EditPageHistoryDiffInfoLikePages

Annotated edit history of cc(1) version 2, including all changes. View license author blame.

Rev	Author	#	Line
1	perry	1	GCC
		2	!!!GCC
		3	NAME
		4	SYNOPSIS
		5	WARNING
		6	DESCRIPTION
		7	OPTIONS
		8	OVERALL OPTIONS
		9	LANGUAGE OPTIONS
		10	PREPROCESSOR OPTIONS
		11	ASSEMBLER OPTION
		12	LINKER OPTIONS
		13	DIRECTORY OPTIONS
		14	WARNING OPTIONS
		15	DEBUGGING OPTIONS
		16	OPTIMIZATION OPTIONS
		17	TARGET OPTIONS
		18	MACHINE DEPENDENT OPTIONS
		19	CODE GENERATION OPTIONS
		20	PRAGMAS
		21	FILES
		22	EXIT STATUS
		23	SEE ALSO
		24	BUGS
		25	COPYING
		26	AUTHORS
		27	----
		28	!!NAME
		29
		30
		31	gcc, g++ - GNU project C and C++ Compiler (gcc-2.95)
		32	!!SYNOPSIS
		33
		34
		35	__gcc__ [[ ''option'' | ''filename'' ]...__
		36	g++__ [[ ''option'' | ''filename'' ]...
		37	!!WARNING
		38
		39
		40	The information in this man page is an extract from the full
		41	documentation of the GNU C compiler, and is limited to the
		42	meaning of the options.
		43
		44
		45	This man page is not kept up to date except when volunteers
		46	want to maintain it. If you find a discrepancy between the
		47	man page and the software, please check the Info file, which
		48	is the authoritative documentation.
		49
		50
		51	If we find that the things in this man page that are out of
		52	date cause significant confusion or complaints, we will stop
		53	distributing the man page. The alternative, updating the man
		54	page when we update the Info file, is impossible because the
		55	rest of the work of maintaining GNU CC leaves us no time for
		56	that. The GNU project regards man pages as obsolete and
		57	should not let them take time away from other
		58	things.
		59
		60
		61	For complete and current documentation, refer to the Info
		62	file `__gcc__' or the manual ''Using and Porting GNU CC
		63	(for version 2.0)''. Both are made from the Texinfo source
		64	file __gcc.texinfo__.
		65	!!DESCRIPTION
		66
		67
		68	The C and C++ compilers are integrated. Both process input
		69	files through one or more of four stages: preprocessing,
		70	compilation, assembly, and linking. Source filename suffixes
		71	identify the source language, but which name you use for the
		72	compiler governs default assumptions:
		73
		74
		75	__gcc__
		76
		77
		78	assumes preprocessed (__.i__) files are C and assumes C
		79	style linking.
		80
		81
		82	__g++__
		83
		84
		85	assumes preprocessed (__.i__) files are C++ and assumes
		86	C++ style linking.
		87
		88
		89	Suffixes of source file names indicate the language and kind
		90	of processing to be done:
		91
		92
		93	__.c__ C source; preprocess, compile, assemble
		94	__.C__ C++ source; preprocess, compile, assemble
		95	__.cc__ C++ source; preprocess, compile, assemble
		96	__.cxx__ C++ source; preprocess, compile, assemble
		97	__.m__ Objective-C source; preprocess, compile, assemble
		98	__.i__ preprocessed C; compile, assemble
		99	__.ii__ preprocessed C++; compile, assemble
		100	__.s__ Assembler source; assemble
		101	__.S__ Assembler source; preprocess, assemble
		102	__.h__ Preprocessor file; not usually named on command line
		103	Files with other suffixes are passed to the linker. Common cases include:
		104
		105
		106	__.o__ Object file
		107	__.a__ Archive file
		108	Linking is always the last stage unless you use one of the __-c__, __-S__, or __-E__ options to avoid it (or unless compilation errors stop the whole process). For the link stage, all __.o__ files corresponding to source files, __-l__ libraries, unrecognized filenames (including named __.o__ object files and __.a__ archives) are passed to the linker in command-line order.
		109	!!OPTIONS
		110
		111
		112	Options must be separate: `__-dr__' is quite different
		113	from `__-d -r__ '.
		114
		115
		116	Most `__-f__' and `__-W__' options have two contrary
		117	forms: __-f__''name'' and __-fno-__''name'' (or
		118	__-W__''name'' and __-Wno-__''name''). Only the
		119	non-default forms are shown here.
		120
		121
		122	Here is a summary of all the options, grouped by type.
		123	Explanations are in the following sections.
		124
		125
		126	__Overall Options__
		127
		128
		129	-c -S -E -o ''file'' -pipe -v -x
		130	''language''
		131
		132
		133	__Language Options__
		134
		135
		136	-ansi -fall-virtual -fcond-mismatch -fdollars-in-identifiers
		137	-fenum-int-equiv -fexternal-templates -fno-asm -fno-builtin
		138	-fhosted -fno-hosted -ffreestanding -fno-freestanding
		139	-fno-strict-prototype -fsigned-bitfields -fsigned-char
		140	-fthis-is-variable -funsigned-bitfields -funsigned-char
		141	-fwritable-strings -traditional -traditional-cpp
		142	-trigraphs
		143
		144
		145	__Warning Options__
		146
		147
		148	-fsyntax-only -pedantic -pedantic-errors -w -W -Wall
		149	-Waggregate-return -Wcast-align -Wcast-qual -Wchar-subscript
		150	-Wcomment -Wconversion -Wenum-clash -Werror -Wformat
		151	-Wid-clash-''len'' -Wimplicit -Wimplicit-int
		152	-Wimplicit-function-declaration -Winline -Wlong-long -Wmain
		153	-Wmissing-prototypes -Wmissing-declarations -Wnested-externs
		154	-Wno-import -Wparentheses -Wpointer-arith -Wredundant-decls
		155	-Wreturn-type -Wshadow -Wstrict-prototypes -Wswitch
		156	-Wtemplate-debugging -Wtraditional -Wtrigraphs
		157	-Wuninitialized -Wunused -Wwrite-strings
		158
		159
		160	__Debugging Options__
		161
		162
		163	-a -d''letters'' -fpretend-float -g -g''level'' -gcoff
		164	-gxcoff -gxcoff+ -gdwarf -gdwarf+ -gstabs -gstabs+ -ggdb -p
		165	-pg -save-temps -print-file-name=''library''
		166	-print-libgcc-file-name
		167	-print-prog-name=''program''
		168
		169
		170	__Optimization Options__
		171
		172
		173	-fcaller-saves -fcse-follow-jumps -fcse-skip-blocks
		174	-fdelayed-branch -felide-constructors
		175	-fexpensive-optimizations -ffast-math -ffloat-store
		176	-fforce-addr -fforce-mem -finline-functions
		177	-fkeep-inline-functions -fmemoize-lookups
		178	-fno-default-inline -fno-defer-pop -fno-function-cse
		179	-fno-inline -fno-peephole -fomit-frame-pointer
		180	-frerun-cse-after-loop -fschedule-insns -fschedule-insns2
		181	-fstrength-reduce -fthread-jumps -funroll-all-loops
		182	-funroll-loops -O -O2 -O3
		183
		184
		185	__Preprocessor Options__
		186
		187
		188	-A''assertion'' -C -dD -dM -dN
		189	-D''macro''[[=''defn''] -E -H -idirafter ''dir''
		190	-include ''file'' -imacros ''file'' -iprefix
		191	''file'' -iwithprefix ''dir'' -M -MD -MM -MMD
		192	-nostdinc -P -U''macro'' -undef
		193
		194
		195	__Assembler Option__
		196
		197
		198	-Wa,''option''
		199
		200
		201	__Linker Options__
		202
		203
		204	-l''library'' -nostartfiles -nostdlib -static -shared
		205	-symbolic -Xlinker ''option'' -Wl,''option'' -u
		206	''symbol''
		207
		208
		209	__Directory Options__
		210
		211
		212	-B''prefix'' -I''dir'' -I- -L''dir''
		213
		214
		215	__Target Options__
		216
		217
		218	-b ''machine'' -V ''version''
		219
		220
		221	__Configuration Dependent Options__
		222
		223
		224	''M680x0 Options''
		225	-m68000 -m68020 -m68020-40 -m68030 -m68040 -m68881
		226	-mbitfield -mc68000 -mc68020 -mfpa -mnobitfield -mrtd
		227	-mshort -msoft-float
		228
		229
		230	''VAX Options''
		231	-mg -mgnu -munix
		232
		233
		234	''SPARC Options''
		235	-mepilogue -mfpu -mhard-float -mno-fpu -mno-epilogue
		236	-msoft-float -msparclite -mv8 -msupersparc
		237	-mcypress
		238
		239
		240	''Convex Options''
		241	-margcount -mc1 -mc2 -mnoargcount
		242
		243
		244	''AMD29K Options''
		245	-m29000 -m29050 -mbw -mdw -mkernel-registers -mlarge -mnbw
		246	-mnodw -msmall -mstack-check -muser-registers
		247
		248
		249	''M88K Options''
		250	-m88000 -m88100 -m88110 -mbig-pic -mcheck-zero-division
		251	-mhandle-large-shift -midentify-revision
		252	-mno-check-zero-division -mno-ocs-debug-info
		253	-mno-ocs-frame-position -mno-optimize-arg-area
		254	-mno-serialize-volatile -mno-underscores -mocs-debug-info
		255	-mocs-frame-position -moptimize-arg-area
		256	-mserialize-volatile -mshort-data-''num'' -msvr3 -msvr4
		257	-mtrap-large-shift -muse-div-instruction -mversion-03.00
		258	-mwarn-passed-structs
		259
		260
		261	''RS6000 Options''
		262	-mfp-in-toc -mno-fop-in-toc
		263
		264
		265	''RT Options''
		266	-mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs
		267	-mfull-fp-blocks -mhc-struct-return -min-line-mul
		268	-mminimum-fp-blocks -mnohc-struct-return
		269
		270
		271	''MIPS Options''
		272	-mcpu=''cpu type'' -mips2 -mips3 -mint64 -mlong64
		273	-mlonglong128 -mmips-as -mgas -mrnames -mno-rnames -mgpopt
		274	-mno-gpopt -mstats -mno-stats -mmemcpy -mno-memcpy
		275	-mno-mips-tfile -mmips-tfile -msoft-float -mhard-float
		276	-mabicalls -mno-abicalls -mhalf-pic -mno-half-pic -G
		277	''num'' -nocpp
		278
		279
		280	''i386 Options''
		281	-m486 -mno-486 -msoft-float -mno-fp-ret-in-387
		282
		283
		284	''HPPA Options''
		285	-mpa-risc-1-0 -mpa-risc-1-1 -mkernel -mshared-libs
		286	-mno-shared-libs -mlong-calls -mdisable-fpregs
		287	-mdisable-indexing -mtrailing-colon
		288
		289
		290	''i960 Options''
		291	-m''cpu-type'' -mnumerics -msoft-float -mleaf-procedures
		292	-mno-leaf-procedures -mtail-call -mno-tail-call
		293	-mcomplex-addr -mno-complex-addr -mcode-align
		294	-mno-code-align -mic-compat -mic2.0-compat -mic3.0-compat
		295	-masm-compat -mintel-asm -mstrict-align -mno-strict-align
		296	-mold-align -mno-old-align
		297
		298
		299	''DEC Alpha Options''
		300	-mfp-regs -mno-fp-regs -mno-soft-float
		301	-msoft-float
		302
		303
		304	''System V Options''
		305	-G -Qy -Qn -YP,''paths'' -Ym,''dir''
		306
		307
		308	__Code Generation Options__
		309
		310
		311	-fcall-saved-''reg'' -fcall-used-''reg''
		312	-ffixed-''reg'' -finhibit-size-directive
		313	-fnonnull-objects -fno-common -fno-ident -fno-gnu-linker
		314	-fpcc-struct-return -fpic -fPIC -freg-struct-return
		315	-fshared-data -fshort-enums -fshort-double -fvolatile
		316	-fvolatile-global -fverbose-asm
		317	!!OVERALL OPTIONS
		318
		319
		320	__-x__ ''language''
		321
		322
		323	Specify explicitly the ''language'' for the following
		324	input files (rather than choosing a default based on the
		325	file name suffix) . This option applies to all following
		326	input files until the next `__-x__' op- tion. Possible
		327	values of ''language'' are `__c__', `__objec-
		328	tive-c__', `__c-header__', `__c++__',
		329	`__cpp-output__', `__assem- bler__', and
		330	`__assembler-with-cpp__'.
		331
		332
		333	__-x none__
		334
		335
		336	Turn off any specification of a language, so that subsequent
		337	files are handled according to their file name suffixes (as
		338	they are if `__-x__' has not been used at
		339	all).
		340
		341
		342	If you want only some of the four stages (preprocess, com-
		343	pile, assemble, link), you can use `__-x__' (or filename
		344	suf- fixes) to tell __gcc__ where to start, and one of
		345	the options `__-c__', `__-S__', or `__-E__' to say
		346	where __gcc__ is to stop. Note that some combinations
		347	(for example, `__-x cpp-output -E__') instruct __gcc__
		348	to do nothing at all.
		349
		350
		351	__-c__
		352
		353
		354	Compile or assemble the source files, but do not link. The
		355	compiler output is an object file corre- sponding to each
		356	source file.
		357
		358
		359	By default, GCC makes the object file name for a source file
		360	by replacing the suffix `__.c__', `__.i__',
		361	`__.s__', etc., with `__.o__'. Use __-o__ to select
		362	another name.
		363
		364
		365	GCC ignores any unrecognized input files (those that do not
		366	require compilation or assembly) with the __-c__
		367	option.
		368
		369
		370	__-S__
		371
		372
		373	Stop after the stage of compilation proper; do not assemble.
		374	The output is an assembler code file for each non-assembler
		375	input file specified.
		376
		377
		378	By default, GCC makes the assembler file name for a source
		379	file by replacing the suffix `__.c__', `__.i__', etc.,
		380	with `__.s__'. Use __-o__ to select another
		381	name.
		382
		383
		384	GCC ignores any input files that don't require com-
		385	pilation.
		386
		387
		388	__-E__
		389
		390
		391	Stop after the preprocessing stage; do not run the compiler
		392	proper. The output is preprocessed source code, which is
		393	sent to the standard output.
		394
		395
		396	GCC ignores input files which don't require prepro-
		397	cessing.
		398
		399
		400	__-o__ ''file''
		401
		402
		403	Place output in file ''file''. This applies regardless to
		404	whatever sort of output GCC is producing, whether it be an
		405	executable file, an object file, an assembler file or
		406	preprocessed C code.
		407
		408
		409	Since only one output file can be specified, it does not
		410	make sense to use `__-o__' when compiling more than one
		411	input file, unless you are producing an executable file as
		412	output.
		413
		414
		415	If you do not specify `__-o__', the default is to put an
		416	executable file in `__a.out__', the object file for
		417	`''source''__.__''suffix''' in
		418	`''source''__.o__', its assembler file in
		419	`''source''__.s__', and all preprocessed C source on
		420	standard output.
		421
		422
		423	__-v__
		424
		425
		426	Print (on standard error output) the commands exe- cuted to
		427	run the stages of compilation. Also print the version number
		428	of the compiler driver program and of the preprocessor and
		429	the compiler proper.
		430
		431
		432	__-pipe__
		433
		434
		435	Use pipes rather than temporary files for communi- cation
		436	between the various stages of compilation. This fails to
		437	work on some systems where the assem- bler cannot read from
		438	a pipe; but the GNU assembler has no trouble.
		439	!!LANGUAGE OPTIONS
		440
		441
		442	The following options control the dialect of C that the
		443	compiler accepts:
		444
		445
		446	__-ansi__
		447
		448
		449	Support all ANSI standard C programs.
		450
		451
		452	This turns off certain features of GNU C that are
		453	incompatible with ANSI C, such as the __asm__,
		454	__inline__ and __typeof__ keywords, and predefined
		455	macros such as __unix__ and __vax__ that identify the
		456	type of system you are using. It also enables the
		457	undesirable and rarely used ANSI trigraph feature, and
		458	disallows `__$__' as part of identifiers.
		459
		460
		461	The alternate keywords ____asm____, ____extension____,
		462	____inline____ and ____typeof____ continue to work
		463	despite `__-ansi__'. You would not want to use them in an
		464	ANSI C program, of course, but it is useful to put them in
		465	header files that might be included in compila- tions done
		466	with `__-ansi__'. Alternate predefined macros such as
		467	____unix____ and ____vax____ are also avail- able,
		468	with or without `__-ansi__'.
		469
		470
		471	The `__-ansi__' option does not cause non-ANSI programs
		472	to be rejected gratuitously. For that, `__-pedantic__' is
		473	required in addition to `__-ansi__'.
		474
		475
		476	The preprocessor predefines a macro ____STRICT_ANSI____
		477	when you use the `__-ansi__' option. Some header files
		478	may notice this macro and refrain from declaring certain
		479	functions or defining certain macros that the ANSI standard
		480	doesn't call for; this is to avoid interfering with any
		481	programs that might use these names for other
		482	things.
		483
		484
		485	__-fno-asm__
		486
		487
		488	Do not recognize __asm__, __inline__ or __typeof__
		489	as a key- word. These words may then be used as identifiers.
		490	You can use ____asm____, ____inline____ and
		491	____typeof____ in- stead. `__-ansi__' implies
		492	`__-fno-asm__'.
		493
		494
		495	__-fno-builtin__
		496
		497
		498	Don't recognize built-in functions that do not be- gin with
		499	two leading underscores. Currently, the functions affected
		500	include ___exit__, __abort__, __abs__, __allo-
		501	ca__, __cos__, __exit__, __fabs__, __labs__,
		502	__memcmp__, __memcpy__, __sin__, __sqrt__,
		503	__strcmp__, __strcpy__, and
		504	__strlen__.
		505
		506
		507	The `__-ansi__' option prevents __alloca__ and
		508	___exit__ from being builtin functions.
		509
		510
		511	__-fhosted__
		512
		513
		514	Compile for a hosted environment; this implies the
		515	`__-fbuiltin__' option, and implies that suspicious
		516	declarations of __main__ should be warned
		517	about.
		518
		519
		520	__-ffreestanding__
		521
		522
		523	Compile for a freestanding environment; this im- plies the
		524	`__-fno-builtin__' option, and implies that __main__
		525	has no special requirements.
		526
		527
		528	__-fno-strict-prototype__
		529
		530
		531	Treat a function declaration with no arguments, such as
		532	`__int foo__ ();', as C would treat it--as say- ing
		533	nothing about the number of arguments or their types (C++
		534	only). Normally, such a declaration in C++ means that the
		535	function __foo__ takes no arguments.
		536
		537
		538	__-trigraphs__
		539
		540
		541	Support ANSI C trigraphs. The `__-ansi__' option im-
		542	plies `__-trigraphs__'.
		543
		544
		545	__-traditional__
		546
		547
		548	Attempt to support some aspects of traditional C compilers.
		549	For details, see the GNU C Manual; the duplicate list here
		550	has been deleted so that we won't get complaints when it is
		551	out of date.
		552
		553
		554	But one note about C++ programs only (not C).
		555	`__-traditional__' has one additional effect for C++:
		556	assignment to __this__ is permitted. This is the same as
		557	the effect of `__-fthis-is-variable__'.
		558
		559
		560	__-traditional-cpp__
		561
		562
		563	Attempt to support some aspects of traditional C
		564	preprocessors. This includes the items that specifically
		565	mention the preprocessor above, but none of the other
		566	effects of `__-traditional__'.
		567
		568
		569	__-fdollars-in-identifiers__
		570
		571
		572	Permit the use of `__$__' in identifiers (C++ only). You
		573	can also use `__-fno-dollars-in-identifiers__' to
		574	explicitly prohibit use of `__$__'. (GNU C++ allows
		575	`__$__' by default on some target systems but not oth-
		576	ers.)
		577
		578
		579	__-fenum-int-equiv__
		580
		581
		582	Permit implicit conversion of __int__ to enumeration
		583	types (C++ only). Normally GNU C++ allows conver- sion of
		584	__enum__ to __int__, but not the other way
		585	around.
		586
		587
		588	__-fexternal-templates__
		589
		590
		591	Produce smaller code for template declarations, by
		592	generating only a single copy of each template function
		593	where it is defined (C++ only). To use this option
		594	successfully, you must also mark all files that use
		595	templates with either `__#pragma im- plementation__' (the
		596	definition) or `__#pragma inter- face__'
		597	(declarations).
		598
		599
		600	When your code is compiled with `__-fexternal-tem-
		601	plates__', all template instantiations are external. You
		602	must arrange for all necessary instantiations to appear in
		603	the implementation file; you can do this with a
		604	__typedef__ that references each instantia- tion needed.
		605	Conversely, when you compile using the default option
		606	`__-fno-external-templates__', all template
		607	instantiations are explicitly internal.
		608
		609
		610	__-fall-virtual__
		611
		612
		613	Treat all possible member functions as virtual, im-
		614	plicitly. All member functions (except for con- structor
		615	functions and __new__ or __delete__ member opera-
		616	tors) are treated as virtual functions of the class where
		617	they appear.
		618
		619
		620	This does not mean that all calls to these member functions
		621	will be made through the internal table of virtual
		622	functions. Under some circumstances, the compiler can
		623	determine that a call to a given virtual function can be
		624	made directly; in these cases the calls are direct in any
		625	case.
		626
		627
		628	__-fcond-mismatch__
		629
		630
		631	Allow conditional expressions with mismatched types in the
		632	second and third arguments. The value of such an expression
		633	is void.
		634
		635
		636	__-fthis-is-variable__
		637
		638
		639	Permit assignment to __this__ (C++ only). The incorpo-
		640	ration of user-defined free store management into C++ has
		641	made assignment to `__this__' an anachronism. Therefore,
		642	by default it is invalid to assign to __this__ within a
		643	class member function. However, for backwards compatibility,
		644	you can make it valid with
		645	`__-fthis-is-variable__'.
		646
		647
		648	__-funsigned-char__
		649
		650
		651	Let the type __char__ be unsigned, like __unsigned
		652	char__.
		653
		654
		655	Each kind of machine has a default for what __char__
		656	should be. It is either like __unsigned char__ by de-
		657	fault or like __signed char__ by default.
		658
		659
		660	Ideally, a portable program should always use __signed
		661	char__ or __unsigned char__ when it depends on the
		662	signedness of an object. But many programs have been written
		663	to use plain __char__ and expect it to be signed, or
		664	expect it to be unsigned, depending on the machines they
		665	were written for. This option, and its inverse, let you make
		666	such a program work with the opposite default.
		667
		668
		669	The type __char__ is always a distinct type from each of
		670	__signed char__ and __unsigned char__, even though its
		671	behavior is always just like one of those two.
		672
		673
		674	__-fsigned-char__
		675
		676
		677	Let the type __char__ be signed, like __signed
		678	char__.
		679
		680
		681	Note that this is equivalent to `__-fno-un-
		682	signed-char__', which is the negative form of `__-fun-
		683	signed-char__'. Likewise, `__-fno-signed-char__' is
		684	equivalent to `__-funsigned-char__'.
		685
		686
		687	__-fsigned-bitfields__
		688
		689
		690	__-funsigned-bitfields__
		691
		692
		693	__-fno-signed-bitfields__
		694
		695
		696	__-fno-unsigned-bitfields__
		697
		698
		699	These options control whether a bitfield is signed or
		700	unsigned, when declared with no explicit `__signed__' or
		701	`__unsigned__' qualifier. By default, such a bitfield is
		702	signed, because this is consistent: the basic integer types
		703	such as __int__ are signed types.
		704
		705
		706	However, when you specify `__-traditional__', bitfields
		707	are all unsigned no matter what.
		708
		709
		710	__-fwritable-strings__
		711
		712
		713	Store string constants in the writable data segment and
		714	don't uniquize them. This is for compatibility with old
		715	programs which assume they can write into string constants.
		716	`__-traditional__' also has this ef- fect.
		717
		718
		719	Writing into string constants is a very bad idea;
		720	``constants'' should be constant.
		721	!!PREPROCESSOR OPTIONS
		722
		723
		724	These options control the C preprocessor, which is run on
		725	each C source file before actual compilation.
		726
		727
		728	If you use the `__-E__' option, GCC does nothing except
		729	pre- processing. Some of these options make sense only
		730	togeth- er with `__-E__' because they cause the
		731	preprocessor output to be unsuitable for actual
		732	compilation.
		733
		734
		735	__-include__ ''file''
		736
		737
		738	Process ''file'' as input before processing the regular
		739	input file. In effect, the contents of ''file'' are
		740	compiled first. Any `__-D__' and `__-U__' options on
		741	the command line are always processed before
		742	`__-include__ ''file''', regardless of the order in
		743	which they are written. All the `__-include__' and
		744	`__-imacros__' options are processed in the order in
		745	which they are writ- ten.
		746
		747
		748	__-imacros__ ''file''
		749
		750
		751	Process ''file'' as input, discarding the resulting
		752	output, before processing the regular input file. Because
		753	the output generated from ''file'' is discard- ed, the
		754	only effect of `__-imacros__ ''file''' is to make the
		755	macros defined in ''file'' available for use in the main
		756	input. The preprocessor evaluates any `__-D__' and
		757	`__-U__' options on the command line before pro- cessing
		758	`__-imacros__''file''', regardless of the order in
		759	which they are written. All the `__-include__' and
		760	`__-imacros__' options are processed in the order in
		761	which they are written.
		762
		763
		764	__-idirafter__ ''dir''
		765
		766
		767	Add the directory ''dir'' to the second include path. The
		768	directories on the second include path are searched when a
		769	header file is not found in any of the directories in the
		770	main include path (the one that `__-I__' adds
		771	to).
		772
		773
		774	__-iprefix__ ''prefix''
		775
		776
		777	Specify ''prefix'' as the prefix for subsequent
		778	`__-iwithprefix__' options.
		779
		780
		781	__-iwithprefix__ ''dir''
		782
		783
		784	Add a directory to the second include path. The directory's
		785	name is made by concatenating ''prefix'' and ''dir'',
		786	where ''prefix'' was specified previously with
		787	`__-iprefix__'.
		788
		789
		790	__-nostdinc__
		791
		792
		793	Do not search the standard system directories for header
		794	files. Only the directories you have speci- fied with
		795	`__-I__' options (and the current directory, if
		796	appropriate) are searched.
		797
		798
		799	By using both `__-nostdinc__' and `__-I-__', you can
		800	limit the include-file search file to only those directo-
		801	ries you specify explicitly.
		802
		803
		804	__-nostdinc++__
		805
		806
		807	Do not search for header files in the C++-specific standard
		808	directories, but do still search the other standard
		809	directories. (This option is used when building
		810	`__libg++__'.)
		811
		812
		813	__-undef__
		814
		815
		816	Do not predefine any nonstandard macros. (Includ- ing
		817	architecture flags).
		818
		819
		820	__-E__
		821
		822
		823	Run only the C preprocessor. Preprocess all the C source
		824	files specified and output the results to standard output or
		825	to the specified output file.
		826
		827
		828	__-C__
		829
		830
		831	Tell the preprocessor not to discard comments. Used with the
		832	`__-E__' option.
		833
		834
		835	__-P__
		836
		837
		838	Tell the preprocessor not to generate `__#line__' com-
		839	mands. Used with the `__-E__' option.
		840
		841
		842	__-M [[ -MG ]__
		843
		844
		845	Tell the preprocessor to output a rule suitable for
		846	__make__ describing the dependencies of each object file.
		847	For each source file, the preprocessor out- puts one
		848	__make__-rule whose target is the object file name for
		849	that source file and whose dependencies are all the files
		850	`__#include__'d in it. This rule may be a single line or
		851	may be continued with `__\__'-new- line if it is long.
		852	The list of rules is printed on standard output instead of
		853	the preprocessed C program.
		854
		855
		856	`__-M__' implies `__-E__'.
		857
		858
		859	`__-MG__' says to treat missing header files as gener-
		860	ated files and assume they live in the same direc- tory as
		861	the source file. It must be specified in addition to
		862	`__-M__'.
		863
		864
		865	__-MM [[ -MG ]__
		866
		867
		868	Like `__-M__' but the output mentions only the user
		869	header files included with `__#include__
		870	''file''
		871	''#include ''''file''
		872	''
		873
		874
		875	__-MD__
		876
		877
		878	Like `__-M__' but the dependency information is written
		879	to files with names made by replacing `__.o__' with
		880	`__.d__' at the end of the output file names. This is in
		881	addition to compiling the file as speci- fied--`__-MD__'
		882	does not inhibit ordinary compilation the way `__-M__'
		883	does.
		884
		885
		886	The Mach utility `__md__' can be used to merge the
		887	`__.d__' files into a single dependency file suitable for
		888	using with the `__make__' command.
		889
		890
		891	__-MMD__
		892
		893
		894	Like `__-MD__' except mention only user header files, not
		895	system header files.
		896
		897
		898	__-H__
		899
		900
		901	Print the name of each header file used, in addi- tion to
		902	other normal activities.
		903
		904
		905	__-A__''question''__(__''answer''__)__
		906
		907
		908	Assert the answer ''answer'' for ''question'', in case
		909	it is tested with a preprocessor conditional such as `__#if
		910	#__''question''__(__''answer''__)__'.
		911	`__-A-__' disables the stan- dard assertions that
		912	normally describe the target machine.
		913
		914
		915	__-A__''question''
		916
		917
		918	(''answer'') Assert the answer ''answer'' for
		919	''question'', in case it is tested with a preprocessor
		920	conditional such as `__#if__
		921	#''question''(''answer'')'. `__-A-__' disables the
		922	standard assertions that normally describe the target
		923	machine.
		924
		925
		926	__-D__''macro''
		927
		928
		929	Define macro ''macro'' with the string `__1__' as its
		930	defi- nition.
		931
		932
		933	__-D__''macro''__=__''defn''
		934
		935
		936	Define macro ''macro'' as ''defn''. All instances of
		937	`__-D__' on the command line are processed before any
		938	`__-U__' options.
		939
		940
		941	__-U__''macro''
		942
		943
		944	Undefine macro ''macro''. `__-U__' options are
		945	evaluated after all `__-D__' options, but before any
		946	`__-include__' and `__-imacros__'
		947	options.
		948
		949
		950	__-dM__
		951
		952
		953	Tell the preprocessor to output only a list of the macro
		954	definitions that are in effect at the end of preprocessing.
		955	Used with the `__-E__' option.
		956
		957
		958	__-dD__
		959
		960
		961	Tell the preprocessor to pass all macro definitions into the
		962	output, in their proper sequence in the rest of the
		963	output.
		964
		965
		966	__-dN__
		967
		968
		969	Like `__-dD__' except that the macro arguments and con-
		970	tents are omitted. Only `__#define__ ''name''' is
		971	included in the output.
		972	!!ASSEMBLER OPTION
		973
		974
		975	__-Wa,__''option''
		976
		977
		978	Pass ''option'' as an option to the assembler. If ''op-
		979	tion'' contains commas, it is split into multiple op-
		980	tions at the commas.
		981	!!LINKER OPTIONS
		982
		983
		984	These options come into play when the compiler links ob-
		985	ject files into an executable output file. They are mean-
		986	ingless if the compiler is not doing a link
		987	step.
		988
		989
		990	''object-file-name''
		991
		992
		993	A file name that does not end in a special recog- nized
		994	suffix is considered to name an object file or library.
		995	(Object files are distinguished from libraries by the linker
		996	according to the file con- tents.) If GCC does a link step,
		997	these object files are used as input to the
		998	linker.
		999
		1000
		1001	__-l__''library''
		1002
		1003
		1004	Use the library named ''library'' when
		1005	linking.
		1006
		1007
		1008	The linker searches a standard list of directories for the
		1009	library, which is actually a file named
		1010	`__lib__''library''.a'. The linker then uses this file
		1011	as if it had been specified precisely by name.
		1012
		1013
		1014	The directories searched include several standard system
		1015	directories plus any that you specify with
		1016	`__-L__'.
		1017
		1018
		1019	Normally the files found this way are library files--archive
		1020	files whose members are object files. The linker handles an
		1021	archive file by scan- ning through it for members which
		1022	define symbols that have so far been referenced but not
		1023	defined. However, if the linker finds an ordinary object
		1024	file rather than a library, the object file is linked in the
		1025	usual fashion. The only difference between using an
		1026	`__-l__' option and specifying a file name is that
		1027	`__-l__' surrounds ''library'' with `__lib__' and
		1028	`__.a__' and searches several directories.
		1029
		1030
		1031	__-lobjc__
		1032
		1033
		1034	You need this special case of the __-l__ option in or-
		1035	der to link an Objective C program.
		1036
		1037
		1038	__-nostartfiles__
		1039
		1040
		1041	Do not use the standard system startup files when linking.
		1042	The standard libraries are used normally.
		1043
		1044
		1045	__-nostdlib__
		1046
		1047
		1048	Don't use the standard system libraries and startup files
		1049	when linking. Only the files you specify will be passed to
		1050	the linker.
		1051
		1052
		1053	__-static__
		1054
		1055
		1056	On systems that support dynamic linking, this pre- vents
		1057	linking with the shared libraries. On other systems, this
		1058	option has no effect.
		1059
		1060
		1061	__-shared__
		1062
		1063
		1064	Produce a shared object which can then be linked with other
		1065	objects to form an executable. Only a few systems support
		1066	this option.
		1067
		1068
		1069	__-symbolic__
		1070
		1071
		1072	Bind references to global symbols when building a shared
		1073	object. Warn about any unresolved refer- ences (unless
		1074	overridden by the link editor option `__-Xlinker -z
		1075	-Xlinker defs__'). Only a few systems support this
		1076	option.
		1077
		1078
		1079	__-Xlinker__ ''option''
		1080
		1081
		1082	Pass ''option'' as an option to the linker. You can use
		1083	this to supply system-specific linker options which GNU CC
		1084	does not know how to recognize.
		1085
		1086
		1087	If you want to pass an option that takes an argu- ment, you
		1088	must use `__-Xlinker__' twice, once for the option and
		1089	once for the argument. For example, to pass `__-assert
		1090	definitions__', you must write `__-Xlink- er -assert
		1091	-Xlinker definitions__'. It does not work to write
		1092	`__-Xlinker __', because
		1093	this passes the entire string as a single argument, which is
		1094	not what the linker expects.
		1095
		1096
		1097	__-Wl,__''option''
		1098
		1099
		1100	Pass ''option'' as an option to the linker. If
		1101	''option'' contains commas, it is split into multiple
		1102	options at the commas.
		1103
		1104
		1105	__-u__ ''symbol''
		1106
		1107
		1108	Pretend the symbol ''symbol'' is undefined, to force
		1109	linking of library modules to define it. You can use
		1110	`__-u__' multiple times with different symbols to force
		1111	loading of additional library modules.
		1112	!!DIRECTORY OPTIONS
		1113
		1114
		1115	These options specify directories to search for header
		1116	files, for libraries and for parts of the
		1117	compiler:
		1118
		1119
		1120	__-I__''dir''
		1121
		1122
		1123	Append directory ''dir'' to the list of directories
		1124	searched for include files.
		1125
		1126
		1127	__-I-__
		1128
		1129
		1130	Any directories you specify with `__-I__' options be-
		1131	fore the `__-I-__' option are searched only for the case
		1132	of `__#include __''file''____'; they
		1133	are not searched for `__#include
		1134	__''file''''
		1135
		1136
		1137	If additional directories are specified with `__-I__'
		1138	options after the `__-I-__', these directories are
		1139	searched for all `__#include__' directives. (Ordinari- ly
		1140	''all'' `__-I__' directories are used this
		1141	way.)
		1142
		1143
		1144	In addition, the `__-I-__' option inhibits the use of the
		1145	current directory (where the current input file came from)
		1146	as the first search directory for `__#in- clude
		1147	__''file''____'. There is no way to
		1148	override this effect of `__-I-__'. With `__-I.__' you
		1149	can specify searching the directory which was current when
		1150	the compiler was invoked. That is not exactly the same as
		1151	what the preprocessor does by default, but it is often
		1152	satisfactory.
		1153
		1154
		1155	`__-I-__' does not inhibit the use of the standard sys-
		1156	tem directories for header files. Thus, `__-I-__' and
		1157	`__-nostdinc__' are independent.
		1158
		1159
		1160	__-L__''dir''
		1161
		1162
		1163	Add directory ''dir'' to the list of directories to be
		1164	searched for `__-l__'.
		1165
		1166
		1167	__-B__''prefix''
		1168
		1169
		1170	This option specifies where to find the executa- bles,
		1171	libraries and data files of the compiler it-
		1172	self.
		1173
		1174
		1175	The compiler driver program runs one or more of the
		1176	subprograms `__cpp__', `__cc1__' (or, for C++,
		1177	`__cc1plus__'), `__as__' and `__ld__'. It tries
		1178	''prefix'' as a prefix for each program it tries to run,
		1179	both with and without
		1180	`''machine''__/__''version''__/__'.
		1181
		1182
		1183	For each subprogram to be run, the compiler driver first
		1184	tries the `__-B__' prefix, if any. If that name is not
		1185	found, or if `__-B__' was not specified, the driver tries
		1186	two standard prefixes, which are `__/usr/lib/gcc/__' and
		1187	`__/usr/local/lib/gcc-lib/__'. If neither of those
		1188	results in a file name that is found, the compiler driver
		1189	searches for the unmodi- fied program name, using the
		1190	directories specified in your `__PATH__' environment
		1191	variable.
		1192
		1193
		1194	The run-time support file `__libgcc.a__' is also searched
		1195	for using the `__-B__' prefix, if needed. If it is not
		1196	found there, the two standard prefixes above are tried, and
		1197	that is all. The file is left out of the link if it is not
		1198	found by those means. Most of the time, on most machines,
		1199	`__libgcc.a__' is not actually necessary.
		1200
		1201
		1202	You can get a similar result from the environment variable
		1203	__GCC_EXEC_PREFIX__; if it is defined, its value is used
		1204	as a prefix in the same way. If both the `__-B__' option
		1205	and the __GCC_EXEC_PREFIX__ variable are present, the
		1206	`__-B__' option is used first and the environment
		1207	variable value second.
		1208	!!WARNING OPTIONS
		1209
		1210
		1211	Warnings are diagnostic messages that report constructions
		1212	which are not inherently erroneous but which are risky or
		1213	suggest there may have been an error.
		1214
		1215
		1216	These options control the amount and kinds of warnings
		1217	produced by GNU CC:
		1218
		1219
		1220	__-fsyntax-only__
		1221
		1222
		1223	Check the code for syntax errors, but don't emit any
		1224	output.
		1225
		1226
		1227	__-w__
		1228
		1229
		1230	Inhibit all warning messages.
		1231
		1232
		1233	__-Wno-import__
		1234
		1235
		1236	Inhibit warning messages about the use of
		1237	__#import__.
		1238
		1239
		1240	__-pedantic__
		1241
		1242
		1243	Issue all the warnings demanded by strict ANSI standard C;
		1244	reject all programs that use forbidden
		1245	extensions.
		1246
		1247
		1248	Valid ANSI standard C programs should compile prop- erly
		1249	with or without this option (though a rare few will require
		1250	`__-ansi__'). However, without this op- tion, certain GNU
		1251	extensions and traditional C fea- tures are supported as
		1252	well. With this option, they are rejected. There is no
		1253	reason to ''use'' this option; it exists only to satisfy
		1254	pedants.
		1255
		1256
		1257	`__-pedantic__' does not cause warning messages for use
		1258	of the alternate keywords whose names begin and end with
		1259	`______'. Pedantic warnings are also disabled in the
		1260	expression that follows ____extension____. Howev- er,
		1261	only system header files should use these es- cape routes;
		1262	application programs should avoid them.
		1263
		1264
		1265	__-pedantic-errors__
		1266
		1267
		1268	Like `__-pedantic__', except that errors are produced
		1269	rather than warnings.
		1270
		1271
		1272	__-W__
		1273
		1274
		1275	Print extra warning messages for these events:
		1276
		1277
		1278	A nonvolatile automatic variable might be changed by a call
		1279	to __longjmp__. These warnings are possible only in
		1280	optimizing compilation.
		1281
		1282
		1283	The compiler sees only the calls to __setjmp__. It cannot
		1284	know where __longjmp__ will be called; in fact, a signal
		1285	handler could call it at any point in the code. As a result,
		1286	you may get a warning even when there is in fact no problem
		1287	because __longjmp__ cannot in fact be called at the place
		1288	which would cause a problem.
		1289
		1290
		1291	A function can return either with or without a val- ue.
		1292	(Falling off the end of the function body is considered
		1293	returning without a value.) For exam- ple, this function
		1294	would evoke such a warning:
		1295
		1296
		1297	foo (a)
		1298	{
		1299	if (a
		1300	Spurious warnings can occur because GNU CC does not realize that certain functions (including __abort__ and __longjmp__) will never return.
		1301
		1302
		1303	An expression-statement or the left-hand side of a comma
		1304	expression contains no side effects. To sup- press the
		1305	warning, cast the unused expression to void. For example, an
		1306	expression such as `__x[[i,j]__' will cause a warning, but
		1307	`__x[[(void)i,j]__' will not.
		1308
		1309
		1310	An unsigned value is compared against zero with
		1311	`____' or `____'.
		1312
		1313
		1314	__-Wimplicit-int__
		1315
		1316
		1317	Warn whenever a declaration does not specify a
		1318	type.
		1319
		1320
		1321	__-Wimplicit-function-declaration__
		1322
		1323
		1324	Warn whenever a function is used before being de-
		1325	clared.
		1326
		1327
		1328	__-Wimplicit__
		1329
		1330
		1331	Same as -Wimplicit-int and -Wimplicit-function-dec-
		1332	laration.
		1333
		1334
		1335	__-Wmain__
		1336
		1337
		1338	Warn if the __main__ function is declared or defined with
		1339	a suspicious type. Typically, it is a func- tion with
		1340	external linkage, returning __int__, and tak- ing zero or
		1341	two arguments.
		1342
		1343
		1344	__-Wreturn-type__
		1345
		1346
		1347	Warn whenever a function is defined with a re- turn-type
		1348	that defaults to __int__. Also warn about any
		1349	__return__ statement with no return-value in a function
		1350	whose return-type is not __void__.
		1351
		1352
		1353	__-Wunused__
		1354
		1355
		1356	Warn whenever a local variable is unused aside from its
		1357	declaration, whenever a function is declared static but
		1358	never defined, and whenever a statement computes a result
		1359	that is explicitly not used.
		1360
		1361
		1362	__-Wswitch__
		1363
		1364
		1365	Warn whenever a __switch__ statement has an index of
		1366	enumeral type and lacks a __case__ for one or more of the
		1367	named codes of that enumeration. (The presence of a
		1368	__default__ label prevents this warning.) __case__
		1369	labels outside the enumeration range also provoke warnings
		1370	when this option is used.
		1371
		1372
		1373	__-Wcomment__
		1374
		1375
		1376	Warn whenever a comment-start sequence `__/__' appears in
		1377	a comment.
		1378
		1379
		1380	__-Wtrigraphs__
		1381
		1382
		1383	Warn if any trigraphs are encountered (assuming they are
		1384	enabled).
		1385
		1386
		1387	__-Wformat__
		1388
		1389
		1390	Check calls to __printf__ and __scanf__, etc., to make
		1391	sure that the arguments supplied have types appropriate to
		1392	the format string specified.
		1393
		1394
		1395	__-Wchar-subscripts__
		1396
		1397
		1398	Warn if an array subscript has type __char__. This is a
		1399	common cause of error, as programmers often for- get that
		1400	this type is signed on some machines.
		1401
		1402
		1403	__-Wuninitialized__
		1404
		1405
		1406	An automatic variable is used without first being
		1407	initialized.
		1408
		1409
		1410	These warnings are possible only in optimizing com-
		1411	pilation, because they require data flow informa- tion that
		1412	is computed only when optimizing. If you don't specify
		1413	`__-O__', you simply won't get these
		1414	warnings.
		1415
		1416
		1417	These warnings occur only for variables that are candidates
		1418	for register allocation. Therefore, they do not occur for a
		1419	variable that is declared __volatile__, or whose address
		1420	is taken, or whose size is other than 1, 2, 4 or 8 bytes.
		1421	Also, they do not occur for structures, unions or arrays,
		1422	even when they are in registers.
		1423
		1424
		1425	Note that there may be no warning about a variable that is
		1426	used only to compute a value that itself is never used,
		1427	because such computations may be delet- ed by data flow
		1428	analysis before the warnings are printed.
		1429
		1430
		1431	These warnings are made optional because GNU CC is not smart
		1432	enough to see all the reasons why the code might be correct
		1433	despite appearing to have an error. Here is one example of
		1434	how this can happen:
		1435
		1436
		1437	{
		1438	int x;
		1439	switch (y)
		1440	{
		1441	case 1: x = 1;
		1442	break;
		1443	case 2: x = 4;
		1444	break;
		1445	case 3: x = 5;
		1446	}
		1447	foo (x);
		1448	}
		1449	If the value of __y__ is always 1, 2 or 3, then __x__ is always initialized, but GNU CC doesn't know this. Here is another common case:
		1450
		1451
		1452	{
		1453	int save_y;
		1454	if (change_y) save_y = y, y = new_y;
		1455	...
		1456	if (change_y) y = save_y;
		1457	}
		1458	This has no bug because __save_y__ is used only if it is set.
		1459
		1460
		1461	Some spurious warnings can be avoided if you de- clare as
		1462	__volatile__ all the functions you use that never
		1463	return.
		1464
		1465
		1466	__-Wparentheses__
		1467
		1468
		1469	Warn if parentheses are omitted in certain con-
		1470	texts.
		1471
		1472
		1473	__-Wtemplate-debugging__
		1474
		1475
		1476	When using templates in a C++ program, warn if de- bugging
		1477	is not yet fully available (C++ only).
		1478
		1479
		1480	__-Wall__
		1481
		1482
		1483	All of the above `__-W__' options combined. These are all
		1484	the options which pertain to usage that we rec- ommend
		1485	avoiding and that we believe is easy to avoid, even in
		1486	conjunction with macros.
		1487
		1488
		1489	The remaining `__-W...__' options are not implied by
		1490	`__-Wall__' because they warn about constructions that we
		1491	consider reasonable to use, on occasion, in clean
		1492	programs.
		1493
		1494
		1495	__-Wtraditional__
		1496
		1497
		1498	Warn about certain constructs that behave differ- ently in
		1499	traditional and ANSI C.
		1500
		1501
		1502	Macro arguments occurring within string constants in the
		1503	macro body. These would substitute the ar- gument in
		1504	traditional C, but are part of the con- stant in ANSI
		1505	C.
		1506
		1507
		1508	A function declared external in one block and then used
		1509	after the end of the block.
		1510
		1511
		1512	A __switch__ statement has an operand of type
		1513	__long__.
		1514
		1515
		1516	__-Wshadow__
		1517
		1518
		1519	Warn whenever a local variable shadows another lo- cal
		1520	variable.
		1521
		1522
		1523	__-Wid-clash-__''len''
		1524
		1525
		1526	Warn whenever two distinct identifiers match in the first
		1527	''len'' characters. This may help you prepare a program
		1528	that will compile with certain obsolete, brain-damaged
		1529	compilers.
		1530
		1531
		1532	__-Wpointer-arith__
		1533
		1534
		1535	Warn about anything that depends on the ``size of'' a
		1536	function type or of __void__. GNU C assigns these types a
		1537	size of 1, for convenience in calculations with __void__
		1538	pointers and pointers to functions.
		1539
		1540
		1541	__-Wcast-qual__
		1542
		1543
		1544	Warn whenever a pointer is cast so as to remove a type
		1545	qualifier from the target type. For example, warn if a
		1546	__const char__ is cast to an ordinary __char__
		1547	.
		1548
		1549
		1550	__-Wcast-align__
		1551
		1552
		1553	Warn whenever a pointer is cast such that the re- quired
		1554	alignment of the target is increased. For example, warn if a
		1555	__char__ is cast to an __int__ on ma- chines where
		1556	integers can only be accessed at two- or four-byte
		1557	boundaries.
		1558
		1559
		1560	__-Wwrite-strings__
		1561
		1562
		1563	Give string constants the type __const
		1564	char[[__''length''__]__ so that copying the address
		1565	of one into a non-__const char__ pointer will get a
		1566	warning. These warnings will help you find at compile time
		1567	code that can try to write into a string constant, but only
		1568	if you have been very careful about using __const__ in
		1569	declarations and prototypes. Otherwise, it will just be a
		1570	nuisance; this is why we did not make `__-Wall__' request
		1571	these warnings.
		1572
		1573
		1574	__-Wconversion__
		1575
		1576
		1577	Warn if a prototype causes a type conversion that is
		1578	different from what would happen to the same ar- gument in
		1579	the absence of a prototype. This in- cludes conversions of
		1580	fixed point to floating and vice versa, and conversions
		1581	changing the width or signedness of a fixed point argument
		1582	except when the same as the default promotion.
		1583
		1584
		1585	__-Waggregate-return__
		1586
		1587
		1588	Warn if any functions that return structures or unions are
		1589	defined or called. (In languages where you can return an
		1590	array, this also elicits a warn- ing.)
		1591
		1592
		1593	__-Wstrict-prototypes__
		1594
		1595
		1596	Warn if a function is declared or defined without specifying
		1597	the argument types. (An old-style func- tion definition is
		1598	permitted without a warning if preceded by a declaration
		1599	which specifies the argu- ment types.)
		1600
		1601
		1602	__-Wmissing-prototypes__
		1603
		1604
		1605	Warn if a global function is defined without a pre- vious
		1606	prototype declaration. This warning is is- sued even if the
		1607	definition itself provides a pro- totype. The aim is to
		1608	detect global functions that fail to be declared in header
		1609	files.
		1610
		1611
		1612	__-Wmissing-declarations__
		1613
		1614
		1615	Warn if a global function is defined without a pre- vious
		1616	declaration. Do so even if the definition itself provides a
		1617	prototype. Use this option to detect global functions that
		1618	are not declared in header files.
		1619
		1620
		1621	__-Wredundant-decls__
		1622
		1623
		1624	Warn if anything is declared more than once in the same
		1625	scope, even in cases where multiple declara- tion is valid
		1626	and changes nothing.
		1627
		1628
		1629	__-Wnested-externs__
		1630
		1631
		1632	Warn if an __extern__ declaration is encountered within
		1633	an function.
		1634
		1635
		1636	__-Wenum-clash__
		1637
		1638
		1639	Warn about conversion between different enumeration types
		1640	(C++ only).
		1641
		1642
		1643	__-Wlong-long__
		1644
		1645
		1646	Warn if __long long__ type is used. This is default. To
		1647	inhibit the warning messages, use flag
		1648	`__-Wno-long-long__'. Flags `__-W-long-long__' and
		1649	`__-Wno-long-long__' are taken into account only when
		1650	flag `__-pedantic__' is used.
		1651
		1652
		1653	__-Woverloaded-virtual__
		1654
		1655
		1656	(C++ only.) In a derived class, the definitions of virtual
		1657	functions must match the type signature of a virtual
		1658	function declared in the base class. Use this option to
		1659	request warnings when a derived class declares a function
		1660	that may be an erroneous attempt to define a virtual
		1661	function: that is, warn when a function with the same name
		1662	as a virtual function in the base class, but with a type
		1663	signa- ture that doesn't match any virtual functions from
		1664	the base class.
		1665
		1666
		1667	__-Winline__
		1668
		1669
		1670	Warn if a function can not be inlined, and either it was
		1671	declared as inline, or else the __-fin- line-functions__
		1672	option was given.
		1673
		1674
		1675	__-Werror__
		1676
		1677
		1678	Treat warnings as errors; abort compilation after any
		1679	warning.
		1680	!!DEBUGGING OPTIONS
		1681
		1682
		1683	GNU CC has various special options that are used for de-
		1684	bugging either your program or GCC:
		1685
		1686
		1687	__-g__
		1688
		1689
		1690	Produce debugging information in the operating sys- tem's
		1691	native format (stabs, COFF, XCOFF, or DWARF). GDB can work
		1692	with this debugging information.
		1693
		1694
		1695	On most systems that use stabs format, `__-g__' enables
		1696	use of extra debugging information that only GDB can use;
		1697	this extra information makes debugging work better in GDB
		1698	but will probably make other de- buggers crash or refuse to
		1699	read the program. If you want to control for certain whether
		1700	to generate the extra information, use `__-gstabs+__',
		1701	`__-gstabs__', `__-gxcoff+__', `__-gxcoff__',
		1702	`__-gdwarf+__', or `__-gdwarf__' (see
		1703	below).
		1704
		1705
		1706	Unlike most other C compilers, GNU CC allows you to use
		1707	`__-g__' with `__-O__'. The shortcuts taken by opti-
		1708	mized code may occasionally produce surprising re- sults:
		1709	some variables you declared may not exist at all; flow of
		1710	control may briefly move where you did not expect it; some
		1711	statements may not be executed because they compute constant
		1712	results or their val- ues were already at hand; some
		1713	statements may exe- cute in different places because they
		1714	were moved out of loops.
		1715
		1716
		1717	Nevertheless it proves possible to debug optimized output.
		1718	This makes it reasonable to use the opti- mizer for programs
		1719	that might have bugs.
		1720
		1721
		1722	The following options are useful when GNU CC is generated
		1723	with the capability for more than one debugging
		1724	format.
		1725
		1726
		1727	__-ggdb__
		1728
		1729
		1730	Produce debugging information in the native format (if that
		1731	is supported), including GDB extensions if at all
		1732	possible.
		1733
		1734
		1735	__-gstabs__
		1736
		1737
		1738	Produce debugging information in stabs format (if that is
		1739	supported), without GDB extensions. This is the format used
		1740	by DBX on most BSD systems.
		1741
		1742
		1743	__-gstabs+__
		1744
		1745
		1746	Produce debugging information in stabs format (if that is
		1747	supported), using GNU extensions understood only by the GNU
		1748	debugger (GDB). The use of these extensions is likely to
		1749	make other debuggers crash or refuse to read the
		1750	program.
		1751
		1752
		1753	__-gcoff__
		1754
		1755
		1756	Produce debugging information in COFF format (if that is
		1757	supported). This is the format used by SDB on most System V
		1758	systems prior to System V Release 4.
		1759
		1760
		1761	__-gxcoff__
		1762
		1763
		1764	Produce debugging information in XCOFF format (if that is
		1765	supported). This is the format used by the DBX debugger on
		1766	IBM RS/6000 systems.
		1767
		1768
		1769	__-gxcoff+__
		1770
		1771
		1772	Produce debugging information in XCOFF format (if that is
		1773	supported), using GNU extensions understood only by the GNU
		1774	debugger (GDB). The use of these extensions is likely to
		1775	make other debuggers crash or refuse to read the
		1776	program.
		1777
		1778
		1779	__-gdwarf__
		1780
		1781
		1782	Produce debugging information in DWARF format (if that is
		1783	supported). This is the format used by SDB on most System V
		1784	Release 4 systems.
		1785
		1786
		1787	__-gdwarf+__
		1788
		1789
		1790	Produce debugging information in DWARF format (if that is
		1791	supported), using GNU extensions understood only by the GNU
		1792	debugger (GDB). The use of these extensions is likely to
		1793	make other debuggers crash or refuse to read the
		1794	program.
		1795
		1796
		1797	__-g__''level''__
		1798	-ggdb__''level''__
		1799	-gstabs__''level''__
		1800	-gcoff__''level''
		1801	__-gxcoff__''level''
		1802
		1803
		1804	__-gdwarf__''level''
		1805
		1806
		1807	Request debugging information and also use ''level'' to
		1808	specify how much information. The default level is
		1809	2.
		1810
		1811
		1812	Level 1 produces minimal information, enough for making
		1813	backtraces in parts of the program that you don't plan to
		1814	debug. This includes descriptions of functions and external
		1815	variables, but no informa- tion about local variables and no
		1816	line numbers.
		1817
		1818
		1819	Level 3 includes extra information, such as all the macro
		1820	definitions present in the program. Some de- buggers support
		1821	macro expansion when you use `__-g3__'.
		1822
		1823
		1824	__-p__
		1825
		1826
		1827	Generate extra code to write profile information suitable
		1828	for the analysis program __prof__.
		1829
		1830
		1831	__-pg__
		1832
		1833
		1834	Generate extra code to write profile information suitable
		1835	for the analysis program __gprof__.
		1836
		1837
		1838	__-a__
		1839
		1840
		1841	Generate extra code to write profile information for basic
		1842	blocks, which will record the number of times each basic
		1843	block is executed. This data could be analyzed by a program
		1844	like __tcov__. Note, however, that the format of the data
		1845	is not what __tcov__ expects. Eventually GNU __gprof__
		1846	should be ex- tended to process this data.
		1847
		1848
		1849	__-ax__
		1850
		1851
		1852	Generate extra code to read basic block profiling parameters
		1853	from file `bb.in' and write profiling results to file
		1854	`bb.out'. `bb.in' contains a list of functions. Whenever a
		1855	function on the list is entered, profiling is turned on.
		1856	When the outmost function is left, profiling is turned off.
		1857	If a function name is prefixed with `-' the function is
		1858	excluded from profiling. If a function name is not unique it
		1859	can be disambiguated by writing
		1860	`/path/filename.d:functionname'. `bb.out' will list some
		1861	available filenames. Four function names have a special
		1862	meaning: `__bb_jumps__' will cause jump frequencies to be
		1863	written to `bb.out'. `__bb_trace__' will cause the sequence
		1864	of basic blocks to be piped into `gzip' and written to file
		1865	`bbtrace.gz'. `__bb_hidecall__' will cause call instructions
		1866	to be excluded from the trace. `__bb_showret__' will cause
		1867	return instructions to be included in the
		1868	trace.
		1869
		1870
		1871	__-d__''letters''
		1872
		1873
		1874	Says to make debugging dumps during compilation at times
		1875	specified by ''letters''. This is used for de- bugging
		1876	the compiler. The file names for most of the dumps are made
		1877	by appending a word to the source file name (e.g.
		1878	`__foo.c.rtl__' or `__foo.c.jump__').
		1879
		1880
		1881	__-dM__
		1882
		1883
		1884	Dump all macro definitions, at the end of prepro- cessing,
		1885	and write no output.
		1886
		1887
		1888	__-dN__
		1889
		1890
		1891	Dump all macro names, at the end of
		1892	preprocessing.
		1893
		1894
		1895	__-dD__
		1896
		1897
		1898	Dump all macro definitions, at the end of prepro- cessing,
		1899	in addition to normal output.
		1900
		1901
		1902	__-dy__
		1903
		1904
		1905	Dump debugging information during parsing, to stan- dard
		1906	error.
		1907
		1908
		1909	__-dr__
		1910
		1911
		1912	Dump after RTL generation, to
		1913	`''file''__.rtl__'.
		1914
		1915
		1916	__-dx__
		1917
		1918
		1919	Just generate RTL for a function instead of compil- ing it.
		1920	Usually used with `__r__'.
		1921
		1922
		1923	__-dj__
		1924
		1925
		1926	Dump after first jump optimization, to
		1927	`''file''__.jump__'.
		1928
		1929
		1930	__-ds__
		1931
		1932
		1933	Dump after CSE (including the jump optimization that
		1934	sometimes follows CSE), to
		1935	`''file''__.cse__'.
		1936
		1937
		1938	__-dL__
		1939
		1940
		1941	Dump after loop optimization, to
		1942	`''file''__.loop__'.
		1943
		1944
		1945	__-dt__
		1946
		1947
		1948	Dump after the second CSE pass (including the jump
		1949	optimization that sometimes follows CSE), to
		1950	`''file''__.cse2__'.
		1951
		1952
		1953	__-df__
		1954
		1955
		1956	Dump after flow analysis, to
		1957	`''file''__.flow__'.
		1958
		1959
		1960	__-dc__
		1961
		1962
		1963	Dump after instruction combination, to `''file''__.com-
		1964	bine__'.
		1965
		1966
		1967	__-dS__
		1968
		1969
		1970	Dump after the first instruction scheduling pass, to
		1971	`''file''__.sched__'.
		1972
		1973
		1974	__-dl__
		1975
		1976
		1977	Dump after local register allocation, to
		1978	`''file''__.lreg__'.
		1979
		1980
		1981	__-dg__
		1982
		1983
		1984	Dump after global register allocation, to
		1985	`''file''__.greg__'.
		1986
		1987
		1988	__-dR__
		1989
		1990
		1991	Dump after the second instruction scheduling pass, to
		1992	`''file''__.sched2__'.
		1993
		1994
		1995	__-dJ__
		1996
		1997
		1998	Dump after last jump optimization, to
		1999	`''file''__.jump2__'.
		2000
		2001
		2002	__-dd__
		2003
		2004
		2005	Dump after delayed branch scheduling, to
		2006	`''file''__.dbr__'.
		2007
		2008
		2009	__-dk__
		2010
		2011
		2012	Dump after conversion from registers to stack, to
		2013	`''file''__.stack__'.
		2014
		2015
		2016	__-da__
		2017
		2018
		2019	Produce all the dumps listed above.
		2020
		2021
		2022	__-dm__
		2023
		2024
		2025	Print statistics on memory usage, at the end of the run, to
		2026	standard error.
		2027
		2028
		2029	__-dp__
		2030
		2031
		2032	Annotate the assembler output with a comment indi- cating
		2033	which pattern and alternative was used.
		2034
		2035
		2036	__-fpretend-float__
		2037
		2038
		2039	When running a cross-compiler, pretend that the target
		2040	machine uses the same floating point format as the host
		2041	machine. This causes incorrect output of the actual floating
		2042	constants, but the actual instruction sequence will probably
		2043	be the same as GNU CC would make when running on the target
		2044	ma- chine.
		2045
		2046
		2047	__-save-temps__
		2048
		2049
		2050	Store the usual ``temporary'' intermediate files
		2051	permanently; place them in the current directory and name
		2052	them based on the source file. Thus, com- piling
		2053	`__foo.c__' with `__-c -save-temps__' would produce
		2054	files `__foo.cpp__' and `__foo.s__', as well as
		2055	`__foo.o__'.
		2056
		2057
		2058	__-print-file-name=__''library''
		2059
		2060
		2061	Print the full absolute name of the library file
		2062	''library'' that would be used when linking--and do not
		2063	do anything else. With this option, GNU CC does not compile
		2064	or link anything; it just prints the file name.
		2065
		2066
		2067	__-print-libgcc-file-name__
		2068
		2069
		2070	Same as `__-print-file-name=libgcc.a__'.
		2071
		2072
		2073	__-print-prog-name=__''program''
		2074
		2075
		2076	Like `__-print-file-name__', but searches for a program
		2077	such as `cpp'.
		2078	!!OPTIMIZATION OPTIONS
		2079
		2080
		2081	These options control various sorts of
		2082	optimizations:
		2083
		2084
		2085	__-O__
		2086
		2087
		2088	__-O1__
		2089
		2090
		2091	Optimize. Optimizing compilation takes somewhat more time,
		2092	and a lot more memory for a large func- tion.
		2093
		2094
		2095	Without `__-O__', the compiler's goal is to reduce the
		2096	cost of compilation and to make debugging produce the
		2097	expected results. Statements are independent: if you stop
		2098	the program with a breakpoint between statements, you can
		2099	then assign a new value to any variable or change the
		2100	program counter to any other statement in the function and
		2101	get exactly the re- sults you would expect from the source
		2102	code.
		2103
		2104
		2105	Without `__-O__', only variables declared __register__
		2106	are allocated in registers. The resulting compiled code is a
		2107	little worse than produced by PCC without
		2108	`__-O__'.
		2109
		2110
		2111	With `__-O__', the compiler tries to reduce code size and
		2112	execution time.
		2113
		2114
		2115	When you specify `__-O__', the two options
		2116	`__-fthread-jumps__' and `__-fdefer-pop__' are turned
		2117	on. On machines that have delay slots, the `__-fde-
		2118	layed-branch__' option is turned on. For those ma- chines
		2119	that can support debugging even without a frame pointer, the
		2120	`__-fomit-frame-pointer__' option is turned on. On some
		2121	machines other flags may also be turned on.
		2122
		2123
		2124	__-O2__
		2125
		2126
		2127	Optimize even more. Nearly all supported optimiza- tions
		2128	that do not involve a space-speed tradeoff are performed.
		2129	Loop unrolling and function inlin- ing are not done, for
		2130	example. As compared to __-O__, this option increases
		2131	both compilation time and the performance of the generated
		2132	code.
		2133
		2134
		2135	__-O3__
		2136
		2137
		2138	Optimize yet more. This turns on everything __-O2__ does,
		2139	along with also turning on __-finline-func-
		2140	tions.__
		2141
		2142
		2143	__-O0__
		2144
		2145
		2146	Do not optimize.
		2147
		2148
		2149	If you use multiple __-O__ options, with or without level
		2150	numbers, the last such option is the one that is
		2151	effective.
		2152
		2153
		2154	Options of the form `__-f__''flag''' specify
		2155	machine-independent flags. Most flags have both positive and
		2156	negative forms; the negative form of `__-ffoo__' would be
		2157	`__-fno-foo__'. The following list shows only one
		2158	form--the one which is not the default. You can figure out
		2159	the other form by either removing `__no-__' or adding
		2160	it.
		2161
		2162
		2163	__-ffloat-store__
		2164
		2165
		2166	Do not store floating point variables in registers. This
		2167	prevents undesirable excess precision on ma- chines such as
		2168	the 68000 where the floating regis- ters (of the 68881) keep
		2169	more precision than a __dou- ble__ is supposed to
		2170	have.
		2171
		2172
		2173	For most programs, the excess precision does only good, but
		2174	a few programs rely on the precise defi- nition of IEEE
		2175	floating point. Use `__-ffloat-store__' for such
		2176	programs.
		2177
		2178
		2179	__-fmemoize-lookups__
		2180
		2181
		2182	__-fsave-memoized__
		2183
		2184
		2185	Use heuristics to compile faster (C++ only). These
		2186	heuristics are not enabled by default, since they are only
		2187	effective for certain input files. Other input files compile
		2188	more slowly.
		2189
		2190
		2191	The first time the compiler must build a call to a member
		2192	function (or reference to a data member), it must (1)
		2193	determine whether the class implements member functions of
		2194	that name; (2) resolve which member function to call (which
		2195	involves figuring out what sorts of type conversions need to
		2196	be made); and (3) check the visibility of the member
		2197	function to the caller. All of this adds up to slower
		2198	compilation. Normally, the second time a call is made to
		2199	that member function (or reference to that data member), it
		2200	must go through the same lengthy process again. This means
		2201	that code like this
		2202
		2203
		2204	cout
		2205
		2206
		2207	makes six passes through all three steps. By using a
		2208	software cache, a ``hit'' significantly reduces this cost.
		2209	Unfortunately, using the cache intro- duces another layer of
		2210	mechanisms which must be im- plemented, and so incurs its
		2211	own overhead. `__-fmem- oize-lookups__' enables the
		2212	software cache.
		2213
		2214
		2215	Because access privileges (visibility) to members and member
		2216	functions may differ from one function context to the next,
		2217	__g++__ may need to flush the cache. With the
		2218	`__-fmemoize-lookups__' flag, the cache is flushed after
		2219	every function that is com- piled. The `-fsave-memoized'
		2220	flag enables the same software cache, but when the compiler
		2221	determines that the context of the last function compiled
		2222	would yield the same access privileges of the next function
		2223	to compile, it preserves the cache. This is most helpful
		2224	when defining many member functions for the same class: with
		2225	the exception of member functions which are friends of other
		2226	classes, each member function has exactly the same access
		2227	privi- leges as every other, and the cache need not be
		2228	flushed.
		2229
		2230
		2231	__-fno-default-inline__
		2232
		2233
		2234	Don't make member functions inline by default mere- ly
		2235	because they are defined inside the class scope (C++
		2236	only).
		2237
		2238
		2239	__-fno-defer-pop__
		2240
		2241
		2242	Always pop the arguments to each function call as soon as
		2243	that function returns. For machines which must pop arguments
		2244	after a function call, the com- piler normally lets
		2245	arguments accumulate on the stack for several function calls
		2246	and pops them all at once.
		2247
		2248
		2249	__-fforce-mem__
		2250
		2251
		2252	Force memory operands to be copied into registers before
		2253	doing arithmetic on them. This may produce better code by
		2254	making all memory references poten- tial common
		2255	subexpressions. When they are not com- mon subexpressions,
		2256	instruction combination should eliminate the separate
		2257	register-load. I am inter- ested in hearing about the
		2258	difference this makes.
		2259
		2260
		2261	__-fforce-addr__
		2262
		2263
		2264	Force memory address constants to be copied into registers
		2265	before doing arithmetic on them. This may produce better
		2266	code just as `__-fforce-mem__' may. I am interested in
		2267	hearing about the difference this makes.
		2268
		2269
		2270	__-fomit-frame-pointer__
		2271
		2272
		2273	Don't keep the frame pointer in a register for functions
		2274	that don't need one. This avoids the in- structions to save,
		2275	set up and restore frame point- ers; it also makes an extra
		2276	register available in many functions. ''It also makes
		2277	debugging impossible on most machines''.
		2278
		2279
		2280	On some machines, such as the Vax, this flag has no effect,
		2281	because the standard calling sequence auto- matically
		2282	handles the frame pointer and nothing is saved by pretending
		2283	it doesn't exist. The ma- chine-description macro
		2284	__FRAME_POINTER_REQUIRED__ con- trols whether a target
		2285	machine supports this flag.
		2286
		2287
		2288	__-finline-functions__
		2289
		2290
		2291	Integrate all simple functions into their callers. The
		2292	compiler heuristically decides which functions are simple
		2293	enough to be worth integrating in this way.
		2294
		2295
		2296	If all calls to a given function are integrated, and the
		2297	function is declared __static__, then GCC nor- mally does
		2298	not output the function as assembler code in its own
		2299	right.
		2300
		2301
		2302	__-fcaller-saves__
		2303
		2304
		2305	Enable values to be allocated in registers that will be
		2306	clobbered by function calls, by emitting extra instructions
		2307	to save and restore the regis- ters around such calls. Such
		2308	allocation is done only when it seems to result in better
		2309	code than would otherwise be produced.
		2310
		2311
		2312	This option is enabled by default on certain ma- chines,
		2313	usually those which have no call-preserved registers to use
		2314	instead.
		2315
		2316
		2317	__-fkeep-inline-functions__
		2318
		2319
		2320	Even if all calls to a given function are integrat- ed, and
		2321	the function is declared __static__, neverthe- less
		2322	output a separate run-time callable version of the
		2323	function.
		2324
		2325
		2326	__-fno-function-cse__
		2327
		2328
		2329	Do not put function addresses in registers; make each
		2330	instruction that calls a constant function contain the
		2331	function's address explicitly.
		2332
		2333
		2334	This option results in less efficient code, but some strange
		2335	hacks that alter the assembler output may be confused by the
		2336	optimizations performed when this option is not
		2337	used.
		2338
		2339
		2340	__-fno-peephole__
		2341
		2342
		2343	Disable any machine-specific peephole optimiza-
		2344	tions.
		2345
		2346
		2347	__-ffast-math__
		2348
		2349
		2350	This option allows GCC to violate some ANSI or IEEE
		2351	rules/specifications in the interest of optimizing code for
		2352	speed. For example, it allows the compil- er to assume
		2353	arguments to the __sqrt__ function are non-negative
		2354	numbers.
		2355
		2356
		2357	This option should never be turned on by any `__-O__'
		2358	option since it can result in incorrect output for programs
		2359	which depend on an exact implementation of IEEE or ANSI
		2360	rules/specifications for math func- tions.
		2361
		2362
		2363	The following options control specific optimizations. The
		2364	`__-O2__' option turns on all of these optimizations
		2365	except `__-funroll-loops__' and
		2366	`__-funroll-all-loops__'.
		2367
		2368
		2369	The `__-O__' option usually turns on the
		2370	`__-fthread-jumps__' and `__-fdelayed-branch__'
		2371	options, but specific machines may change the default
		2372	optimizations.
		2373
		2374
		2375	You can use the following flags in the rare cases when
		2376	``fine-tuning'' of optimizations to be performed is de-
		2377	sired.
		2378
		2379
		2380	__-fstrength-reduce__
		2381
		2382
		2383	Perform the optimizations of loop strength reduc- tion and
		2384	elimination of iteration variables.
		2385
		2386
		2387	__-fthread-jumps__
		2388
		2389
		2390	Perform optimizations where we check to see if a jump
		2391	branches to a location where another compari- son subsumed
		2392	by the first is found. If so, the first branch is redirected
		2393	to either the destina- tion of the second branch or a point
		2394	immediately following it, depending on whether the condition
		2395	is known to be true or false.
		2396
		2397
		2398	__-funroll-loops__
		2399
		2400
		2401	Perform the optimization of loop unrolling. This is only
		2402	done for loops whose number of iterations can be determined
		2403	at compile time or run time.
		2404
		2405
		2406	__-funroll-all-loops__
		2407
		2408
		2409	Perform the optimization of loop unrolling. This is done for
		2410	all loops. This usually makes programs run more
		2411	slowly.
		2412
		2413
		2414	__-fcse-follow-jumps__
		2415
		2416
		2417	In common subexpression elimination, scan through jump
		2418	instructions when the target of the jump is not reached by
		2419	any other path. For example, when CSE encounters an
		2420	__if__ statement with an __else__ clause, CSE will
		2421	follow the jump when the condition tested is
		2422	false.
		2423
		2424
		2425	__-fcse-skip-blocks__
		2426
		2427
		2428	This is similar to `__-fcse-follow-jumps__', but causes
		2429	CSE to follow jumps which conditionally skip over blocks.
		2430	When CSE encounters a simple __if__ statement with no
		2431	else clause, `__-fcse-skip-blocks__' causes CSE to follow
		2432	the jump around the body of the __if__.
		2433
		2434
		2435	__-frerun-cse-after-loop__
		2436
		2437
		2438	Re-run common subexpression elimination after loop
		2439	optimizations has been performed.
		2440
		2441
		2442	__-felide-constructors__
		2443
		2444
		2445	Elide constructors when this seems plausible (C++ only).
		2446	With this flag, GNU C++ initializes __y__ di- rectly from
		2447	the call to __foo__ without going through a temporary in
		2448	the following code:
		2449
		2450
		2451	A foo (); A y = foo ();
		2452
		2453
		2454	Without this option, GNU C++ first initializes __y__ by
		2455	calling the appropriate constructor for type __A__; then
		2456	assigns the result of __foo__ to a temporary; and,
		2457	finally, replaces the initial value of `__y__' with the
		2458	temporary.
		2459
		2460
		2461	The default behavior (`__-fno-elide-constructors__') is
		2462	specified by the draft ANSI C++ standard. If your program's
		2463	constructors have side effects, using
		2464	`__-felide-constructors__' can make your program act
		2465	differently, since some constructor calls may be
		2466	omitted.
		2467
		2468
		2469	__-fexpensive-optimizations__
		2470
		2471
		2472	Perform a number of minor optimizations that are relatively
		2473	expensive.
		2474
		2475
		2476	__-fdelayed-branch__
		2477
		2478
		2479	If supported for the target machine, attempt to re- order
		2480	instructions to exploit instruction slots available after
		2481	delayed branch instructions.
		2482
		2483
		2484	__-fschedule-insns__
		2485
		2486
		2487	If supported for the target machine, attempt to re- order
		2488	instructions to eliminate execution stalls due to required
		2489	data being unavailable. This helps machines that have slow
		2490	floating point or memory load instructions by allowing other
		2491	instructions to be issued until the result of the load or
		2492	floating point instruction is required.
		2493
		2494
		2495	__-fschedule-insns2__
		2496
		2497
		2498	Similar to `__-fschedule-insns__', but requests an ad-
		2499	ditional pass of instruction scheduling after reg- ister
		2500	allocation has been done. This is especially useful on
		2501	machines with a relatively small number of registers and
		2502	where memory load instructions take more than one
		2503	cycle.
		2504	!!TARGET OPTIONS
		2505
		2506
		2507	By default, GNU CC compiles code for the same type of ma-
		2508	chine that you are using. However, it can also be in-
		2509	stalled as a cross-compiler, to compile for some other type
		2510	of machine. In fact, several different configura- tions of
		2511	GNU CC, for different target machines, can be in- stalled
		2512	side by side. Then you specify which one to use with the
		2513	`__-b__' option.
		2514
		2515
		2516	In addition, older and newer versions of GNU CC can be in-
		2517	stalled side by side. One of them (probably the newest) will
		2518	be the default, but you may sometimes wish to use an-
		2519	other.
		2520
		2521
		2522	__-b__ ''machine''
		2523
		2524
		2525	The argument ''machine'' specifies the target machine for
		2526	compilation. This is useful when you have in- stalled GNU CC
		2527	as a cross-compiler.
		2528
		2529
		2530	The value to use for ''machine'' is the same as was
		2531	specified as the machine type when configuring GNU CC as a
		2532	cross-compiler. For example, if a cross-compiler was
		2533	configured with `__configure__ i386v', meaning to compile
		2534	for an 80386 running System V, then you would specify `__-b
		2535	i386v__' to run that cross compiler.
		2536
		2537
		2538	When you do not specify `__-b__', it normally means to
		2539	compile for the same type of machine that you are
		2540	using.
		2541
		2542
		2543	__-V__ ''version''
		2544
		2545
		2546	The argument ''version'' specifies which version of GNU
		2547	CC to run. This is useful when multiple versions are
		2548	installed. For example, ''version'' might be
		2549	`__2.0__', meaning to run GNU CC version
		2550	2.0.
		2551
		2552
		2553	The default version, when you do not specify `__-V__', is
		2554	controlled by the way GNU CC is installed. Nor- mally, it
		2555	will be a version that is recommended for general
		2556	use.
		2557	!!MACHINE DEPENDENT OPTIONS
		2558
		2559
		2560	Each of the target machine types can have its own special
		2561	options, starting with `__-m__', to choose among various
		2562	hard- ware models or configurations--for example, 68010 vs
		2563	68020, floating coprocessor or none. A single installed
		2564	version of the compiler can compile for any model or con-
		2565	figuration, according to the options specified.
		2566
		2567
		2568	Some configurations of the compiler also support addition-
		2569	al special options, usually for command-line compatibility
		2570	with other compilers on the same platform.
		2571
		2572
		2573	These are the `__-m__' options defined for the 68000
		2574	series:
		2575
		2576
		2577	__-m68000__
		2578
		2579
		2580	__-mc68000__
		2581
		2582
		2583	Generate output for a 68000. This is the default when the
		2584	compiler is configured for 68000-based systems.
		2585
		2586
		2587	__-m68020__
		2588
		2589
		2590	__-mc68020__
		2591
		2592
		2593	Generate output for a 68020 (rather than a 68000). This is
		2594	the default when the compiler is configured for 68020-based
		2595	systems.
		2596
		2597
		2598	__-m68881__
		2599
		2600
		2601	Generate output containing 68881 instructions for floating
		2602	point. This is the default for most 68020-based systems
		2603	unless __-nfp__ was specified when the compiler was
		2604	configured.
		2605
		2606
		2607	__-m68030__
		2608
		2609
		2610	Generate output for a 68030. This is the default when the
		2611	compiler is configured for 68030-based systems.
		2612
		2613
		2614	__-m68040__
		2615
		2616
		2617	Generate output for a 68040. This is the default when the
		2618	compiler is configured for 68040-based systems.
		2619
		2620
		2621	__-m68020-40__
		2622
		2623
		2624	Generate output for a 68040, without using any of the new
		2625	instructions. This results in code which can run relatively
		2626	efficiently on either a 68020/68881 or a 68030 or a
		2627	68040.
		2628
		2629
		2630	__-mfpa__
		2631
		2632
		2633	Generate output containing Sun FPA instructions for floating
		2634	point.
		2635
		2636
		2637	__-msoft-float__
		2638
		2639
		2640	Generate output containing library calls for float- ing
		2641	point. ''WARNING:'' the requisite libraries are not part
		2642	of GNU CC. Normally the facilities of the machine's usual C
		2643	compiler are used, but this can't be done directly in
		2644	cross-compilation. You must make your own arrangements to
		2645	provide suitable li- brary functions for
		2646	cross-compilation.
		2647
		2648
		2649	__-mshort__
		2650
		2651
		2652	Consider type __int__ to be 16 bits wide, like __short
		2653	int__.
		2654
		2655
		2656	__-mnobitfield__
		2657
		2658
		2659	Do not use the bit-field instructions. `__-m68000__'
		2660	implies `__-mnobitfield__'.
		2661
		2662
		2663	__-mbitfield__
		2664
		2665
		2666	Do use the bit-field instructions. `__-m68020__' im-
		2667	plies `__-mbitfield__'. This is the default if you use
		2668	the unmodified sources.
		2669
		2670
		2671	__-mrtd__
		2672
		2673
		2674	Use a different function-calling convention, in which
		2675	functions that take a fixed number of argu- ments return
		2676	with the __rtd__ instruction, which pops their arguments
		2677	while returning. This saves one instruction in the caller
		2678	since there is no need to pop the arguments
		2679	there.
		2680
		2681
		2682	This calling convention is incompatible with the one
		2683	normally used on Unix, so you cannot use it if you need to
		2684	call libraries compiled with the Unix compiler.
		2685
		2686
		2687	Also, you must provide function prototypes for all functions
		2688	that take variable numbers of arguments (including
		2689	__printf__); otherwise incorrect code will be generated
		2690	for calls to those functions.
		2691
		2692
		2693	In addition, seriously incorrect code will result if you
		2694	call a function with too many arguments. (Normally, extra
		2695	arguments are harmlessly ignored.)
		2696
		2697
		2698	The __rtd__ instruction is supported by the 68010 and
		2699	68020 processors, but not by the 68000.
		2700
		2701
		2702	These `__-m__' options are defined for the
		2703	Vax:
		2704
		2705
		2706	__-munix__
		2707
		2708
		2709	Do not output certain jump instructions (__aobleq__ and
		2710	so on) that the Unix assembler for the Vax cannot handle
		2711	across long ranges.
		2712
		2713
		2714	__-mgnu__
		2715
		2716
		2717	Do output those jump instructions, on the assump- tion that
		2718	you will assemble with the GNU assembler.
		2719
		2720
		2721	__-mg__
		2722
		2723
		2724	Output code for g-format floating point numbers in- stead of
		2725	d-format.
		2726
		2727
		2728	These `__-m__' switches are supported on the
		2729	SPARC:
		2730
		2731
		2732	__-mfpu__
		2733
		2734
		2735	__-mhard-float__
		2736
		2737
		2738	Generate output containing floating point instruc- tions.
		2739	This is the default.
		2740
		2741
		2742	__-mno-fpu__
		2743
		2744
		2745	__-msoft-float__
		2746
		2747
		2748	Generate output containing library calls for float- ing
		2749	point. ''Warning:'' there is no GNU floating-point
		2750	library for SPARC. Normally the facilities of the machine's
		2751	usual C compiler are used, but this can- not be done
		2752	directly in cross-compilation. You must make your own
		2753	arrangements to provide suitable library functions for
		2754	cross-compilation.
		2755
		2756
		2757	__-msoft-float__ changes the calling convention in the
		2758	output file; therefore, it is only useful if you compile
		2759	''all'' of a program with this option.
		2760
		2761
		2762	__-mno-epilogue__
		2763
		2764
		2765	__-mepilogue__
		2766
		2767
		2768	With __-mepilogue__ (the default), the compiler always
		2769	emits code for function exit at the end of each function.
		2770	Any function exit in the middle of the function (such as a
		2771	return statement in C) will generate a jump to the exit code
		2772	at the end of the function.
		2773
		2774
		2775	With __-mno-epilogue__, the compiler tries to emit exit
		2776	code inline at every function exit.
		2777
		2778
		2779	__-mno-v8__
		2780
		2781
		2782	__-mv8__
		2783
		2784
		2785	__-msparclite__
		2786
		2787
		2788	These three options select variations on the SPARC
		2789	architecture.
		2790
		2791
		2792	By default (unless specifically configured for the Fujitsu
		2793	SPARClite), GCC generates code for the v7 variant of the
		2794	SPARC architecture.
		2795
		2796
		2797	__-mv8__ will give you SPARC v8 code. The only differ-
		2798	ence from v7 code is that the compiler emits the integer
		2799	multiply and integer divide instructions which exist in
		2800	SPARC v8 but not in SPARC v7.
		2801
		2802
		2803	__-msparclite__ will give you SPARClite code. This adds
		2804	the integer multiply, integer divide step and scan (ffs)
		2805	instructions which exist in SPARClite but not in SPARC
		2806	v7.
		2807
		2808
		2809	__-mcypress__
		2810
		2811
		2812	__-msupersparc__
		2813
		2814
		2815	These two options select the processor for which the code is
		2816	optimised.
		2817
		2818
		2819	With __-mcypress__ (the default), the compiler optimis-
		2820	es code for the Cypress CY7C602 chip, as used in the
2	perry	2821	!SparcStation/!SparcServer 3xx series. This is also
		2822	appropriate for the older !SparcStation 1, 2, IPX
1	perry	2823	etc.
		2824
		2825
		2826	With __-msupersparc__ the compiler optimises code for the
2	perry	2827	!SuperSparc cpu, as used in the !SparcStation 10, 1000 and
1	perry	2828	2000 series. This flag also enables use of the full SPARC v8
		2829	instruction set.
		2830
		2831
		2832	These `__-m__' options are defined for the
		2833	Convex:
		2834
		2835
		2836	__-mc1__
		2837
		2838
		2839	Generate output for a C1. This is the default when the
		2840	compiler is configured for a C1.
		2841
		2842
		2843	__-mc2__
		2844
		2845
		2846	Generate output for a C2. This is the default when the
		2847	compiler is configured for a C2.
		2848
		2849
		2850	__-margcount__
		2851
		2852
		2853	Generate code which puts an argument count in the word
		2854	preceding each argument list. Some non- portable Convex and
		2855	Vax programs need this word. (Debuggers don't, except for
		2856	functions with vari- able-length argument lists; this info
		2857	is in the symbol table.)
		2858
		2859
		2860	__-mnoargcount__
		2861
		2862
		2863	Omit the argument count word. This is the default if you use
		2864	the unmodified sources.
		2865
		2866
		2867	These `__-m__' options are defined for the AMD
		2868	Am29000:
		2869
		2870
		2871	__-mdw__
		2872
		2873
		2874	Generate code that assumes the DW bit is set, i.e., that
		2875	byte and halfword operations are directly sup- ported by the
		2876	hardware. This is the default.
		2877
		2878
		2879	__-mnodw__
		2880
		2881
		2882	Generate code that assumes the DW bit is not
		2883	set.
		2884
		2885
		2886	__-mbw__
		2887
		2888
		2889	Generate code that assumes the system supports byte and
		2890	halfword write operations. This is the de-
		2891	fault.
		2892
		2893
		2894	__-mnbw__
		2895
		2896
		2897	Generate code that assumes the systems does not support byte
		2898	and halfword write operations. This implies
		2899	`__-mnodw__'.
		2900
		2901
		2902	__-msmall__
		2903
		2904
		2905	Use a small memory model that assumes that all function
		2906	addresses are either within a single 256 KB segment or at an
		2907	absolute address of less than 256K. This allows the
		2908	__call__ instruction to be used instead of a
		2909	__const__, __consth__, __calli__
		2910	sequence.
		2911
		2912
		2913	__-mlarge__
		2914
		2915
		2916	Do not assume that the __call__ instruction can be used;
		2917	this is the default.
		2918
		2919
		2920	__-m29050__
		2921
		2922
		2923	Generate code for the Am29050.
		2924
		2925
		2926	__-m29000__
		2927
		2928
		2929	Generate code for the Am29000. This is the de-
		2930	fault.
		2931
		2932
		2933	__-mkernel-registers__
		2934
		2935
		2936	Generate references to registers __gr64-gr95__ instead of
		2937	__gr96-gr127__. This option can be used when com- piling
		2938	kernel code that wants a set of global reg- isters disjoint
		2939	from that used by user-mode code.
		2940
		2941
		2942	Note that when this option is used, register names in
		2943	`__-f__' flags must use the normal, user-mode,
		2944	names.
		2945
		2946
		2947	__-muser-registers__
		2948
		2949
		2950	Use the normal set of global registers, __gr96-gr127__.
		2951	This is the default.
		2952
		2953
		2954	__-mstack-check__
		2955
		2956
		2957	Insert a call to ____msp_check__ after each stack ad-
		2958	justment. This is often used for kernel code.
		2959
		2960
		2961	These `__-m__' options are defined for Motorola 88K
		2962	architec- tures:
		2963
		2964
		2965	__-m88000__
		2966
		2967
		2968	Generate code that works well on both the m88100 and the
		2969	m88110.
		2970
		2971
		2972	__-m88100__
		2973
		2974
		2975	Generate code that works best for the m88100, but that also
		2976	runs on the m88110.
		2977
		2978
		2979	__-m88110__
		2980
		2981
		2982	Generate code that works best for the m88110, and may not
		2983	run on the m88100.
		2984
		2985
		2986	__-midentify-revision__
		2987
		2988
		2989	Include an __ident__ directive in the assembler output
		2990	recording the source file name, compiler name and version,
		2991	timestamp, and compilation flags used.
		2992
		2993
		2994	__-mno-underscores__
		2995
		2996
		2997	In assembler output, emit symbol names without adding an
		2998	underscore character at the beginning of each name. The
		2999	default is to use an underscore as prefix on each
		3000	name.
		3001
		3002
		3003	__-mno-check-zero-division__
		3004
		3005
		3006	__-mcheck-zero-division__
		3007
		3008
		3009	Early models of the 88K architecture had problems with
		3010	division by zero; in particular, many of them didn't trap.
		3011	Use these options to avoid including (or to include
		3012	explicitly) additional code to de- tect division by zero and
		3013	signal an exception. All GCC configurations for the 88K use
		3014	`__-mcheck-ze- ro-division__' by default.
		3015
		3016
		3017	__-mocs-debug-info__
		3018
		3019
		3020	__-mno-ocs-debug-info__
		3021
		3022
		3023	Include (or omit) additional debugging information (about
		3024	registers used in each stack frame) as spec- ified in the
		3025	88Open Object Compatibility Standard, ``OCS''. This extra
		3026	information is not needed by GDB. The default for DG/UX,
		3027	SVr4, and Delta 88 SVr3.2 is to include this information;
		3028	other 88k configurations omit this information by
		3029	default.
		3030
		3031
		3032	__-mocs-frame-position__
		3033
		3034
		3035	__-mno-ocs-frame-position__
		3036
		3037
		3038	Force (or do not require) register values to be stored in a
		3039	particular place in stack frames, as specified in OCS. The
		3040	DG/UX, Delta88 SVr3.2, and BCS configurations use
		3041	`__-mocs-frame-position__'; oth- er 88k configurations
		3042	have the default
		3043	`__-mno-ocs-frame-position__'.
		3044
		3045
		3046	__-moptimize-arg-area__
		3047
		3048
		3049	__-mno-optimize-arg-area__
		3050
		3051
		3052	Control how to store function arguments in stack frames.
		3053	`__-moptimize-arg-area__' saves space, but may break some
		3054	debuggers (not GDB). `__-mno-opti- mize-arg-area__'
		3055	conforms better to standards. By default GCC does not
		3056	optimize the argument area.
		3057
		3058
		3059	__-mshort-data-__''num''
		3060
		3061
		3062	''num'' Generate smaller data references by making them
		3063	relative to __r0__, which allows loading a value using a
		3064	single instruction (rather than the usual two). You control
		3065	which data references are affected by specifying ''num''
		3066	with this option. For example, if you specify
		3067	`__-mshort-data-512__', then the data ref- erences
		3068	affected are those involving displacements of less than 512
		3069	bytes. `__-mshort-data-__''num''' is not effective for
		3070	''num'' greater than 64K.
		3071
		3072
		3073	__-mserialize-volatile__
		3074
		3075
		3076	__-mno-serialize-volatile__
		3077
		3078
		3079	Do, or do not, generate code to guarantee sequen- tial
		3080	consistency of volatile memory references.
		3081
		3082
		3083	GNU CC always guarantees consistency by default, for the
		3084	preferred processor submodel. How this is done depends on
		3085	the submodel.
		3086
		3087
		3088	The m88100 processor does not reorder memory refer- ences
		3089	and so always provides sequential consisten- cy. If you use
		3090	`__-m88100__', GNU CC does not generate any special
		3091	instructions for sequential consisten- cy.
		3092
		3093
		3094	The order of memory references made by the m88110 processor
		3095	does not always match the order of the instructions
		3096	requesting those references. In par- ticular, a load
		3097	instruction may execute before a preceding store
		3098	instruction. Such reordering vio- lates sequential
		3099	consistency of volatile memory references, when there are
		3100	multiple processors. When you use `__-m88000__' or
		3101	`__-m88110__', GNU CC gener- ates special instructions
		3102	when appropriate, to force execution in the proper
		3103	order.
		3104
		3105
		3106	The extra code generated to guarantee consistency may affect
		3107	the performance of your application. If you know that you
		3108	can safely forgo this guarantee, you may use the option
		3109	`__-mno-serialize-volatile__'.
		3110
		3111
		3112	If you use the `__-m88100__' option but require sequen-
		3113	tial consistency when running on the m88110 proces- sor, you
		3114	should use `__-mserialize-volatile__'.
		3115
		3116
		3117	__-msvr4__
		3118
		3119
		3120	__-msvr3__
		3121
		3122
		3123	Turn on (`__-msvr4__') or off (`__-msvr3__') compiler
		3124	ex- tensions related to System V release 4 (SVr4). This
		3125	controls the following:
		3126
		3127
		3128	Which variant of the assembler syntax to emit (which you can
		3129	select independently using `__-mver-
		3130	sion-03.00__').
		3131
		3132
		3133	`__-msvr4__' makes the C preprocessor recognize
		3134	`__#prag- ma weak__'
		3135
		3136
		3137	`__-msvr4__' makes GCC issue additional declaration di-
		3138	rectives used in SVr4.
		3139
		3140
		3141	`__-msvr3__' is the default for all m88K configurations
		3142	except the SVr4 configuration.
		3143
		3144
		3145	__-mtrap-large-shift__
		3146
		3147
		3148	__-mhandle-large-shift__
		3149
		3150
		3151	Include code to detect bit-shifts of more than 31 bits;
		3152	respectively, trap such shifts or emit code to handle them
		3153	properly. By default GCC makes no special provision for
		3154	large bit shifts.
		3155
		3156
		3157	__-muse-div-instruction__
		3158
		3159
		3160	Very early models of the 88K architecture didn't have a
		3161	divide instruction, so GCC avoids that in- struction by
		3162	default. Use this option to specify that it's safe to use
		3163	the divide instruction.
		3164
		3165
		3166	__-mversion-03.00__
		3167
		3168
		3169	In the DG/UX configuration, there are two flavors of SVr4.
		3170	This option modifies __-msvr4__ to select whether the
		3171	hybrid-COFF or real-ELF flavor is used. All other
		3172	configurations ignore this option.
		3173
		3174
		3175	__-mwarn-passed-structs__
		3176
		3177
		3178	Warn when a function passes a struct as an argument or
		3179	result. Structure-passing conventions have changed during
		3180	the evolution of the C language, and are often the source of
		3181	portability problems. By default, GCC issues no such
		3182	warning.
		3183
		3184
		3185	These options are defined for the IBM RS6000:
		3186
		3187
		3188	__-mfp-in-toc__
		3189
		3190
		3191	__-mno-fp-in-toc__
		3192
		3193
		3194	Control whether or not floating-point constants go in the
		3195	Table of Contents (TOC), a table of all global variable and
		3196	function addresses. By default GCC puts floating-point
		3197	constants there; if the TOC overflows,
		3198	`__-mno-fp-in-toc__' will reduce the size of the TOC,
		3199	which may avoid the overflow.
		3200
		3201
		3202	These `__-m__' options are defined for the IBM RT
		3203	PC:
		3204
		3205
		3206	__-min-line-mul__
		3207
		3208
		3209	Use an in-line code sequence for integer multi- plies. This
		3210	is the default.
		3211
		3212
		3213	__-mcall-lib-mul__
		3214
		3215
		3216	Call __lmul$$__ for integer multiples.
		3217
		3218
		3219	__-mfull-fp-blocks__
		3220
		3221
		3222	Generate full-size floating point data blocks, in- cluding
		3223	the minimum amount of scratch space recom- mended by IBM.
		3224	This is the default.
		3225
		3226
		3227	__-mminimum-fp-blocks__
		3228
		3229
		3230	Do not include extra scratch space in floating point data
		3231	blocks. This results in smaller code, but slower execution,
		3232	since scratch space must be allocated
		3233	dynamically.
		3234
		3235
		3236	__-mfp-arg-in-fpregs__
		3237
		3238
		3239	Use a calling sequence incompatible with the IBM calling
		3240	convention in which floating point argu- ments are passed in
		3241	floating point registers. Note that __varargs.h__ and
		3242	__stdargs.h__ will not work with floating point operands
		3243	if this option is speci- fied.
		3244
		3245
		3246	__-mfp-arg-in-gregs__
		3247
		3248
		3249	Use the normal calling convention for floating point
		3250	arguments. This is the default.
		3251
		3252
		3253	__-mhc-struct-return__
		3254
		3255
		3256	Return structures of more than one word in memory, rather
		3257	than in a register. This provides compati- bility with the
2	perry	3258	!MetaWare HighC (hc) compiler. Use
1	perry	3259	`__-fpcc-struct-return__' for compatibility with the
		3260	Portable C Compiler (pcc).
		3261
		3262
		3263	__-mnohc-struct-return__
		3264
		3265
		3266	Return some structures of more than one word in registers,
		3267	when convenient. This is the default. For compatibility with
		3268	the IBM-supplied compilers, use either
		3269	`__-fpcc-struct-return__' or
		3270	`__-mhc-struct-return__'.
		3271
		3272
		3273	These `__-m__' options are defined for the MIPS family of
		3274	com- puters:
		3275
		3276
		3277	__-mcpu=__''cpu-type''
		3278
		3279
		3280	Assume the defaults for the machine type ''cpu-type''
		3281	when scheduling instructions. The default ''cpu-type'' is
		3282	__default__, which picks the longest cycles times for any
		3283	of the machines, in order that the code run at reasonable
		3284	rates on all MIPS cpu's. Other choices for ''cpu-type''
		3285	are __r2000__, __r3000__, __r4000__, and
		3286	__r6000__. While picking a specific ''cpu-type'' will
		3287	schedule things appropriately for that particular chip, the
		3288	compiler will not generate any code that does not meet level
		3289	1 of the MIPS ISA (instruction set architecture) without the
		3290	__-mips2__ or __-mips3__ switches being
		3291	used.
		3292
		3293
		3294	__-mips2__
		3295
		3296
		3297	Issue instructions from level 2 of the MIPS ISA (branch
		3298	likely, square root instructions). The __-mcpu=r4000__ or
		3299	__-mcpu=r6000__ switch must be used in conjunction with
		3300	__-mips2__.
		3301
		3302
		3303	__-mips3__
		3304
		3305
		3306	Issue instructions from level 3 of the MIPS ISA (64 bit
		3307	instructions). The __-mcpu=r4000__ switch must be used in
		3308	conjunction with __-mips2__.
		3309
		3310
		3311	__-mint64__
		3312
		3313
		3314	__-mlong64__
		3315
		3316
		3317	__-mlonglong128__
		3318
		3319
		3320	These options don't work at present.
		3321
		3322
		3323	__-mmips-as__
		3324
		3325
		3326	Generate code for the MIPS assembler, and invoke
		3327	__mips-tfile__ to add normal debug information. This is
		3328	the default for all platforms except for the OSF/1 reference
		3329	platform, using the OSF/rose object format. If any of the
		3330	__-ggdb__, __-gstabs__, or __-gstabs+__ switches
		3331	are used, the __mips-tfile__ program will en- capsulate
		3332	the stabs within MIPS ECOFF.
		3333
		3334
		3335	__-mgas__
		3336
		3337
		3338	Generate code for the GNU assembler. This is the default on
		3339	the OSF/1 reference platform, using the OSF/rose object
		3340	format.
		3341
		3342
		3343	__-mrnames__
		3344
		3345
		3346	__-mno-rnames__
		3347
		3348
		3349	The __-mrnames__ switch says to output code using the
		3350	MIPS software names for the registers, instead of the
		3351	hardware names (ie, __a0__ instead of __$4__). The GNU
		3352	assembler does not support the __-mrnames__ switch, and
		3353	the MIPS assembler will be instructed to run the MIPS C
		3354	preprocessor over the source file. The __-mno-rnames__
		3355	switch is default.
		3356
		3357
		3358	__-mgpopt__
		3359
		3360
		3361	__-mno-gpopt__
		3362
		3363
		3364	The __-mgpopt__ switch says to write all of the data
		3365	declarations before the instructions in the text section, to
		3366	all the MIPS assembler to generate one word memory
		3367	references instead of using two words for short global or
		3368	static data items. This is on by default if optimization is
		3369	selected.
		3370
		3371
		3372	__-mstats__
		3373
		3374
		3375	__-mno-stats__
		3376
		3377
		3378	For each non-inline function processed, the __-mstats__
		3379	switch causes the compiler to emit one line to the standard
		3380	error file to print statistics about the program (number of
		3381	registers saved, stack size, etc.).
		3382
		3383
		3384	__-mmemcpy__
		3385
		3386
		3387	__-mno-memcpy__
		3388
		3389
		3390	The __-mmemcpy__ switch makes all block moves call the
		3391	appropriate string function (__memcpy__ or __bcopy__)
		3392	in- stead of possibly generating inline code.
		3393
		3394
		3395	__-mmips-tfile__
		3396
		3397
		3398	__-mno-mips-tfile__
		3399
		3400
		3401	The __-mno-mips-tfile__ switch causes the compiler not
		3402	postprocess the object file with the __mips-tfile__
		3403	program, after the MIPS assembler has generated it to add
		3404	debug support. If __mips-tfile__ is not run, then no
		3405	local variables will be available to the debugger. In
		3406	addition, __stage2__ and __stage3__ objects will have
		3407	the temporary file names passed to the assembler embedded in
		3408	the object file, which means the objects will not compare
		3409	the same.
		3410
		3411
		3412	__-msoft-float__
		3413
		3414
		3415	Generate output containing library calls for float- ing
		3416	point. ''WARNING:'' the requisite libraries are not part
		3417	of GNU CC. Normally the facilities of the machine's usual C
		3418	compiler are used, but this can't be done directly in
		3419	cross-compilation. You must make your own arrangements to
		3420	provide suitable li- brary functions for
		3421	cross-compilation.
		3422
		3423
		3424	__-mhard-float__
		3425
		3426
		3427	Generate output containing floating point instruc- tions.
		3428	This is the default if you use the unmodi- fied
		3429	sources.
		3430
		3431
		3432	__-mfp64__
		3433
		3434
		3435	Assume that the __FR__ bit in the status word is on, and
		3436	that there are 32 64-bit floating point regis- ters, instead
		3437	of 32 32-bit floating point regis- ters. You must also
		3438	specify the __-mcpu=r4000__ and __-mips3__
		3439	switches.
		3440
		3441
		3442	__-mfp32__
		3443
		3444
		3445	Assume that there are 32 32-bit floating point reg- isters.
		3446	This is the default.
		3447
		3448
		3449	__-mabicalls__
		3450
		3451
		3452	__-mno-abicalls__
		3453
		3454
		3455	Emit (or do not emit) the __.abicalls__, __.cpload__,
		3456	and __.cprestore__ pseudo operations that some System V.4
		3457	ports use for position independent code.
		3458
		3459
		3460	__-mhalf-pic__
		3461
		3462
		3463	__-mno-half-pic__
		3464
		3465
		3466	The __-mhalf-pic__ switch says to put pointers to ex-
		3467	tern references into the data section and load them up,
		3468	rather than put the references in the text sec- tion. This
		3469	option does not work at present. __-G__''num'' Put
		3470	global and static items less than or equal to ''num''
		3471	bytes into the small data or bss sections in- stead of the
		3472	normal data or bss section. This al- lows the assembler to
		3473	emit one word memory refer- ence instructions based on the
		3474	global pointer (__gp__ or __$28__), instead of the
		3475	normal two words used. By default, ''num'' is 8 when the
		3476	MIPS assembler is used, and 0 when the GNU assembler is
		3477	used. The __-G__''num'' switch is also passed to the
		3478	assembler and linker. All modules should be compiled with
		3479	the same __-G__''num'' value.
		3480
		3481
		3482	__-nocpp__
		3483
		3484
		3485	Tell the MIPS assembler to not run its preprocessor over
		3486	user assembler files (with a `__.s__' suffix) when
		3487	assembling them.
		3488
		3489
		3490	These `__-m__' options are defined for the Intel 80386
		3491	family of computers: __-m486__
		3492
		3493
		3494	__-mno-486__
		3495
		3496
		3497	Control whether or not code is optimized for a 486 instead
		3498	of an 386. Code generated for a 486 will run on a 386 and
		3499	vice versa.
		3500
		3501
		3502	__-msoft-float__
		3503
		3504
		3505	Generate output containing library calls for float- ing
		3506	point. ''Warning:'' the requisite libraries are not part
		3507	of GNU CC. Normally the facilities of the machine's usual C
		3508	compiler are used, but this can't be done directly in
		3509	cross-compilation. You must make your own arrangements to
		3510	provide suitable li- brary functions for
		3511	cross-compilation.
		3512
		3513
		3514	On machines where a function returns floating point results
		3515	in the 80387 register stack, some floating point opcodes may
		3516	be emitted even if `__-msoft-float__' is
		3517	used.
		3518
		3519
		3520	__-mno-fp-ret-in-387__
		3521
		3522
		3523	Do not use the FPU registers for return values of
		3524	functions.
		3525
		3526
		3527	The usual calling convention has functions return values of
		3528	types __float__ and __double__ in an FPU regis- ter,
		3529	even if there is no FPU. The idea is that the operating
		3530	system should emulate an FPU.
		3531
		3532
		3533	The option `__-mno-fp-ret-in-387__' causes such values to
		3534	be returned in ordinary CPU registers instead.
		3535
		3536
		3537	These `__-m__' options are defined for the HPPA family of
		3538	com- puters:
		3539
		3540
		3541	__-mpa-risc-1-0__
		3542
		3543
		3544	Generate code for a PA 1.0 processor.
		3545
		3546
		3547	__-mpa-risc-1-1__
		3548
		3549
		3550	Generate code for a PA 1.1 processor.
		3551
		3552
		3553	__-mkernel__
		3554
		3555
		3556	Generate code which is suitable for use in kernels.
		3557	Specifically, avoid __add__ instructions in which one of
		3558	the arguments is the DP register; generate __addil__
		3559	instructions instead. This avoids a rather serious bug in
		3560	the HP-UX linker.
		3561
		3562
		3563	__-mshared-libs__
		3564
		3565
		3566	Generate code that can be linked against HP-UX shared
		3567	libraries. This option is not fully func- tion yet, and is
		3568	not on by default for any PA tar- get. Using this option can
		3569	cause incorrect code to be generated by the
		3570	compiler.
		3571
		3572
		3573	__-mno-shared-libs__
		3574
		3575
		3576	Don't generate code that will be linked against shared
		3577	libraries. This is the default for all PA
		3578	targets.
		3579
		3580
		3581	__-mlong-calls__
		3582
		3583
		3584	Generate code which allows calls to functions greater than
		3585	256K away from the caller when the caller and callee are in
		3586	the same source file. Do not turn this option on unless code
		3587	refuses to link with ``branch out of range errors from the
		3588	linker.
		3589
		3590
		3591	__-mdisable-fpregs__
		3592
		3593
		3594	Prevent floating point registers from being used in any
		3595	manner. This is necessary for compiling ker- nels which
		3596	perform lazy context switching of float- ing point
		3597	registers. If you use this option and attempt to perform
		3598	floating point operations, the compiler will
		3599	abort.
		3600
		3601
		3602	__-mdisable-indexing__
		3603
		3604
		3605	Prevent the compiler from using indexing address modes. This
		3606	avoids some rather obscure problems when compiling MIG
		3607	generated code under MACH.
		3608
		3609
		3610	__-mtrailing-colon__
		3611
		3612
		3613	Add a colon to the end of label definitions (for ELF
		3614	assemblers).
		3615
		3616
		3617	These `__-m__' options are defined for the Intel 80960
		3618	family of computers:
		3619
		3620
		3621	__-m__''cpu-type''
		3622
		3623
		3624	Assume the defaults for the machine type ''cpu-type'' for
		3625	instruction and addressing-mode availability and alignment.
		3626	The default ''cpu-type'' is __kb__; other choices are
		3627	__ka__, __mc__, __ca__, __cf__, __sa__, and
		3628	__sb__.
		3629
		3630
		3631	__-mnumerics__
		3632
		3633
		3634	__-msoft-float__
		3635
		3636
		3637	The __-mnumerics__ option indicates that the processor
		3638	does support floating-point instructions. The
		3639	__-msoft-float__ option indicates that floating-point
		3640	support should not be assumed.
		3641
		3642
		3643	__-mleaf-procedures__
		3644
		3645
		3646	__-mno-leaf-procedures__
		3647
		3648
		3649	Do (or do not) attempt to alter leaf procedures to be
		3650	callable with the ''bal'' instruction as well as
		3651	''call''. This will result in more efficient code for
		3652	explicit calls when the ''bal'' instruction can be sub-
		3653	stituted by the assembler or linker, but less effi- cient
		3654	code in other cases, such as calls via func- tion pointers,
		3655	or using a linker that doesn't sup- port this
		3656	optimization.
		3657
		3658
		3659	__-mtail-call__
		3660
		3661
		3662	__-mno-tail-call__
		3663
		3664
		3665	Do (or do not) make additional attempts (beyond those of the
		3666	machine-independent portions of the compiler) to optimize
		3667	tail-recursive calls into branches. You may not want to do
		3668	this because the detection of cases where this is not valid
		3669	is not totally complete. The default is
		3670	__-mno-tail-call__.
		3671
		3672
		3673	__-mcomplex-addr__
		3674
		3675
		3676	__-mno-complex-addr__
		3677
		3678
		3679	Assume (or do not assume) that the use of a complex
		3680	addressing mode is a win on this implementation of the i960.
		3681	Complex addressing modes may not be worthwhile on the
		3682	K-series, but they definitely are on the C-series. The
		3683	default is currently __-mcom- plex-addr__ for all
		3684	processors except the CB and CC.
		3685
		3686
		3687	__-mcode-align__
		3688
		3689
		3690	__-mno-code-align__
		3691
		3692
		3693	Align code to 8-byte boundaries for faster fetching (or
		3694	don't bother). Currently turned on by default for C-series
		3695	implementations only.
		3696
		3697
		3698	__-mic-compat__
		3699
		3700
		3701	__-mic2.0-compat__
		3702
		3703
		3704	__-mic3.0-compat__
		3705
		3706
		3707	Enable compatibility with iC960 v2.0 or v3.0.
		3708
		3709
		3710	__-masm-compat__
		3711
		3712
		3713	__-mintel-asm__
		3714
		3715
		3716	Enable compatibility with the iC960 assembler.
		3717
		3718
		3719	__-mstrict-align__
		3720
		3721
		3722	__-mno-strict-align__
		3723
		3724
		3725	Do not permit (do permit) unaligned accesses.
		3726
		3727
		3728	__-mold-align__
		3729
		3730
		3731	Enable structure-alignment compatibility with In- tel's gcc
		3732	release version 1.3 (based on gcc 1.37). Currently this is
		3733	buggy in that __#pragma align 1__ is always assumed as
		3734	well, and cannot be turned off.
		3735
		3736
		3737	These `__-m__' options are defined for the DEC Alpha
		3738	implemen- tations:
		3739
		3740
		3741	__-mno-soft-float__
		3742
		3743
		3744	__-msoft-float__
		3745
		3746
		3747	Use (do not use) the hardware floating-point in- structions
		3748	for floating-point operations. When __-msoft-float__ is
		3749	specified, functions in `__libgcc1.c__' will be used to
		3750	perform floating-point operations. Unless they are replaced
		3751	by routines that emulate the floating-point operations, or
		3752	compiled in such a way as to call such emulations routines,
		3753	these routines will issue floating-point operations. If you
		3754	are compiling for an Alpha without float- ing-point
		3755	operations, you must ensure that the li- brary is built so
		3756	as not to call them.
		3757
		3758
		3759	Note that Alpha implementations without float- ing-point
		3760	operations are required to have float- ing-point
		3761	registers.
		3762
		3763
		3764	__-mfp-reg__
		3765
		3766
		3767	__-mno-fp-regs__
		3768
		3769
		3770	Generate code that uses (does not use) the float- ing-point
		3771	register set. __-mno-fp-regs__ implies
		3772	__-msoft-float__. If the floating-point register set is
		3773	not used, floating point operands are passed in integer
		3774	registers as if they were integers and floating-point
		3775	results are passed in $0 instead of $f0. This is a
		3776	non-standard calling sequence, so any function with a
		3777	floating-point argument or re- turn value called by code
		3778	compiled with __-mno-fp-regs__ must also be compiled with
		3779	that op- tion.
		3780
		3781
		3782	A typical use of this option is building a kernel that does
		3783	not use, and hence need not save and re- store, any
		3784	floating-point registers.
		3785
		3786
		3787	These additional options are available on System V Release 4
		3788	for compatibility with other compilers on those
		3789	systems:
		3790
		3791
		3792	__-G__
		3793
		3794
		3795	On SVr4 systems, __gcc__ accepts the option `__-G__'
		3796	(and passes it to the system linker), for compatibility with
		3797	other compilers. However, we suggest you use
		3798	`__-symbolic__' or `__-shared__' as appropriate,
		3799	instead of supplying linker options on the __gcc__
		3800	command line.
		3801
		3802
		3803	__-Qy__
		3804
		3805
		3806	Identify the versions of each tool used by the com- piler,
		3807	in a __.ident__ assembler directive in the out-
		3808	put.
		3809
		3810
		3811	__-Qn__
		3812
		3813
		3814	Refrain from adding __.ident__ directives to the output
		3815	file (this is the default).
		3816
		3817
		3818	__-YP,__''dirs''
		3819
		3820
		3821	Search the directories ''dirs'', and no others, for li-
		3822	braries specified with `__-l__'. You can separate di-
		3823	rectory entries in ''dirs'' from one another with
		3824	colons.
		3825
		3826
		3827	__-Ym,__''dir''
		3828
		3829
		3830	Look in the directory ''dir'' to find the M4 preproces-
		3831	sor. The assembler uses this option.
		3832	!!CODE GENERATION OPTIONS
		3833
		3834
		3835	These machine-independent options control the interface
		3836	conventions used in code generation.
		3837
		3838
		3839	Most of them begin with `-f'. These options have both
		3840	positive and negative forms; the negative form of
		3841	`__-ffoo__' would be `__-fno-foo__'. In the table
		3842	below, only one of the forms is listed--the one which is not
		3843	the default. You can figure out the other form by either
		3844	removing `__no-__' or adding it.
		3845
		3846
		3847	__-fnonnull-objects__
		3848
		3849
		3850	Assume that objects reached through references are not null
		3851	(C++ only).
		3852
		3853
		3854	Normally, GNU C++ makes conservative assumptions about
		3855	objects reached through references. For ex- ample, the
		3856	compiler must check that __a__ is not null in code like
		3857	the following:
		3858
		3859
		3860	obj
		3861
		3862
		3863	Checking that references of this sort have non-null values
		3864	requires extra code, however, and it is un- necessary for
		3865	many programs. You can use `__-fnon- null-objects__' to
		3866	omit the checks for null, if your program doesn't require
		3867	checking.
		3868
		3869
		3870	__-fpcc-struct-return__
		3871
		3872
		3873	Use the same convention for returning __struct__ and
		3874	__union__ values that is used by the usual C compiler on
		3875	your system. This convention is less efficient for small
		3876	structures, and on many machines it fails to be reentrant;
		3877	but it has the advantage of allow- ing intercallability
		3878	between GCC-compiled code and PCC-compiled
		3879	code.
		3880
		3881
		3882	__-freg-struct-return__
		3883
		3884
		3885	Use the convention that __struct__ and __union__
		3886	values are returned in registers when possible. This is more
		3887	efficient for small structures than
		3888	__-fpcc-struct-return__.
		3889
		3890
		3891	If you specify neither __-fpcc-struct-return__ nor
		3892	__-freg-struct-return__, GNU CC defaults to whichever
		3893	convention is standard for the target. If there is no
		3894	standard convention, GNU CC defaults to
		3895	__-fpcc-struct-return__.
		3896
		3897
		3898	__-fshort-enums__
		3899
		3900
		3901	Allocate to an __enum__ type only as many bytes as it
		3902	needs for the declared range of possible values.
		3903	Specifically, the __enum__ type will be equivalent to the
		3904	smallest integer type which has enough room.
		3905
		3906
		3907	__-fshort-double__
		3908
		3909
		3910	Use the same size for __double__ as for __float__
		3911	.
		3912
		3913
		3914	__-fshared-data__
		3915
		3916
		3917	Requests that the data and non-__const__ variables of
		3918	this compilation be shared data rather than private data.
		3919	The distinction makes sense only on certain operating
		3920	systems, where shared data is shared be- tween processes
		3921	running the same program, while private data exists in one
		3922	copy per process.
		3923
		3924
		3925	__-fno-common__
		3926
		3927
		3928	Allocate even uninitialized global variables in the bss
		3929	section of the object file, rather than gener- ating them as
		3930	common blocks. This has the effect that if the same variable
		3931	is declared (without __ex- tern__) in two different
		3932	compilations, you will get an error when you link them. The
		3933	only reason this might be useful is if you wish to verify
		3934	that the program will work on other systems which always
		3935	work this way.
		3936
		3937
		3938	__-fno-ident__
		3939
		3940
		3941	Ignore the `__#ident__' directive.
		3942
		3943
		3944	__-fno-gnu-linker__
		3945
		3946
		3947	Do not output global initializations (such as C++
		3948	constructors and destructors) in the form used by the GNU
		3949	linker (on systems where the GNU linker is the standard
		3950	method of handling them). Use this option when you want to
		3951	use a non-GNU linker, which also requires using the
		3952	__collect2__ program to make sure the system linker
		3953	includes constructors and destructors. (__collect2__ is
		3954	included in the GNU CC distribution.) For systems which
		3955	''must'' use __col- lect2__, the compiler driver
		3956	__gcc__ is configured to do this
		3957	automatically.
		3958
		3959
		3960	__-finhibit-size-directive__
		3961
		3962
		3963	Don't output a __.size__ assembler directive, or any-
		3964	thing else that would cause trouble if the function is split
		3965	in the middle, and the two halves are placed at locations
		3966	far apart in memory. This op- tion is used when compiling
		3967	`__crtstuff.c__'; you should not need to use it for
		3968	anything else.
		3969
		3970
		3971	__-fverbose-asm__
		3972
		3973
		3974	Put extra commentary information in the generated assembly
		3975	code to make it more readable. This op- tion is generally
		3976	only of use to those who actually need to read the generated
		3977	assembly code (perhaps while debugging the compiler
		3978	itself).
		3979
		3980
		3981	__-fvolatile__
		3982
		3983
		3984	Consider all memory references through pointers to be
		3985	volatile.
		3986
		3987
		3988	__-fvolatile-global__
		3989
		3990
		3991	Consider all memory references to extern and global data
		3992	items to be volatile.
		3993
		3994
		3995	__-fpic__
		3996
		3997
		3998	If supported for the target machines, generate po-
		3999	sition-independent code, suitable for use in a shared
		4000	library.
		4001
		4002
		4003	__-fPIC__
		4004
		4005
		4006	If supported for the target machine, emit posi-
		4007	tion-independent code, suitable for dynamic link- ing, even
		4008	if branches need large displacements.
		4009
		4010
		4011	__-ffixed-__''reg''
		4012
		4013
		4014	Treat the register named ''reg'' as a fixed register;
		4015	generated code should never refer to it (except perhaps as a
		4016	stack pointer, frame pointer or in some other fixed
		4017	role).
		4018
		4019
		4020	''reg'' must be the name of a register. The register
		4021	names accepted are machine-specific and are defined in the
		4022	__REGISTER_NAMES__ macro in the machine descrip- tion
		4023	macro file.
		4024
		4025
		4026	This flag does not have a negative form, because it
		4027	specifies a three-way choice.
		4028
		4029
		4030	__-fcall-used-__''reg''
		4031
		4032
		4033	Treat the register named ''reg'' as an allocable regis-
		4034	ter that is clobbered by function calls. It may be allocated
		4035	for temporaries or variables that do not live across a call.
		4036	Functions compiled this way will not save and restore the
		4037	register ''reg''.
		4038
		4039
		4040	Use of this flag for a register that has a fixed pervasive
		4041	role in the machine's execution model, such as the stack
		4042	pointer or frame pointer, will produce disastrous
		4043	results.
		4044
		4045
		4046	This flag does not have a negative form, because it
		4047	specifies a three-way choice.
		4048
		4049
		4050	__-fcall-saved-__''reg''
		4051
		4052
		4053	Treat the register named ''reg'' as an allocable regis-
		4054	ter saved by functions. It may be allocated even for
		4055	temporaries or variables that live across a call. Functions
		4056	compiled this way will save and restore the register
		4057	''reg'' if they use it.
		4058
		4059
		4060	Use of this flag for a register that has a fixed pervasive
		4061	role in the machine's execution model, such as the stack
		4062	pointer or frame pointer, will produce disastrous
		4063	results.
		4064
		4065
		4066	A different sort of disaster will result from the use of
		4067	this flag for a register in which function values may be
		4068	returned.
		4069
		4070
		4071	This flag does not have a negative form, because it
		4072	specifies a three-way choice.
		4073	!!PRAGMAS
		4074
		4075
		4076	Two `__#pragma__' directives are supported for GNU C++,
		4077	to permit using the same header file for two purposes: as a
		4078	definition of interfaces to a given object class, and as the
		4079	full definition of the contents of that object
		4080	class.
		4081
		4082
		4083	__#pragma interface__
		4084
		4085
		4086	(C++ only.) Use this directive in header files that define
		4087	object classes, to save space in most of the object files
		4088	that use those classes. Nor- mally, local copies of certain
		4089	information (backup copies of inline member functions,
		4090	debugging infor- mation, and the internal tables that
		4091	implement vir- tual functions) must be kept in each object
		4092	file that includes class definitions. You can use this
		4093	pragma to avoid such duplication. When a header file
		4094	containing `__#pragma interface__' is included in a
		4095	compilation, this auxiliary information will not be
		4096	generated (unless the main input source file it- self uses
		4097	`__#pragma implementation__'). Instead, the object files
		4098	will contain references to be resolved at link
		4099	time.
		4100
		4101
		4102	__#pragma implementation__
		4103
		4104
		4105	__#pragma implementation
		4106	__''objects''__.h__
		4107
		4108
		4109	(C++ only.) Use this pragma in a main input file, when you
		4110	want full output from included header files to be generated
		4111	(and made globally visible). The included header file, in
		4112	turn, should use `__#pragma interface__'. Backup copies
		4113	of inline mem- ber functions, debugging information, and the
		4114	in- ternal tables used to implement virtual functions are
		4115	all generated in implementation files.
		4116
		4117
		4118	If you use `__#pragma implementation__' with no argu-
		4119	ment, it applies to an include file with the same basename
		4120	as your source file; for example, in `__all- class.cc__',
		4121	`__#pragma implementation__' by itself is equivalent to
		4122	`__#pragma implementation __'.
		4123	Use the string argument if you want a single implementation
		4124	file to include code from multiple header
		4125	files.
		4126
		4127
		4128	There is no way to split up the contents of a sin- gle
		4129	header file into multiple implementation files.
		4130	!!FILES
		4131
		4132
		4133	file.c C source file
		4134	file.h C header (preprocessor) file
		4135	file.i preprocessed C source file
		4136	file.C C++ source file
		4137	file.cc C++ source file
		4138	file.cxx C++ source file
		4139	file.m Objective-C source file
		4140	file.s assembly language file
		4141	file.o object file
		4142	a.out link edited output
		4143	''TMPDIR''/cc temporary files
		4144	''LIBDIR''/cpp preprocessor
		4145	''LIBDIR''/cc1 compiler for C
		4146	''LIBDIR''/cc1plus compiler for C++
		4147	''LIBDIR''/collect linker front end needed on some machines
		4148	''LIBDIR''/libgcc.a GCC subroutine library
		4149	/lib/crt[[01n].o start-up routine
		4150	''LIBDIR''/ccrt0 additional start-up routine for C++
		4151	/lib/libc.a standard C library, see
		4152	intro(3)
		4153	/usr/include standard directory for __#include__ files
		4154	''LIBDIR''/include standard gcc directory for __#include__ files
		4155	''LIBDIR''/g++-include additional g++ directory for __#include
		4156	__''LIBDIR'' is usually __/usr/local/lib/__''machine''/''version''.''
		4157	TMPDIR'' comes from the environment variable __TMPDIR__ (default __/usr/tmp__ if available, else __/tmp__).
		4158	!!EXIT STATUS
		4159
		4160
		4161	Normally the exit status is 0, if compilation or link edit
		4162	are successful, and nonzero else. The option __-Werror__
		4163	treats each warning as an error.
		4164	!!SEE ALSO
		4165
		4166
		4167	cpp(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1).
		4168	`__gcc__', `__cpp__', `__as__', `__ld__', and
		4169	`__gdb__' entries in __info__.''
		4170	Using and Porting GNU CC (for version 2.0)'', Richard M.
		4171	Stallman; ''The C Preprocessor'', Richard M. Stallman;
		4172	''Debug- ging with GDB: the GNU Source-Level Debugger'',
		4173	Richard M. Stallman and Roland H. Pesch; ''Using as: the
		4174	GNU Assembler'', Dean Elsner, Jay Fenlason
		4175	''ld: the GNU linker'', Steve Chamberlain and Roland
		4176	Pesch.
		4177	!!BUGS
		4178
		4179
		4180	For instructions on reporting bugs, see the GCC
		4181	manual.
		4182	!!COPYING
		4183
		4184
		4185	Copyright 1991, 1992, 1993 Free Software Foundation,
		4186	Inc.
		4187
		4188
		4189	Permission is granted to make and distribute verbatim copies
		4190	of this manual provided the copyright notice and this
		4191	permission notice are preserved on all copies.
		4192
		4193
		4194	Permission is granted to copy and distribute modified ver-
		4195	sions of this manual under the conditions for verbatim
		4196	copying, provided that the entire resulting derived work is
		4197	distributed under the terms of a permission notice identical
		4198	to this one.
		4199
		4200
		4201	Permission is granted to copy and distribute translations of
		4202	this manual into another language, under the above con-
		4203	ditions for modified versions, except that this permission
		4204	notice may be included in translations approved by the Free
		4205	Software Foundation instead of in the original En-
		4206	glish.
		4207	!!AUTHORS
		4208
		4209
		4210	See the GNU CC Manual for the contributors to GNU
		4211	CC.
		4212	----