dir.c: provide clear_directory() for reclaiming dir_struct memory
[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, int srcpos)
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 x->srcpos = srcpos;
377 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
378 el->excludes[el->nr++] = x;
379 x->el = el;
380 }
381
382 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
383 {
384 int pos, len;
385 unsigned long sz;
386 enum object_type type;
387 void *data;
388 struct index_state *istate = &the_index;
389
390 len = strlen(path);
391 pos = index_name_pos(istate, path, len);
392 if (pos < 0)
393 return NULL;
394 if (!ce_skip_worktree(istate->cache[pos]))
395 return NULL;
396 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
397 if (!data || type != OBJ_BLOB) {
398 free(data);
399 return NULL;
400 }
401 *size = xsize_t(sz);
402 return data;
403 }
404
405 /*
406 * Frees memory within el which was allocated for exclude patterns and
407 * the file buffer. Does not free el itself.
408 */
409 void clear_exclude_list(struct exclude_list *el)
410 {
411 int i;
412
413 for (i = 0; i < el->nr; i++)
414 free(el->excludes[i]);
415 free(el->excludes);
416 free(el->filebuf);
417
418 el->nr = 0;
419 el->excludes = NULL;
420 el->filebuf = NULL;
421 }
422
423 int add_excludes_from_file_to_list(const char *fname,
424 const char *base,
425 int baselen,
426 struct exclude_list *el,
427 int check_index)
428 {
429 struct stat st;
430 int fd, i, lineno = 1;
431 size_t size = 0;
432 char *buf, *entry;
433
434 fd = open(fname, O_RDONLY);
435 if (fd < 0 || fstat(fd, &st) < 0) {
436 if (0 <= fd)
437 close(fd);
438 if (!check_index ||
439 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
440 return -1;
441 if (size == 0) {
442 free(buf);
443 return 0;
444 }
445 if (buf[size-1] != '\n') {
446 buf = xrealloc(buf, size+1);
447 buf[size++] = '\n';
448 }
449 }
450 else {
451 size = xsize_t(st.st_size);
452 if (size == 0) {
453 close(fd);
454 return 0;
455 }
456 buf = xmalloc(size+1);
457 if (read_in_full(fd, buf, size) != size) {
458 free(buf);
459 close(fd);
460 return -1;
461 }
462 buf[size++] = '\n';
463 close(fd);
464 }
465
466 el->filebuf = buf;
467 entry = buf;
468 for (i = 0; i < size; i++) {
469 if (buf[i] == '\n') {
470 if (entry != buf + i && entry[0] != '#') {
471 buf[i - (i && buf[i-1] == '\r')] = 0;
472 add_exclude(entry, base, baselen, el, lineno);
473 }
474 lineno++;
475 entry = buf + i + 1;
476 }
477 }
478 return 0;
479 }
480
481 struct exclude_list *add_exclude_list(struct dir_struct *dir,
482 int group_type, const char *src)
483 {
484 struct exclude_list *el;
485 struct exclude_list_group *group;
486
487 group = &dir->exclude_list_group[group_type];
488 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
489 el = &group->el[group->nr++];
490 memset(el, 0, sizeof(*el));
491 el->src = src;
492 return el;
493 }
494
495 /*
496 * Used to set up core.excludesfile and .git/info/exclude lists.
497 */
498 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
499 {
500 struct exclude_list *el;
501 el = add_exclude_list(dir, EXC_FILE, fname);
502 if (add_excludes_from_file_to_list(fname, "", 0, el, 0) < 0)
503 die("cannot use %s as an exclude file", fname);
504 }
505
506 /*
507 * Loads the per-directory exclude list for the substring of base
508 * which has a char length of baselen.
509 */
510 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
511 {
512 struct exclude_list_group *group;
513 struct exclude_list *el;
514 struct exclude_stack *stk = NULL;
515 int current;
516
517 if ((!dir->exclude_per_dir) ||
518 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
519 return; /* too long a path -- ignore */
520
521 group = &dir->exclude_list_group[EXC_DIRS];
522
523 /* Pop the exclude lists from the EXCL_DIRS exclude_list_group
524 * which originate from directories not in the prefix of the
525 * path being checked. */
526 while ((stk = dir->exclude_stack) != NULL) {
527 if (stk->baselen <= baselen &&
528 !strncmp(dir->basebuf, base, stk->baselen))
529 break;
530 el = &group->el[dir->exclude_stack->exclude_ix];
531 dir->exclude_stack = stk->prev;
532 free((char *)el->src); /* see strdup() below */
533 clear_exclude_list(el);
534 free(stk);
535 group->nr--;
536 }
537
538 /* Read from the parent directories and push them down. */
539 current = stk ? stk->baselen : -1;
540 while (current < baselen) {
541 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
542 const char *cp;
543
544 if (current < 0) {
545 cp = base;
546 current = 0;
547 }
548 else {
549 cp = strchr(base + current + 1, '/');
550 if (!cp)
551 die("oops in prep_exclude");
552 cp++;
553 }
554 stk->prev = dir->exclude_stack;
555 stk->baselen = cp - base;
556 memcpy(dir->basebuf + current, base + current,
557 stk->baselen - current);
558 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
559 /*
560 * dir->basebuf gets reused by the traversal, but we
561 * need fname to remain unchanged to ensure the src
562 * member of each struct exclude correctly
563 * back-references its source file. Other invocations
564 * of add_exclude_list provide stable strings, so we
565 * strdup() and free() here in the caller.
566 */
567 el = add_exclude_list(dir, EXC_DIRS, strdup(dir->basebuf));
568 stk->exclude_ix = group->nr - 1;
569 add_excludes_from_file_to_list(dir->basebuf,
570 dir->basebuf, stk->baselen,
571 el, 1);
572 dir->exclude_stack = stk;
573 current = stk->baselen;
574 }
575 dir->basebuf[baselen] = '\0';
576 }
577
578 int match_basename(const char *basename, int basenamelen,
579 const char *pattern, int prefix, int patternlen,
580 int flags)
581 {
582 if (prefix == patternlen) {
583 if (!strcmp_icase(pattern, basename))
584 return 1;
585 } else if (flags & EXC_FLAG_ENDSWITH) {
586 if (patternlen - 1 <= basenamelen &&
587 !strcmp_icase(pattern + 1,
588 basename + basenamelen - patternlen + 1))
589 return 1;
590 } else {
591 if (fnmatch_icase(pattern, basename, 0) == 0)
592 return 1;
593 }
594 return 0;
595 }
596
597 int match_pathname(const char *pathname, int pathlen,
598 const char *base, int baselen,
599 const char *pattern, int prefix, int patternlen,
600 int flags)
601 {
602 const char *name;
603 int namelen;
604
605 /*
606 * match with FNM_PATHNAME; the pattern has base implicitly
607 * in front of it.
608 */
609 if (*pattern == '/') {
610 pattern++;
611 prefix--;
612 }
613
614 /*
615 * baselen does not count the trailing slash. base[] may or
616 * may not end with a trailing slash though.
617 */
618 if (pathlen < baselen + 1 ||
619 (baselen && pathname[baselen] != '/') ||
620 strncmp_icase(pathname, base, baselen))
621 return 0;
622
623 namelen = baselen ? pathlen - baselen - 1 : pathlen;
624 name = pathname + pathlen - namelen;
625
626 if (prefix) {
627 /*
628 * if the non-wildcard part is longer than the
629 * remaining pathname, surely it cannot match.
630 */
631 if (prefix > namelen)
632 return 0;
633
634 if (strncmp_icase(pattern, name, prefix))
635 return 0;
636 pattern += prefix;
637 name += prefix;
638 namelen -= prefix;
639 }
640
641 return fnmatch_icase(pattern, name, FNM_PATHNAME) == 0;
642 }
643
644 /*
645 * Scan the given exclude list in reverse to see whether pathname
646 * should be ignored. The first match (i.e. the last on the list), if
647 * any, determines the fate. Returns the exclude_list element which
648 * matched, or NULL for undecided.
649 */
650 static struct exclude *last_exclude_matching_from_list(const char *pathname,
651 int pathlen,
652 const char *basename,
653 int *dtype,
654 struct exclude_list *el)
655 {
656 int i;
657
658 if (!el->nr)
659 return NULL; /* undefined */
660
661 for (i = el->nr - 1; 0 <= i; i--) {
662 struct exclude *x = el->excludes[i];
663 const char *exclude = x->pattern;
664 int prefix = x->nowildcardlen;
665
666 if (x->flags & EXC_FLAG_MUSTBEDIR) {
667 if (*dtype == DT_UNKNOWN)
668 *dtype = get_dtype(NULL, pathname, pathlen);
669 if (*dtype != DT_DIR)
670 continue;
671 }
672
673 if (x->flags & EXC_FLAG_NODIR) {
674 if (match_basename(basename,
675 pathlen - (basename - pathname),
676 exclude, prefix, x->patternlen,
677 x->flags))
678 return x;
679 continue;
680 }
681
682 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
683 if (match_pathname(pathname, pathlen,
684 x->base, x->baselen ? x->baselen - 1 : 0,
685 exclude, prefix, x->patternlen, x->flags))
686 return x;
687 }
688 return NULL; /* undecided */
689 }
690
691 /*
692 * Scan the list and let the last match determine the fate.
693 * Return 1 for exclude, 0 for include and -1 for undecided.
694 */
695 int is_excluded_from_list(const char *pathname,
696 int pathlen, const char *basename, int *dtype,
697 struct exclude_list *el)
698 {
699 struct exclude *exclude;
700 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
701 if (exclude)
702 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
703 return -1; /* undecided */
704 }
705
706 /*
707 * Loads the exclude lists for the directory containing pathname, then
708 * scans all exclude lists to determine whether pathname is excluded.
709 * Returns the exclude_list element which matched, or NULL for
710 * undecided.
711 */
712 static struct exclude *last_exclude_matching(struct dir_struct *dir,
713 const char *pathname,
714 int *dtype_p)
715 {
716 int pathlen = strlen(pathname);
717 int i, j;
718 struct exclude_list_group *group;
719 struct exclude *exclude;
720 const char *basename = strrchr(pathname, '/');
721 basename = (basename) ? basename+1 : pathname;
722
723 prep_exclude(dir, pathname, basename-pathname);
724
725 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
726 group = &dir->exclude_list_group[i];
727 for (j = group->nr - 1; j >= 0; j--) {
728 exclude = last_exclude_matching_from_list(
729 pathname, pathlen, basename, dtype_p,
730 &group->el[j]);
731 if (exclude)
732 return exclude;
733 }
734 }
735 return NULL;
736 }
737
738 /*
739 * Loads the exclude lists for the directory containing pathname, then
740 * scans all exclude lists to determine whether pathname is excluded.
741 * Returns 1 if true, otherwise 0.
742 */
743 static int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
744 {
745 struct exclude *exclude =
746 last_exclude_matching(dir, pathname, dtype_p);
747 if (exclude)
748 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
749 return 0;
750 }
751
752 void path_exclude_check_init(struct path_exclude_check *check,
753 struct dir_struct *dir)
754 {
755 check->dir = dir;
756 check->exclude = NULL;
757 strbuf_init(&check->path, 256);
758 }
759
760 void path_exclude_check_clear(struct path_exclude_check *check)
761 {
762 strbuf_release(&check->path);
763 }
764
765 /*
766 * For each subdirectory in name, starting with the top-most, checks
767 * to see if that subdirectory is excluded, and if so, returns the
768 * corresponding exclude structure. Otherwise, checks whether name
769 * itself (which is presumably a file) is excluded.
770 *
771 * A path to a directory known to be excluded is left in check->path to
772 * optimize for repeated checks for files in the same excluded directory.
773 */
774 struct exclude *last_exclude_matching_path(struct path_exclude_check *check,
775 const char *name, int namelen,
776 int *dtype)
777 {
778 int i;
779 struct strbuf *path = &check->path;
780 struct exclude *exclude;
781
782 /*
783 * we allow the caller to pass namelen as an optimization; it
784 * must match the length of the name, as we eventually call
785 * is_excluded() on the whole name string.
786 */
787 if (namelen < 0)
788 namelen = strlen(name);
789
790 /*
791 * If path is non-empty, and name is equal to path or a
792 * subdirectory of path, name should be excluded, because
793 * it's inside a directory which is already known to be
794 * excluded and was previously left in check->path.
795 */
796 if (path->len &&
797 path->len <= namelen &&
798 !memcmp(name, path->buf, path->len) &&
799 (!name[path->len] || name[path->len] == '/'))
800 return check->exclude;
801
802 strbuf_setlen(path, 0);
803 for (i = 0; name[i]; i++) {
804 int ch = name[i];
805
806 if (ch == '/') {
807 int dt = DT_DIR;
808 exclude = last_exclude_matching(check->dir,
809 path->buf, &dt);
810 if (exclude) {
811 check->exclude = exclude;
812 return exclude;
813 }
814 }
815 strbuf_addch(path, ch);
816 }
817
818 /* An entry in the index; cannot be a directory with subentries */
819 strbuf_setlen(path, 0);
820
821 return last_exclude_matching(check->dir, name, dtype);
822 }
823
824 /*
825 * Is this name excluded? This is for a caller like show_files() that
826 * do not honor directory hierarchy and iterate through paths that are
827 * possibly in an ignored directory.
828 */
829 int is_path_excluded(struct path_exclude_check *check,
830 const char *name, int namelen, int *dtype)
831 {
832 struct exclude *exclude =
833 last_exclude_matching_path(check, name, namelen, dtype);
834 if (exclude)
835 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
836 return 0;
837 }
838
839 static struct dir_entry *dir_entry_new(const char *pathname, int len)
840 {
841 struct dir_entry *ent;
842
843 ent = xmalloc(sizeof(*ent) + len + 1);
844 ent->len = len;
845 memcpy(ent->name, pathname, len);
846 ent->name[len] = 0;
847 return ent;
848 }
849
850 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
851 {
852 if (cache_name_exists(pathname, len, ignore_case))
853 return NULL;
854
855 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
856 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
857 }
858
859 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
860 {
861 if (!cache_name_is_other(pathname, len))
862 return NULL;
863
864 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
865 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
866 }
867
868 enum exist_status {
869 index_nonexistent = 0,
870 index_directory,
871 index_gitdir
872 };
873
874 /*
875 * Do not use the alphabetically stored index to look up
876 * the directory name; instead, use the case insensitive
877 * name hash.
878 */
879 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
880 {
881 struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
882 unsigned char endchar;
883
884 if (!ce)
885 return index_nonexistent;
886 endchar = ce->name[len];
887
888 /*
889 * The cache_entry structure returned will contain this dirname
890 * and possibly additional path components.
891 */
892 if (endchar == '/')
893 return index_directory;
894
895 /*
896 * If there are no additional path components, then this cache_entry
897 * represents a submodule. Submodules, despite being directories,
898 * are stored in the cache without a closing slash.
899 */
900 if (!endchar && S_ISGITLINK(ce->ce_mode))
901 return index_gitdir;
902
903 /* This should never be hit, but it exists just in case. */
904 return index_nonexistent;
905 }
906
907 /*
908 * The index sorts alphabetically by entry name, which
909 * means that a gitlink sorts as '\0' at the end, while
910 * a directory (which is defined not as an entry, but as
911 * the files it contains) will sort with the '/' at the
912 * end.
913 */
914 static enum exist_status directory_exists_in_index(const char *dirname, int len)
915 {
916 int pos;
917
918 if (ignore_case)
919 return directory_exists_in_index_icase(dirname, len);
920
921 pos = cache_name_pos(dirname, len);
922 if (pos < 0)
923 pos = -pos-1;
924 while (pos < active_nr) {
925 struct cache_entry *ce = active_cache[pos++];
926 unsigned char endchar;
927
928 if (strncmp(ce->name, dirname, len))
929 break;
930 endchar = ce->name[len];
931 if (endchar > '/')
932 break;
933 if (endchar == '/')
934 return index_directory;
935 if (!endchar && S_ISGITLINK(ce->ce_mode))
936 return index_gitdir;
937 }
938 return index_nonexistent;
939 }
940
941 /*
942 * When we find a directory when traversing the filesystem, we
943 * have three distinct cases:
944 *
945 * - ignore it
946 * - see it as a directory
947 * - recurse into it
948 *
949 * and which one we choose depends on a combination of existing
950 * git index contents and the flags passed into the directory
951 * traversal routine.
952 *
953 * Case 1: If we *already* have entries in the index under that
954 * directory name, we always recurse into the directory to see
955 * all the files.
956 *
957 * Case 2: If we *already* have that directory name as a gitlink,
958 * we always continue to see it as a gitlink, regardless of whether
959 * there is an actual git directory there or not (it might not
960 * be checked out as a subproject!)
961 *
962 * Case 3: if we didn't have it in the index previously, we
963 * have a few sub-cases:
964 *
965 * (a) if "show_other_directories" is true, we show it as
966 * just a directory, unless "hide_empty_directories" is
967 * also true and the directory is empty, in which case
968 * we just ignore it entirely.
969 * (b) if it looks like a git directory, and we don't have
970 * 'no_gitlinks' set we treat it as a gitlink, and show it
971 * as a directory.
972 * (c) otherwise, we recurse into it.
973 */
974 enum directory_treatment {
975 show_directory,
976 ignore_directory,
977 recurse_into_directory
978 };
979
980 static enum directory_treatment treat_directory(struct dir_struct *dir,
981 const char *dirname, int len,
982 const struct path_simplify *simplify)
983 {
984 /* The "len-1" is to strip the final '/' */
985 switch (directory_exists_in_index(dirname, len-1)) {
986 case index_directory:
987 return recurse_into_directory;
988
989 case index_gitdir:
990 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
991 return ignore_directory;
992 return show_directory;
993
994 case index_nonexistent:
995 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
996 break;
997 if (!(dir->flags & DIR_NO_GITLINKS)) {
998 unsigned char sha1[20];
999 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1000 return show_directory;
1001 }
1002 return recurse_into_directory;
1003 }
1004
1005 /* This is the "show_other_directories" case */
1006 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1007 return show_directory;
1008 if (!read_directory_recursive(dir, dirname, len, 1, simplify))
1009 return ignore_directory;
1010 return show_directory;
1011 }
1012
1013 /*
1014 * This is an inexact early pruning of any recursive directory
1015 * reading - if the path cannot possibly be in the pathspec,
1016 * return true, and we'll skip it early.
1017 */
1018 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1019 {
1020 if (simplify) {
1021 for (;;) {
1022 const char *match = simplify->path;
1023 int len = simplify->len;
1024
1025 if (!match)
1026 break;
1027 if (len > pathlen)
1028 len = pathlen;
1029 if (!memcmp(path, match, len))
1030 return 0;
1031 simplify++;
1032 }
1033 return 1;
1034 }
1035 return 0;
1036 }
1037
1038 /*
1039 * This function tells us whether an excluded path matches a
1040 * list of "interesting" pathspecs. That is, whether a path matched
1041 * by any of the pathspecs could possibly be ignored by excluding
1042 * the specified path. This can happen if:
1043 *
1044 * 1. the path is mentioned explicitly in the pathspec
1045 *
1046 * 2. the path is a directory prefix of some element in the
1047 * pathspec
1048 */
1049 static int exclude_matches_pathspec(const char *path, int len,
1050 const struct path_simplify *simplify)
1051 {
1052 if (simplify) {
1053 for (; simplify->path; simplify++) {
1054 if (len == simplify->len
1055 && !memcmp(path, simplify->path, len))
1056 return 1;
1057 if (len < simplify->len
1058 && simplify->path[len] == '/'
1059 && !memcmp(path, simplify->path, len))
1060 return 1;
1061 }
1062 }
1063 return 0;
1064 }
1065
1066 static int get_index_dtype(const char *path, int len)
1067 {
1068 int pos;
1069 struct cache_entry *ce;
1070
1071 ce = cache_name_exists(path, len, 0);
1072 if (ce) {
1073 if (!ce_uptodate(ce))
1074 return DT_UNKNOWN;
1075 if (S_ISGITLINK(ce->ce_mode))
1076 return DT_DIR;
1077 /*
1078 * Nobody actually cares about the
1079 * difference between DT_LNK and DT_REG
1080 */
1081 return DT_REG;
1082 }
1083
1084 /* Try to look it up as a directory */
1085 pos = cache_name_pos(path, len);
1086 if (pos >= 0)
1087 return DT_UNKNOWN;
1088 pos = -pos-1;
1089 while (pos < active_nr) {
1090 ce = active_cache[pos++];
1091 if (strncmp(ce->name, path, len))
1092 break;
1093 if (ce->name[len] > '/')
1094 break;
1095 if (ce->name[len] < '/')
1096 continue;
1097 if (!ce_uptodate(ce))
1098 break; /* continue? */
1099 return DT_DIR;
1100 }
1101 return DT_UNKNOWN;
1102 }
1103
1104 static int get_dtype(struct dirent *de, const char *path, int len)
1105 {
1106 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1107 struct stat st;
1108
1109 if (dtype != DT_UNKNOWN)
1110 return dtype;
1111 dtype = get_index_dtype(path, len);
1112 if (dtype != DT_UNKNOWN)
1113 return dtype;
1114 if (lstat(path, &st))
1115 return dtype;
1116 if (S_ISREG(st.st_mode))
1117 return DT_REG;
1118 if (S_ISDIR(st.st_mode))
1119 return DT_DIR;
1120 if (S_ISLNK(st.st_mode))
1121 return DT_LNK;
1122 return dtype;
1123 }
1124
1125 enum path_treatment {
1126 path_ignored,
1127 path_handled,
1128 path_recurse
1129 };
1130
1131 static enum path_treatment treat_one_path(struct dir_struct *dir,
1132 struct strbuf *path,
1133 const struct path_simplify *simplify,
1134 int dtype, struct dirent *de)
1135 {
1136 int exclude = is_excluded(dir, path->buf, &dtype);
1137 if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
1138 && exclude_matches_pathspec(path->buf, path->len, simplify))
1139 dir_add_ignored(dir, path->buf, path->len);
1140
1141 /*
1142 * Excluded? If we don't explicitly want to show
1143 * ignored files, ignore it
1144 */
1145 if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
1146 return path_ignored;
1147
1148 if (dtype == DT_UNKNOWN)
1149 dtype = get_dtype(de, path->buf, path->len);
1150
1151 /*
1152 * Do we want to see just the ignored files?
1153 * We still need to recurse into directories,
1154 * even if we don't ignore them, since the
1155 * directory may contain files that we do..
1156 */
1157 if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
1158 if (dtype != DT_DIR)
1159 return path_ignored;
1160 }
1161
1162 switch (dtype) {
1163 default:
1164 return path_ignored;
1165 case DT_DIR:
1166 strbuf_addch(path, '/');
1167 switch (treat_directory(dir, path->buf, path->len, simplify)) {
1168 case show_directory:
1169 if (exclude != !!(dir->flags
1170 & DIR_SHOW_IGNORED))
1171 return path_ignored;
1172 break;
1173 case recurse_into_directory:
1174 return path_recurse;
1175 case ignore_directory:
1176 return path_ignored;
1177 }
1178 break;
1179 case DT_REG:
1180 case DT_LNK:
1181 break;
1182 }
1183 return path_handled;
1184 }
1185
1186 static enum path_treatment treat_path(struct dir_struct *dir,
1187 struct dirent *de,
1188 struct strbuf *path,
1189 int baselen,
1190 const struct path_simplify *simplify)
1191 {
1192 int dtype;
1193
1194 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1195 return path_ignored;
1196 strbuf_setlen(path, baselen);
1197 strbuf_addstr(path, de->d_name);
1198 if (simplify_away(path->buf, path->len, simplify))
1199 return path_ignored;
1200
1201 dtype = DTYPE(de);
1202 return treat_one_path(dir, path, simplify, dtype, de);
1203 }
1204
1205 /*
1206 * Read a directory tree. We currently ignore anything but
1207 * directories, regular files and symlinks. That's because git
1208 * doesn't handle them at all yet. Maybe that will change some
1209 * day.
1210 *
1211 * Also, we ignore the name ".git" (even if it is not a directory).
1212 * That likely will not change.
1213 */
1214 static int read_directory_recursive(struct dir_struct *dir,
1215 const char *base, int baselen,
1216 int check_only,
1217 const struct path_simplify *simplify)
1218 {
1219 DIR *fdir;
1220 int contents = 0;
1221 struct dirent *de;
1222 struct strbuf path = STRBUF_INIT;
1223
1224 strbuf_add(&path, base, baselen);
1225
1226 fdir = opendir(path.len ? path.buf : ".");
1227 if (!fdir)
1228 goto out;
1229
1230 while ((de = readdir(fdir)) != NULL) {
1231 switch (treat_path(dir, de, &path, baselen, simplify)) {
1232 case path_recurse:
1233 contents += read_directory_recursive(dir, path.buf,
1234 path.len, 0,
1235 simplify);
1236 continue;
1237 case path_ignored:
1238 continue;
1239 case path_handled:
1240 break;
1241 }
1242 contents++;
1243 if (check_only)
1244 break;
1245 dir_add_name(dir, path.buf, path.len);
1246 }
1247 closedir(fdir);
1248 out:
1249 strbuf_release(&path);
1250
1251 return contents;
1252 }
1253
1254 static int cmp_name(const void *p1, const void *p2)
1255 {
1256 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1257 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1258
1259 return cache_name_compare(e1->name, e1->len,
1260 e2->name, e2->len);
1261 }
1262
1263 static struct path_simplify *create_simplify(const char **pathspec)
1264 {
1265 int nr, alloc = 0;
1266 struct path_simplify *simplify = NULL;
1267
1268 if (!pathspec)
1269 return NULL;
1270
1271 for (nr = 0 ; ; nr++) {
1272 const char *match;
1273 if (nr >= alloc) {
1274 alloc = alloc_nr(alloc);
1275 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
1276 }
1277 match = *pathspec++;
1278 if (!match)
1279 break;
1280 simplify[nr].path = match;
1281 simplify[nr].len = simple_length(match);
1282 }
1283 simplify[nr].path = NULL;
1284 simplify[nr].len = 0;
1285 return simplify;
1286 }
1287
1288 static void free_simplify(struct path_simplify *simplify)
1289 {
1290 free(simplify);
1291 }
1292
1293 static int treat_leading_path(struct dir_struct *dir,
1294 const char *path, int len,
1295 const struct path_simplify *simplify)
1296 {
1297 struct strbuf sb = STRBUF_INIT;
1298 int baselen, rc = 0;
1299 const char *cp;
1300
1301 while (len && path[len - 1] == '/')
1302 len--;
1303 if (!len)
1304 return 1;
1305 baselen = 0;
1306 while (1) {
1307 cp = path + baselen + !!baselen;
1308 cp = memchr(cp, '/', path + len - cp);
1309 if (!cp)
1310 baselen = len;
1311 else
1312 baselen = cp - path;
1313 strbuf_setlen(&sb, 0);
1314 strbuf_add(&sb, path, baselen);
1315 if (!is_directory(sb.buf))
1316 break;
1317 if (simplify_away(sb.buf, sb.len, simplify))
1318 break;
1319 if (treat_one_path(dir, &sb, simplify,
1320 DT_DIR, NULL) == path_ignored)
1321 break; /* do not recurse into it */
1322 if (len <= baselen) {
1323 rc = 1;
1324 break; /* finished checking */
1325 }
1326 }
1327 strbuf_release(&sb);
1328 return rc;
1329 }
1330
1331 int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
1332 {
1333 struct path_simplify *simplify;
1334
1335 if (has_symlink_leading_path(path, len))
1336 return dir->nr;
1337
1338 simplify = create_simplify(pathspec);
1339 if (!len || treat_leading_path(dir, path, len, simplify))
1340 read_directory_recursive(dir, path, len, 0, simplify);
1341 free_simplify(simplify);
1342 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1343 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1344 return dir->nr;
1345 }
1346
1347 int file_exists(const char *f)
1348 {
1349 struct stat sb;
1350 return lstat(f, &sb) == 0;
1351 }
1352
1353 /*
1354 * Given two normalized paths (a trailing slash is ok), if subdir is
1355 * outside dir, return -1. Otherwise return the offset in subdir that
1356 * can be used as relative path to dir.
1357 */
1358 int dir_inside_of(const char *subdir, const char *dir)
1359 {
1360 int offset = 0;
1361
1362 assert(dir && subdir && *dir && *subdir);
1363
1364 while (*dir && *subdir && *dir == *subdir) {
1365 dir++;
1366 subdir++;
1367 offset++;
1368 }
1369
1370 /* hel[p]/me vs hel[l]/yeah */
1371 if (*dir && *subdir)
1372 return -1;
1373
1374 if (!*subdir)
1375 return !*dir ? offset : -1; /* same dir */
1376
1377 /* foo/[b]ar vs foo/[] */
1378 if (is_dir_sep(dir[-1]))
1379 return is_dir_sep(subdir[-1]) ? offset : -1;
1380
1381 /* foo[/]bar vs foo[] */
1382 return is_dir_sep(*subdir) ? offset + 1 : -1;
1383 }
1384
1385 int is_inside_dir(const char *dir)
1386 {
1387 char cwd[PATH_MAX];
1388 if (!dir)
1389 return 0;
1390 if (!getcwd(cwd, sizeof(cwd)))
1391 die_errno("can't find the current directory");
1392 return dir_inside_of(cwd, dir) >= 0;
1393 }
1394
1395 int is_empty_dir(const char *path)
1396 {
1397 DIR *dir = opendir(path);
1398 struct dirent *e;
1399 int ret = 1;
1400
1401 if (!dir)
1402 return 0;
1403
1404 while ((e = readdir(dir)) != NULL)
1405 if (!is_dot_or_dotdot(e->d_name)) {
1406 ret = 0;
1407 break;
1408 }
1409
1410 closedir(dir);
1411 return ret;
1412 }
1413
1414 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1415 {
1416 DIR *dir;
1417 struct dirent *e;
1418 int ret = 0, original_len = path->len, len, kept_down = 0;
1419 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1420 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1421 unsigned char submodule_head[20];
1422
1423 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1424 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1425 /* Do not descend and nuke a nested git work tree. */
1426 if (kept_up)
1427 *kept_up = 1;
1428 return 0;
1429 }
1430
1431 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1432 dir = opendir(path->buf);
1433 if (!dir) {
1434 /* an empty dir could be removed even if it is unreadble */
1435 if (!keep_toplevel)
1436 return rmdir(path->buf);
1437 else
1438 return -1;
1439 }
1440 if (path->buf[original_len - 1] != '/')
1441 strbuf_addch(path, '/');
1442
1443 len = path->len;
1444 while ((e = readdir(dir)) != NULL) {
1445 struct stat st;
1446 if (is_dot_or_dotdot(e->d_name))
1447 continue;
1448
1449 strbuf_setlen(path, len);
1450 strbuf_addstr(path, e->d_name);
1451 if (lstat(path->buf, &st))
1452 ; /* fall thru */
1453 else if (S_ISDIR(st.st_mode)) {
1454 if (!remove_dir_recurse(path, flag, &kept_down))
1455 continue; /* happy */
1456 } else if (!only_empty && !unlink(path->buf))
1457 continue; /* happy, too */
1458
1459 /* path too long, stat fails, or non-directory still exists */
1460 ret = -1;
1461 break;
1462 }
1463 closedir(dir);
1464
1465 strbuf_setlen(path, original_len);
1466 if (!ret && !keep_toplevel && !kept_down)
1467 ret = rmdir(path->buf);
1468 else if (kept_up)
1469 /*
1470 * report the uplevel that it is not an error that we
1471 * did not rmdir() our directory.
1472 */
1473 *kept_up = !ret;
1474 return ret;
1475 }
1476
1477 int remove_dir_recursively(struct strbuf *path, int flag)
1478 {
1479 return remove_dir_recurse(path, flag, NULL);
1480 }
1481
1482 void setup_standard_excludes(struct dir_struct *dir)
1483 {
1484 const char *path;
1485
1486 dir->exclude_per_dir = ".gitignore";
1487 path = git_path("info/exclude");
1488 if (!access(path, R_OK))
1489 add_excludes_from_file(dir, path);
1490 if (excludes_file && !access(excludes_file, R_OK))
1491 add_excludes_from_file(dir, excludes_file);
1492 }
1493
1494 int remove_path(const char *name)
1495 {
1496 char *slash;
1497
1498 if (unlink(name) && errno != ENOENT)
1499 return -1;
1500
1501 slash = strrchr(name, '/');
1502 if (slash) {
1503 char *dirs = xstrdup(name);
1504 slash = dirs + (slash - name);
1505 do {
1506 *slash = '\0';
1507 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1508 free(dirs);
1509 }
1510 return 0;
1511 }
1512
1513 static int pathspec_item_cmp(const void *a_, const void *b_)
1514 {
1515 struct pathspec_item *a, *b;
1516
1517 a = (struct pathspec_item *)a_;
1518 b = (struct pathspec_item *)b_;
1519 return strcmp(a->match, b->match);
1520 }
1521
1522 int init_pathspec(struct pathspec *pathspec, const char **paths)
1523 {
1524 const char **p = paths;
1525 int i;
1526
1527 memset(pathspec, 0, sizeof(*pathspec));
1528 if (!p)
1529 return 0;
1530 while (*p)
1531 p++;
1532 pathspec->raw = paths;
1533 pathspec->nr = p - paths;
1534 if (!pathspec->nr)
1535 return 0;
1536
1537 pathspec->items = xmalloc(sizeof(struct pathspec_item)*pathspec->nr);
1538 for (i = 0; i < pathspec->nr; i++) {
1539 struct pathspec_item *item = pathspec->items+i;
1540 const char *path = paths[i];
1541
1542 item->match = path;
1543 item->len = strlen(path);
1544 item->use_wildcard = !no_wildcard(path);
1545 if (item->use_wildcard)
1546 pathspec->has_wildcard = 1;
1547 }
1548
1549 qsort(pathspec->items, pathspec->nr,
1550 sizeof(struct pathspec_item), pathspec_item_cmp);
1551
1552 return 0;
1553 }
1554
1555 void free_pathspec(struct pathspec *pathspec)
1556 {
1557 free(pathspec->items);
1558 pathspec->items = NULL;
1559 }
1560
1561 /*
1562 * Frees memory within dir which was allocated for exclude lists and
1563 * the exclude_stack. Does not free dir itself.
1564 */
1565 void clear_directory(struct dir_struct *dir)
1566 {
1567 int i, j;
1568 struct exclude_list_group *group;
1569 struct exclude_list *el;
1570 struct exclude_stack *stk;
1571
1572 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1573 group = &dir->exclude_list_group[i];
1574 for (j = 0; j < group->nr; j++) {
1575 el = &group->el[j];
1576 if (i == EXC_DIRS)
1577 free((char *)el->src);
1578 clear_exclude_list(el);
1579 }
1580 free(group->el);
1581 }
1582
1583 stk = dir->exclude_stack;
1584 while (stk) {
1585 struct exclude_stack *prev = stk->prev;
1586 free(stk);
1587 stk = prev;
1588 }
1589 }