![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to gcc.1 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.1 CVS log | Up to [GCC 2.x] / gcc |
1.1.1.7 root 1: .\" Copyright (c) 1991, 1992, 1993, 1994 Free Software Foundation -*-Text-*-
1.1 root 2: .\" See section COPYING for conditions for redistribution
1.1.1.6 root 3: .\"
4: .\" Set up \*(lq, \*(rq if -man hasn't already set it up.
5: .if @@\*(lq@ \{\
6: . ds lq "
7: . if t .ds lq ``
8: . if !@@\(lq@ .ds lq "\(lq
9: .\}
10: .if @@\*(rq@ \{\
11: . ds rq "
12: . if t .ds rq ''
13: . if !@@\(rq@ .ds rq "\(rq
14: .\}
15: .de Id
16: .ds Rv \\$3
17: .ds Dt \\$4
1.1 root 18: ..
1.1.1.6 root 19: .de Sp
20: .if n .sp
21: .if t .sp 0.4
22: ..
23: .Id $Id: gcc.1,v 1.4 1993/10/13 23:19:12 pesch Exp $
24: .TH GCC 1 "\*(Dt" "GNU Tools" "GNU Tools"
1.1 root 25: .SH NAME
1.1.1.8 ! root 26: gcc, g++ \- GNU project C and C++ Compiler (v2.7)
1.1 root 27: .SH SYNOPSIS
1.1.1.6 root 28: .B gcc
29: .RI "[ " option " | " filename " ].\|.\|."
1.1 root 30: .br
1.1.1.6 root 31: .B g++
32: .RI "[ " option " | " filename " ].\|.\|."
1.1 root 33: .SH WARNING
34: The information in this man page is an extract from the full
35: documentation of the GNU C compiler, and is limited to the meaning of
1.1.1.6 root 36: the options.
37: .PP
38: This man page is not kept up to date except when volunteers want to
39: maintain it. If you find a discrepancy between the man page and the
40: software, please check the Info file, which is the authoritative
41: documentation.
42: .PP
43: If we find that the things in this man page that are out of date cause
44: significant confusion or complaints, we will stop distributing the man
45: page. The alternative, updating the man page when we update the Info
46: file, is impossible because the rest of the work of maintaining GNU CC
47: leaves us no time for that. The GNU project regards man pages as
48: obsolete and should not let them take time away from other things.
49: .PP
1.1.1.4 root 50: For complete and current documentation, refer to the Info file `\|\c
1.1 root 51: .B gcc\c
52: \&\|' or the manual
1.1.1.6 root 53: .I
1.1 root 54: Using and Porting GNU CC (for version 2.0)\c
55: \&. Both are made from the Texinfo source file
56: .BR gcc.texinfo .
57: .SH DESCRIPTION
58: The C and C++ compilers are integrated. Both process input files
59: through one or more of four stages: preprocessing, compilation,
60: assembly, and linking. Source filename suffixes identify the source
61: language, but which name you use for the compiler governs default
62: assumptions:
63: .TP
64: .B gcc
65: assumes preprocessed (\c
66: .B .i\c
67: \&) files are C and assumes C style linking.
68: .TP
69: .B g++
70: assumes preprocessed (\c
71: .B .i\c
72: \&) files are C++ and assumes C++ style linking.
1.1.1.6 root 73: .PP
1.1 root 74: Suffixes of source file names indicate the language and kind of
75: processing to be done:
1.1.1.6 root 76: .Sp
77: .nf
78: .ta \w'\fB.cxx\fP 'u
79: \&\fB.c\fP C source; preprocess, compile, assemble
80: \&\fB.C\fP C++ source; preprocess, compile, assemble
81: \&\fB.cc\fP C++ source; preprocess, compile, assemble
82: \&\fB.cxx\fP C++ source; preprocess, compile, assemble
83: \&\fB.m\fP Objective-C source; preprocess, compile, assemble
84: \&\fB.i\fP preprocessed C; compile, assemble
85: \&\fB.ii\fP preprocessed C++; compile, assemble
86: \&\fB.s\fP Assembler source; assemble
87: \&\fB.S\fP Assembler source; preprocess, assemble
88: \&\fB.h\fP Preprocessor file; not usually named on command line
89: .Sp
90: .fi
91: Files with other suffixes are passed to the linker. Common cases include:
92: .Sp
93: .nf
94: \&\fB.o\fP Object file
95: \&\fB.a\fP Archive file
96: .br
97: .fi
98: .Sp
99: Linking is always the last stage unless you use one of the
1.1 root 100: .BR \-c ,
101: .BR \-S ,
1.1.1.6 root 102: or
103: .B \-E
1.1 root 104: options to avoid it (or unless compilation errors stop the whole
1.1.1.6 root 105: process). For the link stage, all
1.1 root 106: .B .o
107: files corresponding to source files,
1.1.1.6 root 108: .B \-l
109: libraries, unrecognized filenames (including named
1.1 root 110: .B .o
1.1.1.6 root 111: object files and
112: .B .a
1.1 root 113: archives)
114: are passed to the linker in command-line order.
115: .SH OPTIONS
116: Options must be separate: `\|\c
117: .B \-dr\c
118: \&\|' is quite different from `\|\c
119: .B \-d \-r
1.1.1.6 root 120: \&\|'.
121: .PP
1.1 root 122: Most `\|\c
123: .B \-f\c
124: \&\|' and `\|\c
125: .B \-W\c
1.1.1.6 root 126: \&\|' options have two contrary forms:
1.1 root 127: .BI \-f name
128: and
129: .BI \-fno\- name\c
1.1.1.6 root 130: \& (or
1.1 root 131: .BI \-W name
132: and
133: .BI \-Wno\- name\c
1.1.1.6 root 134: \&). Only the non-default forms are shown here.
135: .PP
1.1 root 136: Here is a summary of all the options, grouped by type. Explanations are
137: in the following sections.
138: .hy 0
139: .na
140: .TP
141: .B Overall Options
142: .br
143: \-c
144: \-S
145: \-E
146: .RI "\-o " file
147: \-pipe
148: \-v
149: .RI "\-x " language
150: .TP
151: .B Language Options
152: \-ansi
153: \-fall\-virtual
154: \-fcond\-mismatch
155: \-fdollars\-in\-identifiers
156: \-fenum\-int\-equiv
1.1.1.6 root 157: \-fexternal\-templates
1.1 root 158: \-fno\-asm
159: \-fno\-builtin
160: \-fno\-strict\-prototype
161: \-fsigned\-bitfields
162: \-fsigned\-char
163: \-fthis\-is\-variable
164: \-funsigned\-bitfields
165: \-funsigned\-char
166: \-fwritable\-strings
167: \-traditional
168: \-traditional\-cpp
169: \-trigraphs
170: .TP
171: .B Warning Options
172: \-fsyntax\-only
173: \-pedantic
174: \-pedantic\-errors
175: \-w
176: \-W
177: \-Wall
178: \-Waggregate\-return
179: \-Wcast\-align
180: \-Wcast\-qual
1.1.1.5 root 181: \-Wchar\-subscript
1.1 root 182: \-Wcomment
183: \-Wconversion
184: \-Wenum\-clash
185: \-Werror
186: \-Wformat
187: .RI \-Wid\-clash\- len
188: \-Wimplicit
1.1.1.2 root 189: \-Winline
1.1 root 190: \-Wmissing\-prototypes
1.1.1.7 root 191: \-Wmissing\-declarations
1.1.1.5 root 192: \-Wnested\-externs
193: \-Wno\-import
1.1.1.2 root 194: \-Wparentheses
1.1 root 195: \-Wpointer\-arith
1.1.1.5 root 196: \-Wredundant\-decls
1.1 root 197: \-Wreturn\-type
198: \-Wshadow
199: \-Wstrict\-prototypes
200: \-Wswitch
1.1.1.5 root 201: \-Wtemplate\-debugging
1.1 root 202: \-Wtraditional
203: \-Wtrigraphs
204: \-Wuninitialized
205: \-Wunused
206: \-Wwrite\-strings
207: .TP
208: .B Debugging Options
209: \-a
210: .RI \-d letters
211: \-fpretend\-float
212: \-g
1.1.1.5 root 213: .RI \-g level
214: \-gcoff
215: \-gxcoff
216: \-gxcoff+
1.1 root 217: \-gdwarf
1.1.1.5 root 218: \-gdwarf+
219: \-gstabs
220: \-gstabs+
1.1 root 221: \-ggdb
222: \-p
223: \-pg
224: \-save\-temps
1.1.1.7 root 225: .RI \-print\-file\-name= library
1.1.1.5 root 226: \-print\-libgcc\-file\-name
1.1.1.7 root 227: .RI \-print\-prog\-name= program
1.1 root 228: .TP
229: .B Optimization Options
230: \-fcaller\-saves
231: \-fcse\-follow\-jumps
1.1.1.5 root 232: \-fcse\-skip\-blocks
1.1 root 233: \-fdelayed\-branch
234: \-felide\-constructors
235: \-fexpensive\-optimizations
1.1.1.6 root 236: \-ffast\-math
1.1 root 237: \-ffloat\-store
238: \-fforce\-addr
239: \-fforce\-mem
240: \-finline\-functions
241: \-fkeep\-inline\-functions
242: \-fmemoize\-lookups
243: \-fno\-default\-inline
244: \-fno\-defer\-pop
245: \-fno\-function\-cse
1.1.1.5 root 246: \-fno\-inline
247: \-fno\-peephole
1.1 root 248: \-fomit\-frame\-pointer
249: \-frerun\-cse\-after\-loop
250: \-fschedule\-insns
251: \-fschedule\-insns2
252: \-fstrength\-reduce
253: \-fthread\-jumps
254: \-funroll\-all\-loops
255: \-funroll\-loops
256: \-O
257: \-O2
258: .TP
259: .B Preprocessor Options
1.1.1.5 root 260: .RI \-A assertion
1.1 root 261: \-C
262: \-dD
263: \-dM
264: \-dN
265: .RI \-D macro [\|= defn \|]
266: \-E
267: \-H
1.1.1.5 root 268: .RI "\-idirafter " dir
269: .RI "\-include " file
270: .RI "\-imacros " file
271: .RI "\-iprefix " file
272: .RI "\-iwithprefix " dir
1.1 root 273: \-M
274: \-MD
275: \-MM
276: \-MMD
277: \-nostdinc
278: \-P
279: .RI \-U macro
280: \-undef
281: .TP
1.1.1.5 root 282: .B Assembler Option
283: .RI \-Wa, option
284: .TP
1.1 root 285: .B Linker Options
286: .RI \-l library
1.1.1.5 root 287: \-nostartfiles
1.1 root 288: \-nostdlib
289: \-static
1.1.1.5 root 290: \-shared
291: \-symbolic
292: .RI "\-Xlinker\ " option
293: .RI \-Wl, option
294: .RI "\-u " symbol
1.1 root 295: .TP
296: .B Directory Options
297: .RI \-B prefix
298: .RI \-I dir
299: \-I\-
300: .RI \-L dir
301: .TP
302: .B Target Options
303: .RI "\-b " machine
304: .RI "\-V " version
305: .TP
1.1.1.5 root 306: .B Configuration Dependent Options
1.1 root 307: .I M680x0\ Options
308: .br
309: \-m68000
310: \-m68020
1.1.1.5 root 311: \-m68020\-40
312: \-m68030
313: \-m68040
1.1 root 314: \-m68881
315: \-mbitfield
316: \-mc68000
317: \-mc68020
318: \-mfpa
319: \-mnobitfield
320: \-mrtd
321: \-mshort
322: \-msoft\-float
1.1.1.6 root 323: .Sp
1.1 root 324: .I VAX Options
325: .br
326: \-mg
327: \-mgnu
328: \-munix
1.1.1.6 root 329: .Sp
1.1 root 330: .I SPARC Options
331: .br
1.1.1.5 root 332: \-mepilogue
1.1 root 333: \-mfpu
1.1.1.5 root 334: \-mhard\-float
335: \-mno\-fpu
1.1 root 336: \-mno\-epilogue
1.1.1.5 root 337: \-msoft\-float
338: \-msparclite
339: \-mv8
1.1.1.7 root 340: \-msupersparc
341: \-mcypress
1.1.1.6 root 342: .Sp
1.1 root 343: .I Convex Options
344: .br
345: \-margcount
346: \-mc1
347: \-mc2
348: \-mnoargcount
1.1.1.6 root 349: .Sp
1.1 root 350: .I AMD29K Options
351: .br
352: \-m29000
353: \-m29050
354: \-mbw
355: \-mdw
356: \-mkernel\-registers
357: \-mlarge
358: \-mnbw
359: \-mnodw
360: \-msmall
361: \-mstack\-check
362: \-muser\-registers
1.1.1.6 root 363: .Sp
1.1 root 364: .I M88K Options
365: .br
1.1.1.5 root 366: \-m88000
367: \-m88100
368: \-m88110
1.1 root 369: \-mbig\-pic
370: \-mcheck\-zero\-division
371: \-mhandle\-large\-shift
372: \-midentify\-revision
373: \-mno\-check\-zero\-division
374: \-mno\-ocs\-debug\-info
375: \-mno\-ocs\-frame\-position
376: \-mno\-optimize\-arg\-area
1.1.1.8 ! root 377: \-mno\-serialize\-volatile
1.1 root 378: \-mno\-underscores
379: \-mocs\-debug\-info
380: \-mocs\-frame\-position
381: \-moptimize\-arg\-area
1.1.1.5 root 382: \-mserialize\-volatile
1.1 root 383: .RI \-mshort\-data\- num
384: \-msvr3
385: \-msvr4
386: \-mtrap\-large\-shift
387: \-muse\-div\-instruction
388: \-mversion\-03.00
389: \-mwarn\-passed\-structs
1.1.1.6 root 390: .Sp
1.1 root 391: .I RS6000 Options
392: .br
393: \-mfp\-in\-toc
394: \-mno\-fop\-in\-toc
1.1.1.6 root 395: .Sp
1.1 root 396: .I RT Options
397: .br
398: \-mcall\-lib\-mul
399: \-mfp\-arg\-in\-fpregs
400: \-mfp\-arg\-in\-gregs
401: \-mfull\-fp\-blocks
402: \-mhc\-struct\-return
403: \-min\-line\-mul
404: \-mminimum\-fp\-blocks
405: \-mnohc\-struct\-return
1.1.1.6 root 406: .Sp
1.1 root 407: .I MIPS Options
408: .br
409: \-mcpu=\fIcpu type\fP
410: \-mips2
411: \-mips3
412: \-mint64
413: \-mlong64
414: \-mlonglong128
415: \-mmips\-as
416: \-mgas
417: \-mrnames
418: \-mno\-rnames
419: \-mgpopt
420: \-mno\-gpopt
421: \-mstats
422: \-mno\-stats
423: \-mmemcpy
424: \-mno\-memcpy
425: \-mno\-mips\-tfile
426: \-mmips\-tfile
427: \-msoft\-float
428: \-mhard\-float
429: \-mabicalls
430: \-mno\-abicalls
431: \-mhalf\-pic
432: \-mno\-half\-pic
433: \-G \fInum\fP
1.1.1.5 root 434: \-nocpp
1.1.1.6 root 435: .Sp
1.1.1.2 root 436: .I i386 Options
437: .br
438: \-m486
1.1.1.5 root 439: \-mno\-486
440: \-msoft\-float
441: \-mno\-fp\-ret\-in\-387
1.1.1.6 root 442: .Sp
1.1.1.5 root 443: .I HPPA Options
444: .br
445: \-mpa\-risc\-1\-0
446: \-mpa\-risc\-1\-1
447: \-mkernel
448: \-mshared\-libs
449: \-mno\-shared\-libs
450: \-mlong\-calls
451: \-mdisable\-fpregs
452: \-mdisable\-indexing
453: \-mtrailing\-colon
1.1.1.6 root 454: .Sp
1.1.1.5 root 455: .I i960 Options
456: .br
457: \-m\fIcpu-type\fP
458: \-mnumerics
1.1.1.2 root 459: \-msoft\-float
1.1.1.5 root 460: \-mleaf\-procedures
461: \-mno\-leaf\-procedures
462: \-mtail\-call
463: \-mno\-tail\-call
464: \-mcomplex\-addr
465: \-mno\-complex\-addr
466: \-mcode\-align
467: \-mno\-code\-align
468: \-mic\-compat
469: \-mic2.0\-compat
470: \-mic3.0\-compat
471: \-masm\-compat
472: \-mintel\-asm
473: \-mstrict\-align
474: \-mno\-strict\-align
475: \-mold\-align
476: \-mno\-old\-align
1.1.1.6 root 477: .Sp
1.1.1.5 root 478: .I DEC Alpha Options
479: .br
480: \-mfp\-regs
481: \-mno\-fp\-regs
482: \-mno\-soft\-float
483: \-msoft\-float
1.1.1.6 root 484: .Sp
1.1.1.5 root 485: .I System V Options
486: .br
487: \-G
488: \-Qy
489: \-Qn
490: .RI \-YP, paths
491: .RI \-Ym, dir
1.1 root 492: .TP
493: .B Code Generation Options
494: .RI \-fcall\-saved\- reg
495: .RI \-fcall\-used\- reg
496: .RI \-ffixed\- reg
1.1.1.5 root 497: \-finhibit\-size\-directive
1.1 root 498: \-fnonnull\-objects
1.1.1.5 root 499: \-fno\-common
500: \-fno\-ident
501: \-fno\-gnu\-linker
1.1 root 502: \-fpcc\-struct\-return
503: \-fpic
504: \-fPIC
1.1.1.8 ! root 505: \-freg\-struct\-return
1.1 root 506: \-fshared\-data
507: \-fshort\-enums
508: \-fshort\-double
509: \-fvolatile
1.1.1.5 root 510: \-fvolatile\-global
511: \-fverbose\-asm
1.1 root 512: .ad b
513: .hy 1
514: .SH OVERALL OPTIONS
515: .TP
516: .BI "\-x " "language"
1.1.1.6 root 517: Specify explicitly the
1.1 root 518: .I language\c
519: \& for the following input files (rather than choosing a default based
520: on the file name suffix) . This option applies to all following input
521: files until the next `\|\c
522: .B \-x\c
523: \&\|' option. Possible values of \c
524: .I language\c
525: \& are
526: `\|\c
527: .B c\c
528: \&\|', `\|\c
529: .B objective\-c\c
530: \&\|', `\|\c
531: .B c\-header\c
532: \&\|', `\|\c
533: .B c++\c
534: \&\|',
535: `\|\c
536: .B cpp\-output\c
537: \&\|', `\|\c
538: .B assembler\c
539: \&\|', and `\|\c
540: .B assembler\-with\-cpp\c
541: \&\|'.
542: .TP
543: .B \-x none
544: Turn off any specification of a language, so that subsequent files are
545: handled according to their file name suffixes (as they are if `\|\c
546: .B \-x\c
547: \&\|'
548: has not been used at all).
549: .PP
550: If you want only some of the four stages (preprocess, compile,
551: assemble, link), you can use
552: `\|\c
553: .B \-x\c
554: \&\|' (or filename suffixes) to tell \c
555: .B gcc\c
556: \& where to start, and
557: one of the options `\|\c
558: .B \-c\c
559: \&\|', `\|\c
560: .B \-S\c
561: \&\|', or `\|\c
562: .B \-E\c
563: \&\|' to say where
564: .B gcc\c
565: \& is to stop. Note that some combinations (for example,
566: `\|\c
567: .B \-x cpp\-output \-E\c
568: \&\|') instruct \c
569: .B gcc\c
570: \& to do nothing at all.
571: .TP
572: .B \-c
573: Compile or assemble the source files, but do not link. The compiler
574: output is an object file corresponding to each source file.
1.1.1.6 root 575: .Sp
1.1 root 576: By default, GCC makes the object file name for a source file by replacing
577: the suffix `\|\c
578: .B .c\c
579: \&\|', `\|\c
580: .B .i\c
581: \&\|', `\|\c
582: .B .s\c
583: \&\|', etc., with `\|\c
584: .B .o\c
585: \&\|'. Use
586: .B \-o\c
587: \& to select another name.
1.1.1.6 root 588: .Sp
1.1 root 589: GCC ignores any unrecognized input files (those that do not require
590: compilation or assembly) with the
591: .B \-c
592: option.
593: .TP
594: .B \-S
595: Stop after the stage of compilation proper; do not assemble. The output
596: is an assembler code file for each non-assembler input
597: file specified.
1.1.1.6 root 598: .Sp
1.1 root 599: By default, GCC makes the assembler file name for a source file by
600: replacing the suffix `\|\c
601: .B .c\c
602: \&\|', `\|\c
603: .B .i\c
604: \&\|', etc., with `\|\c
605: .B .s\c
606: \&\|'. Use
607: .B \-o\c
608: \& to select another name.
1.1.1.6 root 609: .Sp
1.1 root 610: GCC ignores any input files that don't require compilation.
611: .TP
612: .B \-E
613: Stop after the preprocessing stage; do not run the compiler proper. The
614: output is preprocessed source code, which is sent to the
615: standard output.
1.1.1.6 root 616: .Sp
1.1 root 617: GCC ignores input files which don't require preprocessing.
618: .TP
1.1.1.6 root 619: .BI "\-o " file
1.1 root 620: Place output in file \c
621: .I file\c
622: \&. This applies regardless to whatever
623: sort of output GCC is producing, whether it be an executable file,
624: an object file, an assembler file or preprocessed C code.
1.1.1.6 root 625: .Sp
1.1 root 626: Since only one output file can be specified, it does not make sense to
627: use `\|\c
628: .B \-o\c
629: \&\|' when compiling more than one input file, unless you are
630: producing an executable file as output.
1.1.1.6 root 631: .Sp
1.1 root 632: If you do not specify `\|\c
633: .B \-o\c
634: \&\|', the default is to put an executable file
635: in `\|\c
636: .B a.out\c
637: \&\|', the object file for `\|\c
638: .I source\c
1.1.1.6 root 639: .B \&.\c
1.1 root 640: .I suffix\c
641: \&\c
642: \&\|' in
643: `\|\c
644: .I source\c
1.1.1.6 root 645: .B \&.o\c
1.1 root 646: \&\|', its assembler file in `\|\c
647: .I source\c
1.1.1.6 root 648: .B \&.s\c
1.1 root 649: \&\|', and
650: all preprocessed C source on standard output.
651: .TP
652: .B \-v
653: Print (on standard error output) the commands executed to run the stages
654: of compilation. Also print the version number of the compiler driver
655: program and of the preprocessor and the compiler proper.
656: .TP
657: .B \-pipe
658: Use pipes rather than temporary files for communication between the
659: various stages of compilation. This fails to work on some systems where
660: the assembler cannot read from a pipe; but the GNU assembler has
661: no trouble.
662: .PP
663: .SH LANGUAGE OPTIONS
664: The following options control the dialect of C that the compiler
665: accepts:
666: .TP
667: .B \-ansi
1.1.1.6 root 668: Support all ANSI standard C programs.
669: .Sp
1.1 root 670: This turns off certain features of GNU C that are incompatible with
671: ANSI C, such as the \c
672: .B asm\c
673: \&, \c
674: .B inline\c
675: \& and \c
1.1.1.6 root 676: .B typeof
1.1 root 677: keywords, and predefined macros such as \c
678: .B unix\c
679: \& and \c
1.1.1.6 root 680: .B vax
1.1 root 681: that identify the type of system you are using. It also enables the
1.1.1.5 root 682: undesirable and rarely used ANSI trigraph feature, and disallows `\|\c
1.1 root 683: .B $\c
684: \&\|' as part of identifiers.
1.1.1.6 root 685: .Sp
1.1 root 686: The alternate keywords \c
1.1.1.6 root 687: .B _\|_asm_\|_\c
1.1 root 688: \&, \c
1.1.1.6 root 689: .B _\|_extension_\|_\c
1.1 root 690: \&,
1.1.1.6 root 691: .B _\|_inline_\|_\c
1.1 root 692: \& and \c
1.1.1.6 root 693: .B _\|_typeof_\|_\c
1.1 root 694: \& continue to work despite
695: `\|\c
696: .B \-ansi\c
697: \&\|'. You would not want to use them in an ANSI C program, of
698: course, but it is useful to put them in header files that might be included
699: in compilations done with `\|\c
700: .B \-ansi\c
701: \&\|'. Alternate predefined macros
702: such as \c
1.1.1.6 root 703: .B _\|_unix_\|_\c
1.1 root 704: \& and \c
1.1.1.6 root 705: .B _\|_vax_\|_\c
1.1 root 706: \& are also available, with or
707: without `\|\c
708: .B \-ansi\c
709: \&\|'.
1.1.1.6 root 710: .Sp
1.1 root 711: The `\|\c
712: .B \-ansi\c
713: \&\|' option does not cause non-ANSI programs to be
714: rejected gratuitously. For that, `\|\c
715: .B \-pedantic\c
716: \&\|' is required in
717: addition to `\|\c
718: .B \-ansi\c
1.1.1.6 root 719: \&\|'.
720: .Sp
1.1 root 721: The preprocessor predefines a macro \c
1.1.1.6 root 722: .B _\|_STRICT_ANSI_\|_\c
1.1 root 723: \& when you use the `\|\c
724: .B \-ansi\c
725: \&\|'
726: option. Some header files may notice this macro and refrain
727: from declaring certain functions or defining certain macros that the
728: ANSI standard doesn't call for; this is to avoid interfering with any
729: programs that might use these names for other things.
730: .TP
731: .B \-fno\-asm
732: Do not recognize \c
733: .B asm\c
734: \&, \c
735: .B inline\c
736: \& or \c
737: .B typeof\c
738: \& as a
739: keyword. These words may then be used as identifiers. You can
740: use \c
1.1.1.6 root 741: .B _\|_asm_\|_\c
1.1 root 742: \&, \c
1.1.1.6 root 743: .B _\|_inline_\|_\c
1.1 root 744: \& and \c
1.1.1.6 root 745: .B _\|_typeof_\|_\c
1.1 root 746: \& instead.
747: `\|\c
748: .B \-ansi\c
749: \&\|' implies `\|\c
750: .B \-fno\-asm\c
751: \&\|'.
752: .TP
753: .B \-fno\-builtin
1.1.1.5 root 754: Don't recognize built-in functions that do not begin with two leading
1.1.1.6 root 755: underscores. Currently, the functions affected include \c
1.1.1.5 root 756: .B _exit\c
757: \&,
758: .B abort\c
759: \&, \c
760: .B abs\c
761: \&, \c
1.1 root 762: .B alloca\c
1.1.1.5 root 763: \&, \c
764: .B cos\c
765: \&, \c
766: .B exit\c
767: \&,
768: .B fabs\c
769: \&, \c
770: .B labs\c
771: \&, \c
772: .B memcmp\c
773: \&, \c
774: .B memcpy\c
775: \&, \c
776: .B sin\c
777: \&,
778: .B sqrt\c
779: \&, \c
780: .B strcmp\c
781: \&, \c
782: .B strcpy\c
783: \&, and \c
784: .B strlen\c
1.1 root 785: \&.
1.1.1.6 root 786: .Sp
1.1.1.5 root 787: The `\|\c
788: .B \-ansi\c
789: \&\|' option prevents \c
790: .B alloca\c
791: \& and \c
792: .B _exit\c
793: \& from
794: being builtin functions.
1.1 root 795: .TP
796: .B \-fno\-strict\-prototype
1.1.1.5 root 797: Treat a function declaration with no arguments, such as `\|\c
798: .B int foo
799: ();\c
800: \&\|', as C would treat it\(em\&as saying nothing about the number of
801: arguments or their types (C++ only). Normally, such a declaration in
802: C++ means that the function \c
1.1 root 803: .B foo\c
1.1.1.5 root 804: \& takes no arguments.
1.1 root 805: .TP
806: .B \-trigraphs
807: Support ANSI C trigraphs. The `\|\c
808: .B \-ansi\c
809: \&\|' option implies `\|\c
810: .B \-trigraphs\c
811: \&\|'.
812: .TP
813: .B \-traditional
814: Attempt to support some aspects of traditional C compilers.
1.1.1.3 root 815: For details, see the GNU C Manual; the duplicate list here
816: has been deleted so that we won't get complaints when it
817: is out of date.
1.1.1.6 root 818: .Sp
1.1.1.3 root 819: But one note about C++ programs only (not C). `\|\c
1.1 root 820: .B \-traditional\c
1.1.1.6 root 821: \&\|' has one additional effect for C++: assignment to
1.1 root 822: .B this
823: is permitted. This is the same as the effect of `\|\c
824: .B \-fthis\-is\-variable\c
825: \&\|'.
826: .TP
827: .B \-traditional\-cpp
828: Attempt to support some aspects of traditional C preprocessors.
829: This includes the items that specifically mention the preprocessor above,
830: but none of the other effects of `\|\c
831: .B \-traditional\c
832: \&\|'.
833: .TP
834: .B \-fdollars\-in\-identifiers
835: Permit the use of `\|\c
836: .B $\c
1.1.1.5 root 837: \&\|' in identifiers (C++ only). You can also use
838: `\|\c
839: .B \-fno\-dollars\-in\-identifiers\c
840: \&\|' to explicitly prohibit use of
841: `\|\c
1.1 root 842: .B $\c
1.1.1.5 root 843: \&\|'. (GNU C++ allows `\|\c
1.1 root 844: .B $\c
1.1.1.5 root 845: \&\|' by default on some target systems
846: but not others.)
1.1 root 847: .TP
848: .B \-fenum\-int\-equiv
1.1.1.5 root 849: Permit implicit conversion of \c
1.1 root 850: .B int\c
1.1.1.5 root 851: \& to enumeration types (C++
852: only). Normally GNU C++ allows conversion of \c
853: .B enum\c
854: \& to \c
855: .B int\c
856: \&,
857: but not the other way around.
1.1 root 858: .TP
1.1.1.6 root 859: .B \-fexternal\-templates
860: Produce smaller code for template declarations, by generating only a
861: single copy of each template function where it is defined (C++ only).
862: To use this option successfully, you must also mark all files that
863: use templates with either `\|\c
864: .B #pragma implementation\c
865: \&\|' (the definition) or
866: `\|\c
867: .B #pragma interface\c
868: \&\|' (declarations).
869:
870: When your code is compiled with `\|\c
871: .B \-fexternal\-templates\c
872: \&\|', all
873: template instantiations are external. You must arrange for all
874: necessary instantiations to appear in the implementation file; you can
875: do this with a \c
876: .B typedef\c
877: \& that references each instantiation needed.
878: Conversely, when you compile using the default option
879: `\|\c
880: .B \-fno\-external\-templates\c
881: \&\|', all template instantiations are
882: explicitly internal.
883: .TP
1.1 root 884: .B \-fall\-virtual
1.1.1.5 root 885: Treat all possible member functions as virtual, implicitly. All
886: member functions (except for constructor functions and
887: .B new
888: or
889: .B delete
890: member operators) are treated as virtual functions of the class where
891: they appear.
1.1.1.6 root 892: .Sp
1.1.1.5 root 893: This does not mean that all calls to these member functions will be
894: made through the internal table of virtual functions. Under some
895: circumstances, the compiler can determine that a call to a given
896: virtual function can be made directly; in these cases the calls are
897: direct in any case.
1.1 root 898: .TP
899: .B \-fcond\-mismatch
900: Allow conditional expressions with mismatched types in the second and
901: third arguments. The value of such an expression is void.
902: .TP
903: .B \-fthis\-is\-variable
1.1.1.5 root 904: Permit assignment to \c
1.1 root 905: .B this\c
1.1.1.5 root 906: \& (C++ only). The incorporation of
907: user-defined free store management into C++ has made assignment to
908: `\|\c
1.1 root 909: .B this\c
1.1.1.5 root 910: \&\|' an anachronism. Therefore, by default it is invalid to
911: assign to \c
1.1 root 912: .B this\c
1.1.1.5 root 913: \& within a class member function. However, for
914: backwards compatibility, you can make it valid with
915: `\|\c
916: .B \-fthis-is-variable\c
1.1 root 917: \&\|'.
918: .TP
919: .B \-funsigned\-char
920: Let the type \c
921: .B char\c
922: \& be unsigned, like \c
923: .B unsigned char\c
924: \&.
1.1.1.6 root 925: .Sp
1.1 root 926: Each kind of machine has a default for what \c
927: .B char\c
928: \& should
929: be. It is either like \c
930: .B unsigned char\c
931: \& by default or like
932: .B signed char\c
933: \& by default.
1.1.1.6 root 934: .Sp
1.1 root 935: Ideally, a portable program should always use \c
936: .B signed char\c
937: \& or
938: .B unsigned char\c
939: \& when it depends on the signedness of an object.
940: But many programs have been written to use plain \c
941: .B char\c
942: \& and
943: expect it to be signed, or expect it to be unsigned, depending on the
944: machines they were written for. This option, and its inverse, let you
945: make such a program work with the opposite default.
1.1.1.6 root 946: .Sp
1.1 root 947: The type \c
948: .B char\c
949: \& is always a distinct type from each of
950: .B signed char\c
951: \& and \c
952: .B unsigned char\c
953: \&, even though its behavior
954: is always just like one of those two.
955: .TP
956: .B \-fsigned\-char
957: Let the type \c
958: .B char\c
959: \& be signed, like \c
960: .B signed char\c
961: \&.
1.1.1.6 root 962: .Sp
1.1 root 963: Note that this is equivalent to `\|\c
964: .B \-fno\-unsigned\-char\c
965: \&\|', which is
966: the negative form of `\|\c
967: .B \-funsigned\-char\c
968: \&\|'. Likewise,
969: `\|\c
970: .B \-fno\-signed\-char\c
971: \&\|' is equivalent to `\|\c
972: .B \-funsigned\-char\c
973: \&\|'.
974: .TP
975: .B \-fsigned\-bitfields
976: .TP
977: .B \-funsigned\-bitfields
978: .TP
979: .B \-fno\-signed\-bitfields
980: .TP
981: .B \-fno\-unsigned\-bitfields
982: These options control whether a bitfield is
983: signed or unsigned, when declared with no explicit `\|\c
984: .B signed\c
985: \&\|' or `\|\c
986: .B unsigned\c
987: \&\|' qualifier. By default, such a bitfield is
988: signed, because this is consistent: the basic integer types such as
989: .B int\c
990: \& are signed types.
1.1.1.6 root 991: .Sp
1.1 root 992: However, when you specify `\|\c
993: .B \-traditional\c
994: \&\|', bitfields are all unsigned
995: no matter what.
996: .TP
997: .B \-fwritable\-strings
998: Store string constants in the writable data segment and don't uniquize
999: them. This is for compatibility with old programs which assume they
1000: can write into string constants. `\|\c
1001: .B \-traditional\c
1002: \&\|' also has this
1003: effect.
1.1.1.6 root 1004: .Sp
1005: Writing into string constants is a very bad idea; \*(lqconstants\*(rq should
1.1 root 1006: be constant.
1007: .SH PREPROCESSOR OPTIONS
1008: These options control the C preprocessor, which is run on each C source
1009: file before actual compilation.
1.1.1.6 root 1010: .PP
1.1 root 1011: If you use the `\|\c
1012: .B \-E\c
1013: \&\|' option, GCC does nothing except preprocessing.
1014: Some of these options make sense only together with `\|\c
1015: .B \-E\c
1016: \&\|' because
1017: they cause the preprocessor output to be unsuitable for actual
1018: compilation.
1019: .TP
1.1.1.5 root 1020: .BI "\-include " "file"
1021: Process \c
1022: .I file\c
1023: \& as input before processing the regular input file.
1024: In effect, the contents of \c
1025: .I file\c
1026: \& are compiled first. Any `\|\c
1027: .B \-D\c
1028: \&\|'
1029: and `\|\c
1030: .B \-U\c
1031: \&\|' options on the command line are always processed before
1032: `\|\c
1033: .B \-include \c
1034: .I file\c
1035: \&\c
1036: \&\|', regardless of the order in which they are
1037: written. All the `\|\c
1038: .B \-include\c
1039: \&\|' and `\|\c
1040: .B \-imacros\c
1041: \&\|' options are
1042: processed in the order in which they are written.
1043: .TP
1044: .BI "\-imacros " file
1.1 root 1045: Process \c
1046: .I file\c
1047: \& as input, discarding the resulting output, before
1048: processing the regular input file. Because the output generated from
1049: .I file\c
1050: \& is discarded, the only effect of `\|\c
1.1.1.5 root 1051: .B \-imacros \c
1.1 root 1052: .I file\c
1053: \&\c
1054: \&\|' is to
1055: make the macros defined in \c
1056: .I file\c
1057: \& available for use in the main
1058: input. The preprocessor evaluates any `\|\c
1059: .B \-D\c
1060: \&\|' and `\|\c
1061: .B \-U\c
1062: \&\|' options
1063: on the command line before processing `\|\c
1.1.1.5 root 1064: .B \-imacros\c
1.1 root 1065: .I file\c
1.1.1.5 root 1066: \&\|', regardless of the order in
1067: which they are written. All the `\|\c
1068: .B \-include\c
1069: \&\|' and `\|\c
1070: .B \-imacros\c
1071: \&\|'
1072: options are processed in the order in which they are written.
1073: .TP
1.1.1.6 root 1074: .BI "\-idirafter " "dir"
1.1.1.5 root 1075: Add the directory \c
1076: .I dir\c
1077: \& to the second include path. The directories
1078: on the second include path are searched when a header file is not found
1079: in any of the directories in the main include path (the one that
1080: `\|\c
1081: .B \-I\c
1082: \&\|' adds to).
1083: .TP
1.1.1.6 root 1084: .BI "\-iprefix " "prefix"
1.1.1.5 root 1085: Specify \c
1086: .I prefix\c
1087: \& as the prefix for subsequent `\|\c
1088: .B \-iwithprefix\c
1089: \&\|'
1090: options.
1091: .TP
1.1.1.6 root 1092: .BI "\-iwithprefix " "dir"
1.1.1.5 root 1093: Add a directory to the second include path. The directory's name is
1094: made by concatenating \c
1095: .I prefix\c
1096: \& and \c
1097: .I dir\c
1098: \&, where \c
1.1.1.6 root 1099: .I prefix
1.1.1.5 root 1100: was specified previously with `\|\c
1101: .B \-iprefix\c
1102: \&\|'.
1.1 root 1103: .TP
1104: .B \-nostdinc
1105: Do not search the standard system directories for header files. Only
1106: the directories you have specified with `\|\c
1107: .B \-I\c
1108: \&\|' options (and the
1.1.1.6 root 1109: current directory, if appropriate) are searched.
1110: .Sp
1.1 root 1111: By using both `\|\c
1112: .B \-nostdinc\c
1113: \&\|' and `\|\c
1114: .B \-I\-\c
1115: \&\|', you can limit the include-file search file to only those
1116: directories you specify explicitly.
1117: .TP
1.1.1.5 root 1118: .B \-nostdinc++
1.1.1.6 root 1119: Do not search for header files in the C++\-specific standard directories,
1.1.1.5 root 1120: but do still search the other standard directories.
1121: (This option is used when building `\|\c
1122: .B libg++\c
1123: \&\|'.)
1124: .TP
1.1 root 1125: .B \-undef
1126: Do not predefine any nonstandard macros. (Including architecture flags).
1127: .TP
1128: .B \-E
1129: Run only the C preprocessor. Preprocess all the C source files
1130: specified and output the results to standard output or to the
1131: specified output file.
1132: .TP
1133: .B \-C
1134: Tell the preprocessor not to discard comments. Used with the
1135: `\|\c
1136: .B \-E\c
1137: \&\|' option.
1138: .TP
1139: .B \-P
1140: Tell the preprocessor not to generate `\|\c
1141: .B #line\c
1142: \&\|' commands.
1143: Used with the `\|\c
1144: .B \-E\c
1145: \&\|' option.
1146: .TP
1.1.1.7 root 1147: .B \-M\ [ \-MG ]
1.1 root 1148: Tell the preprocessor to output a rule suitable for \c
1.1.1.6 root 1149: .B make
1.1 root 1150: describing the dependencies of each object file. For each source file,
1151: the preprocessor outputs one \c
1152: .B make\c
1153: \&-rule whose target is the object
1154: file name for that source file and whose dependencies are all the files
1155: `\|\c
1156: .B #include\c
1157: \&\|'d in it. This rule may be a single line or may be
1158: continued with `\|\c
1159: .B \e\c
1160: \&\|'-newline if it is long. The list of rules is
1161: printed on standard output instead of the preprocessed C program.
1.1.1.6 root 1162: .Sp
1.1 root 1163: `\|\c
1164: .B \-M\c
1165: \&\|' implies `\|\c
1166: .B \-E\c
1167: \&\|'.
1.1.1.7 root 1168: .Sp
1169: `\|\c
1170: .B \-MG\c
1171: \&\|' says to treat missing header files as generated files and assume \c
1172: they live in the same directory as the source file. It must be specified \c
1173: in addition to `\|\c
1174: .B \-M\c
1175: \&\|'.
1.1 root 1176: .TP
1.1.1.7 root 1177: .B \-MM\ [ \-MG ]
1.1 root 1178: Like `\|\c
1179: .B \-M\c
1180: \&\|' but the output mentions only the user header files
1181: included with `\|\c
1182: .B #include "\c
1183: .I file\c
1184: \&"\c
1185: \&\|'. System header files
1186: included with `\|\c
1187: .B #include <\c
1188: .I file\c
1189: \&>\c
1190: \&\|' are omitted.
1191: .TP
1192: .B \-MD
1193: Like `\|\c
1194: .B \-M\c
1195: \&\|' but the dependency information is written to files with
1196: names made by replacing `\|\c
1.1.1.5 root 1197: .B .o\c
1.1 root 1198: \&\|' with `\|\c
1199: .B .d\c
1200: \&\|' at the end of the
1.1.1.5 root 1201: output file names. This is in addition to compiling the file as
1.1 root 1202: specified\(em\&`\|\c
1203: .B \-MD\c
1204: \&\|' does not inhibit ordinary compilation the way
1205: `\|\c
1206: .B \-M\c
1207: \&\|' does.
1.1.1.6 root 1208: .Sp
1.1 root 1209: The Mach utility `\|\c
1210: .B md\c
1211: \&\|' can be used to merge the `\|\c
1212: .B .d\c
1213: \&\|' files
1214: into a single dependency file suitable for using with the `\|\c
1215: .B make\c
1216: \&\|'
1217: command.
1218: .TP
1219: .B \-MMD
1220: Like `\|\c
1221: .B \-MD\c
1222: \&\|' except mention only user header files, not system
1223: header files.
1224: .TP
1225: .B \-H
1226: Print the name of each header file used, in addition to other normal
1227: activities.
1228: .TP
1.1.1.6 root 1229: .BI "\-A" "question" ( answer )
1.1.1.5 root 1230: Assert the answer
1231: .I answer
1232: for
1233: .I question\c
1234: \&, in case it is tested
1235: with a preprocessor conditional such as `\|\c
1236: .BI "#if #" question ( answer )\c
1237: \&\|'. `\|\c
1238: .B \-A\-\c
1239: \&\|' disables the standard
1240: assertions that normally describe the target machine.
1241: .TP
1.1.1.6 root 1242: .BI "\-A" "question"\c
1.1.1.5 root 1243: \&(\c
1244: .I answer\c
1245: \&)
1246: Assert the answer \c
1247: .I answer\c
1248: \& for \c
1249: .I question\c
1250: \&, in case it is tested
1251: with a preprocessor conditional such as `\|\c
1252: .B #if
1253: #\c
1254: .I question\c
1255: \&(\c
1256: .I answer\c
1257: \&)\c
1258: \&\|'. `\|\c
1259: .B \-A-\c
1260: \&\|' disables the standard
1261: assertions that normally describe the target machine.
1262: .TP
1.1 root 1263: .BI \-D macro
1264: Define macro \c
1265: .I macro\c
1266: \& with the string `\|\c
1267: .B 1\c
1268: \&\|' as its definition.
1269: .TP
1270: .BI \-D macro = defn
1271: Define macro \c
1272: .I macro\c
1273: \& as \c
1274: .I defn\c
1275: \&. All instances of `\|\c
1276: .B \-D\c
1277: \&\|' on
1278: the command line are processed before any `\|\c
1279: .B \-U\c
1280: \&\|' options.
1281: .TP
1282: .BI \-U macro
1283: Undefine macro \c
1284: .I macro\c
1285: \&. `\|\c
1286: .B \-U\c
1287: \&\|' options are evaluated after all `\|\c
1288: .B \-D\c
1289: \&\|' options, but before any `\|\c
1.1.1.5 root 1290: .B \-include\c
1291: \&\|' and `\|\c
1292: .B \-imacros\c
1.1 root 1293: \&\|' options.
1294: .TP
1295: .B \-dM
1296: Tell the preprocessor to output only a list of the macro definitions
1297: that are in effect at the end of preprocessing. Used with the `\|\c
1298: .B \-E\c
1299: \&\|'
1300: option.
1301: .TP
1302: .B \-dD
1.1.1.5 root 1303: Tell the preprocessor to pass all macro definitions into the output, in
1.1 root 1304: their proper sequence in the rest of the output.
1305: .TP
1306: .B \-dN
1307: Like `\|\c
1308: .B \-dD\c
1309: \&\|' except that the macro arguments and contents are omitted.
1310: Only `\|\c
1311: .B #define \c
1312: .I name\c
1313: \&\c
1314: \&\|' is included in the output.
1.1.1.5 root 1315: .SH ASSEMBLER OPTION
1316: .TP
1.1.1.6 root 1317: .BI "\-Wa," "option"
1.1.1.5 root 1318: Pass \c
1319: .I option\c
1320: \& as an option to the assembler. If \c
1.1.1.6 root 1321: .I option
1.1.1.5 root 1322: contains commas, it is split into multiple options at the commas.
1.1 root 1323: .SH LINKER OPTIONS
1324: These options come into play when the compiler links object files into
1325: an executable output file. They are meaningless if the compiler is
1326: not doing a link step.
1327: .TP
1328: .I object-file-name
1329: A file name that does not end in a special recognized suffix is
1330: considered to name an object file or library. (Object files are
1331: distinguished from libraries by the linker according to the file
1332: contents.) If GCC does a link step, these object files are used as input
1333: to the linker.
1334: .TP
1.1.1.6 root 1335: .BI \-l library
1.1 root 1336: Use the library named \c
1337: .I library\c
1338: \& when linking.
1.1.1.6 root 1339: .Sp
1.1 root 1340: The linker searches a standard list of directories for the library,
1341: which is actually a file named `\|\c
1342: .B lib\c
1343: .I library\c
1344: \&.a\c
1345: \&\|'. The linker
1346: then uses this file as if it had been specified precisely by name.
1.1.1.6 root 1347: .Sp
1.1 root 1348: The directories searched include several standard system directories
1349: plus any that you specify with `\|\c
1350: .B \-L\c
1351: \&\|'.
1.1.1.6 root 1352: .Sp
1.1 root 1353: Normally the files found this way are library files\(em\&archive files
1354: whose members are object files. The linker handles an archive file by
1355: scanning through it for members which define symbols that have so far
1356: been referenced but not defined. However, if the linker finds an
1357: ordinary object file rather than a library, the object file is linked
1358: in the usual fashion. The only difference between using an `\|\c
1359: .B \-l\c
1360: \&\|' option and specifying a file
1361: name is that `\|\c
1362: .B \-l\c
1.1.1.6 root 1363: \&\|' surrounds
1.1 root 1364: .I library
1365: with `\|\c
1366: .B lib\c
1367: \&\|' and `\|\c
1368: .B .a\c
1369: \&\|' and searches several directories.
1370: .TP
1.1.1.5 root 1371: .B \-lobjc
1372: You need this special case of the
1373: .B \-l
1374: option in order to link an Objective C program.
1375: .TP
1376: .B \-nostartfiles
1377: Do not use the standard system startup files when linking.
1378: The standard libraries are used normally.
1379: .TP
1.1 root 1380: .B \-nostdlib
1381: Don't use the standard system libraries and startup files when linking.
1382: Only the files you specify will be passed to the linker.
1383: .TP
1384: .B \-static
1385: On systems that support dynamic linking, this prevents linking with the shared
1.1.1.4 root 1386: libraries. On other systems, this option has no effect.
1.1.1.5 root 1387: .TP
1388: .B \-shared
1389: Produce a shared object which can then be linked with other objects to
1390: form an executable. Only a few systems support this option.
1391: .TP
1392: .B \-symbolic
1393: Bind references to global symbols when building a shared object. Warn
1394: about any unresolved references (unless overridden by the link editor
1395: option `\|\c
1396: .B
1397: \-Xlinker \-z \-Xlinker defs\c
1398: \&\|'). Only a few systems support
1399: this option.
1400: .TP
1.1.1.6 root 1401: .BI "\-Xlinker " "option"
1.1.1.5 root 1402: Pass \c
1403: .I option
1404: as an option to the linker. You can use this to
1405: supply system-specific linker options which GNU CC does not know how to
1406: recognize.
1.1.1.6 root 1407: .Sp
1.1.1.5 root 1408: If you want to pass an option that takes an argument, you must use
1409: `\|\c
1410: .B \-Xlinker\c
1411: \&\|' twice, once for the option and once for the argument.
1412: For example, to pass `\|\c
1413: .B
1414: \-assert definitions\c
1415: \&\|', you must write
1416: `\|\c
1417: .B
1418: \-Xlinker \-assert \-Xlinker definitions\c
1419: \&\|'. It does not work to write
1420: `\|\c
1421: .B
1.1.1.6 root 1422: \-Xlinker "\-assert definitions"\c
1.1.1.5 root 1423: \&\|', because this passes the entire
1424: string as a single argument, which is not what the linker expects.
1425: .TP
1.1.1.6 root 1426: .BI "\-Wl," "option"
1.1.1.5 root 1427: Pass \c
1428: .I option\c
1429: \& as an option to the linker. If \c
1430: .I option\c
1431: \& contains
1432: commas, it is split into multiple options at the commas.
1433: .TP
1.1.1.6 root 1434: .BI "\-u " "symbol"
1.1.1.5 root 1435: Pretend the symbol
1436: .I symbol
1437: is undefined, to force linking of
1438: library modules to define it. You can use `\|\c
1439: .B \-u\c
1440: \&\|' multiple times with
1441: different symbols to force loading of additional library modules.
1.1 root 1442: .SH DIRECTORY OPTIONS
1443: These options specify directories to search for header files, for
1444: libraries and for parts of the compiler:
1445: .TP
1.1.1.6 root 1446: .BI "\-I" "dir"
1.1 root 1447: Append directory \c
1448: .I dir\c
1449: \& to the list of directories searched for include files.
1450: .TP
1451: .B \-I\-
1452: Any directories you specify with `\|\c
1453: .B \-I\c
1454: \&\|' options before the `\|\c
1455: .B \-I\-\c
1456: \&\|'
1457: option are searched only for the case of `\|\c
1.1.1.6 root 1458: .B
1.1 root 1459: #include "\c
1460: .I file\c
1461: .B
1462: \&"\c
1463: \&\|';
1464: they are not searched for `\|\c
1465: .B #include <\c
1466: .I file\c
1467: \&>\c
1468: \&\|'.
1.1.1.6 root 1469: .Sp
1.1 root 1470: If additional directories are specified with `\|\c
1471: .B \-I\c
1472: \&\|' options after
1473: the `\|\c
1474: .B \-I\-\c
1475: \&\|', these directories are searched for all `\|\c
1476: .B #include\c
1477: \&\|'
1478: directives. (Ordinarily \c
1479: .I all\c
1480: \& `\|\c
1481: .B \-I\c
1482: \&\|' directories are used
1483: this way.)
1.1.1.6 root 1484: .Sp
1.1 root 1485: In addition, the `\|\c
1486: .B \-I\-\c
1487: \&\|' option inhibits the use of the current
1488: directory (where the current input file came from) as the first search
1489: directory for `\|\c
1.1.1.6 root 1490: .B
1.1 root 1491: #include "\c
1492: .I file\c
1493: .B
1494: \&"\c
1495: \&\|'. There is no way to
1496: override this effect of `\|\c
1497: .B \-I\-\c
1498: \&\|'. With `\|\c
1499: .B \-I.\c
1500: \&\|' you can specify
1501: searching the directory which was current when the compiler was
1502: invoked. That is not exactly the same as what the preprocessor does
1503: by default, but it is often satisfactory.
1.1.1.6 root 1504: .Sp
1.1 root 1505: `\|\c
1506: .B \-I\-\c
1507: \&\|' does not inhibit the use of the standard system directories
1508: for header files. Thus, `\|\c
1509: .B \-I\-\c
1510: \&\|' and `\|\c
1511: .B \-nostdinc\c
1512: \&\|' are
1513: independent.
1514: .TP
1.1.1.6 root 1515: .BI "\-L" "dir"
1.1 root 1516: Add directory \c
1517: .I dir\c
1518: \& to the list of directories to be searched
1519: for `\|\c
1520: .B \-l\c
1521: \&\|'.
1522: .TP
1.1.1.6 root 1523: .BI "\-B" "prefix"
1.1 root 1524: This option specifies where to find the executables, libraries and
1525: data files of the compiler itself.
1.1.1.6 root 1526: .Sp
1.1 root 1527: The compiler driver program runs one or more of the subprograms
1528: `\|\c
1529: .B cpp\c
1530: \&\|', `\|\c
1531: .B cc1\c
1532: \&\|' (or, for C++, `\|\c
1533: .B cc1plus\c
1534: \&\|'), `\|\c
1535: .B as\c
1536: \&\|' and `\|\c
1537: .B ld\c
1538: \&\|'. It tries
1539: .I prefix\c
1540: \& as a prefix for each program it tries to run, both with and
1541: without `\|\c
1542: .I machine\c
1.1.1.6 root 1543: .B /\c
1.1 root 1544: .I version\c
1.1.1.6 root 1545: .B /\c
1.1 root 1546: \&\|'.
1.1.1.6 root 1547: .Sp
1.1 root 1548: For each subprogram to be run, the compiler driver first tries the
1549: `\|\c
1550: .B \-B\c
1551: \&\|' prefix, if any. If that name is not found, or if `\|\c
1552: .B \-B\c
1553: \&\|'
1554: was not specified, the driver tries two standard prefixes, which are
1555: `\|\c
1556: .B /usr/lib/gcc/\c
1557: \&\|' and `\|\c
1.1.1.2 root 1558: .B /usr/local/lib/gcc-lib/\c
1.1 root 1559: \&\|'. If neither of
1560: those results in a file name that is found, the compiler driver
1561: searches for the unmodified program
1562: name, using the directories specified in your
1563: `\|\c
1564: .B PATH\c
1565: \&\|' environment variable.
1.1.1.6 root 1566: .Sp
1.1 root 1567: The run-time support file `\|\c
1568: .B libgcc.a\c
1569: \&\|' is also searched for using the
1570: `\|\c
1571: .B \-B\c
1572: \&\|' prefix, if needed. If it is not found there, the two
1573: standard prefixes above are tried, and that is all. The file is left
1574: out of the link if it is not found by those means. Most of the time,
1575: on most machines, `\|\c
1576: .B libgcc.a\c
1577: \&\|' is not actually necessary.
1.1.1.6 root 1578: .Sp
1.1 root 1579: You can get a similar result from the environment variable
1580: .B GCC_EXEC_PREFIX\c
1581: \&; if it is defined, its value is used as a prefix
1582: in the same way. If both the `\|\c
1583: .B \-B\c
1584: \&\|' option and the
1585: .B GCC_EXEC_PREFIX\c
1586: \& variable are present, the `\|\c
1587: .B \-B\c
1588: \&\|' option is
1589: used first and the environment variable value second.
1590: .SH WARNING OPTIONS
1591: Warnings are diagnostic messages that report constructions which
1592: are not inherently erroneous but which are risky or suggest there
1593: may have been an error.
1.1.1.6 root 1594: .Sp
1.1 root 1595: These options control the amount and kinds of warnings produced by GNU
1596: CC:
1597: .TP
1598: .B \-fsyntax\-only
1599: Check the code for syntax errors, but don't emit any output.
1600: .TP
1601: .B \-w
1602: Inhibit all warning messages.
1603: .TP
1.1.1.5 root 1604: .B \-Wno\-import
1605: Inhibit warning messages about the use of
1606: .BR #import .
1607: .TP
1.1 root 1608: .B \-pedantic
1609: Issue all the warnings demanded by strict ANSI standard C; reject
1610: all programs that use forbidden extensions.
1.1.1.6 root 1611: .Sp
1.1 root 1612: Valid ANSI standard C programs should compile properly with or without
1613: this option (though a rare few will require `\|\c
1614: .B \-ansi\c
1615: \&\|'). However,
1616: without this option, certain GNU extensions and traditional C features
1617: are supported as well. With this option, they are rejected. There is
1618: no reason to \c
1619: .I use\c
1620: \& this option; it exists only to satisfy pedants.
1.1.1.6 root 1621: .Sp
1.1 root 1622: `\|\c
1623: .B \-pedantic\c
1624: \&\|' does not cause warning messages for use of the
1625: alternate keywords whose names begin and end with `\|\c
1.1.1.6 root 1626: .B _\|_\c
1.1 root 1627: \&\|'. Pedantic
1628: warnings are also disabled in the expression that follows
1.1.1.6 root 1629: .B _\|_extension_\|_\c
1.1 root 1630: \&. However, only system header files should use
1631: these escape routes; application programs should avoid them.
1632: .TP
1633: .B \-pedantic\-errors
1634: Like `\|\c
1635: .B \-pedantic\c
1636: \&\|', except that errors are produced rather than
1637: warnings.
1638: .TP
1639: .B \-W
1640: Print extra warning messages for these events:
1641: .TP
1642: \ \ \ \(bu
1643: A nonvolatile automatic variable might be changed by a call to
1644: .B longjmp\c
1645: \&. These warnings are possible only in
1646: optimizing compilation.
1.1.1.6 root 1647: .Sp
1.1 root 1648: The compiler sees only the calls to \c
1649: .B setjmp\c
1650: \&. It cannot know
1651: where \c
1652: .B longjmp\c
1653: \& will be called; in fact, a signal handler could
1654: call it at any point in the code. As a result, you may get a warning
1655: even when there is in fact no problem because \c
1656: .B longjmp\c
1657: \& cannot
1658: in fact be called at the place which would cause a problem.
1659: .TP
1660: \ \ \ \(bu
1661: A function can return either with or without a value. (Falling
1662: off the end of the function body is considered returning without
1663: a value.) For example, this function would evoke such a
1664: warning:
1.1.1.6 root 1665: .Sp
1666: .nf
1667: foo (a)
1.1 root 1668: {
1.1.1.6 root 1669: if (a > 0)
1670: return a;
1.1 root 1671: }
1.1.1.6 root 1672: .Sp
1673: .fi
1.1 root 1674: Spurious warnings can occur because GNU CC does not realize that
1675: certain functions (including \c
1676: .B abort\c
1677: \& and \c
1678: .B longjmp\c
1679: \&)
1680: will never return.
1681: .TP
1682: \ \ \ \(bu
1.1.1.8 ! root 1683: An expression-statement or the left-hand side of a comma expression
! 1684: contains no side effects.
! 1685: To suppress the warning, cast the unused expression to void.
! 1686: For example, an expression such as `\|\c
! 1687: .B x[i,j]\c
! 1688: \&\|' will cause a warning,
! 1689: but `\|\c
! 1690: .B x[(void)i,j]\c
! 1691: \&\|' will not.
1.1 root 1692: .TP
1693: \ \ \ \(bu
1694: An unsigned value is compared against zero with `\|\c
1695: .B >\c
1696: \&\|' or `\|\c
1697: .B <=\c
1698: \&\|'.
1699: .PP
1700: .TP
1701: .B \-Wimplicit
1702: Warn whenever a function or parameter is implicitly declared.
1703: .TP
1704: .B \-Wreturn\-type
1705: Warn whenever a function is defined with a return-type that defaults
1706: to \c
1707: .B int\c
1708: \&. Also warn about any \c
1709: .B return\c
1710: \& statement with no
1711: return-value in a function whose return-type is not \c
1712: .B void\c
1713: \&.
1714: .TP
1715: .B \-Wunused
1716: Warn whenever a local variable is unused aside from its declaration,
1717: whenever a function is declared static but never defined, and whenever
1718: a statement computes a result that is explicitly not used.
1719: .TP
1720: .B \-Wswitch
1721: Warn whenever a \c
1722: .B switch\c
1723: \& statement has an index of enumeral type
1724: and lacks a \c
1725: .B case\c
1726: \& for one or more of the named codes of that
1727: enumeration. (The presence of a \c
1728: .B default\c
1729: \& label prevents this
1730: warning.) \c
1731: .B case\c
1732: \& labels outside the enumeration range also
1733: provoke warnings when this option is used.
1734: .TP
1735: .B \-Wcomment
1736: Warn whenever a comment-start sequence `\|\c
1.1.1.6 root 1737: .B /\(**\c
1.1 root 1738: \&\|' appears in a comment.
1739: .TP
1740: .B \-Wtrigraphs
1741: Warn if any trigraphs are encountered (assuming they are enabled).
1742: .TP
1743: .B \-Wformat
1744: Check calls to \c
1745: .B printf\c
1746: \& and \c
1747: .B scanf\c
1748: \&, etc., to make sure that
1749: the arguments supplied have types appropriate to the format string
1750: specified.
1751: .TP
1.1.1.5 root 1752: .B \-Wchar\-subscripts
1753: Warn if an array subscript has type
1754: .BR char .
1755: This is a common cause of error, as programmers often forget that this
1756: type is signed on some machines.
1757: .TP
1.1 root 1758: .B \-Wuninitialized
1759: An automatic variable is used without first being initialized.
1.1.1.6 root 1760: .Sp
1.1 root 1761: These warnings are possible only in optimizing compilation,
1762: because they require data flow information that is computed only
1763: when optimizing. If you don't specify `\|\c
1764: .B \-O\c
1765: \&\|', you simply won't
1766: get these warnings.
1.1.1.6 root 1767: .Sp
1.1 root 1768: These warnings occur only for variables that are candidates for
1769: register allocation. Therefore, they do not occur for a variable that
1770: is declared \c
1771: .B volatile\c
1772: \&, or whose address is taken, or whose size
1773: is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
1774: structures, unions or arrays, even when they are in registers.
1.1.1.6 root 1775: .Sp
1.1 root 1776: Note that there may be no warning about a variable that is used only
1777: to compute a value that itself is never used, because such
1778: computations may be deleted by data flow analysis before the warnings
1779: are printed.
1.1.1.6 root 1780: .Sp
1.1 root 1781: These warnings are made optional because GNU CC is not smart
1782: enough to see all the reasons why the code might be correct
1783: despite appearing to have an error. Here is one example of how
1784: this can happen:
1.1.1.6 root 1785: .Sp
1786: .nf
1.1 root 1787: {
1.1.1.6 root 1788: int x;
1789: switch (y)
1790: {
1791: case 1: x = 1;
1792: break;
1793: case 2: x = 4;
1794: break;
1795: case 3: x = 5;
1796: }
1797: foo (x);
1.1 root 1798: }
1.1.1.6 root 1799: .Sp
1800: .fi
1.1 root 1801: If the value of \c
1802: .B y\c
1803: \& is always 1, 2 or 3, then \c
1804: .B x\c
1805: \& is
1806: always initialized, but GNU CC doesn't know this. Here is
1807: another common case:
1.1.1.6 root 1808: .Sp
1809: .nf
1.1 root 1810: {
1.1.1.6 root 1811: int save_y;
1812: if (change_y) save_y = y, y = new_y;
1813: .\|.\|.
1814: if (change_y) y = save_y;
1.1 root 1815: }
1.1.1.6 root 1816: .Sp
1817: .fi
1.1 root 1818: This has no bug because \c
1819: .B save_y\c
1820: \& is used only if it is set.
1.1.1.6 root 1821: .Sp
1.1 root 1822: Some spurious warnings can be avoided if you declare as
1823: .B volatile\c
1824: \& all the functions you use that never return.
1825: .TP
1.1.1.2 root 1826: .B \-Wparentheses
1827: Warn if parentheses are omitted in certain contexts.
1.1 root 1828: .TP
1.1.1.5 root 1829: .B \-Wtemplate\-debugging
1830: When using templates in a C++ program, warn if debugging is not yet
1831: fully available (C++ only).
1832: .TP
1.1 root 1833: .B \-Wall
1834: All of the above `\|\c
1835: .B \-W\c
1836: \&\|' options combined. These are all the
1837: options which pertain to usage that we recommend avoiding and that we
1838: believe is easy to avoid, even in conjunction with macros.
1839: .PP
1840: The remaining `\|\c
1841: .B \-W.\|.\|.\c
1842: \&\|' options are not implied by `\|\c
1843: .B \-Wall\c
1844: \&\|'
1845: because they warn about constructions that we consider reasonable to
1846: use, on occasion, in clean programs.
1847: .TP
1848: .B \-Wtraditional
1849: Warn about certain constructs that behave differently in traditional and
1850: ANSI C.
1851: .TP
1852: \ \ \ \(bu
1853: Macro arguments occurring within string constants in the macro body.
1854: These would substitute the argument in traditional C, but are part of
1855: the constant in ANSI C.
1856: .TP
1857: \ \ \ \(bu
1858: A function declared external in one block and then used after the end of
1859: the block.
1860: .TP
1861: \ \ \ \(bu
1862: A \c
1863: .B switch\c
1864: \& statement has an operand of type \c
1865: .B long\c
1866: \&.
1867: .PP
1868: .TP
1869: .B \-Wshadow
1870: Warn whenever a local variable shadows another local variable.
1871: .TP
1.1.1.6 root 1872: .BI "\-Wid\-clash\-" "len"
1.1 root 1873: Warn whenever two distinct identifiers match in the first \c
1.1.1.6 root 1874: .I len
1.1 root 1875: characters. This may help you prepare a program that will compile
1876: with certain obsolete, brain-damaged compilers.
1877: .TP
1878: .B \-Wpointer\-arith
1.1.1.6 root 1879: Warn about anything that depends on the \*(lqsize of\*(rq a function type or
1.1 root 1880: of \c
1881: .B void\c
1882: \&. GNU C assigns these types a size of 1, for
1883: convenience in calculations with \c
1.1.1.6 root 1884: .B void \(**\c
1.1 root 1885: \& pointers and pointers
1886: to functions.
1887: .TP
1888: .B \-Wcast\-qual
1889: Warn whenever a pointer is cast so as to remove a type qualifier from
1890: the target type. For example, warn if a \c
1.1.1.6 root 1891: .B const char \(**\c
1.1 root 1892: \& is cast
1893: to an ordinary \c
1.1.1.6 root 1894: .B char \(**\c
1.1 root 1895: \&.
1896: .TP
1897: .B \-Wcast\-align
1898: Warn whenever a pointer is cast such that the required alignment of the
1899: target is increased. For example, warn if a \c
1.1.1.6 root 1900: .B char \(**\c
1.1 root 1901: \& is cast to
1902: an \c
1.1.1.6 root 1903: .B int \(**\c
1.1 root 1904: \& on machines where integers can only be accessed at
1905: two- or four-byte boundaries.
1906: .TP
1907: .B \-Wwrite\-strings
1908: Give string constants the type \c
1909: .B const char[\c
1910: .I length\c
1.1.1.6 root 1911: .B ]\c
1.1 root 1912: \& so that
1913: copying the address of one into a non-\c
1914: .B const\c
1915: \& \c
1.1.1.6 root 1916: .B char \(**
1.1 root 1917: pointer will get a warning. These warnings will help you find at
1918: compile time code that can try to write into a string constant, but
1919: only if you have been very careful about using \c
1920: .B const\c
1921: \& in
1922: declarations and prototypes. Otherwise, it will just be a nuisance;
1923: this is why we did not make `\|\c
1924: .B \-Wall\c
1925: \&\|' request these warnings.
1926: .TP
1927: .B \-Wconversion
1928: Warn if a prototype causes a type conversion that is different from what
1929: would happen to the same argument in the absence of a prototype. This
1930: includes conversions of fixed point to floating and vice versa, and
1931: conversions changing the width or signedness of a fixed point argument
1932: except when the same as the default promotion.
1933: .TP
1.1.1.2 root 1934: .B \-Waggregate\-return
1935: Warn if any functions that return structures or unions are defined or
1936: called. (In languages where you can return an array, this also elicits
1937: a warning.)
1938: .TP
1939: .B \-Wstrict\-prototypes
1940: Warn if a function is declared or defined without specifying the
1941: argument types. (An old-style function definition is permitted without
1942: a warning if preceded by a declaration which specifies the argument
1943: types.)
1944: .TP
1945: .B \-Wmissing\-prototypes
1946: Warn if a global function is defined without a previous prototype
1947: declaration. This warning is issued even if the definition itself
1948: provides a prototype. The aim is to detect global functions that fail
1949: to be declared in header files.
1950: .TP
1.1.1.7 root 1951: .B \-Wmissing\-declarations
1952: Warn if a global function is defined without a previous declaration.
1953: Do so even if the definition itself provides a prototype.
1954: Use this option to detect global functions that are not declared in
1955: header files.
1956: .TP
1.1.1.5 root 1957: .B \-Wredundant-decls
1958: Warn if anything is declared more than once in the same scope, even in
1959: cases where multiple declaration is valid and changes nothing.
1960: .TP
1961: .B \-Wnested-externs
1962: Warn if an \c
1963: .B extern\c
1964: \& declaration is encountered within an function.
1965: .TP
1.1 root 1966: .B \-Wenum\-clash
1.1.1.5 root 1967: Warn about conversion between different enumeration types (C++ only).
1.1 root 1968: .TP
1969: .B \-Woverloaded\-virtual
1.1.1.5 root 1970: (C++ only.)
1.1 root 1971: In a derived class, the definitions of virtual functions must match
1972: the type signature of a virtual function declared in the base class.
1973: Use this option to request warnings when a derived class declares a
1974: function that may be an erroneous attempt to define a virtual
1975: function: that is, warn when a function with the same name as a
1976: virtual function in the base class, but with a type signature that
1977: doesn't match any virtual functions from the base class.
1.1.1.2 root 1978: .TP
1979: .B \-Winline
1980: Warn if a function can not be inlined, and either it was declared as inline,
1981: or else the
1982: .B \-finline\-functions
1983: option was given.
1984: .TP
1985: .B \-Werror
1986: Treat warnings as errors; abort compilation after any warning.
1.1 root 1987: .SH DEBUGGING OPTIONS
1988: GNU CC has various special options that are used for debugging
1989: either your program or GCC:
1990: .TP
1991: .B \-g
1992: Produce debugging information in the operating system's native format
1.1.1.5 root 1993: (stabs, COFF, XCOFF, or DWARF). GDB can work with this debugging
1994: information.
1.1.1.6 root 1995: .Sp
1.1.1.5 root 1996: On most systems that use stabs format, `\|\c
1.1 root 1997: .B \-g\c
1.1.1.5 root 1998: \&\|' enables use of extra
1999: debugging information that only GDB can use; this extra information
2000: makes debugging work better in GDB but will probably make other debuggers
2001: crash or
2002: refuse to read the program. If you want to control for certain whether
2003: to generate the extra information, use `\|\c
2004: .B \-gstabs+\c
2005: \&\|', `\|\c
2006: .B \-gstabs\c
2007: \&\|',
1.1 root 2008: `\|\c
1.1.1.5 root 2009: .B \-gxcoff+\c
2010: \&\|', `\|\c
2011: .B \-gxcoff\c
2012: \&\|', `\|\c
2013: .B \-gdwarf+\c
2014: \&\|', or `\|\c
2015: .B \-gdwarf\c
2016: \&\|'
2017: (see below).
1.1.1.6 root 2018: .Sp
1.1 root 2019: Unlike most other C compilers, GNU CC allows you to use `\|\c
2020: .B \-g\c
2021: \&\|' with
2022: `\|\c
2023: .B \-O\c
2024: \&\|'. The shortcuts taken by optimized code may occasionally
2025: produce surprising results: some variables you declared may not exist
2026: at all; flow of control may briefly move where you did not expect it;
2027: some statements may not be executed because they compute constant
2028: results or their values were already at hand; some statements may
2029: execute in different places because they were moved out of loops.
1.1.1.6 root 2030: .Sp
1.1 root 2031: Nevertheless it proves possible to debug optimized output. This makes
2032: it reasonable to use the optimizer for programs that might have bugs.
1.1.1.6 root 2033: .PP
1.1.1.5 root 2034: The following options are useful when GNU CC is generated with the
2035: capability for more than one debugging format.
1.1 root 2036: .TP
2037: .B \-ggdb
1.1.1.5 root 2038: Produce debugging information in the native format (if that is supported),
2039: including GDB extensions if at all possible.
1.1 root 2040: .TP
1.1.1.5 root 2041: .B \-gstabs
2042: Produce debugging information in stabs format (if that is supported),
2043: without GDB extensions. This is the format used by DBX on most BSD
2044: systems.
1.1 root 2045: .TP
1.1.1.5 root 2046: .B \-gstabs+
2047: Produce debugging information in stabs format (if that is supported),
2048: using GNU extensions understood only by the GNU debugger (GDB). The
2049: use of these extensions is likely to make other debuggers crash or
2050: refuse to read the program.
2051: .TP
2052: .B \-gcoff
2053: Produce debugging information in COFF format (if that is supported).
2054: This is the format used by SDB on most System V systems prior to
2055: System V Release 4.
2056: .TP
2057: .B \-gxcoff
2058: Produce debugging information in XCOFF format (if that is supported).
2059: This is the format used by the DBX debugger on IBM RS/6000 systems.
2060: .TP
2061: .B \-gxcoff+
2062: Produce debugging information in XCOFF format (if that is supported),
2063: using GNU extensions understood only by the GNU debugger (GDB). The
2064: use of these extensions is likely to make other debuggers crash or
2065: refuse to read the program.
1.1 root 2066: .TP
2067: .B \-gdwarf
2068: Produce debugging information in DWARF format (if that is supported).
1.1.1.5 root 2069: This is the format used by SDB on most System V Release 4 systems.
2070: .TP
2071: .B \-gdwarf+
2072: Produce debugging information in DWARF format (if that is supported),
2073: using GNU extensions understood only by the GNU debugger (GDB). The
2074: use of these extensions is likely to make other debuggers crash or
2075: refuse to read the program.
1.1 root 2076: .PP
2077: .BI "\-g" "level"
2078: .br
2079: .BI "\-ggdb" "level"
2080: .br
1.1.1.5 root 2081: .BI "\-gstabs" "level"
1.1 root 2082: .br
1.1.1.5 root 2083: .BI "\-gcoff" "level"
2084: .BI "\-gxcoff" "level"
1.1 root 2085: .TP
2086: .BI "\-gdwarf" "level"
2087: Request debugging information and also use \c
2088: .I level\c
2089: \& to specify how
2090: much information. The default level is 2.
1.1.1.6 root 2091: .Sp
1.1 root 2092: Level 1 produces minimal information, enough for making backtraces in
2093: parts of the program that you don't plan to debug. This includes
2094: descriptions of functions and external variables, but no information
2095: about local variables and no line numbers.
1.1.1.6 root 2096: .Sp
1.1.1.5 root 2097: Level 3 includes extra information, such as all the macro definitions
2098: present in the program. Some debuggers support macro expansion when
2099: you use `\|\c
2100: .B \-g3\c
2101: \&\|'.
1.1 root 2102: .TP
2103: .B \-p
2104: Generate extra code to write profile information suitable for the
2105: analysis program \c
2106: .B prof\c
2107: \&.
2108: .TP
2109: .B \-pg
2110: Generate extra code to write profile information suitable for the
2111: analysis program \c
2112: .B gprof\c
2113: \&.
2114: .TP
2115: .B \-a
2116: Generate extra code to write profile information for basic blocks,
2117: which will record the number of times each basic block is executed.
2118: This data could be analyzed by a program like \c
2119: .B tcov\c
2120: \&. Note,
2121: however, that the format of the data is not what \c
2122: .B tcov\c
2123: \& expects.
2124: Eventually GNU \c
2125: .B gprof\c
2126: \& should be extended to process this data.
2127: .TP
1.1.1.6 root 2128: .BI "\-d" "letters"
1.1 root 2129: Says to make debugging dumps during compilation at times specified by
2130: .I letters\c
2131: \&. This is used for debugging the compiler. The file names
2132: for most of the dumps are made by appending a word to the source file
2133: name (e.g. `\|\c
2134: .B foo.c.rtl\c
2135: \&\|' or `\|\c
2136: .B foo.c.jump\c
2137: \&\|').
2138: .TP
2139: .B \-dM
2140: Dump all macro definitions, at the end of preprocessing, and write no
2141: output.
2142: .TP
2143: .B \-dN
2144: Dump all macro names, at the end of preprocessing.
2145: .TP
2146: .B \-dD
2147: Dump all macro definitions, at the end of preprocessing, in addition to
2148: normal output.
2149: .TP
2150: .B \-dy
2151: Dump debugging information during parsing, to standard error.
2152: .TP
2153: .B \-dr
2154: Dump after RTL generation, to `\|\c
2155: .I file\c
1.1.1.6 root 2156: .B \&.rtl\c
1.1 root 2157: \&\|'.
2158: .TP
2159: .B \-dx
2160: Just generate RTL for a function instead of compiling it. Usually used
2161: with `\|\c
2162: .B r\c
2163: \&\|'.
2164: .TP
2165: .B \-dj
2166: Dump after first jump optimization, to `\|\c
2167: .I file\c
1.1.1.6 root 2168: .B \&.jump\c
1.1 root 2169: \&\|'.
2170: .TP
2171: .B \-ds
2172: Dump after CSE (including the jump optimization that sometimes
2173: follows CSE), to `\|\c
2174: .I file\c
1.1.1.6 root 2175: .B \&.cse\c
1.1 root 2176: \&\|'.
2177: .TP
2178: .B \-dL
2179: Dump after loop optimization, to `\|\c
2180: .I file\c
1.1.1.6 root 2181: .B \&.loop\c
1.1 root 2182: \&\|'.
2183: .TP
2184: .B \-dt
2185: Dump after the second CSE pass (including the jump optimization that
2186: sometimes follows CSE), to `\|\c
2187: .I file\c
1.1.1.6 root 2188: .B \&.cse2\c
1.1 root 2189: \&\|'.
2190: .TP
2191: .B \-df
2192: Dump after flow analysis, to `\|\c
2193: .I file\c
1.1.1.6 root 2194: .B \&.flow\c
1.1 root 2195: \&\|'.
2196: .TP
2197: .B \-dc
2198: Dump after instruction combination, to `\|\c
2199: .I file\c
1.1.1.6 root 2200: .B \&.combine\c
1.1 root 2201: \&\|'.
2202: .TP
2203: .B \-dS
2204: Dump after the first instruction scheduling pass, to
2205: `\|\c
2206: .I file\c
1.1.1.6 root 2207: .B \&.sched\c
1.1 root 2208: \&\|'.
2209: .TP
2210: .B \-dl
2211: Dump after local register allocation, to `\|\c
2212: .I file\c
1.1.1.6 root 2213: .B \&.lreg\c
1.1 root 2214: \&\|'.
2215: .TP
2216: .B \-dg
2217: Dump after global register allocation, to `\|\c
2218: .I file\c
1.1.1.6 root 2219: .B \&.greg\c
1.1 root 2220: \&\|'.
2221: .TP
2222: .B \-dR
2223: Dump after the second instruction scheduling pass, to
2224: `\|\c
2225: .I file\c
1.1.1.6 root 2226: .B \&.sched2\c
1.1 root 2227: \&\|'.
2228: .TP
2229: .B \-dJ
2230: Dump after last jump optimization, to `\|\c
2231: .I file\c
1.1.1.6 root 2232: .B \&.jump2\c
1.1 root 2233: \&\|'.
2234: .TP
2235: .B \-dd
2236: Dump after delayed branch scheduling, to `\|\c
2237: .I file\c
1.1.1.6 root 2238: .B \&.dbr\c
1.1 root 2239: \&\|'.
2240: .TP
2241: .B \-dk
2242: Dump after conversion from registers to stack, to `\|\c
2243: .I file\c
1.1.1.6 root 2244: .B \&.stack\c
1.1 root 2245: \&\|'.
2246: .TP
1.1.1.5 root 2247: .B \-da
2248: Produce all the dumps listed above.
2249: .TP
1.1 root 2250: .B \-dm
2251: Print statistics on memory usage, at the end of the run, to
2252: standard error.
2253: .TP
2254: .B \-dp
2255: Annotate the assembler output with a comment indicating which
2256: pattern and alternative was used.
2257: .TP
2258: .B \-fpretend\-float
2259: When running a cross-compiler, pretend that the target machine uses the
2260: same floating point format as the host machine. This causes incorrect
2261: output of the actual floating constants, but the actual instruction
2262: sequence will probably be the same as GNU CC would make when running on
2263: the target machine.
2264: .TP
2265: .B \-save\-temps
1.1.1.6 root 2266: Store the usual \*(lqtemporary\*(rq intermediate files permanently; place them
1.1 root 2267: in the current directory and name them based on the source file. Thus,
2268: compiling `\|\c
2269: .B foo.c\c
2270: \&\|' with `\|\c
2271: .B \-c \-save\-temps\c
2272: \&\|' would produce files
2273: `\|\c
2274: .B foo.cpp\c
2275: \&\|' and `\|\c
2276: .B foo.s\c
2277: \&\|', as well as `\|\c
2278: .B foo.o\c
2279: \&\|'.
1.1.1.5 root 2280: .TP
1.1.1.7 root 2281: .BI "\-print\-file\-name=" "library"
2282: Print the full absolute name of the library file \|\c
2283: .nh
2284: .I library
2285: .hy
2286: \&\| that
1.1.1.5 root 2287: would be used when linking\(em\&and do not do anything else. With this
2288: option, GNU CC does not compile or link anything; it just prints the
2289: file name.
1.1.1.7 root 2290: .TP
2291: .B \-print\-libgcc\-file\-name
2292: Same as `\|\c
2293: .B \-print\-file\-name=libgcc.a\c
2294: \&\|'.
2295: .TP
2296: .BI "\-print\-prog\-name=" "program"
2297: Like `\|\c
2298: .B \-print\-file\-name\c
2299: \&\|', but searches for a program such as `\|\c
2300: cpp\c
2301: \&\|'.
1.1 root 2302: .SH OPTIMIZATION OPTIONS
2303: These options control various sorts of optimizations:
1.1.1.6 root 2304: .TP
1.1 root 2305: .B \-O
1.1.1.5 root 2306: .TP
2307: .B \-O1
1.1 root 2308: Optimize. Optimizing compilation takes somewhat more time, and a lot
2309: more memory for a large function.
1.1.1.6 root 2310: .Sp
1.1 root 2311: Without `\|\c
2312: .B \-O\c
2313: \&\|', the compiler's goal is to reduce the cost of
2314: compilation and to make debugging produce the expected results.
2315: Statements are independent: if you stop the program with a breakpoint
2316: between statements, you can then assign a new value to any variable or
2317: change the program counter to any other statement in the function and
2318: get exactly the results you would expect from the source code.
1.1.1.6 root 2319: .Sp
1.1 root 2320: Without `\|\c
2321: .B \-O\c
2322: \&\|', only variables declared \c
2323: .B register\c
2324: \& are
2325: allocated in registers. The resulting compiled code is a little worse
2326: than produced by PCC without `\|\c
2327: .B \-O\c
2328: \&\|'.
1.1.1.6 root 2329: .Sp
1.1 root 2330: With `\|\c
2331: .B \-O\c
2332: \&\|', the compiler tries to reduce code size and execution
2333: time.
1.1.1.6 root 2334: .Sp
1.1 root 2335: When you specify `\|\c
2336: .B \-O\c
1.1.1.7 root 2337: \&\|', the two options `\|\c
1.1 root 2338: .B \-fthread\-jumps\c
1.1.1.7 root 2339: \&\|' and `\|\c
2340: .B \-fdefer\-pop\c
2341: \&\|' are turned on. On machines that have delay slots, the `\|\c
1.1 root 2342: .B \-fdelayed\-branch\c
1.1.1.7 root 2343: \&\|' option is turned on. For those machines that can support debugging even
2344: without a frame pointer, the `\|\c
2345: .B \-fomit\-frame\-pointer\c
2346: \&\|' option is turned on. On some machines other flags may also be turned on.
1.1 root 2347: .TP
2348: .B \-O2
1.1.1.5 root 2349: Optimize even more. Nearly all supported optimizations that do not
1.1.1.7 root 2350: involve a space-speed tradeoff are performed. Loop unrolling and function
2351: inlining are not done, for example. As compared to
1.1 root 2352: .B \-O\c
1.1.1.5 root 2353: \&,
2354: this option increases both compilation time and the performance of the
1.1 root 2355: generated code.
1.1.1.7 root 2356: .TP
2357: .B \-O3
2358: Optimize yet more. This turns on everything
1.1.1.5 root 2359: .B \-O2
1.1.1.7 root 2360: does, along with also turning on
2361: .B \-finline\-functions.
1.1.1.5 root 2362: .TP
2363: .B \-O0
2364: Do not optimize.
1.1.1.6 root 2365: .Sp
1.1.1.5 root 2366: If you use multiple
2367: .B \-O
2368: options, with or without level numbers, the last such option is the
2369: one that is effective.
1.1 root 2370: .PP
2371: Options of the form `\|\c
2372: .B \-f\c
2373: .I flag\c
2374: \&\c
2375: \&\|' specify machine-independent
2376: flags. Most flags have both positive and negative forms; the negative
2377: form of `\|\c
2378: .B \-ffoo\c
2379: \&\|' would be `\|\c
2380: .B \-fno\-foo\c
2381: \&\|'. The following list shows
2382: only one form\(em\&the one which is not the default.
2383: You can figure out the other form by either removing `\|\c
2384: .B no\-\c
2385: \&\|' or
2386: adding it.
2387: .TP
2388: .B \-ffloat\-store
2389: Do not store floating point variables in registers. This
2390: prevents undesirable excess precision on machines such as the
2391: 68000 where the floating registers (of the 68881) keep more
2392: precision than a \c
2393: .B double\c
2394: \& is supposed to have.
1.1.1.6 root 2395: .Sp
1.1 root 2396: For most programs, the excess precision does only good, but a few
2397: programs rely on the precise definition of IEEE floating point.
2398: Use `\|\c
2399: .B \-ffloat\-store\c
2400: \&\|' for such programs.
2401: .TP
2402: .B \-fmemoize\-lookups
2403: .TP
2404: .B \-fsave\-memoized
1.1.1.5 root 2405: Use heuristics to compile faster (C++ only). These heuristics are not
2406: enabled by default, since they are only effective for certain input
2407: files. Other input files compile more slowly.
1.1.1.6 root 2408: .Sp
1.1 root 2409: The first time the compiler must build a call to a member function (or
2410: reference to a data member), it must (1) determine whether the class
2411: implements member functions of that name; (2) resolve which member
2412: function to call (which involves figuring out what sorts of type
2413: conversions need to be made); and (3) check the visibility of the member
2414: function to the caller. All of this adds up to slower compilation.
2415: Normally, the second time a call is made to that member function (or
2416: reference to that data member), it must go through the same lengthy
2417: process again. This means that code like this
1.1.1.6 root 2418: .Sp
2419: \& cout << "This " << p << " has " << n << " legs.\en";
2420: .Sp
1.1 root 2421: makes six passes through all three steps. By using a software cache,
1.1.1.6 root 2422: a \*(lqhit\*(rq significantly reduces this cost. Unfortunately, using the
1.1 root 2423: cache introduces another layer of mechanisms which must be implemented,
2424: and so incurs its own overhead. `\|\c
2425: .B \-fmemoize\-lookups\c
2426: \&\|' enables
2427: the software cache.
1.1.1.6 root 2428: .Sp
1.1 root 2429: Because access privileges (visibility) to members and member functions
1.1.1.6 root 2430: may differ from one function context to the next,
1.1 root 2431: .B g++
1.1.1.5 root 2432: may need to flush the cache. With the `\|\c
1.1 root 2433: .B \-fmemoize\-lookups\c
2434: \&\|' flag, the cache is flushed after every
2435: function that is compiled. The `\|\c
2436: \-fsave\-memoized\c
2437: \&\|' flag enables the same software cache, but when the compiler
2438: determines that the context of the last function compiled would yield
2439: the same access privileges of the next function to compile, it
1.1.1.6 root 2440: preserves the cache.
1.1 root 2441: This is most helpful when defining many member functions for the same
2442: class: with the exception of member functions which are friends of
2443: other classes, each member function has exactly the same access
2444: privileges as every other, and the cache need not be flushed.
2445: .TP
2446: .B \-fno\-default\-inline
1.1.1.5 root 2447: Don't make member functions inline by default merely because they are
2448: defined inside the class scope (C++ only).
1.1 root 2449: .TP
2450: .B \-fno\-defer\-pop
2451: Always pop the arguments to each function call as soon as that
2452: function returns. For machines which must pop arguments after a
2453: function call, the compiler normally lets arguments accumulate on the
2454: stack for several function calls and pops them all at once.
2455: .TP
2456: .B \-fforce\-mem
2457: Force memory operands to be copied into registers before doing
2458: arithmetic on them. This may produce better code by making all
2459: memory references potential common subexpressions. When they are
2460: not common subexpressions, instruction combination should
2461: eliminate the separate register-load. I am interested in hearing
2462: about the difference this makes.
2463: .TP
2464: .B \-fforce\-addr
2465: Force memory address constants to be copied into registers before
2466: doing arithmetic on them. This may produce better code just as
2467: `\|\c
2468: .B \-fforce\-mem\c
2469: \&\|' may. I am interested in hearing about the
2470: difference this makes.
2471: .TP
2472: .B \-fomit\-frame\-pointer
2473: Don't keep the frame pointer in a register for functions that
2474: don't need one. This avoids the instructions to save, set up and
2475: restore frame pointers; it also makes an extra register available
2476: in many functions. \c
1.1.1.8 ! root 2477: .I It also makes debugging impossible on most machines\c
! 2478: \&.
1.1.1.6 root 2479: .Sp
1.1 root 2480: On some machines, such as the Vax, this flag has no effect, because
2481: the standard calling sequence automatically handles the frame pointer
2482: and nothing is saved by pretending it doesn't exist. The
2483: machine-description macro \c
2484: .B FRAME_POINTER_REQUIRED\c
2485: \& controls
1.1.1.6 root 2486: whether a target machine supports this flag.
1.1 root 2487: .TP
2488: .B \-finline\-functions
2489: Integrate all simple functions into their callers. The compiler
2490: heuristically decides which functions are simple enough to be worth
2491: integrating in this way.
1.1.1.6 root 2492: .Sp
1.1 root 2493: If all calls to a given function are integrated, and the function is
2494: declared \c
2495: .B static\c
2496: \&, then GCC normally does not output the function as
2497: assembler code in its own right.
2498: .TP
2499: .B \-fcaller\-saves
2500: Enable values to be allocated in registers that will be clobbered by
2501: function calls, by emitting extra instructions to save and restore the
2502: registers around such calls. Such allocation is done only when it
2503: seems to result in better code than would otherwise be produced.
1.1.1.6 root 2504: .Sp
1.1 root 2505: This option is enabled by default on certain machines, usually those
2506: which have no call-preserved registers to use instead.
2507: .TP
2508: .B \-fkeep\-inline\-functions
2509: Even if all calls to a given function are integrated, and the function
2510: is declared \c
2511: .B static\c
2512: \&, nevertheless output a separate run-time
2513: callable version of the function.
2514: .TP
2515: .B \-fno\-function\-cse
2516: Do not put function addresses in registers; make each instruction that
2517: calls a constant function contain the function's address explicitly.
1.1.1.6 root 2518: .Sp
1.1 root 2519: This option results in less efficient code, but some strange hacks
2520: that alter the assembler output may be confused by the optimizations
2521: performed when this option is not used.
1.1.1.5 root 2522: .TP
2523: .B \-fno\-peephole
2524: Disable any machine-specific peephole optimizations.
2525: .TP
2526: .B \-ffast-math
2527: This option allows GCC to violate some ANSI or IEEE rules/specifications
2528: in the interest of optimizing code for speed. For example, it allows
2529: the compiler to assume arguments to the \c
2530: .B sqrt\c
1.1.1.6 root 2531: \& function are
2532: non-negative numbers.
2533: .Sp
1.1.1.5 root 2534: This option should never be turned on by any `\|\c
2535: .B \-O\c
1.1.1.6 root 2536: \&\|' option since
2537: it can result in incorrect output for programs which depend on
1.1.1.5 root 2538: an exact implementation of IEEE or ANSI rules/specifications for
2539: math functions.
1.1 root 2540: .PP
2541: The following options control specific optimizations. The `\|\c
2542: .B \-O2\c
2543: \&\|'
2544: option turns on all of these optimizations except `\|\c
2545: .B \-funroll\-loops\c
2546: \&\|'
2547: and `\|\c
2548: .B \-funroll\-all\-loops\c
1.1.1.6 root 2549: \&\|'.
2550: .PP
1.1 root 2551: The `\|\c
2552: .B \-O\c
2553: \&\|' option usually turns on
2554: the `\|\c
2555: .B \-fthread\-jumps\c
2556: \&\|' and `\|\c
2557: .B \-fdelayed\-branch\c
2558: \&\|' options, but
2559: specific machines may change the default optimizations.
1.1.1.6 root 2560: .PP
2561: You can use the following flags in the rare cases when \*(lqfine-tuning\*(rq
1.1 root 2562: of optimizations to be performed is desired.
2563: .TP
2564: .B \-fstrength\-reduce
2565: Perform the optimizations of loop strength reduction and
2566: elimination of iteration variables.
2567: .TP
2568: .B \-fthread\-jumps
2569: Perform optimizations where we check to see if a jump branches to a
2570: location where another comparison subsumed by the first is found. If
2571: so, the first branch is redirected to either the destination of the
2572: second branch or a point immediately following it, depending on whether
2573: the condition is known to be true or false.
2574: .TP
2575: .B \-funroll\-loops
2576: Perform the optimization of loop unrolling. This is only done for loops
2577: whose number of iterations can be determined at compile time or run time.
2578: .TP
2579: .B \-funroll\-all\-loops
2580: Perform the optimization of loop unrolling. This is done for all loops.
2581: This usually makes programs run more slowly.
2582: .TP
2583: .B \-fcse\-follow\-jumps
1.1.1.5 root 2584: In common subexpression elimination, scan through jump instructions
2585: when the target of the jump is not reached by any other path. For
2586: example, when CSE encounters an \c
2587: .B if\c
2588: \& statement with an
2589: .B else\c
2590: \& clause, CSE will follow the jump when the condition
2591: tested is false.
2592: .TP
2593: .B \-fcse\-skip\-blocks
2594: This is similar to `\|\c
2595: .B \-fcse\-follow\-jumps\c
2596: \&\|', but causes CSE to
2597: follow jumps which conditionally skip over blocks. When CSE
2598: encounters a simple \c
2599: .B if\c
2600: \& statement with no else clause,
2601: `\|\c
2602: .B \-fcse\-skip\-blocks\c
2603: \&\|' causes CSE to follow the jump around the
2604: body of the \c
2605: .B if\c
2606: \&.
1.1 root 2607: .TP
2608: .B \-frerun\-cse\-after\-loop
2609: Re-run common subexpression elimination after loop optimizations has been
1.1.1.6 root 2610: performed.
1.1 root 2611: .TP
2612: .B \-felide\-constructors
1.1.1.5 root 2613: Elide constructors when this seems plausible (C++ only). With this
2614: flag, GNU C++ initializes \c
2615: .B y\c
2616: \& directly from the call to \c
1.1.1.6 root 2617: .B foo
1.1.1.5 root 2618: without going through a temporary in the following code:
1.1.1.6 root 2619: .Sp
2620: A foo ();
2621: A y = foo ();
2622: .Sp
1.1.1.5 root 2623: Without this option, GNU C++ first initializes \c
1.1 root 2624: .B y\c
1.1.1.5 root 2625: \& by calling the
2626: appropriate constructor for type \c
2627: .B A\c
2628: \&; then assigns the result of
2629: .B foo\c
1.1.1.8 ! root 2630: \& to a temporary; and, finally, replaces the initial value of
1.1.1.5 root 2631: `\|\c
2632: .B y\c
2633: \&\|' with the temporary.
1.1.1.6 root 2634: .Sp
1.1.1.5 root 2635: The default behavior (`\|\c
2636: .B \-fno\-elide\-constructors\c
2637: \&\|') is specified by
2638: the draft ANSI C++ standard. If your program's constructors have side
2639: effects, using `\|\c
2640: .B \-felide-constructors\c
2641: \&\|' can make your program act
2642: differently, since some constructor calls may be omitted.
1.1 root 2643: .TP
2644: .B \-fexpensive\-optimizations
2645: Perform a number of minor optimizations that are relatively expensive.
2646: .TP
2647: .B \-fdelayed\-branch
2648: If supported for the target machine, attempt to reorder instructions
2649: to exploit instruction slots available after delayed branch
2650: instructions.
2651: .TP
2652: .B \-fschedule\-insns
2653: If supported for the target machine, attempt to reorder instructions to
2654: eliminate execution stalls due to required data being unavailable. This
2655: helps machines that have slow floating point or memory load instructions
2656: by allowing other instructions to be issued until the result of the load
2657: or floating point instruction is required.
2658: .TP
2659: .B \-fschedule\-insns2
2660: Similar to `\|\c
2661: .B \-fschedule\-insns\c
2662: \&\|', but requests an additional pass of
2663: instruction scheduling after register allocation has been done. This is
2664: especially useful on machines with a relatively small number of
2665: registers and where memory load instructions take more than one cycle.
2666: .SH TARGET OPTIONS
2667: By default, GNU CC compiles code for the same type of machine that you
2668: are using. However, it can also be installed as a cross-compiler, to
2669: compile for some other type of machine. In fact, several different
2670: configurations of GNU CC, for different target machines, can be
2671: installed side by side. Then you specify which one to use with the
2672: `\|\c
2673: .B \-b\c
2674: \&\|' option.
1.1.1.6 root 2675: .PP
1.1 root 2676: In addition, older and newer versions of GNU CC can be installed side
2677: by side. One of them (probably the newest) will be the default, but
2678: you may sometimes wish to use another.
2679: .TP
1.1.1.6 root 2680: .BI "\-b " "machine"
1.1 root 2681: The argument \c
2682: .I machine\c
2683: \& specifies the target machine for compilation.
2684: This is useful when you have installed GNU CC as a cross-compiler.
1.1.1.6 root 2685: .Sp
1.1 root 2686: The value to use for \c
2687: .I machine\c
2688: \& is the same as was specified as the
2689: machine type when configuring GNU CC as a cross-compiler. For
2690: example, if a cross-compiler was configured with `\|\c
2691: .B configure
2692: i386v\c
2693: \&\|', meaning to compile for an 80386 running System V, then you
2694: would specify `\|\c
2695: .B \-b i386v\c
2696: \&\|' to run that cross compiler.
1.1.1.6 root 2697: .Sp
1.1 root 2698: When you do not specify `\|\c
2699: .B \-b\c
2700: \&\|', it normally means to compile for
2701: the same type of machine that you are using.
2702: .TP
1.1.1.6 root 2703: .BI "\-V " "version"
1.1 root 2704: The argument \c
2705: .I version\c
2706: \& specifies which version of GNU CC to run.
2707: This is useful when multiple versions are installed. For example,
2708: .I version\c
2709: \& might be `\|\c
2710: .B 2.0\c
2711: \&\|', meaning to run GNU CC version 2.0.
1.1.1.6 root 2712: .Sp
1.1 root 2713: The default version, when you do not specify `\|\c
2714: .B \-V\c
2715: \&\|', is controlled
2716: by the way GNU CC is installed. Normally, it will be a version that
2717: is recommended for general use.
2718: .SH MACHINE DEPENDENT OPTIONS
2719: Each of the target machine types can have its own special options,
2720: starting with `\|\c
2721: .B \-m\c
2722: \&\|', to choose among various hardware models or
2723: configurations\(em\&for example, 68010 vs 68020, floating coprocessor or
2724: none. A single installed version of the compiler can compile for any
2725: model or configuration, according to the options specified.
1.1.1.6 root 2726: .PP
1.1.1.5 root 2727: Some configurations of the compiler also support additional special
2728: options, usually for command-line compatibility with other compilers on
2729: the same platform.
1.1.1.6 root 2730: .PP
1.1 root 2731: These are the `\|\c
2732: .B \-m\c
2733: \&\|' options defined for the 68000 series:
2734: .TP
1.1.1.5 root 2735: .B \-m68000
2736: .TP
2737: .B \-mc68000
2738: Generate output for a 68000. This is the default when the compiler is
2739: configured for 68000-based systems.
2740: .TP
1.1 root 2741: .B \-m68020
2742: .TP
2743: .B \-mc68020
2744: Generate output for a 68020 (rather than a 68000). This is the
1.1.1.5 root 2745: default when the compiler is configured for 68020-based systems.
1.1 root 2746: .TP
2747: .B \-m68881
2748: Generate output containing 68881 instructions for floating point.
1.1.1.6 root 2749: This is the default for most 68020-based systems unless
1.1.1.5 root 2750: .B \-nfp
2751: was specified when the compiler was configured.
2752: .TP
2753: .B \-m68030
2754: Generate output for a 68030. This is the default when the compiler is
2755: configured for 68030-based systems.
2756: .TP
2757: .B \-m68040
2758: Generate output for a 68040. This is the default when the compiler is
2759: configured for 68040-based systems.
2760: .TP
2761: .B \-m68020\-40
2762: Generate output for a 68040, without using any of the new instructions.
2763: This results in code which can run relatively efficiently on either a
2764: 68020/68881 or a 68030 or a 68040.
1.1 root 2765: .TP
2766: .B \-mfpa
2767: Generate output containing Sun FPA instructions for floating point.
2768: .TP
2769: .B \-msoft\-float
2770: Generate output containing library calls for floating point.
2771: .I
1.1.1.6 root 2772: WARNING:
1.1 root 2773: the requisite libraries are not part of GNU CC. Normally the
2774: facilities of the machine's usual C compiler are used, but this can't
2775: be done directly in cross-compilation. You must make your own
2776: arrangements to provide suitable library functions for cross-compilation.
2777: .TP
2778: .B \-mshort
2779: Consider type \c
2780: .B int\c
2781: \& to be 16 bits wide, like \c
2782: .B short int\c
2783: \&.
2784: .TP
2785: .B \-mnobitfield
2786: Do not use the bit-field instructions. `\|\c
2787: .B \-m68000\c
2788: \&\|' implies
2789: `\|\c
2790: .B \-mnobitfield\c
2791: \&\|'.
2792: .TP
2793: .B \-mbitfield
2794: Do use the bit-field instructions. `\|\c
2795: .B \-m68020\c
2796: \&\|' implies
2797: `\|\c
2798: .B \-mbitfield\c
2799: \&\|'. This is the default if you use the unmodified
2800: sources.
2801: .TP
2802: .B \-mrtd
2803: Use a different function-calling convention, in which functions
2804: that take a fixed number of arguments return with the \c
1.1.1.6 root 2805: .B rtd
1.1 root 2806: instruction, which pops their arguments while returning. This
2807: saves one instruction in the caller since there is no need to pop
2808: the arguments there.
1.1.1.6 root 2809: .Sp
1.1 root 2810: This calling convention is incompatible with the one normally
2811: used on Unix, so you cannot use it if you need to call libraries
2812: compiled with the Unix compiler.
1.1.1.6 root 2813: .Sp
1.1 root 2814: Also, you must provide function prototypes for all functions that
2815: take variable numbers of arguments (including \c
2816: .B printf\c
2817: \&);
2818: otherwise incorrect code will be generated for calls to those
2819: functions.
1.1.1.6 root 2820: .Sp
1.1 root 2821: In addition, seriously incorrect code will result if you call a
2822: function with too many arguments. (Normally, extra arguments are
2823: harmlessly ignored.)
1.1.1.6 root 2824: .Sp
1.1 root 2825: The \c
2826: .B rtd\c
2827: \& instruction is supported by the 68010 and 68020
2828: processors, but not by the 68000.
2829: .PP
2830: These `\|\c
2831: .B \-m\c
2832: \&\|' options are defined for the Vax:
2833: .TP
2834: .B \-munix
2835: Do not output certain jump instructions (\c
2836: .B aobleq\c
2837: \& and so on)
2838: that the Unix assembler for the Vax cannot handle across long
2839: ranges.
2840: .TP
2841: .B \-mgnu
2842: Do output those jump instructions, on the assumption that you
2843: will assemble with the GNU assembler.
2844: .TP
2845: .B \-mg
2846: Output code for g-format floating point numbers instead of d-format.
2847: .PP
2848: These `\|\c
2849: .B \-m\c
1.1.1.5 root 2850: \&\|' switches are supported on the SPARC:
2851: .PP
1.1 root 2852: .B \-mfpu
1.1.1.5 root 2853: .TP
2854: .B \-mhard\-float
1.1 root 2855: Generate output containing floating point instructions. This is the
1.1.1.5 root 2856: default.
2857: .PP
2858: .B \-mno\-fpu
1.1 root 2859: .TP
1.1.1.5 root 2860: .B \-msoft\-float
2861: Generate output containing library calls for floating point.
2862: .I Warning:
2863: there is no GNU floating-point library for SPARC.
2864: Normally the facilities of the machine's usual C compiler are used, but
2865: this cannot be done directly in cross-compilation. You must make your
2866: own arrangements to provide suitable library functions for
2867: cross-compilation.
1.1.1.6 root 2868: .Sp
1.1.1.5 root 2869: .B \-msoft\-float
2870: changes the calling convention in the output file;
2871: therefore, it is only useful if you compile
2872: .I all
2873: of a program with this option.
2874: .PP
1.1 root 2875: .B \-mno\-epilogue
1.1.1.5 root 2876: .TP
2877: .B \-mepilogue
2878: With
2879: .B \-mepilogue
2880: (the default), the compiler always emits code for
2881: function exit at the end of each function. Any function exit in
2882: the middle of the function (such as a return statement in C) will
2883: generate a jump to the exit code at the end of the function.
1.1.1.6 root 2884: .Sp
1.1.1.5 root 2885: With
2886: .BR \-mno\-epilogue ,
2887: the compiler tries to emit exit code inline at every function exit.
2888: .PP
1.1.1.7 root 2889: .B \-mno\-v8
2890: .TP
1.1.1.5 root 2891: .B \-mv8
2892: .TP
2893: .B \-msparclite
1.1.1.7 root 2894: These three options select variations on the SPARC architecture.
1.1.1.6 root 2895: .Sp
1.1.1.5 root 2896: By default (unless specifically configured for the Fujitsu SPARClite),
2897: GCC generates code for the v7 variant of the SPARC architecture.
1.1.1.6 root 2898: .Sp
1.1.1.5 root 2899: .B \-mv8
2900: will give you SPARC v8 code. The only difference from v7
2901: code is that the compiler emits the integer multiply and integer
2902: divide instructions which exist in SPARC v8 but not in SPARC v7.
1.1.1.6 root 2903: .Sp
1.1.1.5 root 2904: .B \-msparclite
2905: will give you SPARClite code. This adds the integer
2906: multiply, integer divide step and scan (ffs) instructions which
2907: exist in SPARClite but not in SPARC v7.
1.1 root 2908: .PP
1.1.1.7 root 2909: .B \-mcypress
2910: .TP
2911: .B \-msupersparc
2912: These two options select the processor for which the code is optimised.
2913: .Sp
2914: With
2915: .B \-mcypress
2916: (the default), the compiler optimises code for the Cypress CY7C602 chip, as
1.1.1.8 ! root 2917: used in the SparcStation/SparcServer 3xx series. This is also appropriate for
1.1.1.7 root 2918: the older SparcStation 1, 2, IPX etc.
2919: .Sp
2920: With
2921: .B \-msupersparc
2922: the compiler optimises code for the SuperSparc cpu, as used in the SparcStation
2923: 10, 1000 and 2000 series. This flag also enables use of the full SPARC v8
2924: instruction set.
2925: .PP
1.1 root 2926: These `\|\c
2927: .B \-m\c
2928: \&\|' options are defined for the Convex:
2929: .TP
2930: .B \-mc1
2931: Generate output for a C1. This is the default when the compiler is
2932: configured for a C1.
2933: .TP
2934: .B \-mc2
2935: Generate output for a C2. This is the default when the compiler is
2936: configured for a C2.
2937: .TP
2938: .B \-margcount
2939: Generate code which puts an argument count in the word preceding each
2940: argument list. Some nonportable Convex and Vax programs need this word.
2941: (Debuggers don't, except for functions with variable-length argument
2942: lists; this info is in the symbol table.)
2943: .TP
2944: .B \-mnoargcount
2945: Omit the argument count word. This is the default if you use the
2946: unmodified sources.
2947: .PP
2948: These `\|\c
2949: .B \-m\c
2950: \&\|' options are defined for the AMD Am29000:
2951: .TP
2952: .B \-mdw
2953: Generate code that assumes the DW bit is set, i.e., that byte and
2954: halfword operations are directly supported by the hardware. This is the
2955: default.
2956: .TP
2957: .B \-mnodw
2958: Generate code that assumes the DW bit is not set.
2959: .TP
2960: .B \-mbw
2961: Generate code that assumes the system supports byte and halfword write
2962: operations. This is the default.
2963: .TP
2964: .B \-mnbw
2965: Generate code that assumes the systems does not support byte and
2966: halfword write operations. This implies `\|\c
2967: .B \-mnodw\c
2968: \&\|'.
2969: .TP
2970: .B \-msmall
2971: Use a small memory model that assumes that all function addresses are
2972: either within a single 256 KB segment or at an absolute address of less
2973: than 256K. This allows the \c
2974: .B call\c
2975: \& instruction to be used instead
2976: of a \c
2977: .B const\c
2978: \&, \c
2979: .B consth\c
2980: \&, \c
2981: .B calli\c
2982: \& sequence.
2983: .TP
2984: .B \-mlarge
2985: Do not assume that the \c
2986: .B call\c
2987: \& instruction can be used; this is the
2988: default.
2989: .TP
2990: .B \-m29050
2991: Generate code for the Am29050.
2992: .TP
2993: .B \-m29000
2994: Generate code for the Am29000. This is the default.
2995: .TP
2996: .B \-mkernel\-registers
2997: Generate references to registers \c
2998: .B gr64-gr95\c
2999: \& instead of
3000: .B gr96-gr127\c
3001: \&. This option can be used when compiling kernel code
3002: that wants a set of global registers disjoint from that used by
3003: user-mode code.
1.1.1.6 root 3004: .Sp
1.1 root 3005: Note that when this option is used, register names in `\|\c
3006: .B \-f\c
3007: \&\|' flags
3008: must use the normal, user-mode, names.
3009: .TP
3010: .B \-muser\-registers
3011: Use the normal set of global registers, \c
3012: .B gr96-gr127\c
3013: \&. This is the
3014: default.
3015: .TP
3016: .B \-mstack\-check
3017: Insert a call to \c
1.1.1.6 root 3018: .B _\|_msp_check\c
1.1 root 3019: \& after each stack adjustment. This
3020: is often used for kernel code.
3021: .PP
3022: These `\|\c
3023: .B \-m\c
3024: \&\|' options are defined for Motorola 88K architectures:
3025: .TP
1.1.1.5 root 3026: .B \-m88000
3027: Generate code that works well on both the m88100 and the
3028: m88110.
3029: .TP
3030: .B \-m88100
3031: Generate code that works best for the m88100, but that also
3032: runs on the m88110.
3033: .TP
3034: .B \-m88110
3035: Generate code that works best for the m88110, and may not run
3036: on the m88100.
1.1 root 3037: .TP
3038: .B \-midentify\-revision
3039: Include an \c
3040: .B ident\c
3041: \& directive in the assembler output recording the
3042: source file name, compiler name and version, timestamp, and compilation
3043: flags used.
3044: .TP
3045: .B \-mno\-underscores
3046: In assembler output, emit symbol names without adding an underscore
3047: character at the beginning of each name. The default is to use an
3048: underscore as prefix on each name.
3049: .TP
3050: .B \-mno\-check\-zero\-division
3051: .TP
3052: .B \-mcheck\-zero\-division
3053: Early models of the 88K architecture had problems with division by zero;
3054: in particular, many of them didn't trap. Use these options to avoid
3055: including (or to include explicitly) additional code to detect division
3056: by zero and signal an exception. All GCC configurations for the 88K use
3057: `\|\c
3058: .B \-mcheck\-zero\-division\c
3059: \&\|' by default.
3060: .TP
3061: .B \-mocs\-debug\-info
3062: .TP
3063: .B \-mno\-ocs\-debug\-info
3064: Include (or omit) additional debugging information (about
3065: registers used in each stack frame) as specified in the 88Open Object
1.1.1.6 root 3066: Compatibility Standard, \*(lqOCS\*(rq. This extra information is not needed
1.1 root 3067: by GDB. The default for DG/UX, SVr4, and Delta 88 SVr3.2 is to
3068: include this information; other 88k configurations omit this information
3069: by default.
3070: .TP
3071: .B \-mocs\-frame\-position
3072: .TP
3073: .B \-mno\-ocs\-frame\-position
3074: Force (or do not require) register values to be stored in a particular
3075: place in stack frames, as specified in OCS. The DG/UX, Delta88 SVr3.2,
3076: and BCS configurations use `\|\c
3077: .B \-mocs\-frame\-position\c
3078: \&\|'; other 88k
3079: configurations have the default `\|\c
3080: .B \-mno\-ocs\-frame\-position\c
3081: \&\|'.
3082: .TP
3083: .B \-moptimize\-arg\-area
3084: .TP
3085: .B \-mno\-optimize\-arg\-area
3086: Control how to store function arguments in stack frames.
3087: `\|\c
3088: .B \-moptimize\-arg\-area\c
3089: \&\|' saves space, but may break some
3090: debuggers (not GDB). `\|\c
3091: .B \-mno\-optimize\-arg\-area\c
3092: \&\|' conforms better to
3093: standards. By default GCC does not optimize the argument area.
3094: .TP
1.1.1.6 root 3095: .BI "\-mshort\-data\-" "num"
3096: .I num
1.1 root 3097: Generate smaller data references by making them relative to \c
3098: .B r0\c
3099: \&,
3100: which allows loading a value using a single instruction (rather than the
3101: usual two). You control which data references are affected by
3102: specifying \c
3103: .I num\c
3104: \& with this option. For example, if you specify
3105: `\|\c
3106: .B \-mshort\-data\-512\c
3107: \&\|', then the data references affected are those
3108: involving displacements of less than 512 bytes.
3109: `\|\c
3110: .B \-mshort\-data\-\c
3111: .I num\c
3112: \&\c
3113: \&\|' is not effective for \c
3114: .I num\c
3115: \& greater
3116: than 64K.
1.1.1.5 root 3117: .PP
3118: .B \-mserialize-volatile
1.1 root 3119: .TP
1.1.1.5 root 3120: .B \-mno-serialize-volatile
3121: Do, or do not, generate code to guarantee sequential consistency of
3122: volatile memory references.
1.1.1.6 root 3123: .Sp
1.1.1.5 root 3124: GNU CC always guarantees consistency by default, for the preferred
3125: processor submodel. How this is done depends on the submodel.
1.1.1.6 root 3126: .Sp
1.1.1.5 root 3127: The m88100 processor does not reorder memory references and so always
3128: provides sequential consistency. If you use `\|\c
3129: .B \-m88100\c
3130: \&\|', GNU CC does
3131: not generate any special instructions for sequential consistency.
1.1.1.6 root 3132: .Sp
1.1.1.5 root 3133: The order of memory references made by the m88110 processor does not
3134: always match the order of the instructions requesting those references.
3135: In particular, a load instruction may execute before a preceding store
3136: instruction. Such reordering violates sequential consistency of
3137: volatile memory references, when there are multiple processors. When
3138: you use `\|\c
3139: .B \-m88000\c
3140: \&\|' or `\|\c
3141: .B \-m88110\c
3142: \&\|', GNU CC generates special
3143: instructions when appropriate, to force execution in the proper order.
1.1.1.6 root 3144: .Sp
1.1.1.5 root 3145: The extra code generated to guarantee consistency may affect the
3146: performance of your application. If you know that you can safely forgo
3147: this guarantee, you may use the option `\|\c
3148: .B \-mno-serialize-volatile\c
3149: \&\|'.
1.1.1.6 root 3150: .Sp
1.1.1.5 root 3151: If you use the `\|\c
3152: .B \-m88100\c
3153: \&\|' option but require sequential consistency
3154: when running on the m88110 processor, you should use
3155: `\|\c
3156: .B \-mserialize-volatile\c
3157: \&\|'.
3158: .PP
1.1 root 3159: .B \-msvr4
3160: .TP
3161: .B \-msvr3
3162: Turn on (`\|\c
3163: .B \-msvr4\c
3164: \&\|') or off (`\|\c
3165: .B \-msvr3\c
3166: \&\|') compiler extensions
3167: related to System V release 4 (SVr4). This controls the following:
3168: .TP
1.1.1.6 root 3169: \ \ \ \(bu
1.1 root 3170: Which variant of the assembler syntax to emit (which you can select
3171: independently using `\|\c
1.1.1.5 root 3172: .B \-mversion\-03.00\c
1.1.1.6 root 3173: \&\|').
1.1 root 3174: .TP
3175: \ \ \ \(bu
3176: `\|\c
3177: .B \-msvr4\c
3178: \&\|' makes the C preprocessor recognize `\|\c
3179: .B #pragma weak\c
3180: \&\|'
3181: .TP
3182: \ \ \ \(bu
3183: `\|\c
3184: .B \-msvr4\c
3185: \&\|' makes GCC issue additional declaration directives used in
1.1.1.6 root 3186: SVr4.
1.1 root 3187: .PP
3188: `\|\c
3189: .B \-msvr3\c
3190: \&\|' is the default for all m88K configurations except
3191: the SVr4 configuration.
3192: .TP
3193: .B \-mtrap\-large\-shift
3194: .TP
3195: .B \-mhandle\-large\-shift
3196: Include code to detect bit-shifts of more than 31 bits; respectively,
3197: trap such shifts or emit code to handle them properly. By default GCC
3198: makes no special provision for large bit shifts.
3199: .TP
3200: .B \-muse\-div\-instruction
3201: Very early models of the 88K architecture didn't have a divide
3202: instruction, so GCC avoids that instruction by default. Use this option
1.1.1.6 root 3203: to specify that it's safe to use the divide instruction.
1.1 root 3204: .TP
3205: .B \-mversion\-03.00
1.1.1.5 root 3206: In the DG/UX configuration, there are two flavors of SVr4. This option
3207: modifies
3208: .B \-msvr4
3209: to select whether the hybrid-COFF or real-ELF
3210: flavor is used. All other configurations ignore this option.
1.1 root 3211: .TP
3212: .B \-mwarn\-passed\-structs
3213: Warn when a function passes a struct as an argument or result.
3214: Structure-passing conventions have changed during the evolution of the C
3215: language, and are often the source of portability problems. By default,
3216: GCC issues no such warning.
3217: .PP
3218: These options are defined for the IBM RS6000:
3219: .PP
3220: .B \-mfp\-in\-toc
3221: .TP
3222: .B \-mno\-fp\-in\-toc
3223: Control whether or not floating-point constants go in the Table of
3224: Contents (TOC), a table of all global variable and function addresses. By
3225: default GCC puts floating-point constants there; if the TOC overflows,
3226: `\|\c
3227: .B \-mno\-fp\-in\-toc\c
3228: \&\|' will reduce the size of the TOC, which may avoid
3229: the overflow.
3230: .PP
3231: These `\|\c
3232: .B \-m\c
3233: \&\|' options are defined for the IBM RT PC:
3234: .TP
3235: .B \-min\-line\-mul
3236: Use an in-line code sequence for integer multiplies. This is the
3237: default.
3238: .TP
3239: .B \-mcall\-lib\-mul
3240: Call \c
3241: .B lmul$$\c
3242: \& for integer multiples.
3243: .TP
3244: .B \-mfull\-fp\-blocks
3245: Generate full-size floating point data blocks, including the minimum
3246: amount of scratch space recommended by IBM. This is the default.
3247: .TP
3248: .B \-mminimum\-fp\-blocks
3249: Do not include extra scratch space in floating point data blocks. This
3250: results in smaller code, but slower execution, since scratch space must
3251: be allocated dynamically.
3252: .TP
3253: .B \-mfp\-arg\-in\-fpregs
3254: Use a calling sequence incompatible with the IBM calling convention in
3255: which floating point arguments are passed in floating point registers.
3256: Note that \c
3257: .B varargs.h\c
3258: \& and \c
3259: .B stdargs.h\c
3260: \& will not work with
3261: floating point operands if this option is specified.
3262: .TP
3263: .B \-mfp\-arg\-in\-gregs
3264: Use the normal calling convention for floating point arguments. This is
3265: the default.
3266: .TP
3267: .B \-mhc\-struct\-return
3268: Return structures of more than one word in memory, rather than in a
3269: register. This provides compatibility with the MetaWare HighC (hc)
3270: compiler. Use `\|\c
3271: .B \-fpcc\-struct\-return\c
3272: \&\|' for compatibility with the
3273: Portable C Compiler (pcc).
3274: .TP
3275: .B \-mnohc\-struct\-return
3276: Return some structures of more than one word in registers, when
3277: convenient. This is the default. For compatibility with the
3278: IBM-supplied compilers, use either `\|\c
3279: .B \-fpcc\-struct\-return\c
3280: \&\|' or
3281: `\|\c
3282: .B \-mhc\-struct\-return\c
3283: \&\|'.
3284: .PP
3285: These `\|\c
3286: .B \-m\c
3287: \&\|' options are defined for the MIPS family of computers:
3288: .TP
3289: .BI "\-mcpu=" "cpu-type"
1.1.1.6 root 3290: Assume the defaults for the machine type
1.1 root 3291: .I cpu-type
3292: when
1.1.1.3 root 3293: scheduling instructions. The default
1.1 root 3294: .I cpu-type
3295: is
3296: .BR default ,
3297: which picks the longest cycles times for any of the machines, in order
3298: that the code run at reasonable rates on all MIPS cpu's. Other
3299: choices for
3300: .I cpu-type
3301: are
3302: .BR r2000 ,
3303: .BR r3000 ,
3304: .BR r4000 ,
3305: and
3306: .BR r6000 .
3307: While picking a specific
3308: .I cpu-type
3309: will schedule things appropriately for that particular chip, the
3310: compiler will not generate any code that does not meet level 1 of the
3311: MIPS ISA (instruction set architecture) without the
3312: .B \-mips2
3313: or
3314: .B \-mips3
3315: switches being used.
3316: .TP
3317: .B \-mips2
3318: Issue instructions from level 2 of the MIPS ISA (branch likely, square
1.1.1.6 root 3319: root instructions). The
1.1 root 3320: .B \-mcpu=r4000
3321: or
3322: .B \-mcpu=r6000
1.1.1.3 root 3323: switch must be used in conjunction with
1.1 root 3324: .BR \-mips2 .
3325: .TP
3326: .B \-mips3
3327: Issue instructions from level 3 of the MIPS ISA (64 bit instructions).
3328: The
3329: .B \-mcpu=r4000
1.1.1.3 root 3330: switch must be used in conjunction with
1.1 root 3331: .BR \-mips2 .
3332: .TP
3333: .B \-mint64
3334: .TP
3335: .B \-mlong64
3336: .TP
3337: .B \-mlonglong128
3338: These options don't work at present.
3339: .TP
3340: .B \-mmips\-as
3341: Generate code for the MIPS assembler, and invoke
3342: .B mips\-tfile
3343: to add normal debug information. This is the default for all
3344: platforms except for the OSF/1 reference platform, using the OSF/rose
1.1.1.2 root 3345: object format. If any of the
1.1 root 3346: .BR \-ggdb ,
3347: .BR \-gstabs ,
3348: or
3349: .B \-gstabs+
3350: switches are used, the
3351: .B mips\-tfile
3352: program will encapsulate the stabs within MIPS ECOFF.
3353: .TP
3354: .B \-mgas
3355: Generate code for the GNU assembler. This is the default on the OSF/1
3356: reference platform, using the OSF/rose object format.
3357: .TP
3358: .B \-mrnames
3359: .TP
3360: .B \-mno\-rnames
3361: The
3362: .B \-mrnames
3363: switch says to output code using the MIPS software names for the
1.1.1.6 root 3364: registers, instead of the hardware names (ie,
1.1 root 3365: .B a0
3366: instead of
3367: .BR $4 ).
3368: The GNU assembler does not support the
3369: .B \-mrnames
3370: switch, and the MIPS assembler will be instructed to run the MIPS C
3371: preprocessor over the source file. The
3372: .B \-mno\-rnames
3373: switch is default.
3374: .TP
3375: .B \-mgpopt
3376: .TP
3377: .B \-mno\-gpopt
3378: The
3379: .B \-mgpopt
3380: switch says to write all of the data declarations before the
3381: instructions in the text section, to all the MIPS assembler to
3382: generate one word memory references instead of using two words for
3383: short global or static data items. This is on by default if
3384: optimization is selected.
3385: .TP
3386: .B \-mstats
3387: .TP
3388: .B \-mno\-stats
3389: For each non-inline function processed, the
3390: .B \-mstats
3391: switch causes the compiler to emit one line to the standard error file
3392: to print statistics about the program (number of registers saved,
3393: stack size, etc.).
3394: .TP
3395: .B \-mmemcpy
3396: .TP
3397: .B \-mno\-memcpy
3398: The
3399: .B \-mmemcpy
3400: switch makes all block moves call the appropriate string function
3401: .RB ( memcpy
3402: or
3403: .BR bcopy )
3404: instead of possibly generating inline code.
3405: .TP
3406: .B \-mmips\-tfile
3407: .TP
3408: .B \-mno\-mips\-tfile
3409: The
3410: .B \-mno\-mips\-tfile
3411: switch causes the compiler not postprocess the object file with the
3412: .B mips\-tfile
3413: program, after the MIPS assembler has generated it to add debug
3414: support. If
3415: .B mips\-tfile
3416: is not run, then no local variables will be available to the debugger.
3417: In addition,
3418: .B stage2
3419: and
3420: .B stage3
3421: objects will have the temporary file names passed to the assembler
3422: embedded in the object file, which means the objects will not compare
3423: the same.
3424: .TP
3425: .B \-msoft\-float
3426: Generate output containing library calls for floating point.
3427: .I
1.1.1.6 root 3428: WARNING:
1.1 root 3429: the requisite libraries are not part of GNU CC. Normally the
3430: facilities of the machine's usual C compiler are used, but this can't
3431: be done directly in cross-compilation. You must make your own
3432: arrangements to provide suitable library functions for cross-compilation.
3433: .TP
3434: .B \-mhard\-float
3435: Generate output containing floating point instructions. This is the
3436: default if you use the unmodified sources.
3437: .TP
3438: .B \-mfp64
3439: Assume that the
3440: .B FR
3441: bit in the status word is on, and that there are 32 64-bit floating
3442: point registers, instead of 32 32-bit floating point registers. You
3443: must also specify the
3444: .B \-mcpu=r4000
3445: and
3446: .B \-mips3
3447: switches.
3448: .TP
3449: .B \-mfp32
3450: Assume that there are 32 32-bit floating point registers. This is the
3451: default.
1.1.1.5 root 3452: .PP
1.1 root 3453: .B \-mabicalls
1.1.1.5 root 3454: .TP
3455: .B \-mno\-abicalls
3456: Emit (or do not emit) the
1.1 root 3457: .BR \&.abicalls ,
3458: .BR \&.cpload ,
3459: and
3460: .B \&.cprestore
3461: pseudo operations that some System V.4 ports use for position
3462: independent code.
3463: .TP
3464: .B \-mhalf\-pic
3465: .TP
3466: .B \-mno\-half\-pic
3467: The
3468: .B \-mhalf\-pic
3469: switch says to put pointers to extern references into the data section
3470: and load them up, rather than put the references in the text section.
3471: This option does not work at present.
3472: .B
3473: .BI \-G num
1.1.1.6 root 3474: Put global and static items less than or equal to
1.1 root 3475: .I num
3476: bytes into the small data or bss sections instead of the normal data
3477: or bss section. This allows the assembler to emit one word memory
3478: reference instructions based on the global pointer
3479: .RB ( gp
3480: or
3481: .BR $28 ),
3482: instead of the normal two words used. By default,
3483: .I num
3484: is 8 when the MIPS assembler is used, and 0 when the GNU
3485: assembler is used. The
3486: .BI \-G num
3487: switch is also passed to the assembler and linker. All modules should
3488: be compiled with the same
3489: .BI \-G num
3490: value.
1.1.1.5 root 3491: .TP
3492: .B \-nocpp
3493: Tell the MIPS assembler to not run it's preprocessor over user
3494: assembler files (with a `\|\c
3495: .B .s\c
3496: \&\|' suffix) when assembling them.
3497: .PP
3498: These `\|\c
3499: .B \-m\c
3500: \&\|' options are defined for the Intel 80386 family of computers:
3501: .B \-m486
3502: .TP
3503: .B \-mno\-486
3504: Control whether or not code is optimized for a 486 instead of an
3505: 386. Code generated for a 486 will run on a 386 and vice versa.
3506: .TP
3507: .B \-msoft\-float
3508: Generate output containing library calls for floating point.
3509: .I Warning:
3510: the requisite libraries are not part of GNU CC.
3511: Normally the facilities of the machine's usual C compiler are used, but
3512: this can't be done directly in cross-compilation. You must make your
3513: own arrangements to provide suitable library functions for
3514: cross-compilation.
1.1.1.6 root 3515: .Sp
1.1.1.5 root 3516: On machines where a function returns floating point results in the 80387
3517: register stack, some floating point opcodes may be emitted even if
3518: `\|\c
3519: .B \-msoft-float\c
3520: \&\|' is used.
3521: .TP
3522: .B \-mno-fp-ret-in-387
3523: Do not use the FPU registers for return values of functions.
1.1.1.6 root 3524: .Sp
1.1.1.5 root 3525: The usual calling convention has functions return values of types
3526: .B float\c
3527: \& and \c
3528: .B double\c
3529: \& in an FPU register, even if there
3530: is no FPU. The idea is that the operating system should emulate
3531: an FPU.
1.1.1.6 root 3532: .Sp
1.1.1.5 root 3533: The option `\|\c
3534: .B \-mno-fp-ret-in-387\c
3535: \&\|' causes such values to be returned
3536: in ordinary CPU registers instead.
3537: .PP
3538: These `\|\c
3539: .B \-m\c
3540: \&\|' options are defined for the HPPA family of computers:
3541: .TP
3542: .B \-mpa-risc-1-0
3543: Generate code for a PA 1.0 processor.
3544: .TP
3545: .B \-mpa-risc-1-1
3546: Generate code for a PA 1.1 processor.
3547: .TP
3548: .B \-mkernel
3549: Generate code which is suitable for use in kernels. Specifically, avoid
3550: .B add\c
3551: \& instructions in which one of the arguments is the DP register;
3552: generate \c
3553: .B addil\c
3554: \& instructions instead. This avoids a rather serious
3555: bug in the HP-UX linker.
3556: .TP
3557: .B \-mshared-libs
3558: Generate code that can be linked against HP-UX shared libraries. This option
3559: is not fully function yet, and is not on by default for any PA target. Using
3560: this option can cause incorrect code to be generated by the compiler.
3561: .TP
3562: .B \-mno-shared-libs
3563: Don't generate code that will be linked against shared libraries. This is
3564: the default for all PA targets.
3565: .TP
3566: .B \-mlong-calls
3567: Generate code which allows calls to functions greater than 256K away from
3568: the caller when the caller and callee are in the same source file. Do
1.1.1.6 root 3569: not turn this option on unless code refuses to link with \*(lqbranch out of
3570: range errors\*('' from the linker.
1.1.1.5 root 3571: .TP
3572: .B \-mdisable-fpregs
3573: Prevent floating point registers from being used in any manner. This is
3574: necessary for compiling kernels which perform lazy context switching of
3575: floating point registers. If you use this option and attempt to perform
3576: floating point operations, the compiler will abort.
3577: .TP
3578: .B \-mdisable-indexing
3579: Prevent the compiler from using indexing address modes. This avoids some
3580: rather obscure problems when compiling MIG generated code under MACH.
3581: .TP
3582: .B \-mtrailing-colon
3583: Add a colon to the end of label definitions (for ELF assemblers).
3584: .PP
3585: These `\|\c
3586: .B \-m\c
3587: \&\|' options are defined for the Intel 80960 family of computers:
3588: .TP
3589: .BI "\-m" "cpu-type"
3590: Assume the defaults for the machine type
3591: .I cpu-type
3592: for instruction and addressing-mode availability and alignment.
3593: The default
3594: .I cpu-type
3595: is
3596: .BR kb ;
3597: other choices are
3598: .BR ka ,
3599: .BR mc ,
3600: .BR ca ,
3601: .BR cf ,
3602: .BR sa ,
3603: and
3604: .BR sb .
3605: .TP
3606: .B \-mnumerics
3607: .TP
3608: .B \-msoft\-float
1.1.1.6 root 3609: The
1.1.1.5 root 3610: .B \-mnumerics
3611: option indicates that the processor does support
1.1.1.6 root 3612: floating-point instructions. The
1.1.1.5 root 3613: .B \-msoft\-float
3614: option indicates
3615: that floating-point support should not be assumed.
3616: .TP
3617: .B \-mleaf\-procedures
3618: .TP
3619: .B \-mno\-leaf\-procedures
3620: Do (or do not) attempt to alter leaf procedures to be callable with the
3621: .I bal
3622: instruction as well as
3623: .IR call .
3624: This will result in more
3625: efficient code for explicit calls when the
3626: .I bal
3627: instruction can be
3628: substituted by the assembler or linker, but less efficient code in other
3629: cases, such as calls via function pointers, or using a linker that doesn't
3630: support this optimization.
3631: .TP
3632: .B \-mtail\-call
3633: .TP
3634: .B \-mno\-tail\-call
3635: Do (or do not) make additional attempts (beyond those of the
3636: machine-independent portions of the compiler) to optimize tail-recursive
3637: calls into branches. You may not want to do this because the detection of
3638: cases where this is not valid is not totally complete. The default is
3639: .BR \-mno\-tail\-call .
3640: .TP
3641: .B \-mcomplex\-addr
3642: .TP
3643: .B \-mno\-complex\-addr
3644: Assume (or do not assume) that the use of a complex addressing mode is a
3645: win on this implementation of the i960. Complex addressing modes may not
3646: be worthwhile on the K-series, but they definitely are on the C-series.
1.1.1.6 root 3647: The default is currently
1.1.1.5 root 3648: .B \-mcomplex\-addr
3649: for all processors except
3650: the CB and CC.
3651: .TP
3652: .B \-mcode\-align
3653: .TP
3654: .B \-mno\-code\-align
3655: Align code to 8-byte boundaries for faster fetching (or don't bother).
3656: Currently turned on by default for C-series implementations only.
3657: .TP
3658: .B \-mic\-compat
3659: .TP
3660: .B \-mic2.0\-compat
3661: .TP
3662: .B \-mic3.0\-compat
3663: Enable compatibility with iC960 v2.0 or v3.0.
3664: .TP
3665: .B \-masm\-compat
3666: .TP
3667: .B \-mintel\-asm
3668: Enable compatibility with the iC960 assembler.
3669: .TP
3670: .B \-mstrict\-align
3671: .TP
3672: .B \-mno\-strict\-align
3673: Do not permit (do permit) unaligned accesses.
3674: .TP
3675: .B \-mold\-align
3676: Enable structure-alignment compatibility with Intel's gcc release version
3677: 1.3 (based on gcc 1.37). Currently this is buggy in that
3678: .B #pragma align 1
3679: is always assumed as well, and cannot be turned off.
3680: .PP
3681: These `\|\c
3682: .B \-m\c
3683: \&\|' options are defined for the DEC Alpha implementations:
3684: .TP
3685: .B \-mno-soft-float
3686: .TP
3687: .B \-msoft-float
3688: Use (do not use) the hardware floating-point instructions for
3689: floating-point operations. When \c
3690: .B \-msoft-float\c
3691: \& is specified,
3692: functions in `\|\c
3693: .B libgcc1.c\c
3694: \&\|' will be used to perform floating-point
3695: operations. Unless they are replaced by routines that emulate the
3696: floating-point operations, or compiled in such a way as to call such
3697: emulations routines, these routines will issue floating-point
3698: operations. If you are compiling for an Alpha without floating-point
3699: operations, you must ensure that the library is built so as not to call
3700: them.
1.1.1.6 root 3701: .Sp
1.1.1.5 root 3702: Note that Alpha implementations without floating-point operations are
3703: required to have floating-point registers.
3704: .TP
3705: .B \-mfp-reg
3706: .TP
3707: .B \-mno-fp-regs
3708: Generate code that uses (does not use) the floating-point register set.
3709: .B \-mno-fp-regs\c
3710: \& implies \c
3711: .B \-msoft-float\c
3712: \&. If the floating-point
3713: register set is not used, floating point operands are passed in integer
3714: registers as if they were integers and floating-point results are passed
3715: in $0 instead of $f0. This is a non-standard calling sequence, so any
3716: function with a floating-point argument or return value called by code
3717: compiled with \c
3718: .B \-mno-fp-regs\c
3719: \& must also be compiled with that
3720: option.
1.1.1.6 root 3721: .Sp
1.1.1.5 root 3722: A typical use of this option is building a kernel that does not use,
3723: and hence need not save and restore, any floating-point registers.
3724: .PP
3725: These additional options are available on System V Release 4 for
3726: compatibility with other compilers on those systems:
3727: .TP
3728: .B \-G
3729: On SVr4 systems, \c
3730: .B gcc\c
3731: \& accepts the option `\|\c
3732: .B \-G\c
3733: \&\|' (and passes
3734: it to the system linker), for compatibility with other compilers.
3735: However, we suggest you use `\|\c
3736: .B \-symbolic\c
3737: \&\|' or `\|\c
3738: .B \-shared\c
3739: \&\|' as
3740: appropriate, instead of supplying linker options on the \c
1.1.1.6 root 3741: .B gcc
1.1.1.5 root 3742: command line.
3743: .TP
3744: .B \-Qy
3745: Identify the versions of each tool used by the compiler, in a
3746: .B .ident\c
3747: \& assembler directive in the output.
3748: .TP
3749: .B \-Qn
3750: Refrain from adding \c
3751: .B .ident\c
3752: \& directives to the output file (this is
3753: the default).
3754: .TP
1.1.1.6 root 3755: .BI "\-YP," "dirs"
1.1.1.5 root 3756: Search the directories \c
3757: .I dirs\c
3758: \&, and no others, for libraries
3759: specified with `\|\c
3760: .B \-l\c
3761: \&\|'. You can separate directory entries in
3762: .I dirs\c
3763: \& from one another with colons.
3764: .TP
1.1.1.6 root 3765: .BI "\-Ym," "dir"
1.1.1.5 root 3766: Look in the directory \c
3767: .I dir\c
3768: \& to find the M4 preprocessor.
3769: The assembler uses this option.
1.1 root 3770: .SH CODE GENERATION OPTIONS
3771: These machine-independent options control the interface conventions
3772: used in code generation.
1.1.1.6 root 3773: .PP
1.1 root 3774: Most of them begin with `\|\c
3775: \-f\c
3776: \&\|'. These options have both positive and negative forms; the negative form
3777: of `\|\c
3778: .B \-ffoo\c
3779: \&\|' would be `\|\c
3780: .B \-fno\-foo\c
3781: \&\|'. In the table below, only
3782: one of the forms is listed\(em\&the one which is not the default. You
3783: can figure out the other form by either removing `\|\c
3784: .B no\-\c
3785: \&\|' or adding
3786: it.
3787: .TP
3788: .B \-fnonnull\-objects
1.1.1.5 root 3789: Assume that objects reached through references are not null
3790: (C++ only).
1.1.1.6 root 3791: .Sp
1.1 root 3792: Normally, GNU C++ makes conservative assumptions about objects reached
1.1.1.5 root 3793: through references. For example, the compiler must check that \c
1.1.1.6 root 3794: .B a
1.1.1.5 root 3795: is not null in code like the following:
1.1.1.6 root 3796: .Sp
3797: obj &a = g ();
3798: a.f (2);
3799: .Sp
1.1 root 3800: Checking that references of this sort have non-null values requires
3801: extra code, however, and it is unnecessary for many programs. You can
3802: use `\|\c
1.1.1.5 root 3803: .B \-fnonnull-objects\c
3804: \&\|' to omit the checks for null, if your
3805: program doesn't require checking.
1.1 root 3806: .TP
3807: .B \-fpcc\-struct\-return
3808: Use the same convention for returning \c
3809: .B struct\c
3810: \& and \c
1.1.1.6 root 3811: .B union
1.1 root 3812: values that is used by the usual C compiler on your system. This
3813: convention is less efficient for small structures, and on many
3814: machines it fails to be reentrant; but it has the advantage of
3815: allowing intercallability between GCC-compiled code and PCC-compiled
3816: code.
3817: .TP
1.1.1.5 root 3818: .B \-freg\-struct\-return
3819: Use the convention that
3820: .B struct
3821: and
3822: .B union
3823: values are returned in registers when possible. This is more
3824: efficient for small structures than
3825: .BR \-fpcc\-struct\-return .
1.1.1.6 root 3826: .Sp
1.1.1.5 root 3827: If you specify neither
3828: .B \-fpcc\-struct\-return
3829: nor
3830: .BR \-freg\-struct\-return ,
3831: GNU CC defaults to whichever convention is standard for the target.
3832: If there is no standard convention, GNU CC defaults to
3833: .BR \-fpcc\-struct\-return .
3834: .TP
1.1 root 3835: .B \-fshort\-enums
3836: Allocate to an \c
3837: .B enum\c
3838: \& type only as many bytes as it needs for the
3839: declared range of possible values. Specifically, the \c
3840: .B enum\c
3841: \& type
3842: will be equivalent to the smallest integer type which has enough room.
3843: .TP
3844: .B \-fshort\-double
3845: Use the same size for
3846: .B double
3847: as for
3848: .B float
3849: \&.
3850: .TP
3851: .B \-fshared\-data
3852: Requests that the data and non-\c
3853: .B const\c
3854: \& variables of this
3855: compilation be shared data rather than private data. The distinction
3856: makes sense only on certain operating systems, where shared data is
3857: shared between processes running the same program, while private data
3858: exists in one copy per process.
3859: .TP
3860: .B \-fno\-common
3861: Allocate even uninitialized global variables in the bss section of the
3862: object file, rather than generating them as common blocks. This has the
3863: effect that if the same variable is declared (without \c
3864: .B extern\c
3865: \&) in
3866: two different compilations, you will get an error when you link them.
3867: The only reason this might be useful is if you wish to verify that the
3868: program will work on other systems which always work this way.
3869: .TP
1.1.1.5 root 3870: .B \-fno\-ident
3871: Ignore the `\|\c
3872: .B #ident\c
3873: \&\|' directive.
3874: .TP
3875: .B \-fno\-gnu\-linker
3876: Do not output global initializations (such as C++ constructors and
3877: destructors) in the form used by the GNU linker (on systems where the GNU
3878: linker is the standard method of handling them). Use this option when
3879: you want to use a non-GNU linker, which also requires using the
3880: .B collect2\c
3881: \& program to make sure the system linker includes
3882: constructors and destructors. (\c
3883: .B collect2\c
3884: \& is included in the GNU CC
3885: distribution.) For systems which \c
3886: .I must\c
3887: \& use \c
3888: .B collect2\c
3889: \&, the
3890: compiler driver \c
3891: .B gcc\c
3892: \& is configured to do this automatically.
3893: .TP
3894: .B \-finhibit-size-directive
3895: Don't output a \c
3896: .B .size\c
3897: \& assembler directive, or anything else that
1.1.1.6 root 3898: would cause trouble if the function is split in the middle, and the
1.1.1.5 root 3899: two halves are placed at locations far apart in memory. This option is
3900: used when compiling `\|\c
3901: .B crtstuff.c\c
3902: \&\|'; you should not need to use it
3903: for anything else.
3904: .TP
3905: .B \-fverbose-asm
3906: Put extra commentary information in the generated assembly code to
3907: make it more readable. This option is generally only of use to those
3908: who actually need to read the generated assembly code (perhaps while
3909: debugging the compiler itself).
3910: .TP
1.1 root 3911: .B \-fvolatile
3912: Consider all memory references through pointers to be volatile.
3913: .TP
1.1.1.5 root 3914: .B \-fvolatile\-global
3915: Consider all memory references to extern and global data items to
3916: be volatile.
3917: .TP
1.1 root 3918: .B \-fpic
3919: If supported for the target machines, generate position-independent code,
3920: suitable for use in a shared library.
3921: .TP
3922: .B \-fPIC
3923: If supported for the target machine, emit position-independent code,
3924: suitable for dynamic linking, even if branches need large displacements.
3925: .TP
1.1.1.6 root 3926: .BI "\-ffixed\-" "reg"
1.1 root 3927: Treat the register named \c
3928: .I reg\c
3929: \& as a fixed register; generated code
3930: should never refer to it (except perhaps as a stack pointer, frame
3931: pointer or in some other fixed role).
1.1.1.6 root 3932: .Sp
1.1 root 3933: .I reg\c
3934: \& must be the name of a register. The register names accepted
3935: are machine-specific and are defined in the \c
1.1.1.6 root 3936: .B REGISTER_NAMES
1.1 root 3937: macro in the machine description macro file.
1.1.1.6 root 3938: .Sp
1.1 root 3939: This flag does not have a negative form, because it specifies a
3940: three-way choice.
3941: .TP
1.1.1.6 root 3942: .BI "\-fcall\-used\-" "reg"
1.1 root 3943: Treat the register named \c
3944: .I reg\c
3945: \& as an allocatable register that is
3946: clobbered by function calls. It may be allocated for temporaries or
3947: variables that do not live across a call. Functions compiled this way
3948: will not save and restore the register \c
3949: .I reg\c
3950: \&.
1.1.1.6 root 3951: .Sp
1.1 root 3952: Use of this flag for a register that has a fixed pervasive role in the
3953: machine's execution model, such as the stack pointer or frame pointer,
3954: will produce disastrous results.
1.1.1.6 root 3955: .Sp
1.1 root 3956: This flag does not have a negative form, because it specifies a
3957: three-way choice.
3958: .TP
1.1.1.6 root 3959: .BI "\-fcall\-saved\-" "reg"
1.1 root 3960: Treat the register named \c
3961: .I reg\c
3962: \& as an allocatable register saved by
3963: functions. It may be allocated even for temporaries or variables that
3964: live across a call. Functions compiled this way will save and restore
3965: the register \c
3966: .I reg\c
3967: \& if they use it.
1.1.1.6 root 3968: .Sp
1.1 root 3969: Use of this flag for a register that has a fixed pervasive role in the
3970: machine's execution model, such as the stack pointer or frame pointer,
3971: will produce disastrous results.
1.1.1.6 root 3972: .Sp
1.1 root 3973: A different sort of disaster will result from the use of this flag for
3974: a register in which function values may be returned.
1.1.1.6 root 3975: .Sp
1.1 root 3976: This flag does not have a negative form, because it specifies a
3977: three-way choice.
3978: .SH PRAGMAS
3979: Two `\|\c
3980: .B #pragma\c
3981: \&\|' directives are supported for GNU C++, to permit using the same
3982: header file for two purposes: as a definition of interfaces to a given
3983: object class, and as the full definition of the contents of that object class.
3984: .TP
3985: .B #pragma interface
1.1.1.5 root 3986: (C++ only.)
1.1 root 3987: Use this directive in header files that define object classes, to save
3988: space in most of the object files that use those classes. Normally,
3989: local copies of certain information (backup copies of inline member
3990: functions, debugging information, and the internal tables that
3991: implement virtual functions) must be kept in each object file that
3992: includes class definitions. You can use this pragma to avoid such
3993: duplication. When a header file containing `\|\c
3994: .B #pragma interface\c
3995: \&\|' is included in a compilation, this auxiliary information
3996: will not be generated (unless the main input source file itself uses
3997: `\|\c
3998: .B #pragma implementation\c
3999: \&\|'). Instead, the object files will contain references to be
1.1.1.6 root 4000: resolved at link time.
1.1 root 4001: .TP
4002: .B #pragma implementation
4003: .TP
1.1.1.6 root 4004: \fB#pragma implementation "\fP\fIobjects\fP\fB.h"\fP
1.1.1.5 root 4005: (C++ only.)
1.1 root 4006: Use this pragma in a main input file, when you want full output from
4007: included header files to be generated (and made globally visible).
4008: The included header file, in turn, should use `\|\c
4009: .B #pragma interface\c
1.1.1.6 root 4010: \&\|'.
1.1 root 4011: Backup copies of inline member functions, debugging information, and
4012: the internal tables used to implement virtual functions are all
4013: generated in implementation files.
1.1.1.6 root 4014: .Sp
1.1 root 4015: If you use `\|\c
4016: .B #pragma implementation\c
4017: \&\|' with no argument, it applies to an include file with the same
4018: basename as your source file; for example, in `\|\c
4019: .B allclass.cc\c
4020: \&\|', `\|\c
4021: .B #pragma implementation\c
4022: \&\|' by itself is equivalent to `\|\c
1.1.1.6 root 4023: .B
1.1 root 4024: #pragma implementation "allclass.h"\c
4025: \&\|'. Use the string argument if you want a single implementation
1.1.1.6 root 4026: file to include code from multiple header files.
4027: .Sp
1.1 root 4028: There is no way to split up the contents of a single header file into
1.1.1.6 root 4029: multiple implementation files.
1.1 root 4030: .SH FILES
1.1.1.6 root 4031: .nf
1.1 root 4032: .ta \w'LIBDIR/g++\-include 'u
4033: file.c C source file
4034: file.h C header (preprocessor) file
4035: file.i preprocessed C source file
4036: file.C C++ source file
4037: file.cc C++ source file
4038: file.cxx C++ source file
4039: file.m Objective-C source file
4040: file.s assembly language file
4041: file.o object file
4042: a.out link edited output
4043: \fITMPDIR\fR/cc\(** temporary files
4044: \fILIBDIR\fR/cpp preprocessor
4045: \fILIBDIR\fR/cc1 compiler for C
4046: \fILIBDIR\fR/cc1plus compiler for C++
4047: \fILIBDIR\fR/collect linker front end needed on some machines
4048: \fILIBDIR\fR/libgcc.a GCC subroutine library
4049: /lib/crt[01n].o start-up routine
4050: \fILIBDIR\fR/ccrt0 additional start-up routine for C++
4051: /lib/libc.a standard C library, see
4052: .IR intro (3)
1.1.1.6 root 4053: /usr/include standard directory for \fB#include\fP files
4054: \fILIBDIR\fR/include standard gcc directory for \fB#include\fP files
4055: \fILIBDIR\fR/g++\-include additional g++ directory for \fB#include\fP
4056: .Sp
4057: .fi
1.1 root 4058: .I LIBDIR
4059: is usually
4060: .B /usr/local/lib/\c
4061: .IR machine / version .
4062: .br
4063: .I TMPDIR
1.1.1.6 root 4064: comes from the environment variable
1.1 root 4065: .B TMPDIR
4066: (default
4067: .B /usr/tmp
4068: if available, else
4069: .B /tmp\c
4070: \&).
4071: .SH "SEE ALSO"
4072: cpp(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1).
4073: .br
1.1.1.6 root 4074: .RB "`\|" gcc "\|', `\|" cpp \|',
4075: .RB "`\|" as "\|', `\|" ld \|',
4076: and
1.1 root 4077: .RB `\| gdb \|'
4078: entries in
4079: .B info\c
4080: \&.
4081: .br
1.1.1.6 root 4082: .I
1.1 root 4083: Using and Porting GNU CC (for version 2.0)\c
1.1.1.6 root 4084: , Richard M. Stallman;
1.1 root 4085: .I
4086: The C Preprocessor\c
1.1.1.5 root 4087: , Richard M. Stallman;
1.1.1.6 root 4088: .I
1.1.1.5 root 4089: Debugging with GDB: the GNU Source-Level Debugger\c
4090: , Richard M. Stallman and Roland H. Pesch;
1.1 root 4091: .I
4092: Using as: the GNU Assembler\c
1.1.1.5 root 4093: , Dean Elsner, Jay Fenlason & friends;
1.1 root 4094: .I
1.1.1.5 root 4095: ld: the GNU linker\c
4096: , Steve Chamberlain and Roland Pesch.
1.1 root 4097: .SH BUGS
1.1.1.5 root 4098: For instructions on reporting bugs, see the GCC manual.
1.1 root 4099: .SH COPYING
1.1.1.6 root 4100: Copyright
4101: .if t \(co
4102: 1991, 1992, 1993 Free Software Foundation, Inc.
1.1 root 4103: .PP
4104: Permission is granted to make and distribute verbatim copies of
4105: this manual provided the copyright notice and this permission notice
4106: are preserved on all copies.
4107: .PP
4108: Permission is granted to copy and distribute modified versions of this
4109: manual under the conditions for verbatim copying, provided that the
4110: entire resulting derived work is distributed under the terms of a
4111: permission notice identical to this one.
4112: .PP
4113: Permission is granted to copy and distribute translations of this
4114: manual into another language, under the above conditions for modified
4115: versions, except that this permission notice may be included in
4116: translations approved by the Free Software Foundation instead of in
4117: the original English.
4118: .SH AUTHORS
4119: See the GNU CC Manual for the contributors to GNU CC.
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.