]> git.ipfire.org Git - thirdparty/git.git/blob - dir.c
t1405: explictly delete reflogs for reftable
[thirdparty/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
7 */
8 #include "cache.h"
9 #include "config.h"
10 #include "dir.h"
11 #include "object-store.h"
12 #include "attr.h"
13 #include "refs.h"
14 #include "wildmatch.h"
15 #include "pathspec.h"
16 #include "utf8.h"
17 #include "varint.h"
18 #include "ewah/ewok.h"
19 #include "fsmonitor.h"
20 #include "submodule-config.h"
21
22 /*
23 * Tells read_directory_recursive how a file or directory should be treated.
24 * Values are ordered by significance, e.g. if a directory contains both
25 * excluded and untracked files, it is listed as untracked because
26 * path_untracked > path_excluded.
27 */
28 enum path_treatment {
29 path_none = 0,
30 path_recurse,
31 path_excluded,
32 path_untracked
33 };
34
35 /*
36 * Support data structure for our opendir/readdir/closedir wrappers
37 */
38 struct cached_dir {
39 DIR *fdir;
40 struct untracked_cache_dir *untracked;
41 int nr_files;
42 int nr_dirs;
43
44 const char *d_name;
45 int d_type;
46 const char *file;
47 struct untracked_cache_dir *ucd;
48 };
49
50 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
51 struct index_state *istate, const char *path, int len,
52 struct untracked_cache_dir *untracked,
53 int check_only, int stop_at_first_file, const struct pathspec *pathspec);
54 static int resolve_dtype(int dtype, struct index_state *istate,
55 const char *path, int len);
56 struct dirent *readdir_skip_dot_and_dotdot(DIR *dirp)
57 {
58 struct dirent *e;
59
60 while ((e = readdir(dirp)) != NULL) {
61 if (!is_dot_or_dotdot(e->d_name))
62 break;
63 }
64 return e;
65 }
66
67 int count_slashes(const char *s)
68 {
69 int cnt = 0;
70 while (*s)
71 if (*s++ == '/')
72 cnt++;
73 return cnt;
74 }
75
76 int fspathcmp(const char *a, const char *b)
77 {
78 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
79 }
80
81 int fspatheq(const char *a, const char *b)
82 {
83 return !fspathcmp(a, b);
84 }
85
86 int fspathncmp(const char *a, const char *b, size_t count)
87 {
88 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
89 }
90
91 unsigned int fspathhash(const char *str)
92 {
93 return ignore_case ? strihash(str) : strhash(str);
94 }
95
96 int git_fnmatch(const struct pathspec_item *item,
97 const char *pattern, const char *string,
98 int prefix)
99 {
100 if (prefix > 0) {
101 if (ps_strncmp(item, pattern, string, prefix))
102 return WM_NOMATCH;
103 pattern += prefix;
104 string += prefix;
105 }
106 if (item->flags & PATHSPEC_ONESTAR) {
107 int pattern_len = strlen(++pattern);
108 int string_len = strlen(string);
109 return string_len < pattern_len ||
110 ps_strcmp(item, pattern,
111 string + string_len - pattern_len);
112 }
113 if (item->magic & PATHSPEC_GLOB)
114 return wildmatch(pattern, string,
115 WM_PATHNAME |
116 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
117 else
118 /* wildmatch has not learned no FNM_PATHNAME mode yet */
119 return wildmatch(pattern, string,
120 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
121 }
122
123 static int fnmatch_icase_mem(const char *pattern, int patternlen,
124 const char *string, int stringlen,
125 int flags)
126 {
127 int match_status;
128 struct strbuf pat_buf = STRBUF_INIT;
129 struct strbuf str_buf = STRBUF_INIT;
130 const char *use_pat = pattern;
131 const char *use_str = string;
132
133 if (pattern[patternlen]) {
134 strbuf_add(&pat_buf, pattern, patternlen);
135 use_pat = pat_buf.buf;
136 }
137 if (string[stringlen]) {
138 strbuf_add(&str_buf, string, stringlen);
139 use_str = str_buf.buf;
140 }
141
142 if (ignore_case)
143 flags |= WM_CASEFOLD;
144 match_status = wildmatch(use_pat, use_str, flags);
145
146 strbuf_release(&pat_buf);
147 strbuf_release(&str_buf);
148
149 return match_status;
150 }
151
152 static size_t common_prefix_len(const struct pathspec *pathspec)
153 {
154 int n;
155 size_t max = 0;
156
157 /*
158 * ":(icase)path" is treated as a pathspec full of
159 * wildcard. In other words, only prefix is considered common
160 * prefix. If the pathspec is abc/foo abc/bar, running in
161 * subdir xyz, the common prefix is still xyz, not xyz/abc as
162 * in non-:(icase).
163 */
164 GUARD_PATHSPEC(pathspec,
165 PATHSPEC_FROMTOP |
166 PATHSPEC_MAXDEPTH |
167 PATHSPEC_LITERAL |
168 PATHSPEC_GLOB |
169 PATHSPEC_ICASE |
170 PATHSPEC_EXCLUDE |
171 PATHSPEC_ATTR);
172
173 for (n = 0; n < pathspec->nr; n++) {
174 size_t i = 0, len = 0, item_len;
175 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
176 continue;
177 if (pathspec->items[n].magic & PATHSPEC_ICASE)
178 item_len = pathspec->items[n].prefix;
179 else
180 item_len = pathspec->items[n].nowildcard_len;
181 while (i < item_len && (n == 0 || i < max)) {
182 char c = pathspec->items[n].match[i];
183 if (c != pathspec->items[0].match[i])
184 break;
185 if (c == '/')
186 len = i + 1;
187 i++;
188 }
189 if (n == 0 || len < max) {
190 max = len;
191 if (!max)
192 break;
193 }
194 }
195 return max;
196 }
197
198 /*
199 * Returns a copy of the longest leading path common among all
200 * pathspecs.
201 */
202 char *common_prefix(const struct pathspec *pathspec)
203 {
204 unsigned long len = common_prefix_len(pathspec);
205
206 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
207 }
208
209 int fill_directory(struct dir_struct *dir,
210 struct index_state *istate,
211 const struct pathspec *pathspec)
212 {
213 const char *prefix;
214 size_t prefix_len;
215
216 unsigned exclusive_flags = DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO;
217 if ((dir->flags & exclusive_flags) == exclusive_flags)
218 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
219
220 /*
221 * Calculate common prefix for the pathspec, and
222 * use that to optimize the directory walk
223 */
224 prefix_len = common_prefix_len(pathspec);
225 prefix = prefix_len ? pathspec->items[0].match : "";
226
227 /* Read the directory and prune it */
228 read_directory(dir, istate, prefix, prefix_len, pathspec);
229
230 return prefix_len;
231 }
232
233 int within_depth(const char *name, int namelen,
234 int depth, int max_depth)
235 {
236 const char *cp = name, *cpe = name + namelen;
237
238 while (cp < cpe) {
239 if (*cp++ != '/')
240 continue;
241 depth++;
242 if (depth > max_depth)
243 return 0;
244 }
245 return 1;
246 }
247
248 /*
249 * Read the contents of the blob with the given OID into a buffer.
250 * Append a trailing LF to the end if the last line doesn't have one.
251 *
252 * Returns:
253 * -1 when the OID is invalid or unknown or does not refer to a blob.
254 * 0 when the blob is empty.
255 * 1 along with { data, size } of the (possibly augmented) buffer
256 * when successful.
257 *
258 * Optionally updates the given oid_stat with the given OID (when valid).
259 */
260 static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
261 size_t *size_out, char **data_out)
262 {
263 enum object_type type;
264 unsigned long sz;
265 char *data;
266
267 *size_out = 0;
268 *data_out = NULL;
269
270 data = read_object_file(oid, &type, &sz);
271 if (!data || type != OBJ_BLOB) {
272 free(data);
273 return -1;
274 }
275
276 if (oid_stat) {
277 memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
278 oidcpy(&oid_stat->oid, oid);
279 }
280
281 if (sz == 0) {
282 free(data);
283 return 0;
284 }
285
286 if (data[sz - 1] != '\n') {
287 data = xrealloc(data, st_add(sz, 1));
288 data[sz++] = '\n';
289 }
290
291 *size_out = xsize_t(sz);
292 *data_out = data;
293
294 return 1;
295 }
296
297 #define DO_MATCH_EXCLUDE (1<<0)
298 #define DO_MATCH_DIRECTORY (1<<1)
299 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
300
301 /*
302 * Does the given pathspec match the given name? A match is found if
303 *
304 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
305 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
306 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
307 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
308 *
309 * Return value tells which case it was (1-4), or 0 when there is no match.
310 *
311 * It may be instructive to look at a small table of concrete examples
312 * to understand the differences between 1, 2, and 4:
313 *
314 * Pathspecs
315 * | a/b | a/b/ | a/b/c
316 * ------+-----------+-----------+------------
317 * a/b | EXACT | EXACT[1] | LEADING[2]
318 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
319 * a/b/c | RECURSIVE | RECURSIVE | EXACT
320 *
321 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
322 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
323 */
324 static int match_pathspec_item(struct index_state *istate,
325 const struct pathspec_item *item, int prefix,
326 const char *name, int namelen, unsigned flags)
327 {
328 /* name/namelen has prefix cut off by caller */
329 const char *match = item->match + prefix;
330 int matchlen = item->len - prefix;
331
332 /*
333 * The normal call pattern is:
334 * 1. prefix = common_prefix_len(ps);
335 * 2. prune something, or fill_directory
336 * 3. match_pathspec()
337 *
338 * 'prefix' at #1 may be shorter than the command's prefix and
339 * it's ok for #2 to match extra files. Those extras will be
340 * trimmed at #3.
341 *
342 * Suppose the pathspec is 'foo' and '../bar' running from
343 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
344 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
345 * user does not want XYZ/foo, only the "foo" part should be
346 * case-insensitive. We need to filter out XYZ/foo here. In
347 * other words, we do not trust the caller on comparing the
348 * prefix part when :(icase) is involved. We do exact
349 * comparison ourselves.
350 *
351 * Normally the caller (common_prefix_len() in fact) does
352 * _exact_ matching on name[-prefix+1..-1] and we do not need
353 * to check that part. Be defensive and check it anyway, in
354 * case common_prefix_len is changed, or a new caller is
355 * introduced that does not use common_prefix_len.
356 *
357 * If the penalty turns out too high when prefix is really
358 * long, maybe change it to
359 * strncmp(match, name, item->prefix - prefix)
360 */
361 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
362 strncmp(item->match, name - prefix, item->prefix))
363 return 0;
364
365 if (item->attr_match_nr &&
366 !match_pathspec_attrs(istate, name, namelen, item))
367 return 0;
368
369 /* If the match was just the prefix, we matched */
370 if (!*match)
371 return MATCHED_RECURSIVELY;
372
373 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
374 if (matchlen == namelen)
375 return MATCHED_EXACTLY;
376
377 if (match[matchlen-1] == '/' || name[matchlen] == '/')
378 return MATCHED_RECURSIVELY;
379 } else if ((flags & DO_MATCH_DIRECTORY) &&
380 match[matchlen - 1] == '/' &&
381 namelen == matchlen - 1 &&
382 !ps_strncmp(item, match, name, namelen))
383 return MATCHED_EXACTLY;
384
385 if (item->nowildcard_len < item->len &&
386 !git_fnmatch(item, match, name,
387 item->nowildcard_len - prefix))
388 return MATCHED_FNMATCH;
389
390 /* Perform checks to see if "name" is a leading string of the pathspec */
391 if ( (flags & DO_MATCH_LEADING_PATHSPEC) &&
392 !(flags & DO_MATCH_EXCLUDE)) {
393 /* name is a literal prefix of the pathspec */
394 int offset = name[namelen-1] == '/' ? 1 : 0;
395 if ((namelen < matchlen) &&
396 (match[namelen-offset] == '/') &&
397 !ps_strncmp(item, match, name, namelen))
398 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
399
400 /* name doesn't match up to the first wild character */
401 if (item->nowildcard_len < item->len &&
402 ps_strncmp(item, match, name,
403 item->nowildcard_len - prefix))
404 return 0;
405
406 /*
407 * name has no wildcard, and it didn't match as a leading
408 * pathspec so return.
409 */
410 if (item->nowildcard_len == item->len)
411 return 0;
412
413 /*
414 * Here is where we would perform a wildmatch to check if
415 * "name" can be matched as a directory (or a prefix) against
416 * the pathspec. Since wildmatch doesn't have this capability
417 * at the present we have to punt and say that it is a match,
418 * potentially returning a false positive
419 * The submodules themselves will be able to perform more
420 * accurate matching to determine if the pathspec matches.
421 */
422 return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
423 }
424
425 return 0;
426 }
427
428 /*
429 * do_match_pathspec() is meant to ONLY be called by
430 * match_pathspec_with_flags(); calling it directly risks pathspecs
431 * like ':!unwanted_path' being ignored.
432 *
433 * Given a name and a list of pathspecs, returns the nature of the
434 * closest (i.e. most specific) match of the name to any of the
435 * pathspecs.
436 *
437 * The caller typically calls this multiple times with the same
438 * pathspec and seen[] array but with different name/namelen
439 * (e.g. entries from the index) and is interested in seeing if and
440 * how each pathspec matches all the names it calls this function
441 * with. A mark is left in the seen[] array for each pathspec element
442 * indicating the closest type of match that element achieved, so if
443 * seen[n] remains zero after multiple invocations, that means the nth
444 * pathspec did not match any names, which could indicate that the
445 * user mistyped the nth pathspec.
446 */
447 static int do_match_pathspec(struct index_state *istate,
448 const struct pathspec *ps,
449 const char *name, int namelen,
450 int prefix, char *seen,
451 unsigned flags)
452 {
453 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
454
455 GUARD_PATHSPEC(ps,
456 PATHSPEC_FROMTOP |
457 PATHSPEC_MAXDEPTH |
458 PATHSPEC_LITERAL |
459 PATHSPEC_GLOB |
460 PATHSPEC_ICASE |
461 PATHSPEC_EXCLUDE |
462 PATHSPEC_ATTR);
463
464 if (!ps->nr) {
465 if (!ps->recursive ||
466 !(ps->magic & PATHSPEC_MAXDEPTH) ||
467 ps->max_depth == -1)
468 return MATCHED_RECURSIVELY;
469
470 if (within_depth(name, namelen, 0, ps->max_depth))
471 return MATCHED_EXACTLY;
472 else
473 return 0;
474 }
475
476 name += prefix;
477 namelen -= prefix;
478
479 for (i = ps->nr - 1; i >= 0; i--) {
480 int how;
481
482 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
483 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
484 continue;
485
486 if (seen && seen[i] == MATCHED_EXACTLY)
487 continue;
488 /*
489 * Make exclude patterns optional and never report
490 * "pathspec ':(exclude)foo' matches no files"
491 */
492 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
493 seen[i] = MATCHED_FNMATCH;
494 how = match_pathspec_item(istate, ps->items+i, prefix, name,
495 namelen, flags);
496 if (ps->recursive &&
497 (ps->magic & PATHSPEC_MAXDEPTH) &&
498 ps->max_depth != -1 &&
499 how && how != MATCHED_FNMATCH) {
500 int len = ps->items[i].len;
501 if (name[len] == '/')
502 len++;
503 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
504 how = MATCHED_EXACTLY;
505 else
506 how = 0;
507 }
508 if (how) {
509 if (retval < how)
510 retval = how;
511 if (seen && seen[i] < how)
512 seen[i] = how;
513 }
514 }
515 return retval;
516 }
517
518 static int match_pathspec_with_flags(struct index_state *istate,
519 const struct pathspec *ps,
520 const char *name, int namelen,
521 int prefix, char *seen, unsigned flags)
522 {
523 int positive, negative;
524 positive = do_match_pathspec(istate, ps, name, namelen,
525 prefix, seen, flags);
526 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
527 return positive;
528 negative = do_match_pathspec(istate, ps, name, namelen,
529 prefix, seen,
530 flags | DO_MATCH_EXCLUDE);
531 return negative ? 0 : positive;
532 }
533
534 int match_pathspec(struct index_state *istate,
535 const struct pathspec *ps,
536 const char *name, int namelen,
537 int prefix, char *seen, int is_dir)
538 {
539 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
540 return match_pathspec_with_flags(istate, ps, name, namelen,
541 prefix, seen, flags);
542 }
543
544 /**
545 * Check if a submodule is a superset of the pathspec
546 */
547 int submodule_path_match(struct index_state *istate,
548 const struct pathspec *ps,
549 const char *submodule_name,
550 char *seen)
551 {
552 int matched = match_pathspec_with_flags(istate, ps, submodule_name,
553 strlen(submodule_name),
554 0, seen,
555 DO_MATCH_DIRECTORY |
556 DO_MATCH_LEADING_PATHSPEC);
557 return matched;
558 }
559
560 int report_path_error(const char *ps_matched,
561 const struct pathspec *pathspec)
562 {
563 /*
564 * Make sure all pathspec matched; otherwise it is an error.
565 */
566 int num, errors = 0;
567 for (num = 0; num < pathspec->nr; num++) {
568 int other, found_dup;
569
570 if (ps_matched[num])
571 continue;
572 /*
573 * The caller might have fed identical pathspec
574 * twice. Do not barf on such a mistake.
575 * FIXME: parse_pathspec should have eliminated
576 * duplicate pathspec.
577 */
578 for (found_dup = other = 0;
579 !found_dup && other < pathspec->nr;
580 other++) {
581 if (other == num || !ps_matched[other])
582 continue;
583 if (!strcmp(pathspec->items[other].original,
584 pathspec->items[num].original))
585 /*
586 * Ok, we have a match already.
587 */
588 found_dup = 1;
589 }
590 if (found_dup)
591 continue;
592
593 error(_("pathspec '%s' did not match any file(s) known to git"),
594 pathspec->items[num].original);
595 errors++;
596 }
597 return errors;
598 }
599
600 /*
601 * Return the length of the "simple" part of a path match limiter.
602 */
603 int simple_length(const char *match)
604 {
605 int len = -1;
606
607 for (;;) {
608 unsigned char c = *match++;
609 len++;
610 if (c == '\0' || is_glob_special(c))
611 return len;
612 }
613 }
614
615 int no_wildcard(const char *string)
616 {
617 return string[simple_length(string)] == '\0';
618 }
619
620 void parse_path_pattern(const char **pattern,
621 int *patternlen,
622 unsigned *flags,
623 int *nowildcardlen)
624 {
625 const char *p = *pattern;
626 size_t i, len;
627
628 *flags = 0;
629 if (*p == '!') {
630 *flags |= PATTERN_FLAG_NEGATIVE;
631 p++;
632 }
633 len = strlen(p);
634 if (len && p[len - 1] == '/') {
635 len--;
636 *flags |= PATTERN_FLAG_MUSTBEDIR;
637 }
638 for (i = 0; i < len; i++) {
639 if (p[i] == '/')
640 break;
641 }
642 if (i == len)
643 *flags |= PATTERN_FLAG_NODIR;
644 *nowildcardlen = simple_length(p);
645 /*
646 * we should have excluded the trailing slash from 'p' too,
647 * but that's one more allocation. Instead just make sure
648 * nowildcardlen does not exceed real patternlen
649 */
650 if (*nowildcardlen > len)
651 *nowildcardlen = len;
652 if (*p == '*' && no_wildcard(p + 1))
653 *flags |= PATTERN_FLAG_ENDSWITH;
654 *pattern = p;
655 *patternlen = len;
656 }
657
658 int pl_hashmap_cmp(const void *unused_cmp_data,
659 const struct hashmap_entry *a,
660 const struct hashmap_entry *b,
661 const void *key)
662 {
663 const struct pattern_entry *ee1 =
664 container_of(a, struct pattern_entry, ent);
665 const struct pattern_entry *ee2 =
666 container_of(b, struct pattern_entry, ent);
667
668 size_t min_len = ee1->patternlen <= ee2->patternlen
669 ? ee1->patternlen
670 : ee2->patternlen;
671
672 if (ignore_case)
673 return strncasecmp(ee1->pattern, ee2->pattern, min_len);
674 return strncmp(ee1->pattern, ee2->pattern, min_len);
675 }
676
677 static char *dup_and_filter_pattern(const char *pattern)
678 {
679 char *set, *read;
680 size_t count = 0;
681 char *result = xstrdup(pattern);
682
683 set = result;
684 read = result;
685
686 while (*read) {
687 /* skip escape characters (once) */
688 if (*read == '\\')
689 read++;
690
691 *set = *read;
692
693 set++;
694 read++;
695 count++;
696 }
697 *set = 0;
698
699 if (count > 2 &&
700 *(set - 1) == '*' &&
701 *(set - 2) == '/')
702 *(set - 2) = 0;
703
704 return result;
705 }
706
707 static void add_pattern_to_hashsets(struct pattern_list *pl, struct path_pattern *given)
708 {
709 struct pattern_entry *translated;
710 char *truncated;
711 char *data = NULL;
712 const char *prev, *cur, *next;
713
714 if (!pl->use_cone_patterns)
715 return;
716
717 if (given->flags & PATTERN_FLAG_NEGATIVE &&
718 given->flags & PATTERN_FLAG_MUSTBEDIR &&
719 !strcmp(given->pattern, "/*")) {
720 pl->full_cone = 0;
721 return;
722 }
723
724 if (!given->flags && !strcmp(given->pattern, "/*")) {
725 pl->full_cone = 1;
726 return;
727 }
728
729 if (given->patternlen < 2 ||
730 *given->pattern != '/' ||
731 strstr(given->pattern, "**")) {
732 /* Not a cone pattern. */
733 warning(_("unrecognized pattern: '%s'"), given->pattern);
734 goto clear_hashmaps;
735 }
736
737 prev = given->pattern;
738 cur = given->pattern + 1;
739 next = given->pattern + 2;
740
741 while (*cur) {
742 /* Watch for glob characters '*', '\', '[', '?' */
743 if (!is_glob_special(*cur))
744 goto increment;
745
746 /* But only if *prev != '\\' */
747 if (*prev == '\\')
748 goto increment;
749
750 /* But allow the initial '\' */
751 if (*cur == '\\' &&
752 is_glob_special(*next))
753 goto increment;
754
755 /* But a trailing '/' then '*' is fine */
756 if (*prev == '/' &&
757 *cur == '*' &&
758 *next == 0)
759 goto increment;
760
761 /* Not a cone pattern. */
762 warning(_("unrecognized pattern: '%s'"), given->pattern);
763 goto clear_hashmaps;
764
765 increment:
766 prev++;
767 cur++;
768 next++;
769 }
770
771 if (given->patternlen > 2 &&
772 !strcmp(given->pattern + given->patternlen - 2, "/*")) {
773 if (!(given->flags & PATTERN_FLAG_NEGATIVE)) {
774 /* Not a cone pattern. */
775 warning(_("unrecognized pattern: '%s'"), given->pattern);
776 goto clear_hashmaps;
777 }
778
779 truncated = dup_and_filter_pattern(given->pattern);
780
781 translated = xmalloc(sizeof(struct pattern_entry));
782 translated->pattern = truncated;
783 translated->patternlen = given->patternlen - 2;
784 hashmap_entry_init(&translated->ent,
785 fspathhash(translated->pattern));
786
787 if (!hashmap_get_entry(&pl->recursive_hashmap,
788 translated, ent, NULL)) {
789 /* We did not see the "parent" included */
790 warning(_("unrecognized negative pattern: '%s'"),
791 given->pattern);
792 free(truncated);
793 free(translated);
794 goto clear_hashmaps;
795 }
796
797 hashmap_add(&pl->parent_hashmap, &translated->ent);
798 hashmap_remove(&pl->recursive_hashmap, &translated->ent, &data);
799 free(data);
800 return;
801 }
802
803 if (given->flags & PATTERN_FLAG_NEGATIVE) {
804 warning(_("unrecognized negative pattern: '%s'"),
805 given->pattern);
806 goto clear_hashmaps;
807 }
808
809 translated = xmalloc(sizeof(struct pattern_entry));
810
811 translated->pattern = dup_and_filter_pattern(given->pattern);
812 translated->patternlen = given->patternlen;
813 hashmap_entry_init(&translated->ent,
814 fspathhash(translated->pattern));
815
816 hashmap_add(&pl->recursive_hashmap, &translated->ent);
817
818 if (hashmap_get_entry(&pl->parent_hashmap, translated, ent, NULL)) {
819 /* we already included this at the parent level */
820 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
821 given->pattern);
822 goto clear_hashmaps;
823 }
824
825 return;
826
827 clear_hashmaps:
828 warning(_("disabling cone pattern matching"));
829 hashmap_clear_and_free(&pl->parent_hashmap, struct pattern_entry, ent);
830 hashmap_clear_and_free(&pl->recursive_hashmap, struct pattern_entry, ent);
831 pl->use_cone_patterns = 0;
832 }
833
834 static int hashmap_contains_path(struct hashmap *map,
835 struct strbuf *pattern)
836 {
837 struct pattern_entry p;
838
839 /* Check straight mapping */
840 p.pattern = pattern->buf;
841 p.patternlen = pattern->len;
842 hashmap_entry_init(&p.ent, fspathhash(p.pattern));
843 return !!hashmap_get_entry(map, &p, ent, NULL);
844 }
845
846 int hashmap_contains_parent(struct hashmap *map,
847 const char *path,
848 struct strbuf *buffer)
849 {
850 char *slash_pos;
851
852 strbuf_setlen(buffer, 0);
853
854 if (path[0] != '/')
855 strbuf_addch(buffer, '/');
856
857 strbuf_addstr(buffer, path);
858
859 slash_pos = strrchr(buffer->buf, '/');
860
861 while (slash_pos > buffer->buf) {
862 strbuf_setlen(buffer, slash_pos - buffer->buf);
863
864 if (hashmap_contains_path(map, buffer))
865 return 1;
866
867 slash_pos = strrchr(buffer->buf, '/');
868 }
869
870 return 0;
871 }
872
873 void add_pattern(const char *string, const char *base,
874 int baselen, struct pattern_list *pl, int srcpos)
875 {
876 struct path_pattern *pattern;
877 int patternlen;
878 unsigned flags;
879 int nowildcardlen;
880
881 parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen);
882 if (flags & PATTERN_FLAG_MUSTBEDIR) {
883 FLEXPTR_ALLOC_MEM(pattern, pattern, string, patternlen);
884 } else {
885 pattern = xmalloc(sizeof(*pattern));
886 pattern->pattern = string;
887 }
888 pattern->patternlen = patternlen;
889 pattern->nowildcardlen = nowildcardlen;
890 pattern->base = base;
891 pattern->baselen = baselen;
892 pattern->flags = flags;
893 pattern->srcpos = srcpos;
894 ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc);
895 pl->patterns[pl->nr++] = pattern;
896 pattern->pl = pl;
897
898 add_pattern_to_hashsets(pl, pattern);
899 }
900
901 static int read_skip_worktree_file_from_index(struct index_state *istate,
902 const char *path,
903 size_t *size_out, char **data_out,
904 struct oid_stat *oid_stat)
905 {
906 int pos, len;
907
908 len = strlen(path);
909 pos = index_name_pos(istate, path, len);
910 if (pos < 0)
911 return -1;
912 if (!ce_skip_worktree(istate->cache[pos]))
913 return -1;
914
915 return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
916 }
917
918 /*
919 * Frees memory within pl which was allocated for exclude patterns and
920 * the file buffer. Does not free pl itself.
921 */
922 void clear_pattern_list(struct pattern_list *pl)
923 {
924 int i;
925
926 for (i = 0; i < pl->nr; i++)
927 free(pl->patterns[i]);
928 free(pl->patterns);
929 free(pl->filebuf);
930 hashmap_clear_and_free(&pl->recursive_hashmap, struct pattern_entry, ent);
931 hashmap_clear_and_free(&pl->parent_hashmap, struct pattern_entry, ent);
932
933 memset(pl, 0, sizeof(*pl));
934 }
935
936 static void trim_trailing_spaces(char *buf)
937 {
938 char *p, *last_space = NULL;
939
940 for (p = buf; *p; p++)
941 switch (*p) {
942 case ' ':
943 if (!last_space)
944 last_space = p;
945 break;
946 case '\\':
947 p++;
948 if (!*p)
949 return;
950 /* fallthrough */
951 default:
952 last_space = NULL;
953 }
954
955 if (last_space)
956 *last_space = '\0';
957 }
958
959 /*
960 * Given a subdirectory name and "dir" of the current directory,
961 * search the subdir in "dir" and return it, or create a new one if it
962 * does not exist in "dir".
963 *
964 * If "name" has the trailing slash, it'll be excluded in the search.
965 */
966 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
967 struct untracked_cache_dir *dir,
968 const char *name, int len)
969 {
970 int first, last;
971 struct untracked_cache_dir *d;
972 if (!dir)
973 return NULL;
974 if (len && name[len - 1] == '/')
975 len--;
976 first = 0;
977 last = dir->dirs_nr;
978 while (last > first) {
979 int cmp, next = first + ((last - first) >> 1);
980 d = dir->dirs[next];
981 cmp = strncmp(name, d->name, len);
982 if (!cmp && strlen(d->name) > len)
983 cmp = -1;
984 if (!cmp)
985 return d;
986 if (cmp < 0) {
987 last = next;
988 continue;
989 }
990 first = next+1;
991 }
992
993 uc->dir_created++;
994 FLEX_ALLOC_MEM(d, name, name, len);
995
996 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
997 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
998 dir->dirs_nr - first);
999 dir->dirs_nr++;
1000 dir->dirs[first] = d;
1001 return d;
1002 }
1003
1004 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
1005 {
1006 int i;
1007 dir->valid = 0;
1008 dir->untracked_nr = 0;
1009 for (i = 0; i < dir->dirs_nr; i++)
1010 do_invalidate_gitignore(dir->dirs[i]);
1011 }
1012
1013 static void invalidate_gitignore(struct untracked_cache *uc,
1014 struct untracked_cache_dir *dir)
1015 {
1016 uc->gitignore_invalidated++;
1017 do_invalidate_gitignore(dir);
1018 }
1019
1020 static void invalidate_directory(struct untracked_cache *uc,
1021 struct untracked_cache_dir *dir)
1022 {
1023 int i;
1024
1025 /*
1026 * Invalidation increment here is just roughly correct. If
1027 * untracked_nr or any of dirs[].recurse is non-zero, we
1028 * should increment dir_invalidated too. But that's more
1029 * expensive to do.
1030 */
1031 if (dir->valid)
1032 uc->dir_invalidated++;
1033
1034 dir->valid = 0;
1035 dir->untracked_nr = 0;
1036 for (i = 0; i < dir->dirs_nr; i++)
1037 dir->dirs[i]->recurse = 0;
1038 }
1039
1040 static int add_patterns_from_buffer(char *buf, size_t size,
1041 const char *base, int baselen,
1042 struct pattern_list *pl);
1043
1044 /* Flags for add_patterns() */
1045 #define PATTERN_NOFOLLOW (1<<0)
1046
1047 /*
1048 * Given a file with name "fname", read it (either from disk, or from
1049 * an index if 'istate' is non-null), parse it and store the
1050 * exclude rules in "pl".
1051 *
1052 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1053 * stat data from disk (only valid if add_patterns returns zero). If
1054 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1055 */
1056 static int add_patterns(const char *fname, const char *base, int baselen,
1057 struct pattern_list *pl, struct index_state *istate,
1058 unsigned flags, struct oid_stat *oid_stat)
1059 {
1060 struct stat st;
1061 int r;
1062 int fd;
1063 size_t size = 0;
1064 char *buf;
1065
1066 if (flags & PATTERN_NOFOLLOW)
1067 fd = open_nofollow(fname, O_RDONLY);
1068 else
1069 fd = open(fname, O_RDONLY);
1070
1071 if (fd < 0 || fstat(fd, &st) < 0) {
1072 if (fd < 0)
1073 warn_on_fopen_errors(fname);
1074 else
1075 close(fd);
1076 if (!istate)
1077 return -1;
1078 r = read_skip_worktree_file_from_index(istate, fname,
1079 &size, &buf,
1080 oid_stat);
1081 if (r != 1)
1082 return r;
1083 } else {
1084 size = xsize_t(st.st_size);
1085 if (size == 0) {
1086 if (oid_stat) {
1087 fill_stat_data(&oid_stat->stat, &st);
1088 oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
1089 oid_stat->valid = 1;
1090 }
1091 close(fd);
1092 return 0;
1093 }
1094 buf = xmallocz(size);
1095 if (read_in_full(fd, buf, size) != size) {
1096 free(buf);
1097 close(fd);
1098 return -1;
1099 }
1100 buf[size++] = '\n';
1101 close(fd);
1102 if (oid_stat) {
1103 int pos;
1104 if (oid_stat->valid &&
1105 !match_stat_data_racy(istate, &oid_stat->stat, &st))
1106 ; /* no content change, oid_stat->oid still good */
1107 else if (istate &&
1108 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
1109 !ce_stage(istate->cache[pos]) &&
1110 ce_uptodate(istate->cache[pos]) &&
1111 !would_convert_to_git(istate, fname))
1112 oidcpy(&oid_stat->oid,
1113 &istate->cache[pos]->oid);
1114 else
1115 hash_object_file(the_hash_algo, buf, size,
1116 "blob", &oid_stat->oid);
1117 fill_stat_data(&oid_stat->stat, &st);
1118 oid_stat->valid = 1;
1119 }
1120 }
1121
1122 add_patterns_from_buffer(buf, size, base, baselen, pl);
1123 return 0;
1124 }
1125
1126 static int add_patterns_from_buffer(char *buf, size_t size,
1127 const char *base, int baselen,
1128 struct pattern_list *pl)
1129 {
1130 int i, lineno = 1;
1131 char *entry;
1132
1133 hashmap_init(&pl->recursive_hashmap, pl_hashmap_cmp, NULL, 0);
1134 hashmap_init(&pl->parent_hashmap, pl_hashmap_cmp, NULL, 0);
1135
1136 pl->filebuf = buf;
1137
1138 if (skip_utf8_bom(&buf, size))
1139 size -= buf - pl->filebuf;
1140
1141 entry = buf;
1142
1143 for (i = 0; i < size; i++) {
1144 if (buf[i] == '\n') {
1145 if (entry != buf + i && entry[0] != '#') {
1146 buf[i - (i && buf[i-1] == '\r')] = 0;
1147 trim_trailing_spaces(entry);
1148 add_pattern(entry, base, baselen, pl, lineno);
1149 }
1150 lineno++;
1151 entry = buf + i + 1;
1152 }
1153 }
1154 return 0;
1155 }
1156
1157 int add_patterns_from_file_to_list(const char *fname, const char *base,
1158 int baselen, struct pattern_list *pl,
1159 struct index_state *istate,
1160 unsigned flags)
1161 {
1162 return add_patterns(fname, base, baselen, pl, istate, flags, NULL);
1163 }
1164
1165 int add_patterns_from_blob_to_list(
1166 struct object_id *oid,
1167 const char *base, int baselen,
1168 struct pattern_list *pl)
1169 {
1170 char *buf;
1171 size_t size;
1172 int r;
1173
1174 r = do_read_blob(oid, NULL, &size, &buf);
1175 if (r != 1)
1176 return r;
1177
1178 add_patterns_from_buffer(buf, size, base, baselen, pl);
1179 return 0;
1180 }
1181
1182 struct pattern_list *add_pattern_list(struct dir_struct *dir,
1183 int group_type, const char *src)
1184 {
1185 struct pattern_list *pl;
1186 struct exclude_list_group *group;
1187
1188 group = &dir->exclude_list_group[group_type];
1189 ALLOC_GROW(group->pl, group->nr + 1, group->alloc);
1190 pl = &group->pl[group->nr++];
1191 memset(pl, 0, sizeof(*pl));
1192 pl->src = src;
1193 return pl;
1194 }
1195
1196 /*
1197 * Used to set up core.excludesfile and .git/info/exclude lists.
1198 */
1199 static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname,
1200 struct oid_stat *oid_stat)
1201 {
1202 struct pattern_list *pl;
1203 /*
1204 * catch setup_standard_excludes() that's called before
1205 * dir->untracked is assigned. That function behaves
1206 * differently when dir->untracked is non-NULL.
1207 */
1208 if (!dir->untracked)
1209 dir->unmanaged_exclude_files++;
1210 pl = add_pattern_list(dir, EXC_FILE, fname);
1211 if (add_patterns(fname, "", 0, pl, NULL, 0, oid_stat) < 0)
1212 die(_("cannot use %s as an exclude file"), fname);
1213 }
1214
1215 void add_patterns_from_file(struct dir_struct *dir, const char *fname)
1216 {
1217 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
1218 add_patterns_from_file_1(dir, fname, NULL);
1219 }
1220
1221 int match_basename(const char *basename, int basenamelen,
1222 const char *pattern, int prefix, int patternlen,
1223 unsigned flags)
1224 {
1225 if (prefix == patternlen) {
1226 if (patternlen == basenamelen &&
1227 !fspathncmp(pattern, basename, basenamelen))
1228 return 1;
1229 } else if (flags & PATTERN_FLAG_ENDSWITH) {
1230 /* "*literal" matching against "fooliteral" */
1231 if (patternlen - 1 <= basenamelen &&
1232 !fspathncmp(pattern + 1,
1233 basename + basenamelen - (patternlen - 1),
1234 patternlen - 1))
1235 return 1;
1236 } else {
1237 if (fnmatch_icase_mem(pattern, patternlen,
1238 basename, basenamelen,
1239 0) == 0)
1240 return 1;
1241 }
1242 return 0;
1243 }
1244
1245 int match_pathname(const char *pathname, int pathlen,
1246 const char *base, int baselen,
1247 const char *pattern, int prefix, int patternlen,
1248 unsigned flags)
1249 {
1250 const char *name;
1251 int namelen;
1252
1253 /*
1254 * match with FNM_PATHNAME; the pattern has base implicitly
1255 * in front of it.
1256 */
1257 if (*pattern == '/') {
1258 pattern++;
1259 patternlen--;
1260 prefix--;
1261 }
1262
1263 /*
1264 * baselen does not count the trailing slash. base[] may or
1265 * may not end with a trailing slash though.
1266 */
1267 if (pathlen < baselen + 1 ||
1268 (baselen && pathname[baselen] != '/') ||
1269 fspathncmp(pathname, base, baselen))
1270 return 0;
1271
1272 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1273 name = pathname + pathlen - namelen;
1274
1275 if (prefix) {
1276 /*
1277 * if the non-wildcard part is longer than the
1278 * remaining pathname, surely it cannot match.
1279 */
1280 if (prefix > namelen)
1281 return 0;
1282
1283 if (fspathncmp(pattern, name, prefix))
1284 return 0;
1285 pattern += prefix;
1286 patternlen -= prefix;
1287 name += prefix;
1288 namelen -= prefix;
1289
1290 /*
1291 * If the whole pattern did not have a wildcard,
1292 * then our prefix match is all we need; we
1293 * do not need to call fnmatch at all.
1294 */
1295 if (!patternlen && !namelen)
1296 return 1;
1297 }
1298
1299 return fnmatch_icase_mem(pattern, patternlen,
1300 name, namelen,
1301 WM_PATHNAME) == 0;
1302 }
1303
1304 /*
1305 * Scan the given exclude list in reverse to see whether pathname
1306 * should be ignored. The first match (i.e. the last on the list), if
1307 * any, determines the fate. Returns the exclude_list element which
1308 * matched, or NULL for undecided.
1309 */
1310 static struct path_pattern *last_matching_pattern_from_list(const char *pathname,
1311 int pathlen,
1312 const char *basename,
1313 int *dtype,
1314 struct pattern_list *pl,
1315 struct index_state *istate)
1316 {
1317 struct path_pattern *res = NULL; /* undecided */
1318 int i;
1319
1320 if (!pl->nr)
1321 return NULL; /* undefined */
1322
1323 for (i = pl->nr - 1; 0 <= i; i--) {
1324 struct path_pattern *pattern = pl->patterns[i];
1325 const char *exclude = pattern->pattern;
1326 int prefix = pattern->nowildcardlen;
1327
1328 if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) {
1329 *dtype = resolve_dtype(*dtype, istate, pathname, pathlen);
1330 if (*dtype != DT_DIR)
1331 continue;
1332 }
1333
1334 if (pattern->flags & PATTERN_FLAG_NODIR) {
1335 if (match_basename(basename,
1336 pathlen - (basename - pathname),
1337 exclude, prefix, pattern->patternlen,
1338 pattern->flags)) {
1339 res = pattern;
1340 break;
1341 }
1342 continue;
1343 }
1344
1345 assert(pattern->baselen == 0 ||
1346 pattern->base[pattern->baselen - 1] == '/');
1347 if (match_pathname(pathname, pathlen,
1348 pattern->base,
1349 pattern->baselen ? pattern->baselen - 1 : 0,
1350 exclude, prefix, pattern->patternlen,
1351 pattern->flags)) {
1352 res = pattern;
1353 break;
1354 }
1355 }
1356 return res;
1357 }
1358
1359 /*
1360 * Scan the list of patterns to determine if the ordered list
1361 * of patterns matches on 'pathname'.
1362 *
1363 * Return 1 for a match, 0 for not matched and -1 for undecided.
1364 */
1365 enum pattern_match_result path_matches_pattern_list(
1366 const char *pathname, int pathlen,
1367 const char *basename, int *dtype,
1368 struct pattern_list *pl,
1369 struct index_state *istate)
1370 {
1371 struct path_pattern *pattern;
1372 struct strbuf parent_pathname = STRBUF_INIT;
1373 int result = NOT_MATCHED;
1374 size_t slash_pos;
1375
1376 if (!pl->use_cone_patterns) {
1377 pattern = last_matching_pattern_from_list(pathname, pathlen, basename,
1378 dtype, pl, istate);
1379 if (pattern) {
1380 if (pattern->flags & PATTERN_FLAG_NEGATIVE)
1381 return NOT_MATCHED;
1382 else
1383 return MATCHED;
1384 }
1385
1386 return UNDECIDED;
1387 }
1388
1389 if (pl->full_cone)
1390 return MATCHED;
1391
1392 strbuf_addch(&parent_pathname, '/');
1393 strbuf_add(&parent_pathname, pathname, pathlen);
1394
1395 /*
1396 * Directory entries are matched if and only if a file
1397 * contained immediately within them is matched. For the
1398 * case of a directory entry, modify the path to create
1399 * a fake filename within this directory, allowing us to
1400 * use the file-base matching logic in an equivalent way.
1401 */
1402 if (parent_pathname.len > 0 &&
1403 parent_pathname.buf[parent_pathname.len - 1] == '/') {
1404 slash_pos = parent_pathname.len - 1;
1405 strbuf_add(&parent_pathname, "-", 1);
1406 } else {
1407 const char *slash_ptr = strrchr(parent_pathname.buf, '/');
1408 slash_pos = slash_ptr ? slash_ptr - parent_pathname.buf : 0;
1409 }
1410
1411 if (hashmap_contains_path(&pl->recursive_hashmap,
1412 &parent_pathname)) {
1413 result = MATCHED_RECURSIVE;
1414 goto done;
1415 }
1416
1417 if (!slash_pos) {
1418 /* include every file in root */
1419 result = MATCHED;
1420 goto done;
1421 }
1422
1423 strbuf_setlen(&parent_pathname, slash_pos);
1424
1425 if (hashmap_contains_path(&pl->parent_hashmap, &parent_pathname)) {
1426 result = MATCHED;
1427 goto done;
1428 }
1429
1430 if (hashmap_contains_parent(&pl->recursive_hashmap,
1431 pathname,
1432 &parent_pathname))
1433 result = MATCHED_RECURSIVE;
1434
1435 done:
1436 strbuf_release(&parent_pathname);
1437 return result;
1438 }
1439
1440 int init_sparse_checkout_patterns(struct index_state *istate)
1441 {
1442 if (!core_apply_sparse_checkout)
1443 return 1;
1444 if (istate->sparse_checkout_patterns)
1445 return 0;
1446
1447 CALLOC_ARRAY(istate->sparse_checkout_patterns, 1);
1448
1449 if (get_sparse_checkout_patterns(istate->sparse_checkout_patterns) < 0) {
1450 FREE_AND_NULL(istate->sparse_checkout_patterns);
1451 return -1;
1452 }
1453
1454 return 0;
1455 }
1456
1457 static int path_in_sparse_checkout_1(const char *path,
1458 struct index_state *istate,
1459 int require_cone_mode)
1460 {
1461 int dtype = DT_REG;
1462 enum pattern_match_result match = UNDECIDED;
1463 const char *end, *slash;
1464
1465 /*
1466 * We default to accepting a path if there are no patterns or
1467 * they are of the wrong type.
1468 */
1469 if (init_sparse_checkout_patterns(istate) ||
1470 (require_cone_mode &&
1471 !istate->sparse_checkout_patterns->use_cone_patterns))
1472 return 1;
1473
1474 /*
1475 * If UNDECIDED, use the match from the parent dir (recursively), or
1476 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1477 * never returns UNDECIDED, so we will execute only one iteration in
1478 * this case.
1479 */
1480 for (end = path + strlen(path);
1481 end > path && match == UNDECIDED;
1482 end = slash) {
1483
1484 for (slash = end - 1; slash > path && *slash != '/'; slash--)
1485 ; /* do nothing */
1486
1487 match = path_matches_pattern_list(path, end - path,
1488 slash > path ? slash + 1 : path, &dtype,
1489 istate->sparse_checkout_patterns, istate);
1490
1491 /* We are going to match the parent dir now */
1492 dtype = DT_DIR;
1493 }
1494 return match > 0;
1495 }
1496
1497 int path_in_sparse_checkout(const char *path,
1498 struct index_state *istate)
1499 {
1500 return path_in_sparse_checkout_1(path, istate, 0);
1501 }
1502
1503 int path_in_cone_mode_sparse_checkout(const char *path,
1504 struct index_state *istate)
1505 {
1506 return path_in_sparse_checkout_1(path, istate, 1);
1507 }
1508
1509 static struct path_pattern *last_matching_pattern_from_lists(
1510 struct dir_struct *dir, struct index_state *istate,
1511 const char *pathname, int pathlen,
1512 const char *basename, int *dtype_p)
1513 {
1514 int i, j;
1515 struct exclude_list_group *group;
1516 struct path_pattern *pattern;
1517 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1518 group = &dir->exclude_list_group[i];
1519 for (j = group->nr - 1; j >= 0; j--) {
1520 pattern = last_matching_pattern_from_list(
1521 pathname, pathlen, basename, dtype_p,
1522 &group->pl[j], istate);
1523 if (pattern)
1524 return pattern;
1525 }
1526 }
1527 return NULL;
1528 }
1529
1530 /*
1531 * Loads the per-directory exclude list for the substring of base
1532 * which has a char length of baselen.
1533 */
1534 static void prep_exclude(struct dir_struct *dir,
1535 struct index_state *istate,
1536 const char *base, int baselen)
1537 {
1538 struct exclude_list_group *group;
1539 struct pattern_list *pl;
1540 struct exclude_stack *stk = NULL;
1541 struct untracked_cache_dir *untracked;
1542 int current;
1543
1544 group = &dir->exclude_list_group[EXC_DIRS];
1545
1546 /*
1547 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1548 * which originate from directories not in the prefix of the
1549 * path being checked.
1550 */
1551 while ((stk = dir->exclude_stack) != NULL) {
1552 if (stk->baselen <= baselen &&
1553 !strncmp(dir->basebuf.buf, base, stk->baselen))
1554 break;
1555 pl = &group->pl[dir->exclude_stack->exclude_ix];
1556 dir->exclude_stack = stk->prev;
1557 dir->pattern = NULL;
1558 free((char *)pl->src); /* see strbuf_detach() below */
1559 clear_pattern_list(pl);
1560 free(stk);
1561 group->nr--;
1562 }
1563
1564 /* Skip traversing into sub directories if the parent is excluded */
1565 if (dir->pattern)
1566 return;
1567
1568 /*
1569 * Lazy initialization. All call sites currently just
1570 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1571 * them seems lots of work for little benefit.
1572 */
1573 if (!dir->basebuf.buf)
1574 strbuf_init(&dir->basebuf, PATH_MAX);
1575
1576 /* Read from the parent directories and push them down. */
1577 current = stk ? stk->baselen : -1;
1578 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1579 if (dir->untracked)
1580 untracked = stk ? stk->ucd : dir->untracked->root;
1581 else
1582 untracked = NULL;
1583
1584 while (current < baselen) {
1585 const char *cp;
1586 struct oid_stat oid_stat;
1587
1588 CALLOC_ARRAY(stk, 1);
1589 if (current < 0) {
1590 cp = base;
1591 current = 0;
1592 } else {
1593 cp = strchr(base + current + 1, '/');
1594 if (!cp)
1595 die("oops in prep_exclude");
1596 cp++;
1597 untracked =
1598 lookup_untracked(dir->untracked, untracked,
1599 base + current,
1600 cp - base - current);
1601 }
1602 stk->prev = dir->exclude_stack;
1603 stk->baselen = cp - base;
1604 stk->exclude_ix = group->nr;
1605 stk->ucd = untracked;
1606 pl = add_pattern_list(dir, EXC_DIRS, NULL);
1607 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1608 assert(stk->baselen == dir->basebuf.len);
1609
1610 /* Abort if the directory is excluded */
1611 if (stk->baselen) {
1612 int dt = DT_DIR;
1613 dir->basebuf.buf[stk->baselen - 1] = 0;
1614 dir->pattern = last_matching_pattern_from_lists(dir,
1615 istate,
1616 dir->basebuf.buf, stk->baselen - 1,
1617 dir->basebuf.buf + current, &dt);
1618 dir->basebuf.buf[stk->baselen - 1] = '/';
1619 if (dir->pattern &&
1620 dir->pattern->flags & PATTERN_FLAG_NEGATIVE)
1621 dir->pattern = NULL;
1622 if (dir->pattern) {
1623 dir->exclude_stack = stk;
1624 return;
1625 }
1626 }
1627
1628 /* Try to read per-directory file */
1629 oidclr(&oid_stat.oid);
1630 oid_stat.valid = 0;
1631 if (dir->exclude_per_dir &&
1632 /*
1633 * If we know that no files have been added in
1634 * this directory (i.e. valid_cached_dir() has
1635 * been executed and set untracked->valid) ..
1636 */
1637 (!untracked || !untracked->valid ||
1638 /*
1639 * .. and .gitignore does not exist before
1640 * (i.e. null exclude_oid). Then we can skip
1641 * loading .gitignore, which would result in
1642 * ENOENT anyway.
1643 */
1644 !is_null_oid(&untracked->exclude_oid))) {
1645 /*
1646 * dir->basebuf gets reused by the traversal, but we
1647 * need fname to remain unchanged to ensure the src
1648 * member of each struct path_pattern correctly
1649 * back-references its source file. Other invocations
1650 * of add_pattern_list provide stable strings, so we
1651 * strbuf_detach() and free() here in the caller.
1652 */
1653 struct strbuf sb = STRBUF_INIT;
1654 strbuf_addbuf(&sb, &dir->basebuf);
1655 strbuf_addstr(&sb, dir->exclude_per_dir);
1656 pl->src = strbuf_detach(&sb, NULL);
1657 add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1658 PATTERN_NOFOLLOW,
1659 untracked ? &oid_stat : NULL);
1660 }
1661 /*
1662 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1663 * will first be called in valid_cached_dir() then maybe many
1664 * times more in last_matching_pattern(). When the cache is
1665 * used, last_matching_pattern() will not be called and
1666 * reading .gitignore content will be a waste.
1667 *
1668 * So when it's called by valid_cached_dir() and we can get
1669 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1670 * modified on work tree), we could delay reading the
1671 * .gitignore content until we absolutely need it in
1672 * last_matching_pattern(). Be careful about ignore rule
1673 * order, though, if you do that.
1674 */
1675 if (untracked &&
1676 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1677 invalidate_gitignore(dir->untracked, untracked);
1678 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1679 }
1680 dir->exclude_stack = stk;
1681 current = stk->baselen;
1682 }
1683 strbuf_setlen(&dir->basebuf, baselen);
1684 }
1685
1686 /*
1687 * Loads the exclude lists for the directory containing pathname, then
1688 * scans all exclude lists to determine whether pathname is excluded.
1689 * Returns the exclude_list element which matched, or NULL for
1690 * undecided.
1691 */
1692 struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1693 struct index_state *istate,
1694 const char *pathname,
1695 int *dtype_p)
1696 {
1697 int pathlen = strlen(pathname);
1698 const char *basename = strrchr(pathname, '/');
1699 basename = (basename) ? basename+1 : pathname;
1700
1701 prep_exclude(dir, istate, pathname, basename-pathname);
1702
1703 if (dir->pattern)
1704 return dir->pattern;
1705
1706 return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1707 basename, dtype_p);
1708 }
1709
1710 /*
1711 * Loads the exclude lists for the directory containing pathname, then
1712 * scans all exclude lists to determine whether pathname is excluded.
1713 * Returns 1 if true, otherwise 0.
1714 */
1715 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1716 const char *pathname, int *dtype_p)
1717 {
1718 struct path_pattern *pattern =
1719 last_matching_pattern(dir, istate, pathname, dtype_p);
1720 if (pattern)
1721 return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1722 return 0;
1723 }
1724
1725 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1726 {
1727 struct dir_entry *ent;
1728
1729 FLEX_ALLOC_MEM(ent, name, pathname, len);
1730 ent->len = len;
1731 return ent;
1732 }
1733
1734 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1735 struct index_state *istate,
1736 const char *pathname, int len)
1737 {
1738 if (index_file_exists(istate, pathname, len, ignore_case))
1739 return NULL;
1740
1741 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1742 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1743 }
1744
1745 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1746 struct index_state *istate,
1747 const char *pathname, int len)
1748 {
1749 if (!index_name_is_other(istate, pathname, len))
1750 return NULL;
1751
1752 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1753 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1754 }
1755
1756 enum exist_status {
1757 index_nonexistent = 0,
1758 index_directory,
1759 index_gitdir
1760 };
1761
1762 /*
1763 * Do not use the alphabetically sorted index to look up
1764 * the directory name; instead, use the case insensitive
1765 * directory hash.
1766 */
1767 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1768 const char *dirname, int len)
1769 {
1770 struct cache_entry *ce;
1771
1772 if (index_dir_exists(istate, dirname, len))
1773 return index_directory;
1774
1775 ce = index_file_exists(istate, dirname, len, ignore_case);
1776 if (ce && S_ISGITLINK(ce->ce_mode))
1777 return index_gitdir;
1778
1779 return index_nonexistent;
1780 }
1781
1782 /*
1783 * The index sorts alphabetically by entry name, which
1784 * means that a gitlink sorts as '\0' at the end, while
1785 * a directory (which is defined not as an entry, but as
1786 * the files it contains) will sort with the '/' at the
1787 * end.
1788 */
1789 static enum exist_status directory_exists_in_index(struct index_state *istate,
1790 const char *dirname, int len)
1791 {
1792 int pos;
1793
1794 if (ignore_case)
1795 return directory_exists_in_index_icase(istate, dirname, len);
1796
1797 pos = index_name_pos(istate, dirname, len);
1798 if (pos < 0)
1799 pos = -pos-1;
1800 while (pos < istate->cache_nr) {
1801 const struct cache_entry *ce = istate->cache[pos++];
1802 unsigned char endchar;
1803
1804 if (strncmp(ce->name, dirname, len))
1805 break;
1806 endchar = ce->name[len];
1807 if (endchar > '/')
1808 break;
1809 if (endchar == '/')
1810 return index_directory;
1811 if (!endchar && S_ISGITLINK(ce->ce_mode))
1812 return index_gitdir;
1813 }
1814 return index_nonexistent;
1815 }
1816
1817 /*
1818 * When we find a directory when traversing the filesystem, we
1819 * have three distinct cases:
1820 *
1821 * - ignore it
1822 * - see it as a directory
1823 * - recurse into it
1824 *
1825 * and which one we choose depends on a combination of existing
1826 * git index contents and the flags passed into the directory
1827 * traversal routine.
1828 *
1829 * Case 1: If we *already* have entries in the index under that
1830 * directory name, we always recurse into the directory to see
1831 * all the files.
1832 *
1833 * Case 2: If we *already* have that directory name as a gitlink,
1834 * we always continue to see it as a gitlink, regardless of whether
1835 * there is an actual git directory there or not (it might not
1836 * be checked out as a subproject!)
1837 *
1838 * Case 3: if we didn't have it in the index previously, we
1839 * have a few sub-cases:
1840 *
1841 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1842 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1843 * also true, in which case we need to check if it contains any
1844 * untracked and / or ignored files.
1845 * (b) if it looks like a git directory and we don't have the
1846 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1847 * show it as a directory.
1848 * (c) otherwise, we recurse into it.
1849 */
1850 static enum path_treatment treat_directory(struct dir_struct *dir,
1851 struct index_state *istate,
1852 struct untracked_cache_dir *untracked,
1853 const char *dirname, int len, int baselen, int excluded,
1854 const struct pathspec *pathspec)
1855 {
1856 /*
1857 * WARNING: From this function, you can return path_recurse or you
1858 * can call read_directory_recursive() (or neither), but
1859 * you CAN'T DO BOTH.
1860 */
1861 enum path_treatment state;
1862 int matches_how = 0;
1863 int nested_repo = 0, check_only, stop_early;
1864 int old_ignored_nr, old_untracked_nr;
1865 /* The "len-1" is to strip the final '/' */
1866 enum exist_status status = directory_exists_in_index(istate, dirname, len-1);
1867
1868 if (status == index_directory)
1869 return path_recurse;
1870 if (status == index_gitdir)
1871 return path_none;
1872 if (status != index_nonexistent)
1873 BUG("Unhandled value for directory_exists_in_index: %d\n", status);
1874
1875 /*
1876 * We don't want to descend into paths that don't match the necessary
1877 * patterns. Clearly, if we don't have a pathspec, then we can't check
1878 * for matching patterns. Also, if (excluded) then we know we matched
1879 * the exclusion patterns so as an optimization we can skip checking
1880 * for matching patterns.
1881 */
1882 if (pathspec && !excluded) {
1883 matches_how = match_pathspec_with_flags(istate, pathspec,
1884 dirname, len,
1885 0 /* prefix */,
1886 NULL /* seen */,
1887 DO_MATCH_LEADING_PATHSPEC);
1888 if (!matches_how)
1889 return path_none;
1890 }
1891
1892
1893 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1894 !(dir->flags & DIR_NO_GITLINKS)) {
1895 struct strbuf sb = STRBUF_INIT;
1896 strbuf_addstr(&sb, dirname);
1897 nested_repo = is_nonbare_repository_dir(&sb);
1898 strbuf_release(&sb);
1899 }
1900 if (nested_repo) {
1901 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1902 (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC))
1903 return path_none;
1904 return excluded ? path_excluded : path_untracked;
1905 }
1906
1907 if (!(dir->flags & DIR_SHOW_OTHER_DIRECTORIES)) {
1908 if (excluded &&
1909 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1910 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1911
1912 /*
1913 * This is an excluded directory and we are
1914 * showing ignored paths that match an exclude
1915 * pattern. (e.g. show directory as ignored
1916 * only if it matches an exclude pattern).
1917 * This path will either be 'path_excluded`
1918 * (if we are showing empty directories or if
1919 * the directory is not empty), or will be
1920 * 'path_none' (empty directory, and we are
1921 * not showing empty directories).
1922 */
1923 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1924 return path_excluded;
1925
1926 if (read_directory_recursive(dir, istate, dirname, len,
1927 untracked, 1, 1, pathspec) == path_excluded)
1928 return path_excluded;
1929
1930 return path_none;
1931 }
1932 return path_recurse;
1933 }
1934
1935 assert(dir->flags & DIR_SHOW_OTHER_DIRECTORIES);
1936
1937 /*
1938 * If we have a pathspec which could match something _below_ this
1939 * directory (e.g. when checking 'subdir/' having a pathspec like
1940 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1941 * need to recurse.
1942 */
1943 if (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC)
1944 return path_recurse;
1945
1946 /* Special cases for where this directory is excluded/ignored */
1947 if (excluded) {
1948 /*
1949 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1950 * hiding empty directories, there is no need to
1951 * recurse into an ignored directory.
1952 */
1953 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1954 return path_excluded;
1955
1956 /*
1957 * Even if we are hiding empty directories, we can still avoid
1958 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
1959 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1960 */
1961 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
1962 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1963 return path_excluded;
1964 }
1965
1966 /*
1967 * Other than the path_recurse case above, we only need to
1968 * recurse into untracked directories if any of the following
1969 * bits is set:
1970 * - DIR_SHOW_IGNORED (because then we need to determine if
1971 * there are ignored entries below)
1972 * - DIR_SHOW_IGNORED_TOO (same as above)
1973 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
1974 * the directory is empty)
1975 */
1976 if (!excluded &&
1977 !(dir->flags & (DIR_SHOW_IGNORED |
1978 DIR_SHOW_IGNORED_TOO |
1979 DIR_HIDE_EMPTY_DIRECTORIES))) {
1980 return path_untracked;
1981 }
1982
1983 /*
1984 * Even if we don't want to know all the paths under an untracked or
1985 * ignored directory, we may still need to go into the directory to
1986 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
1987 * an empty directory should be path_none instead of path_excluded or
1988 * path_untracked).
1989 */
1990 check_only = ((dir->flags & DIR_HIDE_EMPTY_DIRECTORIES) &&
1991 !(dir->flags & DIR_SHOW_IGNORED_TOO));
1992
1993 /*
1994 * However, there's another optimization possible as a subset of
1995 * check_only, based on the cases we have to consider:
1996 * A) Directory matches no exclude patterns:
1997 * * Directory is empty => path_none
1998 * * Directory has an untracked file under it => path_untracked
1999 * * Directory has only ignored files under it => path_excluded
2000 * B) Directory matches an exclude pattern:
2001 * * Directory is empty => path_none
2002 * * Directory has an untracked file under it => path_excluded
2003 * * Directory has only ignored files under it => path_excluded
2004 * In case A, we can exit as soon as we've found an untracked
2005 * file but otherwise have to walk all files. In case B, though,
2006 * we can stop at the first file we find under the directory.
2007 */
2008 stop_early = check_only && excluded;
2009
2010 /*
2011 * If /every/ file within an untracked directory is ignored, then
2012 * we want to treat the directory as ignored (for e.g. status
2013 * --porcelain), without listing the individual ignored files
2014 * underneath. To do so, we'll save the current ignored_nr, and
2015 * pop all the ones added after it if it turns out the entire
2016 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2017 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2018 * untracked paths so will need to pop all those off the last
2019 * after we traverse.
2020 */
2021 old_ignored_nr = dir->ignored_nr;
2022 old_untracked_nr = dir->nr;
2023
2024 /* Actually recurse into dirname now, we'll fixup the state later. */
2025 untracked = lookup_untracked(dir->untracked, untracked,
2026 dirname + baselen, len - baselen);
2027 state = read_directory_recursive(dir, istate, dirname, len, untracked,
2028 check_only, stop_early, pathspec);
2029
2030 /* There are a variety of reasons we may need to fixup the state... */
2031 if (state == path_excluded) {
2032 /* state == path_excluded implies all paths under
2033 * dirname were ignored...
2034 *
2035 * if running e.g. `git status --porcelain --ignored=matching`,
2036 * then we want to see the subpaths that are ignored.
2037 *
2038 * if running e.g. just `git status --porcelain`, then
2039 * we just want the directory itself to be listed as ignored
2040 * and not the individual paths underneath.
2041 */
2042 int want_ignored_subpaths =
2043 ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2044 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING));
2045
2046 if (want_ignored_subpaths) {
2047 /*
2048 * with --ignored=matching, we want the subpaths
2049 * INSTEAD of the directory itself.
2050 */
2051 state = path_none;
2052 } else {
2053 int i;
2054 for (i = old_ignored_nr + 1; i<dir->ignored_nr; ++i)
2055 FREE_AND_NULL(dir->ignored[i]);
2056 dir->ignored_nr = old_ignored_nr;
2057 }
2058 }
2059
2060 /*
2061 * We may need to ignore some of the untracked paths we found while
2062 * traversing subdirectories.
2063 */
2064 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2065 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2066 int i;
2067 for (i = old_untracked_nr + 1; i<dir->nr; ++i)
2068 FREE_AND_NULL(dir->entries[i]);
2069 dir->nr = old_untracked_nr;
2070 }
2071
2072 /*
2073 * If there is nothing under the current directory and we are not
2074 * hiding empty directories, then we need to report on the
2075 * untracked or ignored status of the directory itself.
2076 */
2077 if (state == path_none && !(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
2078 state = excluded ? path_excluded : path_untracked;
2079
2080 return state;
2081 }
2082
2083 /*
2084 * This is an inexact early pruning of any recursive directory
2085 * reading - if the path cannot possibly be in the pathspec,
2086 * return true, and we'll skip it early.
2087 */
2088 static int simplify_away(const char *path, int pathlen,
2089 const struct pathspec *pathspec)
2090 {
2091 int i;
2092
2093 if (!pathspec || !pathspec->nr)
2094 return 0;
2095
2096 GUARD_PATHSPEC(pathspec,
2097 PATHSPEC_FROMTOP |
2098 PATHSPEC_MAXDEPTH |
2099 PATHSPEC_LITERAL |
2100 PATHSPEC_GLOB |
2101 PATHSPEC_ICASE |
2102 PATHSPEC_EXCLUDE |
2103 PATHSPEC_ATTR);
2104
2105 for (i = 0; i < pathspec->nr; i++) {
2106 const struct pathspec_item *item = &pathspec->items[i];
2107 int len = item->nowildcard_len;
2108
2109 if (len > pathlen)
2110 len = pathlen;
2111 if (!ps_strncmp(item, item->match, path, len))
2112 return 0;
2113 }
2114
2115 return 1;
2116 }
2117
2118 /*
2119 * This function tells us whether an excluded path matches a
2120 * list of "interesting" pathspecs. That is, whether a path matched
2121 * by any of the pathspecs could possibly be ignored by excluding
2122 * the specified path. This can happen if:
2123 *
2124 * 1. the path is mentioned explicitly in the pathspec
2125 *
2126 * 2. the path is a directory prefix of some element in the
2127 * pathspec
2128 */
2129 static int exclude_matches_pathspec(const char *path, int pathlen,
2130 const struct pathspec *pathspec)
2131 {
2132 int i;
2133
2134 if (!pathspec || !pathspec->nr)
2135 return 0;
2136
2137 GUARD_PATHSPEC(pathspec,
2138 PATHSPEC_FROMTOP |
2139 PATHSPEC_MAXDEPTH |
2140 PATHSPEC_LITERAL |
2141 PATHSPEC_GLOB |
2142 PATHSPEC_ICASE |
2143 PATHSPEC_EXCLUDE);
2144
2145 for (i = 0; i < pathspec->nr; i++) {
2146 const struct pathspec_item *item = &pathspec->items[i];
2147 int len = item->nowildcard_len;
2148
2149 if (len == pathlen &&
2150 !ps_strncmp(item, item->match, path, pathlen))
2151 return 1;
2152 if (len > pathlen &&
2153 item->match[pathlen] == '/' &&
2154 !ps_strncmp(item, item->match, path, pathlen))
2155 return 1;
2156 }
2157 return 0;
2158 }
2159
2160 static int get_index_dtype(struct index_state *istate,
2161 const char *path, int len)
2162 {
2163 int pos;
2164 const struct cache_entry *ce;
2165
2166 ce = index_file_exists(istate, path, len, 0);
2167 if (ce) {
2168 if (!ce_uptodate(ce))
2169 return DT_UNKNOWN;
2170 if (S_ISGITLINK(ce->ce_mode))
2171 return DT_DIR;
2172 /*
2173 * Nobody actually cares about the
2174 * difference between DT_LNK and DT_REG
2175 */
2176 return DT_REG;
2177 }
2178
2179 /* Try to look it up as a directory */
2180 pos = index_name_pos(istate, path, len);
2181 if (pos >= 0)
2182 return DT_UNKNOWN;
2183 pos = -pos-1;
2184 while (pos < istate->cache_nr) {
2185 ce = istate->cache[pos++];
2186 if (strncmp(ce->name, path, len))
2187 break;
2188 if (ce->name[len] > '/')
2189 break;
2190 if (ce->name[len] < '/')
2191 continue;
2192 if (!ce_uptodate(ce))
2193 break; /* continue? */
2194 return DT_DIR;
2195 }
2196 return DT_UNKNOWN;
2197 }
2198
2199 static int resolve_dtype(int dtype, struct index_state *istate,
2200 const char *path, int len)
2201 {
2202 struct stat st;
2203
2204 if (dtype != DT_UNKNOWN)
2205 return dtype;
2206 dtype = get_index_dtype(istate, path, len);
2207 if (dtype != DT_UNKNOWN)
2208 return dtype;
2209 if (lstat(path, &st))
2210 return dtype;
2211 if (S_ISREG(st.st_mode))
2212 return DT_REG;
2213 if (S_ISDIR(st.st_mode))
2214 return DT_DIR;
2215 if (S_ISLNK(st.st_mode))
2216 return DT_LNK;
2217 return dtype;
2218 }
2219
2220 static enum path_treatment treat_path_fast(struct dir_struct *dir,
2221 struct cached_dir *cdir,
2222 struct index_state *istate,
2223 struct strbuf *path,
2224 int baselen,
2225 const struct pathspec *pathspec)
2226 {
2227 /*
2228 * WARNING: From this function, you can return path_recurse or you
2229 * can call read_directory_recursive() (or neither), but
2230 * you CAN'T DO BOTH.
2231 */
2232 strbuf_setlen(path, baselen);
2233 if (!cdir->ucd) {
2234 strbuf_addstr(path, cdir->file);
2235 return path_untracked;
2236 }
2237 strbuf_addstr(path, cdir->ucd->name);
2238 /* treat_one_path() does this before it calls treat_directory() */
2239 strbuf_complete(path, '/');
2240 if (cdir->ucd->check_only)
2241 /*
2242 * check_only is set as a result of treat_directory() getting
2243 * to its bottom. Verify again the same set of directories
2244 * with check_only set.
2245 */
2246 return read_directory_recursive(dir, istate, path->buf, path->len,
2247 cdir->ucd, 1, 0, pathspec);
2248 /*
2249 * We get path_recurse in the first run when
2250 * directory_exists_in_index() returns index_nonexistent. We
2251 * are sure that new changes in the index does not impact the
2252 * outcome. Return now.
2253 */
2254 return path_recurse;
2255 }
2256
2257 static enum path_treatment treat_path(struct dir_struct *dir,
2258 struct untracked_cache_dir *untracked,
2259 struct cached_dir *cdir,
2260 struct index_state *istate,
2261 struct strbuf *path,
2262 int baselen,
2263 const struct pathspec *pathspec)
2264 {
2265 int has_path_in_index, dtype, excluded;
2266
2267 if (!cdir->d_name)
2268 return treat_path_fast(dir, cdir, istate, path,
2269 baselen, pathspec);
2270 if (is_dot_or_dotdot(cdir->d_name) || !fspathcmp(cdir->d_name, ".git"))
2271 return path_none;
2272 strbuf_setlen(path, baselen);
2273 strbuf_addstr(path, cdir->d_name);
2274 if (simplify_away(path->buf, path->len, pathspec))
2275 return path_none;
2276
2277 dtype = resolve_dtype(cdir->d_type, istate, path->buf, path->len);
2278
2279 /* Always exclude indexed files */
2280 has_path_in_index = !!index_file_exists(istate, path->buf, path->len,
2281 ignore_case);
2282 if (dtype != DT_DIR && has_path_in_index)
2283 return path_none;
2284
2285 /*
2286 * When we are looking at a directory P in the working tree,
2287 * there are three cases:
2288 *
2289 * (1) P exists in the index. Everything inside the directory P in
2290 * the working tree needs to go when P is checked out from the
2291 * index.
2292 *
2293 * (2) P does not exist in the index, but there is P/Q in the index.
2294 * We know P will stay a directory when we check out the contents
2295 * of the index, but we do not know yet if there is a directory
2296 * P/Q in the working tree to be killed, so we need to recurse.
2297 *
2298 * (3) P does not exist in the index, and there is no P/Q in the index
2299 * to require P to be a directory, either. Only in this case, we
2300 * know that everything inside P will not be killed without
2301 * recursing.
2302 */
2303 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
2304 (dtype == DT_DIR) &&
2305 !has_path_in_index &&
2306 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
2307 return path_none;
2308
2309 excluded = is_excluded(dir, istate, path->buf, &dtype);
2310
2311 /*
2312 * Excluded? If we don't explicitly want to show
2313 * ignored files, ignore it
2314 */
2315 if (excluded && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
2316 return path_excluded;
2317
2318 switch (dtype) {
2319 default:
2320 return path_none;
2321 case DT_DIR:
2322 /*
2323 * WARNING: Do not ignore/amend the return value from
2324 * treat_directory(), and especially do not change it to return
2325 * path_recurse as that can cause exponential slowdown.
2326 * Instead, modify treat_directory() to return the right value.
2327 */
2328 strbuf_addch(path, '/');
2329 return treat_directory(dir, istate, untracked,
2330 path->buf, path->len,
2331 baselen, excluded, pathspec);
2332 case DT_REG:
2333 case DT_LNK:
2334 if (pathspec &&
2335 !match_pathspec(istate, pathspec, path->buf, path->len,
2336 0 /* prefix */, NULL /* seen */,
2337 0 /* is_dir */))
2338 return path_none;
2339 if (excluded)
2340 return path_excluded;
2341 return path_untracked;
2342 }
2343 }
2344
2345 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
2346 {
2347 if (!dir)
2348 return;
2349 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
2350 dir->untracked_alloc);
2351 dir->untracked[dir->untracked_nr++] = xstrdup(name);
2352 }
2353
2354 static int valid_cached_dir(struct dir_struct *dir,
2355 struct untracked_cache_dir *untracked,
2356 struct index_state *istate,
2357 struct strbuf *path,
2358 int check_only)
2359 {
2360 struct stat st;
2361
2362 if (!untracked)
2363 return 0;
2364
2365 /*
2366 * With fsmonitor, we can trust the untracked cache's valid field.
2367 */
2368 refresh_fsmonitor(istate);
2369 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
2370 if (lstat(path->len ? path->buf : ".", &st)) {
2371 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
2372 return 0;
2373 }
2374 if (!untracked->valid ||
2375 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
2376 fill_stat_data(&untracked->stat_data, &st);
2377 return 0;
2378 }
2379 }
2380
2381 if (untracked->check_only != !!check_only)
2382 return 0;
2383
2384 /*
2385 * prep_exclude will be called eventually on this directory,
2386 * but it's called much later in last_matching_pattern(). We
2387 * need it now to determine the validity of the cache for this
2388 * path. The next calls will be nearly no-op, the way
2389 * prep_exclude() is designed.
2390 */
2391 if (path->len && path->buf[path->len - 1] != '/') {
2392 strbuf_addch(path, '/');
2393 prep_exclude(dir, istate, path->buf, path->len);
2394 strbuf_setlen(path, path->len - 1);
2395 } else
2396 prep_exclude(dir, istate, path->buf, path->len);
2397
2398 /* hopefully prep_exclude() haven't invalidated this entry... */
2399 return untracked->valid;
2400 }
2401
2402 static int open_cached_dir(struct cached_dir *cdir,
2403 struct dir_struct *dir,
2404 struct untracked_cache_dir *untracked,
2405 struct index_state *istate,
2406 struct strbuf *path,
2407 int check_only)
2408 {
2409 const char *c_path;
2410
2411 memset(cdir, 0, sizeof(*cdir));
2412 cdir->untracked = untracked;
2413 if (valid_cached_dir(dir, untracked, istate, path, check_only))
2414 return 0;
2415 c_path = path->len ? path->buf : ".";
2416 cdir->fdir = opendir(c_path);
2417 if (!cdir->fdir)
2418 warning_errno(_("could not open directory '%s'"), c_path);
2419 if (dir->untracked) {
2420 invalidate_directory(dir->untracked, untracked);
2421 dir->untracked->dir_opened++;
2422 }
2423 if (!cdir->fdir)
2424 return -1;
2425 return 0;
2426 }
2427
2428 static int read_cached_dir(struct cached_dir *cdir)
2429 {
2430 struct dirent *de;
2431
2432 if (cdir->fdir) {
2433 de = readdir_skip_dot_and_dotdot(cdir->fdir);
2434 if (!de) {
2435 cdir->d_name = NULL;
2436 cdir->d_type = DT_UNKNOWN;
2437 return -1;
2438 }
2439 cdir->d_name = de->d_name;
2440 cdir->d_type = DTYPE(de);
2441 return 0;
2442 }
2443 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
2444 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
2445 if (!d->recurse) {
2446 cdir->nr_dirs++;
2447 continue;
2448 }
2449 cdir->ucd = d;
2450 cdir->nr_dirs++;
2451 return 0;
2452 }
2453 cdir->ucd = NULL;
2454 if (cdir->nr_files < cdir->untracked->untracked_nr) {
2455 struct untracked_cache_dir *d = cdir->untracked;
2456 cdir->file = d->untracked[cdir->nr_files++];
2457 return 0;
2458 }
2459 return -1;
2460 }
2461
2462 static void close_cached_dir(struct cached_dir *cdir)
2463 {
2464 if (cdir->fdir)
2465 closedir(cdir->fdir);
2466 /*
2467 * We have gone through this directory and found no untracked
2468 * entries. Mark it valid.
2469 */
2470 if (cdir->untracked) {
2471 cdir->untracked->valid = 1;
2472 cdir->untracked->recurse = 1;
2473 }
2474 }
2475
2476 static void add_path_to_appropriate_result_list(struct dir_struct *dir,
2477 struct untracked_cache_dir *untracked,
2478 struct cached_dir *cdir,
2479 struct index_state *istate,
2480 struct strbuf *path,
2481 int baselen,
2482 const struct pathspec *pathspec,
2483 enum path_treatment state)
2484 {
2485 /* add the path to the appropriate result list */
2486 switch (state) {
2487 case path_excluded:
2488 if (dir->flags & DIR_SHOW_IGNORED)
2489 dir_add_name(dir, istate, path->buf, path->len);
2490 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2491 ((dir->flags & DIR_COLLECT_IGNORED) &&
2492 exclude_matches_pathspec(path->buf, path->len,
2493 pathspec)))
2494 dir_add_ignored(dir, istate, path->buf, path->len);
2495 break;
2496
2497 case path_untracked:
2498 if (dir->flags & DIR_SHOW_IGNORED)
2499 break;
2500 dir_add_name(dir, istate, path->buf, path->len);
2501 if (cdir->fdir)
2502 add_untracked(untracked, path->buf + baselen);
2503 break;
2504
2505 default:
2506 break;
2507 }
2508 }
2509
2510 /*
2511 * Read a directory tree. We currently ignore anything but
2512 * directories, regular files and symlinks. That's because git
2513 * doesn't handle them at all yet. Maybe that will change some
2514 * day.
2515 *
2516 * Also, we ignore the name ".git" (even if it is not a directory).
2517 * That likely will not change.
2518 *
2519 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2520 * to signal that a file was found. This is the least significant value that
2521 * indicates that a file was encountered that does not depend on the order of
2522 * whether an untracked or excluded path was encountered first.
2523 *
2524 * Returns the most significant path_treatment value encountered in the scan.
2525 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2526 * significant path_treatment value that will be returned.
2527 */
2528
2529 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
2530 struct index_state *istate, const char *base, int baselen,
2531 struct untracked_cache_dir *untracked, int check_only,
2532 int stop_at_first_file, const struct pathspec *pathspec)
2533 {
2534 /*
2535 * WARNING: Do NOT recurse unless path_recurse is returned from
2536 * treat_path(). Recursing on any other return value
2537 * can result in exponential slowdown.
2538 */
2539 struct cached_dir cdir;
2540 enum path_treatment state, subdir_state, dir_state = path_none;
2541 struct strbuf path = STRBUF_INIT;
2542
2543 strbuf_add(&path, base, baselen);
2544
2545 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
2546 goto out;
2547 dir->visited_directories++;
2548
2549 if (untracked)
2550 untracked->check_only = !!check_only;
2551
2552 while (!read_cached_dir(&cdir)) {
2553 /* check how the file or directory should be treated */
2554 state = treat_path(dir, untracked, &cdir, istate, &path,
2555 baselen, pathspec);
2556 dir->visited_paths++;
2557
2558 if (state > dir_state)
2559 dir_state = state;
2560
2561 /* recurse into subdir if instructed by treat_path */
2562 if (state == path_recurse) {
2563 struct untracked_cache_dir *ud;
2564 ud = lookup_untracked(dir->untracked, untracked,
2565 path.buf + baselen,
2566 path.len - baselen);
2567 subdir_state =
2568 read_directory_recursive(dir, istate, path.buf,
2569 path.len, ud,
2570 check_only, stop_at_first_file, pathspec);
2571 if (subdir_state > dir_state)
2572 dir_state = subdir_state;
2573
2574 if (pathspec &&
2575 !match_pathspec(istate, pathspec, path.buf, path.len,
2576 0 /* prefix */, NULL,
2577 0 /* do NOT special case dirs */))
2578 state = path_none;
2579 }
2580
2581 if (check_only) {
2582 if (stop_at_first_file) {
2583 /*
2584 * If stopping at first file, then
2585 * signal that a file was found by
2586 * returning `path_excluded`. This is
2587 * to return a consistent value
2588 * regardless of whether an ignored or
2589 * excluded file happened to be
2590 * encountered 1st.
2591 *
2592 * In current usage, the
2593 * `stop_at_first_file` is passed when
2594 * an ancestor directory has matched
2595 * an exclude pattern, so any found
2596 * files will be excluded.
2597 */
2598 if (dir_state >= path_excluded) {
2599 dir_state = path_excluded;
2600 break;
2601 }
2602 }
2603
2604 /* abort early if maximum state has been reached */
2605 if (dir_state == path_untracked) {
2606 if (cdir.fdir)
2607 add_untracked(untracked, path.buf + baselen);
2608 break;
2609 }
2610 /* skip the add_path_to_appropriate_result_list() */
2611 continue;
2612 }
2613
2614 add_path_to_appropriate_result_list(dir, untracked, &cdir,
2615 istate, &path, baselen,
2616 pathspec, state);
2617 }
2618 close_cached_dir(&cdir);
2619 out:
2620 strbuf_release(&path);
2621
2622 return dir_state;
2623 }
2624
2625 int cmp_dir_entry(const void *p1, const void *p2)
2626 {
2627 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2628 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2629
2630 return name_compare(e1->name, e1->len, e2->name, e2->len);
2631 }
2632
2633 /* check if *out lexically strictly contains *in */
2634 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2635 {
2636 return (out->len < in->len) &&
2637 (out->name[out->len - 1] == '/') &&
2638 !memcmp(out->name, in->name, out->len);
2639 }
2640
2641 static int treat_leading_path(struct dir_struct *dir,
2642 struct index_state *istate,
2643 const char *path, int len,
2644 const struct pathspec *pathspec)
2645 {
2646 struct strbuf sb = STRBUF_INIT;
2647 struct strbuf subdir = STRBUF_INIT;
2648 int prevlen, baselen;
2649 const char *cp;
2650 struct cached_dir cdir;
2651 enum path_treatment state = path_none;
2652
2653 /*
2654 * For each directory component of path, we are going to check whether
2655 * that path is relevant given the pathspec. For example, if path is
2656 * foo/bar/baz/
2657 * then we will ask treat_path() whether we should go into foo, then
2658 * whether we should go into bar, then whether baz is relevant.
2659 * Checking each is important because e.g. if path is
2660 * .git/info/
2661 * then we need to check .git to know we shouldn't traverse it.
2662 * If the return from treat_path() is:
2663 * * path_none, for any path, we return false.
2664 * * path_recurse, for all path components, we return true
2665 * * <anything else> for some intermediate component, we make sure
2666 * to add that path to the relevant list but return false
2667 * signifying that we shouldn't recurse into it.
2668 */
2669
2670 while (len && path[len - 1] == '/')
2671 len--;
2672 if (!len)
2673 return 1;
2674
2675 memset(&cdir, 0, sizeof(cdir));
2676 cdir.d_type = DT_DIR;
2677 baselen = 0;
2678 prevlen = 0;
2679 while (1) {
2680 prevlen = baselen + !!baselen;
2681 cp = path + prevlen;
2682 cp = memchr(cp, '/', path + len - cp);
2683 if (!cp)
2684 baselen = len;
2685 else
2686 baselen = cp - path;
2687 strbuf_reset(&sb);
2688 strbuf_add(&sb, path, baselen);
2689 if (!is_directory(sb.buf))
2690 break;
2691 strbuf_reset(&sb);
2692 strbuf_add(&sb, path, prevlen);
2693 strbuf_reset(&subdir);
2694 strbuf_add(&subdir, path+prevlen, baselen-prevlen);
2695 cdir.d_name = subdir.buf;
2696 state = treat_path(dir, NULL, &cdir, istate, &sb, prevlen, pathspec);
2697
2698 if (state != path_recurse)
2699 break; /* do not recurse into it */
2700 if (len <= baselen)
2701 break; /* finished checking */
2702 }
2703 add_path_to_appropriate_result_list(dir, NULL, &cdir, istate,
2704 &sb, baselen, pathspec,
2705 state);
2706
2707 strbuf_release(&subdir);
2708 strbuf_release(&sb);
2709 return state == path_recurse;
2710 }
2711
2712 static const char *get_ident_string(void)
2713 {
2714 static struct strbuf sb = STRBUF_INIT;
2715 struct utsname uts;
2716
2717 if (sb.len)
2718 return sb.buf;
2719 if (uname(&uts) < 0)
2720 die_errno(_("failed to get kernel name and information"));
2721 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2722 uts.sysname);
2723 return sb.buf;
2724 }
2725
2726 static int ident_in_untracked(const struct untracked_cache *uc)
2727 {
2728 /*
2729 * Previous git versions may have saved many NUL separated
2730 * strings in the "ident" field, but it is insane to manage
2731 * many locations, so just take care of the first one.
2732 */
2733
2734 return !strcmp(uc->ident.buf, get_ident_string());
2735 }
2736
2737 static void set_untracked_ident(struct untracked_cache *uc)
2738 {
2739 strbuf_reset(&uc->ident);
2740 strbuf_addstr(&uc->ident, get_ident_string());
2741
2742 /*
2743 * This strbuf used to contain a list of NUL separated
2744 * strings, so save NUL too for backward compatibility.
2745 */
2746 strbuf_addch(&uc->ident, 0);
2747 }
2748
2749 static void new_untracked_cache(struct index_state *istate)
2750 {
2751 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2752 strbuf_init(&uc->ident, 100);
2753 uc->exclude_per_dir = ".gitignore";
2754 /* should be the same flags used by git-status */
2755 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2756 set_untracked_ident(uc);
2757 istate->untracked = uc;
2758 istate->cache_changed |= UNTRACKED_CHANGED;
2759 }
2760
2761 void add_untracked_cache(struct index_state *istate)
2762 {
2763 if (!istate->untracked) {
2764 new_untracked_cache(istate);
2765 } else {
2766 if (!ident_in_untracked(istate->untracked)) {
2767 free_untracked_cache(istate->untracked);
2768 new_untracked_cache(istate);
2769 }
2770 }
2771 }
2772
2773 void remove_untracked_cache(struct index_state *istate)
2774 {
2775 if (istate->untracked) {
2776 free_untracked_cache(istate->untracked);
2777 istate->untracked = NULL;
2778 istate->cache_changed |= UNTRACKED_CHANGED;
2779 }
2780 }
2781
2782 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2783 int base_len,
2784 const struct pathspec *pathspec)
2785 {
2786 struct untracked_cache_dir *root;
2787 static int untracked_cache_disabled = -1;
2788
2789 if (!dir->untracked)
2790 return NULL;
2791 if (untracked_cache_disabled < 0)
2792 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2793 if (untracked_cache_disabled)
2794 return NULL;
2795
2796 /*
2797 * We only support $GIT_DIR/info/exclude and core.excludesfile
2798 * as the global ignore rule files. Any other additions
2799 * (e.g. from command line) invalidate the cache. This
2800 * condition also catches running setup_standard_excludes()
2801 * before setting dir->untracked!
2802 */
2803 if (dir->unmanaged_exclude_files)
2804 return NULL;
2805
2806 /*
2807 * Optimize for the main use case only: whole-tree git
2808 * status. More work involved in treat_leading_path() if we
2809 * use cache on just a subset of the worktree. pathspec
2810 * support could make the matter even worse.
2811 */
2812 if (base_len || (pathspec && pathspec->nr))
2813 return NULL;
2814
2815 /* Different set of flags may produce different results */
2816 if (dir->flags != dir->untracked->dir_flags ||
2817 /*
2818 * See treat_directory(), case index_nonexistent. Without
2819 * this flag, we may need to also cache .git file content
2820 * for the resolve_gitlink_ref() call, which we don't.
2821 */
2822 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2823 /* We don't support collecting ignore files */
2824 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2825 DIR_COLLECT_IGNORED)))
2826 return NULL;
2827
2828 /*
2829 * If we use .gitignore in the cache and now you change it to
2830 * .gitexclude, everything will go wrong.
2831 */
2832 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2833 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2834 return NULL;
2835
2836 /*
2837 * EXC_CMDL is not considered in the cache. If people set it,
2838 * skip the cache.
2839 */
2840 if (dir->exclude_list_group[EXC_CMDL].nr)
2841 return NULL;
2842
2843 if (!ident_in_untracked(dir->untracked)) {
2844 warning(_("untracked cache is disabled on this system or location"));
2845 return NULL;
2846 }
2847
2848 if (!dir->untracked->root)
2849 FLEX_ALLOC_STR(dir->untracked->root, name, "");
2850
2851 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2852 root = dir->untracked->root;
2853 if (!oideq(&dir->ss_info_exclude.oid,
2854 &dir->untracked->ss_info_exclude.oid)) {
2855 invalidate_gitignore(dir->untracked, root);
2856 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2857 }
2858 if (!oideq(&dir->ss_excludes_file.oid,
2859 &dir->untracked->ss_excludes_file.oid)) {
2860 invalidate_gitignore(dir->untracked, root);
2861 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2862 }
2863
2864 /* Make sure this directory is not dropped out at saving phase */
2865 root->recurse = 1;
2866 return root;
2867 }
2868
2869 static void emit_traversal_statistics(struct dir_struct *dir,
2870 struct repository *repo,
2871 const char *path,
2872 int path_len)
2873 {
2874 if (!trace2_is_enabled())
2875 return;
2876
2877 if (!path_len) {
2878 trace2_data_string("read_directory", repo, "path", "");
2879 } else {
2880 struct strbuf tmp = STRBUF_INIT;
2881 strbuf_add(&tmp, path, path_len);
2882 trace2_data_string("read_directory", repo, "path", tmp.buf);
2883 strbuf_release(&tmp);
2884 }
2885
2886 trace2_data_intmax("read_directory", repo,
2887 "directories-visited", dir->visited_directories);
2888 trace2_data_intmax("read_directory", repo,
2889 "paths-visited", dir->visited_paths);
2890
2891 if (!dir->untracked)
2892 return;
2893 trace2_data_intmax("read_directory", repo,
2894 "node-creation", dir->untracked->dir_created);
2895 trace2_data_intmax("read_directory", repo,
2896 "gitignore-invalidation",
2897 dir->untracked->gitignore_invalidated);
2898 trace2_data_intmax("read_directory", repo,
2899 "directory-invalidation",
2900 dir->untracked->dir_invalidated);
2901 trace2_data_intmax("read_directory", repo,
2902 "opendir", dir->untracked->dir_opened);
2903 }
2904
2905 int read_directory(struct dir_struct *dir, struct index_state *istate,
2906 const char *path, int len, const struct pathspec *pathspec)
2907 {
2908 struct untracked_cache_dir *untracked;
2909
2910 trace2_region_enter("dir", "read_directory", istate->repo);
2911 dir->visited_paths = 0;
2912 dir->visited_directories = 0;
2913
2914 if (has_symlink_leading_path(path, len)) {
2915 trace2_region_leave("dir", "read_directory", istate->repo);
2916 return dir->nr;
2917 }
2918
2919 untracked = validate_untracked_cache(dir, len, pathspec);
2920 if (!untracked)
2921 /*
2922 * make sure untracked cache code path is disabled,
2923 * e.g. prep_exclude()
2924 */
2925 dir->untracked = NULL;
2926 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2927 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
2928 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2929 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
2930
2931 emit_traversal_statistics(dir, istate->repo, path, len);
2932
2933 trace2_region_leave("dir", "read_directory", istate->repo);
2934 if (dir->untracked) {
2935 static int force_untracked_cache = -1;
2936
2937 if (force_untracked_cache < 0)
2938 force_untracked_cache =
2939 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2940 if (force_untracked_cache &&
2941 dir->untracked == istate->untracked &&
2942 (dir->untracked->dir_opened ||
2943 dir->untracked->gitignore_invalidated ||
2944 dir->untracked->dir_invalidated))
2945 istate->cache_changed |= UNTRACKED_CHANGED;
2946 if (dir->untracked != istate->untracked) {
2947 FREE_AND_NULL(dir->untracked);
2948 }
2949 }
2950
2951 return dir->nr;
2952 }
2953
2954 int file_exists(const char *f)
2955 {
2956 struct stat sb;
2957 return lstat(f, &sb) == 0;
2958 }
2959
2960 int repo_file_exists(struct repository *repo, const char *path)
2961 {
2962 if (repo != the_repository)
2963 BUG("do not know how to check file existence in arbitrary repo");
2964
2965 return file_exists(path);
2966 }
2967
2968 static int cmp_icase(char a, char b)
2969 {
2970 if (a == b)
2971 return 0;
2972 if (ignore_case)
2973 return toupper(a) - toupper(b);
2974 return a - b;
2975 }
2976
2977 /*
2978 * Given two normalized paths (a trailing slash is ok), if subdir is
2979 * outside dir, return -1. Otherwise return the offset in subdir that
2980 * can be used as relative path to dir.
2981 */
2982 int dir_inside_of(const char *subdir, const char *dir)
2983 {
2984 int offset = 0;
2985
2986 assert(dir && subdir && *dir && *subdir);
2987
2988 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2989 dir++;
2990 subdir++;
2991 offset++;
2992 }
2993
2994 /* hel[p]/me vs hel[l]/yeah */
2995 if (*dir && *subdir)
2996 return -1;
2997
2998 if (!*subdir)
2999 return !*dir ? offset : -1; /* same dir */
3000
3001 /* foo/[b]ar vs foo/[] */
3002 if (is_dir_sep(dir[-1]))
3003 return is_dir_sep(subdir[-1]) ? offset : -1;
3004
3005 /* foo[/]bar vs foo[] */
3006 return is_dir_sep(*subdir) ? offset + 1 : -1;
3007 }
3008
3009 int is_inside_dir(const char *dir)
3010 {
3011 char *cwd;
3012 int rc;
3013
3014 if (!dir)
3015 return 0;
3016
3017 cwd = xgetcwd();
3018 rc = (dir_inside_of(cwd, dir) >= 0);
3019 free(cwd);
3020 return rc;
3021 }
3022
3023 int is_empty_dir(const char *path)
3024 {
3025 DIR *dir = opendir(path);
3026 struct dirent *e;
3027 int ret = 1;
3028
3029 if (!dir)
3030 return 0;
3031
3032 e = readdir_skip_dot_and_dotdot(dir);
3033 if (e)
3034 ret = 0;
3035
3036 closedir(dir);
3037 return ret;
3038 }
3039
3040 char *git_url_basename(const char *repo, int is_bundle, int is_bare)
3041 {
3042 const char *end = repo + strlen(repo), *start, *ptr;
3043 size_t len;
3044 char *dir;
3045
3046 /*
3047 * Skip scheme.
3048 */
3049 start = strstr(repo, "://");
3050 if (start == NULL)
3051 start = repo;
3052 else
3053 start += 3;
3054
3055 /*
3056 * Skip authentication data. The stripping does happen
3057 * greedily, such that we strip up to the last '@' inside
3058 * the host part.
3059 */
3060 for (ptr = start; ptr < end && !is_dir_sep(*ptr); ptr++) {
3061 if (*ptr == '@')
3062 start = ptr + 1;
3063 }
3064
3065 /*
3066 * Strip trailing spaces, slashes and /.git
3067 */
3068 while (start < end && (is_dir_sep(end[-1]) || isspace(end[-1])))
3069 end--;
3070 if (end - start > 5 && is_dir_sep(end[-5]) &&
3071 !strncmp(end - 4, ".git", 4)) {
3072 end -= 5;
3073 while (start < end && is_dir_sep(end[-1]))
3074 end--;
3075 }
3076
3077 /*
3078 * Strip trailing port number if we've got only a
3079 * hostname (that is, there is no dir separator but a
3080 * colon). This check is required such that we do not
3081 * strip URI's like '/foo/bar:2222.git', which should
3082 * result in a dir '2222' being guessed due to backwards
3083 * compatibility.
3084 */
3085 if (memchr(start, '/', end - start) == NULL
3086 && memchr(start, ':', end - start) != NULL) {
3087 ptr = end;
3088 while (start < ptr && isdigit(ptr[-1]) && ptr[-1] != ':')
3089 ptr--;
3090 if (start < ptr && ptr[-1] == ':')
3091 end = ptr - 1;
3092 }
3093
3094 /*
3095 * Find last component. To remain backwards compatible we
3096 * also regard colons as path separators, such that
3097 * cloning a repository 'foo:bar.git' would result in a
3098 * directory 'bar' being guessed.
3099 */
3100 ptr = end;
3101 while (start < ptr && !is_dir_sep(ptr[-1]) && ptr[-1] != ':')
3102 ptr--;
3103 start = ptr;
3104
3105 /*
3106 * Strip .{bundle,git}.
3107 */
3108 len = end - start;
3109 strip_suffix_mem(start, &len, is_bundle ? ".bundle" : ".git");
3110
3111 if (!len || (len == 1 && *start == '/'))
3112 die(_("No directory name could be guessed.\n"
3113 "Please specify a directory on the command line"));
3114
3115 if (is_bare)
3116 dir = xstrfmt("%.*s.git", (int)len, start);
3117 else
3118 dir = xstrndup(start, len);
3119 /*
3120 * Replace sequences of 'control' characters and whitespace
3121 * with one ascii space, remove leading and trailing spaces.
3122 */
3123 if (*dir) {
3124 char *out = dir;
3125 int prev_space = 1 /* strip leading whitespace */;
3126 for (end = dir; *end; ++end) {
3127 char ch = *end;
3128 if ((unsigned char)ch < '\x20')
3129 ch = '\x20';
3130 if (isspace(ch)) {
3131 if (prev_space)
3132 continue;
3133 prev_space = 1;
3134 } else
3135 prev_space = 0;
3136 *out++ = ch;
3137 }
3138 *out = '\0';
3139 if (out > dir && prev_space)
3140 out[-1] = '\0';
3141 }
3142 return dir;
3143 }
3144
3145 void strip_dir_trailing_slashes(char *dir)
3146 {
3147 char *end = dir + strlen(dir);
3148
3149 while (dir < end - 1 && is_dir_sep(end[-1]))
3150 end--;
3151 *end = '\0';
3152 }
3153
3154 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
3155 {
3156 DIR *dir;
3157 struct dirent *e;
3158 int ret = 0, original_len = path->len, len, kept_down = 0;
3159 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
3160 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
3161 int purge_original_cwd = (flag & REMOVE_DIR_PURGE_ORIGINAL_CWD);
3162 struct object_id submodule_head;
3163
3164 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
3165 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
3166 /* Do not descend and nuke a nested git work tree. */
3167 if (kept_up)
3168 *kept_up = 1;
3169 return 0;
3170 }
3171
3172 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
3173 dir = opendir(path->buf);
3174 if (!dir) {
3175 if (errno == ENOENT)
3176 return keep_toplevel ? -1 : 0;
3177 else if (errno == EACCES && !keep_toplevel)
3178 /*
3179 * An empty dir could be removable even if it
3180 * is unreadable:
3181 */
3182 return rmdir(path->buf);
3183 else
3184 return -1;
3185 }
3186 strbuf_complete(path, '/');
3187
3188 len = path->len;
3189 while ((e = readdir_skip_dot_and_dotdot(dir)) != NULL) {
3190 struct stat st;
3191
3192 strbuf_setlen(path, len);
3193 strbuf_addstr(path, e->d_name);
3194 if (lstat(path->buf, &st)) {
3195 if (errno == ENOENT)
3196 /*
3197 * file disappeared, which is what we
3198 * wanted anyway
3199 */
3200 continue;
3201 /* fall through */
3202 } else if (S_ISDIR(st.st_mode)) {
3203 if (!remove_dir_recurse(path, flag, &kept_down))
3204 continue; /* happy */
3205 } else if (!only_empty &&
3206 (!unlink(path->buf) || errno == ENOENT)) {
3207 continue; /* happy, too */
3208 }
3209
3210 /* path too long, stat fails, or non-directory still exists */
3211 ret = -1;
3212 break;
3213 }
3214 closedir(dir);
3215
3216 strbuf_setlen(path, original_len);
3217 if (!ret && !keep_toplevel && !kept_down) {
3218 if (!purge_original_cwd &&
3219 startup_info->original_cwd &&
3220 !strcmp(startup_info->original_cwd, path->buf))
3221 ret = -1; /* Do not remove current working directory */
3222 else
3223 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
3224 } else if (kept_up)
3225 /*
3226 * report the uplevel that it is not an error that we
3227 * did not rmdir() our directory.
3228 */
3229 *kept_up = !ret;
3230 return ret;
3231 }
3232
3233 int remove_dir_recursively(struct strbuf *path, int flag)
3234 {
3235 return remove_dir_recurse(path, flag, NULL);
3236 }
3237
3238 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
3239
3240 void setup_standard_excludes(struct dir_struct *dir)
3241 {
3242 dir->exclude_per_dir = ".gitignore";
3243
3244 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3245 if (!excludes_file)
3246 excludes_file = xdg_config_home("ignore");
3247 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
3248 add_patterns_from_file_1(dir, excludes_file,
3249 dir->untracked ? &dir->ss_excludes_file : NULL);
3250
3251 /* per repository user preference */
3252 if (startup_info->have_repository) {
3253 const char *path = git_path_info_exclude();
3254 if (!access_or_warn(path, R_OK, 0))
3255 add_patterns_from_file_1(dir, path,
3256 dir->untracked ? &dir->ss_info_exclude : NULL);
3257 }
3258 }
3259
3260 char *get_sparse_checkout_filename(void)
3261 {
3262 return git_pathdup("info/sparse-checkout");
3263 }
3264
3265 int get_sparse_checkout_patterns(struct pattern_list *pl)
3266 {
3267 int res;
3268 char *sparse_filename = get_sparse_checkout_filename();
3269
3270 pl->use_cone_patterns = core_sparse_checkout_cone;
3271 res = add_patterns_from_file_to_list(sparse_filename, "", 0, pl, NULL, 0);
3272
3273 free(sparse_filename);
3274 return res;
3275 }
3276
3277 int remove_path(const char *name)
3278 {
3279 char *slash;
3280
3281 if (unlink(name) && !is_missing_file_error(errno))
3282 return -1;
3283
3284 slash = strrchr(name, '/');
3285 if (slash) {
3286 char *dirs = xstrdup(name);
3287 slash = dirs + (slash - name);
3288 do {
3289 *slash = '\0';
3290 if (startup_info->original_cwd &&
3291 !strcmp(startup_info->original_cwd, dirs))
3292 break;
3293 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
3294 free(dirs);
3295 }
3296 return 0;
3297 }
3298
3299 /*
3300 * Frees memory within dir which was allocated, and resets fields for further
3301 * use. Does not free dir itself.
3302 */
3303 void dir_clear(struct dir_struct *dir)
3304 {
3305 int i, j;
3306 struct exclude_list_group *group;
3307 struct pattern_list *pl;
3308 struct exclude_stack *stk;
3309 struct dir_struct new = DIR_INIT;
3310
3311 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
3312 group = &dir->exclude_list_group[i];
3313 for (j = 0; j < group->nr; j++) {
3314 pl = &group->pl[j];
3315 if (i == EXC_DIRS)
3316 free((char *)pl->src);
3317 clear_pattern_list(pl);
3318 }
3319 free(group->pl);
3320 }
3321
3322 for (i = 0; i < dir->ignored_nr; i++)
3323 free(dir->ignored[i]);
3324 for (i = 0; i < dir->nr; i++)
3325 free(dir->entries[i]);
3326 free(dir->ignored);
3327 free(dir->entries);
3328
3329 stk = dir->exclude_stack;
3330 while (stk) {
3331 struct exclude_stack *prev = stk->prev;
3332 free(stk);
3333 stk = prev;
3334 }
3335 strbuf_release(&dir->basebuf);
3336
3337 memcpy(dir, &new, sizeof(*dir));
3338 }
3339
3340 struct ondisk_untracked_cache {
3341 struct stat_data info_exclude_stat;
3342 struct stat_data excludes_file_stat;
3343 uint32_t dir_flags;
3344 };
3345
3346 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3347
3348 struct write_data {
3349 int index; /* number of written untracked_cache_dir */
3350 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
3351 struct ewah_bitmap *valid; /* from untracked_cache_dir */
3352 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
3353 struct strbuf out;
3354 struct strbuf sb_stat;
3355 struct strbuf sb_sha1;
3356 };
3357
3358 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
3359 {
3360 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
3361 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
3362 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
3363 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
3364 to->sd_dev = htonl(from->sd_dev);
3365 to->sd_ino = htonl(from->sd_ino);
3366 to->sd_uid = htonl(from->sd_uid);
3367 to->sd_gid = htonl(from->sd_gid);
3368 to->sd_size = htonl(from->sd_size);
3369 }
3370
3371 static void write_one_dir(struct untracked_cache_dir *untracked,
3372 struct write_data *wd)
3373 {
3374 struct stat_data stat_data;
3375 struct strbuf *out = &wd->out;
3376 unsigned char intbuf[16];
3377 unsigned int intlen, value;
3378 int i = wd->index++;
3379
3380 /*
3381 * untracked_nr should be reset whenever valid is clear, but
3382 * for safety..
3383 */
3384 if (!untracked->valid) {
3385 untracked->untracked_nr = 0;
3386 untracked->check_only = 0;
3387 }
3388
3389 if (untracked->check_only)
3390 ewah_set(wd->check_only, i);
3391 if (untracked->valid) {
3392 ewah_set(wd->valid, i);
3393 stat_data_to_disk(&stat_data, &untracked->stat_data);
3394 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
3395 }
3396 if (!is_null_oid(&untracked->exclude_oid)) {
3397 ewah_set(wd->sha1_valid, i);
3398 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
3399 the_hash_algo->rawsz);
3400 }
3401
3402 intlen = encode_varint(untracked->untracked_nr, intbuf);
3403 strbuf_add(out, intbuf, intlen);
3404
3405 /* skip non-recurse directories */
3406 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
3407 if (untracked->dirs[i]->recurse)
3408 value++;
3409 intlen = encode_varint(value, intbuf);
3410 strbuf_add(out, intbuf, intlen);
3411
3412 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
3413
3414 for (i = 0; i < untracked->untracked_nr; i++)
3415 strbuf_add(out, untracked->untracked[i],
3416 strlen(untracked->untracked[i]) + 1);
3417
3418 for (i = 0; i < untracked->dirs_nr; i++)
3419 if (untracked->dirs[i]->recurse)
3420 write_one_dir(untracked->dirs[i], wd);
3421 }
3422
3423 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
3424 {
3425 struct ondisk_untracked_cache *ouc;
3426 struct write_data wd;
3427 unsigned char varbuf[16];
3428 int varint_len;
3429 const unsigned hashsz = the_hash_algo->rawsz;
3430
3431 CALLOC_ARRAY(ouc, 1);
3432 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
3433 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
3434 ouc->dir_flags = htonl(untracked->dir_flags);
3435
3436 varint_len = encode_varint(untracked->ident.len, varbuf);
3437 strbuf_add(out, varbuf, varint_len);
3438 strbuf_addbuf(out, &untracked->ident);
3439
3440 strbuf_add(out, ouc, sizeof(*ouc));
3441 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
3442 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
3443 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
3444 FREE_AND_NULL(ouc);
3445
3446 if (!untracked->root) {
3447 varint_len = encode_varint(0, varbuf);
3448 strbuf_add(out, varbuf, varint_len);
3449 return;
3450 }
3451
3452 wd.index = 0;
3453 wd.check_only = ewah_new();
3454 wd.valid = ewah_new();
3455 wd.sha1_valid = ewah_new();
3456 strbuf_init(&wd.out, 1024);
3457 strbuf_init(&wd.sb_stat, 1024);
3458 strbuf_init(&wd.sb_sha1, 1024);
3459 write_one_dir(untracked->root, &wd);
3460
3461 varint_len = encode_varint(wd.index, varbuf);
3462 strbuf_add(out, varbuf, varint_len);
3463 strbuf_addbuf(out, &wd.out);
3464 ewah_serialize_strbuf(wd.valid, out);
3465 ewah_serialize_strbuf(wd.check_only, out);
3466 ewah_serialize_strbuf(wd.sha1_valid, out);
3467 strbuf_addbuf(out, &wd.sb_stat);
3468 strbuf_addbuf(out, &wd.sb_sha1);
3469 strbuf_addch(out, '\0'); /* safe guard for string lists */
3470
3471 ewah_free(wd.valid);
3472 ewah_free(wd.check_only);
3473 ewah_free(wd.sha1_valid);
3474 strbuf_release(&wd.out);
3475 strbuf_release(&wd.sb_stat);
3476 strbuf_release(&wd.sb_sha1);
3477 }
3478
3479 static void free_untracked(struct untracked_cache_dir *ucd)
3480 {
3481 int i;
3482 if (!ucd)
3483 return;
3484 for (i = 0; i < ucd->dirs_nr; i++)
3485 free_untracked(ucd->dirs[i]);
3486 for (i = 0; i < ucd->untracked_nr; i++)
3487 free(ucd->untracked[i]);
3488 free(ucd->untracked);
3489 free(ucd->dirs);
3490 free(ucd);
3491 }
3492
3493 void free_untracked_cache(struct untracked_cache *uc)
3494 {
3495 if (uc)
3496 free_untracked(uc->root);
3497 free(uc);
3498 }
3499
3500 struct read_data {
3501 int index;
3502 struct untracked_cache_dir **ucd;
3503 struct ewah_bitmap *check_only;
3504 struct ewah_bitmap *valid;
3505 struct ewah_bitmap *sha1_valid;
3506 const unsigned char *data;
3507 const unsigned char *end;
3508 };
3509
3510 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
3511 {
3512 memcpy(to, data, sizeof(*to));
3513 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
3514 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
3515 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
3516 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
3517 to->sd_dev = ntohl(to->sd_dev);
3518 to->sd_ino = ntohl(to->sd_ino);
3519 to->sd_uid = ntohl(to->sd_uid);
3520 to->sd_gid = ntohl(to->sd_gid);
3521 to->sd_size = ntohl(to->sd_size);
3522 }
3523
3524 static int read_one_dir(struct untracked_cache_dir **untracked_,
3525 struct read_data *rd)
3526 {
3527 struct untracked_cache_dir ud, *untracked;
3528 const unsigned char *data = rd->data, *end = rd->end;
3529 const unsigned char *eos;
3530 unsigned int value;
3531 int i;
3532
3533 memset(&ud, 0, sizeof(ud));
3534
3535 value = decode_varint(&data);
3536 if (data > end)
3537 return -1;
3538 ud.recurse = 1;
3539 ud.untracked_alloc = value;
3540 ud.untracked_nr = value;
3541 if (ud.untracked_nr)
3542 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
3543
3544 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
3545 if (data > end)
3546 return -1;
3547 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
3548
3549 eos = memchr(data, '\0', end - data);
3550 if (!eos || eos == end)
3551 return -1;
3552
3553 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
3554 memcpy(untracked, &ud, sizeof(ud));
3555 memcpy(untracked->name, data, eos - data + 1);
3556 data = eos + 1;
3557
3558 for (i = 0; i < untracked->untracked_nr; i++) {
3559 eos = memchr(data, '\0', end - data);
3560 if (!eos || eos == end)
3561 return -1;
3562 untracked->untracked[i] = xmemdupz(data, eos - data);
3563 data = eos + 1;
3564 }
3565
3566 rd->ucd[rd->index++] = untracked;
3567 rd->data = data;
3568
3569 for (i = 0; i < untracked->dirs_nr; i++) {
3570 if (read_one_dir(untracked->dirs + i, rd) < 0)
3571 return -1;
3572 }
3573 return 0;
3574 }
3575
3576 static void set_check_only(size_t pos, void *cb)
3577 {
3578 struct read_data *rd = cb;
3579 struct untracked_cache_dir *ud = rd->ucd[pos];
3580 ud->check_only = 1;
3581 }
3582
3583 static void read_stat(size_t pos, void *cb)
3584 {
3585 struct read_data *rd = cb;
3586 struct untracked_cache_dir *ud = rd->ucd[pos];
3587 if (rd->data + sizeof(struct stat_data) > rd->end) {
3588 rd->data = rd->end + 1;
3589 return;
3590 }
3591 stat_data_from_disk(&ud->stat_data, rd->data);
3592 rd->data += sizeof(struct stat_data);
3593 ud->valid = 1;
3594 }
3595
3596 static void read_oid(size_t pos, void *cb)
3597 {
3598 struct read_data *rd = cb;
3599 struct untracked_cache_dir *ud = rd->ucd[pos];
3600 if (rd->data + the_hash_algo->rawsz > rd->end) {
3601 rd->data = rd->end + 1;
3602 return;
3603 }
3604 oidread(&ud->exclude_oid, rd->data);
3605 rd->data += the_hash_algo->rawsz;
3606 }
3607
3608 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
3609 const unsigned char *sha1)
3610 {
3611 stat_data_from_disk(&oid_stat->stat, data);
3612 oidread(&oid_stat->oid, sha1);
3613 oid_stat->valid = 1;
3614 }
3615
3616 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
3617 {
3618 struct untracked_cache *uc;
3619 struct read_data rd;
3620 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
3621 const char *ident;
3622 int ident_len;
3623 ssize_t len;
3624 const char *exclude_per_dir;
3625 const unsigned hashsz = the_hash_algo->rawsz;
3626 const unsigned offset = sizeof(struct ondisk_untracked_cache);
3627 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
3628
3629 if (sz <= 1 || end[-1] != '\0')
3630 return NULL;
3631 end--;
3632
3633 ident_len = decode_varint(&next);
3634 if (next + ident_len > end)
3635 return NULL;
3636 ident = (const char *)next;
3637 next += ident_len;
3638
3639 if (next + exclude_per_dir_offset + 1 > end)
3640 return NULL;
3641
3642 CALLOC_ARRAY(uc, 1);
3643 strbuf_init(&uc->ident, ident_len);
3644 strbuf_add(&uc->ident, ident, ident_len);
3645 load_oid_stat(&uc->ss_info_exclude,
3646 next + ouc_offset(info_exclude_stat),
3647 next + offset);
3648 load_oid_stat(&uc->ss_excludes_file,
3649 next + ouc_offset(excludes_file_stat),
3650 next + offset + hashsz);
3651 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
3652 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
3653 uc->exclude_per_dir = xstrdup(exclude_per_dir);
3654 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3655 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
3656 if (next >= end)
3657 goto done2;
3658
3659 len = decode_varint(&next);
3660 if (next > end || len == 0)
3661 goto done2;
3662
3663 rd.valid = ewah_new();
3664 rd.check_only = ewah_new();
3665 rd.sha1_valid = ewah_new();
3666 rd.data = next;
3667 rd.end = end;
3668 rd.index = 0;
3669 ALLOC_ARRAY(rd.ucd, len);
3670
3671 if (read_one_dir(&uc->root, &rd) || rd.index != len)
3672 goto done;
3673
3674 next = rd.data;
3675 len = ewah_read_mmap(rd.valid, next, end - next);
3676 if (len < 0)
3677 goto done;
3678
3679 next += len;
3680 len = ewah_read_mmap(rd.check_only, next, end - next);
3681 if (len < 0)
3682 goto done;
3683
3684 next += len;
3685 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
3686 if (len < 0)
3687 goto done;
3688
3689 ewah_each_bit(rd.check_only, set_check_only, &rd);
3690 rd.data = next + len;
3691 ewah_each_bit(rd.valid, read_stat, &rd);
3692 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
3693 next = rd.data;
3694
3695 done:
3696 free(rd.ucd);
3697 ewah_free(rd.valid);
3698 ewah_free(rd.check_only);
3699 ewah_free(rd.sha1_valid);
3700 done2:
3701 if (next != end) {
3702 free_untracked_cache(uc);
3703 uc = NULL;
3704 }
3705 return uc;
3706 }
3707
3708 static void invalidate_one_directory(struct untracked_cache *uc,
3709 struct untracked_cache_dir *ucd)
3710 {
3711 uc->dir_invalidated++;
3712 ucd->valid = 0;
3713 ucd->untracked_nr = 0;
3714 }
3715
3716 /*
3717 * Normally when an entry is added or removed from a directory,
3718 * invalidating that directory is enough. No need to touch its
3719 * ancestors. When a directory is shown as "foo/bar/" in git-status
3720 * however, deleting or adding an entry may have cascading effect.
3721 *
3722 * Say the "foo/bar/file" has become untracked, we need to tell the
3723 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3724 * directory any more (because "bar" is managed by foo as an untracked
3725 * "file").
3726 *
3727 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3728 * was the last untracked entry in the entire "foo", we should show
3729 * "foo/" instead. Which means we have to invalidate past "bar" up to
3730 * "foo".
3731 *
3732 * This function traverses all directories from root to leaf. If there
3733 * is a chance of one of the above cases happening, we invalidate back
3734 * to root. Otherwise we just invalidate the leaf. There may be a more
3735 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3736 * detect these cases and avoid unnecessary invalidation, for example,
3737 * checking for the untracked entry named "bar/" in "foo", but for now
3738 * stick to something safe and simple.
3739 */
3740 static int invalidate_one_component(struct untracked_cache *uc,
3741 struct untracked_cache_dir *dir,
3742 const char *path, int len)
3743 {
3744 const char *rest = strchr(path, '/');
3745
3746 if (rest) {
3747 int component_len = rest - path;
3748 struct untracked_cache_dir *d =
3749 lookup_untracked(uc, dir, path, component_len);
3750 int ret =
3751 invalidate_one_component(uc, d, rest + 1,
3752 len - (component_len + 1));
3753 if (ret)
3754 invalidate_one_directory(uc, dir);
3755 return ret;
3756 }
3757
3758 invalidate_one_directory(uc, dir);
3759 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
3760 }
3761
3762 void untracked_cache_invalidate_path(struct index_state *istate,
3763 const char *path, int safe_path)
3764 {
3765 if (!istate->untracked || !istate->untracked->root)
3766 return;
3767 if (!safe_path && !verify_path(path, 0))
3768 return;
3769 invalidate_one_component(istate->untracked, istate->untracked->root,
3770 path, strlen(path));
3771 }
3772
3773 void untracked_cache_remove_from_index(struct index_state *istate,
3774 const char *path)
3775 {
3776 untracked_cache_invalidate_path(istate, path, 1);
3777 }
3778
3779 void untracked_cache_add_to_index(struct index_state *istate,
3780 const char *path)
3781 {
3782 untracked_cache_invalidate_path(istate, path, 1);
3783 }
3784
3785 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3786 const char *sub_gitdir)
3787 {
3788 int i;
3789 struct repository subrepo;
3790 struct strbuf sub_wt = STRBUF_INIT;
3791 struct strbuf sub_gd = STRBUF_INIT;
3792
3793 const struct submodule *sub;
3794
3795 /* If the submodule has no working tree, we can ignore it. */
3796 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3797 return;
3798
3799 if (repo_read_index(&subrepo) < 0)
3800 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3801
3802 /* TODO: audit for interaction with sparse-index. */
3803 ensure_full_index(subrepo.index);
3804 for (i = 0; i < subrepo.index->cache_nr; i++) {
3805 const struct cache_entry *ce = subrepo.index->cache[i];
3806
3807 if (!S_ISGITLINK(ce->ce_mode))
3808 continue;
3809
3810 while (i + 1 < subrepo.index->cache_nr &&
3811 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3812 /*
3813 * Skip entries with the same name in different stages
3814 * to make sure an entry is returned only once.
3815 */
3816 i++;
3817
3818 sub = submodule_from_path(&subrepo, null_oid(), ce->name);
3819 if (!sub || !is_submodule_active(&subrepo, ce->name))
3820 /* .gitmodules broken or inactive sub */
3821 continue;
3822
3823 strbuf_reset(&sub_wt);
3824 strbuf_reset(&sub_gd);
3825 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3826 submodule_name_to_gitdir(&sub_gd, &subrepo, sub->name);
3827
3828 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3829 }
3830 strbuf_release(&sub_wt);
3831 strbuf_release(&sub_gd);
3832 repo_clear(&subrepo);
3833 }
3834
3835 void connect_work_tree_and_git_dir(const char *work_tree_,
3836 const char *git_dir_,
3837 int recurse_into_nested)
3838 {
3839 struct strbuf gitfile_sb = STRBUF_INIT;
3840 struct strbuf cfg_sb = STRBUF_INIT;
3841 struct strbuf rel_path = STRBUF_INIT;
3842 char *git_dir, *work_tree;
3843
3844 /* Prepare .git file */
3845 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3846 if (safe_create_leading_directories_const(gitfile_sb.buf))
3847 die(_("could not create directories for %s"), gitfile_sb.buf);
3848
3849 /* Prepare config file */
3850 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3851 if (safe_create_leading_directories_const(cfg_sb.buf))
3852 die(_("could not create directories for %s"), cfg_sb.buf);
3853
3854 git_dir = real_pathdup(git_dir_, 1);
3855 work_tree = real_pathdup(work_tree_, 1);
3856
3857 /* Write .git file */
3858 write_file(gitfile_sb.buf, "gitdir: %s",
3859 relative_path(git_dir, work_tree, &rel_path));
3860 /* Update core.worktree setting */
3861 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3862 relative_path(work_tree, git_dir, &rel_path));
3863
3864 strbuf_release(&gitfile_sb);
3865 strbuf_release(&cfg_sb);
3866 strbuf_release(&rel_path);
3867
3868 if (recurse_into_nested)
3869 connect_wt_gitdir_in_nested(work_tree, git_dir);
3870
3871 free(work_tree);
3872 free(git_dir);
3873 }
3874
3875 /*
3876 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3877 */
3878 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3879 {
3880 if (rename(old_git_dir, new_git_dir) < 0)
3881 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3882 old_git_dir, new_git_dir);
3883
3884 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3885 }