untracked-cache: simplify parsing by dropping "len"
[git/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * See Documentation/technical/api-directory-listing.txt
6 *
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10 #include "cache.h"
11 #include "config.h"
12 #include "dir.h"
13 #include "object-store.h"
14 #include "attr.h"
15 #include "refs.h"
16 #include "wildmatch.h"
17 #include "pathspec.h"
18 #include "utf8.h"
19 #include "varint.h"
20 #include "ewah/ewok.h"
21 #include "fsmonitor.h"
22 #include "submodule-config.h"
23
24 /*
25 * Tells read_directory_recursive how a file or directory should be treated.
26 * Values are ordered by significance, e.g. if a directory contains both
27 * excluded and untracked files, it is listed as untracked because
28 * path_untracked > path_excluded.
29 */
30 enum path_treatment {
31 path_none = 0,
32 path_recurse,
33 path_excluded,
34 path_untracked
35 };
36
37 /*
38 * Support data structure for our opendir/readdir/closedir wrappers
39 */
40 struct cached_dir {
41 DIR *fdir;
42 struct untracked_cache_dir *untracked;
43 int nr_files;
44 int nr_dirs;
45
46 struct dirent *de;
47 const char *file;
48 struct untracked_cache_dir *ucd;
49 };
50
51 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
52 struct index_state *istate, const char *path, int len,
53 struct untracked_cache_dir *untracked,
54 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
55 static int get_dtype(struct dirent *de, struct index_state *istate,
56 const char *path, int len);
57
58 int count_slashes(const char *s)
59 {
60 int cnt = 0;
61 while (*s)
62 if (*s++ == '/')
63 cnt++;
64 return cnt;
65 }
66
67 int fspathcmp(const char *a, const char *b)
68 {
69 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
70 }
71
72 int fspathncmp(const char *a, const char *b, size_t count)
73 {
74 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
75 }
76
77 int git_fnmatch(const struct pathspec_item *item,
78 const char *pattern, const char *string,
79 int prefix)
80 {
81 if (prefix > 0) {
82 if (ps_strncmp(item, pattern, string, prefix))
83 return WM_NOMATCH;
84 pattern += prefix;
85 string += prefix;
86 }
87 if (item->flags & PATHSPEC_ONESTAR) {
88 int pattern_len = strlen(++pattern);
89 int string_len = strlen(string);
90 return string_len < pattern_len ||
91 ps_strcmp(item, pattern,
92 string + string_len - pattern_len);
93 }
94 if (item->magic & PATHSPEC_GLOB)
95 return wildmatch(pattern, string,
96 WM_PATHNAME |
97 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
98 else
99 /* wildmatch has not learned no FNM_PATHNAME mode yet */
100 return wildmatch(pattern, string,
101 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
102 }
103
104 static int fnmatch_icase_mem(const char *pattern, int patternlen,
105 const char *string, int stringlen,
106 int flags)
107 {
108 int match_status;
109 struct strbuf pat_buf = STRBUF_INIT;
110 struct strbuf str_buf = STRBUF_INIT;
111 const char *use_pat = pattern;
112 const char *use_str = string;
113
114 if (pattern[patternlen]) {
115 strbuf_add(&pat_buf, pattern, patternlen);
116 use_pat = pat_buf.buf;
117 }
118 if (string[stringlen]) {
119 strbuf_add(&str_buf, string, stringlen);
120 use_str = str_buf.buf;
121 }
122
123 if (ignore_case)
124 flags |= WM_CASEFOLD;
125 match_status = wildmatch(use_pat, use_str, flags);
126
127 strbuf_release(&pat_buf);
128 strbuf_release(&str_buf);
129
130 return match_status;
131 }
132
133 static size_t common_prefix_len(const struct pathspec *pathspec)
134 {
135 int n;
136 size_t max = 0;
137
138 /*
139 * ":(icase)path" is treated as a pathspec full of
140 * wildcard. In other words, only prefix is considered common
141 * prefix. If the pathspec is abc/foo abc/bar, running in
142 * subdir xyz, the common prefix is still xyz, not xuz/abc as
143 * in non-:(icase).
144 */
145 GUARD_PATHSPEC(pathspec,
146 PATHSPEC_FROMTOP |
147 PATHSPEC_MAXDEPTH |
148 PATHSPEC_LITERAL |
149 PATHSPEC_GLOB |
150 PATHSPEC_ICASE |
151 PATHSPEC_EXCLUDE |
152 PATHSPEC_ATTR);
153
154 for (n = 0; n < pathspec->nr; n++) {
155 size_t i = 0, len = 0, item_len;
156 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
157 continue;
158 if (pathspec->items[n].magic & PATHSPEC_ICASE)
159 item_len = pathspec->items[n].prefix;
160 else
161 item_len = pathspec->items[n].nowildcard_len;
162 while (i < item_len && (n == 0 || i < max)) {
163 char c = pathspec->items[n].match[i];
164 if (c != pathspec->items[0].match[i])
165 break;
166 if (c == '/')
167 len = i + 1;
168 i++;
169 }
170 if (n == 0 || len < max) {
171 max = len;
172 if (!max)
173 break;
174 }
175 }
176 return max;
177 }
178
179 /*
180 * Returns a copy of the longest leading path common among all
181 * pathspecs.
182 */
183 char *common_prefix(const struct pathspec *pathspec)
184 {
185 unsigned long len = common_prefix_len(pathspec);
186
187 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
188 }
189
190 int fill_directory(struct dir_struct *dir,
191 struct index_state *istate,
192 const struct pathspec *pathspec)
193 {
194 const char *prefix;
195 size_t prefix_len;
196
197 /*
198 * Calculate common prefix for the pathspec, and
199 * use that to optimize the directory walk
200 */
201 prefix_len = common_prefix_len(pathspec);
202 prefix = prefix_len ? pathspec->items[0].match : "";
203
204 /* Read the directory and prune it */
205 read_directory(dir, istate, prefix, prefix_len, pathspec);
206
207 return prefix_len;
208 }
209
210 int within_depth(const char *name, int namelen,
211 int depth, int max_depth)
212 {
213 const char *cp = name, *cpe = name + namelen;
214
215 while (cp < cpe) {
216 if (*cp++ != '/')
217 continue;
218 depth++;
219 if (depth > max_depth)
220 return 0;
221 }
222 return 1;
223 }
224
225 /*
226 * Read the contents of the blob with the given OID into a buffer.
227 * Append a trailing LF to the end if the last line doesn't have one.
228 *
229 * Returns:
230 * -1 when the OID is invalid or unknown or does not refer to a blob.
231 * 0 when the blob is empty.
232 * 1 along with { data, size } of the (possibly augmented) buffer
233 * when successful.
234 *
235 * Optionally updates the given oid_stat with the given OID (when valid).
236 */
237 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
238 size_t *size_out, char **data_out)
239 {
240 enum object_type type;
241 unsigned long sz;
242 char *data;
243
244 *size_out = 0;
245 *data_out = NULL;
246
247 data = read_object_file(oid, &type, &sz);
248 if (!data || type != OBJ_BLOB) {
249 free(data);
250 return -1;
251 }
252
253 if (oid_stat) {
254 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
255 oidcpy(&oid_stat->oid, oid);
256 }
257
258 if (sz == 0) {
259 free(data);
260 return 0;
261 }
262
263 if (data[sz - 1] != '\n') {
264 data = xrealloc(data, st_add(sz, 1));
265 data[sz++] = '\n';
266 }
267
268 *size_out = xsize_t(sz);
269 *data_out = data;
270
271 return 1;
272 }
273
274 #define DO_MATCH_EXCLUDE (1<<0)
275 #define DO_MATCH_DIRECTORY (1<<1)
276 #define DO_MATCH_SUBMODULE (1<<2)
277
278 /*
279 * Does 'match' match the given name?
280 * A match is found if
281 *
282 * (1) the 'match' string is leading directory of 'name', or
283 * (2) the 'match' string is a wildcard and matches 'name', or
284 * (3) the 'match' string is exactly the same as 'name'.
285 *
286 * and the return value tells which case it was.
287 *
288 * It returns 0 when there is no match.
289 */
290 static int match_pathspec_item(const struct index_state *istate,
291 const struct pathspec_item *item, int prefix,
292 const char *name, int namelen, unsigned flags)
293 {
294 /* name/namelen has prefix cut off by caller */
295 const char *match = item->match + prefix;
296 int matchlen = item->len - prefix;
297
298 /*
299 * The normal call pattern is:
300 * 1. prefix = common_prefix_len(ps);
301 * 2. prune something, or fill_directory
302 * 3. match_pathspec()
303 *
304 * 'prefix' at #1 may be shorter than the command's prefix and
305 * it's ok for #2 to match extra files. Those extras will be
306 * trimmed at #3.
307 *
308 * Suppose the pathspec is 'foo' and '../bar' running from
309 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
310 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
311 * user does not want XYZ/foo, only the "foo" part should be
312 * case-insensitive. We need to filter out XYZ/foo here. In
313 * other words, we do not trust the caller on comparing the
314 * prefix part when :(icase) is involved. We do exact
315 * comparison ourselves.
316 *
317 * Normally the caller (common_prefix_len() in fact) does
318 * _exact_ matching on name[-prefix+1..-1] and we do not need
319 * to check that part. Be defensive and check it anyway, in
320 * case common_prefix_len is changed, or a new caller is
321 * introduced that does not use common_prefix_len.
322 *
323 * If the penalty turns out too high when prefix is really
324 * long, maybe change it to
325 * strncmp(match, name, item->prefix - prefix)
326 */
327 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
328 strncmp(item->match, name - prefix, item->prefix))
329 return 0;
330
331 if (item->attr_match_nr &&
332 !match_pathspec_attrs(istate, name, namelen, item))
333 return 0;
334
335 /* If the match was just the prefix, we matched */
336 if (!*match)
337 return MATCHED_RECURSIVELY;
338
339 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
340 if (matchlen == namelen)
341 return MATCHED_EXACTLY;
342
343 if (match[matchlen-1] == '/' || name[matchlen] == '/')
344 return MATCHED_RECURSIVELY;
345 } else if ((flags & DO_MATCH_DIRECTORY) &&
346 match[matchlen - 1] == '/' &&
347 namelen == matchlen - 1 &&
348 !ps_strncmp(item, match, name, namelen))
349 return MATCHED_EXACTLY;
350
351 if (item->nowildcard_len < item->len &&
352 !git_fnmatch(item, match, name,
353 item->nowildcard_len - prefix))
354 return MATCHED_FNMATCH;
355
356 /* Perform checks to see if "name" is a super set of the pathspec */
357 if (flags & DO_MATCH_SUBMODULE) {
358 /* name is a literal prefix of the pathspec */
359 if ((namelen < matchlen) &&
360 (match[namelen] == '/') &&
361 !ps_strncmp(item, match, name, namelen))
362 return MATCHED_RECURSIVELY;
363
364 /* name" doesn't match up to the first wild character */
365 if (item->nowildcard_len < item->len &&
366 ps_strncmp(item, match, name,
367 item->nowildcard_len - prefix))
368 return 0;
369
370 /*
371 * Here is where we would perform a wildmatch to check if
372 * "name" can be matched as a directory (or a prefix) against
373 * the pathspec. Since wildmatch doesn't have this capability
374 * at the present we have to punt and say that it is a match,
375 * potentially returning a false positive
376 * The submodules themselves will be able to perform more
377 * accurate matching to determine if the pathspec matches.
378 */
379 return MATCHED_RECURSIVELY;
380 }
381
382 return 0;
383 }
384
385 /*
386 * Given a name and a list of pathspecs, returns the nature of the
387 * closest (i.e. most specific) match of the name to any of the
388 * pathspecs.
389 *
390 * The caller typically calls this multiple times with the same
391 * pathspec and seen[] array but with different name/namelen
392 * (e.g. entries from the index) and is interested in seeing if and
393 * how each pathspec matches all the names it calls this function
394 * with. A mark is left in the seen[] array for each pathspec element
395 * indicating the closest type of match that element achieved, so if
396 * seen[n] remains zero after multiple invocations, that means the nth
397 * pathspec did not match any names, which could indicate that the
398 * user mistyped the nth pathspec.
399 */
400 static int do_match_pathspec(const struct index_state *istate,
401 const struct pathspec *ps,
402 const char *name, int namelen,
403 int prefix, char *seen,
404 unsigned flags)
405 {
406 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
407
408 GUARD_PATHSPEC(ps,
409 PATHSPEC_FROMTOP |
410 PATHSPEC_MAXDEPTH |
411 PATHSPEC_LITERAL |
412 PATHSPEC_GLOB |
413 PATHSPEC_ICASE |
414 PATHSPEC_EXCLUDE |
415 PATHSPEC_ATTR);
416
417 if (!ps->nr) {
418 if (!ps->recursive ||
419 !(ps->magic & PATHSPEC_MAXDEPTH) ||
420 ps->max_depth == -1)
421 return MATCHED_RECURSIVELY;
422
423 if (within_depth(name, namelen, 0, ps->max_depth))
424 return MATCHED_EXACTLY;
425 else
426 return 0;
427 }
428
429 name += prefix;
430 namelen -= prefix;
431
432 for (i = ps->nr - 1; i >= 0; i--) {
433 int how;
434
435 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
436 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
437 continue;
438
439 if (seen && seen[i] == MATCHED_EXACTLY)
440 continue;
441 /*
442 * Make exclude patterns optional and never report
443 * "pathspec ':(exclude)foo' matches no files"
444 */
445 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
446 seen[i] = MATCHED_FNMATCH;
447 how = match_pathspec_item(istate, ps->items+i, prefix, name,
448 namelen, flags);
449 if (ps->recursive &&
450 (ps->magic & PATHSPEC_MAXDEPTH) &&
451 ps->max_depth != -1 &&
452 how && how != MATCHED_FNMATCH) {
453 int len = ps->items[i].len;
454 if (name[len] == '/')
455 len++;
456 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
457 how = MATCHED_EXACTLY;
458 else
459 how = 0;
460 }
461 if (how) {
462 if (retval < how)
463 retval = how;
464 if (seen && seen[i] < how)
465 seen[i] = how;
466 }
467 }
468 return retval;
469 }
470
471 int match_pathspec(const struct index_state *istate,
472 const struct pathspec *ps,
473 const char *name, int namelen,
474 int prefix, char *seen, int is_dir)
475 {
476 int positive, negative;
477 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
478 positive = do_match_pathspec(istate, ps, name, namelen,
479 prefix, seen, flags);
480 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
481 return positive;
482 negative = do_match_pathspec(istate, ps, name, namelen,
483 prefix, seen,
484 flags | DO_MATCH_EXCLUDE);
485 return negative ? 0 : positive;
486 }
487
488 /**
489 * Check if a submodule is a superset of the pathspec
490 */
491 int submodule_path_match(const struct index_state *istate,
492 const struct pathspec *ps,
493 const char *submodule_name,
494 char *seen)
495 {
496 int matched = do_match_pathspec(istate, ps, submodule_name,
497 strlen(submodule_name),
498 0, seen,
499 DO_MATCH_DIRECTORY |
500 DO_MATCH_SUBMODULE);
501 return matched;
502 }
503
504 int report_path_error(const char *ps_matched,
505 const struct pathspec *pathspec,
506 const char *prefix)
507 {
508 /*
509 * Make sure all pathspec matched; otherwise it is an error.
510 */
511 int num, errors = 0;
512 for (num = 0; num < pathspec->nr; num++) {
513 int other, found_dup;
514
515 if (ps_matched[num])
516 continue;
517 /*
518 * The caller might have fed identical pathspec
519 * twice. Do not barf on such a mistake.
520 * FIXME: parse_pathspec should have eliminated
521 * duplicate pathspec.
522 */
523 for (found_dup = other = 0;
524 !found_dup && other < pathspec->nr;
525 other++) {
526 if (other == num || !ps_matched[other])
527 continue;
528 if (!strcmp(pathspec->items[other].original,
529 pathspec->items[num].original))
530 /*
531 * Ok, we have a match already.
532 */
533 found_dup = 1;
534 }
535 if (found_dup)
536 continue;
537
538 error(_("pathspec '%s' did not match any file(s) known to git"),
539 pathspec->items[num].original);
540 errors++;
541 }
542 return errors;
543 }
544
545 /*
546 * Return the length of the "simple" part of a path match limiter.
547 */
548 int simple_length(const char *match)
549 {
550 int len = -1;
551
552 for (;;) {
553 unsigned char c = *match++;
554 len++;
555 if (c == '\0' || is_glob_special(c))
556 return len;
557 }
558 }
559
560 int no_wildcard(const char *string)
561 {
562 return string[simple_length(string)] == '\0';
563 }
564
565 void parse_exclude_pattern(const char **pattern,
566 int *patternlen,
567 unsigned *flags,
568 int *nowildcardlen)
569 {
570 const char *p = *pattern;
571 size_t i, len;
572
573 *flags = 0;
574 if (*p == '!') {
575 *flags |= EXC_FLAG_NEGATIVE;
576 p++;
577 }
578 len = strlen(p);
579 if (len && p[len - 1] == '/') {
580 len--;
581 *flags |= EXC_FLAG_MUSTBEDIR;
582 }
583 for (i = 0; i < len; i++) {
584 if (p[i] == '/')
585 break;
586 }
587 if (i == len)
588 *flags |= EXC_FLAG_NODIR;
589 *nowildcardlen = simple_length(p);
590 /*
591 * we should have excluded the trailing slash from 'p' too,
592 * but that's one more allocation. Instead just make sure
593 * nowildcardlen does not exceed real patternlen
594 */
595 if (*nowildcardlen > len)
596 *nowildcardlen = len;
597 if (*p == '*' && no_wildcard(p + 1))
598 *flags |= EXC_FLAG_ENDSWITH;
599 *pattern = p;
600 *patternlen = len;
601 }
602
603 void add_exclude(const char *string, const char *base,
604 int baselen, struct exclude_list *el, int srcpos)
605 {
606 struct exclude *x;
607 int patternlen;
608 unsigned flags;
609 int nowildcardlen;
610
611 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
612 if (flags & EXC_FLAG_MUSTBEDIR) {
613 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
614 } else {
615 x = xmalloc(sizeof(*x));
616 x->pattern = string;
617 }
618 x->patternlen = patternlen;
619 x->nowildcardlen = nowildcardlen;
620 x->base = base;
621 x->baselen = baselen;
622 x->flags = flags;
623 x->srcpos = srcpos;
624 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
625 el->excludes[el->nr++] = x;
626 x->el = el;
627 }
628
629 static int read_skip_worktree_file_from_index(const struct index_state *istate,
630 const char *path,
631 size_t *size_out, char **data_out,
632 struct oid_stat *oid_stat)
633 {
634 int pos, len;
635
636 len = strlen(path);
637 pos = index_name_pos(istate, path, len);
638 if (pos < 0)
639 return -1;
640 if (!ce_skip_worktree(istate->cache[pos]))
641 return -1;
642
643 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
644 }
645
646 /*
647 * Frees memory within el which was allocated for exclude patterns and
648 * the file buffer. Does not free el itself.
649 */
650 void clear_exclude_list(struct exclude_list *el)
651 {
652 int i;
653
654 for (i = 0; i < el->nr; i++)
655 free(el->excludes[i]);
656 free(el->excludes);
657 free(el->filebuf);
658
659 memset(el, 0, sizeof(*el));
660 }
661
662 static void trim_trailing_spaces(char *buf)
663 {
664 char *p, *last_space = NULL;
665
666 for (p = buf; *p; p++)
667 switch (*p) {
668 case ' ':
669 if (!last_space)
670 last_space = p;
671 break;
672 case '\\':
673 p++;
674 if (!*p)
675 return;
676 /* fallthrough */
677 default:
678 last_space = NULL;
679 }
680
681 if (last_space)
682 *last_space = '\0';
683 }
684
685 /*
686 * Given a subdirectory name and "dir" of the current directory,
687 * search the subdir in "dir" and return it, or create a new one if it
688 * does not exist in "dir".
689 *
690 * If "name" has the trailing slash, it'll be excluded in the search.
691 */
692 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
693 struct untracked_cache_dir *dir,
694 const char *name, int len)
695 {
696 int first, last;
697 struct untracked_cache_dir *d;
698 if (!dir)
699 return NULL;
700 if (len && name[len - 1] == '/')
701 len--;
702 first = 0;
703 last = dir->dirs_nr;
704 while (last > first) {
705 int cmp, next = (last + first) >> 1;
706 d = dir->dirs[next];
707 cmp = strncmp(name, d->name, len);
708 if (!cmp && strlen(d->name) > len)
709 cmp = -1;
710 if (!cmp)
711 return d;
712 if (cmp < 0) {
713 last = next;
714 continue;
715 }
716 first = next+1;
717 }
718
719 uc->dir_created++;
720 FLEX_ALLOC_MEM(d, name, name, len);
721
722 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
723 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
724 dir->dirs_nr - first);
725 dir->dirs_nr++;
726 dir->dirs[first] = d;
727 return d;
728 }
729
730 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
731 {
732 int i;
733 dir->valid = 0;
734 dir->untracked_nr = 0;
735 for (i = 0; i < dir->dirs_nr; i++)
736 do_invalidate_gitignore(dir->dirs[i]);
737 }
738
739 static void invalidate_gitignore(struct untracked_cache *uc,
740 struct untracked_cache_dir *dir)
741 {
742 uc->gitignore_invalidated++;
743 do_invalidate_gitignore(dir);
744 }
745
746 static void invalidate_directory(struct untracked_cache *uc,
747 struct untracked_cache_dir *dir)
748 {
749 int i;
750
751 /*
752 * Invalidation increment here is just roughly correct. If
753 * untracked_nr or any of dirs[].recurse is non-zero, we
754 * should increment dir_invalidated too. But that's more
755 * expensive to do.
756 */
757 if (dir->valid)
758 uc->dir_invalidated++;
759
760 dir->valid = 0;
761 dir->untracked_nr = 0;
762 for (i = 0; i < dir->dirs_nr; i++)
763 dir->dirs[i]->recurse = 0;
764 }
765
766 static int add_excludes_from_buffer(char *buf, size_t size,
767 const char *base, int baselen,
768 struct exclude_list *el);
769
770 /*
771 * Given a file with name "fname", read it (either from disk, or from
772 * an index if 'istate' is non-null), parse it and store the
773 * exclude rules in "el".
774 *
775 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
776 * stat data from disk (only valid if add_excludes returns zero). If
777 * ss_valid is non-zero, "ss" must contain good value as input.
778 */
779 static int add_excludes(const char *fname, const char *base, int baselen,
780 struct exclude_list *el, struct index_state *istate,
781 struct oid_stat *oid_stat)
782 {
783 struct stat st;
784 int r;
785 int fd;
786 size_t size = 0;
787 char *buf;
788
789 fd = open(fname, O_RDONLY);
790 if (fd < 0 || fstat(fd, &st) < 0) {
791 if (fd < 0)
792 warn_on_fopen_errors(fname);
793 else
794 close(fd);
795 if (!istate)
796 return -1;
797 r = read_skip_worktree_file_from_index(istate, fname,
798 &size, &buf,
799 oid_stat);
800 if (r != 1)
801 return r;
802 } else {
803 size = xsize_t(st.st_size);
804 if (size == 0) {
805 if (oid_stat) {
806 fill_stat_data(&oid_stat->stat, &st);
807 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
808 oid_stat->valid = 1;
809 }
810 close(fd);
811 return 0;
812 }
813 buf = xmallocz(size);
814 if (read_in_full(fd, buf, size) != size) {
815 free(buf);
816 close(fd);
817 return -1;
818 }
819 buf[size++] = '\n';
820 close(fd);
821 if (oid_stat) {
822 int pos;
823 if (oid_stat->valid &&
824 !match_stat_data_racy(istate, &oid_stat->stat, &st))
825 ; /* no content change, ss->sha1 still good */
826 else if (istate &&
827 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
828 !ce_stage(istate->cache[pos]) &&
829 ce_uptodate(istate->cache[pos]) &&
830 !would_convert_to_git(istate, fname))
831 oidcpy(&oid_stat->oid,
832 &istate->cache[pos]->oid);
833 else
834 hash_object_file(buf, size, "blob",
835 &oid_stat->oid);
836 fill_stat_data(&oid_stat->stat, &st);
837 oid_stat->valid = 1;
838 }
839 }
840
841 add_excludes_from_buffer(buf, size, base, baselen, el);
842 return 0;
843 }
844
845 static int add_excludes_from_buffer(char *buf, size_t size,
846 const char *base, int baselen,
847 struct exclude_list *el)
848 {
849 int i, lineno = 1;
850 char *entry;
851
852 el->filebuf = buf;
853
854 if (skip_utf8_bom(&buf, size))
855 size -= buf - el->filebuf;
856
857 entry = buf;
858
859 for (i = 0; i < size; i++) {
860 if (buf[i] == '\n') {
861 if (entry != buf + i && entry[0] != '#') {
862 buf[i - (i && buf[i-1] == '\r')] = 0;
863 trim_trailing_spaces(entry);
864 add_exclude(entry, base, baselen, el, lineno);
865 }
866 lineno++;
867 entry = buf + i + 1;
868 }
869 }
870 return 0;
871 }
872
873 int add_excludes_from_file_to_list(const char *fname, const char *base,
874 int baselen, struct exclude_list *el,
875 struct index_state *istate)
876 {
877 return add_excludes(fname, base, baselen, el, istate, NULL);
878 }
879
880 int add_excludes_from_blob_to_list(
881 struct object_id *oid,
882 const char *base, int baselen,
883 struct exclude_list *el)
884 {
885 char *buf;
886 size_t size;
887 int r;
888
889 r = do_read_blob(oid, NULL, &size, &buf);
890 if (r != 1)
891 return r;
892
893 add_excludes_from_buffer(buf, size, base, baselen, el);
894 return 0;
895 }
896
897 struct exclude_list *add_exclude_list(struct dir_struct *dir,
898 int group_type, const char *src)
899 {
900 struct exclude_list *el;
901 struct exclude_list_group *group;
902
903 group = &dir->exclude_list_group[group_type];
904 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
905 el = &group->el[group->nr++];
906 memset(el, 0, sizeof(*el));
907 el->src = src;
908 return el;
909 }
910
911 /*
912 * Used to set up core.excludesfile and .git/info/exclude lists.
913 */
914 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
915 struct oid_stat *oid_stat)
916 {
917 struct exclude_list *el;
918 /*
919 * catch setup_standard_excludes() that's called before
920 * dir->untracked is assigned. That function behaves
921 * differently when dir->untracked is non-NULL.
922 */
923 if (!dir->untracked)
924 dir->unmanaged_exclude_files++;
925 el = add_exclude_list(dir, EXC_FILE, fname);
926 if (add_excludes(fname, "", 0, el, NULL, oid_stat) < 0)
927 die(_("cannot use %s as an exclude file"), fname);
928 }
929
930 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
931 {
932 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
933 add_excludes_from_file_1(dir, fname, NULL);
934 }
935
936 int match_basename(const char *basename, int basenamelen,
937 const char *pattern, int prefix, int patternlen,
938 unsigned flags)
939 {
940 if (prefix == patternlen) {
941 if (patternlen == basenamelen &&
942 !fspathncmp(pattern, basename, basenamelen))
943 return 1;
944 } else if (flags & EXC_FLAG_ENDSWITH) {
945 /* "*literal" matching against "fooliteral" */
946 if (patternlen - 1 <= basenamelen &&
947 !fspathncmp(pattern + 1,
948 basename + basenamelen - (patternlen - 1),
949 patternlen - 1))
950 return 1;
951 } else {
952 if (fnmatch_icase_mem(pattern, patternlen,
953 basename, basenamelen,
954 0) == 0)
955 return 1;
956 }
957 return 0;
958 }
959
960 int match_pathname(const char *pathname, int pathlen,
961 const char *base, int baselen,
962 const char *pattern, int prefix, int patternlen,
963 unsigned flags)
964 {
965 const char *name;
966 int namelen;
967
968 /*
969 * match with FNM_PATHNAME; the pattern has base implicitly
970 * in front of it.
971 */
972 if (*pattern == '/') {
973 pattern++;
974 patternlen--;
975 prefix--;
976 }
977
978 /*
979 * baselen does not count the trailing slash. base[] may or
980 * may not end with a trailing slash though.
981 */
982 if (pathlen < baselen + 1 ||
983 (baselen && pathname[baselen] != '/') ||
984 fspathncmp(pathname, base, baselen))
985 return 0;
986
987 namelen = baselen ? pathlen - baselen - 1 : pathlen;
988 name = pathname + pathlen - namelen;
989
990 if (prefix) {
991 /*
992 * if the non-wildcard part is longer than the
993 * remaining pathname, surely it cannot match.
994 */
995 if (prefix > namelen)
996 return 0;
997
998 if (fspathncmp(pattern, name, prefix))
999 return 0;
1000 pattern += prefix;
1001 patternlen -= prefix;
1002 name += prefix;
1003 namelen -= prefix;
1004
1005 /*
1006 * If the whole pattern did not have a wildcard,
1007 * then our prefix match is all we need; we
1008 * do not need to call fnmatch at all.
1009 */
1010 if (!patternlen && !namelen)
1011 return 1;
1012 }
1013
1014 return fnmatch_icase_mem(pattern, patternlen,
1015 name, namelen,
1016 WM_PATHNAME) == 0;
1017 }
1018
1019 /*
1020 * Scan the given exclude list in reverse to see whether pathname
1021 * should be ignored. The first match (i.e. the last on the list), if
1022 * any, determines the fate. Returns the exclude_list element which
1023 * matched, or NULL for undecided.
1024 */
1025 static struct exclude *last_exclude_matching_from_list(const char *pathname,
1026 int pathlen,
1027 const char *basename,
1028 int *dtype,
1029 struct exclude_list *el,
1030 struct index_state *istate)
1031 {
1032 struct exclude *exc = NULL; /* undecided */
1033 int i;
1034
1035 if (!el->nr)
1036 return NULL; /* undefined */
1037
1038 for (i = el->nr - 1; 0 <= i; i--) {
1039 struct exclude *x = el->excludes[i];
1040 const char *exclude = x->pattern;
1041 int prefix = x->nowildcardlen;
1042
1043 if (x->flags & EXC_FLAG_MUSTBEDIR) {
1044 if (*dtype == DT_UNKNOWN)
1045 *dtype = get_dtype(NULL, istate, pathname, pathlen);
1046 if (*dtype != DT_DIR)
1047 continue;
1048 }
1049
1050 if (x->flags & EXC_FLAG_NODIR) {
1051 if (match_basename(basename,
1052 pathlen - (basename - pathname),
1053 exclude, prefix, x->patternlen,
1054 x->flags)) {
1055 exc = x;
1056 break;
1057 }
1058 continue;
1059 }
1060
1061 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1062 if (match_pathname(pathname, pathlen,
1063 x->base, x->baselen ? x->baselen - 1 : 0,
1064 exclude, prefix, x->patternlen, x->flags)) {
1065 exc = x;
1066 break;
1067 }
1068 }
1069 return exc;
1070 }
1071
1072 /*
1073 * Scan the list and let the last match determine the fate.
1074 * Return 1 for exclude, 0 for include and -1 for undecided.
1075 */
1076 int is_excluded_from_list(const char *pathname,
1077 int pathlen, const char *basename, int *dtype,
1078 struct exclude_list *el, struct index_state *istate)
1079 {
1080 struct exclude *exclude;
1081 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1082 dtype, el, istate);
1083 if (exclude)
1084 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1085 return -1; /* undecided */
1086 }
1087
1088 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1089 struct index_state *istate,
1090 const char *pathname, int pathlen, const char *basename,
1091 int *dtype_p)
1092 {
1093 int i, j;
1094 struct exclude_list_group *group;
1095 struct exclude *exclude;
1096 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1097 group = &dir->exclude_list_group[i];
1098 for (j = group->nr - 1; j >= 0; j--) {
1099 exclude = last_exclude_matching_from_list(
1100 pathname, pathlen, basename, dtype_p,
1101 &group->el[j], istate);
1102 if (exclude)
1103 return exclude;
1104 }
1105 }
1106 return NULL;
1107 }
1108
1109 /*
1110 * Loads the per-directory exclude list for the substring of base
1111 * which has a char length of baselen.
1112 */
1113 static void prep_exclude(struct dir_struct *dir,
1114 struct index_state *istate,
1115 const char *base, int baselen)
1116 {
1117 struct exclude_list_group *group;
1118 struct exclude_list *el;
1119 struct exclude_stack *stk = NULL;
1120 struct untracked_cache_dir *untracked;
1121 int current;
1122
1123 group = &dir->exclude_list_group[EXC_DIRS];
1124
1125 /*
1126 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1127 * which originate from directories not in the prefix of the
1128 * path being checked.
1129 */
1130 while ((stk = dir->exclude_stack) != NULL) {
1131 if (stk->baselen <= baselen &&
1132 !strncmp(dir->basebuf.buf, base, stk->baselen))
1133 break;
1134 el = &group->el[dir->exclude_stack->exclude_ix];
1135 dir->exclude_stack = stk->prev;
1136 dir->exclude = NULL;
1137 free((char *)el->src); /* see strbuf_detach() below */
1138 clear_exclude_list(el);
1139 free(stk);
1140 group->nr--;
1141 }
1142
1143 /* Skip traversing into sub directories if the parent is excluded */
1144 if (dir->exclude)
1145 return;
1146
1147 /*
1148 * Lazy initialization. All call sites currently just
1149 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1150 * them seems lots of work for little benefit.
1151 */
1152 if (!dir->basebuf.buf)
1153 strbuf_init(&dir->basebuf, PATH_MAX);
1154
1155 /* Read from the parent directories and push them down. */
1156 current = stk ? stk->baselen : -1;
1157 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1158 if (dir->untracked)
1159 untracked = stk ? stk->ucd : dir->untracked->root;
1160 else
1161 untracked = NULL;
1162
1163 while (current < baselen) {
1164 const char *cp;
1165 struct oid_stat oid_stat;
1166
1167 stk = xcalloc(1, sizeof(*stk));
1168 if (current < 0) {
1169 cp = base;
1170 current = 0;
1171 } else {
1172 cp = strchr(base + current + 1, '/');
1173 if (!cp)
1174 die("oops in prep_exclude");
1175 cp++;
1176 untracked =
1177 lookup_untracked(dir->untracked, untracked,
1178 base + current,
1179 cp - base - current);
1180 }
1181 stk->prev = dir->exclude_stack;
1182 stk->baselen = cp - base;
1183 stk->exclude_ix = group->nr;
1184 stk->ucd = untracked;
1185 el = add_exclude_list(dir, EXC_DIRS, NULL);
1186 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1187 assert(stk->baselen == dir->basebuf.len);
1188
1189 /* Abort if the directory is excluded */
1190 if (stk->baselen) {
1191 int dt = DT_DIR;
1192 dir->basebuf.buf[stk->baselen - 1] = 0;
1193 dir->exclude = last_exclude_matching_from_lists(dir,
1194 istate,
1195 dir->basebuf.buf, stk->baselen - 1,
1196 dir->basebuf.buf + current, &dt);
1197 dir->basebuf.buf[stk->baselen - 1] = '/';
1198 if (dir->exclude &&
1199 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1200 dir->exclude = NULL;
1201 if (dir->exclude) {
1202 dir->exclude_stack = stk;
1203 return;
1204 }
1205 }
1206
1207 /* Try to read per-directory file */
1208 oidclr(&oid_stat.oid);
1209 oid_stat.valid = 0;
1210 if (dir->exclude_per_dir &&
1211 /*
1212 * If we know that no files have been added in
1213 * this directory (i.e. valid_cached_dir() has
1214 * been executed and set untracked->valid) ..
1215 */
1216 (!untracked || !untracked->valid ||
1217 /*
1218 * .. and .gitignore does not exist before
1219 * (i.e. null exclude_oid). Then we can skip
1220 * loading .gitignore, which would result in
1221 * ENOENT anyway.
1222 */
1223 !is_null_oid(&untracked->exclude_oid))) {
1224 /*
1225 * dir->basebuf gets reused by the traversal, but we
1226 * need fname to remain unchanged to ensure the src
1227 * member of each struct exclude correctly
1228 * back-references its source file. Other invocations
1229 * of add_exclude_list provide stable strings, so we
1230 * strbuf_detach() and free() here in the caller.
1231 */
1232 struct strbuf sb = STRBUF_INIT;
1233 strbuf_addbuf(&sb, &dir->basebuf);
1234 strbuf_addstr(&sb, dir->exclude_per_dir);
1235 el->src = strbuf_detach(&sb, NULL);
1236 add_excludes(el->src, el->src, stk->baselen, el, istate,
1237 untracked ? &oid_stat : NULL);
1238 }
1239 /*
1240 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1241 * will first be called in valid_cached_dir() then maybe many
1242 * times more in last_exclude_matching(). When the cache is
1243 * used, last_exclude_matching() will not be called and
1244 * reading .gitignore content will be a waste.
1245 *
1246 * So when it's called by valid_cached_dir() and we can get
1247 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1248 * modified on work tree), we could delay reading the
1249 * .gitignore content until we absolutely need it in
1250 * last_exclude_matching(). Be careful about ignore rule
1251 * order, though, if you do that.
1252 */
1253 if (untracked &&
1254 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1255 invalidate_gitignore(dir->untracked, untracked);
1256 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1257 }
1258 dir->exclude_stack = stk;
1259 current = stk->baselen;
1260 }
1261 strbuf_setlen(&dir->basebuf, baselen);
1262 }
1263
1264 /*
1265 * Loads the exclude lists for the directory containing pathname, then
1266 * scans all exclude lists to determine whether pathname is excluded.
1267 * Returns the exclude_list element which matched, or NULL for
1268 * undecided.
1269 */
1270 struct exclude *last_exclude_matching(struct dir_struct *dir,
1271 struct index_state *istate,
1272 const char *pathname,
1273 int *dtype_p)
1274 {
1275 int pathlen = strlen(pathname);
1276 const char *basename = strrchr(pathname, '/');
1277 basename = (basename) ? basename+1 : pathname;
1278
1279 prep_exclude(dir, istate, pathname, basename-pathname);
1280
1281 if (dir->exclude)
1282 return dir->exclude;
1283
1284 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
1285 basename, dtype_p);
1286 }
1287
1288 /*
1289 * Loads the exclude lists for the directory containing pathname, then
1290 * scans all exclude lists to determine whether pathname is excluded.
1291 * Returns 1 if true, otherwise 0.
1292 */
1293 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1294 const char *pathname, int *dtype_p)
1295 {
1296 struct exclude *exclude =
1297 last_exclude_matching(dir, istate, pathname, dtype_p);
1298 if (exclude)
1299 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1300 return 0;
1301 }
1302
1303 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1304 {
1305 struct dir_entry *ent;
1306
1307 FLEX_ALLOC_MEM(ent, name, pathname, len);
1308 ent->len = len;
1309 return ent;
1310 }
1311
1312 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1313 struct index_state *istate,
1314 const char *pathname, int len)
1315 {
1316 if (index_file_exists(istate, pathname, len, ignore_case))
1317 return NULL;
1318
1319 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1320 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1321 }
1322
1323 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1324 struct index_state *istate,
1325 const char *pathname, int len)
1326 {
1327 if (!index_name_is_other(istate, pathname, len))
1328 return NULL;
1329
1330 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1331 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1332 }
1333
1334 enum exist_status {
1335 index_nonexistent = 0,
1336 index_directory,
1337 index_gitdir
1338 };
1339
1340 /*
1341 * Do not use the alphabetically sorted index to look up
1342 * the directory name; instead, use the case insensitive
1343 * directory hash.
1344 */
1345 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1346 const char *dirname, int len)
1347 {
1348 struct cache_entry *ce;
1349
1350 if (index_dir_exists(istate, dirname, len))
1351 return index_directory;
1352
1353 ce = index_file_exists(istate, dirname, len, ignore_case);
1354 if (ce && S_ISGITLINK(ce->ce_mode))
1355 return index_gitdir;
1356
1357 return index_nonexistent;
1358 }
1359
1360 /*
1361 * The index sorts alphabetically by entry name, which
1362 * means that a gitlink sorts as '\0' at the end, while
1363 * a directory (which is defined not as an entry, but as
1364 * the files it contains) will sort with the '/' at the
1365 * end.
1366 */
1367 static enum exist_status directory_exists_in_index(struct index_state *istate,
1368 const char *dirname, int len)
1369 {
1370 int pos;
1371
1372 if (ignore_case)
1373 return directory_exists_in_index_icase(istate, dirname, len);
1374
1375 pos = index_name_pos(istate, dirname, len);
1376 if (pos < 0)
1377 pos = -pos-1;
1378 while (pos < istate->cache_nr) {
1379 const struct cache_entry *ce = istate->cache[pos++];
1380 unsigned char endchar;
1381
1382 if (strncmp(ce->name, dirname, len))
1383 break;
1384 endchar = ce->name[len];
1385 if (endchar > '/')
1386 break;
1387 if (endchar == '/')
1388 return index_directory;
1389 if (!endchar && S_ISGITLINK(ce->ce_mode))
1390 return index_gitdir;
1391 }
1392 return index_nonexistent;
1393 }
1394
1395 /*
1396 * When we find a directory when traversing the filesystem, we
1397 * have three distinct cases:
1398 *
1399 * - ignore it
1400 * - see it as a directory
1401 * - recurse into it
1402 *
1403 * and which one we choose depends on a combination of existing
1404 * git index contents and the flags passed into the directory
1405 * traversal routine.
1406 *
1407 * Case 1: If we *already* have entries in the index under that
1408 * directory name, we always recurse into the directory to see
1409 * all the files.
1410 *
1411 * Case 2: If we *already* have that directory name as a gitlink,
1412 * we always continue to see it as a gitlink, regardless of whether
1413 * there is an actual git directory there or not (it might not
1414 * be checked out as a subproject!)
1415 *
1416 * Case 3: if we didn't have it in the index previously, we
1417 * have a few sub-cases:
1418 *
1419 * (a) if "show_other_directories" is true, we show it as
1420 * just a directory, unless "hide_empty_directories" is
1421 * also true, in which case we need to check if it contains any
1422 * untracked and / or ignored files.
1423 * (b) if it looks like a git directory, and we don't have
1424 * 'no_gitlinks' set we treat it as a gitlink, and show it
1425 * as a directory.
1426 * (c) otherwise, we recurse into it.
1427 */
1428 static enum path_treatment treat_directory(struct dir_struct *dir,
1429 struct index_state *istate,
1430 struct untracked_cache_dir *untracked,
1431 const char *dirname, int len, int baselen, int exclude,
1432 const struct pathspec *pathspec)
1433 {
1434 /* The "len-1" is to strip the final '/' */
1435 switch (directory_exists_in_index(istate, dirname, len-1)) {
1436 case index_directory:
1437 return path_recurse;
1438
1439 case index_gitdir:
1440 return path_none;
1441
1442 case index_nonexistent:
1443 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1444 break;
1445 if (exclude &&
1446 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1447 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1448
1449 /*
1450 * This is an excluded directory and we are
1451 * showing ignored paths that match an exclude
1452 * pattern. (e.g. show directory as ignored
1453 * only if it matches an exclude pattern).
1454 * This path will either be 'path_excluded`
1455 * (if we are showing empty directories or if
1456 * the directory is not empty), or will be
1457 * 'path_none' (empty directory, and we are
1458 * not showing empty directories).
1459 */
1460 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1461 return path_excluded;
1462
1463 if (read_directory_recursive(dir, istate, dirname, len,
1464 untracked, 1, 1, pathspec) == path_excluded)
1465 return path_excluded;
1466
1467 return path_none;
1468 }
1469 if (!(dir->flags & DIR_NO_GITLINKS)) {
1470 struct object_id oid;
1471 if (resolve_gitlink_ref(dirname, "HEAD", &oid) == 0)
1472 return exclude ? path_excluded : path_untracked;
1473 }
1474 return path_recurse;
1475 }
1476
1477 /* This is the "show_other_directories" case */
1478
1479 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1480 return exclude ? path_excluded : path_untracked;
1481
1482 untracked = lookup_untracked(dir->untracked, untracked,
1483 dirname + baselen, len - baselen);
1484
1485 /*
1486 * If this is an excluded directory, then we only need to check if
1487 * the directory contains any files.
1488 */
1489 return read_directory_recursive(dir, istate, dirname, len,
1490 untracked, 1, exclude, pathspec);
1491 }
1492
1493 /*
1494 * This is an inexact early pruning of any recursive directory
1495 * reading - if the path cannot possibly be in the pathspec,
1496 * return true, and we'll skip it early.
1497 */
1498 static int simplify_away(const char *path, int pathlen,
1499 const struct pathspec *pathspec)
1500 {
1501 int i;
1502
1503 if (!pathspec || !pathspec->nr)
1504 return 0;
1505
1506 GUARD_PATHSPEC(pathspec,
1507 PATHSPEC_FROMTOP |
1508 PATHSPEC_MAXDEPTH |
1509 PATHSPEC_LITERAL |
1510 PATHSPEC_GLOB |
1511 PATHSPEC_ICASE |
1512 PATHSPEC_EXCLUDE |
1513 PATHSPEC_ATTR);
1514
1515 for (i = 0; i < pathspec->nr; i++) {
1516 const struct pathspec_item *item = &pathspec->items[i];
1517 int len = item->nowildcard_len;
1518
1519 if (len > pathlen)
1520 len = pathlen;
1521 if (!ps_strncmp(item, item->match, path, len))
1522 return 0;
1523 }
1524
1525 return 1;
1526 }
1527
1528 /*
1529 * This function tells us whether an excluded path matches a
1530 * list of "interesting" pathspecs. That is, whether a path matched
1531 * by any of the pathspecs could possibly be ignored by excluding
1532 * the specified path. This can happen if:
1533 *
1534 * 1. the path is mentioned explicitly in the pathspec
1535 *
1536 * 2. the path is a directory prefix of some element in the
1537 * pathspec
1538 */
1539 static int exclude_matches_pathspec(const char *path, int pathlen,
1540 const struct pathspec *pathspec)
1541 {
1542 int i;
1543
1544 if (!pathspec || !pathspec->nr)
1545 return 0;
1546
1547 GUARD_PATHSPEC(pathspec,
1548 PATHSPEC_FROMTOP |
1549 PATHSPEC_MAXDEPTH |
1550 PATHSPEC_LITERAL |
1551 PATHSPEC_GLOB |
1552 PATHSPEC_ICASE |
1553 PATHSPEC_EXCLUDE);
1554
1555 for (i = 0; i < pathspec->nr; i++) {
1556 const struct pathspec_item *item = &pathspec->items[i];
1557 int len = item->nowildcard_len;
1558
1559 if (len == pathlen &&
1560 !ps_strncmp(item, item->match, path, pathlen))
1561 return 1;
1562 if (len > pathlen &&
1563 item->match[pathlen] == '/' &&
1564 !ps_strncmp(item, item->match, path, pathlen))
1565 return 1;
1566 }
1567 return 0;
1568 }
1569
1570 static int get_index_dtype(struct index_state *istate,
1571 const char *path, int len)
1572 {
1573 int pos;
1574 const struct cache_entry *ce;
1575
1576 ce = index_file_exists(istate, path, len, 0);
1577 if (ce) {
1578 if (!ce_uptodate(ce))
1579 return DT_UNKNOWN;
1580 if (S_ISGITLINK(ce->ce_mode))
1581 return DT_DIR;
1582 /*
1583 * Nobody actually cares about the
1584 * difference between DT_LNK and DT_REG
1585 */
1586 return DT_REG;
1587 }
1588
1589 /* Try to look it up as a directory */
1590 pos = index_name_pos(istate, path, len);
1591 if (pos >= 0)
1592 return DT_UNKNOWN;
1593 pos = -pos-1;
1594 while (pos < istate->cache_nr) {
1595 ce = istate->cache[pos++];
1596 if (strncmp(ce->name, path, len))
1597 break;
1598 if (ce->name[len] > '/')
1599 break;
1600 if (ce->name[len] < '/')
1601 continue;
1602 if (!ce_uptodate(ce))
1603 break; /* continue? */
1604 return DT_DIR;
1605 }
1606 return DT_UNKNOWN;
1607 }
1608
1609 static int get_dtype(struct dirent *de, struct index_state *istate,
1610 const char *path, int len)
1611 {
1612 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1613 struct stat st;
1614
1615 if (dtype != DT_UNKNOWN)
1616 return dtype;
1617 dtype = get_index_dtype(istate, path, len);
1618 if (dtype != DT_UNKNOWN)
1619 return dtype;
1620 if (lstat(path, &st))
1621 return dtype;
1622 if (S_ISREG(st.st_mode))
1623 return DT_REG;
1624 if (S_ISDIR(st.st_mode))
1625 return DT_DIR;
1626 if (S_ISLNK(st.st_mode))
1627 return DT_LNK;
1628 return dtype;
1629 }
1630
1631 static enum path_treatment treat_one_path(struct dir_struct *dir,
1632 struct untracked_cache_dir *untracked,
1633 struct index_state *istate,
1634 struct strbuf *path,
1635 int baselen,
1636 const struct pathspec *pathspec,
1637 int dtype, struct dirent *de)
1638 {
1639 int exclude;
1640 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1641 enum path_treatment path_treatment;
1642
1643 if (dtype == DT_UNKNOWN)
1644 dtype = get_dtype(de, istate, path->buf, path->len);
1645
1646 /* Always exclude indexed files */
1647 if (dtype != DT_DIR && has_path_in_index)
1648 return path_none;
1649
1650 /*
1651 * When we are looking at a directory P in the working tree,
1652 * there are three cases:
1653 *
1654 * (1) P exists in the index. Everything inside the directory P in
1655 * the working tree needs to go when P is checked out from the
1656 * index.
1657 *
1658 * (2) P does not exist in the index, but there is P/Q in the index.
1659 * We know P will stay a directory when we check out the contents
1660 * of the index, but we do not know yet if there is a directory
1661 * P/Q in the working tree to be killed, so we need to recurse.
1662 *
1663 * (3) P does not exist in the index, and there is no P/Q in the index
1664 * to require P to be a directory, either. Only in this case, we
1665 * know that everything inside P will not be killed without
1666 * recursing.
1667 */
1668 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1669 (dtype == DT_DIR) &&
1670 !has_path_in_index &&
1671 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1672 return path_none;
1673
1674 exclude = is_excluded(dir, istate, path->buf, &dtype);
1675
1676 /*
1677 * Excluded? If we don't explicitly want to show
1678 * ignored files, ignore it
1679 */
1680 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1681 return path_excluded;
1682
1683 switch (dtype) {
1684 default:
1685 return path_none;
1686 case DT_DIR:
1687 strbuf_addch(path, '/');
1688 path_treatment = treat_directory(dir, istate, untracked,
1689 path->buf, path->len,
1690 baselen, exclude, pathspec);
1691 /*
1692 * If 1) we only want to return directories that
1693 * match an exclude pattern and 2) this directory does
1694 * not match an exclude pattern but all of its
1695 * contents are excluded, then indicate that we should
1696 * recurse into this directory (instead of marking the
1697 * directory itself as an ignored path).
1698 */
1699 if (!exclude &&
1700 path_treatment == path_excluded &&
1701 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1702 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1703 return path_recurse;
1704 return path_treatment;
1705 case DT_REG:
1706 case DT_LNK:
1707 return exclude ? path_excluded : path_untracked;
1708 }
1709 }
1710
1711 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1712 struct untracked_cache_dir *untracked,
1713 struct cached_dir *cdir,
1714 struct index_state *istate,
1715 struct strbuf *path,
1716 int baselen,
1717 const struct pathspec *pathspec)
1718 {
1719 strbuf_setlen(path, baselen);
1720 if (!cdir->ucd) {
1721 strbuf_addstr(path, cdir->file);
1722 return path_untracked;
1723 }
1724 strbuf_addstr(path, cdir->ucd->name);
1725 /* treat_one_path() does this before it calls treat_directory() */
1726 strbuf_complete(path, '/');
1727 if (cdir->ucd->check_only)
1728 /*
1729 * check_only is set as a result of treat_directory() getting
1730 * to its bottom. Verify again the same set of directories
1731 * with check_only set.
1732 */
1733 return read_directory_recursive(dir, istate, path->buf, path->len,
1734 cdir->ucd, 1, 0, pathspec);
1735 /*
1736 * We get path_recurse in the first run when
1737 * directory_exists_in_index() returns index_nonexistent. We
1738 * are sure that new changes in the index does not impact the
1739 * outcome. Return now.
1740 */
1741 return path_recurse;
1742 }
1743
1744 static enum path_treatment treat_path(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 int dtype;
1753 struct dirent *de = cdir->de;
1754
1755 if (!de)
1756 return treat_path_fast(dir, untracked, cdir, istate, path,
1757 baselen, pathspec);
1758 if (is_dot_or_dotdot(de->d_name) || !fspathcmp(de->d_name, ".git"))
1759 return path_none;
1760 strbuf_setlen(path, baselen);
1761 strbuf_addstr(path, de->d_name);
1762 if (simplify_away(path->buf, path->len, pathspec))
1763 return path_none;
1764
1765 dtype = DTYPE(de);
1766 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1767 }
1768
1769 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1770 {
1771 if (!dir)
1772 return;
1773 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1774 dir->untracked_alloc);
1775 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1776 }
1777
1778 static int valid_cached_dir(struct dir_struct *dir,
1779 struct untracked_cache_dir *untracked,
1780 struct index_state *istate,
1781 struct strbuf *path,
1782 int check_only)
1783 {
1784 struct stat st;
1785
1786 if (!untracked)
1787 return 0;
1788
1789 /*
1790 * With fsmonitor, we can trust the untracked cache's valid field.
1791 */
1792 refresh_fsmonitor(istate);
1793 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1794 if (lstat(path->len ? path->buf : ".", &st)) {
1795 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1796 return 0;
1797 }
1798 if (!untracked->valid ||
1799 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1800 fill_stat_data(&untracked->stat_data, &st);
1801 return 0;
1802 }
1803 }
1804
1805 if (untracked->check_only != !!check_only)
1806 return 0;
1807
1808 /*
1809 * prep_exclude will be called eventually on this directory,
1810 * but it's called much later in last_exclude_matching(). We
1811 * need it now to determine the validity of the cache for this
1812 * path. The next calls will be nearly no-op, the way
1813 * prep_exclude() is designed.
1814 */
1815 if (path->len && path->buf[path->len - 1] != '/') {
1816 strbuf_addch(path, '/');
1817 prep_exclude(dir, istate, path->buf, path->len);
1818 strbuf_setlen(path, path->len - 1);
1819 } else
1820 prep_exclude(dir, istate, path->buf, path->len);
1821
1822 /* hopefully prep_exclude() haven't invalidated this entry... */
1823 return untracked->valid;
1824 }
1825
1826 static int open_cached_dir(struct cached_dir *cdir,
1827 struct dir_struct *dir,
1828 struct untracked_cache_dir *untracked,
1829 struct index_state *istate,
1830 struct strbuf *path,
1831 int check_only)
1832 {
1833 const char *c_path;
1834
1835 memset(cdir, 0, sizeof(*cdir));
1836 cdir->untracked = untracked;
1837 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1838 return 0;
1839 c_path = path->len ? path->buf : ".";
1840 cdir->fdir = opendir(c_path);
1841 if (!cdir->fdir)
1842 warning_errno(_("could not open directory '%s'"), c_path);
1843 if (dir->untracked) {
1844 invalidate_directory(dir->untracked, untracked);
1845 dir->untracked->dir_opened++;
1846 }
1847 if (!cdir->fdir)
1848 return -1;
1849 return 0;
1850 }
1851
1852 static int read_cached_dir(struct cached_dir *cdir)
1853 {
1854 if (cdir->fdir) {
1855 cdir->de = readdir(cdir->fdir);
1856 if (!cdir->de)
1857 return -1;
1858 return 0;
1859 }
1860 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1861 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1862 if (!d->recurse) {
1863 cdir->nr_dirs++;
1864 continue;
1865 }
1866 cdir->ucd = d;
1867 cdir->nr_dirs++;
1868 return 0;
1869 }
1870 cdir->ucd = NULL;
1871 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1872 struct untracked_cache_dir *d = cdir->untracked;
1873 cdir->file = d->untracked[cdir->nr_files++];
1874 return 0;
1875 }
1876 return -1;
1877 }
1878
1879 static void close_cached_dir(struct cached_dir *cdir)
1880 {
1881 if (cdir->fdir)
1882 closedir(cdir->fdir);
1883 /*
1884 * We have gone through this directory and found no untracked
1885 * entries. Mark it valid.
1886 */
1887 if (cdir->untracked) {
1888 cdir->untracked->valid = 1;
1889 cdir->untracked->recurse = 1;
1890 }
1891 }
1892
1893 /*
1894 * Read a directory tree. We currently ignore anything but
1895 * directories, regular files and symlinks. That's because git
1896 * doesn't handle them at all yet. Maybe that will change some
1897 * day.
1898 *
1899 * Also, we ignore the name ".git" (even if it is not a directory).
1900 * That likely will not change.
1901 *
1902 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1903 * to signal that a file was found. This is the least significant value that
1904 * indicates that a file was encountered that does not depend on the order of
1905 * whether an untracked or exluded path was encountered first.
1906 *
1907 * Returns the most significant path_treatment value encountered in the scan.
1908 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1909 * significant path_treatment value that will be returned.
1910 */
1911
1912 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1913 struct index_state *istate, const char *base, int baselen,
1914 struct untracked_cache_dir *untracked, int check_only,
1915 int stop_at_first_file, const struct pathspec *pathspec)
1916 {
1917 struct cached_dir cdir;
1918 enum path_treatment state, subdir_state, dir_state = path_none;
1919 struct strbuf path = STRBUF_INIT;
1920
1921 strbuf_add(&path, base, baselen);
1922
1923 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1924 goto out;
1925
1926 if (untracked)
1927 untracked->check_only = !!check_only;
1928
1929 while (!read_cached_dir(&cdir)) {
1930 /* check how the file or directory should be treated */
1931 state = treat_path(dir, untracked, &cdir, istate, &path,
1932 baselen, pathspec);
1933
1934 if (state > dir_state)
1935 dir_state = state;
1936
1937 /* recurse into subdir if instructed by treat_path */
1938 if ((state == path_recurse) ||
1939 ((state == path_untracked) &&
1940 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1941 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
1942 struct untracked_cache_dir *ud;
1943 ud = lookup_untracked(dir->untracked, untracked,
1944 path.buf + baselen,
1945 path.len - baselen);
1946 subdir_state =
1947 read_directory_recursive(dir, istate, path.buf,
1948 path.len, ud,
1949 check_only, stop_at_first_file, pathspec);
1950 if (subdir_state > dir_state)
1951 dir_state = subdir_state;
1952 }
1953
1954 if (check_only) {
1955 if (stop_at_first_file) {
1956 /*
1957 * If stopping at first file, then
1958 * signal that a file was found by
1959 * returning `path_excluded`. This is
1960 * to return a consistent value
1961 * regardless of whether an ignored or
1962 * excluded file happened to be
1963 * encountered 1st.
1964 *
1965 * In current usage, the
1966 * `stop_at_first_file` is passed when
1967 * an ancestor directory has matched
1968 * an exclude pattern, so any found
1969 * files will be excluded.
1970 */
1971 if (dir_state >= path_excluded) {
1972 dir_state = path_excluded;
1973 break;
1974 }
1975 }
1976
1977 /* abort early if maximum state has been reached */
1978 if (dir_state == path_untracked) {
1979 if (cdir.fdir)
1980 add_untracked(untracked, path.buf + baselen);
1981 break;
1982 }
1983 /* skip the dir_add_* part */
1984 continue;
1985 }
1986
1987 /* add the path to the appropriate result list */
1988 switch (state) {
1989 case path_excluded:
1990 if (dir->flags & DIR_SHOW_IGNORED)
1991 dir_add_name(dir, istate, path.buf, path.len);
1992 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1993 ((dir->flags & DIR_COLLECT_IGNORED) &&
1994 exclude_matches_pathspec(path.buf, path.len,
1995 pathspec)))
1996 dir_add_ignored(dir, istate, path.buf, path.len);
1997 break;
1998
1999 case path_untracked:
2000 if (dir->flags & DIR_SHOW_IGNORED)
2001 break;
2002 dir_add_name(dir, istate, path.buf, path.len);
2003 if (cdir.fdir)
2004 add_untracked(untracked, path.buf + baselen);
2005 break;
2006
2007 default:
2008 break;
2009 }
2010 }
2011 close_cached_dir(&cdir);
2012 out:
2013 strbuf_release(&path);
2014
2015 return dir_state;
2016 }
2017
2018 int cmp_dir_entry(const void *p1, const void *p2)
2019 {
2020 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2021 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2022
2023 return name_compare(e1->name, e1->len, e2->name, e2->len);
2024 }
2025
2026 /* check if *out lexically strictly contains *in */
2027 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2028 {
2029 return (out->len < in->len) &&
2030 (out->name[out->len - 1] == '/') &&
2031 !memcmp(out->name, in->name, out->len);
2032 }
2033
2034 static int treat_leading_path(struct dir_struct *dir,
2035 struct index_state *istate,
2036 const char *path, int len,
2037 const struct pathspec *pathspec)
2038 {
2039 struct strbuf sb = STRBUF_INIT;
2040 int baselen, rc = 0;
2041 const char *cp;
2042 int old_flags = dir->flags;
2043
2044 while (len && path[len - 1] == '/')
2045 len--;
2046 if (!len)
2047 return 1;
2048 baselen = 0;
2049 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
2050 while (1) {
2051 cp = path + baselen + !!baselen;
2052 cp = memchr(cp, '/', path + len - cp);
2053 if (!cp)
2054 baselen = len;
2055 else
2056 baselen = cp - path;
2057 strbuf_setlen(&sb, 0);
2058 strbuf_add(&sb, path, baselen);
2059 if (!is_directory(sb.buf))
2060 break;
2061 if (simplify_away(sb.buf, sb.len, pathspec))
2062 break;
2063 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
2064 DT_DIR, NULL) == path_none)
2065 break; /* do not recurse into it */
2066 if (len <= baselen) {
2067 rc = 1;
2068 break; /* finished checking */
2069 }
2070 }
2071 strbuf_release(&sb);
2072 dir->flags = old_flags;
2073 return rc;
2074 }
2075
2076 static const char *get_ident_string(void)
2077 {
2078 static struct strbuf sb = STRBUF_INIT;
2079 struct utsname uts;
2080
2081 if (sb.len)
2082 return sb.buf;
2083 if (uname(&uts) < 0)
2084 die_errno(_("failed to get kernel name and information"));
2085 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2086 uts.sysname);
2087 return sb.buf;
2088 }
2089
2090 static int ident_in_untracked(const struct untracked_cache *uc)
2091 {
2092 /*
2093 * Previous git versions may have saved many NUL separated
2094 * strings in the "ident" field, but it is insane to manage
2095 * many locations, so just take care of the first one.
2096 */
2097
2098 return !strcmp(uc->ident.buf, get_ident_string());
2099 }
2100
2101 static void set_untracked_ident(struct untracked_cache *uc)
2102 {
2103 strbuf_reset(&uc->ident);
2104 strbuf_addstr(&uc->ident, get_ident_string());
2105
2106 /*
2107 * This strbuf used to contain a list of NUL separated
2108 * strings, so save NUL too for backward compatibility.
2109 */
2110 strbuf_addch(&uc->ident, 0);
2111 }
2112
2113 static void new_untracked_cache(struct index_state *istate)
2114 {
2115 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2116 strbuf_init(&uc->ident, 100);
2117 uc->exclude_per_dir = ".gitignore";
2118 /* should be the same flags used by git-status */
2119 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2120 set_untracked_ident(uc);
2121 istate->untracked = uc;
2122 istate->cache_changed |= UNTRACKED_CHANGED;
2123 }
2124
2125 void add_untracked_cache(struct index_state *istate)
2126 {
2127 if (!istate->untracked) {
2128 new_untracked_cache(istate);
2129 } else {
2130 if (!ident_in_untracked(istate->untracked)) {
2131 free_untracked_cache(istate->untracked);
2132 new_untracked_cache(istate);
2133 }
2134 }
2135 }
2136
2137 void remove_untracked_cache(struct index_state *istate)
2138 {
2139 if (istate->untracked) {
2140 free_untracked_cache(istate->untracked);
2141 istate->untracked = NULL;
2142 istate->cache_changed |= UNTRACKED_CHANGED;
2143 }
2144 }
2145
2146 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2147 int base_len,
2148 const struct pathspec *pathspec)
2149 {
2150 struct untracked_cache_dir *root;
2151 static int untracked_cache_disabled = -1;
2152
2153 if (!dir->untracked)
2154 return NULL;
2155 if (untracked_cache_disabled < 0)
2156 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2157 if (untracked_cache_disabled)
2158 return NULL;
2159
2160 /*
2161 * We only support $GIT_DIR/info/exclude and core.excludesfile
2162 * as the global ignore rule files. Any other additions
2163 * (e.g. from command line) invalidate the cache. This
2164 * condition also catches running setup_standard_excludes()
2165 * before setting dir->untracked!
2166 */
2167 if (dir->unmanaged_exclude_files)
2168 return NULL;
2169
2170 /*
2171 * Optimize for the main use case only: whole-tree git
2172 * status. More work involved in treat_leading_path() if we
2173 * use cache on just a subset of the worktree. pathspec
2174 * support could make the matter even worse.
2175 */
2176 if (base_len || (pathspec && pathspec->nr))
2177 return NULL;
2178
2179 /* Different set of flags may produce different results */
2180 if (dir->flags != dir->untracked->dir_flags ||
2181 /*
2182 * See treat_directory(), case index_nonexistent. Without
2183 * this flag, we may need to also cache .git file content
2184 * for the resolve_gitlink_ref() call, which we don't.
2185 */
2186 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2187 /* We don't support collecting ignore files */
2188 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2189 DIR_COLLECT_IGNORED)))
2190 return NULL;
2191
2192 /*
2193 * If we use .gitignore in the cache and now you change it to
2194 * .gitexclude, everything will go wrong.
2195 */
2196 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2197 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2198 return NULL;
2199
2200 /*
2201 * EXC_CMDL is not considered in the cache. If people set it,
2202 * skip the cache.
2203 */
2204 if (dir->exclude_list_group[EXC_CMDL].nr)
2205 return NULL;
2206
2207 if (!ident_in_untracked(dir->untracked)) {
2208 warning(_("untracked cache is disabled on this system or location"));
2209 return NULL;
2210 }
2211
2212 if (!dir->untracked->root) {
2213 const int len = sizeof(*dir->untracked->root);
2214 dir->untracked->root = xmalloc(len);
2215 memset(dir->untracked->root, 0, len);
2216 }
2217
2218 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2219 root = dir->untracked->root;
2220 if (!oideq(&dir->ss_info_exclude.oid,
2221 &dir->untracked->ss_info_exclude.oid)) {
2222 invalidate_gitignore(dir->untracked, root);
2223 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2224 }
2225 if (!oideq(&dir->ss_excludes_file.oid,
2226 &dir->untracked->ss_excludes_file.oid)) {
2227 invalidate_gitignore(dir->untracked, root);
2228 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2229 }
2230
2231 /* Make sure this directory is not dropped out at saving phase */
2232 root->recurse = 1;
2233 return root;
2234 }
2235
2236 int read_directory(struct dir_struct *dir, struct index_state *istate,
2237 const char *path, int len, const struct pathspec *pathspec)
2238 {
2239 struct untracked_cache_dir *untracked;
2240
2241 trace_performance_enter();
2242
2243 if (has_symlink_leading_path(path, len)) {
2244 trace_performance_leave("read directory %.*s", len, path);
2245 return dir->nr;
2246 }
2247
2248 untracked = validate_untracked_cache(dir, len, pathspec);
2249 if (!untracked)
2250 /*
2251 * make sure untracked cache code path is disabled,
2252 * e.g. prep_exclude()
2253 */
2254 dir->untracked = NULL;
2255 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2256 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2257 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2258 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2259
2260 /*
2261 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2262 * also pick up untracked contents of untracked dirs; by default
2263 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2264 */
2265 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2266 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2267 int i, j;
2268
2269 /* remove from dir->entries untracked contents of untracked dirs */
2270 for (i = j = 0; j < dir->nr; j++) {
2271 if (i &&
2272 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2273 FREE_AND_NULL(dir->entries[j]);
2274 } else {
2275 dir->entries[i++] = dir->entries[j];
2276 }
2277 }
2278
2279 dir->nr = i;
2280 }
2281
2282 trace_performance_leave("read directory %.*s", len, path);
2283 if (dir->untracked) {
2284 static int force_untracked_cache = -1;
2285 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2286
2287 if (force_untracked_cache < 0)
2288 force_untracked_cache =
2289 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2290 trace_printf_key(&trace_untracked_stats,
2291 "node creation: %u\n"
2292 "gitignore invalidation: %u\n"
2293 "directory invalidation: %u\n"
2294 "opendir: %u\n",
2295 dir->untracked->dir_created,
2296 dir->untracked->gitignore_invalidated,
2297 dir->untracked->dir_invalidated,
2298 dir->untracked->dir_opened);
2299 if (force_untracked_cache &&
2300 dir->untracked == istate->untracked &&
2301 (dir->untracked->dir_opened ||
2302 dir->untracked->gitignore_invalidated ||
2303 dir->untracked->dir_invalidated))
2304 istate->cache_changed |= UNTRACKED_CHANGED;
2305 if (dir->untracked != istate->untracked) {
2306 FREE_AND_NULL(dir->untracked);
2307 }
2308 }
2309 return dir->nr;
2310 }
2311
2312 int file_exists(const char *f)
2313 {
2314 struct stat sb;
2315 return lstat(f, &sb) == 0;
2316 }
2317
2318 static int cmp_icase(char a, char b)
2319 {
2320 if (a == b)
2321 return 0;
2322 if (ignore_case)
2323 return toupper(a) - toupper(b);
2324 return a - b;
2325 }
2326
2327 /*
2328 * Given two normalized paths (a trailing slash is ok), if subdir is
2329 * outside dir, return -1. Otherwise return the offset in subdir that
2330 * can be used as relative path to dir.
2331 */
2332 int dir_inside_of(const char *subdir, const char *dir)
2333 {
2334 int offset = 0;
2335
2336 assert(dir && subdir && *dir && *subdir);
2337
2338 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2339 dir++;
2340 subdir++;
2341 offset++;
2342 }
2343
2344 /* hel[p]/me vs hel[l]/yeah */
2345 if (*dir && *subdir)
2346 return -1;
2347
2348 if (!*subdir)
2349 return !*dir ? offset : -1; /* same dir */
2350
2351 /* foo/[b]ar vs foo/[] */
2352 if (is_dir_sep(dir[-1]))
2353 return is_dir_sep(subdir[-1]) ? offset : -1;
2354
2355 /* foo[/]bar vs foo[] */
2356 return is_dir_sep(*subdir) ? offset + 1 : -1;
2357 }
2358
2359 int is_inside_dir(const char *dir)
2360 {
2361 char *cwd;
2362 int rc;
2363
2364 if (!dir)
2365 return 0;
2366
2367 cwd = xgetcwd();
2368 rc = (dir_inside_of(cwd, dir) >= 0);
2369 free(cwd);
2370 return rc;
2371 }
2372
2373 int is_empty_dir(const char *path)
2374 {
2375 DIR *dir = opendir(path);
2376 struct dirent *e;
2377 int ret = 1;
2378
2379 if (!dir)
2380 return 0;
2381
2382 while ((e = readdir(dir)) != NULL)
2383 if (!is_dot_or_dotdot(e->d_name)) {
2384 ret = 0;
2385 break;
2386 }
2387
2388 closedir(dir);
2389 return ret;
2390 }
2391
2392 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2393 {
2394 DIR *dir;
2395 struct dirent *e;
2396 int ret = 0, original_len = path->len, len, kept_down = 0;
2397 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2398 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2399 struct object_id submodule_head;
2400
2401 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2402 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2403 /* Do not descend and nuke a nested git work tree. */
2404 if (kept_up)
2405 *kept_up = 1;
2406 return 0;
2407 }
2408
2409 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2410 dir = opendir(path->buf);
2411 if (!dir) {
2412 if (errno == ENOENT)
2413 return keep_toplevel ? -1 : 0;
2414 else if (errno == EACCES && !keep_toplevel)
2415 /*
2416 * An empty dir could be removable even if it
2417 * is unreadable:
2418 */
2419 return rmdir(path->buf);
2420 else
2421 return -1;
2422 }
2423 strbuf_complete(path, '/');
2424
2425 len = path->len;
2426 while ((e = readdir(dir)) != NULL) {
2427 struct stat st;
2428 if (is_dot_or_dotdot(e->d_name))
2429 continue;
2430
2431 strbuf_setlen(path, len);
2432 strbuf_addstr(path, e->d_name);
2433 if (lstat(path->buf, &st)) {
2434 if (errno == ENOENT)
2435 /*
2436 * file disappeared, which is what we
2437 * wanted anyway
2438 */
2439 continue;
2440 /* fall thru */
2441 } else if (S_ISDIR(st.st_mode)) {
2442 if (!remove_dir_recurse(path, flag, &kept_down))
2443 continue; /* happy */
2444 } else if (!only_empty &&
2445 (!unlink(path->buf) || errno == ENOENT)) {
2446 continue; /* happy, too */
2447 }
2448
2449 /* path too long, stat fails, or non-directory still exists */
2450 ret = -1;
2451 break;
2452 }
2453 closedir(dir);
2454
2455 strbuf_setlen(path, original_len);
2456 if (!ret && !keep_toplevel && !kept_down)
2457 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2458 else if (kept_up)
2459 /*
2460 * report the uplevel that it is not an error that we
2461 * did not rmdir() our directory.
2462 */
2463 *kept_up = !ret;
2464 return ret;
2465 }
2466
2467 int remove_dir_recursively(struct strbuf *path, int flag)
2468 {
2469 return remove_dir_recurse(path, flag, NULL);
2470 }
2471
2472 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2473
2474 void setup_standard_excludes(struct dir_struct *dir)
2475 {
2476 dir->exclude_per_dir = ".gitignore";
2477
2478 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2479 if (!excludes_file)
2480 excludes_file = xdg_config_home("ignore");
2481 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2482 add_excludes_from_file_1(dir, excludes_file,
2483 dir->untracked ? &dir->ss_excludes_file : NULL);
2484
2485 /* per repository user preference */
2486 if (startup_info->have_repository) {
2487 const char *path = git_path_info_exclude();
2488 if (!access_or_warn(path, R_OK, 0))
2489 add_excludes_from_file_1(dir, path,
2490 dir->untracked ? &dir->ss_info_exclude : NULL);
2491 }
2492 }
2493
2494 int remove_path(const char *name)
2495 {
2496 char *slash;
2497
2498 if (unlink(name) && !is_missing_file_error(errno))
2499 return -1;
2500
2501 slash = strrchr(name, '/');
2502 if (slash) {
2503 char *dirs = xstrdup(name);
2504 slash = dirs + (slash - name);
2505 do {
2506 *slash = '\0';
2507 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2508 free(dirs);
2509 }
2510 return 0;
2511 }
2512
2513 /*
2514 * Frees memory within dir which was allocated for exclude lists and
2515 * the exclude_stack. Does not free dir itself.
2516 */
2517 void clear_directory(struct dir_struct *dir)
2518 {
2519 int i, j;
2520 struct exclude_list_group *group;
2521 struct exclude_list *el;
2522 struct exclude_stack *stk;
2523
2524 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2525 group = &dir->exclude_list_group[i];
2526 for (j = 0; j < group->nr; j++) {
2527 el = &group->el[j];
2528 if (i == EXC_DIRS)
2529 free((char *)el->src);
2530 clear_exclude_list(el);
2531 }
2532 free(group->el);
2533 }
2534
2535 stk = dir->exclude_stack;
2536 while (stk) {
2537 struct exclude_stack *prev = stk->prev;
2538 free(stk);
2539 stk = prev;
2540 }
2541 strbuf_release(&dir->basebuf);
2542 }
2543
2544 struct ondisk_untracked_cache {
2545 struct stat_data info_exclude_stat;
2546 struct stat_data excludes_file_stat;
2547 uint32_t dir_flags;
2548 unsigned char info_exclude_sha1[20];
2549 unsigned char excludes_file_sha1[20];
2550 char exclude_per_dir[FLEX_ARRAY];
2551 };
2552
2553 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2554 #define ouc_size(len) (ouc_offset(exclude_per_dir) + len + 1)
2555
2556 struct write_data {
2557 int index; /* number of written untracked_cache_dir */
2558 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2559 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2560 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2561 struct strbuf out;
2562 struct strbuf sb_stat;
2563 struct strbuf sb_sha1;
2564 };
2565
2566 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2567 {
2568 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2569 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2570 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2571 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2572 to->sd_dev = htonl(from->sd_dev);
2573 to->sd_ino = htonl(from->sd_ino);
2574 to->sd_uid = htonl(from->sd_uid);
2575 to->sd_gid = htonl(from->sd_gid);
2576 to->sd_size = htonl(from->sd_size);
2577 }
2578
2579 static void write_one_dir(struct untracked_cache_dir *untracked,
2580 struct write_data *wd)
2581 {
2582 struct stat_data stat_data;
2583 struct strbuf *out = &wd->out;
2584 unsigned char intbuf[16];
2585 unsigned int intlen, value;
2586 int i = wd->index++;
2587
2588 /*
2589 * untracked_nr should be reset whenever valid is clear, but
2590 * for safety..
2591 */
2592 if (!untracked->valid) {
2593 untracked->untracked_nr = 0;
2594 untracked->check_only = 0;
2595 }
2596
2597 if (untracked->check_only)
2598 ewah_set(wd->check_only, i);
2599 if (untracked->valid) {
2600 ewah_set(wd->valid, i);
2601 stat_data_to_disk(&stat_data, &untracked->stat_data);
2602 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2603 }
2604 if (!is_null_oid(&untracked->exclude_oid)) {
2605 ewah_set(wd->sha1_valid, i);
2606 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
2607 the_hash_algo->rawsz);
2608 }
2609
2610 intlen = encode_varint(untracked->untracked_nr, intbuf);
2611 strbuf_add(out, intbuf, intlen);
2612
2613 /* skip non-recurse directories */
2614 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2615 if (untracked->dirs[i]->recurse)
2616 value++;
2617 intlen = encode_varint(value, intbuf);
2618 strbuf_add(out, intbuf, intlen);
2619
2620 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2621
2622 for (i = 0; i < untracked->untracked_nr; i++)
2623 strbuf_add(out, untracked->untracked[i],
2624 strlen(untracked->untracked[i]) + 1);
2625
2626 for (i = 0; i < untracked->dirs_nr; i++)
2627 if (untracked->dirs[i]->recurse)
2628 write_one_dir(untracked->dirs[i], wd);
2629 }
2630
2631 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2632 {
2633 struct ondisk_untracked_cache *ouc;
2634 struct write_data wd;
2635 unsigned char varbuf[16];
2636 int varint_len;
2637 size_t len = strlen(untracked->exclude_per_dir);
2638
2639 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
2640 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2641 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2642 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.oid.hash);
2643 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.oid.hash);
2644 ouc->dir_flags = htonl(untracked->dir_flags);
2645
2646 varint_len = encode_varint(untracked->ident.len, varbuf);
2647 strbuf_add(out, varbuf, varint_len);
2648 strbuf_addbuf(out, &untracked->ident);
2649
2650 strbuf_add(out, ouc, ouc_size(len));
2651 FREE_AND_NULL(ouc);
2652
2653 if (!untracked->root) {
2654 varint_len = encode_varint(0, varbuf);
2655 strbuf_add(out, varbuf, varint_len);
2656 return;
2657 }
2658
2659 wd.index = 0;
2660 wd.check_only = ewah_new();
2661 wd.valid = ewah_new();
2662 wd.sha1_valid = ewah_new();
2663 strbuf_init(&wd.out, 1024);
2664 strbuf_init(&wd.sb_stat, 1024);
2665 strbuf_init(&wd.sb_sha1, 1024);
2666 write_one_dir(untracked->root, &wd);
2667
2668 varint_len = encode_varint(wd.index, varbuf);
2669 strbuf_add(out, varbuf, varint_len);
2670 strbuf_addbuf(out, &wd.out);
2671 ewah_serialize_strbuf(wd.valid, out);
2672 ewah_serialize_strbuf(wd.check_only, out);
2673 ewah_serialize_strbuf(wd.sha1_valid, out);
2674 strbuf_addbuf(out, &wd.sb_stat);
2675 strbuf_addbuf(out, &wd.sb_sha1);
2676 strbuf_addch(out, '\0'); /* safe guard for string lists */
2677
2678 ewah_free(wd.valid);
2679 ewah_free(wd.check_only);
2680 ewah_free(wd.sha1_valid);
2681 strbuf_release(&wd.out);
2682 strbuf_release(&wd.sb_stat);
2683 strbuf_release(&wd.sb_sha1);
2684 }
2685
2686 static void free_untracked(struct untracked_cache_dir *ucd)
2687 {
2688 int i;
2689 if (!ucd)
2690 return;
2691 for (i = 0; i < ucd->dirs_nr; i++)
2692 free_untracked(ucd->dirs[i]);
2693 for (i = 0; i < ucd->untracked_nr; i++)
2694 free(ucd->untracked[i]);
2695 free(ucd->untracked);
2696 free(ucd->dirs);
2697 free(ucd);
2698 }
2699
2700 void free_untracked_cache(struct untracked_cache *uc)
2701 {
2702 if (uc)
2703 free_untracked(uc->root);
2704 free(uc);
2705 }
2706
2707 struct read_data {
2708 int index;
2709 struct untracked_cache_dir **ucd;
2710 struct ewah_bitmap *check_only;
2711 struct ewah_bitmap *valid;
2712 struct ewah_bitmap *sha1_valid;
2713 const unsigned char *data;
2714 const unsigned char *end;
2715 };
2716
2717 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2718 {
2719 memcpy(to, data, sizeof(*to));
2720 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2721 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2722 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2723 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2724 to->sd_dev = ntohl(to->sd_dev);
2725 to->sd_ino = ntohl(to->sd_ino);
2726 to->sd_uid = ntohl(to->sd_uid);
2727 to->sd_gid = ntohl(to->sd_gid);
2728 to->sd_size = ntohl(to->sd_size);
2729 }
2730
2731 static int read_one_dir(struct untracked_cache_dir **untracked_,
2732 struct read_data *rd)
2733 {
2734 struct untracked_cache_dir ud, *untracked;
2735 const unsigned char *data = rd->data, *end = rd->end;
2736 const unsigned char *eos;
2737 unsigned int value;
2738 int i;
2739
2740 memset(&ud, 0, sizeof(ud));
2741
2742 value = decode_varint(&data);
2743 if (data > end)
2744 return -1;
2745 ud.recurse = 1;
2746 ud.untracked_alloc = value;
2747 ud.untracked_nr = value;
2748 if (ud.untracked_nr)
2749 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2750
2751 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
2752 if (data > end)
2753 return -1;
2754 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2755
2756 eos = memchr(data, '\0', end - data);
2757 if (!eos || eos == end)
2758 return -1;
2759
2760 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
2761 memcpy(untracked, &ud, sizeof(ud));
2762 memcpy(untracked->name, data, eos - data + 1);
2763 data = eos + 1;
2764
2765 for (i = 0; i < untracked->untracked_nr; i++) {
2766 eos = memchr(data, '\0', end - data);
2767 if (!eos || eos == end)
2768 return -1;
2769 untracked->untracked[i] = xmemdupz(data, eos - data);
2770 data = eos + 1;
2771 }
2772
2773 rd->ucd[rd->index++] = untracked;
2774 rd->data = data;
2775
2776 for (i = 0; i < untracked->dirs_nr; i++) {
2777 if (read_one_dir(untracked->dirs + i, rd) < 0)
2778 return -1;
2779 }
2780 return 0;
2781 }
2782
2783 static void set_check_only(size_t pos, void *cb)
2784 {
2785 struct read_data *rd = cb;
2786 struct untracked_cache_dir *ud = rd->ucd[pos];
2787 ud->check_only = 1;
2788 }
2789
2790 static void read_stat(size_t pos, void *cb)
2791 {
2792 struct read_data *rd = cb;
2793 struct untracked_cache_dir *ud = rd->ucd[pos];
2794 if (rd->data + sizeof(struct stat_data) > rd->end) {
2795 rd->data = rd->end + 1;
2796 return;
2797 }
2798 stat_data_from_disk(&ud->stat_data, rd->data);
2799 rd->data += sizeof(struct stat_data);
2800 ud->valid = 1;
2801 }
2802
2803 static void read_oid(size_t pos, void *cb)
2804 {
2805 struct read_data *rd = cb;
2806 struct untracked_cache_dir *ud = rd->ucd[pos];
2807 if (rd->data + the_hash_algo->rawsz > rd->end) {
2808 rd->data = rd->end + 1;
2809 return;
2810 }
2811 hashcpy(ud->exclude_oid.hash, rd->data);
2812 rd->data += the_hash_algo->rawsz;
2813 }
2814
2815 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
2816 const unsigned char *sha1)
2817 {
2818 stat_data_from_disk(&oid_stat->stat, data);
2819 hashcpy(oid_stat->oid.hash, sha1);
2820 oid_stat->valid = 1;
2821 }
2822
2823 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2824 {
2825 struct untracked_cache *uc;
2826 struct read_data rd;
2827 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2828 const char *ident;
2829 int ident_len;
2830 ssize_t len;
2831 const char *exclude_per_dir;
2832
2833 if (sz <= 1 || end[-1] != '\0')
2834 return NULL;
2835 end--;
2836
2837 ident_len = decode_varint(&next);
2838 if (next + ident_len > end)
2839 return NULL;
2840 ident = (const char *)next;
2841 next += ident_len;
2842
2843 if (next + ouc_size(0) > end)
2844 return NULL;
2845
2846 uc = xcalloc(1, sizeof(*uc));
2847 strbuf_init(&uc->ident, ident_len);
2848 strbuf_add(&uc->ident, ident, ident_len);
2849 load_oid_stat(&uc->ss_info_exclude,
2850 next + ouc_offset(info_exclude_stat),
2851 next + ouc_offset(info_exclude_sha1));
2852 load_oid_stat(&uc->ss_excludes_file,
2853 next + ouc_offset(excludes_file_stat),
2854 next + ouc_offset(excludes_file_sha1));
2855 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2856 exclude_per_dir = (const char *)next + ouc_offset(exclude_per_dir);
2857 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2858 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2859 next += ouc_size(strlen(exclude_per_dir));
2860 if (next >= end)
2861 goto done2;
2862
2863 len = decode_varint(&next);
2864 if (next > end || len == 0)
2865 goto done2;
2866
2867 rd.valid = ewah_new();
2868 rd.check_only = ewah_new();
2869 rd.sha1_valid = ewah_new();
2870 rd.data = next;
2871 rd.end = end;
2872 rd.index = 0;
2873 ALLOC_ARRAY(rd.ucd, len);
2874
2875 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2876 goto done;
2877
2878 next = rd.data;
2879 len = ewah_read_mmap(rd.valid, next, end - next);
2880 if (len < 0)
2881 goto done;
2882
2883 next += len;
2884 len = ewah_read_mmap(rd.check_only, next, end - next);
2885 if (len < 0)
2886 goto done;
2887
2888 next += len;
2889 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2890 if (len < 0)
2891 goto done;
2892
2893 ewah_each_bit(rd.check_only, set_check_only, &rd);
2894 rd.data = next + len;
2895 ewah_each_bit(rd.valid, read_stat, &rd);
2896 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
2897 next = rd.data;
2898
2899 done:
2900 free(rd.ucd);
2901 ewah_free(rd.valid);
2902 ewah_free(rd.check_only);
2903 ewah_free(rd.sha1_valid);
2904 done2:
2905 if (next != end) {
2906 free_untracked_cache(uc);
2907 uc = NULL;
2908 }
2909 return uc;
2910 }
2911
2912 static void invalidate_one_directory(struct untracked_cache *uc,
2913 struct untracked_cache_dir *ucd)
2914 {
2915 uc->dir_invalidated++;
2916 ucd->valid = 0;
2917 ucd->untracked_nr = 0;
2918 }
2919
2920 /*
2921 * Normally when an entry is added or removed from a directory,
2922 * invalidating that directory is enough. No need to touch its
2923 * ancestors. When a directory is shown as "foo/bar/" in git-status
2924 * however, deleting or adding an entry may have cascading effect.
2925 *
2926 * Say the "foo/bar/file" has become untracked, we need to tell the
2927 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2928 * directory any more (because "bar" is managed by foo as an untracked
2929 * "file").
2930 *
2931 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2932 * was the last untracked entry in the entire "foo", we should show
2933 * "foo/" instead. Which means we have to invalidate past "bar" up to
2934 * "foo".
2935 *
2936 * This function traverses all directories from root to leaf. If there
2937 * is a chance of one of the above cases happening, we invalidate back
2938 * to root. Otherwise we just invalidate the leaf. There may be a more
2939 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2940 * detect these cases and avoid unnecessary invalidation, for example,
2941 * checking for the untracked entry named "bar/" in "foo", but for now
2942 * stick to something safe and simple.
2943 */
2944 static int invalidate_one_component(struct untracked_cache *uc,
2945 struct untracked_cache_dir *dir,
2946 const char *path, int len)
2947 {
2948 const char *rest = strchr(path, '/');
2949
2950 if (rest) {
2951 int component_len = rest - path;
2952 struct untracked_cache_dir *d =
2953 lookup_untracked(uc, dir, path, component_len);
2954 int ret =
2955 invalidate_one_component(uc, d, rest + 1,
2956 len - (component_len + 1));
2957 if (ret)
2958 invalidate_one_directory(uc, dir);
2959 return ret;
2960 }
2961
2962 invalidate_one_directory(uc, dir);
2963 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2964 }
2965
2966 void untracked_cache_invalidate_path(struct index_state *istate,
2967 const char *path, int safe_path)
2968 {
2969 if (!istate->untracked || !istate->untracked->root)
2970 return;
2971 if (!safe_path && !verify_path(path, 0))
2972 return;
2973 invalidate_one_component(istate->untracked, istate->untracked->root,
2974 path, strlen(path));
2975 }
2976
2977 void untracked_cache_remove_from_index(struct index_state *istate,
2978 const char *path)
2979 {
2980 untracked_cache_invalidate_path(istate, path, 1);
2981 }
2982
2983 void untracked_cache_add_to_index(struct index_state *istate,
2984 const char *path)
2985 {
2986 untracked_cache_invalidate_path(istate, path, 1);
2987 }
2988
2989 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
2990 const char *sub_gitdir)
2991 {
2992 int i;
2993 struct repository subrepo;
2994 struct strbuf sub_wt = STRBUF_INIT;
2995 struct strbuf sub_gd = STRBUF_INIT;
2996
2997 const struct submodule *sub;
2998
2999 /* If the submodule has no working tree, we can ignore it. */
3000 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3001 return;
3002
3003 if (repo_read_index(&subrepo) < 0)
3004 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3005
3006 for (i = 0; i < subrepo.index->cache_nr; i++) {
3007 const struct cache_entry *ce = subrepo.index->cache[i];
3008
3009 if (!S_ISGITLINK(ce->ce_mode))
3010 continue;
3011
3012 while (i + 1 < subrepo.index->cache_nr &&
3013 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3014 /*
3015 * Skip entries with the same name in different stages
3016 * to make sure an entry is returned only once.
3017 */
3018 i++;
3019
3020 sub = submodule_from_path(&subrepo, &null_oid, ce->name);
3021 if (!sub || !is_submodule_active(&subrepo, ce->name))
3022 /* .gitmodules broken or inactive sub */
3023 continue;
3024
3025 strbuf_reset(&sub_wt);
3026 strbuf_reset(&sub_gd);
3027 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3028 strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name);
3029
3030 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3031 }
3032 strbuf_release(&sub_wt);
3033 strbuf_release(&sub_gd);
3034 repo_clear(&subrepo);
3035 }
3036
3037 void connect_work_tree_and_git_dir(const char *work_tree_,
3038 const char *git_dir_,
3039 int recurse_into_nested)
3040 {
3041 struct strbuf gitfile_sb = STRBUF_INIT;
3042 struct strbuf cfg_sb = STRBUF_INIT;
3043 struct strbuf rel_path = STRBUF_INIT;
3044 char *git_dir, *work_tree;
3045
3046 /* Prepare .git file */
3047 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3048 if (safe_create_leading_directories_const(gitfile_sb.buf))
3049 die(_("could not create directories for %s"), gitfile_sb.buf);
3050
3051 /* Prepare config file */
3052 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3053 if (safe_create_leading_directories_const(cfg_sb.buf))
3054 die(_("could not create directories for %s"), cfg_sb.buf);
3055
3056 git_dir = real_pathdup(git_dir_, 1);
3057 work_tree = real_pathdup(work_tree_, 1);
3058
3059 /* Write .git file */
3060 write_file(gitfile_sb.buf, "gitdir: %s",
3061 relative_path(git_dir, work_tree, &rel_path));
3062 /* Update core.worktree setting */
3063 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3064 relative_path(work_tree, git_dir, &rel_path));
3065
3066 strbuf_release(&gitfile_sb);
3067 strbuf_release(&cfg_sb);
3068 strbuf_release(&rel_path);
3069
3070 if (recurse_into_nested)
3071 connect_wt_gitdir_in_nested(work_tree, git_dir);
3072
3073 free(work_tree);
3074 free(git_dir);
3075 }
3076
3077 /*
3078 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3079 */
3080 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3081 {
3082 if (rename(old_git_dir, new_git_dir) < 0)
3083 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3084 old_git_dir, new_git_dir);
3085
3086 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3087 }