?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Applied patch from Jim Meyering; rename dirfd to dir_fd to avoid shadowing problem
[findutils.git] / find / parser.c
blob246c4cee8b1e77407e7e614d5095ab30c2814885
1 /* parser.c -- convert the command line args into an expression tree.
2 Copyright (C) 1990, 1991, 1992, 1993, 1994, 2000, 2001, 2003,
3 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
4
5 This program is free software: you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation, either version 3 of the License, or
8 (at your option) any later version.
9
10 This program is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 GNU General Public License for more details.
14
15 You should have received a copy of the GNU General Public License
16 along with this program. If not, see <http://www.gnu.org/licenses/>.
17 */
18
19 #include <config.h>
20
21 #include "defs.h"
22 #include <ctype.h>
23 #include <math.h>
24 #include <assert.h>
25 #include <pwd.h>
26 #include <errno.h>
27 #include <grp.h>
28 #include <fnmatch.h>
29 #include "modechange.h"
30 #include "modetype.h"
31 #include "xstrtol.h"
32 #include "xalloc.h"
33 #include "quote.h"
34 #include "quotearg.h"
35 #include "buildcmd.h"
36 #include "nextelem.h"
37 #include "stdio-safer.h"
38 #include "regextype.h"
39 #include "stat-time.h"
40 #include "xstrtod.h"
41 #include "fts_.h"
42 #include "getdate.h"
43 #include "error.h"
44 #include "findutils-version.h"
45
46 #include <fcntl.h>
47
48
49 /* The presence of unistd.h is assumed by gnulib these days, so we
50 * might as well assume it too.
51 */
52 /* We need <unistd.h> for isatty(). */
53 #include <unistd.h>
54 #include <sys/stat.h>
55
56 #if ENABLE_NLS
57 # include <libintl.h>
58 # define _(Text) gettext (Text)
59 #else
60 # define _(Text) Text
61 #endif
62 #ifdef gettext_noop
63 # define N_(String) gettext_noop (String)
64 #else
65 /* See locate.c for explanation as to why not use (String) */
66 # define N_(String) String
67 #endif
68
69 #if !defined (isascii) || defined (STDC_HEADERS)
70 #ifdef isascii
71 #undef isascii
72 #endif
73 #define isascii(c) 1
74 #endif
75
76 #define ISDIGIT(c) (isascii ((unsigned char)c) && isdigit ((unsigned char)c))
77 #define ISUPPER(c) (isascii ((unsigned char)c) && isupper ((unsigned char)c))
78
79 #ifndef HAVE_ENDGRENT
80 #define endgrent()
81 #endif
82 #ifndef HAVE_ENDPWENT
83 #define endpwent()
84 #endif
85 \f
86 static boolean parse_accesscheck PARAMS((const struct parser_table* entry, char **argv, int *arg_ptr));
87 static boolean parse_amin PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
88 static boolean parse_and PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
89 static boolean parse_anewer PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
90 static boolean parse_cmin PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
91 static boolean parse_cnewer PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
92 static boolean parse_comma PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
93 static boolean parse_daystart PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
94 static boolean parse_delete PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
95 static boolean parse_d PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
96 static boolean parse_depth PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
97 static boolean parse_empty PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
98 static boolean parse_exec PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
99 static boolean parse_execdir PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
100 static boolean parse_false PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
101 static boolean parse_fls PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
102 static boolean parse_fprintf PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
103 static boolean parse_follow PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
104 static boolean parse_fprint PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
105 static boolean parse_fprint0 PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
106 static boolean parse_fstype PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
107 static boolean parse_gid PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
108 static boolean parse_group PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
109 static boolean parse_help PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
110 static boolean parse_ilname PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
111 static boolean parse_iname PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
112 static boolean parse_inum PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
113 static boolean parse_ipath PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
114 static boolean parse_iregex PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
115 static boolean parse_iwholename PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
116 static boolean parse_links PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
117 static boolean parse_lname PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
118 static boolean parse_ls PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
119 static boolean parse_maxdepth PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
120 static boolean parse_mindepth PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
121 static boolean parse_mmin PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
122 static boolean parse_name PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
123 static boolean parse_negate PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
124 static boolean parse_newer PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
125 static boolean parse_newerXY PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
126 static boolean parse_noleaf PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
127 static boolean parse_nogroup PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
128 static boolean parse_nouser PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
129 static boolean parse_nowarn PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
130 static boolean parse_ok PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
131 static boolean parse_okdir PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
132 static boolean parse_or PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
133 static boolean parse_path PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
134 static boolean parse_perm PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
135 static boolean parse_print0 PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
136 static boolean parse_printf PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
137 static boolean parse_prune PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
138 static boolean parse_regex PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
139 static boolean parse_regextype PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
140 static boolean parse_samefile PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
141 #if 0
142 static boolean parse_show_control_chars PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
143 #endif
144 static boolean parse_size PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
145 static boolean parse_time PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
146 static boolean parse_true PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
147 static boolean parse_type PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
148 static boolean parse_uid PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
149 static boolean parse_used PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
150 static boolean parse_user PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
151 static boolean parse_version PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
152 static boolean parse_wholename PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
153 static boolean parse_xdev PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
154 static boolean parse_ignore_race PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
155 static boolean parse_noignore_race PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
156 static boolean parse_warn PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
157 static boolean parse_xtype PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
158 static boolean parse_quit PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
159
160 boolean parse_print PARAMS((const struct parser_table*, char *argv[], int *arg_ptr));
161
162
163 static boolean insert_type PARAMS((char **argv, int *arg_ptr,
164 const struct parser_table *entry,
165 PRED_FUNC which_pred));
166 static boolean insert_regex PARAMS((char *argv[], int *arg_ptr,
167 const struct parser_table *entry,
168 int regex_options));
169 static boolean insert_fprintf (struct format_val *vec,
170 const struct parser_table *entry,
171 PRED_FUNC func,
172 const char *format);
173
174 static struct segment **make_segment PARAMS((struct segment **segment,
175 char *format, int len,
176 int kind, char format_char,
177 char aux_format_char,
178 struct predicate *pred));
179 static boolean insert_exec_ok PARAMS((const char *action,
180 const struct parser_table *entry,
181 int dir_fd,
182 char *argv[],
183 int *arg_ptr));
184 static boolean get_comp_type PARAMS((const char **str,
185 enum comparison_type *comp_type));
186 static boolean get_relative_timestamp PARAMS((const char *str,
187 struct time_val *tval,
188 struct timespec origin,
189 double sec_per_unit,
190 const char *overflowmessage));
191 static boolean get_num PARAMS((const char *str,
192 uintmax_t *num,
193 enum comparison_type *comp_type));
194 static struct predicate* insert_num PARAMS((char *argv[], int *arg_ptr,
195 const struct parser_table *entry));
196 static void open_output_file (const char *path, struct format_val *p);
197 static void open_stdout (struct format_val *p);
198 static boolean stream_is_tty(FILE *fp);
199 static boolean parse_noop PARAMS((const struct parser_table* entry,
200 char **argv, int *arg_ptr));
201
202 #define PASTE(x,y) x##y
203 #define STRINGIFY(s) #s
204
205 #define PARSE_OPTION(what,suffix) \
206 { (ARG_OPTION), (what), PASTE(parse_,suffix), NULL }
207
208 #define PARSE_POSOPT(what,suffix) \
209 { (ARG_POSITIONAL_OPTION), (what), PASTE(parse_,suffix), NULL }
210
211 #define PARSE_TEST(what,suffix) \
212 { (ARG_TEST), (what), PASTE(parse_,suffix), PASTE(pred_,suffix) }
213
214 #define PARSE_TEST_NP(what,suffix) \
215 { (ARG_TEST), (what), PASTE(parse_,suffix), NULL }
216
217 #define PARSE_ACTION(what,suffix) \
218 { (ARG_ACTION), (what), PASTE(parse_,suffix), PASTE(pred_,suffix) }
219
220 #define PARSE_ACTION_NP(what,suffix) \
221 { (ARG_ACTION), (what), PASTE(parse_,suffix), NULL }
222
223 #define PARSE_PUNCTUATION(what,suffix) \
224 { (ARG_PUNCTUATION), (what), PASTE(parse_,suffix), PASTE(pred_,suffix) }
225
226
227 /* Predicates we cannot handle in the usual way. If you add an entry
228 * to this table, double-check the switch statement in
229 * pred_sanity_check() to make sure that the new case is being
230 * correctly handled.
231 */
232 static struct parser_table const parse_entry_newerXY =
233 {
234 ARG_SPECIAL_PARSE, "newerXY", parse_newerXY, pred_newerXY /* BSD */
235 };
236
237 /* GNU find predicates that are not mentioned in POSIX.2 are marked `GNU'.
238 If they are in some Unix versions of find, they are marked `Unix'. */
239
240 static struct parser_table const parse_table[] =
241 {
242 PARSE_PUNCTUATION("!", negate), /* POSIX */
243 PARSE_PUNCTUATION("not", negate), /* GNU */
244 PARSE_PUNCTUATION("(", openparen), /* POSIX */
245 PARSE_PUNCTUATION(")", closeparen), /* POSIX */
246 PARSE_PUNCTUATION(",", comma), /* GNU */
247 PARSE_PUNCTUATION("a", and), /* POSIX */
248 PARSE_TEST ("amin", amin), /* GNU */
249 PARSE_PUNCTUATION("and", and), /* GNU */
250 PARSE_TEST ("anewer", anewer), /* GNU */
251 {ARG_TEST, "atime", parse_time, pred_atime}, /* POSIX */
252 PARSE_TEST ("cmin", cmin), /* GNU */
253 PARSE_TEST ("cnewer", cnewer), /* GNU */
254 {ARG_TEST, "ctime", parse_time, pred_ctime}, /* POSIX */
255 PARSE_POSOPT ("daystart", daystart), /* GNU */
256 PARSE_ACTION ("delete", delete), /* GNU, Mac OS, FreeBSD */
257 PARSE_OPTION ("d", d), /* Mac OS X, FreeBSD, NetBSD, OpenBSD, but deprecated in favour of -depth */
258 PARSE_OPTION ("depth", depth), /* POSIX */
259 PARSE_TEST ("empty", empty), /* GNU */
260 {ARG_ACTION, "exec", parse_exec, pred_exec}, /* POSIX */
261 {ARG_TEST, "executable", parse_accesscheck, pred_executable}, /* GNU, 4.3.0+ */
262 PARSE_ACTION ("execdir", execdir), /* *BSD, GNU */
263 PARSE_ACTION ("fls", fls), /* GNU */
264 PARSE_POSOPT ("follow", follow), /* GNU, Unix */
265 PARSE_ACTION ("fprint", fprint), /* GNU */
266 PARSE_ACTION ("fprint0", fprint0), /* GNU */
267 {ARG_ACTION, "fprintf", parse_fprintf, pred_fprintf}, /* GNU */
268 PARSE_TEST ("fstype", fstype), /* GNU, Unix */
269 PARSE_TEST ("gid", gid), /* GNU */
270 PARSE_TEST ("group", group), /* POSIX */
271 PARSE_OPTION ("ignore_readdir_race", ignore_race), /* GNU */
272 PARSE_TEST ("ilname", ilname), /* GNU */
273 PARSE_TEST ("iname", iname), /* GNU */
274 PARSE_TEST ("inum", inum), /* GNU, Unix */
275 PARSE_TEST ("ipath", ipath), /* GNU, deprecated in favour of iwholename */
276 PARSE_TEST_NP ("iregex", iregex), /* GNU */
277 PARSE_TEST_NP ("iwholename", iwholename), /* GNU */
278 PARSE_TEST ("links", links), /* POSIX */
279 PARSE_TEST ("lname", lname), /* GNU */
280 PARSE_ACTION ("ls", ls), /* GNU, Unix */
281 PARSE_OPTION ("maxdepth", maxdepth), /* GNU */
282 PARSE_OPTION ("mindepth", mindepth), /* GNU */
283 PARSE_TEST ("mmin", mmin), /* GNU */
284 PARSE_OPTION ("mount", xdev), /* Unix */
285 {ARG_TEST, "mtime", parse_time, pred_mtime}, /* POSIX */
286 PARSE_TEST ("name", name),
287 #ifdef UNIMPLEMENTED_UNIX
288 PARSE(ARG_UNIMPLEMENTED, "ncpio", ncpio), /* Unix */
289 #endif
290 PARSE_TEST ("newer", newer), /* POSIX */
291 {ARG_TEST, "atime", parse_time, pred_atime}, /* POSIX */
292 PARSE_OPTION ("noleaf", noleaf), /* GNU */
293 PARSE_TEST ("nogroup", nogroup), /* POSIX */
294 PARSE_TEST ("nouser", nouser), /* POSIX */
295 PARSE_OPTION ("noignore_readdir_race", noignore_race), /* GNU */
296 PARSE_POSOPT ("nowarn", nowarn), /* GNU */
297 PARSE_PUNCTUATION("o", or), /* POSIX */
298 PARSE_PUNCTUATION("or", or), /* GNU */
299 PARSE_ACTION ("ok", ok), /* POSIX */
300 PARSE_ACTION ("okdir", okdir), /* GNU (-execdir is BSD) */
301 PARSE_TEST ("path", path), /* GNU, HP-UX, RMS prefers wholename, but anyway soon POSIX */
302 PARSE_TEST ("perm", perm), /* POSIX */
303 PARSE_ACTION ("print", print), /* POSIX */
304 PARSE_ACTION ("print0", print0), /* GNU */
305 {ARG_ACTION, "printf", parse_printf, NULL}, /* GNU */
306 PARSE_ACTION ("prune", prune), /* POSIX */
307 PARSE_ACTION ("quit", quit), /* GNU */
308 {ARG_TEST, "readable", parse_accesscheck, pred_readable}, /* GNU, 4.3.0+ */
309 PARSE_TEST ("regex", regex), /* GNU */
310 PARSE_OPTION ("regextype", regextype), /* GNU */
311 PARSE_TEST ("samefile", samefile), /* GNU */
312 #if 0
313 PARSE_OPTION ("show-control-chars", show_control_chars), /* GNU, 4.3.0+ */
314 #endif
315 PARSE_TEST ("size", size), /* POSIX */
316 PARSE_TEST ("type", type), /* POSIX */
317 PARSE_TEST ("uid", uid), /* GNU */
318 PARSE_TEST ("used", used), /* GNU */
319 PARSE_TEST ("user", user), /* POSIX */
320 PARSE_OPTION ("warn", warn), /* GNU */
321 PARSE_TEST_NP ("wholename", wholename), /* GNU, replaced -path, but anyway -path will soon be in POSIX */
322 {ARG_TEST, "writable", parse_accesscheck, pred_writable}, /* GNU, 4.3.0+ */
323 PARSE_OPTION ("xdev", xdev), /* POSIX */
324 PARSE_TEST ("xtype", xtype), /* GNU */
325 #ifdef UNIMPLEMENTED_UNIX
326 /* It's pretty ugly for find to know about archive formats.
327 Plus what it could do with cpio archives is very limited.
328 Better to leave it out. */
329 PARSE(ARG_UNIMPLEMENTED, "cpio", cpio), /* Unix */
330 #endif
331 /* gnulib's stdbool.h might have made true and false into macros,
332 * so we can't leave named 'true' and 'false' tokens, so we have
333 * to expeant the relevant entries longhand.
334 */
335 {ARG_TEST, "false", parse_false, pred_false}, /* GNU */
336 {ARG_TEST, "true", parse_true, pred_true }, /* GNU */
337 {ARG_NOOP, "noop", NULL, pred_true }, /* GNU, internal use only */
338
339 /* Various other cases that don't fit neatly into our macro scheme. */
340 {ARG_TEST, "help", parse_help, NULL}, /* GNU */
341 {ARG_TEST, "-help", parse_help, NULL}, /* GNU */
342 {ARG_TEST, "version", parse_version, NULL}, /* GNU */
343 {ARG_TEST, "-version", parse_version, NULL}, /* GNU */
344 {0, 0, 0, 0}
345 };
346 \f
347
348 static const char *first_nonoption_arg = NULL;
349 static const struct parser_table *noop = NULL;
350
351
352 void
353 check_option_combinations(const struct predicate *p)
354 {
355 enum { seen_delete=1u, seen_prune=2u };
356 unsigned int predicates = 0u;
357
358 while (p)
359 {
360 if (p->pred_func == pred_delete)
361 predicates |= seen_delete;
362 else if (p->pred_func == pred_prune)
363 predicates |= seen_prune;
364 p = p->pred_next;
365 }
366
367 if ((predicates & seen_prune) && (predicates & seen_delete))
368 {
369 /* The user specified both -delete and -prune. One might test
370 * this by first doing
371 * find dirs .... -prune ..... -print
372 * to fnd out what's going to get deleted, and then switch to
373 * find dirs .... -prune ..... -delete
374 * once we are happy. Unfortunately, the -delete action also
375 * implicitly turns on -depth, which will affect the behaviour
376 * of -prune (in fact, it makes it a no-op). In this case we
377 * would like to prevent unfortunate accidents, so we require
378 * the user to have explicitly used -depth.
379 *
380 * We only get away with this because the -delete predicate is not
381 * in POSIX. If it was, we couldn't issue a fatal error here.
382 */
383 if (!options.explicit_depth)
384 {
385 /* This fixes Savannah bug #20865. */
386 error (1, 0, _("The -delete action atomatically turns on -depth, "
387 "but -prune does nothing when -depth is in effect. "
388 "If you want to carry on anyway, just explicitly use "
389 "the -depth option."));
390 }
391 }
392 }
393
394
395 static const struct parser_table*
396 get_noop(void)
397 {
398 int i;
399 if (NULL == noop)
400 {
401 for (i = 0; parse_table[i].parser_name != 0; i++)
402 {
403 if (ARG_NOOP ==parse_table[i].type)
404 {
405 noop = &(parse_table[i]);
406 break;
407 }
408 }
409 }
410 return noop;
411 }
412 \f
413 static int
414 get_stat_Ytime(const struct stat *p,
415 char what,
416 struct timespec *ret)
417 {
418 switch (what)
419 {
420 case 'a':
421 *ret = get_stat_atime(p);
422 return 1;
423 case 'B':
424 *ret = get_stat_birthtime(p);
425 return (ret->tv_nsec >= 0);
426 case 'c':
427 *ret = get_stat_ctime(p);
428 return 1;
429 case 'm':
430 *ret = get_stat_mtime(p);
431 return 1;
432 default:
433 assert (0);
434 abort();
435 }
436 }
437 \f
438 void
439 set_follow_state(enum SymlinkOption opt)
440 {
441 if (options.debug_options & DebugStat)
442 {
443 /* For DebugStat, the choice is made at runtime within debug_stat()
444 * by checking the contents of the symlink_handling variable.
445 */
446 options.xstat = debug_stat;
447 }
448 else
449 {
450 switch (opt)
451 {
452 case SYMLINK_ALWAYS_DEREF: /* -L */
453 options.xstat = optionl_stat;
454 options.no_leaf_check = true;
455 break;
456
457 case SYMLINK_NEVER_DEREF: /* -P (default) */
458 options.xstat = optionp_stat;
459 /* Can't turn no_leaf_check off because the user might have specified
460 * -noleaf anyway
461 */
462 break;
463
464 case SYMLINK_DEREF_ARGSONLY: /* -H */
465 options.xstat = optionh_stat;
466 options.no_leaf_check = true;
467 }
468 }
469 options.symlink_handling = opt;
470 }
471
472
473 void
474 parse_begin_user_args (char **args, int argno,
475 const struct predicate *last,
476 const struct predicate *predicates)
477 {
478 (void) args;
479 (void) argno;
480 (void) last;
481 (void) predicates;
482 first_nonoption_arg = NULL;
483 }
484
485 void
486 parse_end_user_args (char **args, int argno,
487 const struct predicate *last,
488 const struct predicate *predicates)
489 {
490 /* does nothing */
491 (void) args;
492 (void) argno;
493 (void) last;
494 (void) predicates;
495 }
496
497
498 /* Check that it is legal to fid the given primary in its
499 * position and return it.
500 */
501 const struct parser_table*
502 found_parser(const char *original_arg, const struct parser_table *entry)
503 {
504 /* If this is an option, but we have already had a
505 * non-option argument, the user may be under the
506 * impression that the behaviour of the option
507 * argument is conditional on some preceding
508 * tests. This might typically be the case with,
509 * for example, -maxdepth.
510 *
511 * The options -daystart and -follow are exempt
512 * from this treatment, since their positioning
513 * in the command line does have an effect on
514 * subsequent tests but not previous ones. That
515 * might be intentional on the part of the user.
516 */
517 if (entry->type != ARG_POSITIONAL_OPTION)
518 {
519 /* Something other than -follow/-daystart.
520 * If this is an option, check if it followed
521 * a non-option and if so, issue a warning.
522 */
523 if (entry->type == ARG_OPTION)
524 {
525 if ((first_nonoption_arg != NULL)
526 && options.warnings )
527 {
528 /* option which follows a non-option */
529 error (0, 0,
530 _("warning: you have specified the %s "
531 "option after a non-option argument %s, "
532 "but options are not positional (%s affects "
533 "tests specified before it as well as those "
534 "specified after it). Please specify options "
535 "before other arguments.\n"),
536 original_arg,
537 first_nonoption_arg,
538 original_arg);
539 }
540 }
541 else
542 {
543 /* Not an option or a positional option,
544 * so remember we've seen it in order to
545 * use it in a possible future warning message.
546 */
547 if (first_nonoption_arg == NULL)
548 {
549 first_nonoption_arg = original_arg;
550 }
551 }
552 }
553
554 return entry;
555 }
556
557
558 /* Return a pointer to the parser function to invoke for predicate
559 SEARCH_NAME.
560 Return NULL if SEARCH_NAME is not a valid predicate name. */
561
562 const struct parser_table*
563 find_parser (char *search_name)
564 {
565 int i;
566 const char *original_arg = search_name;
567
568 /* Ugh. Special case -newerXY. */
569 if (0 == strncmp("-newer", search_name, 6)
570 && (8 == strlen(search_name)))
571 {
572 return found_parser(original_arg, &parse_entry_newerXY);
573 }
574
575 if (*search_name == '-')
576 search_name++;
577
578 for (i = 0; parse_table[i].parser_name != 0; i++)
579 {
580 if (strcmp (parse_table[i].parser_name, search_name) == 0)
581 {
582 return found_parser(original_arg, &parse_table[i]);
583 }
584 }
585 return NULL;
586 }
587 \f
588 static float
589 estimate_file_age_success_rate(float num_days)
590 {
591 if (num_days < 0.1)
592 {
593 /* Assume 1% of files have timestamps in the future */
594 return 0.01f;
595 }
596 else if (num_days < 1)
597 {
598 /* Assume 30% of files have timestamps today */
599 return 0.3f;
600 }
601 else if (num_days > 100)
602 {
603 /* Assume 30% of files are very old */
604 return 0.3f;
605 }
606 else
607 {
608 /* Assume 39% of files are between 1 and 100 days old. */
609 return 0.39f;
610 }
611 }
612 \f
613 static float
614 estimate_timestamp_success_rate(time_t when)
615 {
616 /* This calculation ignores the nanoseconds field of the
617 * origin, but I don't think that makes much difference
618 * to our estimate.
619 */
620 int num_days = (options.cur_day_start.tv_sec - when) / 86400;
621 return estimate_file_age_success_rate(num_days);
622 }
623
624 /* Collect an argument from the argument list, or
625 * return false.
626 */
627 static boolean
628 collect_arg(char **argv, int *arg_ptr, const char **collected_arg)
629 {
630 if ((argv == NULL) || (argv[*arg_ptr] == NULL))
631 {
632 *collected_arg = NULL;
633 return false;
634 }
635 else
636 {
637 *collected_arg = argv[*arg_ptr];
638 (*arg_ptr)++;
639 return true;
640 }
641 }
642
643 static boolean
644 collect_arg_stat_info(char **argv, int *arg_ptr, struct stat *p)
645 {
646 const char *filename;
647 if (collect_arg(argv, arg_ptr, &filename))
648 {
649 if (0 == (options.xstat)(filename, p))
650 {
651 return true;
652 }
653 else
654 {
655 fatal_file_error(filename);
656 }
657 }
658 else
659 {
660 return false;
661 }
662 }
663
664 /* The parsers are responsible to continue scanning ARGV for
665 their arguments. Each parser knows what is and isn't
666 allowed for itself.
667
668 ARGV is the argument array.
669 *ARG_PTR is the index to start at in ARGV,
670 updated to point beyond the last element consumed.
671
672 The predicate structure is updated with the new information. */
673
674 \f
675 static boolean
676 parse_and (const struct parser_table* entry, char **argv, int *arg_ptr)
677 {
678 struct predicate *our_pred;
679
680 (void) argv;
681 (void) arg_ptr;
682
683 our_pred = get_new_pred (entry);
684 our_pred->pred_func = pred_and;
685 our_pred->p_type = BI_OP;
686 our_pred->p_prec = AND_PREC;
687 our_pred->need_stat = our_pred->need_type = false;
688 return true;
689 }
690
691 static boolean
692 parse_anewer (const struct parser_table* entry, char **argv, int *arg_ptr)
693 {
694 struct stat stat_newer;
695
696 set_stat_placeholders(&stat_newer);
697 if (collect_arg_stat_info(argv, arg_ptr, &stat_newer))
698 {
699 struct predicate *our_pred = insert_primary (entry);
700 our_pred->args.reftime.xval = XVAL_ATIME;
701 our_pred->args.reftime.ts = get_stat_mtime(&stat_newer);
702 our_pred->args.reftime.kind = COMP_GT;
703 our_pred->est_success_rate = estimate_timestamp_success_rate(stat_newer.st_mtime);
704 return true;
705 }
706 return false;
707 }
708
709 boolean
710 parse_closeparen (const struct parser_table* entry, char **argv, int *arg_ptr)
711 {
712 struct predicate *our_pred;
713
714 (void) argv;
715 (void) arg_ptr;
716
717 our_pred = get_new_pred (entry);
718 our_pred->pred_func = pred_closeparen;
719 our_pred->p_type = CLOSE_PAREN;
720 our_pred->p_prec = NO_PREC;
721 our_pred->need_stat = our_pred->need_type = false;
722 return true;
723 }
724
725 static boolean
726 parse_cnewer (const struct parser_table* entry, char **argv, int *arg_ptr)
727 {
728 struct stat stat_newer;
729
730 set_stat_placeholders(&stat_newer);
731 if (collect_arg_stat_info(argv, arg_ptr, &stat_newer))
732 {
733 struct predicate *our_pred = insert_primary (entry);
734 our_pred->args.reftime.xval = XVAL_CTIME; /* like -newercm */
735 our_pred->args.reftime.ts = get_stat_mtime(&stat_newer);
736 our_pred->args.reftime.kind = COMP_GT;
737 our_pred->est_success_rate = estimate_timestamp_success_rate(stat_newer.st_mtime);
738 return true;
739 }
740 return false;
741 }
742
743 static boolean
744 parse_comma (const struct parser_table* entry, char **argv, int *arg_ptr)
745 {
746 struct predicate *our_pred;
747
748 (void) argv;
749 (void) arg_ptr;
750
751 our_pred = get_new_pred (entry);
752 our_pred->pred_func = pred_comma;
753 our_pred->p_type = BI_OP;
754 our_pred->p_prec = COMMA_PREC;
755 our_pred->need_stat = our_pred->need_type = false;
756 our_pred->est_success_rate = 1.0f;
757 return true;
758 }
759
760 static boolean
761 parse_daystart (const struct parser_table* entry, char **argv, int *arg_ptr)
762 {
763 struct tm *local;
764
765 (void) entry;
766 (void) argv;
767 (void) arg_ptr;
768
769 if (options.full_days == false)
770 {
771 options.cur_day_start.tv_sec += DAYSECS;
772 options.cur_day_start.tv_nsec = 0;
773 local = localtime (&options.cur_day_start.tv_sec);
774 options.cur_day_start.tv_sec -= (local
775 ? (local->tm_sec + local->tm_min * 60
776 + local->tm_hour * 3600)
777 : options.cur_day_start.tv_sec % DAYSECS);
778 options.full_days = true;
779 }
780 return true;
781 }
782
783 static boolean
784 parse_delete (const struct parser_table* entry, char *argv[], int *arg_ptr)
785 {
786 struct predicate *our_pred;
787 (void) argv;
788 (void) arg_ptr;
789
790 our_pred = insert_primary (entry);
791 our_pred->side_effects = our_pred->no_default_print = true;
792 /* -delete implies -depth */
793 options.do_dir_first = false;
794
795 /* We do not need stat information because we check for the case
796 * (errno==EISDIR) in pred_delete.
797 */
798 our_pred->need_stat = our_pred->need_type = false;
799
800 our_pred->est_success_rate = 1.0f;
801 return true;
802 }
803
804 static boolean
805 parse_depth (const struct parser_table* entry, char **argv, int *arg_ptr)
806 {
807 (void) entry;
808 (void) argv;
809
810 options.do_dir_first = false;
811 options.explicit_depth = true;
812 return parse_noop(entry, argv, arg_ptr);
813 }
814
815 static boolean
816 parse_d (const struct parser_table* entry, char **argv, int *arg_ptr)
817 {
818 if (options.warnings)
819 {
820 error (0, 0,
821 _("warning: the -d option is deprecated; please use "
822 "-depth instead, because the latter is a "
823 "POSIX-compliant feature."));
824 }
825 return parse_depth(entry, argv, arg_ptr);
826 }
827
828 static boolean
829 parse_empty (const struct parser_table* entry, char **argv, int *arg_ptr)
830 {
831 struct predicate *our_pred;
832 (void) argv;
833 (void) arg_ptr;
834
835 our_pred = insert_primary (entry);
836 our_pred->est_success_rate = 0.01f; /* assume 1% of files are empty. */
837 return true;
838 }
839
840 static boolean
841 parse_exec (const struct parser_table* entry, char **argv, int *arg_ptr)
842 {
843 return insert_exec_ok ("-exec", entry, get_start_dirfd(), argv, arg_ptr);
844 }
845
846 static boolean
847 parse_execdir (const struct parser_table* entry, char **argv, int *arg_ptr)
848 {
849 return insert_exec_ok ("-execdir", entry, -1, argv, arg_ptr);
850 }
851
852 static boolean
853 parse_false (const struct parser_table* entry, char **argv, int *arg_ptr)
854 {
855 struct predicate *our_pred;
856
857 (void) argv;
858 (void) arg_ptr;
859
860 our_pred = insert_primary (entry);
861 our_pred->need_stat = our_pred->need_type = false;
862 our_pred->side_effects = our_pred->no_default_print = false;
863 our_pred->est_success_rate = 0.0f;
864 return true;
865 }
866
867 static boolean
868 insert_fls (const struct parser_table* entry, const char *filename)
869 {
870 struct predicate *our_pred = insert_primary (entry);
871 if (filename)
872 open_output_file (filename, &our_pred->args.printf_vec);
873 else
874 open_stdout (&our_pred->args.printf_vec);
875 our_pred->side_effects = our_pred->no_default_print = true;
876 our_pred->est_success_rate = 1.0f;
877 return true;
878 }
879
880
881 static boolean
882 parse_fls (const struct parser_table* entry, char **argv, int *arg_ptr)
883 {
884 const char *filename;
885 return collect_arg(argv, arg_ptr, &filename)
886 && insert_fls(entry, filename);
887 }
888
889 static boolean
890 parse_follow (const struct parser_table* entry, char **argv, int *arg_ptr)
891 {
892 set_follow_state(SYMLINK_ALWAYS_DEREF);
893 return parse_noop(entry, argv, arg_ptr);
894 }
895
896 static boolean
897 parse_fprint (const struct parser_table* entry, char **argv, int *arg_ptr)
898 {
899 struct predicate *our_pred;
900 const char *filename;
901 if (collect_arg(argv, arg_ptr, &filename))
902 {
903 our_pred = insert_primary (entry);
904 open_output_file (filename, &our_pred->args.printf_vec);
905 our_pred->side_effects = our_pred->no_default_print = true;
906 our_pred->need_stat = our_pred->need_type = false;
907 our_pred->est_success_rate = 1.0f;
908 return true;
909 }
910 else
911 {
912 return false;
913 }
914 }
915
916 static boolean
917 insert_fprint(const struct parser_table* entry, const char *filename)
918 {
919 struct predicate *our_pred = insert_primary (entry);
920 if (filename)
921 open_output_file (filename, &our_pred->args.printf_vec);
922 else
923 open_stdout (&our_pred->args.printf_vec);
924 our_pred->side_effects = our_pred->no_default_print = true;
925 our_pred->need_stat = our_pred->need_type = false;
926 our_pred->est_success_rate = 1.0f;
927 return true;
928 }
929
930
931 static boolean
932 parse_fprint0 (const struct parser_table* entry, char **argv, int *arg_ptr)
933 {
934 const char *filename;
935 if (collect_arg(argv, arg_ptr, &filename))
936 return insert_fprint(entry, filename);
937 else
938 return false;
939 }
940
941 static float estimate_fstype_success_rate(const char *fsname)
942 {
943 struct stat dir_stat;
944 const char *dir = "/";
945 if (0 == stat(dir, &dir_stat))
946 {
947 const char *fstype = filesystem_type(&dir_stat, dir);
948 /* Assume most files are on the same file system type as the root fs. */
949 if (0 == strcmp(fsname, fstype))
950 return 0.7f;
951 else
952 return 0.3f;
953 }
954 return 1.0f;
955 }
956
957
958 static boolean
959 parse_fstype (const struct parser_table* entry, char **argv, int *arg_ptr)
960 {
961 const char *typename;
962 if (collect_arg(argv, arg_ptr, &typename))
963 {
964 struct predicate *our_pred = insert_primary (entry);
965 our_pred->args.str = typename;
966
967 /* This is an expensive operation, so although there are
968 * circumstances where it is selective, we ignore this fact
969 * because we probably don't want to promote this test to the
970 * front anyway.
971 */
972 our_pred->est_success_rate = estimate_fstype_success_rate(typename);
973 return true;
974 }
975 else
976 {
977 return false;
978 }
979 }
980
981 static boolean
982 parse_gid (const struct parser_table* entry, char **argv, int *arg_ptr)
983 {
984 struct predicate *p = insert_num (argv, arg_ptr, entry);
985 if (p)
986 {
987 p->est_success_rate = (p->args.numinfo.l_val < 100) ? 0.99 : 0.2;
988 return true;
989 }
990 else
991 {
992 return false;
993 }
994 }
995
996
997 static int
998 safe_atoi (const char *s)
999 {
1000 long lval;
1001 char *end;
1002
1003 errno = 0;
1004 lval = strtol(s, &end, 10);
1005 if ( (LONG_MAX == lval) || (LONG_MIN == lval) )
1006 {
1007 /* max/min possible value, or an error. */
1008 if (errno == ERANGE)
1009 {
1010 /* too big, or too small. */
1011 error(1, errno, "%s", s);
1012 }
1013 else
1014 {
1015 /* not a valid number */
1016 error(1, errno, "%s", s);
1017 }
1018 /* Otherwise, we do a range chack against INT_MAX and INT_MIN
1019 * below.
1020 */
1021 }
1022
1023 if (lval > INT_MAX || lval < INT_MIN)
1024 {
1025 /* The number was in range for long, but not int. */
1026 errno = ERANGE;
1027 error(1, errno, "%s", s);
1028 }
1029 else if (*end)
1030 {
1031 error(1, errno, "Unexpected suffix %s on %s",
1032 quotearg_n_style(0, options.err_quoting_style, end),
1033 quotearg_n_style(1, options.err_quoting_style, s));
1034 }
1035 else if (end == s)
1036 {
1037 error(1, errno, "Expected an integer: %s",
1038 quotearg_n_style(0, options.err_quoting_style, s));
1039 }
1040 return (int)lval;
1041 }
1042
1043
1044 static boolean
1045 parse_group (const struct parser_table* entry, char **argv, int *arg_ptr)
1046 {
1047 const char *groupname;
1048
1049 if (collect_arg(argv, arg_ptr, &groupname))
1050 {
1051 gid_t gid;
1052 struct predicate *our_pred;
1053 struct group *cur_gr = getgrnam(groupname);
1054 endgrent();
1055 if (cur_gr)
1056 {
1057 gid = cur_gr->gr_gid;
1058 }
1059 else
1060 {
1061 const int gid_len = strspn (groupname, "0123456789");
1062 if (gid_len)
1063 {
1064 if (groupname[gid_len] == 0)
1065 {
1066 gid = safe_atoi (groupname);
1067 }
1068 else
1069 {
1070 /* XXX: no test in test suite for this */
1071 error(1, 0, _("%s is not the name of an existing group and"
1072 " it does not look like a numeric group ID "
1073 "because it has the unexpected suffix %s"),
1074 quotearg_n_style(0, options.err_quoting_style, groupname),
1075 quotearg_n_style(1, options.err_quoting_style, groupname+gid_len));
1076 return false;
1077 }
1078 }
1079 else
1080 {
1081 if (*groupname)
1082 {
1083 /* XXX: no test in test suite for this */
1084 error(1, 0, _("%s is not the name of an existing group"),
1085 quotearg_n_style(0, options.err_quoting_style, groupname));
1086 }
1087 else
1088 {
1089 error(1, 0, _("argument to -group is empty, but should be a group name"));
1090 }
1091 return false;
1092 }
1093 }
1094 our_pred = insert_primary (entry);
1095 our_pred->args.gid = gid;
1096 our_pred->est_success_rate = (our_pred->args.numinfo.l_val < 100) ? 0.99 : 0.2;
1097 return true;
1098 }
1099 return false;
1100 }
1101
1102 static boolean
1103 parse_help (const struct parser_table* entry, char **argv, int *arg_ptr)
1104 {
1105 (void) entry;
1106 (void) argv;
1107 (void) arg_ptr;
1108
1109 usage(stdout, 0, NULL);
1110 puts (_("\n\
1111 default path is the current directory; default expression is -print\n\
1112 expression may consist of: operators, options, tests, and actions:\n"));
1113 puts (_("\
1114 operators (decreasing precedence; -and is implicit where no others are given):\n\
1115 ( EXPR ) ! EXPR -not EXPR EXPR1 -a EXPR2 EXPR1 -and EXPR2\n\
1116 EXPR1 -o EXPR2 EXPR1 -or EXPR2 EXPR1 , EXPR2\n"));
1117 puts (_("\
1118 positional options (always true): -daystart -follow -regextype\n\n\
1119 normal options (always true, specified before other expressions):\n\
1120 -depth --help -maxdepth LEVELS -mindepth LEVELS -mount -noleaf\n\
1121 --version -xdev -ignore_readdir_race -noignore_readdir_race\n"));
1122 puts (_("\
1123 tests (N can be +N or -N or N): -amin N -anewer FILE -atime N -cmin N\n\
1124 -cnewer FILE -ctime N -empty -false -fstype TYPE -gid N -group NAME\n\
1125 -ilname PATTERN -iname PATTERN -inum N -iwholename PATTERN -iregex PATTERN\n\
1126 -links N -lname PATTERN -mmin N -mtime N -name PATTERN -newer FILE"));
1127 puts (_("\
1128 -nouser -nogroup -path PATTERN -perm [+-]MODE -regex PATTERN\n\
1129 -readable -writable -executable\n\
1130 -wholename PATTERN -size N[bcwkMG] -true -type [bcdpflsD] -uid N\n\
1131 -used N -user NAME -xtype [bcdpfls]\n"));
1132 puts (_("\
1133 actions: -delete -print0 -printf FORMAT -fprintf FILE FORMAT -print \n\
1134 -fprint0 FILE -fprint FILE -ls -fls FILE -prune -quit\n\
1135 -exec COMMAND ; -exec COMMAND {} + -ok COMMAND ;\n\
1136 -execdir COMMAND ; -execdir COMMAND {} + -okdir COMMAND ;\n\
1137 "));
1138 puts (_("Report (and track progress on fixing) bugs via the findutils bug-reporting\n\
1139 page at http://savannah.gnu.org/ or, if you have no web access, by sending\n\
1140 email to <bug-findutils@gnu.org>."));
1141 exit (0);
1142 }
1143
1144 static float
1145 estimate_pattern_match_rate(const char *pattern, int is_regex)
1146 {
1147 if (strpbrk(pattern, "*?[") || (is_regex && strpbrk(pattern, ".")))
1148 {
1149 /* A wildcard; assume the pattern matches most files. */
1150 return 0.8f;
1151 }
1152 else
1153 {
1154 return 0.1f;
1155 }
1156 }
1157
1158 static boolean
1159 parse_ilname (const struct parser_table* entry, char **argv, int *arg_ptr)
1160 {
1161 const char *name;
1162 if (collect_arg(argv, arg_ptr, &name))
1163 {
1164 struct predicate *our_pred = insert_primary (entry);
1165 our_pred->args.str = name;
1166 /* Use the generic glob pattern estimator to figure out how many
1167 * links will match, but bear in mind that most files won't be links.
1168 */
1169 our_pred->est_success_rate = 0.1 * estimate_pattern_match_rate(name, 0);
1170 return true;
1171 }
1172 else
1173 {
1174 return false;
1175 }
1176 }
1177
1178
1179 /* sanity check the fnmatch() function to make sure that case folding
1180 * is supported (as opposed to just having the flag ignored).
1181 */
1182 static boolean
1183 fnmatch_sanitycheck(void)
1184 {
1185 static boolean checked = false;
1186 if (!checked)
1187 {
1188 if (0 != fnmatch("foo", "foo", 0)
1189 || 0 == fnmatch("Foo", "foo", 0)
1190 || 0 != fnmatch("Foo", "foo", FNM_CASEFOLD))
1191 {
1192 error (1, 0, _("sanity check of the fnmatch() library function failed."));
1193 return false;
1194 }
1195 checked = true;
1196 }
1197 return checked;
1198 }
1199
1200
1201 static boolean
1202 check_name_arg(const char *pred, const char *arg)
1203 {
1204 if (options.warnings && strchr(arg, '/'))
1205 {
1206 error(0, 0,_("warning: Unix filenames usually don't contain slashes "
1207 "(though pathnames do). That means that '%s %s' will "
1208 "probably evaluate to false all the time on this system. "
1209 "You might find the '-wholename' test more useful, or "
1210 "perhaps '-samefile'. Alternatively, if you are using "
1211 "GNU grep, you could "
1212 "use 'find ... -print0 | grep -FzZ %s'."),
1213 pred,
1214 safely_quote_err_filename(0, arg),
1215 safely_quote_err_filename(1, arg));
1216 }
1217 return true; /* allow it anyway */
1218 }
1219
1220
1221
1222 static boolean
1223 parse_iname (const struct parser_table* entry, char **argv, int *arg_ptr)
1224 {
1225 const char *name;
1226 fnmatch_sanitycheck();
1227 if (collect_arg(argv, arg_ptr, &name))
1228 {
1229 if (check_name_arg("-iname", name))
1230 {
1231 struct predicate *our_pred = insert_primary (entry);
1232 our_pred->need_stat = our_pred->need_type = false;
1233 our_pred->args.str = name;
1234 our_pred->est_success_rate = estimate_pattern_match_rate(name, 0);
1235 return true;
1236 }
1237 }
1238 return false;
1239 }
1240
1241 static boolean
1242 parse_inum (const struct parser_table* entry, char **argv, int *arg_ptr)
1243 {
1244 struct predicate *p = insert_num (argv, arg_ptr, entry);
1245 if (p)
1246 {
1247 /* inode number is exact match only, so very low proportions of
1248 * files match
1249 */
1250 p->est_success_rate = 1e-6;
1251 return true;
1252 }
1253 else
1254 {
1255 return false;
1256 }
1257 }
1258
1259 /* -ipath is deprecated (at RMS's request) in favour of
1260 * -iwholename. See the node "GNU Manuals" in standards.texi
1261 * for the rationale for this (basically, GNU prefers the use
1262 * of the phrase "file name" to "path name"
1263 */
1264 static boolean
1265 parse_ipath (const struct parser_table* entry, char **argv, int *arg_ptr)
1266 {
1267 const char *name;
1268
1269 fnmatch_sanitycheck ();
1270 if (collect_arg (argv, arg_ptr, &name))
1271 {
1272 struct predicate *our_pred = insert_primary_withpred (entry, pred_ipath);
1273 our_pred->need_stat = our_pred->need_type = false;
1274 our_pred->args.str = name;
1275 our_pred->est_success_rate = estimate_pattern_match_rate (name, 0);
1276 return true;
1277 }
1278 return false;
1279 }
1280
1281 static boolean
1282 parse_iwholename (const struct parser_table* entry, char **argv, int *arg_ptr)
1283 {
1284 return parse_ipath (entry, argv, arg_ptr);
1285 }
1286
1287 static boolean
1288 parse_iregex (const struct parser_table* entry, char **argv, int *arg_ptr)
1289 {
1290 return insert_regex (argv, arg_ptr, entry, RE_ICASE|options.regex_options);
1291 }
1292
1293 static boolean
1294 parse_links (const struct parser_table* entry, char **argv, int *arg_ptr)
1295 {
1296 struct predicate *p = insert_num (argv, arg_ptr, entry);
1297 if (p)
1298 {
1299 if (p->args.numinfo.l_val == 1)
1300 p->est_success_rate = 0.99;
1301 else if (p->args.numinfo.l_val == 2)
1302 p->est_success_rate = 0.01;
1303 else
1304 p->est_success_rate = 1e-3;
1305 return true;
1306 }
1307 else
1308 {
1309 return false;
1310 }
1311 }
1312
1313 static boolean
1314 parse_lname (const struct parser_table* entry, char **argv, int *arg_ptr)
1315 {
1316 const char *name;
1317 fnmatch_sanitycheck();
1318 if (collect_arg(argv, arg_ptr, &name))
1319 {
1320 struct predicate *our_pred = insert_primary (entry);
1321 our_pred->args.str = name;
1322 our_pred->est_success_rate = 0.1 * estimate_pattern_match_rate(name, 0);
1323 return true;
1324 }
1325 return false;
1326 }
1327
1328 static boolean
1329 parse_ls (const struct parser_table* entry, char **argv, int *arg_ptr)
1330 {
1331 (void) &argv;
1332 (void) &arg_ptr;
1333 return insert_fls(entry, NULL);
1334 }
1335
1336 static boolean
1337 insert_depthspec(const struct parser_table* entry, char **argv, int *arg_ptr,
1338 int *limitptr)
1339 {
1340 const char *depthstr;
1341 int depth_len;
1342 const char *predicate = argv[(*arg_ptr)-1];
1343 if (collect_arg(argv, arg_ptr, &depthstr))
1344 {
1345 depth_len = strspn (depthstr, "0123456789");
1346 if ((depth_len > 0) && (depthstr[depth_len] == 0))
1347 {
1348 (*limitptr) = safe_atoi (depthstr);
1349 if (*limitptr >= 0)
1350 {
1351 return parse_noop(entry, argv, arg_ptr);
1352 }
1353 }
1354 error(1, 0, _("Expected a positive decimal integer argument to %s, but got %s"),
1355 predicate,
1356 quotearg_n_style(0, options.err_quoting_style, depthstr));
1357 return false;
1358 }
1359 /* missing argument */
1360 return false;
1361 }
1362
1363
1364 static boolean
1365 parse_maxdepth (const struct parser_table* entry, char **argv, int *arg_ptr)
1366 {
1367 return insert_depthspec(entry, argv, arg_ptr, &options.maxdepth);
1368 }
1369
1370 static boolean
1371 parse_mindepth (const struct parser_table* entry, char **argv, int *arg_ptr)
1372 {
1373 return insert_depthspec(entry, argv, arg_ptr, &options.mindepth);
1374 }
1375 \f
1376
1377 static boolean
1378 do_parse_xmin (const struct parser_table* entry,
1379 char **argv,
1380 int *arg_ptr,
1381 enum xval xv)
1382 {
1383 const char *minutes;
1384
1385 if (collect_arg(argv, arg_ptr, &minutes))
1386 {
1387 struct time_val tval;
1388 tval.xval = xv;
1389 struct timespec origin = options.cur_day_start;
1390 origin.tv_sec += DAYSECS;
1391 if (get_relative_timestamp(minutes, &tval, origin, 60,
1392 "arithmetic overflow while converting %s "
1393 "minutes to a number of seconds"))
1394 {
1395 struct predicate *our_pred = insert_primary (entry);
1396 our_pred->args.reftime = tval;
1397 our_pred->est_success_rate = estimate_timestamp_success_rate(tval.ts.tv_sec);
1398 return true;
1399 }
1400 }
1401 return false;
1402 }
1403 static boolean
1404 parse_amin (const struct parser_table* entry, char **argv, int *arg_ptr)
1405 {
1406 return do_parse_xmin(entry, argv, arg_ptr, XVAL_ATIME);
1407 }
1408 \f
1409 static boolean
1410 parse_cmin (const struct parser_table* entry, char **argv, int *arg_ptr)
1411 {
1412 return do_parse_xmin(entry, argv, arg_ptr, XVAL_CTIME);
1413 }
1414
1415 \f
1416 static boolean
1417 parse_mmin (const struct parser_table* entry, char **argv, int *arg_ptr)
1418 {
1419 return do_parse_xmin(entry, argv, arg_ptr, XVAL_MTIME);
1420 }
1421
1422 static boolean
1423 parse_name (const struct parser_table* entry, char **argv, int *arg_ptr)
1424 {
1425 const char *name;
1426 if (collect_arg(argv, arg_ptr, &name))
1427 {
1428 fnmatch_sanitycheck();
1429 if (check_name_arg("-name", name))
1430 {
1431 struct predicate *our_pred = insert_primary (entry);
1432 our_pred->need_stat = our_pred->need_type = false;
1433 our_pred->args.str = name;
1434 our_pred->est_success_rate = estimate_pattern_match_rate(name, 0);
1435 return true;
1436 }
1437 }
1438 return false;
1439 }
1440
1441 static boolean
1442 parse_negate (const struct parser_table* entry, char **argv, int *arg_ptr)
1443 {
1444 struct predicate *our_pred;
1445
1446 (void) &argv;
1447 (void) &arg_ptr;
1448
1449 our_pred = get_new_pred_chk_op (entry);
1450 our_pred->pred_func = pred_negate;
1451 our_pred->p_type = UNI_OP;
1452 our_pred->p_prec = NEGATE_PREC;
1453 our_pred->need_stat = our_pred->need_type = false;
1454 return true;
1455 }
1456
1457 static boolean
1458 parse_newer (const struct parser_table* entry, char **argv, int *arg_ptr)
1459 {
1460 struct predicate *our_pred;
1461 struct stat stat_newer;
1462
1463 set_stat_placeholders(&stat_newer);
1464 if (collect_arg_stat_info(argv, arg_ptr, &stat_newer))
1465 {
1466 our_pred = insert_primary (entry);
1467 our_pred->args.reftime.ts = get_stat_mtime(&stat_newer);
1468 our_pred->args.reftime.xval = XVAL_MTIME;
1469 our_pred->args.reftime.kind = COMP_GT;
1470 our_pred->est_success_rate = estimate_timestamp_success_rate(stat_newer.st_mtime);
1471 return true;
1472 }
1473 return false;
1474 }
1475
1476
1477 static boolean
1478 parse_newerXY (const struct parser_table* entry, char **argv, int *arg_ptr)
1479 {
1480 (void) argv;
1481 (void) arg_ptr;
1482
1483 if ((argv == NULL) || (argv[*arg_ptr] == NULL))
1484 {
1485 return false;
1486 }
1487 else if (8u != strlen(argv[*arg_ptr]))
1488 {
1489 return false;
1490 }
1491 else
1492 {
1493 char x, y;
1494 const char validchars[] = "aBcmt";
1495
1496 assert (0 == strncmp("-newer", argv[*arg_ptr], 6));
1497 x = argv[*arg_ptr][6];
1498 y = argv[*arg_ptr][7];
1499
1500
1501 #if !defined(HAVE_STRUCT_STAT_ST_BIRTHTIME) && !defined(HAVE_STRUCT_STAT_ST_BIRTHTIMENSEC) && !defined(HAVE_STRUCT_STAT_ST_BIRTHTIMESPEC_TV_NSEC)
1502 if ('B' == x || 'B' == y)
1503 {
1504 error(0, 0,
1505 _("This system does not provide a way to find the birth time of a file."));
1506 return 0;
1507 }
1508 #endif
1509
1510 /* -newertY (for any Y) is invalid. */
1511 if (x == 't'
1512 || 0 == strchr(validchars, x)
1513 || 0 == strchr( validchars, y))
1514 {
1515 return false;
1516 }
1517 else
1518 {
1519 struct predicate *our_pred;
1520
1521 /* Because this item is ARG_SPECIAL_PARSE, we have to advance arg_ptr
1522 * past the test name (for most other tests, this is already done)
1523 */
1524 (*arg_ptr)++;
1525
1526 our_pred = insert_primary (entry);
1527
1528
1529 switch (x)
1530 {
1531 case 'a':
1532 our_pred->args.reftime.xval = XVAL_ATIME;
1533 break;
1534 case 'B':
1535 our_pred->args.reftime.xval = XVAL_BIRTHTIME;
1536 break;
1537 case 'c':
1538 our_pred->args.reftime.xval = XVAL_CTIME;
1539 break;
1540 case 'm':
1541 our_pred->args.reftime.xval = XVAL_MTIME;
1542 break;
1543 default:
1544 assert (strchr(validchars, x));
1545 assert (0);
1546 }
1547
1548 if ('t' == y)
1549 {
1550 if (!get_date(&our_pred->args.reftime.ts,
1551 argv[*arg_ptr],
1552 &options.start_time))
1553 {
1554 error(1, 0,
1555 _("I cannot figure out how to interpret %s as a date or time"),
1556 quotearg_n_style(0, options.err_quoting_style, argv[*arg_ptr]));
1557 }
1558 }
1559 else
1560 {
1561 struct stat stat_newer;
1562
1563 /* Stat the named file. */
1564 set_stat_placeholders(&stat_newer);
1565 if ((*options.xstat) (argv[*arg_ptr], &stat_newer))
1566 fatal_file_error(argv[*arg_ptr]);
1567
1568 if (!get_stat_Ytime(&stat_newer, y, &our_pred->args.reftime.ts))
1569 {
1570 /* We cannot extract a timestamp from the struct stat. */
1571 error(1, 0, _("Cannot obtain birth time of file %s"),
1572 safely_quote_err_filename(0, argv[*arg_ptr]));
1573 }
1574 }
1575 our_pred->args.reftime.kind = COMP_GT;
1576 our_pred->est_success_rate = estimate_timestamp_success_rate(our_pred->args.reftime.ts.tv_sec);
1577 (*arg_ptr)++;
1578
1579 assert (our_pred->pred_func != NULL);
1580 assert (our_pred->pred_func == pred_newerXY);
1581 assert (our_pred->need_stat);
1582 return true;
1583 }
1584 }
1585 }
1586
1587
1588 static boolean
1589 parse_noleaf (const struct parser_table* entry, char **argv, int *arg_ptr)
1590 {
1591 options.no_leaf_check = true;
1592 return parse_noop(entry, argv, arg_ptr);
1593 }
1594
1595 #ifdef CACHE_IDS
1596 /* Arbitrary amount by which to increase size
1597 of `uid_unused' and `gid_unused'. */
1598 #define ALLOC_STEP 2048
1599
1600 /* Boolean: if uid_unused[n] is nonzero, then UID n has no passwd entry. */
1601 char *uid_unused = NULL;
1602
1603 /* Number of elements in `uid_unused'. */
1604 unsigned uid_allocated;
1605
1606 /* Similar for GIDs and group entries. */
1607 char *gid_unused = NULL;
1608 unsigned gid_allocated;
1609 #endif
1610
1611 static boolean
1612 parse_nogroup (const struct parser_table* entry, char **argv, int *arg_ptr)
1613 {
1614 struct predicate *our_pred;
1615
1616 (void) &argv;
1617 (void) &arg_ptr;
1618
1619 our_pred = insert_primary (entry);
1620 our_pred->est_success_rate = 1e-4;
1621 #ifdef CACHE_IDS
1622 if (gid_unused == NULL)
1623 {
1624 struct group *gr;
1625
1626 gid_allocated = ALLOC_STEP;
1627 gid_unused = xmalloc (gid_allocated);
1628 memset (gid_unused, 1, gid_allocated);
1629 setgrent ();
1630 while ((gr = getgrent ()) != NULL)
1631 {
1632 if ((unsigned) gr->gr_gid >= gid_allocated)
1633 {
1634 unsigned new_allocated = (unsigned) gr->gr_gid + ALLOC_STEP;
1635 gid_unused = xrealloc (gid_unused, new_allocated);
1636 memset (gid_unused + gid_allocated, 1,
1637 new_allocated - gid_allocated);
1638 gid_allocated = new_allocated;
1639 }
1640 gid_unused[(unsigned) gr->gr_gid] = 0;
1641 }
1642 endgrent ();
1643 }
1644 #endif
1645 return true;
1646 }
1647
1648 static boolean
1649 parse_nouser (const struct parser_table* entry, char **argv, int *arg_ptr)
1650 {
1651 struct predicate *our_pred;
1652 (void) argv;
1653 (void) arg_ptr;
1654
1655
1656 our_pred = insert_primary (entry);
1657 our_pred->est_success_rate = 1e-3;
1658 #ifdef CACHE_IDS
1659 if (uid_unused == NULL)
1660 {
1661 struct passwd *pw;
1662
1663 uid_allocated = ALLOC_STEP;
1664 uid_unused = xmalloc (uid_allocated);
1665 memset (uid_unused, 1, uid_allocated);
1666 setpwent ();
1667 while ((pw = getpwent ()) != NULL)
1668 {
1669 if ((unsigned) pw->pw_uid >= uid_allocated)
1670 {
1671 unsigned new_allocated = (unsigned) pw->pw_uid + ALLOC_STEP;
1672 uid_unused = xrealloc (uid_unused, new_allocated);
1673 memset (uid_unused + uid_allocated, 1,
1674 new_allocated - uid_allocated);
1675 uid_allocated = new_allocated;
1676 }
1677 uid_unused[(unsigned) pw->pw_uid] = 0;
1678 }
1679 endpwent ();
1680 }
1681 #endif
1682 return true;
1683 }
1684
1685 static boolean
1686 parse_nowarn (const struct parser_table* entry, char **argv, int *arg_ptr)
1687 {
1688 options.warnings = false;
1689 return parse_noop(entry, argv, arg_ptr);
1690 }
1691
1692 static boolean
1693 parse_ok (const struct parser_table* entry, char **argv, int *arg_ptr)
1694 {
1695 return insert_exec_ok ("-ok", entry, get_start_dirfd(), argv, arg_ptr);
1696 }
1697
1698 static boolean
1699 parse_okdir (const struct parser_table* entry, char **argv, int *arg_ptr)
1700 {
1701 return insert_exec_ok ("-okdir", entry, -1, argv, arg_ptr);
1702 }
1703
1704 boolean
1705 parse_openparen (const struct parser_table* entry, char **argv, int *arg_ptr)
1706 {
1707 struct predicate *our_pred;
1708
1709 (void) argv;
1710 (void) arg_ptr;
1711
1712 our_pred = get_new_pred_chk_op (entry);
1713 our_pred->pred_func = pred_openparen;
1714 our_pred->p_type = OPEN_PAREN;
1715 our_pred->p_prec = NO_PREC;
1716 our_pred->need_stat = our_pred->need_type = false;
1717 return true;
1718 }
1719
1720 static boolean
1721 parse_or (const struct parser_table* entry, char **argv, int *arg_ptr)
1722 {
1723 struct predicate *our_pred;
1724
1725 (void) argv;
1726 (void) arg_ptr;
1727
1728 our_pred = get_new_pred (entry);
1729 our_pred->pred_func = pred_or;
1730 our_pred->p_type = BI_OP;
1731 our_pred->p_prec = OR_PREC;
1732 our_pred->need_stat = our_pred->need_type = false;
1733 return true;
1734 }
1735
1736 /* For some time, -path was deprecated (at RMS's request) in favour of
1737 * -iwholename. See the node "GNU Manuals" in standards.texi for the
1738 * rationale for this (basically, GNU prefers the use of the phrase
1739 * "file name" to "path name".
1740 *
1741 * We do not issue a warning that this usage is deprecated
1742 * since
1743 * (a) HPUX find supports this predicate also and
1744 * (b) it will soon be in POSIX anyway.
1745 */
1746 static boolean
1747 parse_path (const struct parser_table* entry, char **argv, int *arg_ptr)
1748 {
1749 const char *name;
1750 if (collect_arg(argv, arg_ptr, &name))
1751 {
1752 struct predicate *our_pred = insert_primary_withpred (entry, pred_path);
1753 our_pred->need_stat = our_pred->need_type = false;
1754 our_pred->args.str = name;
1755 our_pred->est_success_rate = estimate_pattern_match_rate (name, 0);
1756 return true;
1757 }
1758 return false;
1759 }
1760
1761 static boolean
1762 parse_wholename (const struct parser_table* entry, char **argv, int *arg_ptr)
1763 {
1764 return parse_path (entry, argv, arg_ptr);
1765 }
1766
1767 static void
1768 non_posix_mode(const char *mode)
1769 {
1770 if (options.posixly_correct)
1771 {
1772 error (1, 0, _("Mode %s is not valid when POSIXLY_CORRECT is on."),
1773 quotearg_n_style(0, options.err_quoting_style, mode));
1774 }
1775 }
1776
1777
1778 static boolean
1779 parse_perm (const struct parser_table* entry, char **argv, int *arg_ptr)
1780 {
1781 mode_t perm_val[2];
1782 float rate;
1783 int mode_start = 0;
1784 boolean havekind = false;
1785 enum permissions_type kind = PERM_EXACT;
1786 struct mode_change *change = NULL;
1787 struct predicate *our_pred;
1788 const char *perm_expr;
1789
1790 if (!collect_arg(argv, arg_ptr, &perm_expr))
1791 return false;
1792
1793 switch (perm_expr[0])
1794 {
1795 case '-':
1796 mode_start = 1;
1797 kind = PERM_AT_LEAST;
1798 havekind = true;
1799 rate = 0.2;
1800 break;
1801
1802 case '+':
1803 change = mode_compile (perm_expr);
1804 if (NULL == change)
1805 {
1806 /* Most likely the caller is an old script that is still
1807 * using the obsolete GNU syntax '-perm +MODE'. This old
1808 * syntax was withdrawn in favor of '-perm /MODE' because
1809 * it is incompatible with POSIX in some cases, but we
1810 * still support uses of it that are not incompatible with
1811 * POSIX.
1812 *
1813 * Example: POSIXLY_CORRECT=y find -perm +a+x
1814 */
1815 non_posix_mode(perm_expr);
1816
1817 /* support the previous behaviour. */
1818 mode_start = 1;
1819 kind = PERM_ANY;
1820 rate = 0.3;
1821 }
1822 else
1823 {
1824 /* This is a POSIX-compatible usage */
1825 mode_start = 0;
1826 kind = PERM_EXACT;
1827 rate = 0.1;
1828 }
1829 havekind = true;
1830 break;
1831
1832 case '/': /* GNU extension */
1833 non_posix_mode(perm_expr);
1834 mode_start = 1;
1835 kind = PERM_ANY;
1836 havekind = true;
1837 rate = 0.3;
1838 break;
1839
1840 default:
1841 /* For example, '-perm 0644', which is valid and matches
1842 * only files whose mode is exactly 0644.
1843 */
1844 mode_start = 0;
1845 kind = PERM_EXACT;
1846 havekind = true;
1847 rate = 0.01;
1848 break;
1849 }
1850
1851 if (NULL == change)
1852 {
1853 change = mode_compile (perm_expr + mode_start);
1854 if (NULL == change)
1855 error (1, 0, _("invalid mode %s"),
1856 quotearg_n_style(0, options.err_quoting_style, perm_expr));
1857 }
1858 perm_val[0] = mode_adjust (0, false, 0, change, NULL);
1859 perm_val[1] = mode_adjust (0, true, 0, change, NULL);
1860 free (change);
1861
1862 if (('/' == perm_expr[0]) && (0 == perm_val[0]) && (0 == perm_val[1]))
1863 {
1864 /* The meaning of -perm /000 will change in the future. It
1865 * currently matches no files, but like -perm -000 it should
1866 * match all files.
1867 *
1868 * Starting in 2005, we used to issue a warning message
1869 * informing the user that the behaviour would change in the
1870 * future. We have now changed the behaviour and issue a
1871 * warning message that the behaviour recently changed.
1872 */
1873 error (0, 0,
1874 _("warning: you have specified a mode pattern %s (which is "
1875 "equivalent to /000). The meaning of -perm /000 has now been "
1876 "changed to be consistent with -perm -000; that is, while it "
1877 "used to match no files, it now matches all files."),
1878 perm_expr);
1879
1880 kind = PERM_AT_LEAST;
1881 havekind = true;
1882
1883 /* The "magic" number below is just the fraction of files on my
1884 * own system that "-type l -xtype l" fails for (i.e. unbroken symlinks).
1885 * Actual totals are 1472 and 1073833.
1886 */
1887 rate = 0.9986; /* probably matches anything but a broken symlink */
1888 }
1889
1890 our_pred = insert_primary (entry);
1891 our_pred->est_success_rate = rate;
1892 if (havekind)
1893 {
1894 our_pred->args.perm.kind = kind;
1895 }
1896 else
1897 {
1898
1899 switch (perm_expr[0])
1900 {
1901 case '-':
1902 our_pred->args.perm.kind = PERM_AT_LEAST;
1903 break;
1904 case '+':
1905 our_pred->args.perm.kind = PERM_ANY;
1906 break;
1907 default:
1908 our_pred->args.perm.kind = PERM_EXACT;
1909 break;
1910 }
1911 }
1912 memcpy (our_pred->args.perm.val, perm_val, sizeof perm_val);
1913 return true;
1914 }
1915
1916 boolean
1917 parse_print (const struct parser_table* entry, char **argv, int *arg_ptr)
1918 {
1919 struct predicate *our_pred;
1920
1921 (void) argv;
1922 (void) arg_ptr;
1923
1924 our_pred = insert_primary (entry);
1925 /* -print has the side effect of printing. This prevents us
1926 from doing undesired multiple printing when the user has
1927 already specified -print. */
1928 our_pred->side_effects = our_pred->no_default_print = true;
1929 our_pred->need_stat = our_pred->need_type = false;
1930 open_stdout(&our_pred->args.printf_vec);
1931 return true;
1932 }
1933
1934 static boolean
1935 parse_print0 (const struct parser_table* entry, char **argv, int *arg_ptr)
1936 {
1937 return insert_fprint(entry, NULL);
1938 }
1939
1940 static boolean
1941 parse_printf (const struct parser_table* entry, char **argv, int *arg_ptr)
1942 {
1943 const char *format;
1944 if (collect_arg(argv, arg_ptr, &format))
1945 {
1946 struct format_val fmt;
1947 open_stdout(&fmt);
1948 return insert_fprintf (&fmt, entry, pred_fprintf, format);
1949 }
1950 return false;
1951 }
1952
1953 static boolean
1954 parse_fprintf (const struct parser_table* entry, char **argv, int *arg_ptr)
1955 {
1956 const char *format, *filename;
1957 if (collect_arg(argv, arg_ptr, &filename))
1958 {
1959 if (collect_arg(argv, arg_ptr, &format))
1960 {
1961 struct format_val fmt;
1962 open_output_file (filename, &fmt);
1963 return insert_fprintf (&fmt, entry, pred_fprintf, format);
1964 }
1965 }
1966 return false;
1967 }
1968
1969 static boolean
1970 parse_prune (const struct parser_table* entry, char **argv, int *arg_ptr)
1971 {
1972 struct predicate *our_pred;
1973
1974 (void) argv;
1975 (void) arg_ptr;
1976
1977 our_pred = insert_primary (entry);
1978 our_pred->need_stat = our_pred->need_type = false;
1979 /* -prune has a side effect that it does not descend into
1980 the current directory. */
1981 our_pred->side_effects = true;
1982 our_pred->no_default_print = false;
1983 return true;
1984 }
1985
1986 static boolean
1987 parse_quit (const struct parser_table* entry, char **argv, int *arg_ptr)
1988 {
1989 struct predicate *our_pred = insert_primary (entry);
1990 (void) argv;
1991 (void) arg_ptr;
1992 our_pred->need_stat = our_pred->need_type = false;
1993 our_pred->side_effects = true; /* Exiting is a side effect... */
1994 our_pred->no_default_print = false; /* Don't inhibit the default print, though. */
1995 our_pred->est_success_rate = 1.0f;
1996 return true;
1997 }
1998
1999
2000 static boolean
2001 parse_regextype (const struct parser_table* entry, char **argv, int *arg_ptr)
2002 {
2003 const char *type_name;
2004 if (collect_arg(argv, arg_ptr, &type_name))
2005 {
2006 /* collect the regex type name */
2007 options.regex_options = get_regex_type(type_name);
2008 return parse_noop(entry, argv, arg_ptr);
2009 }
2010 return false;
2011 }
2012
2013
2014 static boolean
2015 parse_regex (const struct parser_table* entry, char **argv, int *arg_ptr)
2016 {
2017 return insert_regex (argv, arg_ptr, entry, options.regex_options);
2018 }
2019
2020 static boolean
2021 insert_regex (char **argv,
2022 int *arg_ptr,
2023 const struct parser_table *entry,
2024 int regex_options)
2025 {
2026 const char *rx;
2027 if (collect_arg(argv, arg_ptr, &rx))
2028 {
2029 struct re_pattern_buffer *re;
2030 const char *error_message;
2031 struct predicate *our_pred = insert_primary_withpred (entry, pred_regex);
2032 our_pred->need_stat = our_pred->need_type = false;
2033 re = xmalloc (sizeof (struct re_pattern_buffer));
2034 our_pred->args.regex = re;
2035 re->allocated = 100;
2036 re->buffer = xmalloc (re->allocated);
2037 re->fastmap = NULL;
2038
2039 re_set_syntax(regex_options);
2040 re->syntax = regex_options;
2041 re->translate = NULL;
2042
2043 error_message = re_compile_pattern (rx, strlen(rx), re);
2044 if (error_message)
2045 error (1, 0, "%s", error_message);
2046 our_pred->est_success_rate = estimate_pattern_match_rate(rx, 1);
2047 return true;
2048 }
2049 return false;
2050 }
2051
2052 static boolean
2053 parse_size (const struct parser_table* entry, char **argv, int *arg_ptr)
2054 {
2055 struct predicate *our_pred;
2056 uintmax_t num;
2057 char suffix;
2058 enum comparison_type c_type;
2059
2060 int blksize = 512;
2061 int len;
2062
2063 /* XXX: cannot (yet) convert to ue collect_arg() as this
2064 * function modifies the args in-place.
2065 */
2066 if ((argv == NULL) || (argv[*arg_ptr] == NULL))
2067 return false;
2068
2069 len = strlen (argv[*arg_ptr]);
2070 if (len == 0)
2071 error (1, 0, _("invalid null argument to -size"));
2072
2073 suffix = argv[*arg_ptr][len - 1];
2074 switch (suffix)
2075 {
2076 case 'b':
2077 blksize = 512;
2078 argv[*arg_ptr][len - 1] = '\0';
2079 break;
2080
2081 case 'c':
2082 blksize = 1;
2083 argv[*arg_ptr][len - 1] = '\0';
2084 break;
2085
2086 case 'k':
2087 blksize = 1024;
2088 argv[*arg_ptr][len - 1] = '\0';
2089 break;
2090
2091 case 'M': /* Megabytes */
2092 blksize = 1024*1024;
2093 argv[*arg_ptr][len - 1] = '\0';
2094 break;
2095
2096 case 'G': /* Gigabytes */
2097 blksize = 1024*1024*1024;
2098 argv[*arg_ptr][len - 1] = '\0';
2099 break;
2100
2101 case 'w':
2102 blksize = 2;
2103 argv[*arg_ptr][len - 1] = '\0';
2104 break;
2105
2106 case '0':
2107 case '1':
2108 case '2':
2109 case '3':
2110 case '4':
2111 case '5':
2112 case '6':
2113 case '7':
2114 case '8':
2115 case '9':
2116 break;
2117
2118 default:
2119 error (1, 0, _("invalid -size type `%c'"), argv[*arg_ptr][len - 1]);
2120 }
2121 /* TODO: accept fractional megabytes etc. ? */
2122 if (!get_num (argv[*arg_ptr], &num, &c_type))
2123 {
2124 error(1, 0,
2125 _("Invalid argument `%s%c' to -size"),
2126 argv[*arg_ptr], (int)suffix);
2127 return false;
2128 }
2129 our_pred = insert_primary (entry);
2130 our_pred->args.size.kind = c_type;
2131 our_pred->args.size.blocksize = blksize;
2132 our_pred->args.size.size = num;
2133 our_pred->need_stat = true;
2134 our_pred->need_type = false;
2135
2136 if (COMP_GT == c_type)
2137 our_pred->est_success_rate = (num*blksize > 20480) ? 0.1 : 0.9;
2138 else if (COMP_LT == c_type)
2139 our_pred->est_success_rate = (num*blksize > 20480) ? 0.9 : 0.1;
2140 else
2141 our_pred->est_success_rate = 0.01;
2142
2143 (*arg_ptr)++;
2144 return true;
2145 }
2146
2147
2148 static boolean
2149 parse_samefile (const struct parser_table* entry, char **argv, int *arg_ptr)
2150 {
2151 /* General idea: stat the file, remember device and inode numbers.
2152 * If a candidate file matches those, it's the same file.
2153 */
2154 struct predicate *our_pred;
2155 struct stat st, fst;
2156 int fd, openflags;
2157
2158 set_stat_placeholders(&st);
2159 if (!collect_arg_stat_info(argv, arg_ptr, &st))
2160 return false;
2161
2162 set_stat_placeholders(&fst);
2163 /* POSIX systems are free to re-use the inode number of a deleted
2164 * file. To ensure that we are not fooled by inode reuse, we hold
2165 * the file open if we can. This would prevent the system reusing
2166 * the file.
2167 */
2168 fd = -3; /* means, uninitialised */
2169 openflags = O_RDONLY;
2170
2171 if (options.symlink_handling == SYMLINK_NEVER_DEREF)
2172 {
2173 if (options.open_nofollow_available)
2174 {
2175 assert (O_NOFOLLOW != 0);
2176 openflags |= O_NOFOLLOW;
2177 fd = -1; /* safe to open it. */
2178 }
2179 else
2180 {
2181 if (S_ISLNK(st.st_mode))
2182 {
2183 /* no way to ensure that a symlink will not be followed
2184 * by open(2), so fall back on using lstat(). Accept
2185 * the risk that the named file will be deleted and
2186 * replaced with another having the same inode.
2187 *
2188 * Avoid opening the file.
2189 */
2190 fd = -2; /* Do not open it */
2191 }
2192 else
2193 {
2194 fd = -1;
2195 /* Race condition here: the file might become a symlink here. */
2196 }
2197 }
2198 }
2199 else
2200 {
2201 /* We want to dereference the symlink anyway */
2202 fd = -1; /* safe to open it without O_NOFOLLOW */
2203 }
2204
2205 assert (fd != -3); /* check we made a decision */
2206 if (fd == -1)
2207 {
2208 /* Race condition here. The file might become a
2209 * symbolic link in between out call to stat and
2210 * the call to open.
2211 */
2212 fd = open(argv[*arg_ptr], openflags);
2213
2214 if (fd >= 0)
2215 {
2216 /* We stat the file again here to prevent a race condition
2217 * between the first stat and the call to open(2).
2218 */
2219 if (0 != fstat(fd, &fst))
2220 {
2221 fatal_file_error(argv[*arg_ptr]);
2222 }
2223 else
2224 {
2225 /* Worry about the race condition. If the file became a
2226 * symlink after our first stat and before our call to
2227 * open, fst may contain the stat information for the
2228 * destination of the link, not the link itself.
2229 */
2230 if ((*options.xstat) (argv[*arg_ptr], &st))
2231 fatal_file_error(argv[*arg_ptr]);
2232
2233 if ((options.symlink_handling == SYMLINK_NEVER_DEREF)
2234 && (!options.open_nofollow_available))
2235 {
2236 if (S_ISLNK(st.st_mode))
2237 {
2238 /* We lost the race. Leave the data in st. The
2239 * file descriptor points to the wrong thing.
2240 */
2241 close(fd);
2242 fd = -1;
2243 }
2244 else
2245 {
2246 /* Several possibilities here:
2247 * 1. There was no race
2248 * 2. The file changed into a symlink after the stat and
2249 * before the open, and then back into a non-symlink
2250 * before the second stat.
2251 *
2252 * In case (1) there is no problem. In case (2),
2253 * the stat() and fstat() calls will have returned
2254 * different data. O_NOFOLLOW was not available,
2255 * so the open() call may have followed a symlink
2256 * even if the -P option is in effect.
2257 */
2258 if ((st.st_dev == fst.st_dev)
2259 && (st.st_ino == fst.st_ino))
2260 {
2261 /* No race. No need to copy fst to st,
2262 * since they should be identical (modulo
2263 * differences in padding bytes).
2264 */
2265 }
2266 else
2267 {
2268 /* We lost the race. Leave the data in st. The
2269 * file descriptor points to the wrong thing.
2270 */
2271 close(fd);
2272 fd = -1;
2273 }
2274 }
2275 }
2276 else
2277 {
2278 st = fst;
2279 }
2280 }
2281 }
2282 }
2283
2284 our_pred = insert_primary (entry);
2285 our_pred->args.samefileid.ino = st.st_ino;
2286 our_pred->args.samefileid.dev = st.st_dev;
2287 our_pred->args.samefileid.fd = fd;
2288 our_pred->need_type = false;
2289 our_pred->need_stat = true;
2290 our_pred->est_success_rate = 0.01f;
2291 return true;
2292 }
2293
2294 #if 0
2295 /* This function is commented out partly because support for it is
2296 * uneven.
2297 */
2298 static boolean
2299 parse_show_control_chars (const struct parser_table* entry,
2300 char **argv,
2301 int *arg_ptr)
2302 {
2303 const char *arg;
2304 const char *errmsg = _("The -show-control-chars option takes "
2305 "a single argument which "
2306 "must be 'literal' or 'safe'");
2307
2308 if ((argv == NULL) || (argv[*arg_ptr] == NULL))
2309 {
2310 error (1, errno, "%s", errmsg);
2311 return false;
2312 }
2313 else
2314 {
2315 arg = argv[*arg_ptr];
2316
2317 if (0 == strcmp("literal", arg))
2318 {
2319 options.literal_control_chars = true;
2320 }
2321 else if (0 == strcmp("safe", arg))
2322 {
2323 options.literal_control_chars = false;
2324 }
2325 else
2326 {
2327 error (1, errno, "%s", errmsg);
2328 return false;
2329 }
2330 (*arg_ptr)++; /* consume the argument. */
2331 return true;
2332 }
2333 }
2334 #endif
2335
2336
2337 static boolean
2338 parse_true (const struct parser_table* entry, char **argv, int *arg_ptr)
2339 {
2340 struct predicate *our_pred;
2341
2342 (void) argv;
2343 (void) arg_ptr;
2344
2345 our_pred = insert_primary (entry);
2346 our_pred->need_stat = our_pred->need_type = false;
2347 our_pred->est_success_rate = 1.0f;
2348 return true;
2349 }
2350
2351 static boolean
2352 parse_noop (const struct parser_table* entry, char **argv, int *arg_ptr)
2353 {
2354 (void) entry;
2355 return parse_true(get_noop(), argv, arg_ptr);
2356 }
2357
2358 static boolean
2359 parse_accesscheck (const struct parser_table* entry, char **argv, int *arg_ptr)
2360 {
2361 struct predicate *our_pred;
2362 (void) argv;
2363 (void) arg_ptr;
2364 our_pred = insert_primary (entry);
2365 our_pred->need_stat = our_pred->need_type = false;
2366 our_pred->side_effects = our_pred->no_default_print = false;
2367 if (pred_is(our_pred, pred_executable))
2368 our_pred->est_success_rate = 0.2;
2369 else
2370 our_pred->est_success_rate = 0.9;
2371 return true;
2372 }
2373
2374 static boolean
2375 parse_type (const struct parser_table* entry, char **argv, int *arg_ptr)
2376 {
2377 return insert_type (argv, arg_ptr, entry, pred_type);
2378 }
2379
2380 static boolean
2381 parse_uid (const struct parser_table* entry, char **argv, int *arg_ptr)
2382 {
2383 struct predicate *p = insert_num (argv, arg_ptr, entry);
2384 if (p)
2385 {
2386 p->est_success_rate = (p->args.numinfo.l_val < 100) ? 0.99 : 0.2;
2387 return true;
2388 }
2389 else
2390 {
2391 return false;
2392 }
2393 }
2394
2395 static boolean
2396 parse_used (const struct parser_table* entry, char **argv, int *arg_ptr)
2397 {
2398 struct predicate *our_pred;
2399 struct time_val tval;
2400 const char *offset_str;
2401 const char *errmsg = "arithmetic overflow while converting %s days to a number of seconds";
2402
2403 if (collect_arg(argv, arg_ptr, &offset_str))
2404 {
2405 /* The timespec is actually a delta value, so we use an origin of 0. */
2406 struct timespec zero = {0,0};
2407 if (get_relative_timestamp(offset_str, &tval, zero, DAYSECS, errmsg))
2408 {
2409 our_pred = insert_primary (entry);
2410 our_pred->args.reftime = tval;
2411 our_pred->est_success_rate = estimate_file_age_success_rate(tval.ts.tv_sec / DAYSECS);
2412 return true;
2413 }
2414 else
2415 {
2416 error(1, 0, _("Invalid argument %s to -used"), offset_str);
2417 return false;
2418 }
2419 }
2420 else
2421 {
2422 return false; /* missing argument */
2423 }
2424 }
2425
2426 static boolean
2427 parse_user (const struct parser_table* entry, char **argv, int *arg_ptr)
2428 {
2429 const char *username;
2430
2431 if (collect_arg(argv, arg_ptr, &username))
2432 {
2433 struct predicate *our_pred;
2434 uid_t uid;
2435 struct passwd *cur_pwd = getpwnam(username);
2436 endpwent();
2437 if (cur_pwd != NULL)
2438 {
2439 uid = cur_pwd->pw_uid;
2440 }
2441 else
2442 {
2443 int uid_len = strspn (username, "0123456789");
2444 if (uid_len && (username[uid_len]==0))
2445 uid = safe_atoi (username);
2446 else
2447 return false;
2448 }
2449 our_pred = insert_primary (entry);
2450 our_pred->args.uid = uid;
2451 our_pred->est_success_rate = (our_pred->args.uid < 100) ? 0.99 : 0.2;
2452 return true;
2453 }
2454 return false;
2455 }
2456
2457 static boolean
2458 parse_version (const struct parser_table* entry, char **argv, int *arg_ptr)
2459 {
2460 int features = 0;
2461 int flags;
2462
2463 (void) argv;
2464 (void) arg_ptr;
2465 (void) entry;
2466
2467 display_findutils_version("find");
2468 printf (_("Features enabled: "));
2469
2470 #if CACHE_IDS
2471 printf("CACHE_IDS ");
2472 ++features;
2473 #endif
2474 #if DEBUG
2475 printf("DEBUG ");
2476 ++features;
2477 #endif
2478 #if DEBUG_STAT
2479 printf("DEBUG_STAT ");
2480 ++features;
2481 #endif
2482 #if defined USE_STRUCT_DIRENT_D_TYPE && defined HAVE_STRUCT_DIRENT_D_TYPE
2483 printf("D_TYPE ");
2484 ++features;
2485 #endif
2486 #if defined O_NOFOLLOW
2487 printf("O_NOFOLLOW(%s) ",
2488 (options.open_nofollow_available ? "enabled" : "disabled"));
2489 ++features;
2490 #endif
2491 #if defined LEAF_OPTIMISATION
2492 printf("LEAF_OPTIMISATION ");
2493 ++features;
2494 #endif
2495
2496 flags = 0;
2497 if (is_fts_enabled(&flags))
2498 {
2499 int nflags = 0;
2500 printf("FTS(");
2501 ++features;
2502
2503 if (flags & FTS_CWDFD)
2504 {
2505 if (nflags)
2506 {
2507 printf(",");
2508 }
2509 printf("FTS_CWDFD");
2510 ++nflags;
2511 }
2512 printf(") ");
2513 }
2514
2515 printf("CBO(level=%d) ", (int)(options.optimisation_level));
2516 ++features;
2517
2518 if (0 == features)
2519 {
2520 /* For the moment, leave this as English in case someone wants
2521 to parse these strings. */
2522 printf("none");
2523 }
2524 printf("\n");
2525
2526 exit (0);
2527 }
2528
2529 static boolean
2530 parse_xdev (const struct parser_table* entry, char **argv, int *arg_ptr)
2531 {
2532 options.stay_on_filesystem = true;
2533 return parse_noop(entry, argv, arg_ptr);
2534 }
2535
2536 static boolean
2537 parse_ignore_race (const struct parser_table* entry, char **argv, int *arg_ptr)
2538 {
2539 options.ignore_readdir_race = true;
2540 return parse_noop(entry, argv, arg_ptr);
2541 }
2542
2543 static boolean
2544 parse_noignore_race (const struct parser_table* entry, char **argv, int *arg_ptr)
2545 {
2546 options.ignore_readdir_race = false;
2547 return parse_noop(entry, argv, arg_ptr);
2548 }
2549
2550 static boolean
2551 parse_warn (const struct parser_table* entry, char **argv, int *arg_ptr)
2552 {
2553 options.warnings = true;
2554 return parse_noop(entry, argv, arg_ptr);
2555 }
2556
2557 static boolean
2558 parse_xtype (const struct parser_table* entry, char **argv, int *arg_ptr)
2559 {
2560 return insert_type (argv, arg_ptr, entry, pred_xtype);
2561 }
2562 \f
2563 static boolean
2564 insert_type (char **argv, int *arg_ptr,
2565 const struct parser_table *entry,
2566 PRED_FUNC which_pred)
2567 {
2568 mode_t type_cell;
2569 struct predicate *our_pred;
2570 float rate = 0.5;
2571 const char *typeletter;
2572
2573 if (collect_arg(argv, arg_ptr, &typeletter))
2574 {
2575 if (strlen(typeletter) != 1u)
2576 {
2577 error(1, 0, _("Arguments to -type should contain only one letter"));
2578 return false;
2579 }
2580
2581 switch (typeletter[0])
2582 {
2583 case 'b': /* block special */
2584 type_cell = S_IFBLK;
2585 rate = 0.01f;
2586 break;
2587 case 'c': /* character special */
2588 type_cell = S_IFCHR;
2589 rate = 0.01f;
2590 break;
2591 case 'd': /* directory */
2592 type_cell = S_IFDIR;
2593 rate = 0.4f;
2594 break;
2595 case 'f': /* regular file */
2596 type_cell = S_IFREG;
2597 rate = 0.95f;
2598 break;
2599 #ifdef S_IFLNK
2600 case 'l': /* symbolic link */
2601 type_cell = S_IFLNK;
2602 rate = 0.1f;
2603 break;
2604 #endif
2605 #ifdef S_IFIFO
2606 case 'p': /* pipe */
2607 type_cell = S_IFIFO;
2608 rate = 0.01f;
2609 break;
2610 #endif
2611 #ifdef S_IFSOCK
2612 case 's': /* socket */
2613 type_cell = S_IFSOCK;
2614 rate = 0.01f;
2615 break;
2616 #endif
2617 #ifdef S_IFDOOR
2618 case 'D': /* Solaris door */
2619 type_cell = S_IFDOOR;
2620 rate = 0.01f;
2621 break;
2622 #endif
2623 default: /* None of the above ... nuke 'em. */
2624 error(1, 0, _("Unknown argument to -type: %c"), (*typeletter));
2625 return false;
2626 }
2627 our_pred = insert_primary_withpred (entry, which_pred);
2628 our_pred->est_success_rate = rate;
2629
2630 /* Figure out if we will need to stat the file, because if we don't
2631 * need to follow symlinks, we can avoid a stat call by using
2632 * struct dirent.d_type.
2633 */
2634 if (which_pred == pred_xtype)
2635 {
2636 our_pred->need_stat = true;
2637 our_pred->need_type = false;
2638 }
2639 else
2640 {
2641 our_pred->need_stat = false; /* struct dirent is enough */
2642 our_pred->need_type = true;
2643 }
2644 our_pred->args.type = type_cell;
2645 return true;
2646 }
2647 return false;
2648 }
2649
2650
2651 /* Return true if the file accessed via FP is a terminal.
2652 */
2653 static boolean
2654 stream_is_tty(FILE *fp)
2655 {
2656 int fd = fileno(fp);
2657 if (-1 == fd)
2658 {
2659 return false; /* not a valid stream */
2660 }
2661 else
2662 {
2663 return isatty(fd) ? true : false;
2664 }
2665
2666 }
2667
2668
2669
2670
2671 /* XXX: do we need to pass FUNC to this function? */
2672 static boolean
2673 insert_fprintf (struct format_val *vec,
2674 const struct parser_table *entry, PRED_FUNC func,
2675 const char *format_const)
2676 {
2677 char *format = (char*)format_const; /* XXX: casting away constness */
2678 register char *scan; /* Current address in scanning `format'. */
2679 register char *scan2; /* Address inside of element being scanned. */
2680 struct segment **segmentp; /* Address of current segment. */
2681 struct predicate *our_pred;
2682
2683 our_pred = insert_primary_withpred (entry, func);
2684 our_pred->side_effects = our_pred->no_default_print = true;
2685 our_pred->args.printf_vec = *vec;
2686 our_pred->need_type = false;
2687 our_pred->need_stat = false;
2688 our_pred->p_cost = NeedsNothing;
2689
2690 segmentp = &our_pred->args.printf_vec.segment;
2691 *segmentp = NULL;
2692
2693 for (scan = format; *scan; scan++)
2694 {
2695 if (*scan == '\\')
2696 {
2697 scan2 = scan + 1;
2698 if (*scan2 >= '0' && *scan2 <= '7')
2699 {
2700 register int n, i;
2701
2702 for (i = n = 0; i < 3 && (*scan2 >= '0' && *scan2 <= '7');
2703 i++, scan2++)
2704 n = 8 * n + *scan2 - '0';
2705 scan2--;
2706 *scan = n;
2707 }
2708 else
2709 {
2710 switch (*scan2)
2711 {
2712 case 'a':
2713 *scan = 7;
2714 break;
2715 case 'b':
2716 *scan = '\b';
2717 break;
2718 case 'c':
2719 make_segment (segmentp, format, scan - format,
2720 KIND_STOP, 0, 0,
2721 our_pred);
2722 if (our_pred->need_stat && (our_pred->p_cost < NeedsStatInfo))
2723 our_pred->p_cost = NeedsStatInfo;
2724 return true;
2725 case 'f':
2726 *scan = '\f';
2727 break;
2728 case 'n':
2729 *scan = '\n';
2730 break;
2731 case 'r':
2732 *scan = '\r';
2733 break;
2734 case 't':
2735 *scan = '\t';
2736 break;
2737 case 'v':
2738 *scan = '\v';
2739 break;
2740 case '\\':
2741 /* *scan = '\\'; * it already is */
2742 break;
2743 default:
2744 error (0, 0,
2745 _("warning: unrecognized escape `\\%c'"), *scan2);
2746 scan++;
2747 continue;
2748 }
2749 }
2750 segmentp = make_segment (segmentp, format, scan - format + 1,
2751 KIND_PLAIN, 0, 0,
2752 our_pred);
2753 format = scan2 + 1; /* Move past the escape. */
2754 scan = scan2; /* Incremented immediately by `for'. */
2755 }
2756 else if (*scan == '%')
2757 {
2758 if (scan[1] == 0)
2759 {
2760 /* Trailing %. We don't like those. */
2761 error (1, 0, _("error: %s at end of format string"), scan);
2762 }
2763 else if (scan[1] == '%')
2764 {
2765 segmentp = make_segment (segmentp, format, scan - format + 1,
2766 KIND_PLAIN, 0, 0,
2767 our_pred);
2768 scan++;
2769 format = scan + 1;
2770 continue;
2771 }
2772 /* Scan past flags, width and precision, to verify kind. */
2773 for (scan2 = scan; *++scan2 && strchr ("-+ #", *scan2);)
2774 /* Do nothing. */ ;
2775 while (ISDIGIT (*scan2))
2776 scan2++;
2777 if (*scan2 == '.')
2778 for (scan2++; ISDIGIT (*scan2); scan2++)
2779 /* Do nothing. */ ;
2780 if (strchr ("abcdDfFgGhHiklmMnpPsStuUyY", *scan2))
2781 {
2782 segmentp = make_segment (segmentp, format, scan2 - format,
2783 KIND_FORMAT, *scan2, 0,
2784 our_pred);
2785 scan = scan2;
2786 format = scan + 1;
2787 }
2788 else if (strchr ("ABCT", *scan2) && scan2[1])
2789 {
2790 segmentp = make_segment (segmentp, format, scan2 - format,
2791 KIND_FORMAT, scan2[0], scan2[1],
2792 our_pred);
2793 scan = scan2 + 1;
2794 format = scan + 1;
2795 continue;
2796 }
2797 else
2798 {
2799 /* An unrecognized % escape. Print the char after the %. */
2800 error (0, 0, _("warning: unrecognized format directive `%%%c'"),
2801 *scan2);
2802 segmentp = make_segment (segmentp, format, scan - format,
2803 KIND_PLAIN, 0, 0,
2804 our_pred);
2805 format = scan + 1;
2806 continue;
2807 }
2808 }
2809 }
2810
2811 if (scan > format)
2812 make_segment (segmentp, format, scan - format, KIND_PLAIN, 0, 0,
2813 our_pred);
2814 return true;
2815 }
2816
2817 /* Create a new fprintf segment in *SEGMENT, with type KIND,
2818 from the text in FORMAT, which has length LEN.
2819 Return the address of the `next' pointer of the new segment. */
2820
2821 static struct segment **
2822 make_segment (struct segment **segment,
2823 char *format,
2824 int len,
2825 int kind,
2826 char format_char,
2827 char aux_format_char,
2828 struct predicate *pred)
2829 {
2830 enum EvaluationCost mycost = NeedsNothing;
2831 char *fmt;
2832
2833 *segment = xmalloc (sizeof (struct segment));
2834
2835 (*segment)->segkind = kind;
2836 (*segment)->format_char[0] = format_char;
2837 (*segment)->format_char[1] = aux_format_char;
2838 (*segment)->next = NULL;
2839 (*segment)->text_len = len;
2840
2841 fmt = (*segment)->text = xmalloc (len + sizeof "d");
2842 strncpy (fmt, format, len);
2843 fmt += len;
2844
2845 switch (kind)
2846 {
2847 case KIND_PLAIN: /* Plain text string, no % conversion. */
2848 case KIND_STOP: /* Terminate argument, no newline. */
2849 assert (0 == format_char);
2850 assert (0 == aux_format_char);
2851 *fmt = '\0';
2852 if (mycost > pred->p_cost)
2853 pred->p_cost = NeedsNothing;
2854 return &(*segment)->next;
2855 break;
2856 }
2857
2858 assert (kind == KIND_FORMAT);
2859 switch (format_char)
2860 {
2861 case 'l': /* object of symlink */
2862 pred->need_stat = true;
2863 mycost = NeedsLinkName;
2864 *fmt++ = 's';
2865 break;
2866
2867 case 'y': /* file type */
2868 pred->need_type = true;
2869 mycost = NeedsType;
2870 *fmt++ = 's';
2871 break;
2872
2873 case 'a': /* atime in `ctime' format */
2874 case 'A': /* atime in user-specified strftime format */
2875 case 'B': /* birth time in user-specified strftime format */
2876 case 'c': /* ctime in `ctime' format */
2877 case 'C': /* ctime in user-specified strftime format */
2878 case 'F': /* file system type */
2879 case 'g': /* group name */
2880 case 'i': /* inode number */
2881 case 'M': /* mode in `ls -l' format (eg., "drwxr-xr-x") */
2882 case 's': /* size in bytes */
2883 case 't': /* mtime in `ctime' format */
2884 case 'T': /* mtime in user-specified strftime format */
2885 case 'u': /* user name */
2886 pred->need_stat = true;
2887 mycost = NeedsStatInfo;
2888 *fmt++ = 's';
2889 break;
2890
2891 case 'S': /* sparseness */
2892 pred->need_stat = true;
2893 mycost = NeedsStatInfo;
2894 *fmt++ = 'g';
2895 break;
2896
2897 case 'Y': /* symlink pointed file type */
2898 pred->need_stat = true;
2899 mycost = NeedsType; /* true for amortised effect */
2900 *fmt++ = 's';
2901 break;
2902
2903 case 'f': /* basename of path */
2904 case 'h': /* leading directories part of path */
2905 case 'p': /* pathname */
2906 case 'P': /* pathname with ARGV element stripped */
2907 *fmt++ = 's';
2908 break;
2909
2910 case 'H': /* ARGV element file was found under */
2911 *fmt++ = 's';
2912 break;
2913
2914 /* Numeric items that one might expect to honour
2915 * #, 0, + flags but which do not.
2916 */
2917 case 'G': /* GID number */
2918 case 'U': /* UID number */
2919 case 'b': /* size in 512-byte blocks (NOT birthtime in ctime fmt)*/
2920 case 'D': /* Filesystem device on which the file exits */
2921 case 'k': /* size in 1K blocks */
2922 case 'n': /* number of links */
2923 pred->need_stat = true;
2924 mycost = NeedsStatInfo;
2925 *fmt++ = 's';
2926 break;
2927
2928 /* Numeric items that DO honour #, 0, + flags.
2929 */
2930 case 'd': /* depth in search tree (0 = ARGV element) */
2931 *fmt++ = 'd';
2932 break;
2933
2934 case 'm': /* mode as octal number (perms only) */
2935 *fmt++ = 'o';
2936 pred->need_stat = true;
2937 mycost = NeedsStatInfo;
2938 break;
2939
2940 case '{':
2941 case '[':
2942 case '(':
2943 error (1, 0,
2944 _("error: the format directive `%%%c' is reserved for future use"),
2945 (int)kind);
2946 /*NOTREACHED*/
2947 break;
2948 }
2949 *fmt = '\0';
2950
2951 if (mycost > pred->p_cost)
2952 pred->p_cost = mycost;
2953 return &(*segment)->next;
2954 }
2955 \f
2956 static void
2957 check_path_safety(const char *action, char **argv)
2958 {
2959 char *s;
2960 const char *path = getenv("PATH");
2961 if (NULL == path)
2962 {
2963 /* $PATH is not set. Assume the OS default is safe.
2964 * That may not be true on Windows, but I'm not aware
2965 * of a way to get Windows to avoid searching the
2966 * current directory anyway.
2967 */
2968 return;
2969 }
2970
2971 (void)argv;
2972
2973 s = next_element(path, 1);
2974 while ((s = next_element ((char *) NULL, 1)) != NULL)
2975 {
2976 if (0 == strcmp(s, "."))
2977 {
2978 error(1, 0, _("The current directory is included in the PATH "
2979 "environment variable, which is insecure in "
2980 "combination with the %s action of find. "
2981 "Please remove the current directory from your "
2982 "$PATH (that is, remove \".\" or leading or trailing "
2983 "colons)"),
2984 action);
2985 }
2986 else if ('/' != s[0])
2987 {
2988 /* Relative paths are also dangerous in $PATH. */
2989 error(1, 0, _("The relative path %s is included in the PATH "
2990 "environment variable, which is insecure in "
2991 "combination with the %s action of find. "
2992 "Please remove that entry from $PATH"),
2993 safely_quote_err_filename(0, s),
2994 action);
2995 }
2996 }
2997 }
2998
2999 \f
3000 /* handles both exec and ok predicate */
3001 static boolean
3002 new_insert_exec_ok (const char *action,
3003 const struct parser_table *entry,
3004 int dir_fd,
3005 char **argv,
3006 int *arg_ptr)
3007 {
3008 int start, end; /* Indexes in ARGV of start & end of cmd. */
3009 int i; /* Index into cmd args */
3010 int saw_braces; /* True if previous arg was '{}'. */
3011 boolean allow_plus; /* True if + is a valid terminator */
3012 int brace_count; /* Number of instances of {}. */
3013 PRED_FUNC func = entry->pred_func;
3014 enum BC_INIT_STATUS bcstatus;
3015
3016 struct predicate *our_pred;
3017 struct exec_val *execp; /* Pointer for efficiency. */
3018
3019 if ((argv == NULL) || (argv[*arg_ptr] == NULL))
3020 return false;
3021
3022 our_pred = insert_primary_withpred (entry, func);
3023 our_pred->side_effects = our_pred->no_default_print = true;
3024 our_pred->need_type = our_pred->need_stat = false;
3025
3026 execp = &our_pred->args.exec_vec;
3027
3028 if ((func != pred_okdir) && (func != pred_ok))
3029 {
3030 allow_plus = true;
3031 execp->close_stdin = false;
3032 }
3033 else
3034 {
3035 allow_plus = false;
3036 /* If find reads stdin (i.e. for -ok and similar), close stdin
3037 * in the child to prevent some script from consiming the output
3038 * intended for find.
3039 */
3040 execp->close_stdin = true;
3041 }
3042
3043
3044 if ((func == pred_execdir) || (func == pred_okdir))
3045 {
3046 options.ignore_readdir_race = false;
3047 check_path_safety(action, argv);
3048 execp->use_current_dir = true;
3049 }
3050 else
3051 {
3052 execp->use_current_dir = false;
3053 }
3054
3055 our_pred->args.exec_vec.multiple = 0;
3056
3057 /* Count the number of args with path replacements, up until the ';'.
3058 * Also figure out if the command is terminated by ";" or by "+".
3059 */
3060 start = *arg_ptr;
3061 for (end = start, saw_braces=0, brace_count=0;
3062 (argv[end] != NULL)
3063 && ((argv[end][0] != ';') || (argv[end][1] != '\0'));
3064 end++)
3065 {
3066 /* For -exec and -execdir, "{} +" can terminate the command. */
3067 if ( allow_plus
3068 && argv[end][0] == '+' && argv[end][1] == 0
3069 && saw_braces)
3070 {
3071 our_pred->args.exec_vec.multiple = 1;
3072 break;
3073 }
3074
3075 saw_braces = 0;
3076 if (mbsstr (argv[end], "{}"))
3077 {
3078 saw_braces = 1;
3079 ++brace_count;
3080
3081 if (0 == end && (func == pred_execdir || func == pred_okdir))
3082 {
3083 /* The POSIX standard says that {} replacement should
3084 * occur even in the utility name. This is insecure
3085 * since it means we will be executing a command whose
3086 * name is chosen according to whatever find finds in
3087 * the file system. That can be influenced by an
3088 * attacker. Hence for -execdir and -okdir this is not
3089 * allowed. We can specify this as those options are
3090 * not defined by POSIX.
3091 */
3092 error(1, 0, _("You may not use {} within the utility name for "
3093 "-execdir and -okdir, because this is a potential "
3094 "security problem."));
3095 }
3096 }
3097 }
3098
3099 /* Fail if no command given or no semicolon found. */
3100 if ((end == start) || (argv[end] == NULL))
3101 {
3102 *arg_ptr = end;
3103 free(our_pred);
3104 return false;
3105 }
3106
3107 if (our_pred->args.exec_vec.multiple && brace_count > 1)
3108 {
3109
3110 const char *suffix;
3111 if (func == pred_execdir)
3112 suffix = "dir";
3113 else
3114 suffix = "";
3115
3116 error(1, 0,
3117 _("Only one instance of {} is supported with -exec%s ... +"),
3118 suffix);
3119 }
3120
3121 /* We use a switch statement here so that the compiler warns us when
3122 * we forget to handle a newly invented enum value.
3123 *
3124 * Like xargs, we allow 2KiB of headroom for the launched utility to
3125 * export its own environment variables before calling something
3126 * else.
3127 */
3128 bcstatus = bc_init_controlinfo(&execp->ctl, 2048u);
3129 switch (bcstatus)
3130 {
3131 case BC_INIT_ENV_TOO_BIG:
3132 case BC_INIT_CANNOT_ACCOMODATE_HEADROOM:
3133 error(1, 0,
3134 _("The environment is too large for exec()."));
3135 break;
3136 case BC_INIT_OK:
3137 /* Good news. Carry on. */
3138 break;
3139 }
3140 bc_use_sensible_arg_max(&execp->ctl);
3141
3142
3143 execp->ctl.exec_callback = launch;
3144
3145 if (our_pred->args.exec_vec.multiple)
3146 {
3147 /* "+" terminator, so we can just append our arguments after the
3148 * command and initial arguments.
3149 */
3150 execp->replace_vec = NULL;
3151 execp->ctl.replace_pat = NULL;
3152 execp->ctl.rplen = 0;
3153 execp->ctl.lines_per_exec = 0; /* no limit */
3154 execp->ctl.args_per_exec = 0; /* no limit */
3155
3156 /* remember how many arguments there are */
3157 execp->ctl.initial_argc = (end-start) - 1;
3158
3159 /* execp->state = xmalloc(sizeof struct buildcmd_state); */
3160 bc_init_state(&execp->ctl, &execp->state, execp);
3161
3162 /* Gather the initial arguments. Skip the {}. */
3163 for (i=start; i<end-1; ++i)
3164 {
3165 bc_push_arg(&execp->ctl, &execp->state,
3166 argv[i], strlen(argv[i])+1,
3167 NULL, 0,
3168 1);
3169 }
3170 }
3171 else
3172 {
3173 /* Semicolon terminator - more than one {} is supported, so we
3174 * have to do brace-replacement.
3175 */
3176 execp->num_args = end - start;
3177
3178 execp->ctl.replace_pat = "{}";
3179 execp->ctl.rplen = strlen(execp->ctl.replace_pat);
3180 execp->ctl.lines_per_exec = 0; /* no limit */
3181 execp->ctl.args_per_exec = 0; /* no limit */
3182 execp->replace_vec = xmalloc(sizeof(char*)*execp->num_args);
3183
3184
3185 /* execp->state = xmalloc(sizeof(*(execp->state))); */
3186 bc_init_state(&execp->ctl, &execp->state, execp);
3187
3188 /* Remember the (pre-replacement) arguments for later. */
3189 for (i=0; i<execp->num_args; ++i)
3190 {
3191 execp->replace_vec[i] = argv[i+start];
3192 }
3193 }
3194
3195 if (argv[end] == NULL)
3196 *arg_ptr = end;
3197 else
3198 *arg_ptr = end + 1;
3199
3200 return true;
3201 }
3202
3203
3204
3205 static boolean
3206 insert_exec_ok (const char *action,
3207 const struct parser_table *entry,
3208 int dir_fd,
3209 char **argv,
3210 int *arg_ptr)
3211 {
3212 return new_insert_exec_ok(action, entry, dir_fd, argv, arg_ptr);
3213 }
3214
3215
3216 \f
3217 /* Get a timestamp and comparison type.
3218
3219 STR is the ASCII representation.
3220 Set *NUM_DAYS to the number of days/minutes/whatever, taken as being
3221 relative to ORIGIN (usually the current moment or midnight).
3222 Thus the sense of the comparison type appears to be reversed.
3223 Set *COMP_TYPE to the kind of comparison that is requested.
3224 Issue OVERFLOWMESSAGE if overflow occurs.
3225 Return true if all okay, false if input error.
3226
3227 Used by -atime, -ctime and -mtime (parsers) to
3228 get the appropriate information for a time predicate processor. */
3229
3230 static boolean
3231 get_relative_timestamp (const char *str,
3232 struct time_val *result,
3233 struct timespec origin,
3234 double sec_per_unit,
3235 const char *overflowmessage)
3236 {
3237 uintmax_t checkval;
3238 double offset, seconds, nanosec;
3239
3240 if (get_comp_type(&str, &result->kind))
3241 {
3242 /* Invert the sense of the comparison */
3243 switch (result->kind)
3244 {
3245 case COMP_LT: result->kind = COMP_GT; break;
3246 case COMP_GT: result->kind = COMP_LT; break;
3247 default: break;
3248 }
3249
3250 /* Convert the ASCII number into floating-point. */
3251 if (xstrtod(str, NULL, &offset, strtod))
3252 {
3253 /* Separate the floating point number the user specified
3254 * (which is a number of days, or minutes, etc) into an
3255 * integral number of seconds (SECONDS) and a fraction (NANOSEC).
3256 */
3257 nanosec = modf(offset * sec_per_unit, &seconds);
3258 nanosec *= 1.0e9; /* convert from fractional seconds to ns. */
3259
3260 result->ts.tv_sec = origin.tv_sec - seconds;
3261 result->ts.tv_nsec = origin.tv_nsec - nanosec;
3262 checkval = (uintmax_t)origin.tv_sec - seconds;
3263
3264 if (origin.tv_nsec < nanosec)
3265 {
3266 /* Perform a carry operation */
3267 result->ts.tv_nsec += 1000000000;
3268 result->ts.tv_sec -= 1;
3269 checkval -= 1;
3270 }
3271 /* Check for overflow. */
3272 if (checkval != result->ts.tv_sec)
3273 {
3274 /* an overflow has occurred. */
3275 error (1, 0, overflowmessage, str);
3276 }
3277 return true;
3278 }
3279 else
3280 {
3281 /* Conversion from ASCII to double failed. */
3282 return false;
3283 }
3284 }
3285 else
3286 {
3287 return false;
3288 }
3289 }
3290 \f
3291 /* Insert a time predicate based on the information in ENTRY.
3292 ARGV is a pointer to the argument array.
3293 ARG_PTR is a pointer to an index into the array, incremented if
3294 all went well.
3295
3296 Return true if input is valid, false if not.
3297
3298 A new predicate node is assigned, along with an argument node
3299 obtained with malloc.
3300
3301 Used by -atime, -ctime, and -mtime parsers. */
3302
3303 static boolean
3304 parse_time (const struct parser_table* entry, char *argv[], int *arg_ptr)
3305 {
3306 struct predicate *our_pred;
3307 struct time_val tval;
3308 enum comparison_type comp;
3309 const char *timearg, *orig_timearg;
3310 const char *errmsg = "arithmetic overflow while converting %s "
3311 "days to a number of seconds";
3312 struct timespec origin;
3313
3314 if (!collect_arg(argv, arg_ptr, &timearg))
3315 return false;
3316 orig_timearg = timearg;
3317
3318 /* Decide the origin by previewing the comparison type. */
3319 origin = options.cur_day_start;
3320
3321 if (get_comp_type(&timearg, &comp))
3322 {
3323 /* Remember, we invert the sense of the comparison, so this tests
3324 * against COMP_LT instead of COMP_GT...
3325 */
3326 if (COMP_LT == comp)
3327 {
3328 uintmax_t expected = origin.tv_sec + (DAYSECS-1);
3329 origin.tv_sec += (DAYSECS-1);
3330 if (origin.tv_sec != expected)
3331 {
3332 error(1, 0,
3333 _("arithmetic overflow when trying to calculate the end of today"));
3334 }
3335 }
3336 }
3337 /* We discard the value of comp here, as get_relative_timestamp
3338 * will set tval.kind. For that to work, we have to restore
3339 * timearg so that it points to the +/- prefix, if any. get_comp_type()
3340 * will have advanced timearg, so we restore it.
3341 */
3342 timearg = orig_timearg;
3343
3344 if (!get_relative_timestamp(timearg, &tval, origin, DAYSECS, errmsg))
3345 return false;
3346
3347 our_pred = insert_primary (entry);
3348 our_pred->args.reftime = tval;
3349 our_pred->est_success_rate = estimate_timestamp_success_rate(tval.ts.tv_sec);
3350
3351 if (options.debug_options & DebugExpressionTree)
3352 {
3353 time_t t;
3354
3355 fprintf (stderr, "inserting %s\n", our_pred->p_name);
3356 fprintf (stderr, " type: %s %s ",
3357 (tval.kind == COMP_GT) ? "gt" :
3358 ((tval.kind == COMP_LT) ? "lt" : ((tval.kind == COMP_EQ) ? "eq" : "?")),
3359 (tval.kind == COMP_GT) ? " >" :
3360 ((tval.kind == COMP_LT) ? " <" : ((tval.kind == COMP_EQ) ? ">=" : " ?")));
3361 t = our_pred->args.reftime.ts.tv_sec;
3362 fprintf (stderr, "%ju %s",
3363 (uintmax_t) our_pred->args.reftime.ts.tv_sec,
3364 ctime (&t));
3365 if (tval.kind == COMP_EQ)
3366 {
3367 t = our_pred->args.reftime.ts.tv_sec + DAYSECS;
3368 fprintf (stderr, " < %ju %s",
3369 (uintmax_t) t, ctime (&t));
3370 }
3371 }
3372
3373 return true;
3374 }
3375 \f
3376 /* Get the comparison type prefix (if any) from a number argument.
3377 The prefix is at *STR.
3378 Set *COMP_TYPE to the kind of comparison that is requested.
3379 Advance *STR beyond any initial comparison prefix.
3380
3381 Return true if all okay, false if input error. */
3382 static boolean
3383 get_comp_type(const char **str, enum comparison_type *comp_type)
3384 {
3385 switch (**str)
3386 {
3387 case '+':
3388 *comp_type = COMP_GT;
3389 (*str)++;
3390 break;
3391 case '-':
3392 *comp_type = COMP_LT;
3393 (*str)++;
3394 break;
3395 default:
3396 *comp_type = COMP_EQ;
3397 break;
3398 }
3399 return true;
3400 }
3401
3402
3403
3404
3405 \f
3406 /* Get a number with comparison information.
3407 The sense of the comparison information is 'normal'; that is,
3408 '+' looks for a count > than the number and '-' less than.
3409
3410 STR is the ASCII representation of the number.
3411 Set *NUM to the number.
3412 Set *COMP_TYPE to the kind of comparison that is requested.
3413
3414 Return true if all okay, false if input error. */
3415
3416 static boolean
3417 get_num (const char *str,
3418 uintmax_t *num,
3419 enum comparison_type *comp_type)
3420 {
3421 char *pend;
3422
3423 if (str == NULL)
3424 return false;
3425
3426 /* Figure out the comparison type if the caller accepts one. */
3427 if (comp_type)
3428 {
3429 if (!get_comp_type(&str, comp_type))
3430 return false;
3431 }
3432
3433 return xstrtoumax (str, &pend, 10, num, "") == LONGINT_OK;
3434 }
3435 \f
3436 /* Insert a number predicate.
3437 ARGV is a pointer to the argument array.
3438 *ARG_PTR is an index into ARGV, incremented if all went well.
3439 *PRED is the predicate processor to insert.
3440
3441 Return true if input is valid, false if error.
3442
3443 A new predicate node is assigned, along with an argument node
3444 obtained with malloc.
3445
3446 Used by -inum and -links parsers. */
3447
3448 static struct predicate *
3449 insert_num (char **argv, int *arg_ptr, const struct parser_table *entry)
3450 {
3451 const char *numstr;
3452
3453 if (collect_arg(argv, arg_ptr, &numstr))
3454 {
3455 uintmax_t num;
3456 enum comparison_type c_type;
3457
3458 if (get_num (numstr, &num, &c_type))
3459 {
3460 struct predicate *our_pred = insert_primary (entry);
3461 our_pred->args.numinfo.kind = c_type;
3462 our_pred->args.numinfo.l_val = num;
3463
3464 if (options.debug_options & DebugExpressionTree)
3465 {
3466 fprintf (stderr, "inserting %s\n", our_pred->p_name);
3467 fprintf (stderr, " type: %s %s ",
3468 (c_type == COMP_GT) ? "gt" :
3469 ((c_type == COMP_LT) ? "lt" : ((c_type == COMP_EQ) ? "eq" : "?")),
3470 (c_type == COMP_GT) ? " >" :
3471 ((c_type == COMP_LT) ? " <" : ((c_type == COMP_EQ) ? " =" : " ?")));
3472 fprintf (stderr, "%ju\n", our_pred->args.numinfo.l_val);
3473 }
3474 return our_pred;
3475 }
3476 }
3477 return NULL;
3478 }
3479
3480 static void
3481 open_output_file (const char *path, struct format_val *p)
3482 {
3483 p->segment = NULL;
3484 p->quote_opts = clone_quoting_options (NULL);
3485
3486 if (!strcmp (path, "/dev/stderr"))
3487 {
3488 p->stream = stderr;
3489 p->filename = _("standard error");
3490 }
3491 else if (!strcmp (path, "/dev/stdout"))
3492 {
3493 p->stream = stdout;
3494 p->filename = _("standard output");
3495 }
3496 else
3497 {
3498 p->stream = fopen_safer (path, "w");
3499 p->filename = path;
3500
3501 if (p->stream == NULL)
3502 {
3503 fatal_file_error(path);
3504 }
3505 }
3506
3507 p->dest_is_tty = stream_is_tty(p->stream);
3508 }
3509
3510 static void
3511 open_stdout (struct format_val *p)
3512 {
3513 open_output_file("/dev/stdout", p);
3514 }
GNU findutils