Merge git://git.bogomips.org/git-svn
[alt-git.git] / dir.c
blobb439ff0e63481bebb4039901b029bad978b46929
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"
16 struct path_simplify {
17 int len;
18 const char *path;
22 * Tells read_directory_recursive how a file or directory should be treated.
23 * Values are ordered by significance, e.g. if a directory contains both
24 * excluded and untracked files, it is listed as untracked because
25 * path_untracked > path_excluded.
27 enum path_treatment {
28 path_none = 0,
29 path_recurse,
30 path_excluded,
31 path_untracked
34 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
35 const char *path, int len,
36 int check_only, const struct path_simplify *simplify);
37 static int get_dtype(struct dirent *de, const char *path, int len);
39 /* helper string functions with support for the ignore_case flag */
40 int strcmp_icase(const char *a, const char *b)
42 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
45 int strncmp_icase(const char *a, const char *b, size_t count)
47 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
50 int fnmatch_icase(const char *pattern, const char *string, int flags)
52 return fnmatch(pattern, string, flags | (ignore_case ? FNM_CASEFOLD : 0));
55 inline int git_fnmatch(const struct pathspec_item *item,
56 const char *pattern, const char *string,
57 int prefix)
59 if (prefix > 0) {
60 if (ps_strncmp(item, pattern, string, prefix))
61 return FNM_NOMATCH;
62 pattern += prefix;
63 string += prefix;
65 if (item->flags & PATHSPEC_ONESTAR) {
66 int pattern_len = strlen(++pattern);
67 int string_len = strlen(string);
68 return string_len < pattern_len ||
69 ps_strcmp(item, pattern,
70 string + string_len - pattern_len);
72 if (item->magic & PATHSPEC_GLOB)
73 return wildmatch(pattern, string,
74 WM_PATHNAME |
75 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
76 NULL);
77 else
78 /* wildmatch has not learned no FNM_PATHNAME mode yet */
79 return fnmatch(pattern, string,
80 item->magic & PATHSPEC_ICASE ? FNM_CASEFOLD : 0);
83 static int fnmatch_icase_mem(const char *pattern, int patternlen,
84 const char *string, int stringlen,
85 int flags)
87 int match_status;
88 struct strbuf pat_buf = STRBUF_INIT;
89 struct strbuf str_buf = STRBUF_INIT;
90 const char *use_pat = pattern;
91 const char *use_str = string;
93 if (pattern[patternlen]) {
94 strbuf_add(&pat_buf, pattern, patternlen);
95 use_pat = pat_buf.buf;
97 if (string[stringlen]) {
98 strbuf_add(&str_buf, string, stringlen);
99 use_str = str_buf.buf;
102 if (ignore_case)
103 flags |= WM_CASEFOLD;
104 match_status = wildmatch(use_pat, use_str, flags, NULL);
106 strbuf_release(&pat_buf);
107 strbuf_release(&str_buf);
109 return match_status;
112 static size_t common_prefix_len(const struct pathspec *pathspec)
114 int n;
115 size_t max = 0;
118 * ":(icase)path" is treated as a pathspec full of
119 * wildcard. In other words, only prefix is considered common
120 * prefix. If the pathspec is abc/foo abc/bar, running in
121 * subdir xyz, the common prefix is still xyz, not xuz/abc as
122 * in non-:(icase).
124 GUARD_PATHSPEC(pathspec,
125 PATHSPEC_FROMTOP |
126 PATHSPEC_MAXDEPTH |
127 PATHSPEC_LITERAL |
128 PATHSPEC_GLOB |
129 PATHSPEC_ICASE);
131 for (n = 0; n < pathspec->nr; n++) {
132 size_t i = 0, len = 0, item_len;
133 if (pathspec->items[n].magic & PATHSPEC_ICASE)
134 item_len = pathspec->items[n].prefix;
135 else
136 item_len = pathspec->items[n].nowildcard_len;
137 while (i < item_len && (n == 0 || i < max)) {
138 char c = pathspec->items[n].match[i];
139 if (c != pathspec->items[0].match[i])
140 break;
141 if (c == '/')
142 len = i + 1;
143 i++;
145 if (n == 0 || len < max) {
146 max = len;
147 if (!max)
148 break;
151 return max;
155 * Returns a copy of the longest leading path common among all
156 * pathspecs.
158 char *common_prefix(const struct pathspec *pathspec)
160 unsigned long len = common_prefix_len(pathspec);
162 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
165 int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
167 size_t len;
170 * Calculate common prefix for the pathspec, and
171 * use that to optimize the directory walk
173 len = common_prefix_len(pathspec);
175 /* Read the directory and prune it */
176 read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec);
177 return len;
180 int within_depth(const char *name, int namelen,
181 int depth, int max_depth)
183 const char *cp = name, *cpe = name + namelen;
185 while (cp < cpe) {
186 if (*cp++ != '/')
187 continue;
188 depth++;
189 if (depth > max_depth)
190 return 0;
192 return 1;
196 * Does 'match' match the given name?
197 * A match is found if
199 * (1) the 'match' string is leading directory of 'name', or
200 * (2) the 'match' string is a wildcard and matches 'name', or
201 * (3) the 'match' string is exactly the same as 'name'.
203 * and the return value tells which case it was.
205 * It returns 0 when there is no match.
207 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
208 const char *name, int namelen)
210 /* name/namelen has prefix cut off by caller */
211 const char *match = item->match + prefix;
212 int matchlen = item->len - prefix;
215 * The normal call pattern is:
216 * 1. prefix = common_prefix_len(ps);
217 * 2. prune something, or fill_directory
218 * 3. match_pathspec_depth()
220 * 'prefix' at #1 may be shorter than the command's prefix and
221 * it's ok for #2 to match extra files. Those extras will be
222 * trimmed at #3.
224 * Suppose the pathspec is 'foo' and '../bar' running from
225 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
226 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
227 * user does not want XYZ/foo, only the "foo" part should be
228 * case-insensitive. We need to filter out XYZ/foo here. In
229 * other words, we do not trust the caller on comparing the
230 * prefix part when :(icase) is involved. We do exact
231 * comparison ourselves.
233 * Normally the caller (common_prefix_len() in fact) does
234 * _exact_ matching on name[-prefix+1..-1] and we do not need
235 * to check that part. Be defensive and check it anyway, in
236 * case common_prefix_len is changed, or a new caller is
237 * introduced that does not use common_prefix_len.
239 * If the penalty turns out too high when prefix is really
240 * long, maybe change it to
241 * strncmp(match, name, item->prefix - prefix)
243 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
244 strncmp(item->match, name - prefix, item->prefix))
245 return 0;
247 /* If the match was just the prefix, we matched */
248 if (!*match)
249 return MATCHED_RECURSIVELY;
251 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
252 if (matchlen == namelen)
253 return MATCHED_EXACTLY;
255 if (match[matchlen-1] == '/' || name[matchlen] == '/')
256 return MATCHED_RECURSIVELY;
259 if (item->nowildcard_len < item->len &&
260 !git_fnmatch(item, match, name,
261 item->nowildcard_len - prefix))
262 return MATCHED_FNMATCH;
264 return 0;
268 * Given a name and a list of pathspecs, returns the nature of the
269 * closest (i.e. most specific) match of the name to any of the
270 * pathspecs.
272 * The caller typically calls this multiple times with the same
273 * pathspec and seen[] array but with different name/namelen
274 * (e.g. entries from the index) and is interested in seeing if and
275 * how each pathspec matches all the names it calls this function
276 * with. A mark is left in the seen[] array for each pathspec element
277 * indicating the closest type of match that element achieved, so if
278 * seen[n] remains zero after multiple invocations, that means the nth
279 * pathspec did not match any names, which could indicate that the
280 * user mistyped the nth pathspec.
282 int match_pathspec_depth(const struct pathspec *ps,
283 const char *name, int namelen,
284 int prefix, char *seen)
286 int i, retval = 0;
288 GUARD_PATHSPEC(ps,
289 PATHSPEC_FROMTOP |
290 PATHSPEC_MAXDEPTH |
291 PATHSPEC_LITERAL |
292 PATHSPEC_GLOB |
293 PATHSPEC_ICASE);
295 if (!ps->nr) {
296 if (!ps->recursive ||
297 !(ps->magic & PATHSPEC_MAXDEPTH) ||
298 ps->max_depth == -1)
299 return MATCHED_RECURSIVELY;
301 if (within_depth(name, namelen, 0, ps->max_depth))
302 return MATCHED_EXACTLY;
303 else
304 return 0;
307 name += prefix;
308 namelen -= prefix;
310 for (i = ps->nr - 1; i >= 0; i--) {
311 int how;
312 if (seen && seen[i] == MATCHED_EXACTLY)
313 continue;
314 how = match_pathspec_item(ps->items+i, prefix, name, namelen);
315 if (ps->recursive &&
316 (ps->magic & PATHSPEC_MAXDEPTH) &&
317 ps->max_depth != -1 &&
318 how && how != MATCHED_FNMATCH) {
319 int len = ps->items[i].len;
320 if (name[len] == '/')
321 len++;
322 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
323 how = MATCHED_EXACTLY;
324 else
325 how = 0;
327 if (how) {
328 if (retval < how)
329 retval = how;
330 if (seen && seen[i] < how)
331 seen[i] = how;
334 return retval;
338 * Return the length of the "simple" part of a path match limiter.
340 int simple_length(const char *match)
342 int len = -1;
344 for (;;) {
345 unsigned char c = *match++;
346 len++;
347 if (c == '\0' || is_glob_special(c))
348 return len;
352 int no_wildcard(const char *string)
354 return string[simple_length(string)] == '\0';
357 void parse_exclude_pattern(const char **pattern,
358 int *patternlen,
359 int *flags,
360 int *nowildcardlen)
362 const char *p = *pattern;
363 size_t i, len;
365 *flags = 0;
366 if (*p == '!') {
367 *flags |= EXC_FLAG_NEGATIVE;
368 p++;
370 len = strlen(p);
371 if (len && p[len - 1] == '/') {
372 len--;
373 *flags |= EXC_FLAG_MUSTBEDIR;
375 for (i = 0; i < len; i++) {
376 if (p[i] == '/')
377 break;
379 if (i == len)
380 *flags |= EXC_FLAG_NODIR;
381 *nowildcardlen = simple_length(p);
383 * we should have excluded the trailing slash from 'p' too,
384 * but that's one more allocation. Instead just make sure
385 * nowildcardlen does not exceed real patternlen
387 if (*nowildcardlen > len)
388 *nowildcardlen = len;
389 if (*p == '*' && no_wildcard(p + 1))
390 *flags |= EXC_FLAG_ENDSWITH;
391 *pattern = p;
392 *patternlen = len;
395 void add_exclude(const char *string, const char *base,
396 int baselen, struct exclude_list *el, int srcpos)
398 struct exclude *x;
399 int patternlen;
400 int flags;
401 int nowildcardlen;
403 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
404 if (flags & EXC_FLAG_MUSTBEDIR) {
405 char *s;
406 x = xmalloc(sizeof(*x) + patternlen + 1);
407 s = (char *)(x+1);
408 memcpy(s, string, patternlen);
409 s[patternlen] = '\0';
410 x->pattern = s;
411 } else {
412 x = xmalloc(sizeof(*x));
413 x->pattern = string;
415 x->patternlen = patternlen;
416 x->nowildcardlen = nowildcardlen;
417 x->base = base;
418 x->baselen = baselen;
419 x->flags = flags;
420 x->srcpos = srcpos;
421 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
422 el->excludes[el->nr++] = x;
423 x->el = el;
426 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
428 int pos, len;
429 unsigned long sz;
430 enum object_type type;
431 void *data;
433 len = strlen(path);
434 pos = cache_name_pos(path, len);
435 if (pos < 0)
436 return NULL;
437 if (!ce_skip_worktree(active_cache[pos]))
438 return NULL;
439 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
440 if (!data || type != OBJ_BLOB) {
441 free(data);
442 return NULL;
444 *size = xsize_t(sz);
445 return data;
449 * Frees memory within el which was allocated for exclude patterns and
450 * the file buffer. Does not free el itself.
452 void clear_exclude_list(struct exclude_list *el)
454 int i;
456 for (i = 0; i < el->nr; i++)
457 free(el->excludes[i]);
458 free(el->excludes);
459 free(el->filebuf);
461 el->nr = 0;
462 el->excludes = NULL;
463 el->filebuf = NULL;
466 int add_excludes_from_file_to_list(const char *fname,
467 const char *base,
468 int baselen,
469 struct exclude_list *el,
470 int check_index)
472 struct stat st;
473 int fd, i, lineno = 1;
474 size_t size = 0;
475 char *buf, *entry;
477 fd = open(fname, O_RDONLY);
478 if (fd < 0 || fstat(fd, &st) < 0) {
479 if (errno != ENOENT)
480 warn_on_inaccessible(fname);
481 if (0 <= fd)
482 close(fd);
483 if (!check_index ||
484 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
485 return -1;
486 if (size == 0) {
487 free(buf);
488 return 0;
490 if (buf[size-1] != '\n') {
491 buf = xrealloc(buf, size+1);
492 buf[size++] = '\n';
495 else {
496 size = xsize_t(st.st_size);
497 if (size == 0) {
498 close(fd);
499 return 0;
501 buf = xmalloc(size+1);
502 if (read_in_full(fd, buf, size) != size) {
503 free(buf);
504 close(fd);
505 return -1;
507 buf[size++] = '\n';
508 close(fd);
511 el->filebuf = buf;
512 entry = buf;
513 for (i = 0; i < size; i++) {
514 if (buf[i] == '\n') {
515 if (entry != buf + i && entry[0] != '#') {
516 buf[i - (i && buf[i-1] == '\r')] = 0;
517 add_exclude(entry, base, baselen, el, lineno);
519 lineno++;
520 entry = buf + i + 1;
523 return 0;
526 struct exclude_list *add_exclude_list(struct dir_struct *dir,
527 int group_type, const char *src)
529 struct exclude_list *el;
530 struct exclude_list_group *group;
532 group = &dir->exclude_list_group[group_type];
533 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
534 el = &group->el[group->nr++];
535 memset(el, 0, sizeof(*el));
536 el->src = src;
537 return el;
541 * Used to set up core.excludesfile and .git/info/exclude lists.
543 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
545 struct exclude_list *el;
546 el = add_exclude_list(dir, EXC_FILE, fname);
547 if (add_excludes_from_file_to_list(fname, "", 0, el, 0) < 0)
548 die("cannot use %s as an exclude file", fname);
551 int match_basename(const char *basename, int basenamelen,
552 const char *pattern, int prefix, int patternlen,
553 int flags)
555 if (prefix == patternlen) {
556 if (patternlen == basenamelen &&
557 !strncmp_icase(pattern, basename, basenamelen))
558 return 1;
559 } else if (flags & EXC_FLAG_ENDSWITH) {
560 /* "*literal" matching against "fooliteral" */
561 if (patternlen - 1 <= basenamelen &&
562 !strncmp_icase(pattern + 1,
563 basename + basenamelen - (patternlen - 1),
564 patternlen - 1))
565 return 1;
566 } else {
567 if (fnmatch_icase_mem(pattern, patternlen,
568 basename, basenamelen,
569 0) == 0)
570 return 1;
572 return 0;
575 int match_pathname(const char *pathname, int pathlen,
576 const char *base, int baselen,
577 const char *pattern, int prefix, int patternlen,
578 int flags)
580 const char *name;
581 int namelen;
584 * match with FNM_PATHNAME; the pattern has base implicitly
585 * in front of it.
587 if (*pattern == '/') {
588 pattern++;
589 patternlen--;
590 prefix--;
594 * baselen does not count the trailing slash. base[] may or
595 * may not end with a trailing slash though.
597 if (pathlen < baselen + 1 ||
598 (baselen && pathname[baselen] != '/') ||
599 strncmp_icase(pathname, base, baselen))
600 return 0;
602 namelen = baselen ? pathlen - baselen - 1 : pathlen;
603 name = pathname + pathlen - namelen;
605 if (prefix) {
607 * if the non-wildcard part is longer than the
608 * remaining pathname, surely it cannot match.
610 if (prefix > namelen)
611 return 0;
613 if (strncmp_icase(pattern, name, prefix))
614 return 0;
615 pattern += prefix;
616 patternlen -= prefix;
617 name += prefix;
618 namelen -= prefix;
621 * If the whole pattern did not have a wildcard,
622 * then our prefix match is all we need; we
623 * do not need to call fnmatch at all.
625 if (!patternlen && !namelen)
626 return 1;
629 return fnmatch_icase_mem(pattern, patternlen,
630 name, namelen,
631 WM_PATHNAME) == 0;
635 * Scan the given exclude list in reverse to see whether pathname
636 * should be ignored. The first match (i.e. the last on the list), if
637 * any, determines the fate. Returns the exclude_list element which
638 * matched, or NULL for undecided.
640 static struct exclude *last_exclude_matching_from_list(const char *pathname,
641 int pathlen,
642 const char *basename,
643 int *dtype,
644 struct exclude_list *el)
646 int i;
648 if (!el->nr)
649 return NULL; /* undefined */
651 for (i = el->nr - 1; 0 <= i; i--) {
652 struct exclude *x = el->excludes[i];
653 const char *exclude = x->pattern;
654 int prefix = x->nowildcardlen;
656 if (x->flags & EXC_FLAG_MUSTBEDIR) {
657 if (*dtype == DT_UNKNOWN)
658 *dtype = get_dtype(NULL, pathname, pathlen);
659 if (*dtype != DT_DIR)
660 continue;
663 if (x->flags & EXC_FLAG_NODIR) {
664 if (match_basename(basename,
665 pathlen - (basename - pathname),
666 exclude, prefix, x->patternlen,
667 x->flags))
668 return x;
669 continue;
672 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
673 if (match_pathname(pathname, pathlen,
674 x->base, x->baselen ? x->baselen - 1 : 0,
675 exclude, prefix, x->patternlen, x->flags))
676 return x;
678 return NULL; /* undecided */
682 * Scan the list and let the last match determine the fate.
683 * Return 1 for exclude, 0 for include and -1 for undecided.
685 int is_excluded_from_list(const char *pathname,
686 int pathlen, const char *basename, int *dtype,
687 struct exclude_list *el)
689 struct exclude *exclude;
690 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
691 if (exclude)
692 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
693 return -1; /* undecided */
696 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
697 const char *pathname, int pathlen, const char *basename,
698 int *dtype_p)
700 int i, j;
701 struct exclude_list_group *group;
702 struct exclude *exclude;
703 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
704 group = &dir->exclude_list_group[i];
705 for (j = group->nr - 1; j >= 0; j--) {
706 exclude = last_exclude_matching_from_list(
707 pathname, pathlen, basename, dtype_p,
708 &group->el[j]);
709 if (exclude)
710 return exclude;
713 return NULL;
717 * Loads the per-directory exclude list for the substring of base
718 * which has a char length of baselen.
720 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
722 struct exclude_list_group *group;
723 struct exclude_list *el;
724 struct exclude_stack *stk = NULL;
725 int current;
727 group = &dir->exclude_list_group[EXC_DIRS];
729 /* Pop the exclude lists from the EXCL_DIRS exclude_list_group
730 * which originate from directories not in the prefix of the
731 * path being checked. */
732 while ((stk = dir->exclude_stack) != NULL) {
733 if (stk->baselen <= baselen &&
734 !strncmp(dir->basebuf, base, stk->baselen))
735 break;
736 el = &group->el[dir->exclude_stack->exclude_ix];
737 dir->exclude_stack = stk->prev;
738 dir->exclude = NULL;
739 free((char *)el->src); /* see strdup() below */
740 clear_exclude_list(el);
741 free(stk);
742 group->nr--;
745 /* Skip traversing into sub directories if the parent is excluded */
746 if (dir->exclude)
747 return;
749 /* Read from the parent directories and push them down. */
750 current = stk ? stk->baselen : -1;
751 while (current < baselen) {
752 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
753 const char *cp;
755 if (current < 0) {
756 cp = base;
757 current = 0;
759 else {
760 cp = strchr(base + current + 1, '/');
761 if (!cp)
762 die("oops in prep_exclude");
763 cp++;
765 stk->prev = dir->exclude_stack;
766 stk->baselen = cp - base;
767 stk->exclude_ix = group->nr;
768 el = add_exclude_list(dir, EXC_DIRS, NULL);
769 memcpy(dir->basebuf + current, base + current,
770 stk->baselen - current);
772 /* Abort if the directory is excluded */
773 if (stk->baselen) {
774 int dt = DT_DIR;
775 dir->basebuf[stk->baselen - 1] = 0;
776 dir->exclude = last_exclude_matching_from_lists(dir,
777 dir->basebuf, stk->baselen - 1,
778 dir->basebuf + current, &dt);
779 dir->basebuf[stk->baselen - 1] = '/';
780 if (dir->exclude &&
781 dir->exclude->flags & EXC_FLAG_NEGATIVE)
782 dir->exclude = NULL;
783 if (dir->exclude) {
784 dir->basebuf[stk->baselen] = 0;
785 dir->exclude_stack = stk;
786 return;
790 /* Try to read per-directory file unless path is too long */
791 if (dir->exclude_per_dir &&
792 stk->baselen + strlen(dir->exclude_per_dir) < PATH_MAX) {
793 strcpy(dir->basebuf + stk->baselen,
794 dir->exclude_per_dir);
796 * dir->basebuf gets reused by the traversal, but we
797 * need fname to remain unchanged to ensure the src
798 * member of each struct exclude correctly
799 * back-references its source file. Other invocations
800 * of add_exclude_list provide stable strings, so we
801 * strdup() and free() here in the caller.
803 el->src = strdup(dir->basebuf);
804 add_excludes_from_file_to_list(dir->basebuf,
805 dir->basebuf, stk->baselen, el, 1);
807 dir->exclude_stack = stk;
808 current = stk->baselen;
810 dir->basebuf[baselen] = '\0';
814 * Loads the exclude lists for the directory containing pathname, then
815 * scans all exclude lists to determine whether pathname is excluded.
816 * Returns the exclude_list element which matched, or NULL for
817 * undecided.
819 struct exclude *last_exclude_matching(struct dir_struct *dir,
820 const char *pathname,
821 int *dtype_p)
823 int pathlen = strlen(pathname);
824 const char *basename = strrchr(pathname, '/');
825 basename = (basename) ? basename+1 : pathname;
827 prep_exclude(dir, pathname, basename-pathname);
829 if (dir->exclude)
830 return dir->exclude;
832 return last_exclude_matching_from_lists(dir, pathname, pathlen,
833 basename, dtype_p);
837 * Loads the exclude lists for the directory containing pathname, then
838 * scans all exclude lists to determine whether pathname is excluded.
839 * Returns 1 if true, otherwise 0.
841 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
843 struct exclude *exclude =
844 last_exclude_matching(dir, pathname, dtype_p);
845 if (exclude)
846 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
847 return 0;
850 static struct dir_entry *dir_entry_new(const char *pathname, int len)
852 struct dir_entry *ent;
854 ent = xmalloc(sizeof(*ent) + len + 1);
855 ent->len = len;
856 memcpy(ent->name, pathname, len);
857 ent->name[len] = 0;
858 return ent;
861 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
863 if (cache_name_exists(pathname, len, ignore_case))
864 return NULL;
866 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
867 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
870 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
872 if (!cache_name_is_other(pathname, len))
873 return NULL;
875 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
876 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
879 enum exist_status {
880 index_nonexistent = 0,
881 index_directory,
882 index_gitdir
886 * Do not use the alphabetically sorted index to look up
887 * the directory name; instead, use the case insensitive
888 * name hash.
890 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
892 const struct cache_entry *ce = cache_name_exists(dirname, len + 1, ignore_case);
893 unsigned char endchar;
895 if (!ce)
896 return index_nonexistent;
897 endchar = ce->name[len];
900 * The cache_entry structure returned will contain this dirname
901 * and possibly additional path components.
903 if (endchar == '/')
904 return index_directory;
907 * If there are no additional path components, then this cache_entry
908 * represents a submodule. Submodules, despite being directories,
909 * are stored in the cache without a closing slash.
911 if (!endchar && S_ISGITLINK(ce->ce_mode))
912 return index_gitdir;
914 /* This should never be hit, but it exists just in case. */
915 return index_nonexistent;
919 * The index sorts alphabetically by entry name, which
920 * means that a gitlink sorts as '\0' at the end, while
921 * a directory (which is defined not as an entry, but as
922 * the files it contains) will sort with the '/' at the
923 * end.
925 static enum exist_status directory_exists_in_index(const char *dirname, int len)
927 int pos;
929 if (ignore_case)
930 return directory_exists_in_index_icase(dirname, len);
932 pos = cache_name_pos(dirname, len);
933 if (pos < 0)
934 pos = -pos-1;
935 while (pos < active_nr) {
936 const struct cache_entry *ce = active_cache[pos++];
937 unsigned char endchar;
939 if (strncmp(ce->name, dirname, len))
940 break;
941 endchar = ce->name[len];
942 if (endchar > '/')
943 break;
944 if (endchar == '/')
945 return index_directory;
946 if (!endchar && S_ISGITLINK(ce->ce_mode))
947 return index_gitdir;
949 return index_nonexistent;
953 * When we find a directory when traversing the filesystem, we
954 * have three distinct cases:
956 * - ignore it
957 * - see it as a directory
958 * - recurse into it
960 * and which one we choose depends on a combination of existing
961 * git index contents and the flags passed into the directory
962 * traversal routine.
964 * Case 1: If we *already* have entries in the index under that
965 * directory name, we always recurse into the directory to see
966 * all the files.
968 * Case 2: If we *already* have that directory name as a gitlink,
969 * we always continue to see it as a gitlink, regardless of whether
970 * there is an actual git directory there or not (it might not
971 * be checked out as a subproject!)
973 * Case 3: if we didn't have it in the index previously, we
974 * have a few sub-cases:
976 * (a) if "show_other_directories" is true, we show it as
977 * just a directory, unless "hide_empty_directories" is
978 * also true, in which case we need to check if it contains any
979 * untracked and / or ignored files.
980 * (b) if it looks like a git directory, and we don't have
981 * 'no_gitlinks' set we treat it as a gitlink, and show it
982 * as a directory.
983 * (c) otherwise, we recurse into it.
985 static enum path_treatment treat_directory(struct dir_struct *dir,
986 const char *dirname, int len, int exclude,
987 const struct path_simplify *simplify)
989 /* The "len-1" is to strip the final '/' */
990 switch (directory_exists_in_index(dirname, len-1)) {
991 case index_directory:
992 return path_recurse;
994 case index_gitdir:
995 return path_none;
997 case index_nonexistent:
998 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
999 break;
1000 if (!(dir->flags & DIR_NO_GITLINKS)) {
1001 unsigned char sha1[20];
1002 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1003 return path_untracked;
1005 return path_recurse;
1008 /* This is the "show_other_directories" case */
1010 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1011 return exclude ? path_excluded : path_untracked;
1013 return read_directory_recursive(dir, dirname, len, 1, simplify);
1017 * This is an inexact early pruning of any recursive directory
1018 * reading - if the path cannot possibly be in the pathspec,
1019 * return true, and we'll skip it early.
1021 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1023 if (simplify) {
1024 for (;;) {
1025 const char *match = simplify->path;
1026 int len = simplify->len;
1028 if (!match)
1029 break;
1030 if (len > pathlen)
1031 len = pathlen;
1032 if (!memcmp(path, match, len))
1033 return 0;
1034 simplify++;
1036 return 1;
1038 return 0;
1042 * This function tells us whether an excluded path matches a
1043 * list of "interesting" pathspecs. That is, whether a path matched
1044 * by any of the pathspecs could possibly be ignored by excluding
1045 * the specified path. This can happen if:
1047 * 1. the path is mentioned explicitly in the pathspec
1049 * 2. the path is a directory prefix of some element in the
1050 * pathspec
1052 static int exclude_matches_pathspec(const char *path, int len,
1053 const struct path_simplify *simplify)
1055 if (simplify) {
1056 for (; simplify->path; simplify++) {
1057 if (len == simplify->len
1058 && !memcmp(path, simplify->path, len))
1059 return 1;
1060 if (len < simplify->len
1061 && simplify->path[len] == '/'
1062 && !memcmp(path, simplify->path, len))
1063 return 1;
1066 return 0;
1069 static int get_index_dtype(const char *path, int len)
1071 int pos;
1072 const struct cache_entry *ce;
1074 ce = cache_name_exists(path, len, 0);
1075 if (ce) {
1076 if (!ce_uptodate(ce))
1077 return DT_UNKNOWN;
1078 if (S_ISGITLINK(ce->ce_mode))
1079 return DT_DIR;
1081 * Nobody actually cares about the
1082 * difference between DT_LNK and DT_REG
1084 return DT_REG;
1087 /* Try to look it up as a directory */
1088 pos = cache_name_pos(path, len);
1089 if (pos >= 0)
1090 return DT_UNKNOWN;
1091 pos = -pos-1;
1092 while (pos < active_nr) {
1093 ce = active_cache[pos++];
1094 if (strncmp(ce->name, path, len))
1095 break;
1096 if (ce->name[len] > '/')
1097 break;
1098 if (ce->name[len] < '/')
1099 continue;
1100 if (!ce_uptodate(ce))
1101 break; /* continue? */
1102 return DT_DIR;
1104 return DT_UNKNOWN;
1107 static int get_dtype(struct dirent *de, const char *path, int len)
1109 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1110 struct stat st;
1112 if (dtype != DT_UNKNOWN)
1113 return dtype;
1114 dtype = get_index_dtype(path, len);
1115 if (dtype != DT_UNKNOWN)
1116 return dtype;
1117 if (lstat(path, &st))
1118 return dtype;
1119 if (S_ISREG(st.st_mode))
1120 return DT_REG;
1121 if (S_ISDIR(st.st_mode))
1122 return DT_DIR;
1123 if (S_ISLNK(st.st_mode))
1124 return DT_LNK;
1125 return dtype;
1128 static enum path_treatment treat_one_path(struct dir_struct *dir,
1129 struct strbuf *path,
1130 const struct path_simplify *simplify,
1131 int dtype, struct dirent *de)
1133 int exclude;
1134 int has_path_in_index = !!cache_name_exists(path->buf, path->len, ignore_case);
1136 if (dtype == DT_UNKNOWN)
1137 dtype = get_dtype(de, path->buf, path->len);
1139 /* Always exclude indexed files */
1140 if (dtype != DT_DIR && has_path_in_index)
1141 return path_none;
1144 * When we are looking at a directory P in the working tree,
1145 * there are three cases:
1147 * (1) P exists in the index. Everything inside the directory P in
1148 * the working tree needs to go when P is checked out from the
1149 * index.
1151 * (2) P does not exist in the index, but there is P/Q in the index.
1152 * We know P will stay a directory when we check out the contents
1153 * of the index, but we do not know yet if there is a directory
1154 * P/Q in the working tree to be killed, so we need to recurse.
1156 * (3) P does not exist in the index, and there is no P/Q in the index
1157 * to require P to be a directory, either. Only in this case, we
1158 * know that everything inside P will not be killed without
1159 * recursing.
1161 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1162 (dtype == DT_DIR) &&
1163 !has_path_in_index) {
1165 * NEEDSWORK: directory_exists_in_index_icase()
1166 * assumes that one byte past the given path is
1167 * readable and has '/', which needs to be fixed, but
1168 * until then, work it around in the caller.
1170 strbuf_addch(path, '/');
1171 if (directory_exists_in_index(path->buf, path->len - 1) ==
1172 index_nonexistent) {
1173 strbuf_setlen(path, path->len - 1);
1174 return path_none;
1176 strbuf_setlen(path, path->len - 1);
1179 exclude = is_excluded(dir, path->buf, &dtype);
1182 * Excluded? If we don't explicitly want to show
1183 * ignored files, ignore it
1185 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1186 return path_excluded;
1188 switch (dtype) {
1189 default:
1190 return path_none;
1191 case DT_DIR:
1192 strbuf_addch(path, '/');
1193 return treat_directory(dir, path->buf, path->len, exclude,
1194 simplify);
1195 case DT_REG:
1196 case DT_LNK:
1197 return exclude ? path_excluded : path_untracked;
1201 static enum path_treatment treat_path(struct dir_struct *dir,
1202 struct dirent *de,
1203 struct strbuf *path,
1204 int baselen,
1205 const struct path_simplify *simplify)
1207 int dtype;
1209 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1210 return path_none;
1211 strbuf_setlen(path, baselen);
1212 strbuf_addstr(path, de->d_name);
1213 if (simplify_away(path->buf, path->len, simplify))
1214 return path_none;
1216 dtype = DTYPE(de);
1217 return treat_one_path(dir, path, simplify, dtype, de);
1221 * Read a directory tree. We currently ignore anything but
1222 * directories, regular files and symlinks. That's because git
1223 * doesn't handle them at all yet. Maybe that will change some
1224 * day.
1226 * Also, we ignore the name ".git" (even if it is not a directory).
1227 * That likely will not change.
1229 * Returns the most significant path_treatment value encountered in the scan.
1231 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1232 const char *base, int baselen,
1233 int check_only,
1234 const struct path_simplify *simplify)
1236 DIR *fdir;
1237 enum path_treatment state, subdir_state, dir_state = path_none;
1238 struct dirent *de;
1239 struct strbuf path = STRBUF_INIT;
1241 strbuf_add(&path, base, baselen);
1243 fdir = opendir(path.len ? path.buf : ".");
1244 if (!fdir)
1245 goto out;
1247 while ((de = readdir(fdir)) != NULL) {
1248 /* check how the file or directory should be treated */
1249 state = treat_path(dir, de, &path, baselen, simplify);
1250 if (state > dir_state)
1251 dir_state = state;
1253 /* recurse into subdir if instructed by treat_path */
1254 if (state == path_recurse) {
1255 subdir_state = read_directory_recursive(dir, path.buf,
1256 path.len, check_only, simplify);
1257 if (subdir_state > dir_state)
1258 dir_state = subdir_state;
1261 if (check_only) {
1262 /* abort early if maximum state has been reached */
1263 if (dir_state == path_untracked)
1264 break;
1265 /* skip the dir_add_* part */
1266 continue;
1269 /* add the path to the appropriate result list */
1270 switch (state) {
1271 case path_excluded:
1272 if (dir->flags & DIR_SHOW_IGNORED)
1273 dir_add_name(dir, path.buf, path.len);
1274 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1275 ((dir->flags & DIR_COLLECT_IGNORED) &&
1276 exclude_matches_pathspec(path.buf, path.len,
1277 simplify)))
1278 dir_add_ignored(dir, path.buf, path.len);
1279 break;
1281 case path_untracked:
1282 if (!(dir->flags & DIR_SHOW_IGNORED))
1283 dir_add_name(dir, path.buf, path.len);
1284 break;
1286 default:
1287 break;
1290 closedir(fdir);
1291 out:
1292 strbuf_release(&path);
1294 return dir_state;
1297 static int cmp_name(const void *p1, const void *p2)
1299 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1300 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1302 return cache_name_compare(e1->name, e1->len,
1303 e2->name, e2->len);
1306 static struct path_simplify *create_simplify(const char **pathspec)
1308 int nr, alloc = 0;
1309 struct path_simplify *simplify = NULL;
1311 if (!pathspec)
1312 return NULL;
1314 for (nr = 0 ; ; nr++) {
1315 const char *match;
1316 if (nr >= alloc) {
1317 alloc = alloc_nr(alloc);
1318 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
1320 match = *pathspec++;
1321 if (!match)
1322 break;
1323 simplify[nr].path = match;
1324 simplify[nr].len = simple_length(match);
1326 simplify[nr].path = NULL;
1327 simplify[nr].len = 0;
1328 return simplify;
1331 static void free_simplify(struct path_simplify *simplify)
1333 free(simplify);
1336 static int treat_leading_path(struct dir_struct *dir,
1337 const char *path, int len,
1338 const struct path_simplify *simplify)
1340 struct strbuf sb = STRBUF_INIT;
1341 int baselen, rc = 0;
1342 const char *cp;
1343 int old_flags = dir->flags;
1345 while (len && path[len - 1] == '/')
1346 len--;
1347 if (!len)
1348 return 1;
1349 baselen = 0;
1350 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1351 while (1) {
1352 cp = path + baselen + !!baselen;
1353 cp = memchr(cp, '/', path + len - cp);
1354 if (!cp)
1355 baselen = len;
1356 else
1357 baselen = cp - path;
1358 strbuf_setlen(&sb, 0);
1359 strbuf_add(&sb, path, baselen);
1360 if (!is_directory(sb.buf))
1361 break;
1362 if (simplify_away(sb.buf, sb.len, simplify))
1363 break;
1364 if (treat_one_path(dir, &sb, simplify,
1365 DT_DIR, NULL) == path_none)
1366 break; /* do not recurse into it */
1367 if (len <= baselen) {
1368 rc = 1;
1369 break; /* finished checking */
1372 strbuf_release(&sb);
1373 dir->flags = old_flags;
1374 return rc;
1377 int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1379 struct path_simplify *simplify;
1382 * Check out create_simplify()
1384 if (pathspec)
1385 GUARD_PATHSPEC(pathspec,
1386 PATHSPEC_FROMTOP |
1387 PATHSPEC_MAXDEPTH |
1388 PATHSPEC_LITERAL |
1389 PATHSPEC_GLOB |
1390 PATHSPEC_ICASE);
1392 if (has_symlink_leading_path(path, len))
1393 return dir->nr;
1395 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
1396 if (!len || treat_leading_path(dir, path, len, simplify))
1397 read_directory_recursive(dir, path, len, 0, simplify);
1398 free_simplify(simplify);
1399 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1400 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1401 return dir->nr;
1404 int file_exists(const char *f)
1406 struct stat sb;
1407 return lstat(f, &sb) == 0;
1411 * Given two normalized paths (a trailing slash is ok), if subdir is
1412 * outside dir, return -1. Otherwise return the offset in subdir that
1413 * can be used as relative path to dir.
1415 int dir_inside_of(const char *subdir, const char *dir)
1417 int offset = 0;
1419 assert(dir && subdir && *dir && *subdir);
1421 while (*dir && *subdir && *dir == *subdir) {
1422 dir++;
1423 subdir++;
1424 offset++;
1427 /* hel[p]/me vs hel[l]/yeah */
1428 if (*dir && *subdir)
1429 return -1;
1431 if (!*subdir)
1432 return !*dir ? offset : -1; /* same dir */
1434 /* foo/[b]ar vs foo/[] */
1435 if (is_dir_sep(dir[-1]))
1436 return is_dir_sep(subdir[-1]) ? offset : -1;
1438 /* foo[/]bar vs foo[] */
1439 return is_dir_sep(*subdir) ? offset + 1 : -1;
1442 int is_inside_dir(const char *dir)
1444 char cwd[PATH_MAX];
1445 if (!dir)
1446 return 0;
1447 if (!getcwd(cwd, sizeof(cwd)))
1448 die_errno("can't find the current directory");
1449 return dir_inside_of(cwd, dir) >= 0;
1452 int is_empty_dir(const char *path)
1454 DIR *dir = opendir(path);
1455 struct dirent *e;
1456 int ret = 1;
1458 if (!dir)
1459 return 0;
1461 while ((e = readdir(dir)) != NULL)
1462 if (!is_dot_or_dotdot(e->d_name)) {
1463 ret = 0;
1464 break;
1467 closedir(dir);
1468 return ret;
1471 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1473 DIR *dir;
1474 struct dirent *e;
1475 int ret = 0, original_len = path->len, len, kept_down = 0;
1476 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1477 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1478 unsigned char submodule_head[20];
1480 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1481 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1482 /* Do not descend and nuke a nested git work tree. */
1483 if (kept_up)
1484 *kept_up = 1;
1485 return 0;
1488 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1489 dir = opendir(path->buf);
1490 if (!dir) {
1491 /* an empty dir could be removed even if it is unreadble */
1492 if (!keep_toplevel)
1493 return rmdir(path->buf);
1494 else
1495 return -1;
1497 if (path->buf[original_len - 1] != '/')
1498 strbuf_addch(path, '/');
1500 len = path->len;
1501 while ((e = readdir(dir)) != NULL) {
1502 struct stat st;
1503 if (is_dot_or_dotdot(e->d_name))
1504 continue;
1506 strbuf_setlen(path, len);
1507 strbuf_addstr(path, e->d_name);
1508 if (lstat(path->buf, &st))
1509 ; /* fall thru */
1510 else if (S_ISDIR(st.st_mode)) {
1511 if (!remove_dir_recurse(path, flag, &kept_down))
1512 continue; /* happy */
1513 } else if (!only_empty && !unlink(path->buf))
1514 continue; /* happy, too */
1516 /* path too long, stat fails, or non-directory still exists */
1517 ret = -1;
1518 break;
1520 closedir(dir);
1522 strbuf_setlen(path, original_len);
1523 if (!ret && !keep_toplevel && !kept_down)
1524 ret = rmdir(path->buf);
1525 else if (kept_up)
1527 * report the uplevel that it is not an error that we
1528 * did not rmdir() our directory.
1530 *kept_up = !ret;
1531 return ret;
1534 int remove_dir_recursively(struct strbuf *path, int flag)
1536 return remove_dir_recurse(path, flag, NULL);
1539 void setup_standard_excludes(struct dir_struct *dir)
1541 const char *path;
1542 char *xdg_path;
1544 dir->exclude_per_dir = ".gitignore";
1545 path = git_path("info/exclude");
1546 if (!excludes_file) {
1547 home_config_paths(NULL, &xdg_path, "ignore");
1548 excludes_file = xdg_path;
1550 if (!access_or_warn(path, R_OK, 0))
1551 add_excludes_from_file(dir, path);
1552 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
1553 add_excludes_from_file(dir, excludes_file);
1556 int remove_path(const char *name)
1558 char *slash;
1560 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
1561 return -1;
1563 slash = strrchr(name, '/');
1564 if (slash) {
1565 char *dirs = xstrdup(name);
1566 slash = dirs + (slash - name);
1567 do {
1568 *slash = '\0';
1569 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1570 free(dirs);
1572 return 0;
1576 * Frees memory within dir which was allocated for exclude lists and
1577 * the exclude_stack. Does not free dir itself.
1579 void clear_directory(struct dir_struct *dir)
1581 int i, j;
1582 struct exclude_list_group *group;
1583 struct exclude_list *el;
1584 struct exclude_stack *stk;
1586 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1587 group = &dir->exclude_list_group[i];
1588 for (j = 0; j < group->nr; j++) {
1589 el = &group->el[j];
1590 if (i == EXC_DIRS)
1591 free((char *)el->src);
1592 clear_exclude_list(el);
1594 free(group->el);
1597 stk = dir->exclude_stack;
1598 while (stk) {
1599 struct exclude_stack *prev = stk->prev;
1600 free(stk);
1601 stk = prev;