![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to gcc.texi CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Apple XNU] / GNUtools / cc|
Return to gcc.texi CVS log | Up to [Apple XNU] / GNUtools / cc |
1.1 root 1: \input texinfo @c -*-texinfo-*-
2: @c %**start of header
3: @setfilename gcc.info
4: @c @setfilename usegcc.info
5: @c @setfilename portgcc.info
6: @c To produce the full manual, use the "gcc.info" setfilename, and
7: @c make sure the following do NOT begin with '@c' (and the @clear lines DO)
8: @set INTERNALS
9: @set USING
10: @c To produce a user-only manual, use the "usegcc.info" setfilename, and
11: @c make sure the following does NOT begin with '@c':
12: @c @clear INTERNALS
13: @c To produce a porter-only manual, use the "portgcc.info" setfilename,
14: @c and make sure the following does NOT begin with '@c':
15: @c @clear USING
16:
17: @c i have commented out the smallbook command below, and reformatted
18: @c this manual in the regular book size for distribution. in addition,
19: @c i commented out the commands that shift the text to one or the other
20: @c side of the page for smallbook printing (which makes it easier for
21: @c the photocopying people to handle...). -mew, 15june93
22: @c @smallbook
23:
24: @c i also commented out the finalout command, so if there *are* any
25: @c overfulls, you'll (hopefully) see the rectangle in the right hand
26: @c margin. -mew 15june93
27: @c @finalout
28:
29: @c NOTE: checks/things to do:
30: @c
31: @c -have bob do a search in all seven files for "mew" (ideally --mew,
32: @c but i may have forgotten the occasional "--"..).
33: @c -item/itemx, text after all (sub/sub)section titles, etc..
34: @c -consider putting the lists of options on pp 17--> etc in columns or
35: @c somesuch.
36: @c -spellcheck
37: @c -continuity of phrasing; ie, bit-field vs bitfield in rtl.texi
38: @c -overfulls. do a search for "mew" in the files, and you will see
39: @c overfulls that i noted but could not deal with.
40: @c -have to add text: beginning of chapter 8
41:
42: @c
43: @c anything else? --mew 10feb93
44:
45:
46:
47: @ifset INTERNALS
48: @ifset USING
49: @settitle Using and Porting GNU CC
50: @end ifset
51: @end ifset
52: @c seems reasonable to assume at least one of INTERNALS or USING is set...
53: @ifclear INTERNALS
54: @settitle Using GNU CC
55: @end ifclear
56: @ifclear USING
57: @settitle Porting GNU CC
58: @end ifclear
59:
60: @syncodeindex fn cp
61: @syncodeindex vr cp
62: @c %**end of header
63:
64: @c Use with @@smallbook.
65:
66: @c Cause even numbered pages to be printed on the left hand side of
67: @c the page and odd numbered pages to be printed on the right hand
68: @c side of the page. Using this, you can print on both sides of a
69: @c sheet of paper and have the text on the same part of the sheet.
70:
71: @c The text on right hand pages is pushed towards the right hand
72: @c margin and the text on left hand pages is pushed toward the left
73: @c hand margin.
74: @c (To provide the reverse effect, set bindingoffset to -0.75in.)
75:
76: @c @tex
77: @c \global\bindingoffset=0.75in
78: @c \global\normaloffset =0.75in
79: @c @end tex
80:
81: @ifinfo
82: @ifset INTERNALS
83: @ifset USING
84: This file documents the use and the internals of the GNU compiler.
85: @end ifset
86: @end ifset
87: @ifclear USING
88: This file documents the internals of the GNU compiler.
89: @end ifclear
90: @ifclear INTERNALS
91: This file documents the use of the GNU compiler.
92: @end ifclear
93:
94: Published by the Free Software Foundation
95: 675 Massachusetts Avenue
96: Cambridge, MA 02139 USA
97:
98: Copyright (C) 1988, 1989, 1992, 1993 Free Software Foundation, Inc.
99:
100: Permission is granted to make and distribute verbatim copies of
101: this manual provided the copyright notice and this permission notice
102: are preserved on all copies.
103:
104: @ignore
105: Permission is granted to process this file through Tex and print the
106: results, provided the printed document carries copying permission
107: notice identical to this one except for the removal of this paragraph
108: (this paragraph not being relevant to the printed manual).
109:
110: @end ignore
111: Permission is granted to copy and distribute modified versions of this
112: manual under the conditions for verbatim copying, provided also that the
113: sections entitled ``GNU General Public License'' and ``Protect Your
114: Freedom---Fight `Look And Feel'@w{}'' are included exactly as in the
115: original, and provided that the entire resulting derived work is
116: distributed under the terms of a permission notice identical to this
117: one.
118:
119: Permission is granted to copy and distribute translations of this manual
120: into another language, under the above conditions for modified versions,
121: except that the sections entitled ``GNU General Public License'' and
122: ``Protect Your Freedom---Fight `Look And Feel'@w{}'', and this permission
123: notice, may be included in translations approved by the Free Software
124: Foundation instead of in the original English.
125: @end ifinfo
126:
127: @setchapternewpage odd
128:
129: @titlepage
130: @ifset INTERNALS
131: @ifset USING
132: @center @titlefont{Using and Porting GNU CC}
133:
134: @end ifset
135: @end ifset
136: @ifclear INTERNALS
137: @title Using GNU CC
138: @end ifclear
139: @ifclear USING
140: @title Porting GNU CC
141: @end ifclear
142: @sp 2
143: @center Richard M. Stallman
144: @sp 3
145: @center last updated 14 October 1993
146: @sp 1
147: @c The version number appears twice more in this file.
148:
149: @center for version 2.5
150: @c @center (preliminary draft, which will change)
151: @page
152: @vskip 0pt plus 1filll
153: Copyright @copyright{} 1988, 1989, 1992, 1993 Free Software Foundation, Inc.
154: @sp 2
155: For GCC Version 2.5,@*
156: Printed October, 1993.@*
157:
158: ISBN 1-882114-35-3
159: @sp 1
160: Published by the Free Software Foundation @*
161: 675 Massachusetts Avenue @*
162: Cambridge, MA 02139 USA
163: @sp 1
164: Permission is granted to make and distribute verbatim copies of
165: this manual provided the copyright notice and this permission notice
166: are preserved on all copies.
167:
168: Permission is granted to copy and distribute modified versions of this
169: manual under the conditions for verbatim copying, provided also that the
170: sections entitled ``GNU General Public License'' and ``Protect Your
171: Freedom---Fight `Look And Feel'@w{}'' are included exactly as in the
172: original, and provided that the entire resulting derived work is
173: distributed under the terms of a permission notice identical to this
174: one.
175:
176: Permission is granted to copy and distribute translations of this manual
177: into another language, under the above conditions for modified versions,
178: except that the sections entitled ``GNU General Public License'' and
179: ``Protect Your Freedom---Fight `Look And Feel'@w{}'', and this
180: permission notice, may be included in translations approved by the Free
181: Software Foundation instead of in the original English.
182: @end titlepage
183: @page
184:
185: @ifinfo
186:
187: @node Top, Copying,, (DIR)
188: @top Introduction
189: @cindex introduction
190:
191: @ifset INTERNALS
192: @ifset USING
193: This manual documents how to run, install and port the GNU
194: compiler, as well as its new features and incompatibilities, and how to
195: report bugs. It corresponds to GNU CC version 2.5.
196: @end ifset
197: @end ifset
198:
199: @ifclear INTERNALS
200: This manual documents how to run and install the GNU compiler,
201: as well as its new features and incompatibilities, and how to report
202: bugs. It corresponds to GNU CC version 2.5.
203: @end ifclear
204: @ifclear USING
205: This manual documents how to port the GNU compiler,
206: as well as its new features and incompatibilities, and how to report
207: bugs. It corresponds to GNU CC version 2.5.
208: @end ifclear
209:
210: @end ifinfo
211: @menu
212: * Copying:: GNU General Public License says
213: how you can copy and share GNU CC.
214: * Contributors:: People who have contributed to GNU CC.
215: * Boycott:: Protect your freedom---fight ``look and feel''.
216: @ifset USING
217: * G++ and GCC:: You can compile C or C++ programs.
218: * Invoking GCC:: Command options supported by @samp{gcc}.
219: * Installation:: How to configure, compile and install GNU CC.
220: * C Extensions:: GNU extensions to the C language family.
221: * C++ Extensions:: GNU extensions to the C++ language.
222: * Trouble:: If you have trouble installing GNU CC.
223: * Bugs:: How, why and where to report bugs.
224: * Service:: How to find suppliers of support for GNU CC.
225: * VMS:: Using GNU CC on VMS.
226: @end ifset
227: @ifset INTERNALS
228: * Portability:: Goals of GNU CC's portability features.
229: * Interface:: Function-call interface of GNU CC output.
230: * Passes:: Order of passes, what they do, and what each file is for.
231: * RTL:: The intermediate representation that most passes work on.
232: * Machine Desc:: How to write machine description instruction patterns.
233: * Target Macros:: How to write the machine description C macros.
234: * Config:: Writing the @file{xm-@var{machine}.h} file.
235: @end ifset
236:
237: * Index:: Index of concepts and symbol names.
238: @end menu
239:
240: @node Copying, Contributors, Top, Top
241: @unnumbered GNU GENERAL PUBLIC LICENSE
242: @center Version 2, June 1991
243:
244: @display
245: Copyright @copyright{} 1989, 1991 Free Software Foundation, Inc.
246: 675 Mass Ave, Cambridge, MA 02139, USA
247:
248: Everyone is permitted to copy and distribute verbatim copies
249: of this license document, but changing it is not allowed.
250: @end display
251:
252: @unnumberedsec Preamble
253:
254: The licenses for most software are designed to take away your
255: freedom to share and change it. By contrast, the GNU General Public
256: License is intended to guarantee your freedom to share and change free
257: software---to make sure the software is free for all its users. This
258: General Public License applies to most of the Free Software
259: Foundation's software and to any other program whose authors commit to
260: using it. (Some other Free Software Foundation software is covered by
261: the GNU Library General Public License instead.) You can apply it to
262: your programs, too.
263:
264: When we speak of free software, we are referring to freedom, not
265: price. Our General Public Licenses are designed to make sure that you
266: have the freedom to distribute copies of free software (and charge for
267: this service if you wish), that you receive source code or can get it
268: if you want it, that you can change the software or use pieces of it
269: in new free programs; and that you know you can do these things.
270:
271: To protect your rights, we need to make restrictions that forbid
272: anyone to deny you these rights or to ask you to surrender the rights.
273: These restrictions translate to certain responsibilities for you if you
274: distribute copies of the software, or if you modify it.
275:
276: For example, if you distribute copies of such a program, whether
277: gratis or for a fee, you must give the recipients all the rights that
278: you have. You must make sure that they, too, receive or can get the
279: source code. And you must show them these terms so they know their
280: rights.
281:
282: We protect your rights with two steps: (1) copyright the software, and
283: (2) offer you this license which gives you legal permission to copy,
284: distribute and/or modify the software.
285:
286: Also, for each author's protection and ours, we want to make certain
287: that everyone understands that there is no warranty for this free
288: software. If the software is modified by someone else and passed on, we
289: want its recipients to know that what they have is not the original, so
290: that any problems introduced by others will not reflect on the original
291: authors' reputations.
292:
293: Finally, any free program is threatened constantly by software
294: patents. We wish to avoid the danger that redistributors of a free
295: program will individually obtain patent licenses, in effect making the
296: program proprietary. To prevent this, we have made it clear that any
297: patent must be licensed for everyone's free use or not licensed at all.
298:
299: The precise terms and conditions for copying, distribution and
300: modification follow.
301:
302: @iftex
303: @unnumberedsec TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
304: @end iftex
305: @ifinfo
306: @center TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
307: @end ifinfo
308:
309: @enumerate 0
310: @item
311: This License applies to any program or other work which contains
312: a notice placed by the copyright holder saying it may be distributed
313: under the terms of this General Public License. The ``Program'', below,
314: refers to any such program or work, and a ``work based on the Program''
315: means either the Program or any derivative work under copyright law:
316: that is to say, a work containing the Program or a portion of it,
317: either verbatim or with modifications and/or translated into another
318: language. (Hereinafter, translation is included without limitation in
319: the term ``modification''.) Each licensee is addressed as ``you''.
320:
321: Activities other than copying, distribution and modification are not
322: covered by this License; they are outside its scope. The act of
323: running the Program is not restricted, and the output from the Program
324: is covered only if its contents constitute a work based on the
325: Program (independent of having been made by running the Program).
326: Whether that is true depends on what the Program does.
327:
328: @item
329: You may copy and distribute verbatim copies of the Program's
330: source code as you receive it, in any medium, provided that you
331: conspicuously and appropriately publish on each copy an appropriate
332: copyright notice and disclaimer of warranty; keep intact all the
333: notices that refer to this License and to the absence of any warranty;
334: and give any other recipients of the Program a copy of this License
335: along with the Program.
336:
337: You may charge a fee for the physical act of transferring a copy, and
338: you may at your option offer warranty protection in exchange for a fee.
339:
340: @item
341: You may modify your copy or copies of the Program or any portion
342: of it, thus forming a work based on the Program, and copy and
343: distribute such modifications or work under the terms of Section 1
344: above, provided that you also meet all of these conditions:
345:
346: @enumerate a
347: @item
348: You must cause the modified files to carry prominent notices
349: stating that you changed the files and the date of any change.
350:
351: @item
352: You must cause any work that you distribute or publish, that in
353: whole or in part contains or is derived from the Program or any
354: part thereof, to be licensed as a whole at no charge to all third
355: parties under the terms of this License.
356:
357: @item
358: If the modified program normally reads commands interactively
359: when run, you must cause it, when started running for such
360: interactive use in the most ordinary way, to print or display an
361: announcement including an appropriate copyright notice and a
362: notice that there is no warranty (or else, saying that you provide
363: a warranty) and that users may redistribute the program under
364: these conditions, and telling the user how to view a copy of this
365: License. (Exception: if the Program itself is interactive but
366: does not normally print such an announcement, your work based on
367: the Program is not required to print an announcement.)
368: @end enumerate
369:
370: These requirements apply to the modified work as a whole. If
371: identifiable sections of that work are not derived from the Program,
372: and can be reasonably considered independent and separate works in
373: themselves, then this License, and its terms, do not apply to those
374: sections when you distribute them as separate works. But when you
375: distribute the same sections as part of a whole which is a work based
376: on the Program, the distribution of the whole must be on the terms of
377: this License, whose permissions for other licensees extend to the
378: entire whole, and thus to each and every part regardless of who wrote it.
379:
380: Thus, it is not the intent of this section to claim rights or contest
381: your rights to work written entirely by you; rather, the intent is to
382: exercise the right to control the distribution of derivative or
383: collective works based on the Program.
384:
385: In addition, mere aggregation of another work not based on the Program
386: with the Program (or with a work based on the Program) on a volume of
387: a storage or distribution medium does not bring the other work under
388: the scope of this License.
389:
390: @item
391: You may copy and distribute the Program (or a work based on it,
392: under Section 2) in object code or executable form under the terms of
393: Sections 1 and 2 above provided that you also do one of the following:
394:
395: @enumerate a
396: @item
397: Accompany it with the complete corresponding machine-readable
398: source code, which must be distributed under the terms of Sections
399: 1 and 2 above on a medium customarily used for software interchange; or,
400:
401: @item
402: Accompany it with a written offer, valid for at least three
403: years, to give any third party, for a charge no more than your
404: cost of physically performing source distribution, a complete
405: machine-readable copy of the corresponding source code, to be
406: distributed under the terms of Sections 1 and 2 above on a medium
407: customarily used for software interchange; or,
408:
409: @item
410: Accompany it with the information you received as to the offer
411: to distribute corresponding source code. (This alternative is
412: allowed only for noncommercial distribution and only if you
413: received the program in object code or executable form with such
414: an offer, in accord with Subsection b above.)
415: @end enumerate
416:
417: The source code for a work means the preferred form of the work for
418: making modifications to it. For an executable work, complete source
419: code means all the source code for all modules it contains, plus any
420: associated interface definition files, plus the scripts used to
421: control compilation and installation of the executable. However, as a
422: special exception, the source code distributed need not include
423: anything that is normally distributed (in either source or binary
424: form) with the major components (compiler, kernel, and so on) of the
425: operating system on which the executable runs, unless that component
426: itself accompanies the executable.
427:
428: If distribution of executable or object code is made by offering
429: access to copy from a designated place, then offering equivalent
430: access to copy the source code from the same place counts as
431: distribution of the source code, even though third parties are not
432: compelled to copy the source along with the object code.
433:
434: @item
435: You may not copy, modify, sublicense, or distribute the Program
436: except as expressly provided under this License. Any attempt
437: otherwise to copy, modify, sublicense or distribute the Program is
438: void, and will automatically terminate your rights under this License.
439: However, parties who have received copies, or rights, from you under
440: this License will not have their licenses terminated so long as such
441: parties remain in full compliance.
442:
443: @item
444: You are not required to accept this License, since you have not
445: signed it. However, nothing else grants you permission to modify or
446: distribute the Program or its derivative works. These actions are
447: prohibited by law if you do not accept this License. Therefore, by
448: modifying or distributing the Program (or any work based on the
449: Program), you indicate your acceptance of this License to do so, and
450: all its terms and conditions for copying, distributing or modifying
451: the Program or works based on it.
452:
453: @item
454: Each time you redistribute the Program (or any work based on the
455: Program), the recipient automatically receives a license from the
456: original licensor to copy, distribute or modify the Program subject to
457: these terms and conditions. You may not impose any further
458: restrictions on the recipients' exercise of the rights granted herein.
459: You are not responsible for enforcing compliance by third parties to
460: this License.
461:
462: @item
463: If, as a consequence of a court judgment or allegation of patent
464: infringement or for any other reason (not limited to patent issues),
465: conditions are imposed on you (whether by court order, agreement or
466: otherwise) that contradict the conditions of this License, they do not
467: excuse you from the conditions of this License. If you cannot
468: distribute so as to satisfy simultaneously your obligations under this
469: License and any other pertinent obligations, then as a consequence you
470: may not distribute the Program at all. For example, if a patent
471: license would not permit royalty-free redistribution of the Program by
472: all those who receive copies directly or indirectly through you, then
473: the only way you could satisfy both it and this License would be to
474: refrain entirely from distribution of the Program.
475:
476: If any portion of this section is held invalid or unenforceable under
477: any particular circumstance, the balance of the section is intended to
478: apply and the section as a whole is intended to apply in other
479: circumstances.
480:
481: It is not the purpose of this section to induce you to infringe any
482: patents or other property right claims or to contest validity of any
483: such claims; this section has the sole purpose of protecting the
484: integrity of the free software distribution system, which is
485: implemented by public license practices. Many people have made
486: generous contributions to the wide range of software distributed
487: through that system in reliance on consistent application of that
488: system; it is up to the author/donor to decide if he or she is willing
489: to distribute software through any other system and a licensee cannot
490: impose that choice.
491:
492: This section is intended to make thoroughly clear what is believed to
493: be a consequence of the rest of this License.
494:
495: @item
496: If the distribution and/or use of the Program is restricted in
497: certain countries either by patents or by copyrighted interfaces, the
498: original copyright holder who places the Program under this License
499: may add an explicit geographical distribution limitation excluding
500: those countries, so that distribution is permitted only in or among
501: countries not thus excluded. In such case, this License incorporates
502: the limitation as if written in the body of this License.
503:
504: @item
505: The Free Software Foundation may publish revised and/or new versions
506: of the General Public License from time to time. Such new versions will
507: be similar in spirit to the present version, but may differ in detail to
508: address new problems or concerns.
509:
510: Each version is given a distinguishing version number. If the Program
511: specifies a version number of this License which applies to it and ``any
512: later version'', you have the option of following the terms and conditions
513: either of that version or of any later version published by the Free
514: Software Foundation. If the Program does not specify a version number of
515: this License, you may choose any version ever published by the Free Software
516: Foundation.
517:
518: @item
519: If you wish to incorporate parts of the Program into other free
520: programs whose distribution conditions are different, write to the author
521: to ask for permission. For software which is copyrighted by the Free
522: Software Foundation, write to the Free Software Foundation; we sometimes
523: make exceptions for this. Our decision will be guided by the two goals
524: of preserving the free status of all derivatives of our free software and
525: of promoting the sharing and reuse of software generally.
526:
527: @iftex
528: @heading NO WARRANTY
529: @end iftex
530: @ifinfo
531: @center NO WARRANTY
532: @end ifinfo
533:
534: @item
535: BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
536: FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
537: OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
538: PROVIDE THE PROGRAM ``AS IS'' WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
539: OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
540: MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
541: TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
542: PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
543: REPAIR OR CORRECTION.
544:
545: @item
546: IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
547: WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
548: REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
549: INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
550: OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
551: TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
552: YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
553: PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
554: POSSIBILITY OF SUCH DAMAGES.
555: @end enumerate
556:
557: @iftex
558: @heading END OF TERMS AND CONDITIONS
559: @end iftex
560: @ifinfo
561: @center END OF TERMS AND CONDITIONS
562: @end ifinfo
563:
564: @page
565: @unnumberedsec How to Apply These Terms to Your New Programs
566:
567: If you develop a new program, and you want it to be of the greatest
568: possible use to the public, the best way to achieve this is to make it
569: free software which everyone can redistribute and change under these terms.
570:
571: To do so, attach the following notices to the program. It is safest
572: to attach them to the start of each source file to most effectively
573: convey the exclusion of warranty; and each file should have at least
574: the ``copyright'' line and a pointer to where the full notice is found.
575:
576: @smallexample
577: @var{one line to give the program's name and a brief idea of what it does.}
578: Copyright (C) 19@var{yy} @var{name of author}
579:
580: This program is free software; you can redistribute it and/or modify
581: it under the terms of the GNU General Public License as published by
582: the Free Software Foundation; either version 2 of the License, or
583: (at your option) any later version.
584:
585: This program is distributed in the hope that it will be useful,
586: but WITHOUT ANY WARRANTY; without even the implied warranty of
587: MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
588: GNU General Public License for more details.
589:
590: You should have received a copy of the GNU General Public License
591: along with this program; if not, write to the Free Software
592: Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
593: @end smallexample
594:
595: Also add information on how to contact you by electronic and paper mail.
596:
597: If the program is interactive, make it output a short notice like this
598: when it starts in an interactive mode:
599:
600: @smallexample
601: Gnomovision version 69, Copyright (C) 19@var{yy} @var{name of author}
602: Gnomovision comes with ABSOLUTELY NO WARRANTY; for details
603: type `show w'.
604: This is free software, and you are welcome to redistribute it
605: under certain conditions; type `show c' for details.
606: @end smallexample
607:
608: The hypothetical commands @samp{show w} and @samp{show c} should show
609: the appropriate parts of the General Public License. Of course, the
610: commands you use may be called something other than @samp{show w} and
611: @samp{show c}; they could even be mouse-clicks or menu items---whatever
612: suits your program.
613:
614: You should also get your employer (if you work as a programmer) or your
615: school, if any, to sign a ``copyright disclaimer'' for the program, if
616: necessary. Here is a sample; alter the names:
617:
618: @smallexample
619: Yoyodyne, Inc., hereby disclaims all copyright interest in the program
620: `Gnomovision' (which makes passes at compilers) written by James Hacker.
621:
622: @var{signature of Ty Coon}, 1 April 1989
623: Ty Coon, President of Vice
624: @end smallexample
625:
626: This General Public License does not permit incorporating your program into
627: proprietary programs. If your program is a subroutine library, you may
628: consider it more useful to permit linking proprietary applications with the
629: library. If this is what you want to do, use the GNU Library General
630: Public License instead of this License.
631:
632: @node Contributors, Boycott, Copying, Top
633: @unnumbered Contributors to GNU CC
634: @cindex contributors
635:
636: In addition to Richard Stallman, several people have written parts
637: of GNU CC.
638:
639: @itemize @bullet
640: @item
641: The idea of using RTL and some of the optimization ideas came from the
642: program PO written at the University of Arizona by Jack Davidson and
643: Christopher Fraser. See ``Register Allocation and Exhaustive Peephole
644: Optimization'', Software Practice and Experience 14 (9), Sept. 1984,
645: 857-866.
646:
647: @item
648: Paul Rubin wrote most of the preprocessor.
649:
650: @item
651: Leonard Tower wrote parts of the parser, RTL generator, and RTL
652: definitions, and of the Vax machine description.
653:
654: @item
655: Ted Lemon wrote parts of the RTL reader and printer.
656:
657: @item
658: Jim Wilson implemented loop strength reduction and some other
659: loop optimizations.
660:
661: @item
662: Nobuyuki Hikichi of Software Research Associates, Tokyo, contributed
663: the support for the Sony NEWS machine.
664:
665: @item
666: Charles LaBrec contributed the support for the Integrated Solutions
667: 68020 system.
668:
669: @item
670: Michael Tiemann of Cygnus Support wrote the front end for C++, as well
671: as the support for inline functions and instruction scheduling. Also
672: the descriptions of the National Semiconductor 32000 series cpu, the
673: SPARC cpu and part of the Motorola 88000 cpu.
674:
675: @item
676: Jan Stein of the Chalmers Computer Society provided support for
677: Genix, as well as part of the 32000 machine description.
678:
679: @item
680: Randy Smith finished the Sun FPA support.
681:
682: @item
683: Robert Brown implemented the support for Encore 32000 systems.
684:
685: @item
686: David Kashtan of SRI adapted GNU CC to the Vomit-Making System (VMS).
687:
688: @item
689: Alex Crain provided changes for the 3b1.
690:
691: @item
692: Greg Satz and Chris Hanson assisted in making GNU CC work on HP-UX for
693: the 9000 series 300.
694:
695: @item
696: William Schelter did most of the work on the Intel 80386 support.
697:
698: @item
699: Christopher Smith did the port for Convex machines.
700:
701: @item
702: Paul Petersen wrote the machine description for the Alliant FX/8.
703:
704: @item
705: Alain Lichnewsky ported GNU CC to the Mips cpu.
706:
707: @item
708: Devon Bowen, Dale Wiles and Kevin Zachmann ported GNU CC to the Tahoe.
709:
710: @item
711: Jonathan Stone wrote the machine description for the Pyramid computer.
712:
713: @item
714: Gary Miller ported GNU CC to Charles River Data Systems machines.
715:
716: @item
717: Richard Kenner of the New York University Ultracomputer Research
718: Laboratory wrote the machine descriptions for the AMD 29000, the DEC
719: Alpha, the IBM RT PC, and the IBM RS/6000 as well as the support for
720: instruction attributes. He also made changes to better support RISC
721: processors including changes to common subexpression elimination,
722: strength reduction, function calling sequence handling, and condition
723: code support, in addition to generalizing the code for frame pointer
724: elimination.
725:
726: @item
727: Richard Kenner and Michael Tiemann jointly developed reorg.c, the delay
728: slot scheduler.
729:
730: @item
731: Mike Meissner and Tom Wood of Data General finished the port to the
732: Motorola 88000.
733:
734: @item
735: Masanobu Yuhara of Fujitsu Laboratories implemented the machine
736: description for the Tron architecture (specifically, the Gmicro).
737:
738: @item
739: NeXT, Inc.@: donated the front end that supports the Objective C
740: language.
741: @c We need to be careful to make it clear that "Objective C"
742: @c is the name of a language, not that of a program or product.
743:
744: @item
745: James van Artsdalen wrote the code that makes efficient use of
746: the Intel 80387 register stack.
747:
748: @item
749: Mike Meissner at the Open Software Foundation finished the port to the
750: MIPS cpu, including adding ECOFF debug support.
751:
752: @item
753: Ron Guilmette implemented the @code{protoize} and @code{unprotoize}
754: tools, the support for Dwarf symbolic debugging information, and much of
755: the support for System V Release 4. He has also worked heavily on the
756: Intel 386 and 860 support.
757:
758: @item
759: Torbjorn Granlund of the Swedish Institute of Computer Science
760: implemented multiply-by-constant optimization and better long long
761: support, and improved leaf function register allocation.
762:
763: @item
764: Mike Stump implemented the support for Elxsi 64 bit CPU.
765:
766: @item
767: John Wehle added the machine description for the Western Electric 32000
768: processor used in several 3b series machines (no relation to the
769: National Semiconductor 32000 processor).
770:
771: @ignore @c These features aren't advertised yet, since they don't fully work.
772: @item
773: Analog Devices helped implement the support for complex data types
774: and iterators.
775: @end ignore
776:
777: @item
778: Holger Teutsch provided the support for the Clipper cpu.
779:
780: @item
781: Kresten Krab Thorup wrote the run time support for the Objective C
782: language.
783:
784: @item
785: Stephen Moshier contributed the floating point emulator that assists in
786: cross-compilation and permits support for floating point numbers wider
787: than 64 bits.
788:
789: @item
790: David Edelsohn contributed the changes to RS/6000 port to make it
791: support the PowerPC and POWER2 architectures.
792:
793: @item
794: Steve Chamberlain wrote the support for the Hitachi SH processor.
795:
796: @item
797: Peter Schauer wrote the code to allow debugging to work on the Alpha.
798: @end itemize
799:
800: @node Boycott
801: @chapter Protect Your Freedom---Fight ``Look And Feel''
802: @c the above chapter heading overflows onto the next line. --mew 1/26/93
803:
804: @quotation
805: @i{This section is a political message from the League for Programming
806: Freedom to the users of GNU CC. It is included here as an expression
807: of support for the League on the part of the Free Software Foundation.}
808: @end quotation
809:
810: Apple and Lotus are trying to create a new form of legal monopoly: a
811: copyright on a class of user interfaces. These monopolies would cause
812: serious problems for users and developers of computer software and
813: systems. Xerox, too, has tried to make a monopoly for itself on window
814: systems; their suit against Apple was thrown out on a technicality, but
815: Xerox has not said anything to indicate it wouldn't try again.
816:
817: Until a few years ago, the law seemed clear: no one could restrict
818: others from using a user interface; programmers were free to implement
819: any interface they chose. Imitating interfaces, sometimes with changes,
820: was standard practice in the computer field. The interfaces we know
821: evolved gradually in this way; for example, the Macintosh user interface
822: drew ideas from the Xerox interface, which in turn drew on work done at
823: Stanford and SRI. 1-2-3 imitated VisiCalc, and dBase imitated a
824: database program from JPL.
825:
826: Most computer companies, and nearly all computer users, were happy with
827: this state of affairs. The companies that are suing say it does not
828: offer ``enough incentive'' to develop their products, but they must have
829: considered it ``enough'' when they made their decision to do so. It
830: seems they are not satisfied with the opportunity to continue to compete
831: in the marketplace---not even with a head start.
832:
833: If companies like Xerox, Lotus, and Apple are permitted to make law
834: through the courts, the precedent will hobble the software industry:
835:
836: @itemize @bullet
837: @item
838: Gratuitous incompatibilities will burden users. Imagine if each
839: car manufacturer had to arrange the pedals in a different order.
840:
841: @item
842: Software will become and remain more expensive. Users will be
843: ``locked in'' to proprietary interfaces, for which there is no real
844: competition.
845:
846: @item
847: Large companies have an unfair advantage wherever lawsuits become
848: commonplace. Since they can easily afford to sue, they can intimidate
849: small companies with threats even when they don't really have a case.
850:
851: @item
852: User interface improvements will come slower, since incremental
853: evolution through creative imitation will no longer be permitted.
854:
855: @item
856: Even Apple, etc., will find it harder to make improvements if
857: they can no longer adapt the good ideas that others introduce, for
858: fear of weakening their own legal positions. Some users suggest that
859: this stagnation may already have started.
860:
861: @item
862: If you use GNU software, you might find it of some concern that user
863: interface copyright will make it hard for the Free Software Foundation
864: to develop programs compatible with the interfaces that you already
865: know.
866: @end itemize
867:
868: To protect our freedom from lawsuits like these, a group of programmers
869: and users have formed a new grass-roots political organization, the
870: League for Programming Freedom.
871:
872: The purpose of the League is to oppose new monopolistic practices such
873: as user-interface copyright and software patents; it calls for a return
874: to the legal policies of the recent past, in which these practices were
875: not allowed. The League is not concerned with free software as an
876: issue, and not affiliated with the Free Software Foundation.
877:
878: The League's membership rolls include John McCarthy, inventor of Lisp,
879: Marvin Minsky, founder of the Artificial Intelligence lab, Guy L.
880: Steele, Jr., author of well-known books on Lisp and C, as well as
881: Richard Stallman, the developer of GNU CC. Please join and add your
882: name to the list. Membership dues in the League are $42 per year for
883: programmers, managers and professionals; $10.50 for students; $21 for
884: others.
885:
886: The League needs both activist members and members who only pay their
887: dues.
888:
889: To join, or for more information, phone (617) 243-4091 or write to:
890:
891: @display
892: League for Programming Freedom
893: 1 Kendall Square #143
894: P.O. Box 9171
895: Cambridge, MA 02139
896: @end display
897:
898: You can also send electronic mail to @code{league@@prep.ai.mit.edu}.
899:
900: Here are some suggestions from the League for things you can do to
901: protect your freedom to write programs:
902:
903: @itemize @bullet
904: @item
905: Don't buy from Xerox, Lotus or Apple. Buy from their competitors or
906: from the defendants they are suing.
907:
908: @item
909: Don't develop software to work with the systems made by these companies.
910:
911: @item
912: Port your existing software to competing systems, so that you encourage
913: users to switch.
914:
915: @item
916: Write letters to company presidents to let them know their conduct
917: is unacceptable.
918:
919: @item
920: Tell your friends and colleagues about this issue and how it threatens
921: to ruin the computer industry.
922:
923: @item
924: Above all, don't work for the look-and-feel plaintiffs, and don't
925: accept contracts from them.
926:
927: @item
928: Write to Congress to explain the importance of this issue.
929:
930: @display
931: House Subcommittee on Intellectual Property
932: 2137 Rayburn Bldg
933: Washington, DC 20515
934:
935: Senate Subcommittee on Patents, Trademarks and Copyrights
936: United States Senate
937: Washington, DC 20510
938: @end display
939:
940: (These committees have received lots of mail already; let's give them
941: even more.)
942: @end itemize
943:
944: Express your opinion! You can make a difference.
945:
946: @ifset USING
947: @node G++ and GCC
948: @chapter Compile C, C++, or Objective C
949:
950: @cindex Objective C
951: The C, C++, and Objective C versions of the compiler are integrated; the
952: GNU C compiler can compile programs written in C, C++, or Objective C.
953:
954: @cindex GCC
955: ``GCC'' is a common shorthand term for the GNU C compiler. This is both
956: the most general name for the compiler, and the name used when the
957: emphasis is on compiling C programs.
958:
959: @cindex C++
960: @cindex G++
961: When referring to C++ compilation, it is usual to call the compiler
962: ``G++''. Since there is only one compiler, it is also accurate to call
963: it ``GCC'' no matter what the language context; however, the term
964: ``G++'' is more useful when the emphasis is on compiling C++ programs.
965:
966: @cindex compiler compared to C++ preprocessor
967: @cindex intermediate C version, nonexistent
968: @cindex C intermediate output, nonexistent
969: G++ is a @emph{compiler}, not merely a preprocessor. G++ builds object
970: code directly from your C++ program source. There is no intermediate C
971: version of the program. (By contrast, for example, some other
972: implementations use a program that generates a C program from your C++
973: source.) Avoiding an intermediate C representation of the program means
974: that you get better object code, and better debugging information. The
975: GNU debugger, GDB, works with this information in the object code to
976: give you comprehensive C++ source-level editing capabilities
977: (@pxref{C,,C and C++,gdb.info, Debugging with GDB}).
978:
979: @c FIXME! Someone who knows something about Objective C ought to put in
980: @c a paragraph or two about it here, and move the index entry down when
981: @c there is more to point to than the general mention in the 1st par.
982:
983: @include invoke.texi
984:
985: @include install.texi
986:
987: @include extend.texi
988:
989: @node Trouble
990: @chapter Known Causes of Trouble with GNU CC
991: @cindex bugs, known
992: @cindex installation trouble
993: @cindex known causes of trouble
994:
995: This section describes known problems that affect users of GNU CC. Most
996: of these are not GNU CC bugs per se---if they were, we would fix them.
997: But the result for a user may be like the result of a bug.
998:
999: Some of these problems are due to bugs in other software, some are
1000: missing features that are too much work to add, and some are places
1001: where people's opinions differ as to what is best.
1002:
1003: @menu
1004: * Actual Bugs:: Bugs we will fix later.
1005: * Installation Problems:: Problems that manifest when you install GNU CC.
1006: * Cross-Compiler Problems:: Common problems of cross compiling with GNU CC.
1007: * Interoperation:: Problems using GNU CC with other compilers,
1008: and with certain linkers, assemblers and debuggers.
1009: * External Bugs:: Problems compiling certain programs.
1010: * Incompatibilities:: GNU CC is incompatible with traditional C.
1011: * Fixed Headers:: GNU C uses corrected versions of system header files.
1012: This is necessary, but doesn't always work smoothly.
1013: * Disappointments:: Regrettable things we can't change, but not quite bugs.
1014: * C++ Misunderstandings:: Common misunderstandings with GNU C++.
1015: * Protoize Caveats:: Things to watch out for when using @code{protoize}.
1016: * Non-bugs:: Things we think are right, but some others disagree.
1017: * Warnings and Errors:: Which problems in your code get warnings,
1018: and which get errors.
1019: @end menu
1020:
1021: @node Actual Bugs
1022: @section Actual Bugs We Haven't Fixed Yet
1023:
1024: @itemize @bullet
1025: @item
1026: The @code{fixincludes} script interacts badly with automounters; if the
1027: directory of system header files is automounted, it tends to be
1028: unmounted while @code{fixincludes} is running. This would seem to be a
1029: bug in the automounter. We don't know any good way to work around it.
1030:
1031: @item
1032: The @code{fixproto} script will sometimes add prototypes for the
1033: @code{sigsetjmp} and @code{siglongjmp} functions that reference the
1034: @code{jmp_buf} type before that type is defined. To work around this,
1035: edit the offending file and place the typedef in front of the
1036: prototypes.
1037:
1038: @item
1039: Loop unrolling doesn't work properly for certain C++ programs. This is
1040: because of difficulty in updating the debugging information within the
1041: loop being unrolled. We plan to revamp the representation of debugging
1042: information so that this will work properly, but we have not done this
1043: in version 2.5 because we don't want to delay it any further.
1044: @end itemize
1045:
1046: @node Installation Problems
1047: @section Installation Problems
1048:
1049: This is a list of problems (and some apparent problems which don't
1050: really mean anything is wrong) that show up during installation of GNU
1051: CC.
1052:
1053: @itemize @bullet
1054: @item
1055: On certain systems, defining certain environment variables such as
1056: @code{CC} can interfere with the functioning of @code{make}.
1057:
1058: @item
1059: If you encounter seemingly strange errors when trying to build the
1060: compiler in a directory other than the source directory, it could be
1061: because you have previously configured the compiler in the source
1062: directory. Make sure you have done all the necessary preparations.
1063: @xref{Other Dir}.
1064:
1065: @item
1066: If you build GNU CC on a BSD system using a directory stored in a System
1067: V file system, problems may occur in running @code{fixincludes} if the
1068: System V file system doesn't support symbolic links. These problems
1069: result in a failure to fix the declaration of @code{size_t} in
1070: @file{sys/types.h}. If you find that @code{size_t} is a signed type and
1071: that type mismatches occur, this could be the cause.
1072:
1073: The solution is not to use such a directory for building GNU CC.
1074:
1075: @item
1076: In previous versions of GNU CC, the @code{gcc} driver program looked for
1077: @code{as} and @code{ld} in various places; for example, in files
1078: beginning with @file{/usr/local/lib/gcc-}. GNU CC version 2 looks for
1079: them in the directory
1080: @file{/usr/local/lib/gcc-lib/@var{target}/@var{version}}.
1081:
1082: Thus, to use a version of @code{as} or @code{ld} that is not the system
1083: default, for example @code{gas} or GNU @code{ld}, you must put them in
1084: that directory (or make links to them from that directory).
1085:
1086: @item
1087: Some commands executed when making the compiler may fail (return a
1088: non-zero status) and be ignored by @code{make}. These failures, which
1089: are often due to files that were not found, are expected, and can safely
1090: be ignored.
1091:
1092: @item
1093: It is normal to have warnings in compiling certain files about
1094: unreachable code and about enumeration type clashes. These files' names
1095: begin with @samp{insn-}. Also, @file{real.c} may get some warnings that
1096: you can ignore.
1097:
1098: @item
1099: Sometimes @code{make} recompiles parts of the compiler when installing
1100: the compiler. In one case, this was traced down to a bug in
1101: @code{make}. Either ignore the problem or switch to GNU Make.
1102:
1103: @item
1104: If you have installed a program known as purify, you may find that it
1105: causes errors while linking @code{enquire}, which is part of building
1106: GNU CC. The fix is to get rid of the file @code{real-ld} which purify
1107: installs---so that GNU CC won't try to use it.
1108:
1109: @item
1110: On Linux SLS 1.01, there is a problem with @file{libc.a}: it does not
1111: contain the obstack functions. However, GNU CC assumes that the obstack
1112: functions are in @file{libc.a} when it is the GNU C library. To work
1113: around this problem, change the @code{__GNU_LIBRARY__} conditional
1114: around line 31 to @samp{#if 1}.
1115:
1116: @item
1117: On some 386 systems, building the compiler never finishes because
1118: @code{enquire} hangs due to a hardware problem in the motherboard---it
1119: reports floating point exceptions to the kernel incorrectly. You can
1120: install GNU CC except for @file{float.h} by patching out the command to
1121: run @code{enquire}. You may also be able to fix the problem for real by
1122: getting a replacement motherboard. This problem was observed in
1123: Revision E of the Micronics motherboard, and is fixed in Revision F.
1124: It has also been observed in the MYLEX MXA-33 motherboard.
1125:
1126: If you encounter this problem, you may also want to consider removing
1127: the FPU from the socket during the compilation. Alternatively, if you
1128: are running SCO Unix, you can reboot and force the FPU to be ignored.
1129: To do this, type @samp{hd(40)unix auto ignorefpu}.
1130:
1131: @item
1132: On some 386 systems, GNU CC crashes trying to compile @file{enquire.c}.
1133: This happens on machines that don't have a 387 FPU chip. On 386
1134: machines, the system kernel is supposed to emulate the 387 when you
1135: don't have one. The crash is due to a bug in the emulator.
1136:
1137: One of these systems is the Unix from Interactive Systems: 386/ix.
1138: On this system, an alternate emulator is provided, and it does work.
1139: To use it, execute this command as super-user:
1140:
1141: @example
1142: ln /etc/emulator.rel1 /etc/emulator
1143: @end example
1144:
1145: @noindent
1146: and then reboot the system. (The default emulator file remains present
1147: under the name @file{emulator.dflt}.)
1148:
1149: Try using @file{/etc/emulator.att}, if you have such a problem on the
1150: SCO system.
1151:
1152: Another system which has this problem is Esix. We don't know whether it
1153: has an alternate emulator that works.
1154:
1155: On NetBSD 0.8, a similar problem manifests itself as these error messages:
1156:
1157: @example
1158: enquire.c: In function `fprop':
1159: enquire.c:2328: floating overflow
1160: @end example
1161:
1162: @item
1163: On SCO systems, when compiling GNU CC with the system's compiler,
1164: do not use @samp{-O}. Some versions of the system's compiler miscompile
1165: GNU CC with @samp{-O}.
1166:
1167: @cindex @code{genflags}, crash on Sun 4
1168: @item
1169: Sometimes on a Sun 4 you may observe a crash in the program
1170: @code{genflags} or @code{genoutput} while building GNU CC. This is said to
1171: be due to a bug in @code{sh}. You can probably get around it by running
1172: @code{genflags} or @code{genoutput} manually and then retrying the
1173: @code{make}.
1174:
1175: @item
1176: On Solaris 2, executables of GNU CC version 2.0.2 are commonly
1177: available, but they have a bug that shows up when compiling current
1178: versions of GNU CC: undefined symbol errors occur during assembly if you
1179: use @samp{-g}.
1180:
1181: The solution is to compile the current version of GNU CC without
1182: @samp{-g}. That makes a working compiler which you can use to recompile
1183: with @samp{-g}.
1184:
1185: @item
1186: Solaris 2 comes with a number of optional OS packages. Some of these
1187: packages are needed to use GNU CC fully. If you did not install all
1188: optional packages when installing Solaris, you will need to verify that
1189: the packages that GNU CC needs are installed.
1190:
1191: To check whether an optional package is installed, use
1192: the @code{pkginfo} command. To add an optional package, use the
1193: @code{pkgadd} command. For further details, see the Solaris
1194: documentation.
1195:
1196: For Solaris 2.0 and 2.1, GNU CC needs six packages: @samp{SUNWarc},
1197: @samp{SUNWbtool}, @samp{SUNWesu}, @samp{SUNWhea}, @samp{SUNWlibm}, and
1198: @samp{SUNWtoo}.
1199:
1200: For Solaris 2.2, GNU CC needs an additional seventh package: @samp{SUNWsprot}.
1201:
1202: @item
1203: On Solaris 2, trying to use the linker and other tools in
1204: @file{/usr/ucb} to install GNU CC has been observed to cause trouble.
1205: For example, the linker may hang indefinitely. The fix is to remove
1206: @file{/usr/ucb} from your @code{PATH}.
1207:
1208: @item
1209: If you use the 1.31 version of the MIPS assembler (such as was shipped
1210: with Ultrix 3.1), you will need to use the -fno-delayed-branch switch
1211: when optimizing floating point code. Otherwise, the assembler will
1212: complain when the GCC compiler fills a branch delay slot with a
1213: floating point instruction, such as @code{add.d}.
1214:
1215: @item
1216: If on a MIPS system you get an error message saying ``does not have gp
1217: sections for all it's [sic] sectons [sic]'', don't worry about it. This
1218: happens whenever you use GAS with the MIPS linker, but there is not
1219: really anything wrong, and it is okay to use the output file. You can
1220: stop such warnings by installing the GNU linker.
1221:
1222: It would be nice to extend GAS to produce the gp tables, but they are
1223: optional, and there should not be a warning about their absence.
1224:
1225: @item
1226: In Ultrix 4.0 on the MIPS machine, @file{stdio.h} does not work with GNU
1227: CC at all unless it has been fixed with @code{fixincludes}. This causes
1228: problems in building GNU CC. Once GNU CC is installed, the problems go
1229: away.
1230:
1231: To work around this problem, when making the stage 1 compiler, specify
1232: this option to Make:
1233:
1234: @example
1235: GCC_FOR_TARGET="./xgcc -B./ -I./include"
1236: @end example
1237:
1238: When making stage 2 and stage 3, specify this option:
1239:
1240: @example
1241: CFLAGS="-g -I./include"
1242: @end example
1243:
1244: @item
1245: Users have reported some problems with version 2.0 of the MIPS
1246: compiler tools that were shipped with Ultrix 4.1. Version 2.10
1247: which came with Ultrix 4.2 seems to work fine.
1248:
1249: @item
1250: Some versions of the MIPS linker will issue an assertion failure
1251: when linking code that uses @code{alloca} against shared
1252: libraries on RISC-OS 5.0, and DEC's OSF/1 systems. This is a bug
1253: in the linker, that is supposed to be fixed in future revisions.
1254: To protect against this, GNU CC passes @samp{-non_shared} to the
1255: linker unless you pass an explicit @samp{-shared} or
1256: @samp{-call_shared} switch.
1257:
1258: @item
1259: On System V release 3, you may get this error message
1260: while linking:
1261:
1262: @smallexample
1263: ld fatal: failed to write symbol name @var{something}
1264: in strings table for file @var{whatever}
1265: @end smallexample
1266:
1267: This probably indicates that the disk is full or your ULIMIT won't allow
1268: the file to be as large as it needs to be.
1269:
1270: This problem can also result because the kernel parameter @code{MAXUMEM}
1271: is too small. If so, you must regenerate the kernel and make the value
1272: much larger. The default value is reported to be 1024; a value of 32768
1273: is said to work. Smaller values may also work.
1274:
1275: @item
1276: On System V, if you get an error like this,
1277:
1278: @example
1279: /usr/local/lib/bison.simple: In function `yyparse':
1280: /usr/local/lib/bison.simple:625: virtual memory exhausted
1281: @end example
1282:
1283: @noindent
1284: that too indicates a problem with disk space, ULIMIT, or @code{MAXUMEM}.
1285:
1286: @item
1287: Current GNU CC versions probably do not work on version 2 of the NeXT
1288: operating system.
1289:
1290: @item
1291: On NeXTStep 3.0, the Objective C compiler does not work, due,
1292: apparently, to a kernel bug that it happens to trigger. This problem
1293: does not happen on 3.1.
1294:
1295: @item
1296: On the Tower models 4@var{n}0 and 6@var{n}0, by default a process is not
1297: allowed to have more than one megabyte of memory. GNU CC cannot compile
1298: itself (or many other programs) with @samp{-O} in that much memory.
1299:
1300: To solve this problem, reconfigure the kernel adding the following line
1301: to the configuration file:
1302:
1303: @smallexample
1304: MAXUMEM = 4096
1305: @end smallexample
1306:
1307: @item
1308: On HP 9000 series 300 or 400 running HP-UX release 8.0, there is a bug
1309: in the assembler that must be fixed before GNU CC can be built. This
1310: bug manifests itself during the first stage of compilation, while
1311: building @file{libgcc2.a}:
1312:
1313: @smallexample
1314: _floatdisf
1315: cc1: warning: `-g' option not supported on this version of GCC
1316: cc1: warning: `-g1' option not supported on this version of GCC
1317: ./xgcc: Internal compiler error: program as got fatal signal 11
1318: @end smallexample
1319:
1320: A patched version of the assembler is available by anonymous ftp from
1321: @code{altdorf.ai.mit.edu} as the file
1322: @file{archive/cph/hpux-8.0-assembler}. If you have HP software support,
1323: the patch can also be obtained directly from HP, as described in the
1324: following note:
1325:
1326: @quotation
1327: This is the patched assembler, to patch SR#1653-010439, where the
1328: assembler aborts on floating point constants.
1329:
1330: The bug is not really in the assembler, but in the shared library
1331: version of the function ``cvtnum(3c)''. The bug on ``cvtnum(3c)'' is
1332: SR#4701-078451. Anyway, the attached assembler uses the archive
1333: library version of ``cvtnum(3c)'' and thus does not exhibit the bug.
1334: @end quotation
1335:
1336: This patch is also known as PHCO_0800.
1337:
1338: @item
1339: On HP-UX version 8.05, but not on 8.07 or more recent versions,
1340: the @code{fixproto} shell script triggers a bug in the system shell.
1341: If you encounter this problem, upgrade your operating system or
1342: use BASH (the GNU shell) to run @code{fixproto}.
1343:
1344: @item
1345: Some versions of the Pyramid C compiler are reported to be unable to
1346: compile GNU CC. You must use an older version of GNU CC for
1347: bootstrapping. One indication of this problem is if you get a crash
1348: when GNU CC compiles the function @code{muldi3} in file @file{libgcc2.c}.
1349:
1350: You may be able to succeed by getting GNU CC version 1, installing it,
1351: and using it to compile GNU CC version 2. The bug in the Pyramid C
1352: compiler does not seem to affect GNU CC version 1.
1353:
1354: @item
1355: There may be similar problems on System V Release 3.1 on 386 systems.
1356:
1357: @item
1358: On the Intel Paragon (an i860 machine), if you are using operating
1359: system version 1.0, you will get warnings or errors about redefinition
1360: of @code{va_arg} when you build GNU CC.
1361:
1362: If this happens, then you need to link most programs with the library
1363: @file{iclib.a}. You must also modify @file{stdio.h} as follows: before
1364: the lines
1365:
1366: @example
1367: #if defined(__i860__) && !defined(_VA_LIST)
1368: #include <va_list.h>
1369: @end example
1370:
1371: @noindent
1372: insert the line
1373:
1374: @example
1375: #if __PGC__
1376: @end example
1377:
1378: @noindent
1379: and after the lines
1380:
1381: @example
1382: extern int vprintf(const char *, va_list );
1383: extern int vsprintf(char *, const char *, va_list );
1384: #endif
1385: @end example
1386:
1387: @noindent
1388: insert the line
1389:
1390: @example
1391: #endif /* __PGC__ */
1392: @end example
1393:
1394: These problems don't exist in operating system version 1.1.
1395:
1396: @item
1397: On the Altos 3068, programs compiled with GNU CC won't work unless you
1398: fix a kernel bug. This happens using system versions V.2.2 1.0gT1 and
1399: V.2.2 1.0e and perhaps later versions as well. See the file
1400: @file{README.ALTOS}.
1401:
1402: @item
1403: You will get several sorts of compilation and linking errors on the
1404: we32k if you don't follow the special instructions. @xref{WE32K
1405: Install}.
1406: @end itemize
1407:
1408: @node Cross-Compiler Problems
1409: @section Cross-Compiler Problems
1410:
1411: You may run into problems with cross compilation on certain machines,
1412: for several reasons.
1413:
1414: @itemize @bullet
1415: @item
1416: Cross compilation can run into trouble for certain machines because
1417: some target machines' assemblers require floating point numbers to be
1418: written as @emph{integer} constants in certain contexts.
1419:
1420: The compiler writes these integer constants by examining the floating
1421: point value as an integer and printing that integer, because this is
1422: simple to write and independent of the details of the floating point
1423: representation. But this does not work if the compiler is running on
1424: a different machine with an incompatible floating point format, or
1425: even a different byte-ordering.
1426:
1427: In addition, correct constant folding of floating point values
1428: requires representing them in the target machine's format.
1429: (The C standard does not quite require this, but in practice
1430: it is the only way to win.)
1431:
1432: It is now possible to overcome these problems by defining macros such
1433: as @code{REAL_VALUE_TYPE}. But doing so is a substantial amount of
1434: work for each target machine.
1435: @ifset INTERNALS
1436: @xref{Cross-compilation}.
1437: @end ifset
1438: @ifclear INTERNALS
1439: @xref{Cross-compilation,,Cross Compilation and Floating Point Format,
1440: gcc.info, Using and Porting GCC}.
1441: @end ifclear
1442:
1443: @item
1444: At present, the program @file{mips-tfile} which adds debug
1445: support to object files on MIPS systems does not work in a cross
1446: compile environment.
1447: @end itemize
1448:
1449: @node Interoperation
1450: @section Interoperation
1451:
1452: This section lists various difficulties encountered in using GNU C or
1453: GNU C++ together with other compilers or with the assemblers, linkers,
1454: libraries and debuggers on certain systems.
1455:
1456: @itemize @bullet
1457: @item
1458: Objective C does not work on the RS/6000 or the Alpha.
1459:
1460: @item
1461: C++ does not work on the Alpha.
1462:
1463: @item
1464: GNU C++ does not do name mangling in the same way as other C++
1465: compilers. This means that object files compiled with one compiler
1466: cannot be used with another.
1467:
1468: This effect is intentional, to protect you from more subtle problems.
1469: Compilers differ as to many internal details of C++ implementation,
1470: including: how class instances are laid out, how multiple inheritance is
1471: implemented, and how virtual function calls are handled. If the name
1472: encoding were made the same, your programs would link against libraries
1473: provided from other compilers---but the programs would then crash when
1474: run. Incompatible libraries are then detected at link time, rather than
1475: at run time.
1476:
1477: @item
1478: Older GDB versions sometimes fail to read the output of GNU CC version
1479: 2. If you have trouble, get GDB version 4.4 or later.
1480:
1481: @item
1482: @cindex DBX
1483: DBX rejects some files produced by GNU CC, though it accepts similar
1484: constructs in output from PCC. Until someone can supply a coherent
1485: description of what is valid DBX input and what is not, there is
1486: nothing I can do about these problems. You are on your own.
1487:
1488: @item
1489: The GNU assembler (GAS) does not support PIC. To generate PIC code, you
1490: must use some other assembler, such as @file{/bin/as}.
1491:
1492: @item
1493: On some BSD systems, including some versions of Ultrix, use of profiling
1494: causes static variable destructors (currently used only in C++) not to
1495: be run.
1496:
1497: @item
1498: Use of @samp{-I/usr/include} may cause trouble.
1499:
1500: Many systems come with header files that won't work with GNU CC unless
1501: corrected by @code{fixincludes}. The corrected header files go in a new
1502: directory; GNU CC searches this directory before @file{/usr/include}.
1503: If you use @samp{-I/usr/include}, this tells GNU CC to search
1504: @file{/usr/include} earlier on, before the corrected headers. The
1505: result is that you get the uncorrected header files.
1506:
1507: Instead, you should use these options (when compiling C programs):
1508:
1509: @smallexample
1510: -I/usr/local/lib/gcc-lib/@var{target}/@var{version}/include -I/usr/include
1511: @end smallexample
1512:
1513: For C++ programs, GNU CC also uses a special directory that defines C++
1514: interfaces to standard C subroutines. This directory is meant to be
1515: searched @emph{before} other standard include directories, so that it
1516: takes precedence. If you are compiling C++ programs and specifying
1517: include directories explicitly, use this option first, then the two
1518: options above:
1519:
1520: @example
1521: -I/usr/local/lib/g++-include
1522: @end example
1523:
1524: @ignore
1525: @cindex @code{vfork}, for the Sun-4
1526: @item
1527: There is a bug in @code{vfork} on the Sun-4 which causes the registers
1528: of the child process to clobber those of the parent. Because of this,
1529: programs that call @code{vfork} are likely to lose when compiled
1530: optimized with GNU CC when the child code alters registers which contain
1531: C variables in the parent. This affects variables which are live in the
1532: parent across the call to @code{vfork}.
1533:
1534: If you encounter this, you can work around the problem by declaring
1535: variables @code{volatile} in the function that calls @code{vfork}, until
1536: the problem goes away, or by not declaring them @code{register} and not
1537: using @samp{-O} for those source files.
1538: @end ignore
1539:
1540: @item
1541: On some SGI systems, when you use @samp{-lgl_s} as an option,
1542: it gets translated magically to @samp{-lgl_s -lX11_s -lc_s}.
1543: Naturally, this does not happen when you use GNU CC.
1544: You must specify all three options explicitly.
1545:
1546: @item
1547: On a Sparc, GNU CC aligns all values of type @code{double} on an 8-byte
1548: boundary, and it expects every @code{double} to be so aligned. The Sun
1549: compiler usually gives @code{double} values 8-byte alignment, with one
1550: exception: function arguments of type @code{double} may not be aligned.
1551:
1552: As a result, if a function compiled with Sun CC takes the address of an
1553: argument of type @code{double} and passes this pointer of type
1554: @code{double *} to a function compiled with GNU CC, dereferencing the
1555: pointer may cause a fatal signal.
1556:
1557: One way to solve this problem is to compile your entire program with GNU
1558: CC. Another solution is to modify the function that is compiled with
1559: Sun CC to copy the argument into a local variable; local variables
1560: are always properly aligned. A third solution is to modify the function
1561: that uses the pointer to dereference it via the following function
1562: @code{access_double} instead of directly with @samp{*}:
1563:
1564: @smallexample
1565: inline double
1566: access_double (double *unaligned_ptr)
1567: @{
1568: union d2i @{ double d; int i[2]; @};
1569:
1570: union d2i *p = (union d2i *) unaligned_ptr;
1571: union d2i u;
1572:
1573: u.i[0] = p->i[0];
1574: u.i[1] = p->i[1];
1575:
1576: return u.d;
1577: @}
1578: @end smallexample
1579:
1580: @noindent
1581: Storing into the pointer can be done likewise with the same union.
1582:
1583: @item
1584: On Solaris, the @code{malloc} function in the @file{libmalloc.a} library
1585: may allocate memory that is only 4 byte aligned. Since GNU CC on the
1586: Sparc assumes that doubles are 8 byte aligned, this may result in a
1587: fatal signal if doubles are stored in memory allocated by the
1588: @file{libmalloc.a} library.
1589:
1590: The solution is to not use the @file{libmalloc.a} library. Use instead
1591: @code{malloc} and related functions from @file{libc.a}; they do not have
1592: this problem.
1593:
1594: @item
1595: On a Sun, linking using GNU CC fails to find a shared library and
1596: reports that the library doesn't exist at all.
1597:
1598: This happens if you are using the GNU linker, because it does only
1599: static linking and looks only for unshared libraries. If you have a
1600: shared library with no unshared counterpart, the GNU linker won't find
1601: anything.
1602:
1603: We hope to make a linker which supports Sun shared libraries, but please
1604: don't ask when it will be finished---we don't know.
1605:
1606: @item
1607: Sun forgot to include a static version of @file{libdl.a} with some
1608: versions of SunOS (mainly 4.1). This results in undefined symbols when
1609: linking static binaries (that is, if you use @samp{-static}). If you
1610: see undefined symbols @code{_dlclose}, @code{_dlsym} or @code{_dlopen}
1611: when linking, compile and link against the file
1612: @file{mit/util/misc/dlsym.c} from the MIT version of X windows.
1613:
1614: @item
1615: The 128-bit long double format that the Sparc port supports currently
1616: works by using the architecturally defined quad-word floating point
1617: instructions. Since there is no hardware that supports these instructions
1618: they must be emulated by the operating system. Long doubles do not work
1619: in Sun OS versions 4.0.3 and earlier, because the kernel eumulator uses an
1620: obsolete and incompatible format. Long doubles do not work in Sun OS
1621: versions 4.1.1 to 4.1.3 because of emululator bugs that cause random
1622: unpredicatable failures. Long doubles appear to work in Sun OS 5.x
1623: (Solaris 2.x).
1624:
1625: A future implementation of the sparc long double support will use functions
1626: calls to library routines instead of the quad-word floating point instructions.
1627: This will allow long doubles to work in more situtations, since one can
1628: then substitute a working library if the kernel emulator is buggy.
1629:
1630: @item
1631: On HP-UX version 9.01 on the HP PA, the HP compiler @code{cc} does not
1632: compile GNU CC correctly. We do not yet know why. However, GNU CC
1633: compiled on earlier HP-UX versions works properly on HP-UX 9.01 and can
1634: compile itself properly on 9.01.
1635:
1636: @item
1637: On the HP PA machine, ADB sometimes fails to work on functions compiled
1638: with GNU CC. Specifically, it fails to work on functions that use
1639: @code{alloca} or variable-size arrays. This is because GNU CC doesn't
1640: generate HP-UX unwind descriptors for such functions. It may even be
1641: impossible to generate them.
1642:
1643: @item
1644: Debugging (@samp{-g}) is not supported on the HP PA machine, unless you use
1645: the preliminary GNU tools (@pxref{Installation}).
1646:
1647: @item
1648: Taking the address of a label may generate errors from the HP-UX
1649: PA assembler. GAS for the PA does not have this problem.
1650:
1651: @item
1652: Using floating point parameters for indirect calls to static functions
1653: will not work when using the HP assembler. There simply is no way for GCC
1654: to specify what registers hold arguments for static functions when using
1655: the HP assembler. GAS for the PA does not have this problem.
1656:
1657: @item
1658: For some very large functions you may receive errors from the HP linker
1659: complaining about an out of bounds unconditional branch offset. Fixing
1660: this problem correctly requires fixing problems in GNU CC and GAS. We
1661: hope to fix this in time for GNU CC 2.6. Until then you can work around
1662: by making your function smaller, and if you are using GAS, splitting the
1663: function into multiple source files may be necessary.
1664:
1665: @item
1666: GNU CC compiled code sometimes emits warnings from the HP-UX assembler of
1667: the form:
1668:
1669: @smallexample
1670: (warning) Use of GR3 when
1671: frame >= 8192 may cause conflict.
1672: @end smallexample
1673:
1674: These warnings are harmless and can be safely ignored.
1675:
1676: @item
1677: The current version of the assembler (@file{/bin/as}) for the RS/6000
1678: has certain problems that prevent the @samp{-g} option in GCC from
1679: working. Note that @file{Makefile.in} uses @samp{-g} by default when
1680: compiling @file{libgcc2.c}.
1681:
1682: IBM has produced a fixed version of the assembler. The upgraded
1683: assembler unfortunately was not included in any of the AIX 3.2 update
1684: PTF releases (3.2.2, 3.2.3, or 3.2.3e). Users of AIX 3.1 should request
1685: PTF U403044 from IBM and users of AIX 3.2 should request PTF U416277.
1686: See the file @file{README.RS6000} for more details on these updates.
1687:
1688: You can test for the presense of a fixed assembler by using the
1689: command
1690:
1691: @smallexample
1692: as -u < /dev/null
1693: @end smallexample
1694:
1695: @noindent
1696: If the command exits normally, the assembler fix already is installed.
1697: If the assembler complains that "-u" is an unknown flag, you need to
1698: order the fix.
1699:
1700: @item
1701: On the IBM RS/6000, compiling code of the form
1702:
1703: @smallexample
1704: extern int foo;
1705:
1706: @dots{} foo @dots{}
1707:
1708: static int foo;
1709: @end smallexample
1710:
1711: @noindent
1712: will cause the linker to report an undefined symbol @code{foo}.
1713: Although this behavior differs from most other systems, it is not a
1714: bug because redefining an @code{extern} variable as @code{static}
1715: is undefined in ANSI C.
1716:
1717: @item
1718: AIX on the RS/6000 provides support (NLS) for environments outside of
1719: the United States. Compilers and assemblers use NLS to support
1720: locale-specific representations of various objects including
1721: floating-point numbers ("." vs "," for separating decimal fractions).
1722: There have been problems reported where the library linked with GCC does
1723: not produce the same floating-point formats that the assembler accepts.
1724: If you have this problem, set the LANG environment variable to "C" or
1725: "En_US".
1726:
1727: @item
1728: On the RS/6000, XLC version 1.3.0.0 will miscompile @file{jump.c}.
1729: XLC version 1.3.0.1 or later fixes this problem. We do not yet
1730: have a PTF number for this fix.
1731:
1732: @item
1733: There is an assembler bug in versions of DG/UX prior to 5.4.2.01 that
1734: occurs when the @samp{fldcr} instruction is used. GNU CC uses
1735: @samp{fldcr} on the 88100 to serialize volatile memory references. Use
1736: the option @samp{-mno-serialize-volatile} if your version of the
1737: assembler has this bug.
1738:
1739: @item
1740: On VMS, GAS versions 1.38.1 and earlier may cause spurious warning
1741: messages from the linker. These warning messages complain of mismatched
1742: psect attributes. You can ignore them. @xref{VMS Install}.
1743:
1744: @item
1745: On NewsOS version 3, if you include both of the files @file{stddef.h}
1746: and @file{sys/types.h}, you get an error because there are two typedefs
1747: of @code{size_t}. You should change @file{sys/types.h} by adding these
1748: lines around the definition of @code{size_t}:
1749:
1750: @smallexample
1751: #ifndef _SIZE_T
1752: #define _SIZE_T
1753: @var{actual typedef here}
1754: #endif
1755: @end smallexample
1756:
1757: @cindex Alliant
1758: @item
1759: On the Alliant, the system's own convention for returning structures
1760: and unions is unusual, and is not compatible with GNU CC no matter
1761: what options are used.
1762:
1763: @cindex RT PC
1764: @cindex IBM RT PC
1765: @item
1766: On the IBM RT PC, the MetaWare HighC compiler (hc) uses a different
1767: convention for structure and union returning. Use the option
1768: @samp{-mhc-struct-return} to tell GNU CC to use a convention compatible
1769: with it.
1770:
1771: @cindex Vax calling convention
1772: @cindex Ultrix calling convention
1773: @item
1774: On Ultrix, the Fortran compiler expects registers 2 through 5 to be saved
1775: by function calls. However, the C compiler uses conventions compatible
1776: with BSD Unix: registers 2 through 5 may be clobbered by function calls.
1777:
1778: GNU CC uses the same convention as the Ultrix C compiler. You can use
1779: these options to produce code compatible with the Fortran compiler:
1780:
1781: @smallexample
1782: -fcall-saved-r2 -fcall-saved-r3 -fcall-saved-r4 -fcall-saved-r5
1783: @end smallexample
1784:
1785: @item
1786: On the WE32k, you may find that programs compiled with GNU CC do not
1787: work with the standard shared C ilbrary. You may need to link with
1788: the ordinary C compiler. If you do so, you must specify the following
1789: options:
1790:
1791: @smallexample
1792: -L/usr/local/lib/gcc-lib/we32k-att-sysv/2.5 -lgcc -lc_s
1793: @end smallexample
1794:
1795: The first specifies where to find the library @file{libgcc.a}
1796: specified with the @samp{-lgcc} option.
1797:
1798: GNU CC does linking by invoking @code{ld}, just as @code{cc} does, and
1799: there is no reason why it @emph{should} matter which compilation program
1800: you use to invoke @code{ld}. If someone tracks this problem down,
1801: it can probably be fixed easily.
1802:
1803: @item
1804: On the Alpha, you may get assembler errors about invalid syntax as a
1805: result of floating point constants. This is due to a bug in the C
1806: library functions @code{ecvt}, @code{fcvt} and @code{gcvt}. Given valid
1807: floating point numbers, they sometimes print @samp{NaN}.
1808:
1809: @item
1810: On Irix 4.0.5F (and perhaps in some other versions), an assembler bug
1811: sometimes reorders instructions incorrectly when optimization is turned
1812: on. If you think this may be happening to you, try using the GNU
1813: assembler; GAS version 2.1 supports ECOFF on Irix.
1814:
1815: Or use the @samp{-noasmopt} option when you compile GNU CC with itself,
1816: and then again when you compile your program. (This is a temporary
1817: kludge to turn off assembler optimization on Irix.) If this proves to
1818: be what you need, edit the assembler spec in the file @file{specs} so
1819: that it unconditionally passes @samp{-O0} to the assembler, and never
1820: passes @samp{-O2} or @samp{-O3}.
1821: @end itemize
1822:
1823: @node External Bugs
1824: @section Problems Compiling Certain Programs
1825:
1826: @itemize @bullet
1827: @item
1828: Parse errors may occur compiling X11 on a Decstation running Ultrix 4.2
1829: because of problems in DEC's versions of the X11 header files
1830: @file{X11/Xlib.h} and @file{X11/Xutil.h}. People recommend adding
1831: @samp{-I/usr/include/mit} to use the MIT versions of the header files,
1832: using the @samp{-traditional} switch to turn off ANSI C, or fixing the
1833: header files by adding this:
1834:
1835: @example
1836: #ifdef __STDC__
1837: #define NeedFunctionPrototypes 0
1838: #endif
1839: @end example
1840:
1841: @item
1842: If you have trouble compiling Perl on a SunOS 4 system, it may be
1843: because Perl specifies @samp{-I/usr/ucbinclude}. This accesses the
1844: unfixed header files. Perl specifies the options
1845:
1846: @example
1847: -traditional -Dvolatile=__volatile__
1848: -I/usr/include/sun -I/usr/ucbinclude
1849: -fpcc-struct-return
1850: @end example
1851:
1852: @noindent
1853: all of which are unnecessary with GCC 2.4.5 and newer versions. You can
1854: make a properly working Perl by setting @code{ccflags} and
1855: @code{cppflags} to empty values in @file{config.sh}, then typing
1856: @samp{./doSH; make depend; make}.
1857:
1858: @item
1859: On various 386 Unix systems derived from System V, including SCO, ISC,
1860: and ESIX, you may get error messages about running out of virtual memory
1861: while compiling certain programs.
1862:
1863: You can prevent this problem by linking GNU CC with the GNU malloc
1864: (which thus replaces the malloc that comes with the system). GNU malloc
1865: is available as a separate package, and also in the file
1866: @file{src/gmalloc.c} in the GNU Emacs 19 distribution.
1867:
1868: If you have installed GNU malloc as a separate library package, use this
1869: option when you relink GNU CC:
1870:
1871: @example
1872: MALLOC=/usr/local/lib/libgmalloc.a
1873: @end example
1874:
1875: Alternatively, if you have compiled @file{gmalloc.c} from Emacs 19, copy
1876: the object file to @file{gmalloc.o} and use this option when you relink
1877: GNU CC:
1878:
1879: @example
1880: MALLOC=gmalloc.o
1881: @end example
1882: @end itemize
1883:
1884: @node Incompatibilities
1885: @section Incompatibilities of GNU CC
1886: @cindex incompatibilities of GNU CC
1887:
1888: There are several noteworthy incompatibilities between GNU C and most
1889: existing (non-ANSI) versions of C. The @samp{-traditional} option
1890: eliminates many of these incompatibilities, @emph{but not all}, by
1891: telling GNU C to behave like the other C compilers.
1892:
1893: @itemize @bullet
1894: @cindex string constants
1895: @cindex read-only strings
1896: @cindex shared strings
1897: @item
1898: GNU CC normally makes string constants read-only. If several
1899: identical-looking string constants are used, GNU CC stores only one
1900: copy of the string.
1901:
1902: @cindex @code{mktemp}, and constant strings
1903: One consequence is that you cannot call @code{mktemp} with a string
1904: constant argument. The function @code{mktemp} always alters the
1905: string its argument points to.
1906:
1907: @cindex @code{sscanf}, and constant strings
1908: @cindex @code{fscanf}, and constant strings
1909: @cindex @code{scanf}, and constant strings
1910: Another consequence is that @code{sscanf} does not work on some systems
1911: when passed a string constant as its format control string or input.
1912: This is because @code{sscanf} incorrectly tries to write into the string
1913: constant. Likewise @code{fscanf} and @code{scanf}.
1914:
1915: The best solution to these problems is to change the program to use
1916: @code{char}-array variables with initialization strings for these
1917: purposes instead of string constants. But if this is not possible,
1918: you can use the @samp{-fwritable-strings} flag, which directs GNU CC
1919: to handle string constants the same way most C compilers do.
1920: @samp{-traditional} also has this effect, among others.
1921:
1922: @item
1923: @code{-2147483648} is positive.
1924:
1925: This is because 2147483648 cannot fit in the type @code{int}, so
1926: (following the ANSI C rules) its data type is @code{unsigned long int}.
1927: Negating this value yields 2147483648 again.
1928:
1929: @item
1930: GNU CC does not substitute macro arguments when they appear inside of
1931: string constants. For example, the following macro in GNU CC
1932:
1933: @example
1934: #define foo(a) "a"
1935: @end example
1936:
1937: @noindent
1938: will produce output @code{"a"} regardless of what the argument @var{a} is.
1939:
1940: The @samp{-traditional} option directs GNU CC to handle such cases
1941: (among others) in the old-fashioned (non-ANSI) fashion.
1942:
1943: @cindex @code{setjmp} incompatibilities
1944: @cindex @code{longjmp} incompatibilities
1945: @item
1946: When you use @code{setjmp} and @code{longjmp}, the only automatic
1947: variables guaranteed to remain valid are those declared
1948: @code{volatile}. This is a consequence of automatic register
1949: allocation. Consider this function:
1950:
1951: @example
1952: jmp_buf j;
1953:
1954: foo ()
1955: @{
1956: int a, b;
1957:
1958: a = fun1 ();
1959: if (setjmp (j))
1960: return a;
1961:
1962: a = fun2 ();
1963: /* @r{@code{longjmp (j)} may occur in @code{fun3}.} */
1964: return a + fun3 ();
1965: @}
1966: @end example
1967:
1968: Here @code{a} may or may not be restored to its first value when the
1969: @code{longjmp} occurs. If @code{a} is allocated in a register, then
1970: its first value is restored; otherwise, it keeps the last value stored
1971: in it.
1972:
1973: If you use the @samp{-W} option with the @samp{-O} option, you will
1974: get a warning when GNU CC thinks such a problem might be possible.
1975:
1976: The @samp{-traditional} option directs GNU C to put variables in
1977: the stack by default, rather than in registers, in functions that
1978: call @code{setjmp}. This results in the behavior found in
1979: traditional C compilers.
1980:
1981: @item
1982: Programs that use preprocessor directives in the middle of macro
1983: arguments do not work with GNU CC. For example, a program like this
1984: will not work:
1985:
1986: @example
1987: foobar (
1988: #define luser
1989: hack)
1990: @end example
1991:
1992: ANSI C does not permit such a construct. It would make sense to support
1993: it when @samp{-traditional} is used, but it is too much work to
1994: implement.
1995:
1996: @cindex external declaration scope
1997: @cindex scope of external declarations
1998: @cindex declaration scope
1999: @item
2000: Declarations of external variables and functions within a block apply
2001: only to the block containing the declaration. In other words, they
2002: have the same scope as any other declaration in the same place.
2003:
2004: In some other C compilers, a @code{extern} declaration affects all the
2005: rest of the file even if it happens within a block.
2006:
2007: The @samp{-traditional} option directs GNU C to treat all @code{extern}
2008: declarations as global, like traditional compilers.
2009:
2010: @item
2011: In traditional C, you can combine @code{long}, etc., with a typedef name,
2012: as shown here:
2013:
2014: @example
2015: typedef int foo;
2016: typedef long foo bar;
2017: @end example
2018:
2019: In ANSI C, this is not allowed: @code{long} and other type modifiers
2020: require an explicit @code{int}. Because this criterion is expressed
2021: by Bison grammar rules rather than C code, the @samp{-traditional}
2022: flag cannot alter it.
2023:
2024: @cindex typedef names as function parameters
2025: @item
2026: PCC allows typedef names to be used as function parameters. The
2027: difficulty described immediately above applies here too.
2028:
2029: @cindex whitespace
2030: @item
2031: PCC allows whitespace in the middle of compound assignment operators
2032: such as @samp{+=}. GNU CC, following the ANSI standard, does not
2033: allow this. The difficulty described immediately above applies here
2034: too.
2035:
2036: @cindex apostrophes
2037: @cindex '
2038: @item
2039: GNU CC complains about unterminated character constants inside of
2040: preprocessor conditionals that fail. Some programs have English
2041: comments enclosed in conditionals that are guaranteed to fail; if these
2042: comments contain apostrophes, GNU CC will probably report an error. For
2043: example, this code would produce an error:
2044:
2045: @example
2046: #if 0
2047: You can't expect this to work.
2048: #endif
2049: @end example
2050:
2051: The best solution to such a problem is to put the text into an actual
2052: C comment delimited by @samp{/*@dots{}*/}. However,
2053: @samp{-traditional} suppresses these error messages.
2054:
2055: @item
2056: Many user programs contain the declaration @samp{long time ();}. In the
2057: past, the system header files on many systems did not actually declare
2058: @code{time}, so it did not matter what type your program declared it to
2059: return. But in systems with ANSI C headers, @code{time} is declared to
2060: return @code{time_t}, and if that is not the same as @code{long}, then
2061: @samp{long time ();} is erroneous.
2062:
2063: The solution is to change your program to use @code{time_t} as the return
2064: type of @code{time}.
2065:
2066: @cindex @code{float} as function value type
2067: @item
2068: When compiling functions that return @code{float}, PCC converts it to
2069: a double. GNU CC actually returns a @code{float}. If you are concerned
2070: with PCC compatibility, you should declare your functions to return
2071: @code{double}; you might as well say what you mean.
2072:
2073: @cindex structures
2074: @cindex unions
2075: @item
2076: When compiling functions that return structures or unions, GNU CC
2077: output code normally uses a method different from that used on most
2078: versions of Unix. As a result, code compiled with GNU CC cannot call
2079: a structure-returning function compiled with PCC, and vice versa.
2080:
2081: The method used by GNU CC is as follows: a structure or union which is
2082: 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or union
2083: with any other size is stored into an address supplied by the caller
2084: (usually in a special, fixed register, but on some machines it is passed
2085: on the stack). The machine-description macros @code{STRUCT_VALUE} and
2086: @code{STRUCT_INCOMING_VALUE} tell GNU CC where to pass this address.
2087:
2088: By contrast, PCC on most target machines returns structures and unions
2089: of any size by copying the data into an area of static storage, and then
2090: returning the address of that storage as if it were a pointer value.
2091: The caller must copy the data from that memory area to the place where
2092: the value is wanted. GNU CC does not use this method because it is
2093: slower and nonreentrant.
2094:
2095: On some newer machines, PCC uses a reentrant convention for all
2096: structure and union returning. GNU CC on most of these machines uses a
2097: compatible convention when returning structures and unions in memory,
2098: but still returns small structures and unions in registers.
2099:
2100: You can tell GNU CC to use a compatible convention for all structure and
2101: union returning with the option @samp{-fpcc-struct-return}.
2102:
2103: @cindex preprocessing tokens
2104: @cindex preprocessing numbers
2105: @item
2106: GNU C complains about program fragments such as @samp{0x74ae-0x4000}
2107: which appear to be two hexadecimal constants separated by the minus
2108: operator. Actually, this string is a single @dfn{preprocessing token}.
2109: Each such token must correspond to one token in C. Since this does not,
2110: GNU C prints an error message. Although it may appear obvious that what
2111: is meant is an operator and two values, the ANSI C standard specifically
2112: requires that this be treated as erroneous.
2113:
2114: A @dfn{preprocessing token} is a @dfn{preprocessing number} if it
2115: begins with a digit and is followed by letters, underscores, digits,
2116: periods and @samp{e+}, @samp{e-}, @samp{E+}, or @samp{E-} character
2117: sequences.
2118:
2119: To make the above program fragment valid, place whitespace in front of
2120: the minus sign. This whitespace will end the preprocessing number.
2121: @end itemize
2122:
2123: @node Fixed Headers
2124: @section Fixed Header Files
2125:
2126: GNU CC needs to install corrected versions of some system header files.
2127: This is because most target systems have some header files that won't
2128: work with GNU CC unless they are changed. Some have bugs, some are
2129: incompatible with ANSI C, and some depend on special features of other
2130: compilers.
2131:
2132: Installing GNU CC automatically creates and installs the fixed header
2133: files, by running a program called @code{fixincludes} (or for certain
2134: targets an alternative such as @code{fixinc.svr4}). Normally, you
2135: don't need to pay attention to this. But there are cases where it
2136: doesn't do the right thing automatically.
2137:
2138: @itemize @bullet
2139: @item
2140: If you update the system's header files, such as by installing a new
2141: system version, the fixed header files of GNU CC are not automatically
2142: updated. The easiest way to update them is to reinstall GNU CC. (If
2143: you want to be clever, look in the makefile and you can find a
2144: shortcut.)
2145:
2146: @item
2147: On some systems, in particular SunOS 4, header file directories contain
2148: machine-specific symbolic links in certain places. This makes it
2149: possible to share most of the header files among hosts running the
2150: same version of SunOS 4 on different machine models.
2151:
2152: The programs that fix the header files do not understand this special
2153: way of using symbolic links; therefore, the directory of fixed header
2154: files is good only for the machine model used to build it.
2155:
2156: In SunOS 4, only programs that look inside the kernel will notice the
2157: difference between machine models. Therefore, for most purposes, you
2158: need not be concerned about this.
2159:
2160: It is possible to make separate sets of fixed header files for the
2161: different machine models, and arrange a structure of symbolic links so
2162: as to use the proper set, but you'll have to do this by hand.
2163:
2164: @item
2165: On Lynxos, GNU CC by default does not fix the header files. This is
2166: because bugs in the shell cause the @code{fixincludes} script to fail.
2167:
2168: This means you will encounter problems due to bugs in the system header
2169: files. It may be no comfort that they aren't GNU CC's fault, but it
2170: does mean that there's nothing for us to do about them.
2171: @end itemize
2172:
2173: @node Disappointments
2174: @section Disappointments and Misunderstandings
2175:
2176: These problems are perhaps regrettable, but we don't know any practical
2177: way around them.
2178:
2179: @itemize @bullet
2180: @item
2181: Certain local variables aren't recognized by debuggers when you compile
2182: with optimization.
2183:
2184: This occurs because sometimes GNU CC optimizes the variable out of
2185: existence. There is no way to tell the debugger how to compute the
2186: value such a variable ``would have had'', and it is not clear that would
2187: be desirable anyway. So GNU CC simply does not mention the eliminated
2188: variable when it writes debugging information.
2189:
2190: You have to expect a certain amount of disagreement between the
2191: executable and your source code, when you use optimization.
2192:
2193: @cindex conflicting types
2194: @cindex scope of declaration
2195: @item
2196: Users often think it is a bug when GNU CC reports an error for code
2197: like this:
2198:
2199: @example
2200: int foo (struct mumble *);
2201:
2202: struct mumble @{ @dots{} @};
2203:
2204: int foo (struct mumble *x)
2205: @{ @dots{} @}
2206: @end example
2207:
2208: This code really is erroneous, because the scope of @code{struct
2209: mumble} in the prototype is limited to the argument list containing it.
2210: It does not refer to the @code{struct mumble} defined with file scope
2211: immediately below---they are two unrelated types with similar names in
2212: different scopes.
2213:
2214: But in the definition of @code{foo}, the file-scope type is used
2215: because that is available to be inherited. Thus, the definition and
2216: the prototype do not match, and you get an error.
2217:
2218: This behavior may seem silly, but it's what the ANSI standard specifies.
2219: It is easy enough for you to make your code work by moving the
2220: definition of @code{struct mumble} above the prototype. It's not worth
2221: being incompatible with ANSI C just to avoid an error for the example
2222: shown above.
2223:
2224: @item
2225: Accesses to bitfields even in volatile objects works by accessing larger
2226: objects, such as a byte or a word. You cannot rely on what size of
2227: object is accessed in order to read or write the bitfield; it may even
2228: vary for a given bitfield according to the precise usage.
2229:
2230: If you care about controlling the amount of memory that is accessed, use
2231: volatile but do not use bitfields.
2232:
2233: @item
2234: GNU CC comes with shell scripts to fix certain known problems in system
2235: header files. They install corrected copies of various header files in
2236: a special directory where only GNU CC will normally look for them. The
2237: scripts adapt to various systems by searching all the system header
2238: files for the problem cases that we know about.
2239:
2240: If new system header files are installed, nothing automatically arranges
2241: to update the corrected header files. You will have to reinstall GNU CC
2242: to fix the new header files. More specifically, go to the build
2243: directory and delete the files @file{stmp-fixinc} and
2244: @file{stmp-headers}, and the subdirectory @code{include}; then do
2245: @samp{make install} again.
2246:
2247: @item
2248: On 68000 systems, you can get paradoxical results if you test the
2249: precise values of floating point numbers. For example, you can find
2250: that a floating point value which is not a NaN is not equal to itself.
2251: This results from the fact that the the floating point registers hold a
2252: few more bits of precision than fit in a @code{double} in memory.
2253: Compiled code moves values between memory and floating point registers
2254: at its convenience, and moving them into memory truncates them.
2255:
2256: You can partially avoid this problem by using the @samp{-ffloat-store}
2257: option (@pxref{Optimize Options}).
2258:
2259: @item
2260: On the MIPS, variable argument functions using @file{varargs.h}
2261: cannot have a floating point value for the first argument. The
2262: reason for this is that in the absence of a prototype in scope,
2263: if the first argument is a floating point, it is passed in a
2264: floating point register, rather than an integer register.
2265:
2266: If the code is rewritten to use the ANSI standard @file{stdarg.h}
2267: method of variable arguments, and the prototype is in scope at
2268: the time of the call, everything will work fine.
2269: @end itemize
2270:
2271: @node C++ Misunderstandings
2272: @section Common Misunderstandings with GNU C++
2273:
2274: @cindex misunderstandings in C++
2275: @cindex surprises in C++
2276: @cindex C++ misunderstandings
2277: C++ is a complex language and an evolving one, and its standard definition
2278: (the ANSI C++ draft standard) is also evolving. As a result,
2279: your C++ compiler may occasionally surprise you, even when its behavior is
2280: correct. This section discusses some areas that frequently give rise to
2281: questions of this sort.
2282:
2283: @menu
2284: * Static Definitions:: Static member declarations are not definitions
2285: * Temporaries:: Temporaries may vanish before you expect
2286: @end menu
2287:
2288: @node Static Definitions
2289: @subsection Declare @emph{and} Define Static Members
2290:
2291: @cindex C++ static data, declaring and defining
2292: @cindex static data in C++, declaring and defining
2293: @cindex declaring static data in C++
2294: @cindex defining static data in C++
2295: When a class has static data members, it is not enough to @emph{declare}
2296: the static member; you must also @emph{define} it. For example:
2297:
2298: @example
2299: class Foo
2300: @{
2301: @dots{}
2302: void method();
2303: static int bar;
2304: @};
2305: @end example
2306:
2307: This declaration only establishes that the class @code{Foo} has an
2308: @code{int} named @code{Foo::bar}, and a member function named
2309: @code{Foo::method}. But you still need to define @emph{both}
2310: @code{method} and @code{bar} elsewhere. According to the draft ANSI
2311: standard, you must supply an initializer in one (and only one) source
2312: file, such as:
2313:
2314: @example
2315: int Foo::bar = 0;
2316: @end example
2317:
2318: Other C++ compilers may not correctly implement the standard behavior.
2319: As a result, when you switch to @code{g++} from one of these compilers,
2320: you may discover that a program that appeared to work correctly in fact
2321: does not conform to the standard: @code{g++} reports as undefined
2322: symbols any static data members that lack definitions.
2323:
2324: @node Temporaries
2325: @subsection Temporaries May Vanish Before You Expect
2326:
2327: @cindex temporaries, lifetime of
2328: @cindex portions of temporary objects, pointers to
2329: It is dangerous to use pointers or references to @emph{portions} of a
2330: temporary object. The compiler may very well delete the object before
2331: you expect it to, leaving a pointer to garbage. The most common place
2332: where this problem crops up is in classes like the libg++
2333: @code{String} class, that define a conversion function to type
2334: @code{char *} or @code{const char *}. However, any class that returns
2335: a pointer to some internal structure is potentially subject to this
2336: problem.
2337:
2338: For example, a program may use a function @code{strfunc} that returns
2339: @code{String} objects, and another function @code{charfunc} that
2340: operates on pointers to @code{char}:
2341:
2342: @example
2343: String strfunc ();
2344: void charfunc (const char *);
2345: @end example
2346:
2347: @noindent
2348: In this situation, it may seem natural to write @w{@samp{charfunc
2349: (strfunc ());}} based on the knowledge that class @code{String} has an
2350: explicit conversion to @code{char} pointers. However, what really
2351: happens is akin to @samp{charfunc (@w{strfunc ()}.@w{convert ()});},
2352: where the @code{convert} method is a function to do the same data
2353: conversion normally performed by a cast. Since the last use of the
2354: temporary @code{String} object is the call to the conversion function,
2355: the compiler may delete that object before actually calling
2356: @code{charfunc}. The compiler has no way of knowing that deleting the
2357: @code{String} object will invalidate the pointer. The pointer then
2358: points to garbage, so that by the time @code{charfunc} is called, it
2359: gets an invalid argument.
2360:
2361: Code like this may run successfully under some other compilers,
2362: especially those that delete temporaries relatively late. However, the
2363: GNU C++ behavior is also standard-conformant, so if your program depends
2364: on late destruction of temporaries it is not portable.
2365:
2366: If you think this is surprising, you should be aware that the ANSI C++
2367: committee continues to debate the lifetime-of-temporaries problem.
2368:
2369: For now, at least, the safe way to write such code is to give the
2370: temporary a name, which forces it to remain until the end of the scope of
2371: the name. For example:
2372:
2373: @example
2374: String& tmp = strfunc ();
2375: charfunc (tmp);
2376: @end example
2377:
2378: @node Protoize Caveats
2379: @section Caveats of using @code{protoize}
2380:
2381: The conversion programs @code{protoize} and @code{unprotoize} can
2382: sometimes change a source file in a way that won't work unless you
2383: rearrange it.
2384:
2385: @itemize @bullet
2386: @item
2387: @code{protoize} can insert references to a type name or type tag before
2388: the definition, or in a file where they are not defined.
2389:
2390: If this happens, compiler error messages should show you where the new
2391: references are, so fixing the file by hand is straightforward.
2392:
2393: @item
2394: There are some C constructs which @code{protoize} cannot figure out.
2395: For example, it can't determine argument types for declaring a
2396: pointer-to-function variable; this you must do by hand. @code{protoize}
2397: inserts a comment containing @samp{???} each time it finds such a
2398: variable; so you can find all such variables by searching for this
2399: string. ANSI C does not require declaring the argument types of
2400: pointer-to-function types.
2401:
2402: @item
2403: Using @code{unprotoize} can easily introduce bugs. If the program
2404: relied on prototypes to bring about conversion of arguments, these
2405: conversions will not take place in the program without prototypes.
2406: One case in which you can be sure @code{unprotoize} is safe is when
2407: you are removing prototypes that were made with @code{protoize}; if
2408: the program worked before without any prototypes, it will work again
2409: without them.
2410:
2411: You can find all the places where this problem might occur by compiling
2412: the program with the @samp{-Wconversion} option. It prints a warning
2413: whenever an argument is converted.
2414:
2415: @item
2416: Both conversion programs can be confused if there are macro calls in and
2417: around the text to be converted. In other words, the standard syntax
2418: for a declaration or definition must not result from expanding a macro.
2419: This problem is inherent in the design of C and cannot be fixed. If
2420: only a few functions have confusing macro calls, you can easily convert
2421: them manually.
2422:
2423: @item
2424: @code{protoize} cannot get the argument types for a function whose
2425: definition was not actually compiled due to preprocessor conditionals.
2426: When this happens, @code{protoize} changes nothing in regard to such
2427: a function. @code{protoize} tries to detect such instances and warn
2428: about them.
2429:
2430: You can generally work around this problem by using @code{protoize} step
2431: by step, each time specifying a different set of @samp{-D} options for
2432: compilation, until all of the functions have been converted. There is
2433: no automatic way to verify that you have got them all, however.
2434:
2435: @item
2436: Confusion may result if there is an occasion to convert a function
2437: declaration or definition in a region of source code where there is more
2438: than one formal parameter list present. Thus, attempts to convert code
2439: containing multiple (conditionally compiled) versions of a single
2440: function header (in the same vicinity) may not produce the desired (or
2441: expected) results.
2442:
2443: If you plan on converting source files which contain such code, it is
2444: recommended that you first make sure that each conditionally compiled
2445: region of source code which contains an alternative function header also
2446: contains at least one additional follower token (past the final right
2447: parenthesis of the function header). This should circumvent the
2448: problem.
2449:
2450: @item
2451: @code{unprotoize} can become confused when trying to convert a function
2452: definition or declaration which contains a declaration for a
2453: pointer-to-function formal argument which has the same name as the
2454: function being defined or declared. We recommand you avoid such choices
2455: of formal parameter names.
2456:
2457: @item
2458: You might also want to correct some of the indentation by hand and break
2459: long lines. (The conversion programs don't write lines longer than
2460: eighty characters in any case.)
2461: @end itemize
2462:
2463: @node Non-bugs
2464: @section Certain Changes We Don't Want to Make
2465:
2466: This section lists changes that people frequently request, but which
2467: we do not make because we think GNU CC is better without them.
2468:
2469: @itemize @bullet
2470: @item
2471: Checking the number and type of arguments to a function which has an
2472: old-fashioned definition and no prototype.
2473:
2474: Such a feature would work only occasionally---only for calls that appear
2475: in the same file as the called function, following the definition. The
2476: only way to check all calls reliably is to add a prototype for the
2477: function. But adding a prototype eliminates the motivation for this
2478: feature. So the feature is not worthwhile.
2479:
2480: @item
2481: Warning about using an expression whose type is signed as a shift count.
2482:
2483: Shift count operands are probably signed more often than unsigned.
2484: Warning about this would cause far more annoyance than good.
2485:
2486: @item
2487: Warning about assigning a signed value to an unsigned variable.
2488:
2489: Such assignments must be very common; warning about them would cause
2490: more annoyance than good.
2491:
2492: @item
2493: Warning about unreachable code.
2494:
2495: It's very common to have unreachable code in machine-generated
2496: programs. For example, this happens normally in some files of GNU C
2497: itself.
2498:
2499: @item
2500: Warning when a non-void function value is ignored.
2501:
2502: Coming as I do from a Lisp background, I balk at the idea that there is
2503: something dangerous about discarding a value. There are functions that
2504: return values which some callers may find useful; it makes no sense to
2505: clutter the program with a cast to @code{void} whenever the value isn't
2506: useful.
2507:
2508: @item
2509: Assuming (for optimization) that the address of an external symbol is
2510: never zero.
2511:
2512: This assumption is false on certain systems when @samp{#pragma weak} is
2513: used.
2514:
2515: @item
2516: Making @samp{-fshort-enums} the default.
2517:
2518: This would cause storage layout to be incompatible with most other C
2519: compilers. And it doesn't seem very important, given that you can get
2520: the same result in other ways. The case where it matters most is when
2521: the enumeration-valued object is inside a structure, and in that case
2522: you can specify a field width explicitly.
2523:
2524: @item
2525: Making bitfields unsigned by default on particular machines where ``the
2526: ABI standard'' says to do so.
2527:
2528: The ANSI C standard leaves it up to the implementation whether a bitfield
2529: declared plain @code{int} is signed or not. This in effect creates two
2530: alternative dialects of C.
2531:
2532: The GNU C compiler supports both dialects; you can specify the signed
2533: dialect with @samp{-fsigned-bitfields} and the unsigned dialect with
2534: @samp{-funsigned-bitfields}. However, this leaves open the question of
2535: which dialect to use by default.
2536:
2537: Currently, the preferred dialect makes plain bitfields signed, because
2538: this is simplest. Since @code{int} is the same as @code{signed int} in
2539: every other context, it is cleanest for them to be the same in bitfields
2540: as well.
2541:
2542: Some computer manufacturers have published Application Binary Interface
2543: standards which specify that plain bitfields should be unsigned. It is
2544: a mistake, however, to say anything about this issue in an ABI. This is
2545: because the handling of plain bitfields distinguishes two dialects of C.
2546: Both dialects are meaningful on every type of machine. Whether a
2547: particular object file was compiled using signed bitfields or unsigned
2548: is of no concern to other object files, even if they access the same
2549: bitfields in the same data structures.
2550:
2551: A given program is written in one or the other of these two dialects.
2552: The program stands a chance to work on most any machine if it is
2553: compiled with the proper dialect. It is unlikely to work at all if
2554: compiled with the wrong dialect.
2555:
2556: Many users appreciate the GNU C compiler because it provides an
2557: environment that is uniform across machines. These users would be
2558: inconvenienced if the compiler treated plain bitfields differently on
2559: certain machines.
2560:
2561: Occasionally users write programs intended only for a particular machine
2562: type. On these occasions, the users would benefit if the GNU C compiler
2563: were to support by default the same dialect as the other compilers on
2564: that machine. But such applications are rare. And users writing a
2565: program to run on more than one type of machine cannot possibly benefit
2566: from this kind of compatibility.
2567:
2568: This is why GNU CC does and will treat plain bitfields in the same
2569: fashion on all types of machines (by default).
2570:
2571: There are some arguments for making bitfields unsigned by default on all
2572: machines. If, for example, this becomes a universal de facto standard,
2573: it would make sense for GNU CC to go along with it. This is something
2574: to be considered in the future.
2575:
2576: (Of course, users strongly concerned about portability should indicate
2577: explicitly in each bitfield whether it is signed or not. In this way,
2578: they write programs which have the same meaning in both C dialects.)
2579:
2580: @item
2581: Undefining @code{__STDC__} when @samp{-ansi} is not used.
2582:
2583: Currently, GNU CC defines @code{__STDC__} as long as you don't use
2584: @samp{-traditional}. This provides good results in practice.
2585:
2586: Programmers normally use conditionals on @code{__STDC__} to ask whether
2587: it is safe to use certain features of ANSI C, such as function
2588: prototypes or ANSI token concatenation. Since plain @samp{gcc} supports
2589: all the features of ANSI C, the correct answer to these questions is
2590: ``yes''.
2591:
2592: Some users try to use @code{__STDC__} to check for the availability of
2593: certain library facilities. This is actually incorrect usage in an ANSI
2594: C program, because the ANSI C standard says that a conforming
2595: freestanding implementation should define @code{__STDC__} even though it
2596: does not have the library facilities. @samp{gcc -ansi -pedantic} is a
2597: conforming freestanding implementation, and it is therefore required to
2598: define @code{__STDC__}, even though it does not come with an ANSI C
2599: library.
2600:
2601: Sometimes people say that defining @code{__STDC__} in a compiler that
2602: does not completely conform to the ANSI C standard somehow violates the
2603: standard. This is illogical. The standard is a standard for compilers
2604: that claim to support ANSI C, such as @samp{gcc -ansi}---not for other
2605: compilers such as plain @samp{gcc}. Whatever the ANSI C standard says
2606: is relevant to the design of plain @samp{gcc} without @samp{-ansi} only
2607: for pragmatic reasons, not as a requirement.
2608:
2609: @item
2610: Undefining @code{__STDC__} in C++.
2611:
2612: Programs written to compile with C++-to-C translators get the
2613: value of @code{__STDC__} that goes with the C compiler that is
2614: subsequently used. These programs must test @code{__STDC__}
2615: to determine what kind of C preprocessor that compiler uses:
2616: whether they should concatenate tokens in the ANSI C fashion
2617: or in the traditional fashion.
2618:
2619: These programs work properly with GNU C++ if @code{__STDC__} is defined.
2620: They would not work otherwise.
2621:
2622: In addition, many header files are written to provide prototypes in ANSI
2623: C but not in traditional C. Many of these header files can work without
2624: change in C++ provided @code{__STDC__} is defined. If @code{__STDC__}
2625: is not defined, they will all fail, and will all need to be changed to
2626: test explicitly for C++ as well.
2627:
2628: @item
2629: Deleting ``empty'' loops.
2630:
2631: GNU CC does not delete ``empty'' loops because the most likely reason
2632: you would put one in a program is to have a delay. Deleting them will
2633: not make real programs run any faster, so it would be pointless.
2634:
2635: It would be different if optimization of a nonempty loop could produce
2636: an empty one. But this generally can't happen.
2637:
2638: @item
2639: Making side effects happen in the same order as in some other compiler.
2640:
2641: @cindex side effects, order of evaluation
2642: @cindex order of evaluation, side effects
2643: It is never safe to depend on the order of evaluation of side effects.
2644: For example, a function call like this may very well behave differently
2645: from one compiler to another:
2646:
2647: @example
2648: void func (int, int);
2649:
2650: int i = 2;
2651: func (i++, i++);
2652: @end example
2653:
2654: There is no guarantee (in either the C or the C++ standard language
2655: definitions) that the increments will be evaluated in any particular
2656: order. Either increment might happen first. @code{func} might get the
2657: arguments @samp{3, 4}, or it might get @samp{4, 3}, or even @samp{3, 3}.
2658:
2659: @item
2660: Using the ``canonical'' form of the target configuration name as the
2661: directory for installation.
2662:
2663: This would be an improvement in some respects, but it would also cause
2664: problems. For one thing, users might expect to use in the @samp{-b}
2665: option the same name specified at installation; if installation used the
2666: canonical form, that would not work. What's more, the canonical name
2667: might be too long for certain file systems.
2668:
2669: We suggest you make a link to the installation directory under the
2670: canonical name, if you want to use that name in the @samp{-b} option.
2671: @end itemize
2672:
2673: @node Warnings and Errors
2674: @section Warning Messages and Error Messages
2675:
2676: @cindex error messages
2677: @cindex warnings vs errors
2678: @cindex messages, warning and error
2679: The GNU compiler can produce two kinds of diagnostics: errors and
2680: warnings. Each kind has a different purpose:
2681:
2682: @itemize @w{}
2683: @item
2684: @emph{Errors} report problems that make it impossible to compile your
2685: program. GNU CC reports errors with the source file name and line
2686: number where the problem is apparent.
2687:
2688: @item
2689: @emph{Warnings} report other unusual conditions in your code that
2690: @emph{may} indicate a problem, although compilation can (and does)
2691: proceed. Warning messages also report the source file name and line
2692: number, but include the text @samp{warning:} to distinguish them
2693: from error messages.
2694: @end itemize
2695:
2696: Warnings may indicate danger points where you should check to make sure
2697: that your program really does what you intend; or the use of obsolete
2698: features; or the use of nonstandard features of GNU C or C++. Many
2699: warnings are issued only if you ask for them, with one of the @samp{-W}
2700: options (for instance, @samp{-Wall} requests a variety of useful
2701: warnings).
2702:
2703: GNU CC always tries to compile your program if possible; it never
2704: gratuituously rejects a program whose meaning is clear merely because
2705: (for instance) it fails to conform to a standard. In some cases,
2706: however, the C and C++ standards specify that certain extensions are
2707: forbidden, and a diagnostic @emph{must} be issued by a conforming
2708: compiler. The @samp{-pedantic} option tells GNU CC to issue warnings in
2709: such cases; @samp{-pedantic-errors} says to make them errors instead.
2710: This does not mean that @emph{all} non-ANSI constructs get warnings
2711: or errors.
2712:
2713: @xref{Warning Options,,Options to Request or Suppress Warnings}, for
2714: more detail on these and related command-line options.
2715:
2716: @node Bugs
2717: @chapter Reporting Bugs
2718: @cindex bugs
2719: @cindex reporting bugs
2720:
2721: Your bug reports play an essential role in making GNU CC reliable.
2722:
2723: When you encounter a problem, the first thing to do is to see if it is
2724: already known. @xref{Trouble}. If it isn't known, then you should
2725: report the problem.
2726:
2727: Reporting a bug may help you by bringing a solution to your problem, or
2728: it may not. (If it does not, look in the service directory; see
2729: @ref{Service}.) In any case, the principal function of a bug report is
2730: to help the entire community by making the next version of GNU CC work
2731: better. Bug reports are your contribution to the maintenance of GNU CC.
2732:
2733: Since the maintainers are very overloaded, we cannot respond to every
2734: bug report. However, if the bug has not been fixed, we are likely to
2735: send you a patch and ask you to tell us whether it works.
2736:
2737: In order for a bug report to serve its purpose, you must include the
2738: information that makes for fixing the bug.
2739:
2740: @menu
2741: * Criteria: Bug Criteria. Have you really found a bug?
2742: * Where: Bug Lists. Where to send your bug report.
2743: * Reporting: Bug Reporting. How to report a bug effectively.
2744: * Patches: Sending Patches. How to send a patch for GNU CC.
2745: * Known: Trouble. Known problems.
2746: * Help: Service. Where to ask for help.
2747: @end menu
2748:
2749: @node Bug Criteria
2750: @section Have You Found a Bug?
2751: @cindex bug criteria
2752:
2753: If you are not sure whether you have found a bug, here are some guidelines:
2754:
2755: @itemize @bullet
2756: @cindex fatal signal
2757: @cindex core dump
2758: @item
2759: If the compiler gets a fatal signal, for any input whatever, that is a
2760: compiler bug. Reliable compilers never crash.
2761:
2762: @cindex invalid assembly code
2763: @cindex assembly code, invalid
2764: @item
2765: If the compiler produces invalid assembly code, for any input whatever
2766: (except an @code{asm} statement), that is a compiler bug, unless the
2767: compiler reports errors (not just warnings) which would ordinarily
2768: prevent the assembler from being run.
2769:
2770: @cindex undefined behavior
2771: @cindex undefined function value
2772: @cindex increment operators
2773: @item
2774: If the compiler produces valid assembly code that does not correctly
2775: execute the input source code, that is a compiler bug.
2776:
2777: However, you must double-check to make sure, because you may have run
2778: into an incompatibility between GNU C and traditional C
2779: (@pxref{Incompatibilities}). These incompatibilities might be considered
2780: bugs, but they are inescapable consequences of valuable features.
2781:
2782: Or you may have a program whose behavior is undefined, which happened
2783: by chance to give the desired results with another C or C++ compiler.
2784:
2785: For example, in many nonoptimizing compilers, you can write @samp{x;}
2786: at the end of a function instead of @samp{return x;}, with the same
2787: results. But the value of the function is undefined if @code{return}
2788: is omitted; it is not a bug when GNU CC produces different results.
2789:
2790: Problems often result from expressions with two increment operators,
2791: as in @code{f (*p++, *p++)}. Your previous compiler might have
2792: interpreted that expression the way you intended; GNU CC might
2793: interpret it another way. Neither compiler is wrong. The bug is
2794: in your code.
2795:
2796: After you have localized the error to a single source line, it should
2797: be easy to check for these things. If your program is correct and
2798: well defined, you have found a compiler bug.
2799:
2800: @item
2801: If the compiler produces an error message for valid input, that is a
2802: compiler bug.
2803:
2804: @cindex invalid input
2805: @item
2806: If the compiler does not produce an error message for invalid input,
2807: that is a compiler bug. However, you should note that your idea of
2808: ``invalid input'' might be my idea of ``an extension'' or ``support
2809: for traditional practice''.
2810:
2811: @item
2812: If you are an experienced user of C or C++ compilers, your suggestions
2813: for improvement of GNU CC or GNU C++ are welcome in any case.
2814: @end itemize
2815:
2816: @node Bug Lists
2817: @section Where to Report Bugs
2818: @cindex bug report mailing lists
2819:
2820: @kindex bug-gcc@@prep.ai.mit.edu
2821: Send bug reports for GNU C to one of these addresses:
2822:
2823: @example
2824: bug-gcc@@prep.ai.mit.edu
2825: @{ucbvax|mit-eddie|uunet@}!prep.ai.mit.edu!bug-gcc
2826: @end example
2827:
2828: @kindex bug-g++@@prep.ai.mit.edu
2829: Send bug reports for GNU C++ to one of these addresses:
2830:
2831: @example
2832: bug-g++@@prep.ai.mit.edu
2833: @{ucbvax|mit-eddie|uunet@}!prep.ai.mit.edu!bug-g++
2834: @end example
2835:
2836: @kindex bug-libg++@@prep.ai.mit.edu
2837: If your bug involves the C++ class library libg++, send mail to
2838: @samp{bug-lib-g++@@prep.ai.mit.edu}. If you're not sure, you can send
2839: the bug report to both lists.
2840:
2841: @strong{Do not send bug reports to the mailing list @samp{help-gcc}, or
2842: to the newsgroup @samp{gnu.gcc.help}.} Most users of GNU CC do not want
2843: to receive bug reports. Those that do, have asked to be on
2844: @samp{bug-gcc} and/or @samp{bug-g++}.
2845:
2846: The mailing lists @samp{bug-gcc} and @samp{bug-g++} both have newsgroups
2847: which serve as repeaters: @samp{gnu.gcc.bug} and @samp{gnu.g++.bug}.
2848: Each mailing list and its newsgroup carry exactly the same messages.
2849:
2850: Often people think of posting bug reports to the newsgroup instead of
2851: mailing them. This appears to work, but it has one problem which can be
2852: crucial: a newsgroup posting does not contain a mail path back to the
2853: sender. Thus, if maintainers need more information, they may be unable
2854: to reach you. For this reason, you should always send bug reports by
2855: mail to the proper mailing list.
2856:
2857: As a last resort, send bug reports on paper to:
2858:
2859: @example
2860: GNU Compiler Bugs
2861: Free Software Foundation
2862: 675 Mass Ave
2863: Cambridge, MA 02139
2864: @end example
2865:
2866: @node Bug Reporting
2867: @section How to Report Bugs
2868: @cindex compiler bugs, reporting
2869:
2870: The fundamental principle of reporting bugs usefully is this:
2871: @strong{report all the facts}. If you are not sure whether to state a
2872: fact or leave it out, state it!
2873:
2874: Often people omit facts because they think they know what causes the
2875: problem and they conclude that some details don't matter. Thus, you might
2876: assume that the name of the variable you use in an example does not matter.
2877: Well, probably it doesn't, but one cannot be sure. Perhaps the bug is a
2878: stray memory reference which happens to fetch from the location where that
2879: name is stored in memory; perhaps, if the name were different, the contents
2880: of that location would fool the compiler into doing the right thing despite
2881: the bug. Play it safe and give a specific, complete example. That is the
2882: easiest thing for you to do, and the most helpful.
2883:
2884: Keep in mind that the purpose of a bug report is to enable someone to
2885: fix the bug if it is not known. It isn't very important what happens if
2886: the bug is already known. Therefore, always write your bug reports on
2887: the assumption that the bug is not known.
2888:
2889: Sometimes people give a few sketchy facts and ask, ``Does this ring a
2890: bell?'' This cannot help us fix a bug, so it is basically useless. We
2891: respond by asking for enough details to enable us to investigate.
2892: You might as well expedite matters by sending them to begin with.
2893:
2894: Try to make your bug report self-contained. If we have to ask you for
2895: more information, it is best if you include all the previous information
2896: in your response, as well as the information that was missing.
2897:
2898: To enable someone to investigate the bug, you should include all these
2899: things:
2900:
2901: @itemize @bullet
2902: @item
2903: The version of GNU CC. You can get this by running it with the
2904: @samp{-v} option.
2905:
2906: Without this, we won't know whether there is any point in looking for
2907: the bug in the current version of GNU CC.
2908:
2909: @item
2910: A complete input file that will reproduce the bug. If the bug is in the
2911: C preprocessor, send a source file and any header files that it
2912: requires. If the bug is in the compiler proper (@file{cc1}), run your
2913: source file through the C preprocessor by doing @samp{gcc -E
2914: @var{sourcefile} > @var{outfile}}, then include the contents of
2915: @var{outfile} in the bug report. (When you do this, use the same
2916: @samp{-I}, @samp{-D} or @samp{-U} options that you used in actual
2917: compilation.)
2918:
2919: A single statement is not enough of an example. In order to compile it,
2920: it must be embedded in a complete file of compiler input; and the bug
2921: might depend on the details of how this is done.
2922:
2923: Without a real example one can compile, all anyone can do about your bug
2924: report is wish you luck. It would be futile to try to guess how to
2925: provoke the bug. For example, bugs in register allocation and reloading
2926: frequently depend on every little detail of the function they happen in.
2927:
2928: Even if the input file that fails comes from a GNU program, you should
2929: still send the complete test case. Don't ask the GNU CC maintainers to
2930: do the extra work of obtaining the program in question---they are all
2931: overworked as it is. Also, the problem may depend on what is in the
2932: header files on your system; it is unreliable for the GNU CC maintainers
2933: to try the problem with the header files available to them. By sending
2934: CPP output, you can eliminate this source of uncertainty and save us
2935: a certain percentage of wild goose chases.
2936:
2937: @item
2938: The command arguments you gave GNU CC or GNU C++ to compile that example
2939: and observe the bug. For example, did you use @samp{-O}? To guarantee
2940: you won't omit something important, list all the options.
2941:
2942: If we were to try to guess the arguments, we would probably guess wrong
2943: and then we would not encounter the bug.
2944:
2945: @item
2946: The type of machine you are using, and the operating system name and
2947: version number.
2948:
2949: @item
2950: The operands you gave to the @code{configure} command when you installed
2951: the compiler.
2952:
2953: @item
2954: A complete list of any modifications you have made to the compiler
2955: source. (We don't promise to investigate the bug unless it happens in
2956: an unmodified compiler. But if you've made modifications and don't tell
2957: us, then you are sending us on a wild goose chase.)
2958:
2959: Be precise about these changes. A description in English is not
2960: enough---send a context diff for them.
2961:
2962: Adding files of your own (such as a machine description for a machine we
2963: don't support) is a modification of the compiler source.
2964:
2965: @item
2966: Details of any other deviations from the standard procedure for installing
2967: GNU CC.
2968:
2969: @item
2970: A description of what behavior you observe that you believe is
2971: incorrect. For example, ``The compiler gets a fatal signal,'' or,
2972: ``The assembler instruction at line 208 in the output is incorrect.''
2973:
2974: Of course, if the bug is that the compiler gets a fatal signal, then one
2975: can't miss it. But if the bug is incorrect output, the maintainer might
2976: not notice unless it is glaringly wrong. None of us has time to study
2977: all the assembler code from a 50-line C program just on the chance that
2978: one instruction might be wrong. We need @emph{you} to do this part!
2979:
2980: Even if the problem you experience is a fatal signal, you should still
2981: say so explicitly. Suppose something strange is going on, such as, your
2982: copy of the compiler is out of synch, or you have encountered a bug in
2983: the C library on your system. (This has happened!) Your copy might
2984: crash and the copy here would not. If you @i{said} to expect a crash,
2985: then when the compiler here fails to crash, we would know that the bug
2986: was not happening. If you don't say to expect a crash, then we would
2987: not know whether the bug was happening. We would not be able to draw
2988: any conclusion from our observations.
2989:
2990: If the problem is a diagnostic when compiling GNU CC with some other
2991: compiler, say whether it is a warning or an error.
2992:
2993: Often the observed symptom is incorrect output when your program is run.
2994: Sad to say, this is not enough information unless the program is short
2995: and simple. None of us has time to study a large program to figure out
2996: how it would work if compiled correctly, much less which line of it was
2997: compiled wrong. So you will have to do that. Tell us which source line
2998: it is, and what incorrect result happens when that line is executed. A
2999: person who understands the program can find this as easily as finding a
3000: bug in the program itself.
3001:
3002: @item
3003: If you send examples of assembler code output from GNU CC or GNU C++,
3004: please use @samp{-g} when you make them. The debugging information
3005: includes source line numbers which are essential for correlating the
3006: output with the input.
3007:
3008: @item
3009: If you wish to mention something in the GNU CC source, refer to it by
3010: context, not by line number.
3011:
3012: The line numbers in the development sources don't match those in your
3013: sources. Your line numbers would convey no useful information to the
3014: maintainers.
3015:
3016: @item
3017: Additional information from a debugger might enable someone to find a
3018: problem on a machine which he does not have available. However, you
3019: need to think when you collect this information if you want it to have
3020: any chance of being useful.
3021:
3022: @cindex backtrace for bug reports
3023: For example, many people send just a backtrace, but that is never
3024: useful by itself. A simple backtrace with arguments conveys little
3025: about GNU CC because the compiler is largely data-driven; the same
3026: functions are called over and over for different RTL insns, doing
3027: different things depending on the details of the insn.
3028:
3029: Most of the arguments listed in the backtrace are useless because they
3030: are pointers to RTL list structure. The numeric values of the
3031: pointers, which the debugger prints in the backtrace, have no
3032: significance whatever; all that matters is the contents of the objects
3033: they point to (and most of the contents are other such pointers).
3034:
3035: In addition, most compiler passes consist of one or more loops that
3036: scan the RTL insn sequence. The most vital piece of information about
3037: such a loop---which insn it has reached---is usually in a local variable,
3038: not in an argument.
3039:
3040: @findex debug_rtx
3041: What you need to provide in addition to a backtrace are the values of
3042: the local variables for several stack frames up. When a local
3043: variable or an argument is an RTX, first print its value and then use
3044: the GDB command @code{pr} to print the RTL expression that it points
3045: to. (If GDB doesn't run on your machine, use your debugger to call
3046: the function @code{debug_rtx} with the RTX as an argument.) In
3047: general, whenever a variable is a pointer, its value is no use
3048: without the data it points to.
3049: @end itemize
3050:
3051: Here are some things that are not necessary:
3052:
3053: @itemize @bullet
3054: @item
3055: A description of the envelope of the bug.
3056:
3057: Often people who encounter a bug spend a lot of time investigating
3058: which changes to the input file will make the bug go away and which
3059: changes will not affect it.
3060:
3061: This is often time consuming and not very useful, because the way we
3062: will find the bug is by running a single example under the debugger with
3063: breakpoints, not by pure deduction from a series of examples. You might
3064: as well save your time for something else.
3065:
3066: Of course, if you can find a simpler example to report @emph{instead} of
3067: the original one, that is a convenience. Errors in the output will be
3068: easier to spot, running under the debugger will take less time, etc.
3069: Most GNU CC bugs involve just one function, so the most straightforward
3070: way to simplify an example is to delete all the function definitions
3071: except the one where the bug occurs. Those earlier in the file may be
3072: replaced by external declarations if the crucial function depends on
3073: them. (Exception: inline functions may affect compilation of functions
3074: defined later in the file.)
3075:
3076: However, simplification is not vital; if you don't want to do this,
3077: report the bug anyway and send the entire test case you used.
3078:
3079: @item
3080: In particular, some people insert conditionals @samp{#ifdef BUG} around
3081: a statement which, if removed, makes the bug not happen. These are just
3082: clutter; we won't pay any attention to them anyway. Besides, you should
3083: send us cpp output, and that can't have conditionals.
3084:
3085: @item
3086: A patch for the bug.
3087:
3088: A patch for the bug is useful if it is a good one. But don't omit the
3089: necessary information, such as the test case, on the assumption that a
3090: patch is all we need. We might see problems with your patch and decide
3091: to fix the problem another way, or we might not understand it at all.
3092:
3093: Sometimes with a program as complicated as GNU CC it is very hard to
3094: construct an example that will make the program follow a certain path
3095: through the code. If you don't send the example, we won't be able to
3096: construct one, so we won't be able to verify that the bug is fixed.
3097:
3098: And if we can't understand what bug you are trying to fix, or why your
3099: patch should be an improvement, we won't install it. A test case will
3100: help us to understand.
3101:
3102: @xref{Sending Patches}, for guidelines on how to make it easy for us to
3103: understand and install your patches.
3104:
3105: @item
3106: A guess about what the bug is or what it depends on.
3107:
3108: Such guesses are usually wrong. Even I can't guess right about such
3109: things without first using the debugger to find the facts.
3110:
3111: @item
3112: A core dump file.
3113:
3114: We have no way of examining a core dump for your type of machine
3115: unless we have an identical system---and if we do have one,
3116: we should be able to reproduce the crash ourselves.
3117: @end itemize
3118:
3119: @node Sending Patches,, Bug Reporting, Bugs
3120: @section Sending Patches for GNU CC
3121:
3122: If you would like to write bug fixes or improvements for the GNU C
3123: compiler, that is very helpful. When you send your changes, please
3124: follow these guidelines to avoid causing extra work for us in studying
3125: the patches.
3126:
3127: If you don't follow these guidelines, your information might still be
3128: useful, but using it will take extra work. Maintaining GNU C is a lot
3129: of work in the best of circumstances, and we can't keep up unless you do
3130: your best to help.
3131:
3132: @itemize @bullet
3133: @item
3134: Send an explanation with your changes of what problem they fix or what
3135: improvement they bring about. For a bug fix, just include a copy of the
3136: bug report, and explain why the change fixes the bug.
3137:
3138: (Referring to a bug report is not as good as including it, because then
3139: we will have to look it up, and we have probably already deleted it if
3140: we've already fixed the bug.)
3141:
3142: @item
3143: Always include a proper bug report for the problem you think you have
3144: fixed. We need to convince ourselves that the change is right before
3145: installing it. Even if it is right, we might have trouble judging it if
3146: we don't have a way to reproduce the problem.
3147:
3148: @item
3149: Include all the comments that are appropriate to help people reading the
3150: source in the future understand why this change was needed.
3151:
3152: @item
3153: Don't mix together changes made for different reasons.
3154: Send them @emph{individually}.
3155:
3156: If you make two changes for separate reasons, then we might not want to
3157: install them both. We might want to install just one. If you send them
3158: all jumbled together in a single set of diffs, we have to do extra work
3159: to disentangle them---to figure out which parts of the change serve
3160: which purpose. If we don't have time for this, we might have to ignore
3161: your changes entirely.
3162:
3163: If you send each change as soon as you have written it, with its own
3164: explanation, then the two changes never get tangled up, and we can
3165: consider each one properly without any extra work to disentangle them.
3166:
3167: Ideally, each change you send should be impossible to subdivide into
3168: parts that we might want to consider separately, because each of its
3169: parts gets its motivation from the other parts.
3170:
3171: @item
3172: Send each change as soon as that change is finished. Sometimes people
3173: think they are helping us by accumulating many changes to send them all
3174: together. As explained above, this is absolutely the worst thing you
3175: could do.
3176:
3177: Since you should send each change separately, you might as well send it
3178: right away. That gives us the option of installing it immediately if it
3179: is important.
3180:
3181: @item
3182: Use @samp{diff -c} to make your diffs. Diffs without context are hard
3183: for us to install reliably. More than that, they make it hard for us to
3184: study the diffs to decide whether we want to install them. Unidiff
3185: format is better than contextless diffs, but not as easy to read as
3186: @samp{-c} format.
3187:
3188: If you have GNU diff, use @samp{diff -cp}, which shows the name of the
3189: function that each change occurs in.
3190:
3191: @item
3192: Write the change log entries for your changes. We get lots of changes,
3193: and we don't have time to do all the change log writing ourselves.
3194:
3195: Read the @file{ChangeLog} file to see what sorts of information to put
3196: in, and to learn the style that we use. The purpose of the change log
3197: is to show people where to find what was changed. So you need to be
3198: specific about what functions you changed; in large functions, it's
3199: often helpful to indicate where within the function the change was.
3200:
3201: On the other hand, once you have shown people where to find the change,
3202: you need not explain its purpose. Thus, if you add a new function, all
3203: you need to say about it is that it is new. If you feel that the
3204: purpose needs explaining, it probably does---but the explanation will be
3205: much more useful if you put it in comments in the code.
3206:
3207: If you would like your name to appear in the header line for who made
3208: the change, send us the header line.
3209:
3210: @item
3211: When you write the fix, keep in mind that we can't install a change that
3212: would break other systems.
3213:
3214: People often suggest fixing a problem by changing machine-independent
3215: files such as @file{toplev.c} to do something special that a particular
3216: system needs. Sometimes it is totally obvious that such changes would
3217: break GNU CC for almost all users. We can't possibly make a change like
3218: that. At best it might tell us how to write another patch that would
3219: solve the problem acceptably.
3220:
3221: Sometimes people send fixes that @emph{might} be an improvement in
3222: general---but it is hard to be sure of this. It's hard to install
3223: such changes because we have to study them very carefully. Of course,
3224: a good explanation of the reasoning by which you concluded the change
3225: was correct can help convince us.
3226:
3227: The safest changes are changes to the configuration files for a
3228: particular machine. These are safe because they can't create new bugs
3229: on other machines.
3230:
3231: Please help us keep up with the workload by designing the patch in a
3232: form that is good to install.
3233: @end itemize
3234:
3235: @node Service
3236: @chapter How To Get Help with GNU CC
3237:
3238: If you need help installing, using or changing GNU CC, there are two
3239: ways to find it:
3240:
3241: @itemize @bullet
3242: @item
3243: Send a message to a suitable network mailing list. First try
3244: @code{bug-gcc@@prep.ai.mit.edu}, and if that brings no response, try
3245: @code{help-gcc@@prep.ai.mit.edu}.
3246:
3247: @item
3248: Look in the service directory for someone who might help you for a fee.
3249: The service directory is found in the file named @file{SERVICE} in the
3250: GNU CC distribution.
3251: @end itemize
3252:
3253: @node VMS
3254: @chapter Using GNU CC on VMS
3255:
3256: @menu
3257: * Include Files and VMS:: Where the preprocessor looks for the include files.
3258: * Global Declarations:: How to do globaldef, globalref and globalvalue with
3259: GNU CC.
3260: * VMS Misc:: Misc information.
3261: @end menu
3262:
3263: @node Include Files and VMS
3264: @section Include Files and VMS
3265:
3266: @cindex include files and VMS
3267: @cindex VMS and include files
3268: @cindex header files and VMS
3269: Due to the differences between the filesystems of Unix and VMS, GNU CC
3270: attempts to translate file names in @samp{#include} into names that VMS
3271: will understand. The basic strategy is to prepend a prefix to the
3272: specification of the include file, convert the whole filename to a VMS
3273: filename, and then try to open the file. GNU CC tries various prefixes
3274: one by one until one of them succeeds:
3275:
3276: @enumerate
3277: @item
3278: The first prefix is the @samp{GNU_CC_INCLUDE:} logical name: this is
3279: where GNU C header files are traditionally stored. If you wish to store
3280: header files in non-standard locations, then you can assign the logical
3281: @samp{GNU_CC_INCLUDE} to be a search list, where each element of the
3282: list is suitable for use with a rooted logical.
3283:
3284: @item
3285: The next prefix tried is @samp{SYS$SYSROOT:[SYSLIB.]}. This is where
3286: VAX-C header files are traditionally stored.
3287:
3288: @item
3289: If the include file specification by itself is a valid VMS filename, the
3290: preprocessor then uses this name with no prefix in an attempt to open
3291: the include file.
3292:
3293: @item
3294: If the file specification is not a valid VMS filename (i.e. does not
3295: contain a device or a directory specifier, and contains a @samp{/}
3296: character), the preprocessor tries to convert it from Unix syntax to
3297: VMS syntax.
3298:
3299: Conversion works like this: the first directory name becomes a device,
3300: and the rest of the directories are converted into VMS-format directory
3301: names. For example, the name @file{X11/foobar.h} is
3302: translated to @file{X11:[000000]foobar.h} or @file{X11:foobar.h},
3303: whichever one can be opened. This strategy allows you to assign a
3304: logical name to point to the actual location of the header files.
3305:
3306: @item
3307: If none of these strategies succeeds, the @samp{#include} fails.
3308: @end enumerate
3309:
3310: Include directives of the form:
3311:
3312: @example
3313: #include foobar
3314: @end example
3315:
3316: @noindent
3317: are a common source of incompatibility between VAX-C and GNU CC. VAX-C
3318: treats this much like a standard @code{#include <foobar.h>} directive.
3319: That is incompatible with the ANSI C behavior implemented by GNU CC: to
3320: expand the name @code{foobar} as a macro. Macro expansion should
3321: eventually yield one of the two standard formats for @code{#include}:
3322:
3323: @example
3324: #include "@var{file}"
3325: #include <@var{file}>
3326: @end example
3327:
3328: If you have this problem, the best solution is to modify the source to
3329: convert the @code{#include} directives to one of the two standard forms.
3330: That will work with either compiler. If you want a quick and dirty fix,
3331: define the file names as macros with the proper expansion, like this:
3332:
3333: @example
3334: #define stdio <stdio.h>
3335: @end example
3336:
3337: @noindent
3338: This will work, as long as the name doesn't conflict with anything else
3339: in the program.
3340:
3341: Another source of incompatibility is that VAX-C assumes that:
3342:
3343: @example
3344: #include "foobar"
3345: @end example
3346:
3347: @noindent
3348: is actually asking for the file @file{foobar.h}. GNU CC does not
3349: make this assumption, and instead takes what you ask for literally;
3350: it tries to read the file @file{foobar}. The best way to avoid this
3351: problem is to always specify the desired file extension in your include
3352: directives.
3353:
3354: GNU CC for VMS is distributed with a set of include files that is
3355: sufficient to compile most general purpose programs. Even though the
3356: GNU CC distribution does not contain header files to define constants
3357: and structures for some VMS system-specific functions, there is no
3358: reason why you cannot use GNU CC with any of these functions. You first
3359: may have to generate or create header files, either by using the public
3360: domain utility @code{UNSDL} (which can be found on a DECUS tape), or by
3361: extracting the relevant modules from one of the system macro libraries,
3362: and using an editor to construct a C header file.
3363:
3364: A @code{#include} file name cannot contain a DECNET node name. The
3365: preprocessor reports an I/O error if you attempt to use a node name,
3366: whether explicitly, or implicitly via a logical name.
3367:
3368: @node Global Declarations
3369: @section Global Declarations and VMS
3370:
3371: @findex GLOBALREF
3372: @findex GLOBALDEF
3373: @findex GLOBALVALUEDEF
3374: @findex GLOBALVALUEREF
3375: GNU CC does not provide the @code{globalref}, @code{globaldef} and
3376: @code{globalvalue} keywords of VAX-C. You can get the same effect with
3377: an obscure feature of GAS, the GNU assembler. (This requires GAS
3378: version 1.39 or later.) The following macros allow you to use this
3379: feature in a fairly natural way:
3380:
3381: @smallexample
3382: #ifdef __GNUC__
3383: #define GLOBALREF(TYPE,NAME) \
3384: TYPE NAME \
3385: asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME)
3386: #define GLOBALDEF(TYPE,NAME,VALUE) \
3387: TYPE NAME \
3388: asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME) \
3389: = VALUE
3390: #define GLOBALVALUEREF(TYPE,NAME) \
3391: const TYPE NAME[1] \
3392: asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME)
3393: #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
3394: const TYPE NAME[1] \
3395: asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME) \
3396: = @{VALUE@}
3397: #else
3398: #define GLOBALREF(TYPE,NAME) \
3399: globalref TYPE NAME
3400: #define GLOBALDEF(TYPE,NAME,VALUE) \
3401: globaldef TYPE NAME = VALUE
3402: #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
3403: globalvalue TYPE NAME = VALUE
3404: #define GLOBALVALUEREF(TYPE,NAME) \
3405: globalvalue TYPE NAME
3406: #endif
3407: @end smallexample
3408:
3409: @noindent
3410: (The @code{_$$PsectAttributes_GLOBALSYMBOL} prefix at the start of the
3411: name is removed by the assembler, after it has modified the attributes
3412: of the symbol). These macros are provided in the VMS binaries
3413: distribution in a header file @file{GNU_HACKS.H}. An example of the
3414: usage is:
3415:
3416: @example
3417: GLOBALREF (int, ijk);
3418: GLOBALDEF (int, jkl, 0);
3419: @end example
3420:
3421: The macros @code{GLOBALREF} and @code{GLOBALDEF} cannot be used
3422: straightforwardly for arrays, since there is no way to insert the array
3423: dimension into the declaration at the right place. However, you can
3424: declare an array with these macros if you first define a typedef for the
3425: array type, like this:
3426:
3427: @example
3428: typedef int intvector[10];
3429: GLOBALREF (intvector, foo);
3430: @end example
3431:
3432: Array and structure initializers will also break the macros; you can
3433: define the initializer to be a macro of its own, or you can expand the
3434: @code{GLOBALDEF} macro by hand. You may find a case where you wish to
3435: use the @code{GLOBALDEF} macro with a large array, but you are not
3436: interested in explicitly initializing each element of the array. In
3437: such cases you can use an initializer like: @code{@{0,@}}, which will
3438: initialize the entire array to @code{0}.
3439:
3440: A shortcoming of this implementation is that a variable declared with
3441: @code{GLOBALVALUEREF} or @code{GLOBALVALUEDEF} is always an array. For
3442: example, the declaration:
3443:
3444: @example
3445: GLOBALVALUEREF(int, ijk);
3446: @end example
3447:
3448: @noindent
3449: declares the variable @code{ijk} as an array of type @code{int [1]}.
3450: This is done because a globalvalue is actually a constant; its ``value''
3451: is what the linker would normally consider an address. That is not how
3452: an integer value works in C, but it is how an array works. So treating
3453: the symbol as an array name gives consistent results---with the
3454: exception that the value seems to have the wrong type. @strong{Don't
3455: try to access an element of the array.} It doesn't have any elements.
3456: The array ``address'' may not be the address of actual storage.
3457:
3458: The fact that the symbol is an array may lead to warnings where the
3459: variable is used. Insert type casts to avoid the warnings. Here is an
3460: example; it takes advantage of the ANSI C feature allowing macros that
3461: expand to use the same name as the macro itself.
3462:
3463: @example
3464: GLOBALVALUEREF (int, ss$_normal);
3465: GLOBALVALUEDEF (int, xyzzy,123);
3466: #ifdef __GNUC__
3467: #define ss$_normal ((int) ss$_normal)
3468: #define xyzzy ((int) xyzzy)
3469: #endif
3470: @end example
3471:
3472: Don't use @code{globaldef} or @code{globalref} with a variable whose
3473: type is an enumeration type; this is not implemented. Instead, make the
3474: variable an integer, and use a @code{globalvaluedef} for each of the
3475: enumeration values. An example of this would be:
3476:
3477: @example
3478: #ifdef __GNUC__
3479: GLOBALDEF (int, color, 0);
3480: GLOBALVALUEDEF (int, RED, 0);
3481: GLOBALVALUEDEF (int, BLUE, 1);
3482: GLOBALVALUEDEF (int, GREEN, 3);
3483: #else
3484: enum globaldef color @{RED, BLUE, GREEN = 3@};
3485: #endif
3486: @end example
3487:
3488: @node VMS Misc
3489: @section Other VMS Issues
3490:
3491: @cindex exit status and VMS
3492: @cindex return value of @code{main}
3493: @cindex @code{main} and the exit status
3494: GNU CC automatically arranges for @code{main} to return 1 by default if
3495: you fail to specify an explicit return value. This will be interpreted
3496: by VMS as a status code indicating a normal successful completion.
3497: Version 1 of GNU CC did not provide this default.
3498:
3499: GNU CC on VMS works only with the GNU assembler, GAS. You need version
3500: 1.37 or later of GAS in order to produce value debugging information for
3501: the VMS debugger. Use the ordinary VMS linker with the object files
3502: produced by GAS.
3503:
3504: @cindex shared VMS run time system
3505: @cindex @file{VAXCRTL}
3506: Under previous versions of GNU CC, the generated code would occasionally
3507: give strange results when linked to the sharable @file{VAXCRTL} library.
3508: Now this should work.
3509:
3510: A caveat for use of @code{const} global variables: the @code{const}
3511: modifier must be specified in every external declaration of the variable
3512: in all of the source files that use that variable. Otherwise the linker
3513: will issue warnings about conflicting attributes for the variable. Your
3514: program will still work despite the warnings, but the variable will be
3515: placed in writable storage.
3516:
3517: @cindex name augmentation
3518: @cindex case sensitivity and VMS
3519: @cindex VMS and case sensitivity
3520: Although the VMS linker does distinguish between upper and lower case
3521: letters in global symbols, most VMS compilers convert all such symbols
3522: into upper case and most run-time library routines also have upper case
3523: names. To be able to reliably call such routines, GNU CC (by means of
3524: the assembler GAS) converts global symbols into upper case like other
3525: VMS compilers. However, since the usual practice in C is to distinguish
3526: case, GNU CC (via GAS) tries to preserve usual C behavior by augmenting
3527: each name that is not all lower case. This means truncating the name
3528: to at most 23 characters and then adding more characters at the end
3529: which encode the case pattern of those 23. Names which contain at
3530: least one dollar sign are an exception; they are converted directly into
3531: upper case without augmentation.
3532:
3533: Name augmentation yields bad results for programs that use precompiled
3534: libraries (such as Xlib) which were generated by another compiler. You
3535: can use the compiler option @samp{/NOCASE_HACK} to inhibit augmentation;
3536: it makes external C functions and variables case-independent as is usual
3537: on VMS. Alternatively, you could write all references to the functions
3538: and variables in such libraries using lower case; this will work on VMS,
3539: but is not portable to other systems. The compiler option @samp{/NAMES}
3540: also provides control over global name handling.
3541:
3542: Function and variable names are handled somewhat differently with GNU
3543: C++. The GNU C++ compiler performs @dfn{name mangling} on function
3544: names, which means that it adds information to the function name to
3545: describe the data types of the arguments that the function takes. One
3546: result of this is that the name of a function can become very long.
3547: Since the VMS linker only recognizes the first 31 characters in a name,
3548: special action is taken to ensure that each function and variable has a
3549: unique name that can be represented in 31 characters.
3550:
3551: If the name (plus a name augmentation, if required) is less than 32
3552: characters in length, then no special action is performed. If the name
3553: is longer than 31 characters, the assembler (GAS) will generate a
3554: hash string based upon the function name, truncate the function name to
3555: 23 characters, and append the hash string to the truncated name. If the
3556: @samp{/VERBOSE} compiler option is used, the assembler will print both
3557: the full and truncated names of each symbol that is truncated.
3558:
3559: The @samp{/NOCASE_HACK} compiler option should not be used when you are
3560: compiling programs that use libg++. libg++ has several instances of
3561: objects (i.e. @code{Filebuf} and @code{filebuf}) which become
3562: indistinguishable in a case-insensitive environment. This leads to
3563: cases where you need to inhibit augmentation selectively (if you were
3564: using libg++ and Xlib in the same program, for example). There is no
3565: special feature for doing this, but you can get the result by defining a
3566: macro for each mixed case symbol for which you wish to inhibit
3567: augmentation. The macro should expand into the lower case equivalent of
3568: itself. For example:
3569:
3570: @example
3571: #define StuDlyCapS studlycaps
3572: @end example
3573:
3574: These macro definitions can be placed in a header file to minimize the
3575: number of changes to your source code.
3576: @end ifset
3577:
3578: @ifset INTERNALS
3579: @node Portability
3580: @chapter GNU CC and Portability
3581: @cindex portability
3582: @cindex GNU CC and portability
3583:
3584: The main goal of GNU CC was to make a good, fast compiler for machines in
3585: the class that the GNU system aims to run on: 32-bit machines that address
3586: 8-bit bytes and have several general registers. Elegance, theoretical
3587: power and simplicity are only secondary.
3588:
3589: GNU CC gets most of the information about the target machine from a machine
3590: description which gives an algebraic formula for each of the machine's
3591: instructions. This is a very clean way to describe the target. But when
3592: the compiler needs information that is difficult to express in this
3593: fashion, I have not hesitated to define an ad-hoc parameter to the machine
3594: description. The purpose of portability is to reduce the total work needed
3595: on the compiler; it was not of interest for its own sake.
3596:
3597: @cindex endianness
3598: @cindex autoincrement addressing, availability
3599: @findex abort
3600: GNU CC does not contain machine dependent code, but it does contain code
3601: that depends on machine parameters such as endianness (whether the most
3602: significant byte has the highest or lowest address of the bytes in a word)
3603: and the availability of autoincrement addressing. In the RTL-generation
3604: pass, it is often necessary to have multiple strategies for generating code
3605: for a particular kind of syntax tree, strategies that are usable for different
3606: combinations of parameters. Often I have not tried to address all possible
3607: cases, but only the common ones or only the ones that I have encountered.
3608: As a result, a new target may require additional strategies. You will know
3609: if this happens because the compiler will call @code{abort}. Fortunately,
3610: the new strategies can be added in a machine-independent fashion, and will
3611: affect only the target machines that need them.
3612: @end ifset
3613:
3614: @ifset INTERNALS
3615: @node Interface
3616: @chapter Interfacing to GNU CC Output
3617: @cindex interfacing to GNU CC output
3618: @cindex run-time conventions
3619: @cindex function call conventions
3620: @cindex conventions, run-time
3621:
3622: GNU CC is normally configured to use the same function calling convention
3623: normally in use on the target system. This is done with the
3624: machine-description macros described (@pxref{Target Macros}).
3625:
3626: @cindex unions, returning
3627: @cindex structures, returning
3628: @cindex returning structures and unions
3629: However, returning of structure and union values is done differently on
3630: some target machines. As a result, functions compiled with PCC
3631: returning such types cannot be called from code compiled with GNU CC,
3632: and vice versa. This does not cause trouble often because few Unix
3633: library routines return structures or unions.
3634:
3635: GNU CC code returns structures and unions that are 1, 2, 4 or 8 bytes
3636: long in the same registers used for @code{int} or @code{double} return
3637: values. (GNU CC typically allocates variables of such types in
3638: registers also.) Structures and unions of other sizes are returned by
3639: storing them into an address passed by the caller (usually in a
3640: register). The machine-description macros @code{STRUCT_VALUE} and
3641: @code{STRUCT_INCOMING_VALUE} tell GNU CC where to pass this address.
3642:
3643: By contrast, PCC on most target machines returns structures and unions
3644: of any size by copying the data into an area of static storage, and then
3645: returning the address of that storage as if it were a pointer value.
3646: The caller must copy the data from that memory area to the place where
3647: the value is wanted. This is slower than the method used by GNU CC, and
3648: fails to be reentrant.
3649:
3650: On some target machines, such as RISC machines and the 80386, the
3651: standard system convention is to pass to the subroutine the address of
3652: where to return the value. On these machines, GNU CC has been
3653: configured to be compatible with the standard compiler, when this method
3654: is used. It may not be compatible for structures of 1, 2, 4 or 8 bytes.
3655:
3656: @cindex argument passing
3657: @cindex passing arguments
3658: GNU CC uses the system's standard convention for passing arguments. On
3659: some machines, the first few arguments are passed in registers; in
3660: others, all are passed on the stack. It would be possible to use
3661: registers for argument passing on any machine, and this would probably
3662: result in a significant speedup. But the result would be complete
3663: incompatibility with code that follows the standard convention. So this
3664: change is practical only if you are switching to GNU CC as the sole C
3665: compiler for the system. We may implement register argument passing on
3666: certain machines once we have a complete GNU system so that we can
3667: compile the libraries with GNU CC.
3668:
3669: On some machines (particularly the Sparc), certain types of arguments
3670: are passed ``by invisible reference''. This means that the value is
3671: stored in memory, and the address of the memory location is passed to
3672: the subroutine.
3673:
3674: @cindex @code{longjmp} and automatic variables
3675: If you use @code{longjmp}, beware of automatic variables. ANSI C says that
3676: automatic variables that are not declared @code{volatile} have undefined
3677: values after a @code{longjmp}. And this is all GNU CC promises to do,
3678: because it is very difficult to restore register variables correctly, and
3679: one of GNU CC's features is that it can put variables in registers without
3680: your asking it to.
3681:
3682: If you want a variable to be unaltered by @code{longjmp}, and you don't
3683: want to write @code{volatile} because old C compilers don't accept it,
3684: just take the address of the variable. If a variable's address is ever
3685: taken, even if just to compute it and ignore it, then the variable cannot
3686: go in a register:
3687:
3688: @example
3689: @{
3690: int careful;
3691: &careful;
3692: @dots{}
3693: @}
3694: @end example
3695:
3696: @cindex arithmetic libraries
3697: @cindex math libraries
3698: Code compiled with GNU CC may call certain library routines. Most of
3699: them handle arithmetic for which there are no instructions. This
3700: includes multiply and divide on some machines, and floating point
3701: operations on any machine for which floating point support is disabled
3702: with @samp{-msoft-float}. Some standard parts of the C library, such as
3703: @code{bcopy} or @code{memcpy}, are also called automatically. The usual
3704: function call interface is used for calling the library routines.
3705:
3706: These library routines should be defined in the library @file{libgcc.a},
3707: which GNU CC automatically searches whenever it links a program. On
3708: machines that have multiply and divide instructions, if hardware
3709: floating point is in use, normally @file{libgcc.a} is not needed, but it
3710: is searched just in case.
3711:
3712: Each arithmetic function is defined in @file{libgcc1.c} to use the
3713: corresponding C arithmetic operator. As long as the file is compiled
3714: with another C compiler, which supports all the C arithmetic operators,
3715: this file will work portably. However, @file{libgcc1.c} does not work if
3716: compiled with GNU CC, because each arithmetic function would compile
3717: into a call to itself!
3718: @end ifset
3719:
3720: @ifset INTERNALS
3721: @node Passes
3722: @chapter Passes and Files of the Compiler
3723: @cindex passes and files of the compiler
3724: @cindex files and passes of the compiler
3725: @cindex compiler passes and files
3726:
3727: @cindex top level of compiler
3728: The overall control structure of the compiler is in @file{toplev.c}. This
3729: file is responsible for initialization, decoding arguments, opening and
3730: closing files, and sequencing the passes.
3731:
3732: @cindex parsing pass
3733: The parsing pass is invoked only once, to parse the entire input. The RTL
3734: intermediate code for a function is generated as the function is parsed, a
3735: statement at a time. Each statement is read in as a syntax tree and then
3736: converted to RTL; then the storage for the tree for the statement is
3737: reclaimed. Storage for types (and the expressions for their sizes),
3738: declarations, and a representation of the binding contours and how they nest,
3739: remain until the function is finished being compiled; these are all needed
3740: to output the debugging information.
3741:
3742: @findex rest_of_compilation
3743: @findex rest_of_decl_compilation
3744: Each time the parsing pass reads a complete function definition or
3745: top-level declaration, it calls either the function
3746: @code{rest_of_compilation}, or the function
3747: @code{rest_of_decl_compilation} in @file{toplev.c}, which are
3748: responsible for all further processing necessary, ending with output of
3749: the assembler language. All other compiler passes run, in sequence,
3750: within @code{rest_of_compilation}. When that function returns from
3751: compiling a function definition, the storage used for that function
3752: definition's compilation is entirely freed, unless it is an inline
3753: function
3754: @ifset USING
3755: (@pxref{Inline,,An Inline Function is As Fast As a Macro}).
3756: @end ifset
3757: @ifclear USING
3758: (@pxref{Inline,,An Inline Function is As Fast As a Macro,gcc.texi,Using GCC}).
3759: @end ifclear
3760:
3761: Here is a list of all the passes of the compiler and their source files.
3762: Also included is a description of where debugging dumps can be requested
3763: with @samp{-d} options.
3764:
3765: @itemize @bullet
3766: @item
3767: Parsing. This pass reads the entire text of a function definition,
3768: constructing partial syntax trees. This and RTL generation are no longer
3769: truly separate passes (formerly they were), but it is easier to think
3770: of them as separate.
3771:
3772: The tree representation does not entirely follow C syntax, because it is
3773: intended to support other languages as well.
3774:
3775: Language-specific data type analysis is also done in this pass, and every
3776: tree node that represents an expression has a data type attached.
3777: Variables are represented as declaration nodes.
3778:
3779: @cindex constant folding
3780: @cindex arithmetic simplifications
3781: @cindex simplifications, arithmetic
3782: Constant folding and some arithmetic simplifications are also done
3783: during this pass.
3784:
3785: The language-independent source files for parsing are
3786: @file{stor-layout.c}, @file{fold-const.c}, and @file{tree.c}.
3787: There are also header files @file{tree.h} and @file{tree.def}
3788: which define the format of the tree representation.@refill
3789:
3790: @c Avoiding overfull is tricky here.
3791: The source files to parse C are
3792: @file{c-parse.in},
3793: @file{c-decl.c},
3794: @file{c-typeck.c},
3795: @file{c-aux-info.c},
3796: @file{c-convert.c},
3797: and @file{c-lang.c}
3798: along with header files
3799: @file{c-lex.h}, and
3800: @file{c-tree.h}.
3801:
3802: The source files for parsing C++ are @file{cp-parse.y},
3803: @file{cp-class.c},@*
3804: @file{cp-cvt.c}, @file{cp-decl.c}, @file{cp-decl2.c},
3805: @file{cp-dem.c}, @file{cp-except.c},@*
3806: @file{cp-expr.c}, @file{cp-init.c}, @file{cp-lex.c},
3807: @file{cp-method.c}, @file{cp-ptree.c},@*
3808: @file{cp-search.c}, @file{cp-tree.c}, @file{cp-type2.c}, and
3809: @file{cp-typeck.c}, along with header files @file{cp-tree.def},
3810: @file{cp-tree.h}, and @file{cp-decl.h}.
3811:
3812: The special source files for parsing Objective C are
3813: @file{objc-parse.y}, @file{objc-actions.c}, @file{objc-tree.def}, and
3814: @file{objc-actions.h}. Certain C-specific files are used for this as
3815: well.
3816:
3817: The file @file{c-common.c} is also used for all of the above languages.
3818:
3819: @cindex RTL generation
3820: @item
3821: RTL generation. This is the conversion of syntax tree into RTL code.
3822: It is actually done statement-by-statement during parsing, but for
3823: most purposes it can be thought of as a separate pass.
3824:
3825: @cindex target-parameter-dependent code
3826: This is where the bulk of target-parameter-dependent code is found,
3827: since often it is necessary for strategies to apply only when certain
3828: standard kinds of instructions are available. The purpose of named
3829: instruction patterns is to provide this information to the RTL
3830: generation pass.
3831:
3832: @cindex tail recursion optimization
3833: Optimization is done in this pass for @code{if}-conditions that are
3834: comparisons, boolean operations or conditional expressions. Tail
3835: recursion is detected at this time also. Decisions are made about how
3836: best to arrange loops and how to output @code{switch} statements.
3837:
3838: @c Avoiding overfull is tricky here.
3839: The source files for RTL generation include
3840: @file{stmt.c},
3841: @file{calls.c},
3842: @file{expr.c},
3843: @file{explow.c},
3844: @file{expmed.c},
3845: @file{function.c},
3846: @file{optabs.c}
3847: and @file{emit-rtl.c}.
3848: Also, the file
3849: @file{insn-emit.c}, generated from the machine description by the
3850: program @code{genemit}, is used in this pass. The header file
3851: @file{expr.h} is used for communication within this pass.@refill
3852:
3853: @findex genflags
3854: @findex gencodes
3855: The header files @file{insn-flags.h} and @file{insn-codes.h},
3856: generated from the machine description by the programs @code{genflags}
3857: and @code{gencodes}, tell this pass which standard names are available
3858: for use and which patterns correspond to them.@refill
3859:
3860: Aside from debugging information output, none of the following passes
3861: refers to the tree structure representation of the function (only
3862: part of which is saved).
3863:
3864: @cindex inline, automatic
3865: The decision of whether the function can and should be expanded inline
3866: in its subsequent callers is made at the end of rtl generation. The
3867: function must meet certain criteria, currently related to the size of
3868: the function and the types and number of parameters it has. Note that
3869: this function may contain loops, recursive calls to itself
3870: (tail-recursive functions can be inlined!), gotos, in short, all
3871: constructs supported by GNU CC. The file @file{integrate.c} contains
3872: the code to save a function's rtl for later inlining and to inline that
3873: rtl when the function is called. The header file @file{integrate.h}
3874: is also used for this purpose.
3875:
3876: The option @samp{-dr} causes a debugging dump of the RTL code after
3877: this pass. This dump file's name is made by appending @samp{.rtl} to
3878: the input file name.
3879:
3880: @cindex jump optimization
3881: @cindex unreachable code
3882: @cindex dead code
3883: @item
3884: Jump optimization. This pass simplifies jumps to the following
3885: instruction, jumps across jumps, and jumps to jumps. It deletes
3886: unreferenced labels and unreachable code, except that unreachable code
3887: that contains a loop is not recognized as unreachable in this pass.
3888: (Such loops are deleted later in the basic block analysis.) It also
3889: converts some code originally written with jumps into sequences of
3890: instructions that directly set values from the results of comparisons,
3891: if the machine has such instructions.
3892:
3893: Jump optimization is performed two or three times. The first time is
3894: immediately following RTL generation. The second time is after CSE,
3895: but only if CSE says repeated jump optimization is needed. The
3896: last time is right before the final pass. That time, cross-jumping
3897: and deletion of no-op move instructions are done together with the
3898: optimizations described above.
3899:
3900: The source file of this pass is @file{jump.c}.
3901:
3902: The option @samp{-dj} causes a debugging dump of the RTL code after
3903: this pass is run for the first time. This dump file's name is made by
3904: appending @samp{.jump} to the input file name.
3905:
3906: @cindex register use analysis
3907: @item
3908: Register scan. This pass finds the first and last use of each
3909: register, as a guide for common subexpression elimination. Its source
3910: is in @file{regclass.c}.
3911:
3912: @cindex jump threading
3913: @item
3914: Jump threading. This pass detects a condition jump that branches to an
3915: identical or inverse test. Such jumps can be @samp{threaded} through
3916: the second conditional test. The source code for this pass is in
3917: @file{jump.c}. This optimization is only performed if
3918: @samp{-fthread-jumps} is enabled.
3919:
3920: @cindex common subexpression elimination
3921: @cindex constant propagation
3922: @item
3923: Common subexpression elimination. This pass also does constant
3924: propagation. Its source file is @file{cse.c}. If constant
3925: propagation causes conditional jumps to become unconditional or to
3926: become no-ops, jump optimization is run again when CSE is finished.
3927:
3928: The option @samp{-ds} causes a debugging dump of the RTL code after
3929: this pass. This dump file's name is made by appending @samp{.cse} to
3930: the input file name.
3931:
3932: @cindex loop optimization
3933: @cindex code motion
3934: @cindex strength-reduction
3935: @item
3936: Loop optimization. This pass moves constant expressions out of loops,
3937: and optionally does strength-reduction and loop unrolling as well.
3938: Its source files are @file{loop.c} and @file{unroll.c}, plus the header
3939: @file{loop.h} used for communication between them. Loop unrolling uses
3940: some functions in @file{integrate.c} and the header @file{integrate.h}.
3941:
3942: The option @samp{-dL} causes a debugging dump of the RTL code after
3943: this pass. This dump file's name is made by appending @samp{.loop} to
3944: the input file name.
3945:
3946: @item
3947: If @samp{-frerun-cse-after-loop} was enabled, a second common
3948: subexpression elimination pass is performed after the loop optimization
3949: pass. Jump threading is also done again at this time if it was specified.
3950:
3951: The option @samp{-dt} causes a debugging dump of the RTL code after
3952: this pass. This dump file's name is made by appending @samp{.cse2} to
3953: the input file name.
3954:
3955: @cindex register allocation, stupid
3956: @cindex stupid register allocation
3957: @item
3958: Stupid register allocation is performed at this point in a
3959: nonoptimizing compilation. It does a little data flow analysis as
3960: well. When stupid register allocation is in use, the next pass
3961: executed is the reloading pass; the others in between are skipped.
3962: The source file is @file{stupid.c}.
3963:
3964: @cindex data flow analysis
3965: @cindex analysis, data flow
3966: @cindex basic blocks
3967: @item
3968: Data flow analysis (@file{flow.c}). This pass divides the program
3969: into basic blocks (and in the process deletes unreachable loops); then
3970: it computes which pseudo-registers are live at each point in the
3971: program, and makes the first instruction that uses a value point at
3972: the instruction that computed the value.
3973:
3974: @cindex autoincrement/decrement analysis
3975: This pass also deletes computations whose results are never used, and
3976: combines memory references with add or subtract instructions to make
3977: autoincrement or autodecrement addressing.
3978:
3979: The option @samp{-df} causes a debugging dump of the RTL code after
3980: this pass. This dump file's name is made by appending @samp{.flow} to
3981: the input file name. If stupid register allocation is in use, this
3982: dump file reflects the full results of such allocation.
3983:
3984: @cindex instruction combination
3985: @item
3986: Instruction combination (@file{combine.c}). This pass attempts to
3987: combine groups of two or three instructions that are related by data
3988: flow into single instructions. It combines the RTL expressions for
3989: the instructions by substitution, simplifies the result using algebra,
3990: and then attempts to match the result against the machine description.
3991:
3992: The option @samp{-dc} causes a debugging dump of the RTL code after
3993: this pass. This dump file's name is made by appending @samp{.combine}
3994: to the input file name.
3995:
3996: @cindex instruction scheduling
3997: @cindex scheduling, instruction
3998: @item
3999: Instruction scheduling (@file{sched.c}). This pass looks for
4000: instructions whose output will not be available by the time that it is
4001: used in subsequent instructions. (Memory loads and floating point
4002: instructions often have this behavior on RISC machines). It re-orders
4003: instructions within a basic block to try to separate the definition and
4004: use of items that otherwise would cause pipeline stalls.
4005:
4006: Instruction scheduling is performed twice. The first time is immediately
4007: after instruction combination and the second is immediately after reload.
4008:
4009: The option @samp{-dS} causes a debugging dump of the RTL code after this
4010: pass is run for the first time. The dump file's name is made by
4011: appending @samp{.sched} to the input file name.
4012:
4013: @cindex register class preference pass
4014: @item
4015: Register class preferencing. The RTL code is scanned to find out
4016: which register class is best for each pseudo register. The source
4017: file is @file{regclass.c}.
4018:
4019: @cindex register allocation
4020: @cindex local register allocation
4021: @item
4022: Local register allocation (@file{local-alloc.c}). This pass allocates
4023: hard registers to pseudo registers that are used only within one basic
4024: block. Because the basic block is linear, it can use fast and
4025: powerful techniques to do a very good job.
4026:
4027: The option @samp{-dl} causes a debugging dump of the RTL code after
4028: this pass. This dump file's name is made by appending @samp{.lreg} to
4029: the input file name.
4030:
4031: @cindex global register allocation
4032: @item
4033: Global register allocation (@file{global.c}). This pass
4034: allocates hard registers for the remaining pseudo registers (those
4035: whose life spans are not contained in one basic block).
4036:
4037: @cindex reloading
4038: @item
4039: Reloading. This pass renumbers pseudo registers with the hardware
4040: registers numbers they were allocated. Pseudo registers that did not
4041: get hard registers are replaced with stack slots. Then it finds
4042: instructions that are invalid because a value has failed to end up in
4043: a register, or has ended up in a register of the wrong kind. It fixes
4044: up these instructions by reloading the problematical values
4045: temporarily into registers. Additional instructions are generated to
4046: do the copying.
4047:
4048: The reload pass also optionally eliminates the frame pointer and inserts
4049: instructions to save and restore call-clobbered registers around calls.
4050:
4051: Source files are @file{reload.c} and @file{reload1.c}, plus the header
4052: @file{reload.h} used for communication between them.
4053:
4054: The option @samp{-dg} causes a debugging dump of the RTL code after
4055: this pass. This dump file's name is made by appending @samp{.greg} to
4056: the input file name.
4057:
4058: @cindex instruction scheduling
4059: @cindex scheduling, instruction
4060: @item
4061: Instruction scheduling is repeated here to try to avoid pipeline stalls
4062: due to memory loads generated for spilled pseudo registers.
4063:
4064: The option @samp{-dR} causes a debugging dump of the RTL code after
4065: this pass. This dump file's name is made by appending @samp{.sched2}
4066: to the input file name.
4067:
4068: @cindex cross-jumping
4069: @cindex no-op move instructions
4070: @item
4071: Jump optimization is repeated, this time including cross-jumping
4072: and deletion of no-op move instructions.
4073:
4074: The option @samp{-dJ} causes a debugging dump of the RTL code after
4075: this pass. This dump file's name is made by appending @samp{.jump2}
4076: to the input file name.
4077:
4078: @cindex delayed branch scheduling
4079: @cindex scheduling, delayed branch
4080: @item
4081: Delayed branch scheduling. This optional pass attempts to find
4082: instructions that can go into the delay slots of other instructions,
4083: usually jumps and calls. The source file name is @file{reorg.c}.
4084:
4085: The option @samp{-dd} causes a debugging dump of the RTL code after
4086: this pass. This dump file's name is made by appending @samp{.dbr}
4087: to the input file name.
4088:
4089: @cindex register-to-stack conversion
4090: @item
4091: Conversion from usage of some hard registers to usage of a register
4092: stack may be done at this point. Currently, this is supported only
4093: for the floating-point registers of the Intel 80387 coprocessor. The
4094: source file name is @file{reg-stack.c}.
4095:
4096: The options @samp{-dk} causes a debugging dump of the RTL code after
4097: this pass. This dump file's name is made by appending @samp{.stack}
4098: to the input file name.
4099:
4100: @cindex final pass
4101: @cindex peephole optimization
4102: @item
4103: Final. This pass outputs the assembler code for the function. It is
4104: also responsible for identifying spurious test and compare
4105: instructions. Machine-specific peephole optimizations are performed
4106: at the same time. The function entry and exit sequences are generated
4107: directly as assembler code in this pass; they never exist as RTL.
4108:
4109: The source files are @file{final.c} plus @file{insn-output.c}; the
4110: latter is generated automatically from the machine description by the
4111: tool @file{genoutput}. The header file @file{conditions.h} is used
4112: for communication between these files.
4113:
4114: @cindex debugging information generation
4115: @item
4116: Debugging information output. This is run after final because it must
4117: output the stack slot offsets for pseudo registers that did not get
4118: hard registers. Source files are @file{dbxout.c} for DBX symbol table
4119: format, @file{sdbout.c} for SDB symbol table format, and
4120: @file{dwarfout.c} for DWARF symbol table format.
4121: @end itemize
4122:
4123: Some additional files are used by all or many passes:
4124:
4125: @itemize @bullet
4126: @item
4127: Every pass uses @file{machmode.def} and @file{machmode.h} which define
4128: the machine modes.
4129:
4130: @item
4131: Several passes use @file{real.h}, which defines the default
4132: representation of floating point constants and how to operate on them.
4133:
4134: @item
4135: All the passes that work with RTL use the header files @file{rtl.h}
4136: and @file{rtl.def}, and subroutines in file @file{rtl.c}. The tools
4137: @code{gen*} also use these files to read and work with the machine
4138: description RTL.
4139:
4140: @findex genconfig
4141: @item
4142: Several passes refer to the header file @file{insn-config.h} which
4143: contains a few parameters (C macro definitions) generated
4144: automatically from the machine description RTL by the tool
4145: @code{genconfig}.
4146:
4147: @cindex instruction recognizer
4148: @item
4149: Several passes use the instruction recognizer, which consists of
4150: @file{recog.c} and @file{recog.h}, plus the files @file{insn-recog.c}
4151: and @file{insn-extract.c} that are generated automatically from the
4152: machine description by the tools @file{genrecog} and
4153: @file{genextract}.@refill
4154:
4155: @item
4156: Several passes use the header files @file{regs.h} which defines the
4157: information recorded about pseudo register usage, and @file{basic-block.h}
4158: which defines the information recorded about basic blocks.
4159:
4160: @item
4161: @file{hard-reg-set.h} defines the type @code{HARD_REG_SET}, a bit-vector
4162: with a bit for each hard register, and some macros to manipulate it.
4163: This type is just @code{int} if the machine has few enough hard registers;
4164: otherwise it is an array of @code{int} and some of the macros expand
4165: into loops.
4166:
4167: @item
4168: Several passes use instruction attributes. A definition of the
4169: attributes defined for a particular machine is in file
4170: @file{insn-attr.h}, which is generated from the machine description by
4171: the program @file{genattr}. The file @file{insn-attrtab.c} contains
4172: subroutines to obtain the attribute values for insns. It is generated
4173: from the machine description by the program @file{genattrtab}.@refill
4174: @end itemize
4175: @end ifset
4176:
4177: @ifset INTERNALS
4178: @include rtl.texi
4179: @include md.texi
4180: @include tm.texi
4181: @end ifset
4182:
4183: @ifset INTERNALS
4184: @node Config
4185: @chapter The Configuration File
4186: @cindex configuration file
4187: @cindex @file{xm-@var{machine}.h}
4188:
4189: The configuration file @file{xm-@var{machine}.h} contains macro
4190: definitions that describe the machine and system on which the compiler
4191: is running, unlike the definitions in @file{@var{machine}.h}, which
4192: describe the machine for which the compiler is producing output. Most
4193: of the values in @file{xm-@var{machine}.h} are actually the same on all
4194: machines that GNU CC runs on, so large parts of all configuration files
4195: are identical. But there are some macros that vary:
4196:
4197: @table @code
4198: @findex USG
4199: @item USG
4200: Define this macro if the host system is System V.
4201:
4202: @findex VMS
4203: @item VMS
4204: Define this macro if the host system is VMS.
4205:
4206: @findex FAILURE_EXIT_CODE
4207: @item FAILURE_EXIT_CODE
4208: A C expression for the status code to be returned when the compiler
4209: exits after serious errors.
4210:
4211: @findex SUCCESS_EXIT_CODE
4212: @item SUCCESS_EXIT_CODE
4213: A C expression for the status code to be returned when the compiler
4214: exits without serious errors.
4215:
4216: @findex HOST_WORDS_BIG_ENDIAN
4217: @item HOST_WORDS_BIG_ENDIAN
4218: Defined if the host machine stores words of multi-word values in
4219: big-endian order. (GNU CC does not depend on the host byte ordering
4220: within a word.)
4221:
4222: @findex HOST_FLOAT_WORDS_BIG_ENDIAN
4223: @item HOST_FLOAT_WORDS_BIG_ENDIAN
4224: Define this macro to be 1 if the host machine stores @code{DFmode},
4225: @code{XFmode} or @code{TFmode} floating point numbers in memory with the
4226: word containing the sign bit at the lowest address; otherwise, define it
4227: to be zero.
4228:
4229: This macro need not be defined if the ordering is the same as for
4230: multi-word integers.
4231:
4232: @findex HOST_FLOAT_FORMAT
4233: @item HOST_FLOAT_FORMAT
4234: A numeric code distinguishing the floating point format for the host
4235: machine. See @code{TARGET_FLOAT_FORMAT} in @ref{Storage Layout} for the
4236: alternatives and default.
4237:
4238: @findex HOST_BITS_PER_CHAR
4239: @item HOST_BITS_PER_CHAR
4240: A C expression for the number of bits in @code{char} on the host
4241: machine.
4242:
4243: @findex HOST_BITS_PER_SHORT
4244: @item HOST_BITS_PER_SHORT
4245: A C expression for the number of bits in @code{short} on the host
4246: machine.
4247:
4248: @findex HOST_BITS_PER_INT
4249: @item HOST_BITS_PER_INT
4250: A C expression for the number of bits in @code{int} on the host
4251: machine.
4252:
4253: @findex HOST_BITS_PER_LONG
4254: @item HOST_BITS_PER_LONG
4255: A C expression for the number of bits in @code{long} on the host
4256: machine.
4257:
4258: @findex ONLY_INT_FIELDS
4259: @item ONLY_INT_FIELDS
4260: Define this macro to indicate that the host compiler only supports
4261: @code{int} bit fields, rather than other integral types, including
4262: @code{enum}, as do most C compilers.
4263:
4264: @findex EXECUTABLE_SUFFIX
4265: @item EXECUTABLE_SUFFIX
4266: Define this macro if the host system uses a naming convention for
4267: executable files that involves a common suffix (such as, in some
4268: systems, @samp{.exe}) that must be mentioned explicitly when you run
4269: the program.
4270:
4271: @findex OBSTACK_CHUNK_SIZE
4272: @item OBSTACK_CHUNK_SIZE
4273: A C expression for the size of ordinary obstack chunks.
4274: If you don't define this, a usually-reasonable default is used.
4275:
4276: @findex OBSTACK_CHUNK_ALLOC
4277: @item OBSTACK_CHUNK_ALLOC
4278: The function used to allocate obstack chunks.
4279: If you don't define this, @code{xmalloc} is used.
4280:
4281: @findex OBSTACK_CHUNK_FREE
4282: @item OBSTACK_CHUNK_FREE
4283: The function used to free obstack chunks.
4284: If you don't define this, @code{free} is used.
4285:
4286: @findex USE_C_ALLOCA
4287: @item USE_C_ALLOCA
4288: Define this macro to indicate that the compiler is running with the
4289: @code{alloca} implemented in C. This version of @code{alloca} can be
4290: found in the file @file{alloca.c}; to use it, you must also alter the
4291: @file{Makefile} variable @code{ALLOCA}. (This is done automatically
4292: for the systems on which we know it is needed.)
4293:
4294: If you do define this macro, you should probably do it as follows:
4295:
4296: @example
4297: #ifndef __GNUC__
4298: #define USE_C_ALLOCA
4299: #else
4300: #define alloca __builtin_alloca
4301: #endif
4302: @end example
4303:
4304: @noindent
4305: so that when the compiler is compiled with GNU CC it uses the more
4306: efficient built-in @code{alloca} function.
4307:
4308: @item FUNCTION_CONVERSION_BUG
4309: @findex FUNCTION_CONVERSION_BUG
4310: Define this macro to indicate that the host compiler does not properly
4311: handle converting a function value to a pointer-to-function when it is
4312: used in an expression.
4313:
4314: @findex HAVE_VPRINTF
4315: @findex vprintf
4316: @item HAVE_VPRINTF
4317: Define this if the library function @code{vprintf} is available on your
4318: system.
4319:
4320: @findex MULTIBYTE_CHARS
4321: @item MULTIBYTE_CHARS
4322: Define this macro to enable support for multibyte characters in the
4323: input to GNU CC. This requires that the host system support the ANSI C
4324: library functions for converting multibyte characters to wide
4325: characters.
4326:
4327: @findex HAVE_PUTENV
4328: @findex putenv
4329: @item HAVE_PUTENV
4330: Define this if the library function @code{putenv} is available on your
4331: system.
4332:
4333: @findex NO_SYS_SIGLIST
4334: @item NO_SYS_SIGLIST
4335: Define this if your system @emph{does not} provide the variable
4336: @code{sys_siglist}.
4337:
4338: @findex USE_PROTOTYPES
4339: @item USE_PROTOTYPES
4340: Define this to be 1 if you know that the host compiler supports
4341: prototypes, even if it doesn't define __STDC__, or define
4342: it to be 0 if you do not want any prototypes used in compiling
4343: GNU CC. If @samp{USE_PROTOTYPES} is not defined, it will be
4344: determined automatically whether your compiler supports
4345: prototypes by checking if @samp{__STDC__} is defined.
4346:
4347: @findex NO_MD_PROTOTYPES
4348: @item NO_MD_PROTOTYPES
4349: Define this if you wish suppression of prototypes generated from
4350: the machine description file, but to use other prototypes within
4351: GNU CC. If @samp{USE_PROTOTYPES} is defined to be 0, or the
4352: host compiler does not support prototypes, this macro has no
4353: effect.
4354:
4355: @findex MD_CALL_PROTOTYPES
4356: @item MD_CALL_PROTOTYPES
4357: Define this if you wish to generate prototypes for the
4358: @code{gen_call} or @code{gen_call_value} functions generated from
4359: the machine description file. If @samp{USE_PROTOTYPES} is
4360: defined to be 0, or the host compiler does not support
4361: prototypes, or @samp{NO_MD_PROTOTYPES} is defined, this macro has
4362: no effect. As soon as all of the machine descriptions are
4363: modified to have the appropriate number of arguments, this macro
4364: will be removed.
4365:
4366: @vindex sys_siglist
4367: Some systems do provide this variable, but with a different name such
4368: as @code{_sys_siglist}. On these systems, you can define
4369: @code{sys_siglist} as a macro which expands into the name actually
4370: provided.
4371:
4372: @findex NO_STAB_H
4373: @item NO_STAB_H
4374: Define this if your system does not have the include file
4375: @file{stab.h}. If @samp{USG} is defined, @samp{NO_STAB_H} is
4376: assumed.
4377: @end table
4378:
4379: @findex bzero
4380: @findex bcmp
4381: In addition, configuration files for system V define @code{bcopy},
4382: @code{bzero} and @code{bcmp} as aliases. Some files define @code{alloca}
4383: as a macro when compiled with GNU CC, in order to take advantage of the
4384: benefit of GNU CC's built-in @code{alloca}.
4385:
4386:
4387: @node Index
4388: @unnumbered Index
4389: @end ifset
4390:
4391: @ifclear INTERNALS
4392: @node Index
4393: @unnumbered Index
4394: @end ifclear
4395:
4396: @printindex cp
4397: @summarycontents
4398: @contents
4399: @bye
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.