submodule-config: remove submodule_from_cache
[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,
5aaa7fd3 53 int check_only, int stop_at_first_file, 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
578d81d0
JH
224/*
225 * Read the contents of the blob with the given OID into a buffer.
226 * Append a trailing LF to the end if the last line doesn't have one.
227 *
228 * Returns:
229 * -1 when the OID is invalid or unknown or does not refer to a blob.
230 * 0 when the blob is empty.
231 * 1 along with { data, size } of the (possibly augmented) buffer
232 * when successful.
233 *
234 * Optionally updates the given sha1_stat with the given OID (when valid).
235 */
236static int do_read_blob(const struct object_id *oid,
237 struct sha1_stat *sha1_stat,
238 size_t *size_out,
239 char **data_out)
240{
241 enum object_type type;
242 unsigned long sz;
243 char *data;
244
245 *size_out = 0;
246 *data_out = NULL;
247
248 data = read_sha1_file(oid->hash, &type, &sz);
249 if (!data || type != OBJ_BLOB) {
250 free(data);
251 return -1;
252 }
253
254 if (sha1_stat) {
255 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
256 hashcpy(sha1_stat->sha1, oid->hash);
257 }
258
259 if (sz == 0) {
260 free(data);
261 return 0;
262 }
263
264 if (data[sz - 1] != '\n') {
265 data = xrealloc(data, st_add(sz, 1));
266 data[sz++] = '\n';
267 }
268
269 *size_out = xsize_t(sz);
270 *data_out = data;
271
272 return 1;
273}
274
75a6315f
BW
275#define DO_MATCH_EXCLUDE (1<<0)
276#define DO_MATCH_DIRECTORY (1<<1)
277#define DO_MATCH_SUBMODULE (1<<2)
42b0874a 278
b0db7046
BW
279static int match_attrs(const char *name, int namelen,
280 const struct pathspec_item *item)
281{
282 int i;
283
284 git_check_attr(name, item->attr_check);
285 for (i = 0; i < item->attr_match_nr; i++) {
286 const char *value;
287 int matched;
288 enum attr_match_mode match_mode;
289
290 value = item->attr_check->items[i].value;
291 match_mode = item->attr_match[i].match_mode;
292
293 if (ATTR_TRUE(value))
294 matched = (match_mode == MATCH_SET);
295 else if (ATTR_FALSE(value))
296 matched = (match_mode == MATCH_UNSET);
297 else if (ATTR_UNSET(value))
298 matched = (match_mode == MATCH_UNSPECIFIED);
299 else
300 matched = (match_mode == MATCH_VALUE &&
301 !strcmp(item->attr_match[i].value, value));
302 if (!matched)
303 return 0;
304 }
305
306 return 1;
307}
308
61cf2820
NTND
309/*
310 * Does 'match' match the given name?
311 * A match is found if
312 *
313 * (1) the 'match' string is leading directory of 'name', or
314 * (2) the 'match' string is a wildcard and matches 'name', or
315 * (3) the 'match' string is exactly the same as 'name'.
316 *
317 * and the return value tells which case it was.
318 *
319 * It returns 0 when there is no match.
320 */
321static int match_pathspec_item(const struct pathspec_item *item, int prefix,
42b0874a 322 const char *name, int namelen, unsigned flags)
61cf2820
NTND
323{
324 /* name/namelen has prefix cut off by caller */
325 const char *match = item->match + prefix;
326 int matchlen = item->len - prefix;
327
93d93537
NTND
328 /*
329 * The normal call pattern is:
330 * 1. prefix = common_prefix_len(ps);
331 * 2. prune something, or fill_directory
854b0959 332 * 3. match_pathspec()
93d93537
NTND
333 *
334 * 'prefix' at #1 may be shorter than the command's prefix and
335 * it's ok for #2 to match extra files. Those extras will be
336 * trimmed at #3.
337 *
338 * Suppose the pathspec is 'foo' and '../bar' running from
339 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
340 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
341 * user does not want XYZ/foo, only the "foo" part should be
342 * case-insensitive. We need to filter out XYZ/foo here. In
343 * other words, we do not trust the caller on comparing the
344 * prefix part when :(icase) is involved. We do exact
345 * comparison ourselves.
346 *
347 * Normally the caller (common_prefix_len() in fact) does
348 * _exact_ matching on name[-prefix+1..-1] and we do not need
349 * to check that part. Be defensive and check it anyway, in
350 * case common_prefix_len is changed, or a new caller is
351 * introduced that does not use common_prefix_len.
352 *
353 * If the penalty turns out too high when prefix is really
354 * long, maybe change it to
355 * strncmp(match, name, item->prefix - prefix)
356 */
357 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
358 strncmp(item->match, name - prefix, item->prefix))
359 return 0;
360
b0db7046
BW
361 if (item->attr_match_nr && !match_attrs(name, namelen, item))
362 return 0;
363
61cf2820
NTND
364 /* If the match was just the prefix, we matched */
365 if (!*match)
366 return MATCHED_RECURSIVELY;
367
93d93537 368 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
61cf2820
NTND
369 if (matchlen == namelen)
370 return MATCHED_EXACTLY;
371
372 if (match[matchlen-1] == '/' || name[matchlen] == '/')
373 return MATCHED_RECURSIVELY;
68690fdd
NTND
374 } else if ((flags & DO_MATCH_DIRECTORY) &&
375 match[matchlen - 1] == '/' &&
376 namelen == matchlen - 1 &&
377 !ps_strncmp(item, match, name, namelen))
378 return MATCHED_EXACTLY;
61cf2820 379
5d74762d 380 if (item->nowildcard_len < item->len &&
bd30c2e4 381 !git_fnmatch(item, match, name,
8c6abbcd 382 item->nowildcard_len - prefix))
61cf2820
NTND
383 return MATCHED_FNMATCH;
384
75a6315f
BW
385 /* Perform checks to see if "name" is a super set of the pathspec */
386 if (flags & DO_MATCH_SUBMODULE) {
387 /* name is a literal prefix of the pathspec */
388 if ((namelen < matchlen) &&
389 (match[namelen] == '/') &&
390 !ps_strncmp(item, match, name, namelen))
391 return MATCHED_RECURSIVELY;
392
393 /* name" doesn't match up to the first wild character */
394 if (item->nowildcard_len < item->len &&
395 ps_strncmp(item, match, name,
396 item->nowildcard_len - prefix))
397 return 0;
398
399 /*
400 * Here is where we would perform a wildmatch to check if
401 * "name" can be matched as a directory (or a prefix) against
402 * the pathspec. Since wildmatch doesn't have this capability
403 * at the present we have to punt and say that it is a match,
404 * potentially returning a false positive
405 * The submodules themselves will be able to perform more
406 * accurate matching to determine if the pathspec matches.
407 */
408 return MATCHED_RECURSIVELY;
409 }
410
61cf2820
NTND
411 return 0;
412}
413
414/*
52ed1894
AS
415 * Given a name and a list of pathspecs, returns the nature of the
416 * closest (i.e. most specific) match of the name to any of the
417 * pathspecs.
418 *
419 * The caller typically calls this multiple times with the same
420 * pathspec and seen[] array but with different name/namelen
421 * (e.g. entries from the index) and is interested in seeing if and
422 * how each pathspec matches all the names it calls this function
423 * with. A mark is left in the seen[] array for each pathspec element
424 * indicating the closest type of match that element achieved, so if
425 * seen[n] remains zero after multiple invocations, that means the nth
426 * pathspec did not match any names, which could indicate that the
427 * user mistyped the nth pathspec.
61cf2820 428 */
854b0959
NTND
429static int do_match_pathspec(const struct pathspec *ps,
430 const char *name, int namelen,
431 int prefix, char *seen,
42b0874a 432 unsigned flags)
61cf2820 433{
42b0874a 434 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
61cf2820 435
5c6933d2
NTND
436 GUARD_PATHSPEC(ps,
437 PATHSPEC_FROMTOP |
438 PATHSPEC_MAXDEPTH |
bd30c2e4 439 PATHSPEC_LITERAL |
93d93537 440 PATHSPEC_GLOB |
ef79b1f8 441 PATHSPEC_ICASE |
b0db7046
BW
442 PATHSPEC_EXCLUDE |
443 PATHSPEC_ATTR);
8f4f8f45 444
61cf2820 445 if (!ps->nr) {
6330a171
NTND
446 if (!ps->recursive ||
447 !(ps->magic & PATHSPEC_MAXDEPTH) ||
448 ps->max_depth == -1)
61cf2820
NTND
449 return MATCHED_RECURSIVELY;
450
451 if (within_depth(name, namelen, 0, ps->max_depth))
452 return MATCHED_EXACTLY;
453 else
454 return 0;
455 }
456
457 name += prefix;
458 namelen -= prefix;
459
460 for (i = ps->nr - 1; i >= 0; i--) {
461 int how;
ef79b1f8
NTND
462
463 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
464 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
465 continue;
466
61cf2820
NTND
467 if (seen && seen[i] == MATCHED_EXACTLY)
468 continue;
ef79b1f8
NTND
469 /*
470 * Make exclude patterns optional and never report
471 * "pathspec ':(exclude)foo' matches no files"
472 */
473 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
474 seen[i] = MATCHED_FNMATCH;
42b0874a
NTND
475 how = match_pathspec_item(ps->items+i, prefix, name,
476 namelen, flags);
6330a171
NTND
477 if (ps->recursive &&
478 (ps->magic & PATHSPEC_MAXDEPTH) &&
479 ps->max_depth != -1 &&
61cf2820
NTND
480 how && how != MATCHED_FNMATCH) {
481 int len = ps->items[i].len;
482 if (name[len] == '/')
483 len++;
484 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
485 how = MATCHED_EXACTLY;
486 else
487 how = 0;
488 }
489 if (how) {
490 if (retval < how)
491 retval = how;
492 if (seen && seen[i] < how)
493 seen[i] = how;
494 }
495 }
496 return retval;
497}
498
854b0959
NTND
499int match_pathspec(const struct pathspec *ps,
500 const char *name, int namelen,
ae8d0824 501 int prefix, char *seen, int is_dir)
ef79b1f8
NTND
502{
503 int positive, negative;
ae8d0824 504 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
42b0874a
NTND
505 positive = do_match_pathspec(ps, name, namelen,
506 prefix, seen, flags);
ef79b1f8
NTND
507 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
508 return positive;
42b0874a
NTND
509 negative = do_match_pathspec(ps, name, namelen,
510 prefix, seen,
511 flags | DO_MATCH_EXCLUDE);
ef79b1f8
NTND
512 return negative ? 0 : positive;
513}
514
75a6315f
BW
515/**
516 * Check if a submodule is a superset of the pathspec
517 */
518int submodule_path_match(const struct pathspec *ps,
519 const char *submodule_name,
520 char *seen)
521{
522 int matched = do_match_pathspec(ps, submodule_name,
523 strlen(submodule_name),
524 0, seen,
525 DO_MATCH_DIRECTORY |
526 DO_MATCH_SUBMODULE);
527 return matched;
528}
529
777c55a6
JH
530int report_path_error(const char *ps_matched,
531 const struct pathspec *pathspec,
532 const char *prefix)
533{
534 /*
535 * Make sure all pathspec matched; otherwise it is an error.
536 */
777c55a6
JH
537 int num, errors = 0;
538 for (num = 0; num < pathspec->nr; num++) {
539 int other, found_dup;
540
541 if (ps_matched[num])
542 continue;
543 /*
544 * The caller might have fed identical pathspec
545 * twice. Do not barf on such a mistake.
546 * FIXME: parse_pathspec should have eliminated
547 * duplicate pathspec.
548 */
549 for (found_dup = other = 0;
550 !found_dup && other < pathspec->nr;
551 other++) {
552 if (other == num || !ps_matched[other])
553 continue;
554 if (!strcmp(pathspec->items[other].original,
555 pathspec->items[num].original))
556 /*
557 * Ok, we have a match already.
558 */
559 found_dup = 1;
560 }
561 if (found_dup)
562 continue;
563
564 error("pathspec '%s' did not match any file(s) known to git.",
565 pathspec->items[num].original);
566 errors++;
567 }
777c55a6
JH
568 return errors;
569}
570
fcd631ed
NTND
571/*
572 * Return the length of the "simple" part of a path match limiter.
573 */
87323bda 574int simple_length(const char *match)
fcd631ed
NTND
575{
576 int len = -1;
577
578 for (;;) {
579 unsigned char c = *match++;
580 len++;
581 if (c == '\0' || is_glob_special(c))
582 return len;
583 }
584}
585
87323bda 586int no_wildcard(const char *string)
68492fc7 587{
fcd631ed 588 return string[simple_length(string)] == '\0';
68492fc7
LK
589}
590
82dce998
NTND
591void parse_exclude_pattern(const char **pattern,
592 int *patternlen,
f8708998 593 unsigned *flags,
82dce998 594 int *nowildcardlen)
84460eec
NTND
595{
596 const char *p = *pattern;
597 size_t i, len;
598
599 *flags = 0;
600 if (*p == '!') {
601 *flags |= EXC_FLAG_NEGATIVE;
602 p++;
603 }
604 len = strlen(p);
605 if (len && p[len - 1] == '/') {
606 len--;
607 *flags |= EXC_FLAG_MUSTBEDIR;
608 }
609 for (i = 0; i < len; i++) {
610 if (p[i] == '/')
611 break;
612 }
613 if (i == len)
614 *flags |= EXC_FLAG_NODIR;
615 *nowildcardlen = simple_length(p);
616 /*
617 * we should have excluded the trailing slash from 'p' too,
618 * but that's one more allocation. Instead just make sure
619 * nowildcardlen does not exceed real patternlen
620 */
621 if (*nowildcardlen > len)
622 *nowildcardlen = len;
623 if (*p == '*' && no_wildcard(p + 1))
624 *flags |= EXC_FLAG_ENDSWITH;
625 *pattern = p;
626 *patternlen = len;
627}
628
453ec4bd 629void add_exclude(const char *string, const char *base,
c04318e4 630 int baselen, struct exclude_list *el, int srcpos)
453ec4bd 631{
d6b8fc30 632 struct exclude *x;
84460eec 633 int patternlen;
f8708998 634 unsigned flags;
84460eec 635 int nowildcardlen;
453ec4bd 636
84460eec
NTND
637 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
638 if (flags & EXC_FLAG_MUSTBEDIR) {
96ffc06f 639 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
d6b8fc30
JH
640 } else {
641 x = xmalloc(sizeof(*x));
642 x->pattern = string;
643 }
84460eec
NTND
644 x->patternlen = patternlen;
645 x->nowildcardlen = nowildcardlen;
453ec4bd
LT
646 x->base = base;
647 x->baselen = baselen;
d6b8fc30 648 x->flags = flags;
c04318e4 649 x->srcpos = srcpos;
840fc334
AS
650 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
651 el->excludes[el->nr++] = x;
c04318e4 652 x->el = el;
453ec4bd
LT
653}
654
578d81d0
JH
655static int read_skip_worktree_file_from_index(const struct index_state *istate,
656 const char *path,
657 size_t *size_out,
658 char **data_out,
659 struct sha1_stat *sha1_stat)
c28b3d6e
NTND
660{
661 int pos, len;
c28b3d6e
NTND
662
663 len = strlen(path);
6f52b741 664 pos = index_name_pos(istate, path, len);
c28b3d6e 665 if (pos < 0)
578d81d0 666 return -1;
6f52b741 667 if (!ce_skip_worktree(istate->cache[pos]))
578d81d0
JH
668 return -1;
669
670 return do_read_blob(&istate->cache[pos]->oid, sha1_stat, size_out, data_out);
c28b3d6e
NTND
671}
672
f6198812
AS
673/*
674 * Frees memory within el which was allocated for exclude patterns and
675 * the file buffer. Does not free el itself.
676 */
677void clear_exclude_list(struct exclude_list *el)
0fd0e241
NTND
678{
679 int i;
680
5cee3493 681 for (i = 0; i < el->nr; i++)
0fd0e241
NTND
682 free(el->excludes[i]);
683 free(el->excludes);
c082df24 684 free(el->filebuf);
0fd0e241 685
2653a8c6 686 memset(el, 0, sizeof(*el));
0fd0e241
NTND
687}
688
7e2e4b37 689static void trim_trailing_spaces(char *buf)
16402b99 690{
e61a6c1d
PB
691 char *p, *last_space = NULL;
692
693 for (p = buf; *p; p++)
694 switch (*p) {
695 case ' ':
696 if (!last_space)
697 last_space = p;
698 break;
699 case '\\':
700 p++;
701 if (!*p)
702 return;
703 /* fallthrough */
704 default:
705 last_space = NULL;
706 }
707
708 if (last_space)
709 *last_space = '\0';
16402b99
NTND
710}
711
0dcb8d7f
NTND
712/*
713 * Given a subdirectory name and "dir" of the current directory,
714 * search the subdir in "dir" and return it, or create a new one if it
715 * does not exist in "dir".
716 *
717 * If "name" has the trailing slash, it'll be excluded in the search.
718 */
719static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
720 struct untracked_cache_dir *dir,
721 const char *name, int len)
722{
723 int first, last;
724 struct untracked_cache_dir *d;
725 if (!dir)
726 return NULL;
727 if (len && name[len - 1] == '/')
728 len--;
729 first = 0;
730 last = dir->dirs_nr;
731 while (last > first) {
732 int cmp, next = (last + first) >> 1;
733 d = dir->dirs[next];
734 cmp = strncmp(name, d->name, len);
735 if (!cmp && strlen(d->name) > len)
736 cmp = -1;
737 if (!cmp)
738 return d;
739 if (cmp < 0) {
740 last = next;
741 continue;
742 }
743 first = next+1;
744 }
745
746 uc->dir_created++;
96ffc06f 747 FLEX_ALLOC_MEM(d, name, name, len);
0dcb8d7f
NTND
748
749 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
f919ffeb
SG
750 MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
751 dir->dirs_nr - first);
0dcb8d7f
NTND
752 dir->dirs_nr++;
753 dir->dirs[first] = d;
754 return d;
755}
756
ccad261f
NTND
757static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
758{
759 int i;
760 dir->valid = 0;
761 dir->untracked_nr = 0;
762 for (i = 0; i < dir->dirs_nr; i++)
763 do_invalidate_gitignore(dir->dirs[i]);
764}
765
766static void invalidate_gitignore(struct untracked_cache *uc,
767 struct untracked_cache_dir *dir)
768{
769 uc->gitignore_invalidated++;
770 do_invalidate_gitignore(dir);
771}
772
91a2288b
NTND
773static void invalidate_directory(struct untracked_cache *uc,
774 struct untracked_cache_dir *dir)
775{
26cb0182 776 int i;
91a2288b
NTND
777 uc->dir_invalidated++;
778 dir->valid = 0;
779 dir->untracked_nr = 0;
26cb0182
NTND
780 for (i = 0; i < dir->dirs_nr; i++)
781 dir->dirs[i]->recurse = 0;
91a2288b
NTND
782}
783
578d81d0
JH
784static int add_excludes_from_buffer(char *buf, size_t size,
785 const char *base, int baselen,
786 struct exclude_list *el);
787
55fe6f51
NTND
788/*
789 * Given a file with name "fname", read it (either from disk, or from
473e3930 790 * an index if 'istate' is non-null), parse it and store the
55fe6f51
NTND
791 * exclude rules in "el".
792 *
793 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
794 * stat data from disk (only valid if add_excludes returns zero). If
795 * ss_valid is non-zero, "ss" must contain good value as input.
796 */
797static int add_excludes(const char *fname, const char *base, int baselen,
473e3930
BW
798 struct exclude_list *el,
799 struct index_state *istate,
55fe6f51 800 struct sha1_stat *sha1_stat)
453ec4bd 801{
c470701a 802 struct stat st;
578d81d0
JH
803 int r;
804 int fd;
9d14017a 805 size_t size = 0;
578d81d0 806 char *buf;
453ec4bd
LT
807
808 fd = open(fname, O_RDONLY);
c28b3d6e 809 if (fd < 0 || fstat(fd, &st) < 0) {
11dc1fcb
NTND
810 if (fd < 0)
811 warn_on_fopen_errors(fname);
812 else
c28b3d6e 813 close(fd);
578d81d0 814 if (!istate)
c28b3d6e 815 return -1;
578d81d0
JH
816 r = read_skip_worktree_file_from_index(istate, fname,
817 &size, &buf,
818 sha1_stat);
819 if (r != 1)
820 return r;
d961baa8 821 } else {
c28b3d6e
NTND
822 size = xsize_t(st.st_size);
823 if (size == 0) {
55fe6f51
NTND
824 if (sha1_stat) {
825 fill_stat_data(&sha1_stat->stat, &st);
826 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
827 sha1_stat->valid = 1;
828 }
c28b3d6e
NTND
829 close(fd);
830 return 0;
831 }
3733e694 832 buf = xmallocz(size);
c28b3d6e 833 if (read_in_full(fd, buf, size) != size) {
45d76f17 834 free(buf);
c28b3d6e
NTND
835 close(fd);
836 return -1;
837 }
45d76f17 838 buf[size++] = '\n';
c28b3d6e 839 close(fd);
55fe6f51
NTND
840 if (sha1_stat) {
841 int pos;
842 if (sha1_stat->valid &&
473e3930 843 !match_stat_data_racy(istate, &sha1_stat->stat, &st))
55fe6f51 844 ; /* no content change, ss->sha1 still good */
473e3930
BW
845 else if (istate &&
846 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
847 !ce_stage(istate->cache[pos]) &&
848 ce_uptodate(istate->cache[pos]) &&
82b474e0 849 !would_convert_to_git(istate, fname))
99d1a986 850 hashcpy(sha1_stat->sha1,
473e3930 851 istate->cache[pos]->oid.hash);
55fe6f51
NTND
852 else
853 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
854 fill_stat_data(&sha1_stat->stat, &st);
855 sha1_stat->valid = 1;
856 }
6ba78238 857 }
453ec4bd 858
578d81d0
JH
859 add_excludes_from_buffer(buf, size, base, baselen, el);
860 return 0;
861}
862
863static int add_excludes_from_buffer(char *buf, size_t size,
864 const char *base, int baselen,
865 struct exclude_list *el)
866{
867 int i, lineno = 1;
868 char *entry;
869
c082df24 870 el->filebuf = buf;
245e1c19 871
dde843e7
JH
872 if (skip_utf8_bom(&buf, size))
873 size -= buf - el->filebuf;
874
453ec4bd 875 entry = buf;
245e1c19 876
45d76f17
NTND
877 for (i = 0; i < size; i++) {
878 if (buf[i] == '\n') {
453ec4bd
LT
879 if (entry != buf + i && entry[0] != '#') {
880 buf[i - (i && buf[i-1] == '\r')] = 0;
7e2e4b37 881 trim_trailing_spaces(entry);
c04318e4 882 add_exclude(entry, base, baselen, el, lineno);
453ec4bd 883 }
c04318e4 884 lineno++;
453ec4bd
LT
885 entry = buf + i + 1;
886 }
887 }
888 return 0;
453ec4bd
LT
889}
890
55fe6f51
NTND
891int add_excludes_from_file_to_list(const char *fname, const char *base,
892 int baselen, struct exclude_list *el,
473e3930 893 struct index_state *istate)
55fe6f51 894{
473e3930 895 return add_excludes(fname, base, baselen, el, istate, NULL);
55fe6f51
NTND
896}
897
578d81d0
JH
898int add_excludes_from_blob_to_list(
899 struct object_id *oid,
900 const char *base, int baselen,
901 struct exclude_list *el)
902{
903 char *buf;
904 size_t size;
905 int r;
906
907 r = do_read_blob(oid, NULL, &size, &buf);
908 if (r != 1)
909 return r;
910
911 add_excludes_from_buffer(buf, size, base, baselen, el);
912 return 0;
913}
914
c04318e4
AS
915struct exclude_list *add_exclude_list(struct dir_struct *dir,
916 int group_type, const char *src)
c082df24
AS
917{
918 struct exclude_list *el;
919 struct exclude_list_group *group;
920
921 group = &dir->exclude_list_group[group_type];
922 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
923 el = &group->el[group->nr++];
924 memset(el, 0, sizeof(*el));
c04318e4 925 el->src = src;
c082df24
AS
926 return el;
927}
928
929/*
930 * Used to set up core.excludesfile and .git/info/exclude lists.
931 */
0dcb8d7f
NTND
932static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
933 struct sha1_stat *sha1_stat)
453ec4bd 934{
c082df24 935 struct exclude_list *el;
ccad261f
NTND
936 /*
937 * catch setup_standard_excludes() that's called before
938 * dir->untracked is assigned. That function behaves
939 * differently when dir->untracked is non-NULL.
940 */
941 if (!dir->untracked)
942 dir->unmanaged_exclude_files++;
c04318e4 943 el = add_exclude_list(dir, EXC_FILE, fname);
473e3930 944 if (add_excludes(fname, "", 0, el, NULL, sha1_stat) < 0)
453ec4bd
LT
945 die("cannot use %s as an exclude file", fname);
946}
947
0dcb8d7f
NTND
948void add_excludes_from_file(struct dir_struct *dir, const char *fname)
949{
ccad261f 950 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
0dcb8d7f
NTND
951 add_excludes_from_file_1(dir, fname, NULL);
952}
953
82dce998
NTND
954int match_basename(const char *basename, int basenamelen,
955 const char *pattern, int prefix, int patternlen,
f8708998 956 unsigned flags)
593cb880
NTND
957{
958 if (prefix == patternlen) {
0b6e56df 959 if (patternlen == basenamelen &&
ba0897e6 960 !fspathncmp(pattern, basename, basenamelen))
593cb880
NTND
961 return 1;
962 } else if (flags & EXC_FLAG_ENDSWITH) {
0b6e56df 963 /* "*literal" matching against "fooliteral" */
593cb880 964 if (patternlen - 1 <= basenamelen &&
ba0897e6 965 !fspathncmp(pattern + 1,
0b6e56df
JH
966 basename + basenamelen - (patternlen - 1),
967 patternlen - 1))
593cb880
NTND
968 return 1;
969 } else {
0b6e56df
JH
970 if (fnmatch_icase_mem(pattern, patternlen,
971 basename, basenamelen,
972 0) == 0)
593cb880
NTND
973 return 1;
974 }
975 return 0;
976}
977
82dce998
NTND
978int match_pathname(const char *pathname, int pathlen,
979 const char *base, int baselen,
980 const char *pattern, int prefix, int patternlen,
f8708998 981 unsigned flags)
b5592632
NTND
982{
983 const char *name;
984 int namelen;
985
986 /*
987 * match with FNM_PATHNAME; the pattern has base implicitly
988 * in front of it.
989 */
990 if (*pattern == '/') {
991 pattern++;
982ac873 992 patternlen--;
b5592632
NTND
993 prefix--;
994 }
995
996 /*
997 * baselen does not count the trailing slash. base[] may or
998 * may not end with a trailing slash though.
999 */
1000 if (pathlen < baselen + 1 ||
1001 (baselen && pathname[baselen] != '/') ||
ba0897e6 1002 fspathncmp(pathname, base, baselen))
b5592632
NTND
1003 return 0;
1004
1005 namelen = baselen ? pathlen - baselen - 1 : pathlen;
1006 name = pathname + pathlen - namelen;
1007
1008 if (prefix) {
1009 /*
1010 * if the non-wildcard part is longer than the
1011 * remaining pathname, surely it cannot match.
1012 */
1013 if (prefix > namelen)
1014 return 0;
1015
ba0897e6 1016 if (fspathncmp(pattern, name, prefix))
b5592632
NTND
1017 return 0;
1018 pattern += prefix;
ab3aebc1 1019 patternlen -= prefix;
b5592632
NTND
1020 name += prefix;
1021 namelen -= prefix;
ab3aebc1
JK
1022
1023 /*
1024 * If the whole pattern did not have a wildcard,
1025 * then our prefix match is all we need; we
1026 * do not need to call fnmatch at all.
1027 */
5cee3493 1028 if (!patternlen && !namelen)
ab3aebc1 1029 return 1;
b5592632
NTND
1030 }
1031
ab3aebc1
JK
1032 return fnmatch_icase_mem(pattern, patternlen,
1033 name, namelen,
f30366b2 1034 WM_PATHNAME) == 0;
b5592632
NTND
1035}
1036
578cd7c3
AS
1037/*
1038 * Scan the given exclude list in reverse to see whether pathname
1039 * should be ignored. The first match (i.e. the last on the list), if
1040 * any, determines the fate. Returns the exclude_list element which
1041 * matched, or NULL for undecided.
453ec4bd 1042 */
578cd7c3
AS
1043static struct exclude *last_exclude_matching_from_list(const char *pathname,
1044 int pathlen,
1045 const char *basename,
1046 int *dtype,
2b70e88d
BW
1047 struct exclude_list *el,
1048 struct index_state *istate)
453ec4bd 1049{
e6efecc4 1050 struct exclude *exc = NULL; /* undecided */
5cee3493 1051 int i;
453ec4bd 1052
35a94d44 1053 if (!el->nr)
578cd7c3 1054 return NULL; /* undefined */
d6b8fc30 1055
35a94d44
NTND
1056 for (i = el->nr - 1; 0 <= i; i--) {
1057 struct exclude *x = el->excludes[i];
b5592632 1058 const char *exclude = x->pattern;
b5592632 1059 int prefix = x->nowildcardlen;
35a94d44
NTND
1060
1061 if (x->flags & EXC_FLAG_MUSTBEDIR) {
1062 if (*dtype == DT_UNKNOWN)
2b70e88d 1063 *dtype = get_dtype(NULL, istate, pathname, pathlen);
35a94d44
NTND
1064 if (*dtype != DT_DIR)
1065 continue;
1066 }
d6b8fc30 1067
35a94d44 1068 if (x->flags & EXC_FLAG_NODIR) {
593cb880
NTND
1069 if (match_basename(basename,
1070 pathlen - (basename - pathname),
1071 exclude, prefix, x->patternlen,
e6efecc4
NTND
1072 x->flags)) {
1073 exc = x;
1074 break;
1075 }
35a94d44
NTND
1076 continue;
1077 }
1078
b5592632
NTND
1079 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
1080 if (match_pathname(pathname, pathlen,
1081 x->base, x->baselen ? x->baselen - 1 : 0,
e6efecc4
NTND
1082 exclude, prefix, x->patternlen, x->flags)) {
1083 exc = x;
1084 break;
1085 }
453ec4bd 1086 }
e6efecc4 1087 return exc;
578cd7c3
AS
1088}
1089
1090/*
1091 * Scan the list and let the last match determine the fate.
1092 * Return 1 for exclude, 0 for include and -1 for undecided.
1093 */
1094int is_excluded_from_list(const char *pathname,
1095 int pathlen, const char *basename, int *dtype,
fba92be8 1096 struct exclude_list *el, struct index_state *istate)
578cd7c3
AS
1097{
1098 struct exclude *exclude;
fba92be8
BW
1099 exclude = last_exclude_matching_from_list(pathname, pathlen, basename,
1100 dtype, el, istate);
578cd7c3
AS
1101 if (exclude)
1102 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
453ec4bd
LT
1103 return -1; /* undecided */
1104}
1105
46aa2f95 1106static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
2b70e88d 1107 struct index_state *istate,
46aa2f95
KB
1108 const char *pathname, int pathlen, const char *basename,
1109 int *dtype_p)
1110{
1111 int i, j;
1112 struct exclude_list_group *group;
1113 struct exclude *exclude;
1114 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1115 group = &dir->exclude_list_group[i];
1116 for (j = group->nr - 1; j >= 0; j--) {
1117 exclude = last_exclude_matching_from_list(
1118 pathname, pathlen, basename, dtype_p,
2b70e88d 1119 &group->el[j], istate);
46aa2f95
KB
1120 if (exclude)
1121 return exclude;
1122 }
1123 }
1124 return NULL;
1125}
1126
6cd5c582
KB
1127/*
1128 * Loads the per-directory exclude list for the substring of base
1129 * which has a char length of baselen.
1130 */
e799ed40
BW
1131static void prep_exclude(struct dir_struct *dir,
1132 struct index_state *istate,
1133 const char *base, int baselen)
6cd5c582
KB
1134{
1135 struct exclude_list_group *group;
1136 struct exclude_list *el;
1137 struct exclude_stack *stk = NULL;
0dcb8d7f 1138 struct untracked_cache_dir *untracked;
6cd5c582
KB
1139 int current;
1140
6cd5c582
KB
1141 group = &dir->exclude_list_group[EXC_DIRS];
1142
d961baa8
NTND
1143 /*
1144 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
6cd5c582 1145 * which originate from directories not in the prefix of the
d961baa8
NTND
1146 * path being checked.
1147 */
6cd5c582
KB
1148 while ((stk = dir->exclude_stack) != NULL) {
1149 if (stk->baselen <= baselen &&
aceb9429 1150 !strncmp(dir->basebuf.buf, base, stk->baselen))
6cd5c582
KB
1151 break;
1152 el = &group->el[dir->exclude_stack->exclude_ix];
1153 dir->exclude_stack = stk->prev;
95c6f271 1154 dir->exclude = NULL;
aceb9429 1155 free((char *)el->src); /* see strbuf_detach() below */
6cd5c582
KB
1156 clear_exclude_list(el);
1157 free(stk);
1158 group->nr--;
1159 }
1160
95c6f271
KB
1161 /* Skip traversing into sub directories if the parent is excluded */
1162 if (dir->exclude)
1163 return;
1164
aceb9429
NTND
1165 /*
1166 * Lazy initialization. All call sites currently just
1167 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1168 * them seems lots of work for little benefit.
1169 */
1170 if (!dir->basebuf.buf)
1171 strbuf_init(&dir->basebuf, PATH_MAX);
1172
6cd5c582
KB
1173 /* Read from the parent directories and push them down. */
1174 current = stk ? stk->baselen : -1;
aceb9429 1175 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
0dcb8d7f
NTND
1176 if (dir->untracked)
1177 untracked = stk ? stk->ucd : dir->untracked->root;
1178 else
1179 untracked = NULL;
1180
6cd5c582 1181 while (current < baselen) {
6cd5c582 1182 const char *cp;
0dcb8d7f 1183 struct sha1_stat sha1_stat;
6cd5c582 1184
03e11a71 1185 stk = xcalloc(1, sizeof(*stk));
6cd5c582
KB
1186 if (current < 0) {
1187 cp = base;
1188 current = 0;
d961baa8 1189 } else {
6cd5c582
KB
1190 cp = strchr(base + current + 1, '/');
1191 if (!cp)
1192 die("oops in prep_exclude");
1193 cp++;
0dcb8d7f
NTND
1194 untracked =
1195 lookup_untracked(dir->untracked, untracked,
1196 base + current,
1197 cp - base - current);
6cd5c582
KB
1198 }
1199 stk->prev = dir->exclude_stack;
1200 stk->baselen = cp - base;
95c6f271 1201 stk->exclude_ix = group->nr;
0dcb8d7f 1202 stk->ucd = untracked;
95c6f271 1203 el = add_exclude_list(dir, EXC_DIRS, NULL);
aceb9429
NTND
1204 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1205 assert(stk->baselen == dir->basebuf.len);
95c6f271
KB
1206
1207 /* Abort if the directory is excluded */
1208 if (stk->baselen) {
1209 int dt = DT_DIR;
aceb9429 1210 dir->basebuf.buf[stk->baselen - 1] = 0;
95c6f271 1211 dir->exclude = last_exclude_matching_from_lists(dir,
e799ed40 1212 istate,
aceb9429
NTND
1213 dir->basebuf.buf, stk->baselen - 1,
1214 dir->basebuf.buf + current, &dt);
1215 dir->basebuf.buf[stk->baselen - 1] = '/';
c3c327de
KB
1216 if (dir->exclude &&
1217 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1218 dir->exclude = NULL;
95c6f271 1219 if (dir->exclude) {
95c6f271
KB
1220 dir->exclude_stack = stk;
1221 return;
1222 }
1223 }
1224
aceb9429 1225 /* Try to read per-directory file */
0dcb8d7f
NTND
1226 hashclr(sha1_stat.sha1);
1227 sha1_stat.valid = 0;
27b099ae
NTND
1228 if (dir->exclude_per_dir &&
1229 /*
1230 * If we know that no files have been added in
1231 * this directory (i.e. valid_cached_dir() has
1232 * been executed and set untracked->valid) ..
1233 */
1234 (!untracked || !untracked->valid ||
1235 /*
1236 * .. and .gitignore does not exist before
7687252f
DT
1237 * (i.e. null exclude_sha1). Then we can skip
1238 * loading .gitignore, which would result in
1239 * ENOENT anyway.
27b099ae
NTND
1240 */
1241 !is_null_sha1(untracked->exclude_sha1))) {
95c6f271
KB
1242 /*
1243 * dir->basebuf gets reused by the traversal, but we
1244 * need fname to remain unchanged to ensure the src
1245 * member of each struct exclude correctly
1246 * back-references its source file. Other invocations
1247 * of add_exclude_list provide stable strings, so we
aceb9429 1248 * strbuf_detach() and free() here in the caller.
95c6f271 1249 */
aceb9429
NTND
1250 struct strbuf sb = STRBUF_INIT;
1251 strbuf_addbuf(&sb, &dir->basebuf);
1252 strbuf_addstr(&sb, dir->exclude_per_dir);
1253 el->src = strbuf_detach(&sb, NULL);
e799ed40 1254 add_excludes(el->src, el->src, stk->baselen, el, istate,
0dcb8d7f
NTND
1255 untracked ? &sha1_stat : NULL);
1256 }
5ebf79ad
NTND
1257 /*
1258 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1259 * will first be called in valid_cached_dir() then maybe many
1260 * times more in last_exclude_matching(). When the cache is
1261 * used, last_exclude_matching() will not be called and
1262 * reading .gitignore content will be a waste.
1263 *
1264 * So when it's called by valid_cached_dir() and we can get
1265 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1266 * modified on work tree), we could delay reading the
1267 * .gitignore content until we absolutely need it in
1268 * last_exclude_matching(). Be careful about ignore rule
1269 * order, though, if you do that.
1270 */
1271 if (untracked &&
1272 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1273 invalidate_gitignore(dir->untracked, untracked);
0dcb8d7f 1274 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
95c6f271 1275 }
6cd5c582
KB
1276 dir->exclude_stack = stk;
1277 current = stk->baselen;
1278 }
aceb9429 1279 strbuf_setlen(&dir->basebuf, baselen);
6cd5c582
KB
1280}
1281
f4cd69a6
AS
1282/*
1283 * Loads the exclude lists for the directory containing pathname, then
1284 * scans all exclude lists to determine whether pathname is excluded.
1285 * Returns the exclude_list element which matched, or NULL for
1286 * undecided.
1287 */
b07bc8c8 1288struct exclude *last_exclude_matching(struct dir_struct *dir,
a0bba65b
BW
1289 struct index_state *istate,
1290 const char *pathname,
1291 int *dtype_p)
453ec4bd
LT
1292{
1293 int pathlen = strlen(pathname);
68492fc7
LK
1294 const char *basename = strrchr(pathname, '/');
1295 basename = (basename) ? basename+1 : pathname;
453ec4bd 1296
a0bba65b 1297 prep_exclude(dir, istate, pathname, basename-pathname);
c082df24 1298
95c6f271
KB
1299 if (dir->exclude)
1300 return dir->exclude;
1301
a0bba65b 1302 return last_exclude_matching_from_lists(dir, istate, pathname, pathlen,
46aa2f95 1303 basename, dtype_p);
f4cd69a6
AS
1304}
1305
1306/*
1307 * Loads the exclude lists for the directory containing pathname, then
1308 * scans all exclude lists to determine whether pathname is excluded.
1309 * Returns 1 if true, otherwise 0.
1310 */
a0bba65b
BW
1311int is_excluded(struct dir_struct *dir, struct index_state *istate,
1312 const char *pathname, int *dtype_p)
f4cd69a6
AS
1313{
1314 struct exclude *exclude =
a0bba65b 1315 last_exclude_matching(dir, istate, pathname, dtype_p);
f4cd69a6
AS
1316 if (exclude)
1317 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
453ec4bd
LT
1318 return 0;
1319}
1320
f3fa1838
JH
1321static struct dir_entry *dir_entry_new(const char *pathname, int len)
1322{
453ec4bd
LT
1323 struct dir_entry *ent;
1324
96ffc06f 1325 FLEX_ALLOC_MEM(ent, name, pathname, len);
453ec4bd 1326 ent->len = len;
4d06f8ac 1327 return ent;
453ec4bd
LT
1328}
1329
9e58beca
BW
1330static struct dir_entry *dir_add_name(struct dir_struct *dir,
1331 struct index_state *istate,
1332 const char *pathname, int len)
6815e569 1333{
9e58beca 1334 if (index_file_exists(istate, pathname, len, ignore_case))
6815e569
JK
1335 return NULL;
1336
25fd2f7a 1337 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
6815e569
JK
1338 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1339}
1340
9e58beca
BW
1341struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1342 struct index_state *istate,
1343 const char *pathname, int len)
2abd31b0 1344{
9e58beca 1345 if (!index_name_is_other(istate, pathname, len))
2abd31b0
JK
1346 return NULL;
1347
25fd2f7a 1348 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
2abd31b0
JK
1349 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1350}
1351
09595258
LT
1352enum exist_status {
1353 index_nonexistent = 0,
1354 index_directory,
4b05548f 1355 index_gitdir
09595258
LT
1356};
1357
5102c617 1358/*
71261027 1359 * Do not use the alphabetically sorted index to look up
5102c617 1360 * the directory name; instead, use the case insensitive
ebbd7439 1361 * directory hash.
5102c617 1362 */
ae520e36
BW
1363static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1364 const char *dirname, int len)
5102c617 1365{
41284eb0 1366 struct cache_entry *ce;
5102c617 1367
ae520e36 1368 if (index_dir_exists(istate, dirname, len))
5102c617
JJ
1369 return index_directory;
1370
ae520e36 1371 ce = index_file_exists(istate, dirname, len, ignore_case);
41284eb0 1372 if (ce && S_ISGITLINK(ce->ce_mode))
5102c617
JJ
1373 return index_gitdir;
1374
5102c617
JJ
1375 return index_nonexistent;
1376}
1377
09595258
LT
1378/*
1379 * The index sorts alphabetically by entry name, which
1380 * means that a gitlink sorts as '\0' at the end, while
1381 * a directory (which is defined not as an entry, but as
1382 * the files it contains) will sort with the '/' at the
1383 * end.
1384 */
ae520e36
BW
1385static enum exist_status directory_exists_in_index(struct index_state *istate,
1386 const char *dirname, int len)
453ec4bd 1387{
5102c617
JJ
1388 int pos;
1389
1390 if (ignore_case)
ae520e36 1391 return directory_exists_in_index_icase(istate, dirname, len);
5102c617 1392
ae520e36 1393 pos = index_name_pos(istate, dirname, len);
09595258
LT
1394 if (pos < 0)
1395 pos = -pos-1;
ae520e36
BW
1396 while (pos < istate->cache_nr) {
1397 const struct cache_entry *ce = istate->cache[pos++];
09595258
LT
1398 unsigned char endchar;
1399
1400 if (strncmp(ce->name, dirname, len))
1401 break;
1402 endchar = ce->name[len];
1403 if (endchar > '/')
1404 break;
1405 if (endchar == '/')
1406 return index_directory;
7a51ed66 1407 if (!endchar && S_ISGITLINK(ce->ce_mode))
09595258
LT
1408 return index_gitdir;
1409 }
1410 return index_nonexistent;
1411}
1412
1413/*
1414 * When we find a directory when traversing the filesystem, we
1415 * have three distinct cases:
1416 *
1417 * - ignore it
1418 * - see it as a directory
1419 * - recurse into it
1420 *
1421 * and which one we choose depends on a combination of existing
1422 * git index contents and the flags passed into the directory
1423 * traversal routine.
1424 *
1425 * Case 1: If we *already* have entries in the index under that
5bd8e2d8
KB
1426 * directory name, we always recurse into the directory to see
1427 * all the files.
09595258
LT
1428 *
1429 * Case 2: If we *already* have that directory name as a gitlink,
1430 * we always continue to see it as a gitlink, regardless of whether
1431 * there is an actual git directory there or not (it might not
1432 * be checked out as a subproject!)
1433 *
1434 * Case 3: if we didn't have it in the index previously, we
1435 * have a few sub-cases:
1436 *
1437 * (a) if "show_other_directories" is true, we show it as
1438 * just a directory, unless "hide_empty_directories" is
defd7c7b
KB
1439 * also true, in which case we need to check if it contains any
1440 * untracked and / or ignored files.
09595258 1441 * (b) if it looks like a git directory, and we don't have
302b9282 1442 * 'no_gitlinks' set we treat it as a gitlink, and show it
09595258
LT
1443 * as a directory.
1444 * (c) otherwise, we recurse into it.
1445 */
defd7c7b 1446static enum path_treatment treat_directory(struct dir_struct *dir,
0ef8e169 1447 struct index_state *istate,
0dcb8d7f 1448 struct untracked_cache_dir *untracked,
2e5910f2 1449 const char *dirname, int len, int baselen, int exclude,
e1b8c7bd 1450 const struct pathspec *pathspec)
09595258
LT
1451{
1452 /* The "len-1" is to strip the final '/' */
0ef8e169 1453 switch (directory_exists_in_index(istate, dirname, len-1)) {
09595258 1454 case index_directory:
defd7c7b 1455 return path_recurse;
09595258
LT
1456
1457 case index_gitdir:
26c986e1 1458 return path_none;
09595258
LT
1459
1460 case index_nonexistent:
7c4c97c0 1461 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
09595258 1462 break;
eec0f7f2
JM
1463 if (exclude &&
1464 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1465 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1466
1467 /*
1468 * This is an excluded directory and we are
1469 * showing ignored paths that match an exclude
1470 * pattern. (e.g. show directory as ignored
1471 * only if it matches an exclude pattern).
1472 * This path will either be 'path_excluded`
1473 * (if we are showing empty directories or if
1474 * the directory is not empty), or will be
1475 * 'path_none' (empty directory, and we are
1476 * not showing empty directories).
1477 */
1478 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1479 return path_excluded;
1480
1481 if (read_directory_recursive(dir, istate, dirname, len,
1482 untracked, 1, 1, pathspec) == path_excluded)
1483 return path_excluded;
1484
1485 return path_none;
1486 }
7c4c97c0 1487 if (!(dir->flags & DIR_NO_GITLINKS)) {
1053fe82 1488 struct object_id oid;
a98e6101 1489 if (resolve_gitlink_ref(dirname, "HEAD", &oid) == 0)
fadb4820 1490 return exclude ? path_excluded : path_untracked;
09595258 1491 }
defd7c7b 1492 return path_recurse;
09595258
LT
1493 }
1494
1495 /* This is the "show_other_directories" case */
721ac4ed 1496
184d2a8e 1497 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
defd7c7b
KB
1498 return exclude ? path_excluded : path_untracked;
1499
2e5910f2
DT
1500 untracked = lookup_untracked(dir->untracked, untracked,
1501 dirname + baselen, len - baselen);
5aaa7fd3
JM
1502
1503 /*
1504 * If this is an excluded directory, then we only need to check if
1505 * the directory contains any files.
1506 */
0ef8e169 1507 return read_directory_recursive(dir, istate, dirname, len,
5aaa7fd3 1508 untracked, 1, exclude, pathspec);
453ec4bd
LT
1509}
1510
9fc42d60
LT
1511/*
1512 * This is an inexact early pruning of any recursive directory
1513 * reading - if the path cannot possibly be in the pathspec,
1514 * return true, and we'll skip it early.
1515 */
e1b8c7bd
BW
1516static int simplify_away(const char *path, int pathlen,
1517 const struct pathspec *pathspec)
9fc42d60 1518{
e1b8c7bd 1519 int i;
9fc42d60 1520
e1b8c7bd
BW
1521 if (!pathspec || !pathspec->nr)
1522 return 0;
1523
1524 GUARD_PATHSPEC(pathspec,
1525 PATHSPEC_FROMTOP |
1526 PATHSPEC_MAXDEPTH |
1527 PATHSPEC_LITERAL |
1528 PATHSPEC_GLOB |
1529 PATHSPEC_ICASE |
b0db7046
BW
1530 PATHSPEC_EXCLUDE |
1531 PATHSPEC_ATTR);
e1b8c7bd
BW
1532
1533 for (i = 0; i < pathspec->nr; i++) {
1534 const struct pathspec_item *item = &pathspec->items[i];
1535 int len = item->nowildcard_len;
1536
1537 if (len > pathlen)
1538 len = pathlen;
1539 if (!ps_strncmp(item, item->match, path, len))
1540 return 0;
9fc42d60 1541 }
e1b8c7bd
BW
1542
1543 return 1;
9fc42d60
LT
1544}
1545
29209cbe
JK
1546/*
1547 * This function tells us whether an excluded path matches a
1548 * list of "interesting" pathspecs. That is, whether a path matched
1549 * by any of the pathspecs could possibly be ignored by excluding
1550 * the specified path. This can happen if:
1551 *
1552 * 1. the path is mentioned explicitly in the pathspec
1553 *
1554 * 2. the path is a directory prefix of some element in the
1555 * pathspec
1556 */
e1b8c7bd
BW
1557static int exclude_matches_pathspec(const char *path, int pathlen,
1558 const struct pathspec *pathspec)
1559{
1560 int i;
1561
1562 if (!pathspec || !pathspec->nr)
1563 return 0;
1564
1565 GUARD_PATHSPEC(pathspec,
1566 PATHSPEC_FROMTOP |
1567 PATHSPEC_MAXDEPTH |
1568 PATHSPEC_LITERAL |
1569 PATHSPEC_GLOB |
1570 PATHSPEC_ICASE |
1571 PATHSPEC_EXCLUDE);
1572
1573 for (i = 0; i < pathspec->nr; i++) {
1574 const struct pathspec_item *item = &pathspec->items[i];
1575 int len = item->nowildcard_len;
1576
1577 if (len == pathlen &&
1578 !ps_strncmp(item, item->match, path, pathlen))
1579 return 1;
1580 if (len > pathlen &&
1581 item->match[pathlen] == '/' &&
1582 !ps_strncmp(item, item->match, path, pathlen))
1583 return 1;
e96980ef
JK
1584 }
1585 return 0;
1586}
1587
98f2a687
BW
1588static int get_index_dtype(struct index_state *istate,
1589 const char *path, int len)
443e061a
LT
1590{
1591 int pos;
9c5e6c80 1592 const struct cache_entry *ce;
443e061a 1593
98f2a687 1594 ce = index_file_exists(istate, path, len, 0);
443e061a
LT
1595 if (ce) {
1596 if (!ce_uptodate(ce))
1597 return DT_UNKNOWN;
1598 if (S_ISGITLINK(ce->ce_mode))
1599 return DT_DIR;
1600 /*
1601 * Nobody actually cares about the
1602 * difference between DT_LNK and DT_REG
1603 */
1604 return DT_REG;
1605 }
1606
1607 /* Try to look it up as a directory */
98f2a687 1608 pos = index_name_pos(istate, path, len);
443e061a
LT
1609 if (pos >= 0)
1610 return DT_UNKNOWN;
1611 pos = -pos-1;
98f2a687
BW
1612 while (pos < istate->cache_nr) {
1613 ce = istate->cache[pos++];
443e061a
LT
1614 if (strncmp(ce->name, path, len))
1615 break;
1616 if (ce->name[len] > '/')
1617 break;
1618 if (ce->name[len] < '/')
1619 continue;
1620 if (!ce_uptodate(ce))
1621 break; /* continue? */
1622 return DT_DIR;
1623 }
1624 return DT_UNKNOWN;
1625}
1626
98f2a687
BW
1627static int get_dtype(struct dirent *de, struct index_state *istate,
1628 const char *path, int len)
07134421 1629{
6831a88a 1630 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
07134421
LT
1631 struct stat st;
1632
1633 if (dtype != DT_UNKNOWN)
1634 return dtype;
98f2a687 1635 dtype = get_index_dtype(istate, path, len);
443e061a
LT
1636 if (dtype != DT_UNKNOWN)
1637 return dtype;
1638 if (lstat(path, &st))
07134421
LT
1639 return dtype;
1640 if (S_ISREG(st.st_mode))
1641 return DT_REG;
1642 if (S_ISDIR(st.st_mode))
1643 return DT_DIR;
1644 if (S_ISLNK(st.st_mode))
1645 return DT_LNK;
1646 return dtype;
1647}
1648
16e2cfa9 1649static enum path_treatment treat_one_path(struct dir_struct *dir,
0dcb8d7f 1650 struct untracked_cache_dir *untracked,
0ef8e169 1651 struct index_state *istate,
49dc2cc2 1652 struct strbuf *path,
2e5910f2 1653 int baselen,
e1b8c7bd 1654 const struct pathspec *pathspec,
16e2cfa9 1655 int dtype, struct dirent *de)
53cc5356 1656{
8aaf8d77 1657 int exclude;
0ef8e169 1658 int has_path_in_index = !!index_file_exists(istate, path->buf, path->len, ignore_case);
07966ed1 1659 enum path_treatment path_treatment;
2eac2a4c 1660
8aaf8d77 1661 if (dtype == DT_UNKNOWN)
0ef8e169 1662 dtype = get_dtype(de, istate, path->buf, path->len);
8aaf8d77
KB
1663
1664 /* Always exclude indexed files */
2eac2a4c 1665 if (dtype != DT_DIR && has_path_in_index)
defd7c7b 1666 return path_none;
8aaf8d77 1667
2eac2a4c
JH
1668 /*
1669 * When we are looking at a directory P in the working tree,
1670 * there are three cases:
1671 *
1672 * (1) P exists in the index. Everything inside the directory P in
1673 * the working tree needs to go when P is checked out from the
1674 * index.
1675 *
1676 * (2) P does not exist in the index, but there is P/Q in the index.
1677 * We know P will stay a directory when we check out the contents
1678 * of the index, but we do not know yet if there is a directory
1679 * P/Q in the working tree to be killed, so we need to recurse.
1680 *
1681 * (3) P does not exist in the index, and there is no P/Q in the index
1682 * to require P to be a directory, either. Only in this case, we
1683 * know that everything inside P will not be killed without
1684 * recursing.
1685 */
1686 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1687 (dtype == DT_DIR) &&
de372b1b 1688 !has_path_in_index &&
0ef8e169 1689 (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
de372b1b 1690 return path_none;
8aaf8d77 1691
0ef8e169 1692 exclude = is_excluded(dir, istate, path->buf, &dtype);
53cc5356
JH
1693
1694 /*
1695 * Excluded? If we don't explicitly want to show
1696 * ignored files, ignore it
1697 */
0aaf62b6 1698 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
defd7c7b 1699 return path_excluded;
53cc5356 1700
53cc5356
JH
1701 switch (dtype) {
1702 default:
defd7c7b 1703 return path_none;
53cc5356 1704 case DT_DIR:
49dc2cc2 1705 strbuf_addch(path, '/');
07966ed1
JM
1706 path_treatment = treat_directory(dir, istate, untracked,
1707 path->buf, path->len,
1708 baselen, exclude, pathspec);
1709 /*
1710 * If 1) we only want to return directories that
1711 * match an exclude pattern and 2) this directory does
1712 * not match an exclude pattern but all of its
1713 * contents are excluded, then indicate that we should
1714 * recurse into this directory (instead of marking the
1715 * directory itself as an ignored path).
1716 */
1717 if (!exclude &&
1718 path_treatment == path_excluded &&
1719 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1720 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
1721 return path_recurse;
1722 return path_treatment;
53cc5356
JH
1723 case DT_REG:
1724 case DT_LNK:
defd7c7b 1725 return exclude ? path_excluded : path_untracked;
53cc5356 1726 }
53cc5356
JH
1727}
1728
91a2288b
NTND
1729static enum path_treatment treat_path_fast(struct dir_struct *dir,
1730 struct untracked_cache_dir *untracked,
1731 struct cached_dir *cdir,
0ef8e169 1732 struct index_state *istate,
91a2288b
NTND
1733 struct strbuf *path,
1734 int baselen,
e1b8c7bd 1735 const struct pathspec *pathspec)
91a2288b
NTND
1736{
1737 strbuf_setlen(path, baselen);
1738 if (!cdir->ucd) {
1739 strbuf_addstr(path, cdir->file);
1740 return path_untracked;
1741 }
1742 strbuf_addstr(path, cdir->ucd->name);
1743 /* treat_one_path() does this before it calls treat_directory() */
00b6c178 1744 strbuf_complete(path, '/');
91a2288b
NTND
1745 if (cdir->ucd->check_only)
1746 /*
1747 * check_only is set as a result of treat_directory() getting
1748 * to its bottom. Verify again the same set of directories
1749 * with check_only set.
1750 */
0ef8e169 1751 return read_directory_recursive(dir, istate, path->buf, path->len,
5aaa7fd3 1752 cdir->ucd, 1, 0, pathspec);
91a2288b
NTND
1753 /*
1754 * We get path_recurse in the first run when
1755 * directory_exists_in_index() returns index_nonexistent. We
1756 * are sure that new changes in the index does not impact the
1757 * outcome. Return now.
1758 */
1759 return path_recurse;
1760}
1761
16e2cfa9 1762static enum path_treatment treat_path(struct dir_struct *dir,
0dcb8d7f 1763 struct untracked_cache_dir *untracked,
cf7c6148 1764 struct cached_dir *cdir,
0ef8e169 1765 struct index_state *istate,
49dc2cc2 1766 struct strbuf *path,
16e2cfa9 1767 int baselen,
e1b8c7bd 1768 const struct pathspec *pathspec)
16e2cfa9
JH
1769{
1770 int dtype;
cf7c6148 1771 struct dirent *de = cdir->de;
16e2cfa9 1772
91a2288b 1773 if (!de)
0ef8e169 1774 return treat_path_fast(dir, untracked, cdir, istate, path,
e1b8c7bd 1775 baselen, pathspec);
16e2cfa9 1776 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
defd7c7b 1777 return path_none;
49dc2cc2
RS
1778 strbuf_setlen(path, baselen);
1779 strbuf_addstr(path, de->d_name);
e1b8c7bd 1780 if (simplify_away(path->buf, path->len, pathspec))
defd7c7b 1781 return path_none;
16e2cfa9
JH
1782
1783 dtype = DTYPE(de);
0ef8e169 1784 return treat_one_path(dir, untracked, istate, path, baselen, pathspec, dtype, de);
0dcb8d7f
NTND
1785}
1786
1787static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1788{
1789 if (!dir)
1790 return;
1791 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1792 dir->untracked_alloc);
1793 dir->untracked[dir->untracked_nr++] = xstrdup(name);
16e2cfa9
JH
1794}
1795
91a2288b
NTND
1796static int valid_cached_dir(struct dir_struct *dir,
1797 struct untracked_cache_dir *untracked,
207a06ce 1798 struct index_state *istate,
91a2288b
NTND
1799 struct strbuf *path,
1800 int check_only)
1801{
1802 struct stat st;
1803
1804 if (!untracked)
1805 return 0;
1806
883e248b
BP
1807 /*
1808 * With fsmonitor, we can trust the untracked cache's valid field.
1809 */
1810 refresh_fsmonitor(istate);
1811 if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
1812 if (stat(path->len ? path->buf : ".", &st)) {
91a2288b 1813 invalidate_directory(dir->untracked, untracked);
883e248b
BP
1814 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1815 return 0;
1816 }
1817 if (!untracked->valid ||
1818 match_stat_data_racy(istate, &untracked->stat_data, &st)) {
1819 if (untracked->valid)
1820 invalidate_directory(dir->untracked, untracked);
1821 fill_stat_data(&untracked->stat_data, &st);
1822 return 0;
1823 }
91a2288b
NTND
1824 }
1825
1826 if (untracked->check_only != !!check_only) {
1827 invalidate_directory(dir->untracked, untracked);
1828 return 0;
1829 }
1830
1831 /*
1832 * prep_exclude will be called eventually on this directory,
1833 * but it's called much later in last_exclude_matching(). We
1834 * need it now to determine the validity of the cache for this
1835 * path. The next calls will be nearly no-op, the way
1836 * prep_exclude() is designed.
1837 */
1838 if (path->len && path->buf[path->len - 1] != '/') {
1839 strbuf_addch(path, '/');
207a06ce 1840 prep_exclude(dir, istate, path->buf, path->len);
91a2288b
NTND
1841 strbuf_setlen(path, path->len - 1);
1842 } else
207a06ce 1843 prep_exclude(dir, istate, path->buf, path->len);
91a2288b
NTND
1844
1845 /* hopefully prep_exclude() haven't invalidated this entry... */
1846 return untracked->valid;
1847}
1848
cf7c6148
NTND
1849static int open_cached_dir(struct cached_dir *cdir,
1850 struct dir_struct *dir,
1851 struct untracked_cache_dir *untracked,
207a06ce 1852 struct index_state *istate,
cf7c6148
NTND
1853 struct strbuf *path,
1854 int check_only)
1855{
1856 memset(cdir, 0, sizeof(*cdir));
1857 cdir->untracked = untracked;
207a06ce 1858 if (valid_cached_dir(dir, untracked, istate, path, check_only))
91a2288b 1859 return 0;
cf7c6148 1860 cdir->fdir = opendir(path->len ? path->buf : ".");
91a2288b
NTND
1861 if (dir->untracked)
1862 dir->untracked->dir_opened++;
cf7c6148
NTND
1863 if (!cdir->fdir)
1864 return -1;
1865 return 0;
1866}
1867
1868static int read_cached_dir(struct cached_dir *cdir)
1869{
1870 if (cdir->fdir) {
1871 cdir->de = readdir(cdir->fdir);
1872 if (!cdir->de)
1873 return -1;
1874 return 0;
1875 }
91a2288b
NTND
1876 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1877 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
26cb0182
NTND
1878 if (!d->recurse) {
1879 cdir->nr_dirs++;
1880 continue;
1881 }
91a2288b
NTND
1882 cdir->ucd = d;
1883 cdir->nr_dirs++;
1884 return 0;
1885 }
1886 cdir->ucd = NULL;
1887 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1888 struct untracked_cache_dir *d = cdir->untracked;
1889 cdir->file = d->untracked[cdir->nr_files++];
1890 return 0;
1891 }
cf7c6148
NTND
1892 return -1;
1893}
1894
1895static void close_cached_dir(struct cached_dir *cdir)
1896{
1897 if (cdir->fdir)
1898 closedir(cdir->fdir);
91a2288b
NTND
1899 /*
1900 * We have gone through this directory and found no untracked
1901 * entries. Mark it valid.
1902 */
26cb0182 1903 if (cdir->untracked) {
91a2288b 1904 cdir->untracked->valid = 1;
26cb0182
NTND
1905 cdir->untracked->recurse = 1;
1906 }
16e2cfa9
JH
1907}
1908
453ec4bd
LT
1909/*
1910 * Read a directory tree. We currently ignore anything but
1911 * directories, regular files and symlinks. That's because git
1912 * doesn't handle them at all yet. Maybe that will change some
1913 * day.
1914 *
1915 * Also, we ignore the name ".git" (even if it is not a directory).
1916 * That likely will not change.
defd7c7b 1917 *
5aaa7fd3
JM
1918 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1919 * to signal that a file was found. This is the least significant value that
1920 * indicates that a file was encountered that does not depend on the order of
1921 * whether an untracked or exluded path was encountered first.
1922 *
defd7c7b 1923 * Returns the most significant path_treatment value encountered in the scan.
5aaa7fd3
JM
1924 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1925 * significant path_treatment value that will be returned.
453ec4bd 1926 */
5aaa7fd3 1927
defd7c7b 1928static enum path_treatment read_directory_recursive(struct dir_struct *dir,
0ef8e169
BW
1929 struct index_state *istate, const char *base, int baselen,
1930 struct untracked_cache_dir *untracked, int check_only,
5aaa7fd3 1931 int stop_at_first_file, const struct pathspec *pathspec)
453ec4bd 1932{
cf7c6148 1933 struct cached_dir cdir;
defd7c7b 1934 enum path_treatment state, subdir_state, dir_state = path_none;
bef36921 1935 struct strbuf path = STRBUF_INIT;
453ec4bd 1936
bef36921 1937 strbuf_add(&path, base, baselen);
02cb6753 1938
0ef8e169 1939 if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
1528d247
RS
1940 goto out;
1941
0dcb8d7f
NTND
1942 if (untracked)
1943 untracked->check_only = !!check_only;
1944
cf7c6148 1945 while (!read_cached_dir(&cdir)) {
defd7c7b 1946 /* check how the file or directory should be treated */
0ef8e169 1947 state = treat_path(dir, untracked, &cdir, istate, &path,
e1b8c7bd 1948 baselen, pathspec);
0dcb8d7f 1949
defd7c7b
KB
1950 if (state > dir_state)
1951 dir_state = state;
1952
1953 /* recurse into subdir if instructed by treat_path */
df5bcdf8
SL
1954 if ((state == path_recurse) ||
1955 ((state == path_untracked) &&
1956 (dir->flags & DIR_SHOW_IGNORED_TOO) &&
f4fd99bf 1957 (get_dtype(cdir.de, istate, path.buf, path.len) == DT_DIR))) {
0dcb8d7f
NTND
1958 struct untracked_cache_dir *ud;
1959 ud = lookup_untracked(dir->untracked, untracked,
1960 path.buf + baselen,
1961 path.len - baselen);
1962 subdir_state =
0ef8e169 1963 read_directory_recursive(dir, istate, path.buf,
e1b8c7bd 1964 path.len, ud,
5aaa7fd3 1965 check_only, stop_at_first_file, pathspec);
defd7c7b
KB
1966 if (subdir_state > dir_state)
1967 dir_state = subdir_state;
1968 }
1969
1970 if (check_only) {
5aaa7fd3
JM
1971 if (stop_at_first_file) {
1972 /*
1973 * If stopping at first file, then
1974 * signal that a file was found by
1975 * returning `path_excluded`. This is
1976 * to return a consistent value
1977 * regardless of whether an ignored or
1978 * excluded file happened to be
1979 * encountered 1st.
1980 *
1981 * In current usage, the
1982 * `stop_at_first_file` is passed when
1983 * an ancestor directory has matched
1984 * an exclude pattern, so any found
1985 * files will be excluded.
1986 */
1987 if (dir_state >= path_excluded) {
1988 dir_state = path_excluded;
1989 break;
1990 }
1991 }
1992
defd7c7b 1993 /* abort early if maximum state has been reached */
0dcb8d7f 1994 if (dir_state == path_untracked) {
91a2288b 1995 if (cdir.fdir)
0dcb8d7f 1996 add_untracked(untracked, path.buf + baselen);
defd7c7b 1997 break;
0dcb8d7f 1998 }
defd7c7b 1999 /* skip the dir_add_* part */
02cb6753 2000 continue;
453ec4bd 2001 }
defd7c7b
KB
2002
2003 /* add the path to the appropriate result list */
2004 switch (state) {
2005 case path_excluded:
2006 if (dir->flags & DIR_SHOW_IGNORED)
0ef8e169 2007 dir_add_name(dir, istate, path.buf, path.len);
0aaf62b6
KB
2008 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2009 ((dir->flags & DIR_COLLECT_IGNORED) &&
2010 exclude_matches_pathspec(path.buf, path.len,
e1b8c7bd 2011 pathspec)))
0ef8e169 2012 dir_add_ignored(dir, istate, path.buf, path.len);
defd7c7b
KB
2013 break;
2014
2015 case path_untracked:
0dcb8d7f
NTND
2016 if (dir->flags & DIR_SHOW_IGNORED)
2017 break;
0ef8e169 2018 dir_add_name(dir, istate, path.buf, path.len);
91a2288b 2019 if (cdir.fdir)
0dcb8d7f 2020 add_untracked(untracked, path.buf + baselen);
1528d247 2021 break;
defd7c7b
KB
2022
2023 default:
2024 break;
2025 }
453ec4bd 2026 }
cf7c6148 2027 close_cached_dir(&cdir);
1528d247 2028 out:
bef36921 2029 strbuf_release(&path);
453ec4bd 2030
defd7c7b 2031 return dir_state;
453ec4bd
LT
2032}
2033
bbf504a9 2034int cmp_dir_entry(const void *p1, const void *p2)
453ec4bd
LT
2035{
2036 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2037 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2038
ccdd4a0f 2039 return name_compare(e1->name, e1->len, e2->name, e2->len);
453ec4bd
LT
2040}
2041
fb898888 2042/* check if *out lexically strictly contains *in */
bbf504a9 2043int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
fb898888
SL
2044{
2045 return (out->len < in->len) &&
2046 (out->name[out->len - 1] == '/') &&
2047 !memcmp(out->name, in->name, out->len);
2048}
2049
48ffef96 2050static int treat_leading_path(struct dir_struct *dir,
0ef8e169 2051 struct index_state *istate,
48ffef96 2052 const char *path, int len,
e1b8c7bd 2053 const struct pathspec *pathspec)
48ffef96 2054{
49dc2cc2
RS
2055 struct strbuf sb = STRBUF_INIT;
2056 int baselen, rc = 0;
48ffef96 2057 const char *cp;
be8a84c5 2058 int old_flags = dir->flags;
48ffef96
JH
2059
2060 while (len && path[len - 1] == '/')
2061 len--;
2062 if (!len)
2063 return 1;
2064 baselen = 0;
be8a84c5 2065 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
48ffef96
JH
2066 while (1) {
2067 cp = path + baselen + !!baselen;
2068 cp = memchr(cp, '/', path + len - cp);
2069 if (!cp)
2070 baselen = len;
2071 else
2072 baselen = cp - path;
49dc2cc2
RS
2073 strbuf_setlen(&sb, 0);
2074 strbuf_add(&sb, path, baselen);
2075 if (!is_directory(sb.buf))
2076 break;
e1b8c7bd 2077 if (simplify_away(sb.buf, sb.len, pathspec))
49dc2cc2 2078 break;
0ef8e169 2079 if (treat_one_path(dir, NULL, istate, &sb, baselen, pathspec,
defd7c7b 2080 DT_DIR, NULL) == path_none)
49dc2cc2
RS
2081 break; /* do not recurse into it */
2082 if (len <= baselen) {
2083 rc = 1;
2084 break; /* finished checking */
2085 }
48ffef96 2086 }
49dc2cc2 2087 strbuf_release(&sb);
be8a84c5 2088 dir->flags = old_flags;
49dc2cc2 2089 return rc;
48ffef96
JH
2090}
2091
1e8fef60
NTND
2092static const char *get_ident_string(void)
2093{
2094 static struct strbuf sb = STRBUF_INIT;
2095 struct utsname uts;
2096
2097 if (sb.len)
2098 return sb.buf;
100e4337 2099 if (uname(&uts) < 0)
1e8fef60 2100 die_errno(_("failed to get kernel name and information"));
0e0f7618
CC
2101 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2102 uts.sysname);
1e8fef60
NTND
2103 return sb.buf;
2104}
2105
2106static int ident_in_untracked(const struct untracked_cache *uc)
2107{
0e0f7618
CC
2108 /*
2109 * Previous git versions may have saved many NUL separated
2110 * strings in the "ident" field, but it is insane to manage
2111 * many locations, so just take care of the first one.
2112 */
1e8fef60 2113
0e0f7618 2114 return !strcmp(uc->ident.buf, get_ident_string());
1e8fef60
NTND
2115}
2116
0e0f7618 2117static void set_untracked_ident(struct untracked_cache *uc)
1e8fef60 2118{
0e0f7618 2119 strbuf_reset(&uc->ident);
1e8fef60 2120 strbuf_addstr(&uc->ident, get_ident_string());
0e0f7618
CC
2121
2122 /*
2123 * This strbuf used to contain a list of NUL separated
2124 * strings, so save NUL too for backward compatibility.
2125 */
1e8fef60
NTND
2126 strbuf_addch(&uc->ident, 0);
2127}
2128
4a4ca479
CC
2129static void new_untracked_cache(struct index_state *istate)
2130{
2131 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2132 strbuf_init(&uc->ident, 100);
2133 uc->exclude_per_dir = ".gitignore";
2134 /* should be the same flags used by git-status */
2135 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
0e0f7618 2136 set_untracked_ident(uc);
4a4ca479 2137 istate->untracked = uc;
0e0f7618 2138 istate->cache_changed |= UNTRACKED_CHANGED;
4a4ca479
CC
2139}
2140
2141void add_untracked_cache(struct index_state *istate)
2142{
2143 if (!istate->untracked) {
2144 new_untracked_cache(istate);
0e0f7618
CC
2145 } else {
2146 if (!ident_in_untracked(istate->untracked)) {
2147 free_untracked_cache(istate->untracked);
2148 new_untracked_cache(istate);
2149 }
2150 }
4a4ca479
CC
2151}
2152
07b29bfd
CC
2153void remove_untracked_cache(struct index_state *istate)
2154{
2155 if (istate->untracked) {
2156 free_untracked_cache(istate->untracked);
2157 istate->untracked = NULL;
2158 istate->cache_changed |= UNTRACKED_CHANGED;
2159 }
2160}
2161
ccad261f
NTND
2162static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2163 int base_len,
2164 const struct pathspec *pathspec)
2165{
2166 struct untracked_cache_dir *root;
2167
76e6b090 2168 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
ccad261f
NTND
2169 return NULL;
2170
2171 /*
2172 * We only support $GIT_DIR/info/exclude and core.excludesfile
2173 * as the global ignore rule files. Any other additions
2174 * (e.g. from command line) invalidate the cache. This
2175 * condition also catches running setup_standard_excludes()
2176 * before setting dir->untracked!
2177 */
2178 if (dir->unmanaged_exclude_files)
2179 return NULL;
2180
2181 /*
2182 * Optimize for the main use case only: whole-tree git
2183 * status. More work involved in treat_leading_path() if we
2184 * use cache on just a subset of the worktree. pathspec
2185 * support could make the matter even worse.
2186 */
2187 if (base_len || (pathspec && pathspec->nr))
2188 return NULL;
2189
2190 /* Different set of flags may produce different results */
2191 if (dir->flags != dir->untracked->dir_flags ||
2192 /*
2193 * See treat_directory(), case index_nonexistent. Without
2194 * this flag, we may need to also cache .git file content
2195 * for the resolve_gitlink_ref() call, which we don't.
2196 */
2197 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2198 /* We don't support collecting ignore files */
2199 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2200 DIR_COLLECT_IGNORED)))
2201 return NULL;
2202
2203 /*
2204 * If we use .gitignore in the cache and now you change it to
2205 * .gitexclude, everything will go wrong.
2206 */
2207 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2208 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2209 return NULL;
2210
2211 /*
2212 * EXC_CMDL is not considered in the cache. If people set it,
2213 * skip the cache.
2214 */
2215 if (dir->exclude_list_group[EXC_CMDL].nr)
2216 return NULL;
2217
1e8fef60 2218 if (!ident_in_untracked(dir->untracked)) {
0e0f7618 2219 warning(_("Untracked cache is disabled on this system or location."));
1e8fef60
NTND
2220 return NULL;
2221 }
2222
ccad261f
NTND
2223 if (!dir->untracked->root) {
2224 const int len = sizeof(*dir->untracked->root);
2225 dir->untracked->root = xmalloc(len);
2226 memset(dir->untracked->root, 0, len);
2227 }
2228
2229 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2230 root = dir->untracked->root;
2231 if (hashcmp(dir->ss_info_exclude.sha1,
2232 dir->untracked->ss_info_exclude.sha1)) {
2233 invalidate_gitignore(dir->untracked, root);
2234 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2235 }
2236 if (hashcmp(dir->ss_excludes_file.sha1,
2237 dir->untracked->ss_excludes_file.sha1)) {
2238 invalidate_gitignore(dir->untracked, root);
2239 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2240 }
26cb0182
NTND
2241
2242 /* Make sure this directory is not dropped out at saving phase */
2243 root->recurse = 1;
ccad261f
NTND
2244 return root;
2245}
2246
2c1eb104
BW
2247int read_directory(struct dir_struct *dir, struct index_state *istate,
2248 const char *path, int len, const struct pathspec *pathspec)
9fc42d60 2249{
ccad261f 2250 struct untracked_cache_dir *untracked;
b4189aa8 2251
dba2e203 2252 if (has_symlink_leading_path(path, len))
725b0605
JH
2253 return dir->nr;
2254
ccad261f
NTND
2255 untracked = validate_untracked_cache(dir, len, pathspec);
2256 if (!untracked)
2257 /*
2258 * make sure untracked cache code path is disabled,
2259 * e.g. prep_exclude()
2260 */
2261 dir->untracked = NULL;
2c1eb104 2262 if (!len || treat_leading_path(dir, istate, path, len, pathspec))
5aaa7fd3 2263 read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
bbf504a9
SL
2264 QSORT(dir->entries, dir->nr, cmp_dir_entry);
2265 QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
fb898888
SL
2266
2267 /*
2268 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2269 * also pick up untracked contents of untracked dirs; by default
2270 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2271 */
2272 if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2273 !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2274 int i, j;
2275
2276 /* remove from dir->entries untracked contents of untracked dirs */
2277 for (i = j = 0; j < dir->nr; j++) {
bbf504a9
SL
2278 if (i &&
2279 check_dir_entry_contains(dir->entries[i - 1], dir->entries[j])) {
6a83d902 2280 FREE_AND_NULL(dir->entries[j]);
fb898888
SL
2281 } else {
2282 dir->entries[i++] = dir->entries[j];
2283 }
2284 }
2285
2286 dir->nr = i;
2287 }
2288
c9ccb5d3
NTND
2289 if (dir->untracked) {
2290 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2291 trace_printf_key(&trace_untracked_stats,
2292 "node creation: %u\n"
2293 "gitignore invalidation: %u\n"
2294 "directory invalidation: %u\n"
2295 "opendir: %u\n",
2296 dir->untracked->dir_created,
2297 dir->untracked->gitignore_invalidated,
2298 dir->untracked->dir_invalidated,
2299 dir->untracked->dir_opened);
2c1eb104 2300 if (dir->untracked == istate->untracked &&
1bbb3dba
NTND
2301 (dir->untracked->dir_opened ||
2302 dir->untracked->gitignore_invalidated ||
2303 dir->untracked->dir_invalidated))
2c1eb104
BW
2304 istate->cache_changed |= UNTRACKED_CHANGED;
2305 if (dir->untracked != istate->untracked) {
6a83d902 2306 FREE_AND_NULL(dir->untracked);
1bbb3dba 2307 }
c9ccb5d3 2308 }
453ec4bd
LT
2309 return dir->nr;
2310}
c91f0d92 2311
686a4a06 2312int file_exists(const char *f)
c91f0d92 2313{
686a4a06 2314 struct stat sb;
a50f9fc5 2315 return lstat(f, &sb) == 0;
c91f0d92 2316}
e6636747 2317
63ec5e1f
JS
2318static int cmp_icase(char a, char b)
2319{
2320 if (a == b)
2321 return 0;
2322 if (ignore_case)
2323 return toupper(a) - toupper(b);
2324 return a - b;
2325}
2326
e6636747 2327/*
9b125da4
NTND
2328 * Given two normalized paths (a trailing slash is ok), if subdir is
2329 * outside dir, return -1. Otherwise return the offset in subdir that
2330 * can be used as relative path to dir.
e6636747 2331 */
9b125da4 2332int dir_inside_of(const char *subdir, const char *dir)
e6636747 2333{
9b125da4 2334 int offset = 0;
e6636747 2335
9b125da4 2336 assert(dir && subdir && *dir && *subdir);
e6636747 2337
63ec5e1f 2338 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
e6636747 2339 dir++;
9b125da4
NTND
2340 subdir++;
2341 offset++;
490544b1 2342 }
9b125da4
NTND
2343
2344 /* hel[p]/me vs hel[l]/yeah */
2345 if (*dir && *subdir)
2346 return -1;
2347
2348 if (!*subdir)
2349 return !*dir ? offset : -1; /* same dir */
2350
2351 /* foo/[b]ar vs foo/[] */
2352 if (is_dir_sep(dir[-1]))
2353 return is_dir_sep(subdir[-1]) ? offset : -1;
2354
2355 /* foo[/]bar vs foo[] */
2356 return is_dir_sep(*subdir) ? offset + 1 : -1;
e6636747
JS
2357}
2358
2359int is_inside_dir(const char *dir)
2360{
56b9f6e7
RS
2361 char *cwd;
2362 int rc;
2363
b892913d
NTND
2364 if (!dir)
2365 return 0;
56b9f6e7
RS
2366
2367 cwd = xgetcwd();
2368 rc = (dir_inside_of(cwd, dir) >= 0);
2369 free(cwd);
2370 return rc;
e6636747 2371}
7155b727 2372
55892d23
AP
2373int is_empty_dir(const char *path)
2374{
2375 DIR *dir = opendir(path);
2376 struct dirent *e;
2377 int ret = 1;
2378
2379 if (!dir)
2380 return 0;
2381
2382 while ((e = readdir(dir)) != NULL)
2383 if (!is_dot_or_dotdot(e->d_name)) {
2384 ret = 0;
2385 break;
2386 }
2387
2388 closedir(dir);
2389 return ret;
2390}
2391
ae2f203e 2392static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
7155b727 2393{
a0f4afbe 2394 DIR *dir;
7155b727 2395 struct dirent *e;
ae2f203e 2396 int ret = 0, original_len = path->len, len, kept_down = 0;
a0f4afbe 2397 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
c844a803 2398 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1053fe82 2399 struct object_id submodule_head;
7155b727 2400
a0f4afbe 2401 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
a98e6101 2402 !resolve_gitlink_ref(path->buf, "HEAD", &submodule_head)) {
a0f4afbe 2403 /* Do not descend and nuke a nested git work tree. */
ae2f203e
JH
2404 if (kept_up)
2405 *kept_up = 1;
a0f4afbe 2406 return 0;
ae2f203e 2407 }
a0f4afbe 2408
ae2f203e 2409 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
a0f4afbe 2410 dir = opendir(path->buf);
c844a803 2411 if (!dir) {
863808cd
MH
2412 if (errno == ENOENT)
2413 return keep_toplevel ? -1 : 0;
2414 else if (errno == EACCES && !keep_toplevel)
ecb2c282
MH
2415 /*
2416 * An empty dir could be removable even if it
2417 * is unreadable:
2418 */
c844a803
JH
2419 return rmdir(path->buf);
2420 else
2421 return -1;
2422 }
00b6c178 2423 strbuf_complete(path, '/');
7155b727
JS
2424
2425 len = path->len;
2426 while ((e = readdir(dir)) != NULL) {
2427 struct stat st;
8ca12c0d
AP
2428 if (is_dot_or_dotdot(e->d_name))
2429 continue;
7155b727
JS
2430
2431 strbuf_setlen(path, len);
2432 strbuf_addstr(path, e->d_name);
863808cd
MH
2433 if (lstat(path->buf, &st)) {
2434 if (errno == ENOENT)
2435 /*
2436 * file disappeared, which is what we
2437 * wanted anyway
2438 */
2439 continue;
2440 /* fall thru */
2441 } else if (S_ISDIR(st.st_mode)) {
ae2f203e 2442 if (!remove_dir_recurse(path, flag, &kept_down))
7155b727 2443 continue; /* happy */
863808cd
MH
2444 } else if (!only_empty &&
2445 (!unlink(path->buf) || errno == ENOENT)) {
7155b727 2446 continue; /* happy, too */
863808cd 2447 }
7155b727
JS
2448
2449 /* path too long, stat fails, or non-directory still exists */
2450 ret = -1;
2451 break;
2452 }
2453 closedir(dir);
2454
2455 strbuf_setlen(path, original_len);
ae2f203e 2456 if (!ret && !keep_toplevel && !kept_down)
863808cd 2457 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
ae2f203e
JH
2458 else if (kept_up)
2459 /*
2460 * report the uplevel that it is not an error that we
2461 * did not rmdir() our directory.
2462 */
2463 *kept_up = !ret;
7155b727
JS
2464 return ret;
2465}
039bc64e 2466
ae2f203e
JH
2467int remove_dir_recursively(struct strbuf *path, int flag)
2468{
2469 return remove_dir_recurse(path, flag, NULL);
2470}
2471
f932729c
JK
2472static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2473
039bc64e
JH
2474void setup_standard_excludes(struct dir_struct *dir)
2475{
039bc64e 2476 dir->exclude_per_dir = ".gitignore";
099d2d86
JH
2477
2478 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2845ce7f
PT
2479 if (!excludes_file)
2480 excludes_file = xdg_config_home("ignore");
4698c8fe 2481 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
38ccaf93
JH
2482 add_excludes_from_file_1(dir, excludes_file,
2483 dir->untracked ? &dir->ss_excludes_file : NULL);
099d2d86
JH
2484
2485 /* per repository user preference */
f0056f64
JK
2486 if (startup_info->have_repository) {
2487 const char *path = git_path_info_exclude();
2488 if (!access_or_warn(path, R_OK, 0))
2489 add_excludes_from_file_1(dir, path,
2490 dir->untracked ? &dir->ss_info_exclude : NULL);
2491 }
039bc64e 2492}
4a92d1bf
AR
2493
2494int remove_path(const char *name)
2495{
2496 char *slash;
2497
c7054209 2498 if (unlink(name) && !is_missing_file_error(errno))
4a92d1bf
AR
2499 return -1;
2500
2501 slash = strrchr(name, '/');
2502 if (slash) {
2503 char *dirs = xstrdup(name);
2504 slash = dirs + (slash - name);
2505 do {
2506 *slash = '\0';
3fc0d131 2507 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
4a92d1bf
AR
2508 free(dirs);
2509 }
2510 return 0;
2511}
2512
270be816
AS
2513/*
2514 * Frees memory within dir which was allocated for exclude lists and
2515 * the exclude_stack. Does not free dir itself.
2516 */
2517void clear_directory(struct dir_struct *dir)
2518{
2519 int i, j;
2520 struct exclude_list_group *group;
2521 struct exclude_list *el;
2522 struct exclude_stack *stk;
2523
2524 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2525 group = &dir->exclude_list_group[i];
2526 for (j = 0; j < group->nr; j++) {
2527 el = &group->el[j];
2528 if (i == EXC_DIRS)
2529 free((char *)el->src);
2530 clear_exclude_list(el);
2531 }
2532 free(group->el);
2533 }
2534
2535 stk = dir->exclude_stack;
2536 while (stk) {
2537 struct exclude_stack *prev = stk->prev;
2538 free(stk);
2539 stk = prev;
2540 }
aceb9429 2541 strbuf_release(&dir->basebuf);
270be816 2542}
83c094ad
NTND
2543
2544struct ondisk_untracked_cache {
2545 struct stat_data info_exclude_stat;
2546 struct stat_data excludes_file_stat;
2547 uint32_t dir_flags;
2548 unsigned char info_exclude_sha1[20];
2549 unsigned char excludes_file_sha1[20];
2550 char exclude_per_dir[FLEX_ARRAY];
2551};
2552
268ba201
RS
2553#define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2554#define ouc_size(len) (ouc_offset(exclude_per_dir) + len + 1)
83c094ad
NTND
2555
2556struct write_data {
2557 int index; /* number of written untracked_cache_dir */
2558 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2559 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2560 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2561 struct strbuf out;
2562 struct strbuf sb_stat;
2563 struct strbuf sb_sha1;
2564};
2565
2566static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2567{
2568 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2569 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2570 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2571 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2572 to->sd_dev = htonl(from->sd_dev);
2573 to->sd_ino = htonl(from->sd_ino);
2574 to->sd_uid = htonl(from->sd_uid);
2575 to->sd_gid = htonl(from->sd_gid);
2576 to->sd_size = htonl(from->sd_size);
2577}
2578
2579static void write_one_dir(struct untracked_cache_dir *untracked,
2580 struct write_data *wd)
2581{
2582 struct stat_data stat_data;
2583 struct strbuf *out = &wd->out;
2584 unsigned char intbuf[16];
2585 unsigned int intlen, value;
2586 int i = wd->index++;
2587
2588 /*
2589 * untracked_nr should be reset whenever valid is clear, but
2590 * for safety..
2591 */
2592 if (!untracked->valid) {
2593 untracked->untracked_nr = 0;
2594 untracked->check_only = 0;
2595 }
2596
2597 if (untracked->check_only)
2598 ewah_set(wd->check_only, i);
2599 if (untracked->valid) {
2600 ewah_set(wd->valid, i);
2601 stat_data_to_disk(&stat_data, &untracked->stat_data);
2602 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2603 }
2604 if (!is_null_sha1(untracked->exclude_sha1)) {
2605 ewah_set(wd->sha1_valid, i);
2606 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2607 }
2608
2609 intlen = encode_varint(untracked->untracked_nr, intbuf);
2610 strbuf_add(out, intbuf, intlen);
2611
2612 /* skip non-recurse directories */
2613 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2614 if (untracked->dirs[i]->recurse)
2615 value++;
2616 intlen = encode_varint(value, intbuf);
2617 strbuf_add(out, intbuf, intlen);
2618
2619 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2620
2621 for (i = 0; i < untracked->untracked_nr; i++)
2622 strbuf_add(out, untracked->untracked[i],
2623 strlen(untracked->untracked[i]) + 1);
2624
2625 for (i = 0; i < untracked->dirs_nr; i++)
2626 if (untracked->dirs[i]->recurse)
2627 write_one_dir(untracked->dirs[i], wd);
2628}
2629
2630void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2631{
2632 struct ondisk_untracked_cache *ouc;
2633 struct write_data wd;
2634 unsigned char varbuf[16];
e0b83735
JK
2635 int varint_len;
2636 size_t len = strlen(untracked->exclude_per_dir);
2637
2638 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
83c094ad
NTND
2639 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2640 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2641 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2642 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2643 ouc->dir_flags = htonl(untracked->dir_flags);
1e8fef60
NTND
2644
2645 varint_len = encode_varint(untracked->ident.len, varbuf);
2646 strbuf_add(out, varbuf, varint_len);
8109984d 2647 strbuf_addbuf(out, &untracked->ident);
1e8fef60 2648
83c094ad 2649 strbuf_add(out, ouc, ouc_size(len));
6a83d902 2650 FREE_AND_NULL(ouc);
83c094ad
NTND
2651
2652 if (!untracked->root) {
2653 varint_len = encode_varint(0, varbuf);
2654 strbuf_add(out, varbuf, varint_len);
2655 return;
2656 }
2657
2658 wd.index = 0;
2659 wd.check_only = ewah_new();
2660 wd.valid = ewah_new();
2661 wd.sha1_valid = ewah_new();
2662 strbuf_init(&wd.out, 1024);
2663 strbuf_init(&wd.sb_stat, 1024);
2664 strbuf_init(&wd.sb_sha1, 1024);
2665 write_one_dir(untracked->root, &wd);
2666
2667 varint_len = encode_varint(wd.index, varbuf);
2668 strbuf_add(out, varbuf, varint_len);
2669 strbuf_addbuf(out, &wd.out);
2670 ewah_serialize_strbuf(wd.valid, out);
2671 ewah_serialize_strbuf(wd.check_only, out);
2672 ewah_serialize_strbuf(wd.sha1_valid, out);
2673 strbuf_addbuf(out, &wd.sb_stat);
2674 strbuf_addbuf(out, &wd.sb_sha1);
2675 strbuf_addch(out, '\0'); /* safe guard for string lists */
2676
2677 ewah_free(wd.valid);
2678 ewah_free(wd.check_only);
2679 ewah_free(wd.sha1_valid);
2680 strbuf_release(&wd.out);
2681 strbuf_release(&wd.sb_stat);
2682 strbuf_release(&wd.sb_sha1);
2683}
f9e6c649
NTND
2684
2685static void free_untracked(struct untracked_cache_dir *ucd)
2686{
2687 int i;
2688 if (!ucd)
2689 return;
2690 for (i = 0; i < ucd->dirs_nr; i++)
2691 free_untracked(ucd->dirs[i]);
2692 for (i = 0; i < ucd->untracked_nr; i++)
2693 free(ucd->untracked[i]);
2694 free(ucd->untracked);
2695 free(ucd->dirs);
2696 free(ucd);
2697}
2698
2699void free_untracked_cache(struct untracked_cache *uc)
2700{
2701 if (uc)
2702 free_untracked(uc->root);
2703 free(uc);
2704}
2705
2706struct read_data {
2707 int index;
2708 struct untracked_cache_dir **ucd;
2709 struct ewah_bitmap *check_only;
2710 struct ewah_bitmap *valid;
2711 struct ewah_bitmap *sha1_valid;
2712 const unsigned char *data;
2713 const unsigned char *end;
2714};
2715
268ba201 2716static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
f9e6c649 2717{
268ba201
RS
2718 memcpy(to, data, sizeof(*to));
2719 to->sd_ctime.sec = ntohl(to->sd_ctime.sec);
2720 to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
2721 to->sd_mtime.sec = ntohl(to->sd_mtime.sec);
2722 to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
2723 to->sd_dev = ntohl(to->sd_dev);
2724 to->sd_ino = ntohl(to->sd_ino);
2725 to->sd_uid = ntohl(to->sd_uid);
2726 to->sd_gid = ntohl(to->sd_gid);
2727 to->sd_size = ntohl(to->sd_size);
f9e6c649
NTND
2728}
2729
2730static int read_one_dir(struct untracked_cache_dir **untracked_,
2731 struct read_data *rd)
2732{
2733 struct untracked_cache_dir ud, *untracked;
2734 const unsigned char *next, *data = rd->data, *end = rd->end;
2735 unsigned int value;
2736 int i, len;
2737
2738 memset(&ud, 0, sizeof(ud));
2739
2740 next = data;
2741 value = decode_varint(&next);
2742 if (next > end)
2743 return -1;
2744 ud.recurse = 1;
2745 ud.untracked_alloc = value;
2746 ud.untracked_nr = value;
2747 if (ud.untracked_nr)
b32fa95f 2748 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
f9e6c649
NTND
2749 data = next;
2750
2751 next = data;
2752 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2753 if (next > end)
2754 return -1;
b32fa95f 2755 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
f9e6c649
NTND
2756 data = next;
2757
2758 len = strlen((const char *)data);
2759 next = data + len + 1;
2760 if (next > rd->end)
2761 return -1;
50a6c8ef 2762 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
f9e6c649
NTND
2763 memcpy(untracked, &ud, sizeof(ud));
2764 memcpy(untracked->name, data, len + 1);
2765 data = next;
2766
2767 for (i = 0; i < untracked->untracked_nr; i++) {
2768 len = strlen((const char *)data);
2769 next = data + len + 1;
2770 if (next > rd->end)
2771 return -1;
2772 untracked->untracked[i] = xstrdup((const char*)data);
2773 data = next;
2774 }
2775
2776 rd->ucd[rd->index++] = untracked;
2777 rd->data = data;
2778
2779 for (i = 0; i < untracked->dirs_nr; i++) {
2780 len = read_one_dir(untracked->dirs + i, rd);
2781 if (len < 0)
2782 return -1;
2783 }
2784 return 0;
2785}
2786
2787static void set_check_only(size_t pos, void *cb)
2788{
2789 struct read_data *rd = cb;
2790 struct untracked_cache_dir *ud = rd->ucd[pos];
2791 ud->check_only = 1;
2792}
2793
2794static void read_stat(size_t pos, void *cb)
2795{
2796 struct read_data *rd = cb;
2797 struct untracked_cache_dir *ud = rd->ucd[pos];
2798 if (rd->data + sizeof(struct stat_data) > rd->end) {
2799 rd->data = rd->end + 1;
2800 return;
2801 }
268ba201 2802 stat_data_from_disk(&ud->stat_data, rd->data);
f9e6c649
NTND
2803 rd->data += sizeof(struct stat_data);
2804 ud->valid = 1;
2805}
2806
2807static void read_sha1(size_t pos, void *cb)
2808{
2809 struct read_data *rd = cb;
2810 struct untracked_cache_dir *ud = rd->ucd[pos];
2811 if (rd->data + 20 > rd->end) {
2812 rd->data = rd->end + 1;
2813 return;
2814 }
2815 hashcpy(ud->exclude_sha1, rd->data);
2816 rd->data += 20;
2817}
2818
2819static void load_sha1_stat(struct sha1_stat *sha1_stat,
268ba201 2820 const unsigned char *data,
f9e6c649
NTND
2821 const unsigned char *sha1)
2822{
268ba201 2823 stat_data_from_disk(&sha1_stat->stat, data);
f9e6c649
NTND
2824 hashcpy(sha1_stat->sha1, sha1);
2825 sha1_stat->valid = 1;
2826}
2827
2828struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2829{
f9e6c649
NTND
2830 struct untracked_cache *uc;
2831 struct read_data rd;
2832 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
1e8fef60
NTND
2833 const char *ident;
2834 int ident_len, len;
268ba201 2835 const char *exclude_per_dir;
f9e6c649
NTND
2836
2837 if (sz <= 1 || end[-1] != '\0')
2838 return NULL;
2839 end--;
2840
1e8fef60
NTND
2841 ident_len = decode_varint(&next);
2842 if (next + ident_len > end)
2843 return NULL;
2844 ident = (const char *)next;
2845 next += ident_len;
2846
f9e6c649
NTND
2847 if (next + ouc_size(0) > end)
2848 return NULL;
2849
2850 uc = xcalloc(1, sizeof(*uc));
1e8fef60
NTND
2851 strbuf_init(&uc->ident, ident_len);
2852 strbuf_add(&uc->ident, ident, ident_len);
268ba201
RS
2853 load_sha1_stat(&uc->ss_info_exclude,
2854 next + ouc_offset(info_exclude_stat),
2855 next + ouc_offset(info_exclude_sha1));
2856 load_sha1_stat(&uc->ss_excludes_file,
2857 next + ouc_offset(excludes_file_stat),
2858 next + ouc_offset(excludes_file_sha1));
2859 uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
2860 exclude_per_dir = (const char *)next + ouc_offset(exclude_per_dir);
2861 uc->exclude_per_dir = xstrdup(exclude_per_dir);
f9e6c649 2862 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
268ba201 2863 next += ouc_size(strlen(exclude_per_dir));
f9e6c649
NTND
2864 if (next >= end)
2865 goto done2;
2866
2867 len = decode_varint(&next);
2868 if (next > end || len == 0)
2869 goto done2;
2870
2871 rd.valid = ewah_new();
2872 rd.check_only = ewah_new();
2873 rd.sha1_valid = ewah_new();
2874 rd.data = next;
2875 rd.end = end;
2876 rd.index = 0;
b32fa95f 2877 ALLOC_ARRAY(rd.ucd, len);
f9e6c649
NTND
2878
2879 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2880 goto done;
2881
2882 next = rd.data;
2883 len = ewah_read_mmap(rd.valid, next, end - next);
2884 if (len < 0)
2885 goto done;
2886
2887 next += len;
2888 len = ewah_read_mmap(rd.check_only, next, end - next);
2889 if (len < 0)
2890 goto done;
2891
2892 next += len;
2893 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2894 if (len < 0)
2895 goto done;
2896
2897 ewah_each_bit(rd.check_only, set_check_only, &rd);
2898 rd.data = next + len;
2899 ewah_each_bit(rd.valid, read_stat, &rd);
2900 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2901 next = rd.data;
2902
2903done:
2904 free(rd.ucd);
2905 ewah_free(rd.valid);
2906 ewah_free(rd.check_only);
2907 ewah_free(rd.sha1_valid);
2908done2:
2909 if (next != end) {
2910 free_untracked_cache(uc);
2911 uc = NULL;
2912 }
2913 return uc;
2914}
e931371a 2915
73f9145f
NTND
2916static void invalidate_one_directory(struct untracked_cache *uc,
2917 struct untracked_cache_dir *ucd)
2918{
2919 uc->dir_invalidated++;
2920 ucd->valid = 0;
2921 ucd->untracked_nr = 0;
2922}
2923
2924/*
2925 * Normally when an entry is added or removed from a directory,
2926 * invalidating that directory is enough. No need to touch its
2927 * ancestors. When a directory is shown as "foo/bar/" in git-status
2928 * however, deleting or adding an entry may have cascading effect.
2929 *
2930 * Say the "foo/bar/file" has become untracked, we need to tell the
2931 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2932 * directory any more (because "bar" is managed by foo as an untracked
2933 * "file").
2934 *
2935 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2936 * was the last untracked entry in the entire "foo", we should show
2937 * "foo/" instead. Which means we have to invalidate past "bar" up to
2938 * "foo".
2939 *
2940 * This function traverses all directories from root to leaf. If there
2941 * is a chance of one of the above cases happening, we invalidate back
2942 * to root. Otherwise we just invalidate the leaf. There may be a more
2943 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2944 * detect these cases and avoid unnecessary invalidation, for example,
2945 * checking for the untracked entry named "bar/" in "foo", but for now
2946 * stick to something safe and simple.
2947 */
2948static int invalidate_one_component(struct untracked_cache *uc,
2949 struct untracked_cache_dir *dir,
2950 const char *path, int len)
2951{
2952 const char *rest = strchr(path, '/');
2953
2954 if (rest) {
2955 int component_len = rest - path;
2956 struct untracked_cache_dir *d =
2957 lookup_untracked(uc, dir, path, component_len);
2958 int ret =
2959 invalidate_one_component(uc, d, rest + 1,
2960 len - (component_len + 1));
2961 if (ret)
2962 invalidate_one_directory(uc, dir);
2963 return ret;
2964 }
2965
2966 invalidate_one_directory(uc, dir);
2967 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2968}
2969
e931371a
NTND
2970void untracked_cache_invalidate_path(struct index_state *istate,
2971 const char *path)
2972{
e931371a
NTND
2973 if (!istate->untracked || !istate->untracked->root)
2974 return;
73f9145f
NTND
2975 invalidate_one_component(istate->untracked, istate->untracked->root,
2976 path, strlen(path));
e931371a
NTND
2977}
2978
2979void untracked_cache_remove_from_index(struct index_state *istate,
2980 const char *path)
2981{
2982 untracked_cache_invalidate_path(istate, path);
2983}
2984
2985void untracked_cache_add_to_index(struct index_state *istate,
2986 const char *path)
2987{
2988 untracked_cache_invalidate_path(istate, path);
2989}
47e83eb3
SB
2990
2991/* Update gitfile and core.worktree setting to connect work tree and git dir */
2992void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)
2993{
365444a6
SB
2994 struct strbuf gitfile_sb = STRBUF_INIT;
2995 struct strbuf cfg_sb = STRBUF_INIT;
47e83eb3 2996 struct strbuf rel_path = STRBUF_INIT;
365444a6 2997 char *git_dir, *work_tree;
47e83eb3 2998
365444a6
SB
2999 /* Prepare .git file */
3000 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
3001 if (safe_create_leading_directories_const(gitfile_sb.buf))
3002 die(_("could not create directories for %s"), gitfile_sb.buf);
3003
3004 /* Prepare config file */
3005 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
3006 if (safe_create_leading_directories_const(cfg_sb.buf))
3007 die(_("could not create directories for %s"), cfg_sb.buf);
47e83eb3 3008
e394fa01
JH
3009 git_dir = real_pathdup(git_dir_, 1);
3010 work_tree = real_pathdup(work_tree_, 1);
365444a6
SB
3011
3012 /* Write .git file */
3013 write_file(gitfile_sb.buf, "gitdir: %s",
3014 relative_path(git_dir, work_tree, &rel_path));
47e83eb3 3015 /* Update core.worktree setting */
365444a6 3016 git_config_set_in_file(cfg_sb.buf, "core.worktree",
47e83eb3
SB
3017 relative_path(work_tree, git_dir, &rel_path));
3018
365444a6
SB
3019 strbuf_release(&gitfile_sb);
3020 strbuf_release(&cfg_sb);
47e83eb3
SB
3021 strbuf_release(&rel_path);
3022 free(work_tree);
3023 free(git_dir);
3024}
f6f85861
SB
3025
3026/*
3027 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3028 */
3029void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
3030{
3031 if (rename(old_git_dir, new_git_dir) < 0)
3032 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3033 old_git_dir, new_git_dir);
3034
3035 connect_work_tree_and_git_dir(path, new_git_dir);
3036}