1 /* POSIX.2 wordexp implementation.
2 Copyright (C) 1997-2019 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/>. */
33 #include <sys/param.h>
36 #include <sys/types.h>
37 #include <sys/types.h>
42 #include <kernel-features.h>
43 #include <scratch_buffer.h>
45 #include <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
);
68 static int parse_backtick (char **word
, size_t *word_length
,
69 size_t *max_length
, const char *words
,
70 size_t *offset
, int flags
, wordexp_t
*pwordexp
,
71 const char *ifs
, const char *ifs_white
);
72 static int parse_dquote (char **word
, size_t *word_length
, size_t *max_length
,
73 const char *words
, size_t *offset
, int flags
,
74 wordexp_t
*pwordexp
, const char *ifs
,
75 const char *ifs_white
);
76 static int eval_expr (char *expr
, long int *result
);
78 /* The w_*() functions manipulate word lists. */
82 /* Result of w_newword will be ignored if it's the last word. */
84 w_newword (size_t *actlen
, size_t *maxlen
)
86 *actlen
= *maxlen
= 0;
91 w_addchar (char *buffer
, size_t *actlen
, size_t *maxlen
, char ch
)
92 /* (lengths exclude trailing zero) */
94 /* Add a character to the buffer, allocating room for it if needed. */
96 if (*actlen
== *maxlen
)
98 char *old_buffer
= buffer
;
99 assert (buffer
== NULL
|| *maxlen
!= 0);
101 buffer
= (char *) realloc (buffer
, 1 + *maxlen
);
109 buffer
[*actlen
] = ch
;
110 buffer
[++(*actlen
)] = '\0';
117 w_addmem (char *buffer
, size_t *actlen
, size_t *maxlen
, const char *str
,
120 /* Add a string to the buffer, allocating room for it if needed.
122 if (*actlen
+ len
> *maxlen
)
124 char *old_buffer
= buffer
;
125 assert (buffer
== NULL
|| *maxlen
!= 0);
126 *maxlen
+= MAX (2 * len
, W_CHUNK
);
127 buffer
= realloc (old_buffer
, 1 + *maxlen
);
135 *((char *) __mempcpy (&buffer
[*actlen
], str
, len
)) = '\0';
143 w_addstr (char *buffer
, size_t *actlen
, size_t *maxlen
, const char *str
)
144 /* (lengths exclude trailing zero) */
146 /* Add a string to the buffer, allocating room for it if needed.
150 assert (str
!= NULL
); /* w_addstr only called from this file */
153 return w_addmem (buffer
, actlen
, maxlen
, str
, len
);
157 w_addword (wordexp_t
*pwordexp
, char *word
)
159 /* Add a word to the wordlist */
162 bool allocated
= false;
164 /* Internally, NULL acts like "". Convert NULLs to "" before
165 * the caller sees them.
169 word
= __strdup ("");
175 num_p
= 2 + pwordexp
->we_wordc
+ pwordexp
->we_offs
;
176 new_wordv
= realloc (pwordexp
->we_wordv
, sizeof (char *) * num_p
);
177 if (new_wordv
!= NULL
)
179 pwordexp
->we_wordv
= new_wordv
;
180 pwordexp
->we_wordv
[pwordexp
->we_offs
+ pwordexp
->we_wordc
++] = word
;
181 pwordexp
->we_wordv
[pwordexp
->we_offs
+ pwordexp
->we_wordc
] = NULL
;
192 /* The parse_*() functions should leave *offset being the offset in 'words'
193 * to the last character processed.
197 parse_backslash (char **word
, size_t *word_length
, size_t *max_length
,
198 const char *words
, size_t *offset
)
200 /* We are poised _at_ a backslash, not in quotes */
202 switch (words
[1 + *offset
])
205 /* Backslash is last character of input words */
213 *word
= w_addchar (*word
, word_length
, max_length
, words
[1 + *offset
]);
225 parse_qtd_backslash (char **word
, size_t *word_length
, size_t *max_length
,
226 const char *words
, size_t *offset
)
228 /* We are poised _at_ a backslash, inside quotes */
230 switch (words
[1 + *offset
])
233 /* Backslash is last character of input words */
244 *word
= w_addchar (*word
, word_length
, max_length
, words
[1 + *offset
]);
252 *word
= w_addchar (*word
, word_length
, max_length
, words
[*offset
]);
254 *word
= w_addchar (*word
, word_length
, max_length
, words
[1 + *offset
]);
267 parse_tilde (char **word
, size_t *word_length
, size_t *max_length
,
268 const char *words
, size_t *offset
, size_t wordc
)
270 /* We are poised _at_ a tilde */
273 if (*word_length
!= 0)
275 if (!((*word
)[*word_length
- 1] == '=' && wordc
== 0))
277 if (!((*word
)[*word_length
- 1] == ':'
278 && strchr (*word
, '=') && wordc
== 0))
280 *word
= w_addchar (*word
, word_length
, max_length
, '~');
281 return *word
? 0 : WRDE_NOSPACE
;
286 for (i
= 1 + *offset
; words
[i
]; i
++)
288 if (words
[i
] == ':' || words
[i
] == '/' || words
[i
] == ' '
289 || words
[i
] == '\t' || words
[i
] == 0 )
292 if (words
[i
] == '\\')
294 *word
= w_addchar (*word
, word_length
, max_length
, '~');
295 return *word
? 0 : WRDE_NOSPACE
;
299 if (i
== 1 + *offset
)
301 /* Tilde appears on its own */
304 /* POSIX.2 says ~ expands to $HOME and if HOME is unset the
305 results are unspecified. We do a lookup on the uid if
308 home
= getenv ("HOME");
311 *word
= w_addstr (*word
, word_length
, max_length
, home
);
317 struct passwd pwd
, *tpwd
;
318 uid_t uid
= __getuid ();
320 struct scratch_buffer tmpbuf
;
321 scratch_buffer_init (&tmpbuf
);
323 while ((result
= __getpwuid_r (uid
, &pwd
,
324 tmpbuf
.data
, tmpbuf
.length
,
327 if (!scratch_buffer_grow (&tmpbuf
))
330 if (result
== 0 && tpwd
!= NULL
&& pwd
.pw_dir
!= NULL
)
332 *word
= w_addstr (*word
, word_length
, max_length
, pwd
.pw_dir
);
335 scratch_buffer_free (&tmpbuf
);
341 *word
= w_addchar (*word
, word_length
, max_length
, '~');
344 scratch_buffer_free (&tmpbuf
);
348 scratch_buffer_free (&tmpbuf
);
353 /* Look up user name in database to get home directory */
354 char *user
= strndupa (&words
[1 + *offset
], i
- (1 + *offset
));
355 struct passwd pwd
, *tpwd
;
357 struct scratch_buffer tmpbuf
;
358 scratch_buffer_init (&tmpbuf
);
360 while ((result
= __getpwnam_r (user
, &pwd
, tmpbuf
.data
, tmpbuf
.length
,
363 if (!scratch_buffer_grow (&tmpbuf
))
366 if (result
== 0 && tpwd
!= NULL
&& pwd
.pw_dir
)
367 *word
= w_addstr (*word
, word_length
, max_length
, pwd
.pw_dir
);
370 /* (invalid login name) */
371 *word
= w_addchar (*word
, word_length
, max_length
, '~');
373 *word
= w_addstr (*word
, word_length
, max_length
, user
);
376 scratch_buffer_free (&tmpbuf
);
380 return *word
? 0 : WRDE_NOSPACE
;
385 do_parse_glob (const char *glob_word
, char **word
, size_t *word_length
,
386 size_t *max_length
, wordexp_t
*pwordexp
, const char *ifs
,
387 const char *ifs_white
)
393 error
= glob (glob_word
, GLOB_NOCHECK
, NULL
, &globbuf
);
397 /* We can only run into memory problems. */
398 assert (error
== GLOB_NOSPACE
);
404 /* No field splitting allowed. */
405 assert (globbuf
.gl_pathv
[0] != NULL
);
406 *word
= w_addstr (*word
, word_length
, max_length
, globbuf
.gl_pathv
[0]);
407 for (match
= 1; match
< globbuf
.gl_pathc
&& *word
!= NULL
; ++match
)
409 *word
= w_addchar (*word
, word_length
, max_length
, ' ');
411 *word
= w_addstr (*word
, word_length
, max_length
,
412 globbuf
.gl_pathv
[match
]);
416 return *word
? 0 : WRDE_NOSPACE
;
419 assert (ifs
== NULL
|| *ifs
!= '\0');
423 *word
= w_newword (word_length
, max_length
);
426 for (match
= 0; match
< globbuf
.gl_pathc
; ++match
)
428 char *matching_word
= __strdup (globbuf
.gl_pathv
[match
]);
429 if (matching_word
== NULL
|| w_addword (pwordexp
, matching_word
))
441 parse_glob (char **word
, size_t *word_length
, size_t *max_length
,
442 const char *words
, size_t *offset
, int flags
,
443 wordexp_t
*pwordexp
, const char *ifs
, const char *ifs_white
)
445 /* We are poised just after a '*', a '[' or a '?'. */
446 int error
= WRDE_NOSPACE
;
447 int quoted
= 0; /* 1 if singly-quoted, 2 if doubly */
449 wordexp_t glob_list
; /* List of words to glob */
451 glob_list
.we_wordc
= 0;
452 glob_list
.we_wordv
= NULL
;
453 glob_list
.we_offs
= 0;
454 for (; words
[*offset
] != '\0'; ++*offset
)
456 if (strchr (ifs
, words
[*offset
]) != NULL
)
460 /* Sort out quoting */
461 if (words
[*offset
] == '\'')
468 else if (quoted
== 1)
474 else if (words
[*offset
] == '"')
481 else if (quoted
== 2)
488 /* Sort out other special characters */
489 if (quoted
!= 1 && words
[*offset
] == '$')
491 error
= parse_dollars (word
, word_length
, max_length
, words
,
492 offset
, flags
, &glob_list
, ifs
, ifs_white
,
499 else if (words
[*offset
] == '\\')
502 error
= parse_qtd_backslash (word
, word_length
, max_length
,
505 error
= parse_backslash (word
, word_length
, max_length
,
514 *word
= w_addchar (*word
, word_length
, max_length
, words
[*offset
]);
519 /* Don't forget to re-parse the character we stopped at. */
523 error
= w_addword (&glob_list
, *word
);
524 *word
= w_newword (word_length
, max_length
);
525 for (i
= 0; error
== 0 && i
< glob_list
.we_wordc
; i
++)
526 error
= do_parse_glob (glob_list
.we_wordv
[i
], word
, word_length
,
527 max_length
, pwordexp
, ifs
, ifs_white
);
531 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 */
557 eval_expr_val (char **expr
, long int *result
)
561 /* Skip white space */
562 for (digit
= *expr
; digit
&& *digit
&& isspace (*digit
); ++digit
);
566 /* Scan for closing paren */
567 for (++digit
; **expr
&& **expr
!= ')'; ++(*expr
));
575 if (eval_expr (digit
, result
))
581 /* POSIX requires that decimal, octal, and hexadecimal constants are
582 recognized. Therefore we pass 0 as the third parameter to strtol. */
583 *result
= strtol (digit
, expr
, 0);
591 eval_expr_multdiv (char **expr
, long int *result
)
596 if (eval_expr_val (expr
, result
) != 0)
601 /* Skip white space */
602 for (; *expr
&& **expr
&& isspace (**expr
); ++(*expr
));
607 if (eval_expr_val (expr
, &arg
) != 0)
612 else if (**expr
== '/')
615 if (eval_expr_val (expr
, &arg
) != 0)
618 /* Division by zero or integer overflow. */
619 if (arg
== 0 || (arg
== -1 && *result
== LONG_MIN
))
631 eval_expr (char *expr
, long int *result
)
636 if (eval_expr_multdiv (&expr
, result
) != 0)
641 /* Skip white space */
642 for (; expr
&& *expr
&& isspace (*expr
); ++expr
);
647 if (eval_expr_multdiv (&expr
, &arg
) != 0)
652 else if (*expr
== '-')
655 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
;
804 expr
= w_addchar (expr
, &expr_length
, &expr_maxlen
, words
[*offset
]);
815 /* Function called by child process in exec_comm() */
817 __attribute__ ((always_inline
))
818 exec_comm_child (char *comm
, int *fildes
, int showerr
, int noexec
)
820 const char *args
[4] = { _PATH_BSHELL
, "-c", comm
, NULL
};
822 /* Execute the command, or just check syntax? */
826 /* Redirect output. */
827 if (__glibc_likely (fildes
[1] != STDOUT_FILENO
))
829 __dup2 (fildes
[1], STDOUT_FILENO
);
833 /* Reset the close-on-exec flag (if necessary). */
834 __fcntl (fildes
[1], F_SETFD
, 0);
836 /* Redirect stderr to /dev/null if we have to. */
841 __close (STDERR_FILENO
);
842 fd
= __open (_PATH_DEVNULL
, O_WRONLY
);
843 if (fd
>= 0 && fd
!= STDERR_FILENO
)
845 __dup2 (fd
, STDERR_FILENO
);
848 /* Be paranoid. Check that we actually opened the /dev/null
850 if (__builtin_expect (__fxstat64 (_STAT_VER
, STDERR_FILENO
, &st
), 0) != 0
851 || __builtin_expect (S_ISCHR (st
.st_mode
), 1) == 0
852 #if defined DEV_NULL_MAJOR && defined DEV_NULL_MINOR
853 || st
.st_rdev
!= __gnu_dev_makedev (DEV_NULL_MAJOR
, DEV_NULL_MINOR
)
856 /* It's not the /dev/null device. Stop right here. The
857 problem is: how do we stop? We use _exit() with an
858 hopefully unusual exit code. */
862 /* Make sure the subshell doesn't field-split on our behalf. */
866 __execve (_PATH_BSHELL
, (char *const *) args
, __environ
);
872 /* Function to execute a command and retrieve the results */
873 /* pwordexp contains NULL if field-splitting is forbidden */
875 exec_comm (char *comm
, char **word
, size_t *word_length
, size_t *max_length
,
876 int flags
, wordexp_t
*pwordexp
, const char *ifs
,
877 const char *ifs_white
)
884 size_t maxnewlines
= 0;
885 char buffer
[bufsize
];
889 /* Do nothing if command substitution should not succeed. */
890 if (flags
& WRDE_NOCMD
)
893 /* Don't fork() unless necessary */
897 if (__pipe2 (fildes
, O_CLOEXEC
) < 0)
901 if ((pid
= __fork ()) < 0)
910 exec_comm_child (comm
, fildes
, noexec
? 0 : flags
& WRDE_SHOWERR
, noexec
);
914 /* If we are just testing the syntax, only wait. */
916 return (TEMP_FAILURE_RETRY (__waitpid (pid
, &status
, 0)) == pid
917 && status
!= 0) ? WRDE_SYNTAX
: 0;
923 /* Quoted - no field splitting */
927 if ((buflen
= TEMP_FAILURE_RETRY (__read (fildes
[0], buffer
,
930 /* If read returned 0 then the process has closed its
931 stdout. Don't use WNOHANG in that case to avoid busy
932 looping until the process eventually exits. */
933 if (TEMP_FAILURE_RETRY (__waitpid (pid
, &status
,
934 buflen
== 0 ? 0 : WNOHANG
))
937 if ((buflen
= TEMP_FAILURE_RETRY (__read (fildes
[0], buffer
,
942 maxnewlines
+= buflen
;
944 *word
= w_addmem (*word
, word_length
, max_length
, buffer
, buflen
);
950 /* Not quoted - split fields */
954 * 0 when searching for first character in a field not IFS white space
955 * 1 when copying the text of a field
956 * 2 when searching for possible non-whitespace IFS
957 * 3 when searching for non-newline after copying field
962 if ((buflen
= TEMP_FAILURE_RETRY (__read (fildes
[0], buffer
,
965 /* If read returned 0 then the process has closed its
966 stdout. Don't use WNOHANG in that case to avoid busy
967 looping until the process eventually exits. */
968 if (TEMP_FAILURE_RETRY (__waitpid (pid
, &status
,
969 buflen
== 0 ? 0 : WNOHANG
))
972 if ((buflen
= TEMP_FAILURE_RETRY (__read (fildes
[0], buffer
,
977 for (i
= 0; i
< buflen
; ++i
)
979 if (strchr (ifs
, buffer
[i
]) != NULL
)
981 /* Current character is IFS */
982 if (strchr (ifs_white
, buffer
[i
]) == NULL
)
984 /* Current character is IFS but not whitespace */
990 * eg: text<space><comma><space>moretext
992 * So, strip whitespace IFS (like at the start)
999 /* fall through and delimit field.. */
1003 if (buffer
[i
] == '\n')
1005 /* Current character is (IFS) newline */
1007 /* If copying a field, this is the end of it,
1008 but maybe all that's left is trailing newlines.
1009 So start searching for a non-newline. */
1017 /* Current character is IFS white space, but
1020 /* If not either copying a field or searching
1021 for non-newline after a field, ignore it */
1022 if (copying
!= 1 && copying
!= 3)
1025 /* End of field (search for non-ws IFS afterwards) */
1030 /* First IFS white space (non-newline), or IFS non-whitespace.
1031 * Delimit the field. Nulls are converted by w_addword. */
1032 if (w_addword (pwordexp
, *word
) == WRDE_NOSPACE
)
1035 *word
= w_newword (word_length
, max_length
);
1038 /* fall back round the loop.. */
1042 /* Not IFS character */
1046 /* Nothing but (IFS) newlines since the last field,
1047 so delimit it here before starting new word */
1048 if (w_addword (pwordexp
, *word
) == WRDE_NOSPACE
)
1051 *word
= w_newword (word_length
, max_length
);
1056 if (buffer
[i
] == '\n') /* happens if newline not in IFS */
1061 *word
= w_addchar (*word
, word_length
, max_length
,
1070 /* Chop off trailing newlines (required by POSIX.2) */
1071 /* Ensure we don't go back further than the beginning of the
1072 substitution (i.e. remove maxnewlines bytes at most) */
1073 while (maxnewlines
-- != 0
1074 && *word_length
> 0 && (*word
)[*word_length
- 1] == '\n')
1076 (*word
)[--*word_length
] = '\0';
1078 /* If the last word was entirely newlines, turn it into a new word
1079 * which can be ignored if there's nothing following it. */
1080 if (*word_length
== 0)
1083 *word
= w_newword (word_length
, max_length
);
1088 __close (fildes
[0]);
1091 /* Check for syntax error (re-execute but with "-n" flag) */
1092 if (buflen
< 1 && status
!= 0)
1101 __kill (pid
, SIGKILL
);
1102 TEMP_FAILURE_RETRY (__waitpid (pid
, NULL
, 0));
1103 __close (fildes
[0]);
1104 return WRDE_NOSPACE
;
1108 parse_comm (char **word
, size_t *word_length
, size_t *max_length
,
1109 const char *words
, size_t *offset
, int flags
, wordexp_t
*pwordexp
,
1110 const char *ifs
, const char *ifs_white
)
1112 /* We are poised just after "$(" */
1113 int paren_depth
= 1;
1115 int quoted
= 0; /* 1 for singly-quoted, 2 for doubly-quoted */
1118 char *comm
= w_newword (&comm_length
, &comm_maxlen
);
1120 for (; words
[*offset
]; ++(*offset
))
1122 switch (words
[*offset
])
1127 else if (quoted
== 1)
1135 else if (quoted
== 2)
1141 if (!quoted
&& --paren_depth
== 0)
1143 /* Go -- give script to the shell */
1146 #ifdef __libc_ptf_call
1147 /* We do not want the exec_comm call to be cut short
1148 by a thread cancellation since cleanup is very
1149 ugly. Therefore disable cancellation for
1151 // XXX Ideally we do want the thread being cancelable.
1152 // XXX If demand is there we'll change it.
1153 int state
= PTHREAD_CANCEL_ENABLE
;
1154 __libc_ptf_call (__pthread_setcancelstate
,
1155 (PTHREAD_CANCEL_DISABLE
, &state
), 0);
1158 error
= exec_comm (comm
, word
, word_length
, max_length
,
1159 flags
, pwordexp
, ifs
, ifs_white
);
1161 #ifdef __libc_ptf_call
1162 __libc_ptf_call (__pthread_setcancelstate
,
1172 /* This is just part of the script */
1180 comm
= w_addchar (comm
, &comm_length
, &comm_maxlen
, words
[*offset
]);
1182 return WRDE_NOSPACE
;
1185 /* Premature end. */
1191 #define CHAR_IN_SET(ch, char_set) \
1192 (memchr (char_set "", ch, sizeof (char_set) - 1) != NULL)
1195 parse_param (char **word
, size_t *word_length
, size_t *max_length
,
1196 const char *words
, size_t *offset
, int flags
, wordexp_t
*pwordexp
,
1197 const char *ifs
, const char *ifs_white
, int quoted
)
1199 /* We are poised just after "$" */
1203 ACT_RP_SHORT_LEFT
= '#',
1204 ACT_RP_LONG_LEFT
= 'L',
1205 ACT_RP_SHORT_RIGHT
= '%',
1206 ACT_RP_LONG_RIGHT
= 'R',
1207 ACT_NULL_ERROR
= '?',
1208 ACT_NULL_SUBST
= '-',
1209 ACT_NONNULL_SUBST
= '+',
1210 ACT_NULL_ASSIGN
= '='
1216 size_t start
= *offset
;
1220 enum action action
= ACT_NONE
;
1225 int pattern_is_quoted
= 0; /* 1 for singly-quoted, 2 for doubly-quoted */
1229 int brace
= words
[*offset
] == '{';
1231 env
= w_newword (&env_length
, &env_maxlen
);
1232 pattern
= w_newword (&pat_length
, &pat_maxlen
);
1237 /* First collect the parameter name. */
1239 if (words
[*offset
] == '#')
1247 if (isalpha (words
[*offset
]) || words
[*offset
] == '_')
1249 /* Normal parameter name. */
1252 env
= w_addchar (env
, &env_length
, &env_maxlen
,
1257 while (isalnum (words
[++*offset
]) || words
[*offset
] == '_');
1259 else if (isdigit (words
[*offset
]))
1261 /* Numeric parameter name. */
1265 env
= w_addchar (env
, &env_length
, &env_maxlen
,
1272 while (isdigit(words
[++*offset
]));
1274 else if (CHAR_IN_SET (words
[*offset
], "*@$"))
1276 /* Special parameter. */
1278 env
= w_addchar (env
, &env_length
, &env_maxlen
,
1292 /* Check for special action to be applied to the value. */
1293 switch (words
[*offset
])
1300 action
= ACT_RP_SHORT_LEFT
;
1301 if (words
[1 + *offset
] == '#')
1304 action
= ACT_RP_LONG_LEFT
;
1309 action
= ACT_RP_SHORT_RIGHT
;
1310 if (words
[1 + *offset
] == '%')
1313 action
= ACT_RP_LONG_RIGHT
;
1318 if (!CHAR_IN_SET (words
[1 + *offset
], "-=?+"))
1322 action
= words
[++*offset
];
1329 action
= words
[*offset
];
1336 /* Now collect the pattern, but don't expand it yet. */
1338 for (; words
[*offset
]; ++(*offset
))
1340 switch (words
[*offset
])
1343 if (!pattern_is_quoted
)
1348 if (!pattern_is_quoted
)
1357 if (pattern_is_quoted
)
1358 /* Quoted; treat as normal character. */
1361 /* Otherwise, it's an escape: next character is literal. */
1362 if (words
[++*offset
] == '\0')
1365 pattern
= w_addchar (pattern
, &pat_length
, &pat_maxlen
, '\\');
1366 if (pattern
== NULL
)
1372 if (pattern_is_quoted
== 0)
1373 pattern_is_quoted
= 1;
1374 else if (pattern_is_quoted
== 1)
1375 pattern_is_quoted
= 0;
1380 if (pattern_is_quoted
== 0)
1381 pattern_is_quoted
= 2;
1382 else if (pattern_is_quoted
== 2)
1383 pattern_is_quoted
= 0;
1388 pattern
= w_addchar (pattern
, &pat_length
, &pat_maxlen
,
1390 if (pattern
== NULL
)
1395 /* End of input string -- remember to reparse the character that we
1400 if (words
[start
] == '{' && words
[*offset
] != '}')
1407 /* $# expands to the number of positional parameters */
1409 value
= _itoa_word (__libc_argc
- 1, &buffer
[20], 10, 0);
1414 /* Just $ on its own */
1415 *offset
= start
- 1;
1416 *word
= w_addchar (*word
, word_length
, max_length
, '$');
1417 return *word
? 0 : WRDE_NOSPACE
;
1420 /* Is it a numeric parameter? */
1421 else if (isdigit (env
[0]))
1425 if (n
>= __libc_argc
)
1426 /* Substitute NULL. */
1429 /* Replace with appropriate positional parameter. */
1430 value
= __libc_argv
[n
];
1432 /* Is it a special parameter? */
1439 value
= _itoa_word (__getpid (), &buffer
[20], 10, 0);
1441 /* Is it `${#*}' or `${#@}'? */
1442 else if ((*env
== '*' || *env
== '@') && seen_hash
)
1445 value
= _itoa_word (__libc_argc
> 0 ? __libc_argc
- 1 : 0,
1446 &buffer
[20], 10, 0);
1447 *word
= w_addstr (*word
, word_length
, max_length
, value
);
1450 return *word
? 0 : WRDE_NOSPACE
;
1452 /* Is it `$*' or `$@' (unquoted) ? */
1453 else if (*env
== '*' || (*env
== '@' && !quoted
))
1455 size_t plist_len
= 0;
1459 /* Build up value parameter by parameter (copy them) */
1460 for (p
= 1; __libc_argv
[p
]; ++p
)
1461 plist_len
+= strlen (__libc_argv
[p
]) + 1; /* for space */
1462 value
= malloc (plist_len
);
1467 for (p
= 1; __libc_argv
[p
]; ++p
)
1471 end
= __stpcpy (end
, __libc_argv
[p
]);
1478 /* Must be a quoted `$@' */
1479 assert (*env
== '@' && quoted
);
1481 /* Each parameter is a separate word ("$@") */
1482 if (__libc_argc
== 2)
1483 value
= __libc_argv
[1];
1484 else if (__libc_argc
> 2)
1488 /* Append first parameter to current word. */
1489 value
= w_addstr (*word
, word_length
, max_length
,
1491 if (value
== NULL
|| w_addword (pwordexp
, value
))
1494 for (p
= 2; __libc_argv
[p
+ 1]; p
++)
1496 char *newword
= __strdup (__libc_argv
[p
]);
1497 if (newword
== NULL
|| w_addword (pwordexp
, newword
))
1501 /* Start a new word with the last parameter. */
1502 *word
= w_newword (word_length
, max_length
);
1503 value
= __libc_argv
[p
];
1514 value
= getenv (env
);
1516 if (value
== NULL
&& (flags
& WRDE_UNDEF
))
1518 /* Variable not defined. */
1519 error
= WRDE_BADVAL
;
1523 if (action
!= ACT_NONE
)
1525 int expand_pattern
= 0;
1527 /* First, find out if we need to expand pattern (i.e. if we will
1531 case ACT_RP_SHORT_LEFT
:
1532 case ACT_RP_LONG_LEFT
:
1533 case ACT_RP_SHORT_RIGHT
:
1534 case ACT_RP_LONG_RIGHT
:
1535 /* Always expand for these. */
1539 case ACT_NULL_ERROR
:
1540 case ACT_NULL_SUBST
:
1541 case ACT_NULL_ASSIGN
:
1542 if (!value
|| (!*value
&& colon_seen
))
1543 /* If param is unset, or set but null and a colon has been seen,
1544 the expansion of the pattern will be needed. */
1549 case ACT_NONNULL_SUBST
:
1550 /* Expansion of word will be needed if parameter is set and not null,
1551 or set null but no colon has been seen. */
1552 if (value
&& (*value
|| !colon_seen
))
1558 assert (! "Unrecognised action!");
1563 /* We need to perform tilde expansion, parameter expansion,
1564 command substitution, and arithmetic expansion. We also
1565 have to be a bit careful with wildcard characters, as
1566 pattern might be given to fnmatch soon. To do this, we
1567 convert quotes to escapes. */
1573 int quoted
= 0; /* 1: single quotes; 2: double */
1575 expanded
= w_newword (&exp_len
, &exp_maxl
);
1576 for (p
= pattern
; p
&& *p
; p
++)
1585 else if (quoted
== 0)
1594 else if (quoted
== 0)
1604 /* Convert quoted wildchar to escaped wildchar. */
1605 expanded
= w_addchar (expanded
, &exp_len
,
1608 if (expanded
== NULL
)
1615 error
= parse_dollars (&expanded
, &exp_len
, &exp_maxl
, p
,
1616 &offset
, flags
, NULL
, NULL
, NULL
, 1);
1631 if (quoted
|| exp_len
)
1635 error
= parse_tilde (&expanded
, &exp_len
, &exp_maxl
, p
,
1651 expanded
= w_addchar (expanded
, &exp_len
, &exp_maxl
, '\\');
1653 assert (*p
); /* checked when extracted initially */
1654 if (expanded
== NULL
)
1658 expanded
= w_addchar (expanded
, &exp_len
, &exp_maxl
, *p
);
1660 if (expanded
== NULL
)
1671 case ACT_RP_SHORT_LEFT
:
1672 case ACT_RP_LONG_LEFT
:
1673 case ACT_RP_SHORT_RIGHT
:
1674 case ACT_RP_LONG_RIGHT
:
1680 if (value
== NULL
|| pattern
== NULL
|| *pattern
== '\0')
1683 end
= value
+ strlen (value
);
1687 case ACT_RP_SHORT_LEFT
:
1688 for (p
= value
; p
<= end
; ++p
)
1692 if (fnmatch (pattern
, value
, 0) != FNM_NOMATCH
)
1697 char *newval
= __strdup (p
);
1715 case ACT_RP_LONG_LEFT
:
1716 for (p
= end
; p
>= value
; --p
)
1720 if (fnmatch (pattern
, value
, 0) != FNM_NOMATCH
)
1725 char *newval
= __strdup (p
);
1743 case ACT_RP_SHORT_RIGHT
:
1744 for (p
= end
; p
>= value
; --p
)
1746 if (fnmatch (pattern
, p
, 0) != FNM_NOMATCH
)
1749 newval
= malloc (p
- value
+ 1);
1758 *(char *) __mempcpy (newval
, value
, p
- value
) = '\0';
1769 case ACT_RP_LONG_RIGHT
:
1770 for (p
= value
; p
<= end
; ++p
)
1772 if (fnmatch (pattern
, p
, 0) != FNM_NOMATCH
)
1775 newval
= malloc (p
- value
+ 1);
1784 *(char *) __mempcpy (newval
, value
, p
- value
) = '\0';
1802 case ACT_NULL_ERROR
:
1803 if (value
&& *value
)
1804 /* Substitute parameter */
1808 if (!colon_seen
&& value
)
1809 /* Substitute NULL */
1813 const char *str
= pattern
;
1816 str
= _("parameter null or not set");
1818 __fxprintf (NULL
, "%s: %s\n", env
, str
);
1825 case ACT_NULL_SUBST
:
1826 if (value
&& *value
)
1827 /* Substitute parameter */
1833 if (!colon_seen
&& value
)
1834 /* Substitute NULL */
1837 value
= pattern
? __strdup (pattern
) : pattern
;
1840 if (pattern
&& !value
)
1845 case ACT_NONNULL_SUBST
:
1846 if (value
&& (*value
|| !colon_seen
))
1851 value
= pattern
? __strdup (pattern
) : pattern
;
1854 if (pattern
&& !value
)
1860 /* Substitute NULL */
1865 case ACT_NULL_ASSIGN
:
1866 if (value
&& *value
)
1867 /* Substitute parameter */
1870 if (!colon_seen
&& value
)
1872 /* Substitute NULL */
1881 value
= pattern
? __strdup (pattern
) : pattern
;
1884 if (pattern
&& !value
)
1887 __setenv (env
, value
?: "", 1);
1891 assert (! "Unrecognised action!");
1902 char param_length
[21];
1903 param_length
[20] = '\0';
1904 *word
= w_addstr (*word
, word_length
, max_length
,
1905 _itoa_word (value
? strlen (value
) : 0,
1906 ¶m_length
[20], 10, 0));
1909 assert (value
!= NULL
);
1913 return *word
? 0 : WRDE_NOSPACE
;
1919 if (quoted
|| !pwordexp
)
1921 /* Quoted - no field split */
1922 *word
= w_addstr (*word
, word_length
, max_length
, value
);
1926 return *word
? 0 : WRDE_NOSPACE
;
1930 /* Need to field-split */
1931 char *value_copy
= __strdup (value
); /* Don't modify value */
1932 char *field_begin
= value_copy
;
1933 int seen_nonws_ifs
= 0;
1938 if (value_copy
== NULL
)
1943 char *field_end
= field_begin
;
1946 /* If this isn't the first field, start a new word */
1947 if (field_begin
!= value_copy
)
1949 if (w_addword (pwordexp
, *word
) == WRDE_NOSPACE
)
1955 *word
= w_newword (word_length
, max_length
);
1958 /* Skip IFS whitespace before the field */
1959 field_begin
+= strspn (field_begin
, ifs_white
);
1961 if (!seen_nonws_ifs
&& *field_begin
== 0)
1962 /* Nothing but whitespace */
1965 /* Search for the end of the field */
1966 field_end
= field_begin
+ strcspn (field_begin
, ifs
);
1968 /* Set up pointer to the character after end of field and
1969 skip whitespace IFS after it. */
1970 next_field
= field_end
+ strspn (field_end
, ifs_white
);
1972 /* Skip at most one non-whitespace IFS character after the field */
1974 if (*next_field
&& strchr (ifs
, *next_field
))
1980 /* Null-terminate it */
1983 /* Tag a copy onto the current word */
1984 *word
= w_addstr (*word
, word_length
, max_length
, field_begin
);
1986 if (*word
== NULL
&& *field_begin
!= '\0')
1992 field_begin
= next_field
;
1994 while (seen_nonws_ifs
|| *field_begin
);
2006 error
= WRDE_NOSPACE
;
2010 error
= WRDE_SYNTAX
;
2023 parse_dollars (char **word
, size_t *word_length
, size_t *max_length
,
2024 const char *words
, size_t *offset
, int flags
,
2025 wordexp_t
*pwordexp
, const char *ifs
, const char *ifs_white
,
2028 /* We are poised _at_ "$" */
2029 switch (words
[1 + *offset
])
2034 *word
= w_addchar (*word
, word_length
, max_length
, '$');
2035 return *word
? 0 : WRDE_NOSPACE
;
2038 if (words
[2 + *offset
] == '(')
2040 /* Differentiate between $((1+3)) and $((echo);(ls)) */
2041 int i
= 3 + *offset
;
2043 while (words
[i
] && !(depth
== 0 && words
[i
] == ')'))
2045 if (words
[i
] == '(')
2047 else if (words
[i
] == ')')
2053 if (words
[i
] == ')' && words
[i
+ 1] == ')')
2056 /* Call parse_arith -- 0 is for "no brackets" */
2057 return parse_arith (word
, word_length
, max_length
, words
, offset
,
2063 return parse_comm (word
, word_length
, max_length
, words
, offset
, flags
,
2064 quoted
? NULL
: pwordexp
, ifs
, ifs_white
);
2068 /* Call parse_arith -- 1 is for "brackets" */
2069 return parse_arith (word
, word_length
, max_length
, words
, offset
, flags
,
2074 ++(*offset
); /* parse_param needs to know if "{" is there */
2075 return parse_param (word
, word_length
, max_length
, words
, offset
, flags
,
2076 pwordexp
, ifs
, ifs_white
, quoted
);
2081 parse_backtick (char **word
, size_t *word_length
, size_t *max_length
,
2082 const char *words
, size_t *offset
, int flags
,
2083 wordexp_t
*pwordexp
, const char *ifs
, const char *ifs_white
)
2085 /* We are poised just after "`" */
2090 char *comm
= w_newword (&comm_length
, &comm_maxlen
);
2092 for (; words
[*offset
]; ++(*offset
))
2094 switch (words
[*offset
])
2097 /* Go -- give the script to the shell */
2098 error
= exec_comm (comm
, word
, word_length
, max_length
, flags
,
2099 pwordexp
, ifs
, ifs_white
);
2106 error
= parse_qtd_backslash (&comm
, &comm_length
, &comm_maxlen
,
2118 error
= parse_backslash (&comm
, &comm_length
, &comm_maxlen
, words
,
2130 squoting
= 1 - squoting
;
2133 comm
= w_addchar (comm
, &comm_length
, &comm_maxlen
, words
[*offset
]);
2135 return WRDE_NOSPACE
;
2145 parse_dquote (char **word
, size_t *word_length
, size_t *max_length
,
2146 const char *words
, size_t *offset
, int flags
,
2147 wordexp_t
*pwordexp
, const char * ifs
, const char * ifs_white
)
2149 /* We are poised just after a double-quote */
2152 for (; words
[*offset
]; ++(*offset
))
2154 switch (words
[*offset
])
2160 error
= parse_dollars (word
, word_length
, max_length
, words
, offset
,
2161 flags
, pwordexp
, ifs
, ifs_white
, 1);
2162 /* The ``1'' here is to tell parse_dollars not to
2163 * split the fields. It may need to, however ("$@").
2172 error
= parse_backtick (word
, word_length
, max_length
, words
,
2173 offset
, flags
, NULL
, NULL
, NULL
);
2174 /* The first NULL here is to tell parse_backtick not to
2183 error
= parse_qtd_backslash (word
, word_length
, max_length
, words
,
2192 *word
= w_addchar (*word
, word_length
, max_length
, words
[*offset
]);
2194 return WRDE_NOSPACE
;
2198 /* Unterminated string */
2203 * wordfree() is to be called after pwordexp is finished with.
2207 wordfree (wordexp_t
*pwordexp
)
2210 /* wordexp can set pwordexp to NULL */
2211 if (pwordexp
&& pwordexp
->we_wordv
)
2213 char **wordv
= pwordexp
->we_wordv
;
2215 for (wordv
+= pwordexp
->we_offs
; *wordv
; ++wordv
)
2218 free (pwordexp
->we_wordv
);
2219 pwordexp
->we_wordv
= NULL
;
2222 libc_hidden_def (wordfree
)
2229 wordexp (const char *words
, wordexp_t
*pwordexp
, int flags
)
2231 size_t words_offset
;
2234 char *word
= w_newword (&word_length
, &max_length
);
2238 wordexp_t old_word
= *pwordexp
;
2240 if (flags
& WRDE_REUSE
)
2242 /* Minimal implementation of WRDE_REUSE for now */
2243 wordfree (pwordexp
);
2244 old_word
.we_wordv
= NULL
;
2247 if ((flags
& WRDE_APPEND
) == 0)
2249 pwordexp
->we_wordc
= 0;
2251 if (flags
& WRDE_DOOFFS
)
2253 pwordexp
->we_wordv
= calloc (1 + pwordexp
->we_offs
, sizeof (char *));
2254 if (pwordexp
->we_wordv
== NULL
)
2256 error
= WRDE_NOSPACE
;
2262 pwordexp
->we_wordv
= calloc (1, sizeof (char *));
2263 if (pwordexp
->we_wordv
== NULL
)
2265 error
= WRDE_NOSPACE
;
2269 pwordexp
->we_offs
= 0;
2273 /* Find out what the field separators are.
2274 * There are two types: whitespace and non-whitespace.
2276 ifs
= getenv ("IFS");
2279 /* IFS unset - use <space><tab><newline>. */
2280 ifs
= strcpy (ifs_white
, " \t\n");
2284 char *whch
= ifs_white
;
2286 while (*ifsch
!= '\0')
2288 if (*ifsch
== ' ' || *ifsch
== '\t' || *ifsch
== '\n')
2290 /* Whitespace IFS. See first whether it is already in our
2292 char *runp
= ifs_white
;
2294 while (runp
< whch
&& *runp
!= *ifsch
)
2306 for (words_offset
= 0 ; words
[words_offset
] ; ++words_offset
)
2307 switch (words
[words_offset
])
2310 error
= parse_backslash (&word
, &word_length
, &max_length
, words
,
2319 error
= parse_dollars (&word
, &word_length
, &max_length
, words
,
2320 &words_offset
, flags
, pwordexp
, ifs
, ifs_white
,
2330 error
= parse_backtick (&word
, &word_length
, &max_length
, words
,
2331 &words_offset
, flags
, pwordexp
, ifs
,
2341 error
= parse_dquote (&word
, &word_length
, &max_length
, words
,
2342 &words_offset
, flags
, pwordexp
, ifs
, ifs_white
);
2349 error
= w_addword (pwordexp
, NULL
);
2359 error
= parse_squote (&word
, &word_length
, &max_length
, words
,
2367 error
= w_addword (pwordexp
, NULL
);
2376 error
= parse_tilde (&word
, &word_length
, &max_length
, words
,
2377 &words_offset
, pwordexp
->we_wordc
);
2387 error
= parse_glob (&word
, &word_length
, &max_length
, words
,
2388 &words_offset
, flags
, pwordexp
, ifs
, ifs_white
);
2396 /* Is it a word separator? */
2397 if (strchr (" \t", words
[words_offset
]) == NULL
)
2399 char ch
= words
[words_offset
];
2401 /* Not a word separator -- but is it a valid word char? */
2402 if (strchr ("\n|&;<>(){}", ch
))
2405 error
= WRDE_BADCHAR
;
2409 /* "Ordinary" character -- add it to word */
2410 word
= w_addchar (word
, &word_length
, &max_length
,
2414 error
= WRDE_NOSPACE
;
2421 /* If a word has been delimited, add it to the list. */
2424 error
= w_addword (pwordexp
, word
);
2429 word
= w_newword (&word_length
, &max_length
);
2434 /* There was a word separator at the end */
2435 if (word
== NULL
) /* i.e. w_newword */
2438 /* There was no field separator at the end */
2439 return w_addword (pwordexp
, word
);
2443 * free memory used (unless error is WRDE_NOSPACE), and
2444 * set pwordexp members back to what they were.
2449 if (error
== WRDE_NOSPACE
)
2450 return WRDE_NOSPACE
;
2452 if ((flags
& WRDE_APPEND
) == 0)
2453 wordfree (pwordexp
);
2455 *pwordexp
= old_word
;