setup: return correct prefix if worktree is '/'
[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 void free_excludes(struct exclude_list *el)
257 {
258 int i;
259
260 for (i = 0; i < el->nr; i++)
261 free(el->excludes[i]);
262 free(el->excludes);
263
264 el->nr = 0;
265 el->excludes = NULL;
266 }
267
268 int add_excludes_from_file_to_list(const char *fname,
269 const char *base,
270 int baselen,
271 char **buf_p,
272 struct exclude_list *which,
273 int check_index)
274 {
275 struct stat st;
276 int fd, i;
277 size_t size = 0;
278 char *buf, *entry;
279
280 fd = open(fname, O_RDONLY);
281 if (fd < 0 || fstat(fd, &st) < 0) {
282 if (0 <= fd)
283 close(fd);
284 if (!check_index ||
285 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
286 return -1;
287 if (size == 0) {
288 free(buf);
289 return 0;
290 }
291 if (buf[size-1] != '\n') {
292 buf = xrealloc(buf, size+1);
293 buf[size++] = '\n';
294 }
295 }
296 else {
297 size = xsize_t(st.st_size);
298 if (size == 0) {
299 close(fd);
300 return 0;
301 }
302 buf = xmalloc(size+1);
303 if (read_in_full(fd, buf, size) != size) {
304 free(buf);
305 close(fd);
306 return -1;
307 }
308 buf[size++] = '\n';
309 close(fd);
310 }
311
312 if (buf_p)
313 *buf_p = buf;
314 entry = buf;
315 for (i = 0; i < size; i++) {
316 if (buf[i] == '\n') {
317 if (entry != buf + i && entry[0] != '#') {
318 buf[i - (i && buf[i-1] == '\r')] = 0;
319 add_exclude(entry, base, baselen, which);
320 }
321 entry = buf + i + 1;
322 }
323 }
324 return 0;
325 }
326
327 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
328 {
329 if (add_excludes_from_file_to_list(fname, "", 0, NULL,
330 &dir->exclude_list[EXC_FILE], 0) < 0)
331 die("cannot use %s as an exclude file", fname);
332 }
333
334 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
335 {
336 struct exclude_list *el;
337 struct exclude_stack *stk = NULL;
338 int current;
339
340 if ((!dir->exclude_per_dir) ||
341 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
342 return; /* too long a path -- ignore */
343
344 /* Pop the ones that are not the prefix of the path being checked. */
345 el = &dir->exclude_list[EXC_DIRS];
346 while ((stk = dir->exclude_stack) != NULL) {
347 if (stk->baselen <= baselen &&
348 !strncmp(dir->basebuf, base, stk->baselen))
349 break;
350 dir->exclude_stack = stk->prev;
351 while (stk->exclude_ix < el->nr)
352 free(el->excludes[--el->nr]);
353 free(stk->filebuf);
354 free(stk);
355 }
356
357 /* Read from the parent directories and push them down. */
358 current = stk ? stk->baselen : -1;
359 while (current < baselen) {
360 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
361 const char *cp;
362
363 if (current < 0) {
364 cp = base;
365 current = 0;
366 }
367 else {
368 cp = strchr(base + current + 1, '/');
369 if (!cp)
370 die("oops in prep_exclude");
371 cp++;
372 }
373 stk->prev = dir->exclude_stack;
374 stk->baselen = cp - base;
375 stk->exclude_ix = el->nr;
376 memcpy(dir->basebuf + current, base + current,
377 stk->baselen - current);
378 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
379 add_excludes_from_file_to_list(dir->basebuf,
380 dir->basebuf, stk->baselen,
381 &stk->filebuf, el, 1);
382 dir->exclude_stack = stk;
383 current = stk->baselen;
384 }
385 dir->basebuf[baselen] = '\0';
386 }
387
388 /* Scan the list and let the last match determine the fate.
389 * Return 1 for exclude, 0 for include and -1 for undecided.
390 */
391 int excluded_from_list(const char *pathname,
392 int pathlen, const char *basename, int *dtype,
393 struct exclude_list *el)
394 {
395 int i;
396
397 if (el->nr) {
398 for (i = el->nr - 1; 0 <= i; i--) {
399 struct exclude *x = el->excludes[i];
400 const char *exclude = x->pattern;
401 int to_exclude = x->to_exclude;
402
403 if (x->flags & EXC_FLAG_MUSTBEDIR) {
404 if (*dtype == DT_UNKNOWN)
405 *dtype = get_dtype(NULL, pathname, pathlen);
406 if (*dtype != DT_DIR)
407 continue;
408 }
409
410 if (x->flags & EXC_FLAG_NODIR) {
411 /* match basename */
412 if (x->flags & EXC_FLAG_NOWILDCARD) {
413 if (!strcmp_icase(exclude, basename))
414 return to_exclude;
415 } else if (x->flags & EXC_FLAG_ENDSWITH) {
416 if (x->patternlen - 1 <= pathlen &&
417 !strcmp_icase(exclude + 1, pathname + pathlen - x->patternlen + 1))
418 return to_exclude;
419 } else {
420 if (fnmatch_icase(exclude, basename, 0) == 0)
421 return to_exclude;
422 }
423 }
424 else {
425 /* match with FNM_PATHNAME:
426 * exclude has base (baselen long) implicitly
427 * in front of it.
428 */
429 int baselen = x->baselen;
430 if (*exclude == '/')
431 exclude++;
432
433 if (pathlen < baselen ||
434 (baselen && pathname[baselen-1] != '/') ||
435 strncmp_icase(pathname, x->base, baselen))
436 continue;
437
438 if (x->flags & EXC_FLAG_NOWILDCARD) {
439 if (!strcmp_icase(exclude, pathname + baselen))
440 return to_exclude;
441 } else {
442 if (fnmatch_icase(exclude, pathname+baselen,
443 FNM_PATHNAME) == 0)
444 return to_exclude;
445 }
446 }
447 }
448 }
449 return -1; /* undecided */
450 }
451
452 int excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
453 {
454 int pathlen = strlen(pathname);
455 int st;
456 const char *basename = strrchr(pathname, '/');
457 basename = (basename) ? basename+1 : pathname;
458
459 prep_exclude(dir, pathname, basename-pathname);
460 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
461 switch (excluded_from_list(pathname, pathlen, basename,
462 dtype_p, &dir->exclude_list[st])) {
463 case 0:
464 return 0;
465 case 1:
466 return 1;
467 }
468 }
469 return 0;
470 }
471
472 static struct dir_entry *dir_entry_new(const char *pathname, int len)
473 {
474 struct dir_entry *ent;
475
476 ent = xmalloc(sizeof(*ent) + len + 1);
477 ent->len = len;
478 memcpy(ent->name, pathname, len);
479 ent->name[len] = 0;
480 return ent;
481 }
482
483 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
484 {
485 if (cache_name_exists(pathname, len, ignore_case))
486 return NULL;
487
488 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
489 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
490 }
491
492 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
493 {
494 if (!cache_name_is_other(pathname, len))
495 return NULL;
496
497 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
498 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
499 }
500
501 enum exist_status {
502 index_nonexistent = 0,
503 index_directory,
504 index_gitdir
505 };
506
507 /*
508 * Do not use the alphabetically stored index to look up
509 * the directory name; instead, use the case insensitive
510 * name hash.
511 */
512 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
513 {
514 struct cache_entry *ce = index_name_exists(&the_index, dirname, len + 1, ignore_case);
515 unsigned char endchar;
516
517 if (!ce)
518 return index_nonexistent;
519 endchar = ce->name[len];
520
521 /*
522 * The cache_entry structure returned will contain this dirname
523 * and possibly additional path components.
524 */
525 if (endchar == '/')
526 return index_directory;
527
528 /*
529 * If there are no additional path components, then this cache_entry
530 * represents a submodule. Submodules, despite being directories,
531 * are stored in the cache without a closing slash.
532 */
533 if (!endchar && S_ISGITLINK(ce->ce_mode))
534 return index_gitdir;
535
536 /* This should never be hit, but it exists just in case. */
537 return index_nonexistent;
538 }
539
540 /*
541 * The index sorts alphabetically by entry name, which
542 * means that a gitlink sorts as '\0' at the end, while
543 * a directory (which is defined not as an entry, but as
544 * the files it contains) will sort with the '/' at the
545 * end.
546 */
547 static enum exist_status directory_exists_in_index(const char *dirname, int len)
548 {
549 int pos;
550
551 if (ignore_case)
552 return directory_exists_in_index_icase(dirname, len);
553
554 pos = cache_name_pos(dirname, len);
555 if (pos < 0)
556 pos = -pos-1;
557 while (pos < active_nr) {
558 struct cache_entry *ce = active_cache[pos++];
559 unsigned char endchar;
560
561 if (strncmp(ce->name, dirname, len))
562 break;
563 endchar = ce->name[len];
564 if (endchar > '/')
565 break;
566 if (endchar == '/')
567 return index_directory;
568 if (!endchar && S_ISGITLINK(ce->ce_mode))
569 return index_gitdir;
570 }
571 return index_nonexistent;
572 }
573
574 /*
575 * When we find a directory when traversing the filesystem, we
576 * have three distinct cases:
577 *
578 * - ignore it
579 * - see it as a directory
580 * - recurse into it
581 *
582 * and which one we choose depends on a combination of existing
583 * git index contents and the flags passed into the directory
584 * traversal routine.
585 *
586 * Case 1: If we *already* have entries in the index under that
587 * directory name, we always recurse into the directory to see
588 * all the files.
589 *
590 * Case 2: If we *already* have that directory name as a gitlink,
591 * we always continue to see it as a gitlink, regardless of whether
592 * there is an actual git directory there or not (it might not
593 * be checked out as a subproject!)
594 *
595 * Case 3: if we didn't have it in the index previously, we
596 * have a few sub-cases:
597 *
598 * (a) if "show_other_directories" is true, we show it as
599 * just a directory, unless "hide_empty_directories" is
600 * also true and the directory is empty, in which case
601 * we just ignore it entirely.
602 * (b) if it looks like a git directory, and we don't have
603 * 'no_gitlinks' set we treat it as a gitlink, and show it
604 * as a directory.
605 * (c) otherwise, we recurse into it.
606 */
607 enum directory_treatment {
608 show_directory,
609 ignore_directory,
610 recurse_into_directory
611 };
612
613 static enum directory_treatment treat_directory(struct dir_struct *dir,
614 const char *dirname, int len,
615 const struct path_simplify *simplify)
616 {
617 /* The "len-1" is to strip the final '/' */
618 switch (directory_exists_in_index(dirname, len-1)) {
619 case index_directory:
620 return recurse_into_directory;
621
622 case index_gitdir:
623 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
624 return ignore_directory;
625 return show_directory;
626
627 case index_nonexistent:
628 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
629 break;
630 if (!(dir->flags & DIR_NO_GITLINKS)) {
631 unsigned char sha1[20];
632 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
633 return show_directory;
634 }
635 return recurse_into_directory;
636 }
637
638 /* This is the "show_other_directories" case */
639 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
640 return show_directory;
641 if (!read_directory_recursive(dir, dirname, len, 1, simplify))
642 return ignore_directory;
643 return show_directory;
644 }
645
646 /*
647 * This is an inexact early pruning of any recursive directory
648 * reading - if the path cannot possibly be in the pathspec,
649 * return true, and we'll skip it early.
650 */
651 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
652 {
653 if (simplify) {
654 for (;;) {
655 const char *match = simplify->path;
656 int len = simplify->len;
657
658 if (!match)
659 break;
660 if (len > pathlen)
661 len = pathlen;
662 if (!memcmp(path, match, len))
663 return 0;
664 simplify++;
665 }
666 return 1;
667 }
668 return 0;
669 }
670
671 /*
672 * This function tells us whether an excluded path matches a
673 * list of "interesting" pathspecs. That is, whether a path matched
674 * by any of the pathspecs could possibly be ignored by excluding
675 * the specified path. This can happen if:
676 *
677 * 1. the path is mentioned explicitly in the pathspec
678 *
679 * 2. the path is a directory prefix of some element in the
680 * pathspec
681 */
682 static int exclude_matches_pathspec(const char *path, int len,
683 const struct path_simplify *simplify)
684 {
685 if (simplify) {
686 for (; simplify->path; simplify++) {
687 if (len == simplify->len
688 && !memcmp(path, simplify->path, len))
689 return 1;
690 if (len < simplify->len
691 && simplify->path[len] == '/'
692 && !memcmp(path, simplify->path, len))
693 return 1;
694 }
695 }
696 return 0;
697 }
698
699 static int get_index_dtype(const char *path, int len)
700 {
701 int pos;
702 struct cache_entry *ce;
703
704 ce = cache_name_exists(path, len, 0);
705 if (ce) {
706 if (!ce_uptodate(ce))
707 return DT_UNKNOWN;
708 if (S_ISGITLINK(ce->ce_mode))
709 return DT_DIR;
710 /*
711 * Nobody actually cares about the
712 * difference between DT_LNK and DT_REG
713 */
714 return DT_REG;
715 }
716
717 /* Try to look it up as a directory */
718 pos = cache_name_pos(path, len);
719 if (pos >= 0)
720 return DT_UNKNOWN;
721 pos = -pos-1;
722 while (pos < active_nr) {
723 ce = active_cache[pos++];
724 if (strncmp(ce->name, path, len))
725 break;
726 if (ce->name[len] > '/')
727 break;
728 if (ce->name[len] < '/')
729 continue;
730 if (!ce_uptodate(ce))
731 break; /* continue? */
732 return DT_DIR;
733 }
734 return DT_UNKNOWN;
735 }
736
737 static int get_dtype(struct dirent *de, const char *path, int len)
738 {
739 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
740 struct stat st;
741
742 if (dtype != DT_UNKNOWN)
743 return dtype;
744 dtype = get_index_dtype(path, len);
745 if (dtype != DT_UNKNOWN)
746 return dtype;
747 if (lstat(path, &st))
748 return dtype;
749 if (S_ISREG(st.st_mode))
750 return DT_REG;
751 if (S_ISDIR(st.st_mode))
752 return DT_DIR;
753 if (S_ISLNK(st.st_mode))
754 return DT_LNK;
755 return dtype;
756 }
757
758 enum path_treatment {
759 path_ignored,
760 path_handled,
761 path_recurse
762 };
763
764 static enum path_treatment treat_one_path(struct dir_struct *dir,
765 char *path, int *len,
766 const struct path_simplify *simplify,
767 int dtype, struct dirent *de)
768 {
769 int exclude = excluded(dir, path, &dtype);
770 if (exclude && (dir->flags & DIR_COLLECT_IGNORED)
771 && exclude_matches_pathspec(path, *len, simplify))
772 dir_add_ignored(dir, path, *len);
773
774 /*
775 * Excluded? If we don't explicitly want to show
776 * ignored files, ignore it
777 */
778 if (exclude && !(dir->flags & DIR_SHOW_IGNORED))
779 return path_ignored;
780
781 if (dtype == DT_UNKNOWN)
782 dtype = get_dtype(de, path, *len);
783
784 /*
785 * Do we want to see just the ignored files?
786 * We still need to recurse into directories,
787 * even if we don't ignore them, since the
788 * directory may contain files that we do..
789 */
790 if (!exclude && (dir->flags & DIR_SHOW_IGNORED)) {
791 if (dtype != DT_DIR)
792 return path_ignored;
793 }
794
795 switch (dtype) {
796 default:
797 return path_ignored;
798 case DT_DIR:
799 memcpy(path + *len, "/", 2);
800 (*len)++;
801 switch (treat_directory(dir, path, *len, simplify)) {
802 case show_directory:
803 if (exclude != !!(dir->flags
804 & DIR_SHOW_IGNORED))
805 return path_ignored;
806 break;
807 case recurse_into_directory:
808 return path_recurse;
809 case ignore_directory:
810 return path_ignored;
811 }
812 break;
813 case DT_REG:
814 case DT_LNK:
815 break;
816 }
817 return path_handled;
818 }
819
820 static enum path_treatment treat_path(struct dir_struct *dir,
821 struct dirent *de,
822 char *path, int path_max,
823 int baselen,
824 const struct path_simplify *simplify,
825 int *len)
826 {
827 int dtype;
828
829 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
830 return path_ignored;
831 *len = strlen(de->d_name);
832 /* Ignore overly long pathnames! */
833 if (*len + baselen + 8 > path_max)
834 return path_ignored;
835 memcpy(path + baselen, de->d_name, *len + 1);
836 *len += baselen;
837 if (simplify_away(path, *len, simplify))
838 return path_ignored;
839
840 dtype = DTYPE(de);
841 return treat_one_path(dir, path, len, simplify, dtype, de);
842 }
843
844 /*
845 * Read a directory tree. We currently ignore anything but
846 * directories, regular files and symlinks. That's because git
847 * doesn't handle them at all yet. Maybe that will change some
848 * day.
849 *
850 * Also, we ignore the name ".git" (even if it is not a directory).
851 * That likely will not change.
852 */
853 static int read_directory_recursive(struct dir_struct *dir,
854 const char *base, int baselen,
855 int check_only,
856 const struct path_simplify *simplify)
857 {
858 DIR *fdir = opendir(*base ? base : ".");
859 int contents = 0;
860
861 if (fdir) {
862 struct dirent *de;
863 char path[PATH_MAX + 1];
864 memcpy(path, base, baselen);
865
866 while ((de = readdir(fdir)) != NULL) {
867 int len;
868 switch (treat_path(dir, de, path, sizeof(path),
869 baselen, simplify, &len)) {
870 case path_recurse:
871 contents += read_directory_recursive
872 (dir, path, len, 0, simplify);
873 continue;
874 case path_ignored:
875 continue;
876 case path_handled:
877 break;
878 }
879 contents++;
880 if (check_only)
881 goto exit_early;
882 else
883 dir_add_name(dir, path, len);
884 }
885 exit_early:
886 closedir(fdir);
887 }
888
889 return contents;
890 }
891
892 static int cmp_name(const void *p1, const void *p2)
893 {
894 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
895 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
896
897 return cache_name_compare(e1->name, e1->len,
898 e2->name, e2->len);
899 }
900
901 /*
902 * Return the length of the "simple" part of a path match limiter.
903 */
904 static int simple_length(const char *match)
905 {
906 int len = -1;
907
908 for (;;) {
909 unsigned char c = *match++;
910 len++;
911 if (c == '\0' || is_glob_special(c))
912 return len;
913 }
914 }
915
916 static struct path_simplify *create_simplify(const char **pathspec)
917 {
918 int nr, alloc = 0;
919 struct path_simplify *simplify = NULL;
920
921 if (!pathspec)
922 return NULL;
923
924 for (nr = 0 ; ; nr++) {
925 const char *match;
926 if (nr >= alloc) {
927 alloc = alloc_nr(alloc);
928 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
929 }
930 match = *pathspec++;
931 if (!match)
932 break;
933 simplify[nr].path = match;
934 simplify[nr].len = simple_length(match);
935 }
936 simplify[nr].path = NULL;
937 simplify[nr].len = 0;
938 return simplify;
939 }
940
941 static void free_simplify(struct path_simplify *simplify)
942 {
943 free(simplify);
944 }
945
946 static int treat_leading_path(struct dir_struct *dir,
947 const char *path, int len,
948 const struct path_simplify *simplify)
949 {
950 char pathbuf[PATH_MAX];
951 int baselen, blen;
952 const char *cp;
953
954 while (len && path[len - 1] == '/')
955 len--;
956 if (!len)
957 return 1;
958 baselen = 0;
959 while (1) {
960 cp = path + baselen + !!baselen;
961 cp = memchr(cp, '/', path + len - cp);
962 if (!cp)
963 baselen = len;
964 else
965 baselen = cp - path;
966 memcpy(pathbuf, path, baselen);
967 pathbuf[baselen] = '\0';
968 if (!is_directory(pathbuf))
969 return 0;
970 if (simplify_away(pathbuf, baselen, simplify))
971 return 0;
972 blen = baselen;
973 if (treat_one_path(dir, pathbuf, &blen, simplify,
974 DT_DIR, NULL) == path_ignored)
975 return 0; /* do not recurse into it */
976 if (len <= baselen)
977 return 1; /* finished checking */
978 }
979 }
980
981 int read_directory(struct dir_struct *dir, const char *path, int len, const char **pathspec)
982 {
983 struct path_simplify *simplify;
984
985 if (has_symlink_leading_path(path, len))
986 return dir->nr;
987
988 simplify = create_simplify(pathspec);
989 if (!len || treat_leading_path(dir, path, len, simplify))
990 read_directory_recursive(dir, path, len, 0, simplify);
991 free_simplify(simplify);
992 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
993 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
994 return dir->nr;
995 }
996
997 int file_exists(const char *f)
998 {
999 struct stat sb;
1000 return lstat(f, &sb) == 0;
1001 }
1002
1003 /*
1004 * get_relative_cwd() gets the prefix of the current working directory
1005 * relative to 'dir'. If we are not inside 'dir', it returns NULL.
1006 *
1007 * As a convenience, it also returns NULL if 'dir' is already NULL. The
1008 * reason for this behaviour is that it is natural for functions returning
1009 * directory names to return NULL to say "this directory does not exist"
1010 * or "this directory is invalid". These cases are usually handled the
1011 * same as if the cwd is not inside 'dir' at all, so get_relative_cwd()
1012 * returns NULL for both of them.
1013 *
1014 * Most notably, get_relative_cwd(buffer, size, get_git_work_tree())
1015 * unifies the handling of "outside work tree" with "no work tree at all".
1016 */
1017 char *get_relative_cwd(char *buffer, int size, const char *dir)
1018 {
1019 char *cwd = buffer;
1020
1021 if (!dir)
1022 return NULL;
1023 if (!getcwd(buffer, size))
1024 die_errno("can't find the current directory");
1025
1026 if (!is_absolute_path(dir))
1027 dir = make_absolute_path(dir);
1028
1029 while (*dir && *dir == *cwd) {
1030 dir++;
1031 cwd++;
1032 }
1033 if (*dir)
1034 return NULL;
1035 switch (*cwd) {
1036 case '\0':
1037 return cwd;
1038 case '/':
1039 return cwd + 1;
1040 default:
1041 /*
1042 * dir can end with a path separator when it's root
1043 * directory. Return proper prefix in that case.
1044 */
1045 if (dir[-1] == '/')
1046 return cwd;
1047 return NULL;
1048 }
1049 }
1050
1051 * Given two normalized paths (a trailing slash is ok), if subdir is
1052 * outside dir, return -1. Otherwise return the offset in subdir that
1053 * can be used as relative path to dir.
1054 */
1055 int dir_inside_of(const char *subdir, const char *dir)
1056 {
1057 int offset = 0;
1058
1059 assert(dir && subdir && *dir && *subdir);
1060
1061 while (*dir && *subdir && *dir == *subdir) {
1062 dir++;
1063 subdir++;
1064 offset++;
1065 }
1066
1067 /* hel[p]/me vs hel[l]/yeah */
1068 if (*dir && *subdir)
1069 return -1;
1070
1071 if (!*subdir)
1072 return !*dir ? offset : -1; /* same dir */
1073
1074 /* foo/[b]ar vs foo/[] */
1075 if (is_dir_sep(dir[-1]))
1076 return is_dir_sep(subdir[-1]) ? offset : -1;
1077
1078 /* foo[/]bar vs foo[] */
1079 return is_dir_sep(*subdir) ? offset + 1 : -1;
1080 }
1081
1082 int is_inside_dir(const char *dir)
1083 {
1084 char buffer[PATH_MAX];
1085 return get_relative_cwd(buffer, sizeof(buffer), dir) != NULL;
1086 }
1087
1088 int is_empty_dir(const char *path)
1089 {
1090 DIR *dir = opendir(path);
1091 struct dirent *e;
1092 int ret = 1;
1093
1094 if (!dir)
1095 return 0;
1096
1097 while ((e = readdir(dir)) != NULL)
1098 if (!is_dot_or_dotdot(e->d_name)) {
1099 ret = 0;
1100 break;
1101 }
1102
1103 closedir(dir);
1104 return ret;
1105 }
1106
1107 int remove_dir_recursively(struct strbuf *path, int flag)
1108 {
1109 DIR *dir;
1110 struct dirent *e;
1111 int ret = 0, original_len = path->len, len;
1112 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1113 unsigned char submodule_head[20];
1114
1115 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1116 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head))
1117 /* Do not descend and nuke a nested git work tree. */
1118 return 0;
1119
1120 dir = opendir(path->buf);
1121 if (!dir)
1122 return -1;
1123 if (path->buf[original_len - 1] != '/')
1124 strbuf_addch(path, '/');
1125
1126 len = path->len;
1127 while ((e = readdir(dir)) != NULL) {
1128 struct stat st;
1129 if (is_dot_or_dotdot(e->d_name))
1130 continue;
1131
1132 strbuf_setlen(path, len);
1133 strbuf_addstr(path, e->d_name);
1134 if (lstat(path->buf, &st))
1135 ; /* fall thru */
1136 else if (S_ISDIR(st.st_mode)) {
1137 if (!remove_dir_recursively(path, only_empty))
1138 continue; /* happy */
1139 } else if (!only_empty && !unlink(path->buf))
1140 continue; /* happy, too */
1141
1142 /* path too long, stat fails, or non-directory still exists */
1143 ret = -1;
1144 break;
1145 }
1146 closedir(dir);
1147
1148 strbuf_setlen(path, original_len);
1149 if (!ret)
1150 ret = rmdir(path->buf);
1151 return ret;
1152 }
1153
1154 void setup_standard_excludes(struct dir_struct *dir)
1155 {
1156 const char *path;
1157
1158 dir->exclude_per_dir = ".gitignore";
1159 path = git_path("info/exclude");
1160 if (!access(path, R_OK))
1161 add_excludes_from_file(dir, path);
1162 if (excludes_file && !access(excludes_file, R_OK))
1163 add_excludes_from_file(dir, excludes_file);
1164 }
1165
1166 int remove_path(const char *name)
1167 {
1168 char *slash;
1169
1170 if (unlink(name) && errno != ENOENT)
1171 return -1;
1172
1173 slash = strrchr(name, '/');
1174 if (slash) {
1175 char *dirs = xstrdup(name);
1176 slash = dirs + (slash - name);
1177 do {
1178 *slash = '\0';
1179 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1180 free(dirs);
1181 }
1182 return 0;
1183 }
1184