refs: update ref transactions to use struct object_id
[alt-git.git] / dir.c
blob1d17b800cf374d179d7cef3c6bbb180a5a1b4aba
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * See Documentation/technical/api-directory-listing.txt
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10 #define NO_THE_INDEX_COMPATIBILITY_MACROS
11 #include "cache.h"
12 #include "config.h"
13 #include "dir.h"
14 #include "attr.h"
15 #include "refs.h"
16 #include "wildmatch.h"
17 #include "pathspec.h"
18 #include "utf8.h"
19 #include "varint.h"
20 #include "ewah/ewok.h"
23 * Tells read_directory_recursive how a file or directory should be treated.
24 * Values are ordered by significance, e.g. if a directory contains both
25 * excluded and untracked files, it is listed as untracked because
26 * path_untracked > path_excluded.
28 enum path_treatment {
29 path_none = 0,
30 path_recurse,
31 path_excluded,
32 path_untracked
36 * Support data structure for our opendir/readdir/closedir wrappers
38 struct cached_dir {
39 DIR *fdir;
40 struct untracked_cache_dir *untracked;
41 int nr_files;
42 int nr_dirs;
44 struct dirent *de;
45 const char *file;
46 struct untracked_cache_dir *ucd;
49 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
50 struct index_state *istate, const char *path, int len,
51 struct untracked_cache_dir *untracked,
52 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
53 static int get_dtype(struct dirent *de, struct index_state *istate,
54 const char *path, int len);
56 int count_slashes(const char *s)
58 int cnt = 0;
59 while (*s)
60 if (*s++ == '/')
61 cnt++;
62 return cnt;
65 int fspathcmp(const char *a, const char *b)
67 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
70 int fspathncmp(const char *a, const char *b, size_t count)
72 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
75 int git_fnmatch(const struct pathspec_item *item,
76 const char *pattern, const char *string,
77 int prefix)
79 if (prefix > 0) {
80 if (ps_strncmp(item, pattern, string, prefix))
81 return WM_NOMATCH;
82 pattern += prefix;
83 string += prefix;
85 if (item->flags & PATHSPEC_ONESTAR) {
86 int pattern_len = strlen(++pattern);
87 int string_len = strlen(string);
88 return string_len < pattern_len ||
89 ps_strcmp(item, pattern,
90 string + string_len - pattern_len);
92 if (item->magic & PATHSPEC_GLOB)
93 return wildmatch(pattern, string,
94 WM_PATHNAME |
95 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
96 else
97 /* wildmatch has not learned no FNM_PATHNAME mode yet */
98 return wildmatch(pattern, string,
99 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
102 static int fnmatch_icase_mem(const char *pattern, int patternlen,
103 const char *string, int stringlen,
104 int flags)
106 int match_status;
107 struct strbuf pat_buf = STRBUF_INIT;
108 struct strbuf str_buf = STRBUF_INIT;
109 const char *use_pat = pattern;
110 const char *use_str = string;
112 if (pattern[patternlen]) {
113 strbuf_add(&pat_buf, pattern, patternlen);
114 use_pat = pat_buf.buf;
116 if (string[stringlen]) {
117 strbuf_add(&str_buf, string, stringlen);
118 use_str = str_buf.buf;
121 if (ignore_case)
122 flags |= WM_CASEFOLD;
123 match_status = wildmatch(use_pat, use_str, flags);
125 strbuf_release(&pat_buf);
126 strbuf_release(&str_buf);
128 return match_status;
131 static size_t common_prefix_len(const struct pathspec *pathspec)
133 int n;
134 size_t max = 0;
137 * ":(icase)path" is treated as a pathspec full of
138 * wildcard. In other words, only prefix is considered common
139 * prefix. If the pathspec is abc/foo abc/bar, running in
140 * subdir xyz, the common prefix is still xyz, not xuz/abc as
141 * in non-:(icase).
143 GUARD_PATHSPEC(pathspec,
144 PATHSPEC_FROMTOP |
145 PATHSPEC_MAXDEPTH |
146 PATHSPEC_LITERAL |
147 PATHSPEC_GLOB |
148 PATHSPEC_ICASE |
149 PATHSPEC_EXCLUDE |
150 PATHSPEC_ATTR);
152 for (n = 0; n < pathspec->nr; n++) {
153 size_t i = 0, len = 0, item_len;
154 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
155 continue;
156 if (pathspec->items[n].magic & PATHSPEC_ICASE)
157 item_len = pathspec->items[n].prefix;
158 else
159 item_len = pathspec->items[n].nowildcard_len;
160 while (i < item_len && (n == 0 || i < max)) {
161 char c = pathspec->items[n].match[i];
162 if (c != pathspec->items[0].match[i])
163 break;
164 if (c == '/')
165 len = i + 1;
166 i++;
168 if (n == 0 || len < max) {
169 max = len;
170 if (!max)
171 break;
174 return max;
178 * Returns a copy of the longest leading path common among all
179 * pathspecs.
181 char *common_prefix(const struct pathspec *pathspec)
183 unsigned long len = common_prefix_len(pathspec);
185 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
188 int fill_directory(struct dir_struct *dir,
189 struct index_state *istate,
190 const struct pathspec *pathspec)
192 const char *prefix;
193 size_t prefix_len;
196 * Calculate common prefix for the pathspec, and
197 * use that to optimize the directory walk
199 prefix_len = common_prefix_len(pathspec);
200 prefix = prefix_len ? pathspec->items[0].match : "";
202 /* Read the directory and prune it */
203 read_directory(dir, istate, prefix, prefix_len, pathspec);
205 return prefix_len;
208 int within_depth(const char *name, int namelen,
209 int depth, int max_depth)
211 const char *cp = name, *cpe = name + namelen;
213 while (cp < cpe) {
214 if (*cp++ != '/')
215 continue;
216 depth++;
217 if (depth > max_depth)
218 return 0;
220 return 1;
223 #define DO_MATCH_EXCLUDE (1<<0)
224 #define DO_MATCH_DIRECTORY (1<<1)
225 #define DO_MATCH_SUBMODULE (1<<2)
227 static int match_attrs(const char *name, int namelen,
228 const struct pathspec_item *item)
230 int i;
232 git_check_attr(name, item->attr_check);
233 for (i = 0; i < item->attr_match_nr; i++) {
234 const char *value;
235 int matched;
236 enum attr_match_mode match_mode;
238 value = item->attr_check->items[i].value;
239 match_mode = item->attr_match[i].match_mode;
241 if (ATTR_TRUE(value))
242 matched = (match_mode == MATCH_SET);
243 else if (ATTR_FALSE(value))
244 matched = (match_mode == MATCH_UNSET);
245 else if (ATTR_UNSET(value))
246 matched = (match_mode == MATCH_UNSPECIFIED);
247 else
248 matched = (match_mode == MATCH_VALUE &&
249 !strcmp(item->attr_match[i].value, value));
250 if (!matched)
251 return 0;
254 return 1;
258 * Does 'match' match the given name?
259 * A match is found if
261 * (1) the 'match' string is leading directory of 'name', or
262 * (2) the 'match' string is a wildcard and matches 'name', or
263 * (3) the 'match' string is exactly the same as 'name'.
265 * and the return value tells which case it was.
267 * It returns 0 when there is no match.
269 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
270 const char *name, int namelen, unsigned flags)
272 /* name/namelen has prefix cut off by caller */
273 const char *match = item->match + prefix;
274 int matchlen = item->len - prefix;
277 * The normal call pattern is:
278 * 1. prefix = common_prefix_len(ps);
279 * 2. prune something, or fill_directory
280 * 3. match_pathspec()
282 * 'prefix' at #1 may be shorter than the command's prefix and
283 * it's ok for #2 to match extra files. Those extras will be
284 * trimmed at #3.
286 * Suppose the pathspec is 'foo' and '../bar' running from
287 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
288 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
289 * user does not want XYZ/foo, only the "foo" part should be
290 * case-insensitive. We need to filter out XYZ/foo here. In
291 * other words, we do not trust the caller on comparing the
292 * prefix part when :(icase) is involved. We do exact
293 * comparison ourselves.
295 * Normally the caller (common_prefix_len() in fact) does
296 * _exact_ matching on name[-prefix+1..-1] and we do not need
297 * to check that part. Be defensive and check it anyway, in
298 * case common_prefix_len is changed, or a new caller is
299 * introduced that does not use common_prefix_len.
301 * If the penalty turns out too high when prefix is really
302 * long, maybe change it to
303 * strncmp(match, name, item->prefix - prefix)
305 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
306 strncmp(item->match, name - prefix, item->prefix))
307 return 0;
309 if (item->attr_match_nr && !match_attrs(name, namelen, item))
310 return 0;
312 /* If the match was just the prefix, we matched */
313 if (!*match)
314 return MATCHED_RECURSIVELY;
316 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
317 if (matchlen == namelen)
318 return MATCHED_EXACTLY;
320 if (match[matchlen-1] == '/' || name[matchlen] == '/')
321 return MATCHED_RECURSIVELY;
322 } else if ((flags & DO_MATCH_DIRECTORY) &&
323 match[matchlen - 1] == '/' &&
324 namelen == matchlen - 1 &&
325 !ps_strncmp(item, match, name, namelen))
326 return MATCHED_EXACTLY;
328 if (item->nowildcard_len < item->len &&
329 !git_fnmatch(item, match, name,
330 item->nowildcard_len - prefix))
331 return MATCHED_FNMATCH;
333 /* Perform checks to see if "name" is a super set of the pathspec */
334 if (flags & DO_MATCH_SUBMODULE) {
335 /* name is a literal prefix of the pathspec */
336 if ((namelen < matchlen) &&
337 (match[namelen] == '/') &&
338 !ps_strncmp(item, match, name, namelen))
339 return MATCHED_RECURSIVELY;
341 /* name" doesn't match up to the first wild character */
342 if (item->nowildcard_len < item->len &&
343 ps_strncmp(item, match, name,
344 item->nowildcard_len - prefix))
345 return 0;
348 * Here is where we would perform a wildmatch to check if
349 * "name" can be matched as a directory (or a prefix) against
350 * the pathspec. Since wildmatch doesn't have this capability
351 * at the present we have to punt and say that it is a match,
352 * potentially returning a false positive
353 * The submodules themselves will be able to perform more
354 * accurate matching to determine if the pathspec matches.
356 return MATCHED_RECURSIVELY;
359 return 0;
363 * Given a name and a list of pathspecs, returns the nature of the
364 * closest (i.e. most specific) match of the name to any of the
365 * pathspecs.
367 * The caller typically calls this multiple times with the same
368 * pathspec and seen[] array but with different name/namelen
369 * (e.g. entries from the index) and is interested in seeing if and
370 * how each pathspec matches all the names it calls this function
371 * with. A mark is left in the seen[] array for each pathspec element
372 * indicating the closest type of match that element achieved, so if
373 * seen[n] remains zero after multiple invocations, that means the nth
374 * pathspec did not match any names, which could indicate that the
375 * user mistyped the nth pathspec.
377 static int do_match_pathspec(const struct pathspec *ps,
378 const char *name, int namelen,
379 int prefix, char *seen,
380 unsigned flags)
382 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
384 GUARD_PATHSPEC(ps,
385 PATHSPEC_FROMTOP |
386 PATHSPEC_MAXDEPTH |
387 PATHSPEC_LITERAL |
388 PATHSPEC_GLOB |
389 PATHSPEC_ICASE |
390 PATHSPEC_EXCLUDE |
391 PATHSPEC_ATTR);
393 if (!ps->nr) {
394 if (!ps->recursive ||
395 !(ps->magic & PATHSPEC_MAXDEPTH) ||
396 ps->max_depth == -1)
397 return MATCHED_RECURSIVELY;
399 if (within_depth(name, namelen, 0, ps->max_depth))
400 return MATCHED_EXACTLY;
401 else
402 return 0;
405 name += prefix;
406 namelen -= prefix;
408 for (i = ps->nr - 1; i >= 0; i--) {
409 int how;
411 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
412 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
413 continue;
415 if (seen && seen[i] == MATCHED_EXACTLY)
416 continue;
418 * Make exclude patterns optional and never report
419 * "pathspec ':(exclude)foo' matches no files"
421 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
422 seen[i] = MATCHED_FNMATCH;
423 how = match_pathspec_item(ps->items+i, prefix, name,
424 namelen, flags);
425 if (ps->recursive &&
426 (ps->magic & PATHSPEC_MAXDEPTH) &&
427 ps->max_depth != -1 &&
428 how && how != MATCHED_FNMATCH) {
429 int len = ps->items[i].len;
430 if (name[len] == '/')
431 len++;
432 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
433 how = MATCHED_EXACTLY;
434 else
435 how = 0;
437 if (how) {
438 if (retval < how)
439 retval = how;
440 if (seen && seen[i] < how)
441 seen[i] = how;
444 return retval;
447 int match_pathspec(const struct pathspec *ps,
448 const char *name, int namelen,
449 int prefix, char *seen, int is_dir)
451 int positive, negative;
452 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
453 positive = do_match_pathspec(ps, name, namelen,
454 prefix, seen, flags);
455 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
456 return positive;
457 negative = do_match_pathspec(ps, name, namelen,
458 prefix, seen,
459 flags | DO_MATCH_EXCLUDE);
460 return negative ? 0 : positive;
464 * Check if a submodule is a superset of the pathspec
466 int submodule_path_match(const struct pathspec *ps,
467 const char *submodule_name,
468 char *seen)
470 int matched = do_match_pathspec(ps, submodule_name,
471 strlen(submodule_name),
472 0, seen,
473 DO_MATCH_DIRECTORY |
474 DO_MATCH_SUBMODULE);
475 return matched;
478 int report_path_error(const char *ps_matched,
479 const struct pathspec *pathspec,
480 const char *prefix)
483 * Make sure all pathspec matched; otherwise it is an error.
485 int num, errors = 0;
486 for (num = 0; num < pathspec->nr; num++) {
487 int other, found_dup;
489 if (ps_matched[num])
490 continue;
492 * The caller might have fed identical pathspec
493 * twice. Do not barf on such a mistake.
494 * FIXME: parse_pathspec should have eliminated
495 * duplicate pathspec.
497 for (found_dup = other = 0;
498 !found_dup && other < pathspec->nr;
499 other++) {
500 if (other == num || !ps_matched[other])
501 continue;
502 if (!strcmp(pathspec->items[other].original,
503 pathspec->items[num].original))
505 * Ok, we have a match already.
507 found_dup = 1;
509 if (found_dup)
510 continue;
512 error("pathspec '%s' did not match any file(s) known to git.",
513 pathspec->items[num].original);
514 errors++;
516 return errors;
520 * Return the length of the "simple" part of a path match limiter.
522 int simple_length(const char *match)
524 int len = -1;
526 for (;;) {
527 unsigned char c = *match++;
528 len++;
529 if (c == '\0' || is_glob_special(c))
530 return len;
534 int no_wildcard(const char *string)
536 return string[simple_length(string)] == '\0';
539 void parse_exclude_pattern(const char **pattern,
540 int *patternlen,
541 unsigned *flags,
542 int *nowildcardlen)
544 const char *p = *pattern;
545 size_t i, len;
547 *flags = 0;
548 if (*p == '!') {
549 *flags |= EXC_FLAG_NEGATIVE;
550 p++;
552 len = strlen(p);
553 if (len && p[len - 1] == '/') {
554 len--;
555 *flags |= EXC_FLAG_MUSTBEDIR;
557 for (i = 0; i < len; i++) {
558 if (p[i] == '/')
559 break;
561 if (i == len)
562 *flags |= EXC_FLAG_NODIR;
563 *nowildcardlen = simple_length(p);
565 * we should have excluded the trailing slash from 'p' too,
566 * but that's one more allocation. Instead just make sure
567 * nowildcardlen does not exceed real patternlen
569 if (*nowildcardlen > len)
570 *nowildcardlen = len;
571 if (*p == '*' && no_wildcard(p + 1))
572 *flags |= EXC_FLAG_ENDSWITH;
573 *pattern = p;
574 *patternlen = len;
577 void add_exclude(const char *string, const char *base,
578 int baselen, struct exclude_list *el, int srcpos)
580 struct exclude *x;
581 int patternlen;
582 unsigned flags;
583 int nowildcardlen;
585 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
586 if (flags & EXC_FLAG_MUSTBEDIR) {
587 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
588 } else {
589 x = xmalloc(sizeof(*x));
590 x->pattern = string;
592 x->patternlen = patternlen;
593 x->nowildcardlen = nowildcardlen;
594 x->base = base;
595 x->baselen = baselen;
596 x->flags = flags;
597 x->srcpos = srcpos;
598 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
599 el->excludes[el->nr++] = x;
600 x->el = el;
603 static void *read_skip_worktree_file_from_index(const struct index_state *istate,
604 const char *path, size_t *size,
605 struct sha1_stat *sha1_stat)
607 int pos, len;
608 unsigned long sz;
609 enum object_type type;
610 void *data;
612 len = strlen(path);
613 pos = index_name_pos(istate, path, len);
614 if (pos < 0)
615 return NULL;
616 if (!ce_skip_worktree(istate->cache[pos]))
617 return NULL;
618 data = read_sha1_file(istate->cache[pos]->oid.hash, &type, &sz);
619 if (!data || type != OBJ_BLOB) {
620 free(data);
621 return NULL;
623 *size = xsize_t(sz);
624 if (sha1_stat) {
625 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
626 hashcpy(sha1_stat->sha1, istate->cache[pos]->oid.hash);
628 return data;
632 * Frees memory within el which was allocated for exclude patterns and
633 * the file buffer. Does not free el itself.
635 void clear_exclude_list(struct exclude_list *el)
637 int i;
639 for (i = 0; i < el->nr; i++)
640 free(el->excludes[i]);
641 free(el->excludes);
642 free(el->filebuf);
644 memset(el, 0, sizeof(*el));
647 static void trim_trailing_spaces(char *buf)
649 char *p, *last_space = NULL;
651 for (p = buf; *p; p++)
652 switch (*p) {
653 case ' ':
654 if (!last_space)
655 last_space = p;
656 break;
657 case '\\':
658 p++;
659 if (!*p)
660 return;
661 /* fallthrough */
662 default:
663 last_space = NULL;
666 if (last_space)
667 *last_space = '\0';
671 * Given a subdirectory name and "dir" of the current directory,
672 * search the subdir in "dir" and return it, or create a new one if it
673 * does not exist in "dir".
675 * If "name" has the trailing slash, it'll be excluded in the search.
677 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
678 struct untracked_cache_dir *dir,
679 const char *name, int len)
681 int first, last;
682 struct untracked_cache_dir *d;
683 if (!dir)
684 return NULL;
685 if (len && name[len - 1] == '/')
686 len--;
687 first = 0;
688 last = dir->dirs_nr;
689 while (last > first) {
690 int cmp, next = (last + first) >> 1;
691 d = dir->dirs[next];
692 cmp = strncmp(name, d->name, len);
693 if (!cmp && strlen(d->name) > len)
694 cmp = -1;
695 if (!cmp)
696 return d;
697 if (cmp < 0) {
698 last = next;
699 continue;
701 first = next+1;
704 uc->dir_created++;
705 FLEX_ALLOC_MEM(d, name, name, len);
707 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
708 memmove(dir->dirs + first + 1, dir->dirs + first,
709 (dir->dirs_nr - first) * sizeof(*dir->dirs));
710 dir->dirs_nr++;
711 dir->dirs[first] = d;
712 return d;
715 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
717 int i;
718 dir->valid = 0;
719 dir->untracked_nr = 0;
720 for (i = 0; i < dir->dirs_nr; i++)
721 do_invalidate_gitignore(dir->dirs[i]);
724 static void invalidate_gitignore(struct untracked_cache *uc,
725 struct untracked_cache_dir *dir)
727 uc->gitignore_invalidated++;
728 do_invalidate_gitignore(dir);
731 static void invalidate_directory(struct untracked_cache *uc,
732 struct untracked_cache_dir *dir)
734 int i;
735 uc->dir_invalidated++;
736 dir->valid = 0;
737 dir->untracked_nr = 0;
738 for (i = 0; i < dir->dirs_nr; i++)
739 dir->dirs[i]->recurse = 0;
743 * Given a file with name "fname", read it (either from disk, or from
744 * an index if 'istate' is non-null), parse it and store the
745 * exclude rules in "el".
747 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
748 * stat data from disk (only valid if add_excludes returns zero). If
749 * ss_valid is non-zero, "ss" must contain good value as input.
751 static int add_excludes(const char *fname, const char *base, int baselen,
752 struct exclude_list *el,
753 struct index_state *istate,
754 struct sha1_stat *sha1_stat)
756 struct stat st;
757 int fd, i, lineno = 1;
758 size_t size = 0;
759 char *buf, *entry;
761 fd = open(fname, O_RDONLY);
762 if (fd < 0 || fstat(fd, &st) < 0) {
763 if (fd < 0)
764 warn_on_fopen_errors(fname);
765 else
766 close(fd);
767 if (!istate ||
768 (buf = read_skip_worktree_file_from_index(istate, fname, &size, sha1_stat)) == NULL)
769 return -1;
770 if (size == 0) {
771 free(buf);
772 return 0;
774 if (buf[size-1] != '\n') {
775 buf = xrealloc(buf, st_add(size, 1));
776 buf[size++] = '\n';
778 } else {
779 size = xsize_t(st.st_size);
780 if (size == 0) {
781 if (sha1_stat) {
782 fill_stat_data(&sha1_stat->stat, &st);
783 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
784 sha1_stat->valid = 1;
786 close(fd);
787 return 0;
789 buf = xmallocz(size);
790 if (read_in_full(fd, buf, size) != size) {
791 free(buf);
792 close(fd);
793 return -1;
795 buf[size++] = '\n';
796 close(fd);
797 if (sha1_stat) {
798 int pos;
799 if (sha1_stat->valid &&
800 !match_stat_data_racy(istate, &sha1_stat->stat, &st))
801 ; /* no content change, ss->sha1 still good */
802 else if (istate &&
803 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
804 !ce_stage(istate->cache[pos]) &&
805 ce_uptodate(istate->cache[pos]) &&
806 !would_convert_to_git(istate, fname))
807 hashcpy(sha1_stat->sha1,
808 istate->cache[pos]->oid.hash);
809 else
810 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
811 fill_stat_data(&sha1_stat->stat, &st);
812 sha1_stat->valid = 1;
816 el->filebuf = buf;
818 if (skip_utf8_bom(&buf, size))
819 size -= buf - el->filebuf;
821 entry = buf;
823 for (i = 0; i < size; i++) {
824 if (buf[i] == '\n') {
825 if (entry != buf + i && entry[0] != '#') {
826 buf[i - (i && buf[i-1] == '\r')] = 0;
827 trim_trailing_spaces(entry);
828 add_exclude(entry, base, baselen, el, lineno);
830 lineno++;
831 entry = buf + i + 1;
834 return 0;
837 int add_excludes_from_file_to_list(const char *fname, const char *base,
838 int baselen, struct exclude_list *el,
839 struct index_state *istate)
841 return add_excludes(fname, base, baselen, el, istate, NULL);
844 struct exclude_list *add_exclude_list(struct dir_struct *dir,
845 int group_type, const char *src)
847 struct exclude_list *el;
848 struct exclude_list_group *group;
850 group = &dir->exclude_list_group[group_type];
851 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
852 el = &group->el[group->nr++];
853 memset(el, 0, sizeof(*el));
854 el->src = src;
855 return el;
859 * Used to set up core.excludesfile and .git/info/exclude lists.
861 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
862 struct sha1_stat *sha1_stat)
864 struct exclude_list *el;
866 * catch setup_standard_excludes() that's called before
867 * dir->untracked is assigned. That function behaves
868 * differently when dir->untracked is non-NULL.
870 if (!dir->untracked)
871 dir->unmanaged_exclude_files++;
872 el = add_exclude_list(dir, EXC_FILE, fname);
873 if (add_excludes(fname, "", 0, el, NULL, sha1_stat) < 0)
874 die("cannot use %s as an exclude file", fname);
877 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
879 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
880 add_excludes_from_file_1(dir, fname, NULL);
883 int match_basename(const char *basename, int basenamelen,
884 const char *pattern, int prefix, int patternlen,
885 unsigned flags)
887 if (prefix == patternlen) {
888 if (patternlen == basenamelen &&
889 !fspathncmp(pattern, basename, basenamelen))
890 return 1;
891 } else if (flags & EXC_FLAG_ENDSWITH) {
892 /* "*literal" matching against "fooliteral" */
893 if (patternlen - 1 <= basenamelen &&
894 !fspathncmp(pattern + 1,
895 basename + basenamelen - (patternlen - 1),
896 patternlen - 1))
897 return 1;
898 } else {
899 if (fnmatch_icase_mem(pattern, patternlen,
900 basename, basenamelen,
901 0) == 0)
902 return 1;
904 return 0;
907 int match_pathname(const char *pathname, int pathlen,
908 const char *base, int baselen,
909 const char *pattern, int prefix, int patternlen,
910 unsigned flags)
912 const char *name;
913 int namelen;
916 * match with FNM_PATHNAME; the pattern has base implicitly
917 * in front of it.
919 if (*pattern == '/') {
920 pattern++;
921 patternlen--;
922 prefix--;
926 * baselen does not count the trailing slash. base[] may or
927 * may not end with a trailing slash though.
929 if (pathlen < baselen + 1 ||
930 (baselen && pathname[baselen] != '/') ||
931 fspathncmp(pathname, base, baselen))
932 return 0;
934 namelen = baselen ? pathlen - baselen - 1 : pathlen;
935 name = pathname + pathlen - namelen;
937 if (prefix) {
939 * if the non-wildcard part is longer than the
940 * remaining pathname, surely it cannot match.
942 if (prefix > namelen)
943 return 0;
945 if (fspathncmp(pattern, name, prefix))
946 return 0;
947 pattern += prefix;
948 patternlen -= prefix;
949 name += prefix;
950 namelen -= prefix;
953 * If the whole pattern did not have a wildcard,
954 * then our prefix match is all we need; we
955 * do not need to call fnmatch at all.
957 if (!patternlen && !namelen)
958 return 1;
961 return fnmatch_icase_mem(pattern, patternlen,
962 name, namelen,
963 WM_PATHNAME) == 0;
967 * Scan the given exclude list in reverse to see whether pathname
968 * should be ignored. The first match (i.e. the last on the list), if
969 * any, determines the fate. Returns the exclude_list element which
970 * matched, or NULL for undecided.
972 static struct exclude *last_exclude_matching_from_list(const char *pathname,
973 int pathlen,
974 const char *basename,
975 int *dtype,
976 struct exclude_list *el,
977 struct index_state *istate)
979 struct exclude *exc = NULL; /* undecided */
980 int i;
982 if (!el->nr)
983 return NULL; /* undefined */
985 for (i = el->nr - 1; 0 <= i; i--) {
986 struct exclude *x = el->excludes[i];
987 const char *exclude = x->pattern;
988 int prefix = x->nowildcardlen;
990 if (x->flags & EXC_FLAG_MUSTBEDIR) {
991 if (*dtype == DT_UNKNOWN)
992 *dtype = get_dtype(NULL, istate, pathname, pathlen);
993 if (*dtype != DT_DIR)
994 continue;
997 if (x->flags & EXC_FLAG_NODIR) {
998 if (match_basename(basename,
999 pathlen - (basename - pathname),
1000 exclude, prefix, x->patternlen,
1001 x->flags)) {
1002 exc = x;
1003 break;
1005 continue;
1008 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1009 if (match_pathname(pathname, pathlen,
1010 x->base, x->baselen ? x->baselen - 1 : 0,
1011 exclude, prefix, x->patternlen, x->flags)) {
1012 exc = x;
1013 break;
1016 return exc;
1020 * Scan the list and let the last match determine the fate.
1021 * Return 1 for exclude, 0 for include and -1 for undecided.
1023 int is_excluded_from_list(const char *pathname,
1024 int pathlen, const char *basename, int *dtype,
1025 struct exclude_list *el, struct index_state *istate)
1027 struct exclude *exclude;
1028 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1029 dtype, el, istate);
1030 if (exclude)
1031 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1032 return -1; /* undecided */
1035 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1036 struct index_state *istate,
1037 const char *pathname, int pathlen, const char *basename,
1038 int *dtype_p)
1040 int i, j;
1041 struct exclude_list_group *group;
1042 struct exclude *exclude;
1043 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1044 group = &dir->exclude_list_group[i];
1045 for (j = group->nr - 1; j >= 0; j--) {
1046 exclude = last_exclude_matching_from_list(
1047 pathname, pathlen, basename, dtype_p,
1048 &group->el[j], istate);
1049 if (exclude)
1050 return exclude;
1053 return NULL;
1057 * Loads the per-directory exclude list for the substring of base
1058 * which has a char length of baselen.
1060 static void prep_exclude(struct dir_struct *dir,
1061 struct index_state *istate,
1062 const char *base, int baselen)
1064 struct exclude_list_group *group;
1065 struct exclude_list *el;
1066 struct exclude_stack *stk = NULL;
1067 struct untracked_cache_dir *untracked;
1068 int current;
1070 group = &dir->exclude_list_group[EXC_DIRS];
1073 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1074 * which originate from directories not in the prefix of the
1075 * path being checked.
1077 while ((stk = dir->exclude_stack) != NULL) {
1078 if (stk->baselen <= baselen &&
1079 !strncmp(dir->basebuf.buf, base, stk->baselen))
1080 break;
1081 el = &group->el[dir->exclude_stack->exclude_ix];
1082 dir->exclude_stack = stk->prev;
1083 dir->exclude = NULL;
1084 free((char *)el->src); /* see strbuf_detach() below */
1085 clear_exclude_list(el);
1086 free(stk);
1087 group->nr--;
1090 /* Skip traversing into sub directories if the parent is excluded */
1091 if (dir->exclude)
1092 return;
1095 * Lazy initialization. All call sites currently just
1096 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1097 * them seems lots of work for little benefit.
1099 if (!dir->basebuf.buf)
1100 strbuf_init(&dir->basebuf, PATH_MAX);
1102 /* Read from the parent directories and push them down. */
1103 current = stk ? stk->baselen : -1;
1104 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1105 if (dir->untracked)
1106 untracked = stk ? stk->ucd : dir->untracked->root;
1107 else
1108 untracked = NULL;
1110 while (current < baselen) {
1111 const char *cp;
1112 struct sha1_stat sha1_stat;
1114 stk = xcalloc(1, sizeof(*stk));
1115 if (current < 0) {
1116 cp = base;
1117 current = 0;
1118 } else {
1119 cp = strchr(base + current + 1, '/');
1120 if (!cp)
1121 die("oops in prep_exclude");
1122 cp++;
1123 untracked =
1124 lookup_untracked(dir->untracked, untracked,
1125 base + current,
1126 cp - base - current);
1128 stk->prev = dir->exclude_stack;
1129 stk->baselen = cp - base;
1130 stk->exclude_ix = group->nr;
1131 stk->ucd = untracked;
1132 el = add_exclude_list(dir, EXC_DIRS, NULL);
1133 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1134 assert(stk->baselen == dir->basebuf.len);
1136 /* Abort if the directory is excluded */
1137 if (stk->baselen) {
1138 int dt = DT_DIR;
1139 dir->basebuf.buf[stk->baselen - 1] = 0;
1140 dir->exclude = last_exclude_matching_from_lists(dir,
1141 istate,
1142 dir->basebuf.buf, stk->baselen - 1,
1143 dir->basebuf.buf + current, &dt);
1144 dir->basebuf.buf[stk->baselen - 1] = '/';
1145 if (dir->exclude &&
1146 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1147 dir->exclude = NULL;
1148 if (dir->exclude) {
1149 dir->exclude_stack = stk;
1150 return;
1154 /* Try to read per-directory file */
1155 hashclr(sha1_stat.sha1);
1156 sha1_stat.valid = 0;
1157 if (dir->exclude_per_dir &&
1159 * If we know that no files have been added in
1160 * this directory (i.e. valid_cached_dir() has
1161 * been executed and set untracked->valid) ..
1163 (!untracked || !untracked->valid ||
1165 * .. and .gitignore does not exist before
1166 * (i.e. null exclude_sha1). Then we can skip
1167 * loading .gitignore, which would result in
1168 * ENOENT anyway.
1170 !is_null_sha1(untracked->exclude_sha1))) {
1172 * dir->basebuf gets reused by the traversal, but we
1173 * need fname to remain unchanged to ensure the src
1174 * member of each struct exclude correctly
1175 * back-references its source file. Other invocations
1176 * of add_exclude_list provide stable strings, so we
1177 * strbuf_detach() and free() here in the caller.
1179 struct strbuf sb = STRBUF_INIT;
1180 strbuf_addbuf(&sb, &dir->basebuf);
1181 strbuf_addstr(&sb, dir->exclude_per_dir);
1182 el->src = strbuf_detach(&sb, NULL);
1183 add_excludes(el->src, el->src, stk->baselen, el, istate,
1184 untracked ? &sha1_stat : NULL);
1187 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1188 * will first be called in valid_cached_dir() then maybe many
1189 * times more in last_exclude_matching(). When the cache is
1190 * used, last_exclude_matching() will not be called and
1191 * reading .gitignore content will be a waste.
1193 * So when it's called by valid_cached_dir() and we can get
1194 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1195 * modified on work tree), we could delay reading the
1196 * .gitignore content until we absolutely need it in
1197 * last_exclude_matching(). Be careful about ignore rule
1198 * order, though, if you do that.
1200 if (untracked &&
1201 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1202 invalidate_gitignore(dir->untracked, untracked);
1203 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1205 dir->exclude_stack = stk;
1206 current = stk->baselen;
1208 strbuf_setlen(&dir->basebuf, baselen);
1212 * Loads the exclude lists for the directory containing pathname, then
1213 * scans all exclude lists to determine whether pathname is excluded.
1214 * Returns the exclude_list element which matched, or NULL for
1215 * undecided.
1217 struct exclude *last_exclude_matching(struct dir_struct *dir,
1218 struct index_state *istate,
1219 const char *pathname,
1220 int *dtype_p)
1222 int pathlen = strlen(pathname);
1223 const char *basename = strrchr(pathname, '/');
1224 basename = (basename) ? basename+1 : pathname;
1226 prep_exclude(dir, istate, pathname, basename-pathname);
1228 if (dir->exclude)
1229 return dir->exclude;
1231 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
1232 basename, dtype_p);
1236 * Loads the exclude lists for the directory containing pathname, then
1237 * scans all exclude lists to determine whether pathname is excluded.
1238 * Returns 1 if true, otherwise 0.
1240 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1241 const char *pathname, int *dtype_p)
1243 struct exclude *exclude =
1244 last_exclude_matching(dir, istate, pathname, dtype_p);
1245 if (exclude)
1246 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1247 return 0;
1250 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1252 struct dir_entry *ent;
1254 FLEX_ALLOC_MEM(ent, name, pathname, len);
1255 ent->len = len;
1256 return ent;
1259 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1260 struct index_state *istate,
1261 const char *pathname, int len)
1263 if (index_file_exists(istate, pathname, len, ignore_case))
1264 return NULL;
1266 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1267 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1270 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1271 struct index_state *istate,
1272 const char *pathname, int len)
1274 if (!index_name_is_other(istate, pathname, len))
1275 return NULL;
1277 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1278 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1281 enum exist_status {
1282 index_nonexistent = 0,
1283 index_directory,
1284 index_gitdir
1288 * Do not use the alphabetically sorted index to look up
1289 * the directory name; instead, use the case insensitive
1290 * directory hash.
1292 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1293 const char *dirname, int len)
1295 struct cache_entry *ce;
1297 if (index_dir_exists(istate, dirname, len))
1298 return index_directory;
1300 ce = index_file_exists(istate, dirname, len, ignore_case);
1301 if (ce && S_ISGITLINK(ce->ce_mode))
1302 return index_gitdir;
1304 return index_nonexistent;
1308 * The index sorts alphabetically by entry name, which
1309 * means that a gitlink sorts as '\0' at the end, while
1310 * a directory (which is defined not as an entry, but as
1311 * the files it contains) will sort with the '/' at the
1312 * end.
1314 static enum exist_status directory_exists_in_index(struct index_state *istate,
1315 const char *dirname, int len)
1317 int pos;
1319 if (ignore_case)
1320 return directory_exists_in_index_icase(istate, dirname, len);
1322 pos = index_name_pos(istate, dirname, len);
1323 if (pos < 0)
1324 pos = -pos-1;
1325 while (pos < istate->cache_nr) {
1326 const struct cache_entry *ce = istate->cache[pos++];
1327 unsigned char endchar;
1329 if (strncmp(ce->name, dirname, len))
1330 break;
1331 endchar = ce->name[len];
1332 if (endchar > '/')
1333 break;
1334 if (endchar == '/')
1335 return index_directory;
1336 if (!endchar && S_ISGITLINK(ce->ce_mode))
1337 return index_gitdir;
1339 return index_nonexistent;
1343 * When we find a directory when traversing the filesystem, we
1344 * have three distinct cases:
1346 * - ignore it
1347 * - see it as a directory
1348 * - recurse into it
1350 * and which one we choose depends on a combination of existing
1351 * git index contents and the flags passed into the directory
1352 * traversal routine.
1354 * Case 1: If we *already* have entries in the index under that
1355 * directory name, we always recurse into the directory to see
1356 * all the files.
1358 * Case 2: If we *already* have that directory name as a gitlink,
1359 * we always continue to see it as a gitlink, regardless of whether
1360 * there is an actual git directory there or not (it might not
1361 * be checked out as a subproject!)
1363 * Case 3: if we didn't have it in the index previously, we
1364 * have a few sub-cases:
1366 * (a) if "show_other_directories" is true, we show it as
1367 * just a directory, unless "hide_empty_directories" is
1368 * also true, in which case we need to check if it contains any
1369 * untracked and / or ignored files.
1370 * (b) if it looks like a git directory, and we don't have
1371 * 'no_gitlinks' set we treat it as a gitlink, and show it
1372 * as a directory.
1373 * (c) otherwise, we recurse into it.
1375 static enum path_treatment treat_directory(struct dir_struct *dir,
1376 struct index_state *istate,
1377 struct untracked_cache_dir *untracked,
1378 const char *dirname, int len, int baselen, int exclude,
1379 const struct pathspec *pathspec)
1381 /* The "len-1" is to strip the final '/' */
1382 switch (directory_exists_in_index(istate, dirname, len-1)) {
1383 case index_directory:
1384 return path_recurse;
1386 case index_gitdir:
1387 return path_none;
1389 case index_nonexistent:
1390 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1391 break;
1392 if (!(dir->flags & DIR_NO_GITLINKS)) {
1393 unsigned char sha1[20];
1394 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1395 return path_untracked;
1397 return path_recurse;
1400 /* This is the "show_other_directories" case */
1402 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1403 return exclude ? path_excluded : path_untracked;
1405 untracked = lookup_untracked(dir->untracked, untracked,
1406 dirname + baselen, len - baselen);
1409 * If this is an excluded directory, then we only need to check if
1410 * the directory contains any files.
1412 return read_directory_recursive(dir, istate, dirname, len,
1413 untracked, 1, exclude, pathspec);
1417 * This is an inexact early pruning of any recursive directory
1418 * reading - if the path cannot possibly be in the pathspec,
1419 * return true, and we'll skip it early.
1421 static int simplify_away(const char *path, int pathlen,
1422 const struct pathspec *pathspec)
1424 int i;
1426 if (!pathspec || !pathspec->nr)
1427 return 0;
1429 GUARD_PATHSPEC(pathspec,
1430 PATHSPEC_FROMTOP |
1431 PATHSPEC_MAXDEPTH |
1432 PATHSPEC_LITERAL |
1433 PATHSPEC_GLOB |
1434 PATHSPEC_ICASE |
1435 PATHSPEC_EXCLUDE |
1436 PATHSPEC_ATTR);
1438 for (i = 0; i < pathspec->nr; i++) {
1439 const struct pathspec_item *item = &pathspec->items[i];
1440 int len = item->nowildcard_len;
1442 if (len > pathlen)
1443 len = pathlen;
1444 if (!ps_strncmp(item, item->match, path, len))
1445 return 0;
1448 return 1;
1452 * This function tells us whether an excluded path matches a
1453 * list of "interesting" pathspecs. That is, whether a path matched
1454 * by any of the pathspecs could possibly be ignored by excluding
1455 * the specified path. This can happen if:
1457 * 1. the path is mentioned explicitly in the pathspec
1459 * 2. the path is a directory prefix of some element in the
1460 * pathspec
1462 static int exclude_matches_pathspec(const char *path, int pathlen,
1463 const struct pathspec *pathspec)
1465 int i;
1467 if (!pathspec || !pathspec->nr)
1468 return 0;
1470 GUARD_PATHSPEC(pathspec,
1471 PATHSPEC_FROMTOP |
1472 PATHSPEC_MAXDEPTH |
1473 PATHSPEC_LITERAL |
1474 PATHSPEC_GLOB |
1475 PATHSPEC_ICASE |
1476 PATHSPEC_EXCLUDE);
1478 for (i = 0; i < pathspec->nr; i++) {
1479 const struct pathspec_item *item = &pathspec->items[i];
1480 int len = item->nowildcard_len;
1482 if (len == pathlen &&
1483 !ps_strncmp(item, item->match, path, pathlen))
1484 return 1;
1485 if (len > pathlen &&
1486 item->match[pathlen] == '/' &&
1487 !ps_strncmp(item, item->match, path, pathlen))
1488 return 1;
1490 return 0;
1493 static int get_index_dtype(struct index_state *istate,
1494 const char *path, int len)
1496 int pos;
1497 const struct cache_entry *ce;
1499 ce = index_file_exists(istate, path, len, 0);
1500 if (ce) {
1501 if (!ce_uptodate(ce))
1502 return DT_UNKNOWN;
1503 if (S_ISGITLINK(ce->ce_mode))
1504 return DT_DIR;
1506 * Nobody actually cares about the
1507 * difference between DT_LNK and DT_REG
1509 return DT_REG;
1512 /* Try to look it up as a directory */
1513 pos = index_name_pos(istate, path, len);
1514 if (pos >= 0)
1515 return DT_UNKNOWN;
1516 pos = -pos-1;
1517 while (pos < istate->cache_nr) {
1518 ce = istate->cache[pos++];
1519 if (strncmp(ce->name, path, len))
1520 break;
1521 if (ce->name[len] > '/')
1522 break;
1523 if (ce->name[len] < '/')
1524 continue;
1525 if (!ce_uptodate(ce))
1526 break; /* continue? */
1527 return DT_DIR;
1529 return DT_UNKNOWN;
1532 static int get_dtype(struct dirent *de, struct index_state *istate,
1533 const char *path, int len)
1535 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1536 struct stat st;
1538 if (dtype != DT_UNKNOWN)
1539 return dtype;
1540 dtype = get_index_dtype(istate, path, len);
1541 if (dtype != DT_UNKNOWN)
1542 return dtype;
1543 if (lstat(path, &st))
1544 return dtype;
1545 if (S_ISREG(st.st_mode))
1546 return DT_REG;
1547 if (S_ISDIR(st.st_mode))
1548 return DT_DIR;
1549 if (S_ISLNK(st.st_mode))
1550 return DT_LNK;
1551 return dtype;
1554 static enum path_treatment treat_one_path(struct dir_struct *dir,
1555 struct untracked_cache_dir *untracked,
1556 struct index_state *istate,
1557 struct strbuf *path,
1558 int baselen,
1559 const struct pathspec *pathspec,
1560 int dtype, struct dirent *de)
1562 int exclude;
1563 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1565 if (dtype == DT_UNKNOWN)
1566 dtype = get_dtype(de, istate, path->buf, path->len);
1568 /* Always exclude indexed files */
1569 if (dtype != DT_DIR && has_path_in_index)
1570 return path_none;
1573 * When we are looking at a directory P in the working tree,
1574 * there are three cases:
1576 * (1) P exists in the index. Everything inside the directory P in
1577 * the working tree needs to go when P is checked out from the
1578 * index.
1580 * (2) P does not exist in the index, but there is P/Q in the index.
1581 * We know P will stay a directory when we check out the contents
1582 * of the index, but we do not know yet if there is a directory
1583 * P/Q in the working tree to be killed, so we need to recurse.
1585 * (3) P does not exist in the index, and there is no P/Q in the index
1586 * to require P to be a directory, either. Only in this case, we
1587 * know that everything inside P will not be killed without
1588 * recursing.
1590 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1591 (dtype == DT_DIR) &&
1592 !has_path_in_index &&
1593 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1594 return path_none;
1596 exclude = is_excluded(dir, istate, path->buf, &dtype);
1599 * Excluded? If we don't explicitly want to show
1600 * ignored files, ignore it
1602 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1603 return path_excluded;
1605 switch (dtype) {
1606 default:
1607 return path_none;
1608 case DT_DIR:
1609 strbuf_addch(path, '/');
1610 return treat_directory(dir, istate, untracked, path->buf, path->len,
1611 baselen, exclude, pathspec);
1612 case DT_REG:
1613 case DT_LNK:
1614 return exclude ? path_excluded : path_untracked;
1618 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1619 struct untracked_cache_dir *untracked,
1620 struct cached_dir *cdir,
1621 struct index_state *istate,
1622 struct strbuf *path,
1623 int baselen,
1624 const struct pathspec *pathspec)
1626 strbuf_setlen(path, baselen);
1627 if (!cdir->ucd) {
1628 strbuf_addstr(path, cdir->file);
1629 return path_untracked;
1631 strbuf_addstr(path, cdir->ucd->name);
1632 /* treat_one_path() does this before it calls treat_directory() */
1633 strbuf_complete(path, '/');
1634 if (cdir->ucd->check_only)
1636 * check_only is set as a result of treat_directory() getting
1637 * to its bottom. Verify again the same set of directories
1638 * with check_only set.
1640 return read_directory_recursive(dir, istate, path->buf, path->len,
1641 cdir->ucd, 1, 0, pathspec);
1643 * We get path_recurse in the first run when
1644 * directory_exists_in_index() returns index_nonexistent. We
1645 * are sure that new changes in the index does not impact the
1646 * outcome. Return now.
1648 return path_recurse;
1651 static enum path_treatment treat_path(struct dir_struct *dir,
1652 struct untracked_cache_dir *untracked,
1653 struct cached_dir *cdir,
1654 struct index_state *istate,
1655 struct strbuf *path,
1656 int baselen,
1657 const struct pathspec *pathspec)
1659 int dtype;
1660 struct dirent *de = cdir->de;
1662 if (!de)
1663 return treat_path_fast(dir, untracked, cdir, istate, path,
1664 baselen, pathspec);
1665 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1666 return path_none;
1667 strbuf_setlen(path, baselen);
1668 strbuf_addstr(path, de->d_name);
1669 if (simplify_away(path->buf, path->len, pathspec))
1670 return path_none;
1672 dtype = DTYPE(de);
1673 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1676 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1678 if (!dir)
1679 return;
1680 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1681 dir->untracked_alloc);
1682 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1685 static int valid_cached_dir(struct dir_struct *dir,
1686 struct untracked_cache_dir *untracked,
1687 struct index_state *istate,
1688 struct strbuf *path,
1689 int check_only)
1691 struct stat st;
1693 if (!untracked)
1694 return 0;
1696 if (stat(path->len ? path->buf : ".", &st)) {
1697 invalidate_directory(dir->untracked, untracked);
1698 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1699 return 0;
1701 if (!untracked->valid ||
1702 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1703 if (untracked->valid)
1704 invalidate_directory(dir->untracked, untracked);
1705 fill_stat_data(&untracked->stat_data, &st);
1706 return 0;
1709 if (untracked->check_only != !!check_only) {
1710 invalidate_directory(dir->untracked, untracked);
1711 return 0;
1715 * prep_exclude will be called eventually on this directory,
1716 * but it's called much later in last_exclude_matching(). We
1717 * need it now to determine the validity of the cache for this
1718 * path. The next calls will be nearly no-op, the way
1719 * prep_exclude() is designed.
1721 if (path->len && path->buf[path->len - 1] != '/') {
1722 strbuf_addch(path, '/');
1723 prep_exclude(dir, istate, path->buf, path->len);
1724 strbuf_setlen(path, path->len - 1);
1725 } else
1726 prep_exclude(dir, istate, path->buf, path->len);
1728 /* hopefully prep_exclude() haven't invalidated this entry... */
1729 return untracked->valid;
1732 static int open_cached_dir(struct cached_dir *cdir,
1733 struct dir_struct *dir,
1734 struct untracked_cache_dir *untracked,
1735 struct index_state *istate,
1736 struct strbuf *path,
1737 int check_only)
1739 memset(cdir, 0, sizeof(*cdir));
1740 cdir->untracked = untracked;
1741 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1742 return 0;
1743 cdir->fdir = opendir(path->len ? path->buf : ".");
1744 if (dir->untracked)
1745 dir->untracked->dir_opened++;
1746 if (!cdir->fdir)
1747 return -1;
1748 return 0;
1751 static int read_cached_dir(struct cached_dir *cdir)
1753 if (cdir->fdir) {
1754 cdir->de = readdir(cdir->fdir);
1755 if (!cdir->de)
1756 return -1;
1757 return 0;
1759 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1760 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1761 if (!d->recurse) {
1762 cdir->nr_dirs++;
1763 continue;
1765 cdir->ucd = d;
1766 cdir->nr_dirs++;
1767 return 0;
1769 cdir->ucd = NULL;
1770 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1771 struct untracked_cache_dir *d = cdir->untracked;
1772 cdir->file = d->untracked[cdir->nr_files++];
1773 return 0;
1775 return -1;
1778 static void close_cached_dir(struct cached_dir *cdir)
1780 if (cdir->fdir)
1781 closedir(cdir->fdir);
1783 * We have gone through this directory and found no untracked
1784 * entries. Mark it valid.
1786 if (cdir->untracked) {
1787 cdir->untracked->valid = 1;
1788 cdir->untracked->recurse = 1;
1793 * Read a directory tree. We currently ignore anything but
1794 * directories, regular files and symlinks. That's because git
1795 * doesn't handle them at all yet. Maybe that will change some
1796 * day.
1798 * Also, we ignore the name ".git" (even if it is not a directory).
1799 * That likely will not change.
1801 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1802 * to signal that a file was found. This is the least significant value that
1803 * indicates that a file was encountered that does not depend on the order of
1804 * whether an untracked or exluded path was encountered first.
1806 * Returns the most significant path_treatment value encountered in the scan.
1807 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1808 * significant path_treatment value that will be returned.
1811 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1812 struct index_state *istate, const char *base, int baselen,
1813 struct untracked_cache_dir *untracked, int check_only,
1814 int stop_at_first_file, const struct pathspec *pathspec)
1816 struct cached_dir cdir;
1817 enum path_treatment state, subdir_state, dir_state = path_none;
1818 struct strbuf path = STRBUF_INIT;
1820 strbuf_add(&path, base, baselen);
1822 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1823 goto out;
1825 if (untracked)
1826 untracked->check_only = !!check_only;
1828 while (!read_cached_dir(&cdir)) {
1829 /* check how the file or directory should be treated */
1830 state = treat_path(dir, untracked, &cdir, istate, &path,
1831 baselen, pathspec);
1833 if (state > dir_state)
1834 dir_state = state;
1836 /* recurse into subdir if instructed by treat_path */
1837 if ((state == path_recurse) ||
1838 ((state == path_untracked) &&
1839 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1840 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
1841 struct untracked_cache_dir *ud;
1842 ud = lookup_untracked(dir->untracked, untracked,
1843 path.buf + baselen,
1844 path.len - baselen);
1845 subdir_state =
1846 read_directory_recursive(dir, istate, path.buf,
1847 path.len, ud,
1848 check_only, stop_at_first_file, pathspec);
1849 if (subdir_state > dir_state)
1850 dir_state = subdir_state;
1853 if (check_only) {
1854 if (stop_at_first_file) {
1856 * If stopping at first file, then
1857 * signal that a file was found by
1858 * returning `path_excluded`. This is
1859 * to return a consistent value
1860 * regardless of whether an ignored or
1861 * excluded file happened to be
1862 * encountered 1st.
1864 * In current usage, the
1865 * `stop_at_first_file` is passed when
1866 * an ancestor directory has matched
1867 * an exclude pattern, so any found
1868 * files will be excluded.
1870 if (dir_state >= path_excluded) {
1871 dir_state = path_excluded;
1872 break;
1876 /* abort early if maximum state has been reached */
1877 if (dir_state == path_untracked) {
1878 if (cdir.fdir)
1879 add_untracked(untracked, path.buf + baselen);
1880 break;
1882 /* skip the dir_add_* part */
1883 continue;
1886 /* add the path to the appropriate result list */
1887 switch (state) {
1888 case path_excluded:
1889 if (dir->flags & DIR_SHOW_IGNORED)
1890 dir_add_name(dir, istate, path.buf, path.len);
1891 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1892 ((dir->flags & DIR_COLLECT_IGNORED) &&
1893 exclude_matches_pathspec(path.buf, path.len,
1894 pathspec)))
1895 dir_add_ignored(dir, istate, path.buf, path.len);
1896 break;
1898 case path_untracked:
1899 if (dir->flags & DIR_SHOW_IGNORED)
1900 break;
1901 dir_add_name(dir, istate, path.buf, path.len);
1902 if (cdir.fdir)
1903 add_untracked(untracked, path.buf + baselen);
1904 break;
1906 default:
1907 break;
1910 close_cached_dir(&cdir);
1911 out:
1912 strbuf_release(&path);
1914 return dir_state;
1917 int cmp_dir_entry(const void *p1, const void *p2)
1919 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1920 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1922 return name_compare(e1->name, e1->len, e2->name, e2->len);
1925 /* check if *out lexically strictly contains *in */
1926 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
1928 return (out->len < in->len) &&
1929 (out->name[out->len - 1] == '/') &&
1930 !memcmp(out->name, in->name, out->len);
1933 static int treat_leading_path(struct dir_struct *dir,
1934 struct index_state *istate,
1935 const char *path, int len,
1936 const struct pathspec *pathspec)
1938 struct strbuf sb = STRBUF_INIT;
1939 int baselen, rc = 0;
1940 const char *cp;
1941 int old_flags = dir->flags;
1943 while (len && path[len - 1] == '/')
1944 len--;
1945 if (!len)
1946 return 1;
1947 baselen = 0;
1948 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1949 while (1) {
1950 cp = path + baselen + !!baselen;
1951 cp = memchr(cp, '/', path + len - cp);
1952 if (!cp)
1953 baselen = len;
1954 else
1955 baselen = cp - path;
1956 strbuf_setlen(&sb, 0);
1957 strbuf_add(&sb, path, baselen);
1958 if (!is_directory(sb.buf))
1959 break;
1960 if (simplify_away(sb.buf, sb.len, pathspec))
1961 break;
1962 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
1963 DT_DIR, NULL) == path_none)
1964 break; /* do not recurse into it */
1965 if (len <= baselen) {
1966 rc = 1;
1967 break; /* finished checking */
1970 strbuf_release(&sb);
1971 dir->flags = old_flags;
1972 return rc;
1975 static const char *get_ident_string(void)
1977 static struct strbuf sb = STRBUF_INIT;
1978 struct utsname uts;
1980 if (sb.len)
1981 return sb.buf;
1982 if (uname(&uts) < 0)
1983 die_errno(_("failed to get kernel name and information"));
1984 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
1985 uts.sysname);
1986 return sb.buf;
1989 static int ident_in_untracked(const struct untracked_cache *uc)
1992 * Previous git versions may have saved many NUL separated
1993 * strings in the "ident" field, but it is insane to manage
1994 * many locations, so just take care of the first one.
1997 return !strcmp(uc->ident.buf, get_ident_string());
2000 static void set_untracked_ident(struct untracked_cache *uc)
2002 strbuf_reset(&uc->ident);
2003 strbuf_addstr(&uc->ident, get_ident_string());
2006 * This strbuf used to contain a list of NUL separated
2007 * strings, so save NUL too for backward compatibility.
2009 strbuf_addch(&uc->ident, 0);
2012 static void new_untracked_cache(struct index_state *istate)
2014 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2015 strbuf_init(&uc->ident, 100);
2016 uc->exclude_per_dir = ".gitignore";
2017 /* should be the same flags used by git-status */
2018 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2019 set_untracked_ident(uc);
2020 istate->untracked = uc;
2021 istate->cache_changed |= UNTRACKED_CHANGED;
2024 void add_untracked_cache(struct index_state *istate)
2026 if (!istate->untracked) {
2027 new_untracked_cache(istate);
2028 } else {
2029 if (!ident_in_untracked(istate->untracked)) {
2030 free_untracked_cache(istate->untracked);
2031 new_untracked_cache(istate);
2036 void remove_untracked_cache(struct index_state *istate)
2038 if (istate->untracked) {
2039 free_untracked_cache(istate->untracked);
2040 istate->untracked = NULL;
2041 istate->cache_changed |= UNTRACKED_CHANGED;
2045 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2046 int base_len,
2047 const struct pathspec *pathspec)
2049 struct untracked_cache_dir *root;
2051 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
2052 return NULL;
2055 * We only support $GIT_DIR/info/exclude and core.excludesfile
2056 * as the global ignore rule files. Any other additions
2057 * (e.g. from command line) invalidate the cache. This
2058 * condition also catches running setup_standard_excludes()
2059 * before setting dir->untracked!
2061 if (dir->unmanaged_exclude_files)
2062 return NULL;
2065 * Optimize for the main use case only: whole-tree git
2066 * status. More work involved in treat_leading_path() if we
2067 * use cache on just a subset of the worktree. pathspec
2068 * support could make the matter even worse.
2070 if (base_len || (pathspec && pathspec->nr))
2071 return NULL;
2073 /* Different set of flags may produce different results */
2074 if (dir->flags != dir->untracked->dir_flags ||
2076 * See treat_directory(), case index_nonexistent. Without
2077 * this flag, we may need to also cache .git file content
2078 * for the resolve_gitlink_ref() call, which we don't.
2080 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2081 /* We don't support collecting ignore files */
2082 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2083 DIR_COLLECT_IGNORED)))
2084 return NULL;
2087 * If we use .gitignore in the cache and now you change it to
2088 * .gitexclude, everything will go wrong.
2090 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2091 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2092 return NULL;
2095 * EXC_CMDL is not considered in the cache. If people set it,
2096 * skip the cache.
2098 if (dir->exclude_list_group[EXC_CMDL].nr)
2099 return NULL;
2101 if (!ident_in_untracked(dir->untracked)) {
2102 warning(_("Untracked cache is disabled on this system or location."));
2103 return NULL;
2106 if (!dir->untracked->root) {
2107 const int len = sizeof(*dir->untracked->root);
2108 dir->untracked->root = xmalloc(len);
2109 memset(dir->untracked->root, 0, len);
2112 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2113 root = dir->untracked->root;
2114 if (hashcmp(dir->ss_info_exclude.sha1,
2115 dir->untracked->ss_info_exclude.sha1)) {
2116 invalidate_gitignore(dir->untracked, root);
2117 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2119 if (hashcmp(dir->ss_excludes_file.sha1,
2120 dir->untracked->ss_excludes_file.sha1)) {
2121 invalidate_gitignore(dir->untracked, root);
2122 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2125 /* Make sure this directory is not dropped out at saving phase */
2126 root->recurse = 1;
2127 return root;
2130 int read_directory(struct dir_struct *dir, struct index_state *istate,
2131 const char *path, int len, const struct pathspec *pathspec)
2133 struct untracked_cache_dir *untracked;
2135 if (has_symlink_leading_path(path, len))
2136 return dir->nr;
2138 untracked = validate_untracked_cache(dir, len, pathspec);
2139 if (!untracked)
2141 * make sure untracked cache code path is disabled,
2142 * e.g. prep_exclude()
2144 dir->untracked = NULL;
2145 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2146 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2147 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2148 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2151 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2152 * also pick up untracked contents of untracked dirs; by default
2153 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2155 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2156 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2157 int i, j;
2159 /* remove from dir->entries untracked contents of untracked dirs */
2160 for (i = j = 0; j < dir->nr; j++) {
2161 if (i &&
2162 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2163 FREE_AND_NULL(dir->entries[j]);
2164 } else {
2165 dir->entries[i++] = dir->entries[j];
2169 dir->nr = i;
2172 if (dir->untracked) {
2173 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2174 trace_printf_key(&trace_untracked_stats,
2175 "node creation: %u\n"
2176 "gitignore invalidation: %u\n"
2177 "directory invalidation: %u\n"
2178 "opendir: %u\n",
2179 dir->untracked->dir_created,
2180 dir->untracked->gitignore_invalidated,
2181 dir->untracked->dir_invalidated,
2182 dir->untracked->dir_opened);
2183 if (dir->untracked == istate->untracked &&
2184 (dir->untracked->dir_opened ||
2185 dir->untracked->gitignore_invalidated ||
2186 dir->untracked->dir_invalidated))
2187 istate->cache_changed |= UNTRACKED_CHANGED;
2188 if (dir->untracked != istate->untracked) {
2189 FREE_AND_NULL(dir->untracked);
2192 return dir->nr;
2195 int file_exists(const char *f)
2197 struct stat sb;
2198 return lstat(f, &sb) == 0;
2201 static int cmp_icase(char a, char b)
2203 if (a == b)
2204 return 0;
2205 if (ignore_case)
2206 return toupper(a) - toupper(b);
2207 return a - b;
2211 * Given two normalized paths (a trailing slash is ok), if subdir is
2212 * outside dir, return -1. Otherwise return the offset in subdir that
2213 * can be used as relative path to dir.
2215 int dir_inside_of(const char *subdir, const char *dir)
2217 int offset = 0;
2219 assert(dir && subdir && *dir && *subdir);
2221 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2222 dir++;
2223 subdir++;
2224 offset++;
2227 /* hel[p]/me vs hel[l]/yeah */
2228 if (*dir && *subdir)
2229 return -1;
2231 if (!*subdir)
2232 return !*dir ? offset : -1; /* same dir */
2234 /* foo/[b]ar vs foo/[] */
2235 if (is_dir_sep(dir[-1]))
2236 return is_dir_sep(subdir[-1]) ? offset : -1;
2238 /* foo[/]bar vs foo[] */
2239 return is_dir_sep(*subdir) ? offset + 1 : -1;
2242 int is_inside_dir(const char *dir)
2244 char *cwd;
2245 int rc;
2247 if (!dir)
2248 return 0;
2250 cwd = xgetcwd();
2251 rc = (dir_inside_of(cwd, dir) >= 0);
2252 free(cwd);
2253 return rc;
2256 int is_empty_dir(const char *path)
2258 DIR *dir = opendir(path);
2259 struct dirent *e;
2260 int ret = 1;
2262 if (!dir)
2263 return 0;
2265 while ((e = readdir(dir)) != NULL)
2266 if (!is_dot_or_dotdot(e->d_name)) {
2267 ret = 0;
2268 break;
2271 closedir(dir);
2272 return ret;
2275 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2277 DIR *dir;
2278 struct dirent *e;
2279 int ret = 0, original_len = path->len, len, kept_down = 0;
2280 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2281 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2282 unsigned char submodule_head[20];
2284 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2285 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2286 /* Do not descend and nuke a nested git work tree. */
2287 if (kept_up)
2288 *kept_up = 1;
2289 return 0;
2292 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2293 dir = opendir(path->buf);
2294 if (!dir) {
2295 if (errno == ENOENT)
2296 return keep_toplevel ? -1 : 0;
2297 else if (errno == EACCES && !keep_toplevel)
2299 * An empty dir could be removable even if it
2300 * is unreadable:
2302 return rmdir(path->buf);
2303 else
2304 return -1;
2306 strbuf_complete(path, '/');
2308 len = path->len;
2309 while ((e = readdir(dir)) != NULL) {
2310 struct stat st;
2311 if (is_dot_or_dotdot(e->d_name))
2312 continue;
2314 strbuf_setlen(path, len);
2315 strbuf_addstr(path, e->d_name);
2316 if (lstat(path->buf, &st)) {
2317 if (errno == ENOENT)
2319 * file disappeared, which is what we
2320 * wanted anyway
2322 continue;
2323 /* fall thru */
2324 } else if (S_ISDIR(st.st_mode)) {
2325 if (!remove_dir_recurse(path, flag, &kept_down))
2326 continue; /* happy */
2327 } else if (!only_empty &&
2328 (!unlink(path->buf) || errno == ENOENT)) {
2329 continue; /* happy, too */
2332 /* path too long, stat fails, or non-directory still exists */
2333 ret = -1;
2334 break;
2336 closedir(dir);
2338 strbuf_setlen(path, original_len);
2339 if (!ret && !keep_toplevel && !kept_down)
2340 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2341 else if (kept_up)
2343 * report the uplevel that it is not an error that we
2344 * did not rmdir() our directory.
2346 *kept_up = !ret;
2347 return ret;
2350 int remove_dir_recursively(struct strbuf *path, int flag)
2352 return remove_dir_recurse(path, flag, NULL);
2355 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2357 void setup_standard_excludes(struct dir_struct *dir)
2359 dir->exclude_per_dir = ".gitignore";
2361 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2362 if (!excludes_file)
2363 excludes_file = xdg_config_home("ignore");
2364 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2365 add_excludes_from_file_1(dir, excludes_file,
2366 dir->untracked ? &dir->ss_excludes_file : NULL);
2368 /* per repository user preference */
2369 if (startup_info->have_repository) {
2370 const char *path = git_path_info_exclude();
2371 if (!access_or_warn(path, R_OK, 0))
2372 add_excludes_from_file_1(dir, path,
2373 dir->untracked ? &dir->ss_info_exclude : NULL);
2377 int remove_path(const char *name)
2379 char *slash;
2381 if (unlink(name) && !is_missing_file_error(errno))
2382 return -1;
2384 slash = strrchr(name, '/');
2385 if (slash) {
2386 char *dirs = xstrdup(name);
2387 slash = dirs + (slash - name);
2388 do {
2389 *slash = '\0';
2390 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2391 free(dirs);
2393 return 0;
2397 * Frees memory within dir which was allocated for exclude lists and
2398 * the exclude_stack. Does not free dir itself.
2400 void clear_directory(struct dir_struct *dir)
2402 int i, j;
2403 struct exclude_list_group *group;
2404 struct exclude_list *el;
2405 struct exclude_stack *stk;
2407 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2408 group = &dir->exclude_list_group[i];
2409 for (j = 0; j < group->nr; j++) {
2410 el = &group->el[j];
2411 if (i == EXC_DIRS)
2412 free((char *)el->src);
2413 clear_exclude_list(el);
2415 free(group->el);
2418 stk = dir->exclude_stack;
2419 while (stk) {
2420 struct exclude_stack *prev = stk->prev;
2421 free(stk);
2422 stk = prev;
2424 strbuf_release(&dir->basebuf);
2427 struct ondisk_untracked_cache {
2428 struct stat_data info_exclude_stat;
2429 struct stat_data excludes_file_stat;
2430 uint32_t dir_flags;
2431 unsigned char info_exclude_sha1[20];
2432 unsigned char excludes_file_sha1[20];
2433 char exclude_per_dir[FLEX_ARRAY];
2436 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2437 #define ouc_size(len) (ouc_offset(exclude_per_dir) + len + 1)
2439 struct write_data {
2440 int index; /* number of written untracked_cache_dir */
2441 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2442 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2443 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2444 struct strbuf out;
2445 struct strbuf sb_stat;
2446 struct strbuf sb_sha1;
2449 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2451 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2452 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2453 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2454 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2455 to->sd_dev = htonl(from->sd_dev);
2456 to->sd_ino = htonl(from->sd_ino);
2457 to->sd_uid = htonl(from->sd_uid);
2458 to->sd_gid = htonl(from->sd_gid);
2459 to->sd_size = htonl(from->sd_size);
2462 static void write_one_dir(struct untracked_cache_dir *untracked,
2463 struct write_data *wd)
2465 struct stat_data stat_data;
2466 struct strbuf *out = &wd->out;
2467 unsigned char intbuf[16];
2468 unsigned int intlen, value;
2469 int i = wd->index++;
2472 * untracked_nr should be reset whenever valid is clear, but
2473 * for safety..
2475 if (!untracked->valid) {
2476 untracked->untracked_nr = 0;
2477 untracked->check_only = 0;
2480 if (untracked->check_only)
2481 ewah_set(wd->check_only, i);
2482 if (untracked->valid) {
2483 ewah_set(wd->valid, i);
2484 stat_data_to_disk(&stat_data, &untracked->stat_data);
2485 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2487 if (!is_null_sha1(untracked->exclude_sha1)) {
2488 ewah_set(wd->sha1_valid, i);
2489 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2492 intlen = encode_varint(untracked->untracked_nr, intbuf);
2493 strbuf_add(out, intbuf, intlen);
2495 /* skip non-recurse directories */
2496 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2497 if (untracked->dirs[i]->recurse)
2498 value++;
2499 intlen = encode_varint(value, intbuf);
2500 strbuf_add(out, intbuf, intlen);
2502 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2504 for (i = 0; i < untracked->untracked_nr; i++)
2505 strbuf_add(out, untracked->untracked[i],
2506 strlen(untracked->untracked[i]) + 1);
2508 for (i = 0; i < untracked->dirs_nr; i++)
2509 if (untracked->dirs[i]->recurse)
2510 write_one_dir(untracked->dirs[i], wd);
2513 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2515 struct ondisk_untracked_cache *ouc;
2516 struct write_data wd;
2517 unsigned char varbuf[16];
2518 int varint_len;
2519 size_t len = strlen(untracked->exclude_per_dir);
2521 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
2522 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2523 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2524 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2525 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2526 ouc->dir_flags = htonl(untracked->dir_flags);
2528 varint_len = encode_varint(untracked->ident.len, varbuf);
2529 strbuf_add(out, varbuf, varint_len);
2530 strbuf_addbuf(out, &untracked->ident);
2532 strbuf_add(out, ouc, ouc_size(len));
2533 FREE_AND_NULL(ouc);
2535 if (!untracked->root) {
2536 varint_len = encode_varint(0, varbuf);
2537 strbuf_add(out, varbuf, varint_len);
2538 return;
2541 wd.index = 0;
2542 wd.check_only = ewah_new();
2543 wd.valid = ewah_new();
2544 wd.sha1_valid = ewah_new();
2545 strbuf_init(&wd.out, 1024);
2546 strbuf_init(&wd.sb_stat, 1024);
2547 strbuf_init(&wd.sb_sha1, 1024);
2548 write_one_dir(untracked->root, &wd);
2550 varint_len = encode_varint(wd.index, varbuf);
2551 strbuf_add(out, varbuf, varint_len);
2552 strbuf_addbuf(out, &wd.out);
2553 ewah_serialize_strbuf(wd.valid, out);
2554 ewah_serialize_strbuf(wd.check_only, out);
2555 ewah_serialize_strbuf(wd.sha1_valid, out);
2556 strbuf_addbuf(out, &wd.sb_stat);
2557 strbuf_addbuf(out, &wd.sb_sha1);
2558 strbuf_addch(out, '\0'); /* safe guard for string lists */
2560 ewah_free(wd.valid);
2561 ewah_free(wd.check_only);
2562 ewah_free(wd.sha1_valid);
2563 strbuf_release(&wd.out);
2564 strbuf_release(&wd.sb_stat);
2565 strbuf_release(&wd.sb_sha1);
2568 static void free_untracked(struct untracked_cache_dir *ucd)
2570 int i;
2571 if (!ucd)
2572 return;
2573 for (i = 0; i < ucd->dirs_nr; i++)
2574 free_untracked(ucd->dirs[i]);
2575 for (i = 0; i < ucd->untracked_nr; i++)
2576 free(ucd->untracked[i]);
2577 free(ucd->untracked);
2578 free(ucd->dirs);
2579 free(ucd);
2582 void free_untracked_cache(struct untracked_cache *uc)
2584 if (uc)
2585 free_untracked(uc->root);
2586 free(uc);
2589 struct read_data {
2590 int index;
2591 struct untracked_cache_dir **ucd;
2592 struct ewah_bitmap *check_only;
2593 struct ewah_bitmap *valid;
2594 struct ewah_bitmap *sha1_valid;
2595 const unsigned char *data;
2596 const unsigned char *end;
2599 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2601 memcpy(to, data, sizeof(*to));
2602 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2603 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2604 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2605 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2606 to->sd_dev = ntohl(to->sd_dev);
2607 to->sd_ino = ntohl(to->sd_ino);
2608 to->sd_uid = ntohl(to->sd_uid);
2609 to->sd_gid = ntohl(to->sd_gid);
2610 to->sd_size = ntohl(to->sd_size);
2613 static int read_one_dir(struct untracked_cache_dir **untracked_,
2614 struct read_data *rd)
2616 struct untracked_cache_dir ud, *untracked;
2617 const unsigned char *next, *data = rd->data, *end = rd->end;
2618 unsigned int value;
2619 int i, len;
2621 memset(&ud, 0, sizeof(ud));
2623 next = data;
2624 value = decode_varint(&next);
2625 if (next > end)
2626 return -1;
2627 ud.recurse = 1;
2628 ud.untracked_alloc = value;
2629 ud.untracked_nr = value;
2630 if (ud.untracked_nr)
2631 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2632 data = next;
2634 next = data;
2635 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2636 if (next > end)
2637 return -1;
2638 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2639 data = next;
2641 len = strlen((const char *)data);
2642 next = data + len + 1;
2643 if (next > rd->end)
2644 return -1;
2645 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
2646 memcpy(untracked, &ud, sizeof(ud));
2647 memcpy(untracked->name, data, len + 1);
2648 data = next;
2650 for (i = 0; i < untracked->untracked_nr; i++) {
2651 len = strlen((const char *)data);
2652 next = data + len + 1;
2653 if (next > rd->end)
2654 return -1;
2655 untracked->untracked[i] = xstrdup((const char*)data);
2656 data = next;
2659 rd->ucd[rd->index++] = untracked;
2660 rd->data = data;
2662 for (i = 0; i < untracked->dirs_nr; i++) {
2663 len = read_one_dir(untracked->dirs + i, rd);
2664 if (len < 0)
2665 return -1;
2667 return 0;
2670 static void set_check_only(size_t pos, void *cb)
2672 struct read_data *rd = cb;
2673 struct untracked_cache_dir *ud = rd->ucd[pos];
2674 ud->check_only = 1;
2677 static void read_stat(size_t pos, void *cb)
2679 struct read_data *rd = cb;
2680 struct untracked_cache_dir *ud = rd->ucd[pos];
2681 if (rd->data + sizeof(struct stat_data) > rd->end) {
2682 rd->data = rd->end + 1;
2683 return;
2685 stat_data_from_disk(&ud->stat_data, rd->data);
2686 rd->data += sizeof(struct stat_data);
2687 ud->valid = 1;
2690 static void read_sha1(size_t pos, void *cb)
2692 struct read_data *rd = cb;
2693 struct untracked_cache_dir *ud = rd->ucd[pos];
2694 if (rd->data + 20 > rd->end) {
2695 rd->data = rd->end + 1;
2696 return;
2698 hashcpy(ud->exclude_sha1, rd->data);
2699 rd->data += 20;
2702 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2703 const unsigned char *data,
2704 const unsigned char *sha1)
2706 stat_data_from_disk(&sha1_stat->stat, data);
2707 hashcpy(sha1_stat->sha1, sha1);
2708 sha1_stat->valid = 1;
2711 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2713 struct untracked_cache *uc;
2714 struct read_data rd;
2715 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2716 const char *ident;
2717 int ident_len, len;
2718 const char *exclude_per_dir;
2720 if (sz <= 1 || end[-1] != '\0')
2721 return NULL;
2722 end--;
2724 ident_len = decode_varint(&next);
2725 if (next + ident_len > end)
2726 return NULL;
2727 ident = (const char *)next;
2728 next += ident_len;
2730 if (next + ouc_size(0) > end)
2731 return NULL;
2733 uc = xcalloc(1, sizeof(*uc));
2734 strbuf_init(&uc->ident, ident_len);
2735 strbuf_add(&uc->ident, ident, ident_len);
2736 load_sha1_stat(&uc->ss_info_exclude,
2737 next + ouc_offset(info_exclude_stat),
2738 next + ouc_offset(info_exclude_sha1));
2739 load_sha1_stat(&uc->ss_excludes_file,
2740 next + ouc_offset(excludes_file_stat),
2741 next + ouc_offset(excludes_file_sha1));
2742 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2743 exclude_per_dir = (const char *)next + ouc_offset(exclude_per_dir);
2744 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2745 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2746 next += ouc_size(strlen(exclude_per_dir));
2747 if (next >= end)
2748 goto done2;
2750 len = decode_varint(&next);
2751 if (next > end || len == 0)
2752 goto done2;
2754 rd.valid = ewah_new();
2755 rd.check_only = ewah_new();
2756 rd.sha1_valid = ewah_new();
2757 rd.data = next;
2758 rd.end = end;
2759 rd.index = 0;
2760 ALLOC_ARRAY(rd.ucd, len);
2762 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2763 goto done;
2765 next = rd.data;
2766 len = ewah_read_mmap(rd.valid, next, end - next);
2767 if (len < 0)
2768 goto done;
2770 next += len;
2771 len = ewah_read_mmap(rd.check_only, next, end - next);
2772 if (len < 0)
2773 goto done;
2775 next += len;
2776 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2777 if (len < 0)
2778 goto done;
2780 ewah_each_bit(rd.check_only, set_check_only, &rd);
2781 rd.data = next + len;
2782 ewah_each_bit(rd.valid, read_stat, &rd);
2783 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2784 next = rd.data;
2786 done:
2787 free(rd.ucd);
2788 ewah_free(rd.valid);
2789 ewah_free(rd.check_only);
2790 ewah_free(rd.sha1_valid);
2791 done2:
2792 if (next != end) {
2793 free_untracked_cache(uc);
2794 uc = NULL;
2796 return uc;
2799 static void invalidate_one_directory(struct untracked_cache *uc,
2800 struct untracked_cache_dir *ucd)
2802 uc->dir_invalidated++;
2803 ucd->valid = 0;
2804 ucd->untracked_nr = 0;
2808 * Normally when an entry is added or removed from a directory,
2809 * invalidating that directory is enough. No need to touch its
2810 * ancestors. When a directory is shown as "foo/bar/" in git-status
2811 * however, deleting or adding an entry may have cascading effect.
2813 * Say the "foo/bar/file" has become untracked, we need to tell the
2814 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2815 * directory any more (because "bar" is managed by foo as an untracked
2816 * "file").
2818 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2819 * was the last untracked entry in the entire "foo", we should show
2820 * "foo/" instead. Which means we have to invalidate past "bar" up to
2821 * "foo".
2823 * This function traverses all directories from root to leaf. If there
2824 * is a chance of one of the above cases happening, we invalidate back
2825 * to root. Otherwise we just invalidate the leaf. There may be a more
2826 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2827 * detect these cases and avoid unnecessary invalidation, for example,
2828 * checking for the untracked entry named "bar/" in "foo", but for now
2829 * stick to something safe and simple.
2831 static int invalidate_one_component(struct untracked_cache *uc,
2832 struct untracked_cache_dir *dir,
2833 const char *path, int len)
2835 const char *rest = strchr(path, '/');
2837 if (rest) {
2838 int component_len = rest - path;
2839 struct untracked_cache_dir *d =
2840 lookup_untracked(uc, dir, path, component_len);
2841 int ret =
2842 invalidate_one_component(uc, d, rest + 1,
2843 len - (component_len + 1));
2844 if (ret)
2845 invalidate_one_directory(uc, dir);
2846 return ret;
2849 invalidate_one_directory(uc, dir);
2850 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2853 void untracked_cache_invalidate_path(struct index_state *istate,
2854 const char *path)
2856 if (!istate->untracked || !istate->untracked->root)
2857 return;
2858 invalidate_one_component(istate->untracked, istate->untracked->root,
2859 path, strlen(path));
2862 void untracked_cache_remove_from_index(struct index_state *istate,
2863 const char *path)
2865 untracked_cache_invalidate_path(istate, path);
2868 void untracked_cache_add_to_index(struct index_state *istate,
2869 const char *path)
2871 untracked_cache_invalidate_path(istate, path);
2874 /* Update gitfile and core.worktree setting to connect work tree and git dir */
2875 void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)
2877 struct strbuf gitfile_sb = STRBUF_INIT;
2878 struct strbuf cfg_sb = STRBUF_INIT;
2879 struct strbuf rel_path = STRBUF_INIT;
2880 char *git_dir, *work_tree;
2882 /* Prepare .git file */
2883 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
2884 if (safe_create_leading_directories_const(gitfile_sb.buf))
2885 die(_("could not create directories for %s"), gitfile_sb.buf);
2887 /* Prepare config file */
2888 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
2889 if (safe_create_leading_directories_const(cfg_sb.buf))
2890 die(_("could not create directories for %s"), cfg_sb.buf);
2892 git_dir = real_pathdup(git_dir_, 1);
2893 work_tree = real_pathdup(work_tree_, 1);
2895 /* Write .git file */
2896 write_file(gitfile_sb.buf, "gitdir: %s",
2897 relative_path(git_dir, work_tree, &rel_path));
2898 /* Update core.worktree setting */
2899 git_config_set_in_file(cfg_sb.buf, "core.worktree",
2900 relative_path(work_tree, git_dir, &rel_path));
2902 strbuf_release(&gitfile_sb);
2903 strbuf_release(&cfg_sb);
2904 strbuf_release(&rel_path);
2905 free(work_tree);
2906 free(git_dir);
2910 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
2912 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
2914 if (rename(old_git_dir, new_git_dir) < 0)
2915 die_errno(_("could not migrate git directory from '%s' to '%s'"),
2916 old_git_dir, new_git_dir);
2918 connect_work_tree_and_git_dir(path, new_git_dir);