unpack-trees: create working copy of merge entry in merged_entry
[git/jrn.git] / dir.c
bloba5926fbd1aeafd468860da7dbd3d8a5d5999a650
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"
15 struct path_simplify {
16 int len;
17 const char *path;
21 * Tells read_directory_recursive how a file or directory should be treated.
22 * Values are ordered by significance, e.g. if a directory contains both
23 * excluded and untracked files, it is listed as untracked because
24 * path_untracked > path_excluded.
26 enum path_treatment {
27 path_none = 0,
28 path_recurse,
29 path_excluded,
30 path_untracked
33 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
34 const char *path, int len,
35 int check_only, const struct path_simplify *simplify);
36 static int get_dtype(struct dirent *de, const char *path, int len);
38 /* helper string functions with support for the ignore_case flag */
39 int strcmp_icase(const char *a, const char *b)
41 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
44 int strncmp_icase(const char *a, const char *b, size_t count)
46 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
49 int fnmatch_icase(const char *pattern, const char *string, int flags)
51 return fnmatch(pattern, string, flags | (ignore_case ? FNM_CASEFOLD : 0));
54 inline int git_fnmatch(const char *pattern, const char *string,
55 int flags, int prefix)
57 int fnm_flags = 0;
58 if (flags & GFNM_PATHNAME)
59 fnm_flags |= FNM_PATHNAME;
60 if (prefix > 0) {
61 if (strncmp(pattern, string, prefix))
62 return FNM_NOMATCH;
63 pattern += prefix;
64 string += prefix;
66 if (flags & GFNM_ONESTAR) {
67 int pattern_len = strlen(++pattern);
68 int string_len = strlen(string);
69 return string_len < pattern_len ||
70 strcmp(pattern,
71 string + string_len - pattern_len);
73 return fnmatch(pattern, string, fnm_flags);
76 static int fnmatch_icase_mem(const char *pattern, int patternlen,
77 const char *string, int stringlen,
78 int flags)
80 int match_status;
81 struct strbuf pat_buf = STRBUF_INIT;
82 struct strbuf str_buf = STRBUF_INIT;
83 const char *use_pat = pattern;
84 const char *use_str = string;
86 if (pattern[patternlen]) {
87 strbuf_add(&pat_buf, pattern, patternlen);
88 use_pat = pat_buf.buf;
90 if (string[stringlen]) {
91 strbuf_add(&str_buf, string, stringlen);
92 use_str = str_buf.buf;
95 if (ignore_case)
96 flags |= WM_CASEFOLD;
97 match_status = wildmatch(use_pat, use_str, flags, NULL);
99 strbuf_release(&pat_buf);
100 strbuf_release(&str_buf);
102 return match_status;
105 static size_t common_prefix_len(const char **pathspec)
107 const char *n, *first;
108 size_t max = 0;
109 int literal = limit_pathspec_to_literal();
111 if (!pathspec)
112 return max;
114 first = *pathspec;
115 while ((n = *pathspec++)) {
116 size_t i, len = 0;
117 for (i = 0; first == n || i < max; i++) {
118 char c = n[i];
119 if (!c || c != first[i] || (!literal && is_glob_special(c)))
120 break;
121 if (c == '/')
122 len = i + 1;
124 if (first == n || len < max) {
125 max = len;
126 if (!max)
127 break;
130 return max;
134 * Returns a copy of the longest leading path common among all
135 * pathspecs.
137 char *common_prefix(const char **pathspec)
139 unsigned long len = common_prefix_len(pathspec);
141 return len ? xmemdupz(*pathspec, len) : NULL;
144 int fill_directory(struct dir_struct *dir, const char **pathspec)
146 size_t len;
149 * Calculate common prefix for the pathspec, and
150 * use that to optimize the directory walk
152 len = common_prefix_len(pathspec);
154 /* Read the directory and prune it */
155 read_directory(dir, pathspec ? *pathspec : "", len, pathspec);
156 return len;
159 int within_depth(const char *name, int namelen,
160 int depth, int max_depth)
162 const char *cp = name, *cpe = name + namelen;
164 while (cp < cpe) {
165 if (*cp++ != '/')
166 continue;
167 depth++;
168 if (depth > max_depth)
169 return 0;
171 return 1;
175 * Does 'match' match the given name?
176 * A match is found if
178 * (1) the 'match' string is leading directory of 'name', or
179 * (2) the 'match' string is a wildcard and matches 'name', or
180 * (3) the 'match' string is exactly the same as 'name'.
182 * and the return value tells which case it was.
184 * It returns 0 when there is no match.
186 static int match_one(const char *match, const char *name, int namelen)
188 int matchlen;
189 int literal = limit_pathspec_to_literal();
191 /* If the match was just the prefix, we matched */
192 if (!*match)
193 return MATCHED_RECURSIVELY;
195 if (ignore_case) {
196 for (;;) {
197 unsigned char c1 = tolower(*match);
198 unsigned char c2 = tolower(*name);
199 if (c1 == '\0' || (!literal && is_glob_special(c1)))
200 break;
201 if (c1 != c2)
202 return 0;
203 match++;
204 name++;
205 namelen--;
207 } else {
208 for (;;) {
209 unsigned char c1 = *match;
210 unsigned char c2 = *name;
211 if (c1 == '\0' || (!literal && is_glob_special(c1)))
212 break;
213 if (c1 != c2)
214 return 0;
215 match++;
216 name++;
217 namelen--;
222 * If we don't match the matchstring exactly,
223 * we need to match by fnmatch
225 matchlen = strlen(match);
226 if (strncmp_icase(match, name, matchlen)) {
227 if (literal)
228 return 0;
229 return !fnmatch_icase(match, name, 0) ? MATCHED_FNMATCH : 0;
232 if (namelen == matchlen)
233 return MATCHED_EXACTLY;
234 if (match[matchlen-1] == '/' || name[matchlen] == '/')
235 return MATCHED_RECURSIVELY;
236 return 0;
240 * Given a name and a list of pathspecs, returns the nature of the
241 * closest (i.e. most specific) match of the name to any of the
242 * pathspecs.
244 * The caller typically calls this multiple times with the same
245 * pathspec and seen[] array but with different name/namelen
246 * (e.g. entries from the index) and is interested in seeing if and
247 * how each pathspec matches all the names it calls this function
248 * with. A mark is left in the seen[] array for each pathspec element
249 * indicating the closest type of match that element achieved, so if
250 * seen[n] remains zero after multiple invocations, that means the nth
251 * pathspec did not match any names, which could indicate that the
252 * user mistyped the nth pathspec.
254 int match_pathspec(const char **pathspec, const char *name, int namelen,
255 int prefix, char *seen)
257 int i, retval = 0;
259 if (!pathspec)
260 return 1;
262 name += prefix;
263 namelen -= prefix;
265 for (i = 0; pathspec[i] != NULL; i++) {
266 int how;
267 const char *match = pathspec[i] + prefix;
268 if (seen && seen[i] == MATCHED_EXACTLY)
269 continue;
270 how = match_one(match, name, namelen);
271 if (how) {
272 if (retval < how)
273 retval = how;
274 if (seen && seen[i] < how)
275 seen[i] = how;
278 return retval;
282 * Does 'match' match the given name?
283 * A match is found if
285 * (1) the 'match' string is leading directory of 'name', or
286 * (2) the 'match' string is a wildcard and matches 'name', or
287 * (3) the 'match' string is exactly the same as 'name'.
289 * and the return value tells which case it was.
291 * It returns 0 when there is no match.
293 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
294 const char *name, int namelen)
296 /* name/namelen has prefix cut off by caller */
297 const char *match = item->match + prefix;
298 int matchlen = item->len - prefix;
300 /* If the match was just the prefix, we matched */
301 if (!*match)
302 return MATCHED_RECURSIVELY;
304 if (matchlen <= namelen && !strncmp(match, name, matchlen)) {
305 if (matchlen == namelen)
306 return MATCHED_EXACTLY;
308 if (match[matchlen-1] == '/' || name[matchlen] == '/')
309 return MATCHED_RECURSIVELY;
312 if (item->nowildcard_len < item->len &&
313 !git_fnmatch(match, name,
314 item->flags & PATHSPEC_ONESTAR ? GFNM_ONESTAR : 0,
315 item->nowildcard_len - prefix))
316 return MATCHED_FNMATCH;
318 return 0;
322 * Given a name and a list of pathspecs, returns the nature of the
323 * closest (i.e. most specific) match of the name to any of the
324 * pathspecs.
326 * The caller typically calls this multiple times with the same
327 * pathspec and seen[] array but with different name/namelen
328 * (e.g. entries from the index) and is interested in seeing if and
329 * how each pathspec matches all the names it calls this function
330 * with. A mark is left in the seen[] array for each pathspec element
331 * indicating the closest type of match that element achieved, so if
332 * seen[n] remains zero after multiple invocations, that means the nth
333 * pathspec did not match any names, which could indicate that the
334 * user mistyped the nth pathspec.
336 int match_pathspec_depth(const struct pathspec *ps,
337 const char *name, int namelen,
338 int prefix, char *seen)
340 int i, retval = 0;
342 if (!ps->nr) {
343 if (!ps->recursive || ps->max_depth == -1)
344 return MATCHED_RECURSIVELY;
346 if (within_depth(name, namelen, 0, ps->max_depth))
347 return MATCHED_EXACTLY;
348 else
349 return 0;
352 name += prefix;
353 namelen -= prefix;
355 for (i = ps->nr - 1; i >= 0; i--) {
356 int how;
357 if (seen && seen[i] == MATCHED_EXACTLY)
358 continue;
359 how = match_pathspec_item(ps->items+i, prefix, name, namelen);
360 if (ps->recursive && ps->max_depth != -1 &&
361 how && how != MATCHED_FNMATCH) {
362 int len = ps->items[i].len;
363 if (name[len] == '/')
364 len++;
365 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
366 how = MATCHED_EXACTLY;
367 else
368 how = 0;
370 if (how) {
371 if (retval < how)
372 retval = how;
373 if (seen && seen[i] < how)
374 seen[i] = how;
377 return retval;
381 * Return the length of the "simple" part of a path match limiter.
383 static int simple_length(const char *match)
385 int len = -1;
387 for (;;) {
388 unsigned char c = *match++;
389 len++;
390 if (c == '\0' || is_glob_special(c))
391 return len;
395 static int no_wildcard(const char *string)
397 return string[simple_length(string)] == '\0';
400 void parse_exclude_pattern(const char **pattern,
401 int *patternlen,
402 int *flags,
403 int *nowildcardlen)
405 const char *p = *pattern;
406 size_t i, len;
408 *flags = 0;
409 if (*p == '!') {
410 *flags |= EXC_FLAG_NEGATIVE;
411 p++;
413 len = strlen(p);
414 if (len && p[len - 1] == '/') {
415 len--;
416 *flags |= EXC_FLAG_MUSTBEDIR;
418 for (i = 0; i < len; i++) {
419 if (p[i] == '/')
420 break;
422 if (i == len)
423 *flags |= EXC_FLAG_NODIR;
424 *nowildcardlen = simple_length(p);
426 * we should have excluded the trailing slash from 'p' too,
427 * but that's one more allocation. Instead just make sure
428 * nowildcardlen does not exceed real patternlen
430 if (*nowildcardlen > len)
431 *nowildcardlen = len;
432 if (*p == '*' && no_wildcard(p + 1))
433 *flags |= EXC_FLAG_ENDSWITH;
434 *pattern = p;
435 *patternlen = len;
438 void add_exclude(const char *string, const char *base,
439 int baselen, struct exclude_list *el, int srcpos)
441 struct exclude *x;
442 int patternlen;
443 int flags;
444 int nowildcardlen;
446 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
447 if (flags & EXC_FLAG_MUSTBEDIR) {
448 char *s;
449 x = xmalloc(sizeof(*x) + patternlen + 1);
450 s = (char *)(x+1);
451 memcpy(s, string, patternlen);
452 s[patternlen] = '\0';
453 x->pattern = s;
454 } else {
455 x = xmalloc(sizeof(*x));
456 x->pattern = string;
458 x->patternlen = patternlen;
459 x->nowildcardlen = nowildcardlen;
460 x->base = base;
461 x->baselen = baselen;
462 x->flags = flags;
463 x->srcpos = srcpos;
464 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
465 el->excludes[el->nr++] = x;
466 x->el = el;
469 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
471 int pos, len;
472 unsigned long sz;
473 enum object_type type;
474 void *data;
475 struct index_state *istate = &the_index;
477 len = strlen(path);
478 pos = index_name_pos(istate, path, len);
479 if (pos < 0)
480 return NULL;
481 if (!ce_skip_worktree(istate->cache[pos]))
482 return NULL;
483 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
484 if (!data || type != OBJ_BLOB) {
485 free(data);
486 return NULL;
488 *size = xsize_t(sz);
489 return data;
493 * Frees memory within el which was allocated for exclude patterns and
494 * the file buffer. Does not free el itself.
496 void clear_exclude_list(struct exclude_list *el)
498 int i;
500 for (i = 0; i < el->nr; i++)
501 free(el->excludes[i]);
502 free(el->excludes);
503 free(el->filebuf);
505 el->nr = 0;
506 el->excludes = NULL;
507 el->filebuf = NULL;
510 int add_excludes_from_file_to_list(const char *fname,
511 const char *base,
512 int baselen,
513 struct exclude_list *el,
514 int check_index)
516 struct stat st;
517 int fd, i, lineno = 1;
518 size_t size = 0;
519 char *buf, *entry;
521 fd = open(fname, O_RDONLY);
522 if (fd < 0 || fstat(fd, &st) < 0) {
523 if (errno != ENOENT)
524 warn_on_inaccessible(fname);
525 if (0 <= fd)
526 close(fd);
527 if (!check_index ||
528 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
529 return -1;
530 if (size == 0) {
531 free(buf);
532 return 0;
534 if (buf[size-1] != '\n') {
535 buf = xrealloc(buf, size+1);
536 buf[size++] = '\n';
539 else {
540 size = xsize_t(st.st_size);
541 if (size == 0) {
542 close(fd);
543 return 0;
545 buf = xmalloc(size+1);
546 if (read_in_full(fd, buf, size) != size) {
547 free(buf);
548 close(fd);
549 return -1;
551 buf[size++] = '\n';
552 close(fd);
555 el->filebuf = buf;
556 entry = buf;
557 for (i = 0; i < size; i++) {
558 if (buf[i] == '\n') {
559 if (entry != buf + i && entry[0] != '#') {
560 buf[i - (i && buf[i-1] == '\r')] = 0;
561 add_exclude(entry, base, baselen, el, lineno);
563 lineno++;
564 entry = buf + i + 1;
567 return 0;
570 struct exclude_list *add_exclude_list(struct dir_struct *dir,
571 int group_type, const char *src)
573 struct exclude_list *el;
574 struct exclude_list_group *group;
576 group = &dir->exclude_list_group[group_type];
577 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
578 el = &group->el[group->nr++];
579 memset(el, 0, sizeof(*el));
580 el->src = src;
581 return el;
585 * Used to set up core.excludesfile and .git/info/exclude lists.
587 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
589 struct exclude_list *el;
590 el = add_exclude_list(dir, EXC_FILE, fname);
591 if (add_excludes_from_file_to_list(fname, "", 0, el, 0) < 0)
592 die("cannot use %s as an exclude file", fname);
595 int match_basename(const char *basename, int basenamelen,
596 const char *pattern, int prefix, int patternlen,
597 int flags)
599 if (prefix == patternlen) {
600 if (patternlen == basenamelen &&
601 !strncmp_icase(pattern, basename, basenamelen))
602 return 1;
603 } else if (flags & EXC_FLAG_ENDSWITH) {
604 /* "*literal" matching against "fooliteral" */
605 if (patternlen - 1 <= basenamelen &&
606 !strncmp_icase(pattern + 1,
607 basename + basenamelen - (patternlen - 1),
608 patternlen - 1))
609 return 1;
610 } else {
611 if (fnmatch_icase_mem(pattern, patternlen,
612 basename, basenamelen,
613 0) == 0)
614 return 1;
616 return 0;
619 int match_pathname(const char *pathname, int pathlen,
620 const char *base, int baselen,
621 const char *pattern, int prefix, int patternlen,
622 int flags)
624 const char *name;
625 int namelen;
628 * match with FNM_PATHNAME; the pattern has base implicitly
629 * in front of it.
631 if (*pattern == '/') {
632 pattern++;
633 patternlen--;
634 prefix--;
638 * baselen does not count the trailing slash. base[] may or
639 * may not end with a trailing slash though.
641 if (pathlen < baselen + 1 ||
642 (baselen && pathname[baselen] != '/') ||
643 strncmp_icase(pathname, base, baselen))
644 return 0;
646 namelen = baselen ? pathlen - baselen - 1 : pathlen;
647 name = pathname + pathlen - namelen;
649 if (prefix) {
651 * if the non-wildcard part is longer than the
652 * remaining pathname, surely it cannot match.
654 if (prefix > namelen)
655 return 0;
657 if (strncmp_icase(pattern, name, prefix))
658 return 0;
659 pattern += prefix;
660 patternlen -= prefix;
661 name += prefix;
662 namelen -= prefix;
665 * If the whole pattern did not have a wildcard,
666 * then our prefix match is all we need; we
667 * do not need to call fnmatch at all.
669 if (!patternlen && !namelen)
670 return 1;
673 return fnmatch_icase_mem(pattern, patternlen,
674 name, namelen,
675 WM_PATHNAME) == 0;
679 * Scan the given exclude list in reverse to see whether pathname
680 * should be ignored. The first match (i.e. the last on the list), if
681 * any, determines the fate. Returns the exclude_list element which
682 * matched, or NULL for undecided.
684 static struct exclude *last_exclude_matching_from_list(const char *pathname,
685 int pathlen,
686 const char *basename,
687 int *dtype,
688 struct exclude_list *el)
690 int i;
692 if (!el->nr)
693 return NULL; /* undefined */
695 for (i = el->nr - 1; 0 <= i; i--) {
696 struct exclude *x = el->excludes[i];
697 const char *exclude = x->pattern;
698 int prefix = x->nowildcardlen;
700 if (x->flags & EXC_FLAG_MUSTBEDIR) {
701 if (*dtype == DT_UNKNOWN)
702 *dtype = get_dtype(NULL, pathname, pathlen);
703 if (*dtype != DT_DIR)
704 continue;
707 if (x->flags & EXC_FLAG_NODIR) {
708 if (match_basename(basename,
709 pathlen - (basename - pathname),
710 exclude, prefix, x->patternlen,
711 x->flags))
712 return x;
713 continue;
716 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
717 if (match_pathname(pathname, pathlen,
718 x->base, x->baselen ? x->baselen - 1 : 0,
719 exclude, prefix, x->patternlen, x->flags))
720 return x;
722 return NULL; /* undecided */
726 * Scan the list and let the last match determine the fate.
727 * Return 1 for exclude, 0 for include and -1 for undecided.
729 int is_excluded_from_list(const char *pathname,
730 int pathlen, const char *basename, int *dtype,
731 struct exclude_list *el)
733 struct exclude *exclude;
734 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
735 if (exclude)
736 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
737 return -1; /* undecided */
740 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
741 const char *pathname, int pathlen, const char *basename,
742 int *dtype_p)
744 int i, j;
745 struct exclude_list_group *group;
746 struct exclude *exclude;
747 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
748 group = &dir->exclude_list_group[i];
749 for (j = group->nr - 1; j >= 0; j--) {
750 exclude = last_exclude_matching_from_list(
751 pathname, pathlen, basename, dtype_p,
752 &group->el[j]);
753 if (exclude)
754 return exclude;
757 return NULL;
761 * Loads the per-directory exclude list for the substring of base
762 * which has a char length of baselen.
764 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
766 struct exclude_list_group *group;
767 struct exclude_list *el;
768 struct exclude_stack *stk = NULL;
769 int current;
771 group = &dir->exclude_list_group[EXC_DIRS];
773 /* Pop the exclude lists from the EXCL_DIRS exclude_list_group
774 * which originate from directories not in the prefix of the
775 * path being checked. */
776 while ((stk = dir->exclude_stack) != NULL) {
777 if (stk->baselen <= baselen &&
778 !strncmp(dir->basebuf, base, stk->baselen))
779 break;
780 el = &group->el[dir->exclude_stack->exclude_ix];
781 dir->exclude_stack = stk->prev;
782 dir->exclude = NULL;
783 free((char *)el->src); /* see strdup() below */
784 clear_exclude_list(el);
785 free(stk);
786 group->nr--;
789 /* Skip traversing into sub directories if the parent is excluded */
790 if (dir->exclude)
791 return;
793 /* Read from the parent directories and push them down. */
794 current = stk ? stk->baselen : -1;
795 while (current < baselen) {
796 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
797 const char *cp;
799 if (current < 0) {
800 cp = base;
801 current = 0;
803 else {
804 cp = strchr(base + current + 1, '/');
805 if (!cp)
806 die("oops in prep_exclude");
807 cp++;
809 stk->prev = dir->exclude_stack;
810 stk->baselen = cp - base;
811 stk->exclude_ix = group->nr;
812 el = add_exclude_list(dir, EXC_DIRS, NULL);
813 memcpy(dir->basebuf + current, base + current,
814 stk->baselen - current);
816 /* Abort if the directory is excluded */
817 if (stk->baselen) {
818 int dt = DT_DIR;
819 dir->basebuf[stk->baselen - 1] = 0;
820 dir->exclude = last_exclude_matching_from_lists(dir,
821 dir->basebuf, stk->baselen - 1,
822 dir->basebuf + current, &dt);
823 dir->basebuf[stk->baselen - 1] = '/';
824 if (dir->exclude) {
825 dir->basebuf[stk->baselen] = 0;
826 dir->exclude_stack = stk;
827 return;
831 /* Try to read per-directory file unless path is too long */
832 if (dir->exclude_per_dir &&
833 stk->baselen + strlen(dir->exclude_per_dir) < PATH_MAX) {
834 strcpy(dir->basebuf + stk->baselen,
835 dir->exclude_per_dir);
837 * dir->basebuf gets reused by the traversal, but we
838 * need fname to remain unchanged to ensure the src
839 * member of each struct exclude correctly
840 * back-references its source file. Other invocations
841 * of add_exclude_list provide stable strings, so we
842 * strdup() and free() here in the caller.
844 el->src = strdup(dir->basebuf);
845 add_excludes_from_file_to_list(dir->basebuf,
846 dir->basebuf, stk->baselen, el, 1);
848 dir->exclude_stack = stk;
849 current = stk->baselen;
851 dir->basebuf[baselen] = '\0';
855 * Loads the exclude lists for the directory containing pathname, then
856 * scans all exclude lists to determine whether pathname is excluded.
857 * Returns the exclude_list element which matched, or NULL for
858 * undecided.
860 struct exclude *last_exclude_matching(struct dir_struct *dir,
861 const char *pathname,
862 int *dtype_p)
864 int pathlen = strlen(pathname);
865 const char *basename = strrchr(pathname, '/');
866 basename = (basename) ? basename+1 : pathname;
868 prep_exclude(dir, pathname, basename-pathname);
870 if (dir->exclude)
871 return dir->exclude;
873 return last_exclude_matching_from_lists(dir, pathname, pathlen,
874 basename, dtype_p);
878 * Loads the exclude lists for the directory containing pathname, then
879 * scans all exclude lists to determine whether pathname is excluded.
880 * Returns 1 if true, otherwise 0.
882 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
884 struct exclude *exclude =
885 last_exclude_matching(dir, pathname, dtype_p);
886 if (exclude)
887 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
888 return 0;
891 static struct dir_entry *dir_entry_new(const char *pathname, int len)
893 struct dir_entry *ent;
895 ent = xmalloc(sizeof(*ent) + len + 1);
896 ent->len = len;
897 memcpy(ent->name, pathname, len);
898 ent->name[len] = 0;
899 return ent;
902 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
904 if (cache_name_exists(pathname, len, ignore_case))
905 return NULL;
907 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
908 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
911 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
913 if (!cache_name_is_other(pathname, len))
914 return NULL;
916 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
917 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
920 enum exist_status {
921 index_nonexistent = 0,
922 index_directory,
923 index_gitdir
927 * Do not use the alphabetically stored index to look up
928 * the directory name; instead, use the case insensitive
929 * name hash.
931 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
933 struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
934 unsigned char endchar;
936 if (!ce)
937 return index_nonexistent;
938 endchar = ce->name[len];
941 * The cache_entry structure returned will contain this dirname
942 * and possibly additional path components.
944 if (endchar == '/')
945 return index_directory;
948 * If there are no additional path components, then this cache_entry
949 * represents a submodule. Submodules, despite being directories,
950 * are stored in the cache without a closing slash.
952 if (!endchar && S_ISGITLINK(ce->ce_mode))
953 return index_gitdir;
955 /* This should never be hit, but it exists just in case. */
956 return index_nonexistent;
960 * The index sorts alphabetically by entry name, which
961 * means that a gitlink sorts as '\0' at the end, while
962 * a directory (which is defined not as an entry, but as
963 * the files it contains) will sort with the '/' at the
964 * end.
966 static enum exist_status directory_exists_in_index(const char *dirname, int len)
968 int pos;
970 if (ignore_case)
971 return directory_exists_in_index_icase(dirname, len);
973 pos = cache_name_pos(dirname, len);
974 if (pos < 0)
975 pos = -pos-1;
976 while (pos < active_nr) {
977 struct cache_entry *ce = active_cache[pos++];
978 unsigned char endchar;
980 if (strncmp(ce->name, dirname, len))
981 break;
982 endchar = ce->name[len];
983 if (endchar > '/')
984 break;
985 if (endchar == '/')
986 return index_directory;
987 if (!endchar && S_ISGITLINK(ce->ce_mode))
988 return index_gitdir;
990 return index_nonexistent;
994 * When we find a directory when traversing the filesystem, we
995 * have three distinct cases:
997 * - ignore it
998 * - see it as a directory
999 * - recurse into it
1001 * and which one we choose depends on a combination of existing
1002 * git index contents and the flags passed into the directory
1003 * traversal routine.
1005 * Case 1: If we *already* have entries in the index under that
1006 * directory name, we always recurse into the directory to see
1007 * all the files.
1009 * Case 2: If we *already* have that directory name as a gitlink,
1010 * we always continue to see it as a gitlink, regardless of whether
1011 * there is an actual git directory there or not (it might not
1012 * be checked out as a subproject!)
1014 * Case 3: if we didn't have it in the index previously, we
1015 * have a few sub-cases:
1017 * (a) if "show_other_directories" is true, we show it as
1018 * just a directory, unless "hide_empty_directories" is
1019 * also true, in which case we need to check if it contains any
1020 * untracked and / or ignored files.
1021 * (b) if it looks like a git directory, and we don't have
1022 * 'no_gitlinks' set we treat it as a gitlink, and show it
1023 * as a directory.
1024 * (c) otherwise, we recurse into it.
1026 static enum path_treatment treat_directory(struct dir_struct *dir,
1027 const char *dirname, int len, int exclude,
1028 const struct path_simplify *simplify)
1030 /* The "len-1" is to strip the final '/' */
1031 switch (directory_exists_in_index(dirname, len-1)) {
1032 case index_directory:
1033 return path_recurse;
1035 case index_gitdir:
1036 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1037 return path_none;
1038 return path_untracked;
1040 case index_nonexistent:
1041 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1042 break;
1043 if (!(dir->flags & DIR_NO_GITLINKS)) {
1044 unsigned char sha1[20];
1045 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1046 return path_untracked;
1048 return path_recurse;
1051 /* This is the "show_other_directories" case */
1053 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1054 return exclude ? path_excluded : path_untracked;
1056 return read_directory_recursive(dir, dirname, len, 1, simplify);
1060 * This is an inexact early pruning of any recursive directory
1061 * reading - if the path cannot possibly be in the pathspec,
1062 * return true, and we'll skip it early.
1064 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1066 if (simplify) {
1067 for (;;) {
1068 const char *match = simplify->path;
1069 int len = simplify->len;
1071 if (!match)
1072 break;
1073 if (len > pathlen)
1074 len = pathlen;
1075 if (!memcmp(path, match, len))
1076 return 0;
1077 simplify++;
1079 return 1;
1081 return 0;
1085 * This function tells us whether an excluded path matches a
1086 * list of "interesting" pathspecs. That is, whether a path matched
1087 * by any of the pathspecs could possibly be ignored by excluding
1088 * the specified path. This can happen if:
1090 * 1. the path is mentioned explicitly in the pathspec
1092 * 2. the path is a directory prefix of some element in the
1093 * pathspec
1095 static int exclude_matches_pathspec(const char *path, int len,
1096 const struct path_simplify *simplify)
1098 if (simplify) {
1099 for (; simplify->path; simplify++) {
1100 if (len == simplify->len
1101 && !memcmp(path, simplify->path, len))
1102 return 1;
1103 if (len < simplify->len
1104 && simplify->path[len] == '/'
1105 && !memcmp(path, simplify->path, len))
1106 return 1;
1109 return 0;
1112 static int get_index_dtype(const char *path, int len)
1114 int pos;
1115 struct cache_entry *ce;
1117 ce = cache_name_exists(path, len, 0);
1118 if (ce) {
1119 if (!ce_uptodate(ce))
1120 return DT_UNKNOWN;
1121 if (S_ISGITLINK(ce->ce_mode))
1122 return DT_DIR;
1124 * Nobody actually cares about the
1125 * difference between DT_LNK and DT_REG
1127 return DT_REG;
1130 /* Try to look it up as a directory */
1131 pos = cache_name_pos(path, len);
1132 if (pos >= 0)
1133 return DT_UNKNOWN;
1134 pos = -pos-1;
1135 while (pos < active_nr) {
1136 ce = active_cache[pos++];
1137 if (strncmp(ce->name, path, len))
1138 break;
1139 if (ce->name[len] > '/')
1140 break;
1141 if (ce->name[len] < '/')
1142 continue;
1143 if (!ce_uptodate(ce))
1144 break; /* continue? */
1145 return DT_DIR;
1147 return DT_UNKNOWN;
1150 static int get_dtype(struct dirent *de, const char *path, int len)
1152 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1153 struct stat st;
1155 if (dtype != DT_UNKNOWN)
1156 return dtype;
1157 dtype = get_index_dtype(path, len);
1158 if (dtype != DT_UNKNOWN)
1159 return dtype;
1160 if (lstat(path, &st))
1161 return dtype;
1162 if (S_ISREG(st.st_mode))
1163 return DT_REG;
1164 if (S_ISDIR(st.st_mode))
1165 return DT_DIR;
1166 if (S_ISLNK(st.st_mode))
1167 return DT_LNK;
1168 return dtype;
1171 static enum path_treatment treat_one_path(struct dir_struct *dir,
1172 struct strbuf *path,
1173 const struct path_simplify *simplify,
1174 int dtype, struct dirent *de)
1176 int exclude;
1177 if (dtype == DT_UNKNOWN)
1178 dtype = get_dtype(de, path->buf, path->len);
1180 /* Always exclude indexed files */
1181 if (dtype != DT_DIR &&
1182 cache_name_exists(path->buf, path->len, ignore_case))
1183 return path_none;
1185 exclude = is_excluded(dir, path->buf, &dtype);
1188 * Excluded? If we don't explicitly want to show
1189 * ignored files, ignore it
1191 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1192 return path_excluded;
1194 switch (dtype) {
1195 default:
1196 return path_none;
1197 case DT_DIR:
1198 strbuf_addch(path, '/');
1199 return treat_directory(dir, path->buf, path->len, exclude,
1200 simplify);
1201 case DT_REG:
1202 case DT_LNK:
1203 return exclude ? path_excluded : path_untracked;
1207 static enum path_treatment treat_path(struct dir_struct *dir,
1208 struct dirent *de,
1209 struct strbuf *path,
1210 int baselen,
1211 const struct path_simplify *simplify)
1213 int dtype;
1215 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1216 return path_none;
1217 strbuf_setlen(path, baselen);
1218 strbuf_addstr(path, de->d_name);
1219 if (simplify_away(path->buf, path->len, simplify))
1220 return path_none;
1222 dtype = DTYPE(de);
1223 return treat_one_path(dir, path, simplify, dtype, de);
1227 * Read a directory tree. We currently ignore anything but
1228 * directories, regular files and symlinks. That's because git
1229 * doesn't handle them at all yet. Maybe that will change some
1230 * day.
1232 * Also, we ignore the name ".git" (even if it is not a directory).
1233 * That likely will not change.
1235 * Returns the most significant path_treatment value encountered in the scan.
1237 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1238 const char *base, int baselen,
1239 int check_only,
1240 const struct path_simplify *simplify)
1242 DIR *fdir;
1243 enum path_treatment state, subdir_state, dir_state = path_none;
1244 struct dirent *de;
1245 struct strbuf path = STRBUF_INIT;
1247 strbuf_add(&path, base, baselen);
1249 fdir = opendir(path.len ? path.buf : ".");
1250 if (!fdir)
1251 goto out;
1253 while ((de = readdir(fdir)) != NULL) {
1254 /* check how the file or directory should be treated */
1255 state = treat_path(dir, de, &path, baselen, simplify);
1256 if (state > dir_state)
1257 dir_state = state;
1259 /* recurse into subdir if instructed by treat_path */
1260 if (state == path_recurse) {
1261 subdir_state = read_directory_recursive(dir, path.buf,
1262 path.len, check_only, simplify);
1263 if (subdir_state > dir_state)
1264 dir_state = subdir_state;
1267 if (check_only) {
1268 /* abort early if maximum state has been reached */
1269 if (dir_state == path_untracked)
1270 break;
1271 /* skip the dir_add_* part */
1272 continue;
1275 /* add the path to the appropriate result list */
1276 switch (state) {
1277 case path_excluded:
1278 if (dir->flags & DIR_SHOW_IGNORED)
1279 dir_add_name(dir, path.buf, path.len);
1280 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1281 ((dir->flags & DIR_COLLECT_IGNORED) &&
1282 exclude_matches_pathspec(path.buf, path.len,
1283 simplify)))
1284 dir_add_ignored(dir, path.buf, path.len);
1285 break;
1287 case path_untracked:
1288 if (!(dir->flags & DIR_SHOW_IGNORED))
1289 dir_add_name(dir, path.buf, path.len);
1290 break;
1292 default:
1293 break;
1296 closedir(fdir);
1297 out:
1298 strbuf_release(&path);
1300 return dir_state;
1303 static int cmp_name(const void *p1, const void *p2)
1305 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1306 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1308 return cache_name_compare(e1->name, e1->len,
1309 e2->name, e2->len);
1312 static struct path_simplify *create_simplify(const char **pathspec)
1314 int nr, alloc = 0;
1315 struct path_simplify *simplify = NULL;
1317 if (!pathspec)
1318 return NULL;
1320 for (nr = 0 ; ; nr++) {
1321 const char *match;
1322 if (nr >= alloc) {
1323 alloc = alloc_nr(alloc);
1324 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
1326 match = *pathspec++;
1327 if (!match)
1328 break;
1329 simplify[nr].path = match;
1330 simplify[nr].len = simple_length(match);
1332 simplify[nr].path = NULL;
1333 simplify[nr].len = 0;
1334 return simplify;
1337 static void free_simplify(struct path_simplify *simplify)
1339 free(simplify);
1342 static int treat_leading_path(struct dir_struct *dir,
1343 const char *path, int len,
1344 const struct path_simplify *simplify)
1346 struct strbuf sb = STRBUF_INIT;
1347 int baselen, rc = 0;
1348 const char *cp;
1349 int old_flags = dir->flags;
1351 while (len && path[len - 1] == '/')
1352 len--;
1353 if (!len)
1354 return 1;
1355 baselen = 0;
1356 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1357 while (1) {
1358 cp = path + baselen + !!baselen;
1359 cp = memchr(cp, '/', path + len - cp);
1360 if (!cp)
1361 baselen = len;
1362 else
1363 baselen = cp - path;
1364 strbuf_setlen(&sb, 0);
1365 strbuf_add(&sb, path, baselen);
1366 if (!is_directory(sb.buf))
1367 break;
1368 if (simplify_away(sb.buf, sb.len, simplify))
1369 break;
1370 if (treat_one_path(dir, &sb, simplify,
1371 DT_DIR, NULL) == path_none)
1372 break; /* do not recurse into it */
1373 if (len <= baselen) {
1374 rc = 1;
1375 break; /* finished checking */
1378 strbuf_release(&sb);
1379 dir->flags = old_flags;
1380 return rc;
1383 int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
1385 struct path_simplify *simplify;
1387 if (has_symlink_leading_path(path, len))
1388 return dir->nr;
1390 simplify = create_simplify(pathspec);
1391 if (!len || treat_leading_path(dir, path, len, simplify))
1392 read_directory_recursive(dir, path, len, 0, simplify);
1393 free_simplify(simplify);
1394 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1395 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1396 return dir->nr;
1399 int file_exists(const char *f)
1401 struct stat sb;
1402 return lstat(f, &sb) == 0;
1406 * Given two normalized paths (a trailing slash is ok), if subdir is
1407 * outside dir, return -1. Otherwise return the offset in subdir that
1408 * can be used as relative path to dir.
1410 int dir_inside_of(const char *subdir, const char *dir)
1412 int offset = 0;
1414 assert(dir && subdir && *dir && *subdir);
1416 while (*dir && *subdir && *dir == *subdir) {
1417 dir++;
1418 subdir++;
1419 offset++;
1422 /* hel[p]/me vs hel[l]/yeah */
1423 if (*dir && *subdir)
1424 return -1;
1426 if (!*subdir)
1427 return !*dir ? offset : -1; /* same dir */
1429 /* foo/[b]ar vs foo/[] */
1430 if (is_dir_sep(dir[-1]))
1431 return is_dir_sep(subdir[-1]) ? offset : -1;
1433 /* foo[/]bar vs foo[] */
1434 return is_dir_sep(*subdir) ? offset + 1 : -1;
1437 int is_inside_dir(const char *dir)
1439 char cwd[PATH_MAX];
1440 if (!dir)
1441 return 0;
1442 if (!getcwd(cwd, sizeof(cwd)))
1443 die_errno("can't find the current directory");
1444 return dir_inside_of(cwd, dir) >= 0;
1447 int is_empty_dir(const char *path)
1449 DIR *dir = opendir(path);
1450 struct dirent *e;
1451 int ret = 1;
1453 if (!dir)
1454 return 0;
1456 while ((e = readdir(dir)) != NULL)
1457 if (!is_dot_or_dotdot(e->d_name)) {
1458 ret = 0;
1459 break;
1462 closedir(dir);
1463 return ret;
1466 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1468 DIR *dir;
1469 struct dirent *e;
1470 int ret = 0, original_len = path->len, len, kept_down = 0;
1471 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1472 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1473 unsigned char submodule_head[20];
1475 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1476 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1477 /* Do not descend and nuke a nested git work tree. */
1478 if (kept_up)
1479 *kept_up = 1;
1480 return 0;
1483 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1484 dir = opendir(path->buf);
1485 if (!dir) {
1486 /* an empty dir could be removed even if it is unreadble */
1487 if (!keep_toplevel)
1488 return rmdir(path->buf);
1489 else
1490 return -1;
1492 if (path->buf[original_len - 1] != '/')
1493 strbuf_addch(path, '/');
1495 len = path->len;
1496 while ((e = readdir(dir)) != NULL) {
1497 struct stat st;
1498 if (is_dot_or_dotdot(e->d_name))
1499 continue;
1501 strbuf_setlen(path, len);
1502 strbuf_addstr(path, e->d_name);
1503 if (lstat(path->buf, &st))
1504 ; /* fall thru */
1505 else if (S_ISDIR(st.st_mode)) {
1506 if (!remove_dir_recurse(path, flag, &kept_down))
1507 continue; /* happy */
1508 } else if (!only_empty && !unlink(path->buf))
1509 continue; /* happy, too */
1511 /* path too long, stat fails, or non-directory still exists */
1512 ret = -1;
1513 break;
1515 closedir(dir);
1517 strbuf_setlen(path, original_len);
1518 if (!ret && !keep_toplevel && !kept_down)
1519 ret = rmdir(path->buf);
1520 else if (kept_up)
1522 * report the uplevel that it is not an error that we
1523 * did not rmdir() our directory.
1525 *kept_up = !ret;
1526 return ret;
1529 int remove_dir_recursively(struct strbuf *path, int flag)
1531 return remove_dir_recurse(path, flag, NULL);
1534 void setup_standard_excludes(struct dir_struct *dir)
1536 const char *path;
1537 char *xdg_path;
1539 dir->exclude_per_dir = ".gitignore";
1540 path = git_path("info/exclude");
1541 if (!excludes_file) {
1542 home_config_paths(NULL, &xdg_path, "ignore");
1543 excludes_file = xdg_path;
1545 if (!access_or_warn(path, R_OK))
1546 add_excludes_from_file(dir, path);
1547 if (excludes_file && !access_or_warn(excludes_file, R_OK))
1548 add_excludes_from_file(dir, excludes_file);
1551 int remove_path(const char *name)
1553 char *slash;
1555 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
1556 return -1;
1558 slash = strrchr(name, '/');
1559 if (slash) {
1560 char *dirs = xstrdup(name);
1561 slash = dirs + (slash - name);
1562 do {
1563 *slash = '\0';
1564 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1565 free(dirs);
1567 return 0;
1570 static int pathspec_item_cmp(const void *a_, const void *b_)
1572 struct pathspec_item *a, *b;
1574 a = (struct pathspec_item *)a_;
1575 b = (struct pathspec_item *)b_;
1576 return strcmp(a->match, b->match);
1579 int init_pathspec(struct pathspec *pathspec, const char **paths)
1581 const char **p = paths;
1582 int i;
1584 memset(pathspec, 0, sizeof(*pathspec));
1585 if (!p)
1586 return 0;
1587 while (*p)
1588 p++;
1589 pathspec->raw = paths;
1590 pathspec->nr = p - paths;
1591 if (!pathspec->nr)
1592 return 0;
1594 pathspec->items = xmalloc(sizeof(struct pathspec_item)*pathspec->nr);
1595 for (i = 0; i < pathspec->nr; i++) {
1596 struct pathspec_item *item = pathspec->items+i;
1597 const char *path = paths[i];
1599 item->match = path;
1600 item->len = strlen(path);
1601 item->flags = 0;
1602 if (limit_pathspec_to_literal()) {
1603 item->nowildcard_len = item->len;
1604 } else {
1605 item->nowildcard_len = simple_length(path);
1606 if (item->nowildcard_len < item->len) {
1607 pathspec->has_wildcard = 1;
1608 if (path[item->nowildcard_len] == '*' &&
1609 no_wildcard(path + item->nowildcard_len + 1))
1610 item->flags |= PATHSPEC_ONESTAR;
1615 qsort(pathspec->items, pathspec->nr,
1616 sizeof(struct pathspec_item), pathspec_item_cmp);
1618 return 0;
1621 void free_pathspec(struct pathspec *pathspec)
1623 free(pathspec->items);
1624 pathspec->items = NULL;
1627 int limit_pathspec_to_literal(void)
1629 static int flag = -1;
1630 if (flag < 0)
1631 flag = git_env_bool(GIT_LITERAL_PATHSPECS_ENVIRONMENT, 0);
1632 return flag;
1636 * Frees memory within dir which was allocated for exclude lists and
1637 * the exclude_stack. Does not free dir itself.
1639 void clear_directory(struct dir_struct *dir)
1641 int i, j;
1642 struct exclude_list_group *group;
1643 struct exclude_list *el;
1644 struct exclude_stack *stk;
1646 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1647 group = &dir->exclude_list_group[i];
1648 for (j = 0; j < group->nr; j++) {
1649 el = &group->el[j];
1650 if (i == EXC_DIRS)
1651 free((char *)el->src);
1652 clear_exclude_list(el);
1654 free(group->el);
1657 stk = dir->exclude_stack;
1658 while (stk) {
1659 struct exclude_stack *prev = stk->prev;
1660 free(stk);
1661 stk = prev;