Merge branch 'cb/hook-sigpipe' into maint
[git/git-svn.git] / dir.c
blob88e89585eb1634e07a9316ccd6caf8ea3643e817
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;
887 return fnmatch_icase_mem(pattern, patternlen,
888 name, namelen,
889 WM_PATHNAME) == 0;
893 * Scan the given exclude list in reverse to see whether pathname
894 * should be ignored. The first match (i.e. the last on the list), if
895 * any, determines the fate. Returns the exclude_list element which
896 * matched, or NULL for undecided.
898 static struct exclude *last_exclude_matching_from_list(const char *pathname,
899 int pathlen,
900 const char *basename,
901 int *dtype,
902 struct exclude_list *el)
904 int i;
906 if (!el->nr)
907 return NULL; /* undefined */
909 for (i = el->nr - 1; 0 <= i; i--) {
910 struct exclude *x = el->excludes[i];
911 const char *exclude = x->pattern;
912 int prefix = x->nowildcardlen;
914 if (x->flags & EXC_FLAG_MUSTBEDIR) {
915 if (*dtype == DT_UNKNOWN)
916 *dtype = get_dtype(NULL, pathname, pathlen);
917 if (*dtype != DT_DIR)
918 continue;
921 if (x->flags & EXC_FLAG_NODIR) {
922 if (match_basename(basename,
923 pathlen - (basename - pathname),
924 exclude, prefix, x->patternlen,
925 x->flags))
926 return x;
927 continue;
930 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
931 if (match_pathname(pathname, pathlen,
932 x->base, x->baselen ? x->baselen - 1 : 0,
933 exclude, prefix, x->patternlen, x->flags))
934 return x;
936 return NULL; /* undecided */
940 * Scan the list and let the last match determine the fate.
941 * Return 1 for exclude, 0 for include and -1 for undecided.
943 int is_excluded_from_list(const char *pathname,
944 int pathlen, const char *basename, int *dtype,
945 struct exclude_list *el)
947 struct exclude *exclude;
948 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
949 if (exclude)
950 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
951 return -1; /* undecided */
954 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
955 const char *pathname, int pathlen, const char *basename,
956 int *dtype_p)
958 int i, j;
959 struct exclude_list_group *group;
960 struct exclude *exclude;
961 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
962 group = &dir->exclude_list_group[i];
963 for (j = group->nr - 1; j >= 0; j--) {
964 exclude = last_exclude_matching_from_list(
965 pathname, pathlen, basename, dtype_p,
966 &group->el[j]);
967 if (exclude)
968 return exclude;
971 return NULL;
975 * Loads the per-directory exclude list for the substring of base
976 * which has a char length of baselen.
978 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
980 struct exclude_list_group *group;
981 struct exclude_list *el;
982 struct exclude_stack *stk = NULL;
983 struct untracked_cache_dir *untracked;
984 int current;
986 group = &dir->exclude_list_group[EXC_DIRS];
989 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
990 * which originate from directories not in the prefix of the
991 * path being checked.
993 while ((stk = dir->exclude_stack) != NULL) {
994 if (stk->baselen <= baselen &&
995 !strncmp(dir->basebuf.buf, base, stk->baselen))
996 break;
997 el = &group->el[dir->exclude_stack->exclude_ix];
998 dir->exclude_stack = stk->prev;
999 dir->exclude = NULL;
1000 free((char *)el->src); /* see strbuf_detach() below */
1001 clear_exclude_list(el);
1002 free(stk);
1003 group->nr--;
1006 /* Skip traversing into sub directories if the parent is excluded */
1007 if (dir->exclude)
1008 return;
1011 * Lazy initialization. All call sites currently just
1012 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1013 * them seems lots of work for little benefit.
1015 if (!dir->basebuf.buf)
1016 strbuf_init(&dir->basebuf, PATH_MAX);
1018 /* Read from the parent directories and push them down. */
1019 current = stk ? stk->baselen : -1;
1020 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1021 if (dir->untracked)
1022 untracked = stk ? stk->ucd : dir->untracked->root;
1023 else
1024 untracked = NULL;
1026 while (current < baselen) {
1027 const char *cp;
1028 struct sha1_stat sha1_stat;
1030 stk = xcalloc(1, sizeof(*stk));
1031 if (current < 0) {
1032 cp = base;
1033 current = 0;
1034 } else {
1035 cp = strchr(base + current + 1, '/');
1036 if (!cp)
1037 die("oops in prep_exclude");
1038 cp++;
1039 untracked =
1040 lookup_untracked(dir->untracked, untracked,
1041 base + current,
1042 cp - base - current);
1044 stk->prev = dir->exclude_stack;
1045 stk->baselen = cp - base;
1046 stk->exclude_ix = group->nr;
1047 stk->ucd = untracked;
1048 el = add_exclude_list(dir, EXC_DIRS, NULL);
1049 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1050 assert(stk->baselen == dir->basebuf.len);
1052 /* Abort if the directory is excluded */
1053 if (stk->baselen) {
1054 int dt = DT_DIR;
1055 dir->basebuf.buf[stk->baselen - 1] = 0;
1056 dir->exclude = last_exclude_matching_from_lists(dir,
1057 dir->basebuf.buf, stk->baselen - 1,
1058 dir->basebuf.buf + current, &dt);
1059 dir->basebuf.buf[stk->baselen - 1] = '/';
1060 if (dir->exclude &&
1061 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1062 dir->exclude = NULL;
1063 if (dir->exclude) {
1064 dir->exclude_stack = stk;
1065 return;
1069 /* Try to read per-directory file */
1070 hashclr(sha1_stat.sha1);
1071 sha1_stat.valid = 0;
1072 if (dir->exclude_per_dir &&
1074 * If we know that no files have been added in
1075 * this directory (i.e. valid_cached_dir() has
1076 * been executed and set untracked->valid) ..
1078 (!untracked || !untracked->valid ||
1080 * .. and .gitignore does not exist before
1081 * (i.e. null exclude_sha1). Then we can skip
1082 * loading .gitignore, which would result in
1083 * ENOENT anyway.
1085 !is_null_sha1(untracked->exclude_sha1))) {
1087 * dir->basebuf gets reused by the traversal, but we
1088 * need fname to remain unchanged to ensure the src
1089 * member of each struct exclude correctly
1090 * back-references its source file. Other invocations
1091 * of add_exclude_list provide stable strings, so we
1092 * strbuf_detach() and free() here in the caller.
1094 struct strbuf sb = STRBUF_INIT;
1095 strbuf_addbuf(&sb, &dir->basebuf);
1096 strbuf_addstr(&sb, dir->exclude_per_dir);
1097 el->src = strbuf_detach(&sb, NULL);
1098 add_excludes(el->src, el->src, stk->baselen, el, 1,
1099 untracked ? &sha1_stat : NULL);
1102 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1103 * will first be called in valid_cached_dir() then maybe many
1104 * times more in last_exclude_matching(). When the cache is
1105 * used, last_exclude_matching() will not be called and
1106 * reading .gitignore content will be a waste.
1108 * So when it's called by valid_cached_dir() and we can get
1109 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1110 * modified on work tree), we could delay reading the
1111 * .gitignore content until we absolutely need it in
1112 * last_exclude_matching(). Be careful about ignore rule
1113 * order, though, if you do that.
1115 if (untracked &&
1116 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1117 invalidate_gitignore(dir->untracked, untracked);
1118 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1120 dir->exclude_stack = stk;
1121 current = stk->baselen;
1123 strbuf_setlen(&dir->basebuf, baselen);
1127 * Loads the exclude lists for the directory containing pathname, then
1128 * scans all exclude lists to determine whether pathname is excluded.
1129 * Returns the exclude_list element which matched, or NULL for
1130 * undecided.
1132 struct exclude *last_exclude_matching(struct dir_struct *dir,
1133 const char *pathname,
1134 int *dtype_p)
1136 int pathlen = strlen(pathname);
1137 const char *basename = strrchr(pathname, '/');
1138 basename = (basename) ? basename+1 : pathname;
1140 prep_exclude(dir, pathname, basename-pathname);
1142 if (dir->exclude)
1143 return dir->exclude;
1145 return last_exclude_matching_from_lists(dir, pathname, pathlen,
1146 basename, dtype_p);
1150 * Loads the exclude lists for the directory containing pathname, then
1151 * scans all exclude lists to determine whether pathname is excluded.
1152 * Returns 1 if true, otherwise 0.
1154 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1156 struct exclude *exclude =
1157 last_exclude_matching(dir, pathname, dtype_p);
1158 if (exclude)
1159 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1160 return 0;
1163 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1165 struct dir_entry *ent;
1167 ent = xmalloc(sizeof(*ent) + len + 1);
1168 ent->len = len;
1169 memcpy(ent->name, pathname, len);
1170 ent->name[len] = 0;
1171 return ent;
1174 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1176 if (cache_file_exists(pathname, len, ignore_case))
1177 return NULL;
1179 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1180 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1183 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1185 if (!cache_name_is_other(pathname, len))
1186 return NULL;
1188 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1189 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1192 enum exist_status {
1193 index_nonexistent = 0,
1194 index_directory,
1195 index_gitdir
1199 * Do not use the alphabetically sorted index to look up
1200 * the directory name; instead, use the case insensitive
1201 * directory hash.
1203 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1205 struct cache_entry *ce;
1207 if (cache_dir_exists(dirname, len))
1208 return index_directory;
1210 ce = cache_file_exists(dirname, len, ignore_case);
1211 if (ce && S_ISGITLINK(ce->ce_mode))
1212 return index_gitdir;
1214 return index_nonexistent;
1218 * The index sorts alphabetically by entry name, which
1219 * means that a gitlink sorts as '\0' at the end, while
1220 * a directory (which is defined not as an entry, but as
1221 * the files it contains) will sort with the '/' at the
1222 * end.
1224 static enum exist_status directory_exists_in_index(const char *dirname, int len)
1226 int pos;
1228 if (ignore_case)
1229 return directory_exists_in_index_icase(dirname, len);
1231 pos = cache_name_pos(dirname, len);
1232 if (pos < 0)
1233 pos = -pos-1;
1234 while (pos < active_nr) {
1235 const struct cache_entry *ce = active_cache[pos++];
1236 unsigned char endchar;
1238 if (strncmp(ce->name, dirname, len))
1239 break;
1240 endchar = ce->name[len];
1241 if (endchar > '/')
1242 break;
1243 if (endchar == '/')
1244 return index_directory;
1245 if (!endchar && S_ISGITLINK(ce->ce_mode))
1246 return index_gitdir;
1248 return index_nonexistent;
1252 * When we find a directory when traversing the filesystem, we
1253 * have three distinct cases:
1255 * - ignore it
1256 * - see it as a directory
1257 * - recurse into it
1259 * and which one we choose depends on a combination of existing
1260 * git index contents and the flags passed into the directory
1261 * traversal routine.
1263 * Case 1: If we *already* have entries in the index under that
1264 * directory name, we always recurse into the directory to see
1265 * all the files.
1267 * Case 2: If we *already* have that directory name as a gitlink,
1268 * we always continue to see it as a gitlink, regardless of whether
1269 * there is an actual git directory there or not (it might not
1270 * be checked out as a subproject!)
1272 * Case 3: if we didn't have it in the index previously, we
1273 * have a few sub-cases:
1275 * (a) if "show_other_directories" is true, we show it as
1276 * just a directory, unless "hide_empty_directories" is
1277 * also true, in which case we need to check if it contains any
1278 * untracked and / or ignored files.
1279 * (b) if it looks like a git directory, and we don't have
1280 * 'no_gitlinks' set we treat it as a gitlink, and show it
1281 * as a directory.
1282 * (c) otherwise, we recurse into it.
1284 static enum path_treatment treat_directory(struct dir_struct *dir,
1285 struct untracked_cache_dir *untracked,
1286 const char *dirname, int len, int baselen, int exclude,
1287 const struct path_simplify *simplify)
1289 /* The "len-1" is to strip the final '/' */
1290 switch (directory_exists_in_index(dirname, len-1)) {
1291 case index_directory:
1292 return path_recurse;
1294 case index_gitdir:
1295 return path_none;
1297 case index_nonexistent:
1298 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1299 break;
1300 if (!(dir->flags & DIR_NO_GITLINKS)) {
1301 unsigned char sha1[20];
1302 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1303 return path_untracked;
1305 return path_recurse;
1308 /* This is the "show_other_directories" case */
1310 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1311 return exclude ? path_excluded : path_untracked;
1313 untracked = lookup_untracked(dir->untracked, untracked,
1314 dirname + baselen, len - baselen);
1315 return read_directory_recursive(dir, dirname, len,
1316 untracked, 1, simplify);
1320 * This is an inexact early pruning of any recursive directory
1321 * reading - if the path cannot possibly be in the pathspec,
1322 * return true, and we'll skip it early.
1324 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1326 if (simplify) {
1327 for (;;) {
1328 const char *match = simplify->path;
1329 int len = simplify->len;
1331 if (!match)
1332 break;
1333 if (len > pathlen)
1334 len = pathlen;
1335 if (!memcmp(path, match, len))
1336 return 0;
1337 simplify++;
1339 return 1;
1341 return 0;
1345 * This function tells us whether an excluded path matches a
1346 * list of "interesting" pathspecs. That is, whether a path matched
1347 * by any of the pathspecs could possibly be ignored by excluding
1348 * the specified path. This can happen if:
1350 * 1. the path is mentioned explicitly in the pathspec
1352 * 2. the path is a directory prefix of some element in the
1353 * pathspec
1355 static int exclude_matches_pathspec(const char *path, int len,
1356 const struct path_simplify *simplify)
1358 if (simplify) {
1359 for (; simplify->path; simplify++) {
1360 if (len == simplify->len
1361 && !memcmp(path, simplify->path, len))
1362 return 1;
1363 if (len < simplify->len
1364 && simplify->path[len] == '/'
1365 && !memcmp(path, simplify->path, len))
1366 return 1;
1369 return 0;
1372 static int get_index_dtype(const char *path, int len)
1374 int pos;
1375 const struct cache_entry *ce;
1377 ce = cache_file_exists(path, len, 0);
1378 if (ce) {
1379 if (!ce_uptodate(ce))
1380 return DT_UNKNOWN;
1381 if (S_ISGITLINK(ce->ce_mode))
1382 return DT_DIR;
1384 * Nobody actually cares about the
1385 * difference between DT_LNK and DT_REG
1387 return DT_REG;
1390 /* Try to look it up as a directory */
1391 pos = cache_name_pos(path, len);
1392 if (pos >= 0)
1393 return DT_UNKNOWN;
1394 pos = -pos-1;
1395 while (pos < active_nr) {
1396 ce = active_cache[pos++];
1397 if (strncmp(ce->name, path, len))
1398 break;
1399 if (ce->name[len] > '/')
1400 break;
1401 if (ce->name[len] < '/')
1402 continue;
1403 if (!ce_uptodate(ce))
1404 break; /* continue? */
1405 return DT_DIR;
1407 return DT_UNKNOWN;
1410 static int get_dtype(struct dirent *de, const char *path, int len)
1412 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1413 struct stat st;
1415 if (dtype != DT_UNKNOWN)
1416 return dtype;
1417 dtype = get_index_dtype(path, len);
1418 if (dtype != DT_UNKNOWN)
1419 return dtype;
1420 if (lstat(path, &st))
1421 return dtype;
1422 if (S_ISREG(st.st_mode))
1423 return DT_REG;
1424 if (S_ISDIR(st.st_mode))
1425 return DT_DIR;
1426 if (S_ISLNK(st.st_mode))
1427 return DT_LNK;
1428 return dtype;
1431 static enum path_treatment treat_one_path(struct dir_struct *dir,
1432 struct untracked_cache_dir *untracked,
1433 struct strbuf *path,
1434 int baselen,
1435 const struct path_simplify *simplify,
1436 int dtype, struct dirent *de)
1438 int exclude;
1439 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1441 if (dtype == DT_UNKNOWN)
1442 dtype = get_dtype(de, path->buf, path->len);
1444 /* Always exclude indexed files */
1445 if (dtype != DT_DIR && has_path_in_index)
1446 return path_none;
1449 * When we are looking at a directory P in the working tree,
1450 * there are three cases:
1452 * (1) P exists in the index. Everything inside the directory P in
1453 * the working tree needs to go when P is checked out from the
1454 * index.
1456 * (2) P does not exist in the index, but there is P/Q in the index.
1457 * We know P will stay a directory when we check out the contents
1458 * of the index, but we do not know yet if there is a directory
1459 * P/Q in the working tree to be killed, so we need to recurse.
1461 * (3) P does not exist in the index, and there is no P/Q in the index
1462 * to require P to be a directory, either. Only in this case, we
1463 * know that everything inside P will not be killed without
1464 * recursing.
1466 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1467 (dtype == DT_DIR) &&
1468 !has_path_in_index &&
1469 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1470 return path_none;
1472 exclude = is_excluded(dir, path->buf, &dtype);
1475 * Excluded? If we don't explicitly want to show
1476 * ignored files, ignore it
1478 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1479 return path_excluded;
1481 switch (dtype) {
1482 default:
1483 return path_none;
1484 case DT_DIR:
1485 strbuf_addch(path, '/');
1486 return treat_directory(dir, untracked, path->buf, path->len,
1487 baselen, exclude, simplify);
1488 case DT_REG:
1489 case DT_LNK:
1490 return exclude ? path_excluded : path_untracked;
1494 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1495 struct untracked_cache_dir *untracked,
1496 struct cached_dir *cdir,
1497 struct strbuf *path,
1498 int baselen,
1499 const struct path_simplify *simplify)
1501 strbuf_setlen(path, baselen);
1502 if (!cdir->ucd) {
1503 strbuf_addstr(path, cdir->file);
1504 return path_untracked;
1506 strbuf_addstr(path, cdir->ucd->name);
1507 /* treat_one_path() does this before it calls treat_directory() */
1508 if (path->buf[path->len - 1] != '/')
1509 strbuf_addch(path, '/');
1510 if (cdir->ucd->check_only)
1512 * check_only is set as a result of treat_directory() getting
1513 * to its bottom. Verify again the same set of directories
1514 * with check_only set.
1516 return read_directory_recursive(dir, path->buf, path->len,
1517 cdir->ucd, 1, simplify);
1519 * We get path_recurse in the first run when
1520 * directory_exists_in_index() returns index_nonexistent. We
1521 * are sure that new changes in the index does not impact the
1522 * outcome. Return now.
1524 return path_recurse;
1527 static enum path_treatment treat_path(struct dir_struct *dir,
1528 struct untracked_cache_dir *untracked,
1529 struct cached_dir *cdir,
1530 struct strbuf *path,
1531 int baselen,
1532 const struct path_simplify *simplify)
1534 int dtype;
1535 struct dirent *de = cdir->de;
1537 if (!de)
1538 return treat_path_fast(dir, untracked, cdir, path,
1539 baselen, simplify);
1540 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1541 return path_none;
1542 strbuf_setlen(path, baselen);
1543 strbuf_addstr(path, de->d_name);
1544 if (simplify_away(path->buf, path->len, simplify))
1545 return path_none;
1547 dtype = DTYPE(de);
1548 return treat_one_path(dir, untracked, path, baselen, simplify, dtype, de);
1551 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1553 if (!dir)
1554 return;
1555 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1556 dir->untracked_alloc);
1557 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1560 static int valid_cached_dir(struct dir_struct *dir,
1561 struct untracked_cache_dir *untracked,
1562 struct strbuf *path,
1563 int check_only)
1565 struct stat st;
1567 if (!untracked)
1568 return 0;
1570 if (stat(path->len ? path->buf : ".", &st)) {
1571 invalidate_directory(dir->untracked, untracked);
1572 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1573 return 0;
1575 if (!untracked->valid ||
1576 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
1577 if (untracked->valid)
1578 invalidate_directory(dir->untracked, untracked);
1579 fill_stat_data(&untracked->stat_data, &st);
1580 return 0;
1583 if (untracked->check_only != !!check_only) {
1584 invalidate_directory(dir->untracked, untracked);
1585 return 0;
1589 * prep_exclude will be called eventually on this directory,
1590 * but it's called much later in last_exclude_matching(). We
1591 * need it now to determine the validity of the cache for this
1592 * path. The next calls will be nearly no-op, the way
1593 * prep_exclude() is designed.
1595 if (path->len && path->buf[path->len - 1] != '/') {
1596 strbuf_addch(path, '/');
1597 prep_exclude(dir, path->buf, path->len);
1598 strbuf_setlen(path, path->len - 1);
1599 } else
1600 prep_exclude(dir, path->buf, path->len);
1602 /* hopefully prep_exclude() haven't invalidated this entry... */
1603 return untracked->valid;
1606 static int open_cached_dir(struct cached_dir *cdir,
1607 struct dir_struct *dir,
1608 struct untracked_cache_dir *untracked,
1609 struct strbuf *path,
1610 int check_only)
1612 memset(cdir, 0, sizeof(*cdir));
1613 cdir->untracked = untracked;
1614 if (valid_cached_dir(dir, untracked, path, check_only))
1615 return 0;
1616 cdir->fdir = opendir(path->len ? path->buf : ".");
1617 if (dir->untracked)
1618 dir->untracked->dir_opened++;
1619 if (!cdir->fdir)
1620 return -1;
1621 return 0;
1624 static int read_cached_dir(struct cached_dir *cdir)
1626 if (cdir->fdir) {
1627 cdir->de = readdir(cdir->fdir);
1628 if (!cdir->de)
1629 return -1;
1630 return 0;
1632 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1633 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1634 if (!d->recurse) {
1635 cdir->nr_dirs++;
1636 continue;
1638 cdir->ucd = d;
1639 cdir->nr_dirs++;
1640 return 0;
1642 cdir->ucd = NULL;
1643 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1644 struct untracked_cache_dir *d = cdir->untracked;
1645 cdir->file = d->untracked[cdir->nr_files++];
1646 return 0;
1648 return -1;
1651 static void close_cached_dir(struct cached_dir *cdir)
1653 if (cdir->fdir)
1654 closedir(cdir->fdir);
1656 * We have gone through this directory and found no untracked
1657 * entries. Mark it valid.
1659 if (cdir->untracked) {
1660 cdir->untracked->valid = 1;
1661 cdir->untracked->recurse = 1;
1666 * Read a directory tree. We currently ignore anything but
1667 * directories, regular files and symlinks. That's because git
1668 * doesn't handle them at all yet. Maybe that will change some
1669 * day.
1671 * Also, we ignore the name ".git" (even if it is not a directory).
1672 * That likely will not change.
1674 * Returns the most significant path_treatment value encountered in the scan.
1676 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1677 const char *base, int baselen,
1678 struct untracked_cache_dir *untracked, int check_only,
1679 const struct path_simplify *simplify)
1681 struct cached_dir cdir;
1682 enum path_treatment state, subdir_state, dir_state = path_none;
1683 struct strbuf path = STRBUF_INIT;
1685 strbuf_add(&path, base, baselen);
1687 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1688 goto out;
1690 if (untracked)
1691 untracked->check_only = !!check_only;
1693 while (!read_cached_dir(&cdir)) {
1694 /* check how the file or directory should be treated */
1695 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);
1697 if (state > dir_state)
1698 dir_state = state;
1700 /* recurse into subdir if instructed by treat_path */
1701 if (state == path_recurse) {
1702 struct untracked_cache_dir *ud;
1703 ud = lookup_untracked(dir->untracked, untracked,
1704 path.buf + baselen,
1705 path.len - baselen);
1706 subdir_state =
1707 read_directory_recursive(dir, path.buf, path.len,
1708 ud, check_only, simplify);
1709 if (subdir_state > dir_state)
1710 dir_state = subdir_state;
1713 if (check_only) {
1714 /* abort early if maximum state has been reached */
1715 if (dir_state == path_untracked) {
1716 if (cdir.fdir)
1717 add_untracked(untracked, path.buf + baselen);
1718 break;
1720 /* skip the dir_add_* part */
1721 continue;
1724 /* add the path to the appropriate result list */
1725 switch (state) {
1726 case path_excluded:
1727 if (dir->flags & DIR_SHOW_IGNORED)
1728 dir_add_name(dir, path.buf, path.len);
1729 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1730 ((dir->flags & DIR_COLLECT_IGNORED) &&
1731 exclude_matches_pathspec(path.buf, path.len,
1732 simplify)))
1733 dir_add_ignored(dir, path.buf, path.len);
1734 break;
1736 case path_untracked:
1737 if (dir->flags & DIR_SHOW_IGNORED)
1738 break;
1739 dir_add_name(dir, path.buf, path.len);
1740 if (cdir.fdir)
1741 add_untracked(untracked, path.buf + baselen);
1742 break;
1744 default:
1745 break;
1748 close_cached_dir(&cdir);
1749 out:
1750 strbuf_release(&path);
1752 return dir_state;
1755 static int cmp_name(const void *p1, const void *p2)
1757 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1758 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1760 return name_compare(e1->name, e1->len, e2->name, e2->len);
1763 static struct path_simplify *create_simplify(const char **pathspec)
1765 int nr, alloc = 0;
1766 struct path_simplify *simplify = NULL;
1768 if (!pathspec)
1769 return NULL;
1771 for (nr = 0 ; ; nr++) {
1772 const char *match;
1773 ALLOC_GROW(simplify, nr + 1, alloc);
1774 match = *pathspec++;
1775 if (!match)
1776 break;
1777 simplify[nr].path = match;
1778 simplify[nr].len = simple_length(match);
1780 simplify[nr].path = NULL;
1781 simplify[nr].len = 0;
1782 return simplify;
1785 static void free_simplify(struct path_simplify *simplify)
1787 free(simplify);
1790 static int treat_leading_path(struct dir_struct *dir,
1791 const char *path, int len,
1792 const struct path_simplify *simplify)
1794 struct strbuf sb = STRBUF_INIT;
1795 int baselen, rc = 0;
1796 const char *cp;
1797 int old_flags = dir->flags;
1799 while (len && path[len - 1] == '/')
1800 len--;
1801 if (!len)
1802 return 1;
1803 baselen = 0;
1804 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1805 while (1) {
1806 cp = path + baselen + !!baselen;
1807 cp = memchr(cp, '/', path + len - cp);
1808 if (!cp)
1809 baselen = len;
1810 else
1811 baselen = cp - path;
1812 strbuf_setlen(&sb, 0);
1813 strbuf_add(&sb, path, baselen);
1814 if (!is_directory(sb.buf))
1815 break;
1816 if (simplify_away(sb.buf, sb.len, simplify))
1817 break;
1818 if (treat_one_path(dir, NULL, &sb, baselen, simplify,
1819 DT_DIR, NULL) == path_none)
1820 break; /* do not recurse into it */
1821 if (len <= baselen) {
1822 rc = 1;
1823 break; /* finished checking */
1826 strbuf_release(&sb);
1827 dir->flags = old_flags;
1828 return rc;
1831 static const char *get_ident_string(void)
1833 static struct strbuf sb = STRBUF_INIT;
1834 struct utsname uts;
1836 if (sb.len)
1837 return sb.buf;
1838 if (uname(&uts) < 0)
1839 die_errno(_("failed to get kernel name and information"));
1840 strbuf_addf(&sb, "Location %s, system %s %s %s", get_git_work_tree(),
1841 uts.sysname, uts.release, uts.version);
1842 return sb.buf;
1845 static int ident_in_untracked(const struct untracked_cache *uc)
1847 const char *end = uc->ident.buf + uc->ident.len;
1848 const char *p = uc->ident.buf;
1850 for (p = uc->ident.buf; p < end; p += strlen(p) + 1)
1851 if (!strcmp(p, get_ident_string()))
1852 return 1;
1853 return 0;
1856 void add_untracked_ident(struct untracked_cache *uc)
1858 if (ident_in_untracked(uc))
1859 return;
1860 strbuf_addstr(&uc->ident, get_ident_string());
1861 /* this strbuf contains a list of strings, save NUL too */
1862 strbuf_addch(&uc->ident, 0);
1865 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1866 int base_len,
1867 const struct pathspec *pathspec)
1869 struct untracked_cache_dir *root;
1871 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
1872 return NULL;
1875 * We only support $GIT_DIR/info/exclude and core.excludesfile
1876 * as the global ignore rule files. Any other additions
1877 * (e.g. from command line) invalidate the cache. This
1878 * condition also catches running setup_standard_excludes()
1879 * before setting dir->untracked!
1881 if (dir->unmanaged_exclude_files)
1882 return NULL;
1885 * Optimize for the main use case only: whole-tree git
1886 * status. More work involved in treat_leading_path() if we
1887 * use cache on just a subset of the worktree. pathspec
1888 * support could make the matter even worse.
1890 if (base_len || (pathspec && pathspec->nr))
1891 return NULL;
1893 /* Different set of flags may produce different results */
1894 if (dir->flags != dir->untracked->dir_flags ||
1896 * See treat_directory(), case index_nonexistent. Without
1897 * this flag, we may need to also cache .git file content
1898 * for the resolve_gitlink_ref() call, which we don't.
1900 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1901 /* We don't support collecting ignore files */
1902 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1903 DIR_COLLECT_IGNORED)))
1904 return NULL;
1907 * If we use .gitignore in the cache and now you change it to
1908 * .gitexclude, everything will go wrong.
1910 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
1911 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
1912 return NULL;
1915 * EXC_CMDL is not considered in the cache. If people set it,
1916 * skip the cache.
1918 if (dir->exclude_list_group[EXC_CMDL].nr)
1919 return NULL;
1921 if (!ident_in_untracked(dir->untracked)) {
1922 warning(_("Untracked cache is disabled on this system."));
1923 return NULL;
1926 if (!dir->untracked->root) {
1927 const int len = sizeof(*dir->untracked->root);
1928 dir->untracked->root = xmalloc(len);
1929 memset(dir->untracked->root, 0, len);
1932 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
1933 root = dir->untracked->root;
1934 if (hashcmp(dir->ss_info_exclude.sha1,
1935 dir->untracked->ss_info_exclude.sha1)) {
1936 invalidate_gitignore(dir->untracked, root);
1937 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
1939 if (hashcmp(dir->ss_excludes_file.sha1,
1940 dir->untracked->ss_excludes_file.sha1)) {
1941 invalidate_gitignore(dir->untracked, root);
1942 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
1945 /* Make sure this directory is not dropped out at saving phase */
1946 root->recurse = 1;
1947 return root;
1950 int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1952 struct path_simplify *simplify;
1953 struct untracked_cache_dir *untracked;
1956 * Check out create_simplify()
1958 if (pathspec)
1959 GUARD_PATHSPEC(pathspec,
1960 PATHSPEC_FROMTOP |
1961 PATHSPEC_MAXDEPTH |
1962 PATHSPEC_LITERAL |
1963 PATHSPEC_GLOB |
1964 PATHSPEC_ICASE |
1965 PATHSPEC_EXCLUDE);
1967 if (has_symlink_leading_path(path, len))
1968 return dir->nr;
1971 * exclude patterns are treated like positive ones in
1972 * create_simplify. Usually exclude patterns should be a
1973 * subset of positive ones, which has no impacts on
1974 * create_simplify().
1976 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
1977 untracked = validate_untracked_cache(dir, len, pathspec);
1978 if (!untracked)
1980 * make sure untracked cache code path is disabled,
1981 * e.g. prep_exclude()
1983 dir->untracked = NULL;
1984 if (!len || treat_leading_path(dir, path, len, simplify))
1985 read_directory_recursive(dir, path, len, untracked, 0, simplify);
1986 free_simplify(simplify);
1987 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1988 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1989 if (dir->untracked) {
1990 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
1991 trace_printf_key(&trace_untracked_stats,
1992 "node creation: %u\n"
1993 "gitignore invalidation: %u\n"
1994 "directory invalidation: %u\n"
1995 "opendir: %u\n",
1996 dir->untracked->dir_created,
1997 dir->untracked->gitignore_invalidated,
1998 dir->untracked->dir_invalidated,
1999 dir->untracked->dir_opened);
2000 if (dir->untracked == the_index.untracked &&
2001 (dir->untracked->dir_opened ||
2002 dir->untracked->gitignore_invalidated ||
2003 dir->untracked->dir_invalidated))
2004 the_index.cache_changed |= UNTRACKED_CHANGED;
2005 if (dir->untracked != the_index.untracked) {
2006 free(dir->untracked);
2007 dir->untracked = NULL;
2010 return dir->nr;
2013 int file_exists(const char *f)
2015 struct stat sb;
2016 return lstat(f, &sb) == 0;
2019 static int cmp_icase(char a, char b)
2021 if (a == b)
2022 return 0;
2023 if (ignore_case)
2024 return toupper(a) - toupper(b);
2025 return a - b;
2029 * Given two normalized paths (a trailing slash is ok), if subdir is
2030 * outside dir, return -1. Otherwise return the offset in subdir that
2031 * can be used as relative path to dir.
2033 int dir_inside_of(const char *subdir, const char *dir)
2035 int offset = 0;
2037 assert(dir && subdir && *dir && *subdir);
2039 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2040 dir++;
2041 subdir++;
2042 offset++;
2045 /* hel[p]/me vs hel[l]/yeah */
2046 if (*dir && *subdir)
2047 return -1;
2049 if (!*subdir)
2050 return !*dir ? offset : -1; /* same dir */
2052 /* foo/[b]ar vs foo/[] */
2053 if (is_dir_sep(dir[-1]))
2054 return is_dir_sep(subdir[-1]) ? offset : -1;
2056 /* foo[/]bar vs foo[] */
2057 return is_dir_sep(*subdir) ? offset + 1 : -1;
2060 int is_inside_dir(const char *dir)
2062 char *cwd;
2063 int rc;
2065 if (!dir)
2066 return 0;
2068 cwd = xgetcwd();
2069 rc = (dir_inside_of(cwd, dir) >= 0);
2070 free(cwd);
2071 return rc;
2074 int is_empty_dir(const char *path)
2076 DIR *dir = opendir(path);
2077 struct dirent *e;
2078 int ret = 1;
2080 if (!dir)
2081 return 0;
2083 while ((e = readdir(dir)) != NULL)
2084 if (!is_dot_or_dotdot(e->d_name)) {
2085 ret = 0;
2086 break;
2089 closedir(dir);
2090 return ret;
2093 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2095 DIR *dir;
2096 struct dirent *e;
2097 int ret = 0, original_len = path->len, len, kept_down = 0;
2098 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2099 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2100 unsigned char submodule_head[20];
2102 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2103 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2104 /* Do not descend and nuke a nested git work tree. */
2105 if (kept_up)
2106 *kept_up = 1;
2107 return 0;
2110 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2111 dir = opendir(path->buf);
2112 if (!dir) {
2113 if (errno == ENOENT)
2114 return keep_toplevel ? -1 : 0;
2115 else if (errno == EACCES && !keep_toplevel)
2117 * An empty dir could be removable even if it
2118 * is unreadable:
2120 return rmdir(path->buf);
2121 else
2122 return -1;
2124 if (path->buf[original_len - 1] != '/')
2125 strbuf_addch(path, '/');
2127 len = path->len;
2128 while ((e = readdir(dir)) != NULL) {
2129 struct stat st;
2130 if (is_dot_or_dotdot(e->d_name))
2131 continue;
2133 strbuf_setlen(path, len);
2134 strbuf_addstr(path, e->d_name);
2135 if (lstat(path->buf, &st)) {
2136 if (errno == ENOENT)
2138 * file disappeared, which is what we
2139 * wanted anyway
2141 continue;
2142 /* fall thru */
2143 } else if (S_ISDIR(st.st_mode)) {
2144 if (!remove_dir_recurse(path, flag, &kept_down))
2145 continue; /* happy */
2146 } else if (!only_empty &&
2147 (!unlink(path->buf) || errno == ENOENT)) {
2148 continue; /* happy, too */
2151 /* path too long, stat fails, or non-directory still exists */
2152 ret = -1;
2153 break;
2155 closedir(dir);
2157 strbuf_setlen(path, original_len);
2158 if (!ret && !keep_toplevel && !kept_down)
2159 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2160 else if (kept_up)
2162 * report the uplevel that it is not an error that we
2163 * did not rmdir() our directory.
2165 *kept_up = !ret;
2166 return ret;
2169 int remove_dir_recursively(struct strbuf *path, int flag)
2171 return remove_dir_recurse(path, flag, NULL);
2174 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2176 void setup_standard_excludes(struct dir_struct *dir)
2178 const char *path;
2180 dir->exclude_per_dir = ".gitignore";
2182 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2183 if (!excludes_file)
2184 excludes_file = xdg_config_home("ignore");
2185 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2186 add_excludes_from_file_1(dir, excludes_file,
2187 dir->untracked ? &dir->ss_excludes_file : NULL);
2189 /* per repository user preference */
2190 path = git_path_info_exclude();
2191 if (!access_or_warn(path, R_OK, 0))
2192 add_excludes_from_file_1(dir, path,
2193 dir->untracked ? &dir->ss_info_exclude : NULL);
2196 int remove_path(const char *name)
2198 char *slash;
2200 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
2201 return -1;
2203 slash = strrchr(name, '/');
2204 if (slash) {
2205 char *dirs = xstrdup(name);
2206 slash = dirs + (slash - name);
2207 do {
2208 *slash = '\0';
2209 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2210 free(dirs);
2212 return 0;
2216 * Frees memory within dir which was allocated for exclude lists and
2217 * the exclude_stack. Does not free dir itself.
2219 void clear_directory(struct dir_struct *dir)
2221 int i, j;
2222 struct exclude_list_group *group;
2223 struct exclude_list *el;
2224 struct exclude_stack *stk;
2226 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2227 group = &dir->exclude_list_group[i];
2228 for (j = 0; j < group->nr; j++) {
2229 el = &group->el[j];
2230 if (i == EXC_DIRS)
2231 free((char *)el->src);
2232 clear_exclude_list(el);
2234 free(group->el);
2237 stk = dir->exclude_stack;
2238 while (stk) {
2239 struct exclude_stack *prev = stk->prev;
2240 free(stk);
2241 stk = prev;
2243 strbuf_release(&dir->basebuf);
2246 struct ondisk_untracked_cache {
2247 struct stat_data info_exclude_stat;
2248 struct stat_data excludes_file_stat;
2249 uint32_t dir_flags;
2250 unsigned char info_exclude_sha1[20];
2251 unsigned char excludes_file_sha1[20];
2252 char exclude_per_dir[FLEX_ARRAY];
2255 #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2257 struct write_data {
2258 int index; /* number of written untracked_cache_dir */
2259 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2260 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2261 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2262 struct strbuf out;
2263 struct strbuf sb_stat;
2264 struct strbuf sb_sha1;
2267 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2269 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2270 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2271 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2272 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2273 to->sd_dev = htonl(from->sd_dev);
2274 to->sd_ino = htonl(from->sd_ino);
2275 to->sd_uid = htonl(from->sd_uid);
2276 to->sd_gid = htonl(from->sd_gid);
2277 to->sd_size = htonl(from->sd_size);
2280 static void write_one_dir(struct untracked_cache_dir *untracked,
2281 struct write_data *wd)
2283 struct stat_data stat_data;
2284 struct strbuf *out = &wd->out;
2285 unsigned char intbuf[16];
2286 unsigned int intlen, value;
2287 int i = wd->index++;
2290 * untracked_nr should be reset whenever valid is clear, but
2291 * for safety..
2293 if (!untracked->valid) {
2294 untracked->untracked_nr = 0;
2295 untracked->check_only = 0;
2298 if (untracked->check_only)
2299 ewah_set(wd->check_only, i);
2300 if (untracked->valid) {
2301 ewah_set(wd->valid, i);
2302 stat_data_to_disk(&stat_data, &untracked->stat_data);
2303 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2305 if (!is_null_sha1(untracked->exclude_sha1)) {
2306 ewah_set(wd->sha1_valid, i);
2307 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2310 intlen = encode_varint(untracked->untracked_nr, intbuf);
2311 strbuf_add(out, intbuf, intlen);
2313 /* skip non-recurse directories */
2314 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2315 if (untracked->dirs[i]->recurse)
2316 value++;
2317 intlen = encode_varint(value, intbuf);
2318 strbuf_add(out, intbuf, intlen);
2320 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2322 for (i = 0; i < untracked->untracked_nr; i++)
2323 strbuf_add(out, untracked->untracked[i],
2324 strlen(untracked->untracked[i]) + 1);
2326 for (i = 0; i < untracked->dirs_nr; i++)
2327 if (untracked->dirs[i]->recurse)
2328 write_one_dir(untracked->dirs[i], wd);
2331 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2333 struct ondisk_untracked_cache *ouc;
2334 struct write_data wd;
2335 unsigned char varbuf[16];
2336 int len = 0, varint_len;
2337 if (untracked->exclude_per_dir)
2338 len = strlen(untracked->exclude_per_dir);
2339 ouc = xmalloc(sizeof(*ouc) + len + 1);
2340 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2341 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2342 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2343 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2344 ouc->dir_flags = htonl(untracked->dir_flags);
2345 memcpy(ouc->exclude_per_dir, untracked->exclude_per_dir, len + 1);
2347 varint_len = encode_varint(untracked->ident.len, varbuf);
2348 strbuf_add(out, varbuf, varint_len);
2349 strbuf_add(out, untracked->ident.buf, untracked->ident.len);
2351 strbuf_add(out, ouc, ouc_size(len));
2352 free(ouc);
2353 ouc = NULL;
2355 if (!untracked->root) {
2356 varint_len = encode_varint(0, varbuf);
2357 strbuf_add(out, varbuf, varint_len);
2358 return;
2361 wd.index = 0;
2362 wd.check_only = ewah_new();
2363 wd.valid = ewah_new();
2364 wd.sha1_valid = ewah_new();
2365 strbuf_init(&wd.out, 1024);
2366 strbuf_init(&wd.sb_stat, 1024);
2367 strbuf_init(&wd.sb_sha1, 1024);
2368 write_one_dir(untracked->root, &wd);
2370 varint_len = encode_varint(wd.index, varbuf);
2371 strbuf_add(out, varbuf, varint_len);
2372 strbuf_addbuf(out, &wd.out);
2373 ewah_serialize_strbuf(wd.valid, out);
2374 ewah_serialize_strbuf(wd.check_only, out);
2375 ewah_serialize_strbuf(wd.sha1_valid, out);
2376 strbuf_addbuf(out, &wd.sb_stat);
2377 strbuf_addbuf(out, &wd.sb_sha1);
2378 strbuf_addch(out, '\0'); /* safe guard for string lists */
2380 ewah_free(wd.valid);
2381 ewah_free(wd.check_only);
2382 ewah_free(wd.sha1_valid);
2383 strbuf_release(&wd.out);
2384 strbuf_release(&wd.sb_stat);
2385 strbuf_release(&wd.sb_sha1);
2388 static void free_untracked(struct untracked_cache_dir *ucd)
2390 int i;
2391 if (!ucd)
2392 return;
2393 for (i = 0; i < ucd->dirs_nr; i++)
2394 free_untracked(ucd->dirs[i]);
2395 for (i = 0; i < ucd->untracked_nr; i++)
2396 free(ucd->untracked[i]);
2397 free(ucd->untracked);
2398 free(ucd->dirs);
2399 free(ucd);
2402 void free_untracked_cache(struct untracked_cache *uc)
2404 if (uc)
2405 free_untracked(uc->root);
2406 free(uc);
2409 struct read_data {
2410 int index;
2411 struct untracked_cache_dir **ucd;
2412 struct ewah_bitmap *check_only;
2413 struct ewah_bitmap *valid;
2414 struct ewah_bitmap *sha1_valid;
2415 const unsigned char *data;
2416 const unsigned char *end;
2419 static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2421 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2422 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2423 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2424 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2425 to->sd_dev = get_be32(&from->sd_dev);
2426 to->sd_ino = get_be32(&from->sd_ino);
2427 to->sd_uid = get_be32(&from->sd_uid);
2428 to->sd_gid = get_be32(&from->sd_gid);
2429 to->sd_size = get_be32(&from->sd_size);
2432 static int read_one_dir(struct untracked_cache_dir **untracked_,
2433 struct read_data *rd)
2435 struct untracked_cache_dir ud, *untracked;
2436 const unsigned char *next, *data = rd->data, *end = rd->end;
2437 unsigned int value;
2438 int i, len;
2440 memset(&ud, 0, sizeof(ud));
2442 next = data;
2443 value = decode_varint(&next);
2444 if (next > end)
2445 return -1;
2446 ud.recurse = 1;
2447 ud.untracked_alloc = value;
2448 ud.untracked_nr = value;
2449 if (ud.untracked_nr)
2450 ud.untracked = xmalloc(sizeof(*ud.untracked) * ud.untracked_nr);
2451 data = next;
2453 next = data;
2454 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2455 if (next > end)
2456 return -1;
2457 ud.dirs = xmalloc(sizeof(*ud.dirs) * ud.dirs_nr);
2458 data = next;
2460 len = strlen((const char *)data);
2461 next = data + len + 1;
2462 if (next > rd->end)
2463 return -1;
2464 *untracked_ = untracked = xmalloc(sizeof(*untracked) + len);
2465 memcpy(untracked, &ud, sizeof(ud));
2466 memcpy(untracked->name, data, len + 1);
2467 data = next;
2469 for (i = 0; i < untracked->untracked_nr; i++) {
2470 len = strlen((const char *)data);
2471 next = data + len + 1;
2472 if (next > rd->end)
2473 return -1;
2474 untracked->untracked[i] = xstrdup((const char*)data);
2475 data = next;
2478 rd->ucd[rd->index++] = untracked;
2479 rd->data = data;
2481 for (i = 0; i < untracked->dirs_nr; i++) {
2482 len = read_one_dir(untracked->dirs + i, rd);
2483 if (len < 0)
2484 return -1;
2486 return 0;
2489 static void set_check_only(size_t pos, void *cb)
2491 struct read_data *rd = cb;
2492 struct untracked_cache_dir *ud = rd->ucd[pos];
2493 ud->check_only = 1;
2496 static void read_stat(size_t pos, void *cb)
2498 struct read_data *rd = cb;
2499 struct untracked_cache_dir *ud = rd->ucd[pos];
2500 if (rd->data + sizeof(struct stat_data) > rd->end) {
2501 rd->data = rd->end + 1;
2502 return;
2504 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2505 rd->data += sizeof(struct stat_data);
2506 ud->valid = 1;
2509 static void read_sha1(size_t pos, void *cb)
2511 struct read_data *rd = cb;
2512 struct untracked_cache_dir *ud = rd->ucd[pos];
2513 if (rd->data + 20 > rd->end) {
2514 rd->data = rd->end + 1;
2515 return;
2517 hashcpy(ud->exclude_sha1, rd->data);
2518 rd->data += 20;
2521 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2522 const struct stat_data *stat,
2523 const unsigned char *sha1)
2525 stat_data_from_disk(&sha1_stat->stat, stat);
2526 hashcpy(sha1_stat->sha1, sha1);
2527 sha1_stat->valid = 1;
2530 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2532 const struct ondisk_untracked_cache *ouc;
2533 struct untracked_cache *uc;
2534 struct read_data rd;
2535 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2536 const char *ident;
2537 int ident_len, len;
2539 if (sz <= 1 || end[-1] != '\0')
2540 return NULL;
2541 end--;
2543 ident_len = decode_varint(&next);
2544 if (next + ident_len > end)
2545 return NULL;
2546 ident = (const char *)next;
2547 next += ident_len;
2549 ouc = (const struct ondisk_untracked_cache *)next;
2550 if (next + ouc_size(0) > end)
2551 return NULL;
2553 uc = xcalloc(1, sizeof(*uc));
2554 strbuf_init(&uc->ident, ident_len);
2555 strbuf_add(&uc->ident, ident, ident_len);
2556 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2557 ouc->info_exclude_sha1);
2558 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2559 ouc->excludes_file_sha1);
2560 uc->dir_flags = get_be32(&ouc->dir_flags);
2561 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2562 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2563 next += ouc_size(strlen(ouc->exclude_per_dir));
2564 if (next >= end)
2565 goto done2;
2567 len = decode_varint(&next);
2568 if (next > end || len == 0)
2569 goto done2;
2571 rd.valid = ewah_new();
2572 rd.check_only = ewah_new();
2573 rd.sha1_valid = ewah_new();
2574 rd.data = next;
2575 rd.end = end;
2576 rd.index = 0;
2577 rd.ucd = xmalloc(sizeof(*rd.ucd) * len);
2579 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2580 goto done;
2582 next = rd.data;
2583 len = ewah_read_mmap(rd.valid, next, end - next);
2584 if (len < 0)
2585 goto done;
2587 next += len;
2588 len = ewah_read_mmap(rd.check_only, next, end - next);
2589 if (len < 0)
2590 goto done;
2592 next += len;
2593 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2594 if (len < 0)
2595 goto done;
2597 ewah_each_bit(rd.check_only, set_check_only, &rd);
2598 rd.data = next + len;
2599 ewah_each_bit(rd.valid, read_stat, &rd);
2600 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2601 next = rd.data;
2603 done:
2604 free(rd.ucd);
2605 ewah_free(rd.valid);
2606 ewah_free(rd.check_only);
2607 ewah_free(rd.sha1_valid);
2608 done2:
2609 if (next != end) {
2610 free_untracked_cache(uc);
2611 uc = NULL;
2613 return uc;
2616 static void invalidate_one_directory(struct untracked_cache *uc,
2617 struct untracked_cache_dir *ucd)
2619 uc->dir_invalidated++;
2620 ucd->valid = 0;
2621 ucd->untracked_nr = 0;
2625 * Normally when an entry is added or removed from a directory,
2626 * invalidating that directory is enough. No need to touch its
2627 * ancestors. When a directory is shown as "foo/bar/" in git-status
2628 * however, deleting or adding an entry may have cascading effect.
2630 * Say the "foo/bar/file" has become untracked, we need to tell the
2631 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2632 * directory any more (because "bar" is managed by foo as an untracked
2633 * "file").
2635 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2636 * was the last untracked entry in the entire "foo", we should show
2637 * "foo/" instead. Which means we have to invalidate past "bar" up to
2638 * "foo".
2640 * This function traverses all directories from root to leaf. If there
2641 * is a chance of one of the above cases happening, we invalidate back
2642 * to root. Otherwise we just invalidate the leaf. There may be a more
2643 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2644 * detect these cases and avoid unnecessary invalidation, for example,
2645 * checking for the untracked entry named "bar/" in "foo", but for now
2646 * stick to something safe and simple.
2648 static int invalidate_one_component(struct untracked_cache *uc,
2649 struct untracked_cache_dir *dir,
2650 const char *path, int len)
2652 const char *rest = strchr(path, '/');
2654 if (rest) {
2655 int component_len = rest - path;
2656 struct untracked_cache_dir *d =
2657 lookup_untracked(uc, dir, path, component_len);
2658 int ret =
2659 invalidate_one_component(uc, d, rest + 1,
2660 len - (component_len + 1));
2661 if (ret)
2662 invalidate_one_directory(uc, dir);
2663 return ret;
2666 invalidate_one_directory(uc, dir);
2667 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2670 void untracked_cache_invalidate_path(struct index_state *istate,
2671 const char *path)
2673 if (!istate->untracked || !istate->untracked->root)
2674 return;
2675 invalidate_one_component(istate->untracked, istate->untracked->root,
2676 path, strlen(path));
2679 void untracked_cache_remove_from_index(struct index_state *istate,
2680 const char *path)
2682 untracked_cache_invalidate_path(istate, path);
2685 void untracked_cache_add_to_index(struct index_state *istate,
2686 const char *path)
2688 untracked_cache_invalidate_path(istate, path);