. http://gcc.gnu.org/PR14671
8732 32. http://gcc.gnu.org/PR15093
8733 33. http://gcc.gnu.org/PR15178
8734 34. http://gcc.gnu.org/PR12753
8735 35. http://gcc.gnu.org/PR13985
8736 36. http://gcc.gnu.org/PR14810
8737 37. http://gcc.gnu.org/PR14883
8738 38. http://gcc.gnu.org/PR15044
8739 39. http://gcc.gnu.org/PR15057
8740 40. http://gcc.gnu.org/PR15064
8741 41. http://gcc.gnu.org/PR15142
8742 42. http://gcc.gnu.org/PR15159
8743 43. http://gcc.gnu.org/PR15165
8744 44. http://gcc.gnu.org/PR15193
8745 45. http://gcc.gnu.org/PR15209
8746 46. http://gcc.gnu.org/PR15227
8747 47. http://gcc.gnu.org/PR15285
8748 48. http://gcc.gnu.org/PR15299
8749 49. http://gcc.gnu.org/PR15329
8750 50. http://gcc.gnu.org/PR15550
8751 51. http://gcc.gnu.org/PR15554
8752 52. http://gcc.gnu.org/PR15640
8753 53. http://gcc.gnu.org/PR15666
8754 54. http://gcc.gnu.org/PR15696
8755 55. http://gcc.gnu.org/PR15701
8756 56. http://gcc.gnu.org/PR15761
8757 57. http://gcc.gnu.org/PR15829
8758 58. http://gcc.gnu.org/PR14538
8759 59. http://gcc.gnu.org/PR12391
8760 60. http://gcc.gnu.org/PR14649
8761 61. http://gcc.gnu.org/PR15004
8762 62. http://gcc.gnu.org/PR15749
8763 63. http://gcc.gnu.org/PR10646
8764 64. http://gcc.gnu.org/PR12077
8765 65. http://gcc.gnu.org/PR13598
8766 66. http://gcc.gnu.org/PR14211
8767 67. http://gcc.gnu.org/PR14220
8768 68. http://gcc.gnu.org/PR14245
8769 69. http://gcc.gnu.org/PR14340
8770 70. http://gcc.gnu.org/PR14600
8771 71. http://gcc.gnu.org/PR14668
8772 72. http://gcc.gnu.org/PR14775
8773 73. http://gcc.gnu.org/PR14821
8774 74. http://gcc.gnu.org/PR14930
8775 75. http://gcc.gnu.org/PR14932
8776 76. http://gcc.gnu.org/PR14950
8777 77. http://gcc.gnu.org/PR14962
8778 78. http://gcc.gnu.org/PR14975
8779 79. http://gcc.gnu.org/PR15002
8780 80. http://gcc.gnu.org/PR15025
8781 81. http://gcc.gnu.org/PR15046
8782 82. http://gcc.gnu.org/PR15069
8783 83. http://gcc.gnu.org/PR15074
8784 84. http://gcc.gnu.org/PR15083
8785 85. http://gcc.gnu.org/PR15096
8786 86. http://gcc.gnu.org/PR15287
8787 87. http://gcc.gnu.org/PR15317
8788 88. http://gcc.gnu.org/PR15337
8789 89. http://gcc.gnu.org/PR15361
8790 90. http://gcc.gnu.org/PR15412
8791 91. http://gcc.gnu.org/PR15427
8792 92. http://gcc.gnu.org/PR15471
8793 93. http://gcc.gnu.org/PR15503
8794 94. http://gcc.gnu.org/PR15507
8795 95. http://gcc.gnu.org/PR15542
8796 96. http://gcc.gnu.org/PR15565
8797 97. http://gcc.gnu.org/PR15625
8798 98. http://gcc.gnu.org/PR15629
8799 99. http://gcc.gnu.org/PR15742
8800 100. http://gcc.gnu.org/PR15775
8801 101. http://gcc.gnu.org/PR15821
8802 102. http://gcc.gnu.org/PR15862
8803 103. http://gcc.gnu.org/PR15875
8804 104. http://gcc.gnu.org/PR15877
8805 105. http://gcc.gnu.org/PR15947
8806 106. http://gcc.gnu.org/PR16020
8807 107. http://gcc.gnu.org/PR16154
8808 108. http://gcc.gnu.org/PR16174
8809 109. http://gcc.gnu.org/PR14315
8810 110. http://gcc.gnu.org/PR15151
8811 111. http://gcc.gnu.org/PR7993
8812 112. http://gcc.gnu.org/PR15228
8813 113. http://gcc.gnu.org/PR15345
8814 114. http://gcc.gnu.org/PR15945
8815 115. http://gcc.gnu.org/PR15526
8816 116. http://gcc.gnu.org/PR14690
8817 117. http://gcc.gnu.org/PR15112
8818 118. http://gcc.gnu.org/PR15067
8819 119. http://gcc.gnu.org/PR1963
8820 120. http://gcc.gnu.org/PR15717
8821 121. http://gcc.gnu.org/PR14782
8822 122. http://gcc.gnu.org/PR14828
8823 123. http://gcc.gnu.org/PR15202
8824 124. http://gcc.gnu.org/PR14610
8825 125. http://gcc.gnu.org/PR14813
8826 126. http://gcc.gnu.org/PR14857
8827 127. http://gcc.gnu.org/PR15598
8828 128. http://gcc.gnu.org/PR15653
8829 129. http://gcc.gnu.org/PR15189
8830 130. http://gcc.gnu.org/PR15331
8831 131. http://gcc.gnu.org/PR16144
8832 132. http://gcc.gnu.org/PR16176
8833 133. http://gcc.gnu.org/PR11591
8834 134. http://gcc.gnu.org/PR12028
8835 135. http://gcc.gnu.org/PR14478
8836 136. http://gcc.gnu.org/PR14567
8837 137. http://gcc.gnu.org/PR14715
8838 138. http://gcc.gnu.org/PR14902
8839 139. http://gcc.gnu.org/PR14924
8840 140. http://gcc.gnu.org/PR14960
8841 141. http://gcc.gnu.org/PR15106
8842 142. http://gcc.gnu.org/PR16026
8843 143. http://gcc.gnu.org/PR15191
8844 144. http://gcc.gnu.org/PR15662
8845 145. http://gcc.gnu.org/PR15054
8846 146. http://gcc.gnu.org/PR15783
8847 147. http://gcc.gnu.org/PR15626
8848 148. http://gcc.gnu.org/PR14326
8849 149. http://gcc.gnu.org/PR14723
8850 150. http://gcc.gnu.org/PR15290
8851 151. http://gcc.gnu.org/PR15250
8852 152. http://gcc.gnu.org/PR15551
8853 153. http://gcc.gnu.org/PR8309
8854 154. http://gcc.gnu.org/PR13250
8855 155. http://gcc.gnu.org/PR13803
8856 156. http://gcc.gnu.org/PR14093
8857 157. http://gcc.gnu.org/PR14457
8858 158. http://gcc.gnu.org/PR14542
8859 159. http://gcc.gnu.org/PR15100
8860 160. http://gcc.gnu.org/PR15296
8861 161. http://gcc.gnu.org/PR15396
8862 162. http://gcc.gnu.org/PR15782
8863 163. http://gcc.gnu.org/PR11610
8864 164. http://gcc.gnu.org/PR15488
8865 165. http://gcc.gnu.org/PR15489
8866 166. http://gcc.gnu.org/PR13928
8867 167. http://gcc.gnu.org/PR14150
8868 168. http://gcc.gnu.org/PR14949
8869 169. http://gcc.gnu.org/PR15123
8870 170. http://gcc.gnu.org/PR16469
8871 171. http://gcc.gnu.org/PR16344
8872 172. http://gcc.gnu.org/PR16842
8873 173. http://gcc.gnu.org/PR12608
8874 174. http://gcc.gnu.org/PR14492
8875 175. http://gcc.gnu.org/PR15461
8876 176. http://gcc.gnu.org/PR15890
8877 177. http://gcc.gnu.org/PR16180
8878 178. http://gcc.gnu.org/PR16224
8879 179. http://gcc.gnu.org/PR16408
8880 180. http://gcc.gnu.org/PR16529
8881 181. http://gcc.gnu.org/PR16698
8882 182. http://gcc.gnu.org/PR16706
8883 183. http://gcc.gnu.org/PR16810
8884 184. http://gcc.gnu.org/PR16851
8885 185. http://gcc.gnu.org/PR16870
8886 186. http://gcc.gnu.org/PR16904
8887 187. http://gcc.gnu.org/PR16905
8888 188. http://gcc.gnu.org/PR16964
8889 189. http://gcc.gnu.org/PR17068
8890 190. http://gcc.gnu.org/PR16366
8891 191. http://gcc.gnu.org/PR15345
8892 192. http://gcc.gnu.org/PR16590
8893 193. http://gcc.gnu.org/PR16693
8894 194. http://gcc.gnu.org/PR17078
8895 195. http://gcc.gnu.org/PR13956
8896 196. http://gcc.gnu.org/PR16684
8897 197. http://gcc.gnu.org/PR12658
8898 198. http://gcc.gnu.org/PR13092
8899 199. http://gcc.gnu.org/PR15320
8900 200. http://gcc.gnu.org/PR16246
8901 201. http://gcc.gnu.org/PR16273
8902 202. http://gcc.gnu.org/PR16401
8903 203. http://gcc.gnu.org/PR16411
8904 204. http://gcc.gnu.org/PR16489
8905 205. http://gcc.gnu.org/PR16618
8906 206. http://gcc.gnu.org/PR16637
8907 207. http://gcc.gnu.org/PR16717
8908 208. http://gcc.gnu.org/PR16813
8909 209. http://gcc.gnu.org/PR16853
8910 210. http://gcc.gnu.org/PR16889
8911 211. http://gcc.gnu.org/PR16959
8912 212. http://gcc.gnu.org/PR7587
8913 213. http://gcc.gnu.org/PR16473
8914 214. http://gcc.gnu.org/PR16478
8915 215. http://gcc.gnu.org/PR10695
8916 216. http://gcc.gnu.org/PR16974
8917 217. http://gcc.gnu.org/PR16298
8918 218. http://gcc.gnu.org/PR17113
8919 219. http://gcc.gnu.org/PR14697
8920 220. http://gcc.gnu.org/PR15869
8921 221. http://gcc.gnu.org/PR16325
8922 222. http://gcc.gnu.org/PR16357
8923 223. http://gcc.gnu.org/PR16380
8924 224. http://gcc.gnu.org/PR16407
8925 225. http://gcc.gnu.org/PR16643
8926 226. http://gcc.gnu.org/PR15927
8927 227. http://gcc.gnu.org/PR15948
8928 228. http://gcc.gnu.org/PR17019
8929 229. http://gcc.gnu.org/PR16130
8930 230. http://gcc.gnu.org/PR16142
8931 231. http://gcc.gnu.org/PR16278
8932 232. http://gcc.gnu.org/PR16414
8933 233. http://gcc.gnu.org/PR16445
8934 234. http://gcc.gnu.org/PR16490
8935 235. http://gcc.gnu.org/PR16683
8936 236. http://gcc.gnu.org/PR16195
8937 237. http://gcc.gnu.org/PR16239
8938 238. http://gcc.gnu.org/PR16199
8939 239. http://gcc.gnu.org/PR16416
8940 240. http://gcc.gnu.org/PR16430
8941 241. http://gcc.gnu.org/PR16379
8942 242. http://gcc.gnu.org/PR17093
8943 243. http://gcc.gnu.org/PR17119
8944 244. http://gcc.gnu.org/PR15928
8945 245. http://gcc.gnu.org/PR16210
8946 246. http://gcc.gnu.org/PR15488
8947 247. http://gcc.gnu.org/PR16250
8948 248. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3
8949 249. http://gcc.gnu.org/PR17369
8950 250. http://gcc.gnu.org/PR17850
8951 251. http://gcc.gnu.org/PR13948
8952 252. http://gcc.gnu.org/PR14492
8953 253. http://gcc.gnu.org/PR16301
8954 254. http://gcc.gnu.org/PR16566
8955 255. http://gcc.gnu.org/PR17023
8956 256. http://gcc.gnu.org/PR17027
8957 257. http://gcc.gnu.org/PR17524
8958 258. http://gcc.gnu.org/PR17826
8959 259. http://gcc.gnu.org/PR15526
8960 260. http://gcc.gnu.org/PR16999
8961 261. http://gcc.gnu.org/PR17503
8962 262. http://gcc.gnu.org/PR17581
8963 263. http://gcc.gnu.org/PR18129
8964 264. http://gcc.gnu.org/PR10975
8965 265. http://gcc.gnu.org/PR11722
8966 266. http://gcc.gnu.org/PR14534
8967 267. http://gcc.gnu.org/PR15172
8968 268. http://gcc.gnu.org/PR15786
8969 269. http://gcc.gnu.org/PR16162
8970 270. http://gcc.gnu.org/PR16612
8971 271. http://gcc.gnu.org/PR16715
8972 272. http://gcc.gnu.org/PR16848
8973 273. http://gcc.gnu.org/PR17132
8974 274. http://gcc.gnu.org/PR17259
8975 275. http://gcc.gnu.org/PR17327
8976 276. http://gcc.gnu.org/PR17393
8977 277. http://gcc.gnu.org/PR17501
8978 278. http://gcc.gnu.org/PR17537
8979 279. http://gcc.gnu.org/PR17585
8980 280. http://gcc.gnu.org/PR17821
8981 281. http://gcc.gnu.org/PR17829
8982 282. http://gcc.gnu.org/PR17851
8983 283. http://gcc.gnu.org/PR17976
8984 284. http://gcc.gnu.org/PR18020
8985 285. http://gcc.gnu.org/PR18093
8986 286. http://gcc.gnu.org/PR18140
8987 287. http://gcc.gnu.org/PR17541
8988 288. http://gcc.gnu.org/PR17853
8989 289. http://gcc.gnu.org/PR17245
8990 290. http://gcc.gnu.org/PR17167
8991 291. http://gcc.gnu.org/PR17277
8992 292. http://gcc.gnu.org/PR17505
8993 293. http://gcc.gnu.org/PR17684
8994 294. http://gcc.gnu.org/PR17384
8995 295. http://gcc.gnu.org/PR17770
8996 296. http://gcc.gnu.org/PR11476
8997 297. http://gcc.gnu.org/PR14064
8998 298. http://gcc.gnu.org/PR14678
8999 299. http://gcc.gnu.org/PR15583
9000 300. http://gcc.gnu.org/PR15790
9001 301. http://gcc.gnu.org/PR15886
9002 302. http://gcc.gnu.org/PR16884
9003 303. http://gcc.gnu.org/PR13841
9004 304. http://gcc.gnu.org/PR15860
9005 305. http://gcc.gnu.org/PR17465
9006 306. http://gcc.gnu.org/PR17469
9007 307. http://gcc.gnu.org/PR18138
9008 308. http://gcc.gnu.org/PR15498
9009 309. http://gcc.gnu.org/PR15747
9010 310. http://gcc.gnu.org/PR16406
9011 311. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4
9012 312. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5
9013 313. http://gcc.gnu.org/PR24688
9014 314. http://gcc.gnu.org/PR17188
9015 315. http://gcc.gnu.org/PR20187
9016 316. http://gcc.gnu.org/PR21873
9017 317. http://gcc.gnu.org/PR21899
9018 318. http://gcc.gnu.org/PR22061
9019 319. http://gcc.gnu.org/PR22208
9020 320. http://gcc.gnu.org/PR22458
9021 321. http://gcc.gnu.org/PR22589
9022 322. http://gcc.gnu.org/PR24101
9023 323. http://gcc.gnu.org/PR10611
9024 324. http://gcc.gnu.org/PR13377
9025 325. http://gcc.gnu.org/PR16002
9026 326. http://gcc.gnu.org/PR17413
9027 327. http://gcc.gnu.org/PR17609
9028 328. http://gcc.gnu.org/PR17618
9029 329. http://gcc.gnu.org/PR18124
9030 330. http://gcc.gnu.org/PR18155
9031 331. http://gcc.gnu.org/PR18177
9032 332. http://gcc.gnu.org/PR18368
9033 333. http://gcc.gnu.org/PR18378
9034 334. http://gcc.gnu.org/PR18466
9035 335. http://gcc.gnu.org/PR18512
9036 336. http://gcc.gnu.org/PR18545
9037 337. http://gcc.gnu.org/PR18738
9038 338. http://gcc.gnu.org/PR18803
9039 339. http://gcc.gnu.org/PR19004
9040 340. http://gcc.gnu.org/PR19208
9041 341. http://gcc.gnu.org/PR19253
9042 342. http://gcc.gnu.org/PR19608
9043 343. http://gcc.gnu.org/PR19884
9044 344. http://gcc.gnu.org/PR20153
9045 345. http://gcc.gnu.org/PR20563
9046 346. http://gcc.gnu.org/PR20789
9047 347. http://gcc.gnu.org/PR21336
9048 348. http://gcc.gnu.org/PR21768
9049 349. http://gcc.gnu.org/PR21853
9050 350. http://gcc.gnu.org/PR21903
9051 351. http://gcc.gnu.org/PR21983
9052 352. http://gcc.gnu.org/PR21987
9053 353. http://gcc.gnu.org/PR22153
9054 354. http://gcc.gnu.org/PR22172
9055 355. http://gcc.gnu.org/PR21286
9056 356. http://gcc.gnu.org/PR22233
9057 357. http://gcc.gnu.org/PR22508
9058 358. http://gcc.gnu.org/PR22545
9059 359. http://gcc.gnu.org/PR23528
9060 360. http://gcc.gnu.org/PR23550
9061 361. http://gcc.gnu.org/PR23586
9062 362. http://gcc.gnu.org/PR23624
9063 363. http://gcc.gnu.org/PR23639
9064 364. http://gcc.gnu.org/PR23797
9065 365. http://gcc.gnu.org/PR23965
9066 366. http://gcc.gnu.org/PR24052
9067 367. http://gcc.gnu.org/PR24580
9068 368. http://gcc.gnu.org/PR24267
9069 369. http://gcc.gnu.org/PR17810
9070 370. http://gcc.gnu.org/PR17860
9071 371. http://gcc/gnu.org/PR21709
9072 372. http://gcc.gnu.org/PR21964
9073 373. http://gcc.gnu.org/PR22167
9074 374. http://gcc.gnu.org/PR22619
9075 375. http://gcc.gnu.org/PR23241
9076 376. http://gcc.gnu.org/PR23478
9077 377. http://gcc.gnu.org/PR24470
9078 378. http://gcc.gnu.org/PR24950
9079 379. http://gcc.gnu.org/PR14400
9080 380. http://gcc.gnu.org/PR14940
9081 381. http://gcc.gnu.org/PR20239
9082 382. http://gcc.gnu.org/PR15220
9083 383. http://gcc.gnu.org/PR19275
9084 384. http://gcc.gnu.org/PR21888
9085 385. http://gcc.gnu.org/PR15342
9086 386. http://gcc.gnu.org/PR23985
9087 387. http://gcc.gnu.org/PR16719
9088 388. http://gcc.gnu.org/PR21723
9089 389. http://gcc.gnu.org/PR21841
9090 390. http://gcc.gnu.org/PR23644
9091 391. http://gcc.gnu.org/PR24718
9092 392. http://gcc.gnu.org/PR18421
9093 393. http://gcc.gnu.org/PR20621
9094 394. http://gcc.gnu.org/PR18583
9095 395. http://gcc.gnu.org/PR20191
9096 396. http://gcc.gnu.org/PR22083
9097 397. http://gcc.gnu.org/PR23070
9098 398. http://gcc.gnu.org/PR23404
9099 399. http://gcc.gnu.org/PR23539
9100 400. http://gcc.gnu.org/PR24102
9101 401. http://gcc.gnu.org/PR24465
9102 402. http://gcc.gnu.org/PR19933
9103 403. http://gcc.gnu.org/PR21889
9104 404. http://gcc.gnu.org/PR19300
9105 405. http://gcc.gnu.org/PR20301
9106 406. http://gcc.gnu.org/PR20673
9107 407. http://gcc.gnu.org/PR18582
9108 408. http://gcc.gnu.org/PR19340
9109 409. http://gcc.gnu.org/PR21716
9110 410. http://gcc.gnu.org/PR24315
9111 411. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6
9112 412. http://gcc.gnu.org/onlinedocs/
9113 413. mailto:gcc-help@gcc.gnu.org
9114 414. mailto:gcc@gcc.gnu.org
9115 415. http://gcc.gnu.org/lists.html
9116 416. http://www.fsf.org/
9117 417. http://gcc.gnu.org/about.html
9118 418. http://validator.w3.org/check/referer
9119 ======================================================================
9120 http://gcc.gnu.org/gcc-3.3/index.html
9121 GCC 3.3 Release Series
9125 The [1]GNU project and the GCC developers are pleased to announce the
9126 release of GCC 3.3.6.
9128 This release is a bug-fix release, containing fixes for regressions in
9129 GCC 3.3.5 relative to previous releases of GCC.
9131 This release is the last of the series 3.3.x.
9133 The GCC 3.3 release series includes numerous [2]new features,
9134 improvements, bug fixes, and other changes, thanks to an [3]amazing
9135 group of volunteers.
9140 May 3, 2005 ([4]changes)
9143 September 30, 2004 ([5]changes)
9146 May 31, 2004 ([6]changes)
9149 February 14, 2004 ([7]changes)
9152 October 16, 2003 ([8]changes)
9155 August 8, 2003 ([9]changes)
9158 May 14, 2003 ([10]changes)
9160 References and Acknowledgements
9162 GCC used to stand for the GNU C Compiler, but since the compiler
9163 supports several other languages aside from C, it now stands for the
9164 GNU Compiler Collection.
9166 A list of [11]successful builds is updated as new information becomes
9169 The GCC developers would like to thank the numerous people that have
9170 contributed new features, improvements, bug fixes, and other changes as
9171 well as test results to GCC. This [12]amazing group of volunteers is
9172 what makes GCC successful.
9174 For additional information about GCC please refer to the [13]GCC
9175 project web site or contact the [14]GCC development mailing list.
9177 To obtain GCC please use [15]our mirror sites, or our CVS server.
9180 For questions related to the use of GCC, please consult these web
9181 pages and the [16]GCC manuals. If that fails, the
9182 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9183 web pages and the development of GCC are welcome on our developer
9184 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
9187 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
9188 distribution of this entire article is permitted in any medium,
9189 provided this notice is preserved.
9191 These pages are [21]maintained by the GCC team. Last modified
9196 1. http://www.gnu.org/
9197 2. http://gcc.gnu.org/gcc-3.3/changes.html
9198 3. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9199 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
9200 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5
9201 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4
9202 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3
9203 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2
9204 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1
9205 10. http://gcc.gnu.org/gcc-3.3/changes.html
9206 11. http://gcc.gnu.org/gcc-3.3/buildstat.html
9207 12. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9208 13. http://gcc.gnu.org/index.html
9209 14. mailto:gcc@gcc.gnu.org
9210 15. http://gcc.gnu.org/mirrors.html
9211 16. http://gcc.gnu.org/onlinedocs/
9212 17. mailto:gcc-help@gcc.gnu.org
9213 18. mailto:gcc@gcc.gnu.org
9214 19. http://gcc.gnu.org/lists.html
9215 20. http://www.fsf.org/
9216 21. http://gcc.gnu.org/about.html
9217 22. http://validator.w3.org/check/referer
9218 ======================================================================
9219 http://gcc.gnu.org/gcc-3.3/changes.html
9220 GCC 3.3 Release Series
9221 Changes, New Features, and Fixes
9223 The latest release in the 3.3 release series is [1]GCC 3.3.6.
9227 * The preprocessor no longer accepts multi-line string literals. They
9228 were deprecated in 3.0, 3.1, and 3.2.
9229 * The preprocessor no longer supports the -A- switch when appearing
9230 alone. -A- followed by an assertion is still supported.
9231 * Support for all the systems [2]obsoleted in GCC 3.1 has been
9232 removed from GCC 3.3. See below for a [3]list of systems which are
9233 obsoleted in this release.
9234 * Checking for null format arguments has been decoupled from the rest
9235 of the format checking mechanism. Programs which use the format
9236 attribute may regain this functionality by using the new [4]nonnull
9237 function attribute. Note that all functions for which GCC has a
9238 built-in format attribute, an appropriate built-in nonnull
9239 attribute is also applied.
9240 * The DWARF (version 1) debugging format has been deprecated and will
9241 be removed in a future version of GCC. Version 2 of the DWARF
9242 debugging format will continue to be supported for the foreseeable
9244 * The C and Objective-C compilers no longer accept the "Naming Types"
9245 extension (typedef foo = bar); it was already unavailable in C++.
9246 Code which uses it will need to be changed to use the "typeof"
9247 extension instead: typedef typeof(bar) foo. (We have removed this
9248 extension without a period of deprecation because it has caused the
9249 compiler to crash since version 3.0 and no one noticed until very
9250 recently. Thus we conclude it is not in widespread use.)
9251 * The -traditional C compiler option has been removed. It was
9252 deprecated in 3.1 and 3.2. (Traditional preprocessing remains
9253 available.) The <varargs.h> header, used for writing variadic
9254 functions in traditional C, still exists but will produce an error
9256 * GCC 3.3.1 automatically places zero-initialized variables in the
9257 .bss section on some operating systems. Versions of GNU Emacs up to
9258 (and including) 21.3 will not work correctly when using this
9259 optimization; you can use -fno-zero-initialized-in-bss to disable
9262 General Optimizer Improvements
9264 * A new scheme for accurately describing processor pipelines, the
9265 [5]DFA scheduler, has been added.
9266 * Pavel Nejedly, Charles University Prague, has contributed new file
9267 format used by the edge coverage profiler (-fprofile-arcs).
9268 The new format is robust and diagnoses common mistakes where
9269 profiles from different versions (or compilations) of the program
9270 are combined resulting in nonsensical profiles and slow code to
9271 produced with profile feedback. Additionally this format allows
9272 extra data to be gathered. Currently, overall statistics are
9273 produced helping optimizers to identify hot spots of a program
9274 globally replacing the old intra-procedural scheme and resulting in
9275 better code. Note that the gcov tool from older GCC versions will
9276 not be able to parse the profiles generated by GCC 3.3 and vice
9278 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation
9279 pass enabled using -ftracer. This pass simplifies the control flow
9280 of functions allowing other optimizations to do better job.
9281 He also contributed the function reordering pass
9282 (-freorder-functions) to optimize function placement using profile
9285 New Languages and Language specific improvements
9289 * The preprocessor now accepts directives within macro arguments. It
9290 processes them just as if they had not been within macro arguments.
9291 * The separate ISO and traditional preprocessors have been completely
9292 removed. The front end handles either type of preprocessed output
9294 * In C99 mode preprocessor arithmetic is done in the precision of the
9295 target's intmax_t, as required by that standard.
9296 * The preprocessor can now copy comments inside macros to the output
9297 file when the macro is expanded. This feature, enabled using the
9298 -CC option, is intended for use by applications which place
9299 metadata or directives inside comments, such as lint.
9300 * The method of constructing the list of directories to be searched
9301 for header files has been revised. If a directory named by a -I
9302 option is a standard system include directory, the option is
9303 ignored to ensure that the default search order for system
9304 directories and the special treatment of system header files are
9306 * A few more [6]ISO C99 features now work correctly.
9307 * A new function attribute, nonnull, has been added which allows
9308 pointer arguments to functions to be specified as requiring a
9309 non-null value. The compiler currently uses this information to
9310 issue a warning when it detects a null value passed in such an
9312 * A new type attribute, may_alias, has been added. Accesses to
9313 objects with types with this attribute are not subjected to
9314 type-based alias analysis, but are instead assumed to be able to
9315 alias any other type of objects, just like the char type.
9319 * Type based alias analysis has been implemented for C++ aggregate
9324 * Generate an error if Objective-C objects are passed by value in
9325 function and method calls.
9326 * When -Wselector is used, check the whole list of selectors at the
9327 end of compilation, and emit a warning if a @selector() is not
9329 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
9330 * No longer need to include objc/objc-class.h to compile self calls
9331 in class methods (NeXT runtime only).
9332 * New -Wundeclared-selector option.
9333 * Removed selector bloating which was causing object files to be 10%
9334 bigger on average (GNU runtime only).
9335 * Using at run time @protocol() objects has been fixed in certain
9336 situations (GNU runtime only).
9337 * Type checking has been fixed and improved in many situations
9338 involving protocols.
9342 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK
9344 * The JDK 1.4 assert facility has been implemented.
9345 * The bytecode interpreter is now direct threaded and thus faster.
9349 * Fortran improvements are listed in [7]the Fortran documentation.
9353 * Ada tasking now works with glibc 2.3.x threading libraries.
9355 New Targets and Target Specific Improvements
9357 * The following changes have been made to the HP-PA port:
9358 + The port now defaults to scheduling for the PA8000 series of
9360 + Scheduling support for the PA7300 processor has been added.
9361 + The 32-bit port now supports weak symbols under HP-UX 11.
9362 + The handling of initializers and finalizers has been improved
9363 under HP-UX 11. The 64-bit port no longer uses collect2.
9364 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port.
9365 + ABI fixes to correct the passing of small structures by value.
9366 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to
9367 use the DFA processor pipeline description.
9368 * The following NetBSD configurations for the SuperH processor family
9370 + SH3, big-endian, sh-*-netbsdelf*
9371 + SH3, little-endian, shle-*-netbsdelf*
9372 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
9373 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
9374 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
9375 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
9376 * The following changes have been made to the IA-32/x86-64 port:
9377 + SSE2 and 3dNOW! intrinsics are now supported.
9378 + Support for thread local storage has been added to the IA-32
9380 + The x86-64 port has been significantly improved.
9381 * The following changes have been made to the MIPS port:
9382 + All configurations now accept the -mabi switch. Note that you
9383 will need appropriate multilibs for this option to work
9385 + ELF configurations will always pass an ABI flag to the
9386 assembler, except when the MIPS EABI is selected.
9387 + -mabi=64 no longer selects MIPS IV code.
9388 + The -mcpu option, which was deprecated in 3.1 and 3.2, has
9389 been removed from this release.
9390 + -march now changes the core ISA level. In previous releases,
9391 it would change the use of processor-specific extensions, but
9392 would leave the core ISA unchanged. For example, mips64-elf
9393 -march=r8000 will now generate MIPS IV code.
9394 + Under most configurations, -mipsN now acts as a synonym for
9396 + There are some new preprocessor macros to describe the -march
9397 and -mtune settings. See the documentation of those options
9399 + Support for the NEC VR-Series processors has been added. This
9400 includes the 54xx, 5500, and 41xx series.
9401 + Support for the Sandcraft sr71k processor has been added.
9402 * The following changes have been made to the S/390 port:
9403 + Support to build the Java runtime libraries has been added.
9404 Java is now enabled by default on s390-*-linux* and
9405 s390x-*-linux* targets.
9406 + Multilib support for the s390x-*-linux* target has been added;
9407 this allows to build 31-bit binaries using the -m31 option.
9408 + Support for thread local storage has been added.
9409 + Inline assembler code may now use the 'Q' constraint to
9410 specify memory operands without index register.
9411 + Various platform-specific performance improvements have been
9412 implemented; in particular, the compiler now uses the BRANCH
9413 ON COUNT family of instructions and makes more frequent use of
9414 the TEST UNDER MASK family of instructions.
9415 * The following changes have been made to the PowerPC port:
9416 + Support for IBM Power4 processor added.
9417 + Support for Motorola e500 SPE added.
9418 + Support for AIX 5.2 added.
9419 + Function and Data sections now supported on AIX.
9420 + Sibcall optimizations added.
9421 * The support for H8 Tiny is added to the H8/300 port with -mn.
9425 Support for a number of older systems has been declared obsolete in GCC
9426 3.3. Unless there is activity to revive them, the next release of GCC
9427 will have their sources permanently removed.
9429 All configurations of the following processor architectures have been
9431 * Matsushita MN10200, mn10200-*-*
9432 * Motorola 88000, m88k-*-*
9433 * IBM ROMP, romp-*-*
9435 Also, some individual systems have been obsoleted:
9437 + Interix, alpha*-*-interix*
9438 + Linux libc1, alpha*-*-linux*libc1*
9439 + Linux ECOFF, alpha*-*-linux*ecoff*
9441 + Generic a.out, arm*-*-aout*
9442 + Conix, arm*-*-conix*
9443 + "Old ABI," arm*-*-oabi
9444 + StrongARM/COFF, strongarm-*-coff*
9446 + Generic OSF, hppa1.0-*-osf*
9447 + Generic BSD, hppa1.0-*-bsd*
9448 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
9449 + HiUX, hppa*-*-hiux*
9450 + Mach Lites, hppa*-*-lites*
9452 + Windows NT 3.x, i?86-*-win32
9454 + HP systems, m68000-hp-bsd* and m68k-hp-bsd*
9455 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and
9457 + AT&T systems, m68000-att-sysv*
9458 + Atari systems, m68k-atari-sysv*
9459 + Motorola systems, m68k-motorola-sysv*
9460 + NCR systems, m68k-ncr-sysv*
9461 + Plexus systems, m68k-plexus-sysv*
9462 + Commodore systems, m68k-cbm-sysv*
9463 + Citicorp TTI, m68k-tti-*
9464 + Unos, m68k-crds-unos*
9465 + Concurrent RTU, m68k-ccur-rtu*
9466 + Linux a.out, m68k-*-linux*aout*
9467 + Linux libc1, m68k-*-linux*libc1*
9468 + pSOS, m68k-*-psos*
9470 + Generic ECOFF, mips*-*-ecoff*
9471 + SINIX, mips-sni-sysv4
9472 + Orion RTEMS, mips64orion-*-rtems*
9473 * National Semiconductor 32000
9474 + OpenBSD, ns32k-*-openbsd*
9475 * POWER (aka RS/6000) and PowerPC
9476 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
9477 + Bull BOSX, rs6000-bull-bosx
9478 + Generic Mach, rs6000-*-mach*
9479 + Generic SysV, powerpc*-*-sysv*
9480 + Linux libc1, powerpc*-*-linux*libc1*
9482 + Generic a.out, sparc-*-aout*, sparclet-*-aout*,
9483 sparclite-*-aout*, and sparc86x-*-aout*
9484 + NetBSD a.out, sparc-*-netbsd*aout*
9485 + Generic BSD, sparc-*-bsd*
9486 + ChorusOS, sparc-*-chorusos*
9487 + Linux a.out, sparc-*-linux*aout*
9488 + Linux libc1, sparc-*-linux*libc1*
9489 + LynxOS, sparc-*-lynxos*
9490 + Solaris on HAL hardware, sparc-hal-solaris2*
9491 + SunOS versions 3 and 4, sparc-*-sunos[34]*
9493 + RTEMS, v850-*-rtems*
9497 Documentation improvements
9499 Other significant improvements
9501 * Almost all front-end dependencies in the compiler have been
9502 separated out into a set of language hooks. This should make adding
9503 a new front end clearer and easier.
9504 * One effect of removing the separate preprocessor is a small
9505 increase in the robustness of the compiler in general, and the
9506 maintainability of target descriptions. Previously target-specific
9507 built-in macros and others, such as __FAST_MATH__, had to be
9508 handled with so-called specs that were hard to maintain. Often they
9509 would fail to behave properly when conflicting options were
9510 supplied on the command line, and define macros in the user's
9511 namespace even when strict ISO compliance was requested.
9512 Integrating the preprocessor has cleanly solved these issues.
9513 * The Makefile suite now supports redirection of make install by
9514 means of the variable DESTDIR.
9515 __________________________________________________________________
9519 Detailed release notes for the GCC 3.3 release follow.
9525 * [8]10140 cross compiler build failures: missing __mempcpy (DUP:
9528 Internal compiler errors (multi-platform)
9530 * [11]3581 large string causes segmentation fault in cc1
9531 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler
9532 * [13]5533 (c++) ICE when processing std::accumulate(begin, end,
9534 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
9535 * [15]6412 (c++) ICE in retrieve_specialization
9536 * [16]6620 (c++) partial template specialization causes an ICE
9537 (segmentation fault)
9538 * [17]6663 (c++) ICE with attribute aligned
9539 * [18]7068 ICE with incomplete types
9540 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation
9541 * [20]7647 (c++) ICE when data member has the name of the enclosing
9543 * [21]7675 ICE in fixup_var_refs_1
9544 * [22]7718 'complex' template instantiation causes ICE
9545 * [23]8116 (c++) ICE in member template function
9546 * [24]8358 (ada) Ada compiler accesses freed memory, crashes
9547 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation
9549 * [26]8564 (c++) ICE in find_function_data, in function.c
9550 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
9551 * [28]8766 (c++) ICE after failed initialization of static template
9553 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c
9554 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
9555 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class
9557 * [32]9216 (c++) ICE on missing template parameter
9558 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c
9559 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO
9561 * [35]9429 (c++) ICE in template instantiation with a pointered new
9563 * [36]9516 Internal error when using a big array
9564 * [37]9600 (c++) ICE with typedefs in template class
9565 * [38]9629 (c++) virtual inheritance segfault
9566 * [39]9672 (c++) ICE: Error reporting routines re-entered
9567 * [40]9749 (c++) ICE in write_expression on invalid function
9569 * [41]9794 (fortran) ICE: floating point exception during constant
9571 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE
9572 * [43]9916 (c++) ICE with noreturn function in ?: statement
9573 * [44]9936 ICE with local function and variable-length 2d array
9574 * [45]10262 (c++) cc1plus crashes with large generated code
9575 * [46]10278 (c++) ICE in parser for invalid code
9576 * [47]10446 (c++) ICE on definition of nonexistent member function of
9577 nested class in a class template
9578 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable
9580 * [49]10506 (c++) ICE in build_new at cp/init.c with
9581 -fkeep-inline-functions and multiple inheritance
9582 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the
9583 precision of the declared type
9587 * [51]2001 Inordinately long compile times in reload CSE regs
9588 * [52]2391 Exponential compilation time explosion in combine
9589 * [53]2960 Duplicate loop conditions even with -Os
9590 * [54]4046 redundant conditional branch
9591 * [55]6405 Loop-unrolling related performance regressions
9592 * [56]6798 very long compile time with large case-statement
9593 * [57]6871 const objects shouldn't be moved to .bss
9594 * [58]6909 problem w/ -Os on modified loop-2c.c test case
9595 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of
9596 non-void function'' warning
9597 * [60]7642 optimization problem with signbit()
9598 * [61]8634 incorrect code for inlining of memcpy under -O2
9599 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as
9600 regular function call
9604 * [63]2161 long if-else cascade overflows parser stack
9605 * [64]4319 short accepted on typedef'd char
9606 * [65]8602 incorrect line numbers in warning messages when using
9608 * [66]9177 -fdump-translation-unit: C front end deletes function_decl
9609 AST nodes and breaks debugging dumps
9610 * [67]9853 miscompilation of non-constant structure initializer
9612 c++ compiler and library
9614 * [68]45 legal template specialization code is rejected (DUP:
9616 * [70]764 lookup failure: friend operator and dereferencing a pointer
9617 and templates (DUP: [71]5116)
9618 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP:
9620 * [73]3663 G++ doesn't check access control during template
9622 * [74]3797 gcc fails to emit explicit specialization of a template
9624 * [75]3948 Two destructors are called when no copy destructor is
9625 defined (ABI change)
9626 * [76]4137 Conversion operator within template is not accepted
9627 * [77]4361 bogus ambiguity taking the address of a member template
9628 * [78]4802 g++ accepts illegal template code (access to private
9629 member; DUP: [79]5837)
9630 * [80]4803 inline function is used but never defined, and g++ does
9632 * [81]5094 Partial specialization cannot be friend?
9633 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
9634 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run
9636 * [84]7015 certain __asm__ constructs rejected
9637 * [85]7086 compile time regression (quadratic behavior in
9639 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and
9641 * [87]7247 copy constructor missing when inlining enabled (invalid
9643 * [88]7441 string array initialization compilation time regression
9644 from seconds to minutes
9645 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong
9646 * [90]7804 bad printing of floating point constant in warning message
9647 * [91]8099 Friend classes and template specializations
9648 * [92]8117 member function pointers and multiple inheritance
9649 * [93]8205 using declaration and multiple inheritance
9650 * [94]8645 unnecessary non-zero checks in stl_tree.h
9651 * [95]8724 explicit destructor call for incomplete class allowed
9652 * [96]8805 compile time regression with many member variables
9653 * [97]8691 -O3 and -fno-implicit-templates are incompatible
9654 * [98]8700 unhelpful error message for binding temp to reference
9655 * [99]8724 explicit destructor call for incomplete class allowed
9656 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems
9657 * [101]9016 Failure to consistently constant fold "constant" C++
9659 * [102]9053 g++ confused about ambiguity of overloaded function
9661 * [103]9152 undefined virtual thunks
9662 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out
9663 * [105]9297 data corruption due to codegen bug (when copying.)
9664 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken
9665 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf
9666 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in
9668 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters
9669 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439)
9670 * [112]9474 GCC freezes in compiling a weird code mixing <iostream>
9672 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1)
9674 * [115]9555 ostream inserters fail to set badbit on exception
9675 * [116]9561 ostream inserters rethrow exception of wrong type
9676 * [117]9563 ostream::sentry returns true after a failed preparation
9677 * [118]9582 one-definition rule violation in std::allocator
9678 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors
9679 * [120]9683 bug in initialization chains for static const variables
9680 from template classes
9681 * [121]9791 -Woverloaded-virtual reports hiding of destructor
9682 * [122]9817 collate::compare doesn't handle nul characters
9683 * [123]9825 filebuf::sputbackc breaks sbumpc
9684 * [124]9826 operator>>(basic_istream, basic_string) fails to compile
9686 * [125]9924 Multiple using statements for builtin functions not
9688 * [126]9946 destructor is not called for temporary object
9689 * [127]9964 filebuf::close() sometimes fails to close file
9690 * [128]9988 filebuf::overflow writes EOF to file
9691 * [129]10033 optimization breaks polymorphic references w/ typeid
9693 * [130]10097 filebuf::underflow drops characters
9694 * [131]10132 filebuf destructor can throw exceptions
9695 * [132]10180 gcc fails to warn about non-inlined function
9696 * [133]10199 method parametrized by template does not work everywhere
9697 * [134]10300 use of array-new (nothrow) in segfaults on NULL return
9698 * [135]10427 Stack corruption with variable-length automatic arrays
9699 and virtual destructors
9700 * [136]10503 Compilation never stops in fixed_type_or_null
9704 * [137]5956 selectors aren't matched properly when added to the
9707 Fortran compiler and library
9709 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't
9711 * [139]3924 g77 generates code that is rejected by GAS if COFF debug
9713 * [140]5634 doc: explain that configure --prefix=~/... does not work
9714 * [141]6367 multiple repeat counts confuse namelist read into array
9715 * [142]6491 Logical operations error on logicals when using
9717 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C"
9718 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os
9720 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should
9721 assume a direct access file
9722 * [146]7278 g77 "bug"; the executable misbehaves (with -O2
9724 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows
9725 * [148]7388 Incorrect output with 0-based array of characters
9726 * [149]8587 Double complex zero ** double precision number -> NaN
9728 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning:
9729 unknown register name line-length-none
9730 * [151]10197 Direct access files not unformatted by default
9732 Java compiler and library
9734 * [152]6005 gcj fails to build rhug on alpha
9735 * [153]6389 System.getProperty("") should always throw an
9736 IllegalArgumentException
9737 * [154]6576 java.util.ResourceBundle.getResource ignores locale
9738 * [155]6652 new java.io.File("").getCanonicalFile() throws exception
9739 * [156]7060 getMethod() doesn't search super interface
9740 * [157]7073 bytecode interpreter gives wrong answer for interface
9742 * [158]7180 possible bug in
9743 javax.naming.spi.NamingManager.getPlusPath()
9744 * [159]7416 java.security startup refs "GNU libgcj.security"
9745 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent
9746 env (DUP: [161]7578)
9747 * [162]7611 Internal error while compiling libjava with -O
9748 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry
9749 * [164]7766 ZipInputStream.available returns 0 immediately after
9751 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public
9752 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
9753 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens'
9754 * [168]8234 ZipInputStream chokes when InputStream.read() returns
9756 * [169]8415 reflection bug: exception info for Method
9757 * [170]8481 java.Random.nextInt(int) may return negative
9758 * [171]8593 Error reading GZIPped files with BufferedReader
9759 * [172]8759 java.beans.Introspector has no flushCaches() or
9760 flushFromCaches() methods
9761 * [173]8997 spin() calls Thread.sleep
9762 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd
9763 instead of the root content of C:
9764 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns
9766 * [176]9271 Severe bias in java.security.SecureRandom
9768 Ada compiler and library
9770 * [177]6767 make gnatlib-shared fails on -laddr2line
9771 * [178]9911 gnatmake fails to link when GCC configured with
9772 --with-sjlj-exceptions=yes
9773 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled
9774 * [180]10546 Ada tasking not working on Red Hat 9
9778 * [181]7029 preprocessor should ignore #warning with -M
9782 * [182]2903 [arm] Optimization bug with long long arithmetic
9783 * [183]7873 arm-linux-gcc fails when assigning address to a bit field
9787 * [184]7680 float functions undefined in math.h/cmath with #define
9790 HP-UX or HP-PA-specific
9792 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
9793 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to
9795 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen
9799 * [188]6744 Bad assembler code generated: reference to pseudo
9801 * [189]7361 Internal compiler error in reload_cse_simplify_operands,
9806 * [190]9496 [mips-linux] bug in optimizer?
9810 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of
9812 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux
9813 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg
9814 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c
9818 * [195]10267 (documentation) Wrong build instructions for
9821 x86-specific (Intel/AMD)
9823 * [196]7916 ICE in instantiate_virtual_register_1
9824 * [197]7926 (c++) i486 instructions in header files make c++ programs
9826 * [198]8555 ICE in gen_split_1231
9827 * [199]8994 ICE with -O -march=pentium4
9828 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs
9829 * [201]9806 ICE in inline assembly with -fPIC flag
9830 * [202]10077 gcc -msse2 generates movd to move dwords between xmm
9832 * [203]10233 64-bit comparison only comparing bottom 32-bits
9833 * [204]10286 type-punning doesn't work with __m64 and -O
9834 * [205]10308 [x86] ICE with -O -fgcse or -O2
9835 __________________________________________________________________
9841 This section lists the problem reports (PRs) from GCC's bug tracking
9842 system that are known to be fixed in the 3.3.1 release. This list might
9843 not be complete (that is, it is possible that some PRs that have been
9844 fixed are not listed here).
9848 * [206]11272 [Solaris] make bootstrap fails while building libstdc++
9850 Internal compiler errors (multi-platform)
9852 * [207]5754 ICE on invalid nested template class
9853 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64
9854 and --enable-checking
9855 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c
9856 * [210]7053 (c++) ICE when declaring a function already defined as a
9857 friend method of a template class
9858 * [211]8164 (c++) ICE when using different const expressions as
9860 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c
9861 * [213]9559 (c++) ICE with invalid initialization of a static const
9862 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c
9863 when redeclaring a static member variable
9864 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in
9865 dwarfout.c with -g -O -finline-functions
9866 * [216]10432 (c++) ICE in poplevel, in cp/decl.c
9867 * [217]10475 ICE in subreg_highpart_offset for code with long long
9868 * [218]10635 (c++) ICE when dereferencing an incomplete type casted
9870 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while
9871 instantiating static member variables
9872 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets
9873 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c
9874 * [222]10796 (c++) ICE when defining an enum with two values: -1 and
9876 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x
9878 * [224]10939 (c++) ICE with template code
9879 * [225]10956 (c++) ICE when specializing a template member function
9880 of a template class, in tsubst, in cp/pt.c
9881 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*()
9883 * [227]11059 (c++) ICE with empty union
9884 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with
9885 -O2 -fnon-call-exceptions
9886 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type
9887 * [230]11149 (c++) ICE on error when instantiation with call function
9889 * [231]11228 (c++) ICE on new-expression using array operator new and
9890 default-initialization
9891 * [232]11282 (c++) Infinite memory usage after syntax error
9892 * [233]11301 (fortran) ICE with -fno-globals
9893 * [234]11308 (c++) ICE when using an enum type name as if it were a
9895 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from
9897 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR
9898 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c:
9899 template member functions
9903 * [238]11198 -O2 -frename-registers generates wrong code (aliasing
9905 * [239]11304 Wrong code production with -fomit-frame-pointer
9906 * [240]11381 volatile memory access optimized away
9907 * [241]11536 [strength-reduce] -O2 optimization produces wrong code
9908 * [242]11557 constant folding bug generates wrong code
9912 * [243]5897 No warning for statement after return
9913 * [244]11279 DWARF-2 output mishandles large enums
9917 * [245]11022 no warning for non-compatible macro redefinition
9919 C++ compiler and library
9921 * [246]2330 static_cast<>() to a private base is allowed
9922 * [247]5388 Incorrect message "operands to ?: have different types"
9923 * [248]5390 Libiberty fails to demangle multi-digit template
9925 * [249]7877 Incorrect parameter passing to specializations of member
9927 * [250]9393 Anonymous namespaces and compiling the same file twice
9928 * [251]10032 -pedantic converts some errors to warnings
9929 * [252]10468 const typeof(x) is non-const, but only in templates
9930 * [253]10527 confused error message with "new int()" parameter
9932 * [254]10679 parameter MIN_INLINE_INSNS is not honored
9933 * [255]10682 gcc chokes on a typedef for an enum inside a class
9935 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of
9937 * [257]10845 template member function (with nested template as
9938 parameter) cannot be called anymore if another unrelated template
9939 member function is defined
9940 * [258]10849 Cannot define an out-of-class specialization of a
9941 private nested template class
9942 * [259]10888 Suppress -Winline warnings for system headers
9943 * [260]10929 -Winline warns about functions for which no definition
9945 * [261]10931 valid conversion static_cast<const unsigned
9946 int&>(lvalue-of-type-int) is rejected
9947 * [262]10940 Bad code with explicit specialization
9948 * [263]10968 If member function implicitly instantiated, explicit
9949 instantiation of class fails to instantiate it
9950 * [264]10990 Cannot convert with dynamic_cast<> to a private base
9951 class from within a member function
9952 * [265]11039 Bad interaction between implicit typename deprecation
9954 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say
9955 "__unused__" instead
9956 * [267]11095 C++ iostream manipulator causes segfault when called
9957 with negative argument
9958 * [268]11098 g++ doesn't emit complete debugging information for
9959 local variables in destructors
9960 * [269]11137 GNU/Linux shared library constructors not called unless
9961 there's one global object
9962 * [270]11154 spurious ambiguity report for template class
9964 * [271]11329 Compiler cannot find user defined implicit typecast
9965 * [272]11332 Spurious error with casts in ?: expression
9966 * [273]11431 static_cast behavior with subclasses when default
9967 constructor available
9968 * [274]11528 money_get facet does not accept "$.00" as valid
9969 * [275]11546 Type lookup problems in out-of-line definition of a
9970 class doubly nested from a template class
9971 * [276]11567 C++ code containing templated member function with same
9972 name as pure virtual member function results in linking failure
9973 * [277]11645 Failure to deal with using and private inheritance
9975 Java compiler and library
9977 * [278]5179 Qualified static field access doesn't initialize its
9979 * [279]8204 gcj -O2 to native reorders certain instructions
9981 * [280]10838 java.io.ObjectInputStream syntax error
9982 * [281]10886 The RMI registry that comes with GCJ does not work
9984 * [282]11349 JNDI URL context factories not located correctly
9986 x86-specific (Intel/AMD)
9988 * [283]4823 ICE on inline assembly code
9989 * [284]8878 miscompilation with -O and SSE
9990 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3
9992 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads,
9994 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2
9995 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source
9996 * [289]11044 [x86] out of range loop instructions for FP code on K6
9997 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE
9999 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC"
10002 SPARC- or Solaris- specific
10004 * [292]9362 solaris 'as' dies when fed .s and "-gstabs"
10005 * [293]10142 [SPARC64] gcc produces wrong code when passing
10006 structures by value
10007 * [294]10663 New configure check aborts with Sun tools.
10008 * [295]10835 combinatorial explosion in scheduler on HyperSPARC
10009 * [296]10876 ICE in calculate_giv_inc when building KDE
10010 * [297]10955 wrong code at -O3 for structure argument in context of
10012 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25
10013 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x
10018 * [300]10907 gcc violates the ia64 ABI (GP must be preserved)
10019 * [301]11320 scheduler bug (in machine depended reorganization pass)
10020 * [302]11599 bug with conditional and __builtin_prefetch
10024 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem
10026 * [304]10871 error in rs6000_stack_info save_size computation
10027 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse
10032 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx
10033 * [307]10557 [m68k] ICE in subreg_offset_representable_p
10034 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p
10038 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for
10039 functions with __attribute__ ((interrupt ("IRQ")))
10040 * [310]10842 [arm] Clobbered link register is copied to pc under
10041 certain circumstances
10042 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes
10043 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno
10048 * [313]11084 ICE in propagate_one_insn, in flow.c
10052 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf
10053 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c
10054 * [316]11096 i686-linux to sh-linux cross compiler fails to compile
10057 GNU/Linux (or Hurd?) specific
10059 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3
10063 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare
10066 Cygwin (or mingw) specific
10068 * [319]5287 ICE with dllimport attribute
10069 * [320]10148 [MingW/CygWin] Compiler dumps core
10073 * [321]8787 GCC fails to emit .intel_syntax when invoked with
10074 -masm=intel on DJGPP
10076 Darwin (and MacOS X) specific
10078 * [322]10900 trampolines crash
10082 * [323]1607 (c++) Format attributes on methods undocumented
10083 * [324]4252 Invalid option `-fdump-translation-unit'
10084 * [325]4490 Clarify restrictions on -m96bit-long-double,
10085 -m128bit-long-double
10086 * [326]10355 document an issue with regparm attribute on some systems
10088 * [327]10726 (fortran) Documentation for function "IDate Intrinsic
10090 * [328]10805 document bug in old version of Sun assembler
10091 * [329]10815 warn against GNU binutils on AIX
10092 * [330]10877 document need for newer binutils on i?86-*-linux-gnu
10093 * [331]11280 Manual incorrect with respect to -freorder-blocks
10094 * [332]11466 Document -mlittle-endian and its restrictions for the
10097 Testsuite bugs (compiler itself is not affected)
10099 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly
10101 * [334]10810 gcc-3.3 fails make check: buffer overrun in
10103 __________________________________________________________________
10109 This section lists the problem reports (PRs) from [335]GCC's bug
10110 tracking system that are known to be fixed in the 3.3.2 release. This
10111 list might not be complete (that is, it is possible that some PRs that
10112 have been fixed are not listed here).
10114 Bootstrap failures and problems
10116 * [336]8336 [SCO5] bootstrap config still tries to use COFF options
10117 * [337]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with
10118 --enable-threads=posix
10119 * [338]9631 [hppa64-linux] gcc-3.3 fails to bootstrap
10120 * [339]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare
10122 * [340]11687 xstormy16-elf build fails in libf2c
10123 * [341]12263 [SGI IRIX] bootstrap fails during compile of
10124 libf2c/libI77/backspace.c
10125 * [342]12490 buffer overflow in scan-decls.c (during Solaris 9
10126 fix-header processing)
10128 Internal compiler errors (multi-platform)
10130 * [343]7277 Casting integers to vector types causes ICE
10131 * [344]7939 (c++) ICE on invalid function template specialization
10132 * [345]11063 (c++) ICE on parsing initialization list of const array
10134 * [346]11207 ICE with negative index in array element designator
10135 * [347]11522 (fortran) g77 dwarf-2 ICE in
10136 add_abstract_origin_attribute
10137 * [348]11595 (c++) ICE on duplicate label definition
10138 * [349]11646 (c++) ICE in commit_one_edge_insertion with
10139 -fnon-call-exceptions -fgcse -O
10140 * [350]11665 ICE in struct initializer when taking address
10141 * [351]11852 (c++) ICE with bad struct initializer.
10142 * [352]11878 (c++) ICE in cp_expr_size
10143 * [353]11883 ICE with any -O on mercury-generated C code
10144 * [354]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in
10145 cp/typeck2.c when applying typeid operator to template template
10147 * [355]12146 ICE in lookup_template_function, in cp/pt.c
10148 * [356]12215 ICE in make_label_edge with -fnon-call-exceptions
10150 * [357]12369 (c++) ICE with templates and friends
10151 * [358]12446 ICE in emit_move_insn on complicated array reference
10152 * [359]12510 ICE in final_scan_insn
10153 * [360]12544 ICE with large parameters used in nested functions
10155 C and optimization bugs
10157 * [361]9862 spurious warnings with -W -finline-functions
10158 * [362]10962 lookup_field is a linear search on a linked list (can be
10159 slow if large struct)
10160 * [363]11370 -Wunreachable-code gives false complaints
10161 * [364]11637 invalid assembly with -fnon-call-exceptions
10162 * [365]11885 Problem with bitfields in packed structs
10163 * [366]12082 Inappropriate unreachable code warnings
10164 * [367]12180 Inline optimization fails for variadic function
10165 * [368]12340 loop unroller + gcse produces wrong code
10167 C++ compiler and library
10169 * [369]3907 nested template parameter collides with member name
10170 * [370]5293 confusing message when binding a temporary to a reference
10171 * [371]5296 [DR115] Pointers to functions and to template functions
10172 behave differently in deduction
10173 * [372]7939 ICE on function template specialization
10174 * [373]8656 Unable to assign function with __attribute__ and pointer
10175 return type to an appropriate variable
10176 * [374]10147 Confusing error message for invalid template function
10178 * [375]11400 std::search_n() makes assumptions about Size parameter
10179 * [376]11409 issues with using declarations, overloading, and
10181 * [377]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle
10182 multiple bits in mask
10183 * [378]11786 operator() call on variable in other namespace not
10185 * [379]11867 static_cast ignores ambiguity
10186 * [380]11928 bug with conversion operators that are typedefs
10187 * [381]12114 Uninitialized memory accessed in dtor
10188 * [382]12163 static_cast + explicit constructor regression
10189 * [383]12181 Wrong code with comma operator and c++
10190 * [384]12236 regparm and fastcall messes up parameters
10191 * [385]12266 incorrect instantiation of unneeded template during
10192 overload resolution
10193 * [386]12296 istream::peek() doesn't set eofbit
10194 * [387]12298 [sjlj exceptions] Stack unwind destroys
10195 not-yet-constructed object
10196 * [388]12369 ICE with templates and friends
10197 * [389]12337 apparently infinite loop in g++
10198 * [390]12344 stdcall attribute ignored if function returns a pointer
10199 * [391]12451 missing(late) class forward declaration in cxxabi.h
10200 * [392]12486 g++ accepts invalid use of a qualified name
10202 x86 specific (Intel/AMD)
10204 * [393]8869 [x86 MMX] ICE with const variable optimization and MMX
10206 * [394]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions
10208 * [395]11689 g++3.3 emits un-assembleable code for k6 architecture
10209 * [396]12116 [k6] Invalid assembly output values with X-MAME code
10210 * [397]12070 ICE converting between double and long double with
10215 * [398]11184 [ia64 hpux] ICE on __builtin_apply building libobjc
10216 * [399]11535 __builtin_return_address may not work on ia64
10217 * [400]11693 [ia64] ICE in gen_nop_type
10218 * [401]12224 [ia64] Thread-local storage doesn't work
10222 * [402]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux
10224 * [403]11319 loop miscompiled on ppc32
10225 * [404]11949 ICE Compiler segfault with ffmpeg -maltivec code
10229 * [405]11662 wrong code for expr. with cast to long long and
10231 * [406]11965 invalid assembler code for a shift < 32 operation
10232 * [407]12301 (c++) stack corruption when a returned expression throws
10237 * [408]11717 [alpha-linux] unrecognizable insn compiling for.c of
10242 * [409]11313 problem with #pragma weak and static inline functions
10243 * [410]11712 __STDC_EXT__ not defined for C++ by default anymore?
10247 * [411]12166 Profiled programs crash if PROFDIR is set
10249 Solaris-x86 specific
10251 * [412]12101 i386 Solaris no longer works with GNU as?
10253 Miscellaneous embedded target-specific bugs
10255 * [413]10988 [m32r-elf] wrong blockmove code with -O3
10256 * [414]11805 [h8300-unknown-coff] [H8300] ICE for simple code with
10258 * [415]11902 [sh4] spec file improperly inserts rpath even when none
10260 * [416]11903 [sh4] -pthread fails to link due to error in spec file
10262 __________________________________________________________________
10268 In addition to the bug fixes documented below, this release contains
10269 few minor features such as:
10270 * Support for --with-sysroot
10271 * Support for automatic detection of executable stacks
10272 * Support for SSE3 instructions
10273 * Support for thread local storage debugging under GDB on S390
10277 This section lists the problem reports (PRs) from [417]GCC's bug
10278 tracking system that are known to be fixed in the 3.3.3 release. This
10279 list might not be complete (that is, it is possible that some PRs that
10280 have been fixed are not listed here).
10282 Bootstrap failures and issues
10284 * [418]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails
10285 * [419]12399 boehm-gc fails (when building a cross compiler): libtool
10286 unable to infer tagged configuration
10287 * [420]13068 mklibgcc.in doesn't handle multi-level multilib
10288 subdirectories properly
10290 Internal compiler errors (multi-platform)
10292 * [421]10060 ICE (stack overflow) on huge file (300k lines) due to
10293 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c
10294 * [422]10555 (c++) ICE on undefined template argument
10295 * [423]10706 (c++) ICE in mangle_class_name_for_template
10296 * [424]11496 (fortran) error in flow_loops_find when -funroll-loops
10298 * [425]11741 ICE in pre_insert_copy_insn, in gcse.c
10299 * [426]12440 GCC crashes during compilation of quicktime4linux 2.0.0
10300 * [427]12632 (fortran) -fbounds-check ICE
10301 * [428]12712 (c++) ICE on short legit C++ code fragment with gcc
10303 * [429]12726 (c++) ICE (segfault) on trivial code
10304 * [430]12890 (c++) ICE on compilation of class with throwing method
10305 * [431]12900 (c++) ICE in rtl_verify_flow_info_1
10306 * [432]13060 (fortran) ICE in fixup_var_refs_1, in function.c on
10307 correct code with -O2 -fno-force-mem
10308 * [433]13289 (c++) ICE in regenerate_decl_from_template on recursive
10310 * [434]13318 ICE: floating point exception in the loop optimizer
10311 * [435]13392 (c++) ICE in convert_from_eh_region_ranges_1, in
10313 * [436]13574 (c++) invalid array default initializer in class lets
10314 gcc consume all memory and die
10315 * [437]13475 ICE on SIMD variables with partial value initialization
10316 * [438]13797 (c++) ICE on invalid template parameter
10317 * [439]13824 (java) gcj SEGV with simple .java program
10319 C and optimization bugs
10321 * [440]8776 loop invariants are not removed (most likely)
10322 * [441]10339 [sparc,ppc,ppc64] Invalid optimization: replacing
10324 * [442]11350 undefined labels with -Os -fPIC
10325 * [443]12826 Optimizer removes reference through volatile pointer
10326 * [444]12500 stabs debug info: void no longer a predefined / builtin
10328 * [445]12941 builtin-bitops-1.c miscompilation (latent bug)
10329 * [446]12953 tree inliner bug (in inline_forbidden_p) and fix
10330 * [447]13041 linux-2.6/sound/core/oss/rate.c miscompiled
10331 * [448]13507 spurious printf format warning
10332 * [449]13382 Type information for const pointer disappears during
10334 * [450]13394 noreturn attribute ignored on recursive invokation
10335 * [451]13400 Compiled code crashes storing to read-only location
10336 * [452]13521 Endless loop in calculate_global_regs_live
10338 C++ compiler and library
10340 Some of the bug fixes in this list were made to implement decisions
10341 that the ISO C++ standards committee has made concerning several defect
10342 reports (DRs). Links in the list below point to detailed discussion of
10343 the relevant defect report.
10344 * [453]2094 unimplemented: use of `ptrmem_cst' in template type
10346 * [454]2294 using declaration confusion
10347 * [455]5050 template instantiation depth exceeds limit: recursion
10349 * [456]9371 Bad exception handling in
10350 i/ostream::operator>>/<<(streambuf*)
10351 * [457]9546 bad exception handling in ostream members
10352 * [458]10081 basic_ios::_M_cache_locale leaves NULL members in the
10353 face of unknown locales
10354 * [459]10093 [460][DR 61] Setting failbit in exceptions doesn't work
10355 * [461]10095 istream::operator>>(int&) sets ios::badbit when
10356 ios::failbit is set.
10357 * [462]11554 Warning about reordering of initializers doesn't mention
10358 location of constructor
10359 * [463]12297 istream::sentry::sentry() handles eof() incorrectly.
10360 * [464]12352 Exception safety problems in src/localename.cc
10361 * [465]12438 Memory leak in locale::combine()
10362 * [466]12540 Memory leak in locale::locale(const char*)
10363 * [467]12594 DRs [468]60 [TC] and [469]63 [TC] not implemented
10364 * [470]12657 Resolution of [471]DR 292 (WP) still unimplemented
10365 * [472]12696 memory eating infinite loop in diagnostics (error
10367 * [473]12815 Code compiled with optimization behaves unexpectedly
10368 * [474]12862 Conflicts between typedefs/enums and namespace member
10370 * [475]12926 Wrong value after assignment in initialize list using
10372 * [476]12967 Resolution of [477]DR 300 [WP] still unimplemented
10373 * [478]12971 Resolution of [479]DR 328 [WP] still unimplemented
10374 * [480]13007 basic_streambuf::pubimbue, imbue wrong
10375 * [481]13009 Implicitly-defined assignment operator writes to wrong
10377 * [482]13057 regparm attribute not applied to destructor
10378 * [483]13070 -Wformat option ignored in g++
10379 * [484]13081 forward template declarations in <complex> let inlining
10381 * [485]13239 Assertion does not seem to work correctly anymore
10382 * [486]13262 "xxx is private within this context" when initializing a
10383 self-contained template class
10384 * [487]13290 simple typo in concept checking for std::generate_n
10385 * [488]13323 Template code does not compile in presence of typedef
10386 * [489]13369 __verify_grouping (and __add_grouping?) not correct
10387 * [490]13371 infinite loop with packed struct and inlining
10388 * [491]13445 Template argument replacement "dereferences" a typedef
10389 * [492]13461 Fails to access protected-ctor from public constant
10390 * [493]13462 Non-standard-conforming type set::pointer
10391 * [494]13478 gcc uses wrong constructor to initialize a const
10393 * [495]13544 "conflicting types" for enums in different scopes
10394 * [496]13650 string::compare should not (always) use
10395 traits_type::length()
10396 * [497]13683 bogus warning about passing non-PODs through ellipsis
10397 * [498]13688 Derived class is denied access to protected base class
10399 * [499]13774 Member variable cleared in virtual multiple inheritance
10401 * [500]13884 Protect sstream.tcc from extern template use
10403 Java compiler and library
10405 * [501]10746 [win32] garbage collection crash in GCJ
10407 Objective-C compiler and library
10409 * [502]11433 Crash due to dereferencing null pointer when querying
10412 Fortran compiler and library
10414 * [503]12633 logical expression gives incorrect result with
10415 -fugly-logint option
10416 * [504]13037 [gcse-lm] g77 generates incorrect code
10417 * [505]13213 Hex constant problem when compiling with -fugly-logint
10420 x86-specific (Intel/AMD)
10422 * [506]4490 ICE with -m128bit-long-double
10423 * [507]12292 [x86_64] ICE: RTL check: expected code `const_int', have
10424 `reg' in make_field_assignment, in combine.c
10425 * [508]12441 ICE: can't find a register to spill
10426 * [509]12943 array static-init failure under -fpic, -fPIC
10427 * [510]13608 Incorrect code with -O3 -ffast-math
10431 * [511]11598 testcase gcc.dg/20020118-1.c fails runtime check of
10432 __attribute__((aligned(16)))
10433 * [512]11793 ICE in extract_insn, in recog.c (const_vector's)
10434 * [513]12467 vmsumubm emitted when vmsummbm appropriate (typo in
10436 * [514]12537 g++ generates writeable text sections
10440 * [515]12496 wrong result for __atomic_add(&value, -1) when using -O0
10442 * [516]12865 mprotect call to make trampoline executable may fail
10443 * [517]13354 ICE in sparc_emit_set_const32
10447 * [518]10467 [arm] ICE in pre_insert_copy_insn,
10451 * [519]11226 ICE passing struct arg with two floats
10452 * [520]11227 ICE for _Complex float, _Complex long double args
10453 * [521]12644 GCC 3.3.2 fails to compile glibc on ia64
10454 * [522]13149 build gcc-3.3.2 1305 error:unrecognizable insn
10455 * Various fixes for libunwind
10459 * [523]12654 Incorrect comparison code generated for Alpha
10460 * [524]12965 SEGV+ICE in cc1plus on alpha-linux with -O2
10461 * [525]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2
10465 * [526]11634 [hppa] ICE in verify_local_live_at_start, in flow.c
10466 * [527]12158 [hppa] compilation does not terminate at -O1
10470 * [528]11992 Wrong built-in code for memcmp with length 1<<24: only
10471 (1<<24)-1 possible for CLCL-Instruction
10475 * [529]9365 segfault in gen_far_branch (config/sh/sh.c)
10476 * [530]10392 optimizer generates faulty array indexing
10477 * [531]11322 SH profiler outputs multiple definitions of symbol
10478 * [532]13069 gcc/config/sh/rtems.h broken
10479 * [533]13302 Putting a va_list in a struct causes seg fault
10480 * [534]13585 Incorrect optimization of call to sfunc
10481 * Fix inappropriately exported libgcc functions from the shared
10484 Other embedded target specific
10486 * [535]8916 [mcore] unsigned char assign gets hosed.
10487 * [536]11576 [h8300] ICE in change_address_1, in emit-rtl.c
10488 * [537]13122 [h8300] local variable gets corrupted by function call
10489 when -fomit-frame-pointer is given
10490 * [538]13256 [cris] strict_low_part mistreated in delay slots
10491 * [539]13373 [mcore] optimization with -frerun-cse-after-loop
10492 -fexpensive-optimizations produces wrong code on mcore
10496 * [540]12561 gcc/config/t-gnu needs updating to work with
10499 Tru64 Unix specific
10501 * [541]6243 testsuite fails almost all tests due to no libintl in
10502 LD_LIBRARY_PATH during test.
10503 * [542]11397 weak aliases broken on Tru64 UNIX
10507 * [543]12505 build failure due to defines of uchar in cpphash.h and
10509 * [544]13150 WEAK symbols not exported by collect2
10513 * [545]12666 fixincludes problem on IRIX 6.5.19m
10517 * [546]12969 Including sys/byteorder.h breaks configure checks
10519 Testsuite problems (compiler is not affected)
10521 * [547]10819 testsuite creates CR+LF on compiler version lines in
10523 * [548]11612 abi_check not finding correct libgcc_s.so.1
10527 * [549]13211 using -###, incorrect warnings about unused linker file
10529 __________________________________________________________________
10533 This is the [550]list of problem reports (PRs) from GCC's bug tracking
10534 system that are known to be fixed in the 3.3.4 release. This list might
10535 not be complete (that is, it is possible that some PRs that have been
10536 fixed are not listed here).
10537 __________________________________________________________________
10541 This is the [551]list of problem reports (PRs) from GCC's bug tracking
10542 system that are known to be fixed in the 3.3.5 release. This list might
10543 not be complete (that is, it is possible that some PRs that have been
10544 fixed are not listed here).
10545 __________________________________________________________________
10549 This is the [552]list of problem reports (PRs) from GCC's bug tracking
10550 system that are known to be fixed in the 3.3.6 release. This list might
10551 not be complete (that is, it is possible that some PRs that have been
10552 fixed are not listed here).
10555 For questions related to the use of GCC, please consult these web
10556 pages and the [553]GCC manuals. If that fails, the
10557 [554]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10558 web pages and the development of GCC are welcome on our developer
10559 list at [555]gcc@gcc.gnu.org. All of [556]our lists have public
10562 Copyright (C) [557]Free Software Foundation, Inc. Verbatim copying and
10563 distribution of this entire article is permitted in any medium,
10564 provided this notice is preserved.
10566 These pages are [558]maintained by the GCC team. Last modified
10571 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
10572 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems
10573 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
10574 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute
10575 5. http://gcc.gnu.org/news/dfa.html
10576 6. http://gcc.gnu.org/gcc-3.3/c99status.html
10577 7. http://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html
10578 8. http://gcc.gnu.org/PR10140
10579 9. http://gcc.gnu.org/PR10198
10580 10. http://gcc.gnu.org/PR10338
10581 11. http://gcc.gnu.org/PR3581
10582 12. http://gcc.gnu.org/PR4382
10583 13. http://gcc.gnu.org/PR5533
10584 14. http://gcc.gnu.org/PR6387
10585 15. http://gcc.gnu.org/PR6412
10586 16. http://gcc.gnu.org/PR6620
10587 17. http://gcc.gnu.org/PR6663
10588 18. http://gcc.gnu.org/PR7068
10589 19. http://gcc.gnu.org/PR7083
10590 20. http://gcc.gnu.org/PR7647
10591 21. http://gcc.gnu.org/PR7675
10592 22. http://gcc.gnu.org/PR7718
10593 23. http://gcc.gnu.org/PR8116
10594 24. http://gcc.gnu.org/PR8358
10595 25. http://gcc.gnu.org/PR8511
10596 26. http://gcc.gnu.org/PR8564
10597 27. http://gcc.gnu.org/PR8660
10598 28. http://gcc.gnu.org/PR8766
10599 29. http://gcc.gnu.org/PR8803
10600 30. http://gcc.gnu.org/PR8846
10601 31. http://gcc.gnu.org/PR8906
10602 32. http://gcc.gnu.org/PR9216
10603 33. http://gcc.gnu.org/PR9261
10604 34. http://gcc.gnu.org/PR9263
10605 35. http://gcc.gnu.org/PR9429
10606 36. http://gcc.gnu.org/PR9516
10607 37. http://gcc.gnu.org/PR9600
10608 38. http://gcc.gnu.org/PR9629
10609 39. http://gcc.gnu.org/PR9672
10610 40. http://gcc.gnu.org/PR9749
10611 41. http://gcc.gnu.org/PR9794
10612 42. http://gcc.gnu.org/PR9829
10613 43. http://gcc.gnu.org/PR9916
10614 44. http://gcc.gnu.org/PR9936
10615 45. http://gcc.gnu.org/PR10262
10616 46. http://gcc.gnu.org/PR10278
10617 47. http://gcc.gnu.org/PR10446
10618 48. http://gcc.gnu.org/PR10451
10619 49. http://gcc.gnu.org/PR10506
10620 50. http://gcc.gnu.org/PR10549
10621 51. http://gcc.gnu.org/PR2001
10622 52. http://gcc.gnu.org/PR2391
10623 53. http://gcc.gnu.org/PR2960
10624 54. http://gcc.gnu.org/PR4046
10625 55. http://gcc.gnu.org/PR6405
10626 56. http://gcc.gnu.org/PR6798
10627 57. http://gcc.gnu.org/PR6871
10628 58. http://gcc.gnu.org/PR6909
10629 59. http://gcc.gnu.org/PR7189
10630 60. http://gcc.gnu.org/PR7642
10631 61. http://gcc.gnu.org/PR8634
10632 62. http://gcc.gnu.org/PR8750
10633 63. http://gcc.gnu.org/PR2161
10634 64. http://gcc.gnu.org/PR4319
10635 65. http://gcc.gnu.org/PR8602
10636 66. http://gcc.gnu.org/PR9177
10637 67. http://gcc.gnu.org/PR9853
10638 68. http://gcc.gnu.org/PR45
10639 69. http://gcc.gnu.org/PR3784
10640 70. http://gcc.gnu.org/PR764
10641 71. http://gcc.gnu.org/PR5116
10642 72. http://gcc.gnu.org/PR2862
10643 73. http://gcc.gnu.org/PR3663
10644 74. http://gcc.gnu.org/PR3797
10645 75. http://gcc.gnu.org/PR3948
10646 76. http://gcc.gnu.org/PR4137
10647 77. http://gcc.gnu.org/PR4361
10648 78. http://gcc.gnu.org/PR4802
10649 79. http://gcc.gnu.org/PR5837
10650 80. http://gcc.gnu.org/PR4803
10651 81. http://gcc.gnu.org/PR5094
10652 82. http://gcc.gnu.org/PR5730
10653 83. http://gcc.gnu.org/PR6713
10654 84. http://gcc.gnu.org/PR7015
10655 85. http://gcc.gnu.org/PR7086
10656 86. http://gcc.gnu.org/PR7099
10657 87. http://gcc.gnu.org/PR7247
10658 88. http://gcc.gnu.org/PR7441
10659 89. http://gcc.gnu.org/PR7768
10660 90. http://gcc.gnu.org/PR7804
10661 91. http://gcc.gnu.org/PR8099
10662 92. http://gcc.gnu.org/PR8117
10663 93. http://gcc.gnu.org/PR8205
10664 94. http://gcc.gnu.org/PR8645
10665 95. http://gcc.gnu.org/PR8724
10666 96. http://gcc.gnu.org/PR8805
10667 97. http://gcc.gnu.org/PR8691
10668 98. http://gcc.gnu.org/PR8700
10669 99. http://gcc.gnu.org/PR8724
10670 100. http://gcc.gnu.org/PR8949
10671 101. http://gcc.gnu.org/PR9016
10672 102. http://gcc.gnu.org/PR9053
10673 103. http://gcc.gnu.org/PR9152
10674 104. http://gcc.gnu.org/PR9182
10675 105. http://gcc.gnu.org/PR9297
10676 106. http://gcc.gnu.org/PR9318
10677 107. http://gcc.gnu.org/PR9320
10678 108. http://gcc.gnu.org/PR9400
10679 109. http://gcc.gnu.org/PR9424
10680 110. http://gcc.gnu.org/PR9425
10681 111. http://gcc.gnu.org/PR9439
10682 112. http://gcc.gnu.org/PR9474
10683 113. http://gcc.gnu.org/PR9548
10684 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231
10685 115. http://gcc.gnu.org/PR9555
10686 116. http://gcc.gnu.org/PR9561
10687 117. http://gcc.gnu.org/PR9563
10688 118. http://gcc.gnu.org/PR9582
10689 119. http://gcc.gnu.org/PR9622
10690 120. http://gcc.gnu.org/PR9683
10691 121. http://gcc.gnu.org/PR9791
10692 122. http://gcc.gnu.org/PR9817
10693 123. http://gcc.gnu.org/PR9825
10694 124. http://gcc.gnu.org/PR9826
10695 125. http://gcc.gnu.org/PR9924
10696 126. http://gcc.gnu.org/PR9946
10697 127. http://gcc.gnu.org/PR9964
10698 128. http://gcc.gnu.org/PR9988
10699 129. http://gcc.gnu.org/PR10033
10700 130. http://gcc.gnu.org/PR10097
10701 131. http://gcc.gnu.org/PR10132
10702 132. http://gcc.gnu.org/PR10180
10703 133. http://gcc.gnu.org/PR10199
10704 134. http://gcc.gnu.org/PR10300
10705 135. http://gcc.gnu.org/PR10427
10706 136. http://gcc.gnu.org/PR10503
10707 137. http://gcc.gnu.org/PR5956
10708 138. http://gcc.gnu.org/PR1832
10709 139. http://gcc.gnu.org/PR3924
10710 140. http://gcc.gnu.org/PR5634
10711 141. http://gcc.gnu.org/PR6367
10712 142. http://gcc.gnu.org/PR6491
10713 143. http://gcc.gnu.org/PR6742
10714 144. http://gcc.gnu.org/PR7113
10715 145. http://gcc.gnu.org/PR7236
10716 146. http://gcc.gnu.org/PR7278
10717 147. http://gcc.gnu.org/PR7384
10718 148. http://gcc.gnu.org/PR7388
10719 149. http://gcc.gnu.org/PR8587
10720 150. http://gcc.gnu.org/PR9038
10721 151. http://gcc.gnu.org/PR10197
10722 152. http://gcc.gnu.org/PR6005
10723 153. http://gcc.gnu.org/PR6389
10724 154. http://gcc.gnu.org/PR6576
10725 155. http://gcc.gnu.org/PR6652
10726 156. http://gcc.gnu.org/PR7060
10727 157. http://gcc.gnu.org/PR7073
10728 158. http://gcc.gnu.org/PR7180
10729 159. http://gcc.gnu.org/PR7416
10730 160. http://gcc.gnu.org/PR7570
10731 161. http://gcc.gnu.org/PR7578
10732 162. http://gcc.gnu.org/PR7611
10733 163. http://gcc.gnu.org/PR7709
10734 164. http://gcc.gnu.org/PR7766
10735 165. http://gcc.gnu.org/PR7785
10736 166. http://gcc.gnu.org/PR7786
10737 167. http://gcc.gnu.org/PR8142
10738 168. http://gcc.gnu.org/PR8234
10739 169. http://gcc.gnu.org/PR8415
10740 170. http://gcc.gnu.org/PR8481
10741 171. http://gcc.gnu.org/PR8593
10742 172. http://gcc.gnu.org/PR8759
10743 173. http://gcc.gnu.org/PR8997
10744 174. http://gcc.gnu.org/PR9253
10745 175. http://gcc.gnu.org/PR9254
10746 176. http://gcc.gnu.org/PR9271
10747 177. http://gcc.gnu.org/PR6767
10748 178. http://gcc.gnu.org/PR9911
10749 179. http://gcc.gnu.org/PR10020
10750 180. http://gcc.gnu.org/PR10546
10751 181. http://gcc.gnu.org/PR7029
10752 182. http://gcc.gnu.org/PR2903
10753 183. http://gcc.gnu.org/PR7873
10754 184. http://gcc.gnu.org/PR7680
10755 185. http://gcc.gnu.org/PR8705
10756 186. http://gcc.gnu.org/PR9986
10757 187. http://gcc.gnu.org/PR10056
10758 188. http://gcc.gnu.org/PR6744
10759 189. http://gcc.gnu.org/PR7361
10760 190. http://gcc.gnu.org/PR9496
10761 191. http://gcc.gnu.org/PR7067
10762 192. http://gcc.gnu.org/PR8480
10763 193. http://gcc.gnu.org/PR8784
10764 194. http://gcc.gnu.org/PR10315
10765 195. http://gcc.gnu.org/PR10267
10766 196. http://gcc.gnu.org/PR7916
10767 197. http://gcc.gnu.org/PR7926
10768 198. http://gcc.gnu.org/PR8555
10769 199. http://gcc.gnu.org/PR8994
10770 200. http://gcc.gnu.org/PR9426
10771 201. http://gcc.gnu.org/PR9806
10772 202. http://gcc.gnu.org/PR10077
10773 203. http://gcc.gnu.org/PR10233
10774 204. http://gcc.gnu.org/PR10286
10775 205. http://gcc.gnu.org/PR10308
10776 206. http://gcc.gnu.org/PR11272
10777 207. http://gcc.gnu.org/PR5754
10778 208. http://gcc.gnu.org/PR6597
10779 209. http://gcc.gnu.org/PR6949
10780 210. http://gcc.gnu.org/PR7053
10781 211. http://gcc.gnu.org/PR8164
10782 212. http://gcc.gnu.org/PR8384
10783 213. http://gcc.gnu.org/PR9559
10784 214. http://gcc.gnu.org/PR9649
10785 215. http://gcc.gnu.org/PR9864
10786 216. http://gcc.gnu.org/PR10432
10787 217. http://gcc.gnu.org/PR10475
10788 218. http://gcc.gnu.org/PR10635
10789 219. http://gcc.gnu.org/PR10661
10790 220. http://gcc.gnu.org/PR10700
10791 221. http://gcc.gnu.org/PR10712
10792 222. http://gcc.gnu.org/PR10796
10793 223. http://gcc.gnu.org/PR10890
10794 224. http://gcc.gnu.org/PR10939
10795 225. http://gcc.gnu.org/PR10956
10796 226. http://gcc.gnu.org/PR11041
10797 227. http://gcc.gnu.org/PR11059
10798 228. http://gcc.gnu.org/PR11083
10799 229. http://gcc.gnu.org/PR11105
10800 230. http://gcc.gnu.org/PR11149
10801 231. http://gcc.gnu.org/PR11228
10802 232. http://gcc.gnu.org/PR11282
10803 233. http://gcc.gnu.org/PR11301
10804 234. http://gcc.gnu.org/PR11308
10805 235. http://gcc.gnu.org/PR11473
10806 236. http://gcc.gnu.org/PR11503
10807 237. http://gcc.gnu.org/PR11513
10808 238. http://gcc.gnu.org/PR11198
10809 239. http://gcc.gnu.org/PR11304
10810 240. http://gcc.gnu.org/PR11381
10811 241. http://gcc.gnu.org/PR11536
10812 242. http://gcc.gnu.org/PR11557
10813 243. http://gcc.gnu.org/PR5897
10814 244. http://gcc.gnu.org/PR11279
10815 245. http://gcc.gnu.org/PR11022
10816 246. http://gcc.gnu.org/PR2330
10817 247. http://gcc.gnu.org/PR5388
10818 248. http://gcc.gnu.org/PR5390
10819 249. http://gcc.gnu.org/PR7877
10820 250. http://gcc.gnu.org/PR9393
10821 251. http://gcc.gnu.org/PR10032
10822 252. http://gcc.gnu.org/PR10468
10823 253. http://gcc.gnu.org/PR10527
10824 254. http://gcc.gnu.org/PR10679
10825 255. http://gcc.gnu.org/PR10682
10826 256. http://gcc.gnu.org/PR10689
10827 257. http://gcc.gnu.org/PR10845
10828 258. http://gcc.gnu.org/PR10849
10829 259. http://gcc.gnu.org/PR10888
10830 260. http://gcc.gnu.org/PR10929
10831 261. http://gcc.gnu.org/PR10931
10832 262. http://gcc.gnu.org/PR10940
10833 263. http://gcc.gnu.org/PR10968
10834 264. http://gcc.gnu.org/PR10990
10835 265. http://gcc.gnu.org/PR11039
10836 266. http://gcc.gnu.org/PR11062
10837 267. http://gcc.gnu.org/PR11095
10838 268. http://gcc.gnu.org/PR11098
10839 269. http://gcc.gnu.org/PR11137
10840 270. http://gcc.gnu.org/PR11154
10841 271. http://gcc.gnu.org/PR11329
10842 272. http://gcc.gnu.org/PR11332
10843 273. http://gcc.gnu.org/PR11431
10844 274. http://gcc.gnu.org/PR11528
10845 275. http://gcc.gnu.org/PR11546
10846 276. http://gcc.gnu.org/PR11567
10847 277. http://gcc.gnu.org/PR11645
10848 278. http://gcc.gnu.org/PR5179
10849 279. http://gcc.gnu.org/PR8204
10850 280. http://gcc.gnu.org/PR10838
10851 281. http://gcc.gnu.org/PR10886
10852 282. http://gcc.gnu.org/PR11349
10853 283. http://gcc.gnu.org/PR4823
10854 284. http://gcc.gnu.org/PR8878
10855 285. http://gcc.gnu.org/PR9815
10856 286. http://gcc.gnu.org/PR10402
10857 287. http://gcc.gnu.org/PR10504
10858 288. http://gcc.gnu.org/PR10673
10859 289. http://gcc.gnu.org/PR11044
10860 290. http://gcc.gnu.org/PR11089
10861 291. http://gcc.gnu.org/PR11420
10862 292. http://gcc.gnu.org/PR9362
10863 293. http://gcc.gnu.org/PR10142
10864 294. http://gcc.gnu.org/PR10663
10865 295. http://gcc.gnu.org/PR10835
10866 296. http://gcc.gnu.org/PR10876
10867 297. http://gcc.gnu.org/PR10955
10868 298. http://gcc.gnu.org/PR11018
10869 299. http://gcc.gnu.org/PR11556
10870 300. http://gcc.gnu.org/PR10907
10871 301. http://gcc.gnu.org/PR11320
10872 302. http://gcc.gnu.org/PR11599
10873 303. http://gcc.gnu.org/PR9745
10874 304. http://gcc.gnu.org/PR10871
10875 305. http://gcc.gnu.org/PR11440
10876 306. http://gcc.gnu.org/PR7594
10877 307. http://gcc.gnu.org/PR10557
10878 308. http://gcc.gnu.org/PR11054
10879 309. http://gcc.gnu.org/PR10834
10880 310. http://gcc.gnu.org/PR10842
10881 311. http://gcc.gnu.org/PR11052
10882 312. http://gcc.gnu.org/PR11183
10883 313. http://gcc.gnu.org/PR11084
10884 314. http://gcc.gnu.org/PR10331
10885 315. http://gcc.gnu.org/PR10413
10886 316. http://gcc.gnu.org/PR11096
10887 317. http://gcc.gnu.org/PR2873
10888 318. http://gcc.gnu.org/PR3163
10889 319. http://gcc.gnu.org/PR5287
10890 320. http://gcc.gnu.org/PR10148
10891 321. http://gcc.gnu.org/PR8787
10892 322. http://gcc.gnu.org/PR10900
10893 323. http://gcc.gnu.org/PR1607
10894 324. http://gcc.gnu.org/PR4252
10895 325. http://gcc.gnu.org/PR4490
10896 326. http://gcc.gnu.org/PR10355
10897 327. http://gcc.gnu.org/PR10726
10898 328. http://gcc.gnu.org/PR10805
10899 329. http://gcc.gnu.org/PR10815
10900 330. http://gcc.gnu.org/PR10877
10901 331. http://gcc.gnu.org/PR11280
10902 332. http://gcc.gnu.org/PR11466
10903 333. http://gcc.gnu.org/PR10737
10904 334. http://gcc.gnu.org/PR10810
10905 335. http://gcc.gnu.org/bugzilla/
10906 336. http://gcc.gnu.org/PR8336
10907 337. http://gcc.gnu.org/PR9330
10908 338. http://gcc.gnu.org/PR9631
10909 339. http://gcc.gnu.org/PR9877
10910 340. http://gcc.gnu.org/PR11687
10911 341. http://gcc.gnu.org/PR12263
10912 342. http://gcc.gnu.org/PR12490
10913 343. http://gcc.gnu.org/PR7277
10914 344. http://gcc.gnu.org/PR7939
10915 345. http://gcc.gnu.org/PR11063
10916 346. http://gcc.gnu.org/PR11207
10917 347. http://gcc.gnu.org/PR11522
10918 348. http://gcc.gnu.org/PR11595
10919 349. http://gcc.gnu.org/PR11646
10920 350. http://gcc.gnu.org/PR11665
10921 351. http://gcc.gnu.org/PR11852
10922 352. http://gcc.gnu.org/PR11878
10923 353. http://gcc.gnu.org/PR11883
10924 354. http://gcc.gnu.org/PR11991
10925 355. http://gcc.gnu.org/PR12146
10926 356. http://gcc.gnu.org/PR12215
10927 357. http://gcc.gnu.org/PR12369
10928 358. http://gcc.gnu.org/PR12446
10929 359. http://gcc.gnu.org/PR12510
10930 360. http://gcc.gnu.org/PR12544
10931 361. http://gcc.gnu.org/PR9862
10932 362. http://gcc.gnu.org/PR10962
10933 363. http://gcc.gnu.org/PR11370
10934 364. http://gcc.gnu.org/PR11637
10935 365. http://gcc.gnu.org/PR11885
10936 366. http://gcc.gnu.org/PR12082
10937 367. http://gcc.gnu.org/PR12180
10938 368. http://gcc.gnu.org/PR12340
10939 369. http://gcc.gnu.org/PR3907
10940 370. http://gcc.gnu.org/PR5293
10941 371. http://gcc.gnu.org/PR5296
10942 372. http://gcc.gnu.org/PR7939
10943 373. http://gcc.gnu.org/PR8656
10944 374. http://gcc.gnu.org/PR10147
10945 375. http://gcc.gnu.org/PR11400
10946 376. http://gcc.gnu.org/PR11409
10947 377. http://gcc.gnu.org/PR11740
10948 378. http://gcc.gnu.org/PR11786
10949 379. http://gcc.gnu.org/PR11867
10950 380. http://gcc.gnu.org/PR11928
10951 381. http://gcc.gnu.org/PR12114
10952 382. http://gcc.gnu.org/PR12163
10953 383. http://gcc.gnu.org/PR12181
10954 384. http://gcc.gnu.org/PR12236
10955 385. http://gcc.gnu.org/PR12266
10956 386. http://gcc.gnu.org/PR12296
10957 387. http://gcc.gnu.org/PR12298
10958 388. http://gcc.gnu.org/PR12369
10959 389. http://gcc.gnu.org/PR12337
10960 390. http://gcc.gnu.org/PR12344
10961 391. http://gcc.gnu.org/PR12451
10962 392. http://gcc.gnu.org/PR12486
10963 393. http://gcc.gnu.org/PR8869
10964 394. http://gcc.gnu.org/PR9786
10965 395. http://gcc.gnu.org/PR11689
10966 396. http://gcc.gnu.org/PR12116
10967 397. http://gcc.gnu.org/PR12070
10968 398. http://gcc.gnu.org/PR11184
10969 399. http://gcc.gnu.org/PR11535
10970 400. http://gcc.gnu.org/PR11693
10971 401. http://gcc.gnu.org/PR12224
10972 402. http://gcc.gnu.org/PR11087
10973 403. http://gcc.gnu.org/PR11319
10974 404. http://gcc.gnu.org/PR11949
10975 405. http://gcc.gnu.org/PR11662
10976 406. http://gcc.gnu.org/PR11965
10977 407. http://gcc.gnu.org/PR12301
10978 408. http://gcc.gnu.org/PR11717
10979 409. http://gcc.gnu.org/PR11313
10980 410. http://gcc.gnu.org/PR11712
10981 411. http://gcc.gnu.org/PR12166
10982 412. http://gcc.gnu.org/PR12101
10983 413. http://gcc.gnu.org/PR10988
10984 414. http://gcc.gnu.org/PR11805
10985 415. http://gcc.gnu.org/PR11902
10986 416. http://gcc.gnu.org/PR11903
10987 417. http://gcc.gnu.org/bugzilla/
10988 418. http://gcc.gnu.org/PR11890
10989 419. http://gcc.gnu.org/PR12399
10990 420. http://gcc.gnu.org/PR13068
10991 421. http://gcc.gnu.org/PR10060
10992 422. http://gcc.gnu.org/PR10555
10993 423. http://gcc.gnu.org/PR10706
10994 424. http://gcc.gnu.org/PR11496
10995 425. http://gcc.gnu.org/PR11741
10996 426. http://gcc.gnu.org/PR12440
10997 427. http://gcc.gnu.org/PR12632
10998 428. http://gcc.gnu.org/PR12712
10999 429. http://gcc.gnu.org/PR12726
11000 430. http://gcc.gnu.org/PR12890
11001 431. http://gcc.gnu.org/PR12900
11002 432. http://gcc.gnu.org/PR13060
11003 433. http://gcc.gnu.org/PR13289
11004 434. http://gcc.gnu.org/PR13318
11005 435. http://gcc.gnu.org/PR13392
11006 436. http://gcc.gnu.org/PR13574
11007 437. http://gcc.gnu.org/PR13475
11008 438. http://gcc.gnu.org/PR13797
11009 439. http://gcc.gnu.org/PR13824
11010 440. http://gcc.gnu.org/PR8776
11011 441. http://gcc.gnu.org/PR10339
11012 442. http://gcc.gnu.org/PR11350
11013 443. http://gcc.gnu.org/PR12826
11014 444. http://gcc.gnu.org/PR12500
11015 445. http://gcc.gnu.org/PR12941
11016 446. http://gcc.gnu.org/PR12953
11017 447. http://gcc.gnu.org/PR13041
11018 448. http://gcc.gnu.org/PR13507
11019 449. http://gcc.gnu.org/PR13382
11020 450. http://gcc.gnu.org/PR13394
11021 451. http://gcc.gnu.org/PR13400
11022 452. http://gcc.gnu.org/PR13521
11023 453. http://gcc.gnu.org/PR2094
11024 454. http://gcc.gnu.org/PR2294
11025 455. http://gcc.gnu.org/PR5050
11026 456. http://gcc.gnu.org/PR9371
11027 457. http://gcc.gnu.org/PR9546
11028 458. http://gcc.gnu.org/PR10081
11029 459. http://gcc.gnu.org/PR10093
11030 460. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61
11031 461. http://gcc.gnu.org/PR10095
11032 462. http://gcc.gnu.org/PR11554
11033 463. http://gcc.gnu.org/PR12297
11034 464. http://gcc.gnu.org/PR12352
11035 465. http://gcc.gnu.org/PR12438
11036 466. http://gcc.gnu.org/PR12540
11037 467. http://gcc.gnu.org/PR12594
11038 468. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60
11039 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63
11040 470. http://gcc.gnu.org/PR12657
11041 471. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292
11042 472. http://gcc.gnu.org/PR12696
11043 473. http://gcc.gnu.org/PR12815
11044 474. http://gcc.gnu.org/PR12862
11045 475. http://gcc.gnu.org/PR12926
11046 476. http://gcc.gnu.org/PR12967
11047 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html
11048 478. http://gcc.gnu.org/PR12971
11049 479. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328
11050 480. http://gcc.gnu.org/PR13007
11051 481. http://gcc.gnu.org/PR13009
11052 482. http://gcc.gnu.org/PR13057
11053 483. http://gcc.gnu.org/PR13070
11054 484. http://gcc.gnu.org/PR13081
11055 485. http://gcc.gnu.org/PR13239
11056 486. http://gcc.gnu.org/PR13262
11057 487. http://gcc.gnu.org/PR13290
11058 488. http://gcc.gnu.org/PR13323
11059 489. http://gcc.gnu.org/PR13369
11060 490. http://gcc.gnu.org/PR13371
11061 491. http://gcc.gnu.org/PR13445
11062 492. http://gcc.gnu.org/PR13461
11063 493. http://gcc.gnu.org/PR13462
11064 494. http://gcc.gnu.org/PR13478
11065 495. http://gcc.gnu.org/PR13544
11066 496. http://gcc.gnu.org/PR13650
11067 497. http://gcc.gnu.org/PR13683
11068 498. http://gcc.gnu.org/PR13688
11069 499. http://gcc.gnu.org/PR13774
11070 500. http://gcc.gnu.org/PR13884
11071 501. http://gcc.gnu.org/PR10746
11072 502. http://gcc.gnu.org/PR11433
11073 503. http://gcc.gnu.org/PR12633
11074 504. http://gcc.gnu.org/PR13037
11075 505. http://gcc.gnu.org/PR13213
11076 506. http://gcc.gnu.org/PR4490
11077 507. http://gcc.gnu.org/PR12292
11078 508. http://gcc.gnu.org/PR12441
11079 509. http://gcc.gnu.org/PR12943
11080 510. http://gcc.gnu.org/PR13608
11081 511. http://gcc.gnu.org/PR11598
11082 512. http://gcc.gnu.org/PR11793
11083 513. http://gcc.gnu.org/PR12467
11084 514. http://gcc.gnu.org/PR12537
11085 515. http://gcc.gnu.org/PR12496
11086 516. http://gcc.gnu.org/PR12865
11087 517. http://gcc.gnu.org/PR13354
11088 518. http://gcc.gnu.org/PR10467
11089 519. http://gcc.gnu.org/PR11226
11090 520. http://gcc.gnu.org/PR11227
11091 521. http://gcc.gnu.org/PR12644
11092 522. http://gcc.gnu.org/PR13149
11093 523. http://gcc.gnu.org/PR12654
11094 524. http://gcc.gnu.org/PR12965
11095 525. http://gcc.gnu.org/PR13031
11096 526. http://gcc.gnu.org/PR11634
11097 527. http://gcc.gnu.org/PR12158
11098 528. http://gcc.gnu.org/PR11992
11099 529. http://gcc.gnu.org/PR9365
11100 530. http://gcc.gnu.org/PR10392
11101 531. http://gcc.gnu.org/PR11322
11102 532. http://gcc.gnu.org/PR13069
11103 533. http://gcc.gnu.org/PR13302
11104 534. http://gcc.gnu.org/PR13585
11105 535. http://gcc.gnu.org/PR8916
11106 536. http://gcc.gnu.org/PR11576
11107 537. http://gcc.gnu.org/PR13122
11108 538. http://gcc.gnu.org/PR13256
11109 539. http://gcc.gnu.org/PR13373
11110 540. http://gcc.gnu.org/PR12561
11111 541. http://gcc.gnu.org/PR6243
11112 542. http://gcc.gnu.org/PR11397
11113 543. http://gcc.gnu.org/PR12505
11114 544. http://gcc.gnu.org/PR13150
11115 545. http://gcc.gnu.org/PR12666
11116 546. http://gcc.gnu.org/PR12969
11117 547. http://gcc.gnu.org/PR10819
11118 548. http://gcc.gnu.org/PR11612
11119 549. http://gcc.gnu.org/PR13211
11120 550. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4
11121 551. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5
11122 552. http://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6
11123 553. http://gcc.gnu.org/onlinedocs/
11124 554. mailto:gcc-help@gcc.gnu.org
11125 555. mailto:gcc@gcc.gnu.org
11126 556. http://gcc.gnu.org/lists.html
11127 557. http://www.fsf.org/
11128 558. http://gcc.gnu.org/about.html
11129 559. http://validator.w3.org/check/referer
11130 ======================================================================
11131 http://gcc.gnu.org/gcc-3.2/index.html
11132 GCC 3.2 Release Series
11136 The [1]GNU project and the GCC developers are pleased to announce the
11137 release of GCC 3.2.3.
11139 The purpose of the GCC 3.2 release series is to provide a stable
11140 platform for OS distributors to use building their next releases. A
11141 primary objective was to stabilize the C++ ABI; we believe that the
11142 interface to the compiler and the C++ standard library are now
11145 Be aware that C++ code compiled by GCC 3.2.x will (in general) not
11146 interoperate with code compiled by GCC 3.1.1 or earlier.
11148 Please refer to our [2]detailed list of news, caveats, and bug-fixes
11149 for further information.
11154 April 25, 2003 ([3]changes)
11157 February 5, 2003 ([4]changes)
11160 November 19, 2002 ([5]changes)
11163 August 14, 2002 ([6]changes)
11165 References and Acknowledgements
11167 GCC used to stand for the GNU C Compiler, but since the compiler
11168 supports several other languages aside from C, it now stands for the
11169 GNU Compiler Collection.
11171 A list of [7]successful builds is updated as new information becomes
11174 The GCC developers would like to thank the numerous people that have
11175 contributed new features, improvements, bug fixes, and other changes as
11176 well as test results to GCC. This [8]amazing group of volunteers is
11177 what makes GCC successful.
11179 For additional information about GCC please refer to the [9]GCC project
11180 web site or contact the [10]GCC development mailing list.
11182 To obtain GCC please use [11]our mirror sites, or our CVS server.
11185 For questions related to the use of GCC, please consult these web
11186 pages and the [12]GCC manuals. If that fails, the
11187 [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11188 web pages and the development of GCC are welcome on our developer
11189 list at [14]gcc@gcc.gnu.org. All of [15]our lists have public
11192 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and
11193 distribution of this entire article is permitted in any medium,
11194 provided this notice is preserved.
11196 These pages are [17]maintained by the GCC team. Last modified
11201 1. http://www.gnu.org/
11202 2. http://gcc.gnu.org/gcc-3.2/changes.html
11203 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
11204 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2
11205 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1
11206 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2
11207 7. http://gcc.gnu.org/gcc-3.2/buildstat.html
11208 8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11209 9. http://gcc.gnu.org/index.html
11210 10. mailto:gcc@gcc.gnu.org
11211 11. http://gcc.gnu.org/mirrors.html
11212 12. http://gcc.gnu.org/onlinedocs/
11213 13. mailto:gcc-help@gcc.gnu.org
11214 14. mailto:gcc@gcc.gnu.org
11215 15. http://gcc.gnu.org/lists.html
11216 16. http://www.fsf.org/
11217 17. http://gcc.gnu.org/about.html
11218 18. http://validator.w3.org/check/referer
11219 ======================================================================
11220 http://gcc.gnu.org/gcc-3.2/changes.html
11221 GCC 3.2 Release Series
11222 Changes, New Features, and Fixes
11224 The latest release in the 3.2 release series is [1]GCC 3.2.3.
11226 Caveats and New Features
11230 * The C++ compiler does not correctly zero-initialize
11231 pointers-to-data members. You must explicitly initialize them. For
11232 example: int S::*m(0); will work, but depending on
11233 default-initialization to zero will not work. This bug cannot be
11234 fixed in GCC 3.2 without inducing unacceptable risks. It will be
11236 * This GCC release is based on the GCC 3.1 sourcebase, and thus has
11237 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has
11238 a number of C++ ABI fixes which make its C++ compiler generate
11239 binary code which is incompatible with the C++ compilers found in
11240 earlier GCC releases, including GCC 3.1 and GCC 3.1.1.
11242 Frontend Enhancements
11246 * The method of constructing the list of directories to be searched
11247 for header files has been revised. If a directory named by a -I
11248 option is a standard system include directory, the option is
11249 ignored to ensure that the default search order for system
11250 directories and the special treatment of system header files are
11252 * The C and Objective-C compilers no longer accept the "Naming Types"
11253 extension (typedef foo = bar); it was already unavailable in C++.
11254 Code which uses it will need to be changed to use the "typeof"
11255 extension instead: typedef typeof(bar) foo. (We have removed this
11256 extension without a period of deprecation because it has caused the
11257 compiler to crash since version 3.0 and no one noticed until very
11258 recently. Thus we conclude it is not in widespread use.)
11262 * GCC 3.2 fixed serveral differences between the C++ ABI implemented
11263 in GCC and the multi-vendor standard, but more have been found
11264 since the release. 3.2.1 adds a new warning, -Wabi, to warn about
11265 code which is affected by these bugs. We will fix these bugs in
11266 some future release, once we are confident that all have been
11267 found; until then, it is our intention to make changes to the ABI
11268 only if they are necessary for correct compilation of C++, as
11269 opposed to conformance to the ABI documents.
11270 * For details on how to build an ABI compliant compiler for GNU/Linux
11271 systems, check the [3]common C++ ABI page.
11273 New Targets and Target Specific Improvements
11277 * Fixed a number of bugs in SSE and MMX intrinsics.
11278 * Fixed common compiler crashes with SSE instruction set enabled
11279 (implied by -march=pentium3, pentium4, athlon-xp)
11280 * __m128 and __m128i is not 128bit aligned when used in structures.
11284 * A bug whereby the compiler could generate bad code for bzero has
11286 * ABI fixes (implying ABI incompatibilities with previous version in
11288 * Fixed prefetch code generation
11289 __________________________________________________________________
11293 3.2.3 is a bug fix release only; there are no new features that were
11294 not present in GCC 3.2.2.
11298 This section lists the problem reports (PRs) from GCC's bug tracking
11299 system that are known to be fixed in the 3.2.3 release. This list might
11300 not be complete (that is, it is possible that some PRs that have been
11301 fixed are not listed here), and some of the titles have been changed to
11302 make them more clear.
11304 Internal Compiler Errors (multi-platform)
11306 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in
11308 * [5]6440: (c++) template specializations cause ICE
11309 * [6]7050: (c++) ICE on: (i ? get_string() : throw)
11310 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c)
11311 * [8]7982: (c++) ICE due to infinite recursion (using STL set)
11312 * [9]8068: exceedingly high (infinite) memory usage
11313 * [10]8178: ICE with __builtin_ffs
11314 * [11]8396: ICE in copy_to_mode_reg, in explow.c
11315 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c
11316 * [13]9768: ICE when optimizing inline code at -O2
11317 * [14]9798: (c++) Infinite recursion (segfault) in
11318 cp/decl.c:push_using_directive with recursive using directives
11319 * [15]9799: mismatching structure initializer with nested flexible
11321 * [16]9928: ICE on duplicate enum declaration
11322 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects
11324 * [18]10352: ICE in find_reloads_toplev
11325 * [19]10336: ICE with -Wunreachable-code
11329 * [20]8224: Incorrect joining of signed and unsigned division
11330 * [21]8613: -O2 produces wrong code with builtin strlen and
11332 * [22]8828: gcc reports some code is unreachable when it is not
11333 * [23]9226: GCSE breaking argument passing
11334 * [24]9853: miscompilation of non-constant structure initializer
11335 * [25]9797: C99-style struct initializers are miscompiled
11336 * [26]9967: Some standard C function calls should not be replaced
11337 when optimizing for size
11338 * [27]10116: ce2: invalid merge of join_bb in the context of switch
11340 * [28]10171: wrong code for inlined function
11341 * [29]10175: -Wunreachable-code doesn't work for single lines
11343 C++ compiler and library:
11345 * [30]8316: Confusing diagnostic for code that misuses conversion
11347 * [31]9169: filebuf output fails if codecvt<>::out returns noconv
11348 * [32]9420: incomplete type incorrectly reported
11349 * [33]9459: typeof in return type specification of template not
11351 * [34]9507: filebuf::open handles ios_base::ate incorrectly
11352 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc
11353 * [36]9602: Total confusion about template/friend/virtual/abstract
11354 * [37]9993: destructor not called for local object created within and
11355 returned from infinite loop
11356 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2
11359 Java compiler and library:
11361 * [39]9652: libgcj build fails on irix6.5.1[78]
11362 * [40]10144: gas on solaris complains about bad .stabs lines for
11363 java, native as unaffected
11365 x86-specific (Intel/AMD):
11367 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86
11368 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions
11369 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu
11371 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib)
11376 * [45]7784: [Sparc] ICE in extract_insn, in recog.c
11377 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in
11379 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc
11380 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in
11382 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc
11383 * [50]9414: Scheduling bug on Ultrasparc
11384 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64
11388 * [52]7248: broken "inclusive or" code
11389 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1
11393 * [54]9732: Wrong code with -O2 -fPIC
11394 * [55]10073: ICE: powerpc cannot split insn
11398 * [56]7702: optimization problem on a DEC alpha under OSF1
11399 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system
11403 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275)
11404 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10
11406 * [60]10271: Floating point args don't get reloaded across function
11411 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in
11416 * [62]10377: gcc-3.2.2 creates bad assembler code for cris
11418 Miscellaneous and minor bugs:
11420 * [63]6955: collect2 says "core dumped" when there is no core
11421 __________________________________________________________________
11425 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make
11426 install by means of the DESTDIR variable. Parts of the GCC tree have
11427 featured that support long before, but now it is available even from
11430 Other than that, GCC 3.2.2 is a bug fix release only; there are no new
11431 features that were not present in GCC 3.2.1.
11435 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt.
11436 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped
11437 with FreeBSD 5.0 does not have this problem), Interix, a.out-based
11438 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI
11439 change, and thus restores ABI-compatibility with previous releases
11440 (except GCC 3.2.1) on these platforms.
11442 This section lists the problem reports (PRs) from GCC's bug tracking
11443 system that are known to be fixed in the 3.2.2 release. This list might
11444 not be complete (that is, it is possible that some PRs that have been
11445 fixed are not listed here) and some of the titles have been changed to
11446 make them more clear.
11448 Internal Compiler Errors (multi-platform)
11450 * [64]5919: (c++) ICE when passing variable array to template
11452 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=)
11453 * [66]7507: ICE with -O2 when address of called function is a
11454 complicated expression
11455 * [67]7622: ICE with nested inline functions if function's address is
11457 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR
11459 * [70]8031: (c++) ICE in code comparing typeids and casting from
11461 * [71]8275: ICE in simplify_subreg
11462 * [72]8332: (c++) builtin strlen/template interaction causes ICE
11463 * [73]8372: (c++) ICE on explicit call of destructor
11464 * [74]8439: (c, not c++) empty struct causes ICE
11465 * [75]8442: (c++) ICE with nested template classes
11466 * [76]8518: ICE when compiling mplayer ("extern inline" issue)
11467 * [77]8615: (c++) ICE with out-of-range character constant template
11469 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307
11470 * [79]8799: (c++) ICE: error reporting routines re-entered
11471 * [80]9328: (c++) ICE with typeof(X) for overloaded X
11472 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes
11474 C++ (compiler and library) bugs
11476 * [82]47: scoping in nested classes is broken
11477 * [83]6745: problems with iostream rdbuf() member function
11478 * [84]8214: conversion from const char* const to char* sometimes
11480 * [85]8493: builtin strlen and overload resolution (same bug as
11482 * [87]8503: strange behaviour of function types
11483 * [88]8727: compiler confused by inheritance from an anonymous struct
11484 * [89]7445: poor performance of std::locale::classic() in
11485 multi-threaded applications
11486 * [90]8230: mishandling of overflow in vector<T>::resize
11487 * [91]8399: sync_with_stdio(false) breaks unformatted input
11488 * [92]8662: illegal access of private member of unnamed class is
11490 * [93]8707: "make distclean" fails in libstdc++-v3 directory
11491 * [94]8708: __USE_MALLOC doesn't work
11492 * [95]8790: Use of non-thread-safe strtok in src/localename.cc
11493 * [96]8887: Bug in date formats with --enable-clocale=generic
11494 * [97]9076: Call Frame Instructions are not handled correctly during
11496 * [98]9151: std::setprecision limited to 16 digits when outputting a
11498 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers
11499 * [100]9269: libstdc++ headers: explicit specialization of function
11500 must precede its first use
11501 * [101]9322: return value of basic_streambuf<>::getloc affected by
11503 * [102]9433: segfault in runtime support for dynamic_cast
11505 C and optimizer bugs
11507 * [103]8032: GCC incorrectly initializes static structs that have
11509 * [104]8639: simple arithmetic expression broken
11510 * [105]8794: optimization improperly eliminates certain expressions
11511 * [106]8832: traditional "asm volatile" code is illegally optimized
11512 * [107]8988: loop optimizer bug: with -O2, code is generated that
11513 segfaults (found on i386, bug present for all platforms)
11514 * [108]9492: structure copy clobbers subsequent stores to structure
11518 * [109]9267: Objective-C parser won't build with newer bison versions
11523 * [110]8344: Ada build problem due to conflict between gcc/final.o,
11528 * [111]8524: _Pragma within macros is improperly expanded
11529 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with
11534 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95
11536 x86-specific (Intel/AMD)
11538 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction)
11539 * [115]8599: loop unroll bug with -march=k6-3
11540 * [116]9506: ABI breakage in structure return (affects BSD and
11541 Cygwin, but not GNU/Linux)
11543 FreeBSD 5.0 specific
11545 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0
11549 * [118]9292: hppa1.1-rtems configurery problems
11550 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug
11551 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue
11552 * [121]9296: gthr-rtems regression
11553 * [122]9316: powerpc-rtems: extending multilibs
11557 * [123]9493: ICE with -O2 when building a simple function
11561 * [124]7341: hyperlink to gcov in GCC documentation doesn't work
11562 * [125]8947: Please add a warning about "-malign-double" in docs
11563 * [126]7448, [127]8882: typo cleanups
11564 __________________________________________________________________
11568 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++
11569 generates code that is known not to be binary-compatible with the
11570 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included
11571 in the distribution, for details.
11573 This release also removes an old GCC extension, "naming types", and the
11574 documentation now directs users to use a different GCC extension,
11575 __typeof__, instead. The feature had evidently been broken for a while.
11577 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and
11578 the new warning there are no new features that were not present in GCC
11581 In addition, the previous fix for [128]PR 7445 (poor performance of
11582 std::locale::classic() in multi-threaded applications) was reverted
11583 ("unfixed"), because the "fix" was not thread-safe.
11587 This section lists the problem reports (PRs) from GCC's bug tracking
11588 system that are known to be fixed in the 3.2.1 release. This list might
11589 not be complete (that is, it is possible that some PRs that have been
11590 fixed are not listed here). As you can see, the number of bug fixes is
11591 quite large, so it is strongly recommended that users of earlier GCC
11592 3.x releases upgrade to GCC 3.2.1.
11594 Internal Compiler Errors (multi-platform)
11596 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c
11597 * [130]5661: (c++) ICE instantiating template on array of unknown
11599 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on
11601 * [132]6994: (c++) ICE in find_function_data
11602 * [133]7150: preprocessor: GCC -dM -E gives an ICE
11603 * [134]7160: ICE when optimizing branches without a return value
11604 * [135]7228: (c++) ICE when using member template and template
11606 * [136]7266: (c++) ICE with -pedantic on missing typename
11607 * [137]7353: ICE from use of "Naming Types" extension, see above
11608 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c
11609 * [139]7478: (c++) ICE on static_cast inside template
11610 * [140]7526: preprocessor core dump when _Pragma implies #pragma
11612 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803
11614 * [143]7754: (c++) ICE on union with template parameter
11615 * [144]7788: (c++) redeclaring a definition as an incomplete class
11617 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c
11618 * [146]8055: preprocessor dies with SIG11 when building FreeBSD
11620 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related
11622 * [148]8134: (c++) ICE in force_store_init_value on legal code
11623 * [149]8149: (c++) ICE on incomplete type
11624 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array
11627 C++ (compiler and library) bugs
11629 * [151]5607: No pointer adjustment in covariant return types
11630 * [152]6579: Infinite loop with statement expressions in member
11632 * [153]6803: Default copy constructor bug in GCC 3.1
11633 * [154]7176: g++ confused by friend and static member with same name
11634 * [155]7188: Segfault with template class and recursive (incorrect)
11636 * [156]7306: Regression: GCC 3.x fails to compile code with virtual
11637 inheritance if a method has a variable number of arguments
11638 * [157]7461: ctype<char>::classic_table() returns offset array on
11640 * [158]7524: f(const float arg[3]) fails
11641 * [159]7584: Erroneous ambiguous base error on using declaration
11642 * [160]7676: Member template overloading problem
11643 * [161]7679: infinite loop when a right parenthesis is missing
11644 * [162]7811: default locale not taken from environment
11645 * [163]7961: compare( char *) implemented incorrectly in
11647 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if
11648 streambuf::underflow() leaves gptr() NULL (dups: [165]8127,
11650 * [167]8096: deque::at() throws std::range_error instead of
11652 * [168]8127: cout << cin.rdbuf() infinite loop
11653 * [169]8218: Excessively large memory consumed for classes with large
11655 * [170]8287: GCC 3.2: Destructor called for non-constructed local
11657 * [171]8347: empty vector range used in string construction causes
11659 * [172]8348: fail() flag is set in istringstream when eof() flag is
11661 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file)
11663 C and optimizer bugs
11665 * [174]6627: -fno-align-functions doesn't seem to disable function
11667 * [175]6631: life_analysis misoptimizes code to initialize fields of
11669 * [176]7102: unsigned char division results in floating exception
11670 * [177]7120: Run once loop should *always* be unrolled
11672 * [178]7209: Bug involving array referencing and ?: operator
11673 * [179]7515: invalid inlining of global function with -O3
11674 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test
11675 * [181]8467: bug in sibling call optimization
11679 * [182]4890: incorrect line markers from the traditional preprocessor
11680 * [183]7357: -M option omits system headers files (making it the same
11682 * [184]7358: Changes to Sun's make Dependencies
11683 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as
11685 * [186]7862: Interrupting GCC -MD removes .d file but not .o
11686 * [187]8190: Failed compilation deletes -MD dependency file
11687 * [188]8524: _Pragma within macro is improperly expanded
11689 x86 specific (Intel/AMD)
11691 * [189]5351: (i686-only) function pass-by-value structure copy
11692 corrupts stack ([190]7591 is a duplicate)
11693 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with
11694 -march=pentium3/pentium2/athlon (these are all the same underlying
11695 bug, in MMX register use)
11696 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same
11698 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken
11699 * [199]6981: wrong code in 64-bit manipulation on x86
11700 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__
11702 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE
11703 intrinsics are broken
11704 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with
11706 * [203]7693: Typo in i386 mmintrin.h header
11707 * [204]7723: ICE - Pentium3 sse - GCC 3.2
11708 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse
11709 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3
11713 * [207]5967: GCC bug when profiling nested functions on powerpc
11714 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while
11716 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5
11717 * [210]7130: miscompiled code for GCC-3.1 on
11718 powerpc-unknown-linux-gnu with -funroll-all-loops
11719 * [211]7133: PowerPC ICE: unrecognizable insn
11720 * [212]7380: ICE in extract_insn, at recog.c:2148
11721 * [213]8252: ICE on Altivec code with optimization turned on
11722 * [214]8451: Altivec ICE in GCC 3.2
11726 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa
11730 * [216]6668: when using --disable-multilib, libgcc_s.so is installed
11731 in the wrong place on sparc-solaris
11732 * [217]7151: ICE when compiling for UltraSPARC
11733 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long
11735 * [219]7842: [REGRESSION] SPARC code gen bug
11739 * [220]7856: [arm] invalid offset in constant pool reference
11740 * [221]7967: optimization produces wrong code (ARM)
11744 * [222]7374: __builtin_fabsl broken on alpha
11748 * [223]7370: ICE in fixup_var_refs_1 on s390x
11749 * [224]7409: loop optimization bug on s390x-linux-gnu
11750 * [225]8232: s390x: ICE when using bcmp with int length argument
11754 * [226]7623: SCO OpenServer build fails with machmode.def: undefined
11755 symbol: BITS_PER_UNIT
11757 m68k/Coldfire specific
11759 * [227]8314: crtbegin, crtend need to be multilib'ed for this
11764 * [228]761: Document some undocumented options
11765 * [229]5610: Fix documentation about invoking SSE instructions
11767 * [230]7484: List -Wmissing-declarations as C-only option
11768 * [231]7531: -mcmodel not documented for x86-64
11769 * [232]8120: Update documentation of bad use of ##
11770 __________________________________________________________________
11774 3.2 is a small bug fix release, but there is a change to the
11775 application binary interface (ABI), hence the change to the second part
11776 of the version number.
11778 The main purpose of the 3.2 release is to correct a couple of problems
11779 in the C++ ABI, with the intention of providing a stable interface
11780 going forward. Accordingly, 3.2 is only a small change to 3.1.1.
11786 * [233]7320: g++ 3.2 relocation problem
11787 * [234]7470: vtable: virtual function pointers not in declaration
11792 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t
11793 * [236]6503, [237]6642, [238]7186: Problems with comparing or
11794 subtracting various types of const and non-const iterators
11795 * [239]7216: ambiguity with basic_iostream::traits_type
11796 * [240]7220: problem with basic_istream::ignore(0,delimiter)
11797 * [241]7222: locale::operator==() doesn't work on std::locale("")
11798 * [242]7286: placement operator delete issue
11799 * [243]7442: cxxabi.h does not match the C++ ABI
11800 * [244]7445: poor performance of std::locale::classic() in
11801 multi-threaded applications
11805 * [245]7291: off-by-one in generated inline bzero code for x86-64
11808 For questions related to the use of GCC, please consult these web
11809 pages and the [246]GCC manuals. If that fails, the
11810 [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11811 web pages and the development of GCC are welcome on our developer
11812 list at [248]gcc@gcc.gnu.org. All of [249]our lists have public
11815 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and
11816 distribution of this entire article is permitted in any medium,
11817 provided this notice is preserved.
11819 These pages are [251]maintained by the GCC team. Last modified
11824 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
11825 2. http://gcc.gnu.org/gcc-3.1/changes.html
11826 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html
11827 4. http://gcc.gnu.org/PR3782
11828 5. http://gcc.gnu.org/PR6440
11829 6. http://gcc.gnu.org/PR7050
11830 7. http://gcc.gnu.org/PR7741
11831 8. http://gcc.gnu.org/PR7982
11832 9. http://gcc.gnu.org/PR8068
11833 10. http://gcc.gnu.org/PR8178
11834 11. http://gcc.gnu.org/PR8396
11835 12. http://gcc.gnu.org/PR8674
11836 13. http://gcc.gnu.org/PR9768
11837 14. http://gcc.gnu.org/PR9798
11838 15. http://gcc.gnu.org/PR9799
11839 16. http://gcc.gnu.org/PR9928
11840 17. http://gcc.gnu.org/PR10114
11841 18. http://gcc.gnu.org/PR10352
11842 19. http://gcc.gnu.org/PR10336
11843 20. http://gcc.gnu.org/PR8224
11844 21. http://gcc.gnu.org/PR8613
11845 22. http://gcc.gnu.org/PR8828
11846 23. http://gcc.gnu.org/PR9226
11847 24. http://gcc.gnu.org/PR9853
11848 25. http://gcc.gnu.org/PR9797
11849 26. http://gcc.gnu.org/PR9967
11850 27. http://gcc.gnu.org/PR10116
11851 28. http://gcc.gnu.org/PR10171
11852 29. http://gcc.gnu.org/PR10175
11853 30. http://gcc.gnu.org/PR8316
11854 31. http://gcc.gnu.org/PR9169
11855 32. http://gcc.gnu.org/PR9420
11856 33. http://gcc.gnu.org/PR9459
11857 34. http://gcc.gnu.org/PR9507
11858 35. http://gcc.gnu.org/PR9538
11859 36. http://gcc.gnu.org/PR9602
11860 37. http://gcc.gnu.org/PR9993
11861 38. http://gcc.gnu.org/PR10167
11862 39. http://gcc.gnu.org/PR9652
11863 40. http://gcc.gnu.org/PR10144
11864 41. http://gcc.gnu.org/PR8746
11865 42. http://gcc.gnu.org/PR9888
11866 43. http://gcc.gnu.org/PR9638
11867 44. http://gcc.gnu.org/PR9954
11868 45. http://gcc.gnu.org/PR7784
11869 46. http://gcc.gnu.org/PR7796
11870 47. http://gcc.gnu.org/PR8281
11871 48. http://gcc.gnu.org/PR8366
11872 49. http://gcc.gnu.org/PR8726
11873 50. http://gcc.gnu.org/PR9414
11874 51. http://gcc.gnu.org/PR10067
11875 52. http://gcc.gnu.org/PR7248
11876 53. http://gcc.gnu.org/PR8343
11877 54. http://gcc.gnu.org/PR9732
11878 55. http://gcc.gnu.org/PR10073
11879 56. http://gcc.gnu.org/PR7702
11880 57. http://gcc.gnu.org/PR9671
11881 58. http://gcc.gnu.org/PR8694
11882 59. http://gcc.gnu.org/PR9953
11883 60. http://gcc.gnu.org/PR10271
11884 61. http://gcc.gnu.org/PR6362
11885 62. http://gcc.gnu.org/PR10377
11886 63. http://gcc.gnu.org/PR6955
11887 64. http://gcc.gnu.org/PR5919
11888 65. http://gcc.gnu.org/PR7129
11889 66. http://gcc.gnu.org/PR7507
11890 67. http://gcc.gnu.org/PR7622
11891 68. http://gcc.gnu.org/PR7681
11892 69. http://gcc.gnu.org/PR9528
11893 70. http://gcc.gnu.org/PR8031
11894 71. http://gcc.gnu.org/PR8275
11895 72. http://gcc.gnu.org/PR8332
11896 73. http://gcc.gnu.org/PR8372
11897 74. http://gcc.gnu.org/PR8439
11898 75. http://gcc.gnu.org/PR8442
11899 76. http://gcc.gnu.org/PR8518
11900 77. http://gcc.gnu.org/PR8615
11901 78. http://gcc.gnu.org/PR8663
11902 79. http://gcc.gnu.org/PR8799
11903 80. http://gcc.gnu.org/PR9328
11904 81. http://gcc.gnu.org/PR9465
11905 82. http://gcc.gnu.org/PR47
11906 83. http://gcc.gnu.org/PR6745
11907 84. http://gcc.gnu.org/PR8214
11908 85. http://gcc.gnu.org/PR8493
11909 86. http://gcc.gnu.org/PR8332
11910 87. http://gcc.gnu.org/PR8503
11911 88. http://gcc.gnu.org/PR8727
11912 89. http://gcc.gnu.org/PR7445
11913 90. http://gcc.gnu.org/PR8230
11914 91. http://gcc.gnu.org/PR8399
11915 92. http://gcc.gnu.org/PR8662
11916 93. http://gcc.gnu.org/PR8707
11917 94. http://gcc.gnu.org/PR8708
11918 95. http://gcc.gnu.org/PR8790
11919 96. http://gcc.gnu.org/PR8887
11920 97. http://gcc.gnu.org/PR9076
11921 98. http://gcc.gnu.org/PR9151
11922 99. http://gcc.gnu.org/PR9168
11923 100. http://gcc.gnu.org/PR9269
11924 101. http://gcc.gnu.org/PR9322
11925 102. http://gcc.gnu.org/PR9433
11926 103. http://gcc.gnu.org/PR8032
11927 104. http://gcc.gnu.org/PR8639
11928 105. http://gcc.gnu.org/PR8794
11929 106. http://gcc.gnu.org/PR8832
11930 107. http://gcc.gnu.org/PR8988
11931 108. http://gcc.gnu.org/PR9492
11932 109. http://gcc.gnu.org/PR9267
11933 110. http://gcc.gnu.org/PR8344
11934 111. http://gcc.gnu.org/PR8524
11935 112. http://gcc.gnu.org/PR8880
11936 113. http://gcc.gnu.org/PR9090
11937 114. http://gcc.gnu.org/PR8588
11938 115. http://gcc.gnu.org/PR8599
11939 116. http://gcc.gnu.org/PR9506
11940 117. http://gcc.gnu.org/PR9484
11941 118. http://gcc.gnu.org/PR9292
11942 119. http://gcc.gnu.org/PR9293
11943 120. http://gcc.gnu.org/PR9295
11944 121. http://gcc.gnu.org/PR9296
11945 122. http://gcc.gnu.org/PR9316
11946 123. http://gcc.gnu.org/PR9493
11947 124. http://gcc.gnu.org/PR7341
11948 125. http://gcc.gnu.org/PR8947
11949 126. http://gcc.gnu.org/PR7448
11950 127. http://gcc.gnu.org/PR8882
11951 128. http://gcc.gnu.org/PR7445
11952 129. http://gcc.gnu.org/PR2521
11953 130. http://gcc.gnu.org/PR5661
11954 131. http://gcc.gnu.org/PR6419
11955 132. http://gcc.gnu.org/PR6994
11956 133. http://gcc.gnu.org/PR7150
11957 134. http://gcc.gnu.org/PR7160
11958 135. http://gcc.gnu.org/PR7228
11959 136. http://gcc.gnu.org/PR7266
11960 137. http://gcc.gnu.org/PR7353
11961 138. http://gcc.gnu.org/PR7411
11962 139. http://gcc.gnu.org/PR7478
11963 140. http://gcc.gnu.org/PR7526
11964 141. http://gcc.gnu.org/PR7721
11965 142. http://gcc.gnu.org/PR7803
11966 143. http://gcc.gnu.org/PR7754
11967 144. http://gcc.gnu.org/PR7788
11968 145. http://gcc.gnu.org/PR8031
11969 146. http://gcc.gnu.org/PR8055
11970 147. http://gcc.gnu.org/PR8067
11971 148. http://gcc.gnu.org/PR8134
11972 149. http://gcc.gnu.org/PR8149
11973 150. http://gcc.gnu.org/PR8160
11974 151. http://gcc.gnu.org/PR5607
11975 152. http://gcc.gnu.org/PR6579
11976 153. http://gcc.gnu.org/PR6803
11977 154. http://gcc.gnu.org/PR7176
11978 155. http://gcc.gnu.org/PR7188
11979 156. http://gcc.gnu.org/PR7306
11980 157. http://gcc.gnu.org/PR7461
11981 158. http://gcc.gnu.org/PR7524
11982 159. http://gcc.gnu.org/PR7584
11983 160. http://gcc.gnu.org/PR7676
11984 161. http://gcc.gnu.org/PR7679
11985 162. http://gcc.gnu.org/PR7811
11986 163. http://gcc.gnu.org/PR7961
11987 164. http://gcc.gnu.org/PR8071
11988 165. http://gcc.gnu.org/PR8127
11989 166. http://gcc.gnu.org/PR6745
11990 167. http://gcc.gnu.org/PR8096
11991 168. http://gcc.gnu.org/PR8127
11992 169. http://gcc.gnu.org/PR8218
11993 170. http://gcc.gnu.org/PR8287
11994 171. http://gcc.gnu.org/PR8347
11995 172. http://gcc.gnu.org/PR8348
11996 173. http://gcc.gnu.org/PR8391
11997 174. http://gcc.gnu.org/PR6627
11998 175. http://gcc.gnu.org/PR6631
11999 176. http://gcc.gnu.org/PR7102
12000 177. http://gcc.gnu.org/PR7120
12001 178. http://gcc.gnu.org/PR7209
12002 179. http://gcc.gnu.org/PR7515
12003 180. http://gcc.gnu.org/PR7814
12004 181. http://gcc.gnu.org/PR8467
12005 182. http://gcc.gnu.org/PR4890
12006 183. http://gcc.gnu.org/PR7357
12007 184. http://gcc.gnu.org/PR7358
12008 185. http://gcc.gnu.org/PR7602
12009 186. http://gcc.gnu.org/PR7862
12010 187. http://gcc.gnu.org/PR8190
12011 188. http://gcc.gnu.org/PR8524
12012 189. http://gcc.gnu.org/PR5351
12013 190. http://gcc.gnu.org/PR7591
12014 191. http://gcc.gnu.org/PR6845
12015 192. http://gcc.gnu.org/PR7034
12016 193. http://gcc.gnu.org/PR7124
12017 194. http://gcc.gnu.org/PR7174
12018 195. http://gcc.gnu.org/PR7134
12019 196. http://gcc.gnu.org/PR7375
12020 197. http://gcc.gnu.org/PR7390
12021 198. http://gcc.gnu.org/PR6890
12022 199. http://gcc.gnu.org/PR6981
12023 200. http://gcc.gnu.org/PR7242
12024 201. http://gcc.gnu.org/PR7396
12025 202. http://gcc.gnu.org/PR7630
12026 203. http://gcc.gnu.org/PR7693
12027 204. http://gcc.gnu.org/PR7723
12028 205. http://gcc.gnu.org/PR7951
12029 206. http://gcc.gnu.org/PR8146
12030 207. http://gcc.gnu.org/PR5967
12031 208. http://gcc.gnu.org/PR6984
12032 209. http://gcc.gnu.org/PR7114
12033 210. http://gcc.gnu.org/PR7130
12034 211. http://gcc.gnu.org/PR7133
12035 212. http://gcc.gnu.org/PR7380
12036 213. http://gcc.gnu.org/PR8252
12037 214. http://gcc.gnu.org/PR8451
12038 215. http://gcc.gnu.org/PR7250
12039 216. http://gcc.gnu.org/PR6668
12040 217. http://gcc.gnu.org/PR7151
12041 218. http://gcc.gnu.org/PR7335
12042 219. http://gcc.gnu.org/PR7842
12043 220. http://gcc.gnu.org/PR7856
12044 221. http://gcc.gnu.org/PR7967
12045 222. http://gcc.gnu.org/PR7374
12046 223. http://gcc.gnu.org/PR7370
12047 224. http://gcc.gnu.org/PR7409
12048 225. http://gcc.gnu.org/PR8232
12049 226. http://gcc.gnu.org/PR7623
12050 227. http://gcc.gnu.org/PR8314
12051 228. http://gcc.gnu.org/PR761
12052 229. http://gcc.gnu.org/PR5610
12053 230. http://gcc.gnu.org/PR7484
12054 231. http://gcc.gnu.org/PR7531
12055 232. http://gcc.gnu.org/PR8120
12056 233. http://gcc.gnu.org/PR7320
12057 234. http://gcc.gnu.org/PR7470
12058 235. http://gcc.gnu.org/PR6410
12059 236. http://gcc.gnu.org/PR6503
12060 237. http://gcc.gnu.org/PR6642
12061 238. http://gcc.gnu.org/PR7186
12062 239. http://gcc.gnu.org/PR7216
12063 240. http://gcc.gnu.org/PR7220
12064 241. http://gcc.gnu.org/PR7222
12065 242. http://gcc.gnu.org/PR7286
12066 243. http://gcc.gnu.org/PR7442
12067 244. http://gcc.gnu.org/PR7445
12068 245. http://gcc.gnu.org/PR7291
12069 246. http://gcc.gnu.org/onlinedocs/
12070 247. mailto:gcc-help@gcc.gnu.org
12071 248. mailto:gcc@gcc.gnu.org
12072 249. http://gcc.gnu.org/lists.html
12073 250. http://www.fsf.org/
12074 251. http://gcc.gnu.org/about.html
12075 252. http://validator.w3.org/check/referer
12076 ======================================================================
12077 http://gcc.gnu.org/gcc-3.1/index.html
12082 The [1]GNU project and the GCC developers are pleased to announce the
12083 release of GCC 3.1.1.
12085 The links below still apply to GCC 3.1.1.
12089 The [2]GNU project and the GCC developers are pleased to announce the
12090 release of GCC 3.1.
12092 GCC used to stand for the GNU C Compiler, but since the compiler
12093 supports several other languages aside from C, it now stands for the
12094 GNU Compiler Collection.
12096 A list of [3]successful builds is updated as new information becomes
12099 The GCC developers would like to thank the numerous people that have
12100 contributed [4]new features, improvements, bug fixes, and other changes
12101 as well as test results to GCC. This [5]amazing group of volunteers is
12102 what makes GCC successful.
12104 For additional information about GCC please refer to the [6]GCC project
12105 web site or contact the [7]GCC development mailing list.
12107 To obtain GCC please use [8]our mirror sites, or our CVS server.
12108 __________________________________________________________________
12111 For questions related to the use of GCC, please consult these web
12112 pages and the [9]GCC manuals. If that fails, the
12113 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12114 web pages and the development of GCC are welcome on our developer
12115 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
12118 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
12119 distribution of this entire article is permitted in any medium,
12120 provided this notice is preserved.
12122 These pages are [14]maintained by the GCC team. Last modified
12127 1. http://www.gnu.org/
12128 2. http://www.gnu.org/
12129 3. http://gcc.gnu.org/gcc-3.1/buildstat.html
12130 4. http://gcc.gnu.org/gcc-3.1/changes.html
12131 5. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12132 6. http://gcc.gnu.org/index.html
12133 7. mailto:gcc@gcc.gnu.org
12134 8. http://gcc.gnu.org/mirrors.html
12135 9. http://gcc.gnu.org/onlinedocs/
12136 10. mailto:gcc-help@gcc.gnu.org
12137 11. mailto:gcc@gcc.gnu.org
12138 12. http://gcc.gnu.org/lists.html
12139 13. http://www.fsf.org/
12140 14. http://gcc.gnu.org/about.html
12141 15. http://validator.w3.org/check/referer
12142 ======================================================================
12143 http://gcc.gnu.org/gcc-3.1/changes.html
12144 GCC 3.1 Release Series
12145 Changes, New Features, and Fixes
12147 Additional changes in GCC 3.1.1
12149 * A bug related to how structures and unions are returned has been
12150 fixed for powerpc-*-netbsd*.
12151 * An important bug in the implementation of -fprefetch-loop-arrays
12152 has been fixed. Previously the optimization prefetched random
12153 blocks of memory for most targets except for i386.
12154 * The Java compiler now compiles Java programs much faster and also
12155 works with parallel make.
12156 * Nested functions have been fixed for mips*-*-netbsd*.
12157 * Some missing floating point support routines have beed added for
12159 * This [1]message gives additional information about the bugs fixed
12164 * The -traditional C compiler option has been deprecated and will be
12165 removed in GCC 3.3. (It remains possible to preprocess non-C code
12166 with the traditional preprocessor.)
12167 * The default debugging format for most ELF platforms (including
12168 GNU/Linux and FreeBSD; notable exception is Solaris) has changed
12169 from stabs to DWARF2. This requires GDB 5.1.1 or later.
12171 General Optimizer Improvements
12173 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat,
12174 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure
12175 for profile driven optimizations.
12176 Options -fprofile-arcs and -fbranch-probabilities can now be used
12177 to improve speed of the generated code by profiling the actual
12178 program behaviour on typical runs. In the absence of profile info
12179 the compiler attempts to guess the profile statically.
12180 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to
12181 monitor performance of the generated code.
12182 According to the SPECInt2000 results on an AMD Athlon CPU, the code
12183 generated by GCC 3.1 is 6% faster on the average (8.2% faster with
12184 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0
12185 is about 2.1% faster compared to 2.95.3. Tests were done using the
12186 -O2 -march=athlon command-line options.
12187 * Alexandre Oliva, of Red Hat, has generalized the tree inlining
12188 infrastructure developed by CodeSourcery, LLC for the C++ front
12189 end, so that it is now used in the C front end too. Inlining
12190 functions as trees exposes them earlier to the compiler, giving it
12191 more opportunities for optimization.
12192 * Support for data prefetching instructions has been added to the GCC
12193 back end and several targets. A new __builtin_prefetch intrinsic is
12194 available to explicitly insert prefetch instructions and
12195 experimental support for loop array prefetching has been added (see
12196 -fprefetch-loop-array documentation).
12197 * Support for emitting debugging information for macros has been
12198 added for DWARF2. It is activated using -g3.
12200 New Languages and Language specific improvements
12204 * A few more [4]ISO C99 features.
12205 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0.
12206 * The preprocessor's symbol table has been merged with the symbol
12207 table of the C, C++ and Objective-C front ends.
12208 * The preprocessor consumes less memory than the preprocessor in GCC
12209 3.0, often significantly so. On normal input files, it typically
12210 consumes less memory than pre-3.0 cccp-based GCC, too.
12214 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std
12215 was a workaround to allow std compliant code to work with the
12216 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant.
12217 * The C++ ABI has been fixed so that void (A::*)() const is mangled
12218 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only
12219 affects pointer to cv-qualified member function types.
12220 * The C++ ABI has been changed to correctly handle this code:
12222 void operator delete[] (void *, size_t);
12225 struct B : public A {
12230 The amount of storage allocated for the array will be greater than
12231 it was in 3.0, in order to store the number of elements in the
12232 array, so that the correct size can be passed to operator delete[]
12233 when the array is deleted. Previously, the value passed to operator
12234 delete[] was unpredictable.
12235 This change will only affect code that declares a two-argument
12236 operator delete[] with a second parameter of type size_t in a base
12237 class, and does not override that definition in a derived class.
12238 * The C++ ABI has been changed so that:
12240 void operator delete[] (void *, size_t);
12241 void operator delete[] (void *);
12244 does not cause unnecessary storage to be allocated when an array of
12245 A objects is allocated.
12246 This change will only affect code that declares both of these forms
12247 of operator delete[], and declared the two-argument form before the
12249 * The C++ ABI has been changed so that when a parameter is passed by
12250 value, any cleanup for that parameter is performed in the caller,
12251 as specified by the ia64 C++ ABI, rather than the called function
12252 as before. As a result, classes with a non-trivial destructor but a
12253 trivial copy constructor will be passed and returned by invisible
12254 reference, rather than by bitwise copy as before.
12255 * G++ now supports the "named return value optimization": for code
12263 G++ will allocate a in the return value slot, so that the return
12264 becomes a no-op. For this to work, all return statements in the
12265 function must return the same variable.
12266 * Improvements to the C++ library are listed in [5]the libstdc++-v3
12271 * Annoying linker warnings (due to incorrect code being generated)
12273 * If a class method cannot be found, the compiler no longer issues a
12274 warning if a corresponding instance method exists in the root
12276 * Forward @protocol declarations have been fixed.
12277 * Loading of categories has been fixed in certain situations (GNU run
12279 * The class lookup in the run-time library has been rewritten so that
12280 class method dispatch is more than twice as fast as it used to be
12281 (GNU run time only).
12285 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and
12287 * Property files and other system resources can be compiled into
12288 executables which use libgcj using the new gcj --resource feature.
12289 * libgcj has been ported to more platforms. In particular there is
12290 now a mostly-functional mingw32 (Windows) target port.
12291 * JNI and CNI invocation interfaces were implemented, so gcj-compiled
12292 Java code can now be called from a C/C++ application.
12293 * gcj can now use builtin functions for certain known methods, for
12295 * gcj can now automatically remove redundant array-store checks in
12297 * The --no-store-checks optimization option was added. This can be
12298 used to omit runtime store checks for code which is known not to
12299 throw ArrayStoreException
12300 * The following third party interface standards were added to libgcj:
12301 org.w3c.dom and org.xml.sax.
12302 * java.security has been merged with GNU Classpath. The new package
12303 is now JDK 1.2 compliant, and much more complete.
12304 * A bytecode verifier was added to the libgcj interpreter.
12305 * java.lang.Character was rewritten to comply with the Unicode 3.0
12306 standard, and improve performance.
12307 * Partial support for many more locales was added to libgcj.
12308 * Socket timeouts have been implemented.
12309 * libgcj has been merged into a single shared library. There are no
12310 longer separate shared libraries for the garbage collector and
12312 * Several performance improvements were made to gcj and libgcj:
12313 + Hash synchronization (thin locks)
12314 + A special allocation path for finalizer-free objects
12315 + Thread-local allocation
12316 + Parallel GC, and other GC tweaks
12320 Fortran improvements are listed in [6]the Fortran documentation.
12324 [7]Ada Core Technologies, Inc, has contributed its GNAT Ada 95 front
12325 end and associated tools. The GNAT compiler fully implements the Ada
12326 language as defined by the ISO/IEC 8652 standard.
12328 Please note that the integration of the Ada front end is still work in
12331 New Targets and Target Specific Improvements
12333 * Hans-Peter Nilsson has contributed a port to [8]MMIX, the CPU
12334 architecture used in new editions of Donald E. Knuth's The Art of
12335 Computer Programming.
12336 * [9]Axis Communications has contributed its port to the CRIS CPU
12337 architecture, used in the ETRAX system-on-a-chip series. See
12338 [10]Axis' developer site for technical information.
12339 * Alexandre Oliva, of Red Hat, has contributed a port to the
12340 [11]SuperH SH5 64-bit RISC microprocessor architecture, extending
12341 the existing SH port.
12342 * UltraSPARC is fully supported in 64-bit mode. The option -m64
12344 * For compatibility with the Sun compiler #pragma redefine_extname
12345 has been implemented on Solaris.
12346 * The x86 back end has had some noticeable work done to it.
12347 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas
12348 Jaeger have contributed a port to the AMD x86-64 architecture.
12349 For more information on x86-64 see [12]http://www.x86-64.org.
12350 + The compiler now supports MMX, 3DNow!, SSE, and SSE2
12351 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will
12352 enable the respective instruction sets. Intel C++ compatible
12353 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics
12354 will be added in next major release.
12355 + Following those improvements, targets for Pentium MMX, K6-2,
12356 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were
12357 added. Refer to the documentation on -march= and -mcpu=
12358 options for details.
12359 + For those targets that support it, -mfpmath=sse will cause the
12360 compiler to generate SSE/SSE2 instructions for floating point
12361 math instead of x87 instructions. Usually, this will lead to
12362 quicker code -- especially on the Pentium 4. Note that only
12363 scalar floating point instructions are used and GCC does not
12364 exploit SIMD features yet.
12365 + Prefetch support has been added to the Pentium III, Pentium 4,
12366 K6-2, K6-3, and Athlon series.
12367 + Code generated for floating point to integer conversions has
12368 been improved leading to better performance of many 3D
12370 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support.
12371 * C++ support for AIX has been improved.
12372 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the
12373 PowerPC port supporting the AltiVec programming model (SIMD). The
12374 support, though presently useful, is experimental and is expected
12375 to stabilize for 3.2. The support is written to conform to
12376 Motorola's AltiVec specs. See -maltivec.
12380 Support for a number of older systems has been declared obsolete in GCC
12381 3.1. Unless there is activity to revive them, the next release of GCC
12382 will have their sources permanently removed.
12384 All configurations of the following processor architectures have been
12386 * MIL-STD-1750A, 1750a-*-*
12387 * AMD A29k, a29k-*-*
12388 * Convex, c*-convex-*
12389 * Clipper, clipper-*-*
12391 * Intel i860, i860-*-*
12392 * Sun picoJava, pj-*-* and pjl-*-*
12393 * Western Electric 32000, we32k-*-*
12395 Most configurations of the following processor architectures have been
12396 declared obsolete, but we are preserving a few systems which may have
12397 active developers. It is unlikely that the remaining systems will
12398 survive much longer unless we see definite signs of port activity.
12399 * Motorola 88000 except
12400 + Generic a.out, m88k-*-aout*
12401 + Generic SVR4, m88k-*-sysv4
12402 + OpenBSD, m88k-*-openbsd*
12404 + NetBSD, ns32k-*-netbsd*
12405 + OpenBSD, ns32k-*-openbsd*.
12407 + OpenBSD, romp-*-openbsd*.
12409 Finally, only some configurations of these processor architectures are
12412 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka
12413 alpha*-*-osf[45], are still supported.)
12415 + RISCiX, arm-*-riscix*.
12417 + 386BSD, i?86-*-bsd*
12418 + Chorus, i?86-*-chorusos*
12419 + DG/UX, i?86-*-dgux*
12420 + FreeBSD 1.x, i?86-*-freebsd1.*
12421 + IBM AIX, i?86-*-aix*
12422 + ISC UNIX, i?86-*-isc*
12423 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld*
12424 + NEXTstep, i?86-next-*
12425 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose*
12426 + RTEMS/coff, i?86-*-rtemscoff*
12427 + RTEMS/go32, i?86-go32-rtems*
12428 + Sequent/BSD, i?86-sequent-bsd*
12429 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and
12430 i?86-sequent-sysv3*
12431 + SunOS, i?86-*-sunos*
12433 + Altos, m68[k0]*-altos-*
12434 + Apollo, m68[k0]*-apollo-*
12435 + Apple A/UX, m68[k0]*-apple-*
12436 + Bull, m68[k0]*-bull-*
12437 + Convergent, m68[k0]*-convergent-*
12438 + Generic SVR3, m68[k0]*-*-sysv3*
12439 + ISI, m68[k0]*-isi-*
12440 + LynxOS, m68[k0]*-*-lynxos*
12441 + NEXT, m68[k0]*-next-*
12442 + RTEMS/coff, m68[k0]*-*-rtemscoff*
12443 + Sony, m68[k0]*-sony-*
12445 + DEC Ultrix, mips-*-ultrix* and mips-dec-*
12446 + Generic BSD, mips-*-bsd*
12447 + Generic System V, mips-*-sysv*
12448 + IRIX before version 5, mips-sgi-irix[1234]*
12449 + RiscOS, mips-*-riscos*
12450 + Sony, mips-sony-*
12451 + Tandem, mips-tandem-*
12453 + RTEMS/a.out, sparc-*-rtemsaout*.
12455 Documentation improvements
12457 * The old manual ("Using and Porting the GNU Compiler Collection")
12458 has been replaced by a users manual ("Using the GNU Compiler
12459 Collection") and a separate internals reference manual ("GNU
12460 Compiler Collection Internals").
12461 * More complete and much improved documentation about GCC's internal
12462 representation used by the C and C++ front ends.
12463 * Many cleanups and improvements in general.
12466 For questions related to the use of GCC, please consult these web
12467 pages and the [13]GCC manuals. If that fails, the
12468 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12469 web pages and the development of GCC are welcome on our developer
12470 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
12473 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
12474 distribution of this entire article is permitted in any medium,
12475 provided this notice is preserved.
12477 These pages are [18]maintained by the GCC team. Last modified
12482 1. http://gcc.gnu.org/ml/gcc/2002-07/msg01208.html
12483 2. http://gcc.gnu.org/news/profiledriven.html
12484 3. http://gcc.gnu.org/benchmarks/
12485 4. http://gcc.gnu.org/gcc-3.1/c99status.html
12486 5. http://gcc.gnu.org/onlinedocs/libstdc++/faq.html
12487 6. http://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html
12488 7. http://www.adacore.com/
12489 8. http://www-cs-faculty.stanford.edu/~knuth/mmix.html
12490 9. http://www.axis.com/
12491 10. http://developer.axis.com/
12492 11. http://www.superh.com/
12493 12. http://www.x86-64.org/
12494 13. http://gcc.gnu.org/onlinedocs/
12495 14. mailto:gcc-help@gcc.gnu.org
12496 15. mailto:gcc@gcc.gnu.org
12497 16. http://gcc.gnu.org/lists.html
12498 17. http://www.fsf.org/
12499 18. http://gcc.gnu.org/about.html
12500 19. http://validator.w3.org/check/referer
12501 ======================================================================
12502 http://gcc.gnu.org/gcc-3.0/gcc-3.0.html
12507 The [1]GNU project and the GCC developers are pleased to announce the
12508 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0
12511 GCC used to stand for the GNU C Compiler, but since the compiler
12512 supports several other languages aside from C, it now stands for the
12513 GNU Compiler Collection.
12515 GCC 3.0.x has several new optimizations, new targets, new languages and
12516 many other new features, relative to GCC 2.95.x. See the [2]new
12517 features page for a more complete list.
12519 A list of [3]successful builds is updated as new information becomes
12522 The GCC developers would like to thank the numerous people that have
12523 contributed new features, test results, bug fixes, etc to GCC. This
12524 [4]amazing group of volunteers is what makes GCC successful.
12526 And finally, we can't in good conscience fail to mention some
12527 [5]caveats to using GCC 3.0.x.
12529 For additional information about GCC please refer to the [6]GCC project
12530 web site or contact the [7]GCC development mailing list.
12532 To obtain GCC please use [8]our mirror sites, or our CVS server.
12533 __________________________________________________________________
12535 Previous 3.0.x Releases
12537 December 20, 2001: GCC 3.0.3 has been released.
12538 October 25, 2001: GCC 3.0.2 has been released.
12539 August 20, 2001: GCC 3.0.1 has been released.
12540 June 18, 2001: GCC 3.0 has been released.
12543 For questions related to the use of GCC, please consult these web
12544 pages and the [9]GCC manuals. If that fails, the
12545 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12546 web pages and the development of GCC are welcome on our developer
12547 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
12550 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
12551 distribution of this entire article is permitted in any medium,
12552 provided this notice is preserved.
12554 These pages are [14]maintained by the GCC team. Last modified
12559 1. http://www.gnu.org/
12560 2. http://gcc.gnu.org/gcc-3.0/features.html
12561 3. http://gcc.gnu.org/gcc-3.0/buildstat.html
12562 4. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12563 5. http://gcc.gnu.org/gcc-3.0/caveats.html
12564 6. http://gcc.gnu.org/index.html
12565 7. mailto:gcc@gcc.gnu.org
12566 8. http://gcc.gnu.org/mirrors.html
12567 9. http://gcc.gnu.org/onlinedocs/
12568 10. mailto:gcc-help@gcc.gnu.org
12569 11. mailto:gcc@gcc.gnu.org
12570 12. http://gcc.gnu.org/lists.html
12571 13. http://www.fsf.org/
12572 14. http://gcc.gnu.org/about.html
12573 15. http://validator.w3.org/check/referer
12574 ======================================================================
12575 http://gcc.gnu.org/gcc-3.0/features.html
12576 GCC 3.0 New Features
12578 Additional changes in GCC 3.0.4
12580 * GCC 3.0 now supports newer versions of the [1]NetBSD operating
12581 system, which use the ELF object file format, on x86 processors.
12582 * Correct debugging information is generated from functions that have
12583 lines from multiple files (e.g. yacc output).
12584 * A fix for whitespace handling in the -traditional preprocessor,
12585 which can affect Fortran.
12586 * Fixes to the exception handling runtime.
12587 * More fixes for bad code generation in C++.
12588 * A fix for shared library generation under AIX 4.3.
12589 * Documentation updates.
12590 * Port of GCC to Tensilica's Xtensa processor contributed.
12591 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link).
12593 Additional changes in GCC 3.0.3
12595 * A fix to correct an accidental change to the PowerPC ABI.
12596 * Fixes for bad code generation on a variety of architectures.
12597 * Improvements to the debugging information generated for C++
12599 * Fixes for bad code generation in C++.
12600 * A fix to avoid crashes in the C++ demangler.
12601 * A fix to the C++ standard library to avoid buffer overflows.
12602 * Miscellaneous improvements for a variety of architectures.
12604 Additional changes in GCC 3.0.2
12606 * Fixes for bad code generation during loop unrolling.
12607 * Fixes for bad code generation by the sibling call optimization.
12608 * Minor improvements to x86 code generation.
12609 * Implementation of function descriptors in C++ vtables for IA64.
12610 * Numerous minor bug-fixes.
12612 Additional changes in GCC 3.0.1
12614 * C++ fixes for incorrect code-generation.
12615 * Improved cross-compiling support for the C++ standard library.
12616 * Fixes for some embedded targets that worked in GCC 2.95.3, but not
12618 * Fixes for various exception-handling bugs.
12619 * A port to the S/390 architecture.
12621 General Optimizer Improvements
12623 * [2]Basic block reordering pass.
12624 * New if-conversion pass with support for conditional (predicated)
12626 * New tail call and sibling call elimination optimizations.
12627 * New register renaming pass.
12628 * New (experimental) [3]static single assignment (SSA) representation
12630 * New dead-code elimination pass implemented using the SSA
12632 * [4]Global null pointer test elimination.
12633 * [5]Global code hoisting/unification.
12634 * More builtins and optimizations for stdio.h, string.h and old BSD
12635 functions, as well as for ISO C99 functions.
12636 * New builtin __builtin_expect for giving hints to the branch
12639 New Languages and Language specific improvements
12641 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated
12642 and supported, including the run-time library containing most
12643 common non-GUI Java classes, a bytecode interpreter, and the Boehm
12644 conservative garbage collector. Many bugs have been fixed. GCJ can
12645 compile Java source or Java bytecodes to either native code or Java
12646 class files, and supports native methods written in either the
12647 standard JNI or the more efficient and convenient CNI.
12648 * Here is a [6]partial list of C++ improvements, both new features
12649 and those no longer supported.
12650 * New C++ ABI. On the IA-64 platform GCC is capable of
12651 inter-operating with other IA-64 compilers.
12652 * The new ABI also significantly reduces the size of symbol and debug
12654 * New [7]C++ support library and many C++ bug fixes, vastly improving
12655 our conformance to the ISO C++ standard.
12656 * New [8]inliner for C++.
12657 * Rewritten C preprocessor, integrated into the C, C++ and Objective
12658 C compilers, with very many improvements including ISO C99 support
12659 and [9]improvements to dependency generation.
12660 * Support for more [10]ISO C99 features.
12661 * Many improvements to support for checking calls to format functions
12662 such as printf and scanf, including support for ISO C99 format
12663 features, extensions from the Single Unix Specification and GNU
12664 libc 2.2, checking of strfmon formats and features to assist in
12665 auditing for format string security bugs.
12666 * New warnings for C code that may have undefined semantics because
12667 of violations of sequence point rules in the C standard (such as a
12668 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall.
12669 * Additional warning option -Wfloat-equal.
12670 * Improvements to -Wtraditional.
12671 * Fortran improvements are listed in [11]the Fortran documentation.
12673 New Targets and Target Specific Improvements
12675 * New x86 back-end, generating much improved code.
12676 * Support for a generic i386-elf target contributed.
12677 * New option to emit x86 assembly code using Intel style syntax
12679 * HPUX 11 support contributed.
12680 * Improved PowerPC code generation, including scheduled prologue and
12682 * Port of GCC to Intel's IA-64 processor contributed.
12683 * Port of GCC to Motorola's MCore 210 and 340 contributed.
12684 * New unified back-end for Arm, Thumb and StrongArm contributed.
12685 * Port of GCC to Intel's XScale processor contributed.
12686 * Port of GCC to Atmel's AVR microcontrollers contributed.
12687 * Port of GCC to Mitsubishi's D30V processor contributed.
12688 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300
12689 processor family) contributed.
12690 * Port of GCC to Fujitsu's FR30 processor contributed.
12691 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed.
12692 * Port of GCC to Sun's picoJava processor core contributed.
12694 Documentation improvements
12696 * Substantially rewritten and improved C preprocessor manual.
12697 * Many improvements to other documentation.
12698 * Manpages for gcc, cpp and gcov are now generated automatically from
12699 the master Texinfo manual, eliminating the problem of manpages
12700 being out of date. (The generated manpages are only extracts from
12701 the full manual, which is provided in Texinfo form, from which
12702 info, HTML, other formats and a printed manual can be generated.)
12703 * Generated info files are included in the release tarballs alongside
12704 their Texinfo sources, avoiding problems on some platforms with
12705 building makeinfo as part of the GCC distribution.
12707 Other significant improvements
12709 * Garbage collection used internally by the compiler for most memory
12710 allocation instead of obstacks.
12711 * Lengauer and Tarjan algorithm used for computing dominators in the
12712 CFG. This algorithm can be significantly faster and more space
12713 efficient than our older algorithm.
12714 * gccbug script provided to assist in submitting bug reports to our
12715 bug tracking system. (Bug reports previously submitted directly to
12716 our mailing lists, for which you received no bug tracking number,
12717 should be submitted again using gccbug if you can reproduce the
12718 problem with GCC 3.0.)
12719 * The internal libgcc library is [12]built as a shared library on
12720 systems that support it.
12721 * Extensive testsuite included with GCC, with many new tests. In
12722 addition to tests for GCC bugs that have been fixed, many tests
12723 have been added for language features, compiler warnings and
12725 * Additional language-independent warning options -Wpacked, -Wpadded,
12726 -Wunreachable-code and -Wdisabled-optimization.
12727 * Target-independent options -falign-functions, -falign-loops and
12730 Plus a great many bug fixes and almost all the [13]features found in
12734 For questions related to the use of GCC, please consult these web
12735 pages and the [14]GCC manuals. If that fails, the
12736 [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12737 web pages and the development of GCC are welcome on our developer
12738 list at [16]gcc@gcc.gnu.org. All of [17]our lists have public
12741 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and
12742 distribution of this entire article is permitted in any medium,
12743 provided this notice is preserved.
12745 These pages are [19]maintained by the GCC team. Last modified
12750 1. http://www.netbsd.org/
12751 2. http://gcc.gnu.org/news/reorder.html
12752 3. http://gcc.gnu.org/news/ssa.html
12753 4. http://gcc.gnu.org/news/null.html
12754 5. http://gcc.gnu.org/news/unify.html
12755 6. http://gcc.gnu.org/gcc-3.0/c++features.html
12756 7. http://gcc.gnu.org/libstdc++/
12757 8. http://gcc.gnu.org/news/inlining.html
12758 9. http://gcc.gnu.org/news/dependencies.html
12759 10. http://gcc.gnu.org/gcc-3.0/c99status.html
12760 11. http://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
12761 12. http://gcc.gnu.org/gcc-3.0/libgcc.html
12762 13. http://gcc.gnu.org/gcc-2.95/features.html
12763 14. http://gcc.gnu.org/onlinedocs/
12764 15. mailto:gcc-help@gcc.gnu.org
12765 16. mailto:gcc@gcc.gnu.org
12766 17. http://gcc.gnu.org/lists.html
12767 18. http://www.fsf.org/
12768 19. http://gcc.gnu.org/about.html
12769 20. http://validator.w3.org/check/referer
12770 ======================================================================
12771 http://gcc.gnu.org/gcc-3.0/caveats.html
12774 * -fstrict-aliasing is now part of -O2 and higher optimization
12775 levels. This allows the compiler to assume the strictest aliasing
12776 rules applicable to the language being compiled. For C and C++,
12777 this activates optimizations based on the type of expressions. This
12778 optimization may thus break old, non-compliant code.
12779 * Enumerations are now properly promoted to int in function
12780 parameters and function returns. Normally this change is not
12781 visible, but when using -fshort-enums this is an ABI change.
12782 * The undocumented extension that allowed C programs to have a label
12783 at the end of a compound statement has been deprecated and may be
12784 removed in a future version. Programs that now generate a warning
12785 about this may be fixed by adding a null statement (a single
12786 semicolon) after the label.
12787 * The poorly documented extension that allowed string constants in C,
12788 C++ and Objective C to contain unescaped newlines has been
12789 deprecated and may be removed in a future version. Programs using
12790 this extension may be fixed in several ways: the bare newline may
12791 be replaced by \n, or preceded by \n\, or string concatenation may
12792 be used with the bare newline preceded by \n" and " placed at the
12793 start of the next line.
12794 * The Chill compiler is not included in GCC 3.0, because of the lack
12795 of a volunteer to convert it to use garbage collection.
12796 * Certain non-standard iostream methods from earlier versions of
12797 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach,
12798 ostream::form, and istream::gets.
12799 * The new C++ ABI is not yet fully supported by current (as of
12800 2001-07-01) releases and development versions of GDB, or any
12801 earlier versions. There is a problem setting breakpoints by line
12802 number, and other related issues that have been fixed in GCC 3.0
12803 but not yet handled in GDB:
12804 [1]http://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
12807 For questions related to the use of GCC, please consult these web
12808 pages and the [2]GCC manuals. If that fails, the
12809 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12810 web pages and the development of GCC are welcome on our developer
12811 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
12813 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
12814 distribution of this entire article is permitted in any medium,
12815 provided this notice is preserved.
12817 These pages are [7]maintained by the GCC team. Last modified
12822 1. http://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
12823 2. http://gcc.gnu.org/onlinedocs/
12824 3. mailto:gcc-help@gcc.gnu.org
12825 4. mailto:gcc@gcc.gnu.org
12826 5. http://gcc.gnu.org/lists.html
12827 6. http://www.fsf.org/
12828 7. http://gcc.gnu.org/about.html
12829 8. http://validator.w3.org/check/referer
12830 ======================================================================
12831 http://gcc.gnu.org/gcc-2.95/index.html
12834 March 16, 2001: The GNU project and the GCC developers are pleased to
12835 announce the release of GCC version 2.95.3.
12849 July 31, 1999. This is the first release of GCC since the April
12850 1999 GCC/EGCS reunification and includes nearly a year's worth
12851 of new development and bugfixes.
12853 References and Acknowledgements
12855 GCC used to stand for the GNU C Compiler, but since the compiler
12856 supports several other languages aside from C, it now stands for the
12857 GNU Compiler Collection.
12859 The whole suite has been extensively [1]regression tested and
12860 [2]package tested. It should be reliable and suitable for widespread
12863 The compiler has several new optimizations, new targets, new languages
12864 and other new features. See the [3]new features page for a more
12865 complete list of new features found in the GCC 2.95 releases.
12867 The sources include installation instructions in both HTML and
12868 plaintext forms in the install directory in the distribution. However,
12869 the most up to date [4]installation instructions and [5]build/test
12870 status are on the web pages. We will update those pages as new
12871 information becomes available.
12873 The GCC developers would like to thank the numerous people that have
12874 contributed new features, test results, bugfixes, etc to GCC. This
12875 [6]amazing group of volunteers is what makes GCC successful.
12877 And finally, we can't in good conscience fail to mention some
12878 [7]caveats to using GCC 2.95.
12880 Download GCC 2.95 from one of our many [8]mirror sites.
12882 For additional information about GCC please see the [9]GCC project web
12883 server or contact the [10]GCC development mailing list.
12886 For questions related to the use of GCC, please consult these web
12887 pages and the [11]GCC manuals. If that fails, the
12888 [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12889 web pages and the development of GCC are welcome on our developer
12890 list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
12893 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
12894 distribution of this entire article is permitted in any medium,
12895 provided this notice is preserved.
12897 These pages are [16]maintained by the GCC team. Last modified
12902 1. http://gcc.gnu.org/gcc-2.95/regress.html
12903 2. http://gcc.gnu.org/gcc-2.95/othertest.html
12904 3. http://gcc.gnu.org/gcc-2.95/features.html
12905 4. http://gcc.gnu.org/install/
12906 5. http://gcc.gnu.org/gcc-2.95/buildstat.html
12907 6. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12908 7. http://gcc.gnu.org/gcc-2.95/caveats.html
12909 8. http://gcc.gnu.org/mirrors.html
12910 9. http://gcc.gnu.org/index.html
12911 10. mailto:gcc@gcc.gnu.org
12912 11. http://gcc.gnu.org/onlinedocs/
12913 12. mailto:gcc-help@gcc.gnu.org
12914 13. mailto:gcc@gcc.gnu.org
12915 14. http://gcc.gnu.org/lists.html
12916 15. http://www.fsf.org/
12917 16. http://gcc.gnu.org/about.html
12918 17. http://validator.w3.org/check/referer
12919 ======================================================================
12920 http://gcc.gnu.org/gcc-2.95/features.html
12921 GCC 2.95 New Features
12923 * General Optimizer Improvements:
12924 + [1]Localized register spilling to improve speed and code
12925 density especially on small register class machines.
12926 + [2]Global CSE using lazy code motion algorithms.
12927 + [3]Improved global constant/copy propagation.
12928 + [4]Improved control flow graph analysis and manipulation.
12929 + [5]Local dead store elimination.
12930 + [6]Memory Load hoisting/store sinking in loops.
12931 + [7]Type based alias analysis is enabled by default. Note this
12932 feature will expose bugs in the Linux kernel. Please refer to
12933 the FAQ (as shipped with GCC 2.95) for additional information
12935 + Major revamp of GIV detection, combination and simplification
12936 to improve loop performance.
12937 + Major improvements to register allocation and reloading.
12938 * New Languages and Language specific improvements
12939 + [8]Many C++ improvements.
12940 + [9]Many Fortran improvements.
12941 + [10]Java front-end has been integrated. [11]runtime library is
12942 available separately.
12943 + [12]ISO C99 support
12944 + [13]Chill front-end and runtime has been integrated.
12945 + Boehm garbage collector support in libobjc.
12946 + More support for various pragmas which appear in vendor
12948 * New Targets and Target Specific Improvements
12949 + [14]SPARC backend rewrite.
12950 + -mschedule=8000 will optimize code for PA8000 class
12951 processors; -mpa-risc-2-0 will generate code for PA2.0
12953 + Various micro-optimizations for the ia32 port. K6
12955 + Compiler will attempt to align doubles in the stack on the
12957 + Alpha EV6 support
12959 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403.
12960 -mcpu=e603e was added to do -mcpu=603e and -msoft-float.
12965 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix,
12967 + vxWorks targets include support for vxWorks threads
12968 + StrongARM 110 and ARM9 support added. ARM Scheduling
12969 parameters rewritten.
12970 + Various changes to the MIPS port to avoid assembler macros,
12971 which in turn improves performance
12972 + Various performance improvements to the i960 port.
12973 + Major rewrite of ns32k port
12974 * Other significant improvements
12975 + [15]Ability to dump cfg information and display it using vcg.
12976 + The new faster scheme for fixing vendor header files is
12977 enabled by default.
12978 + Experimental internationalization support.
12979 + multibyte character support
12980 + Some compile-time speedups for pathological problems
12981 + Better support for complex types
12982 * Plus the usual mountain of bugfixes
12983 * Core compiler is based on the gcc2 development tree from Sept 30,
12984 1998, so we have all of the [16]features found in GCC 2.8.
12986 Additional Changes in GCC 2.95.1
12988 * Generic bugfixes and improvements
12989 + Various documentation fixes related to the GCC/EGCS merger.
12990 + Fix memory management bug which could lead to spurious aborts,
12991 core dumps or random parsing errors in the compiler.
12992 + Fix a couple bugs in the dwarf1 and dwarf2 debug record
12994 + Fix infinite loop in the CSE optimizer.
12995 + Avoid undefined behavior in compiler FP emulation code
12996 + Fix install problem when prefix is overridden on the make
12998 + Fix problem with unwanted installation of assert.h on some
13000 + Fix problem with finding the wrong assembler in a single tree
13002 + Avoid increasing the known alignment of a register that is
13003 already known to be a pointer.
13004 * Platform specific bugfixes and improvements
13005 + Codegen bugfix for prologue/epilogue for cpu32 target.
13006 + Fix long long code generation bug for the Coldfire target.
13007 + Fix various aborts in the SH compiler.
13008 + Fix bugs in libgcc support library for the SH.
13009 + Fix alpha ev6 code generation bug.
13010 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on
13012 + Fix -fpic code generation bug for rs6000/ppc svr4 targets.
13013 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4
13015 + Fix weak symbol handling for rs6000/ppc svr4 targets.
13016 + Fix various problems with 64bit code generation for the
13018 + Fix codegen bug which caused tetex to be mis-compiled on the
13020 + Fix compiler abort in new cfg code exposed by x86 port.
13021 + Fix out of range array reference in code convert flat
13022 registers to the x87 stacked FP register file.
13023 + Fix minor vxworks configuration bug.
13024 + Fix return type of bsearch for SunOS 4.x.
13025 * Language & Runtime specific fixes.
13026 + The G++ signature extension has been deprecated. It will be
13027 removed in the next major release of G++. Use of signatures
13028 will result in a warning from the compiler.
13029 + Several bugs relating to templates and namespaces were fixed.
13030 + A bug that caused crashes when combining templates with -g on
13031 DWARF1 platforms was fixed.
13032 + Pointers-to-members, virtual functions, and multiple
13033 inheritance should now work together correctly.
13034 + Some code-generation bugs relating to function try blocks were
13036 + G++ is a little bit more lenient with certain archaic
13037 constructs than in GCC 2.95.
13038 + Fix to prevent shared library version #s from bring truncated
13040 + Fix missing std:: in the libstdc++ library.
13041 + Fix stream locking problems in libio.
13042 + Fix problem in java compiler driver.
13044 Additional Changes in GCC 2.95.2
13046 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While
13047 the optimizations performed by -fstrict-aliasing are valid according to
13048 the C and C++ standards, the optimization have caused some problems,
13049 particularly with old non-conforming code.
13051 The GCC developers are experimenting with ways to warn users about code
13052 which violates the C/C++ standards, but those warnings are not ready
13053 for widespread use at this time. Rather than wait for those warnings
13054 the GCC developers have chosen to disable -fstrict-aliasing by default
13055 for the GCC 2.95.2 release.
13057 We strongly encourage developers to find and fix code which violates
13058 the C/C++ standards as -fstrict-aliasing may be enabled by default in
13059 future releases. Use the option -fstrict-aliasing to re-enable these
13061 * Generic bugfixes and improvements
13062 + Fix incorrectly optimized memory reference in global common
13063 subexpression elimination (GCSE) optimization pass.
13064 + Fix code generation bug in regmove.c in which it could
13065 incorrectly change a "const" value.
13066 + Fix bug in optimization of conditionals involving volatile
13068 + Avoid over-allocation of stack space for some procedures.
13069 + Fixed bug in the compiler which caused incorrect optimization
13070 of an obscure series of bit manipulations, shifts and
13072 + Fixed register allocator bug which caused teTeX to be
13073 mis-compiled on SPARC targets.
13074 + Avoid incorrect optimization of degenerate case statements for
13075 certain targets such as the ARM.
13076 + Fix out of range memory reference in the jump optimizer.
13077 + Avoid dereferencing null pointer in fix-header.
13078 + Fix test for GCC specific features so that it is possible to
13079 bootstrap with gcc-2.6.2 and older versions of GCC.
13080 + Fix typo in scheduler which could potentially cause out of
13081 range memory accesses.
13082 + Avoid incorrect loop reversal which caused incorrect code for
13083 certain loops on PowerPC targets.
13084 + Avoid incorrect optimization of switch statements on certain
13085 targets (for example the ARM).
13086 * Platform specific bugfixes and improvements
13087 + Work around bug in Sun V5.0 compilers which caused bootstrap
13088 comparison failures on SPARC targets.
13089 + Fix SPARC backend bug which caused aborts in final.c.
13090 + Fix sparc-hal-solaris2* configuration fragments.
13091 + Fix bug in sparc block profiling.
13092 + Fix obscure code generation bug for the PARISC targets.
13093 + Define __STDC_EXT__ for HPUX configurations.
13094 + Various POWERPC64 code generation bugfixes.
13095 + Fix abort for PPC targets using ELF (ex GNU/Linux).
13096 + Fix collect2 problems for AIX targets.
13097 + Correct handling of .file directive for PPC targets.
13098 + Fix bug in fix_trunc x86 patterns.
13099 + Fix x86 port to correctly pop the FP stack for functions that
13100 return structures in memory.
13101 + Fix minor bug in strlen x86 pattern.
13102 + Use stabs debugging instead of dwarf1 for x86-solaris targets.
13103 + Fix template repository code to handle leading underscore in
13105 + Fix weak/weak alias support for OpenBSD.
13106 + GNU/Linux for the ARM has C++ compatible include files.
13107 * Language & Runtime specific fixes.
13108 + Fix handling of constructor attribute in the C front-end which
13109 caused problems building the Chill runtime library on some
13111 + Fix minor problem merging type qualifiers in the C front-end.
13112 + Fix aliasing bug for pointers and references (C/C++).
13113 + Fix incorrect "non-constant initializer bug" when -traditional
13114 or -fwritable-strings is enabled.
13115 + Fix build error for Chill front-end on SunOS.
13116 + Do not complain about duplicate instantiations when using
13118 + Fix array bounds handling in C++ front-end which caused
13119 problems with dwarf debugging information in some
13121 + Fix minor namespace problem.
13122 + Fix problem linking java programs.
13124 Additional Changes in GCC 2.95.3
13126 * Generic bugfixes and improvements
13127 + Fix numerous problems that caused incorrect optimization in
13128 the register reloading code.
13129 + Fix numerous problems that caused incorrect optimization in
13130 the loop optimizer.
13131 + Fix aborts in the functions build_insn_chain and scan_loops
13132 under some circumstances.
13133 + Fix an alias analysis bug.
13134 + Fix an infinite compilation bug in the combiner.
13135 + A few problems with complex number support have been fixed.
13136 + It is no longer possible for gcc to act as a fork bomb when
13137 installed incorrectly.
13138 + The -fpack-struct option should be recognized now.
13139 + Fixed a bug that caused incorrect code to be generated due to
13140 a lost stack adjustment.
13141 * Platform specific bugfixes and improvements
13142 + Support building ARM toolchains hosted on Windows.
13143 + Fix attribute calculations in ARM toolchains.
13144 + arm-linux support has been improved.
13145 + Fix a PIC failure on sparc targets.
13146 + On ix86 targets, the regparm attribute should now work
13148 + Several updates for the h8300 port.
13149 + Fix problem building libio with glibc 2.2.
13152 For questions related to the use of GCC, please consult these web
13153 pages and the [17]GCC manuals. If that fails, the
13154 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13155 web pages and the development of GCC are welcome on our developer
13156 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
13159 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
13160 distribution of this entire article is permitted in any medium,
13161 provided this notice is preserved.
13163 These pages are [22]maintained by the GCC team. Last modified
13168 1. http://gcc.gnu.org/news/spill.html
13169 2. http://gcc.gnu.org/news/lcm.html
13170 3. http://gcc.gnu.org/news/cprop.html
13171 4. http://gcc.gnu.org/news/cfg.html
13172 5. http://gcc.gnu.org/news/dse.html
13173 6. http://gcc.gnu.org/news/hoist.html
13174 7. http://gcc.gnu.org/news/alias.html
13175 8. http://gcc.gnu.org/gcc-2.95/c++features.html
13176 9. http://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
13177 10. http://gcc.gnu.org/java/gcj-announce.txt
13178 11. http://gcc.gnu.org/news/javaannounce.html
13179 12. http://gcc.gnu.org/c99status.html
13180 13. http://gcc.gnu.org/news/chill.html
13181 14. http://gcc.gnu.org/news/sparc.html
13182 15. http://gcc.gnu.org/news/egcs-vcg.html
13183 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13184 17. http://gcc.gnu.org/onlinedocs/
13185 18. mailto:gcc-help@gcc.gnu.org
13186 19. mailto:gcc@gcc.gnu.org
13187 20. http://gcc.gnu.org/lists.html
13188 21. http://www.fsf.org/
13189 22. http://gcc.gnu.org/about.html
13190 23. http://validator.w3.org/check/referer
13191 ======================================================================
13192 http://gcc.gnu.org/gcc-2.95/caveats.html
13195 * GCC 2.95 will issue an error for invalid asm statements that had
13196 been silently accepted by earlier versions of the compiler. This is
13197 particularly noticeable when compiling older versions of the Linux
13198 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95)
13199 for more information on this issue.
13200 * GCC 2.95 implements type based alias analysis to disambiguate
13201 memory references. Some programs, particularly the Linux kernel
13202 violate ANSI/ISO aliasing rules and therefore may not operate
13203 correctly when compiled with GCC 2.95. Please refer to the FAQ (as
13204 shipped with GCC 2.95) for more information on this issue.
13205 * GCC 2.95 has a known bug in its handling of complex variables for
13206 64bit targets. Instead of silently generating incorrect code, GCC
13207 2.95 will issue a fatal error for situations it can not handle.
13208 This primarily affects the Fortran community as Fortran makes more
13209 use of complex variables than C or C++.
13210 * GCC 2.95 has an integrated libstdc++, but does not have an
13211 integrated libg++. Furthermore old libg++ releases will not work
13212 with GCC 2.95. You can retrieve a recent copy of libg++ from the
13214 Note most C++ programs only need libstdc++.
13215 * Exception handling may not work with shared libraries, particularly
13216 on alphas, hppas, rs6000/powerpc and mips based platforms.
13217 Exception handling is known to work on x86 GNU/Linux platforms with
13219 * In general, GCC 2.95 is more rigorous about rejecting invalid C++
13220 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0,
13221 or EGCS 1.1. As a result it may be necessary to fix C++ code before
13222 it will compile with GCC 2.95.
13223 * G++ is also converting toward the ISO C++ standard; as a result
13224 code which was previously valid (and thus accepted by other
13225 compilers and older versions of g++) may no longer be accepted. The
13226 flag -fpermissive may allow some non-conforming code to compile
13228 * GCC 2.95 compiled C++ code is not binary compatible with EGCS
13229 1.1.x, EGCS 1.0.x or GCC 2.8.x.
13230 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made
13231 between Sept 30, 1998 and April 30, 1999 (the official end of the
13232 GCC 2.8 project). Future GCC releases will include all the changes
13233 from the defunct GCC 2.8 sources.
13236 For questions related to the use of GCC, please consult these web
13237 pages and the [2]GCC manuals. If that fails, the
13238 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13239 web pages and the development of GCC are welcome on our developer
13240 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
13242 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
13243 distribution of this entire article is permitted in any medium,
13244 provided this notice is preserved.
13246 These pages are [7]maintained by the GCC team. Last modified
13251 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz
13252 2. http://gcc.gnu.org/onlinedocs/
13253 3. mailto:gcc-help@gcc.gnu.org
13254 4. mailto:gcc@gcc.gnu.org
13255 5. http://gcc.gnu.org/lists.html
13256 6. http://www.fsf.org/
13257 7. http://gcc.gnu.org/about.html
13258 8. http://validator.w3.org/check/referer
13259 ======================================================================
13260 http://gcc.gnu.org/egcs-1.1/index.html
13263 September 3, 1998: We are pleased to announce the release of EGCS 1.1.
13264 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1.
13265 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2.
13267 EGCS is a free software project to further the development of the GNU
13268 compilers using an open development environment.
13270 EGCS 1.1 is a major new release of the EGCS compiler system. It has
13271 been [1]extensively tested and is believed to be stable and suitable
13272 for widespread use.
13274 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8
13275 development sources; it contains all of the new features found in GCC
13276 2.8.1 as well as all new development from GCC up to June 6, 1998.
13278 EGCS 1.1 also contains many improvements and features not found in GCC
13279 or in older versions of EGCS:
13280 * Global common subexpression elimination and global constant/copy
13281 propagation (aka [2]gcse)
13282 * Ongoing improvements to the [3]alias analysis support to allow for
13283 better optimizations throughout the compiler.
13284 * Vastly improved [4]C++ compiler and integrated C++ runtime
13286 * Fixes for the /tmp symlink race security problems.
13287 * New targets including mips16, arm-thumb and 64 bit PowerPC.
13288 * Improvements to GNU Fortran (g77) compiler and runtime library made
13289 since g77 version 0.5.23.
13291 See the [5]new features page for a more complete list of new features
13292 found in EGCS 1.1 releases.
13294 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS
13296 * General improvements and fixes
13297 + Avoid some stack overflows when compiling large functions.
13298 + Avoid incorrect loop invariant code motions.
13299 + Fix some core dumps on Linux kernel code.
13300 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2.
13301 + Fix code generation problem in gcse.
13302 + Various documentation related fixes.
13303 * g++/libstdc++ improvements and fixes
13304 + MT safe EH fix for setjmp/longjmp based exception handling.
13305 + Fix a few bad interactions between optimization and exception
13307 + Fixes for demangling of template names starting with "__".
13308 + Fix a bug that would fail to run destructors in some cases
13310 + Fix 'new' of classes with virtual bases.
13311 + Fix crash building Qt on the Alpha.
13312 + Fix failure compiling WIFEXITED macro on GNU/Linux.
13313 + Fix some -frepo failures.
13314 * g77 and libf2c improvements and fixes
13315 + Various documentation fixes.
13316 + Avoid compiler crash on RAND intrinsic.
13317 + Fix minor bugs in makefiles exposed by BSD make programs.
13318 + Define _XOPEN_SOURCE for libI77 build to avoid potential
13319 problems on some 64-bit systems.
13320 + Fix problem with implicit endfile on rewind.
13321 + Fix spurious recursive I/O errors.
13322 * platform specific improvements and fixes
13323 + Match all versions of UnixWare7.
13324 + Do not assume x86 SVR4 or UnixWare targets can handle stabs.
13325 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion
13326 from unsigned ints to double precision floats.
13327 + Fix ARM ABI issue with NetBSD.
13328 + Fix a few arm code generation bugs.
13329 + Fixincludes will fix additional broken SCO OpenServer header
13331 + Fix a m68k backend bug which caused invalid offsets in reg+d
13333 + Fix problems with 64bit AIX 4.3 support.
13334 + Fix handling of long longs for varargs/stdarg functions on the
13336 + Minor fixes to CPP predefines for Windows.
13337 + Fix code generation problems with gpr<->fpr copies for 64bit
13339 + Fix a few coldfire code generation bugs.
13340 + Fix some more header file problems on SunOS 4.x.
13341 + Fix assert.h handling for RTEMS.
13342 + Fix Windows handling of TREE_SYMBOL_REFERENCED.
13343 + Fix x86 compiler abort in reg-stack pass.
13344 + Fix cygwin/windows problem with section attributes.
13345 + Fix Alpha code generation problem exposed by SMP Linux
13347 + Fix typo in m68k 32->64bit integer conversion.
13348 + Make sure target libraries build with -fPIC for PPC & Alpha
13351 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS
13353 * General improvements and fixes
13354 + Fix bug in loop optimizer which caused the SPARC (and
13355 potentially other) ports to segfault.
13356 + Fix infinite recursion in alias analysis and combiner code.
13357 + Fix bug in regclass preferencing.
13358 + Fix incorrect loop reversal which caused incorrect code to be
13359 generated for several targets.
13360 + Fix return value for builtin memcpy.
13361 + Reduce compile time for certain loops which exposed quadratic
13362 behavior in the loop optimizer.
13363 + Fix bug which caused volatile memory to be written multiple
13364 times when only one write was needed/desired.
13365 + Fix compiler abort in caller-save.c
13366 + Fix combiner bug which caused incorrect code generation for
13367 certain division by constant operations.
13368 + Fix incorrect code generation due to a bug in range check
13370 + Fix incorrect code generation due to mis-handling of clobbered
13372 + Fix compiler abort/segfault due to incorrect register
13373 splitting when unrolling loops.
13374 + Fix code generation involving autoincremented addresses with
13376 + Work around bug in the scheduler which caused qt to be
13377 mis-compiled on some platforms.
13378 + Fix code generation problems with -fshort-enums.
13379 + Tighten security for temporary files.
13380 + Improve compile time for codes which make heavy use of
13381 overloaded functions.
13382 + Fix multiply defined constructor/destructor symbol problems.
13383 + Avoid setting bogus RPATH environment variable during
13385 + Avoid GNU-make dependencies in the texinfo subdir.
13386 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp.
13387 --enable-cpp=<dirname> can be used to specify an additional
13388 install directory for the cpp wrapper script.
13389 + Fix CSE bug which caused incorrect label-label refs to appear
13391 + Avoid linking in EH routines from libgcc if they are not
13393 + Avoid obscure bug in aliasing code.
13394 + Fix bug in weak symbol handling.
13395 * Platform-specific improvements and fixes
13396 + Fix detection of PPro/PII on Unixware 7.
13397 + Fix compiler segfault when building spec99 and other programs
13399 + Fix code-generation bugs for integer and floating point
13400 conditional move instructions on the PPro/PII.
13401 + Use fixincludes to fix byteorder problems on i?86-*-sysv.
13402 + Fix build failure for the arc port.
13403 + Fix floating point format configuration for i?86-gnu port.
13404 + Fix problems with hppa1.0-hp-hpux10.20 configuration when
13405 threads are enabled.
13406 + Fix coldfire code generation bugs.
13407 + Fix "unrecognized insn" problems for Alpha and PPC ports.
13408 + Fix h8/300 code generation problem with floating point values
13410 + Fix unrecognized insn problems for the m68k port.
13411 + Fix namespace-pollution problem for the x86 port.
13412 + Fix problems with old assembler on x86 NeXT systems.
13413 + Fix PIC code-generation problems for the SPARC port.
13414 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support.
13415 + Fix minor ISO namespace violation in Alpha varargs/stdarg
13417 + Fix incorrect "braf" instruction usage for the SH port.
13418 + Fix minor bug in va-sh which prevented its use with -ansi.
13419 + Fix problems recognizing and supporting FreeBSD.
13420 + Handle OpenBSD systems correctly.
13421 + Minor fixincludes fix for Digital UNIX 4.0B.
13422 + Fix problems with ctors/dtors in SCO shared libraries.
13423 + Abort instead of generating incorrect code for PPro/PII
13424 floating point conditional moves.
13425 + Avoid multiply defined symbols on GNU/Linux systems using
13427 + Fix abort in alpha compiler.
13428 * Fortran-specific fixes
13429 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year
13430 is in the documented, non-Y2K-compliant range of 0-99, instead
13431 of being returned as 100 in the year 2000.
13432 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the
13433 milliseconds value properly in Values(8).
13434 + Fix the `LStat' intrinsic (in libg2c) to return device-ID
13435 information properly in SArray(7).
13437 Each release includes installation instructions in both HTML and
13438 plaintext forms (see the INSTALL directory in the toplevel directory of
13439 the distribution). However, we also keep the most up to date
13440 [6]installation instructions and [7]build/test status on our web page.
13441 We will update those pages as new information becomes available.
13443 The EGCS project would like to thank the numerous people that have
13444 contributed new features, test results, bugfixes, etc. This [8]amazing
13445 group of volunteers is what makes EGCS successful.
13447 And finally, we can't in good conscience fail to mention some
13448 [9]caveats to using EGCS 1.1.
13450 Download EGCS from egcs.cygnus.com (USA California).
13452 The EGCS 1.1 release is also available on many mirror sites.
13453 [10]Goto mirror list to find a closer site.
13456 For questions related to the use of GCC, please consult these web
13457 pages and the [11]GCC manuals. If that fails, the
13458 [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13459 web pages and the development of GCC are welcome on our developer
13460 list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
13463 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
13464 distribution of this entire article is permitted in any medium,
13465 provided this notice is preserved.
13467 These pages are [16]maintained by the GCC team. Last modified
13472 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html
13473 2. http://gcc.gnu.org/news/gcse.html
13474 3. http://gcc.gnu.org/news/alias.html
13475 4. http://gcc.gnu.org/egcs-1.1/c++features.html
13476 5. http://gcc.gnu.org/egcs-1.1/features.html
13477 6. http://gcc.gnu.org/install/
13478 7. http://gcc.gnu.org/egcs-1.1/buildstat.html
13479 8. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13480 9. http://gcc.gnu.org/egcs-1.1/caveats.html
13481 10. http://gcc.gnu.org/mirrors.html
13482 11. http://gcc.gnu.org/onlinedocs/
13483 12. mailto:gcc-help@gcc.gnu.org
13484 13. mailto:gcc@gcc.gnu.org
13485 14. http://gcc.gnu.org/lists.html
13486 15. http://www.fsf.org/
13487 16. http://gcc.gnu.org/about.html
13488 17. http://validator.w3.org/check/referer
13489 ======================================================================
13490 http://gcc.gnu.org/egcs-1.1/features.html
13491 EGCS 1.1 new features
13493 * Integrated GNU Fortran (g77) compiler and runtime library with
13494 improvements, based on g77 version 0.5.23.
13495 * Vast improvements in the C++ compiler; so many they have [1]page of
13497 * Compiler implements [2]global common subexpression elimination and
13498 global copy/constant propagation.
13499 * More major improvements in the [3]alias analysis code.
13500 * More major improvements in the exception handling code to improve
13501 performance, lower static overhead and provide the infrastructure
13502 for future improvements.
13503 * The infamous /tmp symlink race security problems have been fixed.
13504 * The regmove optimization pass has been nearly completely rewritten
13505 to improve performance of generated code.
13506 * The compiler now recomputes register usage information before local
13507 register allocation. By providing more accurate information to the
13508 priority based allocator, we get better register allocation.
13509 * The register reloading phase of the compiler optimizes spill code
13510 much better than in previous releases.
13511 * Some bad interactions between the register allocator and
13512 instruction scheduler have been fixed, resulting in much better
13513 code for certain programs. Additionally, we have tuned the
13514 scheduler in various ways to improve performance of generated code
13515 for some architectures.
13516 * The compiler's branch shortening algorithms have been significantly
13517 improved to work better on targets which align jump targets.
13518 * The compiler now supports -Os to prefer optimizing for code space
13519 over optimizing for code speed.
13520 * The compiler will now totally eliminate library calls which compute
13521 constant values. This primarily helps targets with no integer
13522 div/mul support and targets without floating point support.
13523 * The compiler now supports an extensive "--help" option.
13524 * cpplib has been greatly improved and may be suitable for limited
13526 * Memory footprint for the compiler has been significantly reduced
13527 for some pathological cases.
13528 * The time to build EGCS has been improved for certain targets
13529 (particularly the alpha and mips platforms).
13530 * Many infrastructure improvements throughout the compiler, plus the
13531 usual mountain of bugfixes and minor improvements.
13532 * Target dependent improvements:
13533 + SPARC port now includes V8 plus and V9 support as well as
13534 performance tuning for Ultra class machines. The SPARC port
13535 now uses the Haifa scheduler.
13536 + Alpha port has been tuned for the EV6 processor and has an
13537 optimized expansion of memcpy/bzero. The Alpha port now uses
13538 the Haifa scheduler.
13539 + RS6000/PowerPC: support for the Power64 architecture and AIX
13540 4.3. The RS6000/PowerPC port now uses the Haifa scheduler.
13541 + x86: Alignment of static store data and jump targets is per
13542 Intel recommendations now. Various improvements throughout the
13543 x86 port to improve performance on Pentium processors
13544 (including improved epilogue sequences for Pentium chips and
13545 backend improvements which should help register allocation on
13546 all x86 variants. Conditional move support has been fixed and
13547 enabled for PPro processors. The x86 port also better supports
13548 64bit operations now. Unixware 7, a System V Release 5 target,
13549 is now supported and SCO OpenServer targets can support GAS.
13550 + MIPS has improved multiply/multiply-add support and now
13551 includes mips16 ISA support.
13552 + M68k has many micro-optimizations and Coldfire fixes.
13553 * Core compiler is based on the GCC development tree from June 9,
13554 1998, so we have all of the [4]features found in GCC 2.8.
13557 For questions related to the use of GCC, please consult these web
13558 pages and the [5]GCC manuals. If that fails, the
13559 [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13560 web pages and the development of GCC are welcome on our developer
13561 list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
13563 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
13564 distribution of this entire article is permitted in any medium,
13565 provided this notice is preserved.
13567 These pages are [10]maintained by the GCC team. Last modified
13572 1. http://gcc.gnu.org/egcs-1.1/c++features.html
13573 2. http://gcc.gnu.org/news/gcse.html
13574 3. http://gcc.gnu.org/news/alias.html
13575 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13576 5. http://gcc.gnu.org/onlinedocs/
13577 6. mailto:gcc-help@gcc.gnu.org
13578 7. mailto:gcc@gcc.gnu.org
13579 8. http://gcc.gnu.org/lists.html
13580 9. http://www.fsf.org/
13581 10. http://gcc.gnu.org/about.html
13582 11. http://validator.w3.org/check/referer
13583 ======================================================================
13584 http://gcc.gnu.org/egcs-1.1/caveats.html
13587 * EGCS has an integrated libstdc++, but does not have an integrated
13588 libg++. Furthermore old libg++ releases will not work with EGCS; HJ
13589 Lu has made a libg++-2.8.1.2 snapshot available which may work with
13591 Note most C++ programs only need libstdc++.
13592 * Exception handling may not work with shared libraries, particularly
13593 on alphas, hppas, rs6000/powerpc and mips based platforms.
13594 Exception handling is known to work on x86-linux platforms with
13596 * Some versions of the Linux kernel have bugs which prevent them from
13597 being compiled or from running when compiled by EGCS. See the FAQ
13598 (as shipped with EGCS 1.1) for additional information.
13599 * In general, EGCS is more rigorous about rejecting invalid C++ code
13600 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As
13601 a result it may be necessary to fix C++ code before it will compile
13603 * G++ is also converting toward the ISO C++ standard; as a result
13604 code which was previously valid (and thus accepted by other
13605 compilers and older versions of g++) may no longer be accepted.
13606 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x
13607 or GCC 2.8.x due to changes necessary to support thread safe
13608 exception handling.
13611 For questions related to the use of GCC, please consult these web
13612 pages and the [1]GCC manuals. If that fails, the
13613 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13614 web pages and the development of GCC are welcome on our developer
13615 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
13617 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
13618 distribution of this entire article is permitted in any medium,
13619 provided this notice is preserved.
13621 These pages are [6]maintained by the GCC team. Last modified
13626 1. http://gcc.gnu.org/onlinedocs/
13627 2. mailto:gcc-help@gcc.gnu.org
13628 3. mailto:gcc@gcc.gnu.org
13629 4. http://gcc.gnu.org/lists.html
13630 5. http://www.fsf.org/
13631 6. http://gcc.gnu.org/about.html
13632 7. http://validator.w3.org/check/referer
13633 ======================================================================
13634 http://gcc.gnu.org/egcs-1.0/index.html
13637 December 3, 1997: We are pleased to announce the release of EGCS 1.0.
13638 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1.
13639 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2.
13640 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3.
13642 EGCS is a collaborative effort involving several groups of hackers
13643 using an open development model to accelerate development and testing
13644 of GNU compilers and runtime libraries.
13646 An important goal of EGCS is to allow wide scale testing of
13647 experimental features and optimizations; therefore, EGCS contains some
13648 features and optimizations which are still under development. However,
13649 EGCS has been carefully tested and should be comparable in quality to
13652 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8
13653 development sources; it contains nearly all of the new features found
13656 EGCS 1.0 also contains many improvements and features not found in GCC
13657 2.7 and even the GCC 2.8 series (which was released after the original
13659 * Integrated C++ runtime libraries, including support for most major
13661 * The integrated libstdc++ library includes a verbatim copy of SGI's
13663 * Integrated GNU Fortran compiler.
13664 * New instruction scheduler.
13665 * New alias analysis code.
13667 See the [1]new features page for a more complete list of new features.
13669 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few
13670 critical bugs and add support for Red Hat 5.0 Linux. Changes since the
13672 * Add support for Red Hat 5.0 Linux and better support for Linux
13673 systems using glibc2.
13674 Many programs failed to link when compiled with EGCS 1.0 on Red Hat
13675 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should
13676 fix these problems.
13677 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception
13678 handling interfaces.
13679 To avoid future compatibility problems, we strongly urge anyone who
13680 is planning on distributing shared libraries that contain C++ code
13681 to upgrade to EGCS 1.0.1 first.
13682 Soon after EGCS 1.0 was released, the GCC developers made some
13683 incompatible changes in libgcc's exception handling interfaces.
13684 These changes were needed to solve problems on some platforms. This
13685 means that GCC 2.8.0, when released, will not be seamlessly
13686 compatible with shared libraries built by EGCS 1.0. The reason is
13687 that the libgcc.a in GCC 2.8.0 will not contain a function needed
13688 by the old interface.
13689 The result of this is that there may be compatibility problems with
13690 shared libraries built by EGCS 1.0 when used with GCC 2.8.0.
13691 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface,
13692 and libgcc.a has the support routines for both the old and the new
13693 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed,
13694 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed).
13695 The maintainers of GCC 2.x have decided against including seamless
13696 support for the old interface in 2.8.0, since it was never
13697 "official", so to avoid future compatibility problems we recommend
13698 against distributing any shared libraries built by EGCS 1.0 that
13699 contain C++ code (upgrade to 1.0.1 and use that).
13700 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc backends.
13701 The x86 changes fix code generation errors exposed when building
13702 glibc2 and the usual GNU/Linux dynamic linker (ld.so).
13703 The hppa change fixes a compiler abort when configured for use with
13705 The MIPS changes fix problems with the definition of LONG_MAX on
13706 newer systems, allow for command line selection of the target ABI,
13707 and fix one code generation problem.
13708 The rs6000/ppc change fixes some problems with passing structures
13709 to varargs/stdarg functions.
13710 * A few machine independent bugfixes, mostly to fix code generation
13711 errors when building Linux kernels or glibc.
13712 * Fix a few critical exception handling and template bugs in the C++
13714 * Fix Fortran namelist bug on alphas.
13715 * Fix build problems on x86-solaris systems.
13717 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several
13718 serious problems in EGCS 1.0.1.
13719 * General improvements and fixes
13720 + Memory consumption significantly reduced, especially for
13721 templates and inline functions.
13722 + Fix various problems with glibc2.1.
13723 + Fix loop optimization bug exposed by rs6000/ppc port.
13724 + Fix to avoid potential code generation problems in jump.c.
13725 + Fix some undefined symbol problems in dwarf1 debug support.
13726 * g++/libstdc++ improvements and fixes
13727 + libstdc++ in the EGCS release has been updated and should be
13728 link compatible with libstdc++-2.8.
13729 + Various fixes in libio/libstdc++ to work better on GNU/Linux
13731 + Fix problems with duplicate symbols on systems that do not
13732 support weak symbols.
13733 + Memory corruption bug and undefined symbols in bastring have
13735 + Various exception handling fixes.
13736 + Fix compiler abort for very long thunk names.
13737 * g77 improvements and fixes
13738 + Fix compiler crash for omitted bound in Fortran CASE
13740 + Add missing entries to g77 lang-options.
13741 + Fix problem with -fpedantic in the g77 compiler.
13742 + Fix "backspace" problem with g77 on alphas.
13743 + Fix x86 backend problem with Fortran literals and -fpic.
13744 + Fix some of the problems with negative subscripts for g77 on
13746 + Fixes for Fortran builds on cygwin32/mingw32.
13747 * platform specific improvements and fixes
13748 + Fix long double problems on x86 (exposed by glibc).
13749 + x86 ports define i386 again to keep imake happy.
13750 + Fix exception handling support on NetBSD ports.
13751 + Several changes to collect2 to fix many problems with AIX.
13752 + Define __ELF__ for GNU/Linux on rs6000.
13753 + Fix -mcall-linux problem on GNU/Linux on rs6000.
13754 + Fix stdarg/vararg problem for GNU/Linux on rs6000.
13755 + Allow autoconf to select a proper install problem on AIX 3.1.
13756 + m68k port support includes -mcpu32 option as well as cpu32
13758 + Fix stdarg bug for irix6.
13759 + Allow EGCS to build on irix5 without the gnu assembler.
13760 + Fix problem with static linking on sco5.
13761 + Fix bootstrap on sco5 with native compiler.
13762 + Fix for abort building newlib on H8 target.
13763 + Fix fixincludes handling of math.h on SunOS.
13764 + Minor fix for Motorola 3300 m68k systems.
13766 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few
13767 problems reported by Red Hat for builds of Red Hat 5.1.
13768 * Generic bugfixes:
13769 + Fix a typo in the libio library which resulted in incorrect
13770 behavior of istream::get.
13771 + Fix the Fortran negative array index problem.
13772 + Fix a major problem with the ObjC runtime thread support
13774 + Reduce memory consumption of the Haifa scheduler.
13775 * Target specific bugfixes:
13776 + Fix one x86 floating point code generation bug exposed by
13778 + Fix one x86 internal compiler error exposed by glibc2 builds.
13779 + Fix profiling bugs on the Alpha.
13780 + Fix ImageMagick & emacs 20.2 build problems on the Alpha.
13781 + Fix rs6000/ppc bug when converting values from integer types
13782 to floating point types.
13784 The EGCS 1.0 releases include installation instructions in both HTML
13785 and plaintext forms (see the INSTALL directory in the toplevel
13786 directory of the distribution). However, we also keep the most up to
13787 date [2]installation instructions and [3]build/test status on our web
13788 page. We will update those pages as new information becomes available.
13790 And, we can't in good conscience fail to mention some [4]caveats to
13793 Update: Big thanks to Stanford for providing a high speed link for
13794 downloading EGCS (go.cygnus.com)!
13796 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com
13797 (USA California -- High speed link provided by Stanford).
13799 The EGCS 1.0 release is also available many mirror sites.
13800 [5]Goto mirror list to find a closer site
13802 We'd like to thank the numerous people that have contributed new
13803 features, test results, bugfixes, etc. Unfortunately, they're far too
13804 numerous to mention by name.
13807 For questions related to the use of GCC, please consult these web
13808 pages and the [6]GCC manuals. If that fails, the
13809 [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13810 web pages and the development of GCC are welcome on our developer
13811 list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives.
13813 Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and
13814 distribution of this entire article is permitted in any medium,
13815 provided this notice is preserved.
13817 These pages are [11]maintained by the GCC team. Last modified
13822 1. http://gcc.gnu.org/egcs-1.0/features.html
13823 2. http://gcc.gnu.org/install/
13824 3. http://gcc.gnu.org/egcs-1.0/buildstat.html
13825 4. http://gcc.gnu.org/egcs-1.0/caveats.html
13826 5. http://gcc.gnu.org/mirrors.html
13827 6. http://gcc.gnu.org/onlinedocs/
13828 7. mailto:gcc-help@gcc.gnu.org
13829 8. mailto:gcc@gcc.gnu.org
13830 9. http://gcc.gnu.org/lists.html
13831 10. http://www.fsf.org/
13832 11. http://gcc.gnu.org/about.html
13833 12. http://validator.w3.org/check/referer
13834 ======================================================================
13835 http://gcc.gnu.org/egcs-1.0/features.html
13838 * Core compiler is based on the gcc2 development tree from Aug 2,
13839 1997, so we have most of the [1]features found in GCC 2.8.
13840 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929.
13841 * Vast improvements in the C++ compiler; so many they have [2]page of
13843 * Integrated C++ runtime libraries, including support for most major
13845 * New instruction scheduler from IBM Haifa which includes support for
13846 function wide instruction scheduling as well as superscalar
13848 * Significantly improved alias analysis code.
13849 * Improved register allocation for two address machines.
13850 * Significant code generation improvements for Fortran code on
13852 * Various optimizations from the g77 project as well as improved loop
13854 * Dwarf2 debug format support for some targets.
13855 * egcs libstdc++ includes the SGI STL implementation without changes.
13856 * As a result of these and other changes, egcs libstc++ is not binary
13857 compatible with previous releases of libstdc++.
13858 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO
13859 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and
13860 1.1), Support for RTEMS on several embedded targets, Support for
13861 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and
13862 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc.
13863 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio.
13864 * RS6000/PowerPC ports generate code which can run on all
13865 RS6000/PowerPC variants by default.
13866 * -mcpu= and -march= switches for the x86 port to allow better
13867 control over how the x86 port generates code.
13868 * Includes the template repository patch (aka repo patch); note the
13869 new template code makes repo obsolete for ELF systems using gnu-ld
13871 * Plus the usual assortment of bugfixes and improvements.
13874 For questions related to the use of GCC, please consult these web
13875 pages and the [3]GCC manuals. If that fails, the
13876 [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13877 web pages and the development of GCC are welcome on our developer
13878 list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives.
13880 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and
13881 distribution of this entire article is permitted in any medium,
13882 provided this notice is preserved.
13884 These pages are [8]maintained by the GCC team. Last modified
13889 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13890 2. http://gcc.gnu.org/egcs-1.0/c++features.html
13891 3. http://gcc.gnu.org/onlinedocs/
13892 4. mailto:gcc-help@gcc.gnu.org
13893 5. mailto:gcc@gcc.gnu.org
13894 6. http://gcc.gnu.org/lists.html
13895 7. http://www.fsf.org/
13896 8. http://gcc.gnu.org/about.html
13897 9. http://validator.w3.org/check/referer
13898 ======================================================================
13899 http://gcc.gnu.org/egcs-1.0/caveats.html
13902 * EGCS has an integrated libstdc++, but does not have an integrated
13903 libg++. Furthermore old libg++ releases will not work with egc; HJ
13904 Lu has made a libg++-2.8.1.2 available which may work with EGCS.
13905 Note most C++ programs only need libstdc++.
13906 * Note that using -pedantic or -Wreturn-type can cause an explosion
13907 in the amount of memory needed for template-heavy C++ code, such as
13908 code that uses STL. Also note that -Wall includes -Wreturn-type, so
13909 if you use -Wall you will need to specify -Wno-return-type to turn
13911 * Exception handling may not work with shared libraries, particularly
13912 on alphas, hppas, and mips based platforms. Exception handling is
13913 known to work on x86-linux platforms with shared libraries.
13914 * Some versions of the Linux kernel have bugs which prevent them from
13915 being compiled or from running when compiled by EGCS. See the FAQ
13916 (as shipped with EGCS 1.0) for additional information.
13917 * In general, EGCS is more rigorous about rejecting invalid C++ code
13918 or deprecated C++ constructs than G++ 2.7. As a result it may be
13919 necessary to fix C++ code before it will compile with EGCS.
13920 * G++ is also aggressively tracking the C++ standard; as a result
13921 code which was previously valid (and thus accepted by other
13922 compilers and older versions of G++) may no longer be accepted.
13923 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS
13924 1.0.x and later releases should work with Red Hat Linux 5.0.
13927 For questions related to the use of GCC, please consult these web
13928 pages and the [1]GCC manuals. If that fails, the
13929 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13930 web pages and the development of GCC are welcome on our developer
13931 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
13933 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
13934 distribution of this entire article is permitted in any medium,
13935 provided this notice is preserved.
13937 These pages are [6]maintained by the GCC team. Last modified
13942 1. http://gcc.gnu.org/onlinedocs/
13943 2. mailto:gcc-help@gcc.gnu.org
13944 3. mailto:gcc@gcc.gnu.org
13945 4. http://gcc.gnu.org/lists.html
13946 5. http://www.fsf.org/
13947 6. http://gcc.gnu.org/about.html
13948 7. http://validator.w3.org/check/referer
13949 ======================================================================