dir: convert last_exclude_matching_from_list to take an index
[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
LT
11#include "cache.h"
12#include "dir.h"
b0db7046 13#include "attr.h"
09595258 14#include "refs.h"
237ec6e4 15#include "wildmatch.h"
64acde94 16#include "pathspec.h"
dde843e7 17#include "utf8.h"
83c094ad
NTND
18#include "varint.h"
19#include "ewah/ewok.h"
453ec4bd 20
defd7c7b
KB
21/*
22 * Tells read_directory_recursive how a file or directory should be treated.
23 * Values are ordered by significance, e.g. if a directory contains both
24 * excluded and untracked files, it is listed as untracked because
25 * path_untracked > path_excluded.
26 */
27enum path_treatment {
28 path_none = 0,
29 path_recurse,
30 path_excluded,
31 path_untracked
32};
33
cf7c6148
NTND
34/*
35 * Support data structure for our opendir/readdir/closedir wrappers
36 */
37struct cached_dir {
38 DIR *fdir;
39 struct untracked_cache_dir *untracked;
91a2288b
NTND
40 int nr_files;
41 int nr_dirs;
42
cf7c6148 43 struct dirent *de;
91a2288b
NTND
44 const char *file;
45 struct untracked_cache_dir *ucd;
cf7c6148
NTND
46};
47
defd7c7b 48static enum path_treatment read_directory_recursive(struct dir_struct *dir,
0dcb8d7f 49 const char *path, int len, struct untracked_cache_dir *untracked,
e1b8c7bd 50 int check_only, const struct pathspec *pathspec);
98f2a687
BW
51static int get_dtype(struct dirent *de, struct index_state *istate,
52 const char *path, int len);
09595258 53
ba0897e6 54int fspathcmp(const char *a, const char *b)
8cf2a84e
JJ
55{
56 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
57}
58
ba0897e6 59int fspathncmp(const char *a, const char *b, size_t count)
8cf2a84e
JJ
60{
61 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
62}
63
1f26ce61
CB
64int git_fnmatch(const struct pathspec_item *item,
65 const char *pattern, const char *string,
66 int prefix)
5d74762d 67{
5d74762d 68 if (prefix > 0) {
93d93537 69 if (ps_strncmp(item, pattern, string, prefix))
eb07894f 70 return WM_NOMATCH;
5d74762d
NTND
71 pattern += prefix;
72 string += prefix;
73 }
bd30c2e4 74 if (item->flags & PATHSPEC_ONESTAR) {
8c6abbcd
NTND
75 int pattern_len = strlen(++pattern);
76 int string_len = strlen(string);
77 return string_len < pattern_len ||
93d93537
NTND
78 ps_strcmp(item, pattern,
79 string + string_len - pattern_len);
8c6abbcd 80 }
bd30c2e4 81 if (item->magic & PATHSPEC_GLOB)
93d93537
NTND
82 return wildmatch(pattern, string,
83 WM_PATHNAME |
84 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
85 NULL);
bd30c2e4
NTND
86 else
87 /* wildmatch has not learned no FNM_PATHNAME mode yet */
eb07894f
NTND
88 return wildmatch(pattern, string,
89 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
90 NULL);
5d74762d
NTND
91}
92
0b6e56df
JH
93static int fnmatch_icase_mem(const char *pattern, int patternlen,
94 const char *string, int stringlen,
95 int flags)
96{
97 int match_status;
98 struct strbuf pat_buf = STRBUF_INIT;
99 struct strbuf str_buf = STRBUF_INIT;
100 const char *use_pat = pattern;
101 const char *use_str = string;
102
103 if (pattern[patternlen]) {
104 strbuf_add(&pat_buf, pattern, patternlen);
105 use_pat = pat_buf.buf;
106 }
107 if (string[stringlen]) {
108 strbuf_add(&str_buf, string, stringlen);
109 use_str = str_buf.buf;
110 }
111
f30366b2
JH
112 if (ignore_case)
113 flags |= WM_CASEFOLD;
114 match_status = wildmatch(use_pat, use_str, flags, NULL);
0b6e56df
JH
115
116 strbuf_release(&pat_buf);
117 strbuf_release(&str_buf);
118
119 return match_status;
120}
121
827f4d6c 122static size_t common_prefix_len(const struct pathspec *pathspec)
3c6a370b 123{
827f4d6c 124 int n;
4a085b16 125 size_t max = 0;
3c6a370b 126
93d93537
NTND
127 /*
128 * ":(icase)path" is treated as a pathspec full of
129 * wildcard. In other words, only prefix is considered common
130 * prefix. If the pathspec is abc/foo abc/bar, running in
131 * subdir xyz, the common prefix is still xyz, not xuz/abc as
132 * in non-:(icase).
133 */
5c6933d2
NTND
134 GUARD_PATHSPEC(pathspec,
135 PATHSPEC_FROMTOP |
136 PATHSPEC_MAXDEPTH |
bd30c2e4 137 PATHSPEC_LITERAL |
93d93537 138 PATHSPEC_GLOB |
ef79b1f8 139 PATHSPEC_ICASE |
b0db7046
BW
140 PATHSPEC_EXCLUDE |
141 PATHSPEC_ATTR);
4a085b16 142
827f4d6c 143 for (n = 0; n < pathspec->nr; n++) {
93d93537 144 size_t i = 0, len = 0, item_len;
ef79b1f8
NTND
145 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
146 continue;
93d93537
NTND
147 if (pathspec->items[n].magic & PATHSPEC_ICASE)
148 item_len = pathspec->items[n].prefix;
149 else
150 item_len = pathspec->items[n].nowildcard_len;
151 while (i < item_len && (n == 0 || i < max)) {
827f4d6c
NTND
152 char c = pathspec->items[n].match[i];
153 if (c != pathspec->items[0].match[i])
4a085b16
JH
154 break;
155 if (c == '/')
156 len = i + 1;
827f4d6c 157 i++;
4a085b16 158 }
827f4d6c 159 if (n == 0 || len < max) {
4a085b16
JH
160 max = len;
161 if (!max)
162 break;
163 }
3c6a370b 164 }
4a085b16 165 return max;
3c6a370b
LT
166}
167
f950eb95
CB
168/*
169 * Returns a copy of the longest leading path common among all
170 * pathspecs.
171 */
827f4d6c 172char *common_prefix(const struct pathspec *pathspec)
f950eb95
CB
173{
174 unsigned long len = common_prefix_len(pathspec);
175
827f4d6c 176 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
f950eb95
CB
177}
178
7327d3d1 179int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
1d8842d9 180{
bec5ab89 181 const char *prefix;
966de302 182 size_t prefix_len;
1d8842d9
LT
183
184 /*
185 * Calculate common prefix for the pathspec, and
186 * use that to optimize the directory walk
187 */
bec5ab89
RS
188 prefix_len = common_prefix_len(pathspec);
189 prefix = prefix_len ? pathspec->items[0].match : "";
1d8842d9
LT
190
191 /* Read the directory and prune it */
966de302
BW
192 read_directory(dir, prefix, prefix_len, pathspec);
193
966de302 194 return prefix_len;
1d8842d9
LT
195}
196
bc96cc87
NTND
197int within_depth(const char *name, int namelen,
198 int depth, int max_depth)
199{
200 const char *cp = name, *cpe = name + namelen;
201
202 while (cp < cpe) {
203 if (*cp++ != '/')
204 continue;
205 depth++;
206 if (depth > max_depth)
207 return 0;
208 }
209 return 1;
210}
211
75a6315f
BW
212#define DO_MATCH_EXCLUDE (1<<0)
213#define DO_MATCH_DIRECTORY (1<<1)
214#define DO_MATCH_SUBMODULE (1<<2)
42b0874a 215
b0db7046
BW
216static int match_attrs(const char *name, int namelen,
217 const struct pathspec_item *item)
218{
219 int i;
220
221 git_check_attr(name, item->attr_check);
222 for (i = 0; i < item->attr_match_nr; i++) {
223 const char *value;
224 int matched;
225 enum attr_match_mode match_mode;
226
227 value = item->attr_check->items[i].value;
228 match_mode = item->attr_match[i].match_mode;
229
230 if (ATTR_TRUE(value))
231 matched = (match_mode == MATCH_SET);
232 else if (ATTR_FALSE(value))
233 matched = (match_mode == MATCH_UNSET);
234 else if (ATTR_UNSET(value))
235 matched = (match_mode == MATCH_UNSPECIFIED);
236 else
237 matched = (match_mode == MATCH_VALUE &&
238 !strcmp(item->attr_match[i].value, value));
239 if (!matched)
240 return 0;
241 }
242
243 return 1;
244}
245
61cf2820
NTND
246/*
247 * Does 'match' match the given name?
248 * A match is found if
249 *
250 * (1) the 'match' string is leading directory of 'name', or
251 * (2) the 'match' string is a wildcard and matches 'name', or
252 * (3) the 'match' string is exactly the same as 'name'.
253 *
254 * and the return value tells which case it was.
255 *
256 * It returns 0 when there is no match.
257 */
258static int match_pathspec_item(const struct pathspec_item *item, int prefix,
42b0874a 259 const char *name, int namelen, unsigned flags)
61cf2820
NTND
260{
261 /* name/namelen has prefix cut off by caller */
262 const char *match = item->match + prefix;
263 int matchlen = item->len - prefix;
264
93d93537
NTND
265 /*
266 * The normal call pattern is:
267 * 1. prefix = common_prefix_len(ps);
268 * 2. prune something, or fill_directory
854b0959 269 * 3. match_pathspec()
93d93537
NTND
270 *
271 * 'prefix' at #1 may be shorter than the command's prefix and
272 * it's ok for #2 to match extra files. Those extras will be
273 * trimmed at #3.
274 *
275 * Suppose the pathspec is 'foo' and '../bar' running from
276 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
277 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
278 * user does not want XYZ/foo, only the "foo" part should be
279 * case-insensitive. We need to filter out XYZ/foo here. In
280 * other words, we do not trust the caller on comparing the
281 * prefix part when :(icase) is involved. We do exact
282 * comparison ourselves.
283 *
284 * Normally the caller (common_prefix_len() in fact) does
285 * _exact_ matching on name[-prefix+1..-1] and we do not need
286 * to check that part. Be defensive and check it anyway, in
287 * case common_prefix_len is changed, or a new caller is
288 * introduced that does not use common_prefix_len.
289 *
290 * If the penalty turns out too high when prefix is really
291 * long, maybe change it to
292 * strncmp(match, name, item->prefix - prefix)
293 */
294 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
295 strncmp(item->match, name - prefix, item->prefix))
296 return 0;
297
b0db7046
BW
298 if (item->attr_match_nr && !match_attrs(name, namelen, item))
299 return 0;
300
61cf2820
NTND
301 /* If the match was just the prefix, we matched */
302 if (!*match)
303 return MATCHED_RECURSIVELY;
304
93d93537 305 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
61cf2820
NTND
306 if (matchlen == namelen)
307 return MATCHED_EXACTLY;
308
309 if (match[matchlen-1] == '/' || name[matchlen] == '/')
310 return MATCHED_RECURSIVELY;
68690fdd
NTND
311 } else if ((flags & DO_MATCH_DIRECTORY) &&
312 match[matchlen - 1] == '/' &&
313 namelen == matchlen - 1 &&
314 !ps_strncmp(item, match, name, namelen))
315 return MATCHED_EXACTLY;
61cf2820 316
5d74762d 317 if (item->nowildcard_len < item->len &&
bd30c2e4 318 !git_fnmatch(item, match, name,
8c6abbcd 319 item->nowildcard_len - prefix))
61cf2820
NTND
320 return MATCHED_FNMATCH;
321
75a6315f
BW
322 /* Perform checks to see if "name" is a super set of the pathspec */
323 if (flags & DO_MATCH_SUBMODULE) {
324 /* name is a literal prefix of the pathspec */
325 if ((namelen < matchlen) &&
326 (match[namelen] == '/') &&
327 !ps_strncmp(item, match, name, namelen))
328 return MATCHED_RECURSIVELY;
329
330 /* name" doesn't match up to the first wild character */
331 if (item->nowildcard_len < item->len &&
332 ps_strncmp(item, match, name,
333 item->nowildcard_len - prefix))
334 return 0;
335
336 /*
337 * Here is where we would perform a wildmatch to check if
338 * "name" can be matched as a directory (or a prefix) against
339 * the pathspec. Since wildmatch doesn't have this capability
340 * at the present we have to punt and say that it is a match,
341 * potentially returning a false positive
342 * The submodules themselves will be able to perform more
343 * accurate matching to determine if the pathspec matches.
344 */
345 return MATCHED_RECURSIVELY;
346 }
347
61cf2820
NTND
348 return 0;
349}
350
351/*
52ed1894
AS
352 * Given a name and a list of pathspecs, returns the nature of the
353 * closest (i.e. most specific) match of the name to any of the
354 * pathspecs.
355 *
356 * The caller typically calls this multiple times with the same
357 * pathspec and seen[] array but with different name/namelen
358 * (e.g. entries from the index) and is interested in seeing if and
359 * how each pathspec matches all the names it calls this function
360 * with. A mark is left in the seen[] array for each pathspec element
361 * indicating the closest type of match that element achieved, so if
362 * seen[n] remains zero after multiple invocations, that means the nth
363 * pathspec did not match any names, which could indicate that the
364 * user mistyped the nth pathspec.
61cf2820 365 */
854b0959
NTND
366static int do_match_pathspec(const struct pathspec *ps,
367 const char *name, int namelen,
368 int prefix, char *seen,
42b0874a 369 unsigned flags)
61cf2820 370{
42b0874a 371 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
61cf2820 372
5c6933d2
NTND
373 GUARD_PATHSPEC(ps,
374 PATHSPEC_FROMTOP |
375 PATHSPEC_MAXDEPTH |
bd30c2e4 376 PATHSPEC_LITERAL |
93d93537 377 PATHSPEC_GLOB |
ef79b1f8 378 PATHSPEC_ICASE |
b0db7046
BW
379 PATHSPEC_EXCLUDE |
380 PATHSPEC_ATTR);
8f4f8f45 381
61cf2820 382 if (!ps->nr) {
6330a171
NTND
383 if (!ps->recursive ||
384 !(ps->magic & PATHSPEC_MAXDEPTH) ||
385 ps->max_depth == -1)
61cf2820
NTND
386 return MATCHED_RECURSIVELY;
387
388 if (within_depth(name, namelen, 0, ps->max_depth))
389 return MATCHED_EXACTLY;
390 else
391 return 0;
392 }
393
394 name += prefix;
395 namelen -= prefix;
396
397 for (i = ps->nr - 1; i >= 0; i--) {
398 int how;
ef79b1f8
NTND
399
400 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
401 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
402 continue;
403
61cf2820
NTND
404 if (seen && seen[i] == MATCHED_EXACTLY)
405 continue;
ef79b1f8
NTND
406 /*
407 * Make exclude patterns optional and never report
408 * "pathspec ':(exclude)foo' matches no files"
409 */
410 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
411 seen[i] = MATCHED_FNMATCH;
42b0874a
NTND
412 how = match_pathspec_item(ps->items+i, prefix, name,
413 namelen, flags);
6330a171
NTND
414 if (ps->recursive &&
415 (ps->magic & PATHSPEC_MAXDEPTH) &&
416 ps->max_depth != -1 &&
61cf2820
NTND
417 how && how != MATCHED_FNMATCH) {
418 int len = ps->items[i].len;
419 if (name[len] == '/')
420 len++;
421 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
422 how = MATCHED_EXACTLY;
423 else
424 how = 0;
425 }
426 if (how) {
427 if (retval < how)
428 retval = how;
429 if (seen && seen[i] < how)
430 seen[i] = how;
431 }
432 }
433 return retval;
434}
435
854b0959
NTND
436int match_pathspec(const struct pathspec *ps,
437 const char *name, int namelen,
ae8d0824 438 int prefix, char *seen, int is_dir)
ef79b1f8
NTND
439{
440 int positive, negative;
ae8d0824 441 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
42b0874a
NTND
442 positive = do_match_pathspec(ps, name, namelen,
443 prefix, seen, flags);
ef79b1f8
NTND
444 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
445 return positive;
42b0874a
NTND
446 negative = do_match_pathspec(ps, name, namelen,
447 prefix, seen,
448 flags | DO_MATCH_EXCLUDE);
ef79b1f8
NTND
449 return negative ? 0 : positive;
450}
451
75a6315f
BW
452/**
453 * Check if a submodule is a superset of the pathspec
454 */
455int submodule_path_match(const struct pathspec *ps,
456 const char *submodule_name,
457 char *seen)
458{
459 int matched = do_match_pathspec(ps, submodule_name,
460 strlen(submodule_name),
461 0, seen,
462 DO_MATCH_DIRECTORY |
463 DO_MATCH_SUBMODULE);
464 return matched;
465}
466
777c55a6
JH
467int report_path_error(const char *ps_matched,
468 const struct pathspec *pathspec,
469 const char *prefix)
470{
471 /*
472 * Make sure all pathspec matched; otherwise it is an error.
473 */
777c55a6
JH
474 int num, errors = 0;
475 for (num = 0; num < pathspec->nr; num++) {
476 int other, found_dup;
477
478 if (ps_matched[num])
479 continue;
480 /*
481 * The caller might have fed identical pathspec
482 * twice. Do not barf on such a mistake.
483 * FIXME: parse_pathspec should have eliminated
484 * duplicate pathspec.
485 */
486 for (found_dup = other = 0;
487 !found_dup && other < pathspec->nr;
488 other++) {
489 if (other == num || !ps_matched[other])
490 continue;
491 if (!strcmp(pathspec->items[other].original,
492 pathspec->items[num].original))
493 /*
494 * Ok, we have a match already.
495 */
496 found_dup = 1;
497 }
498 if (found_dup)
499 continue;
500
501 error("pathspec '%s' did not match any file(s) known to git.",
502 pathspec->items[num].original);
503 errors++;
504 }
777c55a6
JH
505 return errors;
506}
507
fcd631ed
NTND
508/*
509 * Return the length of the "simple" part of a path match limiter.
510 */
87323bda 511int simple_length(const char *match)
fcd631ed
NTND
512{
513 int len = -1;
514
515 for (;;) {
516 unsigned char c = *match++;
517 len++;
518 if (c == '\0' || is_glob_special(c))
519 return len;
520 }
521}
522
87323bda 523int no_wildcard(const char *string)
68492fc7 524{
fcd631ed 525 return string[simple_length(string)] == '\0';
68492fc7
LK
526}
527
82dce998
NTND
528void parse_exclude_pattern(const char **pattern,
529 int *patternlen,
f8708998 530 unsigned *flags,
82dce998 531 int *nowildcardlen)
84460eec
NTND
532{
533 const char *p = *pattern;
534 size_t i, len;
535
536 *flags = 0;
537 if (*p == '!') {
538 *flags |= EXC_FLAG_NEGATIVE;
539 p++;
540 }
541 len = strlen(p);
542 if (len && p[len - 1] == '/') {
543 len--;
544 *flags |= EXC_FLAG_MUSTBEDIR;
545 }
546 for (i = 0; i < len; i++) {
547 if (p[i] == '/')
548 break;
549 }
550 if (i == len)
551 *flags |= EXC_FLAG_NODIR;
552 *nowildcardlen = simple_length(p);
553 /*
554 * we should have excluded the trailing slash from 'p' too,
555 * but that's one more allocation. Instead just make sure
556 * nowildcardlen does not exceed real patternlen
557 */
558 if (*nowildcardlen > len)
559 *nowildcardlen = len;
560 if (*p == '*' && no_wildcard(p + 1))
561 *flags |= EXC_FLAG_ENDSWITH;
562 *pattern = p;
563 *patternlen = len;
564}
565
453ec4bd 566void add_exclude(const char *string, const char *base,
c04318e4 567 int baselen, struct exclude_list *el, int srcpos)
453ec4bd 568{
d6b8fc30 569 struct exclude *x;
84460eec 570 int patternlen;
f8708998 571 unsigned flags;
84460eec 572 int nowildcardlen;
453ec4bd 573
84460eec
NTND
574 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
575 if (flags & EXC_FLAG_MUSTBEDIR) {
96ffc06f 576 FLEXPTR_ALLOC_MEM(x, pattern, string, patternlen);
d6b8fc30
JH
577 } else {
578 x = xmalloc(sizeof(*x));
579 x->pattern = string;
580 }
84460eec
NTND
581 x->patternlen = patternlen;
582 x->nowildcardlen = nowildcardlen;
453ec4bd
LT
583 x->base = base;
584 x->baselen = baselen;
d6b8fc30 585 x->flags = flags;
c04318e4 586 x->srcpos = srcpos;
840fc334
AS
587 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
588 el->excludes[el->nr++] = x;
c04318e4 589 x->el = el;
453ec4bd
LT
590}
591
6f52b741
BW
592static void *read_skip_worktree_file_from_index(const struct index_state *istate,
593 const char *path, size_t *size,
55fe6f51 594 struct sha1_stat *sha1_stat)
c28b3d6e
NTND
595{
596 int pos, len;
597 unsigned long sz;
598 enum object_type type;
599 void *data;
c28b3d6e
NTND
600
601 len = strlen(path);
6f52b741 602 pos = index_name_pos(istate, path, len);
c28b3d6e
NTND
603 if (pos < 0)
604 return NULL;
6f52b741 605 if (!ce_skip_worktree(istate->cache[pos]))
c28b3d6e 606 return NULL;
6f52b741 607 data = read_sha1_file(istate->cache[pos]->oid.hash, &type, &sz);
c28b3d6e
NTND
608 if (!data || type != OBJ_BLOB) {
609 free(data);
610 return NULL;
611 }
612 *size = xsize_t(sz);
55fe6f51
NTND
613 if (sha1_stat) {
614 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
6f52b741 615 hashcpy(sha1_stat->sha1, istate->cache[pos]->oid.hash);
55fe6f51 616 }
c28b3d6e
NTND
617 return data;
618}
619
f6198812
AS
620/*
621 * Frees memory within el which was allocated for exclude patterns and
622 * the file buffer. Does not free el itself.
623 */
624void clear_exclude_list(struct exclude_list *el)
0fd0e241
NTND
625{
626 int i;
627
5cee3493 628 for (i = 0; i < el->nr; i++)
0fd0e241
NTND
629 free(el->excludes[i]);
630 free(el->excludes);
c082df24 631 free(el->filebuf);
0fd0e241 632
2653a8c6 633 memset(el, 0, sizeof(*el));
0fd0e241
NTND
634}
635
7e2e4b37 636static void trim_trailing_spaces(char *buf)
16402b99 637{
e61a6c1d
PB
638 char *p, *last_space = NULL;
639
640 for (p = buf; *p; p++)
641 switch (*p) {
642 case ' ':
643 if (!last_space)
644 last_space = p;
645 break;
646 case '\\':
647 p++;
648 if (!*p)
649 return;
650 /* fallthrough */
651 default:
652 last_space = NULL;
653 }
654
655 if (last_space)
656 *last_space = '\0';
16402b99
NTND
657}
658
0dcb8d7f
NTND
659/*
660 * Given a subdirectory name and "dir" of the current directory,
661 * search the subdir in "dir" and return it, or create a new one if it
662 * does not exist in "dir".
663 *
664 * If "name" has the trailing slash, it'll be excluded in the search.
665 */
666static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
667 struct untracked_cache_dir *dir,
668 const char *name, int len)
669{
670 int first, last;
671 struct untracked_cache_dir *d;
672 if (!dir)
673 return NULL;
674 if (len && name[len - 1] == '/')
675 len--;
676 first = 0;
677 last = dir->dirs_nr;
678 while (last > first) {
679 int cmp, next = (last + first) >> 1;
680 d = dir->dirs[next];
681 cmp = strncmp(name, d->name, len);
682 if (!cmp && strlen(d->name) > len)
683 cmp = -1;
684 if (!cmp)
685 return d;
686 if (cmp < 0) {
687 last = next;
688 continue;
689 }
690 first = next+1;
691 }
692
693 uc->dir_created++;
96ffc06f 694 FLEX_ALLOC_MEM(d, name, name, len);
0dcb8d7f
NTND
695
696 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
697 memmove(dir->dirs + first + 1, dir->dirs + first,
698 (dir->dirs_nr - first) * sizeof(*dir->dirs));
699 dir->dirs_nr++;
700 dir->dirs[first] = d;
701 return d;
702}
703
ccad261f
NTND
704static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
705{
706 int i;
707 dir->valid = 0;
708 dir->untracked_nr = 0;
709 for (i = 0; i < dir->dirs_nr; i++)
710 do_invalidate_gitignore(dir->dirs[i]);
711}
712
713static void invalidate_gitignore(struct untracked_cache *uc,
714 struct untracked_cache_dir *dir)
715{
716 uc->gitignore_invalidated++;
717 do_invalidate_gitignore(dir);
718}
719
91a2288b
NTND
720static void invalidate_directory(struct untracked_cache *uc,
721 struct untracked_cache_dir *dir)
722{
26cb0182 723 int i;
91a2288b
NTND
724 uc->dir_invalidated++;
725 dir->valid = 0;
726 dir->untracked_nr = 0;
26cb0182
NTND
727 for (i = 0; i < dir->dirs_nr; i++)
728 dir->dirs[i]->recurse = 0;
91a2288b
NTND
729}
730
55fe6f51
NTND
731/*
732 * Given a file with name "fname", read it (either from disk, or from
733 * the index if "check_index" is non-zero), parse it and store the
734 * exclude rules in "el".
735 *
736 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
737 * stat data from disk (only valid if add_excludes returns zero). If
738 * ss_valid is non-zero, "ss" must contain good value as input.
739 */
740static int add_excludes(const char *fname, const char *base, int baselen,
741 struct exclude_list *el, int check_index,
742 struct sha1_stat *sha1_stat)
453ec4bd 743{
c470701a 744 struct stat st;
c04318e4 745 int fd, i, lineno = 1;
9d14017a 746 size_t size = 0;
453ec4bd
LT
747 char *buf, *entry;
748
749 fd = open(fname, O_RDONLY);
c28b3d6e 750 if (fd < 0 || fstat(fd, &st) < 0) {
69660731 751 if (errno != ENOENT)
55b38a48 752 warn_on_inaccessible(fname);
c28b3d6e
NTND
753 if (0 <= fd)
754 close(fd);
755 if (!check_index ||
6f52b741 756 (buf = read_skip_worktree_file_from_index(&the_index, fname, &size, sha1_stat)) == NULL)
c28b3d6e 757 return -1;
45d76f17
NTND
758 if (size == 0) {
759 free(buf);
760 return 0;
761 }
762 if (buf[size-1] != '\n') {
50a6c8ef 763 buf = xrealloc(buf, st_add(size, 1));
45d76f17
NTND
764 buf[size++] = '\n';
765 }
d961baa8 766 } else {
c28b3d6e
NTND
767 size = xsize_t(st.st_size);
768 if (size == 0) {
55fe6f51
NTND
769 if (sha1_stat) {
770 fill_stat_data(&sha1_stat->stat, &st);
771 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
772 sha1_stat->valid = 1;
773 }
c28b3d6e
NTND
774 close(fd);
775 return 0;
776 }
3733e694 777 buf = xmallocz(size);
c28b3d6e 778 if (read_in_full(fd, buf, size) != size) {
45d76f17 779 free(buf);
c28b3d6e
NTND
780 close(fd);
781 return -1;
782 }
45d76f17 783 buf[size++] = '\n';
c28b3d6e 784 close(fd);
55fe6f51
NTND
785 if (sha1_stat) {
786 int pos;
787 if (sha1_stat->valid &&
ed4efab1 788 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st))
55fe6f51
NTND
789 ; /* no content change, ss->sha1 still good */
790 else if (check_index &&
12cd0bf9
BW
791 (pos = index_name_pos(&the_index, fname, strlen(fname))) >= 0 &&
792 !ce_stage(the_index.cache[pos]) &&
793 ce_uptodate(the_index.cache[pos]) &&
55fe6f51 794 !would_convert_to_git(fname))
99d1a986 795 hashcpy(sha1_stat->sha1,
12cd0bf9 796 the_index.cache[pos]->oid.hash);
55fe6f51
NTND
797 else
798 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
799 fill_stat_data(&sha1_stat->stat, &st);
800 sha1_stat->valid = 1;
801 }
6ba78238 802 }
453ec4bd 803
c082df24 804 el->filebuf = buf;
245e1c19 805
dde843e7
JH
806 if (skip_utf8_bom(&buf, size))
807 size -= buf - el->filebuf;
808
453ec4bd 809 entry = buf;
245e1c19 810
45d76f17
NTND
811 for (i = 0; i < size; i++) {
812 if (buf[i] == '\n') {
453ec4bd
LT
813 if (entry != buf + i && entry[0] != '#') {
814 buf[i - (i && buf[i-1] == '\r')] = 0;
7e2e4b37 815 trim_trailing_spaces(entry);
c04318e4 816 add_exclude(entry, base, baselen, el, lineno);
453ec4bd 817 }
c04318e4 818 lineno++;
453ec4bd
LT
819 entry = buf + i + 1;
820 }
821 }
822 return 0;
453ec4bd
LT
823}
824
55fe6f51
NTND
825int add_excludes_from_file_to_list(const char *fname, const char *base,
826 int baselen, struct exclude_list *el,
827 int check_index)
828{
829 return add_excludes(fname, base, baselen, el, check_index, NULL);
830}
831
c04318e4
AS
832struct exclude_list *add_exclude_list(struct dir_struct *dir,
833 int group_type, const char *src)
c082df24
AS
834{
835 struct exclude_list *el;
836 struct exclude_list_group *group;
837
838 group = &dir->exclude_list_group[group_type];
839 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
840 el = &group->el[group->nr++];
841 memset(el, 0, sizeof(*el));
c04318e4 842 el->src = src;
c082df24
AS
843 return el;
844}
845
846/*
847 * Used to set up core.excludesfile and .git/info/exclude lists.
848 */
0dcb8d7f
NTND
849static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
850 struct sha1_stat *sha1_stat)
453ec4bd 851{
c082df24 852 struct exclude_list *el;
ccad261f
NTND
853 /*
854 * catch setup_standard_excludes() that's called before
855 * dir->untracked is assigned. That function behaves
856 * differently when dir->untracked is non-NULL.
857 */
858 if (!dir->untracked)
859 dir->unmanaged_exclude_files++;
c04318e4 860 el = add_exclude_list(dir, EXC_FILE, fname);
0dcb8d7f 861 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0)
453ec4bd
LT
862 die("cannot use %s as an exclude file", fname);
863}
864
0dcb8d7f
NTND
865void add_excludes_from_file(struct dir_struct *dir, const char *fname)
866{
ccad261f 867 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
0dcb8d7f
NTND
868 add_excludes_from_file_1(dir, fname, NULL);
869}
870
82dce998
NTND
871int match_basename(const char *basename, int basenamelen,
872 const char *pattern, int prefix, int patternlen,
f8708998 873 unsigned flags)
593cb880
NTND
874{
875 if (prefix == patternlen) {
0b6e56df 876 if (patternlen == basenamelen &&
ba0897e6 877 !fspathncmp(pattern, basename, basenamelen))
593cb880
NTND
878 return 1;
879 } else if (flags & EXC_FLAG_ENDSWITH) {
0b6e56df 880 /* "*literal" matching against "fooliteral" */
593cb880 881 if (patternlen - 1 <= basenamelen &&
ba0897e6 882 !fspathncmp(pattern + 1,
0b6e56df
JH
883 basename + basenamelen - (patternlen - 1),
884 patternlen - 1))
593cb880
NTND
885 return 1;
886 } else {
0b6e56df
JH
887 if (fnmatch_icase_mem(pattern, patternlen,
888 basename, basenamelen,
889 0) == 0)
593cb880
NTND
890 return 1;
891 }
892 return 0;
893}
894
82dce998
NTND
895int match_pathname(const char *pathname, int pathlen,
896 const char *base, int baselen,
897 const char *pattern, int prefix, int patternlen,
f8708998 898 unsigned flags)
b5592632
NTND
899{
900 const char *name;
901 int namelen;
902
903 /*
904 * match with FNM_PATHNAME; the pattern has base implicitly
905 * in front of it.
906 */
907 if (*pattern == '/') {
908 pattern++;
982ac873 909 patternlen--;
b5592632
NTND
910 prefix--;
911 }
912
913 /*
914 * baselen does not count the trailing slash. base[] may or
915 * may not end with a trailing slash though.
916 */
917 if (pathlen < baselen + 1 ||
918 (baselen && pathname[baselen] != '/') ||
ba0897e6 919 fspathncmp(pathname, base, baselen))
b5592632
NTND
920 return 0;
921
922 namelen = baselen ? pathlen - baselen - 1 : pathlen;
923 name = pathname + pathlen - namelen;
924
925 if (prefix) {
926 /*
927 * if the non-wildcard part is longer than the
928 * remaining pathname, surely it cannot match.
929 */
930 if (prefix > namelen)
931 return 0;
932
ba0897e6 933 if (fspathncmp(pattern, name, prefix))
b5592632
NTND
934 return 0;
935 pattern += prefix;
ab3aebc1 936 patternlen -= prefix;
b5592632
NTND
937 name += prefix;
938 namelen -= prefix;
ab3aebc1
JK
939
940 /*
941 * If the whole pattern did not have a wildcard,
942 * then our prefix match is all we need; we
943 * do not need to call fnmatch at all.
944 */
5cee3493 945 if (!patternlen && !namelen)
ab3aebc1 946 return 1;
b5592632
NTND
947 }
948
ab3aebc1
JK
949 return fnmatch_icase_mem(pattern, patternlen,
950 name, namelen,
f30366b2 951 WM_PATHNAME) == 0;
b5592632
NTND
952}
953
578cd7c3
AS
954/*
955 * Scan the given exclude list in reverse to see whether pathname
956 * should be ignored. The first match (i.e. the last on the list), if
957 * any, determines the fate. Returns the exclude_list element which
958 * matched, or NULL for undecided.
453ec4bd 959 */
578cd7c3
AS
960static struct exclude *last_exclude_matching_from_list(const char *pathname,
961 int pathlen,
962 const char *basename,
963 int *dtype,
2b70e88d
BW
964 struct exclude_list *el,
965 struct index_state *istate)
453ec4bd 966{
e6efecc4 967 struct exclude *exc = NULL; /* undecided */
5cee3493 968 int i;
453ec4bd 969
35a94d44 970 if (!el->nr)
578cd7c3 971 return NULL; /* undefined */
d6b8fc30 972
35a94d44
NTND
973 for (i = el->nr - 1; 0 <= i; i--) {
974 struct exclude *x = el->excludes[i];
b5592632 975 const char *exclude = x->pattern;
b5592632 976 int prefix = x->nowildcardlen;
35a94d44
NTND
977
978 if (x->flags & EXC_FLAG_MUSTBEDIR) {
979 if (*dtype == DT_UNKNOWN)
2b70e88d 980 *dtype = get_dtype(NULL, istate, pathname, pathlen);
35a94d44
NTND
981 if (*dtype != DT_DIR)
982 continue;
983 }
d6b8fc30 984
35a94d44 985 if (x->flags & EXC_FLAG_NODIR) {
593cb880
NTND
986 if (match_basename(basename,
987 pathlen - (basename - pathname),
988 exclude, prefix, x->patternlen,
e6efecc4
NTND
989 x->flags)) {
990 exc = x;
991 break;
992 }
35a94d44
NTND
993 continue;
994 }
995
b5592632
NTND
996 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
997 if (match_pathname(pathname, pathlen,
998 x->base, x->baselen ? x->baselen - 1 : 0,
e6efecc4
NTND
999 exclude, prefix, x->patternlen, x->flags)) {
1000 exc = x;
1001 break;
1002 }
453ec4bd 1003 }
e6efecc4 1004 return exc;
578cd7c3
AS
1005}
1006
1007/*
1008 * Scan the list and let the last match determine the fate.
1009 * Return 1 for exclude, 0 for include and -1 for undecided.
1010 */
1011int is_excluded_from_list(const char *pathname,
1012 int pathlen, const char *basename, int *dtype,
1013 struct exclude_list *el)
1014{
1015 struct exclude *exclude;
2b70e88d 1016 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el, &the_index);
578cd7c3
AS
1017 if (exclude)
1018 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
453ec4bd
LT
1019 return -1; /* undecided */
1020}
1021
46aa2f95 1022static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
2b70e88d 1023 struct index_state *istate,
46aa2f95
KB
1024 const char *pathname, int pathlen, const char *basename,
1025 int *dtype_p)
1026{
1027 int i, j;
1028 struct exclude_list_group *group;
1029 struct exclude *exclude;
1030 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1031 group = &dir->exclude_list_group[i];
1032 for (j = group->nr - 1; j >= 0; j--) {
1033 exclude = last_exclude_matching_from_list(
1034 pathname, pathlen, basename, dtype_p,
2b70e88d 1035 &group->el[j], istate);
46aa2f95
KB
1036 if (exclude)
1037 return exclude;
1038 }
1039 }
1040 return NULL;
1041}
1042
6cd5c582
KB
1043/*
1044 * Loads the per-directory exclude list for the substring of base
1045 * which has a char length of baselen.
1046 */
1047static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
1048{
1049 struct exclude_list_group *group;
1050 struct exclude_list *el;
1051 struct exclude_stack *stk = NULL;
0dcb8d7f 1052 struct untracked_cache_dir *untracked;
6cd5c582
KB
1053 int current;
1054
6cd5c582
KB
1055 group = &dir->exclude_list_group[EXC_DIRS];
1056
d961baa8
NTND
1057 /*
1058 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
6cd5c582 1059 * which originate from directories not in the prefix of the
d961baa8
NTND
1060 * path being checked.
1061 */
6cd5c582
KB
1062 while ((stk = dir->exclude_stack) != NULL) {
1063 if (stk->baselen <= baselen &&
aceb9429 1064 !strncmp(dir->basebuf.buf, base, stk->baselen))
6cd5c582
KB
1065 break;
1066 el = &group->el[dir->exclude_stack->exclude_ix];
1067 dir->exclude_stack = stk->prev;
95c6f271 1068 dir->exclude = NULL;
aceb9429 1069 free((char *)el->src); /* see strbuf_detach() below */
6cd5c582
KB
1070 clear_exclude_list(el);
1071 free(stk);
1072 group->nr--;
1073 }
1074
95c6f271
KB
1075 /* Skip traversing into sub directories if the parent is excluded */
1076 if (dir->exclude)
1077 return;
1078
aceb9429
NTND
1079 /*
1080 * Lazy initialization. All call sites currently just
1081 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1082 * them seems lots of work for little benefit.
1083 */
1084 if (!dir->basebuf.buf)
1085 strbuf_init(&dir->basebuf, PATH_MAX);
1086
6cd5c582
KB
1087 /* Read from the parent directories and push them down. */
1088 current = stk ? stk->baselen : -1;
aceb9429 1089 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
0dcb8d7f
NTND
1090 if (dir->untracked)
1091 untracked = stk ? stk->ucd : dir->untracked->root;
1092 else
1093 untracked = NULL;
1094
6cd5c582 1095 while (current < baselen) {
6cd5c582 1096 const char *cp;
0dcb8d7f 1097 struct sha1_stat sha1_stat;
6cd5c582 1098
03e11a71 1099 stk = xcalloc(1, sizeof(*stk));
6cd5c582
KB
1100 if (current < 0) {
1101 cp = base;
1102 current = 0;
d961baa8 1103 } else {
6cd5c582
KB
1104 cp = strchr(base + current + 1, '/');
1105 if (!cp)
1106 die("oops in prep_exclude");
1107 cp++;
0dcb8d7f
NTND
1108 untracked =
1109 lookup_untracked(dir->untracked, untracked,
1110 base + current,
1111 cp - base - current);
6cd5c582
KB
1112 }
1113 stk->prev = dir->exclude_stack;
1114 stk->baselen = cp - base;
95c6f271 1115 stk->exclude_ix = group->nr;
0dcb8d7f 1116 stk->ucd = untracked;
95c6f271 1117 el = add_exclude_list(dir, EXC_DIRS, NULL);
aceb9429
NTND
1118 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1119 assert(stk->baselen == dir->basebuf.len);
95c6f271
KB
1120
1121 /* Abort if the directory is excluded */
1122 if (stk->baselen) {
1123 int dt = DT_DIR;
aceb9429 1124 dir->basebuf.buf[stk->baselen - 1] = 0;
95c6f271 1125 dir->exclude = last_exclude_matching_from_lists(dir,
2b70e88d 1126 &the_index,
aceb9429
NTND
1127 dir->basebuf.buf, stk->baselen - 1,
1128 dir->basebuf.buf + current, &dt);
1129 dir->basebuf.buf[stk->baselen - 1] = '/';
c3c327de
KB
1130 if (dir->exclude &&
1131 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1132 dir->exclude = NULL;
95c6f271 1133 if (dir->exclude) {
95c6f271
KB
1134 dir->exclude_stack = stk;
1135 return;
1136 }
1137 }
1138
aceb9429 1139 /* Try to read per-directory file */
0dcb8d7f
NTND
1140 hashclr(sha1_stat.sha1);
1141 sha1_stat.valid = 0;
27b099ae
NTND
1142 if (dir->exclude_per_dir &&
1143 /*
1144 * If we know that no files have been added in
1145 * this directory (i.e. valid_cached_dir() has
1146 * been executed and set untracked->valid) ..
1147 */
1148 (!untracked || !untracked->valid ||
1149 /*
1150 * .. and .gitignore does not exist before
7687252f
DT
1151 * (i.e. null exclude_sha1). Then we can skip
1152 * loading .gitignore, which would result in
1153 * ENOENT anyway.
27b099ae
NTND
1154 */
1155 !is_null_sha1(untracked->exclude_sha1))) {
95c6f271
KB
1156 /*
1157 * dir->basebuf gets reused by the traversal, but we
1158 * need fname to remain unchanged to ensure the src
1159 * member of each struct exclude correctly
1160 * back-references its source file. Other invocations
1161 * of add_exclude_list provide stable strings, so we
aceb9429 1162 * strbuf_detach() and free() here in the caller.
95c6f271 1163 */
aceb9429
NTND
1164 struct strbuf sb = STRBUF_INIT;
1165 strbuf_addbuf(&sb, &dir->basebuf);
1166 strbuf_addstr(&sb, dir->exclude_per_dir);
1167 el->src = strbuf_detach(&sb, NULL);
0dcb8d7f
NTND
1168 add_excludes(el->src, el->src, stk->baselen, el, 1,
1169 untracked ? &sha1_stat : NULL);
1170 }
5ebf79ad
NTND
1171 /*
1172 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1173 * will first be called in valid_cached_dir() then maybe many
1174 * times more in last_exclude_matching(). When the cache is
1175 * used, last_exclude_matching() will not be called and
1176 * reading .gitignore content will be a waste.
1177 *
1178 * So when it's called by valid_cached_dir() and we can get
1179 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1180 * modified on work tree), we could delay reading the
1181 * .gitignore content until we absolutely need it in
1182 * last_exclude_matching(). Be careful about ignore rule
1183 * order, though, if you do that.
1184 */
1185 if (untracked &&
1186 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1187 invalidate_gitignore(dir->untracked, untracked);
0dcb8d7f 1188 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
95c6f271 1189 }
6cd5c582
KB
1190 dir->exclude_stack = stk;
1191 current = stk->baselen;
1192 }
aceb9429 1193 strbuf_setlen(&dir->basebuf, baselen);
6cd5c582
KB
1194}
1195
f4cd69a6
AS
1196/*
1197 * Loads the exclude lists for the directory containing pathname, then
1198 * scans all exclude lists to determine whether pathname is excluded.
1199 * Returns the exclude_list element which matched, or NULL for
1200 * undecided.
1201 */
b07bc8c8 1202struct exclude *last_exclude_matching(struct dir_struct *dir,
f4cd69a6
AS
1203 const char *pathname,
1204 int *dtype_p)
453ec4bd
LT
1205{
1206 int pathlen = strlen(pathname);
68492fc7
LK
1207 const char *basename = strrchr(pathname, '/');
1208 basename = (basename) ? basename+1 : pathname;
453ec4bd 1209
63d285c8 1210 prep_exclude(dir, pathname, basename-pathname);
c082df24 1211
95c6f271
KB
1212 if (dir->exclude)
1213 return dir->exclude;
1214
2b70e88d 1215 return last_exclude_matching_from_lists(dir, &the_index, pathname, pathlen,
46aa2f95 1216 basename, dtype_p);
f4cd69a6
AS
1217}
1218
1219/*
1220 * Loads the exclude lists for the directory containing pathname, then
1221 * scans all exclude lists to determine whether pathname is excluded.
1222 * Returns 1 if true, otherwise 0.
1223 */
b07bc8c8 1224int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
f4cd69a6
AS
1225{
1226 struct exclude *exclude =
1227 last_exclude_matching(dir, pathname, dtype_p);
1228 if (exclude)
1229 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
453ec4bd
LT
1230 return 0;
1231}
1232
f3fa1838
JH
1233static struct dir_entry *dir_entry_new(const char *pathname, int len)
1234{
453ec4bd
LT
1235 struct dir_entry *ent;
1236
96ffc06f 1237 FLEX_ALLOC_MEM(ent, name, pathname, len);
453ec4bd 1238 ent->len = len;
4d06f8ac 1239 return ent;
453ec4bd
LT
1240}
1241
9e58beca
BW
1242static struct dir_entry *dir_add_name(struct dir_struct *dir,
1243 struct index_state *istate,
1244 const char *pathname, int len)
6815e569 1245{
9e58beca 1246 if (index_file_exists(istate, pathname, len, ignore_case))
6815e569
JK
1247 return NULL;
1248
25fd2f7a 1249 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
6815e569
JK
1250 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1251}
1252
9e58beca
BW
1253struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1254 struct index_state *istate,
1255 const char *pathname, int len)
2abd31b0 1256{
9e58beca 1257 if (!index_name_is_other(istate, pathname, len))
2abd31b0
JK
1258 return NULL;
1259
25fd2f7a 1260 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
2abd31b0
JK
1261 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1262}
1263
09595258
LT
1264enum exist_status {
1265 index_nonexistent = 0,
1266 index_directory,
4b05548f 1267 index_gitdir
09595258
LT
1268};
1269
5102c617 1270/*
71261027 1271 * Do not use the alphabetically sorted index to look up
5102c617 1272 * the directory name; instead, use the case insensitive
ebbd7439 1273 * directory hash.
5102c617 1274 */
ae520e36
BW
1275static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1276 const char *dirname, int len)
5102c617 1277{
41284eb0 1278 struct cache_entry *ce;
5102c617 1279
ae520e36 1280 if (index_dir_exists(istate, dirname, len))
5102c617
JJ
1281 return index_directory;
1282
ae520e36 1283 ce = index_file_exists(istate, dirname, len, ignore_case);
41284eb0 1284 if (ce && S_ISGITLINK(ce->ce_mode))
5102c617
JJ
1285 return index_gitdir;
1286
5102c617
JJ
1287 return index_nonexistent;
1288}
1289
09595258
LT
1290/*
1291 * The index sorts alphabetically by entry name, which
1292 * means that a gitlink sorts as '\0' at the end, while
1293 * a directory (which is defined not as an entry, but as
1294 * the files it contains) will sort with the '/' at the
1295 * end.
1296 */
ae520e36
BW
1297static enum exist_status directory_exists_in_index(struct index_state *istate,
1298 const char *dirname, int len)
453ec4bd 1299{
5102c617
JJ
1300 int pos;
1301
1302 if (ignore_case)
ae520e36 1303 return directory_exists_in_index_icase(istate, dirname, len);
5102c617 1304
ae520e36 1305 pos = index_name_pos(istate, dirname, len);
09595258
LT
1306 if (pos < 0)
1307 pos = -pos-1;
ae520e36
BW
1308 while (pos < istate->cache_nr) {
1309 const struct cache_entry *ce = istate->cache[pos++];
09595258
LT
1310 unsigned char endchar;
1311
1312 if (strncmp(ce->name, dirname, len))
1313 break;
1314 endchar = ce->name[len];
1315 if (endchar > '/')
1316 break;
1317 if (endchar == '/')
1318 return index_directory;
7a51ed66 1319 if (!endchar && S_ISGITLINK(ce->ce_mode))
09595258
LT
1320 return index_gitdir;
1321 }
1322 return index_nonexistent;
1323}
1324
1325/*
1326 * When we find a directory when traversing the filesystem, we
1327 * have three distinct cases:
1328 *
1329 * - ignore it
1330 * - see it as a directory
1331 * - recurse into it
1332 *
1333 * and which one we choose depends on a combination of existing
1334 * git index contents and the flags passed into the directory
1335 * traversal routine.
1336 *
1337 * Case 1: If we *already* have entries in the index under that
5bd8e2d8
KB
1338 * directory name, we always recurse into the directory to see
1339 * all the files.
09595258
LT
1340 *
1341 * Case 2: If we *already* have that directory name as a gitlink,
1342 * we always continue to see it as a gitlink, regardless of whether
1343 * there is an actual git directory there or not (it might not
1344 * be checked out as a subproject!)
1345 *
1346 * Case 3: if we didn't have it in the index previously, we
1347 * have a few sub-cases:
1348 *
1349 * (a) if "show_other_directories" is true, we show it as
1350 * just a directory, unless "hide_empty_directories" is
defd7c7b
KB
1351 * also true, in which case we need to check if it contains any
1352 * untracked and / or ignored files.
09595258 1353 * (b) if it looks like a git directory, and we don't have
302b9282 1354 * 'no_gitlinks' set we treat it as a gitlink, and show it
09595258
LT
1355 * as a directory.
1356 * (c) otherwise, we recurse into it.
1357 */
defd7c7b 1358static enum path_treatment treat_directory(struct dir_struct *dir,
0dcb8d7f 1359 struct untracked_cache_dir *untracked,
2e5910f2 1360 const char *dirname, int len, int baselen, int exclude,
e1b8c7bd 1361 const struct pathspec *pathspec)
09595258
LT
1362{
1363 /* The "len-1" is to strip the final '/' */
ae520e36 1364 switch (directory_exists_in_index(&the_index, dirname, len-1)) {
09595258 1365 case index_directory:
defd7c7b 1366 return path_recurse;
09595258
LT
1367
1368 case index_gitdir:
26c986e1 1369 return path_none;
09595258
LT
1370
1371 case index_nonexistent:
7c4c97c0 1372 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
09595258 1373 break;
7c4c97c0 1374 if (!(dir->flags & DIR_NO_GITLINKS)) {
09595258
LT
1375 unsigned char sha1[20];
1376 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
defd7c7b 1377 return path_untracked;
09595258 1378 }
defd7c7b 1379 return path_recurse;
09595258
LT
1380 }
1381
1382 /* This is the "show_other_directories" case */
721ac4ed 1383
184d2a8e 1384 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
defd7c7b
KB
1385 return exclude ? path_excluded : path_untracked;
1386
2e5910f2
DT
1387 untracked = lookup_untracked(dir->untracked, untracked,
1388 dirname + baselen, len - baselen);
0dcb8d7f 1389 return read_directory_recursive(dir, dirname, len,
e1b8c7bd 1390 untracked, 1, pathspec);
453ec4bd
LT
1391}
1392
9fc42d60
LT
1393/*
1394 * This is an inexact early pruning of any recursive directory
1395 * reading - if the path cannot possibly be in the pathspec,
1396 * return true, and we'll skip it early.
1397 */
e1b8c7bd
BW
1398static int simplify_away(const char *path, int pathlen,
1399 const struct pathspec *pathspec)
9fc42d60 1400{
e1b8c7bd 1401 int i;
9fc42d60 1402
e1b8c7bd
BW
1403 if (!pathspec || !pathspec->nr)
1404 return 0;
1405
1406 GUARD_PATHSPEC(pathspec,
1407 PATHSPEC_FROMTOP |
1408 PATHSPEC_MAXDEPTH |
1409 PATHSPEC_LITERAL |
1410 PATHSPEC_GLOB |
1411 PATHSPEC_ICASE |
b0db7046
BW
1412 PATHSPEC_EXCLUDE |
1413 PATHSPEC_ATTR);
e1b8c7bd
BW
1414
1415 for (i = 0; i < pathspec->nr; i++) {
1416 const struct pathspec_item *item = &pathspec->items[i];
1417 int len = item->nowildcard_len;
1418
1419 if (len > pathlen)
1420 len = pathlen;
1421 if (!ps_strncmp(item, item->match, path, len))
1422 return 0;
9fc42d60 1423 }
e1b8c7bd
BW
1424
1425 return 1;
9fc42d60
LT
1426}
1427
29209cbe
JK
1428/*
1429 * This function tells us whether an excluded path matches a
1430 * list of "interesting" pathspecs. That is, whether a path matched
1431 * by any of the pathspecs could possibly be ignored by excluding
1432 * the specified path. This can happen if:
1433 *
1434 * 1. the path is mentioned explicitly in the pathspec
1435 *
1436 * 2. the path is a directory prefix of some element in the
1437 * pathspec
1438 */
e1b8c7bd
BW
1439static int exclude_matches_pathspec(const char *path, int pathlen,
1440 const struct pathspec *pathspec)
1441{
1442 int i;
1443
1444 if (!pathspec || !pathspec->nr)
1445 return 0;
1446
1447 GUARD_PATHSPEC(pathspec,
1448 PATHSPEC_FROMTOP |
1449 PATHSPEC_MAXDEPTH |
1450 PATHSPEC_LITERAL |
1451 PATHSPEC_GLOB |
1452 PATHSPEC_ICASE |
1453 PATHSPEC_EXCLUDE);
1454
1455 for (i = 0; i < pathspec->nr; i++) {
1456 const struct pathspec_item *item = &pathspec->items[i];
1457 int len = item->nowildcard_len;
1458
1459 if (len == pathlen &&
1460 !ps_strncmp(item, item->match, path, pathlen))
1461 return 1;
1462 if (len > pathlen &&
1463 item->match[pathlen] == '/' &&
1464 !ps_strncmp(item, item->match, path, pathlen))
1465 return 1;
e96980ef
JK
1466 }
1467 return 0;
1468}
1469
98f2a687
BW
1470static int get_index_dtype(struct index_state *istate,
1471 const char *path, int len)
443e061a
LT
1472{
1473 int pos;
9c5e6c80 1474 const struct cache_entry *ce;
443e061a 1475
98f2a687 1476 ce = index_file_exists(istate, path, len, 0);
443e061a
LT
1477 if (ce) {
1478 if (!ce_uptodate(ce))
1479 return DT_UNKNOWN;
1480 if (S_ISGITLINK(ce->ce_mode))
1481 return DT_DIR;
1482 /*
1483 * Nobody actually cares about the
1484 * difference between DT_LNK and DT_REG
1485 */
1486 return DT_REG;
1487 }
1488
1489 /* Try to look it up as a directory */
98f2a687 1490 pos = index_name_pos(istate, path, len);
443e061a
LT
1491 if (pos >= 0)
1492 return DT_UNKNOWN;
1493 pos = -pos-1;
98f2a687
BW
1494 while (pos < istate->cache_nr) {
1495 ce = istate->cache[pos++];
443e061a
LT
1496 if (strncmp(ce->name, path, len))
1497 break;
1498 if (ce->name[len] > '/')
1499 break;
1500 if (ce->name[len] < '/')
1501 continue;
1502 if (!ce_uptodate(ce))
1503 break; /* continue? */
1504 return DT_DIR;
1505 }
1506 return DT_UNKNOWN;
1507}
1508
98f2a687
BW
1509static int get_dtype(struct dirent *de, struct index_state *istate,
1510 const char *path, int len)
07134421 1511{
6831a88a 1512 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
07134421
LT
1513 struct stat st;
1514
1515 if (dtype != DT_UNKNOWN)
1516 return dtype;
98f2a687 1517 dtype = get_index_dtype(istate, path, len);
443e061a
LT
1518 if (dtype != DT_UNKNOWN)
1519 return dtype;
1520 if (lstat(path, &st))
07134421
LT
1521 return dtype;
1522 if (S_ISREG(st.st_mode))
1523 return DT_REG;
1524 if (S_ISDIR(st.st_mode))
1525 return DT_DIR;
1526 if (S_ISLNK(st.st_mode))
1527 return DT_LNK;
1528 return dtype;
1529}
1530
16e2cfa9 1531static enum path_treatment treat_one_path(struct dir_struct *dir,
0dcb8d7f 1532 struct untracked_cache_dir *untracked,
49dc2cc2 1533 struct strbuf *path,
2e5910f2 1534 int baselen,
e1b8c7bd 1535 const struct pathspec *pathspec,
16e2cfa9 1536 int dtype, struct dirent *de)
53cc5356 1537{
8aaf8d77 1538 int exclude;
12cd0bf9 1539 int has_path_in_index = !!index_file_exists(&the_index, path->buf, path->len, ignore_case);
2eac2a4c 1540
8aaf8d77 1541 if (dtype == DT_UNKNOWN)
98f2a687 1542 dtype = get_dtype(de, &the_index, path->buf, path->len);
8aaf8d77
KB
1543
1544 /* Always exclude indexed files */
2eac2a4c 1545 if (dtype != DT_DIR && has_path_in_index)
defd7c7b 1546 return path_none;
8aaf8d77 1547
2eac2a4c
JH
1548 /*
1549 * When we are looking at a directory P in the working tree,
1550 * there are three cases:
1551 *
1552 * (1) P exists in the index. Everything inside the directory P in
1553 * the working tree needs to go when P is checked out from the
1554 * index.
1555 *
1556 * (2) P does not exist in the index, but there is P/Q in the index.
1557 * We know P will stay a directory when we check out the contents
1558 * of the index, but we do not know yet if there is a directory
1559 * P/Q in the working tree to be killed, so we need to recurse.
1560 *
1561 * (3) P does not exist in the index, and there is no P/Q in the index
1562 * to require P to be a directory, either. Only in this case, we
1563 * know that everything inside P will not be killed without
1564 * recursing.
1565 */
1566 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1567 (dtype == DT_DIR) &&
de372b1b 1568 !has_path_in_index &&
ae520e36 1569 (directory_exists_in_index(&the_index, path->buf, path->len) == index_nonexistent))
de372b1b 1570 return path_none;
8aaf8d77
KB
1571
1572 exclude = is_excluded(dir, path->buf, &dtype);
53cc5356
JH
1573
1574 /*
1575 * Excluded? If we don't explicitly want to show
1576 * ignored files, ignore it
1577 */
0aaf62b6 1578 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
defd7c7b 1579 return path_excluded;
53cc5356 1580
53cc5356
JH
1581 switch (dtype) {
1582 default:
defd7c7b 1583 return path_none;
53cc5356 1584 case DT_DIR:
49dc2cc2 1585 strbuf_addch(path, '/');
2e5910f2 1586 return treat_directory(dir, untracked, path->buf, path->len,
e1b8c7bd 1587 baselen, exclude, pathspec);
53cc5356
JH
1588 case DT_REG:
1589 case DT_LNK:
defd7c7b 1590 return exclude ? path_excluded : path_untracked;
53cc5356 1591 }
53cc5356
JH
1592}
1593
91a2288b
NTND
1594static enum path_treatment treat_path_fast(struct dir_struct *dir,
1595 struct untracked_cache_dir *untracked,
1596 struct cached_dir *cdir,
1597 struct strbuf *path,
1598 int baselen,
e1b8c7bd 1599 const struct pathspec *pathspec)
91a2288b
NTND
1600{
1601 strbuf_setlen(path, baselen);
1602 if (!cdir->ucd) {
1603 strbuf_addstr(path, cdir->file);
1604 return path_untracked;
1605 }
1606 strbuf_addstr(path, cdir->ucd->name);
1607 /* treat_one_path() does this before it calls treat_directory() */
00b6c178 1608 strbuf_complete(path, '/');
91a2288b
NTND
1609 if (cdir->ucd->check_only)
1610 /*
1611 * check_only is set as a result of treat_directory() getting
1612 * to its bottom. Verify again the same set of directories
1613 * with check_only set.
1614 */
1615 return read_directory_recursive(dir, path->buf, path->len,
e1b8c7bd 1616 cdir->ucd, 1, pathspec);
91a2288b
NTND
1617 /*
1618 * We get path_recurse in the first run when
1619 * directory_exists_in_index() returns index_nonexistent. We
1620 * are sure that new changes in the index does not impact the
1621 * outcome. Return now.
1622 */
1623 return path_recurse;
1624}
1625
16e2cfa9 1626static enum path_treatment treat_path(struct dir_struct *dir,
0dcb8d7f 1627 struct untracked_cache_dir *untracked,
cf7c6148 1628 struct cached_dir *cdir,
49dc2cc2 1629 struct strbuf *path,
16e2cfa9 1630 int baselen,
e1b8c7bd 1631 const struct pathspec *pathspec)
16e2cfa9
JH
1632{
1633 int dtype;
cf7c6148 1634 struct dirent *de = cdir->de;
16e2cfa9 1635
91a2288b
NTND
1636 if (!de)
1637 return treat_path_fast(dir, untracked, cdir, path,
e1b8c7bd 1638 baselen, pathspec);
16e2cfa9 1639 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
defd7c7b 1640 return path_none;
49dc2cc2
RS
1641 strbuf_setlen(path, baselen);
1642 strbuf_addstr(path, de->d_name);
e1b8c7bd 1643 if (simplify_away(path->buf, path->len, pathspec))
defd7c7b 1644 return path_none;
16e2cfa9
JH
1645
1646 dtype = DTYPE(de);
e1b8c7bd 1647 return treat_one_path(dir, untracked, path, baselen, pathspec, dtype, de);
0dcb8d7f
NTND
1648}
1649
1650static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1651{
1652 if (!dir)
1653 return;
1654 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1655 dir->untracked_alloc);
1656 dir->untracked[dir->untracked_nr++] = xstrdup(name);
16e2cfa9
JH
1657}
1658
91a2288b
NTND
1659static int valid_cached_dir(struct dir_struct *dir,
1660 struct untracked_cache_dir *untracked,
1661 struct strbuf *path,
1662 int check_only)
1663{
1664 struct stat st;
1665
1666 if (!untracked)
1667 return 0;
1668
1669 if (stat(path->len ? path->buf : ".", &st)) {
1670 invalidate_directory(dir->untracked, untracked);
1671 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1672 return 0;
1673 }
1674 if (!untracked->valid ||
ed4efab1 1675 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
91a2288b
NTND
1676 if (untracked->valid)
1677 invalidate_directory(dir->untracked, untracked);
1678 fill_stat_data(&untracked->stat_data, &st);
1679 return 0;
1680 }
1681
1682 if (untracked->check_only != !!check_only) {
1683 invalidate_directory(dir->untracked, untracked);
1684 return 0;
1685 }
1686
1687 /*
1688 * prep_exclude will be called eventually on this directory,
1689 * but it's called much later in last_exclude_matching(). We
1690 * need it now to determine the validity of the cache for this
1691 * path. The next calls will be nearly no-op, the way
1692 * prep_exclude() is designed.
1693 */
1694 if (path->len && path->buf[path->len - 1] != '/') {
1695 strbuf_addch(path, '/');
1696 prep_exclude(dir, path->buf, path->len);
1697 strbuf_setlen(path, path->len - 1);
1698 } else
1699 prep_exclude(dir, path->buf, path->len);
1700
1701 /* hopefully prep_exclude() haven't invalidated this entry... */
1702 return untracked->valid;
1703}
1704
cf7c6148
NTND
1705static int open_cached_dir(struct cached_dir *cdir,
1706 struct dir_struct *dir,
1707 struct untracked_cache_dir *untracked,
1708 struct strbuf *path,
1709 int check_only)
1710{
1711 memset(cdir, 0, sizeof(*cdir));
1712 cdir->untracked = untracked;
91a2288b
NTND
1713 if (valid_cached_dir(dir, untracked, path, check_only))
1714 return 0;
cf7c6148 1715 cdir->fdir = opendir(path->len ? path->buf : ".");
91a2288b
NTND
1716 if (dir->untracked)
1717 dir->untracked->dir_opened++;
cf7c6148
NTND
1718 if (!cdir->fdir)
1719 return -1;
1720 return 0;
1721}
1722
1723static int read_cached_dir(struct cached_dir *cdir)
1724{
1725 if (cdir->fdir) {
1726 cdir->de = readdir(cdir->fdir);
1727 if (!cdir->de)
1728 return -1;
1729 return 0;
1730 }
91a2288b
NTND
1731 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1732 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
26cb0182
NTND
1733 if (!d->recurse) {
1734 cdir->nr_dirs++;
1735 continue;
1736 }
91a2288b
NTND
1737 cdir->ucd = d;
1738 cdir->nr_dirs++;
1739 return 0;
1740 }
1741 cdir->ucd = NULL;
1742 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1743 struct untracked_cache_dir *d = cdir->untracked;
1744 cdir->file = d->untracked[cdir->nr_files++];
1745 return 0;
1746 }
cf7c6148
NTND
1747 return -1;
1748}
1749
1750static void close_cached_dir(struct cached_dir *cdir)
1751{
1752 if (cdir->fdir)
1753 closedir(cdir->fdir);
91a2288b
NTND
1754 /*
1755 * We have gone through this directory and found no untracked
1756 * entries. Mark it valid.
1757 */
26cb0182 1758 if (cdir->untracked) {
91a2288b 1759 cdir->untracked->valid = 1;
26cb0182
NTND
1760 cdir->untracked->recurse = 1;
1761 }
16e2cfa9
JH
1762}
1763
453ec4bd
LT
1764/*
1765 * Read a directory tree. We currently ignore anything but
1766 * directories, regular files and symlinks. That's because git
1767 * doesn't handle them at all yet. Maybe that will change some
1768 * day.
1769 *
1770 * Also, we ignore the name ".git" (even if it is not a directory).
1771 * That likely will not change.
defd7c7b
KB
1772 *
1773 * Returns the most significant path_treatment value encountered in the scan.
453ec4bd 1774 */
defd7c7b 1775static enum path_treatment read_directory_recursive(struct dir_struct *dir,
53cc5356 1776 const char *base, int baselen,
0dcb8d7f 1777 struct untracked_cache_dir *untracked, int check_only,
e1b8c7bd 1778 const struct pathspec *pathspec)
453ec4bd 1779{
cf7c6148 1780 struct cached_dir cdir;
defd7c7b 1781 enum path_treatment state, subdir_state, dir_state = path_none;
bef36921 1782 struct strbuf path = STRBUF_INIT;
453ec4bd 1783
bef36921 1784 strbuf_add(&path, base, baselen);
02cb6753 1785
cf7c6148 1786 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1528d247
RS
1787 goto out;
1788
0dcb8d7f
NTND
1789 if (untracked)
1790 untracked->check_only = !!check_only;
1791
cf7c6148 1792 while (!read_cached_dir(&cdir)) {
defd7c7b 1793 /* check how the file or directory should be treated */
e1b8c7bd
BW
1794 state = treat_path(dir, untracked, &cdir, &path,
1795 baselen, pathspec);
0dcb8d7f 1796
defd7c7b
KB
1797 if (state > dir_state)
1798 dir_state = state;
1799
1800 /* recurse into subdir if instructed by treat_path */
1801 if (state == path_recurse) {
0dcb8d7f
NTND
1802 struct untracked_cache_dir *ud;
1803 ud = lookup_untracked(dir->untracked, untracked,
1804 path.buf + baselen,
1805 path.len - baselen);
1806 subdir_state =
e1b8c7bd
BW
1807 read_directory_recursive(dir, path.buf,
1808 path.len, ud,
1809 check_only, pathspec);
defd7c7b
KB
1810 if (subdir_state > dir_state)
1811 dir_state = subdir_state;
1812 }
1813
1814 if (check_only) {
1815 /* abort early if maximum state has been reached */
0dcb8d7f 1816 if (dir_state == path_untracked) {
91a2288b 1817 if (cdir.fdir)
0dcb8d7f 1818 add_untracked(untracked, path.buf + baselen);
defd7c7b 1819 break;
0dcb8d7f 1820 }
defd7c7b 1821 /* skip the dir_add_* part */
02cb6753 1822 continue;
453ec4bd 1823 }
defd7c7b
KB
1824
1825 /* add the path to the appropriate result list */
1826 switch (state) {
1827 case path_excluded:
1828 if (dir->flags & DIR_SHOW_IGNORED)
9e58beca 1829 dir_add_name(dir, &the_index, path.buf, path.len);
0aaf62b6
KB
1830 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1831 ((dir->flags & DIR_COLLECT_IGNORED) &&
1832 exclude_matches_pathspec(path.buf, path.len,
e1b8c7bd 1833 pathspec)))
9e58beca 1834 dir_add_ignored(dir, &the_index, path.buf, path.len);
defd7c7b
KB
1835 break;
1836
1837 case path_untracked:
0dcb8d7f
NTND
1838 if (dir->flags & DIR_SHOW_IGNORED)
1839 break;
9e58beca 1840 dir_add_name(dir, &the_index, path.buf, path.len);
91a2288b 1841 if (cdir.fdir)
0dcb8d7f 1842 add_untracked(untracked, path.buf + baselen);
1528d247 1843 break;
defd7c7b
KB
1844
1845 default:
1846 break;
1847 }
453ec4bd 1848 }
cf7c6148 1849 close_cached_dir(&cdir);
1528d247 1850 out:
bef36921 1851 strbuf_release(&path);
453ec4bd 1852
defd7c7b 1853 return dir_state;
453ec4bd
LT
1854}
1855
1856static int cmp_name(const void *p1, const void *p2)
1857{
1858 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1859 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1860
ccdd4a0f 1861 return name_compare(e1->name, e1->len, e2->name, e2->len);
453ec4bd
LT
1862}
1863
48ffef96
JH
1864static int treat_leading_path(struct dir_struct *dir,
1865 const char *path, int len,
e1b8c7bd 1866 const struct pathspec *pathspec)
48ffef96 1867{
49dc2cc2
RS
1868 struct strbuf sb = STRBUF_INIT;
1869 int baselen, rc = 0;
48ffef96 1870 const char *cp;
be8a84c5 1871 int old_flags = dir->flags;
48ffef96
JH
1872
1873 while (len && path[len - 1] == '/')
1874 len--;
1875 if (!len)
1876 return 1;
1877 baselen = 0;
be8a84c5 1878 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
48ffef96
JH
1879 while (1) {
1880 cp = path + baselen + !!baselen;
1881 cp = memchr(cp, '/', path + len - cp);
1882 if (!cp)
1883 baselen = len;
1884 else
1885 baselen = cp - path;
49dc2cc2
RS
1886 strbuf_setlen(&sb, 0);
1887 strbuf_add(&sb, path, baselen);
1888 if (!is_directory(sb.buf))
1889 break;
e1b8c7bd 1890 if (simplify_away(sb.buf, sb.len, pathspec))
49dc2cc2 1891 break;
e1b8c7bd 1892 if (treat_one_path(dir, NULL, &sb, baselen, pathspec,
defd7c7b 1893 DT_DIR, NULL) == path_none)
49dc2cc2
RS
1894 break; /* do not recurse into it */
1895 if (len <= baselen) {
1896 rc = 1;
1897 break; /* finished checking */
1898 }
48ffef96 1899 }
49dc2cc2 1900 strbuf_release(&sb);
be8a84c5 1901 dir->flags = old_flags;
49dc2cc2 1902 return rc;
48ffef96
JH
1903}
1904
1e8fef60
NTND
1905static const char *get_ident_string(void)
1906{
1907 static struct strbuf sb = STRBUF_INIT;
1908 struct utsname uts;
1909
1910 if (sb.len)
1911 return sb.buf;
100e4337 1912 if (uname(&uts) < 0)
1e8fef60 1913 die_errno(_("failed to get kernel name and information"));
0e0f7618
CC
1914 strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
1915 uts.sysname);
1e8fef60
NTND
1916 return sb.buf;
1917}
1918
1919static int ident_in_untracked(const struct untracked_cache *uc)
1920{
0e0f7618
CC
1921 /*
1922 * Previous git versions may have saved many NUL separated
1923 * strings in the "ident" field, but it is insane to manage
1924 * many locations, so just take care of the first one.
1925 */
1e8fef60 1926
0e0f7618 1927 return !strcmp(uc->ident.buf, get_ident_string());
1e8fef60
NTND
1928}
1929
0e0f7618 1930static void set_untracked_ident(struct untracked_cache *uc)
1e8fef60 1931{
0e0f7618 1932 strbuf_reset(&uc->ident);
1e8fef60 1933 strbuf_addstr(&uc->ident, get_ident_string());
0e0f7618
CC
1934
1935 /*
1936 * This strbuf used to contain a list of NUL separated
1937 * strings, so save NUL too for backward compatibility.
1938 */
1e8fef60
NTND
1939 strbuf_addch(&uc->ident, 0);
1940}
1941
4a4ca479
CC
1942static void new_untracked_cache(struct index_state *istate)
1943{
1944 struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
1945 strbuf_init(&uc->ident, 100);
1946 uc->exclude_per_dir = ".gitignore";
1947 /* should be the same flags used by git-status */
1948 uc->dir_flags = DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
0e0f7618 1949 set_untracked_ident(uc);
4a4ca479 1950 istate->untracked = uc;
0e0f7618 1951 istate->cache_changed |= UNTRACKED_CHANGED;
4a4ca479
CC
1952}
1953
1954void add_untracked_cache(struct index_state *istate)
1955{
1956 if (!istate->untracked) {
1957 new_untracked_cache(istate);
0e0f7618
CC
1958 } else {
1959 if (!ident_in_untracked(istate->untracked)) {
1960 free_untracked_cache(istate->untracked);
1961 new_untracked_cache(istate);
1962 }
1963 }
4a4ca479
CC
1964}
1965
07b29bfd
CC
1966void remove_untracked_cache(struct index_state *istate)
1967{
1968 if (istate->untracked) {
1969 free_untracked_cache(istate->untracked);
1970 istate->untracked = NULL;
1971 istate->cache_changed |= UNTRACKED_CHANGED;
1972 }
1973}
1974
ccad261f
NTND
1975static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1976 int base_len,
1977 const struct pathspec *pathspec)
1978{
1979 struct untracked_cache_dir *root;
1980
76e6b090 1981 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
ccad261f
NTND
1982 return NULL;
1983
1984 /*
1985 * We only support $GIT_DIR/info/exclude and core.excludesfile
1986 * as the global ignore rule files. Any other additions
1987 * (e.g. from command line) invalidate the cache. This
1988 * condition also catches running setup_standard_excludes()
1989 * before setting dir->untracked!
1990 */
1991 if (dir->unmanaged_exclude_files)
1992 return NULL;
1993
1994 /*
1995 * Optimize for the main use case only: whole-tree git
1996 * status. More work involved in treat_leading_path() if we
1997 * use cache on just a subset of the worktree. pathspec
1998 * support could make the matter even worse.
1999 */
2000 if (base_len || (pathspec && pathspec->nr))
2001 return NULL;
2002
2003 /* Different set of flags may produce different results */
2004 if (dir->flags != dir->untracked->dir_flags ||
2005 /*
2006 * See treat_directory(), case index_nonexistent. Without
2007 * this flag, we may need to also cache .git file content
2008 * for the resolve_gitlink_ref() call, which we don't.
2009 */
2010 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
2011 /* We don't support collecting ignore files */
2012 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2013 DIR_COLLECT_IGNORED)))
2014 return NULL;
2015
2016 /*
2017 * If we use .gitignore in the cache and now you change it to
2018 * .gitexclude, everything will go wrong.
2019 */
2020 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2021 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2022 return NULL;
2023
2024 /*
2025 * EXC_CMDL is not considered in the cache. If people set it,
2026 * skip the cache.
2027 */
2028 if (dir->exclude_list_group[EXC_CMDL].nr)
2029 return NULL;
2030
1e8fef60 2031 if (!ident_in_untracked(dir->untracked)) {
0e0f7618 2032 warning(_("Untracked cache is disabled on this system or location."));
1e8fef60
NTND
2033 return NULL;
2034 }
2035
ccad261f
NTND
2036 if (!dir->untracked->root) {
2037 const int len = sizeof(*dir->untracked->root);
2038 dir->untracked->root = xmalloc(len);
2039 memset(dir->untracked->root, 0, len);
2040 }
2041
2042 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2043 root = dir->untracked->root;
2044 if (hashcmp(dir->ss_info_exclude.sha1,
2045 dir->untracked->ss_info_exclude.sha1)) {
2046 invalidate_gitignore(dir->untracked, root);
2047 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
2048 }
2049 if (hashcmp(dir->ss_excludes_file.sha1,
2050 dir->untracked->ss_excludes_file.sha1)) {
2051 invalidate_gitignore(dir->untracked, root);
2052 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
2053 }
26cb0182
NTND
2054
2055 /* Make sure this directory is not dropped out at saving phase */
2056 root->recurse = 1;
ccad261f
NTND
2057 return root;
2058}
2059
e1b8c7bd
BW
2060int read_directory(struct dir_struct *dir, const char *path,
2061 int len, const struct pathspec *pathspec)
9fc42d60 2062{
ccad261f 2063 struct untracked_cache_dir *untracked;
b4189aa8 2064
dba2e203 2065 if (has_symlink_leading_path(path, len))
725b0605
JH
2066 return dir->nr;
2067
ccad261f
NTND
2068 untracked = validate_untracked_cache(dir, len, pathspec);
2069 if (!untracked)
2070 /*
2071 * make sure untracked cache code path is disabled,
2072 * e.g. prep_exclude()
2073 */
2074 dir->untracked = NULL;
e1b8c7bd
BW
2075 if (!len || treat_leading_path(dir, path, len, pathspec))
2076 read_directory_recursive(dir, path, len, untracked, 0, pathspec);
9ed0d8d6
RS
2077 QSORT(dir->entries, dir->nr, cmp_name);
2078 QSORT(dir->ignored, dir->ignored_nr, cmp_name);
c9ccb5d3
NTND
2079 if (dir->untracked) {
2080 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2081 trace_printf_key(&trace_untracked_stats,
2082 "node creation: %u\n"
2083 "gitignore invalidation: %u\n"
2084 "directory invalidation: %u\n"
2085 "opendir: %u\n",
2086 dir->untracked->dir_created,
2087 dir->untracked->gitignore_invalidated,
2088 dir->untracked->dir_invalidated,
2089 dir->untracked->dir_opened);
1bbb3dba
NTND
2090 if (dir->untracked == the_index.untracked &&
2091 (dir->untracked->dir_opened ||
2092 dir->untracked->gitignore_invalidated ||
2093 dir->untracked->dir_invalidated))
2094 the_index.cache_changed |= UNTRACKED_CHANGED;
2095 if (dir->untracked != the_index.untracked) {
2096 free(dir->untracked);
2097 dir->untracked = NULL;
2098 }
c9ccb5d3 2099 }
453ec4bd
LT
2100 return dir->nr;
2101}
c91f0d92 2102
686a4a06 2103int file_exists(const char *f)
c91f0d92 2104{
686a4a06 2105 struct stat sb;
a50f9fc5 2106 return lstat(f, &sb) == 0;
c91f0d92 2107}
e6636747 2108
63ec5e1f
JS
2109static int cmp_icase(char a, char b)
2110{
2111 if (a == b)
2112 return 0;
2113 if (ignore_case)
2114 return toupper(a) - toupper(b);
2115 return a - b;
2116}
2117
e6636747 2118/*
9b125da4
NTND
2119 * Given two normalized paths (a trailing slash is ok), if subdir is
2120 * outside dir, return -1. Otherwise return the offset in subdir that
2121 * can be used as relative path to dir.
e6636747 2122 */
9b125da4 2123int dir_inside_of(const char *subdir, const char *dir)
e6636747 2124{
9b125da4 2125 int offset = 0;
e6636747 2126
9b125da4 2127 assert(dir && subdir && *dir && *subdir);
e6636747 2128
63ec5e1f 2129 while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
e6636747 2130 dir++;
9b125da4
NTND
2131 subdir++;
2132 offset++;
490544b1 2133 }
9b125da4
NTND
2134
2135 /* hel[p]/me vs hel[l]/yeah */
2136 if (*dir && *subdir)
2137 return -1;
2138
2139 if (!*subdir)
2140 return !*dir ? offset : -1; /* same dir */
2141
2142 /* foo/[b]ar vs foo/[] */
2143 if (is_dir_sep(dir[-1]))
2144 return is_dir_sep(subdir[-1]) ? offset : -1;
2145
2146 /* foo[/]bar vs foo[] */
2147 return is_dir_sep(*subdir) ? offset + 1 : -1;
e6636747
JS
2148}
2149
2150int is_inside_dir(const char *dir)
2151{
56b9f6e7
RS
2152 char *cwd;
2153 int rc;
2154
b892913d
NTND
2155 if (!dir)
2156 return 0;
56b9f6e7
RS
2157
2158 cwd = xgetcwd();
2159 rc = (dir_inside_of(cwd, dir) >= 0);
2160 free(cwd);
2161 return rc;
e6636747 2162}
7155b727 2163
55892d23
AP
2164int is_empty_dir(const char *path)
2165{
2166 DIR *dir = opendir(path);
2167 struct dirent *e;
2168 int ret = 1;
2169
2170 if (!dir)
2171 return 0;
2172
2173 while ((e = readdir(dir)) != NULL)
2174 if (!is_dot_or_dotdot(e->d_name)) {
2175 ret = 0;
2176 break;
2177 }
2178
2179 closedir(dir);
2180 return ret;
2181}
2182
ae2f203e 2183static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
7155b727 2184{
a0f4afbe 2185 DIR *dir;
7155b727 2186 struct dirent *e;
ae2f203e 2187 int ret = 0, original_len = path->len, len, kept_down = 0;
a0f4afbe 2188 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
c844a803 2189 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
a0f4afbe 2190 unsigned char submodule_head[20];
7155b727 2191
a0f4afbe 2192 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
ae2f203e 2193 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
a0f4afbe 2194 /* Do not descend and nuke a nested git work tree. */
ae2f203e
JH
2195 if (kept_up)
2196 *kept_up = 1;
a0f4afbe 2197 return 0;
ae2f203e 2198 }
a0f4afbe 2199
ae2f203e 2200 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
a0f4afbe 2201 dir = opendir(path->buf);
c844a803 2202 if (!dir) {
863808cd
MH
2203 if (errno == ENOENT)
2204 return keep_toplevel ? -1 : 0;
2205 else if (errno == EACCES && !keep_toplevel)
ecb2c282
MH
2206 /*
2207 * An empty dir could be removable even if it
2208 * is unreadable:
2209 */
c844a803
JH
2210 return rmdir(path->buf);
2211 else
2212 return -1;
2213 }
00b6c178 2214 strbuf_complete(path, '/');
7155b727
JS
2215
2216 len = path->len;
2217 while ((e = readdir(dir)) != NULL) {
2218 struct stat st;
8ca12c0d
AP
2219 if (is_dot_or_dotdot(e->d_name))
2220 continue;
7155b727
JS
2221
2222 strbuf_setlen(path, len);
2223 strbuf_addstr(path, e->d_name);
863808cd
MH
2224 if (lstat(path->buf, &st)) {
2225 if (errno == ENOENT)
2226 /*
2227 * file disappeared, which is what we
2228 * wanted anyway
2229 */
2230 continue;
2231 /* fall thru */
2232 } else if (S_ISDIR(st.st_mode)) {
ae2f203e 2233 if (!remove_dir_recurse(path, flag, &kept_down))
7155b727 2234 continue; /* happy */
863808cd
MH
2235 } else if (!only_empty &&
2236 (!unlink(path->buf) || errno == ENOENT)) {
7155b727 2237 continue; /* happy, too */
863808cd 2238 }
7155b727
JS
2239
2240 /* path too long, stat fails, or non-directory still exists */
2241 ret = -1;
2242 break;
2243 }
2244 closedir(dir);
2245
2246 strbuf_setlen(path, original_len);
ae2f203e 2247 if (!ret && !keep_toplevel && !kept_down)
863808cd 2248 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
ae2f203e
JH
2249 else if (kept_up)
2250 /*
2251 * report the uplevel that it is not an error that we
2252 * did not rmdir() our directory.
2253 */
2254 *kept_up = !ret;
7155b727
JS
2255 return ret;
2256}
039bc64e 2257
ae2f203e
JH
2258int remove_dir_recursively(struct strbuf *path, int flag)
2259{
2260 return remove_dir_recurse(path, flag, NULL);
2261}
2262
f932729c
JK
2263static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2264
039bc64e
JH
2265void setup_standard_excludes(struct dir_struct *dir)
2266{
039bc64e 2267 dir->exclude_per_dir = ".gitignore";
099d2d86
JH
2268
2269 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2845ce7f
PT
2270 if (!excludes_file)
2271 excludes_file = xdg_config_home("ignore");
4698c8fe 2272 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
38ccaf93
JH
2273 add_excludes_from_file_1(dir, excludes_file,
2274 dir->untracked ? &dir->ss_excludes_file : NULL);
099d2d86
JH
2275
2276 /* per repository user preference */
f0056f64
JK
2277 if (startup_info->have_repository) {
2278 const char *path = git_path_info_exclude();
2279 if (!access_or_warn(path, R_OK, 0))
2280 add_excludes_from_file_1(dir, path,
2281 dir->untracked ? &dir->ss_info_exclude : NULL);
2282 }
039bc64e 2283}
4a92d1bf
AR
2284
2285int remove_path(const char *name)
2286{
2287 char *slash;
2288
9a6728d4 2289 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
4a92d1bf
AR
2290 return -1;
2291
2292 slash = strrchr(name, '/');
2293 if (slash) {
2294 char *dirs = xstrdup(name);
2295 slash = dirs + (slash - name);
2296 do {
2297 *slash = '\0';
3fc0d131 2298 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
4a92d1bf
AR
2299 free(dirs);
2300 }
2301 return 0;
2302}
2303
270be816
AS
2304/*
2305 * Frees memory within dir which was allocated for exclude lists and
2306 * the exclude_stack. Does not free dir itself.
2307 */
2308void clear_directory(struct dir_struct *dir)
2309{
2310 int i, j;
2311 struct exclude_list_group *group;
2312 struct exclude_list *el;
2313 struct exclude_stack *stk;
2314
2315 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2316 group = &dir->exclude_list_group[i];
2317 for (j = 0; j < group->nr; j++) {
2318 el = &group->el[j];
2319 if (i == EXC_DIRS)
2320 free((char *)el->src);
2321 clear_exclude_list(el);
2322 }
2323 free(group->el);
2324 }
2325
2326 stk = dir->exclude_stack;
2327 while (stk) {
2328 struct exclude_stack *prev = stk->prev;
2329 free(stk);
2330 stk = prev;
2331 }
aceb9429 2332 strbuf_release(&dir->basebuf);
270be816 2333}
83c094ad
NTND
2334
2335struct ondisk_untracked_cache {
2336 struct stat_data info_exclude_stat;
2337 struct stat_data excludes_file_stat;
2338 uint32_t dir_flags;
2339 unsigned char info_exclude_sha1[20];
2340 unsigned char excludes_file_sha1[20];
2341 char exclude_per_dir[FLEX_ARRAY];
2342};
2343
2344#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2345
2346struct write_data {
2347 int index; /* number of written untracked_cache_dir */
2348 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2349 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2350 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2351 struct strbuf out;
2352 struct strbuf sb_stat;
2353 struct strbuf sb_sha1;
2354};
2355
2356static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2357{
2358 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2359 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2360 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2361 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2362 to->sd_dev = htonl(from->sd_dev);
2363 to->sd_ino = htonl(from->sd_ino);
2364 to->sd_uid = htonl(from->sd_uid);
2365 to->sd_gid = htonl(from->sd_gid);
2366 to->sd_size = htonl(from->sd_size);
2367}
2368
2369static void write_one_dir(struct untracked_cache_dir *untracked,
2370 struct write_data *wd)
2371{
2372 struct stat_data stat_data;
2373 struct strbuf *out = &wd->out;
2374 unsigned char intbuf[16];
2375 unsigned int intlen, value;
2376 int i = wd->index++;
2377
2378 /*
2379 * untracked_nr should be reset whenever valid is clear, but
2380 * for safety..
2381 */
2382 if (!untracked->valid) {
2383 untracked->untracked_nr = 0;
2384 untracked->check_only = 0;
2385 }
2386
2387 if (untracked->check_only)
2388 ewah_set(wd->check_only, i);
2389 if (untracked->valid) {
2390 ewah_set(wd->valid, i);
2391 stat_data_to_disk(&stat_data, &untracked->stat_data);
2392 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2393 }
2394 if (!is_null_sha1(untracked->exclude_sha1)) {
2395 ewah_set(wd->sha1_valid, i);
2396 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2397 }
2398
2399 intlen = encode_varint(untracked->untracked_nr, intbuf);
2400 strbuf_add(out, intbuf, intlen);
2401
2402 /* skip non-recurse directories */
2403 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2404 if (untracked->dirs[i]->recurse)
2405 value++;
2406 intlen = encode_varint(value, intbuf);
2407 strbuf_add(out, intbuf, intlen);
2408
2409 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2410
2411 for (i = 0; i < untracked->untracked_nr; i++)
2412 strbuf_add(out, untracked->untracked[i],
2413 strlen(untracked->untracked[i]) + 1);
2414
2415 for (i = 0; i < untracked->dirs_nr; i++)
2416 if (untracked->dirs[i]->recurse)
2417 write_one_dir(untracked->dirs[i], wd);
2418}
2419
2420void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2421{
2422 struct ondisk_untracked_cache *ouc;
2423 struct write_data wd;
2424 unsigned char varbuf[16];
e0b83735
JK
2425 int varint_len;
2426 size_t len = strlen(untracked->exclude_per_dir);
2427
2428 FLEX_ALLOC_MEM(ouc, exclude_per_dir, untracked->exclude_per_dir, len);
83c094ad
NTND
2429 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2430 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2431 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2432 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2433 ouc->dir_flags = htonl(untracked->dir_flags);
1e8fef60
NTND
2434
2435 varint_len = encode_varint(untracked->ident.len, varbuf);
2436 strbuf_add(out, varbuf, varint_len);
8109984d 2437 strbuf_addbuf(out, &untracked->ident);
1e8fef60 2438
83c094ad
NTND
2439 strbuf_add(out, ouc, ouc_size(len));
2440 free(ouc);
2441 ouc = NULL;
2442
2443 if (!untracked->root) {
2444 varint_len = encode_varint(0, varbuf);
2445 strbuf_add(out, varbuf, varint_len);
2446 return;
2447 }
2448
2449 wd.index = 0;
2450 wd.check_only = ewah_new();
2451 wd.valid = ewah_new();
2452 wd.sha1_valid = ewah_new();
2453 strbuf_init(&wd.out, 1024);
2454 strbuf_init(&wd.sb_stat, 1024);
2455 strbuf_init(&wd.sb_sha1, 1024);
2456 write_one_dir(untracked->root, &wd);
2457
2458 varint_len = encode_varint(wd.index, varbuf);
2459 strbuf_add(out, varbuf, varint_len);
2460 strbuf_addbuf(out, &wd.out);
2461 ewah_serialize_strbuf(wd.valid, out);
2462 ewah_serialize_strbuf(wd.check_only, out);
2463 ewah_serialize_strbuf(wd.sha1_valid, out);
2464 strbuf_addbuf(out, &wd.sb_stat);
2465 strbuf_addbuf(out, &wd.sb_sha1);
2466 strbuf_addch(out, '\0'); /* safe guard for string lists */
2467
2468 ewah_free(wd.valid);
2469 ewah_free(wd.check_only);
2470 ewah_free(wd.sha1_valid);
2471 strbuf_release(&wd.out);
2472 strbuf_release(&wd.sb_stat);
2473 strbuf_release(&wd.sb_sha1);
2474}
f9e6c649
NTND
2475
2476static void free_untracked(struct untracked_cache_dir *ucd)
2477{
2478 int i;
2479 if (!ucd)
2480 return;
2481 for (i = 0; i < ucd->dirs_nr; i++)
2482 free_untracked(ucd->dirs[i]);
2483 for (i = 0; i < ucd->untracked_nr; i++)
2484 free(ucd->untracked[i]);
2485 free(ucd->untracked);
2486 free(ucd->dirs);
2487 free(ucd);
2488}
2489
2490void free_untracked_cache(struct untracked_cache *uc)
2491{
2492 if (uc)
2493 free_untracked(uc->root);
2494 free(uc);
2495}
2496
2497struct read_data {
2498 int index;
2499 struct untracked_cache_dir **ucd;
2500 struct ewah_bitmap *check_only;
2501 struct ewah_bitmap *valid;
2502 struct ewah_bitmap *sha1_valid;
2503 const unsigned char *data;
2504 const unsigned char *end;
2505};
2506
2507static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2508{
2509 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2510 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2511 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2512 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2513 to->sd_dev = get_be32(&from->sd_dev);
2514 to->sd_ino = get_be32(&from->sd_ino);
2515 to->sd_uid = get_be32(&from->sd_uid);
2516 to->sd_gid = get_be32(&from->sd_gid);
2517 to->sd_size = get_be32(&from->sd_size);
2518}
2519
2520static int read_one_dir(struct untracked_cache_dir **untracked_,
2521 struct read_data *rd)
2522{
2523 struct untracked_cache_dir ud, *untracked;
2524 const unsigned char *next, *data = rd->data, *end = rd->end;
2525 unsigned int value;
2526 int i, len;
2527
2528 memset(&ud, 0, sizeof(ud));
2529
2530 next = data;
2531 value = decode_varint(&next);
2532 if (next > end)
2533 return -1;
2534 ud.recurse = 1;
2535 ud.untracked_alloc = value;
2536 ud.untracked_nr = value;
2537 if (ud.untracked_nr)
b32fa95f 2538 ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
f9e6c649
NTND
2539 data = next;
2540
2541 next = data;
2542 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2543 if (next > end)
2544 return -1;
b32fa95f 2545 ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
f9e6c649
NTND
2546 data = next;
2547
2548 len = strlen((const char *)data);
2549 next = data + len + 1;
2550 if (next > rd->end)
2551 return -1;
50a6c8ef 2552 *untracked_ = untracked = xmalloc(st_add(sizeof(*untracked), len));
f9e6c649
NTND
2553 memcpy(untracked, &ud, sizeof(ud));
2554 memcpy(untracked->name, data, len + 1);
2555 data = next;
2556
2557 for (i = 0; i < untracked->untracked_nr; i++) {
2558 len = strlen((const char *)data);
2559 next = data + len + 1;
2560 if (next > rd->end)
2561 return -1;
2562 untracked->untracked[i] = xstrdup((const char*)data);
2563 data = next;
2564 }
2565
2566 rd->ucd[rd->index++] = untracked;
2567 rd->data = data;
2568
2569 for (i = 0; i < untracked->dirs_nr; i++) {
2570 len = read_one_dir(untracked->dirs + i, rd);
2571 if (len < 0)
2572 return -1;
2573 }
2574 return 0;
2575}
2576
2577static void set_check_only(size_t pos, void *cb)
2578{
2579 struct read_data *rd = cb;
2580 struct untracked_cache_dir *ud = rd->ucd[pos];
2581 ud->check_only = 1;
2582}
2583
2584static void read_stat(size_t pos, void *cb)
2585{
2586 struct read_data *rd = cb;
2587 struct untracked_cache_dir *ud = rd->ucd[pos];
2588 if (rd->data + sizeof(struct stat_data) > rd->end) {
2589 rd->data = rd->end + 1;
2590 return;
2591 }
2592 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2593 rd->data += sizeof(struct stat_data);
2594 ud->valid = 1;
2595}
2596
2597static void read_sha1(size_t pos, void *cb)
2598{
2599 struct read_data *rd = cb;
2600 struct untracked_cache_dir *ud = rd->ucd[pos];
2601 if (rd->data + 20 > rd->end) {
2602 rd->data = rd->end + 1;
2603 return;
2604 }
2605 hashcpy(ud->exclude_sha1, rd->data);
2606 rd->data += 20;
2607}
2608
2609static void load_sha1_stat(struct sha1_stat *sha1_stat,
2610 const struct stat_data *stat,
2611 const unsigned char *sha1)
2612{
2613 stat_data_from_disk(&sha1_stat->stat, stat);
2614 hashcpy(sha1_stat->sha1, sha1);
2615 sha1_stat->valid = 1;
2616}
2617
2618struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2619{
2620 const struct ondisk_untracked_cache *ouc;
2621 struct untracked_cache *uc;
2622 struct read_data rd;
2623 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
1e8fef60
NTND
2624 const char *ident;
2625 int ident_len, len;
f9e6c649
NTND
2626
2627 if (sz <= 1 || end[-1] != '\0')
2628 return NULL;
2629 end--;
2630
1e8fef60
NTND
2631 ident_len = decode_varint(&next);
2632 if (next + ident_len > end)
2633 return NULL;
2634 ident = (const char *)next;
2635 next += ident_len;
2636
f9e6c649
NTND
2637 ouc = (const struct ondisk_untracked_cache *)next;
2638 if (next + ouc_size(0) > end)
2639 return NULL;
2640
2641 uc = xcalloc(1, sizeof(*uc));
1e8fef60
NTND
2642 strbuf_init(&uc->ident, ident_len);
2643 strbuf_add(&uc->ident, ident, ident_len);
f9e6c649
NTND
2644 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2645 ouc->info_exclude_sha1);
2646 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2647 ouc->excludes_file_sha1);
2648 uc->dir_flags = get_be32(&ouc->dir_flags);
2649 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2650 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2651 next += ouc_size(strlen(ouc->exclude_per_dir));
2652 if (next >= end)
2653 goto done2;
2654
2655 len = decode_varint(&next);
2656 if (next > end || len == 0)
2657 goto done2;
2658
2659 rd.valid = ewah_new();
2660 rd.check_only = ewah_new();
2661 rd.sha1_valid = ewah_new();
2662 rd.data = next;
2663 rd.end = end;
2664 rd.index = 0;
b32fa95f 2665 ALLOC_ARRAY(rd.ucd, len);
f9e6c649
NTND
2666
2667 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2668 goto done;
2669
2670 next = rd.data;
2671 len = ewah_read_mmap(rd.valid, next, end - next);
2672 if (len < 0)
2673 goto done;
2674
2675 next += len;
2676 len = ewah_read_mmap(rd.check_only, next, end - next);
2677 if (len < 0)
2678 goto done;
2679
2680 next += len;
2681 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2682 if (len < 0)
2683 goto done;
2684
2685 ewah_each_bit(rd.check_only, set_check_only, &rd);
2686 rd.data = next + len;
2687 ewah_each_bit(rd.valid, read_stat, &rd);
2688 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2689 next = rd.data;
2690
2691done:
2692 free(rd.ucd);
2693 ewah_free(rd.valid);
2694 ewah_free(rd.check_only);
2695 ewah_free(rd.sha1_valid);
2696done2:
2697 if (next != end) {
2698 free_untracked_cache(uc);
2699 uc = NULL;
2700 }
2701 return uc;
2702}
e931371a 2703
73f9145f
NTND
2704static void invalidate_one_directory(struct untracked_cache *uc,
2705 struct untracked_cache_dir *ucd)
2706{
2707 uc->dir_invalidated++;
2708 ucd->valid = 0;
2709 ucd->untracked_nr = 0;
2710}
2711
2712/*
2713 * Normally when an entry is added or removed from a directory,
2714 * invalidating that directory is enough. No need to touch its
2715 * ancestors. When a directory is shown as "foo/bar/" in git-status
2716 * however, deleting or adding an entry may have cascading effect.
2717 *
2718 * Say the "foo/bar/file" has become untracked, we need to tell the
2719 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2720 * directory any more (because "bar" is managed by foo as an untracked
2721 * "file").
2722 *
2723 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2724 * was the last untracked entry in the entire "foo", we should show
2725 * "foo/" instead. Which means we have to invalidate past "bar" up to
2726 * "foo".
2727 *
2728 * This function traverses all directories from root to leaf. If there
2729 * is a chance of one of the above cases happening, we invalidate back
2730 * to root. Otherwise we just invalidate the leaf. There may be a more
2731 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2732 * detect these cases and avoid unnecessary invalidation, for example,
2733 * checking for the untracked entry named "bar/" in "foo", but for now
2734 * stick to something safe and simple.
2735 */
2736static int invalidate_one_component(struct untracked_cache *uc,
2737 struct untracked_cache_dir *dir,
2738 const char *path, int len)
2739{
2740 const char *rest = strchr(path, '/');
2741
2742 if (rest) {
2743 int component_len = rest - path;
2744 struct untracked_cache_dir *d =
2745 lookup_untracked(uc, dir, path, component_len);
2746 int ret =
2747 invalidate_one_component(uc, d, rest + 1,
2748 len - (component_len + 1));
2749 if (ret)
2750 invalidate_one_directory(uc, dir);
2751 return ret;
2752 }
2753
2754 invalidate_one_directory(uc, dir);
2755 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2756}
2757
e931371a
NTND
2758void untracked_cache_invalidate_path(struct index_state *istate,
2759 const char *path)
2760{
e931371a
NTND
2761 if (!istate->untracked || !istate->untracked->root)
2762 return;
73f9145f
NTND
2763 invalidate_one_component(istate->untracked, istate->untracked->root,
2764 path, strlen(path));
e931371a
NTND
2765}
2766
2767void untracked_cache_remove_from_index(struct index_state *istate,
2768 const char *path)
2769{
2770 untracked_cache_invalidate_path(istate, path);
2771}
2772
2773void untracked_cache_add_to_index(struct index_state *istate,
2774 const char *path)
2775{
2776 untracked_cache_invalidate_path(istate, path);
2777}
47e83eb3
SB
2778
2779/* Update gitfile and core.worktree setting to connect work tree and git dir */
2780void connect_work_tree_and_git_dir(const char *work_tree_, const char *git_dir_)
2781{
365444a6
SB
2782 struct strbuf gitfile_sb = STRBUF_INIT;
2783 struct strbuf cfg_sb = STRBUF_INIT;
47e83eb3 2784 struct strbuf rel_path = STRBUF_INIT;
365444a6 2785 char *git_dir, *work_tree;
47e83eb3 2786
365444a6
SB
2787 /* Prepare .git file */
2788 strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
2789 if (safe_create_leading_directories_const(gitfile_sb.buf))
2790 die(_("could not create directories for %s"), gitfile_sb.buf);
2791
2792 /* Prepare config file */
2793 strbuf_addf(&cfg_sb, "%s/config", git_dir_);
2794 if (safe_create_leading_directories_const(cfg_sb.buf))
2795 die(_("could not create directories for %s"), cfg_sb.buf);
47e83eb3 2796
e394fa01
JH
2797 git_dir = real_pathdup(git_dir_, 1);
2798 work_tree = real_pathdup(work_tree_, 1);
365444a6
SB
2799
2800 /* Write .git file */
2801 write_file(gitfile_sb.buf, "gitdir: %s",
2802 relative_path(git_dir, work_tree, &rel_path));
47e83eb3 2803 /* Update core.worktree setting */
365444a6 2804 git_config_set_in_file(cfg_sb.buf, "core.worktree",
47e83eb3
SB
2805 relative_path(work_tree, git_dir, &rel_path));
2806
365444a6
SB
2807 strbuf_release(&gitfile_sb);
2808 strbuf_release(&cfg_sb);
47e83eb3
SB
2809 strbuf_release(&rel_path);
2810 free(work_tree);
2811 free(git_dir);
2812}
f6f85861
SB
2813
2814/*
2815 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
2816 */
2817void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
2818{
2819 if (rename(old_git_dir, new_git_dir) < 0)
2820 die_errno(_("could not migrate git directory from '%s' to '%s'"),
2821 old_git_dir, new_git_dir);
2822
2823 connect_work_tree_and_git_dir(path, new_git_dir);
2824}