]>
Commit | Line | Data |
---|---|---|
f2d76545 | 1 | Noteworthy changes in GCC for EGCS. |
fc5a8790 | 2 | |
956d6950 JL |
3 | A major change in this release is the addition of a framework for |
4 | exception handling, currently used by C++. Many internal changes and | |
5 | optimization improvements have been made. These increase the | |
6 | maintainability and portability of GCC. GCC now uses autoconf to | |
7 | compute many host parameters. | |
c6258ee2 | 8 | |
956d6950 | 9 | The following lists changes that add new features or targets. |
861bb6c1 | 10 | |
956d6950 | 11 | See cp/NEWS for new features of C++ in this release. |
fc5a8790 | 12 | |
956d6950 | 13 | New tools and features: |
fc5a8790 | 14 | |
956d6950 JL |
15 | The Dwarf 2 debugging information format is supported on ELF systems, and |
16 | is the default for -ggdb where available. It can also be used for C++. | |
17 | The Dwarf version 1 debugging format is also permitted for C++, but | |
18 | does not work well. | |
fc5a8790 | 19 | |
956d6950 JL |
20 | gcov.c is provided for test coverage analysis and branch profiling |
21 | analysis is also supported; see -fprofile-arcs, -ftest-coverage, | |
22 | and -fbranch-probabilities. | |
23 | ||
24 | Support for the Checker memory checking tool. | |
25 | ||
26 | New switch, -fstack-check, to check for stack overflow on systems that | |
27 | don't have such built into their ABI. | |
28 | ||
29 | New switches, -Wundef and -Wno-undef to warn if an undefined identifier | |
30 | is evaluated in an #if directive. | |
31 | ||
32 | Options -Wall and -Wimplicit now cause GCC to warn about implicit int | |
33 | in declarations (e.g. `register i;'), since the C Standard committee | |
34 | has decided to disallow this in the next revision of the standard; | |
35 | -Wimplicit-function-declarations and -Wimplicit-int are subsets of | |
36 | this. | |
37 | ||
38 | Option -Wsign-compare causes GCC to warn about comparison of signed and | |
39 | unsigned values. | |
40 | ||
41 | Add -dI option of cccp for cxref. | |
42 | ||
43 | New features in configuration, installation and specs file handling: | |
44 | ||
45 | New option --enable-c-cpplib to configure script. | |
46 | ||
47 | Allow --with-cpu on configure command to specify a default CPU. | |
48 | ||
49 | The -specs=file switch allows you to override default specs used in | |
50 | invoking programs like cc1, as, etc. | |
51 | ||
52 | Allow including one specs file from another and renaming a specs | |
53 | variable. | |
54 | ||
55 | You can now relocate all GCC files with a single environment variable | |
56 | or a registry entry under Windows 95 and Windows NT. | |
57 | ||
58 | Changes in Objective-C: | |
59 | ||
60 | The Objective-C Runtime Library has been made thread-safe. | |
61 | ||
62 | The Objective-C Runtime Library contains an interface for creating | |
63 | mutexes, condition mutexes, and threads; it requires a back-end | |
64 | implementation for the specific platform and/or thread package. | |
65 | Currently supported are DEC/OSF1, IRIX, Mach, OS/2, POSIX, PCThreads, | |
66 | Solaris, and Windows32. The --enable-threads parameter can be used | |
67 | when configuring GCC to enable and select a thread back-end. | |
68 | ||
69 | Objective-C is now configured as separate front-end language to GCC, | |
70 | making it more convenient to conditionally build it. | |
fc5a8790 | 71 | |
956d6950 JL |
72 | The internal structures of the Objective-C Runtime Library have |
73 | changed sufficiently to warrant a new version number; now version 8. | |
74 | Programs compiled with an older version must be recompiled. | |
75 | ||
76 | The Objective-C Runtime Library can be built as a DLL on Windows 95 | |
77 | and Windows NT systems. | |
78 | ||
79 | The Objective-C Runtime Library implements +load. | |
80 | ||
81 | The following new targets are supported (see also list under each | |
82 | individual CPU below): | |
83 | ||
84 | Embedded target m32r-elf. | |
85 | Embedded Hitachi Super-H using ELF. | |
86 | RTEMS real-time system on various CPU targets. | |
87 | ARC processor. | |
88 | NEC V850 processor. | |
89 | Matsushita MN10200 processor. | |
90 | Matsushita MN10300 processor. | |
91 | Sparc and PowerPC running on VxWorks. | |
92 | Support both glibc versions 1 and 2 on Linux-based GNU systems. | |
93 | ||
94 | New features for DEC Alpha systems: | |
95 | ||
96 | Allow detailed specification of IEEE fp support: | |
97 | -mieee, -mieee-with-inexact, and -mieee-conformant | |
98 | -mfp-trap-mode=xxx, -mfp-round-mode=xxx, -mtrap-precision=xxx | |
99 | -mcpu=xxx for CPU selection | |
100 | Support scheduling parameters for EV5. | |
101 | Add support for BWX, CIX, and MAX instruction set extensions. | |
102 | Support Linux-based GNU systems. | |
103 | Support VMS. | |
104 | ||
105 | Additional supported processors and systems for MIPS targets: | |
106 | ||
107 | MIPS4 instruction set. | |
108 | R4100, R4300 and R5000 processors. | |
109 | N32 and N64 ABI. | |
110 | IRIX 6.2. | |
111 | SNI SINIX. | |
112 | ||
113 | New features for Intel x86 family: | |
114 | ||
115 | Add scheduling parameters for Pentium and Pentium Pro. | |
116 | Support stabs on Solaris-x86. | |
117 | Intel x86 processors running the SCO OpenServer 5 family. | |
118 | Intel x86 processors running DG/UX. | |
119 | Intel x86 using Cygwin32 or Mingw32 on Windows 95 and Windows NT. | |
120 | ||
121 | New features for Motorola 68k family: | |
122 | ||
123 | Support for 68060 processor. | |
124 | More consistent switches to specify processor. | |
125 | Motorola 68k family running AUX. | |
126 | 68040 running pSOS, ELF object files, DBX debugging. | |
127 | Coldfire variant of Motorola m68k family. | |
128 | ||
129 | New features for the HP PA RISC: | |
130 | ||
131 | -mspace and m-no-space | |
132 | -mlong-load-store and -mno-long-load-store | |
133 | -mbig-switch -mno-big-switch | |
134 | ||
135 | GCC on the PA requires either gas-2.7 or the HP assembler; for best | |
136 | results using GAS is highly recommended. GAS is required for -g and | |
137 | exception handling support. | |
fc5a8790 JL |
138 | |
139 | New features for SPARC-based systems: | |
140 | ||
956d6950 JL |
141 | The ultrasparc cpu. |
142 | The sparclet cpu, supporting only a.out file format. | |
143 | Sparc running SunOS 4 with the GNU assembler. | |
144 | Sparc running the Linux-based GNU system. | |
145 | Embedded Sparc processors running the ELF object file format. | |
fc5a8790 JL |
146 | -mcpu=xxx |
147 | -mtune=xxx | |
f2d76545 JL |
148 | -malign-loops=xxx |
149 | -malign-jumps=xxx | |
150 | -malign-functions=xxx | |
fc5a8790 JL |
151 | -mimpure-text and -mno-impure-text |
152 | ||
956d6950 JL |
153 | Options -mno-v8 and -mno-sparclite are no longer supported on SPARC |
154 | targets. Options -mcypress, -mv8, -msupersparc, -msparclite, -mf930, | |
155 | and -mf934 are deprecated and will be deleted in GCC 2.9. Use | |
156 | -mcpu=xxx instead. | |
861bb6c1 | 157 | |
956d6950 | 158 | New features for rs6000 and PowerPC systems: |
861bb6c1 | 159 | |
956d6950 JL |
160 | Solaris 2.51 running on PowerPC's. |
161 | The Linux-based GNU system running on PowerPC's. | |
162 | -mcpu=604e,602,603e,620,801,823,mpc505,821,860,power2 | |
163 | -mtune=xxx | |
164 | -mrelocatable-lib, m-no-relocatable-lib | |
165 | -msim, -mmve, -memb | |
166 | -mupdate, -mno-update | |
167 | -mfused-madd, -mno-fused-madd | |
fc5a8790 JL |
168 | |
169 | -mregnames | |
170 | -meabi | |
956d6950 JL |
171 | -mcall-linux, -mcall-solaris, -mcall-sysv-eabi, -mcall-sysv-noeabi |
172 | -msdata, -msdata=none, -msdata=default, -msdata=sysv, -msdata=eabi | |
173 | -memb, -msim, -mmvme | |
174 | -myellowknife, -mads | |
175 | wchar_t is now of type long as per the ABI, not unsigned short. | |
861bb6c1 JL |
176 | -p/-pg support |
177 | -mcpu=403 now implies -mstrict-align. | |
956d6950 JL |
178 | Implement System V profiling. |
179 | ||
180 | Aix 4.1 GCC targets now default to -mcpu=common so that programs | |
181 | compiled can be moved between rs6000 and powerpc based systems. A | |
182 | consequence of this is that -static won't work, and that some programs | |
183 | may be slightly slower. | |
fc5a8790 | 184 | |
956d6950 JL |
185 | You can select the default value to use for -mcpu=xxx on rs6000 and |
186 | powerpc targets by using the --with-cpu=xxx option when configuring the | |
187 | compiler. In addition, a new options, -mtune=xxx was added that | |
188 | selects the machine to schedule for but does not select the | |
189 | architecture level. | |
fc5a8790 | 190 | |
956d6950 JL |
191 | Directory names used for storing the multilib libraries on System V |
192 | and embedded PowerPC systems have been shortened to work with commands | |
193 | like tar that have fixed limits on pathname size. | |
fc5a8790 | 194 | |
956d6950 JL |
195 | New features for the Hitachi H8/300(H): |
196 | ||
197 | -malign-300 | |
198 | -ms (for the Hitachi H8/S processor) | |
199 | -mint32 | |
fc5a8790 | 200 | |
956d6950 | 201 | New features for the ARM: |
fc5a8790 | 202 | |
956d6950 JL |
203 | -march=xxx, -mtune=xxx, -mcpu=xxx |
204 | Support interworking with Thumb code. | |
205 | ARM processor with a.out object format, COFF, or AOF assembler. | |
206 | ARM on "semi-hosted" platform. | |
207 | ARM running NetBSD. | |
208 | ARM running the Linux-based GNU system. | |
fc5a8790 | 209 | |
956d6950 | 210 | New feature for Solaris systems: |
f2d76545 | 211 | |
956d6950 JL |
212 | GCC installation no longer makes a copy of system include files, |
213 | thus insulating GCC better from updates to the operating system. | |
214 | ||
215 | Changes in Objective-C: | |
216 | ||
217 | The Objective-C Runtime Library has been made thread-safe. | |
218 | ||
219 | The Objective-C Runtime Library contains an interface for creating | |
220 | mutexes, condition mutexes, and threads; it requires a back-end | |
221 | implementation for the specific platform and/or thread package. | |
222 | Currently supported are DEC/OSF1, IRIX, Mach, OS/2, POSIX, PCThreads, | |
223 | Solaris, and Windows32. The --enable-threads parameter can be used | |
224 | when configuring GCC to enable and select a thread back-end. | |
225 | ||
226 | Objective-C is now configured as separate front-end language to GCC, | |
227 | making it more convenient to conditionally build it. | |
228 | ||
229 | The internal structures of the Objective-C Runtime Library have | |
230 | changed sufficiently to warrant a new version number; now version 8. | |
231 | Programs compiled with an older version must be recompiled. | |
232 | ||
233 | The Objective-C Runtime Library can be built as a DLL on Windows 95 | |
234 | and Windows NT systems. | |
235 | ||
236 | The Objective-C Runtime Library implements +load. | |
237 | \f | |
238 | Noteworthy changes in GCC version 2.7.2 | |
239 | --------------------------------------- | |
fc5a8790 JL |
240 | |
241 | A few bugs have been fixed (most notably the generation of an | |
242 | invalid assembler opcode on some RS/6000 systems). | |
243 | ||
956d6950 JL |
244 | Noteworthy changes in GCC version 2.7.1 |
245 | --------------------------------------- | |
fc5a8790 JL |
246 | |
247 | This release fixes numerous bugs (mostly minor) in GCC 2.7.0, but | |
248 | also contains a few new features, mostly related to specific targets. | |
249 | ||
250 | Major changes have been made in code to support Windows NT. | |
251 | ||
252 | The following new targets are supported: | |
253 | ||
254 | 2.9 BSD on PDP-11 | |
255 | Linux on m68k | |
256 | HP/UX version 10 on HP PA RISC (treated like version 9) | |
257 | DEC Alpha running Windows NT | |
258 | ||
259 | When parsing C, GCC now recognizes C++ style `//' comments unless you | |
260 | specify `-ansi' or `-traditional'. | |
261 | ||
262 | The PowerPC System V targets (powerpc-*-sysv, powerpc-*-eabi) now use the | |
263 | calling sequence specified in the System V Application Binary Interface | |
264 | Processor Supplement (PowerPC Processor ABI Supplement) rather than the calling | |
265 | sequence used in GCC version 2.7.0. That calling sequence was based on the AIX | |
266 | calling sequence without function descriptors. To compile code for that older | |
267 | calling sequence, either configure the compiler for powerpc-*-eabiaix or use | |
268 | the -mcall-aix switch when compiling and linking. | |
956d6950 JL |
269 | \f |
270 | Noteworthy changes in GCC version 2.7.0 | |
271 | --------------------------------------- | |
fc5a8790 JL |
272 | |
273 | GCC now works better on systems that use ".obj" and ".exe" instead of | |
274 | ".o" and no extension. This involved changes to the driver program, | |
275 | gcc.c, to convert ".o" names to ".obj" and to GCC's Makefile to use | |
276 | ".obj" and ".exe" in filenames that are not targets. In order to | |
277 | build GCC on such systems, you may need versions of GNU make and/or | |
278 | compatible shells. At this point, this support is preliminary. | |
279 | ||
280 | Object file extensions of ".obj" and executable file extensions of | |
281 | ".exe" are allowed when using appropriate version of GNU Make. | |
282 | ||
283 | Numerous enhancements were made to the __attribute__ facility including | |
284 | more attributes and more places that support it. We now support the | |
285 | "packed", "nocommon", "noreturn", "volatile", "const", "unused", | |
286 | "transparent_union", "constructor", "destructor", "mode", "section", | |
287 | "align", "format", "weak", and "alias" attributes. Each of these | |
288 | names may also be specified with added underscores, e.g., "__packed__". | |
289 | __attribute__ may now be applied to parameter definitions, function | |
290 | definitions, and structure, enum, and union definitions. | |
291 | ||
292 | GCC now supports returning more structures in registers, as specified by | |
293 | many calling sequences (ABIs), such as on the HP PA RISC. | |
294 | ||
295 | A new option '-fpack-struct' was added to automatically pack all structure | |
296 | members together without holes. | |
297 | ||
298 | There is a new library (cpplib) and program (cppmain) that at some | |
299 | point will replace cpp (aka cccp). To use cppmain as cpp now, pass | |
300 | the option CCCP=cppmain to make. The library is already used by the | |
301 | fix-header program, which should speed up the fixproto script. | |
302 | ||
303 | New options for supported targets: | |
304 | ||
305 | GNU on many targets. | |
306 | NetBSD on MIPS, m68k, VAX, and x86. | |
307 | LynxOS on x86, m68k, Sparc, and RS/6000. | |
308 | VxWorks on many targets. | |
309 | ||
310 | Windows/NT on x86 architecture. Initial support for Windows/NT on Alpha | |
311 | (not fully working). | |
312 | ||
313 | Many embedded targets, specifically UDI on a29k, aout, coff, elf, | |
314 | and vsta "operating systems" on m68k, m88k, mips, sparc, and x86. | |
315 | ||
316 | Additional support for x86 (i386, i486, and Pentium): | |
317 | ||
318 | Work with old and new linkers for Linux-based GNU systems, | |
319 | supporting both a.out and ELF. | |
320 | FreeBSD on x86. | |
321 | Stdcall convention. | |
322 | -malign-double, -mregparm=, -malign-loops= and -malign-jumps= switches. | |
323 | On ISC systems, support -Xp like -posix. | |
324 | ||
325 | Additions for RS/6000: | |
326 | ||
327 | Instruction scheduling information for PowerPC 403. | |
328 | AIX 4.1 on PowerPC. | |
329 | -mstring and -mno-string. | |
330 | -msoft-float and floating-point emulation included. | |
331 | Preliminary support for PowerPC System V.4 with or without the GNU as. | |
332 | Preliminary support for EABI. | |
333 | Preliminary support for 64-bit systems. | |
334 | Both big and little endian systems. | |
335 | ||
336 | New features for MIPS-based systems: | |
337 | ||
338 | r4650. | |
339 | mips4 and R8000. | |
340 | Irix 6.0. | |
341 | 64-bit ABI. | |
342 | Allow dollar signs in labels on SGI/Irix 5.x. | |
343 | ||
344 | New support for HP PA RISC: | |
345 | ||
346 | Generation of PIC (requires binutils-2.5.2.u6 or later). | |
347 | HP-UX version 9 on HP PA RISC (dynamically links even with -g). | |
348 | Processor variants for HP PA RISC: 700, 7100, and 7100LC. | |
349 | Automatic generation of long calls when needed. | |
350 | -mfast-indirect-calls for kernels and static binaries. | |
351 | ||
352 | The called routine now copies arguments passed by invisible reference, | |
353 | as required by the calling standard. | |
354 | ||
355 | Other new miscellaneous target-specific support: | |
356 | ||
357 | -mno-multm on a29k. | |
358 | -mold-align for i960. | |
359 | Configuration for "semi-hosted" ARM. | |
360 | -momit-leaf-frame-pointer for M88k. | |
361 | SH3 variant of Hitachi Super-H and support both big and little endian. | |
362 | ||
363 | Changes to Objective-C: | |
364 | ||
365 | Bare-bones implementation of NXConstantString has been added, | |
366 | which is invoked by the @"string" directive. | |
367 | ||
368 | Class * has been changed to Class to conform to the NextSTEP and | |
369 | OpenStep runtime. | |
370 | ||
371 | Enhancements to make dynamic loading easier. | |
372 | ||
373 | The module version number has been updated to Version 7, thus existing | |
374 | code will need to be recompiled to use the current run-time library. | |
375 | ||
376 | GCC now supports the ISO Normative Addendum 1 to the C Standard. | |
377 | As a result: | |
378 | ||
379 | The header <iso646.h> defines macros for C programs written | |
380 | in national variants of ISO 646. | |
381 | ||
382 | The following digraph tokens are supported: | |
383 | <: :> <% %> %: %:%: | |
384 | These behave like the following, respectively: | |
385 | [ ] { } # ## | |
386 | ||
387 | Digraph tokens are supported unless you specify the `-traditional' | |
388 | option; you do not need to specify `-ansi' or `-trigraphs'. Except | |
389 | for contrived and unlikely examples involving preprocessor | |
390 | stringizing, digraph interpretation doesn't change the meaning of | |
391 | programs; this is unlike trigraph interpretation, which changes the | |
392 | meanings of relatively common strings. | |
393 | ||
394 | The macro __STDC_VERSION__ has the value 199409L. | |
395 | ||
396 | As usual, for full conformance to the standard, you also need a | |
397 | C library that conforms. | |
398 | ||
399 | The following lists changes that have been made to g++. If some | |
400 | features mentioned below sound unfamiliar, you will probably want to | |
401 | look at the recently-released public review copy of the C++ Working | |
402 | Paper. For PostScript and PDF (Adobe Acrobat) versions, see the | |
403 | archive at ftp://research.att.com/dist/stdc++/WP. For HTML and ASCII | |
404 | versions, see ftp://ftp.cygnus.com/pub/g++. On the web, see | |
405 | http://www.cygnus.com/~mrs/wp-draft. | |
406 | ||
407 | The scope of variables declared in the for-init-statement has been changed | |
408 | to conform to http://www.cygnus.com/~mrs/wp-draft/stmt.html#stmt.for; as a | |
409 | result, packages such as groff 1.09 will not compile unless you specify the | |
410 | -fno-for-scope flag. PLEASE DO NOT REPORT THIS AS A BUG; this is a change | |
411 | mandated by the C++ standardization committee. | |
412 | ||
413 | Binary incompatibilities: | |
414 | ||
415 | The builtin 'bool' type is now the size of a machine word on RISC targets, | |
416 | for code efficiency; it remains one byte long on CISC targets. | |
417 | ||
418 | Code that does not use #pragma interface/implementation will most | |
419 | likely shrink dramatically, as g++ now only emits the vtable for a | |
420 | class in the translation unit where its first non-inline, non-abstract | |
421 | virtual function is defined. | |
422 | ||
423 | Classes that do not define the copy constructor will sometimes be | |
424 | passed and returned in registers. This may illuminate latent bugs in | |
425 | your code. | |
426 | ||
427 | Support for automatic template instantiation has *NOT* been added, due | |
428 | to a disagreement over design philosophies. | |
429 | ||
430 | Support for exception handling has been improved; more targets are now | |
431 | supported, and throws will use the RTTI mechanism to match against the | |
432 | catch parameter type. Optimization is NOT SUPPORTED with | |
433 | -fhandle-exceptions; no need to report this as a bug. | |
434 | ||
435 | Support for Run-Time Type Identification has been added with -frtti. | |
436 | This support is still in alpha; one major restriction is that any file | |
437 | compiled with -frtti must include <typeinfo.h>. | |
438 | ||
439 | Preliminary support for namespaces has been added. This support is far | |
440 | from complete, and probably not useful. | |
441 | ||
442 | Synthesis of compiler-generated constructors, destructors and | |
443 | assignment operators is now deferred until the functions are used. | |
444 | ||
445 | The parsing of expressions such as `a ? b : c = 1' has changed from | |
446 | `(a ? b : c) = 1' to `a : b ? (c = 1)'. | |
447 | ||
448 | The code generated for testing conditions, especially those using || | |
449 | and &&, is now more efficient. | |
450 | ||
451 | The operator keywords and, and_eq, bitand, bitor, compl, not, not_eq, | |
452 | or, or_eq, xor and xor_eq are now supported. Use -ansi or | |
453 | -foperator-names to enable them. | |
454 | ||
455 | The 'explicit' keyword is now supported. 'explicit' is used to mark | |
456 | constructors and type conversion operators that should not be used | |
457 | implicitly. | |
458 | ||
459 | g++ now accepts the typename keyword, though it currently has no | |
460 | semantics; it can be a no-op in the current template implementation. | |
461 | You may want to start using it in your code, however, since the | |
462 | pending rewrite of the template implementation to compile STL properly | |
463 | (perhaps for 2.8.0, perhaps not) will require you to use it as | |
464 | indicated by the current draft. | |
465 | ||
466 | Handling of user-defined type conversion has been overhauled so that | |
467 | type conversion operators are now found and used properly in | |
468 | expressions and function calls. | |
469 | ||
470 | -fno-strict-prototype now only applies to function declarations with | |
471 | "C" linkage. | |
472 | ||
473 | g++ now warns about 'if (x=0)' with -Wparentheses or -Wall. | |
474 | ||
475 | #pragma weak and #pragma pack are supported on System V R4 targets, as | |
476 | are various other target-specific #pragmas supported by gcc. | |
477 | ||
478 | new and delete of const types is now allowed (with no additional | |
479 | semantics). | |
480 | ||
481 | Explicit instantiation of template methods is now supported. Also, | |
482 | 'inline template class foo<int>;' can be used to emit only the vtable | |
483 | for a template class. | |
484 | ||
485 | With -fcheck-new, g++ will check the return value of all calls to | |
486 | operator new, and not attempt to modify a returned null pointer. | |
487 | ||
488 | The template instantiation code now handles more conversions when | |
489 | passing to a parameter that does not depend on template arguments. | |
490 | This means that code like 'string s; cout << s;' now works. | |
491 | ||
492 | Invalid jumps in a switch statement past declarations that require | |
493 | initializations are now caught. | |
494 | ||
495 | Functions declared 'extern inline' now have the same linkage semantics | |
496 | as inline member functions. On supported targets, where previously | |
497 | these functions (and vtables, and template instantiations) would have | |
498 | been defined statically, they will now be defined as weak symbols so | |
499 | that only one out-of-line definition is used. | |
500 | ||
501 | collect2 now demangles linker output, and c++filt has become part of | |
502 | the gcc distribution. | |
503 | \f | |
504 | Noteworthy changes in GCC version 2.6.3: | |
505 | ||
506 | A few more bugs have been fixed. | |
507 | ||
508 | Noteworthy changes in GCC version 2.6.2: | |
509 | ||
510 | A few bugs have been fixed. | |
511 | ||
512 | Names of attributes can now be preceded and followed by double underscores. | |
513 | ||
514 | Noteworthy changes in GCC version 2.6.1: | |
515 | ||
516 | Numerous (mostly minor) bugs have been fixed. | |
517 | ||
518 | The following new configurations are supported: | |
519 | ||
520 | GNU on x86 (instead of treating it like MACH) | |
521 | NetBSD on Sparc and Motorola 68k | |
522 | AIX 4.1 on RS/6000 and PowerPC systems | |
956d6950 | 523 | Sequent DYNIX/ptx 1.x and 2.x. |
fc5a8790 JL |
524 | Both COFF and ELF configurations on AViiON without using /bin/gcc |
525 | Windows/NT on x86 architecture; preliminary | |
526 | AT&T DSP1610 digital signal processor chips | |
527 | i960 systems on bare boards using COFF | |
528 | PDP11; target only and not extensively tested | |
529 | ||
530 | The -pg option is now supported for Alpha under OSF/1 V3.0 or later. | |
531 | ||
532 | Files with an extension of ".c++" are treated as C++ code. | |
533 | ||
534 | The -Xlinker and -Wl arguments are now passed to the linker in the | |
535 | position they were specified on the command line. This makes it | |
536 | possible, for example, to pass flags to the linker about specific | |
537 | object files. | |
538 | ||
539 | The use of positional arguments to the configure script is no longer | |
540 | recommended. Use --target= to specify the target; see the GCC manual. | |
541 | ||
542 | The 386 now supports two new switches: -mreg-alloc=<string> changes | |
543 | the default register allocation order used by the compiler, and | |
544 | -mno-wide-multiply disables the use of the mul/imul instructions that | |
545 | produce 64 bit results in EAX:EDX from 32 bit operands to do long long | |
546 | multiplies and 32-bit division by constants. | |
547 | \f | |
548 | Noteworthy changes in GCC version 2.6.0: | |
549 | ||
550 | Numerous bugs have been fixed, in the C and C++ front-ends, as | |
551 | well as in the common compiler code. | |
552 | ||
553 | This release includes the C, Objective-C, and C++ compilers. However, | |
554 | we have moved the files for the C++ compiler (G++) files to a | |
555 | subdirectory, cp. Subsequent releases of GCC will split these files | |
556 | to a separate TAR file. | |
557 | ||
558 | The G++ team has been tracking the development of the ANSI standard for C++. | |
559 | Here are some new features added from the latest working paper: | |
560 | ||
561 | * built-in boolean type 'bool', with constants 'true' and 'false'. | |
562 | * array new and delete (operator new [] and delete []). | |
563 | * WP-conforming lifetime of temporaries. | |
564 | * explicit instantiation of templates (template class A<int>;), | |
565 | along with an option (-fno-implicit-templates) to disable emission | |
566 | of implicitly instantiated templates, obsoletes -fexternal-templates. | |
567 | * static member constants (static const int foo = 4; within the | |
568 | class declaration). | |
569 | ||
570 | Many error messages have been improved to tell the user more about the | |
571 | problem. Conformance checking with -pedantic-errors has been | |
572 | improved. G++ now compiles Fresco. | |
573 | ||
574 | There is now an experimental implementation of virtual functions using | |
575 | thunks instead of Cfront-style vtables, enabled with -fvtable-thunks. | |
576 | This option also enables a heuristic which causes the compiler to only | |
577 | emit the vtable in the translation unit where its first non-inline | |
578 | virtual function is defined; using this option and | |
579 | -fno-implicit-templates, users should be able to avoid #pragma | |
580 | interface/implementation altogether. | |
581 | ||
582 | Signatures have been added as a GNU C++ extension. Using the option | |
583 | -fhandle-signatures, users are able to turn on recognition of | |
584 | signatures. A short introduction on signatures is in the section | |
585 | `Extension to the C++ Language' in the manual. | |
586 | ||
587 | The `g++' program is now a C program, rather than a shell script. | |
588 | ||
589 | Lots and lots and lots of bugs fixes, in nested types, access control, | |
590 | pointers to member functions, the parser, templates, overload | |
591 | resolution, etc, etc. | |
592 | ||
593 | There have been two major enhancements to the Objective-C compiler: | |
594 | ||
595 | 1) Added portability. It now runs on Alpha, and some problems with | |
596 | message forwarding have been addressed on other platforms. | |
597 | ||
598 | 2) Selectors have been redefined to be pointers to structs like: | |
599 | { void *sel_id, char *sel_types }, where the sel_id is the unique | |
600 | identifier, the selector itself is no longer unique. | |
601 | ||
602 | Programmers should use the new function sel_eq to test selector | |
603 | equivalence. | |
604 | ||
605 | The following major changes have been made to the base compiler and | |
606 | machine-specific files. | |
607 | ||
608 | - The MIL-STD-1750A is a new port, but still preliminary. | |
609 | ||
610 | - The h8/300h is now supported; both the h8/300 and h8/300h ports come | |
611 | with 32 bit IEEE 754 software floating point support. | |
612 | ||
613 | - The 64-bit Sparc (v9) and 64-bit MIPS chips are supported. | |
614 | ||
615 | - NetBSD is supported on m68k, Intel x86, and pc523 systems and FreeBSD | |
616 | on x86. | |
617 | ||
618 | - COFF is supported on x86, m68k, and Sparc systems running LynxOS. | |
619 | ||
620 | - 68K systems from Bull and Concurrent are supported and System V | |
621 | Release 4 is supported on the Atari. | |
622 | ||
623 | - GCC supports GAS on the Motorola 3300 (sysV68) and debugging | |
624 | (assuming GAS) on the Plexus 68K system. (However, GAS does not yet | |
625 | work on those systems). | |
626 | ||
627 | - System V Release 4 is supported on MIPS (Tandem). | |
628 | ||
629 | - For DG/UX, an ELF configuration is now supported, and both the ELF | |
630 | and BCS configurations support ELF and COFF object file formats. | |
631 | ||
632 | - OSF/1 V2.0 is supported on Alpha. | |
633 | ||
634 | - Function profiling is also supported on Alpha. | |
635 | ||
636 | - GAS and GDB is supported for Irix 5 (MIPS). | |
637 | ||
638 | - "common mode" (code that will run on both POWER and PowerPC | |
639 | architectures) is now supported for the RS/6000 family; the | |
640 | compiler knows about more PPC chips. | |
641 | ||
642 | - Both NeXTStep 2.1 and 3 are supported on 68k-based architectures. | |
643 | ||
644 | - On the AMD 29k, the -msoft-float is now supported, as well as | |
645 | -mno-sum-in-toc for RS/6000, -mapp-regs and -mflat for Sparc, and | |
646 | -membedded-pic for MIPS. | |
647 | ||
648 | - GCC can now convert division by integer constants into the equivalent | |
649 | multiplication and shift operations when that is faster than the | |
650 | division. | |
651 | ||
652 | - Two new warning options, -Wbad-function-cast and | |
653 | -Wmissing-declarations have been added. | |
654 | ||
655 | - Configurations may now add machine-specific __attribute__ options on | |
656 | type; many machines support the `section' attribute. | |
657 | ||
658 | - The -ffast-math flag permits some optimization that violate strict | |
659 | IEEE rules, such as converting X * 0.0 to 0.0. | |
660 | \f | |
661 | Noteworthy changes in GCC version 2.5.8: | |
662 | ||
663 | This release only fixes a few serious bugs. These include fixes for a | |
664 | bug that prevented most programs from working on the RS/6000, a bug | |
665 | that caused invalid assembler code for programs with a `switch' | |
666 | statement on the NS32K, a G++ problem that caused undefined names in | |
667 | some configurations, and several less serious problems, some of which | |
668 | can affect most configuration. | |
669 | ||
670 | Noteworthy change in GCC version 2.5.7: | |
671 | ||
672 | This release only fixes a few bugs, one of which was causing bootstrap | |
673 | compare errors on some systems. | |
674 | ||
675 | Noteworthy change in GCC version 2.5.6: | |
676 | ||
677 | A few backend bugs have been fixed, some of which only occur on one | |
678 | machine. | |
679 | ||
680 | The C++ compiler in 2.5.6 includes: | |
681 | ||
682 | * fixes for some common crashes | |
683 | * correct handling of nested types that are referenced as `foo::bar' | |
684 | * spurious warnings about friends being declared static and never | |
685 | defined should no longer appear | |
686 | * enums that are local to a method in a class, or a class that's | |
687 | local to a function, are now handled correctly. For example: | |
688 | class foo { void bar () { enum { x, y } E; x; } }; | |
689 | void bar () { class foo { enum { x, y } E; E baz; }; } | |
690 | ||
691 | Noteworthy change in GCC version 2.5.5: | |
692 | ||
693 | A large number of C++ bugs have been fixed. | |
694 | ||
695 | The fixproto script adds prototypes conditionally on __cplusplus. | |
696 | ||
697 | Noteworthy change in GCC version 2.5.4: | |
698 | ||
699 | A bug fix in passing of structure arguments for the HP-PA architecture | |
700 | makes code compiled with GCC 2.5.4 incompatible with code compiled | |
701 | with earlier versions (if it passes struct arguments of 33 to 64 bits, | |
702 | interspersed with other types of arguments). | |
703 | ||
704 | Noteworthy change in gcc version 2.5.3: | |
705 | ||
706 | The method of "mangling" C++ function names has been changed. So you | |
707 | must recompile all C++ programs completely when you start using GCC | |
708 | 2.5. Also, GCC 2.5 requires libg++ version 2.5. Earlier libg++ | |
709 | versions won't work with GCC 2.5. (This is generally true--GCC | |
710 | version M.N requires libg++ version M.N.) | |
711 | \f | |
712 | Noteworthy GCC changes in version 2.5: | |
713 | ||
714 | * There is now support for the IBM 370 architecture as a target. | |
715 | Currently the only operating system supported is MVS; GCC does not run | |
716 | on MVS, so you must produce .s files using GCC as a cross compiler, | |
717 | then transfer them to MVS to assemble them. This port is not reliable | |
718 | yet. | |
719 | ||
720 | * The Power PC is now supported. | |
721 | ||
722 | * The i860-based Paragon machine is now supported. | |
723 | ||
724 | * The Hitachi 3050 (an HP-PA machine) is now supported. | |
725 | ||
726 | * The variable __GNUC_MINOR__ holds the minor version number of GCC, as | |
727 | an integer. For version 2.5.X, the value is 5. | |
728 | ||
729 | * In C, initializers for static and global variables are now processed | |
730 | an element at a time, so that they don't need a lot of storage. | |
731 | ||
732 | * The C syntax for specifying which structure field comes next in an | |
733 | initializer is now `.FIELDNAME='. The corresponding syntax for | |
734 | array initializers is now `[INDEX]='. For example, | |
735 | ||
736 | char whitespace[256] | |
737 | = { [' '] = 1, ['\t'] = 1, ['\n'] = 1 }; | |
738 | ||
739 | This was changed to accord with the syntax proposed by the Numerical | |
740 | C Extensions Group (NCEG). | |
741 | ||
742 | * Complex numbers are now supported in C. Use the keyword __complex__ | |
743 | to declare complex data types. See the manual for details. | |
744 | ||
745 | * GCC now supports `long double' meaningfully on the Sparc (128-bit | |
746 | floating point) and on the 386 (96-bit floating point). The Sparc | |
747 | support is enabled on on Solaris 2.x because earlier system versions | |
748 | (SunOS 4) have bugs in the emulation. | |
749 | ||
750 | * All targets now have assertions for cpu, machine and system. So you | |
751 | can now use assertions to distinguish among all supported targets. | |
752 | ||
753 | * Nested functions in C may now be inline. Just declare them inline | |
754 | in the usual way. | |
755 | ||
756 | * Packed structure members are now supported fully; it should be possible | |
757 | to access them on any supported target, no matter how little alignment | |
758 | they have. | |
759 | ||
760 | * To declare that a function does not return, you must now write | |
761 | something like this (works only in 2.5): | |
762 | ||
763 | void fatal () __attribute__ ((noreturn)); | |
764 | ||
765 | or like this (works in older versions too): | |
766 | ||
767 | typedef void voidfn (); | |
768 | ||
769 | volatile voidfn fatal; | |
770 | ||
771 | It used to be possible to do so by writing this: | |
772 | ||
773 | volatile void fatal (); | |
774 | ||
775 | but it turns out that ANSI C requires that to mean something | |
776 | else (which is useless). | |
777 | ||
778 | Likewise, to declare that a function is side-effect-free | |
779 | so that calls may be deleted or combined, write | |
780 | something like this (works only in 2.5): | |
781 | ||
782 | int computation () __attribute__ ((const)); | |
783 | ||
784 | or like this (works in older versions too): | |
785 | ||
786 | typedef int intfn (); | |
787 | ||
788 | const intfn computation; | |
789 | ||
790 | * The new option -iwithprefixbefore specifies a directory to add to | |
791 | the search path for include files in the same position where -I would | |
792 | put it, but uses the specified prefix just like -iwithprefix. | |
793 | ||
794 | * Basic block profiling has been enhanced to record the function the | |
795 | basic block comes from, and if the module was compiled for debugging, | |
796 | the line number and filename. A default version of the basic block | |
797 | support module has been added to libgcc2 that appends the basic block | |
798 | information to a text file 'bb.out'. Machine descriptions can now | |
799 | override the basic block support module in the target macro file. | |
800 | ||
801 | New features in g++: | |
802 | ||
803 | * The new flag `-fansi-overloading' for C++. Use a newly implemented | |
804 | scheme of argument matching for C++. It makes g++ more accurately | |
805 | obey the rules set down in Chapter 13 of the Annotated C++ Reference | |
806 | Manual (the ARM). This option will be turned on by default in a | |
807 | future release. | |
808 | ||
809 | * The -finline-debug flag is now gone (it was never really used by the | |
810 | compiler). | |
811 | ||
812 | * Recognizing the syntax for pointers to members, e.g., "foo::*bar", has been | |
813 | dramatically improved. You should not get any syntax errors or incorrect | |
814 | runtime results while using pointers to members correctly; if you do, it's | |
815 | a definite bug. | |
816 | ||
817 | * Forward declaration of an enum is now flagged as an error. | |
818 | ||
819 | * Class-local typedefs are now working properly. | |
820 | ||
821 | * Nested class support has been significantly improved. The compiler | |
822 | will now (in theory) support up to 240 nested classes before hitting | |
823 | other system limits (like memory size). | |
824 | ||
825 | * There is a new C version of the `g++' driver, to replace the old | |
826 | shell script. This should significantly improve the performance of | |
827 | executing g++ on a system where a user's PATH environment variable | |
828 | references many NFS-mounted filesystems. This driver also works | |
829 | under MS-DOS and OS/2. | |
830 | ||
831 | * The ANSI committee working on the C++ standard has adopted a new | |
832 | keyword `mutable'. This will allow you to make a specific member be | |
833 | modifiable in an otherwise const class. | |
834 | \f | |
835 | Noteworthy GCC changes in version 2.4.4: | |
836 | ||
837 | A crash building g++ on various hosts (including m68k) has been | |
838 | fixed. Also the g++ compiler no longer reports incorrect | |
839 | ambiguities in some situations where they do not exist, and | |
840 | const template member functions are now being found properly. | |
841 | ||
842 | Noteworthy GCC changes in version 2.4: | |
843 | ||
844 | * On each target, the default is now to return short structures | |
845 | compatibly with the "usual" compiler on that target. | |
846 | ||
847 | For most targets, this means the default is to return all structures | |
848 | in memory, like long structures, in whatever way is used on that | |
849 | target. Use -freg-struct-return to enable returning short structures | |
850 | (and unions) in registers. | |
851 | ||
852 | This change means that newly compiled binaries are incompatible with | |
853 | binaries compiled with previous versions of GCC. | |
854 | ||
855 | On some targets, GCC is itself the usual compiler. On these targets, | |
856 | the default way to return short structures is still in registers. | |
857 | Use -fpcc-struct-return to tell GCC to return them in memory. | |
858 | ||
859 | * There is now a floating point emulator which can imitate the way all | |
860 | supported target machines do floating point arithmetic. | |
861 | ||
862 | This makes it possible to have cross compilation to and from the VAX, | |
863 | and between machines of different endianness. However, this works | |
864 | only when the target machine description is updated to use the new | |
865 | facilities, and not all have been updated. | |
866 | ||
867 | This also makes possible support for longer floating point types. | |
868 | GCC 2.4 supports extended format on the 68K if you use `long double', | |
869 | for targets that have a 68881. (When we have run time library | |
870 | routines for extended floating point, then `long double' will use | |
871 | extended format on all 68K targets.) | |
872 | ||
873 | We expect to support extended floating point on the i386 and Sparc in | |
874 | future versions. | |
875 | ||
876 | * Building GCC now automatically fixes the system's header files. | |
877 | This should require no attention. | |
878 | ||
879 | * GCC now installs an unsigned data type as size_t when it fixes the | |
880 | header files (on all but a handful of old target machines). | |
881 | Therefore, the bug that size_t failed to be unsigned is fixed. | |
882 | ||
883 | * Building and installation are now completely separate. | |
884 | All new files are constructed during the build process; | |
885 | installation just copies them. | |
886 | ||
887 | * New targets supported: Clipper, Hitachi SH, Hitachi 8300, and Sparc | |
888 | Lite. | |
889 | ||
890 | * A totally new and much better Objective C run time system is included. | |
891 | ||
892 | * Objective C supports many new features. Alas, I can't describe them | |
893 | since I don't use that language; however, they are the same ones | |
894 | supported in recent versions of the NeXT operating system. | |
895 | ||
896 | * The builtin functions __builtin_apply_args, __builtin_apply and | |
897 | __builtin_return let you record the arguments and returned | |
898 | value of a function without knowing their number or type. | |
899 | ||
900 | * The builtin string variables __FUNCTION__ and __PRETTY_FUNCTION__ | |
901 | give the name of the function in the source, and a pretty-printed | |
902 | version of the name. The two are the same in C, but differ in C++. | |
903 | ||
904 | * Casts to union types do not yield lvalues. | |
905 | ||
906 | * ## before an empty rest argument discards the preceding sequence | |
907 | of non-whitespace characters from the macro definition. | |
908 | (This feature is subject to change.) | |
909 | ||
910 | ||
911 | New features specific to C++: | |
912 | ||
913 | * The manual contains a new section ``Common Misunderstandings with | |
914 | GNU C++'' that C++ users should read. | |
915 | ||
916 | * #pragma interface and #pragma implementation let you use the same | |
917 | C++ source file for both interface and implementation. | |
918 | However, this mechanism is still in transition. | |
919 | ||
920 | * Named returned values let you avoid an extra constructor call | |
921 | when a function result has a class type. | |
922 | ||
923 | * The C++ operators <? and >? yield min and max, respectively. | |
924 | ||
925 | * C++ gotos can exit a block safely even if the block has | |
926 | aggregates that require destructors. | |
927 | ||
928 | * gcc defines the macro __GNUG__ when compiling C++ programs. | |
929 | ||
930 | * GNU C++ now correctly distinguishes between the prefix and postfix | |
931 | forms of overloaded operator ++ and --. To avoid breaking old | |
932 | code, if a class defines only the prefix form, the compiler | |
933 | accepts either ++obj or obj++, unless -pedantic is used. | |
934 | ||
935 | * If you are using version 2.3 of libg++, you need to rebuild it with | |
936 | `make CC=gcc' to avoid mismatches in the definition of `size_t'. | |
937 | ||
938 | Newly documented compiler options: | |
939 | ||
940 | -fnostartfiles | |
941 | Omit the standard system startup files when linking. | |
942 | ||
943 | -fvolatile-global | |
944 | Consider memory references to extern and global data items to | |
945 | be volatile. | |
946 | ||
947 | -idirafter DIR | |
948 | Add DIR to the second include path. | |
949 | ||
950 | -iprefix PREFIX | |
951 | Specify PREFIX for later -iwithprefix options. | |
952 | ||
953 | -iwithprefix DIR | |
954 | Add PREFIX/DIR to the second include path. | |
955 | ||
956 | -mv8 | |
957 | Emit Sparc v8 code (with integer multiply and divide). | |
958 | -msparclite | |
959 | Emit Sparclite code (roughly v7.5). | |
960 | ||
961 | -print-libgcc-file-name | |
962 | Search for the libgcc.a file, print its absolute file name, and exit. | |
963 | ||
964 | -Woverloaded-virtual | |
965 | Warn when a derived class function declaration may be an error | |
966 | in defining a C++ virtual function. | |
967 | ||
968 | -Wtemplate-debugging | |
969 | When using templates in a C++ program, warn if debugging is | |
970 | not yet fully available. | |
971 | ||
972 | +eN | |
973 | Control how C++ virtual function definitions are used | |
974 | (like cfront 1.x). | |
975 |