Merge branch 'dl/complete-stash'
[alt-git.git] / dir.c
blobd81b5df8030f55cab6206d12a4b5723253ceaf69
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 "config.h"
10 #include "dir.h"
11 #include "object-store.h"
12 #include "attr.h"
13 #include "refs.h"
14 #include "wildmatch.h"
15 #include "pathspec.h"
16 #include "utf8.h"
17 #include "varint.h"
18 #include "ewah/ewok.h"
19 #include "fsmonitor.h"
20 #include "submodule-config.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 const char *d_name;
45 int d_type;
46 const char *file;
47 struct untracked_cache_dir *ucd;
50 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
51 struct index_state *istate, const char *path, int len,
52 struct untracked_cache_dir *untracked,
53 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
54 static int resolve_dtype(int dtype, struct index_state *istate,
55 const char *path, int len);
57 void dir_init(struct dir_struct *dir)
59 memset(dir, 0, sizeof(*dir));
62 int count_slashes(const char *s)
64 int cnt = 0;
65 while (*s)
66 if (*s++ == '/')
67 cnt++;
68 return cnt;
71 int fspathcmp(const char *a, const char *b)
73 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
76 int fspathncmp(const char *a, const char *b, size_t count)
78 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
81 int git_fnmatch(const struct pathspec_item *item,
82 const char *pattern, const char *string,
83 int prefix)
85 if (prefix > 0) {
86 if (ps_strncmp(item, pattern, string, prefix))
87 return WM_NOMATCH;
88 pattern += prefix;
89 string += prefix;
91 if (item->flags & PATHSPEC_ONESTAR) {
92 int pattern_len = strlen(++pattern);
93 int string_len = strlen(string);
94 return string_len < pattern_len ||
95 ps_strcmp(item, pattern,
96 string + string_len - pattern_len);
98 if (item->magic & PATHSPEC_GLOB)
99 return wildmatch(pattern, string,
100 WM_PATHNAME |
101 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
102 else
103 /* wildmatch has not learned no FNM_PATHNAME mode yet */
104 return wildmatch(pattern, string,
105 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
108 static int fnmatch_icase_mem(const char *pattern, int patternlen,
109 const char *string, int stringlen,
110 int flags)
112 int match_status;
113 struct strbuf pat_buf = STRBUF_INIT;
114 struct strbuf str_buf = STRBUF_INIT;
115 const char *use_pat = pattern;
116 const char *use_str = string;
118 if (pattern[patternlen]) {
119 strbuf_add(&pat_buf, pattern, patternlen);
120 use_pat = pat_buf.buf;
122 if (string[stringlen]) {
123 strbuf_add(&str_buf, string, stringlen);
124 use_str = str_buf.buf;
127 if (ignore_case)
128 flags |= WM_CASEFOLD;
129 match_status = wildmatch(use_pat, use_str, flags);
131 strbuf_release(&pat_buf);
132 strbuf_release(&str_buf);
134 return match_status;
137 static size_t common_prefix_len(const struct pathspec *pathspec)
139 int n;
140 size_t max = 0;
143 * ":(icase)path" is treated as a pathspec full of
144 * wildcard. In other words, only prefix is considered common
145 * prefix. If the pathspec is abc/foo abc/bar, running in
146 * subdir xyz, the common prefix is still xyz, not xyz/abc as
147 * in non-:(icase).
149 GUARD_PATHSPEC(pathspec,
150 PATHSPEC_FROMTOP |
151 PATHSPEC_MAXDEPTH |
152 PATHSPEC_LITERAL |
153 PATHSPEC_GLOB |
154 PATHSPEC_ICASE |
155 PATHSPEC_EXCLUDE |
156 PATHSPEC_ATTR);
158 for (n = 0; n < pathspec->nr; n++) {
159 size_t i = 0, len = 0, item_len;
160 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
161 continue;
162 if (pathspec->items[n].magic & PATHSPEC_ICASE)
163 item_len = pathspec->items[n].prefix;
164 else
165 item_len = pathspec->items[n].nowildcard_len;
166 while (i < item_len && (n == 0 || i < max)) {
167 char c = pathspec->items[n].match[i];
168 if (c != pathspec->items[0].match[i])
169 break;
170 if (c == '/')
171 len = i + 1;
172 i++;
174 if (n == 0 || len < max) {
175 max = len;
176 if (!max)
177 break;
180 return max;
184 * Returns a copy of the longest leading path common among all
185 * pathspecs.
187 char *common_prefix(const struct pathspec *pathspec)
189 unsigned long len = common_prefix_len(pathspec);
191 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
194 int fill_directory(struct dir_struct *dir,
195 struct index_state *istate,
196 const struct pathspec *pathspec)
198 const char *prefix;
199 size_t prefix_len;
201 unsigned exclusive_flags = DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO;
202 if ((dir->flags & exclusive_flags) == exclusive_flags)
203 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
206 * Calculate common prefix for the pathspec, and
207 * use that to optimize the directory walk
209 prefix_len = common_prefix_len(pathspec);
210 prefix = prefix_len ? pathspec->items[0].match : "";
212 /* Read the directory and prune it */
213 read_directory(dir, istate, prefix, prefix_len, pathspec);
215 return prefix_len;
218 int within_depth(const char *name, int namelen,
219 int depth, int max_depth)
221 const char *cp = name, *cpe = name + namelen;
223 while (cp < cpe) {
224 if (*cp++ != '/')
225 continue;
226 depth++;
227 if (depth > max_depth)
228 return 0;
230 return 1;
234 * Read the contents of the blob with the given OID into a buffer.
235 * Append a trailing LF to the end if the last line doesn't have one.
237 * Returns:
238 * -1 when the OID is invalid or unknown or does not refer to a blob.
239 * 0 when the blob is empty.
240 * 1 along with { data, size } of the (possibly augmented) buffer
241 * when successful.
243 * Optionally updates the given oid_stat with the given OID (when valid).
245 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
246 size_t *size_out, char **data_out)
248 enum object_type type;
249 unsigned long sz;
250 char *data;
252 *size_out = 0;
253 *data_out = NULL;
255 data = read_object_file(oid, &type, &sz);
256 if (!data || type != OBJ_BLOB) {
257 free(data);
258 return -1;
261 if (oid_stat) {
262 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
263 oidcpy(&oid_stat->oid, oid);
266 if (sz == 0) {
267 free(data);
268 return 0;
271 if (data[sz - 1] != '\n') {
272 data = xrealloc(data, st_add(sz, 1));
273 data[sz++] = '\n';
276 *size_out = xsize_t(sz);
277 *data_out = data;
279 return 1;
282 #define DO_MATCH_EXCLUDE (1<<0)
283 #define DO_MATCH_DIRECTORY (1<<1)
284 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
287 * Does the given pathspec match the given name? A match is found if
289 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
290 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
291 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
292 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
294 * Return value tells which case it was (1-4), or 0 when there is no match.
296 * It may be instructive to look at a small table of concrete examples
297 * to understand the differences between 1, 2, and 4:
299 * Pathspecs
300 * | a/b | a/b/ | a/b/c
301 * ------+-----------+-----------+------------
302 * a/b | EXACT | EXACT[1] | LEADING[2]
303 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
304 * a/b/c | RECURSIVE | RECURSIVE | EXACT
306 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
307 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
309 static int match_pathspec_item(struct index_state *istate,
310 const struct pathspec_item *item, int prefix,
311 const char *name, int namelen, unsigned flags)
313 /* name/namelen has prefix cut off by caller */
314 const char *match = item->match + prefix;
315 int matchlen = item->len - prefix;
318 * The normal call pattern is:
319 * 1. prefix = common_prefix_len(ps);
320 * 2. prune something, or fill_directory
321 * 3. match_pathspec()
323 * 'prefix' at #1 may be shorter than the command's prefix and
324 * it's ok for #2 to match extra files. Those extras will be
325 * trimmed at #3.
327 * Suppose the pathspec is 'foo' and '../bar' running from
328 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
329 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
330 * user does not want XYZ/foo, only the "foo" part should be
331 * case-insensitive. We need to filter out XYZ/foo here. In
332 * other words, we do not trust the caller on comparing the
333 * prefix part when :(icase) is involved. We do exact
334 * comparison ourselves.
336 * Normally the caller (common_prefix_len() in fact) does
337 * _exact_ matching on name[-prefix+1..-1] and we do not need
338 * to check that part. Be defensive and check it anyway, in
339 * case common_prefix_len is changed, or a new caller is
340 * introduced that does not use common_prefix_len.
342 * If the penalty turns out too high when prefix is really
343 * long, maybe change it to
344 * strncmp(match, name, item->prefix - prefix)
346 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
347 strncmp(item->match, name - prefix, item->prefix))
348 return 0;
350 if (item->attr_match_nr &&
351 !match_pathspec_attrs(istate, name, namelen, item))
352 return 0;
354 /* If the match was just the prefix, we matched */
355 if (!*match)
356 return MATCHED_RECURSIVELY;
358 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
359 if (matchlen == namelen)
360 return MATCHED_EXACTLY;
362 if (match[matchlen-1] == '/' || name[matchlen] == '/')
363 return MATCHED_RECURSIVELY;
364 } else if ((flags & DO_MATCH_DIRECTORY) &&
365 match[matchlen - 1] == '/' &&
366 namelen == matchlen - 1 &&
367 !ps_strncmp(item, match, name, namelen))
368 return MATCHED_EXACTLY;
370 if (item->nowildcard_len < item->len &&
371 !git_fnmatch(item, match, name,
372 item->nowildcard_len - prefix))
373 return MATCHED_FNMATCH;
375 /* Perform checks to see if "name" is a leading string of the pathspec */
376 if ( (flags & DO_MATCH_LEADING_PATHSPEC) &&
377 !(flags & DO_MATCH_EXCLUDE)) {
378 /* name is a literal prefix of the pathspec */
379 int offset = name[namelen-1] == '/' ? 1 : 0;
380 if ((namelen < matchlen) &&
381 (match[namelen-offset] == '/') &&
382 !ps_strncmp(item, match, name, namelen))
383 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
385 /* name doesn't match up to the first wild character */
386 if (item->nowildcard_len < item->len &&
387 ps_strncmp(item, match, name,
388 item->nowildcard_len - prefix))
389 return 0;
392 * name has no wildcard, and it didn't match as a leading
393 * pathspec so return.
395 if (item->nowildcard_len == item->len)
396 return 0;
399 * Here is where we would perform a wildmatch to check if
400 * "name" can be matched as a directory (or a prefix) against
401 * the pathspec. Since wildmatch doesn't have this capability
402 * at the present we have to punt and say that it is a match,
403 * potentially returning a false positive
404 * The submodules themselves will be able to perform more
405 * accurate matching to determine if the pathspec matches.
407 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
410 return 0;
414 * do_match_pathspec() is meant to ONLY be called by
415 * match_pathspec_with_flags(); calling it directly risks pathspecs
416 * like ':!unwanted_path' being ignored.
418 * Given a name and a list of pathspecs, returns the nature of the
419 * closest (i.e. most specific) match of the name to any of the
420 * pathspecs.
422 * The caller typically calls this multiple times with the same
423 * pathspec and seen[] array but with different name/namelen
424 * (e.g. entries from the index) and is interested in seeing if and
425 * how each pathspec matches all the names it calls this function
426 * with. A mark is left in the seen[] array for each pathspec element
427 * indicating the closest type of match that element achieved, so if
428 * seen[n] remains zero after multiple invocations, that means the nth
429 * pathspec did not match any names, which could indicate that the
430 * user mistyped the nth pathspec.
432 static int do_match_pathspec(struct index_state *istate,
433 const struct pathspec *ps,
434 const char *name, int namelen,
435 int prefix, char *seen,
436 unsigned flags)
438 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
440 GUARD_PATHSPEC(ps,
441 PATHSPEC_FROMTOP |
442 PATHSPEC_MAXDEPTH |
443 PATHSPEC_LITERAL |
444 PATHSPEC_GLOB |
445 PATHSPEC_ICASE |
446 PATHSPEC_EXCLUDE |
447 PATHSPEC_ATTR);
449 if (!ps->nr) {
450 if (!ps->recursive ||
451 !(ps->magic & PATHSPEC_MAXDEPTH) ||
452 ps->max_depth == -1)
453 return MATCHED_RECURSIVELY;
455 if (within_depth(name, namelen, 0, ps->max_depth))
456 return MATCHED_EXACTLY;
457 else
458 return 0;
461 name += prefix;
462 namelen -= prefix;
464 for (i = ps->nr - 1; i >= 0; i--) {
465 int how;
467 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
468 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
469 continue;
471 if (seen && seen[i] == MATCHED_EXACTLY)
472 continue;
474 * Make exclude patterns optional and never report
475 * "pathspec ':(exclude)foo' matches no files"
477 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
478 seen[i] = MATCHED_FNMATCH;
479 how = match_pathspec_item(istate, ps->items+i, prefix, name,
480 namelen, flags);
481 if (ps->recursive &&
482 (ps->magic & PATHSPEC_MAXDEPTH) &&
483 ps->max_depth != -1 &&
484 how && how != MATCHED_FNMATCH) {
485 int len = ps->items[i].len;
486 if (name[len] == '/')
487 len++;
488 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
489 how = MATCHED_EXACTLY;
490 else
491 how = 0;
493 if (how) {
494 if (retval < how)
495 retval = how;
496 if (seen && seen[i] < how)
497 seen[i] = how;
500 return retval;
503 static int match_pathspec_with_flags(struct index_state *istate,
504 const struct pathspec *ps,
505 const char *name, int namelen,
506 int prefix, char *seen, unsigned flags)
508 int positive, negative;
509 positive = do_match_pathspec(istate, ps, name, namelen,
510 prefix, seen, flags);
511 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
512 return positive;
513 negative = do_match_pathspec(istate, ps, name, namelen,
514 prefix, seen,
515 flags | DO_MATCH_EXCLUDE);
516 return negative ? 0 : positive;
519 int match_pathspec(struct index_state *istate,
520 const struct pathspec *ps,
521 const char *name, int namelen,
522 int prefix, char *seen, int is_dir)
524 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
525 return match_pathspec_with_flags(istate, ps, name, namelen,
526 prefix, seen, flags);
530 * Check if a submodule is a superset of the pathspec
532 int submodule_path_match(struct index_state *istate,
533 const struct pathspec *ps,
534 const char *submodule_name,
535 char *seen)
537 int matched = match_pathspec_with_flags(istate, ps, submodule_name,
538 strlen(submodule_name),
539 0, seen,
540 DO_MATCH_DIRECTORY |
541 DO_MATCH_LEADING_PATHSPEC);
542 return matched;
545 int report_path_error(const char *ps_matched,
546 const struct pathspec *pathspec)
549 * Make sure all pathspec matched; otherwise it is an error.
551 int num, errors = 0;
552 for (num = 0; num < pathspec->nr; num++) {
553 int other, found_dup;
555 if (ps_matched[num])
556 continue;
558 * The caller might have fed identical pathspec
559 * twice. Do not barf on such a mistake.
560 * FIXME: parse_pathspec should have eliminated
561 * duplicate pathspec.
563 for (found_dup = other = 0;
564 !found_dup && other < pathspec->nr;
565 other++) {
566 if (other == num || !ps_matched[other])
567 continue;
568 if (!strcmp(pathspec->items[other].original,
569 pathspec->items[num].original))
571 * Ok, we have a match already.
573 found_dup = 1;
575 if (found_dup)
576 continue;
578 error(_("pathspec '%s' did not match any file(s) known to git"),
579 pathspec->items[num].original);
580 errors++;
582 return errors;
586 * Return the length of the "simple" part of a path match limiter.
588 int simple_length(const char *match)
590 int len = -1;
592 for (;;) {
593 unsigned char c = *match++;
594 len++;
595 if (c == '\0' || is_glob_special(c))
596 return len;
600 int no_wildcard(const char *string)
602 return string[simple_length(string)] == '\0';
605 void parse_path_pattern(const char **pattern,
606 int *patternlen,
607 unsigned *flags,
608 int *nowildcardlen)
610 const char *p = *pattern;
611 size_t i, len;
613 *flags = 0;
614 if (*p == '!') {
615 *flags |= PATTERN_FLAG_NEGATIVE;
616 p++;
618 len = strlen(p);
619 if (len && p[len - 1] == '/') {
620 len--;
621 *flags |= PATTERN_FLAG_MUSTBEDIR;
623 for (i = 0; i < len; i++) {
624 if (p[i] == '/')
625 break;
627 if (i == len)
628 *flags |= PATTERN_FLAG_NODIR;
629 *nowildcardlen = simple_length(p);
631 * we should have excluded the trailing slash from 'p' too,
632 * but that's one more allocation. Instead just make sure
633 * nowildcardlen does not exceed real patternlen
635 if (*nowildcardlen > len)
636 *nowildcardlen = len;
637 if (*p == '*' && no_wildcard(p + 1))
638 *flags |= PATTERN_FLAG_ENDSWITH;
639 *pattern = p;
640 *patternlen = len;
643 int pl_hashmap_cmp(const void *unused_cmp_data,
644 const struct hashmap_entry *a,
645 const struct hashmap_entry *b,
646 const void *key)
648 const struct pattern_entry *ee1 =
649 container_of(a, struct pattern_entry, ent);
650 const struct pattern_entry *ee2 =
651 container_of(b, struct pattern_entry, ent);
653 size_t min_len = ee1->patternlen <= ee2->patternlen
654 ? ee1->patternlen
655 : ee2->patternlen;
657 if (ignore_case)
658 return strncasecmp(ee1->pattern, ee2->pattern, min_len);
659 return strncmp(ee1->pattern, ee2->pattern, min_len);
662 static char *dup_and_filter_pattern(const char *pattern)
664 char *set, *read;
665 size_t count = 0;
666 char *result = xstrdup(pattern);
668 set = result;
669 read = result;
671 while (*read) {
672 /* skip escape characters (once) */
673 if (*read == '\\')
674 read++;
676 *set = *read;
678 set++;
679 read++;
680 count++;
682 *set = 0;
684 if (count > 2 &&
685 *(set - 1) == '*' &&
686 *(set - 2) == '/')
687 *(set - 2) = 0;
689 return result;
692 static void add_pattern_to_hashsets(struct pattern_list *pl, struct path_pattern *given)
694 struct pattern_entry *translated;
695 char *truncated;
696 char *data = NULL;
697 const char *prev, *cur, *next;
699 if (!pl->use_cone_patterns)
700 return;
702 if (given->flags & PATTERN_FLAG_NEGATIVE &&
703 given->flags & PATTERN_FLAG_MUSTBEDIR &&
704 !strcmp(given->pattern, "/*")) {
705 pl->full_cone = 0;
706 return;
709 if (!given->flags && !strcmp(given->pattern, "/*")) {
710 pl->full_cone = 1;
711 return;
714 if (given->patternlen < 2 ||
715 *given->pattern == '*' ||
716 strstr(given->pattern, "**")) {
717 /* Not a cone pattern. */
718 warning(_("unrecognized pattern: '%s'"), given->pattern);
719 goto clear_hashmaps;
722 prev = given->pattern;
723 cur = given->pattern + 1;
724 next = given->pattern + 2;
726 while (*cur) {
727 /* Watch for glob characters '*', '\', '[', '?' */
728 if (!is_glob_special(*cur))
729 goto increment;
731 /* But only if *prev != '\\' */
732 if (*prev == '\\')
733 goto increment;
735 /* But allow the initial '\' */
736 if (*cur == '\\' &&
737 is_glob_special(*next))
738 goto increment;
740 /* But a trailing '/' then '*' is fine */
741 if (*prev == '/' &&
742 *cur == '*' &&
743 *next == 0)
744 goto increment;
746 /* Not a cone pattern. */
747 warning(_("unrecognized pattern: '%s'"), given->pattern);
748 goto clear_hashmaps;
750 increment:
751 prev++;
752 cur++;
753 next++;
756 if (given->patternlen > 2 &&
757 !strcmp(given->pattern + given->patternlen - 2, "/*")) {
758 if (!(given->flags & PATTERN_FLAG_NEGATIVE)) {
759 /* Not a cone pattern. */
760 warning(_("unrecognized pattern: '%s'"), given->pattern);
761 goto clear_hashmaps;
764 truncated = dup_and_filter_pattern(given->pattern);
766 translated = xmalloc(sizeof(struct pattern_entry));
767 translated->pattern = truncated;
768 translated->patternlen = given->patternlen - 2;
769 hashmap_entry_init(&translated->ent,
770 ignore_case ?
771 strihash(translated->pattern) :
772 strhash(translated->pattern));
774 if (!hashmap_get_entry(&pl->recursive_hashmap,
775 translated, ent, NULL)) {
776 /* We did not see the "parent" included */
777 warning(_("unrecognized negative pattern: '%s'"),
778 given->pattern);
779 free(truncated);
780 free(translated);
781 goto clear_hashmaps;
784 hashmap_add(&pl->parent_hashmap, &translated->ent);
785 hashmap_remove(&pl->recursive_hashmap, &translated->ent, &data);
786 free(data);
787 return;
790 if (given->flags & PATTERN_FLAG_NEGATIVE) {
791 warning(_("unrecognized negative pattern: '%s'"),
792 given->pattern);
793 goto clear_hashmaps;
796 translated = xmalloc(sizeof(struct pattern_entry));
798 translated->pattern = dup_and_filter_pattern(given->pattern);
799 translated->patternlen = given->patternlen;
800 hashmap_entry_init(&translated->ent,
801 ignore_case ?
802 strihash(translated->pattern) :
803 strhash(translated->pattern));
805 hashmap_add(&pl->recursive_hashmap, &translated->ent);
807 if (hashmap_get_entry(&pl->parent_hashmap, translated, ent, NULL)) {
808 /* we already included this at the parent level */
809 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
810 given->pattern);
811 hashmap_remove(&pl->parent_hashmap, &translated->ent, &data);
812 free(data);
813 free(translated);
816 return;
818 clear_hashmaps:
819 warning(_("disabling cone pattern matching"));
820 hashmap_clear_and_free(&pl->parent_hashmap, struct pattern_entry, ent);
821 hashmap_clear_and_free(&pl->recursive_hashmap, struct pattern_entry, ent);
822 pl->use_cone_patterns = 0;
825 static int hashmap_contains_path(struct hashmap *map,
826 struct strbuf *pattern)
828 struct pattern_entry p;
830 /* Check straight mapping */
831 p.pattern = pattern->buf;
832 p.patternlen = pattern->len;
833 hashmap_entry_init(&p.ent,
834 ignore_case ?
835 strihash(p.pattern) :
836 strhash(p.pattern));
837 return !!hashmap_get_entry(map, &p, ent, NULL);
840 int hashmap_contains_parent(struct hashmap *map,
841 const char *path,
842 struct strbuf *buffer)
844 char *slash_pos;
846 strbuf_setlen(buffer, 0);
848 if (path[0] != '/')
849 strbuf_addch(buffer, '/');
851 strbuf_addstr(buffer, path);
853 slash_pos = strrchr(buffer->buf, '/');
855 while (slash_pos > buffer->buf) {
856 strbuf_setlen(buffer, slash_pos - buffer->buf);
858 if (hashmap_contains_path(map, buffer))
859 return 1;
861 slash_pos = strrchr(buffer->buf, '/');
864 return 0;
867 void add_pattern(const char *string, const char *base,
868 int baselen, struct pattern_list *pl, int srcpos)
870 struct path_pattern *pattern;
871 int patternlen;
872 unsigned flags;
873 int nowildcardlen;
875 parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen);
876 if (flags & PATTERN_FLAG_MUSTBEDIR) {
877 FLEXPTR_ALLOC_MEM(pattern, pattern, string, patternlen);
878 } else {
879 pattern = xmalloc(sizeof(*pattern));
880 pattern->pattern = string;
882 pattern->patternlen = patternlen;
883 pattern->nowildcardlen = nowildcardlen;
884 pattern->base = base;
885 pattern->baselen = baselen;
886 pattern->flags = flags;
887 pattern->srcpos = srcpos;
888 ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc);
889 pl->patterns[pl->nr++] = pattern;
890 pattern->pl = pl;
892 add_pattern_to_hashsets(pl, pattern);
895 static int read_skip_worktree_file_from_index(struct index_state *istate,
896 const char *path,
897 size_t *size_out, char **data_out,
898 struct oid_stat *oid_stat)
900 int pos, len;
902 len = strlen(path);
903 pos = index_name_pos(istate, path, len);
904 if (pos < 0)
905 return -1;
906 if (!ce_skip_worktree(istate->cache[pos]))
907 return -1;
909 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
913 * Frees memory within pl which was allocated for exclude patterns and
914 * the file buffer. Does not free pl itself.
916 void clear_pattern_list(struct pattern_list *pl)
918 int i;
920 for (i = 0; i < pl->nr; i++)
921 free(pl->patterns[i]);
922 free(pl->patterns);
923 free(pl->filebuf);
924 hashmap_clear_and_free(&pl->recursive_hashmap, struct pattern_entry, ent);
925 hashmap_clear_and_free(&pl->parent_hashmap, struct pattern_entry, ent);
927 memset(pl, 0, sizeof(*pl));
930 static void trim_trailing_spaces(char *buf)
932 char *p, *last_space = NULL;
934 for (p = buf; *p; p++)
935 switch (*p) {
936 case ' ':
937 if (!last_space)
938 last_space = p;
939 break;
940 case '\\':
941 p++;
942 if (!*p)
943 return;
944 /* fallthrough */
945 default:
946 last_space = NULL;
949 if (last_space)
950 *last_space = '\0';
954 * Given a subdirectory name and "dir" of the current directory,
955 * search the subdir in "dir" and return it, or create a new one if it
956 * does not exist in "dir".
958 * If "name" has the trailing slash, it'll be excluded in the search.
960 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
961 struct untracked_cache_dir *dir,
962 const char *name, int len)
964 int first, last;
965 struct untracked_cache_dir *d;
966 if (!dir)
967 return NULL;
968 if (len && name[len - 1] == '/')
969 len--;
970 first = 0;
971 last = dir->dirs_nr;
972 while (last > first) {
973 int cmp, next = first + ((last - first) >> 1);
974 d = dir->dirs[next];
975 cmp = strncmp(name, d->name, len);
976 if (!cmp && strlen(d->name) > len)
977 cmp = -1;
978 if (!cmp)
979 return d;
980 if (cmp < 0) {
981 last = next;
982 continue;
984 first = next+1;
987 uc->dir_created++;
988 FLEX_ALLOC_MEM(d, name, name, len);
990 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
991 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
992 dir->dirs_nr - first);
993 dir->dirs_nr++;
994 dir->dirs[first] = d;
995 return d;
998 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
1000 int i;
1001 dir->valid = 0;
1002 dir->untracked_nr = 0;
1003 for (i = 0; i < dir->dirs_nr; i++)
1004 do_invalidate_gitignore(dir->dirs[i]);
1007 static void invalidate_gitignore(struct untracked_cache *uc,
1008 struct untracked_cache_dir *dir)
1010 uc->gitignore_invalidated++;
1011 do_invalidate_gitignore(dir);
1014 static void invalidate_directory(struct untracked_cache *uc,
1015 struct untracked_cache_dir *dir)
1017 int i;
1020 * Invalidation increment here is just roughly correct. If
1021 * untracked_nr or any of dirs[].recurse is non-zero, we
1022 * should increment dir_invalidated too. But that's more
1023 * expensive to do.
1025 if (dir->valid)
1026 uc->dir_invalidated++;
1028 dir->valid = 0;
1029 dir->untracked_nr = 0;
1030 for (i = 0; i < dir->dirs_nr; i++)
1031 dir->dirs[i]->recurse = 0;
1034 static int add_patterns_from_buffer(char *buf, size_t size,
1035 const char *base, int baselen,
1036 struct pattern_list *pl);
1038 /* Flags for add_patterns() */
1039 #define PATTERN_NOFOLLOW (1<<0)
1042 * Given a file with name "fname", read it (either from disk, or from
1043 * an index if 'istate' is non-null), parse it and store the
1044 * exclude rules in "pl".
1046 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1047 * stat data from disk (only valid if add_patterns returns zero). If
1048 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1050 static int add_patterns(const char *fname, const char *base, int baselen,
1051 struct pattern_list *pl, struct index_state *istate,
1052 unsigned flags, struct oid_stat *oid_stat)
1054 struct stat st;
1055 int r;
1056 int fd;
1057 size_t size = 0;
1058 char *buf;
1060 if (flags & PATTERN_NOFOLLOW)
1061 fd = open_nofollow(fname, O_RDONLY);
1062 else
1063 fd = open(fname, O_RDONLY);
1065 if (fd < 0 || fstat(fd, &st) < 0) {
1066 if (fd < 0)
1067 warn_on_fopen_errors(fname);
1068 else
1069 close(fd);
1070 if (!istate)
1071 return -1;
1072 r = read_skip_worktree_file_from_index(istate, fname,
1073 &size, &buf,
1074 oid_stat);
1075 if (r != 1)
1076 return r;
1077 } else {
1078 size = xsize_t(st.st_size);
1079 if (size == 0) {
1080 if (oid_stat) {
1081 fill_stat_data(&oid_stat->stat, &st);
1082 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
1083 oid_stat->valid = 1;
1085 close(fd);
1086 return 0;
1088 buf = xmallocz(size);
1089 if (read_in_full(fd, buf, size) != size) {
1090 free(buf);
1091 close(fd);
1092 return -1;
1094 buf[size++] = '\n';
1095 close(fd);
1096 if (oid_stat) {
1097 int pos;
1098 if (oid_stat->valid &&
1099 !match_stat_data_racy(istate, &oid_stat->stat, &st))
1100 ; /* no content change, oid_stat->oid still good */
1101 else if (istate &&
1102 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
1103 !ce_stage(istate->cache[pos]) &&
1104 ce_uptodate(istate->cache[pos]) &&
1105 !would_convert_to_git(istate, fname))
1106 oidcpy(&oid_stat->oid,
1107 &istate->cache[pos]->oid);
1108 else
1109 hash_object_file(the_hash_algo, buf, size,
1110 "blob", &oid_stat->oid);
1111 fill_stat_data(&oid_stat->stat, &st);
1112 oid_stat->valid = 1;
1116 add_patterns_from_buffer(buf, size, base, baselen, pl);
1117 return 0;
1120 static int add_patterns_from_buffer(char *buf, size_t size,
1121 const char *base, int baselen,
1122 struct pattern_list *pl)
1124 int i, lineno = 1;
1125 char *entry;
1127 hashmap_init(&pl->recursive_hashmap, pl_hashmap_cmp, NULL, 0);
1128 hashmap_init(&pl->parent_hashmap, pl_hashmap_cmp, NULL, 0);
1130 pl->filebuf = buf;
1132 if (skip_utf8_bom(&buf, size))
1133 size -= buf - pl->filebuf;
1135 entry = buf;
1137 for (i = 0; i < size; i++) {
1138 if (buf[i] == '\n') {
1139 if (entry != buf + i && entry[0] != '#') {
1140 buf[i - (i && buf[i-1] == '\r')] = 0;
1141 trim_trailing_spaces(entry);
1142 add_pattern(entry, base, baselen, pl, lineno);
1144 lineno++;
1145 entry = buf + i + 1;
1148 return 0;
1151 int add_patterns_from_file_to_list(const char *fname, const char *base,
1152 int baselen, struct pattern_list *pl,
1153 struct index_state *istate,
1154 unsigned flags)
1156 return add_patterns(fname, base, baselen, pl, istate, flags, NULL);
1159 int add_patterns_from_blob_to_list(
1160 struct object_id *oid,
1161 const char *base, int baselen,
1162 struct pattern_list *pl)
1164 char *buf;
1165 size_t size;
1166 int r;
1168 r = do_read_blob(oid, NULL, &size, &buf);
1169 if (r != 1)
1170 return r;
1172 add_patterns_from_buffer(buf, size, base, baselen, pl);
1173 return 0;
1176 struct pattern_list *add_pattern_list(struct dir_struct *dir,
1177 int group_type, const char *src)
1179 struct pattern_list *pl;
1180 struct exclude_list_group *group;
1182 group = &dir->exclude_list_group[group_type];
1183 ALLOC_GROW(group->pl, group->nr + 1, group->alloc);
1184 pl = &group->pl[group->nr++];
1185 memset(pl, 0, sizeof(*pl));
1186 pl->src = src;
1187 return pl;
1191 * Used to set up core.excludesfile and .git/info/exclude lists.
1193 static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname,
1194 struct oid_stat *oid_stat)
1196 struct pattern_list *pl;
1198 * catch setup_standard_excludes() that's called before
1199 * dir->untracked is assigned. That function behaves
1200 * differently when dir->untracked is non-NULL.
1202 if (!dir->untracked)
1203 dir->unmanaged_exclude_files++;
1204 pl = add_pattern_list(dir, EXC_FILE, fname);
1205 if (add_patterns(fname, "", 0, pl, NULL, 0, oid_stat) < 0)
1206 die(_("cannot use %s as an exclude file"), fname);
1209 void add_patterns_from_file(struct dir_struct *dir, const char *fname)
1211 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
1212 add_patterns_from_file_1(dir, fname, NULL);
1215 int match_basename(const char *basename, int basenamelen,
1216 const char *pattern, int prefix, int patternlen,
1217 unsigned flags)
1219 if (prefix == patternlen) {
1220 if (patternlen == basenamelen &&
1221 !fspathncmp(pattern, basename, basenamelen))
1222 return 1;
1223 } else if (flags & PATTERN_FLAG_ENDSWITH) {
1224 /* "*literal" matching against "fooliteral" */
1225 if (patternlen - 1 <= basenamelen &&
1226 !fspathncmp(pattern + 1,
1227 basename + basenamelen - (patternlen - 1),
1228 patternlen - 1))
1229 return 1;
1230 } else {
1231 if (fnmatch_icase_mem(pattern, patternlen,
1232 basename, basenamelen,
1233 0) == 0)
1234 return 1;
1236 return 0;
1239 int match_pathname(const char *pathname, int pathlen,
1240 const char *base, int baselen,
1241 const char *pattern, int prefix, int patternlen,
1242 unsigned flags)
1244 const char *name;
1245 int namelen;
1248 * match with FNM_PATHNAME; the pattern has base implicitly
1249 * in front of it.
1251 if (*pattern == '/') {
1252 pattern++;
1253 patternlen--;
1254 prefix--;
1258 * baselen does not count the trailing slash. base[] may or
1259 * may not end with a trailing slash though.
1261 if (pathlen < baselen + 1 ||
1262 (baselen && pathname[baselen] != '/') ||
1263 fspathncmp(pathname, base, baselen))
1264 return 0;
1266 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1267 name = pathname + pathlen - namelen;
1269 if (prefix) {
1271 * if the non-wildcard part is longer than the
1272 * remaining pathname, surely it cannot match.
1274 if (prefix > namelen)
1275 return 0;
1277 if (fspathncmp(pattern, name, prefix))
1278 return 0;
1279 pattern += prefix;
1280 patternlen -= prefix;
1281 name += prefix;
1282 namelen -= prefix;
1285 * If the whole pattern did not have a wildcard,
1286 * then our prefix match is all we need; we
1287 * do not need to call fnmatch at all.
1289 if (!patternlen && !namelen)
1290 return 1;
1293 return fnmatch_icase_mem(pattern, patternlen,
1294 name, namelen,
1295 WM_PATHNAME) == 0;
1299 * Scan the given exclude list in reverse to see whether pathname
1300 * should be ignored. The first match (i.e. the last on the list), if
1301 * any, determines the fate. Returns the exclude_list element which
1302 * matched, or NULL for undecided.
1304 static struct path_pattern *last_matching_pattern_from_list(const char *pathname,
1305 int pathlen,
1306 const char *basename,
1307 int *dtype,
1308 struct pattern_list *pl,
1309 struct index_state *istate)
1311 struct path_pattern *res = NULL; /* undecided */
1312 int i;
1314 if (!pl->nr)
1315 return NULL; /* undefined */
1317 for (i = pl->nr - 1; 0 <= i; i--) {
1318 struct path_pattern *pattern = pl->patterns[i];
1319 const char *exclude = pattern->pattern;
1320 int prefix = pattern->nowildcardlen;
1322 if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) {
1323 *dtype = resolve_dtype(*dtype, istate, pathname, pathlen);
1324 if (*dtype != DT_DIR)
1325 continue;
1328 if (pattern->flags & PATTERN_FLAG_NODIR) {
1329 if (match_basename(basename,
1330 pathlen - (basename - pathname),
1331 exclude, prefix, pattern->patternlen,
1332 pattern->flags)) {
1333 res = pattern;
1334 break;
1336 continue;
1339 assert(pattern->baselen == 0 ||
1340 pattern->base[pattern->baselen - 1] == '/');
1341 if (match_pathname(pathname, pathlen,
1342 pattern->base,
1343 pattern->baselen ? pattern->baselen - 1 : 0,
1344 exclude, prefix, pattern->patternlen,
1345 pattern->flags)) {
1346 res = pattern;
1347 break;
1350 return res;
1354 * Scan the list of patterns to determine if the ordered list
1355 * of patterns matches on 'pathname'.
1357 * Return 1 for a match, 0 for not matched and -1 for undecided.
1359 enum pattern_match_result path_matches_pattern_list(
1360 const char *pathname, int pathlen,
1361 const char *basename, int *dtype,
1362 struct pattern_list *pl,
1363 struct index_state *istate)
1365 struct path_pattern *pattern;
1366 struct strbuf parent_pathname = STRBUF_INIT;
1367 int result = NOT_MATCHED;
1368 const char *slash_pos;
1370 if (!pl->use_cone_patterns) {
1371 pattern = last_matching_pattern_from_list(pathname, pathlen, basename,
1372 dtype, pl, istate);
1373 if (pattern) {
1374 if (pattern->flags & PATTERN_FLAG_NEGATIVE)
1375 return NOT_MATCHED;
1376 else
1377 return MATCHED;
1380 return UNDECIDED;
1383 if (pl->full_cone)
1384 return MATCHED;
1386 strbuf_addch(&parent_pathname, '/');
1387 strbuf_add(&parent_pathname, pathname, pathlen);
1389 if (hashmap_contains_path(&pl->recursive_hashmap,
1390 &parent_pathname)) {
1391 result = MATCHED_RECURSIVE;
1392 goto done;
1395 slash_pos = strrchr(parent_pathname.buf, '/');
1397 if (slash_pos == parent_pathname.buf) {
1398 /* include every file in root */
1399 result = MATCHED;
1400 goto done;
1403 strbuf_setlen(&parent_pathname, slash_pos - parent_pathname.buf);
1405 if (hashmap_contains_path(&pl->parent_hashmap, &parent_pathname)) {
1406 result = MATCHED;
1407 goto done;
1410 if (hashmap_contains_parent(&pl->recursive_hashmap,
1411 pathname,
1412 &parent_pathname))
1413 result = MATCHED_RECURSIVE;
1415 done:
1416 strbuf_release(&parent_pathname);
1417 return result;
1420 static struct path_pattern *last_matching_pattern_from_lists(
1421 struct dir_struct *dir, struct index_state *istate,
1422 const char *pathname, int pathlen,
1423 const char *basename, int *dtype_p)
1425 int i, j;
1426 struct exclude_list_group *group;
1427 struct path_pattern *pattern;
1428 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1429 group = &dir->exclude_list_group[i];
1430 for (j = group->nr - 1; j >= 0; j--) {
1431 pattern = last_matching_pattern_from_list(
1432 pathname, pathlen, basename, dtype_p,
1433 &group->pl[j], istate);
1434 if (pattern)
1435 return pattern;
1438 return NULL;
1442 * Loads the per-directory exclude list for the substring of base
1443 * which has a char length of baselen.
1445 static void prep_exclude(struct dir_struct *dir,
1446 struct index_state *istate,
1447 const char *base, int baselen)
1449 struct exclude_list_group *group;
1450 struct pattern_list *pl;
1451 struct exclude_stack *stk = NULL;
1452 struct untracked_cache_dir *untracked;
1453 int current;
1455 group = &dir->exclude_list_group[EXC_DIRS];
1458 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1459 * which originate from directories not in the prefix of the
1460 * path being checked.
1462 while ((stk = dir->exclude_stack) != NULL) {
1463 if (stk->baselen <= baselen &&
1464 !strncmp(dir->basebuf.buf, base, stk->baselen))
1465 break;
1466 pl = &group->pl[dir->exclude_stack->exclude_ix];
1467 dir->exclude_stack = stk->prev;
1468 dir->pattern = NULL;
1469 free((char *)pl->src); /* see strbuf_detach() below */
1470 clear_pattern_list(pl);
1471 free(stk);
1472 group->nr--;
1475 /* Skip traversing into sub directories if the parent is excluded */
1476 if (dir->pattern)
1477 return;
1480 * Lazy initialization. All call sites currently just
1481 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1482 * them seems lots of work for little benefit.
1484 if (!dir->basebuf.buf)
1485 strbuf_init(&dir->basebuf, PATH_MAX);
1487 /* Read from the parent directories and push them down. */
1488 current = stk ? stk->baselen : -1;
1489 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1490 if (dir->untracked)
1491 untracked = stk ? stk->ucd : dir->untracked->root;
1492 else
1493 untracked = NULL;
1495 while (current < baselen) {
1496 const char *cp;
1497 struct oid_stat oid_stat;
1499 CALLOC_ARRAY(stk, 1);
1500 if (current < 0) {
1501 cp = base;
1502 current = 0;
1503 } else {
1504 cp = strchr(base + current + 1, '/');
1505 if (!cp)
1506 die("oops in prep_exclude");
1507 cp++;
1508 untracked =
1509 lookup_untracked(dir->untracked, untracked,
1510 base + current,
1511 cp - base - current);
1513 stk->prev = dir->exclude_stack;
1514 stk->baselen = cp - base;
1515 stk->exclude_ix = group->nr;
1516 stk->ucd = untracked;
1517 pl = add_pattern_list(dir, EXC_DIRS, NULL);
1518 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1519 assert(stk->baselen == dir->basebuf.len);
1521 /* Abort if the directory is excluded */
1522 if (stk->baselen) {
1523 int dt = DT_DIR;
1524 dir->basebuf.buf[stk->baselen - 1] = 0;
1525 dir->pattern = last_matching_pattern_from_lists(dir,
1526 istate,
1527 dir->basebuf.buf, stk->baselen - 1,
1528 dir->basebuf.buf + current, &dt);
1529 dir->basebuf.buf[stk->baselen - 1] = '/';
1530 if (dir->pattern &&
1531 dir->pattern->flags & PATTERN_FLAG_NEGATIVE)
1532 dir->pattern = NULL;
1533 if (dir->pattern) {
1534 dir->exclude_stack = stk;
1535 return;
1539 /* Try to read per-directory file */
1540 oidclr(&oid_stat.oid);
1541 oid_stat.valid = 0;
1542 if (dir->exclude_per_dir &&
1544 * If we know that no files have been added in
1545 * this directory (i.e. valid_cached_dir() has
1546 * been executed and set untracked->valid) ..
1548 (!untracked || !untracked->valid ||
1550 * .. and .gitignore does not exist before
1551 * (i.e. null exclude_oid). Then we can skip
1552 * loading .gitignore, which would result in
1553 * ENOENT anyway.
1555 !is_null_oid(&untracked->exclude_oid))) {
1557 * dir->basebuf gets reused by the traversal, but we
1558 * need fname to remain unchanged to ensure the src
1559 * member of each struct path_pattern correctly
1560 * back-references its source file. Other invocations
1561 * of add_pattern_list provide stable strings, so we
1562 * strbuf_detach() and free() here in the caller.
1564 struct strbuf sb = STRBUF_INIT;
1565 strbuf_addbuf(&sb, &dir->basebuf);
1566 strbuf_addstr(&sb, dir->exclude_per_dir);
1567 pl->src = strbuf_detach(&sb, NULL);
1568 add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1569 PATTERN_NOFOLLOW,
1570 untracked ? &oid_stat : NULL);
1573 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1574 * will first be called in valid_cached_dir() then maybe many
1575 * times more in last_matching_pattern(). When the cache is
1576 * used, last_matching_pattern() will not be called and
1577 * reading .gitignore content will be a waste.
1579 * So when it's called by valid_cached_dir() and we can get
1580 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1581 * modified on work tree), we could delay reading the
1582 * .gitignore content until we absolutely need it in
1583 * last_matching_pattern(). Be careful about ignore rule
1584 * order, though, if you do that.
1586 if (untracked &&
1587 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1588 invalidate_gitignore(dir->untracked, untracked);
1589 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1591 dir->exclude_stack = stk;
1592 current = stk->baselen;
1594 strbuf_setlen(&dir->basebuf, baselen);
1598 * Loads the exclude lists for the directory containing pathname, then
1599 * scans all exclude lists to determine whether pathname is excluded.
1600 * Returns the exclude_list element which matched, or NULL for
1601 * undecided.
1603 struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1604 struct index_state *istate,
1605 const char *pathname,
1606 int *dtype_p)
1608 int pathlen = strlen(pathname);
1609 const char *basename = strrchr(pathname, '/');
1610 basename = (basename) ? basename+1 : pathname;
1612 prep_exclude(dir, istate, pathname, basename-pathname);
1614 if (dir->pattern)
1615 return dir->pattern;
1617 return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1618 basename, dtype_p);
1622 * Loads the exclude lists for the directory containing pathname, then
1623 * scans all exclude lists to determine whether pathname is excluded.
1624 * Returns 1 if true, otherwise 0.
1626 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1627 const char *pathname, int *dtype_p)
1629 struct path_pattern *pattern =
1630 last_matching_pattern(dir, istate, pathname, dtype_p);
1631 if (pattern)
1632 return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1633 return 0;
1636 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1638 struct dir_entry *ent;
1640 FLEX_ALLOC_MEM(ent, name, pathname, len);
1641 ent->len = len;
1642 return ent;
1645 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1646 struct index_state *istate,
1647 const char *pathname, int len)
1649 if (index_file_exists(istate, pathname, len, ignore_case))
1650 return NULL;
1652 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1653 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1656 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1657 struct index_state *istate,
1658 const char *pathname, int len)
1660 if (!index_name_is_other(istate, pathname, len))
1661 return NULL;
1663 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1664 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1667 enum exist_status {
1668 index_nonexistent = 0,
1669 index_directory,
1670 index_gitdir
1674 * Do not use the alphabetically sorted index to look up
1675 * the directory name; instead, use the case insensitive
1676 * directory hash.
1678 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1679 const char *dirname, int len)
1681 struct cache_entry *ce;
1683 if (index_dir_exists(istate, dirname, len))
1684 return index_directory;
1686 ce = index_file_exists(istate, dirname, len, ignore_case);
1687 if (ce && S_ISGITLINK(ce->ce_mode))
1688 return index_gitdir;
1690 return index_nonexistent;
1694 * The index sorts alphabetically by entry name, which
1695 * means that a gitlink sorts as '\0' at the end, while
1696 * a directory (which is defined not as an entry, but as
1697 * the files it contains) will sort with the '/' at the
1698 * end.
1700 static enum exist_status directory_exists_in_index(struct index_state *istate,
1701 const char *dirname, int len)
1703 int pos;
1705 if (ignore_case)
1706 return directory_exists_in_index_icase(istate, dirname, len);
1708 pos = index_name_pos(istate, dirname, len);
1709 if (pos < 0)
1710 pos = -pos-1;
1711 while (pos < istate->cache_nr) {
1712 const struct cache_entry *ce = istate->cache[pos++];
1713 unsigned char endchar;
1715 if (strncmp(ce->name, dirname, len))
1716 break;
1717 endchar = ce->name[len];
1718 if (endchar > '/')
1719 break;
1720 if (endchar == '/')
1721 return index_directory;
1722 if (!endchar && S_ISGITLINK(ce->ce_mode))
1723 return index_gitdir;
1725 return index_nonexistent;
1729 * When we find a directory when traversing the filesystem, we
1730 * have three distinct cases:
1732 * - ignore it
1733 * - see it as a directory
1734 * - recurse into it
1736 * and which one we choose depends on a combination of existing
1737 * git index contents and the flags passed into the directory
1738 * traversal routine.
1740 * Case 1: If we *already* have entries in the index under that
1741 * directory name, we always recurse into the directory to see
1742 * all the files.
1744 * Case 2: If we *already* have that directory name as a gitlink,
1745 * we always continue to see it as a gitlink, regardless of whether
1746 * there is an actual git directory there or not (it might not
1747 * be checked out as a subproject!)
1749 * Case 3: if we didn't have it in the index previously, we
1750 * have a few sub-cases:
1752 * (a) if "show_other_directories" is true, we show it as
1753 * just a directory, unless "hide_empty_directories" is
1754 * also true, in which case we need to check if it contains any
1755 * untracked and / or ignored files.
1756 * (b) if it looks like a git directory, and we don't have
1757 * 'no_gitlinks' set we treat it as a gitlink, and show it
1758 * as a directory.
1759 * (c) otherwise, we recurse into it.
1761 static enum path_treatment treat_directory(struct dir_struct *dir,
1762 struct index_state *istate,
1763 struct untracked_cache_dir *untracked,
1764 const char *dirname, int len, int baselen, int excluded,
1765 const struct pathspec *pathspec)
1768 * WARNING: From this function, you can return path_recurse or you
1769 * can call read_directory_recursive() (or neither), but
1770 * you CAN'T DO BOTH.
1772 enum path_treatment state;
1773 int matches_how = 0;
1774 int nested_repo = 0, check_only, stop_early;
1775 int old_ignored_nr, old_untracked_nr;
1776 /* The "len-1" is to strip the final '/' */
1777 enum exist_status status = directory_exists_in_index(istate, dirname, len-1);
1779 if (status == index_directory)
1780 return path_recurse;
1781 if (status == index_gitdir)
1782 return path_none;
1783 if (status != index_nonexistent)
1784 BUG("Unhandled value for directory_exists_in_index: %d\n", status);
1787 * We don't want to descend into paths that don't match the necessary
1788 * patterns. Clearly, if we don't have a pathspec, then we can't check
1789 * for matching patterns. Also, if (excluded) then we know we matched
1790 * the exclusion patterns so as an optimization we can skip checking
1791 * for matching patterns.
1793 if (pathspec && !excluded) {
1794 matches_how = match_pathspec_with_flags(istate, pathspec,
1795 dirname, len,
1796 0 /* prefix */,
1797 NULL /* seen */,
1798 DO_MATCH_LEADING_PATHSPEC);
1799 if (!matches_how)
1800 return path_none;
1804 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1805 !(dir->flags & DIR_NO_GITLINKS)) {
1806 struct strbuf sb = STRBUF_INIT;
1807 strbuf_addstr(&sb, dirname);
1808 nested_repo = is_nonbare_repository_dir(&sb);
1809 strbuf_release(&sb);
1811 if (nested_repo) {
1812 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1813 (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC))
1814 return path_none;
1815 return excluded ? path_excluded : path_untracked;
1818 if (!(dir->flags & DIR_SHOW_OTHER_DIRECTORIES)) {
1819 if (excluded &&
1820 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1821 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1824 * This is an excluded directory and we are
1825 * showing ignored paths that match an exclude
1826 * pattern. (e.g. show directory as ignored
1827 * only if it matches an exclude pattern).
1828 * This path will either be 'path_excluded`
1829 * (if we are showing empty directories or if
1830 * the directory is not empty), or will be
1831 * 'path_none' (empty directory, and we are
1832 * not showing empty directories).
1834 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1835 return path_excluded;
1837 if (read_directory_recursive(dir, istate, dirname, len,
1838 untracked, 1, 1, pathspec) == path_excluded)
1839 return path_excluded;
1841 return path_none;
1843 return path_recurse;
1846 /* This is the "show_other_directories" case */
1849 * If we have a pathspec which could match something _below_ this
1850 * directory (e.g. when checking 'subdir/' having a pathspec like
1851 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1852 * need to recurse.
1854 if (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC)
1855 return path_recurse;
1858 * Other than the path_recurse case immediately above, we only need
1859 * to recurse into untracked/ignored directories if either of the
1860 * following bits is set:
1861 * - DIR_SHOW_IGNORED_TOO (because then we need to determine if
1862 * there are ignored entries below)
1863 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
1864 * the directory is empty)
1866 if (!(dir->flags & (DIR_SHOW_IGNORED_TOO | DIR_HIDE_EMPTY_DIRECTORIES)))
1867 return excluded ? path_excluded : path_untracked;
1870 * ...and even if DIR_SHOW_IGNORED_TOO is set, we can still avoid
1871 * recursing into ignored directories if the path is excluded and
1872 * DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1874 if (excluded &&
1875 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1876 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1877 return path_excluded;
1880 * Even if we don't want to know all the paths under an untracked or
1881 * ignored directory, we may still need to go into the directory to
1882 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
1883 * an empty directory should be path_none instead of path_excluded or
1884 * path_untracked).
1886 check_only = ((dir->flags & DIR_HIDE_EMPTY_DIRECTORIES) &&
1887 !(dir->flags & DIR_SHOW_IGNORED_TOO));
1890 * However, there's another optimization possible as a subset of
1891 * check_only, based on the cases we have to consider:
1892 * A) Directory matches no exclude patterns:
1893 * * Directory is empty => path_none
1894 * * Directory has an untracked file under it => path_untracked
1895 * * Directory has only ignored files under it => path_excluded
1896 * B) Directory matches an exclude pattern:
1897 * * Directory is empty => path_none
1898 * * Directory has an untracked file under it => path_excluded
1899 * * Directory has only ignored files under it => path_excluded
1900 * In case A, we can exit as soon as we've found an untracked
1901 * file but otherwise have to walk all files. In case B, though,
1902 * we can stop at the first file we find under the directory.
1904 stop_early = check_only && excluded;
1907 * If /every/ file within an untracked directory is ignored, then
1908 * we want to treat the directory as ignored (for e.g. status
1909 * --porcelain), without listing the individual ignored files
1910 * underneath. To do so, we'll save the current ignored_nr, and
1911 * pop all the ones added after it if it turns out the entire
1912 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
1913 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
1914 * untracked paths so will need to pop all those off the last
1915 * after we traverse.
1917 old_ignored_nr = dir->ignored_nr;
1918 old_untracked_nr = dir->nr;
1920 /* Actually recurse into dirname now, we'll fixup the state later. */
1921 untracked = lookup_untracked(dir->untracked, untracked,
1922 dirname + baselen, len - baselen);
1923 state = read_directory_recursive(dir, istate, dirname, len, untracked,
1924 check_only, stop_early, pathspec);
1926 /* There are a variety of reasons we may need to fixup the state... */
1927 if (state == path_excluded) {
1928 /* state == path_excluded implies all paths under
1929 * dirname were ignored...
1931 * if running e.g. `git status --porcelain --ignored=matching`,
1932 * then we want to see the subpaths that are ignored.
1934 * if running e.g. just `git status --porcelain`, then
1935 * we just want the directory itself to be listed as ignored
1936 * and not the individual paths underneath.
1938 int want_ignored_subpaths =
1939 ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
1940 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING));
1942 if (want_ignored_subpaths) {
1944 * with --ignored=matching, we want the subpaths
1945 * INSTEAD of the directory itself.
1947 state = path_none;
1948 } else {
1949 int i;
1950 for (i = old_ignored_nr + 1; i<dir->ignored_nr; ++i)
1951 FREE_AND_NULL(dir->ignored[i]);
1952 dir->ignored_nr = old_ignored_nr;
1957 * We may need to ignore some of the untracked paths we found while
1958 * traversing subdirectories.
1960 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
1961 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
1962 int i;
1963 for (i = old_untracked_nr + 1; i<dir->nr; ++i)
1964 FREE_AND_NULL(dir->entries[i]);
1965 dir->nr = old_untracked_nr;
1969 * If there is nothing under the current directory and we are not
1970 * hiding empty directories, then we need to report on the
1971 * untracked or ignored status of the directory itself.
1973 if (state == path_none && !(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1974 state = excluded ? path_excluded : path_untracked;
1976 return state;
1980 * This is an inexact early pruning of any recursive directory
1981 * reading - if the path cannot possibly be in the pathspec,
1982 * return true, and we'll skip it early.
1984 static int simplify_away(const char *path, int pathlen,
1985 const struct pathspec *pathspec)
1987 int i;
1989 if (!pathspec || !pathspec->nr)
1990 return 0;
1992 GUARD_PATHSPEC(pathspec,
1993 PATHSPEC_FROMTOP |
1994 PATHSPEC_MAXDEPTH |
1995 PATHSPEC_LITERAL |
1996 PATHSPEC_GLOB |
1997 PATHSPEC_ICASE |
1998 PATHSPEC_EXCLUDE |
1999 PATHSPEC_ATTR);
2001 for (i = 0; i < pathspec->nr; i++) {
2002 const struct pathspec_item *item = &pathspec->items[i];
2003 int len = item->nowildcard_len;
2005 if (len > pathlen)
2006 len = pathlen;
2007 if (!ps_strncmp(item, item->match, path, len))
2008 return 0;
2011 return 1;
2015 * This function tells us whether an excluded path matches a
2016 * list of "interesting" pathspecs. That is, whether a path matched
2017 * by any of the pathspecs could possibly be ignored by excluding
2018 * the specified path. This can happen if:
2020 * 1. the path is mentioned explicitly in the pathspec
2022 * 2. the path is a directory prefix of some element in the
2023 * pathspec
2025 static int exclude_matches_pathspec(const char *path, int pathlen,
2026 const struct pathspec *pathspec)
2028 int i;
2030 if (!pathspec || !pathspec->nr)
2031 return 0;
2033 GUARD_PATHSPEC(pathspec,
2034 PATHSPEC_FROMTOP |
2035 PATHSPEC_MAXDEPTH |
2036 PATHSPEC_LITERAL |
2037 PATHSPEC_GLOB |
2038 PATHSPEC_ICASE |
2039 PATHSPEC_EXCLUDE);
2041 for (i = 0; i < pathspec->nr; i++) {
2042 const struct pathspec_item *item = &pathspec->items[i];
2043 int len = item->nowildcard_len;
2045 if (len == pathlen &&
2046 !ps_strncmp(item, item->match, path, pathlen))
2047 return 1;
2048 if (len > pathlen &&
2049 item->match[pathlen] == '/' &&
2050 !ps_strncmp(item, item->match, path, pathlen))
2051 return 1;
2053 return 0;
2056 static int get_index_dtype(struct index_state *istate,
2057 const char *path, int len)
2059 int pos;
2060 const struct cache_entry *ce;
2062 ce = index_file_exists(istate, path, len, 0);
2063 if (ce) {
2064 if (!ce_uptodate(ce))
2065 return DT_UNKNOWN;
2066 if (S_ISGITLINK(ce->ce_mode))
2067 return DT_DIR;
2069 * Nobody actually cares about the
2070 * difference between DT_LNK and DT_REG
2072 return DT_REG;
2075 /* Try to look it up as a directory */
2076 pos = index_name_pos(istate, path, len);
2077 if (pos >= 0)
2078 return DT_UNKNOWN;
2079 pos = -pos-1;
2080 while (pos < istate->cache_nr) {
2081 ce = istate->cache[pos++];
2082 if (strncmp(ce->name, path, len))
2083 break;
2084 if (ce->name[len] > '/')
2085 break;
2086 if (ce->name[len] < '/')
2087 continue;
2088 if (!ce_uptodate(ce))
2089 break; /* continue? */
2090 return DT_DIR;
2092 return DT_UNKNOWN;
2095 static int resolve_dtype(int dtype, struct index_state *istate,
2096 const char *path, int len)
2098 struct stat st;
2100 if (dtype != DT_UNKNOWN)
2101 return dtype;
2102 dtype = get_index_dtype(istate, path, len);
2103 if (dtype != DT_UNKNOWN)
2104 return dtype;
2105 if (lstat(path, &st))
2106 return dtype;
2107 if (S_ISREG(st.st_mode))
2108 return DT_REG;
2109 if (S_ISDIR(st.st_mode))
2110 return DT_DIR;
2111 if (S_ISLNK(st.st_mode))
2112 return DT_LNK;
2113 return dtype;
2116 static enum path_treatment treat_path_fast(struct dir_struct *dir,
2117 struct cached_dir *cdir,
2118 struct index_state *istate,
2119 struct strbuf *path,
2120 int baselen,
2121 const struct pathspec *pathspec)
2124 * WARNING: From this function, you can return path_recurse or you
2125 * can call read_directory_recursive() (or neither), but
2126 * you CAN'T DO BOTH.
2128 strbuf_setlen(path, baselen);
2129 if (!cdir->ucd) {
2130 strbuf_addstr(path, cdir->file);
2131 return path_untracked;
2133 strbuf_addstr(path, cdir->ucd->name);
2134 /* treat_one_path() does this before it calls treat_directory() */
2135 strbuf_complete(path, '/');
2136 if (cdir->ucd->check_only)
2138 * check_only is set as a result of treat_directory() getting
2139 * to its bottom. Verify again the same set of directories
2140 * with check_only set.
2142 return read_directory_recursive(dir, istate, path->buf, path->len,
2143 cdir->ucd, 1, 0, pathspec);
2145 * We get path_recurse in the first run when
2146 * directory_exists_in_index() returns index_nonexistent. We
2147 * are sure that new changes in the index does not impact the
2148 * outcome. Return now.
2150 return path_recurse;
2153 static enum path_treatment treat_path(struct dir_struct *dir,
2154 struct untracked_cache_dir *untracked,
2155 struct cached_dir *cdir,
2156 struct index_state *istate,
2157 struct strbuf *path,
2158 int baselen,
2159 const struct pathspec *pathspec)
2161 int has_path_in_index, dtype, excluded;
2163 if (!cdir->d_name)
2164 return treat_path_fast(dir, cdir, istate, path,
2165 baselen, pathspec);
2166 if (is_dot_or_dotdot(cdir->d_name) || !fspathcmp(cdir->d_name, ".git"))
2167 return path_none;
2168 strbuf_setlen(path, baselen);
2169 strbuf_addstr(path, cdir->d_name);
2170 if (simplify_away(path->buf, path->len, pathspec))
2171 return path_none;
2173 dtype = resolve_dtype(cdir->d_type, istate, path->buf, path->len);
2175 /* Always exclude indexed files */
2176 has_path_in_index = !!index_file_exists(istate, path->buf, path->len,
2177 ignore_case);
2178 if (dtype != DT_DIR && has_path_in_index)
2179 return path_none;
2182 * When we are looking at a directory P in the working tree,
2183 * there are three cases:
2185 * (1) P exists in the index. Everything inside the directory P in
2186 * the working tree needs to go when P is checked out from the
2187 * index.
2189 * (2) P does not exist in the index, but there is P/Q in the index.
2190 * We know P will stay a directory when we check out the contents
2191 * of the index, but we do not know yet if there is a directory
2192 * P/Q in the working tree to be killed, so we need to recurse.
2194 * (3) P does not exist in the index, and there is no P/Q in the index
2195 * to require P to be a directory, either. Only in this case, we
2196 * know that everything inside P will not be killed without
2197 * recursing.
2199 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
2200 (dtype == DT_DIR) &&
2201 !has_path_in_index &&
2202 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
2203 return path_none;
2205 excluded = is_excluded(dir, istate, path->buf, &dtype);
2208 * Excluded? If we don't explicitly want to show
2209 * ignored files, ignore it
2211 if (excluded && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
2212 return path_excluded;
2214 switch (dtype) {
2215 default:
2216 return path_none;
2217 case DT_DIR:
2219 * WARNING: Do not ignore/amend the return value from
2220 * treat_directory(), and especially do not change it to return
2221 * path_recurse as that can cause exponential slowdown.
2222 * Instead, modify treat_directory() to return the right value.
2224 strbuf_addch(path, '/');
2225 return treat_directory(dir, istate, untracked,
2226 path->buf, path->len,
2227 baselen, excluded, pathspec);
2228 case DT_REG:
2229 case DT_LNK:
2230 if (pathspec &&
2231 !match_pathspec(istate, pathspec, path->buf, path->len,
2232 0 /* prefix */, NULL /* seen */,
2233 0 /* is_dir */))
2234 return path_none;
2235 if (excluded)
2236 return path_excluded;
2237 return path_untracked;
2241 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
2243 if (!dir)
2244 return;
2245 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
2246 dir->untracked_alloc);
2247 dir->untracked[dir->untracked_nr++] = xstrdup(name);
2250 static int valid_cached_dir(struct dir_struct *dir,
2251 struct untracked_cache_dir *untracked,
2252 struct index_state *istate,
2253 struct strbuf *path,
2254 int check_only)
2256 struct stat st;
2258 if (!untracked)
2259 return 0;
2262 * With fsmonitor, we can trust the untracked cache's valid field.
2264 refresh_fsmonitor(istate);
2265 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
2266 if (lstat(path->len ? path->buf : ".", &st)) {
2267 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
2268 return 0;
2270 if (!untracked->valid ||
2271 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
2272 fill_stat_data(&untracked->stat_data, &st);
2273 return 0;
2277 if (untracked->check_only != !!check_only)
2278 return 0;
2281 * prep_exclude will be called eventually on this directory,
2282 * but it's called much later in last_matching_pattern(). We
2283 * need it now to determine the validity of the cache for this
2284 * path. The next calls will be nearly no-op, the way
2285 * prep_exclude() is designed.
2287 if (path->len && path->buf[path->len - 1] != '/') {
2288 strbuf_addch(path, '/');
2289 prep_exclude(dir, istate, path->buf, path->len);
2290 strbuf_setlen(path, path->len - 1);
2291 } else
2292 prep_exclude(dir, istate, path->buf, path->len);
2294 /* hopefully prep_exclude() haven't invalidated this entry... */
2295 return untracked->valid;
2298 static int open_cached_dir(struct cached_dir *cdir,
2299 struct dir_struct *dir,
2300 struct untracked_cache_dir *untracked,
2301 struct index_state *istate,
2302 struct strbuf *path,
2303 int check_only)
2305 const char *c_path;
2307 memset(cdir, 0, sizeof(*cdir));
2308 cdir->untracked = untracked;
2309 if (valid_cached_dir(dir, untracked, istate, path, check_only))
2310 return 0;
2311 c_path = path->len ? path->buf : ".";
2312 cdir->fdir = opendir(c_path);
2313 if (!cdir->fdir)
2314 warning_errno(_("could not open directory '%s'"), c_path);
2315 if (dir->untracked) {
2316 invalidate_directory(dir->untracked, untracked);
2317 dir->untracked->dir_opened++;
2319 if (!cdir->fdir)
2320 return -1;
2321 return 0;
2324 static int read_cached_dir(struct cached_dir *cdir)
2326 struct dirent *de;
2328 if (cdir->fdir) {
2329 de = readdir(cdir->fdir);
2330 if (!de) {
2331 cdir->d_name = NULL;
2332 cdir->d_type = DT_UNKNOWN;
2333 return -1;
2335 cdir->d_name = de->d_name;
2336 cdir->d_type = DTYPE(de);
2337 return 0;
2339 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
2340 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
2341 if (!d->recurse) {
2342 cdir->nr_dirs++;
2343 continue;
2345 cdir->ucd = d;
2346 cdir->nr_dirs++;
2347 return 0;
2349 cdir->ucd = NULL;
2350 if (cdir->nr_files < cdir->untracked->untracked_nr) {
2351 struct untracked_cache_dir *d = cdir->untracked;
2352 cdir->file = d->untracked[cdir->nr_files++];
2353 return 0;
2355 return -1;
2358 static void close_cached_dir(struct cached_dir *cdir)
2360 if (cdir->fdir)
2361 closedir(cdir->fdir);
2363 * We have gone through this directory and found no untracked
2364 * entries. Mark it valid.
2366 if (cdir->untracked) {
2367 cdir->untracked->valid = 1;
2368 cdir->untracked->recurse = 1;
2372 static void add_path_to_appropriate_result_list(struct dir_struct *dir,
2373 struct untracked_cache_dir *untracked,
2374 struct cached_dir *cdir,
2375 struct index_state *istate,
2376 struct strbuf *path,
2377 int baselen,
2378 const struct pathspec *pathspec,
2379 enum path_treatment state)
2381 /* add the path to the appropriate result list */
2382 switch (state) {
2383 case path_excluded:
2384 if (dir->flags & DIR_SHOW_IGNORED)
2385 dir_add_name(dir, istate, path->buf, path->len);
2386 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2387 ((dir->flags & DIR_COLLECT_IGNORED) &&
2388 exclude_matches_pathspec(path->buf, path->len,
2389 pathspec)))
2390 dir_add_ignored(dir, istate, path->buf, path->len);
2391 break;
2393 case path_untracked:
2394 if (dir->flags & DIR_SHOW_IGNORED)
2395 break;
2396 dir_add_name(dir, istate, path->buf, path->len);
2397 if (cdir->fdir)
2398 add_untracked(untracked, path->buf + baselen);
2399 break;
2401 default:
2402 break;
2407 * Read a directory tree. We currently ignore anything but
2408 * directories, regular files and symlinks. That's because git
2409 * doesn't handle them at all yet. Maybe that will change some
2410 * day.
2412 * Also, we ignore the name ".git" (even if it is not a directory).
2413 * That likely will not change.
2415 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2416 * to signal that a file was found. This is the least significant value that
2417 * indicates that a file was encountered that does not depend on the order of
2418 * whether an untracked or excluded path was encountered first.
2420 * Returns the most significant path_treatment value encountered in the scan.
2421 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2422 * significant path_treatment value that will be returned.
2425 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
2426 struct index_state *istate, const char *base, int baselen,
2427 struct untracked_cache_dir *untracked, int check_only,
2428 int stop_at_first_file, const struct pathspec *pathspec)
2431 * WARNING: Do NOT recurse unless path_recurse is returned from
2432 * treat_path(). Recursing on any other return value
2433 * can result in exponential slowdown.
2435 struct cached_dir cdir;
2436 enum path_treatment state, subdir_state, dir_state = path_none;
2437 struct strbuf path = STRBUF_INIT;
2439 strbuf_add(&path, base, baselen);
2441 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
2442 goto out;
2444 if (untracked)
2445 untracked->check_only = !!check_only;
2447 while (!read_cached_dir(&cdir)) {
2448 /* check how the file or directory should be treated */
2449 state = treat_path(dir, untracked, &cdir, istate, &path,
2450 baselen, pathspec);
2452 if (state > dir_state)
2453 dir_state = state;
2455 /* recurse into subdir if instructed by treat_path */
2456 if (state == path_recurse) {
2457 struct untracked_cache_dir *ud;
2458 ud = lookup_untracked(dir->untracked, untracked,
2459 path.buf + baselen,
2460 path.len - baselen);
2461 subdir_state =
2462 read_directory_recursive(dir, istate, path.buf,
2463 path.len, ud,
2464 check_only, stop_at_first_file, pathspec);
2465 if (subdir_state > dir_state)
2466 dir_state = subdir_state;
2468 if (pathspec &&
2469 !match_pathspec(istate, pathspec, path.buf, path.len,
2470 0 /* prefix */, NULL,
2471 0 /* do NOT special case dirs */))
2472 state = path_none;
2475 if (check_only) {
2476 if (stop_at_first_file) {
2478 * If stopping at first file, then
2479 * signal that a file was found by
2480 * returning `path_excluded`. This is
2481 * to return a consistent value
2482 * regardless of whether an ignored or
2483 * excluded file happened to be
2484 * encountered 1st.
2486 * In current usage, the
2487 * `stop_at_first_file` is passed when
2488 * an ancestor directory has matched
2489 * an exclude pattern, so any found
2490 * files will be excluded.
2492 if (dir_state >= path_excluded) {
2493 dir_state = path_excluded;
2494 break;
2498 /* abort early if maximum state has been reached */
2499 if (dir_state == path_untracked) {
2500 if (cdir.fdir)
2501 add_untracked(untracked, path.buf + baselen);
2502 break;
2504 /* skip the add_path_to_appropriate_result_list() */
2505 continue;
2508 add_path_to_appropriate_result_list(dir, untracked, &cdir,
2509 istate, &path, baselen,
2510 pathspec, state);
2512 close_cached_dir(&cdir);
2513 out:
2514 strbuf_release(&path);
2516 return dir_state;
2519 int cmp_dir_entry(const void *p1, const void *p2)
2521 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2522 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2524 return name_compare(e1->name, e1->len, e2->name, e2->len);
2527 /* check if *out lexically strictly contains *in */
2528 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2530 return (out->len < in->len) &&
2531 (out->name[out->len - 1] == '/') &&
2532 !memcmp(out->name, in->name, out->len);
2535 static int treat_leading_path(struct dir_struct *dir,
2536 struct index_state *istate,
2537 const char *path, int len,
2538 const struct pathspec *pathspec)
2540 struct strbuf sb = STRBUF_INIT;
2541 struct strbuf subdir = STRBUF_INIT;
2542 int prevlen, baselen;
2543 const char *cp;
2544 struct cached_dir cdir;
2545 enum path_treatment state = path_none;
2548 * For each directory component of path, we are going to check whether
2549 * that path is relevant given the pathspec. For example, if path is
2550 * foo/bar/baz/
2551 * then we will ask treat_path() whether we should go into foo, then
2552 * whether we should go into bar, then whether baz is relevant.
2553 * Checking each is important because e.g. if path is
2554 * .git/info/
2555 * then we need to check .git to know we shouldn't traverse it.
2556 * If the return from treat_path() is:
2557 * * path_none, for any path, we return false.
2558 * * path_recurse, for all path components, we return true
2559 * * <anything else> for some intermediate component, we make sure
2560 * to add that path to the relevant list but return false
2561 * signifying that we shouldn't recurse into it.
2564 while (len && path[len - 1] == '/')
2565 len--;
2566 if (!len)
2567 return 1;
2569 memset(&cdir, 0, sizeof(cdir));
2570 cdir.d_type = DT_DIR;
2571 baselen = 0;
2572 prevlen = 0;
2573 while (1) {
2574 prevlen = baselen + !!baselen;
2575 cp = path + prevlen;
2576 cp = memchr(cp, '/', path + len - cp);
2577 if (!cp)
2578 baselen = len;
2579 else
2580 baselen = cp - path;
2581 strbuf_reset(&sb);
2582 strbuf_add(&sb, path, baselen);
2583 if (!is_directory(sb.buf))
2584 break;
2585 strbuf_reset(&sb);
2586 strbuf_add(&sb, path, prevlen);
2587 strbuf_reset(&subdir);
2588 strbuf_add(&subdir, path+prevlen, baselen-prevlen);
2589 cdir.d_name = subdir.buf;
2590 state = treat_path(dir, NULL, &cdir, istate, &sb, prevlen, pathspec);
2592 if (state != path_recurse)
2593 break; /* do not recurse into it */
2594 if (len <= baselen)
2595 break; /* finished checking */
2597 add_path_to_appropriate_result_list(dir, NULL, &cdir, istate,
2598 &sb, baselen, pathspec,
2599 state);
2601 strbuf_release(&subdir);
2602 strbuf_release(&sb);
2603 return state == path_recurse;
2606 static const char *get_ident_string(void)
2608 static struct strbuf sb = STRBUF_INIT;
2609 struct utsname uts;
2611 if (sb.len)
2612 return sb.buf;
2613 if (uname(&uts) < 0)
2614 die_errno(_("failed to get kernel name and information"));
2615 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2616 uts.sysname);
2617 return sb.buf;
2620 static int ident_in_untracked(const struct untracked_cache *uc)
2623 * Previous git versions may have saved many NUL separated
2624 * strings in the "ident" field, but it is insane to manage
2625 * many locations, so just take care of the first one.
2628 return !strcmp(uc->ident.buf, get_ident_string());
2631 static void set_untracked_ident(struct untracked_cache *uc)
2633 strbuf_reset(&uc->ident);
2634 strbuf_addstr(&uc->ident, get_ident_string());
2637 * This strbuf used to contain a list of NUL separated
2638 * strings, so save NUL too for backward compatibility.
2640 strbuf_addch(&uc->ident, 0);
2643 static void new_untracked_cache(struct index_state *istate)
2645 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2646 strbuf_init(&uc->ident, 100);
2647 uc->exclude_per_dir = ".gitignore";
2648 /* should be the same flags used by git-status */
2649 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2650 set_untracked_ident(uc);
2651 istate->untracked = uc;
2652 istate->cache_changed |= UNTRACKED_CHANGED;
2655 void add_untracked_cache(struct index_state *istate)
2657 if (!istate->untracked) {
2658 new_untracked_cache(istate);
2659 } else {
2660 if (!ident_in_untracked(istate->untracked)) {
2661 free_untracked_cache(istate->untracked);
2662 new_untracked_cache(istate);
2667 void remove_untracked_cache(struct index_state *istate)
2669 if (istate->untracked) {
2670 free_untracked_cache(istate->untracked);
2671 istate->untracked = NULL;
2672 istate->cache_changed |= UNTRACKED_CHANGED;
2676 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2677 int base_len,
2678 const struct pathspec *pathspec)
2680 struct untracked_cache_dir *root;
2681 static int untracked_cache_disabled = -1;
2683 if (!dir->untracked)
2684 return NULL;
2685 if (untracked_cache_disabled < 0)
2686 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2687 if (untracked_cache_disabled)
2688 return NULL;
2691 * We only support $GIT_DIR/info/exclude and core.excludesfile
2692 * as the global ignore rule files. Any other additions
2693 * (e.g. from command line) invalidate the cache. This
2694 * condition also catches running setup_standard_excludes()
2695 * before setting dir->untracked!
2697 if (dir->unmanaged_exclude_files)
2698 return NULL;
2701 * Optimize for the main use case only: whole-tree git
2702 * status. More work involved in treat_leading_path() if we
2703 * use cache on just a subset of the worktree. pathspec
2704 * support could make the matter even worse.
2706 if (base_len || (pathspec && pathspec->nr))
2707 return NULL;
2709 /* Different set of flags may produce different results */
2710 if (dir->flags != dir->untracked->dir_flags ||
2712 * See treat_directory(), case index_nonexistent. Without
2713 * this flag, we may need to also cache .git file content
2714 * for the resolve_gitlink_ref() call, which we don't.
2716 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2717 /* We don't support collecting ignore files */
2718 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2719 DIR_COLLECT_IGNORED)))
2720 return NULL;
2723 * If we use .gitignore in the cache and now you change it to
2724 * .gitexclude, everything will go wrong.
2726 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2727 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2728 return NULL;
2731 * EXC_CMDL is not considered in the cache. If people set it,
2732 * skip the cache.
2734 if (dir->exclude_list_group[EXC_CMDL].nr)
2735 return NULL;
2737 if (!ident_in_untracked(dir->untracked)) {
2738 warning(_("untracked cache is disabled on this system or location"));
2739 return NULL;
2742 if (!dir->untracked->root)
2743 FLEX_ALLOC_STR(dir->untracked->root, name, "");
2745 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2746 root = dir->untracked->root;
2747 if (!oideq(&dir->ss_info_exclude.oid,
2748 &dir->untracked->ss_info_exclude.oid)) {
2749 invalidate_gitignore(dir->untracked, root);
2750 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2752 if (!oideq(&dir->ss_excludes_file.oid,
2753 &dir->untracked->ss_excludes_file.oid)) {
2754 invalidate_gitignore(dir->untracked, root);
2755 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2758 /* Make sure this directory is not dropped out at saving phase */
2759 root->recurse = 1;
2760 return root;
2763 int read_directory(struct dir_struct *dir, struct index_state *istate,
2764 const char *path, int len, const struct pathspec *pathspec)
2766 struct untracked_cache_dir *untracked;
2768 trace_performance_enter();
2770 if (has_symlink_leading_path(path, len)) {
2771 trace_performance_leave("read directory %.*s", len, path);
2772 return dir->nr;
2775 untracked = validate_untracked_cache(dir, len, pathspec);
2776 if (!untracked)
2778 * make sure untracked cache code path is disabled,
2779 * e.g. prep_exclude()
2781 dir->untracked = NULL;
2782 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2783 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2784 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2785 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2787 trace_performance_leave("read directory %.*s", len, path);
2788 if (dir->untracked) {
2789 static int force_untracked_cache = -1;
2790 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2792 if (force_untracked_cache < 0)
2793 force_untracked_cache =
2794 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2795 trace_printf_key(&trace_untracked_stats,
2796 "node creation: %u\n"
2797 "gitignore invalidation: %u\n"
2798 "directory invalidation: %u\n"
2799 "opendir: %u\n",
2800 dir->untracked->dir_created,
2801 dir->untracked->gitignore_invalidated,
2802 dir->untracked->dir_invalidated,
2803 dir->untracked->dir_opened);
2804 if (force_untracked_cache &&
2805 dir->untracked == istate->untracked &&
2806 (dir->untracked->dir_opened ||
2807 dir->untracked->gitignore_invalidated ||
2808 dir->untracked->dir_invalidated))
2809 istate->cache_changed |= UNTRACKED_CHANGED;
2810 if (dir->untracked != istate->untracked) {
2811 FREE_AND_NULL(dir->untracked);
2814 return dir->nr;
2817 int file_exists(const char *f)
2819 struct stat sb;
2820 return lstat(f, &sb) == 0;
2823 int repo_file_exists(struct repository *repo, const char *path)
2825 if (repo != the_repository)
2826 BUG("do not know how to check file existence in arbitrary repo");
2828 return file_exists(path);
2831 static int cmp_icase(char a, char b)
2833 if (a == b)
2834 return 0;
2835 if (ignore_case)
2836 return toupper(a) - toupper(b);
2837 return a - b;
2841 * Given two normalized paths (a trailing slash is ok), if subdir is
2842 * outside dir, return -1. Otherwise return the offset in subdir that
2843 * can be used as relative path to dir.
2845 int dir_inside_of(const char *subdir, const char *dir)
2847 int offset = 0;
2849 assert(dir && subdir && *dir && *subdir);
2851 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2852 dir++;
2853 subdir++;
2854 offset++;
2857 /* hel[p]/me vs hel[l]/yeah */
2858 if (*dir && *subdir)
2859 return -1;
2861 if (!*subdir)
2862 return !*dir ? offset : -1; /* same dir */
2864 /* foo/[b]ar vs foo/[] */
2865 if (is_dir_sep(dir[-1]))
2866 return is_dir_sep(subdir[-1]) ? offset : -1;
2868 /* foo[/]bar vs foo[] */
2869 return is_dir_sep(*subdir) ? offset + 1 : -1;
2872 int is_inside_dir(const char *dir)
2874 char *cwd;
2875 int rc;
2877 if (!dir)
2878 return 0;
2880 cwd = xgetcwd();
2881 rc = (dir_inside_of(cwd, dir) >= 0);
2882 free(cwd);
2883 return rc;
2886 int is_empty_dir(const char *path)
2888 DIR *dir = opendir(path);
2889 struct dirent *e;
2890 int ret = 1;
2892 if (!dir)
2893 return 0;
2895 while ((e = readdir(dir)) != NULL)
2896 if (!is_dot_or_dotdot(e->d_name)) {
2897 ret = 0;
2898 break;
2901 closedir(dir);
2902 return ret;
2905 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2907 DIR *dir;
2908 struct dirent *e;
2909 int ret = 0, original_len = path->len, len, kept_down = 0;
2910 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2911 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2912 struct object_id submodule_head;
2914 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2915 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2916 /* Do not descend and nuke a nested git work tree. */
2917 if (kept_up)
2918 *kept_up = 1;
2919 return 0;
2922 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2923 dir = opendir(path->buf);
2924 if (!dir) {
2925 if (errno == ENOENT)
2926 return keep_toplevel ? -1 : 0;
2927 else if (errno == EACCES && !keep_toplevel)
2929 * An empty dir could be removable even if it
2930 * is unreadable:
2932 return rmdir(path->buf);
2933 else
2934 return -1;
2936 strbuf_complete(path, '/');
2938 len = path->len;
2939 while ((e = readdir(dir)) != NULL) {
2940 struct stat st;
2941 if (is_dot_or_dotdot(e->d_name))
2942 continue;
2944 strbuf_setlen(path, len);
2945 strbuf_addstr(path, e->d_name);
2946 if (lstat(path->buf, &st)) {
2947 if (errno == ENOENT)
2949 * file disappeared, which is what we
2950 * wanted anyway
2952 continue;
2953 /* fall through */
2954 } else if (S_ISDIR(st.st_mode)) {
2955 if (!remove_dir_recurse(path, flag, &kept_down))
2956 continue; /* happy */
2957 } else if (!only_empty &&
2958 (!unlink(path->buf) || errno == ENOENT)) {
2959 continue; /* happy, too */
2962 /* path too long, stat fails, or non-directory still exists */
2963 ret = -1;
2964 break;
2966 closedir(dir);
2968 strbuf_setlen(path, original_len);
2969 if (!ret && !keep_toplevel && !kept_down)
2970 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2971 else if (kept_up)
2973 * report the uplevel that it is not an error that we
2974 * did not rmdir() our directory.
2976 *kept_up = !ret;
2977 return ret;
2980 int remove_dir_recursively(struct strbuf *path, int flag)
2982 return remove_dir_recurse(path, flag, NULL);
2985 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2987 void setup_standard_excludes(struct dir_struct *dir)
2989 dir->exclude_per_dir = ".gitignore";
2991 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2992 if (!excludes_file)
2993 excludes_file = xdg_config_home("ignore");
2994 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2995 add_patterns_from_file_1(dir, excludes_file,
2996 dir->untracked ? &dir->ss_excludes_file : NULL);
2998 /* per repository user preference */
2999 if (startup_info->have_repository) {
3000 const char *path = git_path_info_exclude();
3001 if (!access_or_warn(path, R_OK, 0))
3002 add_patterns_from_file_1(dir, path,
3003 dir->untracked ? &dir->ss_info_exclude : NULL);
3007 char *get_sparse_checkout_filename(void)
3009 return git_pathdup("info/sparse-checkout");
3012 int get_sparse_checkout_patterns(struct pattern_list *pl)
3014 int res;
3015 char *sparse_filename = get_sparse_checkout_filename();
3017 pl->use_cone_patterns = core_sparse_checkout_cone;
3018 res = add_patterns_from_file_to_list(sparse_filename, "", 0, pl, NULL, 0);
3020 free(sparse_filename);
3021 return res;
3024 int remove_path(const char *name)
3026 char *slash;
3028 if (unlink(name) && !is_missing_file_error(errno))
3029 return -1;
3031 slash = strrchr(name, '/');
3032 if (slash) {
3033 char *dirs = xstrdup(name);
3034 slash = dirs + (slash - name);
3035 do {
3036 *slash = '\0';
3037 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
3038 free(dirs);
3040 return 0;
3044 * Frees memory within dir which was allocated, and resets fields for further
3045 * use. Does not free dir itself.
3047 void dir_clear(struct dir_struct *dir)
3049 int i, j;
3050 struct exclude_list_group *group;
3051 struct pattern_list *pl;
3052 struct exclude_stack *stk;
3054 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
3055 group = &dir->exclude_list_group[i];
3056 for (j = 0; j < group->nr; j++) {
3057 pl = &group->pl[j];
3058 if (i == EXC_DIRS)
3059 free((char *)pl->src);
3060 clear_pattern_list(pl);
3062 free(group->pl);
3065 for (i = 0; i < dir->ignored_nr; i++)
3066 free(dir->ignored[i]);
3067 for (i = 0; i < dir->nr; i++)
3068 free(dir->entries[i]);
3069 free(dir->ignored);
3070 free(dir->entries);
3072 stk = dir->exclude_stack;
3073 while (stk) {
3074 struct exclude_stack *prev = stk->prev;
3075 free(stk);
3076 stk = prev;
3078 strbuf_release(&dir->basebuf);
3080 dir_init(dir);
3083 struct ondisk_untracked_cache {
3084 struct stat_data info_exclude_stat;
3085 struct stat_data excludes_file_stat;
3086 uint32_t dir_flags;
3089 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3091 struct write_data {
3092 int index; /* number of written untracked_cache_dir */
3093 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
3094 struct ewah_bitmap *valid; /* from untracked_cache_dir */
3095 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
3096 struct strbuf out;
3097 struct strbuf sb_stat;
3098 struct strbuf sb_sha1;
3101 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
3103 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
3104 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
3105 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
3106 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
3107 to->sd_dev = htonl(from->sd_dev);
3108 to->sd_ino = htonl(from->sd_ino);
3109 to->sd_uid = htonl(from->sd_uid);
3110 to->sd_gid = htonl(from->sd_gid);
3111 to->sd_size = htonl(from->sd_size);
3114 static void write_one_dir(struct untracked_cache_dir *untracked,
3115 struct write_data *wd)
3117 struct stat_data stat_data;
3118 struct strbuf *out = &wd->out;
3119 unsigned char intbuf[16];
3120 unsigned int intlen, value;
3121 int i = wd->index++;
3124 * untracked_nr should be reset whenever valid is clear, but
3125 * for safety..
3127 if (!untracked->valid) {
3128 untracked->untracked_nr = 0;
3129 untracked->check_only = 0;
3132 if (untracked->check_only)
3133 ewah_set(wd->check_only, i);
3134 if (untracked->valid) {
3135 ewah_set(wd->valid, i);
3136 stat_data_to_disk(&stat_data, &untracked->stat_data);
3137 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
3139 if (!is_null_oid(&untracked->exclude_oid)) {
3140 ewah_set(wd->sha1_valid, i);
3141 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
3142 the_hash_algo->rawsz);
3145 intlen = encode_varint(untracked->untracked_nr, intbuf);
3146 strbuf_add(out, intbuf, intlen);
3148 /* skip non-recurse directories */
3149 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
3150 if (untracked->dirs[i]->recurse)
3151 value++;
3152 intlen = encode_varint(value, intbuf);
3153 strbuf_add(out, intbuf, intlen);
3155 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
3157 for (i = 0; i < untracked->untracked_nr; i++)
3158 strbuf_add(out, untracked->untracked[i],
3159 strlen(untracked->untracked[i]) + 1);
3161 for (i = 0; i < untracked->dirs_nr; i++)
3162 if (untracked->dirs[i]->recurse)
3163 write_one_dir(untracked->dirs[i], wd);
3166 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
3168 struct ondisk_untracked_cache *ouc;
3169 struct write_data wd;
3170 unsigned char varbuf[16];
3171 int varint_len;
3172 const unsigned hashsz = the_hash_algo->rawsz;
3174 CALLOC_ARRAY(ouc, 1);
3175 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
3176 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
3177 ouc->dir_flags = htonl(untracked->dir_flags);
3179 varint_len = encode_varint(untracked->ident.len, varbuf);
3180 strbuf_add(out, varbuf, varint_len);
3181 strbuf_addbuf(out, &untracked->ident);
3183 strbuf_add(out, ouc, sizeof(*ouc));
3184 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
3185 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
3186 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
3187 FREE_AND_NULL(ouc);
3189 if (!untracked->root) {
3190 varint_len = encode_varint(0, varbuf);
3191 strbuf_add(out, varbuf, varint_len);
3192 return;
3195 wd.index = 0;
3196 wd.check_only = ewah_new();
3197 wd.valid = ewah_new();
3198 wd.sha1_valid = ewah_new();
3199 strbuf_init(&wd.out, 1024);
3200 strbuf_init(&wd.sb_stat, 1024);
3201 strbuf_init(&wd.sb_sha1, 1024);
3202 write_one_dir(untracked->root, &wd);
3204 varint_len = encode_varint(wd.index, varbuf);
3205 strbuf_add(out, varbuf, varint_len);
3206 strbuf_addbuf(out, &wd.out);
3207 ewah_serialize_strbuf(wd.valid, out);
3208 ewah_serialize_strbuf(wd.check_only, out);
3209 ewah_serialize_strbuf(wd.sha1_valid, out);
3210 strbuf_addbuf(out, &wd.sb_stat);
3211 strbuf_addbuf(out, &wd.sb_sha1);
3212 strbuf_addch(out, '\0'); /* safe guard for string lists */
3214 ewah_free(wd.valid);
3215 ewah_free(wd.check_only);
3216 ewah_free(wd.sha1_valid);
3217 strbuf_release(&wd.out);
3218 strbuf_release(&wd.sb_stat);
3219 strbuf_release(&wd.sb_sha1);
3222 static void free_untracked(struct untracked_cache_dir *ucd)
3224 int i;
3225 if (!ucd)
3226 return;
3227 for (i = 0; i < ucd->dirs_nr; i++)
3228 free_untracked(ucd->dirs[i]);
3229 for (i = 0; i < ucd->untracked_nr; i++)
3230 free(ucd->untracked[i]);
3231 free(ucd->untracked);
3232 free(ucd->dirs);
3233 free(ucd);
3236 void free_untracked_cache(struct untracked_cache *uc)
3238 if (uc)
3239 free_untracked(uc->root);
3240 free(uc);
3243 struct read_data {
3244 int index;
3245 struct untracked_cache_dir **ucd;
3246 struct ewah_bitmap *check_only;
3247 struct ewah_bitmap *valid;
3248 struct ewah_bitmap *sha1_valid;
3249 const unsigned char *data;
3250 const unsigned char *end;
3253 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
3255 memcpy(to, data, sizeof(*to));
3256 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
3257 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
3258 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
3259 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
3260 to->sd_dev = ntohl(to->sd_dev);
3261 to->sd_ino = ntohl(to->sd_ino);
3262 to->sd_uid = ntohl(to->sd_uid);
3263 to->sd_gid = ntohl(to->sd_gid);
3264 to->sd_size = ntohl(to->sd_size);
3267 static int read_one_dir(struct untracked_cache_dir **untracked_,
3268 struct read_data *rd)
3270 struct untracked_cache_dir ud, *untracked;
3271 const unsigned char *data = rd->data, *end = rd->end;
3272 const unsigned char *eos;
3273 unsigned int value;
3274 int i;
3276 memset(&ud, 0, sizeof(ud));
3278 value = decode_varint(&data);
3279 if (data > end)
3280 return -1;
3281 ud.recurse = 1;
3282 ud.untracked_alloc = value;
3283 ud.untracked_nr = value;
3284 if (ud.untracked_nr)
3285 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
3287 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
3288 if (data > end)
3289 return -1;
3290 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
3292 eos = memchr(data, '\0', end - data);
3293 if (!eos || eos == end)
3294 return -1;
3296 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
3297 memcpy(untracked, &ud, sizeof(ud));
3298 memcpy(untracked->name, data, eos - data + 1);
3299 data = eos + 1;
3301 for (i = 0; i < untracked->untracked_nr; i++) {
3302 eos = memchr(data, '\0', end - data);
3303 if (!eos || eos == end)
3304 return -1;
3305 untracked->untracked[i] = xmemdupz(data, eos - data);
3306 data = eos + 1;
3309 rd->ucd[rd->index++] = untracked;
3310 rd->data = data;
3312 for (i = 0; i < untracked->dirs_nr; i++) {
3313 if (read_one_dir(untracked->dirs + i, rd) < 0)
3314 return -1;
3316 return 0;
3319 static void set_check_only(size_t pos, void *cb)
3321 struct read_data *rd = cb;
3322 struct untracked_cache_dir *ud = rd->ucd[pos];
3323 ud->check_only = 1;
3326 static void read_stat(size_t pos, void *cb)
3328 struct read_data *rd = cb;
3329 struct untracked_cache_dir *ud = rd->ucd[pos];
3330 if (rd->data + sizeof(struct stat_data) > rd->end) {
3331 rd->data = rd->end + 1;
3332 return;
3334 stat_data_from_disk(&ud->stat_data, rd->data);
3335 rd->data += sizeof(struct stat_data);
3336 ud->valid = 1;
3339 static void read_oid(size_t pos, void *cb)
3341 struct read_data *rd = cb;
3342 struct untracked_cache_dir *ud = rd->ucd[pos];
3343 if (rd->data + the_hash_algo->rawsz > rd->end) {
3344 rd->data = rd->end + 1;
3345 return;
3347 hashcpy(ud->exclude_oid.hash, rd->data);
3348 rd->data += the_hash_algo->rawsz;
3351 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
3352 const unsigned char *sha1)
3354 stat_data_from_disk(&oid_stat->stat, data);
3355 hashcpy(oid_stat->oid.hash, sha1);
3356 oid_stat->valid = 1;
3359 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
3361 struct untracked_cache *uc;
3362 struct read_data rd;
3363 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
3364 const char *ident;
3365 int ident_len;
3366 ssize_t len;
3367 const char *exclude_per_dir;
3368 const unsigned hashsz = the_hash_algo->rawsz;
3369 const unsigned offset = sizeof(struct ondisk_untracked_cache);
3370 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
3372 if (sz <= 1 || end[-1] != '\0')
3373 return NULL;
3374 end--;
3376 ident_len = decode_varint(&next);
3377 if (next + ident_len > end)
3378 return NULL;
3379 ident = (const char *)next;
3380 next += ident_len;
3382 if (next + exclude_per_dir_offset + 1 > end)
3383 return NULL;
3385 CALLOC_ARRAY(uc, 1);
3386 strbuf_init(&uc->ident, ident_len);
3387 strbuf_add(&uc->ident, ident, ident_len);
3388 load_oid_stat(&uc->ss_info_exclude,
3389 next + ouc_offset(info_exclude_stat),
3390 next + offset);
3391 load_oid_stat(&uc->ss_excludes_file,
3392 next + ouc_offset(excludes_file_stat),
3393 next + offset + hashsz);
3394 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
3395 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
3396 uc->exclude_per_dir = xstrdup(exclude_per_dir);
3397 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3398 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
3399 if (next >= end)
3400 goto done2;
3402 len = decode_varint(&next);
3403 if (next > end || len == 0)
3404 goto done2;
3406 rd.valid = ewah_new();
3407 rd.check_only = ewah_new();
3408 rd.sha1_valid = ewah_new();
3409 rd.data = next;
3410 rd.end = end;
3411 rd.index = 0;
3412 ALLOC_ARRAY(rd.ucd, len);
3414 if (read_one_dir(&uc->root, &rd) || rd.index != len)
3415 goto done;
3417 next = rd.data;
3418 len = ewah_read_mmap(rd.valid, next, end - next);
3419 if (len < 0)
3420 goto done;
3422 next += len;
3423 len = ewah_read_mmap(rd.check_only, next, end - next);
3424 if (len < 0)
3425 goto done;
3427 next += len;
3428 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
3429 if (len < 0)
3430 goto done;
3432 ewah_each_bit(rd.check_only, set_check_only, &rd);
3433 rd.data = next + len;
3434 ewah_each_bit(rd.valid, read_stat, &rd);
3435 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
3436 next = rd.data;
3438 done:
3439 free(rd.ucd);
3440 ewah_free(rd.valid);
3441 ewah_free(rd.check_only);
3442 ewah_free(rd.sha1_valid);
3443 done2:
3444 if (next != end) {
3445 free_untracked_cache(uc);
3446 uc = NULL;
3448 return uc;
3451 static void invalidate_one_directory(struct untracked_cache *uc,
3452 struct untracked_cache_dir *ucd)
3454 uc->dir_invalidated++;
3455 ucd->valid = 0;
3456 ucd->untracked_nr = 0;
3460 * Normally when an entry is added or removed from a directory,
3461 * invalidating that directory is enough. No need to touch its
3462 * ancestors. When a directory is shown as "foo/bar/" in git-status
3463 * however, deleting or adding an entry may have cascading effect.
3465 * Say the "foo/bar/file" has become untracked, we need to tell the
3466 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3467 * directory any more (because "bar" is managed by foo as an untracked
3468 * "file").
3470 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3471 * was the last untracked entry in the entire "foo", we should show
3472 * "foo/" instead. Which means we have to invalidate past "bar" up to
3473 * "foo".
3475 * This function traverses all directories from root to leaf. If there
3476 * is a chance of one of the above cases happening, we invalidate back
3477 * to root. Otherwise we just invalidate the leaf. There may be a more
3478 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3479 * detect these cases and avoid unnecessary invalidation, for example,
3480 * checking for the untracked entry named "bar/" in "foo", but for now
3481 * stick to something safe and simple.
3483 static int invalidate_one_component(struct untracked_cache *uc,
3484 struct untracked_cache_dir *dir,
3485 const char *path, int len)
3487 const char *rest = strchr(path, '/');
3489 if (rest) {
3490 int component_len = rest - path;
3491 struct untracked_cache_dir *d =
3492 lookup_untracked(uc, dir, path, component_len);
3493 int ret =
3494 invalidate_one_component(uc, d, rest + 1,
3495 len - (component_len + 1));
3496 if (ret)
3497 invalidate_one_directory(uc, dir);
3498 return ret;
3501 invalidate_one_directory(uc, dir);
3502 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
3505 void untracked_cache_invalidate_path(struct index_state *istate,
3506 const char *path, int safe_path)
3508 if (!istate->untracked || !istate->untracked->root)
3509 return;
3510 if (!safe_path && !verify_path(path, 0))
3511 return;
3512 invalidate_one_component(istate->untracked, istate->untracked->root,
3513 path, strlen(path));
3516 void untracked_cache_remove_from_index(struct index_state *istate,
3517 const char *path)
3519 untracked_cache_invalidate_path(istate, path, 1);
3522 void untracked_cache_add_to_index(struct index_state *istate,
3523 const char *path)
3525 untracked_cache_invalidate_path(istate, path, 1);
3528 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3529 const char *sub_gitdir)
3531 int i;
3532 struct repository subrepo;
3533 struct strbuf sub_wt = STRBUF_INIT;
3534 struct strbuf sub_gd = STRBUF_INIT;
3536 const struct submodule *sub;
3538 /* If the submodule has no working tree, we can ignore it. */
3539 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3540 return;
3542 if (repo_read_index(&subrepo) < 0)
3543 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3545 /* TODO: audit for interaction with sparse-index. */
3546 ensure_full_index(subrepo.index);
3547 for (i = 0; i < subrepo.index->cache_nr; i++) {
3548 const struct cache_entry *ce = subrepo.index->cache[i];
3550 if (!S_ISGITLINK(ce->ce_mode))
3551 continue;
3553 while (i + 1 < subrepo.index->cache_nr &&
3554 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3556 * Skip entries with the same name in different stages
3557 * to make sure an entry is returned only once.
3559 i++;
3561 sub = submodule_from_path(&subrepo, &null_oid, ce->name);
3562 if (!sub || !is_submodule_active(&subrepo, ce->name))
3563 /* .gitmodules broken or inactive sub */
3564 continue;
3566 strbuf_reset(&sub_wt);
3567 strbuf_reset(&sub_gd);
3568 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3569 strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name);
3571 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3573 strbuf_release(&sub_wt);
3574 strbuf_release(&sub_gd);
3575 repo_clear(&subrepo);
3578 void connect_work_tree_and_git_dir(const char *work_tree_,
3579 const char *git_dir_,
3580 int recurse_into_nested)
3582 struct strbuf gitfile_sb = STRBUF_INIT;
3583 struct strbuf cfg_sb = STRBUF_INIT;
3584 struct strbuf rel_path = STRBUF_INIT;
3585 char *git_dir, *work_tree;
3587 /* Prepare .git file */
3588 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3589 if (safe_create_leading_directories_const(gitfile_sb.buf))
3590 die(_("could not create directories for %s"), gitfile_sb.buf);
3592 /* Prepare config file */
3593 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3594 if (safe_create_leading_directories_const(cfg_sb.buf))
3595 die(_("could not create directories for %s"), cfg_sb.buf);
3597 git_dir = real_pathdup(git_dir_, 1);
3598 work_tree = real_pathdup(work_tree_, 1);
3600 /* Write .git file */
3601 write_file(gitfile_sb.buf, "gitdir: %s",
3602 relative_path(git_dir, work_tree, &rel_path));
3603 /* Update core.worktree setting */
3604 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3605 relative_path(work_tree, git_dir, &rel_path));
3607 strbuf_release(&gitfile_sb);
3608 strbuf_release(&cfg_sb);
3609 strbuf_release(&rel_path);
3611 if (recurse_into_nested)
3612 connect_wt_gitdir_in_nested(work_tree, git_dir);
3614 free(work_tree);
3615 free(git_dir);
3619 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3621 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3623 if (rename(old_git_dir, new_git_dir) < 0)
3624 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3625 old_git_dir, new_git_dir);
3627 connect_work_tree_and_git_dir(path, new_git_dir, 0);