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;
307 long size_of_environment
= env_size(environ
);
308 char *default_cmd
= "/bin/echo";
309 int (*read_args
) PARAMS ((void)) = read_line
;
311 program_name
= argv
[0];
313 #ifdef HAVE_SETLOCALE
314 setlocale (LC_ALL
, "");
316 bindtextdomain (PACKAGE
, LOCALEDIR
);
317 textdomain (PACKAGE
);
319 orig_arg_max
= ARG_MAX
;
320 if (orig_arg_max
== -1)
321 orig_arg_max
= LONG_MAX
;
322 orig_arg_max
-= 2048; /* POSIX.2 requires subtracting 2048. */
323 arg_max
= orig_arg_max
;
325 arg_size
= 20 * 1048 + size_of_environment
;
328 /* Take the size of the environment into account. */
329 arg_max
-= env_size (environ
);
331 error (1, 0, _("environment is too large for exec"));
333 while ((optc
= getopt_long (argc
, argv
, "+0e::i::l::n:prs:txP:",
334 longopts
, (int *) 0)) != -1)
339 read_args
= read_string
;
354 replace_pat
= optarg
;
357 /* -i excludes -n -l. */
364 lines_per_exec
= parse_num (optarg
, 'l', 1L, -1L);
367 /* -l excludes -i -n. */
373 args_per_exec
= parse_num (optarg
, 'n', 1L, -1L);
374 /* -n excludes -i -l. */
376 if (args_per_exec
== 1 && replace_pat
)
377 /* ignore -n1 in '-i -n1' */
384 arg_size
= parse_num (optarg
, 's', 1L, orig_arg_max
);
388 print_command
= true;
392 exit_if_size_exceeded
= true;
396 query_before_executing
= true;
397 print_command
= true;
401 always_run_command
= 0;
405 proc_max
= parse_num (optarg
, 'P', 0L, -1L);
409 printf (_("GNU xargs version %s\n"), version_string
);
417 if (replace_pat
|| lines_per_exec
)
418 exit_if_size_exceeded
= true;
427 /* Taking into account the sisze of the environment,
428 * figure out how large a buffer we need to
429 * hold all the arguments. We cannot use ARG_MAX
430 * directly since that may be arbitrarily large.
431 * This is from a patch by Bob Prolux, <bob@proulx.com>.
433 if (arg_max
> arg_size
)
438 linebuf
= (char *) xmalloc (arg_max
+ 1);
439 argbuf
= (char *) xmalloc (arg_max
+ 1);
441 /* Make sure to listen for the kids. */
442 signal (SIGCHLD
, SIG_DFL
);
446 for (; optind
< argc
; optind
++)
447 push_arg (argv
[optind
], strlen (argv
[optind
]) + 1);
448 initial_args
= false;
449 initial_argc
= cmd_argc
;
450 initial_argv_chars
= cmd_argv_chars
;
452 while ((*read_args
) () != -1)
453 if (lines_per_exec
&& lineno
>= lines_per_exec
)
459 /* SYSV xargs seems to do at least one exec, even if the
461 if (cmd_argc
!= initial_argc
462 || (always_run_command
&& procs_executed
== 0))
469 size_t *arglen
= (size_t *) xmalloc (sizeof (size_t) * argc
);
471 for (i
= optind
; i
< argc
; i
++)
472 arglen
[i
] = strlen(argv
[i
]);
473 rplen
= strlen (replace_pat
);
474 while ((len
= (*read_args
) ()) != -1)
476 /* Don't do insert on the command name. */
477 push_arg (argv
[optind
], arglen
[optind
] + 1);
479 for (i
= optind
+ 1; i
< argc
; i
++)
480 do_insert (argv
[i
], arglen
[i
], len
);
485 wait_for_proc (true);
491 append_char_to_buf(char **pbuf
, char **pend
, char **pp
, int c
)
493 char *end_of_buffer
= *pend
;
494 char *start_of_buffer
= *pbuf
;
496 if (p
>= end_of_buffer
)
500 size_t len
= end_of_buffer
- start_of_buffer
;
501 size_t offset
= p
- start_of_buffer
;
503 start_of_buffer
= xrealloc(start_of_buffer
, len
*2);
504 if (NULL
!= start_of_buffer
)
506 end_of_buffer
= start_of_buffer
+ len
;
507 p
= start_of_buffer
+ offset
;
510 /* Update the caller's idea of where the buffer is. */
511 *pbuf
= start_of_buffer
;
512 *pend
= end_of_buffer
;
519 /* Failed to reallocate. */
525 /* I suspect that this can never happen now, because append_char_to_buf()
526 * should only be called wen replace_pat is true.
528 error (1, 0, _("argument line too long"));
535 /* Enough space remains. */
544 /* Read a line of arguments from stdin and add them to the list of
545 arguments to pass to the command. Ignore blank lines and initial blanks.
546 Single and double quotes and backslashes quote metacharacters and blanks
547 as they do in the shell.
548 Return -1 if eof (either physical or logical) is reached,
549 otherwise the length of the last string read (including the null). */
554 static boolean eof
= false;
555 /* Start out in mode SPACE to always strip leading spaces (even with -i). */
556 int state
= SPACE
; /* The type of character we last read. */
557 int prevc
; /* The previous value of c. */
558 int quotc
= 0; /* The last quote character read. */
560 boolean first
= true; /* true if reading first arg on line. */
563 /* Including the NUL, the args must not grow past this point. */
564 char *endbuf
= linebuf
+ arg_max
- initial_argv_chars
- 1;
574 /* COMPAT: SYSV seems to ignore stuff on a line that
575 ends without a \n; we don't. */
581 /* FIXME we don't check for unterminated quotes here. */
582 if (first
&& EOF_STR (linebuf
))
585 push_arg (linebuf
, len
);
599 if (!ISBLANK (prevc
))
600 lineno
++; /* For -l. */
609 if (EOF_STR (linebuf
))
612 return first
? -1 : len
;
615 push_arg (linebuf
, len
);
618 if (!replace_pat
&& ISSPACE (c
))
622 if (EOF_STR (linebuf
))
625 return first
? -1 : len
;
627 push_arg (linebuf
, len
);
649 error (1, 0, _("unmatched %s quote"),
650 quotc
== '"' ? _("double") : _("single"));
664 error (1, 0, _("argument line too long"));
667 append_char_to_buf(&linebuf
, &endbuf
, &p
, c
);
672 /* Read a null-terminated string from stdin and add it to the list of
673 arguments to pass to the command.
674 Return -1 if eof (either physical or logical) is reached,
675 otherwise the length of the string read (including the null). */
680 static boolean eof
= false;
683 /* Including the NUL, the args must not grow past this point. */
684 char *endbuf
= linebuf
+ arg_max
- initial_argv_chars
- 1;
690 int c
= getc (stdin
);
699 push_arg (linebuf
, len
);
704 lineno
++; /* For -l. */
708 push_arg (linebuf
, len
);
712 error (1, 0, _("argument line too long"));
717 /* Finds the first occurrence of the substring NEEDLE in the string
718 HAYSTACK. Both strings can be multibyte strings. */
721 mbstrstr (const char *haystack
, const char *needle
)
726 size_t hlen
= strlen (haystack
);
727 size_t nlen
= strlen (needle
);
731 memset (&mbstate
, 0, sizeof (mbstate_t));
734 if (memcmp (haystack
, needle
, nlen
) == 0)
735 return (char *) haystack
;
736 step
= mbrlen (haystack
, hlen
, &mbstate
);
745 return strstr (haystack
, needle
);
748 /* Replace all instances of `replace_pat' in ARG with `linebuf',
749 and add the resulting string to the list of arguments for the command
751 ARGLEN is the length of ARG, not including the null.
752 LBLEN is the length of `linebuf', not including the null.
754 COMPAT: insertions on the SYSV version are limited to 255 chars per line,
755 and a max of 5 occurrences of replace_pat in the initial-arguments.
756 Those restrictions do not exist here. */
759 do_insert (char *arg
, size_t arglen
, size_t lblen
)
761 /* Temporary copy of each arg with the replace pattern replaced by the
763 static char *insertbuf
;
765 int bytes_left
= arg_max
- 1; /* Bytes left on the command line. */
768 insertbuf
= (char *) xmalloc (arg_max
+ 1);
773 size_t len
; /* Length in ARG before `replace_pat'. */
774 char *s
= mbstrstr (arg
, replace_pat
);
783 strncpy (p
, arg
, len
);
801 error (1, 0, _("command too long"));
803 push_arg (insertbuf
, p
- insertbuf
);
806 /* Add ARG to the end of the list of arguments `cmd_argv' to pass
808 LEN is the length of ARG, including the terminating null.
809 If this brings the list up to its maximum size, execute the command. */
812 push_arg (char *arg
, size_t len
)
816 if (cmd_argv_chars
+ len
> arg_max
)
818 if (initial_args
|| cmd_argc
== initial_argc
)
819 error (1, 0, _("can not fit single argument within argument list size limit"));
820 /* option -i (replace_pat) implies -x (exit_if_size_exceeded) */
822 || (exit_if_size_exceeded
&&
823 (lines_per_exec
|| args_per_exec
)))
824 error (1, 0, _("argument list too long"));
827 if (!initial_args
&& args_per_exec
&&
828 cmd_argc
- initial_argc
== args_per_exec
)
832 if (cmd_argc
>= cmd_argv_alloc
)
837 cmd_argv
= (char **) xmalloc (sizeof (char *) * cmd_argv_alloc
);
842 cmd_argv
= (char **) xrealloc (cmd_argv
,
843 sizeof (char *) * cmd_argv_alloc
);
848 cmd_argv
[cmd_argc
++] = NULL
;
851 cmd_argv
[cmd_argc
++] = argbuf
+ cmd_argv_chars
;
852 strcpy (argbuf
+ cmd_argv_chars
, arg
);
853 cmd_argv_chars
+= len
;
857 /* Print the arguments of the command to execute.
858 If ASK is nonzero, prompt the user for a response, and
859 if the user responds affirmatively, return true;
860 otherwise, return false. */
863 print_args (boolean ask
)
867 for (i
= 0; i
< cmd_argc
- 1; i
++)
868 fprintf (stderr
, "%s ", cmd_argv
[i
]);
871 static FILE *tty_stream
;
876 tty_stream
= fopen ("/dev/tty", "r");
878 error (1, errno
, "/dev/tty");
880 fputs ("?...", stderr
);
882 c
= savec
= getc (tty_stream
);
883 while (c
!= EOF
&& c
!= '\n')
884 c
= getc (tty_stream
);
885 if (savec
== 'y' || savec
== 'Y')
895 /* Close stdin and attach /dev/null to it.
896 * This resolves Savannah bug #3992.
899 prep_child_for_exec (void)
901 const char inputfile
[] = "/dev/null";
902 /* fprintf(stderr, "attaching stdin to /dev/null\n"); */
905 if (open(inputfile
, O_RDONLY
) < 0)
907 /* This is not entirely fatal, since
908 * executing the child with a closed
909 * stdin is almost as good as executing it
910 * with its stdin attached to /dev/null.
912 error (0, errno
, "%s", inputfile
);
917 /* Execute the command that has been built in `cmd_argv'. This may involve
918 waiting for processes that were previously executed. */
925 push_arg ((char *) NULL
, 0); /* Null terminate the arg list. */
926 if (!query_before_executing
|| print_args (true))
928 if (proc_max
&& procs_executing
>= proc_max
)
929 wait_for_proc (false);
930 if (!query_before_executing
&& print_command
)
932 /* If we run out of processes, wait for a child to return and
934 while ((child
= fork ()) < 0 && errno
== EAGAIN
&& procs_executing
)
935 wait_for_proc (false);
939 error (1, errno
, _("cannot fork"));
942 prep_child_for_exec();
943 execvp (cmd_argv
[0], cmd_argv
);
944 error (0, errno
, "%s", cmd_argv
[0]);
945 _exit (errno
== ENOENT
? 127 : 126);
950 cmd_argc
= initial_argc
;
951 cmd_argv_chars
= initial_argv_chars
;
954 /* Add the process with id PID to the list of processes that have
962 /* Find an empty slot. */
963 for (i
= 0; i
< pids_alloc
&& pids
[i
]; i
++)
969 pids_alloc
= proc_max
? proc_max
: 64;
970 pids
= (pid_t
*) xmalloc (sizeof (pid_t
) * pids_alloc
);
975 pids
= (pid_t
*) xrealloc (pids
,
976 sizeof (pid_t
) * pids_alloc
);
978 memset (&pids
[i
], '\0', sizeof (pid_t
) * (pids_alloc
- i
));
985 /* If ALL is true, wait for all child processes to finish;
986 otherwise, wait for one child process to finish.
987 Remove the processes that finish from the list of executing processes. */
990 wait_for_proc (boolean all
)
992 while (procs_executing
)
1000 while ((pid
= wait (&status
)) == (pid_t
) -1)
1002 error (1, errno
, _("error waiting for child process"));
1004 /* Find the entry in `pids' for the child process
1006 for (i
= 0; i
< pids_alloc
&& pid
!= pids
[i
]; i
++)
1009 while (i
== pids_alloc
); /* A child died that we didn't start? */
1011 /* Remove the child from the list. */
1015 if (WEXITSTATUS (status
) == 126 || WEXITSTATUS (status
) == 127)
1016 exit (WEXITSTATUS (status
)); /* Can't find or run the command. */
1017 if (WEXITSTATUS (status
) == 255)
1018 error (124, 0, _("%s: exited with status 255; aborting"), cmd_argv
[0]);
1019 if (WIFSTOPPED (status
))
1020 error (125, 0, _("%s: stopped by signal %d"), cmd_argv
[0], WSTOPSIG (status
));
1021 if (WIFSIGNALED (status
))
1022 error (125, 0, _("%s: terminated by signal %d"), cmd_argv
[0], WTERMSIG (status
));
1023 if (WEXITSTATUS (status
) != 0)
1031 /* Return the value of the number represented in STR.
1032 OPTION is the command line option to which STR is the argument.
1033 If the value does not fall within the boundaries MIN and MAX,
1034 Print an error message mentioning OPTION and exit. */
1037 parse_num (char *str
, int option
, long int min
, long int max
)
1042 val
= strtol (str
, &eptr
, 10);
1043 if (eptr
== str
|| *eptr
)
1045 fprintf (stderr
, _("%s: invalid number for -%c option\n"),
1046 program_name
, option
);
1051 fprintf (stderr
, _("%s: value for -%c option must be >= %ld\n"),
1052 program_name
, option
, min
);
1055 else if (max
>= 0 && val
> max
)
1057 fprintf (stderr
, _("%s: value for -%c option must be < %ld\n"),
1058 program_name
, option
, max
);
1064 /* Return how much of ARG_MAX is used by the environment. */
1067 env_size (char **envp
)
1072 len
+= strlen (*envp
++) + 1;
1078 usage (FILE *stream
, int status
)
1080 fprintf (stream
, _("\
1081 Usage: %s [-0prtx] [-e[eof-str]] [-i[replace-str]] [-l[max-lines]]\n\
1082 [-n max-args] [-s max-chars] [-P max-procs] [--null] [--eof[=eof-str]]\n\
1083 [--replace[=replace-str]] [--max-lines[=max-lines]] [--interactive]\n\
1084 [--max-chars=max-chars] [--verbose] [--exit] [--max-procs=max-procs]\n\
1085 [--max-args=max-args] [--no-run-if-empty] [--version] [--help]\n\
1086 [command [initial-arguments]]\n"),
1088 fputs (_("\nReport bugs to <bug-findutils@gnu.org>.\n"), stream
);