转自一个命令(man 9 style)

挺好的文档。

STYLE(9)               FreeBSD Kernel Developer’s Manual              STYLE(9)

NAME
     style — kernel source file style guide

DESCRIPTION
     This file specifies the preferred style for kernel source files in the
     FreeBSD source tree.  It is also a guide for the preferred userland code
     style.  Many of the style rules are implicit in the examples.  Be careful
     to check the examples before assuming that style is silent on an issue.

     /*
      * Style guide for FreeBSD.  Based on the CSRG’s KNF (Kernel Normal Form).
      *
      *      @(#)style       1.14 (Berkeley) 4/28/95
      * $FreeBSD: src/share/man/man9/style.9,v 1.121 2005/06/28 20:15:18 hmp Exp $
      */

     /*
      * VERY important single-line comments look like this.
      */

     /* Most single-line comments look like this. */

     /*
      * Multi-line comments look like this.  Make them real sentences.  Fill
      * them so they look like real paragraphs.
      */

     The copyright header should be a multi-line comment, with the first line
     of the comment having a dash after the star like so:

     /*-
      * Copyright (c) 1984-2025 John Q. Public.  All Rights Reserved.
      *
      * Long, boring license goes here, but redacted for brevity
      */

     An automatic script collects license information from the tree for all
     comments that start in the first column with “/*-”.  If you desire to
     flag indent(1) to not reformat a comment that starts in the first column
     which is not a license or copyright notice, change the dash to a star for
     those comments.  Comments starting in columns other than the first are
     never considered license statements.

     After any copyright header, there is a blank line, and the $FreeBSD$ for
     non C/C++ language source files.  Version control system ID tags should
     only exist once in a file (unlike in this one).  Non-C/C++ source files
     follow the example above, while C/C++ source files follow the one below.
     All VCS (version control system) revision identification in files
     obtained from elsewhere should be maintained, including, where applica-
     ble, multiple IDs showing a file’s history.  In general, do not edit for-
     eign IDs or their infrastructure.  Unless otherwise wrapped (such as
     “#if defined(LIBC_SCCS)”), enclose both in “#if 0 … #endif” to hide
     any uncompilable bits and to keep the IDs out of object files.  Only add
     “From: ” in front of foreign VCS IDs if the file is renamed.

     #if 0
     #ifndef lint
     static char sccsid[] = "@(#)style       1.14 (Berkeley) 4/28/95";
     #endif /* not lint */
     #endif

     #include
     __FBSDID("$FreeBSD: src/share/man/man9/style.9,v 1.121 2005/06/28 20:15:18 hmp Exp $");

     Leave another blank line before the header files.

     Kernel include files (i.e. sys/*.h) come first; normally, include
     OR , but not both.  includes
     , and it is okay to depend on that.

     #include   /* Non-local includes in angle brackets. */

     For a network program, put the network include files next.

     #include
     #include
     #include
     #include
     #include

     Do not use files in /usr/include for files in the kernel.

     Leave a blank line before the next group, the /usr/include files, which
     should be sorted alphabetically by name.

     #include

     Global pathnames are defined in .  Pathnames local to the pro-
     gram go in "pathnames.h" in the local directory.

     #include

     Leave another blank line before the user include files.

     #include "pathnames.h"          /* Local includes in double quotes. */

     Do not #define or declare names in the implementation namespace except
     for implementing application interfaces.

     The names of “unsafe” macros (ones that have side effects), and the
     names of macros for manifest constants, are all in uppercase.  The expan-
     sions of expression-like macros are either a single token or have outer
     parentheses.  Put a single tab character between the #define and the
     macro name.  If a macro is an inline expansion of a function, the func-
     tion name is all in lowercase and the macro has the same name all in
     uppercase.  Right-justify the backslashes; it makes it easier to read.
     If the macro encapsulates a compound statement, enclose it in a do loop,
     so that it can safely be used in if statements.  Any final statement-ter-
     minating semicolon should be supplied by the macro invocation rather than
     the macro, to make parsing easier for pretty-printers and editors.

     #define MACRO(x, y) do {                                               
             variable = (x) + (y);                                          
             (y) += 2;                                                      
     } while (0)

     When code is conditionally compiled using #ifdef or #if, a comment may be
     added following the matching #endif or #else to permit the reader to eas-
     ily discern where conditionally compiled code regions end.  This comment
     should be used only for (subjectively) long regions, regions greater than
     20 lines, or where a series of nested #ifdef ’s may be confusing to the
     reader.  Exceptions may be made for cases where code is conditionally not
     compiled for the purposes of lint(1), even though the uncompiled region
     may be small.  The comment should be separated from the #endif or #else
     by a single space.  For short conditionally compiled regions, a closing
     comment should not be used.

     The comment for #endif should match the expression used in the corre-
     sponding #if or #ifdef.  The comment for #else and #elif should match the
     inverse of the expression(s) used in the preceding #if and/or #elif
     statements.  In the comments, the subexpression “defined(FOO)” is
     abbreviated as “FOO”.  For the purposes of comments, “#ifndef FOO” is
     treated as “#if !defined(FOO)”.

     #ifdef KTRACE
     #include
     #endif

     #ifdef COMPAT_43
     /* A large region here, or other conditional code. */
     #else /* !COMPAT_43 */
     /* Or here. */
     #endif /* COMPAT_43 */

     #ifndef COMPAT_43
     /* Yet another large region here, or other conditional code. */
     #else /* COMPAT_43 */
     /* Or here. */
     #endif /* !COMPAT_43 */

     The project is slowly moving to use the ISO/IEC 9899:1999 (“ISO C99”)
     unsigned integer identifiers of the form uintXX_t in preference to the
     older BSD-style integer identifiers of the form u_intXX_t.  New code
     should use the former, and old code should be converted to the new form
     if other major work is being done in that area and there is no overriding
     reason to prefer the older BSD-style.  Like white-space commits, care
     should be taken in making uintXX_t only commits.

     Enumeration values are all uppercase.

     enum enumtype { ONE, TWO } et;

     In declarations, do not put any whitespace between asterisks and adjacent
     tokens, except for tokens that are identifiers related to types.  (These
     identifiers are the names of basic types, type qualifiers, and
     typedef-names other than the one being declared.)  Separate these identi-
     fiers from asterisks using a single space.

     When declaring variables in structures, declare them sorted by use, then
     by size (largest to smallest), and then in alphabetical order.  The first
     category normally does not apply, but there are exceptions.  Each one
     gets its own line.  Try to make the structure readable by aligning the
     member names using either one or two tabs depending upon your judgment.
     You should use one tab only if it suffices to align at least 90% of the
     member names.  Names following extremely long types should be separated
     by a single space.

     Major structures should be declared at the top of the file in which they
     are used, or in separate header files if they are used in multiple source
     files.  Use of the structures should be by separate declarations and
     should be extern if they are declared in a header file.

     struct foo {
             struct foo      *next;          /* List of active foo. */
             struct mumble   amumble;        /* Comment for mumble. */
             int             bar;            /* Try to align the comments. */
             struct verylongtypename *baz;   /* Won’t fit in 2 tabs. */
     };
     struct foo *foohead;                    /* Head of global foo list. */

     Use queue(3) macros rather than rolling your own lists, whenever possi-
     ble.  Thus, the previous example would be better written:

     #include

     struct foo {
             LIST_ENTRY(foo) link;           /* Use queue macros for foo lists. */
             struct mumble   amumble;        /* Comment for mumble. */
             int             bar;            /* Try to align the comments. */
             struct verylongtypename *baz;   /* Won’t fit in 2 tabs. */
     };
     LIST_HEAD(, foo) foohead;               /* Head of global foo list. */

     Avoid using typedefs for structure types.  Typedefs are problematic
     because they do not properly hide their underlying type; for example you
     need to know if the typedef is the structure itself or a pointer to the
     structure.  In addition they must be declared exactly once, whereas an
     incomplete structure type can be mentioned as many times as necessary.
     Typedefs are difficult to use in stand-alone header files: the header
     that defines the typedef must be included before the header that uses it,
     or by the header that uses it (which causes namespace pollution), or
     there must be a back-door mechanism for obtaining the typedef.

     When convention requires a typedef, make its name match the struct tag.
     Avoid typedefs ending in “_t”, except as specified in Standard C or by
     POSIX.

     /* Make the structure name match the typedef. */
     typedef struct bar {
             int     level;
     } BAR;
     typedef int             foo;            /* This is foo. */
     typedef const long      baz;            /* This is baz. */

     All functions are prototyped somewhere.

     Function prototypes for private functions (i.e., functions not used else-
     where) go at the top of the first source module.  Functions local to one
     source module should be declared static.

     Functions used from other parts of the kernel are prototyped in the rele-
     vant include file.  Function prototypes should be listed in a logical
     order, preferably alphabetical unless there is a compelling reason to use
     a different ordering.

     Functions that are used locally in more than one module go into a sepa-
     rate header file, e.g. "extern.h".

     Do not use the __P macro.

     In general code can be considered “new code” when it makes up about 50%
     or more of the file(s) involved.  This is enough to break precedents in
     the existing code and use the current style guidelines.

     The kernel has a name associated with parameter types, e.g., in the ker-
     nel use:

     void    function(int fd);

     In header files visible to userland applications, prototypes that are
     visible must use either “protected” names (ones beginning with an
     underscore) or no names with the types.  It is preferable to use pro-
     tected names.  E.g., use:

     void    function(int);

     or:

     void    function(int _fd);

     Prototypes may have an extra space after a tab to enable function names
     to line up:

     static char     *function(int _arg, const char *_arg2, struct foo *_arg3,
                         struct bar *_arg4);
     static void      usage(void);

     /*
      * All major routines should have a comment briefly describing what
      * they do.  The comment before the "main" routine should describe
      * what the program does.
      */
     int
     main(int argc, char *argv[])
     {
             char *ep;
             long num;
             int ch;

     For consistency, getopt(3) should be used to parse options.  Options
     should be sorted in the getopt(3) call and the switch statement, unless
     parts of the switch cascade.  Elements in a switch statement that cascade
     should have a FALLTHROUGH comment.  Numerical arguments should be checked
     for accuracy.  Code that cannot be reached should have a NOTREACHED com-
     ment.

             while ((ch = getopt(argc, argv, "abNn:")) != -1)
                     switch (ch) {           /* Indent the switch. */
                     case ‘a’:               /* Don’t indent the case. */
                             aflag = 1;
                             /* FALLTHROUGH */
                     case ‘b’:
                             bflag = 1;
                             break;
                     case ‘N’:
                             Nflag = 1;
                             break;
                     case ‘n’:
                             num = strtol(optarg, &ep, 10);
                             if (num <= 0 || *ep != ”) {
                                     warnx("illegal number, -n argument — %s",
                                         optarg);
                                     usage();
                             }
                             break;
                     case ‘?’:
                     default:
                             usage();
                             /* NOTREACHED */
                     }
             argc -= optind;
             argv += optind;

     Space after keywords (if, while, for, return, switch).  No braces (`{‘
     and `}’) are used for control statements with zero or only a single
     statement unless that statement is more than a single line in which case
     they are permitted.  Forever loops are done with for’s, not while’s.

             for (p = buf; *p != ”; ++p)
                     ;       /* nothing */
             for (;;)
                     stmt;
             for (;;) {
                     z = a + really + long + statement + that + needs +
                         two + lines + gets + indented + four + spaces +
                         on + the + second + and + subsequent + lines;
             }
             for (;;) {
                     if (cond)
                             stmt;
             }
             if (val != NULL)
                     val = realloc(val, newsize);

     Parts of a for loop may be left empty.  Do not put declarations inside
     blocks unless the routine is unusually complicated.

             for (; cnt < 15; cnt++) {
                     stmt1;
                     stmt2;
             }

     Indentation is an 8 character tab.  Second level indents are four spaces.
     If you have to wrap a long statement, put the operator at the end of the
     line.

             while (cnt < 20 && this_variable_name_is_too_long &&
                 ep != NULL)
                     z = a + really + long + statement + that + needs +
                         two + lines + gets + indented + four + spaces +
                         on + the + second + and + subsequent + lines;

     Do not add whitespace at the end of a line, and only use tabs followed by
     spaces to form the indentation.  Do not use more spaces than a tab will
     produce and do not use spaces in front of tabs.

     Closing and opening braces go on the same line as the else.  Braces that
     are not necessary may be left out.

             if (test)
                     stmt;
             else if (bar) {
                     stmt;
                     stmt;
             } else
                     stmt;

     No spaces after function names.  Commas have a space after them.  No spa-
     ces after `(‘ or `[‘ or preceding `]‘ or `)’ characters.

             error = function(a1, a2);
             if (error != 0)
                     exit(error);

     Unary operators do not require spaces, binary operators do.  Do not use
     parentheses unless they are required for precedence or unless the state-
     ment is confusing without them.  Remember that other people may confuse
     easier than you.  Do YOU understand the following?

             a = b->c[0] + ~d == (e || f) || g && h ? i : j >> 1;
             k = !(l & FLAGS);

     Exits should be 0 on success, or according to the predefined values in
     sysexits(3).

             exit(EX_OK);    /*
                              * Avoid obvious comments such as
                              * "Exit 0 on success."
                              */
     }

     The function type should be on a line by itself preceding the function.
     The opening brace of the function body should be on a line by itself.

     static char *
     function(int a1, int a2, float fl, int a4)
     {

     When declaring variables in functions declare them sorted by size, then
     in alphabetical order; multiple ones per line are okay.  If a line over-
     flows reuse the type keyword.

     Be careful to not obfuscate the code by initializing variables in the
     declarations.  Use this feature only thoughtfully.  DO NOT use function
     calls in initializers.

             struct foo one, *two;
             double three;
             int *four, five;
             char *six, seven, eight, nine, ten, eleven, twelve;

             four = myfunction();

     Do not declare functions inside other functions; ANSI C says that such
     declarations have file scope regardless of the nesting of the declara-
     tion.  Hiding file declarations in what appears to be a local scope is
     undesirable and will elicit complaints from a good compiler.

     Casts and sizeof’s are not followed by a space.  Note that indent(1) does
     not understand this rule.  sizeof’s are written with parenthesis always.
     The redundant parenthesis rules do not apply to sizeof(var) instances.

     NULL is the preferred null pointer constant.  Use NULL instead of (type
     *)0 or (type *)NULL in contexts where the compiler knows the type, e.g.,
     in assignments.  Use (type *)NULL in other contexts, in particular for
     all function args.  (Casting is essential for variadic args and is neces-
     sary for other args if the function prototype might not be in scope.)
     Test pointers against NULL, e.g., use:

     (p = f()) == NULL

     not:

     !(p = f())

     Do not use ! for tests unless it is a boolean, e.g. use:

     if (*p == ”)

     not:

     if (!*p)

     Routines returning void * should not have their return values cast to any
     pointer type.

     Values in return statements should be enclosed in parentheses.

     Use err(3) or warn(3), do not roll your own.

             if ((four = malloc(sizeof(struct foo))) == NULL)
                     err(1, (char *)NULL);
             if ((six = (int *)overflow()) == NULL)
                     errx(1, "number overflowed");
             return (eight);
     }

     Old-style function declarations look like this:

     static char *
     function(a1, a2, fl, a4)
             int a1, a2;     /* Declare ints, too, don’t default them. */
             float fl;       /* Beware double vs. float prototype differences. */
             int a4;         /* List in order declared. */
     {

     Use ANSI function declarations unless you explicitly need K&R compatibil-
     ity.  Long parameter lists are wrapped with a normal four space indent.

     Variable numbers of arguments should look like this:

     #include

     void
     vaf(const char *fmt, …)
     {
             va_list ap;

             va_start(ap, fmt);
             STUFF;
             va_end(ap);
             /* No return needed for void functions. */
     }

     static void
     usage()
     {
             /* Insert an empty line if the function has no local variables. */

     Use printf(3), not fputs(3), puts(3), putchar(3), whatever; it is faster
     and usually cleaner, not to mention avoiding stupid bugs.

     Usage statements should look like the manual pages SYNOPSIS.  The usage
     statement should be structured in the following order:

     1.   Options without operands come first, in alphabetical order, inside a
          single set of brackets (`[‘ and `]‘).

     2.   Options with operands come next, also in alphabetical order, with
          each option and its argument inside its own pair of brackets.

     3.   Required arguments (if any) are next, listed in the order they
          should be specified on the command line.

     4.   Finally, any optional arguments should be listed, listed in the
          order they should be specified, and all inside brackets.

     A bar (`|’) separates “either-or” options/arguments, and multiple
     options/arguments which are specified together are placed in a single set
     of brackets.

         "usage: f [-aDde] [-b b_arg] [-m m_arg] req1 req2 [opt1 [opt2]] "
         "usage: f [-a | -b] [-c [-dEe] [-n number]] "

             (void)fprintf(stderr, "usage: f [-ab] ");
             exit(EX_USAGE);
     }

     Note that the manual page options description should list the options in
     pure alphabetical order.  That is, without regard to whether an option
     takes arguments or not.  The alphabetical ordering should take into
     account the case ordering shown above.

     New core kernel code should be reasonably compliant with the style
     guides.  The guidelines for third-party maintained modules and device
     drivers are more relaxed but at a minimum should be internally consistent
     with their style.

     Stylistic changes (including whitespace changes) are hard on the source
     repository and are to be avoided without good reason.  Code that is
     approximately FreeBSD KNF style compliant in the repository must not
     diverge from compliance.

     Whenever possible, code should be run through a code checker (e.g.,
     lint(1) or gcc -Wall) and produce minimal warnings.

SEE ALSO
     indent(1), lint(1), err(3), sysexits(3), warn(3), style.Makefile(5)

HISTORY
     This manual page is largely based on the src/admin/style/style file from
     the 4.4BSD-Lite2 release, with occasional updates to reflect the current
     practice and desire of the FreeBSD project.

FreeBSD 6.0                    February 10, 2005                   FreeBSD 6.0