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