Merge branch 'esr/maint-doc-fast-import'
[git.git] / dir.c
blob5a83aa7897f270279c403778f43aea6db1efc5af
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 parse_exclude_pattern(const char **pattern,
312 int *patternlen,
313 int *flags,
314 int *nowildcardlen)
316 const char *p = *pattern;
317 size_t i, len;
319 *flags = 0;
320 if (*p == '!') {
321 *flags |= EXC_FLAG_NEGATIVE;
322 p++;
324 len = strlen(p);
325 if (len && p[len - 1] == '/') {
326 len--;
327 *flags |= EXC_FLAG_MUSTBEDIR;
329 for (i = 0; i < len; i++) {
330 if (p[i] == '/')
331 break;
333 if (i == len)
334 *flags |= EXC_FLAG_NODIR;
335 *nowildcardlen = simple_length(p);
337 * we should have excluded the trailing slash from 'p' too,
338 * but that's one more allocation. Instead just make sure
339 * nowildcardlen does not exceed real patternlen
341 if (*nowildcardlen > len)
342 *nowildcardlen = len;
343 if (*p == '*' && no_wildcard(p + 1))
344 *flags |= EXC_FLAG_ENDSWITH;
345 *pattern = p;
346 *patternlen = len;
349 void add_exclude(const char *string, const char *base,
350 int baselen, struct exclude_list *which)
352 struct exclude *x;
353 int patternlen;
354 int flags;
355 int nowildcardlen;
357 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
358 if (flags & EXC_FLAG_MUSTBEDIR) {
359 char *s;
360 x = xmalloc(sizeof(*x) + patternlen + 1);
361 s = (char *)(x+1);
362 memcpy(s, string, patternlen);
363 s[patternlen] = '\0';
364 x->pattern = s;
365 } else {
366 x = xmalloc(sizeof(*x));
367 x->pattern = string;
369 x->patternlen = patternlen;
370 x->nowildcardlen = nowildcardlen;
371 x->base = base;
372 x->baselen = baselen;
373 x->flags = flags;
374 ALLOC_GROW(which->excludes, which->nr + 1, which->alloc);
375 which->excludes[which->nr++] = x;
378 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
380 int pos, len;
381 unsigned long sz;
382 enum object_type type;
383 void *data;
384 struct index_state *istate = &the_index;
386 len = strlen(path);
387 pos = index_name_pos(istate, path, len);
388 if (pos < 0)
389 return NULL;
390 if (!ce_skip_worktree(istate->cache[pos]))
391 return NULL;
392 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
393 if (!data || type != OBJ_BLOB) {
394 free(data);
395 return NULL;
397 *size = xsize_t(sz);
398 return data;
401 void free_excludes(struct exclude_list *el)
403 int i;
405 for (i = 0; i < el->nr; i++)
406 free(el->excludes[i]);
407 free(el->excludes);
409 el->nr = 0;
410 el->excludes = NULL;
413 int add_excludes_from_file_to_list(const char *fname,
414 const char *base,
415 int baselen,
416 char **buf_p,
417 struct exclude_list *which,
418 int check_index)
420 struct stat st;
421 int fd, i;
422 size_t size = 0;
423 char *buf, *entry;
425 fd = open(fname, O_RDONLY);
426 if (fd < 0 || fstat(fd, &st) < 0) {
427 if (errno != ENOENT)
428 warn_on_inaccessible(fname);
429 if (0 <= fd)
430 close(fd);
431 if (!check_index ||
432 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
433 return -1;
434 if (size == 0) {
435 free(buf);
436 return 0;
438 if (buf[size-1] != '\n') {
439 buf = xrealloc(buf, size+1);
440 buf[size++] = '\n';
443 else {
444 size = xsize_t(st.st_size);
445 if (size == 0) {
446 close(fd);
447 return 0;
449 buf = xmalloc(size+1);
450 if (read_in_full(fd, buf, size) != size) {
451 free(buf);
452 close(fd);
453 return -1;
455 buf[size++] = '\n';
456 close(fd);
459 if (buf_p)
460 *buf_p = buf;
461 entry = buf;
462 for (i = 0; i < size; i++) {
463 if (buf[i] == '\n') {
464 if (entry != buf + i && entry[0] != '#') {
465 buf[i - (i && buf[i-1] == '\r')] = 0;
466 add_exclude(entry, base, baselen, which);
468 entry = buf + i + 1;
471 return 0;
474 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
476 if (add_excludes_from_file_to_list(fname, "", 0, NULL,
477 &dir->exclude_list[EXC_FILE], 0) < 0)
478 die("cannot use %s as an exclude file", fname);
481 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
483 struct exclude_list *el;
484 struct exclude_stack *stk = NULL;
485 int current;
487 if ((!dir->exclude_per_dir) ||
488 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
489 return; /* too long a path -- ignore */
491 /* Pop the ones that are not the prefix of the path being checked. */
492 el = &dir->exclude_list[EXC_DIRS];
493 while ((stk = dir->exclude_stack) != NULL) {
494 if (stk->baselen <= baselen &&
495 !strncmp(dir->basebuf, base, stk->baselen))
496 break;
497 dir->exclude_stack = stk->prev;
498 while (stk->exclude_ix < el->nr)
499 free(el->excludes[--el->nr]);
500 free(stk->filebuf);
501 free(stk);
504 /* Read from the parent directories and push them down. */
505 current = stk ? stk->baselen : -1;
506 while (current < baselen) {
507 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
508 const char *cp;
510 if (current < 0) {
511 cp = base;
512 current = 0;
514 else {
515 cp = strchr(base + current + 1, '/');
516 if (!cp)
517 die("oops in prep_exclude");
518 cp++;
520 stk->prev = dir->exclude_stack;
521 stk->baselen = cp - base;
522 stk->exclude_ix = el->nr;
523 memcpy(dir->basebuf + current, base + current,
524 stk->baselen - current);
525 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
526 add_excludes_from_file_to_list(dir->basebuf,
527 dir->basebuf, stk->baselen,
528 &stk->filebuf, el, 1);
529 dir->exclude_stack = stk;
530 current = stk->baselen;
532 dir->basebuf[baselen] = '\0';
535 int match_basename(const char *basename, int basenamelen,
536 const char *pattern, int prefix, int patternlen,
537 int flags)
539 if (prefix == patternlen) {
540 if (!strcmp_icase(pattern, basename))
541 return 1;
542 } else if (flags & EXC_FLAG_ENDSWITH) {
543 if (patternlen - 1 <= basenamelen &&
544 !strcmp_icase(pattern + 1,
545 basename + basenamelen - patternlen + 1))
546 return 1;
547 } else {
548 if (fnmatch_icase(pattern, basename, 0) == 0)
549 return 1;
551 return 0;
554 int match_pathname(const char *pathname, int pathlen,
555 const char *base, int baselen,
556 const char *pattern, int prefix, int patternlen,
557 int flags)
559 const char *name;
560 int namelen;
563 * match with FNM_PATHNAME; the pattern has base implicitly
564 * in front of it.
566 if (*pattern == '/') {
567 pattern++;
568 prefix--;
572 * baselen does not count the trailing slash. base[] may or
573 * may not end with a trailing slash though.
575 if (pathlen < baselen + 1 ||
576 (baselen && pathname[baselen] != '/') ||
577 strncmp_icase(pathname, base, baselen))
578 return 0;
580 namelen = baselen ? pathlen - baselen - 1 : pathlen;
581 name = pathname + pathlen - namelen;
583 if (prefix) {
585 * if the non-wildcard part is longer than the
586 * remaining pathname, surely it cannot match.
588 if (prefix > namelen)
589 return 0;
591 if (strncmp_icase(pattern, name, prefix))
592 return 0;
593 pattern += prefix;
594 name += prefix;
595 namelen -= prefix;
598 return fnmatch_icase(pattern, name, FNM_PATHNAME) == 0;
601 /* Scan the list and let the last match determine the fate.
602 * Return 1 for exclude, 0 for include and -1 for undecided.
604 int excluded_from_list(const char *pathname,
605 int pathlen, const char *basename, int *dtype,
606 struct exclude_list *el)
608 int i;
610 if (!el->nr)
611 return -1; /* undefined */
613 for (i = el->nr - 1; 0 <= i; i--) {
614 struct exclude *x = el->excludes[i];
615 const char *exclude = x->pattern;
616 int to_exclude = x->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
617 int prefix = x->nowildcardlen;
619 if (x->flags & EXC_FLAG_MUSTBEDIR) {
620 if (*dtype == DT_UNKNOWN)
621 *dtype = get_dtype(NULL, pathname, pathlen);
622 if (*dtype != DT_DIR)
623 continue;
626 if (x->flags & EXC_FLAG_NODIR) {
627 if (match_basename(basename,
628 pathlen - (basename - pathname),
629 exclude, prefix, x->patternlen,
630 x->flags))
631 return to_exclude;
632 continue;
635 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
636 if (match_pathname(pathname, pathlen,
637 x->base, x->baselen ? x->baselen - 1 : 0,
638 exclude, prefix, x->patternlen, x->flags))
639 return to_exclude;
641 return -1; /* undecided */
644 static int excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
646 int pathlen = strlen(pathname);
647 int st;
648 const char *basename = strrchr(pathname, '/');
649 basename = (basename) ? basename+1 : pathname;
651 prep_exclude(dir, pathname, basename-pathname);
652 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
653 switch (excluded_from_list(pathname, pathlen, basename,
654 dtype_p, &dir->exclude_list[st])) {
655 case 0:
656 return 0;
657 case 1:
658 return 1;
661 return 0;
664 void path_exclude_check_init(struct path_exclude_check *check,
665 struct dir_struct *dir)
667 check->dir = dir;
668 strbuf_init(&check->path, 256);
671 void path_exclude_check_clear(struct path_exclude_check *check)
673 strbuf_release(&check->path);
677 * Is this name excluded? This is for a caller like show_files() that
678 * do not honor directory hierarchy and iterate through paths that are
679 * possibly in an ignored directory.
681 * A path to a directory known to be excluded is left in check->path to
682 * optimize for repeated checks for files in the same excluded directory.
684 int path_excluded(struct path_exclude_check *check,
685 const char *name, int namelen, int *dtype)
687 int i;
688 struct strbuf *path = &check->path;
691 * we allow the caller to pass namelen as an optimization; it
692 * must match the length of the name, as we eventually call
693 * excluded() on the whole name string.
695 if (namelen < 0)
696 namelen = strlen(name);
698 if (path->len &&
699 path->len <= namelen &&
700 !memcmp(name, path->buf, path->len) &&
701 (!name[path->len] || name[path->len] == '/'))
702 return 1;
704 strbuf_setlen(path, 0);
705 for (i = 0; name[i]; i++) {
706 int ch = name[i];
708 if (ch == '/') {
709 int dt = DT_DIR;
710 if (excluded(check->dir, path->buf, &dt))
711 return 1;
713 strbuf_addch(path, ch);
716 /* An entry in the index; cannot be a directory with subentries */
717 strbuf_setlen(path, 0);
719 return excluded(check->dir, name, dtype);
722 static struct dir_entry *dir_entry_new(const char *pathname, int len)
724 struct dir_entry *ent;
726 ent = xmalloc(sizeof(*ent) + len + 1);
727 ent->len = len;
728 memcpy(ent->name, pathname, len);
729 ent->name[len] = 0;
730 return ent;
733 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
735 if (cache_name_exists(pathname, len, ignore_case))
736 return NULL;
738 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
739 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
742 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
744 if (!cache_name_is_other(pathname, len))
745 return NULL;
747 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
748 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
751 enum exist_status {
752 index_nonexistent = 0,
753 index_directory,
754 index_gitdir
758 * Do not use the alphabetically stored index to look up
759 * the directory name; instead, use the case insensitive
760 * name hash.
762 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
764 struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
765 unsigned char endchar;
767 if (!ce)
768 return index_nonexistent;
769 endchar = ce->name[len];
772 * The cache_entry structure returned will contain this dirname
773 * and possibly additional path components.
775 if (endchar == '/')
776 return index_directory;
779 * If there are no additional path components, then this cache_entry
780 * represents a submodule. Submodules, despite being directories,
781 * are stored in the cache without a closing slash.
783 if (!endchar && S_ISGITLINK(ce->ce_mode))
784 return index_gitdir;
786 /* This should never be hit, but it exists just in case. */
787 return index_nonexistent;
791 * The index sorts alphabetically by entry name, which
792 * means that a gitlink sorts as '\0' at the end, while
793 * a directory (which is defined not as an entry, but as
794 * the files it contains) will sort with the '/' at the
795 * end.
797 static enum exist_status directory_exists_in_index(const char *dirname, int len)
799 int pos;
801 if (ignore_case)
802 return directory_exists_in_index_icase(dirname, len);
804 pos = cache_name_pos(dirname, len);
805 if (pos < 0)
806 pos = -pos-1;
807 while (pos < active_nr) {
808 struct cache_entry *ce = active_cache[pos++];
809 unsigned char endchar;
811 if (strncmp(ce->name, dirname, len))
812 break;
813 endchar = ce->name[len];
814 if (endchar > '/')
815 break;
816 if (endchar == '/')
817 return index_directory;
818 if (!endchar && S_ISGITLINK(ce->ce_mode))
819 return index_gitdir;
821 return index_nonexistent;
825 * When we find a directory when traversing the filesystem, we
826 * have three distinct cases:
828 * - ignore it
829 * - see it as a directory
830 * - recurse into it
832 * and which one we choose depends on a combination of existing
833 * git index contents and the flags passed into the directory
834 * traversal routine.
836 * Case 1: If we *already* have entries in the index under that
837 * directory name, we always recurse into the directory to see
838 * all the files.
840 * Case 2: If we *already* have that directory name as a gitlink,
841 * we always continue to see it as a gitlink, regardless of whether
842 * there is an actual git directory there or not (it might not
843 * be checked out as a subproject!)
845 * Case 3: if we didn't have it in the index previously, we
846 * have a few sub-cases:
848 * (a) if "show_other_directories" is true, we show it as
849 * just a directory, unless "hide_empty_directories" is
850 * also true and the directory is empty, in which case
851 * we just ignore it entirely.
852 * (b) if it looks like a git directory, and we don't have
853 * 'no_gitlinks' set we treat it as a gitlink, and show it
854 * as a directory.
855 * (c) otherwise, we recurse into it.
857 enum directory_treatment {
858 show_directory,
859 ignore_directory,
860 recurse_into_directory
863 static enum directory_treatment treat_directory(struct dir_struct *dir,
864 const char *dirname, int len,
865 const struct path_simplify *simplify)
867 /* The "len-1" is to strip the final '/' */
868 switch (directory_exists_in_index(dirname, len-1)) {
869 case index_directory:
870 return recurse_into_directory;
872 case index_gitdir:
873 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
874 return ignore_directory;
875 return show_directory;
877 case index_nonexistent:
878 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
879 break;
880 if (!(dir->flags & DIR_NO_GITLINKS)) {
881 unsigned char sha1[20];
882 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
883 return show_directory;
885 return recurse_into_directory;
888 /* This is the "show_other_directories" case */
889 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
890 return show_directory;
891 if (!read_directory_recursive(dir, dirname, len, 1, simplify))
892 return ignore_directory;
893 return show_directory;
897 * This is an inexact early pruning of any recursive directory
898 * reading - if the path cannot possibly be in the pathspec,
899 * return true, and we'll skip it early.
901 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
903 if (simplify) {
904 for (;;) {
905 const char *match = simplify->path;
906 int len = simplify->len;
908 if (!match)
909 break;
910 if (len > pathlen)
911 len = pathlen;
912 if (!memcmp(path, match, len))
913 return 0;
914 simplify++;
916 return 1;
918 return 0;
922 * This function tells us whether an excluded path matches a
923 * list of "interesting" pathspecs. That is, whether a path matched
924 * by any of the pathspecs could possibly be ignored by excluding
925 * the specified path. This can happen if:
927 * 1. the path is mentioned explicitly in the pathspec
929 * 2. the path is a directory prefix of some element in the
930 * pathspec
932 static int exclude_matches_pathspec(const char *path, int len,
933 const struct path_simplify *simplify)
935 if (simplify) {
936 for (; simplify->path; simplify++) {
937 if (len == simplify->len
938 && !memcmp(path, simplify->path, len))
939 return 1;
940 if (len < simplify->len
941 && simplify->path[len] == '/'
942 && !memcmp(path, simplify->path, len))
943 return 1;
946 return 0;
949 static int get_index_dtype(const char *path, int len)
951 int pos;
952 struct cache_entry *ce;
954 ce = cache_name_exists(path, len, 0);
955 if (ce) {
956 if (!ce_uptodate(ce))
957 return DT_UNKNOWN;
958 if (S_ISGITLINK(ce->ce_mode))
959 return DT_DIR;
961 * Nobody actually cares about the
962 * difference between DT_LNK and DT_REG
964 return DT_REG;
967 /* Try to look it up as a directory */
968 pos = cache_name_pos(path, len);
969 if (pos >= 0)
970 return DT_UNKNOWN;
971 pos = -pos-1;
972 while (pos < active_nr) {
973 ce = active_cache[pos++];
974 if (strncmp(ce->name, path, len))
975 break;
976 if (ce->name[len] > '/')
977 break;
978 if (ce->name[len] < '/')
979 continue;
980 if (!ce_uptodate(ce))
981 break; /* continue? */
982 return DT_DIR;
984 return DT_UNKNOWN;
987 static int get_dtype(struct dirent *de, const char *path, int len)
989 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
990 struct stat st;
992 if (dtype != DT_UNKNOWN)
993 return dtype;
994 dtype = get_index_dtype(path, len);
995 if (dtype != DT_UNKNOWN)
996 return dtype;
997 if (lstat(path, &st))
998 return dtype;
999 if (S_ISREG(st.st_mode))
1000 return DT_REG;
1001 if (S_ISDIR(st.st_mode))
1002 return DT_DIR;
1003 if (S_ISLNK(st.st_mode))
1004 return DT_LNK;
1005 return dtype;
1008 enum path_treatment {
1009 path_ignored,
1010 path_handled,
1011 path_recurse
1014 static enum path_treatment treat_one_path(struct dir_struct *dir,
1015 struct strbuf *path,
1016 const struct path_simplify *simplify,
1017 int dtype, struct dirent *de)
1019 int exclude = excluded(dir, path->buf, &dtype);
1020 if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
1021 && exclude_matches_pathspec(path->buf, path->len, simplify))
1022 dir_add_ignored(dir, path->buf, path->len);
1025 * Excluded? If we don't explicitly want to show
1026 * ignored files, ignore it
1028 if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
1029 return path_ignored;
1031 if (dtype == DT_UNKNOWN)
1032 dtype = get_dtype(de, path->buf, path->len);
1035 * Do we want to see just the ignored files?
1036 * We still need to recurse into directories,
1037 * even if we don't ignore them, since the
1038 * directory may contain files that we do..
1040 if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
1041 if (dtype != DT_DIR)
1042 return path_ignored;
1045 switch (dtype) {
1046 default:
1047 return path_ignored;
1048 case DT_DIR:
1049 strbuf_addch(path, '/');
1050 switch (treat_directory(dir, path->buf, path->len, simplify)) {
1051 case show_directory:
1052 if (exclude != !!(dir->flags
1053 & DIR_SHOW_IGNORED))
1054 return path_ignored;
1055 break;
1056 case recurse_into_directory:
1057 return path_recurse;
1058 case ignore_directory:
1059 return path_ignored;
1061 break;
1062 case DT_REG:
1063 case DT_LNK:
1064 break;
1066 return path_handled;
1069 static enum path_treatment treat_path(struct dir_struct *dir,
1070 struct dirent *de,
1071 struct strbuf *path,
1072 int baselen,
1073 const struct path_simplify *simplify)
1075 int dtype;
1077 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1078 return path_ignored;
1079 strbuf_setlen(path, baselen);
1080 strbuf_addstr(path, de->d_name);
1081 if (simplify_away(path->buf, path->len, simplify))
1082 return path_ignored;
1084 dtype = DTYPE(de);
1085 return treat_one_path(dir, path, simplify, dtype, de);
1089 * Read a directory tree. We currently ignore anything but
1090 * directories, regular files and symlinks. That's because git
1091 * doesn't handle them at all yet. Maybe that will change some
1092 * day.
1094 * Also, we ignore the name ".git" (even if it is not a directory).
1095 * That likely will not change.
1097 static int read_directory_recursive(struct dir_struct *dir,
1098 const char *base, int baselen,
1099 int check_only,
1100 const struct path_simplify *simplify)
1102 DIR *fdir;
1103 int contents = 0;
1104 struct dirent *de;
1105 struct strbuf path = STRBUF_INIT;
1107 strbuf_add(&path, base, baselen);
1109 fdir = opendir(path.len ? path.buf : ".");
1110 if (!fdir)
1111 goto out;
1113 while ((de = readdir(fdir)) != NULL) {
1114 switch (treat_path(dir, de, &path, baselen, simplify)) {
1115 case path_recurse:
1116 contents += read_directory_recursive(dir, path.buf,
1117 path.len, 0,
1118 simplify);
1119 continue;
1120 case path_ignored:
1121 continue;
1122 case path_handled:
1123 break;
1125 contents++;
1126 if (check_only)
1127 break;
1128 dir_add_name(dir, path.buf, path.len);
1130 closedir(fdir);
1131 out:
1132 strbuf_release(&path);
1134 return contents;
1137 static int cmp_name(const void *p1, const void *p2)
1139 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1140 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1142 return cache_name_compare(e1->name, e1->len,
1143 e2->name, e2->len);
1146 static struct path_simplify *create_simplify(const char **pathspec)
1148 int nr, alloc = 0;
1149 struct path_simplify *simplify = NULL;
1151 if (!pathspec)
1152 return NULL;
1154 for (nr = 0 ; ; nr++) {
1155 const char *match;
1156 if (nr >= alloc) {
1157 alloc = alloc_nr(alloc);
1158 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
1160 match = *pathspec++;
1161 if (!match)
1162 break;
1163 simplify[nr].path = match;
1164 simplify[nr].len = simple_length(match);
1166 simplify[nr].path = NULL;
1167 simplify[nr].len = 0;
1168 return simplify;
1171 static void free_simplify(struct path_simplify *simplify)
1173 free(simplify);
1176 static int treat_leading_path(struct dir_struct *dir,
1177 const char *path, int len,
1178 const struct path_simplify *simplify)
1180 struct strbuf sb = STRBUF_INIT;
1181 int baselen, rc = 0;
1182 const char *cp;
1184 while (len && path[len - 1] == '/')
1185 len--;
1186 if (!len)
1187 return 1;
1188 baselen = 0;
1189 while (1) {
1190 cp = path + baselen + !!baselen;
1191 cp = memchr(cp, '/', path + len - cp);
1192 if (!cp)
1193 baselen = len;
1194 else
1195 baselen = cp - path;
1196 strbuf_setlen(&sb, 0);
1197 strbuf_add(&sb, path, baselen);
1198 if (!is_directory(sb.buf))
1199 break;
1200 if (simplify_away(sb.buf, sb.len, simplify))
1201 break;
1202 if (treat_one_path(dir, &sb, simplify,
1203 DT_DIR, NULL) == path_ignored)
1204 break; /* do not recurse into it */
1205 if (len <= baselen) {
1206 rc = 1;
1207 break; /* finished checking */
1210 strbuf_release(&sb);
1211 return rc;
1214 int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
1216 struct path_simplify *simplify;
1218 if (has_symlink_leading_path(path, len))
1219 return dir->nr;
1221 simplify = create_simplify(pathspec);
1222 if (!len || treat_leading_path(dir, path, len, simplify))
1223 read_directory_recursive(dir, path, len, 0, simplify);
1224 free_simplify(simplify);
1225 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1226 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1227 return dir->nr;
1230 int file_exists(const char *f)
1232 struct stat sb;
1233 return lstat(f, &sb) == 0;
1237 * Given two normalized paths (a trailing slash is ok), if subdir is
1238 * outside dir, return -1. Otherwise return the offset in subdir that
1239 * can be used as relative path to dir.
1241 int dir_inside_of(const char *subdir, const char *dir)
1243 int offset = 0;
1245 assert(dir && subdir && *dir && *subdir);
1247 while (*dir && *subdir && *dir == *subdir) {
1248 dir++;
1249 subdir++;
1250 offset++;
1253 /* hel[p]/me vs hel[l]/yeah */
1254 if (*dir && *subdir)
1255 return -1;
1257 if (!*subdir)
1258 return !*dir ? offset : -1; /* same dir */
1260 /* foo/[b]ar vs foo/[] */
1261 if (is_dir_sep(dir[-1]))
1262 return is_dir_sep(subdir[-1]) ? offset : -1;
1264 /* foo[/]bar vs foo[] */
1265 return is_dir_sep(*subdir) ? offset + 1 : -1;
1268 int is_inside_dir(const char *dir)
1270 char cwd[PATH_MAX];
1271 if (!dir)
1272 return 0;
1273 if (!getcwd(cwd, sizeof(cwd)))
1274 die_errno("can't find the current directory");
1275 return dir_inside_of(cwd, dir) >= 0;
1278 int is_empty_dir(const char *path)
1280 DIR *dir = opendir(path);
1281 struct dirent *e;
1282 int ret = 1;
1284 if (!dir)
1285 return 0;
1287 while ((e = readdir(dir)) != NULL)
1288 if (!is_dot_or_dotdot(e->d_name)) {
1289 ret = 0;
1290 break;
1293 closedir(dir);
1294 return ret;
1297 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1299 DIR *dir;
1300 struct dirent *e;
1301 int ret = 0, original_len = path->len, len, kept_down = 0;
1302 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1303 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1304 unsigned char submodule_head[20];
1306 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1307 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1308 /* Do not descend and nuke a nested git work tree. */
1309 if (kept_up)
1310 *kept_up = 1;
1311 return 0;
1314 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1315 dir = opendir(path->buf);
1316 if (!dir) {
1317 /* an empty dir could be removed even if it is unreadble */
1318 if (!keep_toplevel)
1319 return rmdir(path->buf);
1320 else
1321 return -1;
1323 if (path->buf[original_len - 1] != '/')
1324 strbuf_addch(path, '/');
1326 len = path->len;
1327 while ((e = readdir(dir)) != NULL) {
1328 struct stat st;
1329 if (is_dot_or_dotdot(e->d_name))
1330 continue;
1332 strbuf_setlen(path, len);
1333 strbuf_addstr(path, e->d_name);
1334 if (lstat(path->buf, &st))
1335 ; /* fall thru */
1336 else if (S_ISDIR(st.st_mode)) {
1337 if (!remove_dir_recurse(path, flag, &kept_down))
1338 continue; /* happy */
1339 } else if (!only_empty && !unlink(path->buf))
1340 continue; /* happy, too */
1342 /* path too long, stat fails, or non-directory still exists */
1343 ret = -1;
1344 break;
1346 closedir(dir);
1348 strbuf_setlen(path, original_len);
1349 if (!ret && !keep_toplevel && !kept_down)
1350 ret = rmdir(path->buf);
1351 else if (kept_up)
1353 * report the uplevel that it is not an error that we
1354 * did not rmdir() our directory.
1356 *kept_up = !ret;
1357 return ret;
1360 int remove_dir_recursively(struct strbuf *path, int flag)
1362 return remove_dir_recurse(path, flag, NULL);
1365 void setup_standard_excludes(struct dir_struct *dir)
1367 const char *path;
1368 char *xdg_path;
1370 dir->exclude_per_dir = ".gitignore";
1371 path = git_path("info/exclude");
1372 if (!excludes_file) {
1373 home_config_paths(NULL, &xdg_path, "ignore");
1374 excludes_file = xdg_path;
1376 if (!access_or_warn(path, R_OK))
1377 add_excludes_from_file(dir, path);
1378 if (excludes_file && !access_or_warn(excludes_file, R_OK))
1379 add_excludes_from_file(dir, excludes_file);
1382 int remove_path(const char *name)
1384 char *slash;
1386 if (unlink(name) && errno != ENOENT)
1387 return -1;
1389 slash = strrchr(name, '/');
1390 if (slash) {
1391 char *dirs = xstrdup(name);
1392 slash = dirs + (slash - name);
1393 do {
1394 *slash = '\0';
1395 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1396 free(dirs);
1398 return 0;
1401 static int pathspec_item_cmp(const void *a_, const void *b_)
1403 struct pathspec_item *a, *b;
1405 a = (struct pathspec_item *)a_;
1406 b = (struct pathspec_item *)b_;
1407 return strcmp(a->match, b->match);
1410 int init_pathspec(struct pathspec *pathspec, const char **paths)
1412 const char **p = paths;
1413 int i;
1415 memset(pathspec, 0, sizeof(*pathspec));
1416 if (!p)
1417 return 0;
1418 while (*p)
1419 p++;
1420 pathspec->raw = paths;
1421 pathspec->nr = p - paths;
1422 if (!pathspec->nr)
1423 return 0;
1425 pathspec->items = xmalloc(sizeof(struct pathspec_item)*pathspec->nr);
1426 for (i = 0; i < pathspec->nr; i++) {
1427 struct pathspec_item *item = pathspec->items+i;
1428 const char *path = paths[i];
1430 item->match = path;
1431 item->len = strlen(path);
1432 item->use_wildcard = !no_wildcard(path);
1433 if (item->use_wildcard)
1434 pathspec->has_wildcard = 1;
1437 qsort(pathspec->items, pathspec->nr,
1438 sizeof(struct pathspec_item), pathspec_item_cmp);
1440 return 0;
1443 void free_pathspec(struct pathspec *pathspec)
1445 free(pathspec->items);
1446 pathspec->items = NULL;