1 /* POSIX.2 wordexp implementation.
2 Copyright (C) 1997-2014 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Tim Waugh <tim@cyberelk.demon.co.uk>.
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, see
18 <http://www.gnu.org/licenses/>. */
34 #include <sys/param.h>
37 #include <sys/types.h>
38 #include <sys/types.h>
43 #include <kernel-features.h>
45 #include <bits/libc-lock.h>
48 /* Undefine the following line for the production version. */
49 /* #define NDEBUG 1 */
52 /* Get some device information. */
53 #include <device-nrs.h>
56 * This is a recursive-descent-style word expansion routine.
59 /* These variables are defined and initialized in the startup code. */
60 extern int __libc_argc attribute_hidden
;
61 extern char **__libc_argv attribute_hidden
;
63 /* Some forward declarations */
64 static int parse_dollars (char **word
, size_t *word_length
, size_t *max_length
,
65 const char *words
, size_t *offset
, int flags
,
66 wordexp_t
*pwordexp
, const char *ifs
,
67 const char *ifs_white
, int quoted
)
69 static int parse_backtick (char **word
, size_t *word_length
,
70 size_t *max_length
, const char *words
,
71 size_t *offset
, int flags
, wordexp_t
*pwordexp
,
72 const char *ifs
, const char *ifs_white
)
74 static int parse_dquote (char **word
, size_t *word_length
, size_t *max_length
,
75 const char *words
, size_t *offset
, int flags
,
76 wordexp_t
*pwordexp
, const char *ifs
,
77 const char *ifs_white
)
79 static int eval_expr (char *expr
, long int *result
) internal_function
;
81 /* The w_*() functions manipulate word lists. */
85 /* Result of w_newword will be ignored if it's the last word. */
87 w_newword (size_t *actlen
, size_t *maxlen
)
89 *actlen
= *maxlen
= 0;
94 w_addchar (char *buffer
, size_t *actlen
, size_t *maxlen
, char ch
)
95 /* (lengths exclude trailing zero) */
97 /* Add a character to the buffer, allocating room for it if needed. */
99 if (*actlen
== *maxlen
)
101 char *old_buffer
= buffer
;
102 assert (buffer
== NULL
|| *maxlen
!= 0);
104 buffer
= (char *) realloc (buffer
, 1 + *maxlen
);
112 buffer
[*actlen
] = ch
;
113 buffer
[++(*actlen
)] = '\0';
121 w_addmem (char *buffer
, size_t *actlen
, size_t *maxlen
, const char *str
,
124 /* Add a string to the buffer, allocating room for it if needed.
126 if (*actlen
+ len
> *maxlen
)
128 char *old_buffer
= buffer
;
129 assert (buffer
== NULL
|| *maxlen
!= 0);
130 *maxlen
+= MAX (2 * len
, W_CHUNK
);
131 buffer
= realloc (old_buffer
, 1 + *maxlen
);
139 *((char *) __mempcpy (&buffer
[*actlen
], str
, len
)) = '\0';
148 w_addstr (char *buffer
, size_t *actlen
, size_t *maxlen
, const char *str
)
149 /* (lengths exclude trailing zero) */
151 /* Add a string to the buffer, allocating room for it if needed.
155 assert (str
!= NULL
); /* w_addstr only called from this file */
158 return w_addmem (buffer
, actlen
, maxlen
, str
, len
);
163 w_addword (wordexp_t
*pwordexp
, char *word
)
165 /* Add a word to the wordlist */
168 bool allocated
= false;
170 /* Internally, NULL acts like "". Convert NULLs to "" before
171 * the caller sees them.
175 word
= __strdup ("");
181 num_p
= 2 + pwordexp
->we_wordc
+ pwordexp
->we_offs
;
182 new_wordv
= realloc (pwordexp
->we_wordv
, sizeof (char *) * num_p
);
183 if (new_wordv
!= NULL
)
185 pwordexp
->we_wordv
= new_wordv
;
186 pwordexp
->we_wordv
[pwordexp
->we_offs
+ pwordexp
->we_wordc
++] = word
;
187 pwordexp
->we_wordv
[pwordexp
->we_offs
+ pwordexp
->we_wordc
] = NULL
;
198 /* The parse_*() functions should leave *offset being the offset in 'words'
199 * to the last character processed.
204 parse_backslash (char **word
, size_t *word_length
, size_t *max_length
,
205 const char *words
, size_t *offset
)
207 /* We are poised _at_ a backslash, not in quotes */
209 switch (words
[1 + *offset
])
212 /* Backslash is last character of input words */
220 *word
= w_addchar (*word
, word_length
, max_length
, words
[1 + *offset
]);
233 parse_qtd_backslash (char **word
, size_t *word_length
, size_t *max_length
,
234 const char *words
, size_t *offset
)
236 /* We are poised _at_ a backslash, inside quotes */
238 switch (words
[1 + *offset
])
241 /* Backslash is last character of input words */
252 *word
= w_addchar (*word
, word_length
, max_length
, words
[1 + *offset
]);
260 *word
= w_addchar (*word
, word_length
, max_length
, words
[*offset
]);
262 *word
= w_addchar (*word
, word_length
, max_length
, words
[1 + *offset
]);
276 parse_tilde (char **word
, size_t *word_length
, size_t *max_length
,
277 const char *words
, size_t *offset
, size_t wordc
)
279 /* We are poised _at_ a tilde */
282 if (*word_length
!= 0)
284 if (!((*word
)[*word_length
- 1] == '=' && wordc
== 0))
286 if (!((*word
)[*word_length
- 1] == ':'
287 && strchr (*word
, '=') && wordc
== 0))
289 *word
= w_addchar (*word
, word_length
, max_length
, '~');
290 return *word
? 0 : WRDE_NOSPACE
;
295 for (i
= 1 + *offset
; words
[i
]; i
++)
297 if (words
[i
] == ':' || words
[i
] == '/' || words
[i
] == ' ' ||
298 words
[i
] == '\t' || words
[i
] == 0 )
301 if (words
[i
] == '\\')
303 *word
= w_addchar (*word
, word_length
, max_length
, '~');
304 return *word
? 0 : WRDE_NOSPACE
;
308 if (i
== 1 + *offset
)
310 /* Tilde appears on its own */
312 struct passwd pwd
, *tpwd
;
318 /* POSIX.2 says ~ expands to $HOME and if HOME is unset the
319 results are unspecified. We do a lookup on the uid if
322 home
= getenv ("HOME");
325 *word
= w_addstr (*word
, word_length
, max_length
, home
);
332 buffer
= __alloca (buflen
);
334 while ((result
= __getpwuid_r (uid
, &pwd
, buffer
, buflen
, &tpwd
)) != 0
336 buffer
= extend_alloca (buffer
, buflen
, buflen
+ 1000);
338 if (result
== 0 && tpwd
!= NULL
&& pwd
.pw_dir
!= NULL
)
340 *word
= w_addstr (*word
, word_length
, max_length
, pwd
.pw_dir
);
346 *word
= w_addchar (*word
, word_length
, max_length
, '~');
354 /* Look up user name in database to get home directory */
355 char *user
= strndupa (&words
[1 + *offset
], i
- (1 + *offset
));
356 struct passwd pwd
, *tpwd
;
358 char* buffer
= __alloca (buflen
);
361 while ((result
= __getpwnam_r (user
, &pwd
, buffer
, buflen
, &tpwd
)) != 0
363 buffer
= extend_alloca (buffer
, buflen
, buflen
+ 1000);
365 if (result
== 0 && tpwd
!= NULL
&& pwd
.pw_dir
)
366 *word
= w_addstr (*word
, word_length
, max_length
, pwd
.pw_dir
);
369 /* (invalid login name) */
370 *word
= w_addchar (*word
, word_length
, max_length
, '~');
372 *word
= w_addstr (*word
, word_length
, max_length
, user
);
377 return *word
? 0 : WRDE_NOSPACE
;
383 do_parse_glob (const char *glob_word
, char **word
, size_t *word_length
,
384 size_t *max_length
, wordexp_t
*pwordexp
, const char *ifs
,
385 const char *ifs_white
)
391 error
= glob (glob_word
, GLOB_NOCHECK
, NULL
, &globbuf
);
395 /* We can only run into memory problems. */
396 assert (error
== GLOB_NOSPACE
);
402 /* No field splitting allowed. */
403 assert (globbuf
.gl_pathv
[0] != NULL
);
404 *word
= w_addstr (*word
, word_length
, max_length
, globbuf
.gl_pathv
[0]);
405 for (match
= 1; match
< globbuf
.gl_pathc
&& *word
!= NULL
; ++match
)
407 *word
= w_addchar (*word
, word_length
, max_length
, ' ');
409 *word
= w_addstr (*word
, word_length
, max_length
,
410 globbuf
.gl_pathv
[match
]);
414 return *word
? 0 : WRDE_NOSPACE
;
417 assert (ifs
== NULL
|| *ifs
!= '\0');
421 *word
= w_newword (word_length
, max_length
);
424 for (match
= 0; match
< globbuf
.gl_pathc
; ++match
)
426 char *matching_word
= __strdup (globbuf
.gl_pathv
[match
]);
427 if (matching_word
== NULL
|| w_addword (pwordexp
, matching_word
))
440 parse_glob (char **word
, size_t *word_length
, size_t *max_length
,
441 const char *words
, size_t *offset
, int flags
,
442 wordexp_t
*pwordexp
, const char *ifs
, const char *ifs_white
)
444 /* We are poised just after a '*', a '[' or a '?'. */
445 int error
= WRDE_NOSPACE
;
446 int quoted
= 0; /* 1 if singly-quoted, 2 if doubly */
448 wordexp_t glob_list
; /* List of words to glob */
450 glob_list
.we_wordc
= 0;
451 glob_list
.we_wordv
= NULL
;
452 glob_list
.we_offs
= 0;
453 for (; words
[*offset
] != '\0'; ++*offset
)
455 if (strchr (ifs
, words
[*offset
]) != NULL
)
459 /* Sort out quoting */
460 if (words
[*offset
] == '\'')
467 else if (quoted
== 1)
473 else if (words
[*offset
] == '"')
480 else if (quoted
== 2)
487 /* Sort out other special characters */
488 if (quoted
!= 1 && words
[*offset
] == '$')
490 error
= parse_dollars (word
, word_length
, max_length
, words
,
491 offset
, flags
, &glob_list
, ifs
, ifs_white
,
498 else if (words
[*offset
] == '\\')
501 error
= parse_qtd_backslash (word
, word_length
, max_length
,
504 error
= parse_backslash (word
, word_length
, max_length
,
513 *word
= w_addchar (*word
, word_length
, max_length
, words
[*offset
]);
518 /* Don't forget to re-parse the character we stopped at. */
522 error
= w_addword (&glob_list
, *word
);
523 *word
= w_newword (word_length
, max_length
);
524 for (i
= 0; error
== 0 && i
< glob_list
.we_wordc
; i
++)
525 error
= do_parse_glob (glob_list
.we_wordv
[i
], word
, word_length
,
526 max_length
, pwordexp
, ifs
, ifs_white
);
530 wordfree (&glob_list
);
536 parse_squote (char **word
, size_t *word_length
, size_t *max_length
,
537 const char *words
, size_t *offset
)
539 /* We are poised just after a single quote */
540 for (; words
[*offset
]; ++(*offset
))
542 if (words
[*offset
] != '\'')
544 *word
= w_addchar (*word
, word_length
, max_length
, words
[*offset
]);
551 /* Unterminated string */
555 /* Functions to evaluate an arithmetic expression */
558 eval_expr_val (char **expr
, long int *result
)
562 /* Skip white space */
563 for (digit
= *expr
; digit
&& *digit
&& isspace (*digit
); ++digit
);
567 /* Scan for closing paren */
568 for (++digit
; **expr
&& **expr
!= ')'; ++(*expr
));
576 if (eval_expr (digit
, result
))
582 /* POSIX requires that decimal, octal, and hexadecimal constants are
583 recognized. Therefore we pass 0 as the third parameter to strtol. */
584 *result
= strtol (digit
, expr
, 0);
593 eval_expr_multdiv (char **expr
, long int *result
)
598 if (eval_expr_val (expr
, result
) != 0)
603 /* Skip white space */
604 for (; *expr
&& **expr
&& isspace (**expr
); ++(*expr
));
609 if (eval_expr_val (expr
, &arg
) != 0)
614 else if (**expr
== '/')
617 if (eval_expr_val (expr
, &arg
) != 0)
630 eval_expr (char *expr
, long int *result
)
635 if (eval_expr_multdiv (&expr
, result
) != 0)
640 /* Skip white space */
641 for (; expr
&& *expr
&& isspace (*expr
); ++expr
);
646 if (eval_expr_multdiv (&expr
, &arg
) != 0)
651 else if (*expr
== '-')
654 if (eval_expr_multdiv (&expr
, &arg
) != 0)
667 parse_arith (char **word
, size_t *word_length
, size_t *max_length
,
668 const char *words
, size_t *offset
, int flags
, int bracket
)
670 /* We are poised just after "$((" or "$[" */
677 expr
= w_newword (&expr_length
, &expr_maxlen
);
678 for (; words
[*offset
]; ++(*offset
))
680 switch (words
[*offset
])
683 error
= parse_dollars (&expr
, &expr_length
, &expr_maxlen
,
684 words
, offset
, flags
, NULL
, NULL
, NULL
, 1);
685 /* The ``1'' here is to tell parse_dollars not to
697 error
= parse_backtick (&expr
, &expr_length
, &expr_maxlen
,
698 words
, offset
, flags
, NULL
, NULL
, NULL
);
699 /* The first NULL here is to tell parse_backtick not to
710 error
= parse_qtd_backslash (&expr
, &expr_length
, &expr_maxlen
,
717 /* I think that a backslash within an
718 * arithmetic expansion is bound to
719 * cause an error sooner or later anyway though.
724 if (--paren_depth
== 0)
726 char result
[21]; /* 21 = ceil(log10(2^64)) + 1 */
727 long int numresult
= 0;
728 long long int convertme
;
730 if (bracket
|| words
[1 + *offset
] != ')')
739 if (*expr
&& eval_expr (expr
, &numresult
) != 0)
747 convertme
= -numresult
;
748 *word
= w_addchar (*word
, word_length
, max_length
, '-');
756 convertme
= numresult
;
759 *word
= w_addstr (*word
, word_length
, max_length
,
760 _itoa (convertme
, &result
[20], 10, 0));
762 return *word
? 0 : WRDE_NOSPACE
;
764 expr
= w_addchar (expr
, &expr_length
, &expr_maxlen
, words
[*offset
]);
771 if (bracket
&& paren_depth
== 1)
773 char result
[21]; /* 21 = ceil(log10(2^64)) + 1 */
774 long int numresult
= 0;
777 if (*expr
&& eval_expr (expr
, &numresult
) != 0)
784 *word
= w_addstr (*word
, word_length
, max_length
,
785 _itoa_word (numresult
, &result
[20], 10, 0));
787 return *word
? 0 : WRDE_NOSPACE
;
803 expr
= w_addchar (expr
, &expr_length
, &expr_maxlen
, words
[*offset
]);
814 /* Function called by child process in exec_comm() */
816 internal_function
__attribute__ ((always_inline
))
817 exec_comm_child (char *comm
, int *fildes
, int showerr
, int noexec
)
819 const char *args
[4] = { _PATH_BSHELL
, "-c", comm
, NULL
};
821 /* Execute the command, or just check syntax? */
825 /* Redirect output. */
826 if (__glibc_likely (fildes
[1] != STDOUT_FILENO
))
828 __dup2 (fildes
[1], STDOUT_FILENO
);
834 /* Reset the close-on-exec flag (if necessary). */
835 # ifndef __ASSUME_PIPE2
836 if (__have_pipe2
> 0)
838 __fcntl (fildes
[1], F_SETFD
, 0);
842 /* Redirect stderr to /dev/null if we have to. */
847 __close (STDERR_FILENO
);
848 fd
= __open (_PATH_DEVNULL
, O_WRONLY
);
849 if (fd
>= 0 && fd
!= STDERR_FILENO
)
851 __dup2 (fd
, STDERR_FILENO
);
854 /* Be paranoid. Check that we actually opened the /dev/null
856 if (__builtin_expect (__fxstat64 (_STAT_VER
, STDERR_FILENO
, &st
), 0) != 0
857 || __builtin_expect (S_ISCHR (st
.st_mode
), 1) == 0
858 #if defined DEV_NULL_MAJOR && defined DEV_NULL_MINOR
859 || st
.st_rdev
!= makedev (DEV_NULL_MAJOR
, DEV_NULL_MINOR
)
862 /* It's not the /dev/null device. Stop right here. The
863 problem is: how do we stop? We use _exit() with an
864 hopefully unusual exit code. */
868 /* Make sure the subshell doesn't field-split on our behalf. */
872 __execve (_PATH_BSHELL
, (char *const *) args
, __environ
);
878 /* Function to execute a command and retrieve the results */
879 /* pwordexp contains NULL if field-splitting is forbidden */
882 exec_comm (char *comm
, char **word
, size_t *word_length
, size_t *max_length
,
883 int flags
, wordexp_t
*pwordexp
, const char *ifs
,
884 const char *ifs_white
)
891 size_t maxnewlines
= 0;
892 char buffer
[bufsize
];
896 /* Don't fork() unless necessary */
901 # ifndef __ASSUME_PIPE2
902 if (__have_pipe2
>= 0)
905 int r
= __pipe2 (fildes
, O_CLOEXEC
);
906 # ifndef __ASSUME_PIPE2
907 if (__have_pipe2
== 0)
908 __have_pipe2
= r
!= -1 || errno
!= ENOSYS
? 1 : -1;
910 if (__have_pipe2
> 0)
917 #ifndef __ASSUME_PIPE2
919 if (__have_pipe2
< 0)
921 if (__pipe (fildes
) < 0)
927 if ((pid
= __fork ()) < 0)
936 exec_comm_child (comm
, fildes
, noexec
? 0 : flags
& WRDE_SHOWERR
, noexec
);
940 /* If we are just testing the syntax, only wait. */
942 return (TEMP_FAILURE_RETRY (__waitpid (pid
, &status
, 0)) == pid
943 && status
!= 0) ? WRDE_SYNTAX
: 0;
949 /* Quoted - no field splitting */
953 if ((buflen
= TEMP_FAILURE_RETRY (__read (fildes
[0], buffer
,
956 /* If read returned 0 then the process has closed its
957 stdout. Don't use WNOHANG in that case to avoid busy
958 looping until the process eventually exits. */
959 if (TEMP_FAILURE_RETRY (__waitpid (pid
, &status
,
960 buflen
== 0 ? 0 : WNOHANG
))
963 if ((buflen
= TEMP_FAILURE_RETRY (__read (fildes
[0], buffer
,
968 maxnewlines
+= buflen
;
970 *word
= w_addmem (*word
, word_length
, max_length
, buffer
, buflen
);
976 /* Not quoted - split fields */
980 * 0 when searching for first character in a field not IFS white space
981 * 1 when copying the text of a field
982 * 2 when searching for possible non-whitespace IFS
983 * 3 when searching for non-newline after copying field
988 if ((buflen
= TEMP_FAILURE_RETRY (__read (fildes
[0], buffer
,
991 /* If read returned 0 then the process has closed its
992 stdout. Don't use WNOHANG in that case to avoid busy
993 looping until the process eventually exits. */
994 if (TEMP_FAILURE_RETRY (__waitpid (pid
, &status
,
995 buflen
== 0 ? 0 : WNOHANG
))
998 if ((buflen
= TEMP_FAILURE_RETRY (__read (fildes
[0], buffer
,
1003 for (i
= 0; i
< buflen
; ++i
)
1005 if (strchr (ifs
, buffer
[i
]) != NULL
)
1007 /* Current character is IFS */
1008 if (strchr (ifs_white
, buffer
[i
]) == NULL
)
1010 /* Current character is IFS but not whitespace */
1013 /* current character
1016 * eg: text<space><comma><space>moretext
1018 * So, strip whitespace IFS (like at the start)
1025 /* fall through and delimit field.. */
1029 if (buffer
[i
] == '\n')
1031 /* Current character is (IFS) newline */
1033 /* If copying a field, this is the end of it,
1034 but maybe all that's left is trailing newlines.
1035 So start searching for a non-newline. */
1043 /* Current character is IFS white space, but
1046 /* If not either copying a field or searching
1047 for non-newline after a field, ignore it */
1048 if (copying
!= 1 && copying
!= 3)
1051 /* End of field (search for non-ws IFS afterwards) */
1056 /* First IFS white space (non-newline), or IFS non-whitespace.
1057 * Delimit the field. Nulls are converted by w_addword. */
1058 if (w_addword (pwordexp
, *word
) == WRDE_NOSPACE
)
1061 *word
= w_newword (word_length
, max_length
);
1064 /* fall back round the loop.. */
1068 /* Not IFS character */
1072 /* Nothing but (IFS) newlines since the last field,
1073 so delimit it here before starting new word */
1074 if (w_addword (pwordexp
, *word
) == WRDE_NOSPACE
)
1077 *word
= w_newword (word_length
, max_length
);
1082 if (buffer
[i
] == '\n') /* happens if newline not in IFS */
1087 *word
= w_addchar (*word
, word_length
, max_length
,
1096 /* Chop off trailing newlines (required by POSIX.2) */
1097 /* Ensure we don't go back further than the beginning of the
1098 substitution (i.e. remove maxnewlines bytes at most) */
1099 while (maxnewlines
-- != 0 &&
1100 *word_length
> 0 && (*word
)[*word_length
- 1] == '\n')
1102 (*word
)[--*word_length
] = '\0';
1104 /* If the last word was entirely newlines, turn it into a new word
1105 * which can be ignored if there's nothing following it. */
1106 if (*word_length
== 0)
1109 *word
= w_newword (word_length
, max_length
);
1114 __close (fildes
[0]);
1117 /* Check for syntax error (re-execute but with "-n" flag) */
1118 if (buflen
< 1 && status
!= 0)
1127 __kill (pid
, SIGKILL
);
1128 TEMP_FAILURE_RETRY (__waitpid (pid
, NULL
, 0));
1129 __close (fildes
[0]);
1130 return WRDE_NOSPACE
;
1135 parse_comm (char **word
, size_t *word_length
, size_t *max_length
,
1136 const char *words
, size_t *offset
, int flags
, wordexp_t
*pwordexp
,
1137 const char *ifs
, const char *ifs_white
)
1139 /* We are poised just after "$(" */
1140 int paren_depth
= 1;
1142 int quoted
= 0; /* 1 for singly-quoted, 2 for doubly-quoted */
1145 char *comm
= w_newword (&comm_length
, &comm_maxlen
);
1147 for (; words
[*offset
]; ++(*offset
))
1149 switch (words
[*offset
])
1154 else if (quoted
== 1)
1162 else if (quoted
== 2)
1168 if (!quoted
&& --paren_depth
== 0)
1170 /* Go -- give script to the shell */
1173 #ifdef __libc_ptf_call
1174 /* We do not want the exec_comm call to be cut short
1175 by a thread cancellation since cleanup is very
1176 ugly. Therefore disable cancellation for
1178 // XXX Ideally we do want the thread being cancelable.
1179 // XXX If demand is there we'll change it.
1180 int state
= PTHREAD_CANCEL_ENABLE
;
1181 __libc_ptf_call (pthread_setcancelstate
,
1182 (PTHREAD_CANCEL_DISABLE
, &state
), 0);
1185 error
= exec_comm (comm
, word
, word_length
, max_length
,
1186 flags
, pwordexp
, ifs
, ifs_white
);
1188 #ifdef __libc_ptf_call
1189 __libc_ptf_call (pthread_setcancelstate
, (state
, NULL
), 0);
1198 /* This is just part of the script */
1206 comm
= w_addchar (comm
, &comm_length
, &comm_maxlen
, words
[*offset
]);
1208 return WRDE_NOSPACE
;
1211 /* Premature end. */
1219 parse_param (char **word
, size_t *word_length
, size_t *max_length
,
1220 const char *words
, size_t *offset
, int flags
, wordexp_t
*pwordexp
,
1221 const char *ifs
, const char *ifs_white
, int quoted
)
1223 /* We are poised just after "$" */
1227 ACT_RP_SHORT_LEFT
= '#',
1228 ACT_RP_LONG_LEFT
= 'L',
1229 ACT_RP_SHORT_RIGHT
= '%',
1230 ACT_RP_LONG_RIGHT
= 'R',
1231 ACT_NULL_ERROR
= '?',
1232 ACT_NULL_SUBST
= '-',
1233 ACT_NONNULL_SUBST
= '+',
1234 ACT_NULL_ASSIGN
= '='
1240 size_t start
= *offset
;
1244 enum action action
= ACT_NONE
;
1249 int pattern_is_quoted
= 0; /* 1 for singly-quoted, 2 for doubly-quoted */
1253 int brace
= words
[*offset
] == '{';
1255 env
= w_newword (&env_length
, &env_maxlen
);
1256 pattern
= w_newword (&pat_length
, &pat_maxlen
);
1261 /* First collect the parameter name. */
1263 if (words
[*offset
] == '#')
1271 if (isalpha (words
[*offset
]) || words
[*offset
] == '_')
1273 /* Normal parameter name. */
1276 env
= w_addchar (env
, &env_length
, &env_maxlen
,
1281 while (isalnum (words
[++*offset
]) || words
[*offset
] == '_');
1283 else if (isdigit (words
[*offset
]))
1285 /* Numeric parameter name. */
1289 env
= w_addchar (env
, &env_length
, &env_maxlen
,
1296 while (isdigit(words
[++*offset
]));
1298 else if (strchr ("*@$", words
[*offset
]) != NULL
)
1300 /* Special parameter. */
1302 env
= w_addchar (env
, &env_length
, &env_maxlen
,
1316 /* Check for special action to be applied to the value. */
1317 switch (words
[*offset
])
1324 action
= ACT_RP_SHORT_LEFT
;
1325 if (words
[1 + *offset
] == '#')
1328 action
= ACT_RP_LONG_LEFT
;
1333 action
= ACT_RP_SHORT_RIGHT
;
1334 if (words
[1 + *offset
] == '%')
1337 action
= ACT_RP_LONG_RIGHT
;
1342 if (strchr ("-=?+", words
[1 + *offset
]) == NULL
)
1346 action
= words
[++*offset
];
1353 action
= words
[*offset
];
1360 /* Now collect the pattern, but don't expand it yet. */
1362 for (; words
[*offset
]; ++(*offset
))
1364 switch (words
[*offset
])
1367 if (!pattern_is_quoted
)
1372 if (!pattern_is_quoted
)
1381 if (pattern_is_quoted
)
1382 /* Quoted; treat as normal character. */
1385 /* Otherwise, it's an escape: next character is literal. */
1386 if (words
[++*offset
] == '\0')
1389 pattern
= w_addchar (pattern
, &pat_length
, &pat_maxlen
, '\\');
1390 if (pattern
== NULL
)
1396 if (pattern_is_quoted
== 0)
1397 pattern_is_quoted
= 1;
1398 else if (pattern_is_quoted
== 1)
1399 pattern_is_quoted
= 0;
1404 if (pattern_is_quoted
== 0)
1405 pattern_is_quoted
= 2;
1406 else if (pattern_is_quoted
== 2)
1407 pattern_is_quoted
= 0;
1412 pattern
= w_addchar (pattern
, &pat_length
, &pat_maxlen
,
1414 if (pattern
== NULL
)
1419 /* End of input string -- remember to reparse the character that we
1424 if (words
[start
] == '{' && words
[*offset
] != '}')
1431 /* $# expands to the number of positional parameters */
1433 value
= _itoa_word (__libc_argc
- 1, &buffer
[20], 10, 0);
1438 /* Just $ on its own */
1439 *offset
= start
- 1;
1440 *word
= w_addchar (*word
, word_length
, max_length
, '$');
1441 return *word
? 0 : WRDE_NOSPACE
;
1444 /* Is it a numeric parameter? */
1445 else if (isdigit (env
[0]))
1449 if (n
>= __libc_argc
)
1450 /* Substitute NULL. */
1453 /* Replace with appropriate positional parameter. */
1454 value
= __libc_argv
[n
];
1456 /* Is it a special parameter? */
1463 value
= _itoa_word (__getpid (), &buffer
[20], 10, 0);
1465 /* Is it `${#*}' or `${#@}'? */
1466 else if ((*env
== '*' || *env
== '@') && seen_hash
)
1469 value
= _itoa_word (__libc_argc
> 0 ? __libc_argc
- 1 : 0,
1470 &buffer
[20], 10, 0);
1471 *word
= w_addstr (*word
, word_length
, max_length
, value
);
1474 return *word
? 0 : WRDE_NOSPACE
;
1476 /* Is it `$*' or `$@' (unquoted) ? */
1477 else if (*env
== '*' || (*env
== '@' && !quoted
))
1479 size_t plist_len
= 0;
1483 /* Build up value parameter by parameter (copy them) */
1484 for (p
= 1; __libc_argv
[p
]; ++p
)
1485 plist_len
+= strlen (__libc_argv
[p
]) + 1; /* for space */
1486 value
= malloc (plist_len
);
1491 for (p
= 1; __libc_argv
[p
]; ++p
)
1495 end
= __stpcpy (end
, __libc_argv
[p
]);
1502 /* Must be a quoted `$@' */
1503 assert (*env
== '@' && quoted
);
1505 /* Each parameter is a separate word ("$@") */
1506 if (__libc_argc
== 2)
1507 value
= __libc_argv
[1];
1508 else if (__libc_argc
> 2)
1512 /* Append first parameter to current word. */
1513 value
= w_addstr (*word
, word_length
, max_length
,
1515 if (value
== NULL
|| w_addword (pwordexp
, value
))
1518 for (p
= 2; __libc_argv
[p
+ 1]; p
++)
1520 char *newword
= __strdup (__libc_argv
[p
]);
1521 if (newword
== NULL
|| w_addword (pwordexp
, newword
))
1525 /* Start a new word with the last parameter. */
1526 *word
= w_newword (word_length
, max_length
);
1527 value
= __libc_argv
[p
];
1538 value
= getenv (env
);
1540 if (value
== NULL
&& (flags
& WRDE_UNDEF
))
1542 /* Variable not defined. */
1543 error
= WRDE_BADVAL
;
1547 if (action
!= ACT_NONE
)
1549 int expand_pattern
= 0;
1551 /* First, find out if we need to expand pattern (i.e. if we will
1555 case ACT_RP_SHORT_LEFT
:
1556 case ACT_RP_LONG_LEFT
:
1557 case ACT_RP_SHORT_RIGHT
:
1558 case ACT_RP_LONG_RIGHT
:
1559 /* Always expand for these. */
1563 case ACT_NULL_ERROR
:
1564 case ACT_NULL_SUBST
:
1565 case ACT_NULL_ASSIGN
:
1566 if (!value
|| (!*value
&& colon_seen
))
1567 /* If param is unset, or set but null and a colon has been seen,
1568 the expansion of the pattern will be needed. */
1573 case ACT_NONNULL_SUBST
:
1574 /* Expansion of word will be needed if parameter is set and not null,
1575 or set null but no colon has been seen. */
1576 if (value
&& (*value
|| !colon_seen
))
1582 assert (! "Unrecognised action!");
1587 /* We need to perform tilde expansion, parameter expansion,
1588 command substitution, and arithmetic expansion. We also
1589 have to be a bit careful with wildcard characters, as
1590 pattern might be given to fnmatch soon. To do this, we
1591 convert quotes to escapes. */
1597 int quoted
= 0; /* 1: single quotes; 2: double */
1599 expanded
= w_newword (&exp_len
, &exp_maxl
);
1600 for (p
= pattern
; p
&& *p
; p
++)
1609 else if (quoted
== 0)
1618 else if (quoted
== 0)
1628 /* Convert quoted wildchar to escaped wildchar. */
1629 expanded
= w_addchar (expanded
, &exp_len
,
1632 if (expanded
== NULL
)
1639 error
= parse_dollars (&expanded
, &exp_len
, &exp_maxl
, p
,
1640 &offset
, flags
, NULL
, NULL
, NULL
, 1);
1655 if (quoted
|| exp_len
)
1659 error
= parse_tilde (&expanded
, &exp_len
, &exp_maxl
, p
,
1675 expanded
= w_addchar (expanded
, &exp_len
, &exp_maxl
, '\\');
1677 assert (*p
); /* checked when extracted initially */
1678 if (expanded
== NULL
)
1682 expanded
= w_addchar (expanded
, &exp_len
, &exp_maxl
, *p
);
1684 if (expanded
== NULL
)
1695 case ACT_RP_SHORT_LEFT
:
1696 case ACT_RP_LONG_LEFT
:
1697 case ACT_RP_SHORT_RIGHT
:
1698 case ACT_RP_LONG_RIGHT
:
1704 if (value
== NULL
|| pattern
== NULL
|| *pattern
== '\0')
1707 end
= value
+ strlen (value
);
1711 case ACT_RP_SHORT_LEFT
:
1712 for (p
= value
; p
<= end
; ++p
)
1716 if (fnmatch (pattern
, value
, 0) != FNM_NOMATCH
)
1721 char *newval
= __strdup (p
);
1739 case ACT_RP_LONG_LEFT
:
1740 for (p
= end
; p
>= value
; --p
)
1744 if (fnmatch (pattern
, value
, 0) != FNM_NOMATCH
)
1749 char *newval
= __strdup (p
);
1767 case ACT_RP_SHORT_RIGHT
:
1768 for (p
= end
; p
>= value
; --p
)
1770 if (fnmatch (pattern
, p
, 0) != FNM_NOMATCH
)
1773 newval
= malloc (p
- value
+ 1);
1782 *(char *) __mempcpy (newval
, value
, p
- value
) = '\0';
1793 case ACT_RP_LONG_RIGHT
:
1794 for (p
= value
; p
<= end
; ++p
)
1796 if (fnmatch (pattern
, p
, 0) != FNM_NOMATCH
)
1799 newval
= malloc (p
- value
+ 1);
1808 *(char *) __mempcpy (newval
, value
, p
- value
) = '\0';
1826 case ACT_NULL_ERROR
:
1827 if (value
&& *value
)
1828 /* Substitute parameter */
1832 if (!colon_seen
&& value
)
1833 /* Substitute NULL */
1837 const char *str
= pattern
;
1840 str
= _("parameter null or not set");
1842 __fxprintf (NULL
, "%s: %s\n", env
, str
);
1849 case ACT_NULL_SUBST
:
1850 if (value
&& *value
)
1851 /* Substitute parameter */
1857 if (!colon_seen
&& value
)
1858 /* Substitute NULL */
1861 value
= pattern
? __strdup (pattern
) : pattern
;
1864 if (pattern
&& !value
)
1869 case ACT_NONNULL_SUBST
:
1870 if (value
&& (*value
|| !colon_seen
))
1875 value
= pattern
? __strdup (pattern
) : pattern
;
1878 if (pattern
&& !value
)
1884 /* Substitute NULL */
1889 case ACT_NULL_ASSIGN
:
1890 if (value
&& *value
)
1891 /* Substitute parameter */
1894 if (!colon_seen
&& value
)
1896 /* Substitute NULL */
1905 value
= pattern
? __strdup (pattern
) : pattern
;
1908 if (pattern
&& !value
)
1911 __setenv (env
, value
, 1);
1915 assert (! "Unrecognised action!");
1926 char param_length
[21];
1927 param_length
[20] = '\0';
1928 *word
= w_addstr (*word
, word_length
, max_length
,
1929 _itoa_word (value
? strlen (value
) : 0,
1930 ¶m_length
[20], 10, 0));
1933 assert (value
!= NULL
);
1937 return *word
? 0 : WRDE_NOSPACE
;
1943 if (quoted
|| !pwordexp
)
1945 /* Quoted - no field split */
1946 *word
= w_addstr (*word
, word_length
, max_length
, value
);
1950 return *word
? 0 : WRDE_NOSPACE
;
1954 /* Need to field-split */
1955 char *value_copy
= __strdup (value
); /* Don't modify value */
1956 char *field_begin
= value_copy
;
1957 int seen_nonws_ifs
= 0;
1962 if (value_copy
== NULL
)
1967 char *field_end
= field_begin
;
1970 /* If this isn't the first field, start a new word */
1971 if (field_begin
!= value_copy
)
1973 if (w_addword (pwordexp
, *word
) == WRDE_NOSPACE
)
1979 *word
= w_newword (word_length
, max_length
);
1982 /* Skip IFS whitespace before the field */
1983 field_begin
+= strspn (field_begin
, ifs_white
);
1985 if (!seen_nonws_ifs
&& *field_begin
== 0)
1986 /* Nothing but whitespace */
1989 /* Search for the end of the field */
1990 field_end
= field_begin
+ strcspn (field_begin
, ifs
);
1992 /* Set up pointer to the character after end of field and
1993 skip whitespace IFS after it. */
1994 next_field
= field_end
+ strspn (field_end
, ifs_white
);
1996 /* Skip at most one non-whitespace IFS character after the field */
1998 if (*next_field
&& strchr (ifs
, *next_field
))
2004 /* Null-terminate it */
2007 /* Tag a copy onto the current word */
2008 *word
= w_addstr (*word
, word_length
, max_length
, field_begin
);
2010 if (*word
== NULL
&& *field_begin
!= '\0')
2016 field_begin
= next_field
;
2018 while (seen_nonws_ifs
|| *field_begin
);
2030 error
= WRDE_NOSPACE
;
2034 error
= WRDE_SYNTAX
;
2046 parse_dollars (char **word
, size_t *word_length
, size_t *max_length
,
2047 const char *words
, size_t *offset
, int flags
,
2048 wordexp_t
*pwordexp
, const char *ifs
, const char *ifs_white
,
2051 /* We are poised _at_ "$" */
2052 switch (words
[1 + *offset
])
2057 *word
= w_addchar (*word
, word_length
, max_length
, '$');
2058 return *word
? 0 : WRDE_NOSPACE
;
2061 if (words
[2 + *offset
] == '(')
2063 /* Differentiate between $((1+3)) and $((echo);(ls)) */
2064 int i
= 3 + *offset
;
2066 while (words
[i
] && !(depth
== 0 && words
[i
] == ')'))
2068 if (words
[i
] == '(')
2070 else if (words
[i
] == ')')
2076 if (words
[i
] == ')' && words
[i
+ 1] == ')')
2079 /* Call parse_arith -- 0 is for "no brackets" */
2080 return parse_arith (word
, word_length
, max_length
, words
, offset
,
2085 if (flags
& WRDE_NOCMD
)
2089 return parse_comm (word
, word_length
, max_length
, words
, offset
, flags
,
2090 quoted
? NULL
: pwordexp
, ifs
, ifs_white
);
2094 /* Call parse_arith -- 1 is for "brackets" */
2095 return parse_arith (word
, word_length
, max_length
, words
, offset
, flags
,
2100 ++(*offset
); /* parse_param needs to know if "{" is there */
2101 return parse_param (word
, word_length
, max_length
, words
, offset
, flags
,
2102 pwordexp
, ifs
, ifs_white
, quoted
);
2108 parse_backtick (char **word
, size_t *word_length
, size_t *max_length
,
2109 const char *words
, size_t *offset
, int flags
,
2110 wordexp_t
*pwordexp
, const char *ifs
, const char *ifs_white
)
2112 /* We are poised just after "`" */
2117 char *comm
= w_newword (&comm_length
, &comm_maxlen
);
2119 for (; words
[*offset
]; ++(*offset
))
2121 switch (words
[*offset
])
2124 /* Go -- give the script to the shell */
2125 error
= exec_comm (comm
, word
, word_length
, max_length
, flags
,
2126 pwordexp
, ifs
, ifs_white
);
2133 error
= parse_qtd_backslash (&comm
, &comm_length
, &comm_maxlen
,
2146 error
= parse_backslash (&comm
, &comm_length
, &comm_maxlen
, words
,
2158 squoting
= 1 - squoting
;
2160 comm
= w_addchar (comm
, &comm_length
, &comm_maxlen
, words
[*offset
]);
2162 return WRDE_NOSPACE
;
2173 parse_dquote (char **word
, size_t *word_length
, size_t *max_length
,
2174 const char *words
, size_t *offset
, int flags
,
2175 wordexp_t
*pwordexp
, const char * ifs
, const char * ifs_white
)
2177 /* We are poised just after a double-quote */
2180 for (; words
[*offset
]; ++(*offset
))
2182 switch (words
[*offset
])
2188 error
= parse_dollars (word
, word_length
, max_length
, words
, offset
,
2189 flags
, pwordexp
, ifs
, ifs_white
, 1);
2190 /* The ``1'' here is to tell parse_dollars not to
2191 * split the fields. It may need to, however ("$@").
2199 if (flags
& WRDE_NOCMD
)
2203 error
= parse_backtick (word
, word_length
, max_length
, words
,
2204 offset
, flags
, NULL
, NULL
, NULL
);
2205 /* The first NULL here is to tell parse_backtick not to
2214 error
= parse_qtd_backslash (word
, word_length
, max_length
, words
,
2223 *word
= w_addchar (*word
, word_length
, max_length
, words
[*offset
]);
2225 return WRDE_NOSPACE
;
2229 /* Unterminated string */
2234 * wordfree() is to be called after pwordexp is finished with.
2238 wordfree (wordexp_t
*pwordexp
)
2241 /* wordexp can set pwordexp to NULL */
2242 if (pwordexp
&& pwordexp
->we_wordv
)
2244 char **wordv
= pwordexp
->we_wordv
;
2246 for (wordv
+= pwordexp
->we_offs
; *wordv
; ++wordv
)
2249 free (pwordexp
->we_wordv
);
2250 pwordexp
->we_wordv
= NULL
;
2253 libc_hidden_def (wordfree
)
2260 wordexp (const char *words
, wordexp_t
*pwordexp
, int flags
)
2262 size_t words_offset
;
2265 char *word
= w_newword (&word_length
, &max_length
);
2269 wordexp_t old_word
= *pwordexp
;
2271 if (flags
& WRDE_REUSE
)
2273 /* Minimal implementation of WRDE_REUSE for now */
2274 wordfree (pwordexp
);
2275 old_word
.we_wordv
= NULL
;
2278 if ((flags
& WRDE_APPEND
) == 0)
2280 pwordexp
->we_wordc
= 0;
2282 if (flags
& WRDE_DOOFFS
)
2284 pwordexp
->we_wordv
= calloc (1 + pwordexp
->we_offs
, sizeof (char *));
2285 if (pwordexp
->we_wordv
== NULL
)
2287 error
= WRDE_NOSPACE
;
2293 pwordexp
->we_wordv
= calloc (1, sizeof (char *));
2294 if (pwordexp
->we_wordv
== NULL
)
2296 error
= WRDE_NOSPACE
;
2300 pwordexp
->we_offs
= 0;
2304 /* Find out what the field separators are.
2305 * There are two types: whitespace and non-whitespace.
2307 ifs
= getenv ("IFS");
2310 /* IFS unset - use <space><tab><newline>. */
2311 ifs
= strcpy (ifs_white
, " \t\n");
2315 char *whch
= ifs_white
;
2317 while (*ifsch
!= '\0')
2319 if (*ifsch
== ' ' || *ifsch
== '\t' || *ifsch
== '\n')
2321 /* Whitespace IFS. See first whether it is already in our
2323 char *runp
= ifs_white
;
2325 while (runp
< whch
&& *runp
!= *ifsch
)
2337 for (words_offset
= 0 ; words
[words_offset
] ; ++words_offset
)
2338 switch (words
[words_offset
])
2341 error
= parse_backslash (&word
, &word_length
, &max_length
, words
,
2350 error
= parse_dollars (&word
, &word_length
, &max_length
, words
,
2351 &words_offset
, flags
, pwordexp
, ifs
, ifs_white
,
2360 if (flags
& WRDE_NOCMD
)
2362 error
= WRDE_CMDSUB
;
2367 error
= parse_backtick (&word
, &word_length
, &max_length
, words
,
2368 &words_offset
, flags
, pwordexp
, ifs
,
2378 error
= parse_dquote (&word
, &word_length
, &max_length
, words
,
2379 &words_offset
, flags
, pwordexp
, ifs
, ifs_white
);
2386 error
= w_addword (pwordexp
, NULL
);
2396 error
= parse_squote (&word
, &word_length
, &max_length
, words
,
2404 error
= w_addword (pwordexp
, NULL
);
2413 error
= parse_tilde (&word
, &word_length
, &max_length
, words
,
2414 &words_offset
, pwordexp
->we_wordc
);
2424 error
= parse_glob (&word
, &word_length
, &max_length
, words
,
2425 &words_offset
, flags
, pwordexp
, ifs
, ifs_white
);
2433 /* Is it a word separator? */
2434 if (strchr (" \t", words
[words_offset
]) == NULL
)
2436 char ch
= words
[words_offset
];
2438 /* Not a word separator -- but is it a valid word char? */
2439 if (strchr ("\n|&;<>(){}", ch
))
2442 error
= WRDE_BADCHAR
;
2446 /* "Ordinary" character -- add it to word */
2447 word
= w_addchar (word
, &word_length
, &max_length
,
2451 error
= WRDE_NOSPACE
;
2458 /* If a word has been delimited, add it to the list. */
2461 error
= w_addword (pwordexp
, word
);
2466 word
= w_newword (&word_length
, &max_length
);
2471 /* There was a word separator at the end */
2472 if (word
== NULL
) /* i.e. w_newword */
2475 /* There was no field separator at the end */
2476 return w_addword (pwordexp
, word
);
2480 * free memory used (unless error is WRDE_NOSPACE), and
2481 * set pwordexp members back to what they were.
2486 if (error
== WRDE_NOSPACE
)
2487 return WRDE_NOSPACE
;
2489 if ((flags
& WRDE_APPEND
) == 0)
2490 wordfree (pwordexp
);
2492 *pwordexp
= old_word
;