promisor-remote: skip move_to_tail when no-op
[git/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * See Documentation/technical/api-directory-listing.txt
6 *
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10 #include "cache.h"
11 #include "config.h"
12 #include "dir.h"
13 #include "object-store.h"
14 #include "attr.h"
15 #include "refs.h"
16 #include "wildmatch.h"
17 #include "pathspec.h"
18 #include "utf8.h"
19 #include "varint.h"
20 #include "ewah/ewok.h"
21 #include "fsmonitor.h"
22 #include "submodule-config.h"
23
24 /*
25 * Tells read_directory_recursive how a file or directory should be treated.
26 * Values are ordered by significance, e.g. if a directory contains both
27 * excluded and untracked files, it is listed as untracked because
28 * path_untracked > path_excluded.
29 */
30 enum path_treatment {
31 path_none = 0,
32 path_recurse,
33 path_excluded,
34 path_untracked
35 };
36
37 /*
38 * Support data structure for our opendir/readdir/closedir wrappers
39 */
40 struct cached_dir {
41 DIR *fdir;
42 struct untracked_cache_dir *untracked;
43 int nr_files;
44 int nr_dirs;
45
46 struct dirent *de;
47 const char *file;
48 struct untracked_cache_dir *ucd;
49 };
50
51 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
52 struct index_state *istate, const char *path, int len,
53 struct untracked_cache_dir *untracked,
54 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
55 static int get_dtype(struct dirent *de, struct index_state *istate,
56 const char *path, int len);
57
58 int count_slashes(const char *s)
59 {
60 int cnt = 0;
61 while (*s)
62 if (*s++ == '/')
63 cnt++;
64 return cnt;
65 }
66
67 int fspathcmp(const char *a, const char *b)
68 {
69 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
70 }
71
72 int fspathncmp(const char *a, const char *b, size_t count)
73 {
74 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
75 }
76
77 int git_fnmatch(const struct pathspec_item *item,
78 const char *pattern, const char *string,
79 int prefix)
80 {
81 if (prefix > 0) {
82 if (ps_strncmp(item, pattern, string, prefix))
83 return WM_NOMATCH;
84 pattern += prefix;
85 string += prefix;
86 }
87 if (item->flags & PATHSPEC_ONESTAR) {
88 int pattern_len = strlen(++pattern);
89 int string_len = strlen(string);
90 return string_len < pattern_len ||
91 ps_strcmp(item, pattern,
92 string + string_len - pattern_len);
93 }
94 if (item->magic & PATHSPEC_GLOB)
95 return wildmatch(pattern, string,
96 WM_PATHNAME |
97 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
98 else
99 /* wildmatch has not learned no FNM_PATHNAME mode yet */
100 return wildmatch(pattern, string,
101 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
102 }
103
104 static int fnmatch_icase_mem(const char *pattern, int patternlen,
105 const char *string, int stringlen,
106 int flags)
107 {
108 int match_status;
109 struct strbuf pat_buf = STRBUF_INIT;
110 struct strbuf str_buf = STRBUF_INIT;
111 const char *use_pat = pattern;
112 const char *use_str = string;
113
114 if (pattern[patternlen]) {
115 strbuf_add(&pat_buf, pattern, patternlen);
116 use_pat = pat_buf.buf;
117 }
118 if (string[stringlen]) {
119 strbuf_add(&str_buf, string, stringlen);
120 use_str = str_buf.buf;
121 }
122
123 if (ignore_case)
124 flags |= WM_CASEFOLD;
125 match_status = wildmatch(use_pat, use_str, flags);
126
127 strbuf_release(&pat_buf);
128 strbuf_release(&str_buf);
129
130 return match_status;
131 }
132
133 static size_t common_prefix_len(const struct pathspec *pathspec)
134 {
135 int n;
136 size_t max = 0;
137
138 /*
139 * ":(icase)path" is treated as a pathspec full of
140 * wildcard. In other words, only prefix is considered common
141 * prefix. If the pathspec is abc/foo abc/bar, running in
142 * subdir xyz, the common prefix is still xyz, not xuz/abc as
143 * in non-:(icase).
144 */
145 GUARD_PATHSPEC(pathspec,
146 PATHSPEC_FROMTOP |
147 PATHSPEC_MAXDEPTH |
148 PATHSPEC_LITERAL |
149 PATHSPEC_GLOB |
150 PATHSPEC_ICASE |
151 PATHSPEC_EXCLUDE |
152 PATHSPEC_ATTR);
153
154 for (n = 0; n < pathspec->nr; n++) {
155 size_t i = 0, len = 0, item_len;
156 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
157 continue;
158 if (pathspec->items[n].magic & PATHSPEC_ICASE)
159 item_len = pathspec->items[n].prefix;
160 else
161 item_len = pathspec->items[n].nowildcard_len;
162 while (i < item_len && (n == 0 || i < max)) {
163 char c = pathspec->items[n].match[i];
164 if (c != pathspec->items[0].match[i])
165 break;
166 if (c == '/')
167 len = i + 1;
168 i++;
169 }
170 if (n == 0 || len < max) {
171 max = len;
172 if (!max)
173 break;
174 }
175 }
176 return max;
177 }
178
179 /*
180 * Returns a copy of the longest leading path common among all
181 * pathspecs.
182 */
183 char *common_prefix(const struct pathspec *pathspec)
184 {
185 unsigned long len = common_prefix_len(pathspec);
186
187 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
188 }
189
190 int fill_directory(struct dir_struct *dir,
191 struct index_state *istate,
192 const struct pathspec *pathspec)
193 {
194 const char *prefix;
195 size_t prefix_len;
196
197 /*
198 * Calculate common prefix for the pathspec, and
199 * use that to optimize the directory walk
200 */
201 prefix_len = common_prefix_len(pathspec);
202 prefix = prefix_len ? pathspec->items[0].match : "";
203
204 /* Read the directory and prune it */
205 read_directory(dir, istate, prefix, prefix_len, pathspec);
206
207 return prefix_len;
208 }
209
210 int within_depth(const char *name, int namelen,
211 int depth, int max_depth)
212 {
213 const char *cp = name, *cpe = name + namelen;
214
215 while (cp < cpe) {
216 if (*cp++ != '/')
217 continue;
218 depth++;
219 if (depth > max_depth)
220 return 0;
221 }
222 return 1;
223 }
224
225 /*
226 * Read the contents of the blob with the given OID into a buffer.
227 * Append a trailing LF to the end if the last line doesn't have one.
228 *
229 * Returns:
230 * -1 when the OID is invalid or unknown or does not refer to a blob.
231 * 0 when the blob is empty.
232 * 1 along with { data, size } of the (possibly augmented) buffer
233 * when successful.
234 *
235 * Optionally updates the given oid_stat with the given OID (when valid).
236 */
237 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
238 size_t *size_out, char **data_out)
239 {
240 enum object_type type;
241 unsigned long sz;
242 char *data;
243
244 *size_out = 0;
245 *data_out = NULL;
246
247 data = read_object_file(oid, &type, &sz);
248 if (!data || type != OBJ_BLOB) {
249 free(data);
250 return -1;
251 }
252
253 if (oid_stat) {
254 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
255 oidcpy(&oid_stat->oid, oid);
256 }
257
258 if (sz == 0) {
259 free(data);
260 return 0;
261 }
262
263 if (data[sz - 1] != '\n') {
264 data = xrealloc(data, st_add(sz, 1));
265 data[sz++] = '\n';
266 }
267
268 *size_out = xsize_t(sz);
269 *data_out = data;
270
271 return 1;
272 }
273
274 #define DO_MATCH_EXCLUDE (1<<0)
275 #define DO_MATCH_DIRECTORY (1<<1)
276 #define DO_MATCH_SUBMODULE (1<<2)
277
278 /*
279 * Does 'match' match the given name?
280 * A match is found if
281 *
282 * (1) the 'match' string is leading directory of 'name', or
283 * (2) the 'match' string is a wildcard and matches 'name', or
284 * (3) the 'match' string is exactly the same as 'name'.
285 *
286 * and the return value tells which case it was.
287 *
288 * It returns 0 when there is no match.
289 */
290 static int match_pathspec_item(const struct index_state *istate,
291 const struct pathspec_item *item, int prefix,
292 const char *name, int namelen, unsigned flags)
293 {
294 /* name/namelen has prefix cut off by caller */
295 const char *match = item->match + prefix;
296 int matchlen = item->len - prefix;
297
298 /*
299 * The normal call pattern is:
300 * 1. prefix = common_prefix_len(ps);
301 * 2. prune something, or fill_directory
302 * 3. match_pathspec()
303 *
304 * 'prefix' at #1 may be shorter than the command's prefix and
305 * it's ok for #2 to match extra files. Those extras will be
306 * trimmed at #3.
307 *
308 * Suppose the pathspec is 'foo' and '../bar' running from
309 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
310 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
311 * user does not want XYZ/foo, only the "foo" part should be
312 * case-insensitive. We need to filter out XYZ/foo here. In
313 * other words, we do not trust the caller on comparing the
314 * prefix part when :(icase) is involved. We do exact
315 * comparison ourselves.
316 *
317 * Normally the caller (common_prefix_len() in fact) does
318 * _exact_ matching on name[-prefix+1..-1] and we do not need
319 * to check that part. Be defensive and check it anyway, in
320 * case common_prefix_len is changed, or a new caller is
321 * introduced that does not use common_prefix_len.
322 *
323 * If the penalty turns out too high when prefix is really
324 * long, maybe change it to
325 * strncmp(match, name, item->prefix - prefix)
326 */
327 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
328 strncmp(item->match, name - prefix, item->prefix))
329 return 0;
330
331 if (item->attr_match_nr &&
332 !match_pathspec_attrs(istate, name, namelen, item))
333 return 0;
334
335 /* If the match was just the prefix, we matched */
336 if (!*match)
337 return MATCHED_RECURSIVELY;
338
339 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
340 if (matchlen == namelen)
341 return MATCHED_EXACTLY;
342
343 if (match[matchlen-1] == '/' || name[matchlen] == '/')
344 return MATCHED_RECURSIVELY;
345 } else if ((flags & DO_MATCH_DIRECTORY) &&
346 match[matchlen - 1] == '/' &&
347 namelen == matchlen - 1 &&
348 !ps_strncmp(item, match, name, namelen))
349 return MATCHED_EXACTLY;
350
351 if (item->nowildcard_len < item->len &&
352 !git_fnmatch(item, match, name,
353 item->nowildcard_len - prefix))
354 return MATCHED_FNMATCH;
355
356 /* Perform checks to see if "name" is a super set of the pathspec */
357 if (flags & DO_MATCH_SUBMODULE) {
358 /* name is a literal prefix of the pathspec */
359 if ((namelen < matchlen) &&
360 (match[namelen] == '/') &&
361 !ps_strncmp(item, match, name, namelen))
362 return MATCHED_RECURSIVELY;
363
364 /* name" doesn't match up to the first wild character */
365 if (item->nowildcard_len < item->len &&
366 ps_strncmp(item, match, name,
367 item->nowildcard_len - prefix))
368 return 0;
369
370 /*
371 * Here is where we would perform a wildmatch to check if
372 * "name" can be matched as a directory (or a prefix) against
373 * the pathspec. Since wildmatch doesn't have this capability
374 * at the present we have to punt and say that it is a match,
375 * potentially returning a false positive
376 * The submodules themselves will be able to perform more
377 * accurate matching to determine if the pathspec matches.
378 */
379 return MATCHED_RECURSIVELY;
380 }
381
382 return 0;
383 }
384
385 /*
386 * Given a name and a list of pathspecs, returns the nature of the
387 * closest (i.e. most specific) match of the name to any of the
388 * pathspecs.
389 *
390 * The caller typically calls this multiple times with the same
391 * pathspec and seen[] array but with different name/namelen
392 * (e.g. entries from the index) and is interested in seeing if and
393 * how each pathspec matches all the names it calls this function
394 * with. A mark is left in the seen[] array for each pathspec element
395 * indicating the closest type of match that element achieved, so if
396 * seen[n] remains zero after multiple invocations, that means the nth
397 * pathspec did not match any names, which could indicate that the
398 * user mistyped the nth pathspec.
399 */
400 static int do_match_pathspec(const struct index_state *istate,
401 const struct pathspec *ps,
402 const char *name, int namelen,
403 int prefix, char *seen,
404 unsigned flags)
405 {
406 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
407
408 GUARD_PATHSPEC(ps,
409 PATHSPEC_FROMTOP |
410 PATHSPEC_MAXDEPTH |
411 PATHSPEC_LITERAL |
412 PATHSPEC_GLOB |
413 PATHSPEC_ICASE |
414 PATHSPEC_EXCLUDE |
415 PATHSPEC_ATTR);
416
417 if (!ps->nr) {
418 if (!ps->recursive ||
419 !(ps->magic & PATHSPEC_MAXDEPTH) ||
420 ps->max_depth == -1)
421 return MATCHED_RECURSIVELY;
422
423 if (within_depth(name, namelen, 0, ps->max_depth))
424 return MATCHED_EXACTLY;
425 else
426 return 0;
427 }
428
429 name += prefix;
430 namelen -= prefix;
431
432 for (i = ps->nr - 1; i >= 0; i--) {
433 int how;
434
435 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
436 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
437 continue;
438
439 if (seen && seen[i] == MATCHED_EXACTLY)
440 continue;
441 /*
442 * Make exclude patterns optional and never report
443 * "pathspec ':(exclude)foo' matches no files"
444 */
445 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
446 seen[i] = MATCHED_FNMATCH;
447 how = match_pathspec_item(istate, ps->items+i, prefix, name,
448 namelen, flags);
449 if (ps->recursive &&
450 (ps->magic & PATHSPEC_MAXDEPTH) &&
451 ps->max_depth != -1 &&
452 how && how != MATCHED_FNMATCH) {
453 int len = ps->items[i].len;
454 if (name[len] == '/')
455 len++;
456 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
457 how = MATCHED_EXACTLY;
458 else
459 how = 0;
460 }
461 if (how) {
462 if (retval < how)
463 retval = how;
464 if (seen && seen[i] < how)
465 seen[i] = how;
466 }
467 }
468 return retval;
469 }
470
471 int match_pathspec(const struct index_state *istate,
472 const struct pathspec *ps,
473 const char *name, int namelen,
474 int prefix, char *seen, int is_dir)
475 {
476 int positive, negative;
477 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
478 positive = do_match_pathspec(istate, ps, name, namelen,
479 prefix, seen, flags);
480 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
481 return positive;
482 negative = do_match_pathspec(istate, ps, name, namelen,
483 prefix, seen,
484 flags | DO_MATCH_EXCLUDE);
485 return negative ? 0 : positive;
486 }
487
488 /**
489 * Check if a submodule is a superset of the pathspec
490 */
491 int submodule_path_match(const struct index_state *istate,
492 const struct pathspec *ps,
493 const char *submodule_name,
494 char *seen)
495 {
496 int matched = do_match_pathspec(istate, ps, submodule_name,
497 strlen(submodule_name),
498 0, seen,
499 DO_MATCH_DIRECTORY |
500 DO_MATCH_SUBMODULE);
501 return matched;
502 }
503
504 int report_path_error(const char *ps_matched,
505 const struct pathspec *pathspec)
506 {
507 /*
508 * Make sure all pathspec matched; otherwise it is an error.
509 */
510 int num, errors = 0;
511 for (num = 0; num < pathspec->nr; num++) {
512 int other, found_dup;
513
514 if (ps_matched[num])
515 continue;
516 /*
517 * The caller might have fed identical pathspec
518 * twice. Do not barf on such a mistake.
519 * FIXME: parse_pathspec should have eliminated
520 * duplicate pathspec.
521 */
522 for (found_dup = other = 0;
523 !found_dup && other < pathspec->nr;
524 other++) {
525 if (other == num || !ps_matched[other])
526 continue;
527 if (!strcmp(pathspec->items[other].original,
528 pathspec->items[num].original))
529 /*
530 * Ok, we have a match already.
531 */
532 found_dup = 1;
533 }
534 if (found_dup)
535 continue;
536
537 error(_("pathspec '%s' did not match any file(s) known to git"),
538 pathspec->items[num].original);
539 errors++;
540 }
541 return errors;
542 }
543
544 /*
545 * Return the length of the "simple" part of a path match limiter.
546 */
547 int simple_length(const char *match)
548 {
549 int len = -1;
550
551 for (;;) {
552 unsigned char c = *match++;
553 len++;
554 if (c == '\0' || is_glob_special(c))
555 return len;
556 }
557 }
558
559 int no_wildcard(const char *string)
560 {
561 return string[simple_length(string)] == '\0';
562 }
563
564 void parse_exclude_pattern(const char **pattern,
565 int *patternlen,
566 unsigned *flags,
567 int *nowildcardlen)
568 {
569 const char *p = *pattern;
570 size_t i, len;
571
572 *flags = 0;
573 if (*p == '!') {
574 *flags |= EXC_FLAG_NEGATIVE;
575 p++;
576 }
577 len = strlen(p);
578 if (len && p[len - 1] == '/') {
579 len--;
580 *flags |= EXC_FLAG_MUSTBEDIR;
581 }
582 for (i = 0; i < len; i++) {
583 if (p[i] == '/')
584 break;
585 }
586 if (i == len)
587 *flags |= EXC_FLAG_NODIR;
588 *nowildcardlen = simple_length(p);
589 /*
590 * we should have excluded the trailing slash from 'p' too,
591 * but that's one more allocation. Instead just make sure
592 * nowildcardlen does not exceed real patternlen
593 */
594 if (*nowildcardlen > len)
595 *nowildcardlen = len;
596 if (*p == '*' && no_wildcard(p + 1))
597 *flags |= EXC_FLAG_ENDSWITH;
598 *pattern = p;
599 *patternlen = len;
600 }
601
602 void add_exclude(const char *string, const char *base,
603 int baselen, struct exclude_list *el, int srcpos)
604 {
605 struct exclude *x;
606 int patternlen;
607 unsigned flags;
608 int nowildcardlen;
609
610 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
611 if (flags & EXC_FLAG_MUSTBEDIR) {
612 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
613 } else {
614 x = xmalloc(sizeof(*x));
615 x->pattern = string;
616 }
617 x->patternlen = patternlen;
618 x->nowildcardlen = nowildcardlen;
619 x->base = base;
620 x->baselen = baselen;
621 x->flags = flags;
622 x->srcpos = srcpos;
623 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
624 el->excludes[el->nr++] = x;
625 x->el = el;
626 }
627
628 static int read_skip_worktree_file_from_index(const struct index_state *istate,
629 const char *path,
630 size_t *size_out, char **data_out,
631 struct oid_stat *oid_stat)
632 {
633 int pos, len;
634
635 len = strlen(path);
636 pos = index_name_pos(istate, path, len);
637 if (pos < 0)
638 return -1;
639 if (!ce_skip_worktree(istate->cache[pos]))
640 return -1;
641
642 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
643 }
644
645 /*
646 * Frees memory within el which was allocated for exclude patterns and
647 * the file buffer. Does not free el itself.
648 */
649 void clear_exclude_list(struct exclude_list *el)
650 {
651 int i;
652
653 for (i = 0; i < el->nr; i++)
654 free(el->excludes[i]);
655 free(el->excludes);
656 free(el->filebuf);
657
658 memset(el, 0, sizeof(*el));
659 }
660
661 static void trim_trailing_spaces(char *buf)
662 {
663 char *p, *last_space = NULL;
664
665 for (p = buf; *p; p++)
666 switch (*p) {
667 case ' ':
668 if (!last_space)
669 last_space = p;
670 break;
671 case '\\':
672 p++;
673 if (!*p)
674 return;
675 /* fallthrough */
676 default:
677 last_space = NULL;
678 }
679
680 if (last_space)
681 *last_space = '\0';
682 }
683
684 /*
685 * Given a subdirectory name and "dir" of the current directory,
686 * search the subdir in "dir" and return it, or create a new one if it
687 * does not exist in "dir".
688 *
689 * If "name" has the trailing slash, it'll be excluded in the search.
690 */
691 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
692 struct untracked_cache_dir *dir,
693 const char *name, int len)
694 {
695 int first, last;
696 struct untracked_cache_dir *d;
697 if (!dir)
698 return NULL;
699 if (len && name[len - 1] == '/')
700 len--;
701 first = 0;
702 last = dir->dirs_nr;
703 while (last > first) {
704 int cmp, next = (last + first) >> 1;
705 d = dir->dirs[next];
706 cmp = strncmp(name, d->name, len);
707 if (!cmp && strlen(d->name) > len)
708 cmp = -1;
709 if (!cmp)
710 return d;
711 if (cmp < 0) {
712 last = next;
713 continue;
714 }
715 first = next+1;
716 }
717
718 uc->dir_created++;
719 FLEX_ALLOC_MEM(d, name, name, len);
720
721 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
722 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
723 dir->dirs_nr - first);
724 dir->dirs_nr++;
725 dir->dirs[first] = d;
726 return d;
727 }
728
729 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
730 {
731 int i;
732 dir->valid = 0;
733 dir->untracked_nr = 0;
734 for (i = 0; i < dir->dirs_nr; i++)
735 do_invalidate_gitignore(dir->dirs[i]);
736 }
737
738 static void invalidate_gitignore(struct untracked_cache *uc,
739 struct untracked_cache_dir *dir)
740 {
741 uc->gitignore_invalidated++;
742 do_invalidate_gitignore(dir);
743 }
744
745 static void invalidate_directory(struct untracked_cache *uc,
746 struct untracked_cache_dir *dir)
747 {
748 int i;
749
750 /*
751 * Invalidation increment here is just roughly correct. If
752 * untracked_nr or any of dirs[].recurse is non-zero, we
753 * should increment dir_invalidated too. But that's more
754 * expensive to do.
755 */
756 if (dir->valid)
757 uc->dir_invalidated++;
758
759 dir->valid = 0;
760 dir->untracked_nr = 0;
761 for (i = 0; i < dir->dirs_nr; i++)
762 dir->dirs[i]->recurse = 0;
763 }
764
765 static int add_excludes_from_buffer(char *buf, size_t size,
766 const char *base, int baselen,
767 struct exclude_list *el);
768
769 /*
770 * Given a file with name "fname", read it (either from disk, or from
771 * an index if 'istate' is non-null), parse it and store the
772 * exclude rules in "el".
773 *
774 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
775 * stat data from disk (only valid if add_excludes returns zero). If
776 * ss_valid is non-zero, "ss" must contain good value as input.
777 */
778 static int add_excludes(const char *fname, const char *base, int baselen,
779 struct exclude_list *el, struct index_state *istate,
780 struct oid_stat *oid_stat)
781 {
782 struct stat st;
783 int r;
784 int fd;
785 size_t size = 0;
786 char *buf;
787
788 fd = open(fname, O_RDONLY);
789 if (fd < 0 || fstat(fd, &st) < 0) {
790 if (fd < 0)
791 warn_on_fopen_errors(fname);
792 else
793 close(fd);
794 if (!istate)
795 return -1;
796 r = read_skip_worktree_file_from_index(istate, fname,
797 &size, &buf,
798 oid_stat);
799 if (r != 1)
800 return r;
801 } else {
802 size = xsize_t(st.st_size);
803 if (size == 0) {
804 if (oid_stat) {
805 fill_stat_data(&oid_stat->stat, &st);
806 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
807 oid_stat->valid = 1;
808 }
809 close(fd);
810 return 0;
811 }
812 buf = xmallocz(size);
813 if (read_in_full(fd, buf, size) != size) {
814 free(buf);
815 close(fd);
816 return -1;
817 }
818 buf[size++] = '\n';
819 close(fd);
820 if (oid_stat) {
821 int pos;
822 if (oid_stat->valid &&
823 !match_stat_data_racy(istate, &oid_stat->stat, &st))
824 ; /* no content change, ss->sha1 still good */
825 else if (istate &&
826 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
827 !ce_stage(istate->cache[pos]) &&
828 ce_uptodate(istate->cache[pos]) &&
829 !would_convert_to_git(istate, fname))
830 oidcpy(&oid_stat->oid,
831 &istate->cache[pos]->oid);
832 else
833 hash_object_file(buf, size, "blob",
834 &oid_stat->oid);
835 fill_stat_data(&oid_stat->stat, &st);
836 oid_stat->valid = 1;
837 }
838 }
839
840 add_excludes_from_buffer(buf, size, base, baselen, el);
841 return 0;
842 }
843
844 static int add_excludes_from_buffer(char *buf, size_t size,
845 const char *base, int baselen,
846 struct exclude_list *el)
847 {
848 int i, lineno = 1;
849 char *entry;
850
851 el->filebuf = buf;
852
853 if (skip_utf8_bom(&buf, size))
854 size -= buf - el->filebuf;
855
856 entry = buf;
857
858 for (i = 0; i < size; i++) {
859 if (buf[i] == '\n') {
860 if (entry != buf + i && entry[0] != '#') {
861 buf[i - (i && buf[i-1] == '\r')] = 0;
862 trim_trailing_spaces(entry);
863 add_exclude(entry, base, baselen, el, lineno);
864 }
865 lineno++;
866 entry = buf + i + 1;
867 }
868 }
869 return 0;
870 }
871
872 int add_excludes_from_file_to_list(const char *fname, const char *base,
873 int baselen, struct exclude_list *el,
874 struct index_state *istate)
875 {
876 return add_excludes(fname, base, baselen, el, istate, NULL);
877 }
878
879 int add_excludes_from_blob_to_list(
880 struct object_id *oid,
881 const char *base, int baselen,
882 struct exclude_list *el)
883 {
884 char *buf;
885 size_t size;
886 int r;
887
888 r = do_read_blob(oid, NULL, &size, &buf);
889 if (r != 1)
890 return r;
891
892 add_excludes_from_buffer(buf, size, base, baselen, el);
893 return 0;
894 }
895
896 struct exclude_list *add_exclude_list(struct dir_struct *dir,
897 int group_type, const char *src)
898 {
899 struct exclude_list *el;
900 struct exclude_list_group *group;
901
902 group = &dir->exclude_list_group[group_type];
903 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
904 el = &group->el[group->nr++];
905 memset(el, 0, sizeof(*el));
906 el->src = src;
907 return el;
908 }
909
910 /*
911 * Used to set up core.excludesfile and .git/info/exclude lists.
912 */
913 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
914 struct oid_stat *oid_stat)
915 {
916 struct exclude_list *el;
917 /*
918 * catch setup_standard_excludes() that's called before
919 * dir->untracked is assigned. That function behaves
920 * differently when dir->untracked is non-NULL.
921 */
922 if (!dir->untracked)
923 dir->unmanaged_exclude_files++;
924 el = add_exclude_list(dir, EXC_FILE, fname);
925 if (add_excludes(fname, "", 0, el, NULL, oid_stat) < 0)
926 die(_("cannot use %s as an exclude file"), fname);
927 }
928
929 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
930 {
931 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
932 add_excludes_from_file_1(dir, fname, NULL);
933 }
934
935 int match_basename(const char *basename, int basenamelen,
936 const char *pattern, int prefix, int patternlen,
937 unsigned flags)
938 {
939 if (prefix == patternlen) {
940 if (patternlen == basenamelen &&
941 !fspathncmp(pattern, basename, basenamelen))
942 return 1;
943 } else if (flags & EXC_FLAG_ENDSWITH) {
944 /* "*literal" matching against "fooliteral" */
945 if (patternlen - 1 <= basenamelen &&
946 !fspathncmp(pattern + 1,
947 basename + basenamelen - (patternlen - 1),
948 patternlen - 1))
949 return 1;
950 } else {
951 if (fnmatch_icase_mem(pattern, patternlen,
952 basename, basenamelen,
953 0) == 0)
954 return 1;
955 }
956 return 0;
957 }
958
959 int match_pathname(const char *pathname, int pathlen,
960 const char *base, int baselen,
961 const char *pattern, int prefix, int patternlen,
962 unsigned flags)
963 {
964 const char *name;
965 int namelen;
966
967 /*
968 * match with FNM_PATHNAME; the pattern has base implicitly
969 * in front of it.
970 */
971 if (*pattern == '/') {
972 pattern++;
973 patternlen--;
974 prefix--;
975 }
976
977 /*
978 * baselen does not count the trailing slash. base[] may or
979 * may not end with a trailing slash though.
980 */
981 if (pathlen < baselen + 1 ||
982 (baselen && pathname[baselen] != '/') ||
983 fspathncmp(pathname, base, baselen))
984 return 0;
985
986 namelen = baselen ? pathlen - baselen - 1 : pathlen;
987 name = pathname + pathlen - namelen;
988
989 if (prefix) {
990 /*
991 * if the non-wildcard part is longer than the
992 * remaining pathname, surely it cannot match.
993 */
994 if (prefix > namelen)
995 return 0;
996
997 if (fspathncmp(pattern, name, prefix))
998 return 0;
999 pattern += prefix;
1000 patternlen -= prefix;
1001 name += prefix;
1002 namelen -= prefix;
1003
1004 /*
1005 * If the whole pattern did not have a wildcard,
1006 * then our prefix match is all we need; we
1007 * do not need to call fnmatch at all.
1008 */
1009 if (!patternlen && !namelen)
1010 return 1;
1011 }
1012
1013 return fnmatch_icase_mem(pattern, patternlen,
1014 name, namelen,
1015 WM_PATHNAME) == 0;
1016 }
1017
1018 /*
1019 * Scan the given exclude list in reverse to see whether pathname
1020 * should be ignored. The first match (i.e. the last on the list), if
1021 * any, determines the fate. Returns the exclude_list element which
1022 * matched, or NULL for undecided.
1023 */
1024 static struct exclude *last_exclude_matching_from_list(const char *pathname,
1025 int pathlen,
1026 const char *basename,
1027 int *dtype,
1028 struct exclude_list *el,
1029 struct index_state *istate)
1030 {
1031 struct exclude *exc = NULL; /* undecided */
1032 int i;
1033
1034 if (!el->nr)
1035 return NULL; /* undefined */
1036
1037 for (i = el->nr - 1; 0 <= i; i--) {
1038 struct exclude *x = el->excludes[i];
1039 const char *exclude = x->pattern;
1040 int prefix = x->nowildcardlen;
1041
1042 if (x->flags & EXC_FLAG_MUSTBEDIR) {
1043 if (*dtype == DT_UNKNOWN)
1044 *dtype = get_dtype(NULL, istate, pathname, pathlen);
1045 if (*dtype != DT_DIR)
1046 continue;
1047 }
1048
1049 if (x->flags & EXC_FLAG_NODIR) {
1050 if (match_basename(basename,
1051 pathlen - (basename - pathname),
1052 exclude, prefix, x->patternlen,
1053 x->flags)) {
1054 exc = x;
1055 break;
1056 }
1057 continue;
1058 }
1059
1060 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1061 if (match_pathname(pathname, pathlen,
1062 x->base, x->baselen ? x->baselen - 1 : 0,
1063 exclude, prefix, x->patternlen, x->flags)) {
1064 exc = x;
1065 break;
1066 }
1067 }
1068 return exc;
1069 }
1070
1071 /*
1072 * Scan the list and let the last match determine the fate.
1073 * Return 1 for exclude, 0 for include and -1 for undecided.
1074 */
1075 int is_excluded_from_list(const char *pathname,
1076 int pathlen, const char *basename, int *dtype,
1077 struct exclude_list *el, struct index_state *istate)
1078 {
1079 struct exclude *exclude;
1080 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1081 dtype, el, istate);
1082 if (exclude)
1083 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1084 return -1; /* undecided */
1085 }
1086
1087 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1088 struct index_state *istate,
1089 const char *pathname, int pathlen, const char *basename,
1090 int *dtype_p)
1091 {
1092 int i, j;
1093 struct exclude_list_group *group;
1094 struct exclude *exclude;
1095 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1096 group = &dir->exclude_list_group[i];
1097 for (j = group->nr - 1; j >= 0; j--) {
1098 exclude = last_exclude_matching_from_list(
1099 pathname, pathlen, basename, dtype_p,
1100 &group->el[j], istate);
1101 if (exclude)
1102 return exclude;
1103 }
1104 }
1105 return NULL;
1106 }
1107
1108 /*
1109 * Loads the per-directory exclude list for the substring of base
1110 * which has a char length of baselen.
1111 */
1112 static void prep_exclude(struct dir_struct *dir,
1113 struct index_state *istate,
1114 const char *base, int baselen)
1115 {
1116 struct exclude_list_group *group;
1117 struct exclude_list *el;
1118 struct exclude_stack *stk = NULL;
1119 struct untracked_cache_dir *untracked;
1120 int current;
1121
1122 group = &dir->exclude_list_group[EXC_DIRS];
1123
1124 /*
1125 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1126 * which originate from directories not in the prefix of the
1127 * path being checked.
1128 */
1129 while ((stk = dir->exclude_stack) != NULL) {
1130 if (stk->baselen <= baselen &&
1131 !strncmp(dir->basebuf.buf, base, stk->baselen))
1132 break;
1133 el = &group->el[dir->exclude_stack->exclude_ix];
1134 dir->exclude_stack = stk->prev;
1135 dir->exclude = NULL;
1136 free((char *)el->src); /* see strbuf_detach() below */
1137 clear_exclude_list(el);
1138 free(stk);
1139 group->nr--;
1140 }
1141
1142 /* Skip traversing into sub directories if the parent is excluded */
1143 if (dir->exclude)
1144 return;
1145
1146 /*
1147 * Lazy initialization. All call sites currently just
1148 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1149 * them seems lots of work for little benefit.
1150 */
1151 if (!dir->basebuf.buf)
1152 strbuf_init(&dir->basebuf, PATH_MAX);
1153
1154 /* Read from the parent directories and push them down. */
1155 current = stk ? stk->baselen : -1;
1156 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1157 if (dir->untracked)
1158 untracked = stk ? stk->ucd : dir->untracked->root;
1159 else
1160 untracked = NULL;
1161
1162 while (current < baselen) {
1163 const char *cp;
1164 struct oid_stat oid_stat;
1165
1166 stk = xcalloc(1, sizeof(*stk));
1167 if (current < 0) {
1168 cp = base;
1169 current = 0;
1170 } else {
1171 cp = strchr(base + current + 1, '/');
1172 if (!cp)
1173 die("oops in prep_exclude");
1174 cp++;
1175 untracked =
1176 lookup_untracked(dir->untracked, untracked,
1177 base + current,
1178 cp - base - current);
1179 }
1180 stk->prev = dir->exclude_stack;
1181 stk->baselen = cp - base;
1182 stk->exclude_ix = group->nr;
1183 stk->ucd = untracked;
1184 el = add_exclude_list(dir, EXC_DIRS, NULL);
1185 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1186 assert(stk->baselen == dir->basebuf.len);
1187
1188 /* Abort if the directory is excluded */
1189 if (stk->baselen) {
1190 int dt = DT_DIR;
1191 dir->basebuf.buf[stk->baselen - 1] = 0;
1192 dir->exclude = last_exclude_matching_from_lists(dir,
1193 istate,
1194 dir->basebuf.buf, stk->baselen - 1,
1195 dir->basebuf.buf + current, &dt);
1196 dir->basebuf.buf[stk->baselen - 1] = '/';
1197 if (dir->exclude &&
1198 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1199 dir->exclude = NULL;
1200 if (dir->exclude) {
1201 dir->exclude_stack = stk;
1202 return;
1203 }
1204 }
1205
1206 /* Try to read per-directory file */
1207 oidclr(&oid_stat.oid);
1208 oid_stat.valid = 0;
1209 if (dir->exclude_per_dir &&
1210 /*
1211 * If we know that no files have been added in
1212 * this directory (i.e. valid_cached_dir() has
1213 * been executed and set untracked->valid) ..
1214 */
1215 (!untracked || !untracked->valid ||
1216 /*
1217 * .. and .gitignore does not exist before
1218 * (i.e. null exclude_oid). Then we can skip
1219 * loading .gitignore, which would result in
1220 * ENOENT anyway.
1221 */
1222 !is_null_oid(&untracked->exclude_oid))) {
1223 /*
1224 * dir->basebuf gets reused by the traversal, but we
1225 * need fname to remain unchanged to ensure the src
1226 * member of each struct exclude correctly
1227 * back-references its source file. Other invocations
1228 * of add_exclude_list provide stable strings, so we
1229 * strbuf_detach() and free() here in the caller.
1230 */
1231 struct strbuf sb = STRBUF_INIT;
1232 strbuf_addbuf(&sb, &dir->basebuf);
1233 strbuf_addstr(&sb, dir->exclude_per_dir);
1234 el->src = strbuf_detach(&sb, NULL);
1235 add_excludes(el->src, el->src, stk->baselen, el, istate,
1236 untracked ? &oid_stat : NULL);
1237 }
1238 /*
1239 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1240 * will first be called in valid_cached_dir() then maybe many
1241 * times more in last_exclude_matching(). When the cache is
1242 * used, last_exclude_matching() will not be called and
1243 * reading .gitignore content will be a waste.
1244 *
1245 * So when it's called by valid_cached_dir() and we can get
1246 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1247 * modified on work tree), we could delay reading the
1248 * .gitignore content until we absolutely need it in
1249 * last_exclude_matching(). Be careful about ignore rule
1250 * order, though, if you do that.
1251 */
1252 if (untracked &&
1253 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1254 invalidate_gitignore(dir->untracked, untracked);
1255 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1256 }
1257 dir->exclude_stack = stk;
1258 current = stk->baselen;
1259 }
1260 strbuf_setlen(&dir->basebuf, baselen);
1261 }
1262
1263 /*
1264 * Loads the exclude lists for the directory containing pathname, then
1265 * scans all exclude lists to determine whether pathname is excluded.
1266 * Returns the exclude_list element which matched, or NULL for
1267 * undecided.
1268 */
1269 struct exclude *last_exclude_matching(struct dir_struct *dir,
1270 struct index_state *istate,
1271 const char *pathname,
1272 int *dtype_p)
1273 {
1274 int pathlen = strlen(pathname);
1275 const char *basename = strrchr(pathname, '/');
1276 basename = (basename) ? basename+1 : pathname;
1277
1278 prep_exclude(dir, istate, pathname, basename-pathname);
1279
1280 if (dir->exclude)
1281 return dir->exclude;
1282
1283 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
1284 basename, dtype_p);
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 1 if true, otherwise 0.
1291 */
1292 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1293 const char *pathname, int *dtype_p)
1294 {
1295 struct exclude *exclude =
1296 last_exclude_matching(dir, istate, pathname, dtype_p);
1297 if (exclude)
1298 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1299 return 0;
1300 }
1301
1302 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1303 {
1304 struct dir_entry *ent;
1305
1306 FLEX_ALLOC_MEM(ent, name, pathname, len);
1307 ent->len = len;
1308 return ent;
1309 }
1310
1311 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1312 struct index_state *istate,
1313 const char *pathname, int len)
1314 {
1315 if (index_file_exists(istate, pathname, len, ignore_case))
1316 return NULL;
1317
1318 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1319 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1320 }
1321
1322 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1323 struct index_state *istate,
1324 const char *pathname, int len)
1325 {
1326 if (!index_name_is_other(istate, pathname, len))
1327 return NULL;
1328
1329 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1330 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1331 }
1332
1333 enum exist_status {
1334 index_nonexistent = 0,
1335 index_directory,
1336 index_gitdir
1337 };
1338
1339 /*
1340 * Do not use the alphabetically sorted index to look up
1341 * the directory name; instead, use the case insensitive
1342 * directory hash.
1343 */
1344 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1345 const char *dirname, int len)
1346 {
1347 struct cache_entry *ce;
1348
1349 if (index_dir_exists(istate, dirname, len))
1350 return index_directory;
1351
1352 ce = index_file_exists(istate, dirname, len, ignore_case);
1353 if (ce && S_ISGITLINK(ce->ce_mode))
1354 return index_gitdir;
1355
1356 return index_nonexistent;
1357 }
1358
1359 /*
1360 * The index sorts alphabetically by entry name, which
1361 * means that a gitlink sorts as '\0' at the end, while
1362 * a directory (which is defined not as an entry, but as
1363 * the files it contains) will sort with the '/' at the
1364 * end.
1365 */
1366 static enum exist_status directory_exists_in_index(struct index_state *istate,
1367 const char *dirname, int len)
1368 {
1369 int pos;
1370
1371 if (ignore_case)
1372 return directory_exists_in_index_icase(istate, dirname, len);
1373
1374 pos = index_name_pos(istate, dirname, len);
1375 if (pos < 0)
1376 pos = -pos-1;
1377 while (pos < istate->cache_nr) {
1378 const struct cache_entry *ce = istate->cache[pos++];
1379 unsigned char endchar;
1380
1381 if (strncmp(ce->name, dirname, len))
1382 break;
1383 endchar = ce->name[len];
1384 if (endchar > '/')
1385 break;
1386 if (endchar == '/')
1387 return index_directory;
1388 if (!endchar && S_ISGITLINK(ce->ce_mode))
1389 return index_gitdir;
1390 }
1391 return index_nonexistent;
1392 }
1393
1394 /*
1395 * When we find a directory when traversing the filesystem, we
1396 * have three distinct cases:
1397 *
1398 * - ignore it
1399 * - see it as a directory
1400 * - recurse into it
1401 *
1402 * and which one we choose depends on a combination of existing
1403 * git index contents and the flags passed into the directory
1404 * traversal routine.
1405 *
1406 * Case 1: If we *already* have entries in the index under that
1407 * directory name, we always recurse into the directory to see
1408 * all the files.
1409 *
1410 * Case 2: If we *already* have that directory name as a gitlink,
1411 * we always continue to see it as a gitlink, regardless of whether
1412 * there is an actual git directory there or not (it might not
1413 * be checked out as a subproject!)
1414 *
1415 * Case 3: if we didn't have it in the index previously, we
1416 * have a few sub-cases:
1417 *
1418 * (a) if "show_other_directories" is true, we show it as
1419 * just a directory, unless "hide_empty_directories" is
1420 * also true, in which case we need to check if it contains any
1421 * untracked and / or ignored files.
1422 * (b) if it looks like a git directory, and we don't have
1423 * 'no_gitlinks' set we treat it as a gitlink, and show it
1424 * as a directory.
1425 * (c) otherwise, we recurse into it.
1426 */
1427 static enum path_treatment treat_directory(struct dir_struct *dir,
1428 struct index_state *istate,
1429 struct untracked_cache_dir *untracked,
1430 const char *dirname, int len, int baselen, int exclude,
1431 const struct pathspec *pathspec)
1432 {
1433 /* The "len-1" is to strip the final '/' */
1434 switch (directory_exists_in_index(istate, dirname, len-1)) {
1435 case index_directory:
1436 return path_recurse;
1437
1438 case index_gitdir:
1439 return path_none;
1440
1441 case index_nonexistent:
1442 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1443 break;
1444 if (exclude &&
1445 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1446 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1447
1448 /*
1449 * This is an excluded directory and we are
1450 * showing ignored paths that match an exclude
1451 * pattern. (e.g. show directory as ignored
1452 * only if it matches an exclude pattern).
1453 * This path will either be 'path_excluded`
1454 * (if we are showing empty directories or if
1455 * the directory is not empty), or will be
1456 * 'path_none' (empty directory, and we are
1457 * not showing empty directories).
1458 */
1459 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1460 return path_excluded;
1461
1462 if (read_directory_recursive(dir, istate, dirname, len,
1463 untracked, 1, 1, pathspec) == path_excluded)
1464 return path_excluded;
1465
1466 return path_none;
1467 }
1468 if (!(dir->flags & DIR_NO_GITLINKS)) {
1469 struct strbuf sb = STRBUF_INIT;
1470 strbuf_addstr(&sb, dirname);
1471 if (is_nonbare_repository_dir(&sb))
1472 return exclude ? path_excluded : path_untracked;
1473 strbuf_release(&sb);
1474 }
1475 return path_recurse;
1476 }
1477
1478 /* This is the "show_other_directories" case */
1479
1480 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1481 return exclude ? path_excluded : path_untracked;
1482
1483 untracked = lookup_untracked(dir->untracked, untracked,
1484 dirname + baselen, len - baselen);
1485
1486 /*
1487 * If this is an excluded directory, then we only need to check if
1488 * the directory contains any files.
1489 */
1490 return read_directory_recursive(dir, istate, dirname, len,
1491 untracked, 1, exclude, pathspec);
1492 }
1493
1494 /*
1495 * This is an inexact early pruning of any recursive directory
1496 * reading - if the path cannot possibly be in the pathspec,
1497 * return true, and we'll skip it early.
1498 */
1499 static int simplify_away(const char *path, int pathlen,
1500 const struct pathspec *pathspec)
1501 {
1502 int i;
1503
1504 if (!pathspec || !pathspec->nr)
1505 return 0;
1506
1507 GUARD_PATHSPEC(pathspec,
1508 PATHSPEC_FROMTOP |
1509 PATHSPEC_MAXDEPTH |
1510 PATHSPEC_LITERAL |
1511 PATHSPEC_GLOB |
1512 PATHSPEC_ICASE |
1513 PATHSPEC_EXCLUDE |
1514 PATHSPEC_ATTR);
1515
1516 for (i = 0; i < pathspec->nr; i++) {
1517 const struct pathspec_item *item = &pathspec->items[i];
1518 int len = item->nowildcard_len;
1519
1520 if (len > pathlen)
1521 len = pathlen;
1522 if (!ps_strncmp(item, item->match, path, len))
1523 return 0;
1524 }
1525
1526 return 1;
1527 }
1528
1529 /*
1530 * This function tells us whether an excluded path matches a
1531 * list of "interesting" pathspecs. That is, whether a path matched
1532 * by any of the pathspecs could possibly be ignored by excluding
1533 * the specified path. This can happen if:
1534 *
1535 * 1. the path is mentioned explicitly in the pathspec
1536 *
1537 * 2. the path is a directory prefix of some element in the
1538 * pathspec
1539 */
1540 static int exclude_matches_pathspec(const char *path, int pathlen,
1541 const struct pathspec *pathspec)
1542 {
1543 int i;
1544
1545 if (!pathspec || !pathspec->nr)
1546 return 0;
1547
1548 GUARD_PATHSPEC(pathspec,
1549 PATHSPEC_FROMTOP |
1550 PATHSPEC_MAXDEPTH |
1551 PATHSPEC_LITERAL |
1552 PATHSPEC_GLOB |
1553 PATHSPEC_ICASE |
1554 PATHSPEC_EXCLUDE);
1555
1556 for (i = 0; i < pathspec->nr; i++) {
1557 const struct pathspec_item *item = &pathspec->items[i];
1558 int len = item->nowildcard_len;
1559
1560 if (len == pathlen &&
1561 !ps_strncmp(item, item->match, path, pathlen))
1562 return 1;
1563 if (len > pathlen &&
1564 item->match[pathlen] == '/' &&
1565 !ps_strncmp(item, item->match, path, pathlen))
1566 return 1;
1567 }
1568 return 0;
1569 }
1570
1571 static int get_index_dtype(struct index_state *istate,
1572 const char *path, int len)
1573 {
1574 int pos;
1575 const struct cache_entry *ce;
1576
1577 ce = index_file_exists(istate, path, len, 0);
1578 if (ce) {
1579 if (!ce_uptodate(ce))
1580 return DT_UNKNOWN;
1581 if (S_ISGITLINK(ce->ce_mode))
1582 return DT_DIR;
1583 /*
1584 * Nobody actually cares about the
1585 * difference between DT_LNK and DT_REG
1586 */
1587 return DT_REG;
1588 }
1589
1590 /* Try to look it up as a directory */
1591 pos = index_name_pos(istate, path, len);
1592 if (pos >= 0)
1593 return DT_UNKNOWN;
1594 pos = -pos-1;
1595 while (pos < istate->cache_nr) {
1596 ce = istate->cache[pos++];
1597 if (strncmp(ce->name, path, len))
1598 break;
1599 if (ce->name[len] > '/')
1600 break;
1601 if (ce->name[len] < '/')
1602 continue;
1603 if (!ce_uptodate(ce))
1604 break; /* continue? */
1605 return DT_DIR;
1606 }
1607 return DT_UNKNOWN;
1608 }
1609
1610 static int get_dtype(struct dirent *de, struct index_state *istate,
1611 const char *path, int len)
1612 {
1613 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1614 struct stat st;
1615
1616 if (dtype != DT_UNKNOWN)
1617 return dtype;
1618 dtype = get_index_dtype(istate, path, len);
1619 if (dtype != DT_UNKNOWN)
1620 return dtype;
1621 if (lstat(path, &st))
1622 return dtype;
1623 if (S_ISREG(st.st_mode))
1624 return DT_REG;
1625 if (S_ISDIR(st.st_mode))
1626 return DT_DIR;
1627 if (S_ISLNK(st.st_mode))
1628 return DT_LNK;
1629 return dtype;
1630 }
1631
1632 static enum path_treatment treat_one_path(struct dir_struct *dir,
1633 struct untracked_cache_dir *untracked,
1634 struct index_state *istate,
1635 struct strbuf *path,
1636 int baselen,
1637 const struct pathspec *pathspec,
1638 int dtype, struct dirent *de)
1639 {
1640 int exclude;
1641 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1642 enum path_treatment path_treatment;
1643
1644 if (dtype == DT_UNKNOWN)
1645 dtype = get_dtype(de, istate, path->buf, path->len);
1646
1647 /* Always exclude indexed files */
1648 if (dtype != DT_DIR && has_path_in_index)
1649 return path_none;
1650
1651 /*
1652 * When we are looking at a directory P in the working tree,
1653 * there are three cases:
1654 *
1655 * (1) P exists in the index. Everything inside the directory P in
1656 * the working tree needs to go when P is checked out from the
1657 * index.
1658 *
1659 * (2) P does not exist in the index, but there is P/Q in the index.
1660 * We know P will stay a directory when we check out the contents
1661 * of the index, but we do not know yet if there is a directory
1662 * P/Q in the working tree to be killed, so we need to recurse.
1663 *
1664 * (3) P does not exist in the index, and there is no P/Q in the index
1665 * to require P to be a directory, either. Only in this case, we
1666 * know that everything inside P will not be killed without
1667 * recursing.
1668 */
1669 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1670 (dtype == DT_DIR) &&
1671 !has_path_in_index &&
1672 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1673 return path_none;
1674
1675 exclude = is_excluded(dir, istate, path->buf, &dtype);
1676
1677 /*
1678 * Excluded? If we don't explicitly want to show
1679 * ignored files, ignore it
1680 */
1681 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1682 return path_excluded;
1683
1684 switch (dtype) {
1685 default:
1686 return path_none;
1687 case DT_DIR:
1688 strbuf_addch(path, '/');
1689 path_treatment = treat_directory(dir, istate, untracked,
1690 path->buf, path->len,
1691 baselen, exclude, pathspec);
1692 /*
1693 * If 1) we only want to return directories that
1694 * match an exclude pattern and 2) this directory does
1695 * not match an exclude pattern but all of its
1696 * contents are excluded, then indicate that we should
1697 * recurse into this directory (instead of marking the
1698 * directory itself as an ignored path).
1699 */
1700 if (!exclude &&
1701 path_treatment == path_excluded &&
1702 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1703 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1704 return path_recurse;
1705 return path_treatment;
1706 case DT_REG:
1707 case DT_LNK:
1708 return exclude ? path_excluded : path_untracked;
1709 }
1710 }
1711
1712 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1713 struct untracked_cache_dir *untracked,
1714 struct cached_dir *cdir,
1715 struct index_state *istate,
1716 struct strbuf *path,
1717 int baselen,
1718 const struct pathspec *pathspec)
1719 {
1720 strbuf_setlen(path, baselen);
1721 if (!cdir->ucd) {
1722 strbuf_addstr(path, cdir->file);
1723 return path_untracked;
1724 }
1725 strbuf_addstr(path, cdir->ucd->name);
1726 /* treat_one_path() does this before it calls treat_directory() */
1727 strbuf_complete(path, '/');
1728 if (cdir->ucd->check_only)
1729 /*
1730 * check_only is set as a result of treat_directory() getting
1731 * to its bottom. Verify again the same set of directories
1732 * with check_only set.
1733 */
1734 return read_directory_recursive(dir, istate, path->buf, path->len,
1735 cdir->ucd, 1, 0, pathspec);
1736 /*
1737 * We get path_recurse in the first run when
1738 * directory_exists_in_index() returns index_nonexistent. We
1739 * are sure that new changes in the index does not impact the
1740 * outcome. Return now.
1741 */
1742 return path_recurse;
1743 }
1744
1745 static enum path_treatment treat_path(struct dir_struct *dir,
1746 struct untracked_cache_dir *untracked,
1747 struct cached_dir *cdir,
1748 struct index_state *istate,
1749 struct strbuf *path,
1750 int baselen,
1751 const struct pathspec *pathspec)
1752 {
1753 int dtype;
1754 struct dirent *de = cdir->de;
1755
1756 if (!de)
1757 return treat_path_fast(dir, untracked, cdir, istate, path,
1758 baselen, pathspec);
1759 if (is_dot_or_dotdot(de->d_name) || !fspathcmp(de->d_name, ".git"))
1760 return path_none;
1761 strbuf_setlen(path, baselen);
1762 strbuf_addstr(path, de->d_name);
1763 if (simplify_away(path->buf, path->len, pathspec))
1764 return path_none;
1765
1766 dtype = DTYPE(de);
1767 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1768 }
1769
1770 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1771 {
1772 if (!dir)
1773 return;
1774 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1775 dir->untracked_alloc);
1776 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1777 }
1778
1779 static int valid_cached_dir(struct dir_struct *dir,
1780 struct untracked_cache_dir *untracked,
1781 struct index_state *istate,
1782 struct strbuf *path,
1783 int check_only)
1784 {
1785 struct stat st;
1786
1787 if (!untracked)
1788 return 0;
1789
1790 /*
1791 * With fsmonitor, we can trust the untracked cache's valid field.
1792 */
1793 refresh_fsmonitor(istate);
1794 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1795 if (lstat(path->len ? path->buf : ".", &st)) {
1796 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1797 return 0;
1798 }
1799 if (!untracked->valid ||
1800 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1801 fill_stat_data(&untracked->stat_data, &st);
1802 return 0;
1803 }
1804 }
1805
1806 if (untracked->check_only != !!check_only)
1807 return 0;
1808
1809 /*
1810 * prep_exclude will be called eventually on this directory,
1811 * but it's called much later in last_exclude_matching(). We
1812 * need it now to determine the validity of the cache for this
1813 * path. The next calls will be nearly no-op, the way
1814 * prep_exclude() is designed.
1815 */
1816 if (path->len && path->buf[path->len - 1] != '/') {
1817 strbuf_addch(path, '/');
1818 prep_exclude(dir, istate, path->buf, path->len);
1819 strbuf_setlen(path, path->len - 1);
1820 } else
1821 prep_exclude(dir, istate, path->buf, path->len);
1822
1823 /* hopefully prep_exclude() haven't invalidated this entry... */
1824 return untracked->valid;
1825 }
1826
1827 static int open_cached_dir(struct cached_dir *cdir,
1828 struct dir_struct *dir,
1829 struct untracked_cache_dir *untracked,
1830 struct index_state *istate,
1831 struct strbuf *path,
1832 int check_only)
1833 {
1834 const char *c_path;
1835
1836 memset(cdir, 0, sizeof(*cdir));
1837 cdir->untracked = untracked;
1838 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1839 return 0;
1840 c_path = path->len ? path->buf : ".";
1841 cdir->fdir = opendir(c_path);
1842 if (!cdir->fdir)
1843 warning_errno(_("could not open directory '%s'"), c_path);
1844 if (dir->untracked) {
1845 invalidate_directory(dir->untracked, untracked);
1846 dir->untracked->dir_opened++;
1847 }
1848 if (!cdir->fdir)
1849 return -1;
1850 return 0;
1851 }
1852
1853 static int read_cached_dir(struct cached_dir *cdir)
1854 {
1855 if (cdir->fdir) {
1856 cdir->de = readdir(cdir->fdir);
1857 if (!cdir->de)
1858 return -1;
1859 return 0;
1860 }
1861 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1862 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1863 if (!d->recurse) {
1864 cdir->nr_dirs++;
1865 continue;
1866 }
1867 cdir->ucd = d;
1868 cdir->nr_dirs++;
1869 return 0;
1870 }
1871 cdir->ucd = NULL;
1872 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1873 struct untracked_cache_dir *d = cdir->untracked;
1874 cdir->file = d->untracked[cdir->nr_files++];
1875 return 0;
1876 }
1877 return -1;
1878 }
1879
1880 static void close_cached_dir(struct cached_dir *cdir)
1881 {
1882 if (cdir->fdir)
1883 closedir(cdir->fdir);
1884 /*
1885 * We have gone through this directory and found no untracked
1886 * entries. Mark it valid.
1887 */
1888 if (cdir->untracked) {
1889 cdir->untracked->valid = 1;
1890 cdir->untracked->recurse = 1;
1891 }
1892 }
1893
1894 /*
1895 * Read a directory tree. We currently ignore anything but
1896 * directories, regular files and symlinks. That's because git
1897 * doesn't handle them at all yet. Maybe that will change some
1898 * day.
1899 *
1900 * Also, we ignore the name ".git" (even if it is not a directory).
1901 * That likely will not change.
1902 *
1903 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1904 * to signal that a file was found. This is the least significant value that
1905 * indicates that a file was encountered that does not depend on the order of
1906 * whether an untracked or exluded path was encountered first.
1907 *
1908 * Returns the most significant path_treatment value encountered in the scan.
1909 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1910 * significant path_treatment value that will be returned.
1911 */
1912
1913 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1914 struct index_state *istate, const char *base, int baselen,
1915 struct untracked_cache_dir *untracked, int check_only,
1916 int stop_at_first_file, const struct pathspec *pathspec)
1917 {
1918 struct cached_dir cdir;
1919 enum path_treatment state, subdir_state, dir_state = path_none;
1920 struct strbuf path = STRBUF_INIT;
1921
1922 strbuf_add(&path, base, baselen);
1923
1924 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1925 goto out;
1926
1927 if (untracked)
1928 untracked->check_only = !!check_only;
1929
1930 while (!read_cached_dir(&cdir)) {
1931 /* check how the file or directory should be treated */
1932 state = treat_path(dir, untracked, &cdir, istate, &path,
1933 baselen, pathspec);
1934
1935 if (state > dir_state)
1936 dir_state = state;
1937
1938 /* recurse into subdir if instructed by treat_path */
1939 if ((state == path_recurse) ||
1940 ((state == path_untracked) &&
1941 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1942 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
1943 struct untracked_cache_dir *ud;
1944 ud = lookup_untracked(dir->untracked, untracked,
1945 path.buf + baselen,
1946 path.len - baselen);
1947 subdir_state =
1948 read_directory_recursive(dir, istate, path.buf,
1949 path.len, ud,
1950 check_only, stop_at_first_file, pathspec);
1951 if (subdir_state > dir_state)
1952 dir_state = subdir_state;
1953 }
1954
1955 if (check_only) {
1956 if (stop_at_first_file) {
1957 /*
1958 * If stopping at first file, then
1959 * signal that a file was found by
1960 * returning `path_excluded`. This is
1961 * to return a consistent value
1962 * regardless of whether an ignored or
1963 * excluded file happened to be
1964 * encountered 1st.
1965 *
1966 * In current usage, the
1967 * `stop_at_first_file` is passed when
1968 * an ancestor directory has matched
1969 * an exclude pattern, so any found
1970 * files will be excluded.
1971 */
1972 if (dir_state >= path_excluded) {
1973 dir_state = path_excluded;
1974 break;
1975 }
1976 }
1977
1978 /* abort early if maximum state has been reached */
1979 if (dir_state == path_untracked) {
1980 if (cdir.fdir)
1981 add_untracked(untracked, path.buf + baselen);
1982 break;
1983 }
1984 /* skip the dir_add_* part */
1985 continue;
1986 }
1987
1988 /* add the path to the appropriate result list */
1989 switch (state) {
1990 case path_excluded:
1991 if (dir->flags & DIR_SHOW_IGNORED)
1992 dir_add_name(dir, istate, path.buf, path.len);
1993 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1994 ((dir->flags & DIR_COLLECT_IGNORED) &&
1995 exclude_matches_pathspec(path.buf, path.len,
1996 pathspec)))
1997 dir_add_ignored(dir, istate, path.buf, path.len);
1998 break;
1999
2000 case path_untracked:
2001 if (dir->flags & DIR_SHOW_IGNORED)
2002 break;
2003 dir_add_name(dir, istate, path.buf, path.len);
2004 if (cdir.fdir)
2005 add_untracked(untracked, path.buf + baselen);
2006 break;
2007
2008 default:
2009 break;
2010 }
2011 }
2012 close_cached_dir(&cdir);
2013 out:
2014 strbuf_release(&path);
2015
2016 return dir_state;
2017 }
2018
2019 int cmp_dir_entry(const void *p1, const void *p2)
2020 {
2021 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2022 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2023
2024 return name_compare(e1->name, e1->len, e2->name, e2->len);
2025 }
2026
2027 /* check if *out lexically strictly contains *in */
2028 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2029 {
2030 return (out->len < in->len) &&
2031 (out->name[out->len - 1] == '/') &&
2032 !memcmp(out->name, in->name, out->len);
2033 }
2034
2035 static int treat_leading_path(struct dir_struct *dir,
2036 struct index_state *istate,
2037 const char *path, int len,
2038 const struct pathspec *pathspec)
2039 {
2040 struct strbuf sb = STRBUF_INIT;
2041 int baselen, rc = 0;
2042 const char *cp;
2043 int old_flags = dir->flags;
2044
2045 while (len && path[len - 1] == '/')
2046 len--;
2047 if (!len)
2048 return 1;
2049 baselen = 0;
2050 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
2051 while (1) {
2052 cp = path + baselen + !!baselen;
2053 cp = memchr(cp, '/', path + len - cp);
2054 if (!cp)
2055 baselen = len;
2056 else
2057 baselen = cp - path;
2058 strbuf_setlen(&sb, 0);
2059 strbuf_add(&sb, path, baselen);
2060 if (!is_directory(sb.buf))
2061 break;
2062 if (simplify_away(sb.buf, sb.len, pathspec))
2063 break;
2064 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
2065 DT_DIR, NULL) == path_none)
2066 break; /* do not recurse into it */
2067 if (len <= baselen) {
2068 rc = 1;
2069 break; /* finished checking */
2070 }
2071 }
2072 strbuf_release(&sb);
2073 dir->flags = old_flags;
2074 return rc;
2075 }
2076
2077 static const char *get_ident_string(void)
2078 {
2079 static struct strbuf sb = STRBUF_INIT;
2080 struct utsname uts;
2081
2082 if (sb.len)
2083 return sb.buf;
2084 if (uname(&uts) < 0)
2085 die_errno(_("failed to get kernel name and information"));
2086 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2087 uts.sysname);
2088 return sb.buf;
2089 }
2090
2091 static int ident_in_untracked(const struct untracked_cache *uc)
2092 {
2093 /*
2094 * Previous git versions may have saved many NUL separated
2095 * strings in the "ident" field, but it is insane to manage
2096 * many locations, so just take care of the first one.
2097 */
2098
2099 return !strcmp(uc->ident.buf, get_ident_string());
2100 }
2101
2102 static void set_untracked_ident(struct untracked_cache *uc)
2103 {
2104 strbuf_reset(&uc->ident);
2105 strbuf_addstr(&uc->ident, get_ident_string());
2106
2107 /*
2108 * This strbuf used to contain a list of NUL separated
2109 * strings, so save NUL too for backward compatibility.
2110 */
2111 strbuf_addch(&uc->ident, 0);
2112 }
2113
2114 static void new_untracked_cache(struct index_state *istate)
2115 {
2116 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2117 strbuf_init(&uc->ident, 100);
2118 uc->exclude_per_dir = ".gitignore";
2119 /* should be the same flags used by git-status */
2120 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2121 set_untracked_ident(uc);
2122 istate->untracked = uc;
2123 istate->cache_changed |= UNTRACKED_CHANGED;
2124 }
2125
2126 void add_untracked_cache(struct index_state *istate)
2127 {
2128 if (!istate->untracked) {
2129 new_untracked_cache(istate);
2130 } else {
2131 if (!ident_in_untracked(istate->untracked)) {
2132 free_untracked_cache(istate->untracked);
2133 new_untracked_cache(istate);
2134 }
2135 }
2136 }
2137
2138 void remove_untracked_cache(struct index_state *istate)
2139 {
2140 if (istate->untracked) {
2141 free_untracked_cache(istate->untracked);
2142 istate->untracked = NULL;
2143 istate->cache_changed |= UNTRACKED_CHANGED;
2144 }
2145 }
2146
2147 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2148 int base_len,
2149 const struct pathspec *pathspec)
2150 {
2151 struct untracked_cache_dir *root;
2152 static int untracked_cache_disabled = -1;
2153
2154 if (!dir->untracked)
2155 return NULL;
2156 if (untracked_cache_disabled < 0)
2157 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2158 if (untracked_cache_disabled)
2159 return NULL;
2160
2161 /*
2162 * We only support $GIT_DIR/info/exclude and core.excludesfile
2163 * as the global ignore rule files. Any other additions
2164 * (e.g. from command line) invalidate the cache. This
2165 * condition also catches running setup_standard_excludes()
2166 * before setting dir->untracked!
2167 */
2168 if (dir->unmanaged_exclude_files)
2169 return NULL;
2170
2171 /*
2172 * Optimize for the main use case only: whole-tree git
2173 * status. More work involved in treat_leading_path() if we
2174 * use cache on just a subset of the worktree. pathspec
2175 * support could make the matter even worse.
2176 */
2177 if (base_len || (pathspec && pathspec->nr))
2178 return NULL;
2179
2180 /* Different set of flags may produce different results */
2181 if (dir->flags != dir->untracked->dir_flags ||
2182 /*
2183 * See treat_directory(), case index_nonexistent. Without
2184 * this flag, we may need to also cache .git file content
2185 * for the resolve_gitlink_ref() call, which we don't.
2186 */
2187 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2188 /* We don't support collecting ignore files */
2189 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2190 DIR_COLLECT_IGNORED)))
2191 return NULL;
2192
2193 /*
2194 * If we use .gitignore in the cache and now you change it to
2195 * .gitexclude, everything will go wrong.
2196 */
2197 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2198 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2199 return NULL;
2200
2201 /*
2202 * EXC_CMDL is not considered in the cache. If people set it,
2203 * skip the cache.
2204 */
2205 if (dir->exclude_list_group[EXC_CMDL].nr)
2206 return NULL;
2207
2208 if (!ident_in_untracked(dir->untracked)) {
2209 warning(_("untracked cache is disabled on this system or location"));
2210 return NULL;
2211 }
2212
2213 if (!dir->untracked->root) {
2214 const int len = sizeof(*dir->untracked->root);
2215 dir->untracked->root = xmalloc(len);
2216 memset(dir->untracked->root, 0, len);
2217 }
2218
2219 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2220 root = dir->untracked->root;
2221 if (!oideq(&dir->ss_info_exclude.oid,
2222 &dir->untracked->ss_info_exclude.oid)) {
2223 invalidate_gitignore(dir->untracked, root);
2224 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2225 }
2226 if (!oideq(&dir->ss_excludes_file.oid,
2227 &dir->untracked->ss_excludes_file.oid)) {
2228 invalidate_gitignore(dir->untracked, root);
2229 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2230 }
2231
2232 /* Make sure this directory is not dropped out at saving phase */
2233 root->recurse = 1;
2234 return root;
2235 }
2236
2237 int read_directory(struct dir_struct *dir, struct index_state *istate,
2238 const char *path, int len, const struct pathspec *pathspec)
2239 {
2240 struct untracked_cache_dir *untracked;
2241
2242 trace_performance_enter();
2243
2244 if (has_symlink_leading_path(path, len)) {
2245 trace_performance_leave("read directory %.*s", len, path);
2246 return dir->nr;
2247 }
2248
2249 untracked = validate_untracked_cache(dir, len, pathspec);
2250 if (!untracked)
2251 /*
2252 * make sure untracked cache code path is disabled,
2253 * e.g. prep_exclude()
2254 */
2255 dir->untracked = NULL;
2256 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2257 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2258 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2259 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2260
2261 /*
2262 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2263 * also pick up untracked contents of untracked dirs; by default
2264 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2265 */
2266 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2267 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2268 int i, j;
2269
2270 /* remove from dir->entries untracked contents of untracked dirs */
2271 for (i = j = 0; j < dir->nr; j++) {
2272 if (i &&
2273 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2274 FREE_AND_NULL(dir->entries[j]);
2275 } else {
2276 dir->entries[i++] = dir->entries[j];
2277 }
2278 }
2279
2280 dir->nr = i;
2281 }
2282
2283 trace_performance_leave("read directory %.*s", len, path);
2284 if (dir->untracked) {
2285 static int force_untracked_cache = -1;
2286 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2287
2288 if (force_untracked_cache < 0)
2289 force_untracked_cache =
2290 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2291 trace_printf_key(&trace_untracked_stats,
2292 "node creation: %u\n"
2293 "gitignore invalidation: %u\n"
2294 "directory invalidation: %u\n"
2295 "opendir: %u\n",
2296 dir->untracked->dir_created,
2297 dir->untracked->gitignore_invalidated,
2298 dir->untracked->dir_invalidated,
2299 dir->untracked->dir_opened);
2300 if (force_untracked_cache &&
2301 dir->untracked == istate->untracked &&
2302 (dir->untracked->dir_opened ||
2303 dir->untracked->gitignore_invalidated ||
2304 dir->untracked->dir_invalidated))
2305 istate->cache_changed |= UNTRACKED_CHANGED;
2306 if (dir->untracked != istate->untracked) {
2307 FREE_AND_NULL(dir->untracked);
2308 }
2309 }
2310 return dir->nr;
2311 }
2312
2313 int file_exists(const char *f)
2314 {
2315 struct stat sb;
2316 return lstat(f, &sb) == 0;
2317 }
2318
2319 int repo_file_exists(struct repository *repo, const char *path)
2320 {
2321 if (repo != the_repository)
2322 BUG("do not know how to check file existence in arbitrary repo");
2323
2324 return file_exists(path);
2325 }
2326
2327 static int cmp_icase(char a, char b)
2328 {
2329 if (a == b)
2330 return 0;
2331 if (ignore_case)
2332 return toupper(a) - toupper(b);
2333 return a - b;
2334 }
2335
2336 /*
2337 * Given two normalized paths (a trailing slash is ok), if subdir is
2338 * outside dir, return -1. Otherwise return the offset in subdir that
2339 * can be used as relative path to dir.
2340 */
2341 int dir_inside_of(const char *subdir, const char *dir)
2342 {
2343 int offset = 0;
2344
2345 assert(dir && subdir && *dir && *subdir);
2346
2347 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2348 dir++;
2349 subdir++;
2350 offset++;
2351 }
2352
2353 /* hel[p]/me vs hel[l]/yeah */
2354 if (*dir && *subdir)
2355 return -1;
2356
2357 if (!*subdir)
2358 return !*dir ? offset : -1; /* same dir */
2359
2360 /* foo/[b]ar vs foo/[] */
2361 if (is_dir_sep(dir[-1]))
2362 return is_dir_sep(subdir[-1]) ? offset : -1;
2363
2364 /* foo[/]bar vs foo[] */
2365 return is_dir_sep(*subdir) ? offset + 1 : -1;
2366 }
2367
2368 int is_inside_dir(const char *dir)
2369 {
2370 char *cwd;
2371 int rc;
2372
2373 if (!dir)
2374 return 0;
2375
2376 cwd = xgetcwd();
2377 rc = (dir_inside_of(cwd, dir) >= 0);
2378 free(cwd);
2379 return rc;
2380 }
2381
2382 int is_empty_dir(const char *path)
2383 {
2384 DIR *dir = opendir(path);
2385 struct dirent *e;
2386 int ret = 1;
2387
2388 if (!dir)
2389 return 0;
2390
2391 while ((e = readdir(dir)) != NULL)
2392 if (!is_dot_or_dotdot(e->d_name)) {
2393 ret = 0;
2394 break;
2395 }
2396
2397 closedir(dir);
2398 return ret;
2399 }
2400
2401 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2402 {
2403 DIR *dir;
2404 struct dirent *e;
2405 int ret = 0, original_len = path->len, len, kept_down = 0;
2406 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2407 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2408 struct object_id submodule_head;
2409
2410 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2411 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2412 /* Do not descend and nuke a nested git work tree. */
2413 if (kept_up)
2414 *kept_up = 1;
2415 return 0;
2416 }
2417
2418 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2419 dir = opendir(path->buf);
2420 if (!dir) {
2421 if (errno == ENOENT)
2422 return keep_toplevel ? -1 : 0;
2423 else if (errno == EACCES && !keep_toplevel)
2424 /*
2425 * An empty dir could be removable even if it
2426 * is unreadable:
2427 */
2428 return rmdir(path->buf);
2429 else
2430 return -1;
2431 }
2432 strbuf_complete(path, '/');
2433
2434 len = path->len;
2435 while ((e = readdir(dir)) != NULL) {
2436 struct stat st;
2437 if (is_dot_or_dotdot(e->d_name))
2438 continue;
2439
2440 strbuf_setlen(path, len);
2441 strbuf_addstr(path, e->d_name);
2442 if (lstat(path->buf, &st)) {
2443 if (errno == ENOENT)
2444 /*
2445 * file disappeared, which is what we
2446 * wanted anyway
2447 */
2448 continue;
2449 /* fall thru */
2450 } else if (S_ISDIR(st.st_mode)) {
2451 if (!remove_dir_recurse(path, flag, &kept_down))
2452 continue; /* happy */
2453 } else if (!only_empty &&
2454 (!unlink(path->buf) || errno == ENOENT)) {
2455 continue; /* happy, too */
2456 }
2457
2458 /* path too long, stat fails, or non-directory still exists */
2459 ret = -1;
2460 break;
2461 }
2462 closedir(dir);
2463
2464 strbuf_setlen(path, original_len);
2465 if (!ret && !keep_toplevel && !kept_down)
2466 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2467 else if (kept_up)
2468 /*
2469 * report the uplevel that it is not an error that we
2470 * did not rmdir() our directory.
2471 */
2472 *kept_up = !ret;
2473 return ret;
2474 }
2475
2476 int remove_dir_recursively(struct strbuf *path, int flag)
2477 {
2478 return remove_dir_recurse(path, flag, NULL);
2479 }
2480
2481 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2482
2483 void setup_standard_excludes(struct dir_struct *dir)
2484 {
2485 dir->exclude_per_dir = ".gitignore";
2486
2487 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2488 if (!excludes_file)
2489 excludes_file = xdg_config_home("ignore");
2490 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2491 add_excludes_from_file_1(dir, excludes_file,
2492 dir->untracked ? &dir->ss_excludes_file : NULL);
2493
2494 /* per repository user preference */
2495 if (startup_info->have_repository) {
2496 const char *path = git_path_info_exclude();
2497 if (!access_or_warn(path, R_OK, 0))
2498 add_excludes_from_file_1(dir, path,
2499 dir->untracked ? &dir->ss_info_exclude : NULL);
2500 }
2501 }
2502
2503 int remove_path(const char *name)
2504 {
2505 char *slash;
2506
2507 if (unlink(name) && !is_missing_file_error(errno))
2508 return -1;
2509
2510 slash = strrchr(name, '/');
2511 if (slash) {
2512 char *dirs = xstrdup(name);
2513 slash = dirs + (slash - name);
2514 do {
2515 *slash = '\0';
2516 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2517 free(dirs);
2518 }
2519 return 0;
2520 }
2521
2522 /*
2523 * Frees memory within dir which was allocated for exclude lists and
2524 * the exclude_stack. Does not free dir itself.
2525 */
2526 void clear_directory(struct dir_struct *dir)
2527 {
2528 int i, j;
2529 struct exclude_list_group *group;
2530 struct exclude_list *el;
2531 struct exclude_stack *stk;
2532
2533 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2534 group = &dir->exclude_list_group[i];
2535 for (j = 0; j < group->nr; j++) {
2536 el = &group->el[j];
2537 if (i == EXC_DIRS)
2538 free((char *)el->src);
2539 clear_exclude_list(el);
2540 }
2541 free(group->el);
2542 }
2543
2544 stk = dir->exclude_stack;
2545 while (stk) {
2546 struct exclude_stack *prev = stk->prev;
2547 free(stk);
2548 stk = prev;
2549 }
2550 strbuf_release(&dir->basebuf);
2551 }
2552
2553 struct ondisk_untracked_cache {
2554 struct stat_data info_exclude_stat;
2555 struct stat_data excludes_file_stat;
2556 uint32_t dir_flags;
2557 };
2558
2559 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2560
2561 struct write_data {
2562 int index; /* number of written untracked_cache_dir */
2563 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2564 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2565 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2566 struct strbuf out;
2567 struct strbuf sb_stat;
2568 struct strbuf sb_sha1;
2569 };
2570
2571 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2572 {
2573 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2574 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2575 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2576 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2577 to->sd_dev = htonl(from->sd_dev);
2578 to->sd_ino = htonl(from->sd_ino);
2579 to->sd_uid = htonl(from->sd_uid);
2580 to->sd_gid = htonl(from->sd_gid);
2581 to->sd_size = htonl(from->sd_size);
2582 }
2583
2584 static void write_one_dir(struct untracked_cache_dir *untracked,
2585 struct write_data *wd)
2586 {
2587 struct stat_data stat_data;
2588 struct strbuf *out = &wd->out;
2589 unsigned char intbuf[16];
2590 unsigned int intlen, value;
2591 int i = wd->index++;
2592
2593 /*
2594 * untracked_nr should be reset whenever valid is clear, but
2595 * for safety..
2596 */
2597 if (!untracked->valid) {
2598 untracked->untracked_nr = 0;
2599 untracked->check_only = 0;
2600 }
2601
2602 if (untracked->check_only)
2603 ewah_set(wd->check_only, i);
2604 if (untracked->valid) {
2605 ewah_set(wd->valid, i);
2606 stat_data_to_disk(&stat_data, &untracked->stat_data);
2607 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2608 }
2609 if (!is_null_oid(&untracked->exclude_oid)) {
2610 ewah_set(wd->sha1_valid, i);
2611 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
2612 the_hash_algo->rawsz);
2613 }
2614
2615 intlen = encode_varint(untracked->untracked_nr, intbuf);
2616 strbuf_add(out, intbuf, intlen);
2617
2618 /* skip non-recurse directories */
2619 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2620 if (untracked->dirs[i]->recurse)
2621 value++;
2622 intlen = encode_varint(value, intbuf);
2623 strbuf_add(out, intbuf, intlen);
2624
2625 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2626
2627 for (i = 0; i < untracked->untracked_nr; i++)
2628 strbuf_add(out, untracked->untracked[i],
2629 strlen(untracked->untracked[i]) + 1);
2630
2631 for (i = 0; i < untracked->dirs_nr; i++)
2632 if (untracked->dirs[i]->recurse)
2633 write_one_dir(untracked->dirs[i], wd);
2634 }
2635
2636 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2637 {
2638 struct ondisk_untracked_cache *ouc;
2639 struct write_data wd;
2640 unsigned char varbuf[16];
2641 int varint_len;
2642 const unsigned hashsz = the_hash_algo->rawsz;
2643
2644 ouc = xcalloc(1, sizeof(*ouc));
2645 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2646 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2647 ouc->dir_flags = htonl(untracked->dir_flags);
2648
2649 varint_len = encode_varint(untracked->ident.len, varbuf);
2650 strbuf_add(out, varbuf, varint_len);
2651 strbuf_addbuf(out, &untracked->ident);
2652
2653 strbuf_add(out, ouc, sizeof(*ouc));
2654 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
2655 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
2656 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
2657 FREE_AND_NULL(ouc);
2658
2659 if (!untracked->root) {
2660 varint_len = encode_varint(0, varbuf);
2661 strbuf_add(out, varbuf, varint_len);
2662 return;
2663 }
2664
2665 wd.index = 0;
2666 wd.check_only = ewah_new();
2667 wd.valid = ewah_new();
2668 wd.sha1_valid = ewah_new();
2669 strbuf_init(&wd.out, 1024);
2670 strbuf_init(&wd.sb_stat, 1024);
2671 strbuf_init(&wd.sb_sha1, 1024);
2672 write_one_dir(untracked->root, &wd);
2673
2674 varint_len = encode_varint(wd.index, varbuf);
2675 strbuf_add(out, varbuf, varint_len);
2676 strbuf_addbuf(out, &wd.out);
2677 ewah_serialize_strbuf(wd.valid, out);
2678 ewah_serialize_strbuf(wd.check_only, out);
2679 ewah_serialize_strbuf(wd.sha1_valid, out);
2680 strbuf_addbuf(out, &wd.sb_stat);
2681 strbuf_addbuf(out, &wd.sb_sha1);
2682 strbuf_addch(out, '\0'); /* safe guard for string lists */
2683
2684 ewah_free(wd.valid);
2685 ewah_free(wd.check_only);
2686 ewah_free(wd.sha1_valid);
2687 strbuf_release(&wd.out);
2688 strbuf_release(&wd.sb_stat);
2689 strbuf_release(&wd.sb_sha1);
2690 }
2691
2692 static void free_untracked(struct untracked_cache_dir *ucd)
2693 {
2694 int i;
2695 if (!ucd)
2696 return;
2697 for (i = 0; i < ucd->dirs_nr; i++)
2698 free_untracked(ucd->dirs[i]);
2699 for (i = 0; i < ucd->untracked_nr; i++)
2700 free(ucd->untracked[i]);
2701 free(ucd->untracked);
2702 free(ucd->dirs);
2703 free(ucd);
2704 }
2705
2706 void free_untracked_cache(struct untracked_cache *uc)
2707 {
2708 if (uc)
2709 free_untracked(uc->root);
2710 free(uc);
2711 }
2712
2713 struct read_data {
2714 int index;
2715 struct untracked_cache_dir **ucd;
2716 struct ewah_bitmap *check_only;
2717 struct ewah_bitmap *valid;
2718 struct ewah_bitmap *sha1_valid;
2719 const unsigned char *data;
2720 const unsigned char *end;
2721 };
2722
2723 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2724 {
2725 memcpy(to, data, sizeof(*to));
2726 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2727 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2728 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2729 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2730 to->sd_dev = ntohl(to->sd_dev);
2731 to->sd_ino = ntohl(to->sd_ino);
2732 to->sd_uid = ntohl(to->sd_uid);
2733 to->sd_gid = ntohl(to->sd_gid);
2734 to->sd_size = ntohl(to->sd_size);
2735 }
2736
2737 static int read_one_dir(struct untracked_cache_dir **untracked_,
2738 struct read_data *rd)
2739 {
2740 struct untracked_cache_dir ud, *untracked;
2741 const unsigned char *data = rd->data, *end = rd->end;
2742 const unsigned char *eos;
2743 unsigned int value;
2744 int i;
2745
2746 memset(&ud, 0, sizeof(ud));
2747
2748 value = decode_varint(&data);
2749 if (data > end)
2750 return -1;
2751 ud.recurse = 1;
2752 ud.untracked_alloc = value;
2753 ud.untracked_nr = value;
2754 if (ud.untracked_nr)
2755 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2756
2757 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
2758 if (data > end)
2759 return -1;
2760 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2761
2762 eos = memchr(data, '\0', end - data);
2763 if (!eos || eos == end)
2764 return -1;
2765
2766 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
2767 memcpy(untracked, &ud, sizeof(ud));
2768 memcpy(untracked->name, data, eos - data + 1);
2769 data = eos + 1;
2770
2771 for (i = 0; i < untracked->untracked_nr; i++) {
2772 eos = memchr(data, '\0', end - data);
2773 if (!eos || eos == end)
2774 return -1;
2775 untracked->untracked[i] = xmemdupz(data, eos - data);
2776 data = eos + 1;
2777 }
2778
2779 rd->ucd[rd->index++] = untracked;
2780 rd->data = data;
2781
2782 for (i = 0; i < untracked->dirs_nr; i++) {
2783 if (read_one_dir(untracked->dirs + i, rd) < 0)
2784 return -1;
2785 }
2786 return 0;
2787 }
2788
2789 static void set_check_only(size_t pos, void *cb)
2790 {
2791 struct read_data *rd = cb;
2792 struct untracked_cache_dir *ud = rd->ucd[pos];
2793 ud->check_only = 1;
2794 }
2795
2796 static void read_stat(size_t pos, void *cb)
2797 {
2798 struct read_data *rd = cb;
2799 struct untracked_cache_dir *ud = rd->ucd[pos];
2800 if (rd->data + sizeof(struct stat_data) > rd->end) {
2801 rd->data = rd->end + 1;
2802 return;
2803 }
2804 stat_data_from_disk(&ud->stat_data, rd->data);
2805 rd->data += sizeof(struct stat_data);
2806 ud->valid = 1;
2807 }
2808
2809 static void read_oid(size_t pos, void *cb)
2810 {
2811 struct read_data *rd = cb;
2812 struct untracked_cache_dir *ud = rd->ucd[pos];
2813 if (rd->data + the_hash_algo->rawsz > rd->end) {
2814 rd->data = rd->end + 1;
2815 return;
2816 }
2817 hashcpy(ud->exclude_oid.hash, rd->data);
2818 rd->data += the_hash_algo->rawsz;
2819 }
2820
2821 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
2822 const unsigned char *sha1)
2823 {
2824 stat_data_from_disk(&oid_stat->stat, data);
2825 hashcpy(oid_stat->oid.hash, sha1);
2826 oid_stat->valid = 1;
2827 }
2828
2829 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2830 {
2831 struct untracked_cache *uc;
2832 struct read_data rd;
2833 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2834 const char *ident;
2835 int ident_len;
2836 ssize_t len;
2837 const char *exclude_per_dir;
2838 const unsigned hashsz = the_hash_algo->rawsz;
2839 const unsigned offset = sizeof(struct ondisk_untracked_cache);
2840 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
2841
2842 if (sz <= 1 || end[-1] != '\0')
2843 return NULL;
2844 end--;
2845
2846 ident_len = decode_varint(&next);
2847 if (next + ident_len > end)
2848 return NULL;
2849 ident = (const char *)next;
2850 next += ident_len;
2851
2852 if (next + exclude_per_dir_offset + 1 > end)
2853 return NULL;
2854
2855 uc = xcalloc(1, sizeof(*uc));
2856 strbuf_init(&uc->ident, ident_len);
2857 strbuf_add(&uc->ident, ident, ident_len);
2858 load_oid_stat(&uc->ss_info_exclude,
2859 next + ouc_offset(info_exclude_stat),
2860 next + offset);
2861 load_oid_stat(&uc->ss_excludes_file,
2862 next + ouc_offset(excludes_file_stat),
2863 next + offset + hashsz);
2864 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2865 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
2866 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2867 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2868 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
2869 if (next >= end)
2870 goto done2;
2871
2872 len = decode_varint(&next);
2873 if (next > end || len == 0)
2874 goto done2;
2875
2876 rd.valid = ewah_new();
2877 rd.check_only = ewah_new();
2878 rd.sha1_valid = ewah_new();
2879 rd.data = next;
2880 rd.end = end;
2881 rd.index = 0;
2882 ALLOC_ARRAY(rd.ucd, len);
2883
2884 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2885 goto done;
2886
2887 next = rd.data;
2888 len = ewah_read_mmap(rd.valid, next, end - next);
2889 if (len < 0)
2890 goto done;
2891
2892 next += len;
2893 len = ewah_read_mmap(rd.check_only, next, end - next);
2894 if (len < 0)
2895 goto done;
2896
2897 next += len;
2898 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2899 if (len < 0)
2900 goto done;
2901
2902 ewah_each_bit(rd.check_only, set_check_only, &rd);
2903 rd.data = next + len;
2904 ewah_each_bit(rd.valid, read_stat, &rd);
2905 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
2906 next = rd.data;
2907
2908 done:
2909 free(rd.ucd);
2910 ewah_free(rd.valid);
2911 ewah_free(rd.check_only);
2912 ewah_free(rd.sha1_valid);
2913 done2:
2914 if (next != end) {
2915 free_untracked_cache(uc);
2916 uc = NULL;
2917 }
2918 return uc;
2919 }
2920
2921 static void invalidate_one_directory(struct untracked_cache *uc,
2922 struct untracked_cache_dir *ucd)
2923 {
2924 uc->dir_invalidated++;
2925 ucd->valid = 0;
2926 ucd->untracked_nr = 0;
2927 }
2928
2929 /*
2930 * Normally when an entry is added or removed from a directory,
2931 * invalidating that directory is enough. No need to touch its
2932 * ancestors. When a directory is shown as "foo/bar/" in git-status
2933 * however, deleting or adding an entry may have cascading effect.
2934 *
2935 * Say the "foo/bar/file" has become untracked, we need to tell the
2936 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2937 * directory any more (because "bar" is managed by foo as an untracked
2938 * "file").
2939 *
2940 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2941 * was the last untracked entry in the entire "foo", we should show
2942 * "foo/" instead. Which means we have to invalidate past "bar" up to
2943 * "foo".
2944 *
2945 * This function traverses all directories from root to leaf. If there
2946 * is a chance of one of the above cases happening, we invalidate back
2947 * to root. Otherwise we just invalidate the leaf. There may be a more
2948 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2949 * detect these cases and avoid unnecessary invalidation, for example,
2950 * checking for the untracked entry named "bar/" in "foo", but for now
2951 * stick to something safe and simple.
2952 */
2953 static int invalidate_one_component(struct untracked_cache *uc,
2954 struct untracked_cache_dir *dir,
2955 const char *path, int len)
2956 {
2957 const char *rest = strchr(path, '/');
2958
2959 if (rest) {
2960 int component_len = rest - path;
2961 struct untracked_cache_dir *d =
2962 lookup_untracked(uc, dir, path, component_len);
2963 int ret =
2964 invalidate_one_component(uc, d, rest + 1,
2965 len - (component_len + 1));
2966 if (ret)
2967 invalidate_one_directory(uc, dir);
2968 return ret;
2969 }
2970
2971 invalidate_one_directory(uc, dir);
2972 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2973 }
2974
2975 void untracked_cache_invalidate_path(struct index_state *istate,
2976 const char *path, int safe_path)
2977 {
2978 if (!istate->untracked || !istate->untracked->root)
2979 return;
2980 if (!safe_path && !verify_path(path, 0))
2981 return;
2982 invalidate_one_component(istate->untracked, istate->untracked->root,
2983 path, strlen(path));
2984 }
2985
2986 void untracked_cache_remove_from_index(struct index_state *istate,
2987 const char *path)
2988 {
2989 untracked_cache_invalidate_path(istate, path, 1);
2990 }
2991
2992 void untracked_cache_add_to_index(struct index_state *istate,
2993 const char *path)
2994 {
2995 untracked_cache_invalidate_path(istate, path, 1);
2996 }
2997
2998 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
2999 const char *sub_gitdir)
3000 {
3001 int i;
3002 struct repository subrepo;
3003 struct strbuf sub_wt = STRBUF_INIT;
3004 struct strbuf sub_gd = STRBUF_INIT;
3005
3006 const struct submodule *sub;
3007
3008 /* If the submodule has no working tree, we can ignore it. */
3009 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3010 return;
3011
3012 if (repo_read_index(&subrepo) < 0)
3013 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3014
3015 for (i = 0; i < subrepo.index->cache_nr; i++) {
3016 const struct cache_entry *ce = subrepo.index->cache[i];
3017
3018 if (!S_ISGITLINK(ce->ce_mode))
3019 continue;
3020
3021 while (i + 1 < subrepo.index->cache_nr &&
3022 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3023 /*
3024 * Skip entries with the same name in different stages
3025 * to make sure an entry is returned only once.
3026 */
3027 i++;
3028
3029 sub = submodule_from_path(&subrepo, &null_oid, ce->name);
3030 if (!sub || !is_submodule_active(&subrepo, ce->name))
3031 /* .gitmodules broken or inactive sub */
3032 continue;
3033
3034 strbuf_reset(&sub_wt);
3035 strbuf_reset(&sub_gd);
3036 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3037 strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name);
3038
3039 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3040 }
3041 strbuf_release(&sub_wt);
3042 strbuf_release(&sub_gd);
3043 repo_clear(&subrepo);
3044 }
3045
3046 void connect_work_tree_and_git_dir(const char *work_tree_,
3047 const char *git_dir_,
3048 int recurse_into_nested)
3049 {
3050 struct strbuf gitfile_sb = STRBUF_INIT;
3051 struct strbuf cfg_sb = STRBUF_INIT;
3052 struct strbuf rel_path = STRBUF_INIT;
3053 char *git_dir, *work_tree;
3054
3055 /* Prepare .git file */
3056 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3057 if (safe_create_leading_directories_const(gitfile_sb.buf))
3058 die(_("could not create directories for %s"), gitfile_sb.buf);
3059
3060 /* Prepare config file */
3061 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3062 if (safe_create_leading_directories_const(cfg_sb.buf))
3063 die(_("could not create directories for %s"), cfg_sb.buf);
3064
3065 git_dir = real_pathdup(git_dir_, 1);
3066 work_tree = real_pathdup(work_tree_, 1);
3067
3068 /* Write .git file */
3069 write_file(gitfile_sb.buf, "gitdir: %s",
3070 relative_path(git_dir, work_tree, &rel_path));
3071 /* Update core.worktree setting */
3072 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3073 relative_path(work_tree, git_dir, &rel_path));
3074
3075 strbuf_release(&gitfile_sb);
3076 strbuf_release(&cfg_sb);
3077 strbuf_release(&rel_path);
3078
3079 if (recurse_into_nested)
3080 connect_wt_gitdir_in_nested(work_tree, git_dir);
3081
3082 free(work_tree);
3083 free(git_dir);
3084 }
3085
3086 /*
3087 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3088 */
3089 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3090 {
3091 if (rename(old_git_dir, new_git_dir) < 0)
3092 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3093 old_git_dir, new_git_dir);
3094
3095 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3096 }