convert trivial cases to ALLOC_ARRAY
[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 memset(el, 0, sizeof(*el));
568 }
569
570 static void trim_trailing_spaces(char *buf)
571 {
572 char *p, *last_space = NULL;
573
574 for (p = buf; *p; p++)
575 switch (*p) {
576 case ' ':
577 if (!last_space)
578 last_space = p;
579 break;
580 case '\\':
581 p++;
582 if (!*p)
583 return;
584 /* fallthrough */
585 default:
586 last_space = NULL;
587 }
588
589 if (last_space)
590 *last_space = '\0';
591 }
592
593 /*
594 * Given a subdirectory name and "dir" of the current directory,
595 * search the subdir in "dir" and return it, or create a new one if it
596 * does not exist in "dir".
597 *
598 * If "name" has the trailing slash, it'll be excluded in the search.
599 */
600 static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
601 struct untracked_cache_dir *dir,
602 const char *name, int len)
603 {
604 int first, last;
605 struct untracked_cache_dir *d;
606 if (!dir)
607 return NULL;
608 if (len && name[len - 1] == '/')
609 len--;
610 first = 0;
611 last = dir->dirs_nr;
612 while (last > first) {
613 int cmp, next = (last + first) >> 1;
614 d = dir->dirs[next];
615 cmp = strncmp(name, d->name, len);
616 if (!cmp && strlen(d->name) > len)
617 cmp = -1;
618 if (!cmp)
619 return d;
620 if (cmp < 0) {
621 last = next;
622 continue;
623 }
624 first = next+1;
625 }
626
627 uc->dir_created++;
628 d = xmalloc(sizeof(*d) + len + 1);
629 memset(d, 0, sizeof(*d));
630 memcpy(d->name, name, len);
631 d->name[len] = '\0';
632
633 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
634 memmove(dir->dirs + first + 1, dir->dirs + first,
635 (dir->dirs_nr - first) * sizeof(*dir->dirs));
636 dir->dirs_nr++;
637 dir->dirs[first] = d;
638 return d;
639 }
640
641 static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
642 {
643 int i;
644 dir->valid = 0;
645 dir->untracked_nr = 0;
646 for (i = 0; i < dir->dirs_nr; i++)
647 do_invalidate_gitignore(dir->dirs[i]);
648 }
649
650 static void invalidate_gitignore(struct untracked_cache *uc,
651 struct untracked_cache_dir *dir)
652 {
653 uc->gitignore_invalidated++;
654 do_invalidate_gitignore(dir);
655 }
656
657 static void invalidate_directory(struct untracked_cache *uc,
658 struct untracked_cache_dir *dir)
659 {
660 int i;
661 uc->dir_invalidated++;
662 dir->valid = 0;
663 dir->untracked_nr = 0;
664 for (i = 0; i < dir->dirs_nr; i++)
665 dir->dirs[i]->recurse = 0;
666 }
667
668 /*
669 * Given a file with name "fname", read it (either from disk, or from
670 * the index if "check_index" is non-zero), parse it and store the
671 * exclude rules in "el".
672 *
673 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
674 * stat data from disk (only valid if add_excludes returns zero). If
675 * ss_valid is non-zero, "ss" must contain good value as input.
676 */
677 static int add_excludes(const char *fname, const char *base, int baselen,
678 struct exclude_list *el, int check_index,
679 struct sha1_stat *sha1_stat)
680 {
681 struct stat st;
682 int fd, i, lineno = 1;
683 size_t size = 0;
684 char *buf, *entry;
685
686 fd = open(fname, O_RDONLY);
687 if (fd < 0 || fstat(fd, &st) < 0) {
688 if (errno != ENOENT)
689 warn_on_inaccessible(fname);
690 if (0 <= fd)
691 close(fd);
692 if (!check_index ||
693 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL)
694 return -1;
695 if (size == 0) {
696 free(buf);
697 return 0;
698 }
699 if (buf[size-1] != '\n') {
700 buf = xrealloc(buf, size+1);
701 buf[size++] = '\n';
702 }
703 } else {
704 size = xsize_t(st.st_size);
705 if (size == 0) {
706 if (sha1_stat) {
707 fill_stat_data(&sha1_stat->stat, &st);
708 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
709 sha1_stat->valid = 1;
710 }
711 close(fd);
712 return 0;
713 }
714 buf = xmalloc(size+1);
715 if (read_in_full(fd, buf, size) != size) {
716 free(buf);
717 close(fd);
718 return -1;
719 }
720 buf[size++] = '\n';
721 close(fd);
722 if (sha1_stat) {
723 int pos;
724 if (sha1_stat->valid &&
725 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st))
726 ; /* no content change, ss->sha1 still good */
727 else if (check_index &&
728 (pos = cache_name_pos(fname, strlen(fname))) >= 0 &&
729 !ce_stage(active_cache[pos]) &&
730 ce_uptodate(active_cache[pos]) &&
731 !would_convert_to_git(fname))
732 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
733 else
734 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
735 fill_stat_data(&sha1_stat->stat, &st);
736 sha1_stat->valid = 1;
737 }
738 }
739
740 el->filebuf = buf;
741
742 if (skip_utf8_bom(&buf, size))
743 size -= buf - el->filebuf;
744
745 entry = buf;
746
747 for (i = 0; i < size; i++) {
748 if (buf[i] == '\n') {
749 if (entry != buf + i && entry[0] != '#') {
750 buf[i - (i && buf[i-1] == '\r')] = 0;
751 trim_trailing_spaces(entry);
752 add_exclude(entry, base, baselen, el, lineno);
753 }
754 lineno++;
755 entry = buf + i + 1;
756 }
757 }
758 return 0;
759 }
760
761 int add_excludes_from_file_to_list(const char *fname, const char *base,
762 int baselen, struct exclude_list *el,
763 int check_index)
764 {
765 return add_excludes(fname, base, baselen, el, check_index, NULL);
766 }
767
768 struct exclude_list *add_exclude_list(struct dir_struct *dir,
769 int group_type, const char *src)
770 {
771 struct exclude_list *el;
772 struct exclude_list_group *group;
773
774 group = &dir->exclude_list_group[group_type];
775 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
776 el = &group->el[group->nr++];
777 memset(el, 0, sizeof(*el));
778 el->src = src;
779 return el;
780 }
781
782 /*
783 * Used to set up core.excludesfile and .git/info/exclude lists.
784 */
785 static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
786 struct sha1_stat *sha1_stat)
787 {
788 struct exclude_list *el;
789 /*
790 * catch setup_standard_excludes() that's called before
791 * dir->untracked is assigned. That function behaves
792 * differently when dir->untracked is non-NULL.
793 */
794 if (!dir->untracked)
795 dir->unmanaged_exclude_files++;
796 el = add_exclude_list(dir, EXC_FILE, fname);
797 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0)
798 die("cannot use %s as an exclude file", fname);
799 }
800
801 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
802 {
803 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
804 add_excludes_from_file_1(dir, fname, NULL);
805 }
806
807 int match_basename(const char *basename, int basenamelen,
808 const char *pattern, int prefix, int patternlen,
809 int flags)
810 {
811 if (prefix == patternlen) {
812 if (patternlen == basenamelen &&
813 !strncmp_icase(pattern, basename, basenamelen))
814 return 1;
815 } else if (flags & EXC_FLAG_ENDSWITH) {
816 /* "*literal" matching against "fooliteral" */
817 if (patternlen - 1 <= basenamelen &&
818 !strncmp_icase(pattern + 1,
819 basename + basenamelen - (patternlen - 1),
820 patternlen - 1))
821 return 1;
822 } else {
823 if (fnmatch_icase_mem(pattern, patternlen,
824 basename, basenamelen,
825 0) == 0)
826 return 1;
827 }
828 return 0;
829 }
830
831 int match_pathname(const char *pathname, int pathlen,
832 const char *base, int baselen,
833 const char *pattern, int prefix, int patternlen,
834 int flags)
835 {
836 const char *name;
837 int namelen;
838
839 /*
840 * match with FNM_PATHNAME; the pattern has base implicitly
841 * in front of it.
842 */
843 if (*pattern == '/') {
844 pattern++;
845 patternlen--;
846 prefix--;
847 }
848
849 /*
850 * baselen does not count the trailing slash. base[] may or
851 * may not end with a trailing slash though.
852 */
853 if (pathlen < baselen + 1 ||
854 (baselen && pathname[baselen] != '/') ||
855 strncmp_icase(pathname, base, baselen))
856 return 0;
857
858 namelen = baselen ? pathlen - baselen - 1 : pathlen;
859 name = pathname + pathlen - namelen;
860
861 if (prefix) {
862 /*
863 * if the non-wildcard part is longer than the
864 * remaining pathname, surely it cannot match.
865 */
866 if (prefix > namelen)
867 return 0;
868
869 if (strncmp_icase(pattern, name, prefix))
870 return 0;
871 pattern += prefix;
872 patternlen -= prefix;
873 name += prefix;
874 namelen -= prefix;
875
876 /*
877 * If the whole pattern did not have a wildcard,
878 * then our prefix match is all we need; we
879 * do not need to call fnmatch at all.
880 */
881 if (!patternlen && !namelen)
882 return 1;
883 }
884
885 return fnmatch_icase_mem(pattern, patternlen,
886 name, namelen,
887 WM_PATHNAME) == 0;
888 }
889
890 /*
891 * Scan the given exclude list in reverse to see whether pathname
892 * should be ignored. The first match (i.e. the last on the list), if
893 * any, determines the fate. Returns the exclude_list element which
894 * matched, or NULL for undecided.
895 */
896 static struct exclude *last_exclude_matching_from_list(const char *pathname,
897 int pathlen,
898 const char *basename,
899 int *dtype,
900 struct exclude_list *el)
901 {
902 struct exclude *exc = NULL; /* undecided */
903 int i;
904
905 if (!el->nr)
906 return NULL; /* undefined */
907
908 for (i = el->nr - 1; 0 <= i; i--) {
909 struct exclude *x = el->excludes[i];
910 const char *exclude = x->pattern;
911 int prefix = x->nowildcardlen;
912
913 if (x->flags & EXC_FLAG_MUSTBEDIR) {
914 if (*dtype == DT_UNKNOWN)
915 *dtype = get_dtype(NULL, pathname, pathlen);
916 if (*dtype != DT_DIR)
917 continue;
918 }
919
920 if (x->flags & EXC_FLAG_NODIR) {
921 if (match_basename(basename,
922 pathlen - (basename - pathname),
923 exclude, prefix, x->patternlen,
924 x->flags)) {
925 exc = x;
926 break;
927 }
928 continue;
929 }
930
931 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
932 if (match_pathname(pathname, pathlen,
933 x->base, x->baselen ? x->baselen - 1 : 0,
934 exclude, prefix, x->patternlen, x->flags)) {
935 exc = x;
936 break;
937 }
938 }
939 return exc;
940 }
941
942 /*
943 * Scan the list and let the last match determine the fate.
944 * Return 1 for exclude, 0 for include and -1 for undecided.
945 */
946 int is_excluded_from_list(const char *pathname,
947 int pathlen, const char *basename, int *dtype,
948 struct exclude_list *el)
949 {
950 struct exclude *exclude;
951 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
952 if (exclude)
953 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
954 return -1; /* undecided */
955 }
956
957 static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
958 const char *pathname, int pathlen, const char *basename,
959 int *dtype_p)
960 {
961 int i, j;
962 struct exclude_list_group *group;
963 struct exclude *exclude;
964 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
965 group = &dir->exclude_list_group[i];
966 for (j = group->nr - 1; j >= 0; j--) {
967 exclude = last_exclude_matching_from_list(
968 pathname, pathlen, basename, dtype_p,
969 &group->el[j]);
970 if (exclude)
971 return exclude;
972 }
973 }
974 return NULL;
975 }
976
977 /*
978 * Loads the per-directory exclude list for the substring of base
979 * which has a char length of baselen.
980 */
981 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
982 {
983 struct exclude_list_group *group;
984 struct exclude_list *el;
985 struct exclude_stack *stk = NULL;
986 struct untracked_cache_dir *untracked;
987 int current;
988
989 group = &dir->exclude_list_group[EXC_DIRS];
990
991 /*
992 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
993 * which originate from directories not in the prefix of the
994 * path being checked.
995 */
996 while ((stk = dir->exclude_stack) != NULL) {
997 if (stk->baselen <= baselen &&
998 !strncmp(dir->basebuf.buf, base, stk->baselen))
999 break;
1000 el = &group->el[dir->exclude_stack->exclude_ix];
1001 dir->exclude_stack = stk->prev;
1002 dir->exclude = NULL;
1003 free((char *)el->src); /* see strbuf_detach() below */
1004 clear_exclude_list(el);
1005 free(stk);
1006 group->nr--;
1007 }
1008
1009 /* Skip traversing into sub directories if the parent is excluded */
1010 if (dir->exclude)
1011 return;
1012
1013 /*
1014 * Lazy initialization. All call sites currently just
1015 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1016 * them seems lots of work for little benefit.
1017 */
1018 if (!dir->basebuf.buf)
1019 strbuf_init(&dir->basebuf, PATH_MAX);
1020
1021 /* Read from the parent directories and push them down. */
1022 current = stk ? stk->baselen : -1;
1023 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
1024 if (dir->untracked)
1025 untracked = stk ? stk->ucd : dir->untracked->root;
1026 else
1027 untracked = NULL;
1028
1029 while (current < baselen) {
1030 const char *cp;
1031 struct sha1_stat sha1_stat;
1032
1033 stk = xcalloc(1, sizeof(*stk));
1034 if (current < 0) {
1035 cp = base;
1036 current = 0;
1037 } else {
1038 cp = strchr(base + current + 1, '/');
1039 if (!cp)
1040 die("oops in prep_exclude");
1041 cp++;
1042 untracked =
1043 lookup_untracked(dir->untracked, untracked,
1044 base + current,
1045 cp - base - current);
1046 }
1047 stk->prev = dir->exclude_stack;
1048 stk->baselen = cp - base;
1049 stk->exclude_ix = group->nr;
1050 stk->ucd = untracked;
1051 el = add_exclude_list(dir, EXC_DIRS, NULL);
1052 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1053 assert(stk->baselen == dir->basebuf.len);
1054
1055 /* Abort if the directory is excluded */
1056 if (stk->baselen) {
1057 int dt = DT_DIR;
1058 dir->basebuf.buf[stk->baselen - 1] = 0;
1059 dir->exclude = last_exclude_matching_from_lists(dir,
1060 dir->basebuf.buf, stk->baselen - 1,
1061 dir->basebuf.buf + current, &dt);
1062 dir->basebuf.buf[stk->baselen - 1] = '/';
1063 if (dir->exclude &&
1064 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1065 dir->exclude = NULL;
1066 if (dir->exclude) {
1067 dir->exclude_stack = stk;
1068 return;
1069 }
1070 }
1071
1072 /* Try to read per-directory file */
1073 hashclr(sha1_stat.sha1);
1074 sha1_stat.valid = 0;
1075 if (dir->exclude_per_dir &&
1076 /*
1077 * If we know that no files have been added in
1078 * this directory (i.e. valid_cached_dir() has
1079 * been executed and set untracked->valid) ..
1080 */
1081 (!untracked || !untracked->valid ||
1082 /*
1083 * .. and .gitignore does not exist before
1084 * (i.e. null exclude_sha1). Then we can skip
1085 * loading .gitignore, which would result in
1086 * ENOENT anyway.
1087 */
1088 !is_null_sha1(untracked->exclude_sha1))) {
1089 /*
1090 * dir->basebuf gets reused by the traversal, but we
1091 * need fname to remain unchanged to ensure the src
1092 * member of each struct exclude correctly
1093 * back-references its source file. Other invocations
1094 * of add_exclude_list provide stable strings, so we
1095 * strbuf_detach() and free() here in the caller.
1096 */
1097 struct strbuf sb = STRBUF_INIT;
1098 strbuf_addbuf(&sb, &dir->basebuf);
1099 strbuf_addstr(&sb, dir->exclude_per_dir);
1100 el->src = strbuf_detach(&sb, NULL);
1101 add_excludes(el->src, el->src, stk->baselen, el, 1,
1102 untracked ? &sha1_stat : NULL);
1103 }
1104 /*
1105 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1106 * will first be called in valid_cached_dir() then maybe many
1107 * times more in last_exclude_matching(). When the cache is
1108 * used, last_exclude_matching() will not be called and
1109 * reading .gitignore content will be a waste.
1110 *
1111 * So when it's called by valid_cached_dir() and we can get
1112 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1113 * modified on work tree), we could delay reading the
1114 * .gitignore content until we absolutely need it in
1115 * last_exclude_matching(). Be careful about ignore rule
1116 * order, though, if you do that.
1117 */
1118 if (untracked &&
1119 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1120 invalidate_gitignore(dir->untracked, untracked);
1121 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
1122 }
1123 dir->exclude_stack = stk;
1124 current = stk->baselen;
1125 }
1126 strbuf_setlen(&dir->basebuf, baselen);
1127 }
1128
1129 /*
1130 * Loads the exclude lists for the directory containing pathname, then
1131 * scans all exclude lists to determine whether pathname is excluded.
1132 * Returns the exclude_list element which matched, or NULL for
1133 * undecided.
1134 */
1135 struct exclude *last_exclude_matching(struct dir_struct *dir,
1136 const char *pathname,
1137 int *dtype_p)
1138 {
1139 int pathlen = strlen(pathname);
1140 const char *basename = strrchr(pathname, '/');
1141 basename = (basename) ? basename+1 : pathname;
1142
1143 prep_exclude(dir, pathname, basename-pathname);
1144
1145 if (dir->exclude)
1146 return dir->exclude;
1147
1148 return last_exclude_matching_from_lists(dir, pathname, pathlen,
1149 basename, dtype_p);
1150 }
1151
1152 /*
1153 * Loads the exclude lists for the directory containing pathname, then
1154 * scans all exclude lists to determine whether pathname is excluded.
1155 * Returns 1 if true, otherwise 0.
1156 */
1157 int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
1158 {
1159 struct exclude *exclude =
1160 last_exclude_matching(dir, pathname, dtype_p);
1161 if (exclude)
1162 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
1163 return 0;
1164 }
1165
1166 static struct dir_entry *dir_entry_new(const char *pathname, int len)
1167 {
1168 struct dir_entry *ent;
1169
1170 ent = xmalloc(sizeof(*ent) + len + 1);
1171 ent->len = len;
1172 memcpy(ent->name, pathname, len);
1173 ent->name[len] = 0;
1174 return ent;
1175 }
1176
1177 static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
1178 {
1179 if (cache_file_exists(pathname, len, ignore_case))
1180 return NULL;
1181
1182 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
1183 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1184 }
1185
1186 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
1187 {
1188 if (!cache_name_is_other(pathname, len))
1189 return NULL;
1190
1191 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
1192 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1193 }
1194
1195 enum exist_status {
1196 index_nonexistent = 0,
1197 index_directory,
1198 index_gitdir
1199 };
1200
1201 /*
1202 * Do not use the alphabetically sorted index to look up
1203 * the directory name; instead, use the case insensitive
1204 * directory hash.
1205 */
1206 static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1207 {
1208 struct cache_entry *ce;
1209
1210 if (cache_dir_exists(dirname, len))
1211 return index_directory;
1212
1213 ce = cache_file_exists(dirname, len, ignore_case);
1214 if (ce && S_ISGITLINK(ce->ce_mode))
1215 return index_gitdir;
1216
1217 return index_nonexistent;
1218 }
1219
1220 /*
1221 * The index sorts alphabetically by entry name, which
1222 * means that a gitlink sorts as '\0' at the end, while
1223 * a directory (which is defined not as an entry, but as
1224 * the files it contains) will sort with the '/' at the
1225 * end.
1226 */
1227 static enum exist_status directory_exists_in_index(const char *dirname, int len)
1228 {
1229 int pos;
1230
1231 if (ignore_case)
1232 return directory_exists_in_index_icase(dirname, len);
1233
1234 pos = cache_name_pos(dirname, len);
1235 if (pos < 0)
1236 pos = -pos-1;
1237 while (pos < active_nr) {
1238 const struct cache_entry *ce = active_cache[pos++];
1239 unsigned char endchar;
1240
1241 if (strncmp(ce->name, dirname, len))
1242 break;
1243 endchar = ce->name[len];
1244 if (endchar > '/')
1245 break;
1246 if (endchar == '/')
1247 return index_directory;
1248 if (!endchar && S_ISGITLINK(ce->ce_mode))
1249 return index_gitdir;
1250 }
1251 return index_nonexistent;
1252 }
1253
1254 /*
1255 * When we find a directory when traversing the filesystem, we
1256 * have three distinct cases:
1257 *
1258 * - ignore it
1259 * - see it as a directory
1260 * - recurse into it
1261 *
1262 * and which one we choose depends on a combination of existing
1263 * git index contents and the flags passed into the directory
1264 * traversal routine.
1265 *
1266 * Case 1: If we *already* have entries in the index under that
1267 * directory name, we always recurse into the directory to see
1268 * all the files.
1269 *
1270 * Case 2: If we *already* have that directory name as a gitlink,
1271 * we always continue to see it as a gitlink, regardless of whether
1272 * there is an actual git directory there or not (it might not
1273 * be checked out as a subproject!)
1274 *
1275 * Case 3: if we didn't have it in the index previously, we
1276 * have a few sub-cases:
1277 *
1278 * (a) if "show_other_directories" is true, we show it as
1279 * just a directory, unless "hide_empty_directories" is
1280 * also true, in which case we need to check if it contains any
1281 * untracked and / or ignored files.
1282 * (b) if it looks like a git directory, and we don't have
1283 * 'no_gitlinks' set we treat it as a gitlink, and show it
1284 * as a directory.
1285 * (c) otherwise, we recurse into it.
1286 */
1287 static enum path_treatment treat_directory(struct dir_struct *dir,
1288 struct untracked_cache_dir *untracked,
1289 const char *dirname, int len, int baselen, int exclude,
1290 const struct path_simplify *simplify)
1291 {
1292 /* The "len-1" is to strip the final '/' */
1293 switch (directory_exists_in_index(dirname, len-1)) {
1294 case index_directory:
1295 return path_recurse;
1296
1297 case index_gitdir:
1298 return path_none;
1299
1300 case index_nonexistent:
1301 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1302 break;
1303 if (!(dir->flags & DIR_NO_GITLINKS)) {
1304 unsigned char sha1[20];
1305 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1306 return path_untracked;
1307 }
1308 return path_recurse;
1309 }
1310
1311 /* This is the "show_other_directories" case */
1312
1313 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1314 return exclude ? path_excluded : path_untracked;
1315
1316 untracked = lookup_untracked(dir->untracked, untracked,
1317 dirname + baselen, len - baselen);
1318 return read_directory_recursive(dir, dirname, len,
1319 untracked, 1, simplify);
1320 }
1321
1322 /*
1323 * This is an inexact early pruning of any recursive directory
1324 * reading - if the path cannot possibly be in the pathspec,
1325 * return true, and we'll skip it early.
1326 */
1327 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1328 {
1329 if (simplify) {
1330 for (;;) {
1331 const char *match = simplify->path;
1332 int len = simplify->len;
1333
1334 if (!match)
1335 break;
1336 if (len > pathlen)
1337 len = pathlen;
1338 if (!memcmp(path, match, len))
1339 return 0;
1340 simplify++;
1341 }
1342 return 1;
1343 }
1344 return 0;
1345 }
1346
1347 /*
1348 * This function tells us whether an excluded path matches a
1349 * list of "interesting" pathspecs. That is, whether a path matched
1350 * by any of the pathspecs could possibly be ignored by excluding
1351 * the specified path. This can happen if:
1352 *
1353 * 1. the path is mentioned explicitly in the pathspec
1354 *
1355 * 2. the path is a directory prefix of some element in the
1356 * pathspec
1357 */
1358 static int exclude_matches_pathspec(const char *path, int len,
1359 const struct path_simplify *simplify)
1360 {
1361 if (simplify) {
1362 for (; simplify->path; simplify++) {
1363 if (len == simplify->len
1364 && !memcmp(path, simplify->path, len))
1365 return 1;
1366 if (len < simplify->len
1367 && simplify->path[len] == '/'
1368 && !memcmp(path, simplify->path, len))
1369 return 1;
1370 }
1371 }
1372 return 0;
1373 }
1374
1375 static int get_index_dtype(const char *path, int len)
1376 {
1377 int pos;
1378 const struct cache_entry *ce;
1379
1380 ce = cache_file_exists(path, len, 0);
1381 if (ce) {
1382 if (!ce_uptodate(ce))
1383 return DT_UNKNOWN;
1384 if (S_ISGITLINK(ce->ce_mode))
1385 return DT_DIR;
1386 /*
1387 * Nobody actually cares about the
1388 * difference between DT_LNK and DT_REG
1389 */
1390 return DT_REG;
1391 }
1392
1393 /* Try to look it up as a directory */
1394 pos = cache_name_pos(path, len);
1395 if (pos >= 0)
1396 return DT_UNKNOWN;
1397 pos = -pos-1;
1398 while (pos < active_nr) {
1399 ce = active_cache[pos++];
1400 if (strncmp(ce->name, path, len))
1401 break;
1402 if (ce->name[len] > '/')
1403 break;
1404 if (ce->name[len] < '/')
1405 continue;
1406 if (!ce_uptodate(ce))
1407 break; /* continue? */
1408 return DT_DIR;
1409 }
1410 return DT_UNKNOWN;
1411 }
1412
1413 static int get_dtype(struct dirent *de, const char *path, int len)
1414 {
1415 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1416 struct stat st;
1417
1418 if (dtype != DT_UNKNOWN)
1419 return dtype;
1420 dtype = get_index_dtype(path, len);
1421 if (dtype != DT_UNKNOWN)
1422 return dtype;
1423 if (lstat(path, &st))
1424 return dtype;
1425 if (S_ISREG(st.st_mode))
1426 return DT_REG;
1427 if (S_ISDIR(st.st_mode))
1428 return DT_DIR;
1429 if (S_ISLNK(st.st_mode))
1430 return DT_LNK;
1431 return dtype;
1432 }
1433
1434 static enum path_treatment treat_one_path(struct dir_struct *dir,
1435 struct untracked_cache_dir *untracked,
1436 struct strbuf *path,
1437 int baselen,
1438 const struct path_simplify *simplify,
1439 int dtype, struct dirent *de)
1440 {
1441 int exclude;
1442 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1443
1444 if (dtype == DT_UNKNOWN)
1445 dtype = get_dtype(de, path->buf, path->len);
1446
1447 /* Always exclude indexed files */
1448 if (dtype != DT_DIR && has_path_in_index)
1449 return path_none;
1450
1451 /*
1452 * When we are looking at a directory P in the working tree,
1453 * there are three cases:
1454 *
1455 * (1) P exists in the index. Everything inside the directory P in
1456 * the working tree needs to go when P is checked out from the
1457 * index.
1458 *
1459 * (2) P does not exist in the index, but there is P/Q in the index.
1460 * We know P will stay a directory when we check out the contents
1461 * of the index, but we do not know yet if there is a directory
1462 * P/Q in the working tree to be killed, so we need to recurse.
1463 *
1464 * (3) P does not exist in the index, and there is no P/Q in the index
1465 * to require P to be a directory, either. Only in this case, we
1466 * know that everything inside P will not be killed without
1467 * recursing.
1468 */
1469 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1470 (dtype == DT_DIR) &&
1471 !has_path_in_index &&
1472 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1473 return path_none;
1474
1475 exclude = is_excluded(dir, path->buf, &dtype);
1476
1477 /*
1478 * Excluded? If we don't explicitly want to show
1479 * ignored files, ignore it
1480 */
1481 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1482 return path_excluded;
1483
1484 switch (dtype) {
1485 default:
1486 return path_none;
1487 case DT_DIR:
1488 strbuf_addch(path, '/');
1489 return treat_directory(dir, untracked, path->buf, path->len,
1490 baselen, exclude, simplify);
1491 case DT_REG:
1492 case DT_LNK:
1493 return exclude ? path_excluded : path_untracked;
1494 }
1495 }
1496
1497 static enum path_treatment treat_path_fast(struct dir_struct *dir,
1498 struct untracked_cache_dir *untracked,
1499 struct cached_dir *cdir,
1500 struct strbuf *path,
1501 int baselen,
1502 const struct path_simplify *simplify)
1503 {
1504 strbuf_setlen(path, baselen);
1505 if (!cdir->ucd) {
1506 strbuf_addstr(path, cdir->file);
1507 return path_untracked;
1508 }
1509 strbuf_addstr(path, cdir->ucd->name);
1510 /* treat_one_path() does this before it calls treat_directory() */
1511 strbuf_complete(path, '/');
1512 if (cdir->ucd->check_only)
1513 /*
1514 * check_only is set as a result of treat_directory() getting
1515 * to its bottom. Verify again the same set of directories
1516 * with check_only set.
1517 */
1518 return read_directory_recursive(dir, path->buf, path->len,
1519 cdir->ucd, 1, simplify);
1520 /*
1521 * We get path_recurse in the first run when
1522 * directory_exists_in_index() returns index_nonexistent. We
1523 * are sure that new changes in the index does not impact the
1524 * outcome. Return now.
1525 */
1526 return path_recurse;
1527 }
1528
1529 static enum path_treatment treat_path(struct dir_struct *dir,
1530 struct untracked_cache_dir *untracked,
1531 struct cached_dir *cdir,
1532 struct strbuf *path,
1533 int baselen,
1534 const struct path_simplify *simplify)
1535 {
1536 int dtype;
1537 struct dirent *de = cdir->de;
1538
1539 if (!de)
1540 return treat_path_fast(dir, untracked, cdir, path,
1541 baselen, simplify);
1542 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1543 return path_none;
1544 strbuf_setlen(path, baselen);
1545 strbuf_addstr(path, de->d_name);
1546 if (simplify_away(path->buf, path->len, simplify))
1547 return path_none;
1548
1549 dtype = DTYPE(de);
1550 return treat_one_path(dir, untracked, path, baselen, simplify, dtype, de);
1551 }
1552
1553 static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1554 {
1555 if (!dir)
1556 return;
1557 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1558 dir->untracked_alloc);
1559 dir->untracked[dir->untracked_nr++] = xstrdup(name);
1560 }
1561
1562 static int valid_cached_dir(struct dir_struct *dir,
1563 struct untracked_cache_dir *untracked,
1564 struct strbuf *path,
1565 int check_only)
1566 {
1567 struct stat st;
1568
1569 if (!untracked)
1570 return 0;
1571
1572 if (stat(path->len ? path->buf : ".", &st)) {
1573 invalidate_directory(dir->untracked, untracked);
1574 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1575 return 0;
1576 }
1577 if (!untracked->valid ||
1578 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
1579 if (untracked->valid)
1580 invalidate_directory(dir->untracked, untracked);
1581 fill_stat_data(&untracked->stat_data, &st);
1582 return 0;
1583 }
1584
1585 if (untracked->check_only != !!check_only) {
1586 invalidate_directory(dir->untracked, untracked);
1587 return 0;
1588 }
1589
1590 /*
1591 * prep_exclude will be called eventually on this directory,
1592 * but it's called much later in last_exclude_matching(). We
1593 * need it now to determine the validity of the cache for this
1594 * path. The next calls will be nearly no-op, the way
1595 * prep_exclude() is designed.
1596 */
1597 if (path->len && path->buf[path->len - 1] != '/') {
1598 strbuf_addch(path, '/');
1599 prep_exclude(dir, path->buf, path->len);
1600 strbuf_setlen(path, path->len - 1);
1601 } else
1602 prep_exclude(dir, path->buf, path->len);
1603
1604 /* hopefully prep_exclude() haven't invalidated this entry... */
1605 return untracked->valid;
1606 }
1607
1608 static int open_cached_dir(struct cached_dir *cdir,
1609 struct dir_struct *dir,
1610 struct untracked_cache_dir *untracked,
1611 struct strbuf *path,
1612 int check_only)
1613 {
1614 memset(cdir, 0, sizeof(*cdir));
1615 cdir->untracked = untracked;
1616 if (valid_cached_dir(dir, untracked, path, check_only))
1617 return 0;
1618 cdir->fdir = opendir(path->len ? path->buf : ".");
1619 if (dir->untracked)
1620 dir->untracked->dir_opened++;
1621 if (!cdir->fdir)
1622 return -1;
1623 return 0;
1624 }
1625
1626 static int read_cached_dir(struct cached_dir *cdir)
1627 {
1628 if (cdir->fdir) {
1629 cdir->de = readdir(cdir->fdir);
1630 if (!cdir->de)
1631 return -1;
1632 return 0;
1633 }
1634 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1635 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
1636 if (!d->recurse) {
1637 cdir->nr_dirs++;
1638 continue;
1639 }
1640 cdir->ucd = d;
1641 cdir->nr_dirs++;
1642 return 0;
1643 }
1644 cdir->ucd = NULL;
1645 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1646 struct untracked_cache_dir *d = cdir->untracked;
1647 cdir->file = d->untracked[cdir->nr_files++];
1648 return 0;
1649 }
1650 return -1;
1651 }
1652
1653 static void close_cached_dir(struct cached_dir *cdir)
1654 {
1655 if (cdir->fdir)
1656 closedir(cdir->fdir);
1657 /*
1658 * We have gone through this directory and found no untracked
1659 * entries. Mark it valid.
1660 */
1661 if (cdir->untracked) {
1662 cdir->untracked->valid = 1;
1663 cdir->untracked->recurse = 1;
1664 }
1665 }
1666
1667 /*
1668 * Read a directory tree. We currently ignore anything but
1669 * directories, regular files and symlinks. That's because git
1670 * doesn't handle them at all yet. Maybe that will change some
1671 * day.
1672 *
1673 * Also, we ignore the name ".git" (even if it is not a directory).
1674 * That likely will not change.
1675 *
1676 * Returns the most significant path_treatment value encountered in the scan.
1677 */
1678 static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1679 const char *base, int baselen,
1680 struct untracked_cache_dir *untracked, int check_only,
1681 const struct path_simplify *simplify)
1682 {
1683 struct cached_dir cdir;
1684 enum path_treatment state, subdir_state, dir_state = path_none;
1685 struct strbuf path = STRBUF_INIT;
1686
1687 strbuf_add(&path, base, baselen);
1688
1689 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1690 goto out;
1691
1692 if (untracked)
1693 untracked->check_only = !!check_only;
1694
1695 while (!read_cached_dir(&cdir)) {
1696 /* check how the file or directory should be treated */
1697 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);
1698
1699 if (state > dir_state)
1700 dir_state = state;
1701
1702 /* recurse into subdir if instructed by treat_path */
1703 if (state == path_recurse) {
1704 struct untracked_cache_dir *ud;
1705 ud = lookup_untracked(dir->untracked, untracked,
1706 path.buf + baselen,
1707 path.len - baselen);
1708 subdir_state =
1709 read_directory_recursive(dir, path.buf, path.len,
1710 ud, check_only, simplify);
1711 if (subdir_state > dir_state)
1712 dir_state = subdir_state;
1713 }
1714
1715 if (check_only) {
1716 /* abort early if maximum state has been reached */
1717 if (dir_state == path_untracked) {
1718 if (cdir.fdir)
1719 add_untracked(untracked, path.buf + baselen);
1720 break;
1721 }
1722 /* skip the dir_add_* part */
1723 continue;
1724 }
1725
1726 /* add the path to the appropriate result list */
1727 switch (state) {
1728 case path_excluded:
1729 if (dir->flags & DIR_SHOW_IGNORED)
1730 dir_add_name(dir, path.buf, path.len);
1731 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1732 ((dir->flags & DIR_COLLECT_IGNORED) &&
1733 exclude_matches_pathspec(path.buf, path.len,
1734 simplify)))
1735 dir_add_ignored(dir, path.buf, path.len);
1736 break;
1737
1738 case path_untracked:
1739 if (dir->flags & DIR_SHOW_IGNORED)
1740 break;
1741 dir_add_name(dir, path.buf, path.len);
1742 if (cdir.fdir)
1743 add_untracked(untracked, path.buf + baselen);
1744 break;
1745
1746 default:
1747 break;
1748 }
1749 }
1750 close_cached_dir(&cdir);
1751 out:
1752 strbuf_release(&path);
1753
1754 return dir_state;
1755 }
1756
1757 static int cmp_name(const void *p1, const void *p2)
1758 {
1759 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1760 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1761
1762 return name_compare(e1->name, e1->len, e2->name, e2->len);
1763 }
1764
1765 static struct path_simplify *create_simplify(const char **pathspec)
1766 {
1767 int nr, alloc = 0;
1768 struct path_simplify *simplify = NULL;
1769
1770 if (!pathspec)
1771 return NULL;
1772
1773 for (nr = 0 ; ; nr++) {
1774 const char *match;
1775 ALLOC_GROW(simplify, nr + 1, alloc);
1776 match = *pathspec++;
1777 if (!match)
1778 break;
1779 simplify[nr].path = match;
1780 simplify[nr].len = simple_length(match);
1781 }
1782 simplify[nr].path = NULL;
1783 simplify[nr].len = 0;
1784 return simplify;
1785 }
1786
1787 static void free_simplify(struct path_simplify *simplify)
1788 {
1789 free(simplify);
1790 }
1791
1792 static int treat_leading_path(struct dir_struct *dir,
1793 const char *path, int len,
1794 const struct path_simplify *simplify)
1795 {
1796 struct strbuf sb = STRBUF_INIT;
1797 int baselen, rc = 0;
1798 const char *cp;
1799 int old_flags = dir->flags;
1800
1801 while (len && path[len - 1] == '/')
1802 len--;
1803 if (!len)
1804 return 1;
1805 baselen = 0;
1806 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1807 while (1) {
1808 cp = path + baselen + !!baselen;
1809 cp = memchr(cp, '/', path + len - cp);
1810 if (!cp)
1811 baselen = len;
1812 else
1813 baselen = cp - path;
1814 strbuf_setlen(&sb, 0);
1815 strbuf_add(&sb, path, baselen);
1816 if (!is_directory(sb.buf))
1817 break;
1818 if (simplify_away(sb.buf, sb.len, simplify))
1819 break;
1820 if (treat_one_path(dir, NULL, &sb, baselen, simplify,
1821 DT_DIR, NULL) == path_none)
1822 break; /* do not recurse into it */
1823 if (len <= baselen) {
1824 rc = 1;
1825 break; /* finished checking */
1826 }
1827 }
1828 strbuf_release(&sb);
1829 dir->flags = old_flags;
1830 return rc;
1831 }
1832
1833 static const char *get_ident_string(void)
1834 {
1835 static struct strbuf sb = STRBUF_INIT;
1836 struct utsname uts;
1837
1838 if (sb.len)
1839 return sb.buf;
1840 if (uname(&uts) < 0)
1841 die_errno(_("failed to get kernel name and information"));
1842 strbuf_addf(&sb, "Location %s, system %s %s %s", get_git_work_tree(),
1843 uts.sysname, uts.release, uts.version);
1844 return sb.buf;
1845 }
1846
1847 static int ident_in_untracked(const struct untracked_cache *uc)
1848 {
1849 const char *end = uc->ident.buf + uc->ident.len;
1850 const char *p = uc->ident.buf;
1851
1852 for (p = uc->ident.buf; p < end; p += strlen(p) + 1)
1853 if (!strcmp(p, get_ident_string()))
1854 return 1;
1855 return 0;
1856 }
1857
1858 void add_untracked_ident(struct untracked_cache *uc)
1859 {
1860 if (ident_in_untracked(uc))
1861 return;
1862 strbuf_addstr(&uc->ident, get_ident_string());
1863 /* this strbuf contains a list of strings, save NUL too */
1864 strbuf_addch(&uc->ident, 0);
1865 }
1866
1867 static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1868 int base_len,
1869 const struct pathspec *pathspec)
1870 {
1871 struct untracked_cache_dir *root;
1872
1873 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
1874 return NULL;
1875
1876 /*
1877 * We only support $GIT_DIR/info/exclude and core.excludesfile
1878 * as the global ignore rule files. Any other additions
1879 * (e.g. from command line) invalidate the cache. This
1880 * condition also catches running setup_standard_excludes()
1881 * before setting dir->untracked!
1882 */
1883 if (dir->unmanaged_exclude_files)
1884 return NULL;
1885
1886 /*
1887 * Optimize for the main use case only: whole-tree git
1888 * status. More work involved in treat_leading_path() if we
1889 * use cache on just a subset of the worktree. pathspec
1890 * support could make the matter even worse.
1891 */
1892 if (base_len || (pathspec && pathspec->nr))
1893 return NULL;
1894
1895 /* Different set of flags may produce different results */
1896 if (dir->flags != dir->untracked->dir_flags ||
1897 /*
1898 * See treat_directory(), case index_nonexistent. Without
1899 * this flag, we may need to also cache .git file content
1900 * for the resolve_gitlink_ref() call, which we don't.
1901 */
1902 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1903 /* We don't support collecting ignore files */
1904 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1905 DIR_COLLECT_IGNORED)))
1906 return NULL;
1907
1908 /*
1909 * If we use .gitignore in the cache and now you change it to
1910 * .gitexclude, everything will go wrong.
1911 */
1912 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
1913 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
1914 return NULL;
1915
1916 /*
1917 * EXC_CMDL is not considered in the cache. If people set it,
1918 * skip the cache.
1919 */
1920 if (dir->exclude_list_group[EXC_CMDL].nr)
1921 return NULL;
1922
1923 if (!ident_in_untracked(dir->untracked)) {
1924 warning(_("Untracked cache is disabled on this system."));
1925 return NULL;
1926 }
1927
1928 if (!dir->untracked->root) {
1929 const int len = sizeof(*dir->untracked->root);
1930 dir->untracked->root = xmalloc(len);
1931 memset(dir->untracked->root, 0, len);
1932 }
1933
1934 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
1935 root = dir->untracked->root;
1936 if (hashcmp(dir->ss_info_exclude.sha1,
1937 dir->untracked->ss_info_exclude.sha1)) {
1938 invalidate_gitignore(dir->untracked, root);
1939 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
1940 }
1941 if (hashcmp(dir->ss_excludes_file.sha1,
1942 dir->untracked->ss_excludes_file.sha1)) {
1943 invalidate_gitignore(dir->untracked, root);
1944 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
1945 }
1946
1947 /* Make sure this directory is not dropped out at saving phase */
1948 root->recurse = 1;
1949 return root;
1950 }
1951
1952 int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1953 {
1954 struct path_simplify *simplify;
1955 struct untracked_cache_dir *untracked;
1956
1957 /*
1958 * Check out create_simplify()
1959 */
1960 if (pathspec)
1961 GUARD_PATHSPEC(pathspec,
1962 PATHSPEC_FROMTOP |
1963 PATHSPEC_MAXDEPTH |
1964 PATHSPEC_LITERAL |
1965 PATHSPEC_GLOB |
1966 PATHSPEC_ICASE |
1967 PATHSPEC_EXCLUDE);
1968
1969 if (has_symlink_leading_path(path, len))
1970 return dir->nr;
1971
1972 /*
1973 * exclude patterns are treated like positive ones in
1974 * create_simplify. Usually exclude patterns should be a
1975 * subset of positive ones, which has no impacts on
1976 * create_simplify().
1977 */
1978 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
1979 untracked = validate_untracked_cache(dir, len, pathspec);
1980 if (!untracked)
1981 /*
1982 * make sure untracked cache code path is disabled,
1983 * e.g. prep_exclude()
1984 */
1985 dir->untracked = NULL;
1986 if (!len || treat_leading_path(dir, path, len, simplify))
1987 read_directory_recursive(dir, path, len, untracked, 0, simplify);
1988 free_simplify(simplify);
1989 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1990 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1991 if (dir->untracked) {
1992 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
1993 trace_printf_key(&trace_untracked_stats,
1994 "node creation: %u\n"
1995 "gitignore invalidation: %u\n"
1996 "directory invalidation: %u\n"
1997 "opendir: %u\n",
1998 dir->untracked->dir_created,
1999 dir->untracked->gitignore_invalidated,
2000 dir->untracked->dir_invalidated,
2001 dir->untracked->dir_opened);
2002 if (dir->untracked == the_index.untracked &&
2003 (dir->untracked->dir_opened ||
2004 dir->untracked->gitignore_invalidated ||
2005 dir->untracked->dir_invalidated))
2006 the_index.cache_changed |= UNTRACKED_CHANGED;
2007 if (dir->untracked != the_index.untracked) {
2008 free(dir->untracked);
2009 dir->untracked = NULL;
2010 }
2011 }
2012 return dir->nr;
2013 }
2014
2015 int file_exists(const char *f)
2016 {
2017 struct stat sb;
2018 return lstat(f, &sb) == 0;
2019 }
2020
2021 static int cmp_icase(char a, char b)
2022 {
2023 if (a == b)
2024 return 0;
2025 if (ignore_case)
2026 return toupper(a) - toupper(b);
2027 return a - b;
2028 }
2029
2030 /*
2031 * Given two normalized paths (a trailing slash is ok), if subdir is
2032 * outside dir, return -1. Otherwise return the offset in subdir that
2033 * can be used as relative path to dir.
2034 */
2035 int dir_inside_of(const char *subdir, const char *dir)
2036 {
2037 int offset = 0;
2038
2039 assert(dir && subdir && *dir && *subdir);
2040
2041 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
2042 dir++;
2043 subdir++;
2044 offset++;
2045 }
2046
2047 /* hel[p]/me vs hel[l]/yeah */
2048 if (*dir && *subdir)
2049 return -1;
2050
2051 if (!*subdir)
2052 return !*dir ? offset : -1; /* same dir */
2053
2054 /* foo/[b]ar vs foo/[] */
2055 if (is_dir_sep(dir[-1]))
2056 return is_dir_sep(subdir[-1]) ? offset : -1;
2057
2058 /* foo[/]bar vs foo[] */
2059 return is_dir_sep(*subdir) ? offset + 1 : -1;
2060 }
2061
2062 int is_inside_dir(const char *dir)
2063 {
2064 char *cwd;
2065 int rc;
2066
2067 if (!dir)
2068 return 0;
2069
2070 cwd = xgetcwd();
2071 rc = (dir_inside_of(cwd, dir) >= 0);
2072 free(cwd);
2073 return rc;
2074 }
2075
2076 int is_empty_dir(const char *path)
2077 {
2078 DIR *dir = opendir(path);
2079 struct dirent *e;
2080 int ret = 1;
2081
2082 if (!dir)
2083 return 0;
2084
2085 while ((e = readdir(dir)) != NULL)
2086 if (!is_dot_or_dotdot(e->d_name)) {
2087 ret = 0;
2088 break;
2089 }
2090
2091 closedir(dir);
2092 return ret;
2093 }
2094
2095 static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
2096 {
2097 DIR *dir;
2098 struct dirent *e;
2099 int ret = 0, original_len = path->len, len, kept_down = 0;
2100 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
2101 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
2102 unsigned char submodule_head[20];
2103
2104 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
2105 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
2106 /* Do not descend and nuke a nested git work tree. */
2107 if (kept_up)
2108 *kept_up = 1;
2109 return 0;
2110 }
2111
2112 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
2113 dir = opendir(path->buf);
2114 if (!dir) {
2115 if (errno == ENOENT)
2116 return keep_toplevel ? -1 : 0;
2117 else if (errno == EACCES && !keep_toplevel)
2118 /*
2119 * An empty dir could be removable even if it
2120 * is unreadable:
2121 */
2122 return rmdir(path->buf);
2123 else
2124 return -1;
2125 }
2126 strbuf_complete(path, '/');
2127
2128 len = path->len;
2129 while ((e = readdir(dir)) != NULL) {
2130 struct stat st;
2131 if (is_dot_or_dotdot(e->d_name))
2132 continue;
2133
2134 strbuf_setlen(path, len);
2135 strbuf_addstr(path, e->d_name);
2136 if (lstat(path->buf, &st)) {
2137 if (errno == ENOENT)
2138 /*
2139 * file disappeared, which is what we
2140 * wanted anyway
2141 */
2142 continue;
2143 /* fall thru */
2144 } else if (S_ISDIR(st.st_mode)) {
2145 if (!remove_dir_recurse(path, flag, &kept_down))
2146 continue; /* happy */
2147 } else if (!only_empty &&
2148 (!unlink(path->buf) || errno == ENOENT)) {
2149 continue; /* happy, too */
2150 }
2151
2152 /* path too long, stat fails, or non-directory still exists */
2153 ret = -1;
2154 break;
2155 }
2156 closedir(dir);
2157
2158 strbuf_setlen(path, original_len);
2159 if (!ret && !keep_toplevel && !kept_down)
2160 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
2161 else if (kept_up)
2162 /*
2163 * report the uplevel that it is not an error that we
2164 * did not rmdir() our directory.
2165 */
2166 *kept_up = !ret;
2167 return ret;
2168 }
2169
2170 int remove_dir_recursively(struct strbuf *path, int flag)
2171 {
2172 return remove_dir_recurse(path, flag, NULL);
2173 }
2174
2175 static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2176
2177 void setup_standard_excludes(struct dir_struct *dir)
2178 {
2179 const char *path;
2180
2181 dir->exclude_per_dir = ".gitignore";
2182
2183 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2184 if (!excludes_file)
2185 excludes_file = xdg_config_home("ignore");
2186 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
2187 add_excludes_from_file_1(dir, excludes_file,
2188 dir->untracked ? &dir->ss_excludes_file : NULL);
2189
2190 /* per repository user preference */
2191 path = git_path_info_exclude();
2192 if (!access_or_warn(path, R_OK, 0))
2193 add_excludes_from_file_1(dir, path,
2194 dir->untracked ? &dir->ss_info_exclude : NULL);
2195 }
2196
2197 int remove_path(const char *name)
2198 {
2199 char *slash;
2200
2201 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
2202 return -1;
2203
2204 slash = strrchr(name, '/');
2205 if (slash) {
2206 char *dirs = xstrdup(name);
2207 slash = dirs + (slash - name);
2208 do {
2209 *slash = '\0';
2210 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
2211 free(dirs);
2212 }
2213 return 0;
2214 }
2215
2216 /*
2217 * Frees memory within dir which was allocated for exclude lists and
2218 * the exclude_stack. Does not free dir itself.
2219 */
2220 void clear_directory(struct dir_struct *dir)
2221 {
2222 int i, j;
2223 struct exclude_list_group *group;
2224 struct exclude_list *el;
2225 struct exclude_stack *stk;
2226
2227 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2228 group = &dir->exclude_list_group[i];
2229 for (j = 0; j < group->nr; j++) {
2230 el = &group->el[j];
2231 if (i == EXC_DIRS)
2232 free((char *)el->src);
2233 clear_exclude_list(el);
2234 }
2235 free(group->el);
2236 }
2237
2238 stk = dir->exclude_stack;
2239 while (stk) {
2240 struct exclude_stack *prev = stk->prev;
2241 free(stk);
2242 stk = prev;
2243 }
2244 strbuf_release(&dir->basebuf);
2245 }
2246
2247 struct ondisk_untracked_cache {
2248 struct stat_data info_exclude_stat;
2249 struct stat_data excludes_file_stat;
2250 uint32_t dir_flags;
2251 unsigned char info_exclude_sha1[20];
2252 unsigned char excludes_file_sha1[20];
2253 char exclude_per_dir[FLEX_ARRAY];
2254 };
2255
2256 #define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2257
2258 struct write_data {
2259 int index; /* number of written untracked_cache_dir */
2260 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2261 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2262 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2263 struct strbuf out;
2264 struct strbuf sb_stat;
2265 struct strbuf sb_sha1;
2266 };
2267
2268 static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2269 {
2270 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2271 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2272 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2273 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2274 to->sd_dev = htonl(from->sd_dev);
2275 to->sd_ino = htonl(from->sd_ino);
2276 to->sd_uid = htonl(from->sd_uid);
2277 to->sd_gid = htonl(from->sd_gid);
2278 to->sd_size = htonl(from->sd_size);
2279 }
2280
2281 static void write_one_dir(struct untracked_cache_dir *untracked,
2282 struct write_data *wd)
2283 {
2284 struct stat_data stat_data;
2285 struct strbuf *out = &wd->out;
2286 unsigned char intbuf[16];
2287 unsigned int intlen, value;
2288 int i = wd->index++;
2289
2290 /*
2291 * untracked_nr should be reset whenever valid is clear, but
2292 * for safety..
2293 */
2294 if (!untracked->valid) {
2295 untracked->untracked_nr = 0;
2296 untracked->check_only = 0;
2297 }
2298
2299 if (untracked->check_only)
2300 ewah_set(wd->check_only, i);
2301 if (untracked->valid) {
2302 ewah_set(wd->valid, i);
2303 stat_data_to_disk(&stat_data, &untracked->stat_data);
2304 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2305 }
2306 if (!is_null_sha1(untracked->exclude_sha1)) {
2307 ewah_set(wd->sha1_valid, i);
2308 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2309 }
2310
2311 intlen = encode_varint(untracked->untracked_nr, intbuf);
2312 strbuf_add(out, intbuf, intlen);
2313
2314 /* skip non-recurse directories */
2315 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2316 if (untracked->dirs[i]->recurse)
2317 value++;
2318 intlen = encode_varint(value, intbuf);
2319 strbuf_add(out, intbuf, intlen);
2320
2321 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2322
2323 for (i = 0; i < untracked->untracked_nr; i++)
2324 strbuf_add(out, untracked->untracked[i],
2325 strlen(untracked->untracked[i]) + 1);
2326
2327 for (i = 0; i < untracked->dirs_nr; i++)
2328 if (untracked->dirs[i]->recurse)
2329 write_one_dir(untracked->dirs[i], wd);
2330 }
2331
2332 void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2333 {
2334 struct ondisk_untracked_cache *ouc;
2335 struct write_data wd;
2336 unsigned char varbuf[16];
2337 int len = 0, varint_len;
2338 if (untracked->exclude_per_dir)
2339 len = strlen(untracked->exclude_per_dir);
2340 ouc = xmalloc(sizeof(*ouc) + len + 1);
2341 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2342 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2343 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2344 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2345 ouc->dir_flags = htonl(untracked->dir_flags);
2346 memcpy(ouc->exclude_per_dir, untracked->exclude_per_dir, len + 1);
2347
2348 varint_len = encode_varint(untracked->ident.len, varbuf);
2349 strbuf_add(out, varbuf, varint_len);
2350 strbuf_add(out, untracked->ident.buf, untracked->ident.len);
2351
2352 strbuf_add(out, ouc, ouc_size(len));
2353 free(ouc);
2354 ouc = NULL;
2355
2356 if (!untracked->root) {
2357 varint_len = encode_varint(0, varbuf);
2358 strbuf_add(out, varbuf, varint_len);
2359 return;
2360 }
2361
2362 wd.index = 0;
2363 wd.check_only = ewah_new();
2364 wd.valid = ewah_new();
2365 wd.sha1_valid = ewah_new();
2366 strbuf_init(&wd.out, 1024);
2367 strbuf_init(&wd.sb_stat, 1024);
2368 strbuf_init(&wd.sb_sha1, 1024);
2369 write_one_dir(untracked->root, &wd);
2370
2371 varint_len = encode_varint(wd.index, varbuf);
2372 strbuf_add(out, varbuf, varint_len);
2373 strbuf_addbuf(out, &wd.out);
2374 ewah_serialize_strbuf(wd.valid, out);
2375 ewah_serialize_strbuf(wd.check_only, out);
2376 ewah_serialize_strbuf(wd.sha1_valid, out);
2377 strbuf_addbuf(out, &wd.sb_stat);
2378 strbuf_addbuf(out, &wd.sb_sha1);
2379 strbuf_addch(out, '\0'); /* safe guard for string lists */
2380
2381 ewah_free(wd.valid);
2382 ewah_free(wd.check_only);
2383 ewah_free(wd.sha1_valid);
2384 strbuf_release(&wd.out);
2385 strbuf_release(&wd.sb_stat);
2386 strbuf_release(&wd.sb_sha1);
2387 }
2388
2389 static void free_untracked(struct untracked_cache_dir *ucd)
2390 {
2391 int i;
2392 if (!ucd)
2393 return;
2394 for (i = 0; i < ucd->dirs_nr; i++)
2395 free_untracked(ucd->dirs[i]);
2396 for (i = 0; i < ucd->untracked_nr; i++)
2397 free(ucd->untracked[i]);
2398 free(ucd->untracked);
2399 free(ucd->dirs);
2400 free(ucd);
2401 }
2402
2403 void free_untracked_cache(struct untracked_cache *uc)
2404 {
2405 if (uc)
2406 free_untracked(uc->root);
2407 free(uc);
2408 }
2409
2410 struct read_data {
2411 int index;
2412 struct untracked_cache_dir **ucd;
2413 struct ewah_bitmap *check_only;
2414 struct ewah_bitmap *valid;
2415 struct ewah_bitmap *sha1_valid;
2416 const unsigned char *data;
2417 const unsigned char *end;
2418 };
2419
2420 static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2421 {
2422 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2423 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2424 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2425 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2426 to->sd_dev = get_be32(&from->sd_dev);
2427 to->sd_ino = get_be32(&from->sd_ino);
2428 to->sd_uid = get_be32(&from->sd_uid);
2429 to->sd_gid = get_be32(&from->sd_gid);
2430 to->sd_size = get_be32(&from->sd_size);
2431 }
2432
2433 static int read_one_dir(struct untracked_cache_dir **untracked_,
2434 struct read_data *rd)
2435 {
2436 struct untracked_cache_dir ud, *untracked;
2437 const unsigned char *next, *data = rd->data, *end = rd->end;
2438 unsigned int value;
2439 int i, len;
2440
2441 memset(&ud, 0, sizeof(ud));
2442
2443 next = data;
2444 value = decode_varint(&next);
2445 if (next > end)
2446 return -1;
2447 ud.recurse = 1;
2448 ud.untracked_alloc = value;
2449 ud.untracked_nr = value;
2450 if (ud.untracked_nr)
2451 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
2452 data = next;
2453
2454 next = data;
2455 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2456 if (next > end)
2457 return -1;
2458 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
2459 data = next;
2460
2461 len = strlen((const char *)data);
2462 next = data + len + 1;
2463 if (next > rd->end)
2464 return -1;
2465 *untracked_ = untracked = xmalloc(sizeof(*untracked) + len);
2466 memcpy(untracked, &ud, sizeof(ud));
2467 memcpy(untracked->name, data, len + 1);
2468 data = next;
2469
2470 for (i = 0; i < untracked->untracked_nr; i++) {
2471 len = strlen((const char *)data);
2472 next = data + len + 1;
2473 if (next > rd->end)
2474 return -1;
2475 untracked->untracked[i] = xstrdup((const char*)data);
2476 data = next;
2477 }
2478
2479 rd->ucd[rd->index++] = untracked;
2480 rd->data = data;
2481
2482 for (i = 0; i < untracked->dirs_nr; i++) {
2483 len = read_one_dir(untracked->dirs + i, rd);
2484 if (len < 0)
2485 return -1;
2486 }
2487 return 0;
2488 }
2489
2490 static void set_check_only(size_t pos, void *cb)
2491 {
2492 struct read_data *rd = cb;
2493 struct untracked_cache_dir *ud = rd->ucd[pos];
2494 ud->check_only = 1;
2495 }
2496
2497 static void read_stat(size_t pos, void *cb)
2498 {
2499 struct read_data *rd = cb;
2500 struct untracked_cache_dir *ud = rd->ucd[pos];
2501 if (rd->data + sizeof(struct stat_data) > rd->end) {
2502 rd->data = rd->end + 1;
2503 return;
2504 }
2505 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2506 rd->data += sizeof(struct stat_data);
2507 ud->valid = 1;
2508 }
2509
2510 static void read_sha1(size_t pos, void *cb)
2511 {
2512 struct read_data *rd = cb;
2513 struct untracked_cache_dir *ud = rd->ucd[pos];
2514 if (rd->data + 20 > rd->end) {
2515 rd->data = rd->end + 1;
2516 return;
2517 }
2518 hashcpy(ud->exclude_sha1, rd->data);
2519 rd->data += 20;
2520 }
2521
2522 static void load_sha1_stat(struct sha1_stat *sha1_stat,
2523 const struct stat_data *stat,
2524 const unsigned char *sha1)
2525 {
2526 stat_data_from_disk(&sha1_stat->stat, stat);
2527 hashcpy(sha1_stat->sha1, sha1);
2528 sha1_stat->valid = 1;
2529 }
2530
2531 struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2532 {
2533 const struct ondisk_untracked_cache *ouc;
2534 struct untracked_cache *uc;
2535 struct read_data rd;
2536 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
2537 const char *ident;
2538 int ident_len, len;
2539
2540 if (sz <= 1 || end[-1] != '\0')
2541 return NULL;
2542 end--;
2543
2544 ident_len = decode_varint(&next);
2545 if (next + ident_len > end)
2546 return NULL;
2547 ident = (const char *)next;
2548 next += ident_len;
2549
2550 ouc = (const struct ondisk_untracked_cache *)next;
2551 if (next + ouc_size(0) > end)
2552 return NULL;
2553
2554 uc = xcalloc(1, sizeof(*uc));
2555 strbuf_init(&uc->ident, ident_len);
2556 strbuf_add(&uc->ident, ident, ident_len);
2557 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2558 ouc->info_exclude_sha1);
2559 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2560 ouc->excludes_file_sha1);
2561 uc->dir_flags = get_be32(&ouc->dir_flags);
2562 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2563 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2564 next += ouc_size(strlen(ouc->exclude_per_dir));
2565 if (next >= end)
2566 goto done2;
2567
2568 len = decode_varint(&next);
2569 if (next > end || len == 0)
2570 goto done2;
2571
2572 rd.valid = ewah_new();
2573 rd.check_only = ewah_new();
2574 rd.sha1_valid = ewah_new();
2575 rd.data = next;
2576 rd.end = end;
2577 rd.index = 0;
2578 ALLOC_ARRAY(rd.ucd, len);
2579
2580 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2581 goto done;
2582
2583 next = rd.data;
2584 len = ewah_read_mmap(rd.valid, next, end - next);
2585 if (len < 0)
2586 goto done;
2587
2588 next += len;
2589 len = ewah_read_mmap(rd.check_only, next, end - next);
2590 if (len < 0)
2591 goto done;
2592
2593 next += len;
2594 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2595 if (len < 0)
2596 goto done;
2597
2598 ewah_each_bit(rd.check_only, set_check_only, &rd);
2599 rd.data = next + len;
2600 ewah_each_bit(rd.valid, read_stat, &rd);
2601 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2602 next = rd.data;
2603
2604 done:
2605 free(rd.ucd);
2606 ewah_free(rd.valid);
2607 ewah_free(rd.check_only);
2608 ewah_free(rd.sha1_valid);
2609 done2:
2610 if (next != end) {
2611 free_untracked_cache(uc);
2612 uc = NULL;
2613 }
2614 return uc;
2615 }
2616
2617 static void invalidate_one_directory(struct untracked_cache *uc,
2618 struct untracked_cache_dir *ucd)
2619 {
2620 uc->dir_invalidated++;
2621 ucd->valid = 0;
2622 ucd->untracked_nr = 0;
2623 }
2624
2625 /*
2626 * Normally when an entry is added or removed from a directory,
2627 * invalidating that directory is enough. No need to touch its
2628 * ancestors. When a directory is shown as "foo/bar/" in git-status
2629 * however, deleting or adding an entry may have cascading effect.
2630 *
2631 * Say the "foo/bar/file" has become untracked, we need to tell the
2632 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2633 * directory any more (because "bar" is managed by foo as an untracked
2634 * "file").
2635 *
2636 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2637 * was the last untracked entry in the entire "foo", we should show
2638 * "foo/" instead. Which means we have to invalidate past "bar" up to
2639 * "foo".
2640 *
2641 * This function traverses all directories from root to leaf. If there
2642 * is a chance of one of the above cases happening, we invalidate back
2643 * to root. Otherwise we just invalidate the leaf. There may be a more
2644 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2645 * detect these cases and avoid unnecessary invalidation, for example,
2646 * checking for the untracked entry named "bar/" in "foo", but for now
2647 * stick to something safe and simple.
2648 */
2649 static int invalidate_one_component(struct untracked_cache *uc,
2650 struct untracked_cache_dir *dir,
2651 const char *path, int len)
2652 {
2653 const char *rest = strchr(path, '/');
2654
2655 if (rest) {
2656 int component_len = rest - path;
2657 struct untracked_cache_dir *d =
2658 lookup_untracked(uc, dir, path, component_len);
2659 int ret =
2660 invalidate_one_component(uc, d, rest + 1,
2661 len - (component_len + 1));
2662 if (ret)
2663 invalidate_one_directory(uc, dir);
2664 return ret;
2665 }
2666
2667 invalidate_one_directory(uc, dir);
2668 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2669 }
2670
2671 void untracked_cache_invalidate_path(struct index_state *istate,
2672 const char *path)
2673 {
2674 if (!istate->untracked || !istate->untracked->root)
2675 return;
2676 invalidate_one_component(istate->untracked, istate->untracked->root,
2677 path, strlen(path));
2678 }
2679
2680 void untracked_cache_remove_from_index(struct index_state *istate,
2681 const char *path)
2682 {
2683 untracked_cache_invalidate_path(istate, path);
2684 }
2685
2686 void untracked_cache_add_to_index(struct index_state *istate,
2687 const char *path)
2688 {
2689 untracked_cache_invalidate_path(istate, path);
2690 }