Merge branch 'ab/commit-m-with-fixup' into maint
[git/debian.git] / dir.c
blob7c4b45e30e0ac87527ff0ef3fd8c90670a1e2064
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * See Documentation/technical/api-directory-listing.txt
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10 #define NO_THE_INDEX_COMPATIBILITY_MACROS
11 #include "cache.h"
12 #include "config.h"
13 #include "dir.h"
14 #include "attr.h"
15 #include "refs.h"
16 #include "wildmatch.h"
17 #include "pathspec.h"
18 #include "utf8.h"
19 #include "varint.h"
20 #include "ewah/ewok.h"
21 #include "fsmonitor.h"
24 * Tells read_directory_recursive how a file or directory should be treated.
25 * Values are ordered by significance, e.g. if a directory contains both
26 * excluded and untracked files, it is listed as untracked because
27 * path_untracked > path_excluded.
29 enum path_treatment {
30 path_none = 0,
31 path_recurse,
32 path_excluded,
33 path_untracked
37 * Support data structure for our opendir/readdir/closedir wrappers
39 struct cached_dir {
40 DIR *fdir;
41 struct untracked_cache_dir *untracked;
42 int nr_files;
43 int nr_dirs;
45 struct dirent *de;
46 const char *file;
47 struct untracked_cache_dir *ucd;
50 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
51 struct index_state *istate, const char *path, int len,
52 struct untracked_cache_dir *untracked,
53 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
54 static int get_dtype(struct dirent *de, struct index_state *istate,
55 const char *path, int len);
57 int count_slashes(const char *s)
59 int cnt = 0;
60 while (*s)
61 if (*s++ == '/')
62 cnt++;
63 return cnt;
66 int fspathcmp(const char *a, const char *b)
68 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
71 int fspathncmp(const char *a, const char *b, size_t count)
73 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
76 int git_fnmatch(const struct pathspec_item *item,
77 const char *pattern, const char *string,
78 int prefix)
80 if (prefix > 0) {
81 if (ps_strncmp(item, pattern, string, prefix))
82 return WM_NOMATCH;
83 pattern += prefix;
84 string += prefix;
86 if (item->flags & PATHSPEC_ONESTAR) {
87 int pattern_len = strlen(++pattern);
88 int string_len = strlen(string);
89 return string_len < pattern_len ||
90 ps_strcmp(item, pattern,
91 string + string_len - pattern_len);
93 if (item->magic & PATHSPEC_GLOB)
94 return wildmatch(pattern, string,
95 WM_PATHNAME |
96 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
97 else
98 /* wildmatch has not learned no FNM_PATHNAME mode yet */
99 return wildmatch(pattern, string,
100 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
103 static int fnmatch_icase_mem(const char *pattern, int patternlen,
104 const char *string, int stringlen,
105 int flags)
107 int match_status;
108 struct strbuf pat_buf = STRBUF_INIT;
109 struct strbuf str_buf = STRBUF_INIT;
110 const char *use_pat = pattern;
111 const char *use_str = string;
113 if (pattern[patternlen]) {
114 strbuf_add(&pat_buf, pattern, patternlen);
115 use_pat = pat_buf.buf;
117 if (string[stringlen]) {
118 strbuf_add(&str_buf, string, stringlen);
119 use_str = str_buf.buf;
122 if (ignore_case)
123 flags |= WM_CASEFOLD;
124 match_status = wildmatch(use_pat, use_str, flags);
126 strbuf_release(&pat_buf);
127 strbuf_release(&str_buf);
129 return match_status;
132 static size_t common_prefix_len(const struct pathspec *pathspec)
134 int n;
135 size_t max = 0;
138 * ":(icase)path" is treated as a pathspec full of
139 * wildcard. In other words, only prefix is considered common
140 * prefix. If the pathspec is abc/foo abc/bar, running in
141 * subdir xyz, the common prefix is still xyz, not xuz/abc as
142 * in non-:(icase).
144 GUARD_PATHSPEC(pathspec,
145 PATHSPEC_FROMTOP |
146 PATHSPEC_MAXDEPTH |
147 PATHSPEC_LITERAL |
148 PATHSPEC_GLOB |
149 PATHSPEC_ICASE |
150 PATHSPEC_EXCLUDE |
151 PATHSPEC_ATTR);
153 for (n = 0; n < pathspec->nr; n++) {
154 size_t i = 0, len = 0, item_len;
155 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
156 continue;
157 if (pathspec->items[n].magic & PATHSPEC_ICASE)
158 item_len = pathspec->items[n].prefix;
159 else
160 item_len = pathspec->items[n].nowildcard_len;
161 while (i < item_len && (n == 0 || i < max)) {
162 char c = pathspec->items[n].match[i];
163 if (c != pathspec->items[0].match[i])
164 break;
165 if (c == '/')
166 len = i + 1;
167 i++;
169 if (n == 0 || len < max) {
170 max = len;
171 if (!max)
172 break;
175 return max;
179 * Returns a copy of the longest leading path common among all
180 * pathspecs.
182 char *common_prefix(const struct pathspec *pathspec)
184 unsigned long len = common_prefix_len(pathspec);
186 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
189 int fill_directory(struct dir_struct *dir,
190 struct index_state *istate,
191 const struct pathspec *pathspec)
193 const char *prefix;
194 size_t prefix_len;
197 * Calculate common prefix for the pathspec, and
198 * use that to optimize the directory walk
200 prefix_len = common_prefix_len(pathspec);
201 prefix = prefix_len ? pathspec->items[0].match : "";
203 /* Read the directory and prune it */
204 read_directory(dir, istate, prefix, prefix_len, pathspec);
206 return prefix_len;
209 int within_depth(const char *name, int namelen,
210 int depth, int max_depth)
212 const char *cp = name, *cpe = name + namelen;
214 while (cp < cpe) {
215 if (*cp++ != '/')
216 continue;
217 depth++;
218 if (depth > max_depth)
219 return 0;
221 return 1;
225 * Read the contents of the blob with the given OID into a buffer.
226 * Append a trailing LF to the end if the last line doesn't have one.
228 * Returns:
229 * -1 when the OID is invalid or unknown or does not refer to a blob.
230 * 0 when the blob is empty.
231 * 1 along with { data, size } of the (possibly augmented) buffer
232 * when successful.
234 * Optionally updates the given sha1_stat with the given OID (when valid).
236 static int do_read_blob(const struct object_id *oid,
237 struct sha1_stat *sha1_stat,
238 size_t *size_out,
239 char **data_out)
241 enum object_type type;
242 unsigned long sz;
243 char *data;
245 *size_out = 0;
246 *data_out = NULL;
248 data = read_sha1_file(oid->hash, &type, &sz);
249 if (!data || type != OBJ_BLOB) {
250 free(data);
251 return -1;
254 if (sha1_stat) {
255 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
256 hashcpy(sha1_stat->sha1, oid->hash);
259 if (sz == 0) {
260 free(data);
261 return 0;
264 if (data[sz - 1] != '\n') {
265 data = xrealloc(data, st_add(sz, 1));
266 data[sz++] = '\n';
269 *size_out = xsize_t(sz);
270 *data_out = data;
272 return 1;
275 #define DO_MATCH_EXCLUDE (1<<0)
276 #define DO_MATCH_DIRECTORY (1<<1)
277 #define DO_MATCH_SUBMODULE (1<<2)
279 static int match_attrs(const char *name, int namelen,
280 const struct pathspec_item *item)
282 int i;
284 git_check_attr(name, item->attr_check);
285 for (i = 0; i < item->attr_match_nr; i++) {
286 const char *value;
287 int matched;
288 enum attr_match_mode match_mode;
290 value = item->attr_check->items[i].value;
291 match_mode = item->attr_match[i].match_mode;
293 if (ATTR_TRUE(value))
294 matched = (match_mode == MATCH_SET);
295 else if (ATTR_FALSE(value))
296 matched = (match_mode == MATCH_UNSET);
297 else if (ATTR_UNSET(value))
298 matched = (match_mode == MATCH_UNSPECIFIED);
299 else
300 matched = (match_mode == MATCH_VALUE &&
301 !strcmp(item->attr_match[i].value, value));
302 if (!matched)
303 return 0;
306 return 1;
310 * Does 'match' match the given name?
311 * A match is found if
313 * (1) the 'match' string is leading directory of 'name', or
314 * (2) the 'match' string is a wildcard and matches 'name', or
315 * (3) the 'match' string is exactly the same as 'name'.
317 * and the return value tells which case it was.
319 * It returns 0 when there is no match.
321 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
322 const char *name, int namelen, unsigned flags)
324 /* name/namelen has prefix cut off by caller */
325 const char *match = item->match + prefix;
326 int matchlen = item->len - prefix;
329 * The normal call pattern is:
330 * 1. prefix = common_prefix_len(ps);
331 * 2. prune something, or fill_directory
332 * 3. match_pathspec()
334 * 'prefix' at #1 may be shorter than the command's prefix and
335 * it's ok for #2 to match extra files. Those extras will be
336 * trimmed at #3.
338 * Suppose the pathspec is 'foo' and '../bar' running from
339 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
340 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
341 * user does not want XYZ/foo, only the "foo" part should be
342 * case-insensitive. We need to filter out XYZ/foo here. In
343 * other words, we do not trust the caller on comparing the
344 * prefix part when :(icase) is involved. We do exact
345 * comparison ourselves.
347 * Normally the caller (common_prefix_len() in fact) does
348 * _exact_ matching on name[-prefix+1..-1] and we do not need
349 * to check that part. Be defensive and check it anyway, in
350 * case common_prefix_len is changed, or a new caller is
351 * introduced that does not use common_prefix_len.
353 * If the penalty turns out too high when prefix is really
354 * long, maybe change it to
355 * strncmp(match, name, item->prefix - prefix)
357 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
358 strncmp(item->match, name - prefix, item->prefix))
359 return 0;
361 if (item->attr_match_nr && !match_attrs(name, namelen, item))
362 return 0;
364 /* If the match was just the prefix, we matched */
365 if (!*match)
366 return MATCHED_RECURSIVELY;
368 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
369 if (matchlen == namelen)
370 return MATCHED_EXACTLY;
372 if (match[matchlen-1] == '/' || name[matchlen] == '/')
373 return MATCHED_RECURSIVELY;
374 } else if ((flags & DO_MATCH_DIRECTORY) &&
375 match[matchlen - 1] == '/' &&
376 namelen == matchlen - 1 &&
377 !ps_strncmp(item, match, name, namelen))
378 return MATCHED_EXACTLY;
380 if (item->nowildcard_len < item->len &&
381 !git_fnmatch(item, match, name,
382 item->nowildcard_len - prefix))
383 return MATCHED_FNMATCH;
385 /* Perform checks to see if "name" is a super set of the pathspec */
386 if (flags & DO_MATCH_SUBMODULE) {
387 /* name is a literal prefix of the pathspec */
388 if ((namelen < matchlen) &&
389 (match[namelen] == '/') &&
390 !ps_strncmp(item, match, name, namelen))
391 return MATCHED_RECURSIVELY;
393 /* name" doesn't match up to the first wild character */
394 if (item->nowildcard_len < item->len &&
395 ps_strncmp(item, match, name,
396 item->nowildcard_len - prefix))
397 return 0;
400 * Here is where we would perform a wildmatch to check if
401 * "name" can be matched as a directory (or a prefix) against
402 * the pathspec. Since wildmatch doesn't have this capability
403 * at the present we have to punt and say that it is a match,
404 * potentially returning a false positive
405 * The submodules themselves will be able to perform more
406 * accurate matching to determine if the pathspec matches.
408 return MATCHED_RECURSIVELY;
411 return 0;
415 * Given a name and a list of pathspecs, returns the nature of the
416 * closest (i.e. most specific) match of the name to any of the
417 * pathspecs.
419 * The caller typically calls this multiple times with the same
420 * pathspec and seen[] array but with different name/namelen
421 * (e.g. entries from the index) and is interested in seeing if and
422 * how each pathspec matches all the names it calls this function
423 * with. A mark is left in the seen[] array for each pathspec element
424 * indicating the closest type of match that element achieved, so if
425 * seen[n] remains zero after multiple invocations, that means the nth
426 * pathspec did not match any names, which could indicate that the
427 * user mistyped the nth pathspec.
429 static int do_match_pathspec(const struct pathspec *ps,
430 const char *name, int namelen,
431 int prefix, char *seen,
432 unsigned flags)
434 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
436 GUARD_PATHSPEC(ps,
437 PATHSPEC_FROMTOP |
438 PATHSPEC_MAXDEPTH |
439 PATHSPEC_LITERAL |
440 PATHSPEC_GLOB |
441 PATHSPEC_ICASE |
442 PATHSPEC_EXCLUDE |
443 PATHSPEC_ATTR);
445 if (!ps->nr) {
446 if (!ps->recursive ||
447 !(ps->magic & PATHSPEC_MAXDEPTH) ||
448 ps->max_depth == -1)
449 return MATCHED_RECURSIVELY;
451 if (within_depth(name, namelen, 0, ps->max_depth))
452 return MATCHED_EXACTLY;
453 else
454 return 0;
457 name += prefix;
458 namelen -= prefix;
460 for (i = ps->nr - 1; i >= 0; i--) {
461 int how;
463 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
464 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
465 continue;
467 if (seen && seen[i] == MATCHED_EXACTLY)
468 continue;
470 * Make exclude patterns optional and never report
471 * "pathspec ':(exclude)foo' matches no files"
473 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
474 seen[i] = MATCHED_FNMATCH;
475 how = match_pathspec_item(ps->items+i, prefix, name,
476 namelen, flags);
477 if (ps->recursive &&
478 (ps->magic & PATHSPEC_MAXDEPTH) &&
479 ps->max_depth != -1 &&
480 how && how != MATCHED_FNMATCH) {
481 int len = ps->items[i].len;
482 if (name[len] == '/')
483 len++;
484 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
485 how = MATCHED_EXACTLY;
486 else
487 how = 0;
489 if (how) {
490 if (retval < how)
491 retval = how;
492 if (seen && seen[i] < how)
493 seen[i] = how;
496 return retval;
499 int match_pathspec(const struct pathspec *ps,
500 const char *name, int namelen,
501 int prefix, char *seen, int is_dir)
503 int positive, negative;
504 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
505 positive = do_match_pathspec(ps, name, namelen,
506 prefix, seen, flags);
507 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
508 return positive;
509 negative = do_match_pathspec(ps, name, namelen,
510 prefix, seen,
511 flags | DO_MATCH_EXCLUDE);
512 return negative ? 0 : positive;
516 * Check if a submodule is a superset of the pathspec
518 int submodule_path_match(const struct pathspec *ps,
519 const char *submodule_name,
520 char *seen)
522 int matched = do_match_pathspec(ps, submodule_name,
523 strlen(submodule_name),
524 0, seen,
525 DO_MATCH_DIRECTORY |
526 DO_MATCH_SUBMODULE);
527 return matched;
530 int report_path_error(const char *ps_matched,
531 const struct pathspec *pathspec,
532 const char *prefix)
535 * Make sure all pathspec matched; otherwise it is an error.
537 int num, errors = 0;
538 for (num = 0; num < pathspec->nr; num++) {
539 int other, found_dup;
541 if (ps_matched[num])
542 continue;
544 * The caller might have fed identical pathspec
545 * twice. Do not barf on such a mistake.
546 * FIXME: parse_pathspec should have eliminated
547 * duplicate pathspec.
549 for (found_dup = other = 0;
550 !found_dup && other < pathspec->nr;
551 other++) {
552 if (other == num || !ps_matched[other])
553 continue;
554 if (!strcmp(pathspec->items[other].original,
555 pathspec->items[num].original))
557 * Ok, we have a match already.
559 found_dup = 1;
561 if (found_dup)
562 continue;
564 error("pathspec '%s' did not match any file(s) known to git.",
565 pathspec->items[num].original);
566 errors++;
568 return errors;
572 * Return the length of the "simple" part of a path match limiter.
574 int simple_length(const char *match)
576 int len = -1;
578 for (;;) {
579 unsigned char c = *match++;
580 len++;
581 if (c == '\0' || is_glob_special(c))
582 return len;
586 int no_wildcard(const char *string)
588 return string[simple_length(string)] == '\0';
591 void parse_exclude_pattern(const char **pattern,
592 int *patternlen,
593 unsigned *flags,
594 int *nowildcardlen)
596 const char *p = *pattern;
597 size_t i, len;
599 *flags = 0;
600 if (*p == '!') {
601 *flags |= EXC_FLAG_NEGATIVE;
602 p++;
604 len = strlen(p);
605 if (len && p[len - 1] == '/') {
606 len--;
607 *flags |= EXC_FLAG_MUSTBEDIR;
609 for (i = 0; i < len; i++) {
610 if (p[i] == '/')
611 break;
613 if (i == len)
614 *flags |= EXC_FLAG_NODIR;
615 *nowildcardlen = simple_length(p);
617 * we should have excluded the trailing slash from 'p' too,
618 * but that's one more allocation. Instead just make sure
619 * nowildcardlen does not exceed real patternlen
621 if (*nowildcardlen > len)
622 *nowildcardlen = len;
623 if (*p == '*' && no_wildcard(p + 1))
624 *flags |= EXC_FLAG_ENDSWITH;
625 *pattern = p;
626 *patternlen = len;
629 void add_exclude(const char *string, const char *base,
630 int baselen, struct exclude_list *el, int srcpos)
632 struct exclude *x;
633 int patternlen;
634 unsigned flags;
635 int nowildcardlen;
637 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
638 if (flags & EXC_FLAG_MUSTBEDIR) {
639 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
640 } else {
641 x = xmalloc(sizeof(*x));
642 x->pattern = string;
644 x->patternlen = patternlen;
645 x->nowildcardlen = nowildcardlen;
646 x->base = base;
647 x->baselen = baselen;
648 x->flags = flags;
649 x->srcpos = srcpos;
650 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
651 el->excludes[el->nr++] = x;
652 x->el = el;
655 static int read_skip_worktree_file_from_index(const struct index_state *istate,
656 const char *path,
657 size_t *size_out,
658 char **data_out,
659 struct sha1_stat *sha1_stat)
661 int pos, len;
663 len = strlen(path);
664 pos = index_name_pos(istate, path, len);
665 if (pos < 0)
666 return -1;
667 if (!ce_skip_worktree(istate->cache[pos]))
668 return -1;
670 return do_read_blob(&istate->cache[pos]->oid, sha1_stat, size_out, data_out);
674 * Frees memory within el which was allocated for exclude patterns and
675 * the file buffer. Does not free el itself.
677 void clear_exclude_list(struct exclude_list *el)
679 int i;
681 for (i = 0; i < el->nr; i++)
682 free(el->excludes[i]);
683 free(el->excludes);
684 free(el->filebuf);
686 memset(el, 0, sizeof(*el));
689 static void trim_trailing_spaces(char *buf)
691 char *p, *last_space = NULL;
693 for (p = buf; *p; p++)
694 switch (*p) {
695 case ' ':
696 if (!last_space)
697 last_space = p;
698 break;
699 case '\\':
700 p++;
701 if (!*p)
702 return;
703 /* fallthrough */
704 default:
705 last_space = NULL;
708 if (last_space)
709 *last_space = '\0';
713 * Given a subdirectory name and "dir" of the current directory,
714 * search the subdir in "dir" and return it, or create a new one if it
715 * does not exist in "dir".
717 * If "name" has the trailing slash, it'll be excluded in the search.
719 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
720 struct untracked_cache_dir *dir,
721 const char *name, int len)
723 int first, last;
724 struct untracked_cache_dir *d;
725 if (!dir)
726 return NULL;
727 if (len && name[len - 1] == '/')
728 len--;
729 first = 0;
730 last = dir->dirs_nr;
731 while (last > first) {
732 int cmp, next = (last + first) >> 1;
733 d = dir->dirs[next];
734 cmp = strncmp(name, d->name, len);
735 if (!cmp && strlen(d->name) > len)
736 cmp = -1;
737 if (!cmp)
738 return d;
739 if (cmp < 0) {
740 last = next;
741 continue;
743 first = next+1;
746 uc->dir_created++;
747 FLEX_ALLOC_MEM(d, name, name, len);
749 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
750 memmove(dir->dirs + first + 1, dir->dirs + first,
751 (dir->dirs_nr - first) * sizeof(*dir->dirs));
752 dir->dirs_nr++;
753 dir->dirs[first] = d;
754 return d;
757 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
759 int i;
760 dir->valid = 0;
761 dir->untracked_nr = 0;
762 for (i = 0; i < dir->dirs_nr; i++)
763 do_invalidate_gitignore(dir->dirs[i]);
766 static void invalidate_gitignore(struct untracked_cache *uc,
767 struct untracked_cache_dir *dir)
769 uc->gitignore_invalidated++;
770 do_invalidate_gitignore(dir);
773 static void invalidate_directory(struct untracked_cache *uc,
774 struct untracked_cache_dir *dir)
776 int i;
777 uc->dir_invalidated++;
778 dir->valid = 0;
779 dir->untracked_nr = 0;
780 for (i = 0; i < dir->dirs_nr; i++)
781 dir->dirs[i]->recurse = 0;
784 static int add_excludes_from_buffer(char *buf, size_t size,
785 const char *base, int baselen,
786 struct exclude_list *el);
789 * Given a file with name "fname", read it (either from disk, or from
790 * an index if 'istate' is non-null), parse it and store the
791 * exclude rules in "el".
793 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
794 * stat data from disk (only valid if add_excludes returns zero). If
795 * ss_valid is non-zero, "ss" must contain good value as input.
797 static int add_excludes(const char *fname, const char *base, int baselen,
798 struct exclude_list *el,
799 struct index_state *istate,
800 struct sha1_stat *sha1_stat)
802 struct stat st;
803 int r;
804 int fd;
805 size_t size = 0;
806 char *buf;
808 fd = open(fname, O_RDONLY);
809 if (fd < 0 || fstat(fd, &st) < 0) {
810 if (fd < 0)
811 warn_on_fopen_errors(fname);
812 else
813 close(fd);
814 if (!istate)
815 return -1;
816 r = read_skip_worktree_file_from_index(istate, fname,
817 &size, &buf,
818 sha1_stat);
819 if (r != 1)
820 return r;
821 } else {
822 size = xsize_t(st.st_size);
823 if (size == 0) {
824 if (sha1_stat) {
825 fill_stat_data(&sha1_stat->stat, &st);
826 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
827 sha1_stat->valid = 1;
829 close(fd);
830 return 0;
832 buf = xmallocz(size);
833 if (read_in_full(fd, buf, size) != size) {
834 free(buf);
835 close(fd);
836 return -1;
838 buf[size++] = '\n';
839 close(fd);
840 if (sha1_stat) {
841 int pos;
842 if (sha1_stat->valid &&
843 !match_stat_data_racy(istate, &sha1_stat->stat, &st))
844 ; /* no content change, ss->sha1 still good */
845 else if (istate &&
846 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
847 !ce_stage(istate->cache[pos]) &&
848 ce_uptodate(istate->cache[pos]) &&
849 !would_convert_to_git(istate, fname))
850 hashcpy(sha1_stat->sha1,
851 istate->cache[pos]->oid.hash);
852 else
853 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
854 fill_stat_data(&sha1_stat->stat, &st);
855 sha1_stat->valid = 1;
859 add_excludes_from_buffer(buf, size, base, baselen, el);
860 return 0;
863 static int add_excludes_from_buffer(char *buf, size_t size,
864 const char *base, int baselen,
865 struct exclude_list *el)
867 int i, lineno = 1;
868 char *entry;
870 el->filebuf = buf;
872 if (skip_utf8_bom(&buf, size))
873 size -= buf - el->filebuf;
875 entry = buf;
877 for (i = 0; i < size; i++) {
878 if (buf[i] == '\n') {
879 if (entry != buf + i && entry[0] != '#') {
880 buf[i - (i && buf[i-1] == '\r')] = 0;
881 trim_trailing_spaces(entry);
882 add_exclude(entry, base, baselen, el, lineno);
884 lineno++;
885 entry = buf + i + 1;
888 return 0;
891 int add_excludes_from_file_to_list(const char *fname, const char *base,
892 int baselen, struct exclude_list *el,
893 struct index_state *istate)
895 return add_excludes(fname, base, baselen, el, istate, NULL);
898 int add_excludes_from_blob_to_list(
899 struct object_id *oid,
900 const char *base, int baselen,
901 struct exclude_list *el)
903 char *buf;
904 size_t size;
905 int r;
907 r = do_read_blob(oid, NULL, &size, &buf);
908 if (r != 1)
909 return r;
911 add_excludes_from_buffer(buf, size, base, baselen, el);
912 return 0;
915 struct exclude_list *add_exclude_list(struct dir_struct *dir,
916 int group_type, const char *src)
918 struct exclude_list *el;
919 struct exclude_list_group *group;
921 group = &dir->exclude_list_group[group_type];
922 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
923 el = &group->el[group->nr++];
924 memset(el, 0, sizeof(*el));
925 el->src = src;
926 return el;
930 * Used to set up core.excludesfile and .git/info/exclude lists.
932 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
933 struct sha1_stat *sha1_stat)
935 struct exclude_list *el;
937 * catch setup_standard_excludes() that's called before
938 * dir->untracked is assigned. That function behaves
939 * differently when dir->untracked is non-NULL.
941 if (!dir->untracked)
942 dir->unmanaged_exclude_files++;
943 el = add_exclude_list(dir, EXC_FILE, fname);
944 if (add_excludes(fname, "", 0, el, NULL, sha1_stat) < 0)
945 die("cannot use %s as an exclude file", fname);
948 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
950 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
951 add_excludes_from_file_1(dir, fname, NULL);
954 int match_basename(const char *basename, int basenamelen,
955 const char *pattern, int prefix, int patternlen,
956 unsigned flags)
958 if (prefix == patternlen) {
959 if (patternlen == basenamelen &&
960 !fspathncmp(pattern, basename, basenamelen))
961 return 1;
962 } else if (flags & EXC_FLAG_ENDSWITH) {
963 /* "*literal" matching against "fooliteral" */
964 if (patternlen - 1 <= basenamelen &&
965 !fspathncmp(pattern + 1,
966 basename + basenamelen - (patternlen - 1),
967 patternlen - 1))
968 return 1;
969 } else {
970 if (fnmatch_icase_mem(pattern, patternlen,
971 basename, basenamelen,
972 0) == 0)
973 return 1;
975 return 0;
978 int match_pathname(const char *pathname, int pathlen,
979 const char *base, int baselen,
980 const char *pattern, int prefix, int patternlen,
981 unsigned flags)
983 const char *name;
984 int namelen;
987 * match with FNM_PATHNAME; the pattern has base implicitly
988 * in front of it.
990 if (*pattern == '/') {
991 pattern++;
992 patternlen--;
993 prefix--;
997 * baselen does not count the trailing slash. base[] may or
998 * may not end with a trailing slash though.
1000 if (pathlen < baselen + 1 ||
1001 (baselen && pathname[baselen] != '/') ||
1002 fspathncmp(pathname, base, baselen))
1003 return 0;
1005 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1006 name = pathname + pathlen - namelen;
1008 if (prefix) {
1010 * if the non-wildcard part is longer than the
1011 * remaining pathname, surely it cannot match.
1013 if (prefix > namelen)
1014 return 0;
1016 if (fspathncmp(pattern, name, prefix))
1017 return 0;
1018 pattern += prefix;
1019 patternlen -= prefix;
1020 name += prefix;
1021 namelen -= prefix;
1024 * If the whole pattern did not have a wildcard,
1025 * then our prefix match is all we need; we
1026 * do not need to call fnmatch at all.
1028 if (!patternlen && !namelen)
1029 return 1;
1032 return fnmatch_icase_mem(pattern, patternlen,
1033 name, namelen,
1034 WM_PATHNAME) == 0;
1038 * Scan the given exclude list in reverse to see whether pathname
1039 * should be ignored. The first match (i.e. the last on the list), if
1040 * any, determines the fate. Returns the exclude_list element which
1041 * matched, or NULL for undecided.
1043 static struct exclude *last_exclude_matching_from_list(const char *pathname,
1044 int pathlen,
1045 const char *basename,
1046 int *dtype,
1047 struct exclude_list *el,
1048 struct index_state *istate)
1050 struct exclude *exc = NULL; /* undecided */
1051 int i;
1053 if (!el->nr)
1054 return NULL; /* undefined */
1056 for (i = el->nr - 1; 0 <= i; i--) {
1057 struct exclude *x = el->excludes[i];
1058 const char *exclude = x->pattern;
1059 int prefix = x->nowildcardlen;
1061 if (x->flags & EXC_FLAG_MUSTBEDIR) {
1062 if (*dtype == DT_UNKNOWN)
1063 *dtype = get_dtype(NULL, istate, pathname, pathlen);
1064 if (*dtype != DT_DIR)
1065 continue;
1068 if (x->flags & EXC_FLAG_NODIR) {
1069 if (match_basename(basename,
1070 pathlen - (basename - pathname),
1071 exclude, prefix, x->patternlen,
1072 x->flags)) {
1073 exc = x;
1074 break;
1076 continue;
1079 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1080 if (match_pathname(pathname, pathlen,
1081 x->base, x->baselen ? x->baselen - 1 : 0,
1082 exclude, prefix, x->patternlen, x->flags)) {
1083 exc = x;
1084 break;
1087 return exc;
1091 * Scan the list and let the last match determine the fate.
1092 * Return 1 for exclude, 0 for include and -1 for undecided.
1094 int is_excluded_from_list(const char *pathname,
1095 int pathlen, const char *basename, int *dtype,
1096 struct exclude_list *el, struct index_state *istate)
1098 struct exclude *exclude;
1099 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1100 dtype, el, istate);
1101 if (exclude)
1102 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1103 return -1; /* undecided */
1106 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1107 struct index_state *istate,
1108 const char *pathname, int pathlen, const char *basename,
1109 int *dtype_p)
1111 int i, j;
1112 struct exclude_list_group *group;
1113 struct exclude *exclude;
1114 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1115 group = &dir->exclude_list_group[i];
1116 for (j = group->nr - 1; j >= 0; j--) {
1117 exclude = last_exclude_matching_from_list(
1118 pathname, pathlen, basename, dtype_p,
1119 &group->el[j], istate);
1120 if (exclude)
1121 return exclude;
1124 return NULL;
1128 * Loads the per-directory exclude list for the substring of base
1129 * which has a char length of baselen.
1131 static void prep_exclude(struct dir_struct *dir,
1132 struct index_state *istate,
1133 const char *base, int baselen)
1135 struct exclude_list_group *group;
1136 struct exclude_list *el;
1137 struct exclude_stack *stk = NULL;
1138 struct untracked_cache_dir *untracked;
1139 int current;
1141 group = &dir->exclude_list_group[EXC_DIRS];
1144 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1145 * which originate from directories not in the prefix of the
1146 * path being checked.
1148 while ((stk = dir->exclude_stack) != NULL) {
1149 if (stk->baselen <= baselen &&
1150 !strncmp(dir->basebuf.buf, base, stk->baselen))
1151 break;
1152 el = &group->el[dir->exclude_stack->exclude_ix];
1153 dir->exclude_stack = stk->prev;
1154 dir->exclude = NULL;
1155 free((char *)el->src); /* see strbuf_detach() below */
1156 clear_exclude_list(el);
1157 free(stk);
1158 group->nr--;
1161 /* Skip traversing into sub directories if the parent is excluded */
1162 if (dir->exclude)
1163 return;
1166 * Lazy initialization. All call sites currently just
1167 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1168 * them seems lots of work for little benefit.
1170 if (!dir->basebuf.buf)
1171 strbuf_init(&dir->basebuf, PATH_MAX);
1173 /* Read from the parent directories and push them down. */
1174 current = stk ? stk->baselen : -1;
1175 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1176 if (dir->untracked)
1177 untracked = stk ? stk->ucd : dir->untracked->root;
1178 else
1179 untracked = NULL;
1181 while (current < baselen) {
1182 const char *cp;
1183 struct sha1_stat sha1_stat;
1185 stk = xcalloc(1, sizeof(*stk));
1186 if (current < 0) {
1187 cp = base;
1188 current = 0;
1189 } else {
1190 cp = strchr(base + current + 1, '/');
1191 if (!cp)
1192 die("oops in prep_exclude");
1193 cp++;
1194 untracked =
1195 lookup_untracked(dir->untracked, untracked,
1196 base + current,
1197 cp - base - current);
1199 stk->prev = dir->exclude_stack;
1200 stk->baselen = cp - base;
1201 stk->exclude_ix = group->nr;
1202 stk->ucd = untracked;
1203 el = add_exclude_list(dir, EXC_DIRS, NULL);
1204 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1205 assert(stk->baselen == dir->basebuf.len);
1207 /* Abort if the directory is excluded */
1208 if (stk->baselen) {
1209 int dt = DT_DIR;
1210 dir->basebuf.buf[stk->baselen - 1] = 0;
1211 dir->exclude = last_exclude_matching_from_lists(dir,
1212 istate,
1213 dir->basebuf.buf, stk->baselen - 1,
1214 dir->basebuf.buf + current, &dt);
1215 dir->basebuf.buf[stk->baselen - 1] = '/';
1216 if (dir->exclude &&
1217 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1218 dir->exclude = NULL;
1219 if (dir->exclude) {
1220 dir->exclude_stack = stk;
1221 return;
1225 /* Try to read per-directory file */
1226 hashclr(sha1_stat.sha1);
1227 sha1_stat.valid = 0;
1228 if (dir->exclude_per_dir &&
1230 * If we know that no files have been added in
1231 * this directory (i.e. valid_cached_dir() has
1232 * been executed and set untracked->valid) ..
1234 (!untracked || !untracked->valid ||
1236 * .. and .gitignore does not exist before
1237 * (i.e. null exclude_sha1). Then we can skip
1238 * loading .gitignore, which would result in
1239 * ENOENT anyway.
1241 !is_null_sha1(untracked->exclude_sha1))) {
1243 * dir->basebuf gets reused by the traversal, but we
1244 * need fname to remain unchanged to ensure the src
1245 * member of each struct exclude correctly
1246 * back-references its source file. Other invocations
1247 * of add_exclude_list provide stable strings, so we
1248 * strbuf_detach() and free() here in the caller.
1250 struct strbuf sb = STRBUF_INIT;
1251 strbuf_addbuf(&sb, &dir->basebuf);
1252 strbuf_addstr(&sb, dir->exclude_per_dir);
1253 el->src = strbuf_detach(&sb, NULL);
1254 add_excludes(el->src, el->src, stk->baselen, el, istate,
1255 untracked ? &sha1_stat : NULL);
1258 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1259 * will first be called in valid_cached_dir() then maybe many
1260 * times more in last_exclude_matching(). When the cache is
1261 * used, last_exclude_matching() will not be called and
1262 * reading .gitignore content will be a waste.
1264 * So when it's called by valid_cached_dir() and we can get
1265 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1266 * modified on work tree), we could delay reading the
1267 * .gitignore content until we absolutely need it in
1268 * last_exclude_matching(). Be careful about ignore rule
1269 * order, though, if you do that.
1271 if (untracked &&
1272 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1273 invalidate_gitignore(dir->untracked, untracked);
1274 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1276 dir->exclude_stack = stk;
1277 current = stk->baselen;
1279 strbuf_setlen(&dir->basebuf, baselen);
1283 * Loads the exclude lists for the directory containing pathname, then
1284 * scans all exclude lists to determine whether pathname is excluded.
1285 * Returns the exclude_list element which matched, or NULL for
1286 * undecided.
1288 struct exclude *last_exclude_matching(struct dir_struct *dir,
1289 struct index_state *istate,
1290 const char *pathname,
1291 int *dtype_p)
1293 int pathlen = strlen(pathname);
1294 const char *basename = strrchr(pathname, '/');
1295 basename = (basename) ? basename+1 : pathname;
1297 prep_exclude(dir, istate, pathname, basename-pathname);
1299 if (dir->exclude)
1300 return dir->exclude;
1302 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
1303 basename, dtype_p);
1307 * Loads the exclude lists for the directory containing pathname, then
1308 * scans all exclude lists to determine whether pathname is excluded.
1309 * Returns 1 if true, otherwise 0.
1311 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1312 const char *pathname, int *dtype_p)
1314 struct exclude *exclude =
1315 last_exclude_matching(dir, istate, pathname, dtype_p);
1316 if (exclude)
1317 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1318 return 0;
1321 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1323 struct dir_entry *ent;
1325 FLEX_ALLOC_MEM(ent, name, pathname, len);
1326 ent->len = len;
1327 return ent;
1330 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1331 struct index_state *istate,
1332 const char *pathname, int len)
1334 if (index_file_exists(istate, pathname, len, ignore_case))
1335 return NULL;
1337 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1338 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1341 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1342 struct index_state *istate,
1343 const char *pathname, int len)
1345 if (!index_name_is_other(istate, pathname, len))
1346 return NULL;
1348 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1349 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1352 enum exist_status {
1353 index_nonexistent = 0,
1354 index_directory,
1355 index_gitdir
1359 * Do not use the alphabetically sorted index to look up
1360 * the directory name; instead, use the case insensitive
1361 * directory hash.
1363 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1364 const char *dirname, int len)
1366 struct cache_entry *ce;
1368 if (index_dir_exists(istate, dirname, len))
1369 return index_directory;
1371 ce = index_file_exists(istate, dirname, len, ignore_case);
1372 if (ce && S_ISGITLINK(ce->ce_mode))
1373 return index_gitdir;
1375 return index_nonexistent;
1379 * The index sorts alphabetically by entry name, which
1380 * means that a gitlink sorts as '\0' at the end, while
1381 * a directory (which is defined not as an entry, but as
1382 * the files it contains) will sort with the '/' at the
1383 * end.
1385 static enum exist_status directory_exists_in_index(struct index_state *istate,
1386 const char *dirname, int len)
1388 int pos;
1390 if (ignore_case)
1391 return directory_exists_in_index_icase(istate, dirname, len);
1393 pos = index_name_pos(istate, dirname, len);
1394 if (pos < 0)
1395 pos = -pos-1;
1396 while (pos < istate->cache_nr) {
1397 const struct cache_entry *ce = istate->cache[pos++];
1398 unsigned char endchar;
1400 if (strncmp(ce->name, dirname, len))
1401 break;
1402 endchar = ce->name[len];
1403 if (endchar > '/')
1404 break;
1405 if (endchar == '/')
1406 return index_directory;
1407 if (!endchar && S_ISGITLINK(ce->ce_mode))
1408 return index_gitdir;
1410 return index_nonexistent;
1414 * When we find a directory when traversing the filesystem, we
1415 * have three distinct cases:
1417 * - ignore it
1418 * - see it as a directory
1419 * - recurse into it
1421 * and which one we choose depends on a combination of existing
1422 * git index contents and the flags passed into the directory
1423 * traversal routine.
1425 * Case 1: If we *already* have entries in the index under that
1426 * directory name, we always recurse into the directory to see
1427 * all the files.
1429 * Case 2: If we *already* have that directory name as a gitlink,
1430 * we always continue to see it as a gitlink, regardless of whether
1431 * there is an actual git directory there or not (it might not
1432 * be checked out as a subproject!)
1434 * Case 3: if we didn't have it in the index previously, we
1435 * have a few sub-cases:
1437 * (a) if "show_other_directories" is true, we show it as
1438 * just a directory, unless "hide_empty_directories" is
1439 * also true, in which case we need to check if it contains any
1440 * untracked and / or ignored files.
1441 * (b) if it looks like a git directory, and we don't have
1442 * 'no_gitlinks' set we treat it as a gitlink, and show it
1443 * as a directory.
1444 * (c) otherwise, we recurse into it.
1446 static enum path_treatment treat_directory(struct dir_struct *dir,
1447 struct index_state *istate,
1448 struct untracked_cache_dir *untracked,
1449 const char *dirname, int len, int baselen, int exclude,
1450 const struct pathspec *pathspec)
1452 /* The "len-1" is to strip the final '/' */
1453 switch (directory_exists_in_index(istate, dirname, len-1)) {
1454 case index_directory:
1455 return path_recurse;
1457 case index_gitdir:
1458 return path_none;
1460 case index_nonexistent:
1461 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1462 break;
1463 if (exclude &&
1464 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1465 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1468 * This is an excluded directory and we are
1469 * showing ignored paths that match an exclude
1470 * pattern. (e.g. show directory as ignored
1471 * only if it matches an exclude pattern).
1472 * This path will either be 'path_excluded`
1473 * (if we are showing empty directories or if
1474 * the directory is not empty), or will be
1475 * 'path_none' (empty directory, and we are
1476 * not showing empty directories).
1478 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1479 return path_excluded;
1481 if (read_directory_recursive(dir, istate, dirname, len,
1482 untracked, 1, 1, pathspec) == path_excluded)
1483 return path_excluded;
1485 return path_none;
1487 if (!(dir->flags & DIR_NO_GITLINKS)) {
1488 struct object_id oid;
1489 if (resolve_gitlink_ref(dirname, "HEAD", &oid) == 0)
1490 return exclude ? path_excluded : path_untracked;
1492 return path_recurse;
1495 /* This is the "show_other_directories" case */
1497 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1498 return exclude ? path_excluded : path_untracked;
1500 untracked = lookup_untracked(dir->untracked, untracked,
1501 dirname + baselen, len - baselen);
1504 * If this is an excluded directory, then we only need to check if
1505 * the directory contains any files.
1507 return read_directory_recursive(dir, istate, dirname, len,
1508 untracked, 1, exclude, pathspec);
1512 * This is an inexact early pruning of any recursive directory
1513 * reading - if the path cannot possibly be in the pathspec,
1514 * return true, and we'll skip it early.
1516 static int simplify_away(const char *path, int pathlen,
1517 const struct pathspec *pathspec)
1519 int i;
1521 if (!pathspec || !pathspec->nr)
1522 return 0;
1524 GUARD_PATHSPEC(pathspec,
1525 PATHSPEC_FROMTOP |
1526 PATHSPEC_MAXDEPTH |
1527 PATHSPEC_LITERAL |
1528 PATHSPEC_GLOB |
1529 PATHSPEC_ICASE |
1530 PATHSPEC_EXCLUDE |
1531 PATHSPEC_ATTR);
1533 for (i = 0; i < pathspec->nr; i++) {
1534 const struct pathspec_item *item = &pathspec->items[i];
1535 int len = item->nowildcard_len;
1537 if (len > pathlen)
1538 len = pathlen;
1539 if (!ps_strncmp(item, item->match, path, len))
1540 return 0;
1543 return 1;
1547 * This function tells us whether an excluded path matches a
1548 * list of "interesting" pathspecs. That is, whether a path matched
1549 * by any of the pathspecs could possibly be ignored by excluding
1550 * the specified path. This can happen if:
1552 * 1. the path is mentioned explicitly in the pathspec
1554 * 2. the path is a directory prefix of some element in the
1555 * pathspec
1557 static int exclude_matches_pathspec(const char *path, int pathlen,
1558 const struct pathspec *pathspec)
1560 int i;
1562 if (!pathspec || !pathspec->nr)
1563 return 0;
1565 GUARD_PATHSPEC(pathspec,
1566 PATHSPEC_FROMTOP |
1567 PATHSPEC_MAXDEPTH |
1568 PATHSPEC_LITERAL |
1569 PATHSPEC_GLOB |
1570 PATHSPEC_ICASE |
1571 PATHSPEC_EXCLUDE);
1573 for (i = 0; i < pathspec->nr; i++) {
1574 const struct pathspec_item *item = &pathspec->items[i];
1575 int len = item->nowildcard_len;
1577 if (len == pathlen &&
1578 !ps_strncmp(item, item->match, path, pathlen))
1579 return 1;
1580 if (len > pathlen &&
1581 item->match[pathlen] == '/' &&
1582 !ps_strncmp(item, item->match, path, pathlen))
1583 return 1;
1585 return 0;
1588 static int get_index_dtype(struct index_state *istate,
1589 const char *path, int len)
1591 int pos;
1592 const struct cache_entry *ce;
1594 ce = index_file_exists(istate, path, len, 0);
1595 if (ce) {
1596 if (!ce_uptodate(ce))
1597 return DT_UNKNOWN;
1598 if (S_ISGITLINK(ce->ce_mode))
1599 return DT_DIR;
1601 * Nobody actually cares about the
1602 * difference between DT_LNK and DT_REG
1604 return DT_REG;
1607 /* Try to look it up as a directory */
1608 pos = index_name_pos(istate, path, len);
1609 if (pos >= 0)
1610 return DT_UNKNOWN;
1611 pos = -pos-1;
1612 while (pos < istate->cache_nr) {
1613 ce = istate->cache[pos++];
1614 if (strncmp(ce->name, path, len))
1615 break;
1616 if (ce->name[len] > '/')
1617 break;
1618 if (ce->name[len] < '/')
1619 continue;
1620 if (!ce_uptodate(ce))
1621 break; /* continue? */
1622 return DT_DIR;
1624 return DT_UNKNOWN;
1627 static int get_dtype(struct dirent *de, struct index_state *istate,
1628 const char *path, int len)
1630 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1631 struct stat st;
1633 if (dtype != DT_UNKNOWN)
1634 return dtype;
1635 dtype = get_index_dtype(istate, path, len);
1636 if (dtype != DT_UNKNOWN)
1637 return dtype;
1638 if (lstat(path, &st))
1639 return dtype;
1640 if (S_ISREG(st.st_mode))
1641 return DT_REG;
1642 if (S_ISDIR(st.st_mode))
1643 return DT_DIR;
1644 if (S_ISLNK(st.st_mode))
1645 return DT_LNK;
1646 return dtype;
1649 static enum path_treatment treat_one_path(struct dir_struct *dir,
1650 struct untracked_cache_dir *untracked,
1651 struct index_state *istate,
1652 struct strbuf *path,
1653 int baselen,
1654 const struct pathspec *pathspec,
1655 int dtype, struct dirent *de)
1657 int exclude;
1658 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1659 enum path_treatment path_treatment;
1661 if (dtype == DT_UNKNOWN)
1662 dtype = get_dtype(de, istate, path->buf, path->len);
1664 /* Always exclude indexed files */
1665 if (dtype != DT_DIR && has_path_in_index)
1666 return path_none;
1669 * When we are looking at a directory P in the working tree,
1670 * there are three cases:
1672 * (1) P exists in the index. Everything inside the directory P in
1673 * the working tree needs to go when P is checked out from the
1674 * index.
1676 * (2) P does not exist in the index, but there is P/Q in the index.
1677 * We know P will stay a directory when we check out the contents
1678 * of the index, but we do not know yet if there is a directory
1679 * P/Q in the working tree to be killed, so we need to recurse.
1681 * (3) P does not exist in the index, and there is no P/Q in the index
1682 * to require P to be a directory, either. Only in this case, we
1683 * know that everything inside P will not be killed without
1684 * recursing.
1686 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1687 (dtype == DT_DIR) &&
1688 !has_path_in_index &&
1689 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1690 return path_none;
1692 exclude = is_excluded(dir, istate, path->buf, &dtype);
1695 * Excluded? If we don't explicitly want to show
1696 * ignored files, ignore it
1698 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1699 return path_excluded;
1701 switch (dtype) {
1702 default:
1703 return path_none;
1704 case DT_DIR:
1705 strbuf_addch(path, '/');
1706 path_treatment = treat_directory(dir, istate, untracked,
1707 path->buf, path->len,
1708 baselen, exclude, pathspec);
1710 * If 1) we only want to return directories that
1711 * match an exclude pattern and 2) this directory does
1712 * not match an exclude pattern but all of its
1713 * contents are excluded, then indicate that we should
1714 * recurse into this directory (instead of marking the
1715 * directory itself as an ignored path).
1717 if (!exclude &&
1718 path_treatment == path_excluded &&
1719 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1720 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1721 return path_recurse;
1722 return path_treatment;
1723 case DT_REG:
1724 case DT_LNK:
1725 return exclude ? path_excluded : path_untracked;
1729 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1730 struct untracked_cache_dir *untracked,
1731 struct cached_dir *cdir,
1732 struct index_state *istate,
1733 struct strbuf *path,
1734 int baselen,
1735 const struct pathspec *pathspec)
1737 strbuf_setlen(path, baselen);
1738 if (!cdir->ucd) {
1739 strbuf_addstr(path, cdir->file);
1740 return path_untracked;
1742 strbuf_addstr(path, cdir->ucd->name);
1743 /* treat_one_path() does this before it calls treat_directory() */
1744 strbuf_complete(path, '/');
1745 if (cdir->ucd->check_only)
1747 * check_only is set as a result of treat_directory() getting
1748 * to its bottom. Verify again the same set of directories
1749 * with check_only set.
1751 return read_directory_recursive(dir, istate, path->buf, path->len,
1752 cdir->ucd, 1, 0, pathspec);
1754 * We get path_recurse in the first run when
1755 * directory_exists_in_index() returns index_nonexistent. We
1756 * are sure that new changes in the index does not impact the
1757 * outcome. Return now.
1759 return path_recurse;
1762 static enum path_treatment treat_path(struct dir_struct *dir,
1763 struct untracked_cache_dir *untracked,
1764 struct cached_dir *cdir,
1765 struct index_state *istate,
1766 struct strbuf *path,
1767 int baselen,
1768 const struct pathspec *pathspec)
1770 int dtype;
1771 struct dirent *de = cdir->de;
1773 if (!de)
1774 return treat_path_fast(dir, untracked, cdir, istate, path,
1775 baselen, pathspec);
1776 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1777 return path_none;
1778 strbuf_setlen(path, baselen);
1779 strbuf_addstr(path, de->d_name);
1780 if (simplify_away(path->buf, path->len, pathspec))
1781 return path_none;
1783 dtype = DTYPE(de);
1784 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1787 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1789 if (!dir)
1790 return;
1791 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1792 dir->untracked_alloc);
1793 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1796 static int valid_cached_dir(struct dir_struct *dir,
1797 struct untracked_cache_dir *untracked,
1798 struct index_state *istate,
1799 struct strbuf *path,
1800 int check_only)
1802 struct stat st;
1804 if (!untracked)
1805 return 0;
1808 * With fsmonitor, we can trust the untracked cache's valid field.
1810 refresh_fsmonitor(istate);
1811 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1812 if (stat(path->len ? path->buf : ".", &st)) {
1813 invalidate_directory(dir->untracked, untracked);
1814 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1815 return 0;
1817 if (!untracked->valid ||
1818 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1819 if (untracked->valid)
1820 invalidate_directory(dir->untracked, untracked);
1821 fill_stat_data(&untracked->stat_data, &st);
1822 return 0;
1826 if (untracked->check_only != !!check_only) {
1827 invalidate_directory(dir->untracked, untracked);
1828 return 0;
1832 * prep_exclude will be called eventually on this directory,
1833 * but it's called much later in last_exclude_matching(). We
1834 * need it now to determine the validity of the cache for this
1835 * path. The next calls will be nearly no-op, the way
1836 * prep_exclude() is designed.
1838 if (path->len && path->buf[path->len - 1] != '/') {
1839 strbuf_addch(path, '/');
1840 prep_exclude(dir, istate, path->buf, path->len);
1841 strbuf_setlen(path, path->len - 1);
1842 } else
1843 prep_exclude(dir, istate, path->buf, path->len);
1845 /* hopefully prep_exclude() haven't invalidated this entry... */
1846 return untracked->valid;
1849 static int open_cached_dir(struct cached_dir *cdir,
1850 struct dir_struct *dir,
1851 struct untracked_cache_dir *untracked,
1852 struct index_state *istate,
1853 struct strbuf *path,
1854 int check_only)
1856 memset(cdir, 0, sizeof(*cdir));
1857 cdir->untracked = untracked;
1858 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1859 return 0;
1860 cdir->fdir = opendir(path->len ? path->buf : ".");
1861 if (dir->untracked)
1862 dir->untracked->dir_opened++;
1863 if (!cdir->fdir)
1864 return -1;
1865 return 0;
1868 static int read_cached_dir(struct cached_dir *cdir)
1870 if (cdir->fdir) {
1871 cdir->de = readdir(cdir->fdir);
1872 if (!cdir->de)
1873 return -1;
1874 return 0;
1876 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1877 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1878 if (!d->recurse) {
1879 cdir->nr_dirs++;
1880 continue;
1882 cdir->ucd = d;
1883 cdir->nr_dirs++;
1884 return 0;
1886 cdir->ucd = NULL;
1887 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1888 struct untracked_cache_dir *d = cdir->untracked;
1889 cdir->file = d->untracked[cdir->nr_files++];
1890 return 0;
1892 return -1;
1895 static void close_cached_dir(struct cached_dir *cdir)
1897 if (cdir->fdir)
1898 closedir(cdir->fdir);
1900 * We have gone through this directory and found no untracked
1901 * entries. Mark it valid.
1903 if (cdir->untracked) {
1904 cdir->untracked->valid = 1;
1905 cdir->untracked->recurse = 1;
1910 * Read a directory tree. We currently ignore anything but
1911 * directories, regular files and symlinks. That's because git
1912 * doesn't handle them at all yet. Maybe that will change some
1913 * day.
1915 * Also, we ignore the name ".git" (even if it is not a directory).
1916 * That likely will not change.
1918 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1919 * to signal that a file was found. This is the least significant value that
1920 * indicates that a file was encountered that does not depend on the order of
1921 * whether an untracked or exluded path was encountered first.
1923 * Returns the most significant path_treatment value encountered in the scan.
1924 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1925 * significant path_treatment value that will be returned.
1928 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1929 struct index_state *istate, const char *base, int baselen,
1930 struct untracked_cache_dir *untracked, int check_only,
1931 int stop_at_first_file, const struct pathspec *pathspec)
1933 struct cached_dir cdir;
1934 enum path_treatment state, subdir_state, dir_state = path_none;
1935 struct strbuf path = STRBUF_INIT;
1937 strbuf_add(&path, base, baselen);
1939 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1940 goto out;
1942 if (untracked)
1943 untracked->check_only = !!check_only;
1945 while (!read_cached_dir(&cdir)) {
1946 /* check how the file or directory should be treated */
1947 state = treat_path(dir, untracked, &cdir, istate, &path,
1948 baselen, pathspec);
1950 if (state > dir_state)
1951 dir_state = state;
1953 /* recurse into subdir if instructed by treat_path */
1954 if ((state == path_recurse) ||
1955 ((state == path_untracked) &&
1956 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1957 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
1958 struct untracked_cache_dir *ud;
1959 ud = lookup_untracked(dir->untracked, untracked,
1960 path.buf + baselen,
1961 path.len - baselen);
1962 subdir_state =
1963 read_directory_recursive(dir, istate, path.buf,
1964 path.len, ud,
1965 check_only, stop_at_first_file, pathspec);
1966 if (subdir_state > dir_state)
1967 dir_state = subdir_state;
1970 if (check_only) {
1971 if (stop_at_first_file) {
1973 * If stopping at first file, then
1974 * signal that a file was found by
1975 * returning `path_excluded`. This is
1976 * to return a consistent value
1977 * regardless of whether an ignored or
1978 * excluded file happened to be
1979 * encountered 1st.
1981 * In current usage, the
1982 * `stop_at_first_file` is passed when
1983 * an ancestor directory has matched
1984 * an exclude pattern, so any found
1985 * files will be excluded.
1987 if (dir_state >= path_excluded) {
1988 dir_state = path_excluded;
1989 break;
1993 /* abort early if maximum state has been reached */
1994 if (dir_state == path_untracked) {
1995 if (cdir.fdir)
1996 add_untracked(untracked, path.buf + baselen);
1997 break;
1999 /* skip the dir_add_* part */
2000 continue;
2003 /* add the path to the appropriate result list */
2004 switch (state) {
2005 case path_excluded:
2006 if (dir->flags & DIR_SHOW_IGNORED)
2007 dir_add_name(dir, istate, path.buf, path.len);
2008 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2009 ((dir->flags & DIR_COLLECT_IGNORED) &&
2010 exclude_matches_pathspec(path.buf, path.len,
2011 pathspec)))
2012 dir_add_ignored(dir, istate, path.buf, path.len);
2013 break;
2015 case path_untracked:
2016 if (dir->flags & DIR_SHOW_IGNORED)
2017 break;
2018 dir_add_name(dir, istate, path.buf, path.len);
2019 if (cdir.fdir)
2020 add_untracked(untracked, path.buf + baselen);
2021 break;
2023 default:
2024 break;
2027 close_cached_dir(&cdir);
2028 out:
2029 strbuf_release(&path);
2031 return dir_state;
2034 int cmp_dir_entry(const void *p1, const void *p2)
2036 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2037 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2039 return name_compare(e1->name, e1->len, e2->name, e2->len);
2042 /* check if *out lexically strictly contains *in */
2043 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2045 return (out->len < in->len) &&
2046 (out->name[out->len - 1] == '/') &&
2047 !memcmp(out->name, in->name, out->len);
2050 static int treat_leading_path(struct dir_struct *dir,
2051 struct index_state *istate,
2052 const char *path, int len,
2053 const struct pathspec *pathspec)
2055 struct strbuf sb = STRBUF_INIT;
2056 int baselen, rc = 0;
2057 const char *cp;
2058 int old_flags = dir->flags;
2060 while (len && path[len - 1] == '/')
2061 len--;
2062 if (!len)
2063 return 1;
2064 baselen = 0;
2065 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
2066 while (1) {
2067 cp = path + baselen + !!baselen;
2068 cp = memchr(cp, '/', path + len - cp);
2069 if (!cp)
2070 baselen = len;
2071 else
2072 baselen = cp - path;
2073 strbuf_setlen(&sb, 0);
2074 strbuf_add(&sb, path, baselen);
2075 if (!is_directory(sb.buf))
2076 break;
2077 if (simplify_away(sb.buf, sb.len, pathspec))
2078 break;
2079 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
2080 DT_DIR, NULL) == path_none)
2081 break; /* do not recurse into it */
2082 if (len <= baselen) {
2083 rc = 1;
2084 break; /* finished checking */
2087 strbuf_release(&sb);
2088 dir->flags = old_flags;
2089 return rc;
2092 static const char *get_ident_string(void)
2094 static struct strbuf sb = STRBUF_INIT;
2095 struct utsname uts;
2097 if (sb.len)
2098 return sb.buf;
2099 if (uname(&uts) < 0)
2100 die_errno(_("failed to get kernel name and information"));
2101 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2102 uts.sysname);
2103 return sb.buf;
2106 static int ident_in_untracked(const struct untracked_cache *uc)
2109 * Previous git versions may have saved many NUL separated
2110 * strings in the "ident" field, but it is insane to manage
2111 * many locations, so just take care of the first one.
2114 return !strcmp(uc->ident.buf, get_ident_string());
2117 static void set_untracked_ident(struct untracked_cache *uc)
2119 strbuf_reset(&uc->ident);
2120 strbuf_addstr(&uc->ident, get_ident_string());
2123 * This strbuf used to contain a list of NUL separated
2124 * strings, so save NUL too for backward compatibility.
2126 strbuf_addch(&uc->ident, 0);
2129 static void new_untracked_cache(struct index_state *istate)
2131 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2132 strbuf_init(&uc->ident, 100);
2133 uc->exclude_per_dir = ".gitignore";
2134 /* should be the same flags used by git-status */
2135 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2136 set_untracked_ident(uc);
2137 istate->untracked = uc;
2138 istate->cache_changed |= UNTRACKED_CHANGED;
2141 void add_untracked_cache(struct index_state *istate)
2143 if (!istate->untracked) {
2144 new_untracked_cache(istate);
2145 } else {
2146 if (!ident_in_untracked(istate->untracked)) {
2147 free_untracked_cache(istate->untracked);
2148 new_untracked_cache(istate);
2153 void remove_untracked_cache(struct index_state *istate)
2155 if (istate->untracked) {
2156 free_untracked_cache(istate->untracked);
2157 istate->untracked = NULL;
2158 istate->cache_changed |= UNTRACKED_CHANGED;
2162 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2163 int base_len,
2164 const struct pathspec *pathspec)
2166 struct untracked_cache_dir *root;
2168 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
2169 return NULL;
2172 * We only support $GIT_DIR/info/exclude and core.excludesfile
2173 * as the global ignore rule files. Any other additions
2174 * (e.g. from command line) invalidate the cache. This
2175 * condition also catches running setup_standard_excludes()
2176 * before setting dir->untracked!
2178 if (dir->unmanaged_exclude_files)
2179 return NULL;
2182 * Optimize for the main use case only: whole-tree git
2183 * status. More work involved in treat_leading_path() if we
2184 * use cache on just a subset of the worktree. pathspec
2185 * support could make the matter even worse.
2187 if (base_len || (pathspec && pathspec->nr))
2188 return NULL;
2190 /* Different set of flags may produce different results */
2191 if (dir->flags != dir->untracked->dir_flags ||
2193 * See treat_directory(), case index_nonexistent. Without
2194 * this flag, we may need to also cache .git file content
2195 * for the resolve_gitlink_ref() call, which we don't.
2197 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2198 /* We don't support collecting ignore files */
2199 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2200 DIR_COLLECT_IGNORED)))
2201 return NULL;
2204 * If we use .gitignore in the cache and now you change it to
2205 * .gitexclude, everything will go wrong.
2207 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2208 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2209 return NULL;
2212 * EXC_CMDL is not considered in the cache. If people set it,
2213 * skip the cache.
2215 if (dir->exclude_list_group[EXC_CMDL].nr)
2216 return NULL;
2218 if (!ident_in_untracked(dir->untracked)) {
2219 warning(_("Untracked cache is disabled on this system or location."));
2220 return NULL;
2223 if (!dir->untracked->root) {
2224 const int len = sizeof(*dir->untracked->root);
2225 dir->untracked->root = xmalloc(len);
2226 memset(dir->untracked->root, 0, len);
2229 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2230 root = dir->untracked->root;
2231 if (hashcmp(dir->ss_info_exclude.sha1,
2232 dir->untracked->ss_info_exclude.sha1)) {
2233 invalidate_gitignore(dir->untracked, root);
2234 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2236 if (hashcmp(dir->ss_excludes_file.sha1,
2237 dir->untracked->ss_excludes_file.sha1)) {
2238 invalidate_gitignore(dir->untracked, root);
2239 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2242 /* Make sure this directory is not dropped out at saving phase */
2243 root->recurse = 1;
2244 return root;
2247 int read_directory(struct dir_struct *dir, struct index_state *istate,
2248 const char *path, int len, const struct pathspec *pathspec)
2250 struct untracked_cache_dir *untracked;
2252 if (has_symlink_leading_path(path, len))
2253 return dir->nr;
2255 untracked = validate_untracked_cache(dir, len, pathspec);
2256 if (!untracked)
2258 * make sure untracked cache code path is disabled,
2259 * e.g. prep_exclude()
2261 dir->untracked = NULL;
2262 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2263 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2264 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2265 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2268 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2269 * also pick up untracked contents of untracked dirs; by default
2270 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2272 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2273 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2274 int i, j;
2276 /* remove from dir->entries untracked contents of untracked dirs */
2277 for (i = j = 0; j < dir->nr; j++) {
2278 if (i &&
2279 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2280 FREE_AND_NULL(dir->entries[j]);
2281 } else {
2282 dir->entries[i++] = dir->entries[j];
2286 dir->nr = i;
2289 if (dir->untracked) {
2290 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2291 trace_printf_key(&trace_untracked_stats,
2292 "node creation: %u\n"
2293 "gitignore invalidation: %u\n"
2294 "directory invalidation: %u\n"
2295 "opendir: %u\n",
2296 dir->untracked->dir_created,
2297 dir->untracked->gitignore_invalidated,
2298 dir->untracked->dir_invalidated,
2299 dir->untracked->dir_opened);
2300 if (dir->untracked == istate->untracked &&
2301 (dir->untracked->dir_opened ||
2302 dir->untracked->gitignore_invalidated ||
2303 dir->untracked->dir_invalidated))
2304 istate->cache_changed |= UNTRACKED_CHANGED;
2305 if (dir->untracked != istate->untracked) {
2306 FREE_AND_NULL(dir->untracked);
2309 return dir->nr;
2312 int file_exists(const char *f)
2314 struct stat sb;
2315 return lstat(f, &sb) == 0;
2318 static int cmp_icase(char a, char b)
2320 if (a == b)
2321 return 0;
2322 if (ignore_case)
2323 return toupper(a) - toupper(b);
2324 return a - b;
2328 * Given two normalized paths (a trailing slash is ok), if subdir is
2329 * outside dir, return -1. Otherwise return the offset in subdir that
2330 * can be used as relative path to dir.
2332 int dir_inside_of(const char *subdir, const char *dir)
2334 int offset = 0;
2336 assert(dir && subdir && *dir && *subdir);
2338 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2339 dir++;
2340 subdir++;
2341 offset++;
2344 /* hel[p]/me vs hel[l]/yeah */
2345 if (*dir && *subdir)
2346 return -1;
2348 if (!*subdir)
2349 return !*dir ? offset : -1; /* same dir */
2351 /* foo/[b]ar vs foo/[] */
2352 if (is_dir_sep(dir[-1]))
2353 return is_dir_sep(subdir[-1]) ? offset : -1;
2355 /* foo[/]bar vs foo[] */
2356 return is_dir_sep(*subdir) ? offset + 1 : -1;
2359 int is_inside_dir(const char *dir)
2361 char *cwd;
2362 int rc;
2364 if (!dir)
2365 return 0;
2367 cwd = xgetcwd();
2368 rc = (dir_inside_of(cwd, dir) >= 0);
2369 free(cwd);
2370 return rc;
2373 int is_empty_dir(const char *path)
2375 DIR *dir = opendir(path);
2376 struct dirent *e;
2377 int ret = 1;
2379 if (!dir)
2380 return 0;
2382 while ((e = readdir(dir)) != NULL)
2383 if (!is_dot_or_dotdot(e->d_name)) {
2384 ret = 0;
2385 break;
2388 closedir(dir);
2389 return ret;
2392 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2394 DIR *dir;
2395 struct dirent *e;
2396 int ret = 0, original_len = path->len, len, kept_down = 0;
2397 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2398 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2399 struct object_id submodule_head;
2401 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2402 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2403 /* Do not descend and nuke a nested git work tree. */
2404 if (kept_up)
2405 *kept_up = 1;
2406 return 0;
2409 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2410 dir = opendir(path->buf);
2411 if (!dir) {
2412 if (errno == ENOENT)
2413 return keep_toplevel ? -1 : 0;
2414 else if (errno == EACCES && !keep_toplevel)
2416 * An empty dir could be removable even if it
2417 * is unreadable:
2419 return rmdir(path->buf);
2420 else
2421 return -1;
2423 strbuf_complete(path, '/');
2425 len = path->len;
2426 while ((e = readdir(dir)) != NULL) {
2427 struct stat st;
2428 if (is_dot_or_dotdot(e->d_name))
2429 continue;
2431 strbuf_setlen(path, len);
2432 strbuf_addstr(path, e->d_name);
2433 if (lstat(path->buf, &st)) {
2434 if (errno == ENOENT)
2436 * file disappeared, which is what we
2437 * wanted anyway
2439 continue;
2440 /* fall thru */
2441 } else if (S_ISDIR(st.st_mode)) {
2442 if (!remove_dir_recurse(path, flag, &kept_down))
2443 continue; /* happy */
2444 } else if (!only_empty &&
2445 (!unlink(path->buf) || errno == ENOENT)) {
2446 continue; /* happy, too */
2449 /* path too long, stat fails, or non-directory still exists */
2450 ret = -1;
2451 break;
2453 closedir(dir);
2455 strbuf_setlen(path, original_len);
2456 if (!ret && !keep_toplevel && !kept_down)
2457 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2458 else if (kept_up)
2460 * report the uplevel that it is not an error that we
2461 * did not rmdir() our directory.
2463 *kept_up = !ret;
2464 return ret;
2467 int remove_dir_recursively(struct strbuf *path, int flag)
2469 return remove_dir_recurse(path, flag, NULL);
2472 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2474 void setup_standard_excludes(struct dir_struct *dir)
2476 dir->exclude_per_dir = ".gitignore";
2478 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2479 if (!excludes_file)
2480 excludes_file = xdg_config_home("ignore");
2481 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2482 add_excludes_from_file_1(dir, excludes_file,
2483 dir->untracked ? &dir->ss_excludes_file : NULL);
2485 /* per repository user preference */
2486 if (startup_info->have_repository) {
2487 const char *path = git_path_info_exclude();
2488 if (!access_or_warn(path, R_OK, 0))
2489 add_excludes_from_file_1(dir, path,
2490 dir->untracked ? &dir->ss_info_exclude : NULL);
2494 int remove_path(const char *name)
2496 char *slash;
2498 if (unlink(name) && !is_missing_file_error(errno))
2499 return -1;
2501 slash = strrchr(name, '/');
2502 if (slash) {
2503 char *dirs = xstrdup(name);
2504 slash = dirs + (slash - name);
2505 do {
2506 *slash = '\0';
2507 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2508 free(dirs);
2510 return 0;
2514 * Frees memory within dir which was allocated for exclude lists and
2515 * the exclude_stack. Does not free dir itself.
2517 void clear_directory(struct dir_struct *dir)
2519 int i, j;
2520 struct exclude_list_group *group;
2521 struct exclude_list *el;
2522 struct exclude_stack *stk;
2524 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2525 group = &dir->exclude_list_group[i];
2526 for (j = 0; j < group->nr; j++) {
2527 el = &group->el[j];
2528 if (i == EXC_DIRS)
2529 free((char *)el->src);
2530 clear_exclude_list(el);
2532 free(group->el);
2535 stk = dir->exclude_stack;
2536 while (stk) {
2537 struct exclude_stack *prev = stk->prev;
2538 free(stk);
2539 stk = prev;
2541 strbuf_release(&dir->basebuf);
2544 struct ondisk_untracked_cache {
2545 struct stat_data info_exclude_stat;
2546 struct stat_data excludes_file_stat;
2547 uint32_t dir_flags;
2548 unsigned char info_exclude_sha1[20];
2549 unsigned char excludes_file_sha1[20];
2550 char exclude_per_dir[FLEX_ARRAY];
2553 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2554 #define ouc_size(len) (ouc_offset(exclude_per_dir) + len + 1)
2556 struct write_data {
2557 int index; /* number of written untracked_cache_dir */
2558 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2559 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2560 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2561 struct strbuf out;
2562 struct strbuf sb_stat;
2563 struct strbuf sb_sha1;
2566 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2568 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2569 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2570 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2571 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2572 to->sd_dev = htonl(from->sd_dev);
2573 to->sd_ino = htonl(from->sd_ino);
2574 to->sd_uid = htonl(from->sd_uid);
2575 to->sd_gid = htonl(from->sd_gid);
2576 to->sd_size = htonl(from->sd_size);
2579 static void write_one_dir(struct untracked_cache_dir *untracked,
2580 struct write_data *wd)
2582 struct stat_data stat_data;
2583 struct strbuf *out = &wd->out;
2584 unsigned char intbuf[16];
2585 unsigned int intlen, value;
2586 int i = wd->index++;
2589 * untracked_nr should be reset whenever valid is clear, but
2590 * for safety..
2592 if (!untracked->valid) {
2593 untracked->untracked_nr = 0;
2594 untracked->check_only = 0;
2597 if (untracked->check_only)
2598 ewah_set(wd->check_only, i);
2599 if (untracked->valid) {
2600 ewah_set(wd->valid, i);
2601 stat_data_to_disk(&stat_data, &untracked->stat_data);
2602 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2604 if (!is_null_sha1(untracked->exclude_sha1)) {
2605 ewah_set(wd->sha1_valid, i);
2606 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2609 intlen = encode_varint(untracked->untracked_nr, intbuf);
2610 strbuf_add(out, intbuf, intlen);
2612 /* skip non-recurse directories */
2613 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2614 if (untracked->dirs[i]->recurse)
2615 value++;
2616 intlen = encode_varint(value, intbuf);
2617 strbuf_add(out, intbuf, intlen);
2619 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2621 for (i = 0; i < untracked->untracked_nr; i++)
2622 strbuf_add(out, untracked->untracked[i],
2623 strlen(untracked->untracked[i]) + 1);
2625 for (i = 0; i < untracked->dirs_nr; i++)
2626 if (untracked->dirs[i]->recurse)
2627 write_one_dir(untracked->dirs[i], wd);
2630 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2632 struct ondisk_untracked_cache *ouc;
2633 struct write_data wd;
2634 unsigned char varbuf[16];
2635 int varint_len;
2636 size_t len = strlen(untracked->exclude_per_dir);
2638 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
2639 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2640 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2641 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2642 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2643 ouc->dir_flags = htonl(untracked->dir_flags);
2645 varint_len = encode_varint(untracked->ident.len, varbuf);
2646 strbuf_add(out, varbuf, varint_len);
2647 strbuf_addbuf(out, &untracked->ident);
2649 strbuf_add(out, ouc, ouc_size(len));
2650 FREE_AND_NULL(ouc);
2652 if (!untracked->root) {
2653 varint_len = encode_varint(0, varbuf);
2654 strbuf_add(out, varbuf, varint_len);
2655 return;
2658 wd.index = 0;
2659 wd.check_only = ewah_new();
2660 wd.valid = ewah_new();
2661 wd.sha1_valid = ewah_new();
2662 strbuf_init(&wd.out, 1024);
2663 strbuf_init(&wd.sb_stat, 1024);
2664 strbuf_init(&wd.sb_sha1, 1024);
2665 write_one_dir(untracked->root, &wd);
2667 varint_len = encode_varint(wd.index, varbuf);
2668 strbuf_add(out, varbuf, varint_len);
2669 strbuf_addbuf(out, &wd.out);
2670 ewah_serialize_strbuf(wd.valid, out);
2671 ewah_serialize_strbuf(wd.check_only, out);
2672 ewah_serialize_strbuf(wd.sha1_valid, out);
2673 strbuf_addbuf(out, &wd.sb_stat);
2674 strbuf_addbuf(out, &wd.sb_sha1);
2675 strbuf_addch(out, '\0'); /* safe guard for string lists */
2677 ewah_free(wd.valid);
2678 ewah_free(wd.check_only);
2679 ewah_free(wd.sha1_valid);
2680 strbuf_release(&wd.out);
2681 strbuf_release(&wd.sb_stat);
2682 strbuf_release(&wd.sb_sha1);
2685 static void free_untracked(struct untracked_cache_dir *ucd)
2687 int i;
2688 if (!ucd)
2689 return;
2690 for (i = 0; i < ucd->dirs_nr; i++)
2691 free_untracked(ucd->dirs[i]);
2692 for (i = 0; i < ucd->untracked_nr; i++)
2693 free(ucd->untracked[i]);
2694 free(ucd->untracked);
2695 free(ucd->dirs);
2696 free(ucd);
2699 void free_untracked_cache(struct untracked_cache *uc)
2701 if (uc)
2702 free_untracked(uc->root);
2703 free(uc);
2706 struct read_data {
2707 int index;
2708 struct untracked_cache_dir **ucd;
2709 struct ewah_bitmap *check_only;
2710 struct ewah_bitmap *valid;
2711 struct ewah_bitmap *sha1_valid;
2712 const unsigned char *data;
2713 const unsigned char *end;
2716 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2718 memcpy(to, data, sizeof(*to));
2719 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2720 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2721 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2722 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2723 to->sd_dev = ntohl(to->sd_dev);
2724 to->sd_ino = ntohl(to->sd_ino);
2725 to->sd_uid = ntohl(to->sd_uid);
2726 to->sd_gid = ntohl(to->sd_gid);
2727 to->sd_size = ntohl(to->sd_size);
2730 static int read_one_dir(struct untracked_cache_dir **untracked_,
2731 struct read_data *rd)
2733 struct untracked_cache_dir ud, *untracked;
2734 const unsigned char *next, *data = rd->data, *end = rd->end;
2735 unsigned int value;
2736 int i, len;
2738 memset(&ud, 0, sizeof(ud));
2740 next = data;
2741 value = decode_varint(&next);
2742 if (next > end)
2743 return -1;
2744 ud.recurse = 1;
2745 ud.untracked_alloc = value;
2746 ud.untracked_nr = value;
2747 if (ud.untracked_nr)
2748 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2749 data = next;
2751 next = data;
2752 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2753 if (next > end)
2754 return -1;
2755 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2756 data = next;
2758 len = strlen((const char *)data);
2759 next = data + len + 1;
2760 if (next > rd->end)
2761 return -1;
2762 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
2763 memcpy(untracked, &ud, sizeof(ud));
2764 memcpy(untracked->name, data, len + 1);
2765 data = next;
2767 for (i = 0; i < untracked->untracked_nr; i++) {
2768 len = strlen((const char *)data);
2769 next = data + len + 1;
2770 if (next > rd->end)
2771 return -1;
2772 untracked->untracked[i] = xstrdup((const char*)data);
2773 data = next;
2776 rd->ucd[rd->index++] = untracked;
2777 rd->data = data;
2779 for (i = 0; i < untracked->dirs_nr; i++) {
2780 len = read_one_dir(untracked->dirs + i, rd);
2781 if (len < 0)
2782 return -1;
2784 return 0;
2787 static void set_check_only(size_t pos, void *cb)
2789 struct read_data *rd = cb;
2790 struct untracked_cache_dir *ud = rd->ucd[pos];
2791 ud->check_only = 1;
2794 static void read_stat(size_t pos, void *cb)
2796 struct read_data *rd = cb;
2797 struct untracked_cache_dir *ud = rd->ucd[pos];
2798 if (rd->data + sizeof(struct stat_data) > rd->end) {
2799 rd->data = rd->end + 1;
2800 return;
2802 stat_data_from_disk(&ud->stat_data, rd->data);
2803 rd->data += sizeof(struct stat_data);
2804 ud->valid = 1;
2807 static void read_sha1(size_t pos, void *cb)
2809 struct read_data *rd = cb;
2810 struct untracked_cache_dir *ud = rd->ucd[pos];
2811 if (rd->data + 20 > rd->end) {
2812 rd->data = rd->end + 1;
2813 return;
2815 hashcpy(ud->exclude_sha1, rd->data);
2816 rd->data += 20;
2819 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2820 const unsigned char *data,
2821 const unsigned char *sha1)
2823 stat_data_from_disk(&sha1_stat->stat, data);
2824 hashcpy(sha1_stat->sha1, sha1);
2825 sha1_stat->valid = 1;
2828 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2830 struct untracked_cache *uc;
2831 struct read_data rd;
2832 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2833 const char *ident;
2834 int ident_len, len;
2835 const char *exclude_per_dir;
2837 if (sz <= 1 || end[-1] != '\0')
2838 return NULL;
2839 end--;
2841 ident_len = decode_varint(&next);
2842 if (next + ident_len > end)
2843 return NULL;
2844 ident = (const char *)next;
2845 next += ident_len;
2847 if (next + ouc_size(0) > end)
2848 return NULL;
2850 uc = xcalloc(1, sizeof(*uc));
2851 strbuf_init(&uc->ident, ident_len);
2852 strbuf_add(&uc->ident, ident, ident_len);
2853 load_sha1_stat(&uc->ss_info_exclude,
2854 next + ouc_offset(info_exclude_stat),
2855 next + ouc_offset(info_exclude_sha1));
2856 load_sha1_stat(&uc->ss_excludes_file,
2857 next + ouc_offset(excludes_file_stat),
2858 next + ouc_offset(excludes_file_sha1));
2859 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2860 exclude_per_dir = (const char *)next + ouc_offset(exclude_per_dir);
2861 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2862 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2863 next += ouc_size(strlen(exclude_per_dir));
2864 if (next >= end)
2865 goto done2;
2867 len = decode_varint(&next);
2868 if (next > end || len == 0)
2869 goto done2;
2871 rd.valid = ewah_new();
2872 rd.check_only = ewah_new();
2873 rd.sha1_valid = ewah_new();
2874 rd.data = next;
2875 rd.end = end;
2876 rd.index = 0;
2877 ALLOC_ARRAY(rd.ucd, len);
2879 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2880 goto done;
2882 next = rd.data;
2883 len = ewah_read_mmap(rd.valid, next, end - next);
2884 if (len < 0)
2885 goto done;
2887 next += len;
2888 len = ewah_read_mmap(rd.check_only, next, end - next);
2889 if (len < 0)
2890 goto done;
2892 next += len;
2893 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2894 if (len < 0)
2895 goto done;
2897 ewah_each_bit(rd.check_only, set_check_only, &rd);
2898 rd.data = next + len;
2899 ewah_each_bit(rd.valid, read_stat, &rd);
2900 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2901 next = rd.data;
2903 done:
2904 free(rd.ucd);
2905 ewah_free(rd.valid);
2906 ewah_free(rd.check_only);
2907 ewah_free(rd.sha1_valid);
2908 done2:
2909 if (next != end) {
2910 free_untracked_cache(uc);
2911 uc = NULL;
2913 return uc;
2916 static void invalidate_one_directory(struct untracked_cache *uc,
2917 struct untracked_cache_dir *ucd)
2919 uc->dir_invalidated++;
2920 ucd->valid = 0;
2921 ucd->untracked_nr = 0;
2925 * Normally when an entry is added or removed from a directory,
2926 * invalidating that directory is enough. No need to touch its
2927 * ancestors. When a directory is shown as "foo/bar/" in git-status
2928 * however, deleting or adding an entry may have cascading effect.
2930 * Say the "foo/bar/file" has become untracked, we need to tell the
2931 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2932 * directory any more (because "bar" is managed by foo as an untracked
2933 * "file").
2935 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2936 * was the last untracked entry in the entire "foo", we should show
2937 * "foo/" instead. Which means we have to invalidate past "bar" up to
2938 * "foo".
2940 * This function traverses all directories from root to leaf. If there
2941 * is a chance of one of the above cases happening, we invalidate back
2942 * to root. Otherwise we just invalidate the leaf. There may be a more
2943 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2944 * detect these cases and avoid unnecessary invalidation, for example,
2945 * checking for the untracked entry named "bar/" in "foo", but for now
2946 * stick to something safe and simple.
2948 static int invalidate_one_component(struct untracked_cache *uc,
2949 struct untracked_cache_dir *dir,
2950 const char *path, int len)
2952 const char *rest = strchr(path, '/');
2954 if (rest) {
2955 int component_len = rest - path;
2956 struct untracked_cache_dir *d =
2957 lookup_untracked(uc, dir, path, component_len);
2958 int ret =
2959 invalidate_one_component(uc, d, rest + 1,
2960 len - (component_len + 1));
2961 if (ret)
2962 invalidate_one_directory(uc, dir);
2963 return ret;
2966 invalidate_one_directory(uc, dir);
2967 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2970 void untracked_cache_invalidate_path(struct index_state *istate,
2971 const char *path)
2973 if (!istate->untracked || !istate->untracked->root)
2974 return;
2975 invalidate_one_component(istate->untracked, istate->untracked->root,
2976 path, strlen(path));
2979 void untracked_cache_remove_from_index(struct index_state *istate,
2980 const char *path)
2982 untracked_cache_invalidate_path(istate, path);
2985 void untracked_cache_add_to_index(struct index_state *istate,
2986 const char *path)
2988 untracked_cache_invalidate_path(istate, path);
2991 /* Update gitfile and core.worktree setting to connect work tree and git dir */
2992 void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)
2994 struct strbuf gitfile_sb = STRBUF_INIT;
2995 struct strbuf cfg_sb = STRBUF_INIT;
2996 struct strbuf rel_path = STRBUF_INIT;
2997 char *git_dir, *work_tree;
2999 /* Prepare .git file */
3000 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3001 if (safe_create_leading_directories_const(gitfile_sb.buf))
3002 die(_("could not create directories for %s"), gitfile_sb.buf);
3004 /* Prepare config file */
3005 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3006 if (safe_create_leading_directories_const(cfg_sb.buf))
3007 die(_("could not create directories for %s"), cfg_sb.buf);
3009 git_dir = real_pathdup(git_dir_, 1);
3010 work_tree = real_pathdup(work_tree_, 1);
3012 /* Write .git file */
3013 write_file(gitfile_sb.buf, "gitdir: %s",
3014 relative_path(git_dir, work_tree, &rel_path));
3015 /* Update core.worktree setting */
3016 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3017 relative_path(work_tree, git_dir, &rel_path));
3019 strbuf_release(&gitfile_sb);
3020 strbuf_release(&cfg_sb);
3021 strbuf_release(&rel_path);
3022 free(work_tree);
3023 free(git_dir);
3027 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3029 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3031 if (rename(old_git_dir, new_git_dir) < 0)
3032 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3033 old_git_dir, new_git_dir);
3035 connect_work_tree_and_git_dir(path, new_git_dir);