7b25634832716ade46686d118184c4d43d8c11e7
[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
887 return fnmatch_icase_mem(pattern, patternlen,
888 name, namelen,
889 WM_PATHNAME) == 0;
890 }
891
892 /*
893 * Scan the given exclude list in reverse to see whether pathname
894 * should be ignored. The first match (i.e. the last on the list), if
895 * any, determines the fate. Returns the exclude_list element which
896 * matched, or NULL for undecided.
897 */
898 static struct exclude *last_exclude_matching_from_list(const char *pathname,
899 int pathlen,
900 const char *basename,
901 int *dtype,
902 struct exclude_list *el)
903 {
904 int i;
905
906 if (!el->nr)
907 return NULL; /* undefined */
908
909 for (i = el->nr - 1; 0 <= i; i--) {
910 struct exclude *x = el->excludes[i];
911 const char *exclude = x->pattern;
912 int prefix = x->nowildcardlen;
913
914 if (x->flags & EXC_FLAG_MUSTBEDIR) {
915 if (*dtype == DT_UNKNOWN)
916 *dtype = get_dtype(NULL, pathname, pathlen);
917 if (*dtype != DT_DIR)
918 continue;
919 }
920
921 if (x->flags & EXC_FLAG_NODIR) {
922 if (match_basename(basename,
923 pathlen - (basename - pathname),
924 exclude, prefix, x->patternlen,
925 x->flags))
926 return x;
927 continue;
928 }
929
930 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
931 if (match_pathname(pathname, pathlen,
932 x->base, x->baselen ? x->baselen - 1 : 0,
933 exclude, prefix, x->patternlen, x->flags))
934 return x;
935 }
936 return NULL; /* undecided */
937 }
938
939 /*
940 * Scan the list and let the last match determine the fate.
941 * Return 1 for exclude, 0 for include and -1 for undecided.
942 */
943 int is_excluded_from_list(const char *pathname,
944 int pathlen, const char *basename, int *dtype,
945 struct exclude_list *el)
946 {
947 struct exclude *exclude;
948 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
949 if (exclude)
950 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
951 return -1; /* undecided */
952 }
953
954 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
955 const char *pathname, int pathlen, const char *basename,
956 int *dtype_p)
957 {
958 int i, j;
959 struct exclude_list_group *group;
960 struct exclude *exclude;
961 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
962 group = &dir->exclude_list_group[i];
963 for (j = group->nr - 1; j >= 0; j--) {
964 exclude = last_exclude_matching_from_list(
965 pathname, pathlen, basename, dtype_p,
966 &group->el[j]);
967 if (exclude)
968 return exclude;
969 }
970 }
971 return NULL;
972 }
973
974 /*
975 * Loads the per-directory exclude list for the substring of base
976 * which has a char length of baselen.
977 */
978 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
979 {
980 struct exclude_list_group *group;
981 struct exclude_list *el;
982 struct exclude_stack *stk = NULL;
983 struct untracked_cache_dir *untracked;
984 int current;
985
986 group = &dir->exclude_list_group[EXC_DIRS];
987
988 /*
989 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
990 * which originate from directories not in the prefix of the
991 * path being checked.
992 */
993 while ((stk = dir->exclude_stack) != NULL) {
994 if (stk->baselen <= baselen &&
995 !strncmp(dir->basebuf.buf, base, stk->baselen))
996 break;
997 el = &group->el[dir->exclude_stack->exclude_ix];
998 dir->exclude_stack = stk->prev;
999 dir->exclude = NULL;
1000 free((char *)el->src); /* see strbuf_detach() below */
1001 clear_exclude_list(el);
1002 free(stk);
1003 group->nr--;
1004 }
1005
1006 /* Skip traversing into sub directories if the parent is excluded */
1007 if (dir->exclude)
1008 return;
1009
1010 /*
1011 * Lazy initialization. All call sites currently just
1012 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1013 * them seems lots of work for little benefit.
1014 */
1015 if (!dir->basebuf.buf)
1016 strbuf_init(&dir->basebuf, PATH_MAX);
1017
1018 /* Read from the parent directories and push them down. */
1019 current = stk ? stk->baselen : -1;
1020 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1021 if (dir->untracked)
1022 untracked = stk ? stk->ucd : dir->untracked->root;
1023 else
1024 untracked = NULL;
1025
1026 while (current < baselen) {
1027 const char *cp;
1028 struct sha1_stat sha1_stat;
1029
1030 stk = xcalloc(1, sizeof(*stk));
1031 if (current < 0) {
1032 cp = base;
1033 current = 0;
1034 } else {
1035 cp = strchr(base + current + 1, '/');
1036 if (!cp)
1037 die("oops in prep_exclude");
1038 cp++;
1039 untracked =
1040 lookup_untracked(dir->untracked, untracked,
1041 base + current,
1042 cp - base - current);
1043 }
1044 stk->prev = dir->exclude_stack;
1045 stk->baselen = cp - base;
1046 stk->exclude_ix = group->nr;
1047 stk->ucd = untracked;
1048 el = add_exclude_list(dir, EXC_DIRS, NULL);
1049 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1050 assert(stk->baselen == dir->basebuf.len);
1051
1052 /* Abort if the directory is excluded */
1053 if (stk->baselen) {
1054 int dt = DT_DIR;
1055 dir->basebuf.buf[stk->baselen - 1] = 0;
1056 dir->exclude = last_exclude_matching_from_lists(dir,
1057 dir->basebuf.buf, stk->baselen - 1,
1058 dir->basebuf.buf + current, &dt);
1059 dir->basebuf.buf[stk->baselen - 1] = '/';
1060 if (dir->exclude &&
1061 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1062 dir->exclude = NULL;
1063 if (dir->exclude) {
1064 dir->exclude_stack = stk;
1065 return;
1066 }
1067 }
1068
1069 /* Try to read per-directory file */
1070 hashclr(sha1_stat.sha1);
1071 sha1_stat.valid = 0;
1072 if (dir->exclude_per_dir &&
1073 /*
1074 * If we know that no files have been added in
1075 * this directory (i.e. valid_cached_dir() has
1076 * been executed and set untracked->valid) ..
1077 */
1078 (!untracked || !untracked->valid ||
1079 /*
1080 * .. and .gitignore does not exist before
1081 * (i.e. null exclude_sha1). Then we can skip
1082 * loading .gitignore, which would result in
1083 * ENOENT anyway.
1084 */
1085 !is_null_sha1(untracked->exclude_sha1))) {
1086 /*
1087 * dir->basebuf gets reused by the traversal, but we
1088 * need fname to remain unchanged to ensure the src
1089 * member of each struct exclude correctly
1090 * back-references its source file. Other invocations
1091 * of add_exclude_list provide stable strings, so we
1092 * strbuf_detach() and free() here in the caller.
1093 */
1094 struct strbuf sb = STRBUF_INIT;
1095 strbuf_addbuf(&sb, &dir->basebuf);
1096 strbuf_addstr(&sb, dir->exclude_per_dir);
1097 el->src = strbuf_detach(&sb, NULL);
1098 add_excludes(el->src, el->src, stk->baselen, el, 1,
1099 untracked ? &sha1_stat : NULL);
1100 }
1101 /*
1102 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1103 * will first be called in valid_cached_dir() then maybe many
1104 * times more in last_exclude_matching(). When the cache is
1105 * used, last_exclude_matching() will not be called and
1106 * reading .gitignore content will be a waste.
1107 *
1108 * So when it's called by valid_cached_dir() and we can get
1109 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1110 * modified on work tree), we could delay reading the
1111 * .gitignore content until we absolutely need it in
1112 * last_exclude_matching(). Be careful about ignore rule
1113 * order, though, if you do that.
1114 */
1115 if (untracked &&
1116 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1117 invalidate_gitignore(dir->untracked, untracked);
1118 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1119 }
1120 dir->exclude_stack = stk;
1121 current = stk->baselen;
1122 }
1123 strbuf_setlen(&dir->basebuf, baselen);
1124 }
1125
1126 /*
1127 * Loads the exclude lists for the directory containing pathname, then
1128 * scans all exclude lists to determine whether pathname is excluded.
1129 * Returns the exclude_list element which matched, or NULL for
1130 * undecided.
1131 */
1132 struct exclude *last_exclude_matching(struct dir_struct *dir,
1133 const char *pathname,
1134 int *dtype_p)
1135 {
1136 int pathlen = strlen(pathname);
1137 const char *basename = strrchr(pathname, '/');
1138 basename = (basename) ? basename+1 : pathname;
1139
1140 prep_exclude(dir, pathname, basename-pathname);
1141
1142 if (dir->exclude)
1143 return dir->exclude;
1144
1145 return last_exclude_matching_from_lists(dir, pathname, pathlen,
1146 basename, dtype_p);
1147 }
1148
1149 /*
1150 * Loads the exclude lists for the directory containing pathname, then
1151 * scans all exclude lists to determine whether pathname is excluded.
1152 * Returns 1 if true, otherwise 0.
1153 */
1154 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1155 {
1156 struct exclude *exclude =
1157 last_exclude_matching(dir, pathname, dtype_p);
1158 if (exclude)
1159 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1160 return 0;
1161 }
1162
1163 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1164 {
1165 struct dir_entry *ent;
1166
1167 ent = xmalloc(sizeof(*ent) + len + 1);
1168 ent->len = len;
1169 memcpy(ent->name, pathname, len);
1170 ent->name[len] = 0;
1171 return ent;
1172 }
1173
1174 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1175 {
1176 if (cache_file_exists(pathname, len, ignore_case))
1177 return NULL;
1178
1179 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1180 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1181 }
1182
1183 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1184 {
1185 if (!cache_name_is_other(pathname, len))
1186 return NULL;
1187
1188 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1189 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1190 }
1191
1192 enum exist_status {
1193 index_nonexistent = 0,
1194 index_directory,
1195 index_gitdir
1196 };
1197
1198 /*
1199 * Do not use the alphabetically sorted index to look up
1200 * the directory name; instead, use the case insensitive
1201 * directory hash.
1202 */
1203 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1204 {
1205 const struct cache_entry *ce = cache_dir_exists(dirname, len);
1206 unsigned char endchar;
1207
1208 if (!ce)
1209 return index_nonexistent;
1210 endchar = ce->name[len];
1211
1212 /*
1213 * The cache_entry structure returned will contain this dirname
1214 * and possibly additional path components.
1215 */
1216 if (endchar == '/')
1217 return index_directory;
1218
1219 /*
1220 * If there are no additional path components, then this cache_entry
1221 * represents a submodule. Submodules, despite being directories,
1222 * are stored in the cache without a closing slash.
1223 */
1224 if (!endchar && S_ISGITLINK(ce->ce_mode))
1225 return index_gitdir;
1226
1227 /* This should never be hit, but it exists just in case. */
1228 return index_nonexistent;
1229 }
1230
1231 /*
1232 * The index sorts alphabetically by entry name, which
1233 * means that a gitlink sorts as '\0' at the end, while
1234 * a directory (which is defined not as an entry, but as
1235 * the files it contains) will sort with the '/' at the
1236 * end.
1237 */
1238 static enum exist_status directory_exists_in_index(const char *dirname, int len)
1239 {
1240 int pos;
1241
1242 if (ignore_case)
1243 return directory_exists_in_index_icase(dirname, len);
1244
1245 pos = cache_name_pos(dirname, len);
1246 if (pos < 0)
1247 pos = -pos-1;
1248 while (pos < active_nr) {
1249 const struct cache_entry *ce = active_cache[pos++];
1250 unsigned char endchar;
1251
1252 if (strncmp(ce->name, dirname, len))
1253 break;
1254 endchar = ce->name[len];
1255 if (endchar > '/')
1256 break;
1257 if (endchar == '/')
1258 return index_directory;
1259 if (!endchar && S_ISGITLINK(ce->ce_mode))
1260 return index_gitdir;
1261 }
1262 return index_nonexistent;
1263 }
1264
1265 /*
1266 * When we find a directory when traversing the filesystem, we
1267 * have three distinct cases:
1268 *
1269 * - ignore it
1270 * - see it as a directory
1271 * - recurse into it
1272 *
1273 * and which one we choose depends on a combination of existing
1274 * git index contents and the flags passed into the directory
1275 * traversal routine.
1276 *
1277 * Case 1: If we *already* have entries in the index under that
1278 * directory name, we always recurse into the directory to see
1279 * all the files.
1280 *
1281 * Case 2: If we *already* have that directory name as a gitlink,
1282 * we always continue to see it as a gitlink, regardless of whether
1283 * there is an actual git directory there or not (it might not
1284 * be checked out as a subproject!)
1285 *
1286 * Case 3: if we didn't have it in the index previously, we
1287 * have a few sub-cases:
1288 *
1289 * (a) if "show_other_directories" is true, we show it as
1290 * just a directory, unless "hide_empty_directories" is
1291 * also true, in which case we need to check if it contains any
1292 * untracked and / or ignored files.
1293 * (b) if it looks like a git directory, and we don't have
1294 * 'no_gitlinks' set we treat it as a gitlink, and show it
1295 * as a directory.
1296 * (c) otherwise, we recurse into it.
1297 */
1298 static enum path_treatment treat_directory(struct dir_struct *dir,
1299 struct untracked_cache_dir *untracked,
1300 const char *dirname, int len, int baselen, int exclude,
1301 const struct path_simplify *simplify)
1302 {
1303 /* The "len-1" is to strip the final '/' */
1304 switch (directory_exists_in_index(dirname, len-1)) {
1305 case index_directory:
1306 return path_recurse;
1307
1308 case index_gitdir:
1309 return path_none;
1310
1311 case index_nonexistent:
1312 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1313 break;
1314 if (!(dir->flags & DIR_NO_GITLINKS)) {
1315 unsigned char sha1[20];
1316 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1317 return path_untracked;
1318 }
1319 return path_recurse;
1320 }
1321
1322 /* This is the "show_other_directories" case */
1323
1324 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1325 return exclude ? path_excluded : path_untracked;
1326
1327 untracked = lookup_untracked(dir->untracked, untracked,
1328 dirname + baselen, len - baselen);
1329 return read_directory_recursive(dir, dirname, len,
1330 untracked, 1, simplify);
1331 }
1332
1333 /*
1334 * This is an inexact early pruning of any recursive directory
1335 * reading - if the path cannot possibly be in the pathspec,
1336 * return true, and we'll skip it early.
1337 */
1338 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1339 {
1340 if (simplify) {
1341 for (;;) {
1342 const char *match = simplify->path;
1343 int len = simplify->len;
1344
1345 if (!match)
1346 break;
1347 if (len > pathlen)
1348 len = pathlen;
1349 if (!memcmp(path, match, len))
1350 return 0;
1351 simplify++;
1352 }
1353 return 1;
1354 }
1355 return 0;
1356 }
1357
1358 /*
1359 * This function tells us whether an excluded path matches a
1360 * list of "interesting" pathspecs. That is, whether a path matched
1361 * by any of the pathspecs could possibly be ignored by excluding
1362 * the specified path. This can happen if:
1363 *
1364 * 1. the path is mentioned explicitly in the pathspec
1365 *
1366 * 2. the path is a directory prefix of some element in the
1367 * pathspec
1368 */
1369 static int exclude_matches_pathspec(const char *path, int len,
1370 const struct path_simplify *simplify)
1371 {
1372 if (simplify) {
1373 for (; simplify->path; simplify++) {
1374 if (len == simplify->len
1375 && !memcmp(path, simplify->path, len))
1376 return 1;
1377 if (len < simplify->len
1378 && simplify->path[len] == '/'
1379 && !memcmp(path, simplify->path, len))
1380 return 1;
1381 }
1382 }
1383 return 0;
1384 }
1385
1386 static int get_index_dtype(const char *path, int len)
1387 {
1388 int pos;
1389 const struct cache_entry *ce;
1390
1391 ce = cache_file_exists(path, len, 0);
1392 if (ce) {
1393 if (!ce_uptodate(ce))
1394 return DT_UNKNOWN;
1395 if (S_ISGITLINK(ce->ce_mode))
1396 return DT_DIR;
1397 /*
1398 * Nobody actually cares about the
1399 * difference between DT_LNK and DT_REG
1400 */
1401 return DT_REG;
1402 }
1403
1404 /* Try to look it up as a directory */
1405 pos = cache_name_pos(path, len);
1406 if (pos >= 0)
1407 return DT_UNKNOWN;
1408 pos = -pos-1;
1409 while (pos < active_nr) {
1410 ce = active_cache[pos++];
1411 if (strncmp(ce->name, path, len))
1412 break;
1413 if (ce->name[len] > '/')
1414 break;
1415 if (ce->name[len] < '/')
1416 continue;
1417 if (!ce_uptodate(ce))
1418 break; /* continue? */
1419 return DT_DIR;
1420 }
1421 return DT_UNKNOWN;
1422 }
1423
1424 static int get_dtype(struct dirent *de, const char *path, int len)
1425 {
1426 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1427 struct stat st;
1428
1429 if (dtype != DT_UNKNOWN)
1430 return dtype;
1431 dtype = get_index_dtype(path, len);
1432 if (dtype != DT_UNKNOWN)
1433 return dtype;
1434 if (lstat(path, &st))
1435 return dtype;
1436 if (S_ISREG(st.st_mode))
1437 return DT_REG;
1438 if (S_ISDIR(st.st_mode))
1439 return DT_DIR;
1440 if (S_ISLNK(st.st_mode))
1441 return DT_LNK;
1442 return dtype;
1443 }
1444
1445 static enum path_treatment treat_one_path(struct dir_struct *dir,
1446 struct untracked_cache_dir *untracked,
1447 struct strbuf *path,
1448 int baselen,
1449 const struct path_simplify *simplify,
1450 int dtype, struct dirent *de)
1451 {
1452 int exclude;
1453 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1454
1455 if (dtype == DT_UNKNOWN)
1456 dtype = get_dtype(de, path->buf, path->len);
1457
1458 /* Always exclude indexed files */
1459 if (dtype != DT_DIR && has_path_in_index)
1460 return path_none;
1461
1462 /*
1463 * When we are looking at a directory P in the working tree,
1464 * there are three cases:
1465 *
1466 * (1) P exists in the index. Everything inside the directory P in
1467 * the working tree needs to go when P is checked out from the
1468 * index.
1469 *
1470 * (2) P does not exist in the index, but there is P/Q in the index.
1471 * We know P will stay a directory when we check out the contents
1472 * of the index, but we do not know yet if there is a directory
1473 * P/Q in the working tree to be killed, so we need to recurse.
1474 *
1475 * (3) P does not exist in the index, and there is no P/Q in the index
1476 * to require P to be a directory, either. Only in this case, we
1477 * know that everything inside P will not be killed without
1478 * recursing.
1479 */
1480 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1481 (dtype == DT_DIR) &&
1482 !has_path_in_index &&
1483 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1484 return path_none;
1485
1486 exclude = is_excluded(dir, path->buf, &dtype);
1487
1488 /*
1489 * Excluded? If we don't explicitly want to show
1490 * ignored files, ignore it
1491 */
1492 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1493 return path_excluded;
1494
1495 switch (dtype) {
1496 default:
1497 return path_none;
1498 case DT_DIR:
1499 strbuf_addch(path, '/');
1500 return treat_directory(dir, untracked, path->buf, path->len,
1501 baselen, exclude, simplify);
1502 case DT_REG:
1503 case DT_LNK:
1504 return exclude ? path_excluded : path_untracked;
1505 }
1506 }
1507
1508 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1509 struct untracked_cache_dir *untracked,
1510 struct cached_dir *cdir,
1511 struct strbuf *path,
1512 int baselen,
1513 const struct path_simplify *simplify)
1514 {
1515 strbuf_setlen(path, baselen);
1516 if (!cdir->ucd) {
1517 strbuf_addstr(path, cdir->file);
1518 return path_untracked;
1519 }
1520 strbuf_addstr(path, cdir->ucd->name);
1521 /* treat_one_path() does this before it calls treat_directory() */
1522 if (path->buf[path->len - 1] != '/')
1523 strbuf_addch(path, '/');
1524 if (cdir->ucd->check_only)
1525 /*
1526 * check_only is set as a result of treat_directory() getting
1527 * to its bottom. Verify again the same set of directories
1528 * with check_only set.
1529 */
1530 return read_directory_recursive(dir, path->buf, path->len,
1531 cdir->ucd, 1, simplify);
1532 /*
1533 * We get path_recurse in the first run when
1534 * directory_exists_in_index() returns index_nonexistent. We
1535 * are sure that new changes in the index does not impact the
1536 * outcome. Return now.
1537 */
1538 return path_recurse;
1539 }
1540
1541 static enum path_treatment treat_path(struct dir_struct *dir,
1542 struct untracked_cache_dir *untracked,
1543 struct cached_dir *cdir,
1544 struct strbuf *path,
1545 int baselen,
1546 const struct path_simplify *simplify)
1547 {
1548 int dtype;
1549 struct dirent *de = cdir->de;
1550
1551 if (!de)
1552 return treat_path_fast(dir, untracked, cdir, path,
1553 baselen, simplify);
1554 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1555 return path_none;
1556 strbuf_setlen(path, baselen);
1557 strbuf_addstr(path, de->d_name);
1558 if (simplify_away(path->buf, path->len, simplify))
1559 return path_none;
1560
1561 dtype = DTYPE(de);
1562 return treat_one_path(dir, untracked, path, baselen, simplify, dtype, de);
1563 }
1564
1565 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1566 {
1567 if (!dir)
1568 return;
1569 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1570 dir->untracked_alloc);
1571 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1572 }
1573
1574 static int valid_cached_dir(struct dir_struct *dir,
1575 struct untracked_cache_dir *untracked,
1576 struct strbuf *path,
1577 int check_only)
1578 {
1579 struct stat st;
1580
1581 if (!untracked)
1582 return 0;
1583
1584 if (stat(path->len ? path->buf : ".", &st)) {
1585 invalidate_directory(dir->untracked, untracked);
1586 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1587 return 0;
1588 }
1589 if (!untracked->valid ||
1590 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
1591 if (untracked->valid)
1592 invalidate_directory(dir->untracked, untracked);
1593 fill_stat_data(&untracked->stat_data, &st);
1594 return 0;
1595 }
1596
1597 if (untracked->check_only != !!check_only) {
1598 invalidate_directory(dir->untracked, untracked);
1599 return 0;
1600 }
1601
1602 /*
1603 * prep_exclude will be called eventually on this directory,
1604 * but it's called much later in last_exclude_matching(). We
1605 * need it now to determine the validity of the cache for this
1606 * path. The next calls will be nearly no-op, the way
1607 * prep_exclude() is designed.
1608 */
1609 if (path->len && path->buf[path->len - 1] != '/') {
1610 strbuf_addch(path, '/');
1611 prep_exclude(dir, path->buf, path->len);
1612 strbuf_setlen(path, path->len - 1);
1613 } else
1614 prep_exclude(dir, path->buf, path->len);
1615
1616 /* hopefully prep_exclude() haven't invalidated this entry... */
1617 return untracked->valid;
1618 }
1619
1620 static int open_cached_dir(struct cached_dir *cdir,
1621 struct dir_struct *dir,
1622 struct untracked_cache_dir *untracked,
1623 struct strbuf *path,
1624 int check_only)
1625 {
1626 memset(cdir, 0, sizeof(*cdir));
1627 cdir->untracked = untracked;
1628 if (valid_cached_dir(dir, untracked, path, check_only))
1629 return 0;
1630 cdir->fdir = opendir(path->len ? path->buf : ".");
1631 if (dir->untracked)
1632 dir->untracked->dir_opened++;
1633 if (!cdir->fdir)
1634 return -1;
1635 return 0;
1636 }
1637
1638 static int read_cached_dir(struct cached_dir *cdir)
1639 {
1640 if (cdir->fdir) {
1641 cdir->de = readdir(cdir->fdir);
1642 if (!cdir->de)
1643 return -1;
1644 return 0;
1645 }
1646 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1647 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1648 if (!d->recurse) {
1649 cdir->nr_dirs++;
1650 continue;
1651 }
1652 cdir->ucd = d;
1653 cdir->nr_dirs++;
1654 return 0;
1655 }
1656 cdir->ucd = NULL;
1657 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1658 struct untracked_cache_dir *d = cdir->untracked;
1659 cdir->file = d->untracked[cdir->nr_files++];
1660 return 0;
1661 }
1662 return -1;
1663 }
1664
1665 static void close_cached_dir(struct cached_dir *cdir)
1666 {
1667 if (cdir->fdir)
1668 closedir(cdir->fdir);
1669 /*
1670 * We have gone through this directory and found no untracked
1671 * entries. Mark it valid.
1672 */
1673 if (cdir->untracked) {
1674 cdir->untracked->valid = 1;
1675 cdir->untracked->recurse = 1;
1676 }
1677 }
1678
1679 /*
1680 * Read a directory tree. We currently ignore anything but
1681 * directories, regular files and symlinks. That's because git
1682 * doesn't handle them at all yet. Maybe that will change some
1683 * day.
1684 *
1685 * Also, we ignore the name ".git" (even if it is not a directory).
1686 * That likely will not change.
1687 *
1688 * Returns the most significant path_treatment value encountered in the scan.
1689 */
1690 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1691 const char *base, int baselen,
1692 struct untracked_cache_dir *untracked, int check_only,
1693 const struct path_simplify *simplify)
1694 {
1695 struct cached_dir cdir;
1696 enum path_treatment state, subdir_state, dir_state = path_none;
1697 struct strbuf path = STRBUF_INIT;
1698
1699 strbuf_add(&path, base, baselen);
1700
1701 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1702 goto out;
1703
1704 if (untracked)
1705 untracked->check_only = !!check_only;
1706
1707 while (!read_cached_dir(&cdir)) {
1708 /* check how the file or directory should be treated */
1709 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);
1710
1711 if (state > dir_state)
1712 dir_state = state;
1713
1714 /* recurse into subdir if instructed by treat_path */
1715 if (state == path_recurse) {
1716 struct untracked_cache_dir *ud;
1717 ud = lookup_untracked(dir->untracked, untracked,
1718 path.buf + baselen,
1719 path.len - baselen);
1720 subdir_state =
1721 read_directory_recursive(dir, path.buf, path.len,
1722 ud, check_only, simplify);
1723 if (subdir_state > dir_state)
1724 dir_state = subdir_state;
1725 }
1726
1727 if (check_only) {
1728 /* abort early if maximum state has been reached */
1729 if (dir_state == path_untracked) {
1730 if (cdir.fdir)
1731 add_untracked(untracked, path.buf + baselen);
1732 break;
1733 }
1734 /* skip the dir_add_* part */
1735 continue;
1736 }
1737
1738 /* add the path to the appropriate result list */
1739 switch (state) {
1740 case path_excluded:
1741 if (dir->flags & DIR_SHOW_IGNORED)
1742 dir_add_name(dir, path.buf, path.len);
1743 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1744 ((dir->flags & DIR_COLLECT_IGNORED) &&
1745 exclude_matches_pathspec(path.buf, path.len,
1746 simplify)))
1747 dir_add_ignored(dir, path.buf, path.len);
1748 break;
1749
1750 case path_untracked:
1751 if (dir->flags & DIR_SHOW_IGNORED)
1752 break;
1753 dir_add_name(dir, path.buf, path.len);
1754 if (cdir.fdir)
1755 add_untracked(untracked, path.buf + baselen);
1756 break;
1757
1758 default:
1759 break;
1760 }
1761 }
1762 close_cached_dir(&cdir);
1763 out:
1764 strbuf_release(&path);
1765
1766 return dir_state;
1767 }
1768
1769 static int cmp_name(const void *p1, const void *p2)
1770 {
1771 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1772 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1773
1774 return name_compare(e1->name, e1->len, e2->name, e2->len);
1775 }
1776
1777 static struct path_simplify *create_simplify(const char **pathspec)
1778 {
1779 int nr, alloc = 0;
1780 struct path_simplify *simplify = NULL;
1781
1782 if (!pathspec)
1783 return NULL;
1784
1785 for (nr = 0 ; ; nr++) {
1786 const char *match;
1787 ALLOC_GROW(simplify, nr + 1, alloc);
1788 match = *pathspec++;
1789 if (!match)
1790 break;
1791 simplify[nr].path = match;
1792 simplify[nr].len = simple_length(match);
1793 }
1794 simplify[nr].path = NULL;
1795 simplify[nr].len = 0;
1796 return simplify;
1797 }
1798
1799 static void free_simplify(struct path_simplify *simplify)
1800 {
1801 free(simplify);
1802 }
1803
1804 static int treat_leading_path(struct dir_struct *dir,
1805 const char *path, int len,
1806 const struct path_simplify *simplify)
1807 {
1808 struct strbuf sb = STRBUF_INIT;
1809 int baselen, rc = 0;
1810 const char *cp;
1811 int old_flags = dir->flags;
1812
1813 while (len && path[len - 1] == '/')
1814 len--;
1815 if (!len)
1816 return 1;
1817 baselen = 0;
1818 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1819 while (1) {
1820 cp = path + baselen + !!baselen;
1821 cp = memchr(cp, '/', path + len - cp);
1822 if (!cp)
1823 baselen = len;
1824 else
1825 baselen = cp - path;
1826 strbuf_setlen(&sb, 0);
1827 strbuf_add(&sb, path, baselen);
1828 if (!is_directory(sb.buf))
1829 break;
1830 if (simplify_away(sb.buf, sb.len, simplify))
1831 break;
1832 if (treat_one_path(dir, NULL, &sb, baselen, simplify,
1833 DT_DIR, NULL) == path_none)
1834 break; /* do not recurse into it */
1835 if (len <= baselen) {
1836 rc = 1;
1837 break; /* finished checking */
1838 }
1839 }
1840 strbuf_release(&sb);
1841 dir->flags = old_flags;
1842 return rc;
1843 }
1844
1845 static const char *get_ident_string(void)
1846 {
1847 static struct strbuf sb = STRBUF_INIT;
1848 struct utsname uts;
1849
1850 if (sb.len)
1851 return sb.buf;
1852 if (uname(&uts) < 0)
1853 die_errno(_("failed to get kernel name and information"));
1854 strbuf_addf(&sb, "Location %s, system %s %s %s", get_git_work_tree(),
1855 uts.sysname, uts.release, uts.version);
1856 return sb.buf;
1857 }
1858
1859 static int ident_in_untracked(const struct untracked_cache *uc)
1860 {
1861 const char *end = uc->ident.buf + uc->ident.len;
1862 const char *p = uc->ident.buf;
1863
1864 for (p = uc->ident.buf; p < end; p += strlen(p) + 1)
1865 if (!strcmp(p, get_ident_string()))
1866 return 1;
1867 return 0;
1868 }
1869
1870 void add_untracked_ident(struct untracked_cache *uc)
1871 {
1872 if (ident_in_untracked(uc))
1873 return;
1874 strbuf_addstr(&uc->ident, get_ident_string());
1875 /* this strbuf contains a list of strings, save NUL too */
1876 strbuf_addch(&uc->ident, 0);
1877 }
1878
1879 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1880 int base_len,
1881 const struct pathspec *pathspec)
1882 {
1883 struct untracked_cache_dir *root;
1884
1885 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
1886 return NULL;
1887
1888 /*
1889 * We only support $GIT_DIR/info/exclude and core.excludesfile
1890 * as the global ignore rule files. Any other additions
1891 * (e.g. from command line) invalidate the cache. This
1892 * condition also catches running setup_standard_excludes()
1893 * before setting dir->untracked!
1894 */
1895 if (dir->unmanaged_exclude_files)
1896 return NULL;
1897
1898 /*
1899 * Optimize for the main use case only: whole-tree git
1900 * status. More work involved in treat_leading_path() if we
1901 * use cache on just a subset of the worktree. pathspec
1902 * support could make the matter even worse.
1903 */
1904 if (base_len || (pathspec && pathspec->nr))
1905 return NULL;
1906
1907 /* Different set of flags may produce different results */
1908 if (dir->flags != dir->untracked->dir_flags ||
1909 /*
1910 * See treat_directory(), case index_nonexistent. Without
1911 * this flag, we may need to also cache .git file content
1912 * for the resolve_gitlink_ref() call, which we don't.
1913 */
1914 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1915 /* We don't support collecting ignore files */
1916 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1917 DIR_COLLECT_IGNORED)))
1918 return NULL;
1919
1920 /*
1921 * If we use .gitignore in the cache and now you change it to
1922 * .gitexclude, everything will go wrong.
1923 */
1924 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
1925 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
1926 return NULL;
1927
1928 /*
1929 * EXC_CMDL is not considered in the cache. If people set it,
1930 * skip the cache.
1931 */
1932 if (dir->exclude_list_group[EXC_CMDL].nr)
1933 return NULL;
1934
1935 if (!ident_in_untracked(dir->untracked)) {
1936 warning(_("Untracked cache is disabled on this system."));
1937 return NULL;
1938 }
1939
1940 if (!dir->untracked->root) {
1941 const int len = sizeof(*dir->untracked->root);
1942 dir->untracked->root = xmalloc(len);
1943 memset(dir->untracked->root, 0, len);
1944 }
1945
1946 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
1947 root = dir->untracked->root;
1948 if (hashcmp(dir->ss_info_exclude.sha1,
1949 dir->untracked->ss_info_exclude.sha1)) {
1950 invalidate_gitignore(dir->untracked, root);
1951 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
1952 }
1953 if (hashcmp(dir->ss_excludes_file.sha1,
1954 dir->untracked->ss_excludes_file.sha1)) {
1955 invalidate_gitignore(dir->untracked, root);
1956 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
1957 }
1958
1959 /* Make sure this directory is not dropped out at saving phase */
1960 root->recurse = 1;
1961 return root;
1962 }
1963
1964 int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1965 {
1966 struct path_simplify *simplify;
1967 struct untracked_cache_dir *untracked;
1968
1969 /*
1970 * Check out create_simplify()
1971 */
1972 if (pathspec)
1973 GUARD_PATHSPEC(pathspec,
1974 PATHSPEC_FROMTOP |
1975 PATHSPEC_MAXDEPTH |
1976 PATHSPEC_LITERAL |
1977 PATHSPEC_GLOB |
1978 PATHSPEC_ICASE |
1979 PATHSPEC_EXCLUDE);
1980
1981 if (has_symlink_leading_path(path, len))
1982 return dir->nr;
1983
1984 /*
1985 * exclude patterns are treated like positive ones in
1986 * create_simplify. Usually exclude patterns should be a
1987 * subset of positive ones, which has no impacts on
1988 * create_simplify().
1989 */
1990 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
1991 untracked = validate_untracked_cache(dir, len, pathspec);
1992 if (!untracked)
1993 /*
1994 * make sure untracked cache code path is disabled,
1995 * e.g. prep_exclude()
1996 */
1997 dir->untracked = NULL;
1998 if (!len || treat_leading_path(dir, path, len, simplify))
1999 read_directory_recursive(dir, path, len, untracked, 0, simplify);
2000 free_simplify(simplify);
2001 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
2002 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
2003 if (dir->untracked) {
2004 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2005 trace_printf_key(&trace_untracked_stats,
2006 "node creation: %u\n"
2007 "gitignore invalidation: %u\n"
2008 "directory invalidation: %u\n"
2009 "opendir: %u\n",
2010 dir->untracked->dir_created,
2011 dir->untracked->gitignore_invalidated,
2012 dir->untracked->dir_invalidated,
2013 dir->untracked->dir_opened);
2014 if (dir->untracked == the_index.untracked &&
2015 (dir->untracked->dir_opened ||
2016 dir->untracked->gitignore_invalidated ||
2017 dir->untracked->dir_invalidated))
2018 the_index.cache_changed |= UNTRACKED_CHANGED;
2019 if (dir->untracked != the_index.untracked) {
2020 free(dir->untracked);
2021 dir->untracked = NULL;
2022 }
2023 }
2024 return dir->nr;
2025 }
2026
2027 int file_exists(const char *f)
2028 {
2029 struct stat sb;
2030 return lstat(f, &sb) == 0;
2031 }
2032
2033 /*
2034 * Given two normalized paths (a trailing slash is ok), if subdir is
2035 * outside dir, return -1. Otherwise return the offset in subdir that
2036 * can be used as relative path to dir.
2037 */
2038 int dir_inside_of(const char *subdir, const char *dir)
2039 {
2040 int offset = 0;
2041
2042 assert(dir && subdir && *dir && *subdir);
2043
2044 while (*dir && *subdir && *dir == *subdir) {
2045 dir++;
2046 subdir++;
2047 offset++;
2048 }
2049
2050 /* hel[p]/me vs hel[l]/yeah */
2051 if (*dir && *subdir)
2052 return -1;
2053
2054 if (!*subdir)
2055 return !*dir ? offset : -1; /* same dir */
2056
2057 /* foo/[b]ar vs foo/[] */
2058 if (is_dir_sep(dir[-1]))
2059 return is_dir_sep(subdir[-1]) ? offset : -1;
2060
2061 /* foo[/]bar vs foo[] */
2062 return is_dir_sep(*subdir) ? offset + 1 : -1;
2063 }
2064
2065 int is_inside_dir(const char *dir)
2066 {
2067 char *cwd;
2068 int rc;
2069
2070 if (!dir)
2071 return 0;
2072
2073 cwd = xgetcwd();
2074 rc = (dir_inside_of(cwd, dir) >= 0);
2075 free(cwd);
2076 return rc;
2077 }
2078
2079 int is_empty_dir(const char *path)
2080 {
2081 DIR *dir = opendir(path);
2082 struct dirent *e;
2083 int ret = 1;
2084
2085 if (!dir)
2086 return 0;
2087
2088 while ((e = readdir(dir)) != NULL)
2089 if (!is_dot_or_dotdot(e->d_name)) {
2090 ret = 0;
2091 break;
2092 }
2093
2094 closedir(dir);
2095 return ret;
2096 }
2097
2098 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2099 {
2100 DIR *dir;
2101 struct dirent *e;
2102 int ret = 0, original_len = path->len, len, kept_down = 0;
2103 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2104 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2105 unsigned char submodule_head[20];
2106
2107 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2108 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2109 /* Do not descend and nuke a nested git work tree. */
2110 if (kept_up)
2111 *kept_up = 1;
2112 return 0;
2113 }
2114
2115 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2116 dir = opendir(path->buf);
2117 if (!dir) {
2118 if (errno == ENOENT)
2119 return keep_toplevel ? -1 : 0;
2120 else if (errno == EACCES && !keep_toplevel)
2121 /*
2122 * An empty dir could be removable even if it
2123 * is unreadable:
2124 */
2125 return rmdir(path->buf);
2126 else
2127 return -1;
2128 }
2129 if (path->buf[original_len - 1] != '/')
2130 strbuf_addch(path, '/');
2131
2132 len = path->len;
2133 while ((e = readdir(dir)) != NULL) {
2134 struct stat st;
2135 if (is_dot_or_dotdot(e->d_name))
2136 continue;
2137
2138 strbuf_setlen(path, len);
2139 strbuf_addstr(path, e->d_name);
2140 if (lstat(path->buf, &st)) {
2141 if (errno == ENOENT)
2142 /*
2143 * file disappeared, which is what we
2144 * wanted anyway
2145 */
2146 continue;
2147 /* fall thru */
2148 } else if (S_ISDIR(st.st_mode)) {
2149 if (!remove_dir_recurse(path, flag, &kept_down))
2150 continue; /* happy */
2151 } else if (!only_empty &&
2152 (!unlink(path->buf) || errno == ENOENT)) {
2153 continue; /* happy, too */
2154 }
2155
2156 /* path too long, stat fails, or non-directory still exists */
2157 ret = -1;
2158 break;
2159 }
2160 closedir(dir);
2161
2162 strbuf_setlen(path, original_len);
2163 if (!ret && !keep_toplevel && !kept_down)
2164 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2165 else if (kept_up)
2166 /*
2167 * report the uplevel that it is not an error that we
2168 * did not rmdir() our directory.
2169 */
2170 *kept_up = !ret;
2171 return ret;
2172 }
2173
2174 int remove_dir_recursively(struct strbuf *path, int flag)
2175 {
2176 return remove_dir_recurse(path, flag, NULL);
2177 }
2178
2179 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2180
2181 void setup_standard_excludes(struct dir_struct *dir)
2182 {
2183 const char *path;
2184
2185 dir->exclude_per_dir = ".gitignore";
2186
2187 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2188 if (!excludes_file)
2189 excludes_file = xdg_config_home("ignore");
2190 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2191 add_excludes_from_file_1(dir, excludes_file,
2192 dir->untracked ? &dir->ss_excludes_file : NULL);
2193
2194 /* per repository user preference */
2195 path = git_path_info_exclude();
2196 if (!access_or_warn(path, R_OK, 0))
2197 add_excludes_from_file_1(dir, path,
2198 dir->untracked ? &dir->ss_info_exclude : NULL);
2199 }
2200
2201 int remove_path(const char *name)
2202 {
2203 char *slash;
2204
2205 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
2206 return -1;
2207
2208 slash = strrchr(name, '/');
2209 if (slash) {
2210 char *dirs = xstrdup(name);
2211 slash = dirs + (slash - name);
2212 do {
2213 *slash = '\0';
2214 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2215 free(dirs);
2216 }
2217 return 0;
2218 }
2219
2220 /*
2221 * Frees memory within dir which was allocated for exclude lists and
2222 * the exclude_stack. Does not free dir itself.
2223 */
2224 void clear_directory(struct dir_struct *dir)
2225 {
2226 int i, j;
2227 struct exclude_list_group *group;
2228 struct exclude_list *el;
2229 struct exclude_stack *stk;
2230
2231 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2232 group = &dir->exclude_list_group[i];
2233 for (j = 0; j < group->nr; j++) {
2234 el = &group->el[j];
2235 if (i == EXC_DIRS)
2236 free((char *)el->src);
2237 clear_exclude_list(el);
2238 }
2239 free(group->el);
2240 }
2241
2242 stk = dir->exclude_stack;
2243 while (stk) {
2244 struct exclude_stack *prev = stk->prev;
2245 free(stk);
2246 stk = prev;
2247 }
2248 strbuf_release(&dir->basebuf);
2249 }
2250
2251 struct ondisk_untracked_cache {
2252 struct stat_data info_exclude_stat;
2253 struct stat_data excludes_file_stat;
2254 uint32_t dir_flags;
2255 unsigned char info_exclude_sha1[20];
2256 unsigned char excludes_file_sha1[20];
2257 char exclude_per_dir[FLEX_ARRAY];
2258 };
2259
2260 #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2261
2262 struct write_data {
2263 int index; /* number of written untracked_cache_dir */
2264 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2265 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2266 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2267 struct strbuf out;
2268 struct strbuf sb_stat;
2269 struct strbuf sb_sha1;
2270 };
2271
2272 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2273 {
2274 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2275 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2276 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2277 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2278 to->sd_dev = htonl(from->sd_dev);
2279 to->sd_ino = htonl(from->sd_ino);
2280 to->sd_uid = htonl(from->sd_uid);
2281 to->sd_gid = htonl(from->sd_gid);
2282 to->sd_size = htonl(from->sd_size);
2283 }
2284
2285 static void write_one_dir(struct untracked_cache_dir *untracked,
2286 struct write_data *wd)
2287 {
2288 struct stat_data stat_data;
2289 struct strbuf *out = &wd->out;
2290 unsigned char intbuf[16];
2291 unsigned int intlen, value;
2292 int i = wd->index++;
2293
2294 /*
2295 * untracked_nr should be reset whenever valid is clear, but
2296 * for safety..
2297 */
2298 if (!untracked->valid) {
2299 untracked->untracked_nr = 0;
2300 untracked->check_only = 0;
2301 }
2302
2303 if (untracked->check_only)
2304 ewah_set(wd->check_only, i);
2305 if (untracked->valid) {
2306 ewah_set(wd->valid, i);
2307 stat_data_to_disk(&stat_data, &untracked->stat_data);
2308 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2309 }
2310 if (!is_null_sha1(untracked->exclude_sha1)) {
2311 ewah_set(wd->sha1_valid, i);
2312 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2313 }
2314
2315 intlen = encode_varint(untracked->untracked_nr, intbuf);
2316 strbuf_add(out, intbuf, intlen);
2317
2318 /* skip non-recurse directories */
2319 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2320 if (untracked->dirs[i]->recurse)
2321 value++;
2322 intlen = encode_varint(value, intbuf);
2323 strbuf_add(out, intbuf, intlen);
2324
2325 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2326
2327 for (i = 0; i < untracked->untracked_nr; i++)
2328 strbuf_add(out, untracked->untracked[i],
2329 strlen(untracked->untracked[i]) + 1);
2330
2331 for (i = 0; i < untracked->dirs_nr; i++)
2332 if (untracked->dirs[i]->recurse)
2333 write_one_dir(untracked->dirs[i], wd);
2334 }
2335
2336 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2337 {
2338 struct ondisk_untracked_cache *ouc;
2339 struct write_data wd;
2340 unsigned char varbuf[16];
2341 int len = 0, varint_len;
2342 if (untracked->exclude_per_dir)
2343 len = strlen(untracked->exclude_per_dir);
2344 ouc = xmalloc(sizeof(*ouc) + len + 1);
2345 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2346 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2347 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2348 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2349 ouc->dir_flags = htonl(untracked->dir_flags);
2350 memcpy(ouc->exclude_per_dir, untracked->exclude_per_dir, len + 1);
2351
2352 varint_len = encode_varint(untracked->ident.len, varbuf);
2353 strbuf_add(out, varbuf, varint_len);
2354 strbuf_add(out, untracked->ident.buf, untracked->ident.len);
2355
2356 strbuf_add(out, ouc, ouc_size(len));
2357 free(ouc);
2358 ouc = NULL;
2359
2360 if (!untracked->root) {
2361 varint_len = encode_varint(0, varbuf);
2362 strbuf_add(out, varbuf, varint_len);
2363 return;
2364 }
2365
2366 wd.index = 0;
2367 wd.check_only = ewah_new();
2368 wd.valid = ewah_new();
2369 wd.sha1_valid = ewah_new();
2370 strbuf_init(&wd.out, 1024);
2371 strbuf_init(&wd.sb_stat, 1024);
2372 strbuf_init(&wd.sb_sha1, 1024);
2373 write_one_dir(untracked->root, &wd);
2374
2375 varint_len = encode_varint(wd.index, varbuf);
2376 strbuf_add(out, varbuf, varint_len);
2377 strbuf_addbuf(out, &wd.out);
2378 ewah_serialize_strbuf(wd.valid, out);
2379 ewah_serialize_strbuf(wd.check_only, out);
2380 ewah_serialize_strbuf(wd.sha1_valid, out);
2381 strbuf_addbuf(out, &wd.sb_stat);
2382 strbuf_addbuf(out, &wd.sb_sha1);
2383 strbuf_addch(out, '\0'); /* safe guard for string lists */
2384
2385 ewah_free(wd.valid);
2386 ewah_free(wd.check_only);
2387 ewah_free(wd.sha1_valid);
2388 strbuf_release(&wd.out);
2389 strbuf_release(&wd.sb_stat);
2390 strbuf_release(&wd.sb_sha1);
2391 }
2392
2393 static void free_untracked(struct untracked_cache_dir *ucd)
2394 {
2395 int i;
2396 if (!ucd)
2397 return;
2398 for (i = 0; i < ucd->dirs_nr; i++)
2399 free_untracked(ucd->dirs[i]);
2400 for (i = 0; i < ucd->untracked_nr; i++)
2401 free(ucd->untracked[i]);
2402 free(ucd->untracked);
2403 free(ucd->dirs);
2404 free(ucd);
2405 }
2406
2407 void free_untracked_cache(struct untracked_cache *uc)
2408 {
2409 if (uc)
2410 free_untracked(uc->root);
2411 free(uc);
2412 }
2413
2414 struct read_data {
2415 int index;
2416 struct untracked_cache_dir **ucd;
2417 struct ewah_bitmap *check_only;
2418 struct ewah_bitmap *valid;
2419 struct ewah_bitmap *sha1_valid;
2420 const unsigned char *data;
2421 const unsigned char *end;
2422 };
2423
2424 static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2425 {
2426 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2427 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2428 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2429 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2430 to->sd_dev = get_be32(&from->sd_dev);
2431 to->sd_ino = get_be32(&from->sd_ino);
2432 to->sd_uid = get_be32(&from->sd_uid);
2433 to->sd_gid = get_be32(&from->sd_gid);
2434 to->sd_size = get_be32(&from->sd_size);
2435 }
2436
2437 static int read_one_dir(struct untracked_cache_dir **untracked_,
2438 struct read_data *rd)
2439 {
2440 struct untracked_cache_dir ud, *untracked;
2441 const unsigned char *next, *data = rd->data, *end = rd->end;
2442 unsigned int value;
2443 int i, len;
2444
2445 memset(&ud, 0, sizeof(ud));
2446
2447 next = data;
2448 value = decode_varint(&next);
2449 if (next > end)
2450 return -1;
2451 ud.recurse = 1;
2452 ud.untracked_alloc = value;
2453 ud.untracked_nr = value;
2454 if (ud.untracked_nr)
2455 ud.untracked = xmalloc(sizeof(*ud.untracked) * ud.untracked_nr);
2456 data = next;
2457
2458 next = data;
2459 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2460 if (next > end)
2461 return -1;
2462 ud.dirs = xmalloc(sizeof(*ud.dirs) * ud.dirs_nr);
2463 data = next;
2464
2465 len = strlen((const char *)data);
2466 next = data + len + 1;
2467 if (next > rd->end)
2468 return -1;
2469 *untracked_ = untracked = xmalloc(sizeof(*untracked) + len);
2470 memcpy(untracked, &ud, sizeof(ud));
2471 memcpy(untracked->name, data, len + 1);
2472 data = next;
2473
2474 for (i = 0; i < untracked->untracked_nr; i++) {
2475 len = strlen((const char *)data);
2476 next = data + len + 1;
2477 if (next > rd->end)
2478 return -1;
2479 untracked->untracked[i] = xstrdup((const char*)data);
2480 data = next;
2481 }
2482
2483 rd->ucd[rd->index++] = untracked;
2484 rd->data = data;
2485
2486 for (i = 0; i < untracked->dirs_nr; i++) {
2487 len = read_one_dir(untracked->dirs + i, rd);
2488 if (len < 0)
2489 return -1;
2490 }
2491 return 0;
2492 }
2493
2494 static void set_check_only(size_t pos, void *cb)
2495 {
2496 struct read_data *rd = cb;
2497 struct untracked_cache_dir *ud = rd->ucd[pos];
2498 ud->check_only = 1;
2499 }
2500
2501 static void read_stat(size_t pos, void *cb)
2502 {
2503 struct read_data *rd = cb;
2504 struct untracked_cache_dir *ud = rd->ucd[pos];
2505 if (rd->data + sizeof(struct stat_data) > rd->end) {
2506 rd->data = rd->end + 1;
2507 return;
2508 }
2509 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2510 rd->data += sizeof(struct stat_data);
2511 ud->valid = 1;
2512 }
2513
2514 static void read_sha1(size_t pos, void *cb)
2515 {
2516 struct read_data *rd = cb;
2517 struct untracked_cache_dir *ud = rd->ucd[pos];
2518 if (rd->data + 20 > rd->end) {
2519 rd->data = rd->end + 1;
2520 return;
2521 }
2522 hashcpy(ud->exclude_sha1, rd->data);
2523 rd->data += 20;
2524 }
2525
2526 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2527 const struct stat_data *stat,
2528 const unsigned char *sha1)
2529 {
2530 stat_data_from_disk(&sha1_stat->stat, stat);
2531 hashcpy(sha1_stat->sha1, sha1);
2532 sha1_stat->valid = 1;
2533 }
2534
2535 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2536 {
2537 const struct ondisk_untracked_cache *ouc;
2538 struct untracked_cache *uc;
2539 struct read_data rd;
2540 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2541 const char *ident;
2542 int ident_len, len;
2543
2544 if (sz <= 1 || end[-1] != '\0')
2545 return NULL;
2546 end--;
2547
2548 ident_len = decode_varint(&next);
2549 if (next + ident_len > end)
2550 return NULL;
2551 ident = (const char *)next;
2552 next += ident_len;
2553
2554 ouc = (const struct ondisk_untracked_cache *)next;
2555 if (next + ouc_size(0) > end)
2556 return NULL;
2557
2558 uc = xcalloc(1, sizeof(*uc));
2559 strbuf_init(&uc->ident, ident_len);
2560 strbuf_add(&uc->ident, ident, ident_len);
2561 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2562 ouc->info_exclude_sha1);
2563 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2564 ouc->excludes_file_sha1);
2565 uc->dir_flags = get_be32(&ouc->dir_flags);
2566 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2567 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2568 next += ouc_size(strlen(ouc->exclude_per_dir));
2569 if (next >= end)
2570 goto done2;
2571
2572 len = decode_varint(&next);
2573 if (next > end || len == 0)
2574 goto done2;
2575
2576 rd.valid = ewah_new();
2577 rd.check_only = ewah_new();
2578 rd.sha1_valid = ewah_new();
2579 rd.data = next;
2580 rd.end = end;
2581 rd.index = 0;
2582 rd.ucd = xmalloc(sizeof(*rd.ucd) * len);
2583
2584 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2585 goto done;
2586
2587 next = rd.data;
2588 len = ewah_read_mmap(rd.valid, next, end - next);
2589 if (len < 0)
2590 goto done;
2591
2592 next += len;
2593 len = ewah_read_mmap(rd.check_only, next, end - next);
2594 if (len < 0)
2595 goto done;
2596
2597 next += len;
2598 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2599 if (len < 0)
2600 goto done;
2601
2602 ewah_each_bit(rd.check_only, set_check_only, &rd);
2603 rd.data = next + len;
2604 ewah_each_bit(rd.valid, read_stat, &rd);
2605 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2606 next = rd.data;
2607
2608 done:
2609 free(rd.ucd);
2610 ewah_free(rd.valid);
2611 ewah_free(rd.check_only);
2612 ewah_free(rd.sha1_valid);
2613 done2:
2614 if (next != end) {
2615 free_untracked_cache(uc);
2616 uc = NULL;
2617 }
2618 return uc;
2619 }
2620
2621 static void invalidate_one_directory(struct untracked_cache *uc,
2622 struct untracked_cache_dir *ucd)
2623 {
2624 uc->dir_invalidated++;
2625 ucd->valid = 0;
2626 ucd->untracked_nr = 0;
2627 }
2628
2629 /*
2630 * Normally when an entry is added or removed from a directory,
2631 * invalidating that directory is enough. No need to touch its
2632 * ancestors. When a directory is shown as "foo/bar/" in git-status
2633 * however, deleting or adding an entry may have cascading effect.
2634 *
2635 * Say the "foo/bar/file" has become untracked, we need to tell the
2636 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2637 * directory any more (because "bar" is managed by foo as an untracked
2638 * "file").
2639 *
2640 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2641 * was the last untracked entry in the entire "foo", we should show
2642 * "foo/" instead. Which means we have to invalidate past "bar" up to
2643 * "foo".
2644 *
2645 * This function traverses all directories from root to leaf. If there
2646 * is a chance of one of the above cases happening, we invalidate back
2647 * to root. Otherwise we just invalidate the leaf. There may be a more
2648 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2649 * detect these cases and avoid unnecessary invalidation, for example,
2650 * checking for the untracked entry named "bar/" in "foo", but for now
2651 * stick to something safe and simple.
2652 */
2653 static int invalidate_one_component(struct untracked_cache *uc,
2654 struct untracked_cache_dir *dir,
2655 const char *path, int len)
2656 {
2657 const char *rest = strchr(path, '/');
2658
2659 if (rest) {
2660 int component_len = rest - path;
2661 struct untracked_cache_dir *d =
2662 lookup_untracked(uc, dir, path, component_len);
2663 int ret =
2664 invalidate_one_component(uc, d, rest + 1,
2665 len - (component_len + 1));
2666 if (ret)
2667 invalidate_one_directory(uc, dir);
2668 return ret;
2669 }
2670
2671 invalidate_one_directory(uc, dir);
2672 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2673 }
2674
2675 void untracked_cache_invalidate_path(struct index_state *istate,
2676 const char *path)
2677 {
2678 if (!istate->untracked || !istate->untracked->root)
2679 return;
2680 invalidate_one_component(istate->untracked, istate->untracked->root,
2681 path, strlen(path));
2682 }
2683
2684 void untracked_cache_remove_from_index(struct index_state *istate,
2685 const char *path)
2686 {
2687 untracked_cache_invalidate_path(istate, path);
2688 }
2689
2690 void untracked_cache_add_to_index(struct index_state *istate,
2691 const char *path)
2692 {
2693 untracked_cache_invalidate_path(istate, path);
2694 }