Merge branch 'dl/cocci-everywhere'
[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 "config.h"
12 #include "dir.h"
13 #include "object-store.h"
14 #include "attr.h"
15 #include "refs.h"
16 #include "wildmatch.h"
17 #include "pathspec.h"
18 #include "utf8.h"
19 #include "varint.h"
20 #include "ewah/ewok.h"
21 #include "fsmonitor.h"
22 #include "submodule-config.h"
23
24 /*
25 * Tells read_directory_recursive how a file or directory should be treated.
26 * Values are ordered by significance, e.g. if a directory contains both
27 * excluded and untracked files, it is listed as untracked because
28 * path_untracked > path_excluded.
29 */
30 enum path_treatment {
31 path_none = 0,
32 path_recurse,
33 path_excluded,
34 path_untracked
35 };
36
37 /*
38 * Support data structure for our opendir/readdir/closedir wrappers
39 */
40 struct cached_dir {
41 DIR *fdir;
42 struct untracked_cache_dir *untracked;
43 int nr_files;
44 int nr_dirs;
45
46 struct dirent *de;
47 const char *file;
48 struct untracked_cache_dir *ucd;
49 };
50
51 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
52 struct index_state *istate, const char *path, int len,
53 struct untracked_cache_dir *untracked,
54 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
55 static int get_dtype(struct dirent *de, struct index_state *istate,
56 const char *path, int len);
57
58 int count_slashes(const char *s)
59 {
60 int cnt = 0;
61 while (*s)
62 if (*s++ == '/')
63 cnt++;
64 return cnt;
65 }
66
67 int fspathcmp(const char *a, const char *b)
68 {
69 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
70 }
71
72 int fspathncmp(const char *a, const char *b, size_t count)
73 {
74 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
75 }
76
77 int git_fnmatch(const struct pathspec_item *item,
78 const char *pattern, const char *string,
79 int prefix)
80 {
81 if (prefix > 0) {
82 if (ps_strncmp(item, pattern, string, prefix))
83 return WM_NOMATCH;
84 pattern += prefix;
85 string += prefix;
86 }
87 if (item->flags & PATHSPEC_ONESTAR) {
88 int pattern_len = strlen(++pattern);
89 int string_len = strlen(string);
90 return string_len < pattern_len ||
91 ps_strcmp(item, pattern,
92 string + string_len - pattern_len);
93 }
94 if (item->magic & PATHSPEC_GLOB)
95 return wildmatch(pattern, string,
96 WM_PATHNAME |
97 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
98 else
99 /* wildmatch has not learned no FNM_PATHNAME mode yet */
100 return wildmatch(pattern, string,
101 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
102 }
103
104 static int fnmatch_icase_mem(const char *pattern, int patternlen,
105 const char *string, int stringlen,
106 int flags)
107 {
108 int match_status;
109 struct strbuf pat_buf = STRBUF_INIT;
110 struct strbuf str_buf = STRBUF_INIT;
111 const char *use_pat = pattern;
112 const char *use_str = string;
113
114 if (pattern[patternlen]) {
115 strbuf_add(&pat_buf, pattern, patternlen);
116 use_pat = pat_buf.buf;
117 }
118 if (string[stringlen]) {
119 strbuf_add(&str_buf, string, stringlen);
120 use_str = str_buf.buf;
121 }
122
123 if (ignore_case)
124 flags |= WM_CASEFOLD;
125 match_status = wildmatch(use_pat, use_str, flags);
126
127 strbuf_release(&pat_buf);
128 strbuf_release(&str_buf);
129
130 return match_status;
131 }
132
133 static size_t common_prefix_len(const struct pathspec *pathspec)
134 {
135 int n;
136 size_t max = 0;
137
138 /*
139 * ":(icase)path" is treated as a pathspec full of
140 * wildcard. In other words, only prefix is considered common
141 * prefix. If the pathspec is abc/foo abc/bar, running in
142 * subdir xyz, the common prefix is still xyz, not xuz/abc as
143 * in non-:(icase).
144 */
145 GUARD_PATHSPEC(pathspec,
146 PATHSPEC_FROMTOP |
147 PATHSPEC_MAXDEPTH |
148 PATHSPEC_LITERAL |
149 PATHSPEC_GLOB |
150 PATHSPEC_ICASE |
151 PATHSPEC_EXCLUDE |
152 PATHSPEC_ATTR);
153
154 for (n = 0; n < pathspec->nr; n++) {
155 size_t i = 0, len = 0, item_len;
156 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
157 continue;
158 if (pathspec->items[n].magic & PATHSPEC_ICASE)
159 item_len = pathspec->items[n].prefix;
160 else
161 item_len = pathspec->items[n].nowildcard_len;
162 while (i < item_len && (n == 0 || i < max)) {
163 char c = pathspec->items[n].match[i];
164 if (c != pathspec->items[0].match[i])
165 break;
166 if (c == '/')
167 len = i + 1;
168 i++;
169 }
170 if (n == 0 || len < max) {
171 max = len;
172 if (!max)
173 break;
174 }
175 }
176 return max;
177 }
178
179 /*
180 * Returns a copy of the longest leading path common among all
181 * pathspecs.
182 */
183 char *common_prefix(const struct pathspec *pathspec)
184 {
185 unsigned long len = common_prefix_len(pathspec);
186
187 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
188 }
189
190 int fill_directory(struct dir_struct *dir,
191 struct index_state *istate,
192 const struct pathspec *pathspec)
193 {
194 const char *prefix;
195 size_t prefix_len;
196
197 /*
198 * Calculate common prefix for the pathspec, and
199 * use that to optimize the directory walk
200 */
201 prefix_len = common_prefix_len(pathspec);
202 prefix = prefix_len ? pathspec->items[0].match : "";
203
204 /* Read the directory and prune it */
205 read_directory(dir, istate, prefix, prefix_len, pathspec);
206
207 return prefix_len;
208 }
209
210 int within_depth(const char *name, int namelen,
211 int depth, int max_depth)
212 {
213 const char *cp = name, *cpe = name + namelen;
214
215 while (cp < cpe) {
216 if (*cp++ != '/')
217 continue;
218 depth++;
219 if (depth > max_depth)
220 return 0;
221 }
222 return 1;
223 }
224
225 /*
226 * Read the contents of the blob with the given OID into a buffer.
227 * Append a trailing LF to the end if the last line doesn't have one.
228 *
229 * Returns:
230 * -1 when the OID is invalid or unknown or does not refer to a blob.
231 * 0 when the blob is empty.
232 * 1 along with { data, size } of the (possibly augmented) buffer
233 * when successful.
234 *
235 * Optionally updates the given oid_stat with the given OID (when valid).
236 */
237 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
238 size_t *size_out, char **data_out)
239 {
240 enum object_type type;
241 unsigned long sz;
242 char *data;
243
244 *size_out = 0;
245 *data_out = NULL;
246
247 data = read_object_file(oid, &type, &sz);
248 if (!data || type != OBJ_BLOB) {
249 free(data);
250 return -1;
251 }
252
253 if (oid_stat) {
254 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
255 oidcpy(&oid_stat->oid, oid);
256 }
257
258 if (sz == 0) {
259 free(data);
260 return 0;
261 }
262
263 if (data[sz - 1] != '\n') {
264 data = xrealloc(data, st_add(sz, 1));
265 data[sz++] = '\n';
266 }
267
268 *size_out = xsize_t(sz);
269 *data_out = data;
270
271 return 1;
272 }
273
274 #define DO_MATCH_EXCLUDE (1<<0)
275 #define DO_MATCH_DIRECTORY (1<<1)
276 #define DO_MATCH_SUBMODULE (1<<2)
277
278 /*
279 * Does 'match' match the given name?
280 * A match is found if
281 *
282 * (1) the 'match' string is leading directory of 'name', or
283 * (2) the 'match' string is a wildcard and matches 'name', or
284 * (3) the 'match' string is exactly the same as 'name'.
285 *
286 * and the return value tells which case it was.
287 *
288 * It returns 0 when there is no match.
289 */
290 static int match_pathspec_item(const struct index_state *istate,
291 const struct pathspec_item *item, int prefix,
292 const char *name, int namelen, unsigned flags)
293 {
294 /* name/namelen has prefix cut off by caller */
295 const char *match = item->match + prefix;
296 int matchlen = item->len - prefix;
297
298 /*
299 * The normal call pattern is:
300 * 1. prefix = common_prefix_len(ps);
301 * 2. prune something, or fill_directory
302 * 3. match_pathspec()
303 *
304 * 'prefix' at #1 may be shorter than the command's prefix and
305 * it's ok for #2 to match extra files. Those extras will be
306 * trimmed at #3.
307 *
308 * Suppose the pathspec is 'foo' and '../bar' running from
309 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
310 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
311 * user does not want XYZ/foo, only the "foo" part should be
312 * case-insensitive. We need to filter out XYZ/foo here. In
313 * other words, we do not trust the caller on comparing the
314 * prefix part when :(icase) is involved. We do exact
315 * comparison ourselves.
316 *
317 * Normally the caller (common_prefix_len() in fact) does
318 * _exact_ matching on name[-prefix+1..-1] and we do not need
319 * to check that part. Be defensive and check it anyway, in
320 * case common_prefix_len is changed, or a new caller is
321 * introduced that does not use common_prefix_len.
322 *
323 * If the penalty turns out too high when prefix is really
324 * long, maybe change it to
325 * strncmp(match, name, item->prefix - prefix)
326 */
327 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
328 strncmp(item->match, name - prefix, item->prefix))
329 return 0;
330
331 if (item->attr_match_nr &&
332 !match_pathspec_attrs(istate, name, namelen, item))
333 return 0;
334
335 /* If the match was just the prefix, we matched */
336 if (!*match)
337 return MATCHED_RECURSIVELY;
338
339 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
340 if (matchlen == namelen)
341 return MATCHED_EXACTLY;
342
343 if (match[matchlen-1] == '/' || name[matchlen] == '/')
344 return MATCHED_RECURSIVELY;
345 } else if ((flags & DO_MATCH_DIRECTORY) &&
346 match[matchlen - 1] == '/' &&
347 namelen == matchlen - 1 &&
348 !ps_strncmp(item, match, name, namelen))
349 return MATCHED_EXACTLY;
350
351 if (item->nowildcard_len < item->len &&
352 !git_fnmatch(item, match, name,
353 item->nowildcard_len - prefix))
354 return MATCHED_FNMATCH;
355
356 /* Perform checks to see if "name" is a super set of the pathspec */
357 if (flags & DO_MATCH_SUBMODULE) {
358 /* name is a literal prefix of the pathspec */
359 if ((namelen < matchlen) &&
360 (match[namelen] == '/') &&
361 !ps_strncmp(item, match, name, namelen))
362 return MATCHED_RECURSIVELY;
363
364 /* name" doesn't match up to the first wild character */
365 if (item->nowildcard_len < item->len &&
366 ps_strncmp(item, match, name,
367 item->nowildcard_len - prefix))
368 return 0;
369
370 /*
371 * Here is where we would perform a wildmatch to check if
372 * "name" can be matched as a directory (or a prefix) against
373 * the pathspec. Since wildmatch doesn't have this capability
374 * at the present we have to punt and say that it is a match,
375 * potentially returning a false positive
376 * The submodules themselves will be able to perform more
377 * accurate matching to determine if the pathspec matches.
378 */
379 return MATCHED_RECURSIVELY;
380 }
381
382 return 0;
383 }
384
385 /*
386 * Given a name and a list of pathspecs, returns the nature of the
387 * closest (i.e. most specific) match of the name to any of the
388 * pathspecs.
389 *
390 * The caller typically calls this multiple times with the same
391 * pathspec and seen[] array but with different name/namelen
392 * (e.g. entries from the index) and is interested in seeing if and
393 * how each pathspec matches all the names it calls this function
394 * with. A mark is left in the seen[] array for each pathspec element
395 * indicating the closest type of match that element achieved, so if
396 * seen[n] remains zero after multiple invocations, that means the nth
397 * pathspec did not match any names, which could indicate that the
398 * user mistyped the nth pathspec.
399 */
400 static int do_match_pathspec(const struct index_state *istate,
401 const struct pathspec *ps,
402 const char *name, int namelen,
403 int prefix, char *seen,
404 unsigned flags)
405 {
406 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
407
408 GUARD_PATHSPEC(ps,
409 PATHSPEC_FROMTOP |
410 PATHSPEC_MAXDEPTH |
411 PATHSPEC_LITERAL |
412 PATHSPEC_GLOB |
413 PATHSPEC_ICASE |
414 PATHSPEC_EXCLUDE |
415 PATHSPEC_ATTR);
416
417 if (!ps->nr) {
418 if (!ps->recursive ||
419 !(ps->magic & PATHSPEC_MAXDEPTH) ||
420 ps->max_depth == -1)
421 return MATCHED_RECURSIVELY;
422
423 if (within_depth(name, namelen, 0, ps->max_depth))
424 return MATCHED_EXACTLY;
425 else
426 return 0;
427 }
428
429 name += prefix;
430 namelen -= prefix;
431
432 for (i = ps->nr - 1; i >= 0; i--) {
433 int how;
434
435 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
436 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
437 continue;
438
439 if (seen && seen[i] == MATCHED_EXACTLY)
440 continue;
441 /*
442 * Make exclude patterns optional and never report
443 * "pathspec ':(exclude)foo' matches no files"
444 */
445 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
446 seen[i] = MATCHED_FNMATCH;
447 how = match_pathspec_item(istate, ps->items+i, prefix, name,
448 namelen, flags);
449 if (ps->recursive &&
450 (ps->magic & PATHSPEC_MAXDEPTH) &&
451 ps->max_depth != -1 &&
452 how && how != MATCHED_FNMATCH) {
453 int len = ps->items[i].len;
454 if (name[len] == '/')
455 len++;
456 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
457 how = MATCHED_EXACTLY;
458 else
459 how = 0;
460 }
461 if (how) {
462 if (retval < how)
463 retval = how;
464 if (seen && seen[i] < how)
465 seen[i] = how;
466 }
467 }
468 return retval;
469 }
470
471 int match_pathspec(const struct index_state *istate,
472 const struct pathspec *ps,
473 const char *name, int namelen,
474 int prefix, char *seen, int is_dir)
475 {
476 int positive, negative;
477 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
478 positive = do_match_pathspec(istate, ps, name, namelen,
479 prefix, seen, flags);
480 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
481 return positive;
482 negative = do_match_pathspec(istate, ps, name, namelen,
483 prefix, seen,
484 flags | DO_MATCH_EXCLUDE);
485 return negative ? 0 : positive;
486 }
487
488 /**
489 * Check if a submodule is a superset of the pathspec
490 */
491 int submodule_path_match(const struct index_state *istate,
492 const struct pathspec *ps,
493 const char *submodule_name,
494 char *seen)
495 {
496 int matched = do_match_pathspec(istate, ps, submodule_name,
497 strlen(submodule_name),
498 0, seen,
499 DO_MATCH_DIRECTORY |
500 DO_MATCH_SUBMODULE);
501 return matched;
502 }
503
504 int report_path_error(const char *ps_matched,
505 const struct pathspec *pathspec)
506 {
507 /*
508 * Make sure all pathspec matched; otherwise it is an error.
509 */
510 int num, errors = 0;
511 for (num = 0; num < pathspec->nr; num++) {
512 int other, found_dup;
513
514 if (ps_matched[num])
515 continue;
516 /*
517 * The caller might have fed identical pathspec
518 * twice. Do not barf on such a mistake.
519 * FIXME: parse_pathspec should have eliminated
520 * duplicate pathspec.
521 */
522 for (found_dup = other = 0;
523 !found_dup && other < pathspec->nr;
524 other++) {
525 if (other == num || !ps_matched[other])
526 continue;
527 if (!strcmp(pathspec->items[other].original,
528 pathspec->items[num].original))
529 /*
530 * Ok, we have a match already.
531 */
532 found_dup = 1;
533 }
534 if (found_dup)
535 continue;
536
537 error(_("pathspec '%s' did not match any file(s) known to git"),
538 pathspec->items[num].original);
539 errors++;
540 }
541 return errors;
542 }
543
544 /*
545 * Return the length of the "simple" part of a path match limiter.
546 */
547 int simple_length(const char *match)
548 {
549 int len = -1;
550
551 for (;;) {
552 unsigned char c = *match++;
553 len++;
554 if (c == '\0' || is_glob_special(c))
555 return len;
556 }
557 }
558
559 int no_wildcard(const char *string)
560 {
561 return string[simple_length(string)] == '\0';
562 }
563
564 void parse_path_pattern(const char **pattern,
565 int *patternlen,
566 unsigned *flags,
567 int *nowildcardlen)
568 {
569 const char *p = *pattern;
570 size_t i, len;
571
572 *flags = 0;
573 if (*p == '!') {
574 *flags |= PATTERN_FLAG_NEGATIVE;
575 p++;
576 }
577 len = strlen(p);
578 if (len && p[len - 1] == '/') {
579 len--;
580 *flags |= PATTERN_FLAG_MUSTBEDIR;
581 }
582 for (i = 0; i < len; i++) {
583 if (p[i] == '/')
584 break;
585 }
586 if (i == len)
587 *flags |= PATTERN_FLAG_NODIR;
588 *nowildcardlen = simple_length(p);
589 /*
590 * we should have excluded the trailing slash from 'p' too,
591 * but that's one more allocation. Instead just make sure
592 * nowildcardlen does not exceed real patternlen
593 */
594 if (*nowildcardlen > len)
595 *nowildcardlen = len;
596 if (*p == '*' && no_wildcard(p + 1))
597 *flags |= PATTERN_FLAG_ENDSWITH;
598 *pattern = p;
599 *patternlen = len;
600 }
601
602 void add_pattern(const char *string, const char *base,
603 int baselen, struct pattern_list *pl, int srcpos)
604 {
605 struct path_pattern *pattern;
606 int patternlen;
607 unsigned flags;
608 int nowildcardlen;
609
610 parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen);
611 if (flags & PATTERN_FLAG_MUSTBEDIR) {
612 FLEXPTR_ALLOC_MEM(pattern, pattern, string, patternlen);
613 } else {
614 pattern = xmalloc(sizeof(*pattern));
615 pattern->pattern = string;
616 }
617 pattern->patternlen = patternlen;
618 pattern->nowildcardlen = nowildcardlen;
619 pattern->base = base;
620 pattern->baselen = baselen;
621 pattern->flags = flags;
622 pattern->srcpos = srcpos;
623 ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc);
624 pl->patterns[pl->nr++] = pattern;
625 pattern->pl = pl;
626 }
627
628 static int read_skip_worktree_file_from_index(const struct index_state *istate,
629 const char *path,
630 size_t *size_out, char **data_out,
631 struct oid_stat *oid_stat)
632 {
633 int pos, len;
634
635 len = strlen(path);
636 pos = index_name_pos(istate, path, len);
637 if (pos < 0)
638 return -1;
639 if (!ce_skip_worktree(istate->cache[pos]))
640 return -1;
641
642 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
643 }
644
645 /*
646 * Frees memory within pl which was allocated for exclude patterns and
647 * the file buffer. Does not free pl itself.
648 */
649 void clear_pattern_list(struct pattern_list *pl)
650 {
651 int i;
652
653 for (i = 0; i < pl->nr; i++)
654 free(pl->patterns[i]);
655 free(pl->patterns);
656 free(pl->filebuf);
657
658 memset(pl, 0, sizeof(*pl));
659 }
660
661 static void trim_trailing_spaces(char *buf)
662 {
663 char *p, *last_space = NULL;
664
665 for (p = buf; *p; p++)
666 switch (*p) {
667 case ' ':
668 if (!last_space)
669 last_space = p;
670 break;
671 case '\\':
672 p++;
673 if (!*p)
674 return;
675 /* fallthrough */
676 default:
677 last_space = NULL;
678 }
679
680 if (last_space)
681 *last_space = '\0';
682 }
683
684 /*
685 * Given a subdirectory name and "dir" of the current directory,
686 * search the subdir in "dir" and return it, or create a new one if it
687 * does not exist in "dir".
688 *
689 * If "name" has the trailing slash, it'll be excluded in the search.
690 */
691 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
692 struct untracked_cache_dir *dir,
693 const char *name, int len)
694 {
695 int first, last;
696 struct untracked_cache_dir *d;
697 if (!dir)
698 return NULL;
699 if (len && name[len - 1] == '/')
700 len--;
701 first = 0;
702 last = dir->dirs_nr;
703 while (last > first) {
704 int cmp, next = first + ((last - first) >> 1);
705 d = dir->dirs[next];
706 cmp = strncmp(name, d->name, len);
707 if (!cmp && strlen(d->name) > len)
708 cmp = -1;
709 if (!cmp)
710 return d;
711 if (cmp < 0) {
712 last = next;
713 continue;
714 }
715 first = next+1;
716 }
717
718 uc->dir_created++;
719 FLEX_ALLOC_MEM(d, name, name, len);
720
721 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
722 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
723 dir->dirs_nr - first);
724 dir->dirs_nr++;
725 dir->dirs[first] = d;
726 return d;
727 }
728
729 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
730 {
731 int i;
732 dir->valid = 0;
733 dir->untracked_nr = 0;
734 for (i = 0; i < dir->dirs_nr; i++)
735 do_invalidate_gitignore(dir->dirs[i]);
736 }
737
738 static void invalidate_gitignore(struct untracked_cache *uc,
739 struct untracked_cache_dir *dir)
740 {
741 uc->gitignore_invalidated++;
742 do_invalidate_gitignore(dir);
743 }
744
745 static void invalidate_directory(struct untracked_cache *uc,
746 struct untracked_cache_dir *dir)
747 {
748 int i;
749
750 /*
751 * Invalidation increment here is just roughly correct. If
752 * untracked_nr or any of dirs[].recurse is non-zero, we
753 * should increment dir_invalidated too. But that's more
754 * expensive to do.
755 */
756 if (dir->valid)
757 uc->dir_invalidated++;
758
759 dir->valid = 0;
760 dir->untracked_nr = 0;
761 for (i = 0; i < dir->dirs_nr; i++)
762 dir->dirs[i]->recurse = 0;
763 }
764
765 static int add_patterns_from_buffer(char *buf, size_t size,
766 const char *base, int baselen,
767 struct pattern_list *pl);
768
769 /*
770 * Given a file with name "fname", read it (either from disk, or from
771 * an index if 'istate' is non-null), parse it and store the
772 * exclude rules in "pl".
773 *
774 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
775 * stat data from disk (only valid if add_patterns returns zero). If
776 * ss_valid is non-zero, "ss" must contain good value as input.
777 */
778 static int add_patterns(const char *fname, const char *base, int baselen,
779 struct pattern_list *pl, struct index_state *istate,
780 struct oid_stat *oid_stat)
781 {
782 struct stat st;
783 int r;
784 int fd;
785 size_t size = 0;
786 char *buf;
787
788 fd = open(fname, O_RDONLY);
789 if (fd < 0 || fstat(fd, &st) < 0) {
790 if (fd < 0)
791 warn_on_fopen_errors(fname);
792 else
793 close(fd);
794 if (!istate)
795 return -1;
796 r = read_skip_worktree_file_from_index(istate, fname,
797 &size, &buf,
798 oid_stat);
799 if (r != 1)
800 return r;
801 } else {
802 size = xsize_t(st.st_size);
803 if (size == 0) {
804 if (oid_stat) {
805 fill_stat_data(&oid_stat->stat, &st);
806 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
807 oid_stat->valid = 1;
808 }
809 close(fd);
810 return 0;
811 }
812 buf = xmallocz(size);
813 if (read_in_full(fd, buf, size) != size) {
814 free(buf);
815 close(fd);
816 return -1;
817 }
818 buf[size++] = '\n';
819 close(fd);
820 if (oid_stat) {
821 int pos;
822 if (oid_stat->valid &&
823 !match_stat_data_racy(istate, &oid_stat->stat, &st))
824 ; /* no content change, ss->sha1 still good */
825 else if (istate &&
826 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
827 !ce_stage(istate->cache[pos]) &&
828 ce_uptodate(istate->cache[pos]) &&
829 !would_convert_to_git(istate, fname))
830 oidcpy(&oid_stat->oid,
831 &istate->cache[pos]->oid);
832 else
833 hash_object_file(buf, size, "blob",
834 &oid_stat->oid);
835 fill_stat_data(&oid_stat->stat, &st);
836 oid_stat->valid = 1;
837 }
838 }
839
840 add_patterns_from_buffer(buf, size, base, baselen, pl);
841 return 0;
842 }
843
844 static int add_patterns_from_buffer(char *buf, size_t size,
845 const char *base, int baselen,
846 struct pattern_list *pl)
847 {
848 int i, lineno = 1;
849 char *entry;
850
851 pl->filebuf = buf;
852
853 if (skip_utf8_bom(&buf, size))
854 size -= buf - pl->filebuf;
855
856 entry = buf;
857
858 for (i = 0; i < size; i++) {
859 if (buf[i] == '\n') {
860 if (entry != buf + i && entry[0] != '#') {
861 buf[i - (i && buf[i-1] == '\r')] = 0;
862 trim_trailing_spaces(entry);
863 add_pattern(entry, base, baselen, pl, lineno);
864 }
865 lineno++;
866 entry = buf + i + 1;
867 }
868 }
869 return 0;
870 }
871
872 int add_patterns_from_file_to_list(const char *fname, const char *base,
873 int baselen, struct pattern_list *pl,
874 struct index_state *istate)
875 {
876 return add_patterns(fname, base, baselen, pl, istate, NULL);
877 }
878
879 int add_patterns_from_blob_to_list(
880 struct object_id *oid,
881 const char *base, int baselen,
882 struct pattern_list *pl)
883 {
884 char *buf;
885 size_t size;
886 int r;
887
888 r = do_read_blob(oid, NULL, &size, &buf);
889 if (r != 1)
890 return r;
891
892 add_patterns_from_buffer(buf, size, base, baselen, pl);
893 return 0;
894 }
895
896 struct pattern_list *add_pattern_list(struct dir_struct *dir,
897 int group_type, const char *src)
898 {
899 struct pattern_list *pl;
900 struct exclude_list_group *group;
901
902 group = &dir->exclude_list_group[group_type];
903 ALLOC_GROW(group->pl, group->nr + 1, group->alloc);
904 pl = &group->pl[group->nr++];
905 memset(pl, 0, sizeof(*pl));
906 pl->src = src;
907 return pl;
908 }
909
910 /*
911 * Used to set up core.excludesfile and .git/info/exclude lists.
912 */
913 static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname,
914 struct oid_stat *oid_stat)
915 {
916 struct pattern_list *pl;
917 /*
918 * catch setup_standard_excludes() that's called before
919 * dir->untracked is assigned. That function behaves
920 * differently when dir->untracked is non-NULL.
921 */
922 if (!dir->untracked)
923 dir->unmanaged_exclude_files++;
924 pl = add_pattern_list(dir, EXC_FILE, fname);
925 if (add_patterns(fname, "", 0, pl, NULL, oid_stat) < 0)
926 die(_("cannot use %s as an exclude file"), fname);
927 }
928
929 void add_patterns_from_file(struct dir_struct *dir, const char *fname)
930 {
931 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
932 add_patterns_from_file_1(dir, fname, NULL);
933 }
934
935 int match_basename(const char *basename, int basenamelen,
936 const char *pattern, int prefix, int patternlen,
937 unsigned flags)
938 {
939 if (prefix == patternlen) {
940 if (patternlen == basenamelen &&
941 !fspathncmp(pattern, basename, basenamelen))
942 return 1;
943 } else if (flags & PATTERN_FLAG_ENDSWITH) {
944 /* "*literal" matching against "fooliteral" */
945 if (patternlen - 1 <= basenamelen &&
946 !fspathncmp(pattern + 1,
947 basename + basenamelen - (patternlen - 1),
948 patternlen - 1))
949 return 1;
950 } else {
951 if (fnmatch_icase_mem(pattern, patternlen,
952 basename, basenamelen,
953 0) == 0)
954 return 1;
955 }
956 return 0;
957 }
958
959 int match_pathname(const char *pathname, int pathlen,
960 const char *base, int baselen,
961 const char *pattern, int prefix, int patternlen,
962 unsigned flags)
963 {
964 const char *name;
965 int namelen;
966
967 /*
968 * match with FNM_PATHNAME; the pattern has base implicitly
969 * in front of it.
970 */
971 if (*pattern == '/') {
972 pattern++;
973 patternlen--;
974 prefix--;
975 }
976
977 /*
978 * baselen does not count the trailing slash. base[] may or
979 * may not end with a trailing slash though.
980 */
981 if (pathlen < baselen + 1 ||
982 (baselen && pathname[baselen] != '/') ||
983 fspathncmp(pathname, base, baselen))
984 return 0;
985
986 namelen = baselen ? pathlen - baselen - 1 : pathlen;
987 name = pathname + pathlen - namelen;
988
989 if (prefix) {
990 /*
991 * if the non-wildcard part is longer than the
992 * remaining pathname, surely it cannot match.
993 */
994 if (prefix > namelen)
995 return 0;
996
997 if (fspathncmp(pattern, name, prefix))
998 return 0;
999 pattern += prefix;
1000 patternlen -= prefix;
1001 name += prefix;
1002 namelen -= prefix;
1003
1004 /*
1005 * If the whole pattern did not have a wildcard,
1006 * then our prefix match is all we need; we
1007 * do not need to call fnmatch at all.
1008 */
1009 if (!patternlen && !namelen)
1010 return 1;
1011 }
1012
1013 return fnmatch_icase_mem(pattern, patternlen,
1014 name, namelen,
1015 WM_PATHNAME) == 0;
1016 }
1017
1018 /*
1019 * Scan the given exclude list in reverse to see whether pathname
1020 * should be ignored. The first match (i.e. the last on the list), if
1021 * any, determines the fate. Returns the exclude_list element which
1022 * matched, or NULL for undecided.
1023 */
1024 static struct path_pattern *last_matching_pattern_from_list(const char *pathname,
1025 int pathlen,
1026 const char *basename,
1027 int *dtype,
1028 struct pattern_list *pl,
1029 struct index_state *istate)
1030 {
1031 struct path_pattern *res = NULL; /* undecided */
1032 int i;
1033
1034 if (!pl->nr)
1035 return NULL; /* undefined */
1036
1037 for (i = pl->nr - 1; 0 <= i; i--) {
1038 struct path_pattern *pattern = pl->patterns[i];
1039 const char *exclude = pattern->pattern;
1040 int prefix = pattern->nowildcardlen;
1041
1042 if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) {
1043 if (*dtype == DT_UNKNOWN)
1044 *dtype = get_dtype(NULL, istate, pathname, pathlen);
1045 if (*dtype != DT_DIR)
1046 continue;
1047 }
1048
1049 if (pattern->flags & PATTERN_FLAG_NODIR) {
1050 if (match_basename(basename,
1051 pathlen - (basename - pathname),
1052 exclude, prefix, pattern->patternlen,
1053 pattern->flags)) {
1054 res = pattern;
1055 break;
1056 }
1057 continue;
1058 }
1059
1060 assert(pattern->baselen == 0 ||
1061 pattern->base[pattern->baselen - 1] == '/');
1062 if (match_pathname(pathname, pathlen,
1063 pattern->base,
1064 pattern->baselen ? pattern->baselen - 1 : 0,
1065 exclude, prefix, pattern->patternlen,
1066 pattern->flags)) {
1067 res = pattern;
1068 break;
1069 }
1070 }
1071 return res;
1072 }
1073
1074 /*
1075 * Scan the list of patterns to determine if the ordered list
1076 * of patterns matches on 'pathname'.
1077 *
1078 * Return 1 for a match, 0 for not matched and -1 for undecided.
1079 */
1080 enum pattern_match_result path_matches_pattern_list(
1081 const char *pathname, int pathlen,
1082 const char *basename, int *dtype,
1083 struct pattern_list *pl,
1084 struct index_state *istate)
1085 {
1086 struct path_pattern *pattern;
1087 pattern = last_matching_pattern_from_list(pathname, pathlen, basename,
1088 dtype, pl, istate);
1089 if (pattern) {
1090 if (pattern->flags & PATTERN_FLAG_NEGATIVE)
1091 return NOT_MATCHED;
1092 else
1093 return MATCHED;
1094 }
1095
1096 return UNDECIDED;
1097 }
1098
1099 static struct path_pattern *last_matching_pattern_from_lists(
1100 struct dir_struct *dir, struct index_state *istate,
1101 const char *pathname, int pathlen,
1102 const char *basename, int *dtype_p)
1103 {
1104 int i, j;
1105 struct exclude_list_group *group;
1106 struct path_pattern *pattern;
1107 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1108 group = &dir->exclude_list_group[i];
1109 for (j = group->nr - 1; j >= 0; j--) {
1110 pattern = last_matching_pattern_from_list(
1111 pathname, pathlen, basename, dtype_p,
1112 &group->pl[j], istate);
1113 if (pattern)
1114 return pattern;
1115 }
1116 }
1117 return NULL;
1118 }
1119
1120 /*
1121 * Loads the per-directory exclude list for the substring of base
1122 * which has a char length of baselen.
1123 */
1124 static void prep_exclude(struct dir_struct *dir,
1125 struct index_state *istate,
1126 const char *base, int baselen)
1127 {
1128 struct exclude_list_group *group;
1129 struct pattern_list *pl;
1130 struct exclude_stack *stk = NULL;
1131 struct untracked_cache_dir *untracked;
1132 int current;
1133
1134 group = &dir->exclude_list_group[EXC_DIRS];
1135
1136 /*
1137 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1138 * which originate from directories not in the prefix of the
1139 * path being checked.
1140 */
1141 while ((stk = dir->exclude_stack) != NULL) {
1142 if (stk->baselen <= baselen &&
1143 !strncmp(dir->basebuf.buf, base, stk->baselen))
1144 break;
1145 pl = &group->pl[dir->exclude_stack->exclude_ix];
1146 dir->exclude_stack = stk->prev;
1147 dir->pattern = NULL;
1148 free((char *)pl->src); /* see strbuf_detach() below */
1149 clear_pattern_list(pl);
1150 free(stk);
1151 group->nr--;
1152 }
1153
1154 /* Skip traversing into sub directories if the parent is excluded */
1155 if (dir->pattern)
1156 return;
1157
1158 /*
1159 * Lazy initialization. All call sites currently just
1160 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1161 * them seems lots of work for little benefit.
1162 */
1163 if (!dir->basebuf.buf)
1164 strbuf_init(&dir->basebuf, PATH_MAX);
1165
1166 /* Read from the parent directories and push them down. */
1167 current = stk ? stk->baselen : -1;
1168 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1169 if (dir->untracked)
1170 untracked = stk ? stk->ucd : dir->untracked->root;
1171 else
1172 untracked = NULL;
1173
1174 while (current < baselen) {
1175 const char *cp;
1176 struct oid_stat oid_stat;
1177
1178 stk = xcalloc(1, sizeof(*stk));
1179 if (current < 0) {
1180 cp = base;
1181 current = 0;
1182 } else {
1183 cp = strchr(base + current + 1, '/');
1184 if (!cp)
1185 die("oops in prep_exclude");
1186 cp++;
1187 untracked =
1188 lookup_untracked(dir->untracked, untracked,
1189 base + current,
1190 cp - base - current);
1191 }
1192 stk->prev = dir->exclude_stack;
1193 stk->baselen = cp - base;
1194 stk->exclude_ix = group->nr;
1195 stk->ucd = untracked;
1196 pl = add_pattern_list(dir, EXC_DIRS, NULL);
1197 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1198 assert(stk->baselen == dir->basebuf.len);
1199
1200 /* Abort if the directory is excluded */
1201 if (stk->baselen) {
1202 int dt = DT_DIR;
1203 dir->basebuf.buf[stk->baselen - 1] = 0;
1204 dir->pattern = last_matching_pattern_from_lists(dir,
1205 istate,
1206 dir->basebuf.buf, stk->baselen - 1,
1207 dir->basebuf.buf + current, &dt);
1208 dir->basebuf.buf[stk->baselen - 1] = '/';
1209 if (dir->pattern &&
1210 dir->pattern->flags & PATTERN_FLAG_NEGATIVE)
1211 dir->pattern = NULL;
1212 if (dir->pattern) {
1213 dir->exclude_stack = stk;
1214 return;
1215 }
1216 }
1217
1218 /* Try to read per-directory file */
1219 oidclr(&oid_stat.oid);
1220 oid_stat.valid = 0;
1221 if (dir->exclude_per_dir &&
1222 /*
1223 * If we know that no files have been added in
1224 * this directory (i.e. valid_cached_dir() has
1225 * been executed and set untracked->valid) ..
1226 */
1227 (!untracked || !untracked->valid ||
1228 /*
1229 * .. and .gitignore does not exist before
1230 * (i.e. null exclude_oid). Then we can skip
1231 * loading .gitignore, which would result in
1232 * ENOENT anyway.
1233 */
1234 !is_null_oid(&untracked->exclude_oid))) {
1235 /*
1236 * dir->basebuf gets reused by the traversal, but we
1237 * need fname to remain unchanged to ensure the src
1238 * member of each struct path_pattern correctly
1239 * back-references its source file. Other invocations
1240 * of add_pattern_list provide stable strings, so we
1241 * strbuf_detach() and free() here in the caller.
1242 */
1243 struct strbuf sb = STRBUF_INIT;
1244 strbuf_addbuf(&sb, &dir->basebuf);
1245 strbuf_addstr(&sb, dir->exclude_per_dir);
1246 pl->src = strbuf_detach(&sb, NULL);
1247 add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1248 untracked ? &oid_stat : NULL);
1249 }
1250 /*
1251 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1252 * will first be called in valid_cached_dir() then maybe many
1253 * times more in last_matching_pattern(). When the cache is
1254 * used, last_matching_pattern() will not be called and
1255 * reading .gitignore content will be a waste.
1256 *
1257 * So when it's called by valid_cached_dir() and we can get
1258 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1259 * modified on work tree), we could delay reading the
1260 * .gitignore content until we absolutely need it in
1261 * last_matching_pattern(). Be careful about ignore rule
1262 * order, though, if you do that.
1263 */
1264 if (untracked &&
1265 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1266 invalidate_gitignore(dir->untracked, untracked);
1267 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1268 }
1269 dir->exclude_stack = stk;
1270 current = stk->baselen;
1271 }
1272 strbuf_setlen(&dir->basebuf, baselen);
1273 }
1274
1275 /*
1276 * Loads the exclude lists for the directory containing pathname, then
1277 * scans all exclude lists to determine whether pathname is excluded.
1278 * Returns the exclude_list element which matched, or NULL for
1279 * undecided.
1280 */
1281 struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1282 struct index_state *istate,
1283 const char *pathname,
1284 int *dtype_p)
1285 {
1286 int pathlen = strlen(pathname);
1287 const char *basename = strrchr(pathname, '/');
1288 basename = (basename) ? basename+1 : pathname;
1289
1290 prep_exclude(dir, istate, pathname, basename-pathname);
1291
1292 if (dir->pattern)
1293 return dir->pattern;
1294
1295 return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1296 basename, dtype_p);
1297 }
1298
1299 /*
1300 * Loads the exclude lists for the directory containing pathname, then
1301 * scans all exclude lists to determine whether pathname is excluded.
1302 * Returns 1 if true, otherwise 0.
1303 */
1304 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1305 const char *pathname, int *dtype_p)
1306 {
1307 struct path_pattern *pattern =
1308 last_matching_pattern(dir, istate, pathname, dtype_p);
1309 if (pattern)
1310 return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1311 return 0;
1312 }
1313
1314 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1315 {
1316 struct dir_entry *ent;
1317
1318 FLEX_ALLOC_MEM(ent, name, pathname, len);
1319 ent->len = len;
1320 return ent;
1321 }
1322
1323 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1324 struct index_state *istate,
1325 const char *pathname, int len)
1326 {
1327 if (index_file_exists(istate, pathname, len, ignore_case))
1328 return NULL;
1329
1330 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1331 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1332 }
1333
1334 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1335 struct index_state *istate,
1336 const char *pathname, int len)
1337 {
1338 if (!index_name_is_other(istate, pathname, len))
1339 return NULL;
1340
1341 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1342 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1343 }
1344
1345 enum exist_status {
1346 index_nonexistent = 0,
1347 index_directory,
1348 index_gitdir
1349 };
1350
1351 /*
1352 * Do not use the alphabetically sorted index to look up
1353 * the directory name; instead, use the case insensitive
1354 * directory hash.
1355 */
1356 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1357 const char *dirname, int len)
1358 {
1359 struct cache_entry *ce;
1360
1361 if (index_dir_exists(istate, dirname, len))
1362 return index_directory;
1363
1364 ce = index_file_exists(istate, dirname, len, ignore_case);
1365 if (ce && S_ISGITLINK(ce->ce_mode))
1366 return index_gitdir;
1367
1368 return index_nonexistent;
1369 }
1370
1371 /*
1372 * The index sorts alphabetically by entry name, which
1373 * means that a gitlink sorts as '\0' at the end, while
1374 * a directory (which is defined not as an entry, but as
1375 * the files it contains) will sort with the '/' at the
1376 * end.
1377 */
1378 static enum exist_status directory_exists_in_index(struct index_state *istate,
1379 const char *dirname, int len)
1380 {
1381 int pos;
1382
1383 if (ignore_case)
1384 return directory_exists_in_index_icase(istate, dirname, len);
1385
1386 pos = index_name_pos(istate, dirname, len);
1387 if (pos < 0)
1388 pos = -pos-1;
1389 while (pos < istate->cache_nr) {
1390 const struct cache_entry *ce = istate->cache[pos++];
1391 unsigned char endchar;
1392
1393 if (strncmp(ce->name, dirname, len))
1394 break;
1395 endchar = ce->name[len];
1396 if (endchar > '/')
1397 break;
1398 if (endchar == '/')
1399 return index_directory;
1400 if (!endchar && S_ISGITLINK(ce->ce_mode))
1401 return index_gitdir;
1402 }
1403 return index_nonexistent;
1404 }
1405
1406 /*
1407 * When we find a directory when traversing the filesystem, we
1408 * have three distinct cases:
1409 *
1410 * - ignore it
1411 * - see it as a directory
1412 * - recurse into it
1413 *
1414 * and which one we choose depends on a combination of existing
1415 * git index contents and the flags passed into the directory
1416 * traversal routine.
1417 *
1418 * Case 1: If we *already* have entries in the index under that
1419 * directory name, we always recurse into the directory to see
1420 * all the files.
1421 *
1422 * Case 2: If we *already* have that directory name as a gitlink,
1423 * we always continue to see it as a gitlink, regardless of whether
1424 * there is an actual git directory there or not (it might not
1425 * be checked out as a subproject!)
1426 *
1427 * Case 3: if we didn't have it in the index previously, we
1428 * have a few sub-cases:
1429 *
1430 * (a) if "show_other_directories" is true, we show it as
1431 * just a directory, unless "hide_empty_directories" is
1432 * also true, in which case we need to check if it contains any
1433 * untracked and / or ignored files.
1434 * (b) if it looks like a git directory, and we don't have
1435 * 'no_gitlinks' set we treat it as a gitlink, and show it
1436 * as a directory.
1437 * (c) otherwise, we recurse into it.
1438 */
1439 static enum path_treatment treat_directory(struct dir_struct *dir,
1440 struct index_state *istate,
1441 struct untracked_cache_dir *untracked,
1442 const char *dirname, int len, int baselen, int exclude,
1443 const struct pathspec *pathspec)
1444 {
1445 /* The "len-1" is to strip the final '/' */
1446 switch (directory_exists_in_index(istate, dirname, len-1)) {
1447 case index_directory:
1448 return path_recurse;
1449
1450 case index_gitdir:
1451 return path_none;
1452
1453 case index_nonexistent:
1454 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1455 break;
1456 if (exclude &&
1457 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1458 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1459
1460 /*
1461 * This is an excluded directory and we are
1462 * showing ignored paths that match an exclude
1463 * pattern. (e.g. show directory as ignored
1464 * only if it matches an exclude pattern).
1465 * This path will either be 'path_excluded`
1466 * (if we are showing empty directories or if
1467 * the directory is not empty), or will be
1468 * 'path_none' (empty directory, and we are
1469 * not showing empty directories).
1470 */
1471 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1472 return path_excluded;
1473
1474 if (read_directory_recursive(dir, istate, dirname, len,
1475 untracked, 1, 1, pathspec) == path_excluded)
1476 return path_excluded;
1477
1478 return path_none;
1479 }
1480 if (!(dir->flags & DIR_NO_GITLINKS)) {
1481 struct strbuf sb = STRBUF_INIT;
1482 strbuf_addstr(&sb, dirname);
1483 if (is_nonbare_repository_dir(&sb))
1484 return exclude ? path_excluded : path_untracked;
1485 strbuf_release(&sb);
1486 }
1487 return path_recurse;
1488 }
1489
1490 /* This is the "show_other_directories" case */
1491
1492 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1493 return exclude ? path_excluded : path_untracked;
1494
1495 untracked = lookup_untracked(dir->untracked, untracked,
1496 dirname + baselen, len - baselen);
1497
1498 /*
1499 * If this is an excluded directory, then we only need to check if
1500 * the directory contains any files.
1501 */
1502 return read_directory_recursive(dir, istate, dirname, len,
1503 untracked, 1, exclude, pathspec);
1504 }
1505
1506 /*
1507 * This is an inexact early pruning of any recursive directory
1508 * reading - if the path cannot possibly be in the pathspec,
1509 * return true, and we'll skip it early.
1510 */
1511 static int simplify_away(const char *path, int pathlen,
1512 const struct pathspec *pathspec)
1513 {
1514 int i;
1515
1516 if (!pathspec || !pathspec->nr)
1517 return 0;
1518
1519 GUARD_PATHSPEC(pathspec,
1520 PATHSPEC_FROMTOP |
1521 PATHSPEC_MAXDEPTH |
1522 PATHSPEC_LITERAL |
1523 PATHSPEC_GLOB |
1524 PATHSPEC_ICASE |
1525 PATHSPEC_EXCLUDE |
1526 PATHSPEC_ATTR);
1527
1528 for (i = 0; i < pathspec->nr; i++) {
1529 const struct pathspec_item *item = &pathspec->items[i];
1530 int len = item->nowildcard_len;
1531
1532 if (len > pathlen)
1533 len = pathlen;
1534 if (!ps_strncmp(item, item->match, path, len))
1535 return 0;
1536 }
1537
1538 return 1;
1539 }
1540
1541 /*
1542 * This function tells us whether an excluded path matches a
1543 * list of "interesting" pathspecs. That is, whether a path matched
1544 * by any of the pathspecs could possibly be ignored by excluding
1545 * the specified path. This can happen if:
1546 *
1547 * 1. the path is mentioned explicitly in the pathspec
1548 *
1549 * 2. the path is a directory prefix of some element in the
1550 * pathspec
1551 */
1552 static int exclude_matches_pathspec(const char *path, int pathlen,
1553 const struct pathspec *pathspec)
1554 {
1555 int i;
1556
1557 if (!pathspec || !pathspec->nr)
1558 return 0;
1559
1560 GUARD_PATHSPEC(pathspec,
1561 PATHSPEC_FROMTOP |
1562 PATHSPEC_MAXDEPTH |
1563 PATHSPEC_LITERAL |
1564 PATHSPEC_GLOB |
1565 PATHSPEC_ICASE |
1566 PATHSPEC_EXCLUDE);
1567
1568 for (i = 0; i < pathspec->nr; i++) {
1569 const struct pathspec_item *item = &pathspec->items[i];
1570 int len = item->nowildcard_len;
1571
1572 if (len == pathlen &&
1573 !ps_strncmp(item, item->match, path, pathlen))
1574 return 1;
1575 if (len > pathlen &&
1576 item->match[pathlen] == '/' &&
1577 !ps_strncmp(item, item->match, path, pathlen))
1578 return 1;
1579 }
1580 return 0;
1581 }
1582
1583 static int get_index_dtype(struct index_state *istate,
1584 const char *path, int len)
1585 {
1586 int pos;
1587 const struct cache_entry *ce;
1588
1589 ce = index_file_exists(istate, path, len, 0);
1590 if (ce) {
1591 if (!ce_uptodate(ce))
1592 return DT_UNKNOWN;
1593 if (S_ISGITLINK(ce->ce_mode))
1594 return DT_DIR;
1595 /*
1596 * Nobody actually cares about the
1597 * difference between DT_LNK and DT_REG
1598 */
1599 return DT_REG;
1600 }
1601
1602 /* Try to look it up as a directory */
1603 pos = index_name_pos(istate, path, len);
1604 if (pos >= 0)
1605 return DT_UNKNOWN;
1606 pos = -pos-1;
1607 while (pos < istate->cache_nr) {
1608 ce = istate->cache[pos++];
1609 if (strncmp(ce->name, path, len))
1610 break;
1611 if (ce->name[len] > '/')
1612 break;
1613 if (ce->name[len] < '/')
1614 continue;
1615 if (!ce_uptodate(ce))
1616 break; /* continue? */
1617 return DT_DIR;
1618 }
1619 return DT_UNKNOWN;
1620 }
1621
1622 static int get_dtype(struct dirent *de, struct index_state *istate,
1623 const char *path, int len)
1624 {
1625 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1626 struct stat st;
1627
1628 if (dtype != DT_UNKNOWN)
1629 return dtype;
1630 dtype = get_index_dtype(istate, path, len);
1631 if (dtype != DT_UNKNOWN)
1632 return dtype;
1633 if (lstat(path, &st))
1634 return dtype;
1635 if (S_ISREG(st.st_mode))
1636 return DT_REG;
1637 if (S_ISDIR(st.st_mode))
1638 return DT_DIR;
1639 if (S_ISLNK(st.st_mode))
1640 return DT_LNK;
1641 return dtype;
1642 }
1643
1644 static enum path_treatment treat_one_path(struct dir_struct *dir,
1645 struct untracked_cache_dir *untracked,
1646 struct index_state *istate,
1647 struct strbuf *path,
1648 int baselen,
1649 const struct pathspec *pathspec,
1650 int dtype, struct dirent *de)
1651 {
1652 int exclude;
1653 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1654 enum path_treatment path_treatment;
1655
1656 if (dtype == DT_UNKNOWN)
1657 dtype = get_dtype(de, istate, path->buf, path->len);
1658
1659 /* Always exclude indexed files */
1660 if (dtype != DT_DIR && has_path_in_index)
1661 return path_none;
1662
1663 /*
1664 * When we are looking at a directory P in the working tree,
1665 * there are three cases:
1666 *
1667 * (1) P exists in the index. Everything inside the directory P in
1668 * the working tree needs to go when P is checked out from the
1669 * index.
1670 *
1671 * (2) P does not exist in the index, but there is P/Q in the index.
1672 * We know P will stay a directory when we check out the contents
1673 * of the index, but we do not know yet if there is a directory
1674 * P/Q in the working tree to be killed, so we need to recurse.
1675 *
1676 * (3) P does not exist in the index, and there is no P/Q in the index
1677 * to require P to be a directory, either. Only in this case, we
1678 * know that everything inside P will not be killed without
1679 * recursing.
1680 */
1681 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1682 (dtype == DT_DIR) &&
1683 !has_path_in_index &&
1684 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1685 return path_none;
1686
1687 exclude = is_excluded(dir, istate, path->buf, &dtype);
1688
1689 /*
1690 * Excluded? If we don't explicitly want to show
1691 * ignored files, ignore it
1692 */
1693 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1694 return path_excluded;
1695
1696 switch (dtype) {
1697 default:
1698 return path_none;
1699 case DT_DIR:
1700 strbuf_addch(path, '/');
1701 path_treatment = treat_directory(dir, istate, untracked,
1702 path->buf, path->len,
1703 baselen, exclude, pathspec);
1704 /*
1705 * If 1) we only want to return directories that
1706 * match an exclude pattern and 2) this directory does
1707 * not match an exclude pattern but all of its
1708 * contents are excluded, then indicate that we should
1709 * recurse into this directory (instead of marking the
1710 * directory itself as an ignored path).
1711 */
1712 if (!exclude &&
1713 path_treatment == path_excluded &&
1714 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1715 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1716 return path_recurse;
1717 return path_treatment;
1718 case DT_REG:
1719 case DT_LNK:
1720 return exclude ? path_excluded : path_untracked;
1721 }
1722 }
1723
1724 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1725 struct untracked_cache_dir *untracked,
1726 struct cached_dir *cdir,
1727 struct index_state *istate,
1728 struct strbuf *path,
1729 int baselen,
1730 const struct pathspec *pathspec)
1731 {
1732 strbuf_setlen(path, baselen);
1733 if (!cdir->ucd) {
1734 strbuf_addstr(path, cdir->file);
1735 return path_untracked;
1736 }
1737 strbuf_addstr(path, cdir->ucd->name);
1738 /* treat_one_path() does this before it calls treat_directory() */
1739 strbuf_complete(path, '/');
1740 if (cdir->ucd->check_only)
1741 /*
1742 * check_only is set as a result of treat_directory() getting
1743 * to its bottom. Verify again the same set of directories
1744 * with check_only set.
1745 */
1746 return read_directory_recursive(dir, istate, path->buf, path->len,
1747 cdir->ucd, 1, 0, pathspec);
1748 /*
1749 * We get path_recurse in the first run when
1750 * directory_exists_in_index() returns index_nonexistent. We
1751 * are sure that new changes in the index does not impact the
1752 * outcome. Return now.
1753 */
1754 return path_recurse;
1755 }
1756
1757 static enum path_treatment treat_path(struct dir_struct *dir,
1758 struct untracked_cache_dir *untracked,
1759 struct cached_dir *cdir,
1760 struct index_state *istate,
1761 struct strbuf *path,
1762 int baselen,
1763 const struct pathspec *pathspec)
1764 {
1765 int dtype;
1766 struct dirent *de = cdir->de;
1767
1768 if (!de)
1769 return treat_path_fast(dir, untracked, cdir, istate, path,
1770 baselen, pathspec);
1771 if (is_dot_or_dotdot(de->d_name) || !fspathcmp(de->d_name, ".git"))
1772 return path_none;
1773 strbuf_setlen(path, baselen);
1774 strbuf_addstr(path, de->d_name);
1775 if (simplify_away(path->buf, path->len, pathspec))
1776 return path_none;
1777
1778 dtype = DTYPE(de);
1779 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1780 }
1781
1782 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1783 {
1784 if (!dir)
1785 return;
1786 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1787 dir->untracked_alloc);
1788 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1789 }
1790
1791 static int valid_cached_dir(struct dir_struct *dir,
1792 struct untracked_cache_dir *untracked,
1793 struct index_state *istate,
1794 struct strbuf *path,
1795 int check_only)
1796 {
1797 struct stat st;
1798
1799 if (!untracked)
1800 return 0;
1801
1802 /*
1803 * With fsmonitor, we can trust the untracked cache's valid field.
1804 */
1805 refresh_fsmonitor(istate);
1806 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1807 if (lstat(path->len ? path->buf : ".", &st)) {
1808 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1809 return 0;
1810 }
1811 if (!untracked->valid ||
1812 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1813 fill_stat_data(&untracked->stat_data, &st);
1814 return 0;
1815 }
1816 }
1817
1818 if (untracked->check_only != !!check_only)
1819 return 0;
1820
1821 /*
1822 * prep_exclude will be called eventually on this directory,
1823 * but it's called much later in last_matching_pattern(). We
1824 * need it now to determine the validity of the cache for this
1825 * path. The next calls will be nearly no-op, the way
1826 * prep_exclude() is designed.
1827 */
1828 if (path->len && path->buf[path->len - 1] != '/') {
1829 strbuf_addch(path, '/');
1830 prep_exclude(dir, istate, path->buf, path->len);
1831 strbuf_setlen(path, path->len - 1);
1832 } else
1833 prep_exclude(dir, istate, path->buf, path->len);
1834
1835 /* hopefully prep_exclude() haven't invalidated this entry... */
1836 return untracked->valid;
1837 }
1838
1839 static int open_cached_dir(struct cached_dir *cdir,
1840 struct dir_struct *dir,
1841 struct untracked_cache_dir *untracked,
1842 struct index_state *istate,
1843 struct strbuf *path,
1844 int check_only)
1845 {
1846 const char *c_path;
1847
1848 memset(cdir, 0, sizeof(*cdir));
1849 cdir->untracked = untracked;
1850 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1851 return 0;
1852 c_path = path->len ? path->buf : ".";
1853 cdir->fdir = opendir(c_path);
1854 if (!cdir->fdir)
1855 warning_errno(_("could not open directory '%s'"), c_path);
1856 if (dir->untracked) {
1857 invalidate_directory(dir->untracked, untracked);
1858 dir->untracked->dir_opened++;
1859 }
1860 if (!cdir->fdir)
1861 return -1;
1862 return 0;
1863 }
1864
1865 static int read_cached_dir(struct cached_dir *cdir)
1866 {
1867 if (cdir->fdir) {
1868 cdir->de = readdir(cdir->fdir);
1869 if (!cdir->de)
1870 return -1;
1871 return 0;
1872 }
1873 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1874 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1875 if (!d->recurse) {
1876 cdir->nr_dirs++;
1877 continue;
1878 }
1879 cdir->ucd = d;
1880 cdir->nr_dirs++;
1881 return 0;
1882 }
1883 cdir->ucd = NULL;
1884 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1885 struct untracked_cache_dir *d = cdir->untracked;
1886 cdir->file = d->untracked[cdir->nr_files++];
1887 return 0;
1888 }
1889 return -1;
1890 }
1891
1892 static void close_cached_dir(struct cached_dir *cdir)
1893 {
1894 if (cdir->fdir)
1895 closedir(cdir->fdir);
1896 /*
1897 * We have gone through this directory and found no untracked
1898 * entries. Mark it valid.
1899 */
1900 if (cdir->untracked) {
1901 cdir->untracked->valid = 1;
1902 cdir->untracked->recurse = 1;
1903 }
1904 }
1905
1906 /*
1907 * Read a directory tree. We currently ignore anything but
1908 * directories, regular files and symlinks. That's because git
1909 * doesn't handle them at all yet. Maybe that will change some
1910 * day.
1911 *
1912 * Also, we ignore the name ".git" (even if it is not a directory).
1913 * That likely will not change.
1914 *
1915 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1916 * to signal that a file was found. This is the least significant value that
1917 * indicates that a file was encountered that does not depend on the order of
1918 * whether an untracked or exluded path was encountered first.
1919 *
1920 * Returns the most significant path_treatment value encountered in the scan.
1921 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1922 * significant path_treatment value that will be returned.
1923 */
1924
1925 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1926 struct index_state *istate, const char *base, int baselen,
1927 struct untracked_cache_dir *untracked, int check_only,
1928 int stop_at_first_file, const struct pathspec *pathspec)
1929 {
1930 struct cached_dir cdir;
1931 enum path_treatment state, subdir_state, dir_state = path_none;
1932 struct strbuf path = STRBUF_INIT;
1933
1934 strbuf_add(&path, base, baselen);
1935
1936 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1937 goto out;
1938
1939 if (untracked)
1940 untracked->check_only = !!check_only;
1941
1942 while (!read_cached_dir(&cdir)) {
1943 /* check how the file or directory should be treated */
1944 state = treat_path(dir, untracked, &cdir, istate, &path,
1945 baselen, pathspec);
1946
1947 if (state > dir_state)
1948 dir_state = state;
1949
1950 /* recurse into subdir if instructed by treat_path */
1951 if ((state == path_recurse) ||
1952 ((state == path_untracked) &&
1953 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1954 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
1955 struct untracked_cache_dir *ud;
1956 ud = lookup_untracked(dir->untracked, untracked,
1957 path.buf + baselen,
1958 path.len - baselen);
1959 subdir_state =
1960 read_directory_recursive(dir, istate, path.buf,
1961 path.len, ud,
1962 check_only, stop_at_first_file, pathspec);
1963 if (subdir_state > dir_state)
1964 dir_state = subdir_state;
1965 }
1966
1967 if (check_only) {
1968 if (stop_at_first_file) {
1969 /*
1970 * If stopping at first file, then
1971 * signal that a file was found by
1972 * returning `path_excluded`. This is
1973 * to return a consistent value
1974 * regardless of whether an ignored or
1975 * excluded file happened to be
1976 * encountered 1st.
1977 *
1978 * In current usage, the
1979 * `stop_at_first_file` is passed when
1980 * an ancestor directory has matched
1981 * an exclude pattern, so any found
1982 * files will be excluded.
1983 */
1984 if (dir_state >= path_excluded) {
1985 dir_state = path_excluded;
1986 break;
1987 }
1988 }
1989
1990 /* abort early if maximum state has been reached */
1991 if (dir_state == path_untracked) {
1992 if (cdir.fdir)
1993 add_untracked(untracked, path.buf + baselen);
1994 break;
1995 }
1996 /* skip the dir_add_* part */
1997 continue;
1998 }
1999
2000 /* add the path to the appropriate result list */
2001 switch (state) {
2002 case path_excluded:
2003 if (dir->flags & DIR_SHOW_IGNORED)
2004 dir_add_name(dir, istate, path.buf, path.len);
2005 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2006 ((dir->flags & DIR_COLLECT_IGNORED) &&
2007 exclude_matches_pathspec(path.buf, path.len,
2008 pathspec)))
2009 dir_add_ignored(dir, istate, path.buf, path.len);
2010 break;
2011
2012 case path_untracked:
2013 if (dir->flags & DIR_SHOW_IGNORED)
2014 break;
2015 dir_add_name(dir, istate, path.buf, path.len);
2016 if (cdir.fdir)
2017 add_untracked(untracked, path.buf + baselen);
2018 break;
2019
2020 default:
2021 break;
2022 }
2023 }
2024 close_cached_dir(&cdir);
2025 out:
2026 strbuf_release(&path);
2027
2028 return dir_state;
2029 }
2030
2031 int cmp_dir_entry(const void *p1, const void *p2)
2032 {
2033 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2034 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2035
2036 return name_compare(e1->name, e1->len, e2->name, e2->len);
2037 }
2038
2039 /* check if *out lexically strictly contains *in */
2040 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2041 {
2042 return (out->len < in->len) &&
2043 (out->name[out->len - 1] == '/') &&
2044 !memcmp(out->name, in->name, out->len);
2045 }
2046
2047 static int treat_leading_path(struct dir_struct *dir,
2048 struct index_state *istate,
2049 const char *path, int len,
2050 const struct pathspec *pathspec)
2051 {
2052 struct strbuf sb = STRBUF_INIT;
2053 int baselen, rc = 0;
2054 const char *cp;
2055 int old_flags = dir->flags;
2056
2057 while (len && path[len - 1] == '/')
2058 len--;
2059 if (!len)
2060 return 1;
2061 baselen = 0;
2062 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
2063 while (1) {
2064 cp = path + baselen + !!baselen;
2065 cp = memchr(cp, '/', path + len - cp);
2066 if (!cp)
2067 baselen = len;
2068 else
2069 baselen = cp - path;
2070 strbuf_setlen(&sb, 0);
2071 strbuf_add(&sb, path, baselen);
2072 if (!is_directory(sb.buf))
2073 break;
2074 if (simplify_away(sb.buf, sb.len, pathspec))
2075 break;
2076 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
2077 DT_DIR, NULL) == path_none)
2078 break; /* do not recurse into it */
2079 if (len <= baselen) {
2080 rc = 1;
2081 break; /* finished checking */
2082 }
2083 }
2084 strbuf_release(&sb);
2085 dir->flags = old_flags;
2086 return rc;
2087 }
2088
2089 static const char *get_ident_string(void)
2090 {
2091 static struct strbuf sb = STRBUF_INIT;
2092 struct utsname uts;
2093
2094 if (sb.len)
2095 return sb.buf;
2096 if (uname(&uts) < 0)
2097 die_errno(_("failed to get kernel name and information"));
2098 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2099 uts.sysname);
2100 return sb.buf;
2101 }
2102
2103 static int ident_in_untracked(const struct untracked_cache *uc)
2104 {
2105 /*
2106 * Previous git versions may have saved many NUL separated
2107 * strings in the "ident" field, but it is insane to manage
2108 * many locations, so just take care of the first one.
2109 */
2110
2111 return !strcmp(uc->ident.buf, get_ident_string());
2112 }
2113
2114 static void set_untracked_ident(struct untracked_cache *uc)
2115 {
2116 strbuf_reset(&uc->ident);
2117 strbuf_addstr(&uc->ident, get_ident_string());
2118
2119 /*
2120 * This strbuf used to contain a list of NUL separated
2121 * strings, so save NUL too for backward compatibility.
2122 */
2123 strbuf_addch(&uc->ident, 0);
2124 }
2125
2126 static void new_untracked_cache(struct index_state *istate)
2127 {
2128 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2129 strbuf_init(&uc->ident, 100);
2130 uc->exclude_per_dir = ".gitignore";
2131 /* should be the same flags used by git-status */
2132 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2133 set_untracked_ident(uc);
2134 istate->untracked = uc;
2135 istate->cache_changed |= UNTRACKED_CHANGED;
2136 }
2137
2138 void add_untracked_cache(struct index_state *istate)
2139 {
2140 if (!istate->untracked) {
2141 new_untracked_cache(istate);
2142 } else {
2143 if (!ident_in_untracked(istate->untracked)) {
2144 free_untracked_cache(istate->untracked);
2145 new_untracked_cache(istate);
2146 }
2147 }
2148 }
2149
2150 void remove_untracked_cache(struct index_state *istate)
2151 {
2152 if (istate->untracked) {
2153 free_untracked_cache(istate->untracked);
2154 istate->untracked = NULL;
2155 istate->cache_changed |= UNTRACKED_CHANGED;
2156 }
2157 }
2158
2159 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2160 int base_len,
2161 const struct pathspec *pathspec)
2162 {
2163 struct untracked_cache_dir *root;
2164 static int untracked_cache_disabled = -1;
2165
2166 if (!dir->untracked)
2167 return NULL;
2168 if (untracked_cache_disabled < 0)
2169 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2170 if (untracked_cache_disabled)
2171 return NULL;
2172
2173 /*
2174 * We only support $GIT_DIR/info/exclude and core.excludesfile
2175 * as the global ignore rule files. Any other additions
2176 * (e.g. from command line) invalidate the cache. This
2177 * condition also catches running setup_standard_excludes()
2178 * before setting dir->untracked!
2179 */
2180 if (dir->unmanaged_exclude_files)
2181 return NULL;
2182
2183 /*
2184 * Optimize for the main use case only: whole-tree git
2185 * status. More work involved in treat_leading_path() if we
2186 * use cache on just a subset of the worktree. pathspec
2187 * support could make the matter even worse.
2188 */
2189 if (base_len || (pathspec && pathspec->nr))
2190 return NULL;
2191
2192 /* Different set of flags may produce different results */
2193 if (dir->flags != dir->untracked->dir_flags ||
2194 /*
2195 * See treat_directory(), case index_nonexistent. Without
2196 * this flag, we may need to also cache .git file content
2197 * for the resolve_gitlink_ref() call, which we don't.
2198 */
2199 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2200 /* We don't support collecting ignore files */
2201 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2202 DIR_COLLECT_IGNORED)))
2203 return NULL;
2204
2205 /*
2206 * If we use .gitignore in the cache and now you change it to
2207 * .gitexclude, everything will go wrong.
2208 */
2209 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2210 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2211 return NULL;
2212
2213 /*
2214 * EXC_CMDL is not considered in the cache. If people set it,
2215 * skip the cache.
2216 */
2217 if (dir->exclude_list_group[EXC_CMDL].nr)
2218 return NULL;
2219
2220 if (!ident_in_untracked(dir->untracked)) {
2221 warning(_("untracked cache is disabled on this system or location"));
2222 return NULL;
2223 }
2224
2225 if (!dir->untracked->root) {
2226 const int len = sizeof(*dir->untracked->root);
2227 dir->untracked->root = xmalloc(len);
2228 memset(dir->untracked->root, 0, len);
2229 }
2230
2231 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2232 root = dir->untracked->root;
2233 if (!oideq(&dir->ss_info_exclude.oid,
2234 &dir->untracked->ss_info_exclude.oid)) {
2235 invalidate_gitignore(dir->untracked, root);
2236 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2237 }
2238 if (!oideq(&dir->ss_excludes_file.oid,
2239 &dir->untracked->ss_excludes_file.oid)) {
2240 invalidate_gitignore(dir->untracked, root);
2241 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2242 }
2243
2244 /* Make sure this directory is not dropped out at saving phase */
2245 root->recurse = 1;
2246 return root;
2247 }
2248
2249 int read_directory(struct dir_struct *dir, struct index_state *istate,
2250 const char *path, int len, const struct pathspec *pathspec)
2251 {
2252 struct untracked_cache_dir *untracked;
2253
2254 trace_performance_enter();
2255
2256 if (has_symlink_leading_path(path, len)) {
2257 trace_performance_leave("read directory %.*s", len, path);
2258 return dir->nr;
2259 }
2260
2261 untracked = validate_untracked_cache(dir, len, pathspec);
2262 if (!untracked)
2263 /*
2264 * make sure untracked cache code path is disabled,
2265 * e.g. prep_exclude()
2266 */
2267 dir->untracked = NULL;
2268 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2269 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2270 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2271 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2272
2273 /*
2274 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2275 * also pick up untracked contents of untracked dirs; by default
2276 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2277 */
2278 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2279 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2280 int i, j;
2281
2282 /* remove from dir->entries untracked contents of untracked dirs */
2283 for (i = j = 0; j < dir->nr; j++) {
2284 if (i &&
2285 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2286 FREE_AND_NULL(dir->entries[j]);
2287 } else {
2288 dir->entries[i++] = dir->entries[j];
2289 }
2290 }
2291
2292 dir->nr = i;
2293 }
2294
2295 trace_performance_leave("read directory %.*s", len, path);
2296 if (dir->untracked) {
2297 static int force_untracked_cache = -1;
2298 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2299
2300 if (force_untracked_cache < 0)
2301 force_untracked_cache =
2302 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2303 trace_printf_key(&trace_untracked_stats,
2304 "node creation: %u\n"
2305 "gitignore invalidation: %u\n"
2306 "directory invalidation: %u\n"
2307 "opendir: %u\n",
2308 dir->untracked->dir_created,
2309 dir->untracked->gitignore_invalidated,
2310 dir->untracked->dir_invalidated,
2311 dir->untracked->dir_opened);
2312 if (force_untracked_cache &&
2313 dir->untracked == istate->untracked &&
2314 (dir->untracked->dir_opened ||
2315 dir->untracked->gitignore_invalidated ||
2316 dir->untracked->dir_invalidated))
2317 istate->cache_changed |= UNTRACKED_CHANGED;
2318 if (dir->untracked != istate->untracked) {
2319 FREE_AND_NULL(dir->untracked);
2320 }
2321 }
2322 return dir->nr;
2323 }
2324
2325 int file_exists(const char *f)
2326 {
2327 struct stat sb;
2328 return lstat(f, &sb) == 0;
2329 }
2330
2331 int repo_file_exists(struct repository *repo, const char *path)
2332 {
2333 if (repo != the_repository)
2334 BUG("do not know how to check file existence in arbitrary repo");
2335
2336 return file_exists(path);
2337 }
2338
2339 static int cmp_icase(char a, char b)
2340 {
2341 if (a == b)
2342 return 0;
2343 if (ignore_case)
2344 return toupper(a) - toupper(b);
2345 return a - b;
2346 }
2347
2348 /*
2349 * Given two normalized paths (a trailing slash is ok), if subdir is
2350 * outside dir, return -1. Otherwise return the offset in subdir that
2351 * can be used as relative path to dir.
2352 */
2353 int dir_inside_of(const char *subdir, const char *dir)
2354 {
2355 int offset = 0;
2356
2357 assert(dir && subdir && *dir && *subdir);
2358
2359 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2360 dir++;
2361 subdir++;
2362 offset++;
2363 }
2364
2365 /* hel[p]/me vs hel[l]/yeah */
2366 if (*dir && *subdir)
2367 return -1;
2368
2369 if (!*subdir)
2370 return !*dir ? offset : -1; /* same dir */
2371
2372 /* foo/[b]ar vs foo/[] */
2373 if (is_dir_sep(dir[-1]))
2374 return is_dir_sep(subdir[-1]) ? offset : -1;
2375
2376 /* foo[/]bar vs foo[] */
2377 return is_dir_sep(*subdir) ? offset + 1 : -1;
2378 }
2379
2380 int is_inside_dir(const char *dir)
2381 {
2382 char *cwd;
2383 int rc;
2384
2385 if (!dir)
2386 return 0;
2387
2388 cwd = xgetcwd();
2389 rc = (dir_inside_of(cwd, dir) >= 0);
2390 free(cwd);
2391 return rc;
2392 }
2393
2394 int is_empty_dir(const char *path)
2395 {
2396 DIR *dir = opendir(path);
2397 struct dirent *e;
2398 int ret = 1;
2399
2400 if (!dir)
2401 return 0;
2402
2403 while ((e = readdir(dir)) != NULL)
2404 if (!is_dot_or_dotdot(e->d_name)) {
2405 ret = 0;
2406 break;
2407 }
2408
2409 closedir(dir);
2410 return ret;
2411 }
2412
2413 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2414 {
2415 DIR *dir;
2416 struct dirent *e;
2417 int ret = 0, original_len = path->len, len, kept_down = 0;
2418 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2419 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2420 struct object_id submodule_head;
2421
2422 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2423 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2424 /* Do not descend and nuke a nested git work tree. */
2425 if (kept_up)
2426 *kept_up = 1;
2427 return 0;
2428 }
2429
2430 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2431 dir = opendir(path->buf);
2432 if (!dir) {
2433 if (errno == ENOENT)
2434 return keep_toplevel ? -1 : 0;
2435 else if (errno == EACCES && !keep_toplevel)
2436 /*
2437 * An empty dir could be removable even if it
2438 * is unreadable:
2439 */
2440 return rmdir(path->buf);
2441 else
2442 return -1;
2443 }
2444 strbuf_complete(path, '/');
2445
2446 len = path->len;
2447 while ((e = readdir(dir)) != NULL) {
2448 struct stat st;
2449 if (is_dot_or_dotdot(e->d_name))
2450 continue;
2451
2452 strbuf_setlen(path, len);
2453 strbuf_addstr(path, e->d_name);
2454 if (lstat(path->buf, &st)) {
2455 if (errno == ENOENT)
2456 /*
2457 * file disappeared, which is what we
2458 * wanted anyway
2459 */
2460 continue;
2461 /* fall thru */
2462 } else if (S_ISDIR(st.st_mode)) {
2463 if (!remove_dir_recurse(path, flag, &kept_down))
2464 continue; /* happy */
2465 } else if (!only_empty &&
2466 (!unlink(path->buf) || errno == ENOENT)) {
2467 continue; /* happy, too */
2468 }
2469
2470 /* path too long, stat fails, or non-directory still exists */
2471 ret = -1;
2472 break;
2473 }
2474 closedir(dir);
2475
2476 strbuf_setlen(path, original_len);
2477 if (!ret && !keep_toplevel && !kept_down)
2478 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2479 else if (kept_up)
2480 /*
2481 * report the uplevel that it is not an error that we
2482 * did not rmdir() our directory.
2483 */
2484 *kept_up = !ret;
2485 return ret;
2486 }
2487
2488 int remove_dir_recursively(struct strbuf *path, int flag)
2489 {
2490 return remove_dir_recurse(path, flag, NULL);
2491 }
2492
2493 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2494
2495 void setup_standard_excludes(struct dir_struct *dir)
2496 {
2497 dir->exclude_per_dir = ".gitignore";
2498
2499 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2500 if (!excludes_file)
2501 excludes_file = xdg_config_home("ignore");
2502 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2503 add_patterns_from_file_1(dir, excludes_file,
2504 dir->untracked ? &dir->ss_excludes_file : NULL);
2505
2506 /* per repository user preference */
2507 if (startup_info->have_repository) {
2508 const char *path = git_path_info_exclude();
2509 if (!access_or_warn(path, R_OK, 0))
2510 add_patterns_from_file_1(dir, path,
2511 dir->untracked ? &dir->ss_info_exclude : NULL);
2512 }
2513 }
2514
2515 int remove_path(const char *name)
2516 {
2517 char *slash;
2518
2519 if (unlink(name) && !is_missing_file_error(errno))
2520 return -1;
2521
2522 slash = strrchr(name, '/');
2523 if (slash) {
2524 char *dirs = xstrdup(name);
2525 slash = dirs + (slash - name);
2526 do {
2527 *slash = '\0';
2528 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2529 free(dirs);
2530 }
2531 return 0;
2532 }
2533
2534 /*
2535 * Frees memory within dir which was allocated for exclude lists and
2536 * the exclude_stack. Does not free dir itself.
2537 */
2538 void clear_directory(struct dir_struct *dir)
2539 {
2540 int i, j;
2541 struct exclude_list_group *group;
2542 struct pattern_list *pl;
2543 struct exclude_stack *stk;
2544
2545 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2546 group = &dir->exclude_list_group[i];
2547 for (j = 0; j < group->nr; j++) {
2548 pl = &group->pl[j];
2549 if (i == EXC_DIRS)
2550 free((char *)pl->src);
2551 clear_pattern_list(pl);
2552 }
2553 free(group->pl);
2554 }
2555
2556 stk = dir->exclude_stack;
2557 while (stk) {
2558 struct exclude_stack *prev = stk->prev;
2559 free(stk);
2560 stk = prev;
2561 }
2562 strbuf_release(&dir->basebuf);
2563 }
2564
2565 struct ondisk_untracked_cache {
2566 struct stat_data info_exclude_stat;
2567 struct stat_data excludes_file_stat;
2568 uint32_t dir_flags;
2569 };
2570
2571 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2572
2573 struct write_data {
2574 int index; /* number of written untracked_cache_dir */
2575 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2576 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2577 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2578 struct strbuf out;
2579 struct strbuf sb_stat;
2580 struct strbuf sb_sha1;
2581 };
2582
2583 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2584 {
2585 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2586 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2587 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2588 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2589 to->sd_dev = htonl(from->sd_dev);
2590 to->sd_ino = htonl(from->sd_ino);
2591 to->sd_uid = htonl(from->sd_uid);
2592 to->sd_gid = htonl(from->sd_gid);
2593 to->sd_size = htonl(from->sd_size);
2594 }
2595
2596 static void write_one_dir(struct untracked_cache_dir *untracked,
2597 struct write_data *wd)
2598 {
2599 struct stat_data stat_data;
2600 struct strbuf *out = &wd->out;
2601 unsigned char intbuf[16];
2602 unsigned int intlen, value;
2603 int i = wd->index++;
2604
2605 /*
2606 * untracked_nr should be reset whenever valid is clear, but
2607 * for safety..
2608 */
2609 if (!untracked->valid) {
2610 untracked->untracked_nr = 0;
2611 untracked->check_only = 0;
2612 }
2613
2614 if (untracked->check_only)
2615 ewah_set(wd->check_only, i);
2616 if (untracked->valid) {
2617 ewah_set(wd->valid, i);
2618 stat_data_to_disk(&stat_data, &untracked->stat_data);
2619 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2620 }
2621 if (!is_null_oid(&untracked->exclude_oid)) {
2622 ewah_set(wd->sha1_valid, i);
2623 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
2624 the_hash_algo->rawsz);
2625 }
2626
2627 intlen = encode_varint(untracked->untracked_nr, intbuf);
2628 strbuf_add(out, intbuf, intlen);
2629
2630 /* skip non-recurse directories */
2631 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2632 if (untracked->dirs[i]->recurse)
2633 value++;
2634 intlen = encode_varint(value, intbuf);
2635 strbuf_add(out, intbuf, intlen);
2636
2637 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2638
2639 for (i = 0; i < untracked->untracked_nr; i++)
2640 strbuf_add(out, untracked->untracked[i],
2641 strlen(untracked->untracked[i]) + 1);
2642
2643 for (i = 0; i < untracked->dirs_nr; i++)
2644 if (untracked->dirs[i]->recurse)
2645 write_one_dir(untracked->dirs[i], wd);
2646 }
2647
2648 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2649 {
2650 struct ondisk_untracked_cache *ouc;
2651 struct write_data wd;
2652 unsigned char varbuf[16];
2653 int varint_len;
2654 const unsigned hashsz = the_hash_algo->rawsz;
2655
2656 ouc = xcalloc(1, sizeof(*ouc));
2657 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2658 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2659 ouc->dir_flags = htonl(untracked->dir_flags);
2660
2661 varint_len = encode_varint(untracked->ident.len, varbuf);
2662 strbuf_add(out, varbuf, varint_len);
2663 strbuf_addbuf(out, &untracked->ident);
2664
2665 strbuf_add(out, ouc, sizeof(*ouc));
2666 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
2667 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
2668 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
2669 FREE_AND_NULL(ouc);
2670
2671 if (!untracked->root) {
2672 varint_len = encode_varint(0, varbuf);
2673 strbuf_add(out, varbuf, varint_len);
2674 return;
2675 }
2676
2677 wd.index = 0;
2678 wd.check_only = ewah_new();
2679 wd.valid = ewah_new();
2680 wd.sha1_valid = ewah_new();
2681 strbuf_init(&wd.out, 1024);
2682 strbuf_init(&wd.sb_stat, 1024);
2683 strbuf_init(&wd.sb_sha1, 1024);
2684 write_one_dir(untracked->root, &wd);
2685
2686 varint_len = encode_varint(wd.index, varbuf);
2687 strbuf_add(out, varbuf, varint_len);
2688 strbuf_addbuf(out, &wd.out);
2689 ewah_serialize_strbuf(wd.valid, out);
2690 ewah_serialize_strbuf(wd.check_only, out);
2691 ewah_serialize_strbuf(wd.sha1_valid, out);
2692 strbuf_addbuf(out, &wd.sb_stat);
2693 strbuf_addbuf(out, &wd.sb_sha1);
2694 strbuf_addch(out, '\0'); /* safe guard for string lists */
2695
2696 ewah_free(wd.valid);
2697 ewah_free(wd.check_only);
2698 ewah_free(wd.sha1_valid);
2699 strbuf_release(&wd.out);
2700 strbuf_release(&wd.sb_stat);
2701 strbuf_release(&wd.sb_sha1);
2702 }
2703
2704 static void free_untracked(struct untracked_cache_dir *ucd)
2705 {
2706 int i;
2707 if (!ucd)
2708 return;
2709 for (i = 0; i < ucd->dirs_nr; i++)
2710 free_untracked(ucd->dirs[i]);
2711 for (i = 0; i < ucd->untracked_nr; i++)
2712 free(ucd->untracked[i]);
2713 free(ucd->untracked);
2714 free(ucd->dirs);
2715 free(ucd);
2716 }
2717
2718 void free_untracked_cache(struct untracked_cache *uc)
2719 {
2720 if (uc)
2721 free_untracked(uc->root);
2722 free(uc);
2723 }
2724
2725 struct read_data {
2726 int index;
2727 struct untracked_cache_dir **ucd;
2728 struct ewah_bitmap *check_only;
2729 struct ewah_bitmap *valid;
2730 struct ewah_bitmap *sha1_valid;
2731 const unsigned char *data;
2732 const unsigned char *end;
2733 };
2734
2735 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2736 {
2737 memcpy(to, data, sizeof(*to));
2738 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2739 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2740 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2741 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2742 to->sd_dev = ntohl(to->sd_dev);
2743 to->sd_ino = ntohl(to->sd_ino);
2744 to->sd_uid = ntohl(to->sd_uid);
2745 to->sd_gid = ntohl(to->sd_gid);
2746 to->sd_size = ntohl(to->sd_size);
2747 }
2748
2749 static int read_one_dir(struct untracked_cache_dir **untracked_,
2750 struct read_data *rd)
2751 {
2752 struct untracked_cache_dir ud, *untracked;
2753 const unsigned char *data = rd->data, *end = rd->end;
2754 const unsigned char *eos;
2755 unsigned int value;
2756 int i;
2757
2758 memset(&ud, 0, sizeof(ud));
2759
2760 value = decode_varint(&data);
2761 if (data > end)
2762 return -1;
2763 ud.recurse = 1;
2764 ud.untracked_alloc = value;
2765 ud.untracked_nr = value;
2766 if (ud.untracked_nr)
2767 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2768
2769 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
2770 if (data > end)
2771 return -1;
2772 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2773
2774 eos = memchr(data, '\0', end - data);
2775 if (!eos || eos == end)
2776 return -1;
2777
2778 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
2779 memcpy(untracked, &ud, sizeof(ud));
2780 memcpy(untracked->name, data, eos - data + 1);
2781 data = eos + 1;
2782
2783 for (i = 0; i < untracked->untracked_nr; i++) {
2784 eos = memchr(data, '\0', end - data);
2785 if (!eos || eos == end)
2786 return -1;
2787 untracked->untracked[i] = xmemdupz(data, eos - data);
2788 data = eos + 1;
2789 }
2790
2791 rd->ucd[rd->index++] = untracked;
2792 rd->data = data;
2793
2794 for (i = 0; i < untracked->dirs_nr; i++) {
2795 if (read_one_dir(untracked->dirs + i, rd) < 0)
2796 return -1;
2797 }
2798 return 0;
2799 }
2800
2801 static void set_check_only(size_t pos, void *cb)
2802 {
2803 struct read_data *rd = cb;
2804 struct untracked_cache_dir *ud = rd->ucd[pos];
2805 ud->check_only = 1;
2806 }
2807
2808 static void read_stat(size_t pos, void *cb)
2809 {
2810 struct read_data *rd = cb;
2811 struct untracked_cache_dir *ud = rd->ucd[pos];
2812 if (rd->data + sizeof(struct stat_data) > rd->end) {
2813 rd->data = rd->end + 1;
2814 return;
2815 }
2816 stat_data_from_disk(&ud->stat_data, rd->data);
2817 rd->data += sizeof(struct stat_data);
2818 ud->valid = 1;
2819 }
2820
2821 static void read_oid(size_t pos, void *cb)
2822 {
2823 struct read_data *rd = cb;
2824 struct untracked_cache_dir *ud = rd->ucd[pos];
2825 if (rd->data + the_hash_algo->rawsz > rd->end) {
2826 rd->data = rd->end + 1;
2827 return;
2828 }
2829 hashcpy(ud->exclude_oid.hash, rd->data);
2830 rd->data += the_hash_algo->rawsz;
2831 }
2832
2833 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
2834 const unsigned char *sha1)
2835 {
2836 stat_data_from_disk(&oid_stat->stat, data);
2837 hashcpy(oid_stat->oid.hash, sha1);
2838 oid_stat->valid = 1;
2839 }
2840
2841 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2842 {
2843 struct untracked_cache *uc;
2844 struct read_data rd;
2845 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2846 const char *ident;
2847 int ident_len;
2848 ssize_t len;
2849 const char *exclude_per_dir;
2850 const unsigned hashsz = the_hash_algo->rawsz;
2851 const unsigned offset = sizeof(struct ondisk_untracked_cache);
2852 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
2853
2854 if (sz <= 1 || end[-1] != '\0')
2855 return NULL;
2856 end--;
2857
2858 ident_len = decode_varint(&next);
2859 if (next + ident_len > end)
2860 return NULL;
2861 ident = (const char *)next;
2862 next += ident_len;
2863
2864 if (next + exclude_per_dir_offset + 1 > end)
2865 return NULL;
2866
2867 uc = xcalloc(1, sizeof(*uc));
2868 strbuf_init(&uc->ident, ident_len);
2869 strbuf_add(&uc->ident, ident, ident_len);
2870 load_oid_stat(&uc->ss_info_exclude,
2871 next + ouc_offset(info_exclude_stat),
2872 next + offset);
2873 load_oid_stat(&uc->ss_excludes_file,
2874 next + ouc_offset(excludes_file_stat),
2875 next + offset + hashsz);
2876 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2877 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
2878 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2879 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2880 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
2881 if (next >= end)
2882 goto done2;
2883
2884 len = decode_varint(&next);
2885 if (next > end || len == 0)
2886 goto done2;
2887
2888 rd.valid = ewah_new();
2889 rd.check_only = ewah_new();
2890 rd.sha1_valid = ewah_new();
2891 rd.data = next;
2892 rd.end = end;
2893 rd.index = 0;
2894 ALLOC_ARRAY(rd.ucd, len);
2895
2896 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2897 goto done;
2898
2899 next = rd.data;
2900 len = ewah_read_mmap(rd.valid, next, end - next);
2901 if (len < 0)
2902 goto done;
2903
2904 next += len;
2905 len = ewah_read_mmap(rd.check_only, next, end - next);
2906 if (len < 0)
2907 goto done;
2908
2909 next += len;
2910 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2911 if (len < 0)
2912 goto done;
2913
2914 ewah_each_bit(rd.check_only, set_check_only, &rd);
2915 rd.data = next + len;
2916 ewah_each_bit(rd.valid, read_stat, &rd);
2917 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
2918 next = rd.data;
2919
2920 done:
2921 free(rd.ucd);
2922 ewah_free(rd.valid);
2923 ewah_free(rd.check_only);
2924 ewah_free(rd.sha1_valid);
2925 done2:
2926 if (next != end) {
2927 free_untracked_cache(uc);
2928 uc = NULL;
2929 }
2930 return uc;
2931 }
2932
2933 static void invalidate_one_directory(struct untracked_cache *uc,
2934 struct untracked_cache_dir *ucd)
2935 {
2936 uc->dir_invalidated++;
2937 ucd->valid = 0;
2938 ucd->untracked_nr = 0;
2939 }
2940
2941 /*
2942 * Normally when an entry is added or removed from a directory,
2943 * invalidating that directory is enough. No need to touch its
2944 * ancestors. When a directory is shown as "foo/bar/" in git-status
2945 * however, deleting or adding an entry may have cascading effect.
2946 *
2947 * Say the "foo/bar/file" has become untracked, we need to tell the
2948 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2949 * directory any more (because "bar" is managed by foo as an untracked
2950 * "file").
2951 *
2952 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2953 * was the last untracked entry in the entire "foo", we should show
2954 * "foo/" instead. Which means we have to invalidate past "bar" up to
2955 * "foo".
2956 *
2957 * This function traverses all directories from root to leaf. If there
2958 * is a chance of one of the above cases happening, we invalidate back
2959 * to root. Otherwise we just invalidate the leaf. There may be a more
2960 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2961 * detect these cases and avoid unnecessary invalidation, for example,
2962 * checking for the untracked entry named "bar/" in "foo", but for now
2963 * stick to something safe and simple.
2964 */
2965 static int invalidate_one_component(struct untracked_cache *uc,
2966 struct untracked_cache_dir *dir,
2967 const char *path, int len)
2968 {
2969 const char *rest = strchr(path, '/');
2970
2971 if (rest) {
2972 int component_len = rest - path;
2973 struct untracked_cache_dir *d =
2974 lookup_untracked(uc, dir, path, component_len);
2975 int ret =
2976 invalidate_one_component(uc, d, rest + 1,
2977 len - (component_len + 1));
2978 if (ret)
2979 invalidate_one_directory(uc, dir);
2980 return ret;
2981 }
2982
2983 invalidate_one_directory(uc, dir);
2984 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2985 }
2986
2987 void untracked_cache_invalidate_path(struct index_state *istate,
2988 const char *path, int safe_path)
2989 {
2990 if (!istate->untracked || !istate->untracked->root)
2991 return;
2992 if (!safe_path && !verify_path(path, 0))
2993 return;
2994 invalidate_one_component(istate->untracked, istate->untracked->root,
2995 path, strlen(path));
2996 }
2997
2998 void untracked_cache_remove_from_index(struct index_state *istate,
2999 const char *path)
3000 {
3001 untracked_cache_invalidate_path(istate, path, 1);
3002 }
3003
3004 void untracked_cache_add_to_index(struct index_state *istate,
3005 const char *path)
3006 {
3007 untracked_cache_invalidate_path(istate, path, 1);
3008 }
3009
3010 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3011 const char *sub_gitdir)
3012 {
3013 int i;
3014 struct repository subrepo;
3015 struct strbuf sub_wt = STRBUF_INIT;
3016 struct strbuf sub_gd = STRBUF_INIT;
3017
3018 const struct submodule *sub;
3019
3020 /* If the submodule has no working tree, we can ignore it. */
3021 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3022 return;
3023
3024 if (repo_read_index(&subrepo) < 0)
3025 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3026
3027 for (i = 0; i < subrepo.index->cache_nr; i++) {
3028 const struct cache_entry *ce = subrepo.index->cache[i];
3029
3030 if (!S_ISGITLINK(ce->ce_mode))
3031 continue;
3032
3033 while (i + 1 < subrepo.index->cache_nr &&
3034 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3035 /*
3036 * Skip entries with the same name in different stages
3037 * to make sure an entry is returned only once.
3038 */
3039 i++;
3040
3041 sub = submodule_from_path(&subrepo, &null_oid, ce->name);
3042 if (!sub || !is_submodule_active(&subrepo, ce->name))
3043 /* .gitmodules broken or inactive sub */
3044 continue;
3045
3046 strbuf_reset(&sub_wt);
3047 strbuf_reset(&sub_gd);
3048 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3049 strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name);
3050
3051 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3052 }
3053 strbuf_release(&sub_wt);
3054 strbuf_release(&sub_gd);
3055 repo_clear(&subrepo);
3056 }
3057
3058 void connect_work_tree_and_git_dir(const char *work_tree_,
3059 const char *git_dir_,
3060 int recurse_into_nested)
3061 {
3062 struct strbuf gitfile_sb = STRBUF_INIT;
3063 struct strbuf cfg_sb = STRBUF_INIT;
3064 struct strbuf rel_path = STRBUF_INIT;
3065 char *git_dir, *work_tree;
3066
3067 /* Prepare .git file */
3068 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3069 if (safe_create_leading_directories_const(gitfile_sb.buf))
3070 die(_("could not create directories for %s"), gitfile_sb.buf);
3071
3072 /* Prepare config file */
3073 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3074 if (safe_create_leading_directories_const(cfg_sb.buf))
3075 die(_("could not create directories for %s"), cfg_sb.buf);
3076
3077 git_dir = real_pathdup(git_dir_, 1);
3078 work_tree = real_pathdup(work_tree_, 1);
3079
3080 /* Write .git file */
3081 write_file(gitfile_sb.buf, "gitdir: %s",
3082 relative_path(git_dir, work_tree, &rel_path));
3083 /* Update core.worktree setting */
3084 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3085 relative_path(work_tree, git_dir, &rel_path));
3086
3087 strbuf_release(&gitfile_sb);
3088 strbuf_release(&cfg_sb);
3089 strbuf_release(&rel_path);
3090
3091 if (recurse_into_nested)
3092 connect_wt_gitdir_in_nested(work_tree, git_dir);
3093
3094 free(work_tree);
3095 free(git_dir);
3096 }
3097
3098 /*
3099 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3100 */
3101 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3102 {
3103 if (rename(old_git_dir, new_git_dir) < 0)
3104 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3105 old_git_dir, new_git_dir);
3106
3107 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3108 }