1 Noteworthy changes in GCC version 2.8.0:
3 The following new targets are supported:
5 Solaris 2.5 running on UltraSPARC's.
6 The sparclet cpu (a.out file format).
7 Solaris 2.51 running on PowerPC's.
8 Linux running on PowerPC's.
10 New features for the Hitachi H8/300(H).
15 The -ms switch generates code for the Hitachi H8/S processor.
17 New features for SPARC-based systems:
21 -mimpure-text and -mno-impure-text
23 New features for PowerPC-based systems running either System V, Linux, Solaris,
38 Options -mno-v8 and -mno-sparclite are no longer supported on SPARC
39 targets. Options -mcypress, -mv8, -msupersparc, -msparclite, -mf930,
40 and -mf934 are deprecated and will be deleted in GCC 2.9.
41 Use -mcpu=xxx instead.
43 GCC on the PA requires either gas-2.7 or the HP assembler; for best
44 results using GAS is highly recommended. GAS is required for -g
45 and exception handling support.
47 Aix 4.1 GCC targets now default to -mcpu=common so that programs compiled can
48 be moved between rs6000 and powerpc based systems. A consequence of this is
49 that -static won't work, and that some programs may be slightly slower.
51 You can select the default value to use for -mcpu=xxx on rs6000 and powerpc
52 targets by using the --with-cpu=xxx option when confiuring the compiler. In
53 addition, a new options, -mtune=xxx was added that selects the machine to
54 schedule for but does not select the architecture level.
56 Directory names used for storing the multilib libraries on System V and
57 embedded PowerPC systems have been shortened to work with commands like tar
58 that have fixed limits on pathname size.
60 Noteworthy changes in GCC version 2.7.2:
62 A few bugs have been fixed (most notably the generation of an
63 invalid assembler opcode on some RS/6000 systems).
65 Noteworthy changes in GCC version 2.7.1:
67 This release fixes numerous bugs (mostly minor) in GCC 2.7.0, but
68 also contains a few new features, mostly related to specific targets.
70 Major changes have been made in code to support Windows NT.
72 The following new targets are supported:
76 HP/UX version 10 on HP PA RISC (treated like version 9)
77 DEC Alpha running Windows NT
79 When parsing C, GCC now recognizes C++ style `//' comments unless you
80 specify `-ansi' or `-traditional'.
82 The PowerPC System V targets (powerpc-*-sysv, powerpc-*-eabi) now use the
83 calling sequence specified in the System V Application Binary Interface
84 Processor Supplement (PowerPC Processor ABI Supplement) rather than the calling
85 sequence used in GCC version 2.7.0. That calling sequence was based on the AIX
86 calling sequence without function descriptors. To compile code for that older
87 calling sequence, either configure the compiler for powerpc-*-eabiaix or use
88 the -mcall-aix switch when compiling and linking.
90 Noteworthy changes in GCC version 2.7.0:
92 GCC now works better on systems that use ".obj" and ".exe" instead of
93 ".o" and no extension. This involved changes to the driver program,
94 gcc.c, to convert ".o" names to ".obj" and to GCC's Makefile to use
95 ".obj" and ".exe" in filenames that are not targets. In order to
96 build GCC on such systems, you may need versions of GNU make and/or
97 compatible shells. At this point, this support is preliminary.
99 Object file extensions of ".obj" and executable file extensions of
100 ".exe" are allowed when using appropriate version of GNU Make.
102 Numerous enhancements were made to the __attribute__ facility including
103 more attributes and more places that support it. We now support the
104 "packed", "nocommon", "noreturn", "volatile", "const", "unused",
105 "transparent_union", "constructor", "destructor", "mode", "section",
106 "align", "format", "weak", and "alias" attributes. Each of these
107 names may also be specified with added underscores, e.g., "__packed__".
108 __attribute__ may now be applied to parameter definitions, function
109 definitions, and structure, enum, and union definitions.
111 GCC now supports returning more structures in registers, as specified by
112 many calling sequences (ABIs), such as on the HP PA RISC.
114 A new option '-fpack-struct' was added to automatically pack all structure
115 members together without holes.
117 There is a new library (cpplib) and program (cppmain) that at some
118 point will replace cpp (aka cccp). To use cppmain as cpp now, pass
119 the option CCCP=cppmain to make. The library is already used by the
120 fix-header program, which should speed up the fixproto script.
122 New options for supported targets:
125 NetBSD on MIPS, m68k, VAX, and x86.
126 LynxOS on x86, m68k, Sparc, and RS/6000.
127 VxWorks on many targets.
129 Windows/NT on x86 architecture. Initial support for Windows/NT on Alpha
132 Many embedded targets, specifically UDI on a29k, aout, coff, elf,
133 and vsta "operating systems" on m68k, m88k, mips, sparc, and x86.
135 Additional support for x86 (i386, i486, and Pentium):
137 Work with old and new linkers for Linux-based GNU systems,
138 supporting both a.out and ELF.
141 -malign-double, -mregparm=, -malign-loops= and -malign-jumps= switches.
142 On ISC systems, support -Xp like -posix.
144 Additions for RS/6000:
146 Instruction scheduling information for PowerPC 403.
148 -mstring and -mno-string.
149 -msoft-float and floating-point emulation included.
150 Preliminary support for PowerPC System V.4 with or without the GNU as.
151 Preliminary support for EABI.
152 Preliminary support for 64-bit systems.
153 Both big and little endian systems.
155 New features for MIPS-based systems:
161 Allow dollar signs in labels on SGI/Irix 5.x.
163 New support for HP PA RISC:
165 Generation of PIC (requires binutils-2.5.2.u6 or later).
166 HP-UX version 9 on HP PA RISC (dynamically links even with -g).
167 Processor variants for HP PA RISC: 700, 7100, and 7100LC.
168 Automatic generation of long calls when needed.
169 -mfast-indirect-calls for kernels and static binaries.
171 The called routine now copies arguments passed by invisible reference,
172 as required by the calling standard.
174 Other new miscellaneous target-specific support:
177 -mold-align for i960.
178 Configuration for "semi-hosted" ARM.
179 -momit-leaf-frame-pointer for M88k.
180 SH3 variant of Hitachi Super-H and support both big and little endian.
182 Changes to Objective-C:
184 Bare-bones implementation of NXConstantString has been added,
185 which is invoked by the @"string" directive.
187 Class * has been changed to Class to conform to the NextSTEP and
190 Enhancements to make dynamic loading easier.
192 The module version number has been updated to Version 7, thus existing
193 code will need to be recompiled to use the current run-time library.
195 GCC now supports the ISO Normative Addendum 1 to the C Standard.
198 The header <iso646.h> defines macros for C programs written
199 in national variants of ISO 646.
201 The following digraph tokens are supported:
203 These behave like the following, respectively:
206 Digraph tokens are supported unless you specify the `-traditional'
207 option; you do not need to specify `-ansi' or `-trigraphs'. Except
208 for contrived and unlikely examples involving preprocessor
209 stringizing, digraph interpretation doesn't change the meaning of
210 programs; this is unlike trigraph interpretation, which changes the
211 meanings of relatively common strings.
213 The macro __STDC_VERSION__ has the value 199409L.
215 As usual, for full conformance to the standard, you also need a
216 C library that conforms.
218 The following lists changes that have been made to g++. If some
219 features mentioned below sound unfamiliar, you will probably want to
220 look at the recently-released public review copy of the C++ Working
221 Paper. For PostScript and PDF (Adobe Acrobat) versions, see the
222 archive at ftp://research.att.com/dist/stdc++/WP. For HTML and ASCII
223 versions, see ftp://ftp.cygnus.com/pub/g++. On the web, see
224 http://www.cygnus.com/~mrs/wp-draft.
226 The scope of variables declared in the for-init-statement has been changed
227 to conform to http://www.cygnus.com/~mrs/wp-draft/stmt.html#stmt.for; as a
228 result, packages such as groff 1.09 will not compile unless you specify the
229 -fno-for-scope flag. PLEASE DO NOT REPORT THIS AS A BUG; this is a change
230 mandated by the C++ standardization committee.
232 Binary incompatibilities:
234 The builtin 'bool' type is now the size of a machine word on RISC targets,
235 for code efficiency; it remains one byte long on CISC targets.
237 Code that does not use #pragma interface/implementation will most
238 likely shrink dramatically, as g++ now only emits the vtable for a
239 class in the translation unit where its first non-inline, non-abstract
240 virtual function is defined.
242 Classes that do not define the copy constructor will sometimes be
243 passed and returned in registers. This may illuminate latent bugs in
246 Support for automatic template instantiation has *NOT* been added, due
247 to a disagreement over design philosophies.
249 Support for exception handling has been improved; more targets are now
250 supported, and throws will use the RTTI mechanism to match against the
251 catch parameter type. Optimization is NOT SUPPORTED with
252 -fhandle-exceptions; no need to report this as a bug.
254 Support for Run-Time Type Identification has been added with -frtti.
255 This support is still in alpha; one major restriction is that any file
256 compiled with -frtti must include <typeinfo.h>.
258 Preliminary support for namespaces has been added. This support is far
259 from complete, and probably not useful.
261 Synthesis of compiler-generated constructors, destructors and
262 assignment operators is now deferred until the functions are used.
264 The parsing of expressions such as `a ? b : c = 1' has changed from
265 `(a ? b : c) = 1' to `a : b ? (c = 1)'.
267 The code generated for testing conditions, especially those using ||
268 and &&, is now more efficient.
270 The operator keywords and, and_eq, bitand, bitor, compl, not, not_eq,
271 or, or_eq, xor and xor_eq are now supported. Use -ansi or
272 -foperator-names to enable them.
274 The 'explicit' keyword is now supported. 'explicit' is used to mark
275 constructors and type conversion operators that should not be used
278 g++ now accepts the typename keyword, though it currently has no
279 semantics; it can be a no-op in the current template implementation.
280 You may want to start using it in your code, however, since the
281 pending rewrite of the template implementation to compile STL properly
282 (perhaps for 2.8.0, perhaps not) will require you to use it as
283 indicated by the current draft.
285 Handling of user-defined type conversion has been overhauled so that
286 type conversion operators are now found and used properly in
287 expressions and function calls.
289 -fno-strict-prototype now only applies to function declarations with
292 g++ now warns about 'if (x=0)' with -Wparentheses or -Wall.
294 #pragma weak and #pragma pack are supported on System V R4 targets, as
295 are various other target-specific #pragmas supported by gcc.
297 new and delete of const types is now allowed (with no additional
300 Explicit instantiation of template methods is now supported. Also,
301 'inline template class foo<int>;' can be used to emit only the vtable
302 for a template class.
304 With -fcheck-new, g++ will check the return value of all calls to
305 operator new, and not attempt to modify a returned null pointer.
307 The template instantiation code now handles more conversions when
308 passing to a parameter that does not depend on template arguments.
309 This means that code like 'string s; cout << s;' now works.
311 Invalid jumps in a switch statement past declarations that require
312 initializations are now caught.
314 Functions declared 'extern inline' now have the same linkage semantics
315 as inline member functions. On supported targets, where previously
316 these functions (and vtables, and template instantiations) would have
317 been defined statically, they will now be defined as weak symbols so
318 that only one out-of-line definition is used.
320 collect2 now demangles linker output, and c++filt has become part of
321 the gcc distribution.
323 Noteworthy changes in GCC version 2.6.3:
325 A few more bugs have been fixed.
327 Noteworthy changes in GCC version 2.6.2:
329 A few bugs have been fixed.
331 Names of attributes can now be preceded and followed by double underscores.
333 Noteworthy changes in GCC version 2.6.1:
335 Numerous (mostly minor) bugs have been fixed.
337 The following new configurations are supported:
339 GNU on x86 (instead of treating it like MACH)
340 NetBSD on Sparc and Motorola 68k
341 AIX 4.1 on RS/6000 and PowerPC systems
342 Sequent DYNUX/ptx 1.x and 2.x.
343 Both COFF and ELF configurations on AViiON without using /bin/gcc
344 Windows/NT on x86 architecture; preliminary
345 AT&T DSP1610 digital signal processor chips
346 i960 systems on bare boards using COFF
347 PDP11; target only and not extensively tested
349 The -pg option is now supported for Alpha under OSF/1 V3.0 or later.
351 Files with an extension of ".c++" are treated as C++ code.
353 The -Xlinker and -Wl arguments are now passed to the linker in the
354 position they were specified on the command line. This makes it
355 possible, for example, to pass flags to the linker about specific
358 The use of positional arguments to the configure script is no longer
359 recommended. Use --target= to specify the target; see the GCC manual.
361 The 386 now supports two new switches: -mreg-alloc=<string> changes
362 the default register allocation order used by the compiler, and
363 -mno-wide-multiply disables the use of the mul/imul instructions that
364 produce 64 bit results in EAX:EDX from 32 bit operands to do long long
365 multiplies and 32-bit division by constants.
367 Noteworthy changes in GCC version 2.6.0:
369 Numerous bugs have been fixed, in the C and C++ front-ends, as
370 well as in the common compiler code.
372 This release includes the C, Objective-C, and C++ compilers. However,
373 we have moved the files for the C++ compiler (G++) files to a
374 subdirectory, cp. Subsequent releases of GCC will split these files
375 to a separate TAR file.
377 The G++ team has been tracking the development of the ANSI standard for C++.
378 Here are some new features added from the latest working paper:
380 * built-in boolean type 'bool', with constants 'true' and 'false'.
381 * array new and delete (operator new [] and delete []).
382 * WP-conforming lifetime of temporaries.
383 * explicit instantiation of templates (template class A<int>;),
384 along with an option (-fno-implicit-templates) to disable emission
385 of implicitly instantiated templates, obsoletes -fexternal-templates.
386 * static member constants (static const int foo = 4; within the
389 Many error messages have been improved to tell the user more about the
390 problem. Conformance checking with -pedantic-errors has been
391 improved. G++ now compiles Fresco.
393 There is now an experimental implementation of virtual functions using
394 thunks instead of Cfront-style vtables, enabled with -fvtable-thunks.
395 This option also enables a heuristic which causes the compiler to only
396 emit the vtable in the translation unit where its first non-inline
397 virtual function is defined; using this option and
398 -fno-implicit-templates, users should be able to avoid #pragma
399 interface/implementation altogether.
401 Signatures have been added as a GNU C++ extension. Using the option
402 -fhandle-signatures, users are able to turn on recognition of
403 signatures. A short introduction on signatures is in the section
404 `Extension to the C++ Language' in the manual.
406 The `g++' program is now a C program, rather than a shell script.
408 Lots and lots and lots of bugs fixes, in nested types, access control,
409 pointers to member functions, the parser, templates, overload
410 resolution, etc, etc.
412 There have been two major enhancements to the Objective-C compiler:
414 1) Added portability. It now runs on Alpha, and some problems with
415 message forwarding have been addressed on other platforms.
417 2) Selectors have been redefined to be pointers to structs like:
418 { void *sel_id, char *sel_types }, where the sel_id is the unique
419 identifier, the selector itself is no longer unique.
421 Programmers should use the new function sel_eq to test selector
424 The following major changes have been made to the base compiler and
425 machine-specific files.
427 - The MIL-STD-1750A is a new port, but still preliminary.
429 - The h8/300h is now supported; both the h8/300 and h8/300h ports come
430 with 32 bit IEEE 754 software floating point support.
432 - The 64-bit Sparc (v9) and 64-bit MIPS chips are supported.
434 - NetBSD is supported on m68k, Intel x86, and pc523 systems and FreeBSD
437 - COFF is supported on x86, m68k, and Sparc systems running LynxOS.
439 - 68K systems from Bull and Concurrent are supported and System V
440 Release 4 is supported on the Atari.
442 - GCC supports GAS on the Motorola 3300 (sysV68) and debugging
443 (assuming GAS) on the Plexus 68K system. (However, GAS does not yet
444 work on those systems).
446 - System V Release 4 is supported on MIPS (Tandem).
448 - For DG/UX, an ELF configuration is now supported, and both the ELF
449 and BCS configurations support ELF and COFF object file formats.
451 - OSF/1 V2.0 is supported on Alpha.
453 - Function profiling is also supported on Alpha.
455 - GAS and GDB is supported for Irix 5 (MIPS).
457 - "common mode" (code that will run on both POWER and PowerPC
458 architectures) is now supported for the RS/6000 family; the
459 compiler knows about more PPC chips.
461 - Both NeXTStep 2.1 and 3 are supported on 68k-based architectures.
463 - On the AMD 29k, the -msoft-float is now supported, as well as
464 -mno-sum-in-toc for RS/6000, -mapp-regs and -mflat for Sparc, and
465 -membedded-pic for MIPS.
467 - GCC can now convert division by integer constants into the equivalent
468 multiplication and shift operations when that is faster than the
471 - Two new warning options, -Wbad-function-cast and
472 -Wmissing-declarations have been added.
474 - Configurations may now add machine-specific __attribute__ options on
475 type; many machines support the `section' attribute.
477 - The -ffast-math flag permits some optimization that violate strict
478 IEEE rules, such as converting X * 0.0 to 0.0.
480 Noteworthy changes in GCC version 2.5.8:
482 This release only fixes a few serious bugs. These include fixes for a
483 bug that prevented most programs from working on the RS/6000, a bug
484 that caused invalid assembler code for programs with a `switch'
485 statement on the NS32K, a G++ problem that caused undefined names in
486 some configurations, and several less serious problems, some of which
487 can affect most configuration.
489 Noteworthy change in GCC version 2.5.7:
491 This release only fixes a few bugs, one of which was causing bootstrap
492 compare errors on some systems.
494 Noteworthy change in GCC version 2.5.6:
496 A few backend bugs have been fixed, some of which only occur on one
499 The C++ compiler in 2.5.6 includes:
501 * fixes for some common crashes
502 * correct handling of nested types that are referenced as `foo::bar'
503 * spurious warnings about friends being declared static and never
504 defined should no longer appear
505 * enums that are local to a method in a class, or a class that's
506 local to a function, are now handled correctly. For example:
507 class foo { void bar () { enum { x, y } E; x; } };
508 void bar () { class foo { enum { x, y } E; E baz; }; }
510 Noteworthy change in GCC version 2.5.5:
512 A large number of C++ bugs have been fixed.
514 The fixproto script adds prototypes conditionally on __cplusplus.
516 Noteworthy change in GCC version 2.5.4:
518 A bug fix in passing of structure arguments for the HP-PA architecture
519 makes code compiled with GCC 2.5.4 incompatible with code compiled
520 with earlier versions (if it passes struct arguments of 33 to 64 bits,
521 interspersed with other types of arguments).
523 Noteworthy change in gcc version 2.5.3:
525 The method of "mangling" C++ function names has been changed. So you
526 must recompile all C++ programs completely when you start using GCC
527 2.5. Also, GCC 2.5 requires libg++ version 2.5. Earlier libg++
528 versions won't work with GCC 2.5. (This is generally true--GCC
529 version M.N requires libg++ version M.N.)
531 Noteworthy GCC changes in version 2.5:
533 * There is now support for the IBM 370 architecture as a target.
534 Currently the only operating system supported is MVS; GCC does not run
535 on MVS, so you must produce .s files using GCC as a cross compiler,
536 then transfer them to MVS to assemble them. This port is not reliable
539 * The Power PC is now supported.
541 * The i860-based Paragon machine is now supported.
543 * The Hitachi 3050 (an HP-PA machine) is now supported.
545 * The variable __GNUC_MINOR__ holds the minor version number of GCC, as
546 an integer. For version 2.5.X, the value is 5.
548 * In C, initializers for static and global variables are now processed
549 an element at a time, so that they don't need a lot of storage.
551 * The C syntax for specifying which structure field comes next in an
552 initializer is now `.FIELDNAME='. The corresponding syntax for
553 array initializers is now `[INDEX]='. For example,
556 = { [' '] = 1, ['\t'] = 1, ['\n'] = 1 };
558 This was changed to accord with the syntax proposed by the Numerical
559 C Extensions Group (NCEG).
561 * Complex numbers are now supported in C. Use the keyword __complex__
562 to declare complex data types. See the manual for details.
564 * GCC now supports `long double' meaningfully on the Sparc (128-bit
565 floating point) and on the 386 (96-bit floating point). The Sparc
566 support is enabled on on Solaris 2.x because earlier system versions
567 (SunOS 4) have bugs in the emulation.
569 * All targets now have assertions for cpu, machine and system. So you
570 can now use assertions to distinguish among all supported targets.
572 * Nested functions in C may now be inline. Just declare them inline
575 * Packed structure members are now supported fully; it should be possible
576 to access them on any supported target, no matter how little alignment
579 * To declare that a function does not return, you must now write
580 something like this (works only in 2.5):
582 void fatal () __attribute__ ((noreturn));
584 or like this (works in older versions too):
586 typedef void voidfn ();
588 volatile voidfn fatal;
590 It used to be possible to do so by writing this:
592 volatile void fatal ();
594 but it turns out that ANSI C requires that to mean something
595 else (which is useless).
597 Likewise, to declare that a function is side-effect-free
598 so that calls may be deleted or combined, write
599 something like this (works only in 2.5):
601 int computation () __attribute__ ((const));
603 or like this (works in older versions too):
605 typedef int intfn ();
607 const intfn computation;
609 * The new option -iwithprefixbefore specifies a directory to add to
610 the search path for include files in the same position where -I would
611 put it, but uses the specified prefix just like -iwithprefix.
613 * Basic block profiling has been enhanced to record the function the
614 basic block comes from, and if the module was compiled for debugging,
615 the line number and filename. A default version of the basic block
616 support module has been added to libgcc2 that appends the basic block
617 information to a text file 'bb.out'. Machine descriptions can now
618 override the basic block support module in the target macro file.
622 * The new flag `-fansi-overloading' for C++. Use a newly implemented
623 scheme of argument matching for C++. It makes g++ more accurately
624 obey the rules set down in Chapter 13 of the Annotated C++ Reference
625 Manual (the ARM). This option will be turned on by default in a
628 * The -finline-debug flag is now gone (it was never really used by the
631 * Recognizing the syntax for pointers to members, e.g., "foo::*bar", has been
632 dramatically improved. You should not get any syntax errors or incorrect
633 runtime results while using pointers to members correctly; if you do, it's
636 * Forward declaration of an enum is now flagged as an error.
638 * Class-local typedefs are now working properly.
640 * Nested class support has been significantly improved. The compiler
641 will now (in theory) support up to 240 nested classes before hitting
642 other system limits (like memory size).
644 * There is a new C version of the `g++' driver, to replace the old
645 shell script. This should significantly improve the performance of
646 executing g++ on a system where a user's PATH environment variable
647 references many NFS-mounted filesystems. This driver also works
648 under MS-DOS and OS/2.
650 * The ANSI committee working on the C++ standard has adopted a new
651 keyword `mutable'. This will allow you to make a specific member be
652 modifiable in an otherwise const class.
654 Noteworthy GCC changes in version 2.4.4:
656 A crash building g++ on various hosts (including m68k) has been
657 fixed. Also the g++ compiler no longer reports incorrect
658 ambiguities in some situations where they do not exist, and
659 const template member functions are now being found properly.
661 Noteworthy GCC changes in version 2.4:
663 * On each target, the default is now to return short structures
664 compatibly with the "usual" compiler on that target.
666 For most targets, this means the default is to return all structures
667 in memory, like long structures, in whatever way is used on that
668 target. Use -freg-struct-return to enable returning short structures
669 (and unions) in registers.
671 This change means that newly compiled binaries are incompatible with
672 binaries compiled with previous versions of GCC.
674 On some targets, GCC is itself the usual compiler. On these targets,
675 the default way to return short structures is still in registers.
676 Use -fpcc-struct-return to tell GCC to return them in memory.
678 * There is now a floating point emulator which can imitate the way all
679 supported target machines do floating point arithmetic.
681 This makes it possible to have cross compilation to and from the VAX,
682 and between machines of different endianness. However, this works
683 only when the target machine description is updated to use the new
684 facilities, and not all have been updated.
686 This also makes possible support for longer floating point types.
687 GCC 2.4 supports extended format on the 68K if you use `long double',
688 for targets that have a 68881. (When we have run time library
689 routines for extended floating point, then `long double' will use
690 extended format on all 68K targets.)
692 We expect to support extended floating point on the i386 and Sparc in
695 * Building GCC now automatically fixes the system's header files.
696 This should require no attention.
698 * GCC now installs an unsigned data type as size_t when it fixes the
699 header files (on all but a handful of old target machines).
700 Therefore, the bug that size_t failed to be unsigned is fixed.
702 * Building and installation are now completely separate.
703 All new files are constructed during the build process;
704 installation just copies them.
706 * New targets supported: Clipper, Hitachi SH, Hitachi 8300, and Sparc
709 * A totally new and much better Objective C run time system is included.
711 * Objective C supports many new features. Alas, I can't describe them
712 since I don't use that language; however, they are the same ones
713 supported in recent versions of the NeXT operating system.
715 * The builtin functions __builtin_apply_args, __builtin_apply and
716 __builtin_return let you record the arguments and returned
717 value of a function without knowing their number or type.
719 * The builtin string variables __FUNCTION__ and __PRETTY_FUNCTION__
720 give the name of the function in the source, and a pretty-printed
721 version of the name. The two are the same in C, but differ in C++.
723 * Casts to union types do not yield lvalues.
725 * ## before an empty rest argument discards the preceding sequence
726 of non-whitespace characters from the macro definition.
727 (This feature is subject to change.)
730 New features specific to C++:
732 * The manual contains a new section ``Common Misunderstandings with
733 GNU C++'' that C++ users should read.
735 * #pragma interface and #pragma implementation let you use the same
736 C++ source file for both interface and implementation.
737 However, this mechanism is still in transition.
739 * Named returned values let you avoid an extra constructor call
740 when a function result has a class type.
742 * The C++ operators <? and >? yield min and max, respectively.
744 * C++ gotos can exit a block safely even if the block has
745 aggregates that require destructors.
747 * gcc defines the macro __GNUG__ when compiling C++ programs.
749 * GNU C++ now correctly distinguishes between the prefix and postfix
750 forms of overloaded operator ++ and --. To avoid breaking old
751 code, if a class defines only the prefix form, the compiler
752 accepts either ++obj or obj++, unless -pedantic is used.
754 * If you are using version 2.3 of libg++, you need to rebuild it with
755 `make CC=gcc' to avoid mismatches in the definition of `size_t'.
757 Newly documented compiler options:
760 Omit the standard system startup files when linking.
763 Consider memory references to extern and global data items to
767 Add DIR to the second include path.
770 Specify PREFIX for later -iwithprefix options.
773 Add PREFIX/DIR to the second include path.
776 Emit Sparc v8 code (with integer multiply and divide).
778 Emit Sparclite code (roughly v7.5).
780 -print-libgcc-file-name
781 Search for the libgcc.a file, print its absolute file name, and exit.
784 Warn when a derived class function declaration may be an error
785 in defining a C++ virtual function.
788 When using templates in a C++ program, warn if debugging is
789 not yet fully available.
792 Control how C++ virtual function definitions are used