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
13 #include "wildmatch.h"
17 #include "ewah/ewok.h"
19 struct path_simplify {
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.
38 * Support data structure for our opendir/readdir/closedir wrappers
42 struct untracked_cache_dir *untracked;
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),
74 int git_fnmatch(const struct pathspec_item *item,
75 const char *pattern, const char *string,
79 if (ps_strncmp(item, pattern, 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,
94 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
97 /* wildmatch has not learned no FNM_PATHNAME mode yet */
98 return wildmatch(pattern, string,
99 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
103 static int fnmatch_icase_mem(const char *pattern, int patternlen,
104 const char *string, int stringlen,
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;
123 flags |= WM_CASEFOLD;
124 match_status = wildmatch(use_pat, use_str, flags, NULL);
126 strbuf_release(&pat_buf);
127 strbuf_release(&str_buf);
132 static size_t common_prefix_len(const struct pathspec *pathspec)
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
144 GUARD_PATHSPEC(pathspec,
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)
156 if (pathspec->items[n].magic & PATHSPEC_ICASE)
157 item_len = pathspec->items[n].prefix;
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])
168 if (n == 0 || len < max) {
178 * Returns a copy of the longest leading path common among all
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)
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);
203 int within_depth(const char *name, int namelen,
204 int depth, int max_depth)
206 const char *cp = name, *cpe = name + namelen;
212 if (depth > max_depth)
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
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))
273 /* If the match was just the prefix, we matched */
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;
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
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,
317 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
328 if (!ps->recursive ||
329 !(ps->magic & PATHSPEC_MAXDEPTH) ||
331 return MATCHED_RECURSIVELY;
333 if (within_depth(name, namelen, 0, ps->max_depth))
334 return MATCHED_EXACTLY;
342 for (i = ps->nr - 1; i >= 0; i--) {
345 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
346 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
349 if (seen && seen[i] == MATCHED_EXACTLY)
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,
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] == '/')
366 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
367 how = MATCHED_EXACTLY;
374 if (seen && seen[i] < how)
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)
391 negative = do_match_pathspec(ps, name, namelen,
393 flags | DO_MATCH_EXCLUDE);
394 return negative ? 0 : positive;
397 int report_path_error(const char *ps_matched,
398 const struct pathspec *pathspec,
402 * Make sure all pathspec matched; otherwise it is an error.
405 for (num = 0; num < pathspec->nr; num++) {
406 int other, found_dup;
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;
419 if (other == num || !ps_matched[other])
421 if (!strcmp(pathspec->items[other].original,
422 pathspec->items[num].original))
424 * Ok, we have a match already.
431 error("pathspec '%s' did not match any file(s) known to git.",
432 pathspec->items[num].original);
439 * Return the length of the "simple" part of a path match limiter.
441 int simple_length(const char *match)
446 unsigned char c = *match++;
448 if (c == '\0' || is_glob_special(c))
453 int no_wildcard(const char *string)
455 return string[simple_length(string)] == '\0';
458 void parse_exclude_pattern(const char **pattern,
463 const char *p = *pattern;
468 *flags |= EXC_FLAG_NEGATIVE;
472 if (len && p[len - 1] == '/') {
474 *flags |= EXC_FLAG_MUSTBEDIR;
476 for (i = 0; i < len; i++) {
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;
496 void add_exclude(const char *string, const char *base,
497 int baselen, struct exclude_list *el, int srcpos)
504 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
505 if (flags & EXC_FLAG_MUSTBEDIR) {
507 x = xmalloc(sizeof(*x) + patternlen + 1);
509 memcpy(s, string, patternlen);
510 s[patternlen] = '\0';
513 x = xmalloc(sizeof(*x));
516 x->patternlen = patternlen;
517 x->nowildcardlen = nowildcardlen;
519 x->baselen = baselen;
522 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
523 el->excludes[el->nr++] = x;
527 static void *read_skip_worktree_file_from_index(const char *path, size_t *size,
528 struct sha1_stat *sha1_stat)
532 enum object_type type;
536 pos = cache_name_pos(path, len);
539 if (!ce_skip_worktree(active_cache[pos]))
541 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
542 if (!data || type != OBJ_BLOB) {
548 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
549 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
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)
562 for (i = 0; i < el->nr; i++)
563 free(el->excludes[i]);
572 static void trim_trailing_spaces(char *buf)
574 char *p, *last_space = NULL;
576 for (p = buf; *p; p++)
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)
607 struct untracked_cache_dir *d;
610 if (len && name[len - 1] == '/')
614 while (last > first) {
615 int cmp, next = (last + first) >> 1;
617 cmp = strncmp(name, d->name, len);
618 if (!cmp && strlen(d->name) > len)
630 d = xmalloc(sizeof(*d) + len + 1);
631 memset(d, 0, sizeof(*d));
632 memcpy(d->name, name, len);
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));
639 dir->dirs[first] = d;
643 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
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)
663 uc->dir_invalidated++;
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)
684 int fd, i, lineno = 1;
688 fd = open(fname, O_RDONLY);
689 if (fd < 0 || fstat(fd, &st) < 0) {
691 warn_on_inaccessible(fname);
695 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL)
701 if (buf[size-1] != '\n') {
702 buf = xrealloc(buf, size+1);
706 size = xsize_t(st.st_size);
709 fill_stat_data(&sha1_stat->stat, &st);
710 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
711 sha1_stat->valid = 1;
716 buf = xmalloc(size+1);
717 if (read_in_full(fd, buf, size) != size) {
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);
736 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
737 fill_stat_data(&sha1_stat->stat, &st);
738 sha1_stat->valid = 1;
744 if (skip_utf8_bom(&buf, size))
745 size -= buf - el->filebuf;
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);
763 int add_excludes_from_file_to_list(const char *fname, const char *base,
764 int baselen, struct exclude_list *el,
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));
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.
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,
813 if (prefix == patternlen) {
814 if (patternlen == basenamelen &&
815 !strncmp_icase(pattern, basename, basenamelen))
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),
825 if (fnmatch_icase_mem(pattern, patternlen,
826 basename, basenamelen,
833 int match_pathname(const char *pathname, int pathlen,
834 const char *base, int baselen,
835 const char *pattern, int prefix, int patternlen,
842 * match with FNM_PATHNAME; the pattern has base implicitly
845 if (*pattern == '/') {
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))
860 namelen = baselen ? pathlen - baselen - 1 : pathlen;
861 name = pathname + pathlen - namelen;
865 * if the non-wildcard part is longer than the
866 * remaining pathname, surely it cannot match.
868 if (prefix > namelen)
871 if (strncmp_icase(pattern, name, prefix))
874 patternlen -= 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)
887 return fnmatch_icase_mem(pattern, patternlen,
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,
900 const char *basename,
902 struct exclude_list *el)
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)
921 if (x->flags & EXC_FLAG_NODIR) {
922 if (match_basename(basename,
923 pathlen - (basename - pathname),
924 exclude, prefix, x->patternlen,
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))
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);
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,
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,
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;
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))
997 el = &group->el[dir->exclude_stack->exclude_ix];
998 dir->exclude_stack = stk->prev;
1000 free((char *)el->src); /* see strbuf_detach() below */
1001 clear_exclude_list(el);
1006 /* Skip traversing into sub directories if the parent is excluded */
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);
1022 untracked = stk ? stk->ucd : dir->untracked->root;
1026 while (current < baselen) {
1028 struct sha1_stat sha1_stat;
1030 stk = xcalloc(1, sizeof(*stk));
1035 cp = strchr(base + current + 1, '/');
1037 die("oops in prep_exclude");
1040 lookup_untracked(dir->untracked, untracked,
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 */
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] = '/';
1061 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1062 dir->exclude = NULL;
1064 dir->exclude_stack = stk;
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 and skip_worktree is
1082 * not set). Then we can skip loading .gitignore,
1083 * which would result in ENOENT anyway.
1084 * skip_worktree is taken care in read_directory()
1086 !is_null_sha1(untracked->exclude_sha1))) {
1088 * dir->basebuf gets reused by the traversal, but we
1089 * need fname to remain unchanged to ensure the src
1090 * member of each struct exclude correctly
1091 * back-references its source file. Other invocations
1092 * of add_exclude_list provide stable strings, so we
1093 * strbuf_detach() and free() here in the caller.
1095 struct strbuf sb = STRBUF_INIT;
1096 strbuf_addbuf(&sb, &dir->basebuf);
1097 strbuf_addstr(&sb, dir->exclude_per_dir);
1098 el->src = strbuf_detach(&sb, NULL);
1099 add_excludes(el->src, el->src, stk->baselen, el, 1,
1100 untracked ? &sha1_stat : NULL);
1103 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1104 * will first be called in valid_cached_dir() then maybe many
1105 * times more in last_exclude_matching(). When the cache is
1106 * used, last_exclude_matching() will not be called and
1107 * reading .gitignore content will be a waste.
1109 * So when it's called by valid_cached_dir() and we can get
1110 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1111 * modified on work tree), we could delay reading the
1112 * .gitignore content until we absolutely need it in
1113 * last_exclude_matching(). Be careful about ignore rule
1114 * order, though, if you do that.
1117 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1118 invalidate_gitignore(dir->untracked, untracked);
1119 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1121 dir->exclude_stack = stk;
1122 current = stk->baselen;
1124 strbuf_setlen(&dir->basebuf, baselen);
1128 * Loads the exclude lists for the directory containing pathname, then
1129 * scans all exclude lists to determine whether pathname is excluded.
1130 * Returns the exclude_list element which matched, or NULL for
1133 struct exclude *last_exclude_matching(struct dir_struct *dir,
1134 const char *pathname,
1137 int pathlen = strlen(pathname);
1138 const char *basename = strrchr(pathname, '/');
1139 basename = (basename) ? basename+1 : pathname;
1141 prep_exclude(dir, pathname, basename-pathname);
1144 return dir->exclude;
1146 return last_exclude_matching_from_lists(dir, pathname, pathlen,
1151 * Loads the exclude lists for the directory containing pathname, then
1152 * scans all exclude lists to determine whether pathname is excluded.
1153 * Returns 1 if true, otherwise 0.
1155 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1157 struct exclude *exclude =
1158 last_exclude_matching(dir, pathname, dtype_p);
1160 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1164 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1166 struct dir_entry *ent;
1168 ent = xmalloc(sizeof(*ent) + len + 1);
1170 memcpy(ent->name, pathname, len);
1175 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1177 if (cache_file_exists(pathname, len, ignore_case))
1180 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1181 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1184 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1186 if (!cache_name_is_other(pathname, len))
1189 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1190 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1194 index_nonexistent = 0,
1200 * Do not use the alphabetically sorted index to look up
1201 * the directory name; instead, use the case insensitive
1204 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1206 const struct cache_entry *ce = cache_dir_exists(dirname, len);
1207 unsigned char endchar;
1210 return index_nonexistent;
1211 endchar = ce->name[len];
1214 * The cache_entry structure returned will contain this dirname
1215 * and possibly additional path components.
1218 return index_directory;
1221 * If there are no additional path components, then this cache_entry
1222 * represents a submodule. Submodules, despite being directories,
1223 * are stored in the cache without a closing slash.
1225 if (!endchar && S_ISGITLINK(ce->ce_mode))
1226 return index_gitdir;
1228 /* This should never be hit, but it exists just in case. */
1229 return index_nonexistent;
1233 * The index sorts alphabetically by entry name, which
1234 * means that a gitlink sorts as '\0' at the end, while
1235 * a directory (which is defined not as an entry, but as
1236 * the files it contains) will sort with the '/' at the
1239 static enum exist_status directory_exists_in_index(const char *dirname, int len)
1244 return directory_exists_in_index_icase(dirname, len);
1246 pos = cache_name_pos(dirname, len);
1249 while (pos < active_nr) {
1250 const struct cache_entry *ce = active_cache[pos++];
1251 unsigned char endchar;
1253 if (strncmp(ce->name, dirname, len))
1255 endchar = ce->name[len];
1259 return index_directory;
1260 if (!endchar && S_ISGITLINK(ce->ce_mode))
1261 return index_gitdir;
1263 return index_nonexistent;
1267 * When we find a directory when traversing the filesystem, we
1268 * have three distinct cases:
1271 * - see it as a directory
1274 * and which one we choose depends on a combination of existing
1275 * git index contents and the flags passed into the directory
1276 * traversal routine.
1278 * Case 1: If we *already* have entries in the index under that
1279 * directory name, we always recurse into the directory to see
1282 * Case 2: If we *already* have that directory name as a gitlink,
1283 * we always continue to see it as a gitlink, regardless of whether
1284 * there is an actual git directory there or not (it might not
1285 * be checked out as a subproject!)
1287 * Case 3: if we didn't have it in the index previously, we
1288 * have a few sub-cases:
1290 * (a) if "show_other_directories" is true, we show it as
1291 * just a directory, unless "hide_empty_directories" is
1292 * also true, in which case we need to check if it contains any
1293 * untracked and / or ignored files.
1294 * (b) if it looks like a git directory, and we don't have
1295 * 'no_gitlinks' set we treat it as a gitlink, and show it
1297 * (c) otherwise, we recurse into it.
1299 static enum path_treatment treat_directory(struct dir_struct *dir,
1300 struct untracked_cache_dir *untracked,
1301 const char *dirname, int len, int exclude,
1302 const struct path_simplify *simplify)
1304 /* The "len-1" is to strip the final '/' */
1305 switch (directory_exists_in_index(dirname, len-1)) {
1306 case index_directory:
1307 return path_recurse;
1312 case index_nonexistent:
1313 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1315 if (!(dir->flags & DIR_NO_GITLINKS)) {
1316 unsigned char sha1[20];
1317 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1318 return path_untracked;
1320 return path_recurse;
1323 /* This is the "show_other_directories" case */
1325 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1326 return exclude ? path_excluded : path_untracked;
1328 untracked = lookup_untracked(dir->untracked, untracked, dirname, len);
1329 return read_directory_recursive(dir, dirname, len,
1330 untracked, 1, simplify);
1334 * This is an inexact early pruning of any recursive directory
1335 * reading - if the path cannot possibly be in the pathspec,
1336 * return true, and we'll skip it early.
1338 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1342 const char *match = simplify->path;
1343 int len = simplify->len;
1349 if (!memcmp(path, match, len))
1359 * This function tells us whether an excluded path matches a
1360 * list of "interesting" pathspecs. That is, whether a path matched
1361 * by any of the pathspecs could possibly be ignored by excluding
1362 * the specified path. This can happen if:
1364 * 1. the path is mentioned explicitly in the pathspec
1366 * 2. the path is a directory prefix of some element in the
1369 static int exclude_matches_pathspec(const char *path, int len,
1370 const struct path_simplify *simplify)
1373 for (; simplify->path; simplify++) {
1374 if (len == simplify->len
1375 && !memcmp(path, simplify->path, len))
1377 if (len < simplify->len
1378 && simplify->path[len] == '/'
1379 && !memcmp(path, simplify->path, len))
1386 static int get_index_dtype(const char *path, int len)
1389 const struct cache_entry *ce;
1391 ce = cache_file_exists(path, len, 0);
1393 if (!ce_uptodate(ce))
1395 if (S_ISGITLINK(ce->ce_mode))
1398 * Nobody actually cares about the
1399 * difference between DT_LNK and DT_REG
1404 /* Try to look it up as a directory */
1405 pos = cache_name_pos(path, len);
1409 while (pos < active_nr) {
1410 ce = active_cache[pos++];
1411 if (strncmp(ce->name, path, len))
1413 if (ce->name[len] > '/')
1415 if (ce->name[len] < '/')
1417 if (!ce_uptodate(ce))
1418 break; /* continue? */
1424 static int get_dtype(struct dirent *de, const char *path, int len)
1426 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1429 if (dtype != DT_UNKNOWN)
1431 dtype = get_index_dtype(path, len);
1432 if (dtype != DT_UNKNOWN)
1434 if (lstat(path, &st))
1436 if (S_ISREG(st.st_mode))
1438 if (S_ISDIR(st.st_mode))
1440 if (S_ISLNK(st.st_mode))
1445 static enum path_treatment treat_one_path(struct dir_struct *dir,
1446 struct untracked_cache_dir *untracked,
1447 struct strbuf *path,
1448 const struct path_simplify *simplify,
1449 int dtype, struct dirent *de)
1452 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1454 if (dtype == DT_UNKNOWN)
1455 dtype = get_dtype(de, path->buf, path->len);
1457 /* Always exclude indexed files */
1458 if (dtype != DT_DIR && has_path_in_index)
1462 * When we are looking at a directory P in the working tree,
1463 * there are three cases:
1465 * (1) P exists in the index. Everything inside the directory P in
1466 * the working tree needs to go when P is checked out from the
1469 * (2) P does not exist in the index, but there is P/Q in the index.
1470 * We know P will stay a directory when we check out the contents
1471 * of the index, but we do not know yet if there is a directory
1472 * P/Q in the working tree to be killed, so we need to recurse.
1474 * (3) P does not exist in the index, and there is no P/Q in the index
1475 * to require P to be a directory, either. Only in this case, we
1476 * know that everything inside P will not be killed without
1479 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1480 (dtype == DT_DIR) &&
1481 !has_path_in_index &&
1482 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1485 exclude = is_excluded(dir, path->buf, &dtype);
1488 * Excluded? If we don't explicitly want to show
1489 * ignored files, ignore it
1491 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1492 return path_excluded;
1498 strbuf_addch(path, '/');
1499 return treat_directory(dir, untracked, path->buf, path->len, exclude,
1503 return exclude ? path_excluded : path_untracked;
1507 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1508 struct untracked_cache_dir *untracked,
1509 struct cached_dir *cdir,
1510 struct strbuf *path,
1512 const struct path_simplify *simplify)
1514 strbuf_setlen(path, baselen);
1516 strbuf_addstr(path, cdir->file);
1517 return path_untracked;
1519 strbuf_addstr(path, cdir->ucd->name);
1520 /* treat_one_path() does this before it calls treat_directory() */
1521 if (path->buf[path->len - 1] != '/')
1522 strbuf_addch(path, '/');
1523 if (cdir->ucd->check_only)
1525 * check_only is set as a result of treat_directory() getting
1526 * to its bottom. Verify again the same set of directories
1527 * with check_only set.
1529 return read_directory_recursive(dir, path->buf, path->len,
1530 cdir->ucd, 1, simplify);
1532 * We get path_recurse in the first run when
1533 * directory_exists_in_index() returns index_nonexistent. We
1534 * are sure that new changes in the index does not impact the
1535 * outcome. Return now.
1537 return path_recurse;
1540 static enum path_treatment treat_path(struct dir_struct *dir,
1541 struct untracked_cache_dir *untracked,
1542 struct cached_dir *cdir,
1543 struct strbuf *path,
1545 const struct path_simplify *simplify)
1548 struct dirent *de = cdir->de;
1551 return treat_path_fast(dir, untracked, cdir, path,
1553 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1555 strbuf_setlen(path, baselen);
1556 strbuf_addstr(path, de->d_name);
1557 if (simplify_away(path->buf, path->len, simplify))
1561 return treat_one_path(dir, untracked, path, simplify, dtype, de);
1564 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1568 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1569 dir->untracked_alloc);
1570 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1573 static int valid_cached_dir(struct dir_struct *dir,
1574 struct untracked_cache_dir *untracked,
1575 struct strbuf *path,
1583 if (stat(path->len ? path->buf : ".", &st)) {
1584 invalidate_directory(dir->untracked, untracked);
1585 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1588 if (!untracked->valid ||
1589 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
1590 if (untracked->valid)
1591 invalidate_directory(dir->untracked, untracked);
1592 fill_stat_data(&untracked->stat_data, &st);
1596 if (untracked->check_only != !!check_only) {
1597 invalidate_directory(dir->untracked, untracked);
1602 * prep_exclude will be called eventually on this directory,
1603 * but it's called much later in last_exclude_matching(). We
1604 * need it now to determine the validity of the cache for this
1605 * path. The next calls will be nearly no-op, the way
1606 * prep_exclude() is designed.
1608 if (path->len && path->buf[path->len - 1] != '/') {
1609 strbuf_addch(path, '/');
1610 prep_exclude(dir, path->buf, path->len);
1611 strbuf_setlen(path, path->len - 1);
1613 prep_exclude(dir, path->buf, path->len);
1615 /* hopefully prep_exclude() haven't invalidated this entry... */
1616 return untracked->valid;
1619 static int open_cached_dir(struct cached_dir *cdir,
1620 struct dir_struct *dir,
1621 struct untracked_cache_dir *untracked,
1622 struct strbuf *path,
1625 memset(cdir, 0, sizeof(*cdir));
1626 cdir->untracked = untracked;
1627 if (valid_cached_dir(dir, untracked, path, check_only))
1629 cdir->fdir = opendir(path->len ? path->buf : ".");
1631 dir->untracked->dir_opened++;
1637 static int read_cached_dir(struct cached_dir *cdir)
1640 cdir->de = readdir(cdir->fdir);
1645 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1646 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1656 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1657 struct untracked_cache_dir *d = cdir->untracked;
1658 cdir->file = d->untracked[cdir->nr_files++];
1664 static void close_cached_dir(struct cached_dir *cdir)
1667 closedir(cdir->fdir);
1669 * We have gone through this directory and found no untracked
1670 * entries. Mark it valid.
1672 if (cdir->untracked) {
1673 cdir->untracked->valid = 1;
1674 cdir->untracked->recurse = 1;
1679 * Read a directory tree. We currently ignore anything but
1680 * directories, regular files and symlinks. That's because git
1681 * doesn't handle them at all yet. Maybe that will change some
1684 * Also, we ignore the name ".git" (even if it is not a directory).
1685 * That likely will not change.
1687 * Returns the most significant path_treatment value encountered in the scan.
1689 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1690 const char *base, int baselen,
1691 struct untracked_cache_dir *untracked, int check_only,
1692 const struct path_simplify *simplify)
1694 struct cached_dir cdir;
1695 enum path_treatment state, subdir_state, dir_state = path_none;
1696 struct strbuf path = STRBUF_INIT;
1698 strbuf_add(&path, base, baselen);
1700 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1704 untracked->check_only = !!check_only;
1706 while (!read_cached_dir(&cdir)) {
1707 /* check how the file or directory should be treated */
1708 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);
1710 if (state > dir_state)
1713 /* recurse into subdir if instructed by treat_path */
1714 if (state == path_recurse) {
1715 struct untracked_cache_dir *ud;
1716 ud = lookup_untracked(dir->untracked, untracked,
1718 path.len - baselen);
1720 read_directory_recursive(dir, path.buf, path.len,
1721 ud, check_only, simplify);
1722 if (subdir_state > dir_state)
1723 dir_state = subdir_state;
1727 /* abort early if maximum state has been reached */
1728 if (dir_state == path_untracked) {
1730 add_untracked(untracked, path.buf + baselen);
1733 /* skip the dir_add_* part */
1737 /* add the path to the appropriate result list */
1740 if (dir->flags & DIR_SHOW_IGNORED)
1741 dir_add_name(dir, path.buf, path.len);
1742 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1743 ((dir->flags & DIR_COLLECT_IGNORED) &&
1744 exclude_matches_pathspec(path.buf, path.len,
1746 dir_add_ignored(dir, path.buf, path.len);
1749 case path_untracked:
1750 if (dir->flags & DIR_SHOW_IGNORED)
1752 dir_add_name(dir, path.buf, path.len);
1754 add_untracked(untracked, path.buf + baselen);
1761 close_cached_dir(&cdir);
1763 strbuf_release(&path);
1768 static int cmp_name(const void *p1, const void *p2)
1770 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1771 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1773 return name_compare(e1->name, e1->len, e2->name, e2->len);
1776 static struct path_simplify *create_simplify(const char **pathspec)
1779 struct path_simplify *simplify = NULL;
1784 for (nr = 0 ; ; nr++) {
1786 ALLOC_GROW(simplify, nr + 1, alloc);
1787 match = *pathspec++;
1790 simplify[nr].path = match;
1791 simplify[nr].len = simple_length(match);
1793 simplify[nr].path = NULL;
1794 simplify[nr].len = 0;
1798 static void free_simplify(struct path_simplify *simplify)
1803 static int treat_leading_path(struct dir_struct *dir,
1804 const char *path, int len,
1805 const struct path_simplify *simplify)
1807 struct strbuf sb = STRBUF_INIT;
1808 int baselen, rc = 0;
1810 int old_flags = dir->flags;
1812 while (len && path[len - 1] == '/')
1817 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1819 cp = path + baselen + !!baselen;
1820 cp = memchr(cp, '/', path + len - cp);
1824 baselen = cp - path;
1825 strbuf_setlen(&sb, 0);
1826 strbuf_add(&sb, path, baselen);
1827 if (!is_directory(sb.buf))
1829 if (simplify_away(sb.buf, sb.len, simplify))
1831 if (treat_one_path(dir, NULL, &sb, simplify,
1832 DT_DIR, NULL) == path_none)
1833 break; /* do not recurse into it */
1834 if (len <= baselen) {
1836 break; /* finished checking */
1839 strbuf_release(&sb);
1840 dir->flags = old_flags;
1844 static const char *get_ident_string(void)
1846 static struct strbuf sb = STRBUF_INIT;
1851 if (uname(&uts) < 0)
1852 die_errno(_("failed to get kernel name and information"));
1853 strbuf_addf(&sb, "Location %s, system %s %s %s", get_git_work_tree(),
1854 uts.sysname, uts.release, uts.version);
1858 static int ident_in_untracked(const struct untracked_cache *uc)
1860 const char *end = uc->ident.buf + uc->ident.len;
1861 const char *p = uc->ident.buf;
1863 for (p = uc->ident.buf; p < end; p += strlen(p) + 1)
1864 if (!strcmp(p, get_ident_string()))
1869 void add_untracked_ident(struct untracked_cache *uc)
1871 if (ident_in_untracked(uc))
1873 strbuf_addstr(&uc->ident, get_ident_string());
1874 /* this strbuf contains a list of strings, save NUL too */
1875 strbuf_addch(&uc->ident, 0);
1878 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1880 const struct pathspec *pathspec)
1882 struct untracked_cache_dir *root;
1885 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
1889 * We only support $GIT_DIR/info/exclude and core.excludesfile
1890 * as the global ignore rule files. Any other additions
1891 * (e.g. from command line) invalidate the cache. This
1892 * condition also catches running setup_standard_excludes()
1893 * before setting dir->untracked!
1895 if (dir->unmanaged_exclude_files)
1899 * Optimize for the main use case only: whole-tree git
1900 * status. More work involved in treat_leading_path() if we
1901 * use cache on just a subset of the worktree. pathspec
1902 * support could make the matter even worse.
1904 if (base_len || (pathspec && pathspec->nr))
1907 /* Different set of flags may produce different results */
1908 if (dir->flags != dir->untracked->dir_flags ||
1910 * See treat_directory(), case index_nonexistent. Without
1911 * this flag, we may need to also cache .git file content
1912 * for the resolve_gitlink_ref() call, which we don't.
1914 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1915 /* We don't support collecting ignore files */
1916 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1917 DIR_COLLECT_IGNORED)))
1921 * If we use .gitignore in the cache and now you change it to
1922 * .gitexclude, everything will go wrong.
1924 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
1925 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
1929 * EXC_CMDL is not considered in the cache. If people set it,
1932 if (dir->exclude_list_group[EXC_CMDL].nr)
1936 * An optimization in prep_exclude() does not play well with
1937 * CE_SKIP_WORKTREE. It's a rare case anyway, if a single
1938 * entry has that bit set, disable the whole untracked cache.
1940 for (i = 0; i < active_nr; i++)
1941 if (ce_skip_worktree(active_cache[i]))
1944 if (!ident_in_untracked(dir->untracked)) {
1945 warning(_("Untracked cache is disabled on this system."));
1949 if (!dir->untracked->root) {
1950 const int len = sizeof(*dir->untracked->root);
1951 dir->untracked->root = xmalloc(len);
1952 memset(dir->untracked->root, 0, len);
1955 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
1956 root = dir->untracked->root;
1957 if (hashcmp(dir->ss_info_exclude.sha1,
1958 dir->untracked->ss_info_exclude.sha1)) {
1959 invalidate_gitignore(dir->untracked, root);
1960 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
1962 if (hashcmp(dir->ss_excludes_file.sha1,
1963 dir->untracked->ss_excludes_file.sha1)) {
1964 invalidate_gitignore(dir->untracked, root);
1965 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
1968 /* Make sure this directory is not dropped out at saving phase */
1973 int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1975 struct path_simplify *simplify;
1976 struct untracked_cache_dir *untracked;
1979 * Check out create_simplify()
1982 GUARD_PATHSPEC(pathspec,
1990 if (has_symlink_leading_path(path, len))
1994 * exclude patterns are treated like positive ones in
1995 * create_simplify. Usually exclude patterns should be a
1996 * subset of positive ones, which has no impacts on
1997 * create_simplify().
1999 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
2000 untracked = validate_untracked_cache(dir, len, pathspec);
2003 * make sure untracked cache code path is disabled,
2004 * e.g. prep_exclude()
2006 dir->untracked = NULL;
2007 if (!len || treat_leading_path(dir, path, len, simplify))
2008 read_directory_recursive(dir, path, len, untracked, 0, simplify);
2009 free_simplify(simplify);
2010 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
2011 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
2012 if (dir->untracked) {
2013 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2014 trace_printf_key(&trace_untracked_stats,
2015 "node creation: %u\n"
2016 "gitignore invalidation: %u\n"
2017 "directory invalidation: %u\n"
2019 dir->untracked->dir_created,
2020 dir->untracked->gitignore_invalidated,
2021 dir->untracked->dir_invalidated,
2022 dir->untracked->dir_opened);
2023 if (dir->untracked == the_index.untracked &&
2024 (dir->untracked->dir_opened ||
2025 dir->untracked->gitignore_invalidated ||
2026 dir->untracked->dir_invalidated))
2027 the_index.cache_changed |= UNTRACKED_CHANGED;
2028 if (dir->untracked != the_index.untracked) {
2029 free(dir->untracked);
2030 dir->untracked = NULL;
2036 int file_exists(const char *f)
2039 return lstat(f, &sb) == 0;
2043 * Given two normalized paths (a trailing slash is ok), if subdir is
2044 * outside dir, return -1. Otherwise return the offset in subdir that
2045 * can be used as relative path to dir.
2047 int dir_inside_of(const char *subdir, const char *dir)
2051 assert(dir && subdir && *dir && *subdir);
2053 while (*dir && *subdir && *dir == *subdir) {
2059 /* hel[p]/me vs hel[l]/yeah */
2060 if (*dir && *subdir)
2064 return !*dir ? offset : -1; /* same dir */
2066 /* foo/[b]ar vs foo/[] */
2067 if (is_dir_sep(dir[-1]))
2068 return is_dir_sep(subdir[-1]) ? offset : -1;
2070 /* foo[/]bar vs foo[] */
2071 return is_dir_sep(*subdir) ? offset + 1 : -1;
2074 int is_inside_dir(const char *dir)
2083 rc = (dir_inside_of(cwd, dir) >= 0);
2088 int is_empty_dir(const char *path)
2090 DIR *dir = opendir(path);
2097 while ((e = readdir(dir)) != NULL)
2098 if (!is_dot_or_dotdot(e->d_name)) {
2107 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2111 int ret = 0, original_len = path->len, len, kept_down = 0;
2112 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2113 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2114 unsigned char submodule_head[20];
2116 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2117 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2118 /* Do not descend and nuke a nested git work tree. */
2124 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2125 dir = opendir(path->buf);
2127 if (errno == ENOENT)
2128 return keep_toplevel ? -1 : 0;
2129 else if (errno == EACCES && !keep_toplevel)
2131 * An empty dir could be removable even if it
2134 return rmdir(path->buf);
2138 if (path->buf[original_len - 1] != '/')
2139 strbuf_addch(path, '/');
2142 while ((e = readdir(dir)) != NULL) {
2144 if (is_dot_or_dotdot(e->d_name))
2147 strbuf_setlen(path, len);
2148 strbuf_addstr(path, e->d_name);
2149 if (lstat(path->buf, &st)) {
2150 if (errno == ENOENT)
2152 * file disappeared, which is what we
2157 } else if (S_ISDIR(st.st_mode)) {
2158 if (!remove_dir_recurse(path, flag, &kept_down))
2159 continue; /* happy */
2160 } else if (!only_empty &&
2161 (!unlink(path->buf) || errno == ENOENT)) {
2162 continue; /* happy, too */
2165 /* path too long, stat fails, or non-directory still exists */
2171 strbuf_setlen(path, original_len);
2172 if (!ret && !keep_toplevel && !kept_down)
2173 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2176 * report the uplevel that it is not an error that we
2177 * did not rmdir() our directory.
2183 int remove_dir_recursively(struct strbuf *path, int flag)
2185 return remove_dir_recurse(path, flag, NULL);
2188 void setup_standard_excludes(struct dir_struct *dir)
2192 dir->exclude_per_dir = ".gitignore";
2194 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2196 excludes_file = xdg_config_home("ignore");
2197 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2198 add_excludes_from_file_1(dir, excludes_file,
2199 dir->untracked ? &dir->ss_excludes_file : NULL);
2201 /* per repository user preference */
2202 path = git_path("info/exclude");
2203 if (!access_or_warn(path, R_OK, 0))
2204 add_excludes_from_file_1(dir, path,
2205 dir->untracked ? &dir->ss_info_exclude : NULL);
2208 int remove_path(const char *name)
2212 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
2215 slash = strrchr(name, '/');
2217 char *dirs = xstrdup(name);
2218 slash = dirs + (slash - name);
2221 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2228 * Frees memory within dir which was allocated for exclude lists and
2229 * the exclude_stack. Does not free dir itself.
2231 void clear_directory(struct dir_struct *dir)
2234 struct exclude_list_group *group;
2235 struct exclude_list *el;
2236 struct exclude_stack *stk;
2238 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2239 group = &dir->exclude_list_group[i];
2240 for (j = 0; j < group->nr; j++) {
2243 free((char *)el->src);
2244 clear_exclude_list(el);
2249 stk = dir->exclude_stack;
2251 struct exclude_stack *prev = stk->prev;
2255 strbuf_release(&dir->basebuf);
2258 struct ondisk_untracked_cache {
2259 struct stat_data info_exclude_stat;
2260 struct stat_data excludes_file_stat;
2262 unsigned char info_exclude_sha1[20];
2263 unsigned char excludes_file_sha1[20];
2264 char exclude_per_dir[FLEX_ARRAY];
2267 #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2270 int index; /* number of written untracked_cache_dir */
2271 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2272 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2273 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2275 struct strbuf sb_stat;
2276 struct strbuf sb_sha1;
2279 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2281 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2282 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2283 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2284 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2285 to->sd_dev = htonl(from->sd_dev);
2286 to->sd_ino = htonl(from->sd_ino);
2287 to->sd_uid = htonl(from->sd_uid);
2288 to->sd_gid = htonl(from->sd_gid);
2289 to->sd_size = htonl(from->sd_size);
2292 static void write_one_dir(struct untracked_cache_dir *untracked,
2293 struct write_data *wd)
2295 struct stat_data stat_data;
2296 struct strbuf *out = &wd->out;
2297 unsigned char intbuf[16];
2298 unsigned int intlen, value;
2299 int i = wd->index++;
2302 * untracked_nr should be reset whenever valid is clear, but
2305 if (!untracked->valid) {
2306 untracked->untracked_nr = 0;
2307 untracked->check_only = 0;
2310 if (untracked->check_only)
2311 ewah_set(wd->check_only, i);
2312 if (untracked->valid) {
2313 ewah_set(wd->valid, i);
2314 stat_data_to_disk(&stat_data, &untracked->stat_data);
2315 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2317 if (!is_null_sha1(untracked->exclude_sha1)) {
2318 ewah_set(wd->sha1_valid, i);
2319 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2322 intlen = encode_varint(untracked->untracked_nr, intbuf);
2323 strbuf_add(out, intbuf, intlen);
2325 /* skip non-recurse directories */
2326 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2327 if (untracked->dirs[i]->recurse)
2329 intlen = encode_varint(value, intbuf);
2330 strbuf_add(out, intbuf, intlen);
2332 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2334 for (i = 0; i < untracked->untracked_nr; i++)
2335 strbuf_add(out, untracked->untracked[i],
2336 strlen(untracked->untracked[i]) + 1);
2338 for (i = 0; i < untracked->dirs_nr; i++)
2339 if (untracked->dirs[i]->recurse)
2340 write_one_dir(untracked->dirs[i], wd);
2343 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2345 struct ondisk_untracked_cache *ouc;
2346 struct write_data wd;
2347 unsigned char varbuf[16];
2348 int len = 0, varint_len;
2349 if (untracked->exclude_per_dir)
2350 len = strlen(untracked->exclude_per_dir);
2351 ouc = xmalloc(sizeof(*ouc) + len + 1);
2352 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2353 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2354 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2355 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2356 ouc->dir_flags = htonl(untracked->dir_flags);
2357 memcpy(ouc->exclude_per_dir, untracked->exclude_per_dir, len + 1);
2359 varint_len = encode_varint(untracked->ident.len, varbuf);
2360 strbuf_add(out, varbuf, varint_len);
2361 strbuf_add(out, untracked->ident.buf, untracked->ident.len);
2363 strbuf_add(out, ouc, ouc_size(len));
2367 if (!untracked->root) {
2368 varint_len = encode_varint(0, varbuf);
2369 strbuf_add(out, varbuf, varint_len);
2374 wd.check_only = ewah_new();
2375 wd.valid = ewah_new();
2376 wd.sha1_valid = ewah_new();
2377 strbuf_init(&wd.out, 1024);
2378 strbuf_init(&wd.sb_stat, 1024);
2379 strbuf_init(&wd.sb_sha1, 1024);
2380 write_one_dir(untracked->root, &wd);
2382 varint_len = encode_varint(wd.index, varbuf);
2383 strbuf_add(out, varbuf, varint_len);
2384 strbuf_addbuf(out, &wd.out);
2385 ewah_serialize_strbuf(wd.valid, out);
2386 ewah_serialize_strbuf(wd.check_only, out);
2387 ewah_serialize_strbuf(wd.sha1_valid, out);
2388 strbuf_addbuf(out, &wd.sb_stat);
2389 strbuf_addbuf(out, &wd.sb_sha1);
2390 strbuf_addch(out, '\0'); /* safe guard for string lists */
2392 ewah_free(wd.valid);
2393 ewah_free(wd.check_only);
2394 ewah_free(wd.sha1_valid);
2395 strbuf_release(&wd.out);
2396 strbuf_release(&wd.sb_stat);
2397 strbuf_release(&wd.sb_sha1);
2400 static void free_untracked(struct untracked_cache_dir *ucd)
2405 for (i = 0; i < ucd->dirs_nr; i++)
2406 free_untracked(ucd->dirs[i]);
2407 for (i = 0; i < ucd->untracked_nr; i++)
2408 free(ucd->untracked[i]);
2409 free(ucd->untracked);
2414 void free_untracked_cache(struct untracked_cache *uc)
2417 free_untracked(uc->root);
2423 struct untracked_cache_dir **ucd;
2424 struct ewah_bitmap *check_only;
2425 struct ewah_bitmap *valid;
2426 struct ewah_bitmap *sha1_valid;
2427 const unsigned char *data;
2428 const unsigned char *end;
2431 static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2433 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2434 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2435 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2436 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2437 to->sd_dev = get_be32(&from->sd_dev);
2438 to->sd_ino = get_be32(&from->sd_ino);
2439 to->sd_uid = get_be32(&from->sd_uid);
2440 to->sd_gid = get_be32(&from->sd_gid);
2441 to->sd_size = get_be32(&from->sd_size);
2444 static int read_one_dir(struct untracked_cache_dir **untracked_,
2445 struct read_data *rd)
2447 struct untracked_cache_dir ud, *untracked;
2448 const unsigned char *next, *data = rd->data, *end = rd->end;
2452 memset(&ud, 0, sizeof(ud));
2455 value = decode_varint(&next);
2459 ud.untracked_alloc = value;
2460 ud.untracked_nr = value;
2461 if (ud.untracked_nr)
2462 ud.untracked = xmalloc(sizeof(*ud.untracked) * ud.untracked_nr);
2466 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2469 ud.dirs = xmalloc(sizeof(*ud.dirs) * ud.dirs_nr);
2472 len = strlen((const char *)data);
2473 next = data + len + 1;
2476 *untracked_ = untracked = xmalloc(sizeof(*untracked) + len);
2477 memcpy(untracked, &ud, sizeof(ud));
2478 memcpy(untracked->name, data, len + 1);
2481 for (i = 0; i < untracked->untracked_nr; i++) {
2482 len = strlen((const char *)data);
2483 next = data + len + 1;
2486 untracked->untracked[i] = xstrdup((const char*)data);
2490 rd->ucd[rd->index++] = untracked;
2493 for (i = 0; i < untracked->dirs_nr; i++) {
2494 len = read_one_dir(untracked->dirs + i, rd);
2501 static void set_check_only(size_t pos, void *cb)
2503 struct read_data *rd = cb;
2504 struct untracked_cache_dir *ud = rd->ucd[pos];
2508 static void read_stat(size_t pos, void *cb)
2510 struct read_data *rd = cb;
2511 struct untracked_cache_dir *ud = rd->ucd[pos];
2512 if (rd->data + sizeof(struct stat_data) > rd->end) {
2513 rd->data = rd->end + 1;
2516 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2517 rd->data += sizeof(struct stat_data);
2521 static void read_sha1(size_t pos, void *cb)
2523 struct read_data *rd = cb;
2524 struct untracked_cache_dir *ud = rd->ucd[pos];
2525 if (rd->data + 20 > rd->end) {
2526 rd->data = rd->end + 1;
2529 hashcpy(ud->exclude_sha1, rd->data);
2533 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2534 const struct stat_data *stat,
2535 const unsigned char *sha1)
2537 stat_data_from_disk(&sha1_stat->stat, stat);
2538 hashcpy(sha1_stat->sha1, sha1);
2539 sha1_stat->valid = 1;
2542 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2544 const struct ondisk_untracked_cache *ouc;
2545 struct untracked_cache *uc;
2546 struct read_data rd;
2547 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2551 if (sz <= 1 || end[-1] != '\0')
2555 ident_len = decode_varint(&next);
2556 if (next + ident_len > end)
2558 ident = (const char *)next;
2561 ouc = (const struct ondisk_untracked_cache *)next;
2562 if (next + ouc_size(0) > end)
2565 uc = xcalloc(1, sizeof(*uc));
2566 strbuf_init(&uc->ident, ident_len);
2567 strbuf_add(&uc->ident, ident, ident_len);
2568 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2569 ouc->info_exclude_sha1);
2570 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2571 ouc->excludes_file_sha1);
2572 uc->dir_flags = get_be32(&ouc->dir_flags);
2573 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2574 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2575 next += ouc_size(strlen(ouc->exclude_per_dir));
2579 len = decode_varint(&next);
2580 if (next > end || len == 0)
2583 rd.valid = ewah_new();
2584 rd.check_only = ewah_new();
2585 rd.sha1_valid = ewah_new();
2589 rd.ucd = xmalloc(sizeof(*rd.ucd) * len);
2591 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2595 len = ewah_read_mmap(rd.valid, next, end - next);
2600 len = ewah_read_mmap(rd.check_only, next, end - next);
2605 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2609 ewah_each_bit(rd.check_only, set_check_only, &rd);
2610 rd.data = next + len;
2611 ewah_each_bit(rd.valid, read_stat, &rd);
2612 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2617 ewah_free(rd.valid);
2618 ewah_free(rd.check_only);
2619 ewah_free(rd.sha1_valid);
2622 free_untracked_cache(uc);
2628 void untracked_cache_invalidate_path(struct index_state *istate,
2632 struct untracked_cache_dir *d;
2633 if (!istate->untracked || !istate->untracked->root)
2635 sep = strrchr(path, '/');
2637 d = lookup_untracked(istate->untracked,
2638 istate->untracked->root,
2641 d = istate->untracked->root;
2642 istate->untracked->dir_invalidated++;
2644 d->untracked_nr = 0;
2647 void untracked_cache_remove_from_index(struct index_state *istate,
2650 untracked_cache_invalidate_path(istate, path);
2653 void untracked_cache_add_to_index(struct index_state *istate,
2656 untracked_cache_invalidate_path(istate, path);