]> git.ipfire.org Git - thirdparty/git.git/blob - dir.c
dir: fix checks on common prefix directory
[thirdparty/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 xyz/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_LEADING_PATHSPEC (1<<2)
277
278 /*
279 * Does the given pathspec match the given name? A match is found if
280 *
281 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
282 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
283 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
284 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
285 *
286 * Return value tells which case it was (1-4), or 0 when there is no match.
287 *
288 * It may be instructive to look at a small table of concrete examples
289 * to understand the differences between 1, 2, and 4:
290 *
291 * Pathspecs
292 * | a/b | a/b/ | a/b/c
293 * ------+-----------+-----------+------------
294 * a/b | EXACT | EXACT[1] | LEADING[2]
295 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
296 * a/b/c | RECURSIVE | RECURSIVE | EXACT
297 *
298 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
299 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
300 */
301 static int match_pathspec_item(const struct index_state *istate,
302 const struct pathspec_item *item, int prefix,
303 const char *name, int namelen, unsigned flags)
304 {
305 /* name/namelen has prefix cut off by caller */
306 const char *match = item->match + prefix;
307 int matchlen = item->len - prefix;
308
309 /*
310 * The normal call pattern is:
311 * 1. prefix = common_prefix_len(ps);
312 * 2. prune something, or fill_directory
313 * 3. match_pathspec()
314 *
315 * 'prefix' at #1 may be shorter than the command's prefix and
316 * it's ok for #2 to match extra files. Those extras will be
317 * trimmed at #3.
318 *
319 * Suppose the pathspec is 'foo' and '../bar' running from
320 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
321 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
322 * user does not want XYZ/foo, only the "foo" part should be
323 * case-insensitive. We need to filter out XYZ/foo here. In
324 * other words, we do not trust the caller on comparing the
325 * prefix part when :(icase) is involved. We do exact
326 * comparison ourselves.
327 *
328 * Normally the caller (common_prefix_len() in fact) does
329 * _exact_ matching on name[-prefix+1..-1] and we do not need
330 * to check that part. Be defensive and check it anyway, in
331 * case common_prefix_len is changed, or a new caller is
332 * introduced that does not use common_prefix_len.
333 *
334 * If the penalty turns out too high when prefix is really
335 * long, maybe change it to
336 * strncmp(match, name, item->prefix - prefix)
337 */
338 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
339 strncmp(item->match, name - prefix, item->prefix))
340 return 0;
341
342 if (item->attr_match_nr &&
343 !match_pathspec_attrs(istate, name, namelen, item))
344 return 0;
345
346 /* If the match was just the prefix, we matched */
347 if (!*match)
348 return MATCHED_RECURSIVELY;
349
350 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
351 if (matchlen == namelen)
352 return MATCHED_EXACTLY;
353
354 if (match[matchlen-1] == '/' || name[matchlen] == '/')
355 return MATCHED_RECURSIVELY;
356 } else if ((flags & DO_MATCH_DIRECTORY) &&
357 match[matchlen - 1] == '/' &&
358 namelen == matchlen - 1 &&
359 !ps_strncmp(item, match, name, namelen))
360 return MATCHED_EXACTLY;
361
362 if (item->nowildcard_len < item->len &&
363 !git_fnmatch(item, match, name,
364 item->nowildcard_len - prefix))
365 return MATCHED_FNMATCH;
366
367 /* Perform checks to see if "name" is a leading string of the pathspec */
368 if (flags & DO_MATCH_LEADING_PATHSPEC) {
369 /* name is a literal prefix of the pathspec */
370 int offset = name[namelen-1] == '/' ? 1 : 0;
371 if ((namelen < matchlen) &&
372 (match[namelen-offset] == '/') &&
373 !ps_strncmp(item, match, name, namelen))
374 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
375
376 /* name doesn't match up to the first wild character */
377 if (item->nowildcard_len < item->len &&
378 ps_strncmp(item, match, name,
379 item->nowildcard_len - prefix))
380 return 0;
381
382 /*
383 * name has no wildcard, and it didn't match as a leading
384 * pathspec so return.
385 */
386 if (item->nowildcard_len == item->len)
387 return 0;
388
389 /*
390 * Here is where we would perform a wildmatch to check if
391 * "name" can be matched as a directory (or a prefix) against
392 * the pathspec. Since wildmatch doesn't have this capability
393 * at the present we have to punt and say that it is a match,
394 * potentially returning a false positive
395 * The submodules themselves will be able to perform more
396 * accurate matching to determine if the pathspec matches.
397 */
398 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
399 }
400
401 return 0;
402 }
403
404 /*
405 * Given a name and a list of pathspecs, returns the nature of the
406 * closest (i.e. most specific) match of the name to any of the
407 * pathspecs.
408 *
409 * The caller typically calls this multiple times with the same
410 * pathspec and seen[] array but with different name/namelen
411 * (e.g. entries from the index) and is interested in seeing if and
412 * how each pathspec matches all the names it calls this function
413 * with. A mark is left in the seen[] array for each pathspec element
414 * indicating the closest type of match that element achieved, so if
415 * seen[n] remains zero after multiple invocations, that means the nth
416 * pathspec did not match any names, which could indicate that the
417 * user mistyped the nth pathspec.
418 */
419 static int do_match_pathspec(const struct index_state *istate,
420 const struct pathspec *ps,
421 const char *name, int namelen,
422 int prefix, char *seen,
423 unsigned flags)
424 {
425 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
426
427 GUARD_PATHSPEC(ps,
428 PATHSPEC_FROMTOP |
429 PATHSPEC_MAXDEPTH |
430 PATHSPEC_LITERAL |
431 PATHSPEC_GLOB |
432 PATHSPEC_ICASE |
433 PATHSPEC_EXCLUDE |
434 PATHSPEC_ATTR);
435
436 if (!ps->nr) {
437 if (!ps->recursive ||
438 !(ps->magic & PATHSPEC_MAXDEPTH) ||
439 ps->max_depth == -1)
440 return MATCHED_RECURSIVELY;
441
442 if (within_depth(name, namelen, 0, ps->max_depth))
443 return MATCHED_EXACTLY;
444 else
445 return 0;
446 }
447
448 name += prefix;
449 namelen -= prefix;
450
451 for (i = ps->nr - 1; i >= 0; i--) {
452 int how;
453
454 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
455 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
456 continue;
457
458 if (seen && seen[i] == MATCHED_EXACTLY)
459 continue;
460 /*
461 * Make exclude patterns optional and never report
462 * "pathspec ':(exclude)foo' matches no files"
463 */
464 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
465 seen[i] = MATCHED_FNMATCH;
466 how = match_pathspec_item(istate, ps->items+i, prefix, name,
467 namelen, flags);
468 if (ps->recursive &&
469 (ps->magic & PATHSPEC_MAXDEPTH) &&
470 ps->max_depth != -1 &&
471 how && how != MATCHED_FNMATCH) {
472 int len = ps->items[i].len;
473 if (name[len] == '/')
474 len++;
475 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
476 how = MATCHED_EXACTLY;
477 else
478 how = 0;
479 }
480 if (how) {
481 if (retval < how)
482 retval = how;
483 if (seen && seen[i] < how)
484 seen[i] = how;
485 }
486 }
487 return retval;
488 }
489
490 int match_pathspec(const struct index_state *istate,
491 const struct pathspec *ps,
492 const char *name, int namelen,
493 int prefix, char *seen, int is_dir)
494 {
495 int positive, negative;
496 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
497 positive = do_match_pathspec(istate, ps, name, namelen,
498 prefix, seen, flags);
499 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
500 return positive;
501 negative = do_match_pathspec(istate, ps, name, namelen,
502 prefix, seen,
503 flags | DO_MATCH_EXCLUDE);
504 return negative ? 0 : positive;
505 }
506
507 /**
508 * Check if a submodule is a superset of the pathspec
509 */
510 int submodule_path_match(const struct index_state *istate,
511 const struct pathspec *ps,
512 const char *submodule_name,
513 char *seen)
514 {
515 int matched = do_match_pathspec(istate, ps, submodule_name,
516 strlen(submodule_name),
517 0, seen,
518 DO_MATCH_DIRECTORY |
519 DO_MATCH_LEADING_PATHSPEC);
520 return matched;
521 }
522
523 int report_path_error(const char *ps_matched,
524 const struct pathspec *pathspec)
525 {
526 /*
527 * Make sure all pathspec matched; otherwise it is an error.
528 */
529 int num, errors = 0;
530 for (num = 0; num < pathspec->nr; num++) {
531 int other, found_dup;
532
533 if (ps_matched[num])
534 continue;
535 /*
536 * The caller might have fed identical pathspec
537 * twice. Do not barf on such a mistake.
538 * FIXME: parse_pathspec should have eliminated
539 * duplicate pathspec.
540 */
541 for (found_dup = other = 0;
542 !found_dup && other < pathspec->nr;
543 other++) {
544 if (other == num || !ps_matched[other])
545 continue;
546 if (!strcmp(pathspec->items[other].original,
547 pathspec->items[num].original))
548 /*
549 * Ok, we have a match already.
550 */
551 found_dup = 1;
552 }
553 if (found_dup)
554 continue;
555
556 error(_("pathspec '%s' did not match any file(s) known to git"),
557 pathspec->items[num].original);
558 errors++;
559 }
560 return errors;
561 }
562
563 /*
564 * Return the length of the "simple" part of a path match limiter.
565 */
566 int simple_length(const char *match)
567 {
568 int len = -1;
569
570 for (;;) {
571 unsigned char c = *match++;
572 len++;
573 if (c == '\0' || is_glob_special(c))
574 return len;
575 }
576 }
577
578 int no_wildcard(const char *string)
579 {
580 return string[simple_length(string)] == '\0';
581 }
582
583 void parse_path_pattern(const char **pattern,
584 int *patternlen,
585 unsigned *flags,
586 int *nowildcardlen)
587 {
588 const char *p = *pattern;
589 size_t i, len;
590
591 *flags = 0;
592 if (*p == '!') {
593 *flags |= PATTERN_FLAG_NEGATIVE;
594 p++;
595 }
596 len = strlen(p);
597 if (len && p[len - 1] == '/') {
598 len--;
599 *flags |= PATTERN_FLAG_MUSTBEDIR;
600 }
601 for (i = 0; i < len; i++) {
602 if (p[i] == '/')
603 break;
604 }
605 if (i == len)
606 *flags |= PATTERN_FLAG_NODIR;
607 *nowildcardlen = simple_length(p);
608 /*
609 * we should have excluded the trailing slash from 'p' too,
610 * but that's one more allocation. Instead just make sure
611 * nowildcardlen does not exceed real patternlen
612 */
613 if (*nowildcardlen > len)
614 *nowildcardlen = len;
615 if (*p == '*' && no_wildcard(p + 1))
616 *flags |= PATTERN_FLAG_ENDSWITH;
617 *pattern = p;
618 *patternlen = len;
619 }
620
621 void add_pattern(const char *string, const char *base,
622 int baselen, struct pattern_list *pl, int srcpos)
623 {
624 struct path_pattern *pattern;
625 int patternlen;
626 unsigned flags;
627 int nowildcardlen;
628
629 parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen);
630 if (flags & PATTERN_FLAG_MUSTBEDIR) {
631 FLEXPTR_ALLOC_MEM(pattern, pattern, string, patternlen);
632 } else {
633 pattern = xmalloc(sizeof(*pattern));
634 pattern->pattern = string;
635 }
636 pattern->patternlen = patternlen;
637 pattern->nowildcardlen = nowildcardlen;
638 pattern->base = base;
639 pattern->baselen = baselen;
640 pattern->flags = flags;
641 pattern->srcpos = srcpos;
642 ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc);
643 pl->patterns[pl->nr++] = pattern;
644 pattern->pl = pl;
645 }
646
647 static int read_skip_worktree_file_from_index(const struct index_state *istate,
648 const char *path,
649 size_t *size_out, char **data_out,
650 struct oid_stat *oid_stat)
651 {
652 int pos, len;
653
654 len = strlen(path);
655 pos = index_name_pos(istate, path, len);
656 if (pos < 0)
657 return -1;
658 if (!ce_skip_worktree(istate->cache[pos]))
659 return -1;
660
661 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
662 }
663
664 /*
665 * Frees memory within pl which was allocated for exclude patterns and
666 * the file buffer. Does not free pl itself.
667 */
668 void clear_pattern_list(struct pattern_list *pl)
669 {
670 int i;
671
672 for (i = 0; i < pl->nr; i++)
673 free(pl->patterns[i]);
674 free(pl->patterns);
675 free(pl->filebuf);
676
677 memset(pl, 0, sizeof(*pl));
678 }
679
680 static void trim_trailing_spaces(char *buf)
681 {
682 char *p, *last_space = NULL;
683
684 for (p = buf; *p; p++)
685 switch (*p) {
686 case ' ':
687 if (!last_space)
688 last_space = p;
689 break;
690 case '\\':
691 p++;
692 if (!*p)
693 return;
694 /* fallthrough */
695 default:
696 last_space = NULL;
697 }
698
699 if (last_space)
700 *last_space = '\0';
701 }
702
703 /*
704 * Given a subdirectory name and "dir" of the current directory,
705 * search the subdir in "dir" and return it, or create a new one if it
706 * does not exist in "dir".
707 *
708 * If "name" has the trailing slash, it'll be excluded in the search.
709 */
710 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
711 struct untracked_cache_dir *dir,
712 const char *name, int len)
713 {
714 int first, last;
715 struct untracked_cache_dir *d;
716 if (!dir)
717 return NULL;
718 if (len && name[len - 1] == '/')
719 len--;
720 first = 0;
721 last = dir->dirs_nr;
722 while (last > first) {
723 int cmp, next = first + ((last - first) >> 1);
724 d = dir->dirs[next];
725 cmp = strncmp(name, d->name, len);
726 if (!cmp && strlen(d->name) > len)
727 cmp = -1;
728 if (!cmp)
729 return d;
730 if (cmp < 0) {
731 last = next;
732 continue;
733 }
734 first = next+1;
735 }
736
737 uc->dir_created++;
738 FLEX_ALLOC_MEM(d, name, name, len);
739
740 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
741 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
742 dir->dirs_nr - first);
743 dir->dirs_nr++;
744 dir->dirs[first] = d;
745 return d;
746 }
747
748 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
749 {
750 int i;
751 dir->valid = 0;
752 dir->untracked_nr = 0;
753 for (i = 0; i < dir->dirs_nr; i++)
754 do_invalidate_gitignore(dir->dirs[i]);
755 }
756
757 static void invalidate_gitignore(struct untracked_cache *uc,
758 struct untracked_cache_dir *dir)
759 {
760 uc->gitignore_invalidated++;
761 do_invalidate_gitignore(dir);
762 }
763
764 static void invalidate_directory(struct untracked_cache *uc,
765 struct untracked_cache_dir *dir)
766 {
767 int i;
768
769 /*
770 * Invalidation increment here is just roughly correct. If
771 * untracked_nr or any of dirs[].recurse is non-zero, we
772 * should increment dir_invalidated too. But that's more
773 * expensive to do.
774 */
775 if (dir->valid)
776 uc->dir_invalidated++;
777
778 dir->valid = 0;
779 dir->untracked_nr = 0;
780 for (i = 0; i < dir->dirs_nr; i++)
781 dir->dirs[i]->recurse = 0;
782 }
783
784 static int add_patterns_from_buffer(char *buf, size_t size,
785 const char *base, int baselen,
786 struct pattern_list *pl);
787
788 /*
789 * Given a file with name "fname", read it (either from disk, or from
790 * an index if 'istate' is non-null), parse it and store the
791 * exclude rules in "pl".
792 *
793 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
794 * stat data from disk (only valid if add_patterns returns zero). If
795 * ss_valid is non-zero, "ss" must contain good value as input.
796 */
797 static int add_patterns(const char *fname, const char *base, int baselen,
798 struct pattern_list *pl, struct index_state *istate,
799 struct oid_stat *oid_stat)
800 {
801 struct stat st;
802 int r;
803 int fd;
804 size_t size = 0;
805 char *buf;
806
807 fd = open(fname, O_RDONLY);
808 if (fd < 0 || fstat(fd, &st) < 0) {
809 if (fd < 0)
810 warn_on_fopen_errors(fname);
811 else
812 close(fd);
813 if (!istate)
814 return -1;
815 r = read_skip_worktree_file_from_index(istate, fname,
816 &size, &buf,
817 oid_stat);
818 if (r != 1)
819 return r;
820 } else {
821 size = xsize_t(st.st_size);
822 if (size == 0) {
823 if (oid_stat) {
824 fill_stat_data(&oid_stat->stat, &st);
825 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
826 oid_stat->valid = 1;
827 }
828 close(fd);
829 return 0;
830 }
831 buf = xmallocz(size);
832 if (read_in_full(fd, buf, size) != size) {
833 free(buf);
834 close(fd);
835 return -1;
836 }
837 buf[size++] = '\n';
838 close(fd);
839 if (oid_stat) {
840 int pos;
841 if (oid_stat->valid &&
842 !match_stat_data_racy(istate, &oid_stat->stat, &st))
843 ; /* no content change, ss->sha1 still good */
844 else if (istate &&
845 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
846 !ce_stage(istate->cache[pos]) &&
847 ce_uptodate(istate->cache[pos]) &&
848 !would_convert_to_git(istate, fname))
849 oidcpy(&oid_stat->oid,
850 &istate->cache[pos]->oid);
851 else
852 hash_object_file(buf, size, "blob",
853 &oid_stat->oid);
854 fill_stat_data(&oid_stat->stat, &st);
855 oid_stat->valid = 1;
856 }
857 }
858
859 add_patterns_from_buffer(buf, size, base, baselen, pl);
860 return 0;
861 }
862
863 static int add_patterns_from_buffer(char *buf, size_t size,
864 const char *base, int baselen,
865 struct pattern_list *pl)
866 {
867 int i, lineno = 1;
868 char *entry;
869
870 pl->filebuf = buf;
871
872 if (skip_utf8_bom(&buf, size))
873 size -= buf - pl->filebuf;
874
875 entry = buf;
876
877 for (i = 0; i < size; i++) {
878 if (buf[i] == '\n') {
879 if (entry != buf + i && entry[0] != '#') {
880 buf[i - (i && buf[i-1] == '\r')] = 0;
881 trim_trailing_spaces(entry);
882 add_pattern(entry, base, baselen, pl, lineno);
883 }
884 lineno++;
885 entry = buf + i + 1;
886 }
887 }
888 return 0;
889 }
890
891 int add_patterns_from_file_to_list(const char *fname, const char *base,
892 int baselen, struct pattern_list *pl,
893 struct index_state *istate)
894 {
895 return add_patterns(fname, base, baselen, pl, istate, NULL);
896 }
897
898 int add_patterns_from_blob_to_list(
899 struct object_id *oid,
900 const char *base, int baselen,
901 struct pattern_list *pl)
902 {
903 char *buf;
904 size_t size;
905 int r;
906
907 r = do_read_blob(oid, NULL, &size, &buf);
908 if (r != 1)
909 return r;
910
911 add_patterns_from_buffer(buf, size, base, baselen, pl);
912 return 0;
913 }
914
915 struct pattern_list *add_pattern_list(struct dir_struct *dir,
916 int group_type, const char *src)
917 {
918 struct pattern_list *pl;
919 struct exclude_list_group *group;
920
921 group = &dir->exclude_list_group[group_type];
922 ALLOC_GROW(group->pl, group->nr + 1, group->alloc);
923 pl = &group->pl[group->nr++];
924 memset(pl, 0, sizeof(*pl));
925 pl->src = src;
926 return pl;
927 }
928
929 /*
930 * Used to set up core.excludesfile and .git/info/exclude lists.
931 */
932 static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname,
933 struct oid_stat *oid_stat)
934 {
935 struct pattern_list *pl;
936 /*
937 * catch setup_standard_excludes() that's called before
938 * dir->untracked is assigned. That function behaves
939 * differently when dir->untracked is non-NULL.
940 */
941 if (!dir->untracked)
942 dir->unmanaged_exclude_files++;
943 pl = add_pattern_list(dir, EXC_FILE, fname);
944 if (add_patterns(fname, "", 0, pl, NULL, oid_stat) < 0)
945 die(_("cannot use %s as an exclude file"), fname);
946 }
947
948 void add_patterns_from_file(struct dir_struct *dir, const char *fname)
949 {
950 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
951 add_patterns_from_file_1(dir, fname, NULL);
952 }
953
954 int match_basename(const char *basename, int basenamelen,
955 const char *pattern, int prefix, int patternlen,
956 unsigned flags)
957 {
958 if (prefix == patternlen) {
959 if (patternlen == basenamelen &&
960 !fspathncmp(pattern, basename, basenamelen))
961 return 1;
962 } else if (flags & PATTERN_FLAG_ENDSWITH) {
963 /* "*literal" matching against "fooliteral" */
964 if (patternlen - 1 <= basenamelen &&
965 !fspathncmp(pattern + 1,
966 basename + basenamelen - (patternlen - 1),
967 patternlen - 1))
968 return 1;
969 } else {
970 if (fnmatch_icase_mem(pattern, patternlen,
971 basename, basenamelen,
972 0) == 0)
973 return 1;
974 }
975 return 0;
976 }
977
978 int match_pathname(const char *pathname, int pathlen,
979 const char *base, int baselen,
980 const char *pattern, int prefix, int patternlen,
981 unsigned flags)
982 {
983 const char *name;
984 int namelen;
985
986 /*
987 * match with FNM_PATHNAME; the pattern has base implicitly
988 * in front of it.
989 */
990 if (*pattern == '/') {
991 pattern++;
992 patternlen--;
993 prefix--;
994 }
995
996 /*
997 * baselen does not count the trailing slash. base[] may or
998 * may not end with a trailing slash though.
999 */
1000 if (pathlen < baselen + 1 ||
1001 (baselen && pathname[baselen] != '/') ||
1002 fspathncmp(pathname, base, baselen))
1003 return 0;
1004
1005 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1006 name = pathname + pathlen - namelen;
1007
1008 if (prefix) {
1009 /*
1010 * if the non-wildcard part is longer than the
1011 * remaining pathname, surely it cannot match.
1012 */
1013 if (prefix > namelen)
1014 return 0;
1015
1016 if (fspathncmp(pattern, name, prefix))
1017 return 0;
1018 pattern += prefix;
1019 patternlen -= prefix;
1020 name += prefix;
1021 namelen -= prefix;
1022
1023 /*
1024 * If the whole pattern did not have a wildcard,
1025 * then our prefix match is all we need; we
1026 * do not need to call fnmatch at all.
1027 */
1028 if (!patternlen && !namelen)
1029 return 1;
1030 }
1031
1032 return fnmatch_icase_mem(pattern, patternlen,
1033 name, namelen,
1034 WM_PATHNAME) == 0;
1035 }
1036
1037 /*
1038 * Scan the given exclude list in reverse to see whether pathname
1039 * should be ignored. The first match (i.e. the last on the list), if
1040 * any, determines the fate. Returns the exclude_list element which
1041 * matched, or NULL for undecided.
1042 */
1043 static struct path_pattern *last_matching_pattern_from_list(const char *pathname,
1044 int pathlen,
1045 const char *basename,
1046 int *dtype,
1047 struct pattern_list *pl,
1048 struct index_state *istate)
1049 {
1050 struct path_pattern *res = NULL; /* undecided */
1051 int i;
1052
1053 if (!pl->nr)
1054 return NULL; /* undefined */
1055
1056 for (i = pl->nr - 1; 0 <= i; i--) {
1057 struct path_pattern *pattern = pl->patterns[i];
1058 const char *exclude = pattern->pattern;
1059 int prefix = pattern->nowildcardlen;
1060
1061 if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) {
1062 if (*dtype == DT_UNKNOWN)
1063 *dtype = get_dtype(NULL, istate, pathname, pathlen);
1064 if (*dtype != DT_DIR)
1065 continue;
1066 }
1067
1068 if (pattern->flags & PATTERN_FLAG_NODIR) {
1069 if (match_basename(basename,
1070 pathlen - (basename - pathname),
1071 exclude, prefix, pattern->patternlen,
1072 pattern->flags)) {
1073 res = pattern;
1074 break;
1075 }
1076 continue;
1077 }
1078
1079 assert(pattern->baselen == 0 ||
1080 pattern->base[pattern->baselen - 1] == '/');
1081 if (match_pathname(pathname, pathlen,
1082 pattern->base,
1083 pattern->baselen ? pattern->baselen - 1 : 0,
1084 exclude, prefix, pattern->patternlen,
1085 pattern->flags)) {
1086 res = pattern;
1087 break;
1088 }
1089 }
1090 return res;
1091 }
1092
1093 /*
1094 * Scan the list of patterns to determine if the ordered list
1095 * of patterns matches on 'pathname'.
1096 *
1097 * Return 1 for a match, 0 for not matched and -1 for undecided.
1098 */
1099 enum pattern_match_result path_matches_pattern_list(
1100 const char *pathname, int pathlen,
1101 const char *basename, int *dtype,
1102 struct pattern_list *pl,
1103 struct index_state *istate)
1104 {
1105 struct path_pattern *pattern;
1106 pattern = last_matching_pattern_from_list(pathname, pathlen, basename,
1107 dtype, pl, istate);
1108 if (pattern) {
1109 if (pattern->flags & PATTERN_FLAG_NEGATIVE)
1110 return NOT_MATCHED;
1111 else
1112 return MATCHED;
1113 }
1114
1115 return UNDECIDED;
1116 }
1117
1118 static struct path_pattern *last_matching_pattern_from_lists(
1119 struct dir_struct *dir, struct index_state *istate,
1120 const char *pathname, int pathlen,
1121 const char *basename, int *dtype_p)
1122 {
1123 int i, j;
1124 struct exclude_list_group *group;
1125 struct path_pattern *pattern;
1126 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1127 group = &dir->exclude_list_group[i];
1128 for (j = group->nr - 1; j >= 0; j--) {
1129 pattern = last_matching_pattern_from_list(
1130 pathname, pathlen, basename, dtype_p,
1131 &group->pl[j], istate);
1132 if (pattern)
1133 return pattern;
1134 }
1135 }
1136 return NULL;
1137 }
1138
1139 /*
1140 * Loads the per-directory exclude list for the substring of base
1141 * which has a char length of baselen.
1142 */
1143 static void prep_exclude(struct dir_struct *dir,
1144 struct index_state *istate,
1145 const char *base, int baselen)
1146 {
1147 struct exclude_list_group *group;
1148 struct pattern_list *pl;
1149 struct exclude_stack *stk = NULL;
1150 struct untracked_cache_dir *untracked;
1151 int current;
1152
1153 group = &dir->exclude_list_group[EXC_DIRS];
1154
1155 /*
1156 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1157 * which originate from directories not in the prefix of the
1158 * path being checked.
1159 */
1160 while ((stk = dir->exclude_stack) != NULL) {
1161 if (stk->baselen <= baselen &&
1162 !strncmp(dir->basebuf.buf, base, stk->baselen))
1163 break;
1164 pl = &group->pl[dir->exclude_stack->exclude_ix];
1165 dir->exclude_stack = stk->prev;
1166 dir->pattern = NULL;
1167 free((char *)pl->src); /* see strbuf_detach() below */
1168 clear_pattern_list(pl);
1169 free(stk);
1170 group->nr--;
1171 }
1172
1173 /* Skip traversing into sub directories if the parent is excluded */
1174 if (dir->pattern)
1175 return;
1176
1177 /*
1178 * Lazy initialization. All call sites currently just
1179 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1180 * them seems lots of work for little benefit.
1181 */
1182 if (!dir->basebuf.buf)
1183 strbuf_init(&dir->basebuf, PATH_MAX);
1184
1185 /* Read from the parent directories and push them down. */
1186 current = stk ? stk->baselen : -1;
1187 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1188 if (dir->untracked)
1189 untracked = stk ? stk->ucd : dir->untracked->root;
1190 else
1191 untracked = NULL;
1192
1193 while (current < baselen) {
1194 const char *cp;
1195 struct oid_stat oid_stat;
1196
1197 stk = xcalloc(1, sizeof(*stk));
1198 if (current < 0) {
1199 cp = base;
1200 current = 0;
1201 } else {
1202 cp = strchr(base + current + 1, '/');
1203 if (!cp)
1204 die("oops in prep_exclude");
1205 cp++;
1206 untracked =
1207 lookup_untracked(dir->untracked, untracked,
1208 base + current,
1209 cp - base - current);
1210 }
1211 stk->prev = dir->exclude_stack;
1212 stk->baselen = cp - base;
1213 stk->exclude_ix = group->nr;
1214 stk->ucd = untracked;
1215 pl = add_pattern_list(dir, EXC_DIRS, NULL);
1216 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1217 assert(stk->baselen == dir->basebuf.len);
1218
1219 /* Abort if the directory is excluded */
1220 if (stk->baselen) {
1221 int dt = DT_DIR;
1222 dir->basebuf.buf[stk->baselen - 1] = 0;
1223 dir->pattern = last_matching_pattern_from_lists(dir,
1224 istate,
1225 dir->basebuf.buf, stk->baselen - 1,
1226 dir->basebuf.buf + current, &dt);
1227 dir->basebuf.buf[stk->baselen - 1] = '/';
1228 if (dir->pattern &&
1229 dir->pattern->flags & PATTERN_FLAG_NEGATIVE)
1230 dir->pattern = NULL;
1231 if (dir->pattern) {
1232 dir->exclude_stack = stk;
1233 return;
1234 }
1235 }
1236
1237 /* Try to read per-directory file */
1238 oidclr(&oid_stat.oid);
1239 oid_stat.valid = 0;
1240 if (dir->exclude_per_dir &&
1241 /*
1242 * If we know that no files have been added in
1243 * this directory (i.e. valid_cached_dir() has
1244 * been executed and set untracked->valid) ..
1245 */
1246 (!untracked || !untracked->valid ||
1247 /*
1248 * .. and .gitignore does not exist before
1249 * (i.e. null exclude_oid). Then we can skip
1250 * loading .gitignore, which would result in
1251 * ENOENT anyway.
1252 */
1253 !is_null_oid(&untracked->exclude_oid))) {
1254 /*
1255 * dir->basebuf gets reused by the traversal, but we
1256 * need fname to remain unchanged to ensure the src
1257 * member of each struct path_pattern correctly
1258 * back-references its source file. Other invocations
1259 * of add_pattern_list provide stable strings, so we
1260 * strbuf_detach() and free() here in the caller.
1261 */
1262 struct strbuf sb = STRBUF_INIT;
1263 strbuf_addbuf(&sb, &dir->basebuf);
1264 strbuf_addstr(&sb, dir->exclude_per_dir);
1265 pl->src = strbuf_detach(&sb, NULL);
1266 add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1267 untracked ? &oid_stat : NULL);
1268 }
1269 /*
1270 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1271 * will first be called in valid_cached_dir() then maybe many
1272 * times more in last_matching_pattern(). When the cache is
1273 * used, last_matching_pattern() will not be called and
1274 * reading .gitignore content will be a waste.
1275 *
1276 * So when it's called by valid_cached_dir() and we can get
1277 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1278 * modified on work tree), we could delay reading the
1279 * .gitignore content until we absolutely need it in
1280 * last_matching_pattern(). Be careful about ignore rule
1281 * order, though, if you do that.
1282 */
1283 if (untracked &&
1284 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1285 invalidate_gitignore(dir->untracked, untracked);
1286 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1287 }
1288 dir->exclude_stack = stk;
1289 current = stk->baselen;
1290 }
1291 strbuf_setlen(&dir->basebuf, baselen);
1292 }
1293
1294 /*
1295 * Loads the exclude lists for the directory containing pathname, then
1296 * scans all exclude lists to determine whether pathname is excluded.
1297 * Returns the exclude_list element which matched, or NULL for
1298 * undecided.
1299 */
1300 struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1301 struct index_state *istate,
1302 const char *pathname,
1303 int *dtype_p)
1304 {
1305 int pathlen = strlen(pathname);
1306 const char *basename = strrchr(pathname, '/');
1307 basename = (basename) ? basename+1 : pathname;
1308
1309 prep_exclude(dir, istate, pathname, basename-pathname);
1310
1311 if (dir->pattern)
1312 return dir->pattern;
1313
1314 return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1315 basename, dtype_p);
1316 }
1317
1318 /*
1319 * Loads the exclude lists for the directory containing pathname, then
1320 * scans all exclude lists to determine whether pathname is excluded.
1321 * Returns 1 if true, otherwise 0.
1322 */
1323 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1324 const char *pathname, int *dtype_p)
1325 {
1326 struct path_pattern *pattern =
1327 last_matching_pattern(dir, istate, pathname, dtype_p);
1328 if (pattern)
1329 return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1330 return 0;
1331 }
1332
1333 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1334 {
1335 struct dir_entry *ent;
1336
1337 FLEX_ALLOC_MEM(ent, name, pathname, len);
1338 ent->len = len;
1339 return ent;
1340 }
1341
1342 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1343 struct index_state *istate,
1344 const char *pathname, int len)
1345 {
1346 if (index_file_exists(istate, pathname, len, ignore_case))
1347 return NULL;
1348
1349 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1350 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1351 }
1352
1353 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1354 struct index_state *istate,
1355 const char *pathname, int len)
1356 {
1357 if (!index_name_is_other(istate, pathname, len))
1358 return NULL;
1359
1360 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1361 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1362 }
1363
1364 enum exist_status {
1365 index_nonexistent = 0,
1366 index_directory,
1367 index_gitdir
1368 };
1369
1370 /*
1371 * Do not use the alphabetically sorted index to look up
1372 * the directory name; instead, use the case insensitive
1373 * directory hash.
1374 */
1375 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1376 const char *dirname, int len)
1377 {
1378 struct cache_entry *ce;
1379
1380 if (index_dir_exists(istate, dirname, len))
1381 return index_directory;
1382
1383 ce = index_file_exists(istate, dirname, len, ignore_case);
1384 if (ce && S_ISGITLINK(ce->ce_mode))
1385 return index_gitdir;
1386
1387 return index_nonexistent;
1388 }
1389
1390 /*
1391 * The index sorts alphabetically by entry name, which
1392 * means that a gitlink sorts as '\0' at the end, while
1393 * a directory (which is defined not as an entry, but as
1394 * the files it contains) will sort with the '/' at the
1395 * end.
1396 */
1397 static enum exist_status directory_exists_in_index(struct index_state *istate,
1398 const char *dirname, int len)
1399 {
1400 int pos;
1401
1402 if (ignore_case)
1403 return directory_exists_in_index_icase(istate, dirname, len);
1404
1405 pos = index_name_pos(istate, dirname, len);
1406 if (pos < 0)
1407 pos = -pos-1;
1408 while (pos < istate->cache_nr) {
1409 const struct cache_entry *ce = istate->cache[pos++];
1410 unsigned char endchar;
1411
1412 if (strncmp(ce->name, dirname, len))
1413 break;
1414 endchar = ce->name[len];
1415 if (endchar > '/')
1416 break;
1417 if (endchar == '/')
1418 return index_directory;
1419 if (!endchar && S_ISGITLINK(ce->ce_mode))
1420 return index_gitdir;
1421 }
1422 return index_nonexistent;
1423 }
1424
1425 /*
1426 * When we find a directory when traversing the filesystem, we
1427 * have three distinct cases:
1428 *
1429 * - ignore it
1430 * - see it as a directory
1431 * - recurse into it
1432 *
1433 * and which one we choose depends on a combination of existing
1434 * git index contents and the flags passed into the directory
1435 * traversal routine.
1436 *
1437 * Case 1: If we *already* have entries in the index under that
1438 * directory name, we always recurse into the directory to see
1439 * all the files.
1440 *
1441 * Case 2: If we *already* have that directory name as a gitlink,
1442 * we always continue to see it as a gitlink, regardless of whether
1443 * there is an actual git directory there or not (it might not
1444 * be checked out as a subproject!)
1445 *
1446 * Case 3: if we didn't have it in the index previously, we
1447 * have a few sub-cases:
1448 *
1449 * (a) if "show_other_directories" is true, we show it as
1450 * just a directory, unless "hide_empty_directories" is
1451 * also true, in which case we need to check if it contains any
1452 * untracked and / or ignored files.
1453 * (b) if it looks like a git directory, and we don't have
1454 * 'no_gitlinks' set we treat it as a gitlink, and show it
1455 * as a directory.
1456 * (c) otherwise, we recurse into it.
1457 */
1458 static enum path_treatment treat_directory(struct dir_struct *dir,
1459 struct index_state *istate,
1460 struct untracked_cache_dir *untracked,
1461 const char *dirname, int len, int baselen, int exclude,
1462 const struct pathspec *pathspec)
1463 {
1464 /* The "len-1" is to strip the final '/' */
1465 switch (directory_exists_in_index(istate, dirname, len-1)) {
1466 case index_directory:
1467 return path_recurse;
1468
1469 case index_gitdir:
1470 return path_none;
1471
1472 case index_nonexistent:
1473 if (dir->flags & DIR_SKIP_NESTED_GIT) {
1474 int nested_repo;
1475 struct strbuf sb = STRBUF_INIT;
1476 strbuf_addstr(&sb, dirname);
1477 nested_repo = is_nonbare_repository_dir(&sb);
1478 strbuf_release(&sb);
1479 if (nested_repo)
1480 return path_none;
1481 }
1482
1483 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1484 break;
1485 if (exclude &&
1486 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1487 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1488
1489 /*
1490 * This is an excluded directory and we are
1491 * showing ignored paths that match an exclude
1492 * pattern. (e.g. show directory as ignored
1493 * only if it matches an exclude pattern).
1494 * This path will either be 'path_excluded`
1495 * (if we are showing empty directories or if
1496 * the directory is not empty), or will be
1497 * 'path_none' (empty directory, and we are
1498 * not showing empty directories).
1499 */
1500 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1501 return path_excluded;
1502
1503 if (read_directory_recursive(dir, istate, dirname, len,
1504 untracked, 1, 1, pathspec) == path_excluded)
1505 return path_excluded;
1506
1507 return path_none;
1508 }
1509 if (!(dir->flags & DIR_NO_GITLINKS)) {
1510 struct strbuf sb = STRBUF_INIT;
1511 strbuf_addstr(&sb, dirname);
1512 if (is_nonbare_repository_dir(&sb))
1513 return exclude ? path_excluded : path_untracked;
1514 strbuf_release(&sb);
1515 }
1516 return path_recurse;
1517 }
1518
1519 /* This is the "show_other_directories" case */
1520
1521 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1522 return exclude ? path_excluded : path_untracked;
1523
1524 untracked = lookup_untracked(dir->untracked, untracked,
1525 dirname + baselen, len - baselen);
1526
1527 /*
1528 * If this is an excluded directory, then we only need to check if
1529 * the directory contains any files.
1530 */
1531 return read_directory_recursive(dir, istate, dirname, len,
1532 untracked, 1, exclude, pathspec);
1533 }
1534
1535 /*
1536 * This is an inexact early pruning of any recursive directory
1537 * reading - if the path cannot possibly be in the pathspec,
1538 * return true, and we'll skip it early.
1539 */
1540 static int simplify_away(const char *path, int pathlen,
1541 const struct pathspec *pathspec)
1542 {
1543 int i;
1544
1545 if (!pathspec || !pathspec->nr)
1546 return 0;
1547
1548 GUARD_PATHSPEC(pathspec,
1549 PATHSPEC_FROMTOP |
1550 PATHSPEC_MAXDEPTH |
1551 PATHSPEC_LITERAL |
1552 PATHSPEC_GLOB |
1553 PATHSPEC_ICASE |
1554 PATHSPEC_EXCLUDE |
1555 PATHSPEC_ATTR);
1556
1557 for (i = 0; i < pathspec->nr; i++) {
1558 const struct pathspec_item *item = &pathspec->items[i];
1559 int len = item->nowildcard_len;
1560
1561 if (len > pathlen)
1562 len = pathlen;
1563 if (!ps_strncmp(item, item->match, path, len))
1564 return 0;
1565 }
1566
1567 return 1;
1568 }
1569
1570 /*
1571 * This function tells us whether an excluded path matches a
1572 * list of "interesting" pathspecs. That is, whether a path matched
1573 * by any of the pathspecs could possibly be ignored by excluding
1574 * the specified path. This can happen if:
1575 *
1576 * 1. the path is mentioned explicitly in the pathspec
1577 *
1578 * 2. the path is a directory prefix of some element in the
1579 * pathspec
1580 */
1581 static int exclude_matches_pathspec(const char *path, int pathlen,
1582 const struct pathspec *pathspec)
1583 {
1584 int i;
1585
1586 if (!pathspec || !pathspec->nr)
1587 return 0;
1588
1589 GUARD_PATHSPEC(pathspec,
1590 PATHSPEC_FROMTOP |
1591 PATHSPEC_MAXDEPTH |
1592 PATHSPEC_LITERAL |
1593 PATHSPEC_GLOB |
1594 PATHSPEC_ICASE |
1595 PATHSPEC_EXCLUDE);
1596
1597 for (i = 0; i < pathspec->nr; i++) {
1598 const struct pathspec_item *item = &pathspec->items[i];
1599 int len = item->nowildcard_len;
1600
1601 if (len == pathlen &&
1602 !ps_strncmp(item, item->match, path, pathlen))
1603 return 1;
1604 if (len > pathlen &&
1605 item->match[pathlen] == '/' &&
1606 !ps_strncmp(item, item->match, path, pathlen))
1607 return 1;
1608 }
1609 return 0;
1610 }
1611
1612 static int get_index_dtype(struct index_state *istate,
1613 const char *path, int len)
1614 {
1615 int pos;
1616 const struct cache_entry *ce;
1617
1618 ce = index_file_exists(istate, path, len, 0);
1619 if (ce) {
1620 if (!ce_uptodate(ce))
1621 return DT_UNKNOWN;
1622 if (S_ISGITLINK(ce->ce_mode))
1623 return DT_DIR;
1624 /*
1625 * Nobody actually cares about the
1626 * difference between DT_LNK and DT_REG
1627 */
1628 return DT_REG;
1629 }
1630
1631 /* Try to look it up as a directory */
1632 pos = index_name_pos(istate, path, len);
1633 if (pos >= 0)
1634 return DT_UNKNOWN;
1635 pos = -pos-1;
1636 while (pos < istate->cache_nr) {
1637 ce = istate->cache[pos++];
1638 if (strncmp(ce->name, path, len))
1639 break;
1640 if (ce->name[len] > '/')
1641 break;
1642 if (ce->name[len] < '/')
1643 continue;
1644 if (!ce_uptodate(ce))
1645 break; /* continue? */
1646 return DT_DIR;
1647 }
1648 return DT_UNKNOWN;
1649 }
1650
1651 static int get_dtype(struct dirent *de, struct index_state *istate,
1652 const char *path, int len)
1653 {
1654 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1655 struct stat st;
1656
1657 if (dtype != DT_UNKNOWN)
1658 return dtype;
1659 dtype = get_index_dtype(istate, path, len);
1660 if (dtype != DT_UNKNOWN)
1661 return dtype;
1662 if (lstat(path, &st))
1663 return dtype;
1664 if (S_ISREG(st.st_mode))
1665 return DT_REG;
1666 if (S_ISDIR(st.st_mode))
1667 return DT_DIR;
1668 if (S_ISLNK(st.st_mode))
1669 return DT_LNK;
1670 return dtype;
1671 }
1672
1673 static enum path_treatment treat_one_path(struct dir_struct *dir,
1674 struct untracked_cache_dir *untracked,
1675 struct index_state *istate,
1676 struct strbuf *path,
1677 int baselen,
1678 const struct pathspec *pathspec,
1679 int dtype, struct dirent *de)
1680 {
1681 int exclude;
1682 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1683 enum path_treatment path_treatment;
1684
1685 if (dtype == DT_UNKNOWN)
1686 dtype = get_dtype(de, istate, path->buf, path->len);
1687
1688 /* Always exclude indexed files */
1689 if (dtype != DT_DIR && has_path_in_index)
1690 return path_none;
1691
1692 /*
1693 * When we are looking at a directory P in the working tree,
1694 * there are three cases:
1695 *
1696 * (1) P exists in the index. Everything inside the directory P in
1697 * the working tree needs to go when P is checked out from the
1698 * index.
1699 *
1700 * (2) P does not exist in the index, but there is P/Q in the index.
1701 * We know P will stay a directory when we check out the contents
1702 * of the index, but we do not know yet if there is a directory
1703 * P/Q in the working tree to be killed, so we need to recurse.
1704 *
1705 * (3) P does not exist in the index, and there is no P/Q in the index
1706 * to require P to be a directory, either. Only in this case, we
1707 * know that everything inside P will not be killed without
1708 * recursing.
1709 */
1710 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1711 (dtype == DT_DIR) &&
1712 !has_path_in_index &&
1713 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1714 return path_none;
1715
1716 exclude = is_excluded(dir, istate, path->buf, &dtype);
1717
1718 /*
1719 * Excluded? If we don't explicitly want to show
1720 * ignored files, ignore it
1721 */
1722 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1723 return path_excluded;
1724
1725 switch (dtype) {
1726 default:
1727 return path_none;
1728 case DT_DIR:
1729 strbuf_addch(path, '/');
1730 path_treatment = treat_directory(dir, istate, untracked,
1731 path->buf, path->len,
1732 baselen, exclude, pathspec);
1733 /*
1734 * If 1) we only want to return directories that
1735 * match an exclude pattern and 2) this directory does
1736 * not match an exclude pattern but all of its
1737 * contents are excluded, then indicate that we should
1738 * recurse into this directory (instead of marking the
1739 * directory itself as an ignored path).
1740 */
1741 if (!exclude &&
1742 path_treatment == path_excluded &&
1743 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1744 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1745 return path_recurse;
1746 return path_treatment;
1747 case DT_REG:
1748 case DT_LNK:
1749 return exclude ? path_excluded : path_untracked;
1750 }
1751 }
1752
1753 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1754 struct untracked_cache_dir *untracked,
1755 struct cached_dir *cdir,
1756 struct index_state *istate,
1757 struct strbuf *path,
1758 int baselen,
1759 const struct pathspec *pathspec)
1760 {
1761 strbuf_setlen(path, baselen);
1762 if (!cdir->ucd) {
1763 strbuf_addstr(path, cdir->file);
1764 return path_untracked;
1765 }
1766 strbuf_addstr(path, cdir->ucd->name);
1767 /* treat_one_path() does this before it calls treat_directory() */
1768 strbuf_complete(path, '/');
1769 if (cdir->ucd->check_only)
1770 /*
1771 * check_only is set as a result of treat_directory() getting
1772 * to its bottom. Verify again the same set of directories
1773 * with check_only set.
1774 */
1775 return read_directory_recursive(dir, istate, path->buf, path->len,
1776 cdir->ucd, 1, 0, pathspec);
1777 /*
1778 * We get path_recurse in the first run when
1779 * directory_exists_in_index() returns index_nonexistent. We
1780 * are sure that new changes in the index does not impact the
1781 * outcome. Return now.
1782 */
1783 return path_recurse;
1784 }
1785
1786 static enum path_treatment treat_path(struct dir_struct *dir,
1787 struct untracked_cache_dir *untracked,
1788 struct cached_dir *cdir,
1789 struct index_state *istate,
1790 struct strbuf *path,
1791 int baselen,
1792 const struct pathspec *pathspec)
1793 {
1794 int dtype;
1795 struct dirent *de = cdir->de;
1796
1797 if (!de)
1798 return treat_path_fast(dir, untracked, cdir, istate, path,
1799 baselen, pathspec);
1800 if (is_dot_or_dotdot(de->d_name) || !fspathcmp(de->d_name, ".git"))
1801 return path_none;
1802 strbuf_setlen(path, baselen);
1803 strbuf_addstr(path, de->d_name);
1804 if (simplify_away(path->buf, path->len, pathspec))
1805 return path_none;
1806
1807 dtype = DTYPE(de);
1808 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1809 }
1810
1811 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1812 {
1813 if (!dir)
1814 return;
1815 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1816 dir->untracked_alloc);
1817 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1818 }
1819
1820 static int valid_cached_dir(struct dir_struct *dir,
1821 struct untracked_cache_dir *untracked,
1822 struct index_state *istate,
1823 struct strbuf *path,
1824 int check_only)
1825 {
1826 struct stat st;
1827
1828 if (!untracked)
1829 return 0;
1830
1831 /*
1832 * With fsmonitor, we can trust the untracked cache's valid field.
1833 */
1834 refresh_fsmonitor(istate);
1835 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1836 if (lstat(path->len ? path->buf : ".", &st)) {
1837 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1838 return 0;
1839 }
1840 if (!untracked->valid ||
1841 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1842 fill_stat_data(&untracked->stat_data, &st);
1843 return 0;
1844 }
1845 }
1846
1847 if (untracked->check_only != !!check_only)
1848 return 0;
1849
1850 /*
1851 * prep_exclude will be called eventually on this directory,
1852 * but it's called much later in last_matching_pattern(). We
1853 * need it now to determine the validity of the cache for this
1854 * path. The next calls will be nearly no-op, the way
1855 * prep_exclude() is designed.
1856 */
1857 if (path->len && path->buf[path->len - 1] != '/') {
1858 strbuf_addch(path, '/');
1859 prep_exclude(dir, istate, path->buf, path->len);
1860 strbuf_setlen(path, path->len - 1);
1861 } else
1862 prep_exclude(dir, istate, path->buf, path->len);
1863
1864 /* hopefully prep_exclude() haven't invalidated this entry... */
1865 return untracked->valid;
1866 }
1867
1868 static int open_cached_dir(struct cached_dir *cdir,
1869 struct dir_struct *dir,
1870 struct untracked_cache_dir *untracked,
1871 struct index_state *istate,
1872 struct strbuf *path,
1873 int check_only)
1874 {
1875 const char *c_path;
1876
1877 memset(cdir, 0, sizeof(*cdir));
1878 cdir->untracked = untracked;
1879 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1880 return 0;
1881 c_path = path->len ? path->buf : ".";
1882 cdir->fdir = opendir(c_path);
1883 if (!cdir->fdir)
1884 warning_errno(_("could not open directory '%s'"), c_path);
1885 if (dir->untracked) {
1886 invalidate_directory(dir->untracked, untracked);
1887 dir->untracked->dir_opened++;
1888 }
1889 if (!cdir->fdir)
1890 return -1;
1891 return 0;
1892 }
1893
1894 static int read_cached_dir(struct cached_dir *cdir)
1895 {
1896 if (cdir->fdir) {
1897 cdir->de = readdir(cdir->fdir);
1898 if (!cdir->de)
1899 return -1;
1900 return 0;
1901 }
1902 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1903 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1904 if (!d->recurse) {
1905 cdir->nr_dirs++;
1906 continue;
1907 }
1908 cdir->ucd = d;
1909 cdir->nr_dirs++;
1910 return 0;
1911 }
1912 cdir->ucd = NULL;
1913 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1914 struct untracked_cache_dir *d = cdir->untracked;
1915 cdir->file = d->untracked[cdir->nr_files++];
1916 return 0;
1917 }
1918 return -1;
1919 }
1920
1921 static void close_cached_dir(struct cached_dir *cdir)
1922 {
1923 if (cdir->fdir)
1924 closedir(cdir->fdir);
1925 /*
1926 * We have gone through this directory and found no untracked
1927 * entries. Mark it valid.
1928 */
1929 if (cdir->untracked) {
1930 cdir->untracked->valid = 1;
1931 cdir->untracked->recurse = 1;
1932 }
1933 }
1934
1935 static void add_path_to_appropriate_result_list(struct dir_struct *dir,
1936 struct untracked_cache_dir *untracked,
1937 struct cached_dir *cdir,
1938 struct index_state *istate,
1939 struct strbuf *path,
1940 int baselen,
1941 const struct pathspec *pathspec,
1942 enum path_treatment state)
1943 {
1944 /* add the path to the appropriate result list */
1945 switch (state) {
1946 case path_excluded:
1947 if (dir->flags & DIR_SHOW_IGNORED)
1948 dir_add_name(dir, istate, path->buf, path->len);
1949 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1950 ((dir->flags & DIR_COLLECT_IGNORED) &&
1951 exclude_matches_pathspec(path->buf, path->len,
1952 pathspec)))
1953 dir_add_ignored(dir, istate, path->buf, path->len);
1954 break;
1955
1956 case path_untracked:
1957 if (dir->flags & DIR_SHOW_IGNORED)
1958 break;
1959 dir_add_name(dir, istate, path->buf, path->len);
1960 if (cdir->fdir)
1961 add_untracked(untracked, path->buf + baselen);
1962 break;
1963
1964 default:
1965 break;
1966 }
1967 }
1968
1969 /*
1970 * Read a directory tree. We currently ignore anything but
1971 * directories, regular files and symlinks. That's because git
1972 * doesn't handle them at all yet. Maybe that will change some
1973 * day.
1974 *
1975 * Also, we ignore the name ".git" (even if it is not a directory).
1976 * That likely will not change.
1977 *
1978 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1979 * to signal that a file was found. This is the least significant value that
1980 * indicates that a file was encountered that does not depend on the order of
1981 * whether an untracked or exluded path was encountered first.
1982 *
1983 * Returns the most significant path_treatment value encountered in the scan.
1984 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1985 * significant path_treatment value that will be returned.
1986 */
1987
1988 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1989 struct index_state *istate, const char *base, int baselen,
1990 struct untracked_cache_dir *untracked, int check_only,
1991 int stop_at_first_file, const struct pathspec *pathspec)
1992 {
1993 struct cached_dir cdir;
1994 enum path_treatment state, subdir_state, dir_state = path_none;
1995 struct strbuf path = STRBUF_INIT;
1996
1997 strbuf_add(&path, base, baselen);
1998
1999 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
2000 goto out;
2001
2002 if (untracked)
2003 untracked->check_only = !!check_only;
2004
2005 while (!read_cached_dir(&cdir)) {
2006 /* check how the file or directory should be treated */
2007 state = treat_path(dir, untracked, &cdir, istate, &path,
2008 baselen, pathspec);
2009
2010 if (state > dir_state)
2011 dir_state = state;
2012
2013 /* recurse into subdir if instructed by treat_path */
2014 if ((state == path_recurse) ||
2015 ((state == path_untracked) &&
2016 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR) &&
2017 ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2018 (pathspec &&
2019 do_match_pathspec(istate, pathspec, path.buf, path.len,
2020 baselen, NULL, DO_MATCH_LEADING_PATHSPEC) == MATCHED_RECURSIVELY_LEADING_PATHSPEC)))) {
2021 struct untracked_cache_dir *ud;
2022 ud = lookup_untracked(dir->untracked, untracked,
2023 path.buf + baselen,
2024 path.len - baselen);
2025 subdir_state =
2026 read_directory_recursive(dir, istate, path.buf,
2027 path.len, ud,
2028 check_only, stop_at_first_file, pathspec);
2029 if (subdir_state > dir_state)
2030 dir_state = subdir_state;
2031
2032 if (pathspec &&
2033 !match_pathspec(istate, pathspec, path.buf, path.len,
2034 0 /* prefix */, NULL,
2035 0 /* do NOT special case dirs */))
2036 state = path_none;
2037 }
2038
2039 if (check_only) {
2040 if (stop_at_first_file) {
2041 /*
2042 * If stopping at first file, then
2043 * signal that a file was found by
2044 * returning `path_excluded`. This is
2045 * to return a consistent value
2046 * regardless of whether an ignored or
2047 * excluded file happened to be
2048 * encountered 1st.
2049 *
2050 * In current usage, the
2051 * `stop_at_first_file` is passed when
2052 * an ancestor directory has matched
2053 * an exclude pattern, so any found
2054 * files will be excluded.
2055 */
2056 if (dir_state >= path_excluded) {
2057 dir_state = path_excluded;
2058 break;
2059 }
2060 }
2061
2062 /* abort early if maximum state has been reached */
2063 if (dir_state == path_untracked) {
2064 if (cdir.fdir)
2065 add_untracked(untracked, path.buf + baselen);
2066 break;
2067 }
2068 /* skip the dir_add_* part */
2069 continue;
2070 }
2071
2072 add_path_to_appropriate_result_list(dir, untracked, &cdir,
2073 istate, &path, baselen,
2074 pathspec, state);
2075 }
2076 close_cached_dir(&cdir);
2077 out:
2078 strbuf_release(&path);
2079
2080 return dir_state;
2081 }
2082
2083 int cmp_dir_entry(const void *p1, const void *p2)
2084 {
2085 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2086 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2087
2088 return name_compare(e1->name, e1->len, e2->name, e2->len);
2089 }
2090
2091 /* check if *out lexically strictly contains *in */
2092 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2093 {
2094 return (out->len < in->len) &&
2095 (out->name[out->len - 1] == '/') &&
2096 !memcmp(out->name, in->name, out->len);
2097 }
2098
2099 static int treat_leading_path(struct dir_struct *dir,
2100 struct index_state *istate,
2101 const char *path, int len,
2102 const struct pathspec *pathspec)
2103 {
2104 struct strbuf sb = STRBUF_INIT;
2105 int prevlen, baselen;
2106 const char *cp;
2107 struct cached_dir cdir;
2108 struct dirent *de;
2109 enum path_treatment state = path_none;
2110
2111 /*
2112 * For each directory component of path, we are going to check whether
2113 * that path is relevant given the pathspec. For example, if path is
2114 * foo/bar/baz/
2115 * then we will ask treat_path() whether we should go into foo, then
2116 * whether we should go into bar, then whether baz is relevant.
2117 * Checking each is important because e.g. if path is
2118 * .git/info/
2119 * then we need to check .git to know we shouldn't traverse it.
2120 * If the return from treat_path() is:
2121 * * path_none, for any path, we return false.
2122 * * path_recurse, for all path components, we return true
2123 * * <anything else> for some intermediate component, we make sure
2124 * to add that path to the relevant list but return false
2125 * signifying that we shouldn't recurse into it.
2126 */
2127
2128 while (len && path[len - 1] == '/')
2129 len--;
2130 if (!len)
2131 return 1;
2132
2133 /*
2134 * We need a manufactured dirent with sufficient space to store a
2135 * leading directory component of path in its d_name. Here, we
2136 * assume that the dirent's d_name is either declared as
2137 * char d_name[BIG_ENOUGH]
2138 * or that it is declared at the end of the struct as
2139 * char d_name[]
2140 * For either case, padding with len+1 bytes at the end will ensure
2141 * sufficient storage space.
2142 */
2143 de = xcalloc(1, sizeof(struct dirent)+len+1);
2144 memset(&cdir, 0, sizeof(cdir));
2145 cdir.de = de;
2146 #if defined(DT_UNKNOWN) && !defined(NO_D_TYPE_IN_DIRENT)
2147 de->d_type = DT_DIR;
2148 #endif
2149 baselen = 0;
2150 prevlen = 0;
2151 while (1) {
2152 prevlen = baselen + !!baselen;
2153 cp = path + prevlen;
2154 cp = memchr(cp, '/', path + len - cp);
2155 if (!cp)
2156 baselen = len;
2157 else
2158 baselen = cp - path;
2159 strbuf_reset(&sb);
2160 strbuf_add(&sb, path, baselen);
2161 if (!is_directory(sb.buf))
2162 break;
2163 strbuf_reset(&sb);
2164 strbuf_add(&sb, path, prevlen);
2165 memcpy(de->d_name, path+prevlen, baselen-prevlen);
2166 de->d_name[baselen-prevlen] = '\0';
2167 state = treat_path(dir, NULL, &cdir, istate, &sb, prevlen,
2168 pathspec);
2169 if (state != path_recurse)
2170 break; /* do not recurse into it */
2171 if (len <= baselen)
2172 break; /* finished checking */
2173 }
2174 add_path_to_appropriate_result_list(dir, NULL, &cdir, istate,
2175 &sb, baselen, pathspec,
2176 state);
2177
2178 free(de);
2179 strbuf_release(&sb);
2180 return state == path_recurse;
2181 }
2182
2183 static const char *get_ident_string(void)
2184 {
2185 static struct strbuf sb = STRBUF_INIT;
2186 struct utsname uts;
2187
2188 if (sb.len)
2189 return sb.buf;
2190 if (uname(&uts) < 0)
2191 die_errno(_("failed to get kernel name and information"));
2192 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2193 uts.sysname);
2194 return sb.buf;
2195 }
2196
2197 static int ident_in_untracked(const struct untracked_cache *uc)
2198 {
2199 /*
2200 * Previous git versions may have saved many NUL separated
2201 * strings in the "ident" field, but it is insane to manage
2202 * many locations, so just take care of the first one.
2203 */
2204
2205 return !strcmp(uc->ident.buf, get_ident_string());
2206 }
2207
2208 static void set_untracked_ident(struct untracked_cache *uc)
2209 {
2210 strbuf_reset(&uc->ident);
2211 strbuf_addstr(&uc->ident, get_ident_string());
2212
2213 /*
2214 * This strbuf used to contain a list of NUL separated
2215 * strings, so save NUL too for backward compatibility.
2216 */
2217 strbuf_addch(&uc->ident, 0);
2218 }
2219
2220 static void new_untracked_cache(struct index_state *istate)
2221 {
2222 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2223 strbuf_init(&uc->ident, 100);
2224 uc->exclude_per_dir = ".gitignore";
2225 /* should be the same flags used by git-status */
2226 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2227 set_untracked_ident(uc);
2228 istate->untracked = uc;
2229 istate->cache_changed |= UNTRACKED_CHANGED;
2230 }
2231
2232 void add_untracked_cache(struct index_state *istate)
2233 {
2234 if (!istate->untracked) {
2235 new_untracked_cache(istate);
2236 } else {
2237 if (!ident_in_untracked(istate->untracked)) {
2238 free_untracked_cache(istate->untracked);
2239 new_untracked_cache(istate);
2240 }
2241 }
2242 }
2243
2244 void remove_untracked_cache(struct index_state *istate)
2245 {
2246 if (istate->untracked) {
2247 free_untracked_cache(istate->untracked);
2248 istate->untracked = NULL;
2249 istate->cache_changed |= UNTRACKED_CHANGED;
2250 }
2251 }
2252
2253 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2254 int base_len,
2255 const struct pathspec *pathspec)
2256 {
2257 struct untracked_cache_dir *root;
2258 static int untracked_cache_disabled = -1;
2259
2260 if (!dir->untracked)
2261 return NULL;
2262 if (untracked_cache_disabled < 0)
2263 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2264 if (untracked_cache_disabled)
2265 return NULL;
2266
2267 /*
2268 * We only support $GIT_DIR/info/exclude and core.excludesfile
2269 * as the global ignore rule files. Any other additions
2270 * (e.g. from command line) invalidate the cache. This
2271 * condition also catches running setup_standard_excludes()
2272 * before setting dir->untracked!
2273 */
2274 if (dir->unmanaged_exclude_files)
2275 return NULL;
2276
2277 /*
2278 * Optimize for the main use case only: whole-tree git
2279 * status. More work involved in treat_leading_path() if we
2280 * use cache on just a subset of the worktree. pathspec
2281 * support could make the matter even worse.
2282 */
2283 if (base_len || (pathspec && pathspec->nr))
2284 return NULL;
2285
2286 /* Different set of flags may produce different results */
2287 if (dir->flags != dir->untracked->dir_flags ||
2288 /*
2289 * See treat_directory(), case index_nonexistent. Without
2290 * this flag, we may need to also cache .git file content
2291 * for the resolve_gitlink_ref() call, which we don't.
2292 */
2293 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2294 /* We don't support collecting ignore files */
2295 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2296 DIR_COLLECT_IGNORED)))
2297 return NULL;
2298
2299 /*
2300 * If we use .gitignore in the cache and now you change it to
2301 * .gitexclude, everything will go wrong.
2302 */
2303 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2304 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2305 return NULL;
2306
2307 /*
2308 * EXC_CMDL is not considered in the cache. If people set it,
2309 * skip the cache.
2310 */
2311 if (dir->exclude_list_group[EXC_CMDL].nr)
2312 return NULL;
2313
2314 if (!ident_in_untracked(dir->untracked)) {
2315 warning(_("untracked cache is disabled on this system or location"));
2316 return NULL;
2317 }
2318
2319 if (!dir->untracked->root) {
2320 const int len = sizeof(*dir->untracked->root);
2321 dir->untracked->root = xmalloc(len);
2322 memset(dir->untracked->root, 0, len);
2323 }
2324
2325 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2326 root = dir->untracked->root;
2327 if (!oideq(&dir->ss_info_exclude.oid,
2328 &dir->untracked->ss_info_exclude.oid)) {
2329 invalidate_gitignore(dir->untracked, root);
2330 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2331 }
2332 if (!oideq(&dir->ss_excludes_file.oid,
2333 &dir->untracked->ss_excludes_file.oid)) {
2334 invalidate_gitignore(dir->untracked, root);
2335 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2336 }
2337
2338 /* Make sure this directory is not dropped out at saving phase */
2339 root->recurse = 1;
2340 return root;
2341 }
2342
2343 int read_directory(struct dir_struct *dir, struct index_state *istate,
2344 const char *path, int len, const struct pathspec *pathspec)
2345 {
2346 struct untracked_cache_dir *untracked;
2347
2348 trace_performance_enter();
2349
2350 if (has_symlink_leading_path(path, len)) {
2351 trace_performance_leave("read directory %.*s", len, path);
2352 return dir->nr;
2353 }
2354
2355 untracked = validate_untracked_cache(dir, len, pathspec);
2356 if (!untracked)
2357 /*
2358 * make sure untracked cache code path is disabled,
2359 * e.g. prep_exclude()
2360 */
2361 dir->untracked = NULL;
2362 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2363 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2364 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2365 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2366
2367 /*
2368 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2369 * also pick up untracked contents of untracked dirs; by default
2370 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2371 */
2372 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2373 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2374 int i, j;
2375
2376 /* remove from dir->entries untracked contents of untracked dirs */
2377 for (i = j = 0; j < dir->nr; j++) {
2378 if (i &&
2379 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2380 FREE_AND_NULL(dir->entries[j]);
2381 } else {
2382 dir->entries[i++] = dir->entries[j];
2383 }
2384 }
2385
2386 dir->nr = i;
2387 }
2388
2389 trace_performance_leave("read directory %.*s", len, path);
2390 if (dir->untracked) {
2391 static int force_untracked_cache = -1;
2392 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2393
2394 if (force_untracked_cache < 0)
2395 force_untracked_cache =
2396 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2397 trace_printf_key(&trace_untracked_stats,
2398 "node creation: %u\n"
2399 "gitignore invalidation: %u\n"
2400 "directory invalidation: %u\n"
2401 "opendir: %u\n",
2402 dir->untracked->dir_created,
2403 dir->untracked->gitignore_invalidated,
2404 dir->untracked->dir_invalidated,
2405 dir->untracked->dir_opened);
2406 if (force_untracked_cache &&
2407 dir->untracked == istate->untracked &&
2408 (dir->untracked->dir_opened ||
2409 dir->untracked->gitignore_invalidated ||
2410 dir->untracked->dir_invalidated))
2411 istate->cache_changed |= UNTRACKED_CHANGED;
2412 if (dir->untracked != istate->untracked) {
2413 FREE_AND_NULL(dir->untracked);
2414 }
2415 }
2416 return dir->nr;
2417 }
2418
2419 int file_exists(const char *f)
2420 {
2421 struct stat sb;
2422 return lstat(f, &sb) == 0;
2423 }
2424
2425 int repo_file_exists(struct repository *repo, const char *path)
2426 {
2427 if (repo != the_repository)
2428 BUG("do not know how to check file existence in arbitrary repo");
2429
2430 return file_exists(path);
2431 }
2432
2433 static int cmp_icase(char a, char b)
2434 {
2435 if (a == b)
2436 return 0;
2437 if (ignore_case)
2438 return toupper(a) - toupper(b);
2439 return a - b;
2440 }
2441
2442 /*
2443 * Given two normalized paths (a trailing slash is ok), if subdir is
2444 * outside dir, return -1. Otherwise return the offset in subdir that
2445 * can be used as relative path to dir.
2446 */
2447 int dir_inside_of(const char *subdir, const char *dir)
2448 {
2449 int offset = 0;
2450
2451 assert(dir && subdir && *dir && *subdir);
2452
2453 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2454 dir++;
2455 subdir++;
2456 offset++;
2457 }
2458
2459 /* hel[p]/me vs hel[l]/yeah */
2460 if (*dir && *subdir)
2461 return -1;
2462
2463 if (!*subdir)
2464 return !*dir ? offset : -1; /* same dir */
2465
2466 /* foo/[b]ar vs foo/[] */
2467 if (is_dir_sep(dir[-1]))
2468 return is_dir_sep(subdir[-1]) ? offset : -1;
2469
2470 /* foo[/]bar vs foo[] */
2471 return is_dir_sep(*subdir) ? offset + 1 : -1;
2472 }
2473
2474 int is_inside_dir(const char *dir)
2475 {
2476 char *cwd;
2477 int rc;
2478
2479 if (!dir)
2480 return 0;
2481
2482 cwd = xgetcwd();
2483 rc = (dir_inside_of(cwd, dir) >= 0);
2484 free(cwd);
2485 return rc;
2486 }
2487
2488 int is_empty_dir(const char *path)
2489 {
2490 DIR *dir = opendir(path);
2491 struct dirent *e;
2492 int ret = 1;
2493
2494 if (!dir)
2495 return 0;
2496
2497 while ((e = readdir(dir)) != NULL)
2498 if (!is_dot_or_dotdot(e->d_name)) {
2499 ret = 0;
2500 break;
2501 }
2502
2503 closedir(dir);
2504 return ret;
2505 }
2506
2507 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2508 {
2509 DIR *dir;
2510 struct dirent *e;
2511 int ret = 0, original_len = path->len, len, kept_down = 0;
2512 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2513 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2514 struct object_id submodule_head;
2515
2516 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2517 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2518 /* Do not descend and nuke a nested git work tree. */
2519 if (kept_up)
2520 *kept_up = 1;
2521 return 0;
2522 }
2523
2524 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2525 dir = opendir(path->buf);
2526 if (!dir) {
2527 if (errno == ENOENT)
2528 return keep_toplevel ? -1 : 0;
2529 else if (errno == EACCES && !keep_toplevel)
2530 /*
2531 * An empty dir could be removable even if it
2532 * is unreadable:
2533 */
2534 return rmdir(path->buf);
2535 else
2536 return -1;
2537 }
2538 strbuf_complete(path, '/');
2539
2540 len = path->len;
2541 while ((e = readdir(dir)) != NULL) {
2542 struct stat st;
2543 if (is_dot_or_dotdot(e->d_name))
2544 continue;
2545
2546 strbuf_setlen(path, len);
2547 strbuf_addstr(path, e->d_name);
2548 if (lstat(path->buf, &st)) {
2549 if (errno == ENOENT)
2550 /*
2551 * file disappeared, which is what we
2552 * wanted anyway
2553 */
2554 continue;
2555 /* fall thru */
2556 } else if (S_ISDIR(st.st_mode)) {
2557 if (!remove_dir_recurse(path, flag, &kept_down))
2558 continue; /* happy */
2559 } else if (!only_empty &&
2560 (!unlink(path->buf) || errno == ENOENT)) {
2561 continue; /* happy, too */
2562 }
2563
2564 /* path too long, stat fails, or non-directory still exists */
2565 ret = -1;
2566 break;
2567 }
2568 closedir(dir);
2569
2570 strbuf_setlen(path, original_len);
2571 if (!ret && !keep_toplevel && !kept_down)
2572 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2573 else if (kept_up)
2574 /*
2575 * report the uplevel that it is not an error that we
2576 * did not rmdir() our directory.
2577 */
2578 *kept_up = !ret;
2579 return ret;
2580 }
2581
2582 int remove_dir_recursively(struct strbuf *path, int flag)
2583 {
2584 return remove_dir_recurse(path, flag, NULL);
2585 }
2586
2587 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2588
2589 void setup_standard_excludes(struct dir_struct *dir)
2590 {
2591 dir->exclude_per_dir = ".gitignore";
2592
2593 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2594 if (!excludes_file)
2595 excludes_file = xdg_config_home("ignore");
2596 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2597 add_patterns_from_file_1(dir, excludes_file,
2598 dir->untracked ? &dir->ss_excludes_file : NULL);
2599
2600 /* per repository user preference */
2601 if (startup_info->have_repository) {
2602 const char *path = git_path_info_exclude();
2603 if (!access_or_warn(path, R_OK, 0))
2604 add_patterns_from_file_1(dir, path,
2605 dir->untracked ? &dir->ss_info_exclude : NULL);
2606 }
2607 }
2608
2609 int remove_path(const char *name)
2610 {
2611 char *slash;
2612
2613 if (unlink(name) && !is_missing_file_error(errno))
2614 return -1;
2615
2616 slash = strrchr(name, '/');
2617 if (slash) {
2618 char *dirs = xstrdup(name);
2619 slash = dirs + (slash - name);
2620 do {
2621 *slash = '\0';
2622 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2623 free(dirs);
2624 }
2625 return 0;
2626 }
2627
2628 /*
2629 * Frees memory within dir which was allocated for exclude lists and
2630 * the exclude_stack. Does not free dir itself.
2631 */
2632 void clear_directory(struct dir_struct *dir)
2633 {
2634 int i, j;
2635 struct exclude_list_group *group;
2636 struct pattern_list *pl;
2637 struct exclude_stack *stk;
2638
2639 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2640 group = &dir->exclude_list_group[i];
2641 for (j = 0; j < group->nr; j++) {
2642 pl = &group->pl[j];
2643 if (i == EXC_DIRS)
2644 free((char *)pl->src);
2645 clear_pattern_list(pl);
2646 }
2647 free(group->pl);
2648 }
2649
2650 stk = dir->exclude_stack;
2651 while (stk) {
2652 struct exclude_stack *prev = stk->prev;
2653 free(stk);
2654 stk = prev;
2655 }
2656 strbuf_release(&dir->basebuf);
2657 }
2658
2659 struct ondisk_untracked_cache {
2660 struct stat_data info_exclude_stat;
2661 struct stat_data excludes_file_stat;
2662 uint32_t dir_flags;
2663 };
2664
2665 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2666
2667 struct write_data {
2668 int index; /* number of written untracked_cache_dir */
2669 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2670 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2671 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2672 struct strbuf out;
2673 struct strbuf sb_stat;
2674 struct strbuf sb_sha1;
2675 };
2676
2677 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2678 {
2679 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2680 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2681 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2682 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2683 to->sd_dev = htonl(from->sd_dev);
2684 to->sd_ino = htonl(from->sd_ino);
2685 to->sd_uid = htonl(from->sd_uid);
2686 to->sd_gid = htonl(from->sd_gid);
2687 to->sd_size = htonl(from->sd_size);
2688 }
2689
2690 static void write_one_dir(struct untracked_cache_dir *untracked,
2691 struct write_data *wd)
2692 {
2693 struct stat_data stat_data;
2694 struct strbuf *out = &wd->out;
2695 unsigned char intbuf[16];
2696 unsigned int intlen, value;
2697 int i = wd->index++;
2698
2699 /*
2700 * untracked_nr should be reset whenever valid is clear, but
2701 * for safety..
2702 */
2703 if (!untracked->valid) {
2704 untracked->untracked_nr = 0;
2705 untracked->check_only = 0;
2706 }
2707
2708 if (untracked->check_only)
2709 ewah_set(wd->check_only, i);
2710 if (untracked->valid) {
2711 ewah_set(wd->valid, i);
2712 stat_data_to_disk(&stat_data, &untracked->stat_data);
2713 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2714 }
2715 if (!is_null_oid(&untracked->exclude_oid)) {
2716 ewah_set(wd->sha1_valid, i);
2717 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
2718 the_hash_algo->rawsz);
2719 }
2720
2721 intlen = encode_varint(untracked->untracked_nr, intbuf);
2722 strbuf_add(out, intbuf, intlen);
2723
2724 /* skip non-recurse directories */
2725 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2726 if (untracked->dirs[i]->recurse)
2727 value++;
2728 intlen = encode_varint(value, intbuf);
2729 strbuf_add(out, intbuf, intlen);
2730
2731 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2732
2733 for (i = 0; i < untracked->untracked_nr; i++)
2734 strbuf_add(out, untracked->untracked[i],
2735 strlen(untracked->untracked[i]) + 1);
2736
2737 for (i = 0; i < untracked->dirs_nr; i++)
2738 if (untracked->dirs[i]->recurse)
2739 write_one_dir(untracked->dirs[i], wd);
2740 }
2741
2742 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2743 {
2744 struct ondisk_untracked_cache *ouc;
2745 struct write_data wd;
2746 unsigned char varbuf[16];
2747 int varint_len;
2748 const unsigned hashsz = the_hash_algo->rawsz;
2749
2750 ouc = xcalloc(1, sizeof(*ouc));
2751 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2752 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2753 ouc->dir_flags = htonl(untracked->dir_flags);
2754
2755 varint_len = encode_varint(untracked->ident.len, varbuf);
2756 strbuf_add(out, varbuf, varint_len);
2757 strbuf_addbuf(out, &untracked->ident);
2758
2759 strbuf_add(out, ouc, sizeof(*ouc));
2760 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
2761 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
2762 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
2763 FREE_AND_NULL(ouc);
2764
2765 if (!untracked->root) {
2766 varint_len = encode_varint(0, varbuf);
2767 strbuf_add(out, varbuf, varint_len);
2768 return;
2769 }
2770
2771 wd.index = 0;
2772 wd.check_only = ewah_new();
2773 wd.valid = ewah_new();
2774 wd.sha1_valid = ewah_new();
2775 strbuf_init(&wd.out, 1024);
2776 strbuf_init(&wd.sb_stat, 1024);
2777 strbuf_init(&wd.sb_sha1, 1024);
2778 write_one_dir(untracked->root, &wd);
2779
2780 varint_len = encode_varint(wd.index, varbuf);
2781 strbuf_add(out, varbuf, varint_len);
2782 strbuf_addbuf(out, &wd.out);
2783 ewah_serialize_strbuf(wd.valid, out);
2784 ewah_serialize_strbuf(wd.check_only, out);
2785 ewah_serialize_strbuf(wd.sha1_valid, out);
2786 strbuf_addbuf(out, &wd.sb_stat);
2787 strbuf_addbuf(out, &wd.sb_sha1);
2788 strbuf_addch(out, '\0'); /* safe guard for string lists */
2789
2790 ewah_free(wd.valid);
2791 ewah_free(wd.check_only);
2792 ewah_free(wd.sha1_valid);
2793 strbuf_release(&wd.out);
2794 strbuf_release(&wd.sb_stat);
2795 strbuf_release(&wd.sb_sha1);
2796 }
2797
2798 static void free_untracked(struct untracked_cache_dir *ucd)
2799 {
2800 int i;
2801 if (!ucd)
2802 return;
2803 for (i = 0; i < ucd->dirs_nr; i++)
2804 free_untracked(ucd->dirs[i]);
2805 for (i = 0; i < ucd->untracked_nr; i++)
2806 free(ucd->untracked[i]);
2807 free(ucd->untracked);
2808 free(ucd->dirs);
2809 free(ucd);
2810 }
2811
2812 void free_untracked_cache(struct untracked_cache *uc)
2813 {
2814 if (uc)
2815 free_untracked(uc->root);
2816 free(uc);
2817 }
2818
2819 struct read_data {
2820 int index;
2821 struct untracked_cache_dir **ucd;
2822 struct ewah_bitmap *check_only;
2823 struct ewah_bitmap *valid;
2824 struct ewah_bitmap *sha1_valid;
2825 const unsigned char *data;
2826 const unsigned char *end;
2827 };
2828
2829 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2830 {
2831 memcpy(to, data, sizeof(*to));
2832 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2833 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2834 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2835 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2836 to->sd_dev = ntohl(to->sd_dev);
2837 to->sd_ino = ntohl(to->sd_ino);
2838 to->sd_uid = ntohl(to->sd_uid);
2839 to->sd_gid = ntohl(to->sd_gid);
2840 to->sd_size = ntohl(to->sd_size);
2841 }
2842
2843 static int read_one_dir(struct untracked_cache_dir **untracked_,
2844 struct read_data *rd)
2845 {
2846 struct untracked_cache_dir ud, *untracked;
2847 const unsigned char *data = rd->data, *end = rd->end;
2848 const unsigned char *eos;
2849 unsigned int value;
2850 int i;
2851
2852 memset(&ud, 0, sizeof(ud));
2853
2854 value = decode_varint(&data);
2855 if (data > end)
2856 return -1;
2857 ud.recurse = 1;
2858 ud.untracked_alloc = value;
2859 ud.untracked_nr = value;
2860 if (ud.untracked_nr)
2861 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2862
2863 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
2864 if (data > end)
2865 return -1;
2866 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2867
2868 eos = memchr(data, '\0', end - data);
2869 if (!eos || eos == end)
2870 return -1;
2871
2872 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
2873 memcpy(untracked, &ud, sizeof(ud));
2874 memcpy(untracked->name, data, eos - data + 1);
2875 data = eos + 1;
2876
2877 for (i = 0; i < untracked->untracked_nr; i++) {
2878 eos = memchr(data, '\0', end - data);
2879 if (!eos || eos == end)
2880 return -1;
2881 untracked->untracked[i] = xmemdupz(data, eos - data);
2882 data = eos + 1;
2883 }
2884
2885 rd->ucd[rd->index++] = untracked;
2886 rd->data = data;
2887
2888 for (i = 0; i < untracked->dirs_nr; i++) {
2889 if (read_one_dir(untracked->dirs + i, rd) < 0)
2890 return -1;
2891 }
2892 return 0;
2893 }
2894
2895 static void set_check_only(size_t pos, void *cb)
2896 {
2897 struct read_data *rd = cb;
2898 struct untracked_cache_dir *ud = rd->ucd[pos];
2899 ud->check_only = 1;
2900 }
2901
2902 static void read_stat(size_t pos, void *cb)
2903 {
2904 struct read_data *rd = cb;
2905 struct untracked_cache_dir *ud = rd->ucd[pos];
2906 if (rd->data + sizeof(struct stat_data) > rd->end) {
2907 rd->data = rd->end + 1;
2908 return;
2909 }
2910 stat_data_from_disk(&ud->stat_data, rd->data);
2911 rd->data += sizeof(struct stat_data);
2912 ud->valid = 1;
2913 }
2914
2915 static void read_oid(size_t pos, void *cb)
2916 {
2917 struct read_data *rd = cb;
2918 struct untracked_cache_dir *ud = rd->ucd[pos];
2919 if (rd->data + the_hash_algo->rawsz > rd->end) {
2920 rd->data = rd->end + 1;
2921 return;
2922 }
2923 hashcpy(ud->exclude_oid.hash, rd->data);
2924 rd->data += the_hash_algo->rawsz;
2925 }
2926
2927 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
2928 const unsigned char *sha1)
2929 {
2930 stat_data_from_disk(&oid_stat->stat, data);
2931 hashcpy(oid_stat->oid.hash, sha1);
2932 oid_stat->valid = 1;
2933 }
2934
2935 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2936 {
2937 struct untracked_cache *uc;
2938 struct read_data rd;
2939 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2940 const char *ident;
2941 int ident_len;
2942 ssize_t len;
2943 const char *exclude_per_dir;
2944 const unsigned hashsz = the_hash_algo->rawsz;
2945 const unsigned offset = sizeof(struct ondisk_untracked_cache);
2946 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
2947
2948 if (sz <= 1 || end[-1] != '\0')
2949 return NULL;
2950 end--;
2951
2952 ident_len = decode_varint(&next);
2953 if (next + ident_len > end)
2954 return NULL;
2955 ident = (const char *)next;
2956 next += ident_len;
2957
2958 if (next + exclude_per_dir_offset + 1 > end)
2959 return NULL;
2960
2961 uc = xcalloc(1, sizeof(*uc));
2962 strbuf_init(&uc->ident, ident_len);
2963 strbuf_add(&uc->ident, ident, ident_len);
2964 load_oid_stat(&uc->ss_info_exclude,
2965 next + ouc_offset(info_exclude_stat),
2966 next + offset);
2967 load_oid_stat(&uc->ss_excludes_file,
2968 next + ouc_offset(excludes_file_stat),
2969 next + offset + hashsz);
2970 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2971 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
2972 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2973 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2974 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
2975 if (next >= end)
2976 goto done2;
2977
2978 len = decode_varint(&next);
2979 if (next > end || len == 0)
2980 goto done2;
2981
2982 rd.valid = ewah_new();
2983 rd.check_only = ewah_new();
2984 rd.sha1_valid = ewah_new();
2985 rd.data = next;
2986 rd.end = end;
2987 rd.index = 0;
2988 ALLOC_ARRAY(rd.ucd, len);
2989
2990 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2991 goto done;
2992
2993 next = rd.data;
2994 len = ewah_read_mmap(rd.valid, next, end - next);
2995 if (len < 0)
2996 goto done;
2997
2998 next += len;
2999 len = ewah_read_mmap(rd.check_only, next, end - next);
3000 if (len < 0)
3001 goto done;
3002
3003 next += len;
3004 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
3005 if (len < 0)
3006 goto done;
3007
3008 ewah_each_bit(rd.check_only, set_check_only, &rd);
3009 rd.data = next + len;
3010 ewah_each_bit(rd.valid, read_stat, &rd);
3011 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
3012 next = rd.data;
3013
3014 done:
3015 free(rd.ucd);
3016 ewah_free(rd.valid);
3017 ewah_free(rd.check_only);
3018 ewah_free(rd.sha1_valid);
3019 done2:
3020 if (next != end) {
3021 free_untracked_cache(uc);
3022 uc = NULL;
3023 }
3024 return uc;
3025 }
3026
3027 static void invalidate_one_directory(struct untracked_cache *uc,
3028 struct untracked_cache_dir *ucd)
3029 {
3030 uc->dir_invalidated++;
3031 ucd->valid = 0;
3032 ucd->untracked_nr = 0;
3033 }
3034
3035 /*
3036 * Normally when an entry is added or removed from a directory,
3037 * invalidating that directory is enough. No need to touch its
3038 * ancestors. When a directory is shown as "foo/bar/" in git-status
3039 * however, deleting or adding an entry may have cascading effect.
3040 *
3041 * Say the "foo/bar/file" has become untracked, we need to tell the
3042 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3043 * directory any more (because "bar" is managed by foo as an untracked
3044 * "file").
3045 *
3046 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3047 * was the last untracked entry in the entire "foo", we should show
3048 * "foo/" instead. Which means we have to invalidate past "bar" up to
3049 * "foo".
3050 *
3051 * This function traverses all directories from root to leaf. If there
3052 * is a chance of one of the above cases happening, we invalidate back
3053 * to root. Otherwise we just invalidate the leaf. There may be a more
3054 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3055 * detect these cases and avoid unnecessary invalidation, for example,
3056 * checking for the untracked entry named "bar/" in "foo", but for now
3057 * stick to something safe and simple.
3058 */
3059 static int invalidate_one_component(struct untracked_cache *uc,
3060 struct untracked_cache_dir *dir,
3061 const char *path, int len)
3062 {
3063 const char *rest = strchr(path, '/');
3064
3065 if (rest) {
3066 int component_len = rest - path;
3067 struct untracked_cache_dir *d =
3068 lookup_untracked(uc, dir, path, component_len);
3069 int ret =
3070 invalidate_one_component(uc, d, rest + 1,
3071 len - (component_len + 1));
3072 if (ret)
3073 invalidate_one_directory(uc, dir);
3074 return ret;
3075 }
3076
3077 invalidate_one_directory(uc, dir);
3078 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
3079 }
3080
3081 void untracked_cache_invalidate_path(struct index_state *istate,
3082 const char *path, int safe_path)
3083 {
3084 if (!istate->untracked || !istate->untracked->root)
3085 return;
3086 if (!safe_path && !verify_path(path, 0))
3087 return;
3088 invalidate_one_component(istate->untracked, istate->untracked->root,
3089 path, strlen(path));
3090 }
3091
3092 void untracked_cache_remove_from_index(struct index_state *istate,
3093 const char *path)
3094 {
3095 untracked_cache_invalidate_path(istate, path, 1);
3096 }
3097
3098 void untracked_cache_add_to_index(struct index_state *istate,
3099 const char *path)
3100 {
3101 untracked_cache_invalidate_path(istate, path, 1);
3102 }
3103
3104 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3105 const char *sub_gitdir)
3106 {
3107 int i;
3108 struct repository subrepo;
3109 struct strbuf sub_wt = STRBUF_INIT;
3110 struct strbuf sub_gd = STRBUF_INIT;
3111
3112 const struct submodule *sub;
3113
3114 /* If the submodule has no working tree, we can ignore it. */
3115 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3116 return;
3117
3118 if (repo_read_index(&subrepo) < 0)
3119 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3120
3121 for (i = 0; i < subrepo.index->cache_nr; i++) {
3122 const struct cache_entry *ce = subrepo.index->cache[i];
3123
3124 if (!S_ISGITLINK(ce->ce_mode))
3125 continue;
3126
3127 while (i + 1 < subrepo.index->cache_nr &&
3128 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3129 /*
3130 * Skip entries with the same name in different stages
3131 * to make sure an entry is returned only once.
3132 */
3133 i++;
3134
3135 sub = submodule_from_path(&subrepo, &null_oid, ce->name);
3136 if (!sub || !is_submodule_active(&subrepo, ce->name))
3137 /* .gitmodules broken or inactive sub */
3138 continue;
3139
3140 strbuf_reset(&sub_wt);
3141 strbuf_reset(&sub_gd);
3142 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3143 strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name);
3144
3145 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3146 }
3147 strbuf_release(&sub_wt);
3148 strbuf_release(&sub_gd);
3149 repo_clear(&subrepo);
3150 }
3151
3152 void connect_work_tree_and_git_dir(const char *work_tree_,
3153 const char *git_dir_,
3154 int recurse_into_nested)
3155 {
3156 struct strbuf gitfile_sb = STRBUF_INIT;
3157 struct strbuf cfg_sb = STRBUF_INIT;
3158 struct strbuf rel_path = STRBUF_INIT;
3159 char *git_dir, *work_tree;
3160
3161 /* Prepare .git file */
3162 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3163 if (safe_create_leading_directories_const(gitfile_sb.buf))
3164 die(_("could not create directories for %s"), gitfile_sb.buf);
3165
3166 /* Prepare config file */
3167 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3168 if (safe_create_leading_directories_const(cfg_sb.buf))
3169 die(_("could not create directories for %s"), cfg_sb.buf);
3170
3171 git_dir = real_pathdup(git_dir_, 1);
3172 work_tree = real_pathdup(work_tree_, 1);
3173
3174 /* Write .git file */
3175 write_file(gitfile_sb.buf, "gitdir: %s",
3176 relative_path(git_dir, work_tree, &rel_path));
3177 /* Update core.worktree setting */
3178 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3179 relative_path(work_tree, git_dir, &rel_path));
3180
3181 strbuf_release(&gitfile_sb);
3182 strbuf_release(&cfg_sb);
3183 strbuf_release(&rel_path);
3184
3185 if (recurse_into_nested)
3186 connect_wt_gitdir_in_nested(work_tree, git_dir);
3187
3188 free(work_tree);
3189 free(git_dir);
3190 }
3191
3192 /*
3193 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3194 */
3195 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3196 {
3197 if (rename(old_git_dir, new_git_dir) < 0)
3198 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3199 old_git_dir, new_git_dir);
3200
3201 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3202 }