fsmonitor: teach git to optionally utilize a file system monitor to speed up detectin...
[git/git.git] / dir.c
CommitLineData
453ec4bd
LT
1/*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
95a68344
AS
5 * See Documentation/technical/api-directory-listing.txt
6 *
453ec4bd
LT
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
12cd0bf9 10#define NO_THE_INDEX_COMPATIBILITY_MACROS
453ec4bd 11#include "cache.h"
b2141fc1 12#include "config.h"
453ec4bd 13#include "dir.h"
b0db7046 14#include "attr.h"
09595258 15#include "refs.h"
237ec6e4 16#include "wildmatch.h"
64acde94 17#include "pathspec.h"
dde843e7 18#include "utf8.h"
83c094ad
NTND
19#include "varint.h"
20#include "ewah/ewok.h"
883e248b 21#include "fsmonitor.h"
453ec4bd 22
defd7c7b
KB
23/*
24 * Tells read_directory_recursive how a file or directory should be treated.
25 * Values are ordered by significance, e.g. if a directory contains both
26 * excluded and untracked files, it is listed as untracked because
27 * path_untracked > path_excluded.
28 */
29enum path_treatment {
30 path_none = 0,
31 path_recurse,
32 path_excluded,
33 path_untracked
34};
35
cf7c6148
NTND
36/*
37 * Support data structure for our opendir/readdir/closedir wrappers
38 */
39struct cached_dir {
40 DIR *fdir;
41 struct untracked_cache_dir *untracked;
91a2288b
NTND
42 int nr_files;
43 int nr_dirs;
44
cf7c6148 45 struct dirent *de;
91a2288b
NTND
46 const char *file;
47 struct untracked_cache_dir *ucd;
cf7c6148
NTND
48};
49
defd7c7b 50static enum path_treatment read_directory_recursive(struct dir_struct *dir,
0ef8e169
BW
51 struct index_state *istate, const char *path, int len,
52 struct untracked_cache_dir *untracked,
e1b8c7bd 53 int check_only, const struct pathspec *pathspec);
98f2a687
BW
54static int get_dtype(struct dirent *de, struct index_state *istate,
55 const char *path, int len);
09595258 56
e0556a92
PC
57int count_slashes(const char *s)
58{
59 int cnt = 0;
60 while (*s)
61 if (*s++ == '/')
62 cnt++;
63 return cnt;
64}
65
ba0897e6 66int fspathcmp(const char *a, const char *b)
8cf2a84e
JJ
67{
68 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
69}
70
ba0897e6 71int fspathncmp(const char *a, const char *b, size_t count)
8cf2a84e
JJ
72{
73 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
74}
75
1f26ce61
CB
76int git_fnmatch(const struct pathspec_item *item,
77 const char *pattern, const char *string,
78 int prefix)
5d74762d 79{
5d74762d 80 if (prefix > 0) {
93d93537 81 if (ps_strncmp(item, pattern, string, prefix))
eb07894f 82 return WM_NOMATCH;
5d74762d
NTND
83 pattern += prefix;
84 string += prefix;
85 }
bd30c2e4 86 if (item->flags & PATHSPEC_ONESTAR) {
8c6abbcd
NTND
87 int pattern_len = strlen(++pattern);
88 int string_len = strlen(string);
89 return string_len < pattern_len ||
93d93537
NTND
90 ps_strcmp(item, pattern,
91 string + string_len - pattern_len);
8c6abbcd 92 }
bd30c2e4 93 if (item->magic & PATHSPEC_GLOB)
93d93537
NTND
94 return wildmatch(pattern, string,
95 WM_PATHNAME |
55d34269 96 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
bd30c2e4
NTND
97 else
98 /* wildmatch has not learned no FNM_PATHNAME mode yet */
eb07894f 99 return wildmatch(pattern, string,
55d34269 100 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
5d74762d
NTND
101}
102
0b6e56df
JH
103static 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
f30366b2
JH
122 if (ignore_case)
123 flags |= WM_CASEFOLD;
55d34269 124 match_status = wildmatch(use_pat, use_str, flags);
0b6e56df
JH
125
126 strbuf_release(&pat_buf);
127 strbuf_release(&str_buf);
128
129 return match_status;
130}
131
827f4d6c 132static size_t common_prefix_len(const struct pathspec *pathspec)
3c6a370b 133{
827f4d6c 134 int n;
4a085b16 135 size_t max = 0;
3c6a370b 136
93d93537
NTND
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 */
5c6933d2
NTND
144 GUARD_PATHSPEC(pathspec,
145 PATHSPEC_FROMTOP |
146 PATHSPEC_MAXDEPTH |
bd30c2e4 147 PATHSPEC_LITERAL |
93d93537 148 PATHSPEC_GLOB |
ef79b1f8 149 PATHSPEC_ICASE |
b0db7046
BW
150 PATHSPEC_EXCLUDE |
151 PATHSPEC_ATTR);
4a085b16 152
827f4d6c 153 for (n = 0; n < pathspec->nr; n++) {
93d93537 154 size_t i = 0, len = 0, item_len;
ef79b1f8
NTND
155 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
156 continue;
93d93537
NTND
157 if (pathspec->items[n].magic & PATHSPEC_ICASE)
158 item_len = pathspec->items[n].prefix;
159 else
160 item_len = pathspec->items[n].nowildcard_len;
161 while (i < item_len && (n == 0 || i < max)) {
827f4d6c
NTND
162 char c = pathspec->items[n].match[i];
163 if (c != pathspec->items[0].match[i])
4a085b16
JH
164 break;
165 if (c == '/')
166 len = i + 1;
827f4d6c 167 i++;
4a085b16 168 }
827f4d6c 169 if (n == 0 || len < max) {
4a085b16
JH
170 max = len;
171 if (!max)
172 break;
173 }
3c6a370b 174 }
4a085b16 175 return max;
3c6a370b
LT
176}
177
f950eb95
CB
178/*
179 * Returns a copy of the longest leading path common among all
180 * pathspecs.
181 */
827f4d6c 182char *common_prefix(const struct pathspec *pathspec)
f950eb95
CB
183{
184 unsigned long len = common_prefix_len(pathspec);
185
827f4d6c 186 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
f950eb95
CB
187}
188
0d32c183
BW
189int fill_directory(struct dir_struct *dir,
190 struct index_state *istate,
191 const struct pathspec *pathspec)
1d8842d9 192{
bec5ab89 193 const char *prefix;
966de302 194 size_t prefix_len;
1d8842d9
LT
195
196 /*
197 * Calculate common prefix for the pathspec, and
198 * use that to optimize the directory walk
199 */
bec5ab89
RS
200 prefix_len = common_prefix_len(pathspec);
201 prefix = prefix_len ? pathspec->items[0].match : "";
1d8842d9
LT
202
203 /* Read the directory and prune it */
0d32c183 204 read_directory(dir, istate, prefix, prefix_len, pathspec);
966de302 205
966de302 206 return prefix_len;
1d8842d9
LT
207}
208
bc96cc87
NTND
209int within_depth(const char *name, int namelen,
210 int depth, int max_depth)
211{
212 const char *cp = name, *cpe = name + namelen;
213
214 while (cp < cpe) {
215 if (*cp++ != '/')
216 continue;
217 depth++;
218 if (depth > max_depth)
219 return 0;
220 }
221 return 1;
222}
223
75a6315f
BW
224#define DO_MATCH_EXCLUDE (1<<0)
225#define DO_MATCH_DIRECTORY (1<<1)
226#define DO_MATCH_SUBMODULE (1<<2)
42b0874a 227
b0db7046
BW
228static int match_attrs(const char *name, int namelen,
229 const struct pathspec_item *item)
230{
231 int i;
232
233 git_check_attr(name, item->attr_check);
234 for (i = 0; i < item->attr_match_nr; i++) {
235 const char *value;
236 int matched;
237 enum attr_match_mode match_mode;
238
239 value = item->attr_check->items[i].value;
240 match_mode = item->attr_match[i].match_mode;
241
242 if (ATTR_TRUE(value))
243 matched = (match_mode == MATCH_SET);
244 else if (ATTR_FALSE(value))
245 matched = (match_mode == MATCH_UNSET);
246 else if (ATTR_UNSET(value))
247 matched = (match_mode == MATCH_UNSPECIFIED);
248 else
249 matched = (match_mode == MATCH_VALUE &&
250 !strcmp(item->attr_match[i].value, value));
251 if (!matched)
252 return 0;
253 }
254
255 return 1;
256}
257
61cf2820
NTND
258/*
259 * Does 'match' match the given name?
260 * A match is found if
261 *
262 * (1) the 'match' string is leading directory of 'name', or
263 * (2) the 'match' string is a wildcard and matches 'name', or
264 * (3) the 'match' string is exactly the same as 'name'.
265 *
266 * and the return value tells which case it was.
267 *
268 * It returns 0 when there is no match.
269 */
270static int match_pathspec_item(const struct pathspec_item *item, int prefix,
42b0874a 271 const char *name, int namelen, unsigned flags)
61cf2820
NTND
272{
273 /* name/namelen has prefix cut off by caller */
274 const char *match = item->match + prefix;
275 int matchlen = item->len - prefix;
276
93d93537
NTND
277 /*
278 * The normal call pattern is:
279 * 1. prefix = common_prefix_len(ps);
280 * 2. prune something, or fill_directory
854b0959 281 * 3. match_pathspec()
93d93537
NTND
282 *
283 * 'prefix' at #1 may be shorter than the command's prefix and
284 * it's ok for #2 to match extra files. Those extras will be
285 * trimmed at #3.
286 *
287 * Suppose the pathspec is 'foo' and '../bar' running from
288 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
289 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
290 * user does not want XYZ/foo, only the "foo" part should be
291 * case-insensitive. We need to filter out XYZ/foo here. In
292 * other words, we do not trust the caller on comparing the
293 * prefix part when :(icase) is involved. We do exact
294 * comparison ourselves.
295 *
296 * Normally the caller (common_prefix_len() in fact) does
297 * _exact_ matching on name[-prefix+1..-1] and we do not need
298 * to check that part. Be defensive and check it anyway, in
299 * case common_prefix_len is changed, or a new caller is
300 * introduced that does not use common_prefix_len.
301 *
302 * If the penalty turns out too high when prefix is really
303 * long, maybe change it to
304 * strncmp(match, name, item->prefix - prefix)
305 */
306 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
307 strncmp(item->match, name - prefix, item->prefix))
308 return 0;
309
b0db7046
BW
310 if (item->attr_match_nr && !match_attrs(name, namelen, item))
311 return 0;
312
61cf2820
NTND
313 /* If the match was just the prefix, we matched */
314 if (!*match)
315 return MATCHED_RECURSIVELY;
316
93d93537 317 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
61cf2820
NTND
318 if (matchlen == namelen)
319 return MATCHED_EXACTLY;
320
321 if (match[matchlen-1] == '/' || name[matchlen] == '/')
322 return MATCHED_RECURSIVELY;
68690fdd
NTND
323 } else if ((flags & DO_MATCH_DIRECTORY) &&
324 match[matchlen - 1] == '/' &&
325 namelen == matchlen - 1 &&
326 !ps_strncmp(item, match, name, namelen))
327 return MATCHED_EXACTLY;
61cf2820 328
5d74762d 329 if (item->nowildcard_len < item->len &&
bd30c2e4 330 !git_fnmatch(item, match, name,
8c6abbcd 331 item->nowildcard_len - prefix))
61cf2820
NTND
332 return MATCHED_FNMATCH;
333
75a6315f
BW
334 /* Perform checks to see if "name" is a super set of the pathspec */
335 if (flags & DO_MATCH_SUBMODULE) {
336 /* name is a literal prefix of the pathspec */
337 if ((namelen < matchlen) &&
338 (match[namelen] == '/') &&
339 !ps_strncmp(item, match, name, namelen))
340 return MATCHED_RECURSIVELY;
341
342 /* name" doesn't match up to the first wild character */
343 if (item->nowildcard_len < item->len &&
344 ps_strncmp(item, match, name,
345 item->nowildcard_len - prefix))
346 return 0;
347
348 /*
349 * Here is where we would perform a wildmatch to check if
350 * "name" can be matched as a directory (or a prefix) against
351 * the pathspec. Since wildmatch doesn't have this capability
352 * at the present we have to punt and say that it is a match,
353 * potentially returning a false positive
354 * The submodules themselves will be able to perform more
355 * accurate matching to determine if the pathspec matches.
356 */
357 return MATCHED_RECURSIVELY;
358 }
359
61cf2820
NTND
360 return 0;
361}
362
363/*
52ed1894
AS
364 * Given a name and a list of pathspecs, returns the nature of the
365 * closest (i.e. most specific) match of the name to any of the
366 * pathspecs.
367 *
368 * The caller typically calls this multiple times with the same
369 * pathspec and seen[] array but with different name/namelen
370 * (e.g. entries from the index) and is interested in seeing if and
371 * how each pathspec matches all the names it calls this function
372 * with. A mark is left in the seen[] array for each pathspec element
373 * indicating the closest type of match that element achieved, so if
374 * seen[n] remains zero after multiple invocations, that means the nth
375 * pathspec did not match any names, which could indicate that the
376 * user mistyped the nth pathspec.
61cf2820 377 */
854b0959
NTND
378static int do_match_pathspec(const struct pathspec *ps,
379 const char *name, int namelen,
380 int prefix, char *seen,
42b0874a 381 unsigned flags)
61cf2820 382{
42b0874a 383 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
61cf2820 384
5c6933d2
NTND
385 GUARD_PATHSPEC(ps,
386 PATHSPEC_FROMTOP |
387 PATHSPEC_MAXDEPTH |
bd30c2e4 388 PATHSPEC_LITERAL |
93d93537 389 PATHSPEC_GLOB |
ef79b1f8 390 PATHSPEC_ICASE |
b0db7046
BW
391 PATHSPEC_EXCLUDE |
392 PATHSPEC_ATTR);
8f4f8f45 393
61cf2820 394 if (!ps->nr) {
6330a171
NTND
395 if (!ps->recursive ||
396 !(ps->magic & PATHSPEC_MAXDEPTH) ||
397 ps->max_depth == -1)
61cf2820
NTND
398 return MATCHED_RECURSIVELY;
399
400 if (within_depth(name, namelen, 0, ps->max_depth))
401 return MATCHED_EXACTLY;
402 else
403 return 0;
404 }
405
406 name += prefix;
407 namelen -= prefix;
408
409 for (i = ps->nr - 1; i >= 0; i--) {
410 int how;
ef79b1f8
NTND
411
412 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
413 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
414 continue;
415
61cf2820
NTND
416 if (seen && seen[i] == MATCHED_EXACTLY)
417 continue;
ef79b1f8
NTND
418 /*
419 * Make exclude patterns optional and never report
420 * "pathspec ':(exclude)foo' matches no files"
421 */
422 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
423 seen[i] = MATCHED_FNMATCH;
42b0874a
NTND
424 how = match_pathspec_item(ps->items+i, prefix, name,
425 namelen, flags);
6330a171
NTND
426 if (ps->recursive &&
427 (ps->magic & PATHSPEC_MAXDEPTH) &&
428 ps->max_depth != -1 &&
61cf2820
NTND
429 how && how != MATCHED_FNMATCH) {
430 int len = ps->items[i].len;
431 if (name[len] == '/')
432 len++;
433 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
434 how = MATCHED_EXACTLY;
435 else
436 how = 0;
437 }
438 if (how) {
439 if (retval < how)
440 retval = how;
441 if (seen && seen[i] < how)
442 seen[i] = how;
443 }
444 }
445 return retval;
446}
447
854b0959
NTND
448int match_pathspec(const struct pathspec *ps,
449 const char *name, int namelen,
ae8d0824 450 int prefix, char *seen, int is_dir)
ef79b1f8
NTND
451{
452 int positive, negative;
ae8d0824 453 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
42b0874a
NTND
454 positive = do_match_pathspec(ps, name, namelen,
455 prefix, seen, flags);
ef79b1f8
NTND
456 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
457 return positive;
42b0874a
NTND
458 negative = do_match_pathspec(ps, name, namelen,
459 prefix, seen,
460 flags | DO_MATCH_EXCLUDE);
ef79b1f8
NTND
461 return negative ? 0 : positive;
462}
463
75a6315f
BW
464/**
465 * Check if a submodule is a superset of the pathspec
466 */
467int submodule_path_match(const struct pathspec *ps,
468 const char *submodule_name,
469 char *seen)
470{
471 int matched = do_match_pathspec(ps, submodule_name,
472 strlen(submodule_name),
473 0, seen,
474 DO_MATCH_DIRECTORY |
475 DO_MATCH_SUBMODULE);
476 return matched;
477}
478
777c55a6
JH
479int report_path_error(const char *ps_matched,
480 const struct pathspec *pathspec,
481 const char *prefix)
482{
483 /*
484 * Make sure all pathspec matched; otherwise it is an error.
485 */
777c55a6
JH
486 int num, errors = 0;
487 for (num = 0; num < pathspec->nr; num++) {
488 int other, found_dup;
489
490 if (ps_matched[num])
491 continue;
492 /*
493 * The caller might have fed identical pathspec
494 * twice. Do not barf on such a mistake.
495 * FIXME: parse_pathspec should have eliminated
496 * duplicate pathspec.
497 */
498 for (found_dup = other = 0;
499 !found_dup && other < pathspec->nr;
500 other++) {
501 if (other == num || !ps_matched[other])
502 continue;
503 if (!strcmp(pathspec->items[other].original,
504 pathspec->items[num].original))
505 /*
506 * Ok, we have a match already.
507 */
508 found_dup = 1;
509 }
510 if (found_dup)
511 continue;
512
513 error("pathspec '%s' did not match any file(s) known to git.",
514 pathspec->items[num].original);
515 errors++;
516 }
777c55a6
JH
517 return errors;
518}
519
fcd631ed
NTND
520/*
521 * Return the length of the "simple" part of a path match limiter.
522 */
87323bda 523int simple_length(const char *match)
fcd631ed
NTND
524{
525 int len = -1;
526
527 for (;;) {
528 unsigned char c = *match++;
529 len++;
530 if (c == '\0' || is_glob_special(c))
531 return len;
532 }
533}
534
87323bda 535int no_wildcard(const char *string)
68492fc7 536{
fcd631ed 537 return string[simple_length(string)] == '\0';
68492fc7
LK
538}
539
82dce998
NTND
540void parse_exclude_pattern(const char **pattern,
541 int *patternlen,
f8708998 542 unsigned *flags,
82dce998 543 int *nowildcardlen)
84460eec
NTND
544{
545 const char *p = *pattern;
546 size_t i, len;
547
548 *flags = 0;
549 if (*p == '!') {
550 *flags |= EXC_FLAG_NEGATIVE;
551 p++;
552 }
553 len = strlen(p);
554 if (len && p[len - 1] == '/') {
555 len--;
556 *flags |= EXC_FLAG_MUSTBEDIR;
557 }
558 for (i = 0; i < len; i++) {
559 if (p[i] == '/')
560 break;
561 }
562 if (i == len)
563 *flags |= EXC_FLAG_NODIR;
564 *nowildcardlen = simple_length(p);
565 /*
566 * we should have excluded the trailing slash from 'p' too,
567 * but that's one more allocation. Instead just make sure
568 * nowildcardlen does not exceed real patternlen
569 */
570 if (*nowildcardlen > len)
571 *nowildcardlen = len;
572 if (*p == '*' && no_wildcard(p + 1))
573 *flags |= EXC_FLAG_ENDSWITH;
574 *pattern = p;
575 *patternlen = len;
576}
577
453ec4bd 578void add_exclude(const char *string, const char *base,
c04318e4 579 int baselen, struct exclude_list *el, int srcpos)
453ec4bd 580{
d6b8fc30 581 struct exclude *x;
84460eec 582 int patternlen;
f8708998 583 unsigned flags;
84460eec 584 int nowildcardlen;
453ec4bd 585
84460eec
NTND
586 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
587 if (flags & EXC_FLAG_MUSTBEDIR) {
96ffc06f 588 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
d6b8fc30
JH
589 } else {
590 x = xmalloc(sizeof(*x));
591 x->pattern = string;
592 }
84460eec
NTND
593 x->patternlen = patternlen;
594 x->nowildcardlen = nowildcardlen;
453ec4bd
LT
595 x->base = base;
596 x->baselen = baselen;
d6b8fc30 597 x->flags = flags;
c04318e4 598 x->srcpos = srcpos;
840fc334
AS
599 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
600 el->excludes[el->nr++] = x;
c04318e4 601 x->el = el;
453ec4bd
LT
602}
603
6f52b741
BW
604static void *read_skip_worktree_file_from_index(const struct index_state *istate,
605 const char *path, size_t *size,
55fe6f51 606 struct sha1_stat *sha1_stat)
c28b3d6e
NTND
607{
608 int pos, len;
609 unsigned long sz;
610 enum object_type type;
611 void *data;
c28b3d6e
NTND
612
613 len = strlen(path);
6f52b741 614 pos = index_name_pos(istate, path, len);
c28b3d6e
NTND
615 if (pos < 0)
616 return NULL;
6f52b741 617 if (!ce_skip_worktree(istate->cache[pos]))
c28b3d6e 618 return NULL;
6f52b741 619 data = read_sha1_file(istate->cache[pos]->oid.hash, &type, &sz);
c28b3d6e
NTND
620 if (!data || type != OBJ_BLOB) {
621 free(data);
622 return NULL;
623 }
624 *size = xsize_t(sz);
55fe6f51
NTND
625 if (sha1_stat) {
626 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
6f52b741 627 hashcpy(sha1_stat->sha1, istate->cache[pos]->oid.hash);
55fe6f51 628 }
c28b3d6e
NTND
629 return data;
630}
631
f6198812
AS
632/*
633 * Frees memory within el which was allocated for exclude patterns and
634 * the file buffer. Does not free el itself.
635 */
636void clear_exclude_list(struct exclude_list *el)
0fd0e241
NTND
637{
638 int i;
639
5cee3493 640 for (i = 0; i < el->nr; i++)
0fd0e241
NTND
641 free(el->excludes[i]);
642 free(el->excludes);
c082df24 643 free(el->filebuf);
0fd0e241 644
2653a8c6 645 memset(el, 0, sizeof(*el));
0fd0e241
NTND
646}
647
7e2e4b37 648static void trim_trailing_spaces(char *buf)
16402b99 649{
e61a6c1d
PB
650 char *p, *last_space = NULL;
651
652 for (p = buf; *p; p++)
653 switch (*p) {
654 case ' ':
655 if (!last_space)
656 last_space = p;
657 break;
658 case '\\':
659 p++;
660 if (!*p)
661 return;
662 /* fallthrough */
663 default:
664 last_space = NULL;
665 }
666
667 if (last_space)
668 *last_space = '\0';
16402b99
NTND
669}
670
0dcb8d7f
NTND
671/*
672 * Given a subdirectory name and "dir" of the current directory,
673 * search the subdir in "dir" and return it, or create a new one if it
674 * does not exist in "dir".
675 *
676 * If "name" has the trailing slash, it'll be excluded in the search.
677 */
678static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
679 struct untracked_cache_dir *dir,
680 const char *name, int len)
681{
682 int first, last;
683 struct untracked_cache_dir *d;
684 if (!dir)
685 return NULL;
686 if (len && name[len - 1] == '/')
687 len--;
688 first = 0;
689 last = dir->dirs_nr;
690 while (last > first) {
691 int cmp, next = (last + first) >> 1;
692 d = dir->dirs[next];
693 cmp = strncmp(name, d->name, len);
694 if (!cmp && strlen(d->name) > len)
695 cmp = -1;
696 if (!cmp)
697 return d;
698 if (cmp < 0) {
699 last = next;
700 continue;
701 }
702 first = next+1;
703 }
704
705 uc->dir_created++;
96ffc06f 706 FLEX_ALLOC_MEM(d, name, name, len);
0dcb8d7f
NTND
707
708 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
709 memmove(dir->dirs + first + 1, dir->dirs + first,
710 (dir->dirs_nr - first) * sizeof(*dir->dirs));
711 dir->dirs_nr++;
712 dir->dirs[first] = d;
713 return d;
714}
715
ccad261f
NTND
716static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
717{
718 int i;
719 dir->valid = 0;
720 dir->untracked_nr = 0;
721 for (i = 0; i < dir->dirs_nr; i++)
722 do_invalidate_gitignore(dir->dirs[i]);
723}
724
725static void invalidate_gitignore(struct untracked_cache *uc,
726 struct untracked_cache_dir *dir)
727{
728 uc->gitignore_invalidated++;
729 do_invalidate_gitignore(dir);
730}
731
91a2288b
NTND
732static void invalidate_directory(struct untracked_cache *uc,
733 struct untracked_cache_dir *dir)
734{
26cb0182 735 int i;
91a2288b
NTND
736 uc->dir_invalidated++;
737 dir->valid = 0;
738 dir->untracked_nr = 0;
26cb0182
NTND
739 for (i = 0; i < dir->dirs_nr; i++)
740 dir->dirs[i]->recurse = 0;
91a2288b
NTND
741}
742
55fe6f51
NTND
743/*
744 * Given a file with name "fname", read it (either from disk, or from
473e3930 745 * an index if 'istate' is non-null), parse it and store the
55fe6f51
NTND
746 * exclude rules in "el".
747 *
748 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
749 * stat data from disk (only valid if add_excludes returns zero). If
750 * ss_valid is non-zero, "ss" must contain good value as input.
751 */
752static int add_excludes(const char *fname, const char *base, int baselen,
473e3930
BW
753 struct exclude_list *el,
754 struct index_state *istate,
55fe6f51 755 struct sha1_stat *sha1_stat)
453ec4bd 756{
c470701a 757 struct stat st;
c04318e4 758 int fd, i, lineno = 1;
9d14017a 759 size_t size = 0;
453ec4bd
LT
760 char *buf, *entry;
761
762 fd = open(fname, O_RDONLY);
c28b3d6e 763 if (fd < 0 || fstat(fd, &st) < 0) {
11dc1fcb
NTND
764 if (fd < 0)
765 warn_on_fopen_errors(fname);
766 else
c28b3d6e 767 close(fd);
473e3930
BW
768 if (!istate ||
769 (buf = read_skip_worktree_file_from_index(istate, fname, &size, sha1_stat)) == NULL)
c28b3d6e 770 return -1;
45d76f17
NTND
771 if (size == 0) {
772 free(buf);
773 return 0;
774 }
775 if (buf[size-1] != '\n') {
50a6c8ef 776 buf = xrealloc(buf, st_add(size, 1));
45d76f17
NTND
777 buf[size++] = '\n';
778 }
d961baa8 779 } else {
c28b3d6e
NTND
780 size = xsize_t(st.st_size);
781 if (size == 0) {
55fe6f51
NTND
782 if (sha1_stat) {
783 fill_stat_data(&sha1_stat->stat, &st);
784 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
785 sha1_stat->valid = 1;
786 }
c28b3d6e
NTND
787 close(fd);
788 return 0;
789 }
3733e694 790 buf = xmallocz(size);
c28b3d6e 791 if (read_in_full(fd, buf, size) != size) {
45d76f17 792 free(buf);
c28b3d6e
NTND
793 close(fd);
794 return -1;
795 }
45d76f17 796 buf[size++] = '\n';
c28b3d6e 797 close(fd);
55fe6f51
NTND
798 if (sha1_stat) {
799 int pos;
800 if (sha1_stat->valid &&
473e3930 801 !match_stat_data_racy(istate, &sha1_stat->stat, &st))
55fe6f51 802 ; /* no content change, ss->sha1 still good */
473e3930
BW
803 else if (istate &&
804 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
805 !ce_stage(istate->cache[pos]) &&
806 ce_uptodate(istate->cache[pos]) &&
82b474e0 807 !would_convert_to_git(istate, fname))
99d1a986 808 hashcpy(sha1_stat->sha1,
473e3930 809 istate->cache[pos]->oid.hash);
55fe6f51
NTND
810 else
811 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
812 fill_stat_data(&sha1_stat->stat, &st);
813 sha1_stat->valid = 1;
814 }
6ba78238 815 }
453ec4bd 816
c082df24 817 el->filebuf = buf;
245e1c19 818
dde843e7
JH
819 if (skip_utf8_bom(&buf, size))
820 size -= buf - el->filebuf;
821
453ec4bd 822 entry = buf;
245e1c19 823
45d76f17
NTND
824 for (i = 0; i < size; i++) {
825 if (buf[i] == '\n') {
453ec4bd
LT
826 if (entry != buf + i && entry[0] != '#') {
827 buf[i - (i && buf[i-1] == '\r')] = 0;
7e2e4b37 828 trim_trailing_spaces(entry);
c04318e4 829 add_exclude(entry, base, baselen, el, lineno);
453ec4bd 830 }
c04318e4 831 lineno++;
453ec4bd
LT
832 entry = buf + i + 1;
833 }
834 }
835 return 0;
453ec4bd
LT
836}
837
55fe6f51
NTND
838int add_excludes_from_file_to_list(const char *fname, const char *base,
839 int baselen, struct exclude_list *el,
473e3930 840 struct index_state *istate)
55fe6f51 841{
473e3930 842 return add_excludes(fname, base, baselen, el, istate, NULL);
55fe6f51
NTND
843}
844
c04318e4
AS
845struct exclude_list *add_exclude_list(struct dir_struct *dir,
846 int group_type, const char *src)
c082df24
AS
847{
848 struct exclude_list *el;
849 struct exclude_list_group *group;
850
851 group = &dir->exclude_list_group[group_type];
852 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
853 el = &group->el[group->nr++];
854 memset(el, 0, sizeof(*el));
c04318e4 855 el->src = src;
c082df24
AS
856 return el;
857}
858
859/*
860 * Used to set up core.excludesfile and .git/info/exclude lists.
861 */
0dcb8d7f
NTND
862static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
863 struct sha1_stat *sha1_stat)
453ec4bd 864{
c082df24 865 struct exclude_list *el;
ccad261f
NTND
866 /*
867 * catch setup_standard_excludes() that's called before
868 * dir->untracked is assigned. That function behaves
869 * differently when dir->untracked is non-NULL.
870 */
871 if (!dir->untracked)
872 dir->unmanaged_exclude_files++;
c04318e4 873 el = add_exclude_list(dir, EXC_FILE, fname);
473e3930 874 if (add_excludes(fname, "", 0, el, NULL, sha1_stat) < 0)
453ec4bd
LT
875 die("cannot use %s as an exclude file", fname);
876}
877
0dcb8d7f
NTND
878void add_excludes_from_file(struct dir_struct *dir, const char *fname)
879{
ccad261f 880 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
0dcb8d7f
NTND
881 add_excludes_from_file_1(dir, fname, NULL);
882}
883
82dce998
NTND
884int match_basename(const char *basename, int basenamelen,
885 const char *pattern, int prefix, int patternlen,
f8708998 886 unsigned flags)
593cb880
NTND
887{
888 if (prefix == patternlen) {
0b6e56df 889 if (patternlen == basenamelen &&
ba0897e6 890 !fspathncmp(pattern, basename, basenamelen))
593cb880
NTND
891 return 1;
892 } else if (flags & EXC_FLAG_ENDSWITH) {
0b6e56df 893 /* "*literal" matching against "fooliteral" */
593cb880 894 if (patternlen - 1 <= basenamelen &&
ba0897e6 895 !fspathncmp(pattern + 1,
0b6e56df
JH
896 basename + basenamelen - (patternlen - 1),
897 patternlen - 1))
593cb880
NTND
898 return 1;
899 } else {
0b6e56df
JH
900 if (fnmatch_icase_mem(pattern, patternlen,
901 basename, basenamelen,
902 0) == 0)
593cb880
NTND
903 return 1;
904 }
905 return 0;
906}
907
82dce998
NTND
908int match_pathname(const char *pathname, int pathlen,
909 const char *base, int baselen,
910 const char *pattern, int prefix, int patternlen,
f8708998 911 unsigned flags)
b5592632
NTND
912{
913 const char *name;
914 int namelen;
915
916 /*
917 * match with FNM_PATHNAME; the pattern has base implicitly
918 * in front of it.
919 */
920 if (*pattern == '/') {
921 pattern++;
982ac873 922 patternlen--;
b5592632
NTND
923 prefix--;
924 }
925
926 /*
927 * baselen does not count the trailing slash. base[] may or
928 * may not end with a trailing slash though.
929 */
930 if (pathlen < baselen + 1 ||
931 (baselen && pathname[baselen] != '/') ||
ba0897e6 932 fspathncmp(pathname, base, baselen))
b5592632
NTND
933 return 0;
934
935 namelen = baselen ? pathlen - baselen - 1 : pathlen;
936 name = pathname + pathlen - namelen;
937
938 if (prefix) {
939 /*
940 * if the non-wildcard part is longer than the
941 * remaining pathname, surely it cannot match.
942 */
943 if (prefix > namelen)
944 return 0;
945
ba0897e6 946 if (fspathncmp(pattern, name, prefix))
b5592632
NTND
947 return 0;
948 pattern += prefix;
ab3aebc1 949 patternlen -= prefix;
b5592632
NTND
950 name += prefix;
951 namelen -= prefix;
ab3aebc1
JK
952
953 /*
954 * If the whole pattern did not have a wildcard,
955 * then our prefix match is all we need; we
956 * do not need to call fnmatch at all.
957 */
5cee3493 958 if (!patternlen && !namelen)
ab3aebc1 959 return 1;
b5592632
NTND
960 }
961
ab3aebc1
JK
962 return fnmatch_icase_mem(pattern, patternlen,
963 name, namelen,
f30366b2 964 WM_PATHNAME) == 0;
b5592632
NTND
965}
966
578cd7c3
AS
967/*
968 * Scan the given exclude list in reverse to see whether pathname
969 * should be ignored. The first match (i.e. the last on the list), if
970 * any, determines the fate. Returns the exclude_list element which
971 * matched, or NULL for undecided.
453ec4bd 972 */
578cd7c3
AS
973static struct exclude *last_exclude_matching_from_list(const char *pathname,
974 int pathlen,
975 const char *basename,
976 int *dtype,
2b70e88d
BW
977 struct exclude_list *el,
978 struct index_state *istate)
453ec4bd 979{
e6efecc4 980 struct exclude *exc = NULL; /* undecided */
5cee3493 981 int i;
453ec4bd 982
35a94d44 983 if (!el->nr)
578cd7c3 984 return NULL; /* undefined */
d6b8fc30 985
35a94d44
NTND
986 for (i = el->nr - 1; 0 <= i; i--) {
987 struct exclude *x = el->excludes[i];
b5592632 988 const char *exclude = x->pattern;
b5592632 989 int prefix = x->nowildcardlen;
35a94d44
NTND
990
991 if (x->flags & EXC_FLAG_MUSTBEDIR) {
992 if (*dtype == DT_UNKNOWN)
2b70e88d 993 *dtype = get_dtype(NULL, istate, pathname, pathlen);
35a94d44
NTND
994 if (*dtype != DT_DIR)
995 continue;
996 }
d6b8fc30 997
35a94d44 998 if (x->flags & EXC_FLAG_NODIR) {
593cb880
NTND
999 if (match_basename(basename,
1000 pathlen - (basename - pathname),
1001 exclude, prefix, x->patternlen,
e6efecc4
NTND
1002 x->flags)) {
1003 exc = x;
1004 break;
1005 }
35a94d44
NTND
1006 continue;
1007 }
1008
b5592632
NTND
1009 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1010 if (match_pathname(pathname, pathlen,
1011 x->base, x->baselen ? x->baselen - 1 : 0,
e6efecc4
NTND
1012 exclude, prefix, x->patternlen, x->flags)) {
1013 exc = x;
1014 break;
1015 }
453ec4bd 1016 }
e6efecc4 1017 return exc;
578cd7c3
AS
1018}
1019
1020/*
1021 * Scan the list and let the last match determine the fate.
1022 * Return 1 for exclude, 0 for include and -1 for undecided.
1023 */
1024int is_excluded_from_list(const char *pathname,
1025 int pathlen, const char *basename, int *dtype,
fba92be8 1026 struct exclude_list *el, struct index_state *istate)
578cd7c3
AS
1027{
1028 struct exclude *exclude;
fba92be8
BW
1029 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1030 dtype, el, istate);
578cd7c3
AS
1031 if (exclude)
1032 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
453ec4bd
LT
1033 return -1; /* undecided */
1034}
1035
46aa2f95 1036static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
2b70e88d 1037 struct index_state *istate,
46aa2f95
KB
1038 const char *pathname, int pathlen, const char *basename,
1039 int *dtype_p)
1040{
1041 int i, j;
1042 struct exclude_list_group *group;
1043 struct exclude *exclude;
1044 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1045 group = &dir->exclude_list_group[i];
1046 for (j = group->nr - 1; j >= 0; j--) {
1047 exclude = last_exclude_matching_from_list(
1048 pathname, pathlen, basename, dtype_p,
2b70e88d 1049 &group->el[j], istate);
46aa2f95
KB
1050 if (exclude)
1051 return exclude;
1052 }
1053 }
1054 return NULL;
1055}
1056
6cd5c582
KB
1057/*
1058 * Loads the per-directory exclude list for the substring of base
1059 * which has a char length of baselen.
1060 */
e799ed40
BW
1061static void prep_exclude(struct dir_struct *dir,
1062 struct index_state *istate,
1063 const char *base, int baselen)
6cd5c582
KB
1064{
1065 struct exclude_list_group *group;
1066 struct exclude_list *el;
1067 struct exclude_stack *stk = NULL;
0dcb8d7f 1068 struct untracked_cache_dir *untracked;
6cd5c582
KB
1069 int current;
1070
6cd5c582
KB
1071 group = &dir->exclude_list_group[EXC_DIRS];
1072
d961baa8
NTND
1073 /*
1074 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
6cd5c582 1075 * which originate from directories not in the prefix of the
d961baa8
NTND
1076 * path being checked.
1077 */
6cd5c582
KB
1078 while ((stk = dir->exclude_stack) != NULL) {
1079 if (stk->baselen <= baselen &&
aceb9429 1080 !strncmp(dir->basebuf.buf, base, stk->baselen))
6cd5c582
KB
1081 break;
1082 el = &group->el[dir->exclude_stack->exclude_ix];
1083 dir->exclude_stack = stk->prev;
95c6f271 1084 dir->exclude = NULL;
aceb9429 1085 free((char *)el->src); /* see strbuf_detach() below */
6cd5c582
KB
1086 clear_exclude_list(el);
1087 free(stk);
1088 group->nr--;
1089 }
1090
95c6f271
KB
1091 /* Skip traversing into sub directories if the parent is excluded */
1092 if (dir->exclude)
1093 return;
1094
aceb9429
NTND
1095 /*
1096 * Lazy initialization. All call sites currently just
1097 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1098 * them seems lots of work for little benefit.
1099 */
1100 if (!dir->basebuf.buf)
1101 strbuf_init(&dir->basebuf, PATH_MAX);
1102
6cd5c582
KB
1103 /* Read from the parent directories and push them down. */
1104 current = stk ? stk->baselen : -1;
aceb9429 1105 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
0dcb8d7f
NTND
1106 if (dir->untracked)
1107 untracked = stk ? stk->ucd : dir->untracked->root;
1108 else
1109 untracked = NULL;
1110
6cd5c582 1111 while (current < baselen) {
6cd5c582 1112 const char *cp;
0dcb8d7f 1113 struct sha1_stat sha1_stat;
6cd5c582 1114
03e11a71 1115 stk = xcalloc(1, sizeof(*stk));
6cd5c582
KB
1116 if (current < 0) {
1117 cp = base;
1118 current = 0;
d961baa8 1119 } else {
6cd5c582
KB
1120 cp = strchr(base + current + 1, '/');
1121 if (!cp)
1122 die("oops in prep_exclude");
1123 cp++;
0dcb8d7f
NTND
1124 untracked =
1125 lookup_untracked(dir->untracked, untracked,
1126 base + current,
1127 cp - base - current);
6cd5c582
KB
1128 }
1129 stk->prev = dir->exclude_stack;
1130 stk->baselen = cp - base;
95c6f271 1131 stk->exclude_ix = group->nr;
0dcb8d7f 1132 stk->ucd = untracked;
95c6f271 1133 el = add_exclude_list(dir, EXC_DIRS, NULL);
aceb9429
NTND
1134 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1135 assert(stk->baselen == dir->basebuf.len);
95c6f271
KB
1136
1137 /* Abort if the directory is excluded */
1138 if (stk->baselen) {
1139 int dt = DT_DIR;
aceb9429 1140 dir->basebuf.buf[stk->baselen - 1] = 0;
95c6f271 1141 dir->exclude = last_exclude_matching_from_lists(dir,
e799ed40 1142 istate,
aceb9429
NTND
1143 dir->basebuf.buf, stk->baselen - 1,
1144 dir->basebuf.buf + current, &dt);
1145 dir->basebuf.buf[stk->baselen - 1] = '/';
c3c327de
KB
1146 if (dir->exclude &&
1147 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1148 dir->exclude = NULL;
95c6f271 1149 if (dir->exclude) {
95c6f271
KB
1150 dir->exclude_stack = stk;
1151 return;
1152 }
1153 }
1154
aceb9429 1155 /* Try to read per-directory file */
0dcb8d7f
NTND
1156 hashclr(sha1_stat.sha1);
1157 sha1_stat.valid = 0;
27b099ae
NTND
1158 if (dir->exclude_per_dir &&
1159 /*
1160 * If we know that no files have been added in
1161 * this directory (i.e. valid_cached_dir() has
1162 * been executed and set untracked->valid) ..
1163 */
1164 (!untracked || !untracked->valid ||
1165 /*
1166 * .. and .gitignore does not exist before
7687252f
DT
1167 * (i.e. null exclude_sha1). Then we can skip
1168 * loading .gitignore, which would result in
1169 * ENOENT anyway.
27b099ae
NTND
1170 */
1171 !is_null_sha1(untracked->exclude_sha1))) {
95c6f271
KB
1172 /*
1173 * dir->basebuf gets reused by the traversal, but we
1174 * need fname to remain unchanged to ensure the src
1175 * member of each struct exclude correctly
1176 * back-references its source file. Other invocations
1177 * of add_exclude_list provide stable strings, so we
aceb9429 1178 * strbuf_detach() and free() here in the caller.
95c6f271 1179 */
aceb9429
NTND
1180 struct strbuf sb = STRBUF_INIT;
1181 strbuf_addbuf(&sb, &dir->basebuf);
1182 strbuf_addstr(&sb, dir->exclude_per_dir);
1183 el->src = strbuf_detach(&sb, NULL);
e799ed40 1184 add_excludes(el->src, el->src, stk->baselen, el, istate,
0dcb8d7f
NTND
1185 untracked ? &sha1_stat : NULL);
1186 }
5ebf79ad
NTND
1187 /*
1188 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1189 * will first be called in valid_cached_dir() then maybe many
1190 * times more in last_exclude_matching(). When the cache is
1191 * used, last_exclude_matching() will not be called and
1192 * reading .gitignore content will be a waste.
1193 *
1194 * So when it's called by valid_cached_dir() and we can get
1195 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1196 * modified on work tree), we could delay reading the
1197 * .gitignore content until we absolutely need it in
1198 * last_exclude_matching(). Be careful about ignore rule
1199 * order, though, if you do that.
1200 */
1201 if (untracked &&
1202 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1203 invalidate_gitignore(dir->untracked, untracked);
0dcb8d7f 1204 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
95c6f271 1205 }
6cd5c582
KB
1206 dir->exclude_stack = stk;
1207 current = stk->baselen;
1208 }
aceb9429 1209 strbuf_setlen(&dir->basebuf, baselen);
6cd5c582
KB
1210}
1211
f4cd69a6
AS
1212/*
1213 * Loads the exclude lists for the directory containing pathname, then
1214 * scans all exclude lists to determine whether pathname is excluded.
1215 * Returns the exclude_list element which matched, or NULL for
1216 * undecided.
1217 */
b07bc8c8 1218struct exclude *last_exclude_matching(struct dir_struct *dir,
a0bba65b
BW
1219 struct index_state *istate,
1220 const char *pathname,
1221 int *dtype_p)
453ec4bd
LT
1222{
1223 int pathlen = strlen(pathname);
68492fc7
LK
1224 const char *basename = strrchr(pathname, '/');
1225 basename = (basename) ? basename+1 : pathname;
453ec4bd 1226
a0bba65b 1227 prep_exclude(dir, istate, pathname, basename-pathname);
c082df24 1228
95c6f271
KB
1229 if (dir->exclude)
1230 return dir->exclude;
1231
a0bba65b 1232 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
46aa2f95 1233 basename, dtype_p);
f4cd69a6
AS
1234}
1235
1236/*
1237 * Loads the exclude lists for the directory containing pathname, then
1238 * scans all exclude lists to determine whether pathname is excluded.
1239 * Returns 1 if true, otherwise 0.
1240 */
a0bba65b
BW
1241int is_excluded(struct dir_struct *dir, struct index_state *istate,
1242 const char *pathname, int *dtype_p)
f4cd69a6
AS
1243{
1244 struct exclude *exclude =
a0bba65b 1245 last_exclude_matching(dir, istate, pathname, dtype_p);
f4cd69a6
AS
1246 if (exclude)
1247 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
453ec4bd
LT
1248 return 0;
1249}
1250
f3fa1838
JH
1251static struct dir_entry *dir_entry_new(const char *pathname, int len)
1252{
453ec4bd
LT
1253 struct dir_entry *ent;
1254
96ffc06f 1255 FLEX_ALLOC_MEM(ent, name, pathname, len);
453ec4bd 1256 ent->len = len;
4d06f8ac 1257 return ent;
453ec4bd
LT
1258}
1259
9e58beca
BW
1260static struct dir_entry *dir_add_name(struct dir_struct *dir,
1261 struct index_state *istate,
1262 const char *pathname, int len)
6815e569 1263{
9e58beca 1264 if (index_file_exists(istate, pathname, len, ignore_case))
6815e569
JK
1265 return NULL;
1266
25fd2f7a 1267 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
6815e569
JK
1268 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1269}
1270
9e58beca
BW
1271struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1272 struct index_state *istate,
1273 const char *pathname, int len)
2abd31b0 1274{
9e58beca 1275 if (!index_name_is_other(istate, pathname, len))
2abd31b0
JK
1276 return NULL;
1277
25fd2f7a 1278 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
2abd31b0
JK
1279 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1280}
1281
09595258
LT
1282enum exist_status {
1283 index_nonexistent = 0,
1284 index_directory,
4b05548f 1285 index_gitdir
09595258
LT
1286};
1287
5102c617 1288/*
71261027 1289 * Do not use the alphabetically sorted index to look up
5102c617 1290 * the directory name; instead, use the case insensitive
ebbd7439 1291 * directory hash.
5102c617 1292 */
ae520e36
BW
1293static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1294 const char *dirname, int len)
5102c617 1295{
41284eb0 1296 struct cache_entry *ce;
5102c617 1297
ae520e36 1298 if (index_dir_exists(istate, dirname, len))
5102c617
JJ
1299 return index_directory;
1300
ae520e36 1301 ce = index_file_exists(istate, dirname, len, ignore_case);
41284eb0 1302 if (ce && S_ISGITLINK(ce->ce_mode))
5102c617
JJ
1303 return index_gitdir;
1304
5102c617
JJ
1305 return index_nonexistent;
1306}
1307
09595258
LT
1308/*
1309 * The index sorts alphabetically by entry name, which
1310 * means that a gitlink sorts as '\0' at the end, while
1311 * a directory (which is defined not as an entry, but as
1312 * the files it contains) will sort with the '/' at the
1313 * end.
1314 */
ae520e36
BW
1315static enum exist_status directory_exists_in_index(struct index_state *istate,
1316 const char *dirname, int len)
453ec4bd 1317{
5102c617
JJ
1318 int pos;
1319
1320 if (ignore_case)
ae520e36 1321 return directory_exists_in_index_icase(istate, dirname, len);
5102c617 1322
ae520e36 1323 pos = index_name_pos(istate, dirname, len);
09595258
LT
1324 if (pos < 0)
1325 pos = -pos-1;
ae520e36
BW
1326 while (pos < istate->cache_nr) {
1327 const struct cache_entry *ce = istate->cache[pos++];
09595258
LT
1328 unsigned char endchar;
1329
1330 if (strncmp(ce->name, dirname, len))
1331 break;
1332 endchar = ce->name[len];
1333 if (endchar > '/')
1334 break;
1335 if (endchar == '/')
1336 return index_directory;
7a51ed66 1337 if (!endchar && S_ISGITLINK(ce->ce_mode))
09595258
LT
1338 return index_gitdir;
1339 }
1340 return index_nonexistent;
1341}
1342
1343/*
1344 * When we find a directory when traversing the filesystem, we
1345 * have three distinct cases:
1346 *
1347 * - ignore it
1348 * - see it as a directory
1349 * - recurse into it
1350 *
1351 * and which one we choose depends on a combination of existing
1352 * git index contents and the flags passed into the directory
1353 * traversal routine.
1354 *
1355 * Case 1: If we *already* have entries in the index under that
5bd8e2d8
KB
1356 * directory name, we always recurse into the directory to see
1357 * all the files.
09595258
LT
1358 *
1359 * Case 2: If we *already* have that directory name as a gitlink,
1360 * we always continue to see it as a gitlink, regardless of whether
1361 * there is an actual git directory there or not (it might not
1362 * be checked out as a subproject!)
1363 *
1364 * Case 3: if we didn't have it in the index previously, we
1365 * have a few sub-cases:
1366 *
1367 * (a) if "show_other_directories" is true, we show it as
1368 * just a directory, unless "hide_empty_directories" is
defd7c7b
KB
1369 * also true, in which case we need to check if it contains any
1370 * untracked and / or ignored files.
09595258 1371 * (b) if it looks like a git directory, and we don't have
302b9282 1372 * 'no_gitlinks' set we treat it as a gitlink, and show it
09595258
LT
1373 * as a directory.
1374 * (c) otherwise, we recurse into it.
1375 */
defd7c7b 1376static enum path_treatment treat_directory(struct dir_struct *dir,
0ef8e169 1377 struct index_state *istate,
0dcb8d7f 1378 struct untracked_cache_dir *untracked,
2e5910f2 1379 const char *dirname, int len, int baselen, int exclude,
e1b8c7bd 1380 const struct pathspec *pathspec)
09595258
LT
1381{
1382 /* The "len-1" is to strip the final '/' */
0ef8e169 1383 switch (directory_exists_in_index(istate, dirname, len-1)) {
09595258 1384 case index_directory:
defd7c7b 1385 return path_recurse;
09595258
LT
1386
1387 case index_gitdir:
26c986e1 1388 return path_none;
09595258
LT
1389
1390 case index_nonexistent:
7c4c97c0 1391 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
09595258 1392 break;
7c4c97c0 1393 if (!(dir->flags & DIR_NO_GITLINKS)) {
09595258
LT
1394 unsigned char sha1[20];
1395 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
defd7c7b 1396 return path_untracked;
09595258 1397 }
defd7c7b 1398 return path_recurse;
09595258
LT
1399 }
1400
1401 /* This is the "show_other_directories" case */
721ac4ed 1402
184d2a8e 1403 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
defd7c7b
KB
1404 return exclude ? path_excluded : path_untracked;
1405
2e5910f2
DT
1406 untracked = lookup_untracked(dir->untracked, untracked,
1407 dirname + baselen, len - baselen);
0ef8e169 1408 return read_directory_recursive(dir, istate, dirname, len,
e1b8c7bd 1409 untracked, 1, pathspec);
453ec4bd
LT
1410}
1411
9fc42d60
LT
1412/*
1413 * This is an inexact early pruning of any recursive directory
1414 * reading - if the path cannot possibly be in the pathspec,
1415 * return true, and we'll skip it early.
1416 */
e1b8c7bd
BW
1417static int simplify_away(const char *path, int pathlen,
1418 const struct pathspec *pathspec)
9fc42d60 1419{
e1b8c7bd 1420 int i;
9fc42d60 1421
e1b8c7bd
BW
1422 if (!pathspec || !pathspec->nr)
1423 return 0;
1424
1425 GUARD_PATHSPEC(pathspec,
1426 PATHSPEC_FROMTOP |
1427 PATHSPEC_MAXDEPTH |
1428 PATHSPEC_LITERAL |
1429 PATHSPEC_GLOB |
1430 PATHSPEC_ICASE |
b0db7046
BW
1431 PATHSPEC_EXCLUDE |
1432 PATHSPEC_ATTR);
e1b8c7bd
BW
1433
1434 for (i = 0; i < pathspec->nr; i++) {
1435 const struct pathspec_item *item = &pathspec->items[i];
1436 int len = item->nowildcard_len;
1437
1438 if (len > pathlen)
1439 len = pathlen;
1440 if (!ps_strncmp(item, item->match, path, len))
1441 return 0;
9fc42d60 1442 }
e1b8c7bd
BW
1443
1444 return 1;
9fc42d60
LT
1445}
1446
29209cbe
JK
1447/*
1448 * This function tells us whether an excluded path matches a
1449 * list of "interesting" pathspecs. That is, whether a path matched
1450 * by any of the pathspecs could possibly be ignored by excluding
1451 * the specified path. This can happen if:
1452 *
1453 * 1. the path is mentioned explicitly in the pathspec
1454 *
1455 * 2. the path is a directory prefix of some element in the
1456 * pathspec
1457 */
e1b8c7bd
BW
1458static int exclude_matches_pathspec(const char *path, int pathlen,
1459 const struct pathspec *pathspec)
1460{
1461 int i;
1462
1463 if (!pathspec || !pathspec->nr)
1464 return 0;
1465
1466 GUARD_PATHSPEC(pathspec,
1467 PATHSPEC_FROMTOP |
1468 PATHSPEC_MAXDEPTH |
1469 PATHSPEC_LITERAL |
1470 PATHSPEC_GLOB |
1471 PATHSPEC_ICASE |
1472 PATHSPEC_EXCLUDE);
1473
1474 for (i = 0; i < pathspec->nr; i++) {
1475 const struct pathspec_item *item = &pathspec->items[i];
1476 int len = item->nowildcard_len;
1477
1478 if (len == pathlen &&
1479 !ps_strncmp(item, item->match, path, pathlen))
1480 return 1;
1481 if (len > pathlen &&
1482 item->match[pathlen] == '/' &&
1483 !ps_strncmp(item, item->match, path, pathlen))
1484 return 1;
e96980ef
JK
1485 }
1486 return 0;
1487}
1488
98f2a687
BW
1489static int get_index_dtype(struct index_state *istate,
1490 const char *path, int len)
443e061a
LT
1491{
1492 int pos;
9c5e6c80 1493 const struct cache_entry *ce;
443e061a 1494
98f2a687 1495 ce = index_file_exists(istate, path, len, 0);
443e061a
LT
1496 if (ce) {
1497 if (!ce_uptodate(ce))
1498 return DT_UNKNOWN;
1499 if (S_ISGITLINK(ce->ce_mode))
1500 return DT_DIR;
1501 /*
1502 * Nobody actually cares about the
1503 * difference between DT_LNK and DT_REG
1504 */
1505 return DT_REG;
1506 }
1507
1508 /* Try to look it up as a directory */
98f2a687 1509 pos = index_name_pos(istate, path, len);
443e061a
LT
1510 if (pos >= 0)
1511 return DT_UNKNOWN;
1512 pos = -pos-1;
98f2a687
BW
1513 while (pos < istate->cache_nr) {
1514 ce = istate->cache[pos++];
443e061a
LT
1515 if (strncmp(ce->name, path, len))
1516 break;
1517 if (ce->name[len] > '/')
1518 break;
1519 if (ce->name[len] < '/')
1520 continue;
1521 if (!ce_uptodate(ce))
1522 break; /* continue? */
1523 return DT_DIR;
1524 }
1525 return DT_UNKNOWN;
1526}
1527
98f2a687
BW
1528static int get_dtype(struct dirent *de, struct index_state *istate,
1529 const char *path, int len)
07134421 1530{
6831a88a 1531 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
07134421
LT
1532 struct stat st;
1533
1534 if (dtype != DT_UNKNOWN)
1535 return dtype;
98f2a687 1536 dtype = get_index_dtype(istate, path, len);
443e061a
LT
1537 if (dtype != DT_UNKNOWN)
1538 return dtype;
1539 if (lstat(path, &st))
07134421
LT
1540 return dtype;
1541 if (S_ISREG(st.st_mode))
1542 return DT_REG;
1543 if (S_ISDIR(st.st_mode))
1544 return DT_DIR;
1545 if (S_ISLNK(st.st_mode))
1546 return DT_LNK;
1547 return dtype;
1548}
1549
16e2cfa9 1550static enum path_treatment treat_one_path(struct dir_struct *dir,
0dcb8d7f 1551 struct untracked_cache_dir *untracked,
0ef8e169 1552 struct index_state *istate,
49dc2cc2 1553 struct strbuf *path,
2e5910f2 1554 int baselen,
e1b8c7bd 1555 const struct pathspec *pathspec,
16e2cfa9 1556 int dtype, struct dirent *de)
53cc5356 1557{
8aaf8d77 1558 int exclude;
0ef8e169 1559 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
2eac2a4c 1560
8aaf8d77 1561 if (dtype == DT_UNKNOWN)
0ef8e169 1562 dtype = get_dtype(de, istate, path->buf, path->len);
8aaf8d77
KB
1563
1564 /* Always exclude indexed files */
2eac2a4c 1565 if (dtype != DT_DIR && has_path_in_index)
defd7c7b 1566 return path_none;
8aaf8d77 1567
2eac2a4c
JH
1568 /*
1569 * When we are looking at a directory P in the working tree,
1570 * there are three cases:
1571 *
1572 * (1) P exists in the index. Everything inside the directory P in
1573 * the working tree needs to go when P is checked out from the
1574 * index.
1575 *
1576 * (2) P does not exist in the index, but there is P/Q in the index.
1577 * We know P will stay a directory when we check out the contents
1578 * of the index, but we do not know yet if there is a directory
1579 * P/Q in the working tree to be killed, so we need to recurse.
1580 *
1581 * (3) P does not exist in the index, and there is no P/Q in the index
1582 * to require P to be a directory, either. Only in this case, we
1583 * know that everything inside P will not be killed without
1584 * recursing.
1585 */
1586 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1587 (dtype == DT_DIR) &&
de372b1b 1588 !has_path_in_index &&
0ef8e169 1589 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
de372b1b 1590 return path_none;
8aaf8d77 1591
0ef8e169 1592 exclude = is_excluded(dir, istate, path->buf, &dtype);
53cc5356
JH
1593
1594 /*
1595 * Excluded? If we don't explicitly want to show
1596 * ignored files, ignore it
1597 */
0aaf62b6 1598 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
defd7c7b 1599 return path_excluded;
53cc5356 1600
53cc5356
JH
1601 switch (dtype) {
1602 default:
defd7c7b 1603 return path_none;
53cc5356 1604 case DT_DIR:
49dc2cc2 1605 strbuf_addch(path, '/');
0ef8e169 1606 return treat_directory(dir, istate, untracked, path->buf, path->len,
e1b8c7bd 1607 baselen, exclude, pathspec);
53cc5356
JH
1608 case DT_REG:
1609 case DT_LNK:
defd7c7b 1610 return exclude ? path_excluded : path_untracked;
53cc5356 1611 }
53cc5356
JH
1612}
1613
91a2288b
NTND
1614static enum path_treatment treat_path_fast(struct dir_struct *dir,
1615 struct untracked_cache_dir *untracked,
1616 struct cached_dir *cdir,
0ef8e169 1617 struct index_state *istate,
91a2288b
NTND
1618 struct strbuf *path,
1619 int baselen,
e1b8c7bd 1620 const struct pathspec *pathspec)
91a2288b
NTND
1621{
1622 strbuf_setlen(path, baselen);
1623 if (!cdir->ucd) {
1624 strbuf_addstr(path, cdir->file);
1625 return path_untracked;
1626 }
1627 strbuf_addstr(path, cdir->ucd->name);
1628 /* treat_one_path() does this before it calls treat_directory() */
00b6c178 1629 strbuf_complete(path, '/');
91a2288b
NTND
1630 if (cdir->ucd->check_only)
1631 /*
1632 * check_only is set as a result of treat_directory() getting
1633 * to its bottom. Verify again the same set of directories
1634 * with check_only set.
1635 */
0ef8e169 1636 return read_directory_recursive(dir, istate, path->buf, path->len,
e1b8c7bd 1637 cdir->ucd, 1, pathspec);
91a2288b
NTND
1638 /*
1639 * We get path_recurse in the first run when
1640 * directory_exists_in_index() returns index_nonexistent. We
1641 * are sure that new changes in the index does not impact the
1642 * outcome. Return now.
1643 */
1644 return path_recurse;
1645}
1646
16e2cfa9 1647static enum path_treatment treat_path(struct dir_struct *dir,
0dcb8d7f 1648 struct untracked_cache_dir *untracked,
cf7c6148 1649 struct cached_dir *cdir,
0ef8e169 1650 struct index_state *istate,
49dc2cc2 1651 struct strbuf *path,
16e2cfa9 1652 int baselen,
e1b8c7bd 1653 const struct pathspec *pathspec)
16e2cfa9
JH
1654{
1655 int dtype;
cf7c6148 1656 struct dirent *de = cdir->de;
16e2cfa9 1657
91a2288b 1658 if (!de)
0ef8e169 1659 return treat_path_fast(dir, untracked, cdir, istate, path,
e1b8c7bd 1660 baselen, pathspec);
16e2cfa9 1661 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
defd7c7b 1662 return path_none;
49dc2cc2
RS
1663 strbuf_setlen(path, baselen);
1664 strbuf_addstr(path, de->d_name);
e1b8c7bd 1665 if (simplify_away(path->buf, path->len, pathspec))
defd7c7b 1666 return path_none;
16e2cfa9
JH
1667
1668 dtype = DTYPE(de);
0ef8e169 1669 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
0dcb8d7f
NTND
1670}
1671
1672static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1673{
1674 if (!dir)
1675 return;
1676 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1677 dir->untracked_alloc);
1678 dir->untracked[dir->untracked_nr++] = xstrdup(name);
16e2cfa9
JH
1679}
1680
91a2288b
NTND
1681static int valid_cached_dir(struct dir_struct *dir,
1682 struct untracked_cache_dir *untracked,
207a06ce 1683 struct index_state *istate,
91a2288b
NTND
1684 struct strbuf *path,
1685 int check_only)
1686{
1687 struct stat st;
1688
1689 if (!untracked)
1690 return 0;
1691
883e248b
BP
1692 /*
1693 * With fsmonitor, we can trust the untracked cache's valid field.
1694 */
1695 refresh_fsmonitor(istate);
1696 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1697 if (stat(path->len ? path->buf : ".", &st)) {
91a2288b 1698 invalidate_directory(dir->untracked, untracked);
883e248b
BP
1699 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1700 return 0;
1701 }
1702 if (!untracked->valid ||
1703 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1704 if (untracked->valid)
1705 invalidate_directory(dir->untracked, untracked);
1706 fill_stat_data(&untracked->stat_data, &st);
1707 return 0;
1708 }
91a2288b
NTND
1709 }
1710
1711 if (untracked->check_only != !!check_only) {
1712 invalidate_directory(dir->untracked, untracked);
1713 return 0;
1714 }
1715
1716 /*
1717 * prep_exclude will be called eventually on this directory,
1718 * but it's called much later in last_exclude_matching(). We
1719 * need it now to determine the validity of the cache for this
1720 * path. The next calls will be nearly no-op, the way
1721 * prep_exclude() is designed.
1722 */
1723 if (path->len && path->buf[path->len - 1] != '/') {
1724 strbuf_addch(path, '/');
207a06ce 1725 prep_exclude(dir, istate, path->buf, path->len);
91a2288b
NTND
1726 strbuf_setlen(path, path->len - 1);
1727 } else
207a06ce 1728 prep_exclude(dir, istate, path->buf, path->len);
91a2288b
NTND
1729
1730 /* hopefully prep_exclude() haven't invalidated this entry... */
1731 return untracked->valid;
1732}
1733
cf7c6148
NTND
1734static int open_cached_dir(struct cached_dir *cdir,
1735 struct dir_struct *dir,
1736 struct untracked_cache_dir *untracked,
207a06ce 1737 struct index_state *istate,
cf7c6148
NTND
1738 struct strbuf *path,
1739 int check_only)
1740{
1741 memset(cdir, 0, sizeof(*cdir));
1742 cdir->untracked = untracked;
207a06ce 1743 if (valid_cached_dir(dir, untracked, istate, path, check_only))
91a2288b 1744 return 0;
cf7c6148 1745 cdir->fdir = opendir(path->len ? path->buf : ".");
91a2288b
NTND
1746 if (dir->untracked)
1747 dir->untracked->dir_opened++;
cf7c6148
NTND
1748 if (!cdir->fdir)
1749 return -1;
1750 return 0;
1751}
1752
1753static int read_cached_dir(struct cached_dir *cdir)
1754{
1755 if (cdir->fdir) {
1756 cdir->de = readdir(cdir->fdir);
1757 if (!cdir->de)
1758 return -1;
1759 return 0;
1760 }
91a2288b
NTND
1761 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1762 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
26cb0182
NTND
1763 if (!d->recurse) {
1764 cdir->nr_dirs++;
1765 continue;
1766 }
91a2288b
NTND
1767 cdir->ucd = d;
1768 cdir->nr_dirs++;
1769 return 0;
1770 }
1771 cdir->ucd = NULL;
1772 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1773 struct untracked_cache_dir *d = cdir->untracked;
1774 cdir->file = d->untracked[cdir->nr_files++];
1775 return 0;
1776 }
cf7c6148
NTND
1777 return -1;
1778}
1779
1780static void close_cached_dir(struct cached_dir *cdir)
1781{
1782 if (cdir->fdir)
1783 closedir(cdir->fdir);
91a2288b
NTND
1784 /*
1785 * We have gone through this directory and found no untracked
1786 * entries. Mark it valid.
1787 */
26cb0182 1788 if (cdir->untracked) {
91a2288b 1789 cdir->untracked->valid = 1;
26cb0182
NTND
1790 cdir->untracked->recurse = 1;
1791 }
16e2cfa9
JH
1792}
1793
453ec4bd
LT
1794/*
1795 * Read a directory tree. We currently ignore anything but
1796 * directories, regular files and symlinks. That's because git
1797 * doesn't handle them at all yet. Maybe that will change some
1798 * day.
1799 *
1800 * Also, we ignore the name ".git" (even if it is not a directory).
1801 * That likely will not change.
defd7c7b
KB
1802 *
1803 * Returns the most significant path_treatment value encountered in the scan.
453ec4bd 1804 */
defd7c7b 1805static enum path_treatment read_directory_recursive(struct dir_struct *dir,
0ef8e169
BW
1806 struct index_state *istate, const char *base, int baselen,
1807 struct untracked_cache_dir *untracked, int check_only,
1808 const struct pathspec *pathspec)
453ec4bd 1809{
cf7c6148 1810 struct cached_dir cdir;
defd7c7b 1811 enum path_treatment state, subdir_state, dir_state = path_none;
bef36921 1812 struct strbuf path = STRBUF_INIT;
453ec4bd 1813
bef36921 1814 strbuf_add(&path, base, baselen);
02cb6753 1815
0ef8e169 1816 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1528d247
RS
1817 goto out;
1818
0dcb8d7f
NTND
1819 if (untracked)
1820 untracked->check_only = !!check_only;
1821
cf7c6148 1822 while (!read_cached_dir(&cdir)) {
defd7c7b 1823 /* check how the file or directory should be treated */
0ef8e169 1824 state = treat_path(dir, untracked, &cdir, istate, &path,
e1b8c7bd 1825 baselen, pathspec);
0dcb8d7f 1826
defd7c7b
KB
1827 if (state > dir_state)
1828 dir_state = state;
1829
1830 /* recurse into subdir if instructed by treat_path */
df5bcdf8
SL
1831 if ((state == path_recurse) ||
1832 ((state == path_untracked) &&
1833 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
f4fd99bf 1834 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
0dcb8d7f
NTND
1835 struct untracked_cache_dir *ud;
1836 ud = lookup_untracked(dir->untracked, untracked,
1837 path.buf + baselen,
1838 path.len - baselen);
1839 subdir_state =
0ef8e169 1840 read_directory_recursive(dir, istate, path.buf,
e1b8c7bd
BW
1841 path.len, ud,
1842 check_only, pathspec);
defd7c7b
KB
1843 if (subdir_state > dir_state)
1844 dir_state = subdir_state;
1845 }
1846
1847 if (check_only) {
1848 /* abort early if maximum state has been reached */
0dcb8d7f 1849 if (dir_state == path_untracked) {
91a2288b 1850 if (cdir.fdir)
0dcb8d7f 1851 add_untracked(untracked, path.buf + baselen);
defd7c7b 1852 break;
0dcb8d7f 1853 }
defd7c7b 1854 /* skip the dir_add_* part */
02cb6753 1855 continue;
453ec4bd 1856 }
defd7c7b
KB
1857
1858 /* add the path to the appropriate result list */
1859 switch (state) {
1860 case path_excluded:
1861 if (dir->flags & DIR_SHOW_IGNORED)
0ef8e169 1862 dir_add_name(dir, istate, path.buf, path.len);
0aaf62b6
KB
1863 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1864 ((dir->flags & DIR_COLLECT_IGNORED) &&
1865 exclude_matches_pathspec(path.buf, path.len,
e1b8c7bd 1866 pathspec)))
0ef8e169 1867 dir_add_ignored(dir, istate, path.buf, path.len);
defd7c7b
KB
1868 break;
1869
1870 case path_untracked:
0dcb8d7f
NTND
1871 if (dir->flags & DIR_SHOW_IGNORED)
1872 break;
0ef8e169 1873 dir_add_name(dir, istate, path.buf, path.len);
91a2288b 1874 if (cdir.fdir)
0dcb8d7f 1875 add_untracked(untracked, path.buf + baselen);
1528d247 1876 break;
defd7c7b
KB
1877
1878 default:
1879 break;
1880 }
453ec4bd 1881 }
cf7c6148 1882 close_cached_dir(&cdir);
1528d247 1883 out:
bef36921 1884 strbuf_release(&path);
453ec4bd 1885
defd7c7b 1886 return dir_state;
453ec4bd
LT
1887}
1888
bbf504a9 1889int cmp_dir_entry(const void *p1, const void *p2)
453ec4bd
LT
1890{
1891 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1892 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1893
ccdd4a0f 1894 return name_compare(e1->name, e1->len, e2->name, e2->len);
453ec4bd
LT
1895}
1896
fb898888 1897/* check if *out lexically strictly contains *in */
bbf504a9 1898int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
fb898888
SL
1899{
1900 return (out->len < in->len) &&
1901 (out->name[out->len - 1] == '/') &&
1902 !memcmp(out->name, in->name, out->len);
1903}
1904
48ffef96 1905static int treat_leading_path(struct dir_struct *dir,
0ef8e169 1906 struct index_state *istate,
48ffef96 1907 const char *path, int len,
e1b8c7bd 1908 const struct pathspec *pathspec)
48ffef96 1909{
49dc2cc2
RS
1910 struct strbuf sb = STRBUF_INIT;
1911 int baselen, rc = 0;
48ffef96 1912 const char *cp;
be8a84c5 1913 int old_flags = dir->flags;
48ffef96
JH
1914
1915 while (len && path[len - 1] == '/')
1916 len--;
1917 if (!len)
1918 return 1;
1919 baselen = 0;
be8a84c5 1920 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
48ffef96
JH
1921 while (1) {
1922 cp = path + baselen + !!baselen;
1923 cp = memchr(cp, '/', path + len - cp);
1924 if (!cp)
1925 baselen = len;
1926 else
1927 baselen = cp - path;
49dc2cc2
RS
1928 strbuf_setlen(&sb, 0);
1929 strbuf_add(&sb, path, baselen);
1930 if (!is_directory(sb.buf))
1931 break;
e1b8c7bd 1932 if (simplify_away(sb.buf, sb.len, pathspec))
49dc2cc2 1933 break;
0ef8e169 1934 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
defd7c7b 1935 DT_DIR, NULL) == path_none)
49dc2cc2
RS
1936 break; /* do not recurse into it */
1937 if (len <= baselen) {
1938 rc = 1;
1939 break; /* finished checking */
1940 }
48ffef96 1941 }
49dc2cc2 1942 strbuf_release(&sb);
be8a84c5 1943 dir->flags = old_flags;
49dc2cc2 1944 return rc;
48ffef96
JH
1945}
1946
1e8fef60
NTND
1947static const char *get_ident_string(void)
1948{
1949 static struct strbuf sb = STRBUF_INIT;
1950 struct utsname uts;
1951
1952 if (sb.len)
1953 return sb.buf;
100e4337 1954 if (uname(&uts) < 0)
1e8fef60 1955 die_errno(_("failed to get kernel name and information"));
0e0f7618
CC
1956 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
1957 uts.sysname);
1e8fef60
NTND
1958 return sb.buf;
1959}
1960
1961static int ident_in_untracked(const struct untracked_cache *uc)
1962{
0e0f7618
CC
1963 /*
1964 * Previous git versions may have saved many NUL separated
1965 * strings in the "ident" field, but it is insane to manage
1966 * many locations, so just take care of the first one.
1967 */
1e8fef60 1968
0e0f7618 1969 return !strcmp(uc->ident.buf, get_ident_string());
1e8fef60
NTND
1970}
1971
0e0f7618 1972static void set_untracked_ident(struct untracked_cache *uc)
1e8fef60 1973{
0e0f7618 1974 strbuf_reset(&uc->ident);
1e8fef60 1975 strbuf_addstr(&uc->ident, get_ident_string());
0e0f7618
CC
1976
1977 /*
1978 * This strbuf used to contain a list of NUL separated
1979 * strings, so save NUL too for backward compatibility.
1980 */
1e8fef60
NTND
1981 strbuf_addch(&uc->ident, 0);
1982}
1983
4a4ca479
CC
1984static void new_untracked_cache(struct index_state *istate)
1985{
1986 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
1987 strbuf_init(&uc->ident, 100);
1988 uc->exclude_per_dir = ".gitignore";
1989 /* should be the same flags used by git-status */
1990 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
0e0f7618 1991 set_untracked_ident(uc);
4a4ca479 1992 istate->untracked = uc;
0e0f7618 1993 istate->cache_changed |= UNTRACKED_CHANGED;
4a4ca479
CC
1994}
1995
1996void add_untracked_cache(struct index_state *istate)
1997{
1998 if (!istate->untracked) {
1999 new_untracked_cache(istate);
0e0f7618
CC
2000 } else {
2001 if (!ident_in_untracked(istate->untracked)) {
2002 free_untracked_cache(istate->untracked);
2003 new_untracked_cache(istate);
2004 }
2005 }
4a4ca479
CC
2006}
2007
07b29bfd
CC
2008void remove_untracked_cache(struct index_state *istate)
2009{
2010 if (istate->untracked) {
2011 free_untracked_cache(istate->untracked);
2012 istate->untracked = NULL;
2013 istate->cache_changed |= UNTRACKED_CHANGED;
2014 }
2015}
2016
ccad261f
NTND
2017static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2018 int base_len,
2019 const struct pathspec *pathspec)
2020{
2021 struct untracked_cache_dir *root;
2022
76e6b090 2023 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
ccad261f
NTND
2024 return NULL;
2025
2026 /*
2027 * We only support $GIT_DIR/info/exclude and core.excludesfile
2028 * as the global ignore rule files. Any other additions
2029 * (e.g. from command line) invalidate the cache. This
2030 * condition also catches running setup_standard_excludes()
2031 * before setting dir->untracked!
2032 */
2033 if (dir->unmanaged_exclude_files)
2034 return NULL;
2035
2036 /*
2037 * Optimize for the main use case only: whole-tree git
2038 * status. More work involved in treat_leading_path() if we
2039 * use cache on just a subset of the worktree. pathspec
2040 * support could make the matter even worse.
2041 */
2042 if (base_len || (pathspec && pathspec->nr))
2043 return NULL;
2044
2045 /* Different set of flags may produce different results */
2046 if (dir->flags != dir->untracked->dir_flags ||
2047 /*
2048 * See treat_directory(), case index_nonexistent. Without
2049 * this flag, we may need to also cache .git file content
2050 * for the resolve_gitlink_ref() call, which we don't.
2051 */
2052 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2053 /* We don't support collecting ignore files */
2054 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2055 DIR_COLLECT_IGNORED)))
2056 return NULL;
2057
2058 /*
2059 * If we use .gitignore in the cache and now you change it to
2060 * .gitexclude, everything will go wrong.
2061 */
2062 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2063 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2064 return NULL;
2065
2066 /*
2067 * EXC_CMDL is not considered in the cache. If people set it,
2068 * skip the cache.
2069 */
2070 if (dir->exclude_list_group[EXC_CMDL].nr)
2071 return NULL;
2072
1e8fef60 2073 if (!ident_in_untracked(dir->untracked)) {
0e0f7618 2074 warning(_("Untracked cache is disabled on this system or location."));
1e8fef60
NTND
2075 return NULL;
2076 }
2077
ccad261f
NTND
2078 if (!dir->untracked->root) {
2079 const int len = sizeof(*dir->untracked->root);
2080 dir->untracked->root = xmalloc(len);
2081 memset(dir->untracked->root, 0, len);
2082 }
2083
2084 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2085 root = dir->untracked->root;
2086 if (hashcmp(dir->ss_info_exclude.sha1,
2087 dir->untracked->ss_info_exclude.sha1)) {
2088 invalidate_gitignore(dir->untracked, root);
2089 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2090 }
2091 if (hashcmp(dir->ss_excludes_file.sha1,
2092 dir->untracked->ss_excludes_file.sha1)) {
2093 invalidate_gitignore(dir->untracked, root);
2094 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2095 }
26cb0182
NTND
2096
2097 /* Make sure this directory is not dropped out at saving phase */
2098 root->recurse = 1;
ccad261f
NTND
2099 return root;
2100}
2101
2c1eb104
BW
2102int read_directory(struct dir_struct *dir, struct index_state *istate,
2103 const char *path, int len, const struct pathspec *pathspec)
9fc42d60 2104{
ccad261f 2105 struct untracked_cache_dir *untracked;
b4189aa8 2106
dba2e203 2107 if (has_symlink_leading_path(path, len))
725b0605
JH
2108 return dir->nr;
2109
ccad261f
NTND
2110 untracked = validate_untracked_cache(dir, len, pathspec);
2111 if (!untracked)
2112 /*
2113 * make sure untracked cache code path is disabled,
2114 * e.g. prep_exclude()
2115 */
2116 dir->untracked = NULL;
2c1eb104
BW
2117 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
2118 read_directory_recursive(dir, istate, path, len, untracked, 0, pathspec);
bbf504a9
SL
2119 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2120 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
fb898888
SL
2121
2122 /*
2123 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2124 * also pick up untracked contents of untracked dirs; by default
2125 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2126 */
2127 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2128 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2129 int i, j;
2130
2131 /* remove from dir->entries untracked contents of untracked dirs */
2132 for (i = j = 0; j < dir->nr; j++) {
bbf504a9
SL
2133 if (i &&
2134 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
6a83d902 2135 FREE_AND_NULL(dir->entries[j]);
fb898888
SL
2136 } else {
2137 dir->entries[i++] = dir->entries[j];
2138 }
2139 }
2140
2141 dir->nr = i;
2142 }
2143
c9ccb5d3
NTND
2144 if (dir->untracked) {
2145 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2146 trace_printf_key(&trace_untracked_stats,
2147 "node creation: %u\n"
2148 "gitignore invalidation: %u\n"
2149 "directory invalidation: %u\n"
2150 "opendir: %u\n",
2151 dir->untracked->dir_created,
2152 dir->untracked->gitignore_invalidated,
2153 dir->untracked->dir_invalidated,
2154 dir->untracked->dir_opened);
2c1eb104 2155 if (dir->untracked == istate->untracked &&
1bbb3dba
NTND
2156 (dir->untracked->dir_opened ||
2157 dir->untracked->gitignore_invalidated ||
2158 dir->untracked->dir_invalidated))
2c1eb104
BW
2159 istate->cache_changed |= UNTRACKED_CHANGED;
2160 if (dir->untracked != istate->untracked) {
6a83d902 2161 FREE_AND_NULL(dir->untracked);
1bbb3dba 2162 }
c9ccb5d3 2163 }
453ec4bd
LT
2164 return dir->nr;
2165}
c91f0d92 2166
686a4a06 2167int file_exists(const char *f)
c91f0d92 2168{
686a4a06 2169 struct stat sb;
a50f9fc5 2170 return lstat(f, &sb) == 0;
c91f0d92 2171}
e6636747 2172
63ec5e1f
JS
2173static int cmp_icase(char a, char b)
2174{
2175 if (a == b)
2176 return 0;
2177 if (ignore_case)
2178 return toupper(a) - toupper(b);
2179 return a - b;
2180}
2181
e6636747 2182/*
9b125da4
NTND
2183 * Given two normalized paths (a trailing slash is ok), if subdir is
2184 * outside dir, return -1. Otherwise return the offset in subdir that
2185 * can be used as relative path to dir.
e6636747 2186 */
9b125da4 2187int dir_inside_of(const char *subdir, const char *dir)
e6636747 2188{
9b125da4 2189 int offset = 0;
e6636747 2190
9b125da4 2191 assert(dir && subdir && *dir && *subdir);
e6636747 2192
63ec5e1f 2193 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
e6636747 2194 dir++;
9b125da4
NTND
2195 subdir++;
2196 offset++;
490544b1 2197 }
9b125da4
NTND
2198
2199 /* hel[p]/me vs hel[l]/yeah */
2200 if (*dir && *subdir)
2201 return -1;
2202
2203 if (!*subdir)
2204 return !*dir ? offset : -1; /* same dir */
2205
2206 /* foo/[b]ar vs foo/[] */
2207 if (is_dir_sep(dir[-1]))
2208 return is_dir_sep(subdir[-1]) ? offset : -1;
2209
2210 /* foo[/]bar vs foo[] */
2211 return is_dir_sep(*subdir) ? offset + 1 : -1;
e6636747
JS
2212}
2213
2214int is_inside_dir(const char *dir)
2215{
56b9f6e7
RS
2216 char *cwd;
2217 int rc;
2218
b892913d
NTND
2219 if (!dir)
2220 return 0;
56b9f6e7
RS
2221
2222 cwd = xgetcwd();
2223 rc = (dir_inside_of(cwd, dir) >= 0);
2224 free(cwd);
2225 return rc;
e6636747 2226}
7155b727 2227
55892d23
AP
2228int is_empty_dir(const char *path)
2229{
2230 DIR *dir = opendir(path);
2231 struct dirent *e;
2232 int ret = 1;
2233
2234 if (!dir)
2235 return 0;
2236
2237 while ((e = readdir(dir)) != NULL)
2238 if (!is_dot_or_dotdot(e->d_name)) {
2239 ret = 0;
2240 break;
2241 }
2242
2243 closedir(dir);
2244 return ret;
2245}
2246
ae2f203e 2247static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
7155b727 2248{
a0f4afbe 2249 DIR *dir;
7155b727 2250 struct dirent *e;
ae2f203e 2251 int ret = 0, original_len = path->len, len, kept_down = 0;
a0f4afbe 2252 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
c844a803 2253 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
a0f4afbe 2254 unsigned char submodule_head[20];
7155b727 2255
a0f4afbe 2256 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
ae2f203e 2257 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
a0f4afbe 2258 /* Do not descend and nuke a nested git work tree. */
ae2f203e
JH
2259 if (kept_up)
2260 *kept_up = 1;
a0f4afbe 2261 return 0;
ae2f203e 2262 }
a0f4afbe 2263
ae2f203e 2264 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
a0f4afbe 2265 dir = opendir(path->buf);
c844a803 2266 if (!dir) {
863808cd
MH
2267 if (errno == ENOENT)
2268 return keep_toplevel ? -1 : 0;
2269 else if (errno == EACCES && !keep_toplevel)
ecb2c282
MH
2270 /*
2271 * An empty dir could be removable even if it
2272 * is unreadable:
2273 */
c844a803
JH
2274 return rmdir(path->buf);
2275 else
2276 return -1;
2277 }
00b6c178 2278 strbuf_complete(path, '/');
7155b727
JS
2279
2280 len = path->len;
2281 while ((e = readdir(dir)) != NULL) {
2282 struct stat st;
8ca12c0d
AP
2283 if (is_dot_or_dotdot(e->d_name))
2284 continue;
7155b727
JS
2285
2286 strbuf_setlen(path, len);
2287 strbuf_addstr(path, e->d_name);
863808cd
MH
2288 if (lstat(path->buf, &st)) {
2289 if (errno == ENOENT)
2290 /*
2291 * file disappeared, which is what we
2292 * wanted anyway
2293 */
2294 continue;
2295 /* fall thru */
2296 } else if (S_ISDIR(st.st_mode)) {
ae2f203e 2297 if (!remove_dir_recurse(path, flag, &kept_down))
7155b727 2298 continue; /* happy */
863808cd
MH
2299 } else if (!only_empty &&
2300 (!unlink(path->buf) || errno == ENOENT)) {
7155b727 2301 continue; /* happy, too */
863808cd 2302 }
7155b727
JS
2303
2304 /* path too long, stat fails, or non-directory still exists */
2305 ret = -1;
2306 break;
2307 }
2308 closedir(dir);
2309
2310 strbuf_setlen(path, original_len);
ae2f203e 2311 if (!ret && !keep_toplevel && !kept_down)
863808cd 2312 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
ae2f203e
JH
2313 else if (kept_up)
2314 /*
2315 * report the uplevel that it is not an error that we
2316 * did not rmdir() our directory.
2317 */
2318 *kept_up = !ret;
7155b727
JS
2319 return ret;
2320}
039bc64e 2321
ae2f203e
JH
2322int remove_dir_recursively(struct strbuf *path, int flag)
2323{
2324 return remove_dir_recurse(path, flag, NULL);
2325}
2326
f932729c
JK
2327static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2328
039bc64e
JH
2329void setup_standard_excludes(struct dir_struct *dir)
2330{
039bc64e 2331 dir->exclude_per_dir = ".gitignore";
099d2d86
JH
2332
2333 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2845ce7f
PT
2334 if (!excludes_file)
2335 excludes_file = xdg_config_home("ignore");
4698c8fe 2336 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
38ccaf93
JH
2337 add_excludes_from_file_1(dir, excludes_file,
2338 dir->untracked ? &dir->ss_excludes_file : NULL);
099d2d86
JH
2339
2340 /* per repository user preference */
f0056f64
JK
2341 if (startup_info->have_repository) {
2342 const char *path = git_path_info_exclude();
2343 if (!access_or_warn(path, R_OK, 0))
2344 add_excludes_from_file_1(dir, path,
2345 dir->untracked ? &dir->ss_info_exclude : NULL);
2346 }
039bc64e 2347}
4a92d1bf
AR
2348
2349int remove_path(const char *name)
2350{
2351 char *slash;
2352
c7054209 2353 if (unlink(name) && !is_missing_file_error(errno))
4a92d1bf
AR
2354 return -1;
2355
2356 slash = strrchr(name, '/');
2357 if (slash) {
2358 char *dirs = xstrdup(name);
2359 slash = dirs + (slash - name);
2360 do {
2361 *slash = '\0';
3fc0d131 2362 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
4a92d1bf
AR
2363 free(dirs);
2364 }
2365 return 0;
2366}
2367
270be816
AS
2368/*
2369 * Frees memory within dir which was allocated for exclude lists and
2370 * the exclude_stack. Does not free dir itself.
2371 */
2372void clear_directory(struct dir_struct *dir)
2373{
2374 int i, j;
2375 struct exclude_list_group *group;
2376 struct exclude_list *el;
2377 struct exclude_stack *stk;
2378
2379 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2380 group = &dir->exclude_list_group[i];
2381 for (j = 0; j < group->nr; j++) {
2382 el = &group->el[j];
2383 if (i == EXC_DIRS)
2384 free((char *)el->src);
2385 clear_exclude_list(el);
2386 }
2387 free(group->el);
2388 }
2389
2390 stk = dir->exclude_stack;
2391 while (stk) {
2392 struct exclude_stack *prev = stk->prev;
2393 free(stk);
2394 stk = prev;
2395 }
aceb9429 2396 strbuf_release(&dir->basebuf);
270be816 2397}
83c094ad
NTND
2398
2399struct ondisk_untracked_cache {
2400 struct stat_data info_exclude_stat;
2401 struct stat_data excludes_file_stat;
2402 uint32_t dir_flags;
2403 unsigned char info_exclude_sha1[20];
2404 unsigned char excludes_file_sha1[20];
2405 char exclude_per_dir[FLEX_ARRAY];
2406};
2407
268ba201
RS
2408#define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2409#define ouc_size(len) (ouc_offset(exclude_per_dir) + len + 1)
83c094ad
NTND
2410
2411struct write_data {
2412 int index; /* number of written untracked_cache_dir */
2413 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2414 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2415 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2416 struct strbuf out;
2417 struct strbuf sb_stat;
2418 struct strbuf sb_sha1;
2419};
2420
2421static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2422{
2423 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2424 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2425 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2426 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2427 to->sd_dev = htonl(from->sd_dev);
2428 to->sd_ino = htonl(from->sd_ino);
2429 to->sd_uid = htonl(from->sd_uid);
2430 to->sd_gid = htonl(from->sd_gid);
2431 to->sd_size = htonl(from->sd_size);
2432}
2433
2434static void write_one_dir(struct untracked_cache_dir *untracked,
2435 struct write_data *wd)
2436{
2437 struct stat_data stat_data;
2438 struct strbuf *out = &wd->out;
2439 unsigned char intbuf[16];
2440 unsigned int intlen, value;
2441 int i = wd->index++;
2442
2443 /*
2444 * untracked_nr should be reset whenever valid is clear, but
2445 * for safety..
2446 */
2447 if (!untracked->valid) {
2448 untracked->untracked_nr = 0;
2449 untracked->check_only = 0;
2450 }
2451
2452 if (untracked->check_only)
2453 ewah_set(wd->check_only, i);
2454 if (untracked->valid) {
2455 ewah_set(wd->valid, i);
2456 stat_data_to_disk(&stat_data, &untracked->stat_data);
2457 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2458 }
2459 if (!is_null_sha1(untracked->exclude_sha1)) {
2460 ewah_set(wd->sha1_valid, i);
2461 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2462 }
2463
2464 intlen = encode_varint(untracked->untracked_nr, intbuf);
2465 strbuf_add(out, intbuf, intlen);
2466
2467 /* skip non-recurse directories */
2468 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2469 if (untracked->dirs[i]->recurse)
2470 value++;
2471 intlen = encode_varint(value, intbuf);
2472 strbuf_add(out, intbuf, intlen);
2473
2474 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2475
2476 for (i = 0; i < untracked->untracked_nr; i++)
2477 strbuf_add(out, untracked->untracked[i],
2478 strlen(untracked->untracked[i]) + 1);
2479
2480 for (i = 0; i < untracked->dirs_nr; i++)
2481 if (untracked->dirs[i]->recurse)
2482 write_one_dir(untracked->dirs[i], wd);
2483}
2484
2485void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2486{
2487 struct ondisk_untracked_cache *ouc;
2488 struct write_data wd;
2489 unsigned char varbuf[16];
e0b83735
JK
2490 int varint_len;
2491 size_t len = strlen(untracked->exclude_per_dir);
2492
2493 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
83c094ad
NTND
2494 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2495 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2496 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2497 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2498 ouc->dir_flags = htonl(untracked->dir_flags);
1e8fef60
NTND
2499
2500 varint_len = encode_varint(untracked->ident.len, varbuf);
2501 strbuf_add(out, varbuf, varint_len);
8109984d 2502 strbuf_addbuf(out, &untracked->ident);
1e8fef60 2503
83c094ad 2504 strbuf_add(out, ouc, ouc_size(len));
6a83d902 2505 FREE_AND_NULL(ouc);
83c094ad
NTND
2506
2507 if (!untracked->root) {
2508 varint_len = encode_varint(0, varbuf);
2509 strbuf_add(out, varbuf, varint_len);
2510 return;
2511 }
2512
2513 wd.index = 0;
2514 wd.check_only = ewah_new();
2515 wd.valid = ewah_new();
2516 wd.sha1_valid = ewah_new();
2517 strbuf_init(&wd.out, 1024);
2518 strbuf_init(&wd.sb_stat, 1024);
2519 strbuf_init(&wd.sb_sha1, 1024);
2520 write_one_dir(untracked->root, &wd);
2521
2522 varint_len = encode_varint(wd.index, varbuf);
2523 strbuf_add(out, varbuf, varint_len);
2524 strbuf_addbuf(out, &wd.out);
2525 ewah_serialize_strbuf(wd.valid, out);
2526 ewah_serialize_strbuf(wd.check_only, out);
2527 ewah_serialize_strbuf(wd.sha1_valid, out);
2528 strbuf_addbuf(out, &wd.sb_stat);
2529 strbuf_addbuf(out, &wd.sb_sha1);
2530 strbuf_addch(out, '\0'); /* safe guard for string lists */
2531
2532 ewah_free(wd.valid);
2533 ewah_free(wd.check_only);
2534 ewah_free(wd.sha1_valid);
2535 strbuf_release(&wd.out);
2536 strbuf_release(&wd.sb_stat);
2537 strbuf_release(&wd.sb_sha1);
2538}
f9e6c649
NTND
2539
2540static void free_untracked(struct untracked_cache_dir *ucd)
2541{
2542 int i;
2543 if (!ucd)
2544 return;
2545 for (i = 0; i < ucd->dirs_nr; i++)
2546 free_untracked(ucd->dirs[i]);
2547 for (i = 0; i < ucd->untracked_nr; i++)
2548 free(ucd->untracked[i]);
2549 free(ucd->untracked);
2550 free(ucd->dirs);
2551 free(ucd);
2552}
2553
2554void free_untracked_cache(struct untracked_cache *uc)
2555{
2556 if (uc)
2557 free_untracked(uc->root);
2558 free(uc);
2559}
2560
2561struct read_data {
2562 int index;
2563 struct untracked_cache_dir **ucd;
2564 struct ewah_bitmap *check_only;
2565 struct ewah_bitmap *valid;
2566 struct ewah_bitmap *sha1_valid;
2567 const unsigned char *data;
2568 const unsigned char *end;
2569};
2570
268ba201 2571static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
f9e6c649 2572{
268ba201
RS
2573 memcpy(to, data, sizeof(*to));
2574 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2575 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2576 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2577 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2578 to->sd_dev = ntohl(to->sd_dev);
2579 to->sd_ino = ntohl(to->sd_ino);
2580 to->sd_uid = ntohl(to->sd_uid);
2581 to->sd_gid = ntohl(to->sd_gid);
2582 to->sd_size = ntohl(to->sd_size);
f9e6c649
NTND
2583}
2584
2585static int read_one_dir(struct untracked_cache_dir **untracked_,
2586 struct read_data *rd)
2587{
2588 struct untracked_cache_dir ud, *untracked;
2589 const unsigned char *next, *data = rd->data, *end = rd->end;
2590 unsigned int value;
2591 int i, len;
2592
2593 memset(&ud, 0, sizeof(ud));
2594
2595 next = data;
2596 value = decode_varint(&next);
2597 if (next > end)
2598 return -1;
2599 ud.recurse = 1;
2600 ud.untracked_alloc = value;
2601 ud.untracked_nr = value;
2602 if (ud.untracked_nr)
b32fa95f 2603 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
f9e6c649
NTND
2604 data = next;
2605
2606 next = data;
2607 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2608 if (next > end)
2609 return -1;
b32fa95f 2610 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
f9e6c649
NTND
2611 data = next;
2612
2613 len = strlen((const char *)data);
2614 next = data + len + 1;
2615 if (next > rd->end)
2616 return -1;
50a6c8ef 2617 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
f9e6c649
NTND
2618 memcpy(untracked, &ud, sizeof(ud));
2619 memcpy(untracked->name, data, len + 1);
2620 data = next;
2621
2622 for (i = 0; i < untracked->untracked_nr; i++) {
2623 len = strlen((const char *)data);
2624 next = data + len + 1;
2625 if (next > rd->end)
2626 return -1;
2627 untracked->untracked[i] = xstrdup((const char*)data);
2628 data = next;
2629 }
2630
2631 rd->ucd[rd->index++] = untracked;
2632 rd->data = data;
2633
2634 for (i = 0; i < untracked->dirs_nr; i++) {
2635 len = read_one_dir(untracked->dirs + i, rd);
2636 if (len < 0)
2637 return -1;
2638 }
2639 return 0;
2640}
2641
2642static void set_check_only(size_t pos, void *cb)
2643{
2644 struct read_data *rd = cb;
2645 struct untracked_cache_dir *ud = rd->ucd[pos];
2646 ud->check_only = 1;
2647}
2648
2649static void read_stat(size_t pos, void *cb)
2650{
2651 struct read_data *rd = cb;
2652 struct untracked_cache_dir *ud = rd->ucd[pos];
2653 if (rd->data + sizeof(struct stat_data) > rd->end) {
2654 rd->data = rd->end + 1;
2655 return;
2656 }
268ba201 2657 stat_data_from_disk(&ud->stat_data, rd->data);
f9e6c649
NTND
2658 rd->data += sizeof(struct stat_data);
2659 ud->valid = 1;
2660}
2661
2662static void read_sha1(size_t pos, void *cb)
2663{
2664 struct read_data *rd = cb;
2665 struct untracked_cache_dir *ud = rd->ucd[pos];
2666 if (rd->data + 20 > rd->end) {
2667 rd->data = rd->end + 1;
2668 return;
2669 }
2670 hashcpy(ud->exclude_sha1, rd->data);
2671 rd->data += 20;
2672}
2673
2674static void load_sha1_stat(struct sha1_stat *sha1_stat,
268ba201 2675 const unsigned char *data,
f9e6c649
NTND
2676 const unsigned char *sha1)
2677{
268ba201 2678 stat_data_from_disk(&sha1_stat->stat, data);
f9e6c649
NTND
2679 hashcpy(sha1_stat->sha1, sha1);
2680 sha1_stat->valid = 1;
2681}
2682
2683struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2684{
f9e6c649
NTND
2685 struct untracked_cache *uc;
2686 struct read_data rd;
2687 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
1e8fef60
NTND
2688 const char *ident;
2689 int ident_len, len;
268ba201 2690 const char *exclude_per_dir;
f9e6c649
NTND
2691
2692 if (sz <= 1 || end[-1] != '\0')
2693 return NULL;
2694 end--;
2695
1e8fef60
NTND
2696 ident_len = decode_varint(&next);
2697 if (next + ident_len > end)
2698 return NULL;
2699 ident = (const char *)next;
2700 next += ident_len;
2701
f9e6c649
NTND
2702 if (next + ouc_size(0) > end)
2703 return NULL;
2704
2705 uc = xcalloc(1, sizeof(*uc));
1e8fef60
NTND
2706 strbuf_init(&uc->ident, ident_len);
2707 strbuf_add(&uc->ident, ident, ident_len);
268ba201
RS
2708 load_sha1_stat(&uc->ss_info_exclude,
2709 next + ouc_offset(info_exclude_stat),
2710 next + ouc_offset(info_exclude_sha1));
2711 load_sha1_stat(&uc->ss_excludes_file,
2712 next + ouc_offset(excludes_file_stat),
2713 next + ouc_offset(excludes_file_sha1));
2714 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2715 exclude_per_dir = (const char *)next + ouc_offset(exclude_per_dir);
2716 uc->exclude_per_dir = xstrdup(exclude_per_dir);
f9e6c649 2717 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
268ba201 2718 next += ouc_size(strlen(exclude_per_dir));
f9e6c649
NTND
2719 if (next >= end)
2720 goto done2;
2721
2722 len = decode_varint(&next);
2723 if (next > end || len == 0)
2724 goto done2;
2725
2726 rd.valid = ewah_new();
2727 rd.check_only = ewah_new();
2728 rd.sha1_valid = ewah_new();
2729 rd.data = next;
2730 rd.end = end;
2731 rd.index = 0;
b32fa95f 2732 ALLOC_ARRAY(rd.ucd, len);
f9e6c649
NTND
2733
2734 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2735 goto done;
2736
2737 next = rd.data;
2738 len = ewah_read_mmap(rd.valid, next, end - next);
2739 if (len < 0)
2740 goto done;
2741
2742 next += len;
2743 len = ewah_read_mmap(rd.check_only, next, end - next);
2744 if (len < 0)
2745 goto done;
2746
2747 next += len;
2748 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2749 if (len < 0)
2750 goto done;
2751
2752 ewah_each_bit(rd.check_only, set_check_only, &rd);
2753 rd.data = next + len;
2754 ewah_each_bit(rd.valid, read_stat, &rd);
2755 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2756 next = rd.data;
2757
2758done:
2759 free(rd.ucd);
2760 ewah_free(rd.valid);
2761 ewah_free(rd.check_only);
2762 ewah_free(rd.sha1_valid);
2763done2:
2764 if (next != end) {
2765 free_untracked_cache(uc);
2766 uc = NULL;
2767 }
2768 return uc;
2769}
e931371a 2770
73f9145f
NTND
2771static void invalidate_one_directory(struct untracked_cache *uc,
2772 struct untracked_cache_dir *ucd)
2773{
2774 uc->dir_invalidated++;
2775 ucd->valid = 0;
2776 ucd->untracked_nr = 0;
2777}
2778
2779/*
2780 * Normally when an entry is added or removed from a directory,
2781 * invalidating that directory is enough. No need to touch its
2782 * ancestors. When a directory is shown as "foo/bar/" in git-status
2783 * however, deleting or adding an entry may have cascading effect.
2784 *
2785 * Say the "foo/bar/file" has become untracked, we need to tell the
2786 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2787 * directory any more (because "bar" is managed by foo as an untracked
2788 * "file").
2789 *
2790 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2791 * was the last untracked entry in the entire "foo", we should show
2792 * "foo/" instead. Which means we have to invalidate past "bar" up to
2793 * "foo".
2794 *
2795 * This function traverses all directories from root to leaf. If there
2796 * is a chance of one of the above cases happening, we invalidate back
2797 * to root. Otherwise we just invalidate the leaf. There may be a more
2798 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2799 * detect these cases and avoid unnecessary invalidation, for example,
2800 * checking for the untracked entry named "bar/" in "foo", but for now
2801 * stick to something safe and simple.
2802 */
2803static int invalidate_one_component(struct untracked_cache *uc,
2804 struct untracked_cache_dir *dir,
2805 const char *path, int len)
2806{
2807 const char *rest = strchr(path, '/');
2808
2809 if (rest) {
2810 int component_len = rest - path;
2811 struct untracked_cache_dir *d =
2812 lookup_untracked(uc, dir, path, component_len);
2813 int ret =
2814 invalidate_one_component(uc, d, rest + 1,
2815 len - (component_len + 1));
2816 if (ret)
2817 invalidate_one_directory(uc, dir);
2818 return ret;
2819 }
2820
2821 invalidate_one_directory(uc, dir);
2822 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2823}
2824
e931371a
NTND
2825void untracked_cache_invalidate_path(struct index_state *istate,
2826 const char *path)
2827{
e931371a
NTND
2828 if (!istate->untracked || !istate->untracked->root)
2829 return;
73f9145f
NTND
2830 invalidate_one_component(istate->untracked, istate->untracked->root,
2831 path, strlen(path));
e931371a
NTND
2832}
2833
2834void untracked_cache_remove_from_index(struct index_state *istate,
2835 const char *path)
2836{
2837 untracked_cache_invalidate_path(istate, path);
2838}
2839
2840void untracked_cache_add_to_index(struct index_state *istate,
2841 const char *path)
2842{
2843 untracked_cache_invalidate_path(istate, path);
2844}
47e83eb3
SB
2845
2846/* Update gitfile and core.worktree setting to connect work tree and git dir */
2847void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)
2848{
365444a6
SB
2849 struct strbuf gitfile_sb = STRBUF_INIT;
2850 struct strbuf cfg_sb = STRBUF_INIT;
47e83eb3 2851 struct strbuf rel_path = STRBUF_INIT;
365444a6 2852 char *git_dir, *work_tree;
47e83eb3 2853
365444a6
SB
2854 /* Prepare .git file */
2855 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
2856 if (safe_create_leading_directories_const(gitfile_sb.buf))
2857 die(_("could not create directories for %s"), gitfile_sb.buf);
2858
2859 /* Prepare config file */
2860 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
2861 if (safe_create_leading_directories_const(cfg_sb.buf))
2862 die(_("could not create directories for %s"), cfg_sb.buf);
47e83eb3 2863
e394fa01
JH
2864 git_dir = real_pathdup(git_dir_, 1);
2865 work_tree = real_pathdup(work_tree_, 1);
365444a6
SB
2866
2867 /* Write .git file */
2868 write_file(gitfile_sb.buf, "gitdir: %s",
2869 relative_path(git_dir, work_tree, &rel_path));
47e83eb3 2870 /* Update core.worktree setting */
365444a6 2871 git_config_set_in_file(cfg_sb.buf, "core.worktree",
47e83eb3
SB
2872 relative_path(work_tree, git_dir, &rel_path));
2873
365444a6
SB
2874 strbuf_release(&gitfile_sb);
2875 strbuf_release(&cfg_sb);
47e83eb3
SB
2876 strbuf_release(&rel_path);
2877 free(work_tree);
2878 free(git_dir);
2879}
f6f85861
SB
2880
2881/*
2882 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
2883 */
2884void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
2885{
2886 if (rename(old_git_dir, new_git_dir) < 0)
2887 die_errno(_("could not migrate git directory from '%s' to '%s'"),
2888 old_git_dir, new_git_dir);
2889
2890 connect_work_tree_and_git_dir(path, new_git_dir);
2891}