1 \input texinfo @c -*-texinfo-*-
3 @setfilename standards.text
4 @settitle GNU Coding Standards
10 Copyright (C) 1992 Free Software Foundation
11 Permission is granted to make and distribute verbatim copies of
12 this manual provided the copyright notice and this permission notice
13 are preserved on all copies.
16 Permission is granted to process this file through TeX and print the
17 results, provided the printed document carries copying permission
18 notice identical to this one except for the removal of this paragraph
19 (this paragraph not being relevant to the printed manual).
22 Permission is granted to copy and distribute modified versions of this
23 manual under the conditions for verbatim copying, provided that the entire
24 resulting derived work is distributed under the terms of a permission
25 notice identical to this one.
27 Permission is granted to copy and distribute translations of this manual
28 into another language, under the above conditions for modified versions,
29 except that this permission notice may be stated in a translation approved
30 by the Free Software Foundation.
35 @titlefont{GNU Coding Standards}
36 @author{Richard Stallman}
37 @author{last updated 1 Jul 1992}
38 @c Note date also appears below.
41 @vskip 0pt plus 1filll
42 Copyright @copyright{} 1992 Free Software Foundation
44 Permission is granted to make and distribute verbatim copies of
45 this manual provided the copyright notice and this permission notice
46 are preserved on all copies.
48 Permission is granted to copy and distribute modified versions of this
49 manual under the conditions for verbatim copying, provided that the entire
50 resulting derived work is distributed under the terms of a permission
51 notice identical to this one.
53 Permission is granted to copy and distribute translations of this manual
54 into another language, under the above conditions for modified versions,
55 except that this permission notice may be stated in a translation approved
56 by Free Software Foundation.
60 @node Top, Reading Non-Free Code, (dir), (dir)
63 Last updated 1 Jul 1992.
64 @c Note date also appears above.
68 * Reading Non-Free Code:: Referring to Proprietary Programs
69 * Contributions:: Accepting Contributions
70 * Change Logs:: Recording Changes
71 * Compatibility:: Compatibility with Other Implementations
72 * Makefiles:: Makefile Conventions
73 * Configuration:: How Configuration Should Work
74 * Source Language:: Using Languages Other Than C
75 * Formatting:: Formatting Your Source Code
76 * Comments:: Commenting Your Work
77 * Syntactic Conventions:: Clean Use of C Constructs
78 * Names:: Naming Variables and Functions
79 * Using Extensions:: Using Non-standard Features
80 * Semantics:: Program Behaviour for All Programs
81 * Errors:: Formatting Error Messages
82 * Libraries:: Library Behaviour
83 * Portability:: Portability As It Applies to GNU
84 * User Interfaces:: Standards for Command Line Interfaces
85 * Documentation:: Documenting Programs
86 * Releases:: Making Releases
89 @node Reading Non-Free Code
90 @chapter Referring to Proprietary Programs
92 Don't in any circumstances refer to Unix source code for or during
93 your work on GNU! (Or to any other proprietary programs.)
95 If you have a vague recollection of the internals of a Unix program,
96 this does not absolutely mean you can't write an imitation of it, but
97 do try to organize the imitation internally along different lines,
98 because this is likely to make the details of the Unix version
99 irrelevant and dissimilar to your results.
101 For example, Unix utilities were generally optimized to minimize
102 memory use; if you go for speed instead, your program will be very
103 different. You could keep the entire input file in core and scan it
104 there instead of using stdio. Use a smarter algorithm discovered more
105 recently than the Unix program. Eliminate use of temporary files. Do
106 it in one pass instead of two (we did this in the assembler).
108 Or, on the contrary, emphasize simplicity instead of speed. For some
109 applications, the speed of today's computers makes simpler algorithms
112 Or go for generality. For example, Unix programs often have static
113 tables or fixed-size strings, which make for arbitrary limits; use
114 dynamic allocation instead. Make sure your program handles NULs and
115 other funny characters in the input files. Add a programming language
116 for extensibility and write part of the program in that language.
118 Or turn some parts of the program into independently usable libraries.
119 Or use a simple garbage collector instead of tracking precisely when
120 to free memory, or use a new GNU facility such as obstacks.
124 @chapter Accepting Contributions
126 If someone else sends you a piece of code to add to the program you are
127 working on, we need legal papers to use it---the same sort of legal
128 papers we will need to get from you. @emph{Each} significant
129 contributor to a program must sign some sort of legal papers in order
130 for us to have clear title to the program. The main author alone is not
133 So, before adding in any contributions from other people, tell us
134 so we can arrange to get the papers. Then wait until we tell you
135 that we have received the signed papers, before you actually use the
138 This applies both before you release the program and afterward. If
139 you receive diffs to fix a bug, and they make significant change, we
140 need legal papers for it.
142 You don't need papers for changes of a few lines here or there, since
143 they are not significant for copyright purposes. Also, you don't need
144 papers if all you get from the suggestion is some ideas, not actual code
145 which you use. For example, if you write a different solution to the
146 problem, you don't need to get papers.
148 I know this is frustrating; it's frustrating for us as well. But if
149 you don't wait, you are going out on a limb---for example, what if the
150 contributor's employer won't sign a disclaimer? You might have to take
153 The very worst thing is if you forget to tell us about the other
154 contributor. We could be very embarrassed in court some day as a
160 Keep a change log for each directory, describing the changes made to
161 source files in that directory. The purpose of this is so that people
162 investigating bugs in the future will know about the changes that
163 might have introduced the bug. Often a new bug can be found by
164 looking at what was recently changed. More importantly, change logs
165 can help eliminate conceptual inconsistencies between different parts
166 of a program; they can give you a history of how the conflicting
169 Use the Emacs command @kbd{M-x add-change} to start a new entry in the
170 change log. An entry should have an asterisk, the name of the changed
171 file, and then in parentheses the name of the changed functions,
172 variables or whatever, followed by a colon. Then describe the changes
173 you made to that function or variable.
175 Separate unrelated entries with blank lines. When two entries
176 represent parts of the same change, so that they work together, then
177 don't put blank lines between them. Then you can omit the file name
178 and the asterisk when successive entries are in the same file.
180 Here are some examples:
183 * register.el (insert-register): Return nil.
184 (jump-to-register): Likewise.
186 * sort.el (sort-subr): Return nil.
188 * tex-mode.el (tex-bibtex-file, tex-file, tex-region):
189 Restart the tex shell if process is gone or stopped.
190 (tex-shell-running): New function.
192 * expr.c (store_one_arg): Round size up for move_block_to_reg.
193 (expand_call): Round up when emitting USE insns.
194 * stmt.c (assign_parms): Round size up for move_block_from_reg.
197 There's no need to describe here the full purpose of the changes or how
198 they work together. It is better to put this explanation in comments in
199 the code. That's why just ``New function'' is enough; there is a
200 comment with the function in the source to explain what it does.
202 However, sometimes it is useful to write one line to describe the
203 overall purpose of a large batch of changes.
205 When you change the calling sequence of a function in a simple
206 fashion, and you change all the callers of the function, there is no
207 need to make individual entries for all the callers. Just write in
208 the entry for the function being called, ``All callers changed.''
210 When you change just comments or doc strings, it is enough to write an
211 entry for the file, without mentioning the functions. Write just,
212 ``Doc fix.'' There's no need to keep a change log for documentation
213 files. This is because documentation is not susceptible to bugs that
214 are hard to fix. Documentation does not consist of parts that must
215 interact in a precisely engineered fashion; to correct an error, you
216 need not know the history of the erroneous passage.
220 @chapter Compatibility with Other Implementations
222 With certain exceptions, utility programs and libraries for GNU should
223 be upward compatible with those in Berkeley Unix, and upward compatible
224 with @sc{ANSI} C if @sc{ANSI} C specifies their behavior, and upward
225 compatible with @sc{POSIX} if @sc{POSIX} specifies their behavior.
227 When these standards conflict, it is useful to offer compatibility
228 modes for each of them.
230 @sc{ANSI} C and @sc{POSIX} prohibit many kinds of extensions. Feel
231 free to make the extensions anyway, and include a @samp{--ansi} or
232 @samp{--compatible} option to turn them off. However, if the extension
233 has a significant chance of breaking any real programs or scripts,
234 then it is not really upward compatible. Try to redesign its
237 When a feature is used only by users (not by programs or command
238 files), and it is done poorly in Unix, feel free to replace it
239 completely with something totally different and better. (For example,
240 vi is replaced with Emacs.) But it is nice to offer a compatible
241 feature as well. (There is a free vi clone, so we offer it.)
243 Additional useful features not in Berkeley Unix are welcome.
244 Additional programs with no counterpart in Unix may be useful,
245 but our first priority is usually to duplicate what Unix already
250 @chapter Makefile Conventions
252 This chapter describes conventions for writing Makefiles.
257 * Command Variables::
258 * Directory Variables::
261 @node Makefile Basics
262 @section General Conventions for Makefiles
264 Every Makefile should contain this line:
271 to avoid trouble on systems where the @code{SHELL} variable might be
272 inherited from the environment.
274 Don't assume that @file{.} is in the path for command execution. When
275 you need to run programs that are a part of your package during the
276 make, please make sure that it uses @file{./} if the program is built as
277 part of the make or @file{$(srcdir)/} if the file is an unchanging part
278 of the source code. Without one of these prefixes, the current search
281 The distinction between @file{./} and @file{$(srcdir)/} is important
282 when using the @samp{--srcdir} option to @file{configure}. A rule of
286 foo.1 : foo.man sedscript
287 sed -e sedscript foo.man > foo.1
291 will fail when the current directory is not the source directory,
292 because @file{foo.man} and @file{sedscript} are not in the current
295 Relying on @samp{VPATH} to find the source file will work in the case
296 where there is a single dependency file, since the @file{make} automatic
297 variable @samp{$<} will represent the source file wherever it is. A
302 $(CC) $(CFLAGS) -I. -I$(srcdir) -c bar.c -o foo.o
306 should instead be written as
310 $(CC) $(CFLAGS) $< -o $@
314 in order to allow @samp{VPATH} to work correctly. When the target has
315 multiple dependencies, using an explicit @samp{$(srcdir)} is the easiest
316 way to make the rule work well. For example, the target above for
317 @file{foo.1} is best written as:
320 foo.1 : foo.man sedscript
321 sed -s $(srcdir)/sedscript $(srcdir)/foo.man > foo.1
324 @node Standard Targets
325 @section Standard Targets for Users
327 All GNU programs should have the following targets in their Makefiles:
331 Compile the entire program.
334 Compile the program and copy the executables, libraries, and so on to
335 the file names where they should reside for actual use. If there is a
336 simple test to verify that a program is properly installed then run that
339 Use @samp{-} before any command for installing a man page, so that
340 @code{make} will ignore any errors. This is in case there are systems
341 that don't have the Unix man page documentation system installed.
344 Delete all files from the current directory that are normally created by
345 building the program. Don't delete the files that record the
346 configuration. Also preserve files that could be made by building, but
347 normally aren't because the distribution comes with them.
349 Delete @file{.dvi} files here if they are not part of the distribution.
352 Delete all files from the current directory that are created by
353 configuring or building the program. If you have unpacked the source
354 and built the program without creating any other files, @samp{make
355 distclean} should leave only the files that were in the distribution.
358 Like @samp{clean}, but may refrain from deleting a few files that people
359 normally don't want to recompile. For example, the @samp{mostlyclean}
360 target for GCC does not delete @file{libgcc.a}, because recompiling it
361 is rarely necessary and takes a lot of time.
364 Delete everything from the current directory that can be reconstructed
365 with this Makefile. This typically includes everything deleted by
366 distclean, plus more: C source files produced by Bison, tags tables,
367 info files, and so on.
370 Update a tags table for this program.
373 Create a distribution tar file for this program. The tar file should be
374 set up so that the file names in the tar file start with a subdirectory
375 name which is the name of the package it is a distribution for. This
376 name can include the version number.
378 For example, the distribution tar file of GCC version 1.40 unpacks into
379 a subdirectory named @file{gcc-1.40}.
381 The easiest way to do this is to create a subdirectory appropriately
382 named, use @code{ln} or @code{cp} to install the proper files in it, and
383 then @code{tar} that subdirectory.
385 The @code{dist} target should explicitly depend on all non-source files
386 that are in the distribution, to make sure they are up to date in the
387 distribution. @xref{Releases}.
390 Perform self-tests (if any). The user must build the program before
391 running the tests, but need not install the program; you should write
392 the self-tests so that they work when the program is built but not
396 @node Command Variables
397 @section Variables for Specifying Commands
399 Makefiles should provide variables for overriding certain commands, options,
402 In particular, you should run most utility programs via variables.
403 Thus, if you use Bison, have a variable named @code{BISON} whose default
404 value is set with @samp{BISON = bison}, and refer to it with
405 @code{$(BISON)} whenever you need to use Bison.
407 File management utilities such as @code{ln}, @code{rm}, @code{mv}, and
408 so on, need not be referred to through variables in this way, since users
409 don't need to replace them with other programs.
411 Each program-name variable should come with an options variable that is
412 used to supply options to the program. Append @samp{FLAGS} to the
413 program-name variable name to get the options variable name---for
414 example, @code{BISONFLAGS}. (The name @code{CFLAGS} is an exception to
415 this rule, but we keep it because it is standard.) Use @code{CPPFLAGS}
416 in any compilation command that runs the preprocessor, and use
417 @code{LDFLAGS} in any compilation command that does linking as well as
418 in any direct use of @code{ld}.
420 If there are C compiler options that @emph{must} be used for proper
421 compilation of certain files, do not include them in @code{CFLAGS}.
422 Users expect to be able to specify @code{CFLAGS} freely themselves.
423 Instead, arrange to pass the necessary options to the C compiler
424 independently of @code{CFLAGS}, by writing them explicitly in the
425 compilation commands or by defining an implicit rule, like this:
429 ALL_CFLAGS = $(CFLAGS) -I.
431 $(CC) -c $(ALL_CFLAGS) $(CPPFLAGS) $<
434 Do include the @samp{-g} option in @code{CFLAGS}, because that is not
435 @emph{required} for proper compilation. You can consider it a default
436 that is only recommended. If the package is set up so that it is
437 compiled with GCC by default, then you might as well include @samp{-O}
438 in the default value of @code{CFLAGS} as well.
440 Every Makefile should define the variable @code{INSTALL}, which is the
441 basic command for installing a file into the system.
443 Every Makefile should also define variables @code{INSTALL_PROGRAM} and
444 @code{INSTALL_DATA}. (The default for each of these should be
445 @code{$(INSTALL)}.) Then it should use those variables as the commands
446 for actual installation, for executables and nonexecutables
447 respectively. Use these variables as follows:
450 $(INSTALL_PROGRAM) foo $(bindir)/foo
451 $(INSTALL_DATA) libfoo.a $(libdir)/libfoo.a
455 Always use a file name, not a directory name, as the second argument of
456 the installation commands. Use a separate command for each file to be
459 @node Directory Variables
460 @section Variables for Installation Directories
462 Installation directories should always be named by variables, so it is
463 easy to install in a nonstandard place. The standard names for these
468 A prefix used in constructing the default values of the variables listed
469 below. The default value of @code{prefix} should be @file{/usr/local}
473 A prefix used in constructing the default values of the some of the
474 variables listed below. The default value of @code{exec_prefix} should
477 Generally, @code{$(exec_prefix)} is used for directories that contain
478 machine-specific files (such as executables and subroutine libraries),
479 while @code{$(prefix)} is used directly for other directories.
482 The directory for installing executable programs that users can run.
483 This should normally be @file{/usr/local/bin}, but it should be written
484 as @file{$(exec_prefix)/bin}.
487 The directory for installing executable files to be run by the program
488 rather than by users. Object files and libraries of object code should
489 also go in this directory. The idea is that this directory is used for
490 files that pertain to a specific machine architecture, but need not be
491 in the path for commands. The value of @code{libdir} should normally be
492 @file{/usr/local/lib}, but it should be written as
493 @file{$(exec_prefix)/lib}.
496 The directory for installing read-only data files which the programs
497 refer to while they run. This directory is used for files which are
498 independent of the type of machine being used. This should normally be
499 @file{/usr/local/lib}, but it should be written as
500 @file{$(prefix)/lib}.
503 The directory for installing data files which the programs modify while
504 they run. These files should be independent of the type of machine
505 being used, and it should be possible to share them among machines at a
506 network installation. This should normally be @file{/usr/local/lib},
507 but it should be written as @file{$(prefix)/lib}.
510 The directory for installing @samp{#include} header files to be included
511 by user programs. This should normally be @file{/usr/local/include},
512 but it should be written as @file{$(prefix)/include}.
514 Most compilers other than GCC do not look for header files in
515 @file{/usr/local/include}. So installing the header files this way is
516 only useful with GCC. Sometimes this is not a problem because some
517 libraries are only really intended to work with GCC. But some libraries
518 are intended to work with other compilers. They should install their
519 header files in two places, one specified by @code{includedir} and one
520 specified by @code{oldincludedir}.
523 The directory for installing @samp{#include} header files for use with
524 compilers other than GCC. This should normally be @file{/usr/include}.
526 The Makefile commands should check whether the value of
527 @code{oldincludedir} is empty. If it is, they should not try to use
528 it; they should cancel the second installation of the header files.
531 The directory for installing the man pages (if any) for this package.
532 It should include the suffix for the proper section of the
533 manual---usually @samp{1} for a utility.
536 The directory for installing section 1 man pages.
538 The directory for installing section 2 man pages.
540 Use these names instead of @samp{mandir} if the package needs to install man
541 pages in more than one section of the manual.
543 @strong{Don't make the primary documentation for any GNU software be a
544 man page. Write a manual in Texinfo instead. Man pages are just for
545 the sake of people running GNU software on Unix, which is a secondary
549 The file name extension for the installed man page. This should contain
550 a period followed by the appropriate digit.
553 The directory for installing the info files for this package. By
554 default, it should be @file{/usr/local/info}, but it should be written
555 as @file{$(prefix)/info}.
558 The directory for the sources being compiled. The value of this
559 variable is normally inserted by the @code{configure} shell script.
565 # Common prefix for installation directories.
566 # NOTE: This directory must exist when you start installation.
568 exec_prefix = $(prefix)
569 # Directory in which to put the executable for the command `gcc'
570 bindir = $(exec_prefix)/bin
571 # Directory in which to put the directories used by the compiler.
572 libdir = $(exec_prefix)/lib
573 # Directory in which to put the Info files.
574 infodir = $(prefix)/info
577 If your program installs a large number of files into one of the
578 standard user-specified directories, it might be useful to group them
579 into a subdirectory particular to that program. If you do this, you
580 should write the @code{install} rule to create these subdirectories.
582 Do not expect the user to include the subdirectory name in the value of
583 any of the variables listed above. The idea of having a uniform set of
584 variable names for installation directories is to enable the user to
585 specify the exact same values for several different GNU packages. In
586 order for this to be useful, all the packages must be designed so that
587 they will work sensibly when the user does so.
590 @chapter How Configuration Should Work
592 Each GNU distribution should come with a shell script named
593 @code{configure}. This script is given arguments which describe the
594 kind of machine and system you want to compile the program for.
596 The @code{configure} script must record the configuration options so
597 that they affect compilation.
599 One way to do this is to make a link from a standard name such as
600 @file{config.h} to the proper configuration file for the chosen system.
601 If you use this technique, the distribution should @emph{not} contain a
602 file named @file{config.h}. This is so that people won't be able to
603 build the program without configuring it first.
605 Another thing that @code{configure} can do is to edit the Makefile. If
606 you do this, the distribution should @emph{not} contain a file named
607 @file{Makefile}. Instead, include a file @file{Makefile.in} which
608 contains the input used for editing. Once again, this is so that people
609 won't be able to build the program without configuring it first.
611 If @code{configure} does write the @file{Makefile}, then @file{Makefile}
612 should have a target named @file{Makefile} which causes @code{configure}
613 to be rerun, setting up the same configuration that was set up last
614 time. The files that @code{configure} reads should be listed as
615 dependencies of @file{Makefile}.
617 All the files which are output from the @code{configure} script should
618 have comments at the beginning explaining that they were generated
619 automatically using @code{configure}. This is so that users won't think
620 of trying to edit them by hand.
622 The @code{configure} script should write a file named @file{config.status}
623 which describes which configuration options were specified when the
624 program was last configured. This file should be a shell script which,
625 if run, will recreate the same configuration.
627 The @code{configure} script should accept an option of the form
628 @samp{--srcdir=@var{dirname}} to specify the directory where sources are found
629 (if it is not the current directory). This makes it possible to build
630 the program in a separate directory, so that the actual source directory
633 If the user does not specify @samp{--srcdir}, then @code{configure} should
634 check both @file{.} and @file{..} to see if it can find the sources. If
635 it finds the sources in one of these places, it should use them from
636 there. Otherwise, it should report that it cannot find the sources, and
637 should exit with nonzero status.
639 Usually the easy way to support @samp{--srcdir} is by editing a
640 definition of @code{VPATH} into the Makefile. Some rules may need to
641 refer explicitly to the specified source directory. To make this
642 possible, @code{configure} can add to the Makefile a variable named
643 @code{srcdir} whose value is precisely the specified directory.
645 The @code{configure} script should also take an argument which specifies the
646 type of system to build the program for. This argument should look like
650 @var{cpu}-@var{company}-@var{system}
653 For example, a Sun 3 might be @samp{m68k-sun-sunos4.1}.
655 The @code{configure} script needs to be able to decode all plausible
656 alternatives for how to describe a machine. Thus, @samp{sun3-sunos4.1}
657 would be a valid alias. So would @samp{sun3-bsd4.2}, since SunOS is
658 basically @sc{BSD} and no other @sc{BSD} system is used on a Sun. For many
659 programs, @samp{vax-dec-ultrix} would be an alias for
660 @samp{vax-dec-bsd}, simply because the differences between Ultrix and
661 @sc{BSD} are rarely noticeable, but a few programs might need to distinguish
664 There is a shell script called @file{config.sub} that you can use
665 as a subroutine to validate system types and canonicalize aliases.
667 Other options are permitted to specify in more detail the software
668 or hardware are present on the machine:
671 @item --with-@var{package}
672 The package @var{package} will be installed, so configure this package
673 to work with @var{package}.
675 Possible values of @var{package} include @samp{x}, @samp{gnu-as} (or
676 @samp{gas}), @samp{gnu-ld}, @samp{gnu-libc}, and @samp{gdb}.
679 The target machine has no floating point processor.
682 The target machine assembler is GAS, the GNU assembler.
683 This is obsolete; use @samp{--with-gnu-as} instead.
686 The target machine has the X Window system installed.
687 This is obsolete; use @samp{--with-x} instead.
690 All @code{configure} scripts should accept all of these ``detail''
691 options, whether or not they make any difference to the particular
692 package at hand. In particular, they should accept any option that
693 starts with @samp{--with-}. This is so users will be able to configure
694 an entire GNU source tree at once with a single set of options.
696 Packages that perform part of compilation may support cross-compilation.
697 In such a case, the host and target machines for the program may be
698 different. The @code{configure} script should normally treat the
699 specified type of system as both the host and the target, thus producing
700 a program which works for the same type of machine that it runs on.
702 The way to build a cross-compiler, cross-assembler, or what have you, is
703 to specify the option @samp{--host=@var{hosttype}} when running
704 @code{configure}. This specifies the host system without changing the
705 type of target system. The syntax for @var{hosttype} is the same as
708 Programs for which cross-operation is not meaningful need not accept the
709 @samp{--host} option, because configuring an entire operating system for
710 cross-operation is not a meaningful thing.
712 Some programs have ways of configuring themselves automatically. If
713 your program is set up to do this, your @code{configure} script can simply
714 ignore most of its arguments.
717 @node Source Language
718 @chapter Using Languages Other Than C
720 Using a language other than C is like using a non-standard feature: it
721 will cause trouble for users. Even if GCC supports the other language,
722 users may find it inconvenient to have to install the compiler for that
723 other language in order to build your program. So please write in C.
725 There are three exceptions for this rule:
729 It is okay to use a special language if the same program contains an
730 interpreter for that language.
732 Thus, it is not a problem that GNU Emacs contains code written in Emacs
733 Lisp, because it comes with a Lisp interpreter.
736 It is okay to use another language in a tool specifically intended for
737 use with that language.
739 This is okay because the only people who want to build the tool will be
740 those who have installed the other language anyway.
743 If an application is not of extremely widespread interest, then perhaps
744 it's not important if the application is inconvenient to install.
748 @chapter Formatting Your Source Code
750 It is important to put the open-brace that starts the body of a C
751 function in column zero, and avoid putting any other open-brace or
752 open-parenthesis or open-bracket in column zero. Several tools look
753 for open-braces in column zero to find the beginnings of C functions.
754 These tools will not work on code not formatted that way.
756 It is also important for function definitions to start the name of the
757 function in column zero. This helps people to search for function
758 definitions, and may also help certain tools recognize them. Thus,
759 the proper format is this:
763 concat (s1, s2) /* Name starts in column zero here */
765 @{ /* Open brace in column zero here */
771 or, if you want to use @sc{ANSI} C, format the definition like this:
775 concat (char *s1, char *s2)
781 In @sc{ANSI} C, if the arguments don't fit nicely on one line,
786 lots_of_args (int an_integer, long a_long, short a_short,
787 double a_double, float a_float)
791 For the body of the function, we prefer code formatted like this:
807 We find it easier to read a program when it has spaces before the
808 open-parentheses and after the commas. Especially after the commas.
810 When you split an expression into multiple lines, split it
811 before an operator, not after one. Here is the right way:
814 if (foo_this_is_long && bar > win (x, y, z)
815 && remaining_condition)
818 Try to avoid having two operators of different precedence at the same
819 level of indentation. For example, don't write this:
822 mode = (inmode[j] == VOIDmode
823 || GET_MODE_SIZE (outmode[j]) > GET_MODE_SIZE (inmode[j])
824 ? outmode[j] : inmode[j]);
827 Instead, use extra parentheses so that the indentation shows the nesting:
830 mode = ((inmode[j] == VOIDmode
831 || (GET_MODE_SIZE (outmode[j]) > GET_MODE_SIZE (inmode[j])))
832 ? outmode[j] : inmode[j]);
835 Insert extra parentheses so that Emacs will indent the code properly.
836 For example, the following indentation looks nice if you do it by hand,
837 but Emacs would mess it up:
840 v = rup->ru_utime.tv_sec*1000 + rup->ru_utime.tv_usec/1000
841 + rup->ru_stime.tv_sec*1000 + rup->ru_stime.tv_usec/1000;
844 But adding a set of parentheses solves the problem:
847 v = (rup->ru_utime.tv_sec*1000 + rup->ru_utime.tv_usec/1000
848 + rup->ru_stime.tv_sec*1000 + rup->ru_stime.tv_usec/1000);
851 Format do-while statements like this:
861 Please use formfeed characters (control-L) to divide the program into
862 pages at logical places (but not within a function). It does not matter
863 just how long the pages are, since they do not have to fit on a printed
864 page. The formfeeds should appear alone on lines by themselves.
868 @chapter Commenting Your Work
870 Every program should start with a comment saying briefly what it is for.
871 Example: @samp{fmt - filter for simple filling of text}.
873 Please put a comment on each function saying what the function does,
874 what sorts of arguments it gets, and what the possible values of
875 arguments mean and are used for. It is not necessary to duplicate in
876 words the meaning of the C argument declarations, if a C type is being
877 used in its customary fashion. If there is anything nonstandard about
878 its use (such as an argument of type @code{char *} which is really the
879 address of the second character of a string, not the first), or any
880 possible values that would not work the way one would expect (such as,
881 that strings containing newlines are not guaranteed to work), be sure
884 Also explain the significance of the return value, if there is one.
886 Please put two spaces after the end of a sentence in your comments, so
887 that the Emacs sentence commands will work. Also, please write
888 complete sentences and capitalize the first word. If a lower-case
889 identifer comes at the beginning of a sentence, don't capitalize it!
890 Changing the spelling makes it a different identifier. If you don't
891 like starting a sentence with a lower case letter, write the sentence
892 differently (e.g. ``The identifier lower-case is @dots{}'').
894 The comment on a function is much clearer if you use the argument
895 names to speak about the argument values. The variable name itself
896 should be lower case, but write it in upper case when you are speaking
897 about the value rather than the variable itself. Thus, ``the inode
898 number @var{node_num}'' rather than ``an inode''.
900 There is usually no purpose in restating the name of the function in
901 the comment before it, because the reader can see that for himself.
902 There might be an exception when the comment is so long that the function
903 itself would be off the bottom of the screen.
905 There should be a comment on each static variable as well, like this:
908 /* Nonzero means truncate lines in the display;
909 zero means continue them. */
914 Every @samp{#endif} should have a comment, except in the case of short
915 conditionals (just a few lines) that are not nested. The comment should
916 state the condition of the conditional that is ending, @emph{including
917 its sense}. @samp{#else} should have a comment describing the condition
918 @emph{and sense} of the code that follows. For example:
929 but, by contrast, write the comments this way for a @samp{#ifndef}:
940 @node Syntactic Conventions
941 @chapter Clean Use of C Constructs
943 Please explicitly declare all arguments to functions.
944 Don't omit them just because they are ints.
946 Declarations of external functions and functions to appear later
947 in the source file should all go in one place near the beginning of
948 the file (somewhere before the first function definition in the file),
949 or else should go in a header file. Don't put extern declarations
952 Don't declare multiple variables in one declaration that spans lines.
953 Start a new declaration on each line, instead. For example, instead
977 (If they are global variables, each should have a comment preceding it
980 When you have an if-else statement nested in another if statement,
981 always put braces around the if-else. Thus, never write like this:
1004 If you have an if statement nested inside of an else statement,
1005 either write @code{else if} on one line, like this,
1015 with its then-part indented like the preceding then-part, or write the
1016 nested if within braces like this:
1028 Don't declare both a structure tag and variables or typedefs in the
1029 same declaration. Instead, declare the structure tag separately
1030 and then use it to declare the variables or typedefs.
1032 Try to avoid assignments inside if-conditions. For example, don't
1036 if ((foo = (char *) malloc (sizeof *foo)) == 0)
1037 fatal ("virtual memory exhausted");
1041 instead, write this:
1044 foo = (char *) malloc (sizeof *foo);
1046 fatal ("virtual memory exhausted");
1049 Don't make the program ugly to placate lint. Please don't insert any
1050 casts to void. Zero without a cast is perfectly fine as a null
1055 @chapter Naming Variables and Functions
1057 Please use underscores to separate words in a name, so that the Emacs
1058 word commands can be useful within them. Stick to lower case; reserve
1059 upper case for macros and enum constants, and for name-prefixes that
1060 follow a uniform convention.
1062 For example, you should use names like @code{ignore_space_change_flag};
1063 don't use names like @code{iCantReadThis}.
1065 Variables that indicate whether command-line options have been
1066 specified should be named after the meaning of the option, not after
1067 the option-letter. A comment should state both the exact meaning of
1068 the option and its letter. For example,
1071 /* Ignore changes in horizontal whitespace (-b). */
1072 int ignore_space_change_flag;
1075 When you want to define names with constant integer values, use
1076 @code{enum} rather than @samp{#define}. GDB knows about enumeration
1079 Use file names of 14 characters or less, to avoid creating gratuitous
1080 problems on System V.
1083 @node Using Extensions
1084 @chapter Using Non-standard Features
1086 Many GNU facilities that already exist support a number of convenient
1087 extensions over the comparable Unix facilities. Whether to use these
1088 extensions in implementing your program is a difficult question.
1090 On the one hand, using the extensions can make a cleaner program.
1091 On the other hand, people will not be able to build the program
1092 unless the other GNU tools are available. This might cause the
1093 program to work on fewer kinds of machines.
1095 With some extensions, it might be easy to provide both alternatives.
1096 For example, you can define functions with a ``keyword'' @code{INLINE}
1097 and define that as a macro to expand into either @code{inline} or
1098 nothing, depending on the compiler.
1100 In general, perhaps it is best not to use the extensions if you can
1101 straightforwardly do without them, but to use the extensions if they
1102 are a big improvement.
1104 An exception to this rule are the large, established programs (such as
1105 Emacs) which run on a great variety of systems. Such programs would
1106 be broken by use of GNU extensions.
1108 Another exception is for programs that are used as part of
1109 compilation: anything that must be compiled with other compilers in
1110 order to bootstrap the GNU compilation facilities. If these require
1111 the GNU compiler, then no one can compile them without having them
1112 installed already. That would be no good.
1114 Since most computer systems do not yet implement @sc{ANSI} C, using the
1115 @sc{ANSI} C features is effectively using a GNU extension, so the
1116 same considerations apply. (Except for @sc{ANSI} features that we
1117 discourage, such as trigraphs---don't ever use them.)
1120 @chapter Program Behaviour for All Programs
1122 Avoid arbitrary limits on the length or number of @emph{any} data
1123 structure, including filenames, lines, files, and symbols, by allocating
1124 all data structures dynamically. In most Unix utilities, ``long lines
1125 are silently truncated''. This is not acceptable in a GNU utility.
1127 Utilities reading files should not drop NUL characters, or any other
1128 nonprinting characters @emph{including those with codes above 0177}. The
1129 only sensible exceptions would be utilities specifically intended for
1130 interface to certain types of printers that can't handle those characters.
1132 Check every system call for an error return, unless you know you wish to
1133 ignore errors. Include the system error text (from @code{perror} or
1134 equivalent) in @emph{every} error message resulting from a failing
1135 system call, as well as the name of the file if any and the name of the
1136 utility. Just ``cannot open foo.c'' or ``stat failed'' is not
1139 Check every call to @code{malloc} or @code{realloc} to see if it
1140 returned zero. Check @code{realloc} even if you are making the block
1141 smaller; in a system that rounds block sizes to a power of 2,
1142 @code{realloc} may get a different block if you ask for less space.
1144 In Unix, @code{realloc} can destroy the storage block if it returns
1145 zero. GNU @code{realloc} does not have this bug: if it fails, the
1146 original block is unchanged. Feel free to assume the bug is fixed. If
1147 you wish to run your program on Unix, and wish to avoid lossage in this
1148 case, you can use the GNU @code{malloc}.
1150 You must expect @code{free} to alter the contents of the block that was
1151 freed. Anything you want to fetch from the block, you must fetch before
1152 calling @code{free}.
1154 Use @code{getopt_long} to decode arguments, unless the argument syntax
1155 makes this unreasonable.
1157 When static storage is to be written in during program execution, use
1158 explicit C code to initialize it. Reserve C initialized declarations
1159 for data that will not be changed.
1161 Try to avoid low-level interfaces to obscure Unix data structures (such
1162 as file directories, utmp, or the layout of kernel memory), since these
1163 are less likely to work compatibly. If you need to find all the files
1164 in a directory, use @code{readdir} or some other high-level interface.
1165 These will be supported compatibly by GNU.
1167 By default, the GNU system will provide the signal handling functions of
1168 @sc{BSD} and of @sc{POSIX}. So GNU software should be written to use
1171 In error checks that detect ``impossible'' conditions, just abort.
1172 There is usually no point in printing any message. These checks
1173 indicate the existence of bugs. Whoever wants to fix the bugs will have
1174 to read the source code and run a debugger. So explain the problem with
1175 comments in the source. The relevant data will be in variables, which
1176 are easy to examine with the debugger, so there is no point moving them
1181 @chapter Formatting Error Messages
1183 Error messages from compilers should look like this:
1186 @var{source-file-name}:@var{lineno}: @var{message}
1189 Error messages from other noninteractive programs should look like this:
1192 @var{program}:@var{source-file-name}:@var{lineno}: @var{message}
1196 when there is an appropriate source file, or like this:
1199 @var{program}: @var{message}
1203 when there is no relevant source file.
1205 In an interactive program (one that is reading commands from a
1206 terminal), it is better not to include the program name in an error
1207 message. The place to indicate which program is running is in the
1208 prompt or with the screen layout. (When the same program runs with
1209 input from a source other than a terminal, it is not interactive and
1210 would do best to print error messages using the noninteractive style.)
1212 The string @var{message} should not begin with a capital letter when
1213 it follows a program name and/or filename. Also, it should not end
1216 Error messages from interactive programs, and other messages such as
1217 usage messages, should start with a capital letter. But they should not
1222 @chapter Library Behaviour
1224 Try to make library functions reentrant. If they need to do dynamic
1225 storage allocation, at least try to avoid any nonreentrancy aside from
1226 that of @code{malloc} itself.
1228 Here are certain name conventions for libraries, to avoid name
1231 Choose a name prefix for the library, more than two characters long.
1232 All external function and variable names should start with this
1233 prefix. In addition, there should only be one of these in any given
1234 library member. This usually means putting each one in a separate
1237 An exception can be made when two external symbols are always used
1238 together, so that no reasonable program could use one without the
1239 other; then they can both go in the same file.
1241 External symbols that are not documented entry points for the user
1242 should have names beginning with @samp{_}. They should also contain
1243 the chosen name prefix for the library, to prevent collisions with
1244 other libraries. These can go in the same files with user entry
1247 Static functions and variables can be used as you like and need not
1248 fit any naming convention.
1252 @chapter Portability As It Applies to GNU
1254 Much of what is called ``portability'' in the Unix world refers to
1255 porting to different Unix versions. This is not relevant to GNU
1256 software, because its purpose is to run on top of one and only
1257 one kernel, the GNU kernel, compiled with one and only one C
1258 compiler, the GNU C compiler. The amount and kinds of variation
1259 among GNU systems on different cpu's will be like the variation
1260 among Berkeley 4.3 systems on different cpu's.
1262 It is difficult to be sure exactly what facilities the GNU kernel
1263 will provide, since it isn't finished yet. Therefore, assume you can
1264 use anything in 4.3; just avoid using the format of semi-internal data
1265 bases (e.g., directories) when there is a higher-level alternative
1268 You can freely assume any reasonably standard facilities in the C
1269 language, libraries or kernel, because we will find it necessary to
1270 support these facilities in the full GNU system, whether or not we
1271 have already done so. The fact that there may exist kernels or C
1272 compilers that lack these facilities is irrelevant as long as the GNU
1273 kernel and C compiler support them.
1275 It remains necessary to worry about differences among cpu types, such
1276 as the difference in byte ordering and alignment restrictions. It's
1277 unlikely that 16-bit machines will ever be supported by GNU, so there
1278 is no point in spending any time to consider the possibility that an
1279 int will be less than 32 bits.
1281 You can assume that all pointers have the same format, regardless
1282 of the type they point to, and that this is really an integer.
1283 There are some weird machines where this isn't true, but they aren't
1284 important; don't waste time catering to them. Besides, eventually
1285 we will put function prototypes into all GNU programs, and that will
1286 probably make your program work even on weird machines.
1288 Since some important machines (including the 68000) are big-endian,
1289 it is important not to assume that the address of an int object
1290 is also the address of its least-significant byte. Thus, don't
1291 make the following mistake:
1296 while ((c = getchar()) != EOF)
1297 write(file_descriptor, &c, 1);
1300 You can assume that it is reasonable to use a meg of memory. Don't
1301 strain to reduce memory usage unless it can get to that level. If
1302 your program creates complicated data structures, just make them in
1303 core and give a fatal error if malloc returns zero.
1305 If a program works by lines and could be applied to arbitrary
1306 user-supplied input files, it should keep only a line in memory, because
1307 this is not very hard and users will want to be able to operate on input
1308 files that are bigger than will fit in core all at once.
1311 @node User Interfaces
1312 @chapter Standards for Command Line Interfaces
1314 Please don't make the behavior of a utility depend on the name used
1315 to invoke it. It is useful sometimes to make a link to a utility
1316 with a different name, and that should not change what it does.
1318 Instead, use a run time option or a compilation switch or both
1319 to select among the alternate behaviors.
1321 It is a good idea to follow the @sc{POSIX} guidelines for the
1322 command-line options of a program. The easiest way to do this is to use
1323 @code{getopt} to parse them. Note that the GNU version of @code{getopt}
1324 will normally permit options anywhere among the arguments unless the
1325 special argument @samp{--} is used. This is not what @sc{POSIX}
1326 specifies; it is a GNU extension.
1328 Please define long-named options that are equivalent to the
1329 single-letter Unix-style options. We hope to make GNU more user
1330 friendly this way. This is easy to do with the GNU function
1333 It is usually a good idea for file names given as ordinary arguments
1334 to be input files only; any output files would be specified using
1335 options (preferably @samp{-o}). Even if you allow an output file name
1336 as an ordinary argument for compatibility, try to provide a suitable
1337 option as well. This will lead to more consistency among GNU
1338 utilities, so that there are fewer idiosyncracies for users to
1341 Programs should support an option @samp{--version} which prints the
1342 program's version number, and an option @samp{--help} which prints
1343 option usage information.
1347 @chapter Documenting Programs
1349 Please use Texinfo for documenting GNU programs. See the Texinfo
1350 manual, either the hardcopy or the version in the GNU Emacs Info
1351 sub-system (@kbd{C-h i}).
1353 See existing GNU texinfo files (e.g. those under the @file{man/}
1354 directory in the GNU Emacs Distribution) for examples.
1356 The title page of the manual should state the version of the program
1357 which the manual applies to. The Top node of the manual should also
1358 contain this information. If the manual is changing more frequently
1359 than or independent of the program, also state a version number for
1360 the manual in both of these places.
1362 The manual should document all command-line arguments and all
1363 commands. It should give examples of their use. But don't organize
1364 the manual as a list of features. Instead, organize it by the
1365 concepts a user will have before reaching that point in the manual.
1366 Address the goals that a user will have in mind, and explain how to
1371 @chapter Making Releases
1373 Package the distribution of Foo version 69.96 in a tar file named
1374 @file{foo-69.96.tar}. It should unpack into a subdirectory named
1377 Building and installing the program should never modify any of the files
1378 contained in the distribution. This means that all the files that form
1379 part of the program in any way must be classified into @dfn{source
1380 files} and @dfn{non-source files}. Source files are written by humans
1381 and never changed automatically; non-source files are produced from
1382 source files by programs under the control of the Makefile.
1384 Naturally, all the source files must be in the distribution. It is okay
1385 to include non-source files in the distribution, provided they are
1386 up-to-date and machine-independent, so that building the distribution
1387 normally will never modify them. We commonly included non-source files
1388 produced by Bison, Lex, @TeX{}, and Makeinfo; this helps avoid
1389 unnecessary dependencies between our distributions, so that users can
1390 install whichever packages they want to install.
1392 Non-source files that might actually be modified by building and
1393 installing the program should @strong{never} be included in the
1394 distribution. So if you do distribute non-source files, always make
1395 sure they are up to date when you make a new distribution.
1397 Make sure that no file name in the distribution is no more than 14
1398 characters long. Nowadays, there are systems that adhere to a foolish
1399 interpretation of the POSIX standard which holds that they should refuse
1400 to open a longer name, rather than truncating as they did in the past.
1402 Try to make sure that all the file names will be unique on MS-DOG. A
1403 name on MS-DOG consists of up to 8 characters, optionally followed by a
1404 period and up to three characters. MS-DOG will truncate extra
1405 characters both before and after the period. Thus,
1406 @file{foobarhacker.c} and @file{foobarhacker.o} are not ambiguous; they
1407 are truncated to @file{foobarha.c} and @file{foobarha.o}, which are
1410 Include in your distribution a copy of the @file{texinfo.tex} you used
1411 to test print any @file{*.texinfo} files.
1413 Likewise, if your program uses small GNU software packages like regex,
1414 getopt, obstack, or termcap, include them in the distribution file.
1415 Leaving them out would make the distribution file a little smaller at
1416 the expense of possible inconvenience to a user who doesn't know what