]>
Commit | Line | Data |
---|---|---|
e192f341 CR |
1 | /* subst.c -- The part of the shell that does parameter, command, arithmetic, |
2 | and globbing substitutions. */ | |
3 | ||
4 | /* ``Have a little faith, there's magic in the night. You ain't a | |
5 | beauty, but, hey, you're alright.'' */ | |
6 | ||
7 | /* Copyright (C) 1987-2010 Free Software Foundation, Inc. | |
8 | ||
9 | This file is part of GNU Bash, the Bourne Again SHell. | |
10 | ||
11 | Bash is free software: you can redistribute it and/or modify | |
12 | it under the terms of the GNU General Public License as published by | |
13 | the Free Software Foundation, either version 3 of the License, or | |
14 | (at your option) any later version. | |
15 | ||
16 | Bash is distributed in the hope that it will be useful, | |
17 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
18 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
19 | GNU General Public License for more details. | |
20 | ||
21 | You should have received a copy of the GNU General Public License | |
22 | along with Bash. If not, see <http://www.gnu.org/licenses/>. | |
23 | */ | |
24 | ||
25 | #include "config.h" | |
26 | ||
27 | #include "bashtypes.h" | |
28 | #include <stdio.h> | |
29 | #include "chartypes.h" | |
30 | #if defined (HAVE_PWD_H) | |
31 | # include <pwd.h> | |
32 | #endif | |
33 | #include <signal.h> | |
34 | #include <errno.h> | |
35 | ||
36 | #if defined (HAVE_UNISTD_H) | |
37 | # include <unistd.h> | |
38 | #endif | |
39 | ||
40 | #include "bashansi.h" | |
41 | #include "posixstat.h" | |
42 | #include "bashintl.h" | |
43 | ||
44 | #include "shell.h" | |
45 | #include "parser.h" | |
46 | #include "flags.h" | |
47 | #include "jobs.h" | |
48 | #include "execute_cmd.h" | |
49 | #include "filecntl.h" | |
50 | #include "trap.h" | |
51 | #include "pathexp.h" | |
52 | #include "mailcheck.h" | |
53 | ||
54 | #include "shmbutil.h" | |
55 | #include "typemax.h" | |
56 | ||
57 | #include "builtins/getopt.h" | |
58 | #include "builtins/common.h" | |
59 | ||
60 | #include "builtins/builtext.h" | |
61 | ||
62 | #include <tilde/tilde.h> | |
63 | #include <glob/strmatch.h> | |
64 | ||
65 | #if !defined (errno) | |
66 | extern int errno; | |
67 | #endif /* !errno */ | |
68 | ||
69 | /* The size that strings change by. */ | |
70 | #define DEFAULT_INITIAL_ARRAY_SIZE 112 | |
71 | #define DEFAULT_ARRAY_SIZE 128 | |
72 | ||
73 | /* Variable types. */ | |
74 | #define VT_VARIABLE 0 | |
75 | #define VT_POSPARMS 1 | |
76 | #define VT_ARRAYVAR 2 | |
77 | #define VT_ARRAYMEMBER 3 | |
78 | #define VT_ASSOCVAR 4 | |
79 | ||
80 | #define VT_STARSUB 128 /* $* or ${array[*]} -- used to split */ | |
81 | ||
82 | /* Flags for quoted_strchr */ | |
83 | #define ST_BACKSL 0x01 | |
84 | #define ST_CTLESC 0x02 | |
85 | #define ST_SQUOTE 0x04 /* unused yet */ | |
86 | #define ST_DQUOTE 0x08 /* unused yet */ | |
87 | ||
88 | /* Flags for the `pflags' argument to param_expand() */ | |
89 | #define PF_NOCOMSUB 0x01 /* Do not perform command substitution */ | |
90 | #define PF_IGNUNBOUND 0x02 /* ignore unbound vars even if -u set */ | |
91 | #define PF_NOSPLIT2 0x04 /* same as W_NOSPLIT2 */ | |
92 | ||
93 | /* These defs make it easier to use the editor. */ | |
94 | #define LBRACE '{' | |
95 | #define RBRACE '}' | |
96 | #define LPAREN '(' | |
97 | #define RPAREN ')' | |
98 | ||
99 | #if defined (HANDLE_MULTIBYTE) | |
100 | #define WLPAREN L'(' | |
101 | #define WRPAREN L')' | |
102 | #endif | |
103 | ||
104 | /* Evaluates to 1 if C is one of the shell's special parameters whose length | |
105 | can be taken, but is also one of the special expansion characters. */ | |
106 | #define VALID_SPECIAL_LENGTH_PARAM(c) \ | |
107 | ((c) == '-' || (c) == '?' || (c) == '#') | |
108 | ||
109 | /* Evaluates to 1 if C is one of the shell's special parameters for which an | |
110 | indirect variable reference may be made. */ | |
111 | #define VALID_INDIR_PARAM(c) \ | |
112 | ((posixly_correct == 0 && (c) == '#') || (posixly_correct == 0 && (c) == '?') || (c) == '@' || (c) == '*') | |
113 | ||
114 | /* Evaluates to 1 if C is one of the OP characters that follows the parameter | |
115 | in ${parameter[:]OPword}. */ | |
116 | #define VALID_PARAM_EXPAND_CHAR(c) (sh_syntaxtab[(unsigned char)c] & CSUBSTOP) | |
117 | ||
118 | /* Evaluates to 1 if this is one of the shell's special variables. */ | |
119 | #define SPECIAL_VAR(name, wi) \ | |
120 | ((DIGIT (*name) && all_digits (name)) || \ | |
121 | (name[1] == '\0' && (sh_syntaxtab[(unsigned char)*name] & CSPECVAR)) || \ | |
122 | (wi && name[2] == '\0' && VALID_INDIR_PARAM (name[1]))) | |
123 | ||
124 | /* An expansion function that takes a string and a quoted flag and returns | |
125 | a WORD_LIST *. Used as the type of the third argument to | |
126 | expand_string_if_necessary(). */ | |
127 | typedef WORD_LIST *EXPFUNC __P((char *, int)); | |
128 | ||
129 | /* Process ID of the last command executed within command substitution. */ | |
130 | pid_t last_command_subst_pid = NO_PID; | |
131 | pid_t current_command_subst_pid = NO_PID; | |
132 | ||
133 | /* Variables used to keep track of the characters in IFS. */ | |
134 | SHELL_VAR *ifs_var; | |
135 | char *ifs_value; | |
136 | unsigned char ifs_cmap[UCHAR_MAX + 1]; | |
137 | ||
138 | #if defined (HANDLE_MULTIBYTE) | |
139 | unsigned char ifs_firstc[MB_LEN_MAX]; | |
140 | size_t ifs_firstc_len; | |
141 | #else | |
142 | unsigned char ifs_firstc; | |
143 | #endif | |
144 | ||
145 | /* Sentinel to tell when we are performing variable assignments preceding a | |
146 | command name and putting them into the environment. Used to make sure | |
147 | we use the temporary environment when looking up variable values. */ | |
148 | int assigning_in_environment; | |
149 | ||
150 | /* Used to hold a list of variable assignments preceding a command. Global | |
151 | so the SIGCHLD handler in jobs.c can unwind-protect it when it runs a | |
152 | SIGCHLD trap and so it can be saved and restored by the trap handlers. */ | |
153 | WORD_LIST *subst_assign_varlist = (WORD_LIST *)NULL; | |
154 | ||
155 | /* Extern functions and variables from different files. */ | |
156 | extern int last_command_exit_value, last_command_exit_signal; | |
157 | extern int subshell_environment, line_number; | |
158 | extern int subshell_level, parse_and_execute_level, sourcelevel; | |
159 | extern int eof_encountered; | |
160 | extern int return_catch_flag, return_catch_value; | |
161 | extern pid_t dollar_dollar_pid; | |
162 | extern int posixly_correct; | |
163 | extern char *this_command_name; | |
164 | extern struct fd_bitmap *current_fds_to_close; | |
165 | extern int wordexp_only; | |
166 | extern int expanding_redir; | |
167 | extern int tempenv_assign_error; | |
168 | ||
169 | #if !defined (HAVE_WCSDUP) && defined (HANDLE_MULTIBYTE) | |
170 | extern wchar_t *wcsdup __P((const wchar_t *)); | |
171 | #endif | |
172 | ||
173 | /* Non-zero means to allow unmatched globbed filenames to expand to | |
174 | a null file. */ | |
175 | int allow_null_glob_expansion; | |
176 | ||
177 | /* Non-zero means to throw an error when globbing fails to match anything. */ | |
178 | int fail_glob_expansion; | |
179 | ||
180 | #if 0 | |
181 | /* Variables to keep track of which words in an expanded word list (the | |
182 | output of expand_word_list_internal) are the result of globbing | |
183 | expansions. GLOB_ARGV_FLAGS is used by execute_cmd.c. | |
184 | (CURRENTLY UNUSED). */ | |
185 | char *glob_argv_flags; | |
186 | static int glob_argv_flags_size; | |
187 | #endif | |
188 | ||
189 | static WORD_LIST expand_word_error, expand_word_fatal; | |
190 | static WORD_DESC expand_wdesc_error, expand_wdesc_fatal; | |
191 | static char expand_param_error, expand_param_fatal; | |
192 | static char extract_string_error, extract_string_fatal; | |
193 | ||
194 | /* Tell the expansion functions to not longjmp back to top_level on fatal | |
195 | errors. Enabled when doing completion and prompt string expansion. */ | |
196 | static int no_longjmp_on_fatal_error = 0; | |
197 | ||
198 | /* Set by expand_word_unsplit; used to inhibit splitting and re-joining | |
199 | $* on $IFS, primarily when doing assignment statements. */ | |
200 | static int expand_no_split_dollar_star = 0; | |
201 | ||
202 | /* A WORD_LIST of words to be expanded by expand_word_list_internal, | |
203 | without any leading variable assignments. */ | |
204 | static WORD_LIST *garglist = (WORD_LIST *)NULL; | |
205 | ||
206 | static char *quoted_substring __P((char *, int, int)); | |
207 | static int quoted_strlen __P((char *)); | |
208 | static char *quoted_strchr __P((char *, int, int)); | |
209 | ||
210 | static char *expand_string_if_necessary __P((char *, int, EXPFUNC *)); | |
211 | static inline char *expand_string_to_string_internal __P((char *, int, EXPFUNC *)); | |
212 | static WORD_LIST *call_expand_word_internal __P((WORD_DESC *, int, int, int *, int *)); | |
213 | static WORD_LIST *expand_string_internal __P((char *, int)); | |
214 | static WORD_LIST *expand_string_leave_quoted __P((char *, int)); | |
215 | static WORD_LIST *expand_string_for_rhs __P((char *, int, int *, int *)); | |
216 | ||
217 | static WORD_LIST *list_quote_escapes __P((WORD_LIST *)); | |
218 | static char *make_quoted_char __P((int)); | |
219 | static WORD_LIST *quote_list __P((WORD_LIST *)); | |
220 | ||
221 | static int unquoted_substring __P((char *, char *)); | |
222 | static int unquoted_member __P((int, char *)); | |
223 | ||
224 | #if defined (ARRAY_VARS) | |
225 | static SHELL_VAR *do_compound_assignment __P((char *, char *, int)); | |
226 | #endif | |
227 | static int do_assignment_internal __P((const WORD_DESC *, int)); | |
228 | ||
229 | static char *string_extract_verbatim __P((char *, size_t, int *, char *, int)); | |
230 | static char *string_extract __P((char *, int *, char *, int)); | |
231 | static char *string_extract_double_quoted __P((char *, int *, int)); | |
232 | static inline char *string_extract_single_quoted __P((char *, int *)); | |
233 | static inline int skip_single_quoted __P((const char *, size_t, int)); | |
234 | static int skip_double_quoted __P((char *, size_t, int)); | |
235 | static char *extract_delimited_string __P((char *, int *, char *, char *, char *, int)); | |
236 | static char *extract_dollar_brace_string __P((char *, int *, int, int)); | |
237 | static int skip_matched_pair __P((const char *, int, int, int, int)); | |
238 | ||
239 | static char *pos_params __P((char *, int, int, int)); | |
240 | ||
241 | static unsigned char *mb_getcharlens __P((char *, int)); | |
242 | ||
243 | static char *remove_upattern __P((char *, char *, int)); | |
244 | #if defined (HANDLE_MULTIBYTE) | |
245 | static wchar_t *remove_wpattern __P((wchar_t *, size_t, wchar_t *, int)); | |
246 | #endif | |
247 | static char *remove_pattern __P((char *, char *, int)); | |
248 | ||
249 | static int match_upattern __P((char *, char *, int, char **, char **)); | |
250 | #if defined (HANDLE_MULTIBYTE) | |
251 | static int match_wpattern __P((wchar_t *, char **, size_t, wchar_t *, int, char **, char **)); | |
252 | #endif | |
253 | static int match_pattern __P((char *, char *, int, char **, char **)); | |
254 | static int getpatspec __P((int, char *)); | |
255 | static char *getpattern __P((char *, int, int)); | |
256 | static char *variable_remove_pattern __P((char *, char *, int, int)); | |
257 | static char *list_remove_pattern __P((WORD_LIST *, char *, int, int, int)); | |
258 | static char *parameter_list_remove_pattern __P((int, char *, int, int)); | |
259 | #ifdef ARRAY_VARS | |
260 | static char *array_remove_pattern __P((SHELL_VAR *, char *, int, char *, int)); | |
261 | #endif | |
262 | static char *parameter_brace_remove_pattern __P((char *, char *, int, char *, int, int, int)); | |
263 | ||
264 | static char *process_substitute __P((char *, int)); | |
265 | ||
266 | static char *read_comsub __P((int, int, int *)); | |
267 | ||
268 | #ifdef ARRAY_VARS | |
269 | static arrayind_t array_length_reference __P((char *)); | |
270 | #endif | |
271 | ||
272 | static int valid_brace_expansion_word __P((char *, int)); | |
273 | static int chk_atstar __P((char *, int, int *, int *)); | |
274 | static int chk_arithsub __P((const char *, int)); | |
275 | ||
276 | static WORD_DESC *parameter_brace_expand_word __P((char *, int, int, int, arrayind_t *)); | |
277 | static WORD_DESC *parameter_brace_expand_indir __P((char *, int, int, int *, int *)); | |
278 | static WORD_DESC *parameter_brace_expand_rhs __P((char *, char *, int, int, int *, int *)); | |
279 | static void parameter_brace_expand_error __P((char *, char *)); | |
280 | ||
281 | static int valid_length_expression __P((char *)); | |
282 | static intmax_t parameter_brace_expand_length __P((char *)); | |
283 | ||
284 | static char *skiparith __P((char *, int)); | |
285 | static int verify_substring_values __P((SHELL_VAR *, char *, char *, int, intmax_t *, intmax_t *)); | |
286 | static int get_var_and_type __P((char *, char *, arrayind_t, int, int, SHELL_VAR **, char **)); | |
287 | static char *mb_substring __P((char *, int, int)); | |
288 | static char *parameter_brace_substring __P((char *, char *, int, char *, int, int)); | |
289 | ||
290 | static int shouldexp_replacement __P((char *)); | |
291 | ||
292 | static char *pos_params_pat_subst __P((char *, char *, char *, int)); | |
293 | ||
294 | static char *parameter_brace_patsub __P((char *, char *, int, char *, int, int)); | |
295 | ||
296 | static char *pos_params_casemod __P((char *, char *, int, int)); | |
297 | static char *parameter_brace_casemod __P((char *, char *, int, int, char *, int, int)); | |
298 | ||
299 | static WORD_DESC *parameter_brace_expand __P((char *, int *, int, int, int *, int *)); | |
300 | static WORD_DESC *param_expand __P((char *, int *, int, int *, int *, int *, int *, int)); | |
301 | ||
302 | static WORD_LIST *expand_word_internal __P((WORD_DESC *, int, int, int *, int *)); | |
303 | ||
304 | static WORD_LIST *word_list_split __P((WORD_LIST *)); | |
305 | ||
306 | static void exp_jump_to_top_level __P((int)); | |
307 | ||
308 | static WORD_LIST *separate_out_assignments __P((WORD_LIST *)); | |
309 | static WORD_LIST *glob_expand_word_list __P((WORD_LIST *, int)); | |
310 | #ifdef BRACE_EXPANSION | |
311 | static WORD_LIST *brace_expand_word_list __P((WORD_LIST *, int)); | |
312 | #endif | |
313 | #if defined (ARRAY_VARS) | |
314 | static int make_internal_declare __P((char *, char *)); | |
315 | #endif | |
316 | static WORD_LIST *shell_expand_word_list __P((WORD_LIST *, int)); | |
317 | static WORD_LIST *expand_word_list_internal __P((WORD_LIST *, int)); | |
318 | ||
319 | /* **************************************************************** */ | |
320 | /* */ | |
321 | /* Utility Functions */ | |
322 | /* */ | |
323 | /* **************************************************************** */ | |
324 | ||
325 | #if defined (DEBUG) | |
326 | void | |
327 | dump_word_flags (flags) | |
328 | int flags; | |
329 | { | |
330 | int f; | |
331 | ||
332 | f = flags; | |
333 | fprintf (stderr, "%d -> ", f); | |
334 | if (f & W_ASSIGNASSOC) | |
335 | { | |
336 | f &= ~W_ASSIGNASSOC; | |
337 | fprintf (stderr, "W_ASSIGNASSOC%s", f ? "|" : ""); | |
338 | } | |
339 | if (f & W_HASCTLESC) | |
340 | { | |
341 | f &= ~W_HASCTLESC; | |
342 | fprintf (stderr, "W_HASCTLESC%s", f ? "|" : ""); | |
343 | } | |
344 | if (f & W_NOPROCSUB) | |
345 | { | |
346 | f &= ~W_NOPROCSUB; | |
347 | fprintf (stderr, "W_NOPROCSUB%s", f ? "|" : ""); | |
348 | } | |
349 | if (f & W_DQUOTE) | |
350 | { | |
351 | f &= ~W_DQUOTE; | |
352 | fprintf (stderr, "W_DQUOTE%s", f ? "|" : ""); | |
353 | } | |
354 | if (f & W_HASQUOTEDNULL) | |
355 | { | |
356 | f &= ~W_HASQUOTEDNULL; | |
357 | fprintf (stderr, "W_HASQUOTEDNULL%s", f ? "|" : ""); | |
358 | } | |
359 | if (f & W_ASSIGNARG) | |
360 | { | |
361 | f &= ~W_ASSIGNARG; | |
362 | fprintf (stderr, "W_ASSIGNARG%s", f ? "|" : ""); | |
363 | } | |
364 | if (f & W_ASSNBLTIN) | |
365 | { | |
366 | f &= ~W_ASSNBLTIN; | |
367 | fprintf (stderr, "W_ASSNBLTIN%s", f ? "|" : ""); | |
368 | } | |
369 | if (f & W_COMPASSIGN) | |
370 | { | |
371 | f &= ~W_COMPASSIGN; | |
372 | fprintf (stderr, "W_COMPASSIGN%s", f ? "|" : ""); | |
373 | } | |
374 | if (f & W_NOEXPAND) | |
375 | { | |
376 | f &= ~W_NOEXPAND; | |
377 | fprintf (stderr, "W_NOEXPAND%s", f ? "|" : ""); | |
378 | } | |
379 | if (f & W_ITILDE) | |
380 | { | |
381 | f &= ~W_ITILDE; | |
382 | fprintf (stderr, "W_ITILDE%s", f ? "|" : ""); | |
383 | } | |
384 | if (f & W_NOTILDE) | |
385 | { | |
386 | f &= ~W_NOTILDE; | |
387 | fprintf (stderr, "W_NOTILDE%s", f ? "|" : ""); | |
388 | } | |
389 | if (f & W_ASSIGNRHS) | |
390 | { | |
391 | f &= ~W_ASSIGNRHS; | |
392 | fprintf (stderr, "W_ASSIGNRHS%s", f ? "|" : ""); | |
393 | } | |
394 | if (f & W_NOCOMSUB) | |
395 | { | |
396 | f &= ~W_NOCOMSUB; | |
397 | fprintf (stderr, "W_NOCOMSUB%s", f ? "|" : ""); | |
398 | } | |
399 | if (f & W_DOLLARSTAR) | |
400 | { | |
401 | f &= ~W_DOLLARSTAR; | |
402 | fprintf (stderr, "W_DOLLARSTAR%s", f ? "|" : ""); | |
403 | } | |
404 | if (f & W_DOLLARAT) | |
405 | { | |
406 | f &= ~W_DOLLARAT; | |
407 | fprintf (stderr, "W_DOLLARAT%s", f ? "|" : ""); | |
408 | } | |
409 | if (f & W_TILDEEXP) | |
410 | { | |
411 | f &= ~W_TILDEEXP; | |
412 | fprintf (stderr, "W_TILDEEXP%s", f ? "|" : ""); | |
413 | } | |
414 | if (f & W_NOSPLIT2) | |
415 | { | |
416 | f &= ~W_NOSPLIT2; | |
417 | fprintf (stderr, "W_NOSPLIT2%s", f ? "|" : ""); | |
418 | } | |
419 | if (f & W_NOGLOB) | |
420 | { | |
421 | f &= ~W_NOGLOB; | |
422 | fprintf (stderr, "W_NOGLOB%s", f ? "|" : ""); | |
423 | } | |
424 | if (f & W_NOSPLIT) | |
425 | { | |
426 | f &= ~W_NOSPLIT; | |
427 | fprintf (stderr, "W_NOSPLIT%s", f ? "|" : ""); | |
428 | } | |
429 | if (f & W_GLOBEXP) | |
430 | { | |
431 | f &= ~W_GLOBEXP; | |
432 | fprintf (stderr, "W_GLOBEXP%s", f ? "|" : ""); | |
433 | } | |
434 | if (f & W_ASSIGNMENT) | |
435 | { | |
436 | f &= ~W_ASSIGNMENT; | |
437 | fprintf (stderr, "W_ASSIGNMENT%s", f ? "|" : ""); | |
438 | } | |
439 | if (f & W_QUOTED) | |
440 | { | |
441 | f &= ~W_QUOTED; | |
442 | fprintf (stderr, "W_QUOTED%s", f ? "|" : ""); | |
443 | } | |
444 | if (f & W_HASDOLLAR) | |
445 | { | |
446 | f &= ~W_HASDOLLAR; | |
447 | fprintf (stderr, "W_HASDOLLAR%s", f ? "|" : ""); | |
448 | } | |
449 | fprintf (stderr, "\n"); | |
450 | fflush (stderr); | |
451 | } | |
452 | #endif | |
453 | ||
454 | #ifdef INCLUDE_UNUSED | |
455 | static char * | |
456 | quoted_substring (string, start, end) | |
457 | char *string; | |
458 | int start, end; | |
459 | { | |
460 | register int len, l; | |
461 | register char *result, *s, *r; | |
462 | ||
463 | len = end - start; | |
464 | ||
465 | /* Move to string[start], skipping quoted characters. */ | |
466 | for (s = string, l = 0; *s && l < start; ) | |
467 | { | |
468 | if (*s == CTLESC) | |
469 | { | |
470 | s++; | |
471 | continue; | |
472 | } | |
473 | l++; | |
474 | if (*s == 0) | |
475 | break; | |
476 | } | |
477 | ||
478 | r = result = (char *)xmalloc (2*len + 1); /* save room for quotes */ | |
479 | ||
480 | /* Copy LEN characters, including quote characters. */ | |
481 | s = string + l; | |
482 | for (l = 0; l < len; s++) | |
483 | { | |
484 | if (*s == CTLESC) | |
485 | *r++ = *s++; | |
486 | *r++ = *s; | |
487 | l++; | |
488 | if (*s == 0) | |
489 | break; | |
490 | } | |
491 | *r = '\0'; | |
492 | return result; | |
493 | } | |
494 | #endif | |
495 | ||
496 | #ifdef INCLUDE_UNUSED | |
497 | /* Return the length of S, skipping over quoted characters */ | |
498 | static int | |
499 | quoted_strlen (s) | |
500 | char *s; | |
501 | { | |
502 | register char *p; | |
503 | int i; | |
504 | ||
505 | i = 0; | |
506 | for (p = s; *p; p++) | |
507 | { | |
508 | if (*p == CTLESC) | |
509 | { | |
510 | p++; | |
511 | if (*p == 0) | |
512 | return (i + 1); | |
513 | } | |
514 | i++; | |
515 | } | |
516 | ||
517 | return i; | |
518 | } | |
519 | #endif | |
520 | ||
521 | /* Find the first occurrence of character C in string S, obeying shell | |
522 | quoting rules. If (FLAGS & ST_BACKSL) is non-zero, backslash-escaped | |
523 | characters are skipped. If (FLAGS & ST_CTLESC) is non-zero, characters | |
524 | escaped with CTLESC are skipped. */ | |
525 | static char * | |
526 | quoted_strchr (s, c, flags) | |
527 | char *s; | |
528 | int c, flags; | |
529 | { | |
530 | register char *p; | |
531 | ||
532 | for (p = s; *p; p++) | |
533 | { | |
534 | if (((flags & ST_BACKSL) && *p == '\\') | |
535 | || ((flags & ST_CTLESC) && *p == CTLESC)) | |
536 | { | |
537 | p++; | |
538 | if (*p == '\0') | |
539 | return ((char *)NULL); | |
540 | continue; | |
541 | } | |
542 | else if (*p == c) | |
543 | return p; | |
544 | } | |
545 | return ((char *)NULL); | |
546 | } | |
547 | ||
548 | /* Return 1 if CHARACTER appears in an unquoted portion of | |
549 | STRING. Return 0 otherwise. CHARACTER must be a single-byte character. */ | |
550 | static int | |
551 | unquoted_member (character, string) | |
552 | int character; | |
553 | char *string; | |
554 | { | |
555 | size_t slen; | |
556 | int sindex, c; | |
557 | DECLARE_MBSTATE; | |
558 | ||
559 | slen = strlen (string); | |
560 | sindex = 0; | |
561 | while (c = string[sindex]) | |
562 | { | |
563 | if (c == character) | |
564 | return (1); | |
565 | ||
566 | switch (c) | |
567 | { | |
568 | default: | |
569 | ADVANCE_CHAR (string, slen, sindex); | |
570 | break; | |
571 | ||
572 | case '\\': | |
573 | sindex++; | |
574 | if (string[sindex]) | |
575 | ADVANCE_CHAR (string, slen, sindex); | |
576 | break; | |
577 | ||
578 | case '\'': | |
579 | sindex = skip_single_quoted (string, slen, ++sindex); | |
580 | break; | |
581 | ||
582 | case '"': | |
583 | sindex = skip_double_quoted (string, slen, ++sindex); | |
584 | break; | |
585 | } | |
586 | } | |
587 | return (0); | |
588 | } | |
589 | ||
590 | /* Return 1 if SUBSTR appears in an unquoted portion of STRING. */ | |
591 | static int | |
592 | unquoted_substring (substr, string) | |
593 | char *substr, *string; | |
594 | { | |
595 | size_t slen; | |
596 | int sindex, c, sublen; | |
597 | DECLARE_MBSTATE; | |
598 | ||
599 | if (substr == 0 || *substr == '\0') | |
600 | return (0); | |
601 | ||
602 | slen = strlen (string); | |
603 | sublen = strlen (substr); | |
604 | for (sindex = 0; c = string[sindex]; ) | |
605 | { | |
606 | if (STREQN (string + sindex, substr, sublen)) | |
607 | return (1); | |
608 | ||
609 | switch (c) | |
610 | { | |
611 | case '\\': | |
612 | sindex++; | |
613 | if (string[sindex]) | |
614 | ADVANCE_CHAR (string, slen, sindex); | |
615 | break; | |
616 | ||
617 | case '\'': | |
618 | sindex = skip_single_quoted (string, slen, ++sindex); | |
619 | break; | |
620 | ||
621 | case '"': | |
622 | sindex = skip_double_quoted (string, slen, ++sindex); | |
623 | break; | |
624 | ||
625 | default: | |
626 | ADVANCE_CHAR (string, slen, sindex); | |
627 | break; | |
628 | } | |
629 | } | |
630 | return (0); | |
631 | } | |
632 | ||
633 | /* Most of the substitutions must be done in parallel. In order | |
634 | to avoid using tons of unclear goto's, I have some functions | |
635 | for manipulating malloc'ed strings. They all take INDX, a | |
636 | pointer to an integer which is the offset into the string | |
637 | where manipulation is taking place. They also take SIZE, a | |
638 | pointer to an integer which is the current length of the | |
639 | character array for this string. */ | |
640 | ||
641 | /* Append SOURCE to TARGET at INDEX. SIZE is the current amount | |
642 | of space allocated to TARGET. SOURCE can be NULL, in which | |
643 | case nothing happens. Gets rid of SOURCE by freeing it. | |
644 | Returns TARGET in case the location has changed. */ | |
645 | INLINE char * | |
646 | sub_append_string (source, target, indx, size) | |
647 | char *source, *target; | |
648 | int *indx, *size; | |
649 | { | |
650 | if (source) | |
651 | { | |
652 | int srclen, n; | |
653 | ||
654 | srclen = STRLEN (source); | |
655 | if (srclen >= (int)(*size - *indx)) | |
656 | { | |
657 | n = srclen + *indx; | |
658 | n = (n + DEFAULT_ARRAY_SIZE) - (n % DEFAULT_ARRAY_SIZE); | |
659 | target = (char *)xrealloc (target, (*size = n)); | |
660 | } | |
661 | ||
662 | FASTCOPY (source, target + *indx, srclen); | |
663 | *indx += srclen; | |
664 | target[*indx] = '\0'; | |
665 | ||
666 | free (source); | |
667 | } | |
668 | return (target); | |
669 | } | |
670 | ||
671 | #if 0 | |
672 | /* UNUSED */ | |
673 | /* Append the textual representation of NUMBER to TARGET. | |
674 | INDX and SIZE are as in SUB_APPEND_STRING. */ | |
675 | char * | |
676 | sub_append_number (number, target, indx, size) | |
677 | intmax_t number; | |
678 | int *indx, *size; | |
679 | char *target; | |
680 | { | |
681 | char *temp; | |
682 | ||
683 | temp = itos (number); | |
684 | return (sub_append_string (temp, target, indx, size)); | |
685 | } | |
686 | #endif | |
687 | ||
688 | /* Extract a substring from STRING, starting at SINDEX and ending with | |
689 | one of the characters in CHARLIST. Don't make the ending character | |
690 | part of the string. Leave SINDEX pointing at the ending character. | |
691 | Understand about backslashes in the string. If (flags & SX_VARNAME) | |
692 | is non-zero, and array variables have been compiled into the shell, | |
693 | everything between a `[' and a corresponding `]' is skipped over. | |
694 | If (flags & SX_NOALLOC) is non-zero, don't return the substring, just | |
695 | update SINDEX. If (flags & SX_REQMATCH) is non-zero, the string must | |
696 | contain a closing character from CHARLIST. */ | |
697 | static char * | |
698 | string_extract (string, sindex, charlist, flags) | |
699 | char *string; | |
700 | int *sindex; | |
701 | char *charlist; | |
702 | int flags; | |
703 | { | |
704 | register int c, i; | |
705 | int found; | |
706 | size_t slen; | |
707 | char *temp; | |
708 | DECLARE_MBSTATE; | |
709 | ||
710 | slen = (MB_CUR_MAX > 1) ? strlen (string + *sindex) + *sindex : 0; | |
711 | i = *sindex; | |
712 | found = 0; | |
713 | while (c = string[i]) | |
714 | { | |
715 | if (c == '\\') | |
716 | { | |
717 | if (string[i + 1]) | |
718 | i++; | |
719 | else | |
720 | break; | |
721 | } | |
722 | #if defined (ARRAY_VARS) | |
723 | else if ((flags & SX_VARNAME) && c == '[') | |
724 | { | |
725 | int ni; | |
726 | /* If this is an array subscript, skip over it and continue. */ | |
727 | ni = skipsubscript (string, i, 0); | |
728 | if (string[ni] == ']') | |
729 | i = ni; | |
730 | } | |
731 | #endif | |
732 | else if (MEMBER (c, charlist)) | |
733 | { | |
734 | found = 1; | |
735 | break; | |
736 | } | |
737 | ||
738 | ADVANCE_CHAR (string, slen, i); | |
739 | } | |
740 | ||
741 | /* If we had to have a matching delimiter and didn't find one, return an | |
742 | error and let the caller deal with it. */ | |
743 | if ((flags & SX_REQMATCH) && found == 0) | |
744 | { | |
745 | *sindex = i; | |
746 | return (&extract_string_error); | |
747 | } | |
748 | ||
749 | temp = (flags & SX_NOALLOC) ? (char *)NULL : substring (string, *sindex, i); | |
750 | *sindex = i; | |
751 | ||
752 | return (temp); | |
753 | } | |
754 | ||
755 | /* Extract the contents of STRING as if it is enclosed in double quotes. | |
756 | SINDEX, when passed in, is the offset of the character immediately | |
757 | following the opening double quote; on exit, SINDEX is left pointing after | |
758 | the closing double quote. If STRIPDQ is non-zero, unquoted double | |
759 | quotes are stripped and the string is terminated by a null byte. | |
760 | Backslashes between the embedded double quotes are processed. If STRIPDQ | |
761 | is zero, an unquoted `"' terminates the string. */ | |
762 | static char * | |
763 | string_extract_double_quoted (string, sindex, stripdq) | |
764 | char *string; | |
765 | int *sindex, stripdq; | |
766 | { | |
767 | size_t slen; | |
768 | char *send; | |
769 | int j, i, t; | |
770 | unsigned char c; | |
771 | char *temp, *ret; /* The new string we return. */ | |
772 | int pass_next, backquote, si; /* State variables for the machine. */ | |
773 | int dquote; | |
774 | DECLARE_MBSTATE; | |
775 | ||
776 | slen = strlen (string + *sindex) + *sindex; | |
777 | send = string + slen; | |
778 | ||
779 | pass_next = backquote = dquote = 0; | |
780 | temp = (char *)xmalloc (1 + slen - *sindex); | |
781 | ||
782 | j = 0; | |
783 | i = *sindex; | |
784 | while (c = string[i]) | |
785 | { | |
786 | /* Process a character that was quoted by a backslash. */ | |
787 | if (pass_next) | |
788 | { | |
789 | /* XXX - take another look at this in light of Interp 221 */ | |
790 | /* Posix.2 sez: | |
791 | ||
792 | ``The backslash shall retain its special meaning as an escape | |
793 | character only when followed by one of the characters: | |
794 | $ ` " \ <newline>''. | |
795 | ||
796 | If STRIPDQ is zero, we handle the double quotes here and let | |
797 | expand_word_internal handle the rest. If STRIPDQ is non-zero, | |
798 | we have already been through one round of backslash stripping, | |
799 | and want to strip these backslashes only if DQUOTE is non-zero, | |
800 | indicating that we are inside an embedded double-quoted string. */ | |
801 | ||
802 | /* If we are in an embedded quoted string, then don't strip | |
803 | backslashes before characters for which the backslash | |
804 | retains its special meaning, but remove backslashes in | |
805 | front of other characters. If we are not in an | |
806 | embedded quoted string, don't strip backslashes at all. | |
807 | This mess is necessary because the string was already | |
808 | surrounded by double quotes (and sh has some really weird | |
809 | quoting rules). | |
810 | The returned string will be run through expansion as if | |
811 | it were double-quoted. */ | |
812 | if ((stripdq == 0 && c != '"') || | |
813 | (stripdq && ((dquote && (sh_syntaxtab[c] & CBSDQUOTE)) || dquote == 0))) | |
814 | temp[j++] = '\\'; | |
815 | pass_next = 0; | |
816 | ||
817 | add_one_character: | |
818 | COPY_CHAR_I (temp, j, string, send, i); | |
819 | continue; | |
820 | } | |
821 | ||
822 | /* A backslash protects the next character. The code just above | |
823 | handles preserving the backslash in front of any character but | |
824 | a double quote. */ | |
825 | if (c == '\\') | |
826 | { | |
827 | pass_next++; | |
828 | i++; | |
829 | continue; | |
830 | } | |
831 | ||
832 | /* Inside backquotes, ``the portion of the quoted string from the | |
833 | initial backquote and the characters up to the next backquote | |
834 | that is not preceded by a backslash, having escape characters | |
835 | removed, defines that command''. */ | |
836 | if (backquote) | |
837 | { | |
838 | if (c == '`') | |
839 | backquote = 0; | |
840 | temp[j++] = c; | |
841 | i++; | |
842 | continue; | |
843 | } | |
844 | ||
845 | if (c == '`') | |
846 | { | |
847 | temp[j++] = c; | |
848 | backquote++; | |
849 | i++; | |
850 | continue; | |
851 | } | |
852 | ||
853 | /* Pass everything between `$(' and the matching `)' or a quoted | |
854 | ${ ... } pair through according to the Posix.2 specification. */ | |
855 | if (c == '$' && ((string[i + 1] == LPAREN) || (string[i + 1] == LBRACE))) | |
856 | { | |
857 | int free_ret = 1; | |
858 | ||
859 | si = i + 2; | |
860 | if (string[i + 1] == LPAREN) | |
861 | ret = extract_command_subst (string, &si, 0); | |
862 | else | |
863 | ret = extract_dollar_brace_string (string, &si, Q_DOUBLE_QUOTES, 0); | |
864 | ||
865 | temp[j++] = '$'; | |
866 | temp[j++] = string[i + 1]; | |
867 | ||
868 | /* Just paranoia; ret will not be 0 unless no_longjmp_on_fatal_error | |
869 | is set. */ | |
870 | if (ret == 0 && no_longjmp_on_fatal_error) | |
871 | { | |
872 | free_ret = 0; | |
873 | ret = string + i + 2; | |
874 | } | |
875 | ||
876 | for (t = 0; ret[t]; t++, j++) | |
877 | temp[j] = ret[t]; | |
878 | temp[j] = string[si]; | |
879 | ||
880 | if (string[si]) | |
881 | { | |
882 | j++; | |
883 | i = si + 1; | |
884 | } | |
885 | else | |
886 | i = si; | |
887 | ||
888 | if (free_ret) | |
889 | free (ret); | |
890 | continue; | |
891 | } | |
892 | ||
893 | /* Add any character but a double quote to the quoted string we're | |
894 | accumulating. */ | |
895 | if (c != '"') | |
896 | goto add_one_character; | |
897 | ||
898 | /* c == '"' */ | |
899 | if (stripdq) | |
900 | { | |
901 | dquote ^= 1; | |
902 | i++; | |
903 | continue; | |
904 | } | |
905 | ||
906 | break; | |
907 | } | |
908 | temp[j] = '\0'; | |
909 | ||
910 | /* Point to after the closing quote. */ | |
911 | if (c) | |
912 | i++; | |
913 | *sindex = i; | |
914 | ||
915 | return (temp); | |
916 | } | |
917 | ||
918 | /* This should really be another option to string_extract_double_quoted. */ | |
919 | static int | |
920 | skip_double_quoted (string, slen, sind) | |
921 | char *string; | |
922 | size_t slen; | |
923 | int sind; | |
924 | { | |
925 | int c, i; | |
926 | char *ret; | |
927 | int pass_next, backquote, si; | |
928 | DECLARE_MBSTATE; | |
929 | ||
930 | pass_next = backquote = 0; | |
931 | i = sind; | |
932 | while (c = string[i]) | |
933 | { | |
934 | if (pass_next) | |
935 | { | |
936 | pass_next = 0; | |
937 | ADVANCE_CHAR (string, slen, i); | |
938 | continue; | |
939 | } | |
940 | else if (c == '\\') | |
941 | { | |
942 | pass_next++; | |
943 | i++; | |
944 | continue; | |
945 | } | |
946 | else if (backquote) | |
947 | { | |
948 | if (c == '`') | |
949 | backquote = 0; | |
950 | ADVANCE_CHAR (string, slen, i); | |
951 | continue; | |
952 | } | |
953 | else if (c == '`') | |
954 | { | |
955 | backquote++; | |
956 | i++; | |
957 | continue; | |
958 | } | |
959 | else if (c == '$' && ((string[i + 1] == LPAREN) || (string[i + 1] == LBRACE))) | |
960 | { | |
961 | si = i + 2; | |
962 | if (string[i + 1] == LPAREN) | |
963 | ret = extract_command_subst (string, &si, SX_NOALLOC); | |
964 | else | |
965 | ret = extract_dollar_brace_string (string, &si, Q_DOUBLE_QUOTES, SX_NOALLOC); | |
966 | ||
967 | i = si + 1; | |
968 | continue; | |
969 | } | |
970 | else if (c != '"') | |
971 | { | |
972 | ADVANCE_CHAR (string, slen, i); | |
973 | continue; | |
974 | } | |
975 | else | |
976 | break; | |
977 | } | |
978 | ||
979 | if (c) | |
980 | i++; | |
981 | ||
982 | return (i); | |
983 | } | |
984 | ||
985 | /* Extract the contents of STRING as if it is enclosed in single quotes. | |
986 | SINDEX, when passed in, is the offset of the character immediately | |
987 | following the opening single quote; on exit, SINDEX is left pointing after | |
988 | the closing single quote. */ | |
989 | static inline char * | |
990 | string_extract_single_quoted (string, sindex) | |
991 | char *string; | |
992 | int *sindex; | |
993 | { | |
994 | register int i; | |
995 | size_t slen; | |
996 | char *t; | |
997 | DECLARE_MBSTATE; | |
998 | ||
999 | /* Don't need slen for ADVANCE_CHAR unless multibyte chars possible. */ | |
1000 | slen = (MB_CUR_MAX > 1) ? strlen (string + *sindex) + *sindex : 0; | |
1001 | i = *sindex; | |
1002 | while (string[i] && string[i] != '\'') | |
1003 | ADVANCE_CHAR (string, slen, i); | |
1004 | ||
1005 | t = substring (string, *sindex, i); | |
1006 | ||
1007 | if (string[i]) | |
1008 | i++; | |
1009 | *sindex = i; | |
1010 | ||
1011 | return (t); | |
1012 | } | |
1013 | ||
1014 | static inline int | |
1015 | skip_single_quoted (string, slen, sind) | |
1016 | const char *string; | |
1017 | size_t slen; | |
1018 | int sind; | |
1019 | { | |
1020 | register int c; | |
1021 | DECLARE_MBSTATE; | |
1022 | ||
1023 | c = sind; | |
1024 | while (string[c] && string[c] != '\'') | |
1025 | ADVANCE_CHAR (string, slen, c); | |
1026 | ||
1027 | if (string[c]) | |
1028 | c++; | |
1029 | return c; | |
1030 | } | |
1031 | ||
1032 | /* Just like string_extract, but doesn't hack backslashes or any of | |
1033 | that other stuff. Obeys CTLESC quoting. Used to do splitting on $IFS. */ | |
1034 | static char * | |
1035 | string_extract_verbatim (string, slen, sindex, charlist, flags) | |
1036 | char *string; | |
1037 | size_t slen; | |
1038 | int *sindex; | |
1039 | char *charlist; | |
1040 | int flags; | |
1041 | { | |
1042 | register int i; | |
1043 | #if defined (HANDLE_MULTIBYTE) | |
1044 | size_t clen; | |
1045 | wchar_t *wcharlist; | |
1046 | #endif | |
1047 | int c; | |
1048 | char *temp; | |
1049 | DECLARE_MBSTATE; | |
1050 | ||
1051 | if (charlist[0] == '\'' && charlist[1] == '\0') | |
1052 | { | |
1053 | temp = string_extract_single_quoted (string, sindex); | |
1054 | --*sindex; /* leave *sindex at separator character */ | |
1055 | return temp; | |
1056 | } | |
1057 | ||
1058 | i = *sindex; | |
1059 | #if 0 | |
1060 | /* See how the MBLEN and ADVANCE_CHAR macros work to understand why we need | |
1061 | this only if MB_CUR_MAX > 1. */ | |
1062 | slen = (MB_CUR_MAX > 1) ? strlen (string + *sindex) + *sindex : 1; | |
1063 | #endif | |
1064 | #if defined (HANDLE_MULTIBYTE) | |
1065 | clen = strlen (charlist); | |
1066 | wcharlist = 0; | |
1067 | #endif | |
1068 | while (c = string[i]) | |
1069 | { | |
1070 | #if defined (HANDLE_MULTIBYTE) | |
1071 | size_t mblength; | |
1072 | #endif | |
1073 | if ((flags & SX_NOCTLESC) == 0 && c == CTLESC) | |
1074 | { | |
1075 | i += 2; | |
1076 | continue; | |
1077 | } | |
1078 | /* Even if flags contains SX_NOCTLESC, we let CTLESC quoting CTLNUL | |
1079 | through, to protect the CTLNULs from later calls to | |
1080 | remove_quoted_nulls. */ | |
1081 | else if ((flags & SX_NOESCCTLNUL) == 0 && c == CTLESC && string[i+1] == CTLNUL) | |
1082 | { | |
1083 | i += 2; | |
1084 | continue; | |
1085 | } | |
1086 | ||
1087 | #if defined (HANDLE_MULTIBYTE) | |
1088 | mblength = MBLEN (string + i, slen - i); | |
1089 | if (mblength > 1) | |
1090 | { | |
1091 | wchar_t wc; | |
1092 | mblength = mbtowc (&wc, string + i, slen - i); | |
1093 | if (MB_INVALIDCH (mblength)) | |
1094 | { | |
1095 | if (MEMBER (c, charlist)) | |
1096 | break; | |
1097 | } | |
1098 | else | |
1099 | { | |
1100 | if (wcharlist == 0) | |
1101 | { | |
1102 | size_t len; | |
1103 | len = mbstowcs (wcharlist, charlist, 0); | |
1104 | if (len == -1) | |
1105 | len = 0; | |
1106 | wcharlist = (wchar_t *)xmalloc (sizeof (wchar_t) * (len + 1)); | |
1107 | mbstowcs (wcharlist, charlist, len + 1); | |
1108 | } | |
1109 | ||
1110 | if (wcschr (wcharlist, wc)) | |
1111 | break; | |
1112 | } | |
1113 | } | |
1114 | else | |
1115 | #endif | |
1116 | if (MEMBER (c, charlist)) | |
1117 | break; | |
1118 | ||
1119 | ADVANCE_CHAR (string, slen, i); | |
1120 | } | |
1121 | ||
1122 | #if defined (HANDLE_MULTIBYTE) | |
1123 | FREE (wcharlist); | |
1124 | #endif | |
1125 | ||
1126 | temp = substring (string, *sindex, i); | |
1127 | *sindex = i; | |
1128 | ||
1129 | return (temp); | |
1130 | } | |
1131 | ||
1132 | /* Extract the $( construct in STRING, and return a new string. | |
1133 | Start extracting at (SINDEX) as if we had just seen "$(". | |
1134 | Make (SINDEX) get the position of the matching ")". ) | |
1135 | XFLAGS is additional flags to pass to other extraction functions. */ | |
1136 | char * | |
1137 | extract_command_subst (string, sindex, xflags) | |
1138 | char *string; | |
1139 | int *sindex; | |
1140 | int xflags; | |
1141 | { | |
1142 | if (string[*sindex] == LPAREN) | |
1143 | return (extract_delimited_string (string, sindex, "$(", "(", ")", xflags|SX_COMMAND)); /*)*/ | |
1144 | else | |
1145 | { | |
1146 | xflags |= (no_longjmp_on_fatal_error ? SX_NOLONGJMP : 0); | |
1147 | return (xparse_dolparen (string, string+*sindex, sindex, xflags)); | |
1148 | } | |
1149 | } | |
1150 | ||
1151 | /* Extract the $[ construct in STRING, and return a new string. (]) | |
1152 | Start extracting at (SINDEX) as if we had just seen "$[". | |
1153 | Make (SINDEX) get the position of the matching "]". */ | |
1154 | char * | |
1155 | extract_arithmetic_subst (string, sindex) | |
1156 | char *string; | |
1157 | int *sindex; | |
1158 | { | |
1159 | return (extract_delimited_string (string, sindex, "$[", "[", "]", 0)); /*]*/ | |
1160 | } | |
1161 | ||
1162 | #if defined (PROCESS_SUBSTITUTION) | |
1163 | /* Extract the <( or >( construct in STRING, and return a new string. | |
1164 | Start extracting at (SINDEX) as if we had just seen "<(". | |
1165 | Make (SINDEX) get the position of the matching ")". */ /*))*/ | |
1166 | char * | |
1167 | extract_process_subst (string, starter, sindex) | |
1168 | char *string; | |
1169 | char *starter; | |
1170 | int *sindex; | |
1171 | { | |
1172 | return (extract_delimited_string (string, sindex, starter, "(", ")", 0)); | |
1173 | } | |
1174 | #endif /* PROCESS_SUBSTITUTION */ | |
1175 | ||
1176 | #if defined (ARRAY_VARS) | |
1177 | /* This can be fooled by unquoted right parens in the passed string. If | |
1178 | each caller verifies that the last character in STRING is a right paren, | |
1179 | we don't even need to call extract_delimited_string. */ | |
1180 | char * | |
1181 | extract_array_assignment_list (string, sindex) | |
1182 | char *string; | |
1183 | int *sindex; | |
1184 | { | |
1185 | int slen; | |
1186 | char *ret; | |
1187 | ||
1188 | slen = strlen (string); /* ( */ | |
1189 | if (string[slen - 1] == ')') | |
1190 | { | |
1191 | ret = substring (string, *sindex, slen - 1); | |
1192 | *sindex = slen - 1; | |
1193 | return ret; | |
1194 | } | |
1195 | return 0; | |
1196 | } | |
1197 | #endif | |
1198 | ||
1199 | /* Extract and create a new string from the contents of STRING, a | |
1200 | character string delimited with OPENER and CLOSER. SINDEX is | |
1201 | the address of an int describing the current offset in STRING; | |
1202 | it should point to just after the first OPENER found. On exit, | |
1203 | SINDEX gets the position of the last character of the matching CLOSER. | |
1204 | If OPENER is more than a single character, ALT_OPENER, if non-null, | |
1205 | contains a character string that can also match CLOSER and thus | |
1206 | needs to be skipped. */ | |
1207 | static char * | |
1208 | extract_delimited_string (string, sindex, opener, alt_opener, closer, flags) | |
1209 | char *string; | |
1210 | int *sindex; | |
1211 | char *opener, *alt_opener, *closer; | |
1212 | int flags; | |
1213 | { | |
1214 | int i, c, si; | |
1215 | size_t slen; | |
1216 | char *t, *result; | |
1217 | int pass_character, nesting_level, in_comment; | |
1218 | int len_closer, len_opener, len_alt_opener; | |
1219 | DECLARE_MBSTATE; | |
1220 | ||
1221 | slen = strlen (string + *sindex) + *sindex; | |
1222 | len_opener = STRLEN (opener); | |
1223 | len_alt_opener = STRLEN (alt_opener); | |
1224 | len_closer = STRLEN (closer); | |
1225 | ||
1226 | pass_character = in_comment = 0; | |
1227 | ||
1228 | nesting_level = 1; | |
1229 | i = *sindex; | |
1230 | ||
1231 | while (nesting_level) | |
1232 | { | |
1233 | c = string[i]; | |
1234 | ||
1235 | if (c == 0) | |
1236 | break; | |
1237 | ||
1238 | if (in_comment) | |
1239 | { | |
1240 | if (c == '\n') | |
1241 | in_comment = 0; | |
1242 | ADVANCE_CHAR (string, slen, i); | |
1243 | continue; | |
1244 | } | |
1245 | ||
1246 | if (pass_character) /* previous char was backslash */ | |
1247 | { | |
1248 | pass_character = 0; | |
1249 | ADVANCE_CHAR (string, slen, i); | |
1250 | continue; | |
1251 | } | |
1252 | ||
1253 | /* Not exactly right yet; should handle shell metacharacters and | |
1254 | multibyte characters, too. See COMMENT_BEGIN define in parse.y */ | |
1255 | if ((flags & SX_COMMAND) && c == '#' && (i == 0 || string[i - 1] == '\n' || shellblank (string[i - 1]))) | |
1256 | { | |
1257 | in_comment = 1; | |
1258 | ADVANCE_CHAR (string, slen, i); | |
1259 | continue; | |
1260 | } | |
1261 | ||
1262 | if (c == CTLESC || c == '\\') | |
1263 | { | |
1264 | pass_character++; | |
1265 | i++; | |
1266 | continue; | |
1267 | } | |
1268 | ||
1269 | /* Process a nested command substitution, but only if we're parsing an | |
1270 | arithmetic substitution. */ | |
1271 | if ((flags & SX_COMMAND) && string[i] == '$' && string[i+1] == LPAREN) | |
1272 | { | |
1273 | si = i + 2; | |
1274 | t = extract_command_subst (string, &si, flags|SX_NOALLOC); | |
1275 | i = si + 1; | |
1276 | continue; | |
1277 | } | |
1278 | ||
1279 | /* Process a nested OPENER. */ | |
1280 | if (STREQN (string + i, opener, len_opener)) | |
1281 | { | |
1282 | si = i + len_opener; | |
1283 | t = extract_delimited_string (string, &si, opener, alt_opener, closer, flags|SX_NOALLOC); | |
1284 | i = si + 1; | |
1285 | continue; | |
1286 | } | |
1287 | ||
1288 | /* Process a nested ALT_OPENER */ | |
1289 | if (len_alt_opener && STREQN (string + i, alt_opener, len_alt_opener)) | |
1290 | { | |
1291 | si = i + len_alt_opener; | |
1292 | t = extract_delimited_string (string, &si, alt_opener, alt_opener, closer, flags|SX_NOALLOC); | |
1293 | i = si + 1; | |
1294 | continue; | |
1295 | } | |
1296 | ||
1297 | /* If the current substring terminates the delimited string, decrement | |
1298 | the nesting level. */ | |
1299 | if (STREQN (string + i, closer, len_closer)) | |
1300 | { | |
1301 | i += len_closer - 1; /* move to last byte of the closer */ | |
1302 | nesting_level--; | |
1303 | if (nesting_level == 0) | |
1304 | break; | |
1305 | } | |
1306 | ||
1307 | /* Pass old-style command substitution through verbatim. */ | |
1308 | if (c == '`') | |
1309 | { | |
1310 | si = i + 1; | |
1311 | t = string_extract (string, &si, "`", flags|SX_NOALLOC); | |
1312 | i = si + 1; | |
1313 | continue; | |
1314 | } | |
1315 | ||
1316 | /* Pass single-quoted and double-quoted strings through verbatim. */ | |
1317 | if (c == '\'' || c == '"') | |
1318 | { | |
1319 | si = i + 1; | |
1320 | i = (c == '\'') ? skip_single_quoted (string, slen, si) | |
1321 | : skip_double_quoted (string, slen, si); | |
1322 | continue; | |
1323 | } | |
1324 | ||
1325 | /* move past this character, which was not special. */ | |
1326 | ADVANCE_CHAR (string, slen, i); | |
1327 | } | |
1328 | ||
1329 | if (c == 0 && nesting_level) | |
1330 | { | |
1331 | if (no_longjmp_on_fatal_error == 0) | |
1332 | { | |
1333 | report_error (_("bad substitution: no closing `%s' in %s"), closer, string); | |
1334 | last_command_exit_value = EXECUTION_FAILURE; | |
1335 | exp_jump_to_top_level (DISCARD); | |
1336 | } | |
1337 | else | |
1338 | { | |
1339 | *sindex = i; | |
1340 | return (char *)NULL; | |
1341 | } | |
1342 | } | |
1343 | ||
1344 | si = i - *sindex - len_closer + 1; | |
1345 | if (flags & SX_NOALLOC) | |
1346 | result = (char *)NULL; | |
1347 | else | |
1348 | { | |
1349 | result = (char *)xmalloc (1 + si); | |
1350 | strncpy (result, string + *sindex, si); | |
1351 | result[si] = '\0'; | |
1352 | } | |
1353 | *sindex = i; | |
1354 | ||
1355 | return (result); | |
1356 | } | |
1357 | ||
1358 | /* Extract a parameter expansion expression within ${ and } from STRING. | |
1359 | Obey the Posix.2 rules for finding the ending `}': count braces while | |
1360 | skipping over enclosed quoted strings and command substitutions. | |
1361 | SINDEX is the address of an int describing the current offset in STRING; | |
1362 | it should point to just after the first `{' found. On exit, SINDEX | |
1363 | gets the position of the matching `}'. QUOTED is non-zero if this | |
1364 | occurs inside double quotes. */ | |
1365 | /* XXX -- this is very similar to extract_delimited_string -- XXX */ | |
1366 | static char * | |
1367 | extract_dollar_brace_string (string, sindex, quoted, flags) | |
1368 | char *string; | |
1369 | int *sindex, quoted, flags; | |
1370 | { | |
1371 | register int i, c; | |
1372 | size_t slen; | |
1373 | int pass_character, nesting_level, si, dolbrace_state; | |
1374 | char *result, *t; | |
1375 | DECLARE_MBSTATE; | |
1376 | ||
1377 | pass_character = 0; | |
1378 | nesting_level = 1; | |
1379 | slen = strlen (string + *sindex) + *sindex; | |
1380 | ||
1381 | /* The handling of dolbrace_state needs to agree with the code in parse.y: | |
1382 | parse_matched_pair(). The different initial value is to handle the | |
1383 | case where this function is called to parse the word in | |
9c2db999 CR |
1384 | ${param op word} (SX_WORD). */ |
1385 | dolbrace_state = (flags & SX_WORD) ? DOLBRACE_WORD : DOLBRACE_PARAM; | |
e192f341 CR |
1386 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && (flags & SX_POSIXEXP)) |
1387 | dolbrace_state = DOLBRACE_QUOTE; | |
1388 | ||
1389 | i = *sindex; | |
1390 | while (c = string[i]) | |
1391 | { | |
1392 | if (pass_character) | |
1393 | { | |
1394 | pass_character = 0; | |
1395 | ADVANCE_CHAR (string, slen, i); | |
1396 | continue; | |
1397 | } | |
1398 | ||
1399 | /* CTLESCs and backslashes quote the next character. */ | |
1400 | if (c == CTLESC || c == '\\') | |
1401 | { | |
1402 | pass_character++; | |
1403 | i++; | |
1404 | continue; | |
1405 | } | |
1406 | ||
1407 | if (string[i] == '$' && string[i+1] == LBRACE) | |
1408 | { | |
1409 | nesting_level++; | |
1410 | i += 2; | |
1411 | continue; | |
1412 | } | |
1413 | ||
1414 | if (c == RBRACE) | |
1415 | { | |
1416 | nesting_level--; | |
1417 | if (nesting_level == 0) | |
1418 | break; | |
1419 | i++; | |
1420 | continue; | |
1421 | } | |
1422 | ||
1423 | /* Pass the contents of old-style command substitutions through | |
1424 | verbatim. */ | |
1425 | if (c == '`') | |
1426 | { | |
1427 | si = i + 1; | |
1428 | t = string_extract (string, &si, "`", flags|SX_NOALLOC); | |
1429 | i = si + 1; | |
1430 | continue; | |
1431 | } | |
1432 | ||
1433 | /* Pass the contents of new-style command substitutions and | |
1434 | arithmetic substitutions through verbatim. */ | |
1435 | if (string[i] == '$' && string[i+1] == LPAREN) | |
1436 | { | |
1437 | si = i + 2; | |
1438 | t = extract_command_subst (string, &si, flags|SX_NOALLOC); | |
1439 | i = si + 1; | |
1440 | continue; | |
1441 | } | |
1442 | ||
1443 | #if 0 | |
1444 | /* Pass the contents of single-quoted and double-quoted strings | |
1445 | through verbatim. */ | |
1446 | if (c == '\'' || c == '"') | |
1447 | { | |
1448 | si = i + 1; | |
1449 | i = (c == '\'') ? skip_single_quoted (string, slen, si) | |
1450 | : skip_double_quoted (string, slen, si); | |
1451 | /* skip_XXX_quoted leaves index one past close quote */ | |
1452 | continue; | |
1453 | } | |
1454 | #else /* XXX - bash-4.2 */ | |
1455 | /* Pass the contents of double-quoted strings through verbatim. */ | |
1456 | if (c == '"') | |
1457 | { | |
1458 | si = i + 1; | |
1459 | i = skip_double_quoted (string, slen, si); | |
1460 | /* skip_XXX_quoted leaves index one past close quote */ | |
1461 | continue; | |
1462 | } | |
1463 | ||
1464 | if (c == '\'') | |
1465 | { | |
1466 | /*itrace("extract_dollar_brace_string: c == single quote flags = %d quoted = %d dolbrace_state = %d", flags, quoted, dolbrace_state);*/ | |
1467 | if (posixly_correct && shell_compatibility_level > 41 && dolbrace_state != DOLBRACE_QUOTE && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
1468 | ADVANCE_CHAR (string, slen, i); | |
1469 | else | |
1470 | { | |
1471 | si = i + 1; | |
1472 | i = skip_single_quoted (string, slen, si); | |
1473 | } | |
1474 | ||
1475 | continue; | |
1476 | } | |
1477 | #endif | |
1478 | ||
1479 | /* move past this character, which was not special. */ | |
1480 | ADVANCE_CHAR (string, slen, i); | |
1481 | ||
1482 | /* This logic must agree with parse.y:parse_matched_pair, since they | |
1483 | share the same defines. */ | |
1484 | if (dolbrace_state == DOLBRACE_PARAM && c == '%' && (i - *sindex) > 1) | |
1485 | dolbrace_state = DOLBRACE_QUOTE; | |
1486 | else if (dolbrace_state == DOLBRACE_PARAM && c == '#' && (i - *sindex) > 1) | |
1487 | dolbrace_state = DOLBRACE_QUOTE; | |
1488 | else if (dolbrace_state == DOLBRACE_PARAM && c == '/' && (i - *sindex) > 1) | |
1489 | dolbrace_state = DOLBRACE_QUOTE; | |
1490 | else if (dolbrace_state == DOLBRACE_PARAM && c == '^' && (i - *sindex) > 1) | |
1491 | dolbrace_state = DOLBRACE_QUOTE; | |
1492 | else if (dolbrace_state == DOLBRACE_PARAM && c == ',' && (i - *sindex) > 1) | |
1493 | dolbrace_state = DOLBRACE_QUOTE; | |
1494 | else if (dolbrace_state == DOLBRACE_PARAM && strchr ("#%^,~:-=?+/", c) != 0) | |
1495 | dolbrace_state = DOLBRACE_OP; | |
1496 | else if (dolbrace_state == DOLBRACE_OP && strchr ("#%^,~:-=?+/", c) == 0) | |
1497 | dolbrace_state = DOLBRACE_WORD; | |
1498 | } | |
1499 | ||
1500 | if (c == 0 && nesting_level) | |
1501 | { | |
1502 | if (no_longjmp_on_fatal_error == 0) | |
1503 | { /* { */ | |
1504 | report_error (_("bad substitution: no closing `%s' in %s"), "}", string); | |
1505 | last_command_exit_value = EXECUTION_FAILURE; | |
1506 | exp_jump_to_top_level (DISCARD); | |
1507 | } | |
1508 | else | |
1509 | { | |
1510 | *sindex = i; | |
1511 | return ((char *)NULL); | |
1512 | } | |
1513 | } | |
1514 | ||
1515 | result = (flags & SX_NOALLOC) ? (char *)NULL : substring (string, *sindex, i); | |
1516 | *sindex = i; | |
1517 | ||
1518 | return (result); | |
1519 | } | |
1520 | ||
1521 | /* Remove backslashes which are quoting backquotes from STRING. Modifies | |
1522 | STRING, and returns a pointer to it. */ | |
1523 | char * | |
1524 | de_backslash (string) | |
1525 | char *string; | |
1526 | { | |
1527 | register size_t slen; | |
1528 | register int i, j, prev_i; | |
1529 | DECLARE_MBSTATE; | |
1530 | ||
1531 | slen = strlen (string); | |
1532 | i = j = 0; | |
1533 | ||
1534 | /* Loop copying string[i] to string[j], i >= j. */ | |
1535 | while (i < slen) | |
1536 | { | |
1537 | if (string[i] == '\\' && (string[i + 1] == '`' || string[i + 1] == '\\' || | |
1538 | string[i + 1] == '$')) | |
1539 | i++; | |
1540 | prev_i = i; | |
1541 | ADVANCE_CHAR (string, slen, i); | |
1542 | if (j < prev_i) | |
1543 | do string[j++] = string[prev_i++]; while (prev_i < i); | |
1544 | else | |
1545 | j = i; | |
1546 | } | |
1547 | string[j] = '\0'; | |
1548 | ||
1549 | return (string); | |
1550 | } | |
1551 | ||
1552 | #if 0 | |
1553 | /*UNUSED*/ | |
1554 | /* Replace instances of \! in a string with !. */ | |
1555 | void | |
1556 | unquote_bang (string) | |
1557 | char *string; | |
1558 | { | |
1559 | register int i, j; | |
1560 | register char *temp; | |
1561 | ||
1562 | temp = (char *)xmalloc (1 + strlen (string)); | |
1563 | ||
1564 | for (i = 0, j = 0; (temp[j] = string[i]); i++, j++) | |
1565 | { | |
1566 | if (string[i] == '\\' && string[i + 1] == '!') | |
1567 | { | |
1568 | temp[j] = '!'; | |
1569 | i++; | |
1570 | } | |
1571 | } | |
1572 | strcpy (string, temp); | |
1573 | free (temp); | |
1574 | } | |
1575 | #endif | |
1576 | ||
1577 | #define CQ_RETURN(x) do { no_longjmp_on_fatal_error = 0; return (x); } while (0) | |
1578 | ||
1579 | /* This function assumes s[i] == open; returns with s[ret] == close; used to | |
1580 | parse array subscripts. FLAGS & 1 means to not attempt to skip over | |
1581 | matched pairs of quotes or backquotes, or skip word expansions; it is | |
1582 | intended to be used after expansion has been performed and during final | |
1583 | assignment parsing (see arrayfunc.c:assign_compound_array_list()). */ | |
1584 | static int | |
1585 | skip_matched_pair (string, start, open, close, flags) | |
1586 | const char *string; | |
1587 | int start, open, close, flags; | |
1588 | { | |
1589 | int i, pass_next, backq, si, c, count; | |
1590 | size_t slen; | |
1591 | char *temp, *ss; | |
1592 | DECLARE_MBSTATE; | |
1593 | ||
1594 | slen = strlen (string + start) + start; | |
1595 | no_longjmp_on_fatal_error = 1; | |
1596 | ||
1597 | i = start + 1; /* skip over leading bracket */ | |
1598 | count = 1; | |
1599 | pass_next = backq = 0; | |
1600 | ss = (char *)string; | |
1601 | while (c = string[i]) | |
1602 | { | |
1603 | if (pass_next) | |
1604 | { | |
1605 | pass_next = 0; | |
1606 | if (c == 0) | |
1607 | CQ_RETURN(i); | |
1608 | ADVANCE_CHAR (string, slen, i); | |
1609 | continue; | |
1610 | } | |
1611 | else if (c == '\\') | |
1612 | { | |
1613 | pass_next = 1; | |
1614 | i++; | |
1615 | continue; | |
1616 | } | |
1617 | else if (backq) | |
1618 | { | |
1619 | if (c == '`') | |
1620 | backq = 0; | |
1621 | ADVANCE_CHAR (string, slen, i); | |
1622 | continue; | |
1623 | } | |
1624 | else if ((flags & 1) == 0 && c == '`') | |
1625 | { | |
1626 | backq = 1; | |
1627 | i++; | |
1628 | continue; | |
1629 | } | |
1630 | else if ((flags & 1) == 0 && c == open) | |
1631 | { | |
1632 | count++; | |
1633 | i++; | |
1634 | continue; | |
1635 | } | |
1636 | else if (c == close) | |
1637 | { | |
1638 | count--; | |
1639 | if (count == 0) | |
1640 | break; | |
1641 | i++; | |
1642 | continue; | |
1643 | } | |
1644 | else if ((flags & 1) == 0 && (c == '\'' || c == '"')) | |
1645 | { | |
1646 | i = (c == '\'') ? skip_single_quoted (ss, slen, ++i) | |
1647 | : skip_double_quoted (ss, slen, ++i); | |
1648 | /* no increment, the skip functions increment past the closing quote. */ | |
1649 | } | |
1650 | else if ((flags&1) == 0 && c == '$' && (string[i+1] == LPAREN || string[i+1] == LBRACE)) | |
1651 | { | |
1652 | si = i + 2; | |
1653 | if (string[si] == '\0') | |
1654 | CQ_RETURN(si); | |
1655 | ||
1656 | if (string[i+1] == LPAREN) | |
1657 | temp = extract_delimited_string (ss, &si, "$(", "(", ")", SX_NOALLOC|SX_COMMAND); /* ) */ | |
1658 | else | |
1659 | temp = extract_dollar_brace_string (ss, &si, 0, SX_NOALLOC); | |
1660 | i = si; | |
1661 | if (string[i] == '\0') /* don't increment i past EOS in loop */ | |
1662 | break; | |
1663 | i++; | |
1664 | continue; | |
1665 | } | |
1666 | else | |
1667 | ADVANCE_CHAR (string, slen, i); | |
1668 | } | |
1669 | ||
1670 | CQ_RETURN(i); | |
1671 | } | |
1672 | ||
1673 | #if defined (ARRAY_VARS) | |
1674 | int | |
1675 | skipsubscript (string, start, flags) | |
1676 | const char *string; | |
1677 | int start, flags; | |
1678 | { | |
1679 | return (skip_matched_pair (string, start, '[', ']', flags)); | |
1680 | } | |
1681 | #endif | |
1682 | ||
1683 | /* Skip characters in STRING until we find a character in DELIMS, and return | |
1684 | the index of that character. START is the index into string at which we | |
1685 | begin. This is similar in spirit to strpbrk, but it returns an index into | |
1686 | STRING and takes a starting index. This little piece of code knows quite | |
1687 | a lot of shell syntax. It's very similar to skip_double_quoted and other | |
1688 | functions of that ilk. */ | |
1689 | int | |
1690 | skip_to_delim (string, start, delims, flags) | |
1691 | char *string; | |
1692 | int start; | |
1693 | char *delims; | |
1694 | int flags; | |
1695 | { | |
1696 | int i, pass_next, backq, si, c, invert, skipquote, skipcmd; | |
1697 | size_t slen; | |
1698 | char *temp, open[3]; | |
1699 | DECLARE_MBSTATE; | |
1700 | ||
1701 | slen = strlen (string + start) + start; | |
1702 | if (flags & SD_NOJMP) | |
1703 | no_longjmp_on_fatal_error = 1; | |
1704 | invert = (flags & SD_INVERT); | |
1705 | skipcmd = (flags & SD_NOSKIPCMD) == 0; | |
1706 | ||
1707 | i = start; | |
1708 | pass_next = backq = 0; | |
1709 | while (c = string[i]) | |
1710 | { | |
1711 | /* If this is non-zero, we should not let quote characters be delimiters | |
1712 | and the current character is a single or double quote. We should not | |
1713 | test whether or not it's a delimiter until after we skip single- or | |
1714 | double-quoted strings. */ | |
1715 | skipquote = ((flags & SD_NOQUOTEDELIM) && (c == '\'' || c =='"')); | |
1716 | if (pass_next) | |
1717 | { | |
1718 | pass_next = 0; | |
1719 | if (c == 0) | |
1720 | CQ_RETURN(i); | |
1721 | ADVANCE_CHAR (string, slen, i); | |
1722 | continue; | |
1723 | } | |
1724 | else if (c == '\\') | |
1725 | { | |
1726 | pass_next = 1; | |
1727 | i++; | |
1728 | continue; | |
1729 | } | |
1730 | else if (backq) | |
1731 | { | |
1732 | if (c == '`') | |
1733 | backq = 0; | |
1734 | ADVANCE_CHAR (string, slen, i); | |
1735 | continue; | |
1736 | } | |
1737 | else if (c == '`') | |
1738 | { | |
1739 | backq = 1; | |
1740 | i++; | |
1741 | continue; | |
1742 | } | |
1743 | else if (skipquote == 0 && invert == 0 && member (c, delims)) | |
1744 | break; | |
1745 | else if (c == '\'' || c == '"') | |
1746 | { | |
1747 | i = (c == '\'') ? skip_single_quoted (string, slen, ++i) | |
1748 | : skip_double_quoted (string, slen, ++i); | |
1749 | /* no increment, the skip functions increment past the closing quote. */ | |
1750 | } | |
1751 | else if (c == '$' && ((skipcmd && string[i+1] == LPAREN) || string[i+1] == LBRACE)) | |
1752 | { | |
1753 | si = i + 2; | |
1754 | if (string[si] == '\0') | |
1755 | CQ_RETURN(si); | |
1756 | ||
1757 | if (string[i+1] == LPAREN) | |
1758 | temp = extract_delimited_string (string, &si, "$(", "(", ")", SX_NOALLOC|SX_COMMAND); /* ) */ | |
1759 | else | |
1760 | temp = extract_dollar_brace_string (string, &si, 0, SX_NOALLOC); | |
1761 | i = si; | |
1762 | if (string[i] == '\0') /* don't increment i past EOS in loop */ | |
1763 | break; | |
1764 | i++; | |
1765 | continue; | |
1766 | } | |
1767 | #if defined (PROCESS_SUBSTITUTION) | |
1768 | else if (skipcmd && (c == '<' || c == '>') && string[i+1] == LPAREN) | |
1769 | { | |
1770 | si = i + 2; | |
1771 | if (string[si] == '\0') | |
1772 | CQ_RETURN(si); | |
1773 | temp = extract_process_subst (string, (c == '<') ? "<(" : ">(", &si); | |
1774 | i = si; | |
1775 | if (string[i] == '\0') | |
1776 | break; | |
1777 | i++; | |
1778 | continue; | |
1779 | } | |
1780 | #endif /* PROCESS_SUBSTITUTION */ | |
1781 | #if defined (EXTENDED_GLOB) | |
1782 | else if ((flags & SD_EXTGLOB) && extended_glob && string[i+1] == LPAREN && member (c, "?*+!@")) | |
1783 | { | |
1784 | si = i + 2; | |
1785 | if (string[si] == '\0') | |
1786 | CQ_RETURN(si); | |
1787 | ||
1788 | open[0] = c; | |
1789 | open[1] = LPAREN; | |
1790 | open[2] = '\0'; | |
1791 | temp = extract_delimited_string (string, &si, open, "(", ")", SX_NOALLOC); /* ) */ | |
1792 | ||
1793 | i = si; | |
1794 | if (string[i] == '\0') /* don't increment i past EOS in loop */ | |
1795 | break; | |
1796 | i++; | |
1797 | continue; | |
1798 | } | |
1799 | #endif | |
1800 | else if ((skipquote || invert) && (member (c, delims) == 0)) | |
1801 | break; | |
1802 | else | |
1803 | ADVANCE_CHAR (string, slen, i); | |
1804 | } | |
1805 | ||
1806 | CQ_RETURN(i); | |
1807 | } | |
1808 | ||
1809 | #if defined (READLINE) | |
1810 | /* Return 1 if the portion of STRING ending at EINDEX is quoted (there is | |
1811 | an unclosed quoted string), or if the character at EINDEX is quoted | |
1812 | by a backslash. NO_LONGJMP_ON_FATAL_ERROR is used to flag that the various | |
1813 | single and double-quoted string parsing functions should not return an | |
1814 | error if there are unclosed quotes or braces. The characters that this | |
1815 | recognizes need to be the same as the contents of | |
1816 | rl_completer_quote_characters. */ | |
1817 | ||
1818 | int | |
1819 | char_is_quoted (string, eindex) | |
1820 | char *string; | |
1821 | int eindex; | |
1822 | { | |
1823 | int i, pass_next, c; | |
1824 | size_t slen; | |
1825 | DECLARE_MBSTATE; | |
1826 | ||
1827 | slen = strlen (string); | |
1828 | no_longjmp_on_fatal_error = 1; | |
1829 | i = pass_next = 0; | |
1830 | while (i <= eindex) | |
1831 | { | |
1832 | c = string[i]; | |
1833 | ||
1834 | if (pass_next) | |
1835 | { | |
1836 | pass_next = 0; | |
1837 | if (i >= eindex) /* XXX was if (i >= eindex - 1) */ | |
1838 | CQ_RETURN(1); | |
1839 | ADVANCE_CHAR (string, slen, i); | |
1840 | continue; | |
1841 | } | |
1842 | else if (c == '\\') | |
1843 | { | |
1844 | pass_next = 1; | |
1845 | i++; | |
1846 | continue; | |
1847 | } | |
1848 | else if (c == '\'' || c == '"') | |
1849 | { | |
1850 | i = (c == '\'') ? skip_single_quoted (string, slen, ++i) | |
1851 | : skip_double_quoted (string, slen, ++i); | |
1852 | if (i > eindex) | |
1853 | CQ_RETURN(1); | |
1854 | /* no increment, the skip_xxx functions go one past end */ | |
1855 | } | |
1856 | else | |
1857 | ADVANCE_CHAR (string, slen, i); | |
1858 | } | |
1859 | ||
1860 | CQ_RETURN(0); | |
1861 | } | |
1862 | ||
1863 | int | |
1864 | unclosed_pair (string, eindex, openstr) | |
1865 | char *string; | |
1866 | int eindex; | |
1867 | char *openstr; | |
1868 | { | |
1869 | int i, pass_next, openc, olen; | |
1870 | size_t slen; | |
1871 | DECLARE_MBSTATE; | |
1872 | ||
1873 | slen = strlen (string); | |
1874 | olen = strlen (openstr); | |
1875 | i = pass_next = openc = 0; | |
1876 | while (i <= eindex) | |
1877 | { | |
1878 | if (pass_next) | |
1879 | { | |
1880 | pass_next = 0; | |
1881 | if (i >= eindex) /* XXX was if (i >= eindex - 1) */ | |
1882 | return 0; | |
1883 | ADVANCE_CHAR (string, slen, i); | |
1884 | continue; | |
1885 | } | |
1886 | else if (string[i] == '\\') | |
1887 | { | |
1888 | pass_next = 1; | |
1889 | i++; | |
1890 | continue; | |
1891 | } | |
1892 | else if (STREQN (string + i, openstr, olen)) | |
1893 | { | |
1894 | openc = 1 - openc; | |
1895 | i += olen; | |
1896 | } | |
1897 | else if (string[i] == '\'' || string[i] == '"') | |
1898 | { | |
1899 | i = (string[i] == '\'') ? skip_single_quoted (string, slen, i) | |
1900 | : skip_double_quoted (string, slen, i); | |
1901 | if (i > eindex) | |
1902 | return 0; | |
1903 | } | |
1904 | else | |
1905 | ADVANCE_CHAR (string, slen, i); | |
1906 | } | |
1907 | return (openc); | |
1908 | } | |
1909 | ||
1910 | /* Split STRING (length SLEN) at DELIMS, and return a WORD_LIST with the | |
1911 | individual words. If DELIMS is NULL, the current value of $IFS is used | |
1912 | to split the string, and the function follows the shell field splitting | |
1913 | rules. SENTINEL is an index to look for. NWP, if non-NULL, | |
1914 | gets the number of words in the returned list. CWP, if non-NULL, gets | |
1915 | the index of the word containing SENTINEL. Non-whitespace chars in | |
1916 | DELIMS delimit separate fields. */ | |
1917 | WORD_LIST * | |
1918 | split_at_delims (string, slen, delims, sentinel, flags, nwp, cwp) | |
1919 | char *string; | |
1920 | int slen; | |
1921 | char *delims; | |
1922 | int sentinel, flags; | |
1923 | int *nwp, *cwp; | |
1924 | { | |
1925 | int ts, te, i, nw, cw, ifs_split, dflags; | |
1926 | char *token, *d, *d2; | |
1927 | WORD_LIST *ret, *tl; | |
1928 | ||
1929 | if (string == 0 || *string == '\0') | |
1930 | { | |
1931 | if (nwp) | |
1932 | *nwp = 0; | |
1933 | if (cwp) | |
1934 | *cwp = 0; | |
1935 | return ((WORD_LIST *)NULL); | |
1936 | } | |
1937 | ||
1938 | d = (delims == 0) ? ifs_value : delims; | |
1939 | ifs_split = delims == 0; | |
1940 | ||
1941 | /* Make d2 the non-whitespace characters in delims */ | |
1942 | d2 = 0; | |
1943 | if (delims) | |
1944 | { | |
1945 | size_t slength; | |
1946 | #if defined (HANDLE_MULTIBYTE) | |
1947 | size_t mblength = 1; | |
1948 | #endif | |
1949 | DECLARE_MBSTATE; | |
1950 | ||
1951 | slength = strlen (delims); | |
1952 | d2 = (char *)xmalloc (slength + 1); | |
1953 | i = ts = 0; | |
1954 | while (delims[i]) | |
1955 | { | |
1956 | #if defined (HANDLE_MULTIBYTE) | |
1957 | mbstate_t state_bak; | |
1958 | state_bak = state; | |
1959 | mblength = MBRLEN (delims + i, slength, &state); | |
1960 | if (MB_INVALIDCH (mblength)) | |
1961 | state = state_bak; | |
1962 | else if (mblength > 1) | |
1963 | { | |
1964 | memcpy (d2 + ts, delims + i, mblength); | |
1965 | ts += mblength; | |
1966 | i += mblength; | |
1967 | slength -= mblength; | |
1968 | continue; | |
1969 | } | |
1970 | #endif | |
1971 | if (whitespace (delims[i]) == 0) | |
1972 | d2[ts++] = delims[i]; | |
1973 | ||
1974 | i++; | |
1975 | slength--; | |
1976 | } | |
1977 | d2[ts] = '\0'; | |
1978 | } | |
1979 | ||
1980 | ret = (WORD_LIST *)NULL; | |
1981 | ||
1982 | /* Remove sequences of whitespace characters at the start of the string, as | |
1983 | long as those characters are delimiters. */ | |
1984 | for (i = 0; member (string[i], d) && spctabnl (string[i]); i++) | |
1985 | ; | |
1986 | if (string[i] == '\0') | |
1987 | return (ret); | |
1988 | ||
1989 | ts = i; | |
1990 | nw = 0; | |
1991 | cw = -1; | |
1992 | dflags = flags|SD_NOJMP; | |
1993 | while (1) | |
1994 | { | |
1995 | te = skip_to_delim (string, ts, d, dflags); | |
1996 | ||
1997 | /* If we have a non-whitespace delimiter character, use it to make a | |
1998 | separate field. This is just about what $IFS splitting does and | |
1999 | is closer to the behavior of the shell parser. */ | |
2000 | if (ts == te && d2 && member (string[ts], d2)) | |
2001 | { | |
2002 | te = ts + 1; | |
2003 | /* If we're using IFS splitting, the non-whitespace delimiter char | |
2004 | and any additional IFS whitespace delimits a field. */ | |
2005 | if (ifs_split) | |
2006 | while (member (string[te], d) && spctabnl (string[te])) | |
2007 | te++; | |
2008 | else | |
2009 | while (member (string[te], d2)) | |
2010 | te++; | |
2011 | } | |
2012 | ||
2013 | token = substring (string, ts, te); | |
2014 | ||
2015 | ret = add_string_to_list (token, ret); | |
2016 | free (token); | |
2017 | nw++; | |
2018 | ||
2019 | if (sentinel >= ts && sentinel <= te) | |
2020 | cw = nw; | |
2021 | ||
2022 | /* If the cursor is at whitespace just before word start, set the | |
2023 | sentinel word to the current word. */ | |
2024 | if (cwp && cw == -1 && sentinel == ts-1) | |
2025 | cw = nw; | |
2026 | ||
2027 | /* If the cursor is at whitespace between two words, make a new, empty | |
2028 | word, add it before (well, after, since the list is in reverse order) | |
2029 | the word we just added, and set the current word to that one. */ | |
2030 | if (cwp && cw == -1 && sentinel < ts) | |
2031 | { | |
2032 | tl = make_word_list (make_word (""), ret->next); | |
2033 | ret->next = tl; | |
2034 | cw = nw; | |
2035 | nw++; | |
2036 | } | |
2037 | ||
2038 | if (string[te] == 0) | |
2039 | break; | |
2040 | ||
2041 | i = te; | |
2042 | while (member (string[i], d) && (ifs_split || spctabnl(string[i]))) | |
2043 | i++; | |
2044 | ||
2045 | if (string[i]) | |
2046 | ts = i; | |
2047 | else | |
2048 | break; | |
2049 | } | |
2050 | ||
2051 | /* Special case for SENTINEL at the end of STRING. If we haven't found | |
2052 | the word containing SENTINEL yet, and the index we're looking for is at | |
2053 | the end of STRING (or past the end of the previously-found token, | |
2054 | possible if the end of the line is composed solely of IFS whitespace) | |
2055 | add an additional null argument and set the current word pointer to that. */ | |
2056 | if (cwp && cw == -1 && (sentinel >= slen || sentinel >= te)) | |
2057 | { | |
2058 | if (whitespace (string[sentinel - 1])) | |
2059 | { | |
2060 | token = ""; | |
2061 | ret = add_string_to_list (token, ret); | |
2062 | nw++; | |
2063 | } | |
2064 | cw = nw; | |
2065 | } | |
2066 | ||
2067 | if (nwp) | |
2068 | *nwp = nw; | |
2069 | if (cwp) | |
2070 | *cwp = cw; | |
2071 | ||
2072 | return (REVERSE_LIST (ret, WORD_LIST *)); | |
2073 | } | |
2074 | #endif /* READLINE */ | |
2075 | ||
2076 | #if 0 | |
2077 | /* UNUSED */ | |
2078 | /* Extract the name of the variable to bind to from the assignment string. */ | |
2079 | char * | |
2080 | assignment_name (string) | |
2081 | char *string; | |
2082 | { | |
2083 | int offset; | |
2084 | char *temp; | |
2085 | ||
2086 | offset = assignment (string, 0); | |
2087 | if (offset == 0) | |
2088 | return (char *)NULL; | |
2089 | temp = substring (string, 0, offset); | |
2090 | return (temp); | |
2091 | } | |
2092 | #endif | |
2093 | ||
2094 | /* **************************************************************** */ | |
2095 | /* */ | |
2096 | /* Functions to convert strings to WORD_LISTs and vice versa */ | |
2097 | /* */ | |
2098 | /* **************************************************************** */ | |
2099 | ||
2100 | /* Return a single string of all the words in LIST. SEP is the separator | |
2101 | to put between individual elements of LIST in the output string. */ | |
2102 | char * | |
2103 | string_list_internal (list, sep) | |
2104 | WORD_LIST *list; | |
2105 | char *sep; | |
2106 | { | |
2107 | register WORD_LIST *t; | |
2108 | char *result, *r; | |
2109 | int word_len, sep_len, result_size; | |
2110 | ||
2111 | if (list == 0) | |
2112 | return ((char *)NULL); | |
2113 | ||
2114 | /* Short-circuit quickly if we don't need to separate anything. */ | |
2115 | if (list->next == 0) | |
2116 | return (savestring (list->word->word)); | |
2117 | ||
2118 | /* This is nearly always called with either sep[0] == 0 or sep[1] == 0. */ | |
2119 | sep_len = STRLEN (sep); | |
2120 | result_size = 0; | |
2121 | ||
2122 | for (t = list; t; t = t->next) | |
2123 | { | |
2124 | if (t != list) | |
2125 | result_size += sep_len; | |
2126 | result_size += strlen (t->word->word); | |
2127 | } | |
2128 | ||
2129 | r = result = (char *)xmalloc (result_size + 1); | |
2130 | ||
2131 | for (t = list; t; t = t->next) | |
2132 | { | |
2133 | if (t != list && sep_len) | |
2134 | { | |
2135 | if (sep_len > 1) | |
2136 | { | |
2137 | FASTCOPY (sep, r, sep_len); | |
2138 | r += sep_len; | |
2139 | } | |
2140 | else | |
2141 | *r++ = sep[0]; | |
2142 | } | |
2143 | ||
2144 | word_len = strlen (t->word->word); | |
2145 | FASTCOPY (t->word->word, r, word_len); | |
2146 | r += word_len; | |
2147 | } | |
2148 | ||
2149 | *r = '\0'; | |
2150 | return (result); | |
2151 | } | |
2152 | ||
2153 | /* Return a single string of all the words present in LIST, separating | |
2154 | each word with a space. */ | |
2155 | char * | |
2156 | string_list (list) | |
2157 | WORD_LIST *list; | |
2158 | { | |
2159 | return (string_list_internal (list, " ")); | |
2160 | } | |
2161 | ||
2162 | /* An external interface that can be used by the rest of the shell to | |
2163 | obtain a string containing the first character in $IFS. Handles all | |
2164 | the multibyte complications. If LENP is non-null, it is set to the | |
2165 | length of the returned string. */ | |
2166 | char * | |
2167 | ifs_firstchar (lenp) | |
2168 | int *lenp; | |
2169 | { | |
2170 | char *ret; | |
2171 | int len; | |
2172 | ||
2173 | ret = xmalloc (MB_LEN_MAX + 1); | |
2174 | #if defined (HANDLE_MULTIBYTE) | |
2175 | if (ifs_firstc_len == 1) | |
2176 | { | |
2177 | ret[0] = ifs_firstc[0]; | |
2178 | ret[1] = '\0'; | |
2179 | len = ret[0] ? 1 : 0; | |
2180 | } | |
2181 | else | |
2182 | { | |
2183 | memcpy (ret, ifs_firstc, ifs_firstc_len); | |
2184 | ret[len = ifs_firstc_len] = '\0'; | |
2185 | } | |
2186 | #else | |
2187 | ret[0] = ifs_firstc; | |
2188 | ret[1] = '\0'; | |
2189 | len = ret[0] ? 0 : 1; | |
2190 | #endif | |
2191 | ||
2192 | if (lenp) | |
2193 | *lenp = len; | |
2194 | ||
2195 | return ret; | |
2196 | } | |
2197 | ||
2198 | /* Return a single string of all the words present in LIST, obeying the | |
2199 | quoting rules for "$*", to wit: (P1003.2, draft 11, 3.5.2) "If the | |
2200 | expansion [of $*] appears within a double quoted string, it expands | |
2201 | to a single field with the value of each parameter separated by the | |
2202 | first character of the IFS variable, or by a <space> if IFS is unset." */ | |
2203 | char * | |
2204 | string_list_dollar_star (list) | |
2205 | WORD_LIST *list; | |
2206 | { | |
2207 | char *ret; | |
2208 | #if defined (HANDLE_MULTIBYTE) | |
2209 | # if defined (__GNUC__) | |
2210 | char sep[MB_CUR_MAX + 1]; | |
2211 | # else | |
2212 | char *sep = 0; | |
2213 | # endif | |
2214 | #else | |
2215 | char sep[2]; | |
2216 | #endif | |
2217 | ||
2218 | #if defined (HANDLE_MULTIBYTE) | |
2219 | # if !defined (__GNUC__) | |
2220 | sep = (char *)xmalloc (MB_CUR_MAX + 1); | |
2221 | # endif /* !__GNUC__ */ | |
2222 | if (ifs_firstc_len == 1) | |
2223 | { | |
2224 | sep[0] = ifs_firstc[0]; | |
2225 | sep[1] = '\0'; | |
2226 | } | |
2227 | else | |
2228 | { | |
2229 | memcpy (sep, ifs_firstc, ifs_firstc_len); | |
2230 | sep[ifs_firstc_len] = '\0'; | |
2231 | } | |
2232 | #else | |
2233 | sep[0] = ifs_firstc; | |
2234 | sep[1] = '\0'; | |
2235 | #endif | |
2236 | ||
2237 | ret = string_list_internal (list, sep); | |
2238 | #if defined (HANDLE_MULTIBYTE) && !defined (__GNUC__) | |
2239 | free (sep); | |
2240 | #endif | |
2241 | return ret; | |
2242 | } | |
2243 | ||
2244 | /* Turn $@ into a string. If (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
2245 | is non-zero, the $@ appears within double quotes, and we should quote | |
2246 | the list before converting it into a string. If IFS is unset, and the | |
2247 | word is not quoted, we just need to quote CTLESC and CTLNUL characters | |
2248 | in the words in the list, because the default value of $IFS is | |
2249 | <space><tab><newline>, IFS characters in the words in the list should | |
2250 | also be split. If IFS is null, and the word is not quoted, we need | |
2251 | to quote the words in the list to preserve the positional parameters | |
2252 | exactly. */ | |
2253 | char * | |
2254 | string_list_dollar_at (list, quoted) | |
2255 | WORD_LIST *list; | |
2256 | int quoted; | |
2257 | { | |
2258 | char *ifs, *ret; | |
2259 | #if defined (HANDLE_MULTIBYTE) | |
2260 | # if defined (__GNUC__) | |
2261 | char sep[MB_CUR_MAX + 1]; | |
2262 | # else | |
2263 | char *sep = 0; | |
2264 | # endif /* !__GNUC__ */ | |
2265 | #else | |
2266 | char sep[2]; | |
2267 | #endif | |
2268 | WORD_LIST *tlist; | |
2269 | ||
2270 | /* XXX this could just be ifs = ifs_value; */ | |
2271 | ifs = ifs_var ? value_cell (ifs_var) : (char *)0; | |
2272 | ||
2273 | #if defined (HANDLE_MULTIBYTE) | |
2274 | # if !defined (__GNUC__) | |
2275 | sep = (char *)xmalloc (MB_CUR_MAX + 1); | |
2276 | # endif /* !__GNUC__ */ | |
2277 | if (ifs && *ifs) | |
2278 | { | |
2279 | if (ifs_firstc_len == 1) | |
2280 | { | |
2281 | sep[0] = ifs_firstc[0]; | |
2282 | sep[1] = '\0'; | |
2283 | } | |
2284 | else | |
2285 | { | |
2286 | memcpy (sep, ifs_firstc, ifs_firstc_len); | |
2287 | sep[ifs_firstc_len] = '\0'; | |
2288 | } | |
2289 | } | |
2290 | else | |
2291 | { | |
2292 | sep[0] = ' '; | |
2293 | sep[1] = '\0'; | |
2294 | } | |
2295 | #else | |
2296 | sep[0] = (ifs == 0 || *ifs == 0) ? ' ' : *ifs; | |
2297 | sep[1] = '\0'; | |
2298 | #endif | |
2299 | ||
2300 | /* XXX -- why call quote_list if ifs == 0? we can get away without doing | |
2301 | it now that quote_escapes quotes spaces */ | |
2302 | tlist = (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES|Q_PATQUOTE)) | |
2303 | ? quote_list (list) | |
2304 | : list_quote_escapes (list); | |
2305 | ||
2306 | ret = string_list_internal (tlist, sep); | |
2307 | #if defined (HANDLE_MULTIBYTE) && !defined (__GNUC__) | |
2308 | free (sep); | |
2309 | #endif | |
2310 | return ret; | |
2311 | } | |
2312 | ||
2313 | /* Turn the positional paramters into a string, understanding quoting and | |
2314 | the various subtleties of using the first character of $IFS as the | |
2315 | separator. Calls string_list_dollar_at, string_list_dollar_star, and | |
2316 | string_list as appropriate. */ | |
2317 | char * | |
2318 | string_list_pos_params (pchar, list, quoted) | |
2319 | int pchar; | |
2320 | WORD_LIST *list; | |
2321 | int quoted; | |
2322 | { | |
2323 | char *ret; | |
2324 | WORD_LIST *tlist; | |
2325 | ||
2326 | if (pchar == '*' && (quoted & Q_DOUBLE_QUOTES)) | |
2327 | { | |
2328 | tlist = quote_list (list); | |
2329 | word_list_remove_quoted_nulls (tlist); | |
2330 | ret = string_list_dollar_star (tlist); | |
2331 | } | |
2332 | else if (pchar == '*' && (quoted & Q_HERE_DOCUMENT)) | |
2333 | { | |
2334 | tlist = quote_list (list); | |
2335 | word_list_remove_quoted_nulls (tlist); | |
2336 | ret = string_list (tlist); | |
2337 | } | |
2338 | else if (pchar == '*') | |
2339 | { | |
2340 | /* Even when unquoted, string_list_dollar_star does the right thing | |
2341 | making sure that the first character of $IFS is used as the | |
2342 | separator. */ | |
2343 | ret = string_list_dollar_star (list); | |
2344 | } | |
2345 | else if (pchar == '@' && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
2346 | /* We use string_list_dollar_at, but only if the string is quoted, since | |
2347 | that quotes the escapes if it's not, which we don't want. We could | |
2348 | use string_list (the old code did), but that doesn't do the right | |
2349 | thing if the first character of $IFS is not a space. We use | |
2350 | string_list_dollar_star if the string is unquoted so we make sure that | |
2351 | the elements of $@ are separated by the first character of $IFS for | |
2352 | later splitting. */ | |
2353 | ret = string_list_dollar_at (list, quoted); | |
2354 | else if (pchar == '@') | |
2355 | ret = string_list_dollar_star (list); | |
2356 | else | |
2357 | ret = string_list ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) ? quote_list (list) : list); | |
2358 | ||
2359 | return ret; | |
2360 | } | |
2361 | ||
2362 | /* Return the list of words present in STRING. Separate the string into | |
2363 | words at any of the characters found in SEPARATORS. If QUOTED is | |
2364 | non-zero then word in the list will have its quoted flag set, otherwise | |
2365 | the quoted flag is left as make_word () deemed fit. | |
2366 | ||
2367 | This obeys the P1003.2 word splitting semantics. If `separators' is | |
2368 | exactly <space><tab><newline>, then the splitting algorithm is that of | |
2369 | the Bourne shell, which treats any sequence of characters from `separators' | |
2370 | as a delimiter. If IFS is unset, which results in `separators' being set | |
2371 | to "", no splitting occurs. If separators has some other value, the | |
2372 | following rules are applied (`IFS white space' means zero or more | |
2373 | occurrences of <space>, <tab>, or <newline>, as long as those characters | |
2374 | are in `separators'): | |
2375 | ||
2376 | 1) IFS white space is ignored at the start and the end of the | |
2377 | string. | |
2378 | 2) Each occurrence of a character in `separators' that is not | |
2379 | IFS white space, along with any adjacent occurrences of | |
2380 | IFS white space delimits a field. | |
2381 | 3) Any nonzero-length sequence of IFS white space delimits a field. | |
2382 | */ | |
2383 | ||
2384 | /* BEWARE! list_string strips null arguments. Don't call it twice and | |
2385 | expect to have "" preserved! */ | |
2386 | ||
2387 | /* This performs word splitting and quoted null character removal on | |
2388 | STRING. */ | |
2389 | #define issep(c) \ | |
2390 | (((separators)[0]) ? ((separators)[1] ? isifs(c) \ | |
2391 | : (c) == (separators)[0]) \ | |
2392 | : 0) | |
2393 | ||
2394 | WORD_LIST * | |
2395 | list_string (string, separators, quoted) | |
2396 | register char *string, *separators; | |
2397 | int quoted; | |
2398 | { | |
2399 | WORD_LIST *result; | |
2400 | WORD_DESC *t; | |
2401 | char *current_word, *s; | |
2402 | int sindex, sh_style_split, whitesep, xflags; | |
2403 | size_t slen; | |
2404 | ||
2405 | if (!string || !*string) | |
2406 | return ((WORD_LIST *)NULL); | |
2407 | ||
2408 | sh_style_split = separators && separators[0] == ' ' && | |
2409 | separators[1] == '\t' && | |
2410 | separators[2] == '\n' && | |
2411 | separators[3] == '\0'; | |
2412 | for (xflags = 0, s = ifs_value; s && *s; s++) | |
2413 | { | |
2414 | if (*s == CTLESC) xflags |= SX_NOCTLESC; | |
2415 | else if (*s == CTLNUL) xflags |= SX_NOESCCTLNUL; | |
2416 | } | |
2417 | ||
2418 | slen = 0; | |
2419 | /* Remove sequences of whitespace at the beginning of STRING, as | |
2420 | long as those characters appear in IFS. Do not do this if | |
2421 | STRING is quoted or if there are no separator characters. */ | |
2422 | if (!quoted || !separators || !*separators) | |
2423 | { | |
2424 | for (s = string; *s && spctabnl (*s) && issep (*s); s++); | |
2425 | ||
2426 | if (!*s) | |
2427 | return ((WORD_LIST *)NULL); | |
2428 | ||
2429 | string = s; | |
2430 | } | |
2431 | ||
2432 | /* OK, now STRING points to a word that does not begin with white space. | |
2433 | The splitting algorithm is: | |
2434 | extract a word, stopping at a separator | |
2435 | skip sequences of spc, tab, or nl as long as they are separators | |
2436 | This obeys the field splitting rules in Posix.2. */ | |
2437 | slen = (MB_CUR_MAX > 1) ? strlen (string) : 1; | |
2438 | for (result = (WORD_LIST *)NULL, sindex = 0; string[sindex]; ) | |
2439 | { | |
2440 | /* Don't need string length in ADVANCE_CHAR or string_extract_verbatim | |
2441 | unless multibyte chars are possible. */ | |
2442 | current_word = string_extract_verbatim (string, slen, &sindex, separators, xflags); | |
2443 | if (current_word == 0) | |
2444 | break; | |
2445 | ||
2446 | /* If we have a quoted empty string, add a quoted null argument. We | |
2447 | want to preserve the quoted null character iff this is a quoted | |
2448 | empty string; otherwise the quoted null characters are removed | |
2449 | below. */ | |
2450 | if (QUOTED_NULL (current_word)) | |
2451 | { | |
2452 | t = alloc_word_desc (); | |
2453 | t->word = make_quoted_char ('\0'); | |
2454 | t->flags |= W_QUOTED|W_HASQUOTEDNULL; | |
2455 | result = make_word_list (t, result); | |
2456 | } | |
2457 | else if (current_word[0] != '\0') | |
2458 | { | |
2459 | /* If we have something, then add it regardless. However, | |
2460 | perform quoted null character removal on the current word. */ | |
2461 | remove_quoted_nulls (current_word); | |
2462 | result = add_string_to_list (current_word, result); | |
2463 | result->word->flags &= ~W_HASQUOTEDNULL; /* just to be sure */ | |
2464 | if (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) | |
2465 | result->word->flags |= W_QUOTED; | |
2466 | } | |
2467 | ||
2468 | /* If we're not doing sequences of separators in the traditional | |
2469 | Bourne shell style, then add a quoted null argument. */ | |
2470 | else if (!sh_style_split && !spctabnl (string[sindex])) | |
2471 | { | |
2472 | t = alloc_word_desc (); | |
2473 | t->word = make_quoted_char ('\0'); | |
2474 | t->flags |= W_QUOTED|W_HASQUOTEDNULL; | |
2475 | result = make_word_list (t, result); | |
2476 | } | |
2477 | ||
2478 | free (current_word); | |
2479 | ||
2480 | /* Note whether or not the separator is IFS whitespace, used later. */ | |
2481 | whitesep = string[sindex] && spctabnl (string[sindex]); | |
2482 | ||
2483 | /* Move past the current separator character. */ | |
2484 | if (string[sindex]) | |
2485 | { | |
2486 | DECLARE_MBSTATE; | |
2487 | ADVANCE_CHAR (string, slen, sindex); | |
2488 | } | |
2489 | ||
2490 | /* Now skip sequences of space, tab, or newline characters if they are | |
2491 | in the list of separators. */ | |
2492 | while (string[sindex] && spctabnl (string[sindex]) && issep (string[sindex])) | |
2493 | sindex++; | |
2494 | ||
2495 | /* If the first separator was IFS whitespace and the current character | |
2496 | is a non-whitespace IFS character, it should be part of the current | |
2497 | field delimiter, not a separate delimiter that would result in an | |
2498 | empty field. Look at POSIX.2, 3.6.5, (3)(b). */ | |
2499 | if (string[sindex] && whitesep && issep (string[sindex]) && !spctabnl (string[sindex])) | |
2500 | { | |
2501 | sindex++; | |
2502 | /* An IFS character that is not IFS white space, along with any | |
2503 | adjacent IFS white space, shall delimit a field. (SUSv3) */ | |
2504 | while (string[sindex] && spctabnl (string[sindex]) && isifs (string[sindex])) | |
2505 | sindex++; | |
2506 | } | |
2507 | } | |
2508 | return (REVERSE_LIST (result, WORD_LIST *)); | |
2509 | } | |
2510 | ||
2511 | /* Parse a single word from STRING, using SEPARATORS to separate fields. | |
2512 | ENDPTR is set to the first character after the word. This is used by | |
2513 | the `read' builtin. This is never called with SEPARATORS != $IFS; | |
2514 | it should be simplified. | |
2515 | ||
2516 | XXX - this function is very similar to list_string; they should be | |
2517 | combined - XXX */ | |
2518 | char * | |
2519 | get_word_from_string (stringp, separators, endptr) | |
2520 | char **stringp, *separators, **endptr; | |
2521 | { | |
2522 | register char *s; | |
2523 | char *current_word; | |
2524 | int sindex, sh_style_split, whitesep, xflags; | |
2525 | size_t slen; | |
2526 | ||
2527 | if (!stringp || !*stringp || !**stringp) | |
2528 | return ((char *)NULL); | |
2529 | ||
2530 | sh_style_split = separators && separators[0] == ' ' && | |
2531 | separators[1] == '\t' && | |
2532 | separators[2] == '\n' && | |
2533 | separators[3] == '\0'; | |
2534 | for (xflags = 0, s = ifs_value; s && *s; s++) | |
2535 | { | |
2536 | if (*s == CTLESC) xflags |= SX_NOCTLESC; | |
2537 | if (*s == CTLNUL) xflags |= SX_NOESCCTLNUL; | |
2538 | } | |
2539 | ||
2540 | s = *stringp; | |
2541 | slen = 0; | |
2542 | ||
2543 | /* Remove sequences of whitespace at the beginning of STRING, as | |
2544 | long as those characters appear in IFS. */ | |
2545 | if (sh_style_split || !separators || !*separators) | |
2546 | { | |
2547 | for (; *s && spctabnl (*s) && isifs (*s); s++); | |
2548 | ||
2549 | /* If the string is nothing but whitespace, update it and return. */ | |
2550 | if (!*s) | |
2551 | { | |
2552 | *stringp = s; | |
2553 | if (endptr) | |
2554 | *endptr = s; | |
2555 | return ((char *)NULL); | |
2556 | } | |
2557 | } | |
2558 | ||
2559 | /* OK, S points to a word that does not begin with white space. | |
2560 | Now extract a word, stopping at a separator, save a pointer to | |
2561 | the first character after the word, then skip sequences of spc, | |
2562 | tab, or nl as long as they are separators. | |
2563 | ||
2564 | This obeys the field splitting rules in Posix.2. */ | |
2565 | sindex = 0; | |
2566 | /* Don't need string length in ADVANCE_CHAR or string_extract_verbatim | |
2567 | unless multibyte chars are possible. */ | |
2568 | slen = (MB_CUR_MAX > 1) ? strlen (s) : 1; | |
2569 | current_word = string_extract_verbatim (s, slen, &sindex, separators, xflags); | |
2570 | ||
2571 | /* Set ENDPTR to the first character after the end of the word. */ | |
2572 | if (endptr) | |
2573 | *endptr = s + sindex; | |
2574 | ||
2575 | /* Note whether or not the separator is IFS whitespace, used later. */ | |
2576 | whitesep = s[sindex] && spctabnl (s[sindex]); | |
2577 | ||
2578 | /* Move past the current separator character. */ | |
2579 | if (s[sindex]) | |
2580 | { | |
2581 | DECLARE_MBSTATE; | |
2582 | ADVANCE_CHAR (s, slen, sindex); | |
2583 | } | |
2584 | ||
2585 | /* Now skip sequences of space, tab, or newline characters if they are | |
2586 | in the list of separators. */ | |
2587 | while (s[sindex] && spctabnl (s[sindex]) && isifs (s[sindex])) | |
2588 | sindex++; | |
2589 | ||
2590 | /* If the first separator was IFS whitespace and the current character is | |
2591 | a non-whitespace IFS character, it should be part of the current field | |
2592 | delimiter, not a separate delimiter that would result in an empty field. | |
2593 | Look at POSIX.2, 3.6.5, (3)(b). */ | |
2594 | if (s[sindex] && whitesep && isifs (s[sindex]) && !spctabnl (s[sindex])) | |
2595 | { | |
2596 | sindex++; | |
2597 | /* An IFS character that is not IFS white space, along with any adjacent | |
2598 | IFS white space, shall delimit a field. */ | |
2599 | while (s[sindex] && spctabnl (s[sindex]) && isifs (s[sindex])) | |
2600 | sindex++; | |
2601 | } | |
2602 | ||
2603 | /* Update STRING to point to the next field. */ | |
2604 | *stringp = s + sindex; | |
2605 | return (current_word); | |
2606 | } | |
2607 | ||
2608 | /* Remove IFS white space at the end of STRING. Start at the end | |
2609 | of the string and walk backwards until the beginning of the string | |
2610 | or we find a character that's not IFS white space and not CTLESC. | |
2611 | Only let CTLESC escape a white space character if SAW_ESCAPE is | |
2612 | non-zero. */ | |
2613 | char * | |
2614 | strip_trailing_ifs_whitespace (string, separators, saw_escape) | |
2615 | char *string, *separators; | |
2616 | int saw_escape; | |
2617 | { | |
2618 | char *s; | |
2619 | ||
2620 | s = string + STRLEN (string) - 1; | |
2621 | while (s > string && ((spctabnl (*s) && isifs (*s)) || | |
2622 | (saw_escape && *s == CTLESC && spctabnl (s[1])))) | |
2623 | s--; | |
2624 | *++s = '\0'; | |
2625 | return string; | |
2626 | } | |
2627 | ||
2628 | #if 0 | |
2629 | /* UNUSED */ | |
2630 | /* Split STRING into words at whitespace. Obeys shell-style quoting with | |
2631 | backslashes, single and double quotes. */ | |
2632 | WORD_LIST * | |
2633 | list_string_with_quotes (string) | |
2634 | char *string; | |
2635 | { | |
2636 | WORD_LIST *list; | |
2637 | char *token, *s; | |
2638 | size_t s_len; | |
2639 | int c, i, tokstart, len; | |
2640 | ||
2641 | for (s = string; s && *s && spctabnl (*s); s++) | |
2642 | ; | |
2643 | if (s == 0 || *s == 0) | |
2644 | return ((WORD_LIST *)NULL); | |
2645 | ||
2646 | s_len = strlen (s); | |
2647 | tokstart = i = 0; | |
2648 | list = (WORD_LIST *)NULL; | |
2649 | while (1) | |
2650 | { | |
2651 | c = s[i]; | |
2652 | if (c == '\\') | |
2653 | { | |
2654 | i++; | |
2655 | if (s[i]) | |
2656 | i++; | |
2657 | } | |
2658 | else if (c == '\'') | |
2659 | i = skip_single_quoted (s, s_len, ++i); | |
2660 | else if (c == '"') | |
2661 | i = skip_double_quoted (s, s_len, ++i); | |
2662 | else if (c == 0 || spctabnl (c)) | |
2663 | { | |
2664 | /* We have found the end of a token. Make a word out of it and | |
2665 | add it to the word list. */ | |
2666 | token = substring (s, tokstart, i); | |
2667 | list = add_string_to_list (token, list); | |
2668 | free (token); | |
2669 | while (spctabnl (s[i])) | |
2670 | i++; | |
2671 | if (s[i]) | |
2672 | tokstart = i; | |
2673 | else | |
2674 | break; | |
2675 | } | |
2676 | else | |
2677 | i++; /* normal character */ | |
2678 | } | |
2679 | return (REVERSE_LIST (list, WORD_LIST *)); | |
2680 | } | |
2681 | #endif | |
2682 | ||
2683 | /********************************************************/ | |
2684 | /* */ | |
2685 | /* Functions to perform assignment statements */ | |
2686 | /* */ | |
2687 | /********************************************************/ | |
2688 | ||
2689 | #if defined (ARRAY_VARS) | |
2690 | static SHELL_VAR * | |
2691 | do_compound_assignment (name, value, flags) | |
2692 | char *name, *value; | |
2693 | int flags; | |
2694 | { | |
2695 | SHELL_VAR *v; | |
2696 | int mklocal, mkassoc; | |
2697 | WORD_LIST *list; | |
2698 | ||
2699 | mklocal = flags & ASS_MKLOCAL; | |
2700 | mkassoc = flags & ASS_MKASSOC; | |
2701 | ||
2702 | if (mklocal && variable_context) | |
2703 | { | |
2704 | v = find_variable (name); | |
2705 | list = expand_compound_array_assignment (v, value, flags); | |
2706 | if (mkassoc) | |
2707 | v = make_local_assoc_variable (name); | |
2708 | else if (v == 0 || (array_p (v) == 0 && assoc_p (v) == 0) || v->context != variable_context) | |
2709 | v = make_local_array_variable (name); | |
2710 | assign_compound_array_list (v, list, flags); | |
2711 | } | |
2712 | else | |
2713 | v = assign_array_from_string (name, value, flags); | |
2714 | ||
2715 | return (v); | |
2716 | } | |
2717 | #endif | |
2718 | ||
2719 | /* Given STRING, an assignment string, get the value of the right side | |
2720 | of the `=', and bind it to the left side. If EXPAND is true, then | |
2721 | perform parameter expansion, command substitution, and arithmetic | |
2722 | expansion on the right-hand side. Perform tilde expansion in any | |
2723 | case. Do not perform word splitting on the result of expansion. */ | |
2724 | static int | |
2725 | do_assignment_internal (word, expand) | |
2726 | const WORD_DESC *word; | |
2727 | int expand; | |
2728 | { | |
2729 | int offset, appendop, assign_list, aflags, retval; | |
2730 | char *name, *value, *temp; | |
2731 | SHELL_VAR *entry; | |
2732 | #if defined (ARRAY_VARS) | |
2733 | char *t; | |
2734 | int ni; | |
2735 | #endif | |
2736 | const char *string; | |
2737 | ||
2738 | if (word == 0 || word->word == 0) | |
2739 | return 0; | |
2740 | ||
2741 | appendop = assign_list = aflags = 0; | |
2742 | string = word->word; | |
2743 | offset = assignment (string, 0); | |
2744 | name = savestring (string); | |
2745 | value = (char *)NULL; | |
2746 | ||
2747 | if (name[offset] == '=') | |
2748 | { | |
2749 | if (name[offset - 1] == '+') | |
2750 | { | |
2751 | appendop = 1; | |
2752 | name[offset - 1] = '\0'; | |
2753 | } | |
2754 | ||
2755 | name[offset] = 0; /* might need this set later */ | |
2756 | temp = name + offset + 1; | |
2757 | ||
2758 | #if defined (ARRAY_VARS) | |
2759 | if (expand && (word->flags & W_COMPASSIGN)) | |
2760 | { | |
2761 | assign_list = ni = 1; | |
2762 | value = extract_array_assignment_list (temp, &ni); | |
2763 | } | |
2764 | else | |
2765 | #endif | |
2766 | if (expand && temp[0]) | |
2767 | value = expand_string_if_necessary (temp, 0, expand_string_assignment); | |
2768 | else | |
2769 | value = savestring (temp); | |
2770 | } | |
2771 | ||
2772 | if (value == 0) | |
2773 | { | |
2774 | value = (char *)xmalloc (1); | |
2775 | value[0] = '\0'; | |
2776 | } | |
2777 | ||
2778 | if (echo_command_at_execute) | |
2779 | { | |
2780 | if (appendop) | |
2781 | name[offset - 1] = '+'; | |
2782 | xtrace_print_assignment (name, value, assign_list, 1); | |
2783 | if (appendop) | |
2784 | name[offset - 1] = '\0'; | |
2785 | } | |
2786 | ||
2787 | #define ASSIGN_RETURN(r) do { FREE (value); free (name); return (r); } while (0) | |
2788 | ||
2789 | if (appendop) | |
2790 | aflags |= ASS_APPEND; | |
2791 | ||
2792 | #if defined (ARRAY_VARS) | |
2793 | if (t = mbschr (name, '[')) /*]*/ | |
2794 | { | |
2795 | if (assign_list) | |
2796 | { | |
2797 | report_error (_("%s: cannot assign list to array member"), name); | |
2798 | ASSIGN_RETURN (0); | |
2799 | } | |
2800 | entry = assign_array_element (name, value, aflags); | |
2801 | if (entry == 0) | |
2802 | ASSIGN_RETURN (0); | |
2803 | } | |
2804 | else if (assign_list) | |
2805 | { | |
2806 | if (word->flags & W_ASSIGNARG) | |
2807 | aflags |= ASS_MKLOCAL; | |
2808 | if (word->flags & W_ASSIGNASSOC) | |
2809 | aflags |= ASS_MKASSOC; | |
2810 | entry = do_compound_assignment (name, value, aflags); | |
2811 | } | |
2812 | else | |
2813 | #endif /* ARRAY_VARS */ | |
2814 | entry = bind_variable (name, value, aflags); | |
2815 | ||
2816 | stupidly_hack_special_variables (name); | |
2817 | ||
2818 | #if 1 | |
2819 | /* Return 1 if the assignment seems to have been performed correctly. */ | |
2820 | if (entry == 0 || readonly_p (entry)) | |
2821 | retval = 0; /* assignment failure */ | |
2822 | else if (noassign_p (entry)) | |
2823 | { | |
2824 | last_command_exit_value = EXECUTION_FAILURE; | |
2825 | retval = 1; /* error status, but not assignment failure */ | |
2826 | } | |
2827 | else | |
2828 | retval = 1; | |
2829 | ||
2830 | if (entry && retval != 0 && noassign_p (entry) == 0) | |
2831 | VUNSETATTR (entry, att_invisible); | |
2832 | ||
2833 | ASSIGN_RETURN (retval); | |
2834 | #else | |
2835 | if (entry) | |
2836 | VUNSETATTR (entry, att_invisible); | |
2837 | ||
2838 | ASSIGN_RETURN (entry ? ((readonly_p (entry) == 0) && noassign_p (entry) == 0) : 0); | |
2839 | #endif | |
2840 | } | |
2841 | ||
2842 | /* Perform the assignment statement in STRING, and expand the | |
2843 | right side by doing tilde, command and parameter expansion. */ | |
2844 | int | |
2845 | do_assignment (string) | |
2846 | char *string; | |
2847 | { | |
2848 | WORD_DESC td; | |
2849 | ||
2850 | td.flags = W_ASSIGNMENT; | |
2851 | td.word = string; | |
2852 | ||
2853 | return do_assignment_internal (&td, 1); | |
2854 | } | |
2855 | ||
2856 | int | |
2857 | do_word_assignment (word, flags) | |
2858 | WORD_DESC *word; | |
2859 | int flags; | |
2860 | { | |
2861 | return do_assignment_internal (word, 1); | |
2862 | } | |
2863 | ||
2864 | /* Given STRING, an assignment string, get the value of the right side | |
2865 | of the `=', and bind it to the left side. Do not perform any word | |
2866 | expansions on the right hand side. */ | |
2867 | int | |
2868 | do_assignment_no_expand (string) | |
2869 | char *string; | |
2870 | { | |
2871 | WORD_DESC td; | |
2872 | ||
2873 | td.flags = W_ASSIGNMENT; | |
2874 | td.word = string; | |
2875 | ||
2876 | return (do_assignment_internal (&td, 0)); | |
2877 | } | |
2878 | ||
2879 | /*************************************************** | |
2880 | * * | |
2881 | * Functions to manage the positional parameters * | |
2882 | * * | |
2883 | ***************************************************/ | |
2884 | ||
2885 | /* Return the word list that corresponds to `$*'. */ | |
2886 | WORD_LIST * | |
2887 | list_rest_of_args () | |
2888 | { | |
2889 | register WORD_LIST *list, *args; | |
2890 | int i; | |
2891 | ||
2892 | /* Break out of the loop as soon as one of the dollar variables is null. */ | |
2893 | for (i = 1, list = (WORD_LIST *)NULL; i < 10 && dollar_vars[i]; i++) | |
2894 | list = make_word_list (make_bare_word (dollar_vars[i]), list); | |
2895 | ||
2896 | for (args = rest_of_args; args; args = args->next) | |
2897 | list = make_word_list (make_bare_word (args->word->word), list); | |
2898 | ||
2899 | return (REVERSE_LIST (list, WORD_LIST *)); | |
2900 | } | |
2901 | ||
2902 | int | |
2903 | number_of_args () | |
2904 | { | |
2905 | register WORD_LIST *list; | |
2906 | int n; | |
2907 | ||
2908 | for (n = 0; n < 9 && dollar_vars[n+1]; n++) | |
2909 | ; | |
2910 | for (list = rest_of_args; list; list = list->next) | |
2911 | n++; | |
2912 | return n; | |
2913 | } | |
2914 | ||
2915 | /* Return the value of a positional parameter. This handles values > 10. */ | |
2916 | char * | |
2917 | get_dollar_var_value (ind) | |
2918 | intmax_t ind; | |
2919 | { | |
2920 | char *temp; | |
2921 | WORD_LIST *p; | |
2922 | ||
2923 | if (ind < 10) | |
2924 | temp = dollar_vars[ind] ? savestring (dollar_vars[ind]) : (char *)NULL; | |
2925 | else /* We want something like ${11} */ | |
2926 | { | |
2927 | ind -= 10; | |
2928 | for (p = rest_of_args; p && ind--; p = p->next) | |
2929 | ; | |
2930 | temp = p ? savestring (p->word->word) : (char *)NULL; | |
2931 | } | |
2932 | return (temp); | |
2933 | } | |
2934 | ||
2935 | /* Make a single large string out of the dollar digit variables, | |
2936 | and the rest_of_args. If DOLLAR_STAR is 1, then obey the special | |
2937 | case of "$*" with respect to IFS. */ | |
2938 | char * | |
2939 | string_rest_of_args (dollar_star) | |
2940 | int dollar_star; | |
2941 | { | |
2942 | register WORD_LIST *list; | |
2943 | char *string; | |
2944 | ||
2945 | list = list_rest_of_args (); | |
2946 | string = dollar_star ? string_list_dollar_star (list) : string_list (list); | |
2947 | dispose_words (list); | |
2948 | return (string); | |
2949 | } | |
2950 | ||
2951 | /* Return a string containing the positional parameters from START to | |
2952 | END, inclusive. If STRING[0] == '*', we obey the rules for $*, | |
2953 | which only makes a difference if QUOTED is non-zero. If QUOTED includes | |
2954 | Q_HERE_DOCUMENT or Q_DOUBLE_QUOTES, this returns a quoted list, otherwise | |
2955 | no quoting chars are added. */ | |
2956 | static char * | |
2957 | pos_params (string, start, end, quoted) | |
2958 | char *string; | |
2959 | int start, end, quoted; | |
2960 | { | |
2961 | WORD_LIST *save, *params, *h, *t; | |
2962 | char *ret; | |
2963 | int i; | |
2964 | ||
2965 | /* see if we can short-circuit. if start == end, we want 0 parameters. */ | |
2966 | if (start == end) | |
2967 | return ((char *)NULL); | |
2968 | ||
2969 | save = params = list_rest_of_args (); | |
2970 | if (save == 0) | |
2971 | return ((char *)NULL); | |
2972 | ||
2973 | if (start == 0) /* handle ${@:0[:x]} specially */ | |
2974 | { | |
2975 | t = make_word_list (make_word (dollar_vars[0]), params); | |
2976 | save = params = t; | |
2977 | } | |
2978 | ||
2979 | for (i = start ? 1 : 0; params && i < start; i++) | |
2980 | params = params->next; | |
2981 | if (params == 0) | |
2982 | return ((char *)NULL); | |
2983 | for (h = t = params; params && i < end; i++) | |
2984 | { | |
2985 | t = params; | |
2986 | params = params->next; | |
2987 | } | |
2988 | ||
2989 | t->next = (WORD_LIST *)NULL; | |
2990 | ||
2991 | ret = string_list_pos_params (string[0], h, quoted); | |
2992 | ||
2993 | if (t != params) | |
2994 | t->next = params; | |
2995 | ||
2996 | dispose_words (save); | |
2997 | return (ret); | |
2998 | } | |
2999 | ||
3000 | /******************************************************************/ | |
3001 | /* */ | |
3002 | /* Functions to expand strings to strings or WORD_LISTs */ | |
3003 | /* */ | |
3004 | /******************************************************************/ | |
3005 | ||
3006 | #if defined (PROCESS_SUBSTITUTION) | |
3007 | #define EXP_CHAR(s) (s == '$' || s == '`' || s == '<' || s == '>' || s == CTLESC || s == '~') | |
3008 | #else | |
3009 | #define EXP_CHAR(s) (s == '$' || s == '`' || s == CTLESC || s == '~') | |
3010 | #endif | |
3011 | ||
3012 | /* If there are any characters in STRING that require full expansion, | |
3013 | then call FUNC to expand STRING; otherwise just perform quote | |
3014 | removal if necessary. This returns a new string. */ | |
3015 | static char * | |
3016 | expand_string_if_necessary (string, quoted, func) | |
3017 | char *string; | |
3018 | int quoted; | |
3019 | EXPFUNC *func; | |
3020 | { | |
3021 | WORD_LIST *list; | |
3022 | size_t slen; | |
3023 | int i, saw_quote; | |
3024 | char *ret; | |
3025 | DECLARE_MBSTATE; | |
3026 | ||
3027 | /* Don't need string length for ADVANCE_CHAR unless multibyte chars possible. */ | |
3028 | slen = (MB_CUR_MAX > 1) ? strlen (string) : 0; | |
3029 | i = saw_quote = 0; | |
3030 | while (string[i]) | |
3031 | { | |
3032 | if (EXP_CHAR (string[i])) | |
3033 | break; | |
3034 | else if (string[i] == '\'' || string[i] == '\\' || string[i] == '"') | |
3035 | saw_quote = 1; | |
3036 | ADVANCE_CHAR (string, slen, i); | |
3037 | } | |
3038 | ||
3039 | if (string[i]) | |
3040 | { | |
3041 | list = (*func) (string, quoted); | |
3042 | if (list) | |
3043 | { | |
3044 | ret = string_list (list); | |
3045 | dispose_words (list); | |
3046 | } | |
3047 | else | |
3048 | ret = (char *)NULL; | |
3049 | } | |
3050 | else if (saw_quote && ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) == 0)) | |
3051 | ret = string_quote_removal (string, quoted); | |
3052 | else | |
3053 | ret = savestring (string); | |
3054 | ||
3055 | return ret; | |
3056 | } | |
3057 | ||
3058 | static inline char * | |
3059 | expand_string_to_string_internal (string, quoted, func) | |
3060 | char *string; | |
3061 | int quoted; | |
3062 | EXPFUNC *func; | |
3063 | { | |
3064 | WORD_LIST *list; | |
3065 | char *ret; | |
3066 | ||
3067 | if (string == 0 || *string == '\0') | |
3068 | return ((char *)NULL); | |
3069 | ||
3070 | list = (*func) (string, quoted); | |
3071 | if (list) | |
3072 | { | |
3073 | ret = string_list (list); | |
3074 | dispose_words (list); | |
3075 | } | |
3076 | else | |
3077 | ret = (char *)NULL; | |
3078 | ||
3079 | return (ret); | |
3080 | } | |
3081 | ||
3082 | char * | |
3083 | expand_string_to_string (string, quoted) | |
3084 | char *string; | |
3085 | int quoted; | |
3086 | { | |
3087 | return (expand_string_to_string_internal (string, quoted, expand_string)); | |
3088 | } | |
3089 | ||
3090 | char * | |
3091 | expand_string_unsplit_to_string (string, quoted) | |
3092 | char *string; | |
3093 | int quoted; | |
3094 | { | |
3095 | return (expand_string_to_string_internal (string, quoted, expand_string_unsplit)); | |
3096 | } | |
3097 | ||
3098 | char * | |
3099 | expand_assignment_string_to_string (string, quoted) | |
3100 | char *string; | |
3101 | int quoted; | |
3102 | { | |
3103 | return (expand_string_to_string_internal (string, quoted, expand_string_assignment)); | |
3104 | } | |
3105 | ||
3106 | char * | |
3107 | expand_arith_string (string, quoted) | |
3108 | char *string; | |
3109 | int quoted; | |
3110 | { | |
3111 | return (expand_string_if_necessary (string, quoted, expand_string)); | |
3112 | } | |
3113 | ||
3114 | #if defined (COND_COMMAND) | |
3115 | /* Just remove backslashes in STRING. Returns a new string. */ | |
3116 | char * | |
3117 | remove_backslashes (string) | |
3118 | char *string; | |
3119 | { | |
3120 | char *r, *ret, *s; | |
3121 | ||
3122 | r = ret = (char *)xmalloc (strlen (string) + 1); | |
3123 | for (s = string; s && *s; ) | |
3124 | { | |
3125 | if (*s == '\\') | |
3126 | s++; | |
3127 | if (*s == 0) | |
3128 | break; | |
3129 | *r++ = *s++; | |
3130 | } | |
3131 | *r = '\0'; | |
3132 | return ret; | |
3133 | } | |
3134 | ||
3135 | /* This needs better error handling. */ | |
3136 | /* Expand W for use as an argument to a unary or binary operator in a | |
3137 | [[...]] expression. If SPECIAL is 1, this is the rhs argument | |
3138 | to the != or == operator, and should be treated as a pattern. In | |
3139 | this case, we quote the string specially for the globbing code. If | |
3140 | SPECIAL is 2, this is an rhs argument for the =~ operator, and should | |
3141 | be quoted appropriately for regcomp/regexec. The caller is responsible | |
3142 | for removing the backslashes if the unquoted word is needed later. */ | |
3143 | char * | |
3144 | cond_expand_word (w, special) | |
3145 | WORD_DESC *w; | |
3146 | int special; | |
3147 | { | |
3148 | char *r, *p; | |
3149 | WORD_LIST *l; | |
3150 | int qflags; | |
3151 | ||
3152 | if (w->word == 0 || w->word[0] == '\0') | |
3153 | return ((char *)NULL); | |
3154 | ||
3155 | w->flags |= W_NOSPLIT2; | |
3156 | l = call_expand_word_internal (w, 0, 0, (int *)0, (int *)0); | |
3157 | if (l) | |
3158 | { | |
3159 | if (special == 0) | |
3160 | { | |
3161 | dequote_list (l); | |
3162 | r = string_list (l); | |
3163 | } | |
3164 | else | |
3165 | { | |
3166 | qflags = QGLOB_CVTNULL; | |
3167 | if (special == 2) | |
3168 | qflags |= QGLOB_REGEXP; | |
3169 | p = string_list (l); | |
3170 | r = quote_string_for_globbing (p, qflags); | |
3171 | free (p); | |
3172 | } | |
3173 | dispose_words (l); | |
3174 | } | |
3175 | else | |
3176 | r = (char *)NULL; | |
3177 | ||
3178 | return r; | |
3179 | } | |
3180 | #endif | |
3181 | ||
3182 | /* Call expand_word_internal to expand W and handle error returns. | |
3183 | A convenience function for functions that don't want to handle | |
3184 | any errors or free any memory before aborting. */ | |
3185 | static WORD_LIST * | |
3186 | call_expand_word_internal (w, q, i, c, e) | |
3187 | WORD_DESC *w; | |
3188 | int q, i, *c, *e; | |
3189 | { | |
3190 | WORD_LIST *result; | |
3191 | ||
3192 | result = expand_word_internal (w, q, i, c, e); | |
3193 | if (result == &expand_word_error || result == &expand_word_fatal) | |
3194 | { | |
3195 | /* By convention, each time this error is returned, w->word has | |
3196 | already been freed (it sometimes may not be in the fatal case, | |
3197 | but that doesn't result in a memory leak because we're going | |
3198 | to exit in most cases). */ | |
3199 | w->word = (char *)NULL; | |
3200 | last_command_exit_value = EXECUTION_FAILURE; | |
3201 | exp_jump_to_top_level ((result == &expand_word_error) ? DISCARD : FORCE_EOF); | |
3202 | /* NOTREACHED */ | |
3203 | } | |
3204 | else | |
3205 | return (result); | |
3206 | } | |
3207 | ||
3208 | /* Perform parameter expansion, command substitution, and arithmetic | |
9c2db999 CR |
3209 | expansion on STRING, as if it were a word. Leave the result quoted. |
3210 | Since this does not perform word splitting, it leaves quoted nulls | |
3211 | in the result. */ | |
e192f341 CR |
3212 | static WORD_LIST * |
3213 | expand_string_internal (string, quoted) | |
3214 | char *string; | |
3215 | int quoted; | |
3216 | { | |
3217 | WORD_DESC td; | |
3218 | WORD_LIST *tresult; | |
3219 | ||
3220 | if (string == 0 || *string == 0) | |
3221 | return ((WORD_LIST *)NULL); | |
3222 | ||
3223 | td.flags = 0; | |
3224 | td.word = savestring (string); | |
3225 | ||
3226 | tresult = call_expand_word_internal (&td, quoted, 0, (int *)NULL, (int *)NULL); | |
3227 | ||
3228 | FREE (td.word); | |
3229 | return (tresult); | |
3230 | } | |
3231 | ||
3232 | /* Expand STRING by performing parameter expansion, command substitution, | |
3233 | and arithmetic expansion. Dequote the resulting WORD_LIST before | |
3234 | returning it, but do not perform word splitting. The call to | |
3235 | remove_quoted_nulls () is in here because word splitting normally | |
3236 | takes care of quote removal. */ | |
3237 | WORD_LIST * | |
3238 | expand_string_unsplit (string, quoted) | |
3239 | char *string; | |
3240 | int quoted; | |
3241 | { | |
3242 | WORD_LIST *value; | |
3243 | ||
3244 | if (string == 0 || *string == '\0') | |
3245 | return ((WORD_LIST *)NULL); | |
3246 | ||
3247 | expand_no_split_dollar_star = 1; | |
3248 | value = expand_string_internal (string, quoted); | |
3249 | expand_no_split_dollar_star = 0; | |
3250 | ||
3251 | if (value) | |
3252 | { | |
3253 | if (value->word) | |
3254 | { | |
3255 | remove_quoted_nulls (value->word->word); | |
3256 | value->word->flags &= ~W_HASQUOTEDNULL; | |
3257 | } | |
3258 | dequote_list (value); | |
3259 | } | |
3260 | return (value); | |
3261 | } | |
3262 | ||
3263 | /* Expand the rhs of an assignment statement */ | |
3264 | WORD_LIST * | |
3265 | expand_string_assignment (string, quoted) | |
3266 | char *string; | |
3267 | int quoted; | |
3268 | { | |
3269 | WORD_DESC td; | |
3270 | WORD_LIST *value; | |
3271 | ||
3272 | if (string == 0 || *string == '\0') | |
3273 | return ((WORD_LIST *)NULL); | |
3274 | ||
3275 | expand_no_split_dollar_star = 1; | |
3276 | ||
3277 | td.flags = W_ASSIGNRHS; | |
3278 | td.word = savestring (string); | |
3279 | value = call_expand_word_internal (&td, quoted, 0, (int *)NULL, (int *)NULL); | |
3280 | FREE (td.word); | |
3281 | ||
3282 | expand_no_split_dollar_star = 0; | |
3283 | ||
3284 | if (value) | |
3285 | { | |
3286 | if (value->word) | |
3287 | { | |
3288 | remove_quoted_nulls (value->word->word); | |
3289 | value->word->flags &= ~W_HASQUOTEDNULL; | |
3290 | } | |
3291 | dequote_list (value); | |
3292 | } | |
3293 | return (value); | |
3294 | } | |
3295 | ||
3296 | ||
3297 | /* Expand one of the PS? prompt strings. This is a sort of combination of | |
3298 | expand_string_unsplit and expand_string_internal, but returns the | |
3299 | passed string when an error occurs. Might want to trap other calls | |
3300 | to jump_to_top_level here so we don't endlessly loop. */ | |
3301 | WORD_LIST * | |
3302 | expand_prompt_string (string, quoted, wflags) | |
3303 | char *string; | |
3304 | int quoted; | |
3305 | int wflags; | |
3306 | { | |
3307 | WORD_LIST *value; | |
3308 | WORD_DESC td; | |
3309 | ||
3310 | if (string == 0 || *string == 0) | |
3311 | return ((WORD_LIST *)NULL); | |
3312 | ||
3313 | td.flags = wflags; | |
3314 | td.word = savestring (string); | |
3315 | ||
3316 | no_longjmp_on_fatal_error = 1; | |
3317 | value = expand_word_internal (&td, quoted, 0, (int *)NULL, (int *)NULL); | |
3318 | no_longjmp_on_fatal_error = 0; | |
3319 | ||
3320 | if (value == &expand_word_error || value == &expand_word_fatal) | |
3321 | { | |
3322 | value = make_word_list (make_bare_word (string), (WORD_LIST *)NULL); | |
3323 | return value; | |
3324 | } | |
3325 | FREE (td.word); | |
3326 | if (value) | |
3327 | { | |
3328 | if (value->word) | |
3329 | { | |
3330 | remove_quoted_nulls (value->word->word); | |
3331 | value->word->flags &= ~W_HASQUOTEDNULL; | |
3332 | } | |
3333 | dequote_list (value); | |
3334 | } | |
3335 | return (value); | |
3336 | } | |
3337 | ||
3338 | /* Expand STRING just as if you were expanding a word, but do not dequote | |
3339 | the resultant WORD_LIST. This is called only from within this file, | |
3340 | and is used to correctly preserve quoted characters when expanding | |
3341 | things like ${1+"$@"}. This does parameter expansion, command | |
3342 | substitution, arithmetic expansion, and word splitting. */ | |
3343 | static WORD_LIST * | |
3344 | expand_string_leave_quoted (string, quoted) | |
3345 | char *string; | |
3346 | int quoted; | |
3347 | { | |
3348 | WORD_LIST *tlist; | |
3349 | WORD_LIST *tresult; | |
3350 | ||
3351 | if (string == 0 || *string == '\0') | |
3352 | return ((WORD_LIST *)NULL); | |
3353 | ||
3354 | tlist = expand_string_internal (string, quoted); | |
3355 | ||
3356 | if (tlist) | |
3357 | { | |
3358 | tresult = word_list_split (tlist); | |
3359 | dispose_words (tlist); | |
3360 | return (tresult); | |
3361 | } | |
3362 | return ((WORD_LIST *)NULL); | |
3363 | } | |
3364 | ||
3365 | /* This does not perform word splitting or dequote the WORD_LIST | |
3366 | it returns. */ | |
3367 | static WORD_LIST * | |
3368 | expand_string_for_rhs (string, quoted, dollar_at_p, has_dollar_at) | |
3369 | char *string; | |
3370 | int quoted, *dollar_at_p, *has_dollar_at; | |
3371 | { | |
3372 | WORD_DESC td; | |
3373 | WORD_LIST *tresult; | |
3374 | ||
3375 | if (string == 0 || *string == '\0') | |
3376 | return (WORD_LIST *)NULL; | |
3377 | ||
9c2db999 | 3378 | td.flags = W_NOSPLIT2; /* no splitting, remove "" and '' */ |
e192f341 CR |
3379 | td.word = string; |
3380 | tresult = call_expand_word_internal (&td, quoted, 1, dollar_at_p, has_dollar_at); | |
3381 | return (tresult); | |
3382 | } | |
3383 | ||
3384 | /* Expand STRING just as if you were expanding a word. This also returns | |
3385 | a list of words. Note that filename globbing is *NOT* done for word | |
3386 | or string expansion, just when the shell is expanding a command. This | |
3387 | does parameter expansion, command substitution, arithmetic expansion, | |
3388 | and word splitting. Dequote the resultant WORD_LIST before returning. */ | |
3389 | WORD_LIST * | |
3390 | expand_string (string, quoted) | |
3391 | char *string; | |
3392 | int quoted; | |
3393 | { | |
3394 | WORD_LIST *result; | |
3395 | ||
3396 | if (string == 0 || *string == '\0') | |
3397 | return ((WORD_LIST *)NULL); | |
3398 | ||
3399 | result = expand_string_leave_quoted (string, quoted); | |
3400 | return (result ? dequote_list (result) : result); | |
3401 | } | |
3402 | ||
3403 | /*************************************************** | |
3404 | * * | |
3405 | * Functions to handle quoting chars * | |
3406 | * * | |
3407 | ***************************************************/ | |
3408 | ||
3409 | /* Conventions: | |
3410 | ||
3411 | A string with s[0] == CTLNUL && s[1] == 0 is a quoted null string. | |
3412 | The parser passes CTLNUL as CTLESC CTLNUL. */ | |
3413 | ||
3414 | /* Quote escape characters in string s, but no other characters. This is | |
3415 | used to protect CTLESC and CTLNUL in variable values from the rest of | |
3416 | the word expansion process after the variable is expanded (word splitting | |
3417 | and filename generation). If IFS is null, we quote spaces as well, just | |
3418 | in case we split on spaces later (in the case of unquoted $@, we will | |
3419 | eventually attempt to split the entire word on spaces). Corresponding | |
3420 | code exists in dequote_escapes. Even if we don't end up splitting on | |
3421 | spaces, quoting spaces is not a problem. This should never be called on | |
3422 | a string that is quoted with single or double quotes or part of a here | |
3423 | document (effectively double-quoted). */ | |
3424 | char * | |
3425 | quote_escapes (string) | |
3426 | char *string; | |
3427 | { | |
3428 | register char *s, *t; | |
3429 | size_t slen; | |
3430 | char *result, *send; | |
3431 | int quote_spaces, skip_ctlesc, skip_ctlnul; | |
3432 | DECLARE_MBSTATE; | |
3433 | ||
3434 | slen = strlen (string); | |
3435 | send = string + slen; | |
3436 | ||
3437 | quote_spaces = (ifs_value && *ifs_value == 0); | |
3438 | ||
3439 | for (skip_ctlesc = skip_ctlnul = 0, s = ifs_value; s && *s; s++) | |
3440 | skip_ctlesc |= *s == CTLESC, skip_ctlnul |= *s == CTLNUL; | |
3441 | ||
3442 | t = result = (char *)xmalloc ((slen * 2) + 1); | |
3443 | s = string; | |
3444 | ||
3445 | while (*s) | |
3446 | { | |
3447 | if ((skip_ctlesc == 0 && *s == CTLESC) || (skip_ctlnul == 0 && *s == CTLNUL) || (quote_spaces && *s == ' ')) | |
3448 | *t++ = CTLESC; | |
3449 | COPY_CHAR_P (t, s, send); | |
3450 | } | |
3451 | *t = '\0'; | |
3452 | return (result); | |
3453 | } | |
3454 | ||
3455 | static WORD_LIST * | |
3456 | list_quote_escapes (list) | |
3457 | WORD_LIST *list; | |
3458 | { | |
3459 | register WORD_LIST *w; | |
3460 | char *t; | |
3461 | ||
3462 | for (w = list; w; w = w->next) | |
3463 | { | |
3464 | t = w->word->word; | |
3465 | w->word->word = quote_escapes (t); | |
3466 | free (t); | |
3467 | } | |
3468 | return list; | |
3469 | } | |
3470 | ||
3471 | /* Inverse of quote_escapes; remove CTLESC protecting CTLESC or CTLNUL. | |
3472 | ||
3473 | The parser passes us CTLESC as CTLESC CTLESC and CTLNUL as CTLESC CTLNUL. | |
3474 | This is necessary to make unquoted CTLESC and CTLNUL characters in the | |
3475 | data stream pass through properly. | |
3476 | ||
3477 | We need to remove doubled CTLESC characters inside quoted strings before | |
3478 | quoting the entire string, so we do not double the number of CTLESC | |
3479 | characters. | |
3480 | ||
3481 | Also used by parts of the pattern substitution code. */ | |
3482 | char * | |
3483 | dequote_escapes (string) | |
3484 | char *string; | |
3485 | { | |
3486 | register char *s, *t, *s1; | |
3487 | size_t slen; | |
3488 | char *result, *send; | |
3489 | int quote_spaces; | |
3490 | DECLARE_MBSTATE; | |
3491 | ||
3492 | if (string == 0) | |
3493 | return string; | |
3494 | ||
3495 | slen = strlen (string); | |
3496 | send = string + slen; | |
3497 | ||
3498 | t = result = (char *)xmalloc (slen + 1); | |
3499 | ||
3500 | if (strchr (string, CTLESC) == 0) | |
3501 | return (strcpy (result, string)); | |
3502 | ||
3503 | quote_spaces = (ifs_value && *ifs_value == 0); | |
3504 | ||
3505 | s = string; | |
3506 | while (*s) | |
3507 | { | |
3508 | if (*s == CTLESC && (s[1] == CTLESC || s[1] == CTLNUL || (quote_spaces && s[1] == ' '))) | |
3509 | { | |
3510 | s++; | |
3511 | if (*s == '\0') | |
3512 | break; | |
3513 | } | |
3514 | COPY_CHAR_P (t, s, send); | |
3515 | } | |
3516 | *t = '\0'; | |
3517 | return result; | |
3518 | } | |
3519 | ||
3520 | /* Return a new string with the quoted representation of character C. | |
3521 | This turns "" into QUOTED_NULL, so the W_HASQUOTEDNULL flag needs to be | |
3522 | set in any resultant WORD_DESC where this value is the word. */ | |
3523 | static char * | |
3524 | make_quoted_char (c) | |
3525 | int c; | |
3526 | { | |
3527 | char *temp; | |
3528 | ||
3529 | temp = (char *)xmalloc (3); | |
3530 | if (c == 0) | |
3531 | { | |
3532 | temp[0] = CTLNUL; | |
3533 | temp[1] = '\0'; | |
3534 | } | |
3535 | else | |
3536 | { | |
3537 | temp[0] = CTLESC; | |
3538 | temp[1] = c; | |
3539 | temp[2] = '\0'; | |
3540 | } | |
3541 | return (temp); | |
3542 | } | |
3543 | ||
3544 | /* Quote STRING, returning a new string. This turns "" into QUOTED_NULL, so | |
3545 | the W_HASQUOTEDNULL flag needs to be set in any resultant WORD_DESC where | |
3546 | this value is the word. */ | |
3547 | char * | |
3548 | quote_string (string) | |
3549 | char *string; | |
3550 | { | |
3551 | register char *t; | |
3552 | size_t slen; | |
3553 | char *result, *send; | |
3554 | ||
3555 | if (*string == 0) | |
3556 | { | |
3557 | result = (char *)xmalloc (2); | |
3558 | result[0] = CTLNUL; | |
3559 | result[1] = '\0'; | |
3560 | } | |
3561 | else | |
3562 | { | |
3563 | DECLARE_MBSTATE; | |
3564 | ||
3565 | slen = strlen (string); | |
3566 | send = string + slen; | |
3567 | ||
3568 | result = (char *)xmalloc ((slen * 2) + 1); | |
3569 | ||
3570 | for (t = result; string < send; ) | |
3571 | { | |
3572 | *t++ = CTLESC; | |
3573 | COPY_CHAR_P (t, string, send); | |
3574 | } | |
3575 | *t = '\0'; | |
3576 | } | |
3577 | return (result); | |
3578 | } | |
3579 | ||
3580 | /* De-quote quoted characters in STRING. */ | |
3581 | char * | |
3582 | dequote_string (string) | |
3583 | char *string; | |
3584 | { | |
3585 | register char *s, *t; | |
3586 | size_t slen; | |
3587 | char *result, *send; | |
3588 | DECLARE_MBSTATE; | |
3589 | ||
3590 | slen = strlen (string); | |
3591 | ||
3592 | t = result = (char *)xmalloc (slen + 1); | |
3593 | ||
3594 | if (QUOTED_NULL (string)) | |
3595 | { | |
3596 | result[0] = '\0'; | |
3597 | return (result); | |
3598 | } | |
3599 | ||
3600 | /* If no character in the string can be quoted, don't bother examining | |
3601 | each character. Just return a copy of the string passed to us. */ | |
3602 | if (strchr (string, CTLESC) == NULL) | |
3603 | return (strcpy (result, string)); | |
3604 | ||
3605 | send = string + slen; | |
3606 | s = string; | |
3607 | while (*s) | |
3608 | { | |
3609 | if (*s == CTLESC) | |
3610 | { | |
3611 | s++; | |
3612 | if (*s == '\0') | |
3613 | break; | |
3614 | } | |
3615 | COPY_CHAR_P (t, s, send); | |
3616 | } | |
3617 | ||
3618 | *t = '\0'; | |
3619 | return (result); | |
3620 | } | |
3621 | ||
3622 | /* Quote the entire WORD_LIST list. */ | |
3623 | static WORD_LIST * | |
3624 | quote_list (list) | |
3625 | WORD_LIST *list; | |
3626 | { | |
3627 | register WORD_LIST *w; | |
3628 | char *t; | |
3629 | ||
3630 | for (w = list; w; w = w->next) | |
3631 | { | |
3632 | t = w->word->word; | |
3633 | w->word->word = quote_string (t); | |
3634 | if (*t == 0) | |
3635 | w->word->flags |= W_HASQUOTEDNULL; /* XXX - turn on W_HASQUOTEDNULL here? */ | |
3636 | w->word->flags |= W_QUOTED; | |
3637 | free (t); | |
3638 | } | |
3639 | return list; | |
3640 | } | |
3641 | ||
3642 | /* De-quote quoted characters in each word in LIST. */ | |
3643 | WORD_LIST * | |
3644 | dequote_list (list) | |
3645 | WORD_LIST *list; | |
3646 | { | |
3647 | register char *s; | |
3648 | register WORD_LIST *tlist; | |
3649 | ||
3650 | for (tlist = list; tlist; tlist = tlist->next) | |
3651 | { | |
3652 | s = dequote_string (tlist->word->word); | |
3653 | if (QUOTED_NULL (tlist->word->word)) | |
3654 | tlist->word->flags &= ~W_HASQUOTEDNULL; | |
3655 | free (tlist->word->word); | |
3656 | tlist->word->word = s; | |
3657 | } | |
3658 | return list; | |
3659 | } | |
3660 | ||
3661 | /* Remove CTLESC protecting a CTLESC or CTLNUL in place. Return the passed | |
3662 | string. */ | |
3663 | char * | |
3664 | remove_quoted_escapes (string) | |
3665 | char *string; | |
3666 | { | |
3667 | char *t; | |
3668 | ||
3669 | if (string) | |
3670 | { | |
3671 | t = dequote_escapes (string); | |
3672 | strcpy (string, t); | |
3673 | free (t); | |
3674 | } | |
3675 | ||
3676 | return (string); | |
3677 | } | |
3678 | ||
3679 | /* Perform quoted null character removal on STRING. We don't allow any | |
3680 | quoted null characters in the middle or at the ends of strings because | |
3681 | of how expand_word_internal works. remove_quoted_nulls () turns | |
3682 | STRING into an empty string iff it only consists of a quoted null, | |
3683 | and removes all unquoted CTLNUL characters. */ | |
3684 | char * | |
3685 | remove_quoted_nulls (string) | |
3686 | char *string; | |
3687 | { | |
3688 | register size_t slen; | |
3689 | register int i, j, prev_i; | |
3690 | DECLARE_MBSTATE; | |
3691 | ||
3692 | if (strchr (string, CTLNUL) == 0) /* XXX */ | |
3693 | return string; /* XXX */ | |
3694 | ||
3695 | slen = strlen (string); | |
3696 | i = j = 0; | |
3697 | ||
3698 | while (i < slen) | |
3699 | { | |
3700 | if (string[i] == CTLESC) | |
3701 | { | |
3702 | /* Old code had j++, but we cannot assume that i == j at this | |
3703 | point -- what if a CTLNUL has already been removed from the | |
3704 | string? We don't want to drop the CTLESC or recopy characters | |
3705 | that we've already copied down. */ | |
3706 | i++; string[j++] = CTLESC; | |
3707 | if (i == slen) | |
3708 | break; | |
3709 | } | |
3710 | else if (string[i] == CTLNUL) | |
3711 | i++; | |
3712 | ||
3713 | prev_i = i; | |
3714 | ADVANCE_CHAR (string, slen, i); | |
3715 | if (j < prev_i) | |
3716 | { | |
3717 | do string[j++] = string[prev_i++]; while (prev_i < i); | |
3718 | } | |
3719 | else | |
3720 | j = i; | |
3721 | } | |
3722 | string[j] = '\0'; | |
3723 | ||
3724 | return (string); | |
3725 | } | |
3726 | ||
3727 | /* Perform quoted null character removal on each element of LIST. | |
3728 | This modifies LIST. */ | |
3729 | void | |
3730 | word_list_remove_quoted_nulls (list) | |
3731 | WORD_LIST *list; | |
3732 | { | |
3733 | register WORD_LIST *t; | |
3734 | ||
3735 | for (t = list; t; t = t->next) | |
3736 | { | |
3737 | remove_quoted_nulls (t->word->word); | |
3738 | t->word->flags &= ~W_HASQUOTEDNULL; | |
3739 | } | |
3740 | } | |
3741 | ||
3742 | /* **************************************************************** */ | |
3743 | /* */ | |
3744 | /* Functions for Matching and Removing Patterns */ | |
3745 | /* */ | |
3746 | /* **************************************************************** */ | |
3747 | ||
3748 | #if defined (HANDLE_MULTIBYTE) | |
3749 | #if 0 /* Currently unused */ | |
3750 | static unsigned char * | |
3751 | mb_getcharlens (string, len) | |
3752 | char *string; | |
3753 | int len; | |
3754 | { | |
3755 | int i, offset, last; | |
3756 | unsigned char *ret; | |
3757 | char *p; | |
3758 | DECLARE_MBSTATE; | |
3759 | ||
3760 | i = offset = 0; | |
3761 | last = 0; | |
3762 | ret = (unsigned char *)xmalloc (len); | |
3763 | memset (ret, 0, len); | |
3764 | while (string[last]) | |
3765 | { | |
3766 | ADVANCE_CHAR (string, len, offset); | |
3767 | ret[last] = offset - last; | |
3768 | last = offset; | |
3769 | } | |
3770 | return ret; | |
3771 | } | |
3772 | #endif | |
3773 | #endif | |
3774 | ||
3775 | /* Remove the portion of PARAM matched by PATTERN according to OP, where OP | |
3776 | can have one of 4 values: | |
3777 | RP_LONG_LEFT remove longest matching portion at start of PARAM | |
3778 | RP_SHORT_LEFT remove shortest matching portion at start of PARAM | |
3779 | RP_LONG_RIGHT remove longest matching portion at end of PARAM | |
3780 | RP_SHORT_RIGHT remove shortest matching portion at end of PARAM | |
3781 | */ | |
3782 | ||
3783 | #define RP_LONG_LEFT 1 | |
3784 | #define RP_SHORT_LEFT 2 | |
3785 | #define RP_LONG_RIGHT 3 | |
3786 | #define RP_SHORT_RIGHT 4 | |
3787 | ||
3788 | /* Returns its first argument if nothing matched; new memory otherwise */ | |
3789 | static char * | |
3790 | remove_upattern (param, pattern, op) | |
3791 | char *param, *pattern; | |
3792 | int op; | |
3793 | { | |
3794 | register int len; | |
3795 | register char *end; | |
3796 | register char *p, *ret, c; | |
3797 | ||
3798 | len = STRLEN (param); | |
3799 | end = param + len; | |
3800 | ||
3801 | switch (op) | |
3802 | { | |
3803 | case RP_LONG_LEFT: /* remove longest match at start */ | |
3804 | for (p = end; p >= param; p--) | |
3805 | { | |
3806 | c = *p; *p = '\0'; | |
3807 | if (strmatch (pattern, param, FNMATCH_EXTFLAG) != FNM_NOMATCH) | |
3808 | { | |
3809 | *p = c; | |
3810 | return (savestring (p)); | |
3811 | } | |
3812 | *p = c; | |
3813 | ||
3814 | } | |
3815 | break; | |
3816 | ||
3817 | case RP_SHORT_LEFT: /* remove shortest match at start */ | |
3818 | for (p = param; p <= end; p++) | |
3819 | { | |
3820 | c = *p; *p = '\0'; | |
3821 | if (strmatch (pattern, param, FNMATCH_EXTFLAG) != FNM_NOMATCH) | |
3822 | { | |
3823 | *p = c; | |
3824 | return (savestring (p)); | |
3825 | } | |
3826 | *p = c; | |
3827 | } | |
3828 | break; | |
3829 | ||
3830 | case RP_LONG_RIGHT: /* remove longest match at end */ | |
3831 | for (p = param; p <= end; p++) | |
3832 | { | |
3833 | if (strmatch (pattern, p, FNMATCH_EXTFLAG) != FNM_NOMATCH) | |
3834 | { | |
3835 | c = *p; *p = '\0'; | |
3836 | ret = savestring (param); | |
3837 | *p = c; | |
3838 | return (ret); | |
3839 | } | |
3840 | } | |
3841 | break; | |
3842 | ||
3843 | case RP_SHORT_RIGHT: /* remove shortest match at end */ | |
3844 | for (p = end; p >= param; p--) | |
3845 | { | |
3846 | if (strmatch (pattern, p, FNMATCH_EXTFLAG) != FNM_NOMATCH) | |
3847 | { | |
3848 | c = *p; *p = '\0'; | |
3849 | ret = savestring (param); | |
3850 | *p = c; | |
3851 | return (ret); | |
3852 | } | |
3853 | } | |
3854 | break; | |
3855 | } | |
3856 | ||
3857 | return (param); /* no match, return original string */ | |
3858 | } | |
3859 | ||
3860 | #if defined (HANDLE_MULTIBYTE) | |
3861 | /* Returns its first argument if nothing matched; new memory otherwise */ | |
3862 | static wchar_t * | |
3863 | remove_wpattern (wparam, wstrlen, wpattern, op) | |
3864 | wchar_t *wparam; | |
3865 | size_t wstrlen; | |
3866 | wchar_t *wpattern; | |
3867 | int op; | |
3868 | { | |
3869 | wchar_t wc, *ret; | |
3870 | int n; | |
3871 | ||
3872 | switch (op) | |
3873 | { | |
3874 | case RP_LONG_LEFT: /* remove longest match at start */ | |
3875 | for (n = wstrlen; n >= 0; n--) | |
3876 | { | |
3877 | wc = wparam[n]; wparam[n] = L'\0'; | |
3878 | if (wcsmatch (wpattern, wparam, FNMATCH_EXTFLAG) != FNM_NOMATCH) | |
3879 | { | |
3880 | wparam[n] = wc; | |
3881 | return (wcsdup (wparam + n)); | |
3882 | } | |
3883 | wparam[n] = wc; | |
3884 | } | |
3885 | break; | |
3886 | ||
3887 | case RP_SHORT_LEFT: /* remove shortest match at start */ | |
3888 | for (n = 0; n <= wstrlen; n++) | |
3889 | { | |
3890 | wc = wparam[n]; wparam[n] = L'\0'; | |
3891 | if (wcsmatch (wpattern, wparam, FNMATCH_EXTFLAG) != FNM_NOMATCH) | |
3892 | { | |
3893 | wparam[n] = wc; | |
3894 | return (wcsdup (wparam + n)); | |
3895 | } | |
3896 | wparam[n] = wc; | |
3897 | } | |
3898 | break; | |
3899 | ||
3900 | case RP_LONG_RIGHT: /* remove longest match at end */ | |
3901 | for (n = 0; n <= wstrlen; n++) | |
3902 | { | |
3903 | if (wcsmatch (wpattern, wparam + n, FNMATCH_EXTFLAG) != FNM_NOMATCH) | |
3904 | { | |
3905 | wc = wparam[n]; wparam[n] = L'\0'; | |
3906 | ret = wcsdup (wparam); | |
3907 | wparam[n] = wc; | |
3908 | return (ret); | |
3909 | } | |
3910 | } | |
3911 | break; | |
3912 | ||
3913 | case RP_SHORT_RIGHT: /* remove shortest match at end */ | |
3914 | for (n = wstrlen; n >= 0; n--) | |
3915 | { | |
3916 | if (wcsmatch (wpattern, wparam + n, FNMATCH_EXTFLAG) != FNM_NOMATCH) | |
3917 | { | |
3918 | wc = wparam[n]; wparam[n] = L'\0'; | |
3919 | ret = wcsdup (wparam); | |
3920 | wparam[n] = wc; | |
3921 | return (ret); | |
3922 | } | |
3923 | } | |
3924 | break; | |
3925 | } | |
3926 | ||
3927 | return (wparam); /* no match, return original string */ | |
3928 | } | |
3929 | #endif /* HANDLE_MULTIBYTE */ | |
3930 | ||
3931 | static char * | |
3932 | remove_pattern (param, pattern, op) | |
3933 | char *param, *pattern; | |
3934 | int op; | |
3935 | { | |
3936 | char *xret; | |
3937 | ||
3938 | if (param == NULL) | |
3939 | return (param); | |
3940 | if (*param == '\0' || pattern == NULL || *pattern == '\0') /* minor optimization */ | |
3941 | return (savestring (param)); | |
3942 | ||
3943 | #if defined (HANDLE_MULTIBYTE) | |
3944 | if (MB_CUR_MAX > 1) | |
3945 | { | |
3946 | wchar_t *ret, *oret; | |
3947 | size_t n; | |
3948 | wchar_t *wparam, *wpattern; | |
3949 | mbstate_t ps; | |
3950 | ||
3951 | n = xdupmbstowcs (&wpattern, NULL, pattern); | |
3952 | if (n == (size_t)-1) | |
3953 | { | |
3954 | xret = remove_upattern (param, pattern, op); | |
3955 | return ((xret == param) ? savestring (param) : xret); | |
3956 | } | |
3957 | n = xdupmbstowcs (&wparam, NULL, param); | |
3958 | if (n == (size_t)-1) | |
3959 | { | |
3960 | free (wpattern); | |
3961 | xret = remove_upattern (param, pattern, op); | |
3962 | return ((xret == param) ? savestring (param) : xret); | |
3963 | } | |
3964 | oret = ret = remove_wpattern (wparam, n, wpattern, op); | |
3965 | /* Don't bother to convert wparam back to multibyte string if nothing | |
3966 | matched; just return copy of original string */ | |
3967 | if (ret == wparam) | |
3968 | { | |
3969 | free (wparam); | |
3970 | free (wpattern); | |
3971 | return (savestring (param)); | |
3972 | } | |
3973 | ||
3974 | free (wparam); | |
3975 | free (wpattern); | |
3976 | ||
3977 | n = strlen (param); | |
3978 | xret = (char *)xmalloc (n + 1); | |
3979 | memset (&ps, '\0', sizeof (mbstate_t)); | |
3980 | n = wcsrtombs (xret, (const wchar_t **)&ret, n, &ps); | |
3981 | xret[n] = '\0'; /* just to make sure */ | |
3982 | free (oret); | |
3983 | return xret; | |
3984 | } | |
3985 | else | |
3986 | #endif | |
3987 | { | |
3988 | xret = remove_upattern (param, pattern, op); | |
3989 | return ((xret == param) ? savestring (param) : xret); | |
3990 | } | |
3991 | } | |
3992 | ||
3993 | /* Match PAT anywhere in STRING and return the match boundaries. | |
3994 | This returns 1 in case of a successful match, 0 otherwise. SP | |
3995 | and EP are pointers into the string where the match begins and | |
3996 | ends, respectively. MTYPE controls what kind of match is attempted. | |
3997 | MATCH_BEG and MATCH_END anchor the match at the beginning and end | |
3998 | of the string, respectively. The longest match is returned. */ | |
3999 | static int | |
4000 | match_upattern (string, pat, mtype, sp, ep) | |
4001 | char *string, *pat; | |
4002 | int mtype; | |
4003 | char **sp, **ep; | |
4004 | { | |
4005 | int c, len, mlen; | |
4006 | register char *p, *p1, *npat; | |
4007 | char *end; | |
4008 | int n1; | |
4009 | ||
4010 | /* If the pattern doesn't match anywhere in the string, go ahead and | |
4011 | short-circuit right away. A minor optimization, saves a bunch of | |
4012 | unnecessary calls to strmatch (up to N calls for a string of N | |
4013 | characters) if the match is unsuccessful. To preserve the semantics | |
4014 | of the substring matches below, we make sure that the pattern has | |
4015 | `*' as first and last character, making a new pattern if necessary. */ | |
4016 | /* XXX - check this later if I ever implement `**' with special meaning, | |
4017 | since this will potentially result in `**' at the beginning or end */ | |
4018 | len = STRLEN (pat); | |
4019 | if (pat[0] != '*' || (pat[0] == '*' && pat[1] == LPAREN && extended_glob) || pat[len - 1] != '*') | |
4020 | { | |
4021 | p = npat = (char *)xmalloc (len + 3); | |
4022 | p1 = pat; | |
4023 | if (*p1 != '*' || (*p1 == '*' && p1[1] == LPAREN && extended_glob)) | |
4024 | *p++ = '*'; | |
4025 | while (*p1) | |
4026 | *p++ = *p1++; | |
4027 | if (p1[-1] != '*' || p[-2] == '\\') | |
4028 | *p++ = '*'; | |
4029 | *p = '\0'; | |
4030 | } | |
4031 | else | |
4032 | npat = pat; | |
4033 | c = strmatch (npat, string, FNMATCH_EXTFLAG); | |
4034 | if (npat != pat) | |
4035 | free (npat); | |
4036 | if (c == FNM_NOMATCH) | |
4037 | return (0); | |
4038 | ||
4039 | len = STRLEN (string); | |
4040 | end = string + len; | |
4041 | ||
4042 | mlen = umatchlen (pat, len); | |
4043 | ||
4044 | switch (mtype) | |
4045 | { | |
4046 | case MATCH_ANY: | |
4047 | for (p = string; p <= end; p++) | |
4048 | { | |
4049 | if (match_pattern_char (pat, p)) | |
4050 | { | |
4051 | #if 0 | |
4052 | for (p1 = end; p1 >= p; p1--) | |
4053 | #else | |
4054 | p1 = (mlen == -1) ? end : p + mlen; | |
4055 | /* p1 - p = length of portion of string to be considered | |
4056 | p = current position in string | |
4057 | mlen = number of characters consumed by match (-1 for entire string) | |
4058 | end = end of string | |
4059 | we want to break immediately if the potential match len | |
4060 | is greater than the number of characters remaining in the | |
4061 | string | |
4062 | */ | |
4063 | if (p1 > end) | |
4064 | break; | |
4065 | for ( ; p1 >= p; p1--) | |
4066 | #endif | |
4067 | { | |
4068 | c = *p1; *p1 = '\0'; | |
4069 | if (strmatch (pat, p, FNMATCH_EXTFLAG) == 0) | |
4070 | { | |
4071 | *p1 = c; | |
4072 | *sp = p; | |
4073 | *ep = p1; | |
4074 | return 1; | |
4075 | } | |
4076 | *p1 = c; | |
4077 | #if 1 | |
4078 | /* If MLEN != -1, we have a fixed length pattern. */ | |
4079 | if (mlen != -1) | |
4080 | break; | |
4081 | #endif | |
4082 | } | |
4083 | } | |
4084 | } | |
4085 | ||
4086 | return (0); | |
4087 | ||
4088 | case MATCH_BEG: | |
4089 | if (match_pattern_char (pat, string) == 0) | |
4090 | return (0); | |
4091 | ||
4092 | #if 0 | |
4093 | for (p = end; p >= string; p--) | |
4094 | #else | |
4095 | for (p = (mlen == -1) ? end : string + mlen; p >= string; p--) | |
4096 | #endif | |
4097 | { | |
4098 | c = *p; *p = '\0'; | |
4099 | if (strmatch (pat, string, FNMATCH_EXTFLAG) == 0) | |
4100 | { | |
4101 | *p = c; | |
4102 | *sp = string; | |
4103 | *ep = p; | |
4104 | return 1; | |
4105 | } | |
4106 | *p = c; | |
4107 | #if 1 | |
4108 | /* If MLEN != -1, we have a fixed length pattern. */ | |
4109 | if (mlen != -1) | |
4110 | break; | |
4111 | #endif | |
4112 | } | |
4113 | ||
4114 | return (0); | |
4115 | ||
4116 | case MATCH_END: | |
4117 | #if 0 | |
4118 | for (p = string; p <= end; p++) | |
4119 | #else | |
4120 | for (p = end - ((mlen == -1) ? len : mlen); p <= end; p++) | |
4121 | #endif | |
4122 | { | |
4123 | if (strmatch (pat, p, FNMATCH_EXTFLAG) == 0) | |
4124 | { | |
4125 | *sp = p; | |
4126 | *ep = end; | |
4127 | return 1; | |
4128 | } | |
4129 | #if 1 | |
4130 | /* If MLEN != -1, we have a fixed length pattern. */ | |
4131 | if (mlen != -1) | |
4132 | break; | |
4133 | #endif | |
4134 | } | |
4135 | ||
4136 | return (0); | |
4137 | } | |
4138 | ||
4139 | return (0); | |
4140 | } | |
4141 | ||
4142 | #if defined (HANDLE_MULTIBYTE) | |
4143 | /* Match WPAT anywhere in WSTRING and return the match boundaries. | |
4144 | This returns 1 in case of a successful match, 0 otherwise. Wide | |
4145 | character version. */ | |
4146 | static int | |
4147 | match_wpattern (wstring, indices, wstrlen, wpat, mtype, sp, ep) | |
4148 | wchar_t *wstring; | |
4149 | char **indices; | |
4150 | size_t wstrlen; | |
4151 | wchar_t *wpat; | |
4152 | int mtype; | |
4153 | char **sp, **ep; | |
4154 | { | |
4155 | wchar_t wc, *wp, *nwpat, *wp1; | |
4156 | size_t len; | |
4157 | int mlen; | |
4158 | int n, n1, n2, simple; | |
4159 | ||
4160 | simple = (wpat[0] != L'\\' && wpat[0] != L'*' && wpat[0] != L'?' && wpat[0] != L'['); | |
4161 | #if defined (EXTENDED_GLOB) | |
4162 | if (extended_glob) | |
4163 | simple |= (wpat[1] != L'(' || (wpat[0] != L'*' && wpat[0] != L'?' && wpat[0] != L'+' && wpat[0] != L'!' && wpat[0] != L'@')); /*)*/ | |
4164 | #endif | |
4165 | ||
4166 | /* If the pattern doesn't match anywhere in the string, go ahead and | |
4167 | short-circuit right away. A minor optimization, saves a bunch of | |
4168 | unnecessary calls to strmatch (up to N calls for a string of N | |
4169 | characters) if the match is unsuccessful. To preserve the semantics | |
4170 | of the substring matches below, we make sure that the pattern has | |
4171 | `*' as first and last character, making a new pattern if necessary. */ | |
4172 | len = wcslen (wpat); | |
4173 | if (wpat[0] != L'*' || (wpat[0] == L'*' && wpat[1] == WLPAREN && extended_glob) || wpat[len - 1] != L'*') | |
4174 | { | |
4175 | wp = nwpat = (wchar_t *)xmalloc ((len + 3) * sizeof (wchar_t)); | |
4176 | wp1 = wpat; | |
4177 | if (*wp1 != L'*' || (*wp1 == '*' && wp1[1] == WLPAREN && extended_glob)) | |
4178 | *wp++ = L'*'; | |
4179 | while (*wp1 != L'\0') | |
4180 | *wp++ = *wp1++; | |
4181 | if (wp1[-1] != L'*' || wp1[-2] == L'\\') | |
4182 | *wp++ = L'*'; | |
4183 | *wp = '\0'; | |
4184 | } | |
4185 | else | |
4186 | nwpat = wpat; | |
4187 | len = wcsmatch (nwpat, wstring, FNMATCH_EXTFLAG); | |
4188 | if (nwpat != wpat) | |
4189 | free (nwpat); | |
4190 | if (len == FNM_NOMATCH) | |
4191 | return (0); | |
4192 | ||
4193 | mlen = wmatchlen (wpat, wstrlen); | |
4194 | ||
4195 | /* itrace("wmatchlen (%ls) -> %d", wpat, mlen); */ | |
4196 | switch (mtype) | |
4197 | { | |
4198 | case MATCH_ANY: | |
4199 | for (n = 0; n <= wstrlen; n++) | |
4200 | { | |
4201 | #if 1 | |
4202 | n2 = simple ? (*wpat == wstring[n]) : match_pattern_wchar (wpat, wstring + n); | |
4203 | #else | |
4204 | n2 = match_pattern_wchar (wpat, wstring + n); | |
4205 | #endif | |
4206 | if (n2) | |
4207 | { | |
4208 | #if 0 | |
4209 | for (n1 = wstrlen; n1 >= n; n1--) | |
4210 | #else | |
4211 | n1 = (mlen == -1) ? wstrlen : n + mlen; | |
4212 | if (n1 > wstrlen) | |
4213 | break; | |
4214 | ||
4215 | for ( ; n1 >= n; n1--) | |
4216 | #endif | |
4217 | { | |
4218 | wc = wstring[n1]; wstring[n1] = L'\0'; | |
4219 | if (wcsmatch (wpat, wstring + n, FNMATCH_EXTFLAG) == 0) | |
4220 | { | |
4221 | wstring[n1] = wc; | |
4222 | *sp = indices[n]; | |
4223 | *ep = indices[n1]; | |
4224 | return 1; | |
4225 | } | |
4226 | wstring[n1] = wc; | |
4227 | #if 1 | |
4228 | /* If MLEN != -1, we have a fixed length pattern. */ | |
4229 | if (mlen != -1) | |
4230 | break; | |
4231 | #endif | |
4232 | } | |
4233 | } | |
4234 | } | |
4235 | ||
4236 | return (0); | |
4237 | ||
4238 | case MATCH_BEG: | |
4239 | if (match_pattern_wchar (wpat, wstring) == 0) | |
4240 | return (0); | |
4241 | ||
4242 | #if 0 | |
4243 | for (n = wstrlen; n >= 0; n--) | |
4244 | #else | |
4245 | for (n = (mlen == -1) ? wstrlen : mlen; n >= 0; n--) | |
4246 | #endif | |
4247 | { | |
4248 | wc = wstring[n]; wstring[n] = L'\0'; | |
4249 | if (wcsmatch (wpat, wstring, FNMATCH_EXTFLAG) == 0) | |
4250 | { | |
4251 | wstring[n] = wc; | |
4252 | *sp = indices[0]; | |
4253 | *ep = indices[n]; | |
4254 | return 1; | |
4255 | } | |
4256 | wstring[n] = wc; | |
4257 | #if 1 | |
4258 | /* If MLEN != -1, we have a fixed length pattern. */ | |
4259 | if (mlen != -1) | |
4260 | break; | |
4261 | #endif | |
4262 | } | |
4263 | ||
4264 | return (0); | |
4265 | ||
4266 | case MATCH_END: | |
4267 | #if 0 | |
4268 | for (n = 0; n <= wstrlen; n++) | |
4269 | #else | |
4270 | for (n = wstrlen - ((mlen == -1) ? wstrlen : mlen); n <= wstrlen; n++) | |
4271 | #endif | |
4272 | { | |
4273 | if (wcsmatch (wpat, wstring + n, FNMATCH_EXTFLAG) == 0) | |
4274 | { | |
4275 | *sp = indices[n]; | |
4276 | *ep = indices[wstrlen]; | |
4277 | return 1; | |
4278 | } | |
4279 | #if 1 | |
4280 | /* If MLEN != -1, we have a fixed length pattern. */ | |
4281 | if (mlen != -1) | |
4282 | break; | |
4283 | #endif | |
4284 | } | |
4285 | ||
4286 | return (0); | |
4287 | } | |
4288 | ||
4289 | return (0); | |
4290 | } | |
4291 | #endif /* HANDLE_MULTIBYTE */ | |
4292 | ||
4293 | static int | |
4294 | match_pattern (string, pat, mtype, sp, ep) | |
4295 | char *string, *pat; | |
4296 | int mtype; | |
4297 | char **sp, **ep; | |
4298 | { | |
4299 | #if defined (HANDLE_MULTIBYTE) | |
4300 | int ret; | |
4301 | size_t n; | |
4302 | wchar_t *wstring, *wpat; | |
4303 | char **indices; | |
4304 | size_t slen, plen, mslen, mplen; | |
4305 | #endif | |
4306 | ||
4307 | if (string == 0 || *string == 0 || pat == 0 || *pat == 0) | |
4308 | return (0); | |
4309 | ||
4310 | #if defined (HANDLE_MULTIBYTE) | |
4311 | if (MB_CUR_MAX > 1) | |
4312 | { | |
4313 | #if 0 | |
4314 | slen = STRLEN (string); | |
4315 | mslen = MBSLEN (string); | |
4316 | plen = STRLEN (pat); | |
4317 | mplen = MBSLEN (pat); | |
4318 | if (slen == mslen && plen == mplen) | |
4319 | #else | |
4320 | if (mbsmbchar (string) == 0 && mbsmbchar (pat) == 0) | |
4321 | #endif | |
4322 | return (match_upattern (string, pat, mtype, sp, ep)); | |
4323 | ||
4324 | n = xdupmbstowcs (&wpat, NULL, pat); | |
4325 | if (n == (size_t)-1) | |
4326 | return (match_upattern (string, pat, mtype, sp, ep)); | |
4327 | n = xdupmbstowcs (&wstring, &indices, string); | |
4328 | if (n == (size_t)-1) | |
4329 | { | |
4330 | free (wpat); | |
4331 | return (match_upattern (string, pat, mtype, sp, ep)); | |
4332 | } | |
4333 | ret = match_wpattern (wstring, indices, n, wpat, mtype, sp, ep); | |
4334 | ||
4335 | free (wpat); | |
4336 | free (wstring); | |
4337 | free (indices); | |
4338 | ||
4339 | return (ret); | |
4340 | } | |
4341 | else | |
4342 | #endif | |
4343 | return (match_upattern (string, pat, mtype, sp, ep)); | |
4344 | } | |
4345 | ||
4346 | static int | |
4347 | getpatspec (c, value) | |
4348 | int c; | |
4349 | char *value; | |
4350 | { | |
4351 | if (c == '#') | |
4352 | return ((*value == '#') ? RP_LONG_LEFT : RP_SHORT_LEFT); | |
4353 | else /* c == '%' */ | |
4354 | return ((*value == '%') ? RP_LONG_RIGHT : RP_SHORT_RIGHT); | |
4355 | } | |
4356 | ||
4357 | /* Posix.2 says that the WORD should be run through tilde expansion, | |
4358 | parameter expansion, command substitution and arithmetic expansion. | |
4359 | This leaves the result quoted, so quote_string_for_globbing () has | |
4360 | to be called to fix it up for strmatch (). If QUOTED is non-zero, | |
4361 | it means that the entire expression was enclosed in double quotes. | |
4362 | This means that quoting characters in the pattern do not make any | |
4363 | special pattern characters quoted. For example, the `*' in the | |
4364 | following retains its special meaning: "${foo#'*'}". */ | |
4365 | static char * | |
4366 | getpattern (value, quoted, expandpat) | |
4367 | char *value; | |
4368 | int quoted, expandpat; | |
4369 | { | |
4370 | char *pat, *tword; | |
4371 | WORD_LIST *l; | |
4372 | #if 0 | |
4373 | int i; | |
4374 | #endif | |
4375 | /* There is a problem here: how to handle single or double quotes in the | |
4376 | pattern string when the whole expression is between double quotes? | |
4377 | POSIX.2 says that enclosing double quotes do not cause the pattern to | |
4378 | be quoted, but does that leave us a problem with @ and array[@] and their | |
4379 | expansions inside a pattern? */ | |
4380 | #if 0 | |
4381 | if (expandpat && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && *tword) | |
4382 | { | |
4383 | i = 0; | |
4384 | pat = string_extract_double_quoted (tword, &i, 1); | |
4385 | free (tword); | |
4386 | tword = pat; | |
4387 | } | |
4388 | #endif | |
4389 | ||
4390 | /* expand_string_for_rhs () leaves WORD quoted and does not perform | |
4391 | word splitting. */ | |
4392 | l = *value ? expand_string_for_rhs (value, | |
4393 | (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) ? Q_PATQUOTE : quoted, | |
4394 | (int *)NULL, (int *)NULL) | |
4395 | : (WORD_LIST *)0; | |
4396 | pat = string_list (l); | |
4397 | dispose_words (l); | |
4398 | if (pat) | |
4399 | { | |
4400 | tword = quote_string_for_globbing (pat, QGLOB_CVTNULL); | |
4401 | free (pat); | |
4402 | pat = tword; | |
4403 | } | |
4404 | return (pat); | |
4405 | } | |
4406 | ||
4407 | #if 0 | |
4408 | /* Handle removing a pattern from a string as a result of ${name%[%]value} | |
4409 | or ${name#[#]value}. */ | |
4410 | static char * | |
4411 | variable_remove_pattern (value, pattern, patspec, quoted) | |
4412 | char *value, *pattern; | |
4413 | int patspec, quoted; | |
4414 | { | |
4415 | char *tword; | |
4416 | ||
4417 | tword = remove_pattern (value, pattern, patspec); | |
4418 | ||
4419 | return (tword); | |
4420 | } | |
4421 | #endif | |
4422 | ||
4423 | static char * | |
4424 | list_remove_pattern (list, pattern, patspec, itype, quoted) | |
4425 | WORD_LIST *list; | |
4426 | char *pattern; | |
4427 | int patspec, itype, quoted; | |
4428 | { | |
4429 | WORD_LIST *new, *l; | |
4430 | WORD_DESC *w; | |
4431 | char *tword; | |
4432 | ||
4433 | for (new = (WORD_LIST *)NULL, l = list; l; l = l->next) | |
4434 | { | |
4435 | tword = remove_pattern (l->word->word, pattern, patspec); | |
4436 | w = alloc_word_desc (); | |
4437 | w->word = tword ? tword : savestring (""); | |
4438 | new = make_word_list (w, new); | |
4439 | } | |
4440 | ||
4441 | l = REVERSE_LIST (new, WORD_LIST *); | |
4442 | tword = string_list_pos_params (itype, l, quoted); | |
4443 | dispose_words (l); | |
4444 | ||
4445 | return (tword); | |
4446 | } | |
4447 | ||
4448 | static char * | |
4449 | parameter_list_remove_pattern (itype, pattern, patspec, quoted) | |
4450 | int itype; | |
4451 | char *pattern; | |
4452 | int patspec, quoted; | |
4453 | { | |
4454 | char *ret; | |
4455 | WORD_LIST *list; | |
4456 | ||
4457 | list = list_rest_of_args (); | |
4458 | if (list == 0) | |
4459 | return ((char *)NULL); | |
4460 | ret = list_remove_pattern (list, pattern, patspec, itype, quoted); | |
4461 | dispose_words (list); | |
4462 | return (ret); | |
4463 | } | |
4464 | ||
4465 | #if defined (ARRAY_VARS) | |
4466 | static char * | |
4467 | array_remove_pattern (var, pattern, patspec, varname, quoted) | |
4468 | SHELL_VAR *var; | |
4469 | char *pattern; | |
4470 | int patspec; | |
4471 | char *varname; /* so we can figure out how it's indexed */ | |
4472 | int quoted; | |
4473 | { | |
4474 | ARRAY *a; | |
4475 | HASH_TABLE *h; | |
4476 | int itype; | |
4477 | char *ret; | |
4478 | WORD_LIST *list; | |
4479 | SHELL_VAR *v; | |
4480 | ||
4481 | /* compute itype from varname here */ | |
4482 | v = array_variable_part (varname, &ret, 0); | |
4483 | itype = ret[0]; | |
4484 | ||
4485 | a = (v && array_p (v)) ? array_cell (v) : 0; | |
4486 | h = (v && assoc_p (v)) ? assoc_cell (v) : 0; | |
4487 | ||
4488 | list = a ? array_to_word_list (a) : (h ? assoc_to_word_list (h) : 0); | |
4489 | if (list == 0) | |
4490 | return ((char *)NULL); | |
4491 | ret = list_remove_pattern (list, pattern, patspec, itype, quoted); | |
4492 | dispose_words (list); | |
4493 | ||
4494 | return ret; | |
4495 | } | |
4496 | #endif /* ARRAY_VARS */ | |
4497 | ||
4498 | static char * | |
4499 | parameter_brace_remove_pattern (varname, value, ind, patstr, rtype, quoted, flags) | |
4500 | char *varname, *value; | |
4501 | int ind; | |
4502 | char *patstr; | |
4503 | int rtype, quoted, flags; | |
4504 | { | |
4505 | int vtype, patspec, starsub; | |
4506 | char *temp1, *val, *pattern; | |
4507 | SHELL_VAR *v; | |
4508 | ||
4509 | if (value == 0) | |
4510 | return ((char *)NULL); | |
4511 | ||
4512 | this_command_name = varname; | |
4513 | ||
4514 | vtype = get_var_and_type (varname, value, ind, quoted, flags, &v, &val); | |
4515 | if (vtype == -1) | |
4516 | return ((char *)NULL); | |
4517 | ||
4518 | starsub = vtype & VT_STARSUB; | |
4519 | vtype &= ~VT_STARSUB; | |
4520 | ||
4521 | patspec = getpatspec (rtype, patstr); | |
4522 | if (patspec == RP_LONG_LEFT || patspec == RP_LONG_RIGHT) | |
4523 | patstr++; | |
4524 | ||
4525 | /* Need to pass getpattern newly-allocated memory in case of expansion -- | |
4526 | the expansion code will free the passed string on an error. */ | |
4527 | temp1 = savestring (patstr); | |
4528 | pattern = getpattern (temp1, quoted, 1); | |
4529 | free (temp1); | |
4530 | ||
4531 | temp1 = (char *)NULL; /* shut up gcc */ | |
4532 | switch (vtype) | |
4533 | { | |
4534 | case VT_VARIABLE: | |
4535 | case VT_ARRAYMEMBER: | |
4536 | temp1 = remove_pattern (val, pattern, patspec); | |
4537 | if (vtype == VT_VARIABLE) | |
4538 | FREE (val); | |
4539 | if (temp1) | |
4540 | { | |
4541 | val = (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
4542 | ? quote_string (temp1) | |
4543 | : quote_escapes (temp1); | |
4544 | free (temp1); | |
4545 | temp1 = val; | |
4546 | } | |
4547 | break; | |
4548 | #if defined (ARRAY_VARS) | |
4549 | case VT_ARRAYVAR: | |
4550 | temp1 = array_remove_pattern (v, pattern, patspec, varname, quoted); | |
4551 | if (temp1 && ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) == 0)) | |
4552 | { | |
4553 | val = quote_escapes (temp1); | |
4554 | free (temp1); | |
4555 | temp1 = val; | |
4556 | } | |
4557 | break; | |
4558 | #endif | |
4559 | case VT_POSPARMS: | |
4560 | temp1 = parameter_list_remove_pattern (varname[0], pattern, patspec, quoted); | |
4561 | if (temp1 && ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) == 0)) | |
4562 | { | |
4563 | val = quote_escapes (temp1); | |
4564 | free (temp1); | |
4565 | temp1 = val; | |
4566 | } | |
4567 | break; | |
4568 | } | |
4569 | ||
4570 | FREE (pattern); | |
4571 | return temp1; | |
4572 | } | |
4573 | ||
4574 | /******************************************* | |
4575 | * * | |
4576 | * Functions to expand WORD_DESCs * | |
4577 | * * | |
4578 | *******************************************/ | |
4579 | ||
4580 | /* Expand WORD, performing word splitting on the result. This does | |
4581 | parameter expansion, command substitution, arithmetic expansion, | |
4582 | word splitting, and quote removal. */ | |
4583 | ||
4584 | WORD_LIST * | |
4585 | expand_word (word, quoted) | |
4586 | WORD_DESC *word; | |
4587 | int quoted; | |
4588 | { | |
4589 | WORD_LIST *result, *tresult; | |
4590 | ||
4591 | tresult = call_expand_word_internal (word, quoted, 0, (int *)NULL, (int *)NULL); | |
4592 | result = word_list_split (tresult); | |
4593 | dispose_words (tresult); | |
4594 | return (result ? dequote_list (result) : result); | |
4595 | } | |
4596 | ||
4597 | /* Expand WORD, but do not perform word splitting on the result. This | |
4598 | does parameter expansion, command substitution, arithmetic expansion, | |
4599 | and quote removal. */ | |
4600 | WORD_LIST * | |
4601 | expand_word_unsplit (word, quoted) | |
4602 | WORD_DESC *word; | |
4603 | int quoted; | |
4604 | { | |
4605 | WORD_LIST *result; | |
4606 | ||
4607 | expand_no_split_dollar_star = 1; | |
4608 | #if defined (HANDLE_MULTIBYTE) | |
4609 | if (ifs_firstc[0] == 0) | |
4610 | #else | |
4611 | if (ifs_firstc == 0) | |
4612 | #endif | |
4613 | word->flags |= W_NOSPLIT; | |
4614 | result = call_expand_word_internal (word, quoted, 0, (int *)NULL, (int *)NULL); | |
4615 | expand_no_split_dollar_star = 0; | |
4616 | ||
4617 | return (result ? dequote_list (result) : result); | |
4618 | } | |
4619 | ||
4620 | /* Perform shell expansions on WORD, but do not perform word splitting or | |
4621 | quote removal on the result. Virtually identical to expand_word_unsplit; | |
4622 | could be combined if implementations don't diverge. */ | |
4623 | WORD_LIST * | |
4624 | expand_word_leave_quoted (word, quoted) | |
4625 | WORD_DESC *word; | |
4626 | int quoted; | |
4627 | { | |
4628 | WORD_LIST *result; | |
4629 | ||
4630 | expand_no_split_dollar_star = 1; | |
4631 | #if defined (HANDLE_MULTIBYTE) | |
4632 | if (ifs_firstc[0] == 0) | |
4633 | #else | |
4634 | if (ifs_firstc == 0) | |
4635 | #endif | |
4636 | word->flags |= W_NOSPLIT; | |
4637 | word->flags |= W_NOSPLIT2; | |
4638 | result = call_expand_word_internal (word, quoted, 0, (int *)NULL, (int *)NULL); | |
4639 | expand_no_split_dollar_star = 0; | |
4640 | ||
4641 | return result; | |
4642 | } | |
4643 | ||
4644 | #if defined (PROCESS_SUBSTITUTION) | |
4645 | ||
4646 | /*****************************************************************/ | |
4647 | /* */ | |
4648 | /* Hacking Process Substitution */ | |
4649 | /* */ | |
4650 | /*****************************************************************/ | |
4651 | ||
4652 | #if !defined (HAVE_DEV_FD) | |
4653 | /* Named pipes must be removed explicitly with `unlink'. This keeps a list | |
4654 | of FIFOs the shell has open. unlink_fifo_list will walk the list and | |
4655 | unlink all of them. add_fifo_list adds the name of an open FIFO to the | |
4656 | list. NFIFO is a count of the number of FIFOs in the list. */ | |
4657 | #define FIFO_INCR 20 | |
4658 | ||
4659 | struct temp_fifo { | |
4660 | char *file; | |
4661 | pid_t proc; | |
4662 | }; | |
4663 | ||
4664 | static struct temp_fifo *fifo_list = (struct temp_fifo *)NULL; | |
4665 | static int nfifo; | |
4666 | static int fifo_list_size; | |
4667 | ||
4668 | char * | |
4669 | copy_fifo_list (sizep) | |
4670 | int *sizep; | |
4671 | { | |
4672 | if (sizep) | |
4673 | *sizep = 0; | |
4674 | return (char *)NULL; | |
4675 | } | |
4676 | ||
4677 | static void | |
4678 | add_fifo_list (pathname) | |
4679 | char *pathname; | |
4680 | { | |
4681 | if (nfifo >= fifo_list_size - 1) | |
4682 | { | |
4683 | fifo_list_size += FIFO_INCR; | |
4684 | fifo_list = (struct temp_fifo *)xrealloc (fifo_list, | |
4685 | fifo_list_size * sizeof (struct temp_fifo)); | |
4686 | } | |
4687 | ||
4688 | fifo_list[nfifo].file = savestring (pathname); | |
4689 | nfifo++; | |
4690 | } | |
4691 | ||
4692 | void | |
4693 | unlink_fifo (i) | |
4694 | int i; | |
4695 | { | |
4696 | if ((fifo_list[i].proc == -1) || (kill(fifo_list[i].proc, 0) == -1)) | |
4697 | { | |
4698 | unlink (fifo_list[i].file); | |
4699 | free (fifo_list[i].file); | |
4700 | fifo_list[i].file = (char *)NULL; | |
4701 | fifo_list[i].proc = -1; | |
4702 | } | |
4703 | } | |
4704 | ||
4705 | void | |
4706 | unlink_fifo_list () | |
4707 | { | |
4708 | int saved, i, j; | |
4709 | ||
4710 | if (nfifo == 0) | |
4711 | return; | |
4712 | ||
4713 | for (i = saved = 0; i < nfifo; i++) | |
4714 | { | |
4715 | if ((fifo_list[i].proc == -1) || (kill(fifo_list[i].proc, 0) == -1)) | |
4716 | { | |
4717 | unlink (fifo_list[i].file); | |
4718 | free (fifo_list[i].file); | |
4719 | fifo_list[i].file = (char *)NULL; | |
4720 | fifo_list[i].proc = -1; | |
4721 | } | |
4722 | else | |
4723 | saved++; | |
4724 | } | |
4725 | ||
4726 | /* If we didn't remove some of the FIFOs, compact the list. */ | |
4727 | if (saved) | |
4728 | { | |
4729 | for (i = j = 0; i < nfifo; i++) | |
4730 | if (fifo_list[i].file) | |
4731 | { | |
4732 | fifo_list[j].file = fifo_list[i].file; | |
4733 | fifo_list[j].proc = fifo_list[i].proc; | |
4734 | j++; | |
4735 | } | |
4736 | nfifo = j; | |
4737 | } | |
4738 | else | |
4739 | nfifo = 0; | |
4740 | } | |
4741 | ||
4742 | /* Take LIST, which is a bitmap denoting active FIFOs in fifo_list | |
4743 | from some point in the past, and close all open FIFOs in fifo_list | |
4744 | that are not marked as active in LIST. If LIST is NULL, close | |
4745 | everything in fifo_list. LSIZE is the number of elements in LIST, in | |
4746 | case it's larger than fifo_list_size (size of fifo_list). */ | |
4747 | void | |
4748 | close_new_fifos (list, lsize) | |
4749 | char *list; | |
4750 | int lsize; | |
4751 | { | |
4752 | int i; | |
4753 | ||
4754 | if (list == 0) | |
4755 | { | |
4756 | unlink_fifo_list (); | |
4757 | return; | |
4758 | } | |
4759 | ||
4760 | for (i = 0; i < lsize; i++) | |
4761 | if (list[i] == 0 && i < fifo_list_size && fifo_list[i].proc != -1) | |
4762 | unlink_fifo (i); | |
4763 | ||
4764 | for (i = lsize; i < fifo_list_size; i++) | |
4765 | unlink_fifo (i); | |
4766 | } | |
4767 | ||
4768 | int | |
4769 | fifos_pending () | |
4770 | { | |
4771 | return nfifo; | |
4772 | } | |
4773 | ||
4774 | int | |
4775 | num_fifos () | |
4776 | { | |
4777 | return nfifo; | |
4778 | } | |
4779 | ||
4780 | static char * | |
4781 | make_named_pipe () | |
4782 | { | |
4783 | char *tname; | |
4784 | ||
4785 | tname = sh_mktmpname ("sh-np", MT_USERANDOM|MT_USETMPDIR); | |
4786 | if (mkfifo (tname, 0600) < 0) | |
4787 | { | |
4788 | free (tname); | |
4789 | return ((char *)NULL); | |
4790 | } | |
4791 | ||
4792 | add_fifo_list (tname); | |
4793 | return (tname); | |
4794 | } | |
4795 | ||
4796 | #else /* HAVE_DEV_FD */ | |
4797 | ||
4798 | /* DEV_FD_LIST is a bitmap of file descriptors attached to pipes the shell | |
4799 | has open to children. NFDS is a count of the number of bits currently | |
4800 | set in DEV_FD_LIST. TOTFDS is a count of the highest possible number | |
4801 | of open files. */ | |
4802 | static char *dev_fd_list = (char *)NULL; | |
4803 | static int nfds; | |
4804 | static int totfds; /* The highest possible number of open files. */ | |
4805 | ||
4806 | char * | |
4807 | copy_fifo_list (sizep) | |
4808 | int *sizep; | |
4809 | { | |
4810 | char *ret; | |
4811 | ||
4812 | if (nfds == 0 || totfds == 0) | |
4813 | { | |
4814 | if (sizep) | |
4815 | *sizep = 0; | |
4816 | return (char *)NULL; | |
4817 | } | |
4818 | ||
4819 | if (sizep) | |
4820 | *sizep = totfds; | |
4821 | ret = (char *)xmalloc (totfds); | |
4822 | return (memcpy (ret, dev_fd_list, totfds)); | |
4823 | } | |
4824 | ||
4825 | static void | |
4826 | add_fifo_list (fd) | |
4827 | int fd; | |
4828 | { | |
4829 | if (dev_fd_list == 0 || fd >= totfds) | |
4830 | { | |
4831 | int ofds; | |
4832 | ||
4833 | ofds = totfds; | |
4834 | totfds = getdtablesize (); | |
4835 | if (totfds < 0 || totfds > 256) | |
4836 | totfds = 256; | |
4837 | if (fd >= totfds) | |
4838 | totfds = fd + 2; | |
4839 | ||
4840 | dev_fd_list = (char *)xrealloc (dev_fd_list, totfds); | |
4841 | memset (dev_fd_list + ofds, '\0', totfds - ofds); | |
4842 | } | |
4843 | ||
4844 | dev_fd_list[fd] = 1; | |
4845 | nfds++; | |
4846 | } | |
4847 | ||
4848 | int | |
4849 | fifos_pending () | |
4850 | { | |
4851 | return 0; /* used for cleanup; not needed with /dev/fd */ | |
4852 | } | |
4853 | ||
4854 | int | |
4855 | num_fifos () | |
4856 | { | |
4857 | return nfds; | |
4858 | } | |
4859 | ||
4860 | void | |
4861 | unlink_fifo (fd) | |
4862 | int fd; | |
4863 | { | |
4864 | if (dev_fd_list[fd]) | |
4865 | { | |
4866 | close (fd); | |
4867 | dev_fd_list[fd] = 0; | |
4868 | nfds--; | |
4869 | } | |
4870 | } | |
4871 | ||
4872 | void | |
4873 | unlink_fifo_list () | |
4874 | { | |
4875 | register int i; | |
4876 | ||
4877 | if (nfds == 0) | |
4878 | return; | |
4879 | ||
4880 | for (i = 0; nfds && i < totfds; i++) | |
4881 | unlink_fifo (i); | |
4882 | ||
4883 | nfds = 0; | |
4884 | } | |
4885 | ||
4886 | /* Take LIST, which is a snapshot copy of dev_fd_list from some point in | |
4887 | the past, and close all open fds in dev_fd_list that are not marked | |
4888 | as open in LIST. If LIST is NULL, close everything in dev_fd_list. | |
4889 | LSIZE is the number of elements in LIST, in case it's larger than | |
4890 | totfds (size of dev_fd_list). */ | |
4891 | void | |
4892 | close_new_fifos (list, lsize) | |
4893 | char *list; | |
4894 | int lsize; | |
4895 | { | |
4896 | int i; | |
4897 | ||
4898 | if (list == 0) | |
4899 | { | |
4900 | unlink_fifo_list (); | |
4901 | return; | |
4902 | } | |
4903 | ||
4904 | for (i = 0; i < lsize; i++) | |
4905 | if (list[i] == 0 && i < totfds && dev_fd_list[i]) | |
4906 | unlink_fifo (i); | |
4907 | ||
4908 | for (i = lsize; i < totfds; i++) | |
4909 | unlink_fifo (i); | |
4910 | } | |
4911 | ||
4912 | #if defined (NOTDEF) | |
4913 | print_dev_fd_list () | |
4914 | { | |
4915 | register int i; | |
4916 | ||
4917 | fprintf (stderr, "pid %ld: dev_fd_list:", (long)getpid ()); | |
4918 | fflush (stderr); | |
4919 | ||
4920 | for (i = 0; i < totfds; i++) | |
4921 | { | |
4922 | if (dev_fd_list[i]) | |
4923 | fprintf (stderr, " %d", i); | |
4924 | } | |
4925 | fprintf (stderr, "\n"); | |
4926 | } | |
4927 | #endif /* NOTDEF */ | |
4928 | ||
4929 | static char * | |
4930 | make_dev_fd_filename (fd) | |
4931 | int fd; | |
4932 | { | |
4933 | char *ret, intbuf[INT_STRLEN_BOUND (int) + 1], *p; | |
4934 | ||
4935 | ret = (char *)xmalloc (sizeof (DEV_FD_PREFIX) + 8); | |
4936 | ||
4937 | strcpy (ret, DEV_FD_PREFIX); | |
4938 | p = inttostr (fd, intbuf, sizeof (intbuf)); | |
4939 | strcpy (ret + sizeof (DEV_FD_PREFIX) - 1, p); | |
4940 | ||
4941 | add_fifo_list (fd); | |
4942 | return (ret); | |
4943 | } | |
4944 | ||
4945 | #endif /* HAVE_DEV_FD */ | |
4946 | ||
4947 | /* Return a filename that will open a connection to the process defined by | |
4948 | executing STRING. HAVE_DEV_FD, if defined, means open a pipe and return | |
4949 | a filename in /dev/fd corresponding to a descriptor that is one of the | |
4950 | ends of the pipe. If not defined, we use named pipes on systems that have | |
4951 | them. Systems without /dev/fd and named pipes are out of luck. | |
4952 | ||
4953 | OPEN_FOR_READ_IN_CHILD, if 1, means open the named pipe for reading or | |
4954 | use the read end of the pipe and dup that file descriptor to fd 0 in | |
4955 | the child. If OPEN_FOR_READ_IN_CHILD is 0, we open the named pipe for | |
4956 | writing or use the write end of the pipe in the child, and dup that | |
4957 | file descriptor to fd 1 in the child. The parent does the opposite. */ | |
4958 | ||
4959 | static char * | |
4960 | process_substitute (string, open_for_read_in_child) | |
4961 | char *string; | |
4962 | int open_for_read_in_child; | |
4963 | { | |
4964 | char *pathname; | |
4965 | int fd, result; | |
4966 | pid_t old_pid, pid; | |
4967 | #if defined (HAVE_DEV_FD) | |
4968 | int parent_pipe_fd, child_pipe_fd; | |
4969 | int fildes[2]; | |
4970 | #endif /* HAVE_DEV_FD */ | |
4971 | #if defined (JOB_CONTROL) | |
4972 | pid_t old_pipeline_pgrp; | |
4973 | #endif | |
4974 | ||
4975 | if (!string || !*string || wordexp_only) | |
4976 | return ((char *)NULL); | |
4977 | ||
4978 | #if !defined (HAVE_DEV_FD) | |
4979 | pathname = make_named_pipe (); | |
4980 | #else /* HAVE_DEV_FD */ | |
4981 | if (pipe (fildes) < 0) | |
4982 | { | |
4983 | sys_error (_("cannot make pipe for process substitution")); | |
4984 | return ((char *)NULL); | |
4985 | } | |
4986 | /* If OPEN_FOR_READ_IN_CHILD == 1, we want to use the write end of | |
4987 | the pipe in the parent, otherwise the read end. */ | |
4988 | parent_pipe_fd = fildes[open_for_read_in_child]; | |
4989 | child_pipe_fd = fildes[1 - open_for_read_in_child]; | |
4990 | /* Move the parent end of the pipe to some high file descriptor, to | |
4991 | avoid clashes with FDs used by the script. */ | |
4992 | parent_pipe_fd = move_to_high_fd (parent_pipe_fd, 1, 64); | |
4993 | ||
4994 | pathname = make_dev_fd_filename (parent_pipe_fd); | |
4995 | #endif /* HAVE_DEV_FD */ | |
4996 | ||
4997 | if (pathname == 0) | |
4998 | { | |
4999 | sys_error (_("cannot make pipe for process substitution")); | |
5000 | return ((char *)NULL); | |
5001 | } | |
5002 | ||
5003 | old_pid = last_made_pid; | |
5004 | ||
5005 | #if defined (JOB_CONTROL) | |
5006 | old_pipeline_pgrp = pipeline_pgrp; | |
5007 | pipeline_pgrp = shell_pgrp; | |
5008 | save_pipeline (1); | |
5009 | #endif /* JOB_CONTROL */ | |
5010 | ||
5011 | pid = make_child ((char *)NULL, 1); | |
5012 | if (pid == 0) | |
5013 | { | |
5014 | reset_terminating_signals (); /* XXX */ | |
5015 | free_pushed_string_input (); | |
5016 | /* Cancel traps, in trap.c. */ | |
5017 | restore_original_signals (); /* XXX - what about special builtins? bash-4.2 */ | |
5018 | setup_async_signals (); | |
5019 | subshell_environment |= SUBSHELL_COMSUB|SUBSHELL_PROCSUB; | |
5020 | } | |
5021 | ||
5022 | #if defined (JOB_CONTROL) | |
5023 | set_sigchld_handler (); | |
5024 | stop_making_children (); | |
5025 | /* XXX - should we only do this in the parent? (as in command subst) */ | |
5026 | pipeline_pgrp = old_pipeline_pgrp; | |
5027 | #endif /* JOB_CONTROL */ | |
5028 | ||
5029 | if (pid < 0) | |
5030 | { | |
5031 | sys_error (_("cannot make child for process substitution")); | |
5032 | free (pathname); | |
5033 | #if defined (HAVE_DEV_FD) | |
5034 | close (parent_pipe_fd); | |
5035 | close (child_pipe_fd); | |
5036 | #endif /* HAVE_DEV_FD */ | |
5037 | return ((char *)NULL); | |
5038 | } | |
5039 | ||
5040 | if (pid > 0) | |
5041 | { | |
5042 | #if defined (JOB_CONTROL) | |
5043 | restore_pipeline (1); | |
5044 | #endif | |
5045 | ||
5046 | #if !defined (HAVE_DEV_FD) | |
5047 | fifo_list[nfifo-1].proc = pid; | |
5048 | #endif | |
5049 | ||
5050 | last_made_pid = old_pid; | |
5051 | ||
5052 | #if defined (JOB_CONTROL) && defined (PGRP_PIPE) | |
5053 | close_pgrp_pipe (); | |
5054 | #endif /* JOB_CONTROL && PGRP_PIPE */ | |
5055 | ||
5056 | #if defined (HAVE_DEV_FD) | |
5057 | close (child_pipe_fd); | |
5058 | #endif /* HAVE_DEV_FD */ | |
5059 | ||
5060 | return (pathname); | |
5061 | } | |
5062 | ||
5063 | set_sigint_handler (); | |
5064 | ||
5065 | #if defined (JOB_CONTROL) | |
5066 | set_job_control (0); | |
5067 | #endif /* JOB_CONTROL */ | |
5068 | ||
5069 | #if !defined (HAVE_DEV_FD) | |
5070 | /* Open the named pipe in the child. */ | |
5071 | fd = open (pathname, open_for_read_in_child ? O_RDONLY|O_NONBLOCK : O_WRONLY); | |
5072 | if (fd < 0) | |
5073 | { | |
5074 | /* Two separate strings for ease of translation. */ | |
5075 | if (open_for_read_in_child) | |
5076 | sys_error (_("cannot open named pipe %s for reading"), pathname); | |
5077 | else | |
5078 | sys_error (_("cannot open named pipe %s for writing"), pathname); | |
5079 | ||
5080 | exit (127); | |
5081 | } | |
5082 | if (open_for_read_in_child) | |
5083 | { | |
5084 | if (sh_unset_nodelay_mode (fd) < 0) | |
5085 | { | |
5086 | sys_error (_("cannot reset nodelay mode for fd %d"), fd); | |
5087 | exit (127); | |
5088 | } | |
5089 | } | |
5090 | #else /* HAVE_DEV_FD */ | |
5091 | fd = child_pipe_fd; | |
5092 | #endif /* HAVE_DEV_FD */ | |
5093 | ||
5094 | if (dup2 (fd, open_for_read_in_child ? 0 : 1) < 0) | |
5095 | { | |
5096 | sys_error (_("cannot duplicate named pipe %s as fd %d"), pathname, | |
5097 | open_for_read_in_child ? 0 : 1); | |
5098 | exit (127); | |
5099 | } | |
5100 | ||
5101 | if (fd != (open_for_read_in_child ? 0 : 1)) | |
5102 | close (fd); | |
5103 | ||
5104 | /* Need to close any files that this process has open to pipes inherited | |
5105 | from its parent. */ | |
5106 | if (current_fds_to_close) | |
5107 | { | |
5108 | close_fd_bitmap (current_fds_to_close); | |
5109 | current_fds_to_close = (struct fd_bitmap *)NULL; | |
5110 | } | |
5111 | ||
5112 | #if defined (HAVE_DEV_FD) | |
5113 | /* Make sure we close the parent's end of the pipe and clear the slot | |
5114 | in the fd list so it is not closed later, if reallocated by, for | |
5115 | instance, pipe(2). */ | |
5116 | close (parent_pipe_fd); | |
5117 | dev_fd_list[parent_pipe_fd] = 0; | |
5118 | #endif /* HAVE_DEV_FD */ | |
5119 | ||
5120 | result = parse_and_execute (string, "process substitution", (SEVAL_NONINT|SEVAL_NOHIST)); | |
5121 | ||
5122 | #if !defined (HAVE_DEV_FD) | |
5123 | /* Make sure we close the named pipe in the child before we exit. */ | |
5124 | close (open_for_read_in_child ? 0 : 1); | |
5125 | #endif /* !HAVE_DEV_FD */ | |
5126 | ||
5127 | exit (result); | |
5128 | /*NOTREACHED*/ | |
5129 | } | |
5130 | #endif /* PROCESS_SUBSTITUTION */ | |
5131 | ||
5132 | /***********************************/ | |
5133 | /* */ | |
5134 | /* Command Substitution */ | |
5135 | /* */ | |
5136 | /***********************************/ | |
5137 | ||
5138 | static char * | |
5139 | read_comsub (fd, quoted, rflag) | |
5140 | int fd, quoted; | |
5141 | int *rflag; | |
5142 | { | |
5143 | char *istring, buf[128], *bufp, *s; | |
5144 | int istring_index, istring_size, c, tflag, skip_ctlesc, skip_ctlnul; | |
5145 | ssize_t bufn; | |
5146 | ||
5147 | istring = (char *)NULL; | |
5148 | istring_index = istring_size = bufn = tflag = 0; | |
5149 | ||
5150 | for (skip_ctlesc = skip_ctlnul = 0, s = ifs_value; s && *s; s++) | |
5151 | skip_ctlesc |= *s == CTLESC, skip_ctlnul |= *s == CTLNUL; | |
5152 | ||
5153 | /* Read the output of the command through the pipe. This may need to be | |
5154 | changed to understand multibyte characters in the future. */ | |
5155 | while (1) | |
5156 | { | |
5157 | if (fd < 0) | |
5158 | break; | |
5159 | if (--bufn <= 0) | |
5160 | { | |
5161 | bufn = zread (fd, buf, sizeof (buf)); | |
5162 | if (bufn <= 0) | |
5163 | break; | |
5164 | bufp = buf; | |
5165 | } | |
5166 | c = *bufp++; | |
5167 | ||
5168 | if (c == 0) | |
5169 | { | |
5170 | #if 0 | |
5171 | internal_warning ("read_comsub: ignored null byte in input"); | |
5172 | #endif | |
5173 | continue; | |
5174 | } | |
5175 | ||
5176 | /* Add the character to ISTRING, possibly after resizing it. */ | |
5177 | RESIZE_MALLOCED_BUFFER (istring, istring_index, 2, istring_size, DEFAULT_ARRAY_SIZE); | |
5178 | ||
5179 | /* This is essentially quote_string inline */ | |
5180 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) /* || c == CTLESC || c == CTLNUL */) | |
5181 | istring[istring_index++] = CTLESC; | |
5182 | /* Escape CTLESC and CTLNUL in the output to protect those characters | |
5183 | from the rest of the word expansions (word splitting and globbing.) | |
5184 | This is essentially quote_escapes inline. */ | |
5185 | else if (skip_ctlesc == 0 && c == CTLESC) | |
5186 | { | |
5187 | tflag |= W_HASCTLESC; | |
5188 | istring[istring_index++] = CTLESC; | |
5189 | } | |
5190 | else if ((skip_ctlnul == 0 && c == CTLNUL) || (c == ' ' && (ifs_value && *ifs_value == 0))) | |
5191 | istring[istring_index++] = CTLESC; | |
5192 | ||
5193 | istring[istring_index++] = c; | |
5194 | ||
5195 | #if 0 | |
5196 | #if defined (__CYGWIN__) | |
5197 | if (c == '\n' && istring_index > 1 && istring[istring_index - 2] == '\r') | |
5198 | { | |
5199 | istring_index--; | |
5200 | istring[istring_index - 1] = '\n'; | |
5201 | } | |
5202 | #endif | |
5203 | #endif | |
5204 | } | |
5205 | ||
5206 | if (istring) | |
5207 | istring[istring_index] = '\0'; | |
5208 | ||
5209 | /* If we read no output, just return now and save ourselves some | |
5210 | trouble. */ | |
5211 | if (istring_index == 0) | |
5212 | { | |
5213 | FREE (istring); | |
5214 | if (rflag) | |
5215 | *rflag = tflag; | |
5216 | return (char *)NULL; | |
5217 | } | |
5218 | ||
5219 | /* Strip trailing newlines from the output of the command. */ | |
5220 | if (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
5221 | { | |
5222 | while (istring_index > 0) | |
5223 | { | |
5224 | if (istring[istring_index - 1] == '\n') | |
5225 | { | |
5226 | --istring_index; | |
5227 | ||
5228 | /* If the newline was quoted, remove the quoting char. */ | |
5229 | if (istring[istring_index - 1] == CTLESC) | |
5230 | --istring_index; | |
5231 | } | |
5232 | else | |
5233 | break; | |
5234 | } | |
5235 | istring[istring_index] = '\0'; | |
5236 | } | |
5237 | else | |
5238 | strip_trailing (istring, istring_index - 1, 1); | |
5239 | ||
5240 | if (rflag) | |
5241 | *rflag = tflag; | |
5242 | return istring; | |
5243 | } | |
5244 | ||
5245 | /* Perform command substitution on STRING. This returns a WORD_DESC * with the | |
5246 | contained string possibly quoted. */ | |
5247 | WORD_DESC * | |
5248 | command_substitute (string, quoted) | |
5249 | char *string; | |
5250 | int quoted; | |
5251 | { | |
5252 | pid_t pid, old_pid, old_pipeline_pgrp, old_async_pid; | |
5253 | char *istring; | |
5254 | int result, fildes[2], function_value, pflags, rc, tflag; | |
5255 | WORD_DESC *ret; | |
5256 | ||
5257 | istring = (char *)NULL; | |
5258 | ||
5259 | /* Don't fork () if there is no need to. In the case of no command to | |
5260 | run, just return NULL. */ | |
5261 | if (!string || !*string || (string[0] == '\n' && !string[1])) | |
5262 | return ((WORD_DESC *)NULL); | |
5263 | ||
5264 | if (wordexp_only && read_but_dont_execute) | |
5265 | { | |
5266 | last_command_exit_value = EX_WEXPCOMSUB; | |
5267 | jump_to_top_level (EXITPROG); | |
5268 | } | |
5269 | ||
5270 | /* We're making the assumption here that the command substitution will | |
5271 | eventually run a command from the file system. Since we'll run | |
5272 | maybe_make_export_env in this subshell before executing that command, | |
5273 | the parent shell and any other shells it starts will have to remake | |
5274 | the environment. If we make it before we fork, other shells won't | |
5275 | have to. Don't bother if we have any temporary variable assignments, | |
5276 | though, because the export environment will be remade after this | |
5277 | command completes anyway, but do it if all the words to be expanded | |
5278 | are variable assignments. */ | |
5279 | if (subst_assign_varlist == 0 || garglist == 0) | |
5280 | maybe_make_export_env (); /* XXX */ | |
5281 | ||
5282 | /* Flags to pass to parse_and_execute() */ | |
5283 | pflags = (interactive && sourcelevel == 0) ? SEVAL_RESETLINE : 0; | |
5284 | ||
5285 | /* Pipe the output of executing STRING into the current shell. */ | |
5286 | if (pipe (fildes) < 0) | |
5287 | { | |
5288 | sys_error (_("cannot make pipe for command substitution")); | |
5289 | goto error_exit; | |
5290 | } | |
5291 | ||
5292 | old_pid = last_made_pid; | |
5293 | #if defined (JOB_CONTROL) | |
5294 | old_pipeline_pgrp = pipeline_pgrp; | |
5295 | /* Don't reset the pipeline pgrp if we're already a subshell in a pipeline. */ | |
5296 | if ((subshell_environment & SUBSHELL_PIPE) == 0) | |
5297 | pipeline_pgrp = shell_pgrp; | |
5298 | cleanup_the_pipeline (); | |
5299 | #endif /* JOB_CONTROL */ | |
5300 | ||
5301 | old_async_pid = last_asynchronous_pid; | |
5302 | pid = make_child ((char *)NULL, subshell_environment&SUBSHELL_ASYNC); | |
5303 | last_asynchronous_pid = old_async_pid; | |
5304 | ||
5305 | if (pid == 0) | |
5306 | { | |
5307 | /* Reset the signal handlers in the child, but don't free the | |
5308 | trap strings. Set a flag noting that we have to free the | |
5309 | trap strings if we run trap to change a signal disposition. */ | |
5310 | reset_signal_handlers (); | |
5311 | subshell_environment |= SUBSHELL_RESETTRAP; | |
5312 | } | |
5313 | ||
5314 | #if defined (JOB_CONTROL) | |
5315 | /* XXX DO THIS ONLY IN PARENT ? XXX */ | |
5316 | set_sigchld_handler (); | |
5317 | stop_making_children (); | |
5318 | if (pid != 0) | |
5319 | pipeline_pgrp = old_pipeline_pgrp; | |
5320 | #else | |
5321 | stop_making_children (); | |
5322 | #endif /* JOB_CONTROL */ | |
5323 | ||
5324 | if (pid < 0) | |
5325 | { | |
5326 | sys_error (_("cannot make child for command substitution")); | |
5327 | error_exit: | |
5328 | ||
5329 | FREE (istring); | |
5330 | close (fildes[0]); | |
5331 | close (fildes[1]); | |
5332 | return ((WORD_DESC *)NULL); | |
5333 | } | |
5334 | ||
5335 | if (pid == 0) | |
5336 | { | |
5337 | set_sigint_handler (); /* XXX */ | |
5338 | ||
5339 | free_pushed_string_input (); | |
5340 | ||
5341 | if (dup2 (fildes[1], 1) < 0) | |
5342 | { | |
5343 | sys_error (_("command_substitute: cannot duplicate pipe as fd 1")); | |
5344 | exit (EXECUTION_FAILURE); | |
5345 | } | |
5346 | ||
5347 | /* If standard output is closed in the parent shell | |
5348 | (such as after `exec >&-'), file descriptor 1 will be | |
5349 | the lowest available file descriptor, and end up in | |
5350 | fildes[0]. This can happen for stdin and stderr as well, | |
5351 | but stdout is more important -- it will cause no output | |
5352 | to be generated from this command. */ | |
5353 | if ((fildes[1] != fileno (stdin)) && | |
5354 | (fildes[1] != fileno (stdout)) && | |
5355 | (fildes[1] != fileno (stderr))) | |
5356 | close (fildes[1]); | |
5357 | ||
5358 | if ((fildes[0] != fileno (stdin)) && | |
5359 | (fildes[0] != fileno (stdout)) && | |
5360 | (fildes[0] != fileno (stderr))) | |
5361 | close (fildes[0]); | |
5362 | ||
5363 | #ifdef __CYGWIN__ | |
5364 | /* Let stdio know the fd may have changed from text to binary mode, and | |
5365 | make sure to preserve stdout line buffering. */ | |
5366 | freopen (NULL, "w", stdout); | |
5367 | sh_setlinebuf (stdout); | |
5368 | #endif /* __CYGWIN__ */ | |
5369 | ||
5370 | /* The currently executing shell is not interactive. */ | |
5371 | interactive = 0; | |
5372 | ||
5373 | /* This is a subshell environment. */ | |
5374 | subshell_environment |= SUBSHELL_COMSUB; | |
5375 | ||
5376 | /* When not in POSIX mode, command substitution does not inherit | |
5377 | the -e flag. */ | |
5378 | if (posixly_correct == 0) | |
5379 | exit_immediately_on_error = 0; | |
5380 | ||
5381 | remove_quoted_escapes (string); | |
5382 | ||
5383 | startup_state = 2; /* see if we can avoid a fork */ | |
5384 | /* Give command substitution a place to jump back to on failure, | |
5385 | so we don't go back up to main (). */ | |
5386 | result = setjmp (top_level); | |
5387 | ||
5388 | /* If we're running a command substitution inside a shell function, | |
5389 | trap `return' so we don't return from the function in the subshell | |
5390 | and go off to never-never land. */ | |
5391 | if (result == 0 && return_catch_flag) | |
5392 | function_value = setjmp (return_catch); | |
5393 | else | |
5394 | function_value = 0; | |
5395 | ||
5396 | if (result == ERREXIT) | |
5397 | rc = last_command_exit_value; | |
5398 | else if (result == EXITPROG) | |
5399 | rc = last_command_exit_value; | |
5400 | else if (result) | |
5401 | rc = EXECUTION_FAILURE; | |
5402 | else if (function_value) | |
5403 | rc = return_catch_value; | |
5404 | else | |
5405 | { | |
5406 | subshell_level++; | |
5407 | rc = parse_and_execute (string, "command substitution", pflags|SEVAL_NOHIST); | |
5408 | subshell_level--; | |
5409 | } | |
5410 | ||
5411 | last_command_exit_value = rc; | |
5412 | rc = run_exit_trap (); | |
5413 | #if defined (PROCESS_SUBSTITUTION) | |
5414 | unlink_fifo_list (); | |
5415 | #endif | |
5416 | exit (rc); | |
5417 | } | |
5418 | else | |
5419 | { | |
5420 | #if defined (JOB_CONTROL) && defined (PGRP_PIPE) | |
5421 | close_pgrp_pipe (); | |
5422 | #endif /* JOB_CONTROL && PGRP_PIPE */ | |
5423 | ||
5424 | close (fildes[1]); | |
5425 | ||
5426 | tflag = 0; | |
5427 | istring = read_comsub (fildes[0], quoted, &tflag); | |
5428 | ||
5429 | close (fildes[0]); | |
5430 | ||
5431 | current_command_subst_pid = pid; | |
5432 | last_command_exit_value = wait_for (pid); | |
5433 | last_command_subst_pid = pid; | |
5434 | last_made_pid = old_pid; | |
5435 | ||
5436 | #if defined (JOB_CONTROL) | |
5437 | /* If last_command_exit_value > 128, then the substituted command | |
5438 | was terminated by a signal. If that signal was SIGINT, then send | |
5439 | SIGINT to ourselves. This will break out of loops, for instance. */ | |
5440 | if (last_command_exit_value == (128 + SIGINT) && last_command_exit_signal == SIGINT) | |
5441 | kill (getpid (), SIGINT); | |
5442 | ||
5443 | /* wait_for gives the terminal back to shell_pgrp. If some other | |
5444 | process group should have it, give it away to that group here. | |
5445 | pipeline_pgrp is non-zero only while we are constructing a | |
5446 | pipline, so what we are concerned about is whether or not that | |
5447 | pipeline was started in the background. A pipeline started in | |
5448 | the background should never get the tty back here. */ | |
5449 | if (interactive && pipeline_pgrp != (pid_t)0 && (subshell_environment & SUBSHELL_ASYNC) == 0) | |
5450 | give_terminal_to (pipeline_pgrp, 0); | |
5451 | #endif /* JOB_CONTROL */ | |
5452 | ||
5453 | ret = alloc_word_desc (); | |
5454 | ret->word = istring; | |
5455 | ret->flags = tflag; | |
5456 | ||
5457 | return ret; | |
5458 | } | |
5459 | } | |
5460 | ||
5461 | /******************************************************** | |
5462 | * * | |
5463 | * Utility functions for parameter expansion * | |
5464 | * * | |
5465 | ********************************************************/ | |
5466 | ||
5467 | #if defined (ARRAY_VARS) | |
5468 | ||
5469 | static arrayind_t | |
5470 | array_length_reference (s) | |
5471 | char *s; | |
5472 | { | |
5473 | int len; | |
5474 | arrayind_t ind; | |
5475 | char *akey; | |
5476 | char *t, c; | |
5477 | ARRAY *array; | |
5478 | HASH_TABLE *h; | |
5479 | SHELL_VAR *var; | |
5480 | ||
5481 | var = array_variable_part (s, &t, &len); | |
5482 | ||
5483 | /* If unbound variables should generate an error, report one and return | |
5484 | failure. */ | |
5485 | if ((var == 0 || (assoc_p (var) == 0 && array_p (var) == 0)) && unbound_vars_is_error) | |
5486 | { | |
5487 | c = *--t; | |
5488 | *t = '\0'; | |
5489 | last_command_exit_value = EXECUTION_FAILURE; | |
5490 | err_unboundvar (s); | |
5491 | *t = c; | |
5492 | return (-1); | |
5493 | } | |
5494 | else if (var == 0) | |
5495 | return 0; | |
5496 | ||
5497 | /* We support a couple of expansions for variables that are not arrays. | |
5498 | We'll return the length of the value for v[0], and 1 for v[@] or | |
5499 | v[*]. Return 0 for everything else. */ | |
5500 | ||
5501 | array = array_p (var) ? array_cell (var) : (ARRAY *)NULL; | |
5502 | h = assoc_p (var) ? assoc_cell (var) : (HASH_TABLE *)NULL; | |
5503 | ||
5504 | if (ALL_ELEMENT_SUB (t[0]) && t[1] == ']') | |
5505 | { | |
5506 | if (assoc_p (var)) | |
5507 | return (h ? assoc_num_elements (h) : 0); | |
5508 | else if (array_p (var)) | |
5509 | return (array ? array_num_elements (array) : 0); | |
5510 | else | |
5511 | return (var_isset (var) ? 1 : 0); | |
5512 | } | |
5513 | ||
5514 | if (assoc_p (var)) | |
5515 | { | |
5516 | t[len - 1] = '\0'; | |
5517 | akey = expand_assignment_string_to_string (t, 0); /* [ */ | |
5518 | t[len - 1] = ']'; | |
5519 | if (akey == 0 || *akey == 0) | |
5520 | { | |
5521 | err_badarraysub (t); | |
5522 | return (-1); | |
5523 | } | |
5524 | t = assoc_reference (assoc_cell (var), akey); | |
5525 | } | |
5526 | else | |
5527 | { | |
5528 | ind = array_expand_index (t, len); | |
5529 | if (ind < 0) | |
5530 | { | |
5531 | err_badarraysub (t); | |
5532 | return (-1); | |
5533 | } | |
5534 | if (array_p (var)) | |
5535 | t = array_reference (array, ind); | |
5536 | else | |
5537 | t = (ind == 0) ? value_cell (var) : (char *)NULL; | |
5538 | } | |
5539 | ||
5540 | len = MB_STRLEN (t); | |
5541 | return (len); | |
5542 | } | |
5543 | #endif /* ARRAY_VARS */ | |
5544 | ||
5545 | static int | |
5546 | valid_brace_expansion_word (name, var_is_special) | |
5547 | char *name; | |
5548 | int var_is_special; | |
5549 | { | |
5550 | if (DIGIT (*name) && all_digits (name)) | |
5551 | return 1; | |
5552 | else if (var_is_special) | |
5553 | return 1; | |
5554 | #if defined (ARRAY_VARS) | |
5555 | else if (valid_array_reference (name)) | |
5556 | return 1; | |
5557 | #endif /* ARRAY_VARS */ | |
5558 | else if (legal_identifier (name)) | |
5559 | return 1; | |
5560 | else | |
5561 | return 0; | |
5562 | } | |
5563 | ||
5564 | static int | |
5565 | chk_atstar (name, quoted, quoted_dollar_atp, contains_dollar_at) | |
5566 | char *name; | |
5567 | int quoted; | |
5568 | int *quoted_dollar_atp, *contains_dollar_at; | |
5569 | { | |
5570 | char *temp1; | |
5571 | ||
5572 | if (name == 0) | |
5573 | { | |
5574 | if (quoted_dollar_atp) | |
5575 | *quoted_dollar_atp = 0; | |
5576 | if (contains_dollar_at) | |
5577 | *contains_dollar_at = 0; | |
5578 | return 0; | |
5579 | } | |
5580 | ||
5581 | /* check for $@ and $* */ | |
5582 | if (name[0] == '@' && name[1] == 0) | |
5583 | { | |
5584 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && quoted_dollar_atp) | |
5585 | *quoted_dollar_atp = 1; | |
5586 | if (contains_dollar_at) | |
5587 | *contains_dollar_at = 1; | |
5588 | return 1; | |
5589 | } | |
5590 | else if (name[0] == '*' && name[1] == '\0' && quoted == 0) | |
5591 | { | |
5592 | if (contains_dollar_at) | |
5593 | *contains_dollar_at = 1; | |
5594 | return 1; | |
5595 | } | |
5596 | ||
5597 | /* Now check for ${array[@]} and ${array[*]} */ | |
5598 | #if defined (ARRAY_VARS) | |
5599 | else if (valid_array_reference (name)) | |
5600 | { | |
5601 | temp1 = mbschr (name, '['); | |
5602 | if (temp1 && temp1[1] == '@' && temp1[2] == ']') | |
5603 | { | |
5604 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && quoted_dollar_atp) | |
5605 | *quoted_dollar_atp = 1; | |
5606 | if (contains_dollar_at) | |
5607 | *contains_dollar_at = 1; | |
5608 | return 1; | |
5609 | } /* [ */ | |
5610 | /* ${array[*]}, when unquoted, should be treated like ${array[@]}, | |
5611 | which should result in separate words even when IFS is unset. */ | |
5612 | if (temp1 && temp1[1] == '*' && temp1[2] == ']' && quoted == 0) | |
5613 | { | |
5614 | if (contains_dollar_at) | |
5615 | *contains_dollar_at = 1; | |
5616 | return 1; | |
5617 | } | |
5618 | } | |
5619 | #endif | |
5620 | return 0; | |
5621 | } | |
5622 | ||
5623 | /* Parameter expand NAME, and return a new string which is the expansion, | |
5624 | or NULL if there was no expansion. | |
5625 | VAR_IS_SPECIAL is non-zero if NAME is one of the special variables in | |
5626 | the shell, e.g., "@", "$", "*", etc. QUOTED, if non-zero, means that | |
5627 | NAME was found inside of a double-quoted expression. */ | |
5628 | static WORD_DESC * | |
5629 | parameter_brace_expand_word (name, var_is_special, quoted, pflags, indp) | |
5630 | char *name; | |
5631 | int var_is_special, quoted, pflags; | |
5632 | arrayind_t *indp; | |
5633 | { | |
5634 | WORD_DESC *ret; | |
5635 | char *temp, *tt; | |
5636 | intmax_t arg_index; | |
5637 | SHELL_VAR *var; | |
5638 | int atype, rflags; | |
5639 | arrayind_t ind; | |
5640 | ||
5641 | ret = 0; | |
5642 | temp = 0; | |
5643 | rflags = 0; | |
5644 | ||
5645 | if (indp) | |
5646 | *indp = INTMAX_MIN; | |
5647 | ||
5648 | /* Handle multiple digit arguments, as in ${11}. */ | |
5649 | if (legal_number (name, &arg_index)) | |
5650 | { | |
5651 | tt = get_dollar_var_value (arg_index); | |
5652 | if (tt) | |
5653 | temp = (*tt && (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT))) | |
5654 | ? quote_string (tt) | |
5655 | : quote_escapes (tt); | |
5656 | else | |
5657 | temp = (char *)NULL; | |
5658 | FREE (tt); | |
5659 | } | |
5660 | else if (var_is_special) /* ${@} */ | |
5661 | { | |
5662 | int sindex; | |
5663 | tt = (char *)xmalloc (2 + strlen (name)); | |
5664 | tt[sindex = 0] = '$'; | |
5665 | strcpy (tt + 1, name); | |
5666 | ||
5667 | ret = param_expand (tt, &sindex, quoted, (int *)NULL, (int *)NULL, | |
5668 | (int *)NULL, (int *)NULL, pflags); | |
5669 | free (tt); | |
5670 | } | |
5671 | #if defined (ARRAY_VARS) | |
5672 | else if (valid_array_reference (name)) | |
5673 | { | |
5674 | temp = array_value (name, quoted, 0, &atype, &ind); | |
5675 | if (atype == 0 && temp) | |
5676 | { | |
5677 | temp = (*temp && (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT))) | |
5678 | ? quote_string (temp) | |
5679 | : quote_escapes (temp); | |
5680 | rflags |= W_ARRAYIND; | |
5681 | if (indp) | |
5682 | *indp = ind; | |
5683 | } | |
5684 | else if (atype == 1 && temp && QUOTED_NULL (temp) && (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT))) | |
5685 | rflags |= W_HASQUOTEDNULL; | |
5686 | } | |
5687 | #endif | |
5688 | else if (var = find_variable (name)) | |
5689 | { | |
5690 | if (var_isset (var) && invisible_p (var) == 0) | |
5691 | { | |
5692 | #if defined (ARRAY_VARS) | |
5693 | if (assoc_p (var)) | |
5694 | temp = assoc_reference (assoc_cell (var), "0"); | |
5695 | else if (array_p (var)) | |
5696 | temp = array_reference (array_cell (var), 0); | |
5697 | else | |
5698 | temp = value_cell (var); | |
5699 | #else | |
5700 | temp = value_cell (var); | |
5701 | #endif | |
5702 | ||
5703 | if (temp) | |
5704 | temp = (*temp && (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT))) | |
5705 | ? quote_string (temp) | |
5706 | : quote_escapes (temp); | |
5707 | } | |
5708 | else | |
5709 | temp = (char *)NULL; | |
5710 | } | |
5711 | else | |
5712 | temp = (char *)NULL; | |
5713 | ||
5714 | if (ret == 0) | |
5715 | { | |
5716 | ret = alloc_word_desc (); | |
5717 | ret->word = temp; | |
5718 | ret->flags |= rflags; | |
5719 | } | |
5720 | return ret; | |
5721 | } | |
5722 | ||
5723 | /* Expand an indirect reference to a variable: ${!NAME} expands to the | |
5724 | value of the variable whose name is the value of NAME. */ | |
5725 | static WORD_DESC * | |
5726 | parameter_brace_expand_indir (name, var_is_special, quoted, quoted_dollar_atp, contains_dollar_at) | |
5727 | char *name; | |
5728 | int var_is_special, quoted; | |
5729 | int *quoted_dollar_atp, *contains_dollar_at; | |
5730 | { | |
5731 | char *temp, *t; | |
5732 | WORD_DESC *w; | |
5733 | ||
5734 | w = parameter_brace_expand_word (name, var_is_special, quoted, PF_IGNUNBOUND, 0); | |
5735 | t = w->word; | |
5736 | /* Have to dequote here if necessary */ | |
5737 | if (t) | |
5738 | { | |
5739 | temp = (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) | |
5740 | ? dequote_string (t) | |
5741 | : dequote_escapes (t); | |
5742 | free (t); | |
5743 | t = temp; | |
5744 | } | |
5745 | dispose_word_desc (w); | |
5746 | ||
5747 | chk_atstar (t, quoted, quoted_dollar_atp, contains_dollar_at); | |
5748 | if (t == 0) | |
5749 | return (WORD_DESC *)NULL; | |
5750 | ||
5751 | w = parameter_brace_expand_word (t, SPECIAL_VAR(t, 0), quoted, 0, 0); | |
5752 | free (t); | |
5753 | ||
5754 | return w; | |
5755 | } | |
5756 | ||
5757 | /* Expand the right side of a parameter expansion of the form ${NAMEcVALUE}, | |
5758 | depending on the value of C, the separating character. C can be one of | |
5759 | "-", "+", or "=". QUOTED is true if the entire brace expression occurs | |
5760 | between double quotes. */ | |
5761 | static WORD_DESC * | |
5762 | parameter_brace_expand_rhs (name, value, c, quoted, qdollaratp, hasdollarat) | |
5763 | char *name, *value; | |
5764 | int c, quoted, *qdollaratp, *hasdollarat; | |
5765 | { | |
5766 | WORD_DESC *w; | |
5767 | WORD_LIST *l; | |
5768 | char *t, *t1, *temp; | |
5769 | int hasdol; | |
5770 | ||
5771 | /* If the entire expression is between double quotes, we want to treat | |
5772 | the value as a double-quoted string, with the exception that we strip | |
5773 | embedded unescaped double quotes (for sh backwards compatibility). */ | |
5774 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && *value) | |
5775 | { | |
5776 | hasdol = 0; | |
5777 | temp = string_extract_double_quoted (value, &hasdol, 1); | |
5778 | } | |
5779 | else | |
5780 | temp = value; | |
5781 | ||
5782 | w = alloc_word_desc (); | |
5783 | hasdol = 0; | |
5784 | /* XXX was 0 not quoted */ | |
5785 | l = *temp ? expand_string_for_rhs (temp, quoted, &hasdol, (int *)NULL) | |
5786 | : (WORD_LIST *)0; | |
5787 | if (hasdollarat) | |
5788 | *hasdollarat = hasdol || (l && l->next); | |
5789 | if (temp != value) | |
5790 | free (temp); | |
5791 | if (l) | |
5792 | { | |
5793 | /* The expansion of TEMP returned something. We need to treat things | |
5794 | slightly differently if HASDOL is non-zero. If we have "$@", the | |
5795 | individual words have already been quoted. We need to turn them | |
5796 | into a string with the words separated by the first character of | |
5797 | $IFS without any additional quoting, so string_list_dollar_at won't | |
5798 | do the right thing. We use string_list_dollar_star instead. */ | |
5799 | temp = (hasdol || l->next) ? string_list_dollar_star (l) : string_list (l); | |
5800 | ||
5801 | /* If l->next is not null, we know that TEMP contained "$@", since that | |
5802 | is the only expansion that creates more than one word. */ | |
5803 | if (qdollaratp && ((hasdol && quoted) || l->next)) | |
5804 | *qdollaratp = 1; | |
5805 | dispose_words (l); | |
5806 | } | |
5807 | else if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && hasdol) | |
5808 | { | |
5809 | /* The brace expansion occurred between double quotes and there was | |
5810 | a $@ in TEMP. It does not matter if the $@ is quoted, as long as | |
5811 | it does not expand to anything. In this case, we want to return | |
5812 | a quoted empty string. */ | |
5813 | temp = make_quoted_char ('\0'); | |
5814 | w->flags |= W_HASQUOTEDNULL; | |
5815 | } | |
5816 | else | |
5817 | temp = (char *)NULL; | |
5818 | ||
5819 | if (c == '-' || c == '+') | |
5820 | { | |
5821 | w->word = temp; | |
5822 | return w; | |
5823 | } | |
5824 | ||
5825 | /* c == '=' */ | |
5826 | t = temp ? savestring (temp) : savestring (""); | |
5827 | t1 = dequote_string (t); | |
5828 | free (t); | |
5829 | #if defined (ARRAY_VARS) | |
5830 | if (valid_array_reference (name)) | |
5831 | assign_array_element (name, t1, 0); | |
5832 | else | |
5833 | #endif /* ARRAY_VARS */ | |
5834 | bind_variable (name, t1, 0); | |
5835 | ||
5836 | /* From Posix group discussion Feb-March 2010. Issue 7 0000221 */ | |
5837 | free (temp); | |
5838 | ||
5839 | w->word = t1; | |
5840 | return w; | |
5841 | } | |
5842 | ||
5843 | /* Deal with the right hand side of a ${name:?value} expansion in the case | |
5844 | that NAME is null or not set. If VALUE is non-null it is expanded and | |
5845 | used as the error message to print, otherwise a standard message is | |
5846 | printed. */ | |
5847 | static void | |
5848 | parameter_brace_expand_error (name, value) | |
5849 | char *name, *value; | |
5850 | { | |
5851 | WORD_LIST *l; | |
5852 | char *temp; | |
5853 | ||
5854 | if (value && *value) | |
5855 | { | |
5856 | l = expand_string (value, 0); | |
5857 | temp = string_list (l); | |
5858 | report_error ("%s: %s", name, temp ? temp : ""); /* XXX was value not "" */ | |
5859 | FREE (temp); | |
5860 | dispose_words (l); | |
5861 | } | |
5862 | else | |
5863 | report_error (_("%s: parameter null or not set"), name); | |
5864 | ||
5865 | /* Free the data we have allocated during this expansion, since we | |
5866 | are about to longjmp out. */ | |
5867 | free (name); | |
5868 | FREE (value); | |
5869 | } | |
5870 | ||
5871 | /* Return 1 if NAME is something for which parameter_brace_expand_length is | |
5872 | OK to do. */ | |
5873 | static int | |
5874 | valid_length_expression (name) | |
5875 | char *name; | |
5876 | { | |
5877 | return (name[1] == '\0' || /* ${#} */ | |
5878 | ((sh_syntaxtab[(unsigned char) name[1]] & CSPECVAR) && name[2] == '\0') || /* special param */ | |
5879 | (DIGIT (name[1]) && all_digits (name + 1)) || /* ${#11} */ | |
5880 | #if defined (ARRAY_VARS) | |
5881 | valid_array_reference (name + 1) || /* ${#a[7]} */ | |
5882 | #endif | |
5883 | legal_identifier (name + 1)); /* ${#PS1} */ | |
5884 | } | |
5885 | ||
5886 | /* Handle the parameter brace expansion that requires us to return the | |
5887 | length of a parameter. */ | |
5888 | static intmax_t | |
5889 | parameter_brace_expand_length (name) | |
5890 | char *name; | |
5891 | { | |
5892 | char *t, *newname; | |
5893 | intmax_t number, arg_index; | |
5894 | WORD_LIST *list; | |
5895 | #if defined (ARRAY_VARS) | |
5896 | SHELL_VAR *var; | |
5897 | #endif | |
5898 | ||
5899 | if (name[1] == '\0') /* ${#} */ | |
5900 | number = number_of_args (); | |
5901 | else if ((name[1] == '@' || name[1] == '*') && name[2] == '\0') /* ${#@}, ${#*} */ | |
5902 | number = number_of_args (); | |
5903 | else if ((sh_syntaxtab[(unsigned char) name[1]] & CSPECVAR) && name[2] == '\0') | |
5904 | { | |
5905 | /* Take the lengths of some of the shell's special parameters. */ | |
5906 | switch (name[1]) | |
5907 | { | |
5908 | case '-': | |
5909 | t = which_set_flags (); | |
5910 | break; | |
5911 | case '?': | |
5912 | t = itos (last_command_exit_value); | |
5913 | break; | |
5914 | case '$': | |
5915 | t = itos (dollar_dollar_pid); | |
5916 | break; | |
5917 | case '!': | |
5918 | if (last_asynchronous_pid == NO_PID) | |
5919 | t = (char *)NULL; /* XXX - error if set -u set? */ | |
5920 | else | |
5921 | t = itos (last_asynchronous_pid); | |
5922 | break; | |
5923 | case '#': | |
5924 | t = itos (number_of_args ()); | |
5925 | break; | |
5926 | } | |
5927 | number = STRLEN (t); | |
5928 | FREE (t); | |
5929 | } | |
5930 | #if defined (ARRAY_VARS) | |
5931 | else if (valid_array_reference (name + 1)) | |
5932 | number = array_length_reference (name + 1); | |
5933 | #endif /* ARRAY_VARS */ | |
5934 | else | |
5935 | { | |
5936 | number = 0; | |
5937 | ||
5938 | if (legal_number (name + 1, &arg_index)) /* ${#1} */ | |
5939 | { | |
5940 | t = get_dollar_var_value (arg_index); | |
5941 | if (t == 0 && unbound_vars_is_error) | |
5942 | return INTMAX_MIN; | |
5943 | number = MB_STRLEN (t); | |
5944 | FREE (t); | |
5945 | } | |
5946 | #if defined (ARRAY_VARS) | |
5947 | else if ((var = find_variable (name + 1)) && (invisible_p (var) == 0) && (array_p (var) || assoc_p (var))) | |
5948 | { | |
5949 | if (assoc_p (var)) | |
5950 | t = assoc_reference (assoc_cell (var), "0"); | |
5951 | else | |
5952 | t = array_reference (array_cell (var), 0); | |
5953 | if (t == 0 && unbound_vars_is_error) | |
5954 | return INTMAX_MIN; | |
5955 | number = MB_STRLEN (t); | |
5956 | } | |
5957 | #endif | |
5958 | else /* ${#PS1} */ | |
5959 | { | |
5960 | newname = savestring (name); | |
5961 | newname[0] = '$'; | |
5962 | list = expand_string (newname, Q_DOUBLE_QUOTES); | |
5963 | t = list ? string_list (list) : (char *)NULL; | |
5964 | free (newname); | |
5965 | if (list) | |
5966 | dispose_words (list); | |
5967 | ||
5968 | number = t ? MB_STRLEN (t) : 0; | |
5969 | FREE (t); | |
5970 | } | |
5971 | } | |
5972 | ||
5973 | return (number); | |
5974 | } | |
5975 | ||
5976 | /* Skip characters in SUBSTR until DELIM. SUBSTR is an arithmetic expression, | |
5977 | so we do some ad-hoc parsing of an arithmetic expression to find | |
5978 | the first DELIM, instead of using strchr(3). Two rules: | |
5979 | 1. If the substring contains a `(', read until closing `)'. | |
5980 | 2. If the substring contains a `?', read past one `:' for each `?'. | |
5981 | */ | |
5982 | ||
5983 | static char * | |
5984 | skiparith (substr, delim) | |
5985 | char *substr; | |
5986 | int delim; | |
5987 | { | |
5988 | size_t sublen; | |
5989 | int skipcol, pcount, i; | |
5990 | DECLARE_MBSTATE; | |
5991 | ||
5992 | sublen = strlen (substr); | |
5993 | i = skipcol = pcount = 0; | |
5994 | while (substr[i]) | |
5995 | { | |
5996 | /* Balance parens */ | |
5997 | if (substr[i] == LPAREN) | |
5998 | { | |
5999 | pcount++; | |
6000 | i++; | |
6001 | continue; | |
6002 | } | |
6003 | if (substr[i] == RPAREN && pcount) | |
6004 | { | |
6005 | pcount--; | |
6006 | i++; | |
6007 | continue; | |
6008 | } | |
6009 | if (pcount) | |
6010 | { | |
6011 | ADVANCE_CHAR (substr, sublen, i); | |
6012 | continue; | |
6013 | } | |
6014 | ||
6015 | /* Skip one `:' for each `?' */ | |
6016 | if (substr[i] == ':' && skipcol) | |
6017 | { | |
6018 | skipcol--; | |
6019 | i++; | |
6020 | continue; | |
6021 | } | |
6022 | if (substr[i] == delim) | |
6023 | break; | |
6024 | if (substr[i] == '?') | |
6025 | { | |
6026 | skipcol++; | |
6027 | i++; | |
6028 | continue; | |
6029 | } | |
6030 | ADVANCE_CHAR (substr, sublen, i); | |
6031 | } | |
6032 | ||
6033 | return (substr + i); | |
6034 | } | |
6035 | ||
6036 | /* Verify and limit the start and end of the desired substring. If | |
6037 | VTYPE == 0, a regular shell variable is being used; if it is 1, | |
6038 | then the positional parameters are being used; if it is 2, then | |
6039 | VALUE is really a pointer to an array variable that should be used. | |
6040 | Return value is 1 if both values were OK, 0 if there was a problem | |
6041 | with an invalid expression, or -1 if the values were out of range. */ | |
6042 | static int | |
6043 | verify_substring_values (v, value, substr, vtype, e1p, e2p) | |
6044 | SHELL_VAR *v; | |
6045 | char *value, *substr; | |
6046 | int vtype; | |
6047 | intmax_t *e1p, *e2p; | |
6048 | { | |
6049 | char *t, *temp1, *temp2; | |
6050 | arrayind_t len; | |
6051 | int expok; | |
6052 | #if defined (ARRAY_VARS) | |
6053 | ARRAY *a; | |
6054 | HASH_TABLE *h; | |
6055 | #endif | |
6056 | ||
6057 | /* duplicate behavior of strchr(3) */ | |
6058 | t = skiparith (substr, ':'); | |
6059 | if (*t && *t == ':') | |
6060 | *t = '\0'; | |
6061 | else | |
6062 | t = (char *)0; | |
6063 | ||
6064 | temp1 = expand_arith_string (substr, Q_DOUBLE_QUOTES); | |
6065 | *e1p = evalexp (temp1, &expok); | |
6066 | free (temp1); | |
6067 | if (expok == 0) | |
6068 | return (0); | |
6069 | ||
6070 | len = -1; /* paranoia */ | |
6071 | switch (vtype) | |
6072 | { | |
6073 | case VT_VARIABLE: | |
6074 | case VT_ARRAYMEMBER: | |
6075 | len = MB_STRLEN (value); | |
6076 | break; | |
6077 | case VT_POSPARMS: | |
6078 | len = number_of_args () + 1; | |
6079 | if (*e1p == 0) | |
6080 | len++; /* add one arg if counting from $0 */ | |
6081 | break; | |
6082 | #if defined (ARRAY_VARS) | |
6083 | case VT_ARRAYVAR: | |
6084 | /* For arrays, the first value deals with array indices. Negative | |
6085 | offsets count from one past the array's maximum index. Associative | |
6086 | arrays treat the number of elements as the maximum index. */ | |
6087 | if (assoc_p (v)) | |
6088 | { | |
6089 | h = assoc_cell (v); | |
6090 | len = assoc_num_elements (h) + (*e1p < 0); | |
6091 | } | |
6092 | else | |
6093 | { | |
6094 | a = (ARRAY *)value; | |
6095 | len = array_max_index (a) + (*e1p < 0); /* arrays index from 0 to n - 1 */ | |
6096 | } | |
6097 | break; | |
6098 | #endif | |
6099 | } | |
6100 | ||
6101 | if (len == -1) /* paranoia */ | |
6102 | return -1; | |
6103 | ||
6104 | if (*e1p < 0) /* negative offsets count from end */ | |
6105 | *e1p += len; | |
6106 | ||
6107 | if (*e1p > len || *e1p < 0) | |
6108 | return (-1); | |
6109 | ||
6110 | #if defined (ARRAY_VARS) | |
6111 | /* For arrays, the second offset deals with the number of elements. */ | |
6112 | if (vtype == VT_ARRAYVAR) | |
6113 | len = assoc_p (v) ? assoc_num_elements (h) : array_num_elements (a); | |
6114 | #endif | |
6115 | ||
6116 | if (t) | |
6117 | { | |
6118 | t++; | |
6119 | temp2 = savestring (t); | |
6120 | temp1 = expand_arith_string (temp2, Q_DOUBLE_QUOTES); | |
6121 | free (temp2); | |
6122 | t[-1] = ':'; | |
6123 | *e2p = evalexp (temp1, &expok); | |
6124 | free (temp1); | |
6125 | if (expok == 0) | |
6126 | return (0); | |
6127 | if ((vtype == VT_ARRAYVAR || vtype == VT_POSPARMS) && *e2p < 0) | |
6128 | { | |
6129 | internal_error (_("%s: substring expression < 0"), t); | |
6130 | return (0); | |
6131 | } | |
6132 | #if defined (ARRAY_VARS) | |
6133 | /* In order to deal with sparse arrays, push the intelligence about how | |
6134 | to deal with the number of elements desired down to the array- | |
6135 | specific functions. */ | |
6136 | if (vtype != VT_ARRAYVAR) | |
6137 | #endif | |
6138 | { | |
6139 | if (*e2p < 0) | |
6140 | { | |
6141 | *e2p += len; | |
6142 | if (*e2p < 0 || *e2p < *e1p) | |
6143 | { | |
6144 | internal_error (_("%s: substring expression < 0"), t); | |
6145 | return (0); | |
6146 | } | |
6147 | } | |
6148 | else | |
6149 | *e2p += *e1p; /* want E2 chars starting at E1 */ | |
6150 | if (*e2p > len) | |
6151 | *e2p = len; | |
6152 | } | |
6153 | } | |
6154 | else | |
6155 | *e2p = len; | |
6156 | ||
6157 | return (1); | |
6158 | } | |
6159 | ||
6160 | /* Return the type of variable specified by VARNAME (simple variable, | |
6161 | positional param, or array variable). Also return the value specified | |
6162 | by VARNAME (value of a variable or a reference to an array element). | |
6163 | QUOTED is the standard description of quoting state, using Q_* defines. | |
6164 | FLAGS is currently a set of flags to pass to array_value. If IND is | |
6165 | non-null and not INTMAX_MIN, and FLAGS includes AV_USEIND, IND is | |
6166 | passed to array_value so the array index is not computed again. | |
6167 | If this returns VT_VARIABLE, the caller assumes that CTLESC and CTLNUL | |
6168 | characters in the value are quoted with CTLESC and takes appropriate | |
6169 | steps. For convenience, *VALP is set to the dequoted VALUE. */ | |
6170 | static int | |
6171 | get_var_and_type (varname, value, ind, quoted, flags, varp, valp) | |
6172 | char *varname, *value; | |
6173 | arrayind_t ind; | |
6174 | int quoted, flags; | |
6175 | SHELL_VAR **varp; | |
6176 | char **valp; | |
6177 | { | |
6178 | int vtype; | |
6179 | char *temp; | |
6180 | #if defined (ARRAY_VARS) | |
6181 | SHELL_VAR *v; | |
6182 | #endif | |
6183 | arrayind_t lind; | |
6184 | ||
6185 | /* This sets vtype to VT_VARIABLE or VT_POSPARMS */ | |
6186 | vtype = (varname[0] == '@' || varname[0] == '*') && varname[1] == '\0'; | |
6187 | if (vtype == VT_POSPARMS && varname[0] == '*') | |
6188 | vtype |= VT_STARSUB; | |
6189 | *varp = (SHELL_VAR *)NULL; | |
6190 | ||
6191 | #if defined (ARRAY_VARS) | |
6192 | if (valid_array_reference (varname)) | |
6193 | { | |
6194 | v = array_variable_part (varname, &temp, (int *)0); | |
6195 | /* If we want to signal array_value to use an already-computed index, | |
6196 | set LIND to that index */ | |
6197 | lind = (ind != INTMAX_MIN && (flags & AV_USEIND)) ? ind : 0; | |
6198 | if (v && (array_p (v) || assoc_p (v))) | |
6199 | { /* [ */ | |
6200 | if (ALL_ELEMENT_SUB (temp[0]) && temp[1] == ']') | |
6201 | { | |
6202 | /* Callers have to differentiate betwen indexed and associative */ | |
6203 | vtype = VT_ARRAYVAR; | |
6204 | if (temp[0] == '*') | |
6205 | vtype |= VT_STARSUB; | |
6206 | *valp = array_p (v) ? (char *)array_cell (v) : (char *)assoc_cell (v); | |
6207 | } | |
6208 | else | |
6209 | { | |
6210 | vtype = VT_ARRAYMEMBER; | |
6211 | *valp = array_value (varname, Q_DOUBLE_QUOTES, flags, (int *)NULL, &lind); | |
6212 | } | |
6213 | *varp = v; | |
6214 | } | |
6215 | else if (v && (ALL_ELEMENT_SUB (temp[0]) && temp[1] == ']')) | |
6216 | { | |
6217 | vtype = VT_VARIABLE; | |
6218 | *varp = v; | |
6219 | if (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) | |
6220 | *valp = dequote_string (value); | |
6221 | else | |
6222 | *valp = dequote_escapes (value); | |
6223 | } | |
6224 | else | |
6225 | { | |
6226 | vtype = VT_ARRAYMEMBER; | |
6227 | *varp = v; | |
6228 | *valp = array_value (varname, Q_DOUBLE_QUOTES, flags, (int *)NULL, &lind); | |
6229 | } | |
6230 | } | |
6231 | else if ((v = find_variable (varname)) && (invisible_p (v) == 0) && (assoc_p (v) || array_p (v))) | |
6232 | { | |
6233 | vtype = VT_ARRAYMEMBER; | |
6234 | *varp = v; | |
6235 | *valp = assoc_p (v) ? assoc_reference (assoc_cell (v), "0") : array_reference (array_cell (v), 0); | |
6236 | } | |
6237 | else | |
6238 | #endif | |
6239 | { | |
6240 | if (value && vtype == VT_VARIABLE) | |
6241 | { | |
6242 | if (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) | |
6243 | *valp = dequote_string (value); | |
6244 | else | |
6245 | *valp = dequote_escapes (value); | |
6246 | } | |
6247 | else | |
6248 | *valp = value; | |
6249 | } | |
6250 | ||
6251 | return vtype; | |
6252 | } | |
6253 | ||
6254 | /******************************************************/ | |
6255 | /* */ | |
6256 | /* Functions to extract substrings of variable values */ | |
6257 | /* */ | |
6258 | /******************************************************/ | |
6259 | ||
6260 | #if defined (HANDLE_MULTIBYTE) | |
6261 | /* Character-oriented rather than strictly byte-oriented substrings. S and | |
6262 | E, rather being strict indices into STRING, indicate character (possibly | |
6263 | multibyte character) positions that require calculation. | |
6264 | Used by the ${param:offset[:length]} expansion. */ | |
6265 | static char * | |
6266 | mb_substring (string, s, e) | |
6267 | char *string; | |
6268 | int s, e; | |
6269 | { | |
6270 | char *tt; | |
6271 | int start, stop, i, slen; | |
6272 | DECLARE_MBSTATE; | |
6273 | ||
6274 | start = 0; | |
6275 | /* Don't need string length in ADVANCE_CHAR unless multibyte chars possible. */ | |
6276 | slen = (MB_CUR_MAX > 1) ? STRLEN (string) : 0; | |
6277 | ||
6278 | i = s; | |
6279 | while (string[start] && i--) | |
6280 | ADVANCE_CHAR (string, slen, start); | |
6281 | stop = start; | |
6282 | i = e - s; | |
6283 | while (string[stop] && i--) | |
6284 | ADVANCE_CHAR (string, slen, stop); | |
6285 | tt = substring (string, start, stop); | |
6286 | return tt; | |
6287 | } | |
6288 | #endif | |
6289 | ||
6290 | /* Process a variable substring expansion: ${name:e1[:e2]}. If VARNAME | |
6291 | is `@', use the positional parameters; otherwise, use the value of | |
6292 | VARNAME. If VARNAME is an array variable, use the array elements. */ | |
6293 | ||
6294 | static char * | |
6295 | parameter_brace_substring (varname, value, ind, substr, quoted, flags) | |
6296 | char *varname, *value; | |
6297 | int ind; | |
6298 | char *substr; | |
6299 | int quoted, flags; | |
6300 | { | |
6301 | intmax_t e1, e2; | |
6302 | int vtype, r, starsub; | |
6303 | char *temp, *val, *tt, *oname; | |
6304 | SHELL_VAR *v; | |
6305 | ||
6306 | if (value == 0) | |
6307 | return ((char *)NULL); | |
6308 | ||
6309 | oname = this_command_name; | |
6310 | this_command_name = varname; | |
6311 | ||
6312 | vtype = get_var_and_type (varname, value, ind, quoted, flags, &v, &val); | |
6313 | if (vtype == -1) | |
6314 | { | |
6315 | this_command_name = oname; | |
6316 | return ((char *)NULL); | |
6317 | } | |
6318 | ||
6319 | starsub = vtype & VT_STARSUB; | |
6320 | vtype &= ~VT_STARSUB; | |
6321 | ||
6322 | r = verify_substring_values (v, val, substr, vtype, &e1, &e2); | |
6323 | this_command_name = oname; | |
6324 | if (r <= 0) | |
6325 | return ((r == 0) ? &expand_param_error : (char *)NULL); | |
6326 | ||
6327 | switch (vtype) | |
6328 | { | |
6329 | case VT_VARIABLE: | |
6330 | case VT_ARRAYMEMBER: | |
6331 | #if defined (HANDLE_MULTIBYTE) | |
6332 | if (MB_CUR_MAX > 1) | |
6333 | tt = mb_substring (val, e1, e2); | |
6334 | else | |
6335 | #endif | |
6336 | tt = substring (val, e1, e2); | |
6337 | ||
6338 | if (vtype == VT_VARIABLE) | |
6339 | FREE (val); | |
6340 | if (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) | |
6341 | temp = quote_string (tt); | |
6342 | else | |
6343 | temp = tt ? quote_escapes (tt) : (char *)NULL; | |
6344 | FREE (tt); | |
6345 | break; | |
6346 | case VT_POSPARMS: | |
6347 | tt = pos_params (varname, e1, e2, quoted); | |
6348 | if ((quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) == 0) | |
6349 | { | |
6350 | temp = tt ? quote_escapes (tt) : (char *)NULL; | |
6351 | FREE (tt); | |
6352 | } | |
6353 | else | |
6354 | temp = tt; | |
6355 | break; | |
6356 | #if defined (ARRAY_VARS) | |
6357 | case VT_ARRAYVAR: | |
6358 | if (assoc_p (v)) | |
6359 | /* we convert to list and take first e2 elements starting at e1th | |
6360 | element -- officially undefined for now */ | |
6361 | temp = assoc_subrange (assoc_cell (v), e1, e2, starsub, quoted); | |
6362 | else | |
6363 | /* We want E2 to be the number of elements desired (arrays can be sparse, | |
6364 | so verify_substring_values just returns the numbers specified and we | |
6365 | rely on array_subrange to understand how to deal with them). */ | |
6366 | temp = array_subrange (array_cell (v), e1, e2, starsub, quoted); | |
6367 | /* array_subrange now calls array_quote_escapes as appropriate, so the | |
6368 | caller no longer needs to. */ | |
6369 | break; | |
6370 | #endif | |
6371 | default: | |
6372 | temp = (char *)NULL; | |
6373 | } | |
6374 | ||
6375 | return temp; | |
6376 | } | |
6377 | ||
6378 | /****************************************************************/ | |
6379 | /* */ | |
6380 | /* Functions to perform pattern substitution on variable values */ | |
6381 | /* */ | |
6382 | /****************************************************************/ | |
6383 | ||
6384 | static int | |
6385 | shouldexp_replacement (s) | |
6386 | char *s; | |
6387 | { | |
6388 | register char *p; | |
6389 | ||
6390 | for (p = s; p && *p; p++) | |
6391 | { | |
6392 | if (*p == '\\') | |
6393 | p++; | |
6394 | else if (*p == '&') | |
6395 | return 1; | |
6396 | } | |
6397 | return 0; | |
6398 | } | |
6399 | ||
6400 | char * | |
6401 | pat_subst (string, pat, rep, mflags) | |
6402 | char *string, *pat, *rep; | |
6403 | int mflags; | |
6404 | { | |
6405 | char *ret, *s, *e, *str, *rstr, *mstr; | |
6406 | int rsize, rptr, l, replen, mtype, rxpand, rslen, mlen; | |
6407 | ||
6408 | if (string == 0) | |
6409 | return (savestring ("")); | |
6410 | ||
6411 | mtype = mflags & MATCH_TYPEMASK; | |
6412 | ||
6413 | #if 0 /* bash-4.2 ? */ | |
6414 | rxpand = (rep && *rep) ? shouldexp_replacement (rep) : 0; | |
6415 | #else | |
6416 | rxpand = 0; | |
6417 | #endif | |
6418 | ||
6419 | /* Special cases: | |
6420 | * 1. A null pattern with mtype == MATCH_BEG means to prefix STRING | |
6421 | * with REP and return the result. | |
6422 | * 2. A null pattern with mtype == MATCH_END means to append REP to | |
6423 | * STRING and return the result. | |
6424 | * These don't understand or process `&' in the replacement string. | |
6425 | */ | |
6426 | if ((pat == 0 || *pat == 0) && (mtype == MATCH_BEG || mtype == MATCH_END)) | |
6427 | { | |
6428 | replen = STRLEN (rep); | |
6429 | l = STRLEN (string); | |
6430 | ret = (char *)xmalloc (replen + l + 2); | |
6431 | if (replen == 0) | |
6432 | strcpy (ret, string); | |
6433 | else if (mtype == MATCH_BEG) | |
6434 | { | |
6435 | strcpy (ret, rep); | |
6436 | strcpy (ret + replen, string); | |
6437 | } | |
6438 | else | |
6439 | { | |
6440 | strcpy (ret, string); | |
6441 | strcpy (ret + l, rep); | |
6442 | } | |
6443 | return (ret); | |
6444 | } | |
6445 | ||
6446 | ret = (char *)xmalloc (rsize = 64); | |
6447 | ret[0] = '\0'; | |
6448 | ||
6449 | for (replen = STRLEN (rep), rptr = 0, str = string;;) | |
6450 | { | |
6451 | if (match_pattern (str, pat, mtype, &s, &e) == 0) | |
6452 | break; | |
6453 | l = s - str; | |
6454 | ||
6455 | if (rxpand) | |
6456 | { | |
6457 | int x; | |
6458 | mlen = e - s; | |
6459 | mstr = xmalloc (mlen + 1); | |
6460 | for (x = 0; x < mlen; x++) | |
6461 | mstr[x] = s[x]; | |
6462 | mstr[mlen] = '\0'; | |
6463 | rstr = strcreplace (rep, '&', mstr, 0); | |
6464 | rslen = strlen (rstr); | |
6465 | } | |
6466 | else | |
6467 | { | |
6468 | rstr = rep; | |
6469 | rslen = replen; | |
6470 | } | |
6471 | ||
6472 | RESIZE_MALLOCED_BUFFER (ret, rptr, (l + rslen), rsize, 64); | |
6473 | ||
6474 | /* OK, now copy the leading unmatched portion of the string (from | |
6475 | str to s) to ret starting at rptr (the current offset). Then copy | |
6476 | the replacement string at ret + rptr + (s - str). Increment | |
6477 | rptr (if necessary) and str and go on. */ | |
6478 | if (l) | |
6479 | { | |
6480 | strncpy (ret + rptr, str, l); | |
6481 | rptr += l; | |
6482 | } | |
6483 | if (replen) | |
6484 | { | |
6485 | strncpy (ret + rptr, rstr, rslen); | |
6486 | rptr += rslen; | |
6487 | } | |
6488 | str = e; /* e == end of match */ | |
6489 | ||
6490 | if (rstr != rep) | |
6491 | free (rstr); | |
6492 | ||
6493 | if (((mflags & MATCH_GLOBREP) == 0) || mtype != MATCH_ANY) | |
6494 | break; | |
6495 | ||
6496 | if (s == e) | |
6497 | { | |
6498 | /* On a zero-length match, make sure we copy one character, since | |
6499 | we increment one character to avoid infinite recursion. */ | |
6500 | RESIZE_MALLOCED_BUFFER (ret, rptr, 1, rsize, 64); | |
6501 | ret[rptr++] = *str++; | |
6502 | e++; /* avoid infinite recursion on zero-length match */ | |
6503 | } | |
6504 | } | |
6505 | ||
6506 | /* Now copy the unmatched portion of the input string */ | |
6507 | if (str && *str) | |
6508 | { | |
6509 | RESIZE_MALLOCED_BUFFER (ret, rptr, STRLEN(str) + 1, rsize, 64); | |
6510 | strcpy (ret + rptr, str); | |
6511 | } | |
6512 | else | |
6513 | ret[rptr] = '\0'; | |
6514 | ||
6515 | return ret; | |
6516 | } | |
6517 | ||
6518 | /* Do pattern match and replacement on the positional parameters. */ | |
6519 | static char * | |
6520 | pos_params_pat_subst (string, pat, rep, mflags) | |
6521 | char *string, *pat, *rep; | |
6522 | int mflags; | |
6523 | { | |
6524 | WORD_LIST *save, *params; | |
6525 | WORD_DESC *w; | |
6526 | char *ret; | |
6527 | int pchar, qflags; | |
6528 | ||
6529 | save = params = list_rest_of_args (); | |
6530 | if (save == 0) | |
6531 | return ((char *)NULL); | |
6532 | ||
6533 | for ( ; params; params = params->next) | |
6534 | { | |
6535 | ret = pat_subst (params->word->word, pat, rep, mflags); | |
6536 | w = alloc_word_desc (); | |
6537 | w->word = ret ? ret : savestring (""); | |
6538 | dispose_word (params->word); | |
6539 | params->word = w; | |
6540 | } | |
6541 | ||
6542 | pchar = (mflags & MATCH_STARSUB) == MATCH_STARSUB ? '*' : '@'; | |
6543 | qflags = (mflags & MATCH_QUOTED) == MATCH_QUOTED ? Q_DOUBLE_QUOTES : 0; | |
6544 | ||
6545 | #if 0 | |
6546 | if ((mflags & (MATCH_QUOTED|MATCH_STARSUB)) == (MATCH_QUOTED|MATCH_STARSUB)) | |
6547 | ret = string_list_dollar_star (quote_list (save)); | |
6548 | else if ((mflags & MATCH_STARSUB) == MATCH_STARSUB) | |
6549 | ret = string_list_dollar_star (save); | |
6550 | else if ((mflags & MATCH_QUOTED) == MATCH_QUOTED) | |
6551 | ret = string_list_dollar_at (save, qflags); | |
6552 | else | |
6553 | ret = string_list_dollar_star (save); | |
6554 | #else | |
6555 | ret = string_list_pos_params (pchar, save, qflags); | |
6556 | #endif | |
6557 | ||
6558 | dispose_words (save); | |
6559 | ||
6560 | return (ret); | |
6561 | } | |
6562 | ||
6563 | /* Perform pattern substitution on VALUE, which is the expansion of | |
6564 | VARNAME. PATSUB is an expression supplying the pattern to match | |
6565 | and the string to substitute. QUOTED is a flags word containing | |
6566 | the type of quoting currently in effect. */ | |
6567 | static char * | |
6568 | parameter_brace_patsub (varname, value, ind, patsub, quoted, flags) | |
6569 | char *varname, *value; | |
6570 | int ind; | |
6571 | char *patsub; | |
6572 | int quoted, flags; | |
6573 | { | |
6574 | int vtype, mflags, starsub, delim; | |
6575 | char *val, *temp, *pat, *rep, *p, *lpatsub, *tt; | |
6576 | SHELL_VAR *v; | |
6577 | ||
6578 | if (value == 0) | |
6579 | return ((char *)NULL); | |
6580 | ||
6581 | this_command_name = varname; | |
6582 | ||
6583 | vtype = get_var_and_type (varname, value, ind, quoted, flags, &v, &val); | |
6584 | if (vtype == -1) | |
6585 | return ((char *)NULL); | |
6586 | ||
6587 | starsub = vtype & VT_STARSUB; | |
6588 | vtype &= ~VT_STARSUB; | |
6589 | ||
6590 | mflags = 0; | |
6591 | if (patsub && *patsub == '/') | |
6592 | { | |
6593 | mflags |= MATCH_GLOBREP; | |
6594 | patsub++; | |
6595 | } | |
6596 | ||
6597 | /* Malloc this because expand_string_if_necessary or one of the expansion | |
6598 | functions in its call chain may free it on a substitution error. */ | |
6599 | lpatsub = savestring (patsub); | |
6600 | ||
6601 | if (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
6602 | mflags |= MATCH_QUOTED; | |
6603 | ||
6604 | if (starsub) | |
6605 | mflags |= MATCH_STARSUB; | |
6606 | ||
6607 | /* If the pattern starts with a `/', make sure we skip over it when looking | |
6608 | for the replacement delimiter. */ | |
6609 | #if 0 | |
6610 | if (rep = quoted_strchr ((*patsub == '/') ? lpatsub+1 : lpatsub, '/', ST_BACKSL)) | |
6611 | *rep++ = '\0'; | |
6612 | else | |
6613 | rep = (char *)NULL; | |
6614 | #else | |
6615 | delim = skip_to_delim (lpatsub, ((*patsub == '/') ? 1 : 0), "/", 0); | |
6616 | if (lpatsub[delim] == '/') | |
6617 | { | |
6618 | lpatsub[delim] = 0; | |
6619 | rep = lpatsub + delim + 1; | |
6620 | } | |
6621 | else | |
6622 | rep = (char *)NULL; | |
6623 | #endif | |
6624 | ||
6625 | if (rep && *rep == '\0') | |
6626 | rep = (char *)NULL; | |
6627 | ||
6628 | /* Perform the same expansions on the pattern as performed by the | |
6629 | pattern removal expansions. */ | |
6630 | pat = getpattern (lpatsub, quoted, 1); | |
6631 | ||
6632 | if (rep) | |
6633 | { | |
6634 | if ((mflags & MATCH_QUOTED) == 0) | |
6635 | rep = expand_string_if_necessary (rep, quoted, expand_string_unsplit); | |
6636 | else | |
6637 | rep = expand_string_to_string_internal (rep, quoted, expand_string_unsplit); | |
6638 | } | |
6639 | ||
6640 | /* ksh93 doesn't allow the match specifier to be a part of the expanded | |
6641 | pattern. This is an extension. Make sure we don't anchor the pattern | |
6642 | at the beginning or end of the string if we're doing global replacement, | |
6643 | though. */ | |
6644 | p = pat; | |
6645 | if (mflags & MATCH_GLOBREP) | |
6646 | mflags |= MATCH_ANY; | |
6647 | else if (pat && pat[0] == '#') | |
6648 | { | |
6649 | mflags |= MATCH_BEG; | |
6650 | p++; | |
6651 | } | |
6652 | else if (pat && pat[0] == '%') | |
6653 | { | |
6654 | mflags |= MATCH_END; | |
6655 | p++; | |
6656 | } | |
6657 | else | |
6658 | mflags |= MATCH_ANY; | |
6659 | ||
6660 | /* OK, we now want to substitute REP for PAT in VAL. If | |
6661 | flags & MATCH_GLOBREP is non-zero, the substitution is done | |
6662 | everywhere, otherwise only the first occurrence of PAT is | |
6663 | replaced. The pattern matching code doesn't understand | |
6664 | CTLESC quoting CTLESC and CTLNUL so we use the dequoted variable | |
6665 | values passed in (VT_VARIABLE) so the pattern substitution | |
6666 | code works right. We need to requote special chars after | |
6667 | we're done for VT_VARIABLE and VT_ARRAYMEMBER, and for the | |
6668 | other cases if QUOTED == 0, since the posparams and arrays | |
6669 | indexed by * or @ do special things when QUOTED != 0. */ | |
6670 | ||
6671 | switch (vtype) | |
6672 | { | |
6673 | case VT_VARIABLE: | |
6674 | case VT_ARRAYMEMBER: | |
6675 | temp = pat_subst (val, p, rep, mflags); | |
6676 | if (vtype == VT_VARIABLE) | |
6677 | FREE (val); | |
6678 | if (temp) | |
6679 | { | |
6680 | tt = (mflags & MATCH_QUOTED) ? quote_string (temp) : quote_escapes (temp); | |
6681 | free (temp); | |
6682 | temp = tt; | |
6683 | } | |
6684 | break; | |
6685 | case VT_POSPARMS: | |
6686 | temp = pos_params_pat_subst (val, p, rep, mflags); | |
6687 | if (temp && (mflags & MATCH_QUOTED) == 0) | |
6688 | { | |
6689 | tt = quote_escapes (temp); | |
6690 | free (temp); | |
6691 | temp = tt; | |
6692 | } | |
6693 | break; | |
6694 | #if defined (ARRAY_VARS) | |
6695 | case VT_ARRAYVAR: | |
6696 | temp = assoc_p (v) ? assoc_patsub (assoc_cell (v), p, rep, mflags) | |
6697 | : array_patsub (array_cell (v), p, rep, mflags); | |
6698 | /* Don't call quote_escapes anymore; array_patsub calls | |
6699 | array_quote_escapes as appropriate before adding the | |
6700 | space separators; ditto for assoc_patsub. */ | |
6701 | break; | |
6702 | #endif | |
6703 | } | |
6704 | ||
6705 | FREE (pat); | |
6706 | FREE (rep); | |
6707 | free (lpatsub); | |
6708 | ||
6709 | return temp; | |
6710 | } | |
6711 | ||
6712 | /****************************************************************/ | |
6713 | /* */ | |
6714 | /* Functions to perform case modification on variable values */ | |
6715 | /* */ | |
6716 | /****************************************************************/ | |
6717 | ||
6718 | /* Do case modification on the positional parameters. */ | |
6719 | ||
6720 | static char * | |
6721 | pos_params_modcase (string, pat, modop, mflags) | |
6722 | char *string, *pat; | |
6723 | int modop; | |
6724 | int mflags; | |
6725 | { | |
6726 | WORD_LIST *save, *params; | |
6727 | WORD_DESC *w; | |
6728 | char *ret; | |
6729 | int pchar, qflags; | |
6730 | ||
6731 | save = params = list_rest_of_args (); | |
6732 | if (save == 0) | |
6733 | return ((char *)NULL); | |
6734 | ||
6735 | for ( ; params; params = params->next) | |
6736 | { | |
6737 | ret = sh_modcase (params->word->word, pat, modop); | |
6738 | w = alloc_word_desc (); | |
6739 | w->word = ret ? ret : savestring (""); | |
6740 | dispose_word (params->word); | |
6741 | params->word = w; | |
6742 | } | |
6743 | ||
6744 | pchar = (mflags & MATCH_STARSUB) == MATCH_STARSUB ? '*' : '@'; | |
6745 | qflags = (mflags & MATCH_QUOTED) == MATCH_QUOTED ? Q_DOUBLE_QUOTES : 0; | |
6746 | ||
6747 | ret = string_list_pos_params (pchar, save, qflags); | |
6748 | dispose_words (save); | |
6749 | ||
6750 | return (ret); | |
6751 | } | |
6752 | ||
6753 | /* Perform case modification on VALUE, which is the expansion of | |
6754 | VARNAME. MODSPEC is an expression supplying the type of modification | |
6755 | to perform. QUOTED is a flags word containing the type of quoting | |
6756 | currently in effect. */ | |
6757 | static char * | |
6758 | parameter_brace_casemod (varname, value, ind, modspec, patspec, quoted, flags) | |
6759 | char *varname, *value; | |
6760 | int ind, modspec; | |
6761 | char *patspec; | |
6762 | int quoted, flags; | |
6763 | { | |
6764 | int vtype, starsub, modop, mflags, x; | |
6765 | char *val, *temp, *pat, *p, *lpat, *tt; | |
6766 | SHELL_VAR *v; | |
6767 | ||
6768 | if (value == 0) | |
6769 | return ((char *)NULL); | |
6770 | ||
6771 | this_command_name = varname; | |
6772 | ||
6773 | vtype = get_var_and_type (varname, value, ind, quoted, flags, &v, &val); | |
6774 | if (vtype == -1) | |
6775 | return ((char *)NULL); | |
6776 | ||
6777 | starsub = vtype & VT_STARSUB; | |
6778 | vtype &= ~VT_STARSUB; | |
6779 | ||
6780 | modop = 0; | |
6781 | mflags = 0; | |
6782 | if (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
6783 | mflags |= MATCH_QUOTED; | |
6784 | if (starsub) | |
6785 | mflags |= MATCH_STARSUB; | |
6786 | ||
6787 | p = patspec; | |
6788 | if (modspec == '^') | |
6789 | { | |
6790 | x = p && p[0] == modspec; | |
6791 | modop = x ? CASE_UPPER : CASE_UPFIRST; | |
6792 | p += x; | |
6793 | } | |
6794 | else if (modspec == ',') | |
6795 | { | |
6796 | x = p && p[0] == modspec; | |
6797 | modop = x ? CASE_LOWER : CASE_LOWFIRST; | |
6798 | p += x; | |
6799 | } | |
6800 | else if (modspec == '~') | |
6801 | { | |
6802 | x = p && p[0] == modspec; | |
6803 | modop = x ? CASE_TOGGLEALL : CASE_TOGGLE; | |
6804 | p += x; | |
6805 | } | |
6806 | ||
6807 | lpat = p ? savestring (p) : 0; | |
6808 | /* Perform the same expansions on the pattern as performed by the | |
6809 | pattern removal expansions. FOR LATER */ | |
6810 | pat = lpat ? getpattern (lpat, quoted, 1) : 0; | |
6811 | ||
6812 | /* OK, now we do the case modification. */ | |
6813 | switch (vtype) | |
6814 | { | |
6815 | case VT_VARIABLE: | |
6816 | case VT_ARRAYMEMBER: | |
6817 | temp = sh_modcase (val, pat, modop); | |
6818 | if (vtype == VT_VARIABLE) | |
6819 | FREE (val); | |
6820 | if (temp) | |
6821 | { | |
6822 | tt = (mflags & MATCH_QUOTED) ? quote_string (temp) : quote_escapes (temp); | |
6823 | free (temp); | |
6824 | temp = tt; | |
6825 | } | |
6826 | break; | |
6827 | ||
6828 | case VT_POSPARMS: | |
6829 | temp = pos_params_modcase (val, pat, modop, mflags); | |
6830 | if (temp && (mflags & MATCH_QUOTED) == 0) | |
6831 | { | |
6832 | tt = quote_escapes (temp); | |
6833 | free (temp); | |
6834 | temp = tt; | |
6835 | } | |
6836 | break; | |
6837 | ||
6838 | #if defined (ARRAY_VARS) | |
6839 | case VT_ARRAYVAR: | |
6840 | temp = assoc_p (v) ? assoc_modcase (assoc_cell (v), pat, modop, mflags) | |
6841 | : array_modcase (array_cell (v), pat, modop, mflags); | |
6842 | /* Don't call quote_escapes; array_modcase calls array_quote_escapes | |
6843 | as appropriate before adding the space separators; ditto for | |
6844 | assoc_modcase. */ | |
6845 | break; | |
6846 | #endif | |
6847 | } | |
6848 | ||
6849 | FREE (pat); | |
6850 | free (lpat); | |
6851 | ||
6852 | return temp; | |
6853 | } | |
6854 | ||
6855 | /* Check for unbalanced parens in S, which is the contents of $(( ... )). If | |
6856 | any occur, this must be a nested command substitution, so return 0. | |
6857 | Otherwise, return 1. A valid arithmetic expression must always have a | |
6858 | ( before a matching ), so any cases where there are more right parens | |
6859 | means that this must not be an arithmetic expression, though the parser | |
6860 | will not accept it without a balanced total number of parens. */ | |
6861 | static int | |
6862 | chk_arithsub (s, len) | |
6863 | const char *s; | |
6864 | int len; | |
6865 | { | |
6866 | int i, count; | |
6867 | DECLARE_MBSTATE; | |
6868 | ||
6869 | i = count = 0; | |
6870 | while (i < len) | |
6871 | { | |
6872 | if (s[i] == LPAREN) | |
6873 | count++; | |
6874 | else if (s[i] == RPAREN) | |
6875 | { | |
6876 | count--; | |
6877 | if (count < 0) | |
6878 | return 0; | |
6879 | } | |
6880 | ||
6881 | switch (s[i]) | |
6882 | { | |
6883 | default: | |
6884 | ADVANCE_CHAR (s, len, i); | |
6885 | break; | |
6886 | ||
6887 | case '\\': | |
6888 | i++; | |
6889 | if (s[i]) | |
6890 | ADVANCE_CHAR (s, len, i); | |
6891 | break; | |
6892 | ||
6893 | case '\'': | |
6894 | i = skip_single_quoted (s, len, ++i); | |
6895 | break; | |
6896 | ||
6897 | case '"': | |
6898 | i = skip_double_quoted ((char *)s, len, ++i); | |
6899 | break; | |
6900 | } | |
6901 | } | |
6902 | ||
6903 | return (count == 0); | |
6904 | } | |
6905 | ||
6906 | /****************************************************************/ | |
6907 | /* */ | |
6908 | /* Functions to perform parameter expansion on a string */ | |
6909 | /* */ | |
6910 | /****************************************************************/ | |
6911 | ||
6912 | /* ${[#][!]name[[:][^[^]][,[,]]#[#]%[%]-=?+[word][:e1[:e2]]]} */ | |
6913 | static WORD_DESC * | |
6914 | parameter_brace_expand (string, indexp, quoted, pflags, quoted_dollar_atp, contains_dollar_at) | |
6915 | char *string; | |
6916 | int *indexp, quoted, *quoted_dollar_atp, *contains_dollar_at, pflags; | |
6917 | { | |
6918 | int check_nullness, var_is_set, var_is_null, var_is_special; | |
6919 | int want_substring, want_indir, want_patsub, want_casemod; | |
6920 | char *name, *value, *temp, *temp1; | |
6921 | WORD_DESC *tdesc, *ret; | |
6922 | int t_index, sindex, c, tflag, modspec; | |
6923 | intmax_t number; | |
6924 | arrayind_t ind; | |
6925 | ||
6926 | temp = temp1 = value = (char *)NULL; | |
6927 | var_is_set = var_is_null = var_is_special = check_nullness = 0; | |
6928 | want_substring = want_indir = want_patsub = want_casemod = 0; | |
6929 | ||
6930 | sindex = *indexp; | |
6931 | t_index = ++sindex; | |
6932 | /* ${#var} doesn't have any of the other parameter expansions on it. */ | |
6933 | if (string[t_index] == '#' && legal_variable_starter (string[t_index+1])) /* {{ */ | |
6934 | name = string_extract (string, &t_index, "}", SX_VARNAME); | |
6935 | else | |
6936 | #if defined (CASEMOD_EXPANSIONS) | |
6937 | /* To enable case-toggling expansions using the `~' operator character | |
6938 | change the 1 to 0. */ | |
6939 | # if defined (CASEMOD_CAPCASE) | |
6940 | name = string_extract (string, &t_index, "#%^,~:-=?+/}", SX_VARNAME); | |
6941 | # else | |
6942 | name = string_extract (string, &t_index, "#%^,:-=?+/}", SX_VARNAME); | |
6943 | # endif /* CASEMOD_CAPCASE */ | |
6944 | #else | |
6945 | name = string_extract (string, &t_index, "#%:-=?+/}", SX_VARNAME); | |
6946 | #endif /* CASEMOD_EXPANSIONS */ | |
6947 | ||
6948 | ret = 0; | |
6949 | tflag = 0; | |
6950 | ||
6951 | ind = INTMAX_MIN; | |
6952 | ||
6953 | /* If the name really consists of a special variable, then make sure | |
6954 | that we have the entire name. We don't allow indirect references | |
6955 | to special variables except `#', `?', `@' and `*'. */ | |
6956 | if ((sindex == t_index && VALID_SPECIAL_LENGTH_PARAM (string[t_index])) || | |
6957 | (sindex == t_index - 1 && string[sindex] == '!' && VALID_INDIR_PARAM (string[t_index]))) | |
6958 | { | |
6959 | t_index++; | |
6960 | free (name); | |
6961 | temp1 = string_extract (string, &t_index, "#%:-=?+/}", 0); | |
6962 | name = (char *)xmalloc (3 + (strlen (temp1))); | |
6963 | *name = string[sindex]; | |
6964 | if (string[sindex] == '!') | |
6965 | { | |
6966 | /* indirect reference of $#, $?, $@, or $* */ | |
6967 | name[1] = string[sindex + 1]; | |
6968 | strcpy (name + 2, temp1); | |
6969 | } | |
6970 | else | |
6971 | strcpy (name + 1, temp1); | |
6972 | free (temp1); | |
6973 | } | |
6974 | sindex = t_index; | |
6975 | ||
6976 | /* Find out what character ended the variable name. Then | |
6977 | do the appropriate thing. */ | |
6978 | if (c = string[sindex]) | |
6979 | sindex++; | |
6980 | ||
6981 | /* If c is followed by one of the valid parameter expansion | |
6982 | characters, move past it as normal. If not, assume that | |
6983 | a substring specification is being given, and do not move | |
6984 | past it. */ | |
6985 | if (c == ':' && VALID_PARAM_EXPAND_CHAR (string[sindex])) | |
6986 | { | |
6987 | check_nullness++; | |
6988 | if (c = string[sindex]) | |
6989 | sindex++; | |
6990 | } | |
6991 | else if (c == ':' && string[sindex] != RBRACE) | |
6992 | want_substring = 1; | |
6993 | else if (c == '/' && string[sindex] != RBRACE) | |
6994 | want_patsub = 1; | |
6995 | #if defined (CASEMOD_EXPANSIONS) | |
6996 | else if (c == '^' || c == ',' || c == '~') | |
6997 | { | |
6998 | modspec = c; | |
6999 | want_casemod = 1; | |
7000 | } | |
7001 | #endif | |
7002 | ||
7003 | /* Catch the valid and invalid brace expressions that made it through the | |
7004 | tests above. */ | |
7005 | /* ${#-} is a valid expansion and means to take the length of $-. | |
7006 | Similarly for ${#?} and ${##}... */ | |
7007 | if (name[0] == '#' && name[1] == '\0' && check_nullness == 0 && | |
7008 | VALID_SPECIAL_LENGTH_PARAM (c) && string[sindex] == RBRACE) | |
7009 | { | |
7010 | name = (char *)xrealloc (name, 3); | |
7011 | name[1] = c; | |
7012 | name[2] = '\0'; | |
7013 | c = string[sindex++]; | |
7014 | } | |
7015 | ||
7016 | /* ...but ${#%}, ${#:}, ${#=}, ${#+}, and ${#/} are errors. */ | |
7017 | if (name[0] == '#' && name[1] == '\0' && check_nullness == 0 && | |
7018 | member (c, "%:=+/") && string[sindex] == RBRACE) | |
7019 | { | |
7020 | temp = (char *)NULL; | |
7021 | goto bad_substitution; | |
7022 | } | |
7023 | ||
7024 | /* Indirect expansion begins with a `!'. A valid indirect expansion is | |
7025 | either a variable name, one of the positional parameters or a special | |
7026 | variable that expands to one of the positional parameters. */ | |
7027 | want_indir = *name == '!' && | |
7028 | (legal_variable_starter ((unsigned char)name[1]) || DIGIT (name[1]) | |
7029 | || VALID_INDIR_PARAM (name[1])); | |
7030 | ||
7031 | /* Determine the value of this variable. */ | |
7032 | ||
7033 | /* Check for special variables, directly referenced. */ | |
7034 | if (SPECIAL_VAR (name, want_indir)) | |
7035 | var_is_special++; | |
7036 | ||
7037 | /* Check for special expansion things, like the length of a parameter */ | |
7038 | if (*name == '#' && name[1]) | |
7039 | { | |
7040 | /* If we are not pointing at the character just after the | |
7041 | closing brace, then we haven't gotten all of the name. | |
7042 | Since it begins with a special character, this is a bad | |
7043 | substitution. Also check NAME for validity before trying | |
7044 | to go on. */ | |
7045 | if (string[sindex - 1] != RBRACE || (valid_length_expression (name) == 0)) | |
7046 | { | |
7047 | temp = (char *)NULL; | |
7048 | goto bad_substitution; | |
7049 | } | |
7050 | ||
7051 | number = parameter_brace_expand_length (name); | |
7052 | if (number == INTMAX_MIN && unbound_vars_is_error) | |
7053 | { | |
7054 | last_command_exit_value = EXECUTION_FAILURE; | |
7055 | err_unboundvar (name+1); | |
7056 | free (name); | |
7057 | return (interactive_shell ? &expand_wdesc_error : &expand_wdesc_fatal); | |
7058 | } | |
7059 | free (name); | |
7060 | ||
7061 | *indexp = sindex; | |
7062 | if (number < 0) | |
7063 | return (&expand_wdesc_error); | |
7064 | else | |
7065 | { | |
7066 | ret = alloc_word_desc (); | |
7067 | ret->word = itos (number); | |
7068 | return ret; | |
7069 | } | |
7070 | } | |
7071 | ||
7072 | /* ${@} is identical to $@. */ | |
7073 | if (name[0] == '@' && name[1] == '\0') | |
7074 | { | |
7075 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && quoted_dollar_atp) | |
7076 | *quoted_dollar_atp = 1; | |
7077 | ||
7078 | if (contains_dollar_at) | |
7079 | *contains_dollar_at = 1; | |
7080 | } | |
7081 | ||
7082 | /* Process ${!PREFIX*} expansion. */ | |
7083 | if (want_indir && string[sindex - 1] == RBRACE && | |
7084 | (string[sindex - 2] == '*' || string[sindex - 2] == '@') && | |
7085 | legal_variable_starter ((unsigned char) name[1])) | |
7086 | { | |
7087 | char **x; | |
7088 | WORD_LIST *xlist; | |
7089 | ||
7090 | temp1 = savestring (name + 1); | |
7091 | number = strlen (temp1); | |
7092 | temp1[number - 1] = '\0'; | |
7093 | x = all_variables_matching_prefix (temp1); | |
7094 | xlist = strvec_to_word_list (x, 0, 0); | |
7095 | if (string[sindex - 2] == '*') | |
7096 | temp = string_list_dollar_star (xlist); | |
7097 | else | |
7098 | { | |
7099 | temp = string_list_dollar_at (xlist, quoted); | |
7100 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && quoted_dollar_atp) | |
7101 | *quoted_dollar_atp = 1; | |
7102 | if (contains_dollar_at) | |
7103 | *contains_dollar_at = 1; | |
7104 | } | |
7105 | free (x); | |
7106 | dispose_words (xlist); | |
7107 | free (temp1); | |
7108 | *indexp = sindex; | |
7109 | ||
7110 | ret = alloc_word_desc (); | |
7111 | ret->word = temp; | |
7112 | return ret; | |
7113 | } | |
7114 | ||
7115 | #if defined (ARRAY_VARS) | |
7116 | /* Process ${!ARRAY[@]} and ${!ARRAY[*]} expansion. */ /* [ */ | |
7117 | if (want_indir && string[sindex - 1] == RBRACE && | |
7118 | string[sindex - 2] == ']' && valid_array_reference (name+1)) | |
7119 | { | |
7120 | char *x, *x1; | |
7121 | ||
7122 | temp1 = savestring (name + 1); | |
7123 | x = array_variable_name (temp1, &x1, (int *)0); /* [ */ | |
7124 | FREE (x); | |
7125 | if (ALL_ELEMENT_SUB (x1[0]) && x1[1] == ']') | |
7126 | { | |
7127 | temp = array_keys (temp1, quoted); /* handles assoc vars too */ | |
7128 | if (x1[0] == '@') | |
7129 | { | |
7130 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && quoted_dollar_atp) | |
7131 | *quoted_dollar_atp = 1; | |
7132 | if (contains_dollar_at) | |
7133 | *contains_dollar_at = 1; | |
7134 | } | |
7135 | ||
7136 | free (temp1); | |
7137 | *indexp = sindex; | |
7138 | ||
7139 | ret = alloc_word_desc (); | |
7140 | ret->word = temp; | |
7141 | return ret; | |
7142 | } | |
7143 | ||
7144 | free (temp1); | |
7145 | } | |
7146 | #endif /* ARRAY_VARS */ | |
7147 | ||
7148 | /* Make sure that NAME is valid before trying to go on. */ | |
7149 | if (valid_brace_expansion_word (want_indir ? name + 1 : name, | |
7150 | var_is_special) == 0) | |
7151 | { | |
7152 | temp = (char *)NULL; | |
7153 | goto bad_substitution; | |
7154 | } | |
7155 | ||
7156 | if (want_indir) | |
7157 | tdesc = parameter_brace_expand_indir (name + 1, var_is_special, quoted, quoted_dollar_atp, contains_dollar_at); | |
7158 | else | |
7159 | tdesc = parameter_brace_expand_word (name, var_is_special, quoted, PF_IGNUNBOUND|(pflags&PF_NOSPLIT2), &ind); | |
7160 | ||
7161 | if (tdesc) | |
7162 | { | |
7163 | temp = tdesc->word; | |
7164 | tflag = tdesc->flags; | |
7165 | dispose_word_desc (tdesc); | |
7166 | } | |
7167 | else | |
7168 | temp = (char *)0; | |
7169 | ||
7170 | #if defined (ARRAY_VARS) | |
7171 | if (valid_array_reference (name)) | |
7172 | chk_atstar (name, quoted, quoted_dollar_atp, contains_dollar_at); | |
7173 | #endif | |
7174 | ||
7175 | var_is_set = temp != (char *)0; | |
7176 | var_is_null = check_nullness && (var_is_set == 0 || *temp == 0); | |
7177 | ||
7178 | /* Get the rest of the stuff inside the braces. */ | |
7179 | if (c && c != RBRACE) | |
7180 | { | |
9c2db999 CR |
7181 | /* Extract the contents of the ${ ... } expansion |
7182 | according to the Posix.2 rules. */ | |
7183 | value = extract_dollar_brace_string (string, &sindex, quoted, (c == '%' || c == '#' || c =='/' || c == '^' || c == ',' || c ==':') ? SX_POSIXEXP|SX_WORD : SX_WORD); | |
e192f341 CR |
7184 | if (string[sindex] == RBRACE) |
7185 | sindex++; | |
7186 | else | |
7187 | goto bad_substitution; | |
7188 | } | |
7189 | else | |
7190 | value = (char *)NULL; | |
7191 | ||
7192 | *indexp = sindex; | |
7193 | ||
7194 | /* All the cases where an expansion can possibly generate an unbound | |
7195 | variable error. */ | |
7196 | if (want_substring || want_patsub || want_casemod || c == '#' || c == '%' || c == RBRACE) | |
7197 | { | |
7198 | if (var_is_set == 0 && unbound_vars_is_error && ((name[0] != '@' && name[0] != '*') || name[1])) | |
7199 | { | |
7200 | last_command_exit_value = EXECUTION_FAILURE; | |
7201 | err_unboundvar (name); | |
7202 | FREE (value); | |
7203 | FREE (temp); | |
7204 | free (name); | |
7205 | return (interactive_shell ? &expand_wdesc_error : &expand_wdesc_fatal); | |
7206 | } | |
7207 | } | |
7208 | ||
7209 | /* If this is a substring spec, process it and add the result. */ | |
7210 | if (want_substring) | |
7211 | { | |
7212 | temp1 = parameter_brace_substring (name, temp, ind, value, quoted, (tflag & W_ARRAYIND) ? AV_USEIND : 0); | |
7213 | FREE (name); | |
7214 | FREE (value); | |
7215 | FREE (temp); | |
7216 | ||
7217 | if (temp1 == &expand_param_error) | |
7218 | return (&expand_wdesc_error); | |
7219 | else if (temp1 == &expand_param_fatal) | |
7220 | return (&expand_wdesc_fatal); | |
7221 | ||
7222 | ret = alloc_word_desc (); | |
7223 | ret->word = temp1; | |
7224 | if (temp1 && QUOTED_NULL (temp1) && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
7225 | ret->flags |= W_QUOTED|W_HASQUOTEDNULL; | |
7226 | return ret; | |
7227 | } | |
7228 | else if (want_patsub) | |
7229 | { | |
7230 | temp1 = parameter_brace_patsub (name, temp, ind, value, quoted, (tflag & W_ARRAYIND) ? AV_USEIND : 0); | |
7231 | FREE (name); | |
7232 | FREE (value); | |
7233 | FREE (temp); | |
7234 | ||
7235 | if (temp1 == &expand_param_error) | |
7236 | return (&expand_wdesc_error); | |
7237 | else if (temp1 == &expand_param_fatal) | |
7238 | return (&expand_wdesc_fatal); | |
7239 | ||
7240 | ret = alloc_word_desc (); | |
7241 | ret->word = temp1; | |
7242 | ret = alloc_word_desc (); | |
7243 | ret->word = temp1; | |
7244 | if (temp1 && QUOTED_NULL (temp1) && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
7245 | ret->flags |= W_QUOTED|W_HASQUOTEDNULL; | |
7246 | return ret; | |
7247 | } | |
7248 | #if defined (CASEMOD_EXPANSIONS) | |
7249 | else if (want_casemod) | |
7250 | { | |
7251 | temp1 = parameter_brace_casemod (name, temp, ind, modspec, value, quoted, (tflag & W_ARRAYIND) ? AV_USEIND : 0); | |
7252 | FREE (name); | |
7253 | FREE (value); | |
7254 | FREE (temp); | |
7255 | ||
7256 | if (temp1 == &expand_param_error) | |
7257 | return (&expand_wdesc_error); | |
7258 | else if (temp1 == &expand_param_fatal) | |
7259 | return (&expand_wdesc_fatal); | |
7260 | ||
7261 | ret = alloc_word_desc (); | |
7262 | ret->word = temp1; | |
7263 | if (temp1 && QUOTED_NULL (temp1) && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
7264 | ret->flags |= W_QUOTED|W_HASQUOTEDNULL; | |
7265 | return ret; | |
7266 | } | |
7267 | #endif | |
7268 | ||
7269 | /* Do the right thing based on which character ended the variable name. */ | |
7270 | switch (c) | |
7271 | { | |
7272 | default: | |
7273 | case '\0': | |
7274 | bad_substitution: | |
7275 | report_error (_("%s: bad substitution"), string ? string : "??"); | |
7276 | FREE (value); | |
7277 | FREE (temp); | |
7278 | free (name); | |
7279 | return &expand_wdesc_error; | |
7280 | ||
7281 | case RBRACE: | |
7282 | break; | |
7283 | ||
7284 | case '#': /* ${param#[#]pattern} */ | |
7285 | case '%': /* ${param%[%]pattern} */ | |
7286 | if (value == 0 || *value == '\0' || temp == 0 || *temp == '\0') | |
7287 | { | |
7288 | FREE (value); | |
7289 | break; | |
7290 | } | |
7291 | temp1 = parameter_brace_remove_pattern (name, temp, ind, value, c, quoted, (tflag & W_ARRAYIND) ? AV_USEIND : 0); | |
7292 | free (temp); | |
7293 | free (value); | |
7294 | free (name); | |
7295 | ||
7296 | ret = alloc_word_desc (); | |
7297 | ret->word = temp1; | |
7298 | if (temp1 && QUOTED_NULL (temp1) && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
7299 | ret->flags |= W_QUOTED|W_HASQUOTEDNULL; | |
7300 | return ret; | |
7301 | ||
7302 | case '-': | |
7303 | case '=': | |
7304 | case '?': | |
7305 | case '+': | |
7306 | if (var_is_set && var_is_null == 0) | |
7307 | { | |
7308 | /* If the operator is `+', we don't want the value of the named | |
7309 | variable for anything, just the value of the right hand side. */ | |
7310 | if (c == '+') | |
7311 | { | |
7312 | /* XXX -- if we're double-quoted and the named variable is "$@", | |
7313 | we want to turn off any special handling of "$@" -- | |
7314 | we're not using it, so whatever is on the rhs applies. */ | |
7315 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && quoted_dollar_atp) | |
7316 | *quoted_dollar_atp = 0; | |
7317 | if (contains_dollar_at) | |
7318 | *contains_dollar_at = 0; | |
7319 | ||
7320 | FREE (temp); | |
7321 | if (value) | |
7322 | { | |
7323 | /* From Posix discussion on austin-group list. Issue 221 | |
7324 | requires that backslashes escaping `}' inside | |
7325 | double-quoted ${...} be removed. */ | |
7326 | if (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
7327 | quoted |= Q_DOLBRACE; | |
7328 | ret = parameter_brace_expand_rhs (name, value, c, | |
7329 | quoted, | |
7330 | quoted_dollar_atp, | |
7331 | contains_dollar_at); | |
7332 | /* XXX - fix up later, esp. noting presence of | |
7333 | W_HASQUOTEDNULL in ret->flags */ | |
7334 | free (value); | |
7335 | } | |
7336 | else | |
7337 | temp = (char *)NULL; | |
7338 | } | |
7339 | else | |
7340 | { | |
7341 | FREE (value); | |
7342 | } | |
7343 | /* Otherwise do nothing; just use the value in TEMP. */ | |
7344 | } | |
7345 | else /* VAR not set or VAR is NULL. */ | |
7346 | { | |
7347 | FREE (temp); | |
7348 | temp = (char *)NULL; | |
7349 | if (c == '=' && var_is_special) | |
7350 | { | |
7351 | report_error (_("$%s: cannot assign in this way"), name); | |
7352 | free (name); | |
7353 | free (value); | |
7354 | return &expand_wdesc_error; | |
7355 | } | |
7356 | else if (c == '?') | |
7357 | { | |
7358 | parameter_brace_expand_error (name, value); | |
7359 | return (interactive_shell ? &expand_wdesc_error : &expand_wdesc_fatal); | |
7360 | } | |
7361 | else if (c != '+') | |
7362 | { | |
7363 | /* XXX -- if we're double-quoted and the named variable is "$@", | |
7364 | we want to turn off any special handling of "$@" -- | |
7365 | we're not using it, so whatever is on the rhs applies. */ | |
7366 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && quoted_dollar_atp) | |
7367 | *quoted_dollar_atp = 0; | |
7368 | if (contains_dollar_at) | |
7369 | *contains_dollar_at = 0; | |
7370 | ||
7371 | /* From Posix discussion on austin-group list. Issue 221 requires | |
7372 | that backslashes escaping `}' inside double-quoted ${...} be | |
7373 | removed. */ | |
7374 | if (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
7375 | quoted |= Q_DOLBRACE; | |
7376 | ret = parameter_brace_expand_rhs (name, value, c, quoted, | |
7377 | quoted_dollar_atp, | |
7378 | contains_dollar_at); | |
7379 | /* XXX - fix up later, esp. noting presence of | |
7380 | W_HASQUOTEDNULL in tdesc->flags */ | |
7381 | } | |
7382 | free (value); | |
7383 | } | |
7384 | ||
7385 | break; | |
7386 | } | |
7387 | free (name); | |
7388 | ||
7389 | if (ret == 0) | |
7390 | { | |
7391 | ret = alloc_word_desc (); | |
7392 | ret->flags = tflag; | |
7393 | ret->word = temp; | |
7394 | } | |
7395 | return (ret); | |
7396 | } | |
7397 | ||
7398 | /* Expand a single ${xxx} expansion. The braces are optional. When | |
7399 | the braces are used, parameter_brace_expand() does the work, | |
7400 | possibly calling param_expand recursively. */ | |
7401 | static WORD_DESC * | |
7402 | param_expand (string, sindex, quoted, expanded_something, | |
7403 | contains_dollar_at, quoted_dollar_at_p, had_quoted_null_p, | |
7404 | pflags) | |
7405 | char *string; | |
7406 | int *sindex, quoted, *expanded_something, *contains_dollar_at; | |
7407 | int *quoted_dollar_at_p, *had_quoted_null_p, pflags; | |
7408 | { | |
7409 | char *temp, *temp1, uerror[3]; | |
7410 | int zindex, t_index, expok; | |
7411 | unsigned char c; | |
7412 | intmax_t number; | |
7413 | SHELL_VAR *var; | |
7414 | WORD_LIST *list; | |
7415 | WORD_DESC *tdesc, *ret; | |
7416 | int tflag; | |
7417 | ||
7418 | zindex = *sindex; | |
7419 | c = string[++zindex]; | |
7420 | ||
7421 | temp = (char *)NULL; | |
7422 | ret = tdesc = (WORD_DESC *)NULL; | |
7423 | tflag = 0; | |
7424 | ||
7425 | /* Do simple cases first. Switch on what follows '$'. */ | |
7426 | switch (c) | |
7427 | { | |
7428 | /* $0 .. $9? */ | |
7429 | case '0': | |
7430 | case '1': | |
7431 | case '2': | |
7432 | case '3': | |
7433 | case '4': | |
7434 | case '5': | |
7435 | case '6': | |
7436 | case '7': | |
7437 | case '8': | |
7438 | case '9': | |
7439 | temp1 = dollar_vars[TODIGIT (c)]; | |
7440 | if (unbound_vars_is_error && temp1 == (char *)NULL) | |
7441 | { | |
7442 | uerror[0] = '$'; | |
7443 | uerror[1] = c; | |
7444 | uerror[2] = '\0'; | |
7445 | last_command_exit_value = EXECUTION_FAILURE; | |
7446 | err_unboundvar (uerror); | |
7447 | return (interactive_shell ? &expand_wdesc_error : &expand_wdesc_fatal); | |
7448 | } | |
7449 | if (temp1) | |
7450 | temp = (*temp1 && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
7451 | ? quote_string (temp1) | |
7452 | : quote_escapes (temp1); | |
7453 | else | |
7454 | temp = (char *)NULL; | |
7455 | ||
7456 | break; | |
7457 | ||
7458 | /* $$ -- pid of the invoking shell. */ | |
7459 | case '$': | |
7460 | temp = itos (dollar_dollar_pid); | |
7461 | break; | |
7462 | ||
7463 | /* $# -- number of positional parameters. */ | |
7464 | case '#': | |
7465 | temp = itos (number_of_args ()); | |
7466 | break; | |
7467 | ||
7468 | /* $? -- return value of the last synchronous command. */ | |
7469 | case '?': | |
7470 | temp = itos (last_command_exit_value); | |
7471 | break; | |
7472 | ||
7473 | /* $- -- flags supplied to the shell on invocation or by `set'. */ | |
7474 | case '-': | |
7475 | temp = which_set_flags (); | |
7476 | break; | |
7477 | ||
7478 | /* $! -- Pid of the last asynchronous command. */ | |
7479 | case '!': | |
7480 | /* If no asynchronous pids have been created, expand to nothing. | |
7481 | If `set -u' has been executed, and no async processes have | |
7482 | been created, this is an expansion error. */ | |
7483 | if (last_asynchronous_pid == NO_PID) | |
7484 | { | |
7485 | if (expanded_something) | |
7486 | *expanded_something = 0; | |
7487 | temp = (char *)NULL; | |
7488 | if (unbound_vars_is_error) | |
7489 | { | |
7490 | uerror[0] = '$'; | |
7491 | uerror[1] = c; | |
7492 | uerror[2] = '\0'; | |
7493 | last_command_exit_value = EXECUTION_FAILURE; | |
7494 | err_unboundvar (uerror); | |
7495 | return (interactive_shell ? &expand_wdesc_error : &expand_wdesc_fatal); | |
7496 | } | |
7497 | } | |
7498 | else | |
7499 | temp = itos (last_asynchronous_pid); | |
7500 | break; | |
7501 | ||
7502 | /* The only difference between this and $@ is when the arg is quoted. */ | |
7503 | case '*': /* `$*' */ | |
7504 | list = list_rest_of_args (); | |
7505 | ||
7506 | #if 0 | |
7507 | /* According to austin-group posix proposal by Geoff Clare in | |
7508 | <20090505091501.GA10097@squonk.masqnet> of 5 May 2009: | |
7509 | ||
7510 | "The shell shall write a message to standard error and | |
7511 | immediately exit when it tries to expand an unset parameter | |
7512 | other than the '@' and '*' special parameters." | |
7513 | */ | |
7514 | ||
7515 | if (list == 0 && unbound_vars_is_error && (pflags & PF_IGNUNBOUND) == 0) | |
7516 | { | |
7517 | uerror[0] = '$'; | |
7518 | uerror[1] = '*'; | |
7519 | uerror[2] = '\0'; | |
7520 | last_command_exit_value = EXECUTION_FAILURE; | |
7521 | err_unboundvar (uerror); | |
7522 | return (interactive_shell ? &expand_wdesc_error : &expand_wdesc_fatal); | |
7523 | } | |
7524 | #endif | |
7525 | ||
7526 | /* If there are no command-line arguments, this should just | |
7527 | disappear if there are other characters in the expansion, | |
7528 | even if it's quoted. */ | |
7529 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && list == 0) | |
7530 | temp = (char *)NULL; | |
7531 | else if (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES|Q_PATQUOTE)) | |
7532 | { | |
7533 | /* If we have "$*" we want to make a string of the positional | |
7534 | parameters, separated by the first character of $IFS, and | |
7535 | quote the whole string, including the separators. If IFS | |
7536 | is unset, the parameters are separated by ' '; if $IFS is | |
7537 | null, the parameters are concatenated. */ | |
7538 | temp = (quoted & (Q_DOUBLE_QUOTES|Q_PATQUOTE)) ? string_list_dollar_star (list) : string_list (list); | |
7539 | if (temp) | |
7540 | { | |
7541 | temp1 = quote_string (temp); | |
7542 | if (*temp == 0) | |
7543 | tflag |= W_HASQUOTEDNULL; | |
7544 | free (temp); | |
7545 | temp = temp1; | |
7546 | } | |
7547 | } | |
7548 | else | |
7549 | { | |
7550 | /* We check whether or not we're eventually going to split $* here, | |
7551 | for example when IFS is empty and we are processing the rhs of | |
7552 | an assignment statement. In that case, we don't separate the | |
7553 | arguments at all. Otherwise, if the $* is not quoted it is | |
7554 | identical to $@ */ | |
7555 | #if 1 | |
7556 | # if defined (HANDLE_MULTIBYTE) | |
7557 | if (expand_no_split_dollar_star && ifs_firstc[0] == 0) | |
7558 | # else | |
7559 | if (expand_no_split_dollar_star && ifs_firstc == 0) | |
7560 | # endif | |
7561 | temp = string_list_dollar_star (list); | |
7562 | else | |
7563 | temp = string_list_dollar_at (list, quoted); | |
7564 | #else | |
7565 | temp = string_list_dollar_at (list, quoted); | |
7566 | #endif | |
7567 | if (expand_no_split_dollar_star == 0 && contains_dollar_at) | |
7568 | *contains_dollar_at = 1; | |
7569 | } | |
7570 | ||
7571 | dispose_words (list); | |
7572 | break; | |
7573 | ||
7574 | /* When we have "$@" what we want is "$1" "$2" "$3" ... This | |
7575 | means that we have to turn quoting off after we split into | |
7576 | the individually quoted arguments so that the final split | |
7577 | on the first character of $IFS is still done. */ | |
7578 | case '@': /* `$@' */ | |
7579 | list = list_rest_of_args (); | |
7580 | ||
7581 | #if 0 | |
7582 | /* According to austin-group posix proposal by Geoff Clare in | |
7583 | <20090505091501.GA10097@squonk.masqnet> of 5 May 2009: | |
7584 | ||
7585 | "The shell shall write a message to standard error and | |
7586 | immediately exit when it tries to expand an unset parameter | |
7587 | other than the '@' and '*' special parameters." | |
7588 | */ | |
7589 | ||
7590 | if (list == 0 && unbound_vars_is_error && (pflags & PF_IGNUNBOUND) == 0) | |
7591 | { | |
7592 | uerror[0] = '$'; | |
7593 | uerror[1] = '@'; | |
7594 | uerror[2] = '\0'; | |
7595 | last_command_exit_value = EXECUTION_FAILURE; | |
7596 | err_unboundvar (uerror); | |
7597 | return (interactive_shell ? &expand_wdesc_error : &expand_wdesc_fatal); | |
7598 | } | |
7599 | #endif | |
7600 | ||
7601 | /* We want to flag the fact that we saw this. We can't turn | |
7602 | off quoting entirely, because other characters in the | |
7603 | string might need it (consider "\"$@\""), but we need some | |
7604 | way to signal that the final split on the first character | |
7605 | of $IFS should be done, even though QUOTED is 1. */ | |
7606 | /* XXX - should this test include Q_PATQUOTE? */ | |
7607 | if (quoted_dollar_at_p && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
7608 | *quoted_dollar_at_p = 1; | |
7609 | if (contains_dollar_at) | |
7610 | *contains_dollar_at = 1; | |
7611 | ||
7612 | #if 0 | |
7613 | if (pflags & PF_NOSPLIT2) | |
7614 | temp = string_list_internal (quoted ? quote_list (list) : list, " "); | |
7615 | else | |
7616 | #endif | |
7617 | /* We want to separate the positional parameters with the first | |
7618 | character of $IFS in case $IFS is something other than a space. | |
7619 | We also want to make sure that splitting is done no matter what -- | |
7620 | according to POSIX.2, this expands to a list of the positional | |
7621 | parameters no matter what IFS is set to. */ | |
7622 | temp = string_list_dollar_at (list, quoted); | |
7623 | ||
7624 | dispose_words (list); | |
7625 | break; | |
7626 | ||
7627 | case LBRACE: | |
7628 | tdesc = parameter_brace_expand (string, &zindex, quoted, pflags, | |
7629 | quoted_dollar_at_p, | |
7630 | contains_dollar_at); | |
7631 | ||
7632 | if (tdesc == &expand_wdesc_error || tdesc == &expand_wdesc_fatal) | |
7633 | return (tdesc); | |
7634 | temp = tdesc ? tdesc->word : (char *)0; | |
7635 | ||
7636 | /* XXX */ | |
7637 | /* Quoted nulls should be removed if there is anything else | |
7638 | in the string. */ | |
7639 | /* Note that we saw the quoted null so we can add one back at | |
7640 | the end of this function if there are no other characters | |
7641 | in the string, discard TEMP, and go on. The exception to | |
7642 | this is when we have "${@}" and $1 is '', since $@ needs | |
7643 | special handling. */ | |
7644 | if (tdesc && tdesc->word && (tdesc->flags & W_HASQUOTEDNULL) && QUOTED_NULL (temp)) | |
7645 | { | |
7646 | if (had_quoted_null_p) | |
7647 | *had_quoted_null_p = 1; | |
7648 | if (*quoted_dollar_at_p == 0) | |
7649 | { | |
7650 | free (temp); | |
7651 | tdesc->word = temp = (char *)NULL; | |
7652 | } | |
7653 | ||
7654 | } | |
7655 | ||
7656 | ret = tdesc; | |
7657 | goto return0; | |
7658 | ||
7659 | /* Do command or arithmetic substitution. */ | |
7660 | case LPAREN: | |
7661 | /* We have to extract the contents of this paren substitution. */ | |
7662 | t_index = zindex + 1; | |
7663 | temp = extract_command_subst (string, &t_index, 0); | |
7664 | zindex = t_index; | |
7665 | ||
7666 | /* For Posix.2-style `$(( ))' arithmetic substitution, | |
7667 | extract the expression and pass it to the evaluator. */ | |
7668 | if (temp && *temp == LPAREN) | |
7669 | { | |
7670 | char *temp2; | |
7671 | temp1 = temp + 1; | |
7672 | temp2 = savestring (temp1); | |
7673 | t_index = strlen (temp2) - 1; | |
7674 | ||
7675 | if (temp2[t_index] != RPAREN) | |
7676 | { | |
7677 | free (temp2); | |
7678 | goto comsub; | |
7679 | } | |
7680 | ||
7681 | /* Cut off ending `)' */ | |
7682 | temp2[t_index] = '\0'; | |
7683 | ||
7684 | if (chk_arithsub (temp2, t_index) == 0) | |
7685 | { | |
7686 | free (temp2); | |
7687 | #if 0 | |
7688 | internal_warning (_("future versions of the shell will force evaluation as an arithmetic substitution")); | |
7689 | #endif | |
7690 | goto comsub; | |
7691 | } | |
7692 | ||
7693 | /* Expand variables found inside the expression. */ | |
7694 | temp1 = expand_arith_string (temp2, Q_DOUBLE_QUOTES); | |
7695 | free (temp2); | |
7696 | ||
7697 | arithsub: | |
7698 | /* No error messages. */ | |
7699 | this_command_name = (char *)NULL; | |
7700 | number = evalexp (temp1, &expok); | |
7701 | free (temp); | |
7702 | free (temp1); | |
7703 | if (expok == 0) | |
7704 | { | |
7705 | if (interactive_shell == 0 && posixly_correct) | |
7706 | { | |
7707 | last_command_exit_value = EXECUTION_FAILURE; | |
7708 | return (&expand_wdesc_fatal); | |
7709 | } | |
7710 | else | |
7711 | return (&expand_wdesc_error); | |
7712 | } | |
7713 | temp = itos (number); | |
7714 | break; | |
7715 | } | |
7716 | ||
7717 | comsub: | |
7718 | if (pflags & PF_NOCOMSUB) | |
7719 | /* we need zindex+1 because string[zindex] == RPAREN */ | |
7720 | temp1 = substring (string, *sindex, zindex+1); | |
7721 | else | |
7722 | { | |
7723 | tdesc = command_substitute (temp, quoted); | |
7724 | temp1 = tdesc ? tdesc->word : (char *)NULL; | |
7725 | if (tdesc) | |
7726 | dispose_word_desc (tdesc); | |
7727 | } | |
7728 | FREE (temp); | |
7729 | temp = temp1; | |
7730 | break; | |
7731 | ||
7732 | /* Do POSIX.2d9-style arithmetic substitution. This will probably go | |
7733 | away in a future bash release. */ | |
7734 | case '[': | |
7735 | /* Extract the contents of this arithmetic substitution. */ | |
7736 | t_index = zindex + 1; | |
7737 | temp = extract_arithmetic_subst (string, &t_index); | |
7738 | zindex = t_index; | |
7739 | if (temp == 0) | |
7740 | { | |
7741 | temp = savestring (string); | |
7742 | if (expanded_something) | |
7743 | *expanded_something = 0; | |
7744 | goto return0; | |
7745 | } | |
7746 | ||
7747 | /* Do initial variable expansion. */ | |
7748 | temp1 = expand_arith_string (temp, Q_DOUBLE_QUOTES); | |
7749 | ||
7750 | goto arithsub; | |
7751 | ||
7752 | default: | |
7753 | /* Find the variable in VARIABLE_LIST. */ | |
7754 | temp = (char *)NULL; | |
7755 | ||
7756 | for (t_index = zindex; (c = string[zindex]) && legal_variable_char (c); zindex++) | |
7757 | ; | |
7758 | temp1 = (zindex > t_index) ? substring (string, t_index, zindex) : (char *)NULL; | |
7759 | ||
7760 | /* If this isn't a variable name, then just output the `$'. */ | |
7761 | if (temp1 == 0 || *temp1 == '\0') | |
7762 | { | |
7763 | FREE (temp1); | |
7764 | temp = (char *)xmalloc (2); | |
7765 | temp[0] = '$'; | |
7766 | temp[1] = '\0'; | |
7767 | if (expanded_something) | |
7768 | *expanded_something = 0; | |
7769 | goto return0; | |
7770 | } | |
7771 | ||
7772 | /* If the variable exists, return its value cell. */ | |
7773 | var = find_variable (temp1); | |
7774 | ||
7775 | if (var && invisible_p (var) == 0 && var_isset (var)) | |
7776 | { | |
7777 | #if defined (ARRAY_VARS) | |
7778 | if (assoc_p (var) || array_p (var)) | |
7779 | { | |
7780 | temp = array_p (var) ? array_reference (array_cell (var), 0) | |
7781 | : assoc_reference (assoc_cell (var), "0"); | |
7782 | if (temp) | |
7783 | temp = (*temp && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
7784 | ? quote_string (temp) | |
7785 | : quote_escapes (temp); | |
7786 | else if (unbound_vars_is_error) | |
7787 | goto unbound_variable; | |
7788 | } | |
7789 | else | |
7790 | #endif | |
7791 | { | |
7792 | temp = value_cell (var); | |
7793 | ||
7794 | temp = (*temp && (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES))) | |
7795 | ? quote_string (temp) | |
7796 | : quote_escapes (temp); | |
7797 | } | |
7798 | ||
7799 | free (temp1); | |
7800 | ||
7801 | goto return0; | |
7802 | } | |
7803 | ||
7804 | temp = (char *)NULL; | |
7805 | ||
7806 | unbound_variable: | |
7807 | if (unbound_vars_is_error) | |
7808 | { | |
7809 | last_command_exit_value = EXECUTION_FAILURE; | |
7810 | err_unboundvar (temp1); | |
7811 | } | |
7812 | else | |
7813 | { | |
7814 | free (temp1); | |
7815 | goto return0; | |
7816 | } | |
7817 | ||
7818 | free (temp1); | |
7819 | last_command_exit_value = EXECUTION_FAILURE; | |
7820 | return ((unbound_vars_is_error && interactive_shell == 0) | |
7821 | ? &expand_wdesc_fatal | |
7822 | : &expand_wdesc_error); | |
7823 | } | |
7824 | ||
7825 | if (string[zindex]) | |
7826 | zindex++; | |
7827 | ||
7828 | return0: | |
7829 | *sindex = zindex; | |
7830 | ||
7831 | if (ret == 0) | |
7832 | { | |
7833 | ret = alloc_word_desc (); | |
7834 | ret->flags = tflag; /* XXX */ | |
7835 | ret->word = temp; | |
7836 | } | |
7837 | return ret; | |
7838 | } | |
7839 | ||
7840 | /* Make a word list which is the result of parameter and variable | |
7841 | expansion, command substitution, arithmetic substitution, and | |
7842 | quote removal of WORD. Return a pointer to a WORD_LIST which is | |
7843 | the result of the expansion. If WORD contains a null word, the | |
7844 | word list returned is also null. | |
7845 | ||
7846 | QUOTED contains flag values defined in shell.h. | |
7847 | ||
7848 | ISEXP is used to tell expand_word_internal that the word should be | |
7849 | treated as the result of an expansion. This has implications for | |
7850 | how IFS characters in the word are treated. | |
7851 | ||
7852 | CONTAINS_DOLLAR_AT and EXPANDED_SOMETHING are return values; when non-null | |
7853 | they point to an integer value which receives information about expansion. | |
7854 | CONTAINS_DOLLAR_AT gets non-zero if WORD contained "$@", else zero. | |
7855 | EXPANDED_SOMETHING get non-zero if WORD contained any parameter expansions, | |
7856 | else zero. | |
7857 | ||
7858 | This only does word splitting in the case of $@ expansion. In that | |
7859 | case, we split on ' '. */ | |
7860 | ||
7861 | /* Values for the local variable quoted_state. */ | |
7862 | #define UNQUOTED 0 | |
7863 | #define PARTIALLY_QUOTED 1 | |
7864 | #define WHOLLY_QUOTED 2 | |
7865 | ||
7866 | static WORD_LIST * | |
7867 | expand_word_internal (word, quoted, isexp, contains_dollar_at, expanded_something) | |
7868 | WORD_DESC *word; | |
7869 | int quoted, isexp; | |
7870 | int *contains_dollar_at; | |
7871 | int *expanded_something; | |
7872 | { | |
7873 | WORD_LIST *list; | |
7874 | WORD_DESC *tword; | |
7875 | ||
7876 | /* The intermediate string that we build while expanding. */ | |
7877 | char *istring; | |
7878 | ||
7879 | /* The current size of the above object. */ | |
7880 | int istring_size; | |
7881 | ||
7882 | /* Index into ISTRING. */ | |
7883 | int istring_index; | |
7884 | ||
7885 | /* Temporary string storage. */ | |
7886 | char *temp, *temp1; | |
7887 | ||
7888 | /* The text of WORD. */ | |
7889 | register char *string; | |
7890 | ||
7891 | /* The size of STRING. */ | |
7892 | size_t string_size; | |
7893 | ||
7894 | /* The index into STRING. */ | |
7895 | int sindex; | |
7896 | ||
7897 | /* This gets 1 if we see a $@ while quoted. */ | |
7898 | int quoted_dollar_at; | |
7899 | ||
7900 | /* One of UNQUOTED, PARTIALLY_QUOTED, or WHOLLY_QUOTED, depending on | |
7901 | whether WORD contains no quoting characters, a partially quoted | |
7902 | string (e.g., "xx"ab), or is fully quoted (e.g., "xxab"). */ | |
7903 | int quoted_state; | |
7904 | ||
7905 | /* State flags */ | |
7906 | int had_quoted_null; | |
7907 | int has_dollar_at; | |
7908 | int tflag; | |
7909 | int pflags; /* flags passed to param_expand */ | |
7910 | ||
7911 | int assignoff; /* If assignment, offset of `=' */ | |
7912 | ||
7913 | register unsigned char c; /* Current character. */ | |
7914 | int t_index; /* For calls to string_extract_xxx. */ | |
7915 | ||
7916 | char twochars[2]; | |
7917 | ||
7918 | DECLARE_MBSTATE; | |
7919 | ||
7920 | istring = (char *)xmalloc (istring_size = DEFAULT_INITIAL_ARRAY_SIZE); | |
7921 | istring[istring_index = 0] = '\0'; | |
7922 | quoted_dollar_at = had_quoted_null = has_dollar_at = 0; | |
7923 | quoted_state = UNQUOTED; | |
7924 | ||
7925 | string = word->word; | |
7926 | if (string == 0) | |
7927 | goto finished_with_string; | |
7928 | /* Don't need the string length for the SADD... and COPY_ macros unless | |
7929 | multibyte characters are possible. */ | |
7930 | string_size = (MB_CUR_MAX > 1) ? strlen (string) : 1; | |
7931 | ||
7932 | if (contains_dollar_at) | |
7933 | *contains_dollar_at = 0; | |
7934 | ||
7935 | assignoff = -1; | |
7936 | ||
7937 | /* Begin the expansion. */ | |
7938 | ||
7939 | for (sindex = 0; ;) | |
7940 | { | |
7941 | c = string[sindex]; | |
7942 | ||
7943 | /* Case on toplevel character. */ | |
7944 | switch (c) | |
7945 | { | |
7946 | case '\0': | |
7947 | goto finished_with_string; | |
7948 | ||
7949 | case CTLESC: | |
7950 | sindex++; | |
7951 | #if HANDLE_MULTIBYTE | |
7952 | if (MB_CUR_MAX > 1 && string[sindex]) | |
7953 | { | |
7954 | SADD_MBQCHAR_BODY(temp, string, sindex, string_size); | |
7955 | } | |
7956 | else | |
7957 | #endif | |
7958 | { | |
7959 | temp = (char *)xmalloc (3); | |
7960 | temp[0] = CTLESC; | |
7961 | temp[1] = c = string[sindex]; | |
7962 | temp[2] = '\0'; | |
7963 | } | |
7964 | ||
7965 | dollar_add_string: | |
7966 | if (string[sindex]) | |
7967 | sindex++; | |
7968 | ||
7969 | add_string: | |
7970 | if (temp) | |
7971 | { | |
7972 | istring = sub_append_string (temp, istring, &istring_index, &istring_size); | |
7973 | temp = (char *)0; | |
7974 | } | |
7975 | ||
7976 | break; | |
7977 | ||
7978 | #if defined (PROCESS_SUBSTITUTION) | |
7979 | /* Process substitution. */ | |
7980 | case '<': | |
7981 | case '>': | |
7982 | { | |
7983 | if (string[++sindex] != LPAREN || (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) || (word->flags & (W_DQUOTE|W_NOPROCSUB)) || posixly_correct) | |
7984 | { | |
7985 | sindex--; /* add_character: label increments sindex */ | |
7986 | goto add_character; | |
7987 | } | |
7988 | else | |
7989 | t_index = sindex + 1; /* skip past both '<' and LPAREN */ | |
7990 | ||
7991 | temp1 = extract_process_subst (string, (c == '<') ? "<(" : ">(", &t_index); /*))*/ | |
7992 | sindex = t_index; | |
7993 | ||
7994 | /* If the process substitution specification is `<()', we want to | |
7995 | open the pipe for writing in the child and produce output; if | |
7996 | it is `>()', we want to open the pipe for reading in the child | |
7997 | and consume input. */ | |
7998 | temp = temp1 ? process_substitute (temp1, (c == '>')) : (char *)0; | |
7999 | ||
8000 | FREE (temp1); | |
8001 | ||
8002 | goto dollar_add_string; | |
8003 | } | |
8004 | #endif /* PROCESS_SUBSTITUTION */ | |
8005 | ||
8006 | case '=': | |
8007 | /* Posix.2 section 3.6.1 says that tildes following `=' in words | |
8008 | which are not assignment statements are not expanded. If the | |
8009 | shell isn't in posix mode, though, we perform tilde expansion | |
8010 | on `likely candidate' unquoted assignment statements (flags | |
8011 | include W_ASSIGNMENT but not W_QUOTED). A likely candidate | |
8012 | contains an unquoted :~ or =~. Something to think about: we | |
8013 | now have a flag that says to perform tilde expansion on arguments | |
8014 | to `assignment builtins' like declare and export that look like | |
8015 | assignment statements. We now do tilde expansion on such words | |
8016 | even in POSIX mode. */ | |
8017 | if (word->flags & (W_ASSIGNRHS|W_NOTILDE)) | |
8018 | { | |
8019 | if (isexp == 0 && (word->flags & (W_NOSPLIT|W_NOSPLIT2)) == 0 && isifs (c)) | |
8020 | goto add_ifs_character; | |
8021 | else | |
8022 | goto add_character; | |
8023 | } | |
8024 | /* If we're not in posix mode or forcing assignment-statement tilde | |
8025 | expansion, note where the `=' appears in the word and prepare to | |
8026 | do tilde expansion following the first `='. */ | |
8027 | if ((word->flags & W_ASSIGNMENT) && | |
8028 | (posixly_correct == 0 || (word->flags & W_TILDEEXP)) && | |
8029 | assignoff == -1 && sindex > 0) | |
8030 | assignoff = sindex; | |
8031 | if (sindex == assignoff && string[sindex+1] == '~') /* XXX */ | |
8032 | word->flags |= W_ITILDE; | |
8033 | #if 0 | |
8034 | else if ((word->flags & W_ASSIGNMENT) && | |
8035 | (posixly_correct == 0 || (word->flags & W_TILDEEXP)) && | |
8036 | string[sindex+1] == '~') | |
8037 | word->flags |= W_ITILDE; | |
8038 | #endif | |
8039 | if (isexp == 0 && (word->flags & (W_NOSPLIT|W_NOSPLIT2)) == 0 && isifs (c)) | |
8040 | goto add_ifs_character; | |
8041 | else | |
8042 | goto add_character; | |
8043 | ||
8044 | case ':': | |
8045 | if (word->flags & W_NOTILDE) | |
8046 | { | |
8047 | if (isexp == 0 && (word->flags & (W_NOSPLIT|W_NOSPLIT2)) == 0 && isifs (c)) | |
8048 | goto add_ifs_character; | |
8049 | else | |
8050 | goto add_character; | |
8051 | } | |
8052 | ||
8053 | if ((word->flags & (W_ASSIGNMENT|W_ASSIGNRHS|W_TILDEEXP)) && | |
8054 | string[sindex+1] == '~') | |
8055 | word->flags |= W_ITILDE; | |
8056 | ||
8057 | if (isexp == 0 && (word->flags & (W_NOSPLIT|W_NOSPLIT2)) == 0 && isifs (c)) | |
8058 | goto add_ifs_character; | |
8059 | else | |
8060 | goto add_character; | |
8061 | ||
8062 | case '~': | |
8063 | /* If the word isn't supposed to be tilde expanded, or we're not | |
8064 | at the start of a word or after an unquoted : or = in an | |
8065 | assignment statement, we don't do tilde expansion. */ | |
8066 | if ((word->flags & (W_NOTILDE|W_DQUOTE)) || | |
8067 | (sindex > 0 && ((word->flags & W_ITILDE) == 0)) || | |
8068 | (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT))) | |
8069 | { | |
8070 | word->flags &= ~W_ITILDE; | |
8071 | if (isexp == 0 && (word->flags & (W_NOSPLIT|W_NOSPLIT2)) == 0 && isifs (c) && (quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) == 0) | |
8072 | goto add_ifs_character; | |
8073 | else | |
8074 | goto add_character; | |
8075 | } | |
8076 | ||
8077 | if (word->flags & W_ASSIGNRHS) | |
8078 | tflag = 2; | |
8079 | else if (word->flags & (W_ASSIGNMENT|W_TILDEEXP)) | |
8080 | tflag = 1; | |
8081 | else | |
8082 | tflag = 0; | |
8083 | ||
8084 | temp = bash_tilde_find_word (string + sindex, tflag, &t_index); | |
8085 | ||
8086 | word->flags &= ~W_ITILDE; | |
8087 | ||
8088 | if (temp && *temp && t_index > 0) | |
8089 | { | |
8090 | temp1 = bash_tilde_expand (temp, tflag); | |
8091 | if (temp1 && *temp1 == '~' && STREQ (temp, temp1)) | |
8092 | { | |
8093 | FREE (temp); | |
8094 | FREE (temp1); | |
8095 | goto add_character; /* tilde expansion failed */ | |
8096 | } | |
8097 | free (temp); | |
8098 | temp = temp1; | |
8099 | sindex += t_index; | |
8100 | goto add_quoted_string; /* XXX was add_string */ | |
8101 | } | |
8102 | else | |
8103 | { | |
8104 | FREE (temp); | |
8105 | goto add_character; | |
8106 | } | |
8107 | ||
8108 | case '$': | |
8109 | if (expanded_something) | |
8110 | *expanded_something = 1; | |
8111 | ||
8112 | has_dollar_at = 0; | |
8113 | pflags = (word->flags & W_NOCOMSUB) ? PF_NOCOMSUB : 0; | |
8114 | if (word->flags & W_NOSPLIT2) | |
8115 | pflags |= PF_NOSPLIT2; | |
8116 | tword = param_expand (string, &sindex, quoted, expanded_something, | |
8117 | &has_dollar_at, "ed_dollar_at, | |
8118 | &had_quoted_null, pflags); | |
8119 | ||
8120 | if (tword == &expand_wdesc_error || tword == &expand_wdesc_fatal) | |
8121 | { | |
8122 | free (string); | |
8123 | free (istring); | |
8124 | return ((tword == &expand_wdesc_error) ? &expand_word_error | |
8125 | : &expand_word_fatal); | |
8126 | } | |
8127 | if (contains_dollar_at && has_dollar_at) | |
8128 | *contains_dollar_at = 1; | |
8129 | ||
8130 | if (tword && (tword->flags & W_HASQUOTEDNULL)) | |
8131 | had_quoted_null = 1; | |
8132 | ||
8133 | temp = tword->word; | |
8134 | dispose_word_desc (tword); | |
8135 | ||
8136 | goto add_string; | |
8137 | break; | |
8138 | ||
8139 | case '`': /* Backquoted command substitution. */ | |
8140 | { | |
8141 | t_index = sindex++; | |
8142 | ||
8143 | temp = string_extract (string, &sindex, "`", SX_REQMATCH); | |
8144 | /* The test of sindex against t_index is to allow bare instances of | |
8145 | ` to pass through, for backwards compatibility. */ | |
8146 | if (temp == &extract_string_error || temp == &extract_string_fatal) | |
8147 | { | |
8148 | if (sindex - 1 == t_index) | |
8149 | { | |
8150 | sindex = t_index; | |
8151 | goto add_character; | |
8152 | } | |
8153 | report_error (_("bad substitution: no closing \"`\" in %s") , string+t_index); | |
8154 | free (string); | |
8155 | free (istring); | |
8156 | return ((temp == &extract_string_error) ? &expand_word_error | |
8157 | : &expand_word_fatal); | |
8158 | } | |
8159 | ||
8160 | if (expanded_something) | |
8161 | *expanded_something = 1; | |
8162 | ||
8163 | if (word->flags & W_NOCOMSUB) | |
8164 | /* sindex + 1 because string[sindex] == '`' */ | |
8165 | temp1 = substring (string, t_index, sindex + 1); | |
8166 | else | |
8167 | { | |
8168 | de_backslash (temp); | |
8169 | tword = command_substitute (temp, quoted); | |
8170 | temp1 = tword ? tword->word : (char *)NULL; | |
8171 | if (tword) | |
8172 | dispose_word_desc (tword); | |
8173 | } | |
8174 | FREE (temp); | |
8175 | temp = temp1; | |
8176 | goto dollar_add_string; | |
8177 | } | |
8178 | ||
8179 | case '\\': | |
8180 | if (string[sindex + 1] == '\n') | |
8181 | { | |
8182 | sindex += 2; | |
8183 | continue; | |
8184 | } | |
8185 | ||
8186 | c = string[++sindex]; | |
8187 | ||
8188 | if (quoted & Q_HERE_DOCUMENT) | |
8189 | tflag = CBSHDOC; | |
8190 | else if (quoted & Q_DOUBLE_QUOTES) | |
8191 | tflag = CBSDQUOTE; | |
8192 | else | |
8193 | tflag = 0; | |
8194 | ||
8195 | /* From Posix discussion on austin-group list: Backslash escaping | |
8196 | a } in ${...} is removed. Issue 0000221 */ | |
8197 | if ((quoted & Q_DOLBRACE) && c == RBRACE) | |
8198 | { | |
8199 | SCOPY_CHAR_I (twochars, CTLESC, c, string, sindex, string_size); | |
8200 | } | |
8201 | else if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) && ((sh_syntaxtab[c] & tflag) == 0)) | |
8202 | { | |
8203 | SCOPY_CHAR_I (twochars, '\\', c, string, sindex, string_size); | |
8204 | } | |
8205 | else if (c == 0) | |
8206 | { | |
8207 | c = CTLNUL; | |
8208 | sindex--; /* add_character: label increments sindex */ | |
8209 | goto add_character; | |
8210 | } | |
8211 | else | |
8212 | { | |
8213 | SCOPY_CHAR_I (twochars, CTLESC, c, string, sindex, string_size); | |
8214 | } | |
8215 | ||
8216 | sindex++; | |
8217 | add_twochars: | |
8218 | /* BEFORE jumping here, we need to increment sindex if appropriate */ | |
8219 | RESIZE_MALLOCED_BUFFER (istring, istring_index, 2, istring_size, | |
8220 | DEFAULT_ARRAY_SIZE); | |
8221 | istring[istring_index++] = twochars[0]; | |
8222 | istring[istring_index++] = twochars[1]; | |
8223 | istring[istring_index] = '\0'; | |
8224 | ||
8225 | break; | |
8226 | ||
8227 | case '"': | |
8228 | #if 0 | |
8229 | if ((quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) || (word->flags & W_DQUOTE)) | |
8230 | #else | |
8231 | if ((quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT))) | |
8232 | #endif | |
8233 | goto add_character; | |
8234 | ||
8235 | t_index = ++sindex; | |
8236 | temp = string_extract_double_quoted (string, &sindex, 0); | |
8237 | ||
8238 | /* If the quotes surrounded the entire string, then the | |
8239 | whole word was quoted. */ | |
8240 | quoted_state = (t_index == 1 && string[sindex] == '\0') | |
8241 | ? WHOLLY_QUOTED | |
8242 | : PARTIALLY_QUOTED; | |
8243 | ||
8244 | if (temp && *temp) | |
8245 | { | |
8246 | tword = alloc_word_desc (); | |
8247 | tword->word = temp; | |
8248 | ||
8249 | temp = (char *)NULL; | |
8250 | ||
8251 | has_dollar_at = 0; | |
8252 | /* Need to get W_HASQUOTEDNULL flag through this function. */ | |
8253 | list = expand_word_internal (tword, Q_DOUBLE_QUOTES, 0, &has_dollar_at, (int *)NULL); | |
8254 | ||
8255 | if (list == &expand_word_error || list == &expand_word_fatal) | |
8256 | { | |
8257 | free (istring); | |
8258 | free (string); | |
8259 | /* expand_word_internal has already freed temp_word->word | |
8260 | for us because of the way it prints error messages. */ | |
8261 | tword->word = (char *)NULL; | |
8262 | dispose_word (tword); | |
8263 | return list; | |
8264 | } | |
8265 | ||
8266 | dispose_word (tword); | |
8267 | ||
8268 | /* "$@" (a double-quoted dollar-at) expands into nothing, | |
8269 | not even a NULL word, when there are no positional | |
8270 | parameters. */ | |
8271 | if (list == 0 && has_dollar_at) | |
8272 | { | |
8273 | quoted_dollar_at++; | |
8274 | break; | |
8275 | } | |
8276 | ||
8277 | /* If we get "$@", we know we have expanded something, so we | |
8278 | need to remember it for the final split on $IFS. This is | |
8279 | a special case; it's the only case where a quoted string | |
8280 | can expand into more than one word. It's going to come back | |
8281 | from the above call to expand_word_internal as a list with | |
8282 | a single word, in which all characters are quoted and | |
8283 | separated by blanks. What we want to do is to turn it back | |
8284 | into a list for the next piece of code. */ | |
8285 | if (list) | |
8286 | dequote_list (list); | |
8287 | ||
8288 | if (list && list->word && (list->word->flags & W_HASQUOTEDNULL)) | |
8289 | had_quoted_null = 1; | |
8290 | ||
8291 | if (has_dollar_at) | |
8292 | { | |
8293 | quoted_dollar_at++; | |
8294 | if (contains_dollar_at) | |
8295 | *contains_dollar_at = 1; | |
8296 | if (expanded_something) | |
8297 | *expanded_something = 1; | |
8298 | } | |
8299 | } | |
8300 | else | |
8301 | { | |
8302 | /* What we have is "". This is a minor optimization. */ | |
8303 | FREE (temp); | |
8304 | list = (WORD_LIST *)NULL; | |
8305 | } | |
8306 | ||
8307 | /* The code above *might* return a list (consider the case of "$@", | |
8308 | where it returns "$1", "$2", etc.). We can't throw away the | |
8309 | rest of the list, and we have to make sure each word gets added | |
8310 | as quoted. We test on tresult->next: if it is non-NULL, we | |
8311 | quote the whole list, save it to a string with string_list, and | |
8312 | add that string. We don't need to quote the results of this | |
8313 | (and it would be wrong, since that would quote the separators | |
8314 | as well), so we go directly to add_string. */ | |
8315 | if (list) | |
8316 | { | |
8317 | if (list->next) | |
8318 | { | |
8319 | #if 0 | |
8320 | if (quoted_dollar_at && (word->flags & W_NOSPLIT2)) | |
8321 | temp = string_list_internal (quote_list (list), " "); | |
8322 | else | |
8323 | #endif | |
8324 | /* Testing quoted_dollar_at makes sure that "$@" is | |
8325 | split correctly when $IFS does not contain a space. */ | |
8326 | temp = quoted_dollar_at | |
8327 | ? string_list_dollar_at (list, Q_DOUBLE_QUOTES) | |
8328 | : string_list (quote_list (list)); | |
8329 | dispose_words (list); | |
8330 | goto add_string; | |
8331 | } | |
8332 | else | |
8333 | { | |
8334 | temp = savestring (list->word->word); | |
8335 | tflag = list->word->flags; | |
8336 | dispose_words (list); | |
8337 | ||
8338 | /* If the string is not a quoted null string, we want | |
8339 | to remove any embedded unquoted CTLNUL characters. | |
8340 | We do not want to turn quoted null strings back into | |
8341 | the empty string, though. We do this because we | |
8342 | want to remove any quoted nulls from expansions that | |
8343 | contain other characters. For example, if we have | |
8344 | x"$*"y or "x$*y" and there are no positional parameters, | |
8345 | the $* should expand into nothing. */ | |
8346 | /* We use the W_HASQUOTEDNULL flag to differentiate the | |
8347 | cases: a quoted null character as above and when | |
8348 | CTLNUL is contained in the (non-null) expansion | |
8349 | of some variable. We use the had_quoted_null flag to | |
8350 | pass the value through this function to its caller. */ | |
8351 | if ((tflag & W_HASQUOTEDNULL) && QUOTED_NULL (temp) == 0) | |
8352 | remove_quoted_nulls (temp); /* XXX */ | |
8353 | } | |
8354 | } | |
8355 | else | |
8356 | temp = (char *)NULL; | |
8357 | ||
8358 | /* We do not want to add quoted nulls to strings that are only | |
8359 | partially quoted; we can throw them away. */ | |
8360 | if (temp == 0 && quoted_state == PARTIALLY_QUOTED && (word->flags & (W_NOSPLIT|W_NOSPLIT2))) | |
8361 | continue; | |
8362 | ||
8363 | add_quoted_string: | |
8364 | ||
8365 | if (temp) | |
8366 | { | |
8367 | temp1 = temp; | |
8368 | temp = quote_string (temp); | |
8369 | free (temp1); | |
8370 | goto add_string; | |
8371 | } | |
8372 | else | |
8373 | { | |
8374 | /* Add NULL arg. */ | |
8375 | c = CTLNUL; | |
8376 | sindex--; /* add_character: label increments sindex */ | |
8377 | goto add_character; | |
8378 | } | |
8379 | ||
8380 | /* break; */ | |
8381 | ||
8382 | case '\'': | |
8383 | #if 0 | |
8384 | if ((quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) || (word->flags & W_DQUOTE)) | |
8385 | #else | |
8386 | if ((quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT))) | |
8387 | #endif | |
8388 | goto add_character; | |
8389 | ||
8390 | t_index = ++sindex; | |
8391 | temp = string_extract_single_quoted (string, &sindex); | |
8392 | ||
8393 | /* If the entire STRING was surrounded by single quotes, | |
8394 | then the string is wholly quoted. */ | |
8395 | quoted_state = (t_index == 1 && string[sindex] == '\0') | |
8396 | ? WHOLLY_QUOTED | |
8397 | : PARTIALLY_QUOTED; | |
8398 | ||
8399 | /* If all we had was '', it is a null expansion. */ | |
8400 | if (*temp == '\0') | |
8401 | { | |
8402 | free (temp); | |
8403 | temp = (char *)NULL; | |
8404 | } | |
8405 | else | |
8406 | remove_quoted_escapes (temp); /* ??? */ | |
8407 | ||
8408 | /* We do not want to add quoted nulls to strings that are only | |
8409 | partially quoted; such nulls are discarded. */ | |
8410 | if (temp == 0 && (quoted_state == PARTIALLY_QUOTED)) | |
8411 | continue; | |
8412 | ||
8413 | /* If we have a quoted null expansion, add a quoted NULL to istring. */ | |
8414 | if (temp == 0) | |
8415 | { | |
8416 | c = CTLNUL; | |
8417 | sindex--; /* add_character: label increments sindex */ | |
8418 | goto add_character; | |
8419 | } | |
8420 | else | |
8421 | goto add_quoted_string; | |
8422 | ||
8423 | /* break; */ | |
8424 | ||
8425 | default: | |
8426 | /* This is the fix for " $@ " */ | |
8427 | add_ifs_character: | |
8428 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) || (isexp == 0 && isifs (c))) | |
8429 | { | |
8430 | if (string[sindex]) /* from old goto dollar_add_string */ | |
8431 | sindex++; | |
8432 | if (c == 0) | |
8433 | { | |
8434 | c = CTLNUL; | |
8435 | goto add_character; | |
8436 | } | |
8437 | else | |
8438 | { | |
8439 | #if HANDLE_MULTIBYTE | |
8440 | if (MB_CUR_MAX > 1) | |
8441 | sindex--; | |
8442 | ||
8443 | if (MB_CUR_MAX > 1) | |
8444 | { | |
8445 | SADD_MBQCHAR_BODY(temp, string, sindex, string_size); | |
8446 | } | |
8447 | else | |
8448 | #endif | |
8449 | { | |
8450 | twochars[0] = CTLESC; | |
8451 | twochars[1] = c; | |
8452 | goto add_twochars; | |
8453 | } | |
8454 | } | |
8455 | } | |
8456 | ||
8457 | SADD_MBCHAR (temp, string, sindex, string_size); | |
8458 | ||
8459 | add_character: | |
8460 | RESIZE_MALLOCED_BUFFER (istring, istring_index, 1, istring_size, | |
8461 | DEFAULT_ARRAY_SIZE); | |
8462 | istring[istring_index++] = c; | |
8463 | istring[istring_index] = '\0'; | |
8464 | ||
8465 | /* Next character. */ | |
8466 | sindex++; | |
8467 | } | |
8468 | } | |
8469 | ||
8470 | finished_with_string: | |
8471 | /* OK, we're ready to return. If we have a quoted string, and | |
8472 | quoted_dollar_at is not set, we do no splitting at all; otherwise | |
8473 | we split on ' '. The routines that call this will handle what to | |
8474 | do if nothing has been expanded. */ | |
8475 | ||
8476 | /* Partially and wholly quoted strings which expand to the empty | |
8477 | string are retained as an empty arguments. Unquoted strings | |
8478 | which expand to the empty string are discarded. The single | |
8479 | exception is the case of expanding "$@" when there are no | |
8480 | positional parameters. In that case, we discard the expansion. */ | |
8481 | ||
8482 | /* Because of how the code that handles "" and '' in partially | |
8483 | quoted strings works, we need to make ISTRING into a QUOTED_NULL | |
8484 | if we saw quoting characters, but the expansion was empty. | |
8485 | "" and '' are tossed away before we get to this point when | |
8486 | processing partially quoted strings. This makes "" and $xxx"" | |
8487 | equivalent when xxx is unset. We also look to see whether we | |
8488 | saw a quoted null from a ${} expansion and add one back if we | |
8489 | need to. */ | |
8490 | ||
8491 | /* If we expand to nothing and there were no single or double quotes | |
8492 | in the word, we throw it away. Otherwise, we return a NULL word. | |
8493 | The single exception is for $@ surrounded by double quotes when | |
8494 | there are no positional parameters. In that case, we also throw | |
8495 | the word away. */ | |
8496 | ||
8497 | if (*istring == '\0') | |
8498 | { | |
8499 | if (quoted_dollar_at == 0 && (had_quoted_null || quoted_state == PARTIALLY_QUOTED)) | |
8500 | { | |
8501 | istring[0] = CTLNUL; | |
8502 | istring[1] = '\0'; | |
8503 | tword = make_bare_word (istring); | |
8504 | tword->flags |= W_HASQUOTEDNULL; /* XXX */ | |
8505 | list = make_word_list (tword, (WORD_LIST *)NULL); | |
8506 | if (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
8507 | tword->flags |= W_QUOTED; | |
8508 | } | |
8509 | /* According to sh, ksh, and Posix.2, if a word expands into nothing | |
8510 | and a double-quoted "$@" appears anywhere in it, then the entire | |
8511 | word is removed. */ | |
8512 | else if (quoted_state == UNQUOTED || quoted_dollar_at) | |
8513 | list = (WORD_LIST *)NULL; | |
8514 | #if 0 | |
8515 | else | |
8516 | { | |
8517 | tword = make_bare_word (istring); | |
8518 | if (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
8519 | tword->flags |= W_QUOTED; | |
8520 | list = make_word_list (tword, (WORD_LIST *)NULL); | |
8521 | } | |
8522 | #else | |
8523 | else | |
8524 | list = (WORD_LIST *)NULL; | |
8525 | #endif | |
8526 | } | |
8527 | else if (word->flags & W_NOSPLIT) | |
8528 | { | |
8529 | tword = make_bare_word (istring); | |
8530 | if (word->flags & W_ASSIGNMENT) | |
8531 | tword->flags |= W_ASSIGNMENT; /* XXX */ | |
8532 | if (word->flags & W_COMPASSIGN) | |
8533 | tword->flags |= W_COMPASSIGN; /* XXX */ | |
8534 | if (word->flags & W_NOGLOB) | |
8535 | tword->flags |= W_NOGLOB; /* XXX */ | |
8536 | if (word->flags & W_NOEXPAND) | |
8537 | tword->flags |= W_NOEXPAND; /* XXX */ | |
8538 | if (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) | |
8539 | tword->flags |= W_QUOTED; | |
8540 | if (had_quoted_null) | |
8541 | tword->flags |= W_HASQUOTEDNULL; | |
8542 | list = make_word_list (tword, (WORD_LIST *)NULL); | |
8543 | } | |
8544 | else | |
8545 | { | |
8546 | char *ifs_chars; | |
8547 | ||
8548 | ifs_chars = (quoted_dollar_at || has_dollar_at) ? ifs_value : (char *)NULL; | |
8549 | ||
8550 | /* If we have $@, we need to split the results no matter what. If | |
8551 | IFS is unset or NULL, string_list_dollar_at has separated the | |
8552 | positional parameters with a space, so we split on space (we have | |
8553 | set ifs_chars to " \t\n" above if ifs is unset). If IFS is set, | |
8554 | string_list_dollar_at has separated the positional parameters | |
8555 | with the first character of $IFS, so we split on $IFS. */ | |
8556 | if (has_dollar_at && ifs_chars) | |
8557 | list = list_string (istring, *ifs_chars ? ifs_chars : " ", 1); | |
8558 | else | |
8559 | { | |
8560 | tword = make_bare_word (istring); | |
8561 | if ((quoted & (Q_DOUBLE_QUOTES|Q_HERE_DOCUMENT)) || (quoted_state == WHOLLY_QUOTED)) | |
8562 | tword->flags |= W_QUOTED; | |
8563 | if (word->flags & W_ASSIGNMENT) | |
8564 | tword->flags |= W_ASSIGNMENT; | |
8565 | if (word->flags & W_COMPASSIGN) | |
8566 | tword->flags |= W_COMPASSIGN; | |
8567 | if (word->flags & W_NOGLOB) | |
8568 | tword->flags |= W_NOGLOB; | |
8569 | if (word->flags & W_NOEXPAND) | |
8570 | tword->flags |= W_NOEXPAND; | |
8571 | if (had_quoted_null) | |
8572 | tword->flags |= W_HASQUOTEDNULL; /* XXX */ | |
8573 | list = make_word_list (tword, (WORD_LIST *)NULL); | |
8574 | } | |
8575 | } | |
8576 | ||
8577 | free (istring); | |
8578 | return (list); | |
8579 | } | |
8580 | ||
8581 | /* **************************************************************** */ | |
8582 | /* */ | |
8583 | /* Functions for Quote Removal */ | |
8584 | /* */ | |
8585 | /* **************************************************************** */ | |
8586 | ||
8587 | /* Perform quote removal on STRING. If QUOTED > 0, assume we are obeying the | |
8588 | backslash quoting rules for within double quotes or a here document. */ | |
8589 | char * | |
8590 | string_quote_removal (string, quoted) | |
8591 | char *string; | |
8592 | int quoted; | |
8593 | { | |
8594 | size_t slen; | |
8595 | char *r, *result_string, *temp, *send; | |
8596 | int sindex, tindex, dquote; | |
8597 | unsigned char c; | |
8598 | DECLARE_MBSTATE; | |
8599 | ||
8600 | /* The result can be no longer than the original string. */ | |
8601 | slen = strlen (string); | |
8602 | send = string + slen; | |
8603 | ||
8604 | r = result_string = (char *)xmalloc (slen + 1); | |
8605 | ||
8606 | for (dquote = sindex = 0; c = string[sindex];) | |
8607 | { | |
8608 | switch (c) | |
8609 | { | |
8610 | case '\\': | |
8611 | c = string[++sindex]; | |
8612 | if (c == 0) | |
8613 | { | |
8614 | *r++ = '\\'; | |
8615 | break; | |
8616 | } | |
8617 | if (((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) || dquote) && (sh_syntaxtab[c] & CBSDQUOTE) == 0) | |
8618 | *r++ = '\\'; | |
8619 | /* FALLTHROUGH */ | |
8620 | ||
8621 | default: | |
8622 | SCOPY_CHAR_M (r, string, send, sindex); | |
8623 | break; | |
8624 | ||
8625 | case '\'': | |
8626 | if ((quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) || dquote) | |
8627 | { | |
8628 | *r++ = c; | |
8629 | sindex++; | |
8630 | break; | |
8631 | } | |
8632 | tindex = sindex + 1; | |
8633 | temp = string_extract_single_quoted (string, &tindex); | |
8634 | if (temp) | |
8635 | { | |
8636 | strcpy (r, temp); | |
8637 | r += strlen (r); | |
8638 | free (temp); | |
8639 | } | |
8640 | sindex = tindex; | |
8641 | break; | |
8642 | ||
8643 | case '"': | |
8644 | dquote = 1 - dquote; | |
8645 | sindex++; | |
8646 | break; | |
8647 | } | |
8648 | } | |
8649 | *r = '\0'; | |
8650 | return (result_string); | |
8651 | } | |
8652 | ||
8653 | #if 0 | |
8654 | /* UNUSED */ | |
8655 | /* Perform quote removal on word WORD. This allocates and returns a new | |
8656 | WORD_DESC *. */ | |
8657 | WORD_DESC * | |
8658 | word_quote_removal (word, quoted) | |
8659 | WORD_DESC *word; | |
8660 | int quoted; | |
8661 | { | |
8662 | WORD_DESC *w; | |
8663 | char *t; | |
8664 | ||
8665 | t = string_quote_removal (word->word, quoted); | |
8666 | w = alloc_word_desc (); | |
8667 | w->word = t ? t : savestring (""); | |
8668 | return (w); | |
8669 | } | |
8670 | ||
8671 | /* Perform quote removal on all words in LIST. If QUOTED is non-zero, | |
8672 | the members of the list are treated as if they are surrounded by | |
8673 | double quotes. Return a new list, or NULL if LIST is NULL. */ | |
8674 | WORD_LIST * | |
8675 | word_list_quote_removal (list, quoted) | |
8676 | WORD_LIST *list; | |
8677 | int quoted; | |
8678 | { | |
8679 | WORD_LIST *result, *t, *tresult, *e; | |
8680 | ||
8681 | for (t = list, result = (WORD_LIST *)NULL; t; t = t->next) | |
8682 | { | |
8683 | tresult = make_word_list (word_quote_removal (t->word, quoted), (WORD_LIST *)NULL); | |
8684 | #if 0 | |
8685 | result = (WORD_LIST *) list_append (result, tresult); | |
8686 | #else | |
8687 | if (result == 0) | |
8688 | result = e = tresult; | |
8689 | else | |
8690 | { | |
8691 | e->next = tresult; | |
8692 | while (e->next) | |
8693 | e = e->next; | |
8694 | } | |
8695 | #endif | |
8696 | } | |
8697 | return (result); | |
8698 | } | |
8699 | #endif | |
8700 | ||
8701 | /******************************************* | |
8702 | * * | |
8703 | * Functions to perform word splitting * | |
8704 | * * | |
8705 | *******************************************/ | |
8706 | ||
8707 | void | |
8708 | setifs (v) | |
8709 | SHELL_VAR *v; | |
8710 | { | |
8711 | char *t; | |
8712 | unsigned char uc; | |
8713 | ||
8714 | ifs_var = v; | |
8715 | ifs_value = (v && value_cell (v)) ? value_cell (v) : " \t\n"; | |
8716 | ||
8717 | /* Should really merge ifs_cmap with sh_syntaxtab. XXX - doesn't yet | |
8718 | handle multibyte chars in IFS */ | |
8719 | memset (ifs_cmap, '\0', sizeof (ifs_cmap)); | |
8720 | for (t = ifs_value ; t && *t; t++) | |
8721 | { | |
8722 | uc = *t; | |
8723 | ifs_cmap[uc] = 1; | |
8724 | } | |
8725 | ||
8726 | #if defined (HANDLE_MULTIBYTE) | |
8727 | if (ifs_value == 0) | |
8728 | { | |
8729 | ifs_firstc[0] = '\0'; | |
8730 | ifs_firstc_len = 1; | |
8731 | } | |
8732 | else | |
8733 | { | |
8734 | size_t ifs_len; | |
8735 | ifs_len = strnlen (ifs_value, MB_CUR_MAX); | |
8736 | ifs_firstc_len = MBLEN (ifs_value, ifs_len); | |
8737 | if (ifs_firstc_len == 1 || ifs_firstc_len == 0 || MB_INVALIDCH (ifs_firstc_len)) | |
8738 | { | |
8739 | ifs_firstc[0] = ifs_value[0]; | |
8740 | ifs_firstc[1] = '\0'; | |
8741 | ifs_firstc_len = 1; | |
8742 | } | |
8743 | else | |
8744 | memcpy (ifs_firstc, ifs_value, ifs_firstc_len); | |
8745 | } | |
8746 | #else | |
8747 | ifs_firstc = ifs_value ? *ifs_value : 0; | |
8748 | #endif | |
8749 | } | |
8750 | ||
8751 | char * | |
8752 | getifs () | |
8753 | { | |
8754 | return ifs_value; | |
8755 | } | |
8756 | ||
8757 | /* This splits a single word into a WORD LIST on $IFS, but only if the word | |
8758 | is not quoted. list_string () performs quote removal for us, even if we | |
8759 | don't do any splitting. */ | |
8760 | WORD_LIST * | |
8761 | word_split (w, ifs_chars) | |
8762 | WORD_DESC *w; | |
8763 | char *ifs_chars; | |
8764 | { | |
8765 | WORD_LIST *result; | |
8766 | ||
8767 | if (w) | |
8768 | { | |
8769 | char *xifs; | |
8770 | ||
8771 | xifs = ((w->flags & W_QUOTED) || ifs_chars == 0) ? "" : ifs_chars; | |
8772 | result = list_string (w->word, xifs, w->flags & W_QUOTED); | |
8773 | } | |
8774 | else | |
8775 | result = (WORD_LIST *)NULL; | |
8776 | ||
8777 | return (result); | |
8778 | } | |
8779 | ||
8780 | /* Perform word splitting on LIST and return the RESULT. It is possible | |
8781 | to return (WORD_LIST *)NULL. */ | |
8782 | static WORD_LIST * | |
8783 | word_list_split (list) | |
8784 | WORD_LIST *list; | |
8785 | { | |
8786 | WORD_LIST *result, *t, *tresult, *e; | |
8787 | ||
8788 | for (t = list, result = (WORD_LIST *)NULL; t; t = t->next) | |
8789 | { | |
8790 | tresult = word_split (t->word, ifs_value); | |
8791 | if (result == 0) | |
8792 | result = e = tresult; | |
8793 | else | |
8794 | { | |
8795 | e->next = tresult; | |
8796 | while (e->next) | |
8797 | e = e->next; | |
8798 | } | |
8799 | } | |
8800 | return (result); | |
8801 | } | |
8802 | ||
8803 | /************************************************** | |
8804 | * * | |
8805 | * Functions to expand an entire WORD_LIST * | |
8806 | * * | |
8807 | **************************************************/ | |
8808 | ||
8809 | /* Do any word-expansion-specific cleanup and jump to top_level */ | |
8810 | static void | |
8811 | exp_jump_to_top_level (v) | |
8812 | int v; | |
8813 | { | |
8814 | set_pipestatus_from_exit (last_command_exit_value); | |
8815 | ||
8816 | /* Cleanup code goes here. */ | |
8817 | expand_no_split_dollar_star = 0; /* XXX */ | |
8818 | expanding_redir = 0; | |
8819 | assigning_in_environment = 0; | |
8820 | ||
8821 | if (parse_and_execute_level == 0) | |
8822 | top_level_cleanup (); /* from sig.c */ | |
8823 | ||
8824 | jump_to_top_level (v); | |
8825 | } | |
8826 | ||
8827 | /* Put NLIST (which is a WORD_LIST * of only one element) at the front of | |
8828 | ELIST, and set ELIST to the new list. */ | |
8829 | #define PREPEND_LIST(nlist, elist) \ | |
8830 | do { nlist->next = elist; elist = nlist; } while (0) | |
8831 | ||
8832 | /* Separate out any initial variable assignments from TLIST. If set -k has | |
8833 | been executed, remove all assignment statements from TLIST. Initial | |
8834 | variable assignments and other environment assignments are placed | |
8835 | on SUBST_ASSIGN_VARLIST. */ | |
8836 | static WORD_LIST * | |
8837 | separate_out_assignments (tlist) | |
8838 | WORD_LIST *tlist; | |
8839 | { | |
8840 | register WORD_LIST *vp, *lp; | |
8841 | ||
8842 | if (tlist == 0) | |
8843 | return ((WORD_LIST *)NULL); | |
8844 | ||
8845 | if (subst_assign_varlist) | |
8846 | dispose_words (subst_assign_varlist); /* Clean up after previous error */ | |
8847 | ||
8848 | subst_assign_varlist = (WORD_LIST *)NULL; | |
8849 | vp = lp = tlist; | |
8850 | ||
8851 | /* Separate out variable assignments at the start of the command. | |
8852 | Loop invariant: vp->next == lp | |
8853 | Loop postcondition: | |
8854 | lp = list of words left after assignment statements skipped | |
8855 | tlist = original list of words | |
8856 | */ | |
8857 | while (lp && (lp->word->flags & W_ASSIGNMENT)) | |
8858 | { | |
8859 | vp = lp; | |
8860 | lp = lp->next; | |
8861 | } | |
8862 | ||
8863 | /* If lp != tlist, we have some initial assignment statements. | |
8864 | We make SUBST_ASSIGN_VARLIST point to the list of assignment | |
8865 | words and TLIST point to the remaining words. */ | |
8866 | if (lp != tlist) | |
8867 | { | |
8868 | subst_assign_varlist = tlist; | |
8869 | /* ASSERT(vp->next == lp); */ | |
8870 | vp->next = (WORD_LIST *)NULL; /* terminate variable list */ | |
8871 | tlist = lp; /* remainder of word list */ | |
8872 | } | |
8873 | ||
8874 | /* vp == end of variable list */ | |
8875 | /* tlist == remainder of original word list without variable assignments */ | |
8876 | if (!tlist) | |
8877 | /* All the words in tlist were assignment statements */ | |
8878 | return ((WORD_LIST *)NULL); | |
8879 | ||
8880 | /* ASSERT(tlist != NULL); */ | |
8881 | /* ASSERT((tlist->word->flags & W_ASSIGNMENT) == 0); */ | |
8882 | ||
8883 | /* If the -k option is in effect, we need to go through the remaining | |
8884 | words, separate out the assignment words, and place them on | |
8885 | SUBST_ASSIGN_VARLIST. */ | |
8886 | if (place_keywords_in_env) | |
8887 | { | |
8888 | WORD_LIST *tp; /* tp == running pointer into tlist */ | |
8889 | ||
8890 | tp = tlist; | |
8891 | lp = tlist->next; | |
8892 | ||
8893 | /* Loop Invariant: tp->next == lp */ | |
8894 | /* Loop postcondition: tlist == word list without assignment statements */ | |
8895 | while (lp) | |
8896 | { | |
8897 | if (lp->word->flags & W_ASSIGNMENT) | |
8898 | { | |
8899 | /* Found an assignment statement, add this word to end of | |
8900 | subst_assign_varlist (vp). */ | |
8901 | if (!subst_assign_varlist) | |
8902 | subst_assign_varlist = vp = lp; | |
8903 | else | |
8904 | { | |
8905 | vp->next = lp; | |
8906 | vp = lp; | |
8907 | } | |
8908 | ||
8909 | /* Remove the word pointed to by LP from TLIST. */ | |
8910 | tp->next = lp->next; | |
8911 | /* ASSERT(vp == lp); */ | |
8912 | lp->next = (WORD_LIST *)NULL; | |
8913 | lp = tp->next; | |
8914 | } | |
8915 | else | |
8916 | { | |
8917 | tp = lp; | |
8918 | lp = lp->next; | |
8919 | } | |
8920 | } | |
8921 | } | |
8922 | return (tlist); | |
8923 | } | |
8924 | ||
8925 | #define WEXP_VARASSIGN 0x001 | |
8926 | #define WEXP_BRACEEXP 0x002 | |
8927 | #define WEXP_TILDEEXP 0x004 | |
8928 | #define WEXP_PARAMEXP 0x008 | |
8929 | #define WEXP_PATHEXP 0x010 | |
8930 | ||
8931 | /* All of the expansions, including variable assignments at the start of | |
8932 | the list. */ | |
8933 | #define WEXP_ALL (WEXP_VARASSIGN|WEXP_BRACEEXP|WEXP_TILDEEXP|WEXP_PARAMEXP|WEXP_PATHEXP) | |
8934 | ||
8935 | /* All of the expansions except variable assignments at the start of | |
8936 | the list. */ | |
8937 | #define WEXP_NOVARS (WEXP_BRACEEXP|WEXP_TILDEEXP|WEXP_PARAMEXP|WEXP_PATHEXP) | |
8938 | ||
8939 | /* All of the `shell expansions': brace expansion, tilde expansion, parameter | |
8940 | expansion, command substitution, arithmetic expansion, word splitting, and | |
8941 | quote removal. */ | |
8942 | #define WEXP_SHELLEXP (WEXP_BRACEEXP|WEXP_TILDEEXP|WEXP_PARAMEXP) | |
8943 | ||
8944 | /* Take the list of words in LIST and do the various substitutions. Return | |
8945 | a new list of words which is the expanded list, and without things like | |
8946 | variable assignments. */ | |
8947 | ||
8948 | WORD_LIST * | |
8949 | expand_words (list) | |
8950 | WORD_LIST *list; | |
8951 | { | |
8952 | return (expand_word_list_internal (list, WEXP_ALL)); | |
8953 | } | |
8954 | ||
8955 | /* Same as expand_words (), but doesn't hack variable or environment | |
8956 | variables. */ | |
8957 | WORD_LIST * | |
8958 | expand_words_no_vars (list) | |
8959 | WORD_LIST *list; | |
8960 | { | |
8961 | return (expand_word_list_internal (list, WEXP_NOVARS)); | |
8962 | } | |
8963 | ||
8964 | WORD_LIST * | |
8965 | expand_words_shellexp (list) | |
8966 | WORD_LIST *list; | |
8967 | { | |
8968 | return (expand_word_list_internal (list, WEXP_SHELLEXP)); | |
8969 | } | |
8970 | ||
8971 | static WORD_LIST * | |
8972 | glob_expand_word_list (tlist, eflags) | |
8973 | WORD_LIST *tlist; | |
8974 | int eflags; | |
8975 | { | |
8976 | char **glob_array, *temp_string; | |
8977 | register int glob_index; | |
8978 | WORD_LIST *glob_list, *output_list, *disposables, *next; | |
8979 | WORD_DESC *tword; | |
8980 | ||
8981 | output_list = disposables = (WORD_LIST *)NULL; | |
8982 | glob_array = (char **)NULL; | |
8983 | while (tlist) | |
8984 | { | |
8985 | /* For each word, either globbing is attempted or the word is | |
8986 | added to orig_list. If globbing succeeds, the results are | |
8987 | added to orig_list and the word (tlist) is added to the list | |
8988 | of disposable words. If globbing fails and failed glob | |
8989 | expansions are left unchanged (the shell default), the | |
8990 | original word is added to orig_list. If globbing fails and | |
8991 | failed glob expansions are removed, the original word is | |
8992 | added to the list of disposable words. orig_list ends up | |
8993 | in reverse order and requires a call to REVERSE_LIST to | |
8994 | be set right. After all words are examined, the disposable | |
8995 | words are freed. */ | |
8996 | next = tlist->next; | |
8997 | ||
8998 | /* If the word isn't an assignment and contains an unquoted | |
8999 | pattern matching character, then glob it. */ | |
9000 | if ((tlist->word->flags & W_NOGLOB) == 0 && | |
9001 | unquoted_glob_pattern_p (tlist->word->word)) | |
9002 | { | |
9003 | glob_array = shell_glob_filename (tlist->word->word); | |
9004 | ||
9005 | /* Handle error cases. | |
9006 | I don't think we should report errors like "No such file | |
9007 | or directory". However, I would like to report errors | |
9008 | like "Read failed". */ | |
9009 | ||
9010 | if (glob_array == 0 || GLOB_FAILED (glob_array)) | |
9011 | { | |
9012 | glob_array = (char **)xmalloc (sizeof (char *)); | |
9013 | glob_array[0] = (char *)NULL; | |
9014 | } | |
9015 | ||
9016 | /* Dequote the current word in case we have to use it. */ | |
9017 | if (glob_array[0] == NULL) | |
9018 | { | |
9019 | temp_string = dequote_string (tlist->word->word); | |
9020 | free (tlist->word->word); | |
9021 | tlist->word->word = temp_string; | |
9022 | } | |
9023 | ||
9024 | /* Make the array into a word list. */ | |
9025 | glob_list = (WORD_LIST *)NULL; | |
9026 | for (glob_index = 0; glob_array[glob_index]; glob_index++) | |
9027 | { | |
9028 | tword = make_bare_word (glob_array[glob_index]); | |
9029 | tword->flags |= W_GLOBEXP; /* XXX */ | |
9030 | glob_list = make_word_list (tword, glob_list); | |
9031 | } | |
9032 | ||
9033 | if (glob_list) | |
9034 | { | |
9035 | output_list = (WORD_LIST *)list_append (glob_list, output_list); | |
9036 | PREPEND_LIST (tlist, disposables); | |
9037 | } | |
9038 | else if (fail_glob_expansion != 0) | |
9039 | { | |
9040 | report_error (_("no match: %s"), tlist->word->word); | |
9041 | exp_jump_to_top_level (DISCARD); | |
9042 | } | |
9043 | else if (allow_null_glob_expansion == 0) | |
9044 | { | |
9045 | /* Failed glob expressions are left unchanged. */ | |
9046 | PREPEND_LIST (tlist, output_list); | |
9047 | } | |
9048 | else | |
9049 | { | |
9050 | /* Failed glob expressions are removed. */ | |
9051 | PREPEND_LIST (tlist, disposables); | |
9052 | } | |
9053 | } | |
9054 | else | |
9055 | { | |
9056 | /* Dequote the string. */ | |
9057 | temp_string = dequote_string (tlist->word->word); | |
9058 | free (tlist->word->word); | |
9059 | tlist->word->word = temp_string; | |
9060 | PREPEND_LIST (tlist, output_list); | |
9061 | } | |
9062 | ||
9063 | strvec_dispose (glob_array); | |
9064 | glob_array = (char **)NULL; | |
9065 | ||
9066 | tlist = next; | |
9067 | } | |
9068 | ||
9069 | if (disposables) | |
9070 | dispose_words (disposables); | |
9071 | ||
9072 | if (output_list) | |
9073 | output_list = REVERSE_LIST (output_list, WORD_LIST *); | |
9074 | ||
9075 | return (output_list); | |
9076 | } | |
9077 | ||
9078 | #if defined (BRACE_EXPANSION) | |
9079 | static WORD_LIST * | |
9080 | brace_expand_word_list (tlist, eflags) | |
9081 | WORD_LIST *tlist; | |
9082 | int eflags; | |
9083 | { | |
9084 | register char **expansions; | |
9085 | char *temp_string; | |
9086 | WORD_LIST *disposables, *output_list, *next; | |
9087 | WORD_DESC *w; | |
9088 | int eindex; | |
9089 | ||
9090 | for (disposables = output_list = (WORD_LIST *)NULL; tlist; tlist = next) | |
9091 | { | |
9092 | next = tlist->next; | |
9093 | ||
9094 | if ((tlist->word->flags & (W_COMPASSIGN|W_ASSIGNARG)) == (W_COMPASSIGN|W_ASSIGNARG)) | |
9095 | { | |
9096 | /*itrace("brace_expand_word_list: %s: W_COMPASSIGN|W_ASSIGNARG", tlist->word->word);*/ | |
9097 | PREPEND_LIST (tlist, output_list); | |
9098 | continue; | |
9099 | } | |
9100 | ||
9101 | /* Only do brace expansion if the word has a brace character. If | |
9102 | not, just add the word list element to BRACES and continue. In | |
9103 | the common case, at least when running shell scripts, this will | |
9104 | degenerate to a bunch of calls to `mbschr', and then what is | |
9105 | basically a reversal of TLIST into BRACES, which is corrected | |
9106 | by a call to REVERSE_LIST () on BRACES when the end of TLIST | |
9107 | is reached. */ | |
9108 | if (mbschr (tlist->word->word, LBRACE)) | |
9109 | { | |
9110 | expansions = brace_expand (tlist->word->word); | |
9111 | ||
9112 | for (eindex = 0; temp_string = expansions[eindex]; eindex++) | |
9113 | { | |
9114 | w = make_word (temp_string); | |
9115 | /* If brace expansion didn't change the word, preserve | |
9116 | the flags. We may want to preserve the flags | |
9117 | unconditionally someday -- XXX */ | |
9118 | if (STREQ (temp_string, tlist->word->word)) | |
9119 | w->flags = tlist->word->flags; | |
9120 | output_list = make_word_list (w, output_list); | |
9121 | free (expansions[eindex]); | |
9122 | } | |
9123 | free (expansions); | |
9124 | ||
9125 | /* Add TLIST to the list of words to be freed after brace | |
9126 | expansion has been performed. */ | |
9127 | PREPEND_LIST (tlist, disposables); | |
9128 | } | |
9129 | else | |
9130 | PREPEND_LIST (tlist, output_list); | |
9131 | } | |
9132 | ||
9133 | if (disposables) | |
9134 | dispose_words (disposables); | |
9135 | ||
9136 | if (output_list) | |
9137 | output_list = REVERSE_LIST (output_list, WORD_LIST *); | |
9138 | ||
9139 | return (output_list); | |
9140 | } | |
9141 | #endif | |
9142 | ||
9143 | #if defined (ARRAY_VARS) | |
9144 | /* Take WORD, a compound associative array assignment, and internally run | |
9145 | 'declare -A w', where W is the variable name portion of WORD. */ | |
9146 | static int | |
9147 | make_internal_declare (word, option) | |
9148 | char *word; | |
9149 | char *option; | |
9150 | { | |
9151 | int t; | |
9152 | WORD_LIST *wl; | |
9153 | WORD_DESC *w; | |
9154 | ||
9155 | w = make_word (word); | |
9156 | ||
9157 | t = assignment (w->word, 0); | |
9158 | w->word[t] = '\0'; | |
9159 | ||
9160 | wl = make_word_list (w, (WORD_LIST *)NULL); | |
9161 | wl = make_word_list (make_word (option), wl); | |
9162 | ||
9163 | return (declare_builtin (wl)); | |
9164 | } | |
9165 | #endif | |
9166 | ||
9167 | static WORD_LIST * | |
9168 | shell_expand_word_list (tlist, eflags) | |
9169 | WORD_LIST *tlist; | |
9170 | int eflags; | |
9171 | { | |
9172 | WORD_LIST *expanded, *orig_list, *new_list, *next, *temp_list; | |
9173 | int expanded_something, has_dollar_at; | |
9174 | char *temp_string; | |
9175 | ||
9176 | /* We do tilde expansion all the time. This is what 1003.2 says. */ | |
9177 | new_list = (WORD_LIST *)NULL; | |
9178 | for (orig_list = tlist; tlist; tlist = next) | |
9179 | { | |
9180 | temp_string = tlist->word->word; | |
9181 | ||
9182 | next = tlist->next; | |
9183 | ||
9184 | #if defined (ARRAY_VARS) | |
9185 | /* If this is a compound array assignment to a builtin that accepts | |
9186 | such assignments (e.g., `declare'), take the assignment and perform | |
9187 | it separately, handling the semantics of declarations inside shell | |
9188 | functions. This avoids the double-evaluation of such arguments, | |
9189 | because `declare' does some evaluation of compound assignments on | |
9190 | its own. */ | |
9191 | if ((tlist->word->flags & (W_COMPASSIGN|W_ASSIGNARG)) == (W_COMPASSIGN|W_ASSIGNARG)) | |
9192 | { | |
9193 | int t; | |
9194 | ||
9195 | if (tlist->word->flags & W_ASSIGNASSOC) | |
9196 | make_internal_declare (tlist->word->word, "-A"); | |
9197 | ||
9198 | t = do_word_assignment (tlist->word, 0); | |
9199 | if (t == 0) | |
9200 | { | |
9201 | last_command_exit_value = EXECUTION_FAILURE; | |
9202 | exp_jump_to_top_level (DISCARD); | |
9203 | } | |
9204 | ||
9205 | /* Now transform the word as ksh93 appears to do and go on */ | |
9206 | t = assignment (tlist->word->word, 0); | |
9207 | tlist->word->word[t] = '\0'; | |
9208 | tlist->word->flags &= ~(W_ASSIGNMENT|W_NOSPLIT|W_COMPASSIGN|W_ASSIGNARG|W_ASSIGNASSOC); | |
9209 | } | |
9210 | #endif | |
9211 | ||
9212 | expanded_something = 0; | |
9213 | expanded = expand_word_internal | |
9214 | (tlist->word, 0, 0, &has_dollar_at, &expanded_something); | |
9215 | ||
9216 | if (expanded == &expand_word_error || expanded == &expand_word_fatal) | |
9217 | { | |
9218 | /* By convention, each time this error is returned, | |
9219 | tlist->word->word has already been freed. */ | |
9220 | tlist->word->word = (char *)NULL; | |
9221 | ||
9222 | /* Dispose our copy of the original list. */ | |
9223 | dispose_words (orig_list); | |
9224 | /* Dispose the new list we're building. */ | |
9225 | dispose_words (new_list); | |
9226 | ||
9227 | last_command_exit_value = EXECUTION_FAILURE; | |
9228 | if (expanded == &expand_word_error) | |
9229 | exp_jump_to_top_level (DISCARD); | |
9230 | else | |
9231 | exp_jump_to_top_level (FORCE_EOF); | |
9232 | } | |
9233 | ||
9234 | /* Don't split words marked W_NOSPLIT. */ | |
9235 | if (expanded_something && (tlist->word->flags & W_NOSPLIT) == 0) | |
9236 | { | |
9237 | temp_list = word_list_split (expanded); | |
9238 | dispose_words (expanded); | |
9239 | } | |
9240 | else | |
9241 | { | |
9242 | /* If no parameter expansion, command substitution, process | |
9243 | substitution, or arithmetic substitution took place, then | |
9244 | do not do word splitting. We still have to remove quoted | |
9245 | null characters from the result. */ | |
9246 | word_list_remove_quoted_nulls (expanded); | |
9247 | temp_list = expanded; | |
9248 | } | |
9249 | ||
9250 | expanded = REVERSE_LIST (temp_list, WORD_LIST *); | |
9251 | new_list = (WORD_LIST *)list_append (expanded, new_list); | |
9252 | } | |
9253 | ||
9254 | if (orig_list) | |
9255 | dispose_words (orig_list); | |
9256 | ||
9257 | if (new_list) | |
9258 | new_list = REVERSE_LIST (new_list, WORD_LIST *); | |
9259 | ||
9260 | return (new_list); | |
9261 | } | |
9262 | ||
9263 | /* The workhorse for expand_words () and expand_words_no_vars (). | |
9264 | First arg is LIST, a WORD_LIST of words. | |
9265 | Second arg EFLAGS is a flags word controlling which expansions are | |
9266 | performed. | |
9267 | ||
9268 | This does all of the substitutions: brace expansion, tilde expansion, | |
9269 | parameter expansion, command substitution, arithmetic expansion, | |
9270 | process substitution, word splitting, and pathname expansion, according | |
9271 | to the bits set in EFLAGS. Words with the W_QUOTED or W_NOSPLIT bits | |
9272 | set, or for which no expansion is done, do not undergo word splitting. | |
9273 | Words with the W_NOGLOB bit set do not undergo pathname expansion. */ | |
9274 | static WORD_LIST * | |
9275 | expand_word_list_internal (list, eflags) | |
9276 | WORD_LIST *list; | |
9277 | int eflags; | |
9278 | { | |
9279 | WORD_LIST *new_list, *temp_list; | |
9280 | int tint; | |
9281 | ||
9282 | if (list == 0) | |
9283 | return ((WORD_LIST *)NULL); | |
9284 | ||
9285 | garglist = new_list = copy_word_list (list); | |
9286 | if (eflags & WEXP_VARASSIGN) | |
9287 | { | |
9288 | garglist = new_list = separate_out_assignments (new_list); | |
9289 | if (new_list == 0) | |
9290 | { | |
9291 | if (subst_assign_varlist) | |
9292 | { | |
9293 | /* All the words were variable assignments, so they are placed | |
9294 | into the shell's environment. */ | |
9295 | for (temp_list = subst_assign_varlist; temp_list; temp_list = temp_list->next) | |
9296 | { | |
9297 | this_command_name = (char *)NULL; /* no arithmetic errors */ | |
9298 | tint = do_word_assignment (temp_list->word, 0); | |
9299 | /* Variable assignment errors in non-interactive shells | |
9300 | running in Posix.2 mode cause the shell to exit. */ | |
9301 | if (tint == 0) | |
9302 | { | |
9303 | last_command_exit_value = EXECUTION_FAILURE; | |
9304 | if (interactive_shell == 0 && posixly_correct) | |
9305 | exp_jump_to_top_level (FORCE_EOF); | |
9306 | else | |
9307 | exp_jump_to_top_level (DISCARD); | |
9308 | } | |
9309 | } | |
9310 | dispose_words (subst_assign_varlist); | |
9311 | subst_assign_varlist = (WORD_LIST *)NULL; | |
9312 | } | |
9313 | return ((WORD_LIST *)NULL); | |
9314 | } | |
9315 | } | |
9316 | ||
9317 | /* Begin expanding the words that remain. The expansions take place on | |
9318 | things that aren't really variable assignments. */ | |
9319 | ||
9320 | #if defined (BRACE_EXPANSION) | |
9321 | /* Do brace expansion on this word if there are any brace characters | |
9322 | in the string. */ | |
9323 | if ((eflags & WEXP_BRACEEXP) && brace_expansion && new_list) | |
9324 | new_list = brace_expand_word_list (new_list, eflags); | |
9325 | #endif /* BRACE_EXPANSION */ | |
9326 | ||
9327 | /* Perform the `normal' shell expansions: tilde expansion, parameter and | |
9328 | variable substitution, command substitution, arithmetic expansion, | |
9329 | and word splitting. */ | |
9330 | new_list = shell_expand_word_list (new_list, eflags); | |
9331 | ||
9332 | /* Okay, we're almost done. Now let's just do some filename | |
9333 | globbing. */ | |
9334 | if (new_list) | |
9335 | { | |
9336 | if ((eflags & WEXP_PATHEXP) && disallow_filename_globbing == 0) | |
9337 | /* Glob expand the word list unless globbing has been disabled. */ | |
9338 | new_list = glob_expand_word_list (new_list, eflags); | |
9339 | else | |
9340 | /* Dequote the words, because we're not performing globbing. */ | |
9341 | new_list = dequote_list (new_list); | |
9342 | } | |
9343 | ||
9344 | if ((eflags & WEXP_VARASSIGN) && subst_assign_varlist) | |
9345 | { | |
9346 | sh_wassign_func_t *assign_func; | |
9347 | int is_special_builtin, is_builtin_or_func; | |
9348 | ||
9349 | /* If the remainder of the words expand to nothing, Posix.2 requires | |
9350 | that the variable and environment assignments affect the shell's | |
9351 | environment. */ | |
9352 | assign_func = new_list ? assign_in_env : do_word_assignment; | |
9353 | tempenv_assign_error = 0; | |
9354 | ||
9355 | is_builtin_or_func = (new_list && new_list->word && (find_shell_builtin (new_list->word->word) || find_function (new_list->word->word))); | |
9356 | /* Posix says that special builtins exit if a variable assignment error | |
9357 | occurs in an assignment preceding it. */ | |
9358 | is_special_builtin = (posixly_correct && new_list && new_list->word && find_special_builtin (new_list->word->word)); | |
9359 | ||
9360 | for (temp_list = subst_assign_varlist; temp_list; temp_list = temp_list->next) | |
9361 | { | |
9362 | this_command_name = (char *)NULL; | |
9363 | assigning_in_environment = (assign_func == assign_in_env); | |
9364 | tint = (*assign_func) (temp_list->word, is_builtin_or_func); | |
9365 | assigning_in_environment = 0; | |
9366 | /* Variable assignment errors in non-interactive shells running | |
9367 | in Posix.2 mode cause the shell to exit. */ | |
9368 | if (tint == 0) | |
9369 | { | |
9370 | if (assign_func == do_word_assignment) | |
9371 | { | |
9372 | last_command_exit_value = EXECUTION_FAILURE; | |
9373 | if (interactive_shell == 0 && posixly_correct && is_special_builtin) | |
9374 | exp_jump_to_top_level (FORCE_EOF); | |
9375 | else | |
9376 | exp_jump_to_top_level (DISCARD); | |
9377 | } | |
9378 | else | |
9379 | tempenv_assign_error++; | |
9380 | } | |
9381 | } | |
9382 | ||
9383 | dispose_words (subst_assign_varlist); | |
9384 | subst_assign_varlist = (WORD_LIST *)NULL; | |
9385 | } | |
9386 | ||
9387 | #if 0 | |
9388 | tint = list_length (new_list) + 1; | |
9389 | RESIZE_MALLOCED_BUFFER (glob_argv_flags, 0, tint, glob_argv_flags_size, 16); | |
9390 | for (tint = 0, temp_list = new_list; temp_list; temp_list = temp_list->next) | |
9391 | glob_argv_flags[tint++] = (temp_list->word->flags & W_GLOBEXP) ? '1' : '0'; | |
9392 | glob_argv_flags[tint] = '\0'; | |
9393 | #endif | |
9394 | ||
9395 | return (new_list); | |
9396 | } |