dir.c: rename free_excludes() to clear_exclude_list()
[git/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * See Documentation/technical/api-directory-listing.txt
6 *
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10 #include "cache.h"
11 #include "dir.h"
12 #include "refs.h"
13
14 struct path_simplify {
15 int len;
16 const char *path;
17 };
18
19 static int read_directory_recursive(struct dir_struct *dir, const char *path, int len,
20 int check_only, const struct path_simplify *simplify);
21 static int get_dtype(struct dirent *de, const char *path, int len);
22
23 /* helper string functions with support for the ignore_case flag */
24 int strcmp_icase(const char *a, const char *b)
25 {
26 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
27 }
28
29 int strncmp_icase(const char *a, const char *b, size_t count)
30 {
31 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
32 }
33
34 int fnmatch_icase(const char *pattern, const char *string, int flags)
35 {
36 return fnmatch(pattern, string, flags | (ignore_case ? FNM_CASEFOLD : 0));
37 }
38
39 static size_t common_prefix_len(const char **pathspec)
40 {
41 const char *n, *first;
42 size_t max = 0;
43
44 if (!pathspec)
45 return max;
46
47 first = *pathspec;
48 while ((n = *pathspec++)) {
49 size_t i, len = 0;
50 for (i = 0; first == n || i < max; i++) {
51 char c = n[i];
52 if (!c || c != first[i] || is_glob_special(c))
53 break;
54 if (c == '/')
55 len = i + 1;
56 }
57 if (first == n || len < max) {
58 max = len;
59 if (!max)
60 break;
61 }
62 }
63 return max;
64 }
65
66 /*
67 * Returns a copy of the longest leading path common among all
68 * pathspecs.
69 */
70 char *common_prefix(const char **pathspec)
71 {
72 unsigned long len = common_prefix_len(pathspec);
73
74 return len ? xmemdupz(*pathspec, len) : NULL;
75 }
76
77 int fill_directory(struct dir_struct *dir, const char **pathspec)
78 {
79 size_t len;
80
81 /*
82 * Calculate common prefix for the pathspec, and
83 * use that to optimize the directory walk
84 */
85 len = common_prefix_len(pathspec);
86
87 /* Read the directory and prune it */
88 read_directory(dir, pathspec ? *pathspec : "", len, pathspec);
89 return len;
90 }
91
92 int within_depth(const char *name, int namelen,
93 int depth, int max_depth)
94 {
95 const char *cp = name, *cpe = name + namelen;
96
97 while (cp < cpe) {
98 if (*cp++ != '/')
99 continue;
100 depth++;
101 if (depth > max_depth)
102 return 0;
103 }
104 return 1;
105 }
106
107 /*
108 * Does 'match' match the given name?
109 * A match is found if
110 *
111 * (1) the 'match' string is leading directory of 'name', or
112 * (2) the 'match' string is a wildcard and matches 'name', or
113 * (3) the 'match' string is exactly the same as 'name'.
114 *
115 * and the return value tells which case it was.
116 *
117 * It returns 0 when there is no match.
118 */
119 static int match_one(const char *match, const char *name, int namelen)
120 {
121 int matchlen;
122
123 /* If the match was just the prefix, we matched */
124 if (!*match)
125 return MATCHED_RECURSIVELY;
126
127 if (ignore_case) {
128 for (;;) {
129 unsigned char c1 = tolower(*match);
130 unsigned char c2 = tolower(*name);
131 if (c1 == '\0' || is_glob_special(c1))
132 break;
133 if (c1 != c2)
134 return 0;
135 match++;
136 name++;
137 namelen--;
138 }
139 } else {
140 for (;;) {
141 unsigned char c1 = *match;
142 unsigned char c2 = *name;
143 if (c1 == '\0' || is_glob_special(c1))
144 break;
145 if (c1 != c2)
146 return 0;
147 match++;
148 name++;
149 namelen--;
150 }
151 }
152
153
154 /*
155 * If we don't match the matchstring exactly,
156 * we need to match by fnmatch
157 */
158 matchlen = strlen(match);
159 if (strncmp_icase(match, name, matchlen))
160 return !fnmatch_icase(match, name, 0) ? MATCHED_FNMATCH : 0;
161
162 if (namelen == matchlen)
163 return MATCHED_EXACTLY;
164 if (match[matchlen-1] == '/' || name[matchlen] == '/')
165 return MATCHED_RECURSIVELY;
166 return 0;
167 }
168
169 /*
170 * Given a name and a list of pathspecs, see if the name matches
171 * any of the pathspecs. The caller is also interested in seeing
172 * all pathspec matches some names it calls this function with
173 * (otherwise the user could have mistyped the unmatched pathspec),
174 * and a mark is left in seen[] array for pathspec element that
175 * actually matched anything.
176 */
177 int match_pathspec(const char **pathspec, const char *name, int namelen,
178 int prefix, char *seen)
179 {
180 int i, retval = 0;
181
182 if (!pathspec)
183 return 1;
184
185 name += prefix;
186 namelen -= prefix;
187
188 for (i = 0; pathspec[i] != NULL; i++) {
189 int how;
190 const char *match = pathspec[i] + prefix;
191 if (seen && seen[i] == MATCHED_EXACTLY)
192 continue;
193 how = match_one(match, name, namelen);
194 if (how) {
195 if (retval < how)
196 retval = how;
197 if (seen && seen[i] < how)
198 seen[i] = how;
199 }
200 }
201 return retval;
202 }
203
204 /*
205 * Does 'match' match the given name?
206 * A match is found if
207 *
208 * (1) the 'match' string is leading directory of 'name', or
209 * (2) the 'match' string is a wildcard and matches 'name', or
210 * (3) the 'match' string is exactly the same as 'name'.
211 *
212 * and the return value tells which case it was.
213 *
214 * It returns 0 when there is no match.
215 */
216 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
217 const char *name, int namelen)
218 {
219 /* name/namelen has prefix cut off by caller */
220 const char *match = item->match + prefix;
221 int matchlen = item->len - prefix;
222
223 /* If the match was just the prefix, we matched */
224 if (!*match)
225 return MATCHED_RECURSIVELY;
226
227 if (matchlen <= namelen && !strncmp(match, name, matchlen)) {
228 if (matchlen == namelen)
229 return MATCHED_EXACTLY;
230
231 if (match[matchlen-1] == '/' || name[matchlen] == '/')
232 return MATCHED_RECURSIVELY;
233 }
234
235 if (item->use_wildcard && !fnmatch(match, name, 0))
236 return MATCHED_FNMATCH;
237
238 return 0;
239 }
240
241 /*
242 * Given a name and a list of pathspecs, see if the name matches
243 * any of the pathspecs. The caller is also interested in seeing
244 * all pathspec matches some names it calls this function with
245 * (otherwise the user could have mistyped the unmatched pathspec),
246 * and a mark is left in seen[] array for pathspec element that
247 * actually matched anything.
248 */
249 int match_pathspec_depth(const struct pathspec *ps,
250 const char *name, int namelen,
251 int prefix, char *seen)
252 {
253 int i, retval = 0;
254
255 if (!ps->nr) {
256 if (!ps->recursive || ps->max_depth == -1)
257 return MATCHED_RECURSIVELY;
258
259 if (within_depth(name, namelen, 0, ps->max_depth))
260 return MATCHED_EXACTLY;
261 else
262 return 0;
263 }
264
265 name += prefix;
266 namelen -= prefix;
267
268 for (i = ps->nr - 1; i >= 0; i--) {
269 int how;
270 if (seen && seen[i] == MATCHED_EXACTLY)
271 continue;
272 how = match_pathspec_item(ps->items+i, prefix, name, namelen);
273 if (ps->recursive && ps->max_depth != -1 &&
274 how && how != MATCHED_FNMATCH) {
275 int len = ps->items[i].len;
276 if (name[len] == '/')
277 len++;
278 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
279 how = MATCHED_EXACTLY;
280 else
281 how = 0;
282 }
283 if (how) {
284 if (retval < how)
285 retval = how;
286 if (seen && seen[i] < how)
287 seen[i] = how;
288 }
289 }
290 return retval;
291 }
292
293 /*
294 * Return the length of the "simple" part of a path match limiter.
295 */
296 static int simple_length(const char *match)
297 {
298 int len = -1;
299
300 for (;;) {
301 unsigned char c = *match++;
302 len++;
303 if (c == '\0' || is_glob_special(c))
304 return len;
305 }
306 }
307
308 static int no_wildcard(const char *string)
309 {
310 return string[simple_length(string)] == '\0';
311 }
312
313 void parse_exclude_pattern(const char **pattern,
314 int *patternlen,
315 int *flags,
316 int *nowildcardlen)
317 {
318 const char *p = *pattern;
319 size_t i, len;
320
321 *flags = 0;
322 if (*p == '!') {
323 *flags |= EXC_FLAG_NEGATIVE;
324 p++;
325 }
326 len = strlen(p);
327 if (len && p[len - 1] == '/') {
328 len--;
329 *flags |= EXC_FLAG_MUSTBEDIR;
330 }
331 for (i = 0; i < len; i++) {
332 if (p[i] == '/')
333 break;
334 }
335 if (i == len)
336 *flags |= EXC_FLAG_NODIR;
337 *nowildcardlen = simple_length(p);
338 /*
339 * we should have excluded the trailing slash from 'p' too,
340 * but that's one more allocation. Instead just make sure
341 * nowildcardlen does not exceed real patternlen
342 */
343 if (*nowildcardlen > len)
344 *nowildcardlen = len;
345 if (*p == '*' && no_wildcard(p + 1))
346 *flags |= EXC_FLAG_ENDSWITH;
347 *pattern = p;
348 *patternlen = len;
349 }
350
351 void add_exclude(const char *string, const char *base,
352 int baselen, struct exclude_list *el)
353 {
354 struct exclude *x;
355 int patternlen;
356 int flags;
357 int nowildcardlen;
358
359 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
360 if (flags & EXC_FLAG_MUSTBEDIR) {
361 char *s;
362 x = xmalloc(sizeof(*x) + patternlen + 1);
363 s = (char *)(x+1);
364 memcpy(s, string, patternlen);
365 s[patternlen] = '\0';
366 x->pattern = s;
367 } else {
368 x = xmalloc(sizeof(*x));
369 x->pattern = string;
370 }
371 x->patternlen = patternlen;
372 x->nowildcardlen = nowildcardlen;
373 x->base = base;
374 x->baselen = baselen;
375 x->flags = flags;
376 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
377 el->excludes[el->nr++] = x;
378 }
379
380 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
381 {
382 int pos, len;
383 unsigned long sz;
384 enum object_type type;
385 void *data;
386 struct index_state *istate = &the_index;
387
388 len = strlen(path);
389 pos = index_name_pos(istate, path, len);
390 if (pos < 0)
391 return NULL;
392 if (!ce_skip_worktree(istate->cache[pos]))
393 return NULL;
394 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
395 if (!data || type != OBJ_BLOB) {
396 free(data);
397 return NULL;
398 }
399 *size = xsize_t(sz);
400 return data;
401 }
402
403 /*
404 * Frees memory within el which was allocated for exclude patterns and
405 * the file buffer. Does not free el itself.
406 */
407 void clear_exclude_list(struct exclude_list *el)
408 {
409 int i;
410
411 for (i = 0; i < el->nr; i++)
412 free(el->excludes[i]);
413 free(el->excludes);
414
415 el->nr = 0;
416 el->excludes = NULL;
417 }
418
419 int add_excludes_from_file_to_list(const char *fname,
420 const char *base,
421 int baselen,
422 char **buf_p,
423 struct exclude_list *el,
424 int check_index)
425 {
426 struct stat st;
427 int fd, i;
428 size_t size = 0;
429 char *buf, *entry;
430
431 fd = open(fname, O_RDONLY);
432 if (fd < 0 || fstat(fd, &st) < 0) {
433 if (0 <= fd)
434 close(fd);
435 if (!check_index ||
436 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
437 return -1;
438 if (size == 0) {
439 free(buf);
440 return 0;
441 }
442 if (buf[size-1] != '\n') {
443 buf = xrealloc(buf, size+1);
444 buf[size++] = '\n';
445 }
446 }
447 else {
448 size = xsize_t(st.st_size);
449 if (size == 0) {
450 close(fd);
451 return 0;
452 }
453 buf = xmalloc(size+1);
454 if (read_in_full(fd, buf, size) != size) {
455 free(buf);
456 close(fd);
457 return -1;
458 }
459 buf[size++] = '\n';
460 close(fd);
461 }
462
463 if (buf_p)
464 *buf_p = buf;
465 entry = buf;
466 for (i = 0; i < size; i++) {
467 if (buf[i] == '\n') {
468 if (entry != buf + i && entry[0] != '#') {
469 buf[i - (i && buf[i-1] == '\r')] = 0;
470 add_exclude(entry, base, baselen, el);
471 }
472 entry = buf + i + 1;
473 }
474 }
475 return 0;
476 }
477
478 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
479 {
480 if (add_excludes_from_file_to_list(fname, "", 0, NULL,
481 &dir->exclude_list[EXC_FILE], 0) < 0)
482 die("cannot use %s as an exclude file", fname);
483 }
484
485 /*
486 * Loads the per-directory exclude list for the substring of base
487 * which has a char length of baselen.
488 */
489 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
490 {
491 struct exclude_list *el;
492 struct exclude_stack *stk = NULL;
493 int current;
494
495 if ((!dir->exclude_per_dir) ||
496 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
497 return; /* too long a path -- ignore */
498
499 /* Pop the directories that are not the prefix of the path being checked. */
500 el = &dir->exclude_list[EXC_DIRS];
501 while ((stk = dir->exclude_stack) != NULL) {
502 if (stk->baselen <= baselen &&
503 !strncmp(dir->basebuf, base, stk->baselen))
504 break;
505 dir->exclude_stack = stk->prev;
506 while (stk->exclude_ix < el->nr)
507 free(el->excludes[--el->nr]);
508 free(stk->filebuf);
509 free(stk);
510 }
511
512 /* Read from the parent directories and push them down. */
513 current = stk ? stk->baselen : -1;
514 while (current < baselen) {
515 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
516 const char *cp;
517
518 if (current < 0) {
519 cp = base;
520 current = 0;
521 }
522 else {
523 cp = strchr(base + current + 1, '/');
524 if (!cp)
525 die("oops in prep_exclude");
526 cp++;
527 }
528 stk->prev = dir->exclude_stack;
529 stk->baselen = cp - base;
530 stk->exclude_ix = el->nr;
531 memcpy(dir->basebuf + current, base + current,
532 stk->baselen - current);
533 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
534 add_excludes_from_file_to_list(dir->basebuf,
535 dir->basebuf, stk->baselen,
536 &stk->filebuf, el, 1);
537 dir->exclude_stack = stk;
538 current = stk->baselen;
539 }
540 dir->basebuf[baselen] = '\0';
541 }
542
543 int match_basename(const char *basename, int basenamelen,
544 const char *pattern, int prefix, int patternlen,
545 int flags)
546 {
547 if (prefix == patternlen) {
548 if (!strcmp_icase(pattern, basename))
549 return 1;
550 } else if (flags & EXC_FLAG_ENDSWITH) {
551 if (patternlen - 1 <= basenamelen &&
552 !strcmp_icase(pattern + 1,
553 basename + basenamelen - patternlen + 1))
554 return 1;
555 } else {
556 if (fnmatch_icase(pattern, basename, 0) == 0)
557 return 1;
558 }
559 return 0;
560 }
561
562 int match_pathname(const char *pathname, int pathlen,
563 const char *base, int baselen,
564 const char *pattern, int prefix, int patternlen,
565 int flags)
566 {
567 const char *name;
568 int namelen;
569
570 /*
571 * match with FNM_PATHNAME; the pattern has base implicitly
572 * in front of it.
573 */
574 if (*pattern == '/') {
575 pattern++;
576 prefix--;
577 }
578
579 /*
580 * baselen does not count the trailing slash. base[] may or
581 * may not end with a trailing slash though.
582 */
583 if (pathlen < baselen + 1 ||
584 (baselen && pathname[baselen] != '/') ||
585 strncmp_icase(pathname, base, baselen))
586 return 0;
587
588 namelen = baselen ? pathlen - baselen - 1 : pathlen;
589 name = pathname + pathlen - namelen;
590
591 if (prefix) {
592 /*
593 * if the non-wildcard part is longer than the
594 * remaining pathname, surely it cannot match.
595 */
596 if (prefix > namelen)
597 return 0;
598
599 if (strncmp_icase(pattern, name, prefix))
600 return 0;
601 pattern += prefix;
602 name += prefix;
603 namelen -= prefix;
604 }
605
606 return fnmatch_icase(pattern, name, FNM_PATHNAME) == 0;
607 }
608
609 /*
610 * Scan the given exclude list in reverse to see whether pathname
611 * should be ignored. The first match (i.e. the last on the list), if
612 * any, determines the fate. Returns the exclude_list element which
613 * matched, or NULL for undecided.
614 */
615 static struct exclude *last_exclude_matching_from_list(const char *pathname,
616 int pathlen,
617 const char *basename,
618 int *dtype,
619 struct exclude_list *el)
620 {
621 int i;
622
623 if (!el->nr)
624 return NULL; /* undefined */
625
626 for (i = el->nr - 1; 0 <= i; i--) {
627 struct exclude *x = el->excludes[i];
628 const char *exclude = x->pattern;
629 int prefix = x->nowildcardlen;
630
631 if (x->flags & EXC_FLAG_MUSTBEDIR) {
632 if (*dtype == DT_UNKNOWN)
633 *dtype = get_dtype(NULL, pathname, pathlen);
634 if (*dtype != DT_DIR)
635 continue;
636 }
637
638 if (x->flags & EXC_FLAG_NODIR) {
639 if (match_basename(basename,
640 pathlen - (basename - pathname),
641 exclude, prefix, x->patternlen,
642 x->flags))
643 return x;
644 continue;
645 }
646
647 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
648 if (match_pathname(pathname, pathlen,
649 x->base, x->baselen ? x->baselen - 1 : 0,
650 exclude, prefix, x->patternlen, x->flags))
651 return x;
652 }
653 return NULL; /* undecided */
654 }
655
656 /*
657 * Scan the list and let the last match determine the fate.
658 * Return 1 for exclude, 0 for include and -1 for undecided.
659 */
660 int is_excluded_from_list(const char *pathname,
661 int pathlen, const char *basename, int *dtype,
662 struct exclude_list *el)
663 {
664 struct exclude *exclude;
665 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
666 if (exclude)
667 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
668 return -1; /* undecided */
669 }
670
671 /*
672 * Loads the exclude lists for the directory containing pathname, then
673 * scans all exclude lists to determine whether pathname is excluded.
674 * Returns the exclude_list element which matched, or NULL for
675 * undecided.
676 */
677 static struct exclude *last_exclude_matching(struct dir_struct *dir,
678 const char *pathname,
679 int *dtype_p)
680 {
681 int pathlen = strlen(pathname);
682 int st;
683 struct exclude *exclude;
684 const char *basename = strrchr(pathname, '/');
685 basename = (basename) ? basename+1 : pathname;
686
687 prep_exclude(dir, pathname, basename-pathname);
688 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
689 exclude = last_exclude_matching_from_list(
690 pathname, pathlen, basename, dtype_p,
691 &dir->exclude_list[st]);
692 if (exclude)
693 return exclude;
694 }
695 return NULL;
696 }
697
698 /*
699 * Loads the exclude lists for the directory containing pathname, then
700 * scans all exclude lists to determine whether pathname is excluded.
701 * Returns 1 if true, otherwise 0.
702 */
703 static int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
704 {
705 struct exclude *exclude =
706 last_exclude_matching(dir, pathname, dtype_p);
707 if (exclude)
708 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
709 return 0;
710 }
711
712 void path_exclude_check_init(struct path_exclude_check *check,
713 struct dir_struct *dir)
714 {
715 check->dir = dir;
716 check->exclude = NULL;
717 strbuf_init(&check->path, 256);
718 }
719
720 void path_exclude_check_clear(struct path_exclude_check *check)
721 {
722 strbuf_release(&check->path);
723 }
724
725 /*
726 * For each subdirectory in name, starting with the top-most, checks
727 * to see if that subdirectory is excluded, and if so, returns the
728 * corresponding exclude structure. Otherwise, checks whether name
729 * itself (which is presumably a file) is excluded.
730 *
731 * A path to a directory known to be excluded is left in check->path to
732 * optimize for repeated checks for files in the same excluded directory.
733 */
734 struct exclude *last_exclude_matching_path(struct path_exclude_check *check,
735 const char *name, int namelen,
736 int *dtype)
737 {
738 int i;
739 struct strbuf *path = &check->path;
740 struct exclude *exclude;
741
742 /*
743 * we allow the caller to pass namelen as an optimization; it
744 * must match the length of the name, as we eventually call
745 * is_excluded() on the whole name string.
746 */
747 if (namelen < 0)
748 namelen = strlen(name);
749
750 /*
751 * If path is non-empty, and name is equal to path or a
752 * subdirectory of path, name should be excluded, because
753 * it's inside a directory which is already known to be
754 * excluded and was previously left in check->path.
755 */
756 if (path->len &&
757 path->len <= namelen &&
758 !memcmp(name, path->buf, path->len) &&
759 (!name[path->len] || name[path->len] == '/'))
760 return check->exclude;
761
762 strbuf_setlen(path, 0);
763 for (i = 0; name[i]; i++) {
764 int ch = name[i];
765
766 if (ch == '/') {
767 int dt = DT_DIR;
768 exclude = last_exclude_matching(check->dir,
769 path->buf, &dt);
770 if (exclude) {
771 check->exclude = exclude;
772 return exclude;
773 }
774 }
775 strbuf_addch(path, ch);
776 }
777
778 /* An entry in the index; cannot be a directory with subentries */
779 strbuf_setlen(path, 0);
780
781 return last_exclude_matching(check->dir, name, dtype);
782 }
783
784 /*
785 * Is this name excluded? This is for a caller like show_files() that
786 * do not honor directory hierarchy and iterate through paths that are
787 * possibly in an ignored directory.
788 */
789 int is_path_excluded(struct path_exclude_check *check,
790 const char *name, int namelen, int *dtype)
791 {
792 struct exclude *exclude =
793 last_exclude_matching_path(check, name, namelen, dtype);
794 if (exclude)
795 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
796 return 0;
797 }
798
799 static struct dir_entry *dir_entry_new(const char *pathname, int len)
800 {
801 struct dir_entry *ent;
802
803 ent = xmalloc(sizeof(*ent) + len + 1);
804 ent->len = len;
805 memcpy(ent->name, pathname, len);
806 ent->name[len] = 0;
807 return ent;
808 }
809
810 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
811 {
812 if (cache_name_exists(pathname, len, ignore_case))
813 return NULL;
814
815 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
816 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
817 }
818
819 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
820 {
821 if (!cache_name_is_other(pathname, len))
822 return NULL;
823
824 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
825 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
826 }
827
828 enum exist_status {
829 index_nonexistent = 0,
830 index_directory,
831 index_gitdir
832 };
833
834 /*
835 * Do not use the alphabetically stored index to look up
836 * the directory name; instead, use the case insensitive
837 * name hash.
838 */
839 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
840 {
841 struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
842 unsigned char endchar;
843
844 if (!ce)
845 return index_nonexistent;
846 endchar = ce->name[len];
847
848 /*
849 * The cache_entry structure returned will contain this dirname
850 * and possibly additional path components.
851 */
852 if (endchar == '/')
853 return index_directory;
854
855 /*
856 * If there are no additional path components, then this cache_entry
857 * represents a submodule. Submodules, despite being directories,
858 * are stored in the cache without a closing slash.
859 */
860 if (!endchar && S_ISGITLINK(ce->ce_mode))
861 return index_gitdir;
862
863 /* This should never be hit, but it exists just in case. */
864 return index_nonexistent;
865 }
866
867 /*
868 * The index sorts alphabetically by entry name, which
869 * means that a gitlink sorts as '\0' at the end, while
870 * a directory (which is defined not as an entry, but as
871 * the files it contains) will sort with the '/' at the
872 * end.
873 */
874 static enum exist_status directory_exists_in_index(const char *dirname, int len)
875 {
876 int pos;
877
878 if (ignore_case)
879 return directory_exists_in_index_icase(dirname, len);
880
881 pos = cache_name_pos(dirname, len);
882 if (pos < 0)
883 pos = -pos-1;
884 while (pos < active_nr) {
885 struct cache_entry *ce = active_cache[pos++];
886 unsigned char endchar;
887
888 if (strncmp(ce->name, dirname, len))
889 break;
890 endchar = ce->name[len];
891 if (endchar > '/')
892 break;
893 if (endchar == '/')
894 return index_directory;
895 if (!endchar && S_ISGITLINK(ce->ce_mode))
896 return index_gitdir;
897 }
898 return index_nonexistent;
899 }
900
901 /*
902 * When we find a directory when traversing the filesystem, we
903 * have three distinct cases:
904 *
905 * - ignore it
906 * - see it as a directory
907 * - recurse into it
908 *
909 * and which one we choose depends on a combination of existing
910 * git index contents and the flags passed into the directory
911 * traversal routine.
912 *
913 * Case 1: If we *already* have entries in the index under that
914 * directory name, we always recurse into the directory to see
915 * all the files.
916 *
917 * Case 2: If we *already* have that directory name as a gitlink,
918 * we always continue to see it as a gitlink, regardless of whether
919 * there is an actual git directory there or not (it might not
920 * be checked out as a subproject!)
921 *
922 * Case 3: if we didn't have it in the index previously, we
923 * have a few sub-cases:
924 *
925 * (a) if "show_other_directories" is true, we show it as
926 * just a directory, unless "hide_empty_directories" is
927 * also true and the directory is empty, in which case
928 * we just ignore it entirely.
929 * (b) if it looks like a git directory, and we don't have
930 * 'no_gitlinks' set we treat it as a gitlink, and show it
931 * as a directory.
932 * (c) otherwise, we recurse into it.
933 */
934 enum directory_treatment {
935 show_directory,
936 ignore_directory,
937 recurse_into_directory
938 };
939
940 static enum directory_treatment treat_directory(struct dir_struct *dir,
941 const char *dirname, int len,
942 const struct path_simplify *simplify)
943 {
944 /* The "len-1" is to strip the final '/' */
945 switch (directory_exists_in_index(dirname, len-1)) {
946 case index_directory:
947 return recurse_into_directory;
948
949 case index_gitdir:
950 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
951 return ignore_directory;
952 return show_directory;
953
954 case index_nonexistent:
955 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
956 break;
957 if (!(dir->flags & DIR_NO_GITLINKS)) {
958 unsigned char sha1[20];
959 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
960 return show_directory;
961 }
962 return recurse_into_directory;
963 }
964
965 /* This is the "show_other_directories" case */
966 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
967 return show_directory;
968 if (!read_directory_recursive(dir, dirname, len, 1, simplify))
969 return ignore_directory;
970 return show_directory;
971 }
972
973 /*
974 * This is an inexact early pruning of any recursive directory
975 * reading - if the path cannot possibly be in the pathspec,
976 * return true, and we'll skip it early.
977 */
978 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
979 {
980 if (simplify) {
981 for (;;) {
982 const char *match = simplify->path;
983 int len = simplify->len;
984
985 if (!match)
986 break;
987 if (len > pathlen)
988 len = pathlen;
989 if (!memcmp(path, match, len))
990 return 0;
991 simplify++;
992 }
993 return 1;
994 }
995 return 0;
996 }
997
998 /*
999 * This function tells us whether an excluded path matches a
1000 * list of "interesting" pathspecs. That is, whether a path matched
1001 * by any of the pathspecs could possibly be ignored by excluding
1002 * the specified path. This can happen if:
1003 *
1004 * 1. the path is mentioned explicitly in the pathspec
1005 *
1006 * 2. the path is a directory prefix of some element in the
1007 * pathspec
1008 */
1009 static int exclude_matches_pathspec(const char *path, int len,
1010 const struct path_simplify *simplify)
1011 {
1012 if (simplify) {
1013 for (; simplify->path; simplify++) {
1014 if (len == simplify->len
1015 && !memcmp(path, simplify->path, len))
1016 return 1;
1017 if (len < simplify->len
1018 && simplify->path[len] == '/'
1019 && !memcmp(path, simplify->path, len))
1020 return 1;
1021 }
1022 }
1023 return 0;
1024 }
1025
1026 static int get_index_dtype(const char *path, int len)
1027 {
1028 int pos;
1029 struct cache_entry *ce;
1030
1031 ce = cache_name_exists(path, len, 0);
1032 if (ce) {
1033 if (!ce_uptodate(ce))
1034 return DT_UNKNOWN;
1035 if (S_ISGITLINK(ce->ce_mode))
1036 return DT_DIR;
1037 /*
1038 * Nobody actually cares about the
1039 * difference between DT_LNK and DT_REG
1040 */
1041 return DT_REG;
1042 }
1043
1044 /* Try to look it up as a directory */
1045 pos = cache_name_pos(path, len);
1046 if (pos >= 0)
1047 return DT_UNKNOWN;
1048 pos = -pos-1;
1049 while (pos < active_nr) {
1050 ce = active_cache[pos++];
1051 if (strncmp(ce->name, path, len))
1052 break;
1053 if (ce->name[len] > '/')
1054 break;
1055 if (ce->name[len] < '/')
1056 continue;
1057 if (!ce_uptodate(ce))
1058 break; /* continue? */
1059 return DT_DIR;
1060 }
1061 return DT_UNKNOWN;
1062 }
1063
1064 static int get_dtype(struct dirent *de, const char *path, int len)
1065 {
1066 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1067 struct stat st;
1068
1069 if (dtype != DT_UNKNOWN)
1070 return dtype;
1071 dtype = get_index_dtype(path, len);
1072 if (dtype != DT_UNKNOWN)
1073 return dtype;
1074 if (lstat(path, &st))
1075 return dtype;
1076 if (S_ISREG(st.st_mode))
1077 return DT_REG;
1078 if (S_ISDIR(st.st_mode))
1079 return DT_DIR;
1080 if (S_ISLNK(st.st_mode))
1081 return DT_LNK;
1082 return dtype;
1083 }
1084
1085 enum path_treatment {
1086 path_ignored,
1087 path_handled,
1088 path_recurse
1089 };
1090
1091 static enum path_treatment treat_one_path(struct dir_struct *dir,
1092 struct strbuf *path,
1093 const struct path_simplify *simplify,
1094 int dtype, struct dirent *de)
1095 {
1096 int exclude = is_excluded(dir, path->buf, &dtype);
1097 if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
1098 && exclude_matches_pathspec(path->buf, path->len, simplify))
1099 dir_add_ignored(dir, path->buf, path->len);
1100
1101 /*
1102 * Excluded? If we don't explicitly want to show
1103 * ignored files, ignore it
1104 */
1105 if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
1106 return path_ignored;
1107
1108 if (dtype == DT_UNKNOWN)
1109 dtype = get_dtype(de, path->buf, path->len);
1110
1111 /*
1112 * Do we want to see just the ignored files?
1113 * We still need to recurse into directories,
1114 * even if we don't ignore them, since the
1115 * directory may contain files that we do..
1116 */
1117 if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
1118 if (dtype != DT_DIR)
1119 return path_ignored;
1120 }
1121
1122 switch (dtype) {
1123 default:
1124 return path_ignored;
1125 case DT_DIR:
1126 strbuf_addch(path, '/');
1127 switch (treat_directory(dir, path->buf, path->len, simplify)) {
1128 case show_directory:
1129 if (exclude != !!(dir->flags
1130 & DIR_SHOW_IGNORED))
1131 return path_ignored;
1132 break;
1133 case recurse_into_directory:
1134 return path_recurse;
1135 case ignore_directory:
1136 return path_ignored;
1137 }
1138 break;
1139 case DT_REG:
1140 case DT_LNK:
1141 break;
1142 }
1143 return path_handled;
1144 }
1145
1146 static enum path_treatment treat_path(struct dir_struct *dir,
1147 struct dirent *de,
1148 struct strbuf *path,
1149 int baselen,
1150 const struct path_simplify *simplify)
1151 {
1152 int dtype;
1153
1154 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1155 return path_ignored;
1156 strbuf_setlen(path, baselen);
1157 strbuf_addstr(path, de->d_name);
1158 if (simplify_away(path->buf, path->len, simplify))
1159 return path_ignored;
1160
1161 dtype = DTYPE(de);
1162 return treat_one_path(dir, path, simplify, dtype, de);
1163 }
1164
1165 /*
1166 * Read a directory tree. We currently ignore anything but
1167 * directories, regular files and symlinks. That's because git
1168 * doesn't handle them at all yet. Maybe that will change some
1169 * day.
1170 *
1171 * Also, we ignore the name ".git" (even if it is not a directory).
1172 * That likely will not change.
1173 */
1174 static int read_directory_recursive(struct dir_struct *dir,
1175 const char *base, int baselen,
1176 int check_only,
1177 const struct path_simplify *simplify)
1178 {
1179 DIR *fdir;
1180 int contents = 0;
1181 struct dirent *de;
1182 struct strbuf path = STRBUF_INIT;
1183
1184 strbuf_add(&path, base, baselen);
1185
1186 fdir = opendir(path.len ? path.buf : ".");
1187 if (!fdir)
1188 goto out;
1189
1190 while ((de = readdir(fdir)) != NULL) {
1191 switch (treat_path(dir, de, &path, baselen, simplify)) {
1192 case path_recurse:
1193 contents += read_directory_recursive(dir, path.buf,
1194 path.len, 0,
1195 simplify);
1196 continue;
1197 case path_ignored:
1198 continue;
1199 case path_handled:
1200 break;
1201 }
1202 contents++;
1203 if (check_only)
1204 break;
1205 dir_add_name(dir, path.buf, path.len);
1206 }
1207 closedir(fdir);
1208 out:
1209 strbuf_release(&path);
1210
1211 return contents;
1212 }
1213
1214 static int cmp_name(const void *p1, const void *p2)
1215 {
1216 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1217 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1218
1219 return cache_name_compare(e1->name, e1->len,
1220 e2->name, e2->len);
1221 }
1222
1223 static struct path_simplify *create_simplify(const char **pathspec)
1224 {
1225 int nr, alloc = 0;
1226 struct path_simplify *simplify = NULL;
1227
1228 if (!pathspec)
1229 return NULL;
1230
1231 for (nr = 0 ; ; nr++) {
1232 const char *match;
1233 if (nr >= alloc) {
1234 alloc = alloc_nr(alloc);
1235 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
1236 }
1237 match = *pathspec++;
1238 if (!match)
1239 break;
1240 simplify[nr].path = match;
1241 simplify[nr].len = simple_length(match);
1242 }
1243 simplify[nr].path = NULL;
1244 simplify[nr].len = 0;
1245 return simplify;
1246 }
1247
1248 static void free_simplify(struct path_simplify *simplify)
1249 {
1250 free(simplify);
1251 }
1252
1253 static int treat_leading_path(struct dir_struct *dir,
1254 const char *path, int len,
1255 const struct path_simplify *simplify)
1256 {
1257 struct strbuf sb = STRBUF_INIT;
1258 int baselen, rc = 0;
1259 const char *cp;
1260
1261 while (len && path[len - 1] == '/')
1262 len--;
1263 if (!len)
1264 return 1;
1265 baselen = 0;
1266 while (1) {
1267 cp = path + baselen + !!baselen;
1268 cp = memchr(cp, '/', path + len - cp);
1269 if (!cp)
1270 baselen = len;
1271 else
1272 baselen = cp - path;
1273 strbuf_setlen(&sb, 0);
1274 strbuf_add(&sb, path, baselen);
1275 if (!is_directory(sb.buf))
1276 break;
1277 if (simplify_away(sb.buf, sb.len, simplify))
1278 break;
1279 if (treat_one_path(dir, &sb, simplify,
1280 DT_DIR, NULL) == path_ignored)
1281 break; /* do not recurse into it */
1282 if (len <= baselen) {
1283 rc = 1;
1284 break; /* finished checking */
1285 }
1286 }
1287 strbuf_release(&sb);
1288 return rc;
1289 }
1290
1291 int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
1292 {
1293 struct path_simplify *simplify;
1294
1295 if (has_symlink_leading_path(path, len))
1296 return dir->nr;
1297
1298 simplify = create_simplify(pathspec);
1299 if (!len || treat_leading_path(dir, path, len, simplify))
1300 read_directory_recursive(dir, path, len, 0, simplify);
1301 free_simplify(simplify);
1302 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1303 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1304 return dir->nr;
1305 }
1306
1307 int file_exists(const char *f)
1308 {
1309 struct stat sb;
1310 return lstat(f, &sb) == 0;
1311 }
1312
1313 /*
1314 * Given two normalized paths (a trailing slash is ok), if subdir is
1315 * outside dir, return -1. Otherwise return the offset in subdir that
1316 * can be used as relative path to dir.
1317 */
1318 int dir_inside_of(const char *subdir, const char *dir)
1319 {
1320 int offset = 0;
1321
1322 assert(dir && subdir && *dir && *subdir);
1323
1324 while (*dir && *subdir && *dir == *subdir) {
1325 dir++;
1326 subdir++;
1327 offset++;
1328 }
1329
1330 /* hel[p]/me vs hel[l]/yeah */
1331 if (*dir && *subdir)
1332 return -1;
1333
1334 if (!*subdir)
1335 return !*dir ? offset : -1; /* same dir */
1336
1337 /* foo/[b]ar vs foo/[] */
1338 if (is_dir_sep(dir[-1]))
1339 return is_dir_sep(subdir[-1]) ? offset : -1;
1340
1341 /* foo[/]bar vs foo[] */
1342 return is_dir_sep(*subdir) ? offset + 1 : -1;
1343 }
1344
1345 int is_inside_dir(const char *dir)
1346 {
1347 char cwd[PATH_MAX];
1348 if (!dir)
1349 return 0;
1350 if (!getcwd(cwd, sizeof(cwd)))
1351 die_errno("can't find the current directory");
1352 return dir_inside_of(cwd, dir) >= 0;
1353 }
1354
1355 int is_empty_dir(const char *path)
1356 {
1357 DIR *dir = opendir(path);
1358 struct dirent *e;
1359 int ret = 1;
1360
1361 if (!dir)
1362 return 0;
1363
1364 while ((e = readdir(dir)) != NULL)
1365 if (!is_dot_or_dotdot(e->d_name)) {
1366 ret = 0;
1367 break;
1368 }
1369
1370 closedir(dir);
1371 return ret;
1372 }
1373
1374 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1375 {
1376 DIR *dir;
1377 struct dirent *e;
1378 int ret = 0, original_len = path->len, len, kept_down = 0;
1379 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1380 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1381 unsigned char submodule_head[20];
1382
1383 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1384 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1385 /* Do not descend and nuke a nested git work tree. */
1386 if (kept_up)
1387 *kept_up = 1;
1388 return 0;
1389 }
1390
1391 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1392 dir = opendir(path->buf);
1393 if (!dir) {
1394 /* an empty dir could be removed even if it is unreadble */
1395 if (!keep_toplevel)
1396 return rmdir(path->buf);
1397 else
1398 return -1;
1399 }
1400 if (path->buf[original_len - 1] != '/')
1401 strbuf_addch(path, '/');
1402
1403 len = path->len;
1404 while ((e = readdir(dir)) != NULL) {
1405 struct stat st;
1406 if (is_dot_or_dotdot(e->d_name))
1407 continue;
1408
1409 strbuf_setlen(path, len);
1410 strbuf_addstr(path, e->d_name);
1411 if (lstat(path->buf, &st))
1412 ; /* fall thru */
1413 else if (S_ISDIR(st.st_mode)) {
1414 if (!remove_dir_recurse(path, flag, &kept_down))
1415 continue; /* happy */
1416 } else if (!only_empty && !unlink(path->buf))
1417 continue; /* happy, too */
1418
1419 /* path too long, stat fails, or non-directory still exists */
1420 ret = -1;
1421 break;
1422 }
1423 closedir(dir);
1424
1425 strbuf_setlen(path, original_len);
1426 if (!ret && !keep_toplevel && !kept_down)
1427 ret = rmdir(path->buf);
1428 else if (kept_up)
1429 /*
1430 * report the uplevel that it is not an error that we
1431 * did not rmdir() our directory.
1432 */
1433 *kept_up = !ret;
1434 return ret;
1435 }
1436
1437 int remove_dir_recursively(struct strbuf *path, int flag)
1438 {
1439 return remove_dir_recurse(path, flag, NULL);
1440 }
1441
1442 void setup_standard_excludes(struct dir_struct *dir)
1443 {
1444 const char *path;
1445
1446 dir->exclude_per_dir = ".gitignore";
1447 path = git_path("info/exclude");
1448 if (!access(path, R_OK))
1449 add_excludes_from_file(dir, path);
1450 if (excludes_file && !access(excludes_file, R_OK))
1451 add_excludes_from_file(dir, excludes_file);
1452 }
1453
1454 int remove_path(const char *name)
1455 {
1456 char *slash;
1457
1458 if (unlink(name) && errno != ENOENT)
1459 return -1;
1460
1461 slash = strrchr(name, '/');
1462 if (slash) {
1463 char *dirs = xstrdup(name);
1464 slash = dirs + (slash - name);
1465 do {
1466 *slash = '\0';
1467 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1468 free(dirs);
1469 }
1470 return 0;
1471 }
1472
1473 static int pathspec_item_cmp(const void *a_, const void *b_)
1474 {
1475 struct pathspec_item *a, *b;
1476
1477 a = (struct pathspec_item *)a_;
1478 b = (struct pathspec_item *)b_;
1479 return strcmp(a->match, b->match);
1480 }
1481
1482 int init_pathspec(struct pathspec *pathspec, const char **paths)
1483 {
1484 const char **p = paths;
1485 int i;
1486
1487 memset(pathspec, 0, sizeof(*pathspec));
1488 if (!p)
1489 return 0;
1490 while (*p)
1491 p++;
1492 pathspec->raw = paths;
1493 pathspec->nr = p - paths;
1494 if (!pathspec->nr)
1495 return 0;
1496
1497 pathspec->items = xmalloc(sizeof(struct pathspec_item)*pathspec->nr);
1498 for (i = 0; i < pathspec->nr; i++) {
1499 struct pathspec_item *item = pathspec->items+i;
1500 const char *path = paths[i];
1501
1502 item->match = path;
1503 item->len = strlen(path);
1504 item->use_wildcard = !no_wildcard(path);
1505 if (item->use_wildcard)
1506 pathspec->has_wildcard = 1;
1507 }
1508
1509 qsort(pathspec->items, pathspec->nr,
1510 sizeof(struct pathspec_item), pathspec_item_cmp);
1511
1512 return 0;
1513 }
1514
1515 void free_pathspec(struct pathspec *pathspec)
1516 {
1517 free(pathspec->items);
1518 pathspec->items = NULL;
1519 }