Merge branch 'master' into next
[git/mjg.git] / dir.c
blob0e19b203f2b7a2d8576a1876492314f66f7c642e
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
7 */
8 #include "cache.h"
9 #include "dir.h"
10 #include "refs.h"
11 #include "wildmatch.h"
13 struct path_simplify {
14 int len;
15 const char *path;
18 static int read_directory_recursive(struct dir_struct *dir, const char *path, int len,
19 int check_only, const struct path_simplify *simplify);
20 static int get_dtype(struct dirent *de, const char *path, int len);
22 /* helper string functions with support for the ignore_case flag */
23 int strcmp_icase(const char *a, const char *b)
25 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
28 int strncmp_icase(const char *a, const char *b, size_t count)
30 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
33 int fnmatch_icase(const char *pattern, const char *string, int flags)
35 return fnmatch(pattern, string, flags | (ignore_case ? FNM_CASEFOLD : 0));
38 static size_t common_prefix_len(const char **pathspec)
40 const char *n, *first;
41 size_t max = 0;
43 if (!pathspec)
44 return max;
46 first = *pathspec;
47 while ((n = *pathspec++)) {
48 size_t i, len = 0;
49 for (i = 0; first == n || i < max; i++) {
50 char c = n[i];
51 if (!c || c != first[i] || is_glob_special(c))
52 break;
53 if (c == '/')
54 len = i + 1;
56 if (first == n || len < max) {
57 max = len;
58 if (!max)
59 break;
62 return max;
66 * Returns a copy of the longest leading path common among all
67 * pathspecs.
69 char *common_prefix(const char **pathspec)
71 unsigned long len = common_prefix_len(pathspec);
73 return len ? xmemdupz(*pathspec, len) : NULL;
76 int fill_directory(struct dir_struct *dir, const char **pathspec)
78 size_t len;
81 * Calculate common prefix for the pathspec, and
82 * use that to optimize the directory walk
84 len = common_prefix_len(pathspec);
86 /* Read the directory and prune it */
87 read_directory(dir, pathspec ? *pathspec : "", len, pathspec);
88 return len;
91 int within_depth(const char *name, int namelen,
92 int depth, int max_depth)
94 const char *cp = name, *cpe = name + namelen;
96 while (cp < cpe) {
97 if (*cp++ != '/')
98 continue;
99 depth++;
100 if (depth > max_depth)
101 return 0;
103 return 1;
107 * Does 'match' match the given name?
108 * A match is found if
110 * (1) the 'match' string is leading directory of 'name', or
111 * (2) the 'match' string is a wildcard and matches 'name', or
112 * (3) the 'match' string is exactly the same as 'name'.
114 * and the return value tells which case it was.
116 * It returns 0 when there is no match.
118 static int match_one(const char *match, const char *name, int namelen)
120 int matchlen;
122 /* If the match was just the prefix, we matched */
123 if (!*match)
124 return MATCHED_RECURSIVELY;
126 if (ignore_case) {
127 for (;;) {
128 unsigned char c1 = tolower(*match);
129 unsigned char c2 = tolower(*name);
130 if (c1 == '\0' || is_glob_special(c1))
131 break;
132 if (c1 != c2)
133 return 0;
134 match++;
135 name++;
136 namelen--;
138 } else {
139 for (;;) {
140 unsigned char c1 = *match;
141 unsigned char c2 = *name;
142 if (c1 == '\0' || is_glob_special(c1))
143 break;
144 if (c1 != c2)
145 return 0;
146 match++;
147 name++;
148 namelen--;
154 * If we don't match the matchstring exactly,
155 * we need to match by fnmatch
157 matchlen = strlen(match);
158 if (strncmp_icase(match, name, matchlen))
159 return !fnmatch_icase(match, name, 0) ? MATCHED_FNMATCH : 0;
161 if (namelen == matchlen)
162 return MATCHED_EXACTLY;
163 if (match[matchlen-1] == '/' || name[matchlen] == '/')
164 return MATCHED_RECURSIVELY;
165 return 0;
169 * Given a name and a list of pathspecs, see if the name matches
170 * any of the pathspecs. The caller is also interested in seeing
171 * all pathspec matches some names it calls this function with
172 * (otherwise the user could have mistyped the unmatched pathspec),
173 * and a mark is left in seen[] array for pathspec element that
174 * actually matched anything.
176 int match_pathspec(const char **pathspec, const char *name, int namelen,
177 int prefix, char *seen)
179 int i, retval = 0;
181 if (!pathspec)
182 return 1;
184 name += prefix;
185 namelen -= prefix;
187 for (i = 0; pathspec[i] != NULL; i++) {
188 int how;
189 const char *match = pathspec[i] + prefix;
190 if (seen && seen[i] == MATCHED_EXACTLY)
191 continue;
192 how = match_one(match, name, namelen);
193 if (how) {
194 if (retval < how)
195 retval = how;
196 if (seen && seen[i] < how)
197 seen[i] = how;
200 return retval;
204 * Does 'match' match the given name?
205 * A match is found if
207 * (1) the 'match' string is leading directory of 'name', or
208 * (2) the 'match' string is a wildcard and matches 'name', or
209 * (3) the 'match' string is exactly the same as 'name'.
211 * and the return value tells which case it was.
213 * It returns 0 when there is no match.
215 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
216 const char *name, int namelen)
218 /* name/namelen has prefix cut off by caller */
219 const char *match = item->match + prefix;
220 int matchlen = item->len - prefix;
222 /* If the match was just the prefix, we matched */
223 if (!*match)
224 return MATCHED_RECURSIVELY;
226 if (matchlen <= namelen && !strncmp(match, name, matchlen)) {
227 if (matchlen == namelen)
228 return MATCHED_EXACTLY;
230 if (match[matchlen-1] == '/' || name[matchlen] == '/')
231 return MATCHED_RECURSIVELY;
234 if (item->use_wildcard && !fnmatch(match, name, 0))
235 return MATCHED_FNMATCH;
237 return 0;
241 * Given a name and a list of pathspecs, see if the name matches
242 * any of the pathspecs. The caller is also interested in seeing
243 * all pathspec matches some names it calls this function with
244 * (otherwise the user could have mistyped the unmatched pathspec),
245 * and a mark is left in seen[] array for pathspec element that
246 * actually matched anything.
248 int match_pathspec_depth(const struct pathspec *ps,
249 const char *name, int namelen,
250 int prefix, char *seen)
252 int i, retval = 0;
254 if (!ps->nr) {
255 if (!ps->recursive || ps->max_depth == -1)
256 return MATCHED_RECURSIVELY;
258 if (within_depth(name, namelen, 0, ps->max_depth))
259 return MATCHED_EXACTLY;
260 else
261 return 0;
264 name += prefix;
265 namelen -= prefix;
267 for (i = ps->nr - 1; i >= 0; i--) {
268 int how;
269 if (seen && seen[i] == MATCHED_EXACTLY)
270 continue;
271 how = match_pathspec_item(ps->items+i, prefix, name, namelen);
272 if (ps->recursive && ps->max_depth != -1 &&
273 how && how != MATCHED_FNMATCH) {
274 int len = ps->items[i].len;
275 if (name[len] == '/')
276 len++;
277 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
278 how = MATCHED_EXACTLY;
279 else
280 how = 0;
282 if (how) {
283 if (retval < how)
284 retval = how;
285 if (seen && seen[i] < how)
286 seen[i] = how;
289 return retval;
293 * Return the length of the "simple" part of a path match limiter.
295 static int simple_length(const char *match)
297 int len = -1;
299 for (;;) {
300 unsigned char c = *match++;
301 len++;
302 if (c == '\0' || is_glob_special(c))
303 return len;
307 static int no_wildcard(const char *string)
309 return string[simple_length(string)] == '\0';
312 void parse_exclude_pattern(const char **pattern,
313 int *patternlen,
314 int *flags,
315 int *nowildcardlen)
317 const char *p = *pattern;
318 size_t i, len;
320 *flags = 0;
321 if (*p == '!') {
322 *flags |= EXC_FLAG_NEGATIVE;
323 p++;
325 len = strlen(p);
326 if (len && p[len - 1] == '/') {
327 len--;
328 *flags |= EXC_FLAG_MUSTBEDIR;
330 for (i = 0; i < len; i++) {
331 if (p[i] == '/')
332 break;
334 if (i == len)
335 *flags |= EXC_FLAG_NODIR;
336 *nowildcardlen = simple_length(p);
338 * we should have excluded the trailing slash from 'p' too,
339 * but that's one more allocation. Instead just make sure
340 * nowildcardlen does not exceed real patternlen
342 if (*nowildcardlen > len)
343 *nowildcardlen = len;
344 if (*p == '*' && no_wildcard(p + 1))
345 *flags |= EXC_FLAG_ENDSWITH;
346 *pattern = p;
347 *patternlen = len;
350 void add_exclude(const char *string, const char *base,
351 int baselen, struct exclude_list *which)
353 struct exclude *x;
354 int patternlen;
355 int flags;
356 int nowildcardlen;
358 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
359 if (flags & EXC_FLAG_MUSTBEDIR) {
360 char *s;
361 x = xmalloc(sizeof(*x) + patternlen + 1);
362 s = (char *)(x+1);
363 memcpy(s, string, patternlen);
364 s[patternlen] = '\0';
365 x->pattern = s;
366 } else {
367 x = xmalloc(sizeof(*x));
368 x->pattern = string;
370 x->patternlen = patternlen;
371 x->nowildcardlen = nowildcardlen;
372 x->base = base;
373 x->baselen = baselen;
374 x->flags = flags;
375 ALLOC_GROW(which->excludes, which->nr + 1, which->alloc);
376 which->excludes[which->nr++] = x;
379 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
381 int pos, len;
382 unsigned long sz;
383 enum object_type type;
384 void *data;
385 struct index_state *istate = &the_index;
387 len = strlen(path);
388 pos = index_name_pos(istate, path, len);
389 if (pos < 0)
390 return NULL;
391 if (!ce_skip_worktree(istate->cache[pos]))
392 return NULL;
393 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
394 if (!data || type != OBJ_BLOB) {
395 free(data);
396 return NULL;
398 *size = xsize_t(sz);
399 return data;
402 void free_excludes(struct exclude_list *el)
404 int i;
406 for (i = 0; i < el->nr; i++)
407 free(el->excludes[i]);
408 free(el->excludes);
410 el->nr = 0;
411 el->excludes = NULL;
414 int add_excludes_from_file_to_list(const char *fname,
415 const char *base,
416 int baselen,
417 char **buf_p,
418 struct exclude_list *which,
419 int check_index)
421 struct stat st;
422 int fd, i;
423 size_t size = 0;
424 char *buf, *entry;
426 fd = open(fname, O_RDONLY);
427 if (fd < 0 || fstat(fd, &st) < 0) {
428 if (errno != ENOENT)
429 warn_on_inaccessible(fname);
430 if (0 <= fd)
431 close(fd);
432 if (!check_index ||
433 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
434 return -1;
435 if (size == 0) {
436 free(buf);
437 return 0;
439 if (buf[size-1] != '\n') {
440 buf = xrealloc(buf, size+1);
441 buf[size++] = '\n';
444 else {
445 size = xsize_t(st.st_size);
446 if (size == 0) {
447 close(fd);
448 return 0;
450 buf = xmalloc(size+1);
451 if (read_in_full(fd, buf, size) != size) {
452 free(buf);
453 close(fd);
454 return -1;
456 buf[size++] = '\n';
457 close(fd);
460 if (buf_p)
461 *buf_p = buf;
462 entry = buf;
463 for (i = 0; i < size; i++) {
464 if (buf[i] == '\n') {
465 if (entry != buf + i && entry[0] != '#') {
466 buf[i - (i && buf[i-1] == '\r')] = 0;
467 add_exclude(entry, base, baselen, which);
469 entry = buf + i + 1;
472 return 0;
475 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
477 if (add_excludes_from_file_to_list(fname, "", 0, NULL,
478 &dir->exclude_list[EXC_FILE], 0) < 0)
479 die("cannot use %s as an exclude file", fname);
482 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
484 struct exclude_list *el;
485 struct exclude_stack *stk = NULL;
486 int current;
488 if ((!dir->exclude_per_dir) ||
489 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
490 return; /* too long a path -- ignore */
492 /* Pop the ones that are not the prefix of the path being checked. */
493 el = &dir->exclude_list[EXC_DIRS];
494 while ((stk = dir->exclude_stack) != NULL) {
495 if (stk->baselen <= baselen &&
496 !strncmp(dir->basebuf, base, stk->baselen))
497 break;
498 dir->exclude_stack = stk->prev;
499 while (stk->exclude_ix < el->nr)
500 free(el->excludes[--el->nr]);
501 free(stk->filebuf);
502 free(stk);
505 /* Read from the parent directories and push them down. */
506 current = stk ? stk->baselen : -1;
507 while (current < baselen) {
508 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
509 const char *cp;
511 if (current < 0) {
512 cp = base;
513 current = 0;
515 else {
516 cp = strchr(base + current + 1, '/');
517 if (!cp)
518 die("oops in prep_exclude");
519 cp++;
521 stk->prev = dir->exclude_stack;
522 stk->baselen = cp - base;
523 stk->exclude_ix = el->nr;
524 memcpy(dir->basebuf + current, base + current,
525 stk->baselen - current);
526 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
527 add_excludes_from_file_to_list(dir->basebuf,
528 dir->basebuf, stk->baselen,
529 &stk->filebuf, el, 1);
530 dir->exclude_stack = stk;
531 current = stk->baselen;
533 dir->basebuf[baselen] = '\0';
536 int match_basename(const char *basename, int basenamelen,
537 const char *pattern, int prefix, int patternlen,
538 int flags)
540 if (prefix == patternlen) {
541 if (!strcmp_icase(pattern, basename))
542 return 1;
543 } else if (flags & EXC_FLAG_ENDSWITH) {
544 if (patternlen - 1 <= basenamelen &&
545 !strcmp_icase(pattern + 1,
546 basename + basenamelen - patternlen + 1))
547 return 1;
548 } else {
549 if (fnmatch_icase(pattern, basename, 0) == 0)
550 return 1;
552 return 0;
555 int match_pathname(const char *pathname, int pathlen,
556 const char *base, int baselen,
557 const char *pattern, int prefix, int patternlen,
558 int flags)
560 const char *name;
561 int namelen;
564 * match with FNM_PATHNAME; the pattern has base implicitly
565 * in front of it.
567 if (*pattern == '/') {
568 pattern++;
569 prefix--;
573 * baselen does not count the trailing slash. base[] may or
574 * may not end with a trailing slash though.
576 if (pathlen < baselen + 1 ||
577 (baselen && pathname[baselen] != '/') ||
578 strncmp_icase(pathname, base, baselen))
579 return 0;
581 namelen = baselen ? pathlen - baselen - 1 : pathlen;
582 name = pathname + pathlen - namelen;
584 if (prefix) {
586 * if the non-wildcard part is longer than the
587 * remaining pathname, surely it cannot match.
589 if (prefix > namelen)
590 return 0;
592 if (strncmp_icase(pattern, name, prefix))
593 return 0;
594 pattern += prefix;
595 name += prefix;
596 namelen -= prefix;
599 return wildmatch(pattern, name,
600 ignore_case ? FNM_CASEFOLD : 0) == 0;
603 /* Scan the list and let the last match determine the fate.
604 * Return 1 for exclude, 0 for include and -1 for undecided.
606 int excluded_from_list(const char *pathname,
607 int pathlen, const char *basename, int *dtype,
608 struct exclude_list *el)
610 int i;
612 if (!el->nr)
613 return -1; /* undefined */
615 for (i = el->nr - 1; 0 <= i; i--) {
616 struct exclude *x = el->excludes[i];
617 const char *exclude = x->pattern;
618 int to_exclude = x->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
619 int prefix = x->nowildcardlen;
621 if (x->flags & EXC_FLAG_MUSTBEDIR) {
622 if (*dtype == DT_UNKNOWN)
623 *dtype = get_dtype(NULL, pathname, pathlen);
624 if (*dtype != DT_DIR)
625 continue;
628 if (x->flags & EXC_FLAG_NODIR) {
629 if (match_basename(basename,
630 pathlen - (basename - pathname),
631 exclude, prefix, x->patternlen,
632 x->flags))
633 return to_exclude;
634 continue;
637 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
638 if (match_pathname(pathname, pathlen,
639 x->base, x->baselen ? x->baselen - 1 : 0,
640 exclude, prefix, x->patternlen, x->flags))
641 return to_exclude;
643 return -1; /* undecided */
646 static int excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
648 int pathlen = strlen(pathname);
649 int st;
650 const char *basename = strrchr(pathname, '/');
651 basename = (basename) ? basename+1 : pathname;
653 prep_exclude(dir, pathname, basename-pathname);
654 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
655 switch (excluded_from_list(pathname, pathlen, basename,
656 dtype_p, &dir->exclude_list[st])) {
657 case 0:
658 return 0;
659 case 1:
660 return 1;
663 return 0;
666 void path_exclude_check_init(struct path_exclude_check *check,
667 struct dir_struct *dir)
669 check->dir = dir;
670 strbuf_init(&check->path, 256);
673 void path_exclude_check_clear(struct path_exclude_check *check)
675 strbuf_release(&check->path);
679 * Is this name excluded? This is for a caller like show_files() that
680 * do not honor directory hierarchy and iterate through paths that are
681 * possibly in an ignored directory.
683 * A path to a directory known to be excluded is left in check->path to
684 * optimize for repeated checks for files in the same excluded directory.
686 int path_excluded(struct path_exclude_check *check,
687 const char *name, int namelen, int *dtype)
689 int i;
690 struct strbuf *path = &check->path;
693 * we allow the caller to pass namelen as an optimization; it
694 * must match the length of the name, as we eventually call
695 * excluded() on the whole name string.
697 if (namelen < 0)
698 namelen = strlen(name);
700 if (path->len &&
701 path->len <= namelen &&
702 !memcmp(name, path->buf, path->len) &&
703 (!name[path->len] || name[path->len] == '/'))
704 return 1;
706 strbuf_setlen(path, 0);
707 for (i = 0; name[i]; i++) {
708 int ch = name[i];
710 if (ch == '/') {
711 int dt = DT_DIR;
712 if (excluded(check->dir, path->buf, &dt))
713 return 1;
715 strbuf_addch(path, ch);
718 /* An entry in the index; cannot be a directory with subentries */
719 strbuf_setlen(path, 0);
721 return excluded(check->dir, name, dtype);
724 static struct dir_entry *dir_entry_new(const char *pathname, int len)
726 struct dir_entry *ent;
728 ent = xmalloc(sizeof(*ent) + len + 1);
729 ent->len = len;
730 memcpy(ent->name, pathname, len);
731 ent->name[len] = 0;
732 return ent;
735 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
737 if (cache_name_exists(pathname, len, ignore_case))
738 return NULL;
740 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
741 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
744 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
746 if (!cache_name_is_other(pathname, len))
747 return NULL;
749 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
750 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
753 enum exist_status {
754 index_nonexistent = 0,
755 index_directory,
756 index_gitdir
760 * Do not use the alphabetically stored index to look up
761 * the directory name; instead, use the case insensitive
762 * name hash.
764 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
766 struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
767 unsigned char endchar;
769 if (!ce)
770 return index_nonexistent;
771 endchar = ce->name[len];
774 * The cache_entry structure returned will contain this dirname
775 * and possibly additional path components.
777 if (endchar == '/')
778 return index_directory;
781 * If there are no additional path components, then this cache_entry
782 * represents a submodule. Submodules, despite being directories,
783 * are stored in the cache without a closing slash.
785 if (!endchar && S_ISGITLINK(ce->ce_mode))
786 return index_gitdir;
788 /* This should never be hit, but it exists just in case. */
789 return index_nonexistent;
793 * The index sorts alphabetically by entry name, which
794 * means that a gitlink sorts as '\0' at the end, while
795 * a directory (which is defined not as an entry, but as
796 * the files it contains) will sort with the '/' at the
797 * end.
799 static enum exist_status directory_exists_in_index(const char *dirname, int len)
801 int pos;
803 if (ignore_case)
804 return directory_exists_in_index_icase(dirname, len);
806 pos = cache_name_pos(dirname, len);
807 if (pos < 0)
808 pos = -pos-1;
809 while (pos < active_nr) {
810 struct cache_entry *ce = active_cache[pos++];
811 unsigned char endchar;
813 if (strncmp(ce->name, dirname, len))
814 break;
815 endchar = ce->name[len];
816 if (endchar > '/')
817 break;
818 if (endchar == '/')
819 return index_directory;
820 if (!endchar && S_ISGITLINK(ce->ce_mode))
821 return index_gitdir;
823 return index_nonexistent;
827 * When we find a directory when traversing the filesystem, we
828 * have three distinct cases:
830 * - ignore it
831 * - see it as a directory
832 * - recurse into it
834 * and which one we choose depends on a combination of existing
835 * git index contents and the flags passed into the directory
836 * traversal routine.
838 * Case 1: If we *already* have entries in the index under that
839 * directory name, we always recurse into the directory to see
840 * all the files.
842 * Case 2: If we *already* have that directory name as a gitlink,
843 * we always continue to see it as a gitlink, regardless of whether
844 * there is an actual git directory there or not (it might not
845 * be checked out as a subproject!)
847 * Case 3: if we didn't have it in the index previously, we
848 * have a few sub-cases:
850 * (a) if "show_other_directories" is true, we show it as
851 * just a directory, unless "hide_empty_directories" is
852 * also true and the directory is empty, in which case
853 * we just ignore it entirely.
854 * (b) if it looks like a git directory, and we don't have
855 * 'no_gitlinks' set we treat it as a gitlink, and show it
856 * as a directory.
857 * (c) otherwise, we recurse into it.
859 enum directory_treatment {
860 show_directory,
861 ignore_directory,
862 recurse_into_directory
865 static enum directory_treatment treat_directory(struct dir_struct *dir,
866 const char *dirname, int len,
867 const struct path_simplify *simplify)
869 /* The "len-1" is to strip the final '/' */
870 switch (directory_exists_in_index(dirname, len-1)) {
871 case index_directory:
872 return recurse_into_directory;
874 case index_gitdir:
875 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
876 return ignore_directory;
877 return show_directory;
879 case index_nonexistent:
880 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
881 break;
882 if (!(dir->flags & DIR_NO_GITLINKS)) {
883 unsigned char sha1[20];
884 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
885 return show_directory;
887 return recurse_into_directory;
890 /* This is the "show_other_directories" case */
891 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
892 return show_directory;
893 if (!read_directory_recursive(dir, dirname, len, 1, simplify))
894 return ignore_directory;
895 return show_directory;
899 * This is an inexact early pruning of any recursive directory
900 * reading - if the path cannot possibly be in the pathspec,
901 * return true, and we'll skip it early.
903 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
905 if (simplify) {
906 for (;;) {
907 const char *match = simplify->path;
908 int len = simplify->len;
910 if (!match)
911 break;
912 if (len > pathlen)
913 len = pathlen;
914 if (!memcmp(path, match, len))
915 return 0;
916 simplify++;
918 return 1;
920 return 0;
924 * This function tells us whether an excluded path matches a
925 * list of "interesting" pathspecs. That is, whether a path matched
926 * by any of the pathspecs could possibly be ignored by excluding
927 * the specified path. This can happen if:
929 * 1. the path is mentioned explicitly in the pathspec
931 * 2. the path is a directory prefix of some element in the
932 * pathspec
934 static int exclude_matches_pathspec(const char *path, int len,
935 const struct path_simplify *simplify)
937 if (simplify) {
938 for (; simplify->path; simplify++) {
939 if (len == simplify->len
940 && !memcmp(path, simplify->path, len))
941 return 1;
942 if (len < simplify->len
943 && simplify->path[len] == '/'
944 && !memcmp(path, simplify->path, len))
945 return 1;
948 return 0;
951 static int get_index_dtype(const char *path, int len)
953 int pos;
954 struct cache_entry *ce;
956 ce = cache_name_exists(path, len, 0);
957 if (ce) {
958 if (!ce_uptodate(ce))
959 return DT_UNKNOWN;
960 if (S_ISGITLINK(ce->ce_mode))
961 return DT_DIR;
963 * Nobody actually cares about the
964 * difference between DT_LNK and DT_REG
966 return DT_REG;
969 /* Try to look it up as a directory */
970 pos = cache_name_pos(path, len);
971 if (pos >= 0)
972 return DT_UNKNOWN;
973 pos = -pos-1;
974 while (pos < active_nr) {
975 ce = active_cache[pos++];
976 if (strncmp(ce->name, path, len))
977 break;
978 if (ce->name[len] > '/')
979 break;
980 if (ce->name[len] < '/')
981 continue;
982 if (!ce_uptodate(ce))
983 break; /* continue? */
984 return DT_DIR;
986 return DT_UNKNOWN;
989 static int get_dtype(struct dirent *de, const char *path, int len)
991 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
992 struct stat st;
994 if (dtype != DT_UNKNOWN)
995 return dtype;
996 dtype = get_index_dtype(path, len);
997 if (dtype != DT_UNKNOWN)
998 return dtype;
999 if (lstat(path, &st))
1000 return dtype;
1001 if (S_ISREG(st.st_mode))
1002 return DT_REG;
1003 if (S_ISDIR(st.st_mode))
1004 return DT_DIR;
1005 if (S_ISLNK(st.st_mode))
1006 return DT_LNK;
1007 return dtype;
1010 enum path_treatment {
1011 path_ignored,
1012 path_handled,
1013 path_recurse
1016 static enum path_treatment treat_one_path(struct dir_struct *dir,
1017 struct strbuf *path,
1018 const struct path_simplify *simplify,
1019 int dtype, struct dirent *de)
1021 int exclude = excluded(dir, path->buf, &dtype);
1022 if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
1023 && exclude_matches_pathspec(path->buf, path->len, simplify))
1024 dir_add_ignored(dir, path->buf, path->len);
1027 * Excluded? If we don't explicitly want to show
1028 * ignored files, ignore it
1030 if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
1031 return path_ignored;
1033 if (dtype == DT_UNKNOWN)
1034 dtype = get_dtype(de, path->buf, path->len);
1037 * Do we want to see just the ignored files?
1038 * We still need to recurse into directories,
1039 * even if we don't ignore them, since the
1040 * directory may contain files that we do..
1042 if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
1043 if (dtype != DT_DIR)
1044 return path_ignored;
1047 switch (dtype) {
1048 default:
1049 return path_ignored;
1050 case DT_DIR:
1051 strbuf_addch(path, '/');
1052 switch (treat_directory(dir, path->buf, path->len, simplify)) {
1053 case show_directory:
1054 if (exclude != !!(dir->flags
1055 & DIR_SHOW_IGNORED))
1056 return path_ignored;
1057 break;
1058 case recurse_into_directory:
1059 return path_recurse;
1060 case ignore_directory:
1061 return path_ignored;
1063 break;
1064 case DT_REG:
1065 case DT_LNK:
1066 break;
1068 return path_handled;
1071 static enum path_treatment treat_path(struct dir_struct *dir,
1072 struct dirent *de,
1073 struct strbuf *path,
1074 int baselen,
1075 const struct path_simplify *simplify)
1077 int dtype;
1079 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1080 return path_ignored;
1081 strbuf_setlen(path, baselen);
1082 strbuf_addstr(path, de->d_name);
1083 if (simplify_away(path->buf, path->len, simplify))
1084 return path_ignored;
1086 dtype = DTYPE(de);
1087 return treat_one_path(dir, path, simplify, dtype, de);
1091 * Read a directory tree. We currently ignore anything but
1092 * directories, regular files and symlinks. That's because git
1093 * doesn't handle them at all yet. Maybe that will change some
1094 * day.
1096 * Also, we ignore the name ".git" (even if it is not a directory).
1097 * That likely will not change.
1099 static int read_directory_recursive(struct dir_struct *dir,
1100 const char *base, int baselen,
1101 int check_only,
1102 const struct path_simplify *simplify)
1104 DIR *fdir;
1105 int contents = 0;
1106 struct dirent *de;
1107 struct strbuf path = STRBUF_INIT;
1109 strbuf_add(&path, base, baselen);
1111 fdir = opendir(path.len ? path.buf : ".");
1112 if (!fdir)
1113 goto out;
1115 while ((de = readdir(fdir)) != NULL) {
1116 switch (treat_path(dir, de, &path, baselen, simplify)) {
1117 case path_recurse:
1118 contents += read_directory_recursive(dir, path.buf,
1119 path.len, 0,
1120 simplify);
1121 continue;
1122 case path_ignored:
1123 continue;
1124 case path_handled:
1125 break;
1127 contents++;
1128 if (check_only)
1129 break;
1130 dir_add_name(dir, path.buf, path.len);
1132 closedir(fdir);
1133 out:
1134 strbuf_release(&path);
1136 return contents;
1139 static int cmp_name(const void *p1, const void *p2)
1141 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1142 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1144 return cache_name_compare(e1->name, e1->len,
1145 e2->name, e2->len);
1148 static struct path_simplify *create_simplify(const char **pathspec)
1150 int nr, alloc = 0;
1151 struct path_simplify *simplify = NULL;
1153 if (!pathspec)
1154 return NULL;
1156 for (nr = 0 ; ; nr++) {
1157 const char *match;
1158 if (nr >= alloc) {
1159 alloc = alloc_nr(alloc);
1160 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
1162 match = *pathspec++;
1163 if (!match)
1164 break;
1165 simplify[nr].path = match;
1166 simplify[nr].len = simple_length(match);
1168 simplify[nr].path = NULL;
1169 simplify[nr].len = 0;
1170 return simplify;
1173 static void free_simplify(struct path_simplify *simplify)
1175 free(simplify);
1178 static int treat_leading_path(struct dir_struct *dir,
1179 const char *path, int len,
1180 const struct path_simplify *simplify)
1182 struct strbuf sb = STRBUF_INIT;
1183 int baselen, rc = 0;
1184 const char *cp;
1186 while (len && path[len - 1] == '/')
1187 len--;
1188 if (!len)
1189 return 1;
1190 baselen = 0;
1191 while (1) {
1192 cp = path + baselen + !!baselen;
1193 cp = memchr(cp, '/', path + len - cp);
1194 if (!cp)
1195 baselen = len;
1196 else
1197 baselen = cp - path;
1198 strbuf_setlen(&sb, 0);
1199 strbuf_add(&sb, path, baselen);
1200 if (!is_directory(sb.buf))
1201 break;
1202 if (simplify_away(sb.buf, sb.len, simplify))
1203 break;
1204 if (treat_one_path(dir, &sb, simplify,
1205 DT_DIR, NULL) == path_ignored)
1206 break; /* do not recurse into it */
1207 if (len <= baselen) {
1208 rc = 1;
1209 break; /* finished checking */
1212 strbuf_release(&sb);
1213 return rc;
1216 int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
1218 struct path_simplify *simplify;
1220 if (has_symlink_leading_path(path, len))
1221 return dir->nr;
1223 simplify = create_simplify(pathspec);
1224 if (!len || treat_leading_path(dir, path, len, simplify))
1225 read_directory_recursive(dir, path, len, 0, simplify);
1226 free_simplify(simplify);
1227 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1228 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1229 return dir->nr;
1232 int file_exists(const char *f)
1234 struct stat sb;
1235 return lstat(f, &sb) == 0;
1239 * Given two normalized paths (a trailing slash is ok), if subdir is
1240 * outside dir, return -1. Otherwise return the offset in subdir that
1241 * can be used as relative path to dir.
1243 int dir_inside_of(const char *subdir, const char *dir)
1245 int offset = 0;
1247 assert(dir && subdir && *dir && *subdir);
1249 while (*dir && *subdir && *dir == *subdir) {
1250 dir++;
1251 subdir++;
1252 offset++;
1255 /* hel[p]/me vs hel[l]/yeah */
1256 if (*dir && *subdir)
1257 return -1;
1259 if (!*subdir)
1260 return !*dir ? offset : -1; /* same dir */
1262 /* foo/[b]ar vs foo/[] */
1263 if (is_dir_sep(dir[-1]))
1264 return is_dir_sep(subdir[-1]) ? offset : -1;
1266 /* foo[/]bar vs foo[] */
1267 return is_dir_sep(*subdir) ? offset + 1 : -1;
1270 int is_inside_dir(const char *dir)
1272 char cwd[PATH_MAX];
1273 if (!dir)
1274 return 0;
1275 if (!getcwd(cwd, sizeof(cwd)))
1276 die_errno("can't find the current directory");
1277 return dir_inside_of(cwd, dir) >= 0;
1280 int is_empty_dir(const char *path)
1282 DIR *dir = opendir(path);
1283 struct dirent *e;
1284 int ret = 1;
1286 if (!dir)
1287 return 0;
1289 while ((e = readdir(dir)) != NULL)
1290 if (!is_dot_or_dotdot(e->d_name)) {
1291 ret = 0;
1292 break;
1295 closedir(dir);
1296 return ret;
1299 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1301 DIR *dir;
1302 struct dirent *e;
1303 int ret = 0, original_len = path->len, len, kept_down = 0;
1304 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1305 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1306 unsigned char submodule_head[20];
1308 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1309 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1310 /* Do not descend and nuke a nested git work tree. */
1311 if (kept_up)
1312 *kept_up = 1;
1313 return 0;
1316 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1317 dir = opendir(path->buf);
1318 if (!dir) {
1319 /* an empty dir could be removed even if it is unreadble */
1320 if (!keep_toplevel)
1321 return rmdir(path->buf);
1322 else
1323 return -1;
1325 if (path->buf[original_len - 1] != '/')
1326 strbuf_addch(path, '/');
1328 len = path->len;
1329 while ((e = readdir(dir)) != NULL) {
1330 struct stat st;
1331 if (is_dot_or_dotdot(e->d_name))
1332 continue;
1334 strbuf_setlen(path, len);
1335 strbuf_addstr(path, e->d_name);
1336 if (lstat(path->buf, &st))
1337 ; /* fall thru */
1338 else if (S_ISDIR(st.st_mode)) {
1339 if (!remove_dir_recurse(path, flag, &kept_down))
1340 continue; /* happy */
1341 } else if (!only_empty && !unlink(path->buf))
1342 continue; /* happy, too */
1344 /* path too long, stat fails, or non-directory still exists */
1345 ret = -1;
1346 break;
1348 closedir(dir);
1350 strbuf_setlen(path, original_len);
1351 if (!ret && !keep_toplevel && !kept_down)
1352 ret = rmdir(path->buf);
1353 else if (kept_up)
1355 * report the uplevel that it is not an error that we
1356 * did not rmdir() our directory.
1358 *kept_up = !ret;
1359 return ret;
1362 int remove_dir_recursively(struct strbuf *path, int flag)
1364 return remove_dir_recurse(path, flag, NULL);
1367 void setup_standard_excludes(struct dir_struct *dir)
1369 const char *path;
1370 char *xdg_path;
1372 dir->exclude_per_dir = ".gitignore";
1373 path = git_path("info/exclude");
1374 if (!excludes_file) {
1375 home_config_paths(NULL, &xdg_path, "ignore");
1376 excludes_file = xdg_path;
1378 if (!access_or_warn(path, R_OK))
1379 add_excludes_from_file(dir, path);
1380 if (excludes_file && !access_or_warn(excludes_file, R_OK))
1381 add_excludes_from_file(dir, excludes_file);
1384 int remove_path(const char *name)
1386 char *slash;
1388 if (unlink(name) && errno != ENOENT)
1389 return -1;
1391 slash = strrchr(name, '/');
1392 if (slash) {
1393 char *dirs = xstrdup(name);
1394 slash = dirs + (slash - name);
1395 do {
1396 *slash = '\0';
1397 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1398 free(dirs);
1400 return 0;
1403 static int pathspec_item_cmp(const void *a_, const void *b_)
1405 struct pathspec_item *a, *b;
1407 a = (struct pathspec_item *)a_;
1408 b = (struct pathspec_item *)b_;
1409 return strcmp(a->match, b->match);
1412 int init_pathspec(struct pathspec *pathspec, const char **paths)
1414 const char **p = paths;
1415 int i;
1417 memset(pathspec, 0, sizeof(*pathspec));
1418 if (!p)
1419 return 0;
1420 while (*p)
1421 p++;
1422 pathspec->raw = paths;
1423 pathspec->nr = p - paths;
1424 if (!pathspec->nr)
1425 return 0;
1427 pathspec->items = xmalloc(sizeof(struct pathspec_item)*pathspec->nr);
1428 for (i = 0; i < pathspec->nr; i++) {
1429 struct pathspec_item *item = pathspec->items+i;
1430 const char *path = paths[i];
1432 item->match = path;
1433 item->len = strlen(path);
1434 item->use_wildcard = !no_wildcard(path);
1435 if (item->use_wildcard)
1436 pathspec->has_wildcard = 1;
1439 qsort(pathspec->items, pathspec->nr,
1440 sizeof(struct pathspec_item), pathspec_item_cmp);
1442 return 0;
1445 void free_pathspec(struct pathspec *pathspec)
1447 free(pathspec->items);
1448 pathspec->items = NULL;