]> git.ipfire.org Git - thirdparty/git.git/blob - dir.c
refs: move doc to refs.h
[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 void add_pattern(const char *string, const char *base,
613 int baselen, struct pattern_list *pl, int srcpos)
614 {
615 struct path_pattern *pattern;
616 int patternlen;
617 unsigned flags;
618 int nowildcardlen;
619
620 parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen);
621 if (flags & PATTERN_FLAG_MUSTBEDIR) {
622 FLEXPTR_ALLOC_MEM(pattern, pattern, string, patternlen);
623 } else {
624 pattern = xmalloc(sizeof(*pattern));
625 pattern->pattern = string;
626 }
627 pattern->patternlen = patternlen;
628 pattern->nowildcardlen = nowildcardlen;
629 pattern->base = base;
630 pattern->baselen = baselen;
631 pattern->flags = flags;
632 pattern->srcpos = srcpos;
633 ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc);
634 pl->patterns[pl->nr++] = pattern;
635 pattern->pl = pl;
636 }
637
638 static int read_skip_worktree_file_from_index(const struct index_state *istate,
639 const char *path,
640 size_t *size_out, char **data_out,
641 struct oid_stat *oid_stat)
642 {
643 int pos, len;
644
645 len = strlen(path);
646 pos = index_name_pos(istate, path, len);
647 if (pos < 0)
648 return -1;
649 if (!ce_skip_worktree(istate->cache[pos]))
650 return -1;
651
652 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
653 }
654
655 /*
656 * Frees memory within pl which was allocated for exclude patterns and
657 * the file buffer. Does not free pl itself.
658 */
659 void clear_pattern_list(struct pattern_list *pl)
660 {
661 int i;
662
663 for (i = 0; i < pl->nr; i++)
664 free(pl->patterns[i]);
665 free(pl->patterns);
666 free(pl->filebuf);
667
668 memset(pl, 0, sizeof(*pl));
669 }
670
671 static void trim_trailing_spaces(char *buf)
672 {
673 char *p, *last_space = NULL;
674
675 for (p = buf; *p; p++)
676 switch (*p) {
677 case ' ':
678 if (!last_space)
679 last_space = p;
680 break;
681 case '\\':
682 p++;
683 if (!*p)
684 return;
685 /* fallthrough */
686 default:
687 last_space = NULL;
688 }
689
690 if (last_space)
691 *last_space = '\0';
692 }
693
694 /*
695 * Given a subdirectory name and "dir" of the current directory,
696 * search the subdir in "dir" and return it, or create a new one if it
697 * does not exist in "dir".
698 *
699 * If "name" has the trailing slash, it'll be excluded in the search.
700 */
701 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
702 struct untracked_cache_dir *dir,
703 const char *name, int len)
704 {
705 int first, last;
706 struct untracked_cache_dir *d;
707 if (!dir)
708 return NULL;
709 if (len && name[len - 1] == '/')
710 len--;
711 first = 0;
712 last = dir->dirs_nr;
713 while (last > first) {
714 int cmp, next = first + ((last - first) >> 1);
715 d = dir->dirs[next];
716 cmp = strncmp(name, d->name, len);
717 if (!cmp && strlen(d->name) > len)
718 cmp = -1;
719 if (!cmp)
720 return d;
721 if (cmp < 0) {
722 last = next;
723 continue;
724 }
725 first = next+1;
726 }
727
728 uc->dir_created++;
729 FLEX_ALLOC_MEM(d, name, name, len);
730
731 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
732 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
733 dir->dirs_nr - first);
734 dir->dirs_nr++;
735 dir->dirs[first] = d;
736 return d;
737 }
738
739 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
740 {
741 int i;
742 dir->valid = 0;
743 dir->untracked_nr = 0;
744 for (i = 0; i < dir->dirs_nr; i++)
745 do_invalidate_gitignore(dir->dirs[i]);
746 }
747
748 static void invalidate_gitignore(struct untracked_cache *uc,
749 struct untracked_cache_dir *dir)
750 {
751 uc->gitignore_invalidated++;
752 do_invalidate_gitignore(dir);
753 }
754
755 static void invalidate_directory(struct untracked_cache *uc,
756 struct untracked_cache_dir *dir)
757 {
758 int i;
759
760 /*
761 * Invalidation increment here is just roughly correct. If
762 * untracked_nr or any of dirs[].recurse is non-zero, we
763 * should increment dir_invalidated too. But that's more
764 * expensive to do.
765 */
766 if (dir->valid)
767 uc->dir_invalidated++;
768
769 dir->valid = 0;
770 dir->untracked_nr = 0;
771 for (i = 0; i < dir->dirs_nr; i++)
772 dir->dirs[i]->recurse = 0;
773 }
774
775 static int add_patterns_from_buffer(char *buf, size_t size,
776 const char *base, int baselen,
777 struct pattern_list *pl);
778
779 /*
780 * Given a file with name "fname", read it (either from disk, or from
781 * an index if 'istate' is non-null), parse it and store the
782 * exclude rules in "pl".
783 *
784 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
785 * stat data from disk (only valid if add_patterns returns zero). If
786 * ss_valid is non-zero, "ss" must contain good value as input.
787 */
788 static int add_patterns(const char *fname, const char *base, int baselen,
789 struct pattern_list *pl, struct index_state *istate,
790 struct oid_stat *oid_stat)
791 {
792 struct stat st;
793 int r;
794 int fd;
795 size_t size = 0;
796 char *buf;
797
798 fd = open(fname, O_RDONLY);
799 if (fd < 0 || fstat(fd, &st) < 0) {
800 if (fd < 0)
801 warn_on_fopen_errors(fname);
802 else
803 close(fd);
804 if (!istate)
805 return -1;
806 r = read_skip_worktree_file_from_index(istate, fname,
807 &size, &buf,
808 oid_stat);
809 if (r != 1)
810 return r;
811 } else {
812 size = xsize_t(st.st_size);
813 if (size == 0) {
814 if (oid_stat) {
815 fill_stat_data(&oid_stat->stat, &st);
816 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
817 oid_stat->valid = 1;
818 }
819 close(fd);
820 return 0;
821 }
822 buf = xmallocz(size);
823 if (read_in_full(fd, buf, size) != size) {
824 free(buf);
825 close(fd);
826 return -1;
827 }
828 buf[size++] = '\n';
829 close(fd);
830 if (oid_stat) {
831 int pos;
832 if (oid_stat->valid &&
833 !match_stat_data_racy(istate, &oid_stat->stat, &st))
834 ; /* no content change, ss->sha1 still good */
835 else if (istate &&
836 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
837 !ce_stage(istate->cache[pos]) &&
838 ce_uptodate(istate->cache[pos]) &&
839 !would_convert_to_git(istate, fname))
840 oidcpy(&oid_stat->oid,
841 &istate->cache[pos]->oid);
842 else
843 hash_object_file(buf, size, "blob",
844 &oid_stat->oid);
845 fill_stat_data(&oid_stat->stat, &st);
846 oid_stat->valid = 1;
847 }
848 }
849
850 add_patterns_from_buffer(buf, size, base, baselen, pl);
851 return 0;
852 }
853
854 static int add_patterns_from_buffer(char *buf, size_t size,
855 const char *base, int baselen,
856 struct pattern_list *pl)
857 {
858 int i, lineno = 1;
859 char *entry;
860
861 pl->filebuf = buf;
862
863 if (skip_utf8_bom(&buf, size))
864 size -= buf - pl->filebuf;
865
866 entry = buf;
867
868 for (i = 0; i < size; i++) {
869 if (buf[i] == '\n') {
870 if (entry != buf + i && entry[0] != '#') {
871 buf[i - (i && buf[i-1] == '\r')] = 0;
872 trim_trailing_spaces(entry);
873 add_pattern(entry, base, baselen, pl, lineno);
874 }
875 lineno++;
876 entry = buf + i + 1;
877 }
878 }
879 return 0;
880 }
881
882 int add_patterns_from_file_to_list(const char *fname, const char *base,
883 int baselen, struct pattern_list *pl,
884 struct index_state *istate)
885 {
886 return add_patterns(fname, base, baselen, pl, istate, NULL);
887 }
888
889 int add_patterns_from_blob_to_list(
890 struct object_id *oid,
891 const char *base, int baselen,
892 struct pattern_list *pl)
893 {
894 char *buf;
895 size_t size;
896 int r;
897
898 r = do_read_blob(oid, NULL, &size, &buf);
899 if (r != 1)
900 return r;
901
902 add_patterns_from_buffer(buf, size, base, baselen, pl);
903 return 0;
904 }
905
906 struct pattern_list *add_pattern_list(struct dir_struct *dir,
907 int group_type, const char *src)
908 {
909 struct pattern_list *pl;
910 struct exclude_list_group *group;
911
912 group = &dir->exclude_list_group[group_type];
913 ALLOC_GROW(group->pl, group->nr + 1, group->alloc);
914 pl = &group->pl[group->nr++];
915 memset(pl, 0, sizeof(*pl));
916 pl->src = src;
917 return pl;
918 }
919
920 /*
921 * Used to set up core.excludesfile and .git/info/exclude lists.
922 */
923 static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname,
924 struct oid_stat *oid_stat)
925 {
926 struct pattern_list *pl;
927 /*
928 * catch setup_standard_excludes() that's called before
929 * dir->untracked is assigned. That function behaves
930 * differently when dir->untracked is non-NULL.
931 */
932 if (!dir->untracked)
933 dir->unmanaged_exclude_files++;
934 pl = add_pattern_list(dir, EXC_FILE, fname);
935 if (add_patterns(fname, "", 0, pl, NULL, oid_stat) < 0)
936 die(_("cannot use %s as an exclude file"), fname);
937 }
938
939 void add_patterns_from_file(struct dir_struct *dir, const char *fname)
940 {
941 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
942 add_patterns_from_file_1(dir, fname, NULL);
943 }
944
945 int match_basename(const char *basename, int basenamelen,
946 const char *pattern, int prefix, int patternlen,
947 unsigned flags)
948 {
949 if (prefix == patternlen) {
950 if (patternlen == basenamelen &&
951 !fspathncmp(pattern, basename, basenamelen))
952 return 1;
953 } else if (flags & PATTERN_FLAG_ENDSWITH) {
954 /* "*literal" matching against "fooliteral" */
955 if (patternlen - 1 <= basenamelen &&
956 !fspathncmp(pattern + 1,
957 basename + basenamelen - (patternlen - 1),
958 patternlen - 1))
959 return 1;
960 } else {
961 if (fnmatch_icase_mem(pattern, patternlen,
962 basename, basenamelen,
963 0) == 0)
964 return 1;
965 }
966 return 0;
967 }
968
969 int match_pathname(const char *pathname, int pathlen,
970 const char *base, int baselen,
971 const char *pattern, int prefix, int patternlen,
972 unsigned flags)
973 {
974 const char *name;
975 int namelen;
976
977 /*
978 * match with FNM_PATHNAME; the pattern has base implicitly
979 * in front of it.
980 */
981 if (*pattern == '/') {
982 pattern++;
983 patternlen--;
984 prefix--;
985 }
986
987 /*
988 * baselen does not count the trailing slash. base[] may or
989 * may not end with a trailing slash though.
990 */
991 if (pathlen < baselen + 1 ||
992 (baselen && pathname[baselen] != '/') ||
993 fspathncmp(pathname, base, baselen))
994 return 0;
995
996 namelen = baselen ? pathlen - baselen - 1 : pathlen;
997 name = pathname + pathlen - namelen;
998
999 if (prefix) {
1000 /*
1001 * if the non-wildcard part is longer than the
1002 * remaining pathname, surely it cannot match.
1003 */
1004 if (prefix > namelen)
1005 return 0;
1006
1007 if (fspathncmp(pattern, name, prefix))
1008 return 0;
1009 pattern += prefix;
1010 patternlen -= prefix;
1011 name += prefix;
1012 namelen -= prefix;
1013
1014 /*
1015 * If the whole pattern did not have a wildcard,
1016 * then our prefix match is all we need; we
1017 * do not need to call fnmatch at all.
1018 */
1019 if (!patternlen && !namelen)
1020 return 1;
1021 }
1022
1023 return fnmatch_icase_mem(pattern, patternlen,
1024 name, namelen,
1025 WM_PATHNAME) == 0;
1026 }
1027
1028 /*
1029 * Scan the given exclude list in reverse to see whether pathname
1030 * should be ignored. The first match (i.e. the last on the list), if
1031 * any, determines the fate. Returns the exclude_list element which
1032 * matched, or NULL for undecided.
1033 */
1034 static struct path_pattern *last_matching_pattern_from_list(const char *pathname,
1035 int pathlen,
1036 const char *basename,
1037 int *dtype,
1038 struct pattern_list *pl,
1039 struct index_state *istate)
1040 {
1041 struct path_pattern *res = NULL; /* undecided */
1042 int i;
1043
1044 if (!pl->nr)
1045 return NULL; /* undefined */
1046
1047 for (i = pl->nr - 1; 0 <= i; i--) {
1048 struct path_pattern *pattern = pl->patterns[i];
1049 const char *exclude = pattern->pattern;
1050 int prefix = pattern->nowildcardlen;
1051
1052 if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) {
1053 if (*dtype == DT_UNKNOWN)
1054 *dtype = get_dtype(NULL, istate, pathname, pathlen);
1055 if (*dtype != DT_DIR)
1056 continue;
1057 }
1058
1059 if (pattern->flags & PATTERN_FLAG_NODIR) {
1060 if (match_basename(basename,
1061 pathlen - (basename - pathname),
1062 exclude, prefix, pattern->patternlen,
1063 pattern->flags)) {
1064 res = pattern;
1065 break;
1066 }
1067 continue;
1068 }
1069
1070 assert(pattern->baselen == 0 ||
1071 pattern->base[pattern->baselen - 1] == '/');
1072 if (match_pathname(pathname, pathlen,
1073 pattern->base,
1074 pattern->baselen ? pattern->baselen - 1 : 0,
1075 exclude, prefix, pattern->patternlen,
1076 pattern->flags)) {
1077 res = pattern;
1078 break;
1079 }
1080 }
1081 return res;
1082 }
1083
1084 /*
1085 * Scan the list of patterns to determine if the ordered list
1086 * of patterns matches on 'pathname'.
1087 *
1088 * Return 1 for a match, 0 for not matched and -1 for undecided.
1089 */
1090 enum pattern_match_result path_matches_pattern_list(
1091 const char *pathname, int pathlen,
1092 const char *basename, int *dtype,
1093 struct pattern_list *pl,
1094 struct index_state *istate)
1095 {
1096 struct path_pattern *pattern;
1097 pattern = last_matching_pattern_from_list(pathname, pathlen, basename,
1098 dtype, pl, istate);
1099 if (pattern) {
1100 if (pattern->flags & PATTERN_FLAG_NEGATIVE)
1101 return NOT_MATCHED;
1102 else
1103 return MATCHED;
1104 }
1105
1106 return UNDECIDED;
1107 }
1108
1109 static struct path_pattern *last_matching_pattern_from_lists(
1110 struct dir_struct *dir, struct index_state *istate,
1111 const char *pathname, int pathlen,
1112 const char *basename, int *dtype_p)
1113 {
1114 int i, j;
1115 struct exclude_list_group *group;
1116 struct path_pattern *pattern;
1117 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1118 group = &dir->exclude_list_group[i];
1119 for (j = group->nr - 1; j >= 0; j--) {
1120 pattern = last_matching_pattern_from_list(
1121 pathname, pathlen, basename, dtype_p,
1122 &group->pl[j], istate);
1123 if (pattern)
1124 return pattern;
1125 }
1126 }
1127 return NULL;
1128 }
1129
1130 /*
1131 * Loads the per-directory exclude list for the substring of base
1132 * which has a char length of baselen.
1133 */
1134 static void prep_exclude(struct dir_struct *dir,
1135 struct index_state *istate,
1136 const char *base, int baselen)
1137 {
1138 struct exclude_list_group *group;
1139 struct pattern_list *pl;
1140 struct exclude_stack *stk = NULL;
1141 struct untracked_cache_dir *untracked;
1142 int current;
1143
1144 group = &dir->exclude_list_group[EXC_DIRS];
1145
1146 /*
1147 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1148 * which originate from directories not in the prefix of the
1149 * path being checked.
1150 */
1151 while ((stk = dir->exclude_stack) != NULL) {
1152 if (stk->baselen <= baselen &&
1153 !strncmp(dir->basebuf.buf, base, stk->baselen))
1154 break;
1155 pl = &group->pl[dir->exclude_stack->exclude_ix];
1156 dir->exclude_stack = stk->prev;
1157 dir->pattern = NULL;
1158 free((char *)pl->src); /* see strbuf_detach() below */
1159 clear_pattern_list(pl);
1160 free(stk);
1161 group->nr--;
1162 }
1163
1164 /* Skip traversing into sub directories if the parent is excluded */
1165 if (dir->pattern)
1166 return;
1167
1168 /*
1169 * Lazy initialization. All call sites currently just
1170 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1171 * them seems lots of work for little benefit.
1172 */
1173 if (!dir->basebuf.buf)
1174 strbuf_init(&dir->basebuf, PATH_MAX);
1175
1176 /* Read from the parent directories and push them down. */
1177 current = stk ? stk->baselen : -1;
1178 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1179 if (dir->untracked)
1180 untracked = stk ? stk->ucd : dir->untracked->root;
1181 else
1182 untracked = NULL;
1183
1184 while (current < baselen) {
1185 const char *cp;
1186 struct oid_stat oid_stat;
1187
1188 stk = xcalloc(1, sizeof(*stk));
1189 if (current < 0) {
1190 cp = base;
1191 current = 0;
1192 } else {
1193 cp = strchr(base + current + 1, '/');
1194 if (!cp)
1195 die("oops in prep_exclude");
1196 cp++;
1197 untracked =
1198 lookup_untracked(dir->untracked, untracked,
1199 base + current,
1200 cp - base - current);
1201 }
1202 stk->prev = dir->exclude_stack;
1203 stk->baselen = cp - base;
1204 stk->exclude_ix = group->nr;
1205 stk->ucd = untracked;
1206 pl = add_pattern_list(dir, EXC_DIRS, NULL);
1207 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1208 assert(stk->baselen == dir->basebuf.len);
1209
1210 /* Abort if the directory is excluded */
1211 if (stk->baselen) {
1212 int dt = DT_DIR;
1213 dir->basebuf.buf[stk->baselen - 1] = 0;
1214 dir->pattern = last_matching_pattern_from_lists(dir,
1215 istate,
1216 dir->basebuf.buf, stk->baselen - 1,
1217 dir->basebuf.buf + current, &dt);
1218 dir->basebuf.buf[stk->baselen - 1] = '/';
1219 if (dir->pattern &&
1220 dir->pattern->flags & PATTERN_FLAG_NEGATIVE)
1221 dir->pattern = NULL;
1222 if (dir->pattern) {
1223 dir->exclude_stack = stk;
1224 return;
1225 }
1226 }
1227
1228 /* Try to read per-directory file */
1229 oidclr(&oid_stat.oid);
1230 oid_stat.valid = 0;
1231 if (dir->exclude_per_dir &&
1232 /*
1233 * If we know that no files have been added in
1234 * this directory (i.e. valid_cached_dir() has
1235 * been executed and set untracked->valid) ..
1236 */
1237 (!untracked || !untracked->valid ||
1238 /*
1239 * .. and .gitignore does not exist before
1240 * (i.e. null exclude_oid). Then we can skip
1241 * loading .gitignore, which would result in
1242 * ENOENT anyway.
1243 */
1244 !is_null_oid(&untracked->exclude_oid))) {
1245 /*
1246 * dir->basebuf gets reused by the traversal, but we
1247 * need fname to remain unchanged to ensure the src
1248 * member of each struct path_pattern correctly
1249 * back-references its source file. Other invocations
1250 * of add_pattern_list provide stable strings, so we
1251 * strbuf_detach() and free() here in the caller.
1252 */
1253 struct strbuf sb = STRBUF_INIT;
1254 strbuf_addbuf(&sb, &dir->basebuf);
1255 strbuf_addstr(&sb, dir->exclude_per_dir);
1256 pl->src = strbuf_detach(&sb, NULL);
1257 add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1258 untracked ? &oid_stat : NULL);
1259 }
1260 /*
1261 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1262 * will first be called in valid_cached_dir() then maybe many
1263 * times more in last_matching_pattern(). When the cache is
1264 * used, last_matching_pattern() will not be called and
1265 * reading .gitignore content will be a waste.
1266 *
1267 * So when it's called by valid_cached_dir() and we can get
1268 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1269 * modified on work tree), we could delay reading the
1270 * .gitignore content until we absolutely need it in
1271 * last_matching_pattern(). Be careful about ignore rule
1272 * order, though, if you do that.
1273 */
1274 if (untracked &&
1275 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1276 invalidate_gitignore(dir->untracked, untracked);
1277 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1278 }
1279 dir->exclude_stack = stk;
1280 current = stk->baselen;
1281 }
1282 strbuf_setlen(&dir->basebuf, baselen);
1283 }
1284
1285 /*
1286 * Loads the exclude lists for the directory containing pathname, then
1287 * scans all exclude lists to determine whether pathname is excluded.
1288 * Returns the exclude_list element which matched, or NULL for
1289 * undecided.
1290 */
1291 struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1292 struct index_state *istate,
1293 const char *pathname,
1294 int *dtype_p)
1295 {
1296 int pathlen = strlen(pathname);
1297 const char *basename = strrchr(pathname, '/');
1298 basename = (basename) ? basename+1 : pathname;
1299
1300 prep_exclude(dir, istate, pathname, basename-pathname);
1301
1302 if (dir->pattern)
1303 return dir->pattern;
1304
1305 return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1306 basename, dtype_p);
1307 }
1308
1309 /*
1310 * Loads the exclude lists for the directory containing pathname, then
1311 * scans all exclude lists to determine whether pathname is excluded.
1312 * Returns 1 if true, otherwise 0.
1313 */
1314 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1315 const char *pathname, int *dtype_p)
1316 {
1317 struct path_pattern *pattern =
1318 last_matching_pattern(dir, istate, pathname, dtype_p);
1319 if (pattern)
1320 return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1321 return 0;
1322 }
1323
1324 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1325 {
1326 struct dir_entry *ent;
1327
1328 FLEX_ALLOC_MEM(ent, name, pathname, len);
1329 ent->len = len;
1330 return ent;
1331 }
1332
1333 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1334 struct index_state *istate,
1335 const char *pathname, int len)
1336 {
1337 if (index_file_exists(istate, pathname, len, ignore_case))
1338 return NULL;
1339
1340 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1341 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1342 }
1343
1344 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1345 struct index_state *istate,
1346 const char *pathname, int len)
1347 {
1348 if (!index_name_is_other(istate, pathname, len))
1349 return NULL;
1350
1351 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1352 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1353 }
1354
1355 enum exist_status {
1356 index_nonexistent = 0,
1357 index_directory,
1358 index_gitdir
1359 };
1360
1361 /*
1362 * Do not use the alphabetically sorted index to look up
1363 * the directory name; instead, use the case insensitive
1364 * directory hash.
1365 */
1366 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1367 const char *dirname, int len)
1368 {
1369 struct cache_entry *ce;
1370
1371 if (index_dir_exists(istate, dirname, len))
1372 return index_directory;
1373
1374 ce = index_file_exists(istate, dirname, len, ignore_case);
1375 if (ce && S_ISGITLINK(ce->ce_mode))
1376 return index_gitdir;
1377
1378 return index_nonexistent;
1379 }
1380
1381 /*
1382 * The index sorts alphabetically by entry name, which
1383 * means that a gitlink sorts as '\0' at the end, while
1384 * a directory (which is defined not as an entry, but as
1385 * the files it contains) will sort with the '/' at the
1386 * end.
1387 */
1388 static enum exist_status directory_exists_in_index(struct index_state *istate,
1389 const char *dirname, int len)
1390 {
1391 int pos;
1392
1393 if (ignore_case)
1394 return directory_exists_in_index_icase(istate, dirname, len);
1395
1396 pos = index_name_pos(istate, dirname, len);
1397 if (pos < 0)
1398 pos = -pos-1;
1399 while (pos < istate->cache_nr) {
1400 const struct cache_entry *ce = istate->cache[pos++];
1401 unsigned char endchar;
1402
1403 if (strncmp(ce->name, dirname, len))
1404 break;
1405 endchar = ce->name[len];
1406 if (endchar > '/')
1407 break;
1408 if (endchar == '/')
1409 return index_directory;
1410 if (!endchar && S_ISGITLINK(ce->ce_mode))
1411 return index_gitdir;
1412 }
1413 return index_nonexistent;
1414 }
1415
1416 /*
1417 * When we find a directory when traversing the filesystem, we
1418 * have three distinct cases:
1419 *
1420 * - ignore it
1421 * - see it as a directory
1422 * - recurse into it
1423 *
1424 * and which one we choose depends on a combination of existing
1425 * git index contents and the flags passed into the directory
1426 * traversal routine.
1427 *
1428 * Case 1: If we *already* have entries in the index under that
1429 * directory name, we always recurse into the directory to see
1430 * all the files.
1431 *
1432 * Case 2: If we *already* have that directory name as a gitlink,
1433 * we always continue to see it as a gitlink, regardless of whether
1434 * there is an actual git directory there or not (it might not
1435 * be checked out as a subproject!)
1436 *
1437 * Case 3: if we didn't have it in the index previously, we
1438 * have a few sub-cases:
1439 *
1440 * (a) if "show_other_directories" is true, we show it as
1441 * just a directory, unless "hide_empty_directories" is
1442 * also true, in which case we need to check if it contains any
1443 * untracked and / or ignored files.
1444 * (b) if it looks like a git directory, and we don't have
1445 * 'no_gitlinks' set we treat it as a gitlink, and show it
1446 * as a directory.
1447 * (c) otherwise, we recurse into it.
1448 */
1449 static enum path_treatment treat_directory(struct dir_struct *dir,
1450 struct index_state *istate,
1451 struct untracked_cache_dir *untracked,
1452 const char *dirname, int len, int baselen, int exclude,
1453 const struct pathspec *pathspec)
1454 {
1455 /* The "len-1" is to strip the final '/' */
1456 switch (directory_exists_in_index(istate, dirname, len-1)) {
1457 case index_directory:
1458 return path_recurse;
1459
1460 case index_gitdir:
1461 return path_none;
1462
1463 case index_nonexistent:
1464 if (dir->flags & DIR_SKIP_NESTED_GIT) {
1465 int nested_repo;
1466 struct strbuf sb = STRBUF_INIT;
1467 strbuf_addstr(&sb, dirname);
1468 nested_repo = is_nonbare_repository_dir(&sb);
1469 strbuf_release(&sb);
1470 if (nested_repo)
1471 return path_none;
1472 }
1473
1474 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1475 break;
1476 if (exclude &&
1477 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1478 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1479
1480 /*
1481 * This is an excluded directory and we are
1482 * showing ignored paths that match an exclude
1483 * pattern. (e.g. show directory as ignored
1484 * only if it matches an exclude pattern).
1485 * This path will either be 'path_excluded`
1486 * (if we are showing empty directories or if
1487 * the directory is not empty), or will be
1488 * 'path_none' (empty directory, and we are
1489 * not showing empty directories).
1490 */
1491 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1492 return path_excluded;
1493
1494 if (read_directory_recursive(dir, istate, dirname, len,
1495 untracked, 1, 1, pathspec) == path_excluded)
1496 return path_excluded;
1497
1498 return path_none;
1499 }
1500 if (!(dir->flags & DIR_NO_GITLINKS)) {
1501 struct strbuf sb = STRBUF_INIT;
1502 strbuf_addstr(&sb, dirname);
1503 if (is_nonbare_repository_dir(&sb))
1504 return exclude ? path_excluded : path_untracked;
1505 strbuf_release(&sb);
1506 }
1507 return path_recurse;
1508 }
1509
1510 /* This is the "show_other_directories" case */
1511
1512 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1513 return exclude ? path_excluded : path_untracked;
1514
1515 untracked = lookup_untracked(dir->untracked, untracked,
1516 dirname + baselen, len - baselen);
1517
1518 /*
1519 * If this is an excluded directory, then we only need to check if
1520 * the directory contains any files.
1521 */
1522 return read_directory_recursive(dir, istate, dirname, len,
1523 untracked, 1, exclude, pathspec);
1524 }
1525
1526 /*
1527 * This is an inexact early pruning of any recursive directory
1528 * reading - if the path cannot possibly be in the pathspec,
1529 * return true, and we'll skip it early.
1530 */
1531 static int simplify_away(const char *path, int pathlen,
1532 const struct pathspec *pathspec)
1533 {
1534 int i;
1535
1536 if (!pathspec || !pathspec->nr)
1537 return 0;
1538
1539 GUARD_PATHSPEC(pathspec,
1540 PATHSPEC_FROMTOP |
1541 PATHSPEC_MAXDEPTH |
1542 PATHSPEC_LITERAL |
1543 PATHSPEC_GLOB |
1544 PATHSPEC_ICASE |
1545 PATHSPEC_EXCLUDE |
1546 PATHSPEC_ATTR);
1547
1548 for (i = 0; i < pathspec->nr; i++) {
1549 const struct pathspec_item *item = &pathspec->items[i];
1550 int len = item->nowildcard_len;
1551
1552 if (len > pathlen)
1553 len = pathlen;
1554 if (!ps_strncmp(item, item->match, path, len))
1555 return 0;
1556 }
1557
1558 return 1;
1559 }
1560
1561 /*
1562 * This function tells us whether an excluded path matches a
1563 * list of "interesting" pathspecs. That is, whether a path matched
1564 * by any of the pathspecs could possibly be ignored by excluding
1565 * the specified path. This can happen if:
1566 *
1567 * 1. the path is mentioned explicitly in the pathspec
1568 *
1569 * 2. the path is a directory prefix of some element in the
1570 * pathspec
1571 */
1572 static int exclude_matches_pathspec(const char *path, int pathlen,
1573 const struct pathspec *pathspec)
1574 {
1575 int i;
1576
1577 if (!pathspec || !pathspec->nr)
1578 return 0;
1579
1580 GUARD_PATHSPEC(pathspec,
1581 PATHSPEC_FROMTOP |
1582 PATHSPEC_MAXDEPTH |
1583 PATHSPEC_LITERAL |
1584 PATHSPEC_GLOB |
1585 PATHSPEC_ICASE |
1586 PATHSPEC_EXCLUDE);
1587
1588 for (i = 0; i < pathspec->nr; i++) {
1589 const struct pathspec_item *item = &pathspec->items[i];
1590 int len = item->nowildcard_len;
1591
1592 if (len == pathlen &&
1593 !ps_strncmp(item, item->match, path, pathlen))
1594 return 1;
1595 if (len > pathlen &&
1596 item->match[pathlen] == '/' &&
1597 !ps_strncmp(item, item->match, path, pathlen))
1598 return 1;
1599 }
1600 return 0;
1601 }
1602
1603 static int get_index_dtype(struct index_state *istate,
1604 const char *path, int len)
1605 {
1606 int pos;
1607 const struct cache_entry *ce;
1608
1609 ce = index_file_exists(istate, path, len, 0);
1610 if (ce) {
1611 if (!ce_uptodate(ce))
1612 return DT_UNKNOWN;
1613 if (S_ISGITLINK(ce->ce_mode))
1614 return DT_DIR;
1615 /*
1616 * Nobody actually cares about the
1617 * difference between DT_LNK and DT_REG
1618 */
1619 return DT_REG;
1620 }
1621
1622 /* Try to look it up as a directory */
1623 pos = index_name_pos(istate, path, len);
1624 if (pos >= 0)
1625 return DT_UNKNOWN;
1626 pos = -pos-1;
1627 while (pos < istate->cache_nr) {
1628 ce = istate->cache[pos++];
1629 if (strncmp(ce->name, path, len))
1630 break;
1631 if (ce->name[len] > '/')
1632 break;
1633 if (ce->name[len] < '/')
1634 continue;
1635 if (!ce_uptodate(ce))
1636 break; /* continue? */
1637 return DT_DIR;
1638 }
1639 return DT_UNKNOWN;
1640 }
1641
1642 static int get_dtype(struct dirent *de, struct index_state *istate,
1643 const char *path, int len)
1644 {
1645 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1646 struct stat st;
1647
1648 if (dtype != DT_UNKNOWN)
1649 return dtype;
1650 dtype = get_index_dtype(istate, path, len);
1651 if (dtype != DT_UNKNOWN)
1652 return dtype;
1653 if (lstat(path, &st))
1654 return dtype;
1655 if (S_ISREG(st.st_mode))
1656 return DT_REG;
1657 if (S_ISDIR(st.st_mode))
1658 return DT_DIR;
1659 if (S_ISLNK(st.st_mode))
1660 return DT_LNK;
1661 return dtype;
1662 }
1663
1664 static enum path_treatment treat_one_path(struct dir_struct *dir,
1665 struct untracked_cache_dir *untracked,
1666 struct index_state *istate,
1667 struct strbuf *path,
1668 int baselen,
1669 const struct pathspec *pathspec,
1670 int dtype, struct dirent *de)
1671 {
1672 int exclude;
1673 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1674 enum path_treatment path_treatment;
1675
1676 if (dtype == DT_UNKNOWN)
1677 dtype = get_dtype(de, istate, path->buf, path->len);
1678
1679 /* Always exclude indexed files */
1680 if (dtype != DT_DIR && has_path_in_index)
1681 return path_none;
1682
1683 /*
1684 * When we are looking at a directory P in the working tree,
1685 * there are three cases:
1686 *
1687 * (1) P exists in the index. Everything inside the directory P in
1688 * the working tree needs to go when P is checked out from the
1689 * index.
1690 *
1691 * (2) P does not exist in the index, but there is P/Q in the index.
1692 * We know P will stay a directory when we check out the contents
1693 * of the index, but we do not know yet if there is a directory
1694 * P/Q in the working tree to be killed, so we need to recurse.
1695 *
1696 * (3) P does not exist in the index, and there is no P/Q in the index
1697 * to require P to be a directory, either. Only in this case, we
1698 * know that everything inside P will not be killed without
1699 * recursing.
1700 */
1701 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1702 (dtype == DT_DIR) &&
1703 !has_path_in_index &&
1704 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1705 return path_none;
1706
1707 exclude = is_excluded(dir, istate, path->buf, &dtype);
1708
1709 /*
1710 * Excluded? If we don't explicitly want to show
1711 * ignored files, ignore it
1712 */
1713 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1714 return path_excluded;
1715
1716 switch (dtype) {
1717 default:
1718 return path_none;
1719 case DT_DIR:
1720 strbuf_addch(path, '/');
1721 path_treatment = treat_directory(dir, istate, untracked,
1722 path->buf, path->len,
1723 baselen, exclude, pathspec);
1724 /*
1725 * If 1) we only want to return directories that
1726 * match an exclude pattern and 2) this directory does
1727 * not match an exclude pattern but all of its
1728 * contents are excluded, then indicate that we should
1729 * recurse into this directory (instead of marking the
1730 * directory itself as an ignored path).
1731 */
1732 if (!exclude &&
1733 path_treatment == path_excluded &&
1734 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1735 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1736 return path_recurse;
1737 return path_treatment;
1738 case DT_REG:
1739 case DT_LNK:
1740 return exclude ? path_excluded : path_untracked;
1741 }
1742 }
1743
1744 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1745 struct untracked_cache_dir *untracked,
1746 struct cached_dir *cdir,
1747 struct index_state *istate,
1748 struct strbuf *path,
1749 int baselen,
1750 const struct pathspec *pathspec)
1751 {
1752 strbuf_setlen(path, baselen);
1753 if (!cdir->ucd) {
1754 strbuf_addstr(path, cdir->file);
1755 return path_untracked;
1756 }
1757 strbuf_addstr(path, cdir->ucd->name);
1758 /* treat_one_path() does this before it calls treat_directory() */
1759 strbuf_complete(path, '/');
1760 if (cdir->ucd->check_only)
1761 /*
1762 * check_only is set as a result of treat_directory() getting
1763 * to its bottom. Verify again the same set of directories
1764 * with check_only set.
1765 */
1766 return read_directory_recursive(dir, istate, path->buf, path->len,
1767 cdir->ucd, 1, 0, pathspec);
1768 /*
1769 * We get path_recurse in the first run when
1770 * directory_exists_in_index() returns index_nonexistent. We
1771 * are sure that new changes in the index does not impact the
1772 * outcome. Return now.
1773 */
1774 return path_recurse;
1775 }
1776
1777 static enum path_treatment treat_path(struct dir_struct *dir,
1778 struct untracked_cache_dir *untracked,
1779 struct cached_dir *cdir,
1780 struct index_state *istate,
1781 struct strbuf *path,
1782 int baselen,
1783 const struct pathspec *pathspec)
1784 {
1785 int dtype;
1786 struct dirent *de = cdir->de;
1787
1788 if (!de)
1789 return treat_path_fast(dir, untracked, cdir, istate, path,
1790 baselen, pathspec);
1791 if (is_dot_or_dotdot(de->d_name) || !fspathcmp(de->d_name, ".git"))
1792 return path_none;
1793 strbuf_setlen(path, baselen);
1794 strbuf_addstr(path, de->d_name);
1795 if (simplify_away(path->buf, path->len, pathspec))
1796 return path_none;
1797
1798 dtype = DTYPE(de);
1799 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1800 }
1801
1802 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1803 {
1804 if (!dir)
1805 return;
1806 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1807 dir->untracked_alloc);
1808 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1809 }
1810
1811 static int valid_cached_dir(struct dir_struct *dir,
1812 struct untracked_cache_dir *untracked,
1813 struct index_state *istate,
1814 struct strbuf *path,
1815 int check_only)
1816 {
1817 struct stat st;
1818
1819 if (!untracked)
1820 return 0;
1821
1822 /*
1823 * With fsmonitor, we can trust the untracked cache's valid field.
1824 */
1825 refresh_fsmonitor(istate);
1826 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1827 if (lstat(path->len ? path->buf : ".", &st)) {
1828 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1829 return 0;
1830 }
1831 if (!untracked->valid ||
1832 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1833 fill_stat_data(&untracked->stat_data, &st);
1834 return 0;
1835 }
1836 }
1837
1838 if (untracked->check_only != !!check_only)
1839 return 0;
1840
1841 /*
1842 * prep_exclude will be called eventually on this directory,
1843 * but it's called much later in last_matching_pattern(). We
1844 * need it now to determine the validity of the cache for this
1845 * path. The next calls will be nearly no-op, the way
1846 * prep_exclude() is designed.
1847 */
1848 if (path->len && path->buf[path->len - 1] != '/') {
1849 strbuf_addch(path, '/');
1850 prep_exclude(dir, istate, path->buf, path->len);
1851 strbuf_setlen(path, path->len - 1);
1852 } else
1853 prep_exclude(dir, istate, path->buf, path->len);
1854
1855 /* hopefully prep_exclude() haven't invalidated this entry... */
1856 return untracked->valid;
1857 }
1858
1859 static int open_cached_dir(struct cached_dir *cdir,
1860 struct dir_struct *dir,
1861 struct untracked_cache_dir *untracked,
1862 struct index_state *istate,
1863 struct strbuf *path,
1864 int check_only)
1865 {
1866 const char *c_path;
1867
1868 memset(cdir, 0, sizeof(*cdir));
1869 cdir->untracked = untracked;
1870 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1871 return 0;
1872 c_path = path->len ? path->buf : ".";
1873 cdir->fdir = opendir(c_path);
1874 if (!cdir->fdir)
1875 warning_errno(_("could not open directory '%s'"), c_path);
1876 if (dir->untracked) {
1877 invalidate_directory(dir->untracked, untracked);
1878 dir->untracked->dir_opened++;
1879 }
1880 if (!cdir->fdir)
1881 return -1;
1882 return 0;
1883 }
1884
1885 static int read_cached_dir(struct cached_dir *cdir)
1886 {
1887 if (cdir->fdir) {
1888 cdir->de = readdir(cdir->fdir);
1889 if (!cdir->de)
1890 return -1;
1891 return 0;
1892 }
1893 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1894 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1895 if (!d->recurse) {
1896 cdir->nr_dirs++;
1897 continue;
1898 }
1899 cdir->ucd = d;
1900 cdir->nr_dirs++;
1901 return 0;
1902 }
1903 cdir->ucd = NULL;
1904 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1905 struct untracked_cache_dir *d = cdir->untracked;
1906 cdir->file = d->untracked[cdir->nr_files++];
1907 return 0;
1908 }
1909 return -1;
1910 }
1911
1912 static void close_cached_dir(struct cached_dir *cdir)
1913 {
1914 if (cdir->fdir)
1915 closedir(cdir->fdir);
1916 /*
1917 * We have gone through this directory and found no untracked
1918 * entries. Mark it valid.
1919 */
1920 if (cdir->untracked) {
1921 cdir->untracked->valid = 1;
1922 cdir->untracked->recurse = 1;
1923 }
1924 }
1925
1926 /*
1927 * Read a directory tree. We currently ignore anything but
1928 * directories, regular files and symlinks. That's because git
1929 * doesn't handle them at all yet. Maybe that will change some
1930 * day.
1931 *
1932 * Also, we ignore the name ".git" (even if it is not a directory).
1933 * That likely will not change.
1934 *
1935 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1936 * to signal that a file was found. This is the least significant value that
1937 * indicates that a file was encountered that does not depend on the order of
1938 * whether an untracked or exluded path was encountered first.
1939 *
1940 * Returns the most significant path_treatment value encountered in the scan.
1941 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1942 * significant path_treatment value that will be returned.
1943 */
1944
1945 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1946 struct index_state *istate, const char *base, int baselen,
1947 struct untracked_cache_dir *untracked, int check_only,
1948 int stop_at_first_file, const struct pathspec *pathspec)
1949 {
1950 struct cached_dir cdir;
1951 enum path_treatment state, subdir_state, dir_state = path_none;
1952 struct strbuf path = STRBUF_INIT;
1953
1954 strbuf_add(&path, base, baselen);
1955
1956 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1957 goto out;
1958
1959 if (untracked)
1960 untracked->check_only = !!check_only;
1961
1962 while (!read_cached_dir(&cdir)) {
1963 /* check how the file or directory should be treated */
1964 state = treat_path(dir, untracked, &cdir, istate, &path,
1965 baselen, pathspec);
1966
1967 if (state > dir_state)
1968 dir_state = state;
1969
1970 /* recurse into subdir if instructed by treat_path */
1971 if ((state == path_recurse) ||
1972 ((state == path_untracked) &&
1973 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR) &&
1974 ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1975 (pathspec &&
1976 do_match_pathspec(istate, pathspec, path.buf, path.len,
1977 baselen, NULL, DO_MATCH_LEADING_PATHSPEC) == MATCHED_RECURSIVELY_LEADING_PATHSPEC)))) {
1978 struct untracked_cache_dir *ud;
1979 ud = lookup_untracked(dir->untracked, untracked,
1980 path.buf + baselen,
1981 path.len - baselen);
1982 subdir_state =
1983 read_directory_recursive(dir, istate, path.buf,
1984 path.len, ud,
1985 check_only, stop_at_first_file, pathspec);
1986 if (subdir_state > dir_state)
1987 dir_state = subdir_state;
1988
1989 if (pathspec &&
1990 !match_pathspec(istate, pathspec, path.buf, path.len,
1991 0 /* prefix */, NULL,
1992 0 /* do NOT special case dirs */))
1993 state = path_none;
1994 }
1995
1996 if (check_only) {
1997 if (stop_at_first_file) {
1998 /*
1999 * If stopping at first file, then
2000 * signal that a file was found by
2001 * returning `path_excluded`. This is
2002 * to return a consistent value
2003 * regardless of whether an ignored or
2004 * excluded file happened to be
2005 * encountered 1st.
2006 *
2007 * In current usage, the
2008 * `stop_at_first_file` is passed when
2009 * an ancestor directory has matched
2010 * an exclude pattern, so any found
2011 * files will be excluded.
2012 */
2013 if (dir_state >= path_excluded) {
2014 dir_state = path_excluded;
2015 break;
2016 }
2017 }
2018
2019 /* abort early if maximum state has been reached */
2020 if (dir_state == path_untracked) {
2021 if (cdir.fdir)
2022 add_untracked(untracked, path.buf + baselen);
2023 break;
2024 }
2025 /* skip the dir_add_* part */
2026 continue;
2027 }
2028
2029 /* add the path to the appropriate result list */
2030 switch (state) {
2031 case path_excluded:
2032 if (dir->flags & DIR_SHOW_IGNORED)
2033 dir_add_name(dir, istate, path.buf, path.len);
2034 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2035 ((dir->flags & DIR_COLLECT_IGNORED) &&
2036 exclude_matches_pathspec(path.buf, path.len,
2037 pathspec)))
2038 dir_add_ignored(dir, istate, path.buf, path.len);
2039 break;
2040
2041 case path_untracked:
2042 if (dir->flags & DIR_SHOW_IGNORED)
2043 break;
2044 dir_add_name(dir, istate, path.buf, path.len);
2045 if (cdir.fdir)
2046 add_untracked(untracked, path.buf + baselen);
2047 break;
2048
2049 default:
2050 break;
2051 }
2052 }
2053 close_cached_dir(&cdir);
2054 out:
2055 strbuf_release(&path);
2056
2057 return dir_state;
2058 }
2059
2060 int cmp_dir_entry(const void *p1, const void *p2)
2061 {
2062 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2063 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2064
2065 return name_compare(e1->name, e1->len, e2->name, e2->len);
2066 }
2067
2068 /* check if *out lexically strictly contains *in */
2069 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2070 {
2071 return (out->len < in->len) &&
2072 (out->name[out->len - 1] == '/') &&
2073 !memcmp(out->name, in->name, out->len);
2074 }
2075
2076 static int treat_leading_path(struct dir_struct *dir,
2077 struct index_state *istate,
2078 const char *path, int len,
2079 const struct pathspec *pathspec)
2080 {
2081 struct strbuf sb = STRBUF_INIT;
2082 int baselen, rc = 0;
2083 const char *cp;
2084 int old_flags = dir->flags;
2085
2086 while (len && path[len - 1] == '/')
2087 len--;
2088 if (!len)
2089 return 1;
2090 baselen = 0;
2091 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
2092 while (1) {
2093 cp = path + baselen + !!baselen;
2094 cp = memchr(cp, '/', path + len - cp);
2095 if (!cp)
2096 baselen = len;
2097 else
2098 baselen = cp - path;
2099 strbuf_setlen(&sb, 0);
2100 strbuf_add(&sb, path, baselen);
2101 if (!is_directory(sb.buf))
2102 break;
2103 if (simplify_away(sb.buf, sb.len, pathspec))
2104 break;
2105 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
2106 DT_DIR, NULL) == path_none)
2107 break; /* do not recurse into it */
2108 if (len <= baselen) {
2109 rc = 1;
2110 break; /* finished checking */
2111 }
2112 }
2113 strbuf_release(&sb);
2114 dir->flags = old_flags;
2115 return rc;
2116 }
2117
2118 static const char *get_ident_string(void)
2119 {
2120 static struct strbuf sb = STRBUF_INIT;
2121 struct utsname uts;
2122
2123 if (sb.len)
2124 return sb.buf;
2125 if (uname(&uts) < 0)
2126 die_errno(_("failed to get kernel name and information"));
2127 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2128 uts.sysname);
2129 return sb.buf;
2130 }
2131
2132 static int ident_in_untracked(const struct untracked_cache *uc)
2133 {
2134 /*
2135 * Previous git versions may have saved many NUL separated
2136 * strings in the "ident" field, but it is insane to manage
2137 * many locations, so just take care of the first one.
2138 */
2139
2140 return !strcmp(uc->ident.buf, get_ident_string());
2141 }
2142
2143 static void set_untracked_ident(struct untracked_cache *uc)
2144 {
2145 strbuf_reset(&uc->ident);
2146 strbuf_addstr(&uc->ident, get_ident_string());
2147
2148 /*
2149 * This strbuf used to contain a list of NUL separated
2150 * strings, so save NUL too for backward compatibility.
2151 */
2152 strbuf_addch(&uc->ident, 0);
2153 }
2154
2155 static void new_untracked_cache(struct index_state *istate)
2156 {
2157 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2158 strbuf_init(&uc->ident, 100);
2159 uc->exclude_per_dir = ".gitignore";
2160 /* should be the same flags used by git-status */
2161 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2162 set_untracked_ident(uc);
2163 istate->untracked = uc;
2164 istate->cache_changed |= UNTRACKED_CHANGED;
2165 }
2166
2167 void add_untracked_cache(struct index_state *istate)
2168 {
2169 if (!istate->untracked) {
2170 new_untracked_cache(istate);
2171 } else {
2172 if (!ident_in_untracked(istate->untracked)) {
2173 free_untracked_cache(istate->untracked);
2174 new_untracked_cache(istate);
2175 }
2176 }
2177 }
2178
2179 void remove_untracked_cache(struct index_state *istate)
2180 {
2181 if (istate->untracked) {
2182 free_untracked_cache(istate->untracked);
2183 istate->untracked = NULL;
2184 istate->cache_changed |= UNTRACKED_CHANGED;
2185 }
2186 }
2187
2188 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2189 int base_len,
2190 const struct pathspec *pathspec)
2191 {
2192 struct untracked_cache_dir *root;
2193 static int untracked_cache_disabled = -1;
2194
2195 if (!dir->untracked)
2196 return NULL;
2197 if (untracked_cache_disabled < 0)
2198 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2199 if (untracked_cache_disabled)
2200 return NULL;
2201
2202 /*
2203 * We only support $GIT_DIR/info/exclude and core.excludesfile
2204 * as the global ignore rule files. Any other additions
2205 * (e.g. from command line) invalidate the cache. This
2206 * condition also catches running setup_standard_excludes()
2207 * before setting dir->untracked!
2208 */
2209 if (dir->unmanaged_exclude_files)
2210 return NULL;
2211
2212 /*
2213 * Optimize for the main use case only: whole-tree git
2214 * status. More work involved in treat_leading_path() if we
2215 * use cache on just a subset of the worktree. pathspec
2216 * support could make the matter even worse.
2217 */
2218 if (base_len || (pathspec && pathspec->nr))
2219 return NULL;
2220
2221 /* Different set of flags may produce different results */
2222 if (dir->flags != dir->untracked->dir_flags ||
2223 /*
2224 * See treat_directory(), case index_nonexistent. Without
2225 * this flag, we may need to also cache .git file content
2226 * for the resolve_gitlink_ref() call, which we don't.
2227 */
2228 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2229 /* We don't support collecting ignore files */
2230 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2231 DIR_COLLECT_IGNORED)))
2232 return NULL;
2233
2234 /*
2235 * If we use .gitignore in the cache and now you change it to
2236 * .gitexclude, everything will go wrong.
2237 */
2238 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2239 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2240 return NULL;
2241
2242 /*
2243 * EXC_CMDL is not considered in the cache. If people set it,
2244 * skip the cache.
2245 */
2246 if (dir->exclude_list_group[EXC_CMDL].nr)
2247 return NULL;
2248
2249 if (!ident_in_untracked(dir->untracked)) {
2250 warning(_("untracked cache is disabled on this system or location"));
2251 return NULL;
2252 }
2253
2254 if (!dir->untracked->root) {
2255 const int len = sizeof(*dir->untracked->root);
2256 dir->untracked->root = xmalloc(len);
2257 memset(dir->untracked->root, 0, len);
2258 }
2259
2260 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2261 root = dir->untracked->root;
2262 if (!oideq(&dir->ss_info_exclude.oid,
2263 &dir->untracked->ss_info_exclude.oid)) {
2264 invalidate_gitignore(dir->untracked, root);
2265 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2266 }
2267 if (!oideq(&dir->ss_excludes_file.oid,
2268 &dir->untracked->ss_excludes_file.oid)) {
2269 invalidate_gitignore(dir->untracked, root);
2270 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2271 }
2272
2273 /* Make sure this directory is not dropped out at saving phase */
2274 root->recurse = 1;
2275 return root;
2276 }
2277
2278 int read_directory(struct dir_struct *dir, struct index_state *istate,
2279 const char *path, int len, const struct pathspec *pathspec)
2280 {
2281 struct untracked_cache_dir *untracked;
2282
2283 trace_performance_enter();
2284
2285 if (has_symlink_leading_path(path, len)) {
2286 trace_performance_leave("read directory %.*s", len, path);
2287 return dir->nr;
2288 }
2289
2290 untracked = validate_untracked_cache(dir, len, pathspec);
2291 if (!untracked)
2292 /*
2293 * make sure untracked cache code path is disabled,
2294 * e.g. prep_exclude()
2295 */
2296 dir->untracked = NULL;
2297 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2298 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2299 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2300 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2301
2302 /*
2303 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2304 * also pick up untracked contents of untracked dirs; by default
2305 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2306 */
2307 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2308 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2309 int i, j;
2310
2311 /* remove from dir->entries untracked contents of untracked dirs */
2312 for (i = j = 0; j < dir->nr; j++) {
2313 if (i &&
2314 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2315 FREE_AND_NULL(dir->entries[j]);
2316 } else {
2317 dir->entries[i++] = dir->entries[j];
2318 }
2319 }
2320
2321 dir->nr = i;
2322 }
2323
2324 trace_performance_leave("read directory %.*s", len, path);
2325 if (dir->untracked) {
2326 static int force_untracked_cache = -1;
2327 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2328
2329 if (force_untracked_cache < 0)
2330 force_untracked_cache =
2331 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2332 trace_printf_key(&trace_untracked_stats,
2333 "node creation: %u\n"
2334 "gitignore invalidation: %u\n"
2335 "directory invalidation: %u\n"
2336 "opendir: %u\n",
2337 dir->untracked->dir_created,
2338 dir->untracked->gitignore_invalidated,
2339 dir->untracked->dir_invalidated,
2340 dir->untracked->dir_opened);
2341 if (force_untracked_cache &&
2342 dir->untracked == istate->untracked &&
2343 (dir->untracked->dir_opened ||
2344 dir->untracked->gitignore_invalidated ||
2345 dir->untracked->dir_invalidated))
2346 istate->cache_changed |= UNTRACKED_CHANGED;
2347 if (dir->untracked != istate->untracked) {
2348 FREE_AND_NULL(dir->untracked);
2349 }
2350 }
2351 return dir->nr;
2352 }
2353
2354 int file_exists(const char *f)
2355 {
2356 struct stat sb;
2357 return lstat(f, &sb) == 0;
2358 }
2359
2360 int repo_file_exists(struct repository *repo, const char *path)
2361 {
2362 if (repo != the_repository)
2363 BUG("do not know how to check file existence in arbitrary repo");
2364
2365 return file_exists(path);
2366 }
2367
2368 static int cmp_icase(char a, char b)
2369 {
2370 if (a == b)
2371 return 0;
2372 if (ignore_case)
2373 return toupper(a) - toupper(b);
2374 return a - b;
2375 }
2376
2377 /*
2378 * Given two normalized paths (a trailing slash is ok), if subdir is
2379 * outside dir, return -1. Otherwise return the offset in subdir that
2380 * can be used as relative path to dir.
2381 */
2382 int dir_inside_of(const char *subdir, const char *dir)
2383 {
2384 int offset = 0;
2385
2386 assert(dir && subdir && *dir && *subdir);
2387
2388 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2389 dir++;
2390 subdir++;
2391 offset++;
2392 }
2393
2394 /* hel[p]/me vs hel[l]/yeah */
2395 if (*dir && *subdir)
2396 return -1;
2397
2398 if (!*subdir)
2399 return !*dir ? offset : -1; /* same dir */
2400
2401 /* foo/[b]ar vs foo/[] */
2402 if (is_dir_sep(dir[-1]))
2403 return is_dir_sep(subdir[-1]) ? offset : -1;
2404
2405 /* foo[/]bar vs foo[] */
2406 return is_dir_sep(*subdir) ? offset + 1 : -1;
2407 }
2408
2409 int is_inside_dir(const char *dir)
2410 {
2411 char *cwd;
2412 int rc;
2413
2414 if (!dir)
2415 return 0;
2416
2417 cwd = xgetcwd();
2418 rc = (dir_inside_of(cwd, dir) >= 0);
2419 free(cwd);
2420 return rc;
2421 }
2422
2423 int is_empty_dir(const char *path)
2424 {
2425 DIR *dir = opendir(path);
2426 struct dirent *e;
2427 int ret = 1;
2428
2429 if (!dir)
2430 return 0;
2431
2432 while ((e = readdir(dir)) != NULL)
2433 if (!is_dot_or_dotdot(e->d_name)) {
2434 ret = 0;
2435 break;
2436 }
2437
2438 closedir(dir);
2439 return ret;
2440 }
2441
2442 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2443 {
2444 DIR *dir;
2445 struct dirent *e;
2446 int ret = 0, original_len = path->len, len, kept_down = 0;
2447 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2448 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2449 struct object_id submodule_head;
2450
2451 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2452 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2453 /* Do not descend and nuke a nested git work tree. */
2454 if (kept_up)
2455 *kept_up = 1;
2456 return 0;
2457 }
2458
2459 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2460 dir = opendir(path->buf);
2461 if (!dir) {
2462 if (errno == ENOENT)
2463 return keep_toplevel ? -1 : 0;
2464 else if (errno == EACCES && !keep_toplevel)
2465 /*
2466 * An empty dir could be removable even if it
2467 * is unreadable:
2468 */
2469 return rmdir(path->buf);
2470 else
2471 return -1;
2472 }
2473 strbuf_complete(path, '/');
2474
2475 len = path->len;
2476 while ((e = readdir(dir)) != NULL) {
2477 struct stat st;
2478 if (is_dot_or_dotdot(e->d_name))
2479 continue;
2480
2481 strbuf_setlen(path, len);
2482 strbuf_addstr(path, e->d_name);
2483 if (lstat(path->buf, &st)) {
2484 if (errno == ENOENT)
2485 /*
2486 * file disappeared, which is what we
2487 * wanted anyway
2488 */
2489 continue;
2490 /* fall thru */
2491 } else if (S_ISDIR(st.st_mode)) {
2492 if (!remove_dir_recurse(path, flag, &kept_down))
2493 continue; /* happy */
2494 } else if (!only_empty &&
2495 (!unlink(path->buf) || errno == ENOENT)) {
2496 continue; /* happy, too */
2497 }
2498
2499 /* path too long, stat fails, or non-directory still exists */
2500 ret = -1;
2501 break;
2502 }
2503 closedir(dir);
2504
2505 strbuf_setlen(path, original_len);
2506 if (!ret && !keep_toplevel && !kept_down)
2507 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2508 else if (kept_up)
2509 /*
2510 * report the uplevel that it is not an error that we
2511 * did not rmdir() our directory.
2512 */
2513 *kept_up = !ret;
2514 return ret;
2515 }
2516
2517 int remove_dir_recursively(struct strbuf *path, int flag)
2518 {
2519 return remove_dir_recurse(path, flag, NULL);
2520 }
2521
2522 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2523
2524 void setup_standard_excludes(struct dir_struct *dir)
2525 {
2526 dir->exclude_per_dir = ".gitignore";
2527
2528 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2529 if (!excludes_file)
2530 excludes_file = xdg_config_home("ignore");
2531 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2532 add_patterns_from_file_1(dir, excludes_file,
2533 dir->untracked ? &dir->ss_excludes_file : NULL);
2534
2535 /* per repository user preference */
2536 if (startup_info->have_repository) {
2537 const char *path = git_path_info_exclude();
2538 if (!access_or_warn(path, R_OK, 0))
2539 add_patterns_from_file_1(dir, path,
2540 dir->untracked ? &dir->ss_info_exclude : NULL);
2541 }
2542 }
2543
2544 int remove_path(const char *name)
2545 {
2546 char *slash;
2547
2548 if (unlink(name) && !is_missing_file_error(errno))
2549 return -1;
2550
2551 slash = strrchr(name, '/');
2552 if (slash) {
2553 char *dirs = xstrdup(name);
2554 slash = dirs + (slash - name);
2555 do {
2556 *slash = '\0';
2557 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2558 free(dirs);
2559 }
2560 return 0;
2561 }
2562
2563 /*
2564 * Frees memory within dir which was allocated for exclude lists and
2565 * the exclude_stack. Does not free dir itself.
2566 */
2567 void clear_directory(struct dir_struct *dir)
2568 {
2569 int i, j;
2570 struct exclude_list_group *group;
2571 struct pattern_list *pl;
2572 struct exclude_stack *stk;
2573
2574 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2575 group = &dir->exclude_list_group[i];
2576 for (j = 0; j < group->nr; j++) {
2577 pl = &group->pl[j];
2578 if (i == EXC_DIRS)
2579 free((char *)pl->src);
2580 clear_pattern_list(pl);
2581 }
2582 free(group->pl);
2583 }
2584
2585 stk = dir->exclude_stack;
2586 while (stk) {
2587 struct exclude_stack *prev = stk->prev;
2588 free(stk);
2589 stk = prev;
2590 }
2591 strbuf_release(&dir->basebuf);
2592 }
2593
2594 struct ondisk_untracked_cache {
2595 struct stat_data info_exclude_stat;
2596 struct stat_data excludes_file_stat;
2597 uint32_t dir_flags;
2598 };
2599
2600 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2601
2602 struct write_data {
2603 int index; /* number of written untracked_cache_dir */
2604 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2605 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2606 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2607 struct strbuf out;
2608 struct strbuf sb_stat;
2609 struct strbuf sb_sha1;
2610 };
2611
2612 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2613 {
2614 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2615 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2616 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2617 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2618 to->sd_dev = htonl(from->sd_dev);
2619 to->sd_ino = htonl(from->sd_ino);
2620 to->sd_uid = htonl(from->sd_uid);
2621 to->sd_gid = htonl(from->sd_gid);
2622 to->sd_size = htonl(from->sd_size);
2623 }
2624
2625 static void write_one_dir(struct untracked_cache_dir *untracked,
2626 struct write_data *wd)
2627 {
2628 struct stat_data stat_data;
2629 struct strbuf *out = &wd->out;
2630 unsigned char intbuf[16];
2631 unsigned int intlen, value;
2632 int i = wd->index++;
2633
2634 /*
2635 * untracked_nr should be reset whenever valid is clear, but
2636 * for safety..
2637 */
2638 if (!untracked->valid) {
2639 untracked->untracked_nr = 0;
2640 untracked->check_only = 0;
2641 }
2642
2643 if (untracked->check_only)
2644 ewah_set(wd->check_only, i);
2645 if (untracked->valid) {
2646 ewah_set(wd->valid, i);
2647 stat_data_to_disk(&stat_data, &untracked->stat_data);
2648 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2649 }
2650 if (!is_null_oid(&untracked->exclude_oid)) {
2651 ewah_set(wd->sha1_valid, i);
2652 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
2653 the_hash_algo->rawsz);
2654 }
2655
2656 intlen = encode_varint(untracked->untracked_nr, intbuf);
2657 strbuf_add(out, intbuf, intlen);
2658
2659 /* skip non-recurse directories */
2660 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2661 if (untracked->dirs[i]->recurse)
2662 value++;
2663 intlen = encode_varint(value, intbuf);
2664 strbuf_add(out, intbuf, intlen);
2665
2666 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2667
2668 for (i = 0; i < untracked->untracked_nr; i++)
2669 strbuf_add(out, untracked->untracked[i],
2670 strlen(untracked->untracked[i]) + 1);
2671
2672 for (i = 0; i < untracked->dirs_nr; i++)
2673 if (untracked->dirs[i]->recurse)
2674 write_one_dir(untracked->dirs[i], wd);
2675 }
2676
2677 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2678 {
2679 struct ondisk_untracked_cache *ouc;
2680 struct write_data wd;
2681 unsigned char varbuf[16];
2682 int varint_len;
2683 const unsigned hashsz = the_hash_algo->rawsz;
2684
2685 ouc = xcalloc(1, sizeof(*ouc));
2686 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2687 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2688 ouc->dir_flags = htonl(untracked->dir_flags);
2689
2690 varint_len = encode_varint(untracked->ident.len, varbuf);
2691 strbuf_add(out, varbuf, varint_len);
2692 strbuf_addbuf(out, &untracked->ident);
2693
2694 strbuf_add(out, ouc, sizeof(*ouc));
2695 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
2696 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
2697 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
2698 FREE_AND_NULL(ouc);
2699
2700 if (!untracked->root) {
2701 varint_len = encode_varint(0, varbuf);
2702 strbuf_add(out, varbuf, varint_len);
2703 return;
2704 }
2705
2706 wd.index = 0;
2707 wd.check_only = ewah_new();
2708 wd.valid = ewah_new();
2709 wd.sha1_valid = ewah_new();
2710 strbuf_init(&wd.out, 1024);
2711 strbuf_init(&wd.sb_stat, 1024);
2712 strbuf_init(&wd.sb_sha1, 1024);
2713 write_one_dir(untracked->root, &wd);
2714
2715 varint_len = encode_varint(wd.index, varbuf);
2716 strbuf_add(out, varbuf, varint_len);
2717 strbuf_addbuf(out, &wd.out);
2718 ewah_serialize_strbuf(wd.valid, out);
2719 ewah_serialize_strbuf(wd.check_only, out);
2720 ewah_serialize_strbuf(wd.sha1_valid, out);
2721 strbuf_addbuf(out, &wd.sb_stat);
2722 strbuf_addbuf(out, &wd.sb_sha1);
2723 strbuf_addch(out, '\0'); /* safe guard for string lists */
2724
2725 ewah_free(wd.valid);
2726 ewah_free(wd.check_only);
2727 ewah_free(wd.sha1_valid);
2728 strbuf_release(&wd.out);
2729 strbuf_release(&wd.sb_stat);
2730 strbuf_release(&wd.sb_sha1);
2731 }
2732
2733 static void free_untracked(struct untracked_cache_dir *ucd)
2734 {
2735 int i;
2736 if (!ucd)
2737 return;
2738 for (i = 0; i < ucd->dirs_nr; i++)
2739 free_untracked(ucd->dirs[i]);
2740 for (i = 0; i < ucd->untracked_nr; i++)
2741 free(ucd->untracked[i]);
2742 free(ucd->untracked);
2743 free(ucd->dirs);
2744 free(ucd);
2745 }
2746
2747 void free_untracked_cache(struct untracked_cache *uc)
2748 {
2749 if (uc)
2750 free_untracked(uc->root);
2751 free(uc);
2752 }
2753
2754 struct read_data {
2755 int index;
2756 struct untracked_cache_dir **ucd;
2757 struct ewah_bitmap *check_only;
2758 struct ewah_bitmap *valid;
2759 struct ewah_bitmap *sha1_valid;
2760 const unsigned char *data;
2761 const unsigned char *end;
2762 };
2763
2764 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2765 {
2766 memcpy(to, data, sizeof(*to));
2767 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2768 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2769 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2770 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2771 to->sd_dev = ntohl(to->sd_dev);
2772 to->sd_ino = ntohl(to->sd_ino);
2773 to->sd_uid = ntohl(to->sd_uid);
2774 to->sd_gid = ntohl(to->sd_gid);
2775 to->sd_size = ntohl(to->sd_size);
2776 }
2777
2778 static int read_one_dir(struct untracked_cache_dir **untracked_,
2779 struct read_data *rd)
2780 {
2781 struct untracked_cache_dir ud, *untracked;
2782 const unsigned char *data = rd->data, *end = rd->end;
2783 const unsigned char *eos;
2784 unsigned int value;
2785 int i;
2786
2787 memset(&ud, 0, sizeof(ud));
2788
2789 value = decode_varint(&data);
2790 if (data > end)
2791 return -1;
2792 ud.recurse = 1;
2793 ud.untracked_alloc = value;
2794 ud.untracked_nr = value;
2795 if (ud.untracked_nr)
2796 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2797
2798 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
2799 if (data > end)
2800 return -1;
2801 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2802
2803 eos = memchr(data, '\0', end - data);
2804 if (!eos || eos == end)
2805 return -1;
2806
2807 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
2808 memcpy(untracked, &ud, sizeof(ud));
2809 memcpy(untracked->name, data, eos - data + 1);
2810 data = eos + 1;
2811
2812 for (i = 0; i < untracked->untracked_nr; i++) {
2813 eos = memchr(data, '\0', end - data);
2814 if (!eos || eos == end)
2815 return -1;
2816 untracked->untracked[i] = xmemdupz(data, eos - data);
2817 data = eos + 1;
2818 }
2819
2820 rd->ucd[rd->index++] = untracked;
2821 rd->data = data;
2822
2823 for (i = 0; i < untracked->dirs_nr; i++) {
2824 if (read_one_dir(untracked->dirs + i, rd) < 0)
2825 return -1;
2826 }
2827 return 0;
2828 }
2829
2830 static void set_check_only(size_t pos, void *cb)
2831 {
2832 struct read_data *rd = cb;
2833 struct untracked_cache_dir *ud = rd->ucd[pos];
2834 ud->check_only = 1;
2835 }
2836
2837 static void read_stat(size_t pos, void *cb)
2838 {
2839 struct read_data *rd = cb;
2840 struct untracked_cache_dir *ud = rd->ucd[pos];
2841 if (rd->data + sizeof(struct stat_data) > rd->end) {
2842 rd->data = rd->end + 1;
2843 return;
2844 }
2845 stat_data_from_disk(&ud->stat_data, rd->data);
2846 rd->data += sizeof(struct stat_data);
2847 ud->valid = 1;
2848 }
2849
2850 static void read_oid(size_t pos, void *cb)
2851 {
2852 struct read_data *rd = cb;
2853 struct untracked_cache_dir *ud = rd->ucd[pos];
2854 if (rd->data + the_hash_algo->rawsz > rd->end) {
2855 rd->data = rd->end + 1;
2856 return;
2857 }
2858 hashcpy(ud->exclude_oid.hash, rd->data);
2859 rd->data += the_hash_algo->rawsz;
2860 }
2861
2862 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
2863 const unsigned char *sha1)
2864 {
2865 stat_data_from_disk(&oid_stat->stat, data);
2866 hashcpy(oid_stat->oid.hash, sha1);
2867 oid_stat->valid = 1;
2868 }
2869
2870 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2871 {
2872 struct untracked_cache *uc;
2873 struct read_data rd;
2874 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2875 const char *ident;
2876 int ident_len;
2877 ssize_t len;
2878 const char *exclude_per_dir;
2879 const unsigned hashsz = the_hash_algo->rawsz;
2880 const unsigned offset = sizeof(struct ondisk_untracked_cache);
2881 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
2882
2883 if (sz <= 1 || end[-1] != '\0')
2884 return NULL;
2885 end--;
2886
2887 ident_len = decode_varint(&next);
2888 if (next + ident_len > end)
2889 return NULL;
2890 ident = (const char *)next;
2891 next += ident_len;
2892
2893 if (next + exclude_per_dir_offset + 1 > end)
2894 return NULL;
2895
2896 uc = xcalloc(1, sizeof(*uc));
2897 strbuf_init(&uc->ident, ident_len);
2898 strbuf_add(&uc->ident, ident, ident_len);
2899 load_oid_stat(&uc->ss_info_exclude,
2900 next + ouc_offset(info_exclude_stat),
2901 next + offset);
2902 load_oid_stat(&uc->ss_excludes_file,
2903 next + ouc_offset(excludes_file_stat),
2904 next + offset + hashsz);
2905 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2906 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
2907 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2908 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2909 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
2910 if (next >= end)
2911 goto done2;
2912
2913 len = decode_varint(&next);
2914 if (next > end || len == 0)
2915 goto done2;
2916
2917 rd.valid = ewah_new();
2918 rd.check_only = ewah_new();
2919 rd.sha1_valid = ewah_new();
2920 rd.data = next;
2921 rd.end = end;
2922 rd.index = 0;
2923 ALLOC_ARRAY(rd.ucd, len);
2924
2925 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2926 goto done;
2927
2928 next = rd.data;
2929 len = ewah_read_mmap(rd.valid, next, end - next);
2930 if (len < 0)
2931 goto done;
2932
2933 next += len;
2934 len = ewah_read_mmap(rd.check_only, next, end - next);
2935 if (len < 0)
2936 goto done;
2937
2938 next += len;
2939 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2940 if (len < 0)
2941 goto done;
2942
2943 ewah_each_bit(rd.check_only, set_check_only, &rd);
2944 rd.data = next + len;
2945 ewah_each_bit(rd.valid, read_stat, &rd);
2946 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
2947 next = rd.data;
2948
2949 done:
2950 free(rd.ucd);
2951 ewah_free(rd.valid);
2952 ewah_free(rd.check_only);
2953 ewah_free(rd.sha1_valid);
2954 done2:
2955 if (next != end) {
2956 free_untracked_cache(uc);
2957 uc = NULL;
2958 }
2959 return uc;
2960 }
2961
2962 static void invalidate_one_directory(struct untracked_cache *uc,
2963 struct untracked_cache_dir *ucd)
2964 {
2965 uc->dir_invalidated++;
2966 ucd->valid = 0;
2967 ucd->untracked_nr = 0;
2968 }
2969
2970 /*
2971 * Normally when an entry is added or removed from a directory,
2972 * invalidating that directory is enough. No need to touch its
2973 * ancestors. When a directory is shown as "foo/bar/" in git-status
2974 * however, deleting or adding an entry may have cascading effect.
2975 *
2976 * Say the "foo/bar/file" has become untracked, we need to tell the
2977 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2978 * directory any more (because "bar" is managed by foo as an untracked
2979 * "file").
2980 *
2981 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2982 * was the last untracked entry in the entire "foo", we should show
2983 * "foo/" instead. Which means we have to invalidate past "bar" up to
2984 * "foo".
2985 *
2986 * This function traverses all directories from root to leaf. If there
2987 * is a chance of one of the above cases happening, we invalidate back
2988 * to root. Otherwise we just invalidate the leaf. There may be a more
2989 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2990 * detect these cases and avoid unnecessary invalidation, for example,
2991 * checking for the untracked entry named "bar/" in "foo", but for now
2992 * stick to something safe and simple.
2993 */
2994 static int invalidate_one_component(struct untracked_cache *uc,
2995 struct untracked_cache_dir *dir,
2996 const char *path, int len)
2997 {
2998 const char *rest = strchr(path, '/');
2999
3000 if (rest) {
3001 int component_len = rest - path;
3002 struct untracked_cache_dir *d =
3003 lookup_untracked(uc, dir, path, component_len);
3004 int ret =
3005 invalidate_one_component(uc, d, rest + 1,
3006 len - (component_len + 1));
3007 if (ret)
3008 invalidate_one_directory(uc, dir);
3009 return ret;
3010 }
3011
3012 invalidate_one_directory(uc, dir);
3013 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
3014 }
3015
3016 void untracked_cache_invalidate_path(struct index_state *istate,
3017 const char *path, int safe_path)
3018 {
3019 if (!istate->untracked || !istate->untracked->root)
3020 return;
3021 if (!safe_path && !verify_path(path, 0))
3022 return;
3023 invalidate_one_component(istate->untracked, istate->untracked->root,
3024 path, strlen(path));
3025 }
3026
3027 void untracked_cache_remove_from_index(struct index_state *istate,
3028 const char *path)
3029 {
3030 untracked_cache_invalidate_path(istate, path, 1);
3031 }
3032
3033 void untracked_cache_add_to_index(struct index_state *istate,
3034 const char *path)
3035 {
3036 untracked_cache_invalidate_path(istate, path, 1);
3037 }
3038
3039 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3040 const char *sub_gitdir)
3041 {
3042 int i;
3043 struct repository subrepo;
3044 struct strbuf sub_wt = STRBUF_INIT;
3045 struct strbuf sub_gd = STRBUF_INIT;
3046
3047 const struct submodule *sub;
3048
3049 /* If the submodule has no working tree, we can ignore it. */
3050 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3051 return;
3052
3053 if (repo_read_index(&subrepo) < 0)
3054 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3055
3056 for (i = 0; i < subrepo.index->cache_nr; i++) {
3057 const struct cache_entry *ce = subrepo.index->cache[i];
3058
3059 if (!S_ISGITLINK(ce->ce_mode))
3060 continue;
3061
3062 while (i + 1 < subrepo.index->cache_nr &&
3063 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3064 /*
3065 * Skip entries with the same name in different stages
3066 * to make sure an entry is returned only once.
3067 */
3068 i++;
3069
3070 sub = submodule_from_path(&subrepo, &null_oid, ce->name);
3071 if (!sub || !is_submodule_active(&subrepo, ce->name))
3072 /* .gitmodules broken or inactive sub */
3073 continue;
3074
3075 strbuf_reset(&sub_wt);
3076 strbuf_reset(&sub_gd);
3077 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3078 strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name);
3079
3080 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3081 }
3082 strbuf_release(&sub_wt);
3083 strbuf_release(&sub_gd);
3084 repo_clear(&subrepo);
3085 }
3086
3087 void connect_work_tree_and_git_dir(const char *work_tree_,
3088 const char *git_dir_,
3089 int recurse_into_nested)
3090 {
3091 struct strbuf gitfile_sb = STRBUF_INIT;
3092 struct strbuf cfg_sb = STRBUF_INIT;
3093 struct strbuf rel_path = STRBUF_INIT;
3094 char *git_dir, *work_tree;
3095
3096 /* Prepare .git file */
3097 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3098 if (safe_create_leading_directories_const(gitfile_sb.buf))
3099 die(_("could not create directories for %s"), gitfile_sb.buf);
3100
3101 /* Prepare config file */
3102 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3103 if (safe_create_leading_directories_const(cfg_sb.buf))
3104 die(_("could not create directories for %s"), cfg_sb.buf);
3105
3106 git_dir = real_pathdup(git_dir_, 1);
3107 work_tree = real_pathdup(work_tree_, 1);
3108
3109 /* Write .git file */
3110 write_file(gitfile_sb.buf, "gitdir: %s",
3111 relative_path(git_dir, work_tree, &rel_path));
3112 /* Update core.worktree setting */
3113 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3114 relative_path(work_tree, git_dir, &rel_path));
3115
3116 strbuf_release(&gitfile_sb);
3117 strbuf_release(&cfg_sb);
3118 strbuf_release(&rel_path);
3119
3120 if (recurse_into_nested)
3121 connect_wt_gitdir_in_nested(work_tree, git_dir);
3122
3123 free(work_tree);
3124 free(git_dir);
3125 }
3126
3127 /*
3128 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3129 */
3130 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3131 {
3132 if (rename(old_git_dir, new_git_dir) < 0)
3133 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3134 old_git_dir, new_git_dir);
3135
3136 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3137 }