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>. */
26 # if defined PROTOTYPES || (defined __STDC__ && __STDC__)
27 # define PARAMS(Args) Args
29 # define PARAMS(Args) ()
38 #if !defined (isascii) || defined (STDC_HEADERS)
46 #define ISBLANK(c) (isascii (c) && isblank (c))
48 #define ISBLANK(c) ((c) == ' ' || (c) == '\t')
51 #define ISSPACE(c) (ISBLANK (c) || (c) == '\n' || (c) == '\r' \
52 || (c) == '\f' || (c) == '\v')
54 #include <sys/types.h>
60 #if defined(HAVE_STRING_H) || defined(STDC_HEADERS)
62 #if !defined(STDC_HEADERS)
67 #define memcpy(dest, source, count) (bcopy((source), (dest), (count)))
71 #include <sys/param.h>
79 #define LONG_MAX (~(1 << (sizeof (long) * 8 - 1)))
88 #if !defined(SIGCHLD) && defined(SIGCLD)
89 #define SIGCHLD SIGCLD
92 /* COMPAT: SYSV version defaults size (and has a max value of) to 470.
93 We try to make it as large as possible. */
94 #if !defined(ARG_MAX) && defined(_SC_ARG_MAX)
95 #define ARG_MAX sysconf (_SC_ARG_MAX)
98 #define ARG_MAX NCARGS
103 /* States for read_line. */
119 # include <libintl.h>
120 # define _(Text) gettext (Text)
122 # define _(Text) Text
123 #define textdomain(Domain)
124 #define bindtextdomain(Package, Directory)
127 # define N_(String) gettext_noop (String)
129 /* See locate.c for explanation as to why not use (String) */
130 # define N_(String) String
133 /* Return nonzero if S is the EOF string. */
134 #define EOF_STR(s) (eof_str && *eof_str == *s && !strcmp (eof_str, s))
136 extern char **environ
;
138 /* Do multibyte processing if multibyte characters are supported,
139 unless multibyte sequences are search safe. Multibyte sequences
140 are search safe if searching for a substring using the byte
141 comparison function 'strstr' gives no false positives. All 8-bit
142 encodings and the UTF-8 multibyte encoding are search safe, but
143 the EUC encodings are not.
144 BeOS uses the UTF-8 encoding exclusively, so it is search safe. */
146 # define MULTIBYTE_IS_SEARCH_SAFE 1
148 #define DO_MULTIBYTE (HAVE_MBLEN && ! MULTIBYTE_IS_SEARCH_SAFE)
154 /* Simulate mbrlen with mblen as best we can. */
155 # define mbstate_t int
156 # define mbrlen(s, n, ps) mblen (s, n)
160 /* Not char because of type promotion; NeXT gcc can't handle it. */
172 void error
PARAMS ((int status
, int errnum
, char *message
,...));
174 extern char *version_string
;
176 /* The name this program was run with. */
179 /* Buffer for reading arguments from stdin. */
180 static char *linebuf
;
182 /* Line number in stdin since the last command was executed. */
183 static int lineno
= 0;
185 /* If nonzero, then instead of putting the args from stdin at
186 the end of the command argument list, they are each stuck into the
187 initial args, replacing each occurrence of the `replace_pat' in the
189 static char *replace_pat
= NULL
;
191 /* The length of `replace_pat'. */
192 static size_t rplen
= 0;
194 /* If nonzero, when this string is read on stdin it is treated as
196 I don't like this - it should default to NULL. */
197 static char *eof_str
= "_";
199 /* If nonzero, the maximum number of nonblank lines from stdin to use
201 static long lines_per_exec
= 0;
203 /* The maximum number of arguments to use per command line. */
204 static long args_per_exec
= 1024;
206 /* If true, exit if lines_per_exec or args_per_exec is exceeded. */
207 static boolean exit_if_size_exceeded
= false;
209 /* The maximum number of characters that can be used per command line. */
212 /* Storage for elements of `cmd_argv'. */
215 /* The list of args being built. */
216 static char **cmd_argv
= NULL
;
218 /* Number of elements allocated for `cmd_argv'. */
219 static int cmd_argv_alloc
= 0;
221 /* Number of valid elements in `cmd_argv'. */
222 static int cmd_argc
= 0;
224 /* Number of chars being used in `cmd_argv'. */
225 static int cmd_argv_chars
= 0;
227 /* Number of initial arguments given on the command line. */
228 static int initial_argc
= 0;
230 /* Number of chars in the initial args. */
231 static int initial_argv_chars
= 0;
233 /* true when building up initial arguments in `cmd_argv'. */
234 static boolean initial_args
= true;
236 /* If nonzero, the maximum number of child processes that can be running
238 static int proc_max
= 1;
240 /* Total number of child processes that have been executed. */
241 static int procs_executed
= 0;
243 /* The number of elements in `pids'. */
244 static int procs_executing
= 0;
246 /* List of child processes currently executing. */
247 static pid_t
*pids
= NULL
;
249 /* The number of allocated elements in `pids'. */
250 static int pids_alloc
= 0;
252 /* Exit status; nonzero if any child process exited with a
254 static int child_error
= 0;
256 /* If true, print each command on stderr before executing it. */
257 static boolean print_command
= false;
259 /* If true, query the user before executing each command, and only
260 execute the command if the user responds affirmatively. */
261 static boolean query_before_executing
= false;
263 static struct option
const longopts
[] =
265 {"null", no_argument
, NULL
, '0'},
266 {"eof", optional_argument
, NULL
, 'e'},
267 {"replace", optional_argument
, NULL
, 'i'},
268 {"max-lines", optional_argument
, NULL
, 'l'},
269 {"max-args", required_argument
, NULL
, 'n'},
270 {"interactive", no_argument
, NULL
, 'p'},
271 {"no-run-if-empty", no_argument
, NULL
, 'r'},
272 {"max-chars", required_argument
, NULL
, 's'},
273 {"verbose", no_argument
, NULL
, 't'},
274 {"exit", no_argument
, NULL
, 'x'},
275 {"max-procs", required_argument
, NULL
, 'P'},
276 {"version", no_argument
, NULL
, 'v'},
277 {"help", no_argument
, NULL
, 'h'},
278 {NULL
, no_argument
, NULL
, 0}
281 static int read_line
PARAMS ((void));
282 static int read_string
PARAMS ((void));
283 static char *mbstrstr
PARAMS ((const char *haystack
, const char *needle
));
284 static void do_insert
PARAMS ((char *arg
, size_t arglen
, size_t lblen
));
285 static void push_arg
PARAMS ((char *arg
, size_t len
));
286 static boolean print_args
PARAMS ((boolean ask
));
287 static void do_exec
PARAMS ((void));
288 static void add_proc
PARAMS ((pid_t pid
));
289 static void wait_for_proc
PARAMS ((boolean all
));
290 static long parse_num
PARAMS ((char *str
, int option
, long min
, long max
));
291 static long env_size
PARAMS ((char **envp
));
292 static void usage
PARAMS ((FILE * stream
));
295 main (int argc
, char **argv
)
298 int always_run_command
= 1;
301 long size_of_environment
= env_size(environ
);
302 char *default_cmd
= "/bin/echo";
303 int (*read_args
) PARAMS ((void)) = read_line
;
305 program_name
= argv
[0];
307 #ifdef HAVE_SETLOCALE
308 setlocale (LC_ALL
, "");
310 bindtextdomain (PACKAGE
, LOCALEDIR
);
311 textdomain (PACKAGE
);
313 orig_arg_max
= ARG_MAX
;
314 if (orig_arg_max
== -1)
315 orig_arg_max
= LONG_MAX
;
316 orig_arg_max
-= 2048; /* POSIX.2 requires subtracting 2048. */
317 arg_max
= orig_arg_max
;
319 arg_size
= 20 * 1048 + size_of_environment
;
322 /* Take the size of the environment into account. */
323 arg_max
-= env_size (environ
);
325 error (1, 0, _("environment is too large for exec"));
327 while ((optc
= getopt_long (argc
, argv
, "+0e::i::l::n:prs:txP:",
328 longopts
, (int *) 0)) != -1)
333 read_args
= read_string
;
349 replace_pat
= optarg
;
352 /* -i excludes -n -l. */
359 lines_per_exec
= parse_num (optarg
, 'l', 1L, -1L);
362 /* -l excludes -i -n. */
368 args_per_exec
= parse_num (optarg
, 'n', 1L, -1L);
369 /* -n excludes -i -l. */
371 if (args_per_exec
== 1 && replace_pat
)
372 /* ignore -n1 in '-i -n1' */
379 arg_size
= parse_num (optarg
, 's', 1L, orig_arg_max
);
383 print_command
= true;
387 exit_if_size_exceeded
= true;
391 query_before_executing
= true;
392 print_command
= true;
396 always_run_command
= 0;
400 proc_max
= parse_num (optarg
, 'P', 0L, -1L);
404 printf (_("GNU xargs version %s\n"), version_string
);
413 if (replace_pat
|| lines_per_exec
)
414 exit_if_size_exceeded
= true;
423 /* Taking into account the sisze of the environment,
424 * figure out how large a buffer we need to
425 * hold all the arguments. We cannot use ARG_MAX
426 * directly since that may be arbitrarily large.
427 * This is from a patch by Bob Prolux, <bob@proulx.com>.
429 if (arg_max
> arg_size
)
434 linebuf
= (char *) xmalloc (arg_max
+ 1);
435 argbuf
= (char *) xmalloc (arg_max
+ 1);
437 /* Make sure to listen for the kids. */
438 signal (SIGCHLD
, SIG_DFL
);
442 for (; optind
< argc
; optind
++)
443 push_arg (argv
[optind
], strlen (argv
[optind
]) + 1);
444 initial_args
= false;
445 initial_argc
= cmd_argc
;
446 initial_argv_chars
= cmd_argv_chars
;
448 while ((*read_args
) () != -1)
449 if (lines_per_exec
&& lineno
>= lines_per_exec
)
455 /* SYSV xargs seems to do at least one exec, even if the
457 if (cmd_argc
!= initial_argc
458 || (always_run_command
&& procs_executed
== 0))
465 size_t *arglen
= (size_t *) xmalloc (sizeof (size_t) * argc
);
467 for (i
= optind
; i
< argc
; i
++)
468 arglen
[i
] = strlen(argv
[i
]);
469 rplen
= strlen (replace_pat
);
470 while ((len
= (*read_args
) ()) != -1)
472 /* Don't do insert on the command name. */
473 push_arg (argv
[optind
], arglen
[optind
] + 1);
475 for (i
= optind
+ 1; i
< argc
; i
++)
476 do_insert (argv
[i
], arglen
[i
], len
);
481 wait_for_proc (true);
487 append_char_to_buf(char **pbuf
, char **pend
, char **pp
, int c
)
489 char *end_of_buffer
= *pend
;
490 char *start_of_buffer
= *pbuf
;
492 if (p
>= end_of_buffer
)
496 size_t len
= end_of_buffer
- start_of_buffer
;
497 size_t offset
= p
- start_of_buffer
;
499 start_of_buffer
= xrealloc(start_of_buffer
, len
*2);
500 if (NULL
!= start_of_buffer
)
502 end_of_buffer
= start_of_buffer
+ len
;
503 p
= start_of_buffer
+ offset
;
506 /* Update the caller's idea of where the buffer is. */
507 *pbuf
= start_of_buffer
;
508 *pend
= end_of_buffer
;
515 /* Failed to reallocate. */
521 /* I suspect that this can never happen now, because append_char_to_buf()
522 * should only be called wen replace_pat is true.
524 error (1, 0, _("argument line too long"));
531 /* Enough space remains. */
540 /* Read a line of arguments from stdin and add them to the list of
541 arguments to pass to the command. Ignore blank lines and initial blanks.
542 Single and double quotes and backslashes quote metacharacters and blanks
543 as they do in the shell.
544 Return -1 if eof (either physical or logical) is reached,
545 otherwise the length of the last string read (including the null). */
550 static boolean eof
= false;
551 /* Start out in mode SPACE to always strip leading spaces (even with -i). */
552 int state
= SPACE
; /* The type of character we last read. */
553 int prevc
; /* The previous value of c. */
554 int quotc
= 0; /* The last quote character read. */
556 boolean first
= true; /* true if reading first arg on line. */
559 /* Including the NUL, the args must not grow past this point. */
560 char *endbuf
= linebuf
+ arg_max
- initial_argv_chars
- 1;
570 /* COMPAT: SYSV seems to ignore stuff on a line that
571 ends without a \n; we don't. */
577 /* FIXME we don't check for unterminated quotes here. */
578 if (first
&& EOF_STR (linebuf
))
581 push_arg (linebuf
, len
);
595 if (!ISBLANK (prevc
))
596 lineno
++; /* For -l. */
605 if (EOF_STR (linebuf
))
608 return first
? -1 : len
;
611 push_arg (linebuf
, len
);
614 if (!replace_pat
&& ISSPACE (c
))
618 if (EOF_STR (linebuf
))
621 return first
? -1 : len
;
623 push_arg (linebuf
, len
);
645 error (1, 0, _("unmatched %s quote; by default quotes are special to xargs unless you use the -0 option"),
646 quotc
== '"' ? _("double") : _("single"));
660 error (1, 0, _("argument line too long"));
663 append_char_to_buf(&linebuf
, &endbuf
, &p
, c
);
668 /* Read a null-terminated string from stdin and add it to the list of
669 arguments to pass to the command.
670 Return -1 if eof (either physical or logical) is reached,
671 otherwise the length of the string read (including the null). */
676 static boolean eof
= false;
679 /* Including the NUL, the args must not grow past this point. */
680 char *endbuf
= linebuf
+ arg_max
- initial_argv_chars
- 1;
686 int c
= getc (stdin
);
695 push_arg (linebuf
, len
);
700 lineno
++; /* For -l. */
704 push_arg (linebuf
, len
);
708 error (1, 0, _("argument line too long"));
713 /* Finds the first occurrence of the substring NEEDLE in the string
714 HAYSTACK. Both strings can be multibyte strings. */
717 mbstrstr (const char *haystack
, const char *needle
)
722 size_t hlen
= strlen (haystack
);
723 size_t nlen
= strlen (needle
);
727 memset (&mbstate
, 0, sizeof (mbstate_t));
730 if (memcmp (haystack
, needle
, nlen
) == 0)
731 return (char *) haystack
;
732 step
= mbrlen (haystack
, hlen
, &mbstate
);
741 return strstr (haystack
, needle
);
744 /* Replace all instances of `replace_pat' in ARG with `linebuf',
745 and add the resulting string to the list of arguments for the command
747 ARGLEN is the length of ARG, not including the null.
748 LBLEN is the length of `linebuf', not including the null.
750 COMPAT: insertions on the SYSV version are limited to 255 chars per line,
751 and a max of 5 occurrences of replace_pat in the initial-arguments.
752 Those restrictions do not exist here. */
755 do_insert (char *arg
, size_t arglen
, size_t lblen
)
757 /* Temporary copy of each arg with the replace pattern replaced by the
759 static char *insertbuf
;
761 int bytes_left
= arg_max
- 1; /* Bytes left on the command line. */
764 insertbuf
= (char *) xmalloc (arg_max
+ 1);
769 size_t len
; /* Length in ARG before `replace_pat'. */
770 char *s
= mbstrstr (arg
, replace_pat
);
779 strncpy (p
, arg
, len
);
797 error (1, 0, _("command too long"));
799 push_arg (insertbuf
, p
- insertbuf
);
802 /* Add ARG to the end of the list of arguments `cmd_argv' to pass
804 LEN is the length of ARG, including the terminating null.
805 If this brings the list up to its maximum size, execute the command. */
808 push_arg (char *arg
, size_t len
)
812 if (cmd_argv_chars
+ len
> arg_max
)
814 if (initial_args
|| cmd_argc
== initial_argc
)
815 error (1, 0, _("can not fit single argument within argument list size limit"));
816 /* option -i (replace_pat) implies -x (exit_if_size_exceeded) */
818 || (exit_if_size_exceeded
&&
819 (lines_per_exec
|| args_per_exec
)))
820 error (1, 0, _("argument list too long"));
823 if (!initial_args
&& args_per_exec
&&
824 cmd_argc
- initial_argc
== args_per_exec
)
828 if (cmd_argc
>= cmd_argv_alloc
)
833 cmd_argv
= (char **) xmalloc (sizeof (char *) * cmd_argv_alloc
);
838 cmd_argv
= (char **) xrealloc (cmd_argv
,
839 sizeof (char *) * cmd_argv_alloc
);
844 cmd_argv
[cmd_argc
++] = NULL
;
847 cmd_argv
[cmd_argc
++] = argbuf
+ cmd_argv_chars
;
848 strcpy (argbuf
+ cmd_argv_chars
, arg
);
849 cmd_argv_chars
+= len
;
853 /* Print the arguments of the command to execute.
854 If ASK is nonzero, prompt the user for a response, and
855 if the user responds affirmatively, return true;
856 otherwise, return false. */
859 print_args (boolean ask
)
863 for (i
= 0; i
< cmd_argc
- 1; i
++)
864 fprintf (stderr
, "%s ", cmd_argv
[i
]);
867 static FILE *tty_stream
;
872 tty_stream
= fopen ("/dev/tty", "r");
874 error (1, errno
, "/dev/tty");
876 fputs ("?...", stderr
);
878 c
= savec
= getc (tty_stream
);
879 while (c
!= EOF
&& c
!= '\n')
880 c
= getc (tty_stream
);
881 if (savec
== 'y' || savec
== 'Y')
891 /* Close stdin and attach /dev/null to it.
892 * This resolves Savannah bug #3992.
895 prep_child_for_exec (void)
897 const char inputfile
[] = "/dev/null";
898 /* fprintf(stderr, "attaching stdin to /dev/null\n"); */
901 if (open(inputfile
, O_RDONLY
) < 0)
903 /* This is not entirely fatal, since
904 * executing the child with a closed
905 * stdin is almost as good as executing it
906 * with its stdin attached to /dev/null.
908 error (0, errno
, "%s", inputfile
);
913 /* Execute the command that has been built in `cmd_argv'. This may involve
914 waiting for processes that were previously executed. */
921 push_arg ((char *) NULL
, 0); /* Null terminate the arg list. */
922 if (!query_before_executing
|| print_args (true))
924 if (proc_max
&& procs_executing
>= proc_max
)
925 wait_for_proc (false);
926 if (!query_before_executing
&& print_command
)
928 /* If we run out of processes, wait for a child to return and
930 while ((child
= fork ()) < 0 && errno
== EAGAIN
&& procs_executing
)
931 wait_for_proc (false);
935 error (1, errno
, _("cannot fork"));
938 prep_child_for_exec();
939 execvp (cmd_argv
[0], cmd_argv
);
940 error (0, errno
, "%s", cmd_argv
[0]);
941 _exit (errno
== ENOENT
? 127 : 126);
947 cmd_argc
= initial_argc
;
948 cmd_argv_chars
= initial_argv_chars
;
951 /* Add the process with id PID to the list of processes that have
959 /* Find an empty slot. */
960 for (i
= 0; i
< pids_alloc
&& pids
[i
]; i
++)
966 pids_alloc
= proc_max
? proc_max
: 64;
967 pids
= (pid_t
*) xmalloc (sizeof (pid_t
) * pids_alloc
);
972 pids
= (pid_t
*) xrealloc (pids
,
973 sizeof (pid_t
) * pids_alloc
);
975 memset (&pids
[i
], '\0', sizeof (pid_t
) * (pids_alloc
- i
));
982 /* If ALL is true, wait for all child processes to finish;
983 otherwise, wait for one child process to finish.
984 Remove the processes that finish from the list of executing processes. */
987 wait_for_proc (boolean all
)
989 while (procs_executing
)
997 while ((pid
= wait (&status
)) == (pid_t
) -1)
999 error (1, errno
, _("error waiting for child process"));
1001 /* Find the entry in `pids' for the child process
1003 for (i
= 0; i
< pids_alloc
&& pid
!= pids
[i
]; i
++)
1006 while (i
== pids_alloc
); /* A child died that we didn't start? */
1008 /* Remove the child from the list. */
1012 if (WEXITSTATUS (status
) == 126 || WEXITSTATUS (status
) == 127)
1013 exit (WEXITSTATUS (status
)); /* Can't find or run the command. */
1014 if (WEXITSTATUS (status
) == 255)
1015 error (124, 0, _("%s: exited with status 255; aborting"), cmd_argv
[0]);
1016 if (WIFSTOPPED (status
))
1017 error (125, 0, _("%s: stopped by signal %d"), cmd_argv
[0], WSTOPSIG (status
));
1018 if (WIFSIGNALED (status
))
1019 error (125, 0, _("%s: terminated by signal %d"), cmd_argv
[0], WTERMSIG (status
));
1020 if (WEXITSTATUS (status
) != 0)
1028 /* Return the value of the number represented in STR.
1029 OPTION is the command line option to which STR is the argument.
1030 If the value does not fall within the boundaries MIN and MAX,
1031 Print an error message mentioning OPTION and exit. */
1034 parse_num (char *str
, int option
, long int min
, long int max
)
1039 val
= strtol (str
, &eptr
, 10);
1040 if (eptr
== str
|| *eptr
)
1042 fprintf (stderr
, _("%s: invalid number for -%c option\n"),
1043 program_name
, option
);
1049 fprintf (stderr
, _("%s: value for -%c option must be >= %ld\n"),
1050 program_name
, option
, min
);
1054 else if (max
>= 0 && val
> max
)
1056 fprintf (stderr
, _("%s: value for -%c option must be < %ld\n"),
1057 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
)
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
);