pathspec: allow querying for attributes
[git/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * See Documentation/technical/api-directory-listing.txt
6 *
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10 #include "cache.h"
11 #include "dir.h"
12 #include "attr.h"
13 #include "refs.h"
14 #include "wildmatch.h"
15 #include "pathspec.h"
16 #include "utf8.h"
17 #include "varint.h"
18 #include "ewah/ewok.h"
19
20 /*
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.
25 */
26 enum path_treatment {
27 path_none = 0,
28 path_recurse,
29 path_excluded,
30 path_untracked
31 };
32
33 /*
34 * Support data structure for our opendir/readdir/closedir wrappers
35 */
36 struct cached_dir {
37 DIR *fdir;
38 struct untracked_cache_dir *untracked;
39 int nr_files;
40 int nr_dirs;
41
42 struct dirent *de;
43 const char *file;
44 struct untracked_cache_dir *ucd;
45 };
46
47 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
48 const char *path, int len, struct untracked_cache_dir *untracked,
49 int check_only, const struct pathspec *pathspec);
50 static int get_dtype(struct dirent *de, const char *path, int len);
51
52 int fspathcmp(const char *a, const char *b)
53 {
54 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
55 }
56
57 int fspathncmp(const char *a, const char *b, size_t count)
58 {
59 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
60 }
61
62 int git_fnmatch(const struct pathspec_item *item,
63 const char *pattern, const char *string,
64 int prefix)
65 {
66 if (prefix > 0) {
67 if (ps_strncmp(item, pattern, string, prefix))
68 return WM_NOMATCH;
69 pattern += prefix;
70 string += prefix;
71 }
72 if (item->flags & PATHSPEC_ONESTAR) {
73 int pattern_len = strlen(++pattern);
74 int string_len = strlen(string);
75 return string_len < pattern_len ||
76 ps_strcmp(item, pattern,
77 string + string_len - pattern_len);
78 }
79 if (item->magic & PATHSPEC_GLOB)
80 return wildmatch(pattern, string,
81 WM_PATHNAME |
82 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
83 NULL);
84 else
85 /* wildmatch has not learned no FNM_PATHNAME mode yet */
86 return wildmatch(pattern, string,
87 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
88 NULL);
89 }
90
91 static int fnmatch_icase_mem(const char *pattern, int patternlen,
92 const char *string, int stringlen,
93 int flags)
94 {
95 int match_status;
96 struct strbuf pat_buf = STRBUF_INIT;
97 struct strbuf str_buf = STRBUF_INIT;
98 const char *use_pat = pattern;
99 const char *use_str = string;
100
101 if (pattern[patternlen]) {
102 strbuf_add(&pat_buf, pattern, patternlen);
103 use_pat = pat_buf.buf;
104 }
105 if (string[stringlen]) {
106 strbuf_add(&str_buf, string, stringlen);
107 use_str = str_buf.buf;
108 }
109
110 if (ignore_case)
111 flags |= WM_CASEFOLD;
112 match_status = wildmatch(use_pat, use_str, flags, NULL);
113
114 strbuf_release(&pat_buf);
115 strbuf_release(&str_buf);
116
117 return match_status;
118 }
119
120 static size_t common_prefix_len(const struct pathspec *pathspec)
121 {
122 int n;
123 size_t max = 0;
124
125 /*
126 * ":(icase)path" is treated as a pathspec full of
127 * wildcard. In other words, only prefix is considered common
128 * prefix. If the pathspec is abc/foo abc/bar, running in
129 * subdir xyz, the common prefix is still xyz, not xuz/abc as
130 * in non-:(icase).
131 */
132 GUARD_PATHSPEC(pathspec,
133 PATHSPEC_FROMTOP |
134 PATHSPEC_MAXDEPTH |
135 PATHSPEC_LITERAL |
136 PATHSPEC_GLOB |
137 PATHSPEC_ICASE |
138 PATHSPEC_EXCLUDE |
139 PATHSPEC_ATTR);
140
141 for (n = 0; n < pathspec->nr; n++) {
142 size_t i = 0, len = 0, item_len;
143 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
144 continue;
145 if (pathspec->items[n].magic & PATHSPEC_ICASE)
146 item_len = pathspec->items[n].prefix;
147 else
148 item_len = pathspec->items[n].nowildcard_len;
149 while (i < item_len && (n == 0 || i < max)) {
150 char c = pathspec->items[n].match[i];
151 if (c != pathspec->items[0].match[i])
152 break;
153 if (c == '/')
154 len = i + 1;
155 i++;
156 }
157 if (n == 0 || len < max) {
158 max = len;
159 if (!max)
160 break;
161 }
162 }
163 return max;
164 }
165
166 /*
167 * Returns a copy of the longest leading path common among all
168 * pathspecs.
169 */
170 char *common_prefix(const struct pathspec *pathspec)
171 {
172 unsigned long len = common_prefix_len(pathspec);
173
174 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
175 }
176
177 int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
178 {
179 const char *prefix;
180 size_t prefix_len;
181
182 /*
183 * Calculate common prefix for the pathspec, and
184 * use that to optimize the directory walk
185 */
186 prefix_len = common_prefix_len(pathspec);
187 prefix = prefix_len ? pathspec->items[0].match : "";
188
189 /* Read the directory and prune it */
190 read_directory(dir, prefix, prefix_len, pathspec);
191
192 return prefix_len;
193 }
194
195 int within_depth(const char *name, int namelen,
196 int depth, int max_depth)
197 {
198 const char *cp = name, *cpe = name + namelen;
199
200 while (cp < cpe) {
201 if (*cp++ != '/')
202 continue;
203 depth++;
204 if (depth > max_depth)
205 return 0;
206 }
207 return 1;
208 }
209
210 #define DO_MATCH_EXCLUDE (1<<0)
211 #define DO_MATCH_DIRECTORY (1<<1)
212 #define DO_MATCH_SUBMODULE (1<<2)
213
214 static int match_attrs(const char *name, int namelen,
215 const struct pathspec_item *item)
216 {
217 int i;
218
219 git_check_attr(name, item->attr_check);
220 for (i = 0; i < item->attr_match_nr; i++) {
221 const char *value;
222 int matched;
223 enum attr_match_mode match_mode;
224
225 value = item->attr_check->items[i].value;
226 match_mode = item->attr_match[i].match_mode;
227
228 if (ATTR_TRUE(value))
229 matched = (match_mode == MATCH_SET);
230 else if (ATTR_FALSE(value))
231 matched = (match_mode == MATCH_UNSET);
232 else if (ATTR_UNSET(value))
233 matched = (match_mode == MATCH_UNSPECIFIED);
234 else
235 matched = (match_mode == MATCH_VALUE &&
236 !strcmp(item->attr_match[i].value, value));
237 if (!matched)
238 return 0;
239 }
240
241 return 1;
242 }
243
244 /*
245 * Does 'match' match the given name?
246 * A match is found if
247 *
248 * (1) the 'match' string is leading directory of 'name', or
249 * (2) the 'match' string is a wildcard and matches 'name', or
250 * (3) the 'match' string is exactly the same as 'name'.
251 *
252 * and the return value tells which case it was.
253 *
254 * It returns 0 when there is no match.
255 */
256 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
257 const char *name, int namelen, unsigned flags)
258 {
259 /* name/namelen has prefix cut off by caller */
260 const char *match = item->match + prefix;
261 int matchlen = item->len - prefix;
262
263 /*
264 * The normal call pattern is:
265 * 1. prefix = common_prefix_len(ps);
266 * 2. prune something, or fill_directory
267 * 3. match_pathspec()
268 *
269 * 'prefix' at #1 may be shorter than the command's prefix and
270 * it's ok for #2 to match extra files. Those extras will be
271 * trimmed at #3.
272 *
273 * Suppose the pathspec is 'foo' and '../bar' running from
274 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
275 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
276 * user does not want XYZ/foo, only the "foo" part should be
277 * case-insensitive. We need to filter out XYZ/foo here. In
278 * other words, we do not trust the caller on comparing the
279 * prefix part when :(icase) is involved. We do exact
280 * comparison ourselves.
281 *
282 * Normally the caller (common_prefix_len() in fact) does
283 * _exact_ matching on name[-prefix+1..-1] and we do not need
284 * to check that part. Be defensive and check it anyway, in
285 * case common_prefix_len is changed, or a new caller is
286 * introduced that does not use common_prefix_len.
287 *
288 * If the penalty turns out too high when prefix is really
289 * long, maybe change it to
290 * strncmp(match, name, item->prefix - prefix)
291 */
292 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
293 strncmp(item->match, name - prefix, item->prefix))
294 return 0;
295
296 if (item->attr_match_nr && !match_attrs(name, namelen, item))
297 return 0;
298
299 /* If the match was just the prefix, we matched */
300 if (!*match)
301 return MATCHED_RECURSIVELY;
302
303 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
304 if (matchlen == namelen)
305 return MATCHED_EXACTLY;
306
307 if (match[matchlen-1] == '/' || name[matchlen] == '/')
308 return MATCHED_RECURSIVELY;
309 } else if ((flags & DO_MATCH_DIRECTORY) &&
310 match[matchlen - 1] == '/' &&
311 namelen == matchlen - 1 &&
312 !ps_strncmp(item, match, name, namelen))
313 return MATCHED_EXACTLY;
314
315 if (item->nowildcard_len < item->len &&
316 !git_fnmatch(item, match, name,
317 item->nowildcard_len - prefix))
318 return MATCHED_FNMATCH;
319
320 /* Perform checks to see if "name" is a super set of the pathspec */
321 if (flags & DO_MATCH_SUBMODULE) {
322 /* name is a literal prefix of the pathspec */
323 if ((namelen < matchlen) &&
324 (match[namelen] == '/') &&
325 !ps_strncmp(item, match, name, namelen))
326 return MATCHED_RECURSIVELY;
327
328 /* name" doesn't match up to the first wild character */
329 if (item->nowildcard_len < item->len &&
330 ps_strncmp(item, match, name,
331 item->nowildcard_len - prefix))
332 return 0;
333
334 /*
335 * Here is where we would perform a wildmatch to check if
336 * "name" can be matched as a directory (or a prefix) against
337 * the pathspec. Since wildmatch doesn't have this capability
338 * at the present we have to punt and say that it is a match,
339 * potentially returning a false positive
340 * The submodules themselves will be able to perform more
341 * accurate matching to determine if the pathspec matches.
342 */
343 return MATCHED_RECURSIVELY;
344 }
345
346 return 0;
347 }
348
349 /*
350 * Given a name and a list of pathspecs, returns the nature of the
351 * closest (i.e. most specific) match of the name to any of the
352 * pathspecs.
353 *
354 * The caller typically calls this multiple times with the same
355 * pathspec and seen[] array but with different name/namelen
356 * (e.g. entries from the index) and is interested in seeing if and
357 * how each pathspec matches all the names it calls this function
358 * with. A mark is left in the seen[] array for each pathspec element
359 * indicating the closest type of match that element achieved, so if
360 * seen[n] remains zero after multiple invocations, that means the nth
361 * pathspec did not match any names, which could indicate that the
362 * user mistyped the nth pathspec.
363 */
364 static int do_match_pathspec(const struct pathspec *ps,
365 const char *name, int namelen,
366 int prefix, char *seen,
367 unsigned flags)
368 {
369 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
370
371 GUARD_PATHSPEC(ps,
372 PATHSPEC_FROMTOP |
373 PATHSPEC_MAXDEPTH |
374 PATHSPEC_LITERAL |
375 PATHSPEC_GLOB |
376 PATHSPEC_ICASE |
377 PATHSPEC_EXCLUDE |
378 PATHSPEC_ATTR);
379
380 if (!ps->nr) {
381 if (!ps->recursive ||
382 !(ps->magic & PATHSPEC_MAXDEPTH) ||
383 ps->max_depth == -1)
384 return MATCHED_RECURSIVELY;
385
386 if (within_depth(name, namelen, 0, ps->max_depth))
387 return MATCHED_EXACTLY;
388 else
389 return 0;
390 }
391
392 name += prefix;
393 namelen -= prefix;
394
395 for (i = ps->nr - 1; i >= 0; i--) {
396 int how;
397
398 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
399 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
400 continue;
401
402 if (seen && seen[i] == MATCHED_EXACTLY)
403 continue;
404 /*
405 * Make exclude patterns optional and never report
406 * "pathspec ':(exclude)foo' matches no files"
407 */
408 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
409 seen[i] = MATCHED_FNMATCH;
410 how = match_pathspec_item(ps->items+i, prefix, name,
411 namelen, flags);
412 if (ps->recursive &&
413 (ps->magic & PATHSPEC_MAXDEPTH) &&
414 ps->max_depth != -1 &&
415 how && how != MATCHED_FNMATCH) {
416 int len = ps->items[i].len;
417 if (name[len] == '/')
418 len++;
419 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
420 how = MATCHED_EXACTLY;
421 else
422 how = 0;
423 }
424 if (how) {
425 if (retval < how)
426 retval = how;
427 if (seen && seen[i] < how)
428 seen[i] = how;
429 }
430 }
431 return retval;
432 }
433
434 int match_pathspec(const struct pathspec *ps,
435 const char *name, int namelen,
436 int prefix, char *seen, int is_dir)
437 {
438 int positive, negative;
439 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
440 positive = do_match_pathspec(ps, name, namelen,
441 prefix, seen, flags);
442 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
443 return positive;
444 negative = do_match_pathspec(ps, name, namelen,
445 prefix, seen,
446 flags | DO_MATCH_EXCLUDE);
447 return negative ? 0 : positive;
448 }
449
450 /**
451 * Check if a submodule is a superset of the pathspec
452 */
453 int submodule_path_match(const struct pathspec *ps,
454 const char *submodule_name,
455 char *seen)
456 {
457 int matched = do_match_pathspec(ps, submodule_name,
458 strlen(submodule_name),
459 0, seen,
460 DO_MATCH_DIRECTORY |
461 DO_MATCH_SUBMODULE);
462 return matched;
463 }
464
465 int report_path_error(const char *ps_matched,
466 const struct pathspec *pathspec,
467 const char *prefix)
468 {
469 /*
470 * Make sure all pathspec matched; otherwise it is an error.
471 */
472 int num, errors = 0;
473 for (num = 0; num < pathspec->nr; num++) {
474 int other, found_dup;
475
476 if (ps_matched[num])
477 continue;
478 /*
479 * The caller might have fed identical pathspec
480 * twice. Do not barf on such a mistake.
481 * FIXME: parse_pathspec should have eliminated
482 * duplicate pathspec.
483 */
484 for (found_dup = other = 0;
485 !found_dup && other < pathspec->nr;
486 other++) {
487 if (other == num || !ps_matched[other])
488 continue;
489 if (!strcmp(pathspec->items[other].original,
490 pathspec->items[num].original))
491 /*
492 * Ok, we have a match already.
493 */
494 found_dup = 1;
495 }
496 if (found_dup)
497 continue;
498
499 error("pathspec '%s' did not match any file(s) known to git.",
500 pathspec->items[num].original);
501 errors++;
502 }
503 return errors;
504 }
505
506 /*
507 * Return the length of the "simple" part of a path match limiter.
508 */
509 int simple_length(const char *match)
510 {
511 int len = -1;
512
513 for (;;) {
514 unsigned char c = *match++;
515 len++;
516 if (c == '\0' || is_glob_special(c))
517 return len;
518 }
519 }
520
521 int no_wildcard(const char *string)
522 {
523 return string[simple_length(string)] == '\0';
524 }
525
526 void parse_exclude_pattern(const char **pattern,
527 int *patternlen,
528 unsigned *flags,
529 int *nowildcardlen)
530 {
531 const char *p = *pattern;
532 size_t i, len;
533
534 *flags = 0;
535 if (*p == '!') {
536 *flags |= EXC_FLAG_NEGATIVE;
537 p++;
538 }
539 len = strlen(p);
540 if (len && p[len - 1] == '/') {
541 len--;
542 *flags |= EXC_FLAG_MUSTBEDIR;
543 }
544 for (i = 0; i < len; i++) {
545 if (p[i] == '/')
546 break;
547 }
548 if (i == len)
549 *flags |= EXC_FLAG_NODIR;
550 *nowildcardlen = simple_length(p);
551 /*
552 * we should have excluded the trailing slash from 'p' too,
553 * but that's one more allocation. Instead just make sure
554 * nowildcardlen does not exceed real patternlen
555 */
556 if (*nowildcardlen > len)
557 *nowildcardlen = len;
558 if (*p == '*' && no_wildcard(p + 1))
559 *flags |= EXC_FLAG_ENDSWITH;
560 *pattern = p;
561 *patternlen = len;
562 }
563
564 void add_exclude(const char *string, const char *base,
565 int baselen, struct exclude_list *el, int srcpos)
566 {
567 struct exclude *x;
568 int patternlen;
569 unsigned flags;
570 int nowildcardlen;
571
572 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
573 if (flags & EXC_FLAG_MUSTBEDIR) {
574 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
575 } else {
576 x = xmalloc(sizeof(*x));
577 x->pattern = string;
578 }
579 x->patternlen = patternlen;
580 x->nowildcardlen = nowildcardlen;
581 x->base = base;
582 x->baselen = baselen;
583 x->flags = flags;
584 x->srcpos = srcpos;
585 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
586 el->excludes[el->nr++] = x;
587 x->el = el;
588 }
589
590 static void *read_skip_worktree_file_from_index(const char *path, size_t *size,
591 struct sha1_stat *sha1_stat)
592 {
593 int pos, len;
594 unsigned long sz;
595 enum object_type type;
596 void *data;
597
598 len = strlen(path);
599 pos = cache_name_pos(path, len);
600 if (pos < 0)
601 return NULL;
602 if (!ce_skip_worktree(active_cache[pos]))
603 return NULL;
604 data = read_sha1_file(active_cache[pos]->oid.hash, &type, &sz);
605 if (!data || type != OBJ_BLOB) {
606 free(data);
607 return NULL;
608 }
609 *size = xsize_t(sz);
610 if (sha1_stat) {
611 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
612 hashcpy(sha1_stat->sha1, active_cache[pos]->oid.hash);
613 }
614 return data;
615 }
616
617 /*
618 * Frees memory within el which was allocated for exclude patterns and
619 * the file buffer. Does not free el itself.
620 */
621 void clear_exclude_list(struct exclude_list *el)
622 {
623 int i;
624
625 for (i = 0; i < el->nr; i++)
626 free(el->excludes[i]);
627 free(el->excludes);
628 free(el->filebuf);
629
630 memset(el, 0, sizeof(*el));
631 }
632
633 static void trim_trailing_spaces(char *buf)
634 {
635 char *p, *last_space = NULL;
636
637 for (p = buf; *p; p++)
638 switch (*p) {
639 case ' ':
640 if (!last_space)
641 last_space = p;
642 break;
643 case '\\':
644 p++;
645 if (!*p)
646 return;
647 /* fallthrough */
648 default:
649 last_space = NULL;
650 }
651
652 if (last_space)
653 *last_space = '\0';
654 }
655
656 /*
657 * Given a subdirectory name and "dir" of the current directory,
658 * search the subdir in "dir" and return it, or create a new one if it
659 * does not exist in "dir".
660 *
661 * If "name" has the trailing slash, it'll be excluded in the search.
662 */
663 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
664 struct untracked_cache_dir *dir,
665 const char *name, int len)
666 {
667 int first, last;
668 struct untracked_cache_dir *d;
669 if (!dir)
670 return NULL;
671 if (len && name[len - 1] == '/')
672 len--;
673 first = 0;
674 last = dir->dirs_nr;
675 while (last > first) {
676 int cmp, next = (last + first) >> 1;
677 d = dir->dirs[next];
678 cmp = strncmp(name, d->name, len);
679 if (!cmp && strlen(d->name) > len)
680 cmp = -1;
681 if (!cmp)
682 return d;
683 if (cmp < 0) {
684 last = next;
685 continue;
686 }
687 first = next+1;
688 }
689
690 uc->dir_created++;
691 FLEX_ALLOC_MEM(d, name, name, len);
692
693 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
694 memmove(dir->dirs + first + 1, dir->dirs + first,
695 (dir->dirs_nr - first) * sizeof(*dir->dirs));
696 dir->dirs_nr++;
697 dir->dirs[first] = d;
698 return d;
699 }
700
701 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
702 {
703 int i;
704 dir->valid = 0;
705 dir->untracked_nr = 0;
706 for (i = 0; i < dir->dirs_nr; i++)
707 do_invalidate_gitignore(dir->dirs[i]);
708 }
709
710 static void invalidate_gitignore(struct untracked_cache *uc,
711 struct untracked_cache_dir *dir)
712 {
713 uc->gitignore_invalidated++;
714 do_invalidate_gitignore(dir);
715 }
716
717 static void invalidate_directory(struct untracked_cache *uc,
718 struct untracked_cache_dir *dir)
719 {
720 int i;
721 uc->dir_invalidated++;
722 dir->valid = 0;
723 dir->untracked_nr = 0;
724 for (i = 0; i < dir->dirs_nr; i++)
725 dir->dirs[i]->recurse = 0;
726 }
727
728 /*
729 * Given a file with name "fname", read it (either from disk, or from
730 * the index if "check_index" is non-zero), parse it and store the
731 * exclude rules in "el".
732 *
733 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
734 * stat data from disk (only valid if add_excludes returns zero). If
735 * ss_valid is non-zero, "ss" must contain good value as input.
736 */
737 static int add_excludes(const char *fname, const char *base, int baselen,
738 struct exclude_list *el, int check_index,
739 struct sha1_stat *sha1_stat)
740 {
741 struct stat st;
742 int fd, i, lineno = 1;
743 size_t size = 0;
744 char *buf, *entry;
745
746 fd = open(fname, O_RDONLY);
747 if (fd < 0 || fstat(fd, &st) < 0) {
748 if (errno != ENOENT)
749 warn_on_inaccessible(fname);
750 if (0 <= fd)
751 close(fd);
752 if (!check_index ||
753 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL)
754 return -1;
755 if (size == 0) {
756 free(buf);
757 return 0;
758 }
759 if (buf[size-1] != '\n') {
760 buf = xrealloc(buf, st_add(size, 1));
761 buf[size++] = '\n';
762 }
763 } else {
764 size = xsize_t(st.st_size);
765 if (size == 0) {
766 if (sha1_stat) {
767 fill_stat_data(&sha1_stat->stat, &st);
768 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
769 sha1_stat->valid = 1;
770 }
771 close(fd);
772 return 0;
773 }
774 buf = xmallocz(size);
775 if (read_in_full(fd, buf, size) != size) {
776 free(buf);
777 close(fd);
778 return -1;
779 }
780 buf[size++] = '\n';
781 close(fd);
782 if (sha1_stat) {
783 int pos;
784 if (sha1_stat->valid &&
785 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st))
786 ; /* no content change, ss->sha1 still good */
787 else if (check_index &&
788 (pos = cache_name_pos(fname, strlen(fname))) >= 0 &&
789 !ce_stage(active_cache[pos]) &&
790 ce_uptodate(active_cache[pos]) &&
791 !would_convert_to_git(fname))
792 hashcpy(sha1_stat->sha1,
793 active_cache[pos]->oid.hash);
794 else
795 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
796 fill_stat_data(&sha1_stat->stat, &st);
797 sha1_stat->valid = 1;
798 }
799 }
800
801 el->filebuf = buf;
802
803 if (skip_utf8_bom(&buf, size))
804 size -= buf - el->filebuf;
805
806 entry = buf;
807
808 for (i = 0; i < size; i++) {
809 if (buf[i] == '\n') {
810 if (entry != buf + i && entry[0] != '#') {
811 buf[i - (i && buf[i-1] == '\r')] = 0;
812 trim_trailing_spaces(entry);
813 add_exclude(entry, base, baselen, el, lineno);
814 }
815 lineno++;
816 entry = buf + i + 1;
817 }
818 }
819 return 0;
820 }
821
822 int add_excludes_from_file_to_list(const char *fname, const char *base,
823 int baselen, struct exclude_list *el,
824 int check_index)
825 {
826 return add_excludes(fname, base, baselen, el, check_index, NULL);
827 }
828
829 struct exclude_list *add_exclude_list(struct dir_struct *dir,
830 int group_type, const char *src)
831 {
832 struct exclude_list *el;
833 struct exclude_list_group *group;
834
835 group = &dir->exclude_list_group[group_type];
836 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
837 el = &group->el[group->nr++];
838 memset(el, 0, sizeof(*el));
839 el->src = src;
840 return el;
841 }
842
843 /*
844 * Used to set up core.excludesfile and .git/info/exclude lists.
845 */
846 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
847 struct sha1_stat *sha1_stat)
848 {
849 struct exclude_list *el;
850 /*
851 * catch setup_standard_excludes() that's called before
852 * dir->untracked is assigned. That function behaves
853 * differently when dir->untracked is non-NULL.
854 */
855 if (!dir->untracked)
856 dir->unmanaged_exclude_files++;
857 el = add_exclude_list(dir, EXC_FILE, fname);
858 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0)
859 die("cannot use %s as an exclude file", fname);
860 }
861
862 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
863 {
864 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
865 add_excludes_from_file_1(dir, fname, NULL);
866 }
867
868 int match_basename(const char *basename, int basenamelen,
869 const char *pattern, int prefix, int patternlen,
870 unsigned flags)
871 {
872 if (prefix == patternlen) {
873 if (patternlen == basenamelen &&
874 !fspathncmp(pattern, basename, basenamelen))
875 return 1;
876 } else if (flags & EXC_FLAG_ENDSWITH) {
877 /* "*literal" matching against "fooliteral" */
878 if (patternlen - 1 <= basenamelen &&
879 !fspathncmp(pattern + 1,
880 basename + basenamelen - (patternlen - 1),
881 patternlen - 1))
882 return 1;
883 } else {
884 if (fnmatch_icase_mem(pattern, patternlen,
885 basename, basenamelen,
886 0) == 0)
887 return 1;
888 }
889 return 0;
890 }
891
892 int match_pathname(const char *pathname, int pathlen,
893 const char *base, int baselen,
894 const char *pattern, int prefix, int patternlen,
895 unsigned flags)
896 {
897 const char *name;
898 int namelen;
899
900 /*
901 * match with FNM_PATHNAME; the pattern has base implicitly
902 * in front of it.
903 */
904 if (*pattern == '/') {
905 pattern++;
906 patternlen--;
907 prefix--;
908 }
909
910 /*
911 * baselen does not count the trailing slash. base[] may or
912 * may not end with a trailing slash though.
913 */
914 if (pathlen < baselen + 1 ||
915 (baselen && pathname[baselen] != '/') ||
916 fspathncmp(pathname, base, baselen))
917 return 0;
918
919 namelen = baselen ? pathlen - baselen - 1 : pathlen;
920 name = pathname + pathlen - namelen;
921
922 if (prefix) {
923 /*
924 * if the non-wildcard part is longer than the
925 * remaining pathname, surely it cannot match.
926 */
927 if (prefix > namelen)
928 return 0;
929
930 if (fspathncmp(pattern, name, prefix))
931 return 0;
932 pattern += prefix;
933 patternlen -= prefix;
934 name += prefix;
935 namelen -= prefix;
936
937 /*
938 * If the whole pattern did not have a wildcard,
939 * then our prefix match is all we need; we
940 * do not need to call fnmatch at all.
941 */
942 if (!patternlen && !namelen)
943 return 1;
944 }
945
946 return fnmatch_icase_mem(pattern, patternlen,
947 name, namelen,
948 WM_PATHNAME) == 0;
949 }
950
951 /*
952 * Scan the given exclude list in reverse to see whether pathname
953 * should be ignored. The first match (i.e. the last on the list), if
954 * any, determines the fate. Returns the exclude_list element which
955 * matched, or NULL for undecided.
956 */
957 static struct exclude *last_exclude_matching_from_list(const char *pathname,
958 int pathlen,
959 const char *basename,
960 int *dtype,
961 struct exclude_list *el)
962 {
963 struct exclude *exc = NULL; /* undecided */
964 int i;
965
966 if (!el->nr)
967 return NULL; /* undefined */
968
969 for (i = el->nr - 1; 0 <= i; i--) {
970 struct exclude *x = el->excludes[i];
971 const char *exclude = x->pattern;
972 int prefix = x->nowildcardlen;
973
974 if (x->flags & EXC_FLAG_MUSTBEDIR) {
975 if (*dtype == DT_UNKNOWN)
976 *dtype = get_dtype(NULL, pathname, pathlen);
977 if (*dtype != DT_DIR)
978 continue;
979 }
980
981 if (x->flags & EXC_FLAG_NODIR) {
982 if (match_basename(basename,
983 pathlen - (basename - pathname),
984 exclude, prefix, x->patternlen,
985 x->flags)) {
986 exc = x;
987 break;
988 }
989 continue;
990 }
991
992 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
993 if (match_pathname(pathname, pathlen,
994 x->base, x->baselen ? x->baselen - 1 : 0,
995 exclude, prefix, x->patternlen, x->flags)) {
996 exc = x;
997 break;
998 }
999 }
1000 return exc;
1001 }
1002
1003 /*
1004 * Scan the list and let the last match determine the fate.
1005 * Return 1 for exclude, 0 for include and -1 for undecided.
1006 */
1007 int is_excluded_from_list(const char *pathname,
1008 int pathlen, const char *basename, int *dtype,
1009 struct exclude_list *el)
1010 {
1011 struct exclude *exclude;
1012 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
1013 if (exclude)
1014 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1015 return -1; /* undecided */
1016 }
1017
1018 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1019 const char *pathname, int pathlen, const char *basename,
1020 int *dtype_p)
1021 {
1022 int i, j;
1023 struct exclude_list_group *group;
1024 struct exclude *exclude;
1025 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1026 group = &dir->exclude_list_group[i];
1027 for (j = group->nr - 1; j >= 0; j--) {
1028 exclude = last_exclude_matching_from_list(
1029 pathname, pathlen, basename, dtype_p,
1030 &group->el[j]);
1031 if (exclude)
1032 return exclude;
1033 }
1034 }
1035 return NULL;
1036 }
1037
1038 /*
1039 * Loads the per-directory exclude list for the substring of base
1040 * which has a char length of baselen.
1041 */
1042 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
1043 {
1044 struct exclude_list_group *group;
1045 struct exclude_list *el;
1046 struct exclude_stack *stk = NULL;
1047 struct untracked_cache_dir *untracked;
1048 int current;
1049
1050 group = &dir->exclude_list_group[EXC_DIRS];
1051
1052 /*
1053 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1054 * which originate from directories not in the prefix of the
1055 * path being checked.
1056 */
1057 while ((stk = dir->exclude_stack) != NULL) {
1058 if (stk->baselen <= baselen &&
1059 !strncmp(dir->basebuf.buf, base, stk->baselen))
1060 break;
1061 el = &group->el[dir->exclude_stack->exclude_ix];
1062 dir->exclude_stack = stk->prev;
1063 dir->exclude = NULL;
1064 free((char *)el->src); /* see strbuf_detach() below */
1065 clear_exclude_list(el);
1066 free(stk);
1067 group->nr--;
1068 }
1069
1070 /* Skip traversing into sub directories if the parent is excluded */
1071 if (dir->exclude)
1072 return;
1073
1074 /*
1075 * Lazy initialization. All call sites currently just
1076 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1077 * them seems lots of work for little benefit.
1078 */
1079 if (!dir->basebuf.buf)
1080 strbuf_init(&dir->basebuf, PATH_MAX);
1081
1082 /* Read from the parent directories and push them down. */
1083 current = stk ? stk->baselen : -1;
1084 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1085 if (dir->untracked)
1086 untracked = stk ? stk->ucd : dir->untracked->root;
1087 else
1088 untracked = NULL;
1089
1090 while (current < baselen) {
1091 const char *cp;
1092 struct sha1_stat sha1_stat;
1093
1094 stk = xcalloc(1, sizeof(*stk));
1095 if (current < 0) {
1096 cp = base;
1097 current = 0;
1098 } else {
1099 cp = strchr(base + current + 1, '/');
1100 if (!cp)
1101 die("oops in prep_exclude");
1102 cp++;
1103 untracked =
1104 lookup_untracked(dir->untracked, untracked,
1105 base + current,
1106 cp - base - current);
1107 }
1108 stk->prev = dir->exclude_stack;
1109 stk->baselen = cp - base;
1110 stk->exclude_ix = group->nr;
1111 stk->ucd = untracked;
1112 el = add_exclude_list(dir, EXC_DIRS, NULL);
1113 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1114 assert(stk->baselen == dir->basebuf.len);
1115
1116 /* Abort if the directory is excluded */
1117 if (stk->baselen) {
1118 int dt = DT_DIR;
1119 dir->basebuf.buf[stk->baselen - 1] = 0;
1120 dir->exclude = last_exclude_matching_from_lists(dir,
1121 dir->basebuf.buf, stk->baselen - 1,
1122 dir->basebuf.buf + current, &dt);
1123 dir->basebuf.buf[stk->baselen - 1] = '/';
1124 if (dir->exclude &&
1125 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1126 dir->exclude = NULL;
1127 if (dir->exclude) {
1128 dir->exclude_stack = stk;
1129 return;
1130 }
1131 }
1132
1133 /* Try to read per-directory file */
1134 hashclr(sha1_stat.sha1);
1135 sha1_stat.valid = 0;
1136 if (dir->exclude_per_dir &&
1137 /*
1138 * If we know that no files have been added in
1139 * this directory (i.e. valid_cached_dir() has
1140 * been executed and set untracked->valid) ..
1141 */
1142 (!untracked || !untracked->valid ||
1143 /*
1144 * .. and .gitignore does not exist before
1145 * (i.e. null exclude_sha1). Then we can skip
1146 * loading .gitignore, which would result in
1147 * ENOENT anyway.
1148 */
1149 !is_null_sha1(untracked->exclude_sha1))) {
1150 /*
1151 * dir->basebuf gets reused by the traversal, but we
1152 * need fname to remain unchanged to ensure the src
1153 * member of each struct exclude correctly
1154 * back-references its source file. Other invocations
1155 * of add_exclude_list provide stable strings, so we
1156 * strbuf_detach() and free() here in the caller.
1157 */
1158 struct strbuf sb = STRBUF_INIT;
1159 strbuf_addbuf(&sb, &dir->basebuf);
1160 strbuf_addstr(&sb, dir->exclude_per_dir);
1161 el->src = strbuf_detach(&sb, NULL);
1162 add_excludes(el->src, el->src, stk->baselen, el, 1,
1163 untracked ? &sha1_stat : NULL);
1164 }
1165 /*
1166 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1167 * will first be called in valid_cached_dir() then maybe many
1168 * times more in last_exclude_matching(). When the cache is
1169 * used, last_exclude_matching() will not be called and
1170 * reading .gitignore content will be a waste.
1171 *
1172 * So when it's called by valid_cached_dir() and we can get
1173 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1174 * modified on work tree), we could delay reading the
1175 * .gitignore content until we absolutely need it in
1176 * last_exclude_matching(). Be careful about ignore rule
1177 * order, though, if you do that.
1178 */
1179 if (untracked &&
1180 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1181 invalidate_gitignore(dir->untracked, untracked);
1182 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1183 }
1184 dir->exclude_stack = stk;
1185 current = stk->baselen;
1186 }
1187 strbuf_setlen(&dir->basebuf, baselen);
1188 }
1189
1190 /*
1191 * Loads the exclude lists for the directory containing pathname, then
1192 * scans all exclude lists to determine whether pathname is excluded.
1193 * Returns the exclude_list element which matched, or NULL for
1194 * undecided.
1195 */
1196 struct exclude *last_exclude_matching(struct dir_struct *dir,
1197 const char *pathname,
1198 int *dtype_p)
1199 {
1200 int pathlen = strlen(pathname);
1201 const char *basename = strrchr(pathname, '/');
1202 basename = (basename) ? basename+1 : pathname;
1203
1204 prep_exclude(dir, pathname, basename-pathname);
1205
1206 if (dir->exclude)
1207 return dir->exclude;
1208
1209 return last_exclude_matching_from_lists(dir, pathname, pathlen,
1210 basename, dtype_p);
1211 }
1212
1213 /*
1214 * Loads the exclude lists for the directory containing pathname, then
1215 * scans all exclude lists to determine whether pathname is excluded.
1216 * Returns 1 if true, otherwise 0.
1217 */
1218 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1219 {
1220 struct exclude *exclude =
1221 last_exclude_matching(dir, pathname, dtype_p);
1222 if (exclude)
1223 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1224 return 0;
1225 }
1226
1227 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1228 {
1229 struct dir_entry *ent;
1230
1231 FLEX_ALLOC_MEM(ent, name, pathname, len);
1232 ent->len = len;
1233 return ent;
1234 }
1235
1236 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1237 {
1238 if (cache_file_exists(pathname, len, ignore_case))
1239 return NULL;
1240
1241 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1242 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1243 }
1244
1245 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1246 {
1247 if (!cache_name_is_other(pathname, len))
1248 return NULL;
1249
1250 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1251 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1252 }
1253
1254 enum exist_status {
1255 index_nonexistent = 0,
1256 index_directory,
1257 index_gitdir
1258 };
1259
1260 /*
1261 * Do not use the alphabetically sorted index to look up
1262 * the directory name; instead, use the case insensitive
1263 * directory hash.
1264 */
1265 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1266 {
1267 struct cache_entry *ce;
1268
1269 if (cache_dir_exists(dirname, len))
1270 return index_directory;
1271
1272 ce = cache_file_exists(dirname, len, ignore_case);
1273 if (ce && S_ISGITLINK(ce->ce_mode))
1274 return index_gitdir;
1275
1276 return index_nonexistent;
1277 }
1278
1279 /*
1280 * The index sorts alphabetically by entry name, which
1281 * means that a gitlink sorts as '\0' at the end, while
1282 * a directory (which is defined not as an entry, but as
1283 * the files it contains) will sort with the '/' at the
1284 * end.
1285 */
1286 static enum exist_status directory_exists_in_index(const char *dirname, int len)
1287 {
1288 int pos;
1289
1290 if (ignore_case)
1291 return directory_exists_in_index_icase(dirname, len);
1292
1293 pos = cache_name_pos(dirname, len);
1294 if (pos < 0)
1295 pos = -pos-1;
1296 while (pos < active_nr) {
1297 const struct cache_entry *ce = active_cache[pos++];
1298 unsigned char endchar;
1299
1300 if (strncmp(ce->name, dirname, len))
1301 break;
1302 endchar = ce->name[len];
1303 if (endchar > '/')
1304 break;
1305 if (endchar == '/')
1306 return index_directory;
1307 if (!endchar && S_ISGITLINK(ce->ce_mode))
1308 return index_gitdir;
1309 }
1310 return index_nonexistent;
1311 }
1312
1313 /*
1314 * When we find a directory when traversing the filesystem, we
1315 * have three distinct cases:
1316 *
1317 * - ignore it
1318 * - see it as a directory
1319 * - recurse into it
1320 *
1321 * and which one we choose depends on a combination of existing
1322 * git index contents and the flags passed into the directory
1323 * traversal routine.
1324 *
1325 * Case 1: If we *already* have entries in the index under that
1326 * directory name, we always recurse into the directory to see
1327 * all the files.
1328 *
1329 * Case 2: If we *already* have that directory name as a gitlink,
1330 * we always continue to see it as a gitlink, regardless of whether
1331 * there is an actual git directory there or not (it might not
1332 * be checked out as a subproject!)
1333 *
1334 * Case 3: if we didn't have it in the index previously, we
1335 * have a few sub-cases:
1336 *
1337 * (a) if "show_other_directories" is true, we show it as
1338 * just a directory, unless "hide_empty_directories" is
1339 * also true, in which case we need to check if it contains any
1340 * untracked and / or ignored files.
1341 * (b) if it looks like a git directory, and we don't have
1342 * 'no_gitlinks' set we treat it as a gitlink, and show it
1343 * as a directory.
1344 * (c) otherwise, we recurse into it.
1345 */
1346 static enum path_treatment treat_directory(struct dir_struct *dir,
1347 struct untracked_cache_dir *untracked,
1348 const char *dirname, int len, int baselen, int exclude,
1349 const struct pathspec *pathspec)
1350 {
1351 /* The "len-1" is to strip the final '/' */
1352 switch (directory_exists_in_index(dirname, len-1)) {
1353 case index_directory:
1354 return path_recurse;
1355
1356 case index_gitdir:
1357 return path_none;
1358
1359 case index_nonexistent:
1360 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1361 break;
1362 if (!(dir->flags & DIR_NO_GITLINKS)) {
1363 unsigned char sha1[20];
1364 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1365 return path_untracked;
1366 }
1367 return path_recurse;
1368 }
1369
1370 /* This is the "show_other_directories" case */
1371
1372 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1373 return exclude ? path_excluded : path_untracked;
1374
1375 untracked = lookup_untracked(dir->untracked, untracked,
1376 dirname + baselen, len - baselen);
1377 return read_directory_recursive(dir, dirname, len,
1378 untracked, 1, pathspec);
1379 }
1380
1381 /*
1382 * This is an inexact early pruning of any recursive directory
1383 * reading - if the path cannot possibly be in the pathspec,
1384 * return true, and we'll skip it early.
1385 */
1386 static int simplify_away(const char *path, int pathlen,
1387 const struct pathspec *pathspec)
1388 {
1389 int i;
1390
1391 if (!pathspec || !pathspec->nr)
1392 return 0;
1393
1394 GUARD_PATHSPEC(pathspec,
1395 PATHSPEC_FROMTOP |
1396 PATHSPEC_MAXDEPTH |
1397 PATHSPEC_LITERAL |
1398 PATHSPEC_GLOB |
1399 PATHSPEC_ICASE |
1400 PATHSPEC_EXCLUDE |
1401 PATHSPEC_ATTR);
1402
1403 for (i = 0; i < pathspec->nr; i++) {
1404 const struct pathspec_item *item = &pathspec->items[i];
1405 int len = item->nowildcard_len;
1406
1407 if (len > pathlen)
1408 len = pathlen;
1409 if (!ps_strncmp(item, item->match, path, len))
1410 return 0;
1411 }
1412
1413 return 1;
1414 }
1415
1416 /*
1417 * This function tells us whether an excluded path matches a
1418 * list of "interesting" pathspecs. That is, whether a path matched
1419 * by any of the pathspecs could possibly be ignored by excluding
1420 * the specified path. This can happen if:
1421 *
1422 * 1. the path is mentioned explicitly in the pathspec
1423 *
1424 * 2. the path is a directory prefix of some element in the
1425 * pathspec
1426 */
1427 static int exclude_matches_pathspec(const char *path, int pathlen,
1428 const struct pathspec *pathspec)
1429 {
1430 int i;
1431
1432 if (!pathspec || !pathspec->nr)
1433 return 0;
1434
1435 GUARD_PATHSPEC(pathspec,
1436 PATHSPEC_FROMTOP |
1437 PATHSPEC_MAXDEPTH |
1438 PATHSPEC_LITERAL |
1439 PATHSPEC_GLOB |
1440 PATHSPEC_ICASE |
1441 PATHSPEC_EXCLUDE);
1442
1443 for (i = 0; i < pathspec->nr; i++) {
1444 const struct pathspec_item *item = &pathspec->items[i];
1445 int len = item->nowildcard_len;
1446
1447 if (len == pathlen &&
1448 !ps_strncmp(item, item->match, path, pathlen))
1449 return 1;
1450 if (len > pathlen &&
1451 item->match[pathlen] == '/' &&
1452 !ps_strncmp(item, item->match, path, pathlen))
1453 return 1;
1454 }
1455 return 0;
1456 }
1457
1458 static int get_index_dtype(const char *path, int len)
1459 {
1460 int pos;
1461 const struct cache_entry *ce;
1462
1463 ce = cache_file_exists(path, len, 0);
1464 if (ce) {
1465 if (!ce_uptodate(ce))
1466 return DT_UNKNOWN;
1467 if (S_ISGITLINK(ce->ce_mode))
1468 return DT_DIR;
1469 /*
1470 * Nobody actually cares about the
1471 * difference between DT_LNK and DT_REG
1472 */
1473 return DT_REG;
1474 }
1475
1476 /* Try to look it up as a directory */
1477 pos = cache_name_pos(path, len);
1478 if (pos >= 0)
1479 return DT_UNKNOWN;
1480 pos = -pos-1;
1481 while (pos < active_nr) {
1482 ce = active_cache[pos++];
1483 if (strncmp(ce->name, path, len))
1484 break;
1485 if (ce->name[len] > '/')
1486 break;
1487 if (ce->name[len] < '/')
1488 continue;
1489 if (!ce_uptodate(ce))
1490 break; /* continue? */
1491 return DT_DIR;
1492 }
1493 return DT_UNKNOWN;
1494 }
1495
1496 static int get_dtype(struct dirent *de, const char *path, int len)
1497 {
1498 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1499 struct stat st;
1500
1501 if (dtype != DT_UNKNOWN)
1502 return dtype;
1503 dtype = get_index_dtype(path, len);
1504 if (dtype != DT_UNKNOWN)
1505 return dtype;
1506 if (lstat(path, &st))
1507 return dtype;
1508 if (S_ISREG(st.st_mode))
1509 return DT_REG;
1510 if (S_ISDIR(st.st_mode))
1511 return DT_DIR;
1512 if (S_ISLNK(st.st_mode))
1513 return DT_LNK;
1514 return dtype;
1515 }
1516
1517 static enum path_treatment treat_one_path(struct dir_struct *dir,
1518 struct untracked_cache_dir *untracked,
1519 struct strbuf *path,
1520 int baselen,
1521 const struct pathspec *pathspec,
1522 int dtype, struct dirent *de)
1523 {
1524 int exclude;
1525 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1526
1527 if (dtype == DT_UNKNOWN)
1528 dtype = get_dtype(de, path->buf, path->len);
1529
1530 /* Always exclude indexed files */
1531 if (dtype != DT_DIR && has_path_in_index)
1532 return path_none;
1533
1534 /*
1535 * When we are looking at a directory P in the working tree,
1536 * there are three cases:
1537 *
1538 * (1) P exists in the index. Everything inside the directory P in
1539 * the working tree needs to go when P is checked out from the
1540 * index.
1541 *
1542 * (2) P does not exist in the index, but there is P/Q in the index.
1543 * We know P will stay a directory when we check out the contents
1544 * of the index, but we do not know yet if there is a directory
1545 * P/Q in the working tree to be killed, so we need to recurse.
1546 *
1547 * (3) P does not exist in the index, and there is no P/Q in the index
1548 * to require P to be a directory, either. Only in this case, we
1549 * know that everything inside P will not be killed without
1550 * recursing.
1551 */
1552 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1553 (dtype == DT_DIR) &&
1554 !has_path_in_index &&
1555 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1556 return path_none;
1557
1558 exclude = is_excluded(dir, path->buf, &dtype);
1559
1560 /*
1561 * Excluded? If we don't explicitly want to show
1562 * ignored files, ignore it
1563 */
1564 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1565 return path_excluded;
1566
1567 switch (dtype) {
1568 default:
1569 return path_none;
1570 case DT_DIR:
1571 strbuf_addch(path, '/');
1572 return treat_directory(dir, untracked, path->buf, path->len,
1573 baselen, exclude, pathspec);
1574 case DT_REG:
1575 case DT_LNK:
1576 return exclude ? path_excluded : path_untracked;
1577 }
1578 }
1579
1580 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1581 struct untracked_cache_dir *untracked,
1582 struct cached_dir *cdir,
1583 struct strbuf *path,
1584 int baselen,
1585 const struct pathspec *pathspec)
1586 {
1587 strbuf_setlen(path, baselen);
1588 if (!cdir->ucd) {
1589 strbuf_addstr(path, cdir->file);
1590 return path_untracked;
1591 }
1592 strbuf_addstr(path, cdir->ucd->name);
1593 /* treat_one_path() does this before it calls treat_directory() */
1594 strbuf_complete(path, '/');
1595 if (cdir->ucd->check_only)
1596 /*
1597 * check_only is set as a result of treat_directory() getting
1598 * to its bottom. Verify again the same set of directories
1599 * with check_only set.
1600 */
1601 return read_directory_recursive(dir, path->buf, path->len,
1602 cdir->ucd, 1, pathspec);
1603 /*
1604 * We get path_recurse in the first run when
1605 * directory_exists_in_index() returns index_nonexistent. We
1606 * are sure that new changes in the index does not impact the
1607 * outcome. Return now.
1608 */
1609 return path_recurse;
1610 }
1611
1612 static enum path_treatment treat_path(struct dir_struct *dir,
1613 struct untracked_cache_dir *untracked,
1614 struct cached_dir *cdir,
1615 struct strbuf *path,
1616 int baselen,
1617 const struct pathspec *pathspec)
1618 {
1619 int dtype;
1620 struct dirent *de = cdir->de;
1621
1622 if (!de)
1623 return treat_path_fast(dir, untracked, cdir, path,
1624 baselen, pathspec);
1625 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1626 return path_none;
1627 strbuf_setlen(path, baselen);
1628 strbuf_addstr(path, de->d_name);
1629 if (simplify_away(path->buf, path->len, pathspec))
1630 return path_none;
1631
1632 dtype = DTYPE(de);
1633 return treat_one_path(dir, untracked, path, baselen, pathspec, dtype, de);
1634 }
1635
1636 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1637 {
1638 if (!dir)
1639 return;
1640 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1641 dir->untracked_alloc);
1642 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1643 }
1644
1645 static int valid_cached_dir(struct dir_struct *dir,
1646 struct untracked_cache_dir *untracked,
1647 struct strbuf *path,
1648 int check_only)
1649 {
1650 struct stat st;
1651
1652 if (!untracked)
1653 return 0;
1654
1655 if (stat(path->len ? path->buf : ".", &st)) {
1656 invalidate_directory(dir->untracked, untracked);
1657 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1658 return 0;
1659 }
1660 if (!untracked->valid ||
1661 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
1662 if (untracked->valid)
1663 invalidate_directory(dir->untracked, untracked);
1664 fill_stat_data(&untracked->stat_data, &st);
1665 return 0;
1666 }
1667
1668 if (untracked->check_only != !!check_only) {
1669 invalidate_directory(dir->untracked, untracked);
1670 return 0;
1671 }
1672
1673 /*
1674 * prep_exclude will be called eventually on this directory,
1675 * but it's called much later in last_exclude_matching(). We
1676 * need it now to determine the validity of the cache for this
1677 * path. The next calls will be nearly no-op, the way
1678 * prep_exclude() is designed.
1679 */
1680 if (path->len && path->buf[path->len - 1] != '/') {
1681 strbuf_addch(path, '/');
1682 prep_exclude(dir, path->buf, path->len);
1683 strbuf_setlen(path, path->len - 1);
1684 } else
1685 prep_exclude(dir, path->buf, path->len);
1686
1687 /* hopefully prep_exclude() haven't invalidated this entry... */
1688 return untracked->valid;
1689 }
1690
1691 static int open_cached_dir(struct cached_dir *cdir,
1692 struct dir_struct *dir,
1693 struct untracked_cache_dir *untracked,
1694 struct strbuf *path,
1695 int check_only)
1696 {
1697 memset(cdir, 0, sizeof(*cdir));
1698 cdir->untracked = untracked;
1699 if (valid_cached_dir(dir, untracked, path, check_only))
1700 return 0;
1701 cdir->fdir = opendir(path->len ? path->buf : ".");
1702 if (dir->untracked)
1703 dir->untracked->dir_opened++;
1704 if (!cdir->fdir)
1705 return -1;
1706 return 0;
1707 }
1708
1709 static int read_cached_dir(struct cached_dir *cdir)
1710 {
1711 if (cdir->fdir) {
1712 cdir->de = readdir(cdir->fdir);
1713 if (!cdir->de)
1714 return -1;
1715 return 0;
1716 }
1717 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1718 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1719 if (!d->recurse) {
1720 cdir->nr_dirs++;
1721 continue;
1722 }
1723 cdir->ucd = d;
1724 cdir->nr_dirs++;
1725 return 0;
1726 }
1727 cdir->ucd = NULL;
1728 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1729 struct untracked_cache_dir *d = cdir->untracked;
1730 cdir->file = d->untracked[cdir->nr_files++];
1731 return 0;
1732 }
1733 return -1;
1734 }
1735
1736 static void close_cached_dir(struct cached_dir *cdir)
1737 {
1738 if (cdir->fdir)
1739 closedir(cdir->fdir);
1740 /*
1741 * We have gone through this directory and found no untracked
1742 * entries. Mark it valid.
1743 */
1744 if (cdir->untracked) {
1745 cdir->untracked->valid = 1;
1746 cdir->untracked->recurse = 1;
1747 }
1748 }
1749
1750 /*
1751 * Read a directory tree. We currently ignore anything but
1752 * directories, regular files and symlinks. That's because git
1753 * doesn't handle them at all yet. Maybe that will change some
1754 * day.
1755 *
1756 * Also, we ignore the name ".git" (even if it is not a directory).
1757 * That likely will not change.
1758 *
1759 * Returns the most significant path_treatment value encountered in the scan.
1760 */
1761 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1762 const char *base, int baselen,
1763 struct untracked_cache_dir *untracked, int check_only,
1764 const struct pathspec *pathspec)
1765 {
1766 struct cached_dir cdir;
1767 enum path_treatment state, subdir_state, dir_state = path_none;
1768 struct strbuf path = STRBUF_INIT;
1769
1770 strbuf_add(&path, base, baselen);
1771
1772 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1773 goto out;
1774
1775 if (untracked)
1776 untracked->check_only = !!check_only;
1777
1778 while (!read_cached_dir(&cdir)) {
1779 /* check how the file or directory should be treated */
1780 state = treat_path(dir, untracked, &cdir, &path,
1781 baselen, pathspec);
1782
1783 if (state > dir_state)
1784 dir_state = state;
1785
1786 /* recurse into subdir if instructed by treat_path */
1787 if (state == path_recurse) {
1788 struct untracked_cache_dir *ud;
1789 ud = lookup_untracked(dir->untracked, untracked,
1790 path.buf + baselen,
1791 path.len - baselen);
1792 subdir_state =
1793 read_directory_recursive(dir, path.buf,
1794 path.len, ud,
1795 check_only, pathspec);
1796 if (subdir_state > dir_state)
1797 dir_state = subdir_state;
1798 }
1799
1800 if (check_only) {
1801 /* abort early if maximum state has been reached */
1802 if (dir_state == path_untracked) {
1803 if (cdir.fdir)
1804 add_untracked(untracked, path.buf + baselen);
1805 break;
1806 }
1807 /* skip the dir_add_* part */
1808 continue;
1809 }
1810
1811 /* add the path to the appropriate result list */
1812 switch (state) {
1813 case path_excluded:
1814 if (dir->flags & DIR_SHOW_IGNORED)
1815 dir_add_name(dir, path.buf, path.len);
1816 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1817 ((dir->flags & DIR_COLLECT_IGNORED) &&
1818 exclude_matches_pathspec(path.buf, path.len,
1819 pathspec)))
1820 dir_add_ignored(dir, path.buf, path.len);
1821 break;
1822
1823 case path_untracked:
1824 if (dir->flags & DIR_SHOW_IGNORED)
1825 break;
1826 dir_add_name(dir, path.buf, path.len);
1827 if (cdir.fdir)
1828 add_untracked(untracked, path.buf + baselen);
1829 break;
1830
1831 default:
1832 break;
1833 }
1834 }
1835 close_cached_dir(&cdir);
1836 out:
1837 strbuf_release(&path);
1838
1839 return dir_state;
1840 }
1841
1842 static int cmp_name(const void *p1, const void *p2)
1843 {
1844 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1845 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1846
1847 return name_compare(e1->name, e1->len, e2->name, e2->len);
1848 }
1849
1850 static int treat_leading_path(struct dir_struct *dir,
1851 const char *path, int len,
1852 const struct pathspec *pathspec)
1853 {
1854 struct strbuf sb = STRBUF_INIT;
1855 int baselen, rc = 0;
1856 const char *cp;
1857 int old_flags = dir->flags;
1858
1859 while (len && path[len - 1] == '/')
1860 len--;
1861 if (!len)
1862 return 1;
1863 baselen = 0;
1864 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1865 while (1) {
1866 cp = path + baselen + !!baselen;
1867 cp = memchr(cp, '/', path + len - cp);
1868 if (!cp)
1869 baselen = len;
1870 else
1871 baselen = cp - path;
1872 strbuf_setlen(&sb, 0);
1873 strbuf_add(&sb, path, baselen);
1874 if (!is_directory(sb.buf))
1875 break;
1876 if (simplify_away(sb.buf, sb.len, pathspec))
1877 break;
1878 if (treat_one_path(dir, NULL, &sb, baselen, pathspec,
1879 DT_DIR, NULL) == path_none)
1880 break; /* do not recurse into it */
1881 if (len <= baselen) {
1882 rc = 1;
1883 break; /* finished checking */
1884 }
1885 }
1886 strbuf_release(&sb);
1887 dir->flags = old_flags;
1888 return rc;
1889 }
1890
1891 static const char *get_ident_string(void)
1892 {
1893 static struct strbuf sb = STRBUF_INIT;
1894 struct utsname uts;
1895
1896 if (sb.len)
1897 return sb.buf;
1898 if (uname(&uts) < 0)
1899 die_errno(_("failed to get kernel name and information"));
1900 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
1901 uts.sysname);
1902 return sb.buf;
1903 }
1904
1905 static int ident_in_untracked(const struct untracked_cache *uc)
1906 {
1907 /*
1908 * Previous git versions may have saved many NUL separated
1909 * strings in the "ident" field, but it is insane to manage
1910 * many locations, so just take care of the first one.
1911 */
1912
1913 return !strcmp(uc->ident.buf, get_ident_string());
1914 }
1915
1916 static void set_untracked_ident(struct untracked_cache *uc)
1917 {
1918 strbuf_reset(&uc->ident);
1919 strbuf_addstr(&uc->ident, get_ident_string());
1920
1921 /*
1922 * This strbuf used to contain a list of NUL separated
1923 * strings, so save NUL too for backward compatibility.
1924 */
1925 strbuf_addch(&uc->ident, 0);
1926 }
1927
1928 static void new_untracked_cache(struct index_state *istate)
1929 {
1930 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
1931 strbuf_init(&uc->ident, 100);
1932 uc->exclude_per_dir = ".gitignore";
1933 /* should be the same flags used by git-status */
1934 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
1935 set_untracked_ident(uc);
1936 istate->untracked = uc;
1937 istate->cache_changed |= UNTRACKED_CHANGED;
1938 }
1939
1940 void add_untracked_cache(struct index_state *istate)
1941 {
1942 if (!istate->untracked) {
1943 new_untracked_cache(istate);
1944 } else {
1945 if (!ident_in_untracked(istate->untracked)) {
1946 free_untracked_cache(istate->untracked);
1947 new_untracked_cache(istate);
1948 }
1949 }
1950 }
1951
1952 void remove_untracked_cache(struct index_state *istate)
1953 {
1954 if (istate->untracked) {
1955 free_untracked_cache(istate->untracked);
1956 istate->untracked = NULL;
1957 istate->cache_changed |= UNTRACKED_CHANGED;
1958 }
1959 }
1960
1961 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1962 int base_len,
1963 const struct pathspec *pathspec)
1964 {
1965 struct untracked_cache_dir *root;
1966
1967 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
1968 return NULL;
1969
1970 /*
1971 * We only support $GIT_DIR/info/exclude and core.excludesfile
1972 * as the global ignore rule files. Any other additions
1973 * (e.g. from command line) invalidate the cache. This
1974 * condition also catches running setup_standard_excludes()
1975 * before setting dir->untracked!
1976 */
1977 if (dir->unmanaged_exclude_files)
1978 return NULL;
1979
1980 /*
1981 * Optimize for the main use case only: whole-tree git
1982 * status. More work involved in treat_leading_path() if we
1983 * use cache on just a subset of the worktree. pathspec
1984 * support could make the matter even worse.
1985 */
1986 if (base_len || (pathspec && pathspec->nr))
1987 return NULL;
1988
1989 /* Different set of flags may produce different results */
1990 if (dir->flags != dir->untracked->dir_flags ||
1991 /*
1992 * See treat_directory(), case index_nonexistent. Without
1993 * this flag, we may need to also cache .git file content
1994 * for the resolve_gitlink_ref() call, which we don't.
1995 */
1996 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1997 /* We don't support collecting ignore files */
1998 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1999 DIR_COLLECT_IGNORED)))
2000 return NULL;
2001
2002 /*
2003 * If we use .gitignore in the cache and now you change it to
2004 * .gitexclude, everything will go wrong.
2005 */
2006 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2007 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2008 return NULL;
2009
2010 /*
2011 * EXC_CMDL is not considered in the cache. If people set it,
2012 * skip the cache.
2013 */
2014 if (dir->exclude_list_group[EXC_CMDL].nr)
2015 return NULL;
2016
2017 if (!ident_in_untracked(dir->untracked)) {
2018 warning(_("Untracked cache is disabled on this system or location."));
2019 return NULL;
2020 }
2021
2022 if (!dir->untracked->root) {
2023 const int len = sizeof(*dir->untracked->root);
2024 dir->untracked->root = xmalloc(len);
2025 memset(dir->untracked->root, 0, len);
2026 }
2027
2028 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2029 root = dir->untracked->root;
2030 if (hashcmp(dir->ss_info_exclude.sha1,
2031 dir->untracked->ss_info_exclude.sha1)) {
2032 invalidate_gitignore(dir->untracked, root);
2033 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2034 }
2035 if (hashcmp(dir->ss_excludes_file.sha1,
2036 dir->untracked->ss_excludes_file.sha1)) {
2037 invalidate_gitignore(dir->untracked, root);
2038 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2039 }
2040
2041 /* Make sure this directory is not dropped out at saving phase */
2042 root->recurse = 1;
2043 return root;
2044 }
2045
2046 int read_directory(struct dir_struct *dir, const char *path,
2047 int len, const struct pathspec *pathspec)
2048 {
2049 struct untracked_cache_dir *untracked;
2050
2051 if (has_symlink_leading_path(path, len))
2052 return dir->nr;
2053
2054 untracked = validate_untracked_cache(dir, len, pathspec);
2055 if (!untracked)
2056 /*
2057 * make sure untracked cache code path is disabled,
2058 * e.g. prep_exclude()
2059 */
2060 dir->untracked = NULL;
2061 if (!len || treat_leading_path(dir, path, len, pathspec))
2062 read_directory_recursive(dir, path, len, untracked, 0, pathspec);
2063 QSORT(dir->entries, dir->nr, cmp_name);
2064 QSORT(dir->ignored, dir->ignored_nr, cmp_name);
2065 if (dir->untracked) {
2066 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2067 trace_printf_key(&trace_untracked_stats,
2068 "node creation: %u\n"
2069 "gitignore invalidation: %u\n"
2070 "directory invalidation: %u\n"
2071 "opendir: %u\n",
2072 dir->untracked->dir_created,
2073 dir->untracked->gitignore_invalidated,
2074 dir->untracked->dir_invalidated,
2075 dir->untracked->dir_opened);
2076 if (dir->untracked == the_index.untracked &&
2077 (dir->untracked->dir_opened ||
2078 dir->untracked->gitignore_invalidated ||
2079 dir->untracked->dir_invalidated))
2080 the_index.cache_changed |= UNTRACKED_CHANGED;
2081 if (dir->untracked != the_index.untracked) {
2082 free(dir->untracked);
2083 dir->untracked = NULL;
2084 }
2085 }
2086 return dir->nr;
2087 }
2088
2089 int file_exists(const char *f)
2090 {
2091 struct stat sb;
2092 return lstat(f, &sb) == 0;
2093 }
2094
2095 static int cmp_icase(char a, char b)
2096 {
2097 if (a == b)
2098 return 0;
2099 if (ignore_case)
2100 return toupper(a) - toupper(b);
2101 return a - b;
2102 }
2103
2104 /*
2105 * Given two normalized paths (a trailing slash is ok), if subdir is
2106 * outside dir, return -1. Otherwise return the offset in subdir that
2107 * can be used as relative path to dir.
2108 */
2109 int dir_inside_of(const char *subdir, const char *dir)
2110 {
2111 int offset = 0;
2112
2113 assert(dir && subdir && *dir && *subdir);
2114
2115 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2116 dir++;
2117 subdir++;
2118 offset++;
2119 }
2120
2121 /* hel[p]/me vs hel[l]/yeah */
2122 if (*dir && *subdir)
2123 return -1;
2124
2125 if (!*subdir)
2126 return !*dir ? offset : -1; /* same dir */
2127
2128 /* foo/[b]ar vs foo/[] */
2129 if (is_dir_sep(dir[-1]))
2130 return is_dir_sep(subdir[-1]) ? offset : -1;
2131
2132 /* foo[/]bar vs foo[] */
2133 return is_dir_sep(*subdir) ? offset + 1 : -1;
2134 }
2135
2136 int is_inside_dir(const char *dir)
2137 {
2138 char *cwd;
2139 int rc;
2140
2141 if (!dir)
2142 return 0;
2143
2144 cwd = xgetcwd();
2145 rc = (dir_inside_of(cwd, dir) >= 0);
2146 free(cwd);
2147 return rc;
2148 }
2149
2150 int is_empty_dir(const char *path)
2151 {
2152 DIR *dir = opendir(path);
2153 struct dirent *e;
2154 int ret = 1;
2155
2156 if (!dir)
2157 return 0;
2158
2159 while ((e = readdir(dir)) != NULL)
2160 if (!is_dot_or_dotdot(e->d_name)) {
2161 ret = 0;
2162 break;
2163 }
2164
2165 closedir(dir);
2166 return ret;
2167 }
2168
2169 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2170 {
2171 DIR *dir;
2172 struct dirent *e;
2173 int ret = 0, original_len = path->len, len, kept_down = 0;
2174 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2175 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2176 unsigned char submodule_head[20];
2177
2178 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2179 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2180 /* Do not descend and nuke a nested git work tree. */
2181 if (kept_up)
2182 *kept_up = 1;
2183 return 0;
2184 }
2185
2186 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2187 dir = opendir(path->buf);
2188 if (!dir) {
2189 if (errno == ENOENT)
2190 return keep_toplevel ? -1 : 0;
2191 else if (errno == EACCES && !keep_toplevel)
2192 /*
2193 * An empty dir could be removable even if it
2194 * is unreadable:
2195 */
2196 return rmdir(path->buf);
2197 else
2198 return -1;
2199 }
2200 strbuf_complete(path, '/');
2201
2202 len = path->len;
2203 while ((e = readdir(dir)) != NULL) {
2204 struct stat st;
2205 if (is_dot_or_dotdot(e->d_name))
2206 continue;
2207
2208 strbuf_setlen(path, len);
2209 strbuf_addstr(path, e->d_name);
2210 if (lstat(path->buf, &st)) {
2211 if (errno == ENOENT)
2212 /*
2213 * file disappeared, which is what we
2214 * wanted anyway
2215 */
2216 continue;
2217 /* fall thru */
2218 } else if (S_ISDIR(st.st_mode)) {
2219 if (!remove_dir_recurse(path, flag, &kept_down))
2220 continue; /* happy */
2221 } else if (!only_empty &&
2222 (!unlink(path->buf) || errno == ENOENT)) {
2223 continue; /* happy, too */
2224 }
2225
2226 /* path too long, stat fails, or non-directory still exists */
2227 ret = -1;
2228 break;
2229 }
2230 closedir(dir);
2231
2232 strbuf_setlen(path, original_len);
2233 if (!ret && !keep_toplevel && !kept_down)
2234 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2235 else if (kept_up)
2236 /*
2237 * report the uplevel that it is not an error that we
2238 * did not rmdir() our directory.
2239 */
2240 *kept_up = !ret;
2241 return ret;
2242 }
2243
2244 int remove_dir_recursively(struct strbuf *path, int flag)
2245 {
2246 return remove_dir_recurse(path, flag, NULL);
2247 }
2248
2249 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2250
2251 void setup_standard_excludes(struct dir_struct *dir)
2252 {
2253 dir->exclude_per_dir = ".gitignore";
2254
2255 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2256 if (!excludes_file)
2257 excludes_file = xdg_config_home("ignore");
2258 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2259 add_excludes_from_file_1(dir, excludes_file,
2260 dir->untracked ? &dir->ss_excludes_file : NULL);
2261
2262 /* per repository user preference */
2263 if (startup_info->have_repository) {
2264 const char *path = git_path_info_exclude();
2265 if (!access_or_warn(path, R_OK, 0))
2266 add_excludes_from_file_1(dir, path,
2267 dir->untracked ? &dir->ss_info_exclude : NULL);
2268 }
2269 }
2270
2271 int remove_path(const char *name)
2272 {
2273 char *slash;
2274
2275 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
2276 return -1;
2277
2278 slash = strrchr(name, '/');
2279 if (slash) {
2280 char *dirs = xstrdup(name);
2281 slash = dirs + (slash - name);
2282 do {
2283 *slash = '\0';
2284 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2285 free(dirs);
2286 }
2287 return 0;
2288 }
2289
2290 /*
2291 * Frees memory within dir which was allocated for exclude lists and
2292 * the exclude_stack. Does not free dir itself.
2293 */
2294 void clear_directory(struct dir_struct *dir)
2295 {
2296 int i, j;
2297 struct exclude_list_group *group;
2298 struct exclude_list *el;
2299 struct exclude_stack *stk;
2300
2301 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2302 group = &dir->exclude_list_group[i];
2303 for (j = 0; j < group->nr; j++) {
2304 el = &group->el[j];
2305 if (i == EXC_DIRS)
2306 free((char *)el->src);
2307 clear_exclude_list(el);
2308 }
2309 free(group->el);
2310 }
2311
2312 stk = dir->exclude_stack;
2313 while (stk) {
2314 struct exclude_stack *prev = stk->prev;
2315 free(stk);
2316 stk = prev;
2317 }
2318 strbuf_release(&dir->basebuf);
2319 }
2320
2321 struct ondisk_untracked_cache {
2322 struct stat_data info_exclude_stat;
2323 struct stat_data excludes_file_stat;
2324 uint32_t dir_flags;
2325 unsigned char info_exclude_sha1[20];
2326 unsigned char excludes_file_sha1[20];
2327 char exclude_per_dir[FLEX_ARRAY];
2328 };
2329
2330 #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2331
2332 struct write_data {
2333 int index; /* number of written untracked_cache_dir */
2334 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2335 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2336 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2337 struct strbuf out;
2338 struct strbuf sb_stat;
2339 struct strbuf sb_sha1;
2340 };
2341
2342 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2343 {
2344 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2345 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2346 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2347 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2348 to->sd_dev = htonl(from->sd_dev);
2349 to->sd_ino = htonl(from->sd_ino);
2350 to->sd_uid = htonl(from->sd_uid);
2351 to->sd_gid = htonl(from->sd_gid);
2352 to->sd_size = htonl(from->sd_size);
2353 }
2354
2355 static void write_one_dir(struct untracked_cache_dir *untracked,
2356 struct write_data *wd)
2357 {
2358 struct stat_data stat_data;
2359 struct strbuf *out = &wd->out;
2360 unsigned char intbuf[16];
2361 unsigned int intlen, value;
2362 int i = wd->index++;
2363
2364 /*
2365 * untracked_nr should be reset whenever valid is clear, but
2366 * for safety..
2367 */
2368 if (!untracked->valid) {
2369 untracked->untracked_nr = 0;
2370 untracked->check_only = 0;
2371 }
2372
2373 if (untracked->check_only)
2374 ewah_set(wd->check_only, i);
2375 if (untracked->valid) {
2376 ewah_set(wd->valid, i);
2377 stat_data_to_disk(&stat_data, &untracked->stat_data);
2378 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2379 }
2380 if (!is_null_sha1(untracked->exclude_sha1)) {
2381 ewah_set(wd->sha1_valid, i);
2382 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2383 }
2384
2385 intlen = encode_varint(untracked->untracked_nr, intbuf);
2386 strbuf_add(out, intbuf, intlen);
2387
2388 /* skip non-recurse directories */
2389 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2390 if (untracked->dirs[i]->recurse)
2391 value++;
2392 intlen = encode_varint(value, intbuf);
2393 strbuf_add(out, intbuf, intlen);
2394
2395 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2396
2397 for (i = 0; i < untracked->untracked_nr; i++)
2398 strbuf_add(out, untracked->untracked[i],
2399 strlen(untracked->untracked[i]) + 1);
2400
2401 for (i = 0; i < untracked->dirs_nr; i++)
2402 if (untracked->dirs[i]->recurse)
2403 write_one_dir(untracked->dirs[i], wd);
2404 }
2405
2406 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2407 {
2408 struct ondisk_untracked_cache *ouc;
2409 struct write_data wd;
2410 unsigned char varbuf[16];
2411 int varint_len;
2412 size_t len = strlen(untracked->exclude_per_dir);
2413
2414 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
2415 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2416 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2417 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2418 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2419 ouc->dir_flags = htonl(untracked->dir_flags);
2420
2421 varint_len = encode_varint(untracked->ident.len, varbuf);
2422 strbuf_add(out, varbuf, varint_len);
2423 strbuf_addbuf(out, &untracked->ident);
2424
2425 strbuf_add(out, ouc, ouc_size(len));
2426 free(ouc);
2427 ouc = NULL;
2428
2429 if (!untracked->root) {
2430 varint_len = encode_varint(0, varbuf);
2431 strbuf_add(out, varbuf, varint_len);
2432 return;
2433 }
2434
2435 wd.index = 0;
2436 wd.check_only = ewah_new();
2437 wd.valid = ewah_new();
2438 wd.sha1_valid = ewah_new();
2439 strbuf_init(&wd.out, 1024);
2440 strbuf_init(&wd.sb_stat, 1024);
2441 strbuf_init(&wd.sb_sha1, 1024);
2442 write_one_dir(untracked->root, &wd);
2443
2444 varint_len = encode_varint(wd.index, varbuf);
2445 strbuf_add(out, varbuf, varint_len);
2446 strbuf_addbuf(out, &wd.out);
2447 ewah_serialize_strbuf(wd.valid, out);
2448 ewah_serialize_strbuf(wd.check_only, out);
2449 ewah_serialize_strbuf(wd.sha1_valid, out);
2450 strbuf_addbuf(out, &wd.sb_stat);
2451 strbuf_addbuf(out, &wd.sb_sha1);
2452 strbuf_addch(out, '\0'); /* safe guard for string lists */
2453
2454 ewah_free(wd.valid);
2455 ewah_free(wd.check_only);
2456 ewah_free(wd.sha1_valid);
2457 strbuf_release(&wd.out);
2458 strbuf_release(&wd.sb_stat);
2459 strbuf_release(&wd.sb_sha1);
2460 }
2461
2462 static void free_untracked(struct untracked_cache_dir *ucd)
2463 {
2464 int i;
2465 if (!ucd)
2466 return;
2467 for (i = 0; i < ucd->dirs_nr; i++)
2468 free_untracked(ucd->dirs[i]);
2469 for (i = 0; i < ucd->untracked_nr; i++)
2470 free(ucd->untracked[i]);
2471 free(ucd->untracked);
2472 free(ucd->dirs);
2473 free(ucd);
2474 }
2475
2476 void free_untracked_cache(struct untracked_cache *uc)
2477 {
2478 if (uc)
2479 free_untracked(uc->root);
2480 free(uc);
2481 }
2482
2483 struct read_data {
2484 int index;
2485 struct untracked_cache_dir **ucd;
2486 struct ewah_bitmap *check_only;
2487 struct ewah_bitmap *valid;
2488 struct ewah_bitmap *sha1_valid;
2489 const unsigned char *data;
2490 const unsigned char *end;
2491 };
2492
2493 static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2494 {
2495 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2496 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2497 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2498 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2499 to->sd_dev = get_be32(&from->sd_dev);
2500 to->sd_ino = get_be32(&from->sd_ino);
2501 to->sd_uid = get_be32(&from->sd_uid);
2502 to->sd_gid = get_be32(&from->sd_gid);
2503 to->sd_size = get_be32(&from->sd_size);
2504 }
2505
2506 static int read_one_dir(struct untracked_cache_dir **untracked_,
2507 struct read_data *rd)
2508 {
2509 struct untracked_cache_dir ud, *untracked;
2510 const unsigned char *next, *data = rd->data, *end = rd->end;
2511 unsigned int value;
2512 int i, len;
2513
2514 memset(&ud, 0, sizeof(ud));
2515
2516 next = data;
2517 value = decode_varint(&next);
2518 if (next > end)
2519 return -1;
2520 ud.recurse = 1;
2521 ud.untracked_alloc = value;
2522 ud.untracked_nr = value;
2523 if (ud.untracked_nr)
2524 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2525 data = next;
2526
2527 next = data;
2528 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2529 if (next > end)
2530 return -1;
2531 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2532 data = next;
2533
2534 len = strlen((const char *)data);
2535 next = data + len + 1;
2536 if (next > rd->end)
2537 return -1;
2538 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
2539 memcpy(untracked, &ud, sizeof(ud));
2540 memcpy(untracked->name, data, len + 1);
2541 data = next;
2542
2543 for (i = 0; i < untracked->untracked_nr; i++) {
2544 len = strlen((const char *)data);
2545 next = data + len + 1;
2546 if (next > rd->end)
2547 return -1;
2548 untracked->untracked[i] = xstrdup((const char*)data);
2549 data = next;
2550 }
2551
2552 rd->ucd[rd->index++] = untracked;
2553 rd->data = data;
2554
2555 for (i = 0; i < untracked->dirs_nr; i++) {
2556 len = read_one_dir(untracked->dirs + i, rd);
2557 if (len < 0)
2558 return -1;
2559 }
2560 return 0;
2561 }
2562
2563 static void set_check_only(size_t pos, void *cb)
2564 {
2565 struct read_data *rd = cb;
2566 struct untracked_cache_dir *ud = rd->ucd[pos];
2567 ud->check_only = 1;
2568 }
2569
2570 static void read_stat(size_t pos, void *cb)
2571 {
2572 struct read_data *rd = cb;
2573 struct untracked_cache_dir *ud = rd->ucd[pos];
2574 if (rd->data + sizeof(struct stat_data) > rd->end) {
2575 rd->data = rd->end + 1;
2576 return;
2577 }
2578 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2579 rd->data += sizeof(struct stat_data);
2580 ud->valid = 1;
2581 }
2582
2583 static void read_sha1(size_t pos, void *cb)
2584 {
2585 struct read_data *rd = cb;
2586 struct untracked_cache_dir *ud = rd->ucd[pos];
2587 if (rd->data + 20 > rd->end) {
2588 rd->data = rd->end + 1;
2589 return;
2590 }
2591 hashcpy(ud->exclude_sha1, rd->data);
2592 rd->data += 20;
2593 }
2594
2595 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2596 const struct stat_data *stat,
2597 const unsigned char *sha1)
2598 {
2599 stat_data_from_disk(&sha1_stat->stat, stat);
2600 hashcpy(sha1_stat->sha1, sha1);
2601 sha1_stat->valid = 1;
2602 }
2603
2604 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2605 {
2606 const struct ondisk_untracked_cache *ouc;
2607 struct untracked_cache *uc;
2608 struct read_data rd;
2609 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2610 const char *ident;
2611 int ident_len, len;
2612
2613 if (sz <= 1 || end[-1] != '\0')
2614 return NULL;
2615 end--;
2616
2617 ident_len = decode_varint(&next);
2618 if (next + ident_len > end)
2619 return NULL;
2620 ident = (const char *)next;
2621 next += ident_len;
2622
2623 ouc = (const struct ondisk_untracked_cache *)next;
2624 if (next + ouc_size(0) > end)
2625 return NULL;
2626
2627 uc = xcalloc(1, sizeof(*uc));
2628 strbuf_init(&uc->ident, ident_len);
2629 strbuf_add(&uc->ident, ident, ident_len);
2630 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2631 ouc->info_exclude_sha1);
2632 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2633 ouc->excludes_file_sha1);
2634 uc->dir_flags = get_be32(&ouc->dir_flags);
2635 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2636 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2637 next += ouc_size(strlen(ouc->exclude_per_dir));
2638 if (next >= end)
2639 goto done2;
2640
2641 len = decode_varint(&next);
2642 if (next > end || len == 0)
2643 goto done2;
2644
2645 rd.valid = ewah_new();
2646 rd.check_only = ewah_new();
2647 rd.sha1_valid = ewah_new();
2648 rd.data = next;
2649 rd.end = end;
2650 rd.index = 0;
2651 ALLOC_ARRAY(rd.ucd, len);
2652
2653 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2654 goto done;
2655
2656 next = rd.data;
2657 len = ewah_read_mmap(rd.valid, next, end - next);
2658 if (len < 0)
2659 goto done;
2660
2661 next += len;
2662 len = ewah_read_mmap(rd.check_only, next, end - next);
2663 if (len < 0)
2664 goto done;
2665
2666 next += len;
2667 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2668 if (len < 0)
2669 goto done;
2670
2671 ewah_each_bit(rd.check_only, set_check_only, &rd);
2672 rd.data = next + len;
2673 ewah_each_bit(rd.valid, read_stat, &rd);
2674 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2675 next = rd.data;
2676
2677 done:
2678 free(rd.ucd);
2679 ewah_free(rd.valid);
2680 ewah_free(rd.check_only);
2681 ewah_free(rd.sha1_valid);
2682 done2:
2683 if (next != end) {
2684 free_untracked_cache(uc);
2685 uc = NULL;
2686 }
2687 return uc;
2688 }
2689
2690 static void invalidate_one_directory(struct untracked_cache *uc,
2691 struct untracked_cache_dir *ucd)
2692 {
2693 uc->dir_invalidated++;
2694 ucd->valid = 0;
2695 ucd->untracked_nr = 0;
2696 }
2697
2698 /*
2699 * Normally when an entry is added or removed from a directory,
2700 * invalidating that directory is enough. No need to touch its
2701 * ancestors. When a directory is shown as "foo/bar/" in git-status
2702 * however, deleting or adding an entry may have cascading effect.
2703 *
2704 * Say the "foo/bar/file" has become untracked, we need to tell the
2705 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2706 * directory any more (because "bar" is managed by foo as an untracked
2707 * "file").
2708 *
2709 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2710 * was the last untracked entry in the entire "foo", we should show
2711 * "foo/" instead. Which means we have to invalidate past "bar" up to
2712 * "foo".
2713 *
2714 * This function traverses all directories from root to leaf. If there
2715 * is a chance of one of the above cases happening, we invalidate back
2716 * to root. Otherwise we just invalidate the leaf. There may be a more
2717 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2718 * detect these cases and avoid unnecessary invalidation, for example,
2719 * checking for the untracked entry named "bar/" in "foo", but for now
2720 * stick to something safe and simple.
2721 */
2722 static int invalidate_one_component(struct untracked_cache *uc,
2723 struct untracked_cache_dir *dir,
2724 const char *path, int len)
2725 {
2726 const char *rest = strchr(path, '/');
2727
2728 if (rest) {
2729 int component_len = rest - path;
2730 struct untracked_cache_dir *d =
2731 lookup_untracked(uc, dir, path, component_len);
2732 int ret =
2733 invalidate_one_component(uc, d, rest + 1,
2734 len - (component_len + 1));
2735 if (ret)
2736 invalidate_one_directory(uc, dir);
2737 return ret;
2738 }
2739
2740 invalidate_one_directory(uc, dir);
2741 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2742 }
2743
2744 void untracked_cache_invalidate_path(struct index_state *istate,
2745 const char *path)
2746 {
2747 if (!istate->untracked || !istate->untracked->root)
2748 return;
2749 invalidate_one_component(istate->untracked, istate->untracked->root,
2750 path, strlen(path));
2751 }
2752
2753 void untracked_cache_remove_from_index(struct index_state *istate,
2754 const char *path)
2755 {
2756 untracked_cache_invalidate_path(istate, path);
2757 }
2758
2759 void untracked_cache_add_to_index(struct index_state *istate,
2760 const char *path)
2761 {
2762 untracked_cache_invalidate_path(istate, path);
2763 }
2764
2765 /* Update gitfile and core.worktree setting to connect work tree and git dir */
2766 void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)
2767 {
2768 struct strbuf file_name = STRBUF_INIT;
2769 struct strbuf rel_path = STRBUF_INIT;
2770 char *git_dir = real_pathdup(git_dir_);
2771 char *work_tree = real_pathdup(work_tree_);
2772
2773 /* Update gitfile */
2774 strbuf_addf(&file_name, "%s/.git", work_tree);
2775 write_file(file_name.buf, "gitdir: %s",
2776 relative_path(git_dir, work_tree, &rel_path));
2777
2778 /* Update core.worktree setting */
2779 strbuf_reset(&file_name);
2780 strbuf_addf(&file_name, "%s/config", git_dir);
2781 git_config_set_in_file(file_name.buf, "core.worktree",
2782 relative_path(work_tree, git_dir, &rel_path));
2783
2784 strbuf_release(&file_name);
2785 strbuf_release(&rel_path);
2786 free(work_tree);
2787 free(git_dir);
2788 }
2789
2790 /*
2791 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
2792 */
2793 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
2794 {
2795 if (rename(old_git_dir, new_git_dir) < 0)
2796 die_errno(_("could not migrate git directory from '%s' to '%s'"),
2797 old_git_dir, new_git_dir);
2798
2799 connect_work_tree_and_git_dir(path, new_git_dir);
2800 }