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