1 /* xargs -- build and execute command lines from standard input
2 Copyright (C) 1990, 91, 92, 93, 94, 2000,2003 Free Software Foundation, Inc.
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2, or (at your option)
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 9 Temple Place - Suite 330, Boston, MA 02111-1307,
20 /* Written by Mike Rendell <michael@cs.mun.ca>
21 and David MacKenzie <djm@gnu.ai.mit.edu>. */
23 #include <gnulib/config.h>
25 #undef PACKAGE_VERSION
26 #undef PACKAGE_TARNAME
33 # if defined PROTOTYPES || (defined __STDC__ && __STDC__)
34 # define PARAMS(Args) Args
36 # define PARAMS(Args) ()
45 #if !defined (isascii) || defined (STDC_HEADERS)
53 #define ISBLANK(c) (isascii (c) && isblank (c))
55 #define ISBLANK(c) ((c) == ' ' || (c) == '\t')
58 #define ISSPACE(c) (ISBLANK (c) || (c) == '\n' || (c) == '\r' \
59 || (c) == '\f' || (c) == '\v')
61 #include <sys/types.h>
67 #if defined(HAVE_STRING_H) || defined(STDC_HEADERS)
69 #if !defined(STDC_HEADERS)
74 #define memcpy(dest, source, count) (bcopy((source), (dest), (count)))
78 #include <sys/param.h>
86 #define LONG_MAX (~(1 << (sizeof (long) * 8 - 1)))
95 #if !defined(SIGCHLD) && defined(SIGCLD)
96 #define SIGCHLD SIGCLD
99 /* COMPAT: SYSV version defaults size (and has a max value of) to 470.
100 We try to make it as large as possible. */
101 #if !defined(ARG_MAX) && defined(_SC_ARG_MAX)
102 #define ARG_MAX sysconf (_SC_ARG_MAX)
105 #define ARG_MAX NCARGS
110 /* States for read_line. */
126 # include <libintl.h>
127 # define _(Text) gettext (Text)
129 # define _(Text) Text
130 #define textdomain(Domain)
131 #define bindtextdomain(Package, Directory)
134 # define N_(String) gettext_noop (String)
136 # define N_(String) (String)
139 /* Return nonzero if S is the EOF string. */
140 #define EOF_STR(s) (eof_str && *eof_str == *s && !strcmp (eof_str, s))
142 extern char **environ
;
144 /* Do multibyte processing if multibyte characters are supported,
145 unless multibyte sequences are search safe. Multibyte sequences
146 are search safe if searching for a substring using the byte
147 comparison function 'strstr' gives no false positives. All 8-bit
148 encodings and the UTF-8 multibyte encoding are search safe, but
149 the EUC encodings are not.
150 BeOS uses the UTF-8 encoding exclusively, so it is search safe. */
152 # define MULTIBYTE_IS_SEARCH_SAFE 1
154 #define DO_MULTIBYTE (HAVE_MBLEN && ! MULTIBYTE_IS_SEARCH_SAFE)
160 /* Simulate mbrlen with mblen as best we can. */
161 # define mbstate_t int
162 # define mbrlen(s, n, ps) mblen (s, n)
166 /* Not char because of type promotion; NeXT gcc can't handle it. */
178 void error
PARAMS ((int status
, int errnum
, char *message
,...));
180 extern char *version_string
;
182 /* The name this program was run with. */
185 /* Buffer for reading arguments from stdin. */
186 static char *linebuf
;
188 /* Line number in stdin since the last command was executed. */
189 static int lineno
= 0;
191 /* If nonzero, then instead of putting the args from stdin at
192 the end of the command argument list, they are each stuck into the
193 initial args, replacing each occurrence of the `replace_pat' in the
195 static char *replace_pat
= NULL
;
197 /* The length of `replace_pat'. */
198 static size_t rplen
= 0;
200 /* If nonzero, when this string is read on stdin it is treated as
202 I don't like this - it should default to NULL. */
203 static char *eof_str
= "_";
205 /* If nonzero, the maximum number of nonblank lines from stdin to use
207 static long lines_per_exec
= 0;
209 /* The maximum number of arguments to use per command line. */
210 static long args_per_exec
= 1024;
212 /* If true, exit if lines_per_exec or args_per_exec is exceeded. */
213 static boolean exit_if_size_exceeded
= false;
215 /* The maximum number of characters that can be used per command line. */
218 /* Storage for elements of `cmd_argv'. */
221 /* The list of args being built. */
222 static char **cmd_argv
= NULL
;
224 /* Number of elements allocated for `cmd_argv'. */
225 static int cmd_argv_alloc
= 0;
227 /* Number of valid elements in `cmd_argv'. */
228 static int cmd_argc
= 0;
230 /* Number of chars being used in `cmd_argv'. */
231 static int cmd_argv_chars
= 0;
233 /* Number of initial arguments given on the command line. */
234 static int initial_argc
= 0;
236 /* Number of chars in the initial args. */
237 static int initial_argv_chars
= 0;
239 /* true when building up initial arguments in `cmd_argv'. */
240 static boolean initial_args
= true;
242 /* If nonzero, the maximum number of child processes that can be running
244 static int proc_max
= 1;
246 /* Total number of child processes that have been executed. */
247 static int procs_executed
= 0;
249 /* The number of elements in `pids'. */
250 static int procs_executing
= 0;
252 /* List of child processes currently executing. */
253 static pid_t
*pids
= NULL
;
255 /* The number of allocated elements in `pids'. */
256 static int pids_alloc
= 0;
258 /* Exit status; nonzero if any child process exited with a
260 static int child_error
= 0;
262 /* If true, print each command on stderr before executing it. */
263 static boolean print_command
= false;
265 /* If true, query the user before executing each command, and only
266 execute the command if the user responds affirmatively. */
267 static boolean query_before_executing
= false;
269 static struct option
const longopts
[] =
271 {"null", no_argument
, NULL
, '0'},
272 {"eof", optional_argument
, NULL
, 'e'},
273 {"replace", optional_argument
, NULL
, 'i'},
274 {"max-lines", optional_argument
, NULL
, 'l'},
275 {"max-args", required_argument
, NULL
, 'n'},
276 {"interactive", no_argument
, NULL
, 'p'},
277 {"no-run-if-empty", no_argument
, NULL
, 'r'},
278 {"max-chars", required_argument
, NULL
, 's'},
279 {"verbose", no_argument
, NULL
, 't'},
280 {"exit", no_argument
, NULL
, 'x'},
281 {"max-procs", required_argument
, NULL
, 'P'},
282 {"version", no_argument
, NULL
, 'v'},
283 {"help", no_argument
, NULL
, 'h'},
284 {NULL
, no_argument
, NULL
, 0}
287 static int read_line
PARAMS ((void));
288 static int read_string
PARAMS ((void));
289 static char *mbstrstr
PARAMS ((const char *haystack
, const char *needle
));
290 static void do_insert
PARAMS ((char *arg
, size_t arglen
, size_t lblen
));
291 static void push_arg
PARAMS ((char *arg
, size_t len
));
292 static boolean print_args
PARAMS ((boolean ask
));
293 static void do_exec
PARAMS ((void));
294 static void add_proc
PARAMS ((pid_t pid
));
295 static void wait_for_proc
PARAMS ((boolean all
));
296 static long parse_num
PARAMS ((char *str
, int option
, long min
, long max
));
297 static long env_size
PARAMS ((char **envp
));
298 static void usage
PARAMS ((FILE * stream
, int status
));
301 main (int argc
, char **argv
)
304 int always_run_command
= 1;
306 char *default_cmd
= "/bin/echo";
307 int (*read_args
) PARAMS ((void)) = read_line
;
309 program_name
= argv
[0];
311 #ifdef HAVE_SETLOCALE
312 setlocale (LC_ALL
, "");
314 bindtextdomain (PACKAGE
, LOCALEDIR
);
315 textdomain (PACKAGE
);
317 orig_arg_max
= ARG_MAX
;
318 if (orig_arg_max
== -1)
319 orig_arg_max
= LONG_MAX
;
320 orig_arg_max
-= 2048; /* POSIX.2 requires subtracting 2048. */
321 arg_max
= orig_arg_max
;
323 /* Sanity check for systems with huge ARG_MAX defines (e.g., Suns which
324 have it at 1 meg). Things will work fine with a large ARG_MAX but it
325 will probably hurt the system more than it needs to; an array of this
326 size is allocated. */
327 if (arg_max
> 20 * 1024)
330 /* Take the size of the environment into account. */
331 arg_max
-= env_size (environ
);
333 error (1, 0, _("environment is too large for exec"));
335 while ((optc
= getopt_long (argc
, argv
, "+0e::i::l::n:prs:txP:",
336 longopts
, (int *) 0)) != -1)
341 read_args
= read_string
;
356 replace_pat
= optarg
;
359 /* -i excludes -n -l. */
366 lines_per_exec
= parse_num (optarg
, 'l', 1L, -1L);
369 /* -l excludes -i -n. */
375 args_per_exec
= parse_num (optarg
, 'n', 1L, -1L);
376 /* -n excludes -i -l. */
378 if (args_per_exec
== 1 && replace_pat
)
379 /* ignore -n1 in '-i -n1' */
386 arg_max
= parse_num (optarg
, 's', 1L, orig_arg_max
);
390 print_command
= true;
394 exit_if_size_exceeded
= true;
398 query_before_executing
= true;
399 print_command
= true;
403 always_run_command
= 0;
407 proc_max
= parse_num (optarg
, 'P', 0L, -1L);
411 printf (_("GNU xargs version %s\n"), version_string
);
419 if (replace_pat
|| lines_per_exec
)
420 exit_if_size_exceeded
= true;
429 linebuf
= (char *) xmalloc (arg_max
+ 1);
430 argbuf
= (char *) xmalloc (arg_max
+ 1);
432 /* Make sure to listen for the kids. */
433 signal (SIGCHLD
, SIG_DFL
);
437 for (; optind
< argc
; optind
++)
438 push_arg (argv
[optind
], strlen (argv
[optind
]) + 1);
439 initial_args
= false;
440 initial_argc
= cmd_argc
;
441 initial_argv_chars
= cmd_argv_chars
;
443 while ((*read_args
) () != -1)
444 if (lines_per_exec
&& lineno
>= lines_per_exec
)
450 /* SYSV xargs seems to do at least one exec, even if the
452 if (cmd_argc
!= initial_argc
453 || (always_run_command
&& procs_executed
== 0))
460 size_t *arglen
= (size_t *) xmalloc (sizeof (size_t) * argc
);
462 for (i
= optind
; i
< argc
; i
++)
463 arglen
[i
] = strlen(argv
[i
]);
464 rplen
= strlen (replace_pat
);
465 while ((len
= (*read_args
) ()) != -1)
467 /* Don't do insert on the command name. */
468 push_arg (argv
[optind
], arglen
[optind
] + 1);
470 for (i
= optind
+ 1; i
< argc
; i
++)
471 do_insert (argv
[i
], arglen
[i
], len
);
476 wait_for_proc (true);
480 /* Read a line of arguments from stdin and add them to the list of
481 arguments to pass to the command. Ignore blank lines and initial blanks.
482 Single and double quotes and backslashes quote metacharacters and blanks
483 as they do in the shell.
484 Return -1 if eof (either physical or logical) is reached,
485 otherwise the length of the last string read (including the null). */
490 static boolean eof
= false;
491 /* Start out in mode SPACE to always strip leading spaces (even with -i). */
492 int state
= SPACE
; /* The type of character we last read. */
493 int prevc
; /* The previous value of c. */
494 int quotc
= 0; /* The last quote character read. */
496 boolean first
= true; /* true if reading first arg on line. */
499 /* Including the NUL, the args must not grow past this point. */
500 char *endbuf
= linebuf
+ arg_max
- initial_argv_chars
- 1;
510 /* COMPAT: SYSV seems to ignore stuff on a line that
511 ends without a \n; we don't. */
517 /* FIXME we don't check for unterminated quotes here. */
518 if (first
&& EOF_STR (linebuf
))
521 push_arg (linebuf
, len
);
535 if (!ISBLANK (prevc
))
536 lineno
++; /* For -l. */
545 if (EOF_STR (linebuf
))
548 return first
? -1 : len
;
551 push_arg (linebuf
, len
);
554 if (!replace_pat
&& ISSPACE (c
))
558 if (EOF_STR (linebuf
))
561 return first
? -1 : len
;
563 push_arg (linebuf
, len
);
585 error (1, 0, _("unmatched %s quote"),
586 quotc
== '"' ? _("double") : _("single"));
599 error (1, 0, _("argument line too long"));
604 /* Read a null-terminated string from stdin and add it to the list of
605 arguments to pass to the command.
606 Return -1 if eof (either physical or logical) is reached,
607 otherwise the length of the string read (including the null). */
612 static boolean eof
= false;
615 /* Including the NUL, the args must not grow past this point. */
616 char *endbuf
= linebuf
+ arg_max
- initial_argv_chars
- 1;
622 int c
= getc (stdin
);
631 push_arg (linebuf
, len
);
636 lineno
++; /* For -l. */
640 push_arg (linebuf
, len
);
644 error (1, 0, _("argument line too long"));
649 /* Finds the first occurrence of the substring NEEDLE in the string
650 HAYSTACK. Both strings can be multibyte strings. */
653 mbstrstr (const char *haystack
, const char *needle
)
658 size_t hlen
= strlen (haystack
);
659 size_t nlen
= strlen (needle
);
663 memset (&mbstate
, 0, sizeof (mbstate_t));
666 if (memcmp (haystack
, needle
, nlen
) == 0)
667 return (char *) haystack
;
668 step
= mbrlen (haystack
, hlen
, &mbstate
);
677 return strstr (haystack
, needle
);
680 /* Replace all instances of `replace_pat' in ARG with `linebuf',
681 and add the resulting string to the list of arguments for the command
683 ARGLEN is the length of ARG, not including the null.
684 LBLEN is the length of `linebuf', not including the null.
686 COMPAT: insertions on the SYSV version are limited to 255 chars per line,
687 and a max of 5 occurrences of replace_pat in the initial-arguments.
688 Those restrictions do not exist here. */
691 do_insert (char *arg
, size_t arglen
, size_t lblen
)
693 /* Temporary copy of each arg with the replace pattern replaced by the
695 static char *insertbuf
;
697 int bytes_left
= arg_max
- 1; /* Bytes left on the command line. */
700 insertbuf
= (char *) xmalloc (arg_max
+ 1);
705 size_t len
; /* Length in ARG before `replace_pat'. */
706 char *s
= mbstrstr (arg
, replace_pat
);
715 strncpy (p
, arg
, len
);
733 error (1, 0, _("command too long"));
735 push_arg (insertbuf
, p
- insertbuf
);
738 /* Add ARG to the end of the list of arguments `cmd_argv' to pass
740 LEN is the length of ARG, including the terminating null.
741 If this brings the list up to its maximum size, execute the command. */
744 push_arg (char *arg
, size_t len
)
748 if (cmd_argv_chars
+ len
> arg_max
)
750 if (initial_args
|| cmd_argc
== initial_argc
)
751 error (1, 0, _("can not fit single argument within argument list size limit"));
753 || (exit_if_size_exceeded
&&
754 (lines_per_exec
|| args_per_exec
)))
755 error (1, 0, _("argument list too long"));
758 if (!initial_args
&& args_per_exec
&&
759 cmd_argc
- initial_argc
== args_per_exec
)
763 if (cmd_argc
>= cmd_argv_alloc
)
768 cmd_argv
= (char **) xmalloc (sizeof (char *) * cmd_argv_alloc
);
773 cmd_argv
= (char **) xrealloc (cmd_argv
,
774 sizeof (char *) * cmd_argv_alloc
);
779 cmd_argv
[cmd_argc
++] = NULL
;
782 cmd_argv
[cmd_argc
++] = argbuf
+ cmd_argv_chars
;
783 strcpy (argbuf
+ cmd_argv_chars
, arg
);
784 cmd_argv_chars
+= len
;
788 /* Print the arguments of the command to execute.
789 If ASK is nonzero, prompt the user for a response, and
790 if the user responds affirmatively, return true;
791 otherwise, return false. */
794 print_args (boolean ask
)
798 for (i
= 0; i
< cmd_argc
- 1; i
++)
799 fprintf (stderr
, "%s ", cmd_argv
[i
]);
802 static FILE *tty_stream
;
807 tty_stream
= fopen ("/dev/tty", "r");
809 error (1, errno
, "/dev/tty");
811 fputs ("?...", stderr
);
813 c
= savec
= getc (tty_stream
);
814 while (c
!= EOF
&& c
!= '\n')
815 c
= getc (tty_stream
);
816 if (savec
== 'y' || savec
== 'Y')
826 /* Close stdin and attach /dev/null to it.
827 * This resolves Savannah bug #3992.
830 prep_child_for_exec (void)
832 const char inputfile
[] = "/dev/null";
833 /* fprintf(stderr, "attaching stdin to /dev/null\n"); */
836 if (open(inputfile
, O_RDONLY
) < 0)
838 /* This is not entirely fatal, since
839 * executing the child with a closed
840 * stdin is almost as good as executing it
841 * with its stdin attached to /dev/null.
843 error (0, errno
, "%s", inputfile
);
848 /* Execute the command that has been built in `cmd_argv'. This may involve
849 waiting for processes that were previously executed. */
856 push_arg ((char *) NULL
, 0); /* Null terminate the arg list. */
857 if (!query_before_executing
|| print_args (true))
859 if (proc_max
&& procs_executing
>= proc_max
)
860 wait_for_proc (false);
861 if (!query_before_executing
&& print_command
)
863 /* If we run out of processes, wait for a child to return and
865 while ((child
= fork ()) < 0 && errno
== EAGAIN
&& procs_executing
)
866 wait_for_proc (false);
870 error (1, errno
, _("cannot fork"));
873 prep_child_for_exec();
874 execvp (cmd_argv
[0], cmd_argv
);
875 error (0, errno
, "%s", cmd_argv
[0]);
876 _exit (errno
== ENOENT
? 127 : 126);
881 cmd_argc
= initial_argc
;
882 cmd_argv_chars
= initial_argv_chars
;
885 /* Add the process with id PID to the list of processes that have
893 /* Find an empty slot. */
894 for (i
= 0; i
< pids_alloc
&& pids
[i
]; i
++)
900 pids_alloc
= proc_max
? proc_max
: 64;
901 pids
= (pid_t
*) xmalloc (sizeof (pid_t
) * pids_alloc
);
906 pids
= (pid_t
*) xrealloc (pids
,
907 sizeof (pid_t
) * pids_alloc
);
909 memset (&pids
[i
], '\0', sizeof (pid_t
) * (pids_alloc
- i
));
916 /* If ALL is true, wait for all child processes to finish;
917 otherwise, wait for one child process to finish.
918 Remove the processes that finish from the list of executing processes. */
921 wait_for_proc (boolean all
)
923 while (procs_executing
)
931 while ((pid
= wait (&status
)) == (pid_t
) -1)
933 error (1, errno
, _("error waiting for child process"));
935 /* Find the entry in `pids' for the child process
937 for (i
= 0; i
< pids_alloc
&& pid
!= pids
[i
]; i
++)
940 while (i
== pids_alloc
); /* A child died that we didn't start? */
942 /* Remove the child from the list. */
946 if (WEXITSTATUS (status
) == 126 || WEXITSTATUS (status
) == 127)
947 exit (WEXITSTATUS (status
)); /* Can't find or run the command. */
948 if (WEXITSTATUS (status
) == 255)
949 error (124, 0, _("%s: exited with status 255; aborting"), cmd_argv
[0]);
950 if (WIFSTOPPED (status
))
951 error (125, 0, _("%s: stopped by signal %d"), cmd_argv
[0], WSTOPSIG (status
));
952 if (WIFSIGNALED (status
))
953 error (125, 0, _("%s: terminated by signal %d"), cmd_argv
[0], WTERMSIG (status
));
954 if (WEXITSTATUS (status
) != 0)
962 /* Return the value of the number represented in STR.
963 OPTION is the command line option to which STR is the argument.
964 If the value does not fall within the boundaries MIN and MAX,
965 Print an error message mentioning OPTION and exit. */
968 parse_num (char *str
, int option
, long int min
, long int max
)
973 val
= strtol (str
, &eptr
, 10);
974 if (eptr
== str
|| *eptr
)
976 fprintf (stderr
, _("%s: invalid number for -%c option\n"),
977 program_name
, option
);
982 fprintf (stderr
, _("%s: value for -%c option must be >= %ld\n"),
983 program_name
, option
, min
);
986 else if (max
>= 0 && val
> max
)
988 fprintf (stderr
, _("%s: value for -%c option must be < %ld\n"),
989 program_name
, option
, max
);
995 /* Return how much of ARG_MAX is used by the environment. */
998 env_size (char **envp
)
1003 len
+= strlen (*envp
++) + 1;
1009 usage (FILE *stream
, int status
)
1011 fprintf (stream
, _("\
1012 Usage: %s [-0prtx] [-e[eof-str]] [-i[replace-str]] [-l[max-lines]]\n\
1013 [-n max-args] [-s max-chars] [-P max-procs] [--null] [--eof[=eof-str]]\n\
1014 [--replace[=replace-str]] [--max-lines[=max-lines]] [--interactive]\n\
1015 [--max-chars=max-chars] [--verbose] [--exit] [--max-procs=max-procs]\n\
1016 [--max-args=max-args] [--no-run-if-empty] [--version] [--help]\n\
1017 [command [initial-arguments]]\n"),
1019 fputs (_("\nReport bugs to <bug-findutils@gnu.org>.\n"), stream
);