Add case insensitivity support when using git ls-files
[git/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
7 */
8 #include "cache.h"
9 #include "dir.h"
10 #include "refs.h"
11
12 struct path_simplify {
13 int len;
14 const char *path;
15 };
16
17 static int read_directory_recursive(struct dir_struct *dir, const char *path, int len,
18 int check_only, const struct path_simplify *simplify);
19 static int get_dtype(struct dirent *de, const char *path, int len);
20
21 /* helper string functions with support for the ignore_case flag */
22 int strcmp_icase(const char *a, const char *b)
23 {
24 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
25 }
26
27 int strncmp_icase(const char *a, const char *b, size_t count)
28 {
29 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
30 }
31
32 int fnmatch_icase(const char *pattern, const char *string, int flags)
33 {
34 return fnmatch(pattern, string, flags | (ignore_case ? FNM_CASEFOLD : 0));
35 }
36
37 static int common_prefix(const char **pathspec)
38 {
39 const char *path, *slash, *next;
40 int prefix;
41
42 if (!pathspec)
43 return 0;
44
45 path = *pathspec;
46 slash = strrchr(path, '/');
47 if (!slash)
48 return 0;
49
50 /*
51 * The first 'prefix' characters of 'path' are common leading
52 * path components among the pathspecs we have seen so far,
53 * including the trailing slash.
54 */
55 prefix = slash - path + 1;
56 while ((next = *++pathspec) != NULL) {
57 int len, last_matching_slash = -1;
58 for (len = 0; len < prefix && next[len] == path[len]; len++)
59 if (next[len] == '/')
60 last_matching_slash = len;
61 if (len == prefix)
62 continue;
63 if (last_matching_slash < 0)
64 return 0;
65 prefix = last_matching_slash + 1;
66 }
67 return prefix;
68 }
69
70 int fill_directory(struct dir_struct *dir, const char **pathspec)
71 {
72 const char *path;
73 int len;
74
75 /*
76 * Calculate common prefix for the pathspec, and
77 * use that to optimize the directory walk
78 */
79 len = common_prefix(pathspec);
80 path = "";
81
82 if (len)
83 path = xmemdupz(*pathspec, len);
84
85 /* Read the directory and prune it */
86 read_directory(dir, path, len, pathspec);
87 return len;
88 }
89
90 /*
91 * Does 'match' match the given name?
92 * A match is found if
93 *
94 * (1) the 'match' string is leading directory of 'name', or
95 * (2) the 'match' string is a wildcard and matches 'name', or
96 * (3) the 'match' string is exactly the same as 'name'.
97 *
98 * and the return value tells which case it was.
99 *
100 * It returns 0 when there is no match.
101 */
102 static int match_one(const char *match, const char *name, int namelen)
103 {
104 int matchlen;
105
106 /* If the match was just the prefix, we matched */
107 if (!*match)
108 return MATCHED_RECURSIVELY;
109
110 if (ignore_case) {
111 for (;;) {
112 unsigned char c1 = tolower(*match);
113 unsigned char c2 = tolower(*name);
114 if (c1 == '\0' || is_glob_special(c1))
115 break;
116 if (c1 != c2)
117 return 0;
118 match++;
119 name++;
120 namelen--;
121 }
122 } else {
123 for (;;) {
124 unsigned char c1 = *match;
125 unsigned char c2 = *name;
126 if (c1 == '\0' || is_glob_special(c1))
127 break;
128 if (c1 != c2)
129 return 0;
130 match++;
131 name++;
132 namelen--;
133 }
134 }
135
136
137 /*
138 * If we don't match the matchstring exactly,
139 * we need to match by fnmatch
140 */
141 matchlen = strlen(match);
142 if (strncmp_icase(match, name, matchlen))
143 return !fnmatch_icase(match, name, 0) ? MATCHED_FNMATCH : 0;
144
145 if (namelen == matchlen)
146 return MATCHED_EXACTLY;
147 if (match[matchlen-1] == '/' || name[matchlen] == '/')
148 return MATCHED_RECURSIVELY;
149 return 0;
150 }
151
152 /*
153 * Given a name and a list of pathspecs, see if the name matches
154 * any of the pathspecs. The caller is also interested in seeing
155 * all pathspec matches some names it calls this function with
156 * (otherwise the user could have mistyped the unmatched pathspec),
157 * and a mark is left in seen[] array for pathspec element that
158 * actually matched anything.
159 */
160 int match_pathspec(const char **pathspec, const char *name, int namelen,
161 int prefix, char *seen)
162 {
163 int i, retval = 0;
164
165 if (!pathspec)
166 return 1;
167
168 name += prefix;
169 namelen -= prefix;
170
171 for (i = 0; pathspec[i] != NULL; i++) {
172 int how;
173 const char *match = pathspec[i] + prefix;
174 if (seen && seen[i] == MATCHED_EXACTLY)
175 continue;
176 how = match_one(match, name, namelen);
177 if (how) {
178 if (retval < how)
179 retval = how;
180 if (seen && seen[i] < how)
181 seen[i] = how;
182 }
183 }
184 return retval;
185 }
186
187 static int no_wildcard(const char *string)
188 {
189 return string[strcspn(string, "*?[{\\")] == '\0';
190 }
191
192 void add_exclude(const char *string, const char *base,
193 int baselen, struct exclude_list *which)
194 {
195 struct exclude *x;
196 size_t len;
197 int to_exclude = 1;
198 int flags = 0;
199
200 if (*string == '!') {
201 to_exclude = 0;
202 string++;
203 }
204 len = strlen(string);
205 if (len && string[len - 1] == '/') {
206 char *s;
207 x = xmalloc(sizeof(*x) + len);
208 s = (char *)(x+1);
209 memcpy(s, string, len - 1);
210 s[len - 1] = '\0';
211 string = s;
212 x->pattern = s;
213 flags = EXC_FLAG_MUSTBEDIR;
214 } else {
215 x = xmalloc(sizeof(*x));
216 x->pattern = string;
217 }
218 x->to_exclude = to_exclude;
219 x->patternlen = strlen(string);
220 x->base = base;
221 x->baselen = baselen;
222 x->flags = flags;
223 if (!strchr(string, '/'))
224 x->flags |= EXC_FLAG_NODIR;
225 if (no_wildcard(string))
226 x->flags |= EXC_FLAG_NOWILDCARD;
227 if (*string == '*' && no_wildcard(string+1))
228 x->flags |= EXC_FLAG_ENDSWITH;
229 ALLOC_GROW(which->excludes, which->nr + 1, which->alloc);
230 which->excludes[which->nr++] = x;
231 }
232
233 static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
234 {
235 int pos, len;
236 unsigned long sz;
237 enum object_type type;
238 void *data;
239 struct index_state *istate = &the_index;
240
241 len = strlen(path);
242 pos = index_name_pos(istate, path, len);
243 if (pos < 0)
244 return NULL;
245 if (!ce_skip_worktree(istate->cache[pos]))
246 return NULL;
247 data = read_sha1_file(istate->cache[pos]->sha1, &type, &sz);
248 if (!data || type != OBJ_BLOB) {
249 free(data);
250 return NULL;
251 }
252 *size = xsize_t(sz);
253 return data;
254 }
255
256 int add_excludes_from_file_to_list(const char *fname,
257 const char *base,
258 int baselen,
259 char **buf_p,
260 struct exclude_list *which,
261 int check_index)
262 {
263 struct stat st;
264 int fd, i;
265 size_t size;
266 char *buf, *entry;
267
268 fd = open(fname, O_RDONLY);
269 if (fd < 0 || fstat(fd, &st) < 0) {
270 if (0 <= fd)
271 close(fd);
272 if (!check_index ||
273 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
274 return -1;
275 if (size == 0) {
276 free(buf);
277 return 0;
278 }
279 if (buf[size-1] != '\n') {
280 buf = xrealloc(buf, size+1);
281 buf[size++] = '\n';
282 }
283 }
284 else {
285 size = xsize_t(st.st_size);
286 if (size == 0) {
287 close(fd);
288 return 0;
289 }
290 buf = xmalloc(size+1);
291 if (read_in_full(fd, buf, size) != size) {
292 free(buf);
293 close(fd);
294 return -1;
295 }
296 buf[size++] = '\n';
297 close(fd);
298 }
299
300 if (buf_p)
301 *buf_p = buf;
302 entry = buf;
303 for (i = 0; i < size; i++) {
304 if (buf[i] == '\n') {
305 if (entry != buf + i && entry[0] != '#') {
306 buf[i - (i && buf[i-1] == '\r')] = 0;
307 add_exclude(entry, base, baselen, which);
308 }
309 entry = buf + i + 1;
310 }
311 }
312 return 0;
313 }
314
315 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
316 {
317 if (add_excludes_from_file_to_list(fname, "", 0, NULL,
318 &dir->exclude_list[EXC_FILE], 0) < 0)
319 die("cannot use %s as an exclude file", fname);
320 }
321
322 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
323 {
324 struct exclude_list *el;
325 struct exclude_stack *stk = NULL;
326 int current;
327
328 if ((!dir->exclude_per_dir) ||
329 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
330 return; /* too long a path -- ignore */
331
332 /* Pop the ones that are not the prefix of the path being checked. */
333 el = &dir->exclude_list[EXC_DIRS];
334 while ((stk = dir->exclude_stack) != NULL) {
335 if (stk->baselen <= baselen &&
336 !strncmp(dir->basebuf, base, stk->baselen))
337 break;
338 dir->exclude_stack = stk->prev;
339 while (stk->exclude_ix < el->nr)
340 free(el->excludes[--el->nr]);
341 free(stk->filebuf);
342 free(stk);
343 }
344
345 /* Read from the parent directories and push them down. */
346 current = stk ? stk->baselen : -1;
347 while (current < baselen) {
348 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
349 const char *cp;
350
351 if (current < 0) {
352 cp = base;
353 current = 0;
354 }
355 else {
356 cp = strchr(base + current + 1, '/');
357 if (!cp)
358 die("oops in prep_exclude");
359 cp++;
360 }
361 stk->prev = dir->exclude_stack;
362 stk->baselen = cp - base;
363 stk->exclude_ix = el->nr;
364 memcpy(dir->basebuf + current, base + current,
365 stk->baselen - current);
366 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
367 add_excludes_from_file_to_list(dir->basebuf,
368 dir->basebuf, stk->baselen,
369 &stk->filebuf, el, 1);
370 dir->exclude_stack = stk;
371 current = stk->baselen;
372 }
373 dir->basebuf[baselen] = '\0';
374 }
375
376 /* Scan the list and let the last match determine the fate.
377 * Return 1 for exclude, 0 for include and -1 for undecided.
378 */
379 int excluded_from_list(const char *pathname,
380 int pathlen, const char *basename, int *dtype,
381 struct exclude_list *el)
382 {
383 int i;
384
385 if (el->nr) {
386 for (i = el->nr - 1; 0 <= i; i--) {
387 struct exclude *x = el->excludes[i];
388 const char *exclude = x->pattern;
389 int to_exclude = x->to_exclude;
390
391 if (x->flags & EXC_FLAG_MUSTBEDIR) {
392 if (!dtype) {
393 if (!prefixcmp(pathname, exclude))
394 return to_exclude;
395 else
396 continue;
397 }
398 if (*dtype == DT_UNKNOWN)
399 *dtype = get_dtype(NULL, pathname, pathlen);
400 if (*dtype != DT_DIR)
401 continue;
402 }
403
404 if (x->flags & EXC_FLAG_NODIR) {
405 /* match basename */
406 if (x->flags & EXC_FLAG_NOWILDCARD) {
407 if (!strcmp_icase(exclude, basename))
408 return to_exclude;
409 } else if (x->flags & EXC_FLAG_ENDSWITH) {
410 if (x->patternlen - 1 <= pathlen &&
411 !strcmp_icase(exclude + 1, pathname + pathlen - x->patternlen + 1))
412 return to_exclude;
413 } else {
414 if (fnmatch_icase(exclude, basename, 0) == 0)
415 return to_exclude;
416 }
417 }
418 else {
419 /* match with FNM_PATHNAME:
420 * exclude has base (baselen long) implicitly
421 * in front of it.
422 */
423 int baselen = x->baselen;
424 if (*exclude == '/')
425 exclude++;
426
427 if (pathlen < baselen ||
428 (baselen && pathname[baselen-1] != '/') ||
429 strncmp_icase(pathname, x->base, baselen))
430 continue;
431
432 if (x->flags & EXC_FLAG_NOWILDCARD) {
433 if (!strcmp_icase(exclude, pathname + baselen))
434 return to_exclude;
435 } else {
436 if (fnmatch_icase(exclude, pathname+baselen,
437 FNM_PATHNAME) == 0)
438 return to_exclude;
439 }
440 }
441 }
442 }
443 return -1; /* undecided */
444 }
445
446 int excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
447 {
448 int pathlen = strlen(pathname);
449 int st;
450 const char *basename = strrchr(pathname, '/');
451 basename = (basename) ? basename+1 : pathname;
452
453 prep_exclude(dir, pathname, basename-pathname);
454 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
455 switch (excluded_from_list(pathname, pathlen, basename,
456 dtype_p, &dir->exclude_list[st])) {
457 case 0:
458 return 0;
459 case 1:
460 return 1;
461 }
462 }
463 return 0;
464 }
465
466 static struct dir_entry *dir_entry_new(const char *pathname, int len)
467 {
468 struct dir_entry *ent;
469
470 ent = xmalloc(sizeof(*ent) + len + 1);
471 ent->len = len;
472 memcpy(ent->name, pathname, len);
473 ent->name[len] = 0;
474 return ent;
475 }
476
477 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
478 {
479 if (cache_name_exists(pathname, len, ignore_case))
480 return NULL;
481
482 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
483 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
484 }
485
486 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
487 {
488 if (!cache_name_is_other(pathname, len))
489 return NULL;
490
491 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
492 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
493 }
494
495 enum exist_status {
496 index_nonexistent = 0,
497 index_directory,
498 index_gitdir
499 };
500
501 /*
502 * Do not use the alphabetically stored index to look up
503 * the directory name; instead, use the case insensitive
504 * name hash.
505 */
506 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
507 {
508 struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
509 unsigned char endchar;
510
511 if (!ce)
512 return index_nonexistent;
513 endchar = ce->name[len];
514
515 /*
516 * The cache_entry structure returned will contain this dirname
517 * and possibly additional path components.
518 */
519 if (endchar == '/')
520 return index_directory;
521
522 /*
523 * If there are no additional path components, then this cache_entry
524 * represents a submodule. Submodules, despite being directories,
525 * are stored in the cache without a closing slash.
526 */
527 if (!endchar && S_ISGITLINK(ce->ce_mode))
528 return index_gitdir;
529
530 /* This should never be hit, but it exists just in case. */
531 return index_nonexistent;
532 }
533
534 /*
535 * The index sorts alphabetically by entry name, which
536 * means that a gitlink sorts as '\0' at the end, while
537 * a directory (which is defined not as an entry, but as
538 * the files it contains) will sort with the '/' at the
539 * end.
540 */
541 static enum exist_status directory_exists_in_index(const char *dirname, int len)
542 {
543 int pos;
544
545 if (ignore_case)
546 return directory_exists_in_index_icase(dirname, len);
547
548 pos = cache_name_pos(dirname, len);
549 if (pos < 0)
550 pos = -pos-1;
551 while (pos < active_nr) {
552 struct cache_entry *ce = active_cache[pos++];
553 unsigned char endchar;
554
555 if (strncmp(ce->name, dirname, len))
556 break;
557 endchar = ce->name[len];
558 if (endchar > '/')
559 break;
560 if (endchar == '/')
561 return index_directory;
562 if (!endchar && S_ISGITLINK(ce->ce_mode))
563 return index_gitdir;
564 }
565 return index_nonexistent;
566 }
567
568 /*
569 * When we find a directory when traversing the filesystem, we
570 * have three distinct cases:
571 *
572 * - ignore it
573 * - see it as a directory
574 * - recurse into it
575 *
576 * and which one we choose depends on a combination of existing
577 * git index contents and the flags passed into the directory
578 * traversal routine.
579 *
580 * Case 1: If we *already* have entries in the index under that
581 * directory name, we always recurse into the directory to see
582 * all the files.
583 *
584 * Case 2: If we *already* have that directory name as a gitlink,
585 * we always continue to see it as a gitlink, regardless of whether
586 * there is an actual git directory there or not (it might not
587 * be checked out as a subproject!)
588 *
589 * Case 3: if we didn't have it in the index previously, we
590 * have a few sub-cases:
591 *
592 * (a) if "show_other_directories" is true, we show it as
593 * just a directory, unless "hide_empty_directories" is
594 * also true and the directory is empty, in which case
595 * we just ignore it entirely.
596 * (b) if it looks like a git directory, and we don't have
597 * 'no_gitlinks' set we treat it as a gitlink, and show it
598 * as a directory.
599 * (c) otherwise, we recurse into it.
600 */
601 enum directory_treatment {
602 show_directory,
603 ignore_directory,
604 recurse_into_directory
605 };
606
607 static enum directory_treatment treat_directory(struct dir_struct *dir,
608 const char *dirname, int len,
609 const struct path_simplify *simplify)
610 {
611 /* The "len-1" is to strip the final '/' */
612 switch (directory_exists_in_index(dirname, len-1)) {
613 case index_directory:
614 return recurse_into_directory;
615
616 case index_gitdir:
617 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
618 return ignore_directory;
619 return show_directory;
620
621 case index_nonexistent:
622 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
623 break;
624 if (!(dir->flags & DIR_NO_GITLINKS)) {
625 unsigned char sha1[20];
626 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
627 return show_directory;
628 }
629 return recurse_into_directory;
630 }
631
632 /* This is the "show_other_directories" case */
633 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
634 return show_directory;
635 if (!read_directory_recursive(dir, dirname, len, 1, simplify))
636 return ignore_directory;
637 return show_directory;
638 }
639
640 /*
641 * This is an inexact early pruning of any recursive directory
642 * reading - if the path cannot possibly be in the pathspec,
643 * return true, and we'll skip it early.
644 */
645 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
646 {
647 if (simplify) {
648 for (;;) {
649 const char *match = simplify->path;
650 int len = simplify->len;
651
652 if (!match)
653 break;
654 if (len > pathlen)
655 len = pathlen;
656 if (!memcmp(path, match, len))
657 return 0;
658 simplify++;
659 }
660 return 1;
661 }
662 return 0;
663 }
664
665 /*
666 * This function tells us whether an excluded path matches a
667 * list of "interesting" pathspecs. That is, whether a path matched
668 * by any of the pathspecs could possibly be ignored by excluding
669 * the specified path. This can happen if:
670 *
671 * 1. the path is mentioned explicitly in the pathspec
672 *
673 * 2. the path is a directory prefix of some element in the
674 * pathspec
675 */
676 static int exclude_matches_pathspec(const char *path, int len,
677 const struct path_simplify *simplify)
678 {
679 if (simplify) {
680 for (; simplify->path; simplify++) {
681 if (len == simplify->len
682 && !memcmp(path, simplify->path, len))
683 return 1;
684 if (len < simplify->len
685 && simplify->path[len] == '/'
686 && !memcmp(path, simplify->path, len))
687 return 1;
688 }
689 }
690 return 0;
691 }
692
693 static int get_index_dtype(const char *path, int len)
694 {
695 int pos;
696 struct cache_entry *ce;
697
698 ce = cache_name_exists(path, len, 0);
699 if (ce) {
700 if (!ce_uptodate(ce))
701 return DT_UNKNOWN;
702 if (S_ISGITLINK(ce->ce_mode))
703 return DT_DIR;
704 /*
705 * Nobody actually cares about the
706 * difference between DT_LNK and DT_REG
707 */
708 return DT_REG;
709 }
710
711 /* Try to look it up as a directory */
712 pos = cache_name_pos(path, len);
713 if (pos >= 0)
714 return DT_UNKNOWN;
715 pos = -pos-1;
716 while (pos < active_nr) {
717 ce = active_cache[pos++];
718 if (strncmp(ce->name, path, len))
719 break;
720 if (ce->name[len] > '/')
721 break;
722 if (ce->name[len] < '/')
723 continue;
724 if (!ce_uptodate(ce))
725 break; /* continue? */
726 return DT_DIR;
727 }
728 return DT_UNKNOWN;
729 }
730
731 static int get_dtype(struct dirent *de, const char *path, int len)
732 {
733 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
734 struct stat st;
735
736 if (dtype != DT_UNKNOWN)
737 return dtype;
738 dtype = get_index_dtype(path, len);
739 if (dtype != DT_UNKNOWN)
740 return dtype;
741 if (lstat(path, &st))
742 return dtype;
743 if (S_ISREG(st.st_mode))
744 return DT_REG;
745 if (S_ISDIR(st.st_mode))
746 return DT_DIR;
747 if (S_ISLNK(st.st_mode))
748 return DT_LNK;
749 return dtype;
750 }
751
752 enum path_treatment {
753 path_ignored,
754 path_handled,
755 path_recurse
756 };
757
758 static enum path_treatment treat_one_path(struct dir_struct *dir,
759 char *path, int *len,
760 const struct path_simplify *simplify,
761 int dtype, struct dirent *de)
762 {
763 int exclude = excluded(dir, path, &dtype);
764 if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
765 && exclude_matches_pathspec(path, *len, simplify))
766 dir_add_ignored(dir, path, *len);
767
768 /*
769 * Excluded? If we don't explicitly want to show
770 * ignored files, ignore it
771 */
772 if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
773 return path_ignored;
774
775 if (dtype == DT_UNKNOWN)
776 dtype = get_dtype(de, path, *len);
777
778 /*
779 * Do we want to see just the ignored files?
780 * We still need to recurse into directories,
781 * even if we don't ignore them, since the
782 * directory may contain files that we do..
783 */
784 if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
785 if (dtype != DT_DIR)
786 return path_ignored;
787 }
788
789 switch (dtype) {
790 default:
791 return path_ignored;
792 case DT_DIR:
793 memcpy(path + *len, "/", 2);
794 (*len)++;
795 switch (treat_directory(dir, path, *len, simplify)) {
796 case show_directory:
797 if (exclude != !!(dir->flags
798 & DIR_SHOW_IGNORED))
799 return path_ignored;
800 break;
801 case recurse_into_directory:
802 return path_recurse;
803 case ignore_directory:
804 return path_ignored;
805 }
806 break;
807 case DT_REG:
808 case DT_LNK:
809 break;
810 }
811 return path_handled;
812 }
813
814 static enum path_treatment treat_path(struct dir_struct *dir,
815 struct dirent *de,
816 char *path, int path_max,
817 int baselen,
818 const struct path_simplify *simplify,
819 int *len)
820 {
821 int dtype;
822
823 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
824 return path_ignored;
825 *len = strlen(de->d_name);
826 /* Ignore overly long pathnames! */
827 if (*len + baselen + 8 > path_max)
828 return path_ignored;
829 memcpy(path + baselen, de->d_name, *len + 1);
830 *len += baselen;
831 if (simplify_away(path, *len, simplify))
832 return path_ignored;
833
834 dtype = DTYPE(de);
835 return treat_one_path(dir, path, len, simplify, dtype, de);
836 }
837
838 /*
839 * Read a directory tree. We currently ignore anything but
840 * directories, regular files and symlinks. That's because git
841 * doesn't handle them at all yet. Maybe that will change some
842 * day.
843 *
844 * Also, we ignore the name ".git" (even if it is not a directory).
845 * That likely will not change.
846 */
847 static int read_directory_recursive(struct dir_struct *dir,
848 const char *base, int baselen,
849 int check_only,
850 const struct path_simplify *simplify)
851 {
852 DIR *fdir = opendir(*base ? base : ".");
853 int contents = 0;
854
855 if (fdir) {
856 struct dirent *de;
857 char path[PATH_MAX + 1];
858 memcpy(path, base, baselen);
859
860 while ((de = readdir(fdir)) != NULL) {
861 int len;
862 switch (treat_path(dir, de, path, sizeof(path),
863 baselen, simplify, &len)) {
864 case path_recurse:
865 contents += read_directory_recursive
866 (dir, path, len, 0, simplify);
867 continue;
868 case path_ignored:
869 continue;
870 case path_handled:
871 break;
872 }
873 contents++;
874 if (check_only)
875 goto exit_early;
876 else
877 dir_add_name(dir, path, len);
878 }
879 exit_early:
880 closedir(fdir);
881 }
882
883 return contents;
884 }
885
886 static int cmp_name(const void *p1, const void *p2)
887 {
888 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
889 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
890
891 return cache_name_compare(e1->name, e1->len,
892 e2->name, e2->len);
893 }
894
895 /*
896 * Return the length of the "simple" part of a path match limiter.
897 */
898 static int simple_length(const char *match)
899 {
900 int len = -1;
901
902 for (;;) {
903 unsigned char c = *match++;
904 len++;
905 if (c == '\0' || is_glob_special(c))
906 return len;
907 }
908 }
909
910 static struct path_simplify *create_simplify(const char **pathspec)
911 {
912 int nr, alloc = 0;
913 struct path_simplify *simplify = NULL;
914
915 if (!pathspec)
916 return NULL;
917
918 for (nr = 0 ; ; nr++) {
919 const char *match;
920 if (nr >= alloc) {
921 alloc = alloc_nr(alloc);
922 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
923 }
924 match = *pathspec++;
925 if (!match)
926 break;
927 simplify[nr].path = match;
928 simplify[nr].len = simple_length(match);
929 }
930 simplify[nr].path = NULL;
931 simplify[nr].len = 0;
932 return simplify;
933 }
934
935 static void free_simplify(struct path_simplify *simplify)
936 {
937 free(simplify);
938 }
939
940 static int treat_leading_path(struct dir_struct *dir,
941 const char *path, int len,
942 const struct path_simplify *simplify)
943 {
944 char pathbuf[PATH_MAX];
945 int baselen, blen;
946 const char *cp;
947
948 while (len && path[len - 1] == '/')
949 len--;
950 if (!len)
951 return 1;
952 baselen = 0;
953 while (1) {
954 cp = path + baselen + !!baselen;
955 cp = memchr(cp, '/', path + len - cp);
956 if (!cp)
957 baselen = len;
958 else
959 baselen = cp - path;
960 memcpy(pathbuf, path, baselen);
961 pathbuf[baselen] = '\0';
962 if (!is_directory(pathbuf))
963 return 0;
964 if (simplify_away(pathbuf, baselen, simplify))
965 return 0;
966 blen = baselen;
967 if (treat_one_path(dir, pathbuf, &blen, simplify,
968 DT_DIR, NULL) == path_ignored)
969 return 0; /* do not recurse into it */
970 if (len <= baselen)
971 return 1; /* finished checking */
972 }
973 }
974
975 int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
976 {
977 struct path_simplify *simplify;
978
979 if (has_symlink_leading_path(path, len))
980 return dir->nr;
981
982 simplify = create_simplify(pathspec);
983 if (!len || treat_leading_path(dir, path, len, simplify))
984 read_directory_recursive(dir, path, len, 0, simplify);
985 free_simplify(simplify);
986 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
987 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
988 return dir->nr;
989 }
990
991 int file_exists(const char *f)
992 {
993 struct stat sb;
994 return lstat(f, &sb) == 0;
995 }
996
997 /*
998 * get_relative_cwd() gets the prefix of the current working directory
999 * relative to 'dir'. If we are not inside 'dir', it returns NULL.
1000 *
1001 * As a convenience, it also returns NULL if 'dir' is already NULL. The
1002 * reason for this behaviour is that it is natural for functions returning
1003 * directory names to return NULL to say "this directory does not exist"
1004 * or "this directory is invalid". These cases are usually handled the
1005 * same as if the cwd is not inside 'dir' at all, so get_relative_cwd()
1006 * returns NULL for both of them.
1007 *
1008 * Most notably, get_relative_cwd(buffer, size, get_git_work_tree())
1009 * unifies the handling of "outside work tree" with "no work tree at all".
1010 */
1011 char *get_relative_cwd(char *buffer, int size, const char *dir)
1012 {
1013 char *cwd = buffer;
1014
1015 if (!dir)
1016 return NULL;
1017 if (!getcwd(buffer, size))
1018 die_errno("can't find the current directory");
1019
1020 if (!is_absolute_path(dir))
1021 dir = make_absolute_path(dir);
1022
1023 while (*dir && *dir == *cwd) {
1024 dir++;
1025 cwd++;
1026 }
1027 if (*dir)
1028 return NULL;
1029 switch (*cwd) {
1030 case '\0':
1031 return cwd;
1032 case '/':
1033 return cwd + 1;
1034 default:
1035 return NULL;
1036 }
1037 }
1038
1039 int is_inside_dir(const char *dir)
1040 {
1041 char buffer[PATH_MAX];
1042 return get_relative_cwd(buffer, sizeof(buffer), dir) != NULL;
1043 }
1044
1045 int is_empty_dir(const char *path)
1046 {
1047 DIR *dir = opendir(path);
1048 struct dirent *e;
1049 int ret = 1;
1050
1051 if (!dir)
1052 return 0;
1053
1054 while ((e = readdir(dir)) != NULL)
1055 if (!is_dot_or_dotdot(e->d_name)) {
1056 ret = 0;
1057 break;
1058 }
1059
1060 closedir(dir);
1061 return ret;
1062 }
1063
1064 int remove_dir_recursively(struct strbuf *path, int flag)
1065 {
1066 DIR *dir;
1067 struct dirent *e;
1068 int ret = 0, original_len = path->len, len;
1069 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1070 unsigned char submodule_head[20];
1071
1072 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1073 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head))
1074 /* Do not descend and nuke a nested git work tree. */
1075 return 0;
1076
1077 dir = opendir(path->buf);
1078 if (!dir)
1079 return -1;
1080 if (path->buf[original_len - 1] != '/')
1081 strbuf_addch(path, '/');
1082
1083 len = path->len;
1084 while ((e = readdir(dir)) != NULL) {
1085 struct stat st;
1086 if (is_dot_or_dotdot(e->d_name))
1087 continue;
1088
1089 strbuf_setlen(path, len);
1090 strbuf_addstr(path, e->d_name);
1091 if (lstat(path->buf, &st))
1092 ; /* fall thru */
1093 else if (S_ISDIR(st.st_mode)) {
1094 if (!remove_dir_recursively(path, only_empty))
1095 continue; /* happy */
1096 } else if (!only_empty && !unlink(path->buf))
1097 continue; /* happy, too */
1098
1099 /* path too long, stat fails, or non-directory still exists */
1100 ret = -1;
1101 break;
1102 }
1103 closedir(dir);
1104
1105 strbuf_setlen(path, original_len);
1106 if (!ret)
1107 ret = rmdir(path->buf);
1108 return ret;
1109 }
1110
1111 void setup_standard_excludes(struct dir_struct *dir)
1112 {
1113 const char *path;
1114
1115 dir->exclude_per_dir = ".gitignore";
1116 path = git_path("info/exclude");
1117 if (!access(path, R_OK))
1118 add_excludes_from_file(dir, path);
1119 if (excludes_file && !access(excludes_file, R_OK))
1120 add_excludes_from_file(dir, excludes_file);
1121 }
1122
1123 int remove_path(const char *name)
1124 {
1125 char *slash;
1126
1127 if (unlink(name) && errno != ENOENT)
1128 return -1;
1129
1130 slash = strrchr(name, '/');
1131 if (slash) {
1132 char *dirs = xstrdup(name);
1133 slash = dirs + (slash - name);
1134 do {
1135 *slash = '\0';
1136 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1137 free(dirs);
1138 }
1139 return 0;
1140 }
1141