path: convert do_git_path to take a 'struct repository'
[git/debian.git] / dir.c
blob34445374ba60b5a2e8d1c2b6e96ba7dbdb18eb4c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * See Documentation/technical/api-directory-listing.txt
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10 #define NO_THE_INDEX_COMPATIBILITY_MACROS
11 #include "cache.h"
12 #include "config.h"
13 #include "dir.h"
14 #include "attr.h"
15 #include "refs.h"
16 #include "wildmatch.h"
17 #include "pathspec.h"
18 #include "utf8.h"
19 #include "varint.h"
20 #include "ewah/ewok.h"
23 * Tells read_directory_recursive how a file or directory should be treated.
24 * Values are ordered by significance, e.g. if a directory contains both
25 * excluded and untracked files, it is listed as untracked because
26 * path_untracked > path_excluded.
28 enum path_treatment {
29 path_none = 0,
30 path_recurse,
31 path_excluded,
32 path_untracked
36 * Support data structure for our opendir/readdir/closedir wrappers
38 struct cached_dir {
39 DIR *fdir;
40 struct untracked_cache_dir *untracked;
41 int nr_files;
42 int nr_dirs;
44 struct dirent *de;
45 const char *file;
46 struct untracked_cache_dir *ucd;
49 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
50 struct index_state *istate, const char *path, int len,
51 struct untracked_cache_dir *untracked,
52 int check_only, const struct pathspec *pathspec);
53 static int get_dtype(struct dirent *de, struct index_state *istate,
54 const char *path, int len);
56 int fspathcmp(const char *a, const char *b)
58 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
61 int fspathncmp(const char *a, const char *b, size_t count)
63 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
66 int git_fnmatch(const struct pathspec_item *item,
67 const char *pattern, const char *string,
68 int prefix)
70 if (prefix > 0) {
71 if (ps_strncmp(item, pattern, string, prefix))
72 return WM_NOMATCH;
73 pattern += prefix;
74 string += prefix;
76 if (item->flags & PATHSPEC_ONESTAR) {
77 int pattern_len = strlen(++pattern);
78 int string_len = strlen(string);
79 return string_len < pattern_len ||
80 ps_strcmp(item, pattern,
81 string + string_len - pattern_len);
83 if (item->magic & PATHSPEC_GLOB)
84 return wildmatch(pattern, string,
85 WM_PATHNAME |
86 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
87 NULL);
88 else
89 /* wildmatch has not learned no FNM_PATHNAME mode yet */
90 return wildmatch(pattern, string,
91 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
92 NULL);
95 static int fnmatch_icase_mem(const char *pattern, int patternlen,
96 const char *string, int stringlen,
97 int flags)
99 int match_status;
100 struct strbuf pat_buf = STRBUF_INIT;
101 struct strbuf str_buf = STRBUF_INIT;
102 const char *use_pat = pattern;
103 const char *use_str = string;
105 if (pattern[patternlen]) {
106 strbuf_add(&pat_buf, pattern, patternlen);
107 use_pat = pat_buf.buf;
109 if (string[stringlen]) {
110 strbuf_add(&str_buf, string, stringlen);
111 use_str = str_buf.buf;
114 if (ignore_case)
115 flags |= WM_CASEFOLD;
116 match_status = wildmatch(use_pat, use_str, flags, NULL);
118 strbuf_release(&pat_buf);
119 strbuf_release(&str_buf);
121 return match_status;
124 static size_t common_prefix_len(const struct pathspec *pathspec)
126 int n;
127 size_t max = 0;
130 * ":(icase)path" is treated as a pathspec full of
131 * wildcard. In other words, only prefix is considered common
132 * prefix. If the pathspec is abc/foo abc/bar, running in
133 * subdir xyz, the common prefix is still xyz, not xuz/abc as
134 * in non-:(icase).
136 GUARD_PATHSPEC(pathspec,
137 PATHSPEC_FROMTOP |
138 PATHSPEC_MAXDEPTH |
139 PATHSPEC_LITERAL |
140 PATHSPEC_GLOB |
141 PATHSPEC_ICASE |
142 PATHSPEC_EXCLUDE |
143 PATHSPEC_ATTR);
145 for (n = 0; n < pathspec->nr; n++) {
146 size_t i = 0, len = 0, item_len;
147 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
148 continue;
149 if (pathspec->items[n].magic & PATHSPEC_ICASE)
150 item_len = pathspec->items[n].prefix;
151 else
152 item_len = pathspec->items[n].nowildcard_len;
153 while (i < item_len && (n == 0 || i < max)) {
154 char c = pathspec->items[n].match[i];
155 if (c != pathspec->items[0].match[i])
156 break;
157 if (c == '/')
158 len = i + 1;
159 i++;
161 if (n == 0 || len < max) {
162 max = len;
163 if (!max)
164 break;
167 return max;
171 * Returns a copy of the longest leading path common among all
172 * pathspecs.
174 char *common_prefix(const struct pathspec *pathspec)
176 unsigned long len = common_prefix_len(pathspec);
178 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
181 int fill_directory(struct dir_struct *dir,
182 struct index_state *istate,
183 const struct pathspec *pathspec)
185 const char *prefix;
186 size_t prefix_len;
189 * Calculate common prefix for the pathspec, and
190 * use that to optimize the directory walk
192 prefix_len = common_prefix_len(pathspec);
193 prefix = prefix_len ? pathspec->items[0].match : "";
195 /* Read the directory and prune it */
196 read_directory(dir, istate, prefix, prefix_len, pathspec);
198 return prefix_len;
201 int within_depth(const char *name, int namelen,
202 int depth, int max_depth)
204 const char *cp = name, *cpe = name + namelen;
206 while (cp < cpe) {
207 if (*cp++ != '/')
208 continue;
209 depth++;
210 if (depth > max_depth)
211 return 0;
213 return 1;
216 #define DO_MATCH_EXCLUDE (1<<0)
217 #define DO_MATCH_DIRECTORY (1<<1)
218 #define DO_MATCH_SUBMODULE (1<<2)
220 static int match_attrs(const char *name, int namelen,
221 const struct pathspec_item *item)
223 int i;
225 git_check_attr(name, item->attr_check);
226 for (i = 0; i < item->attr_match_nr; i++) {
227 const char *value;
228 int matched;
229 enum attr_match_mode match_mode;
231 value = item->attr_check->items[i].value;
232 match_mode = item->attr_match[i].match_mode;
234 if (ATTR_TRUE(value))
235 matched = (match_mode == MATCH_SET);
236 else if (ATTR_FALSE(value))
237 matched = (match_mode == MATCH_UNSET);
238 else if (ATTR_UNSET(value))
239 matched = (match_mode == MATCH_UNSPECIFIED);
240 else
241 matched = (match_mode == MATCH_VALUE &&
242 !strcmp(item->attr_match[i].value, value));
243 if (!matched)
244 return 0;
247 return 1;
251 * Does 'match' match the given name?
252 * A match is found if
254 * (1) the 'match' string is leading directory of 'name', or
255 * (2) the 'match' string is a wildcard and matches 'name', or
256 * (3) the 'match' string is exactly the same as 'name'.
258 * and the return value tells which case it was.
260 * It returns 0 when there is no match.
262 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
263 const char *name, int namelen, unsigned flags)
265 /* name/namelen has prefix cut off by caller */
266 const char *match = item->match + prefix;
267 int matchlen = item->len - prefix;
270 * The normal call pattern is:
271 * 1. prefix = common_prefix_len(ps);
272 * 2. prune something, or fill_directory
273 * 3. match_pathspec()
275 * 'prefix' at #1 may be shorter than the command's prefix and
276 * it's ok for #2 to match extra files. Those extras will be
277 * trimmed at #3.
279 * Suppose the pathspec is 'foo' and '../bar' running from
280 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
281 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
282 * user does not want XYZ/foo, only the "foo" part should be
283 * case-insensitive. We need to filter out XYZ/foo here. In
284 * other words, we do not trust the caller on comparing the
285 * prefix part when :(icase) is involved. We do exact
286 * comparison ourselves.
288 * Normally the caller (common_prefix_len() in fact) does
289 * _exact_ matching on name[-prefix+1..-1] and we do not need
290 * to check that part. Be defensive and check it anyway, in
291 * case common_prefix_len is changed, or a new caller is
292 * introduced that does not use common_prefix_len.
294 * If the penalty turns out too high when prefix is really
295 * long, maybe change it to
296 * strncmp(match, name, item->prefix - prefix)
298 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
299 strncmp(item->match, name - prefix, item->prefix))
300 return 0;
302 if (item->attr_match_nr && !match_attrs(name, namelen, item))
303 return 0;
305 /* If the match was just the prefix, we matched */
306 if (!*match)
307 return MATCHED_RECURSIVELY;
309 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
310 if (matchlen == namelen)
311 return MATCHED_EXACTLY;
313 if (match[matchlen-1] == '/' || name[matchlen] == '/')
314 return MATCHED_RECURSIVELY;
315 } else if ((flags & DO_MATCH_DIRECTORY) &&
316 match[matchlen - 1] == '/' &&
317 namelen == matchlen - 1 &&
318 !ps_strncmp(item, match, name, namelen))
319 return MATCHED_EXACTLY;
321 if (item->nowildcard_len < item->len &&
322 !git_fnmatch(item, match, name,
323 item->nowildcard_len - prefix))
324 return MATCHED_FNMATCH;
326 /* Perform checks to see if "name" is a super set of the pathspec */
327 if (flags & DO_MATCH_SUBMODULE) {
328 /* name is a literal prefix of the pathspec */
329 if ((namelen < matchlen) &&
330 (match[namelen] == '/') &&
331 !ps_strncmp(item, match, name, namelen))
332 return MATCHED_RECURSIVELY;
334 /* name" doesn't match up to the first wild character */
335 if (item->nowildcard_len < item->len &&
336 ps_strncmp(item, match, name,
337 item->nowildcard_len - prefix))
338 return 0;
341 * Here is where we would perform a wildmatch to check if
342 * "name" can be matched as a directory (or a prefix) against
343 * the pathspec. Since wildmatch doesn't have this capability
344 * at the present we have to punt and say that it is a match,
345 * potentially returning a false positive
346 * The submodules themselves will be able to perform more
347 * accurate matching to determine if the pathspec matches.
349 return MATCHED_RECURSIVELY;
352 return 0;
356 * Given a name and a list of pathspecs, returns the nature of the
357 * closest (i.e. most specific) match of the name to any of the
358 * pathspecs.
360 * The caller typically calls this multiple times with the same
361 * pathspec and seen[] array but with different name/namelen
362 * (e.g. entries from the index) and is interested in seeing if and
363 * how each pathspec matches all the names it calls this function
364 * with. A mark is left in the seen[] array for each pathspec element
365 * indicating the closest type of match that element achieved, so if
366 * seen[n] remains zero after multiple invocations, that means the nth
367 * pathspec did not match any names, which could indicate that the
368 * user mistyped the nth pathspec.
370 static int do_match_pathspec(const struct pathspec *ps,
371 const char *name, int namelen,
372 int prefix, char *seen,
373 unsigned flags)
375 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
377 GUARD_PATHSPEC(ps,
378 PATHSPEC_FROMTOP |
379 PATHSPEC_MAXDEPTH |
380 PATHSPEC_LITERAL |
381 PATHSPEC_GLOB |
382 PATHSPEC_ICASE |
383 PATHSPEC_EXCLUDE |
384 PATHSPEC_ATTR);
386 if (!ps->nr) {
387 if (!ps->recursive ||
388 !(ps->magic & PATHSPEC_MAXDEPTH) ||
389 ps->max_depth == -1)
390 return MATCHED_RECURSIVELY;
392 if (within_depth(name, namelen, 0, ps->max_depth))
393 return MATCHED_EXACTLY;
394 else
395 return 0;
398 name += prefix;
399 namelen -= prefix;
401 for (i = ps->nr - 1; i >= 0; i--) {
402 int how;
404 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
405 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
406 continue;
408 if (seen && seen[i] == MATCHED_EXACTLY)
409 continue;
411 * Make exclude patterns optional and never report
412 * "pathspec ':(exclude)foo' matches no files"
414 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
415 seen[i] = MATCHED_FNMATCH;
416 how = match_pathspec_item(ps->items+i, prefix, name,
417 namelen, flags);
418 if (ps->recursive &&
419 (ps->magic & PATHSPEC_MAXDEPTH) &&
420 ps->max_depth != -1 &&
421 how && how != MATCHED_FNMATCH) {
422 int len = ps->items[i].len;
423 if (name[len] == '/')
424 len++;
425 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
426 how = MATCHED_EXACTLY;
427 else
428 how = 0;
430 if (how) {
431 if (retval < how)
432 retval = how;
433 if (seen && seen[i] < how)
434 seen[i] = how;
437 return retval;
440 int match_pathspec(const struct pathspec *ps,
441 const char *name, int namelen,
442 int prefix, char *seen, int is_dir)
444 int positive, negative;
445 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
446 positive = do_match_pathspec(ps, name, namelen,
447 prefix, seen, flags);
448 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
449 return positive;
450 negative = do_match_pathspec(ps, name, namelen,
451 prefix, seen,
452 flags | DO_MATCH_EXCLUDE);
453 return negative ? 0 : positive;
457 * Check if a submodule is a superset of the pathspec
459 int submodule_path_match(const struct pathspec *ps,
460 const char *submodule_name,
461 char *seen)
463 int matched = do_match_pathspec(ps, submodule_name,
464 strlen(submodule_name),
465 0, seen,
466 DO_MATCH_DIRECTORY |
467 DO_MATCH_SUBMODULE);
468 return matched;
471 int report_path_error(const char *ps_matched,
472 const struct pathspec *pathspec,
473 const char *prefix)
476 * Make sure all pathspec matched; otherwise it is an error.
478 int num, errors = 0;
479 for (num = 0; num < pathspec->nr; num++) {
480 int other, found_dup;
482 if (ps_matched[num])
483 continue;
485 * The caller might have fed identical pathspec
486 * twice. Do not barf on such a mistake.
487 * FIXME: parse_pathspec should have eliminated
488 * duplicate pathspec.
490 for (found_dup = other = 0;
491 !found_dup && other < pathspec->nr;
492 other++) {
493 if (other == num || !ps_matched[other])
494 continue;
495 if (!strcmp(pathspec->items[other].original,
496 pathspec->items[num].original))
498 * Ok, we have a match already.
500 found_dup = 1;
502 if (found_dup)
503 continue;
505 error("pathspec '%s' did not match any file(s) known to git.",
506 pathspec->items[num].original);
507 errors++;
509 return errors;
513 * Return the length of the "simple" part of a path match limiter.
515 int simple_length(const char *match)
517 int len = -1;
519 for (;;) {
520 unsigned char c = *match++;
521 len++;
522 if (c == '\0' || is_glob_special(c))
523 return len;
527 int no_wildcard(const char *string)
529 return string[simple_length(string)] == '\0';
532 void parse_exclude_pattern(const char **pattern,
533 int *patternlen,
534 unsigned *flags,
535 int *nowildcardlen)
537 const char *p = *pattern;
538 size_t i, len;
540 *flags = 0;
541 if (*p == '!') {
542 *flags |= EXC_FLAG_NEGATIVE;
543 p++;
545 len = strlen(p);
546 if (len && p[len - 1] == '/') {
547 len--;
548 *flags |= EXC_FLAG_MUSTBEDIR;
550 for (i = 0; i < len; i++) {
551 if (p[i] == '/')
552 break;
554 if (i == len)
555 *flags |= EXC_FLAG_NODIR;
556 *nowildcardlen = simple_length(p);
558 * we should have excluded the trailing slash from 'p' too,
559 * but that's one more allocation. Instead just make sure
560 * nowildcardlen does not exceed real patternlen
562 if (*nowildcardlen > len)
563 *nowildcardlen = len;
564 if (*p == '*' && no_wildcard(p + 1))
565 *flags |= EXC_FLAG_ENDSWITH;
566 *pattern = p;
567 *patternlen = len;
570 void add_exclude(const char *string, const char *base,
571 int baselen, struct exclude_list *el, int srcpos)
573 struct exclude *x;
574 int patternlen;
575 unsigned flags;
576 int nowildcardlen;
578 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
579 if (flags & EXC_FLAG_MUSTBEDIR) {
580 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
581 } else {
582 x = xmalloc(sizeof(*x));
583 x->pattern = string;
585 x->patternlen = patternlen;
586 x->nowildcardlen = nowildcardlen;
587 x->base = base;
588 x->baselen = baselen;
589 x->flags = flags;
590 x->srcpos = srcpos;
591 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
592 el->excludes[el->nr++] = x;
593 x->el = el;
596 static void *read_skip_worktree_file_from_index(const struct index_state *istate,
597 const char *path, size_t *size,
598 struct sha1_stat *sha1_stat)
600 int pos, len;
601 unsigned long sz;
602 enum object_type type;
603 void *data;
605 len = strlen(path);
606 pos = index_name_pos(istate, path, len);
607 if (pos < 0)
608 return NULL;
609 if (!ce_skip_worktree(istate->cache[pos]))
610 return NULL;
611 data = read_sha1_file(istate->cache[pos]->oid.hash, &type, &sz);
612 if (!data || type != OBJ_BLOB) {
613 free(data);
614 return NULL;
616 *size = xsize_t(sz);
617 if (sha1_stat) {
618 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
619 hashcpy(sha1_stat->sha1, istate->cache[pos]->oid.hash);
621 return data;
625 * Frees memory within el which was allocated for exclude patterns and
626 * the file buffer. Does not free el itself.
628 void clear_exclude_list(struct exclude_list *el)
630 int i;
632 for (i = 0; i < el->nr; i++)
633 free(el->excludes[i]);
634 free(el->excludes);
635 free(el->filebuf);
637 memset(el, 0, sizeof(*el));
640 static void trim_trailing_spaces(char *buf)
642 char *p, *last_space = NULL;
644 for (p = buf; *p; p++)
645 switch (*p) {
646 case ' ':
647 if (!last_space)
648 last_space = p;
649 break;
650 case '\\':
651 p++;
652 if (!*p)
653 return;
654 /* fallthrough */
655 default:
656 last_space = NULL;
659 if (last_space)
660 *last_space = '\0';
664 * Given a subdirectory name and "dir" of the current directory,
665 * search the subdir in "dir" and return it, or create a new one if it
666 * does not exist in "dir".
668 * If "name" has the trailing slash, it'll be excluded in the search.
670 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
671 struct untracked_cache_dir *dir,
672 const char *name, int len)
674 int first, last;
675 struct untracked_cache_dir *d;
676 if (!dir)
677 return NULL;
678 if (len && name[len - 1] == '/')
679 len--;
680 first = 0;
681 last = dir->dirs_nr;
682 while (last > first) {
683 int cmp, next = (last + first) >> 1;
684 d = dir->dirs[next];
685 cmp = strncmp(name, d->name, len);
686 if (!cmp && strlen(d->name) > len)
687 cmp = -1;
688 if (!cmp)
689 return d;
690 if (cmp < 0) {
691 last = next;
692 continue;
694 first = next+1;
697 uc->dir_created++;
698 FLEX_ALLOC_MEM(d, name, name, len);
700 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
701 memmove(dir->dirs + first + 1, dir->dirs + first,
702 (dir->dirs_nr - first) * sizeof(*dir->dirs));
703 dir->dirs_nr++;
704 dir->dirs[first] = d;
705 return d;
708 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
710 int i;
711 dir->valid = 0;
712 dir->untracked_nr = 0;
713 for (i = 0; i < dir->dirs_nr; i++)
714 do_invalidate_gitignore(dir->dirs[i]);
717 static void invalidate_gitignore(struct untracked_cache *uc,
718 struct untracked_cache_dir *dir)
720 uc->gitignore_invalidated++;
721 do_invalidate_gitignore(dir);
724 static void invalidate_directory(struct untracked_cache *uc,
725 struct untracked_cache_dir *dir)
727 int i;
728 uc->dir_invalidated++;
729 dir->valid = 0;
730 dir->untracked_nr = 0;
731 for (i = 0; i < dir->dirs_nr; i++)
732 dir->dirs[i]->recurse = 0;
736 * Given a file with name "fname", read it (either from disk, or from
737 * an index if 'istate' is non-null), parse it and store the
738 * exclude rules in "el".
740 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
741 * stat data from disk (only valid if add_excludes returns zero). If
742 * ss_valid is non-zero, "ss" must contain good value as input.
744 static int add_excludes(const char *fname, const char *base, int baselen,
745 struct exclude_list *el,
746 struct index_state *istate,
747 struct sha1_stat *sha1_stat)
749 struct stat st;
750 int fd, i, lineno = 1;
751 size_t size = 0;
752 char *buf, *entry;
754 fd = open(fname, O_RDONLY);
755 if (fd < 0 || fstat(fd, &st) < 0) {
756 if (fd < 0)
757 warn_on_fopen_errors(fname);
758 else
759 close(fd);
760 if (!istate ||
761 (buf = read_skip_worktree_file_from_index(istate, fname, &size, sha1_stat)) == NULL)
762 return -1;
763 if (size == 0) {
764 free(buf);
765 return 0;
767 if (buf[size-1] != '\n') {
768 buf = xrealloc(buf, st_add(size, 1));
769 buf[size++] = '\n';
771 } else {
772 size = xsize_t(st.st_size);
773 if (size == 0) {
774 if (sha1_stat) {
775 fill_stat_data(&sha1_stat->stat, &st);
776 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
777 sha1_stat->valid = 1;
779 close(fd);
780 return 0;
782 buf = xmallocz(size);
783 if (read_in_full(fd, buf, size) != size) {
784 free(buf);
785 close(fd);
786 return -1;
788 buf[size++] = '\n';
789 close(fd);
790 if (sha1_stat) {
791 int pos;
792 if (sha1_stat->valid &&
793 !match_stat_data_racy(istate, &sha1_stat->stat, &st))
794 ; /* no content change, ss->sha1 still good */
795 else if (istate &&
796 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
797 !ce_stage(istate->cache[pos]) &&
798 ce_uptodate(istate->cache[pos]) &&
799 !would_convert_to_git(istate, fname))
800 hashcpy(sha1_stat->sha1,
801 istate->cache[pos]->oid.hash);
802 else
803 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
804 fill_stat_data(&sha1_stat->stat, &st);
805 sha1_stat->valid = 1;
809 el->filebuf = buf;
811 if (skip_utf8_bom(&buf, size))
812 size -= buf - el->filebuf;
814 entry = buf;
816 for (i = 0; i < size; i++) {
817 if (buf[i] == '\n') {
818 if (entry != buf + i && entry[0] != '#') {
819 buf[i - (i && buf[i-1] == '\r')] = 0;
820 trim_trailing_spaces(entry);
821 add_exclude(entry, base, baselen, el, lineno);
823 lineno++;
824 entry = buf + i + 1;
827 return 0;
830 int add_excludes_from_file_to_list(const char *fname, const char *base,
831 int baselen, struct exclude_list *el,
832 struct index_state *istate)
834 return add_excludes(fname, base, baselen, el, istate, NULL);
837 struct exclude_list *add_exclude_list(struct dir_struct *dir,
838 int group_type, const char *src)
840 struct exclude_list *el;
841 struct exclude_list_group *group;
843 group = &dir->exclude_list_group[group_type];
844 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
845 el = &group->el[group->nr++];
846 memset(el, 0, sizeof(*el));
847 el->src = src;
848 return el;
852 * Used to set up core.excludesfile and .git/info/exclude lists.
854 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
855 struct sha1_stat *sha1_stat)
857 struct exclude_list *el;
859 * catch setup_standard_excludes() that's called before
860 * dir->untracked is assigned. That function behaves
861 * differently when dir->untracked is non-NULL.
863 if (!dir->untracked)
864 dir->unmanaged_exclude_files++;
865 el = add_exclude_list(dir, EXC_FILE, fname);
866 if (add_excludes(fname, "", 0, el, NULL, sha1_stat) < 0)
867 die("cannot use %s as an exclude file", fname);
870 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
872 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
873 add_excludes_from_file_1(dir, fname, NULL);
876 int match_basename(const char *basename, int basenamelen,
877 const char *pattern, int prefix, int patternlen,
878 unsigned flags)
880 if (prefix == patternlen) {
881 if (patternlen == basenamelen &&
882 !fspathncmp(pattern, basename, basenamelen))
883 return 1;
884 } else if (flags & EXC_FLAG_ENDSWITH) {
885 /* "*literal" matching against "fooliteral" */
886 if (patternlen - 1 <= basenamelen &&
887 !fspathncmp(pattern + 1,
888 basename + basenamelen - (patternlen - 1),
889 patternlen - 1))
890 return 1;
891 } else {
892 if (fnmatch_icase_mem(pattern, patternlen,
893 basename, basenamelen,
894 0) == 0)
895 return 1;
897 return 0;
900 int match_pathname(const char *pathname, int pathlen,
901 const char *base, int baselen,
902 const char *pattern, int prefix, int patternlen,
903 unsigned flags)
905 const char *name;
906 int namelen;
909 * match with FNM_PATHNAME; the pattern has base implicitly
910 * in front of it.
912 if (*pattern == '/') {
913 pattern++;
914 patternlen--;
915 prefix--;
919 * baselen does not count the trailing slash. base[] may or
920 * may not end with a trailing slash though.
922 if (pathlen < baselen + 1 ||
923 (baselen && pathname[baselen] != '/') ||
924 fspathncmp(pathname, base, baselen))
925 return 0;
927 namelen = baselen ? pathlen - baselen - 1 : pathlen;
928 name = pathname + pathlen - namelen;
930 if (prefix) {
932 * if the non-wildcard part is longer than the
933 * remaining pathname, surely it cannot match.
935 if (prefix > namelen)
936 return 0;
938 if (fspathncmp(pattern, name, prefix))
939 return 0;
940 pattern += prefix;
941 patternlen -= prefix;
942 name += prefix;
943 namelen -= prefix;
946 * If the whole pattern did not have a wildcard,
947 * then our prefix match is all we need; we
948 * do not need to call fnmatch at all.
950 if (!patternlen && !namelen)
951 return 1;
954 return fnmatch_icase_mem(pattern, patternlen,
955 name, namelen,
956 WM_PATHNAME) == 0;
960 * Scan the given exclude list in reverse to see whether pathname
961 * should be ignored. The first match (i.e. the last on the list), if
962 * any, determines the fate. Returns the exclude_list element which
963 * matched, or NULL for undecided.
965 static struct exclude *last_exclude_matching_from_list(const char *pathname,
966 int pathlen,
967 const char *basename,
968 int *dtype,
969 struct exclude_list *el,
970 struct index_state *istate)
972 struct exclude *exc = NULL; /* undecided */
973 int i;
975 if (!el->nr)
976 return NULL; /* undefined */
978 for (i = el->nr - 1; 0 <= i; i--) {
979 struct exclude *x = el->excludes[i];
980 const char *exclude = x->pattern;
981 int prefix = x->nowildcardlen;
983 if (x->flags & EXC_FLAG_MUSTBEDIR) {
984 if (*dtype == DT_UNKNOWN)
985 *dtype = get_dtype(NULL, istate, pathname, pathlen);
986 if (*dtype != DT_DIR)
987 continue;
990 if (x->flags & EXC_FLAG_NODIR) {
991 if (match_basename(basename,
992 pathlen - (basename - pathname),
993 exclude, prefix, x->patternlen,
994 x->flags)) {
995 exc = x;
996 break;
998 continue;
1001 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1002 if (match_pathname(pathname, pathlen,
1003 x->base, x->baselen ? x->baselen - 1 : 0,
1004 exclude, prefix, x->patternlen, x->flags)) {
1005 exc = x;
1006 break;
1009 return exc;
1013 * Scan the list and let the last match determine the fate.
1014 * Return 1 for exclude, 0 for include and -1 for undecided.
1016 int is_excluded_from_list(const char *pathname,
1017 int pathlen, const char *basename, int *dtype,
1018 struct exclude_list *el, struct index_state *istate)
1020 struct exclude *exclude;
1021 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1022 dtype, el, istate);
1023 if (exclude)
1024 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1025 return -1; /* undecided */
1028 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1029 struct index_state *istate,
1030 const char *pathname, int pathlen, const char *basename,
1031 int *dtype_p)
1033 int i, j;
1034 struct exclude_list_group *group;
1035 struct exclude *exclude;
1036 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1037 group = &dir->exclude_list_group[i];
1038 for (j = group->nr - 1; j >= 0; j--) {
1039 exclude = last_exclude_matching_from_list(
1040 pathname, pathlen, basename, dtype_p,
1041 &group->el[j], istate);
1042 if (exclude)
1043 return exclude;
1046 return NULL;
1050 * Loads the per-directory exclude list for the substring of base
1051 * which has a char length of baselen.
1053 static void prep_exclude(struct dir_struct *dir,
1054 struct index_state *istate,
1055 const char *base, int baselen)
1057 struct exclude_list_group *group;
1058 struct exclude_list *el;
1059 struct exclude_stack *stk = NULL;
1060 struct untracked_cache_dir *untracked;
1061 int current;
1063 group = &dir->exclude_list_group[EXC_DIRS];
1066 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1067 * which originate from directories not in the prefix of the
1068 * path being checked.
1070 while ((stk = dir->exclude_stack) != NULL) {
1071 if (stk->baselen <= baselen &&
1072 !strncmp(dir->basebuf.buf, base, stk->baselen))
1073 break;
1074 el = &group->el[dir->exclude_stack->exclude_ix];
1075 dir->exclude_stack = stk->prev;
1076 dir->exclude = NULL;
1077 free((char *)el->src); /* see strbuf_detach() below */
1078 clear_exclude_list(el);
1079 free(stk);
1080 group->nr--;
1083 /* Skip traversing into sub directories if the parent is excluded */
1084 if (dir->exclude)
1085 return;
1088 * Lazy initialization. All call sites currently just
1089 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1090 * them seems lots of work for little benefit.
1092 if (!dir->basebuf.buf)
1093 strbuf_init(&dir->basebuf, PATH_MAX);
1095 /* Read from the parent directories and push them down. */
1096 current = stk ? stk->baselen : -1;
1097 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1098 if (dir->untracked)
1099 untracked = stk ? stk->ucd : dir->untracked->root;
1100 else
1101 untracked = NULL;
1103 while (current < baselen) {
1104 const char *cp;
1105 struct sha1_stat sha1_stat;
1107 stk = xcalloc(1, sizeof(*stk));
1108 if (current < 0) {
1109 cp = base;
1110 current = 0;
1111 } else {
1112 cp = strchr(base + current + 1, '/');
1113 if (!cp)
1114 die("oops in prep_exclude");
1115 cp++;
1116 untracked =
1117 lookup_untracked(dir->untracked, untracked,
1118 base + current,
1119 cp - base - current);
1121 stk->prev = dir->exclude_stack;
1122 stk->baselen = cp - base;
1123 stk->exclude_ix = group->nr;
1124 stk->ucd = untracked;
1125 el = add_exclude_list(dir, EXC_DIRS, NULL);
1126 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1127 assert(stk->baselen == dir->basebuf.len);
1129 /* Abort if the directory is excluded */
1130 if (stk->baselen) {
1131 int dt = DT_DIR;
1132 dir->basebuf.buf[stk->baselen - 1] = 0;
1133 dir->exclude = last_exclude_matching_from_lists(dir,
1134 istate,
1135 dir->basebuf.buf, stk->baselen - 1,
1136 dir->basebuf.buf + current, &dt);
1137 dir->basebuf.buf[stk->baselen - 1] = '/';
1138 if (dir->exclude &&
1139 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1140 dir->exclude = NULL;
1141 if (dir->exclude) {
1142 dir->exclude_stack = stk;
1143 return;
1147 /* Try to read per-directory file */
1148 hashclr(sha1_stat.sha1);
1149 sha1_stat.valid = 0;
1150 if (dir->exclude_per_dir &&
1152 * If we know that no files have been added in
1153 * this directory (i.e. valid_cached_dir() has
1154 * been executed and set untracked->valid) ..
1156 (!untracked || !untracked->valid ||
1158 * .. and .gitignore does not exist before
1159 * (i.e. null exclude_sha1). Then we can skip
1160 * loading .gitignore, which would result in
1161 * ENOENT anyway.
1163 !is_null_sha1(untracked->exclude_sha1))) {
1165 * dir->basebuf gets reused by the traversal, but we
1166 * need fname to remain unchanged to ensure the src
1167 * member of each struct exclude correctly
1168 * back-references its source file. Other invocations
1169 * of add_exclude_list provide stable strings, so we
1170 * strbuf_detach() and free() here in the caller.
1172 struct strbuf sb = STRBUF_INIT;
1173 strbuf_addbuf(&sb, &dir->basebuf);
1174 strbuf_addstr(&sb, dir->exclude_per_dir);
1175 el->src = strbuf_detach(&sb, NULL);
1176 add_excludes(el->src, el->src, stk->baselen, el, istate,
1177 untracked ? &sha1_stat : NULL);
1180 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1181 * will first be called in valid_cached_dir() then maybe many
1182 * times more in last_exclude_matching(). When the cache is
1183 * used, last_exclude_matching() will not be called and
1184 * reading .gitignore content will be a waste.
1186 * So when it's called by valid_cached_dir() and we can get
1187 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1188 * modified on work tree), we could delay reading the
1189 * .gitignore content until we absolutely need it in
1190 * last_exclude_matching(). Be careful about ignore rule
1191 * order, though, if you do that.
1193 if (untracked &&
1194 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1195 invalidate_gitignore(dir->untracked, untracked);
1196 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1198 dir->exclude_stack = stk;
1199 current = stk->baselen;
1201 strbuf_setlen(&dir->basebuf, baselen);
1205 * Loads the exclude lists for the directory containing pathname, then
1206 * scans all exclude lists to determine whether pathname is excluded.
1207 * Returns the exclude_list element which matched, or NULL for
1208 * undecided.
1210 struct exclude *last_exclude_matching(struct dir_struct *dir,
1211 struct index_state *istate,
1212 const char *pathname,
1213 int *dtype_p)
1215 int pathlen = strlen(pathname);
1216 const char *basename = strrchr(pathname, '/');
1217 basename = (basename) ? basename+1 : pathname;
1219 prep_exclude(dir, istate, pathname, basename-pathname);
1221 if (dir->exclude)
1222 return dir->exclude;
1224 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
1225 basename, dtype_p);
1229 * Loads the exclude lists for the directory containing pathname, then
1230 * scans all exclude lists to determine whether pathname is excluded.
1231 * Returns 1 if true, otherwise 0.
1233 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1234 const char *pathname, int *dtype_p)
1236 struct exclude *exclude =
1237 last_exclude_matching(dir, istate, pathname, dtype_p);
1238 if (exclude)
1239 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1240 return 0;
1243 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1245 struct dir_entry *ent;
1247 FLEX_ALLOC_MEM(ent, name, pathname, len);
1248 ent->len = len;
1249 return ent;
1252 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1253 struct index_state *istate,
1254 const char *pathname, int len)
1256 if (index_file_exists(istate, pathname, len, ignore_case))
1257 return NULL;
1259 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1260 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1263 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1264 struct index_state *istate,
1265 const char *pathname, int len)
1267 if (!index_name_is_other(istate, pathname, len))
1268 return NULL;
1270 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1271 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1274 enum exist_status {
1275 index_nonexistent = 0,
1276 index_directory,
1277 index_gitdir
1281 * Do not use the alphabetically sorted index to look up
1282 * the directory name; instead, use the case insensitive
1283 * directory hash.
1285 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1286 const char *dirname, int len)
1288 struct cache_entry *ce;
1290 if (index_dir_exists(istate, dirname, len))
1291 return index_directory;
1293 ce = index_file_exists(istate, dirname, len, ignore_case);
1294 if (ce && S_ISGITLINK(ce->ce_mode))
1295 return index_gitdir;
1297 return index_nonexistent;
1301 * The index sorts alphabetically by entry name, which
1302 * means that a gitlink sorts as '\0' at the end, while
1303 * a directory (which is defined not as an entry, but as
1304 * the files it contains) will sort with the '/' at the
1305 * end.
1307 static enum exist_status directory_exists_in_index(struct index_state *istate,
1308 const char *dirname, int len)
1310 int pos;
1312 if (ignore_case)
1313 return directory_exists_in_index_icase(istate, dirname, len);
1315 pos = index_name_pos(istate, dirname, len);
1316 if (pos < 0)
1317 pos = -pos-1;
1318 while (pos < istate->cache_nr) {
1319 const struct cache_entry *ce = istate->cache[pos++];
1320 unsigned char endchar;
1322 if (strncmp(ce->name, dirname, len))
1323 break;
1324 endchar = ce->name[len];
1325 if (endchar > '/')
1326 break;
1327 if (endchar == '/')
1328 return index_directory;
1329 if (!endchar && S_ISGITLINK(ce->ce_mode))
1330 return index_gitdir;
1332 return index_nonexistent;
1336 * When we find a directory when traversing the filesystem, we
1337 * have three distinct cases:
1339 * - ignore it
1340 * - see it as a directory
1341 * - recurse into it
1343 * and which one we choose depends on a combination of existing
1344 * git index contents and the flags passed into the directory
1345 * traversal routine.
1347 * Case 1: If we *already* have entries in the index under that
1348 * directory name, we always recurse into the directory to see
1349 * all the files.
1351 * Case 2: If we *already* have that directory name as a gitlink,
1352 * we always continue to see it as a gitlink, regardless of whether
1353 * there is an actual git directory there or not (it might not
1354 * be checked out as a subproject!)
1356 * Case 3: if we didn't have it in the index previously, we
1357 * have a few sub-cases:
1359 * (a) if "show_other_directories" is true, we show it as
1360 * just a directory, unless "hide_empty_directories" is
1361 * also true, in which case we need to check if it contains any
1362 * untracked and / or ignored files.
1363 * (b) if it looks like a git directory, and we don't have
1364 * 'no_gitlinks' set we treat it as a gitlink, and show it
1365 * as a directory.
1366 * (c) otherwise, we recurse into it.
1368 static enum path_treatment treat_directory(struct dir_struct *dir,
1369 struct index_state *istate,
1370 struct untracked_cache_dir *untracked,
1371 const char *dirname, int len, int baselen, int exclude,
1372 const struct pathspec *pathspec)
1374 /* The "len-1" is to strip the final '/' */
1375 switch (directory_exists_in_index(istate, dirname, len-1)) {
1376 case index_directory:
1377 return path_recurse;
1379 case index_gitdir:
1380 return path_none;
1382 case index_nonexistent:
1383 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1384 break;
1385 if (!(dir->flags & DIR_NO_GITLINKS)) {
1386 unsigned char sha1[20];
1387 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1388 return path_untracked;
1390 return path_recurse;
1393 /* This is the "show_other_directories" case */
1395 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1396 return exclude ? path_excluded : path_untracked;
1398 untracked = lookup_untracked(dir->untracked, untracked,
1399 dirname + baselen, len - baselen);
1400 return read_directory_recursive(dir, istate, dirname, len,
1401 untracked, 1, pathspec);
1405 * This is an inexact early pruning of any recursive directory
1406 * reading - if the path cannot possibly be in the pathspec,
1407 * return true, and we'll skip it early.
1409 static int simplify_away(const char *path, int pathlen,
1410 const struct pathspec *pathspec)
1412 int i;
1414 if (!pathspec || !pathspec->nr)
1415 return 0;
1417 GUARD_PATHSPEC(pathspec,
1418 PATHSPEC_FROMTOP |
1419 PATHSPEC_MAXDEPTH |
1420 PATHSPEC_LITERAL |
1421 PATHSPEC_GLOB |
1422 PATHSPEC_ICASE |
1423 PATHSPEC_EXCLUDE |
1424 PATHSPEC_ATTR);
1426 for (i = 0; i < pathspec->nr; i++) {
1427 const struct pathspec_item *item = &pathspec->items[i];
1428 int len = item->nowildcard_len;
1430 if (len > pathlen)
1431 len = pathlen;
1432 if (!ps_strncmp(item, item->match, path, len))
1433 return 0;
1436 return 1;
1440 * This function tells us whether an excluded path matches a
1441 * list of "interesting" pathspecs. That is, whether a path matched
1442 * by any of the pathspecs could possibly be ignored by excluding
1443 * the specified path. This can happen if:
1445 * 1. the path is mentioned explicitly in the pathspec
1447 * 2. the path is a directory prefix of some element in the
1448 * pathspec
1450 static int exclude_matches_pathspec(const char *path, int pathlen,
1451 const struct pathspec *pathspec)
1453 int i;
1455 if (!pathspec || !pathspec->nr)
1456 return 0;
1458 GUARD_PATHSPEC(pathspec,
1459 PATHSPEC_FROMTOP |
1460 PATHSPEC_MAXDEPTH |
1461 PATHSPEC_LITERAL |
1462 PATHSPEC_GLOB |
1463 PATHSPEC_ICASE |
1464 PATHSPEC_EXCLUDE);
1466 for (i = 0; i < pathspec->nr; i++) {
1467 const struct pathspec_item *item = &pathspec->items[i];
1468 int len = item->nowildcard_len;
1470 if (len == pathlen &&
1471 !ps_strncmp(item, item->match, path, pathlen))
1472 return 1;
1473 if (len > pathlen &&
1474 item->match[pathlen] == '/' &&
1475 !ps_strncmp(item, item->match, path, pathlen))
1476 return 1;
1478 return 0;
1481 static int get_index_dtype(struct index_state *istate,
1482 const char *path, int len)
1484 int pos;
1485 const struct cache_entry *ce;
1487 ce = index_file_exists(istate, path, len, 0);
1488 if (ce) {
1489 if (!ce_uptodate(ce))
1490 return DT_UNKNOWN;
1491 if (S_ISGITLINK(ce->ce_mode))
1492 return DT_DIR;
1494 * Nobody actually cares about the
1495 * difference between DT_LNK and DT_REG
1497 return DT_REG;
1500 /* Try to look it up as a directory */
1501 pos = index_name_pos(istate, path, len);
1502 if (pos >= 0)
1503 return DT_UNKNOWN;
1504 pos = -pos-1;
1505 while (pos < istate->cache_nr) {
1506 ce = istate->cache[pos++];
1507 if (strncmp(ce->name, path, len))
1508 break;
1509 if (ce->name[len] > '/')
1510 break;
1511 if (ce->name[len] < '/')
1512 continue;
1513 if (!ce_uptodate(ce))
1514 break; /* continue? */
1515 return DT_DIR;
1517 return DT_UNKNOWN;
1520 static int get_dtype(struct dirent *de, struct index_state *istate,
1521 const char *path, int len)
1523 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1524 struct stat st;
1526 if (dtype != DT_UNKNOWN)
1527 return dtype;
1528 dtype = get_index_dtype(istate, path, len);
1529 if (dtype != DT_UNKNOWN)
1530 return dtype;
1531 if (lstat(path, &st))
1532 return dtype;
1533 if (S_ISREG(st.st_mode))
1534 return DT_REG;
1535 if (S_ISDIR(st.st_mode))
1536 return DT_DIR;
1537 if (S_ISLNK(st.st_mode))
1538 return DT_LNK;
1539 return dtype;
1542 static enum path_treatment treat_one_path(struct dir_struct *dir,
1543 struct untracked_cache_dir *untracked,
1544 struct index_state *istate,
1545 struct strbuf *path,
1546 int baselen,
1547 const struct pathspec *pathspec,
1548 int dtype, struct dirent *de)
1550 int exclude;
1551 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1553 if (dtype == DT_UNKNOWN)
1554 dtype = get_dtype(de, istate, path->buf, path->len);
1556 /* Always exclude indexed files */
1557 if (dtype != DT_DIR && has_path_in_index)
1558 return path_none;
1561 * When we are looking at a directory P in the working tree,
1562 * there are three cases:
1564 * (1) P exists in the index. Everything inside the directory P in
1565 * the working tree needs to go when P is checked out from the
1566 * index.
1568 * (2) P does not exist in the index, but there is P/Q in the index.
1569 * We know P will stay a directory when we check out the contents
1570 * of the index, but we do not know yet if there is a directory
1571 * P/Q in the working tree to be killed, so we need to recurse.
1573 * (3) P does not exist in the index, and there is no P/Q in the index
1574 * to require P to be a directory, either. Only in this case, we
1575 * know that everything inside P will not be killed without
1576 * recursing.
1578 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1579 (dtype == DT_DIR) &&
1580 !has_path_in_index &&
1581 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1582 return path_none;
1584 exclude = is_excluded(dir, istate, path->buf, &dtype);
1587 * Excluded? If we don't explicitly want to show
1588 * ignored files, ignore it
1590 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1591 return path_excluded;
1593 switch (dtype) {
1594 default:
1595 return path_none;
1596 case DT_DIR:
1597 strbuf_addch(path, '/');
1598 return treat_directory(dir, istate, untracked, path->buf, path->len,
1599 baselen, exclude, pathspec);
1600 case DT_REG:
1601 case DT_LNK:
1602 return exclude ? path_excluded : path_untracked;
1606 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1607 struct untracked_cache_dir *untracked,
1608 struct cached_dir *cdir,
1609 struct index_state *istate,
1610 struct strbuf *path,
1611 int baselen,
1612 const struct pathspec *pathspec)
1614 strbuf_setlen(path, baselen);
1615 if (!cdir->ucd) {
1616 strbuf_addstr(path, cdir->file);
1617 return path_untracked;
1619 strbuf_addstr(path, cdir->ucd->name);
1620 /* treat_one_path() does this before it calls treat_directory() */
1621 strbuf_complete(path, '/');
1622 if (cdir->ucd->check_only)
1624 * check_only is set as a result of treat_directory() getting
1625 * to its bottom. Verify again the same set of directories
1626 * with check_only set.
1628 return read_directory_recursive(dir, istate, path->buf, path->len,
1629 cdir->ucd, 1, pathspec);
1631 * We get path_recurse in the first run when
1632 * directory_exists_in_index() returns index_nonexistent. We
1633 * are sure that new changes in the index does not impact the
1634 * outcome. Return now.
1636 return path_recurse;
1639 static enum path_treatment treat_path(struct dir_struct *dir,
1640 struct untracked_cache_dir *untracked,
1641 struct cached_dir *cdir,
1642 struct index_state *istate,
1643 struct strbuf *path,
1644 int baselen,
1645 const struct pathspec *pathspec)
1647 int dtype;
1648 struct dirent *de = cdir->de;
1650 if (!de)
1651 return treat_path_fast(dir, untracked, cdir, istate, path,
1652 baselen, pathspec);
1653 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1654 return path_none;
1655 strbuf_setlen(path, baselen);
1656 strbuf_addstr(path, de->d_name);
1657 if (simplify_away(path->buf, path->len, pathspec))
1658 return path_none;
1660 dtype = DTYPE(de);
1661 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1664 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1666 if (!dir)
1667 return;
1668 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1669 dir->untracked_alloc);
1670 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1673 static int valid_cached_dir(struct dir_struct *dir,
1674 struct untracked_cache_dir *untracked,
1675 struct index_state *istate,
1676 struct strbuf *path,
1677 int check_only)
1679 struct stat st;
1681 if (!untracked)
1682 return 0;
1684 if (stat(path->len ? path->buf : ".", &st)) {
1685 invalidate_directory(dir->untracked, untracked);
1686 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1687 return 0;
1689 if (!untracked->valid ||
1690 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1691 if (untracked->valid)
1692 invalidate_directory(dir->untracked, untracked);
1693 fill_stat_data(&untracked->stat_data, &st);
1694 return 0;
1697 if (untracked->check_only != !!check_only) {
1698 invalidate_directory(dir->untracked, untracked);
1699 return 0;
1703 * prep_exclude will be called eventually on this directory,
1704 * but it's called much later in last_exclude_matching(). We
1705 * need it now to determine the validity of the cache for this
1706 * path. The next calls will be nearly no-op, the way
1707 * prep_exclude() is designed.
1709 if (path->len && path->buf[path->len - 1] != '/') {
1710 strbuf_addch(path, '/');
1711 prep_exclude(dir, istate, path->buf, path->len);
1712 strbuf_setlen(path, path->len - 1);
1713 } else
1714 prep_exclude(dir, istate, path->buf, path->len);
1716 /* hopefully prep_exclude() haven't invalidated this entry... */
1717 return untracked->valid;
1720 static int open_cached_dir(struct cached_dir *cdir,
1721 struct dir_struct *dir,
1722 struct untracked_cache_dir *untracked,
1723 struct index_state *istate,
1724 struct strbuf *path,
1725 int check_only)
1727 memset(cdir, 0, sizeof(*cdir));
1728 cdir->untracked = untracked;
1729 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1730 return 0;
1731 cdir->fdir = opendir(path->len ? path->buf : ".");
1732 if (dir->untracked)
1733 dir->untracked->dir_opened++;
1734 if (!cdir->fdir)
1735 return -1;
1736 return 0;
1739 static int read_cached_dir(struct cached_dir *cdir)
1741 if (cdir->fdir) {
1742 cdir->de = readdir(cdir->fdir);
1743 if (!cdir->de)
1744 return -1;
1745 return 0;
1747 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1748 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1749 if (!d->recurse) {
1750 cdir->nr_dirs++;
1751 continue;
1753 cdir->ucd = d;
1754 cdir->nr_dirs++;
1755 return 0;
1757 cdir->ucd = NULL;
1758 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1759 struct untracked_cache_dir *d = cdir->untracked;
1760 cdir->file = d->untracked[cdir->nr_files++];
1761 return 0;
1763 return -1;
1766 static void close_cached_dir(struct cached_dir *cdir)
1768 if (cdir->fdir)
1769 closedir(cdir->fdir);
1771 * We have gone through this directory and found no untracked
1772 * entries. Mark it valid.
1774 if (cdir->untracked) {
1775 cdir->untracked->valid = 1;
1776 cdir->untracked->recurse = 1;
1781 * Read a directory tree. We currently ignore anything but
1782 * directories, regular files and symlinks. That's because git
1783 * doesn't handle them at all yet. Maybe that will change some
1784 * day.
1786 * Also, we ignore the name ".git" (even if it is not a directory).
1787 * That likely will not change.
1789 * Returns the most significant path_treatment value encountered in the scan.
1791 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1792 struct index_state *istate, const char *base, int baselen,
1793 struct untracked_cache_dir *untracked, int check_only,
1794 const struct pathspec *pathspec)
1796 struct cached_dir cdir;
1797 enum path_treatment state, subdir_state, dir_state = path_none;
1798 struct strbuf path = STRBUF_INIT;
1800 strbuf_add(&path, base, baselen);
1802 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1803 goto out;
1805 if (untracked)
1806 untracked->check_only = !!check_only;
1808 while (!read_cached_dir(&cdir)) {
1809 /* check how the file or directory should be treated */
1810 state = treat_path(dir, untracked, &cdir, istate, &path,
1811 baselen, pathspec);
1813 if (state > dir_state)
1814 dir_state = state;
1816 /* recurse into subdir if instructed by treat_path */
1817 if ((state == path_recurse) ||
1818 ((state == path_untracked) &&
1819 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1820 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
1821 struct untracked_cache_dir *ud;
1822 ud = lookup_untracked(dir->untracked, untracked,
1823 path.buf + baselen,
1824 path.len - baselen);
1825 subdir_state =
1826 read_directory_recursive(dir, istate, path.buf,
1827 path.len, ud,
1828 check_only, pathspec);
1829 if (subdir_state > dir_state)
1830 dir_state = subdir_state;
1833 if (check_only) {
1834 /* abort early if maximum state has been reached */
1835 if (dir_state == path_untracked) {
1836 if (cdir.fdir)
1837 add_untracked(untracked, path.buf + baselen);
1838 break;
1840 /* skip the dir_add_* part */
1841 continue;
1844 /* add the path to the appropriate result list */
1845 switch (state) {
1846 case path_excluded:
1847 if (dir->flags & DIR_SHOW_IGNORED)
1848 dir_add_name(dir, istate, path.buf, path.len);
1849 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1850 ((dir->flags & DIR_COLLECT_IGNORED) &&
1851 exclude_matches_pathspec(path.buf, path.len,
1852 pathspec)))
1853 dir_add_ignored(dir, istate, path.buf, path.len);
1854 break;
1856 case path_untracked:
1857 if (dir->flags & DIR_SHOW_IGNORED)
1858 break;
1859 dir_add_name(dir, istate, path.buf, path.len);
1860 if (cdir.fdir)
1861 add_untracked(untracked, path.buf + baselen);
1862 break;
1864 default:
1865 break;
1868 close_cached_dir(&cdir);
1869 out:
1870 strbuf_release(&path);
1872 return dir_state;
1875 int cmp_dir_entry(const void *p1, const void *p2)
1877 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1878 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1880 return name_compare(e1->name, e1->len, e2->name, e2->len);
1883 /* check if *out lexically strictly contains *in */
1884 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
1886 return (out->len < in->len) &&
1887 (out->name[out->len - 1] == '/') &&
1888 !memcmp(out->name, in->name, out->len);
1891 static int treat_leading_path(struct dir_struct *dir,
1892 struct index_state *istate,
1893 const char *path, int len,
1894 const struct pathspec *pathspec)
1896 struct strbuf sb = STRBUF_INIT;
1897 int baselen, rc = 0;
1898 const char *cp;
1899 int old_flags = dir->flags;
1901 while (len && path[len - 1] == '/')
1902 len--;
1903 if (!len)
1904 return 1;
1905 baselen = 0;
1906 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1907 while (1) {
1908 cp = path + baselen + !!baselen;
1909 cp = memchr(cp, '/', path + len - cp);
1910 if (!cp)
1911 baselen = len;
1912 else
1913 baselen = cp - path;
1914 strbuf_setlen(&sb, 0);
1915 strbuf_add(&sb, path, baselen);
1916 if (!is_directory(sb.buf))
1917 break;
1918 if (simplify_away(sb.buf, sb.len, pathspec))
1919 break;
1920 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
1921 DT_DIR, NULL) == path_none)
1922 break; /* do not recurse into it */
1923 if (len <= baselen) {
1924 rc = 1;
1925 break; /* finished checking */
1928 strbuf_release(&sb);
1929 dir->flags = old_flags;
1930 return rc;
1933 static const char *get_ident_string(void)
1935 static struct strbuf sb = STRBUF_INIT;
1936 struct utsname uts;
1938 if (sb.len)
1939 return sb.buf;
1940 if (uname(&uts) < 0)
1941 die_errno(_("failed to get kernel name and information"));
1942 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
1943 uts.sysname);
1944 return sb.buf;
1947 static int ident_in_untracked(const struct untracked_cache *uc)
1950 * Previous git versions may have saved many NUL separated
1951 * strings in the "ident" field, but it is insane to manage
1952 * many locations, so just take care of the first one.
1955 return !strcmp(uc->ident.buf, get_ident_string());
1958 static void set_untracked_ident(struct untracked_cache *uc)
1960 strbuf_reset(&uc->ident);
1961 strbuf_addstr(&uc->ident, get_ident_string());
1964 * This strbuf used to contain a list of NUL separated
1965 * strings, so save NUL too for backward compatibility.
1967 strbuf_addch(&uc->ident, 0);
1970 static void new_untracked_cache(struct index_state *istate)
1972 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
1973 strbuf_init(&uc->ident, 100);
1974 uc->exclude_per_dir = ".gitignore";
1975 /* should be the same flags used by git-status */
1976 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
1977 set_untracked_ident(uc);
1978 istate->untracked = uc;
1979 istate->cache_changed |= UNTRACKED_CHANGED;
1982 void add_untracked_cache(struct index_state *istate)
1984 if (!istate->untracked) {
1985 new_untracked_cache(istate);
1986 } else {
1987 if (!ident_in_untracked(istate->untracked)) {
1988 free_untracked_cache(istate->untracked);
1989 new_untracked_cache(istate);
1994 void remove_untracked_cache(struct index_state *istate)
1996 if (istate->untracked) {
1997 free_untracked_cache(istate->untracked);
1998 istate->untracked = NULL;
1999 istate->cache_changed |= UNTRACKED_CHANGED;
2003 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2004 int base_len,
2005 const struct pathspec *pathspec)
2007 struct untracked_cache_dir *root;
2009 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
2010 return NULL;
2013 * We only support $GIT_DIR/info/exclude and core.excludesfile
2014 * as the global ignore rule files. Any other additions
2015 * (e.g. from command line) invalidate the cache. This
2016 * condition also catches running setup_standard_excludes()
2017 * before setting dir->untracked!
2019 if (dir->unmanaged_exclude_files)
2020 return NULL;
2023 * Optimize for the main use case only: whole-tree git
2024 * status. More work involved in treat_leading_path() if we
2025 * use cache on just a subset of the worktree. pathspec
2026 * support could make the matter even worse.
2028 if (base_len || (pathspec && pathspec->nr))
2029 return NULL;
2031 /* Different set of flags may produce different results */
2032 if (dir->flags != dir->untracked->dir_flags ||
2034 * See treat_directory(), case index_nonexistent. Without
2035 * this flag, we may need to also cache .git file content
2036 * for the resolve_gitlink_ref() call, which we don't.
2038 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2039 /* We don't support collecting ignore files */
2040 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2041 DIR_COLLECT_IGNORED)))
2042 return NULL;
2045 * If we use .gitignore in the cache and now you change it to
2046 * .gitexclude, everything will go wrong.
2048 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2049 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2050 return NULL;
2053 * EXC_CMDL is not considered in the cache. If people set it,
2054 * skip the cache.
2056 if (dir->exclude_list_group[EXC_CMDL].nr)
2057 return NULL;
2059 if (!ident_in_untracked(dir->untracked)) {
2060 warning(_("Untracked cache is disabled on this system or location."));
2061 return NULL;
2064 if (!dir->untracked->root) {
2065 const int len = sizeof(*dir->untracked->root);
2066 dir->untracked->root = xmalloc(len);
2067 memset(dir->untracked->root, 0, len);
2070 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2071 root = dir->untracked->root;
2072 if (hashcmp(dir->ss_info_exclude.sha1,
2073 dir->untracked->ss_info_exclude.sha1)) {
2074 invalidate_gitignore(dir->untracked, root);
2075 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2077 if (hashcmp(dir->ss_excludes_file.sha1,
2078 dir->untracked->ss_excludes_file.sha1)) {
2079 invalidate_gitignore(dir->untracked, root);
2080 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2083 /* Make sure this directory is not dropped out at saving phase */
2084 root->recurse = 1;
2085 return root;
2088 int read_directory(struct dir_struct *dir, struct index_state *istate,
2089 const char *path, int len, const struct pathspec *pathspec)
2091 struct untracked_cache_dir *untracked;
2093 if (has_symlink_leading_path(path, len))
2094 return dir->nr;
2096 untracked = validate_untracked_cache(dir, len, pathspec);
2097 if (!untracked)
2099 * make sure untracked cache code path is disabled,
2100 * e.g. prep_exclude()
2102 dir->untracked = NULL;
2103 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2104 read_directory_recursive(dir, istate, path, len, untracked, 0, pathspec);
2105 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2106 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2109 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2110 * also pick up untracked contents of untracked dirs; by default
2111 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2113 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2114 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2115 int i, j;
2117 /* remove from dir->entries untracked contents of untracked dirs */
2118 for (i = j = 0; j < dir->nr; j++) {
2119 if (i &&
2120 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2121 free(dir->entries[j]);
2122 dir->entries[j] = NULL;
2123 } else {
2124 dir->entries[i++] = dir->entries[j];
2128 dir->nr = i;
2131 if (dir->untracked) {
2132 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2133 trace_printf_key(&trace_untracked_stats,
2134 "node creation: %u\n"
2135 "gitignore invalidation: %u\n"
2136 "directory invalidation: %u\n"
2137 "opendir: %u\n",
2138 dir->untracked->dir_created,
2139 dir->untracked->gitignore_invalidated,
2140 dir->untracked->dir_invalidated,
2141 dir->untracked->dir_opened);
2142 if (dir->untracked == istate->untracked &&
2143 (dir->untracked->dir_opened ||
2144 dir->untracked->gitignore_invalidated ||
2145 dir->untracked->dir_invalidated))
2146 istate->cache_changed |= UNTRACKED_CHANGED;
2147 if (dir->untracked != istate->untracked) {
2148 free(dir->untracked);
2149 dir->untracked = NULL;
2152 return dir->nr;
2155 int file_exists(const char *f)
2157 struct stat sb;
2158 return lstat(f, &sb) == 0;
2161 static int cmp_icase(char a, char b)
2163 if (a == b)
2164 return 0;
2165 if (ignore_case)
2166 return toupper(a) - toupper(b);
2167 return a - b;
2171 * Given two normalized paths (a trailing slash is ok), if subdir is
2172 * outside dir, return -1. Otherwise return the offset in subdir that
2173 * can be used as relative path to dir.
2175 int dir_inside_of(const char *subdir, const char *dir)
2177 int offset = 0;
2179 assert(dir && subdir && *dir && *subdir);
2181 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2182 dir++;
2183 subdir++;
2184 offset++;
2187 /* hel[p]/me vs hel[l]/yeah */
2188 if (*dir && *subdir)
2189 return -1;
2191 if (!*subdir)
2192 return !*dir ? offset : -1; /* same dir */
2194 /* foo/[b]ar vs foo/[] */
2195 if (is_dir_sep(dir[-1]))
2196 return is_dir_sep(subdir[-1]) ? offset : -1;
2198 /* foo[/]bar vs foo[] */
2199 return is_dir_sep(*subdir) ? offset + 1 : -1;
2202 int is_inside_dir(const char *dir)
2204 char *cwd;
2205 int rc;
2207 if (!dir)
2208 return 0;
2210 cwd = xgetcwd();
2211 rc = (dir_inside_of(cwd, dir) >= 0);
2212 free(cwd);
2213 return rc;
2216 int is_empty_dir(const char *path)
2218 DIR *dir = opendir(path);
2219 struct dirent *e;
2220 int ret = 1;
2222 if (!dir)
2223 return 0;
2225 while ((e = readdir(dir)) != NULL)
2226 if (!is_dot_or_dotdot(e->d_name)) {
2227 ret = 0;
2228 break;
2231 closedir(dir);
2232 return ret;
2235 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2237 DIR *dir;
2238 struct dirent *e;
2239 int ret = 0, original_len = path->len, len, kept_down = 0;
2240 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2241 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2242 unsigned char submodule_head[20];
2244 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2245 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2246 /* Do not descend and nuke a nested git work tree. */
2247 if (kept_up)
2248 *kept_up = 1;
2249 return 0;
2252 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2253 dir = opendir(path->buf);
2254 if (!dir) {
2255 if (errno == ENOENT)
2256 return keep_toplevel ? -1 : 0;
2257 else if (errno == EACCES && !keep_toplevel)
2259 * An empty dir could be removable even if it
2260 * is unreadable:
2262 return rmdir(path->buf);
2263 else
2264 return -1;
2266 strbuf_complete(path, '/');
2268 len = path->len;
2269 while ((e = readdir(dir)) != NULL) {
2270 struct stat st;
2271 if (is_dot_or_dotdot(e->d_name))
2272 continue;
2274 strbuf_setlen(path, len);
2275 strbuf_addstr(path, e->d_name);
2276 if (lstat(path->buf, &st)) {
2277 if (errno == ENOENT)
2279 * file disappeared, which is what we
2280 * wanted anyway
2282 continue;
2283 /* fall thru */
2284 } else if (S_ISDIR(st.st_mode)) {
2285 if (!remove_dir_recurse(path, flag, &kept_down))
2286 continue; /* happy */
2287 } else if (!only_empty &&
2288 (!unlink(path->buf) || errno == ENOENT)) {
2289 continue; /* happy, too */
2292 /* path too long, stat fails, or non-directory still exists */
2293 ret = -1;
2294 break;
2296 closedir(dir);
2298 strbuf_setlen(path, original_len);
2299 if (!ret && !keep_toplevel && !kept_down)
2300 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2301 else if (kept_up)
2303 * report the uplevel that it is not an error that we
2304 * did not rmdir() our directory.
2306 *kept_up = !ret;
2307 return ret;
2310 int remove_dir_recursively(struct strbuf *path, int flag)
2312 return remove_dir_recurse(path, flag, NULL);
2315 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2317 void setup_standard_excludes(struct dir_struct *dir)
2319 dir->exclude_per_dir = ".gitignore";
2321 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2322 if (!excludes_file)
2323 excludes_file = xdg_config_home("ignore");
2324 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2325 add_excludes_from_file_1(dir, excludes_file,
2326 dir->untracked ? &dir->ss_excludes_file : NULL);
2328 /* per repository user preference */
2329 if (startup_info->have_repository) {
2330 const char *path = git_path_info_exclude();
2331 if (!access_or_warn(path, R_OK, 0))
2332 add_excludes_from_file_1(dir, path,
2333 dir->untracked ? &dir->ss_info_exclude : NULL);
2337 int remove_path(const char *name)
2339 char *slash;
2341 if (unlink(name) && !is_missing_file_error(errno))
2342 return -1;
2344 slash = strrchr(name, '/');
2345 if (slash) {
2346 char *dirs = xstrdup(name);
2347 slash = dirs + (slash - name);
2348 do {
2349 *slash = '\0';
2350 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2351 free(dirs);
2353 return 0;
2357 * Frees memory within dir which was allocated for exclude lists and
2358 * the exclude_stack. Does not free dir itself.
2360 void clear_directory(struct dir_struct *dir)
2362 int i, j;
2363 struct exclude_list_group *group;
2364 struct exclude_list *el;
2365 struct exclude_stack *stk;
2367 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2368 group = &dir->exclude_list_group[i];
2369 for (j = 0; j < group->nr; j++) {
2370 el = &group->el[j];
2371 if (i == EXC_DIRS)
2372 free((char *)el->src);
2373 clear_exclude_list(el);
2375 free(group->el);
2378 stk = dir->exclude_stack;
2379 while (stk) {
2380 struct exclude_stack *prev = stk->prev;
2381 free(stk);
2382 stk = prev;
2384 strbuf_release(&dir->basebuf);
2387 struct ondisk_untracked_cache {
2388 struct stat_data info_exclude_stat;
2389 struct stat_data excludes_file_stat;
2390 uint32_t dir_flags;
2391 unsigned char info_exclude_sha1[20];
2392 unsigned char excludes_file_sha1[20];
2393 char exclude_per_dir[FLEX_ARRAY];
2396 #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2398 struct write_data {
2399 int index; /* number of written untracked_cache_dir */
2400 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2401 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2402 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2403 struct strbuf out;
2404 struct strbuf sb_stat;
2405 struct strbuf sb_sha1;
2408 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2410 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2411 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2412 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2413 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2414 to->sd_dev = htonl(from->sd_dev);
2415 to->sd_ino = htonl(from->sd_ino);
2416 to->sd_uid = htonl(from->sd_uid);
2417 to->sd_gid = htonl(from->sd_gid);
2418 to->sd_size = htonl(from->sd_size);
2421 static void write_one_dir(struct untracked_cache_dir *untracked,
2422 struct write_data *wd)
2424 struct stat_data stat_data;
2425 struct strbuf *out = &wd->out;
2426 unsigned char intbuf[16];
2427 unsigned int intlen, value;
2428 int i = wd->index++;
2431 * untracked_nr should be reset whenever valid is clear, but
2432 * for safety..
2434 if (!untracked->valid) {
2435 untracked->untracked_nr = 0;
2436 untracked->check_only = 0;
2439 if (untracked->check_only)
2440 ewah_set(wd->check_only, i);
2441 if (untracked->valid) {
2442 ewah_set(wd->valid, i);
2443 stat_data_to_disk(&stat_data, &untracked->stat_data);
2444 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2446 if (!is_null_sha1(untracked->exclude_sha1)) {
2447 ewah_set(wd->sha1_valid, i);
2448 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2451 intlen = encode_varint(untracked->untracked_nr, intbuf);
2452 strbuf_add(out, intbuf, intlen);
2454 /* skip non-recurse directories */
2455 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2456 if (untracked->dirs[i]->recurse)
2457 value++;
2458 intlen = encode_varint(value, intbuf);
2459 strbuf_add(out, intbuf, intlen);
2461 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2463 for (i = 0; i < untracked->untracked_nr; i++)
2464 strbuf_add(out, untracked->untracked[i],
2465 strlen(untracked->untracked[i]) + 1);
2467 for (i = 0; i < untracked->dirs_nr; i++)
2468 if (untracked->dirs[i]->recurse)
2469 write_one_dir(untracked->dirs[i], wd);
2472 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2474 struct ondisk_untracked_cache *ouc;
2475 struct write_data wd;
2476 unsigned char varbuf[16];
2477 int varint_len;
2478 size_t len = strlen(untracked->exclude_per_dir);
2480 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
2481 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2482 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2483 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2484 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2485 ouc->dir_flags = htonl(untracked->dir_flags);
2487 varint_len = encode_varint(untracked->ident.len, varbuf);
2488 strbuf_add(out, varbuf, varint_len);
2489 strbuf_addbuf(out, &untracked->ident);
2491 strbuf_add(out, ouc, ouc_size(len));
2492 free(ouc);
2493 ouc = NULL;
2495 if (!untracked->root) {
2496 varint_len = encode_varint(0, varbuf);
2497 strbuf_add(out, varbuf, varint_len);
2498 return;
2501 wd.index = 0;
2502 wd.check_only = ewah_new();
2503 wd.valid = ewah_new();
2504 wd.sha1_valid = ewah_new();
2505 strbuf_init(&wd.out, 1024);
2506 strbuf_init(&wd.sb_stat, 1024);
2507 strbuf_init(&wd.sb_sha1, 1024);
2508 write_one_dir(untracked->root, &wd);
2510 varint_len = encode_varint(wd.index, varbuf);
2511 strbuf_add(out, varbuf, varint_len);
2512 strbuf_addbuf(out, &wd.out);
2513 ewah_serialize_strbuf(wd.valid, out);
2514 ewah_serialize_strbuf(wd.check_only, out);
2515 ewah_serialize_strbuf(wd.sha1_valid, out);
2516 strbuf_addbuf(out, &wd.sb_stat);
2517 strbuf_addbuf(out, &wd.sb_sha1);
2518 strbuf_addch(out, '\0'); /* safe guard for string lists */
2520 ewah_free(wd.valid);
2521 ewah_free(wd.check_only);
2522 ewah_free(wd.sha1_valid);
2523 strbuf_release(&wd.out);
2524 strbuf_release(&wd.sb_stat);
2525 strbuf_release(&wd.sb_sha1);
2528 static void free_untracked(struct untracked_cache_dir *ucd)
2530 int i;
2531 if (!ucd)
2532 return;
2533 for (i = 0; i < ucd->dirs_nr; i++)
2534 free_untracked(ucd->dirs[i]);
2535 for (i = 0; i < ucd->untracked_nr; i++)
2536 free(ucd->untracked[i]);
2537 free(ucd->untracked);
2538 free(ucd->dirs);
2539 free(ucd);
2542 void free_untracked_cache(struct untracked_cache *uc)
2544 if (uc)
2545 free_untracked(uc->root);
2546 free(uc);
2549 struct read_data {
2550 int index;
2551 struct untracked_cache_dir **ucd;
2552 struct ewah_bitmap *check_only;
2553 struct ewah_bitmap *valid;
2554 struct ewah_bitmap *sha1_valid;
2555 const unsigned char *data;
2556 const unsigned char *end;
2559 static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2561 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2562 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2563 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2564 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2565 to->sd_dev = get_be32(&from->sd_dev);
2566 to->sd_ino = get_be32(&from->sd_ino);
2567 to->sd_uid = get_be32(&from->sd_uid);
2568 to->sd_gid = get_be32(&from->sd_gid);
2569 to->sd_size = get_be32(&from->sd_size);
2572 static int read_one_dir(struct untracked_cache_dir **untracked_,
2573 struct read_data *rd)
2575 struct untracked_cache_dir ud, *untracked;
2576 const unsigned char *next, *data = rd->data, *end = rd->end;
2577 unsigned int value;
2578 int i, len;
2580 memset(&ud, 0, sizeof(ud));
2582 next = data;
2583 value = decode_varint(&next);
2584 if (next > end)
2585 return -1;
2586 ud.recurse = 1;
2587 ud.untracked_alloc = value;
2588 ud.untracked_nr = value;
2589 if (ud.untracked_nr)
2590 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2591 data = next;
2593 next = data;
2594 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2595 if (next > end)
2596 return -1;
2597 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2598 data = next;
2600 len = strlen((const char *)data);
2601 next = data + len + 1;
2602 if (next > rd->end)
2603 return -1;
2604 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
2605 memcpy(untracked, &ud, sizeof(ud));
2606 memcpy(untracked->name, data, len + 1);
2607 data = next;
2609 for (i = 0; i < untracked->untracked_nr; i++) {
2610 len = strlen((const char *)data);
2611 next = data + len + 1;
2612 if (next > rd->end)
2613 return -1;
2614 untracked->untracked[i] = xstrdup((const char*)data);
2615 data = next;
2618 rd->ucd[rd->index++] = untracked;
2619 rd->data = data;
2621 for (i = 0; i < untracked->dirs_nr; i++) {
2622 len = read_one_dir(untracked->dirs + i, rd);
2623 if (len < 0)
2624 return -1;
2626 return 0;
2629 static void set_check_only(size_t pos, void *cb)
2631 struct read_data *rd = cb;
2632 struct untracked_cache_dir *ud = rd->ucd[pos];
2633 ud->check_only = 1;
2636 static void read_stat(size_t pos, void *cb)
2638 struct read_data *rd = cb;
2639 struct untracked_cache_dir *ud = rd->ucd[pos];
2640 if (rd->data + sizeof(struct stat_data) > rd->end) {
2641 rd->data = rd->end + 1;
2642 return;
2644 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2645 rd->data += sizeof(struct stat_data);
2646 ud->valid = 1;
2649 static void read_sha1(size_t pos, void *cb)
2651 struct read_data *rd = cb;
2652 struct untracked_cache_dir *ud = rd->ucd[pos];
2653 if (rd->data + 20 > rd->end) {
2654 rd->data = rd->end + 1;
2655 return;
2657 hashcpy(ud->exclude_sha1, rd->data);
2658 rd->data += 20;
2661 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2662 const struct stat_data *stat,
2663 const unsigned char *sha1)
2665 stat_data_from_disk(&sha1_stat->stat, stat);
2666 hashcpy(sha1_stat->sha1, sha1);
2667 sha1_stat->valid = 1;
2670 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2672 const struct ondisk_untracked_cache *ouc;
2673 struct untracked_cache *uc;
2674 struct read_data rd;
2675 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2676 const char *ident;
2677 int ident_len, len;
2679 if (sz <= 1 || end[-1] != '\0')
2680 return NULL;
2681 end--;
2683 ident_len = decode_varint(&next);
2684 if (next + ident_len > end)
2685 return NULL;
2686 ident = (const char *)next;
2687 next += ident_len;
2689 ouc = (const struct ondisk_untracked_cache *)next;
2690 if (next + ouc_size(0) > end)
2691 return NULL;
2693 uc = xcalloc(1, sizeof(*uc));
2694 strbuf_init(&uc->ident, ident_len);
2695 strbuf_add(&uc->ident, ident, ident_len);
2696 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2697 ouc->info_exclude_sha1);
2698 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2699 ouc->excludes_file_sha1);
2700 uc->dir_flags = get_be32(&ouc->dir_flags);
2701 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2702 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2703 next += ouc_size(strlen(ouc->exclude_per_dir));
2704 if (next >= end)
2705 goto done2;
2707 len = decode_varint(&next);
2708 if (next > end || len == 0)
2709 goto done2;
2711 rd.valid = ewah_new();
2712 rd.check_only = ewah_new();
2713 rd.sha1_valid = ewah_new();
2714 rd.data = next;
2715 rd.end = end;
2716 rd.index = 0;
2717 ALLOC_ARRAY(rd.ucd, len);
2719 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2720 goto done;
2722 next = rd.data;
2723 len = ewah_read_mmap(rd.valid, next, end - next);
2724 if (len < 0)
2725 goto done;
2727 next += len;
2728 len = ewah_read_mmap(rd.check_only, next, end - next);
2729 if (len < 0)
2730 goto done;
2732 next += len;
2733 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2734 if (len < 0)
2735 goto done;
2737 ewah_each_bit(rd.check_only, set_check_only, &rd);
2738 rd.data = next + len;
2739 ewah_each_bit(rd.valid, read_stat, &rd);
2740 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2741 next = rd.data;
2743 done:
2744 free(rd.ucd);
2745 ewah_free(rd.valid);
2746 ewah_free(rd.check_only);
2747 ewah_free(rd.sha1_valid);
2748 done2:
2749 if (next != end) {
2750 free_untracked_cache(uc);
2751 uc = NULL;
2753 return uc;
2756 static void invalidate_one_directory(struct untracked_cache *uc,
2757 struct untracked_cache_dir *ucd)
2759 uc->dir_invalidated++;
2760 ucd->valid = 0;
2761 ucd->untracked_nr = 0;
2765 * Normally when an entry is added or removed from a directory,
2766 * invalidating that directory is enough. No need to touch its
2767 * ancestors. When a directory is shown as "foo/bar/" in git-status
2768 * however, deleting or adding an entry may have cascading effect.
2770 * Say the "foo/bar/file" has become untracked, we need to tell the
2771 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2772 * directory any more (because "bar" is managed by foo as an untracked
2773 * "file").
2775 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2776 * was the last untracked entry in the entire "foo", we should show
2777 * "foo/" instead. Which means we have to invalidate past "bar" up to
2778 * "foo".
2780 * This function traverses all directories from root to leaf. If there
2781 * is a chance of one of the above cases happening, we invalidate back
2782 * to root. Otherwise we just invalidate the leaf. There may be a more
2783 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2784 * detect these cases and avoid unnecessary invalidation, for example,
2785 * checking for the untracked entry named "bar/" in "foo", but for now
2786 * stick to something safe and simple.
2788 static int invalidate_one_component(struct untracked_cache *uc,
2789 struct untracked_cache_dir *dir,
2790 const char *path, int len)
2792 const char *rest = strchr(path, '/');
2794 if (rest) {
2795 int component_len = rest - path;
2796 struct untracked_cache_dir *d =
2797 lookup_untracked(uc, dir, path, component_len);
2798 int ret =
2799 invalidate_one_component(uc, d, rest + 1,
2800 len - (component_len + 1));
2801 if (ret)
2802 invalidate_one_directory(uc, dir);
2803 return ret;
2806 invalidate_one_directory(uc, dir);
2807 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2810 void untracked_cache_invalidate_path(struct index_state *istate,
2811 const char *path)
2813 if (!istate->untracked || !istate->untracked->root)
2814 return;
2815 invalidate_one_component(istate->untracked, istate->untracked->root,
2816 path, strlen(path));
2819 void untracked_cache_remove_from_index(struct index_state *istate,
2820 const char *path)
2822 untracked_cache_invalidate_path(istate, path);
2825 void untracked_cache_add_to_index(struct index_state *istate,
2826 const char *path)
2828 untracked_cache_invalidate_path(istate, path);
2831 /* Update gitfile and core.worktree setting to connect work tree and git dir */
2832 void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)
2834 struct strbuf gitfile_sb = STRBUF_INIT;
2835 struct strbuf cfg_sb = STRBUF_INIT;
2836 struct strbuf rel_path = STRBUF_INIT;
2837 char *git_dir, *work_tree;
2839 /* Prepare .git file */
2840 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
2841 if (safe_create_leading_directories_const(gitfile_sb.buf))
2842 die(_("could not create directories for %s"), gitfile_sb.buf);
2844 /* Prepare config file */
2845 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
2846 if (safe_create_leading_directories_const(cfg_sb.buf))
2847 die(_("could not create directories for %s"), cfg_sb.buf);
2849 git_dir = real_pathdup(git_dir_, 1);
2850 work_tree = real_pathdup(work_tree_, 1);
2852 /* Write .git file */
2853 write_file(gitfile_sb.buf, "gitdir: %s",
2854 relative_path(git_dir, work_tree, &rel_path));
2855 /* Update core.worktree setting */
2856 git_config_set_in_file(cfg_sb.buf, "core.worktree",
2857 relative_path(work_tree, git_dir, &rel_path));
2859 strbuf_release(&gitfile_sb);
2860 strbuf_release(&cfg_sb);
2861 strbuf_release(&rel_path);
2862 free(work_tree);
2863 free(git_dir);
2867 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
2869 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
2871 if (rename(old_git_dir, new_git_dir) < 0)
2872 die_errno(_("could not migrate git directory from '%s' to '%s'"),
2873 old_git_dir, new_git_dir);
2875 connect_work_tree_and_git_dir(path, new_git_dir);