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., 59 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) ()
35 #if !defined (isascii) || defined (STDC_HEADERS)
43 #define ISBLANK(c) (isascii (c) && isblank (c))
45 #define ISBLANK(c) ((c) == ' ' || (c) == '\t')
48 #define ISSPACE(c) (ISBLANK (c) || (c) == '\n' || (c) == '\r' \
49 || (c) == '\f' || (c) == '\v')
51 #include <sys/types.h>
57 #if defined(HAVE_STRING_H) || defined(STDC_HEADERS)
59 #if !defined(STDC_HEADERS)
64 #define memcpy(dest, source, count) (bcopy((source), (dest), (count)))
68 #include <sys/param.h>
76 #define LONG_MAX (~(1 << (sizeof (long) * 8 - 1)))
85 #if !defined(SIGCHLD) && defined(SIGCLD)
86 #define SIGCHLD SIGCLD
89 /* COMPAT: SYSV version defaults size (and has a max value of) to 470.
90 We try to make it as large as possible. */
91 #if !defined(ARG_MAX) && defined(_SC_ARG_MAX)
92 #define ARG_MAX sysconf (_SC_ARG_MAX)
95 #define ARG_MAX NCARGS
100 /* States for read_line. */
116 # include <libintl.h>
117 # define _(Text) gettext (Text)
119 # define _(Text) Text
120 #define textdomain(Domain)
121 #define bindtextdomain(Package, Directory)
124 # define N_(String) gettext_noop (String)
126 /* See locate.c for explanation as to why not use (String) */
127 # define N_(String) String
130 /* Return nonzero if S is the EOF string. */
131 #define EOF_STR(s) (eof_str && *eof_str == *s && !strcmp (eof_str, s))
133 extern char **environ
;
135 /* Do multibyte processing if multibyte characters are supported,
136 unless multibyte sequences are search safe. Multibyte sequences
137 are search safe if searching for a substring using the byte
138 comparison function 'strstr' gives no false positives. All 8-bit
139 encodings and the UTF-8 multibyte encoding are search safe, but
140 the EUC encodings are not.
141 BeOS uses the UTF-8 encoding exclusively, so it is search safe. */
143 # define MULTIBYTE_IS_SEARCH_SAFE 1
145 #define DO_MULTIBYTE (HAVE_MBLEN && ! MULTIBYTE_IS_SEARCH_SAFE)
151 /* Simulate mbrlen with mblen as best we can. */
152 # define mbstate_t int
153 # define mbrlen(s, n, ps) mblen (s, n)
157 /* Not char because of type promotion; NeXT gcc can't handle it. */
169 void error
PARAMS ((int status
, int errnum
, char *message
,...));
171 extern char *version_string
;
173 /* The name this program was run with. */
176 /* Buffer for reading arguments from stdin. */
177 static char *linebuf
;
179 /* Line number in stdin since the last command was executed. */
180 static int lineno
= 0;
182 /* If nonzero, then instead of putting the args from stdin at
183 the end of the command argument list, they are each stuck into the
184 initial args, replacing each occurrence of the `replace_pat' in the
186 static char *replace_pat
= NULL
;
188 /* The length of `replace_pat'. */
189 static size_t rplen
= 0;
191 /* If nonzero, when this string is read on stdin it is treated as
193 I don't like this - it should default to NULL. */
194 static char *eof_str
= "_";
196 /* If nonzero, the maximum number of nonblank lines from stdin to use
198 static long lines_per_exec
= 0;
200 /* The maximum number of arguments to use per command line. */
201 static long args_per_exec
= 1024;
203 /* If true, exit if lines_per_exec or args_per_exec is exceeded. */
204 static boolean exit_if_size_exceeded
= false;
206 /* The maximum number of characters that can be used per command line. */
209 /* Storage for elements of `cmd_argv'. */
212 /* The list of args being built. */
213 static char **cmd_argv
= NULL
;
215 /* Number of elements allocated for `cmd_argv'. */
216 static int cmd_argv_alloc
= 0;
218 /* Number of valid elements in `cmd_argv'. */
219 static int cmd_argc
= 0;
221 /* Number of chars being used in `cmd_argv'. */
222 static int cmd_argv_chars
= 0;
224 /* Number of initial arguments given on the command line. */
225 static int initial_argc
= 0;
227 /* Number of chars in the initial args. */
228 static int initial_argv_chars
= 0;
230 /* true when building up initial arguments in `cmd_argv'. */
231 static boolean initial_args
= true;
233 /* If nonzero, the maximum number of child processes that can be running
235 static int proc_max
= 1;
237 /* Total number of child processes that have been executed. */
238 static int procs_executed
= 0;
240 /* The number of elements in `pids'. */
241 static int procs_executing
= 0;
243 /* List of child processes currently executing. */
244 static pid_t
*pids
= NULL
;
246 /* The number of allocated elements in `pids'. */
247 static int pids_alloc
= 0;
249 /* Exit status; nonzero if any child process exited with a
251 static int child_error
= 0;
253 /* If true, print each command on stderr before executing it. */
254 static boolean print_command
= false;
256 /* If true, query the user before executing each command, and only
257 execute the command if the user responds affirmatively. */
258 static boolean query_before_executing
= false;
260 static struct option
const longopts
[] =
262 {"null", no_argument
, NULL
, '0'},
263 {"eof", optional_argument
, NULL
, 'e'},
264 {"replace", optional_argument
, NULL
, 'i'},
265 {"max-lines", optional_argument
, NULL
, 'l'},
266 {"max-args", required_argument
, NULL
, 'n'},
267 {"interactive", no_argument
, NULL
, 'p'},
268 {"no-run-if-empty", no_argument
, NULL
, 'r'},
269 {"max-chars", required_argument
, NULL
, 's'},
270 {"verbose", no_argument
, NULL
, 't'},
271 {"exit", no_argument
, NULL
, 'x'},
272 {"max-procs", required_argument
, NULL
, 'P'},
273 {"version", no_argument
, NULL
, 'v'},
274 {"help", no_argument
, NULL
, 'h'},
275 {NULL
, no_argument
, NULL
, 0}
278 static int read_line
PARAMS ((void));
279 static int read_string
PARAMS ((void));
280 static char *mbstrstr
PARAMS ((const char *haystack
, const char *needle
));
281 static void do_insert
PARAMS ((char *arg
, size_t arglen
, size_t lblen
));
282 static void push_arg
PARAMS ((char *arg
, size_t len
));
283 static boolean print_args
PARAMS ((boolean ask
));
284 static void do_exec
PARAMS ((void));
285 static void add_proc
PARAMS ((pid_t pid
));
286 static void wait_for_proc
PARAMS ((boolean all
));
287 static long parse_num
PARAMS ((char *str
, int option
, long min
, long max
));
288 static long env_size
PARAMS ((char **envp
));
289 static void usage
PARAMS ((FILE * stream
));
292 main (int argc
, char **argv
)
295 int always_run_command
= 1;
298 long size_of_environment
= env_size(environ
);
299 char *default_cmd
= "/bin/echo";
300 int (*read_args
) PARAMS ((void)) = read_line
;
302 program_name
= argv
[0];
304 #ifdef HAVE_SETLOCALE
305 setlocale (LC_ALL
, "");
307 bindtextdomain (PACKAGE
, LOCALEDIR
);
308 textdomain (PACKAGE
);
310 orig_arg_max
= ARG_MAX
;
311 if (orig_arg_max
== -1)
312 orig_arg_max
= LONG_MAX
;
313 orig_arg_max
-= 2048; /* POSIX.2 requires subtracting 2048. */
314 arg_max
= orig_arg_max
;
316 arg_size
= 20 * 1048 + size_of_environment
;
319 /* Take the size of the environment into account. */
320 arg_max
-= env_size (environ
);
322 error (1, 0, _("environment is too large for exec"));
324 while ((optc
= getopt_long (argc
, argv
, "+0e::i::l::n:prs:txP:",
325 longopts
, (int *) 0)) != -1)
330 read_args
= read_string
;
346 replace_pat
= optarg
;
349 /* -i excludes -n -l. */
356 lines_per_exec
= parse_num (optarg
, 'l', 1L, -1L);
359 /* -l excludes -i -n. */
365 args_per_exec
= parse_num (optarg
, 'n', 1L, -1L);
366 /* -n excludes -i -l. */
368 if (args_per_exec
== 1 && replace_pat
)
369 /* ignore -n1 in '-i -n1' */
376 arg_size
= parse_num (optarg
, 's', 1L, orig_arg_max
);
380 print_command
= true;
384 exit_if_size_exceeded
= true;
388 query_before_executing
= true;
389 print_command
= true;
393 always_run_command
= 0;
397 proc_max
= parse_num (optarg
, 'P', 0L, -1L);
401 printf (_("GNU xargs version %s\n"), version_string
);
410 if (replace_pat
|| lines_per_exec
)
411 exit_if_size_exceeded
= true;
420 /* Taking into account the sisze of the environment,
421 * figure out how large a buffer we need to
422 * hold all the arguments. We cannot use ARG_MAX
423 * directly since that may be arbitrarily large.
424 * This is from a patch by Bob Prolux, <bob@proulx.com>.
426 if (arg_max
> arg_size
)
431 linebuf
= (char *) xmalloc (arg_max
+ 1);
432 argbuf
= (char *) xmalloc (arg_max
+ 1);
434 /* Make sure to listen for the kids. */
435 signal (SIGCHLD
, SIG_DFL
);
439 for (; optind
< argc
; optind
++)
440 push_arg (argv
[optind
], strlen (argv
[optind
]) + 1);
441 initial_args
= false;
442 initial_argc
= cmd_argc
;
443 initial_argv_chars
= cmd_argv_chars
;
445 while ((*read_args
) () != -1)
446 if (lines_per_exec
&& lineno
>= lines_per_exec
)
452 /* SYSV xargs seems to do at least one exec, even if the
454 if (cmd_argc
!= initial_argc
455 || (always_run_command
&& procs_executed
== 0))
462 size_t *arglen
= (size_t *) xmalloc (sizeof (size_t) * argc
);
464 for (i
= optind
; i
< argc
; i
++)
465 arglen
[i
] = strlen(argv
[i
]);
466 rplen
= strlen (replace_pat
);
467 while ((len
= (*read_args
) ()) != -1)
469 /* Don't do insert on the command name. */
470 push_arg (argv
[optind
], arglen
[optind
] + 1);
472 for (i
= optind
+ 1; i
< argc
; i
++)
473 do_insert (argv
[i
], arglen
[i
], len
);
478 wait_for_proc (true);
484 append_char_to_buf(char **pbuf
, char **pend
, char **pp
, int c
)
486 char *end_of_buffer
= *pend
;
487 char *start_of_buffer
= *pbuf
;
489 if (p
>= end_of_buffer
)
493 size_t len
= end_of_buffer
- start_of_buffer
;
494 size_t offset
= p
- start_of_buffer
;
496 start_of_buffer
= xrealloc(start_of_buffer
, len
*2);
497 if (NULL
!= start_of_buffer
)
499 end_of_buffer
= start_of_buffer
+ len
;
500 p
= start_of_buffer
+ offset
;
503 /* Update the caller's idea of where the buffer is. */
504 *pbuf
= start_of_buffer
;
505 *pend
= end_of_buffer
;
512 /* Failed to reallocate. */
518 /* I suspect that this can never happen now, because append_char_to_buf()
519 * should only be called wen replace_pat is true.
521 error (1, 0, _("argument line too long"));
528 /* Enough space remains. */
537 /* Read a line of arguments from stdin and add them to the list of
538 arguments to pass to the command. Ignore blank lines and initial blanks.
539 Single and double quotes and backslashes quote metacharacters and blanks
540 as they do in the shell.
541 Return -1 if eof (either physical or logical) is reached,
542 otherwise the length of the last string read (including the null). */
547 static boolean eof
= false;
548 /* Start out in mode SPACE to always strip leading spaces (even with -i). */
549 int state
= SPACE
; /* The type of character we last read. */
550 int prevc
; /* The previous value of c. */
551 int quotc
= 0; /* The last quote character read. */
553 boolean first
= true; /* true if reading first arg on line. */
556 /* Including the NUL, the args must not grow past this point. */
557 char *endbuf
= linebuf
+ arg_max
- initial_argv_chars
- 1;
567 /* COMPAT: SYSV seems to ignore stuff on a line that
568 ends without a \n; we don't. */
574 /* FIXME we don't check for unterminated quotes here. */
575 if (first
&& EOF_STR (linebuf
))
578 push_arg (linebuf
, len
);
592 if (!ISBLANK (prevc
))
593 lineno
++; /* For -l. */
602 if (EOF_STR (linebuf
))
605 return first
? -1 : len
;
608 push_arg (linebuf
, len
);
611 if (!replace_pat
&& ISSPACE (c
))
615 if (EOF_STR (linebuf
))
618 return first
? -1 : len
;
620 push_arg (linebuf
, len
);
642 error (1, 0, _("unmatched %s quote; by default quotes are special to xargs unless you use the -0 option"),
643 quotc
== '"' ? _("double") : _("single"));
657 error (1, 0, _("argument line too long"));
660 append_char_to_buf(&linebuf
, &endbuf
, &p
, c
);
665 /* Read a null-terminated string from stdin and add it to the list of
666 arguments to pass to the command.
667 Return -1 if eof (either physical or logical) is reached,
668 otherwise the length of the string read (including the null). */
673 static boolean eof
= false;
676 /* Including the NUL, the args must not grow past this point. */
677 char *endbuf
= linebuf
+ arg_max
- initial_argv_chars
- 1;
683 int c
= getc (stdin
);
692 push_arg (linebuf
, len
);
697 lineno
++; /* For -l. */
701 push_arg (linebuf
, len
);
705 error (1, 0, _("argument line too long"));
710 /* Finds the first occurrence of the substring NEEDLE in the string
711 HAYSTACK. Both strings can be multibyte strings. */
714 mbstrstr (const char *haystack
, const char *needle
)
719 size_t hlen
= strlen (haystack
);
720 size_t nlen
= strlen (needle
);
724 memset (&mbstate
, 0, sizeof (mbstate_t));
727 if (memcmp (haystack
, needle
, nlen
) == 0)
728 return (char *) haystack
;
729 step
= mbrlen (haystack
, hlen
, &mbstate
);
738 return strstr (haystack
, needle
);
741 /* Replace all instances of `replace_pat' in ARG with `linebuf',
742 and add the resulting string to the list of arguments for the command
744 ARGLEN is the length of ARG, not including the null.
745 LBLEN is the length of `linebuf', not including the null.
747 COMPAT: insertions on the SYSV version are limited to 255 chars per line,
748 and a max of 5 occurrences of replace_pat in the initial-arguments.
749 Those restrictions do not exist here. */
752 do_insert (char *arg
, size_t arglen
, size_t lblen
)
754 /* Temporary copy of each arg with the replace pattern replaced by the
756 static char *insertbuf
;
758 int bytes_left
= arg_max
- 1; /* Bytes left on the command line. */
761 insertbuf
= (char *) xmalloc (arg_max
+ 1);
766 size_t len
; /* Length in ARG before `replace_pat'. */
767 char *s
= mbstrstr (arg
, replace_pat
);
776 strncpy (p
, arg
, len
);
794 error (1, 0, _("command too long"));
796 push_arg (insertbuf
, p
- insertbuf
);
799 /* Add ARG to the end of the list of arguments `cmd_argv' to pass
801 LEN is the length of ARG, including the terminating null.
802 If this brings the list up to its maximum size, execute the command. */
805 push_arg (char *arg
, size_t len
)
809 if (cmd_argv_chars
+ len
> arg_max
)
811 if (initial_args
|| cmd_argc
== initial_argc
)
812 error (1, 0, _("can not fit single argument within argument list size limit"));
813 /* option -i (replace_pat) implies -x (exit_if_size_exceeded) */
815 || (exit_if_size_exceeded
&&
816 (lines_per_exec
|| args_per_exec
)))
817 error (1, 0, _("argument list too long"));
820 if (!initial_args
&& args_per_exec
&&
821 cmd_argc
- initial_argc
== args_per_exec
)
825 if (cmd_argc
>= cmd_argv_alloc
)
830 cmd_argv
= (char **) xmalloc (sizeof (char *) * cmd_argv_alloc
);
835 cmd_argv
= (char **) xrealloc (cmd_argv
,
836 sizeof (char *) * cmd_argv_alloc
);
841 cmd_argv
[cmd_argc
++] = NULL
;
844 cmd_argv
[cmd_argc
++] = argbuf
+ cmd_argv_chars
;
845 strcpy (argbuf
+ cmd_argv_chars
, arg
);
846 cmd_argv_chars
+= len
;
850 /* Print the arguments of the command to execute.
851 If ASK is nonzero, prompt the user for a response, and
852 if the user responds affirmatively, return true;
853 otherwise, return false. */
856 print_args (boolean ask
)
860 for (i
= 0; i
< cmd_argc
- 1; i
++)
861 fprintf (stderr
, "%s ", cmd_argv
[i
]);
864 static FILE *tty_stream
;
869 tty_stream
= fopen ("/dev/tty", "r");
871 error (1, errno
, "/dev/tty");
873 fputs ("?...", stderr
);
875 c
= savec
= getc (tty_stream
);
876 while (c
!= EOF
&& c
!= '\n')
877 c
= getc (tty_stream
);
878 if (savec
== 'y' || savec
== 'Y')
888 /* Close stdin and attach /dev/null to it.
889 * This resolves Savannah bug #3992.
892 prep_child_for_exec (void)
894 const char inputfile
[] = "/dev/null";
895 /* fprintf(stderr, "attaching stdin to /dev/null\n"); */
898 if (open(inputfile
, O_RDONLY
) < 0)
900 /* This is not entirely fatal, since
901 * executing the child with a closed
902 * stdin is almost as good as executing it
903 * with its stdin attached to /dev/null.
905 error (0, errno
, "%s", inputfile
);
910 /* Execute the command that has been built in `cmd_argv'. This may involve
911 waiting for processes that were previously executed. */
918 push_arg ((char *) NULL
, 0); /* Null terminate the arg list. */
919 if (!query_before_executing
|| print_args (true))
921 if (proc_max
&& procs_executing
>= proc_max
)
922 wait_for_proc (false);
923 if (!query_before_executing
&& print_command
)
925 /* If we run out of processes, wait for a child to return and
927 while ((child
= fork ()) < 0 && errno
== EAGAIN
&& procs_executing
)
928 wait_for_proc (false);
932 error (1, errno
, _("cannot fork"));
935 prep_child_for_exec();
936 execvp (cmd_argv
[0], cmd_argv
);
937 error (0, errno
, "%s", cmd_argv
[0]);
938 _exit (errno
== ENOENT
? 127 : 126);
944 cmd_argc
= initial_argc
;
945 cmd_argv_chars
= initial_argv_chars
;
948 /* Add the process with id PID to the list of processes that have
956 /* Find an empty slot. */
957 for (i
= 0; i
< pids_alloc
&& pids
[i
]; i
++)
963 pids_alloc
= proc_max
? proc_max
: 64;
964 pids
= (pid_t
*) xmalloc (sizeof (pid_t
) * pids_alloc
);
969 pids
= (pid_t
*) xrealloc (pids
,
970 sizeof (pid_t
) * pids_alloc
);
972 memset (&pids
[i
], '\0', sizeof (pid_t
) * (pids_alloc
- i
));
979 /* If ALL is true, wait for all child processes to finish;
980 otherwise, wait for one child process to finish.
981 Remove the processes that finish from the list of executing processes. */
984 wait_for_proc (boolean all
)
986 while (procs_executing
)
994 while ((pid
= wait (&status
)) == (pid_t
) -1)
996 error (1, errno
, _("error waiting for child process"));
998 /* Find the entry in `pids' for the child process
1000 for (i
= 0; i
< pids_alloc
&& pid
!= pids
[i
]; i
++)
1003 while (i
== pids_alloc
); /* A child died that we didn't start? */
1005 /* Remove the child from the list. */
1009 if (WEXITSTATUS (status
) == 126 || WEXITSTATUS (status
) == 127)
1010 exit (WEXITSTATUS (status
)); /* Can't find or run the command. */
1011 if (WEXITSTATUS (status
) == 255)
1012 error (124, 0, _("%s: exited with status 255; aborting"), cmd_argv
[0]);
1013 if (WIFSTOPPED (status
))
1014 error (125, 0, _("%s: stopped by signal %d"), cmd_argv
[0], WSTOPSIG (status
));
1015 if (WIFSIGNALED (status
))
1016 error (125, 0, _("%s: terminated by signal %d"), cmd_argv
[0], WTERMSIG (status
));
1017 if (WEXITSTATUS (status
) != 0)
1025 /* Return the value of the number represented in STR.
1026 OPTION is the command line option to which STR is the argument.
1027 If the value does not fall within the boundaries MIN and MAX,
1028 Print an error message mentioning OPTION and exit. */
1031 parse_num (char *str
, int option
, long int min
, long int max
)
1036 val
= strtol (str
, &eptr
, 10);
1037 if (eptr
== str
|| *eptr
)
1039 fprintf (stderr
, _("%s: invalid number for -%c option\n"),
1040 program_name
, option
);
1046 fprintf (stderr
, _("%s: value for -%c option must be >= %ld\n"),
1047 program_name
, option
, min
);
1051 else if (max
>= 0 && val
> max
)
1053 fprintf (stderr
, _("%s: value for -%c option must be < %ld\n"),
1054 program_name
, option
, max
);
1061 /* Return how much of ARG_MAX is used by the environment. */
1064 env_size (char **envp
)
1069 len
+= strlen (*envp
++) + 1;
1075 usage (FILE *stream
)
1077 fprintf (stream
, _("\
1078 Usage: %s [-0prtx] [-e[eof-str]] [-i[replace-str]] [-l[max-lines]]\n\
1079 [-n max-args] [-s max-chars] [-P max-procs] [--null] [--eof[=eof-str]]\n\
1080 [--replace[=replace-str]] [--max-lines[=max-lines]] [--interactive]\n\
1081 [--max-chars=max-chars] [--verbose] [--exit] [--max-procs=max-procs]\n\
1082 [--max-args=max-args] [--no-run-if-empty] [--version] [--help]\n\
1083 [command [initial-arguments]]\n"),
1085 fputs (_("\nReport bugs to <bug-findutils@gnu.org>.\n"), stream
);