Merge branch 'rs/strbuf-read-file-or-whine'
[alt-git.git] / dir.c
blob6dd91be8185d4012e11495a9238636913fedeeef
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 oid_stat with the given OID (when valid).
236 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
237 size_t *size_out, char **data_out)
239 enum object_type type;
240 unsigned long sz;
241 char *data;
243 *size_out = 0;
244 *data_out = NULL;
246 data = read_sha1_file(oid->hash, &type, &sz);
247 if (!data || type != OBJ_BLOB) {
248 free(data);
249 return -1;
252 if (oid_stat) {
253 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
254 oidcpy(&oid_stat->oid, oid);
257 if (sz == 0) {
258 free(data);
259 return 0;
262 if (data[sz - 1] != '\n') {
263 data = xrealloc(data, st_add(sz, 1));
264 data[sz++] = '\n';
267 *size_out = xsize_t(sz);
268 *data_out = data;
270 return 1;
273 #define DO_MATCH_EXCLUDE (1<<0)
274 #define DO_MATCH_DIRECTORY (1<<1)
275 #define DO_MATCH_SUBMODULE (1<<2)
277 static int match_attrs(const char *name, int namelen,
278 const struct pathspec_item *item)
280 int i;
282 git_check_attr(name, item->attr_check);
283 for (i = 0; i < item->attr_match_nr; i++) {
284 const char *value;
285 int matched;
286 enum attr_match_mode match_mode;
288 value = item->attr_check->items[i].value;
289 match_mode = item->attr_match[i].match_mode;
291 if (ATTR_TRUE(value))
292 matched = (match_mode == MATCH_SET);
293 else if (ATTR_FALSE(value))
294 matched = (match_mode == MATCH_UNSET);
295 else if (ATTR_UNSET(value))
296 matched = (match_mode == MATCH_UNSPECIFIED);
297 else
298 matched = (match_mode == MATCH_VALUE &&
299 !strcmp(item->attr_match[i].value, value));
300 if (!matched)
301 return 0;
304 return 1;
308 * Does 'match' match the given name?
309 * A match is found if
311 * (1) the 'match' string is leading directory of 'name', or
312 * (2) the 'match' string is a wildcard and matches 'name', or
313 * (3) the 'match' string is exactly the same as 'name'.
315 * and the return value tells which case it was.
317 * It returns 0 when there is no match.
319 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
320 const char *name, int namelen, unsigned flags)
322 /* name/namelen has prefix cut off by caller */
323 const char *match = item->match + prefix;
324 int matchlen = item->len - prefix;
327 * The normal call pattern is:
328 * 1. prefix = common_prefix_len(ps);
329 * 2. prune something, or fill_directory
330 * 3. match_pathspec()
332 * 'prefix' at #1 may be shorter than the command's prefix and
333 * it's ok for #2 to match extra files. Those extras will be
334 * trimmed at #3.
336 * Suppose the pathspec is 'foo' and '../bar' running from
337 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
338 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
339 * user does not want XYZ/foo, only the "foo" part should be
340 * case-insensitive. We need to filter out XYZ/foo here. In
341 * other words, we do not trust the caller on comparing the
342 * prefix part when :(icase) is involved. We do exact
343 * comparison ourselves.
345 * Normally the caller (common_prefix_len() in fact) does
346 * _exact_ matching on name[-prefix+1..-1] and we do not need
347 * to check that part. Be defensive and check it anyway, in
348 * case common_prefix_len is changed, or a new caller is
349 * introduced that does not use common_prefix_len.
351 * If the penalty turns out too high when prefix is really
352 * long, maybe change it to
353 * strncmp(match, name, item->prefix - prefix)
355 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
356 strncmp(item->match, name - prefix, item->prefix))
357 return 0;
359 if (item->attr_match_nr && !match_attrs(name, namelen, item))
360 return 0;
362 /* If the match was just the prefix, we matched */
363 if (!*match)
364 return MATCHED_RECURSIVELY;
366 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
367 if (matchlen == namelen)
368 return MATCHED_EXACTLY;
370 if (match[matchlen-1] == '/' || name[matchlen] == '/')
371 return MATCHED_RECURSIVELY;
372 } else if ((flags & DO_MATCH_DIRECTORY) &&
373 match[matchlen - 1] == '/' &&
374 namelen == matchlen - 1 &&
375 !ps_strncmp(item, match, name, namelen))
376 return MATCHED_EXACTLY;
378 if (item->nowildcard_len < item->len &&
379 !git_fnmatch(item, match, name,
380 item->nowildcard_len - prefix))
381 return MATCHED_FNMATCH;
383 /* Perform checks to see if "name" is a super set of the pathspec */
384 if (flags & DO_MATCH_SUBMODULE) {
385 /* name is a literal prefix of the pathspec */
386 if ((namelen < matchlen) &&
387 (match[namelen] == '/') &&
388 !ps_strncmp(item, match, name, namelen))
389 return MATCHED_RECURSIVELY;
391 /* name" doesn't match up to the first wild character */
392 if (item->nowildcard_len < item->len &&
393 ps_strncmp(item, match, name,
394 item->nowildcard_len - prefix))
395 return 0;
398 * Here is where we would perform a wildmatch to check if
399 * "name" can be matched as a directory (or a prefix) against
400 * the pathspec. Since wildmatch doesn't have this capability
401 * at the present we have to punt and say that it is a match,
402 * potentially returning a false positive
403 * The submodules themselves will be able to perform more
404 * accurate matching to determine if the pathspec matches.
406 return MATCHED_RECURSIVELY;
409 return 0;
413 * Given a name and a list of pathspecs, returns the nature of the
414 * closest (i.e. most specific) match of the name to any of the
415 * pathspecs.
417 * The caller typically calls this multiple times with the same
418 * pathspec and seen[] array but with different name/namelen
419 * (e.g. entries from the index) and is interested in seeing if and
420 * how each pathspec matches all the names it calls this function
421 * with. A mark is left in the seen[] array for each pathspec element
422 * indicating the closest type of match that element achieved, so if
423 * seen[n] remains zero after multiple invocations, that means the nth
424 * pathspec did not match any names, which could indicate that the
425 * user mistyped the nth pathspec.
427 static int do_match_pathspec(const struct pathspec *ps,
428 const char *name, int namelen,
429 int prefix, char *seen,
430 unsigned flags)
432 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
434 GUARD_PATHSPEC(ps,
435 PATHSPEC_FROMTOP |
436 PATHSPEC_MAXDEPTH |
437 PATHSPEC_LITERAL |
438 PATHSPEC_GLOB |
439 PATHSPEC_ICASE |
440 PATHSPEC_EXCLUDE |
441 PATHSPEC_ATTR);
443 if (!ps->nr) {
444 if (!ps->recursive ||
445 !(ps->magic & PATHSPEC_MAXDEPTH) ||
446 ps->max_depth == -1)
447 return MATCHED_RECURSIVELY;
449 if (within_depth(name, namelen, 0, ps->max_depth))
450 return MATCHED_EXACTLY;
451 else
452 return 0;
455 name += prefix;
456 namelen -= prefix;
458 for (i = ps->nr - 1; i >= 0; i--) {
459 int how;
461 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
462 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
463 continue;
465 if (seen && seen[i] == MATCHED_EXACTLY)
466 continue;
468 * Make exclude patterns optional and never report
469 * "pathspec ':(exclude)foo' matches no files"
471 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
472 seen[i] = MATCHED_FNMATCH;
473 how = match_pathspec_item(ps->items+i, prefix, name,
474 namelen, flags);
475 if (ps->recursive &&
476 (ps->magic & PATHSPEC_MAXDEPTH) &&
477 ps->max_depth != -1 &&
478 how && how != MATCHED_FNMATCH) {
479 int len = ps->items[i].len;
480 if (name[len] == '/')
481 len++;
482 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
483 how = MATCHED_EXACTLY;
484 else
485 how = 0;
487 if (how) {
488 if (retval < how)
489 retval = how;
490 if (seen && seen[i] < how)
491 seen[i] = how;
494 return retval;
497 int match_pathspec(const struct pathspec *ps,
498 const char *name, int namelen,
499 int prefix, char *seen, int is_dir)
501 int positive, negative;
502 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
503 positive = do_match_pathspec(ps, name, namelen,
504 prefix, seen, flags);
505 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
506 return positive;
507 negative = do_match_pathspec(ps, name, namelen,
508 prefix, seen,
509 flags | DO_MATCH_EXCLUDE);
510 return negative ? 0 : positive;
514 * Check if a submodule is a superset of the pathspec
516 int submodule_path_match(const struct pathspec *ps,
517 const char *submodule_name,
518 char *seen)
520 int matched = do_match_pathspec(ps, submodule_name,
521 strlen(submodule_name),
522 0, seen,
523 DO_MATCH_DIRECTORY |
524 DO_MATCH_SUBMODULE);
525 return matched;
528 int report_path_error(const char *ps_matched,
529 const struct pathspec *pathspec,
530 const char *prefix)
533 * Make sure all pathspec matched; otherwise it is an error.
535 int num, errors = 0;
536 for (num = 0; num < pathspec->nr; num++) {
537 int other, found_dup;
539 if (ps_matched[num])
540 continue;
542 * The caller might have fed identical pathspec
543 * twice. Do not barf on such a mistake.
544 * FIXME: parse_pathspec should have eliminated
545 * duplicate pathspec.
547 for (found_dup = other = 0;
548 !found_dup && other < pathspec->nr;
549 other++) {
550 if (other == num || !ps_matched[other])
551 continue;
552 if (!strcmp(pathspec->items[other].original,
553 pathspec->items[num].original))
555 * Ok, we have a match already.
557 found_dup = 1;
559 if (found_dup)
560 continue;
562 error("pathspec '%s' did not match any file(s) known to git.",
563 pathspec->items[num].original);
564 errors++;
566 return errors;
570 * Return the length of the "simple" part of a path match limiter.
572 int simple_length(const char *match)
574 int len = -1;
576 for (;;) {
577 unsigned char c = *match++;
578 len++;
579 if (c == '\0' || is_glob_special(c))
580 return len;
584 int no_wildcard(const char *string)
586 return string[simple_length(string)] == '\0';
589 void parse_exclude_pattern(const char **pattern,
590 int *patternlen,
591 unsigned *flags,
592 int *nowildcardlen)
594 const char *p = *pattern;
595 size_t i, len;
597 *flags = 0;
598 if (*p == '!') {
599 *flags |= EXC_FLAG_NEGATIVE;
600 p++;
602 len = strlen(p);
603 if (len && p[len - 1] == '/') {
604 len--;
605 *flags |= EXC_FLAG_MUSTBEDIR;
607 for (i = 0; i < len; i++) {
608 if (p[i] == '/')
609 break;
611 if (i == len)
612 *flags |= EXC_FLAG_NODIR;
613 *nowildcardlen = simple_length(p);
615 * we should have excluded the trailing slash from 'p' too,
616 * but that's one more allocation. Instead just make sure
617 * nowildcardlen does not exceed real patternlen
619 if (*nowildcardlen > len)
620 *nowildcardlen = len;
621 if (*p == '*' && no_wildcard(p + 1))
622 *flags |= EXC_FLAG_ENDSWITH;
623 *pattern = p;
624 *patternlen = len;
627 void add_exclude(const char *string, const char *base,
628 int baselen, struct exclude_list *el, int srcpos)
630 struct exclude *x;
631 int patternlen;
632 unsigned flags;
633 int nowildcardlen;
635 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
636 if (flags & EXC_FLAG_MUSTBEDIR) {
637 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
638 } else {
639 x = xmalloc(sizeof(*x));
640 x->pattern = string;
642 x->patternlen = patternlen;
643 x->nowildcardlen = nowildcardlen;
644 x->base = base;
645 x->baselen = baselen;
646 x->flags = flags;
647 x->srcpos = srcpos;
648 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
649 el->excludes[el->nr++] = x;
650 x->el = el;
653 static int read_skip_worktree_file_from_index(const struct index_state *istate,
654 const char *path,
655 size_t *size_out, char **data_out,
656 struct oid_stat *oid_stat)
658 int pos, len;
660 len = strlen(path);
661 pos = index_name_pos(istate, path, len);
662 if (pos < 0)
663 return -1;
664 if (!ce_skip_worktree(istate->cache[pos]))
665 return -1;
667 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
671 * Frees memory within el which was allocated for exclude patterns and
672 * the file buffer. Does not free el itself.
674 void clear_exclude_list(struct exclude_list *el)
676 int i;
678 for (i = 0; i < el->nr; i++)
679 free(el->excludes[i]);
680 free(el->excludes);
681 free(el->filebuf);
683 memset(el, 0, sizeof(*el));
686 static void trim_trailing_spaces(char *buf)
688 char *p, *last_space = NULL;
690 for (p = buf; *p; p++)
691 switch (*p) {
692 case ' ':
693 if (!last_space)
694 last_space = p;
695 break;
696 case '\\':
697 p++;
698 if (!*p)
699 return;
700 /* fallthrough */
701 default:
702 last_space = NULL;
705 if (last_space)
706 *last_space = '\0';
710 * Given a subdirectory name and "dir" of the current directory,
711 * search the subdir in "dir" and return it, or create a new one if it
712 * does not exist in "dir".
714 * If "name" has the trailing slash, it'll be excluded in the search.
716 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
717 struct untracked_cache_dir *dir,
718 const char *name, int len)
720 int first, last;
721 struct untracked_cache_dir *d;
722 if (!dir)
723 return NULL;
724 if (len && name[len - 1] == '/')
725 len--;
726 first = 0;
727 last = dir->dirs_nr;
728 while (last > first) {
729 int cmp, next = (last + first) >> 1;
730 d = dir->dirs[next];
731 cmp = strncmp(name, d->name, len);
732 if (!cmp && strlen(d->name) > len)
733 cmp = -1;
734 if (!cmp)
735 return d;
736 if (cmp < 0) {
737 last = next;
738 continue;
740 first = next+1;
743 uc->dir_created++;
744 FLEX_ALLOC_MEM(d, name, name, len);
746 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
747 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
748 dir->dirs_nr - first);
749 dir->dirs_nr++;
750 dir->dirs[first] = d;
751 return d;
754 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
756 int i;
757 dir->valid = 0;
758 dir->untracked_nr = 0;
759 for (i = 0; i < dir->dirs_nr; i++)
760 do_invalidate_gitignore(dir->dirs[i]);
763 static void invalidate_gitignore(struct untracked_cache *uc,
764 struct untracked_cache_dir *dir)
766 uc->gitignore_invalidated++;
767 do_invalidate_gitignore(dir);
770 static void invalidate_directory(struct untracked_cache *uc,
771 struct untracked_cache_dir *dir)
773 int i;
776 * Invalidation increment here is just roughly correct. If
777 * untracked_nr or any of dirs[].recurse is non-zero, we
778 * should increment dir_invalidated too. But that's more
779 * expensive to do.
781 if (dir->valid)
782 uc->dir_invalidated++;
784 dir->valid = 0;
785 dir->untracked_nr = 0;
786 for (i = 0; i < dir->dirs_nr; i++)
787 dir->dirs[i]->recurse = 0;
790 static int add_excludes_from_buffer(char *buf, size_t size,
791 const char *base, int baselen,
792 struct exclude_list *el);
795 * Given a file with name "fname", read it (either from disk, or from
796 * an index if 'istate' is non-null), parse it and store the
797 * exclude rules in "el".
799 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
800 * stat data from disk (only valid if add_excludes returns zero). If
801 * ss_valid is non-zero, "ss" must contain good value as input.
803 static int add_excludes(const char *fname, const char *base, int baselen,
804 struct exclude_list *el, struct index_state *istate,
805 struct oid_stat *oid_stat)
807 struct stat st;
808 int r;
809 int fd;
810 size_t size = 0;
811 char *buf;
813 fd = open(fname, O_RDONLY);
814 if (fd < 0 || fstat(fd, &st) < 0) {
815 if (fd < 0)
816 warn_on_fopen_errors(fname);
817 else
818 close(fd);
819 if (!istate)
820 return -1;
821 r = read_skip_worktree_file_from_index(istate, fname,
822 &size, &buf,
823 oid_stat);
824 if (r != 1)
825 return r;
826 } else {
827 size = xsize_t(st.st_size);
828 if (size == 0) {
829 if (oid_stat) {
830 fill_stat_data(&oid_stat->stat, &st);
831 oidcpy(&oid_stat->oid, &empty_blob_oid);
832 oid_stat->valid = 1;
834 close(fd);
835 return 0;
837 buf = xmallocz(size);
838 if (read_in_full(fd, buf, size) != size) {
839 free(buf);
840 close(fd);
841 return -1;
843 buf[size++] = '\n';
844 close(fd);
845 if (oid_stat) {
846 int pos;
847 if (oid_stat->valid &&
848 !match_stat_data_racy(istate, &oid_stat->stat, &st))
849 ; /* no content change, ss->sha1 still good */
850 else if (istate &&
851 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
852 !ce_stage(istate->cache[pos]) &&
853 ce_uptodate(istate->cache[pos]) &&
854 !would_convert_to_git(istate, fname))
855 oidcpy(&oid_stat->oid,
856 &istate->cache[pos]->oid);
857 else
858 hash_object_file(buf, size, "blob",
859 &oid_stat->oid);
860 fill_stat_data(&oid_stat->stat, &st);
861 oid_stat->valid = 1;
865 add_excludes_from_buffer(buf, size, base, baselen, el);
866 return 0;
869 static int add_excludes_from_buffer(char *buf, size_t size,
870 const char *base, int baselen,
871 struct exclude_list *el)
873 int i, lineno = 1;
874 char *entry;
876 el->filebuf = buf;
878 if (skip_utf8_bom(&buf, size))
879 size -= buf - el->filebuf;
881 entry = buf;
883 for (i = 0; i < size; i++) {
884 if (buf[i] == '\n') {
885 if (entry != buf + i && entry[0] != '#') {
886 buf[i - (i && buf[i-1] == '\r')] = 0;
887 trim_trailing_spaces(entry);
888 add_exclude(entry, base, baselen, el, lineno);
890 lineno++;
891 entry = buf + i + 1;
894 return 0;
897 int add_excludes_from_file_to_list(const char *fname, const char *base,
898 int baselen, struct exclude_list *el,
899 struct index_state *istate)
901 return add_excludes(fname, base, baselen, el, istate, NULL);
904 int add_excludes_from_blob_to_list(
905 struct object_id *oid,
906 const char *base, int baselen,
907 struct exclude_list *el)
909 char *buf;
910 size_t size;
911 int r;
913 r = do_read_blob(oid, NULL, &size, &buf);
914 if (r != 1)
915 return r;
917 add_excludes_from_buffer(buf, size, base, baselen, el);
918 return 0;
921 struct exclude_list *add_exclude_list(struct dir_struct *dir,
922 int group_type, const char *src)
924 struct exclude_list *el;
925 struct exclude_list_group *group;
927 group = &dir->exclude_list_group[group_type];
928 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
929 el = &group->el[group->nr++];
930 memset(el, 0, sizeof(*el));
931 el->src = src;
932 return el;
936 * Used to set up core.excludesfile and .git/info/exclude lists.
938 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
939 struct oid_stat *oid_stat)
941 struct exclude_list *el;
943 * catch setup_standard_excludes() that's called before
944 * dir->untracked is assigned. That function behaves
945 * differently when dir->untracked is non-NULL.
947 if (!dir->untracked)
948 dir->unmanaged_exclude_files++;
949 el = add_exclude_list(dir, EXC_FILE, fname);
950 if (add_excludes(fname, "", 0, el, NULL, oid_stat) < 0)
951 die("cannot use %s as an exclude file", fname);
954 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
956 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
957 add_excludes_from_file_1(dir, fname, NULL);
960 int match_basename(const char *basename, int basenamelen,
961 const char *pattern, int prefix, int patternlen,
962 unsigned flags)
964 if (prefix == patternlen) {
965 if (patternlen == basenamelen &&
966 !fspathncmp(pattern, basename, basenamelen))
967 return 1;
968 } else if (flags & EXC_FLAG_ENDSWITH) {
969 /* "*literal" matching against "fooliteral" */
970 if (patternlen - 1 <= basenamelen &&
971 !fspathncmp(pattern + 1,
972 basename + basenamelen - (patternlen - 1),
973 patternlen - 1))
974 return 1;
975 } else {
976 if (fnmatch_icase_mem(pattern, patternlen,
977 basename, basenamelen,
978 0) == 0)
979 return 1;
981 return 0;
984 int match_pathname(const char *pathname, int pathlen,
985 const char *base, int baselen,
986 const char *pattern, int prefix, int patternlen,
987 unsigned flags)
989 const char *name;
990 int namelen;
993 * match with FNM_PATHNAME; the pattern has base implicitly
994 * in front of it.
996 if (*pattern == '/') {
997 pattern++;
998 patternlen--;
999 prefix--;
1003 * baselen does not count the trailing slash. base[] may or
1004 * may not end with a trailing slash though.
1006 if (pathlen < baselen + 1 ||
1007 (baselen && pathname[baselen] != '/') ||
1008 fspathncmp(pathname, base, baselen))
1009 return 0;
1011 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1012 name = pathname + pathlen - namelen;
1014 if (prefix) {
1016 * if the non-wildcard part is longer than the
1017 * remaining pathname, surely it cannot match.
1019 if (prefix > namelen)
1020 return 0;
1022 if (fspathncmp(pattern, name, prefix))
1023 return 0;
1024 pattern += prefix;
1025 patternlen -= prefix;
1026 name += prefix;
1027 namelen -= prefix;
1030 * If the whole pattern did not have a wildcard,
1031 * then our prefix match is all we need; we
1032 * do not need to call fnmatch at all.
1034 if (!patternlen && !namelen)
1035 return 1;
1038 return fnmatch_icase_mem(pattern, patternlen,
1039 name, namelen,
1040 WM_PATHNAME) == 0;
1044 * Scan the given exclude list in reverse to see whether pathname
1045 * should be ignored. The first match (i.e. the last on the list), if
1046 * any, determines the fate. Returns the exclude_list element which
1047 * matched, or NULL for undecided.
1049 static struct exclude *last_exclude_matching_from_list(const char *pathname,
1050 int pathlen,
1051 const char *basename,
1052 int *dtype,
1053 struct exclude_list *el,
1054 struct index_state *istate)
1056 struct exclude *exc = NULL; /* undecided */
1057 int i;
1059 if (!el->nr)
1060 return NULL; /* undefined */
1062 for (i = el->nr - 1; 0 <= i; i--) {
1063 struct exclude *x = el->excludes[i];
1064 const char *exclude = x->pattern;
1065 int prefix = x->nowildcardlen;
1067 if (x->flags & EXC_FLAG_MUSTBEDIR) {
1068 if (*dtype == DT_UNKNOWN)
1069 *dtype = get_dtype(NULL, istate, pathname, pathlen);
1070 if (*dtype != DT_DIR)
1071 continue;
1074 if (x->flags & EXC_FLAG_NODIR) {
1075 if (match_basename(basename,
1076 pathlen - (basename - pathname),
1077 exclude, prefix, x->patternlen,
1078 x->flags)) {
1079 exc = x;
1080 break;
1082 continue;
1085 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1086 if (match_pathname(pathname, pathlen,
1087 x->base, x->baselen ? x->baselen - 1 : 0,
1088 exclude, prefix, x->patternlen, x->flags)) {
1089 exc = x;
1090 break;
1093 return exc;
1097 * Scan the list and let the last match determine the fate.
1098 * Return 1 for exclude, 0 for include and -1 for undecided.
1100 int is_excluded_from_list(const char *pathname,
1101 int pathlen, const char *basename, int *dtype,
1102 struct exclude_list *el, struct index_state *istate)
1104 struct exclude *exclude;
1105 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1106 dtype, el, istate);
1107 if (exclude)
1108 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1109 return -1; /* undecided */
1112 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1113 struct index_state *istate,
1114 const char *pathname, int pathlen, const char *basename,
1115 int *dtype_p)
1117 int i, j;
1118 struct exclude_list_group *group;
1119 struct exclude *exclude;
1120 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1121 group = &dir->exclude_list_group[i];
1122 for (j = group->nr - 1; j >= 0; j--) {
1123 exclude = last_exclude_matching_from_list(
1124 pathname, pathlen, basename, dtype_p,
1125 &group->el[j], istate);
1126 if (exclude)
1127 return exclude;
1130 return NULL;
1134 * Loads the per-directory exclude list for the substring of base
1135 * which has a char length of baselen.
1137 static void prep_exclude(struct dir_struct *dir,
1138 struct index_state *istate,
1139 const char *base, int baselen)
1141 struct exclude_list_group *group;
1142 struct exclude_list *el;
1143 struct exclude_stack *stk = NULL;
1144 struct untracked_cache_dir *untracked;
1145 int current;
1147 group = &dir->exclude_list_group[EXC_DIRS];
1150 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1151 * which originate from directories not in the prefix of the
1152 * path being checked.
1154 while ((stk = dir->exclude_stack) != NULL) {
1155 if (stk->baselen <= baselen &&
1156 !strncmp(dir->basebuf.buf, base, stk->baselen))
1157 break;
1158 el = &group->el[dir->exclude_stack->exclude_ix];
1159 dir->exclude_stack = stk->prev;
1160 dir->exclude = NULL;
1161 free((char *)el->src); /* see strbuf_detach() below */
1162 clear_exclude_list(el);
1163 free(stk);
1164 group->nr--;
1167 /* Skip traversing into sub directories if the parent is excluded */
1168 if (dir->exclude)
1169 return;
1172 * Lazy initialization. All call sites currently just
1173 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1174 * them seems lots of work for little benefit.
1176 if (!dir->basebuf.buf)
1177 strbuf_init(&dir->basebuf, PATH_MAX);
1179 /* Read from the parent directories and push them down. */
1180 current = stk ? stk->baselen : -1;
1181 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1182 if (dir->untracked)
1183 untracked = stk ? stk->ucd : dir->untracked->root;
1184 else
1185 untracked = NULL;
1187 while (current < baselen) {
1188 const char *cp;
1189 struct oid_stat oid_stat;
1191 stk = xcalloc(1, sizeof(*stk));
1192 if (current < 0) {
1193 cp = base;
1194 current = 0;
1195 } else {
1196 cp = strchr(base + current + 1, '/');
1197 if (!cp)
1198 die("oops in prep_exclude");
1199 cp++;
1200 untracked =
1201 lookup_untracked(dir->untracked, untracked,
1202 base + current,
1203 cp - base - current);
1205 stk->prev = dir->exclude_stack;
1206 stk->baselen = cp - base;
1207 stk->exclude_ix = group->nr;
1208 stk->ucd = untracked;
1209 el = add_exclude_list(dir, EXC_DIRS, NULL);
1210 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1211 assert(stk->baselen == dir->basebuf.len);
1213 /* Abort if the directory is excluded */
1214 if (stk->baselen) {
1215 int dt = DT_DIR;
1216 dir->basebuf.buf[stk->baselen - 1] = 0;
1217 dir->exclude = last_exclude_matching_from_lists(dir,
1218 istate,
1219 dir->basebuf.buf, stk->baselen - 1,
1220 dir->basebuf.buf + current, &dt);
1221 dir->basebuf.buf[stk->baselen - 1] = '/';
1222 if (dir->exclude &&
1223 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1224 dir->exclude = NULL;
1225 if (dir->exclude) {
1226 dir->exclude_stack = stk;
1227 return;
1231 /* Try to read per-directory file */
1232 oidclr(&oid_stat.oid);
1233 oid_stat.valid = 0;
1234 if (dir->exclude_per_dir &&
1236 * If we know that no files have been added in
1237 * this directory (i.e. valid_cached_dir() has
1238 * been executed and set untracked->valid) ..
1240 (!untracked || !untracked->valid ||
1242 * .. and .gitignore does not exist before
1243 * (i.e. null exclude_sha1). Then we can skip
1244 * loading .gitignore, which would result in
1245 * ENOENT anyway.
1247 !is_null_sha1(untracked->exclude_sha1))) {
1249 * dir->basebuf gets reused by the traversal, but we
1250 * need fname to remain unchanged to ensure the src
1251 * member of each struct exclude correctly
1252 * back-references its source file. Other invocations
1253 * of add_exclude_list provide stable strings, so we
1254 * strbuf_detach() and free() here in the caller.
1256 struct strbuf sb = STRBUF_INIT;
1257 strbuf_addbuf(&sb, &dir->basebuf);
1258 strbuf_addstr(&sb, dir->exclude_per_dir);
1259 el->src = strbuf_detach(&sb, NULL);
1260 add_excludes(el->src, el->src, stk->baselen, el, istate,
1261 untracked ? &oid_stat : NULL);
1264 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1265 * will first be called in valid_cached_dir() then maybe many
1266 * times more in last_exclude_matching(). When the cache is
1267 * used, last_exclude_matching() will not be called and
1268 * reading .gitignore content will be a waste.
1270 * So when it's called by valid_cached_dir() and we can get
1271 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1272 * modified on work tree), we could delay reading the
1273 * .gitignore content until we absolutely need it in
1274 * last_exclude_matching(). Be careful about ignore rule
1275 * order, though, if you do that.
1277 if (untracked &&
1278 hashcmp(oid_stat.oid.hash, untracked->exclude_sha1)) {
1279 invalidate_gitignore(dir->untracked, untracked);
1280 hashcpy(untracked->exclude_sha1, oid_stat.oid.hash);
1282 dir->exclude_stack = stk;
1283 current = stk->baselen;
1285 strbuf_setlen(&dir->basebuf, baselen);
1289 * Loads the exclude lists for the directory containing pathname, then
1290 * scans all exclude lists to determine whether pathname is excluded.
1291 * Returns the exclude_list element which matched, or NULL for
1292 * undecided.
1294 struct exclude *last_exclude_matching(struct dir_struct *dir,
1295 struct index_state *istate,
1296 const char *pathname,
1297 int *dtype_p)
1299 int pathlen = strlen(pathname);
1300 const char *basename = strrchr(pathname, '/');
1301 basename = (basename) ? basename+1 : pathname;
1303 prep_exclude(dir, istate, pathname, basename-pathname);
1305 if (dir->exclude)
1306 return dir->exclude;
1308 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
1309 basename, dtype_p);
1313 * Loads the exclude lists for the directory containing pathname, then
1314 * scans all exclude lists to determine whether pathname is excluded.
1315 * Returns 1 if true, otherwise 0.
1317 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1318 const char *pathname, int *dtype_p)
1320 struct exclude *exclude =
1321 last_exclude_matching(dir, istate, pathname, dtype_p);
1322 if (exclude)
1323 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1324 return 0;
1327 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1329 struct dir_entry *ent;
1331 FLEX_ALLOC_MEM(ent, name, pathname, len);
1332 ent->len = len;
1333 return ent;
1336 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1337 struct index_state *istate,
1338 const char *pathname, int len)
1340 if (index_file_exists(istate, pathname, len, ignore_case))
1341 return NULL;
1343 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1344 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1347 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1348 struct index_state *istate,
1349 const char *pathname, int len)
1351 if (!index_name_is_other(istate, pathname, len))
1352 return NULL;
1354 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1355 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1358 enum exist_status {
1359 index_nonexistent = 0,
1360 index_directory,
1361 index_gitdir
1365 * Do not use the alphabetically sorted index to look up
1366 * the directory name; instead, use the case insensitive
1367 * directory hash.
1369 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1370 const char *dirname, int len)
1372 struct cache_entry *ce;
1374 if (index_dir_exists(istate, dirname, len))
1375 return index_directory;
1377 ce = index_file_exists(istate, dirname, len, ignore_case);
1378 if (ce && S_ISGITLINK(ce->ce_mode))
1379 return index_gitdir;
1381 return index_nonexistent;
1385 * The index sorts alphabetically by entry name, which
1386 * means that a gitlink sorts as '\0' at the end, while
1387 * a directory (which is defined not as an entry, but as
1388 * the files it contains) will sort with the '/' at the
1389 * end.
1391 static enum exist_status directory_exists_in_index(struct index_state *istate,
1392 const char *dirname, int len)
1394 int pos;
1396 if (ignore_case)
1397 return directory_exists_in_index_icase(istate, dirname, len);
1399 pos = index_name_pos(istate, dirname, len);
1400 if (pos < 0)
1401 pos = -pos-1;
1402 while (pos < istate->cache_nr) {
1403 const struct cache_entry *ce = istate->cache[pos++];
1404 unsigned char endchar;
1406 if (strncmp(ce->name, dirname, len))
1407 break;
1408 endchar = ce->name[len];
1409 if (endchar > '/')
1410 break;
1411 if (endchar == '/')
1412 return index_directory;
1413 if (!endchar && S_ISGITLINK(ce->ce_mode))
1414 return index_gitdir;
1416 return index_nonexistent;
1420 * When we find a directory when traversing the filesystem, we
1421 * have three distinct cases:
1423 * - ignore it
1424 * - see it as a directory
1425 * - recurse into it
1427 * and which one we choose depends on a combination of existing
1428 * git index contents and the flags passed into the directory
1429 * traversal routine.
1431 * Case 1: If we *already* have entries in the index under that
1432 * directory name, we always recurse into the directory to see
1433 * all the files.
1435 * Case 2: If we *already* have that directory name as a gitlink,
1436 * we always continue to see it as a gitlink, regardless of whether
1437 * there is an actual git directory there or not (it might not
1438 * be checked out as a subproject!)
1440 * Case 3: if we didn't have it in the index previously, we
1441 * have a few sub-cases:
1443 * (a) if "show_other_directories" is true, we show it as
1444 * just a directory, unless "hide_empty_directories" is
1445 * also true, in which case we need to check if it contains any
1446 * untracked and / or ignored files.
1447 * (b) if it looks like a git directory, and we don't have
1448 * 'no_gitlinks' set we treat it as a gitlink, and show it
1449 * as a directory.
1450 * (c) otherwise, we recurse into it.
1452 static enum path_treatment treat_directory(struct dir_struct *dir,
1453 struct index_state *istate,
1454 struct untracked_cache_dir *untracked,
1455 const char *dirname, int len, int baselen, int exclude,
1456 const struct pathspec *pathspec)
1458 /* The "len-1" is to strip the final '/' */
1459 switch (directory_exists_in_index(istate, dirname, len-1)) {
1460 case index_directory:
1461 return path_recurse;
1463 case index_gitdir:
1464 return path_none;
1466 case index_nonexistent:
1467 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1468 break;
1469 if (exclude &&
1470 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1471 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1474 * This is an excluded directory and we are
1475 * showing ignored paths that match an exclude
1476 * pattern. (e.g. show directory as ignored
1477 * only if it matches an exclude pattern).
1478 * This path will either be 'path_excluded`
1479 * (if we are showing empty directories or if
1480 * the directory is not empty), or will be
1481 * 'path_none' (empty directory, and we are
1482 * not showing empty directories).
1484 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1485 return path_excluded;
1487 if (read_directory_recursive(dir, istate, dirname, len,
1488 untracked, 1, 1, pathspec) == path_excluded)
1489 return path_excluded;
1491 return path_none;
1493 if (!(dir->flags & DIR_NO_GITLINKS)) {
1494 struct object_id oid;
1495 if (resolve_gitlink_ref(dirname, "HEAD", &oid) == 0)
1496 return exclude ? path_excluded : path_untracked;
1498 return path_recurse;
1501 /* This is the "show_other_directories" case */
1503 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1504 return exclude ? path_excluded : path_untracked;
1506 untracked = lookup_untracked(dir->untracked, untracked,
1507 dirname + baselen, len - baselen);
1510 * If this is an excluded directory, then we only need to check if
1511 * the directory contains any files.
1513 return read_directory_recursive(dir, istate, dirname, len,
1514 untracked, 1, exclude, pathspec);
1518 * This is an inexact early pruning of any recursive directory
1519 * reading - if the path cannot possibly be in the pathspec,
1520 * return true, and we'll skip it early.
1522 static int simplify_away(const char *path, int pathlen,
1523 const struct pathspec *pathspec)
1525 int i;
1527 if (!pathspec || !pathspec->nr)
1528 return 0;
1530 GUARD_PATHSPEC(pathspec,
1531 PATHSPEC_FROMTOP |
1532 PATHSPEC_MAXDEPTH |
1533 PATHSPEC_LITERAL |
1534 PATHSPEC_GLOB |
1535 PATHSPEC_ICASE |
1536 PATHSPEC_EXCLUDE |
1537 PATHSPEC_ATTR);
1539 for (i = 0; i < pathspec->nr; i++) {
1540 const struct pathspec_item *item = &pathspec->items[i];
1541 int len = item->nowildcard_len;
1543 if (len > pathlen)
1544 len = pathlen;
1545 if (!ps_strncmp(item, item->match, path, len))
1546 return 0;
1549 return 1;
1553 * This function tells us whether an excluded path matches a
1554 * list of "interesting" pathspecs. That is, whether a path matched
1555 * by any of the pathspecs could possibly be ignored by excluding
1556 * the specified path. This can happen if:
1558 * 1. the path is mentioned explicitly in the pathspec
1560 * 2. the path is a directory prefix of some element in the
1561 * pathspec
1563 static int exclude_matches_pathspec(const char *path, int pathlen,
1564 const struct pathspec *pathspec)
1566 int i;
1568 if (!pathspec || !pathspec->nr)
1569 return 0;
1571 GUARD_PATHSPEC(pathspec,
1572 PATHSPEC_FROMTOP |
1573 PATHSPEC_MAXDEPTH |
1574 PATHSPEC_LITERAL |
1575 PATHSPEC_GLOB |
1576 PATHSPEC_ICASE |
1577 PATHSPEC_EXCLUDE);
1579 for (i = 0; i < pathspec->nr; i++) {
1580 const struct pathspec_item *item = &pathspec->items[i];
1581 int len = item->nowildcard_len;
1583 if (len == pathlen &&
1584 !ps_strncmp(item, item->match, path, pathlen))
1585 return 1;
1586 if (len > pathlen &&
1587 item->match[pathlen] == '/' &&
1588 !ps_strncmp(item, item->match, path, pathlen))
1589 return 1;
1591 return 0;
1594 static int get_index_dtype(struct index_state *istate,
1595 const char *path, int len)
1597 int pos;
1598 const struct cache_entry *ce;
1600 ce = index_file_exists(istate, path, len, 0);
1601 if (ce) {
1602 if (!ce_uptodate(ce))
1603 return DT_UNKNOWN;
1604 if (S_ISGITLINK(ce->ce_mode))
1605 return DT_DIR;
1607 * Nobody actually cares about the
1608 * difference between DT_LNK and DT_REG
1610 return DT_REG;
1613 /* Try to look it up as a directory */
1614 pos = index_name_pos(istate, path, len);
1615 if (pos >= 0)
1616 return DT_UNKNOWN;
1617 pos = -pos-1;
1618 while (pos < istate->cache_nr) {
1619 ce = istate->cache[pos++];
1620 if (strncmp(ce->name, path, len))
1621 break;
1622 if (ce->name[len] > '/')
1623 break;
1624 if (ce->name[len] < '/')
1625 continue;
1626 if (!ce_uptodate(ce))
1627 break; /* continue? */
1628 return DT_DIR;
1630 return DT_UNKNOWN;
1633 static int get_dtype(struct dirent *de, struct index_state *istate,
1634 const char *path, int len)
1636 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1637 struct stat st;
1639 if (dtype != DT_UNKNOWN)
1640 return dtype;
1641 dtype = get_index_dtype(istate, path, len);
1642 if (dtype != DT_UNKNOWN)
1643 return dtype;
1644 if (lstat(path, &st))
1645 return dtype;
1646 if (S_ISREG(st.st_mode))
1647 return DT_REG;
1648 if (S_ISDIR(st.st_mode))
1649 return DT_DIR;
1650 if (S_ISLNK(st.st_mode))
1651 return DT_LNK;
1652 return dtype;
1655 static enum path_treatment treat_one_path(struct dir_struct *dir,
1656 struct untracked_cache_dir *untracked,
1657 struct index_state *istate,
1658 struct strbuf *path,
1659 int baselen,
1660 const struct pathspec *pathspec,
1661 int dtype, struct dirent *de)
1663 int exclude;
1664 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1665 enum path_treatment path_treatment;
1667 if (dtype == DT_UNKNOWN)
1668 dtype = get_dtype(de, istate, path->buf, path->len);
1670 /* Always exclude indexed files */
1671 if (dtype != DT_DIR && has_path_in_index)
1672 return path_none;
1675 * When we are looking at a directory P in the working tree,
1676 * there are three cases:
1678 * (1) P exists in the index. Everything inside the directory P in
1679 * the working tree needs to go when P is checked out from the
1680 * index.
1682 * (2) P does not exist in the index, but there is P/Q in the index.
1683 * We know P will stay a directory when we check out the contents
1684 * of the index, but we do not know yet if there is a directory
1685 * P/Q in the working tree to be killed, so we need to recurse.
1687 * (3) P does not exist in the index, and there is no P/Q in the index
1688 * to require P to be a directory, either. Only in this case, we
1689 * know that everything inside P will not be killed without
1690 * recursing.
1692 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1693 (dtype == DT_DIR) &&
1694 !has_path_in_index &&
1695 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1696 return path_none;
1698 exclude = is_excluded(dir, istate, path->buf, &dtype);
1701 * Excluded? If we don't explicitly want to show
1702 * ignored files, ignore it
1704 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1705 return path_excluded;
1707 switch (dtype) {
1708 default:
1709 return path_none;
1710 case DT_DIR:
1711 strbuf_addch(path, '/');
1712 path_treatment = treat_directory(dir, istate, untracked,
1713 path->buf, path->len,
1714 baselen, exclude, pathspec);
1716 * If 1) we only want to return directories that
1717 * match an exclude pattern and 2) this directory does
1718 * not match an exclude pattern but all of its
1719 * contents are excluded, then indicate that we should
1720 * recurse into this directory (instead of marking the
1721 * directory itself as an ignored path).
1723 if (!exclude &&
1724 path_treatment == path_excluded &&
1725 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1726 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1727 return path_recurse;
1728 return path_treatment;
1729 case DT_REG:
1730 case DT_LNK:
1731 return exclude ? path_excluded : path_untracked;
1735 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1736 struct untracked_cache_dir *untracked,
1737 struct cached_dir *cdir,
1738 struct index_state *istate,
1739 struct strbuf *path,
1740 int baselen,
1741 const struct pathspec *pathspec)
1743 strbuf_setlen(path, baselen);
1744 if (!cdir->ucd) {
1745 strbuf_addstr(path, cdir->file);
1746 return path_untracked;
1748 strbuf_addstr(path, cdir->ucd->name);
1749 /* treat_one_path() does this before it calls treat_directory() */
1750 strbuf_complete(path, '/');
1751 if (cdir->ucd->check_only)
1753 * check_only is set as a result of treat_directory() getting
1754 * to its bottom. Verify again the same set of directories
1755 * with check_only set.
1757 return read_directory_recursive(dir, istate, path->buf, path->len,
1758 cdir->ucd, 1, 0, pathspec);
1760 * We get path_recurse in the first run when
1761 * directory_exists_in_index() returns index_nonexistent. We
1762 * are sure that new changes in the index does not impact the
1763 * outcome. Return now.
1765 return path_recurse;
1768 static enum path_treatment treat_path(struct dir_struct *dir,
1769 struct untracked_cache_dir *untracked,
1770 struct cached_dir *cdir,
1771 struct index_state *istate,
1772 struct strbuf *path,
1773 int baselen,
1774 const struct pathspec *pathspec)
1776 int dtype;
1777 struct dirent *de = cdir->de;
1779 if (!de)
1780 return treat_path_fast(dir, untracked, cdir, istate, path,
1781 baselen, pathspec);
1782 if (is_dot_or_dotdot(de->d_name) || !fspathcmp(de->d_name, ".git"))
1783 return path_none;
1784 strbuf_setlen(path, baselen);
1785 strbuf_addstr(path, de->d_name);
1786 if (simplify_away(path->buf, path->len, pathspec))
1787 return path_none;
1789 dtype = DTYPE(de);
1790 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1793 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1795 if (!dir)
1796 return;
1797 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1798 dir->untracked_alloc);
1799 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1802 static int valid_cached_dir(struct dir_struct *dir,
1803 struct untracked_cache_dir *untracked,
1804 struct index_state *istate,
1805 struct strbuf *path,
1806 int check_only)
1808 struct stat st;
1810 if (!untracked)
1811 return 0;
1814 * With fsmonitor, we can trust the untracked cache's valid field.
1816 refresh_fsmonitor(istate);
1817 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1818 if (lstat(path->len ? path->buf : ".", &st)) {
1819 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1820 return 0;
1822 if (!untracked->valid ||
1823 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1824 fill_stat_data(&untracked->stat_data, &st);
1825 return 0;
1829 if (untracked->check_only != !!check_only)
1830 return 0;
1833 * prep_exclude will be called eventually on this directory,
1834 * but it's called much later in last_exclude_matching(). We
1835 * need it now to determine the validity of the cache for this
1836 * path. The next calls will be nearly no-op, the way
1837 * prep_exclude() is designed.
1839 if (path->len && path->buf[path->len - 1] != '/') {
1840 strbuf_addch(path, '/');
1841 prep_exclude(dir, istate, path->buf, path->len);
1842 strbuf_setlen(path, path->len - 1);
1843 } else
1844 prep_exclude(dir, istate, path->buf, path->len);
1846 /* hopefully prep_exclude() haven't invalidated this entry... */
1847 return untracked->valid;
1850 static int open_cached_dir(struct cached_dir *cdir,
1851 struct dir_struct *dir,
1852 struct untracked_cache_dir *untracked,
1853 struct index_state *istate,
1854 struct strbuf *path,
1855 int check_only)
1857 const char *c_path;
1859 memset(cdir, 0, sizeof(*cdir));
1860 cdir->untracked = untracked;
1861 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1862 return 0;
1863 c_path = path->len ? path->buf : ".";
1864 cdir->fdir = opendir(c_path);
1865 if (!cdir->fdir)
1866 warning_errno(_("could not open directory '%s'"), c_path);
1867 if (dir->untracked) {
1868 invalidate_directory(dir->untracked, untracked);
1869 dir->untracked->dir_opened++;
1871 if (!cdir->fdir)
1872 return -1;
1873 return 0;
1876 static int read_cached_dir(struct cached_dir *cdir)
1878 if (cdir->fdir) {
1879 cdir->de = readdir(cdir->fdir);
1880 if (!cdir->de)
1881 return -1;
1882 return 0;
1884 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1885 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1886 if (!d->recurse) {
1887 cdir->nr_dirs++;
1888 continue;
1890 cdir->ucd = d;
1891 cdir->nr_dirs++;
1892 return 0;
1894 cdir->ucd = NULL;
1895 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1896 struct untracked_cache_dir *d = cdir->untracked;
1897 cdir->file = d->untracked[cdir->nr_files++];
1898 return 0;
1900 return -1;
1903 static void close_cached_dir(struct cached_dir *cdir)
1905 if (cdir->fdir)
1906 closedir(cdir->fdir);
1908 * We have gone through this directory and found no untracked
1909 * entries. Mark it valid.
1911 if (cdir->untracked) {
1912 cdir->untracked->valid = 1;
1913 cdir->untracked->recurse = 1;
1918 * Read a directory tree. We currently ignore anything but
1919 * directories, regular files and symlinks. That's because git
1920 * doesn't handle them at all yet. Maybe that will change some
1921 * day.
1923 * Also, we ignore the name ".git" (even if it is not a directory).
1924 * That likely will not change.
1926 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1927 * to signal that a file was found. This is the least significant value that
1928 * indicates that a file was encountered that does not depend on the order of
1929 * whether an untracked or exluded path was encountered first.
1931 * Returns the most significant path_treatment value encountered in the scan.
1932 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1933 * significant path_treatment value that will be returned.
1936 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1937 struct index_state *istate, const char *base, int baselen,
1938 struct untracked_cache_dir *untracked, int check_only,
1939 int stop_at_first_file, const struct pathspec *pathspec)
1941 struct cached_dir cdir;
1942 enum path_treatment state, subdir_state, dir_state = path_none;
1943 struct strbuf path = STRBUF_INIT;
1945 strbuf_add(&path, base, baselen);
1947 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1948 goto out;
1950 if (untracked)
1951 untracked->check_only = !!check_only;
1953 while (!read_cached_dir(&cdir)) {
1954 /* check how the file or directory should be treated */
1955 state = treat_path(dir, untracked, &cdir, istate, &path,
1956 baselen, pathspec);
1958 if (state > dir_state)
1959 dir_state = state;
1961 /* recurse into subdir if instructed by treat_path */
1962 if ((state == path_recurse) ||
1963 ((state == path_untracked) &&
1964 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1965 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
1966 struct untracked_cache_dir *ud;
1967 ud = lookup_untracked(dir->untracked, untracked,
1968 path.buf + baselen,
1969 path.len - baselen);
1970 subdir_state =
1971 read_directory_recursive(dir, istate, path.buf,
1972 path.len, ud,
1973 check_only, stop_at_first_file, pathspec);
1974 if (subdir_state > dir_state)
1975 dir_state = subdir_state;
1978 if (check_only) {
1979 if (stop_at_first_file) {
1981 * If stopping at first file, then
1982 * signal that a file was found by
1983 * returning `path_excluded`. This is
1984 * to return a consistent value
1985 * regardless of whether an ignored or
1986 * excluded file happened to be
1987 * encountered 1st.
1989 * In current usage, the
1990 * `stop_at_first_file` is passed when
1991 * an ancestor directory has matched
1992 * an exclude pattern, so any found
1993 * files will be excluded.
1995 if (dir_state >= path_excluded) {
1996 dir_state = path_excluded;
1997 break;
2001 /* abort early if maximum state has been reached */
2002 if (dir_state == path_untracked) {
2003 if (cdir.fdir)
2004 add_untracked(untracked, path.buf + baselen);
2005 break;
2007 /* skip the dir_add_* part */
2008 continue;
2011 /* add the path to the appropriate result list */
2012 switch (state) {
2013 case path_excluded:
2014 if (dir->flags & DIR_SHOW_IGNORED)
2015 dir_add_name(dir, istate, path.buf, path.len);
2016 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2017 ((dir->flags & DIR_COLLECT_IGNORED) &&
2018 exclude_matches_pathspec(path.buf, path.len,
2019 pathspec)))
2020 dir_add_ignored(dir, istate, path.buf, path.len);
2021 break;
2023 case path_untracked:
2024 if (dir->flags & DIR_SHOW_IGNORED)
2025 break;
2026 dir_add_name(dir, istate, path.buf, path.len);
2027 if (cdir.fdir)
2028 add_untracked(untracked, path.buf + baselen);
2029 break;
2031 default:
2032 break;
2035 close_cached_dir(&cdir);
2036 out:
2037 strbuf_release(&path);
2039 return dir_state;
2042 int cmp_dir_entry(const void *p1, const void *p2)
2044 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2045 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2047 return name_compare(e1->name, e1->len, e2->name, e2->len);
2050 /* check if *out lexically strictly contains *in */
2051 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2053 return (out->len < in->len) &&
2054 (out->name[out->len - 1] == '/') &&
2055 !memcmp(out->name, in->name, out->len);
2058 static int treat_leading_path(struct dir_struct *dir,
2059 struct index_state *istate,
2060 const char *path, int len,
2061 const struct pathspec *pathspec)
2063 struct strbuf sb = STRBUF_INIT;
2064 int baselen, rc = 0;
2065 const char *cp;
2066 int old_flags = dir->flags;
2068 while (len && path[len - 1] == '/')
2069 len--;
2070 if (!len)
2071 return 1;
2072 baselen = 0;
2073 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
2074 while (1) {
2075 cp = path + baselen + !!baselen;
2076 cp = memchr(cp, '/', path + len - cp);
2077 if (!cp)
2078 baselen = len;
2079 else
2080 baselen = cp - path;
2081 strbuf_setlen(&sb, 0);
2082 strbuf_add(&sb, path, baselen);
2083 if (!is_directory(sb.buf))
2084 break;
2085 if (simplify_away(sb.buf, sb.len, pathspec))
2086 break;
2087 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
2088 DT_DIR, NULL) == path_none)
2089 break; /* do not recurse into it */
2090 if (len <= baselen) {
2091 rc = 1;
2092 break; /* finished checking */
2095 strbuf_release(&sb);
2096 dir->flags = old_flags;
2097 return rc;
2100 static const char *get_ident_string(void)
2102 static struct strbuf sb = STRBUF_INIT;
2103 struct utsname uts;
2105 if (sb.len)
2106 return sb.buf;
2107 if (uname(&uts) < 0)
2108 die_errno(_("failed to get kernel name and information"));
2109 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2110 uts.sysname);
2111 return sb.buf;
2114 static int ident_in_untracked(const struct untracked_cache *uc)
2117 * Previous git versions may have saved many NUL separated
2118 * strings in the "ident" field, but it is insane to manage
2119 * many locations, so just take care of the first one.
2122 return !strcmp(uc->ident.buf, get_ident_string());
2125 static void set_untracked_ident(struct untracked_cache *uc)
2127 strbuf_reset(&uc->ident);
2128 strbuf_addstr(&uc->ident, get_ident_string());
2131 * This strbuf used to contain a list of NUL separated
2132 * strings, so save NUL too for backward compatibility.
2134 strbuf_addch(&uc->ident, 0);
2137 static void new_untracked_cache(struct index_state *istate)
2139 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2140 strbuf_init(&uc->ident, 100);
2141 uc->exclude_per_dir = ".gitignore";
2142 /* should be the same flags used by git-status */
2143 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2144 set_untracked_ident(uc);
2145 istate->untracked = uc;
2146 istate->cache_changed |= UNTRACKED_CHANGED;
2149 void add_untracked_cache(struct index_state *istate)
2151 if (!istate->untracked) {
2152 new_untracked_cache(istate);
2153 } else {
2154 if (!ident_in_untracked(istate->untracked)) {
2155 free_untracked_cache(istate->untracked);
2156 new_untracked_cache(istate);
2161 void remove_untracked_cache(struct index_state *istate)
2163 if (istate->untracked) {
2164 free_untracked_cache(istate->untracked);
2165 istate->untracked = NULL;
2166 istate->cache_changed |= UNTRACKED_CHANGED;
2170 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2171 int base_len,
2172 const struct pathspec *pathspec)
2174 struct untracked_cache_dir *root;
2176 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
2177 return NULL;
2180 * We only support $GIT_DIR/info/exclude and core.excludesfile
2181 * as the global ignore rule files. Any other additions
2182 * (e.g. from command line) invalidate the cache. This
2183 * condition also catches running setup_standard_excludes()
2184 * before setting dir->untracked!
2186 if (dir->unmanaged_exclude_files)
2187 return NULL;
2190 * Optimize for the main use case only: whole-tree git
2191 * status. More work involved in treat_leading_path() if we
2192 * use cache on just a subset of the worktree. pathspec
2193 * support could make the matter even worse.
2195 if (base_len || (pathspec && pathspec->nr))
2196 return NULL;
2198 /* Different set of flags may produce different results */
2199 if (dir->flags != dir->untracked->dir_flags ||
2201 * See treat_directory(), case index_nonexistent. Without
2202 * this flag, we may need to also cache .git file content
2203 * for the resolve_gitlink_ref() call, which we don't.
2205 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2206 /* We don't support collecting ignore files */
2207 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2208 DIR_COLLECT_IGNORED)))
2209 return NULL;
2212 * If we use .gitignore in the cache and now you change it to
2213 * .gitexclude, everything will go wrong.
2215 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2216 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2217 return NULL;
2220 * EXC_CMDL is not considered in the cache. If people set it,
2221 * skip the cache.
2223 if (dir->exclude_list_group[EXC_CMDL].nr)
2224 return NULL;
2226 if (!ident_in_untracked(dir->untracked)) {
2227 warning(_("Untracked cache is disabled on this system or location."));
2228 return NULL;
2231 if (!dir->untracked->root) {
2232 const int len = sizeof(*dir->untracked->root);
2233 dir->untracked->root = xmalloc(len);
2234 memset(dir->untracked->root, 0, len);
2237 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2238 root = dir->untracked->root;
2239 if (oidcmp(&dir->ss_info_exclude.oid,
2240 &dir->untracked->ss_info_exclude.oid)) {
2241 invalidate_gitignore(dir->untracked, root);
2242 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2244 if (oidcmp(&dir->ss_excludes_file.oid,
2245 &dir->untracked->ss_excludes_file.oid)) {
2246 invalidate_gitignore(dir->untracked, root);
2247 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2250 /* Make sure this directory is not dropped out at saving phase */
2251 root->recurse = 1;
2252 return root;
2255 int read_directory(struct dir_struct *dir, struct index_state *istate,
2256 const char *path, int len, const struct pathspec *pathspec)
2258 struct untracked_cache_dir *untracked;
2259 uint64_t start = getnanotime();
2261 if (has_symlink_leading_path(path, len))
2262 return dir->nr;
2264 untracked = validate_untracked_cache(dir, len, pathspec);
2265 if (!untracked)
2267 * make sure untracked cache code path is disabled,
2268 * e.g. prep_exclude()
2270 dir->untracked = NULL;
2271 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2272 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2273 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2274 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2277 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2278 * also pick up untracked contents of untracked dirs; by default
2279 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2281 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2282 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2283 int i, j;
2285 /* remove from dir->entries untracked contents of untracked dirs */
2286 for (i = j = 0; j < dir->nr; j++) {
2287 if (i &&
2288 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2289 FREE_AND_NULL(dir->entries[j]);
2290 } else {
2291 dir->entries[i++] = dir->entries[j];
2295 dir->nr = i;
2298 trace_performance_since(start, "read directory %.*s", len, path);
2299 if (dir->untracked) {
2300 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2301 trace_printf_key(&trace_untracked_stats,
2302 "node creation: %u\n"
2303 "gitignore invalidation: %u\n"
2304 "directory invalidation: %u\n"
2305 "opendir: %u\n",
2306 dir->untracked->dir_created,
2307 dir->untracked->gitignore_invalidated,
2308 dir->untracked->dir_invalidated,
2309 dir->untracked->dir_opened);
2310 if (dir->untracked == istate->untracked &&
2311 (dir->untracked->dir_opened ||
2312 dir->untracked->gitignore_invalidated ||
2313 dir->untracked->dir_invalidated))
2314 istate->cache_changed |= UNTRACKED_CHANGED;
2315 if (dir->untracked != istate->untracked) {
2316 FREE_AND_NULL(dir->untracked);
2319 return dir->nr;
2322 int file_exists(const char *f)
2324 struct stat sb;
2325 return lstat(f, &sb) == 0;
2328 static int cmp_icase(char a, char b)
2330 if (a == b)
2331 return 0;
2332 if (ignore_case)
2333 return toupper(a) - toupper(b);
2334 return a - b;
2338 * Given two normalized paths (a trailing slash is ok), if subdir is
2339 * outside dir, return -1. Otherwise return the offset in subdir that
2340 * can be used as relative path to dir.
2342 int dir_inside_of(const char *subdir, const char *dir)
2344 int offset = 0;
2346 assert(dir && subdir && *dir && *subdir);
2348 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2349 dir++;
2350 subdir++;
2351 offset++;
2354 /* hel[p]/me vs hel[l]/yeah */
2355 if (*dir && *subdir)
2356 return -1;
2358 if (!*subdir)
2359 return !*dir ? offset : -1; /* same dir */
2361 /* foo/[b]ar vs foo/[] */
2362 if (is_dir_sep(dir[-1]))
2363 return is_dir_sep(subdir[-1]) ? offset : -1;
2365 /* foo[/]bar vs foo[] */
2366 return is_dir_sep(*subdir) ? offset + 1 : -1;
2369 int is_inside_dir(const char *dir)
2371 char *cwd;
2372 int rc;
2374 if (!dir)
2375 return 0;
2377 cwd = xgetcwd();
2378 rc = (dir_inside_of(cwd, dir) >= 0);
2379 free(cwd);
2380 return rc;
2383 int is_empty_dir(const char *path)
2385 DIR *dir = opendir(path);
2386 struct dirent *e;
2387 int ret = 1;
2389 if (!dir)
2390 return 0;
2392 while ((e = readdir(dir)) != NULL)
2393 if (!is_dot_or_dotdot(e->d_name)) {
2394 ret = 0;
2395 break;
2398 closedir(dir);
2399 return ret;
2402 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2404 DIR *dir;
2405 struct dirent *e;
2406 int ret = 0, original_len = path->len, len, kept_down = 0;
2407 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2408 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2409 struct object_id submodule_head;
2411 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2412 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2413 /* Do not descend and nuke a nested git work tree. */
2414 if (kept_up)
2415 *kept_up = 1;
2416 return 0;
2419 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2420 dir = opendir(path->buf);
2421 if (!dir) {
2422 if (errno == ENOENT)
2423 return keep_toplevel ? -1 : 0;
2424 else if (errno == EACCES && !keep_toplevel)
2426 * An empty dir could be removable even if it
2427 * is unreadable:
2429 return rmdir(path->buf);
2430 else
2431 return -1;
2433 strbuf_complete(path, '/');
2435 len = path->len;
2436 while ((e = readdir(dir)) != NULL) {
2437 struct stat st;
2438 if (is_dot_or_dotdot(e->d_name))
2439 continue;
2441 strbuf_setlen(path, len);
2442 strbuf_addstr(path, e->d_name);
2443 if (lstat(path->buf, &st)) {
2444 if (errno == ENOENT)
2446 * file disappeared, which is what we
2447 * wanted anyway
2449 continue;
2450 /* fall thru */
2451 } else if (S_ISDIR(st.st_mode)) {
2452 if (!remove_dir_recurse(path, flag, &kept_down))
2453 continue; /* happy */
2454 } else if (!only_empty &&
2455 (!unlink(path->buf) || errno == ENOENT)) {
2456 continue; /* happy, too */
2459 /* path too long, stat fails, or non-directory still exists */
2460 ret = -1;
2461 break;
2463 closedir(dir);
2465 strbuf_setlen(path, original_len);
2466 if (!ret && !keep_toplevel && !kept_down)
2467 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2468 else if (kept_up)
2470 * report the uplevel that it is not an error that we
2471 * did not rmdir() our directory.
2473 *kept_up = !ret;
2474 return ret;
2477 int remove_dir_recursively(struct strbuf *path, int flag)
2479 return remove_dir_recurse(path, flag, NULL);
2482 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2484 void setup_standard_excludes(struct dir_struct *dir)
2486 dir->exclude_per_dir = ".gitignore";
2488 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2489 if (!excludes_file)
2490 excludes_file = xdg_config_home("ignore");
2491 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2492 add_excludes_from_file_1(dir, excludes_file,
2493 dir->untracked ? &dir->ss_excludes_file : NULL);
2495 /* per repository user preference */
2496 if (startup_info->have_repository) {
2497 const char *path = git_path_info_exclude();
2498 if (!access_or_warn(path, R_OK, 0))
2499 add_excludes_from_file_1(dir, path,
2500 dir->untracked ? &dir->ss_info_exclude : NULL);
2504 int remove_path(const char *name)
2506 char *slash;
2508 if (unlink(name) && !is_missing_file_error(errno))
2509 return -1;
2511 slash = strrchr(name, '/');
2512 if (slash) {
2513 char *dirs = xstrdup(name);
2514 slash = dirs + (slash - name);
2515 do {
2516 *slash = '\0';
2517 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2518 free(dirs);
2520 return 0;
2524 * Frees memory within dir which was allocated for exclude lists and
2525 * the exclude_stack. Does not free dir itself.
2527 void clear_directory(struct dir_struct *dir)
2529 int i, j;
2530 struct exclude_list_group *group;
2531 struct exclude_list *el;
2532 struct exclude_stack *stk;
2534 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2535 group = &dir->exclude_list_group[i];
2536 for (j = 0; j < group->nr; j++) {
2537 el = &group->el[j];
2538 if (i == EXC_DIRS)
2539 free((char *)el->src);
2540 clear_exclude_list(el);
2542 free(group->el);
2545 stk = dir->exclude_stack;
2546 while (stk) {
2547 struct exclude_stack *prev = stk->prev;
2548 free(stk);
2549 stk = prev;
2551 strbuf_release(&dir->basebuf);
2554 struct ondisk_untracked_cache {
2555 struct stat_data info_exclude_stat;
2556 struct stat_data excludes_file_stat;
2557 uint32_t dir_flags;
2558 unsigned char info_exclude_sha1[20];
2559 unsigned char excludes_file_sha1[20];
2560 char exclude_per_dir[FLEX_ARRAY];
2563 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2564 #define ouc_size(len) (ouc_offset(exclude_per_dir) + len + 1)
2566 struct write_data {
2567 int index; /* number of written untracked_cache_dir */
2568 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2569 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2570 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2571 struct strbuf out;
2572 struct strbuf sb_stat;
2573 struct strbuf sb_sha1;
2576 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2578 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2579 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2580 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2581 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2582 to->sd_dev = htonl(from->sd_dev);
2583 to->sd_ino = htonl(from->sd_ino);
2584 to->sd_uid = htonl(from->sd_uid);
2585 to->sd_gid = htonl(from->sd_gid);
2586 to->sd_size = htonl(from->sd_size);
2589 static void write_one_dir(struct untracked_cache_dir *untracked,
2590 struct write_data *wd)
2592 struct stat_data stat_data;
2593 struct strbuf *out = &wd->out;
2594 unsigned char intbuf[16];
2595 unsigned int intlen, value;
2596 int i = wd->index++;
2599 * untracked_nr should be reset whenever valid is clear, but
2600 * for safety..
2602 if (!untracked->valid) {
2603 untracked->untracked_nr = 0;
2604 untracked->check_only = 0;
2607 if (untracked->check_only)
2608 ewah_set(wd->check_only, i);
2609 if (untracked->valid) {
2610 ewah_set(wd->valid, i);
2611 stat_data_to_disk(&stat_data, &untracked->stat_data);
2612 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2614 if (!is_null_sha1(untracked->exclude_sha1)) {
2615 ewah_set(wd->sha1_valid, i);
2616 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2619 intlen = encode_varint(untracked->untracked_nr, intbuf);
2620 strbuf_add(out, intbuf, intlen);
2622 /* skip non-recurse directories */
2623 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2624 if (untracked->dirs[i]->recurse)
2625 value++;
2626 intlen = encode_varint(value, intbuf);
2627 strbuf_add(out, intbuf, intlen);
2629 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2631 for (i = 0; i < untracked->untracked_nr; i++)
2632 strbuf_add(out, untracked->untracked[i],
2633 strlen(untracked->untracked[i]) + 1);
2635 for (i = 0; i < untracked->dirs_nr; i++)
2636 if (untracked->dirs[i]->recurse)
2637 write_one_dir(untracked->dirs[i], wd);
2640 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2642 struct ondisk_untracked_cache *ouc;
2643 struct write_data wd;
2644 unsigned char varbuf[16];
2645 int varint_len;
2646 size_t len = strlen(untracked->exclude_per_dir);
2648 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
2649 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2650 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2651 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.oid.hash);
2652 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.oid.hash);
2653 ouc->dir_flags = htonl(untracked->dir_flags);
2655 varint_len = encode_varint(untracked->ident.len, varbuf);
2656 strbuf_add(out, varbuf, varint_len);
2657 strbuf_addbuf(out, &untracked->ident);
2659 strbuf_add(out, ouc, ouc_size(len));
2660 FREE_AND_NULL(ouc);
2662 if (!untracked->root) {
2663 varint_len = encode_varint(0, varbuf);
2664 strbuf_add(out, varbuf, varint_len);
2665 return;
2668 wd.index = 0;
2669 wd.check_only = ewah_new();
2670 wd.valid = ewah_new();
2671 wd.sha1_valid = ewah_new();
2672 strbuf_init(&wd.out, 1024);
2673 strbuf_init(&wd.sb_stat, 1024);
2674 strbuf_init(&wd.sb_sha1, 1024);
2675 write_one_dir(untracked->root, &wd);
2677 varint_len = encode_varint(wd.index, varbuf);
2678 strbuf_add(out, varbuf, varint_len);
2679 strbuf_addbuf(out, &wd.out);
2680 ewah_serialize_strbuf(wd.valid, out);
2681 ewah_serialize_strbuf(wd.check_only, out);
2682 ewah_serialize_strbuf(wd.sha1_valid, out);
2683 strbuf_addbuf(out, &wd.sb_stat);
2684 strbuf_addbuf(out, &wd.sb_sha1);
2685 strbuf_addch(out, '\0'); /* safe guard for string lists */
2687 ewah_free(wd.valid);
2688 ewah_free(wd.check_only);
2689 ewah_free(wd.sha1_valid);
2690 strbuf_release(&wd.out);
2691 strbuf_release(&wd.sb_stat);
2692 strbuf_release(&wd.sb_sha1);
2695 static void free_untracked(struct untracked_cache_dir *ucd)
2697 int i;
2698 if (!ucd)
2699 return;
2700 for (i = 0; i < ucd->dirs_nr; i++)
2701 free_untracked(ucd->dirs[i]);
2702 for (i = 0; i < ucd->untracked_nr; i++)
2703 free(ucd->untracked[i]);
2704 free(ucd->untracked);
2705 free(ucd->dirs);
2706 free(ucd);
2709 void free_untracked_cache(struct untracked_cache *uc)
2711 if (uc)
2712 free_untracked(uc->root);
2713 free(uc);
2716 struct read_data {
2717 int index;
2718 struct untracked_cache_dir **ucd;
2719 struct ewah_bitmap *check_only;
2720 struct ewah_bitmap *valid;
2721 struct ewah_bitmap *sha1_valid;
2722 const unsigned char *data;
2723 const unsigned char *end;
2726 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2728 memcpy(to, data, sizeof(*to));
2729 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2730 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2731 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2732 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2733 to->sd_dev = ntohl(to->sd_dev);
2734 to->sd_ino = ntohl(to->sd_ino);
2735 to->sd_uid = ntohl(to->sd_uid);
2736 to->sd_gid = ntohl(to->sd_gid);
2737 to->sd_size = ntohl(to->sd_size);
2740 static int read_one_dir(struct untracked_cache_dir **untracked_,
2741 struct read_data *rd)
2743 struct untracked_cache_dir ud, *untracked;
2744 const unsigned char *next, *data = rd->data, *end = rd->end;
2745 unsigned int value;
2746 int i, len;
2748 memset(&ud, 0, sizeof(ud));
2750 next = data;
2751 value = decode_varint(&next);
2752 if (next > end)
2753 return -1;
2754 ud.recurse = 1;
2755 ud.untracked_alloc = value;
2756 ud.untracked_nr = value;
2757 if (ud.untracked_nr)
2758 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2759 data = next;
2761 next = data;
2762 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2763 if (next > end)
2764 return -1;
2765 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2766 data = next;
2768 len = strlen((const char *)data);
2769 next = data + len + 1;
2770 if (next > rd->end)
2771 return -1;
2772 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
2773 memcpy(untracked, &ud, sizeof(ud));
2774 memcpy(untracked->name, data, len + 1);
2775 data = next;
2777 for (i = 0; i < untracked->untracked_nr; i++) {
2778 len = strlen((const char *)data);
2779 next = data + len + 1;
2780 if (next > rd->end)
2781 return -1;
2782 untracked->untracked[i] = xstrdup((const char*)data);
2783 data = next;
2786 rd->ucd[rd->index++] = untracked;
2787 rd->data = data;
2789 for (i = 0; i < untracked->dirs_nr; i++) {
2790 len = read_one_dir(untracked->dirs + i, rd);
2791 if (len < 0)
2792 return -1;
2794 return 0;
2797 static void set_check_only(size_t pos, void *cb)
2799 struct read_data *rd = cb;
2800 struct untracked_cache_dir *ud = rd->ucd[pos];
2801 ud->check_only = 1;
2804 static void read_stat(size_t pos, void *cb)
2806 struct read_data *rd = cb;
2807 struct untracked_cache_dir *ud = rd->ucd[pos];
2808 if (rd->data + sizeof(struct stat_data) > rd->end) {
2809 rd->data = rd->end + 1;
2810 return;
2812 stat_data_from_disk(&ud->stat_data, rd->data);
2813 rd->data += sizeof(struct stat_data);
2814 ud->valid = 1;
2817 static void read_sha1(size_t pos, void *cb)
2819 struct read_data *rd = cb;
2820 struct untracked_cache_dir *ud = rd->ucd[pos];
2821 if (rd->data + 20 > rd->end) {
2822 rd->data = rd->end + 1;
2823 return;
2825 hashcpy(ud->exclude_sha1, rd->data);
2826 rd->data += 20;
2829 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
2830 const unsigned char *sha1)
2832 stat_data_from_disk(&oid_stat->stat, data);
2833 hashcpy(oid_stat->oid.hash, sha1);
2834 oid_stat->valid = 1;
2837 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2839 struct untracked_cache *uc;
2840 struct read_data rd;
2841 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2842 const char *ident;
2843 int ident_len, len;
2844 const char *exclude_per_dir;
2846 if (sz <= 1 || end[-1] != '\0')
2847 return NULL;
2848 end--;
2850 ident_len = decode_varint(&next);
2851 if (next + ident_len > end)
2852 return NULL;
2853 ident = (const char *)next;
2854 next += ident_len;
2856 if (next + ouc_size(0) > end)
2857 return NULL;
2859 uc = xcalloc(1, sizeof(*uc));
2860 strbuf_init(&uc->ident, ident_len);
2861 strbuf_add(&uc->ident, ident, ident_len);
2862 load_oid_stat(&uc->ss_info_exclude,
2863 next + ouc_offset(info_exclude_stat),
2864 next + ouc_offset(info_exclude_sha1));
2865 load_oid_stat(&uc->ss_excludes_file,
2866 next + ouc_offset(excludes_file_stat),
2867 next + ouc_offset(excludes_file_sha1));
2868 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2869 exclude_per_dir = (const char *)next + ouc_offset(exclude_per_dir);
2870 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2871 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2872 next += ouc_size(strlen(exclude_per_dir));
2873 if (next >= end)
2874 goto done2;
2876 len = decode_varint(&next);
2877 if (next > end || len == 0)
2878 goto done2;
2880 rd.valid = ewah_new();
2881 rd.check_only = ewah_new();
2882 rd.sha1_valid = ewah_new();
2883 rd.data = next;
2884 rd.end = end;
2885 rd.index = 0;
2886 ALLOC_ARRAY(rd.ucd, len);
2888 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2889 goto done;
2891 next = rd.data;
2892 len = ewah_read_mmap(rd.valid, next, end - next);
2893 if (len < 0)
2894 goto done;
2896 next += len;
2897 len = ewah_read_mmap(rd.check_only, next, end - next);
2898 if (len < 0)
2899 goto done;
2901 next += len;
2902 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2903 if (len < 0)
2904 goto done;
2906 ewah_each_bit(rd.check_only, set_check_only, &rd);
2907 rd.data = next + len;
2908 ewah_each_bit(rd.valid, read_stat, &rd);
2909 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2910 next = rd.data;
2912 done:
2913 free(rd.ucd);
2914 ewah_free(rd.valid);
2915 ewah_free(rd.check_only);
2916 ewah_free(rd.sha1_valid);
2917 done2:
2918 if (next != end) {
2919 free_untracked_cache(uc);
2920 uc = NULL;
2922 return uc;
2925 static void invalidate_one_directory(struct untracked_cache *uc,
2926 struct untracked_cache_dir *ucd)
2928 uc->dir_invalidated++;
2929 ucd->valid = 0;
2930 ucd->untracked_nr = 0;
2934 * Normally when an entry is added or removed from a directory,
2935 * invalidating that directory is enough. No need to touch its
2936 * ancestors. When a directory is shown as "foo/bar/" in git-status
2937 * however, deleting or adding an entry may have cascading effect.
2939 * Say the "foo/bar/file" has become untracked, we need to tell the
2940 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2941 * directory any more (because "bar" is managed by foo as an untracked
2942 * "file").
2944 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2945 * was the last untracked entry in the entire "foo", we should show
2946 * "foo/" instead. Which means we have to invalidate past "bar" up to
2947 * "foo".
2949 * This function traverses all directories from root to leaf. If there
2950 * is a chance of one of the above cases happening, we invalidate back
2951 * to root. Otherwise we just invalidate the leaf. There may be a more
2952 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2953 * detect these cases and avoid unnecessary invalidation, for example,
2954 * checking for the untracked entry named "bar/" in "foo", but for now
2955 * stick to something safe and simple.
2957 static int invalidate_one_component(struct untracked_cache *uc,
2958 struct untracked_cache_dir *dir,
2959 const char *path, int len)
2961 const char *rest = strchr(path, '/');
2963 if (rest) {
2964 int component_len = rest - path;
2965 struct untracked_cache_dir *d =
2966 lookup_untracked(uc, dir, path, component_len);
2967 int ret =
2968 invalidate_one_component(uc, d, rest + 1,
2969 len - (component_len + 1));
2970 if (ret)
2971 invalidate_one_directory(uc, dir);
2972 return ret;
2975 invalidate_one_directory(uc, dir);
2976 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2979 void untracked_cache_invalidate_path(struct index_state *istate,
2980 const char *path, int safe_path)
2982 if (!istate->untracked || !istate->untracked->root)
2983 return;
2984 if (!safe_path && !verify_path(path))
2985 return;
2986 invalidate_one_component(istate->untracked, istate->untracked->root,
2987 path, strlen(path));
2990 void untracked_cache_remove_from_index(struct index_state *istate,
2991 const char *path)
2993 untracked_cache_invalidate_path(istate, path, 1);
2996 void untracked_cache_add_to_index(struct index_state *istate,
2997 const char *path)
2999 untracked_cache_invalidate_path(istate, path, 1);
3002 /* Update gitfile and core.worktree setting to connect work tree and git dir */
3003 void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)
3005 struct strbuf gitfile_sb = STRBUF_INIT;
3006 struct strbuf cfg_sb = STRBUF_INIT;
3007 struct strbuf rel_path = STRBUF_INIT;
3008 char *git_dir, *work_tree;
3010 /* Prepare .git file */
3011 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3012 if (safe_create_leading_directories_const(gitfile_sb.buf))
3013 die(_("could not create directories for %s"), gitfile_sb.buf);
3015 /* Prepare config file */
3016 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3017 if (safe_create_leading_directories_const(cfg_sb.buf))
3018 die(_("could not create directories for %s"), cfg_sb.buf);
3020 git_dir = real_pathdup(git_dir_, 1);
3021 work_tree = real_pathdup(work_tree_, 1);
3023 /* Write .git file */
3024 write_file(gitfile_sb.buf, "gitdir: %s",
3025 relative_path(git_dir, work_tree, &rel_path));
3026 /* Update core.worktree setting */
3027 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3028 relative_path(work_tree, git_dir, &rel_path));
3030 strbuf_release(&gitfile_sb);
3031 strbuf_release(&cfg_sb);
3032 strbuf_release(&rel_path);
3033 free(work_tree);
3034 free(git_dir);
3038 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3040 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3042 if (rename(old_git_dir, new_git_dir) < 0)
3043 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3044 old_git_dir, new_git_dir);
3046 connect_work_tree_and_git_dir(path, new_git_dir);