Update draft release notes to 1.7.12
[git/jnareb-git.git] / dir.c
bloba772c6dc6c2bf4dc40f46f0bb5d0899312f3ffd4
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"
12 struct path_simplify {
13 int len;
14 const char *path;
17 static int read_directory_recursive(struct dir_struct *dir, const char *path, int len,
18 int check_only, const struct path_simplify *simplify);
19 static int get_dtype(struct dirent *de, const char *path, int len);
21 /* helper string functions with support for the ignore_case flag */
22 int strcmp_icase(const char *a, const char *b)
24 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
27 int strncmp_icase(const char *a, const char *b, size_t count)
29 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
32 int fnmatch_icase(const char *pattern, const char *string, int flags)
34 return fnmatch(pattern, string, flags | (ignore_case ? FNM_CASEFOLD : 0));
37 static size_t common_prefix_len(const char **pathspec)
39 const char *n, *first;
40 size_t max = 0;
42 if (!pathspec)
43 return max;
45 first = *pathspec;
46 while ((n = *pathspec++)) {
47 size_t i, len = 0;
48 for (i = 0; first == n || i < max; i++) {
49 char c = n[i];
50 if (!c || c != first[i] || is_glob_special(c))
51 break;
52 if (c == '/')
53 len = i + 1;
55 if (first == n || len < max) {
56 max = len;
57 if (!max)
58 break;
61 return max;
65 * Returns a copy of the longest leading path common among all
66 * pathspecs.
68 char *common_prefix(const char **pathspec)
70 unsigned long len = common_prefix_len(pathspec);
72 return len ? xmemdupz(*pathspec, len) : NULL;
75 int fill_directory(struct dir_struct *dir, const char **pathspec)
77 size_t len;
80 * Calculate common prefix for the pathspec, and
81 * use that to optimize the directory walk
83 len = common_prefix_len(pathspec);
85 /* Read the directory and prune it */
86 read_directory(dir, pathspec ? *pathspec : "", len, pathspec);
87 return len;
90 int within_depth(const char *name, int namelen,
91 int depth, int max_depth)
93 const char *cp = name, *cpe = name + namelen;
95 while (cp < cpe) {
96 if (*cp++ != '/')
97 continue;
98 depth++;
99 if (depth > max_depth)
100 return 0;
102 return 1;
106 * Does 'match' match the given name?
107 * A match is found if
109 * (1) the 'match' string is leading directory of 'name', or
110 * (2) the 'match' string is a wildcard and matches 'name', or
111 * (3) the 'match' string is exactly the same as 'name'.
113 * and the return value tells which case it was.
115 * It returns 0 when there is no match.
117 static int match_one(const char *match, const char *name, int namelen)
119 int matchlen;
121 /* If the match was just the prefix, we matched */
122 if (!*match)
123 return MATCHED_RECURSIVELY;
125 if (ignore_case) {
126 for (;;) {
127 unsigned char c1 = tolower(*match);
128 unsigned char c2 = tolower(*name);
129 if (c1 == '\0' || is_glob_special(c1))
130 break;
131 if (c1 != c2)
132 return 0;
133 match++;
134 name++;
135 namelen--;
137 } else {
138 for (;;) {
139 unsigned char c1 = *match;
140 unsigned char c2 = *name;
141 if (c1 == '\0' || is_glob_special(c1))
142 break;
143 if (c1 != c2)
144 return 0;
145 match++;
146 name++;
147 namelen--;
153 * If we don't match the matchstring exactly,
154 * we need to match by fnmatch
156 matchlen = strlen(match);
157 if (strncmp_icase(match, name, matchlen))
158 return !fnmatch_icase(match, name, 0) ? MATCHED_FNMATCH : 0;
160 if (namelen == matchlen)
161 return MATCHED_EXACTLY;
162 if (match[matchlen-1] == '/' || name[matchlen] == '/')
163 return MATCHED_RECURSIVELY;
164 return 0;
168 * Given a name and a list of pathspecs, see if the name matches
169 * any of the pathspecs. The caller is also interested in seeing
170 * all pathspec matches some names it calls this function with
171 * (otherwise the user could have mistyped the unmatched pathspec),
172 * and a mark is left in seen[] array for pathspec element that
173 * actually matched anything.
175 int match_pathspec(const char **pathspec, const char *name, int namelen,
176 int prefix, char *seen)
178 int i, retval = 0;
180 if (!pathspec)
181 return 1;
183 name += prefix;
184 namelen -= prefix;
186 for (i = 0; pathspec[i] != NULL; i++) {
187 int how;
188 const char *match = pathspec[i] + prefix;
189 if (seen && seen[i] == MATCHED_EXACTLY)
190 continue;
191 how = match_one(match, name, namelen);
192 if (how) {
193 if (retval < how)
194 retval = how;
195 if (seen && seen[i] < how)
196 seen[i] = how;
199 return retval;
203 * Does 'match' match the given name?
204 * A match is found if
206 * (1) the 'match' string is leading directory of 'name', or
207 * (2) the 'match' string is a wildcard and matches 'name', or
208 * (3) the 'match' string is exactly the same as 'name'.
210 * and the return value tells which case it was.
212 * It returns 0 when there is no match.
214 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
215 const char *name, int namelen)
217 /* name/namelen has prefix cut off by caller */
218 const char *match = item->match + prefix;
219 int matchlen = item->len - prefix;
221 /* If the match was just the prefix, we matched */
222 if (!*match)
223 return MATCHED_RECURSIVELY;
225 if (matchlen <= namelen && !strncmp(match, name, matchlen)) {
226 if (matchlen == namelen)
227 return MATCHED_EXACTLY;
229 if (match[matchlen-1] == '/' || name[matchlen] == '/')
230 return MATCHED_RECURSIVELY;
233 if (item->use_wildcard && !fnmatch(match, name, 0))
234 return MATCHED_FNMATCH;
236 return 0;
240 * Given a name and a list of pathspecs, see if the name matches
241 * any of the pathspecs. The caller is also interested in seeing
242 * all pathspec matches some names it calls this function with
243 * (otherwise the user could have mistyped the unmatched pathspec),
244 * and a mark is left in seen[] array for pathspec element that
245 * actually matched anything.
247 int match_pathspec_depth(const struct pathspec *ps,
248 const char *name, int namelen,
249 int prefix, char *seen)
251 int i, retval = 0;
253 if (!ps->nr) {
254 if (!ps->recursive || ps->max_depth == -1)
255 return MATCHED_RECURSIVELY;
257 if (within_depth(name, namelen, 0, ps->max_depth))
258 return MATCHED_EXACTLY;
259 else
260 return 0;
263 name += prefix;
264 namelen -= prefix;
266 for (i = ps->nr - 1; i >= 0; i--) {
267 int how;
268 if (seen && seen[i] == MATCHED_EXACTLY)
269 continue;
270 how = match_pathspec_item(ps->items+i, prefix, name, namelen);
271 if (ps->recursive && ps->max_depth != -1 &&
272 how && how != MATCHED_FNMATCH) {
273 int len = ps->items[i].len;
274 if (name[len] == '/')
275 len++;
276 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
277 how = MATCHED_EXACTLY;
278 else
279 how = 0;
281 if (how) {
282 if (retval < how)
283 retval = how;
284 if (seen && seen[i] < how)
285 seen[i] = how;
288 return retval;
292 * Return the length of the "simple" part of a path match limiter.
294 static int simple_length(const char *match)
296 int len = -1;
298 for (;;) {
299 unsigned char c = *match++;
300 len++;
301 if (c == '\0' || is_glob_special(c))
302 return len;
306 static int no_wildcard(const char *string)
308 return string[simple_length(string)] == '\0';
311 void add_exclude(const char *string, const char *base,
312 int baselen, struct exclude_list *which)
314 struct exclude *x;
315 size_t len;
316 int to_exclude = 1;
317 int flags = 0;
319 if (*string == '!') {
320 to_exclude = 0;
321 string++;
323 len = strlen(string);
324 if (len && string[len - 1] == '/') {
325 char *s;
326 x = xmalloc(sizeof(*x) + len);
327 s = (char *)(x+1);
328 memcpy(s, string, len - 1);
329 s[len - 1] = '\0';
330 string = s;
331 x->pattern = s;
332 flags = EXC_FLAG_MUSTBEDIR;
333 } else {
334 x = xmalloc(sizeof(*x));
335 x->pattern = string;
337 x->to_exclude = to_exclude;
338 x->patternlen = strlen(string);
339 x->base = base;
340 x->baselen = baselen;
341 x->flags = flags;
342 if (!strchr(string, '/'))
343 x->flags |= EXC_FLAG_NODIR;
344 x->nowildcardlen = simple_length(string);
345 if (*string == '*' && no_wildcard(string+1))
346 x->flags |= EXC_FLAG_ENDSWITH;
347 ALLOC_GROW(which->excludes, which->nr + 1, which->alloc);
348 which->excludes[which->nr++] = x;
351 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
353 int pos, len;
354 unsigned long sz;
355 enum object_type type;
356 void *data;
357 struct index_state *istate = &the_index;
359 len = strlen(path);
360 pos = index_name_pos(istate, path, len);
361 if (pos < 0)
362 return NULL;
363 if (!ce_skip_worktree(istate->cache[pos]))
364 return NULL;
365 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
366 if (!data || type != OBJ_BLOB) {
367 free(data);
368 return NULL;
370 *size = xsize_t(sz);
371 return data;
374 void free_excludes(struct exclude_list *el)
376 int i;
378 for (i = 0; i < el->nr; i++)
379 free(el->excludes[i]);
380 free(el->excludes);
382 el->nr = 0;
383 el->excludes = NULL;
386 int add_excludes_from_file_to_list(const char *fname,
387 const char *base,
388 int baselen,
389 char **buf_p,
390 struct exclude_list *which,
391 int check_index)
393 struct stat st;
394 int fd, i;
395 size_t size = 0;
396 char *buf, *entry;
398 fd = open(fname, O_RDONLY);
399 if (fd < 0 || fstat(fd, &st) < 0) {
400 if (0 <= fd)
401 close(fd);
402 if (!check_index ||
403 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
404 return -1;
405 if (size == 0) {
406 free(buf);
407 return 0;
409 if (buf[size-1] != '\n') {
410 buf = xrealloc(buf, size+1);
411 buf[size++] = '\n';
414 else {
415 size = xsize_t(st.st_size);
416 if (size == 0) {
417 close(fd);
418 return 0;
420 buf = xmalloc(size+1);
421 if (read_in_full(fd, buf, size) != size) {
422 free(buf);
423 close(fd);
424 return -1;
426 buf[size++] = '\n';
427 close(fd);
430 if (buf_p)
431 *buf_p = buf;
432 entry = buf;
433 for (i = 0; i < size; i++) {
434 if (buf[i] == '\n') {
435 if (entry != buf + i && entry[0] != '#') {
436 buf[i - (i && buf[i-1] == '\r')] = 0;
437 add_exclude(entry, base, baselen, which);
439 entry = buf + i + 1;
442 return 0;
445 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
447 if (add_excludes_from_file_to_list(fname, "", 0, NULL,
448 &dir->exclude_list[EXC_FILE], 0) < 0)
449 die("cannot use %s as an exclude file", fname);
452 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
454 struct exclude_list *el;
455 struct exclude_stack *stk = NULL;
456 int current;
458 if ((!dir->exclude_per_dir) ||
459 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
460 return; /* too long a path -- ignore */
462 /* Pop the ones that are not the prefix of the path being checked. */
463 el = &dir->exclude_list[EXC_DIRS];
464 while ((stk = dir->exclude_stack) != NULL) {
465 if (stk->baselen <= baselen &&
466 !strncmp(dir->basebuf, base, stk->baselen))
467 break;
468 dir->exclude_stack = stk->prev;
469 while (stk->exclude_ix < el->nr)
470 free(el->excludes[--el->nr]);
471 free(stk->filebuf);
472 free(stk);
475 /* Read from the parent directories and push them down. */
476 current = stk ? stk->baselen : -1;
477 while (current < baselen) {
478 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
479 const char *cp;
481 if (current < 0) {
482 cp = base;
483 current = 0;
485 else {
486 cp = strchr(base + current + 1, '/');
487 if (!cp)
488 die("oops in prep_exclude");
489 cp++;
491 stk->prev = dir->exclude_stack;
492 stk->baselen = cp - base;
493 stk->exclude_ix = el->nr;
494 memcpy(dir->basebuf + current, base + current,
495 stk->baselen - current);
496 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
497 add_excludes_from_file_to_list(dir->basebuf,
498 dir->basebuf, stk->baselen,
499 &stk->filebuf, el, 1);
500 dir->exclude_stack = stk;
501 current = stk->baselen;
503 dir->basebuf[baselen] = '\0';
506 /* Scan the list and let the last match determine the fate.
507 * Return 1 for exclude, 0 for include and -1 for undecided.
509 int excluded_from_list(const char *pathname,
510 int pathlen, const char *basename, int *dtype,
511 struct exclude_list *el)
513 int i;
515 if (!el->nr)
516 return -1; /* undefined */
518 for (i = el->nr - 1; 0 <= i; i--) {
519 struct exclude *x = el->excludes[i];
520 const char *name, *exclude = x->pattern;
521 int to_exclude = x->to_exclude;
522 int namelen, prefix = x->nowildcardlen;
524 if (x->flags & EXC_FLAG_MUSTBEDIR) {
525 if (*dtype == DT_UNKNOWN)
526 *dtype = get_dtype(NULL, pathname, pathlen);
527 if (*dtype != DT_DIR)
528 continue;
531 if (x->flags & EXC_FLAG_NODIR) {
532 /* match basename */
533 if (prefix == x->patternlen) {
534 if (!strcmp_icase(exclude, basename))
535 return to_exclude;
536 } else if (x->flags & EXC_FLAG_ENDSWITH) {
537 if (x->patternlen - 1 <= pathlen &&
538 !strcmp_icase(exclude + 1, pathname + pathlen - x->patternlen + 1))
539 return to_exclude;
540 } else {
541 if (fnmatch_icase(exclude, basename, 0) == 0)
542 return to_exclude;
544 continue;
547 /* match with FNM_PATHNAME:
548 * exclude has base (baselen long) implicitly in front of it.
550 if (*exclude == '/') {
551 exclude++;
552 prefix--;
555 if (pathlen < x->baselen ||
556 (x->baselen && pathname[x->baselen-1] != '/') ||
557 strncmp_icase(pathname, x->base, x->baselen))
558 continue;
560 namelen = x->baselen ? pathlen - x->baselen : pathlen;
561 name = pathname + pathlen - namelen;
563 /* if the non-wildcard part is longer than the
564 remaining pathname, surely it cannot match */
565 if (prefix > namelen)
566 continue;
568 if (prefix) {
569 if (strncmp_icase(exclude, name, prefix))
570 continue;
571 exclude += prefix;
572 name += prefix;
573 namelen -= prefix;
576 if (!namelen || !fnmatch_icase(exclude, name, FNM_PATHNAME))
577 return to_exclude;
579 return -1; /* undecided */
582 static int excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
584 int pathlen = strlen(pathname);
585 int st;
586 const char *basename = strrchr(pathname, '/');
587 basename = (basename) ? basename+1 : pathname;
589 prep_exclude(dir, pathname, basename-pathname);
590 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
591 switch (excluded_from_list(pathname, pathlen, basename,
592 dtype_p, &dir->exclude_list[st])) {
593 case 0:
594 return 0;
595 case 1:
596 return 1;
599 return 0;
602 void path_exclude_check_init(struct path_exclude_check *check,
603 struct dir_struct *dir)
605 check->dir = dir;
606 strbuf_init(&check->path, 256);
609 void path_exclude_check_clear(struct path_exclude_check *check)
611 strbuf_release(&check->path);
615 * Is this name excluded? This is for a caller like show_files() that
616 * do not honor directory hierarchy and iterate through paths that are
617 * possibly in an ignored directory.
619 * A path to a directory known to be excluded is left in check->path to
620 * optimize for repeated checks for files in the same excluded directory.
622 int path_excluded(struct path_exclude_check *check,
623 const char *name, int namelen, int *dtype)
625 int i;
626 struct strbuf *path = &check->path;
629 * we allow the caller to pass namelen as an optimization; it
630 * must match the length of the name, as we eventually call
631 * excluded() on the whole name string.
633 if (namelen < 0)
634 namelen = strlen(name);
636 if (path->len &&
637 path->len <= namelen &&
638 !memcmp(name, path->buf, path->len) &&
639 (!name[path->len] || name[path->len] == '/'))
640 return 1;
642 strbuf_setlen(path, 0);
643 for (i = 0; name[i]; i++) {
644 int ch = name[i];
646 if (ch == '/') {
647 int dt = DT_DIR;
648 if (excluded(check->dir, path->buf, &dt))
649 return 1;
651 strbuf_addch(path, ch);
654 /* An entry in the index; cannot be a directory with subentries */
655 strbuf_setlen(path, 0);
657 return excluded(check->dir, name, dtype);
660 static struct dir_entry *dir_entry_new(const char *pathname, int len)
662 struct dir_entry *ent;
664 ent = xmalloc(sizeof(*ent) + len + 1);
665 ent->len = len;
666 memcpy(ent->name, pathname, len);
667 ent->name[len] = 0;
668 return ent;
671 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
673 if (cache_name_exists(pathname, len, ignore_case))
674 return NULL;
676 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
677 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
680 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
682 if (!cache_name_is_other(pathname, len))
683 return NULL;
685 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
686 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
689 enum exist_status {
690 index_nonexistent = 0,
691 index_directory,
692 index_gitdir
696 * Do not use the alphabetically stored index to look up
697 * the directory name; instead, use the case insensitive
698 * name hash.
700 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
702 struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
703 unsigned char endchar;
705 if (!ce)
706 return index_nonexistent;
707 endchar = ce->name[len];
710 * The cache_entry structure returned will contain this dirname
711 * and possibly additional path components.
713 if (endchar == '/')
714 return index_directory;
717 * If there are no additional path components, then this cache_entry
718 * represents a submodule. Submodules, despite being directories,
719 * are stored in the cache without a closing slash.
721 if (!endchar && S_ISGITLINK(ce->ce_mode))
722 return index_gitdir;
724 /* This should never be hit, but it exists just in case. */
725 return index_nonexistent;
729 * The index sorts alphabetically by entry name, which
730 * means that a gitlink sorts as '\0' at the end, while
731 * a directory (which is defined not as an entry, but as
732 * the files it contains) will sort with the '/' at the
733 * end.
735 static enum exist_status directory_exists_in_index(const char *dirname, int len)
737 int pos;
739 if (ignore_case)
740 return directory_exists_in_index_icase(dirname, len);
742 pos = cache_name_pos(dirname, len);
743 if (pos < 0)
744 pos = -pos-1;
745 while (pos < active_nr) {
746 struct cache_entry *ce = active_cache[pos++];
747 unsigned char endchar;
749 if (strncmp(ce->name, dirname, len))
750 break;
751 endchar = ce->name[len];
752 if (endchar > '/')
753 break;
754 if (endchar == '/')
755 return index_directory;
756 if (!endchar && S_ISGITLINK(ce->ce_mode))
757 return index_gitdir;
759 return index_nonexistent;
763 * When we find a directory when traversing the filesystem, we
764 * have three distinct cases:
766 * - ignore it
767 * - see it as a directory
768 * - recurse into it
770 * and which one we choose depends on a combination of existing
771 * git index contents and the flags passed into the directory
772 * traversal routine.
774 * Case 1: If we *already* have entries in the index under that
775 * directory name, we always recurse into the directory to see
776 * all the files.
778 * Case 2: If we *already* have that directory name as a gitlink,
779 * we always continue to see it as a gitlink, regardless of whether
780 * there is an actual git directory there or not (it might not
781 * be checked out as a subproject!)
783 * Case 3: if we didn't have it in the index previously, we
784 * have a few sub-cases:
786 * (a) if "show_other_directories" is true, we show it as
787 * just a directory, unless "hide_empty_directories" is
788 * also true and the directory is empty, in which case
789 * we just ignore it entirely.
790 * (b) if it looks like a git directory, and we don't have
791 * 'no_gitlinks' set we treat it as a gitlink, and show it
792 * as a directory.
793 * (c) otherwise, we recurse into it.
795 enum directory_treatment {
796 show_directory,
797 ignore_directory,
798 recurse_into_directory
801 static enum directory_treatment treat_directory(struct dir_struct *dir,
802 const char *dirname, int len,
803 const struct path_simplify *simplify)
805 /* The "len-1" is to strip the final '/' */
806 switch (directory_exists_in_index(dirname, len-1)) {
807 case index_directory:
808 return recurse_into_directory;
810 case index_gitdir:
811 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
812 return ignore_directory;
813 return show_directory;
815 case index_nonexistent:
816 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
817 break;
818 if (!(dir->flags & DIR_NO_GITLINKS)) {
819 unsigned char sha1[20];
820 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
821 return show_directory;
823 return recurse_into_directory;
826 /* This is the "show_other_directories" case */
827 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
828 return show_directory;
829 if (!read_directory_recursive(dir, dirname, len, 1, simplify))
830 return ignore_directory;
831 return show_directory;
835 * This is an inexact early pruning of any recursive directory
836 * reading - if the path cannot possibly be in the pathspec,
837 * return true, and we'll skip it early.
839 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
841 if (simplify) {
842 for (;;) {
843 const char *match = simplify->path;
844 int len = simplify->len;
846 if (!match)
847 break;
848 if (len > pathlen)
849 len = pathlen;
850 if (!memcmp(path, match, len))
851 return 0;
852 simplify++;
854 return 1;
856 return 0;
860 * This function tells us whether an excluded path matches a
861 * list of "interesting" pathspecs. That is, whether a path matched
862 * by any of the pathspecs could possibly be ignored by excluding
863 * the specified path. This can happen if:
865 * 1. the path is mentioned explicitly in the pathspec
867 * 2. the path is a directory prefix of some element in the
868 * pathspec
870 static int exclude_matches_pathspec(const char *path, int len,
871 const struct path_simplify *simplify)
873 if (simplify) {
874 for (; simplify->path; simplify++) {
875 if (len == simplify->len
876 && !memcmp(path, simplify->path, len))
877 return 1;
878 if (len < simplify->len
879 && simplify->path[len] == '/'
880 && !memcmp(path, simplify->path, len))
881 return 1;
884 return 0;
887 static int get_index_dtype(const char *path, int len)
889 int pos;
890 struct cache_entry *ce;
892 ce = cache_name_exists(path, len, 0);
893 if (ce) {
894 if (!ce_uptodate(ce))
895 return DT_UNKNOWN;
896 if (S_ISGITLINK(ce->ce_mode))
897 return DT_DIR;
899 * Nobody actually cares about the
900 * difference between DT_LNK and DT_REG
902 return DT_REG;
905 /* Try to look it up as a directory */
906 pos = cache_name_pos(path, len);
907 if (pos >= 0)
908 return DT_UNKNOWN;
909 pos = -pos-1;
910 while (pos < active_nr) {
911 ce = active_cache[pos++];
912 if (strncmp(ce->name, path, len))
913 break;
914 if (ce->name[len] > '/')
915 break;
916 if (ce->name[len] < '/')
917 continue;
918 if (!ce_uptodate(ce))
919 break; /* continue? */
920 return DT_DIR;
922 return DT_UNKNOWN;
925 static int get_dtype(struct dirent *de, const char *path, int len)
927 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
928 struct stat st;
930 if (dtype != DT_UNKNOWN)
931 return dtype;
932 dtype = get_index_dtype(path, len);
933 if (dtype != DT_UNKNOWN)
934 return dtype;
935 if (lstat(path, &st))
936 return dtype;
937 if (S_ISREG(st.st_mode))
938 return DT_REG;
939 if (S_ISDIR(st.st_mode))
940 return DT_DIR;
941 if (S_ISLNK(st.st_mode))
942 return DT_LNK;
943 return dtype;
946 enum path_treatment {
947 path_ignored,
948 path_handled,
949 path_recurse
952 static enum path_treatment treat_one_path(struct dir_struct *dir,
953 struct strbuf *path,
954 const struct path_simplify *simplify,
955 int dtype, struct dirent *de)
957 int exclude = excluded(dir, path->buf, &dtype);
958 if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
959 && exclude_matches_pathspec(path->buf, path->len, simplify))
960 dir_add_ignored(dir, path->buf, path->len);
963 * Excluded? If we don't explicitly want to show
964 * ignored files, ignore it
966 if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
967 return path_ignored;
969 if (dtype == DT_UNKNOWN)
970 dtype = get_dtype(de, path->buf, path->len);
973 * Do we want to see just the ignored files?
974 * We still need to recurse into directories,
975 * even if we don't ignore them, since the
976 * directory may contain files that we do..
978 if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
979 if (dtype != DT_DIR)
980 return path_ignored;
983 switch (dtype) {
984 default:
985 return path_ignored;
986 case DT_DIR:
987 strbuf_addch(path, '/');
988 switch (treat_directory(dir, path->buf, path->len, simplify)) {
989 case show_directory:
990 if (exclude != !!(dir->flags
991 & DIR_SHOW_IGNORED))
992 return path_ignored;
993 break;
994 case recurse_into_directory:
995 return path_recurse;
996 case ignore_directory:
997 return path_ignored;
999 break;
1000 case DT_REG:
1001 case DT_LNK:
1002 break;
1004 return path_handled;
1007 static enum path_treatment treat_path(struct dir_struct *dir,
1008 struct dirent *de,
1009 struct strbuf *path,
1010 int baselen,
1011 const struct path_simplify *simplify)
1013 int dtype;
1015 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1016 return path_ignored;
1017 strbuf_setlen(path, baselen);
1018 strbuf_addstr(path, de->d_name);
1019 if (simplify_away(path->buf, path->len, simplify))
1020 return path_ignored;
1022 dtype = DTYPE(de);
1023 return treat_one_path(dir, path, simplify, dtype, de);
1027 * Read a directory tree. We currently ignore anything but
1028 * directories, regular files and symlinks. That's because git
1029 * doesn't handle them at all yet. Maybe that will change some
1030 * day.
1032 * Also, we ignore the name ".git" (even if it is not a directory).
1033 * That likely will not change.
1035 static int read_directory_recursive(struct dir_struct *dir,
1036 const char *base, int baselen,
1037 int check_only,
1038 const struct path_simplify *simplify)
1040 DIR *fdir;
1041 int contents = 0;
1042 struct dirent *de;
1043 struct strbuf path = STRBUF_INIT;
1045 strbuf_add(&path, base, baselen);
1047 fdir = opendir(path.len ? path.buf : ".");
1048 if (!fdir)
1049 goto out;
1051 while ((de = readdir(fdir)) != NULL) {
1052 switch (treat_path(dir, de, &path, baselen, simplify)) {
1053 case path_recurse:
1054 contents += read_directory_recursive(dir, path.buf,
1055 path.len, 0,
1056 simplify);
1057 continue;
1058 case path_ignored:
1059 continue;
1060 case path_handled:
1061 break;
1063 contents++;
1064 if (check_only)
1065 break;
1066 dir_add_name(dir, path.buf, path.len);
1068 closedir(fdir);
1069 out:
1070 strbuf_release(&path);
1072 return contents;
1075 static int cmp_name(const void *p1, const void *p2)
1077 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1078 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1080 return cache_name_compare(e1->name, e1->len,
1081 e2->name, e2->len);
1084 static struct path_simplify *create_simplify(const char **pathspec)
1086 int nr, alloc = 0;
1087 struct path_simplify *simplify = NULL;
1089 if (!pathspec)
1090 return NULL;
1092 for (nr = 0 ; ; nr++) {
1093 const char *match;
1094 if (nr >= alloc) {
1095 alloc = alloc_nr(alloc);
1096 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
1098 match = *pathspec++;
1099 if (!match)
1100 break;
1101 simplify[nr].path = match;
1102 simplify[nr].len = simple_length(match);
1104 simplify[nr].path = NULL;
1105 simplify[nr].len = 0;
1106 return simplify;
1109 static void free_simplify(struct path_simplify *simplify)
1111 free(simplify);
1114 static int treat_leading_path(struct dir_struct *dir,
1115 const char *path, int len,
1116 const struct path_simplify *simplify)
1118 struct strbuf sb = STRBUF_INIT;
1119 int baselen, rc = 0;
1120 const char *cp;
1122 while (len && path[len - 1] == '/')
1123 len--;
1124 if (!len)
1125 return 1;
1126 baselen = 0;
1127 while (1) {
1128 cp = path + baselen + !!baselen;
1129 cp = memchr(cp, '/', path + len - cp);
1130 if (!cp)
1131 baselen = len;
1132 else
1133 baselen = cp - path;
1134 strbuf_setlen(&sb, 0);
1135 strbuf_add(&sb, path, baselen);
1136 if (!is_directory(sb.buf))
1137 break;
1138 if (simplify_away(sb.buf, sb.len, simplify))
1139 break;
1140 if (treat_one_path(dir, &sb, simplify,
1141 DT_DIR, NULL) == path_ignored)
1142 break; /* do not recurse into it */
1143 if (len <= baselen) {
1144 rc = 1;
1145 break; /* finished checking */
1148 strbuf_release(&sb);
1149 return rc;
1152 int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
1154 struct path_simplify *simplify;
1156 if (has_symlink_leading_path(path, len))
1157 return dir->nr;
1159 simplify = create_simplify(pathspec);
1160 if (!len || treat_leading_path(dir, path, len, simplify))
1161 read_directory_recursive(dir, path, len, 0, simplify);
1162 free_simplify(simplify);
1163 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1164 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1165 return dir->nr;
1168 int file_exists(const char *f)
1170 struct stat sb;
1171 return lstat(f, &sb) == 0;
1175 * Given two normalized paths (a trailing slash is ok), if subdir is
1176 * outside dir, return -1. Otherwise return the offset in subdir that
1177 * can be used as relative path to dir.
1179 int dir_inside_of(const char *subdir, const char *dir)
1181 int offset = 0;
1183 assert(dir && subdir && *dir && *subdir);
1185 while (*dir && *subdir && *dir == *subdir) {
1186 dir++;
1187 subdir++;
1188 offset++;
1191 /* hel[p]/me vs hel[l]/yeah */
1192 if (*dir && *subdir)
1193 return -1;
1195 if (!*subdir)
1196 return !*dir ? offset : -1; /* same dir */
1198 /* foo/[b]ar vs foo/[] */
1199 if (is_dir_sep(dir[-1]))
1200 return is_dir_sep(subdir[-1]) ? offset : -1;
1202 /* foo[/]bar vs foo[] */
1203 return is_dir_sep(*subdir) ? offset + 1 : -1;
1206 int is_inside_dir(const char *dir)
1208 char cwd[PATH_MAX];
1209 if (!dir)
1210 return 0;
1211 if (!getcwd(cwd, sizeof(cwd)))
1212 die_errno("can't find the current directory");
1213 return dir_inside_of(cwd, dir) >= 0;
1216 int is_empty_dir(const char *path)
1218 DIR *dir = opendir(path);
1219 struct dirent *e;
1220 int ret = 1;
1222 if (!dir)
1223 return 0;
1225 while ((e = readdir(dir)) != NULL)
1226 if (!is_dot_or_dotdot(e->d_name)) {
1227 ret = 0;
1228 break;
1231 closedir(dir);
1232 return ret;
1235 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1237 DIR *dir;
1238 struct dirent *e;
1239 int ret = 0, original_len = path->len, len, kept_down = 0;
1240 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1241 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1242 unsigned char submodule_head[20];
1244 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1245 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1246 /* Do not descend and nuke a nested git work tree. */
1247 if (kept_up)
1248 *kept_up = 1;
1249 return 0;
1252 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1253 dir = opendir(path->buf);
1254 if (!dir) {
1255 /* an empty dir could be removed even if it is unreadble */
1256 if (!keep_toplevel)
1257 return rmdir(path->buf);
1258 else
1259 return -1;
1261 if (path->buf[original_len - 1] != '/')
1262 strbuf_addch(path, '/');
1264 len = path->len;
1265 while ((e = readdir(dir)) != NULL) {
1266 struct stat st;
1267 if (is_dot_or_dotdot(e->d_name))
1268 continue;
1270 strbuf_setlen(path, len);
1271 strbuf_addstr(path, e->d_name);
1272 if (lstat(path->buf, &st))
1273 ; /* fall thru */
1274 else if (S_ISDIR(st.st_mode)) {
1275 if (!remove_dir_recurse(path, flag, &kept_down))
1276 continue; /* happy */
1277 } else if (!only_empty && !unlink(path->buf))
1278 continue; /* happy, too */
1280 /* path too long, stat fails, or non-directory still exists */
1281 ret = -1;
1282 break;
1284 closedir(dir);
1286 strbuf_setlen(path, original_len);
1287 if (!ret && !keep_toplevel && !kept_down)
1288 ret = rmdir(path->buf);
1289 else if (kept_up)
1291 * report the uplevel that it is not an error that we
1292 * did not rmdir() our directory.
1294 *kept_up = !ret;
1295 return ret;
1298 int remove_dir_recursively(struct strbuf *path, int flag)
1300 return remove_dir_recurse(path, flag, NULL);
1303 void setup_standard_excludes(struct dir_struct *dir)
1305 const char *path;
1306 char *xdg_path;
1308 dir->exclude_per_dir = ".gitignore";
1309 path = git_path("info/exclude");
1310 if (!excludes_file) {
1311 home_config_paths(NULL, &xdg_path, "ignore");
1312 excludes_file = xdg_path;
1314 if (!access(path, R_OK))
1315 add_excludes_from_file(dir, path);
1316 if (!access(excludes_file, R_OK))
1317 add_excludes_from_file(dir, excludes_file);
1320 int remove_path(const char *name)
1322 char *slash;
1324 if (unlink(name) && errno != ENOENT)
1325 return -1;
1327 slash = strrchr(name, '/');
1328 if (slash) {
1329 char *dirs = xstrdup(name);
1330 slash = dirs + (slash - name);
1331 do {
1332 *slash = '\0';
1333 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1334 free(dirs);
1336 return 0;
1339 static int pathspec_item_cmp(const void *a_, const void *b_)
1341 struct pathspec_item *a, *b;
1343 a = (struct pathspec_item *)a_;
1344 b = (struct pathspec_item *)b_;
1345 return strcmp(a->match, b->match);
1348 int init_pathspec(struct pathspec *pathspec, const char **paths)
1350 const char **p = paths;
1351 int i;
1353 memset(pathspec, 0, sizeof(*pathspec));
1354 if (!p)
1355 return 0;
1356 while (*p)
1357 p++;
1358 pathspec->raw = paths;
1359 pathspec->nr = p - paths;
1360 if (!pathspec->nr)
1361 return 0;
1363 pathspec->items = xmalloc(sizeof(struct pathspec_item)*pathspec->nr);
1364 for (i = 0; i < pathspec->nr; i++) {
1365 struct pathspec_item *item = pathspec->items+i;
1366 const char *path = paths[i];
1368 item->match = path;
1369 item->len = strlen(path);
1370 item->use_wildcard = !no_wildcard(path);
1371 if (item->use_wildcard)
1372 pathspec->has_wildcard = 1;
1375 qsort(pathspec->items, pathspec->nr,
1376 sizeof(struct pathspec_item), pathspec_item_cmp);
1378 return 0;
1381 void free_pathspec(struct pathspec *pathspec)
1383 free(pathspec->items);
1384 pathspec->items = NULL;