xref: /netbsd-src/external/gpl3/gcc.old/dist/gcc/doc/cpp.texi (revision 8ecbf5f02b752fcb7debe1a8fab1dc82602bc760)
1\input texinfo
2@setfilename cpp.info
3@settitle The C Preprocessor
4@setchapternewpage off
5@c @smallbook
6@c @cropmarks
7@c @finalout
8
9@include gcc-common.texi
10
11@copying
12@c man begin COPYRIGHT
13Copyright @copyright{} 1987-2018 Free Software Foundation, Inc.
14
15Permission is granted to copy, distribute and/or modify this document
16under the terms of the GNU Free Documentation License, Version 1.3 or
17any later version published by the Free Software Foundation.  A copy of
18the license is included in the
19@c man end
20section entitled ``GNU Free Documentation License''.
21@ignore
22@c man begin COPYRIGHT
23man page gfdl(7).
24@c man end
25@end ignore
26
27@c man begin COPYRIGHT
28This manual contains no Invariant Sections.  The Front-Cover Texts are
29(a) (see below), and the Back-Cover Texts are (b) (see below).
30
31(a) The FSF's Front-Cover Text is:
32
33     A GNU Manual
34
35(b) The FSF's Back-Cover Text is:
36
37     You have freedom to copy and modify this GNU Manual, like GNU
38     software.  Copies published by the Free Software Foundation raise
39     funds for GNU development.
40@c man end
41@end copying
42
43@c Create a separate index for command line options.
44@defcodeindex op
45@syncodeindex vr op
46
47@c Used in cppopts.texi and cppenv.texi.
48@set cppmanual
49
50@ifinfo
51@dircategory Software development
52@direntry
53* Cpp: (cpp).                  The GNU C preprocessor.
54@end direntry
55@end ifinfo
56
57@titlepage
58@title The C Preprocessor
59@versionsubtitle
60@author Richard M. Stallman, Zachary Weinberg
61@page
62@c There is a fill at the bottom of the page, so we need a filll to
63@c override it.
64@vskip 0pt plus 1filll
65@insertcopying
66@end titlepage
67@contents
68@page
69
70@ifnottex
71@node Top
72@top
73The C preprocessor implements the macro language used to transform C,
74C++, and Objective-C programs before they are compiled.  It can also be
75useful on its own.
76
77@menu
78* Overview::
79* Header Files::
80* Macros::
81* Conditionals::
82* Diagnostics::
83* Line Control::
84* Pragmas::
85* Other Directives::
86* Preprocessor Output::
87* Traditional Mode::
88* Implementation Details::
89* Invocation::
90* Environment Variables::
91* GNU Free Documentation License::
92* Index of Directives::
93* Option Index::
94* Concept Index::
95
96@detailmenu
97 --- The Detailed Node Listing ---
98
99Overview
100
101* Character sets::
102* Initial processing::
103* Tokenization::
104* The preprocessing language::
105
106Header Files
107
108* Include Syntax::
109* Include Operation::
110* Search Path::
111* Once-Only Headers::
112* Alternatives to Wrapper #ifndef::
113* Computed Includes::
114* Wrapper Headers::
115* System Headers::
116
117Macros
118
119* Object-like Macros::
120* Function-like Macros::
121* Macro Arguments::
122* Stringizing::
123* Concatenation::
124* Variadic Macros::
125* Predefined Macros::
126* Undefining and Redefining Macros::
127* Directives Within Macro Arguments::
128* Macro Pitfalls::
129
130Predefined Macros
131
132* Standard Predefined Macros::
133* Common Predefined Macros::
134* System-specific Predefined Macros::
135* C++ Named Operators::
136
137Macro Pitfalls
138
139* Misnesting::
140* Operator Precedence Problems::
141* Swallowing the Semicolon::
142* Duplication of Side Effects::
143* Self-Referential Macros::
144* Argument Prescan::
145* Newlines in Arguments::
146
147Conditionals
148
149* Conditional Uses::
150* Conditional Syntax::
151* Deleted Code::
152
153Conditional Syntax
154
155* Ifdef::
156* If::
157* Defined::
158* Else::
159* Elif::
160
161Implementation Details
162
163* Implementation-defined behavior::
164* Implementation limits::
165* Obsolete Features::
166
167Obsolete Features
168
169* Obsolete Features::
170
171@end detailmenu
172@end menu
173
174@insertcopying
175@end ifnottex
176
177@node Overview
178@chapter Overview
179@c man begin DESCRIPTION
180The C preprocessor, often known as @dfn{cpp}, is a @dfn{macro processor}
181that is used automatically by the C compiler to transform your program
182before compilation.  It is called a macro processor because it allows
183you to define @dfn{macros}, which are brief abbreviations for longer
184constructs.
185
186The C preprocessor is intended to be used only with C, C++, and
187Objective-C source code.  In the past, it has been abused as a general
188text processor.  It will choke on input which does not obey C's lexical
189rules.  For example, apostrophes will be interpreted as the beginning of
190character constants, and cause errors.  Also, you cannot rely on it
191preserving characteristics of the input which are not significant to
192C-family languages.  If a Makefile is preprocessed, all the hard tabs
193will be removed, and the Makefile will not work.
194
195Having said that, you can often get away with using cpp on things which
196are not C@.  Other Algol-ish programming languages are often safe
197(Pascal, Ada, etc.) So is assembly, with caution.  @option{-traditional-cpp}
198mode preserves more white space, and is otherwise more permissive.  Many
199of the problems can be avoided by writing C or C++ style comments
200instead of native language comments, and keeping macros simple.
201
202Wherever possible, you should use a preprocessor geared to the language
203you are writing in.  Modern versions of the GNU assembler have macro
204facilities.  Most high level programming languages have their own
205conditional compilation and inclusion mechanism.  If all else fails,
206try a true general text processor, such as GNU M4.
207
208C preprocessors vary in some details.  This manual discusses the GNU C
209preprocessor, which provides a small superset of the features of ISO
210Standard C@.  In its default mode, the GNU C preprocessor does not do a
211few things required by the standard.  These are features which are
212rarely, if ever, used, and may cause surprising changes to the meaning
213of a program which does not expect them.  To get strict ISO Standard C,
214you should use the @option{-std=c90}, @option{-std=c99},
215@option{-std=c11} or @option{-std=c17} options, depending
216on which version of the standard you want.  To get all the mandatory
217diagnostics, you must also use @option{-pedantic}.  @xref{Invocation}.
218
219This manual describes the behavior of the ISO preprocessor.  To
220minimize gratuitous differences, where the ISO preprocessor's
221behavior does not conflict with traditional semantics, the
222traditional preprocessor should behave the same way.  The various
223differences that do exist are detailed in the section @ref{Traditional
224Mode}.
225
226For clarity, unless noted otherwise, references to @samp{CPP} in this
227manual refer to GNU CPP@.
228@c man end
229
230@menu
231* Character sets::
232* Initial processing::
233* Tokenization::
234* The preprocessing language::
235@end menu
236
237@node Character sets
238@section Character sets
239
240Source code character set processing in C and related languages is
241rather complicated.  The C standard discusses two character sets, but
242there are really at least four.
243
244The files input to CPP might be in any character set at all.  CPP's
245very first action, before it even looks for line boundaries, is to
246convert the file into the character set it uses for internal
247processing.  That set is what the C standard calls the @dfn{source}
248character set.  It must be isomorphic with ISO 10646, also known as
249Unicode.  CPP uses the UTF-8 encoding of Unicode.
250
251The character sets of the input files are specified using the
252@option{-finput-charset=} option.
253
254All preprocessing work (the subject of the rest of this manual) is
255carried out in the source character set.  If you request textual
256output from the preprocessor with the @option{-E} option, it will be
257in UTF-8.
258
259After preprocessing is complete, string and character constants are
260converted again, into the @dfn{execution} character set.  This
261character set is under control of the user; the default is UTF-8,
262matching the source character set.  Wide string and character
263constants have their own character set, which is not called out
264specifically in the standard.  Again, it is under control of the user.
265The default is UTF-16 or UTF-32, whichever fits in the target's
266@code{wchar_t} type, in the target machine's byte
267order.@footnote{UTF-16 does not meet the requirements of the C
268standard for a wide character set, but the choice of 16-bit
269@code{wchar_t} is enshrined in some system ABIs so we cannot fix
270this.}  Octal and hexadecimal escape sequences do not undergo
271conversion; @t{'\x12'} has the value 0x12 regardless of the currently
272selected execution character set.  All other escapes are replaced by
273the character in the source character set that they represent, then
274converted to the execution character set, just like unescaped
275characters.
276
277In identifiers, characters outside the ASCII range can only be
278specified with the @samp{\u} and @samp{\U} escapes, not used
279directly.  If strict ISO C90 conformance is specified with an option
280such as @option{-std=c90}, or @option{-fno-extended-identifiers} is
281used, then those escapes are not permitted in identifiers.
282
283@node Initial processing
284@section Initial processing
285
286The preprocessor performs a series of textual transformations on its
287input.  These happen before all other processing.  Conceptually, they
288happen in a rigid order, and the entire file is run through each
289transformation before the next one begins.  CPP actually does them
290all at once, for performance reasons.  These transformations correspond
291roughly to the first three ``phases of translation'' described in the C
292standard.
293
294@enumerate
295@item
296@cindex line endings
297The input file is read into memory and broken into lines.
298
299Different systems use different conventions to indicate the end of a
300line.  GCC accepts the ASCII control sequences @kbd{LF}, @kbd{@w{CR
301LF}} and @kbd{CR} as end-of-line markers.  These are the canonical
302sequences used by Unix, DOS and VMS, and the classic Mac OS (before
303OSX) respectively.  You may therefore safely copy source code written
304on any of those systems to a different one and use it without
305conversion.  (GCC may lose track of the current line number if a file
306doesn't consistently use one convention, as sometimes happens when it
307is edited on computers with different conventions that share a network
308file system.)
309
310If the last line of any input file lacks an end-of-line marker, the end
311of the file is considered to implicitly supply one.  The C standard says
312that this condition provokes undefined behavior, so GCC will emit a
313warning message.
314
315@item
316@cindex trigraphs
317@anchor{trigraphs}If trigraphs are enabled, they are replaced by their
318corresponding single characters.  By default GCC ignores trigraphs,
319but if you request a strictly conforming mode with the @option{-std}
320option, or you specify the @option{-trigraphs} option, then it
321converts them.
322
323These are nine three-character sequences, all starting with @samp{??},
324that are defined by ISO C to stand for single characters.  They permit
325obsolete systems that lack some of C's punctuation to use C@.  For
326example, @samp{??/} stands for @samp{\}, so @t{'??/n'} is a character
327constant for a newline.
328
329Trigraphs are not popular and many compilers implement them
330incorrectly.  Portable code should not rely on trigraphs being either
331converted or ignored.  With @option{-Wtrigraphs} GCC will warn you
332when a trigraph may change the meaning of your program if it were
333converted.  @xref{Wtrigraphs}.
334
335In a string constant, you can prevent a sequence of question marks
336from being confused with a trigraph by inserting a backslash between
337the question marks, or by separating the string literal at the
338trigraph and making use of string literal concatenation.  @t{"(??\?)"}
339is the string @samp{(???)}, not @samp{(?]}.  Traditional C compilers
340do not recognize these idioms.
341
342The nine trigraphs and their replacements are
343
344@smallexample
345Trigraph:       ??(  ??)  ??<  ??>  ??=  ??/  ??'  ??!  ??-
346Replacement:      [    ]    @{    @}    #    \    ^    |    ~
347@end smallexample
348
349@item
350@cindex continued lines
351@cindex backslash-newline
352Continued lines are merged into one long line.
353
354A continued line is a line which ends with a backslash, @samp{\}.  The
355backslash is removed and the following line is joined with the current
356one.  No space is inserted, so you may split a line anywhere, even in
357the middle of a word.  (It is generally more readable to split lines
358only at white space.)
359
360The trailing backslash on a continued line is commonly referred to as a
361@dfn{backslash-newline}.
362
363If there is white space between a backslash and the end of a line, that
364is still a continued line.  However, as this is usually the result of an
365editing mistake, and many compilers will not accept it as a continued
366line, GCC will warn you about it.
367
368@item
369@cindex comments
370@cindex line comments
371@cindex block comments
372All comments are replaced with single spaces.
373
374There are two kinds of comments.  @dfn{Block comments} begin with
375@samp{/*} and continue until the next @samp{*/}.  Block comments do not
376nest:
377
378@smallexample
379/* @r{this is} /* @r{one comment} */ @r{text outside comment}
380@end smallexample
381
382@dfn{Line comments} begin with @samp{//} and continue to the end of the
383current line.  Line comments do not nest either, but it does not matter,
384because they would end in the same place anyway.
385
386@smallexample
387// @r{this is} // @r{one comment}
388@r{text outside comment}
389@end smallexample
390@end enumerate
391
392It is safe to put line comments inside block comments, or vice versa.
393
394@smallexample
395@group
396/* @r{block comment}
397   // @r{contains line comment}
398   @r{yet more comment}
399 */ @r{outside comment}
400
401// @r{line comment} /* @r{contains block comment} */
402@end group
403@end smallexample
404
405But beware of commenting out one end of a block comment with a line
406comment.
407
408@smallexample
409@group
410 // @r{l.c.}  /* @r{block comment begins}
411    @r{oops! this isn't a comment anymore} */
412@end group
413@end smallexample
414
415Comments are not recognized within string literals.
416@t{@w{"/* blah */"}} is the string constant @samp{@w{/* blah */}}, not
417an empty string.
418
419Line comments are not in the 1989 edition of the C standard, but they
420are recognized by GCC as an extension.  In C++ and in the 1999 edition
421of the C standard, they are an official part of the language.
422
423Since these transformations happen before all other processing, you can
424split a line mechanically with backslash-newline anywhere.  You can
425comment out the end of a line.  You can continue a line comment onto the
426next line with backslash-newline.  You can even split @samp{/*},
427@samp{*/}, and @samp{//} onto multiple lines with backslash-newline.
428For example:
429
430@smallexample
431@group
432/\
433*
434*/ # /*
435*/ defi\
436ne FO\
437O 10\
43820
439@end group
440@end smallexample
441
442@noindent
443is equivalent to @code{@w{#define FOO 1020}}.  All these tricks are
444extremely confusing and should not be used in code intended to be
445readable.
446
447There is no way to prevent a backslash at the end of a line from being
448interpreted as a backslash-newline.  This cannot affect any correct
449program, however.
450
451@node Tokenization
452@section Tokenization
453
454@cindex tokens
455@cindex preprocessing tokens
456After the textual transformations are finished, the input file is
457converted into a sequence of @dfn{preprocessing tokens}.  These mostly
458correspond to the syntactic tokens used by the C compiler, but there are
459a few differences.  White space separates tokens; it is not itself a
460token of any kind.  Tokens do not have to be separated by white space,
461but it is often necessary to avoid ambiguities.
462
463When faced with a sequence of characters that has more than one possible
464tokenization, the preprocessor is greedy.  It always makes each token,
465starting from the left, as big as possible before moving on to the next
466token.  For instance, @code{a+++++b} is interpreted as
467@code{@w{a ++ ++ + b}}, not as @code{@w{a ++ + ++ b}}, even though the
468latter tokenization could be part of a valid C program and the former
469could not.
470
471Once the input file is broken into tokens, the token boundaries never
472change, except when the @samp{##} preprocessing operator is used to paste
473tokens together.  @xref{Concatenation}.  For example,
474
475@smallexample
476@group
477#define foo() bar
478foo()baz
479     @expansion{} bar baz
480@emph{not}
481     @expansion{} barbaz
482@end group
483@end smallexample
484
485The compiler does not re-tokenize the preprocessor's output.  Each
486preprocessing token becomes one compiler token.
487
488@cindex identifiers
489Preprocessing tokens fall into five broad classes: identifiers,
490preprocessing numbers, string literals, punctuators, and other.  An
491@dfn{identifier} is the same as an identifier in C: any sequence of
492letters, digits, or underscores, which begins with a letter or
493underscore.  Keywords of C have no significance to the preprocessor;
494they are ordinary identifiers.  You can define a macro whose name is a
495keyword, for instance.  The only identifier which can be considered a
496preprocessing keyword is @code{defined}.  @xref{Defined}.
497
498This is mostly true of other languages which use the C preprocessor.
499However, a few of the keywords of C++ are significant even in the
500preprocessor.  @xref{C++ Named Operators}.
501
502In the 1999 C standard, identifiers may contain letters which are not
503part of the ``basic source character set'', at the implementation's
504discretion (such as accented Latin letters, Greek letters, or Chinese
505ideograms).  This may be done with an extended character set, or the
506@samp{\u} and @samp{\U} escape sequences.  GCC only accepts such
507characters in the @samp{\u} and @samp{\U} forms.
508
509As an extension, GCC treats @samp{$} as a letter.  This is for
510compatibility with some systems, such as VMS, where @samp{$} is commonly
511used in system-defined function and object names.  @samp{$} is not a
512letter in strictly conforming mode, or if you specify the @option{-$}
513option.  @xref{Invocation}.
514
515@cindex numbers
516@cindex preprocessing numbers
517A @dfn{preprocessing number} has a rather bizarre definition.  The
518category includes all the normal integer and floating point constants
519one expects of C, but also a number of other things one might not
520initially recognize as a number.  Formally, preprocessing numbers begin
521with an optional period, a required decimal digit, and then continue
522with any sequence of letters, digits, underscores, periods, and
523exponents.  Exponents are the two-character sequences @samp{e+},
524@samp{e-}, @samp{E+}, @samp{E-}, @samp{p+}, @samp{p-}, @samp{P+}, and
525@samp{P-}.  (The exponents that begin with @samp{p} or @samp{P} are
526used for hexadecimal floating-point constants.)
527
528The purpose of this unusual definition is to isolate the preprocessor
529from the full complexity of numeric constants.  It does not have to
530distinguish between lexically valid and invalid floating-point numbers,
531which is complicated.  The definition also permits you to split an
532identifier at any position and get exactly two tokens, which can then be
533pasted back together with the @samp{##} operator.
534
535It's possible for preprocessing numbers to cause programs to be
536misinterpreted.  For example, @code{0xE+12} is a preprocessing number
537which does not translate to any valid numeric constant, therefore a
538syntax error.  It does not mean @code{@w{0xE + 12}}, which is what you
539might have intended.
540
541@cindex string literals
542@cindex string constants
543@cindex character constants
544@cindex header file names
545@c the @: prevents makeinfo from turning '' into ".
546@dfn{String literals} are string constants, character constants, and
547header file names (the argument of @samp{#include}).@footnote{The C
548standard uses the term @dfn{string literal} to refer only to what we are
549calling @dfn{string constants}.}  String constants and character
550constants are straightforward: @t{"@dots{}"} or @t{'@dots{}'}.  In
551either case embedded quotes should be escaped with a backslash:
552@t{'\'@:'} is the character constant for @samp{'}.  There is no limit on
553the length of a character constant, but the value of a character
554constant that contains more than one character is
555implementation-defined.  @xref{Implementation Details}.
556
557Header file names either look like string constants, @t{"@dots{}"}, or are
558written with angle brackets instead, @t{<@dots{}>}.  In either case,
559backslash is an ordinary character.  There is no way to escape the
560closing quote or angle bracket.  The preprocessor looks for the header
561file in different places depending on which form you use.  @xref{Include
562Operation}.
563
564No string literal may extend past the end of a line.  You may use continued
565lines instead, or string constant concatenation.
566
567@cindex punctuators
568@cindex digraphs
569@cindex alternative tokens
570@dfn{Punctuators} are all the usual bits of punctuation which are
571meaningful to C and C++.  All but three of the punctuation characters in
572ASCII are C punctuators.  The exceptions are @samp{@@}, @samp{$}, and
573@samp{`}.  In addition, all the two- and three-character operators are
574punctuators.  There are also six @dfn{digraphs}, which the C++ standard
575calls @dfn{alternative tokens}, which are merely alternate ways to spell
576other punctuators.  This is a second attempt to work around missing
577punctuation in obsolete systems.  It has no negative side effects,
578unlike trigraphs, but does not cover as much ground.  The digraphs and
579their corresponding normal punctuators are:
580
581@smallexample
582Digraph:        <%  %>  <:  :>  %:  %:%:
583Punctuator:      @{   @}   [   ]   #    ##
584@end smallexample
585
586@cindex other tokens
587Any other single character is considered ``other''.  It is passed on to
588the preprocessor's output unmolested.  The C compiler will almost
589certainly reject source code containing ``other'' tokens.  In ASCII, the
590only other characters are @samp{@@}, @samp{$}, @samp{`}, and control
591characters other than NUL (all bits zero).  (Note that @samp{$} is
592normally considered a letter.)  All characters with the high bit set
593(numeric range 0x7F--0xFF) are also ``other'' in the present
594implementation.  This will change when proper support for international
595character sets is added to GCC@.
596
597NUL is a special case because of the high probability that its
598appearance is accidental, and because it may be invisible to the user
599(many terminals do not display NUL at all).  Within comments, NULs are
600silently ignored, just as any other character would be.  In running
601text, NUL is considered white space.  For example, these two directives
602have the same meaning.
603
604@smallexample
605#define X^@@1
606#define X 1
607@end smallexample
608
609@noindent
610(where @samp{^@@} is ASCII NUL)@.  Within string or character constants,
611NULs are preserved.  In the latter two cases the preprocessor emits a
612warning message.
613
614@node The preprocessing language
615@section The preprocessing language
616@cindex directives
617@cindex preprocessing directives
618@cindex directive line
619@cindex directive name
620
621After tokenization, the stream of tokens may simply be passed straight
622to the compiler's parser.  However, if it contains any operations in the
623@dfn{preprocessing language}, it will be transformed first.  This stage
624corresponds roughly to the standard's ``translation phase 4'' and is
625what most people think of as the preprocessor's job.
626
627The preprocessing language consists of @dfn{directives} to be executed
628and @dfn{macros} to be expanded.  Its primary capabilities are:
629
630@itemize @bullet
631@item
632Inclusion of header files.  These are files of declarations that can be
633substituted into your program.
634
635@item
636Macro expansion.  You can define @dfn{macros}, which are abbreviations
637for arbitrary fragments of C code.  The preprocessor will replace the
638macros with their definitions throughout the program.  Some macros are
639automatically defined for you.
640
641@item
642Conditional compilation.  You can include or exclude parts of the
643program according to various conditions.
644
645@item
646Line control.  If you use a program to combine or rearrange source files
647into an intermediate file which is then compiled, you can use line
648control to inform the compiler where each source line originally came
649from.
650
651@item
652Diagnostics.  You can detect problems at compile time and issue errors
653or warnings.
654@end itemize
655
656There are a few more, less useful, features.
657
658Except for expansion of predefined macros, all these operations are
659triggered with @dfn{preprocessing directives}.  Preprocessing directives
660are lines in your program that start with @samp{#}.  Whitespace is
661allowed before and after the @samp{#}.  The @samp{#} is followed by an
662identifier, the @dfn{directive name}.  It specifies the operation to
663perform.  Directives are commonly referred to as @samp{#@var{name}}
664where @var{name} is the directive name.  For example, @samp{#define} is
665the directive that defines a macro.
666
667The @samp{#} which begins a directive cannot come from a macro
668expansion.  Also, the directive name is not macro expanded.  Thus, if
669@code{foo} is defined as a macro expanding to @code{define}, that does
670not make @samp{#foo} a valid preprocessing directive.
671
672The set of valid directive names is fixed.  Programs cannot define new
673preprocessing directives.
674
675Some directives require arguments; these make up the rest of the
676directive line and must be separated from the directive name by
677whitespace.  For example, @samp{#define} must be followed by a macro
678name and the intended expansion of the macro.
679
680A preprocessing directive cannot cover more than one line.  The line
681may, however, be continued with backslash-newline, or by a block comment
682which extends past the end of the line.  In either case, when the
683directive is processed, the continuations have already been merged with
684the first line to make one long line.
685
686@node Header Files
687@chapter Header Files
688
689@cindex header file
690A header file is a file containing C declarations and macro definitions
691(@pxref{Macros}) to be shared between several source files.  You request
692the use of a header file in your program by @dfn{including} it, with the
693C preprocessing directive @samp{#include}.
694
695Header files serve two purposes.
696
697@itemize @bullet
698@item
699@cindex system header files
700System header files declare the interfaces to parts of the operating
701system.  You include them in your program to supply the definitions and
702declarations you need to invoke system calls and libraries.
703
704@item
705Your own header files contain declarations for interfaces between the
706source files of your program.  Each time you have a group of related
707declarations and macro definitions all or most of which are needed in
708several different source files, it is a good idea to create a header
709file for them.
710@end itemize
711
712Including a header file produces the same results as copying the header
713file into each source file that needs it.  Such copying would be
714time-consuming and error-prone.  With a header file, the related
715declarations appear in only one place.  If they need to be changed, they
716can be changed in one place, and programs that include the header file
717will automatically use the new version when next recompiled.  The header
718file eliminates the labor of finding and changing all the copies as well
719as the risk that a failure to find one copy will result in
720inconsistencies within a program.
721
722In C, the usual convention is to give header files names that end with
723@file{.h}.  It is most portable to use only letters, digits, dashes, and
724underscores in header file names, and at most one dot.
725
726@menu
727* Include Syntax::
728* Include Operation::
729* Search Path::
730* Once-Only Headers::
731* Alternatives to Wrapper #ifndef::
732* Computed Includes::
733* Wrapper Headers::
734* System Headers::
735@end menu
736
737@node Include Syntax
738@section Include Syntax
739
740@findex #include
741Both user and system header files are included using the preprocessing
742directive @samp{#include}.  It has two variants:
743
744@table @code
745@item #include <@var{file}>
746This variant is used for system header files.  It searches for a file
747named @var{file} in a standard list of system directories.  You can prepend
748directories to this list with the @option{-I} option (@pxref{Invocation}).
749
750@item #include "@var{file}"
751This variant is used for header files of your own program.  It
752searches for a file named @var{file} first in the directory containing
753the current file, then in the quote directories and then the same
754directories used for @code{<@var{file}>}.  You can prepend directories
755to the list of quote directories with the @option{-iquote} option.
756@end table
757
758The argument of @samp{#include}, whether delimited with quote marks or
759angle brackets, behaves like a string constant in that comments are not
760recognized, and macro names are not expanded.  Thus, @code{@w{#include
761<x/*y>}} specifies inclusion of a system header file named @file{x/*y}.
762
763However, if backslashes occur within @var{file}, they are considered
764ordinary text characters, not escape characters.  None of the character
765escape sequences appropriate to string constants in C are processed.
766Thus, @code{@w{#include "x\n\\y"}} specifies a filename containing three
767backslashes.  (Some systems interpret @samp{\} as a pathname separator.
768All of these also interpret @samp{/} the same way.  It is most portable
769to use only @samp{/}.)
770
771It is an error if there is anything (other than comments) on the line
772after the file name.
773
774@node Include Operation
775@section Include Operation
776
777The @samp{#include} directive works by directing the C preprocessor to
778scan the specified file as input before continuing with the rest of the
779current file.  The output from the preprocessor contains the output
780already generated, followed by the output resulting from the included
781file, followed by the output that comes from the text after the
782@samp{#include} directive.  For example, if you have a header file
783@file{header.h} as follows,
784
785@smallexample
786char *test (void);
787@end smallexample
788
789@noindent
790and a main program called @file{program.c} that uses the header file,
791like this,
792
793@smallexample
794int x;
795#include "header.h"
796
797int
798main (void)
799@{
800  puts (test ());
801@}
802@end smallexample
803
804@noindent
805the compiler will see the same token stream as it would if
806@file{program.c} read
807
808@smallexample
809int x;
810char *test (void);
811
812int
813main (void)
814@{
815  puts (test ());
816@}
817@end smallexample
818
819Included files are not limited to declarations and macro definitions;
820those are merely the typical uses.  Any fragment of a C program can be
821included from another file.  The include file could even contain the
822beginning of a statement that is concluded in the containing file, or
823the end of a statement that was started in the including file.  However,
824an included file must consist of complete tokens.  Comments and string
825literals which have not been closed by the end of an included file are
826invalid.  For error recovery, they are considered to end at the end of
827the file.
828
829To avoid confusion, it is best if header files contain only complete
830syntactic units---function declarations or definitions, type
831declarations, etc.
832
833The line following the @samp{#include} directive is always treated as a
834separate line by the C preprocessor, even if the included file lacks a
835final newline.
836
837@node Search Path
838@section Search Path
839
840By default, the preprocessor looks for header files included by the quote
841form of the directive @code{@w{#include "@var{file}"}} first relative to
842the directory of the current file, and then in a preconfigured list
843of standard system directories.
844For example, if @file{/usr/include/sys/stat.h} contains
845@code{@w{#include "types.h"}}, GCC looks for @file{types.h} first in
846@file{/usr/include/sys}, then in its usual search path.
847
848For the angle-bracket form @code{@w{#include <@var{file}>}}, the
849preprocessor's default behavior is to look only in the standard system
850directories.  The exact search directory list depends on the target
851system, how GCC is configured, and where it is installed.  You can
852find the default search directory list for your version of CPP by
853invoking it with the @option{-v} option.  For example,
854
855@smallexample
856cpp -v /dev/null -o /dev/null
857@end smallexample
858
859There are a number of command-line options you can use to add
860additional directories to the search path.
861The most commonly-used option is @option{-I@var{dir}}, which causes
862@var{dir} to be searched after the current directory (for the quote
863form of the directive) and ahead of the standard system directories.
864You can specify multiple @option{-I} options on the command line,
865in which case the directories are searched in left-to-right order.
866
867If you need separate control over the search paths for the quote and
868angle-bracket forms of the @samp{#include} directive, you can use the
869@option{-iquote} and/or @option{-isystem} options instead of @option{-I}.
870@xref{Invocation}, for a detailed description of these options, as
871well as others that are less generally useful.
872
873If you specify other options on the command line, such as @option{-I},
874that affect where the preprocessor searches for header files, the
875directory list printed by the @option{-v} option reflects the actual
876search path used by the preprocessor.
877
878Note that you can also prevent the preprocessor from searching any of
879the default system header directories with the @option{-nostdinc}
880option.  This is useful when you are compiling an operating system
881kernel or some other program that does not use the standard C library
882facilities, or the standard C library itself.
883
884@node Once-Only Headers
885@section Once-Only Headers
886@cindex repeated inclusion
887@cindex including just once
888@cindex wrapper @code{#ifndef}
889
890If a header file happens to be included twice, the compiler will process
891its contents twice.  This is very likely to cause an error, e.g.@: when the
892compiler sees the same structure definition twice.  Even if it does not,
893it will certainly waste time.
894
895The standard way to prevent this is to enclose the entire real contents
896of the file in a conditional, like this:
897
898@smallexample
899@group
900/* File foo.  */
901#ifndef FILE_FOO_SEEN
902#define FILE_FOO_SEEN
903
904@var{the entire file}
905
906#endif /* !FILE_FOO_SEEN */
907@end group
908@end smallexample
909
910This construct is commonly known as a @dfn{wrapper #ifndef}.
911When the header is included again, the conditional will be false,
912because @code{FILE_FOO_SEEN} is defined.  The preprocessor will skip
913over the entire contents of the file, and the compiler will not see it
914twice.
915
916CPP optimizes even further.  It remembers when a header file has a
917wrapper @samp{#ifndef}.  If a subsequent @samp{#include} specifies that
918header, and the macro in the @samp{#ifndef} is still defined, it does
919not bother to rescan the file at all.
920
921You can put comments outside the wrapper.  They will not interfere with
922this optimization.
923
924@cindex controlling macro
925@cindex guard macro
926The macro @code{FILE_FOO_SEEN} is called the @dfn{controlling macro} or
927@dfn{guard macro}.  In a user header file, the macro name should not
928begin with @samp{_}.  In a system header file, it should begin with
929@samp{__} to avoid conflicts with user programs.  In any kind of header
930file, the macro name should contain the name of the file and some
931additional text, to avoid conflicts with other header files.
932
933@node Alternatives to Wrapper #ifndef
934@section Alternatives to Wrapper #ifndef
935
936CPP supports two more ways of indicating that a header file should be
937read only once.  Neither one is as portable as a wrapper @samp{#ifndef}
938and we recommend you do not use them in new programs, with the caveat
939that @samp{#import} is standard practice in Objective-C.
940
941@findex #import
942CPP supports a variant of @samp{#include} called @samp{#import} which
943includes a file, but does so at most once.  If you use @samp{#import}
944instead of @samp{#include}, then you don't need the conditionals
945inside the header file to prevent multiple inclusion of the contents.
946@samp{#import} is standard in Objective-C, but is considered a
947deprecated extension in C and C++.
948
949@samp{#import} is not a well designed feature.  It requires the users of
950a header file to know that it should only be included once.  It is much
951better for the header file's implementor to write the file so that users
952don't need to know this.  Using a wrapper @samp{#ifndef} accomplishes
953this goal.
954
955In the present implementation, a single use of @samp{#import} will
956prevent the file from ever being read again, by either @samp{#import} or
957@samp{#include}.  You should not rely on this; do not use both
958@samp{#import} and @samp{#include} to refer to the same header file.
959
960Another way to prevent a header file from being included more than once
961is with the @samp{#pragma once} directive.  If @samp{#pragma once} is
962seen when scanning a header file, that file will never be read again, no
963matter what.
964
965@samp{#pragma once} does not have the problems that @samp{#import} does,
966but it is not recognized by all preprocessors, so you cannot rely on it
967in a portable program.
968
969@node Computed Includes
970@section Computed Includes
971@cindex computed includes
972@cindex macros in include
973
974Sometimes it is necessary to select one of several different header
975files to be included into your program.  They might specify
976configuration parameters to be used on different sorts of operating
977systems, for instance.  You could do this with a series of conditionals,
978
979@smallexample
980#if SYSTEM_1
981# include "system_1.h"
982#elif SYSTEM_2
983# include "system_2.h"
984#elif SYSTEM_3
985@dots{}
986#endif
987@end smallexample
988
989That rapidly becomes tedious.  Instead, the preprocessor offers the
990ability to use a macro for the header name.  This is called a
991@dfn{computed include}.  Instead of writing a header name as the direct
992argument of @samp{#include}, you simply put a macro name there instead:
993
994@smallexample
995#define SYSTEM_H "system_1.h"
996@dots{}
997#include SYSTEM_H
998@end smallexample
999
1000@noindent
1001@code{SYSTEM_H} will be expanded, and the preprocessor will look for
1002@file{system_1.h} as if the @samp{#include} had been written that way
1003originally.  @code{SYSTEM_H} could be defined by your Makefile with a
1004@option{-D} option.
1005
1006You must be careful when you define the macro.  @samp{#define} saves
1007tokens, not text.  The preprocessor has no way of knowing that the macro
1008will be used as the argument of @samp{#include}, so it generates
1009ordinary tokens, not a header name.  This is unlikely to cause problems
1010if you use double-quote includes, which are close enough to string
1011constants.  If you use angle brackets, however, you may have trouble.
1012
1013The syntax of a computed include is actually a bit more general than the
1014above.  If the first non-whitespace character after @samp{#include} is
1015not @samp{"} or @samp{<}, then the entire line is macro-expanded
1016like running text would be.
1017
1018If the line expands to a single string constant, the contents of that
1019string constant are the file to be included.  CPP does not re-examine the
1020string for embedded quotes, but neither does it process backslash
1021escapes in the string.  Therefore
1022
1023@smallexample
1024#define HEADER "a\"b"
1025#include HEADER
1026@end smallexample
1027
1028@noindent
1029looks for a file named @file{a\"b}.  CPP searches for the file according
1030to the rules for double-quoted includes.
1031
1032If the line expands to a token stream beginning with a @samp{<} token
1033and including a @samp{>} token, then the tokens between the @samp{<} and
1034the first @samp{>} are combined to form the filename to be included.
1035Any whitespace between tokens is reduced to a single space; then any
1036space after the initial @samp{<} is retained, but a trailing space
1037before the closing @samp{>} is ignored.  CPP searches for the file
1038according to the rules for angle-bracket includes.
1039
1040In either case, if there are any tokens on the line after the file name,
1041an error occurs and the directive is not processed.  It is also an error
1042if the result of expansion does not match either of the two expected
1043forms.
1044
1045These rules are implementation-defined behavior according to the C
1046standard.  To minimize the risk of different compilers interpreting your
1047computed includes differently, we recommend you use only a single
1048object-like macro which expands to a string constant.  This will also
1049minimize confusion for people reading your program.
1050
1051@node Wrapper Headers
1052@section Wrapper Headers
1053@cindex wrapper headers
1054@cindex overriding a header file
1055@findex #include_next
1056
1057Sometimes it is necessary to adjust the contents of a system-provided
1058header file without editing it directly.  GCC's @command{fixincludes}
1059operation does this, for example.  One way to do that would be to create
1060a new header file with the same name and insert it in the search path
1061before the original header.  That works fine as long as you're willing
1062to replace the old header entirely.  But what if you want to refer to
1063the old header from the new one?
1064
1065You cannot simply include the old header with @samp{#include}.  That
1066will start from the beginning, and find your new header again.  If your
1067header is not protected from multiple inclusion (@pxref{Once-Only
1068Headers}), it will recurse infinitely and cause a fatal error.
1069
1070You could include the old header with an absolute pathname:
1071@smallexample
1072#include "/usr/include/old-header.h"
1073@end smallexample
1074@noindent
1075This works, but is not clean; should the system headers ever move, you
1076would have to edit the new headers to match.
1077
1078There is no way to solve this problem within the C standard, but you can
1079use the GNU extension @samp{#include_next}.  It means, ``Include the
1080@emph{next} file with this name''.  This directive works like
1081@samp{#include} except in searching for the specified file: it starts
1082searching the list of header file directories @emph{after} the directory
1083in which the current file was found.
1084
1085Suppose you specify @option{-I /usr/local/include}, and the list of
1086directories to search also includes @file{/usr/include}; and suppose
1087both directories contain @file{signal.h}.  Ordinary @code{@w{#include
1088<signal.h>}} finds the file under @file{/usr/local/include}.  If that
1089file contains @code{@w{#include_next <signal.h>}}, it starts searching
1090after that directory, and finds the file in @file{/usr/include}.
1091
1092@samp{#include_next} does not distinguish between @code{<@var{file}>}
1093and @code{"@var{file}"} inclusion, nor does it check that the file you
1094specify has the same name as the current file.  It simply looks for the
1095file named, starting with the directory in the search path after the one
1096where the current file was found.
1097
1098The use of @samp{#include_next} can lead to great confusion.  We
1099recommend it be used only when there is no other alternative.  In
1100particular, it should not be used in the headers belonging to a specific
1101program; it should be used only to make global corrections along the
1102lines of @command{fixincludes}.
1103
1104@node System Headers
1105@section System Headers
1106@cindex system header files
1107
1108The header files declaring interfaces to the operating system and
1109runtime libraries often cannot be written in strictly conforming C@.
1110Therefore, GCC gives code found in @dfn{system headers} special
1111treatment.  All warnings, other than those generated by @samp{#warning}
1112(@pxref{Diagnostics}), are suppressed while GCC is processing a system
1113header.  Macros defined in a system header are immune to a few warnings
1114wherever they are expanded.  This immunity is granted on an ad-hoc
1115basis, when we find that a warning generates lots of false positives
1116because of code in macros defined in system headers.
1117
1118Normally, only the headers found in specific directories are considered
1119system headers.  These directories are determined when GCC is compiled.
1120There are, however, two ways to make normal headers into system headers:
1121
1122@itemize @bullet
1123@item
1124Header files found in directories added to the search path with the
1125@option{-isystem} and @option{-idirafter} command-line options are
1126treated as system headers for the purposes of diagnostics.
1127
1128The @option{-cxx-isystem} command line option adds its argument to the
1129list of C++ system headers, similar to @option{-isystem} for C headers.
1130
1131@item
1132@findex #pragma GCC system_header
1133There is also a directive, @code{@w{#pragma GCC system_header}}, which
1134tells GCC to consider the rest of the current include file a system
1135header, no matter where it was found.  Code that comes before the
1136@samp{#pragma} in the file is not affected.  @code{@w{#pragma GCC
1137system_header}} has no effect in the primary source file.
1138@end itemize
1139
1140@node Macros
1141@chapter Macros
1142
1143A @dfn{macro} is a fragment of code which has been given a name.
1144Whenever the name is used, it is replaced by the contents of the macro.
1145There are two kinds of macros.  They differ mostly in what they look
1146like when they are used.  @dfn{Object-like} macros resemble data objects
1147when used, @dfn{function-like} macros resemble function calls.
1148
1149You may define any valid identifier as a macro, even if it is a C
1150keyword.  The preprocessor does not know anything about keywords.  This
1151can be useful if you wish to hide a keyword such as @code{const} from an
1152older compiler that does not understand it.  However, the preprocessor
1153operator @code{defined} (@pxref{Defined}) can never be defined as a
1154macro, and C++'s named operators (@pxref{C++ Named Operators}) cannot be
1155macros when you are compiling C++.
1156
1157@menu
1158* Object-like Macros::
1159* Function-like Macros::
1160* Macro Arguments::
1161* Stringizing::
1162* Concatenation::
1163* Variadic Macros::
1164* Predefined Macros::
1165* Undefining and Redefining Macros::
1166* Directives Within Macro Arguments::
1167* Macro Pitfalls::
1168@end menu
1169
1170@node Object-like Macros
1171@section Object-like Macros
1172@cindex object-like macro
1173@cindex symbolic constants
1174@cindex manifest constants
1175
1176An @dfn{object-like macro} is a simple identifier which will be replaced
1177by a code fragment.  It is called object-like because it looks like a
1178data object in code that uses it.  They are most commonly used to give
1179symbolic names to numeric constants.
1180
1181@findex #define
1182You create macros with the @samp{#define} directive.  @samp{#define} is
1183followed by the name of the macro and then the token sequence it should
1184be an abbreviation for, which is variously referred to as the macro's
1185@dfn{body}, @dfn{expansion} or @dfn{replacement list}.  For example,
1186
1187@smallexample
1188#define BUFFER_SIZE 1024
1189@end smallexample
1190
1191@noindent
1192defines a macro named @code{BUFFER_SIZE} as an abbreviation for the
1193token @code{1024}.  If somewhere after this @samp{#define} directive
1194there comes a C statement of the form
1195
1196@smallexample
1197foo = (char *) malloc (BUFFER_SIZE);
1198@end smallexample
1199
1200@noindent
1201then the C preprocessor will recognize and @dfn{expand} the macro
1202@code{BUFFER_SIZE}.  The C compiler will see the same tokens as it would
1203if you had written
1204
1205@smallexample
1206foo = (char *) malloc (1024);
1207@end smallexample
1208
1209By convention, macro names are written in uppercase.  Programs are
1210easier to read when it is possible to tell at a glance which names are
1211macros.
1212
1213The macro's body ends at the end of the @samp{#define} line.  You may
1214continue the definition onto multiple lines, if necessary, using
1215backslash-newline.  When the macro is expanded, however, it will all
1216come out on one line.  For example,
1217
1218@smallexample
1219#define NUMBERS 1, \
1220                2, \
1221                3
1222int x[] = @{ NUMBERS @};
1223     @expansion{} int x[] = @{ 1, 2, 3 @};
1224@end smallexample
1225
1226@noindent
1227The most common visible consequence of this is surprising line numbers
1228in error messages.
1229
1230There is no restriction on what can go in a macro body provided it
1231decomposes into valid preprocessing tokens.  Parentheses need not
1232balance, and the body need not resemble valid C code.  (If it does not,
1233you may get error messages from the C compiler when you use the macro.)
1234
1235The C preprocessor scans your program sequentially.  Macro definitions
1236take effect at the place you write them.  Therefore, the following input
1237to the C preprocessor
1238
1239@smallexample
1240foo = X;
1241#define X 4
1242bar = X;
1243@end smallexample
1244
1245@noindent
1246produces
1247
1248@smallexample
1249foo = X;
1250bar = 4;
1251@end smallexample
1252
1253When the preprocessor expands a macro name, the macro's expansion
1254replaces the macro invocation, then the expansion is examined for more
1255macros to expand.  For example,
1256
1257@smallexample
1258@group
1259#define TABLESIZE BUFSIZE
1260#define BUFSIZE 1024
1261TABLESIZE
1262     @expansion{} BUFSIZE
1263     @expansion{} 1024
1264@end group
1265@end smallexample
1266
1267@noindent
1268@code{TABLESIZE} is expanded first to produce @code{BUFSIZE}, then that
1269macro is expanded to produce the final result, @code{1024}.
1270
1271Notice that @code{BUFSIZE} was not defined when @code{TABLESIZE} was
1272defined.  The @samp{#define} for @code{TABLESIZE} uses exactly the
1273expansion you specify---in this case, @code{BUFSIZE}---and does not
1274check to see whether it too contains macro names.  Only when you
1275@emph{use} @code{TABLESIZE} is the result of its expansion scanned for
1276more macro names.
1277
1278This makes a difference if you change the definition of @code{BUFSIZE}
1279at some point in the source file.  @code{TABLESIZE}, defined as shown,
1280will always expand using the definition of @code{BUFSIZE} that is
1281currently in effect:
1282
1283@smallexample
1284#define BUFSIZE 1020
1285#define TABLESIZE BUFSIZE
1286#undef BUFSIZE
1287#define BUFSIZE 37
1288@end smallexample
1289
1290@noindent
1291Now @code{TABLESIZE} expands (in two stages) to @code{37}.
1292
1293If the expansion of a macro contains its own name, either directly or
1294via intermediate macros, it is not expanded again when the expansion is
1295examined for more macros.  This prevents infinite recursion.
1296@xref{Self-Referential Macros}, for the precise details.
1297
1298@node Function-like Macros
1299@section Function-like Macros
1300@cindex function-like macros
1301
1302You can also define macros whose use looks like a function call.  These
1303are called @dfn{function-like macros}.  To define a function-like macro,
1304you use the same @samp{#define} directive, but you put a pair of
1305parentheses immediately after the macro name.  For example,
1306
1307@smallexample
1308#define lang_init()  c_init()
1309lang_init()
1310     @expansion{} c_init()
1311@end smallexample
1312
1313A function-like macro is only expanded if its name appears with a pair
1314of parentheses after it.  If you write just the name, it is left alone.
1315This can be useful when you have a function and a macro of the same
1316name, and you wish to use the function sometimes.
1317
1318@smallexample
1319extern void foo(void);
1320#define foo() /* @r{optimized inline version} */
1321@dots{}
1322  foo();
1323  funcptr = foo;
1324@end smallexample
1325
1326Here the call to @code{foo()} will use the macro, but the function
1327pointer will get the address of the real function.  If the macro were to
1328be expanded, it would cause a syntax error.
1329
1330If you put spaces between the macro name and the parentheses in the
1331macro definition, that does not define a function-like macro, it defines
1332an object-like macro whose expansion happens to begin with a pair of
1333parentheses.
1334
1335@smallexample
1336#define lang_init ()    c_init()
1337lang_init()
1338     @expansion{} () c_init()()
1339@end smallexample
1340
1341The first two pairs of parentheses in this expansion come from the
1342macro.  The third is the pair that was originally after the macro
1343invocation.  Since @code{lang_init} is an object-like macro, it does not
1344consume those parentheses.
1345
1346@node Macro Arguments
1347@section Macro Arguments
1348@cindex arguments
1349@cindex macros with arguments
1350@cindex arguments in macro definitions
1351
1352Function-like macros can take @dfn{arguments}, just like true functions.
1353To define a macro that uses arguments, you insert @dfn{parameters}
1354between the pair of parentheses in the macro definition that make the
1355macro function-like.  The parameters must be valid C identifiers,
1356separated by commas and optionally whitespace.
1357
1358To invoke a macro that takes arguments, you write the name of the macro
1359followed by a list of @dfn{actual arguments} in parentheses, separated
1360by commas.  The invocation of the macro need not be restricted to a
1361single logical line---it can cross as many lines in the source file as
1362you wish.  The number of arguments you give must match the number of
1363parameters in the macro definition.  When the macro is expanded, each
1364use of a parameter in its body is replaced by the tokens of the
1365corresponding argument.  (You need not use all of the parameters in the
1366macro body.)
1367
1368As an example, here is a macro that computes the minimum of two numeric
1369values, as it is defined in many C programs, and some uses.
1370
1371@smallexample
1372#define min(X, Y)  ((X) < (Y) ? (X) : (Y))
1373  x = min(a, b);          @expansion{}  x = ((a) < (b) ? (a) : (b));
1374  y = min(1, 2);          @expansion{}  y = ((1) < (2) ? (1) : (2));
1375  z = min(a + 28, *p);    @expansion{}  z = ((a + 28) < (*p) ? (a + 28) : (*p));
1376@end smallexample
1377
1378@noindent
1379(In this small example you can already see several of the dangers of
1380macro arguments.  @xref{Macro Pitfalls}, for detailed explanations.)
1381
1382Leading and trailing whitespace in each argument is dropped, and all
1383whitespace between the tokens of an argument is reduced to a single
1384space.  Parentheses within each argument must balance; a comma within
1385such parentheses does not end the argument.  However, there is no
1386requirement for square brackets or braces to balance, and they do not
1387prevent a comma from separating arguments.  Thus,
1388
1389@smallexample
1390macro (array[x = y, x + 1])
1391@end smallexample
1392
1393@noindent
1394passes two arguments to @code{macro}: @code{array[x = y} and @code{x +
13951]}.  If you want to supply @code{array[x = y, x + 1]} as an argument,
1396you can write it as @code{array[(x = y, x + 1)]}, which is equivalent C
1397code.
1398
1399All arguments to a macro are completely macro-expanded before they are
1400substituted into the macro body.  After substitution, the complete text
1401is scanned again for macros to expand, including the arguments.  This rule
1402may seem strange, but it is carefully designed so you need not worry
1403about whether any function call is actually a macro invocation.  You can
1404run into trouble if you try to be too clever, though.  @xref{Argument
1405Prescan}, for detailed discussion.
1406
1407For example, @code{min (min (a, b), c)} is first expanded to
1408
1409@smallexample
1410  min (((a) < (b) ? (a) : (b)), (c))
1411@end smallexample
1412
1413@noindent
1414and then to
1415
1416@smallexample
1417@group
1418((((a) < (b) ? (a) : (b))) < (c)
1419 ? (((a) < (b) ? (a) : (b)))
1420 : (c))
1421@end group
1422@end smallexample
1423
1424@noindent
1425(Line breaks shown here for clarity would not actually be generated.)
1426
1427@cindex empty macro arguments
1428You can leave macro arguments empty; this is not an error to the
1429preprocessor (but many macros will then expand to invalid code).
1430You cannot leave out arguments entirely; if a macro takes two arguments,
1431there must be exactly one comma at the top level of its argument list.
1432Here are some silly examples using @code{min}:
1433
1434@smallexample
1435min(, b)        @expansion{} ((   ) < (b) ? (   ) : (b))
1436min(a, )        @expansion{} ((a  ) < ( ) ? (a  ) : ( ))
1437min(,)          @expansion{} ((   ) < ( ) ? (   ) : ( ))
1438min((,),)       @expansion{} (((,)) < ( ) ? ((,)) : ( ))
1439
1440min()      @error{} macro "min" requires 2 arguments, but only 1 given
1441min(,,)    @error{} macro "min" passed 3 arguments, but takes just 2
1442@end smallexample
1443
1444Whitespace is not a preprocessing token, so if a macro @code{foo} takes
1445one argument, @code{@w{foo ()}} and @code{@w{foo ( )}} both supply it an
1446empty argument.  Previous GNU preprocessor implementations and
1447documentation were incorrect on this point, insisting that a
1448function-like macro that takes a single argument be passed a space if an
1449empty argument was required.
1450
1451Macro parameters appearing inside string literals are not replaced by
1452their corresponding actual arguments.
1453
1454@smallexample
1455#define foo(x) x, "x"
1456foo(bar)        @expansion{} bar, "x"
1457@end smallexample
1458
1459@node Stringizing
1460@section Stringizing
1461@cindex stringizing
1462@cindex @samp{#} operator
1463
1464Sometimes you may want to convert a macro argument into a string
1465constant.  Parameters are not replaced inside string constants, but you
1466can use the @samp{#} preprocessing operator instead.  When a macro
1467parameter is used with a leading @samp{#}, the preprocessor replaces it
1468with the literal text of the actual argument, converted to a string
1469constant.  Unlike normal parameter replacement, the argument is not
1470macro-expanded first.  This is called @dfn{stringizing}.
1471
1472There is no way to combine an argument with surrounding text and
1473stringize it all together.  Instead, you can write a series of adjacent
1474string constants and stringized arguments.  The preprocessor
1475replaces the stringized arguments with string constants.  The C
1476compiler then combines all the adjacent string constants into one
1477long string.
1478
1479Here is an example of a macro definition that uses stringizing:
1480
1481@smallexample
1482@group
1483#define WARN_IF(EXP) \
1484do @{ if (EXP) \
1485        fprintf (stderr, "Warning: " #EXP "\n"); @} \
1486while (0)
1487WARN_IF (x == 0);
1488     @expansion{} do @{ if (x == 0)
1489           fprintf (stderr, "Warning: " "x == 0" "\n"); @} while (0);
1490@end group
1491@end smallexample
1492
1493@noindent
1494The argument for @code{EXP} is substituted once, as-is, into the
1495@code{if} statement, and once, stringized, into the argument to
1496@code{fprintf}.  If @code{x} were a macro, it would be expanded in the
1497@code{if} statement, but not in the string.
1498
1499The @code{do} and @code{while (0)} are a kludge to make it possible to
1500write @code{WARN_IF (@var{arg});}, which the resemblance of
1501@code{WARN_IF} to a function would make C programmers want to do; see
1502@ref{Swallowing the Semicolon}.
1503
1504Stringizing in C involves more than putting double-quote characters
1505around the fragment.  The preprocessor backslash-escapes the quotes
1506surrounding embedded string constants, and all backslashes within string and
1507character constants, in order to get a valid C string constant with the
1508proper contents.  Thus, stringizing @code{@w{p = "foo\n";}} results in
1509@t{@w{"p = \"foo\\n\";"}}.  However, backslashes that are not inside string
1510or character constants are not duplicated: @samp{\n} by itself
1511stringizes to @t{"\n"}.
1512
1513All leading and trailing whitespace in text being stringized is
1514ignored.  Any sequence of whitespace in the middle of the text is
1515converted to a single space in the stringized result.  Comments are
1516replaced by whitespace long before stringizing happens, so they
1517never appear in stringized text.
1518
1519There is no way to convert a macro argument into a character constant.
1520
1521If you want to stringize the result of expansion of a macro argument,
1522you have to use two levels of macros.
1523
1524@smallexample
1525#define xstr(s) str(s)
1526#define str(s) #s
1527#define foo 4
1528str (foo)
1529     @expansion{} "foo"
1530xstr (foo)
1531     @expansion{} xstr (4)
1532     @expansion{} str (4)
1533     @expansion{} "4"
1534@end smallexample
1535
1536@code{s} is stringized when it is used in @code{str}, so it is not
1537macro-expanded first.  But @code{s} is an ordinary argument to
1538@code{xstr}, so it is completely macro-expanded before @code{xstr}
1539itself is expanded (@pxref{Argument Prescan}).  Therefore, by the time
1540@code{str} gets to its argument, it has already been macro-expanded.
1541
1542@node Concatenation
1543@section Concatenation
1544@cindex concatenation
1545@cindex token pasting
1546@cindex token concatenation
1547@cindex @samp{##} operator
1548
1549It is often useful to merge two tokens into one while expanding macros.
1550This is called @dfn{token pasting} or @dfn{token concatenation}.  The
1551@samp{##} preprocessing operator performs token pasting.  When a macro
1552is expanded, the two tokens on either side of each @samp{##} operator
1553are combined into a single token, which then replaces the @samp{##} and
1554the two original tokens in the macro expansion.  Usually both will be
1555identifiers, or one will be an identifier and the other a preprocessing
1556number.  When pasted, they make a longer identifier.  This isn't the
1557only valid case.  It is also possible to concatenate two numbers (or a
1558number and a name, such as @code{1.5} and @code{e3}) into a number.
1559Also, multi-character operators such as @code{+=} can be formed by
1560token pasting.
1561
1562However, two tokens that don't together form a valid token cannot be
1563pasted together.  For example, you cannot concatenate @code{x} with
1564@code{+} in either order.  If you try, the preprocessor issues a warning
1565and emits the two tokens.  Whether it puts white space between the
1566tokens is undefined.  It is common to find unnecessary uses of @samp{##}
1567in complex macros.  If you get this warning, it is likely that you can
1568simply remove the @samp{##}.
1569
1570Both the tokens combined by @samp{##} could come from the macro body,
1571but you could just as well write them as one token in the first place.
1572Token pasting is most useful when one or both of the tokens comes from a
1573macro argument.  If either of the tokens next to an @samp{##} is a
1574parameter name, it is replaced by its actual argument before @samp{##}
1575executes.  As with stringizing, the actual argument is not
1576macro-expanded first.  If the argument is empty, that @samp{##} has no
1577effect.
1578
1579Keep in mind that the C preprocessor converts comments to whitespace
1580before macros are even considered.  Therefore, you cannot create a
1581comment by concatenating @samp{/} and @samp{*}.  You can put as much
1582whitespace between @samp{##} and its operands as you like, including
1583comments, and you can put comments in arguments that will be
1584concatenated.  However, it is an error if @samp{##} appears at either
1585end of a macro body.
1586
1587Consider a C program that interprets named commands.  There probably
1588needs to be a table of commands, perhaps an array of structures declared
1589as follows:
1590
1591@smallexample
1592@group
1593struct command
1594@{
1595  char *name;
1596  void (*function) (void);
1597@};
1598@end group
1599
1600@group
1601struct command commands[] =
1602@{
1603  @{ "quit", quit_command @},
1604  @{ "help", help_command @},
1605  @dots{}
1606@};
1607@end group
1608@end smallexample
1609
1610It would be cleaner not to have to give each command name twice, once in
1611the string constant and once in the function name.  A macro which takes the
1612name of a command as an argument can make this unnecessary.  The string
1613constant can be created with stringizing, and the function name by
1614concatenating the argument with @samp{_command}.  Here is how it is done:
1615
1616@smallexample
1617#define COMMAND(NAME)  @{ #NAME, NAME ## _command @}
1618
1619struct command commands[] =
1620@{
1621  COMMAND (quit),
1622  COMMAND (help),
1623  @dots{}
1624@};
1625@end smallexample
1626
1627@node Variadic Macros
1628@section Variadic Macros
1629@cindex variable number of arguments
1630@cindex macros with variable arguments
1631@cindex variadic macros
1632
1633A macro can be declared to accept a variable number of arguments much as
1634a function can.  The syntax for defining the macro is similar to that of
1635a function.  Here is an example:
1636
1637@smallexample
1638#define eprintf(@dots{}) fprintf (stderr, __VA_ARGS__)
1639@end smallexample
1640
1641This kind of macro is called @dfn{variadic}.  When the macro is invoked,
1642all the tokens in its argument list after the last named argument (this
1643macro has none), including any commas, become the @dfn{variable
1644argument}.  This sequence of tokens replaces the identifier
1645@code{@w{__VA_ARGS__}} in the macro body wherever it appears.  Thus, we
1646have this expansion:
1647
1648@smallexample
1649eprintf ("%s:%d: ", input_file, lineno)
1650     @expansion{}  fprintf (stderr, "%s:%d: ", input_file, lineno)
1651@end smallexample
1652
1653The variable argument is completely macro-expanded before it is inserted
1654into the macro expansion, just like an ordinary argument.  You may use
1655the @samp{#} and @samp{##} operators to stringize the variable argument
1656or to paste its leading or trailing token with another token.  (But see
1657below for an important special case for @samp{##}.)
1658
1659If your macro is complicated, you may want a more descriptive name for
1660the variable argument than @code{@w{__VA_ARGS__}}.  CPP permits
1661this, as an extension.  You may write an argument name immediately
1662before the @samp{@dots{}}; that name is used for the variable argument.
1663The @code{eprintf} macro above could be written
1664
1665@smallexample
1666#define eprintf(args@dots{}) fprintf (stderr, args)
1667@end smallexample
1668
1669@noindent
1670using this extension.  You cannot use @code{@w{__VA_ARGS__}} and this
1671extension in the same macro.
1672
1673You can have named arguments as well as variable arguments in a variadic
1674macro.  We could define @code{eprintf} like this, instead:
1675
1676@smallexample
1677#define eprintf(format, @dots{}) fprintf (stderr, format, __VA_ARGS__)
1678@end smallexample
1679
1680@noindent
1681This formulation looks more descriptive, but historically it was less
1682flexible: you had to supply at least one argument after the format
1683string.  In standard C, you could not omit the comma separating the
1684named argument from the variable arguments.  (Note that this
1685restriction has been lifted in C++2a, and never existed in GNU C; see
1686below.)
1687
1688Furthermore, if you left the variable argument empty, you would have
1689gotten a syntax error, because there would have been an extra comma
1690after the format string.
1691
1692@smallexample
1693eprintf("success!\n", );
1694     @expansion{} fprintf(stderr, "success!\n", );
1695@end smallexample
1696
1697This has been fixed in C++2a, and GNU CPP also has a pair of
1698extensions which deal with this problem.
1699
1700First, in GNU CPP, and in C++ beginning in C++2a, you are allowed to
1701leave the variable argument out entirely:
1702
1703@smallexample
1704eprintf ("success!\n")
1705     @expansion{} fprintf(stderr, "success!\n", );
1706@end smallexample
1707
1708@noindent
1709Second, C++2a introduces the @code{@w{__VA_OPT__}} function macro.
1710This macro may only appear in the definition of a variadic macro.  If
1711the variable argument has any tokens, then a @code{@w{__VA_OPT__}}
1712invocation expands to its argument; but if the variable argument does
1713not have any tokens, the @code{@w{__VA_OPT__}} expands to nothing:
1714
1715@smallexample
1716#define eprintf(format, @dots{}) \
1717  fprintf (stderr, format __VA_OPT__(,) __VA_ARGS__)
1718@end smallexample
1719
1720@code{@w{__VA_OPT__}} is also available in GNU C and GNU C++.
1721
1722Historically, GNU CPP has also had another extension to handle the
1723trailing comma: the @samp{##} token paste operator has a special
1724meaning when placed between a comma and a variable argument.  Despite
1725the introduction of @code{@w{__VA_OPT__}}, this extension remains
1726supported in GNU CPP, for backward compatibility.  If you write
1727
1728@smallexample
1729#define eprintf(format, @dots{}) fprintf (stderr, format, ##__VA_ARGS__)
1730@end smallexample
1731
1732@noindent
1733and the variable argument is left out when the @code{eprintf} macro is
1734used, then the comma before the @samp{##} will be deleted.  This does
1735@emph{not} happen if you pass an empty argument, nor does it happen if
1736the token preceding @samp{##} is anything other than a comma.
1737
1738@smallexample
1739eprintf ("success!\n")
1740     @expansion{} fprintf(stderr, "success!\n");
1741@end smallexample
1742
1743@noindent
1744The above explanation is ambiguous about the case where the only macro
1745parameter is a variable arguments parameter, as it is meaningless to
1746try to distinguish whether no argument at all is an empty argument or
1747a missing argument.
1748CPP retains the comma when conforming to a specific C
1749standard.  Otherwise the comma is dropped as an extension to the standard.
1750
1751The C standard
1752mandates that the only place the identifier @code{@w{__VA_ARGS__}}
1753can appear is in the replacement list of a variadic macro.  It may not
1754be used as a macro name, macro argument name, or within a different type
1755of macro.  It may also be forbidden in open text; the standard is
1756ambiguous.  We recommend you avoid using it except for its defined
1757purpose.
1758
1759Likewise, C++ forbids @code{@w{__VA_OPT__}} anywhere outside the
1760replacement list of a variadic macro.
1761
1762Variadic macros became a standard part of the C language with C99.
1763GNU CPP previously supported them
1764with a named variable argument
1765(@samp{args@dots{}}, not @samp{@dots{}} and @code{@w{__VA_ARGS__}}), which
1766is still supported for backward compatibility.
1767
1768@node Predefined Macros
1769@section Predefined Macros
1770
1771@cindex predefined macros
1772Several object-like macros are predefined; you use them without
1773supplying their definitions.  They fall into three classes: standard,
1774common, and system-specific.
1775
1776In C++, there is a fourth category, the named operators.  They act like
1777predefined macros, but you cannot undefine them.
1778
1779@menu
1780* Standard Predefined Macros::
1781* Common Predefined Macros::
1782* System-specific Predefined Macros::
1783* C++ Named Operators::
1784@end menu
1785
1786@node Standard Predefined Macros
1787@subsection Standard Predefined Macros
1788@cindex standard predefined macros.
1789
1790The standard predefined macros are specified by the relevant
1791language standards, so they are available with all compilers that
1792implement those standards.  Older compilers may not provide all of
1793them.  Their names all start with double underscores.
1794
1795@table @code
1796@item __FILE__
1797This macro expands to the name of the current input file, in the form of
1798a C string constant.  This is the path by which the preprocessor opened
1799the file, not the short name specified in @samp{#include} or as the
1800input file name argument.  For example,
1801@code{"/usr/local/include/myheader.h"} is a possible expansion of this
1802macro.
1803
1804@item __LINE__
1805This macro expands to the current input line number, in the form of a
1806decimal integer constant.  While we call it a predefined macro, it's
1807a pretty strange macro, since its ``definition'' changes with each
1808new line of source code.
1809@end table
1810
1811@code{__FILE__} and @code{__LINE__} are useful in generating an error
1812message to report an inconsistency detected by the program; the message
1813can state the source line at which the inconsistency was detected.  For
1814example,
1815
1816@smallexample
1817fprintf (stderr, "Internal error: "
1818                 "negative string length "
1819                 "%d at %s, line %d.",
1820         length, __FILE__, __LINE__);
1821@end smallexample
1822
1823An @samp{#include} directive changes the expansions of @code{__FILE__}
1824and @code{__LINE__} to correspond to the included file.  At the end of
1825that file, when processing resumes on the input file that contained
1826the @samp{#include} directive, the expansions of @code{__FILE__} and
1827@code{__LINE__} revert to the values they had before the
1828@samp{#include} (but @code{__LINE__} is then incremented by one as
1829processing moves to the line after the @samp{#include}).
1830
1831A @samp{#line} directive changes @code{__LINE__}, and may change
1832@code{__FILE__} as well.  @xref{Line Control}.
1833
1834C99 introduced @code{__func__}, and GCC has provided @code{__FUNCTION__}
1835for a long time.  Both of these are strings containing the name of the
1836current function (there are slight semantic differences; see the GCC
1837manual).  Neither of them is a macro; the preprocessor does not know the
1838name of the current function.  They tend to be useful in conjunction
1839with @code{__FILE__} and @code{__LINE__}, though.
1840
1841@table @code
1842
1843@item __DATE__
1844This macro expands to a string constant that describes the date on which
1845the preprocessor is being run.  The string constant contains eleven
1846characters and looks like @code{@w{"Feb 12 1996"}}.  If the day of the
1847month is less than 10, it is padded with a space on the left.
1848
1849If GCC cannot determine the current date, it will emit a warning message
1850(once per compilation) and @code{__DATE__} will expand to
1851@code{@w{"??? ?? ????"}}.
1852
1853@item __TIME__
1854This macro expands to a string constant that describes the time at
1855which the preprocessor is being run.  The string constant contains
1856eight characters and looks like @code{"23:59:01"}.
1857
1858If GCC cannot determine the current time, it will emit a warning message
1859(once per compilation) and @code{__TIME__} will expand to
1860@code{"??:??:??"}.
1861
1862@item __STDC__
1863In normal operation, this macro expands to the constant 1, to signify
1864that this compiler conforms to ISO Standard C@.  If GNU CPP is used with
1865a compiler other than GCC, this is not necessarily true; however, the
1866preprocessor always conforms to the standard unless the
1867@option{-traditional-cpp} option is used.
1868
1869This macro is not defined if the @option{-traditional-cpp} option is used.
1870
1871On some hosts, the system compiler uses a different convention, where
1872@code{__STDC__} is normally 0, but is 1 if the user specifies strict
1873conformance to the C Standard.  CPP follows the host convention when
1874processing system header files, but when processing user files
1875@code{__STDC__} is always 1.  This has been reported to cause problems;
1876for instance, some versions of Solaris provide X Windows headers that
1877expect @code{__STDC__} to be either undefined or 1.  @xref{Invocation}.
1878
1879@item __STDC_VERSION__
1880This macro expands to the C Standard's version number, a long integer
1881constant of the form @code{@var{yyyy}@var{mm}L} where @var{yyyy} and
1882@var{mm} are the year and month of the Standard version.  This signifies
1883which version of the C Standard the compiler conforms to.  Like
1884@code{__STDC__}, this is not necessarily accurate for the entire
1885implementation, unless GNU CPP is being used with GCC@.
1886
1887The value @code{199409L} signifies the 1989 C standard as amended in
18881994, which is the current default; the value @code{199901L} signifies
1889the 1999 revision of the C standard; the value @code{201112L}
1890signifies the 2011 revision of the C standard; the value
1891@code{201710L} signifies the 2017 revision of the C standard (which is
1892otherwise identical to the 2011 version apart from correction of
1893defects).
1894
1895This macro is not defined if the @option{-traditional-cpp} option is
1896used, nor when compiling C++ or Objective-C@.
1897
1898@item __STDC_HOSTED__
1899This macro is defined, with value 1, if the compiler's target is a
1900@dfn{hosted environment}.  A hosted environment has the complete
1901facilities of the standard C library available.
1902
1903@item __cplusplus
1904This macro is defined when the C++ compiler is in use.  You can use
1905@code{__cplusplus} to test whether a header is compiled by a C compiler
1906or a C++ compiler.  This macro is similar to @code{__STDC_VERSION__}, in
1907that it expands to a version number.  Depending on the language standard
1908selected, the value of the macro is
1909@code{199711L} for the 1998 C++ standard,
1910@code{201103L} for the 2011 C++ standard,
1911@code{201402L} for the 2014 C++ standard,
1912@code{201703L} for the 2017 C++ standard,
1913or an unspecified value strictly larger than @code{201703L} for the
1914experimental languages enabled by @option{-std=c++2a} and
1915@option{-std=gnu++2a}.
1916
1917@item __OBJC__
1918This macro is defined, with value 1, when the Objective-C compiler is in
1919use.  You can use @code{__OBJC__} to test whether a header is compiled
1920by a C compiler or an Objective-C compiler.
1921
1922@item __ASSEMBLER__
1923This macro is defined with value 1 when preprocessing assembly
1924language.
1925
1926@end table
1927
1928@node Common Predefined Macros
1929@subsection Common Predefined Macros
1930@cindex common predefined macros
1931
1932The common predefined macros are GNU C extensions.  They are available
1933with the same meanings regardless of the machine or operating system on
1934which you are using GNU C or GNU Fortran.  Their names all start with
1935double underscores.
1936
1937@table @code
1938
1939@item __COUNTER__
1940This macro expands to sequential integral values starting from 0.  In
1941conjunction with the @code{##} operator, this provides a convenient means to
1942generate unique identifiers.  Care must be taken to ensure that
1943@code{__COUNTER__} is not expanded prior to inclusion of precompiled headers
1944which use it.  Otherwise, the precompiled headers will not be used.
1945
1946@item __GFORTRAN__
1947The GNU Fortran compiler defines this.
1948
1949@item __GNUC__
1950@itemx __GNUC_MINOR__
1951@itemx __GNUC_PATCHLEVEL__
1952These macros are defined by all GNU compilers that use the C
1953preprocessor: C, C++, Objective-C and Fortran.  Their values are the major
1954version, minor version, and patch level of the compiler, as integer
1955constants.  For example, GCC version @var{x}.@var{y}.@var{z}
1956defines @code{__GNUC__} to @var{x}, @code{__GNUC_MINOR__} to @var{y},
1957and @code{__GNUC_PATCHLEVEL__} to @var{z}.  These
1958macros are also defined if you invoke the preprocessor directly.
1959
1960If all you need to know is whether or not your program is being compiled
1961by GCC, or a non-GCC compiler that claims to accept the GNU C dialects,
1962you can simply test @code{__GNUC__}.  If you need to write code
1963which depends on a specific version, you must be more careful.  Each
1964time the minor version is increased, the patch level is reset to zero;
1965each time the major version is increased, the
1966minor version and patch level are reset.  If you wish to use the
1967predefined macros directly in the conditional, you will need to write it
1968like this:
1969
1970@smallexample
1971/* @r{Test for GCC > 3.2.0} */
1972#if __GNUC__ > 3 || \
1973    (__GNUC__ == 3 && (__GNUC_MINOR__ > 2 || \
1974                       (__GNUC_MINOR__ == 2 && \
1975                        __GNUC_PATCHLEVEL__ > 0))
1976@end smallexample
1977
1978@noindent
1979Another approach is to use the predefined macros to
1980calculate a single number, then compare that against a threshold:
1981
1982@smallexample
1983#define GCC_VERSION (__GNUC__ * 10000 \
1984                     + __GNUC_MINOR__ * 100 \
1985                     + __GNUC_PATCHLEVEL__)
1986@dots{}
1987/* @r{Test for GCC > 3.2.0} */
1988#if GCC_VERSION > 30200
1989@end smallexample
1990
1991@noindent
1992Many people find this form easier to understand.
1993
1994@item __GNUG__
1995The GNU C++ compiler defines this.  Testing it is equivalent to
1996testing @code{@w{(__GNUC__ && __cplusplus)}}.
1997
1998@item __STRICT_ANSI__
1999GCC defines this macro if and only if the @option{-ansi} switch, or a
2000@option{-std} switch specifying strict conformance to some version of ISO C
2001or ISO C++, was specified when GCC was invoked.  It is defined to @samp{1}.
2002This macro exists primarily to direct GNU libc's header files to use only
2003definitions found in standard C.
2004
2005@item __BASE_FILE__
2006This macro expands to the name of the main input file, in the form
2007of a C string constant.  This is the source file that was specified
2008on the command line of the preprocessor or C compiler.
2009
2010@item __INCLUDE_LEVEL__
2011This macro expands to a decimal integer constant that represents the
2012depth of nesting in include files.  The value of this macro is
2013incremented on every @samp{#include} directive and decremented at the
2014end of every included file.  It starts out at 0, its value within the
2015base file specified on the command line.
2016
2017@item __ELF__
2018This macro is defined if the target uses the ELF object format.
2019
2020@item __VERSION__
2021This macro expands to a string constant which describes the version of
2022the compiler in use.  You should not rely on its contents having any
2023particular form, but it can be counted on to contain at least the
2024release number.
2025
2026@item __OPTIMIZE__
2027@itemx __OPTIMIZE_SIZE__
2028@itemx __NO_INLINE__
2029These macros describe the compilation mode.  @code{__OPTIMIZE__} is
2030defined in all optimizing compilations.  @code{__OPTIMIZE_SIZE__} is
2031defined if the compiler is optimizing for size, not speed.
2032@code{__NO_INLINE__} is defined if no functions will be inlined into
2033their callers (when not optimizing, or when inlining has been
2034specifically disabled by @option{-fno-inline}).
2035
2036These macros cause certain GNU header files to provide optimized
2037definitions, using macros or inline functions, of system library
2038functions.  You should not use these macros in any way unless you make
2039sure that programs will execute with the same effect whether or not they
2040are defined.  If they are defined, their value is 1.
2041
2042@item __GNUC_GNU_INLINE__
2043GCC defines this macro if functions declared @code{inline} will be
2044handled in GCC's traditional gnu90 mode.  Object files will contain
2045externally visible definitions of all functions declared @code{inline}
2046without @code{extern} or @code{static}.  They will not contain any
2047definitions of any functions declared @code{extern inline}.
2048
2049@item __GNUC_STDC_INLINE__
2050GCC defines this macro if functions declared @code{inline} will be
2051handled according to the ISO C99 or later standards.  Object files will contain
2052externally visible definitions of all functions declared @code{extern
2053inline}.  They will not contain definitions of any functions declared
2054@code{inline} without @code{extern}.
2055
2056If this macro is defined, GCC supports the @code{gnu_inline} function
2057attribute as a way to always get the gnu90 behavior.
2058
2059@item __CHAR_UNSIGNED__
2060GCC defines this macro if and only if the data type @code{char} is
2061unsigned on the target machine.  It exists to cause the standard header
2062file @file{limits.h} to work correctly.  You should not use this macro
2063yourself; instead, refer to the standard macros defined in @file{limits.h}.
2064
2065@item __WCHAR_UNSIGNED__
2066Like @code{__CHAR_UNSIGNED__}, this macro is defined if and only if the
2067data type @code{wchar_t} is unsigned and the front-end is in C++ mode.
2068
2069@item __REGISTER_PREFIX__
2070This macro expands to a single token (not a string constant) which is
2071the prefix applied to CPU register names in assembly language for this
2072target.  You can use it to write assembly that is usable in multiple
2073environments.  For example, in the @code{m68k-aout} environment it
2074expands to nothing, but in the @code{m68k-coff} environment it expands
2075to a single @samp{%}.
2076
2077@item __USER_LABEL_PREFIX__
2078This macro expands to a single token which is the prefix applied to
2079user labels (symbols visible to C code) in assembly.  For example, in
2080the @code{m68k-aout} environment it expands to an @samp{_}, but in the
2081@code{m68k-coff} environment it expands to nothing.
2082
2083This macro will have the correct definition even if
2084@option{-f(no-)underscores} is in use, but it will not be correct if
2085target-specific options that adjust this prefix are used (e.g.@: the
2086OSF/rose @option{-mno-underscores} option).
2087
2088@item __SIZE_TYPE__
2089@itemx __PTRDIFF_TYPE__
2090@itemx __WCHAR_TYPE__
2091@itemx __WINT_TYPE__
2092@itemx __INTMAX_TYPE__
2093@itemx __UINTMAX_TYPE__
2094@itemx __SIG_ATOMIC_TYPE__
2095@itemx __INT8_TYPE__
2096@itemx __INT16_TYPE__
2097@itemx __INT32_TYPE__
2098@itemx __INT64_TYPE__
2099@itemx __UINT8_TYPE__
2100@itemx __UINT16_TYPE__
2101@itemx __UINT32_TYPE__
2102@itemx __UINT64_TYPE__
2103@itemx __INT_LEAST8_TYPE__
2104@itemx __INT_LEAST16_TYPE__
2105@itemx __INT_LEAST32_TYPE__
2106@itemx __INT_LEAST64_TYPE__
2107@itemx __UINT_LEAST8_TYPE__
2108@itemx __UINT_LEAST16_TYPE__
2109@itemx __UINT_LEAST32_TYPE__
2110@itemx __UINT_LEAST64_TYPE__
2111@itemx __INT_FAST8_TYPE__
2112@itemx __INT_FAST16_TYPE__
2113@itemx __INT_FAST32_TYPE__
2114@itemx __INT_FAST64_TYPE__
2115@itemx __UINT_FAST8_TYPE__
2116@itemx __UINT_FAST16_TYPE__
2117@itemx __UINT_FAST32_TYPE__
2118@itemx __UINT_FAST64_TYPE__
2119@itemx __INTPTR_TYPE__
2120@itemx __UINTPTR_TYPE__
2121These macros are defined to the correct underlying types for the
2122@code{size_t}, @code{ptrdiff_t}, @code{wchar_t}, @code{wint_t},
2123@code{intmax_t}, @code{uintmax_t}, @code{sig_atomic_t}, @code{int8_t},
2124@code{int16_t}, @code{int32_t}, @code{int64_t}, @code{uint8_t},
2125@code{uint16_t}, @code{uint32_t}, @code{uint64_t},
2126@code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
2127@code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
2128@code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
2129@code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
2130@code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
2131@code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t} typedefs,
2132respectively.  They exist to make the standard header files
2133@file{stddef.h}, @file{stdint.h}, and @file{wchar.h} work correctly.
2134You should not use these macros directly; instead, include the
2135appropriate headers and use the typedefs.  Some of these macros may
2136not be defined on particular systems if GCC does not provide a
2137@file{stdint.h} header on those systems.
2138
2139@item __CHAR_BIT__
2140Defined to the number of bits used in the representation of the
2141@code{char} data type.  It exists to make the standard header given
2142numerical limits work correctly.  You should not use
2143this macro directly; instead, include the appropriate headers.
2144
2145@item __SCHAR_MAX__
2146@itemx __WCHAR_MAX__
2147@itemx __SHRT_MAX__
2148@itemx __INT_MAX__
2149@itemx __LONG_MAX__
2150@itemx __LONG_LONG_MAX__
2151@itemx __WINT_MAX__
2152@itemx __SIZE_MAX__
2153@itemx __PTRDIFF_MAX__
2154@itemx __INTMAX_MAX__
2155@itemx __UINTMAX_MAX__
2156@itemx __SIG_ATOMIC_MAX__
2157@itemx __INT8_MAX__
2158@itemx __INT16_MAX__
2159@itemx __INT32_MAX__
2160@itemx __INT64_MAX__
2161@itemx __UINT8_MAX__
2162@itemx __UINT16_MAX__
2163@itemx __UINT32_MAX__
2164@itemx __UINT64_MAX__
2165@itemx __INT_LEAST8_MAX__
2166@itemx __INT_LEAST16_MAX__
2167@itemx __INT_LEAST32_MAX__
2168@itemx __INT_LEAST64_MAX__
2169@itemx __UINT_LEAST8_MAX__
2170@itemx __UINT_LEAST16_MAX__
2171@itemx __UINT_LEAST32_MAX__
2172@itemx __UINT_LEAST64_MAX__
2173@itemx __INT_FAST8_MAX__
2174@itemx __INT_FAST16_MAX__
2175@itemx __INT_FAST32_MAX__
2176@itemx __INT_FAST64_MAX__
2177@itemx __UINT_FAST8_MAX__
2178@itemx __UINT_FAST16_MAX__
2179@itemx __UINT_FAST32_MAX__
2180@itemx __UINT_FAST64_MAX__
2181@itemx __INTPTR_MAX__
2182@itemx __UINTPTR_MAX__
2183@itemx __WCHAR_MIN__
2184@itemx __WINT_MIN__
2185@itemx __SIG_ATOMIC_MIN__
2186Defined to the maximum value of the @code{signed char}, @code{wchar_t},
2187@code{signed short},
2188@code{signed int}, @code{signed long}, @code{signed long long},
2189@code{wint_t}, @code{size_t}, @code{ptrdiff_t},
2190@code{intmax_t}, @code{uintmax_t}, @code{sig_atomic_t}, @code{int8_t},
2191@code{int16_t}, @code{int32_t}, @code{int64_t}, @code{uint8_t},
2192@code{uint16_t}, @code{uint32_t}, @code{uint64_t},
2193@code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
2194@code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
2195@code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
2196@code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
2197@code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
2198@code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t} types and
2199to the minimum value of the @code{wchar_t}, @code{wint_t}, and
2200@code{sig_atomic_t} types respectively.  They exist to make the
2201standard header given numerical limits work correctly.  You should not
2202use these macros directly; instead, include the appropriate headers.
2203Some of these macros may not be defined on particular systems if GCC
2204does not provide a @file{stdint.h} header on those systems.
2205
2206@item __INT8_C
2207@itemx __INT16_C
2208@itemx __INT32_C
2209@itemx __INT64_C
2210@itemx __UINT8_C
2211@itemx __UINT16_C
2212@itemx __UINT32_C
2213@itemx __UINT64_C
2214@itemx __INTMAX_C
2215@itemx __UINTMAX_C
2216Defined to implementations of the standard @file{stdint.h} macros with
2217the same names without the leading @code{__}.  They exist the make the
2218implementation of that header work correctly.  You should not use
2219these macros directly; instead, include the appropriate headers.  Some
2220of these macros may not be defined on particular systems if GCC does
2221not provide a @file{stdint.h} header on those systems.
2222
2223@item __SCHAR_WIDTH__
2224@itemx __SHRT_WIDTH__
2225@itemx __INT_WIDTH__
2226@itemx __LONG_WIDTH__
2227@itemx __LONG_LONG_WIDTH__
2228@itemx __PTRDIFF_WIDTH__
2229@itemx __SIG_ATOMIC_WIDTH__
2230@itemx __SIZE_WIDTH__
2231@itemx __WCHAR_WIDTH__
2232@itemx __WINT_WIDTH__
2233@itemx __INT_LEAST8_WIDTH__
2234@itemx __INT_LEAST16_WIDTH__
2235@itemx __INT_LEAST32_WIDTH__
2236@itemx __INT_LEAST64_WIDTH__
2237@itemx __INT_FAST8_WIDTH__
2238@itemx __INT_FAST16_WIDTH__
2239@itemx __INT_FAST32_WIDTH__
2240@itemx __INT_FAST64_WIDTH__
2241@itemx __INTPTR_WIDTH__
2242@itemx __INTMAX_WIDTH__
2243Defined to the bit widths of the corresponding types.  They exist to
2244make the implementations of @file{limits.h} and @file{stdint.h} behave
2245correctly.  You should not use these macros directly; instead, include
2246the appropriate headers.  Some of these macros may not be defined on
2247particular systems if GCC does not provide a @file{stdint.h} header on
2248those systems.
2249
2250@item __SIZEOF_INT__
2251@itemx __SIZEOF_LONG__
2252@itemx __SIZEOF_LONG_LONG__
2253@itemx __SIZEOF_SHORT__
2254@itemx __SIZEOF_POINTER__
2255@itemx __SIZEOF_FLOAT__
2256@itemx __SIZEOF_DOUBLE__
2257@itemx __SIZEOF_LONG_DOUBLE__
2258@itemx __SIZEOF_SIZE_T__
2259@itemx __SIZEOF_WCHAR_T__
2260@itemx __SIZEOF_WINT_T__
2261@itemx __SIZEOF_PTRDIFF_T__
2262Defined to the number of bytes of the C standard data types: @code{int},
2263@code{long}, @code{long long}, @code{short}, @code{void *}, @code{float},
2264@code{double}, @code{long double}, @code{size_t}, @code{wchar_t}, @code{wint_t}
2265and @code{ptrdiff_t}.
2266
2267@item __BYTE_ORDER__
2268@itemx __ORDER_LITTLE_ENDIAN__
2269@itemx __ORDER_BIG_ENDIAN__
2270@itemx __ORDER_PDP_ENDIAN__
2271@code{__BYTE_ORDER__} is defined to one of the values
2272@code{__ORDER_LITTLE_ENDIAN__}, @code{__ORDER_BIG_ENDIAN__}, or
2273@code{__ORDER_PDP_ENDIAN__} to reflect the layout of multi-byte and
2274multi-word quantities in memory.  If @code{__BYTE_ORDER__} is equal to
2275@code{__ORDER_LITTLE_ENDIAN__} or @code{__ORDER_BIG_ENDIAN__}, then
2276multi-byte and multi-word quantities are laid out identically: the
2277byte (word) at the lowest address is the least significant or most
2278significant byte (word) of the quantity, respectively.  If
2279@code{__BYTE_ORDER__} is equal to @code{__ORDER_PDP_ENDIAN__}, then
2280bytes in 16-bit words are laid out in a little-endian fashion, whereas
2281the 16-bit subwords of a 32-bit quantity are laid out in big-endian
2282fashion.
2283
2284You should use these macros for testing like this:
2285
2286@smallexample
2287/* @r{Test for a little-endian machine} */
2288#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
2289@end smallexample
2290
2291@item __FLOAT_WORD_ORDER__
2292@code{__FLOAT_WORD_ORDER__} is defined to one of the values
2293@code{__ORDER_LITTLE_ENDIAN__} or @code{__ORDER_BIG_ENDIAN__} to reflect
2294the layout of the words of multi-word floating-point quantities.
2295
2296@item __DEPRECATED
2297This macro is defined, with value 1, when compiling a C++ source file
2298with warnings about deprecated constructs enabled.  These warnings are
2299enabled by default, but can be disabled with @option{-Wno-deprecated}.
2300
2301@item __EXCEPTIONS
2302This macro is defined, with value 1, when compiling a C++ source file
2303with exceptions enabled.  If @option{-fno-exceptions} is used when
2304compiling the file, then this macro is not defined.
2305
2306@item __GXX_RTTI
2307This macro is defined, with value 1, when compiling a C++ source file
2308with runtime type identification enabled.  If @option{-fno-rtti} is
2309used when compiling the file, then this macro is not defined.
2310
2311@item __USING_SJLJ_EXCEPTIONS__
2312This macro is defined, with value 1, if the compiler uses the old
2313mechanism based on @code{setjmp} and @code{longjmp} for exception
2314handling.
2315
2316@item __GXX_EXPERIMENTAL_CXX0X__
2317This macro is defined when compiling a C++ source file with the option
2318@option{-std=c++0x} or @option{-std=gnu++0x}. It indicates that some
2319features likely to be included in C++0x are available. Note that these
2320features are experimental, and may change or be removed in future
2321versions of GCC.
2322
2323@item __GXX_WEAK__
2324This macro is defined when compiling a C++ source file.  It has the
2325value 1 if the compiler will use weak symbols, COMDAT sections, or
2326other similar techniques to collapse symbols with ``vague linkage''
2327that are defined in multiple translation units.  If the compiler will
2328not collapse such symbols, this macro is defined with value 0.  In
2329general, user code should not need to make use of this macro; the
2330purpose of this macro is to ease implementation of the C++ runtime
2331library provided with G++.
2332
2333@item __NEXT_RUNTIME__
2334This macro is defined, with value 1, if (and only if) the NeXT runtime
2335(as in @option{-fnext-runtime}) is in use for Objective-C@.  If the GNU
2336runtime is used, this macro is not defined, so that you can use this
2337macro to determine which runtime (NeXT or GNU) is being used.
2338
2339@item __LP64__
2340@itemx _LP64
2341These macros are defined, with value 1, if (and only if) the compilation
2342is for a target where @code{long int} and pointer both use 64-bits and
2343@code{int} uses 32-bit.
2344
2345@item __SSP__
2346This macro is defined, with value 1, when @option{-fstack-protector} is in
2347use.
2348
2349@item __SSP_ALL__
2350This macro is defined, with value 2, when @option{-fstack-protector-all} is
2351in use.
2352
2353@item __SSP_STRONG__
2354This macro is defined, with value 3, when @option{-fstack-protector-strong} is
2355in use.
2356
2357@item __SSP_EXPLICIT__
2358This macro is defined, with value 4, when @option{-fstack-protector-explicit} is
2359in use.
2360
2361@item __SANITIZE_ADDRESS__
2362This macro is defined, with value 1, when @option{-fsanitize=address}
2363or @option{-fsanitize=kernel-address} are in use.
2364
2365@item __SANITIZE_THREAD__
2366This macro is defined, with value 1, when @option{-fsanitize=thread} is in use.
2367
2368@item __TIMESTAMP__
2369This macro expands to a string constant that describes the date and time
2370of the last modification of the current source file. The string constant
2371contains abbreviated day of the week, month, day of the month, time in
2372hh:mm:ss form, year and looks like @code{@w{"Sun Sep 16 01:03:52 1973"}}.
2373If the day of the month is less than 10, it is padded with a space on the left.
2374
2375If GCC cannot determine the current date, it will emit a warning message
2376(once per compilation) and @code{__TIMESTAMP__} will expand to
2377@code{@w{"??? ??? ?? ??:??:?? ????"}}.
2378
2379@item __GCC_HAVE_SYNC_COMPARE_AND_SWAP_1
2380@itemx __GCC_HAVE_SYNC_COMPARE_AND_SWAP_2
2381@itemx __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4
2382@itemx __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8
2383@itemx __GCC_HAVE_SYNC_COMPARE_AND_SWAP_16
2384These macros are defined when the target processor supports atomic compare
2385and swap operations on operands 1, 2, 4, 8 or 16 bytes in length, respectively.
2386
2387@item __GCC_HAVE_DWARF2_CFI_ASM
2388This macro is defined when the compiler is emitting DWARF CFI directives
2389to the assembler.  When this is defined, it is possible to emit those same
2390directives in inline assembly.
2391
2392@item __FP_FAST_FMA
2393@itemx __FP_FAST_FMAF
2394@itemx __FP_FAST_FMAL
2395These macros are defined with value 1 if the backend supports the
2396@code{fma}, @code{fmaf}, and @code{fmal} builtin functions, so that
2397the include file @file{math.h} can define the macros
2398@code{FP_FAST_FMA}, @code{FP_FAST_FMAF}, and @code{FP_FAST_FMAL}
2399for compatibility with the 1999 C standard.
2400
2401@item __FP_FAST_FMAF16
2402@itemx __FP_FAST_FMAF32
2403@itemx __FP_FAST_FMAF64
2404@itemx __FP_FAST_FMAF128
2405@itemx __FP_FAST_FMAF32X
2406@itemx __FP_FAST_FMAF64X
2407@itemx __FP_FAST_FMAF128X
2408These macros are defined with the value 1 if the backend supports the
2409@code{fma} functions using the additional @code{_Float@var{n}} and
2410@code{_Float@var{n}x} types that are defined in ISO/IEC TS
241118661-3:2015.  The include file @file{math.h} can define the
2412@code{FP_FAST_FMAF@var{n}} and @code{FP_FAST_FMAF@var{n}x} macros if
2413the user defined @code{__STDC_WANT_IEC_60559_TYPES_EXT__} before
2414including @file{math.h}.
2415
2416@item __GCC_IEC_559
2417This macro is defined to indicate the intended level of support for
2418IEEE 754 (IEC 60559) floating-point arithmetic.  It expands to a
2419nonnegative integer value.  If 0, it indicates that the combination of
2420the compiler configuration and the command-line options is not
2421intended to support IEEE 754 arithmetic for @code{float} and
2422@code{double} as defined in C99 and C11 Annex F (for example, that the
2423standard rounding modes and exceptions are not supported, or that
2424optimizations are enabled that conflict with IEEE 754 semantics).  If
24251, it indicates that IEEE 754 arithmetic is intended to be supported;
2426this does not mean that all relevant language features are supported
2427by GCC.  If 2 or more, it additionally indicates support for IEEE
2428754-2008 (in particular, that the binary encodings for quiet and
2429signaling NaNs are as specified in IEEE 754-2008).
2430
2431This macro does not indicate the default state of command-line options
2432that control optimizations that C99 and C11 permit to be controlled by
2433standard pragmas, where those standards do not require a particular
2434default state.  It does not indicate whether optimizations respect
2435signaling NaN semantics (the macro for that is
2436@code{__SUPPORT_SNAN__}).  It does not indicate support for decimal
2437floating point or the IEEE 754 binary16 and binary128 types.
2438
2439@item __GCC_IEC_559_COMPLEX
2440This macro is defined to indicate the intended level of support for
2441IEEE 754 (IEC 60559) floating-point arithmetic for complex numbers, as
2442defined in C99 and C11 Annex G.  It expands to a nonnegative integer
2443value.  If 0, it indicates that the combination of the compiler
2444configuration and the command-line options is not intended to support
2445Annex G requirements (for example, because @option{-fcx-limited-range}
2446was used).  If 1 or more, it indicates that it is intended to support
2447those requirements; this does not mean that all relevant language
2448features are supported by GCC.
2449
2450@item __NO_MATH_ERRNO__
2451This macro is defined if @option{-fno-math-errno} is used, or enabled
2452by another option such as @option{-ffast-math} or by default.
2453@end table
2454
2455@node System-specific Predefined Macros
2456@subsection System-specific Predefined Macros
2457
2458@cindex system-specific predefined macros
2459@cindex predefined macros, system-specific
2460@cindex reserved namespace
2461
2462The C preprocessor normally predefines several macros that indicate what
2463type of system and machine is in use.  They are obviously different on
2464each target supported by GCC@.  This manual, being for all systems and
2465machines, cannot tell you what their names are, but you can use
2466@command{cpp -dM} to see them all.  @xref{Invocation}.  All system-specific
2467predefined macros expand to a constant value, so you can test them with
2468either @samp{#ifdef} or @samp{#if}.
2469
2470The C standard requires that all system-specific macros be part of the
2471@dfn{reserved namespace}.  All names which begin with two underscores,
2472or an underscore and a capital letter, are reserved for the compiler and
2473library to use as they wish.  However, historically system-specific
2474macros have had names with no special prefix; for instance, it is common
2475to find @code{unix} defined on Unix systems.  For all such macros, GCC
2476provides a parallel macro with two underscores added at the beginning
2477and the end.  If @code{unix} is defined, @code{__unix__} will be defined
2478too.  There will never be more than two underscores; the parallel of
2479@code{_mips} is @code{__mips__}.
2480
2481When the @option{-ansi} option, or any @option{-std} option that
2482requests strict conformance, is given to the compiler, all the
2483system-specific predefined macros outside the reserved namespace are
2484suppressed.  The parallel macros, inside the reserved namespace, remain
2485defined.
2486
2487We are slowly phasing out all predefined macros which are outside the
2488reserved namespace.  You should never use them in new programs, and we
2489encourage you to correct older code to use the parallel macros whenever
2490you find it.  We don't recommend you use the system-specific macros that
2491are in the reserved namespace, either.  It is better in the long run to
2492check specifically for features you need, using a tool such as
2493@command{autoconf}.
2494
2495@node C++ Named Operators
2496@subsection C++ Named Operators
2497@cindex named operators
2498@cindex C++ named operators
2499@cindex @file{iso646.h}
2500
2501In C++, there are eleven keywords which are simply alternate spellings
2502of operators normally written with punctuation.  These keywords are
2503treated as such even in the preprocessor.  They function as operators in
2504@samp{#if}, and they cannot be defined as macros or poisoned.  In C, you
2505can request that those keywords take their C++ meaning by including
2506@file{iso646.h}.  That header defines each one as a normal object-like
2507macro expanding to the appropriate punctuator.
2508
2509These are the named operators and their corresponding punctuators:
2510
2511@multitable {Named Operator} {Punctuator}
2512@item Named Operator @tab Punctuator
2513@item @code{and}    @tab @code{&&}
2514@item @code{and_eq} @tab @code{&=}
2515@item @code{bitand} @tab @code{&}
2516@item @code{bitor}  @tab @code{|}
2517@item @code{compl}  @tab @code{~}
2518@item @code{not}    @tab @code{!}
2519@item @code{not_eq} @tab @code{!=}
2520@item @code{or}     @tab @code{||}
2521@item @code{or_eq}  @tab @code{|=}
2522@item @code{xor}    @tab @code{^}
2523@item @code{xor_eq} @tab @code{^=}
2524@end multitable
2525
2526@node Undefining and Redefining Macros
2527@section Undefining and Redefining Macros
2528@cindex undefining macros
2529@cindex redefining macros
2530@findex #undef
2531
2532If a macro ceases to be useful, it may be @dfn{undefined} with the
2533@samp{#undef} directive.  @samp{#undef} takes a single argument, the
2534name of the macro to undefine.  You use the bare macro name, even if the
2535macro is function-like.  It is an error if anything appears on the line
2536after the macro name.  @samp{#undef} has no effect if the name is not a
2537macro.
2538
2539@smallexample
2540#define FOO 4
2541x = FOO;        @expansion{} x = 4;
2542#undef FOO
2543x = FOO;        @expansion{} x = FOO;
2544@end smallexample
2545
2546Once a macro has been undefined, that identifier may be @dfn{redefined}
2547as a macro by a subsequent @samp{#define} directive.  The new definition
2548need not have any resemblance to the old definition.
2549
2550However, if an identifier which is currently a macro is redefined, then
2551the new definition must be @dfn{effectively the same} as the old one.
2552Two macro definitions are effectively the same if:
2553@itemize @bullet
2554@item Both are the same type of macro (object- or function-like).
2555@item All the tokens of the replacement list are the same.
2556@item If there are any parameters, they are the same.
2557@item Whitespace appears in the same places in both.  It need not be
2558exactly the same amount of whitespace, though.  Remember that comments
2559count as whitespace.
2560@end itemize
2561
2562@noindent
2563These definitions are effectively the same:
2564@smallexample
2565#define FOUR (2 + 2)
2566#define FOUR         (2    +    2)
2567#define FOUR (2 /* @r{two} */ + 2)
2568@end smallexample
2569@noindent
2570but these are not:
2571@smallexample
2572#define FOUR (2 + 2)
2573#define FOUR ( 2+2 )
2574#define FOUR (2 * 2)
2575#define FOUR(score,and,seven,years,ago) (2 + 2)
2576@end smallexample
2577
2578If a macro is redefined with a definition that is not effectively the
2579same as the old one, the preprocessor issues a warning and changes the
2580macro to use the new definition.  If the new definition is effectively
2581the same, the redefinition is silently ignored.  This allows, for
2582instance, two different headers to define a common macro.  The
2583preprocessor will only complain if the definitions do not match.
2584
2585@node Directives Within Macro Arguments
2586@section Directives Within Macro Arguments
2587@cindex macro arguments and directives
2588
2589Occasionally it is convenient to use preprocessor directives within
2590the arguments of a macro.  The C and C++ standards declare that
2591behavior in these cases is undefined.  GNU CPP
2592processes arbitrary directives within macro arguments in
2593exactly the same way as it would have processed the directive were the
2594function-like macro invocation not present.
2595
2596If, within a macro invocation, that macro is redefined, then the new
2597definition takes effect in time for argument pre-expansion, but the
2598original definition is still used for argument replacement.  Here is a
2599pathological example:
2600
2601@smallexample
2602#define f(x) x x
2603f (1
2604#undef f
2605#define f 2
2606f)
2607@end smallexample
2608
2609@noindent
2610which expands to
2611
2612@smallexample
26131 2 1 2
2614@end smallexample
2615
2616@noindent
2617with the semantics described above.
2618
2619@node Macro Pitfalls
2620@section Macro Pitfalls
2621@cindex problems with macros
2622@cindex pitfalls of macros
2623
2624In this section we describe some special rules that apply to macros and
2625macro expansion, and point out certain cases in which the rules have
2626counter-intuitive consequences that you must watch out for.
2627
2628@menu
2629* Misnesting::
2630* Operator Precedence Problems::
2631* Swallowing the Semicolon::
2632* Duplication of Side Effects::
2633* Self-Referential Macros::
2634* Argument Prescan::
2635* Newlines in Arguments::
2636@end menu
2637
2638@node Misnesting
2639@subsection Misnesting
2640
2641When a macro is called with arguments, the arguments are substituted
2642into the macro body and the result is checked, together with the rest of
2643the input file, for more macro calls.  It is possible to piece together
2644a macro call coming partially from the macro body and partially from the
2645arguments.  For example,
2646
2647@smallexample
2648#define twice(x) (2*(x))
2649#define call_with_1(x) x(1)
2650call_with_1 (twice)
2651     @expansion{} twice(1)
2652     @expansion{} (2*(1))
2653@end smallexample
2654
2655Macro definitions do not have to have balanced parentheses.  By writing
2656an unbalanced open parenthesis in a macro body, it is possible to create
2657a macro call that begins inside the macro body but ends outside of it.
2658For example,
2659
2660@smallexample
2661#define strange(file) fprintf (file, "%s %d",
2662@dots{}
2663strange(stderr) p, 35)
2664     @expansion{} fprintf (stderr, "%s %d", p, 35)
2665@end smallexample
2666
2667The ability to piece together a macro call can be useful, but the use of
2668unbalanced open parentheses in a macro body is just confusing, and
2669should be avoided.
2670
2671@node Operator Precedence Problems
2672@subsection Operator Precedence Problems
2673@cindex parentheses in macro bodies
2674
2675You may have noticed that in most of the macro definition examples shown
2676above, each occurrence of a macro argument name had parentheses around
2677it.  In addition, another pair of parentheses usually surround the
2678entire macro definition.  Here is why it is best to write macros that
2679way.
2680
2681Suppose you define a macro as follows,
2682
2683@smallexample
2684#define ceil_div(x, y) (x + y - 1) / y
2685@end smallexample
2686
2687@noindent
2688whose purpose is to divide, rounding up.  (One use for this operation is
2689to compute how many @code{int} objects are needed to hold a certain
2690number of @code{char} objects.)  Then suppose it is used as follows:
2691
2692@smallexample
2693a = ceil_div (b & c, sizeof (int));
2694     @expansion{} a = (b & c + sizeof (int) - 1) / sizeof (int);
2695@end smallexample
2696
2697@noindent
2698This does not do what is intended.  The operator-precedence rules of
2699C make it equivalent to this:
2700
2701@smallexample
2702a = (b & (c + sizeof (int) - 1)) / sizeof (int);
2703@end smallexample
2704
2705@noindent
2706What we want is this:
2707
2708@smallexample
2709a = ((b & c) + sizeof (int) - 1)) / sizeof (int);
2710@end smallexample
2711
2712@noindent
2713Defining the macro as
2714
2715@smallexample
2716#define ceil_div(x, y) ((x) + (y) - 1) / (y)
2717@end smallexample
2718
2719@noindent
2720provides the desired result.
2721
2722Unintended grouping can result in another way.  Consider @code{sizeof
2723ceil_div(1, 2)}.  That has the appearance of a C expression that would
2724compute the size of the type of @code{ceil_div (1, 2)}, but in fact it
2725means something very different.  Here is what it expands to:
2726
2727@smallexample
2728sizeof ((1) + (2) - 1) / (2)
2729@end smallexample
2730
2731@noindent
2732This would take the size of an integer and divide it by two.  The
2733precedence rules have put the division outside the @code{sizeof} when it
2734was intended to be inside.
2735
2736Parentheses around the entire macro definition prevent such problems.
2737Here, then, is the recommended way to define @code{ceil_div}:
2738
2739@smallexample
2740#define ceil_div(x, y) (((x) + (y) - 1) / (y))
2741@end smallexample
2742
2743@node Swallowing the Semicolon
2744@subsection Swallowing the Semicolon
2745@cindex semicolons (after macro calls)
2746
2747Often it is desirable to define a macro that expands into a compound
2748statement.  Consider, for example, the following macro, that advances a
2749pointer (the argument @code{p} says where to find it) across whitespace
2750characters:
2751
2752@smallexample
2753#define SKIP_SPACES(p, limit)  \
2754@{ char *lim = (limit);         \
2755  while (p < lim) @{            \
2756    if (*p++ != ' ') @{         \
2757      p--; break; @}@}@}
2758@end smallexample
2759
2760@noindent
2761Here backslash-newline is used to split the macro definition, which must
2762be a single logical line, so that it resembles the way such code would
2763be laid out if not part of a macro definition.
2764
2765A call to this macro might be @code{SKIP_SPACES (p, lim)}.  Strictly
2766speaking, the call expands to a compound statement, which is a complete
2767statement with no need for a semicolon to end it.  However, since it
2768looks like a function call, it minimizes confusion if you can use it
2769like a function call, writing a semicolon afterward, as in
2770@code{SKIP_SPACES (p, lim);}
2771
2772This can cause trouble before @code{else} statements, because the
2773semicolon is actually a null statement.  Suppose you write
2774
2775@smallexample
2776if (*p != 0)
2777  SKIP_SPACES (p, lim);
2778else @dots{}
2779@end smallexample
2780
2781@noindent
2782The presence of two statements---the compound statement and a null
2783statement---in between the @code{if} condition and the @code{else}
2784makes invalid C code.
2785
2786The definition of the macro @code{SKIP_SPACES} can be altered to solve
2787this problem, using a @code{do @dots{} while} statement.  Here is how:
2788
2789@smallexample
2790#define SKIP_SPACES(p, limit)     \
2791do @{ char *lim = (limit);         \
2792     while (p < lim) @{            \
2793       if (*p++ != ' ') @{         \
2794         p--; break; @}@}@}          \
2795while (0)
2796@end smallexample
2797
2798Now @code{SKIP_SPACES (p, lim);} expands into
2799
2800@smallexample
2801do @{@dots{}@} while (0);
2802@end smallexample
2803
2804@noindent
2805which is one statement.  The loop executes exactly once; most compilers
2806generate no extra code for it.
2807
2808@node Duplication of Side Effects
2809@subsection Duplication of Side Effects
2810
2811@cindex side effects (in macro arguments)
2812@cindex unsafe macros
2813Many C programs define a macro @code{min}, for ``minimum'', like this:
2814
2815@smallexample
2816#define min(X, Y)  ((X) < (Y) ? (X) : (Y))
2817@end smallexample
2818
2819When you use this macro with an argument containing a side effect,
2820as shown here,
2821
2822@smallexample
2823next = min (x + y, foo (z));
2824@end smallexample
2825
2826@noindent
2827it expands as follows:
2828
2829@smallexample
2830next = ((x + y) < (foo (z)) ? (x + y) : (foo (z)));
2831@end smallexample
2832
2833@noindent
2834where @code{x + y} has been substituted for @code{X} and @code{foo (z)}
2835for @code{Y}.
2836
2837The function @code{foo} is used only once in the statement as it appears
2838in the program, but the expression @code{foo (z)} has been substituted
2839twice into the macro expansion.  As a result, @code{foo} might be called
2840two times when the statement is executed.  If it has side effects or if
2841it takes a long time to compute, the results might not be what you
2842intended.  We say that @code{min} is an @dfn{unsafe} macro.
2843
2844The best solution to this problem is to define @code{min} in a way that
2845computes the value of @code{foo (z)} only once.  The C language offers
2846no standard way to do this, but it can be done with GNU extensions as
2847follows:
2848
2849@smallexample
2850#define min(X, Y)                \
2851(@{ typeof (X) x_ = (X);          \
2852   typeof (Y) y_ = (Y);          \
2853   (x_ < y_) ? x_ : y_; @})
2854@end smallexample
2855
2856The @samp{(@{ @dots{} @})} notation produces a compound statement that
2857acts as an expression.  Its value is the value of its last statement.
2858This permits us to define local variables and assign each argument to
2859one.  The local variables have underscores after their names to reduce
2860the risk of conflict with an identifier of wider scope (it is impossible
2861to avoid this entirely).  Now each argument is evaluated exactly once.
2862
2863If you do not wish to use GNU C extensions, the only solution is to be
2864careful when @emph{using} the macro @code{min}.  For example, you can
2865calculate the value of @code{foo (z)}, save it in a variable, and use
2866that variable in @code{min}:
2867
2868@smallexample
2869@group
2870#define min(X, Y)  ((X) < (Y) ? (X) : (Y))
2871@dots{}
2872@{
2873  int tem = foo (z);
2874  next = min (x + y, tem);
2875@}
2876@end group
2877@end smallexample
2878
2879@noindent
2880(where we assume that @code{foo} returns type @code{int}).
2881
2882@node Self-Referential Macros
2883@subsection Self-Referential Macros
2884@cindex self-reference
2885
2886A @dfn{self-referential} macro is one whose name appears in its
2887definition.  Recall that all macro definitions are rescanned for more
2888macros to replace.  If the self-reference were considered a use of the
2889macro, it would produce an infinitely large expansion.  To prevent this,
2890the self-reference is not considered a macro call.  It is passed into
2891the preprocessor output unchanged.  Consider an example:
2892
2893@smallexample
2894#define foo (4 + foo)
2895@end smallexample
2896
2897@noindent
2898where @code{foo} is also a variable in your program.
2899
2900Following the ordinary rules, each reference to @code{foo} will expand
2901into @code{(4 + foo)}; then this will be rescanned and will expand into
2902@code{(4 + (4 + foo))}; and so on until the computer runs out of memory.
2903
2904The self-reference rule cuts this process short after one step, at
2905@code{(4 + foo)}.  Therefore, this macro definition has the possibly
2906useful effect of causing the program to add 4 to the value of @code{foo}
2907wherever @code{foo} is referred to.
2908
2909In most cases, it is a bad idea to take advantage of this feature.  A
2910person reading the program who sees that @code{foo} is a variable will
2911not expect that it is a macro as well.  The reader will come across the
2912identifier @code{foo} in the program and think its value should be that
2913of the variable @code{foo}, whereas in fact the value is four greater.
2914
2915One common, useful use of self-reference is to create a macro which
2916expands to itself.  If you write
2917
2918@smallexample
2919#define EPERM EPERM
2920@end smallexample
2921
2922@noindent
2923then the macro @code{EPERM} expands to @code{EPERM}.  Effectively, it is
2924left alone by the preprocessor whenever it's used in running text.  You
2925can tell that it's a macro with @samp{#ifdef}.  You might do this if you
2926want to define numeric constants with an @code{enum}, but have
2927@samp{#ifdef} be true for each constant.
2928
2929If a macro @code{x} expands to use a macro @code{y}, and the expansion of
2930@code{y} refers to the macro @code{x}, that is an @dfn{indirect
2931self-reference} of @code{x}.  @code{x} is not expanded in this case
2932either.  Thus, if we have
2933
2934@smallexample
2935#define x (4 + y)
2936#define y (2 * x)
2937@end smallexample
2938
2939@noindent
2940then @code{x} and @code{y} expand as follows:
2941
2942@smallexample
2943@group
2944x    @expansion{} (4 + y)
2945     @expansion{} (4 + (2 * x))
2946
2947y    @expansion{} (2 * x)
2948     @expansion{} (2 * (4 + y))
2949@end group
2950@end smallexample
2951
2952@noindent
2953Each macro is expanded when it appears in the definition of the other
2954macro, but not when it indirectly appears in its own definition.
2955
2956@node Argument Prescan
2957@subsection Argument Prescan
2958@cindex expansion of arguments
2959@cindex macro argument expansion
2960@cindex prescan of macro arguments
2961
2962Macro arguments are completely macro-expanded before they are
2963substituted into a macro body, unless they are stringized or pasted
2964with other tokens.  After substitution, the entire macro body, including
2965the substituted arguments, is scanned again for macros to be expanded.
2966The result is that the arguments are scanned @emph{twice} to expand
2967macro calls in them.
2968
2969Most of the time, this has no effect.  If the argument contained any
2970macro calls, they are expanded during the first scan.  The result
2971therefore contains no macro calls, so the second scan does not change
2972it.  If the argument were substituted as given, with no prescan, the
2973single remaining scan would find the same macro calls and produce the
2974same results.
2975
2976You might expect the double scan to change the results when a
2977self-referential macro is used in an argument of another macro
2978(@pxref{Self-Referential Macros}): the self-referential macro would be
2979expanded once in the first scan, and a second time in the second scan.
2980However, this is not what happens.  The self-references that do not
2981expand in the first scan are marked so that they will not expand in the
2982second scan either.
2983
2984You might wonder, ``Why mention the prescan, if it makes no difference?
2985And why not skip it and make the preprocessor faster?''  The answer is
2986that the prescan does make a difference in three special cases:
2987
2988@itemize @bullet
2989@item
2990Nested calls to a macro.
2991
2992We say that @dfn{nested} calls to a macro occur when a macro's argument
2993contains a call to that very macro.  For example, if @code{f} is a macro
2994that expects one argument, @code{f (f (1))} is a nested pair of calls to
2995@code{f}.  The desired expansion is made by expanding @code{f (1)} and
2996substituting that into the definition of @code{f}.  The prescan causes
2997the expected result to happen.  Without the prescan, @code{f (1)} itself
2998would be substituted as an argument, and the inner use of @code{f} would
2999appear during the main scan as an indirect self-reference and would not
3000be expanded.
3001
3002@item
3003Macros that call other macros that stringize or concatenate.
3004
3005If an argument is stringized or concatenated, the prescan does not
3006occur.  If you @emph{want} to expand a macro, then stringize or
3007concatenate its expansion, you can do that by causing one macro to call
3008another macro that does the stringizing or concatenation.  For
3009instance, if you have
3010
3011@smallexample
3012#define AFTERX(x) X_ ## x
3013#define XAFTERX(x) AFTERX(x)
3014#define TABLESIZE 1024
3015#define BUFSIZE TABLESIZE
3016@end smallexample
3017
3018then @code{AFTERX(BUFSIZE)} expands to @code{X_BUFSIZE}, and
3019@code{XAFTERX(BUFSIZE)} expands to @code{X_1024}.  (Not to
3020@code{X_TABLESIZE}.  Prescan always does a complete expansion.)
3021
3022@item
3023Macros used in arguments, whose expansions contain unshielded commas.
3024
3025This can cause a macro expanded on the second scan to be called with the
3026wrong number of arguments.  Here is an example:
3027
3028@smallexample
3029#define foo  a,b
3030#define bar(x) lose(x)
3031#define lose(x) (1 + (x))
3032@end smallexample
3033
3034We would like @code{bar(foo)} to turn into @code{(1 + (foo))}, which
3035would then turn into @code{(1 + (a,b))}.  Instead, @code{bar(foo)}
3036expands into @code{lose(a,b)}, and you get an error because @code{lose}
3037requires a single argument.  In this case, the problem is easily solved
3038by the same parentheses that ought to be used to prevent misnesting of
3039arithmetic operations:
3040
3041@smallexample
3042#define foo (a,b)
3043@exdent or
3044#define bar(x) lose((x))
3045@end smallexample
3046
3047The extra pair of parentheses prevents the comma in @code{foo}'s
3048definition from being interpreted as an argument separator.
3049
3050@end itemize
3051
3052@node Newlines in Arguments
3053@subsection Newlines in Arguments
3054@cindex newlines in macro arguments
3055
3056The invocation of a function-like macro can extend over many logical
3057lines.  However, in the present implementation, the entire expansion
3058comes out on one line.  Thus line numbers emitted by the compiler or
3059debugger refer to the line the invocation started on, which might be
3060different to the line containing the argument causing the problem.
3061
3062Here is an example illustrating this:
3063
3064@smallexample
3065#define ignore_second_arg(a,b,c) a; c
3066
3067ignore_second_arg (foo (),
3068                   ignored (),
3069                   syntax error);
3070@end smallexample
3071
3072@noindent
3073The syntax error triggered by the tokens @code{syntax error} results in
3074an error message citing line three---the line of ignore_second_arg---
3075even though the problematic code comes from line five.
3076
3077We consider this a bug, and intend to fix it in the near future.
3078
3079@node Conditionals
3080@chapter Conditionals
3081@cindex conditionals
3082
3083A @dfn{conditional} is a directive that instructs the preprocessor to
3084select whether or not to include a chunk of code in the final token
3085stream passed to the compiler.  Preprocessor conditionals can test
3086arithmetic expressions, or whether a name is defined as a macro, or both
3087simultaneously using the special @code{defined} operator.
3088
3089A conditional in the C preprocessor resembles in some ways an @code{if}
3090statement in C, but it is important to understand the difference between
3091them.  The condition in an @code{if} statement is tested during the
3092execution of your program.  Its purpose is to allow your program to
3093behave differently from run to run, depending on the data it is
3094operating on.  The condition in a preprocessing conditional directive is
3095tested when your program is compiled.  Its purpose is to allow different
3096code to be included in the program depending on the situation at the
3097time of compilation.
3098
3099However, the distinction is becoming less clear.  Modern compilers often
3100do test @code{if} statements when a program is compiled, if their
3101conditions are known not to vary at run time, and eliminate code which
3102can never be executed.  If you can count on your compiler to do this,
3103you may find that your program is more readable if you use @code{if}
3104statements with constant conditions (perhaps determined by macros).  Of
3105course, you can only use this to exclude code, not type definitions or
3106other preprocessing directives, and you can only do it if the code
3107remains syntactically valid when it is not to be used.
3108
3109@menu
3110* Conditional Uses::
3111* Conditional Syntax::
3112* Deleted Code::
3113@end menu
3114
3115@node Conditional Uses
3116@section Conditional Uses
3117
3118There are three general reasons to use a conditional.
3119
3120@itemize @bullet
3121@item
3122A program may need to use different code depending on the machine or
3123operating system it is to run on.  In some cases the code for one
3124operating system may be erroneous on another operating system; for
3125example, it might refer to data types or constants that do not exist on
3126the other system.  When this happens, it is not enough to avoid
3127executing the invalid code.  Its mere presence will cause the compiler
3128to reject the program.  With a preprocessing conditional, the offending
3129code can be effectively excised from the program when it is not valid.
3130
3131@item
3132You may want to be able to compile the same source file into two
3133different programs.  One version might make frequent time-consuming
3134consistency checks on its intermediate data, or print the values of
3135those data for debugging, and the other not.
3136
3137@item
3138A conditional whose condition is always false is one way to exclude code
3139from the program but keep it as a sort of comment for future reference.
3140@end itemize
3141
3142Simple programs that do not need system-specific logic or complex
3143debugging hooks generally will not need to use preprocessing
3144conditionals.
3145
3146@node Conditional Syntax
3147@section Conditional Syntax
3148
3149@findex #if
3150A conditional in the C preprocessor begins with a @dfn{conditional
3151directive}: @samp{#if}, @samp{#ifdef} or @samp{#ifndef}.
3152
3153@menu
3154* Ifdef::
3155* If::
3156* Defined::
3157* Else::
3158* Elif::
3159@end menu
3160
3161@node Ifdef
3162@subsection Ifdef
3163@findex #ifdef
3164@findex #endif
3165
3166The simplest sort of conditional is
3167
3168@smallexample
3169@group
3170#ifdef @var{MACRO}
3171
3172@var{controlled text}
3173
3174#endif /* @var{MACRO} */
3175@end group
3176@end smallexample
3177
3178@cindex conditional group
3179This block is called a @dfn{conditional group}.  @var{controlled text}
3180will be included in the output of the preprocessor if and only if
3181@var{MACRO} is defined.  We say that the conditional @dfn{succeeds} if
3182@var{MACRO} is defined, @dfn{fails} if it is not.
3183
3184The @var{controlled text} inside of a conditional can include
3185preprocessing directives.  They are executed only if the conditional
3186succeeds.  You can nest conditional groups inside other conditional
3187groups, but they must be completely nested.  In other words,
3188@samp{#endif} always matches the nearest @samp{#ifdef} (or
3189@samp{#ifndef}, or @samp{#if}).  Also, you cannot start a conditional
3190group in one file and end it in another.
3191
3192Even if a conditional fails, the @var{controlled text} inside it is
3193still run through initial transformations and tokenization.  Therefore,
3194it must all be lexically valid C@.  Normally the only way this matters is
3195that all comments and string literals inside a failing conditional group
3196must still be properly ended.
3197
3198The comment following the @samp{#endif} is not required, but it is a
3199good practice if there is a lot of @var{controlled text}, because it
3200helps people match the @samp{#endif} to the corresponding @samp{#ifdef}.
3201Older programs sometimes put @var{MACRO} directly after the
3202@samp{#endif} without enclosing it in a comment.  This is invalid code
3203according to the C standard.  CPP accepts it with a warning.  It
3204never affects which @samp{#ifndef} the @samp{#endif} matches.
3205
3206@findex #ifndef
3207Sometimes you wish to use some code if a macro is @emph{not} defined.
3208You can do this by writing @samp{#ifndef} instead of @samp{#ifdef}.
3209One common use of @samp{#ifndef} is to include code only the first
3210time a header file is included.  @xref{Once-Only Headers}.
3211
3212Macro definitions can vary between compilations for several reasons.
3213Here are some samples.
3214
3215@itemize @bullet
3216@item
3217Some macros are predefined on each kind of machine
3218(@pxref{System-specific Predefined Macros}).  This allows you to provide
3219code specially tuned for a particular machine.
3220
3221@item
3222System header files define more macros, associated with the features
3223they implement.  You can test these macros with conditionals to avoid
3224using a system feature on a machine where it is not implemented.
3225
3226@item
3227Macros can be defined or undefined with the @option{-D} and @option{-U}
3228command-line options when you compile the program.  You can arrange to
3229compile the same source file into two different programs by choosing a
3230macro name to specify which program you want, writing conditionals to
3231test whether or how this macro is defined, and then controlling the
3232state of the macro with command-line options, perhaps set in the
3233Makefile.  @xref{Invocation}.
3234
3235@item
3236Your program might have a special header file (often called
3237@file{config.h}) that is adjusted when the program is compiled.  It can
3238define or not define macros depending on the features of the system and
3239the desired capabilities of the program.  The adjustment can be
3240automated by a tool such as @command{autoconf}, or done by hand.
3241@end itemize
3242
3243@node If
3244@subsection If
3245
3246The @samp{#if} directive allows you to test the value of an arithmetic
3247expression, rather than the mere existence of one macro.  Its syntax is
3248
3249@smallexample
3250@group
3251#if @var{expression}
3252
3253@var{controlled text}
3254
3255#endif /* @var{expression} */
3256@end group
3257@end smallexample
3258
3259@var{expression} is a C expression of integer type, subject to stringent
3260restrictions.  It may contain
3261
3262@itemize @bullet
3263@item
3264Integer constants.
3265
3266@item
3267Character constants, which are interpreted as they would be in normal
3268code.
3269
3270@item
3271Arithmetic operators for addition, subtraction, multiplication,
3272division, bitwise operations, shifts, comparisons, and logical
3273operations (@code{&&} and @code{||}).  The latter two obey the usual
3274short-circuiting rules of standard C@.
3275
3276@item
3277Macros.  All macros in the expression are expanded before actual
3278computation of the expression's value begins.
3279
3280@item
3281Uses of the @code{defined} operator, which lets you check whether macros
3282are defined in the middle of an @samp{#if}.
3283
3284@item
3285Identifiers that are not macros, which are all considered to be the
3286number zero.  This allows you to write @code{@w{#if MACRO}} instead of
3287@code{@w{#ifdef MACRO}}, if you know that MACRO, when defined, will
3288always have a nonzero value.  Function-like macros used without their
3289function call parentheses are also treated as zero.
3290
3291In some contexts this shortcut is undesirable.  The @option{-Wundef}
3292option causes GCC to warn whenever it encounters an identifier which is
3293not a macro in an @samp{#if}.
3294@end itemize
3295
3296The preprocessor does not know anything about types in the language.
3297Therefore, @code{sizeof} operators are not recognized in @samp{#if}, and
3298neither are @code{enum} constants.  They will be taken as identifiers
3299which are not macros, and replaced by zero.  In the case of
3300@code{sizeof}, this is likely to cause the expression to be invalid.
3301
3302The preprocessor calculates the value of @var{expression}.  It carries
3303out all calculations in the widest integer type known to the compiler;
3304on most machines supported by GCC this is 64 bits.  This is not the same
3305rule as the compiler uses to calculate the value of a constant
3306expression, and may give different results in some cases.  If the value
3307comes out to be nonzero, the @samp{#if} succeeds and the @var{controlled
3308text} is included; otherwise it is skipped.
3309
3310@node Defined
3311@subsection Defined
3312
3313@cindex @code{defined}
3314The special operator @code{defined} is used in @samp{#if} and
3315@samp{#elif} expressions to test whether a certain name is defined as a
3316macro.  @code{defined @var{name}} and @code{defined (@var{name})} are
3317both expressions whose value is 1 if @var{name} is defined as a macro at
3318the current point in the program, and 0 otherwise.  Thus,  @code{@w{#if
3319defined MACRO}} is precisely equivalent to @code{@w{#ifdef MACRO}}.
3320
3321@code{defined} is useful when you wish to test more than one macro for
3322existence at once.  For example,
3323
3324@smallexample
3325#if defined (__vax__) || defined (__ns16000__)
3326@end smallexample
3327
3328@noindent
3329would succeed if either of the names @code{__vax__} or
3330@code{__ns16000__} is defined as a macro.
3331
3332Conditionals written like this:
3333
3334@smallexample
3335#if defined BUFSIZE && BUFSIZE >= 1024
3336@end smallexample
3337
3338@noindent
3339can generally be simplified to just @code{@w{#if BUFSIZE >= 1024}},
3340since if @code{BUFSIZE} is not defined, it will be interpreted as having
3341the value zero.
3342
3343If the @code{defined} operator appears as a result of a macro expansion,
3344the C standard says the behavior is undefined.  GNU cpp treats it as a
3345genuine @code{defined} operator and evaluates it normally.  It will warn
3346wherever your code uses this feature if you use the command-line option
3347@option{-Wpedantic}, since other compilers may handle it differently.  The
3348warning is also enabled by @option{-Wextra}, and can also be enabled
3349individually with @option{-Wexpansion-to-defined}.
3350
3351@node Else
3352@subsection Else
3353
3354@findex #else
3355The @samp{#else} directive can be added to a conditional to provide
3356alternative text to be used if the condition fails.  This is what it
3357looks like:
3358
3359@smallexample
3360@group
3361#if @var{expression}
3362@var{text-if-true}
3363#else /* Not @var{expression} */
3364@var{text-if-false}
3365#endif /* Not @var{expression} */
3366@end group
3367@end smallexample
3368
3369@noindent
3370If @var{expression} is nonzero, the @var{text-if-true} is included and
3371the @var{text-if-false} is skipped.  If @var{expression} is zero, the
3372opposite happens.
3373
3374You can use @samp{#else} with @samp{#ifdef} and @samp{#ifndef}, too.
3375
3376@node Elif
3377@subsection Elif
3378
3379@findex #elif
3380One common case of nested conditionals is used to check for more than two
3381possible alternatives.  For example, you might have
3382
3383@smallexample
3384#if X == 1
3385@dots{}
3386#else /* X != 1 */
3387#if X == 2
3388@dots{}
3389#else /* X != 2 */
3390@dots{}
3391#endif /* X != 2 */
3392#endif /* X != 1 */
3393@end smallexample
3394
3395Another conditional directive, @samp{#elif}, allows this to be
3396abbreviated as follows:
3397
3398@smallexample
3399#if X == 1
3400@dots{}
3401#elif X == 2
3402@dots{}
3403#else /* X != 2 and X != 1*/
3404@dots{}
3405#endif /* X != 2 and X != 1*/
3406@end smallexample
3407
3408@samp{#elif} stands for ``else if''.  Like @samp{#else}, it goes in the
3409middle of a conditional group and subdivides it; it does not require a
3410matching @samp{#endif} of its own.  Like @samp{#if}, the @samp{#elif}
3411directive includes an expression to be tested.  The text following the
3412@samp{#elif} is processed only if the original @samp{#if}-condition
3413failed and the @samp{#elif} condition succeeds.
3414
3415More than one @samp{#elif} can go in the same conditional group.  Then
3416the text after each @samp{#elif} is processed only if the @samp{#elif}
3417condition succeeds after the original @samp{#if} and all previous
3418@samp{#elif} directives within it have failed.
3419
3420@samp{#else} is allowed after any number of @samp{#elif} directives, but
3421@samp{#elif} may not follow @samp{#else}.
3422
3423@node Deleted Code
3424@section Deleted Code
3425@cindex commenting out code
3426
3427If you replace or delete a part of the program but want to keep the old
3428code around for future reference, you often cannot simply comment it
3429out.  Block comments do not nest, so the first comment inside the old
3430code will end the commenting-out.  The probable result is a flood of
3431syntax errors.
3432
3433One way to avoid this problem is to use an always-false conditional
3434instead.  For instance, put @code{#if 0} before the deleted code and
3435@code{#endif} after it.  This works even if the code being turned
3436off contains conditionals, but they must be entire conditionals
3437(balanced @samp{#if} and @samp{#endif}).
3438
3439Some people use @code{#ifdef notdef} instead.  This is risky, because
3440@code{notdef} might be accidentally defined as a macro, and then the
3441conditional would succeed.  @code{#if 0} can be counted on to fail.
3442
3443Do not use @code{#if 0} for comments which are not C code.  Use a real
3444comment, instead.  The interior of @code{#if 0} must consist of complete
3445tokens; in particular, single-quote characters must balance.  Comments
3446often contain unbalanced single-quote characters (known in English as
3447apostrophes).  These confuse @code{#if 0}.  They don't confuse
3448@samp{/*}.
3449
3450@node Diagnostics
3451@chapter Diagnostics
3452@cindex diagnostic
3453@cindex reporting errors
3454@cindex reporting warnings
3455
3456@findex #error
3457The directive @samp{#error} causes the preprocessor to report a fatal
3458error.  The tokens forming the rest of the line following @samp{#error}
3459are used as the error message.
3460
3461You would use @samp{#error} inside of a conditional that detects a
3462combination of parameters which you know the program does not properly
3463support.  For example, if you know that the program will not run
3464properly on a VAX, you might write
3465
3466@smallexample
3467@group
3468#ifdef __vax__
3469#error "Won't work on VAXen.  See comments at get_last_object."
3470#endif
3471@end group
3472@end smallexample
3473
3474If you have several configuration parameters that must be set up by
3475the installation in a consistent way, you can use conditionals to detect
3476an inconsistency and report it with @samp{#error}.  For example,
3477
3478@smallexample
3479#if !defined(FOO) && defined(BAR)
3480#error "BAR requires FOO."
3481#endif
3482@end smallexample
3483
3484@findex #warning
3485The directive @samp{#warning} is like @samp{#error}, but causes the
3486preprocessor to issue a warning and continue preprocessing.  The tokens
3487following @samp{#warning} are used as the warning message.
3488
3489You might use @samp{#warning} in obsolete header files, with a message
3490directing the user to the header file which should be used instead.
3491
3492Neither @samp{#error} nor @samp{#warning} macro-expands its argument.
3493Internal whitespace sequences are each replaced with a single space.
3494The line must consist of complete tokens.  It is wisest to make the
3495argument of these directives be a single string constant; this avoids
3496problems with apostrophes and the like.
3497
3498@node Line Control
3499@chapter Line Control
3500@cindex line control
3501
3502The C preprocessor informs the C compiler of the location in your source
3503code where each token came from.  Presently, this is just the file name
3504and line number.  All the tokens resulting from macro expansion are
3505reported as having appeared on the line of the source file where the
3506outermost macro was used.  We intend to be more accurate in the future.
3507
3508If you write a program which generates source code, such as the
3509@command{bison} parser generator, you may want to adjust the preprocessor's
3510notion of the current file name and line number by hand.  Parts of the
3511output from @command{bison} are generated from scratch, other parts come
3512from a standard parser file.  The rest are copied verbatim from
3513@command{bison}'s input.  You would like compiler error messages and
3514symbolic debuggers to be able to refer to @code{bison}'s input file.
3515
3516@findex #line
3517@command{bison} or any such program can arrange this by writing
3518@samp{#line} directives into the output file.  @samp{#line} is a
3519directive that specifies the original line number and source file name
3520for subsequent input in the current preprocessor input file.
3521@samp{#line} has three variants:
3522
3523@table @code
3524@item #line @var{linenum}
3525@var{linenum} is a non-negative decimal integer constant.  It specifies
3526the line number which should be reported for the following line of
3527input.  Subsequent lines are counted from @var{linenum}.
3528
3529@item #line @var{linenum} @var{filename}
3530@var{linenum} is the same as for the first form, and has the same
3531effect.  In addition, @var{filename} is a string constant.  The
3532following line and all subsequent lines are reported to come from the
3533file it specifies, until something else happens to change that.
3534@var{filename} is interpreted according to the normal rules for a string
3535constant: backslash escapes are interpreted.  This is different from
3536@samp{#include}.
3537
3538@item #line @var{anything else}
3539@var{anything else} is checked for macro calls, which are expanded.
3540The result should match one of the above two forms.
3541@end table
3542
3543@samp{#line} directives alter the results of the @code{__FILE__} and
3544@code{__LINE__} predefined macros from that point on.  @xref{Standard
3545Predefined Macros}.  They do not have any effect on @samp{#include}'s
3546idea of the directory containing the current file.
3547
3548@node Pragmas
3549@chapter Pragmas
3550
3551The @samp{#pragma} directive is the method specified by the C standard
3552for providing additional information to the compiler, beyond what is
3553conveyed in the language itself.  The forms of this directive
3554(commonly known as @dfn{pragmas}) specified by C standard are prefixed with
3555@code{STDC}.  A C compiler is free to attach any meaning it likes to other
3556pragmas.  All GNU-defined, supported pragmas have been given a
3557@code{GCC} prefix.
3558
3559@cindex @code{_Pragma}
3560C99 introduced the @code{@w{_Pragma}} operator.  This feature addresses a
3561major problem with @samp{#pragma}: being a directive, it cannot be
3562produced as the result of macro expansion.  @code{@w{_Pragma}} is an
3563operator, much like @code{sizeof} or @code{defined}, and can be embedded
3564in a macro.
3565
3566Its syntax is @code{@w{_Pragma (@var{string-literal})}}, where
3567@var{string-literal} can be either a normal or wide-character string
3568literal.  It is destringized, by replacing all @samp{\\} with a single
3569@samp{\} and all @samp{\"} with a @samp{"}.  The result is then
3570processed as if it had appeared as the right hand side of a
3571@samp{#pragma} directive.  For example,
3572
3573@smallexample
3574_Pragma ("GCC dependency \"parse.y\"")
3575@end smallexample
3576
3577@noindent
3578has the same effect as @code{#pragma GCC dependency "parse.y"}.  The
3579same effect could be achieved using macros, for example
3580
3581@smallexample
3582#define DO_PRAGMA(x) _Pragma (#x)
3583DO_PRAGMA (GCC dependency "parse.y")
3584@end smallexample
3585
3586The standard is unclear on where a @code{_Pragma} operator can appear.
3587The preprocessor does not accept it within a preprocessing conditional
3588directive like @samp{#if}.  To be safe, you are probably best keeping it
3589out of directives other than @samp{#define}, and putting it on a line of
3590its own.
3591
3592This manual documents the pragmas which are meaningful to the
3593preprocessor itself.  Other pragmas are meaningful to the C or C++
3594compilers.  They are documented in the GCC manual.
3595
3596GCC plugins may provide their own pragmas.
3597
3598@ftable @code
3599@item #pragma GCC dependency
3600@code{#pragma GCC dependency} allows you to check the relative dates of
3601the current file and another file.  If the other file is more recent than
3602the current file, a warning is issued.  This is useful if the current
3603file is derived from the other file, and should be regenerated.  The
3604other file is searched for using the normal include search path.
3605Optional trailing text can be used to give more information in the
3606warning message.
3607
3608@smallexample
3609#pragma GCC dependency "parse.y"
3610#pragma GCC dependency "/usr/include/time.h" rerun fixincludes
3611@end smallexample
3612
3613@item #pragma GCC poison
3614Sometimes, there is an identifier that you want to remove completely
3615from your program, and make sure that it never creeps back in.  To
3616enforce this, you can @dfn{poison} the identifier with this pragma.
3617@code{#pragma GCC poison} is followed by a list of identifiers to
3618poison.  If any of those identifiers appears anywhere in the source
3619after the directive, it is a hard error.  For example,
3620
3621@smallexample
3622#pragma GCC poison printf sprintf fprintf
3623sprintf(some_string, "hello");
3624@end smallexample
3625
3626@noindent
3627will produce an error.
3628
3629If a poisoned identifier appears as part of the expansion of a macro
3630which was defined before the identifier was poisoned, it will @emph{not}
3631cause an error.  This lets you poison an identifier without worrying
3632about system headers defining macros that use it.
3633
3634For example,
3635
3636@smallexample
3637#define strrchr rindex
3638#pragma GCC poison rindex
3639strrchr(some_string, 'h');
3640@end smallexample
3641
3642@noindent
3643will not produce an error.
3644
3645@item #pragma GCC system_header
3646This pragma takes no arguments.  It causes the rest of the code in the
3647current file to be treated as if it came from a system header.
3648@xref{System Headers}.
3649
3650@item #pragma GCC warning
3651@itemx #pragma GCC error
3652@code{#pragma GCC warning "message"} causes the preprocessor to issue
3653a warning diagnostic with the text @samp{message}.  The message
3654contained in the pragma must be a single string literal.  Similarly,
3655@code{#pragma GCC error "message"} issues an error message.  Unlike
3656the @samp{#warning} and @samp{#error} directives, these pragmas can be
3657embedded in preprocessor macros using @samp{_Pragma}.
3658
3659@end ftable
3660
3661@node Other Directives
3662@chapter Other Directives
3663
3664@findex #ident
3665@findex #sccs
3666The @samp{#ident} directive takes one argument, a string constant.  On
3667some systems, that string constant is copied into a special segment of
3668the object file.  On other systems, the directive is ignored.  The
3669@samp{#sccs} directive is a synonym for @samp{#ident}.
3670
3671These directives are not part of the C standard, but they are not
3672official GNU extensions either.  What historical information we have
3673been able to find, suggests they originated with System V@.
3674
3675@cindex null directive
3676The @dfn{null directive} consists of a @samp{#} followed by a newline,
3677with only whitespace (including comments) in between.  A null directive
3678is understood as a preprocessing directive but has no effect on the
3679preprocessor output.  The primary significance of the existence of the
3680null directive is that an input line consisting of just a @samp{#} will
3681produce no output, rather than a line of output containing just a
3682@samp{#}.  Supposedly some old C programs contain such lines.
3683
3684@node Preprocessor Output
3685@chapter Preprocessor Output
3686
3687When the C preprocessor is used with the C, C++, or Objective-C
3688compilers, it is integrated into the compiler and communicates a stream
3689of binary tokens directly to the compiler's parser.  However, it can
3690also be used in the more conventional standalone mode, where it produces
3691textual output.
3692@c FIXME: Document the library interface.
3693
3694@cindex output format
3695The output from the C preprocessor looks much like the input, except
3696that all preprocessing directive lines have been replaced with blank
3697lines and all comments with spaces.  Long runs of blank lines are
3698discarded.
3699
3700The ISO standard specifies that it is implementation defined whether a
3701preprocessor preserves whitespace between tokens, or replaces it with
3702e.g.@: a single space.  In GNU CPP, whitespace between tokens is collapsed
3703to become a single space, with the exception that the first token on a
3704non-directive line is preceded with sufficient spaces that it appears in
3705the same column in the preprocessed output that it appeared in the
3706original source file.  This is so the output is easy to read.
3707CPP does not insert any
3708whitespace where there was none in the original source, except where
3709necessary to prevent an accidental token paste.
3710
3711@cindex linemarkers
3712Source file name and line number information is conveyed by lines
3713of the form
3714
3715@smallexample
3716# @var{linenum} @var{filename} @var{flags}
3717@end smallexample
3718
3719@noindent
3720These are called @dfn{linemarkers}.  They are inserted as needed into
3721the output (but never within a string or character constant).  They mean
3722that the following line originated in file @var{filename} at line
3723@var{linenum}.  @var{filename} will never contain any non-printing
3724characters; they are replaced with octal escape sequences.
3725
3726After the file name comes zero or more flags, which are @samp{1},
3727@samp{2}, @samp{3}, or @samp{4}.  If there are multiple flags, spaces
3728separate them.  Here is what the flags mean:
3729
3730@table @samp
3731@item 1
3732This indicates the start of a new file.
3733@item 2
3734This indicates returning to a file (after having included another file).
3735@item 3
3736This indicates that the following text comes from a system header file,
3737so certain warnings should be suppressed.
3738@item 4
3739This indicates that the following text should be treated as being
3740wrapped in an implicit @code{extern "C"} block.
3741@c maybe cross reference NO_IMPLICIT_EXTERN_C
3742@end table
3743
3744As an extension, the preprocessor accepts linemarkers in non-assembler
3745input files.  They are treated like the corresponding @samp{#line}
3746directive, (@pxref{Line Control}), except that trailing flags are
3747permitted, and are interpreted with the meanings described above.  If
3748multiple flags are given, they must be in ascending order.
3749
3750Some directives may be duplicated in the output of the preprocessor.
3751These are @samp{#ident} (always), @samp{#pragma} (only if the
3752preprocessor does not handle the pragma itself), and @samp{#define} and
3753@samp{#undef} (with certain debugging options).  If this happens, the
3754@samp{#} of the directive will always be in the first column, and there
3755will be no space between the @samp{#} and the directive name.  If macro
3756expansion happens to generate tokens which might be mistaken for a
3757duplicated directive, a space will be inserted between the @samp{#} and
3758the directive name.
3759
3760@node Traditional Mode
3761@chapter Traditional Mode
3762
3763Traditional (pre-standard) C preprocessing is rather different from
3764the preprocessing specified by the standard.  When the preprocessor
3765is invoked with the
3766@option{-traditional-cpp} option, it attempts to emulate a traditional
3767preprocessor.
3768
3769This mode is not useful for compiling C code with GCC,
3770but is intended for use with non-C preprocessing applications.  Thus
3771traditional mode semantics are supported only when invoking
3772the preprocessor explicitly, and not in the compiler front ends.
3773
3774The implementation does not correspond precisely to the behavior of
3775early pre-standard versions of GCC, nor to any true traditional preprocessor.
3776After all, inconsistencies among traditional implementations were a
3777major motivation for C standardization.  However, we intend that it
3778should be compatible with true traditional preprocessors in all ways
3779that actually matter.
3780
3781@menu
3782* Traditional lexical analysis::
3783* Traditional macros::
3784* Traditional miscellany::
3785* Traditional warnings::
3786@end menu
3787
3788@node Traditional lexical analysis
3789@section Traditional lexical analysis
3790
3791The traditional preprocessor does not decompose its input into tokens
3792the same way a standards-conforming preprocessor does.  The input is
3793simply treated as a stream of text with minimal internal form.
3794
3795This implementation does not treat trigraphs (@pxref{trigraphs})
3796specially since they were an invention of the standards committee.  It
3797handles arbitrarily-positioned escaped newlines properly and splices
3798the lines as you would expect; many traditional preprocessors did not
3799do this.
3800
3801The form of horizontal whitespace in the input file is preserved in
3802the output.  In particular, hard tabs remain hard tabs.  This can be
3803useful if, for example, you are preprocessing a Makefile.
3804
3805Traditional CPP only recognizes C-style block comments, and treats the
3806@samp{/*} sequence as introducing a comment only if it lies outside
3807quoted text.  Quoted text is introduced by the usual single and double
3808quotes, and also by an initial @samp{<} in a @code{#include}
3809directive.
3810
3811Traditionally, comments are completely removed and are not replaced
3812with a space.  Since a traditional compiler does its own tokenization
3813of the output of the preprocessor, this means that comments can
3814effectively be used as token paste operators.  However, comments
3815behave like separators for text handled by the preprocessor itself,
3816since it doesn't re-lex its input.  For example, in
3817
3818@smallexample
3819#if foo/**/bar
3820@end smallexample
3821
3822@noindent
3823@samp{foo} and @samp{bar} are distinct identifiers and expanded
3824separately if they happen to be macros.  In other words, this
3825directive is equivalent to
3826
3827@smallexample
3828#if foo bar
3829@end smallexample
3830
3831@noindent
3832rather than
3833
3834@smallexample
3835#if foobar
3836@end smallexample
3837
3838Generally speaking, in traditional mode an opening quote need not have
3839a matching closing quote.  In particular, a macro may be defined with
3840replacement text that contains an unmatched quote.  Of course, if you
3841attempt to compile preprocessed output containing an unmatched quote
3842you will get a syntax error.
3843
3844However, all preprocessing directives other than @code{#define}
3845require matching quotes.  For example:
3846
3847@smallexample
3848#define m This macro's fine and has an unmatched quote
3849"/* This is not a comment.  */
3850/* @r{This is a comment.  The following #include directive
3851   is ill-formed.}  */
3852#include <stdio.h
3853@end smallexample
3854
3855Just as for the ISO preprocessor, what would be a closing quote can be
3856escaped with a backslash to prevent the quoted text from closing.
3857
3858@node Traditional macros
3859@section Traditional macros
3860
3861The major difference between traditional and ISO macros is that the
3862former expand to text rather than to a token sequence.  CPP removes
3863all leading and trailing horizontal whitespace from a macro's
3864replacement text before storing it, but preserves the form of internal
3865whitespace.
3866
3867One consequence is that it is legitimate for the replacement text to
3868contain an unmatched quote (@pxref{Traditional lexical analysis}).  An
3869unclosed string or character constant continues into the text
3870following the macro call.  Similarly, the text at the end of a macro's
3871expansion can run together with the text after the macro invocation to
3872produce a single token.
3873
3874Normally comments are removed from the replacement text after the
3875macro is expanded, but if the @option{-CC} option is passed on the
3876command-line comments are preserved.  (In fact, the current
3877implementation removes comments even before saving the macro
3878replacement text, but it careful to do it in such a way that the
3879observed effect is identical even in the function-like macro case.)
3880
3881The ISO stringizing operator @samp{#} and token paste operator
3882@samp{##} have no special meaning.  As explained later, an effect
3883similar to these operators can be obtained in a different way.  Macro
3884names that are embedded in quotes, either from the main file or after
3885macro replacement, do not expand.
3886
3887CPP replaces an unquoted object-like macro name with its replacement
3888text, and then rescans it for further macros to replace.  Unlike
3889standard macro expansion, traditional macro expansion has no provision
3890to prevent recursion.  If an object-like macro appears unquoted in its
3891replacement text, it will be replaced again during the rescan pass,
3892and so on @emph{ad infinitum}.  GCC detects when it is expanding
3893recursive macros, emits an error message, and continues after the
3894offending macro invocation.
3895
3896@smallexample
3897#define PLUS +
3898#define INC(x) PLUS+x
3899INC(foo);
3900     @expansion{} ++foo;
3901@end smallexample
3902
3903Function-like macros are similar in form but quite different in
3904behavior to their ISO counterparts.  Their arguments are contained
3905within parentheses, are comma-separated, and can cross physical lines.
3906Commas within nested parentheses are not treated as argument
3907separators.  Similarly, a quote in an argument cannot be left
3908unclosed; a following comma or parenthesis that comes before the
3909closing quote is treated like any other character.  There is no
3910facility for handling variadic macros.
3911
3912This implementation removes all comments from macro arguments, unless
3913the @option{-C} option is given.  The form of all other horizontal
3914whitespace in arguments is preserved, including leading and trailing
3915whitespace.  In particular
3916
3917@smallexample
3918f( )
3919@end smallexample
3920
3921@noindent
3922is treated as an invocation of the macro @samp{f} with a single
3923argument consisting of a single space.  If you want to invoke a
3924function-like macro that takes no arguments, you must not leave any
3925whitespace between the parentheses.
3926
3927If a macro argument crosses a new line, the new line is replaced with
3928a space when forming the argument.  If the previous line contained an
3929unterminated quote, the following line inherits the quoted state.
3930
3931Traditional preprocessors replace parameters in the replacement text
3932with their arguments regardless of whether the parameters are within
3933quotes or not.  This provides a way to stringize arguments.  For
3934example
3935
3936@smallexample
3937#define str(x) "x"
3938str(/* @r{A comment} */some text )
3939     @expansion{} "some text "
3940@end smallexample
3941
3942@noindent
3943Note that the comment is removed, but that the trailing space is
3944preserved.  Here is an example of using a comment to effect token
3945pasting.
3946
3947@smallexample
3948#define suffix(x) foo_/**/x
3949suffix(bar)
3950     @expansion{} foo_bar
3951@end smallexample
3952
3953@node Traditional miscellany
3954@section Traditional miscellany
3955
3956Here are some things to be aware of when using the traditional
3957preprocessor.
3958
3959@itemize @bullet
3960@item
3961Preprocessing directives are recognized only when their leading
3962@samp{#} appears in the first column.  There can be no whitespace
3963between the beginning of the line and the @samp{#}, but whitespace can
3964follow the @samp{#}.
3965
3966@item
3967A true traditional C preprocessor does not recognize @samp{#error} or
3968@samp{#pragma}, and may not recognize @samp{#elif}.  CPP supports all
3969the directives in traditional mode that it supports in ISO mode,
3970including extensions, with the exception that the effects of
3971@samp{#pragma GCC poison} are undefined.
3972
3973@item
3974__STDC__ is not defined.
3975
3976@item
3977If you use digraphs the behavior is undefined.
3978
3979@item
3980If a line that looks like a directive appears within macro arguments,
3981the behavior is undefined.
3982
3983@end itemize
3984
3985@node Traditional warnings
3986@section Traditional warnings
3987You can request warnings about features that did not exist, or worked
3988differently, in traditional C with the @option{-Wtraditional} option.
3989GCC does not warn about features of ISO C which you must use when you
3990are using a conforming compiler, such as the @samp{#} and @samp{##}
3991operators.
3992
3993Presently @option{-Wtraditional} warns about:
3994
3995@itemize @bullet
3996@item
3997Macro parameters that appear within string literals in the macro body.
3998In traditional C macro replacement takes place within string literals,
3999but does not in ISO C@.
4000
4001@item
4002In traditional C, some preprocessor directives did not exist.
4003Traditional preprocessors would only consider a line to be a directive
4004if the @samp{#} appeared in column 1 on the line.  Therefore
4005@option{-Wtraditional} warns about directives that traditional C
4006understands but would ignore because the @samp{#} does not appear as the
4007first character on the line.  It also suggests you hide directives like
4008@samp{#pragma} not understood by traditional C by indenting them.  Some
4009traditional implementations would not recognize @samp{#elif}, so it
4010suggests avoiding it altogether.
4011
4012@item
4013A function-like macro that appears without an argument list.  In some
4014traditional preprocessors this was an error.  In ISO C it merely means
4015that the macro is not expanded.
4016
4017@item
4018The unary plus operator.  This did not exist in traditional C@.
4019
4020@item
4021The @samp{U} and @samp{LL} integer constant suffixes, which were not
4022available in traditional C@.  (Traditional C does support the @samp{L}
4023suffix for simple long integer constants.)  You are not warned about
4024uses of these suffixes in macros defined in system headers.  For
4025instance, @code{UINT_MAX} may well be defined as @code{4294967295U}, but
4026you will not be warned if you use @code{UINT_MAX}.
4027
4028You can usually avoid the warning, and the related warning about
4029constants which are so large that they are unsigned, by writing the
4030integer constant in question in hexadecimal, with no U suffix.  Take
4031care, though, because this gives the wrong result in exotic cases.
4032@end itemize
4033
4034@node Implementation Details
4035@chapter Implementation Details
4036
4037Here we document details of how the preprocessor's implementation
4038affects its user-visible behavior.  You should try to avoid undue
4039reliance on behavior described here, as it is possible that it will
4040change subtly in future implementations.
4041
4042Also documented here are obsolete features still supported by CPP@.
4043
4044@menu
4045* Implementation-defined behavior::
4046* Implementation limits::
4047* Obsolete Features::
4048@end menu
4049
4050@node Implementation-defined behavior
4051@section Implementation-defined behavior
4052@cindex implementation-defined behavior
4053
4054This is how CPP behaves in all the cases which the C standard
4055describes as @dfn{implementation-defined}.  This term means that the
4056implementation is free to do what it likes, but must document its choice
4057and stick to it.
4058@c FIXME: Check the C++ standard for more implementation-defined stuff.
4059
4060@itemize @bullet
4061@need 1000
4062@item The mapping of physical source file multi-byte characters to the
4063execution character set.
4064
4065The input character set can be specified using the
4066@option{-finput-charset} option, while the execution character set may
4067be controlled using the @option{-fexec-charset} and
4068@option{-fwide-exec-charset} options.
4069
4070@item Identifier characters.
4071@anchor{Identifier characters}
4072
4073The C and C++ standards allow identifiers to be composed of @samp{_}
4074and the alphanumeric characters.  C++ also allows universal character
4075names.  C99 and later C standards permit both universal character
4076names and implementation-defined characters.
4077
4078GCC allows the @samp{$} character in identifiers as an extension for
4079most targets.  This is true regardless of the @option{std=} switch,
4080since this extension cannot conflict with standards-conforming
4081programs.  When preprocessing assembler, however, dollars are not
4082identifier characters by default.
4083
4084Currently the targets that by default do not permit @samp{$} are AVR,
4085IP2K, MMIX, MIPS Irix 3, ARM aout, and PowerPC targets for the AIX
4086operating system.
4087
4088You can override the default with @option{-fdollars-in-identifiers} or
4089@option{fno-dollars-in-identifiers}.  @xref{fdollars-in-identifiers}.
4090
4091@item Non-empty sequences of whitespace characters.
4092
4093In textual output, each whitespace sequence is collapsed to a single
4094space.  For aesthetic reasons, the first token on each non-directive
4095line of output is preceded with sufficient spaces that it appears in the
4096same column as it did in the original source file.
4097
4098@item The numeric value of character constants in preprocessor expressions.
4099
4100The preprocessor and compiler interpret character constants in the
4101same way; i.e.@: escape sequences such as @samp{\a} are given the
4102values they would have on the target machine.
4103
4104The compiler evaluates a multi-character character constant a character
4105at a time, shifting the previous value left by the number of bits per
4106target character, and then or-ing in the bit-pattern of the new
4107character truncated to the width of a target character.  The final
4108bit-pattern is given type @code{int}, and is therefore signed,
4109regardless of whether single characters are signed or not.
4110If there are more
4111characters in the constant than would fit in the target @code{int} the
4112compiler issues a warning, and the excess leading characters are
4113ignored.
4114
4115For example, @code{'ab'} for a target with an 8-bit @code{char} would be
4116interpreted as @w{@samp{(int) ((unsigned char) 'a' * 256 + (unsigned char)
4117'b')}}, and @code{'\234a'} as @w{@samp{(int) ((unsigned char) '\234' *
4118256 + (unsigned char) 'a')}}.
4119
4120@item Source file inclusion.
4121
4122For a discussion on how the preprocessor locates header files,
4123@ref{Include Operation}.
4124
4125@item Interpretation of the filename resulting from a macro-expanded
4126@samp{#include} directive.
4127
4128@xref{Computed Includes}.
4129
4130@item Treatment of a @samp{#pragma} directive that after macro-expansion
4131results in a standard pragma.
4132
4133No macro expansion occurs on any @samp{#pragma} directive line, so the
4134question does not arise.
4135
4136Note that GCC does not yet implement any of the standard
4137pragmas.
4138
4139@end itemize
4140
4141@node Implementation limits
4142@section Implementation limits
4143@cindex implementation limits
4144
4145CPP has a small number of internal limits.  This section lists the
4146limits which the C standard requires to be no lower than some minimum,
4147and all the others known.  It is intended that there should be as few limits
4148as possible.  If you encounter an undocumented or inconvenient limit,
4149please report that as a bug.  @xref{Bugs, , Reporting Bugs, gcc, Using
4150the GNU Compiler Collection (GCC)}.
4151
4152Where we say something is limited @dfn{only by available memory}, that
4153means that internal data structures impose no intrinsic limit, and space
4154is allocated with @code{malloc} or equivalent.  The actual limit will
4155therefore depend on many things, such as the size of other things
4156allocated by the compiler at the same time, the amount of memory
4157consumed by other processes on the same computer, etc.
4158
4159@itemize @bullet
4160
4161@item Nesting levels of @samp{#include} files.
4162
4163We impose an arbitrary limit of 200 levels, to avoid runaway recursion.
4164The standard requires at least 15 levels.
4165
4166@item Nesting levels of conditional inclusion.
4167
4168The C standard mandates this be at least 63.  CPP is limited only by
4169available memory.
4170
4171@item Levels of parenthesized expressions within a full expression.
4172
4173The C standard requires this to be at least 63.  In preprocessor
4174conditional expressions, it is limited only by available memory.
4175
4176@item Significant initial characters in an identifier or macro name.
4177
4178The preprocessor treats all characters as significant.  The C standard
4179requires only that the first 63 be significant.
4180
4181@item Number of macros simultaneously defined in a single translation unit.
4182
4183The standard requires at least 4095 be possible.  CPP is limited only
4184by available memory.
4185
4186@item Number of parameters in a macro definition and arguments in a macro call.
4187
4188We allow @code{USHRT_MAX}, which is no smaller than 65,535.  The minimum
4189required by the standard is 127.
4190
4191@item Number of characters on a logical source line.
4192
4193The C standard requires a minimum of 4096 be permitted.  CPP places
4194no limits on this, but you may get incorrect column numbers reported in
4195diagnostics for lines longer than 65,535 characters.
4196
4197@item Maximum size of a source file.
4198
4199The standard does not specify any lower limit on the maximum size of a
4200source file.  GNU cpp maps files into memory, so it is limited by the
4201available address space.  This is generally at least two gigabytes.
4202Depending on the operating system, the size of physical memory may or
4203may not be a limitation.
4204
4205@end itemize
4206
4207@node Obsolete Features
4208@section Obsolete Features
4209
4210CPP has some features which are present mainly for compatibility with
4211older programs.  We discourage their use in new code.  In some cases,
4212we plan to remove the feature in a future version of GCC@.
4213
4214@subsection Assertions
4215@cindex assertions
4216
4217@dfn{Assertions} are a deprecated alternative to macros in writing
4218conditionals to test what sort of computer or system the compiled
4219program will run on.  Assertions are usually predefined, but you can
4220define them with preprocessing directives or command-line options.
4221
4222Assertions were intended to provide a more systematic way to describe
4223the compiler's target system and we added them for compatibility with
4224existing compilers.  In practice they are just as unpredictable as the
4225system-specific predefined macros.  In addition, they are not part of
4226any standard, and only a few compilers support them.
4227Therefore, the use of assertions is @strong{less} portable than the use
4228of system-specific predefined macros.  We recommend you do not use them at
4229all.
4230
4231@cindex predicates
4232An assertion looks like this:
4233
4234@smallexample
4235#@var{predicate} (@var{answer})
4236@end smallexample
4237
4238@noindent
4239@var{predicate} must be a single identifier.  @var{answer} can be any
4240sequence of tokens; all characters are significant except for leading
4241and trailing whitespace, and differences in internal whitespace
4242sequences are ignored.  (This is similar to the rules governing macro
4243redefinition.)  Thus, @code{(x + y)} is different from @code{(x+y)} but
4244equivalent to @code{@w{( x + y )}}.  Parentheses do not nest inside an
4245answer.
4246
4247@cindex testing predicates
4248To test an assertion, you write it in an @samp{#if}.  For example, this
4249conditional succeeds if either @code{vax} or @code{ns16000} has been
4250asserted as an answer for @code{machine}.
4251
4252@smallexample
4253#if #machine (vax) || #machine (ns16000)
4254@end smallexample
4255
4256@noindent
4257You can test whether @emph{any} answer is asserted for a predicate by
4258omitting the answer in the conditional:
4259
4260@smallexample
4261#if #machine
4262@end smallexample
4263
4264@findex #assert
4265Assertions are made with the @samp{#assert} directive.  Its sole
4266argument is the assertion to make, without the leading @samp{#} that
4267identifies assertions in conditionals.
4268
4269@smallexample
4270#assert @var{predicate} (@var{answer})
4271@end smallexample
4272
4273@noindent
4274You may make several assertions with the same predicate and different
4275answers.  Subsequent assertions do not override previous ones for the
4276same predicate.  All the answers for any given predicate are
4277simultaneously true.
4278
4279@cindex assertions, canceling
4280@findex #unassert
4281Assertions can be canceled with the @samp{#unassert} directive.  It
4282has the same syntax as @samp{#assert}.  In that form it cancels only the
4283answer which was specified on the @samp{#unassert} line; other answers
4284for that predicate remain true.  You can cancel an entire predicate by
4285leaving out the answer:
4286
4287@smallexample
4288#unassert @var{predicate}
4289@end smallexample
4290
4291@noindent
4292In either form, if no such assertion has been made, @samp{#unassert} has
4293no effect.
4294
4295You can also make or cancel assertions using command-line options.
4296@xref{Invocation}.
4297
4298@node Invocation
4299@chapter Invocation
4300@cindex invocation
4301@cindex command line
4302
4303Most often when you use the C preprocessor you do not have to invoke it
4304explicitly: the C compiler does so automatically.  However, the
4305preprocessor is sometimes useful on its own.  You can invoke the
4306preprocessor either with the @command{cpp} command, or via @command{gcc -E}.
4307In GCC, the preprocessor is actually integrated with the compiler
4308rather than a separate program, and both of these commands invoke
4309GCC and tell it to stop after the preprocessing phase.
4310
4311The @command{cpp} options listed here are also accepted by
4312@command{gcc} and have the same meaning.  Likewise the @command{cpp}
4313command accepts all the usual @command{gcc} driver options, although those
4314pertaining to compilation phases after preprocessing are ignored.
4315
4316Only options specific to preprocessing behavior are documented here.
4317Refer to the GCC manual for full documentation of other driver options.
4318
4319@ignore
4320@c man begin SYNOPSIS
4321cpp [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
4322    [@option{-I}@var{dir}@dots{}] [@option{-iquote}@var{dir}@dots{}]
4323    [@option{-iremap}@var{src}:@var{dst}]
4324    [@option{-W}@var{warn}@dots{}]
4325    [@option{-M}|@option{-MM}] [@option{-MG}] [@option{-MF} @var{filename}]
4326    [@option{-MP}] [@option{-MQ} @var{target}@dots{}]
4327    [@option{-MT} @var{target}@dots{}]
4328    @var{infile} [[@option{-o}] @var{outfile}]
4329
4330Only the most useful options are given above; see below for a more
4331complete list of preprocessor-specific options.
4332In addition, @command{cpp} accepts most @command{gcc} driver options, which
4333are not listed here.  Refer to the GCC documentation for details.
4334@c man end
4335@c man begin SEEALSO
4336gpl(7), gfdl(7), fsf-funding(7),
4337gcc(1), and the Info entries for @file{cpp} and @file{gcc}.
4338@c man end
4339@end ignore
4340
4341@c man begin OPTIONS
4342The @command{cpp} command expects two file names as arguments, @var{infile} and
4343@var{outfile}.  The preprocessor reads @var{infile} together with any
4344other files it specifies with @samp{#include}.  All the output generated
4345by the combined input files is written in @var{outfile}.
4346
4347Either @var{infile} or @var{outfile} may be @option{-}, which as
4348@var{infile} means to read from standard input and as @var{outfile}
4349means to write to standard output.  If either file is omitted, it
4350means the same as if @option{-} had been specified for that file.
4351You can also use the @option{-o @var{outfile}} option to specify the
4352output file.
4353
4354Unless otherwise noted, or the option ends in @samp{=}, all options
4355which take an argument may have that argument appear either immediately
4356after the option, or with a space between option and argument:
4357@option{-Ifoo} and @option{-I foo} have the same effect.
4358
4359@cindex grouping options
4360@cindex options, grouping
4361Many options have multi-letter names; therefore multiple single-letter
4362options may @emph{not} be grouped: @option{-dM} is very different from
4363@w{@samp{-d -M}}.
4364
4365@cindex options
4366
4367@table @gcctabopt
4368@include cppopts.texi
4369@include cppdiropts.texi
4370@include cppwarnopts.texi
4371@end table
4372@c man end
4373
4374@node Environment Variables
4375@chapter Environment Variables
4376@cindex environment variables
4377@c man begin ENVIRONMENT
4378
4379This section describes the environment variables that affect how CPP
4380operates.  You can use them to specify directories or prefixes to use
4381when searching for include files, or to control dependency output.
4382
4383Note that you can also specify places to search using options such as
4384@option{-I}, and control dependency output with options like
4385@option{-M} (@pxref{Invocation}).  These take precedence over
4386environment variables, which in turn take precedence over the
4387configuration of GCC@.
4388
4389@include cppenv.texi
4390@c man end
4391
4392@page
4393@include fdl.texi
4394
4395@page
4396@node Index of Directives
4397@unnumbered Index of Directives
4398@printindex fn
4399
4400@node Option Index
4401@unnumbered Option Index
4402@noindent
4403CPP's command-line options and environment variables are indexed here
4404without any initial @samp{-} or @samp{--}.
4405@printindex op
4406
4407@page
4408@node Concept Index
4409@unnumbered Concept Index
4410@printindex cp
4411
4412@bye
4413