![[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
1.1.1.8 root 39: @c some such.
1.1.1.5 root 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
1.1.1.8 root 99: 59 Temple Place - Suite 330
100: Boston, MA 02111-1307 USA
1.1.1.5 root 101:
1.1.1.8 root 102: Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995 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
1.1.1.9 ! root 150: @center Last updated 29 June 1996
1.1 root 151: @sp 1
1.1.1.5 root 152: @c The version number appears twice more in this file.
153:
1.1.1.9 ! root 154: @center for version 2.7.2.1
1.1 root 155: @page
156: @vskip 0pt plus 1filll
1.1.1.8 root 157: Copyright @copyright{} 1988, 89, 92, 93, 94, 1995 Free Software Foundation, Inc.
1.1.1.5 root 158: @sp 2
1.1.1.8 root 159: For GCC Version 2.7.2@*
1.1.1.5 root 160: @sp 1
161: Published by the Free Software Foundation @*
1.1.1.8 root 162: 59 Temple Place - Suite 330@*
163: Boston, MA 02111-1307, USA@*
164: Last printed November, 1995.@*
165: Printed copies are available for $50 each.@*
166: ISBN 1-882114-66-3
1.1.1.5 root 167: @sp 1
1.1 root 168: Permission is granted to make and distribute verbatim copies of
169: this manual provided the copyright notice and this permission notice
170: are preserved on all copies.
171:
172: Permission is granted to copy and distribute modified versions of this
173: manual under the conditions for verbatim copying, provided also that the
1.1.1.7 root 174: sections entitled ``GNU General Public License,'' ``Funding for Free
175: Software,'' and ``Protect Your Freedom---Fight `Look And Feel'@w{}'' are
176: included exactly as in the original, and provided that the entire
177: resulting derived work is distributed under the terms of a permission
178: notice identical to this one.
1.1 root 179:
180: Permission is granted to copy and distribute translations of this manual
181: into another language, under the above conditions for modified versions,
1.1.1.7 root 182: except that the sections entitled ``GNU General Public License,''
183: ``Funding for Free Software,'' and ``Protect Your Freedom---Fight `Look
184: And Feel'@w{}'', and this permission notice, may be included in
185: translations approved by the Free Software Foundation instead of in the
186: original English.
1.1 root 187: @end titlepage
188: @page
189:
190: @ifinfo
191:
192: @node Top, Copying,, (DIR)
1.1.1.3 root 193: @top Introduction
1.1 root 194: @cindex introduction
195:
196: @ifset INTERNALS
1.1.1.5 root 197: @ifset USING
198: This manual documents how to run, install and port the GNU
199: compiler, as well as its new features and incompatibilities, and how to
1.1.1.8 root 200: report bugs. It corresponds to GNU CC version 2.7.2.
1.1.1.5 root 201: @end ifset
1.1 root 202: @end ifset
203:
204: @ifclear INTERNALS
1.1.1.5 root 205: This manual documents how to run and install the GNU compiler,
206: as well as its new features and incompatibilities, and how to report
1.1.1.8 root 207: bugs. It corresponds to GNU CC version 2.7.2.
1.1.1.5 root 208: @end ifclear
209: @ifclear USING
210: This manual documents how to port the GNU compiler,
211: as well as its new features and incompatibilities, and how to report
1.1.1.8 root 212: bugs. It corresponds to GNU CC version 2.7.1.
1.1 root 213: @end ifclear
214:
215: @end ifinfo
216: @menu
217: * Copying:: GNU General Public License says
218: how you can copy and share GNU CC.
219: * Contributors:: People who have contributed to GNU CC.
1.1.1.7 root 220: * Funding:: How to help assure funding for free software.
221: * Look and Feel:: Protect your freedom---fight ``look and feel''.
1.1.1.5 root 222: @ifset USING
223: * G++ and GCC:: You can compile C or C++ programs.
1.1.1.3 root 224: * Invoking GCC:: Command options supported by @samp{gcc}.
1.1 root 225: * Installation:: How to configure, compile and install GNU CC.
1.1.1.5 root 226: * C Extensions:: GNU extensions to the C language family.
227: * C++ Extensions:: GNU extensions to the C++ language.
1.1 root 228: * Trouble:: If you have trouble installing GNU CC.
1.1.1.3 root 229: * Bugs:: How, why and where to report bugs.
1.1 root 230: * Service:: How to find suppliers of support for GNU CC.
231: * VMS:: Using GNU CC on VMS.
1.1.1.5 root 232: @end ifset
233: @ifset INTERNALS
1.1 root 234: * Portability:: Goals of GNU CC's portability features.
235: * Interface:: Function-call interface of GNU CC output.
236: * Passes:: Order of passes, what they do, and what each file is for.
237: * RTL:: The intermediate representation that most passes work on.
238: * Machine Desc:: How to write machine description instruction patterns.
1.1.1.2 root 239: * Target Macros:: How to write the machine description C macros.
1.1 root 240: * Config:: Writing the @file{xm-@var{machine}.h} file.
1.1.1.8 root 241: * Fragments:: Writing the @file{t-@var{target}} and @file{x-@var{host}} files.
1.1 root 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.
1.1.1.8 root 253: 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
1.1 root 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
1.1.1.8 root 599: Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
1.1 root 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
1.1.1.8 root 696: David Kashtan of SRI adapted GNU CC to VMS.
1.1 root 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
1.1.1.8 root 774: Torbjorn Granlund implemented multiply- and divide-by-constant
775: optimization, improved long long support, and improved leaf function
776: 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.
1.1.1.8 root 817:
818: @item
819: Michael K. Gschwind contributed the port to the PDP-11.
1.1 root 820: @end itemize
821:
1.1.1.7 root 822: @node Funding
823: @chapter Funding Free Software
824:
825: If you want to have more free software a few years from now, it makes
826: sense for you to help encourage people to contribute funds for its
827: development. The most effective approach known is to encourage
828: commercial redistributors to donate.
829:
830: Users of free software systems can boost the pace of development by
831: encouraging for-a-fee distributors to donate part of their selling price
832: to free software developers---the Free Software Foundation, and others.
833:
834: The way to convince distributors to do this is to demand it and expect
835: it from them. So when you compare distributors, judge them partly by
836: how much they give to free software development. Show distributors
837: they must compete to be the one who gives the most.
838:
839: To make this approach work, you must insist on numbers that you can
840: compare, such as, ``We will donate ten dollars to the Frobnitz project
841: for each disk sold.'' Don't be satisfied with a vague promise, such as
842: ``A portion of the profits are donated,'' since it doesn't give a basis
843: for comparison.
844:
845: Even a precise fraction ``of the profits from this disk'' is not very
846: meaningful, since creative accounting and unrelated business decisions
847: can greatly alter what fraction of the sales price counts as profit.
848: If the price you pay is $50, ten percent of the profit is probably
849: less than a dollar; it might be a few cents, or nothing at all.
850:
851: Some redistributors do development work themselves. This is useful too;
852: but to keep everyone honest, you need to inquire how much they do, and
853: what kind. Some kinds of development make much more long-term
854: difference than others. For example, maintaining a separate version of
855: a program contributes very little; maintaining the standard version of a
856: program for the whole community contributes much. Easy new ports
857: contribute little, since someone else would surely do them; difficult
858: ports such as adding a new CPU to the GNU C compiler contribute more;
859: major new features or packages contribute the most.
860:
861: By establishing the idea that supporting further development is ``the
862: proper thing to do'' when distributing free software for a fee, we can
863: assure a steady flow of resources into making more free software.
864:
865: @display
866: Copyright (C) 1994 Free Software Foundation, Inc.
867: Verbatim copying and redistribution of this section is permitted
868: without royalty; alteration is not permitted.
869: @end display
870:
871: @node Look and Feel
1.1 root 872: @chapter Protect Your Freedom---Fight ``Look And Feel''
1.1.1.5 root 873: @c the above chapter heading overflows onto the next line. --mew 1/26/93
1.1 root 874:
875: @quotation
876: @i{This section is a political message from the League for Programming
1.1.1.7 root 877: Freedom to the users of GNU CC. We have included it here because the
878: issue of interface copyright is important to the GNU project.}
1.1 root 879: @end quotation
880:
1.1.1.8 root 881: Apple, Lotus, and now CDC have tried to create a new form of legal
882: monopoly: a copyright on a user interface.
1.1.1.7 root 883:
884: An interface is a kind of language---a set of conventions for
885: communication between two entities, human or machine. Until a few years
886: ago, the law seemed clear: interfaces were outside the domain of
887: copyright, so programmers could program freely and implement whatever
888: interface the users demanded. Imitating de-facto standard interfaces,
889: sometimes with improvements, was standard practice in the computer
890: field. These improvements, if accepted by the users, caught on and
891: became the norm; in this way, much progress took place.
892:
893: Computer users, and most software developers, were happy with this state
894: of affairs. However, large companies such as Apple and Lotus would
895: prefer a different system---one in which they can own interfaces and
896: thereby rid themselves of all serious competitors. They hope that
897: interface copyright will give them, in effect, monopolies on major
898: classes of software.
899:
900: Other large companies such as IBM and Digital also favor interface
901: monopolies, for the same reason: if languages become property, they
902: expect to own many de-facto standard languages. But Apple and Lotus are
1.1.1.8 root 903: the ones who have actually sued. Apple's lawsuit was defeated, for
904: reasons only partly related to the general issue of interface copyright.
905:
906: Lotus won lawsuits against two small companies, which were thus put out
1.1.1.9 ! root 907: of business. Then Lotus sued Borland; Lotus won in the trial court (no
1.1.1.8 root 908: surprise, since it was the same court that had ruled for Lotus twice
1.1.1.9 ! root 909: before), but the court of appeals ruled in favor of Borland, which was
! 910: assisted by a friend-of-the-court brief from the League for Programming
! 911: Freedom.
! 912:
! 913: Lotus appealed the case to the Supreme Court, which heard the case but
! 914: was unable to reach a decision. This failure means that the appeals
! 915: court decision stands, in one portion of the United States, and may
! 916: influence the other appeals courts, but it does not set a nationwide
! 917: precedent. The battle is not over, and it is not limited to the United
! 918: States.
! 919:
! 920: The battle is extending into other areas of software as well. In 1995 a
! 921: company that produced a simulator for a CDC computer was shut down by a
! 922: copyright lawsuit, in which CDC charged that the simulator infringed the
! 923: copyright on the manuals for the computer.
1.1 root 924:
1.1.1.7 root 925: If the monopolists get their way, they will hobble the software field:
1.1 root 926:
927: @itemize @bullet
928: @item
1.1.1.7 root 929: Gratuitous incompatibilities will burden users. Imagine if each car
930: manufacturer had to design a different way to start, stop, and steer a
931: car.
1.1 root 932:
933: @item
1.1.1.7 root 934: Users will be ``locked in'' to whichever interface they learn; then they
935: will be prisoners of one supplier, who will charge a monopolistic price.
1.1 root 936:
937: @item
938: Large companies have an unfair advantage wherever lawsuits become
1.1.1.7 root 939: commonplace. Since they can afford to sue, they can intimidate smaller
940: developers with threats even when they don't really have a case.
1.1 root 941:
942: @item
1.1.1.7 root 943: Interface improvements will come slower, since incremental evolution
944: through creative partial imitation will no longer occur.
945: @end itemize
946:
947: If interface monopolies are accepted, other large companies are waiting
948: to grab theirs:
1.1 root 949:
1.1.1.7 root 950: @itemize @bullet
1.1 root 951: @item
1.1.1.8 root 952: Adobe is expected to claim a monopoly on the interfaces of various
953: popular application programs, if Lotus ultimately wins the case against
954: Borland.
1.1 root 955:
956: @item
1.1.1.7 root 957: Open Computing magazine reported a Microsoft vice president as threatening
1.1.1.8 root 958: to sue people who imitate the interface of Windows.
1.1 root 959: @end itemize
960:
1.1.1.7 root 961: Users invest a great deal of time and money in learning to use computer
962: interfaces. Far more, in fact, than software developers invest in
963: developing @emph{and even implementing} the interfaces. Whoever can own
964: an interface, has made its users into captives, and misappropriated
965: their investment.
966:
967: To protect our freedom from monopolies like these, a group of
968: programmers and users have formed a grass-roots political organization,
969: the League for Programming Freedom.
970:
971: The purpose of the League is to oppose monopolistic practices such as
972: interface copyright and software patents. The League calls for a return
973: to the legal policies of the recent past, in which programmers could
974: program freely. The League is not concerned with free software as an
975: issue, and is not affiliated with the Free Software Foundation.
976:
1.1.1.8 root 977: The League's activities include publicizing the issues, as is being done
1.1.1.7 root 978: here, and filing friend-of-the-court briefs on behalf of defendants sued
1.1.1.8 root 979: by monopolists.
1.1 root 980:
1.1.1.8 root 981: The League's membership rolls include Donald Knuth, the foremost
982: authority on algorithms, John McCarthy, inventor of Lisp, Marvin Minsky,
983: founder of the MIT Artificial Intelligence lab, Guy L. Steele, Jr.,
984: author of well-known books on Lisp and C, as well as Richard Stallman,
985: the developer of GNU CC. Please join and add your name to the list.
986: Membership dues in the League are $42 per year for programmers, managers
987: and professionals; $10.50 for students; $21 for others.
1.1 root 988:
1.1.1.7 root 989: Activist members are especially important, but members who have no time
990: to give are also important. Surveys at major ACM conferences have
1.1.1.8 root 991: indicated a vast majority of attendees agree with the League on both
992: issues (interface copyrights and software patents). If just ten percent
993: of the programmers who agree with the League join the League, we will
994: probably triumph.
1.1 root 995:
1.1.1.4 root 996: To join, or for more information, phone (617) 243-4091 or write to:
1.1 root 997:
998: @display
999: League for Programming Freedom
1000: 1 Kendall Square #143
1001: P.O. Box 9171
1002: Cambridge, MA 02139
1003: @end display
1004:
1.1.1.7 root 1005: You can also send electronic mail to @code{lpf@@uunet.uu.net}.
1.1 root 1006:
1.1.1.7 root 1007: In addition to joining the League, here are some suggestions from the
1008: League for other things you can do to protect your freedom to write
1009: programs:
1.1 root 1010:
1011: @itemize @bullet
1012: @item
1.1.1.7 root 1013: Tell your friends and colleagues about this issue and how it threatens
1014: to ruin the computer industry.
1.1 root 1015:
1016: @item
1.1.1.7 root 1017: Mention that you are a League member in your @file{.signature}, and
1018: mention the League's email address for inquiries.
1.1 root 1019:
1020: @item
1.1.1.7 root 1021: Ask the companies you consider working for or working with to make
1022: statements against software monopolies, and give preference to those
1023: that do.
1.1 root 1024:
1025: @item
1.1.1.8 root 1026: When employers ask you to sign contracts giving them copyright on your
1027: work, insist on a clause saying they will not claim the copyright covers
1028: imitating the interface.
1.1 root 1029:
1030: @item
1.1.1.8 root 1031: When employers ask you to sign contracts giving them patent rights,
1032: insist on clauses saying they can use these rights only defensively.
1033: Don't rely on ``company policy,'' since that can change at any time;
1034: don't rely on an individual executive's private word, since that person
1035: may be replaced. Get a commitment just as binding as the commitment
1036: they get from you.
1037:
1038: @item
1039: Write to Congress to explain the importance of these issues.
1.1 root 1040:
1041: @display
1042: House Subcommittee on Intellectual Property
1043: 2137 Rayburn Bldg
1044: Washington, DC 20515
1045:
1046: Senate Subcommittee on Patents, Trademarks and Copyrights
1047: United States Senate
1048: Washington, DC 20510
1049: @end display
1050:
1051: (These committees have received lots of mail already; let's give them
1052: even more.)
1053: @end itemize
1054:
1.1.1.7 root 1055: Democracy means nothing if you don't use it. Stand up and be counted!
1.1.1.5 root 1056: @ifset USING
1057: @node G++ and GCC
1058: @chapter Compile C, C++, or Objective C
1059:
1060: @cindex Objective C
1061: The C, C++, and Objective C versions of the compiler are integrated; the
1062: GNU C compiler can compile programs written in C, C++, or Objective C.
1063:
1064: @cindex GCC
1065: ``GCC'' is a common shorthand term for the GNU C compiler. This is both
1066: the most general name for the compiler, and the name used when the
1067: emphasis is on compiling C programs.
1068:
1069: @cindex C++
1070: @cindex G++
1071: When referring to C++ compilation, it is usual to call the compiler
1072: ``G++''. Since there is only one compiler, it is also accurate to call
1073: it ``GCC'' no matter what the language context; however, the term
1074: ``G++'' is more useful when the emphasis is on compiling C++ programs.
1075:
1.1.1.7 root 1076: We use the name ``GNU CC'' to refer to the compilation system as a
1077: whole, and more specifically to the language-independent part of the
1078: compiler. For example, we refer to the optimization options as
1079: affecting the behavior of ``GNU CC'' or sometimes just ``the compiler''.
1080:
1081: Front ends for other languages, such as Ada 9X, Fortran, Modula-3, and
1082: Pascal, are under development. These front-ends, like that for C++, are
1083: built in subdirectories of GNU CC and link to it. The result is an
1084: integrated compiler that can compile programs written in C, C++,
1085: Objective C, or any of the languages for which you have installed front
1086: ends.
1087:
1088: In this manual, we only discuss the options for the C, Objective-C, and
1089: C++ compilers and those of the GNU CC core. Consult the documentation
1090: of the other front ends for the options to use when compiling programs
1091: written in other languages.
1092:
1.1.1.5 root 1093: @cindex compiler compared to C++ preprocessor
1094: @cindex intermediate C version, nonexistent
1095: @cindex C intermediate output, nonexistent
1096: G++ is a @emph{compiler}, not merely a preprocessor. G++ builds object
1097: code directly from your C++ program source. There is no intermediate C
1098: version of the program. (By contrast, for example, some other
1099: implementations use a program that generates a C program from your C++
1100: source.) Avoiding an intermediate C representation of the program means
1101: that you get better object code, and better debugging information. The
1102: GNU debugger, GDB, works with this information in the object code to
1103: give you comprehensive C++ source-level editing capabilities
1104: (@pxref{C,,C and C++,gdb.info, Debugging with GDB}).
1105:
1106: @c FIXME! Someone who knows something about Objective C ought to put in
1107: @c a paragraph or two about it here, and move the index entry down when
1108: @c there is more to point to than the general mention in the 1st par.
1109:
1.1 root 1110: @include invoke.texi
1111:
1.1.1.3 root 1112: @include install.texi
1113:
1114: @include extend.texi
1115:
1116: @node Trouble
1117: @chapter Known Causes of Trouble with GNU CC
1118: @cindex bugs, known
1119: @cindex installation trouble
1120: @cindex known causes of trouble
1.1 root 1121:
1.1.1.3 root 1122: This section describes known problems that affect users of GNU CC. Most
1123: of these are not GNU CC bugs per se---if they were, we would fix them.
1124: But the result for a user may be like the result of a bug.
1125:
1126: Some of these problems are due to bugs in other software, some are
1127: missing features that are too much work to add, and some are places
1128: where people's opinions differ as to what is best.
1.1 root 1129:
1130: @menu
1.1.1.3 root 1131: * Actual Bugs:: Bugs we will fix later.
1132: * Installation Problems:: Problems that manifest when you install GNU CC.
1133: * Cross-Compiler Problems:: Common problems of cross compiling with GNU CC.
1134: * Interoperation:: Problems using GNU CC with other compilers,
1135: and with certain linkers, assemblers and debuggers.
1.1.1.5 root 1136: * External Bugs:: Problems compiling certain programs.
1.1.1.3 root 1137: * Incompatibilities:: GNU CC is incompatible with traditional C.
1.1.1.6 root 1138: * Fixed Headers:: GNU C uses corrected versions of system header files.
1139: This is necessary, but doesn't always work smoothly.
1.1.1.8 root 1140: * Standard Libraries:: GNU C uses the system C library, which might not be
1141: compliant with the ISO/ANSI C standard.
1.1.1.3 root 1142: * Disappointments:: Regrettable things we can't change, but not quite bugs.
1.1.1.5 root 1143: * C++ Misunderstandings:: Common misunderstandings with GNU C++.
1.1.1.4 root 1144: * Protoize Caveats:: Things to watch out for when using @code{protoize}.
1.1.1.3 root 1145: * Non-bugs:: Things we think are right, but some others disagree.
1.1.1.5 root 1146: * Warnings and Errors:: Which problems in your code get warnings,
1147: and which get errors.
1.1 root 1148: @end menu
1149:
1.1.1.3 root 1150: @node Actual Bugs
1151: @section Actual Bugs We Haven't Fixed Yet
1.1 root 1152:
1.1.1.3 root 1153: @itemize @bullet
1.1 root 1154: @item
1.1.1.5 root 1155: The @code{fixincludes} script interacts badly with automounters; if the
1156: directory of system header files is automounted, it tends to be
1157: unmounted while @code{fixincludes} is running. This would seem to be a
1158: bug in the automounter. We don't know any good way to work around it.
1159:
1160: @item
1.1.1.6 root 1161: The @code{fixproto} script will sometimes add prototypes for the
1162: @code{sigsetjmp} and @code{siglongjmp} functions that reference the
1163: @code{jmp_buf} type before that type is defined. To work around this,
1164: edit the offending file and place the typedef in front of the
1165: prototypes.
1166:
1167: @item
1.1.1.7 root 1168: There are several obscure case of mis-using struct, union, and
1169: enum tags that are not detected as errors by the compiler.
1170:
1171: @item
1172: When @samp{-pedantic-errors} is specified, GNU C will incorrectly give
1173: an error message when a function name is specified in an expression
1174: involving the comma operator.
1175:
1176: @item
1.1.1.4 root 1177: Loop unrolling doesn't work properly for certain C++ programs. This is
1.1.1.7 root 1178: a bug in the C++ front end. It sometimes emits incorrect debug info, and
1179: the loop unrolling code is unable to recover from this error.
1.1.1.3 root 1180: @end itemize
1.1 root 1181:
1.1.1.3 root 1182: @node Installation Problems
1183: @section Installation Problems
1.1.1.4 root 1184:
1.1.1.3 root 1185: This is a list of problems (and some apparent problems which don't
1186: really mean anything is wrong) that show up during installation of GNU
1187: CC.
1.1 root 1188:
1189: @itemize @bullet
1190: @item
1.1.1.3 root 1191: On certain systems, defining certain environment variables such as
1192: @code{CC} can interfere with the functioning of @code{make}.
1.1 root 1193:
1194: @item
1.1.1.3 root 1195: If you encounter seemingly strange errors when trying to build the
1196: compiler in a directory other than the source directory, it could be
1197: because you have previously configured the compiler in the source
1198: directory. Make sure you have done all the necessary preparations.
1199: @xref{Other Dir}.
1.1 root 1200:
1201: @item
1.1.1.6 root 1202: If you build GNU CC on a BSD system using a directory stored in a System
1203: V file system, problems may occur in running @code{fixincludes} if the
1204: System V file system doesn't support symbolic links. These problems
1205: result in a failure to fix the declaration of @code{size_t} in
1206: @file{sys/types.h}. If you find that @code{size_t} is a signed type and
1207: that type mismatches occur, this could be the cause.
1208:
1209: The solution is not to use such a directory for building GNU CC.
1210:
1211: @item
1.1.1.3 root 1212: In previous versions of GNU CC, the @code{gcc} driver program looked for
1.1.1.5 root 1213: @code{as} and @code{ld} in various places; for example, in files
1214: beginning with @file{/usr/local/lib/gcc-}. GNU CC version 2 looks for
1215: them in the directory
1216: @file{/usr/local/lib/gcc-lib/@var{target}/@var{version}}.
1.1 root 1217:
1.1.1.3 root 1218: Thus, to use a version of @code{as} or @code{ld} that is not the system
1219: default, for example @code{gas} or GNU @code{ld}, you must put them in
1220: that directory (or make links to them from that directory).
1.1 root 1221:
1222: @item
1.1.1.3 root 1223: Some commands executed when making the compiler may fail (return a
1224: non-zero status) and be ignored by @code{make}. These failures, which
1225: are often due to files that were not found, are expected, and can safely
1226: be ignored.
1.1 root 1227:
1228: @item
1.1.1.3 root 1229: It is normal to have warnings in compiling certain files about
1230: unreachable code and about enumeration type clashes. These files' names
1.1.1.5 root 1231: begin with @samp{insn-}. Also, @file{real.c} may get some warnings that
1232: you can ignore.
1.1 root 1233:
1234: @item
1.1.1.3 root 1235: Sometimes @code{make} recompiles parts of the compiler when installing
1236: the compiler. In one case, this was traced down to a bug in
1237: @code{make}. Either ignore the problem or switch to GNU Make.
1.1 root 1238:
1.1.1.4 root 1239: @item
1.1.1.5 root 1240: If you have installed a program known as purify, you may find that it
1241: causes errors while linking @code{enquire}, which is part of building
1242: GNU CC. The fix is to get rid of the file @code{real-ld} which purify
1243: installs---so that GNU CC won't try to use it.
1244:
1245: @item
1.1.1.9 ! root 1246: On SLS 1.01, a Linux-based GNU system, there is a problem with
! 1247: @file{libc.a}: it does not contain the obstack functions. However, GNU
! 1248: CC assumes that the obstack functions are in @file{libc.a} when it is
! 1249: the GNU C library. To work around this problem, change the
! 1250: @code{__GNU_LIBRARY__} conditional around line 31 to @samp{#if 1}.
1.1.1.5 root 1251:
1252: @item
1.1.1.4 root 1253: On some 386 systems, building the compiler never finishes because
1254: @code{enquire} hangs due to a hardware problem in the motherboard---it
1255: reports floating point exceptions to the kernel incorrectly. You can
1256: install GNU CC except for @file{float.h} by patching out the command to
1257: run @code{enquire}. You may also be able to fix the problem for real by
1258: getting a replacement motherboard. This problem was observed in
1259: Revision E of the Micronics motherboard, and is fixed in Revision F.
1.1.1.6 root 1260: It has also been observed in the MYLEX MXA-33 motherboard.
1261:
1262: If you encounter this problem, you may also want to consider removing
1263: the FPU from the socket during the compilation. Alternatively, if you
1264: are running SCO Unix, you can reboot and force the FPU to be ignored.
1265: To do this, type @samp{hd(40)unix auto ignorefpu}.
1.1.1.4 root 1266:
1267: @item
1268: On some 386 systems, GNU CC crashes trying to compile @file{enquire.c}.
1269: This happens on machines that don't have a 387 FPU chip. On 386
1270: machines, the system kernel is supposed to emulate the 387 when you
1271: don't have one. The crash is due to a bug in the emulator.
1272:
1273: One of these systems is the Unix from Interactive Systems: 386/ix.
1274: On this system, an alternate emulator is provided, and it does work.
1275: To use it, execute this command as super-user:
1276:
1277: @example
1278: ln /etc/emulator.rel1 /etc/emulator
1279: @end example
1280:
1281: @noindent
1282: and then reboot the system. (The default emulator file remains present
1283: under the name @file{emulator.dflt}.)
1284:
1.1.1.5 root 1285: Try using @file{/etc/emulator.att}, if you have such a problem on the
1286: SCO system.
1.1.1.4 root 1287:
1288: Another system which has this problem is Esix. We don't know whether it
1289: has an alternate emulator that works.
1290:
1.1.1.5 root 1291: On NetBSD 0.8, a similar problem manifests itself as these error messages:
1292:
1293: @example
1294: enquire.c: In function `fprop':
1295: enquire.c:2328: floating overflow
1296: @end example
1297:
1.1.1.6 root 1298: @item
1299: On SCO systems, when compiling GNU CC with the system's compiler,
1300: do not use @samp{-O}. Some versions of the system's compiler miscompile
1301: GNU CC with @samp{-O}.
1302:
1.1.1.3 root 1303: @cindex @code{genflags}, crash on Sun 4
1.1 root 1304: @item
1.1.1.3 root 1305: Sometimes on a Sun 4 you may observe a crash in the program
1.1.1.4 root 1306: @code{genflags} or @code{genoutput} while building GNU CC. This is said to
1.1.1.3 root 1307: be due to a bug in @code{sh}. You can probably get around it by running
1308: @code{genflags} or @code{genoutput} manually and then retrying the
1309: @code{make}.
1.1 root 1310:
1.1.1.3 root 1311: @item
1.1.1.5 root 1312: On Solaris 2, executables of GNU CC version 2.0.2 are commonly
1313: available, but they have a bug that shows up when compiling current
1314: versions of GNU CC: undefined symbol errors occur during assembly if you
1315: use @samp{-g}.
1316:
1317: The solution is to compile the current version of GNU CC without
1318: @samp{-g}. That makes a working compiler which you can use to recompile
1319: with @samp{-g}.
1320:
1321: @item
1.1.1.6 root 1322: Solaris 2 comes with a number of optional OS packages. Some of these
1.1.1.5 root 1323: packages are needed to use GNU CC fully. If you did not install all
1324: optional packages when installing Solaris, you will need to verify that
1.1.1.6 root 1325: the packages that GNU CC needs are installed.
1.1.1.5 root 1326:
1.1.1.6 root 1327: To check whether an optional package is installed, use
1.1.1.5 root 1328: the @code{pkginfo} command. To add an optional package, use the
1329: @code{pkgadd} command. For further details, see the Solaris
1330: documentation.
1331:
1.1.1.6 root 1332: For Solaris 2.0 and 2.1, GNU CC needs six packages: @samp{SUNWarc},
1333: @samp{SUNWbtool}, @samp{SUNWesu}, @samp{SUNWhea}, @samp{SUNWlibm}, and
1334: @samp{SUNWtoo}.
1335:
1336: For Solaris 2.2, GNU CC needs an additional seventh package: @samp{SUNWsprot}.
1337:
1.1.1.5 root 1338: @item
1339: On Solaris 2, trying to use the linker and other tools in
1340: @file{/usr/ucb} to install GNU CC has been observed to cause trouble.
1341: For example, the linker may hang indefinitely. The fix is to remove
1342: @file{/usr/ucb} from your @code{PATH}.
1343:
1344: @item
1.1.1.3 root 1345: If you use the 1.31 version of the MIPS assembler (such as was shipped
1346: with Ultrix 3.1), you will need to use the -fno-delayed-branch switch
1347: when optimizing floating point code. Otherwise, the assembler will
1348: complain when the GCC compiler fills a branch delay slot with a
1.1.1.6 root 1349: floating point instruction, such as @code{add.d}.
1.1.1.3 root 1350:
1351: @item
1.1.1.5 root 1352: If on a MIPS system you get an error message saying ``does not have gp
1353: sections for all it's [sic] sectons [sic]'', don't worry about it. This
1354: happens whenever you use GAS with the MIPS linker, but there is not
1355: really anything wrong, and it is okay to use the output file. You can
1356: stop such warnings by installing the GNU linker.
1357:
1358: It would be nice to extend GAS to produce the gp tables, but they are
1359: optional, and there should not be a warning about their absence.
1360:
1361: @item
1.1.1.6 root 1362: In Ultrix 4.0 on the MIPS machine, @file{stdio.h} does not work with GNU
1363: CC at all unless it has been fixed with @code{fixincludes}. This causes
1364: problems in building GNU CC. Once GNU CC is installed, the problems go
1365: away.
1366:
1367: To work around this problem, when making the stage 1 compiler, specify
1368: this option to Make:
1369:
1370: @example
1371: GCC_FOR_TARGET="./xgcc -B./ -I./include"
1372: @end example
1373:
1374: When making stage 2 and stage 3, specify this option:
1375:
1376: @example
1377: CFLAGS="-g -I./include"
1378: @end example
1379:
1380: @item
1.1.1.3 root 1381: Users have reported some problems with version 2.0 of the MIPS
1382: compiler tools that were shipped with Ultrix 4.1. Version 2.10
1383: which came with Ultrix 4.2 seems to work fine.
1.1 root 1384:
1.1.1.7 root 1385: Users have also reported some problems with version 2.20 of the
1386: MIPS compiler tools that were shipped with RISC/os 4.x. The earlier
1387: version 2.11 seems to work fine.
1388:
1.1 root 1389: @item
1.1.1.4 root 1390: Some versions of the MIPS linker will issue an assertion failure
1391: when linking code that uses @code{alloca} against shared
1392: libraries on RISC-OS 5.0, and DEC's OSF/1 systems. This is a bug
1393: in the linker, that is supposed to be fixed in future revisions.
1.1.1.6 root 1394: To protect against this, GNU CC passes @samp{-non_shared} to the
1.1.1.4 root 1395: linker unless you pass an explicit @samp{-shared} or
1396: @samp{-call_shared} switch.
1397:
1398: @item
1399: On System V release 3, you may get this error message
1400: while linking:
1401:
1402: @smallexample
1403: ld fatal: failed to write symbol name @var{something}
1404: in strings table for file @var{whatever}
1405: @end smallexample
1406:
1.1.1.5 root 1407: This probably indicates that the disk is full or your ULIMIT won't allow
1.1.1.4 root 1408: the file to be as large as it needs to be.
1409:
1.1.1.5 root 1410: This problem can also result because the kernel parameter @code{MAXUMEM}
1411: is too small. If so, you must regenerate the kernel and make the value
1412: much larger. The default value is reported to be 1024; a value of 32768
1413: is said to work. Smaller values may also work.
1414:
1415: @item
1416: On System V, if you get an error like this,
1417:
1418: @example
1419: /usr/local/lib/bison.simple: In function `yyparse':
1420: /usr/local/lib/bison.simple:625: virtual memory exhausted
1421: @end example
1422:
1423: @noindent
1424: that too indicates a problem with disk space, ULIMIT, or @code{MAXUMEM}.
1425:
1426: @item
1427: Current GNU CC versions probably do not work on version 2 of the NeXT
1428: operating system.
1429:
1430: @item
1.1.1.6 root 1431: On NeXTStep 3.0, the Objective C compiler does not work, due,
1432: apparently, to a kernel bug that it happens to trigger. This problem
1433: does not happen on 3.1.
1434:
1435: @item
1.1.1.5 root 1436: On the Tower models 4@var{n}0 and 6@var{n}0, by default a process is not
1437: allowed to have more than one megabyte of memory. GNU CC cannot compile
1438: itself (or many other programs) with @samp{-O} in that much memory.
1439:
1440: To solve this problem, reconfigure the kernel adding the following line
1441: to the configuration file:
1442:
1443: @smallexample
1444: MAXUMEM = 4096
1445: @end smallexample
1446:
1.1.1.4 root 1447: @item
1.1.1.3 root 1448: On HP 9000 series 300 or 400 running HP-UX release 8.0, there is a bug
1449: in the assembler that must be fixed before GNU CC can be built. This
1450: bug manifests itself during the first stage of compilation, while
1451: building @file{libgcc2.a}:
1.1 root 1452:
1.1.1.5 root 1453: @smallexample
1.1.1.3 root 1454: _floatdisf
1455: cc1: warning: `-g' option not supported on this version of GCC
1456: cc1: warning: `-g1' option not supported on this version of GCC
1.1.1.5 root 1457: ./xgcc: Internal compiler error: program as got fatal signal 11
1458: @end smallexample
1.1 root 1459:
1.1.1.3 root 1460: A patched version of the assembler is available by anonymous ftp from
1461: @code{altdorf.ai.mit.edu} as the file
1462: @file{archive/cph/hpux-8.0-assembler}. If you have HP software support,
1463: the patch can also be obtained directly from HP, as described in the
1464: following note:
1.1 root 1465:
1.1.1.3 root 1466: @quotation
1467: This is the patched assembler, to patch SR#1653-010439, where the
1468: assembler aborts on floating point constants.
1.1 root 1469:
1.1.1.3 root 1470: The bug is not really in the assembler, but in the shared library
1471: version of the function ``cvtnum(3c)''. The bug on ``cvtnum(3c)'' is
1472: SR#4701-078451. Anyway, the attached assembler uses the archive
1473: library version of ``cvtnum(3c)'' and thus does not exhibit the bug.
1474: @end quotation
1.1 root 1475:
1.1.1.7 root 1476: This patch is also known as PHCO_4484.
1.1 root 1477:
1.1.1.3 root 1478: @item
1.1.1.6 root 1479: On HP-UX version 8.05, but not on 8.07 or more recent versions,
1480: the @code{fixproto} shell script triggers a bug in the system shell.
1481: If you encounter this problem, upgrade your operating system or
1482: use BASH (the GNU shell) to run @code{fixproto}.
1.1 root 1483:
1.1.1.3 root 1484: @item
1485: Some versions of the Pyramid C compiler are reported to be unable to
1486: compile GNU CC. You must use an older version of GNU CC for
1487: bootstrapping. One indication of this problem is if you get a crash
1488: when GNU CC compiles the function @code{muldi3} in file @file{libgcc2.c}.
1.1 root 1489:
1.1.1.3 root 1490: You may be able to succeed by getting GNU CC version 1, installing it,
1491: and using it to compile GNU CC version 2. The bug in the Pyramid C
1492: compiler does not seem to affect GNU CC version 1.
1.1 root 1493:
1.1.1.3 root 1494: @item
1.1.1.5 root 1495: There may be similar problems on System V Release 3.1 on 386 systems.
1.1 root 1496:
1497: @item
1.1.1.6 root 1498: On the Intel Paragon (an i860 machine), if you are using operating
1499: system version 1.0, you will get warnings or errors about redefinition
1500: of @code{va_arg} when you build GNU CC.
1501:
1502: If this happens, then you need to link most programs with the library
1503: @file{iclib.a}. You must also modify @file{stdio.h} as follows: before
1504: the lines
1505:
1506: @example
1507: #if defined(__i860__) && !defined(_VA_LIST)
1508: #include <va_list.h>
1509: @end example
1510:
1511: @noindent
1512: insert the line
1513:
1514: @example
1515: #if __PGC__
1516: @end example
1517:
1518: @noindent
1519: and after the lines
1520:
1521: @example
1522: extern int vprintf(const char *, va_list );
1523: extern int vsprintf(char *, const char *, va_list );
1524: #endif
1525: @end example
1526:
1527: @noindent
1528: insert the line
1529:
1530: @example
1531: #endif /* __PGC__ */
1532: @end example
1533:
1534: These problems don't exist in operating system version 1.1.
1535:
1536: @item
1.1.1.3 root 1537: On the Altos 3068, programs compiled with GNU CC won't work unless you
1538: fix a kernel bug. This happens using system versions V.2.2 1.0gT1 and
1539: V.2.2 1.0e and perhaps later versions as well. See the file
1540: @file{README.ALTOS}.
1.1.1.4 root 1541:
1542: @item
1543: You will get several sorts of compilation and linking errors on the
1.1.1.7 root 1544: we32k if you don't follow the special instructions. @xref{Configurations}.
1545:
1546: @item
1547: A bug in the HP-UX 8.05 (and earlier) shell will cause the fixproto
1548: program to report an error of the form:
1549:
1550: @example
1551: ./fixproto: sh internal 1K buffer overflow
1552: @end example
1553:
1554: To fix this, change the first line of the fixproto script to look like:
1555:
1556: @example
1557: #!/bin/ksh
1558: @end example
1.1.1.3 root 1559: @end itemize
1.1 root 1560:
1.1.1.3 root 1561: @node Cross-Compiler Problems
1562: @section Cross-Compiler Problems
1.1 root 1563:
1.1.1.5 root 1564: You may run into problems with cross compilation on certain machines,
1565: for several reasons.
1566:
1.1.1.3 root 1567: @itemize @bullet
1568: @item
1569: Cross compilation can run into trouble for certain machines because
1570: some target machines' assemblers require floating point numbers to be
1571: written as @emph{integer} constants in certain contexts.
1.1 root 1572:
1.1.1.3 root 1573: The compiler writes these integer constants by examining the floating
1574: point value as an integer and printing that integer, because this is
1575: simple to write and independent of the details of the floating point
1576: representation. But this does not work if the compiler is running on
1577: a different machine with an incompatible floating point format, or
1578: even a different byte-ordering.
1.1 root 1579:
1.1.1.3 root 1580: In addition, correct constant folding of floating point values
1581: requires representing them in the target machine's format.
1582: (The C standard does not quite require this, but in practice
1583: it is the only way to win.)
1.1.1.2 root 1584:
1.1.1.3 root 1585: It is now possible to overcome these problems by defining macros such
1586: as @code{REAL_VALUE_TYPE}. But doing so is a substantial amount of
1.1.1.5 root 1587: work for each target machine.
1588: @ifset INTERNALS
1589: @xref{Cross-compilation}.
1.1.1.3 root 1590: @end ifset
1591: @ifclear INTERNALS
1.1.1.5 root 1592: @xref{Cross-compilation,,Cross Compilation and Floating Point Format,
1593: gcc.info, Using and Porting GCC}.
1.1.1.3 root 1594: @end ifclear
1.1 root 1595:
1596: @item
1.1.1.3 root 1597: At present, the program @file{mips-tfile} which adds debug
1598: support to object files on MIPS systems does not work in a cross
1599: compile environment.
1600: @end itemize
1.1 root 1601:
1.1.1.3 root 1602: @node Interoperation
1603: @section Interoperation
1.1 root 1604:
1.1.1.3 root 1605: This section lists various difficulties encountered in using GNU C or
1.1.1.4 root 1606: GNU C++ together with other compilers or with the assemblers, linkers,
1607: libraries and debuggers on certain systems.
1.1 root 1608:
1.1.1.3 root 1609: @itemize @bullet
1610: @item
1.1.1.7 root 1611: Objective C does not work on the RS/6000.
1.1 root 1612:
1613: @item
1.1.1.3 root 1614: GNU C++ does not do name mangling in the same way as other C++
1615: compilers. This means that object files compiled with one compiler
1616: cannot be used with another.
1.1 root 1617:
1.1.1.5 root 1618: This effect is intentional, to protect you from more subtle problems.
1619: Compilers differ as to many internal details of C++ implementation,
1620: including: how class instances are laid out, how multiple inheritance is
1621: implemented, and how virtual function calls are handled. If the name
1622: encoding were made the same, your programs would link against libraries
1623: provided from other compilers---but the programs would then crash when
1624: run. Incompatible libraries are then detected at link time, rather than
1625: at run time.
1.1 root 1626:
1.1.1.3 root 1627: @item
1628: Older GDB versions sometimes fail to read the output of GNU CC version
1629: 2. If you have trouble, get GDB version 4.4 or later.
1.1 root 1630:
1631: @item
1.1.1.3 root 1632: @cindex DBX
1633: DBX rejects some files produced by GNU CC, though it accepts similar
1634: constructs in output from PCC. Until someone can supply a coherent
1635: description of what is valid DBX input and what is not, there is
1636: nothing I can do about these problems. You are on your own.
1.1 root 1637:
1.1.1.3 root 1638: @item
1639: The GNU assembler (GAS) does not support PIC. To generate PIC code, you
1640: must use some other assembler, such as @file{/bin/as}.
1.1 root 1641:
1.1.1.3 root 1642: @item
1.1.1.6 root 1643: On some BSD systems, including some versions of Ultrix, use of profiling
1.1.1.3 root 1644: causes static variable destructors (currently used only in C++) not to
1645: be run.
1.1 root 1646:
1.1.1.4 root 1647: @item
1648: Use of @samp{-I/usr/include} may cause trouble.
1649:
1650: Many systems come with header files that won't work with GNU CC unless
1651: corrected by @code{fixincludes}. The corrected header files go in a new
1652: directory; GNU CC searches this directory before @file{/usr/include}.
1653: If you use @samp{-I/usr/include}, this tells GNU CC to search
1654: @file{/usr/include} earlier on, before the corrected headers. The
1655: result is that you get the uncorrected header files.
1656:
1.1.1.5 root 1657: Instead, you should use these options (when compiling C programs):
1.1.1.4 root 1658:
1.1.1.5 root 1659: @smallexample
1.1.1.4 root 1660: -I/usr/local/lib/gcc-lib/@var{target}/@var{version}/include -I/usr/include
1.1.1.5 root 1661: @end smallexample
1662:
1663: For C++ programs, GNU CC also uses a special directory that defines C++
1664: interfaces to standard C subroutines. This directory is meant to be
1665: searched @emph{before} other standard include directories, so that it
1666: takes precedence. If you are compiling C++ programs and specifying
1667: include directories explicitly, use this option first, then the two
1668: options above:
1669:
1670: @example
1671: -I/usr/local/lib/g++-include
1.1.1.4 root 1672: @end example
1673:
1.1.1.3 root 1674: @ignore
1675: @cindex @code{vfork}, for the Sun-4
1676: @item
1677: There is a bug in @code{vfork} on the Sun-4 which causes the registers
1678: of the child process to clobber those of the parent. Because of this,
1679: programs that call @code{vfork} are likely to lose when compiled
1680: optimized with GNU CC when the child code alters registers which contain
1681: C variables in the parent. This affects variables which are live in the
1682: parent across the call to @code{vfork}.
1.1 root 1683:
1.1.1.3 root 1684: If you encounter this, you can work around the problem by declaring
1685: variables @code{volatile} in the function that calls @code{vfork}, until
1686: the problem goes away, or by not declaring them @code{register} and not
1687: using @samp{-O} for those source files.
1688: @end ignore
1.1 root 1689:
1.1.1.3 root 1690: @item
1.1.1.6 root 1691: On some SGI systems, when you use @samp{-lgl_s} as an option,
1692: it gets translated magically to @samp{-lgl_s -lX11_s -lc_s}.
1693: Naturally, this does not happen when you use GNU CC.
1694: You must specify all three options explicitly.
1695:
1696: @item
1.1.1.4 root 1697: On a Sparc, GNU CC aligns all values of type @code{double} on an 8-byte
1698: boundary, and it expects every @code{double} to be so aligned. The Sun
1699: compiler usually gives @code{double} values 8-byte alignment, with one
1700: exception: function arguments of type @code{double} may not be aligned.
1701:
1702: As a result, if a function compiled with Sun CC takes the address of an
1703: argument of type @code{double} and passes this pointer of type
1704: @code{double *} to a function compiled with GNU CC, dereferencing the
1705: pointer may cause a fatal signal.
1706:
1707: One way to solve this problem is to compile your entire program with GNU
1708: CC. Another solution is to modify the function that is compiled with
1709: Sun CC to copy the argument into a local variable; local variables
1710: are always properly aligned. A third solution is to modify the function
1711: that uses the pointer to dereference it via the following function
1712: @code{access_double} instead of directly with @samp{*}:
1713:
1.1.1.5 root 1714: @smallexample
1.1.1.4 root 1715: inline double
1716: access_double (double *unaligned_ptr)
1717: @{
1718: union d2i @{ double d; int i[2]; @};
1719:
1720: union d2i *p = (union d2i *) unaligned_ptr;
1721: union d2i u;
1722:
1723: u.i[0] = p->i[0];
1724: u.i[1] = p->i[1];
1725:
1726: return u.d;
1727: @}
1.1.1.5 root 1728: @end smallexample
1.1.1.4 root 1729:
1730: @noindent
1731: Storing into the pointer can be done likewise with the same union.
1732:
1733: @item
1.1.1.5 root 1734: On Solaris, the @code{malloc} function in the @file{libmalloc.a} library
1735: may allocate memory that is only 4 byte aligned. Since GNU CC on the
1736: Sparc assumes that doubles are 8 byte aligned, this may result in a
1737: fatal signal if doubles are stored in memory allocated by the
1738: @file{libmalloc.a} library.
1739:
1740: The solution is to not use the @file{libmalloc.a} library. Use instead
1741: @code{malloc} and related functions from @file{libc.a}; they do not have
1742: this problem.
1743:
1744: @item
1.1.1.3 root 1745: Sun forgot to include a static version of @file{libdl.a} with some
1746: versions of SunOS (mainly 4.1). This results in undefined symbols when
1747: linking static binaries (that is, if you use @samp{-static}). If you
1748: see undefined symbols @code{_dlclose}, @code{_dlsym} or @code{_dlopen}
1749: when linking, compile and link against the file
1750: @file{mit/util/misc/dlsym.c} from the MIT version of X windows.
1.1 root 1751:
1752: @item
1.1.1.6 root 1753: The 128-bit long double format that the Sparc port supports currently
1754: works by using the architecturally defined quad-word floating point
1.1.1.8 root 1755: instructions. Since there is no hardware that supports these
1756: instructions they must be emulated by the operating system. Long
1757: doubles do not work in Sun OS versions 4.0.3 and earlier, because the
1758: kernel emulator uses an obsolete and incompatible format. Long doubles
1759: do not work in Sun OS version 4.1.1 due to a problem in a Sun library.
1760: Long doubles do work on Sun OS versions 4.1.2 and higher, but GNU CC
1761: does not enable them by default. Long doubles appear to work in Sun OS
1762: 5.x (Solaris 2.x).
1.1.1.6 root 1763:
1764: @item
1765: On HP-UX version 9.01 on the HP PA, the HP compiler @code{cc} does not
1766: compile GNU CC correctly. We do not yet know why. However, GNU CC
1767: compiled on earlier HP-UX versions works properly on HP-UX 9.01 and can
1768: compile itself properly on 9.01.
1769:
1770: @item
1.1.1.3 root 1771: On the HP PA machine, ADB sometimes fails to work on functions compiled
1772: with GNU CC. Specifically, it fails to work on functions that use
1773: @code{alloca} or variable-size arrays. This is because GNU CC doesn't
1.1.1.4 root 1774: generate HP-UX unwind descriptors for such functions. It may even be
1.1.1.3 root 1775: impossible to generate them.
1.1.1.2 root 1776:
1.1.1.3 root 1777: @item
1.1.1.4 root 1778: Debugging (@samp{-g}) is not supported on the HP PA machine, unless you use
1779: the preliminary GNU tools (@pxref{Installation}).
1.1.1.2 root 1780:
1781: @item
1.1.1.4 root 1782: Taking the address of a label may generate errors from the HP-UX
1783: PA assembler. GAS for the PA does not have this problem.
1784:
1785: @item
1.1.1.6 root 1786: Using floating point parameters for indirect calls to static functions
1787: will not work when using the HP assembler. There simply is no way for GCC
1788: to specify what registers hold arguments for static functions when using
1789: the HP assembler. GAS for the PA does not have this problem.
1790:
1791: @item
1.1.1.8 root 1792: In extremely rare cases involving some very large functions you may
1793: receive errors from the HP linker complaining about an out of bounds
1794: unconditional branch offset. This used to occur more often in previous
1795: versions of GNU CC, but is now exceptionally rare. If you should run
1796: into it, you can work around by making your function smaller.
1.1 root 1797:
1.1.1.3 root 1798: @item
1.1.1.5 root 1799: GNU CC compiled code sometimes emits warnings from the HP-UX assembler of
1800: the form:
1801:
1802: @smallexample
1803: (warning) Use of GR3 when
1804: frame >= 8192 may cause conflict.
1805: @end smallexample
1806:
1807: These warnings are harmless and can be safely ignored.
1808:
1809: @item
1.1.1.3 root 1810: The current version of the assembler (@file{/bin/as}) for the RS/6000
1811: has certain problems that prevent the @samp{-g} option in GCC from
1.1.1.5 root 1812: working. Note that @file{Makefile.in} uses @samp{-g} by default when
1813: compiling @file{libgcc2.c}.
1814:
1815: IBM has produced a fixed version of the assembler. The upgraded
1816: assembler unfortunately was not included in any of the AIX 3.2 update
1817: PTF releases (3.2.2, 3.2.3, or 3.2.3e). Users of AIX 3.1 should request
1818: PTF U403044 from IBM and users of AIX 3.2 should request PTF U416277.
1819: See the file @file{README.RS6000} for more details on these updates.
1820:
1821: You can test for the presense of a fixed assembler by using the
1822: command
1.1 root 1823:
1.1.1.5 root 1824: @smallexample
1825: as -u < /dev/null
1826: @end smallexample
1827:
1828: @noindent
1829: If the command exits normally, the assembler fix already is installed.
1830: If the assembler complains that "-u" is an unknown flag, you need to
1831: order the fix.
1.1 root 1832:
1833: @item
1.1.1.3 root 1834: On the IBM RS/6000, compiling code of the form
1.1 root 1835:
1.1.1.5 root 1836: @smallexample
1.1.1.3 root 1837: extern int foo;
1.1 root 1838:
1.1.1.3 root 1839: @dots{} foo @dots{}
1.1 root 1840:
1.1.1.3 root 1841: static int foo;
1.1.1.5 root 1842: @end smallexample
1.1 root 1843:
1.1.1.3 root 1844: @noindent
1845: will cause the linker to report an undefined symbol @code{foo}.
1846: Although this behavior differs from most other systems, it is not a
1847: bug because redefining an @code{extern} variable as @code{static}
1848: is undefined in ANSI C.
1.1 root 1849:
1850: @item
1.1.1.5 root 1851: AIX on the RS/6000 provides support (NLS) for environments outside of
1852: the United States. Compilers and assemblers use NLS to support
1853: locale-specific representations of various objects including
1854: floating-point numbers ("." vs "," for separating decimal fractions).
1855: There have been problems reported where the library linked with GCC does
1856: not produce the same floating-point formats that the assembler accepts.
1857: If you have this problem, set the LANG environment variable to "C" or
1858: "En_US".
1859:
1860: @item
1.1.1.7 root 1861: Even if you specify @samp{-fdollars-in-identifiers},
1862: you cannot successfully use @samp{$} in identifiers on the RS/6000 due
1863: to a restriction in the IBM assembler. GAS supports these
1864: identifiers.
1865:
1866: @item
1867: On the RS/6000, XLC version 1.3.0.0 will miscompile @file{jump.c}. XLC
1868: version 1.3.0.1 or later fixes this problem. You can obtain XLC-1.3.0.2
1869: by requesting PTF 421749 from IBM.
1.1.1.6 root 1870:
1871: @item
1.1.1.5 root 1872: There is an assembler bug in versions of DG/UX prior to 5.4.2.01 that
1873: occurs when the @samp{fldcr} instruction is used. GNU CC uses
1874: @samp{fldcr} on the 88100 to serialize volatile memory references. Use
1.1.1.6 root 1875: the option @samp{-mno-serialize-volatile} if your version of the
1.1.1.5 root 1876: assembler has this bug.
1877:
1878: @item
1.1.1.3 root 1879: On VMS, GAS versions 1.38.1 and earlier may cause spurious warning
1880: messages from the linker. These warning messages complain of mismatched
1881: psect attributes. You can ignore them. @xref{VMS Install}.
1.1 root 1882:
1.1.1.4 root 1883: @item
1.1.1.5 root 1884: On NewsOS version 3, if you include both of the files @file{stddef.h}
1885: and @file{sys/types.h}, you get an error because there are two typedefs
1886: of @code{size_t}. You should change @file{sys/types.h} by adding these
1.1.1.4 root 1887: lines around the definition of @code{size_t}:
1888:
1.1.1.5 root 1889: @smallexample
1.1.1.4 root 1890: #ifndef _SIZE_T
1891: #define _SIZE_T
1892: @var{actual typedef here}
1893: #endif
1.1.1.5 root 1894: @end smallexample
1.1.1.4 root 1895:
1.1.1.3 root 1896: @cindex Alliant
1.1 root 1897: @item
1.1.1.3 root 1898: On the Alliant, the system's own convention for returning structures
1899: and unions is unusual, and is not compatible with GNU CC no matter
1900: what options are used.
1.1 root 1901:
1.1.1.3 root 1902: @cindex RT PC
1903: @cindex IBM RT PC
1.1 root 1904: @item
1.1.1.5 root 1905: On the IBM RT PC, the MetaWare HighC compiler (hc) uses a different
1906: convention for structure and union returning. Use the option
1.1.1.3 root 1907: @samp{-mhc-struct-return} to tell GNU CC to use a convention compatible
1908: with it.
1.1.1.2 root 1909:
1.1.1.3 root 1910: @cindex Vax calling convention
1911: @cindex Ultrix calling convention
1.1.1.2 root 1912: @item
1.1.1.3 root 1913: On Ultrix, the Fortran compiler expects registers 2 through 5 to be saved
1914: by function calls. However, the C compiler uses conventions compatible
1915: with BSD Unix: registers 2 through 5 may be clobbered by function calls.
1.1 root 1916:
1.1.1.3 root 1917: GNU CC uses the same convention as the Ultrix C compiler. You can use
1918: these options to produce code compatible with the Fortran compiler:
1.1 root 1919:
1.1.1.3 root 1920: @smallexample
1921: -fcall-saved-r2 -fcall-saved-r3 -fcall-saved-r4 -fcall-saved-r5
1922: @end smallexample
1.1.1.4 root 1923:
1924: @item
1925: On the WE32k, you may find that programs compiled with GNU CC do not
1.1.1.8 root 1926: work with the standard shared C library. You may need to link with
1.1.1.4 root 1927: the ordinary C compiler. If you do so, you must specify the following
1928: options:
1929:
1930: @smallexample
1.1.1.8 root 1931: -L/usr/local/lib/gcc-lib/we32k-att-sysv/2.7.1 -lgcc -lc_s
1.1.1.4 root 1932: @end smallexample
1933:
1934: The first specifies where to find the library @file{libgcc.a}
1935: specified with the @samp{-lgcc} option.
1936:
1937: GNU CC does linking by invoking @code{ld}, just as @code{cc} does, and
1938: there is no reason why it @emph{should} matter which compilation program
1939: you use to invoke @code{ld}. If someone tracks this problem down,
1940: it can probably be fixed easily.
1.1.1.5 root 1941:
1942: @item
1943: On the Alpha, you may get assembler errors about invalid syntax as a
1944: result of floating point constants. This is due to a bug in the C
1945: library functions @code{ecvt}, @code{fcvt} and @code{gcvt}. Given valid
1946: floating point numbers, they sometimes print @samp{NaN}.
1947:
1948: @item
1949: On Irix 4.0.5F (and perhaps in some other versions), an assembler bug
1950: sometimes reorders instructions incorrectly when optimization is turned
1951: on. If you think this may be happening to you, try using the GNU
1952: assembler; GAS version 2.1 supports ECOFF on Irix.
1953:
1954: Or use the @samp{-noasmopt} option when you compile GNU CC with itself,
1955: and then again when you compile your program. (This is a temporary
1956: kludge to turn off assembler optimization on Irix.) If this proves to
1957: be what you need, edit the assembler spec in the file @file{specs} so
1958: that it unconditionally passes @samp{-O0} to the assembler, and never
1959: passes @samp{-O2} or @samp{-O3}.
1960: @end itemize
1961:
1962: @node External Bugs
1963: @section Problems Compiling Certain Programs
1964:
1.1.1.7 root 1965: @c prevent bad page break with this line
1966: Certain programs have problems compiling.
1967:
1.1.1.5 root 1968: @itemize @bullet
1969: @item
1970: Parse errors may occur compiling X11 on a Decstation running Ultrix 4.2
1971: because of problems in DEC's versions of the X11 header files
1972: @file{X11/Xlib.h} and @file{X11/Xutil.h}. People recommend adding
1973: @samp{-I/usr/include/mit} to use the MIT versions of the header files,
1974: using the @samp{-traditional} switch to turn off ANSI C, or fixing the
1975: header files by adding this:
1976:
1977: @example
1978: #ifdef __STDC__
1979: #define NeedFunctionPrototypes 0
1980: #endif
1981: @end example
1982:
1983: @item
1.1.1.6 root 1984: If you have trouble compiling Perl on a SunOS 4 system, it may be
1985: because Perl specifies @samp{-I/usr/ucbinclude}. This accesses the
1986: unfixed header files. Perl specifies the options
1987:
1988: @example
1989: -traditional -Dvolatile=__volatile__
1990: -I/usr/include/sun -I/usr/ucbinclude
1991: -fpcc-struct-return
1992: @end example
1993:
1994: @noindent
1.1.1.7 root 1995: most of which are unnecessary with GCC 2.4.5 and newer versions. You
1996: can make a properly working Perl by setting @code{ccflags} to
1997: @samp{-fwritable-strings} (implied by the @samp{-traditional} in the
1998: original options) and @code{cppflags} to empty in @file{config.sh}, then
1999: typing @samp{./doSH; make depend; make}.
1.1.1.6 root 2000:
2001: @item
1.1.1.5 root 2002: On various 386 Unix systems derived from System V, including SCO, ISC,
2003: and ESIX, you may get error messages about running out of virtual memory
2004: while compiling certain programs.
2005:
2006: You can prevent this problem by linking GNU CC with the GNU malloc
2007: (which thus replaces the malloc that comes with the system). GNU malloc
2008: is available as a separate package, and also in the file
2009: @file{src/gmalloc.c} in the GNU Emacs 19 distribution.
2010:
2011: If you have installed GNU malloc as a separate library package, use this
2012: option when you relink GNU CC:
2013:
2014: @example
2015: MALLOC=/usr/local/lib/libgmalloc.a
2016: @end example
2017:
2018: Alternatively, if you have compiled @file{gmalloc.c} from Emacs 19, copy
2019: the object file to @file{gmalloc.o} and use this option when you relink
2020: GNU CC:
2021:
2022: @example
2023: MALLOC=gmalloc.o
2024: @end example
1.1 root 2025: @end itemize
2026:
1.1.1.3 root 2027: @node Incompatibilities
2028: @section Incompatibilities of GNU CC
1.1 root 2029: @cindex incompatibilities of GNU CC
2030:
2031: There are several noteworthy incompatibilities between GNU C and most
2032: existing (non-ANSI) versions of C. The @samp{-traditional} option
1.1.1.3 root 2033: eliminates many of these incompatibilities, @emph{but not all}, by
1.1 root 2034: telling GNU C to behave like the other C compilers.
2035:
2036: @itemize @bullet
2037: @cindex string constants
2038: @cindex read-only strings
2039: @cindex shared strings
2040: @item
2041: GNU CC normally makes string constants read-only. If several
2042: identical-looking string constants are used, GNU CC stores only one
2043: copy of the string.
2044:
2045: @cindex @code{mktemp}, and constant strings
2046: One consequence is that you cannot call @code{mktemp} with a string
2047: constant argument. The function @code{mktemp} always alters the
2048: string its argument points to.
2049:
2050: @cindex @code{sscanf}, and constant strings
2051: @cindex @code{fscanf}, and constant strings
2052: @cindex @code{scanf}, and constant strings
2053: Another consequence is that @code{sscanf} does not work on some systems
2054: when passed a string constant as its format control string or input.
2055: This is because @code{sscanf} incorrectly tries to write into the string
2056: constant. Likewise @code{fscanf} and @code{scanf}.
2057:
2058: The best solution to these problems is to change the program to use
2059: @code{char}-array variables with initialization strings for these
2060: purposes instead of string constants. But if this is not possible,
2061: you can use the @samp{-fwritable-strings} flag, which directs GNU CC
2062: to handle string constants the same way most C compilers do.
2063: @samp{-traditional} also has this effect, among others.
2064:
2065: @item
1.1.1.3 root 2066: @code{-2147483648} is positive.
2067:
2068: This is because 2147483648 cannot fit in the type @code{int}, so
2069: (following the ANSI C rules) its data type is @code{unsigned long int}.
2070: Negating this value yields 2147483648 again.
2071:
2072: @item
1.1 root 2073: GNU CC does not substitute macro arguments when they appear inside of
2074: string constants. For example, the following macro in GNU CC
2075:
2076: @example
2077: #define foo(a) "a"
2078: @end example
2079:
2080: @noindent
2081: will produce output @code{"a"} regardless of what the argument @var{a} is.
2082:
2083: The @samp{-traditional} option directs GNU CC to handle such cases
2084: (among others) in the old-fashioned (non-ANSI) fashion.
2085:
2086: @cindex @code{setjmp} incompatibilities
2087: @cindex @code{longjmp} incompatibilities
2088: @item
2089: When you use @code{setjmp} and @code{longjmp}, the only automatic
2090: variables guaranteed to remain valid are those declared
2091: @code{volatile}. This is a consequence of automatic register
2092: allocation. Consider this function:
2093:
2094: @example
2095: jmp_buf j;
2096:
2097: foo ()
2098: @{
2099: int a, b;
2100:
2101: a = fun1 ();
2102: if (setjmp (j))
2103: return a;
2104:
2105: a = fun2 ();
2106: /* @r{@code{longjmp (j)} may occur in @code{fun3}.} */
2107: return a + fun3 ();
2108: @}
2109: @end example
2110:
2111: Here @code{a} may or may not be restored to its first value when the
2112: @code{longjmp} occurs. If @code{a} is allocated in a register, then
2113: its first value is restored; otherwise, it keeps the last value stored
2114: in it.
2115:
2116: If you use the @samp{-W} option with the @samp{-O} option, you will
2117: get a warning when GNU CC thinks such a problem might be possible.
2118:
2119: The @samp{-traditional} option directs GNU C to put variables in
2120: the stack by default, rather than in registers, in functions that
2121: call @code{setjmp}. This results in the behavior found in
2122: traditional C compilers.
2123:
1.1.1.3 root 2124: @item
1.1.1.8 root 2125: Programs that use preprocessing directives in the middle of macro
1.1.1.3 root 2126: arguments do not work with GNU CC. For example, a program like this
2127: will not work:
2128:
2129: @example
2130: foobar (
2131: #define luser
2132: hack)
2133: @end example
2134:
2135: ANSI C does not permit such a construct. It would make sense to support
2136: it when @samp{-traditional} is used, but it is too much work to
2137: implement.
2138:
1.1 root 2139: @cindex external declaration scope
2140: @cindex scope of external declarations
2141: @cindex declaration scope
2142: @item
2143: Declarations of external variables and functions within a block apply
2144: only to the block containing the declaration. In other words, they
2145: have the same scope as any other declaration in the same place.
2146:
2147: In some other C compilers, a @code{extern} declaration affects all the
2148: rest of the file even if it happens within a block.
2149:
2150: The @samp{-traditional} option directs GNU C to treat all @code{extern}
2151: declarations as global, like traditional compilers.
2152:
2153: @item
2154: In traditional C, you can combine @code{long}, etc., with a typedef name,
2155: as shown here:
2156:
2157: @example
2158: typedef int foo;
2159: typedef long foo bar;
2160: @end example
2161:
2162: In ANSI C, this is not allowed: @code{long} and other type modifiers
2163: require an explicit @code{int}. Because this criterion is expressed
2164: by Bison grammar rules rather than C code, the @samp{-traditional}
2165: flag cannot alter it.
2166:
2167: @cindex typedef names as function parameters
2168: @item
2169: PCC allows typedef names to be used as function parameters. The
2170: difficulty described immediately above applies here too.
2171:
2172: @cindex whitespace
2173: @item
2174: PCC allows whitespace in the middle of compound assignment operators
2175: such as @samp{+=}. GNU CC, following the ANSI standard, does not
2176: allow this. The difficulty described immediately above applies here
2177: too.
2178:
2179: @cindex apostrophes
2180: @cindex '
2181: @item
1.1.1.4 root 2182: GNU CC complains about unterminated character constants inside of
1.1.1.8 root 2183: preprocessing conditionals that fail. Some programs have English
1.1.1.4 root 2184: comments enclosed in conditionals that are guaranteed to fail; if these
2185: comments contain apostrophes, GNU CC will probably report an error. For
2186: example, this code would produce an error:
1.1 root 2187:
2188: @example
2189: #if 0
2190: You can't expect this to work.
2191: #endif
2192: @end example
2193:
2194: The best solution to such a problem is to put the text into an actual
2195: C comment delimited by @samp{/*@dots{}*/}. However,
2196: @samp{-traditional} suppresses these error messages.
2197:
1.1.1.4 root 2198: @item
2199: Many user programs contain the declaration @samp{long time ();}. In the
2200: past, the system header files on many systems did not actually declare
2201: @code{time}, so it did not matter what type your program declared it to
2202: return. But in systems with ANSI C headers, @code{time} is declared to
2203: return @code{time_t}, and if that is not the same as @code{long}, then
2204: @samp{long time ();} is erroneous.
2205:
2206: The solution is to change your program to use @code{time_t} as the return
2207: type of @code{time}.
2208:
1.1 root 2209: @cindex @code{float} as function value type
2210: @item
2211: When compiling functions that return @code{float}, PCC converts it to
2212: a double. GNU CC actually returns a @code{float}. If you are concerned
2213: with PCC compatibility, you should declare your functions to return
2214: @code{double}; you might as well say what you mean.
2215:
2216: @cindex structures
2217: @cindex unions
2218: @item
2219: When compiling functions that return structures or unions, GNU CC
2220: output code normally uses a method different from that used on most
2221: versions of Unix. As a result, code compiled with GNU CC cannot call
2222: a structure-returning function compiled with PCC, and vice versa.
2223:
2224: The method used by GNU CC is as follows: a structure or union which is
2225: 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or union
2226: with any other size is stored into an address supplied by the caller
2227: (usually in a special, fixed register, but on some machines it is passed
2228: on the stack). The machine-description macros @code{STRUCT_VALUE} and
2229: @code{STRUCT_INCOMING_VALUE} tell GNU CC where to pass this address.
2230:
1.1.1.3 root 2231: By contrast, PCC on most target machines returns structures and unions
2232: of any size by copying the data into an area of static storage, and then
2233: returning the address of that storage as if it were a pointer value.
2234: The caller must copy the data from that memory area to the place where
2235: the value is wanted. GNU CC does not use this method because it is
2236: slower and nonreentrant.
2237:
2238: On some newer machines, PCC uses a reentrant convention for all
2239: structure and union returning. GNU CC on most of these machines uses a
2240: compatible convention when returning structures and unions in memory,
2241: but still returns small structures and unions in registers.
2242:
2243: You can tell GNU CC to use a compatible convention for all structure and
2244: union returning with the option @samp{-fpcc-struct-return}.
1.1.1.6 root 2245:
2246: @cindex preprocessing tokens
2247: @cindex preprocessing numbers
2248: @item
2249: GNU C complains about program fragments such as @samp{0x74ae-0x4000}
2250: which appear to be two hexadecimal constants separated by the minus
2251: operator. Actually, this string is a single @dfn{preprocessing token}.
2252: Each such token must correspond to one token in C. Since this does not,
2253: GNU C prints an error message. Although it may appear obvious that what
2254: is meant is an operator and two values, the ANSI C standard specifically
2255: requires that this be treated as erroneous.
2256:
2257: A @dfn{preprocessing token} is a @dfn{preprocessing number} if it
2258: begins with a digit and is followed by letters, underscores, digits,
2259: periods and @samp{e+}, @samp{e-}, @samp{E+}, or @samp{E-} character
2260: sequences.
2261:
2262: To make the above program fragment valid, place whitespace in front of
2263: the minus sign. This whitespace will end the preprocessing number.
2264: @end itemize
2265:
2266: @node Fixed Headers
2267: @section Fixed Header Files
2268:
2269: GNU CC needs to install corrected versions of some system header files.
2270: This is because most target systems have some header files that won't
2271: work with GNU CC unless they are changed. Some have bugs, some are
2272: incompatible with ANSI C, and some depend on special features of other
2273: compilers.
2274:
2275: Installing GNU CC automatically creates and installs the fixed header
2276: files, by running a program called @code{fixincludes} (or for certain
2277: targets an alternative such as @code{fixinc.svr4}). Normally, you
2278: don't need to pay attention to this. But there are cases where it
2279: doesn't do the right thing automatically.
2280:
2281: @itemize @bullet
2282: @item
2283: If you update the system's header files, such as by installing a new
2284: system version, the fixed header files of GNU CC are not automatically
2285: updated. The easiest way to update them is to reinstall GNU CC. (If
2286: you want to be clever, look in the makefile and you can find a
2287: shortcut.)
2288:
2289: @item
2290: On some systems, in particular SunOS 4, header file directories contain
2291: machine-specific symbolic links in certain places. This makes it
2292: possible to share most of the header files among hosts running the
2293: same version of SunOS 4 on different machine models.
2294:
2295: The programs that fix the header files do not understand this special
2296: way of using symbolic links; therefore, the directory of fixed header
2297: files is good only for the machine model used to build it.
2298:
2299: In SunOS 4, only programs that look inside the kernel will notice the
2300: difference between machine models. Therefore, for most purposes, you
2301: need not be concerned about this.
2302:
2303: It is possible to make separate sets of fixed header files for the
2304: different machine models, and arrange a structure of symbolic links so
2305: as to use the proper set, but you'll have to do this by hand.
2306:
2307: @item
2308: On Lynxos, GNU CC by default does not fix the header files. This is
2309: because bugs in the shell cause the @code{fixincludes} script to fail.
2310:
2311: This means you will encounter problems due to bugs in the system header
2312: files. It may be no comfort that they aren't GNU CC's fault, but it
2313: does mean that there's nothing for us to do about them.
1.1.1.3 root 2314: @end itemize
2315:
1.1.1.8 root 2316: @node Standard Libraries
2317: @section Standard Libraries
2318:
2319: GNU CC by itself attempts to be what the ISO/ANSI C standard calls a
2320: @dfn{conforming freestanding implementation}. This means all ANSI
2321: C language features are available, as well as the contents of
2322: @file{float.h}, @file{limits.h}, @file{stdarg.h}, and
2323: @file{stddef.h}. The rest of the C library is supplied by the
2324: vendor of the operating system. If that C library doesn't conform to
2325: the C standards, then your programs might get warnings (especially when
2326: using @samp{-Wall}) that you don't expect.
2327:
2328: For example, the @code{sprintf} function on SunOS 4.1.3 returns
2329: @code{char *} while the C standard says that @code{sprintf} returns an
2330: @code{int}. The @code{fixincludes} program could make the prototype for
2331: this function match the Standard, but that would be wrong, since the
2332: function will still return @code{char *}.
2333:
2334: If you need a Standard compliant library, then you need to find one, as
2335: GNU CC does not provide one. The GNU C library (called @code{glibc})
2336: has been ported to a number of operating systems, and provides ANSI/ISO,
2337: POSIX, BSD and SystemV compatibility. You could also ask your operating
2338: system vendor if newer libraries are available.
2339:
1.1.1.3 root 2340: @node Disappointments
2341: @section Disappointments and Misunderstandings
2342:
2343: These problems are perhaps regrettable, but we don't know any practical
2344: way around them.
2345:
2346: @itemize @bullet
2347: @item
2348: Certain local variables aren't recognized by debuggers when you compile
2349: with optimization.
2350:
2351: This occurs because sometimes GNU CC optimizes the variable out of
2352: existence. There is no way to tell the debugger how to compute the
2353: value such a variable ``would have had'', and it is not clear that would
2354: be desirable anyway. So GNU CC simply does not mention the eliminated
2355: variable when it writes debugging information.
2356:
2357: You have to expect a certain amount of disagreement between the
2358: executable and your source code, when you use optimization.
2359:
2360: @cindex conflicting types
2361: @cindex scope of declaration
2362: @item
2363: Users often think it is a bug when GNU CC reports an error for code
2364: like this:
2365:
2366: @example
2367: int foo (struct mumble *);
2368:
2369: struct mumble @{ @dots{} @};
2370:
2371: int foo (struct mumble *x)
2372: @{ @dots{} @}
2373: @end example
2374:
2375: This code really is erroneous, because the scope of @code{struct
1.1.1.4 root 2376: mumble} in the prototype is limited to the argument list containing it.
1.1.1.3 root 2377: It does not refer to the @code{struct mumble} defined with file scope
2378: immediately below---they are two unrelated types with similar names in
2379: different scopes.
2380:
2381: But in the definition of @code{foo}, the file-scope type is used
2382: because that is available to be inherited. Thus, the definition and
2383: the prototype do not match, and you get an error.
2384:
2385: This behavior may seem silly, but it's what the ANSI standard specifies.
2386: It is easy enough for you to make your code work by moving the
2387: definition of @code{struct mumble} above the prototype. It's not worth
2388: being incompatible with ANSI C just to avoid an error for the example
2389: shown above.
1.1.1.4 root 2390:
2391: @item
2392: Accesses to bitfields even in volatile objects works by accessing larger
2393: objects, such as a byte or a word. You cannot rely on what size of
2394: object is accessed in order to read or write the bitfield; it may even
2395: vary for a given bitfield according to the precise usage.
2396:
2397: If you care about controlling the amount of memory that is accessed, use
2398: volatile but do not use bitfields.
2399:
2400: @item
1.1.1.5 root 2401: GNU CC comes with shell scripts to fix certain known problems in system
2402: header files. They install corrected copies of various header files in
2403: a special directory where only GNU CC will normally look for them. The
2404: scripts adapt to various systems by searching all the system header
2405: files for the problem cases that we know about.
2406:
2407: If new system header files are installed, nothing automatically arranges
2408: to update the corrected header files. You will have to reinstall GNU CC
2409: to fix the new header files. More specifically, go to the build
2410: directory and delete the files @file{stmp-fixinc} and
2411: @file{stmp-headers}, and the subdirectory @code{include}; then do
2412: @samp{make install} again.
2413:
2414: @item
1.1.1.4 root 2415: On 68000 systems, you can get paradoxical results if you test the
2416: precise values of floating point numbers. For example, you can find
2417: that a floating point value which is not a NaN is not equal to itself.
2418: This results from the fact that the the floating point registers hold a
2419: few more bits of precision than fit in a @code{double} in memory.
2420: Compiled code moves values between memory and floating point registers
2421: at its convenience, and moving them into memory truncates them.
2422:
1.1.1.5 root 2423: You can partially avoid this problem by using the @samp{-ffloat-store}
2424: option (@pxref{Optimize Options}).
1.1.1.4 root 2425:
2426: @item
2427: On the MIPS, variable argument functions using @file{varargs.h}
2428: cannot have a floating point value for the first argument. The
2429: reason for this is that in the absence of a prototype in scope,
2430: if the first argument is a floating point, it is passed in a
1.1.1.5 root 2431: floating point register, rather than an integer register.
1.1.1.4 root 2432:
2433: If the code is rewritten to use the ANSI standard @file{stdarg.h}
2434: method of variable arguments, and the prototype is in scope at
2435: the time of the call, everything will work fine.
1.1.1.3 root 2436: @end itemize
2437:
1.1.1.5 root 2438: @node C++ Misunderstandings
2439: @section Common Misunderstandings with GNU C++
2440:
2441: @cindex misunderstandings in C++
2442: @cindex surprises in C++
2443: @cindex C++ misunderstandings
2444: C++ is a complex language and an evolving one, and its standard definition
2445: (the ANSI C++ draft standard) is also evolving. As a result,
2446: your C++ compiler may occasionally surprise you, even when its behavior is
2447: correct. This section discusses some areas that frequently give rise to
2448: questions of this sort.
2449:
2450: @menu
2451: * Static Definitions:: Static member declarations are not definitions
2452: * Temporaries:: Temporaries may vanish before you expect
2453: @end menu
2454:
2455: @node Static Definitions
2456: @subsection Declare @emph{and} Define Static Members
2457:
2458: @cindex C++ static data, declaring and defining
2459: @cindex static data in C++, declaring and defining
2460: @cindex declaring static data in C++
2461: @cindex defining static data in C++
2462: When a class has static data members, it is not enough to @emph{declare}
2463: the static member; you must also @emph{define} it. For example:
2464:
2465: @example
2466: class Foo
2467: @{
2468: @dots{}
2469: void method();
2470: static int bar;
2471: @};
2472: @end example
2473:
2474: This declaration only establishes that the class @code{Foo} has an
2475: @code{int} named @code{Foo::bar}, and a member function named
2476: @code{Foo::method}. But you still need to define @emph{both}
2477: @code{method} and @code{bar} elsewhere. According to the draft ANSI
2478: standard, you must supply an initializer in one (and only one) source
2479: file, such as:
2480:
2481: @example
2482: int Foo::bar = 0;
2483: @end example
2484:
2485: Other C++ compilers may not correctly implement the standard behavior.
2486: As a result, when you switch to @code{g++} from one of these compilers,
2487: you may discover that a program that appeared to work correctly in fact
2488: does not conform to the standard: @code{g++} reports as undefined
2489: symbols any static data members that lack definitions.
2490:
2491: @node Temporaries
2492: @subsection Temporaries May Vanish Before You Expect
2493:
2494: @cindex temporaries, lifetime of
2495: @cindex portions of temporary objects, pointers to
2496: It is dangerous to use pointers or references to @emph{portions} of a
2497: temporary object. The compiler may very well delete the object before
2498: you expect it to, leaving a pointer to garbage. The most common place
2499: where this problem crops up is in classes like the libg++
2500: @code{String} class, that define a conversion function to type
2501: @code{char *} or @code{const char *}. However, any class that returns
2502: a pointer to some internal structure is potentially subject to this
2503: problem.
2504:
2505: For example, a program may use a function @code{strfunc} that returns
2506: @code{String} objects, and another function @code{charfunc} that
2507: operates on pointers to @code{char}:
2508:
2509: @example
2510: String strfunc ();
2511: void charfunc (const char *);
2512: @end example
2513:
2514: @noindent
2515: In this situation, it may seem natural to write @w{@samp{charfunc
2516: (strfunc ());}} based on the knowledge that class @code{String} has an
2517: explicit conversion to @code{char} pointers. However, what really
2518: happens is akin to @samp{charfunc (@w{strfunc ()}.@w{convert ()});},
2519: where the @code{convert} method is a function to do the same data
2520: conversion normally performed by a cast. Since the last use of the
2521: temporary @code{String} object is the call to the conversion function,
2522: the compiler may delete that object before actually calling
2523: @code{charfunc}. The compiler has no way of knowing that deleting the
2524: @code{String} object will invalidate the pointer. The pointer then
2525: points to garbage, so that by the time @code{charfunc} is called, it
2526: gets an invalid argument.
2527:
2528: Code like this may run successfully under some other compilers,
2529: especially those that delete temporaries relatively late. However, the
1.1.1.8 root 2530: GNU C++ behavior is also standard-conforming, so if your program depends
1.1.1.5 root 2531: on late destruction of temporaries it is not portable.
2532:
2533: If you think this is surprising, you should be aware that the ANSI C++
2534: committee continues to debate the lifetime-of-temporaries problem.
2535:
2536: For now, at least, the safe way to write such code is to give the
2537: temporary a name, which forces it to remain until the end of the scope of
2538: the name. For example:
2539:
2540: @example
2541: String& tmp = strfunc ();
2542: charfunc (tmp);
2543: @end example
2544:
1.1.1.4 root 2545: @node Protoize Caveats
2546: @section Caveats of using @code{protoize}
2547:
2548: The conversion programs @code{protoize} and @code{unprotoize} can
2549: sometimes change a source file in a way that won't work unless you
2550: rearrange it.
2551:
2552: @itemize @bullet
2553: @item
2554: @code{protoize} can insert references to a type name or type tag before
2555: the definition, or in a file where they are not defined.
2556:
2557: If this happens, compiler error messages should show you where the new
2558: references are, so fixing the file by hand is straightforward.
2559:
2560: @item
2561: There are some C constructs which @code{protoize} cannot figure out.
2562: For example, it can't determine argument types for declaring a
2563: pointer-to-function variable; this you must do by hand. @code{protoize}
2564: inserts a comment containing @samp{???} each time it finds such a
2565: variable; so you can find all such variables by searching for this
2566: string. ANSI C does not require declaring the argument types of
2567: pointer-to-function types.
2568:
2569: @item
2570: Using @code{unprotoize} can easily introduce bugs. If the program
2571: relied on prototypes to bring about conversion of arguments, these
2572: conversions will not take place in the program without prototypes.
2573: One case in which you can be sure @code{unprotoize} is safe is when
2574: you are removing prototypes that were made with @code{protoize}; if
2575: the program worked before without any prototypes, it will work again
2576: without them.
2577:
2578: You can find all the places where this problem might occur by compiling
2579: the program with the @samp{-Wconversion} option. It prints a warning
2580: whenever an argument is converted.
2581:
2582: @item
2583: Both conversion programs can be confused if there are macro calls in and
2584: around the text to be converted. In other words, the standard syntax
2585: for a declaration or definition must not result from expanding a macro.
2586: This problem is inherent in the design of C and cannot be fixed. If
2587: only a few functions have confusing macro calls, you can easily convert
2588: them manually.
2589:
2590: @item
2591: @code{protoize} cannot get the argument types for a function whose
1.1.1.8 root 2592: definition was not actually compiled due to preprocessing conditionals.
1.1.1.4 root 2593: When this happens, @code{protoize} changes nothing in regard to such
2594: a function. @code{protoize} tries to detect such instances and warn
2595: about them.
2596:
2597: You can generally work around this problem by using @code{protoize} step
2598: by step, each time specifying a different set of @samp{-D} options for
2599: compilation, until all of the functions have been converted. There is
2600: no automatic way to verify that you have got them all, however.
2601:
2602: @item
2603: Confusion may result if there is an occasion to convert a function
2604: declaration or definition in a region of source code where there is more
2605: than one formal parameter list present. Thus, attempts to convert code
2606: containing multiple (conditionally compiled) versions of a single
2607: function header (in the same vicinity) may not produce the desired (or
2608: expected) results.
2609:
2610: If you plan on converting source files which contain such code, it is
2611: recommended that you first make sure that each conditionally compiled
2612: region of source code which contains an alternative function header also
2613: contains at least one additional follower token (past the final right
2614: parenthesis of the function header). This should circumvent the
2615: problem.
2616:
2617: @item
2618: @code{unprotoize} can become confused when trying to convert a function
2619: definition or declaration which contains a declaration for a
2620: pointer-to-function formal argument which has the same name as the
2621: function being defined or declared. We recommand you avoid such choices
2622: of formal parameter names.
2623:
2624: @item
2625: You might also want to correct some of the indentation by hand and break
2626: long lines. (The conversion programs don't write lines longer than
2627: eighty characters in any case.)
2628: @end itemize
2629:
2630: @node Non-bugs
1.1.1.3 root 2631: @section Certain Changes We Don't Want to Make
2632:
2633: This section lists changes that people frequently request, but which
2634: we do not make because we think GNU CC is better without them.
2635:
2636: @itemize @bullet
2637: @item
2638: Checking the number and type of arguments to a function which has an
2639: old-fashioned definition and no prototype.
2640:
2641: Such a feature would work only occasionally---only for calls that appear
2642: in the same file as the called function, following the definition. The
2643: only way to check all calls reliably is to add a prototype for the
2644: function. But adding a prototype eliminates the motivation for this
2645: feature. So the feature is not worthwhile.
2646:
2647: @item
2648: Warning about using an expression whose type is signed as a shift count.
2649:
2650: Shift count operands are probably signed more often than unsigned.
2651: Warning about this would cause far more annoyance than good.
2652:
2653: @item
2654: Warning about assigning a signed value to an unsigned variable.
2655:
2656: Such assignments must be very common; warning about them would cause
2657: more annoyance than good.
2658:
1.1.1.4 root 2659: @item
2660: Warning about unreachable code.
2661:
2662: It's very common to have unreachable code in machine-generated
2663: programs. For example, this happens normally in some files of GNU C
2664: itself.
2665:
1.1.1.3 root 2666: @item
2667: Warning when a non-void function value is ignored.
2668:
2669: Coming as I do from a Lisp background, I balk at the idea that there is
2670: something dangerous about discarding a value. There are functions that
2671: return values which some callers may find useful; it makes no sense to
2672: clutter the program with a cast to @code{void} whenever the value isn't
2673: useful.
2674:
2675: @item
2676: Assuming (for optimization) that the address of an external symbol is
2677: never zero.
2678:
2679: This assumption is false on certain systems when @samp{#pragma weak} is
2680: used.
2681:
2682: @item
2683: Making @samp{-fshort-enums} the default.
2684:
2685: This would cause storage layout to be incompatible with most other C
2686: compilers. And it doesn't seem very important, given that you can get
2687: the same result in other ways. The case where it matters most is when
2688: the enumeration-valued object is inside a structure, and in that case
2689: you can specify a field width explicitly.
2690:
2691: @item
2692: Making bitfields unsigned by default on particular machines where ``the
2693: ABI standard'' says to do so.
2694:
2695: The ANSI C standard leaves it up to the implementation whether a bitfield
2696: declared plain @code{int} is signed or not. This in effect creates two
2697: alternative dialects of C.
2698:
1.1.1.5 root 2699: The GNU C compiler supports both dialects; you can specify the signed
2700: dialect with @samp{-fsigned-bitfields} and the unsigned dialect with
2701: @samp{-funsigned-bitfields}. However, this leaves open the question of
2702: which dialect to use by default.
1.1.1.3 root 2703:
2704: Currently, the preferred dialect makes plain bitfields signed, because
2705: this is simplest. Since @code{int} is the same as @code{signed int} in
2706: every other context, it is cleanest for them to be the same in bitfields
2707: as well.
2708:
2709: Some computer manufacturers have published Application Binary Interface
2710: standards which specify that plain bitfields should be unsigned. It is
2711: a mistake, however, to say anything about this issue in an ABI. This is
2712: because the handling of plain bitfields distinguishes two dialects of C.
2713: Both dialects are meaningful on every type of machine. Whether a
2714: particular object file was compiled using signed bitfields or unsigned
2715: is of no concern to other object files, even if they access the same
2716: bitfields in the same data structures.
2717:
2718: A given program is written in one or the other of these two dialects.
2719: The program stands a chance to work on most any machine if it is
2720: compiled with the proper dialect. It is unlikely to work at all if
2721: compiled with the wrong dialect.
2722:
2723: Many users appreciate the GNU C compiler because it provides an
2724: environment that is uniform across machines. These users would be
2725: inconvenienced if the compiler treated plain bitfields differently on
2726: certain machines.
2727:
2728: Occasionally users write programs intended only for a particular machine
2729: type. On these occasions, the users would benefit if the GNU C compiler
2730: were to support by default the same dialect as the other compilers on
2731: that machine. But such applications are rare. And users writing a
2732: program to run on more than one type of machine cannot possibly benefit
2733: from this kind of compatibility.
2734:
2735: This is why GNU CC does and will treat plain bitfields in the same
2736: fashion on all types of machines (by default).
2737:
2738: There are some arguments for making bitfields unsigned by default on all
2739: machines. If, for example, this becomes a universal de facto standard,
2740: it would make sense for GNU CC to go along with it. This is something
2741: to be considered in the future.
1.1 root 2742:
1.1.1.3 root 2743: (Of course, users strongly concerned about portability should indicate
2744: explicitly in each bitfield whether it is signed or not. In this way,
2745: they write programs which have the same meaning in both C dialects.)
1.1 root 2746:
1.1.1.3 root 2747: @item
2748: Undefining @code{__STDC__} when @samp{-ansi} is not used.
1.1 root 2749:
1.1.1.3 root 2750: Currently, GNU CC defines @code{__STDC__} as long as you don't use
2751: @samp{-traditional}. This provides good results in practice.
1.1 root 2752:
1.1.1.3 root 2753: Programmers normally use conditionals on @code{__STDC__} to ask whether
2754: it is safe to use certain features of ANSI C, such as function
2755: prototypes or ANSI token concatenation. Since plain @samp{gcc} supports
2756: all the features of ANSI C, the correct answer to these questions is
2757: ``yes''.
1.1 root 2758:
1.1.1.3 root 2759: Some users try to use @code{__STDC__} to check for the availability of
2760: certain library facilities. This is actually incorrect usage in an ANSI
2761: C program, because the ANSI C standard says that a conforming
2762: freestanding implementation should define @code{__STDC__} even though it
2763: does not have the library facilities. @samp{gcc -ansi -pedantic} is a
2764: conforming freestanding implementation, and it is therefore required to
2765: define @code{__STDC__}, even though it does not come with an ANSI C
2766: library.
1.1 root 2767:
1.1.1.3 root 2768: Sometimes people say that defining @code{__STDC__} in a compiler that
2769: does not completely conform to the ANSI C standard somehow violates the
2770: standard. This is illogical. The standard is a standard for compilers
2771: that claim to support ANSI C, such as @samp{gcc -ansi}---not for other
2772: compilers such as plain @samp{gcc}. Whatever the ANSI C standard says
2773: is relevant to the design of plain @samp{gcc} without @samp{-ansi} only
2774: for pragmatic reasons, not as a requirement.
1.1 root 2775:
2776: @item
1.1.1.3 root 2777: Undefining @code{__STDC__} in C++.
1.1 root 2778:
1.1.1.3 root 2779: Programs written to compile with C++-to-C translators get the
2780: value of @code{__STDC__} that goes with the C compiler that is
2781: subsequently used. These programs must test @code{__STDC__}
2782: to determine what kind of C preprocessor that compiler uses:
2783: whether they should concatenate tokens in the ANSI C fashion
2784: or in the traditional fashion.
1.1 root 2785:
1.1.1.3 root 2786: These programs work properly with GNU C++ if @code{__STDC__} is defined.
2787: They would not work otherwise.
2788:
2789: In addition, many header files are written to provide prototypes in ANSI
2790: C but not in traditional C. Many of these header files can work without
2791: change in C++ provided @code{__STDC__} is defined. If @code{__STDC__}
2792: is not defined, they will all fail, and will all need to be changed to
2793: test explicitly for C++ as well.
1.1 root 2794:
2795: @item
1.1.1.3 root 2796: Deleting ``empty'' loops.
1.1 root 2797:
1.1.1.3 root 2798: GNU CC does not delete ``empty'' loops because the most likely reason
2799: you would put one in a program is to have a delay. Deleting them will
2800: not make real programs run any faster, so it would be pointless.
1.1 root 2801:
1.1.1.3 root 2802: It would be different if optimization of a nonempty loop could produce
2803: an empty one. But this generally can't happen.
1.1.1.5 root 2804:
2805: @item
2806: Making side effects happen in the same order as in some other compiler.
2807:
2808: @cindex side effects, order of evaluation
2809: @cindex order of evaluation, side effects
2810: It is never safe to depend on the order of evaluation of side effects.
2811: For example, a function call like this may very well behave differently
2812: from one compiler to another:
2813:
2814: @example
2815: void func (int, int);
2816:
2817: int i = 2;
2818: func (i++, i++);
2819: @end example
2820:
2821: There is no guarantee (in either the C or the C++ standard language
2822: definitions) that the increments will be evaluated in any particular
2823: order. Either increment might happen first. @code{func} might get the
1.1.1.8 root 2824: arguments @samp{2, 3}, or it might get @samp{3, 2}, or even @samp{2, 2}.
1.1.1.5 root 2825:
2826: @item
1.1.1.7 root 2827: Not allowing structures with volatile fields in registers.
1.1.1.5 root 2828:
1.1.1.7 root 2829: Strictly speaking, there is no prohibition in the ANSI C standard
2830: against allowing structures with volatile fields in registers, but
2831: it does not seem to make any sense and is probably not what you wanted
2832: to do. So the compiler will give an error message in this case.
1.1.1.5 root 2833: @end itemize
2834:
2835: @node Warnings and Errors
2836: @section Warning Messages and Error Messages
2837:
2838: @cindex error messages
2839: @cindex warnings vs errors
2840: @cindex messages, warning and error
2841: The GNU compiler can produce two kinds of diagnostics: errors and
2842: warnings. Each kind has a different purpose:
2843:
2844: @itemize @w{}
2845: @item
2846: @emph{Errors} report problems that make it impossible to compile your
2847: program. GNU CC reports errors with the source file name and line
2848: number where the problem is apparent.
2849:
2850: @item
2851: @emph{Warnings} report other unusual conditions in your code that
2852: @emph{may} indicate a problem, although compilation can (and does)
2853: proceed. Warning messages also report the source file name and line
2854: number, but include the text @samp{warning:} to distinguish them
2855: from error messages.
1.1.1.3 root 2856: @end itemize
2857:
1.1.1.5 root 2858: Warnings may indicate danger points where you should check to make sure
2859: that your program really does what you intend; or the use of obsolete
2860: features; or the use of nonstandard features of GNU C or C++. Many
2861: warnings are issued only if you ask for them, with one of the @samp{-W}
2862: options (for instance, @samp{-Wall} requests a variety of useful
2863: warnings).
2864:
2865: GNU CC always tries to compile your program if possible; it never
1.1.1.8 root 2866: gratuitously rejects a program whose meaning is clear merely because
1.1.1.5 root 2867: (for instance) it fails to conform to a standard. In some cases,
2868: however, the C and C++ standards specify that certain extensions are
2869: forbidden, and a diagnostic @emph{must} be issued by a conforming
2870: compiler. The @samp{-pedantic} option tells GNU CC to issue warnings in
2871: such cases; @samp{-pedantic-errors} says to make them errors instead.
2872: This does not mean that @emph{all} non-ANSI constructs get warnings
2873: or errors.
2874:
2875: @xref{Warning Options,,Options to Request or Suppress Warnings}, for
2876: more detail on these and related command-line options.
2877:
1.1.1.3 root 2878: @node Bugs
1.1 root 2879: @chapter Reporting Bugs
2880: @cindex bugs
2881: @cindex reporting bugs
2882:
2883: Your bug reports play an essential role in making GNU CC reliable.
2884:
2885: When you encounter a problem, the first thing to do is to see if it is
1.1.1.3 root 2886: already known. @xref{Trouble}. If it isn't known, then you should
2887: report the problem.
1.1 root 2888:
2889: Reporting a bug may help you by bringing a solution to your problem, or
2890: it may not. (If it does not, look in the service directory; see
2891: @ref{Service}.) In any case, the principal function of a bug report is
2892: to help the entire community by making the next version of GNU CC work
2893: better. Bug reports are your contribution to the maintenance of GNU CC.
2894:
1.1.1.6 root 2895: Since the maintainers are very overloaded, we cannot respond to every
2896: bug report. However, if the bug has not been fixed, we are likely to
2897: send you a patch and ask you to tell us whether it works.
2898:
1.1 root 2899: In order for a bug report to serve its purpose, you must include the
2900: information that makes for fixing the bug.
2901:
2902: @menu
2903: * Criteria: Bug Criteria. Have you really found a bug?
1.1.1.3 root 2904: * Where: Bug Lists. Where to send your bug report.
1.1 root 2905: * Reporting: Bug Reporting. How to report a bug effectively.
1.1.1.3 root 2906: * Patches: Sending Patches. How to send a patch for GNU CC.
1.1 root 2907: * Known: Trouble. Known problems.
2908: * Help: Service. Where to ask for help.
2909: @end menu
2910:
1.1.1.3 root 2911: @node Bug Criteria
1.1 root 2912: @section Have You Found a Bug?
2913: @cindex bug criteria
2914:
2915: If you are not sure whether you have found a bug, here are some guidelines:
2916:
2917: @itemize @bullet
2918: @cindex fatal signal
2919: @cindex core dump
2920: @item
2921: If the compiler gets a fatal signal, for any input whatever, that is a
2922: compiler bug. Reliable compilers never crash.
2923:
2924: @cindex invalid assembly code
2925: @cindex assembly code, invalid
2926: @item
2927: If the compiler produces invalid assembly code, for any input whatever
2928: (except an @code{asm} statement), that is a compiler bug, unless the
2929: compiler reports errors (not just warnings) which would ordinarily
2930: prevent the assembler from being run.
2931:
2932: @cindex undefined behavior
2933: @cindex undefined function value
2934: @cindex increment operators
2935: @item
2936: If the compiler produces valid assembly code that does not correctly
2937: execute the input source code, that is a compiler bug.
2938:
2939: However, you must double-check to make sure, because you may have run
2940: into an incompatibility between GNU C and traditional C
2941: (@pxref{Incompatibilities}). These incompatibilities might be considered
2942: bugs, but they are inescapable consequences of valuable features.
2943:
2944: Or you may have a program whose behavior is undefined, which happened
1.1.1.3 root 2945: by chance to give the desired results with another C or C++ compiler.
1.1 root 2946:
2947: For example, in many nonoptimizing compilers, you can write @samp{x;}
2948: at the end of a function instead of @samp{return x;}, with the same
2949: results. But the value of the function is undefined if @code{return}
2950: is omitted; it is not a bug when GNU CC produces different results.
2951:
2952: Problems often result from expressions with two increment operators,
2953: as in @code{f (*p++, *p++)}. Your previous compiler might have
2954: interpreted that expression the way you intended; GNU CC might
2955: interpret it another way. Neither compiler is wrong. The bug is
2956: in your code.
2957:
2958: After you have localized the error to a single source line, it should
2959: be easy to check for these things. If your program is correct and
2960: well defined, you have found a compiler bug.
2961:
2962: @item
2963: If the compiler produces an error message for valid input, that is a
2964: compiler bug.
2965:
2966: @cindex invalid input
2967: @item
2968: If the compiler does not produce an error message for invalid input,
2969: that is a compiler bug. However, you should note that your idea of
2970: ``invalid input'' might be my idea of ``an extension'' or ``support
2971: for traditional practice''.
2972:
2973: @item
1.1.1.3 root 2974: If you are an experienced user of C or C++ compilers, your suggestions
2975: for improvement of GNU CC or GNU C++ are welcome in any case.
1.1 root 2976: @end itemize
2977:
1.1.1.3 root 2978: @node Bug Lists
2979: @section Where to Report Bugs
2980: @cindex bug report mailing lists
1.1.1.5 root 2981: @kindex bug-gcc@@prep.ai.mit.edu
1.1.1.7 root 2982: Send bug reports for GNU C to @samp{bug-gcc@@prep.ai.mit.edu}.
1.1 root 2983:
1.1.1.5 root 2984: @kindex bug-g++@@prep.ai.mit.edu
2985: @kindex bug-libg++@@prep.ai.mit.edu
1.1.1.7 root 2986: Send bug reports for GNU C++ to @samp{bug-g++@@prep.ai.mit.edu}.
1.1.1.5 root 2987: If your bug involves the C++ class library libg++, send mail to
2988: @samp{bug-lib-g++@@prep.ai.mit.edu}. If you're not sure, you can send
2989: the bug report to both lists.
2990:
1.1.1.7 root 2991: @strong{Do not send bug reports to @samp{help-gcc@@prep.ai.mit.edu} or
1.1.1.5 root 2992: to the newsgroup @samp{gnu.gcc.help}.} Most users of GNU CC do not want
2993: to receive bug reports. Those that do, have asked to be on
2994: @samp{bug-gcc} and/or @samp{bug-g++}.
1.1.1.3 root 2995:
2996: The mailing lists @samp{bug-gcc} and @samp{bug-g++} both have newsgroups
2997: which serve as repeaters: @samp{gnu.gcc.bug} and @samp{gnu.g++.bug}.
2998: Each mailing list and its newsgroup carry exactly the same messages.
2999:
3000: Often people think of posting bug reports to the newsgroup instead of
3001: mailing them. This appears to work, but it has one problem which can be
3002: crucial: a newsgroup posting does not contain a mail path back to the
1.1.1.4 root 3003: sender. Thus, if maintainers need more information, they may be unable
1.1.1.3 root 3004: to reach you. For this reason, you should always send bug reports by
3005: mail to the proper mailing list.
1.1 root 3006:
3007: As a last resort, send bug reports on paper to:
3008:
3009: @example
3010: GNU Compiler Bugs
3011: Free Software Foundation
1.1.1.8 root 3012: 59 Temple Place - Suite 330
3013: Boston, MA 02111-1307, USA
1.1 root 3014: @end example
3015:
1.1.1.3 root 3016: @node Bug Reporting
3017: @section How to Report Bugs
3018: @cindex compiler bugs, reporting
3019:
1.1 root 3020: The fundamental principle of reporting bugs usefully is this:
3021: @strong{report all the facts}. If you are not sure whether to state a
3022: fact or leave it out, state it!
3023:
3024: Often people omit facts because they think they know what causes the
3025: problem and they conclude that some details don't matter. Thus, you might
3026: assume that the name of the variable you use in an example does not matter.
3027: Well, probably it doesn't, but one cannot be sure. Perhaps the bug is a
3028: stray memory reference which happens to fetch from the location where that
3029: name is stored in memory; perhaps, if the name were different, the contents
3030: of that location would fool the compiler into doing the right thing despite
3031: the bug. Play it safe and give a specific, complete example. That is the
3032: easiest thing for you to do, and the most helpful.
3033:
1.1.1.3 root 3034: Keep in mind that the purpose of a bug report is to enable someone to
3035: fix the bug if it is not known. It isn't very important what happens if
1.1 root 3036: the bug is already known. Therefore, always write your bug reports on
3037: the assumption that the bug is not known.
3038:
3039: Sometimes people give a few sketchy facts and ask, ``Does this ring a
1.1.1.3 root 3040: bell?'' This cannot help us fix a bug, so it is basically useless. We
3041: respond by asking for enough details to enable us to investigate.
3042: You might as well expedite matters by sending them to begin with.
3043:
3044: Try to make your bug report self-contained. If we have to ask you for
3045: more information, it is best if you include all the previous information
3046: in your response, as well as the information that was missing.
1.1 root 3047:
1.1.1.7 root 3048: Please report each bug in a separate message. This makes it easier for
3049: us to track which bugs have been fixed and to forward your bugs reports
3050: to the appropriate maintainer.
3051:
1.1.1.8 root 3052: Do not compress and encode any part of your bug report using programs
3053: such as @file{uuencode}. If you do so it will slow down the processing
3054: of your bug. If you must submit multiple large files, use @file{shar},
3055: which allows us to read your message without having to run any
3056: decompression programs.
3057:
1.1.1.3 root 3058: To enable someone to investigate the bug, you should include all these
3059: things:
1.1 root 3060:
3061: @itemize @bullet
3062: @item
3063: The version of GNU CC. You can get this by running it with the
3064: @samp{-v} option.
3065:
1.1.1.3 root 3066: Without this, we won't know whether there is any point in looking for
1.1 root 3067: the bug in the current version of GNU CC.
3068:
3069: @item
1.1.1.3 root 3070: A complete input file that will reproduce the bug. If the bug is in the
3071: C preprocessor, send a source file and any header files that it
1.1 root 3072: requires. If the bug is in the compiler proper (@file{cc1}), run your
3073: source file through the C preprocessor by doing @samp{gcc -E
3074: @var{sourcefile} > @var{outfile}}, then include the contents of
1.1.1.3 root 3075: @var{outfile} in the bug report. (When you do this, use the same
3076: @samp{-I}, @samp{-D} or @samp{-U} options that you used in actual
3077: compilation.)
1.1 root 3078:
1.1.1.4 root 3079: A single statement is not enough of an example. In order to compile it,
3080: it must be embedded in a complete file of compiler input; and the bug
3081: might depend on the details of how this is done.
1.1 root 3082:
1.1.1.3 root 3083: Without a real example one can compile, all anyone can do about your bug
1.1 root 3084: report is wish you luck. It would be futile to try to guess how to
1.1.1.3 root 3085: provoke the bug. For example, bugs in register allocation and reloading
3086: frequently depend on every little detail of the function they happen in.
1.1 root 3087:
1.1.1.4 root 3088: Even if the input file that fails comes from a GNU program, you should
3089: still send the complete test case. Don't ask the GNU CC maintainers to
3090: do the extra work of obtaining the program in question---they are all
3091: overworked as it is. Also, the problem may depend on what is in the
3092: header files on your system; it is unreliable for the GNU CC maintainers
3093: to try the problem with the header files available to them. By sending
1.1.1.5 root 3094: CPP output, you can eliminate this source of uncertainty and save us
3095: a certain percentage of wild goose chases.
1.1.1.4 root 3096:
1.1 root 3097: @item
1.1.1.3 root 3098: The command arguments you gave GNU CC or GNU C++ to compile that example
3099: and observe the bug. For example, did you use @samp{-O}? To guarantee
3100: you won't omit something important, list all the options.
1.1 root 3101:
1.1.1.3 root 3102: If we were to try to guess the arguments, we would probably guess wrong
3103: and then we would not encounter the bug.
1.1 root 3104:
3105: @item
3106: The type of machine you are using, and the operating system name and
3107: version number.
3108:
3109: @item
3110: The operands you gave to the @code{configure} command when you installed
3111: the compiler.
3112:
3113: @item
1.1.1.3 root 3114: A complete list of any modifications you have made to the compiler
3115: source. (We don't promise to investigate the bug unless it happens in
3116: an unmodified compiler. But if you've made modifications and don't tell
3117: us, then you are sending us on a wild goose chase.)
1.1 root 3118:
1.1.1.4 root 3119: Be precise about these changes. A description in English is not
3120: enough---send a context diff for them.
1.1 root 3121:
1.1.1.3 root 3122: Adding files of your own (such as a machine description for a machine we
3123: don't support) is a modification of the compiler source.
1.1 root 3124:
3125: @item
1.1.1.3 root 3126: Details of any other deviations from the standard procedure for installing
3127: GNU CC.
1.1 root 3128:
3129: @item
1.1.1.3 root 3130: A description of what behavior you observe that you believe is
3131: incorrect. For example, ``The compiler gets a fatal signal,'' or,
3132: ``The assembler instruction at line 208 in the output is incorrect.''
1.1 root 3133:
1.1.1.3 root 3134: Of course, if the bug is that the compiler gets a fatal signal, then one
3135: can't miss it. But if the bug is incorrect output, the maintainer might
3136: not notice unless it is glaringly wrong. None of us has time to study
3137: all the assembler code from a 50-line C program just on the chance that
1.1.1.4 root 3138: one instruction might be wrong. We need @emph{you} to do this part!
1.1 root 3139:
1.1.1.3 root 3140: Even if the problem you experience is a fatal signal, you should still
3141: say so explicitly. Suppose something strange is going on, such as, your
3142: copy of the compiler is out of synch, or you have encountered a bug in
3143: the C library on your system. (This has happened!) Your copy might
3144: crash and the copy here would not. If you @i{said} to expect a crash,
3145: then when the compiler here fails to crash, we would know that the bug
3146: was not happening. If you don't say to expect a crash, then we would
3147: not know whether the bug was happening. We would not be able to draw
3148: any conclusion from our observations.
3149:
1.1.1.4 root 3150: If the problem is a diagnostic when compiling GNU CC with some other
3151: compiler, say whether it is a warning or an error.
3152:
1.1.1.3 root 3153: Often the observed symptom is incorrect output when your program is run.
3154: Sad to say, this is not enough information unless the program is short
3155: and simple. None of us has time to study a large program to figure out
3156: how it would work if compiled correctly, much less which line of it was
3157: compiled wrong. So you will have to do that. Tell us which source line
3158: it is, and what incorrect result happens when that line is executed. A
3159: person who understands the program can find this as easily as finding a
3160: bug in the program itself.
3161:
3162: @item
3163: If you send examples of assembler code output from GNU CC or GNU C++,
3164: please use @samp{-g} when you make them. The debugging information
3165: includes source line numbers which are essential for correlating the
3166: output with the input.
3167:
3168: @item
1.1.1.4 root 3169: If you wish to mention something in the GNU CC source, refer to it by
3170: context, not by line number.
1.1.1.3 root 3171:
3172: The line numbers in the development sources don't match those in your
3173: sources. Your line numbers would convey no useful information to the
3174: maintainers.
3175:
3176: @item
3177: Additional information from a debugger might enable someone to find a
3178: problem on a machine which he does not have available. However, you
3179: need to think when you collect this information if you want it to have
3180: any chance of being useful.
1.1 root 3181:
3182: @cindex backtrace for bug reports
3183: For example, many people send just a backtrace, but that is never
3184: useful by itself. A simple backtrace with arguments conveys little
3185: about GNU CC because the compiler is largely data-driven; the same
3186: functions are called over and over for different RTL insns, doing
3187: different things depending on the details of the insn.
3188:
3189: Most of the arguments listed in the backtrace are useless because they
3190: are pointers to RTL list structure. The numeric values of the
3191: pointers, which the debugger prints in the backtrace, have no
3192: significance whatever; all that matters is the contents of the objects
3193: they point to (and most of the contents are other such pointers).
3194:
3195: In addition, most compiler passes consist of one or more loops that
3196: scan the RTL insn sequence. The most vital piece of information about
3197: such a loop---which insn it has reached---is usually in a local variable,
3198: not in an argument.
3199:
3200: @findex debug_rtx
3201: What you need to provide in addition to a backtrace are the values of
3202: the local variables for several stack frames up. When a local
3203: variable or an argument is an RTX, first print its value and then use
3204: the GDB command @code{pr} to print the RTL expression that it points
3205: to. (If GDB doesn't run on your machine, use your debugger to call
3206: the function @code{debug_rtx} with the RTX as an argument.) In
3207: general, whenever a variable is a pointer, its value is no use
3208: without the data it points to.
3209: @end itemize
3210:
3211: Here are some things that are not necessary:
3212:
3213: @itemize @bullet
3214: @item
3215: A description of the envelope of the bug.
3216:
3217: Often people who encounter a bug spend a lot of time investigating
3218: which changes to the input file will make the bug go away and which
3219: changes will not affect it.
3220:
1.1.1.3 root 3221: This is often time consuming and not very useful, because the way we
3222: will find the bug is by running a single example under the debugger with
3223: breakpoints, not by pure deduction from a series of examples. You might
3224: as well save your time for something else.
3225:
3226: Of course, if you can find a simpler example to report @emph{instead} of
3227: the original one, that is a convenience. Errors in the output will be
3228: easier to spot, running under the debugger will take less time, etc.
3229: Most GNU CC bugs involve just one function, so the most straightforward
3230: way to simplify an example is to delete all the function definitions
3231: except the one where the bug occurs. Those earlier in the file may be
3232: replaced by external declarations if the crucial function depends on
3233: them. (Exception: inline functions may affect compilation of functions
3234: defined later in the file.)
1.1 root 3235:
3236: However, simplification is not vital; if you don't want to do this,
1.1.1.3 root 3237: report the bug anyway and send the entire test case you used.
1.1 root 3238:
3239: @item
1.1.1.4 root 3240: In particular, some people insert conditionals @samp{#ifdef BUG} around
3241: a statement which, if removed, makes the bug not happen. These are just
3242: clutter; we won't pay any attention to them anyway. Besides, you should
3243: send us cpp output, and that can't have conditionals.
3244:
3245: @item
1.1 root 3246: A patch for the bug.
3247:
1.1.1.3 root 3248: A patch for the bug is useful if it is a good one. But don't omit the
3249: necessary information, such as the test case, on the assumption that a
3250: patch is all we need. We might see problems with your patch and decide
3251: to fix the problem another way, or we might not understand it at all.
1.1 root 3252:
3253: Sometimes with a program as complicated as GNU CC it is very hard to
3254: construct an example that will make the program follow a certain path
1.1.1.3 root 3255: through the code. If you don't send the example, we won't be able to
3256: construct one, so we won't be able to verify that the bug is fixed.
3257:
3258: And if we can't understand what bug you are trying to fix, or why your
3259: patch should be an improvement, we won't install it. A test case will
3260: help us to understand.
1.1 root 3261:
1.1.1.3 root 3262: @xref{Sending Patches}, for guidelines on how to make it easy for us to
3263: understand and install your patches.
1.1 root 3264:
3265: @item
3266: A guess about what the bug is or what it depends on.
3267:
3268: Such guesses are usually wrong. Even I can't guess right about such
3269: things without first using the debugger to find the facts.
1.1.1.4 root 3270:
3271: @item
3272: A core dump file.
3273:
3274: We have no way of examining a core dump for your type of machine
3275: unless we have an identical system---and if we do have one,
3276: we should be able to reproduce the crash ourselves.
1.1 root 3277: @end itemize
3278:
1.1.1.3 root 3279: @node Sending Patches,, Bug Reporting, Bugs
3280: @section Sending Patches for GNU CC
1.1 root 3281:
1.1.1.3 root 3282: If you would like to write bug fixes or improvements for the GNU C
1.1.1.8 root 3283: compiler, that is very helpful. Send suggested fixes to the bug report
3284: mailing list, @code{bug-gcc@@prep.ai.mit.edu}.
1.1.1.3 root 3285:
1.1.1.8 root 3286: Please follow these guidelines so we can study your patches efficiently.
1.1.1.3 root 3287: If you don't follow these guidelines, your information might still be
3288: useful, but using it will take extra work. Maintaining GNU C is a lot
3289: of work in the best of circumstances, and we can't keep up unless you do
3290: your best to help.
1.1 root 3291:
3292: @itemize @bullet
3293: @item
1.1.1.3 root 3294: Send an explanation with your changes of what problem they fix or what
3295: improvement they bring about. For a bug fix, just include a copy of the
3296: bug report, and explain why the change fixes the bug.
1.1 root 3297:
1.1.1.3 root 3298: (Referring to a bug report is not as good as including it, because then
3299: we will have to look it up, and we have probably already deleted it if
3300: we've already fixed the bug.)
1.1 root 3301:
3302: @item
1.1.1.3 root 3303: Always include a proper bug report for the problem you think you have
3304: fixed. We need to convince ourselves that the change is right before
3305: installing it. Even if it is right, we might have trouble judging it if
3306: we don't have a way to reproduce the problem.
1.1.1.2 root 3307:
1.1.1.3 root 3308: @item
3309: Include all the comments that are appropriate to help people reading the
3310: source in the future understand why this change was needed.
1.1.1.2 root 3311:
3312: @item
1.1.1.3 root 3313: Don't mix together changes made for different reasons.
3314: Send them @emph{individually}.
1.1 root 3315:
1.1.1.3 root 3316: If you make two changes for separate reasons, then we might not want to
3317: install them both. We might want to install just one. If you send them
3318: all jumbled together in a single set of diffs, we have to do extra work
3319: to disentangle them---to figure out which parts of the change serve
3320: which purpose. If we don't have time for this, we might have to ignore
3321: your changes entirely.
1.1 root 3322:
1.1.1.3 root 3323: If you send each change as soon as you have written it, with its own
3324: explanation, then the two changes never get tangled up, and we can
3325: consider each one properly without any extra work to disentangle them.
1.1 root 3326:
1.1.1.3 root 3327: Ideally, each change you send should be impossible to subdivide into
3328: parts that we might want to consider separately, because each of its
3329: parts gets its motivation from the other parts.
1.1 root 3330:
1.1.1.3 root 3331: @item
3332: Send each change as soon as that change is finished. Sometimes people
3333: think they are helping us by accumulating many changes to send them all
3334: together. As explained above, this is absolutely the worst thing you
3335: could do.
1.1 root 3336:
1.1.1.3 root 3337: Since you should send each change separately, you might as well send it
3338: right away. That gives us the option of installing it immediately if it
3339: is important.
1.1 root 3340:
1.1.1.3 root 3341: @item
3342: Use @samp{diff -c} to make your diffs. Diffs without context are hard
3343: for us to install reliably. More than that, they make it hard for us to
3344: study the diffs to decide whether we want to install them. Unidiff
3345: format is better than contextless diffs, but not as easy to read as
3346: @samp{-c} format.
1.1 root 3347:
1.1.1.3 root 3348: If you have GNU diff, use @samp{diff -cp}, which shows the name of the
3349: function that each change occurs in.
1.1 root 3350:
1.1.1.3 root 3351: @item
3352: Write the change log entries for your changes. We get lots of changes,
3353: and we don't have time to do all the change log writing ourselves.
1.1 root 3354:
1.1.1.3 root 3355: Read the @file{ChangeLog} file to see what sorts of information to put
3356: in, and to learn the style that we use. The purpose of the change log
3357: is to show people where to find what was changed. So you need to be
3358: specific about what functions you changed; in large functions, it's
3359: often helpful to indicate where within the function the change was.
1.1.1.2 root 3360:
1.1.1.3 root 3361: On the other hand, once you have shown people where to find the change,
1.1.1.5 root 3362: you need not explain its purpose. Thus, if you add a new function, all
1.1.1.3 root 3363: you need to say about it is that it is new. If you feel that the
3364: purpose needs explaining, it probably does---but the explanation will be
3365: much more useful if you put it in comments in the code.
3366:
3367: If you would like your name to appear in the header line for who made
3368: the change, send us the header line.
1.1 root 3369:
3370: @item
1.1.1.4 root 3371: When you write the fix, keep in mind that we can't install a change that
1.1.1.3 root 3372: would break other systems.
1.1 root 3373:
1.1.1.3 root 3374: People often suggest fixing a problem by changing machine-independent
3375: files such as @file{toplev.c} to do something special that a particular
3376: system needs. Sometimes it is totally obvious that such changes would
3377: break GNU CC for almost all users. We can't possibly make a change like
3378: that. At best it might tell us how to write another patch that would
3379: solve the problem acceptably.
1.1 root 3380:
1.1.1.3 root 3381: Sometimes people send fixes that @emph{might} be an improvement in
3382: general---but it is hard to be sure of this. It's hard to install
3383: such changes because we have to study them very carefully. Of course,
3384: a good explanation of the reasoning by which you concluded the change
3385: was correct can help convince us.
1.1 root 3386:
1.1.1.3 root 3387: The safest changes are changes to the configuration files for a
3388: particular machine. These are safe because they can't create new bugs
3389: on other machines.
1.1 root 3390:
1.1.1.3 root 3391: Please help us keep up with the workload by designing the patch in a
3392: form that is good to install.
3393: @end itemize
1.1 root 3394:
1.1.1.3 root 3395: @node Service
3396: @chapter How To Get Help with GNU CC
1.1 root 3397:
1.1.1.3 root 3398: If you need help installing, using or changing GNU CC, there are two
3399: ways to find it:
1.1 root 3400:
1.1.1.3 root 3401: @itemize @bullet
3402: @item
3403: Send a message to a suitable network mailing list. First try
3404: @code{bug-gcc@@prep.ai.mit.edu}, and if that brings no response, try
3405: @code{help-gcc@@prep.ai.mit.edu}.
1.1 root 3406:
1.1.1.3 root 3407: @item
3408: Look in the service directory for someone who might help you for a fee.
3409: The service directory is found in the file named @file{SERVICE} in the
3410: GNU CC distribution.
1.1 root 3411: @end itemize
3412:
1.1.1.3 root 3413: @node VMS
1.1 root 3414: @chapter Using GNU CC on VMS
3415:
1.1.1.7 root 3416: @c prevent bad page break with this line
3417: Here is how to use GNU CC on VMS.
3418:
1.1 root 3419: @menu
3420: * Include Files and VMS:: Where the preprocessor looks for the include files.
3421: * Global Declarations:: How to do globaldef, globalref and globalvalue with
3422: GNU CC.
3423: * VMS Misc:: Misc information.
3424: @end menu
3425:
1.1.1.3 root 3426: @node Include Files and VMS
1.1 root 3427: @section Include Files and VMS
3428:
3429: @cindex include files and VMS
3430: @cindex VMS and include files
3431: @cindex header files and VMS
3432: Due to the differences between the filesystems of Unix and VMS, GNU CC
3433: attempts to translate file names in @samp{#include} into names that VMS
3434: will understand. The basic strategy is to prepend a prefix to the
3435: specification of the include file, convert the whole filename to a VMS
3436: filename, and then try to open the file. GNU CC tries various prefixes
3437: one by one until one of them succeeds:
3438:
3439: @enumerate
3440: @item
3441: The first prefix is the @samp{GNU_CC_INCLUDE:} logical name: this is
3442: where GNU C header files are traditionally stored. If you wish to store
3443: header files in non-standard locations, then you can assign the logical
3444: @samp{GNU_CC_INCLUDE} to be a search list, where each element of the
3445: list is suitable for use with a rooted logical.
3446:
3447: @item
3448: The next prefix tried is @samp{SYS$SYSROOT:[SYSLIB.]}. This is where
3449: VAX-C header files are traditionally stored.
3450:
3451: @item
3452: If the include file specification by itself is a valid VMS filename, the
3453: preprocessor then uses this name with no prefix in an attempt to open
3454: the include file.
3455:
3456: @item
3457: If the file specification is not a valid VMS filename (i.e. does not
3458: contain a device or a directory specifier, and contains a @samp{/}
3459: character), the preprocessor tries to convert it from Unix syntax to
3460: VMS syntax.
3461:
3462: Conversion works like this: the first directory name becomes a device,
3463: and the rest of the directories are converted into VMS-format directory
1.1.1.5 root 3464: names. For example, the name @file{X11/foobar.h} is
3465: translated to @file{X11:[000000]foobar.h} or @file{X11:foobar.h},
3466: whichever one can be opened. This strategy allows you to assign a
3467: logical name to point to the actual location of the header files.
1.1 root 3468:
3469: @item
3470: If none of these strategies succeeds, the @samp{#include} fails.
3471: @end enumerate
3472:
3473: Include directives of the form:
3474:
3475: @example
3476: #include foobar
3477: @end example
3478:
3479: @noindent
3480: are a common source of incompatibility between VAX-C and GNU CC. VAX-C
3481: treats this much like a standard @code{#include <foobar.h>} directive.
3482: That is incompatible with the ANSI C behavior implemented by GNU CC: to
3483: expand the name @code{foobar} as a macro. Macro expansion should
3484: eventually yield one of the two standard formats for @code{#include}:
3485:
3486: @example
3487: #include "@var{file}"
3488: #include <@var{file}>
3489: @end example
3490:
3491: If you have this problem, the best solution is to modify the source to
3492: convert the @code{#include} directives to one of the two standard forms.
3493: That will work with either compiler. If you want a quick and dirty fix,
3494: define the file names as macros with the proper expansion, like this:
3495:
3496: @example
3497: #define stdio <stdio.h>
3498: @end example
3499:
3500: @noindent
3501: This will work, as long as the name doesn't conflict with anything else
3502: in the program.
3503:
3504: Another source of incompatibility is that VAX-C assumes that:
3505:
3506: @example
3507: #include "foobar"
3508: @end example
3509:
3510: @noindent
3511: is actually asking for the file @file{foobar.h}. GNU CC does not
3512: make this assumption, and instead takes what you ask for literally;
3513: it tries to read the file @file{foobar}. The best way to avoid this
3514: problem is to always specify the desired file extension in your include
3515: directives.
3516:
3517: GNU CC for VMS is distributed with a set of include files that is
3518: sufficient to compile most general purpose programs. Even though the
3519: GNU CC distribution does not contain header files to define constants
3520: and structures for some VMS system-specific functions, there is no
3521: reason why you cannot use GNU CC with any of these functions. You first
3522: may have to generate or create header files, either by using the public
3523: domain utility @code{UNSDL} (which can be found on a DECUS tape), or by
3524: extracting the relevant modules from one of the system macro libraries,
3525: and using an editor to construct a C header file.
3526:
1.1.1.4 root 3527: A @code{#include} file name cannot contain a DECNET node name. The
3528: preprocessor reports an I/O error if you attempt to use a node name,
3529: whether explicitly, or implicitly via a logical name.
3530:
1.1.1.3 root 3531: @node Global Declarations
1.1 root 3532: @section Global Declarations and VMS
3533:
3534: @findex GLOBALREF
3535: @findex GLOBALDEF
3536: @findex GLOBALVALUEDEF
3537: @findex GLOBALVALUEREF
3538: GNU CC does not provide the @code{globalref}, @code{globaldef} and
3539: @code{globalvalue} keywords of VAX-C. You can get the same effect with
3540: an obscure feature of GAS, the GNU assembler. (This requires GAS
3541: version 1.39 or later.) The following macros allow you to use this
3542: feature in a fairly natural way:
3543:
3544: @smallexample
3545: #ifdef __GNUC__
1.1.1.2 root 3546: #define GLOBALREF(TYPE,NAME) \
3547: TYPE NAME \
3548: asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME)
3549: #define GLOBALDEF(TYPE,NAME,VALUE) \
3550: TYPE NAME \
3551: asm ("_$$PsectAttributes_GLOBALSYMBOL$$" #NAME) \
3552: = VALUE
3553: #define GLOBALVALUEREF(TYPE,NAME) \
3554: const TYPE NAME[1] \
3555: asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME)
3556: #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
3557: const TYPE NAME[1] \
3558: asm ("_$$PsectAttributes_GLOBALVALUE$$" #NAME) \
3559: = @{VALUE@}
1.1 root 3560: #else
1.1.1.2 root 3561: #define GLOBALREF(TYPE,NAME) \
3562: globalref TYPE NAME
3563: #define GLOBALDEF(TYPE,NAME,VALUE) \
3564: globaldef TYPE NAME = VALUE
3565: #define GLOBALVALUEDEF(TYPE,NAME,VALUE) \
3566: globalvalue TYPE NAME = VALUE
3567: #define GLOBALVALUEREF(TYPE,NAME) \
3568: globalvalue TYPE NAME
1.1 root 3569: #endif
3570: @end smallexample
3571:
3572: @noindent
3573: (The @code{_$$PsectAttributes_GLOBALSYMBOL} prefix at the start of the
3574: name is removed by the assembler, after it has modified the attributes
3575: of the symbol). These macros are provided in the VMS binaries
3576: distribution in a header file @file{GNU_HACKS.H}. An example of the
3577: usage is:
3578:
3579: @example
1.1.1.2 root 3580: GLOBALREF (int, ijk);
3581: GLOBALDEF (int, jkl, 0);
1.1 root 3582: @end example
3583:
3584: The macros @code{GLOBALREF} and @code{GLOBALDEF} cannot be used
3585: straightforwardly for arrays, since there is no way to insert the array
3586: dimension into the declaration at the right place. However, you can
3587: declare an array with these macros if you first define a typedef for the
3588: array type, like this:
3589:
3590: @example
3591: typedef int intvector[10];
1.1.1.2 root 3592: GLOBALREF (intvector, foo);
1.1 root 3593: @end example
3594:
3595: Array and structure initializers will also break the macros; you can
3596: define the initializer to be a macro of its own, or you can expand the
3597: @code{GLOBALDEF} macro by hand. You may find a case where you wish to
3598: use the @code{GLOBALDEF} macro with a large array, but you are not
3599: interested in explicitly initializing each element of the array. In
3600: such cases you can use an initializer like: @code{@{0,@}}, which will
3601: initialize the entire array to @code{0}.
3602:
3603: A shortcoming of this implementation is that a variable declared with
3604: @code{GLOBALVALUEREF} or @code{GLOBALVALUEDEF} is always an array. For
3605: example, the declaration:
3606:
3607: @example
1.1.1.2 root 3608: GLOBALVALUEREF(int, ijk);
1.1 root 3609: @end example
3610:
3611: @noindent
3612: declares the variable @code{ijk} as an array of type @code{int [1]}.
3613: This is done because a globalvalue is actually a constant; its ``value''
3614: is what the linker would normally consider an address. That is not how
3615: an integer value works in C, but it is how an array works. So treating
3616: the symbol as an array name gives consistent results---with the
3617: exception that the value seems to have the wrong type. @strong{Don't
3618: try to access an element of the array.} It doesn't have any elements.
3619: The array ``address'' may not be the address of actual storage.
3620:
3621: The fact that the symbol is an array may lead to warnings where the
3622: variable is used. Insert type casts to avoid the warnings. Here is an
3623: example; it takes advantage of the ANSI C feature allowing macros that
3624: expand to use the same name as the macro itself.
3625:
3626: @example
1.1.1.2 root 3627: GLOBALVALUEREF (int, ss$_normal);
3628: GLOBALVALUEDEF (int, xyzzy,123);
1.1 root 3629: #ifdef __GNUC__
3630: #define ss$_normal ((int) ss$_normal)
3631: #define xyzzy ((int) xyzzy)
3632: #endif
3633: @end example
3634:
3635: Don't use @code{globaldef} or @code{globalref} with a variable whose
3636: type is an enumeration type; this is not implemented. Instead, make the
3637: variable an integer, and use a @code{globalvaluedef} for each of the
3638: enumeration values. An example of this would be:
3639:
3640: @example
3641: #ifdef __GNUC__
1.1.1.2 root 3642: GLOBALDEF (int, color, 0);
3643: GLOBALVALUEDEF (int, RED, 0);
3644: GLOBALVALUEDEF (int, BLUE, 1);
3645: GLOBALVALUEDEF (int, GREEN, 3);
1.1 root 3646: #else
3647: enum globaldef color @{RED, BLUE, GREEN = 3@};
3648: #endif
3649: @end example
3650:
1.1.1.3 root 3651: @node VMS Misc
1.1 root 3652: @section Other VMS Issues
3653:
3654: @cindex exit status and VMS
3655: @cindex return value of @code{main}
3656: @cindex @code{main} and the exit status
3657: GNU CC automatically arranges for @code{main} to return 1 by default if
3658: you fail to specify an explicit return value. This will be interpreted
3659: by VMS as a status code indicating a normal successful completion.
3660: Version 1 of GNU CC did not provide this default.
3661:
3662: GNU CC on VMS works only with the GNU assembler, GAS. You need version
3663: 1.37 or later of GAS in order to produce value debugging information for
3664: the VMS debugger. Use the ordinary VMS linker with the object files
3665: produced by GAS.
3666:
3667: @cindex shared VMS run time system
3668: @cindex @file{VAXCRTL}
3669: Under previous versions of GNU CC, the generated code would occasionally
3670: give strange results when linked to the sharable @file{VAXCRTL} library.
3671: Now this should work.
3672:
3673: A caveat for use of @code{const} global variables: the @code{const}
3674: modifier must be specified in every external declaration of the variable
3675: in all of the source files that use that variable. Otherwise the linker
3676: will issue warnings about conflicting attributes for the variable. Your
3677: program will still work despite the warnings, but the variable will be
3678: placed in writable storage.
3679:
3680: @cindex name augmentation
3681: @cindex case sensitivity and VMS
3682: @cindex VMS and case sensitivity
1.1.1.4 root 3683: Although the VMS linker does distinguish between upper and lower case
3684: letters in global symbols, most VMS compilers convert all such symbols
3685: into upper case and most run-time library routines also have upper case
3686: names. To be able to reliably call such routines, GNU CC (by means of
3687: the assembler GAS) converts global symbols into upper case like other
3688: VMS compilers. However, since the usual practice in C is to distinguish
3689: case, GNU CC (via GAS) tries to preserve usual C behavior by augmenting
3690: each name that is not all lower case. This means truncating the name
3691: to at most 23 characters and then adding more characters at the end
3692: which encode the case pattern of those 23. Names which contain at
3693: least one dollar sign are an exception; they are converted directly into
3694: upper case without augmentation.
1.1 root 3695:
3696: Name augmentation yields bad results for programs that use precompiled
3697: libraries (such as Xlib) which were generated by another compiler. You
3698: can use the compiler option @samp{/NOCASE_HACK} to inhibit augmentation;
3699: it makes external C functions and variables case-independent as is usual
3700: on VMS. Alternatively, you could write all references to the functions
3701: and variables in such libraries using lower case; this will work on VMS,
1.1.1.4 root 3702: but is not portable to other systems. The compiler option @samp{/NAMES}
3703: also provides control over global name handling.
1.1 root 3704:
3705: Function and variable names are handled somewhat differently with GNU
3706: C++. The GNU C++ compiler performs @dfn{name mangling} on function
3707: names, which means that it adds information to the function name to
1.1.1.5 root 3708: describe the data types of the arguments that the function takes. One
1.1 root 3709: result of this is that the name of a function can become very long.
3710: Since the VMS linker only recognizes the first 31 characters in a name,
3711: special action is taken to ensure that each function and variable has a
3712: unique name that can be represented in 31 characters.
3713:
3714: If the name (plus a name augmentation, if required) is less than 32
1.1.1.5 root 3715: characters in length, then no special action is performed. If the name
1.1 root 3716: is longer than 31 characters, the assembler (GAS) will generate a
3717: hash string based upon the function name, truncate the function name to
3718: 23 characters, and append the hash string to the truncated name. If the
3719: @samp{/VERBOSE} compiler option is used, the assembler will print both
3720: the full and truncated names of each symbol that is truncated.
3721:
3722: The @samp{/NOCASE_HACK} compiler option should not be used when you are
1.1.1.5 root 3723: compiling programs that use libg++. libg++ has several instances of
1.1 root 3724: objects (i.e. @code{Filebuf} and @code{filebuf}) which become
3725: indistinguishable in a case-insensitive environment. This leads to
3726: cases where you need to inhibit augmentation selectively (if you were
3727: using libg++ and Xlib in the same program, for example). There is no
3728: special feature for doing this, but you can get the result by defining a
3729: macro for each mixed case symbol for which you wish to inhibit
3730: augmentation. The macro should expand into the lower case equivalent of
3731: itself. For example:
3732:
3733: @example
3734: #define StuDlyCapS studlycaps
3735: @end example
3736:
3737: These macro definitions can be placed in a header file to minimize the
3738: number of changes to your source code.
1.1.1.5 root 3739: @end ifset
1.1 root 3740:
3741: @ifset INTERNALS
1.1.1.3 root 3742: @node Portability
1.1 root 3743: @chapter GNU CC and Portability
3744: @cindex portability
3745: @cindex GNU CC and portability
3746:
3747: The main goal of GNU CC was to make a good, fast compiler for machines in
3748: the class that the GNU system aims to run on: 32-bit machines that address
3749: 8-bit bytes and have several general registers. Elegance, theoretical
3750: power and simplicity are only secondary.
3751:
3752: GNU CC gets most of the information about the target machine from a machine
3753: description which gives an algebraic formula for each of the machine's
3754: instructions. This is a very clean way to describe the target. But when
3755: the compiler needs information that is difficult to express in this
3756: fashion, I have not hesitated to define an ad-hoc parameter to the machine
3757: description. The purpose of portability is to reduce the total work needed
3758: on the compiler; it was not of interest for its own sake.
3759:
3760: @cindex endianness
3761: @cindex autoincrement addressing, availability
3762: @findex abort
3763: GNU CC does not contain machine dependent code, but it does contain code
3764: that depends on machine parameters such as endianness (whether the most
3765: significant byte has the highest or lowest address of the bytes in a word)
3766: and the availability of autoincrement addressing. In the RTL-generation
3767: pass, it is often necessary to have multiple strategies for generating code
3768: for a particular kind of syntax tree, strategies that are usable for different
3769: combinations of parameters. Often I have not tried to address all possible
3770: cases, but only the common ones or only the ones that I have encountered.
3771: As a result, a new target may require additional strategies. You will know
3772: if this happens because the compiler will call @code{abort}. Fortunately,
3773: the new strategies can be added in a machine-independent fashion, and will
3774: affect only the target machines that need them.
3775: @end ifset
3776:
3777: @ifset INTERNALS
1.1.1.3 root 3778: @node Interface
1.1 root 3779: @chapter Interfacing to GNU CC Output
3780: @cindex interfacing to GNU CC output
3781: @cindex run-time conventions
3782: @cindex function call conventions
3783: @cindex conventions, run-time
3784:
3785: GNU CC is normally configured to use the same function calling convention
3786: normally in use on the target system. This is done with the
1.1.1.2 root 3787: machine-description macros described (@pxref{Target Macros}).
1.1 root 3788:
3789: @cindex unions, returning
3790: @cindex structures, returning
3791: @cindex returning structures and unions
3792: However, returning of structure and union values is done differently on
3793: some target machines. As a result, functions compiled with PCC
3794: returning such types cannot be called from code compiled with GNU CC,
3795: and vice versa. This does not cause trouble often because few Unix
3796: library routines return structures or unions.
3797:
3798: GNU CC code returns structures and unions that are 1, 2, 4 or 8 bytes
3799: long in the same registers used for @code{int} or @code{double} return
3800: values. (GNU CC typically allocates variables of such types in
3801: registers also.) Structures and unions of other sizes are returned by
3802: storing them into an address passed by the caller (usually in a
3803: register). The machine-description macros @code{STRUCT_VALUE} and
3804: @code{STRUCT_INCOMING_VALUE} tell GNU CC where to pass this address.
3805:
3806: By contrast, PCC on most target machines returns structures and unions
3807: of any size by copying the data into an area of static storage, and then
3808: returning the address of that storage as if it were a pointer value.
3809: The caller must copy the data from that memory area to the place where
3810: the value is wanted. This is slower than the method used by GNU CC, and
3811: fails to be reentrant.
3812:
3813: On some target machines, such as RISC machines and the 80386, the
3814: standard system convention is to pass to the subroutine the address of
3815: where to return the value. On these machines, GNU CC has been
3816: configured to be compatible with the standard compiler, when this method
3817: is used. It may not be compatible for structures of 1, 2, 4 or 8 bytes.
3818:
3819: @cindex argument passing
3820: @cindex passing arguments
3821: GNU CC uses the system's standard convention for passing arguments. On
3822: some machines, the first few arguments are passed in registers; in
3823: others, all are passed on the stack. It would be possible to use
3824: registers for argument passing on any machine, and this would probably
3825: result in a significant speedup. But the result would be complete
3826: incompatibility with code that follows the standard convention. So this
3827: change is practical only if you are switching to GNU CC as the sole C
3828: compiler for the system. We may implement register argument passing on
3829: certain machines once we have a complete GNU system so that we can
3830: compile the libraries with GNU CC.
3831:
3832: On some machines (particularly the Sparc), certain types of arguments
3833: are passed ``by invisible reference''. This means that the value is
3834: stored in memory, and the address of the memory location is passed to
3835: the subroutine.
3836:
3837: @cindex @code{longjmp} and automatic variables
3838: If you use @code{longjmp}, beware of automatic variables. ANSI C says that
3839: automatic variables that are not declared @code{volatile} have undefined
3840: values after a @code{longjmp}. And this is all GNU CC promises to do,
3841: because it is very difficult to restore register variables correctly, and
3842: one of GNU CC's features is that it can put variables in registers without
3843: your asking it to.
3844:
3845: If you want a variable to be unaltered by @code{longjmp}, and you don't
3846: want to write @code{volatile} because old C compilers don't accept it,
3847: just take the address of the variable. If a variable's address is ever
3848: taken, even if just to compute it and ignore it, then the variable cannot
3849: go in a register:
3850:
3851: @example
3852: @{
3853: int careful;
3854: &careful;
3855: @dots{}
3856: @}
3857: @end example
3858:
3859: @cindex arithmetic libraries
3860: @cindex math libraries
3861: Code compiled with GNU CC may call certain library routines. Most of
3862: them handle arithmetic for which there are no instructions. This
3863: includes multiply and divide on some machines, and floating point
3864: operations on any machine for which floating point support is disabled
3865: with @samp{-msoft-float}. Some standard parts of the C library, such as
3866: @code{bcopy} or @code{memcpy}, are also called automatically. The usual
3867: function call interface is used for calling the library routines.
3868:
3869: These library routines should be defined in the library @file{libgcc.a},
3870: which GNU CC automatically searches whenever it links a program. On
3871: machines that have multiply and divide instructions, if hardware
3872: floating point is in use, normally @file{libgcc.a} is not needed, but it
3873: is searched just in case.
3874:
3875: Each arithmetic function is defined in @file{libgcc1.c} to use the
3876: corresponding C arithmetic operator. As long as the file is compiled
3877: with another C compiler, which supports all the C arithmetic operators,
3878: this file will work portably. However, @file{libgcc1.c} does not work if
3879: compiled with GNU CC, because each arithmetic function would compile
3880: into a call to itself!
3881: @end ifset
3882:
3883: @ifset INTERNALS
1.1.1.3 root 3884: @node Passes
1.1 root 3885: @chapter Passes and Files of the Compiler
3886: @cindex passes and files of the compiler
3887: @cindex files and passes of the compiler
3888: @cindex compiler passes and files
3889:
3890: @cindex top level of compiler
3891: The overall control structure of the compiler is in @file{toplev.c}. This
3892: file is responsible for initialization, decoding arguments, opening and
3893: closing files, and sequencing the passes.
3894:
3895: @cindex parsing pass
3896: The parsing pass is invoked only once, to parse the entire input. The RTL
3897: intermediate code for a function is generated as the function is parsed, a
3898: statement at a time. Each statement is read in as a syntax tree and then
3899: converted to RTL; then the storage for the tree for the statement is
3900: reclaimed. Storage for types (and the expressions for their sizes),
3901: declarations, and a representation of the binding contours and how they nest,
3902: remain until the function is finished being compiled; these are all needed
3903: to output the debugging information.
3904:
3905: @findex rest_of_compilation
3906: @findex rest_of_decl_compilation
3907: Each time the parsing pass reads a complete function definition or
1.1.1.5 root 3908: top-level declaration, it calls either the function
3909: @code{rest_of_compilation}, or the function
3910: @code{rest_of_decl_compilation} in @file{toplev.c}, which are
3911: responsible for all further processing necessary, ending with output of
3912: the assembler language. All other compiler passes run, in sequence,
3913: within @code{rest_of_compilation}. When that function returns from
3914: compiling a function definition, the storage used for that function
3915: definition's compilation is entirely freed, unless it is an inline
3916: function
3917: @ifset USING
3918: (@pxref{Inline,,An Inline Function is As Fast As a Macro}).
3919: @end ifset
3920: @ifclear USING
3921: (@pxref{Inline,,An Inline Function is As Fast As a Macro,gcc.texi,Using GCC}).
3922: @end ifclear
1.1 root 3923:
3924: Here is a list of all the passes of the compiler and their source files.
3925: Also included is a description of where debugging dumps can be requested
3926: with @samp{-d} options.
3927:
3928: @itemize @bullet
3929: @item
3930: Parsing. This pass reads the entire text of a function definition,
3931: constructing partial syntax trees. This and RTL generation are no longer
3932: truly separate passes (formerly they were), but it is easier to think
3933: of them as separate.
3934:
3935: The tree representation does not entirely follow C syntax, because it is
3936: intended to support other languages as well.
3937:
3938: Language-specific data type analysis is also done in this pass, and every
3939: tree node that represents an expression has a data type attached.
3940: Variables are represented as declaration nodes.
3941:
3942: @cindex constant folding
3943: @cindex arithmetic simplifications
3944: @cindex simplifications, arithmetic
3945: Constant folding and some arithmetic simplifications are also done
3946: during this pass.
3947:
3948: The language-independent source files for parsing are
3949: @file{stor-layout.c}, @file{fold-const.c}, and @file{tree.c}.
3950: There are also header files @file{tree.h} and @file{tree.def}
3951: which define the format of the tree representation.@refill
3952:
1.1.1.5 root 3953: @c Avoiding overfull is tricky here.
3954: The source files to parse C are
3955: @file{c-parse.in},
3956: @file{c-decl.c},
3957: @file{c-typeck.c},
3958: @file{c-aux-info.c},
3959: @file{c-convert.c},
3960: and @file{c-lang.c}
3961: along with header files
3962: @file{c-lex.h}, and
1.1 root 3963: @file{c-tree.h}.
3964:
3965: The source files for parsing C++ are @file{cp-parse.y},
1.1.1.4 root 3966: @file{cp-class.c},@*
3967: @file{cp-cvt.c}, @file{cp-decl.c}, @file{cp-decl2.c},
1.1 root 3968: @file{cp-dem.c}, @file{cp-except.c},@*
3969: @file{cp-expr.c}, @file{cp-init.c}, @file{cp-lex.c},
3970: @file{cp-method.c}, @file{cp-ptree.c},@*
3971: @file{cp-search.c}, @file{cp-tree.c}, @file{cp-type2.c}, and
3972: @file{cp-typeck.c}, along with header files @file{cp-tree.def},
3973: @file{cp-tree.h}, and @file{cp-decl.h}.
3974:
3975: The special source files for parsing Objective C are
3976: @file{objc-parse.y}, @file{objc-actions.c}, @file{objc-tree.def}, and
3977: @file{objc-actions.h}. Certain C-specific files are used for this as
3978: well.
3979:
3980: The file @file{c-common.c} is also used for all of the above languages.
3981:
3982: @cindex RTL generation
3983: @item
3984: RTL generation. This is the conversion of syntax tree into RTL code.
3985: It is actually done statement-by-statement during parsing, but for
3986: most purposes it can be thought of as a separate pass.
3987:
3988: @cindex target-parameter-dependent code
3989: This is where the bulk of target-parameter-dependent code is found,
3990: since often it is necessary for strategies to apply only when certain
3991: standard kinds of instructions are available. The purpose of named
3992: instruction patterns is to provide this information to the RTL
3993: generation pass.
3994:
3995: @cindex tail recursion optimization
3996: Optimization is done in this pass for @code{if}-conditions that are
3997: comparisons, boolean operations or conditional expressions. Tail
3998: recursion is detected at this time also. Decisions are made about how
3999: best to arrange loops and how to output @code{switch} statements.
4000:
1.1.1.5 root 4001: @c Avoiding overfull is tricky here.
4002: The source files for RTL generation include
4003: @file{stmt.c},
4004: @file{calls.c},
4005: @file{expr.c},
4006: @file{explow.c},
4007: @file{expmed.c},
4008: @file{function.c},
4009: @file{optabs.c}
4010: and @file{emit-rtl.c}.
4011: Also, the file
1.1 root 4012: @file{insn-emit.c}, generated from the machine description by the
4013: program @code{genemit}, is used in this pass. The header file
4014: @file{expr.h} is used for communication within this pass.@refill
4015:
4016: @findex genflags
4017: @findex gencodes
4018: The header files @file{insn-flags.h} and @file{insn-codes.h},
4019: generated from the machine description by the programs @code{genflags}
4020: and @code{gencodes}, tell this pass which standard names are available
4021: for use and which patterns correspond to them.@refill
4022:
4023: Aside from debugging information output, none of the following passes
4024: refers to the tree structure representation of the function (only
4025: part of which is saved).
4026:
4027: @cindex inline, automatic
4028: The decision of whether the function can and should be expanded inline
4029: in its subsequent callers is made at the end of rtl generation. The
4030: function must meet certain criteria, currently related to the size of
4031: the function and the types and number of parameters it has. Note that
4032: this function may contain loops, recursive calls to itself
4033: (tail-recursive functions can be inlined!), gotos, in short, all
4034: constructs supported by GNU CC. The file @file{integrate.c} contains
4035: the code to save a function's rtl for later inlining and to inline that
4036: rtl when the function is called. The header file @file{integrate.h}
4037: is also used for this purpose.
4038:
4039: The option @samp{-dr} causes a debugging dump of the RTL code after
4040: this pass. This dump file's name is made by appending @samp{.rtl} to
4041: the input file name.
4042:
4043: @cindex jump optimization
4044: @cindex unreachable code
4045: @cindex dead code
4046: @item
4047: Jump optimization. This pass simplifies jumps to the following
4048: instruction, jumps across jumps, and jumps to jumps. It deletes
4049: unreferenced labels and unreachable code, except that unreachable code
4050: that contains a loop is not recognized as unreachable in this pass.
4051: (Such loops are deleted later in the basic block analysis.) It also
4052: converts some code originally written with jumps into sequences of
4053: instructions that directly set values from the results of comparisons,
4054: if the machine has such instructions.
4055:
4056: Jump optimization is performed two or three times. The first time is
4057: immediately following RTL generation. The second time is after CSE,
4058: but only if CSE says repeated jump optimization is needed. The
4059: last time is right before the final pass. That time, cross-jumping
4060: and deletion of no-op move instructions are done together with the
4061: optimizations described above.
4062:
4063: The source file of this pass is @file{jump.c}.
4064:
4065: The option @samp{-dj} causes a debugging dump of the RTL code after
4066: this pass is run for the first time. This dump file's name is made by
4067: appending @samp{.jump} to the input file name.
4068:
4069: @cindex register use analysis
4070: @item
4071: Register scan. This pass finds the first and last use of each
4072: register, as a guide for common subexpression elimination. Its source
4073: is in @file{regclass.c}.
4074:
4075: @cindex jump threading
4076: @item
4077: Jump threading. This pass detects a condition jump that branches to an
4078: identical or inverse test. Such jumps can be @samp{threaded} through
4079: the second conditional test. The source code for this pass is in
4080: @file{jump.c}. This optimization is only performed if
4081: @samp{-fthread-jumps} is enabled.
4082:
4083: @cindex common subexpression elimination
4084: @cindex constant propagation
4085: @item
4086: Common subexpression elimination. This pass also does constant
4087: propagation. Its source file is @file{cse.c}. If constant
4088: propagation causes conditional jumps to become unconditional or to
4089: become no-ops, jump optimization is run again when CSE is finished.
4090:
4091: The option @samp{-ds} causes a debugging dump of the RTL code after
4092: this pass. This dump file's name is made by appending @samp{.cse} to
4093: the input file name.
4094:
4095: @cindex loop optimization
4096: @cindex code motion
4097: @cindex strength-reduction
4098: @item
4099: Loop optimization. This pass moves constant expressions out of loops,
4100: and optionally does strength-reduction and loop unrolling as well.
4101: Its source files are @file{loop.c} and @file{unroll.c}, plus the header
4102: @file{loop.h} used for communication between them. Loop unrolling uses
4103: some functions in @file{integrate.c} and the header @file{integrate.h}.
4104:
4105: The option @samp{-dL} causes a debugging dump of the RTL code after
4106: this pass. This dump file's name is made by appending @samp{.loop} to
4107: the input file name.
4108:
4109: @item
4110: If @samp{-frerun-cse-after-loop} was enabled, a second common
4111: subexpression elimination pass is performed after the loop optimization
4112: pass. Jump threading is also done again at this time if it was specified.
4113:
4114: The option @samp{-dt} causes a debugging dump of the RTL code after
4115: this pass. This dump file's name is made by appending @samp{.cse2} to
4116: the input file name.
4117:
4118: @cindex register allocation, stupid
4119: @cindex stupid register allocation
4120: @item
4121: Stupid register allocation is performed at this point in a
4122: nonoptimizing compilation. It does a little data flow analysis as
4123: well. When stupid register allocation is in use, the next pass
4124: executed is the reloading pass; the others in between are skipped.
4125: The source file is @file{stupid.c}.
4126:
4127: @cindex data flow analysis
4128: @cindex analysis, data flow
4129: @cindex basic blocks
4130: @item
4131: Data flow analysis (@file{flow.c}). This pass divides the program
4132: into basic blocks (and in the process deletes unreachable loops); then
4133: it computes which pseudo-registers are live at each point in the
4134: program, and makes the first instruction that uses a value point at
4135: the instruction that computed the value.
4136:
4137: @cindex autoincrement/decrement analysis
4138: This pass also deletes computations whose results are never used, and
4139: combines memory references with add or subtract instructions to make
4140: autoincrement or autodecrement addressing.
4141:
4142: The option @samp{-df} causes a debugging dump of the RTL code after
4143: this pass. This dump file's name is made by appending @samp{.flow} to
4144: the input file name. If stupid register allocation is in use, this
4145: dump file reflects the full results of such allocation.
4146:
4147: @cindex instruction combination
4148: @item
4149: Instruction combination (@file{combine.c}). This pass attempts to
4150: combine groups of two or three instructions that are related by data
4151: flow into single instructions. It combines the RTL expressions for
4152: the instructions by substitution, simplifies the result using algebra,
4153: and then attempts to match the result against the machine description.
4154:
4155: The option @samp{-dc} causes a debugging dump of the RTL code after
4156: this pass. This dump file's name is made by appending @samp{.combine}
4157: to the input file name.
4158:
4159: @cindex instruction scheduling
4160: @cindex scheduling, instruction
4161: @item
4162: Instruction scheduling (@file{sched.c}). This pass looks for
4163: instructions whose output will not be available by the time that it is
4164: used in subsequent instructions. (Memory loads and floating point
4165: instructions often have this behavior on RISC machines). It re-orders
4166: instructions within a basic block to try to separate the definition and
4167: use of items that otherwise would cause pipeline stalls.
4168:
4169: Instruction scheduling is performed twice. The first time is immediately
4170: after instruction combination and the second is immediately after reload.
4171:
4172: The option @samp{-dS} causes a debugging dump of the RTL code after this
4173: pass is run for the first time. The dump file's name is made by
4174: appending @samp{.sched} to the input file name.
4175:
4176: @cindex register class preference pass
4177: @item
4178: Register class preferencing. The RTL code is scanned to find out
4179: which register class is best for each pseudo register. The source
4180: file is @file{regclass.c}.
4181:
4182: @cindex register allocation
4183: @cindex local register allocation
4184: @item
4185: Local register allocation (@file{local-alloc.c}). This pass allocates
4186: hard registers to pseudo registers that are used only within one basic
4187: block. Because the basic block is linear, it can use fast and
4188: powerful techniques to do a very good job.
4189:
4190: The option @samp{-dl} causes a debugging dump of the RTL code after
4191: this pass. This dump file's name is made by appending @samp{.lreg} to
4192: the input file name.
4193:
4194: @cindex global register allocation
4195: @item
1.1.1.4 root 4196: Global register allocation (@file{global.c}). This pass
1.1 root 4197: allocates hard registers for the remaining pseudo registers (those
4198: whose life spans are not contained in one basic block).
4199:
4200: @cindex reloading
4201: @item
4202: Reloading. This pass renumbers pseudo registers with the hardware
4203: registers numbers they were allocated. Pseudo registers that did not
4204: get hard registers are replaced with stack slots. Then it finds
4205: instructions that are invalid because a value has failed to end up in
4206: a register, or has ended up in a register of the wrong kind. It fixes
4207: up these instructions by reloading the problematical values
4208: temporarily into registers. Additional instructions are generated to
4209: do the copying.
4210:
4211: The reload pass also optionally eliminates the frame pointer and inserts
4212: instructions to save and restore call-clobbered registers around calls.
4213:
4214: Source files are @file{reload.c} and @file{reload1.c}, plus the header
4215: @file{reload.h} used for communication between them.
4216:
4217: The option @samp{-dg} causes a debugging dump of the RTL code after
4218: this pass. This dump file's name is made by appending @samp{.greg} to
4219: the input file name.
4220:
4221: @cindex instruction scheduling
4222: @cindex scheduling, instruction
4223: @item
4224: Instruction scheduling is repeated here to try to avoid pipeline stalls
4225: due to memory loads generated for spilled pseudo registers.
4226:
4227: The option @samp{-dR} causes a debugging dump of the RTL code after
4228: this pass. This dump file's name is made by appending @samp{.sched2}
4229: to the input file name.
4230:
4231: @cindex cross-jumping
4232: @cindex no-op move instructions
4233: @item
4234: Jump optimization is repeated, this time including cross-jumping
4235: and deletion of no-op move instructions.
4236:
4237: The option @samp{-dJ} causes a debugging dump of the RTL code after
4238: this pass. This dump file's name is made by appending @samp{.jump2}
4239: to the input file name.
4240:
4241: @cindex delayed branch scheduling
4242: @cindex scheduling, delayed branch
4243: @item
4244: Delayed branch scheduling. This optional pass attempts to find
4245: instructions that can go into the delay slots of other instructions,
4246: usually jumps and calls. The source file name is @file{reorg.c}.
4247:
4248: The option @samp{-dd} causes a debugging dump of the RTL code after
4249: this pass. This dump file's name is made by appending @samp{.dbr}
4250: to the input file name.
4251:
4252: @cindex register-to-stack conversion
4253: @item
4254: Conversion from usage of some hard registers to usage of a register
4255: stack may be done at this point. Currently, this is supported only
4256: for the floating-point registers of the Intel 80387 coprocessor. The
4257: source file name is @file{reg-stack.c}.
4258:
4259: The options @samp{-dk} causes a debugging dump of the RTL code after
4260: this pass. This dump file's name is made by appending @samp{.stack}
4261: to the input file name.
4262:
4263: @cindex final pass
4264: @cindex peephole optimization
4265: @item
4266: Final. This pass outputs the assembler code for the function. It is
4267: also responsible for identifying spurious test and compare
4268: instructions. Machine-specific peephole optimizations are performed
4269: at the same time. The function entry and exit sequences are generated
4270: directly as assembler code in this pass; they never exist as RTL.
4271:
4272: The source files are @file{final.c} plus @file{insn-output.c}; the
4273: latter is generated automatically from the machine description by the
4274: tool @file{genoutput}. The header file @file{conditions.h} is used
4275: for communication between these files.
4276:
4277: @cindex debugging information generation
4278: @item
4279: Debugging information output. This is run after final because it must
4280: output the stack slot offsets for pseudo registers that did not get
4281: hard registers. Source files are @file{dbxout.c} for DBX symbol table
4282: format, @file{sdbout.c} for SDB symbol table format, and
4283: @file{dwarfout.c} for DWARF symbol table format.
4284: @end itemize
4285:
4286: Some additional files are used by all or many passes:
4287:
4288: @itemize @bullet
4289: @item
4290: Every pass uses @file{machmode.def} and @file{machmode.h} which define
4291: the machine modes.
4292:
4293: @item
4294: Several passes use @file{real.h}, which defines the default
4295: representation of floating point constants and how to operate on them.
4296:
4297: @item
4298: All the passes that work with RTL use the header files @file{rtl.h}
4299: and @file{rtl.def}, and subroutines in file @file{rtl.c}. The tools
4300: @code{gen*} also use these files to read and work with the machine
4301: description RTL.
4302:
4303: @findex genconfig
4304: @item
4305: Several passes refer to the header file @file{insn-config.h} which
4306: contains a few parameters (C macro definitions) generated
4307: automatically from the machine description RTL by the tool
4308: @code{genconfig}.
4309:
4310: @cindex instruction recognizer
4311: @item
4312: Several passes use the instruction recognizer, which consists of
4313: @file{recog.c} and @file{recog.h}, plus the files @file{insn-recog.c}
4314: and @file{insn-extract.c} that are generated automatically from the
4315: machine description by the tools @file{genrecog} and
4316: @file{genextract}.@refill
4317:
4318: @item
4319: Several passes use the header files @file{regs.h} which defines the
4320: information recorded about pseudo register usage, and @file{basic-block.h}
4321: which defines the information recorded about basic blocks.
4322:
4323: @item
4324: @file{hard-reg-set.h} defines the type @code{HARD_REG_SET}, a bit-vector
4325: with a bit for each hard register, and some macros to manipulate it.
4326: This type is just @code{int} if the machine has few enough hard registers;
4327: otherwise it is an array of @code{int} and some of the macros expand
4328: into loops.
4329:
4330: @item
4331: Several passes use instruction attributes. A definition of the
4332: attributes defined for a particular machine is in file
4333: @file{insn-attr.h}, which is generated from the machine description by
4334: the program @file{genattr}. The file @file{insn-attrtab.c} contains
4335: subroutines to obtain the attribute values for insns. It is generated
4336: from the machine description by the program @file{genattrtab}.@refill
4337: @end itemize
4338: @end ifset
4339:
1.1.1.5 root 4340: @ifset INTERNALS
1.1 root 4341: @include rtl.texi
4342: @include md.texi
4343: @include tm.texi
1.1.1.5 root 4344: @end ifset
1.1 root 4345:
4346: @ifset INTERNALS
1.1.1.3 root 4347: @node Config
1.1 root 4348: @chapter The Configuration File
4349: @cindex configuration file
4350: @cindex @file{xm-@var{machine}.h}
4351:
4352: The configuration file @file{xm-@var{machine}.h} contains macro
4353: definitions that describe the machine and system on which the compiler
4354: is running, unlike the definitions in @file{@var{machine}.h}, which
4355: describe the machine for which the compiler is producing output. Most
4356: of the values in @file{xm-@var{machine}.h} are actually the same on all
4357: machines that GNU CC runs on, so large parts of all configuration files
4358: are identical. But there are some macros that vary:
4359:
4360: @table @code
4361: @findex USG
4362: @item USG
4363: Define this macro if the host system is System V.
4364:
4365: @findex VMS
4366: @item VMS
4367: Define this macro if the host system is VMS.
4368:
1.1.1.8 root 4369: @findex FATAL_EXIT_CODE
4370: @item FATAL_EXIT_CODE
1.1 root 4371: A C expression for the status code to be returned when the compiler
4372: exits after serious errors.
4373:
4374: @findex SUCCESS_EXIT_CODE
4375: @item SUCCESS_EXIT_CODE
4376: A C expression for the status code to be returned when the compiler
4377: exits without serious errors.
4378:
4379: @findex HOST_WORDS_BIG_ENDIAN
4380: @item HOST_WORDS_BIG_ENDIAN
4381: Defined if the host machine stores words of multi-word values in
4382: big-endian order. (GNU CC does not depend on the host byte ordering
4383: within a word.)
4384:
1.1.1.6 root 4385: @findex HOST_FLOAT_WORDS_BIG_ENDIAN
4386: @item HOST_FLOAT_WORDS_BIG_ENDIAN
4387: Define this macro to be 1 if the host machine stores @code{DFmode},
4388: @code{XFmode} or @code{TFmode} floating point numbers in memory with the
4389: word containing the sign bit at the lowest address; otherwise, define it
4390: to be zero.
4391:
4392: This macro need not be defined if the ordering is the same as for
4393: multi-word integers.
4394:
1.1 root 4395: @findex HOST_FLOAT_FORMAT
4396: @item HOST_FLOAT_FORMAT
4397: A numeric code distinguishing the floating point format for the host
4398: machine. See @code{TARGET_FLOAT_FORMAT} in @ref{Storage Layout} for the
4399: alternatives and default.
4400:
4401: @findex HOST_BITS_PER_CHAR
4402: @item HOST_BITS_PER_CHAR
4403: A C expression for the number of bits in @code{char} on the host
4404: machine.
4405:
4406: @findex HOST_BITS_PER_SHORT
4407: @item HOST_BITS_PER_SHORT
4408: A C expression for the number of bits in @code{short} on the host
4409: machine.
4410:
4411: @findex HOST_BITS_PER_INT
4412: @item HOST_BITS_PER_INT
4413: A C expression for the number of bits in @code{int} on the host
4414: machine.
4415:
4416: @findex HOST_BITS_PER_LONG
4417: @item HOST_BITS_PER_LONG
4418: A C expression for the number of bits in @code{long} on the host
4419: machine.
4420:
4421: @findex ONLY_INT_FIELDS
4422: @item ONLY_INT_FIELDS
4423: Define this macro to indicate that the host compiler only supports
4424: @code{int} bit fields, rather than other integral types, including
4425: @code{enum}, as do most C compilers.
4426:
4427: @findex OBSTACK_CHUNK_SIZE
4428: @item OBSTACK_CHUNK_SIZE
4429: A C expression for the size of ordinary obstack chunks.
4430: If you don't define this, a usually-reasonable default is used.
4431:
4432: @findex OBSTACK_CHUNK_ALLOC
4433: @item OBSTACK_CHUNK_ALLOC
4434: The function used to allocate obstack chunks.
4435: If you don't define this, @code{xmalloc} is used.
4436:
4437: @findex OBSTACK_CHUNK_FREE
4438: @item OBSTACK_CHUNK_FREE
4439: The function used to free obstack chunks.
4440: If you don't define this, @code{free} is used.
4441:
4442: @findex USE_C_ALLOCA
4443: @item USE_C_ALLOCA
4444: Define this macro to indicate that the compiler is running with the
4445: @code{alloca} implemented in C. This version of @code{alloca} can be
4446: found in the file @file{alloca.c}; to use it, you must also alter the
4447: @file{Makefile} variable @code{ALLOCA}. (This is done automatically
4448: for the systems on which we know it is needed.)
4449:
4450: If you do define this macro, you should probably do it as follows:
4451:
4452: @example
4453: #ifndef __GNUC__
4454: #define USE_C_ALLOCA
4455: #else
4456: #define alloca __builtin_alloca
4457: #endif
4458: @end example
4459:
4460: @noindent
4461: so that when the compiler is compiled with GNU CC it uses the more
4462: efficient built-in @code{alloca} function.
4463:
4464: @item FUNCTION_CONVERSION_BUG
4465: @findex FUNCTION_CONVERSION_BUG
4466: Define this macro to indicate that the host compiler does not properly
4467: handle converting a function value to a pointer-to-function when it is
4468: used in an expression.
1.1.1.2 root 4469:
4470: @findex HAVE_VPRINTF
4471: @findex vprintf
4472: @item HAVE_VPRINTF
4473: Define this if the library function @code{vprintf} is available on your
4474: system.
4475:
1.1.1.3 root 4476: @findex MULTIBYTE_CHARS
4477: @item MULTIBYTE_CHARS
4478: Define this macro to enable support for multibyte characters in the
4479: input to GNU CC. This requires that the host system support the ANSI C
4480: library functions for converting multibyte characters to wide
4481: characters.
4482:
1.1.1.2 root 4483: @findex HAVE_PUTENV
4484: @findex putenv
4485: @item HAVE_PUTENV
4486: Define this if the library function @code{putenv} is available on your
4487: system.
4488:
1.1.1.8 root 4489: @findex POSIX
4490: @item POSIX
4491: Define this if your system is POSIX.1 compliant.
4492:
1.1.1.2 root 4493: @findex NO_SYS_SIGLIST
4494: @item NO_SYS_SIGLIST
4495: Define this if your system @emph{does not} provide the variable
4496: @code{sys_siglist}.
4497:
1.1.1.7 root 4498: @findex DONT_DECLARE_SYS_SIGLIST
4499: @item DONT_DECLARE_SYS_SIGLIST
4500: Define this if your system has the variable @code{sys_siglist}, and
4501: there is already a declaration of it in the system header files.
4502:
1.1.1.4 root 4503: @findex USE_PROTOTYPES
4504: @item USE_PROTOTYPES
4505: Define this to be 1 if you know that the host compiler supports
4506: prototypes, even if it doesn't define __STDC__, or define
4507: it to be 0 if you do not want any prototypes used in compiling
4508: GNU CC. If @samp{USE_PROTOTYPES} is not defined, it will be
4509: determined automatically whether your compiler supports
4510: prototypes by checking if @samp{__STDC__} is defined.
4511:
4512: @findex NO_MD_PROTOTYPES
4513: @item NO_MD_PROTOTYPES
4514: Define this if you wish suppression of prototypes generated from
4515: the machine description file, but to use other prototypes within
4516: GNU CC. If @samp{USE_PROTOTYPES} is defined to be 0, or the
4517: host compiler does not support prototypes, this macro has no
4518: effect.
4519:
4520: @findex MD_CALL_PROTOTYPES
4521: @item MD_CALL_PROTOTYPES
4522: Define this if you wish to generate prototypes for the
4523: @code{gen_call} or @code{gen_call_value} functions generated from
4524: the machine description file. If @samp{USE_PROTOTYPES} is
4525: defined to be 0, or the host compiler does not support
4526: prototypes, or @samp{NO_MD_PROTOTYPES} is defined, this macro has
4527: no effect. As soon as all of the machine descriptions are
4528: modified to have the appropriate number of arguments, this macro
4529: will be removed.
4530:
1.1.1.2 root 4531: @vindex sys_siglist
4532: Some systems do provide this variable, but with a different name such
4533: as @code{_sys_siglist}. On these systems, you can define
4534: @code{sys_siglist} as a macro which expands into the name actually
4535: provided.
4536:
4537: @findex NO_STAB_H
4538: @item NO_STAB_H
4539: Define this if your system does not have the include file
4540: @file{stab.h}. If @samp{USG} is defined, @samp{NO_STAB_H} is
4541: assumed.
1.1.1.7 root 4542:
4543: @findex PATH_SEPARATOR
4544: @item PATH_SEPARATOR
4545: Define this macro to be a C character constant representing the
4546: character used to separate components in paths. The default value is.
4547: the colon character
4548:
4549: @findex DIR_SEPARATOR
4550: @item DIR_SEPARATOR
4551: If your system uses some character other than slash to separate
4552: directory names within a file specification, define this macro to be a C
4553: character constant specifying that character. When GNU CC displays file
4554: names, the character you specify will be used. GNU CC will test for
4555: both slash and the character you specify when parsing filenames.
1.1.1.8 root 4556:
4557: @findex OBJECT_SUFFIX
4558: @item OBJECT_SUFFIX
4559: Define this macro to be a C string representing the suffix for object
4560: files on your machine. If you do not define this macro, GNU CC will use
4561: @samp{.o} as the suffix for object files.
4562:
4563: @findex EXECUTABLE_SUFFIX
4564: @item EXECUTABLE_SUFFIX
4565: Define this macro to be a C string representing the suffix for executable
4566: files on your machine. If you do not define this macro, GNU CC will use
4567: the null string as the suffix for object files.
4568:
4569: @findex COLLECT_EXPORT_LIST
4570: @item COLLECT_EXPORT_LIST
4571: If defined, @code{collect2} will scan the individual object files
4572: specified on its command line and create an export list for the linker.
4573: Define this macro for systems like AIX, where the linker discards
4574: object files that are not referenced from @code{main} and uses export
4575: lists.
1.1 root 4576: @end table
4577:
4578: @findex bzero
4579: @findex bcmp
4580: In addition, configuration files for system V define @code{bcopy},
4581: @code{bzero} and @code{bcmp} as aliases. Some files define @code{alloca}
4582: as a macro when compiled with GNU CC, in order to take advantage of the
4583: benefit of GNU CC's built-in @code{alloca}.
4584:
1.1.1.8 root 4585: @node Fragments
4586: @chapter Makefile Fragments
4587: @cindex makefile fragment
4588:
4589: When you configure GNU CC using the @file{configure} script
4590: (@pxref{Installation}), it will construct the file @file{Makefile} from
4591: the template file @file{Makefile.in}. When it does this, it will
4592: incorporate makefile fragment files from the @file{config} directory,
4593: named @file{t-@var{target}} and @file{x-@var{host}}. If these files do
4594: not exist, it means nothing needs to be added for a given target or
4595: host.
4596:
4597: @menu
4598: * Target Fragment:: Writing the @file{t-@var{target}} file.
4599: * Host Fragment:: Writing the @file{x-@var{host}} file.
4600: @end menu
4601:
4602: @node Target Fragment
4603: @section The Target Makefile Fragment
4604: @cindex target makefile fragment
4605: @cindex @file{t-@var{target}}
4606:
4607: The target makefile fragment, @file{t-@var{target}}, defines special
4608: target dependent variables and targets used in the @file{Makefile}:
4609:
4610: @table @code
4611: @findex LIBGCC1
4612: @item LIBGCC1
4613: The rule to use to build @file{libgcc1.a}.
4614: If your target does not need to use the functions in @file{libgcc1.a},
4615: set this to empty.
4616: @xref{Interface}.
4617:
4618: @findex CROSS_LIBGCC1
4619: @item CROSS_LIBGCC1
4620: The rule to use to build @file{libgcc1.a} when building a cross
4621: compiler. If your target does not need to use the functions in
4622: @file{libgcc1.a}, set this to empty. @xref{Cross Runtime}.
4623:
4624: @findex LIBGCC2_CFLAGS
4625: @item LIBGCC2_CFLAGS
4626: Compiler flags to use when compiling @file{libgcc2.c}.
4627:
4628: @findex LIB2FUNCS_EXTRA
4629: @item LIB2FUNCS_EXTRA
4630: A list of source file names to be compiled or assembled and inserted
4631: into @file{libgcc.a}.
4632:
4633: @findex CRTSTUFF_T_CFLAGS
4634: @item CRTSTUFF_T_CFLAGS
4635: Special flags used when compiling @file{crtstuff.c}.
4636: @xref{Initialization}.
4637:
4638: @findex MULTILIB_OPTIONS
4639: @item MULTILIB_OPTIONS
4640: For some targets, invoking GNU CC in different ways produces objects
4641: that can not be linked together. For example, for some targets GNU CC
4642: produces both big and little endian code. For these targets, you must
4643: arrange for multiple versions of @file{libgcc.a} to be compiled, one for
4644: each set of incompatible options. When GNU CC invokes the linker, it
4645: arranges to link in the right version of @file{libgcc.a}, based on
4646: the command line options used.
4647:
4648: The @code{MULTILIB_OPTIONS} macro lists the set of options for which
4649: special versions of @file{libgcc.a} must be built. Write options that
4650: are mutually incompatible side by side, separated by a slash. Write
4651: options that may be used together separated by a space. The build
4652: procedure will build all combinations of compatible options.
4653:
4654: For example, if you set @code{MULTILIB_OPTIONS} to @samp{m68000/m68020
4655: msoft-float}, @file{Makefile} will build special versions of
4656: @file{libgcc.a} using the options @samp{-m68000}, @samp{-m68020},
4657: @samp{-msoft-float}, @samp{-m68000 -msoft-float}, and @samp{-m68020
4658: -msoft-float}.
4659:
4660: @findex MULTILIB_DIRNAMES
4661: @item MULTILIB_DIRNAMES
4662: If @code{MULTILIB_OPTIONS} is used, this variable specifies the
4663: directory names that should be used to hold the various libraries.
4664: Write one element in @code{MULTILIB_DIRNAMES} for each element in
4665: @code{MULTILIB_OPTIONS}. If @code{MULTILIB_DIRNAMES} is not used, the
4666: default value will be @code{MULTILIB_OPTIONS}, with all slashes treated
4667: as spaces.
4668:
4669: For example, if @code{MULTILIB_OPTIONS} is @samp{m68000/m68020
4670: msoft-float}, then the default value of @code{MULTILIB_DIRNAMES} is
4671: @samp{m68000 m68020 msoft-float}. You may specify a different value if
4672: you desire a different set of directory names.
4673:
4674: @findex MULTILIB_MATCHES
4675: @item MULTILIB_MATCHES
4676: Sometimes the same option may be written in two different ways. If an
4677: option is listed in @code{MULTILIB_OPTIONS}, GNU CC needs to know about
4678: any synonyms. In that case, set @code{MULTILIB_MATCHES} to a list of
4679: items of the form @samp{option=option} to describe all relevant
4680: synonyms. For example, @samp{m68000=mc68000 m68020=mc68020}.
4681: @end table
4682:
4683: @node Host Fragment
4684: @section The Host Makefile Fragment
4685: @cindex host makefile fragment
4686: @cindex @file{x-@var{host}}
4687:
4688: The host makefile fragment, @file{x-@var{host}}, defines special host
4689: dependent variables and targets used in the @file{Makefile}:
4690:
4691: @table @code
4692: @findex CC
4693: @item CC
4694: The compiler to use when building the first stage.
4695:
4696: @findex CLIB
4697: @item CLIB
4698: Additional host libraries to link with.
4699:
4700: @findex OLDCC
4701: @item OLDCC
4702: The compiler to use when building @file{libgcc1.a} for a native
4703: compilation.
4704:
4705: @findex OLDAR
4706: @item OLDAR
4707: The version of @code{ar} to use when building @file{libgcc1.a} for a native
4708: compilation.
4709:
4710: @findex INSTALL
4711: @item INSTALL
4712: The install program to use.
4713: @end table
1.1 root 4714:
1.1.1.3 root 4715: @node Index
1.1 root 4716: @unnumbered Index
4717: @end ifset
4718:
4719: @ifclear INTERNALS
1.1.1.3 root 4720: @node Index
1.1 root 4721: @unnumbered Index
4722: @end ifclear
4723:
4724: @printindex cp
1.1.1.5 root 4725: @summarycontents
1.1 root 4726: @contents
4727: @bye
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.