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