2014-12-22 Steve Ellcey <sellcey@imgtec.com>
[glibc.git] / posix / wordexp.c
blob26f3a2653feba2b1a5904937d9d6b58c32109e24
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/>. */
20 #include <alloca.h>
21 #include <ctype.h>
22 #include <errno.h>
23 #include <fcntl.h>
24 #include <fnmatch.h>
25 #include <glob.h>
26 #include <libintl.h>
27 #include <paths.h>
28 #include <pwd.h>
29 #include <signal.h>
30 #include <stdbool.h>
31 #include <stdio.h>
32 #include <stdlib.h>
33 #include <string.h>
34 #include <sys/param.h>
35 #include <sys/stat.h>
36 #include <sys/time.h>
37 #include <sys/types.h>
38 #include <sys/types.h>
39 #include <sys/wait.h>
40 #include <unistd.h>
41 #include <wchar.h>
42 #include <wordexp.h>
43 #include <kernel-features.h>
45 #include <bits/libc-lock.h>
46 #include <_itoa.h>
48 /* Undefine the following line for the production version. */
49 /* #define NDEBUG 1 */
50 #include <assert.h>
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 internal_function;
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)
73 internal_function;
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)
78 internal_function;
79 static int eval_expr (char *expr, long int *result) internal_function;
81 /* The w_*() functions manipulate word lists. */
83 #define W_CHUNK (100)
85 /* Result of w_newword will be ignored if it's the last word. */
86 static inline char *
87 w_newword (size_t *actlen, size_t *maxlen)
89 *actlen = *maxlen = 0;
90 return NULL;
93 static char *
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);
103 *maxlen += W_CHUNK;
104 buffer = (char *) realloc (buffer, 1 + *maxlen);
106 if (buffer == NULL)
107 free (old_buffer);
110 if (buffer != NULL)
112 buffer[*actlen] = ch;
113 buffer[++(*actlen)] = '\0';
116 return buffer;
119 static char *
120 internal_function
121 w_addmem (char *buffer, size_t *actlen, size_t *maxlen, const char *str,
122 size_t len)
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);
133 if (buffer == NULL)
134 free (old_buffer);
137 if (buffer != NULL)
139 *((char *) __mempcpy (&buffer[*actlen], str, len)) = '\0';
140 *actlen += len;
143 return buffer;
146 static char *
147 internal_function
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.
153 size_t len;
155 assert (str != NULL); /* w_addstr only called from this file */
156 len = strlen (str);
158 return w_addmem (buffer, actlen, maxlen, str, len);
161 static int
162 internal_function
163 w_addword (wordexp_t *pwordexp, char *word)
165 /* Add a word to the wordlist */
166 size_t num_p;
167 char **new_wordv;
168 bool allocated = false;
170 /* Internally, NULL acts like "". Convert NULLs to "" before
171 * the caller sees them.
173 if (word == NULL)
175 word = __strdup ("");
176 if (word == NULL)
177 goto no_space;
178 allocated = true;
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;
188 return 0;
191 if (allocated)
192 free (word);
194 no_space:
195 return WRDE_NOSPACE;
198 /* The parse_*() functions should leave *offset being the offset in 'words'
199 * to the last character processed.
202 static int
203 internal_function
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])
211 case 0:
212 /* Backslash is last character of input words */
213 return WRDE_SYNTAX;
215 case '\n':
216 ++(*offset);
217 break;
219 default:
220 *word = w_addchar (*word, word_length, max_length, words[1 + *offset]);
221 if (*word == NULL)
222 return WRDE_NOSPACE;
224 ++(*offset);
225 break;
228 return 0;
231 static int
232 internal_function
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])
240 case 0:
241 /* Backslash is last character of input words */
242 return WRDE_SYNTAX;
244 case '\n':
245 ++(*offset);
246 break;
248 case '$':
249 case '`':
250 case '"':
251 case '\\':
252 *word = w_addchar (*word, word_length, max_length, words[1 + *offset]);
253 if (*word == NULL)
254 return WRDE_NOSPACE;
256 ++(*offset);
257 break;
259 default:
260 *word = w_addchar (*word, word_length, max_length, words[*offset]);
261 if (*word != NULL)
262 *word = w_addchar (*word, word_length, max_length, words[1 + *offset]);
264 if (*word == NULL)
265 return WRDE_NOSPACE;
267 ++(*offset);
268 break;
271 return 0;
274 static int
275 internal_function
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 */
280 size_t i;
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 )
299 break;
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 */
311 uid_t uid;
312 struct passwd pwd, *tpwd;
313 int buflen = 1000;
314 char* home;
315 char* buffer;
316 int result;
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
320 HOME is unset. */
322 home = getenv ("HOME");
323 if (home != NULL)
325 *word = w_addstr (*word, word_length, max_length, home);
326 if (*word == NULL)
327 return WRDE_NOSPACE;
329 else
331 uid = __getuid ();
332 buffer = __alloca (buflen);
334 while ((result = __getpwuid_r (uid, &pwd, buffer, buflen, &tpwd)) != 0
335 && errno == ERANGE)
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);
341 if (*word == NULL)
342 return WRDE_NOSPACE;
344 else
346 *word = w_addchar (*word, word_length, max_length, '~');
347 if (*word == NULL)
348 return WRDE_NOSPACE;
352 else
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;
357 int buflen = 1000;
358 char* buffer = __alloca (buflen);
359 int result;
361 while ((result = __getpwnam_r (user, &pwd, buffer, buflen, &tpwd)) != 0
362 && errno == ERANGE)
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);
367 else
369 /* (invalid login name) */
370 *word = w_addchar (*word, word_length, max_length, '~');
371 if (*word != NULL)
372 *word = w_addstr (*word, word_length, max_length, user);
375 *offset = i - 1;
377 return *word ? 0 : WRDE_NOSPACE;
381 static int
382 internal_function
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)
387 int error;
388 unsigned int match;
389 glob_t globbuf;
391 error = glob (glob_word, GLOB_NOCHECK, NULL, &globbuf);
393 if (error != 0)
395 /* We can only run into memory problems. */
396 assert (error == GLOB_NOSPACE);
397 return WRDE_NOSPACE;
400 if (ifs && !*ifs)
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, ' ');
408 if (*word != NULL)
409 *word = w_addstr (*word, word_length, max_length,
410 globbuf.gl_pathv[match]);
413 globfree (&globbuf);
414 return *word ? 0 : WRDE_NOSPACE;
417 assert (ifs == NULL || *ifs != '\0');
418 if (*word != NULL)
420 free (*word);
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))
429 globfree (&globbuf);
430 return WRDE_NOSPACE;
434 globfree (&globbuf);
435 return 0;
438 static int
439 internal_function
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 */
447 size_t i;
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)
456 /* Reached IFS */
457 break;
459 /* Sort out quoting */
460 if (words[*offset] == '\'')
462 if (quoted == 0)
464 quoted = 1;
465 continue;
467 else if (quoted == 1)
469 quoted = 0;
470 continue;
473 else if (words[*offset] == '"')
475 if (quoted == 0)
477 quoted = 2;
478 continue;
480 else if (quoted == 2)
482 quoted = 0;
483 continue;
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,
492 quoted == 2);
493 if (error)
494 goto tidy_up;
496 continue;
498 else if (words[*offset] == '\\')
500 if (quoted)
501 error = parse_qtd_backslash (word, word_length, max_length,
502 words, offset);
503 else
504 error = parse_backslash (word, word_length, max_length,
505 words, offset);
507 if (error)
508 goto tidy_up;
510 continue;
513 *word = w_addchar (*word, word_length, max_length, words[*offset]);
514 if (*word == NULL)
515 goto tidy_up;
518 /* Don't forget to re-parse the character we stopped at. */
519 --*offset;
521 /* Glob the words */
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);
528 /* Now tidy up */
529 tidy_up:
530 wordfree (&glob_list);
531 return error;
534 static int
535 internal_function
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]);
545 if (*word == NULL)
546 return WRDE_NOSPACE;
548 else return 0;
551 /* Unterminated string */
552 return WRDE_SYNTAX;
555 /* Functions to evaluate an arithmetic expression */
556 static int
557 internal_function
558 eval_expr_val (char **expr, long int *result)
560 char *digit;
562 /* Skip white space */
563 for (digit = *expr; digit && *digit && isspace (*digit); ++digit);
565 if (*digit == '(')
567 /* Scan for closing paren */
568 for (++digit; **expr && **expr != ')'; ++(*expr));
570 /* Is there one? */
571 if (!**expr)
572 return WRDE_SYNTAX;
574 *(*expr)++ = 0;
576 if (eval_expr (digit, result))
577 return WRDE_SYNTAX;
579 return 0;
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);
585 if (digit == *expr)
586 return WRDE_SYNTAX;
588 return 0;
591 static int
592 internal_function
593 eval_expr_multdiv (char **expr, long int *result)
595 long int arg;
597 /* Read a Value */
598 if (eval_expr_val (expr, result) != 0)
599 return WRDE_SYNTAX;
601 while (**expr)
603 /* Skip white space */
604 for (; *expr && **expr && isspace (**expr); ++(*expr));
606 if (**expr == '*')
608 ++(*expr);
609 if (eval_expr_val (expr, &arg) != 0)
610 return WRDE_SYNTAX;
612 *result *= arg;
614 else if (**expr == '/')
616 ++(*expr);
617 if (eval_expr_val (expr, &arg) != 0)
618 return WRDE_SYNTAX;
620 *result /= arg;
622 else break;
625 return 0;
628 static int
629 internal_function
630 eval_expr (char *expr, long int *result)
632 long int arg;
634 /* Read a Multdiv */
635 if (eval_expr_multdiv (&expr, result) != 0)
636 return WRDE_SYNTAX;
638 while (*expr)
640 /* Skip white space */
641 for (; expr && *expr && isspace (*expr); ++expr);
643 if (*expr == '+')
645 ++expr;
646 if (eval_expr_multdiv (&expr, &arg) != 0)
647 return WRDE_SYNTAX;
649 *result += arg;
651 else if (*expr == '-')
653 ++expr;
654 if (eval_expr_multdiv (&expr, &arg) != 0)
655 return WRDE_SYNTAX;
657 *result -= arg;
659 else break;
662 return 0;
665 static int
666 internal_function
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 "$((" or "$[" */
671 int error;
672 int paren_depth = 1;
673 size_t expr_length;
674 size_t expr_maxlen;
675 char *expr;
677 expr = w_newword (&expr_length, &expr_maxlen);
678 for (; words[*offset]; ++(*offset))
680 switch (words[*offset])
682 case '$':
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
686 * split the fields.
688 if (error)
690 free (expr);
691 return error;
693 break;
695 case '`':
696 (*offset)++;
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
700 * split the fields.
702 if (error)
704 free (expr);
705 return error;
707 break;
709 case '\\':
710 error = parse_qtd_backslash (&expr, &expr_length, &expr_maxlen,
711 words, offset);
712 if (error)
714 free (expr);
715 return error;
717 /* I think that a backslash within an
718 * arithmetic expansion is bound to
719 * cause an error sooner or later anyway though.
721 break;
723 case ')':
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] != ')')
732 free (expr);
733 return WRDE_SYNTAX;
736 ++(*offset);
738 /* Go - evaluate. */
739 if (*expr && eval_expr (expr, &numresult) != 0)
741 free (expr);
742 return WRDE_SYNTAX;
745 if (numresult < 0)
747 convertme = -numresult;
748 *word = w_addchar (*word, word_length, max_length, '-');
749 if (!*word)
751 free (expr);
752 return WRDE_NOSPACE;
755 else
756 convertme = numresult;
758 result[20] = '\0';
759 *word = w_addstr (*word, word_length, max_length,
760 _itoa (convertme, &result[20], 10, 0));
761 free (expr);
762 return *word ? 0 : WRDE_NOSPACE;
764 expr = w_addchar (expr, &expr_length, &expr_maxlen, words[*offset]);
765 if (expr == NULL)
766 return WRDE_NOSPACE;
768 break;
770 case ']':
771 if (bracket && paren_depth == 1)
773 char result[21]; /* 21 = ceil(log10(2^64)) + 1 */
774 long int numresult = 0;
776 /* Go - evaluate. */
777 if (*expr && eval_expr (expr, &numresult) != 0)
779 free (expr);
780 return WRDE_SYNTAX;
783 result[20] = '\0';
784 *word = w_addstr (*word, word_length, max_length,
785 _itoa_word (numresult, &result[20], 10, 0));
786 free (expr);
787 return *word ? 0 : WRDE_NOSPACE;
790 free (expr);
791 return WRDE_SYNTAX;
793 case '\n':
794 case ';':
795 case '{':
796 case '}':
797 free (expr);
798 return WRDE_BADCHAR;
800 case '(':
801 ++paren_depth;
802 default:
803 expr = w_addchar (expr, &expr_length, &expr_maxlen, words[*offset]);
804 if (expr == NULL)
805 return WRDE_NOSPACE;
809 /* Premature end */
810 free (expr);
811 return WRDE_SYNTAX;
814 /* Function called by child process in exec_comm() */
815 static inline void
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? */
822 if (noexec)
823 args[1] = "-nc";
825 /* Redirect output. */
826 if (__glibc_likely (fildes[1] != STDOUT_FILENO))
828 __dup2 (fildes[1], STDOUT_FILENO);
829 __close (fildes[1]);
831 else
833 #ifdef O_CLOEXEC
834 /* Reset the close-on-exec flag (if necessary). */
835 # ifndef __ASSUME_PIPE2
836 if (__have_pipe2 > 0)
837 # endif
838 __fcntl (fildes[1], F_SETFD, 0);
839 #endif
842 /* Redirect stderr to /dev/null if we have to. */
843 if (showerr == 0)
845 struct stat64 st;
846 int fd;
847 __close (STDERR_FILENO);
848 fd = __open (_PATH_DEVNULL, O_WRONLY);
849 if (fd >= 0 && fd != STDERR_FILENO)
851 __dup2 (fd, STDERR_FILENO);
852 __close (fd);
854 /* Be paranoid. Check that we actually opened the /dev/null
855 device. */
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)
860 #endif
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. */
865 _exit (90);
868 /* Make sure the subshell doesn't field-split on our behalf. */
869 __unsetenv ("IFS");
871 __close (fildes[0]);
872 __execve (_PATH_BSHELL, (char *const *) args, __environ);
874 /* Bad. What now? */
875 abort ();
878 /* Function to execute a command and retrieve the results */
879 /* pwordexp contains NULL if field-splitting is forbidden */
880 static int
881 internal_function
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)
886 int fildes[2];
887 #define bufsize 128
888 int buflen;
889 int i;
890 int status = 0;
891 size_t maxnewlines = 0;
892 char buffer[bufsize];
893 pid_t pid;
894 int noexec = 0;
896 /* Do nothing if command substitution should not succeed. */
897 if (flags & WRDE_NOCMD)
898 return WRDE_CMDSUB;
900 /* Don't fork() unless necessary */
901 if (!comm || !*comm)
902 return 0;
904 #ifdef O_CLOEXEC
905 # ifndef __ASSUME_PIPE2
906 if (__have_pipe2 >= 0)
907 # endif
909 int r = __pipe2 (fildes, O_CLOEXEC);
910 # ifndef __ASSUME_PIPE2
911 if (__have_pipe2 == 0)
912 __have_pipe2 = r != -1 || errno != ENOSYS ? 1 : -1;
914 if (__have_pipe2 > 0)
915 # endif
916 if (r < 0)
917 /* Bad */
918 return WRDE_NOSPACE;
920 #endif
921 #ifndef __ASSUME_PIPE2
922 # ifdef O_CLOEXEC
923 if (__have_pipe2 < 0)
924 # endif
925 if (__pipe (fildes) < 0)
926 /* Bad */
927 return WRDE_NOSPACE;
928 #endif
930 again:
931 if ((pid = __fork ()) < 0)
933 /* Bad */
934 __close (fildes[0]);
935 __close (fildes[1]);
936 return WRDE_NOSPACE;
939 if (pid == 0)
940 exec_comm_child (comm, fildes, noexec ? 0 : flags & WRDE_SHOWERR, noexec);
942 /* Parent */
944 /* If we are just testing the syntax, only wait. */
945 if (noexec)
946 return (TEMP_FAILURE_RETRY (__waitpid (pid, &status, 0)) == pid
947 && status != 0) ? WRDE_SYNTAX : 0;
949 __close (fildes[1]);
950 fildes[1] = -1;
952 if (!pwordexp)
953 /* Quoted - no field splitting */
955 while (1)
957 if ((buflen = TEMP_FAILURE_RETRY (__read (fildes[0], buffer,
958 bufsize))) < 1)
960 /* If read returned 0 then the process has closed its
961 stdout. Don't use WNOHANG in that case to avoid busy
962 looping until the process eventually exits. */
963 if (TEMP_FAILURE_RETRY (__waitpid (pid, &status,
964 buflen == 0 ? 0 : WNOHANG))
965 == 0)
966 continue;
967 if ((buflen = TEMP_FAILURE_RETRY (__read (fildes[0], buffer,
968 bufsize))) < 1)
969 break;
972 maxnewlines += buflen;
974 *word = w_addmem (*word, word_length, max_length, buffer, buflen);
975 if (*word == NULL)
976 goto no_space;
979 else
980 /* Not quoted - split fields */
982 int copying = 0;
983 /* 'copying' is:
984 * 0 when searching for first character in a field not IFS white space
985 * 1 when copying the text of a field
986 * 2 when searching for possible non-whitespace IFS
987 * 3 when searching for non-newline after copying field
990 while (1)
992 if ((buflen = TEMP_FAILURE_RETRY (__read (fildes[0], buffer,
993 bufsize))) < 1)
995 /* If read returned 0 then the process has closed its
996 stdout. Don't use WNOHANG in that case to avoid busy
997 looping until the process eventually exits. */
998 if (TEMP_FAILURE_RETRY (__waitpid (pid, &status,
999 buflen == 0 ? 0 : WNOHANG))
1000 == 0)
1001 continue;
1002 if ((buflen = TEMP_FAILURE_RETRY (__read (fildes[0], buffer,
1003 bufsize))) < 1)
1004 break;
1007 for (i = 0; i < buflen; ++i)
1009 if (strchr (ifs, buffer[i]) != NULL)
1011 /* Current character is IFS */
1012 if (strchr (ifs_white, buffer[i]) == NULL)
1014 /* Current character is IFS but not whitespace */
1015 if (copying == 2)
1017 /* current character
1020 * eg: text<space><comma><space>moretext
1022 * So, strip whitespace IFS (like at the start)
1024 copying = 0;
1025 continue;
1028 copying = 0;
1029 /* fall through and delimit field.. */
1031 else
1033 if (buffer[i] == '\n')
1035 /* Current character is (IFS) newline */
1037 /* If copying a field, this is the end of it,
1038 but maybe all that's left is trailing newlines.
1039 So start searching for a non-newline. */
1040 if (copying == 1)
1041 copying = 3;
1043 continue;
1045 else
1047 /* Current character is IFS white space, but
1048 not a newline */
1050 /* If not either copying a field or searching
1051 for non-newline after a field, ignore it */
1052 if (copying != 1 && copying != 3)
1053 continue;
1055 /* End of field (search for non-ws IFS afterwards) */
1056 copying = 2;
1060 /* First IFS white space (non-newline), or IFS non-whitespace.
1061 * Delimit the field. Nulls are converted by w_addword. */
1062 if (w_addword (pwordexp, *word) == WRDE_NOSPACE)
1063 goto no_space;
1065 *word = w_newword (word_length, max_length);
1067 maxnewlines = 0;
1068 /* fall back round the loop.. */
1070 else
1072 /* Not IFS character */
1074 if (copying == 3)
1076 /* Nothing but (IFS) newlines since the last field,
1077 so delimit it here before starting new word */
1078 if (w_addword (pwordexp, *word) == WRDE_NOSPACE)
1079 goto no_space;
1081 *word = w_newword (word_length, max_length);
1084 copying = 1;
1086 if (buffer[i] == '\n') /* happens if newline not in IFS */
1087 maxnewlines++;
1088 else
1089 maxnewlines = 0;
1091 *word = w_addchar (*word, word_length, max_length,
1092 buffer[i]);
1093 if (*word == NULL)
1094 goto no_space;
1100 /* Chop off trailing newlines (required by POSIX.2) */
1101 /* Ensure we don't go back further than the beginning of the
1102 substitution (i.e. remove maxnewlines bytes at most) */
1103 while (maxnewlines-- != 0 &&
1104 *word_length > 0 && (*word)[*word_length - 1] == '\n')
1106 (*word)[--*word_length] = '\0';
1108 /* If the last word was entirely newlines, turn it into a new word
1109 * which can be ignored if there's nothing following it. */
1110 if (*word_length == 0)
1112 free (*word);
1113 *word = w_newword (word_length, max_length);
1114 break;
1118 __close (fildes[0]);
1119 fildes[0] = -1;
1121 /* Check for syntax error (re-execute but with "-n" flag) */
1122 if (buflen < 1 && status != 0)
1124 noexec = 1;
1125 goto again;
1128 return 0;
1130 no_space:
1131 __kill (pid, SIGKILL);
1132 TEMP_FAILURE_RETRY (__waitpid (pid, NULL, 0));
1133 __close (fildes[0]);
1134 return WRDE_NOSPACE;
1137 static int
1138 internal_function
1139 parse_comm (char **word, size_t *word_length, size_t *max_length,
1140 const char *words, size_t *offset, int flags, wordexp_t *pwordexp,
1141 const char *ifs, const char *ifs_white)
1143 /* We are poised just after "$(" */
1144 int paren_depth = 1;
1145 int error = 0;
1146 int quoted = 0; /* 1 for singly-quoted, 2 for doubly-quoted */
1147 size_t comm_length;
1148 size_t comm_maxlen;
1149 char *comm = w_newword (&comm_length, &comm_maxlen);
1151 for (; words[*offset]; ++(*offset))
1153 switch (words[*offset])
1155 case '\'':
1156 if (quoted == 0)
1157 quoted = 1;
1158 else if (quoted == 1)
1159 quoted = 0;
1161 break;
1163 case '"':
1164 if (quoted == 0)
1165 quoted = 2;
1166 else if (quoted == 2)
1167 quoted = 0;
1169 break;
1171 case ')':
1172 if (!quoted && --paren_depth == 0)
1174 /* Go -- give script to the shell */
1175 if (comm)
1177 #ifdef __libc_ptf_call
1178 /* We do not want the exec_comm call to be cut short
1179 by a thread cancellation since cleanup is very
1180 ugly. Therefore disable cancellation for
1181 now. */
1182 // XXX Ideally we do want the thread being cancelable.
1183 // XXX If demand is there we'll change it.
1184 int state = PTHREAD_CANCEL_ENABLE;
1185 __libc_ptf_call (pthread_setcancelstate,
1186 (PTHREAD_CANCEL_DISABLE, &state), 0);
1187 #endif
1189 error = exec_comm (comm, word, word_length, max_length,
1190 flags, pwordexp, ifs, ifs_white);
1192 #ifdef __libc_ptf_call
1193 __libc_ptf_call (pthread_setcancelstate, (state, NULL), 0);
1194 #endif
1196 free (comm);
1199 return error;
1202 /* This is just part of the script */
1203 break;
1205 case '(':
1206 if (!quoted)
1207 ++paren_depth;
1210 comm = w_addchar (comm, &comm_length, &comm_maxlen, words[*offset]);
1211 if (comm == NULL)
1212 return WRDE_NOSPACE;
1215 /* Premature end. */
1216 free (comm);
1218 return WRDE_SYNTAX;
1221 static int
1222 internal_function
1223 parse_param (char **word, size_t *word_length, size_t *max_length,
1224 const char *words, size_t *offset, int flags, wordexp_t *pwordexp,
1225 const char *ifs, const char *ifs_white, int quoted)
1227 /* We are poised just after "$" */
1228 enum action
1230 ACT_NONE,
1231 ACT_RP_SHORT_LEFT = '#',
1232 ACT_RP_LONG_LEFT = 'L',
1233 ACT_RP_SHORT_RIGHT = '%',
1234 ACT_RP_LONG_RIGHT = 'R',
1235 ACT_NULL_ERROR = '?',
1236 ACT_NULL_SUBST = '-',
1237 ACT_NONNULL_SUBST = '+',
1238 ACT_NULL_ASSIGN = '='
1240 size_t env_length;
1241 size_t env_maxlen;
1242 size_t pat_length;
1243 size_t pat_maxlen;
1244 size_t start = *offset;
1245 char *env;
1246 char *pattern;
1247 char *value = NULL;
1248 enum action action = ACT_NONE;
1249 int depth = 0;
1250 int colon_seen = 0;
1251 int seen_hash = 0;
1252 int free_value = 0;
1253 int pattern_is_quoted = 0; /* 1 for singly-quoted, 2 for doubly-quoted */
1254 int error;
1255 int special = 0;
1256 char buffer[21];
1257 int brace = words[*offset] == '{';
1259 env = w_newword (&env_length, &env_maxlen);
1260 pattern = w_newword (&pat_length, &pat_maxlen);
1262 if (brace)
1263 ++*offset;
1265 /* First collect the parameter name. */
1267 if (words[*offset] == '#')
1269 seen_hash = 1;
1270 if (!brace)
1271 goto envsubst;
1272 ++*offset;
1275 if (isalpha (words[*offset]) || words[*offset] == '_')
1277 /* Normal parameter name. */
1280 env = w_addchar (env, &env_length, &env_maxlen,
1281 words[*offset]);
1282 if (env == NULL)
1283 goto no_space;
1285 while (isalnum (words[++*offset]) || words[*offset] == '_');
1287 else if (isdigit (words[*offset]))
1289 /* Numeric parameter name. */
1290 special = 1;
1293 env = w_addchar (env, &env_length, &env_maxlen,
1294 words[*offset]);
1295 if (env == NULL)
1296 goto no_space;
1297 if (!brace)
1298 goto envsubst;
1300 while (isdigit(words[++*offset]));
1302 else if (strchr ("*@$", words[*offset]) != NULL)
1304 /* Special parameter. */
1305 special = 1;
1306 env = w_addchar (env, &env_length, &env_maxlen,
1307 words[*offset]);
1308 if (env == NULL)
1309 goto no_space;
1310 ++*offset;
1312 else
1314 if (brace)
1315 goto syntax;
1318 if (brace)
1320 /* Check for special action to be applied to the value. */
1321 switch (words[*offset])
1323 case '}':
1324 /* Evaluate. */
1325 goto envsubst;
1327 case '#':
1328 action = ACT_RP_SHORT_LEFT;
1329 if (words[1 + *offset] == '#')
1331 ++*offset;
1332 action = ACT_RP_LONG_LEFT;
1334 break;
1336 case '%':
1337 action = ACT_RP_SHORT_RIGHT;
1338 if (words[1 + *offset] == '%')
1340 ++*offset;
1341 action = ACT_RP_LONG_RIGHT;
1343 break;
1345 case ':':
1346 if (strchr ("-=?+", words[1 + *offset]) == NULL)
1347 goto syntax;
1349 colon_seen = 1;
1350 action = words[++*offset];
1351 break;
1353 case '-':
1354 case '=':
1355 case '?':
1356 case '+':
1357 action = words[*offset];
1358 break;
1360 default:
1361 goto syntax;
1364 /* Now collect the pattern, but don't expand it yet. */
1365 ++*offset;
1366 for (; words[*offset]; ++(*offset))
1368 switch (words[*offset])
1370 case '{':
1371 if (!pattern_is_quoted)
1372 ++depth;
1373 break;
1375 case '}':
1376 if (!pattern_is_quoted)
1378 if (depth == 0)
1379 goto envsubst;
1380 --depth;
1382 break;
1384 case '\\':
1385 if (pattern_is_quoted)
1386 /* Quoted; treat as normal character. */
1387 break;
1389 /* Otherwise, it's an escape: next character is literal. */
1390 if (words[++*offset] == '\0')
1391 goto syntax;
1393 pattern = w_addchar (pattern, &pat_length, &pat_maxlen, '\\');
1394 if (pattern == NULL)
1395 goto no_space;
1397 break;
1399 case '\'':
1400 if (pattern_is_quoted == 0)
1401 pattern_is_quoted = 1;
1402 else if (pattern_is_quoted == 1)
1403 pattern_is_quoted = 0;
1405 break;
1407 case '"':
1408 if (pattern_is_quoted == 0)
1409 pattern_is_quoted = 2;
1410 else if (pattern_is_quoted == 2)
1411 pattern_is_quoted = 0;
1413 break;
1416 pattern = w_addchar (pattern, &pat_length, &pat_maxlen,
1417 words[*offset]);
1418 if (pattern == NULL)
1419 goto no_space;
1423 /* End of input string -- remember to reparse the character that we
1424 * stopped at. */
1425 --(*offset);
1427 envsubst:
1428 if (words[start] == '{' && words[*offset] != '}')
1429 goto syntax;
1431 if (env == NULL)
1433 if (seen_hash)
1435 /* $# expands to the number of positional parameters */
1436 buffer[20] = '\0';
1437 value = _itoa_word (__libc_argc - 1, &buffer[20], 10, 0);
1438 seen_hash = 0;
1440 else
1442 /* Just $ on its own */
1443 *offset = start - 1;
1444 *word = w_addchar (*word, word_length, max_length, '$');
1445 return *word ? 0 : WRDE_NOSPACE;
1448 /* Is it a numeric parameter? */
1449 else if (isdigit (env[0]))
1451 int n = atoi (env);
1453 if (n >= __libc_argc)
1454 /* Substitute NULL. */
1455 value = NULL;
1456 else
1457 /* Replace with appropriate positional parameter. */
1458 value = __libc_argv[n];
1460 /* Is it a special parameter? */
1461 else if (special)
1463 /* Is it `$$'? */
1464 if (*env == '$')
1466 buffer[20] = '\0';
1467 value = _itoa_word (__getpid (), &buffer[20], 10, 0);
1469 /* Is it `${#*}' or `${#@}'? */
1470 else if ((*env == '*' || *env == '@') && seen_hash)
1472 buffer[20] = '\0';
1473 value = _itoa_word (__libc_argc > 0 ? __libc_argc - 1 : 0,
1474 &buffer[20], 10, 0);
1475 *word = w_addstr (*word, word_length, max_length, value);
1476 free (env);
1477 free (pattern);
1478 return *word ? 0 : WRDE_NOSPACE;
1480 /* Is it `$*' or `$@' (unquoted) ? */
1481 else if (*env == '*' || (*env == '@' && !quoted))
1483 size_t plist_len = 0;
1484 int p;
1485 char *end;
1487 /* Build up value parameter by parameter (copy them) */
1488 for (p = 1; __libc_argv[p]; ++p)
1489 plist_len += strlen (__libc_argv[p]) + 1; /* for space */
1490 value = malloc (plist_len);
1491 if (value == NULL)
1492 goto no_space;
1493 end = value;
1494 *end = 0;
1495 for (p = 1; __libc_argv[p]; ++p)
1497 if (p > 1)
1498 *end++ = ' ';
1499 end = __stpcpy (end, __libc_argv[p]);
1502 free_value = 1;
1504 else
1506 /* Must be a quoted `$@' */
1507 assert (*env == '@' && quoted);
1509 /* Each parameter is a separate word ("$@") */
1510 if (__libc_argc == 2)
1511 value = __libc_argv[1];
1512 else if (__libc_argc > 2)
1514 int p;
1516 /* Append first parameter to current word. */
1517 value = w_addstr (*word, word_length, max_length,
1518 __libc_argv[1]);
1519 if (value == NULL || w_addword (pwordexp, value))
1520 goto no_space;
1522 for (p = 2; __libc_argv[p + 1]; p++)
1524 char *newword = __strdup (__libc_argv[p]);
1525 if (newword == NULL || w_addword (pwordexp, newword))
1526 goto no_space;
1529 /* Start a new word with the last parameter. */
1530 *word = w_newword (word_length, max_length);
1531 value = __libc_argv[p];
1533 else
1535 free (env);
1536 free (pattern);
1537 return 0;
1541 else
1542 value = getenv (env);
1544 if (value == NULL && (flags & WRDE_UNDEF))
1546 /* Variable not defined. */
1547 error = WRDE_BADVAL;
1548 goto do_error;
1551 if (action != ACT_NONE)
1553 int expand_pattern = 0;
1555 /* First, find out if we need to expand pattern (i.e. if we will
1556 * use it). */
1557 switch (action)
1559 case ACT_RP_SHORT_LEFT:
1560 case ACT_RP_LONG_LEFT:
1561 case ACT_RP_SHORT_RIGHT:
1562 case ACT_RP_LONG_RIGHT:
1563 /* Always expand for these. */
1564 expand_pattern = 1;
1565 break;
1567 case ACT_NULL_ERROR:
1568 case ACT_NULL_SUBST:
1569 case ACT_NULL_ASSIGN:
1570 if (!value || (!*value && colon_seen))
1571 /* If param is unset, or set but null and a colon has been seen,
1572 the expansion of the pattern will be needed. */
1573 expand_pattern = 1;
1575 break;
1577 case ACT_NONNULL_SUBST:
1578 /* Expansion of word will be needed if parameter is set and not null,
1579 or set null but no colon has been seen. */
1580 if (value && (*value || !colon_seen))
1581 expand_pattern = 1;
1583 break;
1585 default:
1586 assert (! "Unrecognised action!");
1589 if (expand_pattern)
1591 /* We need to perform tilde expansion, parameter expansion,
1592 command substitution, and arithmetic expansion. We also
1593 have to be a bit careful with wildcard characters, as
1594 pattern might be given to fnmatch soon. To do this, we
1595 convert quotes to escapes. */
1597 char *expanded;
1598 size_t exp_len;
1599 size_t exp_maxl;
1600 char *p;
1601 int quoted = 0; /* 1: single quotes; 2: double */
1603 expanded = w_newword (&exp_len, &exp_maxl);
1604 for (p = pattern; p && *p; p++)
1606 size_t offset;
1608 switch (*p)
1610 case '"':
1611 if (quoted == 2)
1612 quoted = 0;
1613 else if (quoted == 0)
1614 quoted = 2;
1615 else break;
1617 continue;
1619 case '\'':
1620 if (quoted == 1)
1621 quoted = 0;
1622 else if (quoted == 0)
1623 quoted = 1;
1624 else break;
1626 continue;
1628 case '*':
1629 case '?':
1630 if (quoted)
1632 /* Convert quoted wildchar to escaped wildchar. */
1633 expanded = w_addchar (expanded, &exp_len,
1634 &exp_maxl, '\\');
1636 if (expanded == NULL)
1637 goto no_space;
1639 break;
1641 case '$':
1642 offset = 0;
1643 error = parse_dollars (&expanded, &exp_len, &exp_maxl, p,
1644 &offset, flags, NULL, NULL, NULL, 1);
1645 if (error)
1647 if (free_value)
1648 free (value);
1650 free (expanded);
1652 goto do_error;
1655 p += offset;
1656 continue;
1658 case '~':
1659 if (quoted || exp_len)
1660 break;
1662 offset = 0;
1663 error = parse_tilde (&expanded, &exp_len, &exp_maxl, p,
1664 &offset, 0);
1665 if (error)
1667 if (free_value)
1668 free (value);
1670 free (expanded);
1672 goto do_error;
1675 p += offset;
1676 continue;
1678 case '\\':
1679 expanded = w_addchar (expanded, &exp_len, &exp_maxl, '\\');
1680 ++p;
1681 assert (*p); /* checked when extracted initially */
1682 if (expanded == NULL)
1683 goto no_space;
1686 expanded = w_addchar (expanded, &exp_len, &exp_maxl, *p);
1688 if (expanded == NULL)
1689 goto no_space;
1692 free (pattern);
1694 pattern = expanded;
1697 switch (action)
1699 case ACT_RP_SHORT_LEFT:
1700 case ACT_RP_LONG_LEFT:
1701 case ACT_RP_SHORT_RIGHT:
1702 case ACT_RP_LONG_RIGHT:
1704 char *p;
1705 char c;
1706 char *end;
1708 if (value == NULL || pattern == NULL || *pattern == '\0')
1709 break;
1711 end = value + strlen (value);
1713 switch (action)
1715 case ACT_RP_SHORT_LEFT:
1716 for (p = value; p <= end; ++p)
1718 c = *p;
1719 *p = '\0';
1720 if (fnmatch (pattern, value, 0) != FNM_NOMATCH)
1722 *p = c;
1723 if (free_value)
1725 char *newval = __strdup (p);
1726 if (newval == NULL)
1728 free (value);
1729 goto no_space;
1731 free (value);
1732 value = newval;
1734 else
1735 value = p;
1736 break;
1738 *p = c;
1741 break;
1743 case ACT_RP_LONG_LEFT:
1744 for (p = end; p >= value; --p)
1746 c = *p;
1747 *p = '\0';
1748 if (fnmatch (pattern, value, 0) != FNM_NOMATCH)
1750 *p = c;
1751 if (free_value)
1753 char *newval = __strdup (p);
1754 if (newval == NULL)
1756 free (value);
1757 goto no_space;
1759 free (value);
1760 value = newval;
1762 else
1763 value = p;
1764 break;
1766 *p = c;
1769 break;
1771 case ACT_RP_SHORT_RIGHT:
1772 for (p = end; p >= value; --p)
1774 if (fnmatch (pattern, p, 0) != FNM_NOMATCH)
1776 char *newval;
1777 newval = malloc (p - value + 1);
1779 if (newval == NULL)
1781 if (free_value)
1782 free (value);
1783 goto no_space;
1786 *(char *) __mempcpy (newval, value, p - value) = '\0';
1787 if (free_value)
1788 free (value);
1789 value = newval;
1790 free_value = 1;
1791 break;
1795 break;
1797 case ACT_RP_LONG_RIGHT:
1798 for (p = value; p <= end; ++p)
1800 if (fnmatch (pattern, p, 0) != FNM_NOMATCH)
1802 char *newval;
1803 newval = malloc (p - value + 1);
1805 if (newval == NULL)
1807 if (free_value)
1808 free (value);
1809 goto no_space;
1812 *(char *) __mempcpy (newval, value, p - value) = '\0';
1813 if (free_value)
1814 free (value);
1815 value = newval;
1816 free_value = 1;
1817 break;
1821 break;
1823 default:
1824 break;
1827 break;
1830 case ACT_NULL_ERROR:
1831 if (value && *value)
1832 /* Substitute parameter */
1833 break;
1835 error = 0;
1836 if (!colon_seen && value)
1837 /* Substitute NULL */
1839 else
1841 const char *str = pattern;
1843 if (str[0] == '\0')
1844 str = _("parameter null or not set");
1846 __fxprintf (NULL, "%s: %s\n", env, str);
1849 if (free_value)
1850 free (value);
1851 goto do_error;
1853 case ACT_NULL_SUBST:
1854 if (value && *value)
1855 /* Substitute parameter */
1856 break;
1858 if (free_value)
1859 free (value);
1861 if (!colon_seen && value)
1862 /* Substitute NULL */
1863 goto success;
1865 value = pattern ? __strdup (pattern) : pattern;
1866 free_value = 1;
1868 if (pattern && !value)
1869 goto no_space;
1871 break;
1873 case ACT_NONNULL_SUBST:
1874 if (value && (*value || !colon_seen))
1876 if (free_value)
1877 free (value);
1879 value = pattern ? __strdup (pattern) : pattern;
1880 free_value = 1;
1882 if (pattern && !value)
1883 goto no_space;
1885 break;
1888 /* Substitute NULL */
1889 if (free_value)
1890 free (value);
1891 goto success;
1893 case ACT_NULL_ASSIGN:
1894 if (value && *value)
1895 /* Substitute parameter */
1896 break;
1898 if (!colon_seen && value)
1900 /* Substitute NULL */
1901 if (free_value)
1902 free (value);
1903 goto success;
1906 if (free_value)
1907 free (value);
1909 value = pattern ? __strdup (pattern) : pattern;
1910 free_value = 1;
1912 if (pattern && !value)
1913 goto no_space;
1915 __setenv (env, value, 1);
1916 break;
1918 default:
1919 assert (! "Unrecognised action!");
1923 free (env);
1924 env = NULL;
1925 free (pattern);
1926 pattern = NULL;
1928 if (seen_hash)
1930 char param_length[21];
1931 param_length[20] = '\0';
1932 *word = w_addstr (*word, word_length, max_length,
1933 _itoa_word (value ? strlen (value) : 0,
1934 &param_length[20], 10, 0));
1935 if (free_value)
1937 assert (value != NULL);
1938 free (value);
1941 return *word ? 0 : WRDE_NOSPACE;
1944 if (value == NULL)
1945 return 0;
1947 if (quoted || !pwordexp)
1949 /* Quoted - no field split */
1950 *word = w_addstr (*word, word_length, max_length, value);
1951 if (free_value)
1952 free (value);
1954 return *word ? 0 : WRDE_NOSPACE;
1956 else
1958 /* Need to field-split */
1959 char *value_copy = __strdup (value); /* Don't modify value */
1960 char *field_begin = value_copy;
1961 int seen_nonws_ifs = 0;
1963 if (free_value)
1964 free (value);
1966 if (value_copy == NULL)
1967 goto no_space;
1971 char *field_end = field_begin;
1972 char *next_field;
1974 /* If this isn't the first field, start a new word */
1975 if (field_begin != value_copy)
1977 if (w_addword (pwordexp, *word) == WRDE_NOSPACE)
1979 free (value_copy);
1980 goto no_space;
1983 *word = w_newword (word_length, max_length);
1986 /* Skip IFS whitespace before the field */
1987 field_begin += strspn (field_begin, ifs_white);
1989 if (!seen_nonws_ifs && *field_begin == 0)
1990 /* Nothing but whitespace */
1991 break;
1993 /* Search for the end of the field */
1994 field_end = field_begin + strcspn (field_begin, ifs);
1996 /* Set up pointer to the character after end of field and
1997 skip whitespace IFS after it. */
1998 next_field = field_end + strspn (field_end, ifs_white);
2000 /* Skip at most one non-whitespace IFS character after the field */
2001 seen_nonws_ifs = 0;
2002 if (*next_field && strchr (ifs, *next_field))
2004 seen_nonws_ifs = 1;
2005 next_field++;
2008 /* Null-terminate it */
2009 *field_end = 0;
2011 /* Tag a copy onto the current word */
2012 *word = w_addstr (*word, word_length, max_length, field_begin);
2014 if (*word == NULL && *field_begin != '\0')
2016 free (value_copy);
2017 goto no_space;
2020 field_begin = next_field;
2022 while (seen_nonws_ifs || *field_begin);
2024 free (value_copy);
2027 return 0;
2029 success:
2030 error = 0;
2031 goto do_error;
2033 no_space:
2034 error = WRDE_NOSPACE;
2035 goto do_error;
2037 syntax:
2038 error = WRDE_SYNTAX;
2040 do_error:
2041 free (env);
2043 free (pattern);
2045 return error;
2048 static int
2049 internal_function
2050 parse_dollars (char **word, size_t *word_length, size_t *max_length,
2051 const char *words, size_t *offset, int flags,
2052 wordexp_t *pwordexp, const char *ifs, const char *ifs_white,
2053 int quoted)
2055 /* We are poised _at_ "$" */
2056 switch (words[1 + *offset])
2058 case '"':
2059 case '\'':
2060 case 0:
2061 *word = w_addchar (*word, word_length, max_length, '$');
2062 return *word ? 0 : WRDE_NOSPACE;
2064 case '(':
2065 if (words[2 + *offset] == '(')
2067 /* Differentiate between $((1+3)) and $((echo);(ls)) */
2068 int i = 3 + *offset;
2069 int depth = 0;
2070 while (words[i] && !(depth == 0 && words[i] == ')'))
2072 if (words[i] == '(')
2073 ++depth;
2074 else if (words[i] == ')')
2075 --depth;
2077 ++i;
2080 if (words[i] == ')' && words[i + 1] == ')')
2082 (*offset) += 3;
2083 /* Call parse_arith -- 0 is for "no brackets" */
2084 return parse_arith (word, word_length, max_length, words, offset,
2085 flags, 0);
2089 (*offset) += 2;
2090 return parse_comm (word, word_length, max_length, words, offset, flags,
2091 quoted? NULL : pwordexp, ifs, ifs_white);
2093 case '[':
2094 (*offset) += 2;
2095 /* Call parse_arith -- 1 is for "brackets" */
2096 return parse_arith (word, word_length, max_length, words, offset, flags,
2099 case '{':
2100 default:
2101 ++(*offset); /* parse_param needs to know if "{" is there */
2102 return parse_param (word, word_length, max_length, words, offset, flags,
2103 pwordexp, ifs, ifs_white, quoted);
2107 static int
2108 internal_function
2109 parse_backtick (char **word, size_t *word_length, size_t *max_length,
2110 const char *words, size_t *offset, int flags,
2111 wordexp_t *pwordexp, const char *ifs, const char *ifs_white)
2113 /* We are poised just after "`" */
2114 int error;
2115 int squoting = 0;
2116 size_t comm_length;
2117 size_t comm_maxlen;
2118 char *comm = w_newword (&comm_length, &comm_maxlen);
2120 for (; words[*offset]; ++(*offset))
2122 switch (words[*offset])
2124 case '`':
2125 /* Go -- give the script to the shell */
2126 error = exec_comm (comm, word, word_length, max_length, flags,
2127 pwordexp, ifs, ifs_white);
2128 free (comm);
2129 return error;
2131 case '\\':
2132 if (squoting)
2134 error = parse_qtd_backslash (&comm, &comm_length, &comm_maxlen,
2135 words, offset);
2137 if (error)
2139 free (comm);
2140 return error;
2143 break;
2146 ++(*offset);
2147 error = parse_backslash (&comm, &comm_length, &comm_maxlen, words,
2148 offset);
2150 if (error)
2152 free (comm);
2153 return error;
2156 break;
2158 case '\'':
2159 squoting = 1 - squoting;
2160 default:
2161 comm = w_addchar (comm, &comm_length, &comm_maxlen, words[*offset]);
2162 if (comm == NULL)
2163 return WRDE_NOSPACE;
2167 /* Premature end */
2168 free (comm);
2169 return WRDE_SYNTAX;
2172 static int
2173 internal_function
2174 parse_dquote (char **word, size_t *word_length, size_t *max_length,
2175 const char *words, size_t *offset, int flags,
2176 wordexp_t *pwordexp, const char * ifs, const char * ifs_white)
2178 /* We are poised just after a double-quote */
2179 int error;
2181 for (; words[*offset]; ++(*offset))
2183 switch (words[*offset])
2185 case '"':
2186 return 0;
2188 case '$':
2189 error = parse_dollars (word, word_length, max_length, words, offset,
2190 flags, pwordexp, ifs, ifs_white, 1);
2191 /* The ``1'' here is to tell parse_dollars not to
2192 * split the fields. It may need to, however ("$@").
2194 if (error)
2195 return error;
2197 break;
2199 case '`':
2200 ++(*offset);
2201 error = parse_backtick (word, word_length, max_length, words,
2202 offset, flags, NULL, NULL, NULL);
2203 /* The first NULL here is to tell parse_backtick not to
2204 * split the fields.
2206 if (error)
2207 return error;
2209 break;
2211 case '\\':
2212 error = parse_qtd_backslash (word, word_length, max_length, words,
2213 offset);
2215 if (error)
2216 return error;
2218 break;
2220 default:
2221 *word = w_addchar (*word, word_length, max_length, words[*offset]);
2222 if (*word == NULL)
2223 return WRDE_NOSPACE;
2227 /* Unterminated string */
2228 return WRDE_SYNTAX;
2232 * wordfree() is to be called after pwordexp is finished with.
2235 void
2236 wordfree (wordexp_t *pwordexp)
2239 /* wordexp can set pwordexp to NULL */
2240 if (pwordexp && pwordexp->we_wordv)
2242 char **wordv = pwordexp->we_wordv;
2244 for (wordv += pwordexp->we_offs; *wordv; ++wordv)
2245 free (*wordv);
2247 free (pwordexp->we_wordv);
2248 pwordexp->we_wordv = NULL;
2251 libc_hidden_def (wordfree)
2254 * wordexp()
2258 wordexp (const char *words, wordexp_t *pwordexp, int flags)
2260 size_t words_offset;
2261 size_t word_length;
2262 size_t max_length;
2263 char *word = w_newword (&word_length, &max_length);
2264 int error;
2265 char *ifs;
2266 char ifs_white[4];
2267 wordexp_t old_word = *pwordexp;
2269 if (flags & WRDE_REUSE)
2271 /* Minimal implementation of WRDE_REUSE for now */
2272 wordfree (pwordexp);
2273 old_word.we_wordv = NULL;
2276 if ((flags & WRDE_APPEND) == 0)
2278 pwordexp->we_wordc = 0;
2280 if (flags & WRDE_DOOFFS)
2282 pwordexp->we_wordv = calloc (1 + pwordexp->we_offs, sizeof (char *));
2283 if (pwordexp->we_wordv == NULL)
2285 error = WRDE_NOSPACE;
2286 goto do_error;
2289 else
2291 pwordexp->we_wordv = calloc (1, sizeof (char *));
2292 if (pwordexp->we_wordv == NULL)
2294 error = WRDE_NOSPACE;
2295 goto do_error;
2298 pwordexp->we_offs = 0;
2302 /* Find out what the field separators are.
2303 * There are two types: whitespace and non-whitespace.
2305 ifs = getenv ("IFS");
2307 if (ifs == NULL)
2308 /* IFS unset - use <space><tab><newline>. */
2309 ifs = strcpy (ifs_white, " \t\n");
2310 else
2312 char *ifsch = ifs;
2313 char *whch = ifs_white;
2315 while (*ifsch != '\0')
2317 if (*ifsch == ' ' || *ifsch == '\t' || *ifsch == '\n')
2319 /* Whitespace IFS. See first whether it is already in our
2320 collection. */
2321 char *runp = ifs_white;
2323 while (runp < whch && *runp != *ifsch)
2324 ++runp;
2326 if (runp == whch)
2327 *whch++ = *ifsch;
2330 ++ifsch;
2332 *whch = '\0';
2335 for (words_offset = 0 ; words[words_offset] ; ++words_offset)
2336 switch (words[words_offset])
2338 case '\\':
2339 error = parse_backslash (&word, &word_length, &max_length, words,
2340 &words_offset);
2342 if (error)
2343 goto do_error;
2345 break;
2347 case '$':
2348 error = parse_dollars (&word, &word_length, &max_length, words,
2349 &words_offset, flags, pwordexp, ifs, ifs_white,
2352 if (error)
2353 goto do_error;
2355 break;
2357 case '`':
2358 ++words_offset;
2359 error = parse_backtick (&word, &word_length, &max_length, words,
2360 &words_offset, flags, pwordexp, ifs,
2361 ifs_white);
2363 if (error)
2364 goto do_error;
2366 break;
2368 case '"':
2369 ++words_offset;
2370 error = parse_dquote (&word, &word_length, &max_length, words,
2371 &words_offset, flags, pwordexp, ifs, ifs_white);
2373 if (error)
2374 goto do_error;
2376 if (!word_length)
2378 error = w_addword (pwordexp, NULL);
2380 if (error)
2381 return error;
2384 break;
2386 case '\'':
2387 ++words_offset;
2388 error = parse_squote (&word, &word_length, &max_length, words,
2389 &words_offset);
2391 if (error)
2392 goto do_error;
2394 if (!word_length)
2396 error = w_addword (pwordexp, NULL);
2398 if (error)
2399 return error;
2402 break;
2404 case '~':
2405 error = parse_tilde (&word, &word_length, &max_length, words,
2406 &words_offset, pwordexp->we_wordc);
2408 if (error)
2409 goto do_error;
2411 break;
2413 case '*':
2414 case '[':
2415 case '?':
2416 error = parse_glob (&word, &word_length, &max_length, words,
2417 &words_offset, flags, pwordexp, ifs, ifs_white);
2419 if (error)
2420 goto do_error;
2422 break;
2424 default:
2425 /* Is it a word separator? */
2426 if (strchr (" \t", words[words_offset]) == NULL)
2428 char ch = words[words_offset];
2430 /* Not a word separator -- but is it a valid word char? */
2431 if (strchr ("\n|&;<>(){}", ch))
2433 /* Fail */
2434 error = WRDE_BADCHAR;
2435 goto do_error;
2438 /* "Ordinary" character -- add it to word */
2439 word = w_addchar (word, &word_length, &max_length,
2440 ch);
2441 if (word == NULL)
2443 error = WRDE_NOSPACE;
2444 goto do_error;
2447 break;
2450 /* If a word has been delimited, add it to the list. */
2451 if (word != NULL)
2453 error = w_addword (pwordexp, word);
2454 if (error)
2455 goto do_error;
2458 word = w_newword (&word_length, &max_length);
2461 /* End of string */
2463 /* There was a word separator at the end */
2464 if (word == NULL) /* i.e. w_newword */
2465 return 0;
2467 /* There was no field separator at the end */
2468 return w_addword (pwordexp, word);
2470 do_error:
2471 /* Error:
2472 * free memory used (unless error is WRDE_NOSPACE), and
2473 * set pwordexp members back to what they were.
2476 free (word);
2478 if (error == WRDE_NOSPACE)
2479 return WRDE_NOSPACE;
2481 if ((flags & WRDE_APPEND) == 0)
2482 wordfree (pwordexp);
2484 *pwordexp = old_word;
2485 return error;