1 Noteworthy changes in GCC for EGCS 1.1.
2 ---------------------------------------
4 The compiler now implements global common subexpression elimination (gcse) as
5 well as global constant/copy propagation. (link to gcse page).
7 More major improvements have been made to the alias analysis code. A new
8 option to allow front-ends to provide alias information to the optimizers
9 has also been added (-fstrict-aliasing). -fstrict-aliasing is off by default
10 now, but will be enabled by default in the future. (link to alias page)
12 Major changes continue in the exception handling support. This release
13 includes some changes to reduce static overhead for exception handling. It
14 also includes some major changes to the setjmp/longjmp based EH mechanism to
15 make it less pessimistic. And finally, major infrastructure improvements
16 to the dwarf2 EH mechanism have been made to make our EH support extensible.
18 We have fixed the infamous security problems with temporary files.
20 The "regmove" optimization pass has been nearly completely rewritten. It now
21 uses much more information about the target to determine profitability of
24 The compiler now recomputes register usage information immediately before
25 register allocation. Previously such information was only not kept up to
26 date after instruction combination which led to poor register allocation
27 choices by our priority based register allocator.
29 The register reloading phase of the compiler has been improved to better
30 optimize spill code. This primarily helps targets which generate lots of
31 spills (like the x86 ports and many register poor embedded ports).
33 A few changes in the heuristics used by the register allocator and scheduler
34 have been made which can significantly improve performance for certain
37 The compiler's branch shortening algorithms have been significantly improved
38 to work better on targets which align jump targets.
40 The compiler now supports the "ADDRESSOF" optimization which can significantly
41 reduce the overhead for certain inline calls (and inline calls in general).
43 The compiler now supports a code size optimization switch (-Os). When enabled
44 the compiler will prefer optimizations which improve code size over those
45 which improve code speed.
47 The compiler has been improved to completely eliminate library calls which
48 compute constant values. This is particularly useful on machines which
49 do not have integer mul/div or floating point support on-chip.
51 GCC now supports a "--help" option to print detailed help information.
53 cpplib has been greatly improved. It is probably useable for some sites now
54 (major missing feature is trigraphs).
56 Memory footprint for the compiler has been significantly reduced for certain
59 Build time improvements for targets which support lots of sched parameters
60 (alpha and mips primarily).
62 Compile time for certain programs using large constant initializers has been
63 improved (effects glibc significantly).
65 Plus an incredible number of infrastructure changes, warning fixes, bugfixes
66 and local optimizations.
68 Various improvements have been made to better support cross compilations. They
69 are still not easy, but they are improving.
73 Sparc: Now includes V8 plus and V9 support, lots of tuning for Ultrasparcs
74 and uses the Haifa scheduler by default.
76 Alpha: EV6 tuned, optimized expansion of memcpy/bzero.
78 x86: Data in the static store is aligned per Intel recommendations. Jump
79 targets are aligned per Intel recommendations. Improved epilogue
80 sequences for Pentium chips. Backend improvements which should help
81 register allocation on all x86 variants. Support for PPro conditional
82 move instructions has been fixed and enabled. Random changes
83 throughout the port to make generated code more Pentium friendly.
84 Improved support for 64bit integer operations.
86 RS6000/PowerPC: Includes AIX4.3 support as well as PowerPC64 support.
87 Haifa instruction scheduling is enabled by default now.
89 MIPS: Multiply/Multiply-Add support has been largely rewritten to generate
90 more efficient code. Includes mips16 support.
92 M68K: Various micro-optimizations and Coldfire fixes.
94 M32r: Major improvements to this port.
96 Arm: Includes Thumb and super interworking support.
98 EGCS includes all gcc2 changes up to and including the June 9, 1998 snapshot.
101 Noteworthy changes in GCC version 2.8.1
102 ---------------------------------------
104 Numerous bugs have been fixed and some minor performance
105 improvements (compilation speed) have been made.
107 Noteworthy changes in GCC version 2.8.0
108 ---------------------------------------
110 A major change in this release is the addition of a framework for
111 exception handling, currently used by C++. Many internal changes and
112 optimization improvements have been made. These increase the
113 maintainability and portability of GCC. GCC now uses autoconf to
114 compute many host parameters.
116 The following lists changes that add new features or targets.
118 See cp/NEWS for new features of C++ in this release.
120 New tools and features:
122 The Dwarf 2 debugging information format is supported on ELF systems, and
123 is the default for -ggdb where available. It can also be used for C++.
124 The Dwarf version 1 debugging format is also permitted for C++, but
127 gcov.c is provided for test coverage analysis and branch profiling
128 analysis is also supported; see -fprofile-arcs, -ftest-coverage,
129 and -fbranch-probabilities.
131 Support for the Checker memory checking tool.
133 New switch, -fstack-check, to check for stack overflow on systems that
134 don't have such built into their ABI.
136 New switches, -Wundef and -Wno-undef to warn if an undefined identifier
137 is evaluated in an #if directive.
139 Options -Wall and -Wimplicit now cause GCC to warn about implicit int
140 in declarations (e.g. `register i;'), since the C Standard committee
141 has decided to disallow this in the next revision of the standard;
142 -Wimplicit-function-declarations and -Wimplicit-int are subsets of
145 Option -Wsign-compare causes GCC to warn about comparison of signed and
148 Add -dI option of cccp for cxref.
150 New features in configuration, installation and specs file handling:
152 New option --enable-c-cpplib to configure script.
154 You can use --with-cpu on the configure command to specify the default
155 CPU that GCC should generate code for.
157 The -specs=file switch allows you to override default specs used in
158 invoking programs like cc1, as, etc.
160 Allow including one specs file from another and renaming a specs
163 You can now relocate all GCC files with a single environment variable
164 or a registry entry under Windows 95 and Windows NT.
166 Changes in Objective-C:
168 The Objective-C Runtime Library has been made thread-safe.
170 The Objective-C Runtime Library contains an interface for creating
171 mutexes, condition mutexes, and threads; it requires a back-end
172 implementation for the specific platform and/or thread package.
173 Currently supported are DEC/OSF1, IRIX, Mach, OS/2, POSIX, PCThreads,
174 Solaris, and Windows32. The --enable-threads parameter can be used
175 when configuring GCC to enable and select a thread back-end.
177 Objective-C is now configured as separate front-end language to GCC,
178 making it more convenient to conditionally build it.
180 The internal structures of the Objective-C Runtime Library have
181 changed sufficiently to warrant a new version number; now version 8.
182 Programs compiled with an older version must be recompiled.
184 The Objective-C Runtime Library can be built as a DLL on Windows 95
185 and Windows NT systems.
187 The Objective-C Runtime Library implements +load.
189 The following new targets are supported (see also list under each
190 individual CPU below):
192 Embedded target m32r-elf.
193 Embedded Hitachi Super-H using ELF.
194 RTEMS real-time system on various CPU targets.
197 Matsushita MN10200 processor.
198 Matsushita MN10300 processor.
199 Sparc and PowerPC running on VxWorks.
200 Support both glibc versions 1 and 2 on Linux-based GNU systems.
202 New features for DEC Alpha systems:
204 Allow detailed specification of IEEE fp support:
205 -mieee, -mieee-with-inexact, and -mieee-conformant
206 -mfp-trap-mode=xxx, -mfp-round-mode=xxx, -mtrap-precision=xxx
207 -mcpu=xxx for CPU selection
208 Support scheduling parameters for EV5.
209 Add support for BWX, CIX, and MAX instruction set extensions.
210 Support Linux-based GNU systems.
213 Additional supported processors and systems for MIPS targets:
215 MIPS4 instruction set.
216 R4100, R4300 and R5000 processors.
221 New features for Intel x86 family:
223 Add scheduling parameters for Pentium and Pentium Pro.
224 Support stabs on Solaris-x86.
225 Intel x86 processors running the SCO OpenServer 5 family.
226 Intel x86 processors running DG/UX.
227 Intel x86 using Cygwin32 or Mingw32 on Windows 95 and Windows NT.
229 New features for Motorola 68k family:
231 Support for 68060 processor.
232 More consistent switches to specify processor.
233 Motorola 68k family running AUX.
234 68040 running pSOS, ELF object files, DBX debugging.
235 Coldfire variant of Motorola m68k family.
237 New features for the HP PA RISC:
239 -mspace and -mno-space
240 -mlong-load-store and -mno-long-load-store
241 -mbig-switch -mno-big-switch
243 GCC on the PA requires either gas-2.7 or the HP assembler; for best
244 results using GAS is highly recommended. GAS is required for -g and
245 exception handling support.
247 New features for SPARC-based systems:
250 The sparclet cpu, supporting only a.out file format.
251 Sparc running SunOS 4 with the GNU assembler.
252 Sparc running the Linux-based GNU system.
253 Embedded Sparc processors running the ELF object file format.
258 -malign-functions=xxx
259 -mimpure-text and -mno-impure-text
261 Options -mno-v8 and -mno-sparclite are no longer supported on SPARC
262 targets. Options -mcypress, -mv8, -msupersparc, -msparclite, -mf930,
263 and -mf934 are deprecated and will be deleted in GCC 2.9. Use
266 New features for rs6000 and PowerPC systems:
268 Solaris 2.51 running on PowerPC's.
269 The Linux-based GNU system running on PowerPC's.
270 -mcpu=604e,602,603e,620,801,823,mpc505,821,860,power2
272 -mrelocatable-lib, -mno-relocatable-lib
274 -mupdate, -mno-update
275 -mfused-madd, -mno-fused-madd
279 -mcall-linux, -mcall-solaris, -mcall-sysv-eabi, -mcall-sysv-noeabi
280 -msdata, -msdata=none, -msdata=default, -msdata=sysv, -msdata=eabi
283 wchar_t is now of type long as per the ABI, not unsigned short.
285 -mcpu=403 now implies -mstrict-align.
286 Implement System V profiling.
288 Aix 4.1 GCC targets now default to -mcpu=common so that programs
289 compiled can be moved between rs6000 and powerpc based systems. A
290 consequence of this is that -static won't work, and that some programs
291 may be slightly slower.
293 You can select the default value to use for -mcpu=xxx on rs6000 and
294 powerpc targets by using the --with-cpu=xxx option when configuring the
295 compiler. In addition, a new options, -mtune=xxx was added that
296 selects the machine to schedule for but does not select the
299 Directory names used for storing the multilib libraries on System V
300 and embedded PowerPC systems have been shortened to work with commands
301 like tar that have fixed limits on pathname size.
303 New features for the Hitachi H8/300(H):
306 -ms (for the Hitachi H8/S processor)
309 New features for the ARM:
311 -march=xxx, -mtune=xxx, -mcpu=xxx
312 Support interworking with Thumb code.
313 ARM processor with a.out object format, COFF, or AOF assembler.
314 ARM on "semi-hosted" platform.
316 ARM running the Linux-based GNU system.
318 New feature for Solaris systems:
320 GCC installation no longer makes a copy of system include files,
321 thus insulating GCC better from updates to the operating system.
324 Noteworthy changes in GCC version 2.7.2
325 ---------------------------------------
327 A few bugs have been fixed (most notably the generation of an
328 invalid assembler opcode on some RS/6000 systems).
330 Noteworthy changes in GCC version 2.7.1
331 ---------------------------------------
333 This release fixes numerous bugs (mostly minor) in GCC 2.7.0, but
334 also contains a few new features, mostly related to specific targets.
336 Major changes have been made in code to support Windows NT.
338 The following new targets are supported:
342 HP/UX version 10 on HP PA RISC (treated like version 9)
343 DEC Alpha running Windows NT
345 When parsing C, GCC now recognizes C++ style `//' comments unless you
346 specify `-ansi' or `-traditional'.
348 The PowerPC System V targets (powerpc-*-sysv, powerpc-*-eabi) now use the
349 calling sequence specified in the System V Application Binary Interface
350 Processor Supplement (PowerPC Processor ABI Supplement) rather than the calling
351 sequence used in GCC version 2.7.0. That calling sequence was based on the AIX
352 calling sequence without function descriptors. To compile code for that older
353 calling sequence, either configure the compiler for powerpc-*-eabiaix or use
354 the -mcall-aix switch when compiling and linking.
356 Noteworthy changes in GCC version 2.7.0
357 ---------------------------------------
359 GCC now works better on systems that use ".obj" and ".exe" instead of
360 ".o" and no extension. This involved changes to the driver program,
361 gcc.c, to convert ".o" names to ".obj" and to GCC's Makefile to use
362 ".obj" and ".exe" in filenames that are not targets. In order to
363 build GCC on such systems, you may need versions of GNU make and/or
364 compatible shells. At this point, this support is preliminary.
366 Object file extensions of ".obj" and executable file extensions of
367 ".exe" are allowed when using appropriate version of GNU Make.
369 Numerous enhancements were made to the __attribute__ facility including
370 more attributes and more places that support it. We now support the
371 "packed", "nocommon", "noreturn", "volatile", "const", "unused",
372 "transparent_union", "constructor", "destructor", "mode", "section",
373 "align", "format", "weak", and "alias" attributes. Each of these
374 names may also be specified with added underscores, e.g., "__packed__".
375 __attribute__ may now be applied to parameter definitions, function
376 definitions, and structure, enum, and union definitions.
378 GCC now supports returning more structures in registers, as specified by
379 many calling sequences (ABIs), such as on the HP PA RISC.
381 A new option '-fpack-struct' was added to automatically pack all structure
382 members together without holes.
384 There is a new library (cpplib) and program (cppmain) that at some
385 point will replace cpp (aka cccp). To use cppmain as cpp now, pass
386 the option CCCP=cppmain to make. The library is already used by the
387 fix-header program, which should speed up the fixproto script.
389 New options for supported targets:
392 NetBSD on MIPS, m68k, VAX, and x86.
393 LynxOS on x86, m68k, Sparc, and RS/6000.
394 VxWorks on many targets.
396 Windows/NT on x86 architecture. Initial support for Windows/NT on Alpha
399 Many embedded targets, specifically UDI on a29k, aout, coff, elf,
400 and vsta "operating systems" on m68k, m88k, mips, sparc, and x86.
402 Additional support for x86 (i386, i486, and Pentium):
404 Work with old and new linkers for Linux-based GNU systems,
405 supporting both a.out and ELF.
408 -malign-double, -mregparm=, -malign-loops= and -malign-jumps= switches.
409 On ISC systems, support -Xp like -posix.
411 Additions for RS/6000:
413 Instruction scheduling information for PowerPC 403.
415 -mstring and -mno-string.
416 -msoft-float and floating-point emulation included.
417 Preliminary support for PowerPC System V.4 with or without the GNU as.
418 Preliminary support for EABI.
419 Preliminary support for 64-bit systems.
420 Both big and little endian systems.
422 New features for MIPS-based systems:
428 Allow dollar signs in labels on SGI/Irix 5.x.
430 New support for HP PA RISC:
432 Generation of PIC (requires binutils-2.5.2.u6 or later).
433 HP-UX version 9 on HP PA RISC (dynamically links even with -g).
434 Processor variants for HP PA RISC: 700, 7100, and 7100LC.
435 Automatic generation of long calls when needed.
436 -mfast-indirect-calls for kernels and static binaries.
438 The called routine now copies arguments passed by invisible reference,
439 as required by the calling standard.
441 Other new miscellaneous target-specific support:
444 -mold-align for i960.
445 Configuration for "semi-hosted" ARM.
446 -momit-leaf-frame-pointer for M88k.
447 SH3 variant of Hitachi Super-H and support both big and little endian.
449 Changes to Objective-C:
451 Bare-bones implementation of NXConstantString has been added,
452 which is invoked by the @"string" directive.
454 Class * has been changed to Class to conform to the NextSTEP and
457 Enhancements to make dynamic loading easier.
459 The module version number has been updated to Version 7, thus existing
460 code will need to be recompiled to use the current run-time library.
462 GCC now supports the ISO Normative Addendum 1 to the C Standard.
465 The header <iso646.h> defines macros for C programs written
466 in national variants of ISO 646.
468 The following digraph tokens are supported:
470 These behave like the following, respectively:
473 Digraph tokens are supported unless you specify the `-traditional'
474 option; you do not need to specify `-ansi' or `-trigraphs'. Except
475 for contrived and unlikely examples involving preprocessor
476 stringizing, digraph interpretation doesn't change the meaning of
477 programs; this is unlike trigraph interpretation, which changes the
478 meanings of relatively common strings.
480 The macro __STDC_VERSION__ has the value 199409L.
482 As usual, for full conformance to the standard, you also need a
483 C library that conforms.
485 The following lists changes that have been made to g++. If some
486 features mentioned below sound unfamiliar, you will probably want to
487 look at the recently-released public review copy of the C++ Working
488 Paper. For PostScript and PDF (Adobe Acrobat) versions, see the
489 archive at ftp://research.att.com/dist/stdc++/WP. For HTML and ASCII
490 versions, see ftp://ftp.cygnus.com/pub/g++. On the web, see
491 http://www.cygnus.com/~mrs/wp-draft.
493 The scope of variables declared in the for-init-statement has been changed
494 to conform to http://www.cygnus.com/~mrs/wp-draft/stmt.html#stmt.for; as a
495 result, packages such as groff 1.09 will not compile unless you specify the
496 -fno-for-scope flag. PLEASE DO NOT REPORT THIS AS A BUG; this is a change
497 mandated by the C++ standardization committee.
499 Binary incompatibilities:
501 The builtin 'bool' type is now the size of a machine word on RISC targets,
502 for code efficiency; it remains one byte long on CISC targets.
504 Code that does not use #pragma interface/implementation will most
505 likely shrink dramatically, as g++ now only emits the vtable for a
506 class in the translation unit where its first non-inline, non-abstract
507 virtual function is defined.
509 Classes that do not define the copy constructor will sometimes be
510 passed and returned in registers. This may illuminate latent bugs in
513 Support for automatic template instantiation has *NOT* been added, due
514 to a disagreement over design philosophies.
516 Support for exception handling has been improved; more targets are now
517 supported, and throws will use the RTTI mechanism to match against the
518 catch parameter type. Optimization is NOT SUPPORTED with
519 -fhandle-exceptions; no need to report this as a bug.
521 Support for Run-Time Type Identification has been added with -frtti.
522 This support is still in alpha; one major restriction is that any file
523 compiled with -frtti must include <typeinfo.h>.
525 Preliminary support for namespaces has been added. This support is far
526 from complete, and probably not useful.
528 Synthesis of compiler-generated constructors, destructors and
529 assignment operators is now deferred until the functions are used.
531 The parsing of expressions such as `a ? b : c = 1' has changed from
532 `(a ? b : c) = 1' to `a : b ? (c = 1)'.
534 The code generated for testing conditions, especially those using ||
535 and &&, is now more efficient.
537 The operator keywords and, and_eq, bitand, bitor, compl, not, not_eq,
538 or, or_eq, xor and xor_eq are now supported. Use -ansi or
539 -foperator-names to enable them.
541 The 'explicit' keyword is now supported. 'explicit' is used to mark
542 constructors and type conversion operators that should not be used
545 g++ now accepts the typename keyword, though it currently has no
546 semantics; it can be a no-op in the current template implementation.
547 You may want to start using it in your code, however, since the
548 pending rewrite of the template implementation to compile STL properly
549 (perhaps for 2.8.0, perhaps not) will require you to use it as
550 indicated by the current draft.
552 Handling of user-defined type conversion has been overhauled so that
553 type conversion operators are now found and used properly in
554 expressions and function calls.
556 -fno-strict-prototype now only applies to function declarations with
559 g++ now warns about 'if (x=0)' with -Wparentheses or -Wall.
561 #pragma weak and #pragma pack are supported on System V R4 targets, as
562 are various other target-specific #pragmas supported by gcc.
564 new and delete of const types is now allowed (with no additional
567 Explicit instantiation of template methods is now supported. Also,
568 'inline template class foo<int>;' can be used to emit only the vtable
569 for a template class.
571 With -fcheck-new, g++ will check the return value of all calls to
572 operator new, and not attempt to modify a returned null pointer.
574 The template instantiation code now handles more conversions when
575 passing to a parameter that does not depend on template arguments.
576 This means that code like 'string s; cout << s;' now works.
578 Invalid jumps in a switch statement past declarations that require
579 initializations are now caught.
581 Functions declared 'extern inline' now have the same linkage semantics
582 as inline member functions. On supported targets, where previously
583 these functions (and vtables, and template instantiations) would have
584 been defined statically, they will now be defined as weak symbols so
585 that only one out-of-line definition is used.
587 collect2 now demangles linker output, and c++filt has become part of
588 the gcc distribution.
590 Noteworthy changes in GCC version 2.6.3:
592 A few more bugs have been fixed.
594 Noteworthy changes in GCC version 2.6.2:
596 A few bugs have been fixed.
598 Names of attributes can now be preceded and followed by double underscores.
600 Noteworthy changes in GCC version 2.6.1:
602 Numerous (mostly minor) bugs have been fixed.
604 The following new configurations are supported:
606 GNU on x86 (instead of treating it like MACH)
607 NetBSD on Sparc and Motorola 68k
608 AIX 4.1 on RS/6000 and PowerPC systems
609 Sequent DYNIX/ptx 1.x and 2.x.
610 Both COFF and ELF configurations on AViiON without using /bin/gcc
611 Windows/NT on x86 architecture; preliminary
612 AT&T DSP1610 digital signal processor chips
613 i960 systems on bare boards using COFF
614 PDP11; target only and not extensively tested
616 The -pg option is now supported for Alpha under OSF/1 V3.0 or later.
618 Files with an extension of ".c++" are treated as C++ code.
620 The -Xlinker and -Wl arguments are now passed to the linker in the
621 position they were specified on the command line. This makes it
622 possible, for example, to pass flags to the linker about specific
625 The use of positional arguments to the configure script is no longer
626 recommended. Use --target= to specify the target; see the GCC manual.
628 The 386 now supports two new switches: -mreg-alloc=<string> changes
629 the default register allocation order used by the compiler, and
630 -mno-wide-multiply disables the use of the mul/imul instructions that
631 produce 64 bit results in EAX:EDX from 32 bit operands to do long long
632 multiplies and 32-bit division by constants.
634 Noteworthy changes in GCC version 2.6.0:
636 Numerous bugs have been fixed, in the C and C++ front-ends, as
637 well as in the common compiler code.
639 This release includes the C, Objective-C, and C++ compilers. However,
640 we have moved the files for the C++ compiler (G++) files to a
641 subdirectory, cp. Subsequent releases of GCC will split these files
642 to a separate TAR file.
644 The G++ team has been tracking the development of the ANSI standard for C++.
645 Here are some new features added from the latest working paper:
647 * built-in boolean type 'bool', with constants 'true' and 'false'.
648 * array new and delete (operator new [] and delete []).
649 * WP-conforming lifetime of temporaries.
650 * explicit instantiation of templates (template class A<int>;),
651 along with an option (-fno-implicit-templates) to disable emission
652 of implicitly instantiated templates, obsoletes -fexternal-templates.
653 * static member constants (static const int foo = 4; within the
656 Many error messages have been improved to tell the user more about the
657 problem. Conformance checking with -pedantic-errors has been
658 improved. G++ now compiles Fresco.
660 There is now an experimental implementation of virtual functions using
661 thunks instead of Cfront-style vtables, enabled with -fvtable-thunks.
662 This option also enables a heuristic which causes the compiler to only
663 emit the vtable in the translation unit where its first non-inline
664 virtual function is defined; using this option and
665 -fno-implicit-templates, users should be able to avoid #pragma
666 interface/implementation altogether.
668 Signatures have been added as a GNU C++ extension. Using the option
669 -fhandle-signatures, users are able to turn on recognition of
670 signatures. A short introduction on signatures is in the section
671 `Extension to the C++ Language' in the manual.
673 The `g++' program is now a C program, rather than a shell script.
675 Lots and lots and lots of bugs fixes, in nested types, access control,
676 pointers to member functions, the parser, templates, overload
677 resolution, etc, etc.
679 There have been two major enhancements to the Objective-C compiler:
681 1) Added portability. It now runs on Alpha, and some problems with
682 message forwarding have been addressed on other platforms.
684 2) Selectors have been redefined to be pointers to structs like:
685 { void *sel_id, char *sel_types }, where the sel_id is the unique
686 identifier, the selector itself is no longer unique.
688 Programmers should use the new function sel_eq to test selector
691 The following major changes have been made to the base compiler and
692 machine-specific files.
694 - The MIL-STD-1750A is a new port, but still preliminary.
696 - The h8/300h is now supported; both the h8/300 and h8/300h ports come
697 with 32 bit IEEE 754 software floating point support.
699 - The 64-bit Sparc (v9) and 64-bit MIPS chips are supported.
701 - NetBSD is supported on m68k, Intel x86, and pc523 systems and FreeBSD
704 - COFF is supported on x86, m68k, and Sparc systems running LynxOS.
706 - 68K systems from Bull and Concurrent are supported and System V
707 Release 4 is supported on the Atari.
709 - GCC supports GAS on the Motorola 3300 (sysV68) and debugging
710 (assuming GAS) on the Plexus 68K system. (However, GAS does not yet
711 work on those systems).
713 - System V Release 4 is supported on MIPS (Tandem).
715 - For DG/UX, an ELF configuration is now supported, and both the ELF
716 and BCS configurations support ELF and COFF object file formats.
718 - OSF/1 V2.0 is supported on Alpha.
720 - Function profiling is also supported on Alpha.
722 - GAS and GDB is supported for Irix 5 (MIPS).
724 - "common mode" (code that will run on both POWER and PowerPC
725 architectures) is now supported for the RS/6000 family; the
726 compiler knows about more PPC chips.
728 - Both NeXTStep 2.1 and 3 are supported on 68k-based architectures.
730 - On the AMD 29k, the -msoft-float is now supported, as well as
731 -mno-sum-in-toc for RS/6000, -mapp-regs and -mflat for Sparc, and
732 -membedded-pic for MIPS.
734 - GCC can now convert division by integer constants into the equivalent
735 multiplication and shift operations when that is faster than the
738 - Two new warning options, -Wbad-function-cast and
739 -Wmissing-declarations have been added.
741 - Configurations may now add machine-specific __attribute__ options on
742 type; many machines support the `section' attribute.
744 - The -ffast-math flag permits some optimization that violate strict
745 IEEE rules, such as converting X * 0.0 to 0.0.
747 Noteworthy changes in GCC version 2.5.8:
749 This release only fixes a few serious bugs. These include fixes for a
750 bug that prevented most programs from working on the RS/6000, a bug
751 that caused invalid assembler code for programs with a `switch'
752 statement on the NS32K, a G++ problem that caused undefined names in
753 some configurations, and several less serious problems, some of which
754 can affect most configuration.
756 Noteworthy change in GCC version 2.5.7:
758 This release only fixes a few bugs, one of which was causing bootstrap
759 compare errors on some systems.
761 Noteworthy change in GCC version 2.5.6:
763 A few backend bugs have been fixed, some of which only occur on one
766 The C++ compiler in 2.5.6 includes:
768 * fixes for some common crashes
769 * correct handling of nested types that are referenced as `foo::bar'
770 * spurious warnings about friends being declared static and never
771 defined should no longer appear
772 * enums that are local to a method in a class, or a class that's
773 local to a function, are now handled correctly. For example:
774 class foo { void bar () { enum { x, y } E; x; } };
775 void bar () { class foo { enum { x, y } E; E baz; }; }
777 Noteworthy change in GCC version 2.5.5:
779 A large number of C++ bugs have been fixed.
781 The fixproto script adds prototypes conditionally on __cplusplus.
783 Noteworthy change in GCC version 2.5.4:
785 A bug fix in passing of structure arguments for the HP-PA architecture
786 makes code compiled with GCC 2.5.4 incompatible with code compiled
787 with earlier versions (if it passes struct arguments of 33 to 64 bits,
788 interspersed with other types of arguments).
790 Noteworthy change in gcc version 2.5.3:
792 The method of "mangling" C++ function names has been changed. So you
793 must recompile all C++ programs completely when you start using GCC
794 2.5. Also, GCC 2.5 requires libg++ version 2.5. Earlier libg++
795 versions won't work with GCC 2.5. (This is generally true--GCC
796 version M.N requires libg++ version M.N.)
798 Noteworthy GCC changes in version 2.5:
800 * There is now support for the IBM 370 architecture as a target.
801 Currently the only operating system supported is MVS; GCC does not run
802 on MVS, so you must produce .s files using GCC as a cross compiler,
803 then transfer them to MVS to assemble them. This port is not reliable
806 * The Power PC is now supported.
808 * The i860-based Paragon machine is now supported.
810 * The Hitachi 3050 (an HP-PA machine) is now supported.
812 * The variable __GNUC_MINOR__ holds the minor version number of GCC, as
813 an integer. For version 2.5.X, the value is 5.
815 * In C, initializers for static and global variables are now processed
816 an element at a time, so that they don't need a lot of storage.
818 * The C syntax for specifying which structure field comes next in an
819 initializer is now `.FIELDNAME='. The corresponding syntax for
820 array initializers is now `[INDEX]='. For example,
823 = { [' '] = 1, ['\t'] = 1, ['\n'] = 1 };
825 This was changed to accord with the syntax proposed by the Numerical
826 C Extensions Group (NCEG).
828 * Complex numbers are now supported in C. Use the keyword __complex__
829 to declare complex data types. See the manual for details.
831 * GCC now supports `long double' meaningfully on the Sparc (128-bit
832 floating point) and on the 386 (96-bit floating point). The Sparc
833 support is enabled on Solaris 2.x because earlier system versions
834 (SunOS 4) have bugs in the emulation.
836 * All targets now have assertions for cpu, machine and system. So you
837 can now use assertions to distinguish among all supported targets.
839 * Nested functions in C may now be inline. Just declare them inline
842 * Packed structure members are now supported fully; it should be possible
843 to access them on any supported target, no matter how little alignment
846 * To declare that a function does not return, you must now write
847 something like this (works only in 2.5):
849 void fatal () __attribute__ ((noreturn));
851 or like this (works in older versions too):
853 typedef void voidfn ();
855 volatile voidfn fatal;
857 It used to be possible to do so by writing this:
859 volatile void fatal ();
861 but it turns out that ANSI C requires that to mean something
862 else (which is useless).
864 Likewise, to declare that a function is side-effect-free
865 so that calls may be deleted or combined, write
866 something like this (works only in 2.5):
868 int computation () __attribute__ ((const));
870 or like this (works in older versions too):
872 typedef int intfn ();
874 const intfn computation;
876 * The new option -iwithprefixbefore specifies a directory to add to
877 the search path for include files in the same position where -I would
878 put it, but uses the specified prefix just like -iwithprefix.
880 * Basic block profiling has been enhanced to record the function the
881 basic block comes from, and if the module was compiled for debugging,
882 the line number and filename. A default version of the basic block
883 support module has been added to libgcc2 that appends the basic block
884 information to a text file 'bb.out'. Machine descriptions can now
885 override the basic block support module in the target macro file.
889 * The new flag `-fansi-overloading' for C++. Use a newly implemented
890 scheme of argument matching for C++. It makes g++ more accurately
891 obey the rules set down in Chapter 13 of the Annotated C++ Reference
892 Manual (the ARM). This option will be turned on by default in a
895 * The -finline-debug flag is now gone (it was never really used by the
898 * Recognizing the syntax for pointers to members, e.g., "foo::*bar", has been
899 dramatically improved. You should not get any syntax errors or incorrect
900 runtime results while using pointers to members correctly; if you do, it's
903 * Forward declaration of an enum is now flagged as an error.
905 * Class-local typedefs are now working properly.
907 * Nested class support has been significantly improved. The compiler
908 will now (in theory) support up to 240 nested classes before hitting
909 other system limits (like memory size).
911 * There is a new C version of the `g++' driver, to replace the old
912 shell script. This should significantly improve the performance of
913 executing g++ on a system where a user's PATH environment variable
914 references many NFS-mounted filesystems. This driver also works
915 under MS-DOS and OS/2.
917 * The ANSI committee working on the C++ standard has adopted a new
918 keyword `mutable'. This will allow you to make a specific member be
919 modifiable in an otherwise const class.
921 Noteworthy GCC changes in version 2.4.4:
923 A crash building g++ on various hosts (including m68k) has been
924 fixed. Also the g++ compiler no longer reports incorrect
925 ambiguities in some situations where they do not exist, and
926 const template member functions are now being found properly.
928 Noteworthy GCC changes in version 2.4:
930 * On each target, the default is now to return short structures
931 compatibly with the "usual" compiler on that target.
933 For most targets, this means the default is to return all structures
934 in memory, like long structures, in whatever way is used on that
935 target. Use -freg-struct-return to enable returning short structures
936 (and unions) in registers.
938 This change means that newly compiled binaries are incompatible with
939 binaries compiled with previous versions of GCC.
941 On some targets, GCC is itself the usual compiler. On these targets,
942 the default way to return short structures is still in registers.
943 Use -fpcc-struct-return to tell GCC to return them in memory.
945 * There is now a floating point emulator which can imitate the way all
946 supported target machines do floating point arithmetic.
948 This makes it possible to have cross compilation to and from the VAX,
949 and between machines of different endianness. However, this works
950 only when the target machine description is updated to use the new
951 facilities, and not all have been updated.
953 This also makes possible support for longer floating point types.
954 GCC 2.4 supports extended format on the 68K if you use `long double',
955 for targets that have a 68881. (When we have run time library
956 routines for extended floating point, then `long double' will use
957 extended format on all 68K targets.)
959 We expect to support extended floating point on the i386 and Sparc in
962 * Building GCC now automatically fixes the system's header files.
963 This should require no attention.
965 * GCC now installs an unsigned data type as size_t when it fixes the
966 header files (on all but a handful of old target machines).
967 Therefore, the bug that size_t failed to be unsigned is fixed.
969 * Building and installation are now completely separate.
970 All new files are constructed during the build process;
971 installation just copies them.
973 * New targets supported: Clipper, Hitachi SH, Hitachi 8300, and Sparc
976 * A totally new and much better Objective C run time system is included.
978 * Objective C supports many new features. Alas, I can't describe them
979 since I don't use that language; however, they are the same ones
980 supported in recent versions of the NeXT operating system.
982 * The builtin functions __builtin_apply_args, __builtin_apply and
983 __builtin_return let you record the arguments and returned
984 value of a function without knowing their number or type.
986 * The builtin string variables __FUNCTION__ and __PRETTY_FUNCTION__
987 give the name of the function in the source, and a pretty-printed
988 version of the name. The two are the same in C, but differ in C++.
990 * Casts to union types do not yield lvalues.
992 * ## before an empty rest argument discards the preceding sequence
993 of non-whitespace characters from the macro definition.
994 (This feature is subject to change.)
997 New features specific to C++:
999 * The manual contains a new section ``Common Misunderstandings with
1000 GNU C++'' that C++ users should read.
1002 * #pragma interface and #pragma implementation let you use the same
1003 C++ source file for both interface and implementation.
1004 However, this mechanism is still in transition.
1006 * Named returned values let you avoid an extra constructor call
1007 when a function result has a class type.
1009 * The C++ operators <? and >? yield min and max, respectively.
1011 * C++ gotos can exit a block safely even if the block has
1012 aggregates that require destructors.
1014 * gcc defines the macro __GNUG__ when compiling C++ programs.
1016 * GNU C++ now correctly distinguishes between the prefix and postfix
1017 forms of overloaded operator ++ and --. To avoid breaking old
1018 code, if a class defines only the prefix form, the compiler
1019 accepts either ++obj or obj++, unless -pedantic is used.
1021 * If you are using version 2.3 of libg++, you need to rebuild it with
1022 `make CC=gcc' to avoid mismatches in the definition of `size_t'.
1024 Newly documented compiler options:
1027 Omit the standard system startup files when linking.
1030 Consider memory references to extern and global data items to
1034 Add DIR to the second include path.
1037 Specify PREFIX for later -iwithprefix options.
1040 Add PREFIX/DIR to the second include path.
1043 Emit Sparc v8 code (with integer multiply and divide).
1045 Emit Sparclite code (roughly v7.5).
1047 -print-libgcc-file-name
1048 Search for the libgcc.a file, print its absolute file name, and exit.
1050 -Woverloaded-virtual
1051 Warn when a derived class function declaration may be an error
1052 in defining a C++ virtual function.
1054 -Wtemplate-debugging
1055 When using templates in a C++ program, warn if debugging is
1056 not yet fully available.
1059 Control how C++ virtual function definitions are used