]> git.ipfire.org Git - thirdparty/git.git/blob - dir.c
clone: allow "--bare" with "-o"
[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 OBJ_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 the path is empty, there are no
1467 * patterns, or the patterns are of the wrong type.
1468 */
1469 if (!*path ||
1470 init_sparse_checkout_patterns(istate) ||
1471 (require_cone_mode &&
1472 !istate->sparse_checkout_patterns->use_cone_patterns))
1473 return 1;
1474
1475 /*
1476 * If UNDECIDED, use the match from the parent dir (recursively), or
1477 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1478 * never returns UNDECIDED, so we will execute only one iteration in
1479 * this case.
1480 */
1481 for (end = path + strlen(path);
1482 end > path && match == UNDECIDED;
1483 end = slash) {
1484
1485 for (slash = end - 1; slash > path && *slash != '/'; slash--)
1486 ; /* do nothing */
1487
1488 match = path_matches_pattern_list(path, end - path,
1489 slash > path ? slash + 1 : path, &dtype,
1490 istate->sparse_checkout_patterns, istate);
1491
1492 /* We are going to match the parent dir now */
1493 dtype = DT_DIR;
1494 }
1495 return match > 0;
1496 }
1497
1498 int path_in_sparse_checkout(const char *path,
1499 struct index_state *istate)
1500 {
1501 return path_in_sparse_checkout_1(path, istate, 0);
1502 }
1503
1504 int path_in_cone_mode_sparse_checkout(const char *path,
1505 struct index_state *istate)
1506 {
1507 return path_in_sparse_checkout_1(path, istate, 1);
1508 }
1509
1510 static struct path_pattern *last_matching_pattern_from_lists(
1511 struct dir_struct *dir, struct index_state *istate,
1512 const char *pathname, int pathlen,
1513 const char *basename, int *dtype_p)
1514 {
1515 int i, j;
1516 struct exclude_list_group *group;
1517 struct path_pattern *pattern;
1518 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1519 group = &dir->exclude_list_group[i];
1520 for (j = group->nr - 1; j >= 0; j--) {
1521 pattern = last_matching_pattern_from_list(
1522 pathname, pathlen, basename, dtype_p,
1523 &group->pl[j], istate);
1524 if (pattern)
1525 return pattern;
1526 }
1527 }
1528 return NULL;
1529 }
1530
1531 /*
1532 * Loads the per-directory exclude list for the substring of base
1533 * which has a char length of baselen.
1534 */
1535 static void prep_exclude(struct dir_struct *dir,
1536 struct index_state *istate,
1537 const char *base, int baselen)
1538 {
1539 struct exclude_list_group *group;
1540 struct pattern_list *pl;
1541 struct exclude_stack *stk = NULL;
1542 struct untracked_cache_dir *untracked;
1543 int current;
1544
1545 group = &dir->exclude_list_group[EXC_DIRS];
1546
1547 /*
1548 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1549 * which originate from directories not in the prefix of the
1550 * path being checked.
1551 */
1552 while ((stk = dir->exclude_stack) != NULL) {
1553 if (stk->baselen <= baselen &&
1554 !strncmp(dir->basebuf.buf, base, stk->baselen))
1555 break;
1556 pl = &group->pl[dir->exclude_stack->exclude_ix];
1557 dir->exclude_stack = stk->prev;
1558 dir->pattern = NULL;
1559 free((char *)pl->src); /* see strbuf_detach() below */
1560 clear_pattern_list(pl);
1561 free(stk);
1562 group->nr--;
1563 }
1564
1565 /* Skip traversing into sub directories if the parent is excluded */
1566 if (dir->pattern)
1567 return;
1568
1569 /*
1570 * Lazy initialization. All call sites currently just
1571 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1572 * them seems lots of work for little benefit.
1573 */
1574 if (!dir->basebuf.buf)
1575 strbuf_init(&dir->basebuf, PATH_MAX);
1576
1577 /* Read from the parent directories and push them down. */
1578 current = stk ? stk->baselen : -1;
1579 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1580 if (dir->untracked)
1581 untracked = stk ? stk->ucd : dir->untracked->root;
1582 else
1583 untracked = NULL;
1584
1585 while (current < baselen) {
1586 const char *cp;
1587 struct oid_stat oid_stat;
1588
1589 CALLOC_ARRAY(stk, 1);
1590 if (current < 0) {
1591 cp = base;
1592 current = 0;
1593 } else {
1594 cp = strchr(base + current + 1, '/');
1595 if (!cp)
1596 die("oops in prep_exclude");
1597 cp++;
1598 untracked =
1599 lookup_untracked(dir->untracked, untracked,
1600 base + current,
1601 cp - base - current);
1602 }
1603 stk->prev = dir->exclude_stack;
1604 stk->baselen = cp - base;
1605 stk->exclude_ix = group->nr;
1606 stk->ucd = untracked;
1607 pl = add_pattern_list(dir, EXC_DIRS, NULL);
1608 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1609 assert(stk->baselen == dir->basebuf.len);
1610
1611 /* Abort if the directory is excluded */
1612 if (stk->baselen) {
1613 int dt = DT_DIR;
1614 dir->basebuf.buf[stk->baselen - 1] = 0;
1615 dir->pattern = last_matching_pattern_from_lists(dir,
1616 istate,
1617 dir->basebuf.buf, stk->baselen - 1,
1618 dir->basebuf.buf + current, &dt);
1619 dir->basebuf.buf[stk->baselen - 1] = '/';
1620 if (dir->pattern &&
1621 dir->pattern->flags & PATTERN_FLAG_NEGATIVE)
1622 dir->pattern = NULL;
1623 if (dir->pattern) {
1624 dir->exclude_stack = stk;
1625 return;
1626 }
1627 }
1628
1629 /* Try to read per-directory file */
1630 oidclr(&oid_stat.oid);
1631 oid_stat.valid = 0;
1632 if (dir->exclude_per_dir &&
1633 /*
1634 * If we know that no files have been added in
1635 * this directory (i.e. valid_cached_dir() has
1636 * been executed and set untracked->valid) ..
1637 */
1638 (!untracked || !untracked->valid ||
1639 /*
1640 * .. and .gitignore does not exist before
1641 * (i.e. null exclude_oid). Then we can skip
1642 * loading .gitignore, which would result in
1643 * ENOENT anyway.
1644 */
1645 !is_null_oid(&untracked->exclude_oid))) {
1646 /*
1647 * dir->basebuf gets reused by the traversal, but we
1648 * need fname to remain unchanged to ensure the src
1649 * member of each struct path_pattern correctly
1650 * back-references its source file. Other invocations
1651 * of add_pattern_list provide stable strings, so we
1652 * strbuf_detach() and free() here in the caller.
1653 */
1654 struct strbuf sb = STRBUF_INIT;
1655 strbuf_addbuf(&sb, &dir->basebuf);
1656 strbuf_addstr(&sb, dir->exclude_per_dir);
1657 pl->src = strbuf_detach(&sb, NULL);
1658 add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1659 PATTERN_NOFOLLOW,
1660 untracked ? &oid_stat : NULL);
1661 }
1662 /*
1663 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1664 * will first be called in valid_cached_dir() then maybe many
1665 * times more in last_matching_pattern(). When the cache is
1666 * used, last_matching_pattern() will not be called and
1667 * reading .gitignore content will be a waste.
1668 *
1669 * So when it's called by valid_cached_dir() and we can get
1670 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1671 * modified on work tree), we could delay reading the
1672 * .gitignore content until we absolutely need it in
1673 * last_matching_pattern(). Be careful about ignore rule
1674 * order, though, if you do that.
1675 */
1676 if (untracked &&
1677 !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1678 invalidate_gitignore(dir->untracked, untracked);
1679 oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1680 }
1681 dir->exclude_stack = stk;
1682 current = stk->baselen;
1683 }
1684 strbuf_setlen(&dir->basebuf, baselen);
1685 }
1686
1687 /*
1688 * Loads the exclude lists for the directory containing pathname, then
1689 * scans all exclude lists to determine whether pathname is excluded.
1690 * Returns the exclude_list element which matched, or NULL for
1691 * undecided.
1692 */
1693 struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1694 struct index_state *istate,
1695 const char *pathname,
1696 int *dtype_p)
1697 {
1698 int pathlen = strlen(pathname);
1699 const char *basename = strrchr(pathname, '/');
1700 basename = (basename) ? basename+1 : pathname;
1701
1702 prep_exclude(dir, istate, pathname, basename-pathname);
1703
1704 if (dir->pattern)
1705 return dir->pattern;
1706
1707 return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1708 basename, dtype_p);
1709 }
1710
1711 /*
1712 * Loads the exclude lists for the directory containing pathname, then
1713 * scans all exclude lists to determine whether pathname is excluded.
1714 * Returns 1 if true, otherwise 0.
1715 */
1716 int is_excluded(struct dir_struct *dir, struct index_state *istate,
1717 const char *pathname, int *dtype_p)
1718 {
1719 struct path_pattern *pattern =
1720 last_matching_pattern(dir, istate, pathname, dtype_p);
1721 if (pattern)
1722 return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1723 return 0;
1724 }
1725
1726 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1727 {
1728 struct dir_entry *ent;
1729
1730 FLEX_ALLOC_MEM(ent, name, pathname, len);
1731 ent->len = len;
1732 return ent;
1733 }
1734
1735 static struct dir_entry *dir_add_name(struct dir_struct *dir,
1736 struct index_state *istate,
1737 const char *pathname, int len)
1738 {
1739 if (index_file_exists(istate, pathname, len, ignore_case))
1740 return NULL;
1741
1742 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1743 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1744 }
1745
1746 struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1747 struct index_state *istate,
1748 const char *pathname, int len)
1749 {
1750 if (!index_name_is_other(istate, pathname, len))
1751 return NULL;
1752
1753 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1754 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1755 }
1756
1757 enum exist_status {
1758 index_nonexistent = 0,
1759 index_directory,
1760 index_gitdir
1761 };
1762
1763 /*
1764 * Do not use the alphabetically sorted index to look up
1765 * the directory name; instead, use the case insensitive
1766 * directory hash.
1767 */
1768 static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1769 const char *dirname, int len)
1770 {
1771 struct cache_entry *ce;
1772
1773 if (index_dir_exists(istate, dirname, len))
1774 return index_directory;
1775
1776 ce = index_file_exists(istate, dirname, len, ignore_case);
1777 if (ce && S_ISGITLINK(ce->ce_mode))
1778 return index_gitdir;
1779
1780 return index_nonexistent;
1781 }
1782
1783 /*
1784 * The index sorts alphabetically by entry name, which
1785 * means that a gitlink sorts as '\0' at the end, while
1786 * a directory (which is defined not as an entry, but as
1787 * the files it contains) will sort with the '/' at the
1788 * end.
1789 */
1790 static enum exist_status directory_exists_in_index(struct index_state *istate,
1791 const char *dirname, int len)
1792 {
1793 int pos;
1794
1795 if (ignore_case)
1796 return directory_exists_in_index_icase(istate, dirname, len);
1797
1798 pos = index_name_pos(istate, dirname, len);
1799 if (pos < 0)
1800 pos = -pos-1;
1801 while (pos < istate->cache_nr) {
1802 const struct cache_entry *ce = istate->cache[pos++];
1803 unsigned char endchar;
1804
1805 if (strncmp(ce->name, dirname, len))
1806 break;
1807 endchar = ce->name[len];
1808 if (endchar > '/')
1809 break;
1810 if (endchar == '/')
1811 return index_directory;
1812 if (!endchar && S_ISGITLINK(ce->ce_mode))
1813 return index_gitdir;
1814 }
1815 return index_nonexistent;
1816 }
1817
1818 /*
1819 * When we find a directory when traversing the filesystem, we
1820 * have three distinct cases:
1821 *
1822 * - ignore it
1823 * - see it as a directory
1824 * - recurse into it
1825 *
1826 * and which one we choose depends on a combination of existing
1827 * git index contents and the flags passed into the directory
1828 * traversal routine.
1829 *
1830 * Case 1: If we *already* have entries in the index under that
1831 * directory name, we always recurse into the directory to see
1832 * all the files.
1833 *
1834 * Case 2: If we *already* have that directory name as a gitlink,
1835 * we always continue to see it as a gitlink, regardless of whether
1836 * there is an actual git directory there or not (it might not
1837 * be checked out as a subproject!)
1838 *
1839 * Case 3: if we didn't have it in the index previously, we
1840 * have a few sub-cases:
1841 *
1842 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1843 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1844 * also true, in which case we need to check if it contains any
1845 * untracked and / or ignored files.
1846 * (b) if it looks like a git directory and we don't have the
1847 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1848 * show it as a directory.
1849 * (c) otherwise, we recurse into it.
1850 */
1851 static enum path_treatment treat_directory(struct dir_struct *dir,
1852 struct index_state *istate,
1853 struct untracked_cache_dir *untracked,
1854 const char *dirname, int len, int baselen, int excluded,
1855 const struct pathspec *pathspec)
1856 {
1857 /*
1858 * WARNING: From this function, you can return path_recurse or you
1859 * can call read_directory_recursive() (or neither), but
1860 * you CAN'T DO BOTH.
1861 */
1862 enum path_treatment state;
1863 int matches_how = 0;
1864 int check_only, stop_early;
1865 int old_ignored_nr, old_untracked_nr;
1866 /* The "len-1" is to strip the final '/' */
1867 enum exist_status status = directory_exists_in_index(istate, dirname, len-1);
1868
1869 if (status == index_directory)
1870 return path_recurse;
1871 if (status == index_gitdir)
1872 return path_none;
1873 if (status != index_nonexistent)
1874 BUG("Unhandled value for directory_exists_in_index: %d\n", status);
1875
1876 /*
1877 * We don't want to descend into paths that don't match the necessary
1878 * patterns. Clearly, if we don't have a pathspec, then we can't check
1879 * for matching patterns. Also, if (excluded) then we know we matched
1880 * the exclusion patterns so as an optimization we can skip checking
1881 * for matching patterns.
1882 */
1883 if (pathspec && !excluded) {
1884 matches_how = match_pathspec_with_flags(istate, pathspec,
1885 dirname, len,
1886 0 /* prefix */,
1887 NULL /* seen */,
1888 DO_MATCH_LEADING_PATHSPEC);
1889 if (!matches_how)
1890 return path_none;
1891 }
1892
1893
1894 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1895 !(dir->flags & DIR_NO_GITLINKS)) {
1896 /*
1897 * Determine if `dirname` is a nested repo by confirming that:
1898 * 1) we are in a nonbare repository, and
1899 * 2) `dirname` is not an immediate parent of `the_repository->gitdir`,
1900 * which could occur if the git_dir or worktree location was
1901 * manually configured by the user; see t2205 testcases 1-3 for
1902 * examples where this matters
1903 */
1904 int nested_repo;
1905 struct strbuf sb = STRBUF_INIT;
1906 strbuf_addstr(&sb, dirname);
1907 nested_repo = is_nonbare_repository_dir(&sb);
1908
1909 if (nested_repo) {
1910 char *real_dirname, *real_gitdir;
1911 strbuf_addstr(&sb, ".git");
1912 real_dirname = real_pathdup(sb.buf, 1);
1913 real_gitdir = real_pathdup(the_repository->gitdir, 1);
1914
1915 nested_repo = !!strcmp(real_dirname, real_gitdir);
1916 free(real_gitdir);
1917 free(real_dirname);
1918 }
1919 strbuf_release(&sb);
1920
1921 if (nested_repo) {
1922 if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1923 (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC))
1924 return path_none;
1925 return excluded ? path_excluded : path_untracked;
1926 }
1927 }
1928
1929 if (!(dir->flags & DIR_SHOW_OTHER_DIRECTORIES)) {
1930 if (excluded &&
1931 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1932 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1933
1934 /*
1935 * This is an excluded directory and we are
1936 * showing ignored paths that match an exclude
1937 * pattern. (e.g. show directory as ignored
1938 * only if it matches an exclude pattern).
1939 * This path will either be 'path_excluded`
1940 * (if we are showing empty directories or if
1941 * the directory is not empty), or will be
1942 * 'path_none' (empty directory, and we are
1943 * not showing empty directories).
1944 */
1945 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1946 return path_excluded;
1947
1948 if (read_directory_recursive(dir, istate, dirname, len,
1949 untracked, 1, 1, pathspec) == path_excluded)
1950 return path_excluded;
1951
1952 return path_none;
1953 }
1954 return path_recurse;
1955 }
1956
1957 assert(dir->flags & DIR_SHOW_OTHER_DIRECTORIES);
1958
1959 /*
1960 * If we have a pathspec which could match something _below_ this
1961 * directory (e.g. when checking 'subdir/' having a pathspec like
1962 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1963 * need to recurse.
1964 */
1965 if (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC)
1966 return path_recurse;
1967
1968 /* Special cases for where this directory is excluded/ignored */
1969 if (excluded) {
1970 /*
1971 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1972 * hiding empty directories, there is no need to
1973 * recurse into an ignored directory.
1974 */
1975 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1976 return path_excluded;
1977
1978 /*
1979 * Even if we are hiding empty directories, we can still avoid
1980 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
1981 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1982 */
1983 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
1984 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1985 return path_excluded;
1986 }
1987
1988 /*
1989 * Other than the path_recurse case above, we only need to
1990 * recurse into untracked directories if any of the following
1991 * bits is set:
1992 * - DIR_SHOW_IGNORED (because then we need to determine if
1993 * there are ignored entries below)
1994 * - DIR_SHOW_IGNORED_TOO (same as above)
1995 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
1996 * the directory is empty)
1997 */
1998 if (!excluded &&
1999 !(dir->flags & (DIR_SHOW_IGNORED |
2000 DIR_SHOW_IGNORED_TOO |
2001 DIR_HIDE_EMPTY_DIRECTORIES))) {
2002 return path_untracked;
2003 }
2004
2005 /*
2006 * Even if we don't want to know all the paths under an untracked or
2007 * ignored directory, we may still need to go into the directory to
2008 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2009 * an empty directory should be path_none instead of path_excluded or
2010 * path_untracked).
2011 */
2012 check_only = ((dir->flags & DIR_HIDE_EMPTY_DIRECTORIES) &&
2013 !(dir->flags & DIR_SHOW_IGNORED_TOO));
2014
2015 /*
2016 * However, there's another optimization possible as a subset of
2017 * check_only, based on the cases we have to consider:
2018 * A) Directory matches no exclude patterns:
2019 * * Directory is empty => path_none
2020 * * Directory has an untracked file under it => path_untracked
2021 * * Directory has only ignored files under it => path_excluded
2022 * B) Directory matches an exclude pattern:
2023 * * Directory is empty => path_none
2024 * * Directory has an untracked file under it => path_excluded
2025 * * Directory has only ignored files under it => path_excluded
2026 * In case A, we can exit as soon as we've found an untracked
2027 * file but otherwise have to walk all files. In case B, though,
2028 * we can stop at the first file we find under the directory.
2029 */
2030 stop_early = check_only && excluded;
2031
2032 /*
2033 * If /every/ file within an untracked directory is ignored, then
2034 * we want to treat the directory as ignored (for e.g. status
2035 * --porcelain), without listing the individual ignored files
2036 * underneath. To do so, we'll save the current ignored_nr, and
2037 * pop all the ones added after it if it turns out the entire
2038 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2039 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2040 * untracked paths so will need to pop all those off the last
2041 * after we traverse.
2042 */
2043 old_ignored_nr = dir->ignored_nr;
2044 old_untracked_nr = dir->nr;
2045
2046 /* Actually recurse into dirname now, we'll fixup the state later. */
2047 untracked = lookup_untracked(dir->untracked, untracked,
2048 dirname + baselen, len - baselen);
2049 state = read_directory_recursive(dir, istate, dirname, len, untracked,
2050 check_only, stop_early, pathspec);
2051
2052 /* There are a variety of reasons we may need to fixup the state... */
2053 if (state == path_excluded) {
2054 /* state == path_excluded implies all paths under
2055 * dirname were ignored...
2056 *
2057 * if running e.g. `git status --porcelain --ignored=matching`,
2058 * then we want to see the subpaths that are ignored.
2059 *
2060 * if running e.g. just `git status --porcelain`, then
2061 * we just want the directory itself to be listed as ignored
2062 * and not the individual paths underneath.
2063 */
2064 int want_ignored_subpaths =
2065 ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2066 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING));
2067
2068 if (want_ignored_subpaths) {
2069 /*
2070 * with --ignored=matching, we want the subpaths
2071 * INSTEAD of the directory itself.
2072 */
2073 state = path_none;
2074 } else {
2075 int i;
2076 for (i = old_ignored_nr + 1; i<dir->ignored_nr; ++i)
2077 FREE_AND_NULL(dir->ignored[i]);
2078 dir->ignored_nr = old_ignored_nr;
2079 }
2080 }
2081
2082 /*
2083 * We may need to ignore some of the untracked paths we found while
2084 * traversing subdirectories.
2085 */
2086 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2087 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2088 int i;
2089 for (i = old_untracked_nr + 1; i<dir->nr; ++i)
2090 FREE_AND_NULL(dir->entries[i]);
2091 dir->nr = old_untracked_nr;
2092 }
2093
2094 /*
2095 * If there is nothing under the current directory and we are not
2096 * hiding empty directories, then we need to report on the
2097 * untracked or ignored status of the directory itself.
2098 */
2099 if (state == path_none && !(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
2100 state = excluded ? path_excluded : path_untracked;
2101
2102 return state;
2103 }
2104
2105 /*
2106 * This is an inexact early pruning of any recursive directory
2107 * reading - if the path cannot possibly be in the pathspec,
2108 * return true, and we'll skip it early.
2109 */
2110 static int simplify_away(const char *path, int pathlen,
2111 const struct pathspec *pathspec)
2112 {
2113 int i;
2114
2115 if (!pathspec || !pathspec->nr)
2116 return 0;
2117
2118 GUARD_PATHSPEC(pathspec,
2119 PATHSPEC_FROMTOP |
2120 PATHSPEC_MAXDEPTH |
2121 PATHSPEC_LITERAL |
2122 PATHSPEC_GLOB |
2123 PATHSPEC_ICASE |
2124 PATHSPEC_EXCLUDE |
2125 PATHSPEC_ATTR);
2126
2127 for (i = 0; i < pathspec->nr; i++) {
2128 const struct pathspec_item *item = &pathspec->items[i];
2129 int len = item->nowildcard_len;
2130
2131 if (len > pathlen)
2132 len = pathlen;
2133 if (!ps_strncmp(item, item->match, path, len))
2134 return 0;
2135 }
2136
2137 return 1;
2138 }
2139
2140 /*
2141 * This function tells us whether an excluded path matches a
2142 * list of "interesting" pathspecs. That is, whether a path matched
2143 * by any of the pathspecs could possibly be ignored by excluding
2144 * the specified path. This can happen if:
2145 *
2146 * 1. the path is mentioned explicitly in the pathspec
2147 *
2148 * 2. the path is a directory prefix of some element in the
2149 * pathspec
2150 */
2151 static int exclude_matches_pathspec(const char *path, int pathlen,
2152 const struct pathspec *pathspec)
2153 {
2154 int i;
2155
2156 if (!pathspec || !pathspec->nr)
2157 return 0;
2158
2159 GUARD_PATHSPEC(pathspec,
2160 PATHSPEC_FROMTOP |
2161 PATHSPEC_MAXDEPTH |
2162 PATHSPEC_LITERAL |
2163 PATHSPEC_GLOB |
2164 PATHSPEC_ICASE |
2165 PATHSPEC_EXCLUDE);
2166
2167 for (i = 0; i < pathspec->nr; i++) {
2168 const struct pathspec_item *item = &pathspec->items[i];
2169 int len = item->nowildcard_len;
2170
2171 if (len == pathlen &&
2172 !ps_strncmp(item, item->match, path, pathlen))
2173 return 1;
2174 if (len > pathlen &&
2175 item->match[pathlen] == '/' &&
2176 !ps_strncmp(item, item->match, path, pathlen))
2177 return 1;
2178 }
2179 return 0;
2180 }
2181
2182 static int get_index_dtype(struct index_state *istate,
2183 const char *path, int len)
2184 {
2185 int pos;
2186 const struct cache_entry *ce;
2187
2188 ce = index_file_exists(istate, path, len, 0);
2189 if (ce) {
2190 if (!ce_uptodate(ce))
2191 return DT_UNKNOWN;
2192 if (S_ISGITLINK(ce->ce_mode))
2193 return DT_DIR;
2194 /*
2195 * Nobody actually cares about the
2196 * difference between DT_LNK and DT_REG
2197 */
2198 return DT_REG;
2199 }
2200
2201 /* Try to look it up as a directory */
2202 pos = index_name_pos(istate, path, len);
2203 if (pos >= 0)
2204 return DT_UNKNOWN;
2205 pos = -pos-1;
2206 while (pos < istate->cache_nr) {
2207 ce = istate->cache[pos++];
2208 if (strncmp(ce->name, path, len))
2209 break;
2210 if (ce->name[len] > '/')
2211 break;
2212 if (ce->name[len] < '/')
2213 continue;
2214 if (!ce_uptodate(ce))
2215 break; /* continue? */
2216 return DT_DIR;
2217 }
2218 return DT_UNKNOWN;
2219 }
2220
2221 static int resolve_dtype(int dtype, struct index_state *istate,
2222 const char *path, int len)
2223 {
2224 struct stat st;
2225
2226 if (dtype != DT_UNKNOWN)
2227 return dtype;
2228 dtype = get_index_dtype(istate, path, len);
2229 if (dtype != DT_UNKNOWN)
2230 return dtype;
2231 if (lstat(path, &st))
2232 return dtype;
2233 if (S_ISREG(st.st_mode))
2234 return DT_REG;
2235 if (S_ISDIR(st.st_mode))
2236 return DT_DIR;
2237 if (S_ISLNK(st.st_mode))
2238 return DT_LNK;
2239 return dtype;
2240 }
2241
2242 static enum path_treatment treat_path_fast(struct dir_struct *dir,
2243 struct cached_dir *cdir,
2244 struct index_state *istate,
2245 struct strbuf *path,
2246 int baselen,
2247 const struct pathspec *pathspec)
2248 {
2249 /*
2250 * WARNING: From this function, you can return path_recurse or you
2251 * can call read_directory_recursive() (or neither), but
2252 * you CAN'T DO BOTH.
2253 */
2254 strbuf_setlen(path, baselen);
2255 if (!cdir->ucd) {
2256 strbuf_addstr(path, cdir->file);
2257 return path_untracked;
2258 }
2259 strbuf_addstr(path, cdir->ucd->name);
2260 /* treat_one_path() does this before it calls treat_directory() */
2261 strbuf_complete(path, '/');
2262 if (cdir->ucd->check_only)
2263 /*
2264 * check_only is set as a result of treat_directory() getting
2265 * to its bottom. Verify again the same set of directories
2266 * with check_only set.
2267 */
2268 return read_directory_recursive(dir, istate, path->buf, path->len,
2269 cdir->ucd, 1, 0, pathspec);
2270 /*
2271 * We get path_recurse in the first run when
2272 * directory_exists_in_index() returns index_nonexistent. We
2273 * are sure that new changes in the index does not impact the
2274 * outcome. Return now.
2275 */
2276 return path_recurse;
2277 }
2278
2279 static enum path_treatment treat_path(struct dir_struct *dir,
2280 struct untracked_cache_dir *untracked,
2281 struct cached_dir *cdir,
2282 struct index_state *istate,
2283 struct strbuf *path,
2284 int baselen,
2285 const struct pathspec *pathspec)
2286 {
2287 int has_path_in_index, dtype, excluded;
2288
2289 if (!cdir->d_name)
2290 return treat_path_fast(dir, cdir, istate, path,
2291 baselen, pathspec);
2292 if (is_dot_or_dotdot(cdir->d_name) || !fspathcmp(cdir->d_name, ".git"))
2293 return path_none;
2294 strbuf_setlen(path, baselen);
2295 strbuf_addstr(path, cdir->d_name);
2296 if (simplify_away(path->buf, path->len, pathspec))
2297 return path_none;
2298
2299 dtype = resolve_dtype(cdir->d_type, istate, path->buf, path->len);
2300
2301 /* Always exclude indexed files */
2302 has_path_in_index = !!index_file_exists(istate, path->buf, path->len,
2303 ignore_case);
2304 if (dtype != DT_DIR && has_path_in_index)
2305 return path_none;
2306
2307 /*
2308 * When we are looking at a directory P in the working tree,
2309 * there are three cases:
2310 *
2311 * (1) P exists in the index. Everything inside the directory P in
2312 * the working tree needs to go when P is checked out from the
2313 * index.
2314 *
2315 * (2) P does not exist in the index, but there is P/Q in the index.
2316 * We know P will stay a directory when we check out the contents
2317 * of the index, but we do not know yet if there is a directory
2318 * P/Q in the working tree to be killed, so we need to recurse.
2319 *
2320 * (3) P does not exist in the index, and there is no P/Q in the index
2321 * to require P to be a directory, either. Only in this case, we
2322 * know that everything inside P will not be killed without
2323 * recursing.
2324 */
2325 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
2326 (dtype == DT_DIR) &&
2327 !has_path_in_index &&
2328 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
2329 return path_none;
2330
2331 excluded = is_excluded(dir, istate, path->buf, &dtype);
2332
2333 /*
2334 * Excluded? If we don't explicitly want to show
2335 * ignored files, ignore it
2336 */
2337 if (excluded && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
2338 return path_excluded;
2339
2340 switch (dtype) {
2341 default:
2342 return path_none;
2343 case DT_DIR:
2344 /*
2345 * WARNING: Do not ignore/amend the return value from
2346 * treat_directory(), and especially do not change it to return
2347 * path_recurse as that can cause exponential slowdown.
2348 * Instead, modify treat_directory() to return the right value.
2349 */
2350 strbuf_addch(path, '/');
2351 return treat_directory(dir, istate, untracked,
2352 path->buf, path->len,
2353 baselen, excluded, pathspec);
2354 case DT_REG:
2355 case DT_LNK:
2356 if (pathspec &&
2357 !match_pathspec(istate, pathspec, path->buf, path->len,
2358 0 /* prefix */, NULL /* seen */,
2359 0 /* is_dir */))
2360 return path_none;
2361 if (excluded)
2362 return path_excluded;
2363 return path_untracked;
2364 }
2365 }
2366
2367 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
2368 {
2369 if (!dir)
2370 return;
2371 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
2372 dir->untracked_alloc);
2373 dir->untracked[dir->untracked_nr++] = xstrdup(name);
2374 }
2375
2376 static int valid_cached_dir(struct dir_struct *dir,
2377 struct untracked_cache_dir *untracked,
2378 struct index_state *istate,
2379 struct strbuf *path,
2380 int check_only)
2381 {
2382 struct stat st;
2383
2384 if (!untracked)
2385 return 0;
2386
2387 /*
2388 * With fsmonitor, we can trust the untracked cache's valid field.
2389 */
2390 refresh_fsmonitor(istate);
2391 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
2392 if (lstat(path->len ? path->buf : ".", &st)) {
2393 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
2394 return 0;
2395 }
2396 if (!untracked->valid ||
2397 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
2398 fill_stat_data(&untracked->stat_data, &st);
2399 return 0;
2400 }
2401 }
2402
2403 if (untracked->check_only != !!check_only)
2404 return 0;
2405
2406 /*
2407 * prep_exclude will be called eventually on this directory,
2408 * but it's called much later in last_matching_pattern(). We
2409 * need it now to determine the validity of the cache for this
2410 * path. The next calls will be nearly no-op, the way
2411 * prep_exclude() is designed.
2412 */
2413 if (path->len && path->buf[path->len - 1] != '/') {
2414 strbuf_addch(path, '/');
2415 prep_exclude(dir, istate, path->buf, path->len);
2416 strbuf_setlen(path, path->len - 1);
2417 } else
2418 prep_exclude(dir, istate, path->buf, path->len);
2419
2420 /* hopefully prep_exclude() haven't invalidated this entry... */
2421 return untracked->valid;
2422 }
2423
2424 static int open_cached_dir(struct cached_dir *cdir,
2425 struct dir_struct *dir,
2426 struct untracked_cache_dir *untracked,
2427 struct index_state *istate,
2428 struct strbuf *path,
2429 int check_only)
2430 {
2431 const char *c_path;
2432
2433 memset(cdir, 0, sizeof(*cdir));
2434 cdir->untracked = untracked;
2435 if (valid_cached_dir(dir, untracked, istate, path, check_only))
2436 return 0;
2437 c_path = path->len ? path->buf : ".";
2438 cdir->fdir = opendir(c_path);
2439 if (!cdir->fdir)
2440 warning_errno(_("could not open directory '%s'"), c_path);
2441 if (dir->untracked) {
2442 invalidate_directory(dir->untracked, untracked);
2443 dir->untracked->dir_opened++;
2444 }
2445 if (!cdir->fdir)
2446 return -1;
2447 return 0;
2448 }
2449
2450 static int read_cached_dir(struct cached_dir *cdir)
2451 {
2452 struct dirent *de;
2453
2454 if (cdir->fdir) {
2455 de = readdir_skip_dot_and_dotdot(cdir->fdir);
2456 if (!de) {
2457 cdir->d_name = NULL;
2458 cdir->d_type = DT_UNKNOWN;
2459 return -1;
2460 }
2461 cdir->d_name = de->d_name;
2462 cdir->d_type = DTYPE(de);
2463 return 0;
2464 }
2465 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
2466 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
2467 if (!d->recurse) {
2468 cdir->nr_dirs++;
2469 continue;
2470 }
2471 cdir->ucd = d;
2472 cdir->nr_dirs++;
2473 return 0;
2474 }
2475 cdir->ucd = NULL;
2476 if (cdir->nr_files < cdir->untracked->untracked_nr) {
2477 struct untracked_cache_dir *d = cdir->untracked;
2478 cdir->file = d->untracked[cdir->nr_files++];
2479 return 0;
2480 }
2481 return -1;
2482 }
2483
2484 static void close_cached_dir(struct cached_dir *cdir)
2485 {
2486 if (cdir->fdir)
2487 closedir(cdir->fdir);
2488 /*
2489 * We have gone through this directory and found no untracked
2490 * entries. Mark it valid.
2491 */
2492 if (cdir->untracked) {
2493 cdir->untracked->valid = 1;
2494 cdir->untracked->recurse = 1;
2495 }
2496 }
2497
2498 static void add_path_to_appropriate_result_list(struct dir_struct *dir,
2499 struct untracked_cache_dir *untracked,
2500 struct cached_dir *cdir,
2501 struct index_state *istate,
2502 struct strbuf *path,
2503 int baselen,
2504 const struct pathspec *pathspec,
2505 enum path_treatment state)
2506 {
2507 /* add the path to the appropriate result list */
2508 switch (state) {
2509 case path_excluded:
2510 if (dir->flags & DIR_SHOW_IGNORED)
2511 dir_add_name(dir, istate, path->buf, path->len);
2512 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2513 ((dir->flags & DIR_COLLECT_IGNORED) &&
2514 exclude_matches_pathspec(path->buf, path->len,
2515 pathspec)))
2516 dir_add_ignored(dir, istate, path->buf, path->len);
2517 break;
2518
2519 case path_untracked:
2520 if (dir->flags & DIR_SHOW_IGNORED)
2521 break;
2522 dir_add_name(dir, istate, path->buf, path->len);
2523 if (cdir->fdir)
2524 add_untracked(untracked, path->buf + baselen);
2525 break;
2526
2527 default:
2528 break;
2529 }
2530 }
2531
2532 /*
2533 * Read a directory tree. We currently ignore anything but
2534 * directories, regular files and symlinks. That's because git
2535 * doesn't handle them at all yet. Maybe that will change some
2536 * day.
2537 *
2538 * Also, we ignore the name ".git" (even if it is not a directory).
2539 * That likely will not change.
2540 *
2541 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2542 * to signal that a file was found. This is the least significant value that
2543 * indicates that a file was encountered that does not depend on the order of
2544 * whether an untracked or excluded path was encountered first.
2545 *
2546 * Returns the most significant path_treatment value encountered in the scan.
2547 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2548 * significant path_treatment value that will be returned.
2549 */
2550
2551 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
2552 struct index_state *istate, const char *base, int baselen,
2553 struct untracked_cache_dir *untracked, int check_only,
2554 int stop_at_first_file, const struct pathspec *pathspec)
2555 {
2556 /*
2557 * WARNING: Do NOT recurse unless path_recurse is returned from
2558 * treat_path(). Recursing on any other return value
2559 * can result in exponential slowdown.
2560 */
2561 struct cached_dir cdir;
2562 enum path_treatment state, subdir_state, dir_state = path_none;
2563 struct strbuf path = STRBUF_INIT;
2564
2565 strbuf_add(&path, base, baselen);
2566
2567 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
2568 goto out;
2569 dir->visited_directories++;
2570
2571 if (untracked)
2572 untracked->check_only = !!check_only;
2573
2574 while (!read_cached_dir(&cdir)) {
2575 /* check how the file or directory should be treated */
2576 state = treat_path(dir, untracked, &cdir, istate, &path,
2577 baselen, pathspec);
2578 dir->visited_paths++;
2579
2580 if (state > dir_state)
2581 dir_state = state;
2582
2583 /* recurse into subdir if instructed by treat_path */
2584 if (state == path_recurse) {
2585 struct untracked_cache_dir *ud;
2586 ud = lookup_untracked(dir->untracked, untracked,
2587 path.buf + baselen,
2588 path.len - baselen);
2589 subdir_state =
2590 read_directory_recursive(dir, istate, path.buf,
2591 path.len, ud,
2592 check_only, stop_at_first_file, pathspec);
2593 if (subdir_state > dir_state)
2594 dir_state = subdir_state;
2595
2596 if (pathspec &&
2597 !match_pathspec(istate, pathspec, path.buf, path.len,
2598 0 /* prefix */, NULL,
2599 0 /* do NOT special case dirs */))
2600 state = path_none;
2601 }
2602
2603 if (check_only) {
2604 if (stop_at_first_file) {
2605 /*
2606 * If stopping at first file, then
2607 * signal that a file was found by
2608 * returning `path_excluded`. This is
2609 * to return a consistent value
2610 * regardless of whether an ignored or
2611 * excluded file happened to be
2612 * encountered 1st.
2613 *
2614 * In current usage, the
2615 * `stop_at_first_file` is passed when
2616 * an ancestor directory has matched
2617 * an exclude pattern, so any found
2618 * files will be excluded.
2619 */
2620 if (dir_state >= path_excluded) {
2621 dir_state = path_excluded;
2622 break;
2623 }
2624 }
2625
2626 /* abort early if maximum state has been reached */
2627 if (dir_state == path_untracked) {
2628 if (cdir.fdir)
2629 add_untracked(untracked, path.buf + baselen);
2630 break;
2631 }
2632 /* skip the add_path_to_appropriate_result_list() */
2633 continue;
2634 }
2635
2636 add_path_to_appropriate_result_list(dir, untracked, &cdir,
2637 istate, &path, baselen,
2638 pathspec, state);
2639 }
2640 close_cached_dir(&cdir);
2641 out:
2642 strbuf_release(&path);
2643
2644 return dir_state;
2645 }
2646
2647 int cmp_dir_entry(const void *p1, const void *p2)
2648 {
2649 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2650 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2651
2652 return name_compare(e1->name, e1->len, e2->name, e2->len);
2653 }
2654
2655 /* check if *out lexically strictly contains *in */
2656 int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2657 {
2658 return (out->len < in->len) &&
2659 (out->name[out->len - 1] == '/') &&
2660 !memcmp(out->name, in->name, out->len);
2661 }
2662
2663 static int treat_leading_path(struct dir_struct *dir,
2664 struct index_state *istate,
2665 const char *path, int len,
2666 const struct pathspec *pathspec)
2667 {
2668 struct strbuf sb = STRBUF_INIT;
2669 struct strbuf subdir = STRBUF_INIT;
2670 int prevlen, baselen;
2671 const char *cp;
2672 struct cached_dir cdir;
2673 enum path_treatment state = path_none;
2674
2675 /*
2676 * For each directory component of path, we are going to check whether
2677 * that path is relevant given the pathspec. For example, if path is
2678 * foo/bar/baz/
2679 * then we will ask treat_path() whether we should go into foo, then
2680 * whether we should go into bar, then whether baz is relevant.
2681 * Checking each is important because e.g. if path is
2682 * .git/info/
2683 * then we need to check .git to know we shouldn't traverse it.
2684 * If the return from treat_path() is:
2685 * * path_none, for any path, we return false.
2686 * * path_recurse, for all path components, we return true
2687 * * <anything else> for some intermediate component, we make sure
2688 * to add that path to the relevant list but return false
2689 * signifying that we shouldn't recurse into it.
2690 */
2691
2692 while (len && path[len - 1] == '/')
2693 len--;
2694 if (!len)
2695 return 1;
2696
2697 memset(&cdir, 0, sizeof(cdir));
2698 cdir.d_type = DT_DIR;
2699 baselen = 0;
2700 prevlen = 0;
2701 while (1) {
2702 prevlen = baselen + !!baselen;
2703 cp = path + prevlen;
2704 cp = memchr(cp, '/', path + len - cp);
2705 if (!cp)
2706 baselen = len;
2707 else
2708 baselen = cp - path;
2709 strbuf_reset(&sb);
2710 strbuf_add(&sb, path, baselen);
2711 if (!is_directory(sb.buf))
2712 break;
2713 strbuf_reset(&sb);
2714 strbuf_add(&sb, path, prevlen);
2715 strbuf_reset(&subdir);
2716 strbuf_add(&subdir, path+prevlen, baselen-prevlen);
2717 cdir.d_name = subdir.buf;
2718 state = treat_path(dir, NULL, &cdir, istate, &sb, prevlen, pathspec);
2719
2720 if (state != path_recurse)
2721 break; /* do not recurse into it */
2722 if (len <= baselen)
2723 break; /* finished checking */
2724 }
2725 add_path_to_appropriate_result_list(dir, NULL, &cdir, istate,
2726 &sb, baselen, pathspec,
2727 state);
2728
2729 strbuf_release(&subdir);
2730 strbuf_release(&sb);
2731 return state == path_recurse;
2732 }
2733
2734 static const char *get_ident_string(void)
2735 {
2736 static struct strbuf sb = STRBUF_INIT;
2737 struct utsname uts;
2738
2739 if (sb.len)
2740 return sb.buf;
2741 if (uname(&uts) < 0)
2742 die_errno(_("failed to get kernel name and information"));
2743 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2744 uts.sysname);
2745 return sb.buf;
2746 }
2747
2748 static int ident_in_untracked(const struct untracked_cache *uc)
2749 {
2750 /*
2751 * Previous git versions may have saved many NUL separated
2752 * strings in the "ident" field, but it is insane to manage
2753 * many locations, so just take care of the first one.
2754 */
2755
2756 return !strcmp(uc->ident.buf, get_ident_string());
2757 }
2758
2759 static void set_untracked_ident(struct untracked_cache *uc)
2760 {
2761 strbuf_reset(&uc->ident);
2762 strbuf_addstr(&uc->ident, get_ident_string());
2763
2764 /*
2765 * This strbuf used to contain a list of NUL separated
2766 * strings, so save NUL too for backward compatibility.
2767 */
2768 strbuf_addch(&uc->ident, 0);
2769 }
2770
2771 static unsigned new_untracked_cache_flags(struct index_state *istate)
2772 {
2773 struct repository *repo = istate->repo;
2774 char *val;
2775
2776 /*
2777 * This logic is coordinated with the setting of these flags in
2778 * wt-status.c#wt_status_collect_untracked(), and the evaluation
2779 * of the config setting in commit.c#git_status_config()
2780 */
2781 if (!repo_config_get_string(repo, "status.showuntrackedfiles", &val) &&
2782 !strcmp(val, "all"))
2783 return 0;
2784
2785 /*
2786 * The default, if "all" is not set, is "normal" - leading us here.
2787 * If the value is "none" then it really doesn't matter.
2788 */
2789 return DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2790 }
2791
2792 static void new_untracked_cache(struct index_state *istate, int flags)
2793 {
2794 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2795 strbuf_init(&uc->ident, 100);
2796 uc->exclude_per_dir = ".gitignore";
2797 uc->dir_flags = flags >= 0 ? flags : new_untracked_cache_flags(istate);
2798 set_untracked_ident(uc);
2799 istate->untracked = uc;
2800 istate->cache_changed |= UNTRACKED_CHANGED;
2801 }
2802
2803 void add_untracked_cache(struct index_state *istate)
2804 {
2805 if (!istate->untracked) {
2806 new_untracked_cache(istate, -1);
2807 } else {
2808 if (!ident_in_untracked(istate->untracked)) {
2809 free_untracked_cache(istate->untracked);
2810 new_untracked_cache(istate, -1);
2811 }
2812 }
2813 }
2814
2815 void remove_untracked_cache(struct index_state *istate)
2816 {
2817 if (istate->untracked) {
2818 free_untracked_cache(istate->untracked);
2819 istate->untracked = NULL;
2820 istate->cache_changed |= UNTRACKED_CHANGED;
2821 }
2822 }
2823
2824 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2825 int base_len,
2826 const struct pathspec *pathspec,
2827 struct index_state *istate)
2828 {
2829 struct untracked_cache_dir *root;
2830 static int untracked_cache_disabled = -1;
2831
2832 if (!dir->untracked)
2833 return NULL;
2834 if (untracked_cache_disabled < 0)
2835 untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2836 if (untracked_cache_disabled)
2837 return NULL;
2838
2839 /*
2840 * We only support $GIT_DIR/info/exclude and core.excludesfile
2841 * as the global ignore rule files. Any other additions
2842 * (e.g. from command line) invalidate the cache. This
2843 * condition also catches running setup_standard_excludes()
2844 * before setting dir->untracked!
2845 */
2846 if (dir->unmanaged_exclude_files)
2847 return NULL;
2848
2849 /*
2850 * Optimize for the main use case only: whole-tree git
2851 * status. More work involved in treat_leading_path() if we
2852 * use cache on just a subset of the worktree. pathspec
2853 * support could make the matter even worse.
2854 */
2855 if (base_len || (pathspec && pathspec->nr))
2856 return NULL;
2857
2858 /* We don't support collecting ignore files */
2859 if (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2860 DIR_COLLECT_IGNORED))
2861 return NULL;
2862
2863 /*
2864 * If we use .gitignore in the cache and now you change it to
2865 * .gitexclude, everything will go wrong.
2866 */
2867 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2868 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2869 return NULL;
2870
2871 /*
2872 * EXC_CMDL is not considered in the cache. If people set it,
2873 * skip the cache.
2874 */
2875 if (dir->exclude_list_group[EXC_CMDL].nr)
2876 return NULL;
2877
2878 if (!ident_in_untracked(dir->untracked)) {
2879 warning(_("untracked cache is disabled on this system or location"));
2880 return NULL;
2881 }
2882
2883 /*
2884 * If the untracked structure we received does not have the same flags
2885 * as requested in this run, we're going to need to either discard the
2886 * existing structure (and potentially later recreate), or bypass the
2887 * untracked cache mechanism for this run.
2888 */
2889 if (dir->flags != dir->untracked->dir_flags) {
2890 /*
2891 * If the untracked structure we received does not have the same flags
2892 * as configured, then we need to reset / create a new "untracked"
2893 * structure to match the new config.
2894 *
2895 * Keeping the saved and used untracked cache consistent with the
2896 * configuration provides an opportunity for frequent users of
2897 * "git status -uall" to leverage the untracked cache by aligning their
2898 * configuration - setting "status.showuntrackedfiles" to "all" or
2899 * "normal" as appropriate.
2900 *
2901 * Previously using -uall (or setting "status.showuntrackedfiles" to
2902 * "all") was incompatible with untracked cache and *consistently*
2903 * caused surprisingly bad performance (with fscache and fsmonitor
2904 * enabled) on Windows.
2905 *
2906 * IMPROVEMENT OPPORTUNITY: If we reworked the untracked cache storage
2907 * to not be as bound up with the desired output in a given run,
2908 * and instead iterated through and stored enough information to
2909 * correctly serve both "modes", then users could get peak performance
2910 * with or without '-uall' regardless of their
2911 * "status.showuntrackedfiles" config.
2912 */
2913 if (dir->untracked->dir_flags != new_untracked_cache_flags(istate)) {
2914 free_untracked_cache(istate->untracked);
2915 new_untracked_cache(istate, dir->flags);
2916 dir->untracked = istate->untracked;
2917 }
2918 else {
2919 /*
2920 * Current untracked cache data is consistent with config, but not
2921 * usable in this request/run; just bypass untracked cache.
2922 */
2923 return NULL;
2924 }
2925 }
2926
2927 if (!dir->untracked->root) {
2928 /* Untracked cache existed but is not initialized; fix that */
2929 FLEX_ALLOC_STR(dir->untracked->root, name, "");
2930 istate->cache_changed |= UNTRACKED_CHANGED;
2931 }
2932
2933 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2934 root = dir->untracked->root;
2935 if (!oideq(&dir->ss_info_exclude.oid,
2936 &dir->untracked->ss_info_exclude.oid)) {
2937 invalidate_gitignore(dir->untracked, root);
2938 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2939 }
2940 if (!oideq(&dir->ss_excludes_file.oid,
2941 &dir->untracked->ss_excludes_file.oid)) {
2942 invalidate_gitignore(dir->untracked, root);
2943 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2944 }
2945
2946 /* Make sure this directory is not dropped out at saving phase */
2947 root->recurse = 1;
2948 return root;
2949 }
2950
2951 static void emit_traversal_statistics(struct dir_struct *dir,
2952 struct repository *repo,
2953 const char *path,
2954 int path_len)
2955 {
2956 if (!trace2_is_enabled())
2957 return;
2958
2959 if (!path_len) {
2960 trace2_data_string("read_directory", repo, "path", "");
2961 } else {
2962 struct strbuf tmp = STRBUF_INIT;
2963 strbuf_add(&tmp, path, path_len);
2964 trace2_data_string("read_directory", repo, "path", tmp.buf);
2965 strbuf_release(&tmp);
2966 }
2967
2968 trace2_data_intmax("read_directory", repo,
2969 "directories-visited", dir->visited_directories);
2970 trace2_data_intmax("read_directory", repo,
2971 "paths-visited", dir->visited_paths);
2972
2973 if (!dir->untracked)
2974 return;
2975 trace2_data_intmax("read_directory", repo,
2976 "node-creation", dir->untracked->dir_created);
2977 trace2_data_intmax("read_directory", repo,
2978 "gitignore-invalidation",
2979 dir->untracked->gitignore_invalidated);
2980 trace2_data_intmax("read_directory", repo,
2981 "directory-invalidation",
2982 dir->untracked->dir_invalidated);
2983 trace2_data_intmax("read_directory", repo,
2984 "opendir", dir->untracked->dir_opened);
2985 }
2986
2987 int read_directory(struct dir_struct *dir, struct index_state *istate,
2988 const char *path, int len, const struct pathspec *pathspec)
2989 {
2990 struct untracked_cache_dir *untracked;
2991
2992 trace2_region_enter("dir", "read_directory", istate->repo);
2993 dir->visited_paths = 0;
2994 dir->visited_directories = 0;
2995
2996 if (has_symlink_leading_path(path, len)) {
2997 trace2_region_leave("dir", "read_directory", istate->repo);
2998 return dir->nr;
2999 }
3000
3001 untracked = validate_untracked_cache(dir, len, pathspec, istate);
3002 if (!untracked)
3003 /*
3004 * make sure untracked cache code path is disabled,
3005 * e.g. prep_exclude()
3006 */
3007 dir->untracked = NULL;
3008 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
3009 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
3010 QSORT(dir->entries, dir->nr, cmp_dir_entry);
3011 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
3012
3013 emit_traversal_statistics(dir, istate->repo, path, len);
3014
3015 trace2_region_leave("dir", "read_directory", istate->repo);
3016 if (dir->untracked) {
3017 static int force_untracked_cache = -1;
3018
3019 if (force_untracked_cache < 0)
3020 force_untracked_cache =
3021 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", -1);
3022 if (force_untracked_cache < 0)
3023 force_untracked_cache = (istate->repo->settings.core_untracked_cache == UNTRACKED_CACHE_WRITE);
3024 if (force_untracked_cache &&
3025 dir->untracked == istate->untracked &&
3026 (dir->untracked->dir_opened ||
3027 dir->untracked->gitignore_invalidated ||
3028 dir->untracked->dir_invalidated))
3029 istate->cache_changed |= UNTRACKED_CHANGED;
3030 if (dir->untracked != istate->untracked) {
3031 FREE_AND_NULL(dir->untracked);
3032 }
3033 }
3034
3035 return dir->nr;
3036 }
3037
3038 int file_exists(const char *f)
3039 {
3040 struct stat sb;
3041 return lstat(f, &sb) == 0;
3042 }
3043
3044 int repo_file_exists(struct repository *repo, const char *path)
3045 {
3046 if (repo != the_repository)
3047 BUG("do not know how to check file existence in arbitrary repo");
3048
3049 return file_exists(path);
3050 }
3051
3052 static int cmp_icase(char a, char b)
3053 {
3054 if (a == b)
3055 return 0;
3056 if (ignore_case)
3057 return toupper(a) - toupper(b);
3058 return a - b;
3059 }
3060
3061 /*
3062 * Given two normalized paths (a trailing slash is ok), if subdir is
3063 * outside dir, return -1. Otherwise return the offset in subdir that
3064 * can be used as relative path to dir.
3065 */
3066 int dir_inside_of(const char *subdir, const char *dir)
3067 {
3068 int offset = 0;
3069
3070 assert(dir && subdir && *dir && *subdir);
3071
3072 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
3073 dir++;
3074 subdir++;
3075 offset++;
3076 }
3077
3078 /* hel[p]/me vs hel[l]/yeah */
3079 if (*dir && *subdir)
3080 return -1;
3081
3082 if (!*subdir)
3083 return !*dir ? offset : -1; /* same dir */
3084
3085 /* foo/[b]ar vs foo/[] */
3086 if (is_dir_sep(dir[-1]))
3087 return is_dir_sep(subdir[-1]) ? offset : -1;
3088
3089 /* foo[/]bar vs foo[] */
3090 return is_dir_sep(*subdir) ? offset + 1 : -1;
3091 }
3092
3093 int is_inside_dir(const char *dir)
3094 {
3095 char *cwd;
3096 int rc;
3097
3098 if (!dir)
3099 return 0;
3100
3101 cwd = xgetcwd();
3102 rc = (dir_inside_of(cwd, dir) >= 0);
3103 free(cwd);
3104 return rc;
3105 }
3106
3107 int is_empty_dir(const char *path)
3108 {
3109 DIR *dir = opendir(path);
3110 struct dirent *e;
3111 int ret = 1;
3112
3113 if (!dir)
3114 return 0;
3115
3116 e = readdir_skip_dot_and_dotdot(dir);
3117 if (e)
3118 ret = 0;
3119
3120 closedir(dir);
3121 return ret;
3122 }
3123
3124 char *git_url_basename(const char *repo, int is_bundle, int is_bare)
3125 {
3126 const char *end = repo + strlen(repo), *start, *ptr;
3127 size_t len;
3128 char *dir;
3129
3130 /*
3131 * Skip scheme.
3132 */
3133 start = strstr(repo, "://");
3134 if (!start)
3135 start = repo;
3136 else
3137 start += 3;
3138
3139 /*
3140 * Skip authentication data. The stripping does happen
3141 * greedily, such that we strip up to the last '@' inside
3142 * the host part.
3143 */
3144 for (ptr = start; ptr < end && !is_dir_sep(*ptr); ptr++) {
3145 if (*ptr == '@')
3146 start = ptr + 1;
3147 }
3148
3149 /*
3150 * Strip trailing spaces, slashes and /.git
3151 */
3152 while (start < end && (is_dir_sep(end[-1]) || isspace(end[-1])))
3153 end--;
3154 if (end - start > 5 && is_dir_sep(end[-5]) &&
3155 !strncmp(end - 4, ".git", 4)) {
3156 end -= 5;
3157 while (start < end && is_dir_sep(end[-1]))
3158 end--;
3159 }
3160
3161 /*
3162 * It should not be possible to overflow `ptrdiff_t` by passing in an
3163 * insanely long URL, but GCC does not know that and will complain
3164 * without this check.
3165 */
3166 if (end - start < 0)
3167 die(_("No directory name could be guessed.\n"
3168 "Please specify a directory on the command line"));
3169
3170 /*
3171 * Strip trailing port number if we've got only a
3172 * hostname (that is, there is no dir separator but a
3173 * colon). This check is required such that we do not
3174 * strip URI's like '/foo/bar:2222.git', which should
3175 * result in a dir '2222' being guessed due to backwards
3176 * compatibility.
3177 */
3178 if (memchr(start, '/', end - start) == NULL
3179 && memchr(start, ':', end - start) != NULL) {
3180 ptr = end;
3181 while (start < ptr && isdigit(ptr[-1]) && ptr[-1] != ':')
3182 ptr--;
3183 if (start < ptr && ptr[-1] == ':')
3184 end = ptr - 1;
3185 }
3186
3187 /*
3188 * Find last component. To remain backwards compatible we
3189 * also regard colons as path separators, such that
3190 * cloning a repository 'foo:bar.git' would result in a
3191 * directory 'bar' being guessed.
3192 */
3193 ptr = end;
3194 while (start < ptr && !is_dir_sep(ptr[-1]) && ptr[-1] != ':')
3195 ptr--;
3196 start = ptr;
3197
3198 /*
3199 * Strip .{bundle,git}.
3200 */
3201 len = end - start;
3202 strip_suffix_mem(start, &len, is_bundle ? ".bundle" : ".git");
3203
3204 if (!len || (len == 1 && *start == '/'))
3205 die(_("No directory name could be guessed.\n"
3206 "Please specify a directory on the command line"));
3207
3208 if (is_bare)
3209 dir = xstrfmt("%.*s.git", (int)len, start);
3210 else
3211 dir = xstrndup(start, len);
3212 /*
3213 * Replace sequences of 'control' characters and whitespace
3214 * with one ascii space, remove leading and trailing spaces.
3215 */
3216 if (*dir) {
3217 char *out = dir;
3218 int prev_space = 1 /* strip leading whitespace */;
3219 for (end = dir; *end; ++end) {
3220 char ch = *end;
3221 if ((unsigned char)ch < '\x20')
3222 ch = '\x20';
3223 if (isspace(ch)) {
3224 if (prev_space)
3225 continue;
3226 prev_space = 1;
3227 } else
3228 prev_space = 0;
3229 *out++ = ch;
3230 }
3231 *out = '\0';
3232 if (out > dir && prev_space)
3233 out[-1] = '\0';
3234 }
3235 return dir;
3236 }
3237
3238 void strip_dir_trailing_slashes(char *dir)
3239 {
3240 char *end = dir + strlen(dir);
3241
3242 while (dir < end - 1 && is_dir_sep(end[-1]))
3243 end--;
3244 *end = '\0';
3245 }
3246
3247 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
3248 {
3249 DIR *dir;
3250 struct dirent *e;
3251 int ret = 0, original_len = path->len, len, kept_down = 0;
3252 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
3253 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
3254 int purge_original_cwd = (flag & REMOVE_DIR_PURGE_ORIGINAL_CWD);
3255 struct object_id submodule_head;
3256
3257 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
3258 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
3259 /* Do not descend and nuke a nested git work tree. */
3260 if (kept_up)
3261 *kept_up = 1;
3262 return 0;
3263 }
3264
3265 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
3266 dir = opendir(path->buf);
3267 if (!dir) {
3268 if (errno == ENOENT)
3269 return keep_toplevel ? -1 : 0;
3270 else if (errno == EACCES && !keep_toplevel)
3271 /*
3272 * An empty dir could be removable even if it
3273 * is unreadable:
3274 */
3275 return rmdir(path->buf);
3276 else
3277 return -1;
3278 }
3279 strbuf_complete(path, '/');
3280
3281 len = path->len;
3282 while ((e = readdir_skip_dot_and_dotdot(dir)) != NULL) {
3283 struct stat st;
3284
3285 strbuf_setlen(path, len);
3286 strbuf_addstr(path, e->d_name);
3287 if (lstat(path->buf, &st)) {
3288 if (errno == ENOENT)
3289 /*
3290 * file disappeared, which is what we
3291 * wanted anyway
3292 */
3293 continue;
3294 /* fall through */
3295 } else if (S_ISDIR(st.st_mode)) {
3296 if (!remove_dir_recurse(path, flag, &kept_down))
3297 continue; /* happy */
3298 } else if (!only_empty &&
3299 (!unlink(path->buf) || errno == ENOENT)) {
3300 continue; /* happy, too */
3301 }
3302
3303 /* path too long, stat fails, or non-directory still exists */
3304 ret = -1;
3305 break;
3306 }
3307 closedir(dir);
3308
3309 strbuf_setlen(path, original_len);
3310 if (!ret && !keep_toplevel && !kept_down) {
3311 if (!purge_original_cwd &&
3312 startup_info->original_cwd &&
3313 !strcmp(startup_info->original_cwd, path->buf))
3314 ret = -1; /* Do not remove current working directory */
3315 else
3316 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
3317 } else if (kept_up)
3318 /*
3319 * report the uplevel that it is not an error that we
3320 * did not rmdir() our directory.
3321 */
3322 *kept_up = !ret;
3323 return ret;
3324 }
3325
3326 int remove_dir_recursively(struct strbuf *path, int flag)
3327 {
3328 return remove_dir_recurse(path, flag, NULL);
3329 }
3330
3331 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
3332
3333 void setup_standard_excludes(struct dir_struct *dir)
3334 {
3335 dir->exclude_per_dir = ".gitignore";
3336
3337 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3338 if (!excludes_file)
3339 excludes_file = xdg_config_home("ignore");
3340 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
3341 add_patterns_from_file_1(dir, excludes_file,
3342 dir->untracked ? &dir->ss_excludes_file : NULL);
3343
3344 /* per repository user preference */
3345 if (startup_info->have_repository) {
3346 const char *path = git_path_info_exclude();
3347 if (!access_or_warn(path, R_OK, 0))
3348 add_patterns_from_file_1(dir, path,
3349 dir->untracked ? &dir->ss_info_exclude : NULL);
3350 }
3351 }
3352
3353 char *get_sparse_checkout_filename(void)
3354 {
3355 return git_pathdup("info/sparse-checkout");
3356 }
3357
3358 int get_sparse_checkout_patterns(struct pattern_list *pl)
3359 {
3360 int res;
3361 char *sparse_filename = get_sparse_checkout_filename();
3362
3363 pl->use_cone_patterns = core_sparse_checkout_cone;
3364 res = add_patterns_from_file_to_list(sparse_filename, "", 0, pl, NULL, 0);
3365
3366 free(sparse_filename);
3367 return res;
3368 }
3369
3370 int remove_path(const char *name)
3371 {
3372 char *slash;
3373
3374 if (unlink(name) && !is_missing_file_error(errno))
3375 return -1;
3376
3377 slash = strrchr(name, '/');
3378 if (slash) {
3379 char *dirs = xstrdup(name);
3380 slash = dirs + (slash - name);
3381 do {
3382 *slash = '\0';
3383 if (startup_info->original_cwd &&
3384 !strcmp(startup_info->original_cwd, dirs))
3385 break;
3386 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
3387 free(dirs);
3388 }
3389 return 0;
3390 }
3391
3392 /*
3393 * Frees memory within dir which was allocated, and resets fields for further
3394 * use. Does not free dir itself.
3395 */
3396 void dir_clear(struct dir_struct *dir)
3397 {
3398 int i, j;
3399 struct exclude_list_group *group;
3400 struct pattern_list *pl;
3401 struct exclude_stack *stk;
3402 struct dir_struct new = DIR_INIT;
3403
3404 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
3405 group = &dir->exclude_list_group[i];
3406 for (j = 0; j < group->nr; j++) {
3407 pl = &group->pl[j];
3408 if (i == EXC_DIRS)
3409 free((char *)pl->src);
3410 clear_pattern_list(pl);
3411 }
3412 free(group->pl);
3413 }
3414
3415 for (i = 0; i < dir->ignored_nr; i++)
3416 free(dir->ignored[i]);
3417 for (i = 0; i < dir->nr; i++)
3418 free(dir->entries[i]);
3419 free(dir->ignored);
3420 free(dir->entries);
3421
3422 stk = dir->exclude_stack;
3423 while (stk) {
3424 struct exclude_stack *prev = stk->prev;
3425 free(stk);
3426 stk = prev;
3427 }
3428 strbuf_release(&dir->basebuf);
3429
3430 memcpy(dir, &new, sizeof(*dir));
3431 }
3432
3433 struct ondisk_untracked_cache {
3434 struct stat_data info_exclude_stat;
3435 struct stat_data excludes_file_stat;
3436 uint32_t dir_flags;
3437 };
3438
3439 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3440
3441 struct write_data {
3442 int index; /* number of written untracked_cache_dir */
3443 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
3444 struct ewah_bitmap *valid; /* from untracked_cache_dir */
3445 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
3446 struct strbuf out;
3447 struct strbuf sb_stat;
3448 struct strbuf sb_sha1;
3449 };
3450
3451 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
3452 {
3453 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
3454 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
3455 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
3456 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
3457 to->sd_dev = htonl(from->sd_dev);
3458 to->sd_ino = htonl(from->sd_ino);
3459 to->sd_uid = htonl(from->sd_uid);
3460 to->sd_gid = htonl(from->sd_gid);
3461 to->sd_size = htonl(from->sd_size);
3462 }
3463
3464 static void write_one_dir(struct untracked_cache_dir *untracked,
3465 struct write_data *wd)
3466 {
3467 struct stat_data stat_data;
3468 struct strbuf *out = &wd->out;
3469 unsigned char intbuf[16];
3470 unsigned int intlen, value;
3471 int i = wd->index++;
3472
3473 /*
3474 * untracked_nr should be reset whenever valid is clear, but
3475 * for safety..
3476 */
3477 if (!untracked->valid) {
3478 untracked->untracked_nr = 0;
3479 untracked->check_only = 0;
3480 }
3481
3482 if (untracked->check_only)
3483 ewah_set(wd->check_only, i);
3484 if (untracked->valid) {
3485 ewah_set(wd->valid, i);
3486 stat_data_to_disk(&stat_data, &untracked->stat_data);
3487 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
3488 }
3489 if (!is_null_oid(&untracked->exclude_oid)) {
3490 ewah_set(wd->sha1_valid, i);
3491 strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
3492 the_hash_algo->rawsz);
3493 }
3494
3495 intlen = encode_varint(untracked->untracked_nr, intbuf);
3496 strbuf_add(out, intbuf, intlen);
3497
3498 /* skip non-recurse directories */
3499 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
3500 if (untracked->dirs[i]->recurse)
3501 value++;
3502 intlen = encode_varint(value, intbuf);
3503 strbuf_add(out, intbuf, intlen);
3504
3505 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
3506
3507 for (i = 0; i < untracked->untracked_nr; i++)
3508 strbuf_add(out, untracked->untracked[i],
3509 strlen(untracked->untracked[i]) + 1);
3510
3511 for (i = 0; i < untracked->dirs_nr; i++)
3512 if (untracked->dirs[i]->recurse)
3513 write_one_dir(untracked->dirs[i], wd);
3514 }
3515
3516 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
3517 {
3518 struct ondisk_untracked_cache *ouc;
3519 struct write_data wd;
3520 unsigned char varbuf[16];
3521 int varint_len;
3522 const unsigned hashsz = the_hash_algo->rawsz;
3523
3524 CALLOC_ARRAY(ouc, 1);
3525 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
3526 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
3527 ouc->dir_flags = htonl(untracked->dir_flags);
3528
3529 varint_len = encode_varint(untracked->ident.len, varbuf);
3530 strbuf_add(out, varbuf, varint_len);
3531 strbuf_addbuf(out, &untracked->ident);
3532
3533 strbuf_add(out, ouc, sizeof(*ouc));
3534 strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
3535 strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
3536 strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
3537 FREE_AND_NULL(ouc);
3538
3539 if (!untracked->root) {
3540 varint_len = encode_varint(0, varbuf);
3541 strbuf_add(out, varbuf, varint_len);
3542 return;
3543 }
3544
3545 wd.index = 0;
3546 wd.check_only = ewah_new();
3547 wd.valid = ewah_new();
3548 wd.sha1_valid = ewah_new();
3549 strbuf_init(&wd.out, 1024);
3550 strbuf_init(&wd.sb_stat, 1024);
3551 strbuf_init(&wd.sb_sha1, 1024);
3552 write_one_dir(untracked->root, &wd);
3553
3554 varint_len = encode_varint(wd.index, varbuf);
3555 strbuf_add(out, varbuf, varint_len);
3556 strbuf_addbuf(out, &wd.out);
3557 ewah_serialize_strbuf(wd.valid, out);
3558 ewah_serialize_strbuf(wd.check_only, out);
3559 ewah_serialize_strbuf(wd.sha1_valid, out);
3560 strbuf_addbuf(out, &wd.sb_stat);
3561 strbuf_addbuf(out, &wd.sb_sha1);
3562 strbuf_addch(out, '\0'); /* safe guard for string lists */
3563
3564 ewah_free(wd.valid);
3565 ewah_free(wd.check_only);
3566 ewah_free(wd.sha1_valid);
3567 strbuf_release(&wd.out);
3568 strbuf_release(&wd.sb_stat);
3569 strbuf_release(&wd.sb_sha1);
3570 }
3571
3572 static void free_untracked(struct untracked_cache_dir *ucd)
3573 {
3574 int i;
3575 if (!ucd)
3576 return;
3577 for (i = 0; i < ucd->dirs_nr; i++)
3578 free_untracked(ucd->dirs[i]);
3579 for (i = 0; i < ucd->untracked_nr; i++)
3580 free(ucd->untracked[i]);
3581 free(ucd->untracked);
3582 free(ucd->dirs);
3583 free(ucd);
3584 }
3585
3586 void free_untracked_cache(struct untracked_cache *uc)
3587 {
3588 if (uc)
3589 free_untracked(uc->root);
3590 free(uc);
3591 }
3592
3593 struct read_data {
3594 int index;
3595 struct untracked_cache_dir **ucd;
3596 struct ewah_bitmap *check_only;
3597 struct ewah_bitmap *valid;
3598 struct ewah_bitmap *sha1_valid;
3599 const unsigned char *data;
3600 const unsigned char *end;
3601 };
3602
3603 static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
3604 {
3605 memcpy(to, data, sizeof(*to));
3606 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
3607 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
3608 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
3609 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
3610 to->sd_dev = ntohl(to->sd_dev);
3611 to->sd_ino = ntohl(to->sd_ino);
3612 to->sd_uid = ntohl(to->sd_uid);
3613 to->sd_gid = ntohl(to->sd_gid);
3614 to->sd_size = ntohl(to->sd_size);
3615 }
3616
3617 static int read_one_dir(struct untracked_cache_dir **untracked_,
3618 struct read_data *rd)
3619 {
3620 struct untracked_cache_dir ud, *untracked;
3621 const unsigned char *data = rd->data, *end = rd->end;
3622 const unsigned char *eos;
3623 unsigned int value;
3624 int i;
3625
3626 memset(&ud, 0, sizeof(ud));
3627
3628 value = decode_varint(&data);
3629 if (data > end)
3630 return -1;
3631 ud.recurse = 1;
3632 ud.untracked_alloc = value;
3633 ud.untracked_nr = value;
3634 if (ud.untracked_nr)
3635 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
3636
3637 ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
3638 if (data > end)
3639 return -1;
3640 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
3641
3642 eos = memchr(data, '\0', end - data);
3643 if (!eos || eos == end)
3644 return -1;
3645
3646 *untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
3647 memcpy(untracked, &ud, sizeof(ud));
3648 memcpy(untracked->name, data, eos - data + 1);
3649 data = eos + 1;
3650
3651 for (i = 0; i < untracked->untracked_nr; i++) {
3652 eos = memchr(data, '\0', end - data);
3653 if (!eos || eos == end)
3654 return -1;
3655 untracked->untracked[i] = xmemdupz(data, eos - data);
3656 data = eos + 1;
3657 }
3658
3659 rd->ucd[rd->index++] = untracked;
3660 rd->data = data;
3661
3662 for (i = 0; i < untracked->dirs_nr; i++) {
3663 if (read_one_dir(untracked->dirs + i, rd) < 0)
3664 return -1;
3665 }
3666 return 0;
3667 }
3668
3669 static void set_check_only(size_t pos, void *cb)
3670 {
3671 struct read_data *rd = cb;
3672 struct untracked_cache_dir *ud = rd->ucd[pos];
3673 ud->check_only = 1;
3674 }
3675
3676 static void read_stat(size_t pos, void *cb)
3677 {
3678 struct read_data *rd = cb;
3679 struct untracked_cache_dir *ud = rd->ucd[pos];
3680 if (rd->data + sizeof(struct stat_data) > rd->end) {
3681 rd->data = rd->end + 1;
3682 return;
3683 }
3684 stat_data_from_disk(&ud->stat_data, rd->data);
3685 rd->data += sizeof(struct stat_data);
3686 ud->valid = 1;
3687 }
3688
3689 static void read_oid(size_t pos, void *cb)
3690 {
3691 struct read_data *rd = cb;
3692 struct untracked_cache_dir *ud = rd->ucd[pos];
3693 if (rd->data + the_hash_algo->rawsz > rd->end) {
3694 rd->data = rd->end + 1;
3695 return;
3696 }
3697 oidread(&ud->exclude_oid, rd->data);
3698 rd->data += the_hash_algo->rawsz;
3699 }
3700
3701 static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
3702 const unsigned char *sha1)
3703 {
3704 stat_data_from_disk(&oid_stat->stat, data);
3705 oidread(&oid_stat->oid, sha1);
3706 oid_stat->valid = 1;
3707 }
3708
3709 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
3710 {
3711 struct untracked_cache *uc;
3712 struct read_data rd;
3713 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
3714 const char *ident;
3715 int ident_len;
3716 ssize_t len;
3717 const char *exclude_per_dir;
3718 const unsigned hashsz = the_hash_algo->rawsz;
3719 const unsigned offset = sizeof(struct ondisk_untracked_cache);
3720 const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
3721
3722 if (sz <= 1 || end[-1] != '\0')
3723 return NULL;
3724 end--;
3725
3726 ident_len = decode_varint(&next);
3727 if (next + ident_len > end)
3728 return NULL;
3729 ident = (const char *)next;
3730 next += ident_len;
3731
3732 if (next + exclude_per_dir_offset + 1 > end)
3733 return NULL;
3734
3735 CALLOC_ARRAY(uc, 1);
3736 strbuf_init(&uc->ident, ident_len);
3737 strbuf_add(&uc->ident, ident, ident_len);
3738 load_oid_stat(&uc->ss_info_exclude,
3739 next + ouc_offset(info_exclude_stat),
3740 next + offset);
3741 load_oid_stat(&uc->ss_excludes_file,
3742 next + ouc_offset(excludes_file_stat),
3743 next + offset + hashsz);
3744 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
3745 exclude_per_dir = (const char *)next + exclude_per_dir_offset;
3746 uc->exclude_per_dir = xstrdup(exclude_per_dir);
3747 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3748 next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
3749 if (next >= end)
3750 goto done2;
3751
3752 len = decode_varint(&next);
3753 if (next > end || len == 0)
3754 goto done2;
3755
3756 rd.valid = ewah_new();
3757 rd.check_only = ewah_new();
3758 rd.sha1_valid = ewah_new();
3759 rd.data = next;
3760 rd.end = end;
3761 rd.index = 0;
3762 ALLOC_ARRAY(rd.ucd, len);
3763
3764 if (read_one_dir(&uc->root, &rd) || rd.index != len)
3765 goto done;
3766
3767 next = rd.data;
3768 len = ewah_read_mmap(rd.valid, next, end - next);
3769 if (len < 0)
3770 goto done;
3771
3772 next += len;
3773 len = ewah_read_mmap(rd.check_only, next, end - next);
3774 if (len < 0)
3775 goto done;
3776
3777 next += len;
3778 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
3779 if (len < 0)
3780 goto done;
3781
3782 ewah_each_bit(rd.check_only, set_check_only, &rd);
3783 rd.data = next + len;
3784 ewah_each_bit(rd.valid, read_stat, &rd);
3785 ewah_each_bit(rd.sha1_valid, read_oid, &rd);
3786 next = rd.data;
3787
3788 done:
3789 free(rd.ucd);
3790 ewah_free(rd.valid);
3791 ewah_free(rd.check_only);
3792 ewah_free(rd.sha1_valid);
3793 done2:
3794 if (next != end) {
3795 free_untracked_cache(uc);
3796 uc = NULL;
3797 }
3798 return uc;
3799 }
3800
3801 static void invalidate_one_directory(struct untracked_cache *uc,
3802 struct untracked_cache_dir *ucd)
3803 {
3804 uc->dir_invalidated++;
3805 ucd->valid = 0;
3806 ucd->untracked_nr = 0;
3807 }
3808
3809 /*
3810 * Normally when an entry is added or removed from a directory,
3811 * invalidating that directory is enough. No need to touch its
3812 * ancestors. When a directory is shown as "foo/bar/" in git-status
3813 * however, deleting or adding an entry may have cascading effect.
3814 *
3815 * Say the "foo/bar/file" has become untracked, we need to tell the
3816 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3817 * directory any more (because "bar" is managed by foo as an untracked
3818 * "file").
3819 *
3820 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3821 * was the last untracked entry in the entire "foo", we should show
3822 * "foo/" instead. Which means we have to invalidate past "bar" up to
3823 * "foo".
3824 *
3825 * This function traverses all directories from root to leaf. If there
3826 * is a chance of one of the above cases happening, we invalidate back
3827 * to root. Otherwise we just invalidate the leaf. There may be a more
3828 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3829 * detect these cases and avoid unnecessary invalidation, for example,
3830 * checking for the untracked entry named "bar/" in "foo", but for now
3831 * stick to something safe and simple.
3832 */
3833 static int invalidate_one_component(struct untracked_cache *uc,
3834 struct untracked_cache_dir *dir,
3835 const char *path, int len)
3836 {
3837 const char *rest = strchr(path, '/');
3838
3839 if (rest) {
3840 int component_len = rest - path;
3841 struct untracked_cache_dir *d =
3842 lookup_untracked(uc, dir, path, component_len);
3843 int ret =
3844 invalidate_one_component(uc, d, rest + 1,
3845 len - (component_len + 1));
3846 if (ret)
3847 invalidate_one_directory(uc, dir);
3848 return ret;
3849 }
3850
3851 invalidate_one_directory(uc, dir);
3852 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
3853 }
3854
3855 void untracked_cache_invalidate_path(struct index_state *istate,
3856 const char *path, int safe_path)
3857 {
3858 if (!istate->untracked || !istate->untracked->root)
3859 return;
3860 if (!safe_path && !verify_path(path, 0))
3861 return;
3862 invalidate_one_component(istate->untracked, istate->untracked->root,
3863 path, strlen(path));
3864 }
3865
3866 void untracked_cache_remove_from_index(struct index_state *istate,
3867 const char *path)
3868 {
3869 untracked_cache_invalidate_path(istate, path, 1);
3870 }
3871
3872 void untracked_cache_add_to_index(struct index_state *istate,
3873 const char *path)
3874 {
3875 untracked_cache_invalidate_path(istate, path, 1);
3876 }
3877
3878 static void connect_wt_gitdir_in_nested(const char *sub_worktree,
3879 const char *sub_gitdir)
3880 {
3881 int i;
3882 struct repository subrepo;
3883 struct strbuf sub_wt = STRBUF_INIT;
3884 struct strbuf sub_gd = STRBUF_INIT;
3885
3886 const struct submodule *sub;
3887
3888 /* If the submodule has no working tree, we can ignore it. */
3889 if (repo_init(&subrepo, sub_gitdir, sub_worktree))
3890 return;
3891
3892 if (repo_read_index(&subrepo) < 0)
3893 die(_("index file corrupt in repo %s"), subrepo.gitdir);
3894
3895 /* TODO: audit for interaction with sparse-index. */
3896 ensure_full_index(subrepo.index);
3897 for (i = 0; i < subrepo.index->cache_nr; i++) {
3898 const struct cache_entry *ce = subrepo.index->cache[i];
3899
3900 if (!S_ISGITLINK(ce->ce_mode))
3901 continue;
3902
3903 while (i + 1 < subrepo.index->cache_nr &&
3904 !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
3905 /*
3906 * Skip entries with the same name in different stages
3907 * to make sure an entry is returned only once.
3908 */
3909 i++;
3910
3911 sub = submodule_from_path(&subrepo, null_oid(), ce->name);
3912 if (!sub || !is_submodule_active(&subrepo, ce->name))
3913 /* .gitmodules broken or inactive sub */
3914 continue;
3915
3916 strbuf_reset(&sub_wt);
3917 strbuf_reset(&sub_gd);
3918 strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
3919 submodule_name_to_gitdir(&sub_gd, &subrepo, sub->name);
3920
3921 connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
3922 }
3923 strbuf_release(&sub_wt);
3924 strbuf_release(&sub_gd);
3925 repo_clear(&subrepo);
3926 }
3927
3928 void connect_work_tree_and_git_dir(const char *work_tree_,
3929 const char *git_dir_,
3930 int recurse_into_nested)
3931 {
3932 struct strbuf gitfile_sb = STRBUF_INIT;
3933 struct strbuf cfg_sb = STRBUF_INIT;
3934 struct strbuf rel_path = STRBUF_INIT;
3935 char *git_dir, *work_tree;
3936
3937 /* Prepare .git file */
3938 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3939 if (safe_create_leading_directories_const(gitfile_sb.buf))
3940 die(_("could not create directories for %s"), gitfile_sb.buf);
3941
3942 /* Prepare config file */
3943 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3944 if (safe_create_leading_directories_const(cfg_sb.buf))
3945 die(_("could not create directories for %s"), cfg_sb.buf);
3946
3947 git_dir = real_pathdup(git_dir_, 1);
3948 work_tree = real_pathdup(work_tree_, 1);
3949
3950 /* Write .git file */
3951 write_file(gitfile_sb.buf, "gitdir: %s",
3952 relative_path(git_dir, work_tree, &rel_path));
3953 /* Update core.worktree setting */
3954 git_config_set_in_file(cfg_sb.buf, "core.worktree",
3955 relative_path(work_tree, git_dir, &rel_path));
3956
3957 strbuf_release(&gitfile_sb);
3958 strbuf_release(&cfg_sb);
3959 strbuf_release(&rel_path);
3960
3961 if (recurse_into_nested)
3962 connect_wt_gitdir_in_nested(work_tree, git_dir);
3963
3964 free(work_tree);
3965 free(git_dir);
3966 }
3967
3968 /*
3969 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3970 */
3971 void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3972 {
3973 if (rename(old_git_dir, new_git_dir) < 0)
3974 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3975 old_git_dir, new_git_dir);
3976
3977 connect_work_tree_and_git_dir(path, new_git_dir, 0);
3978 }
3979
3980 int path_match_flags(const char *const str, const enum path_match_flags flags)
3981 {
3982 const char *p = str;
3983
3984 if (flags & PATH_MATCH_NATIVE &&
3985 flags & PATH_MATCH_XPLATFORM)
3986 BUG("path_match_flags() must get one match kind, not multiple!");
3987 else if (!(flags & PATH_MATCH_KINDS_MASK))
3988 BUG("path_match_flags() must get at least one match kind!");
3989
3990 if (flags & PATH_MATCH_STARTS_WITH_DOT_SLASH &&
3991 flags & PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH)
3992 BUG("path_match_flags() must get one platform kind, not multiple!");
3993 else if (!(flags & PATH_MATCH_PLATFORM_MASK))
3994 BUG("path_match_flags() must get at least one platform kind!");
3995
3996 if (*p++ != '.')
3997 return 0;
3998 if (flags & PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH &&
3999 *p++ != '.')
4000 return 0;
4001
4002 if (flags & PATH_MATCH_NATIVE)
4003 return is_dir_sep(*p);
4004 else if (flags & PATH_MATCH_XPLATFORM)
4005 return is_xplatform_dir_sep(*p);
4006 BUG("unreachable");
4007 }