Merge branch 'nd/no-the-index'
[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 #define NO_THE_INDEX_COMPATIBILITY_MACROS
11 #include "cache.h"
12 #include "config.h"
13 #include "dir.h"
14 #include "object-store.h"
15 #include "attr.h"
16 #include "refs.h"
17 #include "wildmatch.h"
18 #include "pathspec.h"
19 #include "utf8.h"
20 #include "varint.h"
21 #include "ewah/ewok.h"
22 #include "fsmonitor.h"
23 #include "submodule-config.h"
24
25 /*
26 * Tells read_directory_recursive how a file or directory should be treated.
27 * Values are ordered by significance, e.g. if a directory contains both
28 * excluded and untracked files, it is listed as untracked because
29 * path_untracked > path_excluded.
30 */
31 enum path_treatment {
32 path_none = 0,
33 path_recurse,
34 path_excluded,
35 path_untracked
36 };
37
38 /*
39 * Support data structure for our opendir/readdir/closedir wrappers
40 */
41 struct cached_dir {
42 DIR *fdir;
43 struct untracked_cache_dir *untracked;
44 int nr_files;
45 int nr_dirs;
46
47 struct dirent *de;
48 const char *file;
49 struct untracked_cache_dir *ucd;
50 };
51
52 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
53 struct index_state *istate, const char *path, int len,
54 struct untracked_cache_dir *untracked,
55 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
56 static int get_dtype(struct dirent *de, struct index_state *istate,
57 const char *path, int len);
58
59 int count_slashes(const char *s)
60 {
61 int cnt = 0;
62 while (*s)
63 if (*s++ == '/')
64 cnt++;
65 return cnt;
66 }
67
68 int fspathcmp(const char *a, const char *b)
69 {
70 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
71 }
72
73 int fspathncmp(const char *a, const char *b, size_t count)
74 {
75 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
76 }
77
78 int git_fnmatch(const struct pathspec_item *item,
79 const char *pattern, const char *string,
80 int prefix)
81 {
82 if (prefix > 0) {
83 if (ps_strncmp(item, pattern, string, prefix))
84 return WM_NOMATCH;
85 pattern += prefix;
86 string += prefix;
87 }
88 if (item->flags & PATHSPEC_ONESTAR) {
89 int pattern_len = strlen(++pattern);
90 int string_len = strlen(string);
91 return string_len < pattern_len ||
92 ps_strcmp(item, pattern,
93 string + string_len - pattern_len);
94 }
95 if (item->magic & PATHSPEC_GLOB)
96 return wildmatch(pattern, string,
97 WM_PATHNAME |
98 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
99 else
100 /* wildmatch has not learned no FNM_PATHNAME mode yet */
101 return wildmatch(pattern, string,
102 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
103 }
104
105 static int fnmatch_icase_mem(const char *pattern, int patternlen,
106 const char *string, int stringlen,
107 int flags)
108 {
109 int match_status;
110 struct strbuf pat_buf = STRBUF_INIT;
111 struct strbuf str_buf = STRBUF_INIT;
112 const char *use_pat = pattern;
113 const char *use_str = string;
114
115 if (pattern[patternlen]) {
116 strbuf_add(&pat_buf, pattern, patternlen);
117 use_pat = pat_buf.buf;
118 }
119 if (string[stringlen]) {
120 strbuf_add(&str_buf, string, stringlen);
121 use_str = str_buf.buf;
122 }
123
124 if (ignore_case)
125 flags |= WM_CASEFOLD;
126 match_status = wildmatch(use_pat, use_str, flags);
127
128 strbuf_release(&pat_buf);
129 strbuf_release(&str_buf);
130
131 return match_status;
132 }
133
134 static size_t common_prefix_len(const struct pathspec *pathspec)
135 {
136 int n;
137 size_t max = 0;
138
139 /*
140 * ":(icase)path" is treated as a pathspec full of
141 * wildcard. In other words, only prefix is considered common
142 * prefix. If the pathspec is abc/foo abc/bar, running in
143 * subdir xyz, the common prefix is still xyz, not xuz/abc as
144 * in non-:(icase).
145 */
146 GUARD_PATHSPEC(pathspec,
147 PATHSPEC_FROMTOP |
148 PATHSPEC_MAXDEPTH |
149 PATHSPEC_LITERAL |
150 PATHSPEC_GLOB |
151 PATHSPEC_ICASE |
152 PATHSPEC_EXCLUDE |
153 PATHSPEC_ATTR);
154
155 for (n = 0; n < pathspec->nr; n++) {
156 size_t i = 0, len = 0, item_len;
157 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
158 continue;
159 if (pathspec->items[n].magic & PATHSPEC_ICASE)
160 item_len = pathspec->items[n].prefix;
161 else
162 item_len = pathspec->items[n].nowildcard_len;
163 while (i < item_len && (n == 0 || i < max)) {
164 char c = pathspec->items[n].match[i];
165 if (c != pathspec->items[0].match[i])
166 break;
167 if (c == '/')
168 len = i + 1;
169 i++;
170 }
171 if (n == 0 || len < max) {
172 max = len;
173 if (!max)
174 break;
175 }
176 }
177 return max;
178 }
179
180 /*
181 * Returns a copy of the longest leading path common among all
182 * pathspecs.
183 */
184 char *common_prefix(const struct pathspec *pathspec)
185 {
186 unsigned long len = common_prefix_len(pathspec);
187
188 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
189 }
190
191 int fill_directory(struct dir_struct *dir,
192 struct index_state *istate,
193 const struct pathspec *pathspec)
194 {
195 const char *prefix;
196 size_t prefix_len;
197
198 /*
199 * Calculate common prefix for the pathspec, and
200 * use that to optimize the directory walk
201 */
202 prefix_len = common_prefix_len(pathspec);
203 prefix = prefix_len ? pathspec->items[0].match : "";
204
205 /* Read the directory and prune it */
206 read_directory(dir, istate, prefix, prefix_len, pathspec);
207
208 return prefix_len;
209 }
210
211 int within_depth(const char *name, int namelen,
212 int depth, int max_depth)
213 {
214 const char *cp = name, *cpe = name + namelen;
215
216 while (cp < cpe) {
217 if (*cp++ != '/')
218 continue;
219 depth++;
220 if (depth > max_depth)
221 return 0;
222 }
223 return 1;
224 }
225
226 /*
227 * Read the contents of the blob with the given OID into a buffer.
228 * Append a trailing LF to the end if the last line doesn't have one.
229 *
230 * Returns:
231 * -1 when the OID is invalid or unknown or does not refer to a blob.
232 * 0 when the blob is empty.
233 * 1 along with { data, size } of the (possibly augmented) buffer
234 * when successful.
235 *
236 * Optionally updates the given oid_stat with the given OID (when valid).
237 */
238 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
239 size_t *size_out, char **data_out)
240 {
241 enum object_type type;
242 unsigned long sz;
243 char *data;
244
245 *size_out = 0;
246 *data_out = NULL;
247
248 data = read_object_file(oid, &type, &sz);
249 if (!data || type != OBJ_BLOB) {
250 free(data);
251 return -1;
252 }
253
254 if (oid_stat) {
255 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
256 oidcpy(&oid_stat->oid, oid);
257 }
258
259 if (sz == 0) {
260 free(data);
261 return 0;
262 }
263
264 if (data[sz - 1] != '\n') {
265 data = xrealloc(data, st_add(sz, 1));
266 data[sz++] = '\n';
267 }
268
269 *size_out = xsize_t(sz);
270 *data_out = data;
271
272 return 1;
273 }
274
275 #define DO_MATCH_EXCLUDE (1<<0)
276 #define DO_MATCH_DIRECTORY (1<<1)
277 #define DO_MATCH_SUBMODULE (1<<2)
278
279 static int match_attrs(const struct index_state *istate,
280 const char *name, int namelen,
281 const struct pathspec_item *item)
282 {
283 int i;
284
285 git_check_attr(istate, name, item->attr_check);
286 for (i = 0; i < item->attr_match_nr; i++) {
287 const char *value;
288 int matched;
289 enum attr_match_mode match_mode;
290
291 value = item->attr_check->items[i].value;
292 match_mode = item->attr_match[i].match_mode;
293
294 if (ATTR_TRUE(value))
295 matched = (match_mode == MATCH_SET);
296 else if (ATTR_FALSE(value))
297 matched = (match_mode == MATCH_UNSET);
298 else if (ATTR_UNSET(value))
299 matched = (match_mode == MATCH_UNSPECIFIED);
300 else
301 matched = (match_mode == MATCH_VALUE &&
302 !strcmp(item->attr_match[i].value, value));
303 if (!matched)
304 return 0;
305 }
306
307 return 1;
308 }
309
310 /*
311 * Does 'match' match the given name?
312 * A match is found if
313 *
314 * (1) the 'match' string is leading directory of 'name', or
315 * (2) the 'match' string is a wildcard and matches 'name', or
316 * (3) the 'match' string is exactly the same as 'name'.
317 *
318 * and the return value tells which case it was.
319 *
320 * It returns 0 when there is no match.
321 */
322 static int match_pathspec_item(const struct index_state *istate,
323 const struct pathspec_item *item, int prefix,
324 const char *name, int namelen, unsigned flags)
325 {
326 /* name/namelen has prefix cut off by caller */
327 const char *match = item->match + prefix;
328 int matchlen = item->len - prefix;
329
330 /*
331 * The normal call pattern is:
332 * 1. prefix = common_prefix_len(ps);
333 * 2. prune something, or fill_directory
334 * 3. match_pathspec()
335 *
336 * 'prefix' at #1 may be shorter than the command's prefix and
337 * it's ok for #2 to match extra files. Those extras will be
338 * trimmed at #3.
339 *
340 * Suppose the pathspec is 'foo' and '../bar' running from
341 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
342 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
343 * user does not want XYZ/foo, only the "foo" part should be
344 * case-insensitive. We need to filter out XYZ/foo here. In
345 * other words, we do not trust the caller on comparing the
346 * prefix part when :(icase) is involved. We do exact
347 * comparison ourselves.
348 *
349 * Normally the caller (common_prefix_len() in fact) does
350 * _exact_ matching on name[-prefix+1..-1] and we do not need
351 * to check that part. Be defensive and check it anyway, in
352 * case common_prefix_len is changed, or a new caller is
353 * introduced that does not use common_prefix_len.
354 *
355 * If the penalty turns out too high when prefix is really
356 * long, maybe change it to
357 * strncmp(match, name, item->prefix - prefix)
358 */
359 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
360 strncmp(item->match, name - prefix, item->prefix))
361 return 0;
362
363 if (item->attr_match_nr && !match_attrs(istate, name, namelen, item))
364 return 0;
365
366 /* If the match was just the prefix, we matched */
367 if (!*match)
368 return MATCHED_RECURSIVELY;
369
370 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
371 if (matchlen == namelen)
372 return MATCHED_EXACTLY;
373
374 if (match[matchlen-1] == '/' || name[matchlen] == '/')
375 return MATCHED_RECURSIVELY;
376 } else if ((flags & DO_MATCH_DIRECTORY) &&
377 match[matchlen - 1] == '/' &&
378 namelen == matchlen - 1 &&
379 !ps_strncmp(item, match, name, namelen))
380 return MATCHED_EXACTLY;
381
382 if (item->nowildcard_len < item->len &&
383 !git_fnmatch(item, match, name,
384 item->nowildcard_len - prefix))
385 return MATCHED_FNMATCH;
386
387 /* Perform checks to see if "name" is a super set of the pathspec */
388 if (flags & DO_MATCH_SUBMODULE) {
389 /* name is a literal prefix of the pathspec */
390 if ((namelen < matchlen) &&
391 (match[namelen] == '/') &&
392 !ps_strncmp(item, match, name, namelen))
393 return MATCHED_RECURSIVELY;
394
395 /* name" doesn't match up to the first wild character */
396 if (item->nowildcard_len < item->len &&
397 ps_strncmp(item, match, name,
398 item->nowildcard_len - prefix))
399 return 0;
400
401 /*
402 * Here is where we would perform a wildmatch to check if
403 * "name" can be matched as a directory (or a prefix) against
404 * the pathspec. Since wildmatch doesn't have this capability
405 * at the present we have to punt and say that it is a match,
406 * potentially returning a false positive
407 * The submodules themselves will be able to perform more
408 * accurate matching to determine if the pathspec matches.
409 */
410 return MATCHED_RECURSIVELY;
411 }
412
413 return 0;
414 }
415
416 /*
417 * Given a name and a list of pathspecs, returns the nature of the
418 * closest (i.e. most specific) match of the name to any of the
419 * pathspecs.
420 *
421 * The caller typically calls this multiple times with the same
422 * pathspec and seen[] array but with different name/namelen
423 * (e.g. entries from the index) and is interested in seeing if and
424 * how each pathspec matches all the names it calls this function
425 * with. A mark is left in the seen[] array for each pathspec element
426 * indicating the closest type of match that element achieved, so if
427 * seen[n] remains zero after multiple invocations, that means the nth
428 * pathspec did not match any names, which could indicate that the
429 * user mistyped the nth pathspec.
430 */
431 static int do_match_pathspec(const struct index_state *istate,
432 const struct pathspec *ps,
433 const char *name, int namelen,
434 int prefix, char *seen,
435 unsigned flags)
436 {
437 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
438
439 GUARD_PATHSPEC(ps,
440 PATHSPEC_FROMTOP |
441 PATHSPEC_MAXDEPTH |
442 PATHSPEC_LITERAL |
443 PATHSPEC_GLOB |
444 PATHSPEC_ICASE |
445 PATHSPEC_EXCLUDE |
446 PATHSPEC_ATTR);
447
448 if (!ps->nr) {
449 if (!ps->recursive ||
450 !(ps->magic & PATHSPEC_MAXDEPTH) ||
451 ps->max_depth == -1)
452 return MATCHED_RECURSIVELY;
453
454 if (within_depth(name, namelen, 0, ps->max_depth))
455 return MATCHED_EXACTLY;
456 else
457 return 0;
458 }
459
460 name += prefix;
461 namelen -= prefix;
462
463 for (i = ps->nr - 1; i >= 0; i--) {
464 int how;
465
466 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
467 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
468 continue;
469
470 if (seen && seen[i] == MATCHED_EXACTLY)
471 continue;
472 /*
473 * Make exclude patterns optional and never report
474 * "pathspec ':(exclude)foo' matches no files"
475 */
476 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
477 seen[i] = MATCHED_FNMATCH;
478 how = match_pathspec_item(istate, ps->items+i, prefix, name,
479 namelen, flags);
480 if (ps->recursive &&
481 (ps->magic & PATHSPEC_MAXDEPTH) &&
482 ps->max_depth != -1 &&
483 how && how != MATCHED_FNMATCH) {
484 int len = ps->items[i].len;
485 if (name[len] == '/')
486 len++;
487 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
488 how = MATCHED_EXACTLY;
489 else
490 how = 0;
491 }
492 if (how) {
493 if (retval < how)
494 retval = how;
495 if (seen && seen[i] < how)
496 seen[i] = how;
497 }
498 }
499 return retval;
500 }
501
502 int match_pathspec(const struct index_state *istate,
503 const struct pathspec *ps,
504 const char *name, int namelen,
505 int prefix, char *seen, int is_dir)
506 {
507 int positive, negative;
508 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
509 positive = do_match_pathspec(istate, ps, name, namelen,
510 prefix, seen, flags);
511 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
512 return positive;
513 negative = do_match_pathspec(istate, ps, name, namelen,
514 prefix, seen,
515 flags | DO_MATCH_EXCLUDE);
516 return negative ? 0 : positive;
517 }
518
519 /**
520 * Check if a submodule is a superset of the pathspec
521 */
522 int submodule_path_match(const struct index_state *istate,
523 const struct pathspec *ps,
524 const char *submodule_name,
525 char *seen)
526 {
527 int matched = do_match_pathspec(istate, ps, submodule_name,
528 strlen(submodule_name),
529 0, seen,
530 DO_MATCH_DIRECTORY |
531 DO_MATCH_SUBMODULE);
532 return matched;
533 }
534
535 int report_path_error(const char *ps_matched,
536 const struct pathspec *pathspec,
537 const char *prefix)
538 {
539 /*
540 * Make sure all pathspec matched; otherwise it is an error.
541 */
542 int num, errors = 0;
543 for (num = 0; num < pathspec->nr; num++) {
544 int other, found_dup;
545
546 if (ps_matched[num])
547 continue;
548 /*
549 * The caller might have fed identical pathspec
550 * twice. Do not barf on such a mistake.
551 * FIXME: parse_pathspec should have eliminated
552 * duplicate pathspec.
553 */
554 for (found_dup = other = 0;
555 !found_dup && other < pathspec->nr;
556 other++) {
557 if (other == num || !ps_matched[other])
558 continue;
559 if (!strcmp(pathspec->items[other].original,
560 pathspec->items[num].original))
561 /*
562 * Ok, we have a match already.
563 */
564 found_dup = 1;
565 }
566 if (found_dup)
567 continue;
568
569 error(_("pathspec '%s' did not match any file(s) known to git"),
570 pathspec->items[num].original);
571 errors++;
572 }
573 return errors;
574 }
575
576 /*
577 * Return the length of the "simple" part of a path match limiter.
578 */
579 int simple_length(const char *match)
580 {
581 int len = -1;
582
583 for (;;) {
584 unsigned char c = *match++;
585 len++;
586 if (c == '\0' || is_glob_special(c))
587 return len;
588 }
589 }
590
591 int no_wildcard(const char *string)
592 {
593 return string[simple_length(string)] == '\0';
594 }
595
596 void parse_exclude_pattern(const char **pattern,
597 int *patternlen,
598 unsigned *flags,
599 int *nowildcardlen)
600 {
601 const char *p = *pattern;
602 size_t i, len;
603
604 *flags = 0;
605 if (*p == '!') {
606 *flags |= EXC_FLAG_NEGATIVE;
607 p++;
608 }
609 len = strlen(p);
610 if (len && p[len - 1] == '/') {
611 len--;
612 *flags |= EXC_FLAG_MUSTBEDIR;
613 }
614 for (i = 0; i < len; i++) {
615 if (p[i] == '/')
616 break;
617 }
618 if (i == len)
619 *flags |= EXC_FLAG_NODIR;
620 *nowildcardlen = simple_length(p);
621 /*
622 * we should have excluded the trailing slash from 'p' too,
623 * but that's one more allocation. Instead just make sure
624 * nowildcardlen does not exceed real patternlen
625 */
626 if (*nowildcardlen > len)
627 *nowildcardlen = len;
628 if (*p == '*' && no_wildcard(p + 1))
629 *flags |= EXC_FLAG_ENDSWITH;
630 *pattern = p;
631 *patternlen = len;
632 }
633
634 void add_exclude(const char *string, const char *base,
635 int baselen, struct exclude_list *el, int srcpos)
636 {
637 struct exclude *x;
638 int patternlen;
639 unsigned flags;
640 int nowildcardlen;
641
642 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
643 if (flags & EXC_FLAG_MUSTBEDIR) {
644 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
645 } else {
646 x = xmalloc(sizeof(*x));
647 x->pattern = string;
648 }
649 x->patternlen = patternlen;
650 x->nowildcardlen = nowildcardlen;
651 x->base = base;
652 x->baselen = baselen;
653 x->flags = flags;
654 x->srcpos = srcpos;
655 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
656 el->excludes[el->nr++] = x;
657 x->el = el;
658 }
659
660 static int read_skip_worktree_file_from_index(const struct index_state *istate,
661 const char *path,
662 size_t *size_out, char **data_out,
663 struct oid_stat *oid_stat)
664 {
665 int pos, len;
666
667 len = strlen(path);
668 pos = index_name_pos(istate, path, len);
669 if (pos < 0)
670 return -1;
671 if (!ce_skip_worktree(istate->cache[pos]))
672 return -1;
673
674 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
675 }
676
677 /*
678 * Frees memory within el which was allocated for exclude patterns and
679 * the file buffer. Does not free el itself.
680 */
681 void clear_exclude_list(struct exclude_list *el)
682 {
683 int i;
684
685 for (i = 0; i < el->nr; i++)
686 free(el->excludes[i]);
687 free(el->excludes);
688 free(el->filebuf);
689
690 memset(el, 0, sizeof(*el));
691 }
692
693 static void trim_trailing_spaces(char *buf)
694 {
695 char *p, *last_space = NULL;
696
697 for (p = buf; *p; p++)
698 switch (*p) {
699 case ' ':
700 if (!last_space)
701 last_space = p;
702 break;
703 case '\\':
704 p++;
705 if (!*p)
706 return;
707 /* fallthrough */
708 default:
709 last_space = NULL;
710 }
711
712 if (last_space)
713 *last_space = '\0';
714 }
715
716 /*
717 * Given a subdirectory name and "dir" of the current directory,
718 * search the subdir in "dir" and return it, or create a new one if it
719 * does not exist in "dir".
720 *
721 * If "name" has the trailing slash, it'll be excluded in the search.
722 */
723 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
724 struct untracked_cache_dir *dir,
725 const char *name, int len)
726 {
727 int first, last;
728 struct untracked_cache_dir *d;
729 if (!dir)
730 return NULL;
731 if (len && name[len - 1] == '/')
732 len--;
733 first = 0;
734 last = dir->dirs_nr;
735 while (last > first) {
736 int cmp, next = (last + first) >> 1;
737 d = dir->dirs[next];
738 cmp = strncmp(name, d->name, len);
739 if (!cmp && strlen(d->name) > len)
740 cmp = -1;
741 if (!cmp)
742 return d;
743 if (cmp < 0) {
744 last = next;
745 continue;
746 }
747 first = next+1;
748 }
749
750 uc->dir_created++;
751 FLEX_ALLOC_MEM(d, name, name, len);
752
753 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
754 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
755 dir->dirs_nr - first);
756 dir->dirs_nr++;
757 dir->dirs[first] = d;
758 return d;
759 }
760
761 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
762 {
763 int i;
764 dir->valid = 0;
765 dir->untracked_nr = 0;
766 for (i = 0; i < dir->dirs_nr; i++)
767 do_invalidate_gitignore(dir->dirs[i]);
768 }
769
770 static void invalidate_gitignore(struct untracked_cache *uc,
771 struct untracked_cache_dir *dir)
772 {
773 uc->gitignore_invalidated++;
774 do_invalidate_gitignore(dir);
775 }
776
777 static void invalidate_directory(struct untracked_cache *uc,
778 struct untracked_cache_dir *dir)
779 {
780 int i;
781
782 /*
783 * Invalidation increment here is just roughly correct. If
784 * untracked_nr or any of dirs[].recurse is non-zero, we
785 * should increment dir_invalidated too. But that's more
786 * expensive to do.
787 */
788 if (dir->valid)
789 uc->dir_invalidated++;
790
791 dir->valid = 0;
792 dir->untracked_nr = 0;
793 for (i = 0; i < dir->dirs_nr; i++)
794 dir->dirs[i]->recurse = 0;
795 }
796
797 static int add_excludes_from_buffer(char *buf, size_t size,
798 const char *base, int baselen,
799 struct exclude_list *el);
800
801 /*
802 * Given a file with name "fname", read it (either from disk, or from
803 * an index if 'istate' is non-null), parse it and store the
804 * exclude rules in "el".
805 *
806 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
807 * stat data from disk (only valid if add_excludes returns zero). If
808 * ss_valid is non-zero, "ss" must contain good value as input.
809 */
810 static int add_excludes(const char *fname, const char *base, int baselen,
811 struct exclude_list *el, struct index_state *istate,
812 struct oid_stat *oid_stat)
813 {
814 struct stat st;
815 int r;
816 int fd;
817 size_t size = 0;
818 char *buf;
819
820 fd = open(fname, O_RDONLY);
821 if (fd < 0 || fstat(fd, &st) < 0) {
822 if (fd < 0)
823 warn_on_fopen_errors(fname);
824 else
825 close(fd);
826 if (!istate)
827 return -1;
828 r = read_skip_worktree_file_from_index(istate, fname,
829 &size, &buf,
830 oid_stat);
831 if (r != 1)
832 return r;
833 } else {
834 size = xsize_t(st.st_size);
835 if (size == 0) {
836 if (oid_stat) {
837 fill_stat_data(&oid_stat->stat, &st);
838 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
839 oid_stat->valid = 1;
840 }
841 close(fd);
842 return 0;
843 }
844 buf = xmallocz(size);
845 if (read_in_full(fd, buf, size) != size) {
846 free(buf);
847 close(fd);
848 return -1;
849 }
850 buf[size++] = '\n';
851 close(fd);
852 if (oid_stat) {
853 int pos;
854 if (oid_stat->valid &&
855 !match_stat_data_racy(istate, &oid_stat->stat, &st))
856 ; /* no content change, ss->sha1 still good */
857 else if (istate &&
858 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
859 !ce_stage(istate->cache[pos]) &&
860 ce_uptodate(istate->cache[pos]) &&
861 !would_convert_to_git(istate, fname))
862 oidcpy(&oid_stat->oid,
863 &istate->cache[pos]->oid);
864 else
865 hash_object_file(buf, size, "blob",
866 &oid_stat->oid);
867 fill_stat_data(&oid_stat->stat, &st);
868 oid_stat->valid = 1;
869 }
870 }
871
872 add_excludes_from_buffer(buf, size, base, baselen, el);
873 return 0;
874 }
875
876 static int add_excludes_from_buffer(char *buf, size_t size,
877 const char *base, int baselen,
878 struct exclude_list *el)
879 {
880 int i, lineno = 1;
881 char *entry;
882
883 el->filebuf = buf;
884
885 if (skip_utf8_bom(&buf, size))
886 size -= buf - el->filebuf;
887
888 entry = buf;
889
890 for (i = 0; i < size; i++) {
891 if (buf[i] == '\n') {
892 if (entry != buf + i && entry[0] != '#') {
893 buf[i - (i && buf[i-1] == '\r')] = 0;
894 trim_trailing_spaces(entry);
895 add_exclude(entry, base, baselen, el, lineno);
896 }
897 lineno++;
898 entry = buf + i + 1;
899 }
900 }
901 return 0;
902 }
903
904 int add_excludes_from_file_to_list(const char *fname, const char *base,
905 int baselen, struct exclude_list *el,
906 struct index_state *istate)
907 {
908 return add_excludes(fname, base, baselen, el, istate, NULL);
909 }
910
911 int add_excludes_from_blob_to_list(
912 struct object_id *oid,
913 const char *base, int baselen,
914 struct exclude_list *el)
915 {
916 char *buf;
917 size_t size;
918 int r;
919
920 r = do_read_blob(oid, NULL, &size, &buf);
921 if (r != 1)
922 return r;
923
924 add_excludes_from_buffer(buf, size, base, baselen, el);
925 return 0;
926 }
927
928 struct exclude_list *add_exclude_list(struct dir_struct *dir,
929 int group_type, const char *src)
930 {
931 struct exclude_list *el;
932 struct exclude_list_group *group;
933
934 group = &dir->exclude_list_group[group_type];
935 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
936 el = &group->el[group->nr++];
937 memset(el, 0, sizeof(*el));
938 el->src = src;
939 return el;
940 }
941
942 /*
943 * Used to set up core.excludesfile and .git/info/exclude lists.
944 */
945 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
946 struct oid_stat *oid_stat)
947 {
948 struct exclude_list *el;
949 /*
950 * catch setup_standard_excludes() that's called before
951 * dir->untracked is assigned. That function behaves
952 * differently when dir->untracked is non-NULL.
953 */
954 if (!dir->untracked)
955 dir->unmanaged_exclude_files++;
956 el = add_exclude_list(dir, EXC_FILE, fname);
957 if (add_excludes(fname, "", 0, el, NULL, oid_stat) < 0)
958 die(_("cannot use %s as an exclude file"), fname);
959 }
960
961 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
962 {
963 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
964 add_excludes_from_file_1(dir, fname, NULL);
965 }
966
967 int match_basename(const char *basename, int basenamelen,
968 const char *pattern, int prefix, int patternlen,
969 unsigned flags)
970 {
971 if (prefix == patternlen) {
972 if (patternlen == basenamelen &&
973 !fspathncmp(pattern, basename, basenamelen))
974 return 1;
975 } else if (flags & EXC_FLAG_ENDSWITH) {
976 /* "*literal" matching against "fooliteral" */
977 if (patternlen - 1 <= basenamelen &&
978 !fspathncmp(pattern + 1,
979 basename + basenamelen - (patternlen - 1),
980 patternlen - 1))
981 return 1;
982 } else {
983 if (fnmatch_icase_mem(pattern, patternlen,
984 basename, basenamelen,
985 0) == 0)
986 return 1;
987 }
988 return 0;
989 }
990
991 int match_pathname(const char *pathname, int pathlen,
992 const char *base, int baselen,
993 const char *pattern, int prefix, int patternlen,
994 unsigned flags)
995 {
996 const char *name;
997 int namelen;
998
999 /*
1000 * match with FNM_PATHNAME; the pattern has base implicitly
1001 * in front of it.
1002 */
1003 if (*pattern == '/') {
1004 pattern++;
1005 patternlen--;
1006 prefix--;
1007 }
1008
1009 /*
1010 * baselen does not count the trailing slash. base[] may or
1011 * may not end with a trailing slash though.
1012 */
1013 if (pathlen < baselen + 1 ||
1014 (baselen && pathname[baselen] != '/') ||
1015 fspathncmp(pathname, base, baselen))
1016 return 0;
1017
1018 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1019 name = pathname + pathlen - namelen;
1020
1021 if (prefix) {
1022 /*
1023 * if the non-wildcard part is longer than the
1024 * remaining pathname, surely it cannot match.
1025 */
1026 if (prefix > namelen)
1027 return 0;
1028
1029 if (fspathncmp(pattern, name, prefix))
1030 return 0;
1031 pattern += prefix;
1032 patternlen -= prefix;
1033 name += prefix;
1034 namelen -= prefix;
1035
1036 /*
1037 * If the whole pattern did not have a wildcard,
1038 * then our prefix match is all we need; we
1039 * do not need to call fnmatch at all.
1040 */
1041 if (!patternlen && !namelen)
1042 return 1;
1043 }
1044
1045 return fnmatch_icase_mem(pattern, patternlen,
1046 name, namelen,
1047 WM_PATHNAME) == 0;
1048 }
1049
1050 /*
1051 * Scan the given exclude list in reverse to see whether pathname
1052 * should be ignored. The first match (i.e. the last on the list), if
1053 * any, determines the fate. Returns the exclude_list element which
1054 * matched, or NULL for undecided.
1055 */
1056 static struct exclude *last_exclude_matching_from_list(const char *pathname,
1057 int pathlen,
1058 const char *basename,
1059 int *dtype,
1060 struct exclude_list *el,
1061 struct index_state *istate)
1062 {
1063 struct exclude *exc = NULL; /* undecided */
1064 int i;
1065
1066 if (!el->nr)
1067 return NULL; /* undefined */
1068
1069 for (i = el->nr - 1; 0 <= i; i--) {
1070 struct exclude *x = el->excludes[i];
1071 const char *exclude = x->pattern;
1072 int prefix = x->nowildcardlen;
1073
1074 if (x->flags & EXC_FLAG_MUSTBEDIR) {
1075 if (*dtype == DT_UNKNOWN)
1076 *dtype = get_dtype(NULL, istate, pathname, pathlen);
1077 if (*dtype != DT_DIR)
1078 continue;
1079 }
1080
1081 if (x->flags & EXC_FLAG_NODIR) {
1082 if (match_basename(basename,
1083 pathlen - (basename - pathname),
1084 exclude, prefix, x->patternlen,
1085 x->flags)) {
1086 exc = x;
1087 break;
1088 }
1089 continue;
1090 }
1091
1092 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1093 if (match_pathname(pathname, pathlen,
1094 x->base, x->baselen ? x->baselen - 1 : 0,
1095 exclude, prefix, x->patternlen, x->flags)) {
1096 exc = x;
1097 break;
1098 }
1099 }
1100 return exc;
1101 }
1102
1103 /*
1104 * Scan the list and let the last match determine the fate.
1105 * Return 1 for exclude, 0 for include and -1 for undecided.
1106 */
1107 int is_excluded_from_list(const char *pathname,
1108 int pathlen, const char *basename, int *dtype,
1109 struct exclude_list *el, struct index_state *istate)
1110 {
1111 struct exclude *exclude;
1112 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1113 dtype, el, istate);
1114 if (exclude)
1115 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1116 return -1; /* undecided */
1117 }
1118
1119 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1120 struct index_state *istate,
1121 const char *pathname, int pathlen, const char *basename,
1122 int *dtype_p)
1123 {
1124 int i, j;
1125 struct exclude_list_group *group;
1126 struct exclude *exclude;
1127 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1128 group = &dir->exclude_list_group[i];
1129 for (j = group->nr - 1; j >= 0; j--) {
1130 exclude = last_exclude_matching_from_list(
1131 pathname, pathlen, basename, dtype_p,
1132 &group->el[j], istate);
1133 if (exclude)
1134 return exclude;
1135 }
1136 }
1137 return NULL;
1138 }
1139
1140 /*
1141 * Loads the per-directory exclude list for the substring of base
1142 * which has a char length of baselen.
1143 */
1144 static void prep_exclude(struct dir_struct *dir,
1145 struct index_state *istate,
1146 const char *base, int baselen)
1147 {
1148 struct exclude_list_group *group;
1149 struct exclude_list *el;
1150 struct exclude_stack *stk = NULL;
1151 struct untracked_cache_dir *untracked;
1152 int current;
1153
1154 group = &dir->exclude_list_group[EXC_DIRS];
1155
1156 /*
1157 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1158 * which originate from directories not in the prefix of the
1159 * path being checked.
1160 */
1161 while ((stk = dir->exclude_stack) != NULL) {
1162 if (stk->baselen <= baselen &&
1163 !strncmp(dir->basebuf.buf, base, stk->baselen))
1164 break;
1165 el = &group->el[dir->exclude_stack->exclude_ix];
1166 dir->exclude_stack = stk->prev;
1167 dir->exclude = NULL;
1168 free((char *)el->src); /* see strbuf_detach() below */
1169 clear_exclude_list(el);
1170 free(stk);
1171 group->nr--;
1172 }
1173
1174 /* Skip traversing into sub directories if the parent is excluded */
1175 if (dir->exclude)
1176 return;
1177
1178 /*
1179 * Lazy initialization. All call sites currently just
1180 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1181 * them seems lots of work for little benefit.
1182 */
1183 if (!dir->basebuf.buf)
1184 strbuf_init(&dir->basebuf, PATH_MAX);
1185
1186 /* Read from the parent directories and push them down. */
1187 current = stk ? stk->baselen : -1;
1188 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1189 if (dir->untracked)
1190 untracked = stk ? stk->ucd : dir->untracked->root;
1191 else
1192 untracked = NULL;
1193
1194 while (current < baselen) {
1195 const char *cp;
1196 struct oid_stat oid_stat;
1197
1198 stk = xcalloc(1, sizeof(*stk));
1199 if (current < 0) {
1200 cp = base;
1201 current = 0;
1202 } else {
1203 cp = strchr(base + current + 1, '/');
1204 if (!cp)
1205 die("oops in prep_exclude");
1206 cp++;
1207 untracked =
1208 lookup_untracked(dir->untracked, untracked,
1209 base + current,
1210 cp - base - current);
1211 }
1212 stk->prev = dir->exclude_stack;
1213 stk->baselen = cp - base;
1214 stk->exclude_ix = group->nr;
1215 stk->ucd = untracked;
1216 el = add_exclude_list(dir, EXC_DIRS, NULL);
1217 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1218 assert(stk->baselen == dir->basebuf.len);
1219
1220 /* Abort if the directory is excluded */
1221 if (stk->baselen) {
1222 int dt = DT_DIR;
1223 dir->basebuf.buf[stk->baselen - 1] = 0;
1224 dir->exclude = last_exclude_matching_from_lists(dir,
1225 istate,
1226 dir->basebuf.buf, stk->baselen - 1,
1227 dir->basebuf.buf + current, &dt);
1228 dir->basebuf.buf[stk->baselen - 1] = '/';
1229 if (dir->exclude &&
1230 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1231 dir->exclude = NULL;
1232 if (dir->exclude) {
1233 dir->exclude_stack = stk;
1234 return;
1235 }
1236 }
1237
1238 /* Try to read per-directory file */
1239 oidclr(&oid_stat.oid);
1240 oid_stat.valid = 0;
1241 if (dir->exclude_per_dir &&
1242 /*
1243 * If we know that no files have been added in
1244 * this directory (i.e. valid_cached_dir() has
1245 * been executed and set untracked->valid) ..
1246 */
1247 (!untracked || !untracked->valid ||
1248 /*
1249 * .. and .gitignore does not exist before
1250 * (i.e. null exclude_oid). Then we can skip
1251 * loading .gitignore, which would result in
1252 * ENOENT anyway.
1253 */
1254 !is_null_oid(&untracked->exclude_oid))) {
1255 /*
1256 * dir->basebuf gets reused by the traversal, but we
1257 * need fname to remain unchanged to ensure the src
1258 * member of each struct exclude correctly
1259 * back-references its source file. Other invocations
1260 * of add_exclude_list provide stable strings, so we
1261 * strbuf_detach() and free() here in the caller.
1262 */
1263 struct strbuf sb = STRBUF_INIT;
1264 strbuf_addbuf(&sb, &dir->basebuf);
1265 strbuf_addstr(&sb, dir->exclude_per_dir);
1266 el->src = strbuf_detach(&sb, NULL);
1267 add_excludes(el->src, el->src, stk->baselen, el, istate,
1268 untracked ? &oid_stat : NULL);
1269 }
1270 /*
1271 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1272 * will first be called in valid_cached_dir() then maybe many
1273 * times more in last_exclude_matching(). When the cache is
1274 * used, last_exclude_matching() will not be called and
1275 * reading .gitignore content will be a waste.
1276 *
1277 * So when it's called by valid_cached_dir() and we can get
1278 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1279 * modified on work tree), we could delay reading the
1280 * .gitignore content until we absolutely need it in
1281 * last_exclude_matching(). Be careful about ignore rule
1282 * order, though, if you do that.
1283 */
1284 if (untracked &&
1285 oidcmp(&oid_stat.oid, &untracked->exclude_oid)) {
1286 invalidate_gitignore(dir->untracked, untracked);
1287 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1288 }
1289 dir->exclude_stack = stk;
1290 current = stk->baselen;
1291 }
1292 strbuf_setlen(&dir->basebuf, baselen);
1293 }
1294
1295 /*
1296 * Loads the exclude lists for the directory containing pathname, then
1297 * scans all exclude lists to determine whether pathname is excluded.
1298 * Returns the exclude_list element which matched, or NULL for
1299 * undecided.
1300 */
1301 struct exclude *last_exclude_matching(struct dir_struct *dir,
1302 struct index_state *istate,
1303 const char *pathname,
1304 int *dtype_p)
1305 {
1306 int pathlen = strlen(pathname);
1307 const char *basename = strrchr(pathname, '/');
1308 basename = (basename) ? basename+1 : pathname;
1309
1310 prep_exclude(dir, istate, pathname, basename-pathname);
1311
1312 if (dir->exclude)
1313 return dir->exclude;
1314
1315 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
1316 basename, dtype_p);
1317 }
1318
1319 /*
1320 * Loads the exclude lists for the directory containing pathname, then
1321 * scans all exclude lists to determine whether pathname is excluded.
1322 * Returns 1 if true, otherwise 0.
1323 */
1324 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1325 const char *pathname, int *dtype_p)
1326 {
1327 struct exclude *exclude =
1328 last_exclude_matching(dir, istate, pathname, dtype_p);
1329 if (exclude)
1330 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1331 return 0;
1332 }
1333
1334 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1335 {
1336 struct dir_entry *ent;
1337
1338 FLEX_ALLOC_MEM(ent, name, pathname, len);
1339 ent->len = len;
1340 return ent;
1341 }
1342
1343 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1344 struct index_state *istate,
1345 const char *pathname, int len)
1346 {
1347 if (index_file_exists(istate, pathname, len, ignore_case))
1348 return NULL;
1349
1350 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1351 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1352 }
1353
1354 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1355 struct index_state *istate,
1356 const char *pathname, int len)
1357 {
1358 if (!index_name_is_other(istate, pathname, len))
1359 return NULL;
1360
1361 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1362 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1363 }
1364
1365 enum exist_status {
1366 index_nonexistent = 0,
1367 index_directory,
1368 index_gitdir
1369 };
1370
1371 /*
1372 * Do not use the alphabetically sorted index to look up
1373 * the directory name; instead, use the case insensitive
1374 * directory hash.
1375 */
1376 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1377 const char *dirname, int len)
1378 {
1379 struct cache_entry *ce;
1380
1381 if (index_dir_exists(istate, dirname, len))
1382 return index_directory;
1383
1384 ce = index_file_exists(istate, dirname, len, ignore_case);
1385 if (ce && S_ISGITLINK(ce->ce_mode))
1386 return index_gitdir;
1387
1388 return index_nonexistent;
1389 }
1390
1391 /*
1392 * The index sorts alphabetically by entry name, which
1393 * means that a gitlink sorts as '\0' at the end, while
1394 * a directory (which is defined not as an entry, but as
1395 * the files it contains) will sort with the '/' at the
1396 * end.
1397 */
1398 static enum exist_status directory_exists_in_index(struct index_state *istate,
1399 const char *dirname, int len)
1400 {
1401 int pos;
1402
1403 if (ignore_case)
1404 return directory_exists_in_index_icase(istate, dirname, len);
1405
1406 pos = index_name_pos(istate, dirname, len);
1407 if (pos < 0)
1408 pos = -pos-1;
1409 while (pos < istate->cache_nr) {
1410 const struct cache_entry *ce = istate->cache[pos++];
1411 unsigned char endchar;
1412
1413 if (strncmp(ce->name, dirname, len))
1414 break;
1415 endchar = ce->name[len];
1416 if (endchar > '/')
1417 break;
1418 if (endchar == '/')
1419 return index_directory;
1420 if (!endchar && S_ISGITLINK(ce->ce_mode))
1421 return index_gitdir;
1422 }
1423 return index_nonexistent;
1424 }
1425
1426 /*
1427 * When we find a directory when traversing the filesystem, we
1428 * have three distinct cases:
1429 *
1430 * - ignore it
1431 * - see it as a directory
1432 * - recurse into it
1433 *
1434 * and which one we choose depends on a combination of existing
1435 * git index contents and the flags passed into the directory
1436 * traversal routine.
1437 *
1438 * Case 1: If we *already* have entries in the index under that
1439 * directory name, we always recurse into the directory to see
1440 * all the files.
1441 *
1442 * Case 2: If we *already* have that directory name as a gitlink,
1443 * we always continue to see it as a gitlink, regardless of whether
1444 * there is an actual git directory there or not (it might not
1445 * be checked out as a subproject!)
1446 *
1447 * Case 3: if we didn't have it in the index previously, we
1448 * have a few sub-cases:
1449 *
1450 * (a) if "show_other_directories" is true, we show it as
1451 * just a directory, unless "hide_empty_directories" is
1452 * also true, in which case we need to check if it contains any
1453 * untracked and / or ignored files.
1454 * (b) if it looks like a git directory, and we don't have
1455 * 'no_gitlinks' set we treat it as a gitlink, and show it
1456 * as a directory.
1457 * (c) otherwise, we recurse into it.
1458 */
1459 static enum path_treatment treat_directory(struct dir_struct *dir,
1460 struct index_state *istate,
1461 struct untracked_cache_dir *untracked,
1462 const char *dirname, int len, int baselen, int exclude,
1463 const struct pathspec *pathspec)
1464 {
1465 /* The "len-1" is to strip the final '/' */
1466 switch (directory_exists_in_index(istate, dirname, len-1)) {
1467 case index_directory:
1468 return path_recurse;
1469
1470 case index_gitdir:
1471 return path_none;
1472
1473 case index_nonexistent:
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 object_id oid;
1502 if (resolve_gitlink_ref(dirname, "HEAD", &oid) == 0)
1503 return exclude ? path_excluded : path_untracked;
1504 }
1505 return path_recurse;
1506 }
1507
1508 /* This is the "show_other_directories" case */
1509
1510 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1511 return exclude ? path_excluded : path_untracked;
1512
1513 untracked = lookup_untracked(dir->untracked, untracked,
1514 dirname + baselen, len - baselen);
1515
1516 /*
1517 * If this is an excluded directory, then we only need to check if
1518 * the directory contains any files.
1519 */
1520 return read_directory_recursive(dir, istate, dirname, len,
1521 untracked, 1, exclude, pathspec);
1522 }
1523
1524 /*
1525 * This is an inexact early pruning of any recursive directory
1526 * reading - if the path cannot possibly be in the pathspec,
1527 * return true, and we'll skip it early.
1528 */
1529 static int simplify_away(const char *path, int pathlen,
1530 const struct pathspec *pathspec)
1531 {
1532 int i;
1533
1534 if (!pathspec || !pathspec->nr)
1535 return 0;
1536
1537 GUARD_PATHSPEC(pathspec,
1538 PATHSPEC_FROMTOP |
1539 PATHSPEC_MAXDEPTH |
1540 PATHSPEC_LITERAL |
1541 PATHSPEC_GLOB |
1542 PATHSPEC_ICASE |
1543 PATHSPEC_EXCLUDE |
1544 PATHSPEC_ATTR);
1545
1546 for (i = 0; i < pathspec->nr; i++) {
1547 const struct pathspec_item *item = &pathspec->items[i];
1548 int len = item->nowildcard_len;
1549
1550 if (len > pathlen)
1551 len = pathlen;
1552 if (!ps_strncmp(item, item->match, path, len))
1553 return 0;
1554 }
1555
1556 return 1;
1557 }
1558
1559 /*
1560 * This function tells us whether an excluded path matches a
1561 * list of "interesting" pathspecs. That is, whether a path matched
1562 * by any of the pathspecs could possibly be ignored by excluding
1563 * the specified path. This can happen if:
1564 *
1565 * 1. the path is mentioned explicitly in the pathspec
1566 *
1567 * 2. the path is a directory prefix of some element in the
1568 * pathspec
1569 */
1570 static int exclude_matches_pathspec(const char *path, int pathlen,
1571 const struct pathspec *pathspec)
1572 {
1573 int i;
1574
1575 if (!pathspec || !pathspec->nr)
1576 return 0;
1577
1578 GUARD_PATHSPEC(pathspec,
1579 PATHSPEC_FROMTOP |
1580 PATHSPEC_MAXDEPTH |
1581 PATHSPEC_LITERAL |
1582 PATHSPEC_GLOB |
1583 PATHSPEC_ICASE |
1584 PATHSPEC_EXCLUDE);
1585
1586 for (i = 0; i < pathspec->nr; i++) {
1587 const struct pathspec_item *item = &pathspec->items[i];
1588 int len = item->nowildcard_len;
1589
1590 if (len == pathlen &&
1591 !ps_strncmp(item, item->match, path, pathlen))
1592 return 1;
1593 if (len > pathlen &&
1594 item->match[pathlen] == '/' &&
1595 !ps_strncmp(item, item->match, path, pathlen))
1596 return 1;
1597 }
1598 return 0;
1599 }
1600
1601 static int get_index_dtype(struct index_state *istate,
1602 const char *path, int len)
1603 {
1604 int pos;
1605 const struct cache_entry *ce;
1606
1607 ce = index_file_exists(istate, path, len, 0);
1608 if (ce) {
1609 if (!ce_uptodate(ce))
1610 return DT_UNKNOWN;
1611 if (S_ISGITLINK(ce->ce_mode))
1612 return DT_DIR;
1613 /*
1614 * Nobody actually cares about the
1615 * difference between DT_LNK and DT_REG
1616 */
1617 return DT_REG;
1618 }
1619
1620 /* Try to look it up as a directory */
1621 pos = index_name_pos(istate, path, len);
1622 if (pos >= 0)
1623 return DT_UNKNOWN;
1624 pos = -pos-1;
1625 while (pos < istate->cache_nr) {
1626 ce = istate->cache[pos++];
1627 if (strncmp(ce->name, path, len))
1628 break;
1629 if (ce->name[len] > '/')
1630 break;
1631 if (ce->name[len] < '/')
1632 continue;
1633 if (!ce_uptodate(ce))
1634 break; /* continue? */
1635 return DT_DIR;
1636 }
1637 return DT_UNKNOWN;
1638 }
1639
1640 static int get_dtype(struct dirent *de, struct index_state *istate,
1641 const char *path, int len)
1642 {
1643 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1644 struct stat st;
1645
1646 if (dtype != DT_UNKNOWN)
1647 return dtype;
1648 dtype = get_index_dtype(istate, path, len);
1649 if (dtype != DT_UNKNOWN)
1650 return dtype;
1651 if (lstat(path, &st))
1652 return dtype;
1653 if (S_ISREG(st.st_mode))
1654 return DT_REG;
1655 if (S_ISDIR(st.st_mode))
1656 return DT_DIR;
1657 if (S_ISLNK(st.st_mode))
1658 return DT_LNK;
1659 return dtype;
1660 }
1661
1662 static enum path_treatment treat_one_path(struct dir_struct *dir,
1663 struct untracked_cache_dir *untracked,
1664 struct index_state *istate,
1665 struct strbuf *path,
1666 int baselen,
1667 const struct pathspec *pathspec,
1668 int dtype, struct dirent *de)
1669 {
1670 int exclude;
1671 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
1672 enum path_treatment path_treatment;
1673
1674 if (dtype == DT_UNKNOWN)
1675 dtype = get_dtype(de, istate, path->buf, path->len);
1676
1677 /* Always exclude indexed files */
1678 if (dtype != DT_DIR && has_path_in_index)
1679 return path_none;
1680
1681 /*
1682 * When we are looking at a directory P in the working tree,
1683 * there are three cases:
1684 *
1685 * (1) P exists in the index. Everything inside the directory P in
1686 * the working tree needs to go when P is checked out from the
1687 * index.
1688 *
1689 * (2) P does not exist in the index, but there is P/Q in the index.
1690 * We know P will stay a directory when we check out the contents
1691 * of the index, but we do not know yet if there is a directory
1692 * P/Q in the working tree to be killed, so we need to recurse.
1693 *
1694 * (3) P does not exist in the index, and there is no P/Q in the index
1695 * to require P to be a directory, either. Only in this case, we
1696 * know that everything inside P will not be killed without
1697 * recursing.
1698 */
1699 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1700 (dtype == DT_DIR) &&
1701 !has_path_in_index &&
1702 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
1703 return path_none;
1704
1705 exclude = is_excluded(dir, istate, path->buf, &dtype);
1706
1707 /*
1708 * Excluded? If we don't explicitly want to show
1709 * ignored files, ignore it
1710 */
1711 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1712 return path_excluded;
1713
1714 switch (dtype) {
1715 default:
1716 return path_none;
1717 case DT_DIR:
1718 strbuf_addch(path, '/');
1719 path_treatment = treat_directory(dir, istate, untracked,
1720 path->buf, path->len,
1721 baselen, exclude, pathspec);
1722 /*
1723 * If 1) we only want to return directories that
1724 * match an exclude pattern and 2) this directory does
1725 * not match an exclude pattern but all of its
1726 * contents are excluded, then indicate that we should
1727 * recurse into this directory (instead of marking the
1728 * directory itself as an ignored path).
1729 */
1730 if (!exclude &&
1731 path_treatment == path_excluded &&
1732 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1733 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1734 return path_recurse;
1735 return path_treatment;
1736 case DT_REG:
1737 case DT_LNK:
1738 return exclude ? path_excluded : path_untracked;
1739 }
1740 }
1741
1742 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1743 struct untracked_cache_dir *untracked,
1744 struct cached_dir *cdir,
1745 struct index_state *istate,
1746 struct strbuf *path,
1747 int baselen,
1748 const struct pathspec *pathspec)
1749 {
1750 strbuf_setlen(path, baselen);
1751 if (!cdir->ucd) {
1752 strbuf_addstr(path, cdir->file);
1753 return path_untracked;
1754 }
1755 strbuf_addstr(path, cdir->ucd->name);
1756 /* treat_one_path() does this before it calls treat_directory() */
1757 strbuf_complete(path, '/');
1758 if (cdir->ucd->check_only)
1759 /*
1760 * check_only is set as a result of treat_directory() getting
1761 * to its bottom. Verify again the same set of directories
1762 * with check_only set.
1763 */
1764 return read_directory_recursive(dir, istate, path->buf, path->len,
1765 cdir->ucd, 1, 0, pathspec);
1766 /*
1767 * We get path_recurse in the first run when
1768 * directory_exists_in_index() returns index_nonexistent. We
1769 * are sure that new changes in the index does not impact the
1770 * outcome. Return now.
1771 */
1772 return path_recurse;
1773 }
1774
1775 static enum path_treatment treat_path(struct dir_struct *dir,
1776 struct untracked_cache_dir *untracked,
1777 struct cached_dir *cdir,
1778 struct index_state *istate,
1779 struct strbuf *path,
1780 int baselen,
1781 const struct pathspec *pathspec)
1782 {
1783 int dtype;
1784 struct dirent *de = cdir->de;
1785
1786 if (!de)
1787 return treat_path_fast(dir, untracked, cdir, istate, path,
1788 baselen, pathspec);
1789 if (is_dot_or_dotdot(de->d_name) || !fspathcmp(de->d_name, ".git"))
1790 return path_none;
1791 strbuf_setlen(path, baselen);
1792 strbuf_addstr(path, de->d_name);
1793 if (simplify_away(path->buf, path->len, pathspec))
1794 return path_none;
1795
1796 dtype = DTYPE(de);
1797 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
1798 }
1799
1800 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1801 {
1802 if (!dir)
1803 return;
1804 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1805 dir->untracked_alloc);
1806 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1807 }
1808
1809 static int valid_cached_dir(struct dir_struct *dir,
1810 struct untracked_cache_dir *untracked,
1811 struct index_state *istate,
1812 struct strbuf *path,
1813 int check_only)
1814 {
1815 struct stat st;
1816
1817 if (!untracked)
1818 return 0;
1819
1820 /*
1821 * With fsmonitor, we can trust the untracked cache's valid field.
1822 */
1823 refresh_fsmonitor(istate);
1824 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1825 if (lstat(path->len ? path->buf : ".", &st)) {
1826 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1827 return 0;
1828 }
1829 if (!untracked->valid ||
1830 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1831 fill_stat_data(&untracked->stat_data, &st);
1832 return 0;
1833 }
1834 }
1835
1836 if (untracked->check_only != !!check_only)
1837 return 0;
1838
1839 /*
1840 * prep_exclude will be called eventually on this directory,
1841 * but it's called much later in last_exclude_matching(). We
1842 * need it now to determine the validity of the cache for this
1843 * path. The next calls will be nearly no-op, the way
1844 * prep_exclude() is designed.
1845 */
1846 if (path->len && path->buf[path->len - 1] != '/') {
1847 strbuf_addch(path, '/');
1848 prep_exclude(dir, istate, path->buf, path->len);
1849 strbuf_setlen(path, path->len - 1);
1850 } else
1851 prep_exclude(dir, istate, path->buf, path->len);
1852
1853 /* hopefully prep_exclude() haven't invalidated this entry... */
1854 return untracked->valid;
1855 }
1856
1857 static int open_cached_dir(struct cached_dir *cdir,
1858 struct dir_struct *dir,
1859 struct untracked_cache_dir *untracked,
1860 struct index_state *istate,
1861 struct strbuf *path,
1862 int check_only)
1863 {
1864 const char *c_path;
1865
1866 memset(cdir, 0, sizeof(*cdir));
1867 cdir->untracked = untracked;
1868 if (valid_cached_dir(dir, untracked, istate, path, check_only))
1869 return 0;
1870 c_path = path->len ? path->buf : ".";
1871 cdir->fdir = opendir(c_path);
1872 if (!cdir->fdir)
1873 warning_errno(_("could not open directory '%s'"), c_path);
1874 if (dir->untracked) {
1875 invalidate_directory(dir->untracked, untracked);
1876 dir->untracked->dir_opened++;
1877 }
1878 if (!cdir->fdir)
1879 return -1;
1880 return 0;
1881 }
1882
1883 static int read_cached_dir(struct cached_dir *cdir)
1884 {
1885 if (cdir->fdir) {
1886 cdir->de = readdir(cdir->fdir);
1887 if (!cdir->de)
1888 return -1;
1889 return 0;
1890 }
1891 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1892 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1893 if (!d->recurse) {
1894 cdir->nr_dirs++;
1895 continue;
1896 }
1897 cdir->ucd = d;
1898 cdir->nr_dirs++;
1899 return 0;
1900 }
1901 cdir->ucd = NULL;
1902 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1903 struct untracked_cache_dir *d = cdir->untracked;
1904 cdir->file = d->untracked[cdir->nr_files++];
1905 return 0;
1906 }
1907 return -1;
1908 }
1909
1910 static void close_cached_dir(struct cached_dir *cdir)
1911 {
1912 if (cdir->fdir)
1913 closedir(cdir->fdir);
1914 /*
1915 * We have gone through this directory and found no untracked
1916 * entries. Mark it valid.
1917 */
1918 if (cdir->untracked) {
1919 cdir->untracked->valid = 1;
1920 cdir->untracked->recurse = 1;
1921 }
1922 }
1923
1924 /*
1925 * Read a directory tree. We currently ignore anything but
1926 * directories, regular files and symlinks. That's because git
1927 * doesn't handle them at all yet. Maybe that will change some
1928 * day.
1929 *
1930 * Also, we ignore the name ".git" (even if it is not a directory).
1931 * That likely will not change.
1932 *
1933 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1934 * to signal that a file was found. This is the least significant value that
1935 * indicates that a file was encountered that does not depend on the order of
1936 * whether an untracked or exluded path was encountered first.
1937 *
1938 * Returns the most significant path_treatment value encountered in the scan.
1939 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1940 * significant path_treatment value that will be returned.
1941 */
1942
1943 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1944 struct index_state *istate, const char *base, int baselen,
1945 struct untracked_cache_dir *untracked, int check_only,
1946 int stop_at_first_file, const struct pathspec *pathspec)
1947 {
1948 struct cached_dir cdir;
1949 enum path_treatment state, subdir_state, dir_state = path_none;
1950 struct strbuf path = STRBUF_INIT;
1951
1952 strbuf_add(&path, base, baselen);
1953
1954 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1955 goto out;
1956
1957 if (untracked)
1958 untracked->check_only = !!check_only;
1959
1960 while (!read_cached_dir(&cdir)) {
1961 /* check how the file or directory should be treated */
1962 state = treat_path(dir, untracked, &cdir, istate, &path,
1963 baselen, pathspec);
1964
1965 if (state > dir_state)
1966 dir_state = state;
1967
1968 /* recurse into subdir if instructed by treat_path */
1969 if ((state == path_recurse) ||
1970 ((state == path_untracked) &&
1971 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1972 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
1973 struct untracked_cache_dir *ud;
1974 ud = lookup_untracked(dir->untracked, untracked,
1975 path.buf + baselen,
1976 path.len - baselen);
1977 subdir_state =
1978 read_directory_recursive(dir, istate, path.buf,
1979 path.len, ud,
1980 check_only, stop_at_first_file, pathspec);
1981 if (subdir_state > dir_state)
1982 dir_state = subdir_state;
1983 }
1984
1985 if (check_only) {
1986 if (stop_at_first_file) {
1987 /*
1988 * If stopping at first file, then
1989 * signal that a file was found by
1990 * returning `path_excluded`. This is
1991 * to return a consistent value
1992 * regardless of whether an ignored or
1993 * excluded file happened to be
1994 * encountered 1st.
1995 *
1996 * In current usage, the
1997 * `stop_at_first_file` is passed when
1998 * an ancestor directory has matched
1999 * an exclude pattern, so any found
2000 * files will be excluded.
2001 */
2002 if (dir_state >= path_excluded) {
2003 dir_state = path_excluded;
2004 break;
2005 }
2006 }
2007
2008 /* abort early if maximum state has been reached */
2009 if (dir_state == path_untracked) {
2010 if (cdir.fdir)
2011 add_untracked(untracked, path.buf + baselen);
2012 break;
2013 }
2014 /* skip the dir_add_* part */
2015 continue;
2016 }
2017
2018 /* add the path to the appropriate result list */
2019 switch (state) {
2020 case path_excluded:
2021 if (dir->flags & DIR_SHOW_IGNORED)
2022 dir_add_name(dir, istate, path.buf, path.len);
2023 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2024 ((dir->flags & DIR_COLLECT_IGNORED) &&
2025 exclude_matches_pathspec(path.buf, path.len,
2026 pathspec)))
2027 dir_add_ignored(dir, istate, path.buf, path.len);
2028 break;
2029
2030 case path_untracked:
2031 if (dir->flags & DIR_SHOW_IGNORED)
2032 break;
2033 dir_add_name(dir, istate, path.buf, path.len);
2034 if (cdir.fdir)
2035 add_untracked(untracked, path.buf + baselen);
2036 break;
2037
2038 default:
2039 break;
2040 }
2041 }
2042 close_cached_dir(&cdir);
2043 out:
2044 strbuf_release(&path);
2045
2046 return dir_state;
2047 }
2048
2049 int cmp_dir_entry(const void *p1, const void *p2)
2050 {
2051 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2052 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2053
2054 return name_compare(e1->name, e1->len, e2->name, e2->len);
2055 }
2056
2057 /* check if *out lexically strictly contains *in */
2058 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2059 {
2060 return (out->len < in->len) &&
2061 (out->name[out->len - 1] == '/') &&
2062 !memcmp(out->name, in->name, out->len);
2063 }
2064
2065 static int treat_leading_path(struct dir_struct *dir,
2066 struct index_state *istate,
2067 const char *path, int len,
2068 const struct pathspec *pathspec)
2069 {
2070 struct strbuf sb = STRBUF_INIT;
2071 int baselen, rc = 0;
2072 const char *cp;
2073 int old_flags = dir->flags;
2074
2075 while (len && path[len - 1] == '/')
2076 len--;
2077 if (!len)
2078 return 1;
2079 baselen = 0;
2080 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
2081 while (1) {
2082 cp = path + baselen + !!baselen;
2083 cp = memchr(cp, '/', path + len - cp);
2084 if (!cp)
2085 baselen = len;
2086 else
2087 baselen = cp - path;
2088 strbuf_setlen(&sb, 0);
2089 strbuf_add(&sb, path, baselen);
2090 if (!is_directory(sb.buf))
2091 break;
2092 if (simplify_away(sb.buf, sb.len, pathspec))
2093 break;
2094 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
2095 DT_DIR, NULL) == path_none)
2096 break; /* do not recurse into it */
2097 if (len <= baselen) {
2098 rc = 1;
2099 break; /* finished checking */
2100 }
2101 }
2102 strbuf_release(&sb);
2103 dir->flags = old_flags;
2104 return rc;
2105 }
2106
2107 static const char *get_ident_string(void)
2108 {
2109 static struct strbuf sb = STRBUF_INIT;
2110 struct utsname uts;
2111
2112 if (sb.len)
2113 return sb.buf;
2114 if (uname(&uts) < 0)
2115 die_errno(_("failed to get kernel name and information"));
2116 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2117 uts.sysname);
2118 return sb.buf;
2119 }
2120
2121 static int ident_in_untracked(const struct untracked_cache *uc)
2122 {
2123 /*
2124 * Previous git versions may have saved many NUL separated
2125 * strings in the "ident" field, but it is insane to manage
2126 * many locations, so just take care of the first one.
2127 */
2128
2129 return !strcmp(uc->ident.buf, get_ident_string());
2130 }
2131
2132 static void set_untracked_ident(struct untracked_cache *uc)
2133 {
2134 strbuf_reset(&uc->ident);
2135 strbuf_addstr(&uc->ident, get_ident_string());
2136
2137 /*
2138 * This strbuf used to contain a list of NUL separated
2139 * strings, so save NUL too for backward compatibility.
2140 */
2141 strbuf_addch(&uc->ident, 0);
2142 }
2143
2144 static void new_untracked_cache(struct index_state *istate)
2145 {
2146 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2147 strbuf_init(&uc->ident, 100);
2148 uc->exclude_per_dir = ".gitignore";
2149 /* should be the same flags used by git-status */
2150 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2151 set_untracked_ident(uc);
2152 istate->untracked = uc;
2153 istate->cache_changed |= UNTRACKED_CHANGED;
2154 }
2155
2156 void add_untracked_cache(struct index_state *istate)
2157 {
2158 if (!istate->untracked) {
2159 new_untracked_cache(istate);
2160 } else {
2161 if (!ident_in_untracked(istate->untracked)) {
2162 free_untracked_cache(istate->untracked);
2163 new_untracked_cache(istate);
2164 }
2165 }
2166 }
2167
2168 void remove_untracked_cache(struct index_state *istate)
2169 {
2170 if (istate->untracked) {
2171 free_untracked_cache(istate->untracked);
2172 istate->untracked = NULL;
2173 istate->cache_changed |= UNTRACKED_CHANGED;
2174 }
2175 }
2176
2177 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2178 int base_len,
2179 const struct pathspec *pathspec)
2180 {
2181 struct untracked_cache_dir *root;
2182 static int untracked_cache_disabled = -1;
2183
2184 if (!dir->untracked)
2185 return NULL;
2186 if (untracked_cache_disabled < 0)
2187 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2188 if (untracked_cache_disabled)
2189 return NULL;
2190
2191 /*
2192 * We only support $GIT_DIR/info/exclude and core.excludesfile
2193 * as the global ignore rule files. Any other additions
2194 * (e.g. from command line) invalidate the cache. This
2195 * condition also catches running setup_standard_excludes()
2196 * before setting dir->untracked!
2197 */
2198 if (dir->unmanaged_exclude_files)
2199 return NULL;
2200
2201 /*
2202 * Optimize for the main use case only: whole-tree git
2203 * status. More work involved in treat_leading_path() if we
2204 * use cache on just a subset of the worktree. pathspec
2205 * support could make the matter even worse.
2206 */
2207 if (base_len || (pathspec && pathspec->nr))
2208 return NULL;
2209
2210 /* Different set of flags may produce different results */
2211 if (dir->flags != dir->untracked->dir_flags ||
2212 /*
2213 * See treat_directory(), case index_nonexistent. Without
2214 * this flag, we may need to also cache .git file content
2215 * for the resolve_gitlink_ref() call, which we don't.
2216 */
2217 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2218 /* We don't support collecting ignore files */
2219 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2220 DIR_COLLECT_IGNORED)))
2221 return NULL;
2222
2223 /*
2224 * If we use .gitignore in the cache and now you change it to
2225 * .gitexclude, everything will go wrong.
2226 */
2227 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2228 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2229 return NULL;
2230
2231 /*
2232 * EXC_CMDL is not considered in the cache. If people set it,
2233 * skip the cache.
2234 */
2235 if (dir->exclude_list_group[EXC_CMDL].nr)
2236 return NULL;
2237
2238 if (!ident_in_untracked(dir->untracked)) {
2239 warning(_("untracked cache is disabled on this system or location"));
2240 return NULL;
2241 }
2242
2243 if (!dir->untracked->root) {
2244 const int len = sizeof(*dir->untracked->root);
2245 dir->untracked->root = xmalloc(len);
2246 memset(dir->untracked->root, 0, len);
2247 }
2248
2249 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2250 root = dir->untracked->root;
2251 if (oidcmp(&dir->ss_info_exclude.oid,
2252 &dir->untracked->ss_info_exclude.oid)) {
2253 invalidate_gitignore(dir->untracked, root);
2254 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2255 }
2256 if (oidcmp(&dir->ss_excludes_file.oid,
2257 &dir->untracked->ss_excludes_file.oid)) {
2258 invalidate_gitignore(dir->untracked, root);
2259 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2260 }
2261
2262 /* Make sure this directory is not dropped out at saving phase */
2263 root->recurse = 1;
2264 return root;
2265 }
2266
2267 int read_directory(struct dir_struct *dir, struct index_state *istate,
2268 const char *path, int len, const struct pathspec *pathspec)
2269 {
2270 struct untracked_cache_dir *untracked;
2271 uint64_t start = getnanotime();
2272
2273 if (has_symlink_leading_path(path, len))
2274 return dir->nr;
2275
2276 untracked = validate_untracked_cache(dir, len, pathspec);
2277 if (!untracked)
2278 /*
2279 * make sure untracked cache code path is disabled,
2280 * e.g. prep_exclude()
2281 */
2282 dir->untracked = NULL;
2283 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2284 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2285 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2286 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2287
2288 /*
2289 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2290 * also pick up untracked contents of untracked dirs; by default
2291 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2292 */
2293 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2294 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2295 int i, j;
2296
2297 /* remove from dir->entries untracked contents of untracked dirs */
2298 for (i = j = 0; j < dir->nr; j++) {
2299 if (i &&
2300 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
2301 FREE_AND_NULL(dir->entries[j]);
2302 } else {
2303 dir->entries[i++] = dir->entries[j];
2304 }
2305 }
2306
2307 dir->nr = i;
2308 }
2309
2310 trace_performance_since(start, "read directory %.*s", len, path);
2311 if (dir->untracked) {
2312 static int force_untracked_cache = -1;
2313 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2314
2315 if (force_untracked_cache < 0)
2316 force_untracked_cache =
2317 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2318 trace_printf_key(&trace_untracked_stats,
2319 "node creation: %u\n"
2320 "gitignore invalidation: %u\n"
2321 "directory invalidation: %u\n"
2322 "opendir: %u\n",
2323 dir->untracked->dir_created,
2324 dir->untracked->gitignore_invalidated,
2325 dir->untracked->dir_invalidated,
2326 dir->untracked->dir_opened);
2327 if (force_untracked_cache &&
2328 dir->untracked == istate->untracked &&
2329 (dir->untracked->dir_opened ||
2330 dir->untracked->gitignore_invalidated ||
2331 dir->untracked->dir_invalidated))
2332 istate->cache_changed |= UNTRACKED_CHANGED;
2333 if (dir->untracked != istate->untracked) {
2334 FREE_AND_NULL(dir->untracked);
2335 }
2336 }
2337 return dir->nr;
2338 }
2339
2340 int file_exists(const char *f)
2341 {
2342 struct stat sb;
2343 return lstat(f, &sb) == 0;
2344 }
2345
2346 static int cmp_icase(char a, char b)
2347 {
2348 if (a == b)
2349 return 0;
2350 if (ignore_case)
2351 return toupper(a) - toupper(b);
2352 return a - b;
2353 }
2354
2355 /*
2356 * Given two normalized paths (a trailing slash is ok), if subdir is
2357 * outside dir, return -1. Otherwise return the offset in subdir that
2358 * can be used as relative path to dir.
2359 */
2360 int dir_inside_of(const char *subdir, const char *dir)
2361 {
2362 int offset = 0;
2363
2364 assert(dir && subdir && *dir && *subdir);
2365
2366 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2367 dir++;
2368 subdir++;
2369 offset++;
2370 }
2371
2372 /* hel[p]/me vs hel[l]/yeah */
2373 if (*dir && *subdir)
2374 return -1;
2375
2376 if (!*subdir)
2377 return !*dir ? offset : -1; /* same dir */
2378
2379 /* foo/[b]ar vs foo/[] */
2380 if (is_dir_sep(dir[-1]))
2381 return is_dir_sep(subdir[-1]) ? offset : -1;
2382
2383 /* foo[/]bar vs foo[] */
2384 return is_dir_sep(*subdir) ? offset + 1 : -1;
2385 }
2386
2387 int is_inside_dir(const char *dir)
2388 {
2389 char *cwd;
2390 int rc;
2391
2392 if (!dir)
2393 return 0;
2394
2395 cwd = xgetcwd();
2396 rc = (dir_inside_of(cwd, dir) >= 0);
2397 free(cwd);
2398 return rc;
2399 }
2400
2401 int is_empty_dir(const char *path)
2402 {
2403 DIR *dir = opendir(path);
2404 struct dirent *e;
2405 int ret = 1;
2406
2407 if (!dir)
2408 return 0;
2409
2410 while ((e = readdir(dir)) != NULL)
2411 if (!is_dot_or_dotdot(e->d_name)) {
2412 ret = 0;
2413 break;
2414 }
2415
2416 closedir(dir);
2417 return ret;
2418 }
2419
2420 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2421 {
2422 DIR *dir;
2423 struct dirent *e;
2424 int ret = 0, original_len = path->len, len, kept_down = 0;
2425 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2426 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2427 struct object_id submodule_head;
2428
2429 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2430 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
2431 /* Do not descend and nuke a nested git work tree. */
2432 if (kept_up)
2433 *kept_up = 1;
2434 return 0;
2435 }
2436
2437 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2438 dir = opendir(path->buf);
2439 if (!dir) {
2440 if (errno == ENOENT)
2441 return keep_toplevel ? -1 : 0;
2442 else if (errno == EACCES && !keep_toplevel)
2443 /*
2444 * An empty dir could be removable even if it
2445 * is unreadable:
2446 */
2447 return rmdir(path->buf);
2448 else
2449 return -1;
2450 }
2451 strbuf_complete(path, '/');
2452
2453 len = path->len;
2454 while ((e = readdir(dir)) != NULL) {
2455 struct stat st;
2456 if (is_dot_or_dotdot(e->d_name))
2457 continue;
2458
2459 strbuf_setlen(path, len);
2460 strbuf_addstr(path, e->d_name);
2461 if (lstat(path->buf, &st)) {
2462 if (errno == ENOENT)
2463 /*
2464 * file disappeared, which is what we
2465 * wanted anyway
2466 */
2467 continue;
2468 /* fall thru */
2469 } else if (S_ISDIR(st.st_mode)) {
2470 if (!remove_dir_recurse(path, flag, &kept_down))
2471 continue; /* happy */
2472 } else if (!only_empty &&
2473 (!unlink(path->buf) || errno == ENOENT)) {
2474 continue; /* happy, too */
2475 }
2476
2477 /* path too long, stat fails, or non-directory still exists */
2478 ret = -1;
2479 break;
2480 }
2481 closedir(dir);
2482
2483 strbuf_setlen(path, original_len);
2484 if (!ret && !keep_toplevel && !kept_down)
2485 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2486 else if (kept_up)
2487 /*
2488 * report the uplevel that it is not an error that we
2489 * did not rmdir() our directory.
2490 */
2491 *kept_up = !ret;
2492 return ret;
2493 }
2494
2495 int remove_dir_recursively(struct strbuf *path, int flag)
2496 {
2497 return remove_dir_recurse(path, flag, NULL);
2498 }
2499
2500 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2501
2502 void setup_standard_excludes(struct dir_struct *dir)
2503 {
2504 dir->exclude_per_dir = ".gitignore";
2505
2506 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2507 if (!excludes_file)
2508 excludes_file = xdg_config_home("ignore");
2509 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2510 add_excludes_from_file_1(dir, excludes_file,
2511 dir->untracked ? &dir->ss_excludes_file : NULL);
2512
2513 /* per repository user preference */
2514 if (startup_info->have_repository) {
2515 const char *path = git_path_info_exclude();
2516 if (!access_or_warn(path, R_OK, 0))
2517 add_excludes_from_file_1(dir, path,
2518 dir->untracked ? &dir->ss_info_exclude : NULL);
2519 }
2520 }
2521
2522 int remove_path(const char *name)
2523 {
2524 char *slash;
2525
2526 if (unlink(name) && !is_missing_file_error(errno))
2527 return -1;
2528
2529 slash = strrchr(name, '/');
2530 if (slash) {
2531 char *dirs = xstrdup(name);
2532 slash = dirs + (slash - name);
2533 do {
2534 *slash = '\0';
2535 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2536 free(dirs);
2537 }
2538 return 0;
2539 }
2540
2541 /*
2542 * Frees memory within dir which was allocated for exclude lists and
2543 * the exclude_stack. Does not free dir itself.
2544 */
2545 void clear_directory(struct dir_struct *dir)
2546 {
2547 int i, j;
2548 struct exclude_list_group *group;
2549 struct exclude_list *el;
2550 struct exclude_stack *stk;
2551
2552 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2553 group = &dir->exclude_list_group[i];
2554 for (j = 0; j < group->nr; j++) {
2555 el = &group->el[j];
2556 if (i == EXC_DIRS)
2557 free((char *)el->src);
2558 clear_exclude_list(el);
2559 }
2560 free(group->el);
2561 }
2562
2563 stk = dir->exclude_stack;
2564 while (stk) {
2565 struct exclude_stack *prev = stk->prev;
2566 free(stk);
2567 stk = prev;
2568 }
2569 strbuf_release(&dir->basebuf);
2570 }
2571
2572 struct ondisk_untracked_cache {
2573 struct stat_data info_exclude_stat;
2574 struct stat_data excludes_file_stat;
2575 uint32_t dir_flags;
2576 unsigned char info_exclude_sha1[20];
2577 unsigned char excludes_file_sha1[20];
2578 char exclude_per_dir[FLEX_ARRAY];
2579 };
2580
2581 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2582 #define ouc_size(len) (ouc_offset(exclude_per_dir) + len + 1)
2583
2584 struct write_data {
2585 int index; /* number of written untracked_cache_dir */
2586 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2587 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2588 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2589 struct strbuf out;
2590 struct strbuf sb_stat;
2591 struct strbuf sb_sha1;
2592 };
2593
2594 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2595 {
2596 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2597 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2598 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2599 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2600 to->sd_dev = htonl(from->sd_dev);
2601 to->sd_ino = htonl(from->sd_ino);
2602 to->sd_uid = htonl(from->sd_uid);
2603 to->sd_gid = htonl(from->sd_gid);
2604 to->sd_size = htonl(from->sd_size);
2605 }
2606
2607 static void write_one_dir(struct untracked_cache_dir *untracked,
2608 struct write_data *wd)
2609 {
2610 struct stat_data stat_data;
2611 struct strbuf *out = &wd->out;
2612 unsigned char intbuf[16];
2613 unsigned int intlen, value;
2614 int i = wd->index++;
2615
2616 /*
2617 * untracked_nr should be reset whenever valid is clear, but
2618 * for safety..
2619 */
2620 if (!untracked->valid) {
2621 untracked->untracked_nr = 0;
2622 untracked->check_only = 0;
2623 }
2624
2625 if (untracked->check_only)
2626 ewah_set(wd->check_only, i);
2627 if (untracked->valid) {
2628 ewah_set(wd->valid, i);
2629 stat_data_to_disk(&stat_data, &untracked->stat_data);
2630 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2631 }
2632 if (!is_null_oid(&untracked->exclude_oid)) {
2633 ewah_set(wd->sha1_valid, i);
2634 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
2635 the_hash_algo->rawsz);
2636 }
2637
2638 intlen = encode_varint(untracked->untracked_nr, intbuf);
2639 strbuf_add(out, intbuf, intlen);
2640
2641 /* skip non-recurse directories */
2642 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2643 if (untracked->dirs[i]->recurse)
2644 value++;
2645 intlen = encode_varint(value, intbuf);
2646 strbuf_add(out, intbuf, intlen);
2647
2648 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2649
2650 for (i = 0; i < untracked->untracked_nr; i++)
2651 strbuf_add(out, untracked->untracked[i],
2652 strlen(untracked->untracked[i]) + 1);
2653
2654 for (i = 0; i < untracked->dirs_nr; i++)
2655 if (untracked->dirs[i]->recurse)
2656 write_one_dir(untracked->dirs[i], wd);
2657 }
2658
2659 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2660 {
2661 struct ondisk_untracked_cache *ouc;
2662 struct write_data wd;
2663 unsigned char varbuf[16];
2664 int varint_len;
2665 size_t len = strlen(untracked->exclude_per_dir);
2666
2667 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
2668 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2669 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2670 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.oid.hash);
2671 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.oid.hash);
2672 ouc->dir_flags = htonl(untracked->dir_flags);
2673
2674 varint_len = encode_varint(untracked->ident.len, varbuf);
2675 strbuf_add(out, varbuf, varint_len);
2676 strbuf_addbuf(out, &untracked->ident);
2677
2678 strbuf_add(out, ouc, ouc_size(len));
2679 FREE_AND_NULL(ouc);
2680
2681 if (!untracked->root) {
2682 varint_len = encode_varint(0, varbuf);
2683 strbuf_add(out, varbuf, varint_len);
2684 return;
2685 }
2686
2687 wd.index = 0;
2688 wd.check_only = ewah_new();
2689 wd.valid = ewah_new();
2690 wd.sha1_valid = ewah_new();
2691 strbuf_init(&wd.out, 1024);
2692 strbuf_init(&wd.sb_stat, 1024);
2693 strbuf_init(&wd.sb_sha1, 1024);
2694 write_one_dir(untracked->root, &wd);
2695
2696 varint_len = encode_varint(wd.index, varbuf);
2697 strbuf_add(out, varbuf, varint_len);
2698 strbuf_addbuf(out, &wd.out);
2699 ewah_serialize_strbuf(wd.valid, out);
2700 ewah_serialize_strbuf(wd.check_only, out);
2701 ewah_serialize_strbuf(wd.sha1_valid, out);
2702 strbuf_addbuf(out, &wd.sb_stat);
2703 strbuf_addbuf(out, &wd.sb_sha1);
2704 strbuf_addch(out, '\0'); /* safe guard for string lists */
2705
2706 ewah_free(wd.valid);
2707 ewah_free(wd.check_only);
2708 ewah_free(wd.sha1_valid);
2709 strbuf_release(&wd.out);
2710 strbuf_release(&wd.sb_stat);
2711 strbuf_release(&wd.sb_sha1);
2712 }
2713
2714 static void free_untracked(struct untracked_cache_dir *ucd)
2715 {
2716 int i;
2717 if (!ucd)
2718 return;
2719 for (i = 0; i < ucd->dirs_nr; i++)
2720 free_untracked(ucd->dirs[i]);
2721 for (i = 0; i < ucd->untracked_nr; i++)
2722 free(ucd->untracked[i]);
2723 free(ucd->untracked);
2724 free(ucd->dirs);
2725 free(ucd);
2726 }
2727
2728 void free_untracked_cache(struct untracked_cache *uc)
2729 {
2730 if (uc)
2731 free_untracked(uc->root);
2732 free(uc);
2733 }
2734
2735 struct read_data {
2736 int index;
2737 struct untracked_cache_dir **ucd;
2738 struct ewah_bitmap *check_only;
2739 struct ewah_bitmap *valid;
2740 struct ewah_bitmap *sha1_valid;
2741 const unsigned char *data;
2742 const unsigned char *end;
2743 };
2744
2745 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
2746 {
2747 memcpy(to, data, sizeof(*to));
2748 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2749 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2750 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2751 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2752 to->sd_dev = ntohl(to->sd_dev);
2753 to->sd_ino = ntohl(to->sd_ino);
2754 to->sd_uid = ntohl(to->sd_uid);
2755 to->sd_gid = ntohl(to->sd_gid);
2756 to->sd_size = ntohl(to->sd_size);
2757 }
2758
2759 static int read_one_dir(struct untracked_cache_dir **untracked_,
2760 struct read_data *rd)
2761 {
2762 struct untracked_cache_dir ud, *untracked;
2763 const unsigned char *next, *data = rd->data, *end = rd->end;
2764 unsigned int value;
2765 int i, len;
2766
2767 memset(&ud, 0, sizeof(ud));
2768
2769 next = data;
2770 value = decode_varint(&next);
2771 if (next > end)
2772 return -1;
2773 ud.recurse = 1;
2774 ud.untracked_alloc = value;
2775 ud.untracked_nr = value;
2776 if (ud.untracked_nr)
2777 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2778 data = next;
2779
2780 next = data;
2781 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2782 if (next > end)
2783 return -1;
2784 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2785 data = next;
2786
2787 len = strlen((const char *)data);
2788 next = data + len + 1;
2789 if (next > rd->end)
2790 return -1;
2791 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
2792 memcpy(untracked, &ud, sizeof(ud));
2793 memcpy(untracked->name, data, len + 1);
2794 data = next;
2795
2796 for (i = 0; i < untracked->untracked_nr; i++) {
2797 len = strlen((const char *)data);
2798 next = data + len + 1;
2799 if (next > rd->end)
2800 return -1;
2801 untracked->untracked[i] = xstrdup((const char*)data);
2802 data = next;
2803 }
2804
2805 rd->ucd[rd->index++] = untracked;
2806 rd->data = data;
2807
2808 for (i = 0; i < untracked->dirs_nr; i++) {
2809 len = read_one_dir(untracked->dirs + i, rd);
2810 if (len < 0)
2811 return -1;
2812 }
2813 return 0;
2814 }
2815
2816 static void set_check_only(size_t pos, void *cb)
2817 {
2818 struct read_data *rd = cb;
2819 struct untracked_cache_dir *ud = rd->ucd[pos];
2820 ud->check_only = 1;
2821 }
2822
2823 static void read_stat(size_t pos, void *cb)
2824 {
2825 struct read_data *rd = cb;
2826 struct untracked_cache_dir *ud = rd->ucd[pos];
2827 if (rd->data + sizeof(struct stat_data) > rd->end) {
2828 rd->data = rd->end + 1;
2829 return;
2830 }
2831 stat_data_from_disk(&ud->stat_data, rd->data);
2832 rd->data += sizeof(struct stat_data);
2833 ud->valid = 1;
2834 }
2835
2836 static void read_oid(size_t pos, void *cb)
2837 {
2838 struct read_data *rd = cb;
2839 struct untracked_cache_dir *ud = rd->ucd[pos];
2840 if (rd->data + the_hash_algo->rawsz > rd->end) {
2841 rd->data = rd->end + 1;
2842 return;
2843 }
2844 hashcpy(ud->exclude_oid.hash, rd->data);
2845 rd->data += the_hash_algo->rawsz;
2846 }
2847
2848 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
2849 const unsigned char *sha1)
2850 {
2851 stat_data_from_disk(&oid_stat->stat, data);
2852 hashcpy(oid_stat->oid.hash, sha1);
2853 oid_stat->valid = 1;
2854 }
2855
2856 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2857 {
2858 struct untracked_cache *uc;
2859 struct read_data rd;
2860 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2861 const char *ident;
2862 int ident_len;
2863 ssize_t len;
2864 const char *exclude_per_dir;
2865
2866 if (sz <= 1 || end[-1] != '\0')
2867 return NULL;
2868 end--;
2869
2870 ident_len = decode_varint(&next);
2871 if (next + ident_len > end)
2872 return NULL;
2873 ident = (const char *)next;
2874 next += ident_len;
2875
2876 if (next + ouc_size(0) > end)
2877 return NULL;
2878
2879 uc = xcalloc(1, sizeof(*uc));
2880 strbuf_init(&uc->ident, ident_len);
2881 strbuf_add(&uc->ident, ident, ident_len);
2882 load_oid_stat(&uc->ss_info_exclude,
2883 next + ouc_offset(info_exclude_stat),
2884 next + ouc_offset(info_exclude_sha1));
2885 load_oid_stat(&uc->ss_excludes_file,
2886 next + ouc_offset(excludes_file_stat),
2887 next + ouc_offset(excludes_file_sha1));
2888 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2889 exclude_per_dir = (const char *)next + ouc_offset(exclude_per_dir);
2890 uc->exclude_per_dir = xstrdup(exclude_per_dir);
2891 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2892 next += ouc_size(strlen(exclude_per_dir));
2893 if (next >= end)
2894 goto done2;
2895
2896 len = decode_varint(&next);
2897 if (next > end || len == 0)
2898 goto done2;
2899
2900 rd.valid = ewah_new();
2901 rd.check_only = ewah_new();
2902 rd.sha1_valid = ewah_new();
2903 rd.data = next;
2904 rd.end = end;
2905 rd.index = 0;
2906 ALLOC_ARRAY(rd.ucd, len);
2907
2908 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2909 goto done;
2910
2911 next = rd.data;
2912 len = ewah_read_mmap(rd.valid, next, end - next);
2913 if (len < 0)
2914 goto done;
2915
2916 next += len;
2917 len = ewah_read_mmap(rd.check_only, next, end - next);
2918 if (len < 0)
2919 goto done;
2920
2921 next += len;
2922 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2923 if (len < 0)
2924 goto done;
2925
2926 ewah_each_bit(rd.check_only, set_check_only, &rd);
2927 rd.data = next + len;
2928 ewah_each_bit(rd.valid, read_stat, &rd);
2929 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
2930 next = rd.data;
2931
2932 done:
2933 free(rd.ucd);
2934 ewah_free(rd.valid);
2935 ewah_free(rd.check_only);
2936 ewah_free(rd.sha1_valid);
2937 done2:
2938 if (next != end) {
2939 free_untracked_cache(uc);
2940 uc = NULL;
2941 }
2942 return uc;
2943 }
2944
2945 static void invalidate_one_directory(struct untracked_cache *uc,
2946 struct untracked_cache_dir *ucd)
2947 {
2948 uc->dir_invalidated++;
2949 ucd->valid = 0;
2950 ucd->untracked_nr = 0;
2951 }
2952
2953 /*
2954 * Normally when an entry is added or removed from a directory,
2955 * invalidating that directory is enough. No need to touch its
2956 * ancestors. When a directory is shown as "foo/bar/" in git-status
2957 * however, deleting or adding an entry may have cascading effect.
2958 *
2959 * Say the "foo/bar/file" has become untracked, we need to tell the
2960 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2961 * directory any more (because "bar" is managed by foo as an untracked
2962 * "file").
2963 *
2964 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2965 * was the last untracked entry in the entire "foo", we should show
2966 * "foo/" instead. Which means we have to invalidate past "bar" up to
2967 * "foo".
2968 *
2969 * This function traverses all directories from root to leaf. If there
2970 * is a chance of one of the above cases happening, we invalidate back
2971 * to root. Otherwise we just invalidate the leaf. There may be a more
2972 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2973 * detect these cases and avoid unnecessary invalidation, for example,
2974 * checking for the untracked entry named "bar/" in "foo", but for now
2975 * stick to something safe and simple.
2976 */
2977 static int invalidate_one_component(struct untracked_cache *uc,
2978 struct untracked_cache_dir *dir,
2979 const char *path, int len)
2980 {
2981 const char *rest = strchr(path, '/');
2982
2983 if (rest) {
2984 int component_len = rest - path;
2985 struct untracked_cache_dir *d =
2986 lookup_untracked(uc, dir, path, component_len);
2987 int ret =
2988 invalidate_one_component(uc, d, rest + 1,
2989 len - (component_len + 1));
2990 if (ret)
2991 invalidate_one_directory(uc, dir);
2992 return ret;
2993 }
2994
2995 invalidate_one_directory(uc, dir);
2996 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2997 }
2998
2999 void untracked_cache_invalidate_path(struct index_state *istate,
3000 const char *path, int safe_path)
3001 {
3002 if (!istate->untracked || !istate->untracked->root)
3003 return;
3004 if (!safe_path && !verify_path(path, 0))
3005 return;
3006 invalidate_one_component(istate->untracked, istate->untracked->root,
3007 path, strlen(path));
3008 }
3009
3010 void untracked_cache_remove_from_index(struct index_state *istate,
3011 const char *path)
3012 {
3013 untracked_cache_invalidate_path(istate, path, 1);
3014 }
3015
3016 void untracked_cache_add_to_index(struct index_state *istate,
3017 const char *path)
3018 {
3019 untracked_cache_invalidate_path(istate, path, 1);
3020 }
3021
3022 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3023 const char *sub_gitdir)
3024 {
3025 int i;
3026 struct repository subrepo;
3027 struct strbuf sub_wt = STRBUF_INIT;
3028 struct strbuf sub_gd = STRBUF_INIT;
3029
3030 const struct submodule *sub;
3031
3032 /* If the submodule has no working tree, we can ignore it. */
3033 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3034 return;
3035
3036 if (repo_read_index(&subrepo) < 0)
3037 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3038
3039 for (i = 0; i < subrepo.index->cache_nr; i++) {
3040 const struct cache_entry *ce = subrepo.index->cache[i];
3041
3042 if (!S_ISGITLINK(ce->ce_mode))
3043 continue;
3044
3045 while (i + 1 < subrepo.index->cache_nr &&
3046 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3047 /*
3048 * Skip entries with the same name in different stages
3049 * to make sure an entry is returned only once.
3050 */
3051 i++;
3052
3053 sub = submodule_from_path(&subrepo, &null_oid, ce->name);
3054 if (!sub || !is_submodule_active(&subrepo, ce->name))
3055 /* .gitmodules broken or inactive sub */
3056 continue;
3057
3058 strbuf_reset(&sub_wt);
3059 strbuf_reset(&sub_gd);
3060 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3061 strbuf_addf(&sub_gd, "%s/modules/%s", sub_gitdir, sub->name);
3062
3063 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3064 }
3065 strbuf_release(&sub_wt);
3066 strbuf_release(&sub_gd);
3067 repo_clear(&subrepo);
3068 }
3069
3070 void connect_work_tree_and_git_dir(const char *work_tree_,
3071 const char *git_dir_,
3072 int recurse_into_nested)
3073 {
3074 struct strbuf gitfile_sb = STRBUF_INIT;
3075 struct strbuf cfg_sb = STRBUF_INIT;
3076 struct strbuf rel_path = STRBUF_INIT;
3077 char *git_dir, *work_tree;
3078
3079 /* Prepare .git file */
3080 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3081 if (safe_create_leading_directories_const(gitfile_sb.buf))
3082 die(_("could not create directories for %s"), gitfile_sb.buf);
3083
3084 /* Prepare config file */
3085 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3086 if (safe_create_leading_directories_const(cfg_sb.buf))
3087 die(_("could not create directories for %s"), cfg_sb.buf);
3088
3089 git_dir = real_pathdup(git_dir_, 1);
3090 work_tree = real_pathdup(work_tree_, 1);
3091
3092 /* Write .git file */
3093 write_file(gitfile_sb.buf, "gitdir: %s",
3094 relative_path(git_dir, work_tree, &rel_path));
3095 /* Update core.worktree setting */
3096 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3097 relative_path(work_tree, git_dir, &rel_path));
3098
3099 strbuf_release(&gitfile_sb);
3100 strbuf_release(&cfg_sb);
3101 strbuf_release(&rel_path);
3102
3103 if (recurse_into_nested)
3104 connect_wt_gitdir_in_nested(work_tree, git_dir);
3105
3106 free(work_tree);
3107 free(git_dir);
3108 }
3109
3110 /*
3111 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3112 */
3113 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3114 {
3115 if (rename(old_git_dir, new_git_dir) < 0)
3116 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3117 old_git_dir, new_git_dir);
3118
3119 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3120 }