Revert "test-wildmatch: add "perf" command to compare wildmatch and fnmatch"
[git/raj.git] / dir.c
blob59ccb5218ad27f35093b1f9abced9bf8a7dd9374
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 wildmatch(pattern, string,
53 flags | (ignore_case ? WM_CASEFOLD : 0),
54 NULL);
57 inline int git_fnmatch(const struct pathspec_item *item,
58 const char *pattern, const char *string,
59 int prefix)
61 if (prefix > 0) {
62 if (ps_strncmp(item, pattern, string, prefix))
63 return WM_NOMATCH;
64 pattern += prefix;
65 string += prefix;
67 if (item->flags & PATHSPEC_ONESTAR) {
68 int pattern_len = strlen(++pattern);
69 int string_len = strlen(string);
70 return string_len < pattern_len ||
71 ps_strcmp(item, pattern,
72 string + string_len - pattern_len);
74 if (item->magic & PATHSPEC_GLOB)
75 return wildmatch(pattern, string,
76 WM_PATHNAME |
77 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
78 NULL);
79 else
80 /* wildmatch has not learned no FNM_PATHNAME mode yet */
81 return wildmatch(pattern, string,
82 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
83 NULL);
86 static int fnmatch_icase_mem(const char *pattern, int patternlen,
87 const char *string, int stringlen,
88 int flags)
90 int match_status;
91 struct strbuf pat_buf = STRBUF_INIT;
92 struct strbuf str_buf = STRBUF_INIT;
93 const char *use_pat = pattern;
94 const char *use_str = string;
96 if (pattern[patternlen]) {
97 strbuf_add(&pat_buf, pattern, patternlen);
98 use_pat = pat_buf.buf;
100 if (string[stringlen]) {
101 strbuf_add(&str_buf, string, stringlen);
102 use_str = str_buf.buf;
105 if (ignore_case)
106 flags |= WM_CASEFOLD;
107 match_status = wildmatch(use_pat, use_str, flags, NULL);
109 strbuf_release(&pat_buf);
110 strbuf_release(&str_buf);
112 return match_status;
115 static size_t common_prefix_len(const struct pathspec *pathspec)
117 int n;
118 size_t max = 0;
121 * ":(icase)path" is treated as a pathspec full of
122 * wildcard. In other words, only prefix is considered common
123 * prefix. If the pathspec is abc/foo abc/bar, running in
124 * subdir xyz, the common prefix is still xyz, not xuz/abc as
125 * in non-:(icase).
127 GUARD_PATHSPEC(pathspec,
128 PATHSPEC_FROMTOP |
129 PATHSPEC_MAXDEPTH |
130 PATHSPEC_LITERAL |
131 PATHSPEC_GLOB |
132 PATHSPEC_ICASE |
133 PATHSPEC_EXCLUDE);
135 for (n = 0; n < pathspec->nr; n++) {
136 size_t i = 0, len = 0, item_len;
137 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
138 continue;
139 if (pathspec->items[n].magic & PATHSPEC_ICASE)
140 item_len = pathspec->items[n].prefix;
141 else
142 item_len = pathspec->items[n].nowildcard_len;
143 while (i < item_len && (n == 0 || i < max)) {
144 char c = pathspec->items[n].match[i];
145 if (c != pathspec->items[0].match[i])
146 break;
147 if (c == '/')
148 len = i + 1;
149 i++;
151 if (n == 0 || len < max) {
152 max = len;
153 if (!max)
154 break;
157 return max;
161 * Returns a copy of the longest leading path common among all
162 * pathspecs.
164 char *common_prefix(const struct pathspec *pathspec)
166 unsigned long len = common_prefix_len(pathspec);
168 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
171 int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
173 size_t len;
176 * Calculate common prefix for the pathspec, and
177 * use that to optimize the directory walk
179 len = common_prefix_len(pathspec);
181 /* Read the directory and prune it */
182 read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec);
183 return len;
186 int within_depth(const char *name, int namelen,
187 int depth, int max_depth)
189 const char *cp = name, *cpe = name + namelen;
191 while (cp < cpe) {
192 if (*cp++ != '/')
193 continue;
194 depth++;
195 if (depth > max_depth)
196 return 0;
198 return 1;
202 * Does 'match' match the given name?
203 * A match is found if
205 * (1) the 'match' string is leading directory of 'name', or
206 * (2) the 'match' string is a wildcard and matches 'name', or
207 * (3) the 'match' string is exactly the same as 'name'.
209 * and the return value tells which case it was.
211 * It returns 0 when there is no match.
213 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
214 const char *name, int namelen)
216 /* name/namelen has prefix cut off by caller */
217 const char *match = item->match + prefix;
218 int matchlen = item->len - prefix;
221 * The normal call pattern is:
222 * 1. prefix = common_prefix_len(ps);
223 * 2. prune something, or fill_directory
224 * 3. match_pathspec_depth()
226 * 'prefix' at #1 may be shorter than the command's prefix and
227 * it's ok for #2 to match extra files. Those extras will be
228 * trimmed at #3.
230 * Suppose the pathspec is 'foo' and '../bar' running from
231 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
232 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
233 * user does not want XYZ/foo, only the "foo" part should be
234 * case-insensitive. We need to filter out XYZ/foo here. In
235 * other words, we do not trust the caller on comparing the
236 * prefix part when :(icase) is involved. We do exact
237 * comparison ourselves.
239 * Normally the caller (common_prefix_len() in fact) does
240 * _exact_ matching on name[-prefix+1..-1] and we do not need
241 * to check that part. Be defensive and check it anyway, in
242 * case common_prefix_len is changed, or a new caller is
243 * introduced that does not use common_prefix_len.
245 * If the penalty turns out too high when prefix is really
246 * long, maybe change it to
247 * strncmp(match, name, item->prefix - prefix)
249 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
250 strncmp(item->match, name - prefix, item->prefix))
251 return 0;
253 /* If the match was just the prefix, we matched */
254 if (!*match)
255 return MATCHED_RECURSIVELY;
257 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
258 if (matchlen == namelen)
259 return MATCHED_EXACTLY;
261 if (match[matchlen-1] == '/' || name[matchlen] == '/')
262 return MATCHED_RECURSIVELY;
265 if (item->nowildcard_len < item->len &&
266 !git_fnmatch(item, match, name,
267 item->nowildcard_len - prefix))
268 return MATCHED_FNMATCH;
270 return 0;
274 * Given a name and a list of pathspecs, returns the nature of the
275 * closest (i.e. most specific) match of the name to any of the
276 * pathspecs.
278 * The caller typically calls this multiple times with the same
279 * pathspec and seen[] array but with different name/namelen
280 * (e.g. entries from the index) and is interested in seeing if and
281 * how each pathspec matches all the names it calls this function
282 * with. A mark is left in the seen[] array for each pathspec element
283 * indicating the closest type of match that element achieved, so if
284 * seen[n] remains zero after multiple invocations, that means the nth
285 * pathspec did not match any names, which could indicate that the
286 * user mistyped the nth pathspec.
288 static int match_pathspec_depth_1(const struct pathspec *ps,
289 const char *name, int namelen,
290 int prefix, char *seen,
291 int exclude)
293 int i, retval = 0;
295 GUARD_PATHSPEC(ps,
296 PATHSPEC_FROMTOP |
297 PATHSPEC_MAXDEPTH |
298 PATHSPEC_LITERAL |
299 PATHSPEC_GLOB |
300 PATHSPEC_ICASE |
301 PATHSPEC_EXCLUDE);
303 if (!ps->nr) {
304 if (!ps->recursive ||
305 !(ps->magic & PATHSPEC_MAXDEPTH) ||
306 ps->max_depth == -1)
307 return MATCHED_RECURSIVELY;
309 if (within_depth(name, namelen, 0, ps->max_depth))
310 return MATCHED_EXACTLY;
311 else
312 return 0;
315 name += prefix;
316 namelen -= prefix;
318 for (i = ps->nr - 1; i >= 0; i--) {
319 int how;
321 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
322 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
323 continue;
325 if (seen && seen[i] == MATCHED_EXACTLY)
326 continue;
328 * Make exclude patterns optional and never report
329 * "pathspec ':(exclude)foo' matches no files"
331 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
332 seen[i] = MATCHED_FNMATCH;
333 how = match_pathspec_item(ps->items+i, prefix, name, namelen);
334 if (ps->recursive &&
335 (ps->magic & PATHSPEC_MAXDEPTH) &&
336 ps->max_depth != -1 &&
337 how && how != MATCHED_FNMATCH) {
338 int len = ps->items[i].len;
339 if (name[len] == '/')
340 len++;
341 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
342 how = MATCHED_EXACTLY;
343 else
344 how = 0;
346 if (how) {
347 if (retval < how)
348 retval = how;
349 if (seen && seen[i] < how)
350 seen[i] = how;
353 return retval;
356 int match_pathspec_depth(const struct pathspec *ps,
357 const char *name, int namelen,
358 int prefix, char *seen)
360 int positive, negative;
361 positive = match_pathspec_depth_1(ps, name, namelen, prefix, seen, 0);
362 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
363 return positive;
364 negative = match_pathspec_depth_1(ps, name, namelen, prefix, seen, 1);
365 return negative ? 0 : positive;
369 * Return the length of the "simple" part of a path match limiter.
371 int simple_length(const char *match)
373 int len = -1;
375 for (;;) {
376 unsigned char c = *match++;
377 len++;
378 if (c == '\0' || is_glob_special(c))
379 return len;
383 int no_wildcard(const char *string)
385 return string[simple_length(string)] == '\0';
388 void parse_exclude_pattern(const char **pattern,
389 int *patternlen,
390 int *flags,
391 int *nowildcardlen)
393 const char *p = *pattern;
394 size_t i, len;
396 *flags = 0;
397 if (*p == '!') {
398 *flags |= EXC_FLAG_NEGATIVE;
399 p++;
401 len = strlen(p);
402 if (len && p[len - 1] == '/') {
403 len--;
404 *flags |= EXC_FLAG_MUSTBEDIR;
406 for (i = 0; i < len; i++) {
407 if (p[i] == '/')
408 break;
410 if (i == len)
411 *flags |= EXC_FLAG_NODIR;
412 *nowildcardlen = simple_length(p);
414 * we should have excluded the trailing slash from 'p' too,
415 * but that's one more allocation. Instead just make sure
416 * nowildcardlen does not exceed real patternlen
418 if (*nowildcardlen > len)
419 *nowildcardlen = len;
420 if (*p == '*' && no_wildcard(p + 1))
421 *flags |= EXC_FLAG_ENDSWITH;
422 *pattern = p;
423 *patternlen = len;
426 void add_exclude(const char *string, const char *base,
427 int baselen, struct exclude_list *el, int srcpos)
429 struct exclude *x;
430 int patternlen;
431 int flags;
432 int nowildcardlen;
434 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
435 if (flags & EXC_FLAG_MUSTBEDIR) {
436 char *s;
437 x = xmalloc(sizeof(*x) + patternlen + 1);
438 s = (char *)(x+1);
439 memcpy(s, string, patternlen);
440 s[patternlen] = '\0';
441 x->pattern = s;
442 } else {
443 x = xmalloc(sizeof(*x));
444 x->pattern = string;
446 x->patternlen = patternlen;
447 x->nowildcardlen = nowildcardlen;
448 x->base = base;
449 x->baselen = baselen;
450 x->flags = flags;
451 x->srcpos = srcpos;
452 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
453 el->excludes[el->nr++] = x;
454 x->el = el;
457 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
459 int pos, len;
460 unsigned long sz;
461 enum object_type type;
462 void *data;
464 len = strlen(path);
465 pos = cache_name_pos(path, len);
466 if (pos < 0)
467 return NULL;
468 if (!ce_skip_worktree(active_cache[pos]))
469 return NULL;
470 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
471 if (!data || type != OBJ_BLOB) {
472 free(data);
473 return NULL;
475 *size = xsize_t(sz);
476 return data;
480 * Frees memory within el which was allocated for exclude patterns and
481 * the file buffer. Does not free el itself.
483 void clear_exclude_list(struct exclude_list *el)
485 int i;
487 for (i = 0; i < el->nr; i++)
488 free(el->excludes[i]);
489 free(el->excludes);
490 free(el->filebuf);
492 el->nr = 0;
493 el->excludes = NULL;
494 el->filebuf = NULL;
497 int add_excludes_from_file_to_list(const char *fname,
498 const char *base,
499 int baselen,
500 struct exclude_list *el,
501 int check_index)
503 struct stat st;
504 int fd, i, lineno = 1;
505 size_t size = 0;
506 char *buf, *entry;
508 fd = open(fname, O_RDONLY);
509 if (fd < 0 || fstat(fd, &st) < 0) {
510 if (errno != ENOENT)
511 warn_on_inaccessible(fname);
512 if (0 <= fd)
513 close(fd);
514 if (!check_index ||
515 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
516 return -1;
517 if (size == 0) {
518 free(buf);
519 return 0;
521 if (buf[size-1] != '\n') {
522 buf = xrealloc(buf, size+1);
523 buf[size++] = '\n';
526 else {
527 size = xsize_t(st.st_size);
528 if (size == 0) {
529 close(fd);
530 return 0;
532 buf = xmalloc(size+1);
533 if (read_in_full(fd, buf, size) != size) {
534 free(buf);
535 close(fd);
536 return -1;
538 buf[size++] = '\n';
539 close(fd);
542 el->filebuf = buf;
543 entry = buf;
544 for (i = 0; i < size; i++) {
545 if (buf[i] == '\n') {
546 if (entry != buf + i && entry[0] != '#') {
547 buf[i - (i && buf[i-1] == '\r')] = 0;
548 add_exclude(entry, base, baselen, el, lineno);
550 lineno++;
551 entry = buf + i + 1;
554 return 0;
557 struct exclude_list *add_exclude_list(struct dir_struct *dir,
558 int group_type, const char *src)
560 struct exclude_list *el;
561 struct exclude_list_group *group;
563 group = &dir->exclude_list_group[group_type];
564 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
565 el = &group->el[group->nr++];
566 memset(el, 0, sizeof(*el));
567 el->src = src;
568 return el;
572 * Used to set up core.excludesfile and .git/info/exclude lists.
574 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
576 struct exclude_list *el;
577 el = add_exclude_list(dir, EXC_FILE, fname);
578 if (add_excludes_from_file_to_list(fname, "", 0, el, 0) < 0)
579 die("cannot use %s as an exclude file", fname);
582 int match_basename(const char *basename, int basenamelen,
583 const char *pattern, int prefix, int patternlen,
584 int flags)
586 if (prefix == patternlen) {
587 if (patternlen == basenamelen &&
588 !strncmp_icase(pattern, basename, basenamelen))
589 return 1;
590 } else if (flags & EXC_FLAG_ENDSWITH) {
591 /* "*literal" matching against "fooliteral" */
592 if (patternlen - 1 <= basenamelen &&
593 !strncmp_icase(pattern + 1,
594 basename + basenamelen - (patternlen - 1),
595 patternlen - 1))
596 return 1;
597 } else {
598 if (fnmatch_icase_mem(pattern, patternlen,
599 basename, basenamelen,
600 0) == 0)
601 return 1;
603 return 0;
606 int match_pathname(const char *pathname, int pathlen,
607 const char *base, int baselen,
608 const char *pattern, int prefix, int patternlen,
609 int flags)
611 const char *name;
612 int namelen;
615 * match with FNM_PATHNAME; the pattern has base implicitly
616 * in front of it.
618 if (*pattern == '/') {
619 pattern++;
620 patternlen--;
621 prefix--;
625 * baselen does not count the trailing slash. base[] may or
626 * may not end with a trailing slash though.
628 if (pathlen < baselen + 1 ||
629 (baselen && pathname[baselen] != '/') ||
630 strncmp_icase(pathname, base, baselen))
631 return 0;
633 namelen = baselen ? pathlen - baselen - 1 : pathlen;
634 name = pathname + pathlen - namelen;
636 if (prefix) {
638 * if the non-wildcard part is longer than the
639 * remaining pathname, surely it cannot match.
641 if (prefix > namelen)
642 return 0;
644 if (strncmp_icase(pattern, name, prefix))
645 return 0;
646 pattern += prefix;
647 patternlen -= prefix;
648 name += prefix;
649 namelen -= prefix;
652 * If the whole pattern did not have a wildcard,
653 * then our prefix match is all we need; we
654 * do not need to call fnmatch at all.
656 if (!patternlen && !namelen)
657 return 1;
660 return fnmatch_icase_mem(pattern, patternlen,
661 name, namelen,
662 WM_PATHNAME) == 0;
666 * Scan the given exclude list in reverse to see whether pathname
667 * should be ignored. The first match (i.e. the last on the list), if
668 * any, determines the fate. Returns the exclude_list element which
669 * matched, or NULL for undecided.
671 static struct exclude *last_exclude_matching_from_list(const char *pathname,
672 int pathlen,
673 const char *basename,
674 int *dtype,
675 struct exclude_list *el)
677 int i;
679 if (!el->nr)
680 return NULL; /* undefined */
682 for (i = el->nr - 1; 0 <= i; i--) {
683 struct exclude *x = el->excludes[i];
684 const char *exclude = x->pattern;
685 int prefix = x->nowildcardlen;
687 if (x->flags & EXC_FLAG_MUSTBEDIR) {
688 if (*dtype == DT_UNKNOWN)
689 *dtype = get_dtype(NULL, pathname, pathlen);
690 if (*dtype != DT_DIR)
691 continue;
694 if (x->flags & EXC_FLAG_NODIR) {
695 if (match_basename(basename,
696 pathlen - (basename - pathname),
697 exclude, prefix, x->patternlen,
698 x->flags))
699 return x;
700 continue;
703 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
704 if (match_pathname(pathname, pathlen,
705 x->base, x->baselen ? x->baselen - 1 : 0,
706 exclude, prefix, x->patternlen, x->flags))
707 return x;
709 return NULL; /* undecided */
713 * Scan the list and let the last match determine the fate.
714 * Return 1 for exclude, 0 for include and -1 for undecided.
716 int is_excluded_from_list(const char *pathname,
717 int pathlen, const char *basename, int *dtype,
718 struct exclude_list *el)
720 struct exclude *exclude;
721 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
722 if (exclude)
723 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
724 return -1; /* undecided */
727 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
728 const char *pathname, int pathlen, const char *basename,
729 int *dtype_p)
731 int i, j;
732 struct exclude_list_group *group;
733 struct exclude *exclude;
734 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
735 group = &dir->exclude_list_group[i];
736 for (j = group->nr - 1; j >= 0; j--) {
737 exclude = last_exclude_matching_from_list(
738 pathname, pathlen, basename, dtype_p,
739 &group->el[j]);
740 if (exclude)
741 return exclude;
744 return NULL;
748 * Loads the per-directory exclude list for the substring of base
749 * which has a char length of baselen.
751 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
753 struct exclude_list_group *group;
754 struct exclude_list *el;
755 struct exclude_stack *stk = NULL;
756 int current;
758 group = &dir->exclude_list_group[EXC_DIRS];
760 /* Pop the exclude lists from the EXCL_DIRS exclude_list_group
761 * which originate from directories not in the prefix of the
762 * path being checked. */
763 while ((stk = dir->exclude_stack) != NULL) {
764 if (stk->baselen <= baselen &&
765 !strncmp(dir->basebuf, base, stk->baselen))
766 break;
767 el = &group->el[dir->exclude_stack->exclude_ix];
768 dir->exclude_stack = stk->prev;
769 dir->exclude = NULL;
770 free((char *)el->src); /* see strdup() below */
771 clear_exclude_list(el);
772 free(stk);
773 group->nr--;
776 /* Skip traversing into sub directories if the parent is excluded */
777 if (dir->exclude)
778 return;
780 /* Read from the parent directories and push them down. */
781 current = stk ? stk->baselen : -1;
782 while (current < baselen) {
783 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
784 const char *cp;
786 if (current < 0) {
787 cp = base;
788 current = 0;
790 else {
791 cp = strchr(base + current + 1, '/');
792 if (!cp)
793 die("oops in prep_exclude");
794 cp++;
796 stk->prev = dir->exclude_stack;
797 stk->baselen = cp - base;
798 stk->exclude_ix = group->nr;
799 el = add_exclude_list(dir, EXC_DIRS, NULL);
800 memcpy(dir->basebuf + current, base + current,
801 stk->baselen - current);
803 /* Abort if the directory is excluded */
804 if (stk->baselen) {
805 int dt = DT_DIR;
806 dir->basebuf[stk->baselen - 1] = 0;
807 dir->exclude = last_exclude_matching_from_lists(dir,
808 dir->basebuf, stk->baselen - 1,
809 dir->basebuf + current, &dt);
810 dir->basebuf[stk->baselen - 1] = '/';
811 if (dir->exclude &&
812 dir->exclude->flags & EXC_FLAG_NEGATIVE)
813 dir->exclude = NULL;
814 if (dir->exclude) {
815 dir->basebuf[stk->baselen] = 0;
816 dir->exclude_stack = stk;
817 return;
821 /* Try to read per-directory file unless path is too long */
822 if (dir->exclude_per_dir &&
823 stk->baselen + strlen(dir->exclude_per_dir) < PATH_MAX) {
824 strcpy(dir->basebuf + stk->baselen,
825 dir->exclude_per_dir);
827 * dir->basebuf gets reused by the traversal, but we
828 * need fname to remain unchanged to ensure the src
829 * member of each struct exclude correctly
830 * back-references its source file. Other invocations
831 * of add_exclude_list provide stable strings, so we
832 * strdup() and free() here in the caller.
834 el->src = strdup(dir->basebuf);
835 add_excludes_from_file_to_list(dir->basebuf,
836 dir->basebuf, stk->baselen, el, 1);
838 dir->exclude_stack = stk;
839 current = stk->baselen;
841 dir->basebuf[baselen] = '\0';
845 * Loads the exclude lists for the directory containing pathname, then
846 * scans all exclude lists to determine whether pathname is excluded.
847 * Returns the exclude_list element which matched, or NULL for
848 * undecided.
850 struct exclude *last_exclude_matching(struct dir_struct *dir,
851 const char *pathname,
852 int *dtype_p)
854 int pathlen = strlen(pathname);
855 const char *basename = strrchr(pathname, '/');
856 basename = (basename) ? basename+1 : pathname;
858 prep_exclude(dir, pathname, basename-pathname);
860 if (dir->exclude)
861 return dir->exclude;
863 return last_exclude_matching_from_lists(dir, pathname, pathlen,
864 basename, dtype_p);
868 * Loads the exclude lists for the directory containing pathname, then
869 * scans all exclude lists to determine whether pathname is excluded.
870 * Returns 1 if true, otherwise 0.
872 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
874 struct exclude *exclude =
875 last_exclude_matching(dir, pathname, dtype_p);
876 if (exclude)
877 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
878 return 0;
881 static struct dir_entry *dir_entry_new(const char *pathname, int len)
883 struct dir_entry *ent;
885 ent = xmalloc(sizeof(*ent) + len + 1);
886 ent->len = len;
887 memcpy(ent->name, pathname, len);
888 ent->name[len] = 0;
889 return ent;
892 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
894 if (cache_file_exists(pathname, len, ignore_case))
895 return NULL;
897 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
898 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
901 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
903 if (!cache_name_is_other(pathname, len))
904 return NULL;
906 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
907 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
910 enum exist_status {
911 index_nonexistent = 0,
912 index_directory,
913 index_gitdir
917 * Do not use the alphabetically sorted index to look up
918 * the directory name; instead, use the case insensitive
919 * directory hash.
921 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
923 const struct cache_entry *ce = cache_dir_exists(dirname, len);
924 unsigned char endchar;
926 if (!ce)
927 return index_nonexistent;
928 endchar = ce->name[len];
931 * The cache_entry structure returned will contain this dirname
932 * and possibly additional path components.
934 if (endchar == '/')
935 return index_directory;
938 * If there are no additional path components, then this cache_entry
939 * represents a submodule. Submodules, despite being directories,
940 * are stored in the cache without a closing slash.
942 if (!endchar && S_ISGITLINK(ce->ce_mode))
943 return index_gitdir;
945 /* This should never be hit, but it exists just in case. */
946 return index_nonexistent;
950 * The index sorts alphabetically by entry name, which
951 * means that a gitlink sorts as '\0' at the end, while
952 * a directory (which is defined not as an entry, but as
953 * the files it contains) will sort with the '/' at the
954 * end.
956 static enum exist_status directory_exists_in_index(const char *dirname, int len)
958 int pos;
960 if (ignore_case)
961 return directory_exists_in_index_icase(dirname, len);
963 pos = cache_name_pos(dirname, len);
964 if (pos < 0)
965 pos = -pos-1;
966 while (pos < active_nr) {
967 const struct cache_entry *ce = active_cache[pos++];
968 unsigned char endchar;
970 if (strncmp(ce->name, dirname, len))
971 break;
972 endchar = ce->name[len];
973 if (endchar > '/')
974 break;
975 if (endchar == '/')
976 return index_directory;
977 if (!endchar && S_ISGITLINK(ce->ce_mode))
978 return index_gitdir;
980 return index_nonexistent;
984 * When we find a directory when traversing the filesystem, we
985 * have three distinct cases:
987 * - ignore it
988 * - see it as a directory
989 * - recurse into it
991 * and which one we choose depends on a combination of existing
992 * git index contents and the flags passed into the directory
993 * traversal routine.
995 * Case 1: If we *already* have entries in the index under that
996 * directory name, we always recurse into the directory to see
997 * all the files.
999 * Case 2: If we *already* have that directory name as a gitlink,
1000 * we always continue to see it as a gitlink, regardless of whether
1001 * there is an actual git directory there or not (it might not
1002 * be checked out as a subproject!)
1004 * Case 3: if we didn't have it in the index previously, we
1005 * have a few sub-cases:
1007 * (a) if "show_other_directories" is true, we show it as
1008 * just a directory, unless "hide_empty_directories" is
1009 * also true, in which case we need to check if it contains any
1010 * untracked and / or ignored files.
1011 * (b) if it looks like a git directory, and we don't have
1012 * 'no_gitlinks' set we treat it as a gitlink, and show it
1013 * as a directory.
1014 * (c) otherwise, we recurse into it.
1016 static enum path_treatment treat_directory(struct dir_struct *dir,
1017 const char *dirname, int len, int exclude,
1018 const struct path_simplify *simplify)
1020 /* The "len-1" is to strip the final '/' */
1021 switch (directory_exists_in_index(dirname, len-1)) {
1022 case index_directory:
1023 return path_recurse;
1025 case index_gitdir:
1026 return path_none;
1028 case index_nonexistent:
1029 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1030 break;
1031 if (!(dir->flags & DIR_NO_GITLINKS)) {
1032 unsigned char sha1[20];
1033 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1034 return path_untracked;
1036 return path_recurse;
1039 /* This is the "show_other_directories" case */
1041 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1042 return exclude ? path_excluded : path_untracked;
1044 return read_directory_recursive(dir, dirname, len, 1, simplify);
1048 * This is an inexact early pruning of any recursive directory
1049 * reading - if the path cannot possibly be in the pathspec,
1050 * return true, and we'll skip it early.
1052 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1054 if (simplify) {
1055 for (;;) {
1056 const char *match = simplify->path;
1057 int len = simplify->len;
1059 if (!match)
1060 break;
1061 if (len > pathlen)
1062 len = pathlen;
1063 if (!memcmp(path, match, len))
1064 return 0;
1065 simplify++;
1067 return 1;
1069 return 0;
1073 * This function tells us whether an excluded path matches a
1074 * list of "interesting" pathspecs. That is, whether a path matched
1075 * by any of the pathspecs could possibly be ignored by excluding
1076 * the specified path. This can happen if:
1078 * 1. the path is mentioned explicitly in the pathspec
1080 * 2. the path is a directory prefix of some element in the
1081 * pathspec
1083 static int exclude_matches_pathspec(const char *path, int len,
1084 const struct path_simplify *simplify)
1086 if (simplify) {
1087 for (; simplify->path; simplify++) {
1088 if (len == simplify->len
1089 && !memcmp(path, simplify->path, len))
1090 return 1;
1091 if (len < simplify->len
1092 && simplify->path[len] == '/'
1093 && !memcmp(path, simplify->path, len))
1094 return 1;
1097 return 0;
1100 static int get_index_dtype(const char *path, int len)
1102 int pos;
1103 const struct cache_entry *ce;
1105 ce = cache_file_exists(path, len, 0);
1106 if (ce) {
1107 if (!ce_uptodate(ce))
1108 return DT_UNKNOWN;
1109 if (S_ISGITLINK(ce->ce_mode))
1110 return DT_DIR;
1112 * Nobody actually cares about the
1113 * difference between DT_LNK and DT_REG
1115 return DT_REG;
1118 /* Try to look it up as a directory */
1119 pos = cache_name_pos(path, len);
1120 if (pos >= 0)
1121 return DT_UNKNOWN;
1122 pos = -pos-1;
1123 while (pos < active_nr) {
1124 ce = active_cache[pos++];
1125 if (strncmp(ce->name, path, len))
1126 break;
1127 if (ce->name[len] > '/')
1128 break;
1129 if (ce->name[len] < '/')
1130 continue;
1131 if (!ce_uptodate(ce))
1132 break; /* continue? */
1133 return DT_DIR;
1135 return DT_UNKNOWN;
1138 static int get_dtype(struct dirent *de, const char *path, int len)
1140 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1141 struct stat st;
1143 if (dtype != DT_UNKNOWN)
1144 return dtype;
1145 dtype = get_index_dtype(path, len);
1146 if (dtype != DT_UNKNOWN)
1147 return dtype;
1148 if (lstat(path, &st))
1149 return dtype;
1150 if (S_ISREG(st.st_mode))
1151 return DT_REG;
1152 if (S_ISDIR(st.st_mode))
1153 return DT_DIR;
1154 if (S_ISLNK(st.st_mode))
1155 return DT_LNK;
1156 return dtype;
1159 static enum path_treatment treat_one_path(struct dir_struct *dir,
1160 struct strbuf *path,
1161 const struct path_simplify *simplify,
1162 int dtype, struct dirent *de)
1164 int exclude;
1165 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1167 if (dtype == DT_UNKNOWN)
1168 dtype = get_dtype(de, path->buf, path->len);
1170 /* Always exclude indexed files */
1171 if (dtype != DT_DIR && has_path_in_index)
1172 return path_none;
1175 * When we are looking at a directory P in the working tree,
1176 * there are three cases:
1178 * (1) P exists in the index. Everything inside the directory P in
1179 * the working tree needs to go when P is checked out from the
1180 * index.
1182 * (2) P does not exist in the index, but there is P/Q in the index.
1183 * We know P will stay a directory when we check out the contents
1184 * of the index, but we do not know yet if there is a directory
1185 * P/Q in the working tree to be killed, so we need to recurse.
1187 * (3) P does not exist in the index, and there is no P/Q in the index
1188 * to require P to be a directory, either. Only in this case, we
1189 * know that everything inside P will not be killed without
1190 * recursing.
1192 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1193 (dtype == DT_DIR) &&
1194 !has_path_in_index &&
1195 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1196 return path_none;
1198 exclude = is_excluded(dir, path->buf, &dtype);
1201 * Excluded? If we don't explicitly want to show
1202 * ignored files, ignore it
1204 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1205 return path_excluded;
1207 switch (dtype) {
1208 default:
1209 return path_none;
1210 case DT_DIR:
1211 strbuf_addch(path, '/');
1212 return treat_directory(dir, path->buf, path->len, exclude,
1213 simplify);
1214 case DT_REG:
1215 case DT_LNK:
1216 return exclude ? path_excluded : path_untracked;
1220 static enum path_treatment treat_path(struct dir_struct *dir,
1221 struct dirent *de,
1222 struct strbuf *path,
1223 int baselen,
1224 const struct path_simplify *simplify)
1226 int dtype;
1228 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1229 return path_none;
1230 strbuf_setlen(path, baselen);
1231 strbuf_addstr(path, de->d_name);
1232 if (simplify_away(path->buf, path->len, simplify))
1233 return path_none;
1235 dtype = DTYPE(de);
1236 return treat_one_path(dir, path, simplify, dtype, de);
1240 * Read a directory tree. We currently ignore anything but
1241 * directories, regular files and symlinks. That's because git
1242 * doesn't handle them at all yet. Maybe that will change some
1243 * day.
1245 * Also, we ignore the name ".git" (even if it is not a directory).
1246 * That likely will not change.
1248 * Returns the most significant path_treatment value encountered in the scan.
1250 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1251 const char *base, int baselen,
1252 int check_only,
1253 const struct path_simplify *simplify)
1255 DIR *fdir;
1256 enum path_treatment state, subdir_state, dir_state = path_none;
1257 struct dirent *de;
1258 struct strbuf path = STRBUF_INIT;
1260 strbuf_add(&path, base, baselen);
1262 fdir = opendir(path.len ? path.buf : ".");
1263 if (!fdir)
1264 goto out;
1266 while ((de = readdir(fdir)) != NULL) {
1267 /* check how the file or directory should be treated */
1268 state = treat_path(dir, de, &path, baselen, simplify);
1269 if (state > dir_state)
1270 dir_state = state;
1272 /* recurse into subdir if instructed by treat_path */
1273 if (state == path_recurse) {
1274 subdir_state = read_directory_recursive(dir, path.buf,
1275 path.len, check_only, simplify);
1276 if (subdir_state > dir_state)
1277 dir_state = subdir_state;
1280 if (check_only) {
1281 /* abort early if maximum state has been reached */
1282 if (dir_state == path_untracked)
1283 break;
1284 /* skip the dir_add_* part */
1285 continue;
1288 /* add the path to the appropriate result list */
1289 switch (state) {
1290 case path_excluded:
1291 if (dir->flags & DIR_SHOW_IGNORED)
1292 dir_add_name(dir, path.buf, path.len);
1293 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1294 ((dir->flags & DIR_COLLECT_IGNORED) &&
1295 exclude_matches_pathspec(path.buf, path.len,
1296 simplify)))
1297 dir_add_ignored(dir, path.buf, path.len);
1298 break;
1300 case path_untracked:
1301 if (!(dir->flags & DIR_SHOW_IGNORED))
1302 dir_add_name(dir, path.buf, path.len);
1303 break;
1305 default:
1306 break;
1309 closedir(fdir);
1310 out:
1311 strbuf_release(&path);
1313 return dir_state;
1316 static int cmp_name(const void *p1, const void *p2)
1318 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1319 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1321 return cache_name_compare(e1->name, e1->len,
1322 e2->name, e2->len);
1325 static struct path_simplify *create_simplify(const char **pathspec)
1327 int nr, alloc = 0;
1328 struct path_simplify *simplify = NULL;
1330 if (!pathspec)
1331 return NULL;
1333 for (nr = 0 ; ; nr++) {
1334 const char *match;
1335 if (nr >= alloc) {
1336 alloc = alloc_nr(alloc);
1337 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
1339 match = *pathspec++;
1340 if (!match)
1341 break;
1342 simplify[nr].path = match;
1343 simplify[nr].len = simple_length(match);
1345 simplify[nr].path = NULL;
1346 simplify[nr].len = 0;
1347 return simplify;
1350 static void free_simplify(struct path_simplify *simplify)
1352 free(simplify);
1355 static int treat_leading_path(struct dir_struct *dir,
1356 const char *path, int len,
1357 const struct path_simplify *simplify)
1359 struct strbuf sb = STRBUF_INIT;
1360 int baselen, rc = 0;
1361 const char *cp;
1362 int old_flags = dir->flags;
1364 while (len && path[len - 1] == '/')
1365 len--;
1366 if (!len)
1367 return 1;
1368 baselen = 0;
1369 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1370 while (1) {
1371 cp = path + baselen + !!baselen;
1372 cp = memchr(cp, '/', path + len - cp);
1373 if (!cp)
1374 baselen = len;
1375 else
1376 baselen = cp - path;
1377 strbuf_setlen(&sb, 0);
1378 strbuf_add(&sb, path, baselen);
1379 if (!is_directory(sb.buf))
1380 break;
1381 if (simplify_away(sb.buf, sb.len, simplify))
1382 break;
1383 if (treat_one_path(dir, &sb, simplify,
1384 DT_DIR, NULL) == path_none)
1385 break; /* do not recurse into it */
1386 if (len <= baselen) {
1387 rc = 1;
1388 break; /* finished checking */
1391 strbuf_release(&sb);
1392 dir->flags = old_flags;
1393 return rc;
1396 int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1398 struct path_simplify *simplify;
1401 * Check out create_simplify()
1403 if (pathspec)
1404 GUARD_PATHSPEC(pathspec,
1405 PATHSPEC_FROMTOP |
1406 PATHSPEC_MAXDEPTH |
1407 PATHSPEC_LITERAL |
1408 PATHSPEC_GLOB |
1409 PATHSPEC_ICASE |
1410 PATHSPEC_EXCLUDE);
1412 if (has_symlink_leading_path(path, len))
1413 return dir->nr;
1416 * exclude patterns are treated like positive ones in
1417 * create_simplify. Usually exclude patterns should be a
1418 * subset of positive ones, which has no impacts on
1419 * create_simplify().
1421 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
1422 if (!len || treat_leading_path(dir, path, len, simplify))
1423 read_directory_recursive(dir, path, len, 0, simplify);
1424 free_simplify(simplify);
1425 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1426 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1427 return dir->nr;
1430 int file_exists(const char *f)
1432 struct stat sb;
1433 return lstat(f, &sb) == 0;
1437 * Given two normalized paths (a trailing slash is ok), if subdir is
1438 * outside dir, return -1. Otherwise return the offset in subdir that
1439 * can be used as relative path to dir.
1441 int dir_inside_of(const char *subdir, const char *dir)
1443 int offset = 0;
1445 assert(dir && subdir && *dir && *subdir);
1447 while (*dir && *subdir && *dir == *subdir) {
1448 dir++;
1449 subdir++;
1450 offset++;
1453 /* hel[p]/me vs hel[l]/yeah */
1454 if (*dir && *subdir)
1455 return -1;
1457 if (!*subdir)
1458 return !*dir ? offset : -1; /* same dir */
1460 /* foo/[b]ar vs foo/[] */
1461 if (is_dir_sep(dir[-1]))
1462 return is_dir_sep(subdir[-1]) ? offset : -1;
1464 /* foo[/]bar vs foo[] */
1465 return is_dir_sep(*subdir) ? offset + 1 : -1;
1468 int is_inside_dir(const char *dir)
1470 char cwd[PATH_MAX];
1471 if (!dir)
1472 return 0;
1473 if (!getcwd(cwd, sizeof(cwd)))
1474 die_errno("can't find the current directory");
1475 return dir_inside_of(cwd, dir) >= 0;
1478 int is_empty_dir(const char *path)
1480 DIR *dir = opendir(path);
1481 struct dirent *e;
1482 int ret = 1;
1484 if (!dir)
1485 return 0;
1487 while ((e = readdir(dir)) != NULL)
1488 if (!is_dot_or_dotdot(e->d_name)) {
1489 ret = 0;
1490 break;
1493 closedir(dir);
1494 return ret;
1497 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1499 DIR *dir;
1500 struct dirent *e;
1501 int ret = 0, original_len = path->len, len, kept_down = 0;
1502 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1503 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1504 unsigned char submodule_head[20];
1506 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1507 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1508 /* Do not descend and nuke a nested git work tree. */
1509 if (kept_up)
1510 *kept_up = 1;
1511 return 0;
1514 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1515 dir = opendir(path->buf);
1516 if (!dir) {
1517 if (errno == ENOENT)
1518 return keep_toplevel ? -1 : 0;
1519 else if (errno == EACCES && !keep_toplevel)
1521 * An empty dir could be removable even if it
1522 * is unreadable:
1524 return rmdir(path->buf);
1525 else
1526 return -1;
1528 if (path->buf[original_len - 1] != '/')
1529 strbuf_addch(path, '/');
1531 len = path->len;
1532 while ((e = readdir(dir)) != NULL) {
1533 struct stat st;
1534 if (is_dot_or_dotdot(e->d_name))
1535 continue;
1537 strbuf_setlen(path, len);
1538 strbuf_addstr(path, e->d_name);
1539 if (lstat(path->buf, &st)) {
1540 if (errno == ENOENT)
1542 * file disappeared, which is what we
1543 * wanted anyway
1545 continue;
1546 /* fall thru */
1547 } else if (S_ISDIR(st.st_mode)) {
1548 if (!remove_dir_recurse(path, flag, &kept_down))
1549 continue; /* happy */
1550 } else if (!only_empty &&
1551 (!unlink(path->buf) || errno == ENOENT)) {
1552 continue; /* happy, too */
1555 /* path too long, stat fails, or non-directory still exists */
1556 ret = -1;
1557 break;
1559 closedir(dir);
1561 strbuf_setlen(path, original_len);
1562 if (!ret && !keep_toplevel && !kept_down)
1563 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
1564 else if (kept_up)
1566 * report the uplevel that it is not an error that we
1567 * did not rmdir() our directory.
1569 *kept_up = !ret;
1570 return ret;
1573 int remove_dir_recursively(struct strbuf *path, int flag)
1575 return remove_dir_recurse(path, flag, NULL);
1578 void setup_standard_excludes(struct dir_struct *dir)
1580 const char *path;
1581 char *xdg_path;
1583 dir->exclude_per_dir = ".gitignore";
1584 path = git_path("info/exclude");
1585 if (!excludes_file) {
1586 home_config_paths(NULL, &xdg_path, "ignore");
1587 excludes_file = xdg_path;
1589 if (!access_or_warn(path, R_OK, 0))
1590 add_excludes_from_file(dir, path);
1591 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
1592 add_excludes_from_file(dir, excludes_file);
1595 int remove_path(const char *name)
1597 char *slash;
1599 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
1600 return -1;
1602 slash = strrchr(name, '/');
1603 if (slash) {
1604 char *dirs = xstrdup(name);
1605 slash = dirs + (slash - name);
1606 do {
1607 *slash = '\0';
1608 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1609 free(dirs);
1611 return 0;
1615 * Frees memory within dir which was allocated for exclude lists and
1616 * the exclude_stack. Does not free dir itself.
1618 void clear_directory(struct dir_struct *dir)
1620 int i, j;
1621 struct exclude_list_group *group;
1622 struct exclude_list *el;
1623 struct exclude_stack *stk;
1625 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1626 group = &dir->exclude_list_group[i];
1627 for (j = 0; j < group->nr; j++) {
1628 el = &group->el[j];
1629 if (i == EXC_DIRS)
1630 free((char *)el->src);
1631 clear_exclude_list(el);
1633 free(group->el);
1636 stk = dir->exclude_stack;
1637 while (stk) {
1638 struct exclude_stack *prev = stk->prev;
1639 free(stk);
1640 stk = prev;