]> git.ipfire.org Git - thirdparty/git.git/blob - dir.c
The sixth batch for 2.26
[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 int count_slashes(const char *s)
58 {
59 int cnt = 0;
60 while (*s)
61 if (*s++ == '/')
62 cnt++;
63 return cnt;
64 }
65
66 int fspathcmp(const char *a, const char *b)
67 {
68 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
69 }
70
71 int fspathncmp(const char *a, const char *b, size_t count)
72 {
73 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
74 }
75
76 int git_fnmatch(const struct pathspec_item *item,
77 const char *pattern, const char *string,
78 int prefix)
79 {
80 if (prefix > 0) {
81 if (ps_strncmp(item, pattern, string, prefix))
82 return WM_NOMATCH;
83 pattern += prefix;
84 string += prefix;
85 }
86 if (item->flags & PATHSPEC_ONESTAR) {
87 int pattern_len = strlen(++pattern);
88 int string_len = strlen(string);
89 return string_len < pattern_len ||
90 ps_strcmp(item, pattern,
91 string + string_len - pattern_len);
92 }
93 if (item->magic & PATHSPEC_GLOB)
94 return wildmatch(pattern, string,
95 WM_PATHNAME |
96 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
97 else
98 /* wildmatch has not learned no FNM_PATHNAME mode yet */
99 return wildmatch(pattern, string,
100 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
101 }
102
103 static int fnmatch_icase_mem(const char *pattern, int patternlen,
104 const char *string, int stringlen,
105 int flags)
106 {
107 int match_status;
108 struct strbuf pat_buf = STRBUF_INIT;
109 struct strbuf str_buf = STRBUF_INIT;
110 const char *use_pat = pattern;
111 const char *use_str = string;
112
113 if (pattern[patternlen]) {
114 strbuf_add(&pat_buf, pattern, patternlen);
115 use_pat = pat_buf.buf;
116 }
117 if (string[stringlen]) {
118 strbuf_add(&str_buf, string, stringlen);
119 use_str = str_buf.buf;
120 }
121
122 if (ignore_case)
123 flags |= WM_CASEFOLD;
124 match_status = wildmatch(use_pat, use_str, flags);
125
126 strbuf_release(&pat_buf);
127 strbuf_release(&str_buf);
128
129 return match_status;
130 }
131
132 static size_t common_prefix_len(const struct pathspec *pathspec)
133 {
134 int n;
135 size_t max = 0;
136
137 /*
138 * ":(icase)path" is treated as a pathspec full of
139 * wildcard. In other words, only prefix is considered common
140 * prefix. If the pathspec is abc/foo abc/bar, running in
141 * subdir xyz, the common prefix is still xyz, not xyz/abc as
142 * in non-:(icase).
143 */
144 GUARD_PATHSPEC(pathspec,
145 PATHSPEC_FROMTOP |
146 PATHSPEC_MAXDEPTH |
147 PATHSPEC_LITERAL |
148 PATHSPEC_GLOB |
149 PATHSPEC_ICASE |
150 PATHSPEC_EXCLUDE |
151 PATHSPEC_ATTR);
152
153 for (n = 0; n < pathspec->nr; n++) {
154 size_t i = 0, len = 0, item_len;
155 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
156 continue;
157 if (pathspec->items[n].magic & PATHSPEC_ICASE)
158 item_len = pathspec->items[n].prefix;
159 else
160 item_len = pathspec->items[n].nowildcard_len;
161 while (i < item_len && (n == 0 || i < max)) {
162 char c = pathspec->items[n].match[i];
163 if (c != pathspec->items[0].match[i])
164 break;
165 if (c == '/')
166 len = i + 1;
167 i++;
168 }
169 if (n == 0 || len < max) {
170 max = len;
171 if (!max)
172 break;
173 }
174 }
175 return max;
176 }
177
178 /*
179 * Returns a copy of the longest leading path common among all
180 * pathspecs.
181 */
182 char *common_prefix(const struct pathspec *pathspec)
183 {
184 unsigned long len = common_prefix_len(pathspec);
185
186 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
187 }
188
189 int fill_directory(struct dir_struct *dir,
190 struct index_state *istate,
191 const struct pathspec *pathspec)
192 {
193 const char *prefix;
194 size_t prefix_len;
195
196 /*
197 * Calculate common prefix for the pathspec, and
198 * use that to optimize the directory walk
199 */
200 prefix_len = common_prefix_len(pathspec);
201 prefix = prefix_len ? pathspec->items[0].match : "";
202
203 /* Read the directory and prune it */
204 read_directory(dir, istate, prefix, prefix_len, pathspec);
205
206 return prefix_len;
207 }
208
209 int within_depth(const char *name, int namelen,
210 int depth, int max_depth)
211 {
212 const char *cp = name, *cpe = name + namelen;
213
214 while (cp < cpe) {
215 if (*cp++ != '/')
216 continue;
217 depth++;
218 if (depth > max_depth)
219 return 0;
220 }
221 return 1;
222 }
223
224 /*
225 * Read the contents of the blob with the given OID into a buffer.
226 * Append a trailing LF to the end if the last line doesn't have one.
227 *
228 * Returns:
229 * -1 when the OID is invalid or unknown or does not refer to a blob.
230 * 0 when the blob is empty.
231 * 1 along with { data, size } of the (possibly augmented) buffer
232 * when successful.
233 *
234 * Optionally updates the given oid_stat with the given OID (when valid).
235 */
236 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
237 size_t *size_out, char **data_out)
238 {
239 enum object_type type;
240 unsigned long sz;
241 char *data;
242
243 *size_out = 0;
244 *data_out = NULL;
245
246 data = read_object_file(oid, &type, &sz);
247 if (!data || type != OBJ_BLOB) {
248 free(data);
249 return -1;
250 }
251
252 if (oid_stat) {
253 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
254 oidcpy(&oid_stat->oid, oid);
255 }
256
257 if (sz == 0) {
258 free(data);
259 return 0;
260 }
261
262 if (data[sz - 1] != '\n') {
263 data = xrealloc(data, st_add(sz, 1));
264 data[sz++] = '\n';
265 }
266
267 *size_out = xsize_t(sz);
268 *data_out = data;
269
270 return 1;
271 }
272
273 #define DO_MATCH_EXCLUDE (1<<0)
274 #define DO_MATCH_DIRECTORY (1<<1)
275 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
276
277 /*
278 * Does the given pathspec match the given name? A match is found if
279 *
280 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
281 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
282 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
283 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
284 *
285 * Return value tells which case it was (1-4), or 0 when there is no match.
286 *
287 * It may be instructive to look at a small table of concrete examples
288 * to understand the differences between 1, 2, and 4:
289 *
290 * Pathspecs
291 * | a/b | a/b/ | a/b/c
292 * ------+-----------+-----------+------------
293 * a/b | EXACT | EXACT[1] | LEADING[2]
294 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
295 * a/b/c | RECURSIVE | RECURSIVE | EXACT
296 *
297 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
298 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
299 */
300 static int match_pathspec_item(const struct index_state *istate,
301 const struct pathspec_item *item, int prefix,
302 const char *name, int namelen, unsigned flags)
303 {
304 /* name/namelen has prefix cut off by caller */
305 const char *match = item->match + prefix;
306 int matchlen = item->len - prefix;
307
308 /*
309 * The normal call pattern is:
310 * 1. prefix = common_prefix_len(ps);
311 * 2. prune something, or fill_directory
312 * 3. match_pathspec()
313 *
314 * 'prefix' at #1 may be shorter than the command's prefix and
315 * it's ok for #2 to match extra files. Those extras will be
316 * trimmed at #3.
317 *
318 * Suppose the pathspec is 'foo' and '../bar' running from
319 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
320 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
321 * user does not want XYZ/foo, only the "foo" part should be
322 * case-insensitive. We need to filter out XYZ/foo here. In
323 * other words, we do not trust the caller on comparing the
324 * prefix part when :(icase) is involved. We do exact
325 * comparison ourselves.
326 *
327 * Normally the caller (common_prefix_len() in fact) does
328 * _exact_ matching on name[-prefix+1..-1] and we do not need
329 * to check that part. Be defensive and check it anyway, in
330 * case common_prefix_len is changed, or a new caller is
331 * introduced that does not use common_prefix_len.
332 *
333 * If the penalty turns out too high when prefix is really
334 * long, maybe change it to
335 * strncmp(match, name, item->prefix - prefix)
336 */
337 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
338 strncmp(item->match, name - prefix, item->prefix))
339 return 0;
340
341 if (item->attr_match_nr &&
342 !match_pathspec_attrs(istate, name, namelen, item))
343 return 0;
344
345 /* If the match was just the prefix, we matched */
346 if (!*match)
347 return MATCHED_RECURSIVELY;
348
349 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
350 if (matchlen == namelen)
351 return MATCHED_EXACTLY;
352
353 if (match[matchlen-1] == '/' || name[matchlen] == '/')
354 return MATCHED_RECURSIVELY;
355 } else if ((flags & DO_MATCH_DIRECTORY) &&
356 match[matchlen - 1] == '/' &&
357 namelen == matchlen - 1 &&
358 !ps_strncmp(item, match, name, namelen))
359 return MATCHED_EXACTLY;
360
361 if (item->nowildcard_len < item->len &&
362 !git_fnmatch(item, match, name,
363 item->nowildcard_len - prefix))
364 return MATCHED_FNMATCH;
365
366 /* Perform checks to see if "name" is a leading string of the pathspec */
367 if (flags & DO_MATCH_LEADING_PATHSPEC) {
368 /* name is a literal prefix of the pathspec */
369 int offset = name[namelen-1] == '/' ? 1 : 0;
370 if ((namelen < matchlen) &&
371 (match[namelen-offset] == '/') &&
372 !ps_strncmp(item, match, name, namelen))
373 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
374
375 /* name doesn't match up to the first wild character */
376 if (item->nowildcard_len < item->len &&
377 ps_strncmp(item, match, name,
378 item->nowildcard_len - prefix))
379 return 0;
380
381 /*
382 * name has no wildcard, and it didn't match as a leading
383 * pathspec so return.
384 */
385 if (item->nowildcard_len == item->len)
386 return 0;
387
388 /*
389 * Here is where we would perform a wildmatch to check if
390 * "name" can be matched as a directory (or a prefix) against
391 * the pathspec. Since wildmatch doesn't have this capability
392 * at the present we have to punt and say that it is a match,
393 * potentially returning a false positive
394 * The submodules themselves will be able to perform more
395 * accurate matching to determine if the pathspec matches.
396 */
397 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
398 }
399
400 return 0;
401 }
402
403 /*
404 * Given a name and a list of pathspecs, returns the nature of the
405 * closest (i.e. most specific) match of the name to any of the
406 * pathspecs.
407 *
408 * The caller typically calls this multiple times with the same
409 * pathspec and seen[] array but with different name/namelen
410 * (e.g. entries from the index) and is interested in seeing if and
411 * how each pathspec matches all the names it calls this function
412 * with. A mark is left in the seen[] array for each pathspec element
413 * indicating the closest type of match that element achieved, so if
414 * seen[n] remains zero after multiple invocations, that means the nth
415 * pathspec did not match any names, which could indicate that the
416 * user mistyped the nth pathspec.
417 */
418 static int do_match_pathspec(const struct index_state *istate,
419 const struct pathspec *ps,
420 const char *name, int namelen,
421 int prefix, char *seen,
422 unsigned flags)
423 {
424 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
425
426 GUARD_PATHSPEC(ps,
427 PATHSPEC_FROMTOP |
428 PATHSPEC_MAXDEPTH |
429 PATHSPEC_LITERAL |
430 PATHSPEC_GLOB |
431 PATHSPEC_ICASE |
432 PATHSPEC_EXCLUDE |
433 PATHSPEC_ATTR);
434
435 if (!ps->nr) {
436 if (!ps->recursive ||
437 !(ps->magic & PATHSPEC_MAXDEPTH) ||
438 ps->max_depth == -1)
439 return MATCHED_RECURSIVELY;
440
441 if (within_depth(name, namelen, 0, ps->max_depth))
442 return MATCHED_EXACTLY;
443 else
444 return 0;
445 }
446
447 name += prefix;
448 namelen -= prefix;
449
450 for (i = ps->nr - 1; i >= 0; i--) {
451 int how;
452
453 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
454 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
455 continue;
456
457 if (seen && seen[i] == MATCHED_EXACTLY)
458 continue;
459 /*
460 * Make exclude patterns optional and never report
461 * "pathspec ':(exclude)foo' matches no files"
462 */
463 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
464 seen[i] = MATCHED_FNMATCH;
465 how = match_pathspec_item(istate, ps->items+i, prefix, name,
466 namelen, flags);
467 if (ps->recursive &&
468 (ps->magic & PATHSPEC_MAXDEPTH) &&
469 ps->max_depth != -1 &&
470 how && how != MATCHED_FNMATCH) {
471 int len = ps->items[i].len;
472 if (name[len] == '/')
473 len++;
474 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
475 how = MATCHED_EXACTLY;
476 else
477 how = 0;
478 }
479 if (how) {
480 if (retval < how)
481 retval = how;
482 if (seen && seen[i] < how)
483 seen[i] = how;
484 }
485 }
486 return retval;
487 }
488
489 int match_pathspec(const struct index_state *istate,
490 const struct pathspec *ps,
491 const char *name, int namelen,
492 int prefix, char *seen, int is_dir)
493 {
494 int positive, negative;
495 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
496 positive = do_match_pathspec(istate, ps, name, namelen,
497 prefix, seen, flags);
498 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
499 return positive;
500 negative = do_match_pathspec(istate, ps, name, namelen,
501 prefix, seen,
502 flags | DO_MATCH_EXCLUDE);
503 return negative ? 0 : positive;
504 }
505
506 /**
507 * Check if a submodule is a superset of the pathspec
508 */
509 int submodule_path_match(const struct index_state *istate,
510 const struct pathspec *ps,
511 const char *submodule_name,
512 char *seen)
513 {
514 int matched = do_match_pathspec(istate, ps, submodule_name,
515 strlen(submodule_name),
516 0, seen,
517 DO_MATCH_DIRECTORY |
518 DO_MATCH_LEADING_PATHSPEC);
519 return matched;
520 }
521
522 int report_path_error(const char *ps_matched,
523 const struct pathspec *pathspec)
524 {
525 /*
526 * Make sure all pathspec matched; otherwise it is an error.
527 */
528 int num, errors = 0;
529 for (num = 0; num < pathspec->nr; num++) {
530 int other, found_dup;
531
532 if (ps_matched[num])
533 continue;
534 /*
535 * The caller might have fed identical pathspec
536 * twice. Do not barf on such a mistake.
537 * FIXME: parse_pathspec should have eliminated
538 * duplicate pathspec.
539 */
540 for (found_dup = other = 0;
541 !found_dup && other < pathspec->nr;
542 other++) {
543 if (other == num || !ps_matched[other])
544 continue;
545 if (!strcmp(pathspec->items[other].original,
546 pathspec->items[num].original))
547 /*
548 * Ok, we have a match already.
549 */
550 found_dup = 1;
551 }
552 if (found_dup)
553 continue;
554
555 error(_("pathspec '%s' did not match any file(s) known to git"),
556 pathspec->items[num].original);
557 errors++;
558 }
559 return errors;
560 }
561
562 /*
563 * Return the length of the "simple" part of a path match limiter.
564 */
565 int simple_length(const char *match)
566 {
567 int len = -1;
568
569 for (;;) {
570 unsigned char c = *match++;
571 len++;
572 if (c == '\0' || is_glob_special(c))
573 return len;
574 }
575 }
576
577 int no_wildcard(const char *string)
578 {
579 return string[simple_length(string)] == '\0';
580 }
581
582 void parse_path_pattern(const char **pattern,
583 int *patternlen,
584 unsigned *flags,
585 int *nowildcardlen)
586 {
587 const char *p = *pattern;
588 size_t i, len;
589
590 *flags = 0;
591 if (*p == '!') {
592 *flags |= PATTERN_FLAG_NEGATIVE;
593 p++;
594 }
595 len = strlen(p);
596 if (len && p[len - 1] == '/') {
597 len--;
598 *flags |= PATTERN_FLAG_MUSTBEDIR;
599 }
600 for (i = 0; i < len; i++) {
601 if (p[i] == '/')
602 break;
603 }
604 if (i == len)
605 *flags |= PATTERN_FLAG_NODIR;
606 *nowildcardlen = simple_length(p);
607 /*
608 * we should have excluded the trailing slash from 'p' too,
609 * but that's one more allocation. Instead just make sure
610 * nowildcardlen does not exceed real patternlen
611 */
612 if (*nowildcardlen > len)
613 *nowildcardlen = len;
614 if (*p == '*' && no_wildcard(p + 1))
615 *flags |= PATTERN_FLAG_ENDSWITH;
616 *pattern = p;
617 *patternlen = len;
618 }
619
620 int pl_hashmap_cmp(const void *unused_cmp_data,
621 const struct hashmap_entry *a,
622 const struct hashmap_entry *b,
623 const void *key)
624 {
625 const struct pattern_entry *ee1 =
626 container_of(a, struct pattern_entry, ent);
627 const struct pattern_entry *ee2 =
628 container_of(b, struct pattern_entry, ent);
629
630 size_t min_len = ee1->patternlen <= ee2->patternlen
631 ? ee1->patternlen
632 : ee2->patternlen;
633
634 if (ignore_case)
635 return strncasecmp(ee1->pattern, ee2->pattern, min_len);
636 return strncmp(ee1->pattern, ee2->pattern, min_len);
637 }
638
639 static char *dup_and_filter_pattern(const char *pattern)
640 {
641 char *set, *read;
642 size_t count = 0;
643 char *result = xstrdup(pattern);
644
645 set = result;
646 read = result;
647
648 while (*read) {
649 /* skip escape characters (once) */
650 if (*read == '\\')
651 read++;
652
653 *set = *read;
654
655 set++;
656 read++;
657 count++;
658 }
659 *set = 0;
660
661 if (count > 2 &&
662 *(set - 1) == '*' &&
663 *(set - 2) == '/')
664 *(set - 2) = 0;
665
666 return result;
667 }
668
669 static void add_pattern_to_hashsets(struct pattern_list *pl, struct path_pattern *given)
670 {
671 struct pattern_entry *translated;
672 char *truncated;
673 char *data = NULL;
674 const char *prev, *cur, *next;
675
676 if (!pl->use_cone_patterns)
677 return;
678
679 if (given->flags & PATTERN_FLAG_NEGATIVE &&
680 given->flags & PATTERN_FLAG_MUSTBEDIR &&
681 !strcmp(given->pattern, "/*")) {
682 pl->full_cone = 0;
683 return;
684 }
685
686 if (!given->flags && !strcmp(given->pattern, "/*")) {
687 pl->full_cone = 1;
688 return;
689 }
690
691 if (given->patternlen <= 2 ||
692 *given->pattern == '*' ||
693 strstr(given->pattern, "**")) {
694 /* Not a cone pattern. */
695 warning(_("unrecognized pattern: '%s'"), given->pattern);
696 goto clear_hashmaps;
697 }
698
699 prev = given->pattern;
700 cur = given->pattern + 1;
701 next = given->pattern + 2;
702
703 while (*cur) {
704 /* Watch for glob characters '*', '\', '[', '?' */
705 if (!is_glob_special(*cur))
706 goto increment;
707
708 /* But only if *prev != '\\' */
709 if (*prev == '\\')
710 goto increment;
711
712 /* But allow the initial '\' */
713 if (*cur == '\\' &&
714 is_glob_special(*next))
715 goto increment;
716
717 /* But a trailing '/' then '*' is fine */
718 if (*prev == '/' &&
719 *cur == '*' &&
720 *next == 0)
721 goto increment;
722
723 /* Not a cone pattern. */
724 warning(_("unrecognized pattern: '%s'"), given->pattern);
725 goto clear_hashmaps;
726
727 increment:
728 prev++;
729 cur++;
730 next++;
731 }
732
733 if (given->patternlen > 2 &&
734 !strcmp(given->pattern + given->patternlen - 2, "/*")) {
735 if (!(given->flags & PATTERN_FLAG_NEGATIVE)) {
736 /* Not a cone pattern. */
737 warning(_("unrecognized pattern: '%s'"), given->pattern);
738 goto clear_hashmaps;
739 }
740
741 truncated = dup_and_filter_pattern(given->pattern);
742
743 translated = xmalloc(sizeof(struct pattern_entry));
744 translated->pattern = truncated;
745 translated->patternlen = given->patternlen - 2;
746 hashmap_entry_init(&translated->ent,
747 ignore_case ?
748 strihash(translated->pattern) :
749 strhash(translated->pattern));
750
751 if (!hashmap_get_entry(&pl->recursive_hashmap,
752 translated, ent, NULL)) {
753 /* We did not see the "parent" included */
754 warning(_("unrecognized negative pattern: '%s'"),
755 given->pattern);
756 free(truncated);
757 free(translated);
758 goto clear_hashmaps;
759 }
760
761 hashmap_add(&pl->parent_hashmap, &translated->ent);
762 hashmap_remove(&pl->recursive_hashmap, &translated->ent, &data);
763 free(data);
764 return;
765 }
766
767 if (given->flags & PATTERN_FLAG_NEGATIVE) {
768 warning(_("unrecognized negative pattern: '%s'"),
769 given->pattern);
770 goto clear_hashmaps;
771 }
772
773 translated = xmalloc(sizeof(struct pattern_entry));
774
775 translated->pattern = dup_and_filter_pattern(given->pattern);
776 translated->patternlen = given->patternlen;
777 hashmap_entry_init(&translated->ent,
778 ignore_case ?
779 strihash(translated->pattern) :
780 strhash(translated->pattern));
781
782 hashmap_add(&pl->recursive_hashmap, &translated->ent);
783
784 if (hashmap_get_entry(&pl->parent_hashmap, translated, ent, NULL)) {
785 /* we already included this at the parent level */
786 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
787 given->pattern);
788 hashmap_remove(&pl->parent_hashmap, &translated->ent, &data);
789 free(data);
790 free(translated);
791 }
792
793 return;
794
795 clear_hashmaps:
796 warning(_("disabling cone pattern matching"));
797 hashmap_free_entries(&pl->parent_hashmap, struct pattern_entry, ent);
798 hashmap_free_entries(&pl->recursive_hashmap, struct pattern_entry, ent);
799 pl->use_cone_patterns = 0;
800 }
801
802 static int hashmap_contains_path(struct hashmap *map,
803 struct strbuf *pattern)
804 {
805 struct pattern_entry p;
806
807 /* Check straight mapping */
808 p.pattern = pattern->buf;
809 p.patternlen = pattern->len;
810 hashmap_entry_init(&p.ent,
811 ignore_case ?
812 strihash(p.pattern) :
813 strhash(p.pattern));
814 return !!hashmap_get_entry(map, &p, ent, NULL);
815 }
816
817 int hashmap_contains_parent(struct hashmap *map,
818 const char *path,
819 struct strbuf *buffer)
820 {
821 char *slash_pos;
822
823 strbuf_setlen(buffer, 0);
824
825 if (path[0] != '/')
826 strbuf_addch(buffer, '/');
827
828 strbuf_addstr(buffer, path);
829
830 slash_pos = strrchr(buffer->buf, '/');
831
832 while (slash_pos > buffer->buf) {
833 strbuf_setlen(buffer, slash_pos - buffer->buf);
834
835 if (hashmap_contains_path(map, buffer))
836 return 1;
837
838 slash_pos = strrchr(buffer->buf, '/');
839 }
840
841 return 0;
842 }
843
844 void add_pattern(const char *string, const char *base,
845 int baselen, struct pattern_list *pl, int srcpos)
846 {
847 struct path_pattern *pattern;
848 int patternlen;
849 unsigned flags;
850 int nowildcardlen;
851
852 parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen);
853 if (flags & PATTERN_FLAG_MUSTBEDIR) {
854 FLEXPTR_ALLOC_MEM(pattern, pattern, string, patternlen);
855 } else {
856 pattern = xmalloc(sizeof(*pattern));
857 pattern->pattern = string;
858 }
859 pattern->patternlen = patternlen;
860 pattern->nowildcardlen = nowildcardlen;
861 pattern->base = base;
862 pattern->baselen = baselen;
863 pattern->flags = flags;
864 pattern->srcpos = srcpos;
865 ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc);
866 pl->patterns[pl->nr++] = pattern;
867 pattern->pl = pl;
868
869 add_pattern_to_hashsets(pl, pattern);
870 }
871
872 static int read_skip_worktree_file_from_index(const struct index_state *istate,
873 const char *path,
874 size_t *size_out, char **data_out,
875 struct oid_stat *oid_stat)
876 {
877 int pos, len;
878
879 len = strlen(path);
880 pos = index_name_pos(istate, path, len);
881 if (pos < 0)
882 return -1;
883 if (!ce_skip_worktree(istate->cache[pos]))
884 return -1;
885
886 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
887 }
888
889 /*
890 * Frees memory within pl which was allocated for exclude patterns and
891 * the file buffer. Does not free pl itself.
892 */
893 void clear_pattern_list(struct pattern_list *pl)
894 {
895 int i;
896
897 for (i = 0; i < pl->nr; i++)
898 free(pl->patterns[i]);
899 free(pl->patterns);
900 free(pl->filebuf);
901
902 memset(pl, 0, sizeof(*pl));
903 }
904
905 static void trim_trailing_spaces(char *buf)
906 {
907 char *p, *last_space = NULL;
908
909 for (p = buf; *p; p++)
910 switch (*p) {
911 case ' ':
912 if (!last_space)
913 last_space = p;
914 break;
915 case '\\':
916 p++;
917 if (!*p)
918 return;
919 /* fallthrough */
920 default:
921 last_space = NULL;
922 }
923
924 if (last_space)
925 *last_space = '\0';
926 }
927
928 /*
929 * Given a subdirectory name and "dir" of the current directory,
930 * search the subdir in "dir" and return it, or create a new one if it
931 * does not exist in "dir".
932 *
933 * If "name" has the trailing slash, it'll be excluded in the search.
934 */
935 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
936 struct untracked_cache_dir *dir,
937 const char *name, int len)
938 {
939 int first, last;
940 struct untracked_cache_dir *d;
941 if (!dir)
942 return NULL;
943 if (len && name[len - 1] == '/')
944 len--;
945 first = 0;
946 last = dir->dirs_nr;
947 while (last > first) {
948 int cmp, next = first + ((last - first) >> 1);
949 d = dir->dirs[next];
950 cmp = strncmp(name, d->name, len);
951 if (!cmp && strlen(d->name) > len)
952 cmp = -1;
953 if (!cmp)
954 return d;
955 if (cmp < 0) {
956 last = next;
957 continue;
958 }
959 first = next+1;
960 }
961
962 uc->dir_created++;
963 FLEX_ALLOC_MEM(d, name, name, len);
964
965 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
966 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
967 dir->dirs_nr - first);
968 dir->dirs_nr++;
969 dir->dirs[first] = d;
970 return d;
971 }
972
973 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
974 {
975 int i;
976 dir->valid = 0;
977 dir->untracked_nr = 0;
978 for (i = 0; i < dir->dirs_nr; i++)
979 do_invalidate_gitignore(dir->dirs[i]);
980 }
981
982 static void invalidate_gitignore(struct untracked_cache *uc,
983 struct untracked_cache_dir *dir)
984 {
985 uc->gitignore_invalidated++;
986 do_invalidate_gitignore(dir);
987 }
988
989 static void invalidate_directory(struct untracked_cache *uc,
990 struct untracked_cache_dir *dir)
991 {
992 int i;
993
994 /*
995 * Invalidation increment here is just roughly correct. If
996 * untracked_nr or any of dirs[].recurse is non-zero, we
997 * should increment dir_invalidated too. But that's more
998 * expensive to do.
999 */
1000 if (dir->valid)
1001 uc->dir_invalidated++;
1002
1003 dir->valid = 0;
1004 dir->untracked_nr = 0;
1005 for (i = 0; i < dir->dirs_nr; i++)
1006 dir->dirs[i]->recurse = 0;
1007 }
1008
1009 static int add_patterns_from_buffer(char *buf, size_t size,
1010 const char *base, int baselen,
1011 struct pattern_list *pl);
1012
1013 /*
1014 * Given a file with name "fname", read it (either from disk, or from
1015 * an index if 'istate' is non-null), parse it and store the
1016 * exclude rules in "pl".
1017 *
1018 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
1019 * stat data from disk (only valid if add_patterns returns zero). If
1020 * ss_valid is non-zero, "ss" must contain good value as input.
1021 */
1022 static int add_patterns(const char *fname, const char *base, int baselen,
1023 struct pattern_list *pl, struct index_state *istate,
1024 struct oid_stat *oid_stat)
1025 {
1026 struct stat st;
1027 int r;
1028 int fd;
1029 size_t size = 0;
1030 char *buf;
1031
1032 fd = open(fname, O_RDONLY);
1033 if (fd < 0 || fstat(fd, &st) < 0) {
1034 if (fd < 0)
1035 warn_on_fopen_errors(fname);
1036 else
1037 close(fd);
1038 if (!istate)
1039 return -1;
1040 r = read_skip_worktree_file_from_index(istate, fname,
1041 &size, &buf,
1042 oid_stat);
1043 if (r != 1)
1044 return r;
1045 } else {
1046 size = xsize_t(st.st_size);
1047 if (size == 0) {
1048 if (oid_stat) {
1049 fill_stat_data(&oid_stat->stat, &st);
1050 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
1051 oid_stat->valid = 1;
1052 }
1053 close(fd);
1054 return 0;
1055 }
1056 buf = xmallocz(size);
1057 if (read_in_full(fd, buf, size) != size) {
1058 free(buf);
1059 close(fd);
1060 return -1;
1061 }
1062 buf[size++] = '\n';
1063 close(fd);
1064 if (oid_stat) {
1065 int pos;
1066 if (oid_stat->valid &&
1067 !match_stat_data_racy(istate, &oid_stat->stat, &st))
1068 ; /* no content change, ss->sha1 still good */
1069 else if (istate &&
1070 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
1071 !ce_stage(istate->cache[pos]) &&
1072 ce_uptodate(istate->cache[pos]) &&
1073 !would_convert_to_git(istate, fname))
1074 oidcpy(&oid_stat->oid,
1075 &istate->cache[pos]->oid);
1076 else
1077 hash_object_file(the_hash_algo, buf, size,
1078 "blob", &oid_stat->oid);
1079 fill_stat_data(&oid_stat->stat, &st);
1080 oid_stat->valid = 1;
1081 }
1082 }
1083
1084 add_patterns_from_buffer(buf, size, base, baselen, pl);
1085 return 0;
1086 }
1087
1088 static int add_patterns_from_buffer(char *buf, size_t size,
1089 const char *base, int baselen,
1090 struct pattern_list *pl)
1091 {
1092 int i, lineno = 1;
1093 char *entry;
1094
1095 hashmap_init(&pl->recursive_hashmap, pl_hashmap_cmp, NULL, 0);
1096 hashmap_init(&pl->parent_hashmap, pl_hashmap_cmp, NULL, 0);
1097
1098 pl->filebuf = buf;
1099
1100 if (skip_utf8_bom(&buf, size))
1101 size -= buf - pl->filebuf;
1102
1103 entry = buf;
1104
1105 for (i = 0; i < size; i++) {
1106 if (buf[i] == '\n') {
1107 if (entry != buf + i && entry[0] != '#') {
1108 buf[i - (i && buf[i-1] == '\r')] = 0;
1109 trim_trailing_spaces(entry);
1110 add_pattern(entry, base, baselen, pl, lineno);
1111 }
1112 lineno++;
1113 entry = buf + i + 1;
1114 }
1115 }
1116 return 0;
1117 }
1118
1119 int add_patterns_from_file_to_list(const char *fname, const char *base,
1120 int baselen, struct pattern_list *pl,
1121 struct index_state *istate)
1122 {
1123 return add_patterns(fname, base, baselen, pl, istate, NULL);
1124 }
1125
1126 int add_patterns_from_blob_to_list(
1127 struct object_id *oid,
1128 const char *base, int baselen,
1129 struct pattern_list *pl)
1130 {
1131 char *buf;
1132 size_t size;
1133 int r;
1134
1135 r = do_read_blob(oid, NULL, &size, &buf);
1136 if (r != 1)
1137 return r;
1138
1139 add_patterns_from_buffer(buf, size, base, baselen, pl);
1140 return 0;
1141 }
1142
1143 struct pattern_list *add_pattern_list(struct dir_struct *dir,
1144 int group_type, const char *src)
1145 {
1146 struct pattern_list *pl;
1147 struct exclude_list_group *group;
1148
1149 group = &dir->exclude_list_group[group_type];
1150 ALLOC_GROW(group->pl, group->nr + 1, group->alloc);
1151 pl = &group->pl[group->nr++];
1152 memset(pl, 0, sizeof(*pl));
1153 pl->src = src;
1154 return pl;
1155 }
1156
1157 /*
1158 * Used to set up core.excludesfile and .git/info/exclude lists.
1159 */
1160 static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname,
1161 struct oid_stat *oid_stat)
1162 {
1163 struct pattern_list *pl;
1164 /*
1165 * catch setup_standard_excludes() that's called before
1166 * dir->untracked is assigned. That function behaves
1167 * differently when dir->untracked is non-NULL.
1168 */
1169 if (!dir->untracked)
1170 dir->unmanaged_exclude_files++;
1171 pl = add_pattern_list(dir, EXC_FILE, fname);
1172 if (add_patterns(fname, "", 0, pl, NULL, oid_stat) < 0)
1173 die(_("cannot use %s as an exclude file"), fname);
1174 }
1175
1176 void add_patterns_from_file(struct dir_struct *dir, const char *fname)
1177 {
1178 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
1179 add_patterns_from_file_1(dir, fname, NULL);
1180 }
1181
1182 int match_basename(const char *basename, int basenamelen,
1183 const char *pattern, int prefix, int patternlen,
1184 unsigned flags)
1185 {
1186 if (prefix == patternlen) {
1187 if (patternlen == basenamelen &&
1188 !fspathncmp(pattern, basename, basenamelen))
1189 return 1;
1190 } else if (flags & PATTERN_FLAG_ENDSWITH) {
1191 /* "*literal" matching against "fooliteral" */
1192 if (patternlen - 1 <= basenamelen &&
1193 !fspathncmp(pattern + 1,
1194 basename + basenamelen - (patternlen - 1),
1195 patternlen - 1))
1196 return 1;
1197 } else {
1198 if (fnmatch_icase_mem(pattern, patternlen,
1199 basename, basenamelen,
1200 0) == 0)
1201 return 1;
1202 }
1203 return 0;
1204 }
1205
1206 int match_pathname(const char *pathname, int pathlen,
1207 const char *base, int baselen,
1208 const char *pattern, int prefix, int patternlen,
1209 unsigned flags)
1210 {
1211 const char *name;
1212 int namelen;
1213
1214 /*
1215 * match with FNM_PATHNAME; the pattern has base implicitly
1216 * in front of it.
1217 */
1218 if (*pattern == '/') {
1219 pattern++;
1220 patternlen--;
1221 prefix--;
1222 }
1223
1224 /*
1225 * baselen does not count the trailing slash. base[] may or
1226 * may not end with a trailing slash though.
1227 */
1228 if (pathlen < baselen + 1 ||
1229 (baselen && pathname[baselen] != '/') ||
1230 fspathncmp(pathname, base, baselen))
1231 return 0;
1232
1233 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1234 name = pathname + pathlen - namelen;
1235
1236 if (prefix) {
1237 /*
1238 * if the non-wildcard part is longer than the
1239 * remaining pathname, surely it cannot match.
1240 */
1241 if (prefix > namelen)
1242 return 0;
1243
1244 if (fspathncmp(pattern, name, prefix))
1245 return 0;
1246 pattern += prefix;
1247 patternlen -= prefix;
1248 name += prefix;
1249 namelen -= prefix;
1250
1251 /*
1252 * If the whole pattern did not have a wildcard,
1253 * then our prefix match is all we need; we
1254 * do not need to call fnmatch at all.
1255 */
1256 if (!patternlen && !namelen)
1257 return 1;
1258 }
1259
1260 return fnmatch_icase_mem(pattern, patternlen,
1261 name, namelen,
1262 WM_PATHNAME) == 0;
1263 }
1264
1265 /*
1266 * Scan the given exclude list in reverse to see whether pathname
1267 * should be ignored. The first match (i.e. the last on the list), if
1268 * any, determines the fate. Returns the exclude_list element which
1269 * matched, or NULL for undecided.
1270 */
1271 static struct path_pattern *last_matching_pattern_from_list(const char *pathname,
1272 int pathlen,
1273 const char *basename,
1274 int *dtype,
1275 struct pattern_list *pl,
1276 struct index_state *istate)
1277 {
1278 struct path_pattern *res = NULL; /* undecided */
1279 int i;
1280
1281 if (!pl->nr)
1282 return NULL; /* undefined */
1283
1284 for (i = pl->nr - 1; 0 <= i; i--) {
1285 struct path_pattern *pattern = pl->patterns[i];
1286 const char *exclude = pattern->pattern;
1287 int prefix = pattern->nowildcardlen;
1288
1289 if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) {
1290 *dtype = resolve_dtype(*dtype, istate, pathname, pathlen);
1291 if (*dtype != DT_DIR)
1292 continue;
1293 }
1294
1295 if (pattern->flags & PATTERN_FLAG_NODIR) {
1296 if (match_basename(basename,
1297 pathlen - (basename - pathname),
1298 exclude, prefix, pattern->patternlen,
1299 pattern->flags)) {
1300 res = pattern;
1301 break;
1302 }
1303 continue;
1304 }
1305
1306 assert(pattern->baselen == 0 ||
1307 pattern->base[pattern->baselen - 1] == '/');
1308 if (match_pathname(pathname, pathlen,
1309 pattern->base,
1310 pattern->baselen ? pattern->baselen - 1 : 0,
1311 exclude, prefix, pattern->patternlen,
1312 pattern->flags)) {
1313 res = pattern;
1314 break;
1315 }
1316 }
1317 return res;
1318 }
1319
1320 /*
1321 * Scan the list of patterns to determine if the ordered list
1322 * of patterns matches on 'pathname'.
1323 *
1324 * Return 1 for a match, 0 for not matched and -1 for undecided.
1325 */
1326 enum pattern_match_result path_matches_pattern_list(
1327 const char *pathname, int pathlen,
1328 const char *basename, int *dtype,
1329 struct pattern_list *pl,
1330 struct index_state *istate)
1331 {
1332 struct path_pattern *pattern;
1333 struct strbuf parent_pathname = STRBUF_INIT;
1334 int result = NOT_MATCHED;
1335 const char *slash_pos;
1336
1337 if (!pl->use_cone_patterns) {
1338 pattern = last_matching_pattern_from_list(pathname, pathlen, basename,
1339 dtype, pl, istate);
1340 if (pattern) {
1341 if (pattern->flags & PATTERN_FLAG_NEGATIVE)
1342 return NOT_MATCHED;
1343 else
1344 return MATCHED;
1345 }
1346
1347 return UNDECIDED;
1348 }
1349
1350 if (pl->full_cone)
1351 return MATCHED;
1352
1353 strbuf_addch(&parent_pathname, '/');
1354 strbuf_add(&parent_pathname, pathname, pathlen);
1355
1356 if (hashmap_contains_path(&pl->recursive_hashmap,
1357 &parent_pathname)) {
1358 result = MATCHED_RECURSIVE;
1359 goto done;
1360 }
1361
1362 slash_pos = strrchr(parent_pathname.buf, '/');
1363
1364 if (slash_pos == parent_pathname.buf) {
1365 /* include every file in root */
1366 result = MATCHED;
1367 goto done;
1368 }
1369
1370 strbuf_setlen(&parent_pathname, slash_pos - parent_pathname.buf);
1371
1372 if (hashmap_contains_path(&pl->parent_hashmap, &parent_pathname)) {
1373 result = MATCHED;
1374 goto done;
1375 }
1376
1377 if (hashmap_contains_parent(&pl->recursive_hashmap,
1378 pathname,
1379 &parent_pathname))
1380 result = MATCHED_RECURSIVE;
1381
1382 done:
1383 strbuf_release(&parent_pathname);
1384 return result;
1385 }
1386
1387 static struct path_pattern *last_matching_pattern_from_lists(
1388 struct dir_struct *dir, struct index_state *istate,
1389 const char *pathname, int pathlen,
1390 const char *basename, int *dtype_p)
1391 {
1392 int i, j;
1393 struct exclude_list_group *group;
1394 struct path_pattern *pattern;
1395 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1396 group = &dir->exclude_list_group[i];
1397 for (j = group->nr - 1; j >= 0; j--) {
1398 pattern = last_matching_pattern_from_list(
1399 pathname, pathlen, basename, dtype_p,
1400 &group->pl[j], istate);
1401 if (pattern)
1402 return pattern;
1403 }
1404 }
1405 return NULL;
1406 }
1407
1408 /*
1409 * Loads the per-directory exclude list for the substring of base
1410 * which has a char length of baselen.
1411 */
1412 static void prep_exclude(struct dir_struct *dir,
1413 struct index_state *istate,
1414 const char *base, int baselen)
1415 {
1416 struct exclude_list_group *group;
1417 struct pattern_list *pl;
1418 struct exclude_stack *stk = NULL;
1419 struct untracked_cache_dir *untracked;
1420 int current;
1421
1422 group = &dir->exclude_list_group[EXC_DIRS];
1423
1424 /*
1425 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1426 * which originate from directories not in the prefix of the
1427 * path being checked.
1428 */
1429 while ((stk = dir->exclude_stack) != NULL) {
1430 if (stk->baselen <= baselen &&
1431 !strncmp(dir->basebuf.buf, base, stk->baselen))
1432 break;
1433 pl = &group->pl[dir->exclude_stack->exclude_ix];
1434 dir->exclude_stack = stk->prev;
1435 dir->pattern = NULL;
1436 free((char *)pl->src); /* see strbuf_detach() below */
1437 clear_pattern_list(pl);
1438 free(stk);
1439 group->nr--;
1440 }
1441
1442 /* Skip traversing into sub directories if the parent is excluded */
1443 if (dir->pattern)
1444 return;
1445
1446 /*
1447 * Lazy initialization. All call sites currently just
1448 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1449 * them seems lots of work for little benefit.
1450 */
1451 if (!dir->basebuf.buf)
1452 strbuf_init(&dir->basebuf, PATH_MAX);
1453
1454 /* Read from the parent directories and push them down. */
1455 current = stk ? stk->baselen : -1;
1456 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1457 if (dir->untracked)
1458 untracked = stk ? stk->ucd : dir->untracked->root;
1459 else
1460 untracked = NULL;
1461
1462 while (current < baselen) {
1463 const char *cp;
1464 struct oid_stat oid_stat;
1465
1466 stk = xcalloc(1, sizeof(*stk));
1467 if (current < 0) {
1468 cp = base;
1469 current = 0;
1470 } else {
1471 cp = strchr(base + current + 1, '/');
1472 if (!cp)
1473 die("oops in prep_exclude");
1474 cp++;
1475 untracked =
1476 lookup_untracked(dir->untracked, untracked,
1477 base + current,
1478 cp - base - current);
1479 }
1480 stk->prev = dir->exclude_stack;
1481 stk->baselen = cp - base;
1482 stk->exclude_ix = group->nr;
1483 stk->ucd = untracked;
1484 pl = add_pattern_list(dir, EXC_DIRS, NULL);
1485 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1486 assert(stk->baselen == dir->basebuf.len);
1487
1488 /* Abort if the directory is excluded */
1489 if (stk->baselen) {
1490 int dt = DT_DIR;
1491 dir->basebuf.buf[stk->baselen - 1] = 0;
1492 dir->pattern = last_matching_pattern_from_lists(dir,
1493 istate,
1494 dir->basebuf.buf, stk->baselen - 1,
1495 dir->basebuf.buf + current, &dt);
1496 dir->basebuf.buf[stk->baselen - 1] = '/';
1497 if (dir->pattern &&
1498 dir->pattern->flags & PATTERN_FLAG_NEGATIVE)
1499 dir->pattern = NULL;
1500 if (dir->pattern) {
1501 dir->exclude_stack = stk;
1502 return;
1503 }
1504 }
1505
1506 /* Try to read per-directory file */
1507 oidclr(&oid_stat.oid);
1508 oid_stat.valid = 0;
1509 if (dir->exclude_per_dir &&
1510 /*
1511 * If we know that no files have been added in
1512 * this directory (i.e. valid_cached_dir() has
1513 * been executed and set untracked->valid) ..
1514 */
1515 (!untracked || !untracked->valid ||
1516 /*
1517 * .. and .gitignore does not exist before
1518 * (i.e. null exclude_oid). Then we can skip
1519 * loading .gitignore, which would result in
1520 * ENOENT anyway.
1521 */
1522 !is_null_oid(&untracked->exclude_oid))) {
1523 /*
1524 * dir->basebuf gets reused by the traversal, but we
1525 * need fname to remain unchanged to ensure the src
1526 * member of each struct path_pattern correctly
1527 * back-references its source file. Other invocations
1528 * of add_pattern_list provide stable strings, so we
1529 * strbuf_detach() and free() here in the caller.
1530 */
1531 struct strbuf sb = STRBUF_INIT;
1532 strbuf_addbuf(&sb, &dir->basebuf);
1533 strbuf_addstr(&sb, dir->exclude_per_dir);
1534 pl->src = strbuf_detach(&sb, NULL);
1535 add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1536 untracked ? &oid_stat : NULL);
1537 }
1538 /*
1539 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1540 * will first be called in valid_cached_dir() then maybe many
1541 * times more in last_matching_pattern(). When the cache is
1542 * used, last_matching_pattern() will not be called and
1543 * reading .gitignore content will be a waste.
1544 *
1545 * So when it's called by valid_cached_dir() and we can get
1546 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1547 * modified on work tree), we could delay reading the
1548 * .gitignore content until we absolutely need it in
1549 * last_matching_pattern(). Be careful about ignore rule
1550 * order, though, if you do that.
1551 */
1552 if (untracked &&
1553 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1554 invalidate_gitignore(dir->untracked, untracked);
1555 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1556 }
1557 dir->exclude_stack = stk;
1558 current = stk->baselen;
1559 }
1560 strbuf_setlen(&dir->basebuf, baselen);
1561 }
1562
1563 /*
1564 * Loads the exclude lists for the directory containing pathname, then
1565 * scans all exclude lists to determine whether pathname is excluded.
1566 * Returns the exclude_list element which matched, or NULL for
1567 * undecided.
1568 */
1569 struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1570 struct index_state *istate,
1571 const char *pathname,
1572 int *dtype_p)
1573 {
1574 int pathlen = strlen(pathname);
1575 const char *basename = strrchr(pathname, '/');
1576 basename = (basename) ? basename+1 : pathname;
1577
1578 prep_exclude(dir, istate, pathname, basename-pathname);
1579
1580 if (dir->pattern)
1581 return dir->pattern;
1582
1583 return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1584 basename, dtype_p);
1585 }
1586
1587 /*
1588 * Loads the exclude lists for the directory containing pathname, then
1589 * scans all exclude lists to determine whether pathname is excluded.
1590 * Returns 1 if true, otherwise 0.
1591 */
1592 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1593 const char *pathname, int *dtype_p)
1594 {
1595 struct path_pattern *pattern =
1596 last_matching_pattern(dir, istate, pathname, dtype_p);
1597 if (pattern)
1598 return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1599 return 0;
1600 }
1601
1602 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1603 {
1604 struct dir_entry *ent;
1605
1606 FLEX_ALLOC_MEM(ent, name, pathname, len);
1607 ent->len = len;
1608 return ent;
1609 }
1610
1611 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1612 struct index_state *istate,
1613 const char *pathname, int len)
1614 {
1615 if (index_file_exists(istate, pathname, len, ignore_case))
1616 return NULL;
1617
1618 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1619 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1620 }
1621
1622 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1623 struct index_state *istate,
1624 const char *pathname, int len)
1625 {
1626 if (!index_name_is_other(istate, pathname, len))
1627 return NULL;
1628
1629 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1630 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1631 }
1632
1633 enum exist_status {
1634 index_nonexistent = 0,
1635 index_directory,
1636 index_gitdir
1637 };
1638
1639 /*
1640 * Do not use the alphabetically sorted index to look up
1641 * the directory name; instead, use the case insensitive
1642 * directory hash.
1643 */
1644 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1645 const char *dirname, int len)
1646 {
1647 struct cache_entry *ce;
1648
1649 if (index_dir_exists(istate, dirname, len))
1650 return index_directory;
1651
1652 ce = index_file_exists(istate, dirname, len, ignore_case);
1653 if (ce && S_ISGITLINK(ce->ce_mode))
1654 return index_gitdir;
1655
1656 return index_nonexistent;
1657 }
1658
1659 /*
1660 * The index sorts alphabetically by entry name, which
1661 * means that a gitlink sorts as '\0' at the end, while
1662 * a directory (which is defined not as an entry, but as
1663 * the files it contains) will sort with the '/' at the
1664 * end.
1665 */
1666 static enum exist_status directory_exists_in_index(struct index_state *istate,
1667 const char *dirname, int len)
1668 {
1669 int pos;
1670
1671 if (ignore_case)
1672 return directory_exists_in_index_icase(istate, dirname, len);
1673
1674 pos = index_name_pos(istate, dirname, len);
1675 if (pos < 0)
1676 pos = -pos-1;
1677 while (pos < istate->cache_nr) {
1678 const struct cache_entry *ce = istate->cache[pos++];
1679 unsigned char endchar;
1680
1681 if (strncmp(ce->name, dirname, len))
1682 break;
1683 endchar = ce->name[len];
1684 if (endchar > '/')
1685 break;
1686 if (endchar == '/')
1687 return index_directory;
1688 if (!endchar && S_ISGITLINK(ce->ce_mode))
1689 return index_gitdir;
1690 }
1691 return index_nonexistent;
1692 }
1693
1694 /*
1695 * When we find a directory when traversing the filesystem, we
1696 * have three distinct cases:
1697 *
1698 * - ignore it
1699 * - see it as a directory
1700 * - recurse into it
1701 *
1702 * and which one we choose depends on a combination of existing
1703 * git index contents and the flags passed into the directory
1704 * traversal routine.
1705 *
1706 * Case 1: If we *already* have entries in the index under that
1707 * directory name, we always recurse into the directory to see
1708 * all the files.
1709 *
1710 * Case 2: If we *already* have that directory name as a gitlink,
1711 * we always continue to see it as a gitlink, regardless of whether
1712 * there is an actual git directory there or not (it might not
1713 * be checked out as a subproject!)
1714 *
1715 * Case 3: if we didn't have it in the index previously, we
1716 * have a few sub-cases:
1717 *
1718 * (a) if "show_other_directories" is true, we show it as
1719 * just a directory, unless "hide_empty_directories" is
1720 * also true, in which case we need to check if it contains any
1721 * untracked and / or ignored files.
1722 * (b) if it looks like a git directory, and we don't have
1723 * 'no_gitlinks' set we treat it as a gitlink, and show it
1724 * as a directory.
1725 * (c) otherwise, we recurse into it.
1726 */
1727 static enum path_treatment treat_directory(struct dir_struct *dir,
1728 struct index_state *istate,
1729 struct untracked_cache_dir *untracked,
1730 const char *dirname, int len, int baselen, int exclude,
1731 const struct pathspec *pathspec)
1732 {
1733 int nested_repo = 0;
1734
1735 /* The "len-1" is to strip the final '/' */
1736 switch (directory_exists_in_index(istate, dirname, len-1)) {
1737 case index_directory:
1738 return path_recurse;
1739
1740 case index_gitdir:
1741 return path_none;
1742
1743 case index_nonexistent:
1744 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1745 !(dir->flags & DIR_NO_GITLINKS)) {
1746 struct strbuf sb = STRBUF_INIT;
1747 strbuf_addstr(&sb, dirname);
1748 nested_repo = is_nonbare_repository_dir(&sb);
1749 strbuf_release(&sb);
1750 }
1751 if (nested_repo)
1752 return ((dir->flags & DIR_SKIP_NESTED_GIT) ? path_none :
1753 (exclude ? path_excluded : path_untracked));
1754
1755 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1756 break;
1757 if (exclude &&
1758 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1759 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1760
1761 /*
1762 * This is an excluded directory and we are
1763 * showing ignored paths that match an exclude
1764 * pattern. (e.g. show directory as ignored
1765 * only if it matches an exclude pattern).
1766 * This path will either be 'path_excluded`
1767 * (if we are showing empty directories or if
1768 * the directory is not empty), or will be
1769 * 'path_none' (empty directory, and we are
1770 * not showing empty directories).
1771 */
1772 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1773 return path_excluded;
1774
1775 if (read_directory_recursive(dir, istate, dirname, len,
1776 untracked, 1, 1, pathspec) == path_excluded)
1777 return path_excluded;
1778
1779 return path_none;
1780 }
1781 return path_recurse;
1782 }
1783
1784 /* This is the "show_other_directories" case */
1785
1786 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1787 return exclude ? path_excluded : path_untracked;
1788
1789 untracked = lookup_untracked(dir->untracked, untracked,
1790 dirname + baselen, len - baselen);
1791
1792 /*
1793 * If this is an excluded directory, then we only need to check if
1794 * the directory contains any files.
1795 */
1796 return read_directory_recursive(dir, istate, dirname, len,
1797 untracked, 1, exclude, pathspec);
1798 }
1799
1800 /*
1801 * This is an inexact early pruning of any recursive directory
1802 * reading - if the path cannot possibly be in the pathspec,
1803 * return true, and we'll skip it early.
1804 */
1805 static int simplify_away(const char *path, int pathlen,
1806 const struct pathspec *pathspec)
1807 {
1808 int i;
1809
1810 if (!pathspec || !pathspec->nr)
1811 return 0;
1812
1813 GUARD_PATHSPEC(pathspec,
1814 PATHSPEC_FROMTOP |
1815 PATHSPEC_MAXDEPTH |
1816 PATHSPEC_LITERAL |
1817 PATHSPEC_GLOB |
1818 PATHSPEC_ICASE |
1819 PATHSPEC_EXCLUDE |
1820 PATHSPEC_ATTR);
1821
1822 for (i = 0; i < pathspec->nr; i++) {
1823 const struct pathspec_item *item = &pathspec->items[i];
1824 int len = item->nowildcard_len;
1825
1826 if (len > pathlen)
1827 len = pathlen;
1828 if (!ps_strncmp(item, item->match, path, len))
1829 return 0;
1830 }
1831
1832 return 1;
1833 }
1834
1835 /*
1836 * This function tells us whether an excluded path matches a
1837 * list of "interesting" pathspecs. That is, whether a path matched
1838 * by any of the pathspecs could possibly be ignored by excluding
1839 * the specified path. This can happen if:
1840 *
1841 * 1. the path is mentioned explicitly in the pathspec
1842 *
1843 * 2. the path is a directory prefix of some element in the
1844 * pathspec
1845 */
1846 static int exclude_matches_pathspec(const char *path, int pathlen,
1847 const struct pathspec *pathspec)
1848 {
1849 int i;
1850
1851 if (!pathspec || !pathspec->nr)
1852 return 0;
1853
1854 GUARD_PATHSPEC(pathspec,
1855 PATHSPEC_FROMTOP |
1856 PATHSPEC_MAXDEPTH |
1857 PATHSPEC_LITERAL |
1858 PATHSPEC_GLOB |
1859 PATHSPEC_ICASE |
1860 PATHSPEC_EXCLUDE);
1861
1862 for (i = 0; i < pathspec->nr; i++) {
1863 const struct pathspec_item *item = &pathspec->items[i];
1864 int len = item->nowildcard_len;
1865
1866 if (len == pathlen &&
1867 !ps_strncmp(item, item->match, path, pathlen))
1868 return 1;
1869 if (len > pathlen &&
1870 item->match[pathlen] == '/' &&
1871 !ps_strncmp(item, item->match, path, pathlen))
1872 return 1;
1873 }
1874 return 0;
1875 }
1876
1877 static int get_index_dtype(struct index_state *istate,
1878 const char *path, int len)
1879 {
1880 int pos;
1881 const struct cache_entry *ce;
1882
1883 ce = index_file_exists(istate, path, len, 0);
1884 if (ce) {
1885 if (!ce_uptodate(ce))
1886 return DT_UNKNOWN;
1887 if (S_ISGITLINK(ce->ce_mode))
1888 return DT_DIR;
1889 /*
1890 * Nobody actually cares about the
1891 * difference between DT_LNK and DT_REG
1892 */
1893 return DT_REG;
1894 }
1895
1896 /* Try to look it up as a directory */
1897 pos = index_name_pos(istate, path, len);
1898 if (pos >= 0)
1899 return DT_UNKNOWN;
1900 pos = -pos-1;
1901 while (pos < istate->cache_nr) {
1902 ce = istate->cache[pos++];
1903 if (strncmp(ce->name, path, len))
1904 break;
1905 if (ce->name[len] > '/')
1906 break;
1907 if (ce->name[len] < '/')
1908 continue;
1909 if (!ce_uptodate(ce))
1910 break; /* continue? */
1911 return DT_DIR;
1912 }
1913 return DT_UNKNOWN;
1914 }
1915
1916 static int resolve_dtype(int dtype, struct index_state *istate,
1917 const char *path, int len)
1918 {
1919 struct stat st;
1920
1921 if (dtype != DT_UNKNOWN)
1922 return dtype;
1923 dtype = get_index_dtype(istate, path, len);
1924 if (dtype != DT_UNKNOWN)
1925 return dtype;
1926 if (lstat(path, &st))
1927 return dtype;
1928 if (S_ISREG(st.st_mode))
1929 return DT_REG;
1930 if (S_ISDIR(st.st_mode))
1931 return DT_DIR;
1932 if (S_ISLNK(st.st_mode))
1933 return DT_LNK;
1934 return dtype;
1935 }
1936
1937 static enum path_treatment treat_one_path(struct dir_struct *dir,
1938 struct untracked_cache_dir *untracked,
1939 struct index_state *istate,
1940 struct strbuf *path,
1941 int baselen,
1942 const struct pathspec *pathspec,
1943 int dtype)
1944 {
1945 int exclude;
1946 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1947 enum path_treatment path_treatment;
1948
1949 dtype = resolve_dtype(dtype, istate, path->buf, path->len);
1950
1951 /* Always exclude indexed files */
1952 if (dtype != DT_DIR && has_path_in_index)
1953 return path_none;
1954
1955 /*
1956 * When we are looking at a directory P in the working tree,
1957 * there are three cases:
1958 *
1959 * (1) P exists in the index. Everything inside the directory P in
1960 * the working tree needs to go when P is checked out from the
1961 * index.
1962 *
1963 * (2) P does not exist in the index, but there is P/Q in the index.
1964 * We know P will stay a directory when we check out the contents
1965 * of the index, but we do not know yet if there is a directory
1966 * P/Q in the working tree to be killed, so we need to recurse.
1967 *
1968 * (3) P does not exist in the index, and there is no P/Q in the index
1969 * to require P to be a directory, either. Only in this case, we
1970 * know that everything inside P will not be killed without
1971 * recursing.
1972 */
1973 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1974 (dtype == DT_DIR) &&
1975 !has_path_in_index &&
1976 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1977 return path_none;
1978
1979 exclude = is_excluded(dir, istate, path->buf, &dtype);
1980
1981 /*
1982 * Excluded? If we don't explicitly want to show
1983 * ignored files, ignore it
1984 */
1985 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1986 return path_excluded;
1987
1988 switch (dtype) {
1989 default:
1990 return path_none;
1991 case DT_DIR:
1992 strbuf_addch(path, '/');
1993 path_treatment = treat_directory(dir, istate, untracked,
1994 path->buf, path->len,
1995 baselen, exclude, pathspec);
1996 /*
1997 * If 1) we only want to return directories that
1998 * match an exclude pattern and 2) this directory does
1999 * not match an exclude pattern but all of its
2000 * contents are excluded, then indicate that we should
2001 * recurse into this directory (instead of marking the
2002 * directory itself as an ignored path).
2003 */
2004 if (!exclude &&
2005 path_treatment == path_excluded &&
2006 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
2007 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
2008 return path_recurse;
2009 return path_treatment;
2010 case DT_REG:
2011 case DT_LNK:
2012 return exclude ? path_excluded : path_untracked;
2013 }
2014 }
2015
2016 static enum path_treatment treat_path_fast(struct dir_struct *dir,
2017 struct untracked_cache_dir *untracked,
2018 struct cached_dir *cdir,
2019 struct index_state *istate,
2020 struct strbuf *path,
2021 int baselen,
2022 const struct pathspec *pathspec)
2023 {
2024 strbuf_setlen(path, baselen);
2025 if (!cdir->ucd) {
2026 strbuf_addstr(path, cdir->file);
2027 return path_untracked;
2028 }
2029 strbuf_addstr(path, cdir->ucd->name);
2030 /* treat_one_path() does this before it calls treat_directory() */
2031 strbuf_complete(path, '/');
2032 if (cdir->ucd->check_only)
2033 /*
2034 * check_only is set as a result of treat_directory() getting
2035 * to its bottom. Verify again the same set of directories
2036 * with check_only set.
2037 */
2038 return read_directory_recursive(dir, istate, path->buf, path->len,
2039 cdir->ucd, 1, 0, pathspec);
2040 /*
2041 * We get path_recurse in the first run when
2042 * directory_exists_in_index() returns index_nonexistent. We
2043 * are sure that new changes in the index does not impact the
2044 * outcome. Return now.
2045 */
2046 return path_recurse;
2047 }
2048
2049 static enum path_treatment treat_path(struct dir_struct *dir,
2050 struct untracked_cache_dir *untracked,
2051 struct cached_dir *cdir,
2052 struct index_state *istate,
2053 struct strbuf *path,
2054 int baselen,
2055 const struct pathspec *pathspec)
2056 {
2057 if (!cdir->d_name)
2058 return treat_path_fast(dir, untracked, cdir, istate, path,
2059 baselen, pathspec);
2060 if (is_dot_or_dotdot(cdir->d_name) || !fspathcmp(cdir->d_name, ".git"))
2061 return path_none;
2062 strbuf_setlen(path, baselen);
2063 strbuf_addstr(path, cdir->d_name);
2064 if (simplify_away(path->buf, path->len, pathspec))
2065 return path_none;
2066
2067 return treat_one_path(dir, untracked, istate, path, baselen, pathspec,
2068 cdir->d_type);
2069 }
2070
2071 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
2072 {
2073 if (!dir)
2074 return;
2075 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
2076 dir->untracked_alloc);
2077 dir->untracked[dir->untracked_nr++] = xstrdup(name);
2078 }
2079
2080 static int valid_cached_dir(struct dir_struct *dir,
2081 struct untracked_cache_dir *untracked,
2082 struct index_state *istate,
2083 struct strbuf *path,
2084 int check_only)
2085 {
2086 struct stat st;
2087
2088 if (!untracked)
2089 return 0;
2090
2091 /*
2092 * With fsmonitor, we can trust the untracked cache's valid field.
2093 */
2094 refresh_fsmonitor(istate);
2095 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
2096 if (lstat(path->len ? path->buf : ".", &st)) {
2097 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
2098 return 0;
2099 }
2100 if (!untracked->valid ||
2101 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
2102 fill_stat_data(&untracked->stat_data, &st);
2103 return 0;
2104 }
2105 }
2106
2107 if (untracked->check_only != !!check_only)
2108 return 0;
2109
2110 /*
2111 * prep_exclude will be called eventually on this directory,
2112 * but it's called much later in last_matching_pattern(). We
2113 * need it now to determine the validity of the cache for this
2114 * path. The next calls will be nearly no-op, the way
2115 * prep_exclude() is designed.
2116 */
2117 if (path->len && path->buf[path->len - 1] != '/') {
2118 strbuf_addch(path, '/');
2119 prep_exclude(dir, istate, path->buf, path->len);
2120 strbuf_setlen(path, path->len - 1);
2121 } else
2122 prep_exclude(dir, istate, path->buf, path->len);
2123
2124 /* hopefully prep_exclude() haven't invalidated this entry... */
2125 return untracked->valid;
2126 }
2127
2128 static int open_cached_dir(struct cached_dir *cdir,
2129 struct dir_struct *dir,
2130 struct untracked_cache_dir *untracked,
2131 struct index_state *istate,
2132 struct strbuf *path,
2133 int check_only)
2134 {
2135 const char *c_path;
2136
2137 memset(cdir, 0, sizeof(*cdir));
2138 cdir->untracked = untracked;
2139 if (valid_cached_dir(dir, untracked, istate, path, check_only))
2140 return 0;
2141 c_path = path->len ? path->buf : ".";
2142 cdir->fdir = opendir(c_path);
2143 if (!cdir->fdir)
2144 warning_errno(_("could not open directory '%s'"), c_path);
2145 if (dir->untracked) {
2146 invalidate_directory(dir->untracked, untracked);
2147 dir->untracked->dir_opened++;
2148 }
2149 if (!cdir->fdir)
2150 return -1;
2151 return 0;
2152 }
2153
2154 static int read_cached_dir(struct cached_dir *cdir)
2155 {
2156 struct dirent *de;
2157
2158 if (cdir->fdir) {
2159 de = readdir(cdir->fdir);
2160 if (!de) {
2161 cdir->d_name = NULL;
2162 cdir->d_type = DT_UNKNOWN;
2163 return -1;
2164 }
2165 cdir->d_name = de->d_name;
2166 cdir->d_type = DTYPE(de);
2167 return 0;
2168 }
2169 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
2170 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
2171 if (!d->recurse) {
2172 cdir->nr_dirs++;
2173 continue;
2174 }
2175 cdir->ucd = d;
2176 cdir->nr_dirs++;
2177 return 0;
2178 }
2179 cdir->ucd = NULL;
2180 if (cdir->nr_files < cdir->untracked->untracked_nr) {
2181 struct untracked_cache_dir *d = cdir->untracked;
2182 cdir->file = d->untracked[cdir->nr_files++];
2183 return 0;
2184 }
2185 return -1;
2186 }
2187
2188 static void close_cached_dir(struct cached_dir *cdir)
2189 {
2190 if (cdir->fdir)
2191 closedir(cdir->fdir);
2192 /*
2193 * We have gone through this directory and found no untracked
2194 * entries. Mark it valid.
2195 */
2196 if (cdir->untracked) {
2197 cdir->untracked->valid = 1;
2198 cdir->untracked->recurse = 1;
2199 }
2200 }
2201
2202 static void add_path_to_appropriate_result_list(struct dir_struct *dir,
2203 struct untracked_cache_dir *untracked,
2204 struct cached_dir *cdir,
2205 struct index_state *istate,
2206 struct strbuf *path,
2207 int baselen,
2208 const struct pathspec *pathspec,
2209 enum path_treatment state)
2210 {
2211 /* add the path to the appropriate result list */
2212 switch (state) {
2213 case path_excluded:
2214 if (dir->flags & DIR_SHOW_IGNORED)
2215 dir_add_name(dir, istate, path->buf, path->len);
2216 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2217 ((dir->flags & DIR_COLLECT_IGNORED) &&
2218 exclude_matches_pathspec(path->buf, path->len,
2219 pathspec)))
2220 dir_add_ignored(dir, istate, path->buf, path->len);
2221 break;
2222
2223 case path_untracked:
2224 if (dir->flags & DIR_SHOW_IGNORED)
2225 break;
2226 dir_add_name(dir, istate, path->buf, path->len);
2227 if (cdir->fdir)
2228 add_untracked(untracked, path->buf + baselen);
2229 break;
2230
2231 default:
2232 break;
2233 }
2234 }
2235
2236 /*
2237 * Read a directory tree. We currently ignore anything but
2238 * directories, regular files and symlinks. That's because git
2239 * doesn't handle them at all yet. Maybe that will change some
2240 * day.
2241 *
2242 * Also, we ignore the name ".git" (even if it is not a directory).
2243 * That likely will not change.
2244 *
2245 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2246 * to signal that a file was found. This is the least significant value that
2247 * indicates that a file was encountered that does not depend on the order of
2248 * whether an untracked or exluded path was encountered first.
2249 *
2250 * Returns the most significant path_treatment value encountered in the scan.
2251 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2252 * significant path_treatment value that will be returned.
2253 */
2254
2255 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
2256 struct index_state *istate, const char *base, int baselen,
2257 struct untracked_cache_dir *untracked, int check_only,
2258 int stop_at_first_file, const struct pathspec *pathspec)
2259 {
2260 /*
2261 * WARNING WARNING WARNING:
2262 *
2263 * Any updates to the traversal logic here may need corresponding
2264 * updates in treat_leading_path(). See the commit message for the
2265 * commit adding this warning as well as the commit preceding it
2266 * for details.
2267 */
2268
2269 struct cached_dir cdir;
2270 enum path_treatment state, subdir_state, dir_state = path_none;
2271 struct strbuf path = STRBUF_INIT;
2272
2273 strbuf_add(&path, base, baselen);
2274
2275 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
2276 goto out;
2277
2278 if (untracked)
2279 untracked->check_only = !!check_only;
2280
2281 while (!read_cached_dir(&cdir)) {
2282 /* check how the file or directory should be treated */
2283 state = treat_path(dir, untracked, &cdir, istate, &path,
2284 baselen, pathspec);
2285
2286 if (state > dir_state)
2287 dir_state = state;
2288
2289 /* recurse into subdir if instructed by treat_path */
2290 if ((state == path_recurse) ||
2291 ((state == path_untracked) &&
2292 (resolve_dtype(cdir.d_type, istate, path.buf, path.len) == DT_DIR) &&
2293 ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2294 (pathspec &&
2295 do_match_pathspec(istate, pathspec, path.buf, path.len,
2296 baselen, NULL, DO_MATCH_LEADING_PATHSPEC) == MATCHED_RECURSIVELY_LEADING_PATHSPEC)))) {
2297 struct untracked_cache_dir *ud;
2298 ud = lookup_untracked(dir->untracked, untracked,
2299 path.buf + baselen,
2300 path.len - baselen);
2301 subdir_state =
2302 read_directory_recursive(dir, istate, path.buf,
2303 path.len, ud,
2304 check_only, stop_at_first_file, pathspec);
2305 if (subdir_state > dir_state)
2306 dir_state = subdir_state;
2307
2308 if (pathspec &&
2309 !match_pathspec(istate, pathspec, path.buf, path.len,
2310 0 /* prefix */, NULL,
2311 0 /* do NOT special case dirs */))
2312 state = path_none;
2313 }
2314
2315 if (check_only) {
2316 if (stop_at_first_file) {
2317 /*
2318 * If stopping at first file, then
2319 * signal that a file was found by
2320 * returning `path_excluded`. This is
2321 * to return a consistent value
2322 * regardless of whether an ignored or
2323 * excluded file happened to be
2324 * encountered 1st.
2325 *
2326 * In current usage, the
2327 * `stop_at_first_file` is passed when
2328 * an ancestor directory has matched
2329 * an exclude pattern, so any found
2330 * files will be excluded.
2331 */
2332 if (dir_state >= path_excluded) {
2333 dir_state = path_excluded;
2334 break;
2335 }
2336 }
2337
2338 /* abort early if maximum state has been reached */
2339 if (dir_state == path_untracked) {
2340 if (cdir.fdir)
2341 add_untracked(untracked, path.buf + baselen);
2342 break;
2343 }
2344 /* skip the dir_add_* part */
2345 continue;
2346 }
2347
2348 add_path_to_appropriate_result_list(dir, untracked, &cdir,
2349 istate, &path, baselen,
2350 pathspec, state);
2351 }
2352 close_cached_dir(&cdir);
2353 out:
2354 strbuf_release(&path);
2355
2356 return dir_state;
2357 }
2358
2359 int cmp_dir_entry(const void *p1, const void *p2)
2360 {
2361 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2362 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2363
2364 return name_compare(e1->name, e1->len, e2->name, e2->len);
2365 }
2366
2367 /* check if *out lexically strictly contains *in */
2368 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2369 {
2370 return (out->len < in->len) &&
2371 (out->name[out->len - 1] == '/') &&
2372 !memcmp(out->name, in->name, out->len);
2373 }
2374
2375 static int treat_leading_path(struct dir_struct *dir,
2376 struct index_state *istate,
2377 const char *path, int len,
2378 const struct pathspec *pathspec)
2379 {
2380 /*
2381 * WARNING WARNING WARNING:
2382 *
2383 * Any updates to the traversal logic here may need corresponding
2384 * updates in read_directory_recursive(). See 777b420347 (dir:
2385 * synchronize treat_leading_path() and read_directory_recursive(),
2386 * 2019-12-19) and its parent commit for details.
2387 */
2388
2389 struct strbuf sb = STRBUF_INIT;
2390 struct strbuf subdir = STRBUF_INIT;
2391 int prevlen, baselen;
2392 const char *cp;
2393 struct cached_dir cdir;
2394 enum path_treatment state = path_none;
2395
2396 /*
2397 * For each directory component of path, we are going to check whether
2398 * that path is relevant given the pathspec. For example, if path is
2399 * foo/bar/baz/
2400 * then we will ask treat_path() whether we should go into foo, then
2401 * whether we should go into bar, then whether baz is relevant.
2402 * Checking each is important because e.g. if path is
2403 * .git/info/
2404 * then we need to check .git to know we shouldn't traverse it.
2405 * If the return from treat_path() is:
2406 * * path_none, for any path, we return false.
2407 * * path_recurse, for all path components, we return true
2408 * * <anything else> for some intermediate component, we make sure
2409 * to add that path to the relevant list but return false
2410 * signifying that we shouldn't recurse into it.
2411 */
2412
2413 while (len && path[len - 1] == '/')
2414 len--;
2415 if (!len)
2416 return 1;
2417
2418 memset(&cdir, 0, sizeof(cdir));
2419 cdir.d_type = DT_DIR;
2420 baselen = 0;
2421 prevlen = 0;
2422 while (1) {
2423 prevlen = baselen + !!baselen;
2424 cp = path + prevlen;
2425 cp = memchr(cp, '/', path + len - cp);
2426 if (!cp)
2427 baselen = len;
2428 else
2429 baselen = cp - path;
2430 strbuf_reset(&sb);
2431 strbuf_add(&sb, path, baselen);
2432 if (!is_directory(sb.buf))
2433 break;
2434 strbuf_reset(&sb);
2435 strbuf_add(&sb, path, prevlen);
2436 strbuf_reset(&subdir);
2437 strbuf_add(&subdir, path+prevlen, baselen-prevlen);
2438 cdir.d_name = subdir.buf;
2439 state = treat_path(dir, NULL, &cdir, istate, &sb, prevlen,
2440 pathspec);
2441 if (state == path_untracked &&
2442 resolve_dtype(cdir.d_type, istate, sb.buf, sb.len) == DT_DIR &&
2443 (dir->flags & DIR_SHOW_IGNORED_TOO ||
2444 do_match_pathspec(istate, pathspec, sb.buf, sb.len,
2445 baselen, NULL, DO_MATCH_LEADING_PATHSPEC) == MATCHED_RECURSIVELY_LEADING_PATHSPEC)) {
2446 if (!match_pathspec(istate, pathspec, sb.buf, sb.len,
2447 0 /* prefix */, NULL,
2448 0 /* do NOT special case dirs */))
2449 state = path_none;
2450 add_path_to_appropriate_result_list(dir, NULL, &cdir,
2451 istate,
2452 &sb, baselen,
2453 pathspec, state);
2454 state = path_recurse;
2455 }
2456
2457 if (state != path_recurse)
2458 break; /* do not recurse into it */
2459 if (len <= baselen)
2460 break; /* finished checking */
2461 }
2462 add_path_to_appropriate_result_list(dir, NULL, &cdir, istate,
2463 &sb, baselen, pathspec,
2464 state);
2465
2466 strbuf_release(&subdir);
2467 strbuf_release(&sb);
2468 return state == path_recurse;
2469 }
2470
2471 static const char *get_ident_string(void)
2472 {
2473 static struct strbuf sb = STRBUF_INIT;
2474 struct utsname uts;
2475
2476 if (sb.len)
2477 return sb.buf;
2478 if (uname(&uts) < 0)
2479 die_errno(_("failed to get kernel name and information"));
2480 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2481 uts.sysname);
2482 return sb.buf;
2483 }
2484
2485 static int ident_in_untracked(const struct untracked_cache *uc)
2486 {
2487 /*
2488 * Previous git versions may have saved many NUL separated
2489 * strings in the "ident" field, but it is insane to manage
2490 * many locations, so just take care of the first one.
2491 */
2492
2493 return !strcmp(uc->ident.buf, get_ident_string());
2494 }
2495
2496 static void set_untracked_ident(struct untracked_cache *uc)
2497 {
2498 strbuf_reset(&uc->ident);
2499 strbuf_addstr(&uc->ident, get_ident_string());
2500
2501 /*
2502 * This strbuf used to contain a list of NUL separated
2503 * strings, so save NUL too for backward compatibility.
2504 */
2505 strbuf_addch(&uc->ident, 0);
2506 }
2507
2508 static void new_untracked_cache(struct index_state *istate)
2509 {
2510 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2511 strbuf_init(&uc->ident, 100);
2512 uc->exclude_per_dir = ".gitignore";
2513 /* should be the same flags used by git-status */
2514 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2515 set_untracked_ident(uc);
2516 istate->untracked = uc;
2517 istate->cache_changed |= UNTRACKED_CHANGED;
2518 }
2519
2520 void add_untracked_cache(struct index_state *istate)
2521 {
2522 if (!istate->untracked) {
2523 new_untracked_cache(istate);
2524 } else {
2525 if (!ident_in_untracked(istate->untracked)) {
2526 free_untracked_cache(istate->untracked);
2527 new_untracked_cache(istate);
2528 }
2529 }
2530 }
2531
2532 void remove_untracked_cache(struct index_state *istate)
2533 {
2534 if (istate->untracked) {
2535 free_untracked_cache(istate->untracked);
2536 istate->untracked = NULL;
2537 istate->cache_changed |= UNTRACKED_CHANGED;
2538 }
2539 }
2540
2541 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2542 int base_len,
2543 const struct pathspec *pathspec)
2544 {
2545 struct untracked_cache_dir *root;
2546 static int untracked_cache_disabled = -1;
2547
2548 if (!dir->untracked)
2549 return NULL;
2550 if (untracked_cache_disabled < 0)
2551 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2552 if (untracked_cache_disabled)
2553 return NULL;
2554
2555 /*
2556 * We only support $GIT_DIR/info/exclude and core.excludesfile
2557 * as the global ignore rule files. Any other additions
2558 * (e.g. from command line) invalidate the cache. This
2559 * condition also catches running setup_standard_excludes()
2560 * before setting dir->untracked!
2561 */
2562 if (dir->unmanaged_exclude_files)
2563 return NULL;
2564
2565 /*
2566 * Optimize for the main use case only: whole-tree git
2567 * status. More work involved in treat_leading_path() if we
2568 * use cache on just a subset of the worktree. pathspec
2569 * support could make the matter even worse.
2570 */
2571 if (base_len || (pathspec && pathspec->nr))
2572 return NULL;
2573
2574 /* Different set of flags may produce different results */
2575 if (dir->flags != dir->untracked->dir_flags ||
2576 /*
2577 * See treat_directory(), case index_nonexistent. Without
2578 * this flag, we may need to also cache .git file content
2579 * for the resolve_gitlink_ref() call, which we don't.
2580 */
2581 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2582 /* We don't support collecting ignore files */
2583 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2584 DIR_COLLECT_IGNORED)))
2585 return NULL;
2586
2587 /*
2588 * If we use .gitignore in the cache and now you change it to
2589 * .gitexclude, everything will go wrong.
2590 */
2591 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2592 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2593 return NULL;
2594
2595 /*
2596 * EXC_CMDL is not considered in the cache. If people set it,
2597 * skip the cache.
2598 */
2599 if (dir->exclude_list_group[EXC_CMDL].nr)
2600 return NULL;
2601
2602 if (!ident_in_untracked(dir->untracked)) {
2603 warning(_("untracked cache is disabled on this system or location"));
2604 return NULL;
2605 }
2606
2607 if (!dir->untracked->root) {
2608 const int len = sizeof(*dir->untracked->root);
2609 dir->untracked->root = xmalloc(len);
2610 memset(dir->untracked->root, 0, len);
2611 }
2612
2613 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2614 root = dir->untracked->root;
2615 if (!oideq(&dir->ss_info_exclude.oid,
2616 &dir->untracked->ss_info_exclude.oid)) {
2617 invalidate_gitignore(dir->untracked, root);
2618 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2619 }
2620 if (!oideq(&dir->ss_excludes_file.oid,
2621 &dir->untracked->ss_excludes_file.oid)) {
2622 invalidate_gitignore(dir->untracked, root);
2623 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2624 }
2625
2626 /* Make sure this directory is not dropped out at saving phase */
2627 root->recurse = 1;
2628 return root;
2629 }
2630
2631 int read_directory(struct dir_struct *dir, struct index_state *istate,
2632 const char *path, int len, const struct pathspec *pathspec)
2633 {
2634 struct untracked_cache_dir *untracked;
2635
2636 trace_performance_enter();
2637
2638 if (has_symlink_leading_path(path, len)) {
2639 trace_performance_leave("read directory %.*s", len, path);
2640 return dir->nr;
2641 }
2642
2643 untracked = validate_untracked_cache(dir, len, pathspec);
2644 if (!untracked)
2645 /*
2646 * make sure untracked cache code path is disabled,
2647 * e.g. prep_exclude()
2648 */
2649 dir->untracked = NULL;
2650 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2651 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2652 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2653 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2654
2655 /*
2656 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2657 * also pick up untracked contents of untracked dirs; by default
2658 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2659 */
2660 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2661 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2662 int i, j;
2663
2664 /* remove from dir->entries untracked contents of untracked dirs */
2665 for (i = j = 0; j < dir->nr; j++) {
2666 if (i &&
2667 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2668 FREE_AND_NULL(dir->entries[j]);
2669 } else {
2670 dir->entries[i++] = dir->entries[j];
2671 }
2672 }
2673
2674 dir->nr = i;
2675 }
2676
2677 trace_performance_leave("read directory %.*s", len, path);
2678 if (dir->untracked) {
2679 static int force_untracked_cache = -1;
2680 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2681
2682 if (force_untracked_cache < 0)
2683 force_untracked_cache =
2684 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2685 trace_printf_key(&trace_untracked_stats,
2686 "node creation: %u\n"
2687 "gitignore invalidation: %u\n"
2688 "directory invalidation: %u\n"
2689 "opendir: %u\n",
2690 dir->untracked->dir_created,
2691 dir->untracked->gitignore_invalidated,
2692 dir->untracked->dir_invalidated,
2693 dir->untracked->dir_opened);
2694 if (force_untracked_cache &&
2695 dir->untracked == istate->untracked &&
2696 (dir->untracked->dir_opened ||
2697 dir->untracked->gitignore_invalidated ||
2698 dir->untracked->dir_invalidated))
2699 istate->cache_changed |= UNTRACKED_CHANGED;
2700 if (dir->untracked != istate->untracked) {
2701 FREE_AND_NULL(dir->untracked);
2702 }
2703 }
2704 return dir->nr;
2705 }
2706
2707 int file_exists(const char *f)
2708 {
2709 struct stat sb;
2710 return lstat(f, &sb) == 0;
2711 }
2712
2713 int repo_file_exists(struct repository *repo, const char *path)
2714 {
2715 if (repo != the_repository)
2716 BUG("do not know how to check file existence in arbitrary repo");
2717
2718 return file_exists(path);
2719 }
2720
2721 static int cmp_icase(char a, char b)
2722 {
2723 if (a == b)
2724 return 0;
2725 if (ignore_case)
2726 return toupper(a) - toupper(b);
2727 return a - b;
2728 }
2729
2730 /*
2731 * Given two normalized paths (a trailing slash is ok), if subdir is
2732 * outside dir, return -1. Otherwise return the offset in subdir that
2733 * can be used as relative path to dir.
2734 */
2735 int dir_inside_of(const char *subdir, const char *dir)
2736 {
2737 int offset = 0;
2738
2739 assert(dir && subdir && *dir && *subdir);
2740
2741 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2742 dir++;
2743 subdir++;
2744 offset++;
2745 }
2746
2747 /* hel[p]/me vs hel[l]/yeah */
2748 if (*dir && *subdir)
2749 return -1;
2750
2751 if (!*subdir)
2752 return !*dir ? offset : -1; /* same dir */
2753
2754 /* foo/[b]ar vs foo/[] */
2755 if (is_dir_sep(dir[-1]))
2756 return is_dir_sep(subdir[-1]) ? offset : -1;
2757
2758 /* foo[/]bar vs foo[] */
2759 return is_dir_sep(*subdir) ? offset + 1 : -1;
2760 }
2761
2762 int is_inside_dir(const char *dir)
2763 {
2764 char *cwd;
2765 int rc;
2766
2767 if (!dir)
2768 return 0;
2769
2770 cwd = xgetcwd();
2771 rc = (dir_inside_of(cwd, dir) >= 0);
2772 free(cwd);
2773 return rc;
2774 }
2775
2776 int is_empty_dir(const char *path)
2777 {
2778 DIR *dir = opendir(path);
2779 struct dirent *e;
2780 int ret = 1;
2781
2782 if (!dir)
2783 return 0;
2784
2785 while ((e = readdir(dir)) != NULL)
2786 if (!is_dot_or_dotdot(e->d_name)) {
2787 ret = 0;
2788 break;
2789 }
2790
2791 closedir(dir);
2792 return ret;
2793 }
2794
2795 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2796 {
2797 DIR *dir;
2798 struct dirent *e;
2799 int ret = 0, original_len = path->len, len, kept_down = 0;
2800 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2801 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2802 struct object_id submodule_head;
2803
2804 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2805 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2806 /* Do not descend and nuke a nested git work tree. */
2807 if (kept_up)
2808 *kept_up = 1;
2809 return 0;
2810 }
2811
2812 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2813 dir = opendir(path->buf);
2814 if (!dir) {
2815 if (errno == ENOENT)
2816 return keep_toplevel ? -1 : 0;
2817 else if (errno == EACCES && !keep_toplevel)
2818 /*
2819 * An empty dir could be removable even if it
2820 * is unreadable:
2821 */
2822 return rmdir(path->buf);
2823 else
2824 return -1;
2825 }
2826 strbuf_complete(path, '/');
2827
2828 len = path->len;
2829 while ((e = readdir(dir)) != NULL) {
2830 struct stat st;
2831 if (is_dot_or_dotdot(e->d_name))
2832 continue;
2833
2834 strbuf_setlen(path, len);
2835 strbuf_addstr(path, e->d_name);
2836 if (lstat(path->buf, &st)) {
2837 if (errno == ENOENT)
2838 /*
2839 * file disappeared, which is what we
2840 * wanted anyway
2841 */
2842 continue;
2843 /* fall through */
2844 } else if (S_ISDIR(st.st_mode)) {
2845 if (!remove_dir_recurse(path, flag, &kept_down))
2846 continue; /* happy */
2847 } else if (!only_empty &&
2848 (!unlink(path->buf) || errno == ENOENT)) {
2849 continue; /* happy, too */
2850 }
2851
2852 /* path too long, stat fails, or non-directory still exists */
2853 ret = -1;
2854 break;
2855 }
2856 closedir(dir);
2857
2858 strbuf_setlen(path, original_len);
2859 if (!ret && !keep_toplevel && !kept_down)
2860 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2861 else if (kept_up)
2862 /*
2863 * report the uplevel that it is not an error that we
2864 * did not rmdir() our directory.
2865 */
2866 *kept_up = !ret;
2867 return ret;
2868 }
2869
2870 int remove_dir_recursively(struct strbuf *path, int flag)
2871 {
2872 return remove_dir_recurse(path, flag, NULL);
2873 }
2874
2875 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2876
2877 void setup_standard_excludes(struct dir_struct *dir)
2878 {
2879 dir->exclude_per_dir = ".gitignore";
2880
2881 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2882 if (!excludes_file)
2883 excludes_file = xdg_config_home("ignore");
2884 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2885 add_patterns_from_file_1(dir, excludes_file,
2886 dir->untracked ? &dir->ss_excludes_file : NULL);
2887
2888 /* per repository user preference */
2889 if (startup_info->have_repository) {
2890 const char *path = git_path_info_exclude();
2891 if (!access_or_warn(path, R_OK, 0))
2892 add_patterns_from_file_1(dir, path,
2893 dir->untracked ? &dir->ss_info_exclude : NULL);
2894 }
2895 }
2896
2897 int remove_path(const char *name)
2898 {
2899 char *slash;
2900
2901 if (unlink(name) && !is_missing_file_error(errno))
2902 return -1;
2903
2904 slash = strrchr(name, '/');
2905 if (slash) {
2906 char *dirs = xstrdup(name);
2907 slash = dirs + (slash - name);
2908 do {
2909 *slash = '\0';
2910 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2911 free(dirs);
2912 }
2913 return 0;
2914 }
2915
2916 /*
2917 * Frees memory within dir which was allocated for exclude lists and
2918 * the exclude_stack. Does not free dir itself.
2919 */
2920 void clear_directory(struct dir_struct *dir)
2921 {
2922 int i, j;
2923 struct exclude_list_group *group;
2924 struct pattern_list *pl;
2925 struct exclude_stack *stk;
2926
2927 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2928 group = &dir->exclude_list_group[i];
2929 for (j = 0; j < group->nr; j++) {
2930 pl = &group->pl[j];
2931 if (i == EXC_DIRS)
2932 free((char *)pl->src);
2933 clear_pattern_list(pl);
2934 }
2935 free(group->pl);
2936 }
2937
2938 stk = dir->exclude_stack;
2939 while (stk) {
2940 struct exclude_stack *prev = stk->prev;
2941 free(stk);
2942 stk = prev;
2943 }
2944 strbuf_release(&dir->basebuf);
2945 }
2946
2947 struct ondisk_untracked_cache {
2948 struct stat_data info_exclude_stat;
2949 struct stat_data excludes_file_stat;
2950 uint32_t dir_flags;
2951 };
2952
2953 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2954
2955 struct write_data {
2956 int index; /* number of written untracked_cache_dir */
2957 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2958 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2959 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2960 struct strbuf out;
2961 struct strbuf sb_stat;
2962 struct strbuf sb_sha1;
2963 };
2964
2965 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2966 {
2967 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2968 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2969 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2970 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2971 to->sd_dev = htonl(from->sd_dev);
2972 to->sd_ino = htonl(from->sd_ino);
2973 to->sd_uid = htonl(from->sd_uid);
2974 to->sd_gid = htonl(from->sd_gid);
2975 to->sd_size = htonl(from->sd_size);
2976 }
2977
2978 static void write_one_dir(struct untracked_cache_dir *untracked,
2979 struct write_data *wd)
2980 {
2981 struct stat_data stat_data;
2982 struct strbuf *out = &wd->out;
2983 unsigned char intbuf[16];
2984 unsigned int intlen, value;
2985 int i = wd->index++;
2986
2987 /*
2988 * untracked_nr should be reset whenever valid is clear, but
2989 * for safety..
2990 */
2991 if (!untracked->valid) {
2992 untracked->untracked_nr = 0;
2993 untracked->check_only = 0;
2994 }
2995
2996 if (untracked->check_only)
2997 ewah_set(wd->check_only, i);
2998 if (untracked->valid) {
2999 ewah_set(wd->valid, i);
3000 stat_data_to_disk(&stat_data, &untracked->stat_data);
3001 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
3002 }
3003 if (!is_null_oid(&untracked->exclude_oid)) {
3004 ewah_set(wd->sha1_valid, i);
3005 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
3006 the_hash_algo->rawsz);
3007 }
3008
3009 intlen = encode_varint(untracked->untracked_nr, intbuf);
3010 strbuf_add(out, intbuf, intlen);
3011
3012 /* skip non-recurse directories */
3013 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
3014 if (untracked->dirs[i]->recurse)
3015 value++;
3016 intlen = encode_varint(value, intbuf);
3017 strbuf_add(out, intbuf, intlen);
3018
3019 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
3020
3021 for (i = 0; i < untracked->untracked_nr; i++)
3022 strbuf_add(out, untracked->untracked[i],
3023 strlen(untracked->untracked[i]) + 1);
3024
3025 for (i = 0; i < untracked->dirs_nr; i++)
3026 if (untracked->dirs[i]->recurse)
3027 write_one_dir(untracked->dirs[i], wd);
3028 }
3029
3030 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
3031 {
3032 struct ondisk_untracked_cache *ouc;
3033 struct write_data wd;
3034 unsigned char varbuf[16];
3035 int varint_len;
3036 const unsigned hashsz = the_hash_algo->rawsz;
3037
3038 ouc = xcalloc(1, sizeof(*ouc));
3039 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
3040 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
3041 ouc->dir_flags = htonl(untracked->dir_flags);
3042
3043 varint_len = encode_varint(untracked->ident.len, varbuf);
3044 strbuf_add(out, varbuf, varint_len);
3045 strbuf_addbuf(out, &untracked->ident);
3046
3047 strbuf_add(out, ouc, sizeof(*ouc));
3048 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
3049 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
3050 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
3051 FREE_AND_NULL(ouc);
3052
3053 if (!untracked->root) {
3054 varint_len = encode_varint(0, varbuf);
3055 strbuf_add(out, varbuf, varint_len);
3056 return;
3057 }
3058
3059 wd.index = 0;
3060 wd.check_only = ewah_new();
3061 wd.valid = ewah_new();
3062 wd.sha1_valid = ewah_new();
3063 strbuf_init(&wd.out, 1024);
3064 strbuf_init(&wd.sb_stat, 1024);
3065 strbuf_init(&wd.sb_sha1, 1024);
3066 write_one_dir(untracked->root, &wd);
3067
3068 varint_len = encode_varint(wd.index, varbuf);
3069 strbuf_add(out, varbuf, varint_len);
3070 strbuf_addbuf(out, &wd.out);
3071 ewah_serialize_strbuf(wd.valid, out);
3072 ewah_serialize_strbuf(wd.check_only, out);
3073 ewah_serialize_strbuf(wd.sha1_valid, out);
3074 strbuf_addbuf(out, &wd.sb_stat);
3075 strbuf_addbuf(out, &wd.sb_sha1);
3076 strbuf_addch(out, '\0'); /* safe guard for string lists */
3077
3078 ewah_free(wd.valid);
3079 ewah_free(wd.check_only);
3080 ewah_free(wd.sha1_valid);
3081 strbuf_release(&wd.out);
3082 strbuf_release(&wd.sb_stat);
3083 strbuf_release(&wd.sb_sha1);
3084 }
3085
3086 static void free_untracked(struct untracked_cache_dir *ucd)
3087 {
3088 int i;
3089 if (!ucd)
3090 return;
3091 for (i = 0; i < ucd->dirs_nr; i++)
3092 free_untracked(ucd->dirs[i]);
3093 for (i = 0; i < ucd->untracked_nr; i++)
3094 free(ucd->untracked[i]);
3095 free(ucd->untracked);
3096 free(ucd->dirs);
3097 free(ucd);
3098 }
3099
3100 void free_untracked_cache(struct untracked_cache *uc)
3101 {
3102 if (uc)
3103 free_untracked(uc->root);
3104 free(uc);
3105 }
3106
3107 struct read_data {
3108 int index;
3109 struct untracked_cache_dir **ucd;
3110 struct ewah_bitmap *check_only;
3111 struct ewah_bitmap *valid;
3112 struct ewah_bitmap *sha1_valid;
3113 const unsigned char *data;
3114 const unsigned char *end;
3115 };
3116
3117 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
3118 {
3119 memcpy(to, data, sizeof(*to));
3120 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
3121 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
3122 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
3123 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
3124 to->sd_dev = ntohl(to->sd_dev);
3125 to->sd_ino = ntohl(to->sd_ino);
3126 to->sd_uid = ntohl(to->sd_uid);
3127 to->sd_gid = ntohl(to->sd_gid);
3128 to->sd_size = ntohl(to->sd_size);
3129 }
3130
3131 static int read_one_dir(struct untracked_cache_dir **untracked_,
3132 struct read_data *rd)
3133 {
3134 struct untracked_cache_dir ud, *untracked;
3135 const unsigned char *data = rd->data, *end = rd->end;
3136 const unsigned char *eos;
3137 unsigned int value;
3138 int i;
3139
3140 memset(&ud, 0, sizeof(ud));
3141
3142 value = decode_varint(&data);
3143 if (data > end)
3144 return -1;
3145 ud.recurse = 1;
3146 ud.untracked_alloc = value;
3147 ud.untracked_nr = value;
3148 if (ud.untracked_nr)
3149 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
3150
3151 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
3152 if (data > end)
3153 return -1;
3154 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
3155
3156 eos = memchr(data, '\0', end - data);
3157 if (!eos || eos == end)
3158 return -1;
3159
3160 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
3161 memcpy(untracked, &ud, sizeof(ud));
3162 memcpy(untracked->name, data, eos - data + 1);
3163 data = eos + 1;
3164
3165 for (i = 0; i < untracked->untracked_nr; i++) {
3166 eos = memchr(data, '\0', end - data);
3167 if (!eos || eos == end)
3168 return -1;
3169 untracked->untracked[i] = xmemdupz(data, eos - data);
3170 data = eos + 1;
3171 }
3172
3173 rd->ucd[rd->index++] = untracked;
3174 rd->data = data;
3175
3176 for (i = 0; i < untracked->dirs_nr; i++) {
3177 if (read_one_dir(untracked->dirs + i, rd) < 0)
3178 return -1;
3179 }
3180 return 0;
3181 }
3182
3183 static void set_check_only(size_t pos, void *cb)
3184 {
3185 struct read_data *rd = cb;
3186 struct untracked_cache_dir *ud = rd->ucd[pos];
3187 ud->check_only = 1;
3188 }
3189
3190 static void read_stat(size_t pos, void *cb)
3191 {
3192 struct read_data *rd = cb;
3193 struct untracked_cache_dir *ud = rd->ucd[pos];
3194 if (rd->data + sizeof(struct stat_data) > rd->end) {
3195 rd->data = rd->end + 1;
3196 return;
3197 }
3198 stat_data_from_disk(&ud->stat_data, rd->data);
3199 rd->data += sizeof(struct stat_data);
3200 ud->valid = 1;
3201 }
3202
3203 static void read_oid(size_t pos, void *cb)
3204 {
3205 struct read_data *rd = cb;
3206 struct untracked_cache_dir *ud = rd->ucd[pos];
3207 if (rd->data + the_hash_algo->rawsz > rd->end) {
3208 rd->data = rd->end + 1;
3209 return;
3210 }
3211 hashcpy(ud->exclude_oid.hash, rd->data);
3212 rd->data += the_hash_algo->rawsz;
3213 }
3214
3215 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
3216 const unsigned char *sha1)
3217 {
3218 stat_data_from_disk(&oid_stat->stat, data);
3219 hashcpy(oid_stat->oid.hash, sha1);
3220 oid_stat->valid = 1;
3221 }
3222
3223 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
3224 {
3225 struct untracked_cache *uc;
3226 struct read_data rd;
3227 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
3228 const char *ident;
3229 int ident_len;
3230 ssize_t len;
3231 const char *exclude_per_dir;
3232 const unsigned hashsz = the_hash_algo->rawsz;
3233 const unsigned offset = sizeof(struct ondisk_untracked_cache);
3234 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
3235
3236 if (sz <= 1 || end[-1] != '\0')
3237 return NULL;
3238 end--;
3239
3240 ident_len = decode_varint(&next);
3241 if (next + ident_len > end)
3242 return NULL;
3243 ident = (const char *)next;
3244 next += ident_len;
3245
3246 if (next + exclude_per_dir_offset + 1 > end)
3247 return NULL;
3248
3249 uc = xcalloc(1, sizeof(*uc));
3250 strbuf_init(&uc->ident, ident_len);
3251 strbuf_add(&uc->ident, ident, ident_len);
3252 load_oid_stat(&uc->ss_info_exclude,
3253 next + ouc_offset(info_exclude_stat),
3254 next + offset);
3255 load_oid_stat(&uc->ss_excludes_file,
3256 next + ouc_offset(excludes_file_stat),
3257 next + offset + hashsz);
3258 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
3259 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
3260 uc->exclude_per_dir = xstrdup(exclude_per_dir);
3261 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3262 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
3263 if (next >= end)
3264 goto done2;
3265
3266 len = decode_varint(&next);
3267 if (next > end || len == 0)
3268 goto done2;
3269
3270 rd.valid = ewah_new();
3271 rd.check_only = ewah_new();
3272 rd.sha1_valid = ewah_new();
3273 rd.data = next;
3274 rd.end = end;
3275 rd.index = 0;
3276 ALLOC_ARRAY(rd.ucd, len);
3277
3278 if (read_one_dir(&uc->root, &rd) || rd.index != len)
3279 goto done;
3280
3281 next = rd.data;
3282 len = ewah_read_mmap(rd.valid, next, end - next);
3283 if (len < 0)
3284 goto done;
3285
3286 next += len;
3287 len = ewah_read_mmap(rd.check_only, next, end - next);
3288 if (len < 0)
3289 goto done;
3290
3291 next += len;
3292 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
3293 if (len < 0)
3294 goto done;
3295
3296 ewah_each_bit(rd.check_only, set_check_only, &rd);
3297 rd.data = next + len;
3298 ewah_each_bit(rd.valid, read_stat, &rd);
3299 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
3300 next = rd.data;
3301
3302 done:
3303 free(rd.ucd);
3304 ewah_free(rd.valid);
3305 ewah_free(rd.check_only);
3306 ewah_free(rd.sha1_valid);
3307 done2:
3308 if (next != end) {
3309 free_untracked_cache(uc);
3310 uc = NULL;
3311 }
3312 return uc;
3313 }
3314
3315 static void invalidate_one_directory(struct untracked_cache *uc,
3316 struct untracked_cache_dir *ucd)
3317 {
3318 uc->dir_invalidated++;
3319 ucd->valid = 0;
3320 ucd->untracked_nr = 0;
3321 }
3322
3323 /*
3324 * Normally when an entry is added or removed from a directory,
3325 * invalidating that directory is enough. No need to touch its
3326 * ancestors. When a directory is shown as "foo/bar/" in git-status
3327 * however, deleting or adding an entry may have cascading effect.
3328 *
3329 * Say the "foo/bar/file" has become untracked, we need to tell the
3330 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3331 * directory any more (because "bar" is managed by foo as an untracked
3332 * "file").
3333 *
3334 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3335 * was the last untracked entry in the entire "foo", we should show
3336 * "foo/" instead. Which means we have to invalidate past "bar" up to
3337 * "foo".
3338 *
3339 * This function traverses all directories from root to leaf. If there
3340 * is a chance of one of the above cases happening, we invalidate back
3341 * to root. Otherwise we just invalidate the leaf. There may be a more
3342 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3343 * detect these cases and avoid unnecessary invalidation, for example,
3344 * checking for the untracked entry named "bar/" in "foo", but for now
3345 * stick to something safe and simple.
3346 */
3347 static int invalidate_one_component(struct untracked_cache *uc,
3348 struct untracked_cache_dir *dir,
3349 const char *path, int len)
3350 {
3351 const char *rest = strchr(path, '/');
3352
3353 if (rest) {
3354 int component_len = rest - path;
3355 struct untracked_cache_dir *d =
3356 lookup_untracked(uc, dir, path, component_len);
3357 int ret =
3358 invalidate_one_component(uc, d, rest + 1,
3359 len - (component_len + 1));
3360 if (ret)
3361 invalidate_one_directory(uc, dir);
3362 return ret;
3363 }
3364
3365 invalidate_one_directory(uc, dir);
3366 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
3367 }
3368
3369 void untracked_cache_invalidate_path(struct index_state *istate,
3370 const char *path, int safe_path)
3371 {
3372 if (!istate->untracked || !istate->untracked->root)
3373 return;
3374 if (!safe_path && !verify_path(path, 0))
3375 return;
3376 invalidate_one_component(istate->untracked, istate->untracked->root,
3377 path, strlen(path));
3378 }
3379
3380 void untracked_cache_remove_from_index(struct index_state *istate,
3381 const char *path)
3382 {
3383 untracked_cache_invalidate_path(istate, path, 1);
3384 }
3385
3386 void untracked_cache_add_to_index(struct index_state *istate,
3387 const char *path)
3388 {
3389 untracked_cache_invalidate_path(istate, path, 1);
3390 }
3391
3392 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3393 const char *sub_gitdir)
3394 {
3395 int i;
3396 struct repository subrepo;
3397 struct strbuf sub_wt = STRBUF_INIT;
3398 struct strbuf sub_gd = STRBUF_INIT;
3399
3400 const struct submodule *sub;
3401
3402 /* If the submodule has no working tree, we can ignore it. */
3403 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3404 return;
3405
3406 if (repo_read_index(&subrepo) < 0)
3407 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3408
3409 for (i = 0; i < subrepo.index->cache_nr; i++) {
3410 const struct cache_entry *ce = subrepo.index->cache[i];
3411
3412 if (!S_ISGITLINK(ce->ce_mode))
3413 continue;
3414
3415 while (i + 1 < subrepo.index->cache_nr &&
3416 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3417 /*
3418 * Skip entries with the same name in different stages
3419 * to make sure an entry is returned only once.
3420 */
3421 i++;
3422
3423 sub = submodule_from_path(&subrepo, &null_oid, ce->name);
3424 if (!sub || !is_submodule_active(&subrepo, ce->name))
3425 /* .gitmodules broken or inactive sub */
3426 continue;
3427
3428 strbuf_reset(&sub_wt);
3429 strbuf_reset(&sub_gd);
3430 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3431 strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name);
3432
3433 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3434 }
3435 strbuf_release(&sub_wt);
3436 strbuf_release(&sub_gd);
3437 repo_clear(&subrepo);
3438 }
3439
3440 void connect_work_tree_and_git_dir(const char *work_tree_,
3441 const char *git_dir_,
3442 int recurse_into_nested)
3443 {
3444 struct strbuf gitfile_sb = STRBUF_INIT;
3445 struct strbuf cfg_sb = STRBUF_INIT;
3446 struct strbuf rel_path = STRBUF_INIT;
3447 char *git_dir, *work_tree;
3448
3449 /* Prepare .git file */
3450 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3451 if (safe_create_leading_directories_const(gitfile_sb.buf))
3452 die(_("could not create directories for %s"), gitfile_sb.buf);
3453
3454 /* Prepare config file */
3455 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3456 if (safe_create_leading_directories_const(cfg_sb.buf))
3457 die(_("could not create directories for %s"), cfg_sb.buf);
3458
3459 git_dir = real_pathdup(git_dir_, 1);
3460 work_tree = real_pathdup(work_tree_, 1);
3461
3462 /* Write .git file */
3463 write_file(gitfile_sb.buf, "gitdir: %s",
3464 relative_path(git_dir, work_tree, &rel_path));
3465 /* Update core.worktree setting */
3466 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3467 relative_path(work_tree, git_dir, &rel_path));
3468
3469 strbuf_release(&gitfile_sb);
3470 strbuf_release(&cfg_sb);
3471 strbuf_release(&rel_path);
3472
3473 if (recurse_into_nested)
3474 connect_wt_gitdir_in_nested(work_tree, git_dir);
3475
3476 free(work_tree);
3477 free(git_dir);
3478 }
3479
3480 /*
3481 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3482 */
3483 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3484 {
3485 if (rename(old_git_dir, new_git_dir) < 0)
3486 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3487 old_git_dir, new_git_dir);
3488
3489 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3490 }