![[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 [GCC 2.x] / gcc|
Return to gcc.texi CVS log | Up to [GCC 2.x] / gcc |
1.1 root 1: \input texinfo @c -*-texinfo-*-
1.1.1.5 root 2: @c %**start of header
1.1 root 3: @setfilename gcc.info
4: @c @setfilename usegcc.info
1.1.1.5 root 5: @c @setfilename portgcc.info
1.1 root 6: @c To produce the full manual, use the "gcc.info" setfilename, and
1.1.1.5 root 7: @c make sure the following do NOT begin with '@c' (and the @clear lines DO)
1.1 root 8: @set INTERNALS
1.1.1.5 root 9: @set USING
1.1 root 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
1.1.1.5 root 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
1.1.1.7 ! root 22:
! 23: @c (For FSF printing, turn on smallbook, comment out finalout below;
! 24: @c that is all that is needed.)
! 25:
! 26: @c smallbook
1.1.1.5 root 27:
28: @c i also commented out the finalout command, so if there *are* any
29: @c overfulls, you'll (hopefully) see the rectangle in the right hand
30: @c margin. -mew 15june93
1.1.1.7 ! root 31: @c finalout
1.1.1.5 root 32:
33: @c NOTE: checks/things to do:
34: @c
35: @c -have bob do a search in all seven files for "mew" (ideally --mew,
36: @c but i may have forgotten the occasional "--"..).
37: @c -item/itemx, text after all (sub/sub)section titles, etc..
38: @c -consider putting the lists of options on pp 17--> etc in columns or
39: @c somesuch.
40: @c -spellcheck
41: @c -continuity of phrasing; ie, bit-field vs bitfield in rtl.texi
42: @c -overfulls. do a search for "mew" in the files, and you will see
43: @c overfulls that i noted but could not deal with.
44: @c -have to add text: beginning of chapter 8
45:
46: @c
47: @c anything else? --mew 10feb93
48:
49:
1.1 root 50:
51: @ifset INTERNALS
1.1.1.5 root 52: @ifset USING
1.1 root 53: @settitle Using and Porting GNU CC
54: @end ifset
1.1.1.5 root 55: @end ifset
56: @c seems reasonable to assume at least one of INTERNALS or USING is set...
1.1 root 57: @ifclear INTERNALS
58: @settitle Using GNU CC
59: @end ifclear
1.1.1.5 root 60: @ifclear USING
61: @settitle Porting GNU CC
62: @end ifclear
1.1 root 63:
64: @syncodeindex fn cp
1.1.1.2 root 65: @syncodeindex vr cp
1.1.1.5 root 66: @c %**end of header
67:
68: @c Use with @@smallbook.
69:
70: @c Cause even numbered pages to be printed on the left hand side of
71: @c the page and odd numbered pages to be printed on the right hand
72: @c side of the page. Using this, you can print on both sides of a
73: @c sheet of paper and have the text on the same part of the sheet.
74:
75: @c The text on right hand pages is pushed towards the right hand
76: @c margin and the text on left hand pages is pushed toward the left
77: @c hand margin.
78: @c (To provide the reverse effect, set bindingoffset to -0.75in.)
79:
80: @c @tex
81: @c \global\bindingoffset=0.75in
82: @c \global\normaloffset =0.75in
83: @c @end tex
1.1 root 84:
85: @ifinfo
86: @ifset INTERNALS
1.1.1.5 root 87: @ifset USING
88: This file documents the use and the internals of the GNU compiler.
1.1 root 89: @end ifset
1.1.1.5 root 90: @end ifset
91: @ifclear USING
92: This file documents the internals of the GNU compiler.
93: @end ifclear
94: @ifclear INTERNALS
95: This file documents the use of the GNU compiler.
96: @end ifclear
1.1 root 97:
1.1.1.5 root 98: Published by the Free Software Foundation
99: 675 Massachusetts Avenue
100: Cambridge, MA 02139 USA
101:
1.1.1.7 ! root 102: Copyright (C) 1988, 1989, 1992, 1993, 1994 Free Software Foundation, Inc.
1.1 root 103:
104: Permission is granted to make and distribute verbatim copies of
105: this manual provided the copyright notice and this permission notice
106: are preserved on all copies.
107:
108: @ignore
109: Permission is granted to process this file through Tex and print the
110: results, provided the printed document carries copying permission
111: notice identical to this one except for the removal of this paragraph
112: (this paragraph not being relevant to the printed manual).
113:
114: @end ignore
115: Permission is granted to copy and distribute modified versions of this
116: manual under the conditions for verbatim copying, provided also that the
1.1.1.7 ! root 117: sections entitled ``GNU General Public License,'' ``Funding for Free
! 118: Software,'' and ``Protect Your Freedom---Fight `Look And Feel'@w{}'' are
! 119: included exactly as in the original, and provided that the entire
! 120: resulting derived work is distributed under the terms of a permission
! 121: notice identical to this one.
1.1 root 122:
123: Permission is granted to copy and distribute translations of this manual
124: into another language, under the above conditions for modified versions,
1.1.1.7 ! root 125: except that the sections entitled ``GNU General Public License,''
! 126: ``Funding for Free Software,'' and ``Protect Your Freedom---Fight `Look
! 127: And Feel'@w{}'', and this permission notice, may be included in
! 128: translations approved by the Free Software Foundation instead of in the
! 129: original English.
1.1 root 130: @end ifinfo
131:
132: @setchapternewpage odd
133:
134: @titlepage
135: @ifset INTERNALS
1.1.1.5 root 136: @ifset USING
1.1 root 137: @center @titlefont{Using and Porting GNU CC}
1.1.1.5 root 138:
139: @end ifset
1.1 root 140: @end ifset
141: @ifclear INTERNALS
142: @title Using GNU CC
143: @end ifclear
1.1.1.5 root 144: @ifclear USING
145: @title Porting GNU CC
146: @end ifclear
1.1 root 147: @sp 2
148: @center Richard M. Stallman
149: @sp 3
1.1.1.7 ! root 150: @center Last updated 19 September 1994
1.1 root 151: @sp 1
1.1.1.5 root 152: @c The version number appears twice more in this file.
153:
1.1.1.7 ! root 154: @center for version 2.6
1.1.1.5 root 155: @c @center (preliminary draft, which will change)
1.1 root 156: @page
157: @vskip 0pt plus 1filll
1.1.1.7 ! root 158: Copyright @copyright{} 1988, 89, 92, 93, 1994 Free Software Foundation, Inc.
1.1.1.5 root 159: @sp 2
1.1.1.7 ! root 160: For GCC Version 2.6.@*
! 161: @c Printed November, 1994.@*
1.1 root 162:
1.1.1.7 ! root 163: @c ISBN 1-882114-35-3
1.1.1.5 root 164: @sp 1
165: Published by the Free Software Foundation @*
166: 675 Massachusetts Avenue @*
167: Cambridge, MA 02139 USA
168: @sp 1
1.1 root 169: Permission is granted to make and distribute verbatim copies of
170: this manual provided the copyright notice and this permission notice
171: are preserved on all copies.
172:
173: Permission is granted to copy and distribute modified versions of this
174: manual under the conditions for verbatim copying, provided also that the
1.1.1.7 ! root 175: sections entitled ``GNU General Public License,'' ``Funding for Free
! 176: Software,'' and ``Protect Your Freedom---Fight `Look And Feel'@w{}'' are
! 177: included exactly as in the original, and provided that the entire
! 178: resulting derived work is distributed under the terms of a permission
! 179: notice identical to this one.
1.1 root 180:
181: Permission is granted to copy and distribute translations of this manual
182: into another language, under the above conditions for modified versions,
1.1.1.7 ! root 183: except that the sections entitled ``GNU General Public License,''
! 184: ``Funding for Free Software,'' and ``Protect Your Freedom---Fight `Look
! 185: And Feel'@w{}'', and this permission notice, may be included in
! 186: translations approved by the Free Software Foundation instead of in the
! 187: original English.
1.1 root 188: @end titlepage
189: @page
190:
191: @ifinfo
192:
193: @node Top, Copying,, (DIR)
1.1.1.3 root 194: @top Introduction
1.1 root 195: @cindex introduction
196:
197: @ifset INTERNALS
1.1.1.5 root 198: @ifset USING
199: This manual documents how to run, install and port the GNU
200: compiler, as well as its new features and incompatibilities, and how to
1.1.1.7 ! root 201: report bugs. It corresponds to GNU CC version 2.6.
1.1.1.5 root 202: @end ifset
1.1 root 203: @end ifset
204:
205: @ifclear INTERNALS
1.1.1.5 root 206: This manual documents how to run and install the GNU compiler,
207: as well as its new features and incompatibilities, and how to report
1.1.1.7 ! root 208: bugs. It corresponds to GNU CC version 2.6.
1.1.1.5 root 209: @end ifclear
210: @ifclear USING
211: This manual documents how to port the GNU compiler,
212: as well as its new features and incompatibilities, and how to report
1.1.1.7 ! root 213: bugs. It corresponds to GNU CC version 2.6.
1.1 root 214: @end ifclear
215:
216: @end ifinfo
217: @menu
218: * Copying:: GNU General Public License says
219: how you can copy and share GNU CC.
220: * Contributors:: People who have contributed to GNU CC.
1.1.1.7 ! root 221: * Funding:: How to help assure funding for free software.
! 222: * Look and Feel:: Protect your freedom---fight ``look and feel''.
1.1.1.5 root 223: @ifset USING
224: * G++ and GCC:: You can compile C or C++ programs.
1.1.1.3 root 225: * Invoking GCC:: Command options supported by @samp{gcc}.
1.1 root 226: * Installation:: How to configure, compile and install GNU CC.
1.1.1.5 root 227: * C Extensions:: GNU extensions to the C language family.
228: * C++ Extensions:: GNU extensions to the C++ language.
1.1 root 229: * Trouble:: If you have trouble installing GNU CC.
1.1.1.3 root 230: * Bugs:: How, why and where to report bugs.
1.1 root 231: * Service:: How to find suppliers of support for GNU CC.
232: * VMS:: Using GNU CC on VMS.
1.1.1.5 root 233: @end ifset
234: @ifset INTERNALS
1.1 root 235: * Portability:: Goals of GNU CC's portability features.
236: * Interface:: Function-call interface of GNU CC output.
237: * Passes:: Order of passes, what they do, and what each file is for.
238: * RTL:: The intermediate representation that most passes work on.
239: * Machine Desc:: How to write machine description instruction patterns.
1.1.1.2 root 240: * Target Macros:: How to write the machine description C macros.
1.1 root 241: * Config:: Writing the @file{xm-@var{machine}.h} file.
242: @end ifset
1.1.1.5 root 243:
1.1 root 244: * Index:: Index of concepts and symbol names.
245: @end menu
246:
1.1.1.7 ! root 247: @node Copying
1.1 root 248: @unnumbered GNU GENERAL PUBLIC LICENSE
249: @center Version 2, June 1991
250:
251: @display
252: Copyright @copyright{} 1989, 1991 Free Software Foundation, Inc.
253: 675 Mass Ave, Cambridge, MA 02139, USA
254:
255: Everyone is permitted to copy and distribute verbatim copies
256: of this license document, but changing it is not allowed.
257: @end display
258:
259: @unnumberedsec Preamble
260:
261: The licenses for most software are designed to take away your
262: freedom to share and change it. By contrast, the GNU General Public
263: License is intended to guarantee your freedom to share and change free
264: software---to make sure the software is free for all its users. This
265: General Public License applies to most of the Free Software
266: Foundation's software and to any other program whose authors commit to
267: using it. (Some other Free Software Foundation software is covered by
268: the GNU Library General Public License instead.) You can apply it to
269: your programs, too.
270:
271: When we speak of free software, we are referring to freedom, not
272: price. Our General Public Licenses are designed to make sure that you
273: have the freedom to distribute copies of free software (and charge for
274: this service if you wish), that you receive source code or can get it
275: if you want it, that you can change the software or use pieces of it
276: in new free programs; and that you know you can do these things.
277:
278: To protect your rights, we need to make restrictions that forbid
279: anyone to deny you these rights or to ask you to surrender the rights.
280: These restrictions translate to certain responsibilities for you if you
281: distribute copies of the software, or if you modify it.
282:
283: For example, if you distribute copies of such a program, whether
284: gratis or for a fee, you must give the recipients all the rights that
285: you have. You must make sure that they, too, receive or can get the
286: source code. And you must show them these terms so they know their
287: rights.
288:
289: We protect your rights with two steps: (1) copyright the software, and
290: (2) offer you this license which gives you legal permission to copy,
291: distribute and/or modify the software.
292:
293: Also, for each author's protection and ours, we want to make certain
294: that everyone understands that there is no warranty for this free
295: software. If the software is modified by someone else and passed on, we
296: want its recipients to know that what they have is not the original, so
297: that any problems introduced by others will not reflect on the original
298: authors' reputations.
299:
300: Finally, any free program is threatened constantly by software
301: patents. We wish to avoid the danger that redistributors of a free
302: program will individually obtain patent licenses, in effect making the
303: program proprietary. To prevent this, we have made it clear that any
304: patent must be licensed for everyone's free use or not licensed at all.
305:
306: The precise terms and conditions for copying, distribution and
307: modification follow.
308:
309: @iftex
310: @unnumberedsec TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
311: @end iftex
312: @ifinfo
313: @center TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
314: @end ifinfo
315:
1.1.1.5 root 316: @enumerate 0
1.1 root 317: @item
318: This License applies to any program or other work which contains
319: a notice placed by the copyright holder saying it may be distributed
320: under the terms of this General Public License. The ``Program'', below,
321: refers to any such program or work, and a ``work based on the Program''
322: means either the Program or any derivative work under copyright law:
323: that is to say, a work containing the Program or a portion of it,
324: either verbatim or with modifications and/or translated into another
325: language. (Hereinafter, translation is included without limitation in
326: the term ``modification''.) Each licensee is addressed as ``you''.
327:
328: Activities other than copying, distribution and modification are not
329: covered by this License; they are outside its scope. The act of
330: running the Program is not restricted, and the output from the Program
331: is covered only if its contents constitute a work based on the
332: Program (independent of having been made by running the Program).
333: Whether that is true depends on what the Program does.
334:
335: @item
336: You may copy and distribute verbatim copies of the Program's
337: source code as you receive it, in any medium, provided that you
338: conspicuously and appropriately publish on each copy an appropriate
339: copyright notice and disclaimer of warranty; keep intact all the
340: notices that refer to this License and to the absence of any warranty;
341: and give any other recipients of the Program a copy of this License
342: along with the Program.
343:
344: You may charge a fee for the physical act of transferring a copy, and
345: you may at your option offer warranty protection in exchange for a fee.
346:
347: @item
348: You may modify your copy or copies of the Program or any portion
349: of it, thus forming a work based on the Program, and copy and
350: distribute such modifications or work under the terms of Section 1
351: above, provided that you also meet all of these conditions:
352:
353: @enumerate a
354: @item
355: You must cause the modified files to carry prominent notices
356: stating that you changed the files and the date of any change.
357:
358: @item
359: You must cause any work that you distribute or publish, that in
360: whole or in part contains or is derived from the Program or any
361: part thereof, to be licensed as a whole at no charge to all third
362: parties under the terms of this License.
363:
364: @item
365: If the modified program normally reads commands interactively
366: when run, you must cause it, when started running for such
367: interactive use in the most ordinary way, to print or display an
368: announcement including an appropriate copyright notice and a
369: notice that there is no warranty (or else, saying that you provide
370: a warranty) and that users may redistribute the program under
371: these conditions, and telling the user how to view a copy of this
372: License. (Exception: if the Program itself is interactive but
373: does not normally print such an announcement, your work based on
374: the Program is not required to print an announcement.)
375: @end enumerate
376:
377: These requirements apply to the modified work as a whole. If
378: identifiable sections of that work are not derived from the Program,
379: and can be reasonably considered independent and separate works in
380: themselves, then this License, and its terms, do not apply to those
381: sections when you distribute them as separate works. But when you
382: distribute the same sections as part of a whole which is a work based
383: on the Program, the distribution of the whole must be on the terms of
384: this License, whose permissions for other licensees extend to the
385: entire whole, and thus to each and every part regardless of who wrote it.
386:
387: Thus, it is not the intent of this section to claim rights or contest
388: your rights to work written entirely by you; rather, the intent is to
389: exercise the right to control the distribution of derivative or
390: collective works based on the Program.
391:
392: In addition, mere aggregation of another work not based on the Program
393: with the Program (or with a work based on the Program) on a volume of
394: a storage or distribution medium does not bring the other work under
395: the scope of this License.
396:
397: @item
398: You may copy and distribute the Program (or a work based on it,
399: under Section 2) in object code or executable form under the terms of
400: Sections 1 and 2 above provided that you also do one of the following:
401:
402: @enumerate a
403: @item
404: Accompany it with the complete corresponding machine-readable
405: source code, which must be distributed under the terms of Sections
406: 1 and 2 above on a medium customarily used for software interchange; or,
407:
408: @item
409: Accompany it with a written offer, valid for at least three
410: years, to give any third party, for a charge no more than your
411: cost of physically performing source distribution, a complete
412: machine-readable copy of the corresponding source code, to be
413: distributed under the terms of Sections 1 and 2 above on a medium
414: customarily used for software interchange; or,
415:
416: @item
417: Accompany it with the information you received as to the offer
418: to distribute corresponding source code. (This alternative is
419: allowed only for noncommercial distribution and only if you
420: received the program in object code or executable form with such
421: an offer, in accord with Subsection b above.)
422: @end enumerate
423:
424: The source code for a work means the preferred form of the work for
425: making modifications to it. For an executable work, complete source
426: code means all the source code for all modules it contains, plus any
427: associated interface definition files, plus the scripts used to
428: control compilation and installation of the executable. However, as a
429: special exception, the source code distributed need not include
430: anything that is normally distributed (in either source or binary
431: form) with the major components (compiler, kernel, and so on) of the
432: operating system on which the executable runs, unless that component
433: itself accompanies the executable.
434:
435: If distribution of executable or object code is made by offering
436: access to copy from a designated place, then offering equivalent
437: access to copy the source code from the same place counts as
438: distribution of the source code, even though third parties are not
439: compelled to copy the source along with the object code.
440:
441: @item
442: You may not copy, modify, sublicense, or distribute the Program
443: except as expressly provided under this License. Any attempt
444: otherwise to copy, modify, sublicense or distribute the Program is
445: void, and will automatically terminate your rights under this License.
446: However, parties who have received copies, or rights, from you under
447: this License will not have their licenses terminated so long as such
448: parties remain in full compliance.
449:
450: @item
451: You are not required to accept this License, since you have not
452: signed it. However, nothing else grants you permission to modify or
453: distribute the Program or its derivative works. These actions are
454: prohibited by law if you do not accept this License. Therefore, by
455: modifying or distributing the Program (or any work based on the
456: Program), you indicate your acceptance of this License to do so, and
457: all its terms and conditions for copying, distributing or modifying
458: the Program or works based on it.
459:
460: @item
461: Each time you redistribute the Program (or any work based on the
462: Program), the recipient automatically receives a license from the
463: original licensor to copy, distribute or modify the Program subject to
464: these terms and conditions. You may not impose any further
465: restrictions on the recipients' exercise of the rights granted herein.
466: You are not responsible for enforcing compliance by third parties to
467: this License.
468:
469: @item
470: If, as a consequence of a court judgment or allegation of patent
471: infringement or for any other reason (not limited to patent issues),
472: conditions are imposed on you (whether by court order, agreement or
473: otherwise) that contradict the conditions of this License, they do not
474: excuse you from the conditions of this License. If you cannot
475: distribute so as to satisfy simultaneously your obligations under this
476: License and any other pertinent obligations, then as a consequence you
477: may not distribute the Program at all. For example, if a patent
478: license would not permit royalty-free redistribution of the Program by
479: all those who receive copies directly or indirectly through you, then
480: the only way you could satisfy both it and this License would be to
481: refrain entirely from distribution of the Program.
482:
483: If any portion of this section is held invalid or unenforceable under
484: any particular circumstance, the balance of the section is intended to
485: apply and the section as a whole is intended to apply in other
486: circumstances.
487:
488: It is not the purpose of this section to induce you to infringe any
489: patents or other property right claims or to contest validity of any
490: such claims; this section has the sole purpose of protecting the
491: integrity of the free software distribution system, which is
492: implemented by public license practices. Many people have made
493: generous contributions to the wide range of software distributed
494: through that system in reliance on consistent application of that
495: system; it is up to the author/donor to decide if he or she is willing
496: to distribute software through any other system and a licensee cannot
497: impose that choice.
498:
499: This section is intended to make thoroughly clear what is believed to
500: be a consequence of the rest of this License.
501:
502: @item
503: If the distribution and/or use of the Program is restricted in
504: certain countries either by patents or by copyrighted interfaces, the
505: original copyright holder who places the Program under this License
506: may add an explicit geographical distribution limitation excluding
507: those countries, so that distribution is permitted only in or among
508: countries not thus excluded. In such case, this License incorporates
509: the limitation as if written in the body of this License.
510:
511: @item
512: The Free Software Foundation may publish revised and/or new versions
513: of the General Public License from time to time. Such new versions will
514: be similar in spirit to the present version, but may differ in detail to
515: address new problems or concerns.
516:
517: Each version is given a distinguishing version number. If the Program
518: specifies a version number of this License which applies to it and ``any
519: later version'', you have the option of following the terms and conditions
520: either of that version or of any later version published by the Free
521: Software Foundation. If the Program does not specify a version number of
522: this License, you may choose any version ever published by the Free Software
523: Foundation.
524:
525: @item
526: If you wish to incorporate parts of the Program into other free
527: programs whose distribution conditions are different, write to the author
528: to ask for permission. For software which is copyrighted by the Free
529: Software Foundation, write to the Free Software Foundation; we sometimes
530: make exceptions for this. Our decision will be guided by the two goals
531: of preserving the free status of all derivatives of our free software and
532: of promoting the sharing and reuse of software generally.
533:
534: @iftex
535: @heading NO WARRANTY
536: @end iftex
537: @ifinfo
538: @center NO WARRANTY
539: @end ifinfo
540:
541: @item
542: BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
543: FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
544: OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
545: PROVIDE THE PROGRAM ``AS IS'' WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
546: OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
547: MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
548: TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
549: PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
550: REPAIR OR CORRECTION.
551:
552: @item
553: IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
554: WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
555: REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
556: INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
557: OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
558: TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
559: YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
560: PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
561: POSSIBILITY OF SUCH DAMAGES.
562: @end enumerate
563:
564: @iftex
565: @heading END OF TERMS AND CONDITIONS
566: @end iftex
567: @ifinfo
568: @center END OF TERMS AND CONDITIONS
569: @end ifinfo
570:
571: @page
1.1.1.4 root 572: @unnumberedsec How to Apply These Terms to Your New Programs
1.1 root 573:
574: If you develop a new program, and you want it to be of the greatest
575: possible use to the public, the best way to achieve this is to make it
576: free software which everyone can redistribute and change under these terms.
577:
578: To do so, attach the following notices to the program. It is safest
579: to attach them to the start of each source file to most effectively
580: convey the exclusion of warranty; and each file should have at least
581: the ``copyright'' line and a pointer to where the full notice is found.
582:
583: @smallexample
1.1.1.5 root 584: @var{one line to give the program's name and a brief idea of what it does.}
1.1 root 585: Copyright (C) 19@var{yy} @var{name of author}
586:
1.1.1.5 root 587: This program is free software; you can redistribute it and/or modify
588: it under the terms of the GNU General Public License as published by
589: the Free Software Foundation; either version 2 of the License, or
590: (at your option) any later version.
1.1 root 591:
592: This program is distributed in the hope that it will be useful,
593: but WITHOUT ANY WARRANTY; without even the implied warranty of
594: MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
595: GNU General Public License for more details.
596:
597: You should have received a copy of the GNU General Public License
598: along with this program; if not, write to the Free Software
599: Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
600: @end smallexample
601:
602: Also add information on how to contact you by electronic and paper mail.
603:
604: If the program is interactive, make it output a short notice like this
605: when it starts in an interactive mode:
606:
607: @smallexample
608: Gnomovision version 69, Copyright (C) 19@var{yy} @var{name of author}
1.1.1.4 root 609: Gnomovision comes with ABSOLUTELY NO WARRANTY; for details
1.1.1.5 root 610: type `show w'.
611: This is free software, and you are welcome to redistribute it
612: under certain conditions; type `show c' for details.
1.1 root 613: @end smallexample
614:
615: The hypothetical commands @samp{show w} and @samp{show c} should show
616: the appropriate parts of the General Public License. Of course, the
617: commands you use may be called something other than @samp{show w} and
618: @samp{show c}; they could even be mouse-clicks or menu items---whatever
619: suits your program.
620:
621: You should also get your employer (if you work as a programmer) or your
622: school, if any, to sign a ``copyright disclaimer'' for the program, if
623: necessary. Here is a sample; alter the names:
624:
1.1.1.5 root 625: @smallexample
626: Yoyodyne, Inc., hereby disclaims all copyright interest in the program
627: `Gnomovision' (which makes passes at compilers) written by James Hacker.
1.1 root 628:
629: @var{signature of Ty Coon}, 1 April 1989
630: Ty Coon, President of Vice
1.1.1.5 root 631: @end smallexample
1.1 root 632:
633: This General Public License does not permit incorporating your program into
634: proprietary programs. If your program is a subroutine library, you may
635: consider it more useful to permit linking proprietary applications with the
636: library. If this is what you want to do, use the GNU Library General
637: Public License instead of this License.
638:
1.1.1.7 ! root 639: @node Contributors
1.1 root 640: @unnumbered Contributors to GNU CC
641: @cindex contributors
642:
643: In addition to Richard Stallman, several people have written parts
644: of GNU CC.
645:
646: @itemize @bullet
647: @item
648: The idea of using RTL and some of the optimization ideas came from the
1.1.1.2 root 649: program PO written at the University of Arizona by Jack Davidson and
1.1 root 650: Christopher Fraser. See ``Register Allocation and Exhaustive Peephole
651: Optimization'', Software Practice and Experience 14 (9), Sept. 1984,
652: 857-866.
653:
654: @item
655: Paul Rubin wrote most of the preprocessor.
656:
657: @item
658: Leonard Tower wrote parts of the parser, RTL generator, and RTL
659: definitions, and of the Vax machine description.
660:
661: @item
662: Ted Lemon wrote parts of the RTL reader and printer.
663:
664: @item
665: Jim Wilson implemented loop strength reduction and some other
666: loop optimizations.
667:
668: @item
669: Nobuyuki Hikichi of Software Research Associates, Tokyo, contributed
670: the support for the Sony NEWS machine.
671:
672: @item
673: Charles LaBrec contributed the support for the Integrated Solutions
674: 68020 system.
675:
676: @item
677: Michael Tiemann of Cygnus Support wrote the front end for C++, as well
678: as the support for inline functions and instruction scheduling. Also
679: the descriptions of the National Semiconductor 32000 series cpu, the
680: SPARC cpu and part of the Motorola 88000 cpu.
681:
682: @item
1.1.1.7 ! root 683: Gerald Baumgartner added the signature extension to the C++ front-end.
! 684:
! 685: @item
1.1 root 686: Jan Stein of the Chalmers Computer Society provided support for
687: Genix, as well as part of the 32000 machine description.
688:
689: @item
690: Randy Smith finished the Sun FPA support.
691:
692: @item
693: Robert Brown implemented the support for Encore 32000 systems.
694:
695: @item
696: David Kashtan of SRI adapted GNU CC to the Vomit-Making System (VMS).
697:
698: @item
699: Alex Crain provided changes for the 3b1.
700:
701: @item
702: Greg Satz and Chris Hanson assisted in making GNU CC work on HP-UX for
703: the 9000 series 300.
704:
705: @item
706: William Schelter did most of the work on the Intel 80386 support.
707:
708: @item
709: Christopher Smith did the port for Convex machines.
710:
711: @item
712: Paul Petersen wrote the machine description for the Alliant FX/8.
713:
714: @item
1.1.1.7 ! root 715: Dario Dariol contributed the four varieties of sample programs
! 716: that print a copy of their source.
! 717:
! 718: @item
1.1 root 719: Alain Lichnewsky ported GNU CC to the Mips cpu.
720:
721: @item
722: Devon Bowen, Dale Wiles and Kevin Zachmann ported GNU CC to the Tahoe.
723:
724: @item
725: Jonathan Stone wrote the machine description for the Pyramid computer.
726:
727: @item
1.1.1.3 root 728: Gary Miller ported GNU CC to Charles River Data Systems machines.
729:
730: @item
1.1.1.4 root 731: Richard Kenner of the New York University Ultracomputer Research
732: Laboratory wrote the machine descriptions for the AMD 29000, the DEC
733: Alpha, the IBM RT PC, and the IBM RS/6000 as well as the support for
734: instruction attributes. He also made changes to better support RISC
1.1 root 735: processors including changes to common subexpression elimination,
736: strength reduction, function calling sequence handling, and condition
737: code support, in addition to generalizing the code for frame pointer
738: elimination.
739:
740: @item
741: Richard Kenner and Michael Tiemann jointly developed reorg.c, the delay
742: slot scheduler.
743:
744: @item
745: Mike Meissner and Tom Wood of Data General finished the port to the
746: Motorola 88000.
747:
748: @item
749: Masanobu Yuhara of Fujitsu Laboratories implemented the machine
750: description for the Tron architecture (specifically, the Gmicro).
751:
752: @item
753: NeXT, Inc.@: donated the front end that supports the Objective C
754: language.
755: @c We need to be careful to make it clear that "Objective C"
756: @c is the name of a language, not that of a program or product.
757:
758: @item
759: James van Artsdalen wrote the code that makes efficient use of
760: the Intel 80387 register stack.
761:
762: @item
763: Mike Meissner at the Open Software Foundation finished the port to the
1.1.1.7 ! root 764: MIPS cpu, including adding ECOFF debug support, and worked on the
! 765: Intel port for the Intel 80386 cpu.
1.1 root 766:
767: @item
1.1.1.4 root 768: Ron Guilmette implemented the @code{protoize} and @code{unprotoize}
769: tools, the support for Dwarf symbolic debugging information, and much of
770: the support for System V Release 4. He has also worked heavily on the
771: Intel 386 and 860 support.
1.1.1.2 root 772:
773: @item
774: Torbjorn Granlund of the Swedish Institute of Computer Science
775: implemented multiply-by-constant optimization and better long long
776: support, and improved leaf function register allocation.
1.1.1.4 root 777:
778: @item
779: Mike Stump implemented the support for Elxsi 64 bit CPU.
1.1.1.5 root 780:
781: @item
782: John Wehle added the machine description for the Western Electric 32000
783: processor used in several 3b series machines (no relation to the
784: National Semiconductor 32000 processor).
785:
786: @ignore @c These features aren't advertised yet, since they don't fully work.
787: @item
788: Analog Devices helped implement the support for complex data types
789: and iterators.
790: @end ignore
791:
792: @item
793: Holger Teutsch provided the support for the Clipper cpu.
794:
795: @item
796: Kresten Krab Thorup wrote the run time support for the Objective C
797: language.
798:
799: @item
800: Stephen Moshier contributed the floating point emulator that assists in
801: cross-compilation and permits support for floating point numbers wider
802: than 64 bits.
803:
804: @item
1.1.1.6 root 805: David Edelsohn contributed the changes to RS/6000 port to make it
806: support the PowerPC and POWER2 architectures.
807:
808: @item
1.1.1.5 root 809: Steve Chamberlain wrote the support for the Hitachi SH processor.
1.1.1.6 root 810:
811: @item
812: Peter Schauer wrote the code to allow debugging to work on the Alpha.
1.1.1.7 ! root 813:
! 814: @item
! 815: Oliver M. Kellogg of Deutsche Aerospace contributed the port to the
! 816: MIL-STD-1750A.
1.1 root 817: @end itemize
818:
1.1.1.7 ! root 819: @node Funding
! 820: @chapter Funding Free Software
! 821:
! 822: If you want to have more free software a few years from now, it makes
! 823: sense for you to help encourage people to contribute funds for its
! 824: development. The most effective approach known is to encourage
! 825: commercial redistributors to donate.
! 826:
! 827: Users of free software systems can boost the pace of development by
! 828: encouraging for-a-fee distributors to donate part of their selling price
! 829: to free software developers---the Free Software Foundation, and others.
! 830:
! 831: The way to convince distributors to do this is to demand it and expect
! 832: it from them. So when you compare distributors, judge them partly by
! 833: how much they give to free software development. Show distributors
! 834: they must compete to be the one who gives the most.
! 835:
! 836: To make this approach work, you must insist on numbers that you can
! 837: compare, such as, ``We will donate ten dollars to the Frobnitz project
! 838: for each disk sold.'' Don't be satisfied with a vague promise, such as
! 839: ``A portion of the profits are donated,'' since it doesn't give a basis
! 840: for comparison.
! 841:
! 842: Even a precise fraction ``of the profits from this disk'' is not very
! 843: meaningful, since creative accounting and unrelated business decisions
! 844: can greatly alter what fraction of the sales price counts as profit.
! 845: If the price you pay is $50, ten percent of the profit is probably
! 846: less than a dollar; it might be a few cents, or nothing at all.
! 847:
! 848: Some redistributors do development work themselves. This is useful too;
! 849: but to keep everyone honest, you need to inquire how much they do, and
! 850: what kind. Some kinds of development make much more long-term
! 851: difference than others. For example, maintaining a separate version of
! 852: a program contributes very little; maintaining the standard version of a
! 853: program for the whole community contributes much. Easy new ports
! 854: contribute little, since someone else would surely do them; difficult
! 855: ports such as adding a new CPU to the GNU C compiler contribute more;
! 856: major new features or packages contribute the most.
! 857:
! 858: By establishing the idea that supporting further development is ``the
! 859: proper thing to do'' when distributing free software for a fee, we can
! 860: assure a steady flow of resources into making more free software.
! 861:
! 862: @display
! 863: Copyright (C) 1994 Free Software Foundation, Inc.
! 864: Verbatim copying and redistribution of this section is permitted
! 865: without royalty; alteration is not permitted.
! 866: @end display
! 867:
! 868: @node Look and Feel
1.1 root 869: @chapter Protect Your Freedom---Fight ``Look And Feel''
1.1.1.5 root 870: @c the above chapter heading overflows onto the next line. --mew 1/26/93
1.1 root 871:
872: @quotation
873: @i{This section is a political message from the League for Programming
1.1.1.7 ! root 874: Freedom to the users of GNU CC. We have included it here because the
! 875: issue of interface copyright is important to the GNU project.}
1.1 root 876: @end quotation
877:
1.1.1.7 ! root 878: Apple and Lotus have tried to create a new form of legal monopoly: a
! 879: copyright on a user interface.
! 880:
! 881: An interface is a kind of language---a set of conventions for
! 882: communication between two entities, human or machine. Until a few years
! 883: ago, the law seemed clear: interfaces were outside the domain of
! 884: copyright, so programmers could program freely and implement whatever
! 885: interface the users demanded. Imitating de-facto standard interfaces,
! 886: sometimes with improvements, was standard practice in the computer
! 887: field. These improvements, if accepted by the users, caught on and
! 888: became the norm; in this way, much progress took place.
! 889:
! 890: Computer users, and most software developers, were happy with this state
! 891: of affairs. However, large companies such as Apple and Lotus would
! 892: prefer a different system---one in which they can own interfaces and
! 893: thereby rid themselves of all serious competitors. They hope that
! 894: interface copyright will give them, in effect, monopolies on major
! 895: classes of software.
! 896:
! 897: Other large companies such as IBM and Digital also favor interface
! 898: monopolies, for the same reason: if languages become property, they
! 899: expect to own many de-facto standard languages. But Apple and Lotus are
! 900: the ones who have actually sued. Lotus has won lawsuits against two
! 901: small companies, which were thus put out of business. Then they sued
! 902: Borland; this case is now before the court of appeals. Apple's lawsuit
! 903: against HP and Microsoft is also being decided by an appeals court.
! 904: Widespread rumors that Apple had lost the case are untrue; as of July
! 905: 1994, the final outcome is unknown.
1.1 root 906:
1.1.1.7 ! root 907: If the monopolists get their way, they will hobble the software field:
1.1 root 908:
909: @itemize @bullet
910: @item
1.1.1.7 ! root 911: Gratuitous incompatibilities will burden users. Imagine if each car
! 912: manufacturer had to design a different way to start, stop, and steer a
! 913: car.
1.1 root 914:
915: @item
1.1.1.7 ! root 916: Users will be ``locked in'' to whichever interface they learn; then they
! 917: will be prisoners of one supplier, who will charge a monopolistic price.
1.1 root 918:
919: @item
920: Large companies have an unfair advantage wherever lawsuits become
1.1.1.7 ! root 921: commonplace. Since they can afford to sue, they can intimidate smaller
! 922: developers with threats even when they don't really have a case.
1.1 root 923:
924: @item
1.1.1.7 ! root 925: Interface improvements will come slower, since incremental evolution
! 926: through creative partial imitation will no longer occur.
! 927: @end itemize
! 928:
! 929: If interface monopolies are accepted, other large companies are waiting
! 930: to grab theirs:
1.1 root 931:
1.1.1.7 ! root 932: @itemize @bullet
1.1 root 933: @item
1.1.1.7 ! root 934: Adobe is expected to claim a monopoly on the interfaces of various popular
! 935: application programs, if Borland's appeal against Lotus fails.
1.1 root 936:
937: @item
1.1.1.7 ! root 938: Open Computing magazine reported a Microsoft vice president as threatening
! 939: to sue people who copy the interface of Windows.
1.1 root 940: @end itemize
941:
1.1.1.7 ! root 942: Users invest a great deal of time and money in learning to use computer
! 943: interfaces. Far more, in fact, than software developers invest in
! 944: developing @emph{and even implementing} the interfaces. Whoever can own
! 945: an interface, has made its users into captives, and misappropriated
! 946: their investment.
! 947:
! 948: To protect our freedom from monopolies like these, a group of
! 949: programmers and users have formed a grass-roots political organization,
! 950: the League for Programming Freedom.
! 951:
! 952: The purpose of the League is to oppose monopolistic practices such as
! 953: interface copyright and software patents. The League calls for a return
! 954: to the legal policies of the recent past, in which programmers could
! 955: program freely. The League is not concerned with free software as an
! 956: issue, and is not affiliated with the Free Software Foundation.
! 957:
! 958: The League's activities include publicizing the issue, as is being done
! 959: here, and filing friend-of-the-court briefs on behalf of defendants sued
! 960: by monopolists. Recently the League filed a friend-of-the-court brief
! 961: for Borland in its appeal against Lotus.
1.1 root 962:
963: The League's membership rolls include John McCarthy, inventor of Lisp,
1.1.1.7 ! root 964: Marvin Minsky, founder of the MIT Artificial Intelligence lab, Guy L.
1.1 root 965: Steele, Jr., author of well-known books on Lisp and C, as well as
966: Richard Stallman, the developer of GNU CC. Please join and add your
967: name to the list. Membership dues in the League are $42 per year for
968: programmers, managers and professionals; $10.50 for students; $21 for
969: others.
970:
1.1.1.7 ! root 971: Activist members are especially important, but members who have no time
! 972: to give are also important. Surveys at major ACM conferences have
! 973: indicated a vast majority of attendees agree with the League. If just
! 974: ten percent of the programmers who agree with the League join the
! 975: League, we will probably triumph.
1.1 root 976:
1.1.1.4 root 977: To join, or for more information, phone (617) 243-4091 or write to:
1.1 root 978:
979: @display
980: League for Programming Freedom
981: 1 Kendall Square #143
982: P.O. Box 9171
983: Cambridge, MA 02139
984: @end display
985:
1.1.1.7 ! root 986: You can also send electronic mail to @code{lpf@@uunet.uu.net}.
1.1 root 987:
1.1.1.7 ! root 988: In addition to joining the League, here are some suggestions from the
! 989: League for other things you can do to protect your freedom to write
! 990: programs:
1.1 root 991:
992: @itemize @bullet
993: @item
1.1.1.7 ! root 994: Tell your friends and colleagues about this issue and how it threatens
! 995: to ruin the computer industry.
1.1 root 996:
997: @item
1.1.1.7 ! root 998: Mention that you are a League member in your @file{.signature}, and
! 999: mention the League's email address for inquiries.
1.1 root 1000:
1001: @item
1.1.1.7 ! root 1002: Ask the companies you consider working for or working with to make
! 1003: statements against software monopolies, and give preference to those
! 1004: that do.
1.1 root 1005:
1006: @item
1.1.1.7 ! root 1007: When employers ask you to sign contracts giving them copyright or patent
! 1008: rights, insist on clauses saying they can use these rights only
! 1009: defensively. Don't rely on ``company policy,'' since that can change at
! 1010: any time; don't rely on an individual executive's private word, since
! 1011: that person may be replaced. Get a commitment just as binding as the
! 1012: commitment they get from you.
1.1 root 1013:
1014: @item
1015: Write to Congress to explain the importance of this issue.
1016:
1017: @display
1018: House Subcommittee on Intellectual Property
1019: 2137 Rayburn Bldg
1020: Washington, DC 20515
1021:
1022: Senate Subcommittee on Patents, Trademarks and Copyrights
1023: United States Senate
1024: Washington, DC 20510
1025: @end display
1026:
1027: (These committees have received lots of mail already; let's give them
1028: even more.)
1029: @end itemize
1030:
1.1.1.7 ! root 1031: Democracy means nothing if you don't use it. Stand up and be counted!
1.1.1.5 root 1032: @ifset USING
1033: @node G++ and GCC
1034: @chapter Compile C, C++, or Objective C
1035:
1036: @cindex Objective C
1037: The C, C++, and Objective C versions of the compiler are integrated; the
1038: GNU C compiler can compile programs written in C, C++, or Objective C.
1039:
1040: @cindex GCC
1041: ``GCC'' is a common shorthand term for the GNU C compiler. This is both
1042: the most general name for the compiler, and the name used when the
1043: emphasis is on compiling C programs.
1044:
1045: @cindex C++
1046: @cindex G++
1047: When referring to C++ compilation, it is usual to call the compiler
1048: ``G++''. Since there is only one compiler, it is also accurate to call
1049: it ``GCC'' no matter what the language context; however, the term
1050: ``G++'' is more useful when the emphasis is on compiling C++ programs.
1051:
1.1.1.7 ! root 1052: We use the name ``GNU CC'' to refer to the compilation system as a
! 1053: whole, and more specifically to the language-independent part of the
! 1054: compiler. For example, we refer to the optimization options as
! 1055: affecting the behavior of ``GNU CC'' or sometimes just ``the compiler''.
! 1056:
! 1057: Front ends for other languages, such as Ada 9X, Fortran, Modula-3, and
! 1058: Pascal, are under development. These front-ends, like that for C++, are
! 1059: built in subdirectories of GNU CC and link to it. The result is an
! 1060: integrated compiler that can compile programs written in C, C++,
! 1061: Objective C, or any of the languages for which you have installed front
! 1062: ends.
! 1063:
! 1064: In this manual, we only discuss the options for the C, Objective-C, and
! 1065: C++ compilers and those of the GNU CC core. Consult the documentation
! 1066: of the other front ends for the options to use when compiling programs
! 1067: written in other languages.
! 1068:
1.1.1.5 root 1069: @cindex compiler compared to C++ preprocessor
1070: @cindex intermediate C version, nonexistent
1071: @cindex C intermediate output, nonexistent
1072: G++ is a @emph{compiler}, not merely a preprocessor. G++ builds object
1073: code directly from your C++ program source. There is no intermediate C
1074: version of the program. (By contrast, for example, some other
1075: implementations use a program that generates a C program from your C++
1076: source.) Avoiding an intermediate C representation of the program means
1077: that you get better object code, and better debugging information. The
1078: GNU debugger, GDB, works with this information in the object code to
1079: give you comprehensive C++ source-level editing capabilities
1080: (@pxref{C,,C and C++,gdb.info, Debugging with GDB}).
1081:
1082: @c FIXME! Someone who knows something about Objective C ought to put in
1083: @c a paragraph or two about it here, and move the index entry down when
1084: @c there is more to point to than the general mention in the 1st par.
1085:
1.1 root 1086: @include invoke.texi
1087:
1.1.1.3 root 1088: @include install.texi
1089:
1090: @include extend.texi
1091:
1092: @node Trouble
1093: @chapter Known Causes of Trouble with GNU CC
1094: @cindex bugs, known
1095: @cindex installation trouble
1096: @cindex known causes of trouble
1.1 root 1097:
1.1.1.3 root 1098: This section describes known problems that affect users of GNU CC. Most
1099: of these are not GNU CC bugs per se---if they were, we would fix them.
1100: But the result for a user may be like the result of a bug.
1101:
1102: Some of these problems are due to bugs in other software, some are
1103: missing features that are too much work to add, and some are places
1104: where people's opinions differ as to what is best.
1.1 root 1105:
1106: @menu
1.1.1.3 root 1107: * Actual Bugs:: Bugs we will fix later.
1108: * Installation Problems:: Problems that manifest when you install GNU CC.
1109: * Cross-Compiler Problems:: Common problems of cross compiling with GNU CC.
1110: * Interoperation:: Problems using GNU CC with other compilers,
1111: and with certain linkers, assemblers and debuggers.
1.1.1.5 root 1112: * External Bugs:: Problems compiling certain programs.
1.1.1.3 root 1113: * Incompatibilities:: GNU CC is incompatible with traditional C.
1.1.1.6 root 1114: * Fixed Headers:: GNU C uses corrected versions of system header files.
1115: This is necessary, but doesn't always work smoothly.
1.1.1.3 root 1116: * Disappointments:: Regrettable things we can't change, but not quite bugs.
1.1.1.5 root 1117: * C++ Misunderstandings:: Common misunderstandings with GNU C++.
1.1.1.4 root 1118: * Protoize Caveats:: Things to watch out for when using @code{protoize}.
1.1.1.3 root 1119: * Non-bugs:: Things we think are right, but some others disagree.
1.1.1.5 root 1120: * Warnings and Errors:: Which problems in your code get warnings,
1121: and which get errors.
1.1 root 1122: @end menu
1123:
1.1.1.3 root 1124: @node Actual Bugs
1125: @section Actual Bugs We Haven't Fixed Yet
1.1 root 1126:
1.1.1.3 root 1127: @itemize @bullet
1.1 root 1128: @item
1.1.1.5 root 1129: The @code{fixincludes} script interacts badly with automounters; if the
1130: directory of system header files is automounted, it tends to be
1131: unmounted while @code{fixincludes} is running. This would seem to be a
1132: bug in the automounter. We don't know any good way to work around it.
1133:
1134: @item
1.1.1.6 root 1135: The @code{fixproto} script will sometimes add prototypes for the
1136: @code{sigsetjmp} and @code{siglongjmp} functions that reference the
1137: @code{jmp_buf} type before that type is defined. To work around this,
1138: edit the offending file and place the typedef in front of the
1139: prototypes.
1140:
1141: @item
1.1.1.7 ! root 1142: There are several obscure case of mis-using struct, union, and
! 1143: enum tags that are not detected as errors by the compiler.
! 1144:
! 1145: @item
! 1146: When @samp{-pedantic-errors} is specified, GNU C will incorrectly give
! 1147: an error message when a function name is specified in an expression
! 1148: involving the comma operator.
! 1149:
! 1150: @item
1.1.1.4 root 1151: Loop unrolling doesn't work properly for certain C++ programs. This is
1.1.1.7 ! root 1152: a bug in the C++ front end. It sometimes emits incorrect debug info, and
! 1153: the loop unrolling code is unable to recover from this error.
1.1.1.3 root 1154: @end itemize
1.1 root 1155:
1.1.1.3 root 1156: @node Installation Problems
1157: @section Installation Problems
1.1.1.4 root 1158:
1.1.1.3 root 1159: This is a list of problems (and some apparent problems which don't
1160: really mean anything is wrong) that show up during installation of GNU
1161: CC.
1.1 root 1162:
1163: @itemize @bullet
1164: @item
1.1.1.3 root 1165: On certain systems, defining certain environment variables such as
1166: @code{CC} can interfere with the functioning of @code{make}.
1.1 root 1167:
1168: @item
1.1.1.3 root 1169: If you encounter seemingly strange errors when trying to build the
1170: compiler in a directory other than the source directory, it could be
1171: because you have previously configured the compiler in the source
1172: directory. Make sure you have done all the necessary preparations.
1173: @xref{Other Dir}.
1.1 root 1174:
1175: @item
1.1.1.6 root 1176: If you build GNU CC on a BSD system using a directory stored in a System
1177: V file system, problems may occur in running @code{fixincludes} if the
1178: System V file system doesn't support symbolic links. These problems
1179: result in a failure to fix the declaration of @code{size_t} in
1180: @file{sys/types.h}. If you find that @code{size_t} is a signed type and
1181: that type mismatches occur, this could be the cause.
1182:
1183: The solution is not to use such a directory for building GNU CC.
1184:
1185: @item
1.1.1.3 root 1186: In previous versions of GNU CC, the @code{gcc} driver program looked for
1.1.1.5 root 1187: @code{as} and @code{ld} in various places; for example, in files
1188: beginning with @file{/usr/local/lib/gcc-}. GNU CC version 2 looks for
1189: them in the directory
1190: @file{/usr/local/lib/gcc-lib/@var{target}/@var{version}}.
1.1 root 1191:
1.1.1.3 root 1192: Thus, to use a version of @code{as} or @code{ld} that is not the system
1193: default, for example @code{gas} or GNU @code{ld}, you must put them in
1194: that directory (or make links to them from that directory).
1.1 root 1195:
1196: @item
1.1.1.3 root 1197: Some commands executed when making the compiler may fail (return a
1198: non-zero status) and be ignored by @code{make}. These failures, which
1199: are often due to files that were not found, are expected, and can safely
1200: be ignored.
1.1 root 1201:
1202: @item
1.1.1.3 root 1203: It is normal to have warnings in compiling certain files about
1204: unreachable code and about enumeration type clashes. These files' names
1.1.1.5 root 1205: begin with @samp{insn-}. Also, @file{real.c} may get some warnings that
1206: you can ignore.
1.1 root 1207:
1208: @item
1.1.1.3 root 1209: Sometimes @code{make} recompiles parts of the compiler when installing
1210: the compiler. In one case, this was traced down to a bug in
1211: @code{make}. Either ignore the problem or switch to GNU Make.
1.1 root 1212:
1.1.1.4 root 1213: @item
1.1.1.5 root 1214: If you have installed a program known as purify, you may find that it
1215: causes errors while linking @code{enquire}, which is part of building
1216: GNU CC. The fix is to get rid of the file @code{real-ld} which purify
1217: installs---so that GNU CC won't try to use it.
1218:
1219: @item
1220: On Linux SLS 1.01, there is a problem with @file{libc.a}: it does not
1221: contain the obstack functions. However, GNU CC assumes that the obstack
1222: functions are in @file{libc.a} when it is the GNU C library. To work
1223: around this problem, change the @code{__GNU_LIBRARY__} conditional
1224: around line 31 to @samp{#if 1}.
1225:
1226: @item
1.1.1.4 root 1227: On some 386 systems, building the compiler never finishes because
1228: @code{enquire} hangs due to a hardware problem in the motherboard---it
1229: reports floating point exceptions to the kernel incorrectly. You can
1230: install GNU CC except for @file{float.h} by patching out the command to
1231: run @code{enquire}. You may also be able to fix the problem for real by
1232: getting a replacement motherboard. This problem was observed in
1233: Revision E of the Micronics motherboard, and is fixed in Revision F.
1.1.1.6 root 1234: It has also been observed in the MYLEX MXA-33 motherboard.
1235:
1236: If you encounter this problem, you may also want to consider removing
1237: the FPU from the socket during the compilation. Alternatively, if you
1238: are running SCO Unix, you can reboot and force the FPU to be ignored.
1239: To do this, type @samp{hd(40)unix auto ignorefpu}.
1.1.1.4 root 1240:
1241: @item
1242: On some 386 systems, GNU CC crashes trying to compile @file{enquire.c}.
1243: This happens on machines that don't have a 387 FPU chip. On 386
1244: machines, the system kernel is supposed to emulate the 387 when you
1245: don't have one. The crash is due to a bug in the emulator.
1246:
1247: One of these systems is the Unix from Interactive Systems: 386/ix.
1248: On this system, an alternate emulator is provided, and it does work.
1249: To use it, execute this command as super-user:
1250:
1251: @example
1252: ln /etc/emulator.rel1 /etc/emulator
1253: @end example
1254:
1255: @noindent
1256: and then reboot the system. (The default emulator file remains present
1257: under the name @file{emulator.dflt}.)
1258:
1.1.1.5 root 1259: Try using @file{/etc/emulator.att}, if you have such a problem on the
1260: SCO system.
1.1.1.4 root 1261:
1262: Another system which has this problem is Esix. We don't know whether it
1263: has an alternate emulator that works.
1264:
1.1.1.5 root 1265: On NetBSD 0.8, a similar problem manifests itself as these error messages:
1266:
1267: @example
1268: enquire.c: In function `fprop':
1269: enquire.c:2328: floating overflow
1270: @end example
1271:
1.1.1.6 root 1272: @item
1273: On SCO systems, when compiling GNU CC with the system's compiler,
1274: do not use @samp{-O}. Some versions of the system's compiler miscompile
1275: GNU CC with @samp{-O}.
1276:
1.1.1.3 root 1277: @cindex @code{genflags}, crash on Sun 4
1.1 root 1278: @item
1.1.1.3 root 1279: Sometimes on a Sun 4 you may observe a crash in the program
1.1.1.4 root 1280: @code{genflags} or @code{genoutput} while building GNU CC. This is said to
1.1.1.3 root 1281: be due to a bug in @code{sh}. You can probably get around it by running
1282: @code{genflags} or @code{genoutput} manually and then retrying the
1283: @code{make}.
1.1 root 1284:
1.1.1.3 root 1285: @item
1.1.1.5 root 1286: On Solaris 2, executables of GNU CC version 2.0.2 are commonly
1287: available, but they have a bug that shows up when compiling current
1288: versions of GNU CC: undefined symbol errors occur during assembly if you
1289: use @samp{-g}.
1290:
1291: The solution is to compile the current version of GNU CC without
1292: @samp{-g}. That makes a working compiler which you can use to recompile
1293: with @samp{-g}.
1294:
1295: @item
1.1.1.6 root 1296: Solaris 2 comes with a number of optional OS packages. Some of these
1.1.1.5 root 1297: packages are needed to use GNU CC fully. If you did not install all
1298: optional packages when installing Solaris, you will need to verify that
1.1.1.6 root 1299: the packages that GNU CC needs are installed.
1.1.1.5 root 1300:
1.1.1.6 root 1301: To check whether an optional package is installed, use
1.1.1.5 root 1302: the @code{pkginfo} command. To add an optional package, use the
1303: @code{pkgadd} command. For further details, see the Solaris
1304: documentation.
1305:
1.1.1.6 root 1306: For Solaris 2.0 and 2.1, GNU CC needs six packages: @samp{SUNWarc},
1307: @samp{SUNWbtool}, @samp{SUNWesu}, @samp{SUNWhea}, @samp{SUNWlibm}, and
1308: @samp{SUNWtoo}.
1309:
1310: For Solaris 2.2, GNU CC needs an additional seventh package: @samp{SUNWsprot}.
1311:
1.1.1.5 root 1312: @item
1313: On Solaris 2, trying to use the linker and other tools in
1314: @file{/usr/ucb} to install GNU CC has been observed to cause trouble.
1315: For example, the linker may hang indefinitely. The fix is to remove
1316: @file{/usr/ucb} from your @code{PATH}.
1317:
1318: @item
1.1.1.3 root 1319: If you use the 1.31 version of the MIPS assembler (such as was shipped
1320: with Ultrix 3.1), you will need to use the -fno-delayed-branch switch
1321: when optimizing floating point code. Otherwise, the assembler will
1322: complain when the GCC compiler fills a branch delay slot with a
1.1.1.6 root 1323: floating point instruction, such as @code{add.d}.
1.1.1.3 root 1324:
1325: @item
1.1.1.5 root 1326: If on a MIPS system you get an error message saying ``does not have gp
1327: sections for all it's [sic] sectons [sic]'', don't worry about it. This
1328: happens whenever you use GAS with the MIPS linker, but there is not
1329: really anything wrong, and it is okay to use the output file. You can
1330: stop such warnings by installing the GNU linker.
1331:
1332: It would be nice to extend GAS to produce the gp tables, but they are
1333: optional, and there should not be a warning about their absence.
1334:
1335: @item
1.1.1.6 root 1336: In Ultrix 4.0 on the MIPS machine, @file{stdio.h} does not work with GNU
1337: CC at all unless it has been fixed with @code{fixincludes}. This causes
1338: problems in building GNU CC. Once GNU CC is installed, the problems go
1339: away.
1340:
1341: To work around this problem, when making the stage 1 compiler, specify
1342: this option to Make:
1343:
1344: @example
1345: GCC_FOR_TARGET="./xgcc -B./ -I./include"
1346: @end example
1347:
1348: When making stage 2 and stage 3, specify this option:
1349:
1350: @example
1351: CFLAGS="-g -I./include"
1352: @end example
1353:
1354: @item
1.1.1.3 root 1355: Users have reported some problems with version 2.0 of the MIPS
1356: compiler tools that were shipped with Ultrix 4.1. Version 2.10
1357: which came with Ultrix 4.2 seems to work fine.
1.1 root 1358:
1.1.1.7 ! root 1359: Users have also reported some problems with version 2.20 of the
! 1360: MIPS compiler tools that were shipped with RISC/os 4.x. The earlier
! 1361: version 2.11 seems to work fine.
! 1362:
1.1 root 1363: @item
1.1.1.4 root 1364: Some versions of the MIPS linker will issue an assertion failure
1365: when linking code that uses @code{alloca} against shared
1366: libraries on RISC-OS 5.0, and DEC's OSF/1 systems. This is a bug
1367: in the linker, that is supposed to be fixed in future revisions.
1.1.1.6 root 1368: To protect against this, GNU CC passes @samp{-non_shared} to the
1.1.1.4 root 1369: linker unless you pass an explicit @samp{-shared} or
1370: @samp{-call_shared} switch.
1371:
1372: @item
1373: On System V release 3, you may get this error message
1374: while linking:
1375:
1376: @smallexample
1377: ld fatal: failed to write symbol name @var{something}
1378: in strings table for file @var{whatever}
1379: @end smallexample
1380:
1.1.1.5 root 1381: This probably indicates that the disk is full or your ULIMIT won't allow
1.1.1.4 root 1382: the file to be as large as it needs to be.
1383:
1.1.1.5 root 1384: This problem can also result because the kernel parameter @code{MAXUMEM}
1385: is too small. If so, you must regenerate the kernel and make the value
1386: much larger. The default value is reported to be 1024; a value of 32768
1387: is said to work. Smaller values may also work.
1388:
1389: @item
1390: On System V, if you get an error like this,
1391:
1392: @example
1393: /usr/local/lib/bison.simple: In function `yyparse':
1394: /usr/local/lib/bison.simple:625: virtual memory exhausted
1395: @end example
1396:
1397: @noindent
1398: that too indicates a problem with disk space, ULIMIT, or @code{MAXUMEM}.
1399:
1400: @item
1401: Current GNU CC versions probably do not work on version 2 of the NeXT
1402: operating system.
1403:
1404: @item
1.1.1.6 root 1405: On NeXTStep 3.0, the Objective C compiler does not work, due,
1406: apparently, to a kernel bug that it happens to trigger. This problem
1407: does not happen on 3.1.
1408:
1409: @item
1.1.1.5 root 1410: On the Tower models 4@var{n}0 and 6@var{n}0, by default a process is not
1411: allowed to have more than one megabyte of memory. GNU CC cannot compile
1412: itself (or many other programs) with @samp{-O} in that much memory.
1413:
1414: To solve this problem, reconfigure the kernel adding the following line
1415: to the configuration file:
1416:
1417: @smallexample
1418: MAXUMEM = 4096
1419: @end smallexample
1420:
1.1.1.4 root 1421: @item
1.1.1.3 root 1422: On HP 9000 series 300 or 400 running HP-UX release 8.0, there is a bug
1423: in the assembler that must be fixed before GNU CC can be built. This
1424: bug manifests itself during the first stage of compilation, while
1425: building @file{libgcc2.a}:
1.1 root 1426:
1.1.1.5 root 1427: @smallexample
1.1.1.3 root 1428: _floatdisf
1429: cc1: warning: `-g' option not supported on this version of GCC
1430: cc1: warning: `-g1' option not supported on this version of GCC
1.1.1.5 root 1431: ./xgcc: Internal compiler error: program as got fatal signal 11
1432: @end smallexample
1.1 root 1433:
1.1.1.3 root 1434: A patched version of the assembler is available by anonymous ftp from
1435: @code{altdorf.ai.mit.edu} as the file
1436: @file{archive/cph/hpux-8.0-assembler}. If you have HP software support,
1437: the patch can also be obtained directly from HP, as described in the
1438: following note:
1.1 root 1439:
1.1.1.3 root 1440: @quotation
1441: This is the patched assembler, to patch SR#1653-010439, where the
1442: assembler aborts on floating point constants.
1.1 root 1443:
1.1.1.3 root 1444: The bug is not really in the assembler, but in the shared library
1445: version of the function ``cvtnum(3c)''. The bug on ``cvtnum(3c)'' is
1446: SR#4701-078451. Anyway, the attached assembler uses the archive
1447: library version of ``cvtnum(3c)'' and thus does not exhibit the bug.
1448: @end quotation
1.1 root 1449:
1.1.1.7 ! root 1450: This patch is also known as PHCO_4484.
1.1 root 1451:
1.1.1.3 root 1452: @item
1.1.1.6 root 1453: On HP-UX version 8.05, but not on 8.07 or more recent versions,
1454: the @code{fixproto} shell script triggers a bug in the system shell.
1455: If you encounter this problem, upgrade your operating system or
1456: use BASH (the GNU shell) to run @code{fixproto}.
1.1 root 1457:
1.1.1.3 root 1458: @item
1459: Some versions of the Pyramid C compiler are reported to be unable to
1460: compile GNU CC. You must use an older version of GNU CC for
1461: bootstrapping. One indication of this problem is if you get a crash
1462: when GNU CC compiles the function @code{muldi3} in file @file{libgcc2.c}.
1.1 root 1463:
1.1.1.3 root 1464: You may be able to succeed by getting GNU CC version 1, installing it,
1465: and using it to compile GNU CC version 2. The bug in the Pyramid C
1466: compiler does not seem to affect GNU CC version 1.
1.1 root 1467:
1.1.1.3 root 1468: @item
1.1.1.5 root 1469: There may be similar problems on System V Release 3.1 on 386 systems.
1.1 root 1470:
1471: @item
1.1.1.6 root 1472: On the Intel Paragon (an i860 machine), if you are using operating
1473: system version 1.0, you will get warnings or errors about redefinition
1474: of @code{va_arg} when you build GNU CC.
1475:
1476: If this happens, then you need to link most programs with the library
1477: @file{iclib.a}. You must also modify @file{stdio.h} as follows: before
1478: the lines
1479:
1480: @example
1481: #if defined(__i860__) && !defined(_VA_LIST)
1482: #include <va_list.h>
1483: @end example
1484:
1485: @noindent
1486: insert the line
1487:
1488: @example
1489: #if __PGC__
1490: @end example
1491:
1492: @noindent
1493: and after the lines
1494:
1495: @example
1496: extern int vprintf(const char *, va_list );
1497: extern int vsprintf(char *, const char *, va_list );
1498: #endif
1499: @end example
1500:
1501: @noindent
1502: insert the line
1503:
1504: @example
1505: #endif /* __PGC__ */
1506: @end example
1507:
1508: These problems don't exist in operating system version 1.1.
1509:
1510: @item
1.1.1.3 root 1511: On the Altos 3068, programs compiled with GNU CC won't work unless you
1512: fix a kernel bug. This happens using system versions V.2.2 1.0gT1 and
1513: V.2.2 1.0e and perhaps later versions as well. See the file
1514: @file{README.ALTOS}.
1.1.1.4 root 1515:
1516: @item
1517: You will get several sorts of compilation and linking errors on the
1.1.1.7 ! root 1518: we32k if you don't follow the special instructions. @xref{Configurations}.
! 1519:
! 1520: @item
! 1521: A bug in the HP-UX 8.05 (and earlier) shell will cause the fixproto
! 1522: program to report an error of the form:
! 1523:
! 1524: @example
! 1525: ./fixproto: sh internal 1K buffer overflow
! 1526: @end example
! 1527:
! 1528: To fix this, change the first line of the fixproto script to look like:
! 1529:
! 1530: @example
! 1531: #!/bin/ksh
! 1532: @end example
1.1.1.3 root 1533: @end itemize
1.1 root 1534:
1.1.1.3 root 1535: @node Cross-Compiler Problems
1536: @section Cross-Compiler Problems
1.1 root 1537:
1.1.1.5 root 1538: You may run into problems with cross compilation on certain machines,
1539: for several reasons.
1540:
1.1.1.3 root 1541: @itemize @bullet
1542: @item
1543: Cross compilation can run into trouble for certain machines because
1544: some target machines' assemblers require floating point numbers to be
1545: written as @emph{integer} constants in certain contexts.
1.1 root 1546:
1.1.1.3 root 1547: The compiler writes these integer constants by examining the floating
1548: point value as an integer and printing that integer, because this is
1549: simple to write and independent of the details of the floating point
1550: representation. But this does not work if the compiler is running on
1551: a different machine with an incompatible floating point format, or
1552: even a different byte-ordering.
1.1 root 1553:
1.1.1.3 root 1554: In addition, correct constant folding of floating point values
1555: requires representing them in the target machine's format.
1556: (The C standard does not quite require this, but in practice
1557: it is the only way to win.)
1.1.1.2 root 1558:
1.1.1.3 root 1559: It is now possible to overcome these problems by defining macros such
1560: as @code{REAL_VALUE_TYPE}. But doing so is a substantial amount of
1.1.1.5 root 1561: work for each target machine.
1562: @ifset INTERNALS
1563: @xref{Cross-compilation}.
1.1.1.3 root 1564: @end ifset
1565: @ifclear INTERNALS
1.1.1.5 root 1566: @xref{Cross-compilation,,Cross Compilation and Floating Point Format,
1567: gcc.info, Using and Porting GCC}.
1.1.1.3 root 1568: @end ifclear
1.1 root 1569:
1570: @item
1.1.1.3 root 1571: At present, the program @file{mips-tfile} which adds debug
1572: support to object files on MIPS systems does not work in a cross
1573: compile environment.
1574: @end itemize
1.1 root 1575:
1.1.1.3 root 1576: @node Interoperation
1577: @section Interoperation
1.1 root 1578:
1.1.1.3 root 1579: This section lists various difficulties encountered in using GNU C or
1.1.1.4 root 1580: GNU C++ together with other compilers or with the assemblers, linkers,
1581: libraries and debuggers on certain systems.
1.1 root 1582:
1.1.1.3 root 1583: @itemize @bullet
1584: @item
1.1.1.7 ! root 1585: Objective C does not work on the RS/6000.
1.1 root 1586:
1587: @item
1.1.1.3 root 1588: GNU C++ does not do name mangling in the same way as other C++
1589: compilers. This means that object files compiled with one compiler
1590: cannot be used with another.
1.1 root 1591:
1.1.1.5 root 1592: This effect is intentional, to protect you from more subtle problems.
1593: Compilers differ as to many internal details of C++ implementation,
1594: including: how class instances are laid out, how multiple inheritance is
1595: implemented, and how virtual function calls are handled. If the name
1596: encoding were made the same, your programs would link against libraries
1597: provided from other compilers---but the programs would then crash when
1598: run. Incompatible libraries are then detected at link time, rather than
1599: at run time.
1.1 root 1600:
1.1.1.3 root 1601: @item
1602: Older GDB versions sometimes fail to read the output of GNU CC version
1603: 2. If you have trouble, get GDB version 4.4 or later.
1.1 root 1604:
1605: @item
1.1.1.3 root 1606: @cindex DBX
1607: DBX rejects some files produced by GNU CC, though it accepts similar
1608: constructs in output from PCC. Until someone can supply a coherent
1609: description of what is valid DBX input and what is not, there is
1610: nothing I can do about these problems. You are on your own.
1.1 root 1611:
1.1.1.3 root 1612: @item
1613: The GNU assembler (GAS) does not support PIC. To generate PIC code, you
1614: must use some other assembler, such as @file{/bin/as}.
1.1 root 1615:
1.1.1.3 root 1616: @item
1.1.1.6 root 1617: On some BSD systems, including some versions of Ultrix, use of profiling
1.1.1.3 root 1618: causes static variable destructors (currently used only in C++) not to
1619: be run.
1.1 root 1620:
1.1.1.4 root 1621: @item
1622: Use of @samp{-I/usr/include} may cause trouble.
1623:
1624: Many systems come with header files that won't work with GNU CC unless
1625: corrected by @code{fixincludes}. The corrected header files go in a new
1626: directory; GNU CC searches this directory before @file{/usr/include}.
1627: If you use @samp{-I/usr/include}, this tells GNU CC to search
1628: @file{/usr/include} earlier on, before the corrected headers. The
1629: result is that you get the uncorrected header files.
1630:
1.1.1.5 root 1631: Instead, you should use these options (when compiling C programs):
1.1.1.4 root 1632:
1.1.1.5 root 1633: @smallexample
1.1.1.4 root 1634: -I/usr/local/lib/gcc-lib/@var{target}/@var{version}/include -I/usr/include
1.1.1.5 root 1635: @end smallexample
1636:
1637: For C++ programs, GNU CC also uses a special directory that defines C++
1638: interfaces to standard C subroutines. This directory is meant to be
1639: searched @emph{before} other standard include directories, so that it
1640: takes precedence. If you are compiling C++ programs and specifying
1641: include directories explicitly, use this option first, then the two
1642: options above:
1643:
1644: @example
1645: -I/usr/local/lib/g++-include
1.1.1.4 root 1646: @end example
1647:
1.1.1.3 root 1648: @ignore
1649: @cindex @code{vfork}, for the Sun-4
1650: @item
1651: There is a bug in @code{vfork} on the Sun-4 which causes the registers
1652: of the child process to clobber those of the parent. Because of this,
1653: programs that call @code{vfork} are likely to lose when compiled
1654: optimized with GNU CC when the child code alters registers which contain
1655: C variables in the parent. This affects variables which are live in the
1656: parent across the call to @code{vfork}.
1.1 root 1657:
1.1.1.3 root 1658: If you encounter this, you can work around the problem by declaring
1659: variables @code{volatile} in the function that calls @code{vfork}, until
1660: the problem goes away, or by not declaring them @code{register} and not
1661: using @samp{-O} for those source files.
1662: @end ignore
1.1 root 1663:
1.1.1.3 root 1664: @item
1.1.1.6 root 1665: On some SGI systems, when you use @samp{-lgl_s} as an option,
1666: it gets translated magically to @samp{-lgl_s -lX11_s -lc_s}.
1667: Naturally, this does not happen when you use GNU CC.
1668: You must specify all three options explicitly.
1669:
1670: @item
1.1.1.4 root 1671: On a Sparc, GNU CC aligns all values of type @code{double} on an 8-byte
1672: boundary, and it expects every @code{double} to be so aligned. The Sun
1673: compiler usually gives @code{double} values 8-byte alignment, with one
1674: exception: function arguments of type @code{double} may not be aligned.
1675:
1676: As a result, if a function compiled with Sun CC takes the address of an
1677: argument of type @code{double} and passes this pointer of type
1678: @code{double *} to a function compiled with GNU CC, dereferencing the
1679: pointer may cause a fatal signal.
1680:
1681: One way to solve this problem is to compile your entire program with GNU
1682: CC. Another solution is to modify the function that is compiled with
1683: Sun CC to copy the argument into a local variable; local variables
1684: are always properly aligned. A third solution is to modify the function
1685: that uses the pointer to dereference it via the following function
1686: @code{access_double} instead of directly with @samp{*}:
1687:
1.1.1.5 root 1688: @smallexample
1.1.1.4 root 1689: inline double
1690: access_double (double *unaligned_ptr)
1691: @{
1692: union d2i @{ double d; int i[2]; @};
1693:
1694: union d2i *p = (union d2i *) unaligned_ptr;
1695: union d2i u;
1696:
1697: u.i[0] = p->i[0];
1698: u.i[1] = p->i[1];
1699:
1700: return u.d;
1701: @}
1.1.1.5 root 1702: @end smallexample
1.1.1.4 root 1703:
1704: @noindent
1705: Storing into the pointer can be done likewise with the same union.
1706:
1707: @item
1.1.1.5 root 1708: On Solaris, the @code{malloc} function in the @file{libmalloc.a} library
1709: may allocate memory that is only 4 byte aligned. Since GNU CC on the
1710: Sparc assumes that doubles are 8 byte aligned, this may result in a
1711: fatal signal if doubles are stored in memory allocated by the
1712: @file{libmalloc.a} library.
1713:
1714: The solution is to not use the @file{libmalloc.a} library. Use instead
1715: @code{malloc} and related functions from @file{libc.a}; they do not have
1716: this problem.
1717:
1718: @item
1.1.1.3 root 1719: Sun forgot to include a static version of @file{libdl.a} with some
1720: versions of SunOS (mainly 4.1). This results in undefined symbols when
1721: linking static binaries (that is, if you use @samp{-static}). If you
1722: see undefined symbols @code{_dlclose}, @code{_dlsym} or @code{_dlopen}
1723: when linking, compile and link against the file
1724: @file{mit/util/misc/dlsym.c} from the MIT version of X windows.
1.1 root 1725:
1726: @item
1.1.1.6 root 1727: The 128-bit long double format that the Sparc port supports currently
1728: works by using the architecturally defined quad-word floating point
1729: instructions. Since there is no hardware that supports these instructions
1730: they must be emulated by the operating system. Long doubles do not work
1731: in Sun OS versions 4.0.3 and earlier, because the kernel eumulator uses an
1732: obsolete and incompatible format. Long doubles do not work in Sun OS
1733: versions 4.1.1 to 4.1.3 because of emululator bugs that cause random
1734: unpredicatable failures. Long doubles appear to work in Sun OS 5.x
1735: (Solaris 2.x).
1736:
1737: @item
1738: On HP-UX version 9.01 on the HP PA, the HP compiler @code{cc} does not
1739: compile GNU CC correctly. We do not yet know why. However, GNU CC
1740: compiled on earlier HP-UX versions works properly on HP-UX 9.01 and can
1741: compile itself properly on 9.01.
1742:
1743: @item
1.1.1.3 root 1744: On the HP PA machine, ADB sometimes fails to work on functions compiled
1745: with GNU CC. Specifically, it fails to work on functions that use
1746: @code{alloca} or variable-size arrays. This is because GNU CC doesn't
1.1.1.4 root 1747: generate HP-UX unwind descriptors for such functions. It may even be
1.1.1.3 root 1748: impossible to generate them.
1.1.1.2 root 1749:
1.1.1.3 root 1750: @item
1.1.1.4 root 1751: Debugging (@samp{-g}) is not supported on the HP PA machine, unless you use
1752: the preliminary GNU tools (@pxref{Installation}).
1.1.1.2 root 1753:
1754: @item
1.1.1.4 root 1755: Taking the address of a label may generate errors from the HP-UX
1756: PA assembler. GAS for the PA does not have this problem.
1757:
1758: @item
1.1.1.6 root 1759: Using floating point parameters for indirect calls to static functions
1760: will not work when using the HP assembler. There simply is no way for GCC
1761: to specify what registers hold arguments for static functions when using
1762: the HP assembler. GAS for the PA does not have this problem.
1763:
1764: @item
1765: For some very large functions you may receive errors from the HP linker
1766: complaining about an out of bounds unconditional branch offset. Fixing
1767: this problem correctly requires fixing problems in GNU CC and GAS. We
1768: hope to fix this in time for GNU CC 2.6. Until then you can work around
1769: by making your function smaller, and if you are using GAS, splitting the
1770: function into multiple source files may be necessary.
1.1 root 1771:
1.1.1.3 root 1772: @item
1.1.1.5 root 1773: GNU CC compiled code sometimes emits warnings from the HP-UX assembler of
1774: the form:
1775:
1776: @smallexample
1777: (warning) Use of GR3 when
1778: frame >= 8192 may cause conflict.
1779: @end smallexample
1780:
1781: These warnings are harmless and can be safely ignored.
1782:
1783: @item
1.1.1.3 root 1784: The current version of the assembler (@file{/bin/as}) for the RS/6000
1785: has certain problems that prevent the @samp{-g} option in GCC from
1.1.1.5 root 1786: working. Note that @file{Makefile.in} uses @samp{-g} by default when
1787: compiling @file{libgcc2.c}.
1788:
1789: IBM has produced a fixed version of the assembler. The upgraded
1790: assembler unfortunately was not included in any of the AIX 3.2 update
1791: PTF releases (3.2.2, 3.2.3, or 3.2.3e). Users of AIX 3.1 should request
1792: PTF U403044 from IBM and users of AIX 3.2 should request PTF U416277.
1793: See the file @file{README.RS6000} for more details on these updates.
1794:
1795: You can test for the presense of a fixed assembler by using the
1796: command
1.1 root 1797:
1.1.1.5 root 1798: @smallexample
1799: as -u < /dev/null
1800: @end smallexample
1801:
1802: @noindent
1803: If the command exits normally, the assembler fix already is installed.
1804: If the assembler complains that "-u" is an unknown flag, you need to
1805: order the fix.
1.1 root 1806:
1807: @item
1.1.1.3 root 1808: On the IBM RS/6000, compiling code of the form
1.1 root 1809:
1.1.1.5 root 1810: @smallexample
1.1.1.3 root 1811: extern int foo;
1.1 root 1812:
1.1.1.3 root 1813: @dots{} foo @dots{}
1.1 root 1814:
1.1.1.3 root 1815: static int foo;
1.1.1.5 root 1816: @end smallexample
1.1 root 1817:
1.1.1.3 root 1818: @noindent
1819: will cause the linker to report an undefined symbol @code{foo}.
1820: Although this behavior differs from most other systems, it is not a
1821: bug because redefining an @code{extern} variable as @code{static}
1822: is undefined in ANSI C.
1.1 root 1823:
1824: @item
1.1.1.5 root 1825: AIX on the RS/6000 provides support (NLS) for environments outside of
1826: the United States. Compilers and assemblers use NLS to support
1827: locale-specific representations of various objects including
1828: floating-point numbers ("." vs "," for separating decimal fractions).
1829: There have been problems reported where the library linked with GCC does
1830: not produce the same floating-point formats that the assembler accepts.
1831: If you have this problem, set the LANG environment variable to "C" or
1832: "En_US".
1833:
1834: @item
1.1.1.7 ! root 1835: Even if you specify @samp{-fdollars-in-identifiers},
! 1836: you cannot successfully use @samp{$} in identifiers on the RS/6000 due
! 1837: to a restriction in the IBM assembler. GAS supports these
! 1838: identifiers.
! 1839:
! 1840: @item
! 1841: On the RS/6000, XLC version 1.3.0.0 will miscompile @file{jump.c}. XLC
! 1842: version 1.3.0.1 or later fixes this problem. You can obtain XLC-1.3.0.2
! 1843: by requesting PTF 421749 from IBM.
1.1.1.6 root 1844:
1845: @item
1.1.1.5 root 1846: There is an assembler bug in versions of DG/UX prior to 5.4.2.01 that
1847: occurs when the @samp{fldcr} instruction is used. GNU CC uses
1848: @samp{fldcr} on the 88100 to serialize volatile memory references. Use
1.1.1.6 root 1849: the option @samp{-mno-serialize-volatile} if your version of the
1.1.1.5 root 1850: assembler has this bug.
1851:
1852: @item
1.1.1.3 root 1853: On VMS, GAS versions 1.38.1 and earlier may cause spurious warning
1854: messages from the linker. These warning messages complain of mismatched
1855: psect attributes. You can ignore them. @xref{VMS Install}.
1.1 root 1856:
1.1.1.4 root 1857: @item
1.1.1.5 root 1858: On NewsOS version 3, if you include both of the files @file{stddef.h}
1859: and @file{sys/types.h}, you get an error because there are two typedefs
1860: of @code{size_t}. You should change @file{sys/types.h} by adding these
1.1.1.4 root 1861: lines around the definition of @code{size_t}:
1862:
1.1.1.5 root 1863: @smallexample
1.1.1.4 root 1864: #ifndef _SIZE_T
1865: #define _SIZE_T
1866: @var{actual typedef here}
1867: #endif
1.1.1.5 root 1868: @end smallexample
1.1.1.4 root 1869:
1.1.1.3 root 1870: @cindex Alliant
1.1 root 1871: @item
1.1.1.3 root 1872: On the Alliant, the system's own convention for returning structures
1873: and unions is unusual, and is not compatible with GNU CC no matter
1874: what options are used.
1.1 root 1875:
1.1.1.3 root 1876: @cindex RT PC
1877: @cindex IBM RT PC
1.1 root 1878: @item
1.1.1.5 root 1879: On the IBM RT PC, the MetaWare HighC compiler (hc) uses a different
1880: convention for structure and union returning. Use the option
1.1.1.3 root 1881: @samp{-mhc-struct-return} to tell GNU CC to use a convention compatible
1882: with it.
1.1.1.2 root 1883:
1.1.1.3 root 1884: @cindex Vax calling convention
1885: @cindex Ultrix calling convention
1.1.1.2 root 1886: @item
1.1.1.3 root 1887: On Ultrix, the Fortran compiler expects registers 2 through 5 to be saved
1888: by function calls. However, the C compiler uses conventions compatible
1889: with BSD Unix: registers 2 through 5 may be clobbered by function calls.
1.1 root 1890:
1.1.1.3 root 1891: GNU CC uses the same convention as the Ultrix C compiler. You can use
1892: these options to produce code compatible with the Fortran compiler:
1.1 root 1893:
1.1.1.3 root 1894: @smallexample
1895: -fcall-saved-r2 -fcall-saved-r3 -fcall-saved-r4 -fcall-saved-r5
1896: @end smallexample
1.1.1.4 root 1897:
1898: @item
1899: On the WE32k, you may find that programs compiled with GNU CC do not
1900: work with the standard shared C ilbrary. You may need to link with
1901: the ordinary C compiler. If you do so, you must specify the following
1902: options:
1903:
1904: @smallexample
1.1.1.7 ! root 1905: -L/usr/local/lib/gcc-lib/we32k-att-sysv/2.6.0 -lgcc -lc_s
1.1.1.4 root 1906: @end smallexample
1907:
1908: The first specifies where to find the library @file{libgcc.a}
1909: specified with the @samp{-lgcc} option.
1910:
1911: GNU CC does linking by invoking @code{ld}, just as @code{cc} does, and
1912: there is no reason why it @emph{should} matter which compilation program
1913: you use to invoke @code{ld}. If someone tracks this problem down,
1914: it can probably be fixed easily.
1.1.1.5 root 1915:
1916: @item
1917: On the Alpha, you may get assembler errors about invalid syntax as a
1918: result of floating point constants. This is due to a bug in the C
1919: library functions @code{ecvt}, @code{fcvt} and @code{gcvt}. Given valid
1920: floating point numbers, they sometimes print @samp{NaN}.
1921:
1922: @item
1923: On Irix 4.0.5F (and perhaps in some other versions), an assembler bug
1924: sometimes reorders instructions incorrectly when optimization is turned
1925: on. If you think this may be happening to you, try using the GNU
1926: assembler; GAS version 2.1 supports ECOFF on Irix.
1927:
1928: Or use the @samp{-noasmopt} option when you compile GNU CC with itself,
1929: and then again when you compile your program. (This is a temporary
1930: kludge to turn off assembler optimization on Irix.) If this proves to
1931: be what you need, edit the assembler spec in the file @file{specs} so
1932: that it unconditionally passes @samp{-O0} to the assembler, and never
1933: passes @samp{-O2} or @samp{-O3}.
1934: @end itemize
1935:
1936: @node External Bugs
1937: @section Problems Compiling Certain Programs
1938:
1.1.1.7 ! root 1939: @c prevent bad page break with this line
! 1940: Certain programs have problems compiling.
! 1941:
1.1.1.5 root 1942: @itemize @bullet
1943: @item
1944: Parse errors may occur compiling X11 on a Decstation running Ultrix 4.2
1945: because of problems in DEC's versions of the X11 header files
1946: @file{X11/Xlib.h} and @file{X11/Xutil.h}. People recommend adding
1947: @samp{-I/usr/include/mit} to use the MIT versions of the header files,
1948: using the @samp{-traditional} switch to turn off ANSI C, or fixing the
1949: header files by adding this:
1950:
1951: @example
1952: #ifdef __STDC__
1953: #define NeedFunctionPrototypes 0
1954: #endif
1955: @end example
1956:
1957: @item
1.1.1.6 root 1958: If you have trouble compiling Perl on a SunOS 4 system, it may be
1959: because Perl specifies @samp{-I/usr/ucbinclude}. This accesses the
1960: unfixed header files. Perl specifies the options
1961:
1962: @example
1963: -traditional -Dvolatile=__volatile__
1964: -I/usr/include/sun -I/usr/ucbinclude
1965: -fpcc-struct-return
1966: @end example
1967:
1968: @noindent
1.1.1.7 ! root 1969: most of which are unnecessary with GCC 2.4.5 and newer versions. You
! 1970: can make a properly working Perl by setting @code{ccflags} to
! 1971: @samp{-fwritable-strings} (implied by the @samp{-traditional} in the
! 1972: original options) and @code{cppflags} to empty in @file{config.sh}, then
! 1973: typing @samp{./doSH; make depend; make}.
1.1.1.6 root 1974:
1975: @item
1.1.1.5 root 1976: On various 386 Unix systems derived from System V, including SCO, ISC,
1977: and ESIX, you may get error messages about running out of virtual memory
1978: while compiling certain programs.
1979:
1980: You can prevent this problem by linking GNU CC with the GNU malloc
1981: (which thus replaces the malloc that comes with the system). GNU malloc
1982: is available as a separate package, and also in the file
1983: @file{src/gmalloc.c} in the GNU Emacs 19 distribution.
1984:
1985: If you have installed GNU malloc as a separate library package, use this
1986: option when you relink GNU CC:
1987:
1988: @example
1989: MALLOC=/usr/local/lib/libgmalloc.a
1990: @end example
1991:
1992: Alternatively, if you have compiled @file{gmalloc.c} from Emacs 19, copy
1993: the object file to @file{gmalloc.o} and use this option when you relink
1994: GNU CC:
1995:
1996: @example
1997: MALLOC=gmalloc.o
1998: @end example
1.1 root 1999: @end itemize
2000:
1.1.1.3 root 2001: @node Incompatibilities
2002: @section Incompatibilities of GNU CC
1.1 root 2003: @cindex incompatibilities of GNU CC
2004:
2005: There are several noteworthy incompatibilities between GNU C and most
2006: existing (non-ANSI) versions of C. The @samp{-traditional} option
1.1.1.3 root 2007: eliminates many of these incompatibilities, @emph{but not all}, by
1.1 root 2008: telling GNU C to behave like the other C compilers.
2009:
2010: @itemize @bullet
2011: @cindex string constants
2012: @cindex read-only strings
2013: @cindex shared strings
2014: @item
2015: GNU CC normally makes string constants read-only. If several
2016: identical-looking string constants are used, GNU CC stores only one
2017: copy of the string.
2018:
2019: @cindex @code{mktemp}, and constant strings
2020: One consequence is that you cannot call @code{mktemp} with a string
2021: constant argument. The function @code{mktemp} always alters the
2022: string its argument points to.
2023:
2024: @cindex @code{sscanf}, and constant strings
2025: @cindex @code{fscanf}, and constant strings
2026: @cindex @code{scanf}, and constant strings
2027: Another consequence is that @code{sscanf} does not work on some systems
2028: when passed a string constant as its format control string or input.
2029: This is because @code{sscanf} incorrectly tries to write into the string
2030: constant. Likewise @code{fscanf} and @code{scanf}.
2031:
2032: The best solution to these problems is to change the program to use
2033: @code{char}-array variables with initialization strings for these
2034: purposes instead of string constants. But if this is not possible,
2035: you can use the @samp{-fwritable-strings} flag, which directs GNU CC
2036: to handle string constants the same way most C compilers do.
2037: @samp{-traditional} also has this effect, among others.
2038:
2039: @item
1.1.1.3 root 2040: @code{-2147483648} is positive.
2041:
2042: This is because 2147483648 cannot fit in the type @code{int}, so
2043: (following the ANSI C rules) its data type is @code{unsigned long int}.
2044: Negating this value yields 2147483648 again.
2045:
2046: @item
1.1 root 2047: GNU CC does not substitute macro arguments when they appear inside of
2048: string constants. For example, the following macro in GNU CC
2049:
2050: @example
2051: #define foo(a) "a"
2052: @end example
2053:
2054: @noindent
2055: will produce output @code{"a"} regardless of what the argument @var{a} is.
2056:
2057: The @samp{-traditional} option directs GNU CC to handle such cases
2058: (among others) in the old-fashioned (non-ANSI) fashion.
2059:
2060: @cindex @code{setjmp} incompatibilities
2061: @cindex @code{longjmp} incompatibilities
2062: @item
2063: When you use @code{setjmp} and @code{longjmp}, the only automatic
2064: variables guaranteed to remain valid are those declared
2065: @code{volatile}. This is a consequence of automatic register
2066: allocation. Consider this function:
2067:
2068: @example
2069: jmp_buf j;
2070:
2071: foo ()
2072: @{
2073: int a, b;
2074:
2075: a = fun1 ();
2076: if (setjmp (j))
2077: return a;
2078:
2079: a = fun2 ();
2080: /* @r{@code{longjmp (j)} may occur in @code{fun3}.} */
2081: return a + fun3 ();
2082: @}
2083: @end example
2084:
2085: Here @code{a} may or may not be restored to its first value when the
2086: @code{longjmp} occurs. If @code{a} is allocated in a register, then
2087: its first value is restored; otherwise, it keeps the last value stored
2088: in it.
2089:
2090: If you use the @samp{-W} option with the @samp{-O} option, you will
2091: get a warning when GNU CC thinks such a problem might be possible.
2092:
2093: The @samp{-traditional} option directs GNU C to put variables in
2094: the stack by default, rather than in registers, in functions that
2095: call @code{setjmp}. This results in the behavior found in
2096: traditional C compilers.
2097:
1.1.1.3 root 2098: @item
2099: Programs that use preprocessor directives in the middle of macro
2100: arguments do not work with GNU CC. For example, a program like this
2101: will not work:
2102:
2103: @example
2104: foobar (
2105: #define luser
2106: hack)
2107: @end example
2108:
2109: ANSI C does not permit such a construct. It would make sense to support
2110: it when @samp{-traditional} is used, but it is too much work to
2111: implement.
2112:
1.1 root 2113: @cindex external declaration scope
2114: @cindex scope of external declarations
2115: @cindex declaration scope
2116: @item
2117: Declarations of external variables and functions within a block apply
2118: only to the block containing the declaration. In other words, they
2119: have the same scope as any other declaration in the same place.
2120:
2121: In some other C compilers, a @code{extern} declaration affects all the
2122: rest of the file even if it happens within a block.
2123:
2124: The @samp{-traditional} option directs GNU C to treat all @code{extern}
2125: declarations as global, like traditional compilers.
2126:
2127: @item
2128: In traditional C, you can combine @code{long}, etc., with a typedef name,
2129: as shown here:
2130:
2131: @example
2132: typedef int foo;
2133: typedef long foo bar;
2134: @end example
2135:
2136: In ANSI C, this is not allowed: @code{long} and other type modifiers
2137: require an explicit @code{int}. Because this criterion is expressed
2138: by Bison grammar rules rather than C code, the @samp{-traditional}
2139: flag cannot alter it.
2140:
2141: @cindex typedef names as function parameters
2142: @item
2143: PCC allows typedef names to be used as function parameters. The
2144: difficulty described immediately above applies here too.
2145:
2146: @cindex whitespace
2147: @item
2148: PCC allows whitespace in the middle of compound assignment operators
2149: such as @samp{+=}. GNU CC, following the ANSI standard, does not
2150: allow this. The difficulty described immediately above applies here
2151: too.
2152:
2153: @cindex apostrophes
2154: @cindex '
2155: @item
1.1.1.4 root 2156: GNU CC complains about unterminated character constants inside of
2157: preprocessor conditionals that fail. Some programs have English
2158: comments enclosed in conditionals that are guaranteed to fail; if these
2159: comments contain apostrophes, GNU CC will probably report an error. For
2160: example, this code would produce an error:
1.1 root 2161:
2162: @example
2163: #if 0
2164: You can't expect this to work.
2165: #endif
2166: @end example
2167:
2168: The best solution to such a problem is to put the text into an actual
2169: C comment delimited by @samp{/*@dots{}*/}. However,
2170: @samp{-traditional} suppresses these error messages.
2171:
1.1.1.4 root 2172: @item
2173: Many user programs contain the declaration @samp{long time ();}. In the
2174: past, the system header files on many systems did not actually declare
2175: @code{time}, so it did not matter what type your program declared it to
2176: return. But in systems with ANSI C headers, @code{time} is declared to
2177: return @code{time_t}, and if that is not the same as @code{long}, then
2178: @samp{long time ();} is erroneous.
2179:
2180: The solution is to change your program to use @code{time_t} as the return
2181: type of @code{time}.
2182:
1.1 root 2183: @cindex @code{float} as function value type
2184: @item
2185: When compiling functions that return @code{float}, PCC converts it to
2186: a double. GNU CC actually returns a @code{float}. If you are concerned
2187: with PCC compatibility, you should declare your functions to return
2188: @code{double}; you might as well say what you mean.
2189:
2190: @cindex structures
2191: @cindex unions
2192: @item
2193: When compiling functions that return structures or unions, GNU CC
2194: output code normally uses a method different from that used on most
2195: versions of Unix. As a result, code compiled with GNU CC cannot call
2196: a structure-returning function compiled with PCC, and vice versa.
2197:
2198: The method used by GNU CC is as follows: a structure or union which is
2199: 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or union
2200: with any other size is stored into an address supplied by the caller
2201: (usually in a special, fixed register, but on some machines it is passed
2202: on the stack). The machine-description macros @code{STRUCT_VALUE} and
2203: @code{STRUCT_INCOMING_VALUE} tell GNU CC where to pass this address.
2204:
1.1.1.3 root 2205: By contrast, PCC on most target machines returns structures and unions
2206: of any size by copying the data into an area of static storage, and then
2207: returning the address of that storage as if it were a pointer value.
2208: The caller must copy the data from that memory area to the place where
2209: the value is wanted. GNU CC does not use this method because it is
2210: slower and nonreentrant.
2211:
2212: On some newer machines, PCC uses a reentrant convention for all
2213: structure and union returning. GNU CC on most of these machines uses a
2214: compatible convention when returning structures and unions in memory,
2215: but still returns small structures and unions in registers.
2216:
2217: You can tell GNU CC to use a compatible convention for all structure and
2218: union returning with the option @samp{-fpcc-struct-return}.
1.1.1.6 root 2219:
2220: @cindex preprocessing tokens
2221: @cindex preprocessing numbers
2222: @item
2223: GNU C complains about program fragments such as @samp{0x74ae-0x4000}
2224: which appear to be two hexadecimal constants separated by the minus
2225: operator. Actually, this string is a single @dfn{preprocessing token}.
2226: Each such token must correspond to one token in C. Since this does not,
2227: GNU C prints an error message. Although it may appear obvious that what
2228: is meant is an operator and two values, the ANSI C standard specifically
2229: requires that this be treated as erroneous.
2230:
2231: A @dfn{preprocessing token} is a @dfn{preprocessing number} if it
2232: begins with a digit and is followed by letters, underscores, digits,
2233: periods and @samp{e+}, @samp{e-}, @samp{E+}, or @samp{E-} character
2234: sequences.
2235:
2236: To make the above program fragment valid, place whitespace in front of
2237: the minus sign. This whitespace will end the preprocessing number.
2238: @end itemize
2239:
2240: @node Fixed Headers
2241: @section Fixed Header Files
2242:
2243: GNU CC needs to install corrected versions of some system header files.
2244: This is because most target systems have some header files that won't
2245: work with GNU CC unless they are changed. Some have bugs, some are
2246: incompatible with ANSI C, and some depend on special features of other
2247: compilers.
2248:
2249: Installing GNU CC automatically creates and installs the fixed header
2250: files, by running a program called @code{fixincludes} (or for certain
2251: targets an alternative such as @code{fixinc.svr4}). Normally, you
2252: don't need to pay attention to this. But there are cases where it
2253: doesn't do the right thing automatically.
2254:
2255: @itemize @bullet
2256: @item
2257: If you update the system's header files, such as by installing a new
2258: system version, the fixed header files of GNU CC are not automatically
2259: updated. The easiest way to update them is to reinstall GNU CC. (If
2260: you want to be clever, look in the makefile and you can find a
2261: shortcut.)
2262:
2263: @item
2264: On some systems, in particular SunOS 4, header file directories contain
2265: machine-specific symbolic links in certain places. This makes it
2266: possible to share most of the header files among hosts running the
2267: same version of SunOS 4 on different machine models.
2268:
2269: The programs that fix the header files do not understand this special
2270: way of using symbolic links; therefore, the directory of fixed header
2271: files is good only for the machine model used to build it.
2272:
2273: In SunOS 4, only programs that look inside the kernel will notice the
2274: difference between machine models. Therefore, for most purposes, you
2275: need not be concerned about this.
2276:
2277: It is possible to make separate sets of fixed header files for the
2278: different machine models, and arrange a structure of symbolic links so
2279: as to use the proper set, but you'll have to do this by hand.
2280:
2281: @item
2282: On Lynxos, GNU CC by default does not fix the header files. This is
2283: because bugs in the shell cause the @code{fixincludes} script to fail.
2284:
2285: This means you will encounter problems due to bugs in the system header
2286: files. It may be no comfort that they aren't GNU CC's fault, but it
2287: does mean that there's nothing for us to do about them.
1.1.1.3 root 2288: @end itemize
2289:
2290: @node Disappointments
2291: @section Disappointments and Misunderstandings
2292:
2293: These problems are perhaps regrettable, but we don't know any practical
2294: way around them.
2295:
2296: @itemize @bullet
2297: @item
2298: Certain local variables aren't recognized by debuggers when you compile
2299: with optimization.
2300:
2301: This occurs because sometimes GNU CC optimizes the variable out of
2302: existence. There is no way to tell the debugger how to compute the
2303: value such a variable ``would have had'', and it is not clear that would
2304: be desirable anyway. So GNU CC simply does not mention the eliminated
2305: variable when it writes debugging information.
2306:
2307: You have to expect a certain amount of disagreement between the
2308: executable and your source code, when you use optimization.
2309:
2310: @cindex conflicting types
2311: @cindex scope of declaration
2312: @item
2313: Users often think it is a bug when GNU CC reports an error for code
2314: like this:
2315:
2316: @example
2317: int foo (struct mumble *);
2318:
2319: struct mumble @{ @dots{} @};
2320:
2321: int foo (struct mumble *x)
2322: @{ @dots{} @}
2323: @end example
2324:
2325: This code really is erroneous, because the scope of @code{struct
1.1.1.4 root 2326: mumble} in the prototype is limited to the argument list containing it.
1.1.1.3 root 2327: It does not refer to the @code{struct mumble} defined with file scope
2328: immediately below---they are two unrelated types with similar names in
2329: different scopes.
2330:
2331: But in the definition of @code{foo}, the file-scope type is used
2332: because that is available to be inherited. Thus, the definition and
2333: the prototype do not match, and you get an error.
2334:
2335: This behavior may seem silly, but it's what the ANSI standard specifies.
2336: It is easy enough for you to make your code work by moving the
2337: definition of @code{struct mumble} above the prototype. It's not worth
2338: being incompatible with ANSI C just to avoid an error for the example
2339: shown above.
1.1.1.4 root 2340:
2341: @item
2342: Accesses to bitfields even in volatile objects works by accessing larger
2343: objects, such as a byte or a word. You cannot rely on what size of
2344: object is accessed in order to read or write the bitfield; it may even
2345: vary for a given bitfield according to the precise usage.
2346:
2347: If you care about controlling the amount of memory that is accessed, use
2348: volatile but do not use bitfields.
2349:
2350: @item
1.1.1.5 root 2351: GNU CC comes with shell scripts to fix certain known problems in system
2352: header files. They install corrected copies of various header files in
2353: a special directory where only GNU CC will normally look for them. The
2354: scripts adapt to various systems by searching all the system header
2355: files for the problem cases that we know about.
2356:
2357: If new system header files are installed, nothing automatically arranges
2358: to update the corrected header files. You will have to reinstall GNU CC
2359: to fix the new header files. More specifically, go to the build
2360: directory and delete the files @file{stmp-fixinc} and
2361: @file{stmp-headers}, and the subdirectory @code{include}; then do
2362: @samp{make install} again.
2363:
2364: @item
1.1.1.4 root 2365: On 68000 systems, you can get paradoxical results if you test the
2366: precise values of floating point numbers. For example, you can find
2367: that a floating point value which is not a NaN is not equal to itself.
2368: This results from the fact that the the floating point registers hold a
2369: few more bits of precision than fit in a @code{double} in memory.
2370: Compiled code moves values between memory and floating point registers
2371: at its convenience, and moving them into memory truncates them.
2372:
1.1.1.5 root 2373: You can partially avoid this problem by using the @samp{-ffloat-store}
2374: option (@pxref{Optimize Options}).
1.1.1.4 root 2375:
2376: @item
2377: On the MIPS, variable argument functions using @file{varargs.h}
2378: cannot have a floating point value for the first argument. The
2379: reason for this is that in the absence of a prototype in scope,
2380: if the first argument is a floating point, it is passed in a
1.1.1.5 root 2381: floating point register, rather than an integer register.
1.1.1.4 root 2382:
2383: If the code is rewritten to use the ANSI standard @file{stdarg.h}
2384: method of variable arguments, and the prototype is in scope at
2385: the time of the call, everything will work fine.
1.1.1.3 root 2386: @end itemize
2387:
1.1.1.5 root 2388: @node C++ Misunderstandings
2389: @section Common Misunderstandings with GNU C++
2390:
2391: @cindex misunderstandings in C++
2392: @cindex surprises in C++
2393: @cindex C++ misunderstandings
2394: C++ is a complex language and an evolving one, and its standard definition
2395: (the ANSI C++ draft standard) is also evolving. As a result,
2396: your C++ compiler may occasionally surprise you, even when its behavior is
2397: correct. This section discusses some areas that frequently give rise to
2398: questions of this sort.
2399:
2400: @menu
2401: * Static Definitions:: Static member declarations are not definitions
2402: * Temporaries:: Temporaries may vanish before you expect
2403: @end menu
2404:
2405: @node Static Definitions
2406: @subsection Declare @emph{and} Define Static Members
2407:
2408: @cindex C++ static data, declaring and defining
2409: @cindex static data in C++, declaring and defining
2410: @cindex declaring static data in C++
2411: @cindex defining static data in C++
2412: When a class has static data members, it is not enough to @emph{declare}
2413: the static member; you must also @emph{define} it. For example:
2414:
2415: @example
2416: class Foo
2417: @{
2418: @dots{}
2419: void method();
2420: static int bar;
2421: @};
2422: @end example
2423:
2424: This declaration only establishes that the class @code{Foo} has an
2425: @code{int} named @code{Foo::bar}, and a member function named
2426: @code{Foo::method}. But you still need to define @emph{both}
2427: @code{method} and @code{bar} elsewhere. According to the draft ANSI
2428: standard, you must supply an initializer in one (and only one) source
2429: file, such as:
2430:
2431: @example
2432: int Foo::bar = 0;
2433: @end example
2434:
2435: Other C++ compilers may not correctly implement the standard behavior.
2436: As a result, when you switch to @code{g++} from one of these compilers,
2437: you may discover that a program that appeared to work correctly in fact
2438: does not conform to the standard: @code{g++} reports as undefined
2439: symbols any static data members that lack definitions.
2440:
2441: @node Temporaries
2442: @subsection Temporaries May Vanish Before You Expect
2443:
2444: @cindex temporaries, lifetime of
2445: @cindex portions of temporary objects, pointers to
2446: It is dangerous to use pointers or references to @emph{portions} of a
2447: temporary object. The compiler may very well delete the object before
2448: you expect it to, leaving a pointer to garbage. The most common place
2449: where this problem crops up is in classes like the libg++
2450: @code{String} class, that define a conversion function to type
2451: @code{char *} or @code{const char *}. However, any class that returns
2452: a pointer to some internal structure is potentially subject to this
2453: problem.
2454:
2455: For example, a program may use a function @code{strfunc} that returns
2456: @code{String} objects, and another function @code{charfunc} that
2457: operates on pointers to @code{char}:
2458:
2459: @example
2460: String strfunc ();
2461: void charfunc (const char *);
2462: @end example
2463:
2464: @noindent
2465: In this situation, it may seem natural to write @w{@samp{charfunc
2466: (strfunc ());}} based on the knowledge that class @code{String} has an
2467: explicit conversion to @code{char} pointers. However, what really
2468: happens is akin to @samp{charfunc (@w{strfunc ()}.@w{convert ()});},
2469: where the @code{convert} method is a function to do the same data
2470: conversion normally performed by a cast. Since the last use of the
2471: temporary @code{String} object is the call to the conversion function,
2472: the compiler may delete that object before actually calling
2473: @code{charfunc}. The compiler has no way of knowing that deleting the
2474: @code{String} object will invalidate the pointer. The pointer then
2475: points to garbage, so that by the time @code{charfunc} is called, it
2476: gets an invalid argument.
2477:
2478: Code like this may run successfully under some other compilers,
2479: especially those that delete temporaries relatively late. However, the
2480: GNU C++ behavior is also standard-conformant, so if your program depends
2481: on late destruction of temporaries it is not portable.
2482:
2483: If you think this is surprising, you should be aware that the ANSI C++
2484: committee continues to debate the lifetime-of-temporaries problem.
2485:
2486: For now, at least, the safe way to write such code is to give the
2487: temporary a name, which forces it to remain until the end of the scope of
2488: the name. For example:
2489:
2490: @example
2491: String& tmp = strfunc ();
2492: charfunc (tmp);
2493: @end example
2494:
1.1.1.4 root 2495: @node Protoize Caveats
2496: @section Caveats of using @code{protoize}
2497:
2498: The conversion programs @code{protoize} and @code{unprotoize} can
2499: sometimes change a source file in a way that won't work unless you
2500: rearrange it.
2501:
2502: @itemize @bullet
2503: @item
2504: @code{protoize} can insert references to a type name or type tag before
2505: the definition, or in a file where they are not defined.
2506:
2507: If this happens, compiler error messages should show you where the new
2508: references are, so fixing the file by hand is straightforward.
2509:
2510: @item
2511: There are some C constructs which @code{protoize} cannot figure out.
2512: For example, it can't determine argument types for declaring a
2513: pointer-to-function variable; this you must do by hand. @code{protoize}
2514: inserts a comment containing @samp{???} each time it finds such a
2515: variable; so you can find all such variables by searching for this
2516: string. ANSI C does not require declaring the argument types of
2517: pointer-to-function types.
2518:
2519: @item
2520: Using @code{unprotoize} can easily introduce bugs. If the program
2521: relied on prototypes to bring about conversion of arguments, these
2522: conversions will not take place in the program without prototypes.
2523: One case in which you can be sure @code{unprotoize} is safe is when
2524: you are removing prototypes that were made with @code{protoize}; if
2525: the program worked before without any prototypes, it will work again
2526: without them.
2527:
2528: You can find all the places where this problem might occur by compiling
2529: the program with the @samp{-Wconversion} option. It prints a warning
2530: whenever an argument is converted.
2531:
2532: @item
2533: Both conversion programs can be confused if there are macro calls in and
2534: around the text to be converted. In other words, the standard syntax
2535: for a declaration or definition must not result from expanding a macro.
2536: This problem is inherent in the design of C and cannot be fixed. If
2537: only a few functions have confusing macro calls, you can easily convert
2538: them manually.
2539:
2540: @item
2541: @code{protoize} cannot get the argument types for a function whose
2542: definition was not actually compiled due to preprocessor conditionals.
2543: When this happens, @code{protoize} changes nothing in regard to such
2544: a function. @code{protoize} tries to detect such instances and warn
2545: about them.
2546:
2547: You can generally work around this problem by using @code{protoize} step
2548: by step, each time specifying a different set of @samp{-D} options for
2549: compilation, until all of the functions have been converted. There is
2550: no automatic way to verify that you have got them all, however.
2551:
2552: @item
2553: Confusion may result if there is an occasion to convert a function
2554: declaration or definition in a region of source code where there is more
2555: than one formal parameter list present. Thus, attempts to convert code
2556: containing multiple (conditionally compiled) versions of a single
2557: function header (in the same vicinity) may not produce the desired (or
2558: expected) results.
2559:
2560: If you plan on converting source files which contain such code, it is
2561: recommended that you first make sure that each conditionally compiled
2562: region of source code which contains an alternative function header also
2563: contains at least one additional follower token (past the final right
2564: parenthesis of the function header). This should circumvent the
2565: problem.
2566:
2567: @item
2568: @code{unprotoize} can become confused when trying to convert a function
2569: definition or declaration which contains a declaration for a
2570: pointer-to-function formal argument which has the same name as the
2571: function being defined or declared. We recommand you avoid such choices
2572: of formal parameter names.
2573:
2574: @item
2575: You might also want to correct some of the indentation by hand and break
2576: long lines. (The conversion programs don't write lines longer than
2577: eighty characters in any case.)
2578: @end itemize
2579:
2580: @node Non-bugs
1.1.1.3 root 2581: @section Certain Changes We Don't Want to Make
2582:
2583: This section lists changes that people frequently request, but which
2584: we do not make because we think GNU CC is better without them.
2585:
2586: @itemize @bullet
2587: @item
2588: Checking the number and type of arguments to a function which has an
2589: old-fashioned definition and no prototype.
2590:
2591: Such a feature would work only occasionally---only for calls that appear
2592: in the same file as the called function, following the definition. The
2593: only way to check all calls reliably is to add a prototype for the
2594: function. But adding a prototype eliminates the motivation for this
2595: feature. So the feature is not worthwhile.
2596:
2597: @item
2598: Warning about using an expression whose type is signed as a shift count.
2599:
2600: Shift count operands are probably signed more often than unsigned.
2601: Warning about this would cause far more annoyance than good.
2602:
2603: @item
2604: Warning about assigning a signed value to an unsigned variable.
2605:
2606: Such assignments must be very common; warning about them would cause
2607: more annoyance than good.
2608:
1.1.1.4 root 2609: @item
2610: Warning about unreachable code.
2611:
2612: It's very common to have unreachable code in machine-generated
2613: programs. For example, this happens normally in some files of GNU C
2614: itself.
2615:
1.1.1.3 root 2616: @item
2617: Warning when a non-void function value is ignored.
2618:
2619: Coming as I do from a Lisp background, I balk at the idea that there is
2620: something dangerous about discarding a value. There are functions that
2621: return values which some callers may find useful; it makes no sense to
2622: clutter the program with a cast to @code{void} whenever the value isn't
2623: useful.
2624:
2625: @item
2626: Assuming (for optimization) that the address of an external symbol is
2627: never zero.
2628:
2629: This assumption is false on certain systems when @samp{#pragma weak} is
2630: used.
2631:
2632: @item
2633: Making @samp{-fshort-enums} the default.
2634:
2635: This would cause storage layout to be incompatible with most other C
2636: compilers. And it doesn't seem very important, given that you can get
2637: the same result in other ways. The case where it matters most is when
2638: the enumeration-valued object is inside a structure, and in that case
2639: you can specify a field width explicitly.
2640:
2641: @item
2642: Making bitfields unsigned by default on particular machines where ``the
2643: ABI standard'' says to do so.
2644:
2645: The ANSI C standard leaves it up to the implementation whether a bitfield
2646: declared plain @code{int} is signed or not. This in effect creates two
2647: alternative dialects of C.
2648:
1.1.1.5 root 2649: The GNU C compiler supports both dialects; you can specify the signed
2650: dialect with @samp{-fsigned-bitfields} and the unsigned dialect with
2651: @samp{-funsigned-bitfields}. However, this leaves open the question of
2652: which dialect to use by default.
1.1.1.3 root 2653:
2654: Currently, the preferred dialect makes plain bitfields signed, because
2655: this is simplest. Since @code{int} is the same as @code{signed int} in
2656: every other context, it is cleanest for them to be the same in bitfields
2657: as well.
2658:
2659: Some computer manufacturers have published Application Binary Interface
2660: standards which specify that plain bitfields should be unsigned. It is
2661: a mistake, however, to say anything about this issue in an ABI. This is
2662: because the handling of plain bitfields distinguishes two dialects of C.
2663: Both dialects are meaningful on every type of machine. Whether a
2664: particular object file was compiled using signed bitfields or unsigned
2665: is of no concern to other object files, even if they access the same
2666: bitfields in the same data structures.
2667:
2668: A given program is written in one or the other of these two dialects.
2669: The program stands a chance to work on most any machine if it is
2670: compiled with the proper dialect. It is unlikely to work at all if
2671: compiled with the wrong dialect.
2672:
2673: Many users appreciate the GNU C compiler because it provides an
2674: environment that is uniform across machines. These users would be
2675: inconvenienced if the compiler treated plain bitfields differently on
2676: certain machines.
2677:
2678: Occasionally users write programs intended only for a particular machine
2679: type. On these occasions, the users would benefit if the GNU C compiler
2680: were to support by default the same dialect as the other compilers on
2681: that machine. But such applications are rare. And users writing a
2682: program to run on more than one type of machine cannot possibly benefit
2683: from this kind of compatibility.
2684:
2685: This is why GNU CC does and will treat plain bitfields in the same
2686: fashion on all types of machines (by default).
2687:
2688: There are some arguments for making bitfields unsigned by default on all
2689: machines. If, for example, this becomes a universal de facto standard,
2690: it would make sense for GNU CC to go along with it. This is something
2691: to be considered in the future.
1.1 root 2692:
1.1.1.3 root 2693: (Of course, users strongly concerned about portability should indicate
2694: explicitly in each bitfield whether it is signed or not. In this way,
2695: they write programs which have the same meaning in both C dialects.)
1.1 root 2696:
1.1.1.3 root 2697: @item
2698: Undefining @code{__STDC__} when @samp{-ansi} is not used.
1.1 root 2699:
1.1.1.3 root 2700: Currently, GNU CC defines @code{__STDC__} as long as you don't use
2701: @samp{-traditional}. This provides good results in practice.
1.1 root 2702:
1.1.1.3 root 2703: Programmers normally use conditionals on @code{__STDC__} to ask whether
2704: it is safe to use certain features of ANSI C, such as function
2705: prototypes or ANSI token concatenation. Since plain @samp{gcc} supports
2706: all the features of ANSI C, the correct answer to these questions is
2707: ``yes''.
1.1 root 2708:
1.1.1.3 root 2709: Some users try to use @code{__STDC__} to check for the availability of
2710: certain library facilities. This is actually incorrect usage in an ANSI
2711: C program, because the ANSI C standard says that a conforming
2712: freestanding implementation should define @code{__STDC__} even though it
2713: does not have the library facilities. @samp{gcc -ansi -pedantic} is a
2714: conforming freestanding implementation, and it is therefore required to
2715: define @code{__STDC__}, even though it does not come with an ANSI C
2716: library.
1.1 root 2717:
1.1.1.3 root 2718: Sometimes people say that defining @code{__STDC__} in a compiler that
2719: does not completely conform to the ANSI C standard somehow violates the
2720: standard. This is illogical. The standard is a standard for compilers
2721: that claim to support ANSI C, such as @samp{gcc -ansi}---not for other
2722: compilers such as plain @samp{gcc}. Whatever the ANSI C standard says
2723: is relevant to the design of plain @samp{gcc} without @samp{-ansi} only
2724: for pragmatic reasons, not as a requirement.
1.1 root 2725:
2726: @item
1.1.1.3 root 2727: Undefining @code{__STDC__} in C++.
1.1 root 2728:
1.1.1.3 root 2729: Programs written to compile with C++-to-C translators get the
2730: value of @code{__STDC__} that goes with the C compiler that is
2731: subsequently used. These programs must test @code{__STDC__}
2732: to determine what kind of C preprocessor that compiler uses:
2733: whether they should concatenate tokens in the ANSI C fashion
2734: or in the traditional fashion.
1.1 root 2735:
1.1.1.3 root 2736: These programs work properly with GNU C++ if @code{__STDC__} is defined.
2737: They would not work otherwise.
2738:
2739: In addition, many header files are written to provide prototypes in ANSI
2740: C but not in traditional C. Many of these header files can work without
2741: change in C++ provided @code{__STDC__} is defined. If @code{__STDC__}
2742: is not defined, they will all fail, and will all need to be changed to
2743: test explicitly for C++ as well.
1.1 root 2744:
2745: @item
1.1.1.3 root 2746: Deleting ``empty'' loops.
1.1 root 2747:
1.1.1.3 root 2748: GNU CC does not delete ``empty'' loops because the most likely reason
2749: you would put one in a program is to have a delay. Deleting them will
2750: not make real programs run any faster, so it would be pointless.
1.1 root 2751:
1.1.1.3 root 2752: It would be different if optimization of a nonempty loop could produce
2753: an empty one. But this generally can't happen.
1.1.1.5 root 2754:
2755: @item
2756: Making side effects happen in the same order as in some other compiler.
2757:
2758: @cindex side effects, order of evaluation
2759: @cindex order of evaluation, side effects
2760: It is never safe to depend on the order of evaluation of side effects.
2761: For example, a function call like this may very well behave differently
2762: from one compiler to another:
2763:
2764: @example
2765: void func (int, int);
2766:
2767: int i = 2;
2768: func (i++, i++);
2769: @end example
2770:
2771: There is no guarantee (in either the C or the C++ standard language
2772: definitions) that the increments will be evaluated in any particular
2773: order. Either increment might happen first. @code{func} might get the
2774: arguments @samp{3, 4}, or it might get @samp{4, 3}, or even @samp{3, 3}.
2775:
2776: @item
1.1.1.7 ! root 2777: Not allowing structures with volatile fields in registers.
1.1.1.5 root 2778:
1.1.1.7 ! root 2779: Strictly speaking, there is no prohibition in the ANSI C standard
! 2780: against allowing structures with volatile fields in registers, but
! 2781: it does not seem to make any sense and is probably not what you wanted
! 2782: to do. So the compiler will give an error message in this case.
1.1.1.5 root 2783: @end itemize
2784:
2785: @node Warnings and Errors
2786: @section Warning Messages and Error Messages
2787:
2788: @cindex error messages
2789: @cindex warnings vs errors
2790: @cindex messages, warning and error
2791: The GNU compiler can produce two kinds of diagnostics: errors and
2792: warnings. Each kind has a different purpose:
2793:
2794: @itemize @w{}
2795: @item
2796: @emph{Errors} report problems that make it impossible to compile your
2797: program. GNU CC reports errors with the source file name and line
2798: number where the problem is apparent.
2799:
2800: @item
2801: @emph{Warnings} report other unusual conditions in your code that
2802: @emph{may} indicate a problem, although compilation can (and does)
2803: proceed. Warning messages also report the source file name and line
2804: number, but include the text @samp{warning:} to distinguish them
2805: from error messages.
1.1.1.3 root 2806: @end itemize
2807:
1.1.1.5 root 2808: Warnings may indicate danger points where you should check to make sure
2809: that your program really does what you intend; or the use of obsolete
2810: features; or the use of nonstandard features of GNU C or C++. Many
2811: warnings are issued only if you ask for them, with one of the @samp{-W}
2812: options (for instance, @samp{-Wall} requests a variety of useful
2813: warnings).
2814:
2815: GNU CC always tries to compile your program if possible; it never
2816: gratuituously rejects a program whose meaning is clear merely because
2817: (for instance) it fails to conform to a standard. In some cases,
2818: however, the C and C++ standards specify that certain extensions are
2819: forbidden, and a diagnostic @emph{must} be issued by a conforming
2820: compiler. The @samp{-pedantic} option tells GNU CC to issue warnings in
2821: such cases; @samp{-pedantic-errors} says to make them errors instead.
2822: This does not mean that @emph{all} non-ANSI constructs get warnings
2823: or errors.
2824:
2825: @xref{Warning Options,,Options to Request or Suppress Warnings}, for
2826: more detail on these and related command-line options.
2827:
1.1.1.3 root 2828: @node Bugs
1.1 root 2829: @chapter Reporting Bugs
2830: @cindex bugs
2831: @cindex reporting bugs
2832:
2833: Your bug reports play an essential role in making GNU CC reliable.
2834:
2835: When you encounter a problem, the first thing to do is to see if it is
1.1.1.3 root 2836: already known. @xref{Trouble}. If it isn't known, then you should
2837: report the problem.
1.1 root 2838:
2839: Reporting a bug may help you by bringing a solution to your problem, or
2840: it may not. (If it does not, look in the service directory; see
2841: @ref{Service}.) In any case, the principal function of a bug report is
2842: to help the entire community by making the next version of GNU CC work
2843: better. Bug reports are your contribution to the maintenance of GNU CC.
2844:
1.1.1.6 root 2845: Since the maintainers are very overloaded, we cannot respond to every
2846: bug report. However, if the bug has not been fixed, we are likely to
2847: send you a patch and ask you to tell us whether it works.
2848:
1.1 root 2849: In order for a bug report to serve its purpose, you must include the
2850: information that makes for fixing the bug.
2851:
2852: @menu
2853: * Criteria: Bug Criteria. Have you really found a bug?
1.1.1.3 root 2854: * Where: Bug Lists. Where to send your bug report.
1.1 root 2855: * Reporting: Bug Reporting. How to report a bug effectively.
1.1.1.3 root 2856: * Patches: Sending Patches. How to send a patch for GNU CC.
1.1 root 2857: * Known: Trouble. Known problems.
2858: * Help: Service. Where to ask for help.
2859: @end menu
2860:
1.1.1.3 root 2861: @node Bug Criteria
1.1 root 2862: @section Have You Found a Bug?
2863: @cindex bug criteria
2864:
2865: If you are not sure whether you have found a bug, here are some guidelines:
2866:
2867: @itemize @bullet
2868: @cindex fatal signal
2869: @cindex core dump
2870: @item
2871: If the compiler gets a fatal signal, for any input whatever, that is a
2872: compiler bug. Reliable compilers never crash.
2873:
2874: @cindex invalid assembly code
2875: @cindex assembly code, invalid
2876: @item
2877: If the compiler produces invalid assembly code, for any input whatever
2878: (except an @code{asm} statement), that is a compiler bug, unless the
2879: compiler reports errors (not just warnings) which would ordinarily
2880: prevent the assembler from being run.
2881:
2882: @cindex undefined behavior
2883: @cindex undefined function value
2884: @cindex increment operators
2885: @item
2886: If the compiler produces valid assembly code that does not correctly
2887: execute the input source code, that is a compiler bug.
2888:
2889: However, you must double-check to make sure, because you may have run
2890: into an incompatibility between GNU C and traditional C
2891: (@pxref{Incompatibilities}). These incompatibilities might be considered
2892: bugs, but they are inescapable consequences of valuable features.
2893:
2894: Or you may have a program whose behavior is undefined, which happened
1.1.1.3 root 2895: by chance to give the desired results with another C or C++ compiler.
1.1 root 2896:
2897: For example, in many nonoptimizing compilers, you can write @samp{x;}
2898: at the end of a function instead of @samp{return x;}, with the same
2899: results. But the value of the function is undefined if @code{return}
2900: is omitted; it is not a bug when GNU CC produces different results.
2901:
2902: Problems often result from expressions with two increment operators,
2903: as in @code{f (*p++, *p++)}. Your previous compiler might have
2904: interpreted that expression the way you intended; GNU CC might
2905: interpret it another way. Neither compiler is wrong. The bug is
2906: in your code.
2907:
2908: After you have localized the error to a single source line, it should
2909: be easy to check for these things. If your program is correct and
2910: well defined, you have found a compiler bug.
2911:
2912: @item
2913: If the compiler produces an error message for valid input, that is a
2914: compiler bug.
2915:
2916: @cindex invalid input
2917: @item
2918: If the compiler does not produce an error message for invalid input,
2919: that is a compiler bug. However, you should note that your idea of
2920: ``invalid input'' might be my idea of ``an extension'' or ``support
2921: for traditional practice''.
2922:
2923: @item
1.1.1.3 root 2924: If you are an experienced user of C or C++ compilers, your suggestions
2925: for improvement of GNU CC or GNU C++ are welcome in any case.
1.1 root 2926: @end itemize
2927:
1.1.1.3 root 2928: @node Bug Lists
2929: @section Where to Report Bugs
2930: @cindex bug report mailing lists
1.1.1.5 root 2931: @kindex bug-gcc@@prep.ai.mit.edu
1.1.1.7 ! root 2932: Send bug reports for GNU C to @samp{bug-gcc@@prep.ai.mit.edu}.
1.1 root 2933:
1.1.1.5 root 2934: @kindex bug-g++@@prep.ai.mit.edu
2935: @kindex bug-libg++@@prep.ai.mit.edu
1.1.1.7 ! root 2936: Send bug reports for GNU C++ to @samp{bug-g++@@prep.ai.mit.edu}.
1.1.1.5 root 2937: If your bug involves the C++ class library libg++, send mail to
2938: @samp{bug-lib-g++@@prep.ai.mit.edu}. If you're not sure, you can send
2939: the bug report to both lists.
2940:
1.1.1.7 ! root 2941: @strong{Do not send bug reports to @samp{help-gcc@@prep.ai.mit.edu} or
1.1.1.5 root 2942: to the newsgroup @samp{gnu.gcc.help}.} Most users of GNU CC do not want
2943: to receive bug reports. Those that do, have asked to be on
2944: @samp{bug-gcc} and/or @samp{bug-g++}.
1.1.1.3 root 2945:
2946: The mailing lists @samp{bug-gcc} and @samp{bug-g++} both have newsgroups
2947: which serve as repeaters: @samp{gnu.gcc.bug} and @samp{gnu.g++.bug}.
2948: Each mailing list and its newsgroup carry exactly the same messages.
2949:
2950: Often people think of posting bug reports to the newsgroup instead of
2951: mailing them. This appears to work, but it has one problem which can be
2952: crucial: a newsgroup posting does not contain a mail path back to the
1.1.1.4 root 2953: sender. Thus, if maintainers need more information, they may be unable
1.1.1.3 root 2954: to reach you. For this reason, you should always send bug reports by
2955: mail to the proper mailing list.
1.1 root 2956:
2957: As a last resort, send bug reports on paper to:
2958:
2959: @example
2960: GNU Compiler Bugs
2961: Free Software Foundation
2962: 675 Mass Ave
2963: Cambridge, MA 02139
2964: @end example
2965:
1.1.1.3 root 2966: @node Bug Reporting
2967: @section How to Report Bugs
2968: @cindex compiler bugs, reporting
2969:
1.1 root 2970: The fundamental principle of reporting bugs usefully is this:
2971: @strong{report all the facts}. If you are not sure whether to state a
2972: fact or leave it out, state it!
2973:
2974: Often people omit facts because they think they know what causes the
2975: problem and they conclude that some details don't matter. Thus, you might
2976: assume that the name of the variable you use in an example does not matter.
2977: Well, probably it doesn't, but one cannot be sure. Perhaps the bug is a
2978: stray memory reference which happens to fetch from the location where that
2979: name is stored in memory; perhaps, if the name were different, the contents
2980: of that location would fool the compiler into doing the right thing despite
2981: the bug. Play it safe and give a specific, complete example. That is the
2982: easiest thing for you to do, and the most helpful.
2983:
1.1.1.3 root 2984: Keep in mind that the purpose of a bug report is to enable someone to
2985: fix the bug if it is not known. It isn't very important what happens if
1.1 root 2986: the bug is already known. Therefore, always write your bug reports on
2987: the assumption that the bug is not known.
2988:
2989: Sometimes people give a few sketchy facts and ask, ``Does this ring a
1.1.1.3 root 2990: bell?'' This cannot help us fix a bug, so it is basically useless. We
2991: respond by asking for enough details to enable us to investigate.
2992: You might as well expedite matters by sending them to begin with.
2993:
2994: Try to make your bug report self-contained. If we have to ask you for
2995: more information, it is best if you include all the previous information
2996: in your response, as well as the information that was missing.
1.1 root 2997:
1.1.1.7 ! root 2998: Please report each bug in a separate message. This makes it easier for
! 2999: us to track which bugs have been fixed and to forward your bugs reports
! 3000: to the appropriate maintainer.
! 3001:
1.1.1.3 root 3002: To enable someone to investigate the bug, you should include all these
3003: things:
1.1 root 3004:
3005: @itemize @bullet
3006: @item
3007: The version of GNU CC. You can get this by running it with the
3008: @samp{-v} option.
3009:
1.1.1.3 root 3010: Without this, we won't know whether there is any point in looking for
1.1 root 3011: the bug in the current version of GNU CC.
3012:
3013: @item
1.1.1.3 root 3014: A complete input file that will reproduce the bug. If the bug is in the
3015: C preprocessor, send a source file and any header files that it
1.1 root 3016: requires. If the bug is in the compiler proper (@file{cc1}), run your
3017: source file through the C preprocessor by doing @samp{gcc -E
3018: @var{sourcefile} > @var{outfile}}, then include the contents of
1.1.1.3 root 3019: @var{outfile} in the bug report. (When you do this, use the same
3020: @samp{-I}, @samp{-D} or @samp{-U} options that you used in actual
3021: compilation.)
1.1 root 3022:
1.1.1.4 root 3023: A single statement is not enough of an example. In order to compile it,
3024: it must be embedded in a complete file of compiler input; and the bug
3025: might depend on the details of how this is done.
1.1 root 3026:
1.1.1.3 root 3027: Without a real example one can compile, all anyone can do about your bug
1.1 root 3028: report is wish you luck. It would be futile to try to guess how to
1.1.1.3 root 3029: provoke the bug. For example, bugs in register allocation and reloading
3030: frequently depend on every little detail of the function they happen in.
1.1 root 3031:
1.1.1.4 root 3032: Even if the input file that fails comes from a GNU program, you should
3033: still send the complete test case. Don't ask the GNU CC maintainers to
3034: do the extra work of obtaining the program in question---they are all
3035: overworked as it is. Also, the problem may depend on what is in the
3036: header files on your system; it is unreliable for the GNU CC maintainers
3037: to try the problem with the header files available to them. By sending
1.1.1.5 root 3038: CPP output, you can eliminate this source of uncertainty and save us
3039: a certain percentage of wild goose chases.
1.1.1.4 root 3040:
1.1 root 3041: @item
1.1.1.3 root 3042: The command arguments you gave GNU CC or GNU C++ to compile that example
3043: and observe the bug. For example, did you use @samp{-O}? To guarantee
3044: you won't omit something important, list all the options.
1.1 root 3045:
1.1.1.3 root 3046: If we were to try to guess the arguments, we would probably guess wrong
3047: and then we would not encounter the bug.
1.1 root 3048:
3049: @item
3050: The type of machine you are using, and the operating system name and
3051: version number.
3052:
3053: @item
3054: The operands you gave to the @code{configure} command when you installed
3055: the compiler.
3056:
3057: @item
1.1.1.3 root 3058: A complete list of any modifications you have made to the compiler
3059: source. (We don't promise to investigate the bug unless it happens in
3060: an unmodified compiler. But if you've made modifications and don't tell
3061: us, then you are sending us on a wild goose chase.)
1.1 root 3062:
1.1.1.4 root 3063: Be precise about these changes. A description in English is not
3064: enough---send a context diff for them.
1.1 root 3065:
1.1.1.3 root 3066: Adding files of your own (such as a machine description for a machine we
3067: don't support) is a modification of the compiler source.
1.1 root 3068:
3069: @item
1.1.1.3 root 3070: Details of any other deviations from the standard procedure for installing
3071: GNU CC.
1.1 root 3072:
3073: @item
1.1.1.3 root 3074: A description of what behavior you observe that you believe is
3075: incorrect. For example, ``The compiler gets a fatal signal,'' or,
3076: ``The assembler instruction at line 208 in the output is incorrect.''
1.1 root 3077:
1.1.1.3 root 3078: Of course, if the bug is that the compiler gets a fatal signal, then one
3079: can't miss it. But if the bug is incorrect output, the maintainer might
3080: not notice unless it is glaringly wrong. None of us has time to study
3081: all the assembler code from a 50-line C program just on the chance that
1.1.1.4 root 3082: one instruction might be wrong. We need @emph{you} to do this part!
1.1 root 3083:
1.1.1.3 root 3084: Even if the problem you experience is a fatal signal, you should still
3085: say so explicitly. Suppose something strange is going on, such as, your
3086: copy of the compiler is out of synch, or you have encountered a bug in
3087: the C library on your system. (This has happened!) Your copy might
3088: crash and the copy here would not. If you @i{said} to expect a crash,
3089: then when the compiler here fails to crash, we would know that the bug
3090: was not happening. If you don't say to expect a crash, then we would
3091: not know whether the bug was happening. We would not be able to draw
3092: any conclusion from our observations.
3093:
1.1.1.4 root 3094: If the problem is a diagnostic when compiling GNU CC with some other
3095: compiler, say whether it is a warning or an error.
3096:
1.1.1.3 root 3097: Often the observed symptom is incorrect output when your program is run.
3098: Sad to say, this is not enough information unless the program is short
3099: and simple. None of us has time to study a large program to figure out
3100: how it would work if compiled correctly, much less which line of it was
3101: compiled wrong. So you will have to do that. Tell us which source line
3102: it is, and what incorrect result happens when that line is executed. A
3103: person who understands the program can find this as easily as finding a
3104: bug in the program itself.
3105:
3106: @item
3107: If you send examples of assembler code output from GNU CC or GNU C++,
3108: please use @samp{-g} when you make them. The debugging information
3109: includes source line numbers which are essential for correlating the
3110: output with the input.
3111:
3112: @item
1.1.1.4 root 3113: If you wish to mention something in the GNU CC source, refer to it by
3114: context, not by line number.
1.1.1.3 root 3115:
3116: The line numbers in the development sources don't match those in your
3117: sources. Your line numbers would convey no useful information to the
3118: maintainers.
3119:
3120: @item
3121: Additional information from a debugger might enable someone to find a
3122: problem on a machine which he does not have available. However, you
3123: need to think when you collect this information if you want it to have
3124: any chance of being useful.
1.1 root 3125:
3126: @cindex backtrace for bug reports
3127: For example, many people send just a backtrace, but that is never
3128: useful by itself. A simple backtrace with arguments conveys little
3129: about GNU CC because the compiler is largely data-driven; the same
3130: functions are called over and over for different RTL insns, doing
3131: different things depending on the details of the insn.
3132:
3133: Most of the arguments listed in the backtrace are useless because they
3134: are pointers to RTL list structure. The numeric values of the
3135: pointers, which the debugger prints in the backtrace, have no
3136: significance whatever; all that matters is the contents of the objects
3137: they point to (and most of the contents are other such pointers).
3138:
3139: In addition, most compiler passes consist of one or more loops that
3140: scan the RTL insn sequence. The most vital piece of information about
3141: such a loop---which insn it has reached---is usually in a local variable,
3142: not in an argument.
3143:
3144: @findex debug_rtx
3145: What you need to provide in addition to a backtrace are the values of
3146: the local variables for several stack frames up. When a local
3147: variable or an argument is an RTX, first print its value and then use
3148: the GDB command @code{pr} to print the RTL expression that it points
3149: to. (If GDB doesn't run on your machine, use your debugger to call
3150: the function @code{debug_rtx} with the RTX as an argument.) In
3151: general, whenever a variable is a pointer, its value is no use
3152: without the data it points to.
3153: @end itemize
3154:
3155: Here are some things that are not necessary:
3156:
3157: @itemize @bullet
3158: @item
3159: A description of the envelope of the bug.
3160:
3161: Often people who encounter a bug spend a lot of time investigating
3162: which changes to the input file will make the bug go away and which
3163: changes will not affect it.
3164:
1.1.1.3 root 3165: This is often time consuming and not very useful, because the way we
3166: will find the bug is by running a single example under the debugger with
3167: breakpoints, not by pure deduction from a series of examples. You might
3168: as well save your time for something else.
3169:
3170: Of course, if you can find a simpler example to report @emph{instead} of
3171: the original one, that is a convenience. Errors in the output will be
3172: easier to spot, running under the debugger will take less time, etc.
3173: Most GNU CC bugs involve just one function, so the most straightforward
3174: way to simplify an example is to delete all the function definitions
3175: except the one where the bug occurs. Those earlier in the file may be
3176: replaced by external declarations if the crucial function depends on
3177: them. (Exception: inline functions may affect compilation of functions
3178: defined later in the file.)
1.1 root 3179:
3180: However, simplification is not vital; if you don't want to do this,
1.1.1.3 root 3181: report the bug anyway and send the entire test case you used.
1.1 root 3182:
3183: @item
1.1.1.4 root 3184: In particular, some people insert conditionals @samp{#ifdef BUG} around
3185: a statement which, if removed, makes the bug not happen. These are just
3186: clutter; we won't pay any attention to them anyway. Besides, you should
3187: send us cpp output, and that can't have conditionals.
3188:
3189: @item
1.1 root 3190: A patch for the bug.
3191:
1.1.1.3 root 3192: A patch for the bug is useful if it is a good one. But don't omit the
3193: necessary information, such as the test case, on the assumption that a
3194: patch is all we need. We might see problems with your patch and decide
3195: to fix the problem another way, or we might not understand it at all.
1.1 root 3196:
3197: Sometimes with a program as complicated as GNU CC it is very hard to
3198: construct an example that will make the program follow a certain path
1.1.1.3 root 3199: through the code. If you don't send the example, we won't be able to
3200: construct one, so we won't be able to verify that the bug is fixed.
3201:
3202: And if we can't understand what bug you are trying to fix, or why your
3203: patch should be an improvement, we won't install it. A test case will
3204: help us to understand.
1.1 root 3205:
1.1.1.3 root 3206: @xref{Sending Patches}, for guidelines on how to make it easy for us to
3207: understand and install your patches.
1.1 root 3208:
3209: @item
3210: A guess about what the bug is or what it depends on.
3211:
3212: Such guesses are usually wrong. Even I can't guess right about such
3213: things without first using the debugger to find the facts.
1.1.1.4 root 3214:
3215: @item
3216: A core dump file.
3217:
3218: We have no way of examining a core dump for your type of machine
3219: unless we have an identical system---and if we do have one,
3220: we should be able to reproduce the crash ourselves.
1.1 root 3221: @end itemize
3222:
1.1.1.3 root 3223: @node Sending Patches,, Bug Reporting, Bugs
3224: @section Sending Patches for GNU CC
1.1 root 3225:
1.1.1.3 root 3226: If you would like to write bug fixes or improvements for the GNU C
3227: compiler, that is very helpful. When you send your changes, please
3228: follow these guidelines to avoid causing extra work for us in studying
3229: the patches.
3230:
3231: If you don't follow these guidelines, your information might still be
3232: useful, but using it will take extra work. Maintaining GNU C is a lot
3233: of work in the best of circumstances, and we can't keep up unless you do
3234: your best to help.
1.1 root 3235:
3236: @itemize @bullet
3237: @item
1.1.1.3 root 3238: Send an explanation with your changes of what problem they fix or what
3239: improvement they bring about. For a bug fix, just include a copy of the
3240: bug report, and explain why the change fixes the bug.
1.1 root 3241:
1.1.1.3 root 3242: (Referring to a bug report is not as good as including it, because then
3243: we will have to look it up, and we have probably already deleted it if
3244: we've already fixed the bug.)
1.1 root 3245:
3246: @item
1.1.1.3 root 3247: Always include a proper bug report for the problem you think you have
3248: fixed. We need to convince ourselves that the change is right before
3249: installing it. Even if it is right, we might have trouble judging it if
3250: we don't have a way to reproduce the problem.
1.1.1.2 root 3251:
1.1.1.3 root 3252: @item
3253: Include all the comments that are appropriate to help people reading the
3254: source in the future understand why this change was needed.
1.1.1.2 root 3255:
3256: @item
1.1.1.3 root 3257: Don't mix together changes made for different reasons.
3258: Send them @emph{individually}.
1.1 root 3259:
1.1.1.3 root 3260: If you make two changes for separate reasons, then we might not want to
3261: install them both. We might want to install just one. If you send them
3262: all jumbled together in a single set of diffs, we have to do extra work
3263: to disentangle them---to figure out which parts of the change serve
3264: which purpose. If we don't have time for this, we might have to ignore
3265: your changes entirely.
1.1 root 3266:
1.1.1.3 root 3267: If you send each change as soon as you have written it, with its own
3268: explanation, then the two changes never get tangled up, and we can
3269: consider each one properly without any extra work to disentangle them.
1.1 root 3270:
1.1.1.3 root 3271: Ideally, each change you send should be impossible to subdivide into
3272: parts that we might want to consider separately, because each of its
3273: parts gets its motivation from the other parts.
1.1 root 3274:
1.1.1.3 root 3275: @item
3276: Send each change as soon as that change is finished. Sometimes people
3277: think they are helping us by accumulating many changes to send them all
3278: together. As explained above, this is absolutely the worst thing you
3279: could do.
1.1 root 3280:
1.1.1.3 root 3281: Since you should send each change separately, you might as well send it
3282: right away. That gives us the option of installing it immediately if it
3283: is important.
1.1 root 3284:
1.1.1.3 root 3285: @item
3286: Use @samp{diff -c} to make your diffs. Diffs without context are hard
3287: for us to install reliably. More than that, they make it hard for us to
3288: study the diffs to decide whether we want to install them. Unidiff
3289: format is better than contextless diffs, but not as easy to read as
3290: @samp{-c} format.
1.1 root 3291:
1.1.1.3 root 3292: If you have GNU diff, use @samp{diff -cp}, which shows the name of the
3293: function that each change occurs in.
1.1 root 3294:
1.1.1.3 root 3295: @item
3296: Write the change log entries for your changes. We get lots of changes,
3297: and we don't have time to do all the change log writing ourselves.
1.1 root 3298:
1.1.1.3 root 3299: Read the @file{ChangeLog} file to see what sorts of information to put
3300: in, and to learn the style that we use. The purpose of the change log
3301: is to show people where to find what was changed. So you need to be
3302: specific about what functions you changed; in large functions, it's
3303: often helpful to indicate where within the function the change was.
1.1.1.2 root 3304:
1.1.1.3 root 3305: On the other hand, once you have shown people where to find the change,
1.1.1.5 root 3306: you need not explain its purpose. Thus, if you add a new function, all
1.1.1.3 root 3307: you need to say about it is that it is new. If you feel that the
3308: purpose needs explaining, it probably does---but the explanation will be
3309: much more useful if you put it in comments in the code.
3310:
3311: If you would like your name to appear in the header line for who made
3312: the change, send us the header line.
1.1 root 3313:
3314: @item
1.1.1.4 root 3315: When you write the fix, keep in mind that we can't install a change that
1.1.1.3 root 3316: would break other systems.
1.1 root 3317:
1.1.1.3 root 3318: People often suggest fixing a problem by changing machine-independent
3319: files such as @file{toplev.c} to do something special that a particular
3320: system needs. Sometimes it is totally obvious that such changes would
3321: break GNU CC for almost all users. We can't possibly make a change like
3322: that. At best it might tell us how to write another patch that would
3323: solve the problem acceptably.
1.1 root 3324:
1.1.1.3 root 3325: Sometimes people send fixes that @emph{might} be an improvement in
3326: general---but it is hard to be sure of this. It's hard to install
3327: such changes because we have to study them very carefully. Of course,
3328: a good explanation of the reasoning by which you concluded the change
3329: was correct can help convince us.
1.1 root 3330:
1.1.1.3 root 3331: The safest changes are changes to the configuration files for a
3332: particular machine. These are safe because they can't create new bugs
3333: on other machines.
1.1 root 3334:
1.1.1.3 root 3335: Please help us keep up with the workload by designing the patch in a
3336: form that is good to install.
3337: @end itemize
1.1 root 3338:
1.1.1.3 root 3339: @node Service
3340: @chapter How To Get Help with GNU CC
1.1 root 3341:
1.1.1.3 root 3342: If you need help installing, using or changing GNU CC, there are two
3343: ways to find it:
1.1 root 3344:
1.1.1.3 root 3345: @itemize @bullet
3346: @item
3347: Send a message to a suitable network mailing list. First try
3348: @code{bug-gcc@@prep.ai.mit.edu}, and if that brings no response, try
3349: @code{help-gcc@@prep.ai.mit.edu}.
1.1 root 3350:
1.1.1.3 root 3351: @item
3352: Look in the service directory for someone who might help you for a fee.
3353: The service directory is found in the file named @file{SERVICE} in the
3354: GNU CC distribution.
1.1 root 3355: @end itemize
3356:
1.1.1.3 root 3357: @node VMS
1.1 root 3358: @chapter Using GNU CC on VMS
3359:
1.1.1.7 ! root 3360: @c prevent bad page break with this line
! 3361: Here is how to use GNU CC on VMS.
! 3362:
1.1 root 3363: @menu
3364: * Include Files and VMS:: Where the preprocessor looks for the include files.
3365: * Global Declarations:: How to do globaldef, globalref and globalvalue with
3366: GNU CC.
3367: * VMS Misc:: Misc information.
3368: @end menu
3369:
1.1.1.3 root 3370: @node Include Files and VMS
1.1 root 3371: @section Include Files and VMS
3372:
3373: @cindex include files and VMS
3374: @cindex VMS and include files
3375: @cindex header files and VMS
3376: Due to the differences between the filesystems of Unix and VMS, GNU CC
3377: attempts to translate file names in @samp{#include} into names that VMS
3378: will understand. The basic strategy is to prepend a prefix to the
3379: specification of the include file, convert the whole filename to a VMS
3380: filename, and then try to open the file. GNU CC tries various prefixes
3381: one by one until one of them succeeds:
3382:
3383: @enumerate
3384: @item
3385: The first prefix is the @samp{GNU_CC_INCLUDE:} logical name: this is
3386: where GNU C header files are traditionally stored. If you wish to store
3387: header files in non-standard locations, then you can assign the logical
3388: @samp{GNU_CC_INCLUDE} to be a search list, where each element of the
3389: list is suitable for use with a rooted logical.
3390:
3391: @item
3392: The next prefix tried is @samp{SYS$SYSROOT:[SYSLIB.]}. This is where
3393: VAX-C header files are traditionally stored.
3394:
3395: @item
3396: If the include file specification by itself is a valid VMS filename, the
3397: preprocessor then uses this name with no prefix in an attempt to open
3398: the include file.
3399:
3400: @item
3401: If the file specification is not a valid VMS filename (i.e. does not
3402: contain a device or a directory specifier, and contains a @samp{/}
3403: character), the preprocessor tries to convert it from Unix syntax to
3404: VMS syntax.
3405:
3406: Conversion works like this: the first directory name becomes a device,
3407: and the rest of the directories are converted into VMS-format directory
1.1.1.5 root 3408: names. For example, the name @file{X11/foobar.h} is
3409: translated to @file{X11:[000000]foobar.h} or @file{X11:foobar.h},
3410: whichever one can be opened. This strategy allows you to assign a
3411: logical name to point to the actual location of the header files.
1.1 root 3412:
3413: @item
3414: If none of these strategies succeeds, the @samp{#include} fails.
3415: @end enumerate
3416:
3417: Include directives of the form:
3418:
3419: @example
3420: #include foobar
3421: @end example
3422:
3423: @noindent
3424: are a common source of incompatibility between VAX-C and GNU CC. VAX-C
3425: treats this much like a standard @code{#include <foobar.h>} directive.
3426: That is incompatible with the ANSI C behavior implemented by GNU CC: to
3427: expand the name @code{foobar} as a macro. Macro expansion should
3428: eventually yield one of the two standard formats for @code{#include}:
3429:
3430: @example
3431: #include "@var{file}"
3432: #include <@var{file}>
3433: @end example
3434:
3435: If you have this problem, the best solution is to modify the source to
3436: convert the @code{#include} directives to one of the two standard forms.
3437: That will work with either compiler. If you want a quick and dirty fix,
3438: define the file names as macros with the proper expansion, like this:
3439:
3440: @example
3441: #define stdio <stdio.h>
3442: @end example
3443:
3444: @noindent
3445: This will work, as long as the name doesn't conflict with anything else
3446: in the program.
3447:
3448: Another source of incompatibility is that VAX-C assumes that:
3449:
3450: @example
3451: #include "foobar"
3452: @end example
3453:
3454: @noindent
3455: is actually asking for the file @file{foobar.h}. GNU CC does not
3456: make this assumption, and instead takes what you ask for literally;
3457: it tries to read the file @file{foobar}. The best way to avoid this
3458: problem is to always specify the desired file extension in your include
3459: directives.
3460:
3461: GNU CC for VMS is distributed with a set of include files that is
3462: sufficient to compile most general purpose programs. Even though the
3463: GNU CC distribution does not contain header files to define constants
3464: and structures for some VMS system-specific functions, there is no
3465: reason why you cannot use GNU CC with any of these functions. You first
3466: may have to generate or create header files, either by using the public
3467: domain utility @code{UNSDL} (which can be found on a DECUS tape), or by
3468: extracting the relevant modules from one of the system macro libraries,
3469: and using an editor to construct a C header file.
3470:
1.1.1.4 root 3471: A @code{#include} file name cannot contain a DECNET node name. The
3472: preprocessor reports an I/O error if you attempt to use a node name,
3473: whether explicitly, or implicitly via a logical name.
3474:
1.1.1.3 root 3475: @node Global Declarations
1.1 root 3476: @section Global Declarations and VMS
3477:
3478: @findex GLOBALREF
3479: @findex GLOBALDEF
3480: @findex GLOBALVALUEDEF
3481: @findex GLOBALVALUEREF
3482: GNU CC does not provide the @code{globalref}, @code{globaldef} and
3483: @code{globalvalue} keywords of VAX-C. You can get the same effect with
3484: an obscure feature of GAS, the GNU assembler. (This requires GAS
3485: version 1.39 or later.) The following macros allow you to use this
3486: feature in a fairly natural way:
3487:
3488: @smallexample
3489: #ifdef __GNUC__
1.1.1.2 root 3490: #define GLOBALREF(TYPE,NAME) \
3491: TYPE NAME \
3492: asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME)
3493: #define GLOBALDEF(TYPE,NAME,VALUE) \
3494: TYPE NAME \
3495: asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME) \
3496: = VALUE
3497: #define GLOBALVALUEREF(TYPE,NAME) \
3498: const TYPE NAME[1] \
3499: asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME)
3500: #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
3501: const TYPE NAME[1] \
3502: asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME) \
3503: = @{VALUE@}
1.1 root 3504: #else
1.1.1.2 root 3505: #define GLOBALREF(TYPE,NAME) \
3506: globalref TYPE NAME
3507: #define GLOBALDEF(TYPE,NAME,VALUE) \
3508: globaldef TYPE NAME = VALUE
3509: #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
3510: globalvalue TYPE NAME = VALUE
3511: #define GLOBALVALUEREF(TYPE,NAME) \
3512: globalvalue TYPE NAME
1.1 root 3513: #endif
3514: @end smallexample
3515:
3516: @noindent
3517: (The @code{_$$PsectAttributes_GLOBALSYMBOL} prefix at the start of the
3518: name is removed by the assembler, after it has modified the attributes
3519: of the symbol). These macros are provided in the VMS binaries
3520: distribution in a header file @file{GNU_HACKS.H}. An example of the
3521: usage is:
3522:
3523: @example
1.1.1.2 root 3524: GLOBALREF (int, ijk);
3525: GLOBALDEF (int, jkl, 0);
1.1 root 3526: @end example
3527:
3528: The macros @code{GLOBALREF} and @code{GLOBALDEF} cannot be used
3529: straightforwardly for arrays, since there is no way to insert the array
3530: dimension into the declaration at the right place. However, you can
3531: declare an array with these macros if you first define a typedef for the
3532: array type, like this:
3533:
3534: @example
3535: typedef int intvector[10];
1.1.1.2 root 3536: GLOBALREF (intvector, foo);
1.1 root 3537: @end example
3538:
3539: Array and structure initializers will also break the macros; you can
3540: define the initializer to be a macro of its own, or you can expand the
3541: @code{GLOBALDEF} macro by hand. You may find a case where you wish to
3542: use the @code{GLOBALDEF} macro with a large array, but you are not
3543: interested in explicitly initializing each element of the array. In
3544: such cases you can use an initializer like: @code{@{0,@}}, which will
3545: initialize the entire array to @code{0}.
3546:
3547: A shortcoming of this implementation is that a variable declared with
3548: @code{GLOBALVALUEREF} or @code{GLOBALVALUEDEF} is always an array. For
3549: example, the declaration:
3550:
3551: @example
1.1.1.2 root 3552: GLOBALVALUEREF(int, ijk);
1.1 root 3553: @end example
3554:
3555: @noindent
3556: declares the variable @code{ijk} as an array of type @code{int [1]}.
3557: This is done because a globalvalue is actually a constant; its ``value''
3558: is what the linker would normally consider an address. That is not how
3559: an integer value works in C, but it is how an array works. So treating
3560: the symbol as an array name gives consistent results---with the
3561: exception that the value seems to have the wrong type. @strong{Don't
3562: try to access an element of the array.} It doesn't have any elements.
3563: The array ``address'' may not be the address of actual storage.
3564:
3565: The fact that the symbol is an array may lead to warnings where the
3566: variable is used. Insert type casts to avoid the warnings. Here is an
3567: example; it takes advantage of the ANSI C feature allowing macros that
3568: expand to use the same name as the macro itself.
3569:
3570: @example
1.1.1.2 root 3571: GLOBALVALUEREF (int, ss$_normal);
3572: GLOBALVALUEDEF (int, xyzzy,123);
1.1 root 3573: #ifdef __GNUC__
3574: #define ss$_normal ((int) ss$_normal)
3575: #define xyzzy ((int) xyzzy)
3576: #endif
3577: @end example
3578:
3579: Don't use @code{globaldef} or @code{globalref} with a variable whose
3580: type is an enumeration type; this is not implemented. Instead, make the
3581: variable an integer, and use a @code{globalvaluedef} for each of the
3582: enumeration values. An example of this would be:
3583:
3584: @example
3585: #ifdef __GNUC__
1.1.1.2 root 3586: GLOBALDEF (int, color, 0);
3587: GLOBALVALUEDEF (int, RED, 0);
3588: GLOBALVALUEDEF (int, BLUE, 1);
3589: GLOBALVALUEDEF (int, GREEN, 3);
1.1 root 3590: #else
3591: enum globaldef color @{RED, BLUE, GREEN = 3@};
3592: #endif
3593: @end example
3594:
1.1.1.3 root 3595: @node VMS Misc
1.1 root 3596: @section Other VMS Issues
3597:
3598: @cindex exit status and VMS
3599: @cindex return value of @code{main}
3600: @cindex @code{main} and the exit status
3601: GNU CC automatically arranges for @code{main} to return 1 by default if
3602: you fail to specify an explicit return value. This will be interpreted
3603: by VMS as a status code indicating a normal successful completion.
3604: Version 1 of GNU CC did not provide this default.
3605:
3606: GNU CC on VMS works only with the GNU assembler, GAS. You need version
3607: 1.37 or later of GAS in order to produce value debugging information for
3608: the VMS debugger. Use the ordinary VMS linker with the object files
3609: produced by GAS.
3610:
3611: @cindex shared VMS run time system
3612: @cindex @file{VAXCRTL}
3613: Under previous versions of GNU CC, the generated code would occasionally
3614: give strange results when linked to the sharable @file{VAXCRTL} library.
3615: Now this should work.
3616:
3617: A caveat for use of @code{const} global variables: the @code{const}
3618: modifier must be specified in every external declaration of the variable
3619: in all of the source files that use that variable. Otherwise the linker
3620: will issue warnings about conflicting attributes for the variable. Your
3621: program will still work despite the warnings, but the variable will be
3622: placed in writable storage.
3623:
3624: @cindex name augmentation
3625: @cindex case sensitivity and VMS
3626: @cindex VMS and case sensitivity
1.1.1.4 root 3627: Although the VMS linker does distinguish between upper and lower case
3628: letters in global symbols, most VMS compilers convert all such symbols
3629: into upper case and most run-time library routines also have upper case
3630: names. To be able to reliably call such routines, GNU CC (by means of
3631: the assembler GAS) converts global symbols into upper case like other
3632: VMS compilers. However, since the usual practice in C is to distinguish
3633: case, GNU CC (via GAS) tries to preserve usual C behavior by augmenting
3634: each name that is not all lower case. This means truncating the name
3635: to at most 23 characters and then adding more characters at the end
3636: which encode the case pattern of those 23. Names which contain at
3637: least one dollar sign are an exception; they are converted directly into
3638: upper case without augmentation.
1.1 root 3639:
3640: Name augmentation yields bad results for programs that use precompiled
3641: libraries (such as Xlib) which were generated by another compiler. You
3642: can use the compiler option @samp{/NOCASE_HACK} to inhibit augmentation;
3643: it makes external C functions and variables case-independent as is usual
3644: on VMS. Alternatively, you could write all references to the functions
3645: and variables in such libraries using lower case; this will work on VMS,
1.1.1.4 root 3646: but is not portable to other systems. The compiler option @samp{/NAMES}
3647: also provides control over global name handling.
1.1 root 3648:
3649: Function and variable names are handled somewhat differently with GNU
3650: C++. The GNU C++ compiler performs @dfn{name mangling} on function
3651: names, which means that it adds information to the function name to
1.1.1.5 root 3652: describe the data types of the arguments that the function takes. One
1.1 root 3653: result of this is that the name of a function can become very long.
3654: Since the VMS linker only recognizes the first 31 characters in a name,
3655: special action is taken to ensure that each function and variable has a
3656: unique name that can be represented in 31 characters.
3657:
3658: If the name (plus a name augmentation, if required) is less than 32
1.1.1.5 root 3659: characters in length, then no special action is performed. If the name
1.1 root 3660: is longer than 31 characters, the assembler (GAS) will generate a
3661: hash string based upon the function name, truncate the function name to
3662: 23 characters, and append the hash string to the truncated name. If the
3663: @samp{/VERBOSE} compiler option is used, the assembler will print both
3664: the full and truncated names of each symbol that is truncated.
3665:
3666: The @samp{/NOCASE_HACK} compiler option should not be used when you are
1.1.1.5 root 3667: compiling programs that use libg++. libg++ has several instances of
1.1 root 3668: objects (i.e. @code{Filebuf} and @code{filebuf}) which become
3669: indistinguishable in a case-insensitive environment. This leads to
3670: cases where you need to inhibit augmentation selectively (if you were
3671: using libg++ and Xlib in the same program, for example). There is no
3672: special feature for doing this, but you can get the result by defining a
3673: macro for each mixed case symbol for which you wish to inhibit
3674: augmentation. The macro should expand into the lower case equivalent of
3675: itself. For example:
3676:
3677: @example
3678: #define StuDlyCapS studlycaps
3679: @end example
3680:
3681: These macro definitions can be placed in a header file to minimize the
3682: number of changes to your source code.
1.1.1.5 root 3683: @end ifset
1.1 root 3684:
3685: @ifset INTERNALS
1.1.1.3 root 3686: @node Portability
1.1 root 3687: @chapter GNU CC and Portability
3688: @cindex portability
3689: @cindex GNU CC and portability
3690:
3691: The main goal of GNU CC was to make a good, fast compiler for machines in
3692: the class that the GNU system aims to run on: 32-bit machines that address
3693: 8-bit bytes and have several general registers. Elegance, theoretical
3694: power and simplicity are only secondary.
3695:
3696: GNU CC gets most of the information about the target machine from a machine
3697: description which gives an algebraic formula for each of the machine's
3698: instructions. This is a very clean way to describe the target. But when
3699: the compiler needs information that is difficult to express in this
3700: fashion, I have not hesitated to define an ad-hoc parameter to the machine
3701: description. The purpose of portability is to reduce the total work needed
3702: on the compiler; it was not of interest for its own sake.
3703:
3704: @cindex endianness
3705: @cindex autoincrement addressing, availability
3706: @findex abort
3707: GNU CC does not contain machine dependent code, but it does contain code
3708: that depends on machine parameters such as endianness (whether the most
3709: significant byte has the highest or lowest address of the bytes in a word)
3710: and the availability of autoincrement addressing. In the RTL-generation
3711: pass, it is often necessary to have multiple strategies for generating code
3712: for a particular kind of syntax tree, strategies that are usable for different
3713: combinations of parameters. Often I have not tried to address all possible
3714: cases, but only the common ones or only the ones that I have encountered.
3715: As a result, a new target may require additional strategies. You will know
3716: if this happens because the compiler will call @code{abort}. Fortunately,
3717: the new strategies can be added in a machine-independent fashion, and will
3718: affect only the target machines that need them.
3719: @end ifset
3720:
3721: @ifset INTERNALS
1.1.1.3 root 3722: @node Interface
1.1 root 3723: @chapter Interfacing to GNU CC Output
3724: @cindex interfacing to GNU CC output
3725: @cindex run-time conventions
3726: @cindex function call conventions
3727: @cindex conventions, run-time
3728:
3729: GNU CC is normally configured to use the same function calling convention
3730: normally in use on the target system. This is done with the
1.1.1.2 root 3731: machine-description macros described (@pxref{Target Macros}).
1.1 root 3732:
3733: @cindex unions, returning
3734: @cindex structures, returning
3735: @cindex returning structures and unions
3736: However, returning of structure and union values is done differently on
3737: some target machines. As a result, functions compiled with PCC
3738: returning such types cannot be called from code compiled with GNU CC,
3739: and vice versa. This does not cause trouble often because few Unix
3740: library routines return structures or unions.
3741:
3742: GNU CC code returns structures and unions that are 1, 2, 4 or 8 bytes
3743: long in the same registers used for @code{int} or @code{double} return
3744: values. (GNU CC typically allocates variables of such types in
3745: registers also.) Structures and unions of other sizes are returned by
3746: storing them into an address passed by the caller (usually in a
3747: register). The machine-description macros @code{STRUCT_VALUE} and
3748: @code{STRUCT_INCOMING_VALUE} tell GNU CC where to pass this address.
3749:
3750: By contrast, PCC on most target machines returns structures and unions
3751: of any size by copying the data into an area of static storage, and then
3752: returning the address of that storage as if it were a pointer value.
3753: The caller must copy the data from that memory area to the place where
3754: the value is wanted. This is slower than the method used by GNU CC, and
3755: fails to be reentrant.
3756:
3757: On some target machines, such as RISC machines and the 80386, the
3758: standard system convention is to pass to the subroutine the address of
3759: where to return the value. On these machines, GNU CC has been
3760: configured to be compatible with the standard compiler, when this method
3761: is used. It may not be compatible for structures of 1, 2, 4 or 8 bytes.
3762:
3763: @cindex argument passing
3764: @cindex passing arguments
3765: GNU CC uses the system's standard convention for passing arguments. On
3766: some machines, the first few arguments are passed in registers; in
3767: others, all are passed on the stack. It would be possible to use
3768: registers for argument passing on any machine, and this would probably
3769: result in a significant speedup. But the result would be complete
3770: incompatibility with code that follows the standard convention. So this
3771: change is practical only if you are switching to GNU CC as the sole C
3772: compiler for the system. We may implement register argument passing on
3773: certain machines once we have a complete GNU system so that we can
3774: compile the libraries with GNU CC.
3775:
3776: On some machines (particularly the Sparc), certain types of arguments
3777: are passed ``by invisible reference''. This means that the value is
3778: stored in memory, and the address of the memory location is passed to
3779: the subroutine.
3780:
3781: @cindex @code{longjmp} and automatic variables
3782: If you use @code{longjmp}, beware of automatic variables. ANSI C says that
3783: automatic variables that are not declared @code{volatile} have undefined
3784: values after a @code{longjmp}. And this is all GNU CC promises to do,
3785: because it is very difficult to restore register variables correctly, and
3786: one of GNU CC's features is that it can put variables in registers without
3787: your asking it to.
3788:
3789: If you want a variable to be unaltered by @code{longjmp}, and you don't
3790: want to write @code{volatile} because old C compilers don't accept it,
3791: just take the address of the variable. If a variable's address is ever
3792: taken, even if just to compute it and ignore it, then the variable cannot
3793: go in a register:
3794:
3795: @example
3796: @{
3797: int careful;
3798: &careful;
3799: @dots{}
3800: @}
3801: @end example
3802:
3803: @cindex arithmetic libraries
3804: @cindex math libraries
3805: Code compiled with GNU CC may call certain library routines. Most of
3806: them handle arithmetic for which there are no instructions. This
3807: includes multiply and divide on some machines, and floating point
3808: operations on any machine for which floating point support is disabled
3809: with @samp{-msoft-float}. Some standard parts of the C library, such as
3810: @code{bcopy} or @code{memcpy}, are also called automatically. The usual
3811: function call interface is used for calling the library routines.
3812:
3813: These library routines should be defined in the library @file{libgcc.a},
3814: which GNU CC automatically searches whenever it links a program. On
3815: machines that have multiply and divide instructions, if hardware
3816: floating point is in use, normally @file{libgcc.a} is not needed, but it
3817: is searched just in case.
3818:
3819: Each arithmetic function is defined in @file{libgcc1.c} to use the
3820: corresponding C arithmetic operator. As long as the file is compiled
3821: with another C compiler, which supports all the C arithmetic operators,
3822: this file will work portably. However, @file{libgcc1.c} does not work if
3823: compiled with GNU CC, because each arithmetic function would compile
3824: into a call to itself!
3825: @end ifset
3826:
3827: @ifset INTERNALS
1.1.1.3 root 3828: @node Passes
1.1 root 3829: @chapter Passes and Files of the Compiler
3830: @cindex passes and files of the compiler
3831: @cindex files and passes of the compiler
3832: @cindex compiler passes and files
3833:
3834: @cindex top level of compiler
3835: The overall control structure of the compiler is in @file{toplev.c}. This
3836: file is responsible for initialization, decoding arguments, opening and
3837: closing files, and sequencing the passes.
3838:
3839: @cindex parsing pass
3840: The parsing pass is invoked only once, to parse the entire input. The RTL
3841: intermediate code for a function is generated as the function is parsed, a
3842: statement at a time. Each statement is read in as a syntax tree and then
3843: converted to RTL; then the storage for the tree for the statement is
3844: reclaimed. Storage for types (and the expressions for their sizes),
3845: declarations, and a representation of the binding contours and how they nest,
3846: remain until the function is finished being compiled; these are all needed
3847: to output the debugging information.
3848:
3849: @findex rest_of_compilation
3850: @findex rest_of_decl_compilation
3851: Each time the parsing pass reads a complete function definition or
1.1.1.5 root 3852: top-level declaration, it calls either the function
3853: @code{rest_of_compilation}, or the function
3854: @code{rest_of_decl_compilation} in @file{toplev.c}, which are
3855: responsible for all further processing necessary, ending with output of
3856: the assembler language. All other compiler passes run, in sequence,
3857: within @code{rest_of_compilation}. When that function returns from
3858: compiling a function definition, the storage used for that function
3859: definition's compilation is entirely freed, unless it is an inline
3860: function
3861: @ifset USING
3862: (@pxref{Inline,,An Inline Function is As Fast As a Macro}).
3863: @end ifset
3864: @ifclear USING
3865: (@pxref{Inline,,An Inline Function is As Fast As a Macro,gcc.texi,Using GCC}).
3866: @end ifclear
1.1 root 3867:
3868: Here is a list of all the passes of the compiler and their source files.
3869: Also included is a description of where debugging dumps can be requested
3870: with @samp{-d} options.
3871:
3872: @itemize @bullet
3873: @item
3874: Parsing. This pass reads the entire text of a function definition,
3875: constructing partial syntax trees. This and RTL generation are no longer
3876: truly separate passes (formerly they were), but it is easier to think
3877: of them as separate.
3878:
3879: The tree representation does not entirely follow C syntax, because it is
3880: intended to support other languages as well.
3881:
3882: Language-specific data type analysis is also done in this pass, and every
3883: tree node that represents an expression has a data type attached.
3884: Variables are represented as declaration nodes.
3885:
3886: @cindex constant folding
3887: @cindex arithmetic simplifications
3888: @cindex simplifications, arithmetic
3889: Constant folding and some arithmetic simplifications are also done
3890: during this pass.
3891:
3892: The language-independent source files for parsing are
3893: @file{stor-layout.c}, @file{fold-const.c}, and @file{tree.c}.
3894: There are also header files @file{tree.h} and @file{tree.def}
3895: which define the format of the tree representation.@refill
3896:
1.1.1.5 root 3897: @c Avoiding overfull is tricky here.
3898: The source files to parse C are
3899: @file{c-parse.in},
3900: @file{c-decl.c},
3901: @file{c-typeck.c},
3902: @file{c-aux-info.c},
3903: @file{c-convert.c},
3904: and @file{c-lang.c}
3905: along with header files
3906: @file{c-lex.h}, and
1.1 root 3907: @file{c-tree.h}.
3908:
3909: The source files for parsing C++ are @file{cp-parse.y},
1.1.1.4 root 3910: @file{cp-class.c},@*
3911: @file{cp-cvt.c}, @file{cp-decl.c}, @file{cp-decl2.c},
1.1 root 3912: @file{cp-dem.c}, @file{cp-except.c},@*
3913: @file{cp-expr.c}, @file{cp-init.c}, @file{cp-lex.c},
3914: @file{cp-method.c}, @file{cp-ptree.c},@*
3915: @file{cp-search.c}, @file{cp-tree.c}, @file{cp-type2.c}, and
3916: @file{cp-typeck.c}, along with header files @file{cp-tree.def},
3917: @file{cp-tree.h}, and @file{cp-decl.h}.
3918:
3919: The special source files for parsing Objective C are
3920: @file{objc-parse.y}, @file{objc-actions.c}, @file{objc-tree.def}, and
3921: @file{objc-actions.h}. Certain C-specific files are used for this as
3922: well.
3923:
3924: The file @file{c-common.c} is also used for all of the above languages.
3925:
3926: @cindex RTL generation
3927: @item
3928: RTL generation. This is the conversion of syntax tree into RTL code.
3929: It is actually done statement-by-statement during parsing, but for
3930: most purposes it can be thought of as a separate pass.
3931:
3932: @cindex target-parameter-dependent code
3933: This is where the bulk of target-parameter-dependent code is found,
3934: since often it is necessary for strategies to apply only when certain
3935: standard kinds of instructions are available. The purpose of named
3936: instruction patterns is to provide this information to the RTL
3937: generation pass.
3938:
3939: @cindex tail recursion optimization
3940: Optimization is done in this pass for @code{if}-conditions that are
3941: comparisons, boolean operations or conditional expressions. Tail
3942: recursion is detected at this time also. Decisions are made about how
3943: best to arrange loops and how to output @code{switch} statements.
3944:
1.1.1.5 root 3945: @c Avoiding overfull is tricky here.
3946: The source files for RTL generation include
3947: @file{stmt.c},
3948: @file{calls.c},
3949: @file{expr.c},
3950: @file{explow.c},
3951: @file{expmed.c},
3952: @file{function.c},
3953: @file{optabs.c}
3954: and @file{emit-rtl.c}.
3955: Also, the file
1.1 root 3956: @file{insn-emit.c}, generated from the machine description by the
3957: program @code{genemit}, is used in this pass. The header file
3958: @file{expr.h} is used for communication within this pass.@refill
3959:
3960: @findex genflags
3961: @findex gencodes
3962: The header files @file{insn-flags.h} and @file{insn-codes.h},
3963: generated from the machine description by the programs @code{genflags}
3964: and @code{gencodes}, tell this pass which standard names are available
3965: for use and which patterns correspond to them.@refill
3966:
3967: Aside from debugging information output, none of the following passes
3968: refers to the tree structure representation of the function (only
3969: part of which is saved).
3970:
3971: @cindex inline, automatic
3972: The decision of whether the function can and should be expanded inline
3973: in its subsequent callers is made at the end of rtl generation. The
3974: function must meet certain criteria, currently related to the size of
3975: the function and the types and number of parameters it has. Note that
3976: this function may contain loops, recursive calls to itself
3977: (tail-recursive functions can be inlined!), gotos, in short, all
3978: constructs supported by GNU CC. The file @file{integrate.c} contains
3979: the code to save a function's rtl for later inlining and to inline that
3980: rtl when the function is called. The header file @file{integrate.h}
3981: is also used for this purpose.
3982:
3983: The option @samp{-dr} causes a debugging dump of the RTL code after
3984: this pass. This dump file's name is made by appending @samp{.rtl} to
3985: the input file name.
3986:
3987: @cindex jump optimization
3988: @cindex unreachable code
3989: @cindex dead code
3990: @item
3991: Jump optimization. This pass simplifies jumps to the following
3992: instruction, jumps across jumps, and jumps to jumps. It deletes
3993: unreferenced labels and unreachable code, except that unreachable code
3994: that contains a loop is not recognized as unreachable in this pass.
3995: (Such loops are deleted later in the basic block analysis.) It also
3996: converts some code originally written with jumps into sequences of
3997: instructions that directly set values from the results of comparisons,
3998: if the machine has such instructions.
3999:
4000: Jump optimization is performed two or three times. The first time is
4001: immediately following RTL generation. The second time is after CSE,
4002: but only if CSE says repeated jump optimization is needed. The
4003: last time is right before the final pass. That time, cross-jumping
4004: and deletion of no-op move instructions are done together with the
4005: optimizations described above.
4006:
4007: The source file of this pass is @file{jump.c}.
4008:
4009: The option @samp{-dj} causes a debugging dump of the RTL code after
4010: this pass is run for the first time. This dump file's name is made by
4011: appending @samp{.jump} to the input file name.
4012:
4013: @cindex register use analysis
4014: @item
4015: Register scan. This pass finds the first and last use of each
4016: register, as a guide for common subexpression elimination. Its source
4017: is in @file{regclass.c}.
4018:
4019: @cindex jump threading
4020: @item
4021: Jump threading. This pass detects a condition jump that branches to an
4022: identical or inverse test. Such jumps can be @samp{threaded} through
4023: the second conditional test. The source code for this pass is in
4024: @file{jump.c}. This optimization is only performed if
4025: @samp{-fthread-jumps} is enabled.
4026:
4027: @cindex common subexpression elimination
4028: @cindex constant propagation
4029: @item
4030: Common subexpression elimination. This pass also does constant
4031: propagation. Its source file is @file{cse.c}. If constant
4032: propagation causes conditional jumps to become unconditional or to
4033: become no-ops, jump optimization is run again when CSE is finished.
4034:
4035: The option @samp{-ds} causes a debugging dump of the RTL code after
4036: this pass. This dump file's name is made by appending @samp{.cse} to
4037: the input file name.
4038:
4039: @cindex loop optimization
4040: @cindex code motion
4041: @cindex strength-reduction
4042: @item
4043: Loop optimization. This pass moves constant expressions out of loops,
4044: and optionally does strength-reduction and loop unrolling as well.
4045: Its source files are @file{loop.c} and @file{unroll.c}, plus the header
4046: @file{loop.h} used for communication between them. Loop unrolling uses
4047: some functions in @file{integrate.c} and the header @file{integrate.h}.
4048:
4049: The option @samp{-dL} causes a debugging dump of the RTL code after
4050: this pass. This dump file's name is made by appending @samp{.loop} to
4051: the input file name.
4052:
4053: @item
4054: If @samp{-frerun-cse-after-loop} was enabled, a second common
4055: subexpression elimination pass is performed after the loop optimization
4056: pass. Jump threading is also done again at this time if it was specified.
4057:
4058: The option @samp{-dt} causes a debugging dump of the RTL code after
4059: this pass. This dump file's name is made by appending @samp{.cse2} to
4060: the input file name.
4061:
4062: @cindex register allocation, stupid
4063: @cindex stupid register allocation
4064: @item
4065: Stupid register allocation is performed at this point in a
4066: nonoptimizing compilation. It does a little data flow analysis as
4067: well. When stupid register allocation is in use, the next pass
4068: executed is the reloading pass; the others in between are skipped.
4069: The source file is @file{stupid.c}.
4070:
4071: @cindex data flow analysis
4072: @cindex analysis, data flow
4073: @cindex basic blocks
4074: @item
4075: Data flow analysis (@file{flow.c}). This pass divides the program
4076: into basic blocks (and in the process deletes unreachable loops); then
4077: it computes which pseudo-registers are live at each point in the
4078: program, and makes the first instruction that uses a value point at
4079: the instruction that computed the value.
4080:
4081: @cindex autoincrement/decrement analysis
4082: This pass also deletes computations whose results are never used, and
4083: combines memory references with add or subtract instructions to make
4084: autoincrement or autodecrement addressing.
4085:
4086: The option @samp{-df} causes a debugging dump of the RTL code after
4087: this pass. This dump file's name is made by appending @samp{.flow} to
4088: the input file name. If stupid register allocation is in use, this
4089: dump file reflects the full results of such allocation.
4090:
4091: @cindex instruction combination
4092: @item
4093: Instruction combination (@file{combine.c}). This pass attempts to
4094: combine groups of two or three instructions that are related by data
4095: flow into single instructions. It combines the RTL expressions for
4096: the instructions by substitution, simplifies the result using algebra,
4097: and then attempts to match the result against the machine description.
4098:
4099: The option @samp{-dc} causes a debugging dump of the RTL code after
4100: this pass. This dump file's name is made by appending @samp{.combine}
4101: to the input file name.
4102:
4103: @cindex instruction scheduling
4104: @cindex scheduling, instruction
4105: @item
4106: Instruction scheduling (@file{sched.c}). This pass looks for
4107: instructions whose output will not be available by the time that it is
4108: used in subsequent instructions. (Memory loads and floating point
4109: instructions often have this behavior on RISC machines). It re-orders
4110: instructions within a basic block to try to separate the definition and
4111: use of items that otherwise would cause pipeline stalls.
4112:
4113: Instruction scheduling is performed twice. The first time is immediately
4114: after instruction combination and the second is immediately after reload.
4115:
4116: The option @samp{-dS} causes a debugging dump of the RTL code after this
4117: pass is run for the first time. The dump file's name is made by
4118: appending @samp{.sched} to the input file name.
4119:
4120: @cindex register class preference pass
4121: @item
4122: Register class preferencing. The RTL code is scanned to find out
4123: which register class is best for each pseudo register. The source
4124: file is @file{regclass.c}.
4125:
4126: @cindex register allocation
4127: @cindex local register allocation
4128: @item
4129: Local register allocation (@file{local-alloc.c}). This pass allocates
4130: hard registers to pseudo registers that are used only within one basic
4131: block. Because the basic block is linear, it can use fast and
4132: powerful techniques to do a very good job.
4133:
4134: The option @samp{-dl} causes a debugging dump of the RTL code after
4135: this pass. This dump file's name is made by appending @samp{.lreg} to
4136: the input file name.
4137:
4138: @cindex global register allocation
4139: @item
1.1.1.4 root 4140: Global register allocation (@file{global.c}). This pass
1.1 root 4141: allocates hard registers for the remaining pseudo registers (those
4142: whose life spans are not contained in one basic block).
4143:
4144: @cindex reloading
4145: @item
4146: Reloading. This pass renumbers pseudo registers with the hardware
4147: registers numbers they were allocated. Pseudo registers that did not
4148: get hard registers are replaced with stack slots. Then it finds
4149: instructions that are invalid because a value has failed to end up in
4150: a register, or has ended up in a register of the wrong kind. It fixes
4151: up these instructions by reloading the problematical values
4152: temporarily into registers. Additional instructions are generated to
4153: do the copying.
4154:
4155: The reload pass also optionally eliminates the frame pointer and inserts
4156: instructions to save and restore call-clobbered registers around calls.
4157:
4158: Source files are @file{reload.c} and @file{reload1.c}, plus the header
4159: @file{reload.h} used for communication between them.
4160:
4161: The option @samp{-dg} causes a debugging dump of the RTL code after
4162: this pass. This dump file's name is made by appending @samp{.greg} to
4163: the input file name.
4164:
4165: @cindex instruction scheduling
4166: @cindex scheduling, instruction
4167: @item
4168: Instruction scheduling is repeated here to try to avoid pipeline stalls
4169: due to memory loads generated for spilled pseudo registers.
4170:
4171: The option @samp{-dR} causes a debugging dump of the RTL code after
4172: this pass. This dump file's name is made by appending @samp{.sched2}
4173: to the input file name.
4174:
4175: @cindex cross-jumping
4176: @cindex no-op move instructions
4177: @item
4178: Jump optimization is repeated, this time including cross-jumping
4179: and deletion of no-op move instructions.
4180:
4181: The option @samp{-dJ} causes a debugging dump of the RTL code after
4182: this pass. This dump file's name is made by appending @samp{.jump2}
4183: to the input file name.
4184:
4185: @cindex delayed branch scheduling
4186: @cindex scheduling, delayed branch
4187: @item
4188: Delayed branch scheduling. This optional pass attempts to find
4189: instructions that can go into the delay slots of other instructions,
4190: usually jumps and calls. The source file name is @file{reorg.c}.
4191:
4192: The option @samp{-dd} causes a debugging dump of the RTL code after
4193: this pass. This dump file's name is made by appending @samp{.dbr}
4194: to the input file name.
4195:
4196: @cindex register-to-stack conversion
4197: @item
4198: Conversion from usage of some hard registers to usage of a register
4199: stack may be done at this point. Currently, this is supported only
4200: for the floating-point registers of the Intel 80387 coprocessor. The
4201: source file name is @file{reg-stack.c}.
4202:
4203: The options @samp{-dk} causes a debugging dump of the RTL code after
4204: this pass. This dump file's name is made by appending @samp{.stack}
4205: to the input file name.
4206:
4207: @cindex final pass
4208: @cindex peephole optimization
4209: @item
4210: Final. This pass outputs the assembler code for the function. It is
4211: also responsible for identifying spurious test and compare
4212: instructions. Machine-specific peephole optimizations are performed
4213: at the same time. The function entry and exit sequences are generated
4214: directly as assembler code in this pass; they never exist as RTL.
4215:
4216: The source files are @file{final.c} plus @file{insn-output.c}; the
4217: latter is generated automatically from the machine description by the
4218: tool @file{genoutput}. The header file @file{conditions.h} is used
4219: for communication between these files.
4220:
4221: @cindex debugging information generation
4222: @item
4223: Debugging information output. This is run after final because it must
4224: output the stack slot offsets for pseudo registers that did not get
4225: hard registers. Source files are @file{dbxout.c} for DBX symbol table
4226: format, @file{sdbout.c} for SDB symbol table format, and
4227: @file{dwarfout.c} for DWARF symbol table format.
4228: @end itemize
4229:
4230: Some additional files are used by all or many passes:
4231:
4232: @itemize @bullet
4233: @item
4234: Every pass uses @file{machmode.def} and @file{machmode.h} which define
4235: the machine modes.
4236:
4237: @item
4238: Several passes use @file{real.h}, which defines the default
4239: representation of floating point constants and how to operate on them.
4240:
4241: @item
4242: All the passes that work with RTL use the header files @file{rtl.h}
4243: and @file{rtl.def}, and subroutines in file @file{rtl.c}. The tools
4244: @code{gen*} also use these files to read and work with the machine
4245: description RTL.
4246:
4247: @findex genconfig
4248: @item
4249: Several passes refer to the header file @file{insn-config.h} which
4250: contains a few parameters (C macro definitions) generated
4251: automatically from the machine description RTL by the tool
4252: @code{genconfig}.
4253:
4254: @cindex instruction recognizer
4255: @item
4256: Several passes use the instruction recognizer, which consists of
4257: @file{recog.c} and @file{recog.h}, plus the files @file{insn-recog.c}
4258: and @file{insn-extract.c} that are generated automatically from the
4259: machine description by the tools @file{genrecog} and
4260: @file{genextract}.@refill
4261:
4262: @item
4263: Several passes use the header files @file{regs.h} which defines the
4264: information recorded about pseudo register usage, and @file{basic-block.h}
4265: which defines the information recorded about basic blocks.
4266:
4267: @item
4268: @file{hard-reg-set.h} defines the type @code{HARD_REG_SET}, a bit-vector
4269: with a bit for each hard register, and some macros to manipulate it.
4270: This type is just @code{int} if the machine has few enough hard registers;
4271: otherwise it is an array of @code{int} and some of the macros expand
4272: into loops.
4273:
4274: @item
4275: Several passes use instruction attributes. A definition of the
4276: attributes defined for a particular machine is in file
4277: @file{insn-attr.h}, which is generated from the machine description by
4278: the program @file{genattr}. The file @file{insn-attrtab.c} contains
4279: subroutines to obtain the attribute values for insns. It is generated
4280: from the machine description by the program @file{genattrtab}.@refill
4281: @end itemize
4282: @end ifset
4283:
1.1.1.5 root 4284: @ifset INTERNALS
1.1 root 4285: @include rtl.texi
4286: @include md.texi
4287: @include tm.texi
1.1.1.5 root 4288: @end ifset
1.1 root 4289:
4290: @ifset INTERNALS
1.1.1.3 root 4291: @node Config
1.1 root 4292: @chapter The Configuration File
4293: @cindex configuration file
4294: @cindex @file{xm-@var{machine}.h}
4295:
4296: The configuration file @file{xm-@var{machine}.h} contains macro
4297: definitions that describe the machine and system on which the compiler
4298: is running, unlike the definitions in @file{@var{machine}.h}, which
4299: describe the machine for which the compiler is producing output. Most
4300: of the values in @file{xm-@var{machine}.h} are actually the same on all
4301: machines that GNU CC runs on, so large parts of all configuration files
4302: are identical. But there are some macros that vary:
4303:
4304: @table @code
4305: @findex USG
4306: @item USG
4307: Define this macro if the host system is System V.
4308:
4309: @findex VMS
4310: @item VMS
4311: Define this macro if the host system is VMS.
4312:
4313: @findex FAILURE_EXIT_CODE
4314: @item FAILURE_EXIT_CODE
4315: A C expression for the status code to be returned when the compiler
4316: exits after serious errors.
4317:
4318: @findex SUCCESS_EXIT_CODE
4319: @item SUCCESS_EXIT_CODE
4320: A C expression for the status code to be returned when the compiler
4321: exits without serious errors.
4322:
4323: @findex HOST_WORDS_BIG_ENDIAN
4324: @item HOST_WORDS_BIG_ENDIAN
4325: Defined if the host machine stores words of multi-word values in
4326: big-endian order. (GNU CC does not depend on the host byte ordering
4327: within a word.)
4328:
1.1.1.6 root 4329: @findex HOST_FLOAT_WORDS_BIG_ENDIAN
4330: @item HOST_FLOAT_WORDS_BIG_ENDIAN
4331: Define this macro to be 1 if the host machine stores @code{DFmode},
4332: @code{XFmode} or @code{TFmode} floating point numbers in memory with the
4333: word containing the sign bit at the lowest address; otherwise, define it
4334: to be zero.
4335:
4336: This macro need not be defined if the ordering is the same as for
4337: multi-word integers.
4338:
1.1 root 4339: @findex HOST_FLOAT_FORMAT
4340: @item HOST_FLOAT_FORMAT
4341: A numeric code distinguishing the floating point format for the host
4342: machine. See @code{TARGET_FLOAT_FORMAT} in @ref{Storage Layout} for the
4343: alternatives and default.
4344:
4345: @findex HOST_BITS_PER_CHAR
4346: @item HOST_BITS_PER_CHAR
4347: A C expression for the number of bits in @code{char} on the host
4348: machine.
4349:
4350: @findex HOST_BITS_PER_SHORT
4351: @item HOST_BITS_PER_SHORT
4352: A C expression for the number of bits in @code{short} on the host
4353: machine.
4354:
4355: @findex HOST_BITS_PER_INT
4356: @item HOST_BITS_PER_INT
4357: A C expression for the number of bits in @code{int} on the host
4358: machine.
4359:
4360: @findex HOST_BITS_PER_LONG
4361: @item HOST_BITS_PER_LONG
4362: A C expression for the number of bits in @code{long} on the host
4363: machine.
4364:
4365: @findex ONLY_INT_FIELDS
4366: @item ONLY_INT_FIELDS
4367: Define this macro to indicate that the host compiler only supports
4368: @code{int} bit fields, rather than other integral types, including
4369: @code{enum}, as do most C compilers.
4370:
4371: @findex EXECUTABLE_SUFFIX
4372: @item EXECUTABLE_SUFFIX
4373: Define this macro if the host system uses a naming convention for
4374: executable files that involves a common suffix (such as, in some
4375: systems, @samp{.exe}) that must be mentioned explicitly when you run
4376: the program.
4377:
4378: @findex OBSTACK_CHUNK_SIZE
4379: @item OBSTACK_CHUNK_SIZE
4380: A C expression for the size of ordinary obstack chunks.
4381: If you don't define this, a usually-reasonable default is used.
4382:
4383: @findex OBSTACK_CHUNK_ALLOC
4384: @item OBSTACK_CHUNK_ALLOC
4385: The function used to allocate obstack chunks.
4386: If you don't define this, @code{xmalloc} is used.
4387:
4388: @findex OBSTACK_CHUNK_FREE
4389: @item OBSTACK_CHUNK_FREE
4390: The function used to free obstack chunks.
4391: If you don't define this, @code{free} is used.
4392:
4393: @findex USE_C_ALLOCA
4394: @item USE_C_ALLOCA
4395: Define this macro to indicate that the compiler is running with the
4396: @code{alloca} implemented in C. This version of @code{alloca} can be
4397: found in the file @file{alloca.c}; to use it, you must also alter the
4398: @file{Makefile} variable @code{ALLOCA}. (This is done automatically
4399: for the systems on which we know it is needed.)
4400:
4401: If you do define this macro, you should probably do it as follows:
4402:
4403: @example
4404: #ifndef __GNUC__
4405: #define USE_C_ALLOCA
4406: #else
4407: #define alloca __builtin_alloca
4408: #endif
4409: @end example
4410:
4411: @noindent
4412: so that when the compiler is compiled with GNU CC it uses the more
4413: efficient built-in @code{alloca} function.
4414:
4415: @item FUNCTION_CONVERSION_BUG
4416: @findex FUNCTION_CONVERSION_BUG
4417: Define this macro to indicate that the host compiler does not properly
4418: handle converting a function value to a pointer-to-function when it is
4419: used in an expression.
1.1.1.2 root 4420:
4421: @findex HAVE_VPRINTF
4422: @findex vprintf
4423: @item HAVE_VPRINTF
4424: Define this if the library function @code{vprintf} is available on your
4425: system.
4426:
1.1.1.3 root 4427: @findex MULTIBYTE_CHARS
4428: @item MULTIBYTE_CHARS
4429: Define this macro to enable support for multibyte characters in the
4430: input to GNU CC. This requires that the host system support the ANSI C
4431: library functions for converting multibyte characters to wide
4432: characters.
4433:
1.1.1.2 root 4434: @findex HAVE_PUTENV
4435: @findex putenv
4436: @item HAVE_PUTENV
4437: Define this if the library function @code{putenv} is available on your
4438: system.
4439:
4440: @findex NO_SYS_SIGLIST
4441: @item NO_SYS_SIGLIST
4442: Define this if your system @emph{does not} provide the variable
4443: @code{sys_siglist}.
4444:
1.1.1.7 ! root 4445: @findex DONT_DECLARE_SYS_SIGLIST
! 4446: @item DONT_DECLARE_SYS_SIGLIST
! 4447: Define this if your system has the variable @code{sys_siglist}, and
! 4448: there is already a declaration of it in the system header files.
! 4449:
1.1.1.4 root 4450: @findex USE_PROTOTYPES
4451: @item USE_PROTOTYPES
4452: Define this to be 1 if you know that the host compiler supports
4453: prototypes, even if it doesn't define __STDC__, or define
4454: it to be 0 if you do not want any prototypes used in compiling
4455: GNU CC. If @samp{USE_PROTOTYPES} is not defined, it will be
4456: determined automatically whether your compiler supports
4457: prototypes by checking if @samp{__STDC__} is defined.
4458:
4459: @findex NO_MD_PROTOTYPES
4460: @item NO_MD_PROTOTYPES
4461: Define this if you wish suppression of prototypes generated from
4462: the machine description file, but to use other prototypes within
4463: GNU CC. If @samp{USE_PROTOTYPES} is defined to be 0, or the
4464: host compiler does not support prototypes, this macro has no
4465: effect.
4466:
4467: @findex MD_CALL_PROTOTYPES
4468: @item MD_CALL_PROTOTYPES
4469: Define this if you wish to generate prototypes for the
4470: @code{gen_call} or @code{gen_call_value} functions generated from
4471: the machine description file. If @samp{USE_PROTOTYPES} is
4472: defined to be 0, or the host compiler does not support
4473: prototypes, or @samp{NO_MD_PROTOTYPES} is defined, this macro has
4474: no effect. As soon as all of the machine descriptions are
4475: modified to have the appropriate number of arguments, this macro
4476: will be removed.
4477:
1.1.1.2 root 4478: @vindex sys_siglist
4479: Some systems do provide this variable, but with a different name such
4480: as @code{_sys_siglist}. On these systems, you can define
4481: @code{sys_siglist} as a macro which expands into the name actually
4482: provided.
4483:
4484: @findex NO_STAB_H
4485: @item NO_STAB_H
4486: Define this if your system does not have the include file
4487: @file{stab.h}. If @samp{USG} is defined, @samp{NO_STAB_H} is
4488: assumed.
1.1.1.7 ! root 4489:
! 4490: @findex PATH_SEPARATOR
! 4491: @item PATH_SEPARATOR
! 4492: Define this macro to be a C character constant representing the
! 4493: character used to separate components in paths. The default value is.
! 4494: the colon character
! 4495:
! 4496: @findex DIR_SEPARATOR
! 4497: @item DIR_SEPARATOR
! 4498: If your system uses some character other than slash to separate
! 4499: directory names within a file specification, define this macro to be a C
! 4500: character constant specifying that character. When GNU CC displays file
! 4501: names, the character you specify will be used. GNU CC will test for
! 4502: both slash and the character you specify when parsing filenames.
1.1 root 4503: @end table
4504:
4505: @findex bzero
4506: @findex bcmp
4507: In addition, configuration files for system V define @code{bcopy},
4508: @code{bzero} and @code{bcmp} as aliases. Some files define @code{alloca}
4509: as a macro when compiled with GNU CC, in order to take advantage of the
4510: benefit of GNU CC's built-in @code{alloca}.
4511:
4512:
1.1.1.3 root 4513: @node Index
1.1 root 4514: @unnumbered Index
4515: @end ifset
4516:
4517: @ifclear INTERNALS
1.1.1.3 root 4518: @node Index
1.1 root 4519: @unnumbered Index
4520: @end ifclear
4521:
4522: @printindex cp
1.1.1.5 root 4523: @summarycontents
1.1 root 4524: @contents
4525: @bye
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.