2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
12 struct path_simplify {
17 static int read_directory_recursive(struct dir_struct *dir, const char *path, int len,
18 int check_only, const struct path_simplify *simplify);
19 static int get_dtype(struct dirent *de, const char *path, int len);
21 /* helper string functions with support for the ignore_case flag */
22 int strcmp_icase(const char *a, const char *b)
24 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
27 int strncmp_icase(const char *a, const char *b, size_t count)
29 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
32 int fnmatch_icase(const char *pattern, const char *string, int flags)
34 return fnmatch(pattern, string, flags | (ignore_case ? FNM_CASEFOLD : 0));
37 static int common_prefix(const char **pathspec)
39 const char *path, *slash, *next;
46 slash = strrchr(path, '/');
51 * The first 'prefix' characters of 'path' are common leading
52 * path components among the pathspecs we have seen so far,
53 * including the trailing slash.
55 prefix = slash - path + 1;
56 while ((next = *++pathspec) != NULL) {
57 int len, last_matching_slash = -1;
58 for (len = 0; len < prefix && next[len] == path[len]; len++)
60 last_matching_slash = len;
63 if (last_matching_slash < 0)
65 prefix = last_matching_slash + 1;
70 int fill_directory(struct dir_struct *dir, const char **pathspec)
76 * Calculate common prefix for the pathspec, and
77 * use that to optimize the directory walk
79 len = common_prefix(pathspec);
83 path = xmemdupz(*pathspec, len);
85 /* Read the directory and prune it */
86 read_directory(dir, path, len, pathspec);
91 * Does 'match' match the given name?
94 * (1) the 'match' string is leading directory of 'name', or
95 * (2) the 'match' string is a wildcard and matches 'name', or
96 * (3) the 'match' string is exactly the same as 'name'.
98 * and the return value tells which case it was.
100 * It returns 0 when there is no match.
102 static int match_one(const char *match, const char *name, int namelen)
106 /* If the match was just the prefix, we matched */
108 return MATCHED_RECURSIVELY;
112 unsigned char c1 = tolower(*match);
113 unsigned char c2 = tolower(*name);
114 if (c1 == '\0' || is_glob_special(c1))
124 unsigned char c1 = *match;
125 unsigned char c2 = *name;
126 if (c1 == '\0' || is_glob_special(c1))
138 * If we don't match the matchstring exactly,
139 * we need to match by fnmatch
141 matchlen = strlen(match);
142 if (strncmp_icase(match, name, matchlen))
143 return !fnmatch_icase(match, name, 0) ? MATCHED_FNMATCH : 0;
145 if (namelen == matchlen)
146 return MATCHED_EXACTLY;
147 if (match[matchlen-1] == '/' || name[matchlen] == '/')
148 return MATCHED_RECURSIVELY;
153 * Given a name and a list of pathspecs, see if the name matches
154 * any of the pathspecs. The caller is also interested in seeing
155 * all pathspec matches some names it calls this function with
156 * (otherwise the user could have mistyped the unmatched pathspec),
157 * and a mark is left in seen[] array for pathspec element that
158 * actually matched anything.
160 int match_pathspec(const char **pathspec, const char *name, int namelen,
161 int prefix, char *seen)
171 for (i = 0; pathspec[i] != NULL; i++) {
173 const char *match = pathspec[i] + prefix;
174 if (seen && seen[i] == MATCHED_EXACTLY)
176 how = match_one(match, name, namelen);
180 if (seen && seen[i] < how)
187 static int no_wildcard(const char *string)
189 return string[strcspn(string, "*?[{\\")] == '\0';
192 void add_exclude(const char *string, const char *base,
193 int baselen, struct exclude_list *which)
200 if (*string == '!') {
204 len = strlen(string);
205 if (len && string[len - 1] == '/') {
207 x = xmalloc(sizeof(*x) + len);
209 memcpy(s, string, len - 1);
213 flags = EXC_FLAG_MUSTBEDIR;
215 x = xmalloc(sizeof(*x));
218 x->to_exclude = to_exclude;
219 x->patternlen = strlen(string);
221 x->baselen = baselen;
223 if (!strchr(string, '/'))
224 x->flags |= EXC_FLAG_NODIR;
225 if (no_wildcard(string))
226 x->flags |= EXC_FLAG_NOWILDCARD;
227 if (*string == '*' && no_wildcard(string+1))
228 x->flags |= EXC_FLAG_ENDSWITH;
229 ALLOC_GROW(which->excludes, which->nr + 1, which->alloc);
230 which->excludes[which->nr++] = x;
233 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
237 enum object_type type;
239 struct index_state *istate = &the_index;
242 pos = index_name_pos(istate, path, len);
245 if (!ce_skip_worktree(istate->cache[pos]))
247 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
248 if (!data || type != OBJ_BLOB) {
256 int add_excludes_from_file_to_list(const char *fname,
260 struct exclude_list *which,
268 fd = open(fname, O_RDONLY);
269 if (fd < 0 || fstat(fd, &st) < 0) {
273 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
279 if (buf[size-1] != '\n') {
280 buf = xrealloc(buf, size+1);
285 size = xsize_t(st.st_size);
290 buf = xmalloc(size+1);
291 if (read_in_full(fd, buf, size) != size) {
303 for (i = 0; i < size; i++) {
304 if (buf[i] == '\n') {
305 if (entry != buf + i && entry[0] != '#') {
306 buf[i - (i && buf[i-1] == '\r')] = 0;
307 add_exclude(entry, base, baselen, which);
315 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
317 if (add_excludes_from_file_to_list(fname, "", 0, NULL,
318 &dir->exclude_list[EXC_FILE], 0) < 0)
319 die("cannot use %s as an exclude file", fname);
322 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
324 struct exclude_list *el;
325 struct exclude_stack *stk = NULL;
328 if ((!dir->exclude_per_dir) ||
329 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
330 return; /* too long a path -- ignore */
332 /* Pop the ones that are not the prefix of the path being checked. */
333 el = &dir->exclude_list[EXC_DIRS];
334 while ((stk = dir->exclude_stack) != NULL) {
335 if (stk->baselen <= baselen &&
336 !strncmp(dir->basebuf, base, stk->baselen))
338 dir->exclude_stack = stk->prev;
339 while (stk->exclude_ix < el->nr)
340 free(el->excludes[--el->nr]);
345 /* Read from the parent directories and push them down. */
346 current = stk ? stk->baselen : -1;
347 while (current < baselen) {
348 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
356 cp = strchr(base + current + 1, '/');
358 die("oops in prep_exclude");
361 stk->prev = dir->exclude_stack;
362 stk->baselen = cp - base;
363 stk->exclude_ix = el->nr;
364 memcpy(dir->basebuf + current, base + current,
365 stk->baselen - current);
366 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
367 add_excludes_from_file_to_list(dir->basebuf,
368 dir->basebuf, stk->baselen,
369 &stk->filebuf, el, 1);
370 dir->exclude_stack = stk;
371 current = stk->baselen;
373 dir->basebuf[baselen] = '\0';
376 /* Scan the list and let the last match determine the fate.
377 * Return 1 for exclude, 0 for include and -1 for undecided.
379 int excluded_from_list(const char *pathname,
380 int pathlen, const char *basename, int *dtype,
381 struct exclude_list *el)
386 for (i = el->nr - 1; 0 <= i; i--) {
387 struct exclude *x = el->excludes[i];
388 const char *exclude = x->pattern;
389 int to_exclude = x->to_exclude;
391 if (x->flags & EXC_FLAG_MUSTBEDIR) {
393 if (!prefixcmp(pathname, exclude) &&
394 pathname[x->patternlen] == '/')
399 if (*dtype == DT_UNKNOWN)
400 *dtype = get_dtype(NULL, pathname, pathlen);
401 if (*dtype != DT_DIR)
405 if (x->flags & EXC_FLAG_NODIR) {
407 if (x->flags & EXC_FLAG_NOWILDCARD) {
408 if (!strcmp_icase(exclude, basename))
410 } else if (x->flags & EXC_FLAG_ENDSWITH) {
411 if (x->patternlen - 1 <= pathlen &&
412 !strcmp_icase(exclude + 1, pathname + pathlen - x->patternlen + 1))
415 if (fnmatch_icase(exclude, basename, 0) == 0)
420 /* match with FNM_PATHNAME:
421 * exclude has base (baselen long) implicitly
424 int baselen = x->baselen;
428 if (pathlen < baselen ||
429 (baselen && pathname[baselen-1] != '/') ||
430 strncmp_icase(pathname, x->base, baselen))
433 if (x->flags & EXC_FLAG_NOWILDCARD) {
434 if (!strcmp_icase(exclude, pathname + baselen))
437 if (fnmatch_icase(exclude, pathname+baselen,
444 return -1; /* undecided */
447 int excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
449 int pathlen = strlen(pathname);
451 const char *basename = strrchr(pathname, '/');
452 basename = (basename) ? basename+1 : pathname;
454 prep_exclude(dir, pathname, basename-pathname);
455 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
456 switch (excluded_from_list(pathname, pathlen, basename,
457 dtype_p, &dir->exclude_list[st])) {
467 static struct dir_entry *dir_entry_new(const char *pathname, int len)
469 struct dir_entry *ent;
471 ent = xmalloc(sizeof(*ent) + len + 1);
473 memcpy(ent->name, pathname, len);
478 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
480 if (cache_name_exists(pathname, len, ignore_case))
483 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
484 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
487 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
489 if (!cache_name_is_other(pathname, len))
492 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
493 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
497 index_nonexistent = 0,
503 * Do not use the alphabetically stored index to look up
504 * the directory name; instead, use the case insensitive
507 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
509 struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
510 unsigned char endchar;
513 return index_nonexistent;
514 endchar = ce->name[len];
517 * The cache_entry structure returned will contain this dirname
518 * and possibly additional path components.
521 return index_directory;
524 * If there are no additional path components, then this cache_entry
525 * represents a submodule. Submodules, despite being directories,
526 * are stored in the cache without a closing slash.
528 if (!endchar && S_ISGITLINK(ce->ce_mode))
531 /* This should never be hit, but it exists just in case. */
532 return index_nonexistent;
536 * The index sorts alphabetically by entry name, which
537 * means that a gitlink sorts as '\0' at the end, while
538 * a directory (which is defined not as an entry, but as
539 * the files it contains) will sort with the '/' at the
542 static enum exist_status directory_exists_in_index(const char *dirname, int len)
547 return directory_exists_in_index_icase(dirname, len);
549 pos = cache_name_pos(dirname, len);
552 while (pos < active_nr) {
553 struct cache_entry *ce = active_cache[pos++];
554 unsigned char endchar;
556 if (strncmp(ce->name, dirname, len))
558 endchar = ce->name[len];
562 return index_directory;
563 if (!endchar && S_ISGITLINK(ce->ce_mode))
566 return index_nonexistent;
570 * When we find a directory when traversing the filesystem, we
571 * have three distinct cases:
574 * - see it as a directory
577 * and which one we choose depends on a combination of existing
578 * git index contents and the flags passed into the directory
581 * Case 1: If we *already* have entries in the index under that
582 * directory name, we always recurse into the directory to see
585 * Case 2: If we *already* have that directory name as a gitlink,
586 * we always continue to see it as a gitlink, regardless of whether
587 * there is an actual git directory there or not (it might not
588 * be checked out as a subproject!)
590 * Case 3: if we didn't have it in the index previously, we
591 * have a few sub-cases:
593 * (a) if "show_other_directories" is true, we show it as
594 * just a directory, unless "hide_empty_directories" is
595 * also true and the directory is empty, in which case
596 * we just ignore it entirely.
597 * (b) if it looks like a git directory, and we don't have
598 * 'no_gitlinks' set we treat it as a gitlink, and show it
600 * (c) otherwise, we recurse into it.
602 enum directory_treatment {
605 recurse_into_directory
608 static enum directory_treatment treat_directory(struct dir_struct *dir,
609 const char *dirname, int len,
610 const struct path_simplify *simplify)
612 /* The "len-1" is to strip the final '/' */
613 switch (directory_exists_in_index(dirname, len-1)) {
614 case index_directory:
615 return recurse_into_directory;
618 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
619 return ignore_directory;
620 return show_directory;
622 case index_nonexistent:
623 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
625 if (!(dir->flags & DIR_NO_GITLINKS)) {
626 unsigned char sha1[20];
627 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
628 return show_directory;
630 return recurse_into_directory;
633 /* This is the "show_other_directories" case */
634 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
635 return show_directory;
636 if (!read_directory_recursive(dir, dirname, len, 1, simplify))
637 return ignore_directory;
638 return show_directory;
642 * This is an inexact early pruning of any recursive directory
643 * reading - if the path cannot possibly be in the pathspec,
644 * return true, and we'll skip it early.
646 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
650 const char *match = simplify->path;
651 int len = simplify->len;
657 if (!memcmp(path, match, len))
667 * This function tells us whether an excluded path matches a
668 * list of "interesting" pathspecs. That is, whether a path matched
669 * by any of the pathspecs could possibly be ignored by excluding
670 * the specified path. This can happen if:
672 * 1. the path is mentioned explicitly in the pathspec
674 * 2. the path is a directory prefix of some element in the
677 static int exclude_matches_pathspec(const char *path, int len,
678 const struct path_simplify *simplify)
681 for (; simplify->path; simplify++) {
682 if (len == simplify->len
683 && !memcmp(path, simplify->path, len))
685 if (len < simplify->len
686 && simplify->path[len] == '/'
687 && !memcmp(path, simplify->path, len))
694 static int get_index_dtype(const char *path, int len)
697 struct cache_entry *ce;
699 ce = cache_name_exists(path, len, 0);
701 if (!ce_uptodate(ce))
703 if (S_ISGITLINK(ce->ce_mode))
706 * Nobody actually cares about the
707 * difference between DT_LNK and DT_REG
712 /* Try to look it up as a directory */
713 pos = cache_name_pos(path, len);
717 while (pos < active_nr) {
718 ce = active_cache[pos++];
719 if (strncmp(ce->name, path, len))
721 if (ce->name[len] > '/')
723 if (ce->name[len] < '/')
725 if (!ce_uptodate(ce))
726 break; /* continue? */
732 static int get_dtype(struct dirent *de, const char *path, int len)
734 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
737 if (dtype != DT_UNKNOWN)
739 dtype = get_index_dtype(path, len);
740 if (dtype != DT_UNKNOWN)
742 if (lstat(path, &st))
744 if (S_ISREG(st.st_mode))
746 if (S_ISDIR(st.st_mode))
748 if (S_ISLNK(st.st_mode))
753 enum path_treatment {
759 static enum path_treatment treat_one_path(struct dir_struct *dir,
760 char *path, int *len,
761 const struct path_simplify *simplify,
762 int dtype, struct dirent *de)
764 int exclude = excluded(dir, path, &dtype);
765 if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
766 && exclude_matches_pathspec(path, *len, simplify))
767 dir_add_ignored(dir, path, *len);
770 * Excluded? If we don't explicitly want to show
771 * ignored files, ignore it
773 if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
776 if (dtype == DT_UNKNOWN)
777 dtype = get_dtype(de, path, *len);
780 * Do we want to see just the ignored files?
781 * We still need to recurse into directories,
782 * even if we don't ignore them, since the
783 * directory may contain files that we do..
785 if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
794 memcpy(path + *len, "/", 2);
796 switch (treat_directory(dir, path, *len, simplify)) {
798 if (exclude != !!(dir->flags
802 case recurse_into_directory:
804 case ignore_directory:
815 static enum path_treatment treat_path(struct dir_struct *dir,
817 char *path, int path_max,
819 const struct path_simplify *simplify,
824 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
826 *len = strlen(de->d_name);
827 /* Ignore overly long pathnames! */
828 if (*len + baselen + 8 > path_max)
830 memcpy(path + baselen, de->d_name, *len + 1);
832 if (simplify_away(path, *len, simplify))
836 return treat_one_path(dir, path, len, simplify, dtype, de);
840 * Read a directory tree. We currently ignore anything but
841 * directories, regular files and symlinks. That's because git
842 * doesn't handle them at all yet. Maybe that will change some
845 * Also, we ignore the name ".git" (even if it is not a directory).
846 * That likely will not change.
848 static int read_directory_recursive(struct dir_struct *dir,
849 const char *base, int baselen,
851 const struct path_simplify *simplify)
853 DIR *fdir = opendir(*base ? base : ".");
858 char path[PATH_MAX + 1];
859 memcpy(path, base, baselen);
861 while ((de = readdir(fdir)) != NULL) {
863 switch (treat_path(dir, de, path, sizeof(path),
864 baselen, simplify, &len)) {
866 contents += read_directory_recursive
867 (dir, path, len, 0, simplify);
878 dir_add_name(dir, path, len);
887 static int cmp_name(const void *p1, const void *p2)
889 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
890 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
892 return cache_name_compare(e1->name, e1->len,
897 * Return the length of the "simple" part of a path match limiter.
899 static int simple_length(const char *match)
904 unsigned char c = *match++;
906 if (c == '\0' || is_glob_special(c))
911 static struct path_simplify *create_simplify(const char **pathspec)
914 struct path_simplify *simplify = NULL;
919 for (nr = 0 ; ; nr++) {
922 alloc = alloc_nr(alloc);
923 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
928 simplify[nr].path = match;
929 simplify[nr].len = simple_length(match);
931 simplify[nr].path = NULL;
932 simplify[nr].len = 0;
936 static void free_simplify(struct path_simplify *simplify)
941 static int treat_leading_path(struct dir_struct *dir,
942 const char *path, int len,
943 const struct path_simplify *simplify)
945 char pathbuf[PATH_MAX];
949 while (len && path[len - 1] == '/')
955 cp = path + baselen + !!baselen;
956 cp = memchr(cp, '/', path + len - cp);
961 memcpy(pathbuf, path, baselen);
962 pathbuf[baselen] = '\0';
963 if (!is_directory(pathbuf))
965 if (simplify_away(pathbuf, baselen, simplify))
968 if (treat_one_path(dir, pathbuf, &blen, simplify,
969 DT_DIR, NULL) == path_ignored)
970 return 0; /* do not recurse into it */
972 return 1; /* finished checking */
976 int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
978 struct path_simplify *simplify;
980 if (has_symlink_leading_path(path, len))
983 simplify = create_simplify(pathspec);
984 if (!len || treat_leading_path(dir, path, len, simplify))
985 read_directory_recursive(dir, path, len, 0, simplify);
986 free_simplify(simplify);
987 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
988 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
992 int file_exists(const char *f)
995 return lstat(f, &sb) == 0;
999 * get_relative_cwd() gets the prefix of the current working directory
1000 * relative to 'dir'. If we are not inside 'dir', it returns NULL.
1002 * As a convenience, it also returns NULL if 'dir' is already NULL. The
1003 * reason for this behaviour is that it is natural for functions returning
1004 * directory names to return NULL to say "this directory does not exist"
1005 * or "this directory is invalid". These cases are usually handled the
1006 * same as if the cwd is not inside 'dir' at all, so get_relative_cwd()
1007 * returns NULL for both of them.
1009 * Most notably, get_relative_cwd(buffer, size, get_git_work_tree())
1010 * unifies the handling of "outside work tree" with "no work tree at all".
1012 char *get_relative_cwd(char *buffer, int size, const char *dir)
1018 if (!getcwd(buffer, size))
1019 die_errno("can't find the current directory");
1021 if (!is_absolute_path(dir))
1022 dir = make_absolute_path(dir);
1024 while (*dir && *dir == *cwd) {
1040 int is_inside_dir(const char *dir)
1042 char buffer[PATH_MAX];
1043 return get_relative_cwd(buffer, sizeof(buffer), dir) != NULL;
1046 int is_empty_dir(const char *path)
1048 DIR *dir = opendir(path);
1055 while ((e = readdir(dir)) != NULL)
1056 if (!is_dot_or_dotdot(e->d_name)) {
1065 int remove_dir_recursively(struct strbuf *path, int flag)
1069 int ret = 0, original_len = path->len, len;
1070 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1071 unsigned char submodule_head[20];
1073 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1074 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head))
1075 /* Do not descend and nuke a nested git work tree. */
1078 dir = opendir(path->buf);
1081 if (path->buf[original_len - 1] != '/')
1082 strbuf_addch(path, '/');
1085 while ((e = readdir(dir)) != NULL) {
1087 if (is_dot_or_dotdot(e->d_name))
1090 strbuf_setlen(path, len);
1091 strbuf_addstr(path, e->d_name);
1092 if (lstat(path->buf, &st))
1094 else if (S_ISDIR(st.st_mode)) {
1095 if (!remove_dir_recursively(path, only_empty))
1096 continue; /* happy */
1097 } else if (!only_empty && !unlink(path->buf))
1098 continue; /* happy, too */
1100 /* path too long, stat fails, or non-directory still exists */
1106 strbuf_setlen(path, original_len);
1108 ret = rmdir(path->buf);
1112 void setup_standard_excludes(struct dir_struct *dir)
1116 dir->exclude_per_dir = ".gitignore";
1117 path = git_path("info/exclude");
1118 if (!access(path, R_OK))
1119 add_excludes_from_file(dir, path);
1120 if (excludes_file && !access(excludes_file, R_OK))
1121 add_excludes_from_file(dir, excludes_file);
1124 int remove_path(const char *name)
1128 if (unlink(name) && errno != ENOENT)
1131 slash = strrchr(name, '/');
1133 char *dirs = xstrdup(name);
1134 slash = dirs + (slash - name);
1137 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));