pack-objects: fix --no-keep-true-parents
[git/debian.git] / dir.c
bloba7469df3ac7f95109c48ced53259aac021e14b8b
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 "git-compat-util.h"
9 #include "abspath.h"
10 #include "alloc.h"
11 #include "config.h"
12 #include "convert.h"
13 #include "dir.h"
14 #include "environment.h"
15 #include "gettext.h"
16 #include "object-file.h"
17 #include "object-store.h"
18 #include "attr.h"
19 #include "refs.h"
20 #include "wildmatch.h"
21 #include "pathspec.h"
22 #include "utf8.h"
23 #include "varint.h"
24 #include "ewah/ewok.h"
25 #include "fsmonitor.h"
26 #include "setup.h"
27 #include "submodule-config.h"
28 #include "symlinks.h"
29 #include "trace2.h"
30 #include "tree.h"
31 #include "wrapper.h"
34 * Tells read_directory_recursive how a file or directory should be treated.
35 * Values are ordered by significance, e.g. if a directory contains both
36 * excluded and untracked files, it is listed as untracked because
37 * path_untracked > path_excluded.
39 enum path_treatment {
40 path_none = 0,
41 path_recurse,
42 path_excluded,
43 path_untracked
47 * Support data structure for our opendir/readdir/closedir wrappers
49 struct cached_dir {
50 DIR *fdir;
51 struct untracked_cache_dir *untracked;
52 int nr_files;
53 int nr_dirs;
55 const char *d_name;
56 int d_type;
57 const char *file;
58 struct untracked_cache_dir *ucd;
61 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
62 struct index_state *istate, const char *path, int len,
63 struct untracked_cache_dir *untracked,
64 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
65 static int resolve_dtype(int dtype, struct index_state *istate,
66 const char *path, int len);
67 struct dirent *readdir_skip_dot_and_dotdot(DIR *dirp)
69 struct dirent *e;
71 while ((e = readdir(dirp)) != NULL) {
72 if (!is_dot_or_dotdot(e->d_name))
73 break;
75 return e;
78 int count_slashes(const char *s)
80 int cnt = 0;
81 while (*s)
82 if (*s++ == '/')
83 cnt++;
84 return cnt;
87 int fspathcmp(const char *a, const char *b)
89 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
92 int fspatheq(const char *a, const char *b)
94 return !fspathcmp(a, b);
97 int fspathncmp(const char *a, const char *b, size_t count)
99 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
102 unsigned int fspathhash(const char *str)
104 return ignore_case ? strihash(str) : strhash(str);
107 int git_fnmatch(const struct pathspec_item *item,
108 const char *pattern, const char *string,
109 int prefix)
111 if (prefix > 0) {
112 if (ps_strncmp(item, pattern, string, prefix))
113 return WM_NOMATCH;
114 pattern += prefix;
115 string += prefix;
117 if (item->flags & PATHSPEC_ONESTAR) {
118 int pattern_len = strlen(++pattern);
119 int string_len = strlen(string);
120 return string_len < pattern_len ||
121 ps_strcmp(item, pattern,
122 string + string_len - pattern_len);
124 if (item->magic & PATHSPEC_GLOB)
125 return wildmatch(pattern, string,
126 WM_PATHNAME |
127 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
128 else
129 /* wildmatch has not learned no FNM_PATHNAME mode yet */
130 return wildmatch(pattern, string,
131 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
134 static int fnmatch_icase_mem(const char *pattern, int patternlen,
135 const char *string, int stringlen,
136 int flags)
138 int match_status;
139 struct strbuf pat_buf = STRBUF_INIT;
140 struct strbuf str_buf = STRBUF_INIT;
141 const char *use_pat = pattern;
142 const char *use_str = string;
144 if (pattern[patternlen]) {
145 strbuf_add(&pat_buf, pattern, patternlen);
146 use_pat = pat_buf.buf;
148 if (string[stringlen]) {
149 strbuf_add(&str_buf, string, stringlen);
150 use_str = str_buf.buf;
153 if (ignore_case)
154 flags |= WM_CASEFOLD;
155 match_status = wildmatch(use_pat, use_str, flags);
157 strbuf_release(&pat_buf);
158 strbuf_release(&str_buf);
160 return match_status;
163 static size_t common_prefix_len(const struct pathspec *pathspec)
165 int n;
166 size_t max = 0;
169 * ":(icase)path" is treated as a pathspec full of
170 * wildcard. In other words, only prefix is considered common
171 * prefix. If the pathspec is abc/foo abc/bar, running in
172 * subdir xyz, the common prefix is still xyz, not xyz/abc as
173 * in non-:(icase).
175 GUARD_PATHSPEC(pathspec,
176 PATHSPEC_FROMTOP |
177 PATHSPEC_MAXDEPTH |
178 PATHSPEC_LITERAL |
179 PATHSPEC_GLOB |
180 PATHSPEC_ICASE |
181 PATHSPEC_EXCLUDE |
182 PATHSPEC_ATTR);
184 for (n = 0; n < pathspec->nr; n++) {
185 size_t i = 0, len = 0, item_len;
186 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
187 continue;
188 if (pathspec->items[n].magic & PATHSPEC_ICASE)
189 item_len = pathspec->items[n].prefix;
190 else
191 item_len = pathspec->items[n].nowildcard_len;
192 while (i < item_len && (n == 0 || i < max)) {
193 char c = pathspec->items[n].match[i];
194 if (c != pathspec->items[0].match[i])
195 break;
196 if (c == '/')
197 len = i + 1;
198 i++;
200 if (n == 0 || len < max) {
201 max = len;
202 if (!max)
203 break;
206 return max;
210 * Returns a copy of the longest leading path common among all
211 * pathspecs.
213 char *common_prefix(const struct pathspec *pathspec)
215 unsigned long len = common_prefix_len(pathspec);
217 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
220 int fill_directory(struct dir_struct *dir,
221 struct index_state *istate,
222 const struct pathspec *pathspec)
224 const char *prefix;
225 size_t prefix_len;
227 unsigned exclusive_flags = DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO;
228 if ((dir->flags & exclusive_flags) == exclusive_flags)
229 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
232 * Calculate common prefix for the pathspec, and
233 * use that to optimize the directory walk
235 prefix_len = common_prefix_len(pathspec);
236 prefix = prefix_len ? pathspec->items[0].match : "";
238 /* Read the directory and prune it */
239 read_directory(dir, istate, prefix, prefix_len, pathspec);
241 return prefix_len;
244 int within_depth(const char *name, int namelen,
245 int depth, int max_depth)
247 const char *cp = name, *cpe = name + namelen;
249 while (cp < cpe) {
250 if (*cp++ != '/')
251 continue;
252 depth++;
253 if (depth > max_depth)
254 return 0;
256 return 1;
260 * Read the contents of the blob with the given OID into a buffer.
261 * Append a trailing LF to the end if the last line doesn't have one.
263 * Returns:
264 * -1 when the OID is invalid or unknown or does not refer to a blob.
265 * 0 when the blob is empty.
266 * 1 along with { data, size } of the (possibly augmented) buffer
267 * when successful.
269 * Optionally updates the given oid_stat with the given OID (when valid).
271 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
272 size_t *size_out, char **data_out)
274 enum object_type type;
275 unsigned long sz;
276 char *data;
278 *size_out = 0;
279 *data_out = NULL;
281 data = repo_read_object_file(the_repository, oid, &type, &sz);
282 if (!data || type != OBJ_BLOB) {
283 free(data);
284 return -1;
287 if (oid_stat) {
288 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
289 oidcpy(&oid_stat->oid, oid);
292 if (sz == 0) {
293 free(data);
294 return 0;
297 if (data[sz - 1] != '\n') {
298 data = xrealloc(data, st_add(sz, 1));
299 data[sz++] = '\n';
302 *size_out = xsize_t(sz);
303 *data_out = data;
305 return 1;
308 #define DO_MATCH_EXCLUDE (1<<0)
309 #define DO_MATCH_DIRECTORY (1<<1)
310 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
313 * Does the given pathspec match the given name? A match is found if
315 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
316 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
317 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
318 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
320 * Return value tells which case it was (1-4), or 0 when there is no match.
322 * It may be instructive to look at a small table of concrete examples
323 * to understand the differences between 1, 2, and 4:
325 * Pathspecs
326 * | a/b | a/b/ | a/b/c
327 * ------+-----------+-----------+------------
328 * a/b | EXACT | EXACT[1] | LEADING[2]
329 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
330 * a/b/c | RECURSIVE | RECURSIVE | EXACT
332 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
333 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
335 static int match_pathspec_item(struct index_state *istate,
336 const struct pathspec_item *item, int prefix,
337 const char *name, int namelen, unsigned flags)
339 /* name/namelen has prefix cut off by caller */
340 const char *match = item->match + prefix;
341 int matchlen = item->len - prefix;
344 * The normal call pattern is:
345 * 1. prefix = common_prefix_len(ps);
346 * 2. prune something, or fill_directory
347 * 3. match_pathspec()
349 * 'prefix' at #1 may be shorter than the command's prefix and
350 * it's ok for #2 to match extra files. Those extras will be
351 * trimmed at #3.
353 * Suppose the pathspec is 'foo' and '../bar' running from
354 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
355 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
356 * user does not want XYZ/foo, only the "foo" part should be
357 * case-insensitive. We need to filter out XYZ/foo here. In
358 * other words, we do not trust the caller on comparing the
359 * prefix part when :(icase) is involved. We do exact
360 * comparison ourselves.
362 * Normally the caller (common_prefix_len() in fact) does
363 * _exact_ matching on name[-prefix+1..-1] and we do not need
364 * to check that part. Be defensive and check it anyway, in
365 * case common_prefix_len is changed, or a new caller is
366 * introduced that does not use common_prefix_len.
368 * If the penalty turns out too high when prefix is really
369 * long, maybe change it to
370 * strncmp(match, name, item->prefix - prefix)
372 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
373 strncmp(item->match, name - prefix, item->prefix))
374 return 0;
376 if (item->attr_match_nr &&
377 !match_pathspec_attrs(istate, name, namelen, item))
378 return 0;
380 /* If the match was just the prefix, we matched */
381 if (!*match)
382 return MATCHED_RECURSIVELY;
384 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
385 if (matchlen == namelen)
386 return MATCHED_EXACTLY;
388 if (match[matchlen-1] == '/' || name[matchlen] == '/')
389 return MATCHED_RECURSIVELY;
390 } else if ((flags & DO_MATCH_DIRECTORY) &&
391 match[matchlen - 1] == '/' &&
392 namelen == matchlen - 1 &&
393 !ps_strncmp(item, match, name, namelen))
394 return MATCHED_EXACTLY;
396 if (item->nowildcard_len < item->len &&
397 !git_fnmatch(item, match, name,
398 item->nowildcard_len - prefix))
399 return MATCHED_FNMATCH;
401 /* Perform checks to see if "name" is a leading string of the pathspec */
402 if ( (flags & DO_MATCH_LEADING_PATHSPEC) &&
403 !(flags & DO_MATCH_EXCLUDE)) {
404 /* name is a literal prefix of the pathspec */
405 int offset = name[namelen-1] == '/' ? 1 : 0;
406 if ((namelen < matchlen) &&
407 (match[namelen-offset] == '/') &&
408 !ps_strncmp(item, match, name, namelen))
409 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
411 /* name doesn't match up to the first wild character */
412 if (item->nowildcard_len < item->len &&
413 ps_strncmp(item, match, name,
414 item->nowildcard_len - prefix))
415 return 0;
418 * name has no wildcard, and it didn't match as a leading
419 * pathspec so return.
421 if (item->nowildcard_len == item->len)
422 return 0;
425 * Here is where we would perform a wildmatch to check if
426 * "name" can be matched as a directory (or a prefix) against
427 * the pathspec. Since wildmatch doesn't have this capability
428 * at the present we have to punt and say that it is a match,
429 * potentially returning a false positive
430 * The submodules themselves will be able to perform more
431 * accurate matching to determine if the pathspec matches.
433 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
436 return 0;
440 * do_match_pathspec() is meant to ONLY be called by
441 * match_pathspec_with_flags(); calling it directly risks pathspecs
442 * like ':!unwanted_path' being ignored.
444 * Given a name and a list of pathspecs, returns the nature of the
445 * closest (i.e. most specific) match of the name to any of the
446 * pathspecs.
448 * The caller typically calls this multiple times with the same
449 * pathspec and seen[] array but with different name/namelen
450 * (e.g. entries from the index) and is interested in seeing if and
451 * how each pathspec matches all the names it calls this function
452 * with. A mark is left in the seen[] array for each pathspec element
453 * indicating the closest type of match that element achieved, so if
454 * seen[n] remains zero after multiple invocations, that means the nth
455 * pathspec did not match any names, which could indicate that the
456 * user mistyped the nth pathspec.
458 static int do_match_pathspec(struct index_state *istate,
459 const struct pathspec *ps,
460 const char *name, int namelen,
461 int prefix, char *seen,
462 unsigned flags)
464 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
466 GUARD_PATHSPEC(ps,
467 PATHSPEC_FROMTOP |
468 PATHSPEC_MAXDEPTH |
469 PATHSPEC_LITERAL |
470 PATHSPEC_GLOB |
471 PATHSPEC_ICASE |
472 PATHSPEC_EXCLUDE |
473 PATHSPEC_ATTR);
475 if (!ps->nr) {
476 if (!ps->recursive ||
477 !(ps->magic & PATHSPEC_MAXDEPTH) ||
478 ps->max_depth == -1)
479 return MATCHED_RECURSIVELY;
481 if (within_depth(name, namelen, 0, ps->max_depth))
482 return MATCHED_EXACTLY;
483 else
484 return 0;
487 name += prefix;
488 namelen -= prefix;
490 for (i = ps->nr - 1; i >= 0; i--) {
491 int how;
493 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
494 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
495 continue;
497 if (seen && seen[i] == MATCHED_EXACTLY)
498 continue;
500 * Make exclude patterns optional and never report
501 * "pathspec ':(exclude)foo' matches no files"
503 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
504 seen[i] = MATCHED_FNMATCH;
505 how = match_pathspec_item(istate, ps->items+i, prefix, name,
506 namelen, flags);
507 if (ps->recursive &&
508 (ps->magic & PATHSPEC_MAXDEPTH) &&
509 ps->max_depth != -1 &&
510 how && how != MATCHED_FNMATCH) {
511 int len = ps->items[i].len;
512 if (name[len] == '/')
513 len++;
514 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
515 how = MATCHED_EXACTLY;
516 else
517 how = 0;
519 if (how) {
520 if (retval < how)
521 retval = how;
522 if (seen && seen[i] < how)
523 seen[i] = how;
526 return retval;
529 static int match_pathspec_with_flags(struct index_state *istate,
530 const struct pathspec *ps,
531 const char *name, int namelen,
532 int prefix, char *seen, unsigned flags)
534 int positive, negative;
535 positive = do_match_pathspec(istate, ps, name, namelen,
536 prefix, seen, flags);
537 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
538 return positive;
539 negative = do_match_pathspec(istate, ps, name, namelen,
540 prefix, seen,
541 flags | DO_MATCH_EXCLUDE);
542 return negative ? 0 : positive;
545 int match_pathspec(struct index_state *istate,
546 const struct pathspec *ps,
547 const char *name, int namelen,
548 int prefix, char *seen, int is_dir)
550 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
551 return match_pathspec_with_flags(istate, ps, name, namelen,
552 prefix, seen, flags);
556 * Check if a submodule is a superset of the pathspec
558 int submodule_path_match(struct index_state *istate,
559 const struct pathspec *ps,
560 const char *submodule_name,
561 char *seen)
563 int matched = match_pathspec_with_flags(istate, ps, submodule_name,
564 strlen(submodule_name),
565 0, seen,
566 DO_MATCH_DIRECTORY |
567 DO_MATCH_LEADING_PATHSPEC);
568 return matched;
571 int report_path_error(const char *ps_matched,
572 const struct pathspec *pathspec)
575 * Make sure all pathspec matched; otherwise it is an error.
577 int num, errors = 0;
578 for (num = 0; num < pathspec->nr; num++) {
579 int other, found_dup;
581 if (ps_matched[num])
582 continue;
584 * The caller might have fed identical pathspec
585 * twice. Do not barf on such a mistake.
586 * FIXME: parse_pathspec should have eliminated
587 * duplicate pathspec.
589 for (found_dup = other = 0;
590 !found_dup && other < pathspec->nr;
591 other++) {
592 if (other == num || !ps_matched[other])
593 continue;
594 if (!strcmp(pathspec->items[other].original,
595 pathspec->items[num].original))
597 * Ok, we have a match already.
599 found_dup = 1;
601 if (found_dup)
602 continue;
604 error(_("pathspec '%s' did not match any file(s) known to git"),
605 pathspec->items[num].original);
606 errors++;
608 return errors;
612 * Return the length of the "simple" part of a path match limiter.
614 int simple_length(const char *match)
616 int len = -1;
618 for (;;) {
619 unsigned char c = *match++;
620 len++;
621 if (c == '\0' || is_glob_special(c))
622 return len;
626 int no_wildcard(const char *string)
628 return string[simple_length(string)] == '\0';
631 void parse_path_pattern(const char **pattern,
632 int *patternlen,
633 unsigned *flags,
634 int *nowildcardlen)
636 const char *p = *pattern;
637 size_t i, len;
639 *flags = 0;
640 if (*p == '!') {
641 *flags |= PATTERN_FLAG_NEGATIVE;
642 p++;
644 len = strlen(p);
645 if (len && p[len - 1] == '/') {
646 len--;
647 *flags |= PATTERN_FLAG_MUSTBEDIR;
649 for (i = 0; i < len; i++) {
650 if (p[i] == '/')
651 break;
653 if (i == len)
654 *flags |= PATTERN_FLAG_NODIR;
655 *nowildcardlen = simple_length(p);
657 * we should have excluded the trailing slash from 'p' too,
658 * but that's one more allocation. Instead just make sure
659 * nowildcardlen does not exceed real patternlen
661 if (*nowildcardlen > len)
662 *nowildcardlen = len;
663 if (*p == '*' && no_wildcard(p + 1))
664 *flags |= PATTERN_FLAG_ENDSWITH;
665 *pattern = p;
666 *patternlen = len;
669 int pl_hashmap_cmp(const void *cmp_data UNUSED,
670 const struct hashmap_entry *a,
671 const struct hashmap_entry *b,
672 const void *key UNUSED)
674 const struct pattern_entry *ee1 =
675 container_of(a, struct pattern_entry, ent);
676 const struct pattern_entry *ee2 =
677 container_of(b, struct pattern_entry, ent);
679 size_t min_len = ee1->patternlen <= ee2->patternlen
680 ? ee1->patternlen
681 : ee2->patternlen;
683 return fspathncmp(ee1->pattern, ee2->pattern, min_len);
686 static char *dup_and_filter_pattern(const char *pattern)
688 char *set, *read;
689 size_t count = 0;
690 char *result = xstrdup(pattern);
692 set = result;
693 read = result;
695 while (*read) {
696 /* skip escape characters (once) */
697 if (*read == '\\')
698 read++;
700 *set = *read;
702 set++;
703 read++;
704 count++;
706 *set = 0;
708 if (count > 2 &&
709 *(set - 1) == '*' &&
710 *(set - 2) == '/')
711 *(set - 2) = 0;
713 return result;
716 static void add_pattern_to_hashsets(struct pattern_list *pl, struct path_pattern *given)
718 struct pattern_entry *translated;
719 char *truncated;
720 char *data = NULL;
721 const char *prev, *cur, *next;
723 if (!pl->use_cone_patterns)
724 return;
726 if (given->flags & PATTERN_FLAG_NEGATIVE &&
727 given->flags & PATTERN_FLAG_MUSTBEDIR &&
728 !strcmp(given->pattern, "/*")) {
729 pl->full_cone = 0;
730 return;
733 if (!given->flags && !strcmp(given->pattern, "/*")) {
734 pl->full_cone = 1;
735 return;
738 if (given->patternlen < 2 ||
739 *given->pattern != '/' ||
740 strstr(given->pattern, "**")) {
741 /* Not a cone pattern. */
742 warning(_("unrecognized pattern: '%s'"), given->pattern);
743 goto clear_hashmaps;
746 if (!(given->flags & PATTERN_FLAG_MUSTBEDIR) &&
747 strcmp(given->pattern, "/*")) {
748 /* Not a cone pattern. */
749 warning(_("unrecognized pattern: '%s'"), given->pattern);
750 goto clear_hashmaps;
753 prev = given->pattern;
754 cur = given->pattern + 1;
755 next = given->pattern + 2;
757 while (*cur) {
758 /* Watch for glob characters '*', '\', '[', '?' */
759 if (!is_glob_special(*cur))
760 goto increment;
762 /* But only if *prev != '\\' */
763 if (*prev == '\\')
764 goto increment;
766 /* But allow the initial '\' */
767 if (*cur == '\\' &&
768 is_glob_special(*next))
769 goto increment;
771 /* But a trailing '/' then '*' is fine */
772 if (*prev == '/' &&
773 *cur == '*' &&
774 *next == 0)
775 goto increment;
777 /* Not a cone pattern. */
778 warning(_("unrecognized pattern: '%s'"), given->pattern);
779 goto clear_hashmaps;
781 increment:
782 prev++;
783 cur++;
784 next++;
787 if (given->patternlen > 2 &&
788 !strcmp(given->pattern + given->patternlen - 2, "/*")) {
789 if (!(given->flags & PATTERN_FLAG_NEGATIVE)) {
790 /* Not a cone pattern. */
791 warning(_("unrecognized pattern: '%s'"), given->pattern);
792 goto clear_hashmaps;
795 truncated = dup_and_filter_pattern(given->pattern);
797 translated = xmalloc(sizeof(struct pattern_entry));
798 translated->pattern = truncated;
799 translated->patternlen = given->patternlen - 2;
800 hashmap_entry_init(&translated->ent,
801 fspathhash(translated->pattern));
803 if (!hashmap_get_entry(&pl->recursive_hashmap,
804 translated, ent, NULL)) {
805 /* We did not see the "parent" included */
806 warning(_("unrecognized negative pattern: '%s'"),
807 given->pattern);
808 free(truncated);
809 free(translated);
810 goto clear_hashmaps;
813 hashmap_add(&pl->parent_hashmap, &translated->ent);
814 hashmap_remove(&pl->recursive_hashmap, &translated->ent, &data);
815 free(data);
816 return;
819 if (given->flags & PATTERN_FLAG_NEGATIVE) {
820 warning(_("unrecognized negative pattern: '%s'"),
821 given->pattern);
822 goto clear_hashmaps;
825 translated = xmalloc(sizeof(struct pattern_entry));
827 translated->pattern = dup_and_filter_pattern(given->pattern);
828 translated->patternlen = given->patternlen;
829 hashmap_entry_init(&translated->ent,
830 fspathhash(translated->pattern));
832 hashmap_add(&pl->recursive_hashmap, &translated->ent);
834 if (hashmap_get_entry(&pl->parent_hashmap, translated, ent, NULL)) {
835 /* we already included this at the parent level */
836 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
837 given->pattern);
838 goto clear_hashmaps;
841 return;
843 clear_hashmaps:
844 warning(_("disabling cone pattern matching"));
845 hashmap_clear_and_free(&pl->parent_hashmap, struct pattern_entry, ent);
846 hashmap_clear_and_free(&pl->recursive_hashmap, struct pattern_entry, ent);
847 pl->use_cone_patterns = 0;
850 static int hashmap_contains_path(struct hashmap *map,
851 struct strbuf *pattern)
853 struct pattern_entry p;
855 /* Check straight mapping */
856 p.pattern = pattern->buf;
857 p.patternlen = pattern->len;
858 hashmap_entry_init(&p.ent, fspathhash(p.pattern));
859 return !!hashmap_get_entry(map, &p, ent, NULL);
862 int hashmap_contains_parent(struct hashmap *map,
863 const char *path,
864 struct strbuf *buffer)
866 char *slash_pos;
868 strbuf_setlen(buffer, 0);
870 if (path[0] != '/')
871 strbuf_addch(buffer, '/');
873 strbuf_addstr(buffer, path);
875 slash_pos = strrchr(buffer->buf, '/');
877 while (slash_pos > buffer->buf) {
878 strbuf_setlen(buffer, slash_pos - buffer->buf);
880 if (hashmap_contains_path(map, buffer))
881 return 1;
883 slash_pos = strrchr(buffer->buf, '/');
886 return 0;
889 void add_pattern(const char *string, const char *base,
890 int baselen, struct pattern_list *pl, int srcpos)
892 struct path_pattern *pattern;
893 int patternlen;
894 unsigned flags;
895 int nowildcardlen;
897 parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen);
898 if (flags & PATTERN_FLAG_MUSTBEDIR) {
899 FLEXPTR_ALLOC_MEM(pattern, pattern, string, patternlen);
900 } else {
901 pattern = xmalloc(sizeof(*pattern));
902 pattern->pattern = string;
904 pattern->patternlen = patternlen;
905 pattern->nowildcardlen = nowildcardlen;
906 pattern->base = base;
907 pattern->baselen = baselen;
908 pattern->flags = flags;
909 pattern->srcpos = srcpos;
910 ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc);
911 pl->patterns[pl->nr++] = pattern;
912 pattern->pl = pl;
914 add_pattern_to_hashsets(pl, pattern);
917 static int read_skip_worktree_file_from_index(struct index_state *istate,
918 const char *path,
919 size_t *size_out, char **data_out,
920 struct oid_stat *oid_stat)
922 int pos, len;
924 len = strlen(path);
925 pos = index_name_pos(istate, path, len);
926 if (pos < 0)
927 return -1;
928 if (!ce_skip_worktree(istate->cache[pos]))
929 return -1;
931 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
935 * Frees memory within pl which was allocated for exclude patterns and
936 * the file buffer. Does not free pl itself.
938 void clear_pattern_list(struct pattern_list *pl)
940 int i;
942 for (i = 0; i < pl->nr; i++)
943 free(pl->patterns[i]);
944 free(pl->patterns);
945 free(pl->filebuf);
946 hashmap_clear_and_free(&pl->recursive_hashmap, struct pattern_entry, ent);
947 hashmap_clear_and_free(&pl->parent_hashmap, struct pattern_entry, ent);
949 memset(pl, 0, sizeof(*pl));
952 static void trim_trailing_spaces(char *buf)
954 char *p, *last_space = NULL;
956 for (p = buf; *p; p++)
957 switch (*p) {
958 case ' ':
959 if (!last_space)
960 last_space = p;
961 break;
962 case '\\':
963 p++;
964 if (!*p)
965 return;
966 /* fallthrough */
967 default:
968 last_space = NULL;
971 if (last_space)
972 *last_space = '\0';
976 * Given a subdirectory name and "dir" of the current directory,
977 * search the subdir in "dir" and return it, or create a new one if it
978 * does not exist in "dir".
980 * If "name" has the trailing slash, it'll be excluded in the search.
982 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
983 struct untracked_cache_dir *dir,
984 const char *name, int len)
986 int first, last;
987 struct untracked_cache_dir *d;
988 if (!dir)
989 return NULL;
990 if (len && name[len - 1] == '/')
991 len--;
992 first = 0;
993 last = dir->dirs_nr;
994 while (last > first) {
995 int cmp, next = first + ((last - first) >> 1);
996 d = dir->dirs[next];
997 cmp = strncmp(name, d->name, len);
998 if (!cmp && strlen(d->name) > len)
999 cmp = -1;
1000 if (!cmp)
1001 return d;
1002 if (cmp < 0) {
1003 last = next;
1004 continue;
1006 first = next+1;
1009 uc->dir_created++;
1010 FLEX_ALLOC_MEM(d, name, name, len);
1012 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
1013 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
1014 dir->dirs_nr - first);
1015 dir->dirs_nr++;
1016 dir->dirs[first] = d;
1017 return d;
1020 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
1022 int i;
1023 dir->valid = 0;
1024 dir->untracked_nr = 0;
1025 for (i = 0; i < dir->dirs_nr; i++)
1026 do_invalidate_gitignore(dir->dirs[i]);
1029 static void invalidate_gitignore(struct untracked_cache *uc,
1030 struct untracked_cache_dir *dir)
1032 uc->gitignore_invalidated++;
1033 do_invalidate_gitignore(dir);
1036 static void invalidate_directory(struct untracked_cache *uc,
1037 struct untracked_cache_dir *dir)
1039 int i;
1042 * Invalidation increment here is just roughly correct. If
1043 * untracked_nr or any of dirs[].recurse is non-zero, we
1044 * should increment dir_invalidated too. But that's more
1045 * expensive to do.
1047 if (dir->valid)
1048 uc->dir_invalidated++;
1050 dir->valid = 0;
1051 dir->untracked_nr = 0;
1052 for (i = 0; i < dir->dirs_nr; i++)
1053 dir->dirs[i]->recurse = 0;
1056 static int add_patterns_from_buffer(char *buf, size_t size,
1057 const char *base, int baselen,
1058 struct pattern_list *pl);
1060 /* Flags for add_patterns() */
1061 #define PATTERN_NOFOLLOW (1<<0)
1064 * Given a file with name "fname", read it (either from disk, or from
1065 * an index if 'istate' is non-null), parse it and store the
1066 * exclude rules in "pl".
1068 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1069 * stat data from disk (only valid if add_patterns returns zero). If
1070 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1072 static int add_patterns(const char *fname, const char *base, int baselen,
1073 struct pattern_list *pl, struct index_state *istate,
1074 unsigned flags, struct oid_stat *oid_stat)
1076 struct stat st;
1077 int r;
1078 int fd;
1079 size_t size = 0;
1080 char *buf;
1082 if (flags & PATTERN_NOFOLLOW)
1083 fd = open_nofollow(fname, O_RDONLY);
1084 else
1085 fd = open(fname, O_RDONLY);
1087 if (fd < 0 || fstat(fd, &st) < 0) {
1088 if (fd < 0)
1089 warn_on_fopen_errors(fname);
1090 else
1091 close(fd);
1092 if (!istate)
1093 return -1;
1094 r = read_skip_worktree_file_from_index(istate, fname,
1095 &size, &buf,
1096 oid_stat);
1097 if (r != 1)
1098 return r;
1099 } else {
1100 size = xsize_t(st.st_size);
1101 if (size == 0) {
1102 if (oid_stat) {
1103 fill_stat_data(&oid_stat->stat, &st);
1104 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
1105 oid_stat->valid = 1;
1107 close(fd);
1108 return 0;
1110 buf = xmallocz(size);
1111 if (read_in_full(fd, buf, size) != size) {
1112 free(buf);
1113 close(fd);
1114 return -1;
1116 buf[size++] = '\n';
1117 close(fd);
1118 if (oid_stat) {
1119 int pos;
1120 if (oid_stat->valid &&
1121 !match_stat_data_racy(istate, &oid_stat->stat, &st))
1122 ; /* no content change, oid_stat->oid still good */
1123 else if (istate &&
1124 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
1125 !ce_stage(istate->cache[pos]) &&
1126 ce_uptodate(istate->cache[pos]) &&
1127 !would_convert_to_git(istate, fname))
1128 oidcpy(&oid_stat->oid,
1129 &istate->cache[pos]->oid);
1130 else
1131 hash_object_file(the_hash_algo, buf, size,
1132 OBJ_BLOB, &oid_stat->oid);
1133 fill_stat_data(&oid_stat->stat, &st);
1134 oid_stat->valid = 1;
1138 add_patterns_from_buffer(buf, size, base, baselen, pl);
1139 return 0;
1142 static int add_patterns_from_buffer(char *buf, size_t size,
1143 const char *base, int baselen,
1144 struct pattern_list *pl)
1146 int i, lineno = 1;
1147 char *entry;
1149 hashmap_init(&pl->recursive_hashmap, pl_hashmap_cmp, NULL, 0);
1150 hashmap_init(&pl->parent_hashmap, pl_hashmap_cmp, NULL, 0);
1152 pl->filebuf = buf;
1154 if (skip_utf8_bom(&buf, size))
1155 size -= buf - pl->filebuf;
1157 entry = buf;
1159 for (i = 0; i < size; i++) {
1160 if (buf[i] == '\n') {
1161 if (entry != buf + i && entry[0] != '#') {
1162 buf[i - (i && buf[i-1] == '\r')] = 0;
1163 trim_trailing_spaces(entry);
1164 add_pattern(entry, base, baselen, pl, lineno);
1166 lineno++;
1167 entry = buf + i + 1;
1170 return 0;
1173 int add_patterns_from_file_to_list(const char *fname, const char *base,
1174 int baselen, struct pattern_list *pl,
1175 struct index_state *istate,
1176 unsigned flags)
1178 return add_patterns(fname, base, baselen, pl, istate, flags, NULL);
1181 int add_patterns_from_blob_to_list(
1182 struct object_id *oid,
1183 const char *base, int baselen,
1184 struct pattern_list *pl)
1186 char *buf;
1187 size_t size;
1188 int r;
1190 r = do_read_blob(oid, NULL, &size, &buf);
1191 if (r != 1)
1192 return r;
1194 add_patterns_from_buffer(buf, size, base, baselen, pl);
1195 return 0;
1198 struct pattern_list *add_pattern_list(struct dir_struct *dir,
1199 int group_type, const char *src)
1201 struct pattern_list *pl;
1202 struct exclude_list_group *group;
1204 group = &dir->internal.exclude_list_group[group_type];
1205 ALLOC_GROW(group->pl, group->nr + 1, group->alloc);
1206 pl = &group->pl[group->nr++];
1207 memset(pl, 0, sizeof(*pl));
1208 pl->src = src;
1209 return pl;
1213 * Used to set up core.excludesfile and .git/info/exclude lists.
1215 static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname,
1216 struct oid_stat *oid_stat)
1218 struct pattern_list *pl;
1220 * catch setup_standard_excludes() that's called before
1221 * dir->untracked is assigned. That function behaves
1222 * differently when dir->untracked is non-NULL.
1224 if (!dir->untracked)
1225 dir->internal.unmanaged_exclude_files++;
1226 pl = add_pattern_list(dir, EXC_FILE, fname);
1227 if (add_patterns(fname, "", 0, pl, NULL, 0, oid_stat) < 0)
1228 die(_("cannot use %s as an exclude file"), fname);
1231 void add_patterns_from_file(struct dir_struct *dir, const char *fname)
1233 dir->internal.unmanaged_exclude_files++; /* see validate_untracked_cache() */
1234 add_patterns_from_file_1(dir, fname, NULL);
1237 int match_basename(const char *basename, int basenamelen,
1238 const char *pattern, int prefix, int patternlen,
1239 unsigned flags)
1241 if (prefix == patternlen) {
1242 if (patternlen == basenamelen &&
1243 !fspathncmp(pattern, basename, basenamelen))
1244 return 1;
1245 } else if (flags & PATTERN_FLAG_ENDSWITH) {
1246 /* "*literal" matching against "fooliteral" */
1247 if (patternlen - 1 <= basenamelen &&
1248 !fspathncmp(pattern + 1,
1249 basename + basenamelen - (patternlen - 1),
1250 patternlen - 1))
1251 return 1;
1252 } else {
1253 if (fnmatch_icase_mem(pattern, patternlen,
1254 basename, basenamelen,
1255 0) == 0)
1256 return 1;
1258 return 0;
1261 int match_pathname(const char *pathname, int pathlen,
1262 const char *base, int baselen,
1263 const char *pattern, int prefix, int patternlen)
1265 const char *name;
1266 int namelen;
1269 * match with FNM_PATHNAME; the pattern has base implicitly
1270 * in front of it.
1272 if (*pattern == '/') {
1273 pattern++;
1274 patternlen--;
1275 prefix--;
1279 * baselen does not count the trailing slash. base[] may or
1280 * may not end with a trailing slash though.
1282 if (pathlen < baselen + 1 ||
1283 (baselen && pathname[baselen] != '/') ||
1284 fspathncmp(pathname, base, baselen))
1285 return 0;
1287 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1288 name = pathname + pathlen - namelen;
1290 if (prefix) {
1292 * if the non-wildcard part is longer than the
1293 * remaining pathname, surely it cannot match.
1295 if (prefix > namelen)
1296 return 0;
1298 if (fspathncmp(pattern, name, prefix))
1299 return 0;
1300 pattern += prefix;
1301 patternlen -= prefix;
1302 name += prefix;
1303 namelen -= prefix;
1306 * If the whole pattern did not have a wildcard,
1307 * then our prefix match is all we need; we
1308 * do not need to call fnmatch at all.
1310 if (!patternlen && !namelen)
1311 return 1;
1314 return fnmatch_icase_mem(pattern, patternlen,
1315 name, namelen,
1316 WM_PATHNAME) == 0;
1320 * Scan the given exclude list in reverse to see whether pathname
1321 * should be ignored. The first match (i.e. the last on the list), if
1322 * any, determines the fate. Returns the exclude_list element which
1323 * matched, or NULL for undecided.
1325 static struct path_pattern *last_matching_pattern_from_list(const char *pathname,
1326 int pathlen,
1327 const char *basename,
1328 int *dtype,
1329 struct pattern_list *pl,
1330 struct index_state *istate)
1332 struct path_pattern *res = NULL; /* undecided */
1333 int i;
1335 if (!pl->nr)
1336 return NULL; /* undefined */
1338 for (i = pl->nr - 1; 0 <= i; i--) {
1339 struct path_pattern *pattern = pl->patterns[i];
1340 const char *exclude = pattern->pattern;
1341 int prefix = pattern->nowildcardlen;
1343 if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) {
1344 *dtype = resolve_dtype(*dtype, istate, pathname, pathlen);
1345 if (*dtype != DT_DIR)
1346 continue;
1349 if (pattern->flags & PATTERN_FLAG_NODIR) {
1350 if (match_basename(basename,
1351 pathlen - (basename - pathname),
1352 exclude, prefix, pattern->patternlen,
1353 pattern->flags)) {
1354 res = pattern;
1355 break;
1357 continue;
1360 assert(pattern->baselen == 0 ||
1361 pattern->base[pattern->baselen - 1] == '/');
1362 if (match_pathname(pathname, pathlen,
1363 pattern->base,
1364 pattern->baselen ? pattern->baselen - 1 : 0,
1365 exclude, prefix, pattern->patternlen)) {
1366 res = pattern;
1367 break;
1370 return res;
1374 * Scan the list of patterns to determine if the ordered list
1375 * of patterns matches on 'pathname'.
1377 * Return 1 for a match, 0 for not matched and -1 for undecided.
1379 enum pattern_match_result path_matches_pattern_list(
1380 const char *pathname, int pathlen,
1381 const char *basename, int *dtype,
1382 struct pattern_list *pl,
1383 struct index_state *istate)
1385 struct path_pattern *pattern;
1386 struct strbuf parent_pathname = STRBUF_INIT;
1387 int result = NOT_MATCHED;
1388 size_t slash_pos;
1390 if (!pl->use_cone_patterns) {
1391 pattern = last_matching_pattern_from_list(pathname, pathlen, basename,
1392 dtype, pl, istate);
1393 if (pattern) {
1394 if (pattern->flags & PATTERN_FLAG_NEGATIVE)
1395 return NOT_MATCHED;
1396 else
1397 return MATCHED;
1400 return UNDECIDED;
1403 if (pl->full_cone)
1404 return MATCHED;
1406 strbuf_addch(&parent_pathname, '/');
1407 strbuf_add(&parent_pathname, pathname, pathlen);
1410 * Directory entries are matched if and only if a file
1411 * contained immediately within them is matched. For the
1412 * case of a directory entry, modify the path to create
1413 * a fake filename within this directory, allowing us to
1414 * use the file-base matching logic in an equivalent way.
1416 if (parent_pathname.len > 0 &&
1417 parent_pathname.buf[parent_pathname.len - 1] == '/') {
1418 slash_pos = parent_pathname.len - 1;
1419 strbuf_add(&parent_pathname, "-", 1);
1420 } else {
1421 const char *slash_ptr = strrchr(parent_pathname.buf, '/');
1422 slash_pos = slash_ptr ? slash_ptr - parent_pathname.buf : 0;
1425 if (hashmap_contains_path(&pl->recursive_hashmap,
1426 &parent_pathname)) {
1427 result = MATCHED_RECURSIVE;
1428 goto done;
1431 if (!slash_pos) {
1432 /* include every file in root */
1433 result = MATCHED;
1434 goto done;
1437 strbuf_setlen(&parent_pathname, slash_pos);
1439 if (hashmap_contains_path(&pl->parent_hashmap, &parent_pathname)) {
1440 result = MATCHED;
1441 goto done;
1444 if (hashmap_contains_parent(&pl->recursive_hashmap,
1445 pathname,
1446 &parent_pathname))
1447 result = MATCHED_RECURSIVE;
1449 done:
1450 strbuf_release(&parent_pathname);
1451 return result;
1454 int init_sparse_checkout_patterns(struct index_state *istate)
1456 if (!core_apply_sparse_checkout)
1457 return 1;
1458 if (istate->sparse_checkout_patterns)
1459 return 0;
1461 CALLOC_ARRAY(istate->sparse_checkout_patterns, 1);
1463 if (get_sparse_checkout_patterns(istate->sparse_checkout_patterns) < 0) {
1464 FREE_AND_NULL(istate->sparse_checkout_patterns);
1465 return -1;
1468 return 0;
1471 static int path_in_sparse_checkout_1(const char *path,
1472 struct index_state *istate,
1473 int require_cone_mode)
1475 int dtype = DT_REG;
1476 enum pattern_match_result match = UNDECIDED;
1477 const char *end, *slash;
1480 * We default to accepting a path if the path is empty, there are no
1481 * patterns, or the patterns are of the wrong type.
1483 if (!*path ||
1484 init_sparse_checkout_patterns(istate) ||
1485 (require_cone_mode &&
1486 !istate->sparse_checkout_patterns->use_cone_patterns))
1487 return 1;
1490 * If UNDECIDED, use the match from the parent dir (recursively), or
1491 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1492 * never returns UNDECIDED, so we will execute only one iteration in
1493 * this case.
1495 for (end = path + strlen(path);
1496 end > path && match == UNDECIDED;
1497 end = slash) {
1499 for (slash = end - 1; slash > path && *slash != '/'; slash--)
1500 ; /* do nothing */
1502 match = path_matches_pattern_list(path, end - path,
1503 slash > path ? slash + 1 : path, &dtype,
1504 istate->sparse_checkout_patterns, istate);
1506 /* We are going to match the parent dir now */
1507 dtype = DT_DIR;
1509 return match > 0;
1512 int path_in_sparse_checkout(const char *path,
1513 struct index_state *istate)
1515 return path_in_sparse_checkout_1(path, istate, 0);
1518 int path_in_cone_mode_sparse_checkout(const char *path,
1519 struct index_state *istate)
1521 return path_in_sparse_checkout_1(path, istate, 1);
1524 static struct path_pattern *last_matching_pattern_from_lists(
1525 struct dir_struct *dir, struct index_state *istate,
1526 const char *pathname, int pathlen,
1527 const char *basename, int *dtype_p)
1529 int i, j;
1530 struct exclude_list_group *group;
1531 struct path_pattern *pattern;
1532 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1533 group = &dir->internal.exclude_list_group[i];
1534 for (j = group->nr - 1; j >= 0; j--) {
1535 pattern = last_matching_pattern_from_list(
1536 pathname, pathlen, basename, dtype_p,
1537 &group->pl[j], istate);
1538 if (pattern)
1539 return pattern;
1542 return NULL;
1546 * Loads the per-directory exclude list for the substring of base
1547 * which has a char length of baselen.
1549 static void prep_exclude(struct dir_struct *dir,
1550 struct index_state *istate,
1551 const char *base, int baselen)
1553 struct exclude_list_group *group;
1554 struct pattern_list *pl;
1555 struct exclude_stack *stk = NULL;
1556 struct untracked_cache_dir *untracked;
1557 int current;
1559 group = &dir->internal.exclude_list_group[EXC_DIRS];
1562 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1563 * which originate from directories not in the prefix of the
1564 * path being checked.
1566 while ((stk = dir->internal.exclude_stack) != NULL) {
1567 if (stk->baselen <= baselen &&
1568 !strncmp(dir->internal.basebuf.buf, base, stk->baselen))
1569 break;
1570 pl = &group->pl[dir->internal.exclude_stack->exclude_ix];
1571 dir->internal.exclude_stack = stk->prev;
1572 dir->internal.pattern = NULL;
1573 free((char *)pl->src); /* see strbuf_detach() below */
1574 clear_pattern_list(pl);
1575 free(stk);
1576 group->nr--;
1579 /* Skip traversing into sub directories if the parent is excluded */
1580 if (dir->internal.pattern)
1581 return;
1584 * Lazy initialization. All call sites currently just
1585 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1586 * them seems lots of work for little benefit.
1588 if (!dir->internal.basebuf.buf)
1589 strbuf_init(&dir->internal.basebuf, PATH_MAX);
1591 /* Read from the parent directories and push them down. */
1592 current = stk ? stk->baselen : -1;
1593 strbuf_setlen(&dir->internal.basebuf, current < 0 ? 0 : current);
1594 if (dir->untracked)
1595 untracked = stk ? stk->ucd : dir->untracked->root;
1596 else
1597 untracked = NULL;
1599 while (current < baselen) {
1600 const char *cp;
1601 struct oid_stat oid_stat;
1603 CALLOC_ARRAY(stk, 1);
1604 if (current < 0) {
1605 cp = base;
1606 current = 0;
1607 } else {
1608 cp = strchr(base + current + 1, '/');
1609 if (!cp)
1610 die("oops in prep_exclude");
1611 cp++;
1612 untracked =
1613 lookup_untracked(dir->untracked,
1614 untracked,
1615 base + current,
1616 cp - base - current);
1618 stk->prev = dir->internal.exclude_stack;
1619 stk->baselen = cp - base;
1620 stk->exclude_ix = group->nr;
1621 stk->ucd = untracked;
1622 pl = add_pattern_list(dir, EXC_DIRS, NULL);
1623 strbuf_add(&dir->internal.basebuf, base + current, stk->baselen - current);
1624 assert(stk->baselen == dir->internal.basebuf.len);
1626 /* Abort if the directory is excluded */
1627 if (stk->baselen) {
1628 int dt = DT_DIR;
1629 dir->internal.basebuf.buf[stk->baselen - 1] = 0;
1630 dir->internal.pattern = last_matching_pattern_from_lists(dir,
1631 istate,
1632 dir->internal.basebuf.buf, stk->baselen - 1,
1633 dir->internal.basebuf.buf + current, &dt);
1634 dir->internal.basebuf.buf[stk->baselen - 1] = '/';
1635 if (dir->internal.pattern &&
1636 dir->internal.pattern->flags & PATTERN_FLAG_NEGATIVE)
1637 dir->internal.pattern = NULL;
1638 if (dir->internal.pattern) {
1639 dir->internal.exclude_stack = stk;
1640 return;
1644 /* Try to read per-directory file */
1645 oidclr(&oid_stat.oid);
1646 oid_stat.valid = 0;
1647 if (dir->exclude_per_dir &&
1649 * If we know that no files have been added in
1650 * this directory (i.e. valid_cached_dir() has
1651 * been executed and set untracked->valid) ..
1653 (!untracked || !untracked->valid ||
1655 * .. and .gitignore does not exist before
1656 * (i.e. null exclude_oid). Then we can skip
1657 * loading .gitignore, which would result in
1658 * ENOENT anyway.
1660 !is_null_oid(&untracked->exclude_oid))) {
1662 * dir->internal.basebuf gets reused by the traversal,
1663 * but we need fname to remain unchanged to ensure the
1664 * src member of each struct path_pattern correctly
1665 * back-references its source file. Other invocations
1666 * of add_pattern_list provide stable strings, so we
1667 * strbuf_detach() and free() here in the caller.
1669 struct strbuf sb = STRBUF_INIT;
1670 strbuf_addbuf(&sb, &dir->internal.basebuf);
1671 strbuf_addstr(&sb, dir->exclude_per_dir);
1672 pl->src = strbuf_detach(&sb, NULL);
1673 add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1674 PATTERN_NOFOLLOW,
1675 untracked ? &oid_stat : NULL);
1678 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1679 * will first be called in valid_cached_dir() then maybe many
1680 * times more in last_matching_pattern(). When the cache is
1681 * used, last_matching_pattern() will not be called and
1682 * reading .gitignore content will be a waste.
1684 * So when it's called by valid_cached_dir() and we can get
1685 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1686 * modified on work tree), we could delay reading the
1687 * .gitignore content until we absolutely need it in
1688 * last_matching_pattern(). Be careful about ignore rule
1689 * order, though, if you do that.
1691 if (untracked &&
1692 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1693 invalidate_gitignore(dir->untracked, untracked);
1694 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1696 dir->internal.exclude_stack = stk;
1697 current = stk->baselen;
1699 strbuf_setlen(&dir->internal.basebuf, baselen);
1703 * Loads the exclude lists for the directory containing pathname, then
1704 * scans all exclude lists to determine whether pathname is excluded.
1705 * Returns the exclude_list element which matched, or NULL for
1706 * undecided.
1708 struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1709 struct index_state *istate,
1710 const char *pathname,
1711 int *dtype_p)
1713 int pathlen = strlen(pathname);
1714 const char *basename = strrchr(pathname, '/');
1715 basename = (basename) ? basename+1 : pathname;
1717 prep_exclude(dir, istate, pathname, basename-pathname);
1719 if (dir->internal.pattern)
1720 return dir->internal.pattern;
1722 return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1723 basename, dtype_p);
1727 * Loads the exclude lists for the directory containing pathname, then
1728 * scans all exclude lists to determine whether pathname is excluded.
1729 * Returns 1 if true, otherwise 0.
1731 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1732 const char *pathname, int *dtype_p)
1734 struct path_pattern *pattern =
1735 last_matching_pattern(dir, istate, pathname, dtype_p);
1736 if (pattern)
1737 return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1738 return 0;
1741 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1743 struct dir_entry *ent;
1745 FLEX_ALLOC_MEM(ent, name, pathname, len);
1746 ent->len = len;
1747 return ent;
1750 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1751 struct index_state *istate,
1752 const char *pathname, int len)
1754 if (index_file_exists(istate, pathname, len, ignore_case))
1755 return NULL;
1757 ALLOC_GROW(dir->entries, dir->nr+1, dir->internal.alloc);
1758 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1761 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1762 struct index_state *istate,
1763 const char *pathname, int len)
1765 if (!index_name_is_other(istate, pathname, len))
1766 return NULL;
1768 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->internal.ignored_alloc);
1769 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1772 enum exist_status {
1773 index_nonexistent = 0,
1774 index_directory,
1775 index_gitdir
1779 * Do not use the alphabetically sorted index to look up
1780 * the directory name; instead, use the case insensitive
1781 * directory hash.
1783 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1784 const char *dirname, int len)
1786 struct cache_entry *ce;
1788 if (index_dir_exists(istate, dirname, len))
1789 return index_directory;
1791 ce = index_file_exists(istate, dirname, len, ignore_case);
1792 if (ce && S_ISGITLINK(ce->ce_mode))
1793 return index_gitdir;
1795 return index_nonexistent;
1799 * The index sorts alphabetically by entry name, which
1800 * means that a gitlink sorts as '\0' at the end, while
1801 * a directory (which is defined not as an entry, but as
1802 * the files it contains) will sort with the '/' at the
1803 * end.
1805 static enum exist_status directory_exists_in_index(struct index_state *istate,
1806 const char *dirname, int len)
1808 int pos;
1810 if (ignore_case)
1811 return directory_exists_in_index_icase(istate, dirname, len);
1813 pos = index_name_pos(istate, dirname, len);
1814 if (pos < 0)
1815 pos = -pos-1;
1816 while (pos < istate->cache_nr) {
1817 const struct cache_entry *ce = istate->cache[pos++];
1818 unsigned char endchar;
1820 if (strncmp(ce->name, dirname, len))
1821 break;
1822 endchar = ce->name[len];
1823 if (endchar > '/')
1824 break;
1825 if (endchar == '/')
1826 return index_directory;
1827 if (!endchar && S_ISGITLINK(ce->ce_mode))
1828 return index_gitdir;
1830 return index_nonexistent;
1834 * When we find a directory when traversing the filesystem, we
1835 * have three distinct cases:
1837 * - ignore it
1838 * - see it as a directory
1839 * - recurse into it
1841 * and which one we choose depends on a combination of existing
1842 * git index contents and the flags passed into the directory
1843 * traversal routine.
1845 * Case 1: If we *already* have entries in the index under that
1846 * directory name, we always recurse into the directory to see
1847 * all the files.
1849 * Case 2: If we *already* have that directory name as a gitlink,
1850 * we always continue to see it as a gitlink, regardless of whether
1851 * there is an actual git directory there or not (it might not
1852 * be checked out as a subproject!)
1854 * Case 3: if we didn't have it in the index previously, we
1855 * have a few sub-cases:
1857 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1858 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1859 * also true, in which case we need to check if it contains any
1860 * untracked and / or ignored files.
1861 * (b) if it looks like a git directory and we don't have the
1862 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1863 * show it as a directory.
1864 * (c) otherwise, we recurse into it.
1866 static enum path_treatment treat_directory(struct dir_struct *dir,
1867 struct index_state *istate,
1868 struct untracked_cache_dir *untracked,
1869 const char *dirname, int len, int baselen, int excluded,
1870 const struct pathspec *pathspec)
1873 * WARNING: From this function, you can return path_recurse or you
1874 * can call read_directory_recursive() (or neither), but
1875 * you CAN'T DO BOTH.
1877 enum path_treatment state;
1878 int matches_how = 0;
1879 int check_only, stop_early;
1880 int old_ignored_nr, old_untracked_nr;
1881 /* The "len-1" is to strip the final '/' */
1882 enum exist_status status = directory_exists_in_index(istate, dirname, len-1);
1884 if (status == index_directory)
1885 return path_recurse;
1886 if (status == index_gitdir)
1887 return path_none;
1888 if (status != index_nonexistent)
1889 BUG("Unhandled value for directory_exists_in_index: %d\n", status);
1892 * We don't want to descend into paths that don't match the necessary
1893 * patterns. Clearly, if we don't have a pathspec, then we can't check
1894 * for matching patterns. Also, if (excluded) then we know we matched
1895 * the exclusion patterns so as an optimization we can skip checking
1896 * for matching patterns.
1898 if (pathspec && !excluded) {
1899 matches_how = match_pathspec_with_flags(istate, pathspec,
1900 dirname, len,
1901 0 /* prefix */,
1902 NULL /* seen */,
1903 DO_MATCH_LEADING_PATHSPEC);
1904 if (!matches_how)
1905 return path_none;
1909 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1910 !(dir->flags & DIR_NO_GITLINKS)) {
1912 * Determine if `dirname` is a nested repo by confirming that:
1913 * 1) we are in a nonbare repository, and
1914 * 2) `dirname` is not an immediate parent of `the_repository->gitdir`,
1915 * which could occur if the git_dir or worktree location was
1916 * manually configured by the user; see t2205 testcases 1-3 for
1917 * examples where this matters
1919 int nested_repo;
1920 struct strbuf sb = STRBUF_INIT;
1921 strbuf_addstr(&sb, dirname);
1922 nested_repo = is_nonbare_repository_dir(&sb);
1924 if (nested_repo) {
1925 char *real_dirname, *real_gitdir;
1926 strbuf_addstr(&sb, ".git");
1927 real_dirname = real_pathdup(sb.buf, 1);
1928 real_gitdir = real_pathdup(the_repository->gitdir, 1);
1930 nested_repo = !!strcmp(real_dirname, real_gitdir);
1931 free(real_gitdir);
1932 free(real_dirname);
1934 strbuf_release(&sb);
1936 if (nested_repo) {
1937 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1938 (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC))
1939 return path_none;
1940 return excluded ? path_excluded : path_untracked;
1944 if (!(dir->flags & DIR_SHOW_OTHER_DIRECTORIES)) {
1945 if (excluded &&
1946 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1947 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1950 * This is an excluded directory and we are
1951 * showing ignored paths that match an exclude
1952 * pattern. (e.g. show directory as ignored
1953 * only if it matches an exclude pattern).
1954 * This path will either be 'path_excluded`
1955 * (if we are showing empty directories or if
1956 * the directory is not empty), or will be
1957 * 'path_none' (empty directory, and we are
1958 * not showing empty directories).
1960 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1961 return path_excluded;
1963 if (read_directory_recursive(dir, istate, dirname, len,
1964 untracked, 1, 1, pathspec) == path_excluded)
1965 return path_excluded;
1967 return path_none;
1969 return path_recurse;
1972 assert(dir->flags & DIR_SHOW_OTHER_DIRECTORIES);
1975 * If we have a pathspec which could match something _below_ this
1976 * directory (e.g. when checking 'subdir/' having a pathspec like
1977 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1978 * need to recurse.
1980 if (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC)
1981 return path_recurse;
1983 /* Special cases for where this directory is excluded/ignored */
1984 if (excluded) {
1986 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1987 * hiding empty directories, there is no need to
1988 * recurse into an ignored directory.
1990 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1991 return path_excluded;
1994 * Even if we are hiding empty directories, we can still avoid
1995 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
1996 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1998 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
1999 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
2000 return path_excluded;
2004 * Other than the path_recurse case above, we only need to
2005 * recurse into untracked directories if any of the following
2006 * bits is set:
2007 * - DIR_SHOW_IGNORED (because then we need to determine if
2008 * there are ignored entries below)
2009 * - DIR_SHOW_IGNORED_TOO (same as above)
2010 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
2011 * the directory is empty)
2013 if (!excluded &&
2014 !(dir->flags & (DIR_SHOW_IGNORED |
2015 DIR_SHOW_IGNORED_TOO |
2016 DIR_HIDE_EMPTY_DIRECTORIES))) {
2017 return path_untracked;
2021 * Even if we don't want to know all the paths under an untracked or
2022 * ignored directory, we may still need to go into the directory to
2023 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2024 * an empty directory should be path_none instead of path_excluded or
2025 * path_untracked).
2027 check_only = ((dir->flags & DIR_HIDE_EMPTY_DIRECTORIES) &&
2028 !(dir->flags & DIR_SHOW_IGNORED_TOO));
2031 * However, there's another optimization possible as a subset of
2032 * check_only, based on the cases we have to consider:
2033 * A) Directory matches no exclude patterns:
2034 * * Directory is empty => path_none
2035 * * Directory has an untracked file under it => path_untracked
2036 * * Directory has only ignored files under it => path_excluded
2037 * B) Directory matches an exclude pattern:
2038 * * Directory is empty => path_none
2039 * * Directory has an untracked file under it => path_excluded
2040 * * Directory has only ignored files under it => path_excluded
2041 * In case A, we can exit as soon as we've found an untracked
2042 * file but otherwise have to walk all files. In case B, though,
2043 * we can stop at the first file we find under the directory.
2045 stop_early = check_only && excluded;
2048 * If /every/ file within an untracked directory is ignored, then
2049 * we want to treat the directory as ignored (for e.g. status
2050 * --porcelain), without listing the individual ignored files
2051 * underneath. To do so, we'll save the current ignored_nr, and
2052 * pop all the ones added after it if it turns out the entire
2053 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2054 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2055 * untracked paths so will need to pop all those off the last
2056 * after we traverse.
2058 old_ignored_nr = dir->ignored_nr;
2059 old_untracked_nr = dir->nr;
2061 /* Actually recurse into dirname now, we'll fixup the state later. */
2062 untracked = lookup_untracked(dir->untracked, untracked,
2063 dirname + baselen, len - baselen);
2064 state = read_directory_recursive(dir, istate, dirname, len, untracked,
2065 check_only, stop_early, pathspec);
2067 /* There are a variety of reasons we may need to fixup the state... */
2068 if (state == path_excluded) {
2069 /* state == path_excluded implies all paths under
2070 * dirname were ignored...
2072 * if running e.g. `git status --porcelain --ignored=matching`,
2073 * then we want to see the subpaths that are ignored.
2075 * if running e.g. just `git status --porcelain`, then
2076 * we just want the directory itself to be listed as ignored
2077 * and not the individual paths underneath.
2079 int want_ignored_subpaths =
2080 ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2081 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING));
2083 if (want_ignored_subpaths) {
2085 * with --ignored=matching, we want the subpaths
2086 * INSTEAD of the directory itself.
2088 state = path_none;
2089 } else {
2090 int i;
2091 for (i = old_ignored_nr + 1; i<dir->ignored_nr; ++i)
2092 FREE_AND_NULL(dir->ignored[i]);
2093 dir->ignored_nr = old_ignored_nr;
2098 * We may need to ignore some of the untracked paths we found while
2099 * traversing subdirectories.
2101 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2102 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2103 int i;
2104 for (i = old_untracked_nr + 1; i<dir->nr; ++i)
2105 FREE_AND_NULL(dir->entries[i]);
2106 dir->nr = old_untracked_nr;
2110 * If there is nothing under the current directory and we are not
2111 * hiding empty directories, then we need to report on the
2112 * untracked or ignored status of the directory itself.
2114 if (state == path_none && !(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
2115 state = excluded ? path_excluded : path_untracked;
2117 return state;
2121 * This is an inexact early pruning of any recursive directory
2122 * reading - if the path cannot possibly be in the pathspec,
2123 * return true, and we'll skip it early.
2125 static int simplify_away(const char *path, int pathlen,
2126 const struct pathspec *pathspec)
2128 int i;
2130 if (!pathspec || !pathspec->nr)
2131 return 0;
2133 GUARD_PATHSPEC(pathspec,
2134 PATHSPEC_FROMTOP |
2135 PATHSPEC_MAXDEPTH |
2136 PATHSPEC_LITERAL |
2137 PATHSPEC_GLOB |
2138 PATHSPEC_ICASE |
2139 PATHSPEC_EXCLUDE |
2140 PATHSPEC_ATTR);
2142 for (i = 0; i < pathspec->nr; i++) {
2143 const struct pathspec_item *item = &pathspec->items[i];
2144 int len = item->nowildcard_len;
2146 if (len > pathlen)
2147 len = pathlen;
2148 if (!ps_strncmp(item, item->match, path, len))
2149 return 0;
2152 return 1;
2156 * This function tells us whether an excluded path matches a
2157 * list of "interesting" pathspecs. That is, whether a path matched
2158 * by any of the pathspecs could possibly be ignored by excluding
2159 * the specified path. This can happen if:
2161 * 1. the path is mentioned explicitly in the pathspec
2163 * 2. the path is a directory prefix of some element in the
2164 * pathspec
2166 static int exclude_matches_pathspec(const char *path, int pathlen,
2167 const struct pathspec *pathspec)
2169 int i;
2171 if (!pathspec || !pathspec->nr)
2172 return 0;
2174 GUARD_PATHSPEC(pathspec,
2175 PATHSPEC_FROMTOP |
2176 PATHSPEC_MAXDEPTH |
2177 PATHSPEC_LITERAL |
2178 PATHSPEC_GLOB |
2179 PATHSPEC_ICASE |
2180 PATHSPEC_EXCLUDE);
2182 for (i = 0; i < pathspec->nr; i++) {
2183 const struct pathspec_item *item = &pathspec->items[i];
2184 int len = item->nowildcard_len;
2186 if (len == pathlen &&
2187 !ps_strncmp(item, item->match, path, pathlen))
2188 return 1;
2189 if (len > pathlen &&
2190 item->match[pathlen] == '/' &&
2191 !ps_strncmp(item, item->match, path, pathlen))
2192 return 1;
2194 return 0;
2197 static int get_index_dtype(struct index_state *istate,
2198 const char *path, int len)
2200 int pos;
2201 const struct cache_entry *ce;
2203 ce = index_file_exists(istate, path, len, 0);
2204 if (ce) {
2205 if (!ce_uptodate(ce))
2206 return DT_UNKNOWN;
2207 if (S_ISGITLINK(ce->ce_mode))
2208 return DT_DIR;
2210 * Nobody actually cares about the
2211 * difference between DT_LNK and DT_REG
2213 return DT_REG;
2216 /* Try to look it up as a directory */
2217 pos = index_name_pos(istate, path, len);
2218 if (pos >= 0)
2219 return DT_UNKNOWN;
2220 pos = -pos-1;
2221 while (pos < istate->cache_nr) {
2222 ce = istate->cache[pos++];
2223 if (strncmp(ce->name, path, len))
2224 break;
2225 if (ce->name[len] > '/')
2226 break;
2227 if (ce->name[len] < '/')
2228 continue;
2229 if (!ce_uptodate(ce))
2230 break; /* continue? */
2231 return DT_DIR;
2233 return DT_UNKNOWN;
2236 static int resolve_dtype(int dtype, struct index_state *istate,
2237 const char *path, int len)
2239 struct stat st;
2241 if (dtype != DT_UNKNOWN)
2242 return dtype;
2243 dtype = get_index_dtype(istate, path, len);
2244 if (dtype != DT_UNKNOWN)
2245 return dtype;
2246 if (lstat(path, &st))
2247 return dtype;
2248 if (S_ISREG(st.st_mode))
2249 return DT_REG;
2250 if (S_ISDIR(st.st_mode))
2251 return DT_DIR;
2252 if (S_ISLNK(st.st_mode))
2253 return DT_LNK;
2254 return dtype;
2257 static enum path_treatment treat_path_fast(struct dir_struct *dir,
2258 struct cached_dir *cdir,
2259 struct index_state *istate,
2260 struct strbuf *path,
2261 int baselen,
2262 const struct pathspec *pathspec)
2265 * WARNING: From this function, you can return path_recurse or you
2266 * can call read_directory_recursive() (or neither), but
2267 * you CAN'T DO BOTH.
2269 strbuf_setlen(path, baselen);
2270 if (!cdir->ucd) {
2271 strbuf_addstr(path, cdir->file);
2272 return path_untracked;
2274 strbuf_addstr(path, cdir->ucd->name);
2275 /* treat_one_path() does this before it calls treat_directory() */
2276 strbuf_complete(path, '/');
2277 if (cdir->ucd->check_only)
2279 * check_only is set as a result of treat_directory() getting
2280 * to its bottom. Verify again the same set of directories
2281 * with check_only set.
2283 return read_directory_recursive(dir, istate, path->buf, path->len,
2284 cdir->ucd, 1, 0, pathspec);
2286 * We get path_recurse in the first run when
2287 * directory_exists_in_index() returns index_nonexistent. We
2288 * are sure that new changes in the index does not impact the
2289 * outcome. Return now.
2291 return path_recurse;
2294 static enum path_treatment treat_path(struct dir_struct *dir,
2295 struct untracked_cache_dir *untracked,
2296 struct cached_dir *cdir,
2297 struct index_state *istate,
2298 struct strbuf *path,
2299 int baselen,
2300 const struct pathspec *pathspec)
2302 int has_path_in_index, dtype, excluded;
2304 if (!cdir->d_name)
2305 return treat_path_fast(dir, cdir, istate, path,
2306 baselen, pathspec);
2307 if (is_dot_or_dotdot(cdir->d_name) || !fspathcmp(cdir->d_name, ".git"))
2308 return path_none;
2309 strbuf_setlen(path, baselen);
2310 strbuf_addstr(path, cdir->d_name);
2311 if (simplify_away(path->buf, path->len, pathspec))
2312 return path_none;
2314 dtype = resolve_dtype(cdir->d_type, istate, path->buf, path->len);
2316 /* Always exclude indexed files */
2317 has_path_in_index = !!index_file_exists(istate, path->buf, path->len,
2318 ignore_case);
2319 if (dtype != DT_DIR && has_path_in_index)
2320 return path_none;
2323 * When we are looking at a directory P in the working tree,
2324 * there are three cases:
2326 * (1) P exists in the index. Everything inside the directory P in
2327 * the working tree needs to go when P is checked out from the
2328 * index.
2330 * (2) P does not exist in the index, but there is P/Q in the index.
2331 * We know P will stay a directory when we check out the contents
2332 * of the index, but we do not know yet if there is a directory
2333 * P/Q in the working tree to be killed, so we need to recurse.
2335 * (3) P does not exist in the index, and there is no P/Q in the index
2336 * to require P to be a directory, either. Only in this case, we
2337 * know that everything inside P will not be killed without
2338 * recursing.
2340 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
2341 (dtype == DT_DIR) &&
2342 !has_path_in_index &&
2343 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
2344 return path_none;
2346 excluded = is_excluded(dir, istate, path->buf, &dtype);
2349 * Excluded? If we don't explicitly want to show
2350 * ignored files, ignore it
2352 if (excluded && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
2353 return path_excluded;
2355 switch (dtype) {
2356 default:
2357 return path_none;
2358 case DT_DIR:
2360 * WARNING: Do not ignore/amend the return value from
2361 * treat_directory(), and especially do not change it to return
2362 * path_recurse as that can cause exponential slowdown.
2363 * Instead, modify treat_directory() to return the right value.
2365 strbuf_addch(path, '/');
2366 return treat_directory(dir, istate, untracked,
2367 path->buf, path->len,
2368 baselen, excluded, pathspec);
2369 case DT_REG:
2370 case DT_LNK:
2371 if (pathspec &&
2372 !match_pathspec(istate, pathspec, path->buf, path->len,
2373 0 /* prefix */, NULL /* seen */,
2374 0 /* is_dir */))
2375 return path_none;
2376 if (excluded)
2377 return path_excluded;
2378 return path_untracked;
2382 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
2384 if (!dir)
2385 return;
2386 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
2387 dir->untracked_alloc);
2388 dir->untracked[dir->untracked_nr++] = xstrdup(name);
2391 static int valid_cached_dir(struct dir_struct *dir,
2392 struct untracked_cache_dir *untracked,
2393 struct index_state *istate,
2394 struct strbuf *path,
2395 int check_only)
2397 struct stat st;
2399 if (!untracked)
2400 return 0;
2403 * With fsmonitor, we can trust the untracked cache's valid field.
2405 refresh_fsmonitor(istate);
2406 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
2407 if (lstat(path->len ? path->buf : ".", &st)) {
2408 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
2409 return 0;
2411 if (!untracked->valid ||
2412 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
2413 fill_stat_data(&untracked->stat_data, &st);
2414 return 0;
2418 if (untracked->check_only != !!check_only)
2419 return 0;
2422 * prep_exclude will be called eventually on this directory,
2423 * but it's called much later in last_matching_pattern(). We
2424 * need it now to determine the validity of the cache for this
2425 * path. The next calls will be nearly no-op, the way
2426 * prep_exclude() is designed.
2428 if (path->len && path->buf[path->len - 1] != '/') {
2429 strbuf_addch(path, '/');
2430 prep_exclude(dir, istate, path->buf, path->len);
2431 strbuf_setlen(path, path->len - 1);
2432 } else
2433 prep_exclude(dir, istate, path->buf, path->len);
2435 /* hopefully prep_exclude() haven't invalidated this entry... */
2436 return untracked->valid;
2439 static int open_cached_dir(struct cached_dir *cdir,
2440 struct dir_struct *dir,
2441 struct untracked_cache_dir *untracked,
2442 struct index_state *istate,
2443 struct strbuf *path,
2444 int check_only)
2446 const char *c_path;
2448 memset(cdir, 0, sizeof(*cdir));
2449 cdir->untracked = untracked;
2450 if (valid_cached_dir(dir, untracked, istate, path, check_only))
2451 return 0;
2452 c_path = path->len ? path->buf : ".";
2453 cdir->fdir = opendir(c_path);
2454 if (!cdir->fdir)
2455 warning_errno(_("could not open directory '%s'"), c_path);
2456 if (dir->untracked) {
2457 invalidate_directory(dir->untracked, untracked);
2458 dir->untracked->dir_opened++;
2460 if (!cdir->fdir)
2461 return -1;
2462 return 0;
2465 static int read_cached_dir(struct cached_dir *cdir)
2467 struct dirent *de;
2469 if (cdir->fdir) {
2470 de = readdir_skip_dot_and_dotdot(cdir->fdir);
2471 if (!de) {
2472 cdir->d_name = NULL;
2473 cdir->d_type = DT_UNKNOWN;
2474 return -1;
2476 cdir->d_name = de->d_name;
2477 cdir->d_type = DTYPE(de);
2478 return 0;
2480 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
2481 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
2482 if (!d->recurse) {
2483 cdir->nr_dirs++;
2484 continue;
2486 cdir->ucd = d;
2487 cdir->nr_dirs++;
2488 return 0;
2490 cdir->ucd = NULL;
2491 if (cdir->nr_files < cdir->untracked->untracked_nr) {
2492 struct untracked_cache_dir *d = cdir->untracked;
2493 cdir->file = d->untracked[cdir->nr_files++];
2494 return 0;
2496 return -1;
2499 static void close_cached_dir(struct cached_dir *cdir)
2501 if (cdir->fdir)
2502 closedir(cdir->fdir);
2504 * We have gone through this directory and found no untracked
2505 * entries. Mark it valid.
2507 if (cdir->untracked) {
2508 cdir->untracked->valid = 1;
2509 cdir->untracked->recurse = 1;
2513 static void add_path_to_appropriate_result_list(struct dir_struct *dir,
2514 struct untracked_cache_dir *untracked,
2515 struct cached_dir *cdir,
2516 struct index_state *istate,
2517 struct strbuf *path,
2518 int baselen,
2519 const struct pathspec *pathspec,
2520 enum path_treatment state)
2522 /* add the path to the appropriate result list */
2523 switch (state) {
2524 case path_excluded:
2525 if (dir->flags & DIR_SHOW_IGNORED)
2526 dir_add_name(dir, istate, path->buf, path->len);
2527 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2528 ((dir->flags & DIR_COLLECT_IGNORED) &&
2529 exclude_matches_pathspec(path->buf, path->len,
2530 pathspec)))
2531 dir_add_ignored(dir, istate, path->buf, path->len);
2532 break;
2534 case path_untracked:
2535 if (dir->flags & DIR_SHOW_IGNORED)
2536 break;
2537 dir_add_name(dir, istate, path->buf, path->len);
2538 if (cdir->fdir)
2539 add_untracked(untracked, path->buf + baselen);
2540 break;
2542 default:
2543 break;
2548 * Read a directory tree. We currently ignore anything but
2549 * directories, regular files and symlinks. That's because git
2550 * doesn't handle them at all yet. Maybe that will change some
2551 * day.
2553 * Also, we ignore the name ".git" (even if it is not a directory).
2554 * That likely will not change.
2556 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2557 * to signal that a file was found. This is the least significant value that
2558 * indicates that a file was encountered that does not depend on the order of
2559 * whether an untracked or excluded path was encountered first.
2561 * Returns the most significant path_treatment value encountered in the scan.
2562 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2563 * significant path_treatment value that will be returned.
2566 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
2567 struct index_state *istate, const char *base, int baselen,
2568 struct untracked_cache_dir *untracked, int check_only,
2569 int stop_at_first_file, const struct pathspec *pathspec)
2572 * WARNING: Do NOT recurse unless path_recurse is returned from
2573 * treat_path(). Recursing on any other return value
2574 * can result in exponential slowdown.
2576 struct cached_dir cdir;
2577 enum path_treatment state, subdir_state, dir_state = path_none;
2578 struct strbuf path = STRBUF_INIT;
2580 strbuf_add(&path, base, baselen);
2582 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
2583 goto out;
2584 dir->internal.visited_directories++;
2586 if (untracked)
2587 untracked->check_only = !!check_only;
2589 while (!read_cached_dir(&cdir)) {
2590 /* check how the file or directory should be treated */
2591 state = treat_path(dir, untracked, &cdir, istate, &path,
2592 baselen, pathspec);
2593 dir->internal.visited_paths++;
2595 if (state > dir_state)
2596 dir_state = state;
2598 /* recurse into subdir if instructed by treat_path */
2599 if (state == path_recurse) {
2600 struct untracked_cache_dir *ud;
2601 ud = lookup_untracked(dir->untracked,
2602 untracked,
2603 path.buf + baselen,
2604 path.len - baselen);
2605 subdir_state =
2606 read_directory_recursive(dir, istate, path.buf,
2607 path.len, ud,
2608 check_only, stop_at_first_file, pathspec);
2609 if (subdir_state > dir_state)
2610 dir_state = subdir_state;
2612 if (pathspec &&
2613 !match_pathspec(istate, pathspec, path.buf, path.len,
2614 0 /* prefix */, NULL,
2615 0 /* do NOT special case dirs */))
2616 state = path_none;
2619 if (check_only) {
2620 if (stop_at_first_file) {
2622 * If stopping at first file, then
2623 * signal that a file was found by
2624 * returning `path_excluded`. This is
2625 * to return a consistent value
2626 * regardless of whether an ignored or
2627 * excluded file happened to be
2628 * encountered 1st.
2630 * In current usage, the
2631 * `stop_at_first_file` is passed when
2632 * an ancestor directory has matched
2633 * an exclude pattern, so any found
2634 * files will be excluded.
2636 if (dir_state >= path_excluded) {
2637 dir_state = path_excluded;
2638 break;
2642 /* abort early if maximum state has been reached */
2643 if (dir_state == path_untracked) {
2644 if (cdir.fdir)
2645 add_untracked(untracked, path.buf + baselen);
2646 break;
2648 /* skip the add_path_to_appropriate_result_list() */
2649 continue;
2652 add_path_to_appropriate_result_list(dir, untracked, &cdir,
2653 istate, &path, baselen,
2654 pathspec, state);
2656 close_cached_dir(&cdir);
2657 out:
2658 strbuf_release(&path);
2660 return dir_state;
2663 int cmp_dir_entry(const void *p1, const void *p2)
2665 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2666 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2668 return name_compare(e1->name, e1->len, e2->name, e2->len);
2671 /* check if *out lexically strictly contains *in */
2672 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2674 return (out->len < in->len) &&
2675 (out->name[out->len - 1] == '/') &&
2676 !memcmp(out->name, in->name, out->len);
2679 static int treat_leading_path(struct dir_struct *dir,
2680 struct index_state *istate,
2681 const char *path, int len,
2682 const struct pathspec *pathspec)
2684 struct strbuf sb = STRBUF_INIT;
2685 struct strbuf subdir = STRBUF_INIT;
2686 int prevlen, baselen;
2687 const char *cp;
2688 struct cached_dir cdir;
2689 enum path_treatment state = path_none;
2692 * For each directory component of path, we are going to check whether
2693 * that path is relevant given the pathspec. For example, if path is
2694 * foo/bar/baz/
2695 * then we will ask treat_path() whether we should go into foo, then
2696 * whether we should go into bar, then whether baz is relevant.
2697 * Checking each is important because e.g. if path is
2698 * .git/info/
2699 * then we need to check .git to know we shouldn't traverse it.
2700 * If the return from treat_path() is:
2701 * * path_none, for any path, we return false.
2702 * * path_recurse, for all path components, we return true
2703 * * <anything else> for some intermediate component, we make sure
2704 * to add that path to the relevant list but return false
2705 * signifying that we shouldn't recurse into it.
2708 while (len && path[len - 1] == '/')
2709 len--;
2710 if (!len)
2711 return 1;
2713 memset(&cdir, 0, sizeof(cdir));
2714 cdir.d_type = DT_DIR;
2715 baselen = 0;
2716 prevlen = 0;
2717 while (1) {
2718 prevlen = baselen + !!baselen;
2719 cp = path + prevlen;
2720 cp = memchr(cp, '/', path + len - cp);
2721 if (!cp)
2722 baselen = len;
2723 else
2724 baselen = cp - path;
2725 strbuf_reset(&sb);
2726 strbuf_add(&sb, path, baselen);
2727 if (!is_directory(sb.buf))
2728 break;
2729 strbuf_reset(&sb);
2730 strbuf_add(&sb, path, prevlen);
2731 strbuf_reset(&subdir);
2732 strbuf_add(&subdir, path+prevlen, baselen-prevlen);
2733 cdir.d_name = subdir.buf;
2734 state = treat_path(dir, NULL, &cdir, istate, &sb, prevlen, pathspec);
2736 if (state != path_recurse)
2737 break; /* do not recurse into it */
2738 if (len <= baselen)
2739 break; /* finished checking */
2741 add_path_to_appropriate_result_list(dir, NULL, &cdir, istate,
2742 &sb, baselen, pathspec,
2743 state);
2745 strbuf_release(&subdir);
2746 strbuf_release(&sb);
2747 return state == path_recurse;
2750 static const char *get_ident_string(void)
2752 static struct strbuf sb = STRBUF_INIT;
2753 struct utsname uts;
2755 if (sb.len)
2756 return sb.buf;
2757 if (uname(&uts) < 0)
2758 die_errno(_("failed to get kernel name and information"));
2759 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2760 uts.sysname);
2761 return sb.buf;
2764 static int ident_in_untracked(const struct untracked_cache *uc)
2767 * Previous git versions may have saved many NUL separated
2768 * strings in the "ident" field, but it is insane to manage
2769 * many locations, so just take care of the first one.
2772 return !strcmp(uc->ident.buf, get_ident_string());
2775 static void set_untracked_ident(struct untracked_cache *uc)
2777 strbuf_reset(&uc->ident);
2778 strbuf_addstr(&uc->ident, get_ident_string());
2781 * This strbuf used to contain a list of NUL separated
2782 * strings, so save NUL too for backward compatibility.
2784 strbuf_addch(&uc->ident, 0);
2787 static unsigned new_untracked_cache_flags(struct index_state *istate)
2789 struct repository *repo = istate->repo;
2790 char *val;
2793 * This logic is coordinated with the setting of these flags in
2794 * wt-status.c#wt_status_collect_untracked(), and the evaluation
2795 * of the config setting in commit.c#git_status_config()
2797 if (!repo_config_get_string(repo, "status.showuntrackedfiles", &val) &&
2798 !strcmp(val, "all"))
2799 return 0;
2802 * The default, if "all" is not set, is "normal" - leading us here.
2803 * If the value is "none" then it really doesn't matter.
2805 return DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2808 static void new_untracked_cache(struct index_state *istate, int flags)
2810 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2811 strbuf_init(&uc->ident, 100);
2812 uc->exclude_per_dir = ".gitignore";
2813 uc->dir_flags = flags >= 0 ? flags : new_untracked_cache_flags(istate);
2814 set_untracked_ident(uc);
2815 istate->untracked = uc;
2816 istate->cache_changed |= UNTRACKED_CHANGED;
2819 void add_untracked_cache(struct index_state *istate)
2821 if (!istate->untracked) {
2822 new_untracked_cache(istate, -1);
2823 } else {
2824 if (!ident_in_untracked(istate->untracked)) {
2825 free_untracked_cache(istate->untracked);
2826 new_untracked_cache(istate, -1);
2831 void remove_untracked_cache(struct index_state *istate)
2833 if (istate->untracked) {
2834 free_untracked_cache(istate->untracked);
2835 istate->untracked = NULL;
2836 istate->cache_changed |= UNTRACKED_CHANGED;
2840 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2841 int base_len,
2842 const struct pathspec *pathspec,
2843 struct index_state *istate)
2845 struct untracked_cache_dir *root;
2846 static int untracked_cache_disabled = -1;
2848 if (!dir->untracked)
2849 return NULL;
2850 if (untracked_cache_disabled < 0)
2851 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2852 if (untracked_cache_disabled)
2853 return NULL;
2856 * We only support $GIT_DIR/info/exclude and core.excludesfile
2857 * as the global ignore rule files. Any other additions
2858 * (e.g. from command line) invalidate the cache. This
2859 * condition also catches running setup_standard_excludes()
2860 * before setting dir->untracked!
2862 if (dir->internal.unmanaged_exclude_files)
2863 return NULL;
2866 * Optimize for the main use case only: whole-tree git
2867 * status. More work involved in treat_leading_path() if we
2868 * use cache on just a subset of the worktree. pathspec
2869 * support could make the matter even worse.
2871 if (base_len || (pathspec && pathspec->nr))
2872 return NULL;
2874 /* We don't support collecting ignore files */
2875 if (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2876 DIR_COLLECT_IGNORED))
2877 return NULL;
2880 * If we use .gitignore in the cache and now you change it to
2881 * .gitexclude, everything will go wrong.
2883 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2884 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2885 return NULL;
2888 * EXC_CMDL is not considered in the cache. If people set it,
2889 * skip the cache.
2891 if (dir->internal.exclude_list_group[EXC_CMDL].nr)
2892 return NULL;
2894 if (!ident_in_untracked(dir->untracked)) {
2895 warning(_("untracked cache is disabled on this system or location"));
2896 return NULL;
2900 * If the untracked structure we received does not have the same flags
2901 * as requested in this run, we're going to need to either discard the
2902 * existing structure (and potentially later recreate), or bypass the
2903 * untracked cache mechanism for this run.
2905 if (dir->flags != dir->untracked->dir_flags) {
2907 * If the untracked structure we received does not have the same flags
2908 * as configured, then we need to reset / create a new "untracked"
2909 * structure to match the new config.
2911 * Keeping the saved and used untracked cache consistent with the
2912 * configuration provides an opportunity for frequent users of
2913 * "git status -uall" to leverage the untracked cache by aligning their
2914 * configuration - setting "status.showuntrackedfiles" to "all" or
2915 * "normal" as appropriate.
2917 * Previously using -uall (or setting "status.showuntrackedfiles" to
2918 * "all") was incompatible with untracked cache and *consistently*
2919 * caused surprisingly bad performance (with fscache and fsmonitor
2920 * enabled) on Windows.
2922 * IMPROVEMENT OPPORTUNITY: If we reworked the untracked cache storage
2923 * to not be as bound up with the desired output in a given run,
2924 * and instead iterated through and stored enough information to
2925 * correctly serve both "modes", then users could get peak performance
2926 * with or without '-uall' regardless of their
2927 * "status.showuntrackedfiles" config.
2929 if (dir->untracked->dir_flags != new_untracked_cache_flags(istate)) {
2930 free_untracked_cache(istate->untracked);
2931 new_untracked_cache(istate, dir->flags);
2932 dir->untracked = istate->untracked;
2934 else {
2936 * Current untracked cache data is consistent with config, but not
2937 * usable in this request/run; just bypass untracked cache.
2939 return NULL;
2943 if (!dir->untracked->root) {
2944 /* Untracked cache existed but is not initialized; fix that */
2945 FLEX_ALLOC_STR(dir->untracked->root, name, "");
2946 istate->cache_changed |= UNTRACKED_CHANGED;
2949 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2950 root = dir->untracked->root;
2951 if (!oideq(&dir->internal.ss_info_exclude.oid,
2952 &dir->untracked->ss_info_exclude.oid)) {
2953 invalidate_gitignore(dir->untracked, root);
2954 dir->untracked->ss_info_exclude = dir->internal.ss_info_exclude;
2956 if (!oideq(&dir->internal.ss_excludes_file.oid,
2957 &dir->untracked->ss_excludes_file.oid)) {
2958 invalidate_gitignore(dir->untracked, root);
2959 dir->untracked->ss_excludes_file = dir->internal.ss_excludes_file;
2962 /* Make sure this directory is not dropped out at saving phase */
2963 root->recurse = 1;
2964 return root;
2967 static void emit_traversal_statistics(struct dir_struct *dir,
2968 struct repository *repo,
2969 const char *path,
2970 int path_len)
2972 if (!trace2_is_enabled())
2973 return;
2975 if (!path_len) {
2976 trace2_data_string("read_directory", repo, "path", "");
2977 } else {
2978 struct strbuf tmp = STRBUF_INIT;
2979 strbuf_add(&tmp, path, path_len);
2980 trace2_data_string("read_directory", repo, "path", tmp.buf);
2981 strbuf_release(&tmp);
2984 trace2_data_intmax("read_directory", repo,
2985 "directories-visited", dir->internal.visited_directories);
2986 trace2_data_intmax("read_directory", repo,
2987 "paths-visited", dir->internal.visited_paths);
2989 if (!dir->untracked)
2990 return;
2991 trace2_data_intmax("read_directory", repo,
2992 "node-creation", dir->untracked->dir_created);
2993 trace2_data_intmax("read_directory", repo,
2994 "gitignore-invalidation",
2995 dir->untracked->gitignore_invalidated);
2996 trace2_data_intmax("read_directory", repo,
2997 "directory-invalidation",
2998 dir->untracked->dir_invalidated);
2999 trace2_data_intmax("read_directory", repo,
3000 "opendir", dir->untracked->dir_opened);
3003 int read_directory(struct dir_struct *dir, struct index_state *istate,
3004 const char *path, int len, const struct pathspec *pathspec)
3006 struct untracked_cache_dir *untracked;
3008 trace2_region_enter("dir", "read_directory", istate->repo);
3009 dir->internal.visited_paths = 0;
3010 dir->internal.visited_directories = 0;
3012 if (has_symlink_leading_path(path, len)) {
3013 trace2_region_leave("dir", "read_directory", istate->repo);
3014 return dir->nr;
3017 untracked = validate_untracked_cache(dir, len, pathspec, istate);
3018 if (!untracked)
3020 * make sure untracked cache code path is disabled,
3021 * e.g. prep_exclude()
3023 dir->untracked = NULL;
3024 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
3025 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
3026 QSORT(dir->entries, dir->nr, cmp_dir_entry);
3027 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
3029 emit_traversal_statistics(dir, istate->repo, path, len);
3031 trace2_region_leave("dir", "read_directory", istate->repo);
3032 if (dir->untracked) {
3033 static int force_untracked_cache = -1;
3035 if (force_untracked_cache < 0)
3036 force_untracked_cache =
3037 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", -1);
3038 if (force_untracked_cache < 0)
3039 force_untracked_cache = (istate->repo->settings.core_untracked_cache == UNTRACKED_CACHE_WRITE);
3040 if (force_untracked_cache &&
3041 dir->untracked == istate->untracked &&
3042 (dir->untracked->dir_opened ||
3043 dir->untracked->gitignore_invalidated ||
3044 dir->untracked->dir_invalidated))
3045 istate->cache_changed |= UNTRACKED_CHANGED;
3046 if (dir->untracked != istate->untracked) {
3047 FREE_AND_NULL(dir->untracked);
3051 return dir->nr;
3054 int file_exists(const char *f)
3056 struct stat sb;
3057 return lstat(f, &sb) == 0;
3060 int repo_file_exists(struct repository *repo, const char *path)
3062 if (repo != the_repository)
3063 BUG("do not know how to check file existence in arbitrary repo");
3065 return file_exists(path);
3068 static int cmp_icase(char a, char b)
3070 if (a == b)
3071 return 0;
3072 if (ignore_case)
3073 return toupper(a) - toupper(b);
3074 return a - b;
3078 * Given two normalized paths (a trailing slash is ok), if subdir is
3079 * outside dir, return -1. Otherwise return the offset in subdir that
3080 * can be used as relative path to dir.
3082 int dir_inside_of(const char *subdir, const char *dir)
3084 int offset = 0;
3086 assert(dir && subdir && *dir && *subdir);
3088 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
3089 dir++;
3090 subdir++;
3091 offset++;
3094 /* hel[p]/me vs hel[l]/yeah */
3095 if (*dir && *subdir)
3096 return -1;
3098 if (!*subdir)
3099 return !*dir ? offset : -1; /* same dir */
3101 /* foo/[b]ar vs foo/[] */
3102 if (is_dir_sep(dir[-1]))
3103 return is_dir_sep(subdir[-1]) ? offset : -1;
3105 /* foo[/]bar vs foo[] */
3106 return is_dir_sep(*subdir) ? offset + 1 : -1;
3109 int is_inside_dir(const char *dir)
3111 char *cwd;
3112 int rc;
3114 if (!dir)
3115 return 0;
3117 cwd = xgetcwd();
3118 rc = (dir_inside_of(cwd, dir) >= 0);
3119 free(cwd);
3120 return rc;
3123 int is_empty_dir(const char *path)
3125 DIR *dir = opendir(path);
3126 struct dirent *e;
3127 int ret = 1;
3129 if (!dir)
3130 return 0;
3132 e = readdir_skip_dot_and_dotdot(dir);
3133 if (e)
3134 ret = 0;
3136 closedir(dir);
3137 return ret;
3140 char *git_url_basename(const char *repo, int is_bundle, int is_bare)
3142 const char *end = repo + strlen(repo), *start, *ptr;
3143 size_t len;
3144 char *dir;
3147 * Skip scheme.
3149 start = strstr(repo, "://");
3150 if (!start)
3151 start = repo;
3152 else
3153 start += 3;
3156 * Skip authentication data. The stripping does happen
3157 * greedily, such that we strip up to the last '@' inside
3158 * the host part.
3160 for (ptr = start; ptr < end && !is_dir_sep(*ptr); ptr++) {
3161 if (*ptr == '@')
3162 start = ptr + 1;
3166 * Strip trailing spaces, slashes and /.git
3168 while (start < end && (is_dir_sep(end[-1]) || isspace(end[-1])))
3169 end--;
3170 if (end - start > 5 && is_dir_sep(end[-5]) &&
3171 !strncmp(end - 4, ".git", 4)) {
3172 end -= 5;
3173 while (start < end && is_dir_sep(end[-1]))
3174 end--;
3178 * It should not be possible to overflow `ptrdiff_t` by passing in an
3179 * insanely long URL, but GCC does not know that and will complain
3180 * without this check.
3182 if (end - start < 0)
3183 die(_("No directory name could be guessed.\n"
3184 "Please specify a directory on the command line"));
3187 * Strip trailing port number if we've got only a
3188 * hostname (that is, there is no dir separator but a
3189 * colon). This check is required such that we do not
3190 * strip URI's like '/foo/bar:2222.git', which should
3191 * result in a dir '2222' being guessed due to backwards
3192 * compatibility.
3194 if (memchr(start, '/', end - start) == NULL
3195 && memchr(start, ':', end - start) != NULL) {
3196 ptr = end;
3197 while (start < ptr && isdigit(ptr[-1]) && ptr[-1] != ':')
3198 ptr--;
3199 if (start < ptr && ptr[-1] == ':')
3200 end = ptr - 1;
3204 * Find last component. To remain backwards compatible we
3205 * also regard colons as path separators, such that
3206 * cloning a repository 'foo:bar.git' would result in a
3207 * directory 'bar' being guessed.
3209 ptr = end;
3210 while (start < ptr && !is_dir_sep(ptr[-1]) && ptr[-1] != ':')
3211 ptr--;
3212 start = ptr;
3215 * Strip .{bundle,git}.
3217 len = end - start;
3218 strip_suffix_mem(start, &len, is_bundle ? ".bundle" : ".git");
3220 if (!len || (len == 1 && *start == '/'))
3221 die(_("No directory name could be guessed.\n"
3222 "Please specify a directory on the command line"));
3224 if (is_bare)
3225 dir = xstrfmt("%.*s.git", (int)len, start);
3226 else
3227 dir = xstrndup(start, len);
3229 * Replace sequences of 'control' characters and whitespace
3230 * with one ascii space, remove leading and trailing spaces.
3232 if (*dir) {
3233 char *out = dir;
3234 int prev_space = 1 /* strip leading whitespace */;
3235 for (end = dir; *end; ++end) {
3236 char ch = *end;
3237 if ((unsigned char)ch < '\x20')
3238 ch = '\x20';
3239 if (isspace(ch)) {
3240 if (prev_space)
3241 continue;
3242 prev_space = 1;
3243 } else
3244 prev_space = 0;
3245 *out++ = ch;
3247 *out = '\0';
3248 if (out > dir && prev_space)
3249 out[-1] = '\0';
3251 return dir;
3254 void strip_dir_trailing_slashes(char *dir)
3256 char *end = dir + strlen(dir);
3258 while (dir < end - 1 && is_dir_sep(end[-1]))
3259 end--;
3260 *end = '\0';
3263 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
3265 DIR *dir;
3266 struct dirent *e;
3267 int ret = 0, original_len = path->len, len, kept_down = 0;
3268 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
3269 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
3270 int purge_original_cwd = (flag & REMOVE_DIR_PURGE_ORIGINAL_CWD);
3271 struct object_id submodule_head;
3273 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
3274 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
3275 /* Do not descend and nuke a nested git work tree. */
3276 if (kept_up)
3277 *kept_up = 1;
3278 return 0;
3281 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
3282 dir = opendir(path->buf);
3283 if (!dir) {
3284 if (errno == ENOENT)
3285 return keep_toplevel ? -1 : 0;
3286 else if (errno == EACCES && !keep_toplevel)
3288 * An empty dir could be removable even if it
3289 * is unreadable:
3291 return rmdir(path->buf);
3292 else
3293 return -1;
3295 strbuf_complete(path, '/');
3297 len = path->len;
3298 while ((e = readdir_skip_dot_and_dotdot(dir)) != NULL) {
3299 struct stat st;
3301 strbuf_setlen(path, len);
3302 strbuf_addstr(path, e->d_name);
3303 if (lstat(path->buf, &st)) {
3304 if (errno == ENOENT)
3306 * file disappeared, which is what we
3307 * wanted anyway
3309 continue;
3310 /* fall through */
3311 } else if (S_ISDIR(st.st_mode)) {
3312 if (!remove_dir_recurse(path, flag, &kept_down))
3313 continue; /* happy */
3314 } else if (!only_empty &&
3315 (!unlink(path->buf) || errno == ENOENT)) {
3316 continue; /* happy, too */
3319 /* path too long, stat fails, or non-directory still exists */
3320 ret = -1;
3321 break;
3323 closedir(dir);
3325 strbuf_setlen(path, original_len);
3326 if (!ret && !keep_toplevel && !kept_down) {
3327 if (!purge_original_cwd &&
3328 startup_info->original_cwd &&
3329 !strcmp(startup_info->original_cwd, path->buf))
3330 ret = -1; /* Do not remove current working directory */
3331 else
3332 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
3333 } else if (kept_up)
3335 * report the uplevel that it is not an error that we
3336 * did not rmdir() our directory.
3338 *kept_up = !ret;
3339 return ret;
3342 int remove_dir_recursively(struct strbuf *path, int flag)
3344 return remove_dir_recurse(path, flag, NULL);
3347 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
3349 void setup_standard_excludes(struct dir_struct *dir)
3351 dir->exclude_per_dir = ".gitignore";
3353 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3354 if (!excludes_file)
3355 excludes_file = xdg_config_home("ignore");
3356 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
3357 add_patterns_from_file_1(dir, excludes_file,
3358 dir->untracked ? &dir->internal.ss_excludes_file : NULL);
3360 /* per repository user preference */
3361 if (startup_info->have_repository) {
3362 const char *path = git_path_info_exclude();
3363 if (!access_or_warn(path, R_OK, 0))
3364 add_patterns_from_file_1(dir, path,
3365 dir->untracked ? &dir->internal.ss_info_exclude : NULL);
3369 char *get_sparse_checkout_filename(void)
3371 return git_pathdup("info/sparse-checkout");
3374 int get_sparse_checkout_patterns(struct pattern_list *pl)
3376 int res;
3377 char *sparse_filename = get_sparse_checkout_filename();
3379 pl->use_cone_patterns = core_sparse_checkout_cone;
3380 res = add_patterns_from_file_to_list(sparse_filename, "", 0, pl, NULL, 0);
3382 free(sparse_filename);
3383 return res;
3386 int remove_path(const char *name)
3388 char *slash;
3390 if (unlink(name) && !is_missing_file_error(errno))
3391 return -1;
3393 slash = strrchr(name, '/');
3394 if (slash) {
3395 char *dirs = xstrdup(name);
3396 slash = dirs + (slash - name);
3397 do {
3398 *slash = '\0';
3399 if (startup_info->original_cwd &&
3400 !strcmp(startup_info->original_cwd, dirs))
3401 break;
3402 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
3403 free(dirs);
3405 return 0;
3409 * Frees memory within dir which was allocated, and resets fields for further
3410 * use. Does not free dir itself.
3412 void dir_clear(struct dir_struct *dir)
3414 int i, j;
3415 struct exclude_list_group *group;
3416 struct pattern_list *pl;
3417 struct exclude_stack *stk;
3418 struct dir_struct new = DIR_INIT;
3420 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
3421 group = &dir->internal.exclude_list_group[i];
3422 for (j = 0; j < group->nr; j++) {
3423 pl = &group->pl[j];
3424 if (i == EXC_DIRS)
3425 free((char *)pl->src);
3426 clear_pattern_list(pl);
3428 free(group->pl);
3431 for (i = 0; i < dir->ignored_nr; i++)
3432 free(dir->ignored[i]);
3433 for (i = 0; i < dir->nr; i++)
3434 free(dir->entries[i]);
3435 free(dir->ignored);
3436 free(dir->entries);
3438 stk = dir->internal.exclude_stack;
3439 while (stk) {
3440 struct exclude_stack *prev = stk->prev;
3441 free(stk);
3442 stk = prev;
3444 strbuf_release(&dir->internal.basebuf);
3446 memcpy(dir, &new, sizeof(*dir));
3449 struct ondisk_untracked_cache {
3450 struct stat_data info_exclude_stat;
3451 struct stat_data excludes_file_stat;
3452 uint32_t dir_flags;
3455 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3457 struct write_data {
3458 int index; /* number of written untracked_cache_dir */
3459 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
3460 struct ewah_bitmap *valid; /* from untracked_cache_dir */
3461 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
3462 struct strbuf out;
3463 struct strbuf sb_stat;
3464 struct strbuf sb_sha1;
3467 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
3469 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
3470 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
3471 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
3472 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
3473 to->sd_dev = htonl(from->sd_dev);
3474 to->sd_ino = htonl(from->sd_ino);
3475 to->sd_uid = htonl(from->sd_uid);
3476 to->sd_gid = htonl(from->sd_gid);
3477 to->sd_size = htonl(from->sd_size);
3480 static void write_one_dir(struct untracked_cache_dir *untracked,
3481 struct write_data *wd)
3483 struct stat_data stat_data;
3484 struct strbuf *out = &wd->out;
3485 unsigned char intbuf[16];
3486 unsigned int intlen, value;
3487 int i = wd->index++;
3490 * untracked_nr should be reset whenever valid is clear, but
3491 * for safety..
3493 if (!untracked->valid) {
3494 untracked->untracked_nr = 0;
3495 untracked->check_only = 0;
3498 if (untracked->check_only)
3499 ewah_set(wd->check_only, i);
3500 if (untracked->valid) {
3501 ewah_set(wd->valid, i);
3502 stat_data_to_disk(&stat_data, &untracked->stat_data);
3503 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
3505 if (!is_null_oid(&untracked->exclude_oid)) {
3506 ewah_set(wd->sha1_valid, i);
3507 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
3508 the_hash_algo->rawsz);
3511 intlen = encode_varint(untracked->untracked_nr, intbuf);
3512 strbuf_add(out, intbuf, intlen);
3514 /* skip non-recurse directories */
3515 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
3516 if (untracked->dirs[i]->recurse)
3517 value++;
3518 intlen = encode_varint(value, intbuf);
3519 strbuf_add(out, intbuf, intlen);
3521 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
3523 for (i = 0; i < untracked->untracked_nr; i++)
3524 strbuf_add(out, untracked->untracked[i],
3525 strlen(untracked->untracked[i]) + 1);
3527 for (i = 0; i < untracked->dirs_nr; i++)
3528 if (untracked->dirs[i]->recurse)
3529 write_one_dir(untracked->dirs[i], wd);
3532 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
3534 struct ondisk_untracked_cache *ouc;
3535 struct write_data wd;
3536 unsigned char varbuf[16];
3537 int varint_len;
3538 const unsigned hashsz = the_hash_algo->rawsz;
3540 CALLOC_ARRAY(ouc, 1);
3541 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
3542 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
3543 ouc->dir_flags = htonl(untracked->dir_flags);
3545 varint_len = encode_varint(untracked->ident.len, varbuf);
3546 strbuf_add(out, varbuf, varint_len);
3547 strbuf_addbuf(out, &untracked->ident);
3549 strbuf_add(out, ouc, sizeof(*ouc));
3550 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
3551 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
3552 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
3553 FREE_AND_NULL(ouc);
3555 if (!untracked->root) {
3556 varint_len = encode_varint(0, varbuf);
3557 strbuf_add(out, varbuf, varint_len);
3558 return;
3561 wd.index = 0;
3562 wd.check_only = ewah_new();
3563 wd.valid = ewah_new();
3564 wd.sha1_valid = ewah_new();
3565 strbuf_init(&wd.out, 1024);
3566 strbuf_init(&wd.sb_stat, 1024);
3567 strbuf_init(&wd.sb_sha1, 1024);
3568 write_one_dir(untracked->root, &wd);
3570 varint_len = encode_varint(wd.index, varbuf);
3571 strbuf_add(out, varbuf, varint_len);
3572 strbuf_addbuf(out, &wd.out);
3573 ewah_serialize_strbuf(wd.valid, out);
3574 ewah_serialize_strbuf(wd.check_only, out);
3575 ewah_serialize_strbuf(wd.sha1_valid, out);
3576 strbuf_addbuf(out, &wd.sb_stat);
3577 strbuf_addbuf(out, &wd.sb_sha1);
3578 strbuf_addch(out, '\0'); /* safe guard for string lists */
3580 ewah_free(wd.valid);
3581 ewah_free(wd.check_only);
3582 ewah_free(wd.sha1_valid);
3583 strbuf_release(&wd.out);
3584 strbuf_release(&wd.sb_stat);
3585 strbuf_release(&wd.sb_sha1);
3588 static void free_untracked(struct untracked_cache_dir *ucd)
3590 int i;
3591 if (!ucd)
3592 return;
3593 for (i = 0; i < ucd->dirs_nr; i++)
3594 free_untracked(ucd->dirs[i]);
3595 for (i = 0; i < ucd->untracked_nr; i++)
3596 free(ucd->untracked[i]);
3597 free(ucd->untracked);
3598 free(ucd->dirs);
3599 free(ucd);
3602 void free_untracked_cache(struct untracked_cache *uc)
3604 if (!uc)
3605 return;
3607 free(uc->exclude_per_dir_to_free);
3608 strbuf_release(&uc->ident);
3609 free_untracked(uc->root);
3610 free(uc);
3613 struct read_data {
3614 int index;
3615 struct untracked_cache_dir **ucd;
3616 struct ewah_bitmap *check_only;
3617 struct ewah_bitmap *valid;
3618 struct ewah_bitmap *sha1_valid;
3619 const unsigned char *data;
3620 const unsigned char *end;
3623 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
3625 memcpy(to, data, sizeof(*to));
3626 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
3627 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
3628 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
3629 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
3630 to->sd_dev = ntohl(to->sd_dev);
3631 to->sd_ino = ntohl(to->sd_ino);
3632 to->sd_uid = ntohl(to->sd_uid);
3633 to->sd_gid = ntohl(to->sd_gid);
3634 to->sd_size = ntohl(to->sd_size);
3637 static int read_one_dir(struct untracked_cache_dir **untracked_,
3638 struct read_data *rd)
3640 struct untracked_cache_dir ud, *untracked;
3641 const unsigned char *data = rd->data, *end = rd->end;
3642 const unsigned char *eos;
3643 unsigned int value;
3644 int i;
3646 memset(&ud, 0, sizeof(ud));
3648 value = decode_varint(&data);
3649 if (data > end)
3650 return -1;
3651 ud.recurse = 1;
3652 ud.untracked_alloc = value;
3653 ud.untracked_nr = value;
3654 if (ud.untracked_nr)
3655 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
3657 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
3658 if (data > end)
3659 return -1;
3660 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
3662 eos = memchr(data, '\0', end - data);
3663 if (!eos || eos == end)
3664 return -1;
3666 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
3667 memcpy(untracked, &ud, sizeof(ud));
3668 memcpy(untracked->name, data, eos - data + 1);
3669 data = eos + 1;
3671 for (i = 0; i < untracked->untracked_nr; i++) {
3672 eos = memchr(data, '\0', end - data);
3673 if (!eos || eos == end)
3674 return -1;
3675 untracked->untracked[i] = xmemdupz(data, eos - data);
3676 data = eos + 1;
3679 rd->ucd[rd->index++] = untracked;
3680 rd->data = data;
3682 for (i = 0; i < untracked->dirs_nr; i++) {
3683 if (read_one_dir(untracked->dirs + i, rd) < 0)
3684 return -1;
3686 return 0;
3689 static void set_check_only(size_t pos, void *cb)
3691 struct read_data *rd = cb;
3692 struct untracked_cache_dir *ud = rd->ucd[pos];
3693 ud->check_only = 1;
3696 static void read_stat(size_t pos, void *cb)
3698 struct read_data *rd = cb;
3699 struct untracked_cache_dir *ud = rd->ucd[pos];
3700 if (rd->data + sizeof(struct stat_data) > rd->end) {
3701 rd->data = rd->end + 1;
3702 return;
3704 stat_data_from_disk(&ud->stat_data, rd->data);
3705 rd->data += sizeof(struct stat_data);
3706 ud->valid = 1;
3709 static void read_oid(size_t pos, void *cb)
3711 struct read_data *rd = cb;
3712 struct untracked_cache_dir *ud = rd->ucd[pos];
3713 if (rd->data + the_hash_algo->rawsz > rd->end) {
3714 rd->data = rd->end + 1;
3715 return;
3717 oidread(&ud->exclude_oid, rd->data);
3718 rd->data += the_hash_algo->rawsz;
3721 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
3722 const unsigned char *sha1)
3724 stat_data_from_disk(&oid_stat->stat, data);
3725 oidread(&oid_stat->oid, sha1);
3726 oid_stat->valid = 1;
3729 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
3731 struct untracked_cache *uc;
3732 struct read_data rd;
3733 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
3734 const char *ident;
3735 int ident_len;
3736 ssize_t len;
3737 const char *exclude_per_dir;
3738 const unsigned hashsz = the_hash_algo->rawsz;
3739 const unsigned offset = sizeof(struct ondisk_untracked_cache);
3740 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
3742 if (sz <= 1 || end[-1] != '\0')
3743 return NULL;
3744 end--;
3746 ident_len = decode_varint(&next);
3747 if (next + ident_len > end)
3748 return NULL;
3749 ident = (const char *)next;
3750 next += ident_len;
3752 if (next + exclude_per_dir_offset + 1 > end)
3753 return NULL;
3755 CALLOC_ARRAY(uc, 1);
3756 strbuf_init(&uc->ident, ident_len);
3757 strbuf_add(&uc->ident, ident, ident_len);
3758 load_oid_stat(&uc->ss_info_exclude,
3759 next + ouc_offset(info_exclude_stat),
3760 next + offset);
3761 load_oid_stat(&uc->ss_excludes_file,
3762 next + ouc_offset(excludes_file_stat),
3763 next + offset + hashsz);
3764 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
3765 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
3766 uc->exclude_per_dir = uc->exclude_per_dir_to_free = xstrdup(exclude_per_dir);
3767 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3768 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
3769 if (next >= end)
3770 goto done2;
3772 len = decode_varint(&next);
3773 if (next > end || len == 0)
3774 goto done2;
3776 rd.valid = ewah_new();
3777 rd.check_only = ewah_new();
3778 rd.sha1_valid = ewah_new();
3779 rd.data = next;
3780 rd.end = end;
3781 rd.index = 0;
3782 ALLOC_ARRAY(rd.ucd, len);
3784 if (read_one_dir(&uc->root, &rd) || rd.index != len)
3785 goto done;
3787 next = rd.data;
3788 len = ewah_read_mmap(rd.valid, next, end - next);
3789 if (len < 0)
3790 goto done;
3792 next += len;
3793 len = ewah_read_mmap(rd.check_only, next, end - next);
3794 if (len < 0)
3795 goto done;
3797 next += len;
3798 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
3799 if (len < 0)
3800 goto done;
3802 ewah_each_bit(rd.check_only, set_check_only, &rd);
3803 rd.data = next + len;
3804 ewah_each_bit(rd.valid, read_stat, &rd);
3805 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
3806 next = rd.data;
3808 done:
3809 free(rd.ucd);
3810 ewah_free(rd.valid);
3811 ewah_free(rd.check_only);
3812 ewah_free(rd.sha1_valid);
3813 done2:
3814 if (next != end) {
3815 free_untracked_cache(uc);
3816 uc = NULL;
3818 return uc;
3821 static void invalidate_one_directory(struct untracked_cache *uc,
3822 struct untracked_cache_dir *ucd)
3824 uc->dir_invalidated++;
3825 ucd->valid = 0;
3826 ucd->untracked_nr = 0;
3830 * Normally when an entry is added or removed from a directory,
3831 * invalidating that directory is enough. No need to touch its
3832 * ancestors. When a directory is shown as "foo/bar/" in git-status
3833 * however, deleting or adding an entry may have cascading effect.
3835 * Say the "foo/bar/file" has become untracked, we need to tell the
3836 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3837 * directory any more (because "bar" is managed by foo as an untracked
3838 * "file").
3840 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3841 * was the last untracked entry in the entire "foo", we should show
3842 * "foo/" instead. Which means we have to invalidate past "bar" up to
3843 * "foo".
3845 * This function traverses all directories from root to leaf. If there
3846 * is a chance of one of the above cases happening, we invalidate back
3847 * to root. Otherwise we just invalidate the leaf. There may be a more
3848 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3849 * detect these cases and avoid unnecessary invalidation, for example,
3850 * checking for the untracked entry named "bar/" in "foo", but for now
3851 * stick to something safe and simple.
3853 static int invalidate_one_component(struct untracked_cache *uc,
3854 struct untracked_cache_dir *dir,
3855 const char *path, int len)
3857 const char *rest = strchr(path, '/');
3859 if (rest) {
3860 int component_len = rest - path;
3861 struct untracked_cache_dir *d =
3862 lookup_untracked(uc, dir, path, component_len);
3863 int ret =
3864 invalidate_one_component(uc, d, rest + 1,
3865 len - (component_len + 1));
3866 if (ret)
3867 invalidate_one_directory(uc, dir);
3868 return ret;
3871 invalidate_one_directory(uc, dir);
3872 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
3875 void untracked_cache_invalidate_path(struct index_state *istate,
3876 const char *path, int safe_path)
3878 if (!istate->untracked || !istate->untracked->root)
3879 return;
3880 if (!safe_path && !verify_path(path, 0))
3881 return;
3882 invalidate_one_component(istate->untracked, istate->untracked->root,
3883 path, strlen(path));
3886 void untracked_cache_remove_from_index(struct index_state *istate,
3887 const char *path)
3889 untracked_cache_invalidate_path(istate, path, 1);
3892 void untracked_cache_add_to_index(struct index_state *istate,
3893 const char *path)
3895 untracked_cache_invalidate_path(istate, path, 1);
3898 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3899 const char *sub_gitdir)
3901 int i;
3902 struct repository subrepo;
3903 struct strbuf sub_wt = STRBUF_INIT;
3904 struct strbuf sub_gd = STRBUF_INIT;
3906 const struct submodule *sub;
3908 /* If the submodule has no working tree, we can ignore it. */
3909 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3910 return;
3912 if (repo_read_index(&subrepo) < 0)
3913 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3915 /* TODO: audit for interaction with sparse-index. */
3916 ensure_full_index(subrepo.index);
3917 for (i = 0; i < subrepo.index->cache_nr; i++) {
3918 const struct cache_entry *ce = subrepo.index->cache[i];
3920 if (!S_ISGITLINK(ce->ce_mode))
3921 continue;
3923 while (i + 1 < subrepo.index->cache_nr &&
3924 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3926 * Skip entries with the same name in different stages
3927 * to make sure an entry is returned only once.
3929 i++;
3931 sub = submodule_from_path(&subrepo, null_oid(), ce->name);
3932 if (!sub || !is_submodule_active(&subrepo, ce->name))
3933 /* .gitmodules broken or inactive sub */
3934 continue;
3936 strbuf_reset(&sub_wt);
3937 strbuf_reset(&sub_gd);
3938 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3939 submodule_name_to_gitdir(&sub_gd, &subrepo, sub->name);
3941 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3943 strbuf_release(&sub_wt);
3944 strbuf_release(&sub_gd);
3945 repo_clear(&subrepo);
3948 void connect_work_tree_and_git_dir(const char *work_tree_,
3949 const char *git_dir_,
3950 int recurse_into_nested)
3952 struct strbuf gitfile_sb = STRBUF_INIT;
3953 struct strbuf cfg_sb = STRBUF_INIT;
3954 struct strbuf rel_path = STRBUF_INIT;
3955 char *git_dir, *work_tree;
3957 /* Prepare .git file */
3958 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3959 if (safe_create_leading_directories_const(gitfile_sb.buf))
3960 die(_("could not create directories for %s"), gitfile_sb.buf);
3962 /* Prepare config file */
3963 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3964 if (safe_create_leading_directories_const(cfg_sb.buf))
3965 die(_("could not create directories for %s"), cfg_sb.buf);
3967 git_dir = real_pathdup(git_dir_, 1);
3968 work_tree = real_pathdup(work_tree_, 1);
3970 /* Write .git file */
3971 write_file(gitfile_sb.buf, "gitdir: %s",
3972 relative_path(git_dir, work_tree, &rel_path));
3973 /* Update core.worktree setting */
3974 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3975 relative_path(work_tree, git_dir, &rel_path));
3977 strbuf_release(&gitfile_sb);
3978 strbuf_release(&cfg_sb);
3979 strbuf_release(&rel_path);
3981 if (recurse_into_nested)
3982 connect_wt_gitdir_in_nested(work_tree, git_dir);
3984 free(work_tree);
3985 free(git_dir);
3989 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3991 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3993 if (rename(old_git_dir, new_git_dir) < 0)
3994 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3995 old_git_dir, new_git_dir);
3997 connect_work_tree_and_git_dir(path, new_git_dir, 0);
4000 int path_match_flags(const char *const str, const enum path_match_flags flags)
4002 const char *p = str;
4004 if (flags & PATH_MATCH_NATIVE &&
4005 flags & PATH_MATCH_XPLATFORM)
4006 BUG("path_match_flags() must get one match kind, not multiple!");
4007 else if (!(flags & PATH_MATCH_KINDS_MASK))
4008 BUG("path_match_flags() must get at least one match kind!");
4010 if (flags & PATH_MATCH_STARTS_WITH_DOT_SLASH &&
4011 flags & PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH)
4012 BUG("path_match_flags() must get one platform kind, not multiple!");
4013 else if (!(flags & PATH_MATCH_PLATFORM_MASK))
4014 BUG("path_match_flags() must get at least one platform kind!");
4016 if (*p++ != '.')
4017 return 0;
4018 if (flags & PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH &&
4019 *p++ != '.')
4020 return 0;
4022 if (flags & PATH_MATCH_NATIVE)
4023 return is_dir_sep(*p);
4024 else if (flags & PATH_MATCH_XPLATFORM)
4025 return is_xplatform_dir_sep(*p);
4026 BUG("unreachable");