git-multimail: update to release 1.2.0
[git.git] / dir.c
blobb90484acf8831150bce34670315e00d2f6f6480b
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 #include "cache.h"
11 #include "dir.h"
12 #include "refs.h"
13 #include "wildmatch.h"
14 #include "pathspec.h"
15 #include "utf8.h"
16 #include "varint.h"
17 #include "ewah/ewok.h"
19 struct path_simplify {
20 int len;
21 const char *path;
25 * Tells read_directory_recursive how a file or directory should be treated.
26 * Values are ordered by significance, e.g. if a directory contains both
27 * excluded and untracked files, it is listed as untracked because
28 * path_untracked > path_excluded.
30 enum path_treatment {
31 path_none = 0,
32 path_recurse,
33 path_excluded,
34 path_untracked
38 * Support data structure for our opendir/readdir/closedir wrappers
40 struct cached_dir {
41 DIR *fdir;
42 struct untracked_cache_dir *untracked;
43 int nr_files;
44 int nr_dirs;
46 struct dirent *de;
47 const char *file;
48 struct untracked_cache_dir *ucd;
51 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
52 const char *path, int len, struct untracked_cache_dir *untracked,
53 int check_only, const struct path_simplify *simplify);
54 static int get_dtype(struct dirent *de, const char *path, int len);
56 /* helper string functions with support for the ignore_case flag */
57 int strcmp_icase(const char *a, const char *b)
59 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
62 int strncmp_icase(const char *a, const char *b, size_t count)
64 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
67 int fnmatch_icase(const char *pattern, const char *string, int flags)
69 return wildmatch(pattern, string,
70 flags | (ignore_case ? WM_CASEFOLD : 0),
71 NULL);
74 int git_fnmatch(const struct pathspec_item *item,
75 const char *pattern, const char *string,
76 int prefix)
78 if (prefix > 0) {
79 if (ps_strncmp(item, pattern, string, prefix))
80 return WM_NOMATCH;
81 pattern += prefix;
82 string += prefix;
84 if (item->flags & PATHSPEC_ONESTAR) {
85 int pattern_len = strlen(++pattern);
86 int string_len = strlen(string);
87 return string_len < pattern_len ||
88 ps_strcmp(item, pattern,
89 string + string_len - pattern_len);
91 if (item->magic & PATHSPEC_GLOB)
92 return wildmatch(pattern, string,
93 WM_PATHNAME |
94 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
95 NULL);
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,
100 NULL);
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, NULL);
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);
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, const struct pathspec *pathspec)
190 size_t len;
193 * Calculate common prefix for the pathspec, and
194 * use that to optimize the directory walk
196 len = common_prefix_len(pathspec);
198 /* Read the directory and prune it */
199 read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec);
200 return len;
203 int within_depth(const char *name, int namelen,
204 int depth, int max_depth)
206 const char *cp = name, *cpe = name + namelen;
208 while (cp < cpe) {
209 if (*cp++ != '/')
210 continue;
211 depth++;
212 if (depth > max_depth)
213 return 0;
215 return 1;
218 #define DO_MATCH_EXCLUDE 1
219 #define DO_MATCH_DIRECTORY 2
222 * Does 'match' match the given name?
223 * A match is found if
225 * (1) the 'match' string is leading directory of 'name', or
226 * (2) the 'match' string is a wildcard and matches 'name', or
227 * (3) the 'match' string is exactly the same as 'name'.
229 * and the return value tells which case it was.
231 * It returns 0 when there is no match.
233 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
234 const char *name, int namelen, unsigned flags)
236 /* name/namelen has prefix cut off by caller */
237 const char *match = item->match + prefix;
238 int matchlen = item->len - prefix;
241 * The normal call pattern is:
242 * 1. prefix = common_prefix_len(ps);
243 * 2. prune something, or fill_directory
244 * 3. match_pathspec()
246 * 'prefix' at #1 may be shorter than the command's prefix and
247 * it's ok for #2 to match extra files. Those extras will be
248 * trimmed at #3.
250 * Suppose the pathspec is 'foo' and '../bar' running from
251 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
252 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
253 * user does not want XYZ/foo, only the "foo" part should be
254 * case-insensitive. We need to filter out XYZ/foo here. In
255 * other words, we do not trust the caller on comparing the
256 * prefix part when :(icase) is involved. We do exact
257 * comparison ourselves.
259 * Normally the caller (common_prefix_len() in fact) does
260 * _exact_ matching on name[-prefix+1..-1] and we do not need
261 * to check that part. Be defensive and check it anyway, in
262 * case common_prefix_len is changed, or a new caller is
263 * introduced that does not use common_prefix_len.
265 * If the penalty turns out too high when prefix is really
266 * long, maybe change it to
267 * strncmp(match, name, item->prefix - prefix)
269 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
270 strncmp(item->match, name - prefix, item->prefix))
271 return 0;
273 /* If the match was just the prefix, we matched */
274 if (!*match)
275 return MATCHED_RECURSIVELY;
277 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
278 if (matchlen == namelen)
279 return MATCHED_EXACTLY;
281 if (match[matchlen-1] == '/' || name[matchlen] == '/')
282 return MATCHED_RECURSIVELY;
283 } else if ((flags & DO_MATCH_DIRECTORY) &&
284 match[matchlen - 1] == '/' &&
285 namelen == matchlen - 1 &&
286 !ps_strncmp(item, match, name, namelen))
287 return MATCHED_EXACTLY;
289 if (item->nowildcard_len < item->len &&
290 !git_fnmatch(item, match, name,
291 item->nowildcard_len - prefix))
292 return MATCHED_FNMATCH;
294 return 0;
298 * Given a name and a list of pathspecs, returns the nature of the
299 * closest (i.e. most specific) match of the name to any of the
300 * pathspecs.
302 * The caller typically calls this multiple times with the same
303 * pathspec and seen[] array but with different name/namelen
304 * (e.g. entries from the index) and is interested in seeing if and
305 * how each pathspec matches all the names it calls this function
306 * with. A mark is left in the seen[] array for each pathspec element
307 * indicating the closest type of match that element achieved, so if
308 * seen[n] remains zero after multiple invocations, that means the nth
309 * pathspec did not match any names, which could indicate that the
310 * user mistyped the nth pathspec.
312 static int do_match_pathspec(const struct pathspec *ps,
313 const char *name, int namelen,
314 int prefix, char *seen,
315 unsigned flags)
317 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
319 GUARD_PATHSPEC(ps,
320 PATHSPEC_FROMTOP |
321 PATHSPEC_MAXDEPTH |
322 PATHSPEC_LITERAL |
323 PATHSPEC_GLOB |
324 PATHSPEC_ICASE |
325 PATHSPEC_EXCLUDE);
327 if (!ps->nr) {
328 if (!ps->recursive ||
329 !(ps->magic & PATHSPEC_MAXDEPTH) ||
330 ps->max_depth == -1)
331 return MATCHED_RECURSIVELY;
333 if (within_depth(name, namelen, 0, ps->max_depth))
334 return MATCHED_EXACTLY;
335 else
336 return 0;
339 name += prefix;
340 namelen -= prefix;
342 for (i = ps->nr - 1; i >= 0; i--) {
343 int how;
345 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
346 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
347 continue;
349 if (seen && seen[i] == MATCHED_EXACTLY)
350 continue;
352 * Make exclude patterns optional and never report
353 * "pathspec ':(exclude)foo' matches no files"
355 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
356 seen[i] = MATCHED_FNMATCH;
357 how = match_pathspec_item(ps->items+i, prefix, name,
358 namelen, flags);
359 if (ps->recursive &&
360 (ps->magic & PATHSPEC_MAXDEPTH) &&
361 ps->max_depth != -1 &&
362 how && how != MATCHED_FNMATCH) {
363 int len = ps->items[i].len;
364 if (name[len] == '/')
365 len++;
366 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
367 how = MATCHED_EXACTLY;
368 else
369 how = 0;
371 if (how) {
372 if (retval < how)
373 retval = how;
374 if (seen && seen[i] < how)
375 seen[i] = how;
378 return retval;
381 int match_pathspec(const struct pathspec *ps,
382 const char *name, int namelen,
383 int prefix, char *seen, int is_dir)
385 int positive, negative;
386 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
387 positive = do_match_pathspec(ps, name, namelen,
388 prefix, seen, flags);
389 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
390 return positive;
391 negative = do_match_pathspec(ps, name, namelen,
392 prefix, seen,
393 flags | DO_MATCH_EXCLUDE);
394 return negative ? 0 : positive;
397 int report_path_error(const char *ps_matched,
398 const struct pathspec *pathspec,
399 const char *prefix)
402 * Make sure all pathspec matched; otherwise it is an error.
404 int num, errors = 0;
405 for (num = 0; num < pathspec->nr; num++) {
406 int other, found_dup;
408 if (ps_matched[num])
409 continue;
411 * The caller might have fed identical pathspec
412 * twice. Do not barf on such a mistake.
413 * FIXME: parse_pathspec should have eliminated
414 * duplicate pathspec.
416 for (found_dup = other = 0;
417 !found_dup && other < pathspec->nr;
418 other++) {
419 if (other == num || !ps_matched[other])
420 continue;
421 if (!strcmp(pathspec->items[other].original,
422 pathspec->items[num].original))
424 * Ok, we have a match already.
426 found_dup = 1;
428 if (found_dup)
429 continue;
431 error("pathspec '%s' did not match any file(s) known to git.",
432 pathspec->items[num].original);
433 errors++;
435 return errors;
439 * Return the length of the "simple" part of a path match limiter.
441 int simple_length(const char *match)
443 int len = -1;
445 for (;;) {
446 unsigned char c = *match++;
447 len++;
448 if (c == '\0' || is_glob_special(c))
449 return len;
453 int no_wildcard(const char *string)
455 return string[simple_length(string)] == '\0';
458 void parse_exclude_pattern(const char **pattern,
459 int *patternlen,
460 int *flags,
461 int *nowildcardlen)
463 const char *p = *pattern;
464 size_t i, len;
466 *flags = 0;
467 if (*p == '!') {
468 *flags |= EXC_FLAG_NEGATIVE;
469 p++;
471 len = strlen(p);
472 if (len && p[len - 1] == '/') {
473 len--;
474 *flags |= EXC_FLAG_MUSTBEDIR;
476 for (i = 0; i < len; i++) {
477 if (p[i] == '/')
478 break;
480 if (i == len)
481 *flags |= EXC_FLAG_NODIR;
482 *nowildcardlen = simple_length(p);
484 * we should have excluded the trailing slash from 'p' too,
485 * but that's one more allocation. Instead just make sure
486 * nowildcardlen does not exceed real patternlen
488 if (*nowildcardlen > len)
489 *nowildcardlen = len;
490 if (*p == '*' && no_wildcard(p + 1))
491 *flags |= EXC_FLAG_ENDSWITH;
492 *pattern = p;
493 *patternlen = len;
496 void add_exclude(const char *string, const char *base,
497 int baselen, struct exclude_list *el, int srcpos)
499 struct exclude *x;
500 int patternlen;
501 int flags;
502 int nowildcardlen;
504 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
505 if (flags & EXC_FLAG_MUSTBEDIR) {
506 char *s;
507 x = xmalloc(sizeof(*x) + patternlen + 1);
508 s = (char *)(x+1);
509 memcpy(s, string, patternlen);
510 s[patternlen] = '\0';
511 x->pattern = s;
512 } else {
513 x = xmalloc(sizeof(*x));
514 x->pattern = string;
516 x->patternlen = patternlen;
517 x->nowildcardlen = nowildcardlen;
518 x->base = base;
519 x->baselen = baselen;
520 x->flags = flags;
521 x->srcpos = srcpos;
522 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
523 el->excludes[el->nr++] = x;
524 x->el = el;
527 static void *read_skip_worktree_file_from_index(const char *path, size_t *size,
528 struct sha1_stat *sha1_stat)
530 int pos, len;
531 unsigned long sz;
532 enum object_type type;
533 void *data;
535 len = strlen(path);
536 pos = cache_name_pos(path, len);
537 if (pos < 0)
538 return NULL;
539 if (!ce_skip_worktree(active_cache[pos]))
540 return NULL;
541 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
542 if (!data || type != OBJ_BLOB) {
543 free(data);
544 return NULL;
546 *size = xsize_t(sz);
547 if (sha1_stat) {
548 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
549 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
551 return data;
555 * Frees memory within el which was allocated for exclude patterns and
556 * the file buffer. Does not free el itself.
558 void clear_exclude_list(struct exclude_list *el)
560 int i;
562 for (i = 0; i < el->nr; i++)
563 free(el->excludes[i]);
564 free(el->excludes);
565 free(el->filebuf);
567 el->nr = 0;
568 el->excludes = NULL;
569 el->filebuf = NULL;
572 static void trim_trailing_spaces(char *buf)
574 char *p, *last_space = NULL;
576 for (p = buf; *p; p++)
577 switch (*p) {
578 case ' ':
579 if (!last_space)
580 last_space = p;
581 break;
582 case '\\':
583 p++;
584 if (!*p)
585 return;
586 /* fallthrough */
587 default:
588 last_space = NULL;
591 if (last_space)
592 *last_space = '\0';
596 * Given a subdirectory name and "dir" of the current directory,
597 * search the subdir in "dir" and return it, or create a new one if it
598 * does not exist in "dir".
600 * If "name" has the trailing slash, it'll be excluded in the search.
602 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
603 struct untracked_cache_dir *dir,
604 const char *name, int len)
606 int first, last;
607 struct untracked_cache_dir *d;
608 if (!dir)
609 return NULL;
610 if (len && name[len - 1] == '/')
611 len--;
612 first = 0;
613 last = dir->dirs_nr;
614 while (last > first) {
615 int cmp, next = (last + first) >> 1;
616 d = dir->dirs[next];
617 cmp = strncmp(name, d->name, len);
618 if (!cmp && strlen(d->name) > len)
619 cmp = -1;
620 if (!cmp)
621 return d;
622 if (cmp < 0) {
623 last = next;
624 continue;
626 first = next+1;
629 uc->dir_created++;
630 d = xmalloc(sizeof(*d) + len + 1);
631 memset(d, 0, sizeof(*d));
632 memcpy(d->name, name, len);
633 d->name[len] = '\0';
635 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
636 memmove(dir->dirs + first + 1, dir->dirs + first,
637 (dir->dirs_nr - first) * sizeof(*dir->dirs));
638 dir->dirs_nr++;
639 dir->dirs[first] = d;
640 return d;
643 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
645 int i;
646 dir->valid = 0;
647 dir->untracked_nr = 0;
648 for (i = 0; i < dir->dirs_nr; i++)
649 do_invalidate_gitignore(dir->dirs[i]);
652 static void invalidate_gitignore(struct untracked_cache *uc,
653 struct untracked_cache_dir *dir)
655 uc->gitignore_invalidated++;
656 do_invalidate_gitignore(dir);
659 static void invalidate_directory(struct untracked_cache *uc,
660 struct untracked_cache_dir *dir)
662 int i;
663 uc->dir_invalidated++;
664 dir->valid = 0;
665 dir->untracked_nr = 0;
666 for (i = 0; i < dir->dirs_nr; i++)
667 dir->dirs[i]->recurse = 0;
671 * Given a file with name "fname", read it (either from disk, or from
672 * the index if "check_index" is non-zero), parse it and store the
673 * exclude rules in "el".
675 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
676 * stat data from disk (only valid if add_excludes returns zero). If
677 * ss_valid is non-zero, "ss" must contain good value as input.
679 static int add_excludes(const char *fname, const char *base, int baselen,
680 struct exclude_list *el, int check_index,
681 struct sha1_stat *sha1_stat)
683 struct stat st;
684 int fd, i, lineno = 1;
685 size_t size = 0;
686 char *buf, *entry;
688 fd = open(fname, O_RDONLY);
689 if (fd < 0 || fstat(fd, &st) < 0) {
690 if (errno != ENOENT)
691 warn_on_inaccessible(fname);
692 if (0 <= fd)
693 close(fd);
694 if (!check_index ||
695 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL)
696 return -1;
697 if (size == 0) {
698 free(buf);
699 return 0;
701 if (buf[size-1] != '\n') {
702 buf = xrealloc(buf, size+1);
703 buf[size++] = '\n';
705 } else {
706 size = xsize_t(st.st_size);
707 if (size == 0) {
708 if (sha1_stat) {
709 fill_stat_data(&sha1_stat->stat, &st);
710 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
711 sha1_stat->valid = 1;
713 close(fd);
714 return 0;
716 buf = xmalloc(size+1);
717 if (read_in_full(fd, buf, size) != size) {
718 free(buf);
719 close(fd);
720 return -1;
722 buf[size++] = '\n';
723 close(fd);
724 if (sha1_stat) {
725 int pos;
726 if (sha1_stat->valid &&
727 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st))
728 ; /* no content change, ss->sha1 still good */
729 else if (check_index &&
730 (pos = cache_name_pos(fname, strlen(fname))) >= 0 &&
731 !ce_stage(active_cache[pos]) &&
732 ce_uptodate(active_cache[pos]) &&
733 !would_convert_to_git(fname))
734 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
735 else
736 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
737 fill_stat_data(&sha1_stat->stat, &st);
738 sha1_stat->valid = 1;
742 el->filebuf = buf;
744 if (skip_utf8_bom(&buf, size))
745 size -= buf - el->filebuf;
747 entry = buf;
749 for (i = 0; i < size; i++) {
750 if (buf[i] == '\n') {
751 if (entry != buf + i && entry[0] != '#') {
752 buf[i - (i && buf[i-1] == '\r')] = 0;
753 trim_trailing_spaces(entry);
754 add_exclude(entry, base, baselen, el, lineno);
756 lineno++;
757 entry = buf + i + 1;
760 return 0;
763 int add_excludes_from_file_to_list(const char *fname, const char *base,
764 int baselen, struct exclude_list *el,
765 int check_index)
767 return add_excludes(fname, base, baselen, el, check_index, NULL);
770 struct exclude_list *add_exclude_list(struct dir_struct *dir,
771 int group_type, const char *src)
773 struct exclude_list *el;
774 struct exclude_list_group *group;
776 group = &dir->exclude_list_group[group_type];
777 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
778 el = &group->el[group->nr++];
779 memset(el, 0, sizeof(*el));
780 el->src = src;
781 return el;
785 * Used to set up core.excludesfile and .git/info/exclude lists.
787 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
788 struct sha1_stat *sha1_stat)
790 struct exclude_list *el;
792 * catch setup_standard_excludes() that's called before
793 * dir->untracked is assigned. That function behaves
794 * differently when dir->untracked is non-NULL.
796 if (!dir->untracked)
797 dir->unmanaged_exclude_files++;
798 el = add_exclude_list(dir, EXC_FILE, fname);
799 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0)
800 die("cannot use %s as an exclude file", fname);
803 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
805 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
806 add_excludes_from_file_1(dir, fname, NULL);
809 int match_basename(const char *basename, int basenamelen,
810 const char *pattern, int prefix, int patternlen,
811 int flags)
813 if (prefix == patternlen) {
814 if (patternlen == basenamelen &&
815 !strncmp_icase(pattern, basename, basenamelen))
816 return 1;
817 } else if (flags & EXC_FLAG_ENDSWITH) {
818 /* "*literal" matching against "fooliteral" */
819 if (patternlen - 1 <= basenamelen &&
820 !strncmp_icase(pattern + 1,
821 basename + basenamelen - (patternlen - 1),
822 patternlen - 1))
823 return 1;
824 } else {
825 if (fnmatch_icase_mem(pattern, patternlen,
826 basename, basenamelen,
827 0) == 0)
828 return 1;
830 return 0;
833 int match_pathname(const char *pathname, int pathlen,
834 const char *base, int baselen,
835 const char *pattern, int prefix, int patternlen,
836 int flags)
838 const char *name;
839 int namelen;
842 * match with FNM_PATHNAME; the pattern has base implicitly
843 * in front of it.
845 if (*pattern == '/') {
846 pattern++;
847 patternlen--;
848 prefix--;
852 * baselen does not count the trailing slash. base[] may or
853 * may not end with a trailing slash though.
855 if (pathlen < baselen + 1 ||
856 (baselen && pathname[baselen] != '/') ||
857 strncmp_icase(pathname, base, baselen))
858 return 0;
860 namelen = baselen ? pathlen - baselen - 1 : pathlen;
861 name = pathname + pathlen - namelen;
863 if (prefix) {
865 * if the non-wildcard part is longer than the
866 * remaining pathname, surely it cannot match.
868 if (prefix > namelen)
869 return 0;
871 if (strncmp_icase(pattern, name, prefix))
872 return 0;
873 pattern += prefix;
874 patternlen -= prefix;
875 name += prefix;
876 namelen -= prefix;
879 * If the whole pattern did not have a wildcard,
880 * then our prefix match is all we need; we
881 * do not need to call fnmatch at all.
883 if (!patternlen && !namelen)
884 return 1;
886 * This can happen when we ignore some exclude rules
887 * on directories in other to see if negative rules
888 * may match. E.g.
890 * /abc
891 * !/abc/def/ghi
893 * The pattern of interest is "/abc". On the first
894 * try, we should match path "abc" with this pattern
895 * in the "if" statement right above, but the caller
896 * ignores it.
898 * On the second try with paths within "abc",
899 * e.g. "abc/xyz", we come here and try to match it
900 * with "/abc".
902 if (!patternlen && namelen && *name == '/')
903 return 1;
906 return fnmatch_icase_mem(pattern, patternlen,
907 name, namelen,
908 WM_PATHNAME) == 0;
912 * Return non-zero if pathname is a directory and an ancestor of the
913 * literal path in a (negative) pattern. This is used to keep
914 * descending in "foo" and "foo/bar" when the pattern is
915 * "!foo/bar/.gitignore". "foo/notbar" will not be descended however.
917 static int match_neg_path(const char *pathname, int pathlen, int *dtype,
918 const char *base, int baselen,
919 const char *pattern, int prefix, int patternlen,
920 int flags)
922 assert((flags & EXC_FLAG_NEGATIVE) && !(flags & EXC_FLAG_NODIR));
924 if (*dtype == DT_UNKNOWN)
925 *dtype = get_dtype(NULL, pathname, pathlen);
926 if (*dtype != DT_DIR)
927 return 0;
929 if (*pattern == '/') {
930 pattern++;
931 patternlen--;
932 prefix--;
935 if (baselen) {
936 if (((pathlen < baselen && base[pathlen] == '/') ||
937 pathlen == baselen) &&
938 !strncmp_icase(pathname, base, pathlen))
939 return 1;
940 pathname += baselen + 1;
941 pathlen -= baselen + 1;
945 if (prefix &&
946 ((pathlen < prefix && pattern[pathlen] == '/') &&
947 !strncmp_icase(pathname, pattern, pathlen)))
948 return 1;
950 return 0;
954 * Scan the given exclude list in reverse to see whether pathname
955 * should be ignored. The first match (i.e. the last on the list), if
956 * any, determines the fate. Returns the exclude_list element which
957 * matched, or NULL for undecided.
959 static struct exclude *last_exclude_matching_from_list(const char *pathname,
960 int pathlen,
961 const char *basename,
962 int *dtype,
963 struct exclude_list *el)
965 struct exclude *exc = NULL; /* undecided */
966 int i, matched_negative_path = 0;
968 if (!el->nr)
969 return NULL; /* undefined */
971 for (i = el->nr - 1; 0 <= i; i--) {
972 struct exclude *x = el->excludes[i];
973 const char *exclude = x->pattern;
974 int prefix = x->nowildcardlen;
976 if (x->flags & EXC_FLAG_MUSTBEDIR) {
977 if (*dtype == DT_UNKNOWN)
978 *dtype = get_dtype(NULL, pathname, pathlen);
979 if (*dtype != DT_DIR)
980 continue;
983 if (x->flags & EXC_FLAG_NODIR) {
984 if (match_basename(basename,
985 pathlen - (basename - pathname),
986 exclude, prefix, x->patternlen,
987 x->flags)) {
988 exc = x;
989 break;
991 continue;
994 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
995 if (match_pathname(pathname, pathlen,
996 x->base, x->baselen ? x->baselen - 1 : 0,
997 exclude, prefix, x->patternlen, x->flags)) {
998 exc = x;
999 break;
1002 if ((x->flags & EXC_FLAG_NEGATIVE) && !matched_negative_path &&
1003 match_neg_path(pathname, pathlen, dtype, x->base,
1004 x->baselen ? x->baselen - 1 : 0,
1005 exclude, prefix, x->patternlen, x->flags))
1006 matched_negative_path = 1;
1008 if (exc &&
1009 !(exc->flags & EXC_FLAG_NEGATIVE) &&
1010 !(exc->flags & EXC_FLAG_NODIR) &&
1011 matched_negative_path)
1012 exc = NULL;
1013 return exc;
1017 * Scan the list and let the last match determine the fate.
1018 * Return 1 for exclude, 0 for include and -1 for undecided.
1020 int is_excluded_from_list(const char *pathname,
1021 int pathlen, const char *basename, int *dtype,
1022 struct exclude_list *el)
1024 struct exclude *exclude;
1025 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
1026 if (exclude)
1027 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1028 return -1; /* undecided */
1031 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1032 const char *pathname, int pathlen, const char *basename,
1033 int *dtype_p)
1035 int i, j;
1036 struct exclude_list_group *group;
1037 struct exclude *exclude;
1038 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1039 group = &dir->exclude_list_group[i];
1040 for (j = group->nr - 1; j >= 0; j--) {
1041 exclude = last_exclude_matching_from_list(
1042 pathname, pathlen, basename, dtype_p,
1043 &group->el[j]);
1044 if (exclude)
1045 return exclude;
1048 return NULL;
1052 * Loads the per-directory exclude list for the substring of base
1053 * which has a char length of baselen.
1055 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
1057 struct exclude_list_group *group;
1058 struct exclude_list *el;
1059 struct exclude_stack *stk = NULL;
1060 struct untracked_cache_dir *untracked;
1061 int current;
1063 group = &dir->exclude_list_group[EXC_DIRS];
1066 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1067 * which originate from directories not in the prefix of the
1068 * path being checked.
1070 while ((stk = dir->exclude_stack) != NULL) {
1071 if (stk->baselen <= baselen &&
1072 !strncmp(dir->basebuf.buf, base, stk->baselen))
1073 break;
1074 el = &group->el[dir->exclude_stack->exclude_ix];
1075 dir->exclude_stack = stk->prev;
1076 dir->exclude = NULL;
1077 free((char *)el->src); /* see strbuf_detach() below */
1078 clear_exclude_list(el);
1079 free(stk);
1080 group->nr--;
1083 /* Skip traversing into sub directories if the parent is excluded */
1084 if (dir->exclude)
1085 return;
1088 * Lazy initialization. All call sites currently just
1089 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1090 * them seems lots of work for little benefit.
1092 if (!dir->basebuf.buf)
1093 strbuf_init(&dir->basebuf, PATH_MAX);
1095 /* Read from the parent directories and push them down. */
1096 current = stk ? stk->baselen : -1;
1097 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1098 if (dir->untracked)
1099 untracked = stk ? stk->ucd : dir->untracked->root;
1100 else
1101 untracked = NULL;
1103 while (current < baselen) {
1104 const char *cp;
1105 struct sha1_stat sha1_stat;
1107 stk = xcalloc(1, sizeof(*stk));
1108 if (current < 0) {
1109 cp = base;
1110 current = 0;
1111 } else {
1112 cp = strchr(base + current + 1, '/');
1113 if (!cp)
1114 die("oops in prep_exclude");
1115 cp++;
1116 untracked =
1117 lookup_untracked(dir->untracked, untracked,
1118 base + current,
1119 cp - base - current);
1121 stk->prev = dir->exclude_stack;
1122 stk->baselen = cp - base;
1123 stk->exclude_ix = group->nr;
1124 stk->ucd = untracked;
1125 el = add_exclude_list(dir, EXC_DIRS, NULL);
1126 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1127 assert(stk->baselen == dir->basebuf.len);
1129 /* Abort if the directory is excluded */
1130 if (stk->baselen) {
1131 int dt = DT_DIR;
1132 dir->basebuf.buf[stk->baselen - 1] = 0;
1133 dir->exclude = last_exclude_matching_from_lists(dir,
1134 dir->basebuf.buf, stk->baselen - 1,
1135 dir->basebuf.buf + current, &dt);
1136 dir->basebuf.buf[stk->baselen - 1] = '/';
1137 if (dir->exclude &&
1138 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1139 dir->exclude = NULL;
1140 if (dir->exclude) {
1141 dir->exclude_stack = stk;
1142 return;
1146 /* Try to read per-directory file */
1147 hashclr(sha1_stat.sha1);
1148 sha1_stat.valid = 0;
1149 if (dir->exclude_per_dir &&
1151 * If we know that no files have been added in
1152 * this directory (i.e. valid_cached_dir() has
1153 * been executed and set untracked->valid) ..
1155 (!untracked || !untracked->valid ||
1157 * .. and .gitignore does not exist before
1158 * (i.e. null exclude_sha1). Then we can skip
1159 * loading .gitignore, which would result in
1160 * ENOENT anyway.
1162 !is_null_sha1(untracked->exclude_sha1))) {
1164 * dir->basebuf gets reused by the traversal, but we
1165 * need fname to remain unchanged to ensure the src
1166 * member of each struct exclude correctly
1167 * back-references its source file. Other invocations
1168 * of add_exclude_list provide stable strings, so we
1169 * strbuf_detach() and free() here in the caller.
1171 struct strbuf sb = STRBUF_INIT;
1172 strbuf_addbuf(&sb, &dir->basebuf);
1173 strbuf_addstr(&sb, dir->exclude_per_dir);
1174 el->src = strbuf_detach(&sb, NULL);
1175 add_excludes(el->src, el->src, stk->baselen, el, 1,
1176 untracked ? &sha1_stat : NULL);
1179 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1180 * will first be called in valid_cached_dir() then maybe many
1181 * times more in last_exclude_matching(). When the cache is
1182 * used, last_exclude_matching() will not be called and
1183 * reading .gitignore content will be a waste.
1185 * So when it's called by valid_cached_dir() and we can get
1186 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1187 * modified on work tree), we could delay reading the
1188 * .gitignore content until we absolutely need it in
1189 * last_exclude_matching(). Be careful about ignore rule
1190 * order, though, if you do that.
1192 if (untracked &&
1193 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1194 invalidate_gitignore(dir->untracked, untracked);
1195 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1197 dir->exclude_stack = stk;
1198 current = stk->baselen;
1200 strbuf_setlen(&dir->basebuf, baselen);
1204 * Loads the exclude lists for the directory containing pathname, then
1205 * scans all exclude lists to determine whether pathname is excluded.
1206 * Returns the exclude_list element which matched, or NULL for
1207 * undecided.
1209 struct exclude *last_exclude_matching(struct dir_struct *dir,
1210 const char *pathname,
1211 int *dtype_p)
1213 int pathlen = strlen(pathname);
1214 const char *basename = strrchr(pathname, '/');
1215 basename = (basename) ? basename+1 : pathname;
1217 prep_exclude(dir, pathname, basename-pathname);
1219 if (dir->exclude)
1220 return dir->exclude;
1222 return last_exclude_matching_from_lists(dir, pathname, pathlen,
1223 basename, dtype_p);
1227 * Loads the exclude lists for the directory containing pathname, then
1228 * scans all exclude lists to determine whether pathname is excluded.
1229 * Returns 1 if true, otherwise 0.
1231 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1233 struct exclude *exclude =
1234 last_exclude_matching(dir, pathname, dtype_p);
1235 if (exclude)
1236 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1237 return 0;
1240 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1242 struct dir_entry *ent;
1244 ent = xmalloc(sizeof(*ent) + len + 1);
1245 ent->len = len;
1246 memcpy(ent->name, pathname, len);
1247 ent->name[len] = 0;
1248 return ent;
1251 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1253 if (cache_file_exists(pathname, len, ignore_case))
1254 return NULL;
1256 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1257 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1260 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1262 if (!cache_name_is_other(pathname, len))
1263 return NULL;
1265 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1266 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1269 enum exist_status {
1270 index_nonexistent = 0,
1271 index_directory,
1272 index_gitdir
1276 * Do not use the alphabetically sorted index to look up
1277 * the directory name; instead, use the case insensitive
1278 * directory hash.
1280 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1282 const struct cache_entry *ce = cache_dir_exists(dirname, len);
1283 unsigned char endchar;
1285 if (!ce)
1286 return index_nonexistent;
1287 endchar = ce->name[len];
1290 * The cache_entry structure returned will contain this dirname
1291 * and possibly additional path components.
1293 if (endchar == '/')
1294 return index_directory;
1297 * If there are no additional path components, then this cache_entry
1298 * represents a submodule. Submodules, despite being directories,
1299 * are stored in the cache without a closing slash.
1301 if (!endchar && S_ISGITLINK(ce->ce_mode))
1302 return index_gitdir;
1304 /* This should never be hit, but it exists just in case. */
1305 return index_nonexistent;
1309 * The index sorts alphabetically by entry name, which
1310 * means that a gitlink sorts as '\0' at the end, while
1311 * a directory (which is defined not as an entry, but as
1312 * the files it contains) will sort with the '/' at the
1313 * end.
1315 static enum exist_status directory_exists_in_index(const char *dirname, int len)
1317 int pos;
1319 if (ignore_case)
1320 return directory_exists_in_index_icase(dirname, len);
1322 pos = cache_name_pos(dirname, len);
1323 if (pos < 0)
1324 pos = -pos-1;
1325 while (pos < active_nr) {
1326 const struct cache_entry *ce = active_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 untracked_cache_dir *untracked,
1377 const char *dirname, int len, int baselen, int exclude,
1378 const struct path_simplify *simplify)
1380 /* The "len-1" is to strip the final '/' */
1381 switch (directory_exists_in_index(dirname, len-1)) {
1382 case index_directory:
1383 return path_recurse;
1385 case index_gitdir:
1386 return path_none;
1388 case index_nonexistent:
1389 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1390 break;
1391 if (!(dir->flags & DIR_NO_GITLINKS)) {
1392 unsigned char sha1[20];
1393 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1394 return path_untracked;
1396 return path_recurse;
1399 /* This is the "show_other_directories" case */
1401 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1402 return exclude ? path_excluded : path_untracked;
1404 untracked = lookup_untracked(dir->untracked, untracked,
1405 dirname + baselen, len - baselen);
1406 return read_directory_recursive(dir, dirname, len,
1407 untracked, 1, simplify);
1411 * This is an inexact early pruning of any recursive directory
1412 * reading - if the path cannot possibly be in the pathspec,
1413 * return true, and we'll skip it early.
1415 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1417 if (simplify) {
1418 for (;;) {
1419 const char *match = simplify->path;
1420 int len = simplify->len;
1422 if (!match)
1423 break;
1424 if (len > pathlen)
1425 len = pathlen;
1426 if (!memcmp(path, match, len))
1427 return 0;
1428 simplify++;
1430 return 1;
1432 return 0;
1436 * This function tells us whether an excluded path matches a
1437 * list of "interesting" pathspecs. That is, whether a path matched
1438 * by any of the pathspecs could possibly be ignored by excluding
1439 * the specified path. This can happen if:
1441 * 1. the path is mentioned explicitly in the pathspec
1443 * 2. the path is a directory prefix of some element in the
1444 * pathspec
1446 static int exclude_matches_pathspec(const char *path, int len,
1447 const struct path_simplify *simplify)
1449 if (simplify) {
1450 for (; simplify->path; simplify++) {
1451 if (len == simplify->len
1452 && !memcmp(path, simplify->path, len))
1453 return 1;
1454 if (len < simplify->len
1455 && simplify->path[len] == '/'
1456 && !memcmp(path, simplify->path, len))
1457 return 1;
1460 return 0;
1463 static int get_index_dtype(const char *path, int len)
1465 int pos;
1466 const struct cache_entry *ce;
1468 ce = cache_file_exists(path, len, 0);
1469 if (ce) {
1470 if (!ce_uptodate(ce))
1471 return DT_UNKNOWN;
1472 if (S_ISGITLINK(ce->ce_mode))
1473 return DT_DIR;
1475 * Nobody actually cares about the
1476 * difference between DT_LNK and DT_REG
1478 return DT_REG;
1481 /* Try to look it up as a directory */
1482 pos = cache_name_pos(path, len);
1483 if (pos >= 0)
1484 return DT_UNKNOWN;
1485 pos = -pos-1;
1486 while (pos < active_nr) {
1487 ce = active_cache[pos++];
1488 if (strncmp(ce->name, path, len))
1489 break;
1490 if (ce->name[len] > '/')
1491 break;
1492 if (ce->name[len] < '/')
1493 continue;
1494 if (!ce_uptodate(ce))
1495 break; /* continue? */
1496 return DT_DIR;
1498 return DT_UNKNOWN;
1501 static int get_dtype(struct dirent *de, const char *path, int len)
1503 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1504 struct stat st;
1506 if (dtype != DT_UNKNOWN)
1507 return dtype;
1508 dtype = get_index_dtype(path, len);
1509 if (dtype != DT_UNKNOWN)
1510 return dtype;
1511 if (lstat(path, &st))
1512 return dtype;
1513 if (S_ISREG(st.st_mode))
1514 return DT_REG;
1515 if (S_ISDIR(st.st_mode))
1516 return DT_DIR;
1517 if (S_ISLNK(st.st_mode))
1518 return DT_LNK;
1519 return dtype;
1522 static enum path_treatment treat_one_path(struct dir_struct *dir,
1523 struct untracked_cache_dir *untracked,
1524 struct strbuf *path,
1525 int baselen,
1526 const struct path_simplify *simplify,
1527 int dtype, struct dirent *de)
1529 int exclude;
1530 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1532 if (dtype == DT_UNKNOWN)
1533 dtype = get_dtype(de, path->buf, path->len);
1535 /* Always exclude indexed files */
1536 if (dtype != DT_DIR && has_path_in_index)
1537 return path_none;
1540 * When we are looking at a directory P in the working tree,
1541 * there are three cases:
1543 * (1) P exists in the index. Everything inside the directory P in
1544 * the working tree needs to go when P is checked out from the
1545 * index.
1547 * (2) P does not exist in the index, but there is P/Q in the index.
1548 * We know P will stay a directory when we check out the contents
1549 * of the index, but we do not know yet if there is a directory
1550 * P/Q in the working tree to be killed, so we need to recurse.
1552 * (3) P does not exist in the index, and there is no P/Q in the index
1553 * to require P to be a directory, either. Only in this case, we
1554 * know that everything inside P will not be killed without
1555 * recursing.
1557 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1558 (dtype == DT_DIR) &&
1559 !has_path_in_index &&
1560 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1561 return path_none;
1563 exclude = is_excluded(dir, path->buf, &dtype);
1566 * Excluded? If we don't explicitly want to show
1567 * ignored files, ignore it
1569 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1570 return path_excluded;
1572 switch (dtype) {
1573 default:
1574 return path_none;
1575 case DT_DIR:
1576 strbuf_addch(path, '/');
1577 return treat_directory(dir, untracked, path->buf, path->len,
1578 baselen, exclude, simplify);
1579 case DT_REG:
1580 case DT_LNK:
1581 return exclude ? path_excluded : path_untracked;
1585 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1586 struct untracked_cache_dir *untracked,
1587 struct cached_dir *cdir,
1588 struct strbuf *path,
1589 int baselen,
1590 const struct path_simplify *simplify)
1592 strbuf_setlen(path, baselen);
1593 if (!cdir->ucd) {
1594 strbuf_addstr(path, cdir->file);
1595 return path_untracked;
1597 strbuf_addstr(path, cdir->ucd->name);
1598 /* treat_one_path() does this before it calls treat_directory() */
1599 if (path->buf[path->len - 1] != '/')
1600 strbuf_addch(path, '/');
1601 if (cdir->ucd->check_only)
1603 * check_only is set as a result of treat_directory() getting
1604 * to its bottom. Verify again the same set of directories
1605 * with check_only set.
1607 return read_directory_recursive(dir, path->buf, path->len,
1608 cdir->ucd, 1, simplify);
1610 * We get path_recurse in the first run when
1611 * directory_exists_in_index() returns index_nonexistent. We
1612 * are sure that new changes in the index does not impact the
1613 * outcome. Return now.
1615 return path_recurse;
1618 static enum path_treatment treat_path(struct dir_struct *dir,
1619 struct untracked_cache_dir *untracked,
1620 struct cached_dir *cdir,
1621 struct strbuf *path,
1622 int baselen,
1623 const struct path_simplify *simplify)
1625 int dtype;
1626 struct dirent *de = cdir->de;
1628 if (!de)
1629 return treat_path_fast(dir, untracked, cdir, path,
1630 baselen, simplify);
1631 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1632 return path_none;
1633 strbuf_setlen(path, baselen);
1634 strbuf_addstr(path, de->d_name);
1635 if (simplify_away(path->buf, path->len, simplify))
1636 return path_none;
1638 dtype = DTYPE(de);
1639 return treat_one_path(dir, untracked, path, baselen, simplify, dtype, de);
1642 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1644 if (!dir)
1645 return;
1646 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1647 dir->untracked_alloc);
1648 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1651 static int valid_cached_dir(struct dir_struct *dir,
1652 struct untracked_cache_dir *untracked,
1653 struct strbuf *path,
1654 int check_only)
1656 struct stat st;
1658 if (!untracked)
1659 return 0;
1661 if (stat(path->len ? path->buf : ".", &st)) {
1662 invalidate_directory(dir->untracked, untracked);
1663 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1664 return 0;
1666 if (!untracked->valid ||
1667 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
1668 if (untracked->valid)
1669 invalidate_directory(dir->untracked, untracked);
1670 fill_stat_data(&untracked->stat_data, &st);
1671 return 0;
1674 if (untracked->check_only != !!check_only) {
1675 invalidate_directory(dir->untracked, untracked);
1676 return 0;
1680 * prep_exclude will be called eventually on this directory,
1681 * but it's called much later in last_exclude_matching(). We
1682 * need it now to determine the validity of the cache for this
1683 * path. The next calls will be nearly no-op, the way
1684 * prep_exclude() is designed.
1686 if (path->len && path->buf[path->len - 1] != '/') {
1687 strbuf_addch(path, '/');
1688 prep_exclude(dir, path->buf, path->len);
1689 strbuf_setlen(path, path->len - 1);
1690 } else
1691 prep_exclude(dir, path->buf, path->len);
1693 /* hopefully prep_exclude() haven't invalidated this entry... */
1694 return untracked->valid;
1697 static int open_cached_dir(struct cached_dir *cdir,
1698 struct dir_struct *dir,
1699 struct untracked_cache_dir *untracked,
1700 struct strbuf *path,
1701 int check_only)
1703 memset(cdir, 0, sizeof(*cdir));
1704 cdir->untracked = untracked;
1705 if (valid_cached_dir(dir, untracked, path, check_only))
1706 return 0;
1707 cdir->fdir = opendir(path->len ? path->buf : ".");
1708 if (dir->untracked)
1709 dir->untracked->dir_opened++;
1710 if (!cdir->fdir)
1711 return -1;
1712 return 0;
1715 static int read_cached_dir(struct cached_dir *cdir)
1717 if (cdir->fdir) {
1718 cdir->de = readdir(cdir->fdir);
1719 if (!cdir->de)
1720 return -1;
1721 return 0;
1723 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1724 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1725 if (!d->recurse) {
1726 cdir->nr_dirs++;
1727 continue;
1729 cdir->ucd = d;
1730 cdir->nr_dirs++;
1731 return 0;
1733 cdir->ucd = NULL;
1734 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1735 struct untracked_cache_dir *d = cdir->untracked;
1736 cdir->file = d->untracked[cdir->nr_files++];
1737 return 0;
1739 return -1;
1742 static void close_cached_dir(struct cached_dir *cdir)
1744 if (cdir->fdir)
1745 closedir(cdir->fdir);
1747 * We have gone through this directory and found no untracked
1748 * entries. Mark it valid.
1750 if (cdir->untracked) {
1751 cdir->untracked->valid = 1;
1752 cdir->untracked->recurse = 1;
1757 * Read a directory tree. We currently ignore anything but
1758 * directories, regular files and symlinks. That's because git
1759 * doesn't handle them at all yet. Maybe that will change some
1760 * day.
1762 * Also, we ignore the name ".git" (even if it is not a directory).
1763 * That likely will not change.
1765 * Returns the most significant path_treatment value encountered in the scan.
1767 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1768 const char *base, int baselen,
1769 struct untracked_cache_dir *untracked, int check_only,
1770 const struct path_simplify *simplify)
1772 struct cached_dir cdir;
1773 enum path_treatment state, subdir_state, dir_state = path_none;
1774 struct strbuf path = STRBUF_INIT;
1776 strbuf_add(&path, base, baselen);
1778 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1779 goto out;
1781 if (untracked)
1782 untracked->check_only = !!check_only;
1784 while (!read_cached_dir(&cdir)) {
1785 /* check how the file or directory should be treated */
1786 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);
1788 if (state > dir_state)
1789 dir_state = state;
1791 /* recurse into subdir if instructed by treat_path */
1792 if (state == path_recurse) {
1793 struct untracked_cache_dir *ud;
1794 ud = lookup_untracked(dir->untracked, untracked,
1795 path.buf + baselen,
1796 path.len - baselen);
1797 subdir_state =
1798 read_directory_recursive(dir, path.buf, path.len,
1799 ud, check_only, simplify);
1800 if (subdir_state > dir_state)
1801 dir_state = subdir_state;
1804 if (check_only) {
1805 /* abort early if maximum state has been reached */
1806 if (dir_state == path_untracked) {
1807 if (cdir.fdir)
1808 add_untracked(untracked, path.buf + baselen);
1809 break;
1811 /* skip the dir_add_* part */
1812 continue;
1815 /* add the path to the appropriate result list */
1816 switch (state) {
1817 case path_excluded:
1818 if (dir->flags & DIR_SHOW_IGNORED)
1819 dir_add_name(dir, path.buf, path.len);
1820 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1821 ((dir->flags & DIR_COLLECT_IGNORED) &&
1822 exclude_matches_pathspec(path.buf, path.len,
1823 simplify)))
1824 dir_add_ignored(dir, path.buf, path.len);
1825 break;
1827 case path_untracked:
1828 if (dir->flags & DIR_SHOW_IGNORED)
1829 break;
1830 dir_add_name(dir, path.buf, path.len);
1831 if (cdir.fdir)
1832 add_untracked(untracked, path.buf + baselen);
1833 break;
1835 default:
1836 break;
1839 close_cached_dir(&cdir);
1840 out:
1841 strbuf_release(&path);
1843 return dir_state;
1846 static int cmp_name(const void *p1, const void *p2)
1848 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1849 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1851 return name_compare(e1->name, e1->len, e2->name, e2->len);
1854 static struct path_simplify *create_simplify(const char **pathspec)
1856 int nr, alloc = 0;
1857 struct path_simplify *simplify = NULL;
1859 if (!pathspec)
1860 return NULL;
1862 for (nr = 0 ; ; nr++) {
1863 const char *match;
1864 ALLOC_GROW(simplify, nr + 1, alloc);
1865 match = *pathspec++;
1866 if (!match)
1867 break;
1868 simplify[nr].path = match;
1869 simplify[nr].len = simple_length(match);
1871 simplify[nr].path = NULL;
1872 simplify[nr].len = 0;
1873 return simplify;
1876 static void free_simplify(struct path_simplify *simplify)
1878 free(simplify);
1881 static int treat_leading_path(struct dir_struct *dir,
1882 const char *path, int len,
1883 const struct path_simplify *simplify)
1885 struct strbuf sb = STRBUF_INIT;
1886 int baselen, rc = 0;
1887 const char *cp;
1888 int old_flags = dir->flags;
1890 while (len && path[len - 1] == '/')
1891 len--;
1892 if (!len)
1893 return 1;
1894 baselen = 0;
1895 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1896 while (1) {
1897 cp = path + baselen + !!baselen;
1898 cp = memchr(cp, '/', path + len - cp);
1899 if (!cp)
1900 baselen = len;
1901 else
1902 baselen = cp - path;
1903 strbuf_setlen(&sb, 0);
1904 strbuf_add(&sb, path, baselen);
1905 if (!is_directory(sb.buf))
1906 break;
1907 if (simplify_away(sb.buf, sb.len, simplify))
1908 break;
1909 if (treat_one_path(dir, NULL, &sb, baselen, simplify,
1910 DT_DIR, NULL) == path_none)
1911 break; /* do not recurse into it */
1912 if (len <= baselen) {
1913 rc = 1;
1914 break; /* finished checking */
1917 strbuf_release(&sb);
1918 dir->flags = old_flags;
1919 return rc;
1922 static const char *get_ident_string(void)
1924 static struct strbuf sb = STRBUF_INIT;
1925 struct utsname uts;
1927 if (sb.len)
1928 return sb.buf;
1929 if (uname(&uts) < 0)
1930 die_errno(_("failed to get kernel name and information"));
1931 strbuf_addf(&sb, "Location %s, system %s %s %s", get_git_work_tree(),
1932 uts.sysname, uts.release, uts.version);
1933 return sb.buf;
1936 static int ident_in_untracked(const struct untracked_cache *uc)
1938 const char *end = uc->ident.buf + uc->ident.len;
1939 const char *p = uc->ident.buf;
1941 for (p = uc->ident.buf; p < end; p += strlen(p) + 1)
1942 if (!strcmp(p, get_ident_string()))
1943 return 1;
1944 return 0;
1947 void add_untracked_ident(struct untracked_cache *uc)
1949 if (ident_in_untracked(uc))
1950 return;
1951 strbuf_addstr(&uc->ident, get_ident_string());
1952 /* this strbuf contains a list of strings, save NUL too */
1953 strbuf_addch(&uc->ident, 0);
1956 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1957 int base_len,
1958 const struct pathspec *pathspec)
1960 struct untracked_cache_dir *root;
1962 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
1963 return NULL;
1966 * We only support $GIT_DIR/info/exclude and core.excludesfile
1967 * as the global ignore rule files. Any other additions
1968 * (e.g. from command line) invalidate the cache. This
1969 * condition also catches running setup_standard_excludes()
1970 * before setting dir->untracked!
1972 if (dir->unmanaged_exclude_files)
1973 return NULL;
1976 * Optimize for the main use case only: whole-tree git
1977 * status. More work involved in treat_leading_path() if we
1978 * use cache on just a subset of the worktree. pathspec
1979 * support could make the matter even worse.
1981 if (base_len || (pathspec && pathspec->nr))
1982 return NULL;
1984 /* Different set of flags may produce different results */
1985 if (dir->flags != dir->untracked->dir_flags ||
1987 * See treat_directory(), case index_nonexistent. Without
1988 * this flag, we may need to also cache .git file content
1989 * for the resolve_gitlink_ref() call, which we don't.
1991 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1992 /* We don't support collecting ignore files */
1993 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1994 DIR_COLLECT_IGNORED)))
1995 return NULL;
1998 * If we use .gitignore in the cache and now you change it to
1999 * .gitexclude, everything will go wrong.
2001 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2002 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2003 return NULL;
2006 * EXC_CMDL is not considered in the cache. If people set it,
2007 * skip the cache.
2009 if (dir->exclude_list_group[EXC_CMDL].nr)
2010 return NULL;
2012 if (!ident_in_untracked(dir->untracked)) {
2013 warning(_("Untracked cache is disabled on this system."));
2014 return NULL;
2017 if (!dir->untracked->root) {
2018 const int len = sizeof(*dir->untracked->root);
2019 dir->untracked->root = xmalloc(len);
2020 memset(dir->untracked->root, 0, len);
2023 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2024 root = dir->untracked->root;
2025 if (hashcmp(dir->ss_info_exclude.sha1,
2026 dir->untracked->ss_info_exclude.sha1)) {
2027 invalidate_gitignore(dir->untracked, root);
2028 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2030 if (hashcmp(dir->ss_excludes_file.sha1,
2031 dir->untracked->ss_excludes_file.sha1)) {
2032 invalidate_gitignore(dir->untracked, root);
2033 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2036 /* Make sure this directory is not dropped out at saving phase */
2037 root->recurse = 1;
2038 return root;
2041 int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
2043 struct path_simplify *simplify;
2044 struct untracked_cache_dir *untracked;
2047 * Check out create_simplify()
2049 if (pathspec)
2050 GUARD_PATHSPEC(pathspec,
2051 PATHSPEC_FROMTOP |
2052 PATHSPEC_MAXDEPTH |
2053 PATHSPEC_LITERAL |
2054 PATHSPEC_GLOB |
2055 PATHSPEC_ICASE |
2056 PATHSPEC_EXCLUDE);
2058 if (has_symlink_leading_path(path, len))
2059 return dir->nr;
2062 * exclude patterns are treated like positive ones in
2063 * create_simplify. Usually exclude patterns should be a
2064 * subset of positive ones, which has no impacts on
2065 * create_simplify().
2067 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
2068 untracked = validate_untracked_cache(dir, len, pathspec);
2069 if (!untracked)
2071 * make sure untracked cache code path is disabled,
2072 * e.g. prep_exclude()
2074 dir->untracked = NULL;
2075 if (!len || treat_leading_path(dir, path, len, simplify))
2076 read_directory_recursive(dir, path, len, untracked, 0, simplify);
2077 free_simplify(simplify);
2078 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
2079 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
2080 if (dir->untracked) {
2081 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2082 trace_printf_key(&trace_untracked_stats,
2083 "node creation: %u\n"
2084 "gitignore invalidation: %u\n"
2085 "directory invalidation: %u\n"
2086 "opendir: %u\n",
2087 dir->untracked->dir_created,
2088 dir->untracked->gitignore_invalidated,
2089 dir->untracked->dir_invalidated,
2090 dir->untracked->dir_opened);
2091 if (dir->untracked == the_index.untracked &&
2092 (dir->untracked->dir_opened ||
2093 dir->untracked->gitignore_invalidated ||
2094 dir->untracked->dir_invalidated))
2095 the_index.cache_changed |= UNTRACKED_CHANGED;
2096 if (dir->untracked != the_index.untracked) {
2097 free(dir->untracked);
2098 dir->untracked = NULL;
2101 return dir->nr;
2104 int file_exists(const char *f)
2106 struct stat sb;
2107 return lstat(f, &sb) == 0;
2111 * Given two normalized paths (a trailing slash is ok), if subdir is
2112 * outside dir, return -1. Otherwise return the offset in subdir that
2113 * can be used as relative path to dir.
2115 int dir_inside_of(const char *subdir, const char *dir)
2117 int offset = 0;
2119 assert(dir && subdir && *dir && *subdir);
2121 while (*dir && *subdir && *dir == *subdir) {
2122 dir++;
2123 subdir++;
2124 offset++;
2127 /* hel[p]/me vs hel[l]/yeah */
2128 if (*dir && *subdir)
2129 return -1;
2131 if (!*subdir)
2132 return !*dir ? offset : -1; /* same dir */
2134 /* foo/[b]ar vs foo/[] */
2135 if (is_dir_sep(dir[-1]))
2136 return is_dir_sep(subdir[-1]) ? offset : -1;
2138 /* foo[/]bar vs foo[] */
2139 return is_dir_sep(*subdir) ? offset + 1 : -1;
2142 int is_inside_dir(const char *dir)
2144 char *cwd;
2145 int rc;
2147 if (!dir)
2148 return 0;
2150 cwd = xgetcwd();
2151 rc = (dir_inside_of(cwd, dir) >= 0);
2152 free(cwd);
2153 return rc;
2156 int is_empty_dir(const char *path)
2158 DIR *dir = opendir(path);
2159 struct dirent *e;
2160 int ret = 1;
2162 if (!dir)
2163 return 0;
2165 while ((e = readdir(dir)) != NULL)
2166 if (!is_dot_or_dotdot(e->d_name)) {
2167 ret = 0;
2168 break;
2171 closedir(dir);
2172 return ret;
2175 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2177 DIR *dir;
2178 struct dirent *e;
2179 int ret = 0, original_len = path->len, len, kept_down = 0;
2180 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2181 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2182 unsigned char submodule_head[20];
2184 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2185 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2186 /* Do not descend and nuke a nested git work tree. */
2187 if (kept_up)
2188 *kept_up = 1;
2189 return 0;
2192 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2193 dir = opendir(path->buf);
2194 if (!dir) {
2195 if (errno == ENOENT)
2196 return keep_toplevel ? -1 : 0;
2197 else if (errno == EACCES && !keep_toplevel)
2199 * An empty dir could be removable even if it
2200 * is unreadable:
2202 return rmdir(path->buf);
2203 else
2204 return -1;
2206 if (path->buf[original_len - 1] != '/')
2207 strbuf_addch(path, '/');
2209 len = path->len;
2210 while ((e = readdir(dir)) != NULL) {
2211 struct stat st;
2212 if (is_dot_or_dotdot(e->d_name))
2213 continue;
2215 strbuf_setlen(path, len);
2216 strbuf_addstr(path, e->d_name);
2217 if (lstat(path->buf, &st)) {
2218 if (errno == ENOENT)
2220 * file disappeared, which is what we
2221 * wanted anyway
2223 continue;
2224 /* fall thru */
2225 } else if (S_ISDIR(st.st_mode)) {
2226 if (!remove_dir_recurse(path, flag, &kept_down))
2227 continue; /* happy */
2228 } else if (!only_empty &&
2229 (!unlink(path->buf) || errno == ENOENT)) {
2230 continue; /* happy, too */
2233 /* path too long, stat fails, or non-directory still exists */
2234 ret = -1;
2235 break;
2237 closedir(dir);
2239 strbuf_setlen(path, original_len);
2240 if (!ret && !keep_toplevel && !kept_down)
2241 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2242 else if (kept_up)
2244 * report the uplevel that it is not an error that we
2245 * did not rmdir() our directory.
2247 *kept_up = !ret;
2248 return ret;
2251 int remove_dir_recursively(struct strbuf *path, int flag)
2253 return remove_dir_recurse(path, flag, NULL);
2256 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2258 void setup_standard_excludes(struct dir_struct *dir)
2260 const char *path;
2262 dir->exclude_per_dir = ".gitignore";
2264 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2265 if (!excludes_file)
2266 excludes_file = xdg_config_home("ignore");
2267 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2268 add_excludes_from_file_1(dir, excludes_file,
2269 dir->untracked ? &dir->ss_excludes_file : NULL);
2271 /* per repository user preference */
2272 path = git_path_info_exclude();
2273 if (!access_or_warn(path, R_OK, 0))
2274 add_excludes_from_file_1(dir, path,
2275 dir->untracked ? &dir->ss_info_exclude : NULL);
2278 int remove_path(const char *name)
2280 char *slash;
2282 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
2283 return -1;
2285 slash = strrchr(name, '/');
2286 if (slash) {
2287 char *dirs = xstrdup(name);
2288 slash = dirs + (slash - name);
2289 do {
2290 *slash = '\0';
2291 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2292 free(dirs);
2294 return 0;
2298 * Frees memory within dir which was allocated for exclude lists and
2299 * the exclude_stack. Does not free dir itself.
2301 void clear_directory(struct dir_struct *dir)
2303 int i, j;
2304 struct exclude_list_group *group;
2305 struct exclude_list *el;
2306 struct exclude_stack *stk;
2308 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2309 group = &dir->exclude_list_group[i];
2310 for (j = 0; j < group->nr; j++) {
2311 el = &group->el[j];
2312 if (i == EXC_DIRS)
2313 free((char *)el->src);
2314 clear_exclude_list(el);
2316 free(group->el);
2319 stk = dir->exclude_stack;
2320 while (stk) {
2321 struct exclude_stack *prev = stk->prev;
2322 free(stk);
2323 stk = prev;
2325 strbuf_release(&dir->basebuf);
2328 struct ondisk_untracked_cache {
2329 struct stat_data info_exclude_stat;
2330 struct stat_data excludes_file_stat;
2331 uint32_t dir_flags;
2332 unsigned char info_exclude_sha1[20];
2333 unsigned char excludes_file_sha1[20];
2334 char exclude_per_dir[FLEX_ARRAY];
2337 #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2339 struct write_data {
2340 int index; /* number of written untracked_cache_dir */
2341 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2342 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2343 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2344 struct strbuf out;
2345 struct strbuf sb_stat;
2346 struct strbuf sb_sha1;
2349 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2351 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2352 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2353 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2354 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2355 to->sd_dev = htonl(from->sd_dev);
2356 to->sd_ino = htonl(from->sd_ino);
2357 to->sd_uid = htonl(from->sd_uid);
2358 to->sd_gid = htonl(from->sd_gid);
2359 to->sd_size = htonl(from->sd_size);
2362 static void write_one_dir(struct untracked_cache_dir *untracked,
2363 struct write_data *wd)
2365 struct stat_data stat_data;
2366 struct strbuf *out = &wd->out;
2367 unsigned char intbuf[16];
2368 unsigned int intlen, value;
2369 int i = wd->index++;
2372 * untracked_nr should be reset whenever valid is clear, but
2373 * for safety..
2375 if (!untracked->valid) {
2376 untracked->untracked_nr = 0;
2377 untracked->check_only = 0;
2380 if (untracked->check_only)
2381 ewah_set(wd->check_only, i);
2382 if (untracked->valid) {
2383 ewah_set(wd->valid, i);
2384 stat_data_to_disk(&stat_data, &untracked->stat_data);
2385 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2387 if (!is_null_sha1(untracked->exclude_sha1)) {
2388 ewah_set(wd->sha1_valid, i);
2389 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2392 intlen = encode_varint(untracked->untracked_nr, intbuf);
2393 strbuf_add(out, intbuf, intlen);
2395 /* skip non-recurse directories */
2396 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2397 if (untracked->dirs[i]->recurse)
2398 value++;
2399 intlen = encode_varint(value, intbuf);
2400 strbuf_add(out, intbuf, intlen);
2402 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2404 for (i = 0; i < untracked->untracked_nr; i++)
2405 strbuf_add(out, untracked->untracked[i],
2406 strlen(untracked->untracked[i]) + 1);
2408 for (i = 0; i < untracked->dirs_nr; i++)
2409 if (untracked->dirs[i]->recurse)
2410 write_one_dir(untracked->dirs[i], wd);
2413 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2415 struct ondisk_untracked_cache *ouc;
2416 struct write_data wd;
2417 unsigned char varbuf[16];
2418 int len = 0, varint_len;
2419 if (untracked->exclude_per_dir)
2420 len = strlen(untracked->exclude_per_dir);
2421 ouc = xmalloc(sizeof(*ouc) + len + 1);
2422 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2423 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2424 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2425 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2426 ouc->dir_flags = htonl(untracked->dir_flags);
2427 memcpy(ouc->exclude_per_dir, untracked->exclude_per_dir, len + 1);
2429 varint_len = encode_varint(untracked->ident.len, varbuf);
2430 strbuf_add(out, varbuf, varint_len);
2431 strbuf_add(out, untracked->ident.buf, untracked->ident.len);
2433 strbuf_add(out, ouc, ouc_size(len));
2434 free(ouc);
2435 ouc = NULL;
2437 if (!untracked->root) {
2438 varint_len = encode_varint(0, varbuf);
2439 strbuf_add(out, varbuf, varint_len);
2440 return;
2443 wd.index = 0;
2444 wd.check_only = ewah_new();
2445 wd.valid = ewah_new();
2446 wd.sha1_valid = ewah_new();
2447 strbuf_init(&wd.out, 1024);
2448 strbuf_init(&wd.sb_stat, 1024);
2449 strbuf_init(&wd.sb_sha1, 1024);
2450 write_one_dir(untracked->root, &wd);
2452 varint_len = encode_varint(wd.index, varbuf);
2453 strbuf_add(out, varbuf, varint_len);
2454 strbuf_addbuf(out, &wd.out);
2455 ewah_serialize_strbuf(wd.valid, out);
2456 ewah_serialize_strbuf(wd.check_only, out);
2457 ewah_serialize_strbuf(wd.sha1_valid, out);
2458 strbuf_addbuf(out, &wd.sb_stat);
2459 strbuf_addbuf(out, &wd.sb_sha1);
2460 strbuf_addch(out, '\0'); /* safe guard for string lists */
2462 ewah_free(wd.valid);
2463 ewah_free(wd.check_only);
2464 ewah_free(wd.sha1_valid);
2465 strbuf_release(&wd.out);
2466 strbuf_release(&wd.sb_stat);
2467 strbuf_release(&wd.sb_sha1);
2470 static void free_untracked(struct untracked_cache_dir *ucd)
2472 int i;
2473 if (!ucd)
2474 return;
2475 for (i = 0; i < ucd->dirs_nr; i++)
2476 free_untracked(ucd->dirs[i]);
2477 for (i = 0; i < ucd->untracked_nr; i++)
2478 free(ucd->untracked[i]);
2479 free(ucd->untracked);
2480 free(ucd->dirs);
2481 free(ucd);
2484 void free_untracked_cache(struct untracked_cache *uc)
2486 if (uc)
2487 free_untracked(uc->root);
2488 free(uc);
2491 struct read_data {
2492 int index;
2493 struct untracked_cache_dir **ucd;
2494 struct ewah_bitmap *check_only;
2495 struct ewah_bitmap *valid;
2496 struct ewah_bitmap *sha1_valid;
2497 const unsigned char *data;
2498 const unsigned char *end;
2501 static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2503 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2504 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2505 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2506 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2507 to->sd_dev = get_be32(&from->sd_dev);
2508 to->sd_ino = get_be32(&from->sd_ino);
2509 to->sd_uid = get_be32(&from->sd_uid);
2510 to->sd_gid = get_be32(&from->sd_gid);
2511 to->sd_size = get_be32(&from->sd_size);
2514 static int read_one_dir(struct untracked_cache_dir **untracked_,
2515 struct read_data *rd)
2517 struct untracked_cache_dir ud, *untracked;
2518 const unsigned char *next, *data = rd->data, *end = rd->end;
2519 unsigned int value;
2520 int i, len;
2522 memset(&ud, 0, sizeof(ud));
2524 next = data;
2525 value = decode_varint(&next);
2526 if (next > end)
2527 return -1;
2528 ud.recurse = 1;
2529 ud.untracked_alloc = value;
2530 ud.untracked_nr = value;
2531 if (ud.untracked_nr)
2532 ud.untracked = xmalloc(sizeof(*ud.untracked) * ud.untracked_nr);
2533 data = next;
2535 next = data;
2536 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2537 if (next > end)
2538 return -1;
2539 ud.dirs = xmalloc(sizeof(*ud.dirs) * ud.dirs_nr);
2540 data = next;
2542 len = strlen((const char *)data);
2543 next = data + len + 1;
2544 if (next > rd->end)
2545 return -1;
2546 *untracked_ = untracked = xmalloc(sizeof(*untracked) + len);
2547 memcpy(untracked, &ud, sizeof(ud));
2548 memcpy(untracked->name, data, len + 1);
2549 data = next;
2551 for (i = 0; i < untracked->untracked_nr; i++) {
2552 len = strlen((const char *)data);
2553 next = data + len + 1;
2554 if (next > rd->end)
2555 return -1;
2556 untracked->untracked[i] = xstrdup((const char*)data);
2557 data = next;
2560 rd->ucd[rd->index++] = untracked;
2561 rd->data = data;
2563 for (i = 0; i < untracked->dirs_nr; i++) {
2564 len = read_one_dir(untracked->dirs + i, rd);
2565 if (len < 0)
2566 return -1;
2568 return 0;
2571 static void set_check_only(size_t pos, void *cb)
2573 struct read_data *rd = cb;
2574 struct untracked_cache_dir *ud = rd->ucd[pos];
2575 ud->check_only = 1;
2578 static void read_stat(size_t pos, void *cb)
2580 struct read_data *rd = cb;
2581 struct untracked_cache_dir *ud = rd->ucd[pos];
2582 if (rd->data + sizeof(struct stat_data) > rd->end) {
2583 rd->data = rd->end + 1;
2584 return;
2586 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2587 rd->data += sizeof(struct stat_data);
2588 ud->valid = 1;
2591 static void read_sha1(size_t pos, void *cb)
2593 struct read_data *rd = cb;
2594 struct untracked_cache_dir *ud = rd->ucd[pos];
2595 if (rd->data + 20 > rd->end) {
2596 rd->data = rd->end + 1;
2597 return;
2599 hashcpy(ud->exclude_sha1, rd->data);
2600 rd->data += 20;
2603 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2604 const struct stat_data *stat,
2605 const unsigned char *sha1)
2607 stat_data_from_disk(&sha1_stat->stat, stat);
2608 hashcpy(sha1_stat->sha1, sha1);
2609 sha1_stat->valid = 1;
2612 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2614 const struct ondisk_untracked_cache *ouc;
2615 struct untracked_cache *uc;
2616 struct read_data rd;
2617 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2618 const char *ident;
2619 int ident_len, len;
2621 if (sz <= 1 || end[-1] != '\0')
2622 return NULL;
2623 end--;
2625 ident_len = decode_varint(&next);
2626 if (next + ident_len > end)
2627 return NULL;
2628 ident = (const char *)next;
2629 next += ident_len;
2631 ouc = (const struct ondisk_untracked_cache *)next;
2632 if (next + ouc_size(0) > end)
2633 return NULL;
2635 uc = xcalloc(1, sizeof(*uc));
2636 strbuf_init(&uc->ident, ident_len);
2637 strbuf_add(&uc->ident, ident, ident_len);
2638 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2639 ouc->info_exclude_sha1);
2640 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2641 ouc->excludes_file_sha1);
2642 uc->dir_flags = get_be32(&ouc->dir_flags);
2643 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2644 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2645 next += ouc_size(strlen(ouc->exclude_per_dir));
2646 if (next >= end)
2647 goto done2;
2649 len = decode_varint(&next);
2650 if (next > end || len == 0)
2651 goto done2;
2653 rd.valid = ewah_new();
2654 rd.check_only = ewah_new();
2655 rd.sha1_valid = ewah_new();
2656 rd.data = next;
2657 rd.end = end;
2658 rd.index = 0;
2659 rd.ucd = xmalloc(sizeof(*rd.ucd) * len);
2661 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2662 goto done;
2664 next = rd.data;
2665 len = ewah_read_mmap(rd.valid, next, end - next);
2666 if (len < 0)
2667 goto done;
2669 next += len;
2670 len = ewah_read_mmap(rd.check_only, next, end - next);
2671 if (len < 0)
2672 goto done;
2674 next += len;
2675 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2676 if (len < 0)
2677 goto done;
2679 ewah_each_bit(rd.check_only, set_check_only, &rd);
2680 rd.data = next + len;
2681 ewah_each_bit(rd.valid, read_stat, &rd);
2682 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2683 next = rd.data;
2685 done:
2686 free(rd.ucd);
2687 ewah_free(rd.valid);
2688 ewah_free(rd.check_only);
2689 ewah_free(rd.sha1_valid);
2690 done2:
2691 if (next != end) {
2692 free_untracked_cache(uc);
2693 uc = NULL;
2695 return uc;
2698 static void invalidate_one_directory(struct untracked_cache *uc,
2699 struct untracked_cache_dir *ucd)
2701 uc->dir_invalidated++;
2702 ucd->valid = 0;
2703 ucd->untracked_nr = 0;
2707 * Normally when an entry is added or removed from a directory,
2708 * invalidating that directory is enough. No need to touch its
2709 * ancestors. When a directory is shown as "foo/bar/" in git-status
2710 * however, deleting or adding an entry may have cascading effect.
2712 * Say the "foo/bar/file" has become untracked, we need to tell the
2713 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2714 * directory any more (because "bar" is managed by foo as an untracked
2715 * "file").
2717 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2718 * was the last untracked entry in the entire "foo", we should show
2719 * "foo/" instead. Which means we have to invalidate past "bar" up to
2720 * "foo".
2722 * This function traverses all directories from root to leaf. If there
2723 * is a chance of one of the above cases happening, we invalidate back
2724 * to root. Otherwise we just invalidate the leaf. There may be a more
2725 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2726 * detect these cases and avoid unnecessary invalidation, for example,
2727 * checking for the untracked entry named "bar/" in "foo", but for now
2728 * stick to something safe and simple.
2730 static int invalidate_one_component(struct untracked_cache *uc,
2731 struct untracked_cache_dir *dir,
2732 const char *path, int len)
2734 const char *rest = strchr(path, '/');
2736 if (rest) {
2737 int component_len = rest - path;
2738 struct untracked_cache_dir *d =
2739 lookup_untracked(uc, dir, path, component_len);
2740 int ret =
2741 invalidate_one_component(uc, d, rest + 1,
2742 len - (component_len + 1));
2743 if (ret)
2744 invalidate_one_directory(uc, dir);
2745 return ret;
2748 invalidate_one_directory(uc, dir);
2749 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2752 void untracked_cache_invalidate_path(struct index_state *istate,
2753 const char *path)
2755 if (!istate->untracked || !istate->untracked->root)
2756 return;
2757 invalidate_one_component(istate->untracked, istate->untracked->root,
2758 path, strlen(path));
2761 void untracked_cache_remove_from_index(struct index_state *istate,
2762 const char *path)
2764 untracked_cache_invalidate_path(istate, path);
2767 void untracked_cache_add_to_index(struct index_state *istate,
2768 const char *path)
2770 untracked_cache_invalidate_path(istate, path);