Merge branch 'dt/name-hash-dir-entry-fix'
[git/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * See Documentation/technical/api-directory-listing.txt
6 *
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10 #include "cache.h"
11 #include "dir.h"
12 #include "refs.h"
13 #include "wildmatch.h"
14 #include "pathspec.h"
15 #include "utf8.h"
16 #include "varint.h"
17 #include "ewah/ewok.h"
18
19 struct path_simplify {
20 int len;
21 const char *path;
22 };
23
24 /*
25 * Tells read_directory_recursive how a file or directory should be treated.
26 * Values are ordered by significance, e.g. if a directory contains both
27 * excluded and untracked files, it is listed as untracked because
28 * path_untracked > path_excluded.
29 */
30 enum path_treatment {
31 path_none = 0,
32 path_recurse,
33 path_excluded,
34 path_untracked
35 };
36
37 /*
38 * Support data structure for our opendir/readdir/closedir wrappers
39 */
40 struct cached_dir {
41 DIR *fdir;
42 struct untracked_cache_dir *untracked;
43 int nr_files;
44 int nr_dirs;
45
46 struct dirent *de;
47 const char *file;
48 struct untracked_cache_dir *ucd;
49 };
50
51 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
52 const char *path, int len, struct untracked_cache_dir *untracked,
53 int check_only, const struct path_simplify *simplify);
54 static int get_dtype(struct dirent *de, const char *path, int len);
55
56 /* helper string functions with support for the ignore_case flag */
57 int strcmp_icase(const char *a, const char *b)
58 {
59 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
60 }
61
62 int strncmp_icase(const char *a, const char *b, size_t count)
63 {
64 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
65 }
66
67 int fnmatch_icase(const char *pattern, const char *string, int flags)
68 {
69 return wildmatch(pattern, string,
70 flags | (ignore_case ? WM_CASEFOLD : 0),
71 NULL);
72 }
73
74 int git_fnmatch(const struct pathspec_item *item,
75 const char *pattern, const char *string,
76 int prefix)
77 {
78 if (prefix > 0) {
79 if (ps_strncmp(item, pattern, string, prefix))
80 return WM_NOMATCH;
81 pattern += prefix;
82 string += prefix;
83 }
84 if (item->flags & PATHSPEC_ONESTAR) {
85 int pattern_len = strlen(++pattern);
86 int string_len = strlen(string);
87 return string_len < pattern_len ||
88 ps_strcmp(item, pattern,
89 string + string_len - pattern_len);
90 }
91 if (item->magic & PATHSPEC_GLOB)
92 return wildmatch(pattern, string,
93 WM_PATHNAME |
94 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
95 NULL);
96 else
97 /* wildmatch has not learned no FNM_PATHNAME mode yet */
98 return wildmatch(pattern, string,
99 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
100 NULL);
101 }
102
103 static int fnmatch_icase_mem(const char *pattern, int patternlen,
104 const char *string, int stringlen,
105 int flags)
106 {
107 int match_status;
108 struct strbuf pat_buf = STRBUF_INIT;
109 struct strbuf str_buf = STRBUF_INIT;
110 const char *use_pat = pattern;
111 const char *use_str = string;
112
113 if (pattern[patternlen]) {
114 strbuf_add(&pat_buf, pattern, patternlen);
115 use_pat = pat_buf.buf;
116 }
117 if (string[stringlen]) {
118 strbuf_add(&str_buf, string, stringlen);
119 use_str = str_buf.buf;
120 }
121
122 if (ignore_case)
123 flags |= WM_CASEFOLD;
124 match_status = wildmatch(use_pat, use_str, flags, NULL);
125
126 strbuf_release(&pat_buf);
127 strbuf_release(&str_buf);
128
129 return match_status;
130 }
131
132 static size_t common_prefix_len(const struct pathspec *pathspec)
133 {
134 int n;
135 size_t max = 0;
136
137 /*
138 * ":(icase)path" is treated as a pathspec full of
139 * wildcard. In other words, only prefix is considered common
140 * prefix. If the pathspec is abc/foo abc/bar, running in
141 * subdir xyz, the common prefix is still xyz, not xuz/abc as
142 * in non-:(icase).
143 */
144 GUARD_PATHSPEC(pathspec,
145 PATHSPEC_FROMTOP |
146 PATHSPEC_MAXDEPTH |
147 PATHSPEC_LITERAL |
148 PATHSPEC_GLOB |
149 PATHSPEC_ICASE |
150 PATHSPEC_EXCLUDE);
151
152 for (n = 0; n < pathspec->nr; n++) {
153 size_t i = 0, len = 0, item_len;
154 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
155 continue;
156 if (pathspec->items[n].magic & PATHSPEC_ICASE)
157 item_len = pathspec->items[n].prefix;
158 else
159 item_len = pathspec->items[n].nowildcard_len;
160 while (i < item_len && (n == 0 || i < max)) {
161 char c = pathspec->items[n].match[i];
162 if (c != pathspec->items[0].match[i])
163 break;
164 if (c == '/')
165 len = i + 1;
166 i++;
167 }
168 if (n == 0 || len < max) {
169 max = len;
170 if (!max)
171 break;
172 }
173 }
174 return max;
175 }
176
177 /*
178 * Returns a copy of the longest leading path common among all
179 * pathspecs.
180 */
181 char *common_prefix(const struct pathspec *pathspec)
182 {
183 unsigned long len = common_prefix_len(pathspec);
184
185 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
186 }
187
188 int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
189 {
190 size_t len;
191
192 /*
193 * Calculate common prefix for the pathspec, and
194 * use that to optimize the directory walk
195 */
196 len = common_prefix_len(pathspec);
197
198 /* Read the directory and prune it */
199 read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec);
200 return len;
201 }
202
203 int within_depth(const char *name, int namelen,
204 int depth, int max_depth)
205 {
206 const char *cp = name, *cpe = name + namelen;
207
208 while (cp < cpe) {
209 if (*cp++ != '/')
210 continue;
211 depth++;
212 if (depth > max_depth)
213 return 0;
214 }
215 return 1;
216 }
217
218 #define DO_MATCH_EXCLUDE 1
219 #define DO_MATCH_DIRECTORY 2
220
221 /*
222 * Does 'match' match the given name?
223 * A match is found if
224 *
225 * (1) the 'match' string is leading directory of 'name', or
226 * (2) the 'match' string is a wildcard and matches 'name', or
227 * (3) the 'match' string is exactly the same as 'name'.
228 *
229 * and the return value tells which case it was.
230 *
231 * It returns 0 when there is no match.
232 */
233 static int match_pathspec_item(const struct pathspec_item *item, int prefix,
234 const char *name, int namelen, unsigned flags)
235 {
236 /* name/namelen has prefix cut off by caller */
237 const char *match = item->match + prefix;
238 int matchlen = item->len - prefix;
239
240 /*
241 * The normal call pattern is:
242 * 1. prefix = common_prefix_len(ps);
243 * 2. prune something, or fill_directory
244 * 3. match_pathspec()
245 *
246 * 'prefix' at #1 may be shorter than the command's prefix and
247 * it's ok for #2 to match extra files. Those extras will be
248 * trimmed at #3.
249 *
250 * Suppose the pathspec is 'foo' and '../bar' running from
251 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
252 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
253 * user does not want XYZ/foo, only the "foo" part should be
254 * case-insensitive. We need to filter out XYZ/foo here. In
255 * other words, we do not trust the caller on comparing the
256 * prefix part when :(icase) is involved. We do exact
257 * comparison ourselves.
258 *
259 * Normally the caller (common_prefix_len() in fact) does
260 * _exact_ matching on name[-prefix+1..-1] and we do not need
261 * to check that part. Be defensive and check it anyway, in
262 * case common_prefix_len is changed, or a new caller is
263 * introduced that does not use common_prefix_len.
264 *
265 * If the penalty turns out too high when prefix is really
266 * long, maybe change it to
267 * strncmp(match, name, item->prefix - prefix)
268 */
269 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
270 strncmp(item->match, name - prefix, item->prefix))
271 return 0;
272
273 /* If the match was just the prefix, we matched */
274 if (!*match)
275 return MATCHED_RECURSIVELY;
276
277 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
278 if (matchlen == namelen)
279 return MATCHED_EXACTLY;
280
281 if (match[matchlen-1] == '/' || name[matchlen] == '/')
282 return MATCHED_RECURSIVELY;
283 } else if ((flags & DO_MATCH_DIRECTORY) &&
284 match[matchlen - 1] == '/' &&
285 namelen == matchlen - 1 &&
286 !ps_strncmp(item, match, name, namelen))
287 return MATCHED_EXACTLY;
288
289 if (item->nowildcard_len < item->len &&
290 !git_fnmatch(item, match, name,
291 item->nowildcard_len - prefix))
292 return MATCHED_FNMATCH;
293
294 return 0;
295 }
296
297 /*
298 * Given a name and a list of pathspecs, returns the nature of the
299 * closest (i.e. most specific) match of the name to any of the
300 * pathspecs.
301 *
302 * The caller typically calls this multiple times with the same
303 * pathspec and seen[] array but with different name/namelen
304 * (e.g. entries from the index) and is interested in seeing if and
305 * how each pathspec matches all the names it calls this function
306 * with. A mark is left in the seen[] array for each pathspec element
307 * indicating the closest type of match that element achieved, so if
308 * seen[n] remains zero after multiple invocations, that means the nth
309 * pathspec did not match any names, which could indicate that the
310 * user mistyped the nth pathspec.
311 */
312 static int do_match_pathspec(const struct pathspec *ps,
313 const char *name, int namelen,
314 int prefix, char *seen,
315 unsigned flags)
316 {
317 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
318
319 GUARD_PATHSPEC(ps,
320 PATHSPEC_FROMTOP |
321 PATHSPEC_MAXDEPTH |
322 PATHSPEC_LITERAL |
323 PATHSPEC_GLOB |
324 PATHSPEC_ICASE |
325 PATHSPEC_EXCLUDE);
326
327 if (!ps->nr) {
328 if (!ps->recursive ||
329 !(ps->magic & PATHSPEC_MAXDEPTH) ||
330 ps->max_depth == -1)
331 return MATCHED_RECURSIVELY;
332
333 if (within_depth(name, namelen, 0, ps->max_depth))
334 return MATCHED_EXACTLY;
335 else
336 return 0;
337 }
338
339 name += prefix;
340 namelen -= prefix;
341
342 for (i = ps->nr - 1; i >= 0; i--) {
343 int how;
344
345 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
346 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
347 continue;
348
349 if (seen && seen[i] == MATCHED_EXACTLY)
350 continue;
351 /*
352 * Make exclude patterns optional and never report
353 * "pathspec ':(exclude)foo' matches no files"
354 */
355 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
356 seen[i] = MATCHED_FNMATCH;
357 how = match_pathspec_item(ps->items+i, prefix, name,
358 namelen, flags);
359 if (ps->recursive &&
360 (ps->magic & PATHSPEC_MAXDEPTH) &&
361 ps->max_depth != -1 &&
362 how && how != MATCHED_FNMATCH) {
363 int len = ps->items[i].len;
364 if (name[len] == '/')
365 len++;
366 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
367 how = MATCHED_EXACTLY;
368 else
369 how = 0;
370 }
371 if (how) {
372 if (retval < how)
373 retval = how;
374 if (seen && seen[i] < how)
375 seen[i] = how;
376 }
377 }
378 return retval;
379 }
380
381 int match_pathspec(const struct pathspec *ps,
382 const char *name, int namelen,
383 int prefix, char *seen, int is_dir)
384 {
385 int positive, negative;
386 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
387 positive = do_match_pathspec(ps, name, namelen,
388 prefix, seen, flags);
389 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
390 return positive;
391 negative = do_match_pathspec(ps, name, namelen,
392 prefix, seen,
393 flags | DO_MATCH_EXCLUDE);
394 return negative ? 0 : positive;
395 }
396
397 int report_path_error(const char *ps_matched,
398 const struct pathspec *pathspec,
399 const char *prefix)
400 {
401 /*
402 * Make sure all pathspec matched; otherwise it is an error.
403 */
404 int num, errors = 0;
405 for (num = 0; num < pathspec->nr; num++) {
406 int other, found_dup;
407
408 if (ps_matched[num])
409 continue;
410 /*
411 * The caller might have fed identical pathspec
412 * twice. Do not barf on such a mistake.
413 * FIXME: parse_pathspec should have eliminated
414 * duplicate pathspec.
415 */
416 for (found_dup = other = 0;
417 !found_dup && other < pathspec->nr;
418 other++) {
419 if (other == num || !ps_matched[other])
420 continue;
421 if (!strcmp(pathspec->items[other].original,
422 pathspec->items[num].original))
423 /*
424 * Ok, we have a match already.
425 */
426 found_dup = 1;
427 }
428 if (found_dup)
429 continue;
430
431 error("pathspec '%s' did not match any file(s) known to git.",
432 pathspec->items[num].original);
433 errors++;
434 }
435 return errors;
436 }
437
438 /*
439 * Return the length of the "simple" part of a path match limiter.
440 */
441 int simple_length(const char *match)
442 {
443 int len = -1;
444
445 for (;;) {
446 unsigned char c = *match++;
447 len++;
448 if (c == '\0' || is_glob_special(c))
449 return len;
450 }
451 }
452
453 int no_wildcard(const char *string)
454 {
455 return string[simple_length(string)] == '\0';
456 }
457
458 void parse_exclude_pattern(const char **pattern,
459 int *patternlen,
460 int *flags,
461 int *nowildcardlen)
462 {
463 const char *p = *pattern;
464 size_t i, len;
465
466 *flags = 0;
467 if (*p == '!') {
468 *flags |= EXC_FLAG_NEGATIVE;
469 p++;
470 }
471 len = strlen(p);
472 if (len && p[len - 1] == '/') {
473 len--;
474 *flags |= EXC_FLAG_MUSTBEDIR;
475 }
476 for (i = 0; i < len; i++) {
477 if (p[i] == '/')
478 break;
479 }
480 if (i == len)
481 *flags |= EXC_FLAG_NODIR;
482 *nowildcardlen = simple_length(p);
483 /*
484 * we should have excluded the trailing slash from 'p' too,
485 * but that's one more allocation. Instead just make sure
486 * nowildcardlen does not exceed real patternlen
487 */
488 if (*nowildcardlen > len)
489 *nowildcardlen = len;
490 if (*p == '*' && no_wildcard(p + 1))
491 *flags |= EXC_FLAG_ENDSWITH;
492 *pattern = p;
493 *patternlen = len;
494 }
495
496 void add_exclude(const char *string, const char *base,
497 int baselen, struct exclude_list *el, int srcpos)
498 {
499 struct exclude *x;
500 int patternlen;
501 int flags;
502 int nowildcardlen;
503
504 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
505 if (flags & EXC_FLAG_MUSTBEDIR) {
506 char *s;
507 x = xmalloc(sizeof(*x) + patternlen + 1);
508 s = (char *)(x+1);
509 memcpy(s, string, patternlen);
510 s[patternlen] = '\0';
511 x->pattern = s;
512 } else {
513 x = xmalloc(sizeof(*x));
514 x->pattern = string;
515 }
516 x->patternlen = patternlen;
517 x->nowildcardlen = nowildcardlen;
518 x->base = base;
519 x->baselen = baselen;
520 x->flags = flags;
521 x->srcpos = srcpos;
522 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
523 el->excludes[el->nr++] = x;
524 x->el = el;
525 }
526
527 static void *read_skip_worktree_file_from_index(const char *path, size_t *size,
528 struct sha1_stat *sha1_stat)
529 {
530 int pos, len;
531 unsigned long sz;
532 enum object_type type;
533 void *data;
534
535 len = strlen(path);
536 pos = cache_name_pos(path, len);
537 if (pos < 0)
538 return NULL;
539 if (!ce_skip_worktree(active_cache[pos]))
540 return NULL;
541 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
542 if (!data || type != OBJ_BLOB) {
543 free(data);
544 return NULL;
545 }
546 *size = xsize_t(sz);
547 if (sha1_stat) {
548 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
549 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
550 }
551 return data;
552 }
553
554 /*
555 * Frees memory within el which was allocated for exclude patterns and
556 * the file buffer. Does not free el itself.
557 */
558 void clear_exclude_list(struct exclude_list *el)
559 {
560 int i;
561
562 for (i = 0; i < el->nr; i++)
563 free(el->excludes[i]);
564 free(el->excludes);
565 free(el->filebuf);
566
567 el->nr = 0;
568 el->excludes = NULL;
569 el->filebuf = NULL;
570 }
571
572 static void trim_trailing_spaces(char *buf)
573 {
574 char *p, *last_space = NULL;
575
576 for (p = buf; *p; p++)
577 switch (*p) {
578 case ' ':
579 if (!last_space)
580 last_space = p;
581 break;
582 case '\\':
583 p++;
584 if (!*p)
585 return;
586 /* fallthrough */
587 default:
588 last_space = NULL;
589 }
590
591 if (last_space)
592 *last_space = '\0';
593 }
594
595 /*
596 * Given a subdirectory name and "dir" of the current directory,
597 * search the subdir in "dir" and return it, or create a new one if it
598 * does not exist in "dir".
599 *
600 * If "name" has the trailing slash, it'll be excluded in the search.
601 */
602 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
603 struct untracked_cache_dir *dir,
604 const char *name, int len)
605 {
606 int first, last;
607 struct untracked_cache_dir *d;
608 if (!dir)
609 return NULL;
610 if (len && name[len - 1] == '/')
611 len--;
612 first = 0;
613 last = dir->dirs_nr;
614 while (last > first) {
615 int cmp, next = (last + first) >> 1;
616 d = dir->dirs[next];
617 cmp = strncmp(name, d->name, len);
618 if (!cmp && strlen(d->name) > len)
619 cmp = -1;
620 if (!cmp)
621 return d;
622 if (cmp < 0) {
623 last = next;
624 continue;
625 }
626 first = next+1;
627 }
628
629 uc->dir_created++;
630 d = xmalloc(sizeof(*d) + len + 1);
631 memset(d, 0, sizeof(*d));
632 memcpy(d->name, name, len);
633 d->name[len] = '\0';
634
635 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
636 memmove(dir->dirs + first + 1, dir->dirs + first,
637 (dir->dirs_nr - first) * sizeof(*dir->dirs));
638 dir->dirs_nr++;
639 dir->dirs[first] = d;
640 return d;
641 }
642
643 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
644 {
645 int i;
646 dir->valid = 0;
647 dir->untracked_nr = 0;
648 for (i = 0; i < dir->dirs_nr; i++)
649 do_invalidate_gitignore(dir->dirs[i]);
650 }
651
652 static void invalidate_gitignore(struct untracked_cache *uc,
653 struct untracked_cache_dir *dir)
654 {
655 uc->gitignore_invalidated++;
656 do_invalidate_gitignore(dir);
657 }
658
659 static void invalidate_directory(struct untracked_cache *uc,
660 struct untracked_cache_dir *dir)
661 {
662 int i;
663 uc->dir_invalidated++;
664 dir->valid = 0;
665 dir->untracked_nr = 0;
666 for (i = 0; i < dir->dirs_nr; i++)
667 dir->dirs[i]->recurse = 0;
668 }
669
670 /*
671 * Given a file with name "fname", read it (either from disk, or from
672 * the index if "check_index" is non-zero), parse it and store the
673 * exclude rules in "el".
674 *
675 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
676 * stat data from disk (only valid if add_excludes returns zero). If
677 * ss_valid is non-zero, "ss" must contain good value as input.
678 */
679 static int add_excludes(const char *fname, const char *base, int baselen,
680 struct exclude_list *el, int check_index,
681 struct sha1_stat *sha1_stat)
682 {
683 struct stat st;
684 int fd, i, lineno = 1;
685 size_t size = 0;
686 char *buf, *entry;
687
688 fd = open(fname, O_RDONLY);
689 if (fd < 0 || fstat(fd, &st) < 0) {
690 if (errno != ENOENT)
691 warn_on_inaccessible(fname);
692 if (0 <= fd)
693 close(fd);
694 if (!check_index ||
695 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL)
696 return -1;
697 if (size == 0) {
698 free(buf);
699 return 0;
700 }
701 if (buf[size-1] != '\n') {
702 buf = xrealloc(buf, size+1);
703 buf[size++] = '\n';
704 }
705 } else {
706 size = xsize_t(st.st_size);
707 if (size == 0) {
708 if (sha1_stat) {
709 fill_stat_data(&sha1_stat->stat, &st);
710 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
711 sha1_stat->valid = 1;
712 }
713 close(fd);
714 return 0;
715 }
716 buf = xmalloc(size+1);
717 if (read_in_full(fd, buf, size) != size) {
718 free(buf);
719 close(fd);
720 return -1;
721 }
722 buf[size++] = '\n';
723 close(fd);
724 if (sha1_stat) {
725 int pos;
726 if (sha1_stat->valid &&
727 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st))
728 ; /* no content change, ss->sha1 still good */
729 else if (check_index &&
730 (pos = cache_name_pos(fname, strlen(fname))) >= 0 &&
731 !ce_stage(active_cache[pos]) &&
732 ce_uptodate(active_cache[pos]) &&
733 !would_convert_to_git(fname))
734 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
735 else
736 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
737 fill_stat_data(&sha1_stat->stat, &st);
738 sha1_stat->valid = 1;
739 }
740 }
741
742 el->filebuf = buf;
743
744 if (skip_utf8_bom(&buf, size))
745 size -= buf - el->filebuf;
746
747 entry = buf;
748
749 for (i = 0; i < size; i++) {
750 if (buf[i] == '\n') {
751 if (entry != buf + i && entry[0] != '#') {
752 buf[i - (i && buf[i-1] == '\r')] = 0;
753 trim_trailing_spaces(entry);
754 add_exclude(entry, base, baselen, el, lineno);
755 }
756 lineno++;
757 entry = buf + i + 1;
758 }
759 }
760 return 0;
761 }
762
763 int add_excludes_from_file_to_list(const char *fname, const char *base,
764 int baselen, struct exclude_list *el,
765 int check_index)
766 {
767 return add_excludes(fname, base, baselen, el, check_index, NULL);
768 }
769
770 struct exclude_list *add_exclude_list(struct dir_struct *dir,
771 int group_type, const char *src)
772 {
773 struct exclude_list *el;
774 struct exclude_list_group *group;
775
776 group = &dir->exclude_list_group[group_type];
777 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
778 el = &group->el[group->nr++];
779 memset(el, 0, sizeof(*el));
780 el->src = src;
781 return el;
782 }
783
784 /*
785 * Used to set up core.excludesfile and .git/info/exclude lists.
786 */
787 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
788 struct sha1_stat *sha1_stat)
789 {
790 struct exclude_list *el;
791 /*
792 * catch setup_standard_excludes() that's called before
793 * dir->untracked is assigned. That function behaves
794 * differently when dir->untracked is non-NULL.
795 */
796 if (!dir->untracked)
797 dir->unmanaged_exclude_files++;
798 el = add_exclude_list(dir, EXC_FILE, fname);
799 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0)
800 die("cannot use %s as an exclude file", fname);
801 }
802
803 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
804 {
805 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
806 add_excludes_from_file_1(dir, fname, NULL);
807 }
808
809 int match_basename(const char *basename, int basenamelen,
810 const char *pattern, int prefix, int patternlen,
811 int flags)
812 {
813 if (prefix == patternlen) {
814 if (patternlen == basenamelen &&
815 !strncmp_icase(pattern, basename, basenamelen))
816 return 1;
817 } else if (flags & EXC_FLAG_ENDSWITH) {
818 /* "*literal" matching against "fooliteral" */
819 if (patternlen - 1 <= basenamelen &&
820 !strncmp_icase(pattern + 1,
821 basename + basenamelen - (patternlen - 1),
822 patternlen - 1))
823 return 1;
824 } else {
825 if (fnmatch_icase_mem(pattern, patternlen,
826 basename, basenamelen,
827 0) == 0)
828 return 1;
829 }
830 return 0;
831 }
832
833 int match_pathname(const char *pathname, int pathlen,
834 const char *base, int baselen,
835 const char *pattern, int prefix, int patternlen,
836 int flags)
837 {
838 const char *name;
839 int namelen;
840
841 /*
842 * match with FNM_PATHNAME; the pattern has base implicitly
843 * in front of it.
844 */
845 if (*pattern == '/') {
846 pattern++;
847 patternlen--;
848 prefix--;
849 }
850
851 /*
852 * baselen does not count the trailing slash. base[] may or
853 * may not end with a trailing slash though.
854 */
855 if (pathlen < baselen + 1 ||
856 (baselen && pathname[baselen] != '/') ||
857 strncmp_icase(pathname, base, baselen))
858 return 0;
859
860 namelen = baselen ? pathlen - baselen - 1 : pathlen;
861 name = pathname + pathlen - namelen;
862
863 if (prefix) {
864 /*
865 * if the non-wildcard part is longer than the
866 * remaining pathname, surely it cannot match.
867 */
868 if (prefix > namelen)
869 return 0;
870
871 if (strncmp_icase(pattern, name, prefix))
872 return 0;
873 pattern += prefix;
874 patternlen -= prefix;
875 name += prefix;
876 namelen -= prefix;
877
878 /*
879 * If the whole pattern did not have a wildcard,
880 * then our prefix match is all we need; we
881 * do not need to call fnmatch at all.
882 */
883 if (!patternlen && !namelen)
884 return 1;
885 /*
886 * This can happen when we ignore some exclude rules
887 * on directories in other to see if negative rules
888 * may match. E.g.
889 *
890 * /abc
891 * !/abc/def/ghi
892 *
893 * The pattern of interest is "/abc". On the first
894 * try, we should match path "abc" with this pattern
895 * in the "if" statement right above, but the caller
896 * ignores it.
897 *
898 * On the second try with paths within "abc",
899 * e.g. "abc/xyz", we come here and try to match it
900 * with "/abc".
901 */
902 if (!patternlen && namelen && *name == '/')
903 return 1;
904 }
905
906 return fnmatch_icase_mem(pattern, patternlen,
907 name, namelen,
908 WM_PATHNAME) == 0;
909 }
910
911 /*
912 * Return non-zero if pathname is a directory and an ancestor of the
913 * literal path in a (negative) pattern. This is used to keep
914 * descending in "foo" and "foo/bar" when the pattern is
915 * "!foo/bar/.gitignore". "foo/notbar" will not be descended however.
916 */
917 static int match_neg_path(const char *pathname, int pathlen, int *dtype,
918 const char *base, int baselen,
919 const char *pattern, int prefix, int patternlen,
920 int flags)
921 {
922 assert((flags & EXC_FLAG_NEGATIVE) && !(flags & EXC_FLAG_NODIR));
923
924 if (*dtype == DT_UNKNOWN)
925 *dtype = get_dtype(NULL, pathname, pathlen);
926 if (*dtype != DT_DIR)
927 return 0;
928
929 if (*pattern == '/') {
930 pattern++;
931 patternlen--;
932 prefix--;
933 }
934
935 if (baselen) {
936 if (((pathlen < baselen && base[pathlen] == '/') ||
937 pathlen == baselen) &&
938 !strncmp_icase(pathname, base, pathlen))
939 return 1;
940 pathname += baselen + 1;
941 pathlen -= baselen + 1;
942 }
943
944
945 if (prefix &&
946 ((pathlen < prefix && pattern[pathlen] == '/') &&
947 !strncmp_icase(pathname, pattern, pathlen)))
948 return 1;
949
950 return 0;
951 }
952
953 /*
954 * Scan the given exclude list in reverse to see whether pathname
955 * should be ignored. The first match (i.e. the last on the list), if
956 * any, determines the fate. Returns the exclude_list element which
957 * matched, or NULL for undecided.
958 */
959 static struct exclude *last_exclude_matching_from_list(const char *pathname,
960 int pathlen,
961 const char *basename,
962 int *dtype,
963 struct exclude_list *el)
964 {
965 struct exclude *exc = NULL; /* undecided */
966 int i, matched_negative_path = 0;
967
968 if (!el->nr)
969 return NULL; /* undefined */
970
971 for (i = el->nr - 1; 0 <= i; i--) {
972 struct exclude *x = el->excludes[i];
973 const char *exclude = x->pattern;
974 int prefix = x->nowildcardlen;
975
976 if (x->flags & EXC_FLAG_MUSTBEDIR) {
977 if (*dtype == DT_UNKNOWN)
978 *dtype = get_dtype(NULL, pathname, pathlen);
979 if (*dtype != DT_DIR)
980 continue;
981 }
982
983 if (x->flags & EXC_FLAG_NODIR) {
984 if (match_basename(basename,
985 pathlen - (basename - pathname),
986 exclude, prefix, x->patternlen,
987 x->flags)) {
988 exc = x;
989 break;
990 }
991 continue;
992 }
993
994 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
995 if (match_pathname(pathname, pathlen,
996 x->base, x->baselen ? x->baselen - 1 : 0,
997 exclude, prefix, x->patternlen, x->flags)) {
998 exc = x;
999 break;
1000 }
1001
1002 if ((x->flags & EXC_FLAG_NEGATIVE) && !matched_negative_path &&
1003 match_neg_path(pathname, pathlen, dtype, x->base,
1004 x->baselen ? x->baselen - 1 : 0,
1005 exclude, prefix, x->patternlen, x->flags))
1006 matched_negative_path = 1;
1007 }
1008 if (exc &&
1009 !(exc->flags & EXC_FLAG_NEGATIVE) &&
1010 !(exc->flags & EXC_FLAG_NODIR) &&
1011 matched_negative_path)
1012 exc = NULL;
1013 return exc;
1014 }
1015
1016 /*
1017 * Scan the list and let the last match determine the fate.
1018 * Return 1 for exclude, 0 for include and -1 for undecided.
1019 */
1020 int is_excluded_from_list(const char *pathname,
1021 int pathlen, const char *basename, int *dtype,
1022 struct exclude_list *el)
1023 {
1024 struct exclude *exclude;
1025 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
1026 if (exclude)
1027 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1028 return -1; /* undecided */
1029 }
1030
1031 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
1032 const char *pathname, int pathlen, const char *basename,
1033 int *dtype_p)
1034 {
1035 int i, j;
1036 struct exclude_list_group *group;
1037 struct exclude *exclude;
1038 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1039 group = &dir->exclude_list_group[i];
1040 for (j = group->nr - 1; j >= 0; j--) {
1041 exclude = last_exclude_matching_from_list(
1042 pathname, pathlen, basename, dtype_p,
1043 &group->el[j]);
1044 if (exclude)
1045 return exclude;
1046 }
1047 }
1048 return NULL;
1049 }
1050
1051 /*
1052 * Loads the per-directory exclude list for the substring of base
1053 * which has a char length of baselen.
1054 */
1055 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
1056 {
1057 struct exclude_list_group *group;
1058 struct exclude_list *el;
1059 struct exclude_stack *stk = NULL;
1060 struct untracked_cache_dir *untracked;
1061 int current;
1062
1063 group = &dir->exclude_list_group[EXC_DIRS];
1064
1065 /*
1066 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1067 * which originate from directories not in the prefix of the
1068 * path being checked.
1069 */
1070 while ((stk = dir->exclude_stack) != NULL) {
1071 if (stk->baselen <= baselen &&
1072 !strncmp(dir->basebuf.buf, base, stk->baselen))
1073 break;
1074 el = &group->el[dir->exclude_stack->exclude_ix];
1075 dir->exclude_stack = stk->prev;
1076 dir->exclude = NULL;
1077 free((char *)el->src); /* see strbuf_detach() below */
1078 clear_exclude_list(el);
1079 free(stk);
1080 group->nr--;
1081 }
1082
1083 /* Skip traversing into sub directories if the parent is excluded */
1084 if (dir->exclude)
1085 return;
1086
1087 /*
1088 * Lazy initialization. All call sites currently just
1089 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1090 * them seems lots of work for little benefit.
1091 */
1092 if (!dir->basebuf.buf)
1093 strbuf_init(&dir->basebuf, PATH_MAX);
1094
1095 /* Read from the parent directories and push them down. */
1096 current = stk ? stk->baselen : -1;
1097 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1098 if (dir->untracked)
1099 untracked = stk ? stk->ucd : dir->untracked->root;
1100 else
1101 untracked = NULL;
1102
1103 while (current < baselen) {
1104 const char *cp;
1105 struct sha1_stat sha1_stat;
1106
1107 stk = xcalloc(1, sizeof(*stk));
1108 if (current < 0) {
1109 cp = base;
1110 current = 0;
1111 } else {
1112 cp = strchr(base + current + 1, '/');
1113 if (!cp)
1114 die("oops in prep_exclude");
1115 cp++;
1116 untracked =
1117 lookup_untracked(dir->untracked, untracked,
1118 base + current,
1119 cp - base - current);
1120 }
1121 stk->prev = dir->exclude_stack;
1122 stk->baselen = cp - base;
1123 stk->exclude_ix = group->nr;
1124 stk->ucd = untracked;
1125 el = add_exclude_list(dir, EXC_DIRS, NULL);
1126 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1127 assert(stk->baselen == dir->basebuf.len);
1128
1129 /* Abort if the directory is excluded */
1130 if (stk->baselen) {
1131 int dt = DT_DIR;
1132 dir->basebuf.buf[stk->baselen - 1] = 0;
1133 dir->exclude = last_exclude_matching_from_lists(dir,
1134 dir->basebuf.buf, stk->baselen - 1,
1135 dir->basebuf.buf + current, &dt);
1136 dir->basebuf.buf[stk->baselen - 1] = '/';
1137 if (dir->exclude &&
1138 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1139 dir->exclude = NULL;
1140 if (dir->exclude) {
1141 dir->exclude_stack = stk;
1142 return;
1143 }
1144 }
1145
1146 /* Try to read per-directory file */
1147 hashclr(sha1_stat.sha1);
1148 sha1_stat.valid = 0;
1149 if (dir->exclude_per_dir &&
1150 /*
1151 * If we know that no files have been added in
1152 * this directory (i.e. valid_cached_dir() has
1153 * been executed and set untracked->valid) ..
1154 */
1155 (!untracked || !untracked->valid ||
1156 /*
1157 * .. and .gitignore does not exist before
1158 * (i.e. null exclude_sha1). Then we can skip
1159 * loading .gitignore, which would result in
1160 * ENOENT anyway.
1161 */
1162 !is_null_sha1(untracked->exclude_sha1))) {
1163 /*
1164 * dir->basebuf gets reused by the traversal, but we
1165 * need fname to remain unchanged to ensure the src
1166 * member of each struct exclude correctly
1167 * back-references its source file. Other invocations
1168 * of add_exclude_list provide stable strings, so we
1169 * strbuf_detach() and free() here in the caller.
1170 */
1171 struct strbuf sb = STRBUF_INIT;
1172 strbuf_addbuf(&sb, &dir->basebuf);
1173 strbuf_addstr(&sb, dir->exclude_per_dir);
1174 el->src = strbuf_detach(&sb, NULL);
1175 add_excludes(el->src, el->src, stk->baselen, el, 1,
1176 untracked ? &sha1_stat : NULL);
1177 }
1178 /*
1179 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1180 * will first be called in valid_cached_dir() then maybe many
1181 * times more in last_exclude_matching(). When the cache is
1182 * used, last_exclude_matching() will not be called and
1183 * reading .gitignore content will be a waste.
1184 *
1185 * So when it's called by valid_cached_dir() and we can get
1186 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1187 * modified on work tree), we could delay reading the
1188 * .gitignore content until we absolutely need it in
1189 * last_exclude_matching(). Be careful about ignore rule
1190 * order, though, if you do that.
1191 */
1192 if (untracked &&
1193 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1194 invalidate_gitignore(dir->untracked, untracked);
1195 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1196 }
1197 dir->exclude_stack = stk;
1198 current = stk->baselen;
1199 }
1200 strbuf_setlen(&dir->basebuf, baselen);
1201 }
1202
1203 /*
1204 * Loads the exclude lists for the directory containing pathname, then
1205 * scans all exclude lists to determine whether pathname is excluded.
1206 * Returns the exclude_list element which matched, or NULL for
1207 * undecided.
1208 */
1209 struct exclude *last_exclude_matching(struct dir_struct *dir,
1210 const char *pathname,
1211 int *dtype_p)
1212 {
1213 int pathlen = strlen(pathname);
1214 const char *basename = strrchr(pathname, '/');
1215 basename = (basename) ? basename+1 : pathname;
1216
1217 prep_exclude(dir, pathname, basename-pathname);
1218
1219 if (dir->exclude)
1220 return dir->exclude;
1221
1222 return last_exclude_matching_from_lists(dir, pathname, pathlen,
1223 basename, dtype_p);
1224 }
1225
1226 /*
1227 * Loads the exclude lists for the directory containing pathname, then
1228 * scans all exclude lists to determine whether pathname is excluded.
1229 * Returns 1 if true, otherwise 0.
1230 */
1231 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1232 {
1233 struct exclude *exclude =
1234 last_exclude_matching(dir, pathname, dtype_p);
1235 if (exclude)
1236 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1237 return 0;
1238 }
1239
1240 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1241 {
1242 struct dir_entry *ent;
1243
1244 ent = xmalloc(sizeof(*ent) + len + 1);
1245 ent->len = len;
1246 memcpy(ent->name, pathname, len);
1247 ent->name[len] = 0;
1248 return ent;
1249 }
1250
1251 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1252 {
1253 if (cache_file_exists(pathname, len, ignore_case))
1254 return NULL;
1255
1256 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1257 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1258 }
1259
1260 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1261 {
1262 if (!cache_name_is_other(pathname, len))
1263 return NULL;
1264
1265 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1266 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1267 }
1268
1269 enum exist_status {
1270 index_nonexistent = 0,
1271 index_directory,
1272 index_gitdir
1273 };
1274
1275 /*
1276 * Do not use the alphabetically sorted index to look up
1277 * the directory name; instead, use the case insensitive
1278 * directory hash.
1279 */
1280 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1281 {
1282 struct cache_entry *ce;
1283
1284 if (cache_dir_exists(dirname, len))
1285 return index_directory;
1286
1287 ce = cache_file_exists(dirname, len, ignore_case);
1288 if (ce && S_ISGITLINK(ce->ce_mode))
1289 return index_gitdir;
1290
1291 return index_nonexistent;
1292 }
1293
1294 /*
1295 * The index sorts alphabetically by entry name, which
1296 * means that a gitlink sorts as '\0' at the end, while
1297 * a directory (which is defined not as an entry, but as
1298 * the files it contains) will sort with the '/' at the
1299 * end.
1300 */
1301 static enum exist_status directory_exists_in_index(const char *dirname, int len)
1302 {
1303 int pos;
1304
1305 if (ignore_case)
1306 return directory_exists_in_index_icase(dirname, len);
1307
1308 pos = cache_name_pos(dirname, len);
1309 if (pos < 0)
1310 pos = -pos-1;
1311 while (pos < active_nr) {
1312 const struct cache_entry *ce = active_cache[pos++];
1313 unsigned char endchar;
1314
1315 if (strncmp(ce->name, dirname, len))
1316 break;
1317 endchar = ce->name[len];
1318 if (endchar > '/')
1319 break;
1320 if (endchar == '/')
1321 return index_directory;
1322 if (!endchar && S_ISGITLINK(ce->ce_mode))
1323 return index_gitdir;
1324 }
1325 return index_nonexistent;
1326 }
1327
1328 /*
1329 * When we find a directory when traversing the filesystem, we
1330 * have three distinct cases:
1331 *
1332 * - ignore it
1333 * - see it as a directory
1334 * - recurse into it
1335 *
1336 * and which one we choose depends on a combination of existing
1337 * git index contents and the flags passed into the directory
1338 * traversal routine.
1339 *
1340 * Case 1: If we *already* have entries in the index under that
1341 * directory name, we always recurse into the directory to see
1342 * all the files.
1343 *
1344 * Case 2: If we *already* have that directory name as a gitlink,
1345 * we always continue to see it as a gitlink, regardless of whether
1346 * there is an actual git directory there or not (it might not
1347 * be checked out as a subproject!)
1348 *
1349 * Case 3: if we didn't have it in the index previously, we
1350 * have a few sub-cases:
1351 *
1352 * (a) if "show_other_directories" is true, we show it as
1353 * just a directory, unless "hide_empty_directories" is
1354 * also true, in which case we need to check if it contains any
1355 * untracked and / or ignored files.
1356 * (b) if it looks like a git directory, and we don't have
1357 * 'no_gitlinks' set we treat it as a gitlink, and show it
1358 * as a directory.
1359 * (c) otherwise, we recurse into it.
1360 */
1361 static enum path_treatment treat_directory(struct dir_struct *dir,
1362 struct untracked_cache_dir *untracked,
1363 const char *dirname, int len, int baselen, int exclude,
1364 const struct path_simplify *simplify)
1365 {
1366 /* The "len-1" is to strip the final '/' */
1367 switch (directory_exists_in_index(dirname, len-1)) {
1368 case index_directory:
1369 return path_recurse;
1370
1371 case index_gitdir:
1372 return path_none;
1373
1374 case index_nonexistent:
1375 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1376 break;
1377 if (!(dir->flags & DIR_NO_GITLINKS)) {
1378 unsigned char sha1[20];
1379 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1380 return path_untracked;
1381 }
1382 return path_recurse;
1383 }
1384
1385 /* This is the "show_other_directories" case */
1386
1387 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1388 return exclude ? path_excluded : path_untracked;
1389
1390 untracked = lookup_untracked(dir->untracked, untracked,
1391 dirname + baselen, len - baselen);
1392 return read_directory_recursive(dir, dirname, len,
1393 untracked, 1, simplify);
1394 }
1395
1396 /*
1397 * This is an inexact early pruning of any recursive directory
1398 * reading - if the path cannot possibly be in the pathspec,
1399 * return true, and we'll skip it early.
1400 */
1401 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1402 {
1403 if (simplify) {
1404 for (;;) {
1405 const char *match = simplify->path;
1406 int len = simplify->len;
1407
1408 if (!match)
1409 break;
1410 if (len > pathlen)
1411 len = pathlen;
1412 if (!memcmp(path, match, len))
1413 return 0;
1414 simplify++;
1415 }
1416 return 1;
1417 }
1418 return 0;
1419 }
1420
1421 /*
1422 * This function tells us whether an excluded path matches a
1423 * list of "interesting" pathspecs. That is, whether a path matched
1424 * by any of the pathspecs could possibly be ignored by excluding
1425 * the specified path. This can happen if:
1426 *
1427 * 1. the path is mentioned explicitly in the pathspec
1428 *
1429 * 2. the path is a directory prefix of some element in the
1430 * pathspec
1431 */
1432 static int exclude_matches_pathspec(const char *path, int len,
1433 const struct path_simplify *simplify)
1434 {
1435 if (simplify) {
1436 for (; simplify->path; simplify++) {
1437 if (len == simplify->len
1438 && !memcmp(path, simplify->path, len))
1439 return 1;
1440 if (len < simplify->len
1441 && simplify->path[len] == '/'
1442 && !memcmp(path, simplify->path, len))
1443 return 1;
1444 }
1445 }
1446 return 0;
1447 }
1448
1449 static int get_index_dtype(const char *path, int len)
1450 {
1451 int pos;
1452 const struct cache_entry *ce;
1453
1454 ce = cache_file_exists(path, len, 0);
1455 if (ce) {
1456 if (!ce_uptodate(ce))
1457 return DT_UNKNOWN;
1458 if (S_ISGITLINK(ce->ce_mode))
1459 return DT_DIR;
1460 /*
1461 * Nobody actually cares about the
1462 * difference between DT_LNK and DT_REG
1463 */
1464 return DT_REG;
1465 }
1466
1467 /* Try to look it up as a directory */
1468 pos = cache_name_pos(path, len);
1469 if (pos >= 0)
1470 return DT_UNKNOWN;
1471 pos = -pos-1;
1472 while (pos < active_nr) {
1473 ce = active_cache[pos++];
1474 if (strncmp(ce->name, path, len))
1475 break;
1476 if (ce->name[len] > '/')
1477 break;
1478 if (ce->name[len] < '/')
1479 continue;
1480 if (!ce_uptodate(ce))
1481 break; /* continue? */
1482 return DT_DIR;
1483 }
1484 return DT_UNKNOWN;
1485 }
1486
1487 static int get_dtype(struct dirent *de, const char *path, int len)
1488 {
1489 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1490 struct stat st;
1491
1492 if (dtype != DT_UNKNOWN)
1493 return dtype;
1494 dtype = get_index_dtype(path, len);
1495 if (dtype != DT_UNKNOWN)
1496 return dtype;
1497 if (lstat(path, &st))
1498 return dtype;
1499 if (S_ISREG(st.st_mode))
1500 return DT_REG;
1501 if (S_ISDIR(st.st_mode))
1502 return DT_DIR;
1503 if (S_ISLNK(st.st_mode))
1504 return DT_LNK;
1505 return dtype;
1506 }
1507
1508 static enum path_treatment treat_one_path(struct dir_struct *dir,
1509 struct untracked_cache_dir *untracked,
1510 struct strbuf *path,
1511 int baselen,
1512 const struct path_simplify *simplify,
1513 int dtype, struct dirent *de)
1514 {
1515 int exclude;
1516 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1517
1518 if (dtype == DT_UNKNOWN)
1519 dtype = get_dtype(de, path->buf, path->len);
1520
1521 /* Always exclude indexed files */
1522 if (dtype != DT_DIR && has_path_in_index)
1523 return path_none;
1524
1525 /*
1526 * When we are looking at a directory P in the working tree,
1527 * there are three cases:
1528 *
1529 * (1) P exists in the index. Everything inside the directory P in
1530 * the working tree needs to go when P is checked out from the
1531 * index.
1532 *
1533 * (2) P does not exist in the index, but there is P/Q in the index.
1534 * We know P will stay a directory when we check out the contents
1535 * of the index, but we do not know yet if there is a directory
1536 * P/Q in the working tree to be killed, so we need to recurse.
1537 *
1538 * (3) P does not exist in the index, and there is no P/Q in the index
1539 * to require P to be a directory, either. Only in this case, we
1540 * know that everything inside P will not be killed without
1541 * recursing.
1542 */
1543 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1544 (dtype == DT_DIR) &&
1545 !has_path_in_index &&
1546 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1547 return path_none;
1548
1549 exclude = is_excluded(dir, path->buf, &dtype);
1550
1551 /*
1552 * Excluded? If we don't explicitly want to show
1553 * ignored files, ignore it
1554 */
1555 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1556 return path_excluded;
1557
1558 switch (dtype) {
1559 default:
1560 return path_none;
1561 case DT_DIR:
1562 strbuf_addch(path, '/');
1563 return treat_directory(dir, untracked, path->buf, path->len,
1564 baselen, exclude, simplify);
1565 case DT_REG:
1566 case DT_LNK:
1567 return exclude ? path_excluded : path_untracked;
1568 }
1569 }
1570
1571 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1572 struct untracked_cache_dir *untracked,
1573 struct cached_dir *cdir,
1574 struct strbuf *path,
1575 int baselen,
1576 const struct path_simplify *simplify)
1577 {
1578 strbuf_setlen(path, baselen);
1579 if (!cdir->ucd) {
1580 strbuf_addstr(path, cdir->file);
1581 return path_untracked;
1582 }
1583 strbuf_addstr(path, cdir->ucd->name);
1584 /* treat_one_path() does this before it calls treat_directory() */
1585 strbuf_complete(path, '/');
1586 if (cdir->ucd->check_only)
1587 /*
1588 * check_only is set as a result of treat_directory() getting
1589 * to its bottom. Verify again the same set of directories
1590 * with check_only set.
1591 */
1592 return read_directory_recursive(dir, path->buf, path->len,
1593 cdir->ucd, 1, simplify);
1594 /*
1595 * We get path_recurse in the first run when
1596 * directory_exists_in_index() returns index_nonexistent. We
1597 * are sure that new changes in the index does not impact the
1598 * outcome. Return now.
1599 */
1600 return path_recurse;
1601 }
1602
1603 static enum path_treatment treat_path(struct dir_struct *dir,
1604 struct untracked_cache_dir *untracked,
1605 struct cached_dir *cdir,
1606 struct strbuf *path,
1607 int baselen,
1608 const struct path_simplify *simplify)
1609 {
1610 int dtype;
1611 struct dirent *de = cdir->de;
1612
1613 if (!de)
1614 return treat_path_fast(dir, untracked, cdir, path,
1615 baselen, simplify);
1616 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1617 return path_none;
1618 strbuf_setlen(path, baselen);
1619 strbuf_addstr(path, de->d_name);
1620 if (simplify_away(path->buf, path->len, simplify))
1621 return path_none;
1622
1623 dtype = DTYPE(de);
1624 return treat_one_path(dir, untracked, path, baselen, simplify, dtype, de);
1625 }
1626
1627 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1628 {
1629 if (!dir)
1630 return;
1631 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1632 dir->untracked_alloc);
1633 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1634 }
1635
1636 static int valid_cached_dir(struct dir_struct *dir,
1637 struct untracked_cache_dir *untracked,
1638 struct strbuf *path,
1639 int check_only)
1640 {
1641 struct stat st;
1642
1643 if (!untracked)
1644 return 0;
1645
1646 if (stat(path->len ? path->buf : ".", &st)) {
1647 invalidate_directory(dir->untracked, untracked);
1648 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1649 return 0;
1650 }
1651 if (!untracked->valid ||
1652 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
1653 if (untracked->valid)
1654 invalidate_directory(dir->untracked, untracked);
1655 fill_stat_data(&untracked->stat_data, &st);
1656 return 0;
1657 }
1658
1659 if (untracked->check_only != !!check_only) {
1660 invalidate_directory(dir->untracked, untracked);
1661 return 0;
1662 }
1663
1664 /*
1665 * prep_exclude will be called eventually on this directory,
1666 * but it's called much later in last_exclude_matching(). We
1667 * need it now to determine the validity of the cache for this
1668 * path. The next calls will be nearly no-op, the way
1669 * prep_exclude() is designed.
1670 */
1671 if (path->len && path->buf[path->len - 1] != '/') {
1672 strbuf_addch(path, '/');
1673 prep_exclude(dir, path->buf, path->len);
1674 strbuf_setlen(path, path->len - 1);
1675 } else
1676 prep_exclude(dir, path->buf, path->len);
1677
1678 /* hopefully prep_exclude() haven't invalidated this entry... */
1679 return untracked->valid;
1680 }
1681
1682 static int open_cached_dir(struct cached_dir *cdir,
1683 struct dir_struct *dir,
1684 struct untracked_cache_dir *untracked,
1685 struct strbuf *path,
1686 int check_only)
1687 {
1688 memset(cdir, 0, sizeof(*cdir));
1689 cdir->untracked = untracked;
1690 if (valid_cached_dir(dir, untracked, path, check_only))
1691 return 0;
1692 cdir->fdir = opendir(path->len ? path->buf : ".");
1693 if (dir->untracked)
1694 dir->untracked->dir_opened++;
1695 if (!cdir->fdir)
1696 return -1;
1697 return 0;
1698 }
1699
1700 static int read_cached_dir(struct cached_dir *cdir)
1701 {
1702 if (cdir->fdir) {
1703 cdir->de = readdir(cdir->fdir);
1704 if (!cdir->de)
1705 return -1;
1706 return 0;
1707 }
1708 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1709 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1710 if (!d->recurse) {
1711 cdir->nr_dirs++;
1712 continue;
1713 }
1714 cdir->ucd = d;
1715 cdir->nr_dirs++;
1716 return 0;
1717 }
1718 cdir->ucd = NULL;
1719 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1720 struct untracked_cache_dir *d = cdir->untracked;
1721 cdir->file = d->untracked[cdir->nr_files++];
1722 return 0;
1723 }
1724 return -1;
1725 }
1726
1727 static void close_cached_dir(struct cached_dir *cdir)
1728 {
1729 if (cdir->fdir)
1730 closedir(cdir->fdir);
1731 /*
1732 * We have gone through this directory and found no untracked
1733 * entries. Mark it valid.
1734 */
1735 if (cdir->untracked) {
1736 cdir->untracked->valid = 1;
1737 cdir->untracked->recurse = 1;
1738 }
1739 }
1740
1741 /*
1742 * Read a directory tree. We currently ignore anything but
1743 * directories, regular files and symlinks. That's because git
1744 * doesn't handle them at all yet. Maybe that will change some
1745 * day.
1746 *
1747 * Also, we ignore the name ".git" (even if it is not a directory).
1748 * That likely will not change.
1749 *
1750 * Returns the most significant path_treatment value encountered in the scan.
1751 */
1752 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1753 const char *base, int baselen,
1754 struct untracked_cache_dir *untracked, int check_only,
1755 const struct path_simplify *simplify)
1756 {
1757 struct cached_dir cdir;
1758 enum path_treatment state, subdir_state, dir_state = path_none;
1759 struct strbuf path = STRBUF_INIT;
1760
1761 strbuf_add(&path, base, baselen);
1762
1763 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1764 goto out;
1765
1766 if (untracked)
1767 untracked->check_only = !!check_only;
1768
1769 while (!read_cached_dir(&cdir)) {
1770 /* check how the file or directory should be treated */
1771 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);
1772
1773 if (state > dir_state)
1774 dir_state = state;
1775
1776 /* recurse into subdir if instructed by treat_path */
1777 if (state == path_recurse) {
1778 struct untracked_cache_dir *ud;
1779 ud = lookup_untracked(dir->untracked, untracked,
1780 path.buf + baselen,
1781 path.len - baselen);
1782 subdir_state =
1783 read_directory_recursive(dir, path.buf, path.len,
1784 ud, check_only, simplify);
1785 if (subdir_state > dir_state)
1786 dir_state = subdir_state;
1787 }
1788
1789 if (check_only) {
1790 /* abort early if maximum state has been reached */
1791 if (dir_state == path_untracked) {
1792 if (cdir.fdir)
1793 add_untracked(untracked, path.buf + baselen);
1794 break;
1795 }
1796 /* skip the dir_add_* part */
1797 continue;
1798 }
1799
1800 /* add the path to the appropriate result list */
1801 switch (state) {
1802 case path_excluded:
1803 if (dir->flags & DIR_SHOW_IGNORED)
1804 dir_add_name(dir, path.buf, path.len);
1805 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1806 ((dir->flags & DIR_COLLECT_IGNORED) &&
1807 exclude_matches_pathspec(path.buf, path.len,
1808 simplify)))
1809 dir_add_ignored(dir, path.buf, path.len);
1810 break;
1811
1812 case path_untracked:
1813 if (dir->flags & DIR_SHOW_IGNORED)
1814 break;
1815 dir_add_name(dir, path.buf, path.len);
1816 if (cdir.fdir)
1817 add_untracked(untracked, path.buf + baselen);
1818 break;
1819
1820 default:
1821 break;
1822 }
1823 }
1824 close_cached_dir(&cdir);
1825 out:
1826 strbuf_release(&path);
1827
1828 return dir_state;
1829 }
1830
1831 static int cmp_name(const void *p1, const void *p2)
1832 {
1833 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1834 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1835
1836 return name_compare(e1->name, e1->len, e2->name, e2->len);
1837 }
1838
1839 static struct path_simplify *create_simplify(const char **pathspec)
1840 {
1841 int nr, alloc = 0;
1842 struct path_simplify *simplify = NULL;
1843
1844 if (!pathspec)
1845 return NULL;
1846
1847 for (nr = 0 ; ; nr++) {
1848 const char *match;
1849 ALLOC_GROW(simplify, nr + 1, alloc);
1850 match = *pathspec++;
1851 if (!match)
1852 break;
1853 simplify[nr].path = match;
1854 simplify[nr].len = simple_length(match);
1855 }
1856 simplify[nr].path = NULL;
1857 simplify[nr].len = 0;
1858 return simplify;
1859 }
1860
1861 static void free_simplify(struct path_simplify *simplify)
1862 {
1863 free(simplify);
1864 }
1865
1866 static int treat_leading_path(struct dir_struct *dir,
1867 const char *path, int len,
1868 const struct path_simplify *simplify)
1869 {
1870 struct strbuf sb = STRBUF_INIT;
1871 int baselen, rc = 0;
1872 const char *cp;
1873 int old_flags = dir->flags;
1874
1875 while (len && path[len - 1] == '/')
1876 len--;
1877 if (!len)
1878 return 1;
1879 baselen = 0;
1880 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1881 while (1) {
1882 cp = path + baselen + !!baselen;
1883 cp = memchr(cp, '/', path + len - cp);
1884 if (!cp)
1885 baselen = len;
1886 else
1887 baselen = cp - path;
1888 strbuf_setlen(&sb, 0);
1889 strbuf_add(&sb, path, baselen);
1890 if (!is_directory(sb.buf))
1891 break;
1892 if (simplify_away(sb.buf, sb.len, simplify))
1893 break;
1894 if (treat_one_path(dir, NULL, &sb, baselen, simplify,
1895 DT_DIR, NULL) == path_none)
1896 break; /* do not recurse into it */
1897 if (len <= baselen) {
1898 rc = 1;
1899 break; /* finished checking */
1900 }
1901 }
1902 strbuf_release(&sb);
1903 dir->flags = old_flags;
1904 return rc;
1905 }
1906
1907 static const char *get_ident_string(void)
1908 {
1909 static struct strbuf sb = STRBUF_INIT;
1910 struct utsname uts;
1911
1912 if (sb.len)
1913 return sb.buf;
1914 if (uname(&uts) < 0)
1915 die_errno(_("failed to get kernel name and information"));
1916 strbuf_addf(&sb, "Location %s, system %s %s %s", get_git_work_tree(),
1917 uts.sysname, uts.release, uts.version);
1918 return sb.buf;
1919 }
1920
1921 static int ident_in_untracked(const struct untracked_cache *uc)
1922 {
1923 const char *end = uc->ident.buf + uc->ident.len;
1924 const char *p = uc->ident.buf;
1925
1926 for (p = uc->ident.buf; p < end; p += strlen(p) + 1)
1927 if (!strcmp(p, get_ident_string()))
1928 return 1;
1929 return 0;
1930 }
1931
1932 void add_untracked_ident(struct untracked_cache *uc)
1933 {
1934 if (ident_in_untracked(uc))
1935 return;
1936 strbuf_addstr(&uc->ident, get_ident_string());
1937 /* this strbuf contains a list of strings, save NUL too */
1938 strbuf_addch(&uc->ident, 0);
1939 }
1940
1941 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1942 int base_len,
1943 const struct pathspec *pathspec)
1944 {
1945 struct untracked_cache_dir *root;
1946
1947 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
1948 return NULL;
1949
1950 /*
1951 * We only support $GIT_DIR/info/exclude and core.excludesfile
1952 * as the global ignore rule files. Any other additions
1953 * (e.g. from command line) invalidate the cache. This
1954 * condition also catches running setup_standard_excludes()
1955 * before setting dir->untracked!
1956 */
1957 if (dir->unmanaged_exclude_files)
1958 return NULL;
1959
1960 /*
1961 * Optimize for the main use case only: whole-tree git
1962 * status. More work involved in treat_leading_path() if we
1963 * use cache on just a subset of the worktree. pathspec
1964 * support could make the matter even worse.
1965 */
1966 if (base_len || (pathspec && pathspec->nr))
1967 return NULL;
1968
1969 /* Different set of flags may produce different results */
1970 if (dir->flags != dir->untracked->dir_flags ||
1971 /*
1972 * See treat_directory(), case index_nonexistent. Without
1973 * this flag, we may need to also cache .git file content
1974 * for the resolve_gitlink_ref() call, which we don't.
1975 */
1976 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1977 /* We don't support collecting ignore files */
1978 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1979 DIR_COLLECT_IGNORED)))
1980 return NULL;
1981
1982 /*
1983 * If we use .gitignore in the cache and now you change it to
1984 * .gitexclude, everything will go wrong.
1985 */
1986 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
1987 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
1988 return NULL;
1989
1990 /*
1991 * EXC_CMDL is not considered in the cache. If people set it,
1992 * skip the cache.
1993 */
1994 if (dir->exclude_list_group[EXC_CMDL].nr)
1995 return NULL;
1996
1997 if (!ident_in_untracked(dir->untracked)) {
1998 warning(_("Untracked cache is disabled on this system."));
1999 return NULL;
2000 }
2001
2002 if (!dir->untracked->root) {
2003 const int len = sizeof(*dir->untracked->root);
2004 dir->untracked->root = xmalloc(len);
2005 memset(dir->untracked->root, 0, len);
2006 }
2007
2008 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2009 root = dir->untracked->root;
2010 if (hashcmp(dir->ss_info_exclude.sha1,
2011 dir->untracked->ss_info_exclude.sha1)) {
2012 invalidate_gitignore(dir->untracked, root);
2013 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2014 }
2015 if (hashcmp(dir->ss_excludes_file.sha1,
2016 dir->untracked->ss_excludes_file.sha1)) {
2017 invalidate_gitignore(dir->untracked, root);
2018 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2019 }
2020
2021 /* Make sure this directory is not dropped out at saving phase */
2022 root->recurse = 1;
2023 return root;
2024 }
2025
2026 int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
2027 {
2028 struct path_simplify *simplify;
2029 struct untracked_cache_dir *untracked;
2030
2031 /*
2032 * Check out create_simplify()
2033 */
2034 if (pathspec)
2035 GUARD_PATHSPEC(pathspec,
2036 PATHSPEC_FROMTOP |
2037 PATHSPEC_MAXDEPTH |
2038 PATHSPEC_LITERAL |
2039 PATHSPEC_GLOB |
2040 PATHSPEC_ICASE |
2041 PATHSPEC_EXCLUDE);
2042
2043 if (has_symlink_leading_path(path, len))
2044 return dir->nr;
2045
2046 /*
2047 * exclude patterns are treated like positive ones in
2048 * create_simplify. Usually exclude patterns should be a
2049 * subset of positive ones, which has no impacts on
2050 * create_simplify().
2051 */
2052 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
2053 untracked = validate_untracked_cache(dir, len, pathspec);
2054 if (!untracked)
2055 /*
2056 * make sure untracked cache code path is disabled,
2057 * e.g. prep_exclude()
2058 */
2059 dir->untracked = NULL;
2060 if (!len || treat_leading_path(dir, path, len, simplify))
2061 read_directory_recursive(dir, path, len, untracked, 0, simplify);
2062 free_simplify(simplify);
2063 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
2064 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
2065 if (dir->untracked) {
2066 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2067 trace_printf_key(&trace_untracked_stats,
2068 "node creation: %u\n"
2069 "gitignore invalidation: %u\n"
2070 "directory invalidation: %u\n"
2071 "opendir: %u\n",
2072 dir->untracked->dir_created,
2073 dir->untracked->gitignore_invalidated,
2074 dir->untracked->dir_invalidated,
2075 dir->untracked->dir_opened);
2076 if (dir->untracked == the_index.untracked &&
2077 (dir->untracked->dir_opened ||
2078 dir->untracked->gitignore_invalidated ||
2079 dir->untracked->dir_invalidated))
2080 the_index.cache_changed |= UNTRACKED_CHANGED;
2081 if (dir->untracked != the_index.untracked) {
2082 free(dir->untracked);
2083 dir->untracked = NULL;
2084 }
2085 }
2086 return dir->nr;
2087 }
2088
2089 int file_exists(const char *f)
2090 {
2091 struct stat sb;
2092 return lstat(f, &sb) == 0;
2093 }
2094
2095 static int cmp_icase(char a, char b)
2096 {
2097 if (a == b)
2098 return 0;
2099 if (ignore_case)
2100 return toupper(a) - toupper(b);
2101 return a - b;
2102 }
2103
2104 /*
2105 * Given two normalized paths (a trailing slash is ok), if subdir is
2106 * outside dir, return -1. Otherwise return the offset in subdir that
2107 * can be used as relative path to dir.
2108 */
2109 int dir_inside_of(const char *subdir, const char *dir)
2110 {
2111 int offset = 0;
2112
2113 assert(dir && subdir && *dir && *subdir);
2114
2115 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2116 dir++;
2117 subdir++;
2118 offset++;
2119 }
2120
2121 /* hel[p]/me vs hel[l]/yeah */
2122 if (*dir && *subdir)
2123 return -1;
2124
2125 if (!*subdir)
2126 return !*dir ? offset : -1; /* same dir */
2127
2128 /* foo/[b]ar vs foo/[] */
2129 if (is_dir_sep(dir[-1]))
2130 return is_dir_sep(subdir[-1]) ? offset : -1;
2131
2132 /* foo[/]bar vs foo[] */
2133 return is_dir_sep(*subdir) ? offset + 1 : -1;
2134 }
2135
2136 int is_inside_dir(const char *dir)
2137 {
2138 char *cwd;
2139 int rc;
2140
2141 if (!dir)
2142 return 0;
2143
2144 cwd = xgetcwd();
2145 rc = (dir_inside_of(cwd, dir) >= 0);
2146 free(cwd);
2147 return rc;
2148 }
2149
2150 int is_empty_dir(const char *path)
2151 {
2152 DIR *dir = opendir(path);
2153 struct dirent *e;
2154 int ret = 1;
2155
2156 if (!dir)
2157 return 0;
2158
2159 while ((e = readdir(dir)) != NULL)
2160 if (!is_dot_or_dotdot(e->d_name)) {
2161 ret = 0;
2162 break;
2163 }
2164
2165 closedir(dir);
2166 return ret;
2167 }
2168
2169 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2170 {
2171 DIR *dir;
2172 struct dirent *e;
2173 int ret = 0, original_len = path->len, len, kept_down = 0;
2174 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2175 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2176 unsigned char submodule_head[20];
2177
2178 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2179 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2180 /* Do not descend and nuke a nested git work tree. */
2181 if (kept_up)
2182 *kept_up = 1;
2183 return 0;
2184 }
2185
2186 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2187 dir = opendir(path->buf);
2188 if (!dir) {
2189 if (errno == ENOENT)
2190 return keep_toplevel ? -1 : 0;
2191 else if (errno == EACCES && !keep_toplevel)
2192 /*
2193 * An empty dir could be removable even if it
2194 * is unreadable:
2195 */
2196 return rmdir(path->buf);
2197 else
2198 return -1;
2199 }
2200 strbuf_complete(path, '/');
2201
2202 len = path->len;
2203 while ((e = readdir(dir)) != NULL) {
2204 struct stat st;
2205 if (is_dot_or_dotdot(e->d_name))
2206 continue;
2207
2208 strbuf_setlen(path, len);
2209 strbuf_addstr(path, e->d_name);
2210 if (lstat(path->buf, &st)) {
2211 if (errno == ENOENT)
2212 /*
2213 * file disappeared, which is what we
2214 * wanted anyway
2215 */
2216 continue;
2217 /* fall thru */
2218 } else if (S_ISDIR(st.st_mode)) {
2219 if (!remove_dir_recurse(path, flag, &kept_down))
2220 continue; /* happy */
2221 } else if (!only_empty &&
2222 (!unlink(path->buf) || errno == ENOENT)) {
2223 continue; /* happy, too */
2224 }
2225
2226 /* path too long, stat fails, or non-directory still exists */
2227 ret = -1;
2228 break;
2229 }
2230 closedir(dir);
2231
2232 strbuf_setlen(path, original_len);
2233 if (!ret && !keep_toplevel && !kept_down)
2234 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2235 else if (kept_up)
2236 /*
2237 * report the uplevel that it is not an error that we
2238 * did not rmdir() our directory.
2239 */
2240 *kept_up = !ret;
2241 return ret;
2242 }
2243
2244 int remove_dir_recursively(struct strbuf *path, int flag)
2245 {
2246 return remove_dir_recurse(path, flag, NULL);
2247 }
2248
2249 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2250
2251 void setup_standard_excludes(struct dir_struct *dir)
2252 {
2253 const char *path;
2254
2255 dir->exclude_per_dir = ".gitignore";
2256
2257 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2258 if (!excludes_file)
2259 excludes_file = xdg_config_home("ignore");
2260 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2261 add_excludes_from_file_1(dir, excludes_file,
2262 dir->untracked ? &dir->ss_excludes_file : NULL);
2263
2264 /* per repository user preference */
2265 path = git_path_info_exclude();
2266 if (!access_or_warn(path, R_OK, 0))
2267 add_excludes_from_file_1(dir, path,
2268 dir->untracked ? &dir->ss_info_exclude : NULL);
2269 }
2270
2271 int remove_path(const char *name)
2272 {
2273 char *slash;
2274
2275 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
2276 return -1;
2277
2278 slash = strrchr(name, '/');
2279 if (slash) {
2280 char *dirs = xstrdup(name);
2281 slash = dirs + (slash - name);
2282 do {
2283 *slash = '\0';
2284 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2285 free(dirs);
2286 }
2287 return 0;
2288 }
2289
2290 /*
2291 * Frees memory within dir which was allocated for exclude lists and
2292 * the exclude_stack. Does not free dir itself.
2293 */
2294 void clear_directory(struct dir_struct *dir)
2295 {
2296 int i, j;
2297 struct exclude_list_group *group;
2298 struct exclude_list *el;
2299 struct exclude_stack *stk;
2300
2301 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2302 group = &dir->exclude_list_group[i];
2303 for (j = 0; j < group->nr; j++) {
2304 el = &group->el[j];
2305 if (i == EXC_DIRS)
2306 free((char *)el->src);
2307 clear_exclude_list(el);
2308 }
2309 free(group->el);
2310 }
2311
2312 stk = dir->exclude_stack;
2313 while (stk) {
2314 struct exclude_stack *prev = stk->prev;
2315 free(stk);
2316 stk = prev;
2317 }
2318 strbuf_release(&dir->basebuf);
2319 }
2320
2321 struct ondisk_untracked_cache {
2322 struct stat_data info_exclude_stat;
2323 struct stat_data excludes_file_stat;
2324 uint32_t dir_flags;
2325 unsigned char info_exclude_sha1[20];
2326 unsigned char excludes_file_sha1[20];
2327 char exclude_per_dir[FLEX_ARRAY];
2328 };
2329
2330 #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2331
2332 struct write_data {
2333 int index; /* number of written untracked_cache_dir */
2334 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2335 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2336 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2337 struct strbuf out;
2338 struct strbuf sb_stat;
2339 struct strbuf sb_sha1;
2340 };
2341
2342 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2343 {
2344 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2345 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2346 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2347 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2348 to->sd_dev = htonl(from->sd_dev);
2349 to->sd_ino = htonl(from->sd_ino);
2350 to->sd_uid = htonl(from->sd_uid);
2351 to->sd_gid = htonl(from->sd_gid);
2352 to->sd_size = htonl(from->sd_size);
2353 }
2354
2355 static void write_one_dir(struct untracked_cache_dir *untracked,
2356 struct write_data *wd)
2357 {
2358 struct stat_data stat_data;
2359 struct strbuf *out = &wd->out;
2360 unsigned char intbuf[16];
2361 unsigned int intlen, value;
2362 int i = wd->index++;
2363
2364 /*
2365 * untracked_nr should be reset whenever valid is clear, but
2366 * for safety..
2367 */
2368 if (!untracked->valid) {
2369 untracked->untracked_nr = 0;
2370 untracked->check_only = 0;
2371 }
2372
2373 if (untracked->check_only)
2374 ewah_set(wd->check_only, i);
2375 if (untracked->valid) {
2376 ewah_set(wd->valid, i);
2377 stat_data_to_disk(&stat_data, &untracked->stat_data);
2378 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2379 }
2380 if (!is_null_sha1(untracked->exclude_sha1)) {
2381 ewah_set(wd->sha1_valid, i);
2382 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2383 }
2384
2385 intlen = encode_varint(untracked->untracked_nr, intbuf);
2386 strbuf_add(out, intbuf, intlen);
2387
2388 /* skip non-recurse directories */
2389 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2390 if (untracked->dirs[i]->recurse)
2391 value++;
2392 intlen = encode_varint(value, intbuf);
2393 strbuf_add(out, intbuf, intlen);
2394
2395 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2396
2397 for (i = 0; i < untracked->untracked_nr; i++)
2398 strbuf_add(out, untracked->untracked[i],
2399 strlen(untracked->untracked[i]) + 1);
2400
2401 for (i = 0; i < untracked->dirs_nr; i++)
2402 if (untracked->dirs[i]->recurse)
2403 write_one_dir(untracked->dirs[i], wd);
2404 }
2405
2406 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2407 {
2408 struct ondisk_untracked_cache *ouc;
2409 struct write_data wd;
2410 unsigned char varbuf[16];
2411 int len = 0, varint_len;
2412 if (untracked->exclude_per_dir)
2413 len = strlen(untracked->exclude_per_dir);
2414 ouc = xmalloc(sizeof(*ouc) + len + 1);
2415 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2416 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2417 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2418 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2419 ouc->dir_flags = htonl(untracked->dir_flags);
2420 memcpy(ouc->exclude_per_dir, untracked->exclude_per_dir, len + 1);
2421
2422 varint_len = encode_varint(untracked->ident.len, varbuf);
2423 strbuf_add(out, varbuf, varint_len);
2424 strbuf_add(out, untracked->ident.buf, untracked->ident.len);
2425
2426 strbuf_add(out, ouc, ouc_size(len));
2427 free(ouc);
2428 ouc = NULL;
2429
2430 if (!untracked->root) {
2431 varint_len = encode_varint(0, varbuf);
2432 strbuf_add(out, varbuf, varint_len);
2433 return;
2434 }
2435
2436 wd.index = 0;
2437 wd.check_only = ewah_new();
2438 wd.valid = ewah_new();
2439 wd.sha1_valid = ewah_new();
2440 strbuf_init(&wd.out, 1024);
2441 strbuf_init(&wd.sb_stat, 1024);
2442 strbuf_init(&wd.sb_sha1, 1024);
2443 write_one_dir(untracked->root, &wd);
2444
2445 varint_len = encode_varint(wd.index, varbuf);
2446 strbuf_add(out, varbuf, varint_len);
2447 strbuf_addbuf(out, &wd.out);
2448 ewah_serialize_strbuf(wd.valid, out);
2449 ewah_serialize_strbuf(wd.check_only, out);
2450 ewah_serialize_strbuf(wd.sha1_valid, out);
2451 strbuf_addbuf(out, &wd.sb_stat);
2452 strbuf_addbuf(out, &wd.sb_sha1);
2453 strbuf_addch(out, '\0'); /* safe guard for string lists */
2454
2455 ewah_free(wd.valid);
2456 ewah_free(wd.check_only);
2457 ewah_free(wd.sha1_valid);
2458 strbuf_release(&wd.out);
2459 strbuf_release(&wd.sb_stat);
2460 strbuf_release(&wd.sb_sha1);
2461 }
2462
2463 static void free_untracked(struct untracked_cache_dir *ucd)
2464 {
2465 int i;
2466 if (!ucd)
2467 return;
2468 for (i = 0; i < ucd->dirs_nr; i++)
2469 free_untracked(ucd->dirs[i]);
2470 for (i = 0; i < ucd->untracked_nr; i++)
2471 free(ucd->untracked[i]);
2472 free(ucd->untracked);
2473 free(ucd->dirs);
2474 free(ucd);
2475 }
2476
2477 void free_untracked_cache(struct untracked_cache *uc)
2478 {
2479 if (uc)
2480 free_untracked(uc->root);
2481 free(uc);
2482 }
2483
2484 struct read_data {
2485 int index;
2486 struct untracked_cache_dir **ucd;
2487 struct ewah_bitmap *check_only;
2488 struct ewah_bitmap *valid;
2489 struct ewah_bitmap *sha1_valid;
2490 const unsigned char *data;
2491 const unsigned char *end;
2492 };
2493
2494 static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2495 {
2496 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2497 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2498 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2499 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2500 to->sd_dev = get_be32(&from->sd_dev);
2501 to->sd_ino = get_be32(&from->sd_ino);
2502 to->sd_uid = get_be32(&from->sd_uid);
2503 to->sd_gid = get_be32(&from->sd_gid);
2504 to->sd_size = get_be32(&from->sd_size);
2505 }
2506
2507 static int read_one_dir(struct untracked_cache_dir **untracked_,
2508 struct read_data *rd)
2509 {
2510 struct untracked_cache_dir ud, *untracked;
2511 const unsigned char *next, *data = rd->data, *end = rd->end;
2512 unsigned int value;
2513 int i, len;
2514
2515 memset(&ud, 0, sizeof(ud));
2516
2517 next = data;
2518 value = decode_varint(&next);
2519 if (next > end)
2520 return -1;
2521 ud.recurse = 1;
2522 ud.untracked_alloc = value;
2523 ud.untracked_nr = value;
2524 if (ud.untracked_nr)
2525 ud.untracked = xmalloc(sizeof(*ud.untracked) * ud.untracked_nr);
2526 data = next;
2527
2528 next = data;
2529 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2530 if (next > end)
2531 return -1;
2532 ud.dirs = xmalloc(sizeof(*ud.dirs) * ud.dirs_nr);
2533 data = next;
2534
2535 len = strlen((const char *)data);
2536 next = data + len + 1;
2537 if (next > rd->end)
2538 return -1;
2539 *untracked_ = untracked = xmalloc(sizeof(*untracked) + len);
2540 memcpy(untracked, &ud, sizeof(ud));
2541 memcpy(untracked->name, data, len + 1);
2542 data = next;
2543
2544 for (i = 0; i < untracked->untracked_nr; i++) {
2545 len = strlen((const char *)data);
2546 next = data + len + 1;
2547 if (next > rd->end)
2548 return -1;
2549 untracked->untracked[i] = xstrdup((const char*)data);
2550 data = next;
2551 }
2552
2553 rd->ucd[rd->index++] = untracked;
2554 rd->data = data;
2555
2556 for (i = 0; i < untracked->dirs_nr; i++) {
2557 len = read_one_dir(untracked->dirs + i, rd);
2558 if (len < 0)
2559 return -1;
2560 }
2561 return 0;
2562 }
2563
2564 static void set_check_only(size_t pos, void *cb)
2565 {
2566 struct read_data *rd = cb;
2567 struct untracked_cache_dir *ud = rd->ucd[pos];
2568 ud->check_only = 1;
2569 }
2570
2571 static void read_stat(size_t pos, void *cb)
2572 {
2573 struct read_data *rd = cb;
2574 struct untracked_cache_dir *ud = rd->ucd[pos];
2575 if (rd->data + sizeof(struct stat_data) > rd->end) {
2576 rd->data = rd->end + 1;
2577 return;
2578 }
2579 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2580 rd->data += sizeof(struct stat_data);
2581 ud->valid = 1;
2582 }
2583
2584 static void read_sha1(size_t pos, void *cb)
2585 {
2586 struct read_data *rd = cb;
2587 struct untracked_cache_dir *ud = rd->ucd[pos];
2588 if (rd->data + 20 > rd->end) {
2589 rd->data = rd->end + 1;
2590 return;
2591 }
2592 hashcpy(ud->exclude_sha1, rd->data);
2593 rd->data += 20;
2594 }
2595
2596 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2597 const struct stat_data *stat,
2598 const unsigned char *sha1)
2599 {
2600 stat_data_from_disk(&sha1_stat->stat, stat);
2601 hashcpy(sha1_stat->sha1, sha1);
2602 sha1_stat->valid = 1;
2603 }
2604
2605 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2606 {
2607 const struct ondisk_untracked_cache *ouc;
2608 struct untracked_cache *uc;
2609 struct read_data rd;
2610 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2611 const char *ident;
2612 int ident_len, len;
2613
2614 if (sz <= 1 || end[-1] != '\0')
2615 return NULL;
2616 end--;
2617
2618 ident_len = decode_varint(&next);
2619 if (next + ident_len > end)
2620 return NULL;
2621 ident = (const char *)next;
2622 next += ident_len;
2623
2624 ouc = (const struct ondisk_untracked_cache *)next;
2625 if (next + ouc_size(0) > end)
2626 return NULL;
2627
2628 uc = xcalloc(1, sizeof(*uc));
2629 strbuf_init(&uc->ident, ident_len);
2630 strbuf_add(&uc->ident, ident, ident_len);
2631 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2632 ouc->info_exclude_sha1);
2633 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2634 ouc->excludes_file_sha1);
2635 uc->dir_flags = get_be32(&ouc->dir_flags);
2636 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2637 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2638 next += ouc_size(strlen(ouc->exclude_per_dir));
2639 if (next >= end)
2640 goto done2;
2641
2642 len = decode_varint(&next);
2643 if (next > end || len == 0)
2644 goto done2;
2645
2646 rd.valid = ewah_new();
2647 rd.check_only = ewah_new();
2648 rd.sha1_valid = ewah_new();
2649 rd.data = next;
2650 rd.end = end;
2651 rd.index = 0;
2652 rd.ucd = xmalloc(sizeof(*rd.ucd) * len);
2653
2654 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2655 goto done;
2656
2657 next = rd.data;
2658 len = ewah_read_mmap(rd.valid, next, end - next);
2659 if (len < 0)
2660 goto done;
2661
2662 next += len;
2663 len = ewah_read_mmap(rd.check_only, next, end - next);
2664 if (len < 0)
2665 goto done;
2666
2667 next += len;
2668 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2669 if (len < 0)
2670 goto done;
2671
2672 ewah_each_bit(rd.check_only, set_check_only, &rd);
2673 rd.data = next + len;
2674 ewah_each_bit(rd.valid, read_stat, &rd);
2675 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2676 next = rd.data;
2677
2678 done:
2679 free(rd.ucd);
2680 ewah_free(rd.valid);
2681 ewah_free(rd.check_only);
2682 ewah_free(rd.sha1_valid);
2683 done2:
2684 if (next != end) {
2685 free_untracked_cache(uc);
2686 uc = NULL;
2687 }
2688 return uc;
2689 }
2690
2691 static void invalidate_one_directory(struct untracked_cache *uc,
2692 struct untracked_cache_dir *ucd)
2693 {
2694 uc->dir_invalidated++;
2695 ucd->valid = 0;
2696 ucd->untracked_nr = 0;
2697 }
2698
2699 /*
2700 * Normally when an entry is added or removed from a directory,
2701 * invalidating that directory is enough. No need to touch its
2702 * ancestors. When a directory is shown as "foo/bar/" in git-status
2703 * however, deleting or adding an entry may have cascading effect.
2704 *
2705 * Say the "foo/bar/file" has become untracked, we need to tell the
2706 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2707 * directory any more (because "bar" is managed by foo as an untracked
2708 * "file").
2709 *
2710 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2711 * was the last untracked entry in the entire "foo", we should show
2712 * "foo/" instead. Which means we have to invalidate past "bar" up to
2713 * "foo".
2714 *
2715 * This function traverses all directories from root to leaf. If there
2716 * is a chance of one of the above cases happening, we invalidate back
2717 * to root. Otherwise we just invalidate the leaf. There may be a more
2718 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2719 * detect these cases and avoid unnecessary invalidation, for example,
2720 * checking for the untracked entry named "bar/" in "foo", but for now
2721 * stick to something safe and simple.
2722 */
2723 static int invalidate_one_component(struct untracked_cache *uc,
2724 struct untracked_cache_dir *dir,
2725 const char *path, int len)
2726 {
2727 const char *rest = strchr(path, '/');
2728
2729 if (rest) {
2730 int component_len = rest - path;
2731 struct untracked_cache_dir *d =
2732 lookup_untracked(uc, dir, path, component_len);
2733 int ret =
2734 invalidate_one_component(uc, d, rest + 1,
2735 len - (component_len + 1));
2736 if (ret)
2737 invalidate_one_directory(uc, dir);
2738 return ret;
2739 }
2740
2741 invalidate_one_directory(uc, dir);
2742 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2743 }
2744
2745 void untracked_cache_invalidate_path(struct index_state *istate,
2746 const char *path)
2747 {
2748 if (!istate->untracked || !istate->untracked->root)
2749 return;
2750 invalidate_one_component(istate->untracked, istate->untracked->root,
2751 path, strlen(path));
2752 }
2753
2754 void untracked_cache_remove_from_index(struct index_state *istate,
2755 const char *path)
2756 {
2757 untracked_cache_invalidate_path(istate, path);
2758 }
2759
2760 void untracked_cache_add_to_index(struct index_state *istate,
2761 const char *path)
2762 {
2763 untracked_cache_invalidate_path(istate, path);
2764 }