gpg-interface: prefer "long" key format output when verifying pgp signatures
[git/debian.git] / dir.c
blob656f272adc69b633ddd06277bf9cc77205a7e8ee
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 git_fnmatch(const struct pathspec_item *item,
68 const char *pattern, const char *string,
69 int prefix)
71 if (prefix > 0) {
72 if (ps_strncmp(item, pattern, string, prefix))
73 return WM_NOMATCH;
74 pattern += prefix;
75 string += prefix;
77 if (item->flags & PATHSPEC_ONESTAR) {
78 int pattern_len = strlen(++pattern);
79 int string_len = strlen(string);
80 return string_len < pattern_len ||
81 ps_strcmp(item, pattern,
82 string + string_len - pattern_len);
84 if (item->magic & PATHSPEC_GLOB)
85 return wildmatch(pattern, string,
86 WM_PATHNAME |
87 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
88 NULL);
89 else
90 /* wildmatch has not learned no FNM_PATHNAME mode yet */
91 return wildmatch(pattern, string,
92 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
93 NULL);
96 static int fnmatch_icase_mem(const char *pattern, int patternlen,
97 const char *string, int stringlen,
98 int flags)
100 int match_status;
101 struct strbuf pat_buf = STRBUF_INIT;
102 struct strbuf str_buf = STRBUF_INIT;
103 const char *use_pat = pattern;
104 const char *use_str = string;
106 if (pattern[patternlen]) {
107 strbuf_add(&pat_buf, pattern, patternlen);
108 use_pat = pat_buf.buf;
110 if (string[stringlen]) {
111 strbuf_add(&str_buf, string, stringlen);
112 use_str = str_buf.buf;
115 if (ignore_case)
116 flags |= WM_CASEFOLD;
117 match_status = wildmatch(use_pat, use_str, flags, NULL);
119 strbuf_release(&pat_buf);
120 strbuf_release(&str_buf);
122 return match_status;
125 static size_t common_prefix_len(const struct pathspec *pathspec)
127 int n;
128 size_t max = 0;
131 * ":(icase)path" is treated as a pathspec full of
132 * wildcard. In other words, only prefix is considered common
133 * prefix. If the pathspec is abc/foo abc/bar, running in
134 * subdir xyz, the common prefix is still xyz, not xuz/abc as
135 * in non-:(icase).
137 GUARD_PATHSPEC(pathspec,
138 PATHSPEC_FROMTOP |
139 PATHSPEC_MAXDEPTH |
140 PATHSPEC_LITERAL |
141 PATHSPEC_GLOB |
142 PATHSPEC_ICASE |
143 PATHSPEC_EXCLUDE);
145 for (n = 0; n < pathspec->nr; n++) {
146 size_t i = 0, len = 0, item_len;
147 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
148 continue;
149 if (pathspec->items[n].magic & PATHSPEC_ICASE)
150 item_len = pathspec->items[n].prefix;
151 else
152 item_len = pathspec->items[n].nowildcard_len;
153 while (i < item_len && (n == 0 || i < max)) {
154 char c = pathspec->items[n].match[i];
155 if (c != pathspec->items[0].match[i])
156 break;
157 if (c == '/')
158 len = i + 1;
159 i++;
161 if (n == 0 || len < max) {
162 max = len;
163 if (!max)
164 break;
167 return max;
171 * Returns a copy of the longest leading path common among all
172 * pathspecs.
174 char *common_prefix(const struct pathspec *pathspec)
176 unsigned long len = common_prefix_len(pathspec);
178 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
181 int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
183 size_t len;
186 * Calculate common prefix for the pathspec, and
187 * use that to optimize the directory walk
189 len = common_prefix_len(pathspec);
191 /* Read the directory and prune it */
192 read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec);
193 return len;
196 int within_depth(const char *name, int namelen,
197 int depth, int max_depth)
199 const char *cp = name, *cpe = name + namelen;
201 while (cp < cpe) {
202 if (*cp++ != '/')
203 continue;
204 depth++;
205 if (depth > max_depth)
206 return 0;
208 return 1;
211 #define DO_MATCH_EXCLUDE 1
212 #define DO_MATCH_DIRECTORY 2
215 * Does 'match' match the given name?
216 * A match is found if
218 * (1) the 'match' string is leading directory of 'name', or
219 * (2) the 'match' string is a wildcard and matches 'name', or
220 * (3) the 'match' string is exactly the same as 'name'.
222 * and the return value tells which case it was.
224 * It returns 0 when there is no match.
226 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
227 const char *name, int namelen, unsigned flags)
229 /* name/namelen has prefix cut off by caller */
230 const char *match = item->match + prefix;
231 int matchlen = item->len - prefix;
234 * The normal call pattern is:
235 * 1. prefix = common_prefix_len(ps);
236 * 2. prune something, or fill_directory
237 * 3. match_pathspec()
239 * 'prefix' at #1 may be shorter than the command's prefix and
240 * it's ok for #2 to match extra files. Those extras will be
241 * trimmed at #3.
243 * Suppose the pathspec is 'foo' and '../bar' running from
244 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
245 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
246 * user does not want XYZ/foo, only the "foo" part should be
247 * case-insensitive. We need to filter out XYZ/foo here. In
248 * other words, we do not trust the caller on comparing the
249 * prefix part when :(icase) is involved. We do exact
250 * comparison ourselves.
252 * Normally the caller (common_prefix_len() in fact) does
253 * _exact_ matching on name[-prefix+1..-1] and we do not need
254 * to check that part. Be defensive and check it anyway, in
255 * case common_prefix_len is changed, or a new caller is
256 * introduced that does not use common_prefix_len.
258 * If the penalty turns out too high when prefix is really
259 * long, maybe change it to
260 * strncmp(match, name, item->prefix - prefix)
262 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
263 strncmp(item->match, name - prefix, item->prefix))
264 return 0;
266 /* If the match was just the prefix, we matched */
267 if (!*match)
268 return MATCHED_RECURSIVELY;
270 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
271 if (matchlen == namelen)
272 return MATCHED_EXACTLY;
274 if (match[matchlen-1] == '/' || name[matchlen] == '/')
275 return MATCHED_RECURSIVELY;
276 } else if ((flags & DO_MATCH_DIRECTORY) &&
277 match[matchlen - 1] == '/' &&
278 namelen == matchlen - 1 &&
279 !ps_strncmp(item, match, name, namelen))
280 return MATCHED_EXACTLY;
282 if (item->nowildcard_len < item->len &&
283 !git_fnmatch(item, match, name,
284 item->nowildcard_len - prefix))
285 return MATCHED_FNMATCH;
287 return 0;
291 * Given a name and a list of pathspecs, returns the nature of the
292 * closest (i.e. most specific) match of the name to any of the
293 * pathspecs.
295 * The caller typically calls this multiple times with the same
296 * pathspec and seen[] array but with different name/namelen
297 * (e.g. entries from the index) and is interested in seeing if and
298 * how each pathspec matches all the names it calls this function
299 * with. A mark is left in the seen[] array for each pathspec element
300 * indicating the closest type of match that element achieved, so if
301 * seen[n] remains zero after multiple invocations, that means the nth
302 * pathspec did not match any names, which could indicate that the
303 * user mistyped the nth pathspec.
305 static int do_match_pathspec(const struct pathspec *ps,
306 const char *name, int namelen,
307 int prefix, char *seen,
308 unsigned flags)
310 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
312 GUARD_PATHSPEC(ps,
313 PATHSPEC_FROMTOP |
314 PATHSPEC_MAXDEPTH |
315 PATHSPEC_LITERAL |
316 PATHSPEC_GLOB |
317 PATHSPEC_ICASE |
318 PATHSPEC_EXCLUDE);
320 if (!ps->nr) {
321 if (!ps->recursive ||
322 !(ps->magic & PATHSPEC_MAXDEPTH) ||
323 ps->max_depth == -1)
324 return MATCHED_RECURSIVELY;
326 if (within_depth(name, namelen, 0, ps->max_depth))
327 return MATCHED_EXACTLY;
328 else
329 return 0;
332 name += prefix;
333 namelen -= prefix;
335 for (i = ps->nr - 1; i >= 0; i--) {
336 int how;
338 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
339 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
340 continue;
342 if (seen && seen[i] == MATCHED_EXACTLY)
343 continue;
345 * Make exclude patterns optional and never report
346 * "pathspec ':(exclude)foo' matches no files"
348 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
349 seen[i] = MATCHED_FNMATCH;
350 how = match_pathspec_item(ps->items+i, prefix, name,
351 namelen, flags);
352 if (ps->recursive &&
353 (ps->magic & PATHSPEC_MAXDEPTH) &&
354 ps->max_depth != -1 &&
355 how && how != MATCHED_FNMATCH) {
356 int len = ps->items[i].len;
357 if (name[len] == '/')
358 len++;
359 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
360 how = MATCHED_EXACTLY;
361 else
362 how = 0;
364 if (how) {
365 if (retval < how)
366 retval = how;
367 if (seen && seen[i] < how)
368 seen[i] = how;
371 return retval;
374 int match_pathspec(const struct pathspec *ps,
375 const char *name, int namelen,
376 int prefix, char *seen, int is_dir)
378 int positive, negative;
379 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
380 positive = do_match_pathspec(ps, name, namelen,
381 prefix, seen, flags);
382 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
383 return positive;
384 negative = do_match_pathspec(ps, name, namelen,
385 prefix, seen,
386 flags | DO_MATCH_EXCLUDE);
387 return negative ? 0 : positive;
390 int report_path_error(const char *ps_matched,
391 const struct pathspec *pathspec,
392 const char *prefix)
395 * Make sure all pathspec matched; otherwise it is an error.
397 int num, errors = 0;
398 for (num = 0; num < pathspec->nr; num++) {
399 int other, found_dup;
401 if (ps_matched[num])
402 continue;
404 * The caller might have fed identical pathspec
405 * twice. Do not barf on such a mistake.
406 * FIXME: parse_pathspec should have eliminated
407 * duplicate pathspec.
409 for (found_dup = other = 0;
410 !found_dup && other < pathspec->nr;
411 other++) {
412 if (other == num || !ps_matched[other])
413 continue;
414 if (!strcmp(pathspec->items[other].original,
415 pathspec->items[num].original))
417 * Ok, we have a match already.
419 found_dup = 1;
421 if (found_dup)
422 continue;
424 error("pathspec '%s' did not match any file(s) known to git.",
425 pathspec->items[num].original);
426 errors++;
428 return errors;
432 * Return the length of the "simple" part of a path match limiter.
434 int simple_length(const char *match)
436 int len = -1;
438 for (;;) {
439 unsigned char c = *match++;
440 len++;
441 if (c == '\0' || is_glob_special(c))
442 return len;
446 int no_wildcard(const char *string)
448 return string[simple_length(string)] == '\0';
451 void parse_exclude_pattern(const char **pattern,
452 int *patternlen,
453 unsigned *flags,
454 int *nowildcardlen)
456 const char *p = *pattern;
457 size_t i, len;
459 *flags = 0;
460 if (*p == '!') {
461 *flags |= EXC_FLAG_NEGATIVE;
462 p++;
464 len = strlen(p);
465 if (len && p[len - 1] == '/') {
466 len--;
467 *flags |= EXC_FLAG_MUSTBEDIR;
469 for (i = 0; i < len; i++) {
470 if (p[i] == '/')
471 break;
473 if (i == len)
474 *flags |= EXC_FLAG_NODIR;
475 *nowildcardlen = simple_length(p);
477 * we should have excluded the trailing slash from 'p' too,
478 * but that's one more allocation. Instead just make sure
479 * nowildcardlen does not exceed real patternlen
481 if (*nowildcardlen > len)
482 *nowildcardlen = len;
483 if (*p == '*' && no_wildcard(p + 1))
484 *flags |= EXC_FLAG_ENDSWITH;
485 *pattern = p;
486 *patternlen = len;
489 void add_exclude(const char *string, const char *base,
490 int baselen, struct exclude_list *el, int srcpos)
492 struct exclude *x;
493 int patternlen;
494 unsigned flags;
495 int nowildcardlen;
497 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
498 if (flags & EXC_FLAG_MUSTBEDIR) {
499 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
500 } else {
501 x = xmalloc(sizeof(*x));
502 x->pattern = string;
504 x->patternlen = patternlen;
505 x->nowildcardlen = nowildcardlen;
506 x->base = base;
507 x->baselen = baselen;
508 x->flags = flags;
509 x->srcpos = srcpos;
510 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
511 el->excludes[el->nr++] = x;
512 x->el = el;
515 static void *read_skip_worktree_file_from_index(const char *path, size_t *size,
516 struct sha1_stat *sha1_stat)
518 int pos, len;
519 unsigned long sz;
520 enum object_type type;
521 void *data;
523 len = strlen(path);
524 pos = cache_name_pos(path, len);
525 if (pos < 0)
526 return NULL;
527 if (!ce_skip_worktree(active_cache[pos]))
528 return NULL;
529 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
530 if (!data || type != OBJ_BLOB) {
531 free(data);
532 return NULL;
534 *size = xsize_t(sz);
535 if (sha1_stat) {
536 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
537 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
539 return data;
543 * Frees memory within el which was allocated for exclude patterns and
544 * the file buffer. Does not free el itself.
546 void clear_exclude_list(struct exclude_list *el)
548 int i;
550 for (i = 0; i < el->nr; i++)
551 free(el->excludes[i]);
552 free(el->excludes);
553 free(el->filebuf);
555 memset(el, 0, sizeof(*el));
558 static void trim_trailing_spaces(char *buf)
560 char *p, *last_space = NULL;
562 for (p = buf; *p; p++)
563 switch (*p) {
564 case ' ':
565 if (!last_space)
566 last_space = p;
567 break;
568 case '\\':
569 p++;
570 if (!*p)
571 return;
572 /* fallthrough */
573 default:
574 last_space = NULL;
577 if (last_space)
578 *last_space = '\0';
582 * Given a subdirectory name and "dir" of the current directory,
583 * search the subdir in "dir" and return it, or create a new one if it
584 * does not exist in "dir".
586 * If "name" has the trailing slash, it'll be excluded in the search.
588 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
589 struct untracked_cache_dir *dir,
590 const char *name, int len)
592 int first, last;
593 struct untracked_cache_dir *d;
594 if (!dir)
595 return NULL;
596 if (len && name[len - 1] == '/')
597 len--;
598 first = 0;
599 last = dir->dirs_nr;
600 while (last > first) {
601 int cmp, next = (last + first) >> 1;
602 d = dir->dirs[next];
603 cmp = strncmp(name, d->name, len);
604 if (!cmp && strlen(d->name) > len)
605 cmp = -1;
606 if (!cmp)
607 return d;
608 if (cmp < 0) {
609 last = next;
610 continue;
612 first = next+1;
615 uc->dir_created++;
616 FLEX_ALLOC_MEM(d, name, name, len);
618 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
619 memmove(dir->dirs + first + 1, dir->dirs + first,
620 (dir->dirs_nr - first) * sizeof(*dir->dirs));
621 dir->dirs_nr++;
622 dir->dirs[first] = d;
623 return d;
626 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
628 int i;
629 dir->valid = 0;
630 dir->untracked_nr = 0;
631 for (i = 0; i < dir->dirs_nr; i++)
632 do_invalidate_gitignore(dir->dirs[i]);
635 static void invalidate_gitignore(struct untracked_cache *uc,
636 struct untracked_cache_dir *dir)
638 uc->gitignore_invalidated++;
639 do_invalidate_gitignore(dir);
642 static void invalidate_directory(struct untracked_cache *uc,
643 struct untracked_cache_dir *dir)
645 int i;
646 uc->dir_invalidated++;
647 dir->valid = 0;
648 dir->untracked_nr = 0;
649 for (i = 0; i < dir->dirs_nr; i++)
650 dir->dirs[i]->recurse = 0;
654 * Given a file with name "fname", read it (either from disk, or from
655 * the index if "check_index" is non-zero), parse it and store the
656 * exclude rules in "el".
658 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
659 * stat data from disk (only valid if add_excludes returns zero). If
660 * ss_valid is non-zero, "ss" must contain good value as input.
662 static int add_excludes(const char *fname, const char *base, int baselen,
663 struct exclude_list *el, int check_index,
664 struct sha1_stat *sha1_stat)
666 struct stat st;
667 int fd, i, lineno = 1;
668 size_t size = 0;
669 char *buf, *entry;
671 fd = open(fname, O_RDONLY);
672 if (fd < 0 || fstat(fd, &st) < 0) {
673 if (errno != ENOENT)
674 warn_on_inaccessible(fname);
675 if (0 <= fd)
676 close(fd);
677 if (!check_index ||
678 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL)
679 return -1;
680 if (size == 0) {
681 free(buf);
682 return 0;
684 if (buf[size-1] != '\n') {
685 buf = xrealloc(buf, st_add(size, 1));
686 buf[size++] = '\n';
688 } else {
689 size = xsize_t(st.st_size);
690 if (size == 0) {
691 if (sha1_stat) {
692 fill_stat_data(&sha1_stat->stat, &st);
693 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
694 sha1_stat->valid = 1;
696 close(fd);
697 return 0;
699 buf = xmallocz(size);
700 if (read_in_full(fd, buf, size) != size) {
701 free(buf);
702 close(fd);
703 return -1;
705 buf[size++] = '\n';
706 close(fd);
707 if (sha1_stat) {
708 int pos;
709 if (sha1_stat->valid &&
710 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st))
711 ; /* no content change, ss->sha1 still good */
712 else if (check_index &&
713 (pos = cache_name_pos(fname, strlen(fname))) >= 0 &&
714 !ce_stage(active_cache[pos]) &&
715 ce_uptodate(active_cache[pos]) &&
716 !would_convert_to_git(fname))
717 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
718 else
719 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
720 fill_stat_data(&sha1_stat->stat, &st);
721 sha1_stat->valid = 1;
725 el->filebuf = buf;
727 if (skip_utf8_bom(&buf, size))
728 size -= buf - el->filebuf;
730 entry = buf;
732 for (i = 0; i < size; i++) {
733 if (buf[i] == '\n') {
734 if (entry != buf + i && entry[0] != '#') {
735 buf[i - (i && buf[i-1] == '\r')] = 0;
736 trim_trailing_spaces(entry);
737 add_exclude(entry, base, baselen, el, lineno);
739 lineno++;
740 entry = buf + i + 1;
743 return 0;
746 int add_excludes_from_file_to_list(const char *fname, const char *base,
747 int baselen, struct exclude_list *el,
748 int check_index)
750 return add_excludes(fname, base, baselen, el, check_index, NULL);
753 struct exclude_list *add_exclude_list(struct dir_struct *dir,
754 int group_type, const char *src)
756 struct exclude_list *el;
757 struct exclude_list_group *group;
759 group = &dir->exclude_list_group[group_type];
760 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
761 el = &group->el[group->nr++];
762 memset(el, 0, sizeof(*el));
763 el->src = src;
764 return el;
768 * Used to set up core.excludesfile and .git/info/exclude lists.
770 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
771 struct sha1_stat *sha1_stat)
773 struct exclude_list *el;
775 * catch setup_standard_excludes() that's called before
776 * dir->untracked is assigned. That function behaves
777 * differently when dir->untracked is non-NULL.
779 if (!dir->untracked)
780 dir->unmanaged_exclude_files++;
781 el = add_exclude_list(dir, EXC_FILE, fname);
782 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0)
783 die("cannot use %s as an exclude file", fname);
786 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
788 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
789 add_excludes_from_file_1(dir, fname, NULL);
792 int match_basename(const char *basename, int basenamelen,
793 const char *pattern, int prefix, int patternlen,
794 unsigned flags)
796 if (prefix == patternlen) {
797 if (patternlen == basenamelen &&
798 !strncmp_icase(pattern, basename, basenamelen))
799 return 1;
800 } else if (flags & EXC_FLAG_ENDSWITH) {
801 /* "*literal" matching against "fooliteral" */
802 if (patternlen - 1 <= basenamelen &&
803 !strncmp_icase(pattern + 1,
804 basename + basenamelen - (patternlen - 1),
805 patternlen - 1))
806 return 1;
807 } else {
808 if (fnmatch_icase_mem(pattern, patternlen,
809 basename, basenamelen,
810 0) == 0)
811 return 1;
813 return 0;
816 int match_pathname(const char *pathname, int pathlen,
817 const char *base, int baselen,
818 const char *pattern, int prefix, int patternlen,
819 unsigned flags)
821 const char *name;
822 int namelen;
825 * match with FNM_PATHNAME; the pattern has base implicitly
826 * in front of it.
828 if (*pattern == '/') {
829 pattern++;
830 patternlen--;
831 prefix--;
835 * baselen does not count the trailing slash. base[] may or
836 * may not end with a trailing slash though.
838 if (pathlen < baselen + 1 ||
839 (baselen && pathname[baselen] != '/') ||
840 strncmp_icase(pathname, base, baselen))
841 return 0;
843 namelen = baselen ? pathlen - baselen - 1 : pathlen;
844 name = pathname + pathlen - namelen;
846 if (prefix) {
848 * if the non-wildcard part is longer than the
849 * remaining pathname, surely it cannot match.
851 if (prefix > namelen)
852 return 0;
854 if (strncmp_icase(pattern, name, prefix))
855 return 0;
856 pattern += prefix;
857 patternlen -= prefix;
858 name += prefix;
859 namelen -= prefix;
862 * If the whole pattern did not have a wildcard,
863 * then our prefix match is all we need; we
864 * do not need to call fnmatch at all.
866 if (!patternlen && !namelen)
867 return 1;
870 return fnmatch_icase_mem(pattern, patternlen,
871 name, namelen,
872 WM_PATHNAME) == 0;
876 * Scan the given exclude list in reverse to see whether pathname
877 * should be ignored. The first match (i.e. the last on the list), if
878 * any, determines the fate. Returns the exclude_list element which
879 * matched, or NULL for undecided.
881 static struct exclude *last_exclude_matching_from_list(const char *pathname,
882 int pathlen,
883 const char *basename,
884 int *dtype,
885 struct exclude_list *el)
887 struct exclude *exc = NULL; /* undecided */
888 int i;
890 if (!el->nr)
891 return NULL; /* undefined */
893 for (i = el->nr - 1; 0 <= i; i--) {
894 struct exclude *x = el->excludes[i];
895 const char *exclude = x->pattern;
896 int prefix = x->nowildcardlen;
898 if (x->flags & EXC_FLAG_MUSTBEDIR) {
899 if (*dtype == DT_UNKNOWN)
900 *dtype = get_dtype(NULL, pathname, pathlen);
901 if (*dtype != DT_DIR)
902 continue;
905 if (x->flags & EXC_FLAG_NODIR) {
906 if (match_basename(basename,
907 pathlen - (basename - pathname),
908 exclude, prefix, x->patternlen,
909 x->flags)) {
910 exc = x;
911 break;
913 continue;
916 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
917 if (match_pathname(pathname, pathlen,
918 x->base, x->baselen ? x->baselen - 1 : 0,
919 exclude, prefix, x->patternlen, x->flags)) {
920 exc = x;
921 break;
924 return exc;
928 * Scan the list and let the last match determine the fate.
929 * Return 1 for exclude, 0 for include and -1 for undecided.
931 int is_excluded_from_list(const char *pathname,
932 int pathlen, const char *basename, int *dtype,
933 struct exclude_list *el)
935 struct exclude *exclude;
936 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
937 if (exclude)
938 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
939 return -1; /* undecided */
942 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
943 const char *pathname, int pathlen, const char *basename,
944 int *dtype_p)
946 int i, j;
947 struct exclude_list_group *group;
948 struct exclude *exclude;
949 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
950 group = &dir->exclude_list_group[i];
951 for (j = group->nr - 1; j >= 0; j--) {
952 exclude = last_exclude_matching_from_list(
953 pathname, pathlen, basename, dtype_p,
954 &group->el[j]);
955 if (exclude)
956 return exclude;
959 return NULL;
963 * Loads the per-directory exclude list for the substring of base
964 * which has a char length of baselen.
966 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
968 struct exclude_list_group *group;
969 struct exclude_list *el;
970 struct exclude_stack *stk = NULL;
971 struct untracked_cache_dir *untracked;
972 int current;
974 group = &dir->exclude_list_group[EXC_DIRS];
977 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
978 * which originate from directories not in the prefix of the
979 * path being checked.
981 while ((stk = dir->exclude_stack) != NULL) {
982 if (stk->baselen <= baselen &&
983 !strncmp(dir->basebuf.buf, base, stk->baselen))
984 break;
985 el = &group->el[dir->exclude_stack->exclude_ix];
986 dir->exclude_stack = stk->prev;
987 dir->exclude = NULL;
988 free((char *)el->src); /* see strbuf_detach() below */
989 clear_exclude_list(el);
990 free(stk);
991 group->nr--;
994 /* Skip traversing into sub directories if the parent is excluded */
995 if (dir->exclude)
996 return;
999 * Lazy initialization. All call sites currently just
1000 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1001 * them seems lots of work for little benefit.
1003 if (!dir->basebuf.buf)
1004 strbuf_init(&dir->basebuf, PATH_MAX);
1006 /* Read from the parent directories and push them down. */
1007 current = stk ? stk->baselen : -1;
1008 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1009 if (dir->untracked)
1010 untracked = stk ? stk->ucd : dir->untracked->root;
1011 else
1012 untracked = NULL;
1014 while (current < baselen) {
1015 const char *cp;
1016 struct sha1_stat sha1_stat;
1018 stk = xcalloc(1, sizeof(*stk));
1019 if (current < 0) {
1020 cp = base;
1021 current = 0;
1022 } else {
1023 cp = strchr(base + current + 1, '/');
1024 if (!cp)
1025 die("oops in prep_exclude");
1026 cp++;
1027 untracked =
1028 lookup_untracked(dir->untracked, untracked,
1029 base + current,
1030 cp - base - current);
1032 stk->prev = dir->exclude_stack;
1033 stk->baselen = cp - base;
1034 stk->exclude_ix = group->nr;
1035 stk->ucd = untracked;
1036 el = add_exclude_list(dir, EXC_DIRS, NULL);
1037 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1038 assert(stk->baselen == dir->basebuf.len);
1040 /* Abort if the directory is excluded */
1041 if (stk->baselen) {
1042 int dt = DT_DIR;
1043 dir->basebuf.buf[stk->baselen - 1] = 0;
1044 dir->exclude = last_exclude_matching_from_lists(dir,
1045 dir->basebuf.buf, stk->baselen - 1,
1046 dir->basebuf.buf + current, &dt);
1047 dir->basebuf.buf[stk->baselen - 1] = '/';
1048 if (dir->exclude &&
1049 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1050 dir->exclude = NULL;
1051 if (dir->exclude) {
1052 dir->exclude_stack = stk;
1053 return;
1057 /* Try to read per-directory file */
1058 hashclr(sha1_stat.sha1);
1059 sha1_stat.valid = 0;
1060 if (dir->exclude_per_dir &&
1062 * If we know that no files have been added in
1063 * this directory (i.e. valid_cached_dir() has
1064 * been executed and set untracked->valid) ..
1066 (!untracked || !untracked->valid ||
1068 * .. and .gitignore does not exist before
1069 * (i.e. null exclude_sha1). Then we can skip
1070 * loading .gitignore, which would result in
1071 * ENOENT anyway.
1073 !is_null_sha1(untracked->exclude_sha1))) {
1075 * dir->basebuf gets reused by the traversal, but we
1076 * need fname to remain unchanged to ensure the src
1077 * member of each struct exclude correctly
1078 * back-references its source file. Other invocations
1079 * of add_exclude_list provide stable strings, so we
1080 * strbuf_detach() and free() here in the caller.
1082 struct strbuf sb = STRBUF_INIT;
1083 strbuf_addbuf(&sb, &dir->basebuf);
1084 strbuf_addstr(&sb, dir->exclude_per_dir);
1085 el->src = strbuf_detach(&sb, NULL);
1086 add_excludes(el->src, el->src, stk->baselen, el, 1,
1087 untracked ? &sha1_stat : NULL);
1090 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1091 * will first be called in valid_cached_dir() then maybe many
1092 * times more in last_exclude_matching(). When the cache is
1093 * used, last_exclude_matching() will not be called and
1094 * reading .gitignore content will be a waste.
1096 * So when it's called by valid_cached_dir() and we can get
1097 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1098 * modified on work tree), we could delay reading the
1099 * .gitignore content until we absolutely need it in
1100 * last_exclude_matching(). Be careful about ignore rule
1101 * order, though, if you do that.
1103 if (untracked &&
1104 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1105 invalidate_gitignore(dir->untracked, untracked);
1106 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1108 dir->exclude_stack = stk;
1109 current = stk->baselen;
1111 strbuf_setlen(&dir->basebuf, baselen);
1115 * Loads the exclude lists for the directory containing pathname, then
1116 * scans all exclude lists to determine whether pathname is excluded.
1117 * Returns the exclude_list element which matched, or NULL for
1118 * undecided.
1120 struct exclude *last_exclude_matching(struct dir_struct *dir,
1121 const char *pathname,
1122 int *dtype_p)
1124 int pathlen = strlen(pathname);
1125 const char *basename = strrchr(pathname, '/');
1126 basename = (basename) ? basename+1 : pathname;
1128 prep_exclude(dir, pathname, basename-pathname);
1130 if (dir->exclude)
1131 return dir->exclude;
1133 return last_exclude_matching_from_lists(dir, pathname, pathlen,
1134 basename, dtype_p);
1138 * Loads the exclude lists for the directory containing pathname, then
1139 * scans all exclude lists to determine whether pathname is excluded.
1140 * Returns 1 if true, otherwise 0.
1142 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1144 struct exclude *exclude =
1145 last_exclude_matching(dir, pathname, dtype_p);
1146 if (exclude)
1147 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1148 return 0;
1151 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1153 struct dir_entry *ent;
1155 FLEX_ALLOC_MEM(ent, name, pathname, len);
1156 ent->len = len;
1157 return ent;
1160 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1162 if (cache_file_exists(pathname, len, ignore_case))
1163 return NULL;
1165 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1166 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1169 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1171 if (!cache_name_is_other(pathname, len))
1172 return NULL;
1174 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1175 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1178 enum exist_status {
1179 index_nonexistent = 0,
1180 index_directory,
1181 index_gitdir
1185 * Do not use the alphabetically sorted index to look up
1186 * the directory name; instead, use the case insensitive
1187 * directory hash.
1189 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1191 struct cache_entry *ce;
1193 if (cache_dir_exists(dirname, len))
1194 return index_directory;
1196 ce = cache_file_exists(dirname, len, ignore_case);
1197 if (ce && S_ISGITLINK(ce->ce_mode))
1198 return index_gitdir;
1200 return index_nonexistent;
1204 * The index sorts alphabetically by entry name, which
1205 * means that a gitlink sorts as '\0' at the end, while
1206 * a directory (which is defined not as an entry, but as
1207 * the files it contains) will sort with the '/' at the
1208 * end.
1210 static enum exist_status directory_exists_in_index(const char *dirname, int len)
1212 int pos;
1214 if (ignore_case)
1215 return directory_exists_in_index_icase(dirname, len);
1217 pos = cache_name_pos(dirname, len);
1218 if (pos < 0)
1219 pos = -pos-1;
1220 while (pos < active_nr) {
1221 const struct cache_entry *ce = active_cache[pos++];
1222 unsigned char endchar;
1224 if (strncmp(ce->name, dirname, len))
1225 break;
1226 endchar = ce->name[len];
1227 if (endchar > '/')
1228 break;
1229 if (endchar == '/')
1230 return index_directory;
1231 if (!endchar && S_ISGITLINK(ce->ce_mode))
1232 return index_gitdir;
1234 return index_nonexistent;
1238 * When we find a directory when traversing the filesystem, we
1239 * have three distinct cases:
1241 * - ignore it
1242 * - see it as a directory
1243 * - recurse into it
1245 * and which one we choose depends on a combination of existing
1246 * git index contents and the flags passed into the directory
1247 * traversal routine.
1249 * Case 1: If we *already* have entries in the index under that
1250 * directory name, we always recurse into the directory to see
1251 * all the files.
1253 * Case 2: If we *already* have that directory name as a gitlink,
1254 * we always continue to see it as a gitlink, regardless of whether
1255 * there is an actual git directory there or not (it might not
1256 * be checked out as a subproject!)
1258 * Case 3: if we didn't have it in the index previously, we
1259 * have a few sub-cases:
1261 * (a) if "show_other_directories" is true, we show it as
1262 * just a directory, unless "hide_empty_directories" is
1263 * also true, in which case we need to check if it contains any
1264 * untracked and / or ignored files.
1265 * (b) if it looks like a git directory, and we don't have
1266 * 'no_gitlinks' set we treat it as a gitlink, and show it
1267 * as a directory.
1268 * (c) otherwise, we recurse into it.
1270 static enum path_treatment treat_directory(struct dir_struct *dir,
1271 struct untracked_cache_dir *untracked,
1272 const char *dirname, int len, int baselen, int exclude,
1273 const struct path_simplify *simplify)
1275 /* The "len-1" is to strip the final '/' */
1276 switch (directory_exists_in_index(dirname, len-1)) {
1277 case index_directory:
1278 return path_recurse;
1280 case index_gitdir:
1281 return path_none;
1283 case index_nonexistent:
1284 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1285 break;
1286 if (!(dir->flags & DIR_NO_GITLINKS)) {
1287 unsigned char sha1[20];
1288 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1289 return path_untracked;
1291 return path_recurse;
1294 /* This is the "show_other_directories" case */
1296 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1297 return exclude ? path_excluded : path_untracked;
1299 untracked = lookup_untracked(dir->untracked, untracked,
1300 dirname + baselen, len - baselen);
1301 return read_directory_recursive(dir, dirname, len,
1302 untracked, 1, simplify);
1306 * This is an inexact early pruning of any recursive directory
1307 * reading - if the path cannot possibly be in the pathspec,
1308 * return true, and we'll skip it early.
1310 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1312 if (simplify) {
1313 for (;;) {
1314 const char *match = simplify->path;
1315 int len = simplify->len;
1317 if (!match)
1318 break;
1319 if (len > pathlen)
1320 len = pathlen;
1321 if (!memcmp(path, match, len))
1322 return 0;
1323 simplify++;
1325 return 1;
1327 return 0;
1331 * This function tells us whether an excluded path matches a
1332 * list of "interesting" pathspecs. That is, whether a path matched
1333 * by any of the pathspecs could possibly be ignored by excluding
1334 * the specified path. This can happen if:
1336 * 1. the path is mentioned explicitly in the pathspec
1338 * 2. the path is a directory prefix of some element in the
1339 * pathspec
1341 static int exclude_matches_pathspec(const char *path, int len,
1342 const struct path_simplify *simplify)
1344 if (simplify) {
1345 for (; simplify->path; simplify++) {
1346 if (len == simplify->len
1347 && !memcmp(path, simplify->path, len))
1348 return 1;
1349 if (len < simplify->len
1350 && simplify->path[len] == '/'
1351 && !memcmp(path, simplify->path, len))
1352 return 1;
1355 return 0;
1358 static int get_index_dtype(const char *path, int len)
1360 int pos;
1361 const struct cache_entry *ce;
1363 ce = cache_file_exists(path, len, 0);
1364 if (ce) {
1365 if (!ce_uptodate(ce))
1366 return DT_UNKNOWN;
1367 if (S_ISGITLINK(ce->ce_mode))
1368 return DT_DIR;
1370 * Nobody actually cares about the
1371 * difference between DT_LNK and DT_REG
1373 return DT_REG;
1376 /* Try to look it up as a directory */
1377 pos = cache_name_pos(path, len);
1378 if (pos >= 0)
1379 return DT_UNKNOWN;
1380 pos = -pos-1;
1381 while (pos < active_nr) {
1382 ce = active_cache[pos++];
1383 if (strncmp(ce->name, path, len))
1384 break;
1385 if (ce->name[len] > '/')
1386 break;
1387 if (ce->name[len] < '/')
1388 continue;
1389 if (!ce_uptodate(ce))
1390 break; /* continue? */
1391 return DT_DIR;
1393 return DT_UNKNOWN;
1396 static int get_dtype(struct dirent *de, const char *path, int len)
1398 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1399 struct stat st;
1401 if (dtype != DT_UNKNOWN)
1402 return dtype;
1403 dtype = get_index_dtype(path, len);
1404 if (dtype != DT_UNKNOWN)
1405 return dtype;
1406 if (lstat(path, &st))
1407 return dtype;
1408 if (S_ISREG(st.st_mode))
1409 return DT_REG;
1410 if (S_ISDIR(st.st_mode))
1411 return DT_DIR;
1412 if (S_ISLNK(st.st_mode))
1413 return DT_LNK;
1414 return dtype;
1417 static enum path_treatment treat_one_path(struct dir_struct *dir,
1418 struct untracked_cache_dir *untracked,
1419 struct strbuf *path,
1420 int baselen,
1421 const struct path_simplify *simplify,
1422 int dtype, struct dirent *de)
1424 int exclude;
1425 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1427 if (dtype == DT_UNKNOWN)
1428 dtype = get_dtype(de, path->buf, path->len);
1430 /* Always exclude indexed files */
1431 if (dtype != DT_DIR && has_path_in_index)
1432 return path_none;
1435 * When we are looking at a directory P in the working tree,
1436 * there are three cases:
1438 * (1) P exists in the index. Everything inside the directory P in
1439 * the working tree needs to go when P is checked out from the
1440 * index.
1442 * (2) P does not exist in the index, but there is P/Q in the index.
1443 * We know P will stay a directory when we check out the contents
1444 * of the index, but we do not know yet if there is a directory
1445 * P/Q in the working tree to be killed, so we need to recurse.
1447 * (3) P does not exist in the index, and there is no P/Q in the index
1448 * to require P to be a directory, either. Only in this case, we
1449 * know that everything inside P will not be killed without
1450 * recursing.
1452 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1453 (dtype == DT_DIR) &&
1454 !has_path_in_index &&
1455 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1456 return path_none;
1458 exclude = is_excluded(dir, path->buf, &dtype);
1461 * Excluded? If we don't explicitly want to show
1462 * ignored files, ignore it
1464 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1465 return path_excluded;
1467 switch (dtype) {
1468 default:
1469 return path_none;
1470 case DT_DIR:
1471 strbuf_addch(path, '/');
1472 return treat_directory(dir, untracked, path->buf, path->len,
1473 baselen, exclude, simplify);
1474 case DT_REG:
1475 case DT_LNK:
1476 return exclude ? path_excluded : path_untracked;
1480 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1481 struct untracked_cache_dir *untracked,
1482 struct cached_dir *cdir,
1483 struct strbuf *path,
1484 int baselen,
1485 const struct path_simplify *simplify)
1487 strbuf_setlen(path, baselen);
1488 if (!cdir->ucd) {
1489 strbuf_addstr(path, cdir->file);
1490 return path_untracked;
1492 strbuf_addstr(path, cdir->ucd->name);
1493 /* treat_one_path() does this before it calls treat_directory() */
1494 strbuf_complete(path, '/');
1495 if (cdir->ucd->check_only)
1497 * check_only is set as a result of treat_directory() getting
1498 * to its bottom. Verify again the same set of directories
1499 * with check_only set.
1501 return read_directory_recursive(dir, path->buf, path->len,
1502 cdir->ucd, 1, simplify);
1504 * We get path_recurse in the first run when
1505 * directory_exists_in_index() returns index_nonexistent. We
1506 * are sure that new changes in the index does not impact the
1507 * outcome. Return now.
1509 return path_recurse;
1512 static enum path_treatment treat_path(struct dir_struct *dir,
1513 struct untracked_cache_dir *untracked,
1514 struct cached_dir *cdir,
1515 struct strbuf *path,
1516 int baselen,
1517 const struct path_simplify *simplify)
1519 int dtype;
1520 struct dirent *de = cdir->de;
1522 if (!de)
1523 return treat_path_fast(dir, untracked, cdir, path,
1524 baselen, simplify);
1525 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1526 return path_none;
1527 strbuf_setlen(path, baselen);
1528 strbuf_addstr(path, de->d_name);
1529 if (simplify_away(path->buf, path->len, simplify))
1530 return path_none;
1532 dtype = DTYPE(de);
1533 return treat_one_path(dir, untracked, path, baselen, simplify, dtype, de);
1536 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1538 if (!dir)
1539 return;
1540 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1541 dir->untracked_alloc);
1542 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1545 static int valid_cached_dir(struct dir_struct *dir,
1546 struct untracked_cache_dir *untracked,
1547 struct strbuf *path,
1548 int check_only)
1550 struct stat st;
1552 if (!untracked)
1553 return 0;
1555 if (stat(path->len ? path->buf : ".", &st)) {
1556 invalidate_directory(dir->untracked, untracked);
1557 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1558 return 0;
1560 if (!untracked->valid ||
1561 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
1562 if (untracked->valid)
1563 invalidate_directory(dir->untracked, untracked);
1564 fill_stat_data(&untracked->stat_data, &st);
1565 return 0;
1568 if (untracked->check_only != !!check_only) {
1569 invalidate_directory(dir->untracked, untracked);
1570 return 0;
1574 * prep_exclude will be called eventually on this directory,
1575 * but it's called much later in last_exclude_matching(). We
1576 * need it now to determine the validity of the cache for this
1577 * path. The next calls will be nearly no-op, the way
1578 * prep_exclude() is designed.
1580 if (path->len && path->buf[path->len - 1] != '/') {
1581 strbuf_addch(path, '/');
1582 prep_exclude(dir, path->buf, path->len);
1583 strbuf_setlen(path, path->len - 1);
1584 } else
1585 prep_exclude(dir, path->buf, path->len);
1587 /* hopefully prep_exclude() haven't invalidated this entry... */
1588 return untracked->valid;
1591 static int open_cached_dir(struct cached_dir *cdir,
1592 struct dir_struct *dir,
1593 struct untracked_cache_dir *untracked,
1594 struct strbuf *path,
1595 int check_only)
1597 memset(cdir, 0, sizeof(*cdir));
1598 cdir->untracked = untracked;
1599 if (valid_cached_dir(dir, untracked, path, check_only))
1600 return 0;
1601 cdir->fdir = opendir(path->len ? path->buf : ".");
1602 if (dir->untracked)
1603 dir->untracked->dir_opened++;
1604 if (!cdir->fdir)
1605 return -1;
1606 return 0;
1609 static int read_cached_dir(struct cached_dir *cdir)
1611 if (cdir->fdir) {
1612 cdir->de = readdir(cdir->fdir);
1613 if (!cdir->de)
1614 return -1;
1615 return 0;
1617 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1618 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1619 if (!d->recurse) {
1620 cdir->nr_dirs++;
1621 continue;
1623 cdir->ucd = d;
1624 cdir->nr_dirs++;
1625 return 0;
1627 cdir->ucd = NULL;
1628 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1629 struct untracked_cache_dir *d = cdir->untracked;
1630 cdir->file = d->untracked[cdir->nr_files++];
1631 return 0;
1633 return -1;
1636 static void close_cached_dir(struct cached_dir *cdir)
1638 if (cdir->fdir)
1639 closedir(cdir->fdir);
1641 * We have gone through this directory and found no untracked
1642 * entries. Mark it valid.
1644 if (cdir->untracked) {
1645 cdir->untracked->valid = 1;
1646 cdir->untracked->recurse = 1;
1651 * Read a directory tree. We currently ignore anything but
1652 * directories, regular files and symlinks. That's because git
1653 * doesn't handle them at all yet. Maybe that will change some
1654 * day.
1656 * Also, we ignore the name ".git" (even if it is not a directory).
1657 * That likely will not change.
1659 * Returns the most significant path_treatment value encountered in the scan.
1661 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1662 const char *base, int baselen,
1663 struct untracked_cache_dir *untracked, int check_only,
1664 const struct path_simplify *simplify)
1666 struct cached_dir cdir;
1667 enum path_treatment state, subdir_state, dir_state = path_none;
1668 struct strbuf path = STRBUF_INIT;
1670 strbuf_add(&path, base, baselen);
1672 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1673 goto out;
1675 if (untracked)
1676 untracked->check_only = !!check_only;
1678 while (!read_cached_dir(&cdir)) {
1679 /* check how the file or directory should be treated */
1680 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);
1682 if (state > dir_state)
1683 dir_state = state;
1685 /* recurse into subdir if instructed by treat_path */
1686 if (state == path_recurse) {
1687 struct untracked_cache_dir *ud;
1688 ud = lookup_untracked(dir->untracked, untracked,
1689 path.buf + baselen,
1690 path.len - baselen);
1691 subdir_state =
1692 read_directory_recursive(dir, path.buf, path.len,
1693 ud, check_only, simplify);
1694 if (subdir_state > dir_state)
1695 dir_state = subdir_state;
1698 if (check_only) {
1699 /* abort early if maximum state has been reached */
1700 if (dir_state == path_untracked) {
1701 if (cdir.fdir)
1702 add_untracked(untracked, path.buf + baselen);
1703 break;
1705 /* skip the dir_add_* part */
1706 continue;
1709 /* add the path to the appropriate result list */
1710 switch (state) {
1711 case path_excluded:
1712 if (dir->flags & DIR_SHOW_IGNORED)
1713 dir_add_name(dir, path.buf, path.len);
1714 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1715 ((dir->flags & DIR_COLLECT_IGNORED) &&
1716 exclude_matches_pathspec(path.buf, path.len,
1717 simplify)))
1718 dir_add_ignored(dir, path.buf, path.len);
1719 break;
1721 case path_untracked:
1722 if (dir->flags & DIR_SHOW_IGNORED)
1723 break;
1724 dir_add_name(dir, path.buf, path.len);
1725 if (cdir.fdir)
1726 add_untracked(untracked, path.buf + baselen);
1727 break;
1729 default:
1730 break;
1733 close_cached_dir(&cdir);
1734 out:
1735 strbuf_release(&path);
1737 return dir_state;
1740 static int cmp_name(const void *p1, const void *p2)
1742 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1743 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1745 return name_compare(e1->name, e1->len, e2->name, e2->len);
1748 static struct path_simplify *create_simplify(const char **pathspec)
1750 int nr, alloc = 0;
1751 struct path_simplify *simplify = NULL;
1753 if (!pathspec)
1754 return NULL;
1756 for (nr = 0 ; ; nr++) {
1757 const char *match;
1758 ALLOC_GROW(simplify, nr + 1, alloc);
1759 match = *pathspec++;
1760 if (!match)
1761 break;
1762 simplify[nr].path = match;
1763 simplify[nr].len = simple_length(match);
1765 simplify[nr].path = NULL;
1766 simplify[nr].len = 0;
1767 return simplify;
1770 static void free_simplify(struct path_simplify *simplify)
1772 free(simplify);
1775 static int treat_leading_path(struct dir_struct *dir,
1776 const char *path, int len,
1777 const struct path_simplify *simplify)
1779 struct strbuf sb = STRBUF_INIT;
1780 int baselen, rc = 0;
1781 const char *cp;
1782 int old_flags = dir->flags;
1784 while (len && path[len - 1] == '/')
1785 len--;
1786 if (!len)
1787 return 1;
1788 baselen = 0;
1789 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1790 while (1) {
1791 cp = path + baselen + !!baselen;
1792 cp = memchr(cp, '/', path + len - cp);
1793 if (!cp)
1794 baselen = len;
1795 else
1796 baselen = cp - path;
1797 strbuf_setlen(&sb, 0);
1798 strbuf_add(&sb, path, baselen);
1799 if (!is_directory(sb.buf))
1800 break;
1801 if (simplify_away(sb.buf, sb.len, simplify))
1802 break;
1803 if (treat_one_path(dir, NULL, &sb, baselen, simplify,
1804 DT_DIR, NULL) == path_none)
1805 break; /* do not recurse into it */
1806 if (len <= baselen) {
1807 rc = 1;
1808 break; /* finished checking */
1811 strbuf_release(&sb);
1812 dir->flags = old_flags;
1813 return rc;
1816 static const char *get_ident_string(void)
1818 static struct strbuf sb = STRBUF_INIT;
1819 struct utsname uts;
1821 if (sb.len)
1822 return sb.buf;
1823 if (uname(&uts) < 0)
1824 die_errno(_("failed to get kernel name and information"));
1825 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
1826 uts.sysname);
1827 return sb.buf;
1830 static int ident_in_untracked(const struct untracked_cache *uc)
1833 * Previous git versions may have saved many NUL separated
1834 * strings in the "ident" field, but it is insane to manage
1835 * many locations, so just take care of the first one.
1838 return !strcmp(uc->ident.buf, get_ident_string());
1841 static void set_untracked_ident(struct untracked_cache *uc)
1843 strbuf_reset(&uc->ident);
1844 strbuf_addstr(&uc->ident, get_ident_string());
1847 * This strbuf used to contain a list of NUL separated
1848 * strings, so save NUL too for backward compatibility.
1850 strbuf_addch(&uc->ident, 0);
1853 static void new_untracked_cache(struct index_state *istate)
1855 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
1856 strbuf_init(&uc->ident, 100);
1857 uc->exclude_per_dir = ".gitignore";
1858 /* should be the same flags used by git-status */
1859 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
1860 set_untracked_ident(uc);
1861 istate->untracked = uc;
1862 istate->cache_changed |= UNTRACKED_CHANGED;
1865 void add_untracked_cache(struct index_state *istate)
1867 if (!istate->untracked) {
1868 new_untracked_cache(istate);
1869 } else {
1870 if (!ident_in_untracked(istate->untracked)) {
1871 free_untracked_cache(istate->untracked);
1872 new_untracked_cache(istate);
1877 void remove_untracked_cache(struct index_state *istate)
1879 if (istate->untracked) {
1880 free_untracked_cache(istate->untracked);
1881 istate->untracked = NULL;
1882 istate->cache_changed |= UNTRACKED_CHANGED;
1886 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1887 int base_len,
1888 const struct pathspec *pathspec)
1890 struct untracked_cache_dir *root;
1892 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
1893 return NULL;
1896 * We only support $GIT_DIR/info/exclude and core.excludesfile
1897 * as the global ignore rule files. Any other additions
1898 * (e.g. from command line) invalidate the cache. This
1899 * condition also catches running setup_standard_excludes()
1900 * before setting dir->untracked!
1902 if (dir->unmanaged_exclude_files)
1903 return NULL;
1906 * Optimize for the main use case only: whole-tree git
1907 * status. More work involved in treat_leading_path() if we
1908 * use cache on just a subset of the worktree. pathspec
1909 * support could make the matter even worse.
1911 if (base_len || (pathspec && pathspec->nr))
1912 return NULL;
1914 /* Different set of flags may produce different results */
1915 if (dir->flags != dir->untracked->dir_flags ||
1917 * See treat_directory(), case index_nonexistent. Without
1918 * this flag, we may need to also cache .git file content
1919 * for the resolve_gitlink_ref() call, which we don't.
1921 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1922 /* We don't support collecting ignore files */
1923 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1924 DIR_COLLECT_IGNORED)))
1925 return NULL;
1928 * If we use .gitignore in the cache and now you change it to
1929 * .gitexclude, everything will go wrong.
1931 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
1932 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
1933 return NULL;
1936 * EXC_CMDL is not considered in the cache. If people set it,
1937 * skip the cache.
1939 if (dir->exclude_list_group[EXC_CMDL].nr)
1940 return NULL;
1942 if (!ident_in_untracked(dir->untracked)) {
1943 warning(_("Untracked cache is disabled on this system or location."));
1944 return NULL;
1947 if (!dir->untracked->root) {
1948 const int len = sizeof(*dir->untracked->root);
1949 dir->untracked->root = xmalloc(len);
1950 memset(dir->untracked->root, 0, len);
1953 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
1954 root = dir->untracked->root;
1955 if (hashcmp(dir->ss_info_exclude.sha1,
1956 dir->untracked->ss_info_exclude.sha1)) {
1957 invalidate_gitignore(dir->untracked, root);
1958 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
1960 if (hashcmp(dir->ss_excludes_file.sha1,
1961 dir->untracked->ss_excludes_file.sha1)) {
1962 invalidate_gitignore(dir->untracked, root);
1963 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
1966 /* Make sure this directory is not dropped out at saving phase */
1967 root->recurse = 1;
1968 return root;
1971 int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1973 struct path_simplify *simplify;
1974 struct untracked_cache_dir *untracked;
1977 * Check out create_simplify()
1979 if (pathspec)
1980 GUARD_PATHSPEC(pathspec,
1981 PATHSPEC_FROMTOP |
1982 PATHSPEC_MAXDEPTH |
1983 PATHSPEC_LITERAL |
1984 PATHSPEC_GLOB |
1985 PATHSPEC_ICASE |
1986 PATHSPEC_EXCLUDE);
1988 if (has_symlink_leading_path(path, len))
1989 return dir->nr;
1992 * exclude patterns are treated like positive ones in
1993 * create_simplify. Usually exclude patterns should be a
1994 * subset of positive ones, which has no impacts on
1995 * create_simplify().
1997 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
1998 untracked = validate_untracked_cache(dir, len, pathspec);
1999 if (!untracked)
2001 * make sure untracked cache code path is disabled,
2002 * e.g. prep_exclude()
2004 dir->untracked = NULL;
2005 if (!len || treat_leading_path(dir, path, len, simplify))
2006 read_directory_recursive(dir, path, len, untracked, 0, simplify);
2007 free_simplify(simplify);
2008 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
2009 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
2010 if (dir->untracked) {
2011 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2012 trace_printf_key(&trace_untracked_stats,
2013 "node creation: %u\n"
2014 "gitignore invalidation: %u\n"
2015 "directory invalidation: %u\n"
2016 "opendir: %u\n",
2017 dir->untracked->dir_created,
2018 dir->untracked->gitignore_invalidated,
2019 dir->untracked->dir_invalidated,
2020 dir->untracked->dir_opened);
2021 if (dir->untracked == the_index.untracked &&
2022 (dir->untracked->dir_opened ||
2023 dir->untracked->gitignore_invalidated ||
2024 dir->untracked->dir_invalidated))
2025 the_index.cache_changed |= UNTRACKED_CHANGED;
2026 if (dir->untracked != the_index.untracked) {
2027 free(dir->untracked);
2028 dir->untracked = NULL;
2031 return dir->nr;
2034 int file_exists(const char *f)
2036 struct stat sb;
2037 return lstat(f, &sb) == 0;
2040 static int cmp_icase(char a, char b)
2042 if (a == b)
2043 return 0;
2044 if (ignore_case)
2045 return toupper(a) - toupper(b);
2046 return a - b;
2050 * Given two normalized paths (a trailing slash is ok), if subdir is
2051 * outside dir, return -1. Otherwise return the offset in subdir that
2052 * can be used as relative path to dir.
2054 int dir_inside_of(const char *subdir, const char *dir)
2056 int offset = 0;
2058 assert(dir && subdir && *dir && *subdir);
2060 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2061 dir++;
2062 subdir++;
2063 offset++;
2066 /* hel[p]/me vs hel[l]/yeah */
2067 if (*dir && *subdir)
2068 return -1;
2070 if (!*subdir)
2071 return !*dir ? offset : -1; /* same dir */
2073 /* foo/[b]ar vs foo/[] */
2074 if (is_dir_sep(dir[-1]))
2075 return is_dir_sep(subdir[-1]) ? offset : -1;
2077 /* foo[/]bar vs foo[] */
2078 return is_dir_sep(*subdir) ? offset + 1 : -1;
2081 int is_inside_dir(const char *dir)
2083 char *cwd;
2084 int rc;
2086 if (!dir)
2087 return 0;
2089 cwd = xgetcwd();
2090 rc = (dir_inside_of(cwd, dir) >= 0);
2091 free(cwd);
2092 return rc;
2095 int is_empty_dir(const char *path)
2097 DIR *dir = opendir(path);
2098 struct dirent *e;
2099 int ret = 1;
2101 if (!dir)
2102 return 0;
2104 while ((e = readdir(dir)) != NULL)
2105 if (!is_dot_or_dotdot(e->d_name)) {
2106 ret = 0;
2107 break;
2110 closedir(dir);
2111 return ret;
2114 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2116 DIR *dir;
2117 struct dirent *e;
2118 int ret = 0, original_len = path->len, len, kept_down = 0;
2119 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2120 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2121 unsigned char submodule_head[20];
2123 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2124 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2125 /* Do not descend and nuke a nested git work tree. */
2126 if (kept_up)
2127 *kept_up = 1;
2128 return 0;
2131 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2132 dir = opendir(path->buf);
2133 if (!dir) {
2134 if (errno == ENOENT)
2135 return keep_toplevel ? -1 : 0;
2136 else if (errno == EACCES && !keep_toplevel)
2138 * An empty dir could be removable even if it
2139 * is unreadable:
2141 return rmdir(path->buf);
2142 else
2143 return -1;
2145 strbuf_complete(path, '/');
2147 len = path->len;
2148 while ((e = readdir(dir)) != NULL) {
2149 struct stat st;
2150 if (is_dot_or_dotdot(e->d_name))
2151 continue;
2153 strbuf_setlen(path, len);
2154 strbuf_addstr(path, e->d_name);
2155 if (lstat(path->buf, &st)) {
2156 if (errno == ENOENT)
2158 * file disappeared, which is what we
2159 * wanted anyway
2161 continue;
2162 /* fall thru */
2163 } else if (S_ISDIR(st.st_mode)) {
2164 if (!remove_dir_recurse(path, flag, &kept_down))
2165 continue; /* happy */
2166 } else if (!only_empty &&
2167 (!unlink(path->buf) || errno == ENOENT)) {
2168 continue; /* happy, too */
2171 /* path too long, stat fails, or non-directory still exists */
2172 ret = -1;
2173 break;
2175 closedir(dir);
2177 strbuf_setlen(path, original_len);
2178 if (!ret && !keep_toplevel && !kept_down)
2179 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2180 else if (kept_up)
2182 * report the uplevel that it is not an error that we
2183 * did not rmdir() our directory.
2185 *kept_up = !ret;
2186 return ret;
2189 int remove_dir_recursively(struct strbuf *path, int flag)
2191 return remove_dir_recurse(path, flag, NULL);
2194 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2196 void setup_standard_excludes(struct dir_struct *dir)
2198 const char *path;
2200 dir->exclude_per_dir = ".gitignore";
2202 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2203 if (!excludes_file)
2204 excludes_file = xdg_config_home("ignore");
2205 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2206 add_excludes_from_file_1(dir, excludes_file,
2207 dir->untracked ? &dir->ss_excludes_file : NULL);
2209 /* per repository user preference */
2210 path = git_path_info_exclude();
2211 if (!access_or_warn(path, R_OK, 0))
2212 add_excludes_from_file_1(dir, path,
2213 dir->untracked ? &dir->ss_info_exclude : NULL);
2216 int remove_path(const char *name)
2218 char *slash;
2220 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
2221 return -1;
2223 slash = strrchr(name, '/');
2224 if (slash) {
2225 char *dirs = xstrdup(name);
2226 slash = dirs + (slash - name);
2227 do {
2228 *slash = '\0';
2229 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2230 free(dirs);
2232 return 0;
2236 * Frees memory within dir which was allocated for exclude lists and
2237 * the exclude_stack. Does not free dir itself.
2239 void clear_directory(struct dir_struct *dir)
2241 int i, j;
2242 struct exclude_list_group *group;
2243 struct exclude_list *el;
2244 struct exclude_stack *stk;
2246 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2247 group = &dir->exclude_list_group[i];
2248 for (j = 0; j < group->nr; j++) {
2249 el = &group->el[j];
2250 if (i == EXC_DIRS)
2251 free((char *)el->src);
2252 clear_exclude_list(el);
2254 free(group->el);
2257 stk = dir->exclude_stack;
2258 while (stk) {
2259 struct exclude_stack *prev = stk->prev;
2260 free(stk);
2261 stk = prev;
2263 strbuf_release(&dir->basebuf);
2266 struct ondisk_untracked_cache {
2267 struct stat_data info_exclude_stat;
2268 struct stat_data excludes_file_stat;
2269 uint32_t dir_flags;
2270 unsigned char info_exclude_sha1[20];
2271 unsigned char excludes_file_sha1[20];
2272 char exclude_per_dir[FLEX_ARRAY];
2275 #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2277 struct write_data {
2278 int index; /* number of written untracked_cache_dir */
2279 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2280 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2281 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2282 struct strbuf out;
2283 struct strbuf sb_stat;
2284 struct strbuf sb_sha1;
2287 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2289 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2290 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2291 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2292 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2293 to->sd_dev = htonl(from->sd_dev);
2294 to->sd_ino = htonl(from->sd_ino);
2295 to->sd_uid = htonl(from->sd_uid);
2296 to->sd_gid = htonl(from->sd_gid);
2297 to->sd_size = htonl(from->sd_size);
2300 static void write_one_dir(struct untracked_cache_dir *untracked,
2301 struct write_data *wd)
2303 struct stat_data stat_data;
2304 struct strbuf *out = &wd->out;
2305 unsigned char intbuf[16];
2306 unsigned int intlen, value;
2307 int i = wd->index++;
2310 * untracked_nr should be reset whenever valid is clear, but
2311 * for safety..
2313 if (!untracked->valid) {
2314 untracked->untracked_nr = 0;
2315 untracked->check_only = 0;
2318 if (untracked->check_only)
2319 ewah_set(wd->check_only, i);
2320 if (untracked->valid) {
2321 ewah_set(wd->valid, i);
2322 stat_data_to_disk(&stat_data, &untracked->stat_data);
2323 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2325 if (!is_null_sha1(untracked->exclude_sha1)) {
2326 ewah_set(wd->sha1_valid, i);
2327 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2330 intlen = encode_varint(untracked->untracked_nr, intbuf);
2331 strbuf_add(out, intbuf, intlen);
2333 /* skip non-recurse directories */
2334 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2335 if (untracked->dirs[i]->recurse)
2336 value++;
2337 intlen = encode_varint(value, intbuf);
2338 strbuf_add(out, intbuf, intlen);
2340 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2342 for (i = 0; i < untracked->untracked_nr; i++)
2343 strbuf_add(out, untracked->untracked[i],
2344 strlen(untracked->untracked[i]) + 1);
2346 for (i = 0; i < untracked->dirs_nr; i++)
2347 if (untracked->dirs[i]->recurse)
2348 write_one_dir(untracked->dirs[i], wd);
2351 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2353 struct ondisk_untracked_cache *ouc;
2354 struct write_data wd;
2355 unsigned char varbuf[16];
2356 int varint_len;
2357 size_t len = strlen(untracked->exclude_per_dir);
2359 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
2360 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2361 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2362 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2363 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2364 ouc->dir_flags = htonl(untracked->dir_flags);
2366 varint_len = encode_varint(untracked->ident.len, varbuf);
2367 strbuf_add(out, varbuf, varint_len);
2368 strbuf_add(out, untracked->ident.buf, untracked->ident.len);
2370 strbuf_add(out, ouc, ouc_size(len));
2371 free(ouc);
2372 ouc = NULL;
2374 if (!untracked->root) {
2375 varint_len = encode_varint(0, varbuf);
2376 strbuf_add(out, varbuf, varint_len);
2377 return;
2380 wd.index = 0;
2381 wd.check_only = ewah_new();
2382 wd.valid = ewah_new();
2383 wd.sha1_valid = ewah_new();
2384 strbuf_init(&wd.out, 1024);
2385 strbuf_init(&wd.sb_stat, 1024);
2386 strbuf_init(&wd.sb_sha1, 1024);
2387 write_one_dir(untracked->root, &wd);
2389 varint_len = encode_varint(wd.index, varbuf);
2390 strbuf_add(out, varbuf, varint_len);
2391 strbuf_addbuf(out, &wd.out);
2392 ewah_serialize_strbuf(wd.valid, out);
2393 ewah_serialize_strbuf(wd.check_only, out);
2394 ewah_serialize_strbuf(wd.sha1_valid, out);
2395 strbuf_addbuf(out, &wd.sb_stat);
2396 strbuf_addbuf(out, &wd.sb_sha1);
2397 strbuf_addch(out, '\0'); /* safe guard for string lists */
2399 ewah_free(wd.valid);
2400 ewah_free(wd.check_only);
2401 ewah_free(wd.sha1_valid);
2402 strbuf_release(&wd.out);
2403 strbuf_release(&wd.sb_stat);
2404 strbuf_release(&wd.sb_sha1);
2407 static void free_untracked(struct untracked_cache_dir *ucd)
2409 int i;
2410 if (!ucd)
2411 return;
2412 for (i = 0; i < ucd->dirs_nr; i++)
2413 free_untracked(ucd->dirs[i]);
2414 for (i = 0; i < ucd->untracked_nr; i++)
2415 free(ucd->untracked[i]);
2416 free(ucd->untracked);
2417 free(ucd->dirs);
2418 free(ucd);
2421 void free_untracked_cache(struct untracked_cache *uc)
2423 if (uc)
2424 free_untracked(uc->root);
2425 free(uc);
2428 struct read_data {
2429 int index;
2430 struct untracked_cache_dir **ucd;
2431 struct ewah_bitmap *check_only;
2432 struct ewah_bitmap *valid;
2433 struct ewah_bitmap *sha1_valid;
2434 const unsigned char *data;
2435 const unsigned char *end;
2438 static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2440 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2441 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2442 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2443 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2444 to->sd_dev = get_be32(&from->sd_dev);
2445 to->sd_ino = get_be32(&from->sd_ino);
2446 to->sd_uid = get_be32(&from->sd_uid);
2447 to->sd_gid = get_be32(&from->sd_gid);
2448 to->sd_size = get_be32(&from->sd_size);
2451 static int read_one_dir(struct untracked_cache_dir **untracked_,
2452 struct read_data *rd)
2454 struct untracked_cache_dir ud, *untracked;
2455 const unsigned char *next, *data = rd->data, *end = rd->end;
2456 unsigned int value;
2457 int i, len;
2459 memset(&ud, 0, sizeof(ud));
2461 next = data;
2462 value = decode_varint(&next);
2463 if (next > end)
2464 return -1;
2465 ud.recurse = 1;
2466 ud.untracked_alloc = value;
2467 ud.untracked_nr = value;
2468 if (ud.untracked_nr)
2469 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2470 data = next;
2472 next = data;
2473 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2474 if (next > end)
2475 return -1;
2476 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2477 data = next;
2479 len = strlen((const char *)data);
2480 next = data + len + 1;
2481 if (next > rd->end)
2482 return -1;
2483 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
2484 memcpy(untracked, &ud, sizeof(ud));
2485 memcpy(untracked->name, data, len + 1);
2486 data = next;
2488 for (i = 0; i < untracked->untracked_nr; i++) {
2489 len = strlen((const char *)data);
2490 next = data + len + 1;
2491 if (next > rd->end)
2492 return -1;
2493 untracked->untracked[i] = xstrdup((const char*)data);
2494 data = next;
2497 rd->ucd[rd->index++] = untracked;
2498 rd->data = data;
2500 for (i = 0; i < untracked->dirs_nr; i++) {
2501 len = read_one_dir(untracked->dirs + i, rd);
2502 if (len < 0)
2503 return -1;
2505 return 0;
2508 static void set_check_only(size_t pos, void *cb)
2510 struct read_data *rd = cb;
2511 struct untracked_cache_dir *ud = rd->ucd[pos];
2512 ud->check_only = 1;
2515 static void read_stat(size_t pos, void *cb)
2517 struct read_data *rd = cb;
2518 struct untracked_cache_dir *ud = rd->ucd[pos];
2519 if (rd->data + sizeof(struct stat_data) > rd->end) {
2520 rd->data = rd->end + 1;
2521 return;
2523 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2524 rd->data += sizeof(struct stat_data);
2525 ud->valid = 1;
2528 static void read_sha1(size_t pos, void *cb)
2530 struct read_data *rd = cb;
2531 struct untracked_cache_dir *ud = rd->ucd[pos];
2532 if (rd->data + 20 > rd->end) {
2533 rd->data = rd->end + 1;
2534 return;
2536 hashcpy(ud->exclude_sha1, rd->data);
2537 rd->data += 20;
2540 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2541 const struct stat_data *stat,
2542 const unsigned char *sha1)
2544 stat_data_from_disk(&sha1_stat->stat, stat);
2545 hashcpy(sha1_stat->sha1, sha1);
2546 sha1_stat->valid = 1;
2549 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2551 const struct ondisk_untracked_cache *ouc;
2552 struct untracked_cache *uc;
2553 struct read_data rd;
2554 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2555 const char *ident;
2556 int ident_len, len;
2558 if (sz <= 1 || end[-1] != '\0')
2559 return NULL;
2560 end--;
2562 ident_len = decode_varint(&next);
2563 if (next + ident_len > end)
2564 return NULL;
2565 ident = (const char *)next;
2566 next += ident_len;
2568 ouc = (const struct ondisk_untracked_cache *)next;
2569 if (next + ouc_size(0) > end)
2570 return NULL;
2572 uc = xcalloc(1, sizeof(*uc));
2573 strbuf_init(&uc->ident, ident_len);
2574 strbuf_add(&uc->ident, ident, ident_len);
2575 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2576 ouc->info_exclude_sha1);
2577 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2578 ouc->excludes_file_sha1);
2579 uc->dir_flags = get_be32(&ouc->dir_flags);
2580 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2581 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2582 next += ouc_size(strlen(ouc->exclude_per_dir));
2583 if (next >= end)
2584 goto done2;
2586 len = decode_varint(&next);
2587 if (next > end || len == 0)
2588 goto done2;
2590 rd.valid = ewah_new();
2591 rd.check_only = ewah_new();
2592 rd.sha1_valid = ewah_new();
2593 rd.data = next;
2594 rd.end = end;
2595 rd.index = 0;
2596 ALLOC_ARRAY(rd.ucd, len);
2598 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2599 goto done;
2601 next = rd.data;
2602 len = ewah_read_mmap(rd.valid, next, end - next);
2603 if (len < 0)
2604 goto done;
2606 next += len;
2607 len = ewah_read_mmap(rd.check_only, next, end - next);
2608 if (len < 0)
2609 goto done;
2611 next += len;
2612 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2613 if (len < 0)
2614 goto done;
2616 ewah_each_bit(rd.check_only, set_check_only, &rd);
2617 rd.data = next + len;
2618 ewah_each_bit(rd.valid, read_stat, &rd);
2619 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2620 next = rd.data;
2622 done:
2623 free(rd.ucd);
2624 ewah_free(rd.valid);
2625 ewah_free(rd.check_only);
2626 ewah_free(rd.sha1_valid);
2627 done2:
2628 if (next != end) {
2629 free_untracked_cache(uc);
2630 uc = NULL;
2632 return uc;
2635 static void invalidate_one_directory(struct untracked_cache *uc,
2636 struct untracked_cache_dir *ucd)
2638 uc->dir_invalidated++;
2639 ucd->valid = 0;
2640 ucd->untracked_nr = 0;
2644 * Normally when an entry is added or removed from a directory,
2645 * invalidating that directory is enough. No need to touch its
2646 * ancestors. When a directory is shown as "foo/bar/" in git-status
2647 * however, deleting or adding an entry may have cascading effect.
2649 * Say the "foo/bar/file" has become untracked, we need to tell the
2650 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2651 * directory any more (because "bar" is managed by foo as an untracked
2652 * "file").
2654 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2655 * was the last untracked entry in the entire "foo", we should show
2656 * "foo/" instead. Which means we have to invalidate past "bar" up to
2657 * "foo".
2659 * This function traverses all directories from root to leaf. If there
2660 * is a chance of one of the above cases happening, we invalidate back
2661 * to root. Otherwise we just invalidate the leaf. There may be a more
2662 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2663 * detect these cases and avoid unnecessary invalidation, for example,
2664 * checking for the untracked entry named "bar/" in "foo", but for now
2665 * stick to something safe and simple.
2667 static int invalidate_one_component(struct untracked_cache *uc,
2668 struct untracked_cache_dir *dir,
2669 const char *path, int len)
2671 const char *rest = strchr(path, '/');
2673 if (rest) {
2674 int component_len = rest - path;
2675 struct untracked_cache_dir *d =
2676 lookup_untracked(uc, dir, path, component_len);
2677 int ret =
2678 invalidate_one_component(uc, d, rest + 1,
2679 len - (component_len + 1));
2680 if (ret)
2681 invalidate_one_directory(uc, dir);
2682 return ret;
2685 invalidate_one_directory(uc, dir);
2686 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2689 void untracked_cache_invalidate_path(struct index_state *istate,
2690 const char *path)
2692 if (!istate->untracked || !istate->untracked->root)
2693 return;
2694 invalidate_one_component(istate->untracked, istate->untracked->root,
2695 path, strlen(path));
2698 void untracked_cache_remove_from_index(struct index_state *istate,
2699 const char *path)
2701 untracked_cache_invalidate_path(istate, path);
2704 void untracked_cache_add_to_index(struct index_state *istate,
2705 const char *path)
2707 untracked_cache_invalidate_path(istate, path);