]> git.ipfire.org Git - thirdparty/git.git/blob - dir.c
Merge branch 'jk/test-chainlint-softer'
[thirdparty/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
7 */
8 #include "cache.h"
9 #include "config.h"
10 #include "dir.h"
11 #include "object-store.h"
12 #include "attr.h"
13 #include "refs.h"
14 #include "wildmatch.h"
15 #include "pathspec.h"
16 #include "utf8.h"
17 #include "varint.h"
18 #include "ewah/ewok.h"
19 #include "fsmonitor.h"
20 #include "submodule-config.h"
21
22 /*
23 * Tells read_directory_recursive how a file or directory should be treated.
24 * Values are ordered by significance, e.g. if a directory contains both
25 * excluded and untracked files, it is listed as untracked because
26 * path_untracked > path_excluded.
27 */
28 enum path_treatment {
29 path_none = 0,
30 path_recurse,
31 path_excluded,
32 path_untracked
33 };
34
35 /*
36 * Support data structure for our opendir/readdir/closedir wrappers
37 */
38 struct cached_dir {
39 DIR *fdir;
40 struct untracked_cache_dir *untracked;
41 int nr_files;
42 int nr_dirs;
43
44 const char *d_name;
45 int d_type;
46 const char *file;
47 struct untracked_cache_dir *ucd;
48 };
49
50 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
51 struct index_state *istate, const char *path, int len,
52 struct untracked_cache_dir *untracked,
53 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
54 static int resolve_dtype(int dtype, struct index_state *istate,
55 const char *path, int len);
56
57 void dir_init(struct dir_struct *dir)
58 {
59 memset(dir, 0, sizeof(*dir));
60 }
61
62 struct dirent *
63 readdir_skip_dot_and_dotdot(DIR *dirp)
64 {
65 struct dirent *e;
66
67 while ((e = readdir(dirp)) != NULL) {
68 if (!is_dot_or_dotdot(e->d_name))
69 break;
70 }
71 return e;
72 }
73
74 int count_slashes(const char *s)
75 {
76 int cnt = 0;
77 while (*s)
78 if (*s++ == '/')
79 cnt++;
80 return cnt;
81 }
82
83 int fspathcmp(const char *a, const char *b)
84 {
85 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
86 }
87
88 int fspathncmp(const char *a, const char *b, size_t count)
89 {
90 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
91 }
92
93 int git_fnmatch(const struct pathspec_item *item,
94 const char *pattern, const char *string,
95 int prefix)
96 {
97 if (prefix > 0) {
98 if (ps_strncmp(item, pattern, string, prefix))
99 return WM_NOMATCH;
100 pattern += prefix;
101 string += prefix;
102 }
103 if (item->flags & PATHSPEC_ONESTAR) {
104 int pattern_len = strlen(++pattern);
105 int string_len = strlen(string);
106 return string_len < pattern_len ||
107 ps_strcmp(item, pattern,
108 string + string_len - pattern_len);
109 }
110 if (item->magic & PATHSPEC_GLOB)
111 return wildmatch(pattern, string,
112 WM_PATHNAME |
113 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
114 else
115 /* wildmatch has not learned no FNM_PATHNAME mode yet */
116 return wildmatch(pattern, string,
117 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
118 }
119
120 static int fnmatch_icase_mem(const char *pattern, int patternlen,
121 const char *string, int stringlen,
122 int flags)
123 {
124 int match_status;
125 struct strbuf pat_buf = STRBUF_INIT;
126 struct strbuf str_buf = STRBUF_INIT;
127 const char *use_pat = pattern;
128 const char *use_str = string;
129
130 if (pattern[patternlen]) {
131 strbuf_add(&pat_buf, pattern, patternlen);
132 use_pat = pat_buf.buf;
133 }
134 if (string[stringlen]) {
135 strbuf_add(&str_buf, string, stringlen);
136 use_str = str_buf.buf;
137 }
138
139 if (ignore_case)
140 flags |= WM_CASEFOLD;
141 match_status = wildmatch(use_pat, use_str, flags);
142
143 strbuf_release(&pat_buf);
144 strbuf_release(&str_buf);
145
146 return match_status;
147 }
148
149 static size_t common_prefix_len(const struct pathspec *pathspec)
150 {
151 int n;
152 size_t max = 0;
153
154 /*
155 * ":(icase)path" is treated as a pathspec full of
156 * wildcard. In other words, only prefix is considered common
157 * prefix. If the pathspec is abc/foo abc/bar, running in
158 * subdir xyz, the common prefix is still xyz, not xyz/abc as
159 * in non-:(icase).
160 */
161 GUARD_PATHSPEC(pathspec,
162 PATHSPEC_FROMTOP |
163 PATHSPEC_MAXDEPTH |
164 PATHSPEC_LITERAL |
165 PATHSPEC_GLOB |
166 PATHSPEC_ICASE |
167 PATHSPEC_EXCLUDE |
168 PATHSPEC_ATTR);
169
170 for (n = 0; n < pathspec->nr; n++) {
171 size_t i = 0, len = 0, item_len;
172 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
173 continue;
174 if (pathspec->items[n].magic & PATHSPEC_ICASE)
175 item_len = pathspec->items[n].prefix;
176 else
177 item_len = pathspec->items[n].nowildcard_len;
178 while (i < item_len && (n == 0 || i < max)) {
179 char c = pathspec->items[n].match[i];
180 if (c != pathspec->items[0].match[i])
181 break;
182 if (c == '/')
183 len = i + 1;
184 i++;
185 }
186 if (n == 0 || len < max) {
187 max = len;
188 if (!max)
189 break;
190 }
191 }
192 return max;
193 }
194
195 /*
196 * Returns a copy of the longest leading path common among all
197 * pathspecs.
198 */
199 char *common_prefix(const struct pathspec *pathspec)
200 {
201 unsigned long len = common_prefix_len(pathspec);
202
203 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
204 }
205
206 int fill_directory(struct dir_struct *dir,
207 struct index_state *istate,
208 const struct pathspec *pathspec)
209 {
210 const char *prefix;
211 size_t prefix_len;
212
213 unsigned exclusive_flags = DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO;
214 if ((dir->flags & exclusive_flags) == exclusive_flags)
215 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
216
217 /*
218 * Calculate common prefix for the pathspec, and
219 * use that to optimize the directory walk
220 */
221 prefix_len = common_prefix_len(pathspec);
222 prefix = prefix_len ? pathspec->items[0].match : "";
223
224 /* Read the directory and prune it */
225 read_directory(dir, istate, prefix, prefix_len, pathspec);
226
227 return prefix_len;
228 }
229
230 int within_depth(const char *name, int namelen,
231 int depth, int max_depth)
232 {
233 const char *cp = name, *cpe = name + namelen;
234
235 while (cp < cpe) {
236 if (*cp++ != '/')
237 continue;
238 depth++;
239 if (depth > max_depth)
240 return 0;
241 }
242 return 1;
243 }
244
245 /*
246 * Read the contents of the blob with the given OID into a buffer.
247 * Append a trailing LF to the end if the last line doesn't have one.
248 *
249 * Returns:
250 * -1 when the OID is invalid or unknown or does not refer to a blob.
251 * 0 when the blob is empty.
252 * 1 along with { data, size } of the (possibly augmented) buffer
253 * when successful.
254 *
255 * Optionally updates the given oid_stat with the given OID (when valid).
256 */
257 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
258 size_t *size_out, char **data_out)
259 {
260 enum object_type type;
261 unsigned long sz;
262 char *data;
263
264 *size_out = 0;
265 *data_out = NULL;
266
267 data = read_object_file(oid, &type, &sz);
268 if (!data || type != OBJ_BLOB) {
269 free(data);
270 return -1;
271 }
272
273 if (oid_stat) {
274 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
275 oidcpy(&oid_stat->oid, oid);
276 }
277
278 if (sz == 0) {
279 free(data);
280 return 0;
281 }
282
283 if (data[sz - 1] != '\n') {
284 data = xrealloc(data, st_add(sz, 1));
285 data[sz++] = '\n';
286 }
287
288 *size_out = xsize_t(sz);
289 *data_out = data;
290
291 return 1;
292 }
293
294 #define DO_MATCH_EXCLUDE (1<<0)
295 #define DO_MATCH_DIRECTORY (1<<1)
296 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
297
298 /*
299 * Does the given pathspec match the given name? A match is found if
300 *
301 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
302 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
303 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
304 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
305 *
306 * Return value tells which case it was (1-4), or 0 when there is no match.
307 *
308 * It may be instructive to look at a small table of concrete examples
309 * to understand the differences between 1, 2, and 4:
310 *
311 * Pathspecs
312 * | a/b | a/b/ | a/b/c
313 * ------+-----------+-----------+------------
314 * a/b | EXACT | EXACT[1] | LEADING[2]
315 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
316 * a/b/c | RECURSIVE | RECURSIVE | EXACT
317 *
318 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
319 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
320 */
321 static int match_pathspec_item(struct index_state *istate,
322 const struct pathspec_item *item, int prefix,
323 const char *name, int namelen, unsigned flags)
324 {
325 /* name/namelen has prefix cut off by caller */
326 const char *match = item->match + prefix;
327 int matchlen = item->len - prefix;
328
329 /*
330 * The normal call pattern is:
331 * 1. prefix = common_prefix_len(ps);
332 * 2. prune something, or fill_directory
333 * 3. match_pathspec()
334 *
335 * 'prefix' at #1 may be shorter than the command's prefix and
336 * it's ok for #2 to match extra files. Those extras will be
337 * trimmed at #3.
338 *
339 * Suppose the pathspec is 'foo' and '../bar' running from
340 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
341 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
342 * user does not want XYZ/foo, only the "foo" part should be
343 * case-insensitive. We need to filter out XYZ/foo here. In
344 * other words, we do not trust the caller on comparing the
345 * prefix part when :(icase) is involved. We do exact
346 * comparison ourselves.
347 *
348 * Normally the caller (common_prefix_len() in fact) does
349 * _exact_ matching on name[-prefix+1..-1] and we do not need
350 * to check that part. Be defensive and check it anyway, in
351 * case common_prefix_len is changed, or a new caller is
352 * introduced that does not use common_prefix_len.
353 *
354 * If the penalty turns out too high when prefix is really
355 * long, maybe change it to
356 * strncmp(match, name, item->prefix - prefix)
357 */
358 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
359 strncmp(item->match, name - prefix, item->prefix))
360 return 0;
361
362 if (item->attr_match_nr &&
363 !match_pathspec_attrs(istate, name, namelen, item))
364 return 0;
365
366 /* If the match was just the prefix, we matched */
367 if (!*match)
368 return MATCHED_RECURSIVELY;
369
370 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
371 if (matchlen == namelen)
372 return MATCHED_EXACTLY;
373
374 if (match[matchlen-1] == '/' || name[matchlen] == '/')
375 return MATCHED_RECURSIVELY;
376 } else if ((flags & DO_MATCH_DIRECTORY) &&
377 match[matchlen - 1] == '/' &&
378 namelen == matchlen - 1 &&
379 !ps_strncmp(item, match, name, namelen))
380 return MATCHED_EXACTLY;
381
382 if (item->nowildcard_len < item->len &&
383 !git_fnmatch(item, match, name,
384 item->nowildcard_len - prefix))
385 return MATCHED_FNMATCH;
386
387 /* Perform checks to see if "name" is a leading string of the pathspec */
388 if ( (flags & DO_MATCH_LEADING_PATHSPEC) &&
389 !(flags & DO_MATCH_EXCLUDE)) {
390 /* name is a literal prefix of the pathspec */
391 int offset = name[namelen-1] == '/' ? 1 : 0;
392 if ((namelen < matchlen) &&
393 (match[namelen-offset] == '/') &&
394 !ps_strncmp(item, match, name, namelen))
395 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
396
397 /* name doesn't match up to the first wild character */
398 if (item->nowildcard_len < item->len &&
399 ps_strncmp(item, match, name,
400 item->nowildcard_len - prefix))
401 return 0;
402
403 /*
404 * name has no wildcard, and it didn't match as a leading
405 * pathspec so return.
406 */
407 if (item->nowildcard_len == item->len)
408 return 0;
409
410 /*
411 * Here is where we would perform a wildmatch to check if
412 * "name" can be matched as a directory (or a prefix) against
413 * the pathspec. Since wildmatch doesn't have this capability
414 * at the present we have to punt and say that it is a match,
415 * potentially returning a false positive
416 * The submodules themselves will be able to perform more
417 * accurate matching to determine if the pathspec matches.
418 */
419 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
420 }
421
422 return 0;
423 }
424
425 /*
426 * do_match_pathspec() is meant to ONLY be called by
427 * match_pathspec_with_flags(); calling it directly risks pathspecs
428 * like ':!unwanted_path' being ignored.
429 *
430 * Given a name and a list of pathspecs, returns the nature of the
431 * closest (i.e. most specific) match of the name to any of the
432 * pathspecs.
433 *
434 * The caller typically calls this multiple times with the same
435 * pathspec and seen[] array but with different name/namelen
436 * (e.g. entries from the index) and is interested in seeing if and
437 * how each pathspec matches all the names it calls this function
438 * with. A mark is left in the seen[] array for each pathspec element
439 * indicating the closest type of match that element achieved, so if
440 * seen[n] remains zero after multiple invocations, that means the nth
441 * pathspec did not match any names, which could indicate that the
442 * user mistyped the nth pathspec.
443 */
444 static int do_match_pathspec(struct index_state *istate,
445 const struct pathspec *ps,
446 const char *name, int namelen,
447 int prefix, char *seen,
448 unsigned flags)
449 {
450 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
451
452 GUARD_PATHSPEC(ps,
453 PATHSPEC_FROMTOP |
454 PATHSPEC_MAXDEPTH |
455 PATHSPEC_LITERAL |
456 PATHSPEC_GLOB |
457 PATHSPEC_ICASE |
458 PATHSPEC_EXCLUDE |
459 PATHSPEC_ATTR);
460
461 if (!ps->nr) {
462 if (!ps->recursive ||
463 !(ps->magic & PATHSPEC_MAXDEPTH) ||
464 ps->max_depth == -1)
465 return MATCHED_RECURSIVELY;
466
467 if (within_depth(name, namelen, 0, ps->max_depth))
468 return MATCHED_EXACTLY;
469 else
470 return 0;
471 }
472
473 name += prefix;
474 namelen -= prefix;
475
476 for (i = ps->nr - 1; i >= 0; i--) {
477 int how;
478
479 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
480 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
481 continue;
482
483 if (seen && seen[i] == MATCHED_EXACTLY)
484 continue;
485 /*
486 * Make exclude patterns optional and never report
487 * "pathspec ':(exclude)foo' matches no files"
488 */
489 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
490 seen[i] = MATCHED_FNMATCH;
491 how = match_pathspec_item(istate, ps->items+i, prefix, name,
492 namelen, flags);
493 if (ps->recursive &&
494 (ps->magic & PATHSPEC_MAXDEPTH) &&
495 ps->max_depth != -1 &&
496 how && how != MATCHED_FNMATCH) {
497 int len = ps->items[i].len;
498 if (name[len] == '/')
499 len++;
500 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
501 how = MATCHED_EXACTLY;
502 else
503 how = 0;
504 }
505 if (how) {
506 if (retval < how)
507 retval = how;
508 if (seen && seen[i] < how)
509 seen[i] = how;
510 }
511 }
512 return retval;
513 }
514
515 static int match_pathspec_with_flags(struct index_state *istate,
516 const struct pathspec *ps,
517 const char *name, int namelen,
518 int prefix, char *seen, unsigned flags)
519 {
520 int positive, negative;
521 positive = do_match_pathspec(istate, ps, name, namelen,
522 prefix, seen, flags);
523 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
524 return positive;
525 negative = do_match_pathspec(istate, ps, name, namelen,
526 prefix, seen,
527 flags | DO_MATCH_EXCLUDE);
528 return negative ? 0 : positive;
529 }
530
531 int match_pathspec(struct index_state *istate,
532 const struct pathspec *ps,
533 const char *name, int namelen,
534 int prefix, char *seen, int is_dir)
535 {
536 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
537 return match_pathspec_with_flags(istate, ps, name, namelen,
538 prefix, seen, flags);
539 }
540
541 /**
542 * Check if a submodule is a superset of the pathspec
543 */
544 int submodule_path_match(struct index_state *istate,
545 const struct pathspec *ps,
546 const char *submodule_name,
547 char *seen)
548 {
549 int matched = match_pathspec_with_flags(istate, ps, submodule_name,
550 strlen(submodule_name),
551 0, seen,
552 DO_MATCH_DIRECTORY |
553 DO_MATCH_LEADING_PATHSPEC);
554 return matched;
555 }
556
557 int report_path_error(const char *ps_matched,
558 const struct pathspec *pathspec)
559 {
560 /*
561 * Make sure all pathspec matched; otherwise it is an error.
562 */
563 int num, errors = 0;
564 for (num = 0; num < pathspec->nr; num++) {
565 int other, found_dup;
566
567 if (ps_matched[num])
568 continue;
569 /*
570 * The caller might have fed identical pathspec
571 * twice. Do not barf on such a mistake.
572 * FIXME: parse_pathspec should have eliminated
573 * duplicate pathspec.
574 */
575 for (found_dup = other = 0;
576 !found_dup && other < pathspec->nr;
577 other++) {
578 if (other == num || !ps_matched[other])
579 continue;
580 if (!strcmp(pathspec->items[other].original,
581 pathspec->items[num].original))
582 /*
583 * Ok, we have a match already.
584 */
585 found_dup = 1;
586 }
587 if (found_dup)
588 continue;
589
590 error(_("pathspec '%s' did not match any file(s) known to git"),
591 pathspec->items[num].original);
592 errors++;
593 }
594 return errors;
595 }
596
597 /*
598 * Return the length of the "simple" part of a path match limiter.
599 */
600 int simple_length(const char *match)
601 {
602 int len = -1;
603
604 for (;;) {
605 unsigned char c = *match++;
606 len++;
607 if (c == '\0' || is_glob_special(c))
608 return len;
609 }
610 }
611
612 int no_wildcard(const char *string)
613 {
614 return string[simple_length(string)] == '\0';
615 }
616
617 void parse_path_pattern(const char **pattern,
618 int *patternlen,
619 unsigned *flags,
620 int *nowildcardlen)
621 {
622 const char *p = *pattern;
623 size_t i, len;
624
625 *flags = 0;
626 if (*p == '!') {
627 *flags |= PATTERN_FLAG_NEGATIVE;
628 p++;
629 }
630 len = strlen(p);
631 if (len && p[len - 1] == '/') {
632 len--;
633 *flags |= PATTERN_FLAG_MUSTBEDIR;
634 }
635 for (i = 0; i < len; i++) {
636 if (p[i] == '/')
637 break;
638 }
639 if (i == len)
640 *flags |= PATTERN_FLAG_NODIR;
641 *nowildcardlen = simple_length(p);
642 /*
643 * we should have excluded the trailing slash from 'p' too,
644 * but that's one more allocation. Instead just make sure
645 * nowildcardlen does not exceed real patternlen
646 */
647 if (*nowildcardlen > len)
648 *nowildcardlen = len;
649 if (*p == '*' && no_wildcard(p + 1))
650 *flags |= PATTERN_FLAG_ENDSWITH;
651 *pattern = p;
652 *patternlen = len;
653 }
654
655 int pl_hashmap_cmp(const void *unused_cmp_data,
656 const struct hashmap_entry *a,
657 const struct hashmap_entry *b,
658 const void *key)
659 {
660 const struct pattern_entry *ee1 =
661 container_of(a, struct pattern_entry, ent);
662 const struct pattern_entry *ee2 =
663 container_of(b, struct pattern_entry, ent);
664
665 size_t min_len = ee1->patternlen <= ee2->patternlen
666 ? ee1->patternlen
667 : ee2->patternlen;
668
669 if (ignore_case)
670 return strncasecmp(ee1->pattern, ee2->pattern, min_len);
671 return strncmp(ee1->pattern, ee2->pattern, min_len);
672 }
673
674 static char *dup_and_filter_pattern(const char *pattern)
675 {
676 char *set, *read;
677 size_t count = 0;
678 char *result = xstrdup(pattern);
679
680 set = result;
681 read = result;
682
683 while (*read) {
684 /* skip escape characters (once) */
685 if (*read == '\\')
686 read++;
687
688 *set = *read;
689
690 set++;
691 read++;
692 count++;
693 }
694 *set = 0;
695
696 if (count > 2 &&
697 *(set - 1) == '*' &&
698 *(set - 2) == '/')
699 *(set - 2) = 0;
700
701 return result;
702 }
703
704 static void add_pattern_to_hashsets(struct pattern_list *pl, struct path_pattern *given)
705 {
706 struct pattern_entry *translated;
707 char *truncated;
708 char *data = NULL;
709 const char *prev, *cur, *next;
710
711 if (!pl->use_cone_patterns)
712 return;
713
714 if (given->flags & PATTERN_FLAG_NEGATIVE &&
715 given->flags & PATTERN_FLAG_MUSTBEDIR &&
716 !strcmp(given->pattern, "/*")) {
717 pl->full_cone = 0;
718 return;
719 }
720
721 if (!given->flags && !strcmp(given->pattern, "/*")) {
722 pl->full_cone = 1;
723 return;
724 }
725
726 if (given->patternlen < 2 ||
727 *given->pattern == '*' ||
728 strstr(given->pattern, "**")) {
729 /* Not a cone pattern. */
730 warning(_("unrecognized pattern: '%s'"), given->pattern);
731 goto clear_hashmaps;
732 }
733
734 prev = given->pattern;
735 cur = given->pattern + 1;
736 next = given->pattern + 2;
737
738 while (*cur) {
739 /* Watch for glob characters '*', '\', '[', '?' */
740 if (!is_glob_special(*cur))
741 goto increment;
742
743 /* But only if *prev != '\\' */
744 if (*prev == '\\')
745 goto increment;
746
747 /* But allow the initial '\' */
748 if (*cur == '\\' &&
749 is_glob_special(*next))
750 goto increment;
751
752 /* But a trailing '/' then '*' is fine */
753 if (*prev == '/' &&
754 *cur == '*' &&
755 *next == 0)
756 goto increment;
757
758 /* Not a cone pattern. */
759 warning(_("unrecognized pattern: '%s'"), given->pattern);
760 goto clear_hashmaps;
761
762 increment:
763 prev++;
764 cur++;
765 next++;
766 }
767
768 if (given->patternlen > 2 &&
769 !strcmp(given->pattern + given->patternlen - 2, "/*")) {
770 if (!(given->flags & PATTERN_FLAG_NEGATIVE)) {
771 /* Not a cone pattern. */
772 warning(_("unrecognized pattern: '%s'"), given->pattern);
773 goto clear_hashmaps;
774 }
775
776 truncated = dup_and_filter_pattern(given->pattern);
777
778 translated = xmalloc(sizeof(struct pattern_entry));
779 translated->pattern = truncated;
780 translated->patternlen = given->patternlen - 2;
781 hashmap_entry_init(&translated->ent,
782 ignore_case ?
783 strihash(translated->pattern) :
784 strhash(translated->pattern));
785
786 if (!hashmap_get_entry(&pl->recursive_hashmap,
787 translated, ent, NULL)) {
788 /* We did not see the "parent" included */
789 warning(_("unrecognized negative pattern: '%s'"),
790 given->pattern);
791 free(truncated);
792 free(translated);
793 goto clear_hashmaps;
794 }
795
796 hashmap_add(&pl->parent_hashmap, &translated->ent);
797 hashmap_remove(&pl->recursive_hashmap, &translated->ent, &data);
798 free(data);
799 return;
800 }
801
802 if (given->flags & PATTERN_FLAG_NEGATIVE) {
803 warning(_("unrecognized negative pattern: '%s'"),
804 given->pattern);
805 goto clear_hashmaps;
806 }
807
808 translated = xmalloc(sizeof(struct pattern_entry));
809
810 translated->pattern = dup_and_filter_pattern(given->pattern);
811 translated->patternlen = given->patternlen;
812 hashmap_entry_init(&translated->ent,
813 ignore_case ?
814 strihash(translated->pattern) :
815 strhash(translated->pattern));
816
817 hashmap_add(&pl->recursive_hashmap, &translated->ent);
818
819 if (hashmap_get_entry(&pl->parent_hashmap, translated, ent, NULL)) {
820 /* we already included this at the parent level */
821 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
822 given->pattern);
823 hashmap_remove(&pl->parent_hashmap, &translated->ent, &data);
824 free(data);
825 free(translated);
826 }
827
828 return;
829
830 clear_hashmaps:
831 warning(_("disabling cone pattern matching"));
832 hashmap_clear_and_free(&pl->parent_hashmap, struct pattern_entry, ent);
833 hashmap_clear_and_free(&pl->recursive_hashmap, struct pattern_entry, ent);
834 pl->use_cone_patterns = 0;
835 }
836
837 static int hashmap_contains_path(struct hashmap *map,
838 struct strbuf *pattern)
839 {
840 struct pattern_entry p;
841
842 /* Check straight mapping */
843 p.pattern = pattern->buf;
844 p.patternlen = pattern->len;
845 hashmap_entry_init(&p.ent,
846 ignore_case ?
847 strihash(p.pattern) :
848 strhash(p.pattern));
849 return !!hashmap_get_entry(map, &p, ent, NULL);
850 }
851
852 int hashmap_contains_parent(struct hashmap *map,
853 const char *path,
854 struct strbuf *buffer)
855 {
856 char *slash_pos;
857
858 strbuf_setlen(buffer, 0);
859
860 if (path[0] != '/')
861 strbuf_addch(buffer, '/');
862
863 strbuf_addstr(buffer, path);
864
865 slash_pos = strrchr(buffer->buf, '/');
866
867 while (slash_pos > buffer->buf) {
868 strbuf_setlen(buffer, slash_pos - buffer->buf);
869
870 if (hashmap_contains_path(map, buffer))
871 return 1;
872
873 slash_pos = strrchr(buffer->buf, '/');
874 }
875
876 return 0;
877 }
878
879 void add_pattern(const char *string, const char *base,
880 int baselen, struct pattern_list *pl, int srcpos)
881 {
882 struct path_pattern *pattern;
883 int patternlen;
884 unsigned flags;
885 int nowildcardlen;
886
887 parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen);
888 if (flags & PATTERN_FLAG_MUSTBEDIR) {
889 FLEXPTR_ALLOC_MEM(pattern, pattern, string, patternlen);
890 } else {
891 pattern = xmalloc(sizeof(*pattern));
892 pattern->pattern = string;
893 }
894 pattern->patternlen = patternlen;
895 pattern->nowildcardlen = nowildcardlen;
896 pattern->base = base;
897 pattern->baselen = baselen;
898 pattern->flags = flags;
899 pattern->srcpos = srcpos;
900 ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc);
901 pl->patterns[pl->nr++] = pattern;
902 pattern->pl = pl;
903
904 add_pattern_to_hashsets(pl, pattern);
905 }
906
907 static int read_skip_worktree_file_from_index(struct index_state *istate,
908 const char *path,
909 size_t *size_out, char **data_out,
910 struct oid_stat *oid_stat)
911 {
912 int pos, len;
913
914 len = strlen(path);
915 pos = index_name_pos(istate, path, len);
916 if (pos < 0)
917 return -1;
918 if (!ce_skip_worktree(istate->cache[pos]))
919 return -1;
920
921 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
922 }
923
924 /*
925 * Frees memory within pl which was allocated for exclude patterns and
926 * the file buffer. Does not free pl itself.
927 */
928 void clear_pattern_list(struct pattern_list *pl)
929 {
930 int i;
931
932 for (i = 0; i < pl->nr; i++)
933 free(pl->patterns[i]);
934 free(pl->patterns);
935 free(pl->filebuf);
936 hashmap_clear_and_free(&pl->recursive_hashmap, struct pattern_entry, ent);
937 hashmap_clear_and_free(&pl->parent_hashmap, struct pattern_entry, ent);
938
939 memset(pl, 0, sizeof(*pl));
940 }
941
942 static void trim_trailing_spaces(char *buf)
943 {
944 char *p, *last_space = NULL;
945
946 for (p = buf; *p; p++)
947 switch (*p) {
948 case ' ':
949 if (!last_space)
950 last_space = p;
951 break;
952 case '\\':
953 p++;
954 if (!*p)
955 return;
956 /* fallthrough */
957 default:
958 last_space = NULL;
959 }
960
961 if (last_space)
962 *last_space = '\0';
963 }
964
965 /*
966 * Given a subdirectory name and "dir" of the current directory,
967 * search the subdir in "dir" and return it, or create a new one if it
968 * does not exist in "dir".
969 *
970 * If "name" has the trailing slash, it'll be excluded in the search.
971 */
972 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
973 struct untracked_cache_dir *dir,
974 const char *name, int len)
975 {
976 int first, last;
977 struct untracked_cache_dir *d;
978 if (!dir)
979 return NULL;
980 if (len && name[len - 1] == '/')
981 len--;
982 first = 0;
983 last = dir->dirs_nr;
984 while (last > first) {
985 int cmp, next = first + ((last - first) >> 1);
986 d = dir->dirs[next];
987 cmp = strncmp(name, d->name, len);
988 if (!cmp && strlen(d->name) > len)
989 cmp = -1;
990 if (!cmp)
991 return d;
992 if (cmp < 0) {
993 last = next;
994 continue;
995 }
996 first = next+1;
997 }
998
999 uc->dir_created++;
1000 FLEX_ALLOC_MEM(d, name, name, len);
1001
1002 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
1003 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
1004 dir->dirs_nr - first);
1005 dir->dirs_nr++;
1006 dir->dirs[first] = d;
1007 return d;
1008 }
1009
1010 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
1011 {
1012 int i;
1013 dir->valid = 0;
1014 dir->untracked_nr = 0;
1015 for (i = 0; i < dir->dirs_nr; i++)
1016 do_invalidate_gitignore(dir->dirs[i]);
1017 }
1018
1019 static void invalidate_gitignore(struct untracked_cache *uc,
1020 struct untracked_cache_dir *dir)
1021 {
1022 uc->gitignore_invalidated++;
1023 do_invalidate_gitignore(dir);
1024 }
1025
1026 static void invalidate_directory(struct untracked_cache *uc,
1027 struct untracked_cache_dir *dir)
1028 {
1029 int i;
1030
1031 /*
1032 * Invalidation increment here is just roughly correct. If
1033 * untracked_nr or any of dirs[].recurse is non-zero, we
1034 * should increment dir_invalidated too. But that's more
1035 * expensive to do.
1036 */
1037 if (dir->valid)
1038 uc->dir_invalidated++;
1039
1040 dir->valid = 0;
1041 dir->untracked_nr = 0;
1042 for (i = 0; i < dir->dirs_nr; i++)
1043 dir->dirs[i]->recurse = 0;
1044 }
1045
1046 static int add_patterns_from_buffer(char *buf, size_t size,
1047 const char *base, int baselen,
1048 struct pattern_list *pl);
1049
1050 /* Flags for add_patterns() */
1051 #define PATTERN_NOFOLLOW (1<<0)
1052
1053 /*
1054 * Given a file with name "fname", read it (either from disk, or from
1055 * an index if 'istate' is non-null), parse it and store the
1056 * exclude rules in "pl".
1057 *
1058 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1059 * stat data from disk (only valid if add_patterns returns zero). If
1060 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1061 */
1062 static int add_patterns(const char *fname, const char *base, int baselen,
1063 struct pattern_list *pl, struct index_state *istate,
1064 unsigned flags, struct oid_stat *oid_stat)
1065 {
1066 struct stat st;
1067 int r;
1068 int fd;
1069 size_t size = 0;
1070 char *buf;
1071
1072 if (flags & PATTERN_NOFOLLOW)
1073 fd = open_nofollow(fname, O_RDONLY);
1074 else
1075 fd = open(fname, O_RDONLY);
1076
1077 if (fd < 0 || fstat(fd, &st) < 0) {
1078 if (fd < 0)
1079 warn_on_fopen_errors(fname);
1080 else
1081 close(fd);
1082 if (!istate)
1083 return -1;
1084 r = read_skip_worktree_file_from_index(istate, fname,
1085 &size, &buf,
1086 oid_stat);
1087 if (r != 1)
1088 return r;
1089 } else {
1090 size = xsize_t(st.st_size);
1091 if (size == 0) {
1092 if (oid_stat) {
1093 fill_stat_data(&oid_stat->stat, &st);
1094 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
1095 oid_stat->valid = 1;
1096 }
1097 close(fd);
1098 return 0;
1099 }
1100 buf = xmallocz(size);
1101 if (read_in_full(fd, buf, size) != size) {
1102 free(buf);
1103 close(fd);
1104 return -1;
1105 }
1106 buf[size++] = '\n';
1107 close(fd);
1108 if (oid_stat) {
1109 int pos;
1110 if (oid_stat->valid &&
1111 !match_stat_data_racy(istate, &oid_stat->stat, &st))
1112 ; /* no content change, oid_stat->oid still good */
1113 else if (istate &&
1114 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
1115 !ce_stage(istate->cache[pos]) &&
1116 ce_uptodate(istate->cache[pos]) &&
1117 !would_convert_to_git(istate, fname))
1118 oidcpy(&oid_stat->oid,
1119 &istate->cache[pos]->oid);
1120 else
1121 hash_object_file(the_hash_algo, buf, size,
1122 "blob", &oid_stat->oid);
1123 fill_stat_data(&oid_stat->stat, &st);
1124 oid_stat->valid = 1;
1125 }
1126 }
1127
1128 add_patterns_from_buffer(buf, size, base, baselen, pl);
1129 return 0;
1130 }
1131
1132 static int add_patterns_from_buffer(char *buf, size_t size,
1133 const char *base, int baselen,
1134 struct pattern_list *pl)
1135 {
1136 int i, lineno = 1;
1137 char *entry;
1138
1139 hashmap_init(&pl->recursive_hashmap, pl_hashmap_cmp, NULL, 0);
1140 hashmap_init(&pl->parent_hashmap, pl_hashmap_cmp, NULL, 0);
1141
1142 pl->filebuf = buf;
1143
1144 if (skip_utf8_bom(&buf, size))
1145 size -= buf - pl->filebuf;
1146
1147 entry = buf;
1148
1149 for (i = 0; i < size; i++) {
1150 if (buf[i] == '\n') {
1151 if (entry != buf + i && entry[0] != '#') {
1152 buf[i - (i && buf[i-1] == '\r')] = 0;
1153 trim_trailing_spaces(entry);
1154 add_pattern(entry, base, baselen, pl, lineno);
1155 }
1156 lineno++;
1157 entry = buf + i + 1;
1158 }
1159 }
1160 return 0;
1161 }
1162
1163 int add_patterns_from_file_to_list(const char *fname, const char *base,
1164 int baselen, struct pattern_list *pl,
1165 struct index_state *istate,
1166 unsigned flags)
1167 {
1168 return add_patterns(fname, base, baselen, pl, istate, flags, NULL);
1169 }
1170
1171 int add_patterns_from_blob_to_list(
1172 struct object_id *oid,
1173 const char *base, int baselen,
1174 struct pattern_list *pl)
1175 {
1176 char *buf;
1177 size_t size;
1178 int r;
1179
1180 r = do_read_blob(oid, NULL, &size, &buf);
1181 if (r != 1)
1182 return r;
1183
1184 add_patterns_from_buffer(buf, size, base, baselen, pl);
1185 return 0;
1186 }
1187
1188 struct pattern_list *add_pattern_list(struct dir_struct *dir,
1189 int group_type, const char *src)
1190 {
1191 struct pattern_list *pl;
1192 struct exclude_list_group *group;
1193
1194 group = &dir->exclude_list_group[group_type];
1195 ALLOC_GROW(group->pl, group->nr + 1, group->alloc);
1196 pl = &group->pl[group->nr++];
1197 memset(pl, 0, sizeof(*pl));
1198 pl->src = src;
1199 return pl;
1200 }
1201
1202 /*
1203 * Used to set up core.excludesfile and .git/info/exclude lists.
1204 */
1205 static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname,
1206 struct oid_stat *oid_stat)
1207 {
1208 struct pattern_list *pl;
1209 /*
1210 * catch setup_standard_excludes() that's called before
1211 * dir->untracked is assigned. That function behaves
1212 * differently when dir->untracked is non-NULL.
1213 */
1214 if (!dir->untracked)
1215 dir->unmanaged_exclude_files++;
1216 pl = add_pattern_list(dir, EXC_FILE, fname);
1217 if (add_patterns(fname, "", 0, pl, NULL, 0, oid_stat) < 0)
1218 die(_("cannot use %s as an exclude file"), fname);
1219 }
1220
1221 void add_patterns_from_file(struct dir_struct *dir, const char *fname)
1222 {
1223 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
1224 add_patterns_from_file_1(dir, fname, NULL);
1225 }
1226
1227 int match_basename(const char *basename, int basenamelen,
1228 const char *pattern, int prefix, int patternlen,
1229 unsigned flags)
1230 {
1231 if (prefix == patternlen) {
1232 if (patternlen == basenamelen &&
1233 !fspathncmp(pattern, basename, basenamelen))
1234 return 1;
1235 } else if (flags & PATTERN_FLAG_ENDSWITH) {
1236 /* "*literal" matching against "fooliteral" */
1237 if (patternlen - 1 <= basenamelen &&
1238 !fspathncmp(pattern + 1,
1239 basename + basenamelen - (patternlen - 1),
1240 patternlen - 1))
1241 return 1;
1242 } else {
1243 if (fnmatch_icase_mem(pattern, patternlen,
1244 basename, basenamelen,
1245 0) == 0)
1246 return 1;
1247 }
1248 return 0;
1249 }
1250
1251 int match_pathname(const char *pathname, int pathlen,
1252 const char *base, int baselen,
1253 const char *pattern, int prefix, int patternlen,
1254 unsigned flags)
1255 {
1256 const char *name;
1257 int namelen;
1258
1259 /*
1260 * match with FNM_PATHNAME; the pattern has base implicitly
1261 * in front of it.
1262 */
1263 if (*pattern == '/') {
1264 pattern++;
1265 patternlen--;
1266 prefix--;
1267 }
1268
1269 /*
1270 * baselen does not count the trailing slash. base[] may or
1271 * may not end with a trailing slash though.
1272 */
1273 if (pathlen < baselen + 1 ||
1274 (baselen && pathname[baselen] != '/') ||
1275 fspathncmp(pathname, base, baselen))
1276 return 0;
1277
1278 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1279 name = pathname + pathlen - namelen;
1280
1281 if (prefix) {
1282 /*
1283 * if the non-wildcard part is longer than the
1284 * remaining pathname, surely it cannot match.
1285 */
1286 if (prefix > namelen)
1287 return 0;
1288
1289 if (fspathncmp(pattern, name, prefix))
1290 return 0;
1291 pattern += prefix;
1292 patternlen -= prefix;
1293 name += prefix;
1294 namelen -= prefix;
1295
1296 /*
1297 * If the whole pattern did not have a wildcard,
1298 * then our prefix match is all we need; we
1299 * do not need to call fnmatch at all.
1300 */
1301 if (!patternlen && !namelen)
1302 return 1;
1303 }
1304
1305 return fnmatch_icase_mem(pattern, patternlen,
1306 name, namelen,
1307 WM_PATHNAME) == 0;
1308 }
1309
1310 /*
1311 * Scan the given exclude list in reverse to see whether pathname
1312 * should be ignored. The first match (i.e. the last on the list), if
1313 * any, determines the fate. Returns the exclude_list element which
1314 * matched, or NULL for undecided.
1315 */
1316 static struct path_pattern *last_matching_pattern_from_list(const char *pathname,
1317 int pathlen,
1318 const char *basename,
1319 int *dtype,
1320 struct pattern_list *pl,
1321 struct index_state *istate)
1322 {
1323 struct path_pattern *res = NULL; /* undecided */
1324 int i;
1325
1326 if (!pl->nr)
1327 return NULL; /* undefined */
1328
1329 for (i = pl->nr - 1; 0 <= i; i--) {
1330 struct path_pattern *pattern = pl->patterns[i];
1331 const char *exclude = pattern->pattern;
1332 int prefix = pattern->nowildcardlen;
1333
1334 if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) {
1335 *dtype = resolve_dtype(*dtype, istate, pathname, pathlen);
1336 if (*dtype != DT_DIR)
1337 continue;
1338 }
1339
1340 if (pattern->flags & PATTERN_FLAG_NODIR) {
1341 if (match_basename(basename,
1342 pathlen - (basename - pathname),
1343 exclude, prefix, pattern->patternlen,
1344 pattern->flags)) {
1345 res = pattern;
1346 break;
1347 }
1348 continue;
1349 }
1350
1351 assert(pattern->baselen == 0 ||
1352 pattern->base[pattern->baselen - 1] == '/');
1353 if (match_pathname(pathname, pathlen,
1354 pattern->base,
1355 pattern->baselen ? pattern->baselen - 1 : 0,
1356 exclude, prefix, pattern->patternlen,
1357 pattern->flags)) {
1358 res = pattern;
1359 break;
1360 }
1361 }
1362 return res;
1363 }
1364
1365 /*
1366 * Scan the list of patterns to determine if the ordered list
1367 * of patterns matches on 'pathname'.
1368 *
1369 * Return 1 for a match, 0 for not matched and -1 for undecided.
1370 */
1371 enum pattern_match_result path_matches_pattern_list(
1372 const char *pathname, int pathlen,
1373 const char *basename, int *dtype,
1374 struct pattern_list *pl,
1375 struct index_state *istate)
1376 {
1377 struct path_pattern *pattern;
1378 struct strbuf parent_pathname = STRBUF_INIT;
1379 int result = NOT_MATCHED;
1380 const char *slash_pos;
1381
1382 if (!pl->use_cone_patterns) {
1383 pattern = last_matching_pattern_from_list(pathname, pathlen, basename,
1384 dtype, pl, istate);
1385 if (pattern) {
1386 if (pattern->flags & PATTERN_FLAG_NEGATIVE)
1387 return NOT_MATCHED;
1388 else
1389 return MATCHED;
1390 }
1391
1392 return UNDECIDED;
1393 }
1394
1395 if (pl->full_cone)
1396 return MATCHED;
1397
1398 strbuf_addch(&parent_pathname, '/');
1399 strbuf_add(&parent_pathname, pathname, pathlen);
1400
1401 if (hashmap_contains_path(&pl->recursive_hashmap,
1402 &parent_pathname)) {
1403 result = MATCHED_RECURSIVE;
1404 goto done;
1405 }
1406
1407 slash_pos = strrchr(parent_pathname.buf, '/');
1408
1409 if (slash_pos == parent_pathname.buf) {
1410 /* include every file in root */
1411 result = MATCHED;
1412 goto done;
1413 }
1414
1415 strbuf_setlen(&parent_pathname, slash_pos - parent_pathname.buf);
1416
1417 if (hashmap_contains_path(&pl->parent_hashmap, &parent_pathname)) {
1418 result = MATCHED;
1419 goto done;
1420 }
1421
1422 if (hashmap_contains_parent(&pl->recursive_hashmap,
1423 pathname,
1424 &parent_pathname))
1425 result = MATCHED_RECURSIVE;
1426
1427 done:
1428 strbuf_release(&parent_pathname);
1429 return result;
1430 }
1431
1432 static struct path_pattern *last_matching_pattern_from_lists(
1433 struct dir_struct *dir, struct index_state *istate,
1434 const char *pathname, int pathlen,
1435 const char *basename, int *dtype_p)
1436 {
1437 int i, j;
1438 struct exclude_list_group *group;
1439 struct path_pattern *pattern;
1440 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1441 group = &dir->exclude_list_group[i];
1442 for (j = group->nr - 1; j >= 0; j--) {
1443 pattern = last_matching_pattern_from_list(
1444 pathname, pathlen, basename, dtype_p,
1445 &group->pl[j], istate);
1446 if (pattern)
1447 return pattern;
1448 }
1449 }
1450 return NULL;
1451 }
1452
1453 /*
1454 * Loads the per-directory exclude list for the substring of base
1455 * which has a char length of baselen.
1456 */
1457 static void prep_exclude(struct dir_struct *dir,
1458 struct index_state *istate,
1459 const char *base, int baselen)
1460 {
1461 struct exclude_list_group *group;
1462 struct pattern_list *pl;
1463 struct exclude_stack *stk = NULL;
1464 struct untracked_cache_dir *untracked;
1465 int current;
1466
1467 group = &dir->exclude_list_group[EXC_DIRS];
1468
1469 /*
1470 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1471 * which originate from directories not in the prefix of the
1472 * path being checked.
1473 */
1474 while ((stk = dir->exclude_stack) != NULL) {
1475 if (stk->baselen <= baselen &&
1476 !strncmp(dir->basebuf.buf, base, stk->baselen))
1477 break;
1478 pl = &group->pl[dir->exclude_stack->exclude_ix];
1479 dir->exclude_stack = stk->prev;
1480 dir->pattern = NULL;
1481 free((char *)pl->src); /* see strbuf_detach() below */
1482 clear_pattern_list(pl);
1483 free(stk);
1484 group->nr--;
1485 }
1486
1487 /* Skip traversing into sub directories if the parent is excluded */
1488 if (dir->pattern)
1489 return;
1490
1491 /*
1492 * Lazy initialization. All call sites currently just
1493 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1494 * them seems lots of work for little benefit.
1495 */
1496 if (!dir->basebuf.buf)
1497 strbuf_init(&dir->basebuf, PATH_MAX);
1498
1499 /* Read from the parent directories and push them down. */
1500 current = stk ? stk->baselen : -1;
1501 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1502 if (dir->untracked)
1503 untracked = stk ? stk->ucd : dir->untracked->root;
1504 else
1505 untracked = NULL;
1506
1507 while (current < baselen) {
1508 const char *cp;
1509 struct oid_stat oid_stat;
1510
1511 CALLOC_ARRAY(stk, 1);
1512 if (current < 0) {
1513 cp = base;
1514 current = 0;
1515 } else {
1516 cp = strchr(base + current + 1, '/');
1517 if (!cp)
1518 die("oops in prep_exclude");
1519 cp++;
1520 untracked =
1521 lookup_untracked(dir->untracked, untracked,
1522 base + current,
1523 cp - base - current);
1524 }
1525 stk->prev = dir->exclude_stack;
1526 stk->baselen = cp - base;
1527 stk->exclude_ix = group->nr;
1528 stk->ucd = untracked;
1529 pl = add_pattern_list(dir, EXC_DIRS, NULL);
1530 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1531 assert(stk->baselen == dir->basebuf.len);
1532
1533 /* Abort if the directory is excluded */
1534 if (stk->baselen) {
1535 int dt = DT_DIR;
1536 dir->basebuf.buf[stk->baselen - 1] = 0;
1537 dir->pattern = last_matching_pattern_from_lists(dir,
1538 istate,
1539 dir->basebuf.buf, stk->baselen - 1,
1540 dir->basebuf.buf + current, &dt);
1541 dir->basebuf.buf[stk->baselen - 1] = '/';
1542 if (dir->pattern &&
1543 dir->pattern->flags & PATTERN_FLAG_NEGATIVE)
1544 dir->pattern = NULL;
1545 if (dir->pattern) {
1546 dir->exclude_stack = stk;
1547 return;
1548 }
1549 }
1550
1551 /* Try to read per-directory file */
1552 oidclr(&oid_stat.oid);
1553 oid_stat.valid = 0;
1554 if (dir->exclude_per_dir &&
1555 /*
1556 * If we know that no files have been added in
1557 * this directory (i.e. valid_cached_dir() has
1558 * been executed and set untracked->valid) ..
1559 */
1560 (!untracked || !untracked->valid ||
1561 /*
1562 * .. and .gitignore does not exist before
1563 * (i.e. null exclude_oid). Then we can skip
1564 * loading .gitignore, which would result in
1565 * ENOENT anyway.
1566 */
1567 !is_null_oid(&untracked->exclude_oid))) {
1568 /*
1569 * dir->basebuf gets reused by the traversal, but we
1570 * need fname to remain unchanged to ensure the src
1571 * member of each struct path_pattern correctly
1572 * back-references its source file. Other invocations
1573 * of add_pattern_list provide stable strings, so we
1574 * strbuf_detach() and free() here in the caller.
1575 */
1576 struct strbuf sb = STRBUF_INIT;
1577 strbuf_addbuf(&sb, &dir->basebuf);
1578 strbuf_addstr(&sb, dir->exclude_per_dir);
1579 pl->src = strbuf_detach(&sb, NULL);
1580 add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1581 PATTERN_NOFOLLOW,
1582 untracked ? &oid_stat : NULL);
1583 }
1584 /*
1585 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1586 * will first be called in valid_cached_dir() then maybe many
1587 * times more in last_matching_pattern(). When the cache is
1588 * used, last_matching_pattern() will not be called and
1589 * reading .gitignore content will be a waste.
1590 *
1591 * So when it's called by valid_cached_dir() and we can get
1592 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1593 * modified on work tree), we could delay reading the
1594 * .gitignore content until we absolutely need it in
1595 * last_matching_pattern(). Be careful about ignore rule
1596 * order, though, if you do that.
1597 */
1598 if (untracked &&
1599 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1600 invalidate_gitignore(dir->untracked, untracked);
1601 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1602 }
1603 dir->exclude_stack = stk;
1604 current = stk->baselen;
1605 }
1606 strbuf_setlen(&dir->basebuf, baselen);
1607 }
1608
1609 /*
1610 * Loads the exclude lists for the directory containing pathname, then
1611 * scans all exclude lists to determine whether pathname is excluded.
1612 * Returns the exclude_list element which matched, or NULL for
1613 * undecided.
1614 */
1615 struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1616 struct index_state *istate,
1617 const char *pathname,
1618 int *dtype_p)
1619 {
1620 int pathlen = strlen(pathname);
1621 const char *basename = strrchr(pathname, '/');
1622 basename = (basename) ? basename+1 : pathname;
1623
1624 prep_exclude(dir, istate, pathname, basename-pathname);
1625
1626 if (dir->pattern)
1627 return dir->pattern;
1628
1629 return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1630 basename, dtype_p);
1631 }
1632
1633 /*
1634 * Loads the exclude lists for the directory containing pathname, then
1635 * scans all exclude lists to determine whether pathname is excluded.
1636 * Returns 1 if true, otherwise 0.
1637 */
1638 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1639 const char *pathname, int *dtype_p)
1640 {
1641 struct path_pattern *pattern =
1642 last_matching_pattern(dir, istate, pathname, dtype_p);
1643 if (pattern)
1644 return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1645 return 0;
1646 }
1647
1648 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1649 {
1650 struct dir_entry *ent;
1651
1652 FLEX_ALLOC_MEM(ent, name, pathname, len);
1653 ent->len = len;
1654 return ent;
1655 }
1656
1657 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1658 struct index_state *istate,
1659 const char *pathname, int len)
1660 {
1661 if (index_file_exists(istate, pathname, len, ignore_case))
1662 return NULL;
1663
1664 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1665 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1666 }
1667
1668 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1669 struct index_state *istate,
1670 const char *pathname, int len)
1671 {
1672 if (!index_name_is_other(istate, pathname, len))
1673 return NULL;
1674
1675 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1676 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1677 }
1678
1679 enum exist_status {
1680 index_nonexistent = 0,
1681 index_directory,
1682 index_gitdir
1683 };
1684
1685 /*
1686 * Do not use the alphabetically sorted index to look up
1687 * the directory name; instead, use the case insensitive
1688 * directory hash.
1689 */
1690 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1691 const char *dirname, int len)
1692 {
1693 struct cache_entry *ce;
1694
1695 if (index_dir_exists(istate, dirname, len))
1696 return index_directory;
1697
1698 ce = index_file_exists(istate, dirname, len, ignore_case);
1699 if (ce && S_ISGITLINK(ce->ce_mode))
1700 return index_gitdir;
1701
1702 return index_nonexistent;
1703 }
1704
1705 /*
1706 * The index sorts alphabetically by entry name, which
1707 * means that a gitlink sorts as '\0' at the end, while
1708 * a directory (which is defined not as an entry, but as
1709 * the files it contains) will sort with the '/' at the
1710 * end.
1711 */
1712 static enum exist_status directory_exists_in_index(struct index_state *istate,
1713 const char *dirname, int len)
1714 {
1715 int pos;
1716
1717 if (ignore_case)
1718 return directory_exists_in_index_icase(istate, dirname, len);
1719
1720 pos = index_name_pos(istate, dirname, len);
1721 if (pos < 0)
1722 pos = -pos-1;
1723 while (pos < istate->cache_nr) {
1724 const struct cache_entry *ce = istate->cache[pos++];
1725 unsigned char endchar;
1726
1727 if (strncmp(ce->name, dirname, len))
1728 break;
1729 endchar = ce->name[len];
1730 if (endchar > '/')
1731 break;
1732 if (endchar == '/')
1733 return index_directory;
1734 if (!endchar && S_ISGITLINK(ce->ce_mode))
1735 return index_gitdir;
1736 }
1737 return index_nonexistent;
1738 }
1739
1740 /*
1741 * When we find a directory when traversing the filesystem, we
1742 * have three distinct cases:
1743 *
1744 * - ignore it
1745 * - see it as a directory
1746 * - recurse into it
1747 *
1748 * and which one we choose depends on a combination of existing
1749 * git index contents and the flags passed into the directory
1750 * traversal routine.
1751 *
1752 * Case 1: If we *already* have entries in the index under that
1753 * directory name, we always recurse into the directory to see
1754 * all the files.
1755 *
1756 * Case 2: If we *already* have that directory name as a gitlink,
1757 * we always continue to see it as a gitlink, regardless of whether
1758 * there is an actual git directory there or not (it might not
1759 * be checked out as a subproject!)
1760 *
1761 * Case 3: if we didn't have it in the index previously, we
1762 * have a few sub-cases:
1763 *
1764 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1765 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1766 * also true, in which case we need to check if it contains any
1767 * untracked and / or ignored files.
1768 * (b) if it looks like a git directory and we don't have the
1769 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1770 * show it as a directory.
1771 * (c) otherwise, we recurse into it.
1772 */
1773 static enum path_treatment treat_directory(struct dir_struct *dir,
1774 struct index_state *istate,
1775 struct untracked_cache_dir *untracked,
1776 const char *dirname, int len, int baselen, int excluded,
1777 const struct pathspec *pathspec)
1778 {
1779 /*
1780 * WARNING: From this function, you can return path_recurse or you
1781 * can call read_directory_recursive() (or neither), but
1782 * you CAN'T DO BOTH.
1783 */
1784 enum path_treatment state;
1785 int matches_how = 0;
1786 int nested_repo = 0, check_only, stop_early;
1787 int old_ignored_nr, old_untracked_nr;
1788 /* The "len-1" is to strip the final '/' */
1789 enum exist_status status = directory_exists_in_index(istate, dirname, len-1);
1790
1791 if (status == index_directory)
1792 return path_recurse;
1793 if (status == index_gitdir)
1794 return path_none;
1795 if (status != index_nonexistent)
1796 BUG("Unhandled value for directory_exists_in_index: %d\n", status);
1797
1798 /*
1799 * We don't want to descend into paths that don't match the necessary
1800 * patterns. Clearly, if we don't have a pathspec, then we can't check
1801 * for matching patterns. Also, if (excluded) then we know we matched
1802 * the exclusion patterns so as an optimization we can skip checking
1803 * for matching patterns.
1804 */
1805 if (pathspec && !excluded) {
1806 matches_how = match_pathspec_with_flags(istate, pathspec,
1807 dirname, len,
1808 0 /* prefix */,
1809 NULL /* seen */,
1810 DO_MATCH_LEADING_PATHSPEC);
1811 if (!matches_how)
1812 return path_none;
1813 }
1814
1815
1816 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1817 !(dir->flags & DIR_NO_GITLINKS)) {
1818 struct strbuf sb = STRBUF_INIT;
1819 strbuf_addstr(&sb, dirname);
1820 nested_repo = is_nonbare_repository_dir(&sb);
1821 strbuf_release(&sb);
1822 }
1823 if (nested_repo) {
1824 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1825 (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC))
1826 return path_none;
1827 return excluded ? path_excluded : path_untracked;
1828 }
1829
1830 if (!(dir->flags & DIR_SHOW_OTHER_DIRECTORIES)) {
1831 if (excluded &&
1832 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1833 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1834
1835 /*
1836 * This is an excluded directory and we are
1837 * showing ignored paths that match an exclude
1838 * pattern. (e.g. show directory as ignored
1839 * only if it matches an exclude pattern).
1840 * This path will either be 'path_excluded`
1841 * (if we are showing empty directories or if
1842 * the directory is not empty), or will be
1843 * 'path_none' (empty directory, and we are
1844 * not showing empty directories).
1845 */
1846 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1847 return path_excluded;
1848
1849 if (read_directory_recursive(dir, istate, dirname, len,
1850 untracked, 1, 1, pathspec) == path_excluded)
1851 return path_excluded;
1852
1853 return path_none;
1854 }
1855 return path_recurse;
1856 }
1857
1858 assert(dir->flags & DIR_SHOW_OTHER_DIRECTORIES);
1859
1860 /*
1861 * If we have a pathspec which could match something _below_ this
1862 * directory (e.g. when checking 'subdir/' having a pathspec like
1863 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1864 * need to recurse.
1865 */
1866 if (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC)
1867 return path_recurse;
1868
1869 /* Special cases for where this directory is excluded/ignored */
1870 if (excluded) {
1871 /*
1872 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1873 * hiding empty directories, there is no need to
1874 * recurse into an ignored directory.
1875 */
1876 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1877 return path_excluded;
1878
1879 /*
1880 * Even if we are hiding empty directories, we can still avoid
1881 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
1882 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1883 */
1884 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
1885 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1886 return path_excluded;
1887 }
1888
1889 /*
1890 * Other than the path_recurse case above, we only need to
1891 * recurse into untracked directories if any of the following
1892 * bits is set:
1893 * - DIR_SHOW_IGNORED (because then we need to determine if
1894 * there are ignored entries below)
1895 * - DIR_SHOW_IGNORED_TOO (same as above)
1896 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
1897 * the directory is empty)
1898 */
1899 if (!excluded &&
1900 !(dir->flags & (DIR_SHOW_IGNORED |
1901 DIR_SHOW_IGNORED_TOO |
1902 DIR_HIDE_EMPTY_DIRECTORIES))) {
1903 return path_untracked;
1904 }
1905
1906 /*
1907 * Even if we don't want to know all the paths under an untracked or
1908 * ignored directory, we may still need to go into the directory to
1909 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
1910 * an empty directory should be path_none instead of path_excluded or
1911 * path_untracked).
1912 */
1913 check_only = ((dir->flags & DIR_HIDE_EMPTY_DIRECTORIES) &&
1914 !(dir->flags & DIR_SHOW_IGNORED_TOO));
1915
1916 /*
1917 * However, there's another optimization possible as a subset of
1918 * check_only, based on the cases we have to consider:
1919 * A) Directory matches no exclude patterns:
1920 * * Directory is empty => path_none
1921 * * Directory has an untracked file under it => path_untracked
1922 * * Directory has only ignored files under it => path_excluded
1923 * B) Directory matches an exclude pattern:
1924 * * Directory is empty => path_none
1925 * * Directory has an untracked file under it => path_excluded
1926 * * Directory has only ignored files under it => path_excluded
1927 * In case A, we can exit as soon as we've found an untracked
1928 * file but otherwise have to walk all files. In case B, though,
1929 * we can stop at the first file we find under the directory.
1930 */
1931 stop_early = check_only && excluded;
1932
1933 /*
1934 * If /every/ file within an untracked directory is ignored, then
1935 * we want to treat the directory as ignored (for e.g. status
1936 * --porcelain), without listing the individual ignored files
1937 * underneath. To do so, we'll save the current ignored_nr, and
1938 * pop all the ones added after it if it turns out the entire
1939 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
1940 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
1941 * untracked paths so will need to pop all those off the last
1942 * after we traverse.
1943 */
1944 old_ignored_nr = dir->ignored_nr;
1945 old_untracked_nr = dir->nr;
1946
1947 /* Actually recurse into dirname now, we'll fixup the state later. */
1948 untracked = lookup_untracked(dir->untracked, untracked,
1949 dirname + baselen, len - baselen);
1950 state = read_directory_recursive(dir, istate, dirname, len, untracked,
1951 check_only, stop_early, pathspec);
1952
1953 /* There are a variety of reasons we may need to fixup the state... */
1954 if (state == path_excluded) {
1955 /* state == path_excluded implies all paths under
1956 * dirname were ignored...
1957 *
1958 * if running e.g. `git status --porcelain --ignored=matching`,
1959 * then we want to see the subpaths that are ignored.
1960 *
1961 * if running e.g. just `git status --porcelain`, then
1962 * we just want the directory itself to be listed as ignored
1963 * and not the individual paths underneath.
1964 */
1965 int want_ignored_subpaths =
1966 ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
1967 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING));
1968
1969 if (want_ignored_subpaths) {
1970 /*
1971 * with --ignored=matching, we want the subpaths
1972 * INSTEAD of the directory itself.
1973 */
1974 state = path_none;
1975 } else {
1976 int i;
1977 for (i = old_ignored_nr + 1; i<dir->ignored_nr; ++i)
1978 FREE_AND_NULL(dir->ignored[i]);
1979 dir->ignored_nr = old_ignored_nr;
1980 }
1981 }
1982
1983 /*
1984 * We may need to ignore some of the untracked paths we found while
1985 * traversing subdirectories.
1986 */
1987 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
1988 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
1989 int i;
1990 for (i = old_untracked_nr + 1; i<dir->nr; ++i)
1991 FREE_AND_NULL(dir->entries[i]);
1992 dir->nr = old_untracked_nr;
1993 }
1994
1995 /*
1996 * If there is nothing under the current directory and we are not
1997 * hiding empty directories, then we need to report on the
1998 * untracked or ignored status of the directory itself.
1999 */
2000 if (state == path_none && !(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
2001 state = excluded ? path_excluded : path_untracked;
2002
2003 return state;
2004 }
2005
2006 /*
2007 * This is an inexact early pruning of any recursive directory
2008 * reading - if the path cannot possibly be in the pathspec,
2009 * return true, and we'll skip it early.
2010 */
2011 static int simplify_away(const char *path, int pathlen,
2012 const struct pathspec *pathspec)
2013 {
2014 int i;
2015
2016 if (!pathspec || !pathspec->nr)
2017 return 0;
2018
2019 GUARD_PATHSPEC(pathspec,
2020 PATHSPEC_FROMTOP |
2021 PATHSPEC_MAXDEPTH |
2022 PATHSPEC_LITERAL |
2023 PATHSPEC_GLOB |
2024 PATHSPEC_ICASE |
2025 PATHSPEC_EXCLUDE |
2026 PATHSPEC_ATTR);
2027
2028 for (i = 0; i < pathspec->nr; i++) {
2029 const struct pathspec_item *item = &pathspec->items[i];
2030 int len = item->nowildcard_len;
2031
2032 if (len > pathlen)
2033 len = pathlen;
2034 if (!ps_strncmp(item, item->match, path, len))
2035 return 0;
2036 }
2037
2038 return 1;
2039 }
2040
2041 /*
2042 * This function tells us whether an excluded path matches a
2043 * list of "interesting" pathspecs. That is, whether a path matched
2044 * by any of the pathspecs could possibly be ignored by excluding
2045 * the specified path. This can happen if:
2046 *
2047 * 1. the path is mentioned explicitly in the pathspec
2048 *
2049 * 2. the path is a directory prefix of some element in the
2050 * pathspec
2051 */
2052 static int exclude_matches_pathspec(const char *path, int pathlen,
2053 const struct pathspec *pathspec)
2054 {
2055 int i;
2056
2057 if (!pathspec || !pathspec->nr)
2058 return 0;
2059
2060 GUARD_PATHSPEC(pathspec,
2061 PATHSPEC_FROMTOP |
2062 PATHSPEC_MAXDEPTH |
2063 PATHSPEC_LITERAL |
2064 PATHSPEC_GLOB |
2065 PATHSPEC_ICASE |
2066 PATHSPEC_EXCLUDE);
2067
2068 for (i = 0; i < pathspec->nr; i++) {
2069 const struct pathspec_item *item = &pathspec->items[i];
2070 int len = item->nowildcard_len;
2071
2072 if (len == pathlen &&
2073 !ps_strncmp(item, item->match, path, pathlen))
2074 return 1;
2075 if (len > pathlen &&
2076 item->match[pathlen] == '/' &&
2077 !ps_strncmp(item, item->match, path, pathlen))
2078 return 1;
2079 }
2080 return 0;
2081 }
2082
2083 static int get_index_dtype(struct index_state *istate,
2084 const char *path, int len)
2085 {
2086 int pos;
2087 const struct cache_entry *ce;
2088
2089 ce = index_file_exists(istate, path, len, 0);
2090 if (ce) {
2091 if (!ce_uptodate(ce))
2092 return DT_UNKNOWN;
2093 if (S_ISGITLINK(ce->ce_mode))
2094 return DT_DIR;
2095 /*
2096 * Nobody actually cares about the
2097 * difference between DT_LNK and DT_REG
2098 */
2099 return DT_REG;
2100 }
2101
2102 /* Try to look it up as a directory */
2103 pos = index_name_pos(istate, path, len);
2104 if (pos >= 0)
2105 return DT_UNKNOWN;
2106 pos = -pos-1;
2107 while (pos < istate->cache_nr) {
2108 ce = istate->cache[pos++];
2109 if (strncmp(ce->name, path, len))
2110 break;
2111 if (ce->name[len] > '/')
2112 break;
2113 if (ce->name[len] < '/')
2114 continue;
2115 if (!ce_uptodate(ce))
2116 break; /* continue? */
2117 return DT_DIR;
2118 }
2119 return DT_UNKNOWN;
2120 }
2121
2122 static int resolve_dtype(int dtype, struct index_state *istate,
2123 const char *path, int len)
2124 {
2125 struct stat st;
2126
2127 if (dtype != DT_UNKNOWN)
2128 return dtype;
2129 dtype = get_index_dtype(istate, path, len);
2130 if (dtype != DT_UNKNOWN)
2131 return dtype;
2132 if (lstat(path, &st))
2133 return dtype;
2134 if (S_ISREG(st.st_mode))
2135 return DT_REG;
2136 if (S_ISDIR(st.st_mode))
2137 return DT_DIR;
2138 if (S_ISLNK(st.st_mode))
2139 return DT_LNK;
2140 return dtype;
2141 }
2142
2143 static enum path_treatment treat_path_fast(struct dir_struct *dir,
2144 struct cached_dir *cdir,
2145 struct index_state *istate,
2146 struct strbuf *path,
2147 int baselen,
2148 const struct pathspec *pathspec)
2149 {
2150 /*
2151 * WARNING: From this function, you can return path_recurse or you
2152 * can call read_directory_recursive() (or neither), but
2153 * you CAN'T DO BOTH.
2154 */
2155 strbuf_setlen(path, baselen);
2156 if (!cdir->ucd) {
2157 strbuf_addstr(path, cdir->file);
2158 return path_untracked;
2159 }
2160 strbuf_addstr(path, cdir->ucd->name);
2161 /* treat_one_path() does this before it calls treat_directory() */
2162 strbuf_complete(path, '/');
2163 if (cdir->ucd->check_only)
2164 /*
2165 * check_only is set as a result of treat_directory() getting
2166 * to its bottom. Verify again the same set of directories
2167 * with check_only set.
2168 */
2169 return read_directory_recursive(dir, istate, path->buf, path->len,
2170 cdir->ucd, 1, 0, pathspec);
2171 /*
2172 * We get path_recurse in the first run when
2173 * directory_exists_in_index() returns index_nonexistent. We
2174 * are sure that new changes in the index does not impact the
2175 * outcome. Return now.
2176 */
2177 return path_recurse;
2178 }
2179
2180 static enum path_treatment treat_path(struct dir_struct *dir,
2181 struct untracked_cache_dir *untracked,
2182 struct cached_dir *cdir,
2183 struct index_state *istate,
2184 struct strbuf *path,
2185 int baselen,
2186 const struct pathspec *pathspec)
2187 {
2188 int has_path_in_index, dtype, excluded;
2189
2190 if (!cdir->d_name)
2191 return treat_path_fast(dir, cdir, istate, path,
2192 baselen, pathspec);
2193 if (is_dot_or_dotdot(cdir->d_name) || !fspathcmp(cdir->d_name, ".git"))
2194 return path_none;
2195 strbuf_setlen(path, baselen);
2196 strbuf_addstr(path, cdir->d_name);
2197 if (simplify_away(path->buf, path->len, pathspec))
2198 return path_none;
2199
2200 dtype = resolve_dtype(cdir->d_type, istate, path->buf, path->len);
2201
2202 /* Always exclude indexed files */
2203 has_path_in_index = !!index_file_exists(istate, path->buf, path->len,
2204 ignore_case);
2205 if (dtype != DT_DIR && has_path_in_index)
2206 return path_none;
2207
2208 /*
2209 * When we are looking at a directory P in the working tree,
2210 * there are three cases:
2211 *
2212 * (1) P exists in the index. Everything inside the directory P in
2213 * the working tree needs to go when P is checked out from the
2214 * index.
2215 *
2216 * (2) P does not exist in the index, but there is P/Q in the index.
2217 * We know P will stay a directory when we check out the contents
2218 * of the index, but we do not know yet if there is a directory
2219 * P/Q in the working tree to be killed, so we need to recurse.
2220 *
2221 * (3) P does not exist in the index, and there is no P/Q in the index
2222 * to require P to be a directory, either. Only in this case, we
2223 * know that everything inside P will not be killed without
2224 * recursing.
2225 */
2226 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
2227 (dtype == DT_DIR) &&
2228 !has_path_in_index &&
2229 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
2230 return path_none;
2231
2232 excluded = is_excluded(dir, istate, path->buf, &dtype);
2233
2234 /*
2235 * Excluded? If we don't explicitly want to show
2236 * ignored files, ignore it
2237 */
2238 if (excluded && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
2239 return path_excluded;
2240
2241 switch (dtype) {
2242 default:
2243 return path_none;
2244 case DT_DIR:
2245 /*
2246 * WARNING: Do not ignore/amend the return value from
2247 * treat_directory(), and especially do not change it to return
2248 * path_recurse as that can cause exponential slowdown.
2249 * Instead, modify treat_directory() to return the right value.
2250 */
2251 strbuf_addch(path, '/');
2252 return treat_directory(dir, istate, untracked,
2253 path->buf, path->len,
2254 baselen, excluded, pathspec);
2255 case DT_REG:
2256 case DT_LNK:
2257 if (pathspec &&
2258 !match_pathspec(istate, pathspec, path->buf, path->len,
2259 0 /* prefix */, NULL /* seen */,
2260 0 /* is_dir */))
2261 return path_none;
2262 if (excluded)
2263 return path_excluded;
2264 return path_untracked;
2265 }
2266 }
2267
2268 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
2269 {
2270 if (!dir)
2271 return;
2272 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
2273 dir->untracked_alloc);
2274 dir->untracked[dir->untracked_nr++] = xstrdup(name);
2275 }
2276
2277 static int valid_cached_dir(struct dir_struct *dir,
2278 struct untracked_cache_dir *untracked,
2279 struct index_state *istate,
2280 struct strbuf *path,
2281 int check_only)
2282 {
2283 struct stat st;
2284
2285 if (!untracked)
2286 return 0;
2287
2288 /*
2289 * With fsmonitor, we can trust the untracked cache's valid field.
2290 */
2291 refresh_fsmonitor(istate);
2292 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
2293 if (lstat(path->len ? path->buf : ".", &st)) {
2294 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
2295 return 0;
2296 }
2297 if (!untracked->valid ||
2298 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
2299 fill_stat_data(&untracked->stat_data, &st);
2300 return 0;
2301 }
2302 }
2303
2304 if (untracked->check_only != !!check_only)
2305 return 0;
2306
2307 /*
2308 * prep_exclude will be called eventually on this directory,
2309 * but it's called much later in last_matching_pattern(). We
2310 * need it now to determine the validity of the cache for this
2311 * path. The next calls will be nearly no-op, the way
2312 * prep_exclude() is designed.
2313 */
2314 if (path->len && path->buf[path->len - 1] != '/') {
2315 strbuf_addch(path, '/');
2316 prep_exclude(dir, istate, path->buf, path->len);
2317 strbuf_setlen(path, path->len - 1);
2318 } else
2319 prep_exclude(dir, istate, path->buf, path->len);
2320
2321 /* hopefully prep_exclude() haven't invalidated this entry... */
2322 return untracked->valid;
2323 }
2324
2325 static int open_cached_dir(struct cached_dir *cdir,
2326 struct dir_struct *dir,
2327 struct untracked_cache_dir *untracked,
2328 struct index_state *istate,
2329 struct strbuf *path,
2330 int check_only)
2331 {
2332 const char *c_path;
2333
2334 memset(cdir, 0, sizeof(*cdir));
2335 cdir->untracked = untracked;
2336 if (valid_cached_dir(dir, untracked, istate, path, check_only))
2337 return 0;
2338 c_path = path->len ? path->buf : ".";
2339 cdir->fdir = opendir(c_path);
2340 if (!cdir->fdir)
2341 warning_errno(_("could not open directory '%s'"), c_path);
2342 if (dir->untracked) {
2343 invalidate_directory(dir->untracked, untracked);
2344 dir->untracked->dir_opened++;
2345 }
2346 if (!cdir->fdir)
2347 return -1;
2348 return 0;
2349 }
2350
2351 static int read_cached_dir(struct cached_dir *cdir)
2352 {
2353 struct dirent *de;
2354
2355 if (cdir->fdir) {
2356 de = readdir_skip_dot_and_dotdot(cdir->fdir);
2357 if (!de) {
2358 cdir->d_name = NULL;
2359 cdir->d_type = DT_UNKNOWN;
2360 return -1;
2361 }
2362 cdir->d_name = de->d_name;
2363 cdir->d_type = DTYPE(de);
2364 return 0;
2365 }
2366 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
2367 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
2368 if (!d->recurse) {
2369 cdir->nr_dirs++;
2370 continue;
2371 }
2372 cdir->ucd = d;
2373 cdir->nr_dirs++;
2374 return 0;
2375 }
2376 cdir->ucd = NULL;
2377 if (cdir->nr_files < cdir->untracked->untracked_nr) {
2378 struct untracked_cache_dir *d = cdir->untracked;
2379 cdir->file = d->untracked[cdir->nr_files++];
2380 return 0;
2381 }
2382 return -1;
2383 }
2384
2385 static void close_cached_dir(struct cached_dir *cdir)
2386 {
2387 if (cdir->fdir)
2388 closedir(cdir->fdir);
2389 /*
2390 * We have gone through this directory and found no untracked
2391 * entries. Mark it valid.
2392 */
2393 if (cdir->untracked) {
2394 cdir->untracked->valid = 1;
2395 cdir->untracked->recurse = 1;
2396 }
2397 }
2398
2399 static void add_path_to_appropriate_result_list(struct dir_struct *dir,
2400 struct untracked_cache_dir *untracked,
2401 struct cached_dir *cdir,
2402 struct index_state *istate,
2403 struct strbuf *path,
2404 int baselen,
2405 const struct pathspec *pathspec,
2406 enum path_treatment state)
2407 {
2408 /* add the path to the appropriate result list */
2409 switch (state) {
2410 case path_excluded:
2411 if (dir->flags & DIR_SHOW_IGNORED)
2412 dir_add_name(dir, istate, path->buf, path->len);
2413 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2414 ((dir->flags & DIR_COLLECT_IGNORED) &&
2415 exclude_matches_pathspec(path->buf, path->len,
2416 pathspec)))
2417 dir_add_ignored(dir, istate, path->buf, path->len);
2418 break;
2419
2420 case path_untracked:
2421 if (dir->flags & DIR_SHOW_IGNORED)
2422 break;
2423 dir_add_name(dir, istate, path->buf, path->len);
2424 if (cdir->fdir)
2425 add_untracked(untracked, path->buf + baselen);
2426 break;
2427
2428 default:
2429 break;
2430 }
2431 }
2432
2433 /*
2434 * Read a directory tree. We currently ignore anything but
2435 * directories, regular files and symlinks. That's because git
2436 * doesn't handle them at all yet. Maybe that will change some
2437 * day.
2438 *
2439 * Also, we ignore the name ".git" (even if it is not a directory).
2440 * That likely will not change.
2441 *
2442 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2443 * to signal that a file was found. This is the least significant value that
2444 * indicates that a file was encountered that does not depend on the order of
2445 * whether an untracked or excluded path was encountered first.
2446 *
2447 * Returns the most significant path_treatment value encountered in the scan.
2448 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2449 * significant path_treatment value that will be returned.
2450 */
2451
2452 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
2453 struct index_state *istate, const char *base, int baselen,
2454 struct untracked_cache_dir *untracked, int check_only,
2455 int stop_at_first_file, const struct pathspec *pathspec)
2456 {
2457 /*
2458 * WARNING: Do NOT recurse unless path_recurse is returned from
2459 * treat_path(). Recursing on any other return value
2460 * can result in exponential slowdown.
2461 */
2462 struct cached_dir cdir;
2463 enum path_treatment state, subdir_state, dir_state = path_none;
2464 struct strbuf path = STRBUF_INIT;
2465
2466 strbuf_add(&path, base, baselen);
2467
2468 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
2469 goto out;
2470 dir->visited_directories++;
2471
2472 if (untracked)
2473 untracked->check_only = !!check_only;
2474
2475 while (!read_cached_dir(&cdir)) {
2476 /* check how the file or directory should be treated */
2477 state = treat_path(dir, untracked, &cdir, istate, &path,
2478 baselen, pathspec);
2479 dir->visited_paths++;
2480
2481 if (state > dir_state)
2482 dir_state = state;
2483
2484 /* recurse into subdir if instructed by treat_path */
2485 if (state == path_recurse) {
2486 struct untracked_cache_dir *ud;
2487 ud = lookup_untracked(dir->untracked, untracked,
2488 path.buf + baselen,
2489 path.len - baselen);
2490 subdir_state =
2491 read_directory_recursive(dir, istate, path.buf,
2492 path.len, ud,
2493 check_only, stop_at_first_file, pathspec);
2494 if (subdir_state > dir_state)
2495 dir_state = subdir_state;
2496
2497 if (pathspec &&
2498 !match_pathspec(istate, pathspec, path.buf, path.len,
2499 0 /* prefix */, NULL,
2500 0 /* do NOT special case dirs */))
2501 state = path_none;
2502 }
2503
2504 if (check_only) {
2505 if (stop_at_first_file) {
2506 /*
2507 * If stopping at first file, then
2508 * signal that a file was found by
2509 * returning `path_excluded`. This is
2510 * to return a consistent value
2511 * regardless of whether an ignored or
2512 * excluded file happened to be
2513 * encountered 1st.
2514 *
2515 * In current usage, the
2516 * `stop_at_first_file` is passed when
2517 * an ancestor directory has matched
2518 * an exclude pattern, so any found
2519 * files will be excluded.
2520 */
2521 if (dir_state >= path_excluded) {
2522 dir_state = path_excluded;
2523 break;
2524 }
2525 }
2526
2527 /* abort early if maximum state has been reached */
2528 if (dir_state == path_untracked) {
2529 if (cdir.fdir)
2530 add_untracked(untracked, path.buf + baselen);
2531 break;
2532 }
2533 /* skip the add_path_to_appropriate_result_list() */
2534 continue;
2535 }
2536
2537 add_path_to_appropriate_result_list(dir, untracked, &cdir,
2538 istate, &path, baselen,
2539 pathspec, state);
2540 }
2541 close_cached_dir(&cdir);
2542 out:
2543 strbuf_release(&path);
2544
2545 return dir_state;
2546 }
2547
2548 int cmp_dir_entry(const void *p1, const void *p2)
2549 {
2550 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2551 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2552
2553 return name_compare(e1->name, e1->len, e2->name, e2->len);
2554 }
2555
2556 /* check if *out lexically strictly contains *in */
2557 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2558 {
2559 return (out->len < in->len) &&
2560 (out->name[out->len - 1] == '/') &&
2561 !memcmp(out->name, in->name, out->len);
2562 }
2563
2564 static int treat_leading_path(struct dir_struct *dir,
2565 struct index_state *istate,
2566 const char *path, int len,
2567 const struct pathspec *pathspec)
2568 {
2569 struct strbuf sb = STRBUF_INIT;
2570 struct strbuf subdir = STRBUF_INIT;
2571 int prevlen, baselen;
2572 const char *cp;
2573 struct cached_dir cdir;
2574 enum path_treatment state = path_none;
2575
2576 /*
2577 * For each directory component of path, we are going to check whether
2578 * that path is relevant given the pathspec. For example, if path is
2579 * foo/bar/baz/
2580 * then we will ask treat_path() whether we should go into foo, then
2581 * whether we should go into bar, then whether baz is relevant.
2582 * Checking each is important because e.g. if path is
2583 * .git/info/
2584 * then we need to check .git to know we shouldn't traverse it.
2585 * If the return from treat_path() is:
2586 * * path_none, for any path, we return false.
2587 * * path_recurse, for all path components, we return true
2588 * * <anything else> for some intermediate component, we make sure
2589 * to add that path to the relevant list but return false
2590 * signifying that we shouldn't recurse into it.
2591 */
2592
2593 while (len && path[len - 1] == '/')
2594 len--;
2595 if (!len)
2596 return 1;
2597
2598 memset(&cdir, 0, sizeof(cdir));
2599 cdir.d_type = DT_DIR;
2600 baselen = 0;
2601 prevlen = 0;
2602 while (1) {
2603 prevlen = baselen + !!baselen;
2604 cp = path + prevlen;
2605 cp = memchr(cp, '/', path + len - cp);
2606 if (!cp)
2607 baselen = len;
2608 else
2609 baselen = cp - path;
2610 strbuf_reset(&sb);
2611 strbuf_add(&sb, path, baselen);
2612 if (!is_directory(sb.buf))
2613 break;
2614 strbuf_reset(&sb);
2615 strbuf_add(&sb, path, prevlen);
2616 strbuf_reset(&subdir);
2617 strbuf_add(&subdir, path+prevlen, baselen-prevlen);
2618 cdir.d_name = subdir.buf;
2619 state = treat_path(dir, NULL, &cdir, istate, &sb, prevlen, pathspec);
2620
2621 if (state != path_recurse)
2622 break; /* do not recurse into it */
2623 if (len <= baselen)
2624 break; /* finished checking */
2625 }
2626 add_path_to_appropriate_result_list(dir, NULL, &cdir, istate,
2627 &sb, baselen, pathspec,
2628 state);
2629
2630 strbuf_release(&subdir);
2631 strbuf_release(&sb);
2632 return state == path_recurse;
2633 }
2634
2635 static const char *get_ident_string(void)
2636 {
2637 static struct strbuf sb = STRBUF_INIT;
2638 struct utsname uts;
2639
2640 if (sb.len)
2641 return sb.buf;
2642 if (uname(&uts) < 0)
2643 die_errno(_("failed to get kernel name and information"));
2644 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2645 uts.sysname);
2646 return sb.buf;
2647 }
2648
2649 static int ident_in_untracked(const struct untracked_cache *uc)
2650 {
2651 /*
2652 * Previous git versions may have saved many NUL separated
2653 * strings in the "ident" field, but it is insane to manage
2654 * many locations, so just take care of the first one.
2655 */
2656
2657 return !strcmp(uc->ident.buf, get_ident_string());
2658 }
2659
2660 static void set_untracked_ident(struct untracked_cache *uc)
2661 {
2662 strbuf_reset(&uc->ident);
2663 strbuf_addstr(&uc->ident, get_ident_string());
2664
2665 /*
2666 * This strbuf used to contain a list of NUL separated
2667 * strings, so save NUL too for backward compatibility.
2668 */
2669 strbuf_addch(&uc->ident, 0);
2670 }
2671
2672 static void new_untracked_cache(struct index_state *istate)
2673 {
2674 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2675 strbuf_init(&uc->ident, 100);
2676 uc->exclude_per_dir = ".gitignore";
2677 /* should be the same flags used by git-status */
2678 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2679 set_untracked_ident(uc);
2680 istate->untracked = uc;
2681 istate->cache_changed |= UNTRACKED_CHANGED;
2682 }
2683
2684 void add_untracked_cache(struct index_state *istate)
2685 {
2686 if (!istate->untracked) {
2687 new_untracked_cache(istate);
2688 } else {
2689 if (!ident_in_untracked(istate->untracked)) {
2690 free_untracked_cache(istate->untracked);
2691 new_untracked_cache(istate);
2692 }
2693 }
2694 }
2695
2696 void remove_untracked_cache(struct index_state *istate)
2697 {
2698 if (istate->untracked) {
2699 free_untracked_cache(istate->untracked);
2700 istate->untracked = NULL;
2701 istate->cache_changed |= UNTRACKED_CHANGED;
2702 }
2703 }
2704
2705 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2706 int base_len,
2707 const struct pathspec *pathspec)
2708 {
2709 struct untracked_cache_dir *root;
2710 static int untracked_cache_disabled = -1;
2711
2712 if (!dir->untracked)
2713 return NULL;
2714 if (untracked_cache_disabled < 0)
2715 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2716 if (untracked_cache_disabled)
2717 return NULL;
2718
2719 /*
2720 * We only support $GIT_DIR/info/exclude and core.excludesfile
2721 * as the global ignore rule files. Any other additions
2722 * (e.g. from command line) invalidate the cache. This
2723 * condition also catches running setup_standard_excludes()
2724 * before setting dir->untracked!
2725 */
2726 if (dir->unmanaged_exclude_files)
2727 return NULL;
2728
2729 /*
2730 * Optimize for the main use case only: whole-tree git
2731 * status. More work involved in treat_leading_path() if we
2732 * use cache on just a subset of the worktree. pathspec
2733 * support could make the matter even worse.
2734 */
2735 if (base_len || (pathspec && pathspec->nr))
2736 return NULL;
2737
2738 /* Different set of flags may produce different results */
2739 if (dir->flags != dir->untracked->dir_flags ||
2740 /*
2741 * See treat_directory(), case index_nonexistent. Without
2742 * this flag, we may need to also cache .git file content
2743 * for the resolve_gitlink_ref() call, which we don't.
2744 */
2745 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2746 /* We don't support collecting ignore files */
2747 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2748 DIR_COLLECT_IGNORED)))
2749 return NULL;
2750
2751 /*
2752 * If we use .gitignore in the cache and now you change it to
2753 * .gitexclude, everything will go wrong.
2754 */
2755 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2756 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2757 return NULL;
2758
2759 /*
2760 * EXC_CMDL is not considered in the cache. If people set it,
2761 * skip the cache.
2762 */
2763 if (dir->exclude_list_group[EXC_CMDL].nr)
2764 return NULL;
2765
2766 if (!ident_in_untracked(dir->untracked)) {
2767 warning(_("untracked cache is disabled on this system or location"));
2768 return NULL;
2769 }
2770
2771 if (!dir->untracked->root)
2772 FLEX_ALLOC_STR(dir->untracked->root, name, "");
2773
2774 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2775 root = dir->untracked->root;
2776 if (!oideq(&dir->ss_info_exclude.oid,
2777 &dir->untracked->ss_info_exclude.oid)) {
2778 invalidate_gitignore(dir->untracked, root);
2779 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2780 }
2781 if (!oideq(&dir->ss_excludes_file.oid,
2782 &dir->untracked->ss_excludes_file.oid)) {
2783 invalidate_gitignore(dir->untracked, root);
2784 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2785 }
2786
2787 /* Make sure this directory is not dropped out at saving phase */
2788 root->recurse = 1;
2789 return root;
2790 }
2791
2792 static void emit_traversal_statistics(struct dir_struct *dir,
2793 struct repository *repo,
2794 const char *path,
2795 int path_len)
2796 {
2797 if (!trace2_is_enabled())
2798 return;
2799
2800 if (!path_len) {
2801 trace2_data_string("read_directory", repo, "path", "");
2802 } else {
2803 struct strbuf tmp = STRBUF_INIT;
2804 strbuf_add(&tmp, path, path_len);
2805 trace2_data_string("read_directory", repo, "path", tmp.buf);
2806 strbuf_release(&tmp);
2807 }
2808
2809 trace2_data_intmax("read_directory", repo,
2810 "directories-visited", dir->visited_directories);
2811 trace2_data_intmax("read_directory", repo,
2812 "paths-visited", dir->visited_paths);
2813
2814 if (!dir->untracked)
2815 return;
2816 trace2_data_intmax("read_directory", repo,
2817 "node-creation", dir->untracked->dir_created);
2818 trace2_data_intmax("read_directory", repo,
2819 "gitignore-invalidation",
2820 dir->untracked->gitignore_invalidated);
2821 trace2_data_intmax("read_directory", repo,
2822 "directory-invalidation",
2823 dir->untracked->dir_invalidated);
2824 trace2_data_intmax("read_directory", repo,
2825 "opendir", dir->untracked->dir_opened);
2826 }
2827
2828 int read_directory(struct dir_struct *dir, struct index_state *istate,
2829 const char *path, int len, const struct pathspec *pathspec)
2830 {
2831 struct untracked_cache_dir *untracked;
2832
2833 trace2_region_enter("dir", "read_directory", istate->repo);
2834 dir->visited_paths = 0;
2835 dir->visited_directories = 0;
2836
2837 if (has_symlink_leading_path(path, len)) {
2838 trace2_region_leave("dir", "read_directory", istate->repo);
2839 return dir->nr;
2840 }
2841
2842 untracked = validate_untracked_cache(dir, len, pathspec);
2843 if (!untracked)
2844 /*
2845 * make sure untracked cache code path is disabled,
2846 * e.g. prep_exclude()
2847 */
2848 dir->untracked = NULL;
2849 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2850 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2851 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2852 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2853
2854 emit_traversal_statistics(dir, istate->repo, path, len);
2855
2856 trace2_region_leave("dir", "read_directory", istate->repo);
2857 if (dir->untracked) {
2858 static int force_untracked_cache = -1;
2859
2860 if (force_untracked_cache < 0)
2861 force_untracked_cache =
2862 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2863 if (force_untracked_cache &&
2864 dir->untracked == istate->untracked &&
2865 (dir->untracked->dir_opened ||
2866 dir->untracked->gitignore_invalidated ||
2867 dir->untracked->dir_invalidated))
2868 istate->cache_changed |= UNTRACKED_CHANGED;
2869 if (dir->untracked != istate->untracked) {
2870 FREE_AND_NULL(dir->untracked);
2871 }
2872 }
2873
2874 return dir->nr;
2875 }
2876
2877 int file_exists(const char *f)
2878 {
2879 struct stat sb;
2880 return lstat(f, &sb) == 0;
2881 }
2882
2883 int repo_file_exists(struct repository *repo, const char *path)
2884 {
2885 if (repo != the_repository)
2886 BUG("do not know how to check file existence in arbitrary repo");
2887
2888 return file_exists(path);
2889 }
2890
2891 static int cmp_icase(char a, char b)
2892 {
2893 if (a == b)
2894 return 0;
2895 if (ignore_case)
2896 return toupper(a) - toupper(b);
2897 return a - b;
2898 }
2899
2900 /*
2901 * Given two normalized paths (a trailing slash is ok), if subdir is
2902 * outside dir, return -1. Otherwise return the offset in subdir that
2903 * can be used as relative path to dir.
2904 */
2905 int dir_inside_of(const char *subdir, const char *dir)
2906 {
2907 int offset = 0;
2908
2909 assert(dir && subdir && *dir && *subdir);
2910
2911 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2912 dir++;
2913 subdir++;
2914 offset++;
2915 }
2916
2917 /* hel[p]/me vs hel[l]/yeah */
2918 if (*dir && *subdir)
2919 return -1;
2920
2921 if (!*subdir)
2922 return !*dir ? offset : -1; /* same dir */
2923
2924 /* foo/[b]ar vs foo/[] */
2925 if (is_dir_sep(dir[-1]))
2926 return is_dir_sep(subdir[-1]) ? offset : -1;
2927
2928 /* foo[/]bar vs foo[] */
2929 return is_dir_sep(*subdir) ? offset + 1 : -1;
2930 }
2931
2932 int is_inside_dir(const char *dir)
2933 {
2934 char *cwd;
2935 int rc;
2936
2937 if (!dir)
2938 return 0;
2939
2940 cwd = xgetcwd();
2941 rc = (dir_inside_of(cwd, dir) >= 0);
2942 free(cwd);
2943 return rc;
2944 }
2945
2946 int is_empty_dir(const char *path)
2947 {
2948 DIR *dir = opendir(path);
2949 struct dirent *e;
2950 int ret = 1;
2951
2952 if (!dir)
2953 return 0;
2954
2955 e = readdir_skip_dot_and_dotdot(dir);
2956 if (e)
2957 ret = 0;
2958
2959 closedir(dir);
2960 return ret;
2961 }
2962
2963 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2964 {
2965 DIR *dir;
2966 struct dirent *e;
2967 int ret = 0, original_len = path->len, len, kept_down = 0;
2968 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2969 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2970 struct object_id submodule_head;
2971
2972 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2973 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2974 /* Do not descend and nuke a nested git work tree. */
2975 if (kept_up)
2976 *kept_up = 1;
2977 return 0;
2978 }
2979
2980 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2981 dir = opendir(path->buf);
2982 if (!dir) {
2983 if (errno == ENOENT)
2984 return keep_toplevel ? -1 : 0;
2985 else if (errno == EACCES && !keep_toplevel)
2986 /*
2987 * An empty dir could be removable even if it
2988 * is unreadable:
2989 */
2990 return rmdir(path->buf);
2991 else
2992 return -1;
2993 }
2994 strbuf_complete(path, '/');
2995
2996 len = path->len;
2997 while ((e = readdir_skip_dot_and_dotdot(dir)) != NULL) {
2998 struct stat st;
2999
3000 strbuf_setlen(path, len);
3001 strbuf_addstr(path, e->d_name);
3002 if (lstat(path->buf, &st)) {
3003 if (errno == ENOENT)
3004 /*
3005 * file disappeared, which is what we
3006 * wanted anyway
3007 */
3008 continue;
3009 /* fall through */
3010 } else if (S_ISDIR(st.st_mode)) {
3011 if (!remove_dir_recurse(path, flag, &kept_down))
3012 continue; /* happy */
3013 } else if (!only_empty &&
3014 (!unlink(path->buf) || errno == ENOENT)) {
3015 continue; /* happy, too */
3016 }
3017
3018 /* path too long, stat fails, or non-directory still exists */
3019 ret = -1;
3020 break;
3021 }
3022 closedir(dir);
3023
3024 strbuf_setlen(path, original_len);
3025 if (!ret && !keep_toplevel && !kept_down)
3026 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
3027 else if (kept_up)
3028 /*
3029 * report the uplevel that it is not an error that we
3030 * did not rmdir() our directory.
3031 */
3032 *kept_up = !ret;
3033 return ret;
3034 }
3035
3036 int remove_dir_recursively(struct strbuf *path, int flag)
3037 {
3038 return remove_dir_recurse(path, flag, NULL);
3039 }
3040
3041 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
3042
3043 void setup_standard_excludes(struct dir_struct *dir)
3044 {
3045 dir->exclude_per_dir = ".gitignore";
3046
3047 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3048 if (!excludes_file)
3049 excludes_file = xdg_config_home("ignore");
3050 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
3051 add_patterns_from_file_1(dir, excludes_file,
3052 dir->untracked ? &dir->ss_excludes_file : NULL);
3053
3054 /* per repository user preference */
3055 if (startup_info->have_repository) {
3056 const char *path = git_path_info_exclude();
3057 if (!access_or_warn(path, R_OK, 0))
3058 add_patterns_from_file_1(dir, path,
3059 dir->untracked ? &dir->ss_info_exclude : NULL);
3060 }
3061 }
3062
3063 char *get_sparse_checkout_filename(void)
3064 {
3065 return git_pathdup("info/sparse-checkout");
3066 }
3067
3068 int get_sparse_checkout_patterns(struct pattern_list *pl)
3069 {
3070 int res;
3071 char *sparse_filename = get_sparse_checkout_filename();
3072
3073 pl->use_cone_patterns = core_sparse_checkout_cone;
3074 res = add_patterns_from_file_to_list(sparse_filename, "", 0, pl, NULL, 0);
3075
3076 free(sparse_filename);
3077 return res;
3078 }
3079
3080 int remove_path(const char *name)
3081 {
3082 char *slash;
3083
3084 if (unlink(name) && !is_missing_file_error(errno))
3085 return -1;
3086
3087 slash = strrchr(name, '/');
3088 if (slash) {
3089 char *dirs = xstrdup(name);
3090 slash = dirs + (slash - name);
3091 do {
3092 *slash = '\0';
3093 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
3094 free(dirs);
3095 }
3096 return 0;
3097 }
3098
3099 /*
3100 * Frees memory within dir which was allocated, and resets fields for further
3101 * use. Does not free dir itself.
3102 */
3103 void dir_clear(struct dir_struct *dir)
3104 {
3105 int i, j;
3106 struct exclude_list_group *group;
3107 struct pattern_list *pl;
3108 struct exclude_stack *stk;
3109
3110 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
3111 group = &dir->exclude_list_group[i];
3112 for (j = 0; j < group->nr; j++) {
3113 pl = &group->pl[j];
3114 if (i == EXC_DIRS)
3115 free((char *)pl->src);
3116 clear_pattern_list(pl);
3117 }
3118 free(group->pl);
3119 }
3120
3121 for (i = 0; i < dir->ignored_nr; i++)
3122 free(dir->ignored[i]);
3123 for (i = 0; i < dir->nr; i++)
3124 free(dir->entries[i]);
3125 free(dir->ignored);
3126 free(dir->entries);
3127
3128 stk = dir->exclude_stack;
3129 while (stk) {
3130 struct exclude_stack *prev = stk->prev;
3131 free(stk);
3132 stk = prev;
3133 }
3134 strbuf_release(&dir->basebuf);
3135
3136 dir_init(dir);
3137 }
3138
3139 struct ondisk_untracked_cache {
3140 struct stat_data info_exclude_stat;
3141 struct stat_data excludes_file_stat;
3142 uint32_t dir_flags;
3143 };
3144
3145 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3146
3147 struct write_data {
3148 int index; /* number of written untracked_cache_dir */
3149 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
3150 struct ewah_bitmap *valid; /* from untracked_cache_dir */
3151 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
3152 struct strbuf out;
3153 struct strbuf sb_stat;
3154 struct strbuf sb_sha1;
3155 };
3156
3157 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
3158 {
3159 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
3160 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
3161 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
3162 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
3163 to->sd_dev = htonl(from->sd_dev);
3164 to->sd_ino = htonl(from->sd_ino);
3165 to->sd_uid = htonl(from->sd_uid);
3166 to->sd_gid = htonl(from->sd_gid);
3167 to->sd_size = htonl(from->sd_size);
3168 }
3169
3170 static void write_one_dir(struct untracked_cache_dir *untracked,
3171 struct write_data *wd)
3172 {
3173 struct stat_data stat_data;
3174 struct strbuf *out = &wd->out;
3175 unsigned char intbuf[16];
3176 unsigned int intlen, value;
3177 int i = wd->index++;
3178
3179 /*
3180 * untracked_nr should be reset whenever valid is clear, but
3181 * for safety..
3182 */
3183 if (!untracked->valid) {
3184 untracked->untracked_nr = 0;
3185 untracked->check_only = 0;
3186 }
3187
3188 if (untracked->check_only)
3189 ewah_set(wd->check_only, i);
3190 if (untracked->valid) {
3191 ewah_set(wd->valid, i);
3192 stat_data_to_disk(&stat_data, &untracked->stat_data);
3193 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
3194 }
3195 if (!is_null_oid(&untracked->exclude_oid)) {
3196 ewah_set(wd->sha1_valid, i);
3197 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
3198 the_hash_algo->rawsz);
3199 }
3200
3201 intlen = encode_varint(untracked->untracked_nr, intbuf);
3202 strbuf_add(out, intbuf, intlen);
3203
3204 /* skip non-recurse directories */
3205 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
3206 if (untracked->dirs[i]->recurse)
3207 value++;
3208 intlen = encode_varint(value, intbuf);
3209 strbuf_add(out, intbuf, intlen);
3210
3211 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
3212
3213 for (i = 0; i < untracked->untracked_nr; i++)
3214 strbuf_add(out, untracked->untracked[i],
3215 strlen(untracked->untracked[i]) + 1);
3216
3217 for (i = 0; i < untracked->dirs_nr; i++)
3218 if (untracked->dirs[i]->recurse)
3219 write_one_dir(untracked->dirs[i], wd);
3220 }
3221
3222 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
3223 {
3224 struct ondisk_untracked_cache *ouc;
3225 struct write_data wd;
3226 unsigned char varbuf[16];
3227 int varint_len;
3228 const unsigned hashsz = the_hash_algo->rawsz;
3229
3230 CALLOC_ARRAY(ouc, 1);
3231 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
3232 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
3233 ouc->dir_flags = htonl(untracked->dir_flags);
3234
3235 varint_len = encode_varint(untracked->ident.len, varbuf);
3236 strbuf_add(out, varbuf, varint_len);
3237 strbuf_addbuf(out, &untracked->ident);
3238
3239 strbuf_add(out, ouc, sizeof(*ouc));
3240 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
3241 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
3242 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
3243 FREE_AND_NULL(ouc);
3244
3245 if (!untracked->root) {
3246 varint_len = encode_varint(0, varbuf);
3247 strbuf_add(out, varbuf, varint_len);
3248 return;
3249 }
3250
3251 wd.index = 0;
3252 wd.check_only = ewah_new();
3253 wd.valid = ewah_new();
3254 wd.sha1_valid = ewah_new();
3255 strbuf_init(&wd.out, 1024);
3256 strbuf_init(&wd.sb_stat, 1024);
3257 strbuf_init(&wd.sb_sha1, 1024);
3258 write_one_dir(untracked->root, &wd);
3259
3260 varint_len = encode_varint(wd.index, varbuf);
3261 strbuf_add(out, varbuf, varint_len);
3262 strbuf_addbuf(out, &wd.out);
3263 ewah_serialize_strbuf(wd.valid, out);
3264 ewah_serialize_strbuf(wd.check_only, out);
3265 ewah_serialize_strbuf(wd.sha1_valid, out);
3266 strbuf_addbuf(out, &wd.sb_stat);
3267 strbuf_addbuf(out, &wd.sb_sha1);
3268 strbuf_addch(out, '\0'); /* safe guard for string lists */
3269
3270 ewah_free(wd.valid);
3271 ewah_free(wd.check_only);
3272 ewah_free(wd.sha1_valid);
3273 strbuf_release(&wd.out);
3274 strbuf_release(&wd.sb_stat);
3275 strbuf_release(&wd.sb_sha1);
3276 }
3277
3278 static void free_untracked(struct untracked_cache_dir *ucd)
3279 {
3280 int i;
3281 if (!ucd)
3282 return;
3283 for (i = 0; i < ucd->dirs_nr; i++)
3284 free_untracked(ucd->dirs[i]);
3285 for (i = 0; i < ucd->untracked_nr; i++)
3286 free(ucd->untracked[i]);
3287 free(ucd->untracked);
3288 free(ucd->dirs);
3289 free(ucd);
3290 }
3291
3292 void free_untracked_cache(struct untracked_cache *uc)
3293 {
3294 if (uc)
3295 free_untracked(uc->root);
3296 free(uc);
3297 }
3298
3299 struct read_data {
3300 int index;
3301 struct untracked_cache_dir **ucd;
3302 struct ewah_bitmap *check_only;
3303 struct ewah_bitmap *valid;
3304 struct ewah_bitmap *sha1_valid;
3305 const unsigned char *data;
3306 const unsigned char *end;
3307 };
3308
3309 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
3310 {
3311 memcpy(to, data, sizeof(*to));
3312 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
3313 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
3314 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
3315 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
3316 to->sd_dev = ntohl(to->sd_dev);
3317 to->sd_ino = ntohl(to->sd_ino);
3318 to->sd_uid = ntohl(to->sd_uid);
3319 to->sd_gid = ntohl(to->sd_gid);
3320 to->sd_size = ntohl(to->sd_size);
3321 }
3322
3323 static int read_one_dir(struct untracked_cache_dir **untracked_,
3324 struct read_data *rd)
3325 {
3326 struct untracked_cache_dir ud, *untracked;
3327 const unsigned char *data = rd->data, *end = rd->end;
3328 const unsigned char *eos;
3329 unsigned int value;
3330 int i;
3331
3332 memset(&ud, 0, sizeof(ud));
3333
3334 value = decode_varint(&data);
3335 if (data > end)
3336 return -1;
3337 ud.recurse = 1;
3338 ud.untracked_alloc = value;
3339 ud.untracked_nr = value;
3340 if (ud.untracked_nr)
3341 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
3342
3343 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
3344 if (data > end)
3345 return -1;
3346 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
3347
3348 eos = memchr(data, '\0', end - data);
3349 if (!eos || eos == end)
3350 return -1;
3351
3352 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
3353 memcpy(untracked, &ud, sizeof(ud));
3354 memcpy(untracked->name, data, eos - data + 1);
3355 data = eos + 1;
3356
3357 for (i = 0; i < untracked->untracked_nr; i++) {
3358 eos = memchr(data, '\0', end - data);
3359 if (!eos || eos == end)
3360 return -1;
3361 untracked->untracked[i] = xmemdupz(data, eos - data);
3362 data = eos + 1;
3363 }
3364
3365 rd->ucd[rd->index++] = untracked;
3366 rd->data = data;
3367
3368 for (i = 0; i < untracked->dirs_nr; i++) {
3369 if (read_one_dir(untracked->dirs + i, rd) < 0)
3370 return -1;
3371 }
3372 return 0;
3373 }
3374
3375 static void set_check_only(size_t pos, void *cb)
3376 {
3377 struct read_data *rd = cb;
3378 struct untracked_cache_dir *ud = rd->ucd[pos];
3379 ud->check_only = 1;
3380 }
3381
3382 static void read_stat(size_t pos, void *cb)
3383 {
3384 struct read_data *rd = cb;
3385 struct untracked_cache_dir *ud = rd->ucd[pos];
3386 if (rd->data + sizeof(struct stat_data) > rd->end) {
3387 rd->data = rd->end + 1;
3388 return;
3389 }
3390 stat_data_from_disk(&ud->stat_data, rd->data);
3391 rd->data += sizeof(struct stat_data);
3392 ud->valid = 1;
3393 }
3394
3395 static void read_oid(size_t pos, void *cb)
3396 {
3397 struct read_data *rd = cb;
3398 struct untracked_cache_dir *ud = rd->ucd[pos];
3399 if (rd->data + the_hash_algo->rawsz > rd->end) {
3400 rd->data = rd->end + 1;
3401 return;
3402 }
3403 oidread(&ud->exclude_oid, rd->data);
3404 rd->data += the_hash_algo->rawsz;
3405 }
3406
3407 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
3408 const unsigned char *sha1)
3409 {
3410 stat_data_from_disk(&oid_stat->stat, data);
3411 oidread(&oid_stat->oid, sha1);
3412 oid_stat->valid = 1;
3413 }
3414
3415 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
3416 {
3417 struct untracked_cache *uc;
3418 struct read_data rd;
3419 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
3420 const char *ident;
3421 int ident_len;
3422 ssize_t len;
3423 const char *exclude_per_dir;
3424 const unsigned hashsz = the_hash_algo->rawsz;
3425 const unsigned offset = sizeof(struct ondisk_untracked_cache);
3426 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
3427
3428 if (sz <= 1 || end[-1] != '\0')
3429 return NULL;
3430 end--;
3431
3432 ident_len = decode_varint(&next);
3433 if (next + ident_len > end)
3434 return NULL;
3435 ident = (const char *)next;
3436 next += ident_len;
3437
3438 if (next + exclude_per_dir_offset + 1 > end)
3439 return NULL;
3440
3441 CALLOC_ARRAY(uc, 1);
3442 strbuf_init(&uc->ident, ident_len);
3443 strbuf_add(&uc->ident, ident, ident_len);
3444 load_oid_stat(&uc->ss_info_exclude,
3445 next + ouc_offset(info_exclude_stat),
3446 next + offset);
3447 load_oid_stat(&uc->ss_excludes_file,
3448 next + ouc_offset(excludes_file_stat),
3449 next + offset + hashsz);
3450 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
3451 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
3452 uc->exclude_per_dir = xstrdup(exclude_per_dir);
3453 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3454 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
3455 if (next >= end)
3456 goto done2;
3457
3458 len = decode_varint(&next);
3459 if (next > end || len == 0)
3460 goto done2;
3461
3462 rd.valid = ewah_new();
3463 rd.check_only = ewah_new();
3464 rd.sha1_valid = ewah_new();
3465 rd.data = next;
3466 rd.end = end;
3467 rd.index = 0;
3468 ALLOC_ARRAY(rd.ucd, len);
3469
3470 if (read_one_dir(&uc->root, &rd) || rd.index != len)
3471 goto done;
3472
3473 next = rd.data;
3474 len = ewah_read_mmap(rd.valid, next, end - next);
3475 if (len < 0)
3476 goto done;
3477
3478 next += len;
3479 len = ewah_read_mmap(rd.check_only, next, end - next);
3480 if (len < 0)
3481 goto done;
3482
3483 next += len;
3484 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
3485 if (len < 0)
3486 goto done;
3487
3488 ewah_each_bit(rd.check_only, set_check_only, &rd);
3489 rd.data = next + len;
3490 ewah_each_bit(rd.valid, read_stat, &rd);
3491 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
3492 next = rd.data;
3493
3494 done:
3495 free(rd.ucd);
3496 ewah_free(rd.valid);
3497 ewah_free(rd.check_only);
3498 ewah_free(rd.sha1_valid);
3499 done2:
3500 if (next != end) {
3501 free_untracked_cache(uc);
3502 uc = NULL;
3503 }
3504 return uc;
3505 }
3506
3507 static void invalidate_one_directory(struct untracked_cache *uc,
3508 struct untracked_cache_dir *ucd)
3509 {
3510 uc->dir_invalidated++;
3511 ucd->valid = 0;
3512 ucd->untracked_nr = 0;
3513 }
3514
3515 /*
3516 * Normally when an entry is added or removed from a directory,
3517 * invalidating that directory is enough. No need to touch its
3518 * ancestors. When a directory is shown as "foo/bar/" in git-status
3519 * however, deleting or adding an entry may have cascading effect.
3520 *
3521 * Say the "foo/bar/file" has become untracked, we need to tell the
3522 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3523 * directory any more (because "bar" is managed by foo as an untracked
3524 * "file").
3525 *
3526 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3527 * was the last untracked entry in the entire "foo", we should show
3528 * "foo/" instead. Which means we have to invalidate past "bar" up to
3529 * "foo".
3530 *
3531 * This function traverses all directories from root to leaf. If there
3532 * is a chance of one of the above cases happening, we invalidate back
3533 * to root. Otherwise we just invalidate the leaf. There may be a more
3534 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3535 * detect these cases and avoid unnecessary invalidation, for example,
3536 * checking for the untracked entry named "bar/" in "foo", but for now
3537 * stick to something safe and simple.
3538 */
3539 static int invalidate_one_component(struct untracked_cache *uc,
3540 struct untracked_cache_dir *dir,
3541 const char *path, int len)
3542 {
3543 const char *rest = strchr(path, '/');
3544
3545 if (rest) {
3546 int component_len = rest - path;
3547 struct untracked_cache_dir *d =
3548 lookup_untracked(uc, dir, path, component_len);
3549 int ret =
3550 invalidate_one_component(uc, d, rest + 1,
3551 len - (component_len + 1));
3552 if (ret)
3553 invalidate_one_directory(uc, dir);
3554 return ret;
3555 }
3556
3557 invalidate_one_directory(uc, dir);
3558 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
3559 }
3560
3561 void untracked_cache_invalidate_path(struct index_state *istate,
3562 const char *path, int safe_path)
3563 {
3564 if (!istate->untracked || !istate->untracked->root)
3565 return;
3566 if (!safe_path && !verify_path(path, 0))
3567 return;
3568 invalidate_one_component(istate->untracked, istate->untracked->root,
3569 path, strlen(path));
3570 }
3571
3572 void untracked_cache_remove_from_index(struct index_state *istate,
3573 const char *path)
3574 {
3575 untracked_cache_invalidate_path(istate, path, 1);
3576 }
3577
3578 void untracked_cache_add_to_index(struct index_state *istate,
3579 const char *path)
3580 {
3581 untracked_cache_invalidate_path(istate, path, 1);
3582 }
3583
3584 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3585 const char *sub_gitdir)
3586 {
3587 int i;
3588 struct repository subrepo;
3589 struct strbuf sub_wt = STRBUF_INIT;
3590 struct strbuf sub_gd = STRBUF_INIT;
3591
3592 const struct submodule *sub;
3593
3594 /* If the submodule has no working tree, we can ignore it. */
3595 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3596 return;
3597
3598 if (repo_read_index(&subrepo) < 0)
3599 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3600
3601 /* TODO: audit for interaction with sparse-index. */
3602 ensure_full_index(subrepo.index);
3603 for (i = 0; i < subrepo.index->cache_nr; i++) {
3604 const struct cache_entry *ce = subrepo.index->cache[i];
3605
3606 if (!S_ISGITLINK(ce->ce_mode))
3607 continue;
3608
3609 while (i + 1 < subrepo.index->cache_nr &&
3610 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3611 /*
3612 * Skip entries with the same name in different stages
3613 * to make sure an entry is returned only once.
3614 */
3615 i++;
3616
3617 sub = submodule_from_path(&subrepo, null_oid(), ce->name);
3618 if (!sub || !is_submodule_active(&subrepo, ce->name))
3619 /* .gitmodules broken or inactive sub */
3620 continue;
3621
3622 strbuf_reset(&sub_wt);
3623 strbuf_reset(&sub_gd);
3624 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3625 strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name);
3626
3627 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3628 }
3629 strbuf_release(&sub_wt);
3630 strbuf_release(&sub_gd);
3631 repo_clear(&subrepo);
3632 }
3633
3634 void connect_work_tree_and_git_dir(const char *work_tree_,
3635 const char *git_dir_,
3636 int recurse_into_nested)
3637 {
3638 struct strbuf gitfile_sb = STRBUF_INIT;
3639 struct strbuf cfg_sb = STRBUF_INIT;
3640 struct strbuf rel_path = STRBUF_INIT;
3641 char *git_dir, *work_tree;
3642
3643 /* Prepare .git file */
3644 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3645 if (safe_create_leading_directories_const(gitfile_sb.buf))
3646 die(_("could not create directories for %s"), gitfile_sb.buf);
3647
3648 /* Prepare config file */
3649 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3650 if (safe_create_leading_directories_const(cfg_sb.buf))
3651 die(_("could not create directories for %s"), cfg_sb.buf);
3652
3653 git_dir = real_pathdup(git_dir_, 1);
3654 work_tree = real_pathdup(work_tree_, 1);
3655
3656 /* Write .git file */
3657 write_file(gitfile_sb.buf, "gitdir: %s",
3658 relative_path(git_dir, work_tree, &rel_path));
3659 /* Update core.worktree setting */
3660 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3661 relative_path(work_tree, git_dir, &rel_path));
3662
3663 strbuf_release(&gitfile_sb);
3664 strbuf_release(&cfg_sb);
3665 strbuf_release(&rel_path);
3666
3667 if (recurse_into_nested)
3668 connect_wt_gitdir_in_nested(work_tree, git_dir);
3669
3670 free(work_tree);
3671 free(git_dir);
3672 }
3673
3674 /*
3675 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3676 */
3677 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3678 {
3679 if (rename(old_git_dir, new_git_dir) < 0)
3680 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3681 old_git_dir, new_git_dir);
3682
3683 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3684 }