use strbuf_complete to conditionally append slash
[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 */
453ec4bd
LT
10#include "cache.h"
11#include "dir.h"
09595258 12#include "refs.h"
237ec6e4 13#include "wildmatch.h"
64acde94 14#include "pathspec.h"
dde843e7 15#include "utf8.h"
83c094ad
NTND
16#include "varint.h"
17#include "ewah/ewok.h"
453ec4bd 18
9fc42d60
LT
19struct path_simplify {
20 int len;
21 const char *path;
22};
23
defd7c7b
KB
24/*
25 * Tells read_directory_recursive how a file or directory should be treated.
26 * Values are ordered by significance, e.g. if a directory contains both
27 * excluded and untracked files, it is listed as untracked because
28 * path_untracked > path_excluded.
29 */
30enum path_treatment {
31 path_none = 0,
32 path_recurse,
33 path_excluded,
34 path_untracked
35};
36
cf7c6148
NTND
37/*
38 * Support data structure for our opendir/readdir/closedir wrappers
39 */
40struct cached_dir {
41 DIR *fdir;
42 struct untracked_cache_dir *untracked;
91a2288b
NTND
43 int nr_files;
44 int nr_dirs;
45
cf7c6148 46 struct dirent *de;
91a2288b
NTND
47 const char *file;
48 struct untracked_cache_dir *ucd;
cf7c6148
NTND
49};
50
defd7c7b 51static enum path_treatment read_directory_recursive(struct dir_struct *dir,
0dcb8d7f 52 const char *path, int len, struct untracked_cache_dir *untracked,
09595258 53 int check_only, const struct path_simplify *simplify);
caa6b782 54static int get_dtype(struct dirent *de, const char *path, int len);
09595258 55
8cf2a84e
JJ
56/* helper string functions with support for the ignore_case flag */
57int strcmp_icase(const char *a, const char *b)
58{
59 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
60}
61
62int strncmp_icase(const char *a, const char *b, size_t count)
63{
64 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
65}
66
67int fnmatch_icase(const char *pattern, const char *string, int flags)
68{
eb07894f
NTND
69 return wildmatch(pattern, string,
70 flags | (ignore_case ? WM_CASEFOLD : 0),
71 NULL);
8cf2a84e
JJ
72}
73
1f26ce61
CB
74int git_fnmatch(const struct pathspec_item *item,
75 const char *pattern, const char *string,
76 int prefix)
5d74762d 77{
5d74762d 78 if (prefix > 0) {
93d93537 79 if (ps_strncmp(item, pattern, string, prefix))
eb07894f 80 return WM_NOMATCH;
5d74762d
NTND
81 pattern += prefix;
82 string += prefix;
83 }
bd30c2e4 84 if (item->flags & PATHSPEC_ONESTAR) {
8c6abbcd
NTND
85 int pattern_len = strlen(++pattern);
86 int string_len = strlen(string);
87 return string_len < pattern_len ||
93d93537
NTND
88 ps_strcmp(item, pattern,
89 string + string_len - pattern_len);
8c6abbcd 90 }
bd30c2e4 91 if (item->magic & PATHSPEC_GLOB)
93d93537
NTND
92 return wildmatch(pattern, string,
93 WM_PATHNAME |
94 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
95 NULL);
bd30c2e4
NTND
96 else
97 /* wildmatch has not learned no FNM_PATHNAME mode yet */
eb07894f
NTND
98 return wildmatch(pattern, string,
99 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
100 NULL);
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;
124 match_status = wildmatch(use_pat, use_str, flags, NULL);
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
NTND
149 PATHSPEC_ICASE |
150 PATHSPEC_EXCLUDE);
4a085b16 151
827f4d6c 152 for (n = 0; n < pathspec->nr; n++) {
93d93537 153 size_t i = 0, len = 0, item_len;
ef79b1f8
NTND
154 if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
155 continue;
93d93537
NTND
156 if (pathspec->items[n].magic & PATHSPEC_ICASE)
157 item_len = pathspec->items[n].prefix;
158 else
159 item_len = pathspec->items[n].nowildcard_len;
160 while (i < item_len && (n == 0 || i < max)) {
827f4d6c
NTND
161 char c = pathspec->items[n].match[i];
162 if (c != pathspec->items[0].match[i])
4a085b16
JH
163 break;
164 if (c == '/')
165 len = i + 1;
827f4d6c 166 i++;
4a085b16 167 }
827f4d6c 168 if (n == 0 || len < max) {
4a085b16
JH
169 max = len;
170 if (!max)
171 break;
172 }
3c6a370b 173 }
4a085b16 174 return max;
3c6a370b
LT
175}
176
f950eb95
CB
177/*
178 * Returns a copy of the longest leading path common among all
179 * pathspecs.
180 */
827f4d6c 181char *common_prefix(const struct pathspec *pathspec)
f950eb95
CB
182{
183 unsigned long len = common_prefix_len(pathspec);
184
827f4d6c 185 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
f950eb95
CB
186}
187
7327d3d1 188int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
1d8842d9 189{
4a085b16 190 size_t len;
1d8842d9
LT
191
192 /*
193 * Calculate common prefix for the pathspec, and
194 * use that to optimize the directory walk
195 */
4a085b16 196 len = common_prefix_len(pathspec);
1d8842d9
LT
197
198 /* Read the directory and prune it */
b3920bbd 199 read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec);
dba2e203 200 return len;
1d8842d9
LT
201}
202
bc96cc87
NTND
203int within_depth(const char *name, int namelen,
204 int depth, int max_depth)
205{
206 const char *cp = name, *cpe = name + namelen;
207
208 while (cp < cpe) {
209 if (*cp++ != '/')
210 continue;
211 depth++;
212 if (depth > max_depth)
213 return 0;
214 }
215 return 1;
216}
217
42b0874a 218#define DO_MATCH_EXCLUDE 1
68690fdd 219#define DO_MATCH_DIRECTORY 2
42b0874a 220
61cf2820
NTND
221/*
222 * Does 'match' match the given name?
223 * A match is found if
224 *
225 * (1) the 'match' string is leading directory of 'name', or
226 * (2) the 'match' string is a wildcard and matches 'name', or
227 * (3) the 'match' string is exactly the same as 'name'.
228 *
229 * and the return value tells which case it was.
230 *
231 * It returns 0 when there is no match.
232 */
233static int match_pathspec_item(const struct pathspec_item *item, int prefix,
42b0874a 234 const char *name, int namelen, unsigned flags)
61cf2820
NTND
235{
236 /* name/namelen has prefix cut off by caller */
237 const char *match = item->match + prefix;
238 int matchlen = item->len - prefix;
239
93d93537
NTND
240 /*
241 * The normal call pattern is:
242 * 1. prefix = common_prefix_len(ps);
243 * 2. prune something, or fill_directory
854b0959 244 * 3. match_pathspec()
93d93537
NTND
245 *
246 * 'prefix' at #1 may be shorter than the command's prefix and
247 * it's ok for #2 to match extra files. Those extras will be
248 * trimmed at #3.
249 *
250 * Suppose the pathspec is 'foo' and '../bar' running from
251 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
252 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
253 * user does not want XYZ/foo, only the "foo" part should be
254 * case-insensitive. We need to filter out XYZ/foo here. In
255 * other words, we do not trust the caller on comparing the
256 * prefix part when :(icase) is involved. We do exact
257 * comparison ourselves.
258 *
259 * Normally the caller (common_prefix_len() in fact) does
260 * _exact_ matching on name[-prefix+1..-1] and we do not need
261 * to check that part. Be defensive and check it anyway, in
262 * case common_prefix_len is changed, or a new caller is
263 * introduced that does not use common_prefix_len.
264 *
265 * If the penalty turns out too high when prefix is really
266 * long, maybe change it to
267 * strncmp(match, name, item->prefix - prefix)
268 */
269 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
270 strncmp(item->match, name - prefix, item->prefix))
271 return 0;
272
61cf2820
NTND
273 /* If the match was just the prefix, we matched */
274 if (!*match)
275 return MATCHED_RECURSIVELY;
276
93d93537 277 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
61cf2820
NTND
278 if (matchlen == namelen)
279 return MATCHED_EXACTLY;
280
281 if (match[matchlen-1] == '/' || name[matchlen] == '/')
282 return MATCHED_RECURSIVELY;
68690fdd
NTND
283 } else if ((flags & DO_MATCH_DIRECTORY) &&
284 match[matchlen - 1] == '/' &&
285 namelen == matchlen - 1 &&
286 !ps_strncmp(item, match, name, namelen))
287 return MATCHED_EXACTLY;
61cf2820 288
5d74762d 289 if (item->nowildcard_len < item->len &&
bd30c2e4 290 !git_fnmatch(item, match, name,
8c6abbcd 291 item->nowildcard_len - prefix))
61cf2820
NTND
292 return MATCHED_FNMATCH;
293
294 return 0;
295}
296
297/*
52ed1894
AS
298 * Given a name and a list of pathspecs, returns the nature of the
299 * closest (i.e. most specific) match of the name to any of the
300 * pathspecs.
301 *
302 * The caller typically calls this multiple times with the same
303 * pathspec and seen[] array but with different name/namelen
304 * (e.g. entries from the index) and is interested in seeing if and
305 * how each pathspec matches all the names it calls this function
306 * with. A mark is left in the seen[] array for each pathspec element
307 * indicating the closest type of match that element achieved, so if
308 * seen[n] remains zero after multiple invocations, that means the nth
309 * pathspec did not match any names, which could indicate that the
310 * user mistyped the nth pathspec.
61cf2820 311 */
854b0959
NTND
312static int do_match_pathspec(const struct pathspec *ps,
313 const char *name, int namelen,
314 int prefix, char *seen,
42b0874a 315 unsigned flags)
61cf2820 316{
42b0874a 317 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
61cf2820 318
5c6933d2
NTND
319 GUARD_PATHSPEC(ps,
320 PATHSPEC_FROMTOP |
321 PATHSPEC_MAXDEPTH |
bd30c2e4 322 PATHSPEC_LITERAL |
93d93537 323 PATHSPEC_GLOB |
ef79b1f8
NTND
324 PATHSPEC_ICASE |
325 PATHSPEC_EXCLUDE);
8f4f8f45 326
61cf2820 327 if (!ps->nr) {
6330a171
NTND
328 if (!ps->recursive ||
329 !(ps->magic & PATHSPEC_MAXDEPTH) ||
330 ps->max_depth == -1)
61cf2820
NTND
331 return MATCHED_RECURSIVELY;
332
333 if (within_depth(name, namelen, 0, ps->max_depth))
334 return MATCHED_EXACTLY;
335 else
336 return 0;
337 }
338
339 name += prefix;
340 namelen -= prefix;
341
342 for (i = ps->nr - 1; i >= 0; i--) {
343 int how;
ef79b1f8
NTND
344
345 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
346 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
347 continue;
348
61cf2820
NTND
349 if (seen && seen[i] == MATCHED_EXACTLY)
350 continue;
ef79b1f8
NTND
351 /*
352 * Make exclude patterns optional and never report
353 * "pathspec ':(exclude)foo' matches no files"
354 */
355 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
356 seen[i] = MATCHED_FNMATCH;
42b0874a
NTND
357 how = match_pathspec_item(ps->items+i, prefix, name,
358 namelen, flags);
6330a171
NTND
359 if (ps->recursive &&
360 (ps->magic & PATHSPEC_MAXDEPTH) &&
361 ps->max_depth != -1 &&
61cf2820
NTND
362 how && how != MATCHED_FNMATCH) {
363 int len = ps->items[i].len;
364 if (name[len] == '/')
365 len++;
366 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
367 how = MATCHED_EXACTLY;
368 else
369 how = 0;
370 }
371 if (how) {
372 if (retval < how)
373 retval = how;
374 if (seen && seen[i] < how)
375 seen[i] = how;
376 }
377 }
378 return retval;
379}
380
854b0959
NTND
381int match_pathspec(const struct pathspec *ps,
382 const char *name, int namelen,
ae8d0824 383 int prefix, char *seen, int is_dir)
ef79b1f8
NTND
384{
385 int positive, negative;
ae8d0824 386 unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
42b0874a
NTND
387 positive = do_match_pathspec(ps, name, namelen,
388 prefix, seen, flags);
ef79b1f8
NTND
389 if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
390 return positive;
42b0874a
NTND
391 negative = do_match_pathspec(ps, name, namelen,
392 prefix, seen,
393 flags | DO_MATCH_EXCLUDE);
ef79b1f8
NTND
394 return negative ? 0 : positive;
395}
396
777c55a6
JH
397int report_path_error(const char *ps_matched,
398 const struct pathspec *pathspec,
399 const char *prefix)
400{
401 /*
402 * Make sure all pathspec matched; otherwise it is an error.
403 */
777c55a6
JH
404 int num, errors = 0;
405 for (num = 0; num < pathspec->nr; num++) {
406 int other, found_dup;
407
408 if (ps_matched[num])
409 continue;
410 /*
411 * The caller might have fed identical pathspec
412 * twice. Do not barf on such a mistake.
413 * FIXME: parse_pathspec should have eliminated
414 * duplicate pathspec.
415 */
416 for (found_dup = other = 0;
417 !found_dup && other < pathspec->nr;
418 other++) {
419 if (other == num || !ps_matched[other])
420 continue;
421 if (!strcmp(pathspec->items[other].original,
422 pathspec->items[num].original))
423 /*
424 * Ok, we have a match already.
425 */
426 found_dup = 1;
427 }
428 if (found_dup)
429 continue;
430
431 error("pathspec '%s' did not match any file(s) known to git.",
432 pathspec->items[num].original);
433 errors++;
434 }
777c55a6
JH
435 return errors;
436}
437
fcd631ed
NTND
438/*
439 * Return the length of the "simple" part of a path match limiter.
440 */
87323bda 441int simple_length(const char *match)
fcd631ed
NTND
442{
443 int len = -1;
444
445 for (;;) {
446 unsigned char c = *match++;
447 len++;
448 if (c == '\0' || is_glob_special(c))
449 return len;
450 }
451}
452
87323bda 453int no_wildcard(const char *string)
68492fc7 454{
fcd631ed 455 return string[simple_length(string)] == '\0';
68492fc7
LK
456}
457
82dce998
NTND
458void parse_exclude_pattern(const char **pattern,
459 int *patternlen,
460 int *flags,
461 int *nowildcardlen)
84460eec
NTND
462{
463 const char *p = *pattern;
464 size_t i, len;
465
466 *flags = 0;
467 if (*p == '!') {
468 *flags |= EXC_FLAG_NEGATIVE;
469 p++;
470 }
471 len = strlen(p);
472 if (len && p[len - 1] == '/') {
473 len--;
474 *flags |= EXC_FLAG_MUSTBEDIR;
475 }
476 for (i = 0; i < len; i++) {
477 if (p[i] == '/')
478 break;
479 }
480 if (i == len)
481 *flags |= EXC_FLAG_NODIR;
482 *nowildcardlen = simple_length(p);
483 /*
484 * we should have excluded the trailing slash from 'p' too,
485 * but that's one more allocation. Instead just make sure
486 * nowildcardlen does not exceed real patternlen
487 */
488 if (*nowildcardlen > len)
489 *nowildcardlen = len;
490 if (*p == '*' && no_wildcard(p + 1))
491 *flags |= EXC_FLAG_ENDSWITH;
492 *pattern = p;
493 *patternlen = len;
494}
495
453ec4bd 496void add_exclude(const char *string, const char *base,
c04318e4 497 int baselen, struct exclude_list *el, int srcpos)
453ec4bd 498{
d6b8fc30 499 struct exclude *x;
84460eec
NTND
500 int patternlen;
501 int flags;
502 int nowildcardlen;
453ec4bd 503
84460eec
NTND
504 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
505 if (flags & EXC_FLAG_MUSTBEDIR) {
d6b8fc30 506 char *s;
84460eec 507 x = xmalloc(sizeof(*x) + patternlen + 1);
4b25d091 508 s = (char *)(x+1);
84460eec
NTND
509 memcpy(s, string, patternlen);
510 s[patternlen] = '\0';
d6b8fc30 511 x->pattern = s;
d6b8fc30
JH
512 } else {
513 x = xmalloc(sizeof(*x));
514 x->pattern = string;
515 }
84460eec
NTND
516 x->patternlen = patternlen;
517 x->nowildcardlen = nowildcardlen;
453ec4bd
LT
518 x->base = base;
519 x->baselen = baselen;
d6b8fc30 520 x->flags = flags;
c04318e4 521 x->srcpos = srcpos;
840fc334
AS
522 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
523 el->excludes[el->nr++] = x;
c04318e4 524 x->el = el;
453ec4bd
LT
525}
526
55fe6f51
NTND
527static void *read_skip_worktree_file_from_index(const char *path, size_t *size,
528 struct sha1_stat *sha1_stat)
c28b3d6e
NTND
529{
530 int pos, len;
531 unsigned long sz;
532 enum object_type type;
533 void *data;
c28b3d6e
NTND
534
535 len = strlen(path);
71261027 536 pos = cache_name_pos(path, len);
c28b3d6e
NTND
537 if (pos < 0)
538 return NULL;
71261027 539 if (!ce_skip_worktree(active_cache[pos]))
c28b3d6e 540 return NULL;
71261027 541 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
c28b3d6e
NTND
542 if (!data || type != OBJ_BLOB) {
543 free(data);
544 return NULL;
545 }
546 *size = xsize_t(sz);
55fe6f51
NTND
547 if (sha1_stat) {
548 memset(&sha1_stat->stat, 0, sizeof(sha1_stat->stat));
549 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
550 }
c28b3d6e
NTND
551 return data;
552}
553
f6198812
AS
554/*
555 * Frees memory within el which was allocated for exclude patterns and
556 * the file buffer. Does not free el itself.
557 */
558void clear_exclude_list(struct exclude_list *el)
0fd0e241
NTND
559{
560 int i;
561
562 for (i = 0; i < el->nr; i++)
563 free(el->excludes[i]);
564 free(el->excludes);
c082df24 565 free(el->filebuf);
0fd0e241
NTND
566
567 el->nr = 0;
568 el->excludes = NULL;
c082df24 569 el->filebuf = NULL;
0fd0e241
NTND
570}
571
7e2e4b37 572static void trim_trailing_spaces(char *buf)
16402b99 573{
e61a6c1d
PB
574 char *p, *last_space = NULL;
575
576 for (p = buf; *p; p++)
577 switch (*p) {
578 case ' ':
579 if (!last_space)
580 last_space = p;
581 break;
582 case '\\':
583 p++;
584 if (!*p)
585 return;
586 /* fallthrough */
587 default:
588 last_space = NULL;
589 }
590
591 if (last_space)
592 *last_space = '\0';
16402b99
NTND
593}
594
0dcb8d7f
NTND
595/*
596 * Given a subdirectory name and "dir" of the current directory,
597 * search the subdir in "dir" and return it, or create a new one if it
598 * does not exist in "dir".
599 *
600 * If "name" has the trailing slash, it'll be excluded in the search.
601 */
602static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
603 struct untracked_cache_dir *dir,
604 const char *name, int len)
605{
606 int first, last;
607 struct untracked_cache_dir *d;
608 if (!dir)
609 return NULL;
610 if (len && name[len - 1] == '/')
611 len--;
612 first = 0;
613 last = dir->dirs_nr;
614 while (last > first) {
615 int cmp, next = (last + first) >> 1;
616 d = dir->dirs[next];
617 cmp = strncmp(name, d->name, len);
618 if (!cmp && strlen(d->name) > len)
619 cmp = -1;
620 if (!cmp)
621 return d;
622 if (cmp < 0) {
623 last = next;
624 continue;
625 }
626 first = next+1;
627 }
628
629 uc->dir_created++;
630 d = xmalloc(sizeof(*d) + len + 1);
631 memset(d, 0, sizeof(*d));
632 memcpy(d->name, name, len);
633 d->name[len] = '\0';
634
635 ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
636 memmove(dir->dirs + first + 1, dir->dirs + first,
637 (dir->dirs_nr - first) * sizeof(*dir->dirs));
638 dir->dirs_nr++;
639 dir->dirs[first] = d;
640 return d;
641}
642
ccad261f
NTND
643static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
644{
645 int i;
646 dir->valid = 0;
647 dir->untracked_nr = 0;
648 for (i = 0; i < dir->dirs_nr; i++)
649 do_invalidate_gitignore(dir->dirs[i]);
650}
651
652static void invalidate_gitignore(struct untracked_cache *uc,
653 struct untracked_cache_dir *dir)
654{
655 uc->gitignore_invalidated++;
656 do_invalidate_gitignore(dir);
657}
658
91a2288b
NTND
659static void invalidate_directory(struct untracked_cache *uc,
660 struct untracked_cache_dir *dir)
661{
26cb0182 662 int i;
91a2288b
NTND
663 uc->dir_invalidated++;
664 dir->valid = 0;
665 dir->untracked_nr = 0;
26cb0182
NTND
666 for (i = 0; i < dir->dirs_nr; i++)
667 dir->dirs[i]->recurse = 0;
91a2288b
NTND
668}
669
55fe6f51
NTND
670/*
671 * Given a file with name "fname", read it (either from disk, or from
672 * the index if "check_index" is non-zero), parse it and store the
673 * exclude rules in "el".
674 *
675 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
676 * stat data from disk (only valid if add_excludes returns zero). If
677 * ss_valid is non-zero, "ss" must contain good value as input.
678 */
679static int add_excludes(const char *fname, const char *base, int baselen,
680 struct exclude_list *el, int check_index,
681 struct sha1_stat *sha1_stat)
453ec4bd 682{
c470701a 683 struct stat st;
c04318e4 684 int fd, i, lineno = 1;
9d14017a 685 size_t size = 0;
453ec4bd
LT
686 char *buf, *entry;
687
688 fd = open(fname, O_RDONLY);
c28b3d6e 689 if (fd < 0 || fstat(fd, &st) < 0) {
69660731 690 if (errno != ENOENT)
55b38a48 691 warn_on_inaccessible(fname);
c28b3d6e
NTND
692 if (0 <= fd)
693 close(fd);
694 if (!check_index ||
55fe6f51 695 (buf = read_skip_worktree_file_from_index(fname, &size, sha1_stat)) == NULL)
c28b3d6e 696 return -1;
45d76f17
NTND
697 if (size == 0) {
698 free(buf);
699 return 0;
700 }
701 if (buf[size-1] != '\n') {
702 buf = xrealloc(buf, size+1);
703 buf[size++] = '\n';
704 }
d961baa8 705 } else {
c28b3d6e
NTND
706 size = xsize_t(st.st_size);
707 if (size == 0) {
55fe6f51
NTND
708 if (sha1_stat) {
709 fill_stat_data(&sha1_stat->stat, &st);
710 hashcpy(sha1_stat->sha1, EMPTY_BLOB_SHA1_BIN);
711 sha1_stat->valid = 1;
712 }
c28b3d6e
NTND
713 close(fd);
714 return 0;
715 }
45d76f17 716 buf = xmalloc(size+1);
c28b3d6e 717 if (read_in_full(fd, buf, size) != size) {
45d76f17 718 free(buf);
c28b3d6e
NTND
719 close(fd);
720 return -1;
721 }
45d76f17 722 buf[size++] = '\n';
c28b3d6e 723 close(fd);
55fe6f51
NTND
724 if (sha1_stat) {
725 int pos;
726 if (sha1_stat->valid &&
ed4efab1 727 !match_stat_data_racy(&the_index, &sha1_stat->stat, &st))
55fe6f51
NTND
728 ; /* no content change, ss->sha1 still good */
729 else if (check_index &&
730 (pos = cache_name_pos(fname, strlen(fname))) >= 0 &&
731 !ce_stage(active_cache[pos]) &&
732 ce_uptodate(active_cache[pos]) &&
733 !would_convert_to_git(fname))
734 hashcpy(sha1_stat->sha1, active_cache[pos]->sha1);
735 else
736 hash_sha1_file(buf, size, "blob", sha1_stat->sha1);
737 fill_stat_data(&sha1_stat->stat, &st);
738 sha1_stat->valid = 1;
739 }
6ba78238 740 }
453ec4bd 741
c082df24 742 el->filebuf = buf;
245e1c19 743
dde843e7
JH
744 if (skip_utf8_bom(&buf, size))
745 size -= buf - el->filebuf;
746
453ec4bd 747 entry = buf;
245e1c19 748
45d76f17
NTND
749 for (i = 0; i < size; i++) {
750 if (buf[i] == '\n') {
453ec4bd
LT
751 if (entry != buf + i && entry[0] != '#') {
752 buf[i - (i && buf[i-1] == '\r')] = 0;
7e2e4b37 753 trim_trailing_spaces(entry);
c04318e4 754 add_exclude(entry, base, baselen, el, lineno);
453ec4bd 755 }
c04318e4 756 lineno++;
453ec4bd
LT
757 entry = buf + i + 1;
758 }
759 }
760 return 0;
453ec4bd
LT
761}
762
55fe6f51
NTND
763int add_excludes_from_file_to_list(const char *fname, const char *base,
764 int baselen, struct exclude_list *el,
765 int check_index)
766{
767 return add_excludes(fname, base, baselen, el, check_index, NULL);
768}
769
c04318e4
AS
770struct exclude_list *add_exclude_list(struct dir_struct *dir,
771 int group_type, const char *src)
c082df24
AS
772{
773 struct exclude_list *el;
774 struct exclude_list_group *group;
775
776 group = &dir->exclude_list_group[group_type];
777 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
778 el = &group->el[group->nr++];
779 memset(el, 0, sizeof(*el));
c04318e4 780 el->src = src;
c082df24
AS
781 return el;
782}
783
784/*
785 * Used to set up core.excludesfile and .git/info/exclude lists.
786 */
0dcb8d7f
NTND
787static void add_excludes_from_file_1(struct dir_struct *dir, const char *fname,
788 struct sha1_stat *sha1_stat)
453ec4bd 789{
c082df24 790 struct exclude_list *el;
ccad261f
NTND
791 /*
792 * catch setup_standard_excludes() that's called before
793 * dir->untracked is assigned. That function behaves
794 * differently when dir->untracked is non-NULL.
795 */
796 if (!dir->untracked)
797 dir->unmanaged_exclude_files++;
c04318e4 798 el = add_exclude_list(dir, EXC_FILE, fname);
0dcb8d7f 799 if (add_excludes(fname, "", 0, el, 0, sha1_stat) < 0)
453ec4bd
LT
800 die("cannot use %s as an exclude file", fname);
801}
802
0dcb8d7f
NTND
803void add_excludes_from_file(struct dir_struct *dir, const char *fname)
804{
ccad261f 805 dir->unmanaged_exclude_files++; /* see validate_untracked_cache() */
0dcb8d7f
NTND
806 add_excludes_from_file_1(dir, fname, NULL);
807}
808
82dce998
NTND
809int match_basename(const char *basename, int basenamelen,
810 const char *pattern, int prefix, int patternlen,
811 int flags)
593cb880
NTND
812{
813 if (prefix == patternlen) {
0b6e56df
JH
814 if (patternlen == basenamelen &&
815 !strncmp_icase(pattern, basename, basenamelen))
593cb880
NTND
816 return 1;
817 } else if (flags & EXC_FLAG_ENDSWITH) {
0b6e56df 818 /* "*literal" matching against "fooliteral" */
593cb880 819 if (patternlen - 1 <= basenamelen &&
0b6e56df
JH
820 !strncmp_icase(pattern + 1,
821 basename + basenamelen - (patternlen - 1),
822 patternlen - 1))
593cb880
NTND
823 return 1;
824 } else {
0b6e56df
JH
825 if (fnmatch_icase_mem(pattern, patternlen,
826 basename, basenamelen,
827 0) == 0)
593cb880
NTND
828 return 1;
829 }
830 return 0;
831}
832
82dce998
NTND
833int match_pathname(const char *pathname, int pathlen,
834 const char *base, int baselen,
835 const char *pattern, int prefix, int patternlen,
836 int flags)
b5592632
NTND
837{
838 const char *name;
839 int namelen;
840
841 /*
842 * match with FNM_PATHNAME; the pattern has base implicitly
843 * in front of it.
844 */
845 if (*pattern == '/') {
846 pattern++;
982ac873 847 patternlen--;
b5592632
NTND
848 prefix--;
849 }
850
851 /*
852 * baselen does not count the trailing slash. base[] may or
853 * may not end with a trailing slash though.
854 */
855 if (pathlen < baselen + 1 ||
856 (baselen && pathname[baselen] != '/') ||
857 strncmp_icase(pathname, base, baselen))
858 return 0;
859
860 namelen = baselen ? pathlen - baselen - 1 : pathlen;
861 name = pathname + pathlen - namelen;
862
863 if (prefix) {
864 /*
865 * if the non-wildcard part is longer than the
866 * remaining pathname, surely it cannot match.
867 */
868 if (prefix > namelen)
869 return 0;
870
871 if (strncmp_icase(pattern, name, prefix))
872 return 0;
873 pattern += prefix;
ab3aebc1 874 patternlen -= prefix;
b5592632
NTND
875 name += prefix;
876 namelen -= prefix;
ab3aebc1
JK
877
878 /*
879 * If the whole pattern did not have a wildcard,
880 * then our prefix match is all we need; we
881 * do not need to call fnmatch at all.
882 */
883 if (!patternlen && !namelen)
884 return 1;
b5592632
NTND
885 }
886
ab3aebc1
JK
887 return fnmatch_icase_mem(pattern, patternlen,
888 name, namelen,
f30366b2 889 WM_PATHNAME) == 0;
b5592632
NTND
890}
891
578cd7c3
AS
892/*
893 * Scan the given exclude list in reverse to see whether pathname
894 * should be ignored. The first match (i.e. the last on the list), if
895 * any, determines the fate. Returns the exclude_list element which
896 * matched, or NULL for undecided.
453ec4bd 897 */
578cd7c3
AS
898static struct exclude *last_exclude_matching_from_list(const char *pathname,
899 int pathlen,
900 const char *basename,
901 int *dtype,
902 struct exclude_list *el)
453ec4bd
LT
903{
904 int i;
905
35a94d44 906 if (!el->nr)
578cd7c3 907 return NULL; /* undefined */
d6b8fc30 908
35a94d44
NTND
909 for (i = el->nr - 1; 0 <= i; i--) {
910 struct exclude *x = el->excludes[i];
b5592632 911 const char *exclude = x->pattern;
b5592632 912 int prefix = x->nowildcardlen;
35a94d44
NTND
913
914 if (x->flags & EXC_FLAG_MUSTBEDIR) {
915 if (*dtype == DT_UNKNOWN)
916 *dtype = get_dtype(NULL, pathname, pathlen);
917 if (*dtype != DT_DIR)
918 continue;
919 }
d6b8fc30 920
35a94d44 921 if (x->flags & EXC_FLAG_NODIR) {
593cb880
NTND
922 if (match_basename(basename,
923 pathlen - (basename - pathname),
924 exclude, prefix, x->patternlen,
925 x->flags))
578cd7c3 926 return x;
35a94d44
NTND
927 continue;
928 }
929
b5592632
NTND
930 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
931 if (match_pathname(pathname, pathlen,
932 x->base, x->baselen ? x->baselen - 1 : 0,
933 exclude, prefix, x->patternlen, x->flags))
578cd7c3 934 return x;
453ec4bd 935 }
578cd7c3
AS
936 return NULL; /* undecided */
937}
938
939/*
940 * Scan the list and let the last match determine the fate.
941 * Return 1 for exclude, 0 for include and -1 for undecided.
942 */
943int is_excluded_from_list(const char *pathname,
944 int pathlen, const char *basename, int *dtype,
945 struct exclude_list *el)
946{
947 struct exclude *exclude;
948 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
949 if (exclude)
950 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
453ec4bd
LT
951 return -1; /* undecided */
952}
953
46aa2f95
KB
954static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
955 const char *pathname, int pathlen, const char *basename,
956 int *dtype_p)
957{
958 int i, j;
959 struct exclude_list_group *group;
960 struct exclude *exclude;
961 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
962 group = &dir->exclude_list_group[i];
963 for (j = group->nr - 1; j >= 0; j--) {
964 exclude = last_exclude_matching_from_list(
965 pathname, pathlen, basename, dtype_p,
966 &group->el[j]);
967 if (exclude)
968 return exclude;
969 }
970 }
971 return NULL;
972}
973
6cd5c582
KB
974/*
975 * Loads the per-directory exclude list for the substring of base
976 * which has a char length of baselen.
977 */
978static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
979{
980 struct exclude_list_group *group;
981 struct exclude_list *el;
982 struct exclude_stack *stk = NULL;
0dcb8d7f 983 struct untracked_cache_dir *untracked;
6cd5c582
KB
984 int current;
985
6cd5c582
KB
986 group = &dir->exclude_list_group[EXC_DIRS];
987
d961baa8
NTND
988 /*
989 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
6cd5c582 990 * which originate from directories not in the prefix of the
d961baa8
NTND
991 * path being checked.
992 */
6cd5c582
KB
993 while ((stk = dir->exclude_stack) != NULL) {
994 if (stk->baselen <= baselen &&
aceb9429 995 !strncmp(dir->basebuf.buf, base, stk->baselen))
6cd5c582
KB
996 break;
997 el = &group->el[dir->exclude_stack->exclude_ix];
998 dir->exclude_stack = stk->prev;
95c6f271 999 dir->exclude = NULL;
aceb9429 1000 free((char *)el->src); /* see strbuf_detach() below */
6cd5c582
KB
1001 clear_exclude_list(el);
1002 free(stk);
1003 group->nr--;
1004 }
1005
95c6f271
KB
1006 /* Skip traversing into sub directories if the parent is excluded */
1007 if (dir->exclude)
1008 return;
1009
aceb9429
NTND
1010 /*
1011 * Lazy initialization. All call sites currently just
1012 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1013 * them seems lots of work for little benefit.
1014 */
1015 if (!dir->basebuf.buf)
1016 strbuf_init(&dir->basebuf, PATH_MAX);
1017
6cd5c582
KB
1018 /* Read from the parent directories and push them down. */
1019 current = stk ? stk->baselen : -1;
aceb9429 1020 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
0dcb8d7f
NTND
1021 if (dir->untracked)
1022 untracked = stk ? stk->ucd : dir->untracked->root;
1023 else
1024 untracked = NULL;
1025
6cd5c582 1026 while (current < baselen) {
6cd5c582 1027 const char *cp;
0dcb8d7f 1028 struct sha1_stat sha1_stat;
6cd5c582 1029
03e11a71 1030 stk = xcalloc(1, sizeof(*stk));
6cd5c582
KB
1031 if (current < 0) {
1032 cp = base;
1033 current = 0;
d961baa8 1034 } else {
6cd5c582
KB
1035 cp = strchr(base + current + 1, '/');
1036 if (!cp)
1037 die("oops in prep_exclude");
1038 cp++;
0dcb8d7f
NTND
1039 untracked =
1040 lookup_untracked(dir->untracked, untracked,
1041 base + current,
1042 cp - base - current);
6cd5c582
KB
1043 }
1044 stk->prev = dir->exclude_stack;
1045 stk->baselen = cp - base;
95c6f271 1046 stk->exclude_ix = group->nr;
0dcb8d7f 1047 stk->ucd = untracked;
95c6f271 1048 el = add_exclude_list(dir, EXC_DIRS, NULL);
aceb9429
NTND
1049 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
1050 assert(stk->baselen == dir->basebuf.len);
95c6f271
KB
1051
1052 /* Abort if the directory is excluded */
1053 if (stk->baselen) {
1054 int dt = DT_DIR;
aceb9429 1055 dir->basebuf.buf[stk->baselen - 1] = 0;
95c6f271 1056 dir->exclude = last_exclude_matching_from_lists(dir,
aceb9429
NTND
1057 dir->basebuf.buf, stk->baselen - 1,
1058 dir->basebuf.buf + current, &dt);
1059 dir->basebuf.buf[stk->baselen - 1] = '/';
c3c327de
KB
1060 if (dir->exclude &&
1061 dir->exclude->flags & EXC_FLAG_NEGATIVE)
1062 dir->exclude = NULL;
95c6f271 1063 if (dir->exclude) {
95c6f271
KB
1064 dir->exclude_stack = stk;
1065 return;
1066 }
1067 }
1068
aceb9429 1069 /* Try to read per-directory file */
0dcb8d7f
NTND
1070 hashclr(sha1_stat.sha1);
1071 sha1_stat.valid = 0;
27b099ae
NTND
1072 if (dir->exclude_per_dir &&
1073 /*
1074 * If we know that no files have been added in
1075 * this directory (i.e. valid_cached_dir() has
1076 * been executed and set untracked->valid) ..
1077 */
1078 (!untracked || !untracked->valid ||
1079 /*
1080 * .. and .gitignore does not exist before
7687252f
DT
1081 * (i.e. null exclude_sha1). Then we can skip
1082 * loading .gitignore, which would result in
1083 * ENOENT anyway.
27b099ae
NTND
1084 */
1085 !is_null_sha1(untracked->exclude_sha1))) {
95c6f271
KB
1086 /*
1087 * dir->basebuf gets reused by the traversal, but we
1088 * need fname to remain unchanged to ensure the src
1089 * member of each struct exclude correctly
1090 * back-references its source file. Other invocations
1091 * of add_exclude_list provide stable strings, so we
aceb9429 1092 * strbuf_detach() and free() here in the caller.
95c6f271 1093 */
aceb9429
NTND
1094 struct strbuf sb = STRBUF_INIT;
1095 strbuf_addbuf(&sb, &dir->basebuf);
1096 strbuf_addstr(&sb, dir->exclude_per_dir);
1097 el->src = strbuf_detach(&sb, NULL);
0dcb8d7f
NTND
1098 add_excludes(el->src, el->src, stk->baselen, el, 1,
1099 untracked ? &sha1_stat : NULL);
1100 }
5ebf79ad
NTND
1101 /*
1102 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1103 * will first be called in valid_cached_dir() then maybe many
1104 * times more in last_exclude_matching(). When the cache is
1105 * used, last_exclude_matching() will not be called and
1106 * reading .gitignore content will be a waste.
1107 *
1108 * So when it's called by valid_cached_dir() and we can get
1109 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1110 * modified on work tree), we could delay reading the
1111 * .gitignore content until we absolutely need it in
1112 * last_exclude_matching(). Be careful about ignore rule
1113 * order, though, if you do that.
1114 */
1115 if (untracked &&
1116 hashcmp(sha1_stat.sha1, untracked->exclude_sha1)) {
1117 invalidate_gitignore(dir->untracked, untracked);
0dcb8d7f 1118 hashcpy(untracked->exclude_sha1, sha1_stat.sha1);
95c6f271 1119 }
6cd5c582
KB
1120 dir->exclude_stack = stk;
1121 current = stk->baselen;
1122 }
aceb9429 1123 strbuf_setlen(&dir->basebuf, baselen);
6cd5c582
KB
1124}
1125
f4cd69a6
AS
1126/*
1127 * Loads the exclude lists for the directory containing pathname, then
1128 * scans all exclude lists to determine whether pathname is excluded.
1129 * Returns the exclude_list element which matched, or NULL for
1130 * undecided.
1131 */
b07bc8c8 1132struct exclude *last_exclude_matching(struct dir_struct *dir,
f4cd69a6
AS
1133 const char *pathname,
1134 int *dtype_p)
453ec4bd
LT
1135{
1136 int pathlen = strlen(pathname);
68492fc7
LK
1137 const char *basename = strrchr(pathname, '/');
1138 basename = (basename) ? basename+1 : pathname;
453ec4bd 1139
63d285c8 1140 prep_exclude(dir, pathname, basename-pathname);
c082df24 1141
95c6f271
KB
1142 if (dir->exclude)
1143 return dir->exclude;
1144
46aa2f95
KB
1145 return last_exclude_matching_from_lists(dir, pathname, pathlen,
1146 basename, dtype_p);
f4cd69a6
AS
1147}
1148
1149/*
1150 * Loads the exclude lists for the directory containing pathname, then
1151 * scans all exclude lists to determine whether pathname is excluded.
1152 * Returns 1 if true, otherwise 0.
1153 */
b07bc8c8 1154int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
f4cd69a6
AS
1155{
1156 struct exclude *exclude =
1157 last_exclude_matching(dir, pathname, dtype_p);
1158 if (exclude)
1159 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
453ec4bd
LT
1160 return 0;
1161}
1162
f3fa1838
JH
1163static struct dir_entry *dir_entry_new(const char *pathname, int len)
1164{
453ec4bd
LT
1165 struct dir_entry *ent;
1166
453ec4bd
LT
1167 ent = xmalloc(sizeof(*ent) + len + 1);
1168 ent->len = len;
1169 memcpy(ent->name, pathname, len);
1170 ent->name[len] = 0;
4d06f8ac 1171 return ent;
453ec4bd
LT
1172}
1173
159b3212 1174static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
6815e569 1175{
ebbd7439 1176 if (cache_file_exists(pathname, len, ignore_case))
6815e569
JK
1177 return NULL;
1178
25fd2f7a 1179 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
6815e569
JK
1180 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1181}
1182
108da0db 1183struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
2abd31b0 1184{
6e4f981f 1185 if (!cache_name_is_other(pathname, len))
2abd31b0
JK
1186 return NULL;
1187
25fd2f7a 1188 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
2abd31b0
JK
1189 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1190}
1191
09595258
LT
1192enum exist_status {
1193 index_nonexistent = 0,
1194 index_directory,
4b05548f 1195 index_gitdir
09595258
LT
1196};
1197
5102c617 1198/*
71261027 1199 * Do not use the alphabetically sorted index to look up
5102c617 1200 * the directory name; instead, use the case insensitive
ebbd7439 1201 * directory hash.
5102c617
JJ
1202 */
1203static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
1204{
d28eec26 1205 const struct cache_entry *ce = cache_dir_exists(dirname, len);
5102c617
JJ
1206 unsigned char endchar;
1207
1208 if (!ce)
1209 return index_nonexistent;
1210 endchar = ce->name[len];
1211
1212 /*
1213 * The cache_entry structure returned will contain this dirname
1214 * and possibly additional path components.
1215 */
1216 if (endchar == '/')
1217 return index_directory;
1218
1219 /*
1220 * If there are no additional path components, then this cache_entry
1221 * represents a submodule. Submodules, despite being directories,
1222 * are stored in the cache without a closing slash.
1223 */
1224 if (!endchar && S_ISGITLINK(ce->ce_mode))
1225 return index_gitdir;
1226
1227 /* This should never be hit, but it exists just in case. */
1228 return index_nonexistent;
1229}
1230
09595258
LT
1231/*
1232 * The index sorts alphabetically by entry name, which
1233 * means that a gitlink sorts as '\0' at the end, while
1234 * a directory (which is defined not as an entry, but as
1235 * the files it contains) will sort with the '/' at the
1236 * end.
1237 */
1238static enum exist_status directory_exists_in_index(const char *dirname, int len)
453ec4bd 1239{
5102c617
JJ
1240 int pos;
1241
1242 if (ignore_case)
1243 return directory_exists_in_index_icase(dirname, len);
1244
1245 pos = cache_name_pos(dirname, len);
09595258
LT
1246 if (pos < 0)
1247 pos = -pos-1;
1248 while (pos < active_nr) {
9c5e6c80 1249 const struct cache_entry *ce = active_cache[pos++];
09595258
LT
1250 unsigned char endchar;
1251
1252 if (strncmp(ce->name, dirname, len))
1253 break;
1254 endchar = ce->name[len];
1255 if (endchar > '/')
1256 break;
1257 if (endchar == '/')
1258 return index_directory;
7a51ed66 1259 if (!endchar && S_ISGITLINK(ce->ce_mode))
09595258
LT
1260 return index_gitdir;
1261 }
1262 return index_nonexistent;
1263}
1264
1265/*
1266 * When we find a directory when traversing the filesystem, we
1267 * have three distinct cases:
1268 *
1269 * - ignore it
1270 * - see it as a directory
1271 * - recurse into it
1272 *
1273 * and which one we choose depends on a combination of existing
1274 * git index contents and the flags passed into the directory
1275 * traversal routine.
1276 *
1277 * Case 1: If we *already* have entries in the index under that
5bd8e2d8
KB
1278 * directory name, we always recurse into the directory to see
1279 * all the files.
09595258
LT
1280 *
1281 * Case 2: If we *already* have that directory name as a gitlink,
1282 * we always continue to see it as a gitlink, regardless of whether
1283 * there is an actual git directory there or not (it might not
1284 * be checked out as a subproject!)
1285 *
1286 * Case 3: if we didn't have it in the index previously, we
1287 * have a few sub-cases:
1288 *
1289 * (a) if "show_other_directories" is true, we show it as
1290 * just a directory, unless "hide_empty_directories" is
defd7c7b
KB
1291 * also true, in which case we need to check if it contains any
1292 * untracked and / or ignored files.
09595258 1293 * (b) if it looks like a git directory, and we don't have
302b9282 1294 * 'no_gitlinks' set we treat it as a gitlink, and show it
09595258
LT
1295 * as a directory.
1296 * (c) otherwise, we recurse into it.
1297 */
defd7c7b 1298static enum path_treatment treat_directory(struct dir_struct *dir,
0dcb8d7f 1299 struct untracked_cache_dir *untracked,
2e5910f2 1300 const char *dirname, int len, int baselen, int exclude,
09595258
LT
1301 const struct path_simplify *simplify)
1302{
1303 /* The "len-1" is to strip the final '/' */
1304 switch (directory_exists_in_index(dirname, len-1)) {
1305 case index_directory:
defd7c7b 1306 return path_recurse;
09595258
LT
1307
1308 case index_gitdir:
26c986e1 1309 return path_none;
09595258
LT
1310
1311 case index_nonexistent:
7c4c97c0 1312 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
09595258 1313 break;
7c4c97c0 1314 if (!(dir->flags & DIR_NO_GITLINKS)) {
09595258
LT
1315 unsigned char sha1[20];
1316 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
defd7c7b 1317 return path_untracked;
09595258 1318 }
defd7c7b 1319 return path_recurse;
09595258
LT
1320 }
1321
1322 /* This is the "show_other_directories" case */
721ac4ed 1323
184d2a8e 1324 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
defd7c7b
KB
1325 return exclude ? path_excluded : path_untracked;
1326
2e5910f2
DT
1327 untracked = lookup_untracked(dir->untracked, untracked,
1328 dirname + baselen, len - baselen);
0dcb8d7f
NTND
1329 return read_directory_recursive(dir, dirname, len,
1330 untracked, 1, simplify);
453ec4bd
LT
1331}
1332
9fc42d60
LT
1333/*
1334 * This is an inexact early pruning of any recursive directory
1335 * reading - if the path cannot possibly be in the pathspec,
1336 * return true, and we'll skip it early.
1337 */
1338static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1339{
1340 if (simplify) {
1341 for (;;) {
1342 const char *match = simplify->path;
1343 int len = simplify->len;
1344
1345 if (!match)
1346 break;
1347 if (len > pathlen)
1348 len = pathlen;
1349 if (!memcmp(path, match, len))
1350 return 0;
1351 simplify++;
1352 }
1353 return 1;
1354 }
1355 return 0;
1356}
1357
29209cbe
JK
1358/*
1359 * This function tells us whether an excluded path matches a
1360 * list of "interesting" pathspecs. That is, whether a path matched
1361 * by any of the pathspecs could possibly be ignored by excluding
1362 * the specified path. This can happen if:
1363 *
1364 * 1. the path is mentioned explicitly in the pathspec
1365 *
1366 * 2. the path is a directory prefix of some element in the
1367 * pathspec
1368 */
1369static int exclude_matches_pathspec(const char *path, int len,
1370 const struct path_simplify *simplify)
e96980ef
JK
1371{
1372 if (simplify) {
1373 for (; simplify->path; simplify++) {
1374 if (len == simplify->len
1375 && !memcmp(path, simplify->path, len))
1376 return 1;
29209cbe
JK
1377 if (len < simplify->len
1378 && simplify->path[len] == '/'
1379 && !memcmp(path, simplify->path, len))
1380 return 1;
e96980ef
JK
1381 }
1382 }
1383 return 0;
1384}
1385
443e061a
LT
1386static int get_index_dtype(const char *path, int len)
1387{
1388 int pos;
9c5e6c80 1389 const struct cache_entry *ce;
443e061a 1390
ebbd7439 1391 ce = cache_file_exists(path, len, 0);
443e061a
LT
1392 if (ce) {
1393 if (!ce_uptodate(ce))
1394 return DT_UNKNOWN;
1395 if (S_ISGITLINK(ce->ce_mode))
1396 return DT_DIR;
1397 /*
1398 * Nobody actually cares about the
1399 * difference between DT_LNK and DT_REG
1400 */
1401 return DT_REG;
1402 }
1403
1404 /* Try to look it up as a directory */
1405 pos = cache_name_pos(path, len);
1406 if (pos >= 0)
1407 return DT_UNKNOWN;
1408 pos = -pos-1;
1409 while (pos < active_nr) {
1410 ce = active_cache[pos++];
1411 if (strncmp(ce->name, path, len))
1412 break;
1413 if (ce->name[len] > '/')
1414 break;
1415 if (ce->name[len] < '/')
1416 continue;
1417 if (!ce_uptodate(ce))
1418 break; /* continue? */
1419 return DT_DIR;
1420 }
1421 return DT_UNKNOWN;
1422}
1423
caa6b782 1424static int get_dtype(struct dirent *de, const char *path, int len)
07134421 1425{
6831a88a 1426 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
07134421
LT
1427 struct stat st;
1428
1429 if (dtype != DT_UNKNOWN)
1430 return dtype;
443e061a
LT
1431 dtype = get_index_dtype(path, len);
1432 if (dtype != DT_UNKNOWN)
1433 return dtype;
1434 if (lstat(path, &st))
07134421
LT
1435 return dtype;
1436 if (S_ISREG(st.st_mode))
1437 return DT_REG;
1438 if (S_ISDIR(st.st_mode))
1439 return DT_DIR;
1440 if (S_ISLNK(st.st_mode))
1441 return DT_LNK;
1442 return dtype;
1443}
1444
16e2cfa9 1445static enum path_treatment treat_one_path(struct dir_struct *dir,
0dcb8d7f 1446 struct untracked_cache_dir *untracked,
49dc2cc2 1447 struct strbuf *path,
2e5910f2 1448 int baselen,
16e2cfa9
JH
1449 const struct path_simplify *simplify,
1450 int dtype, struct dirent *de)
53cc5356 1451{
8aaf8d77 1452 int exclude;
ebbd7439 1453 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
2eac2a4c 1454
8aaf8d77
KB
1455 if (dtype == DT_UNKNOWN)
1456 dtype = get_dtype(de, path->buf, path->len);
1457
1458 /* Always exclude indexed files */
2eac2a4c 1459 if (dtype != DT_DIR && has_path_in_index)
defd7c7b 1460 return path_none;
8aaf8d77 1461
2eac2a4c
JH
1462 /*
1463 * When we are looking at a directory P in the working tree,
1464 * there are three cases:
1465 *
1466 * (1) P exists in the index. Everything inside the directory P in
1467 * the working tree needs to go when P is checked out from the
1468 * index.
1469 *
1470 * (2) P does not exist in the index, but there is P/Q in the index.
1471 * We know P will stay a directory when we check out the contents
1472 * of the index, but we do not know yet if there is a directory
1473 * P/Q in the working tree to be killed, so we need to recurse.
1474 *
1475 * (3) P does not exist in the index, and there is no P/Q in the index
1476 * to require P to be a directory, either. Only in this case, we
1477 * know that everything inside P will not be killed without
1478 * recursing.
1479 */
1480 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1481 (dtype == DT_DIR) &&
de372b1b
ES
1482 !has_path_in_index &&
1483 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1484 return path_none;
8aaf8d77
KB
1485
1486 exclude = is_excluded(dir, path->buf, &dtype);
53cc5356
JH
1487
1488 /*
1489 * Excluded? If we don't explicitly want to show
1490 * ignored files, ignore it
1491 */
0aaf62b6 1492 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
defd7c7b 1493 return path_excluded;
53cc5356 1494
53cc5356
JH
1495 switch (dtype) {
1496 default:
defd7c7b 1497 return path_none;
53cc5356 1498 case DT_DIR:
49dc2cc2 1499 strbuf_addch(path, '/');
2e5910f2
DT
1500 return treat_directory(dir, untracked, path->buf, path->len,
1501 baselen, exclude, simplify);
53cc5356
JH
1502 case DT_REG:
1503 case DT_LNK:
defd7c7b 1504 return exclude ? path_excluded : path_untracked;
53cc5356 1505 }
53cc5356
JH
1506}
1507
91a2288b
NTND
1508static enum path_treatment treat_path_fast(struct dir_struct *dir,
1509 struct untracked_cache_dir *untracked,
1510 struct cached_dir *cdir,
1511 struct strbuf *path,
1512 int baselen,
1513 const struct path_simplify *simplify)
1514{
1515 strbuf_setlen(path, baselen);
1516 if (!cdir->ucd) {
1517 strbuf_addstr(path, cdir->file);
1518 return path_untracked;
1519 }
1520 strbuf_addstr(path, cdir->ucd->name);
1521 /* treat_one_path() does this before it calls treat_directory() */
00b6c178 1522 strbuf_complete(path, '/');
91a2288b
NTND
1523 if (cdir->ucd->check_only)
1524 /*
1525 * check_only is set as a result of treat_directory() getting
1526 * to its bottom. Verify again the same set of directories
1527 * with check_only set.
1528 */
1529 return read_directory_recursive(dir, path->buf, path->len,
1530 cdir->ucd, 1, simplify);
1531 /*
1532 * We get path_recurse in the first run when
1533 * directory_exists_in_index() returns index_nonexistent. We
1534 * are sure that new changes in the index does not impact the
1535 * outcome. Return now.
1536 */
1537 return path_recurse;
1538}
1539
16e2cfa9 1540static enum path_treatment treat_path(struct dir_struct *dir,
0dcb8d7f 1541 struct untracked_cache_dir *untracked,
cf7c6148 1542 struct cached_dir *cdir,
49dc2cc2 1543 struct strbuf *path,
16e2cfa9 1544 int baselen,
49dc2cc2 1545 const struct path_simplify *simplify)
16e2cfa9
JH
1546{
1547 int dtype;
cf7c6148 1548 struct dirent *de = cdir->de;
16e2cfa9 1549
91a2288b
NTND
1550 if (!de)
1551 return treat_path_fast(dir, untracked, cdir, path,
1552 baselen, simplify);
16e2cfa9 1553 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
defd7c7b 1554 return path_none;
49dc2cc2
RS
1555 strbuf_setlen(path, baselen);
1556 strbuf_addstr(path, de->d_name);
1557 if (simplify_away(path->buf, path->len, simplify))
defd7c7b 1558 return path_none;
16e2cfa9
JH
1559
1560 dtype = DTYPE(de);
2e5910f2 1561 return treat_one_path(dir, untracked, path, baselen, simplify, dtype, de);
0dcb8d7f
NTND
1562}
1563
1564static void add_untracked(struct untracked_cache_dir *dir, const char *name)
1565{
1566 if (!dir)
1567 return;
1568 ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
1569 dir->untracked_alloc);
1570 dir->untracked[dir->untracked_nr++] = xstrdup(name);
16e2cfa9
JH
1571}
1572
91a2288b
NTND
1573static int valid_cached_dir(struct dir_struct *dir,
1574 struct untracked_cache_dir *untracked,
1575 struct strbuf *path,
1576 int check_only)
1577{
1578 struct stat st;
1579
1580 if (!untracked)
1581 return 0;
1582
1583 if (stat(path->len ? path->buf : ".", &st)) {
1584 invalidate_directory(dir->untracked, untracked);
1585 memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
1586 return 0;
1587 }
1588 if (!untracked->valid ||
ed4efab1 1589 match_stat_data_racy(&the_index, &untracked->stat_data, &st)) {
91a2288b
NTND
1590 if (untracked->valid)
1591 invalidate_directory(dir->untracked, untracked);
1592 fill_stat_data(&untracked->stat_data, &st);
1593 return 0;
1594 }
1595
1596 if (untracked->check_only != !!check_only) {
1597 invalidate_directory(dir->untracked, untracked);
1598 return 0;
1599 }
1600
1601 /*
1602 * prep_exclude will be called eventually on this directory,
1603 * but it's called much later in last_exclude_matching(). We
1604 * need it now to determine the validity of the cache for this
1605 * path. The next calls will be nearly no-op, the way
1606 * prep_exclude() is designed.
1607 */
1608 if (path->len && path->buf[path->len - 1] != '/') {
1609 strbuf_addch(path, '/');
1610 prep_exclude(dir, path->buf, path->len);
1611 strbuf_setlen(path, path->len - 1);
1612 } else
1613 prep_exclude(dir, path->buf, path->len);
1614
1615 /* hopefully prep_exclude() haven't invalidated this entry... */
1616 return untracked->valid;
1617}
1618
cf7c6148
NTND
1619static int open_cached_dir(struct cached_dir *cdir,
1620 struct dir_struct *dir,
1621 struct untracked_cache_dir *untracked,
1622 struct strbuf *path,
1623 int check_only)
1624{
1625 memset(cdir, 0, sizeof(*cdir));
1626 cdir->untracked = untracked;
91a2288b
NTND
1627 if (valid_cached_dir(dir, untracked, path, check_only))
1628 return 0;
cf7c6148 1629 cdir->fdir = opendir(path->len ? path->buf : ".");
91a2288b
NTND
1630 if (dir->untracked)
1631 dir->untracked->dir_opened++;
cf7c6148
NTND
1632 if (!cdir->fdir)
1633 return -1;
1634 return 0;
1635}
1636
1637static int read_cached_dir(struct cached_dir *cdir)
1638{
1639 if (cdir->fdir) {
1640 cdir->de = readdir(cdir->fdir);
1641 if (!cdir->de)
1642 return -1;
1643 return 0;
1644 }
91a2288b
NTND
1645 while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
1646 struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
26cb0182
NTND
1647 if (!d->recurse) {
1648 cdir->nr_dirs++;
1649 continue;
1650 }
91a2288b
NTND
1651 cdir->ucd = d;
1652 cdir->nr_dirs++;
1653 return 0;
1654 }
1655 cdir->ucd = NULL;
1656 if (cdir->nr_files < cdir->untracked->untracked_nr) {
1657 struct untracked_cache_dir *d = cdir->untracked;
1658 cdir->file = d->untracked[cdir->nr_files++];
1659 return 0;
1660 }
cf7c6148
NTND
1661 return -1;
1662}
1663
1664static void close_cached_dir(struct cached_dir *cdir)
1665{
1666 if (cdir->fdir)
1667 closedir(cdir->fdir);
91a2288b
NTND
1668 /*
1669 * We have gone through this directory and found no untracked
1670 * entries. Mark it valid.
1671 */
26cb0182 1672 if (cdir->untracked) {
91a2288b 1673 cdir->untracked->valid = 1;
26cb0182
NTND
1674 cdir->untracked->recurse = 1;
1675 }
16e2cfa9
JH
1676}
1677
453ec4bd
LT
1678/*
1679 * Read a directory tree. We currently ignore anything but
1680 * directories, regular files and symlinks. That's because git
1681 * doesn't handle them at all yet. Maybe that will change some
1682 * day.
1683 *
1684 * Also, we ignore the name ".git" (even if it is not a directory).
1685 * That likely will not change.
defd7c7b
KB
1686 *
1687 * Returns the most significant path_treatment value encountered in the scan.
453ec4bd 1688 */
defd7c7b 1689static enum path_treatment read_directory_recursive(struct dir_struct *dir,
53cc5356 1690 const char *base, int baselen,
0dcb8d7f 1691 struct untracked_cache_dir *untracked, int check_only,
53cc5356 1692 const struct path_simplify *simplify)
453ec4bd 1693{
cf7c6148 1694 struct cached_dir cdir;
defd7c7b 1695 enum path_treatment state, subdir_state, dir_state = path_none;
bef36921 1696 struct strbuf path = STRBUF_INIT;
453ec4bd 1697
bef36921 1698 strbuf_add(&path, base, baselen);
02cb6753 1699
cf7c6148 1700 if (open_cached_dir(&cdir, dir, untracked, &path, check_only))
1528d247
RS
1701 goto out;
1702
0dcb8d7f
NTND
1703 if (untracked)
1704 untracked->check_only = !!check_only;
1705
cf7c6148 1706 while (!read_cached_dir(&cdir)) {
defd7c7b 1707 /* check how the file or directory should be treated */
cf7c6148 1708 state = treat_path(dir, untracked, &cdir, &path, baselen, simplify);
0dcb8d7f 1709
defd7c7b
KB
1710 if (state > dir_state)
1711 dir_state = state;
1712
1713 /* recurse into subdir if instructed by treat_path */
1714 if (state == path_recurse) {
0dcb8d7f
NTND
1715 struct untracked_cache_dir *ud;
1716 ud = lookup_untracked(dir->untracked, untracked,
1717 path.buf + baselen,
1718 path.len - baselen);
1719 subdir_state =
1720 read_directory_recursive(dir, path.buf, path.len,
1721 ud, check_only, simplify);
defd7c7b
KB
1722 if (subdir_state > dir_state)
1723 dir_state = subdir_state;
1724 }
1725
1726 if (check_only) {
1727 /* abort early if maximum state has been reached */
0dcb8d7f 1728 if (dir_state == path_untracked) {
91a2288b 1729 if (cdir.fdir)
0dcb8d7f 1730 add_untracked(untracked, path.buf + baselen);
defd7c7b 1731 break;
0dcb8d7f 1732 }
defd7c7b 1733 /* skip the dir_add_* part */
02cb6753 1734 continue;
453ec4bd 1735 }
defd7c7b
KB
1736
1737 /* add the path to the appropriate result list */
1738 switch (state) {
1739 case path_excluded:
1740 if (dir->flags & DIR_SHOW_IGNORED)
1741 dir_add_name(dir, path.buf, path.len);
0aaf62b6
KB
1742 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1743 ((dir->flags & DIR_COLLECT_IGNORED) &&
1744 exclude_matches_pathspec(path.buf, path.len,
1745 simplify)))
1746 dir_add_ignored(dir, path.buf, path.len);
defd7c7b
KB
1747 break;
1748
1749 case path_untracked:
0dcb8d7f
NTND
1750 if (dir->flags & DIR_SHOW_IGNORED)
1751 break;
1752 dir_add_name(dir, path.buf, path.len);
91a2288b 1753 if (cdir.fdir)
0dcb8d7f 1754 add_untracked(untracked, path.buf + baselen);
1528d247 1755 break;
defd7c7b
KB
1756
1757 default:
1758 break;
1759 }
453ec4bd 1760 }
cf7c6148 1761 close_cached_dir(&cdir);
1528d247 1762 out:
bef36921 1763 strbuf_release(&path);
453ec4bd 1764
defd7c7b 1765 return dir_state;
453ec4bd
LT
1766}
1767
1768static int cmp_name(const void *p1, const void *p2)
1769{
1770 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1771 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1772
ccdd4a0f 1773 return name_compare(e1->name, e1->len, e2->name, e2->len);
453ec4bd
LT
1774}
1775
9fc42d60
LT
1776static struct path_simplify *create_simplify(const char **pathspec)
1777{
1778 int nr, alloc = 0;
1779 struct path_simplify *simplify = NULL;
1780
1781 if (!pathspec)
1782 return NULL;
1783
1784 for (nr = 0 ; ; nr++) {
1785 const char *match;
9af49f82 1786 ALLOC_GROW(simplify, nr + 1, alloc);
9fc42d60
LT
1787 match = *pathspec++;
1788 if (!match)
1789 break;
1790 simplify[nr].path = match;
1791 simplify[nr].len = simple_length(match);
1792 }
1793 simplify[nr].path = NULL;
1794 simplify[nr].len = 0;
1795 return simplify;
1796}
1797
1798static void free_simplify(struct path_simplify *simplify)
1799{
8e0f7003 1800 free(simplify);
9fc42d60
LT
1801}
1802
48ffef96
JH
1803static int treat_leading_path(struct dir_struct *dir,
1804 const char *path, int len,
1805 const struct path_simplify *simplify)
1806{
49dc2cc2
RS
1807 struct strbuf sb = STRBUF_INIT;
1808 int baselen, rc = 0;
48ffef96 1809 const char *cp;
be8a84c5 1810 int old_flags = dir->flags;
48ffef96
JH
1811
1812 while (len && path[len - 1] == '/')
1813 len--;
1814 if (!len)
1815 return 1;
1816 baselen = 0;
be8a84c5 1817 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
48ffef96
JH
1818 while (1) {
1819 cp = path + baselen + !!baselen;
1820 cp = memchr(cp, '/', path + len - cp);
1821 if (!cp)
1822 baselen = len;
1823 else
1824 baselen = cp - path;
49dc2cc2
RS
1825 strbuf_setlen(&sb, 0);
1826 strbuf_add(&sb, path, baselen);
1827 if (!is_directory(sb.buf))
1828 break;
1829 if (simplify_away(sb.buf, sb.len, simplify))
1830 break;
2e5910f2 1831 if (treat_one_path(dir, NULL, &sb, baselen, simplify,
defd7c7b 1832 DT_DIR, NULL) == path_none)
49dc2cc2
RS
1833 break; /* do not recurse into it */
1834 if (len <= baselen) {
1835 rc = 1;
1836 break; /* finished checking */
1837 }
48ffef96 1838 }
49dc2cc2 1839 strbuf_release(&sb);
be8a84c5 1840 dir->flags = old_flags;
49dc2cc2 1841 return rc;
48ffef96
JH
1842}
1843
1e8fef60
NTND
1844static const char *get_ident_string(void)
1845{
1846 static struct strbuf sb = STRBUF_INIT;
1847 struct utsname uts;
1848
1849 if (sb.len)
1850 return sb.buf;
100e4337 1851 if (uname(&uts) < 0)
1e8fef60
NTND
1852 die_errno(_("failed to get kernel name and information"));
1853 strbuf_addf(&sb, "Location %s, system %s %s %s", get_git_work_tree(),
1854 uts.sysname, uts.release, uts.version);
1855 return sb.buf;
1856}
1857
1858static int ident_in_untracked(const struct untracked_cache *uc)
1859{
1860 const char *end = uc->ident.buf + uc->ident.len;
1861 const char *p = uc->ident.buf;
1862
1863 for (p = uc->ident.buf; p < end; p += strlen(p) + 1)
1864 if (!strcmp(p, get_ident_string()))
1865 return 1;
1866 return 0;
1867}
1868
1869void add_untracked_ident(struct untracked_cache *uc)
1870{
1871 if (ident_in_untracked(uc))
1872 return;
1873 strbuf_addstr(&uc->ident, get_ident_string());
1874 /* this strbuf contains a list of strings, save NUL too */
1875 strbuf_addch(&uc->ident, 0);
1876}
1877
ccad261f
NTND
1878static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
1879 int base_len,
1880 const struct pathspec *pathspec)
1881{
1882 struct untracked_cache_dir *root;
1883
76e6b090 1884 if (!dir->untracked || getenv("GIT_DISABLE_UNTRACKED_CACHE"))
ccad261f
NTND
1885 return NULL;
1886
1887 /*
1888 * We only support $GIT_DIR/info/exclude and core.excludesfile
1889 * as the global ignore rule files. Any other additions
1890 * (e.g. from command line) invalidate the cache. This
1891 * condition also catches running setup_standard_excludes()
1892 * before setting dir->untracked!
1893 */
1894 if (dir->unmanaged_exclude_files)
1895 return NULL;
1896
1897 /*
1898 * Optimize for the main use case only: whole-tree git
1899 * status. More work involved in treat_leading_path() if we
1900 * use cache on just a subset of the worktree. pathspec
1901 * support could make the matter even worse.
1902 */
1903 if (base_len || (pathspec && pathspec->nr))
1904 return NULL;
1905
1906 /* Different set of flags may produce different results */
1907 if (dir->flags != dir->untracked->dir_flags ||
1908 /*
1909 * See treat_directory(), case index_nonexistent. Without
1910 * this flag, we may need to also cache .git file content
1911 * for the resolve_gitlink_ref() call, which we don't.
1912 */
1913 !(dir->flags & DIR_SHOW_OTHER_DIRECTORIES) ||
1914 /* We don't support collecting ignore files */
1915 (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
1916 DIR_COLLECT_IGNORED)))
1917 return NULL;
1918
1919 /*
1920 * If we use .gitignore in the cache and now you change it to
1921 * .gitexclude, everything will go wrong.
1922 */
1923 if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
1924 strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
1925 return NULL;
1926
1927 /*
1928 * EXC_CMDL is not considered in the cache. If people set it,
1929 * skip the cache.
1930 */
1931 if (dir->exclude_list_group[EXC_CMDL].nr)
1932 return NULL;
1933
1e8fef60
NTND
1934 if (!ident_in_untracked(dir->untracked)) {
1935 warning(_("Untracked cache is disabled on this system."));
1936 return NULL;
1937 }
1938
ccad261f
NTND
1939 if (!dir->untracked->root) {
1940 const int len = sizeof(*dir->untracked->root);
1941 dir->untracked->root = xmalloc(len);
1942 memset(dir->untracked->root, 0, len);
1943 }
1944
1945 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
1946 root = dir->untracked->root;
1947 if (hashcmp(dir->ss_info_exclude.sha1,
1948 dir->untracked->ss_info_exclude.sha1)) {
1949 invalidate_gitignore(dir->untracked, root);
1950 dir->untracked->ss_info_exclude = dir->ss_info_exclude;
1951 }
1952 if (hashcmp(dir->ss_excludes_file.sha1,
1953 dir->untracked->ss_excludes_file.sha1)) {
1954 invalidate_gitignore(dir->untracked, root);
1955 dir->untracked->ss_excludes_file = dir->ss_excludes_file;
1956 }
26cb0182
NTND
1957
1958 /* Make sure this directory is not dropped out at saving phase */
1959 root->recurse = 1;
ccad261f
NTND
1960 return root;
1961}
1962
7327d3d1 1963int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
9fc42d60 1964{
725b0605 1965 struct path_simplify *simplify;
ccad261f 1966 struct untracked_cache_dir *untracked;
b4189aa8 1967
7327d3d1
NTND
1968 /*
1969 * Check out create_simplify()
1970 */
1971 if (pathspec)
5c6933d2
NTND
1972 GUARD_PATHSPEC(pathspec,
1973 PATHSPEC_FROMTOP |
1974 PATHSPEC_MAXDEPTH |
bd30c2e4 1975 PATHSPEC_LITERAL |
93d93537 1976 PATHSPEC_GLOB |
ef79b1f8
NTND
1977 PATHSPEC_ICASE |
1978 PATHSPEC_EXCLUDE);
7327d3d1 1979
dba2e203 1980 if (has_symlink_leading_path(path, len))
725b0605
JH
1981 return dir->nr;
1982
ef79b1f8
NTND
1983 /*
1984 * exclude patterns are treated like positive ones in
1985 * create_simplify. Usually exclude patterns should be a
1986 * subset of positive ones, which has no impacts on
1987 * create_simplify().
1988 */
b3920bbd 1989 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
ccad261f
NTND
1990 untracked = validate_untracked_cache(dir, len, pathspec);
1991 if (!untracked)
1992 /*
1993 * make sure untracked cache code path is disabled,
1994 * e.g. prep_exclude()
1995 */
1996 dir->untracked = NULL;
48ffef96 1997 if (!len || treat_leading_path(dir, path, len, simplify))
ccad261f 1998 read_directory_recursive(dir, path, len, untracked, 0, simplify);
9fc42d60 1999 free_simplify(simplify);
453ec4bd 2000 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
2abd31b0 2001 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
c9ccb5d3
NTND
2002 if (dir->untracked) {
2003 static struct trace_key trace_untracked_stats = TRACE_KEY_INIT(UNTRACKED_STATS);
2004 trace_printf_key(&trace_untracked_stats,
2005 "node creation: %u\n"
2006 "gitignore invalidation: %u\n"
2007 "directory invalidation: %u\n"
2008 "opendir: %u\n",
2009 dir->untracked->dir_created,
2010 dir->untracked->gitignore_invalidated,
2011 dir->untracked->dir_invalidated,
2012 dir->untracked->dir_opened);
1bbb3dba
NTND
2013 if (dir->untracked == the_index.untracked &&
2014 (dir->untracked->dir_opened ||
2015 dir->untracked->gitignore_invalidated ||
2016 dir->untracked->dir_invalidated))
2017 the_index.cache_changed |= UNTRACKED_CHANGED;
2018 if (dir->untracked != the_index.untracked) {
2019 free(dir->untracked);
2020 dir->untracked = NULL;
2021 }
c9ccb5d3 2022 }
453ec4bd
LT
2023 return dir->nr;
2024}
c91f0d92 2025
686a4a06 2026int file_exists(const char *f)
c91f0d92 2027{
686a4a06 2028 struct stat sb;
a50f9fc5 2029 return lstat(f, &sb) == 0;
c91f0d92 2030}
e6636747
JS
2031
2032/*
9b125da4
NTND
2033 * Given two normalized paths (a trailing slash is ok), if subdir is
2034 * outside dir, return -1. Otherwise return the offset in subdir that
2035 * can be used as relative path to dir.
e6636747 2036 */
9b125da4 2037int dir_inside_of(const char *subdir, const char *dir)
e6636747 2038{
9b125da4 2039 int offset = 0;
e6636747 2040
9b125da4 2041 assert(dir && subdir && *dir && *subdir);
e6636747 2042
9b125da4 2043 while (*dir && *subdir && *dir == *subdir) {
e6636747 2044 dir++;
9b125da4
NTND
2045 subdir++;
2046 offset++;
490544b1 2047 }
9b125da4
NTND
2048
2049 /* hel[p]/me vs hel[l]/yeah */
2050 if (*dir && *subdir)
2051 return -1;
2052
2053 if (!*subdir)
2054 return !*dir ? offset : -1; /* same dir */
2055
2056 /* foo/[b]ar vs foo/[] */
2057 if (is_dir_sep(dir[-1]))
2058 return is_dir_sep(subdir[-1]) ? offset : -1;
2059
2060 /* foo[/]bar vs foo[] */
2061 return is_dir_sep(*subdir) ? offset + 1 : -1;
e6636747
JS
2062}
2063
2064int is_inside_dir(const char *dir)
2065{
56b9f6e7
RS
2066 char *cwd;
2067 int rc;
2068
b892913d
NTND
2069 if (!dir)
2070 return 0;
56b9f6e7
RS
2071
2072 cwd = xgetcwd();
2073 rc = (dir_inside_of(cwd, dir) >= 0);
2074 free(cwd);
2075 return rc;
e6636747 2076}
7155b727 2077
55892d23
AP
2078int is_empty_dir(const char *path)
2079{
2080 DIR *dir = opendir(path);
2081 struct dirent *e;
2082 int ret = 1;
2083
2084 if (!dir)
2085 return 0;
2086
2087 while ((e = readdir(dir)) != NULL)
2088 if (!is_dot_or_dotdot(e->d_name)) {
2089 ret = 0;
2090 break;
2091 }
2092
2093 closedir(dir);
2094 return ret;
2095}
2096
ae2f203e 2097static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
7155b727 2098{
a0f4afbe 2099 DIR *dir;
7155b727 2100 struct dirent *e;
ae2f203e 2101 int ret = 0, original_len = path->len, len, kept_down = 0;
a0f4afbe 2102 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
c844a803 2103 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
a0f4afbe 2104 unsigned char submodule_head[20];
7155b727 2105
a0f4afbe 2106 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
ae2f203e 2107 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
a0f4afbe 2108 /* Do not descend and nuke a nested git work tree. */
ae2f203e
JH
2109 if (kept_up)
2110 *kept_up = 1;
a0f4afbe 2111 return 0;
ae2f203e 2112 }
a0f4afbe 2113
ae2f203e 2114 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
a0f4afbe 2115 dir = opendir(path->buf);
c844a803 2116 if (!dir) {
863808cd
MH
2117 if (errno == ENOENT)
2118 return keep_toplevel ? -1 : 0;
2119 else if (errno == EACCES && !keep_toplevel)
ecb2c282
MH
2120 /*
2121 * An empty dir could be removable even if it
2122 * is unreadable:
2123 */
c844a803
JH
2124 return rmdir(path->buf);
2125 else
2126 return -1;
2127 }
00b6c178 2128 strbuf_complete(path, '/');
7155b727
JS
2129
2130 len = path->len;
2131 while ((e = readdir(dir)) != NULL) {
2132 struct stat st;
8ca12c0d
AP
2133 if (is_dot_or_dotdot(e->d_name))
2134 continue;
7155b727
JS
2135
2136 strbuf_setlen(path, len);
2137 strbuf_addstr(path, e->d_name);
863808cd
MH
2138 if (lstat(path->buf, &st)) {
2139 if (errno == ENOENT)
2140 /*
2141 * file disappeared, which is what we
2142 * wanted anyway
2143 */
2144 continue;
2145 /* fall thru */
2146 } else if (S_ISDIR(st.st_mode)) {
ae2f203e 2147 if (!remove_dir_recurse(path, flag, &kept_down))
7155b727 2148 continue; /* happy */
863808cd
MH
2149 } else if (!only_empty &&
2150 (!unlink(path->buf) || errno == ENOENT)) {
7155b727 2151 continue; /* happy, too */
863808cd 2152 }
7155b727
JS
2153
2154 /* path too long, stat fails, or non-directory still exists */
2155 ret = -1;
2156 break;
2157 }
2158 closedir(dir);
2159
2160 strbuf_setlen(path, original_len);
ae2f203e 2161 if (!ret && !keep_toplevel && !kept_down)
863808cd 2162 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
ae2f203e
JH
2163 else if (kept_up)
2164 /*
2165 * report the uplevel that it is not an error that we
2166 * did not rmdir() our directory.
2167 */
2168 *kept_up = !ret;
7155b727
JS
2169 return ret;
2170}
039bc64e 2171
ae2f203e
JH
2172int remove_dir_recursively(struct strbuf *path, int flag)
2173{
2174 return remove_dir_recurse(path, flag, NULL);
2175}
2176
f932729c
JK
2177static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
2178
039bc64e
JH
2179void setup_standard_excludes(struct dir_struct *dir)
2180{
2181 const char *path;
2182
2183 dir->exclude_per_dir = ".gitignore";
099d2d86
JH
2184
2185 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
2845ce7f
PT
2186 if (!excludes_file)
2187 excludes_file = xdg_config_home("ignore");
4698c8fe 2188 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
38ccaf93
JH
2189 add_excludes_from_file_1(dir, excludes_file,
2190 dir->untracked ? &dir->ss_excludes_file : NULL);
099d2d86
JH
2191
2192 /* per repository user preference */
f932729c 2193 path = git_path_info_exclude();
099d2d86 2194 if (!access_or_warn(path, R_OK, 0))
0dcb8d7f
NTND
2195 add_excludes_from_file_1(dir, path,
2196 dir->untracked ? &dir->ss_info_exclude : NULL);
039bc64e 2197}
4a92d1bf
AR
2198
2199int remove_path(const char *name)
2200{
2201 char *slash;
2202
9a6728d4 2203 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
4a92d1bf
AR
2204 return -1;
2205
2206 slash = strrchr(name, '/');
2207 if (slash) {
2208 char *dirs = xstrdup(name);
2209 slash = dirs + (slash - name);
2210 do {
2211 *slash = '\0';
3fc0d131 2212 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
4a92d1bf
AR
2213 free(dirs);
2214 }
2215 return 0;
2216}
2217
270be816
AS
2218/*
2219 * Frees memory within dir which was allocated for exclude lists and
2220 * the exclude_stack. Does not free dir itself.
2221 */
2222void clear_directory(struct dir_struct *dir)
2223{
2224 int i, j;
2225 struct exclude_list_group *group;
2226 struct exclude_list *el;
2227 struct exclude_stack *stk;
2228
2229 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
2230 group = &dir->exclude_list_group[i];
2231 for (j = 0; j < group->nr; j++) {
2232 el = &group->el[j];
2233 if (i == EXC_DIRS)
2234 free((char *)el->src);
2235 clear_exclude_list(el);
2236 }
2237 free(group->el);
2238 }
2239
2240 stk = dir->exclude_stack;
2241 while (stk) {
2242 struct exclude_stack *prev = stk->prev;
2243 free(stk);
2244 stk = prev;
2245 }
aceb9429 2246 strbuf_release(&dir->basebuf);
270be816 2247}
83c094ad
NTND
2248
2249struct ondisk_untracked_cache {
2250 struct stat_data info_exclude_stat;
2251 struct stat_data excludes_file_stat;
2252 uint32_t dir_flags;
2253 unsigned char info_exclude_sha1[20];
2254 unsigned char excludes_file_sha1[20];
2255 char exclude_per_dir[FLEX_ARRAY];
2256};
2257
2258#define ouc_size(len) (offsetof(struct ondisk_untracked_cache, exclude_per_dir) + len + 1)
2259
2260struct write_data {
2261 int index; /* number of written untracked_cache_dir */
2262 struct ewah_bitmap *check_only; /* from untracked_cache_dir */
2263 struct ewah_bitmap *valid; /* from untracked_cache_dir */
2264 struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
2265 struct strbuf out;
2266 struct strbuf sb_stat;
2267 struct strbuf sb_sha1;
2268};
2269
2270static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
2271{
2272 to->sd_ctime.sec = htonl(from->sd_ctime.sec);
2273 to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
2274 to->sd_mtime.sec = htonl(from->sd_mtime.sec);
2275 to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
2276 to->sd_dev = htonl(from->sd_dev);
2277 to->sd_ino = htonl(from->sd_ino);
2278 to->sd_uid = htonl(from->sd_uid);
2279 to->sd_gid = htonl(from->sd_gid);
2280 to->sd_size = htonl(from->sd_size);
2281}
2282
2283static void write_one_dir(struct untracked_cache_dir *untracked,
2284 struct write_data *wd)
2285{
2286 struct stat_data stat_data;
2287 struct strbuf *out = &wd->out;
2288 unsigned char intbuf[16];
2289 unsigned int intlen, value;
2290 int i = wd->index++;
2291
2292 /*
2293 * untracked_nr should be reset whenever valid is clear, but
2294 * for safety..
2295 */
2296 if (!untracked->valid) {
2297 untracked->untracked_nr = 0;
2298 untracked->check_only = 0;
2299 }
2300
2301 if (untracked->check_only)
2302 ewah_set(wd->check_only, i);
2303 if (untracked->valid) {
2304 ewah_set(wd->valid, i);
2305 stat_data_to_disk(&stat_data, &untracked->stat_data);
2306 strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
2307 }
2308 if (!is_null_sha1(untracked->exclude_sha1)) {
2309 ewah_set(wd->sha1_valid, i);
2310 strbuf_add(&wd->sb_sha1, untracked->exclude_sha1, 20);
2311 }
2312
2313 intlen = encode_varint(untracked->untracked_nr, intbuf);
2314 strbuf_add(out, intbuf, intlen);
2315
2316 /* skip non-recurse directories */
2317 for (i = 0, value = 0; i < untracked->dirs_nr; i++)
2318 if (untracked->dirs[i]->recurse)
2319 value++;
2320 intlen = encode_varint(value, intbuf);
2321 strbuf_add(out, intbuf, intlen);
2322
2323 strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
2324
2325 for (i = 0; i < untracked->untracked_nr; i++)
2326 strbuf_add(out, untracked->untracked[i],
2327 strlen(untracked->untracked[i]) + 1);
2328
2329 for (i = 0; i < untracked->dirs_nr; i++)
2330 if (untracked->dirs[i]->recurse)
2331 write_one_dir(untracked->dirs[i], wd);
2332}
2333
2334void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
2335{
2336 struct ondisk_untracked_cache *ouc;
2337 struct write_data wd;
2338 unsigned char varbuf[16];
2339 int len = 0, varint_len;
2340 if (untracked->exclude_per_dir)
2341 len = strlen(untracked->exclude_per_dir);
2342 ouc = xmalloc(sizeof(*ouc) + len + 1);
2343 stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
2344 stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
2345 hashcpy(ouc->info_exclude_sha1, untracked->ss_info_exclude.sha1);
2346 hashcpy(ouc->excludes_file_sha1, untracked->ss_excludes_file.sha1);
2347 ouc->dir_flags = htonl(untracked->dir_flags);
2348 memcpy(ouc->exclude_per_dir, untracked->exclude_per_dir, len + 1);
1e8fef60
NTND
2349
2350 varint_len = encode_varint(untracked->ident.len, varbuf);
2351 strbuf_add(out, varbuf, varint_len);
2352 strbuf_add(out, untracked->ident.buf, untracked->ident.len);
2353
83c094ad
NTND
2354 strbuf_add(out, ouc, ouc_size(len));
2355 free(ouc);
2356 ouc = NULL;
2357
2358 if (!untracked->root) {
2359 varint_len = encode_varint(0, varbuf);
2360 strbuf_add(out, varbuf, varint_len);
2361 return;
2362 }
2363
2364 wd.index = 0;
2365 wd.check_only = ewah_new();
2366 wd.valid = ewah_new();
2367 wd.sha1_valid = ewah_new();
2368 strbuf_init(&wd.out, 1024);
2369 strbuf_init(&wd.sb_stat, 1024);
2370 strbuf_init(&wd.sb_sha1, 1024);
2371 write_one_dir(untracked->root, &wd);
2372
2373 varint_len = encode_varint(wd.index, varbuf);
2374 strbuf_add(out, varbuf, varint_len);
2375 strbuf_addbuf(out, &wd.out);
2376 ewah_serialize_strbuf(wd.valid, out);
2377 ewah_serialize_strbuf(wd.check_only, out);
2378 ewah_serialize_strbuf(wd.sha1_valid, out);
2379 strbuf_addbuf(out, &wd.sb_stat);
2380 strbuf_addbuf(out, &wd.sb_sha1);
2381 strbuf_addch(out, '\0'); /* safe guard for string lists */
2382
2383 ewah_free(wd.valid);
2384 ewah_free(wd.check_only);
2385 ewah_free(wd.sha1_valid);
2386 strbuf_release(&wd.out);
2387 strbuf_release(&wd.sb_stat);
2388 strbuf_release(&wd.sb_sha1);
2389}
f9e6c649
NTND
2390
2391static void free_untracked(struct untracked_cache_dir *ucd)
2392{
2393 int i;
2394 if (!ucd)
2395 return;
2396 for (i = 0; i < ucd->dirs_nr; i++)
2397 free_untracked(ucd->dirs[i]);
2398 for (i = 0; i < ucd->untracked_nr; i++)
2399 free(ucd->untracked[i]);
2400 free(ucd->untracked);
2401 free(ucd->dirs);
2402 free(ucd);
2403}
2404
2405void free_untracked_cache(struct untracked_cache *uc)
2406{
2407 if (uc)
2408 free_untracked(uc->root);
2409 free(uc);
2410}
2411
2412struct read_data {
2413 int index;
2414 struct untracked_cache_dir **ucd;
2415 struct ewah_bitmap *check_only;
2416 struct ewah_bitmap *valid;
2417 struct ewah_bitmap *sha1_valid;
2418 const unsigned char *data;
2419 const unsigned char *end;
2420};
2421
2422static void stat_data_from_disk(struct stat_data *to, const struct stat_data *from)
2423{
2424 to->sd_ctime.sec = get_be32(&from->sd_ctime.sec);
2425 to->sd_ctime.nsec = get_be32(&from->sd_ctime.nsec);
2426 to->sd_mtime.sec = get_be32(&from->sd_mtime.sec);
2427 to->sd_mtime.nsec = get_be32(&from->sd_mtime.nsec);
2428 to->sd_dev = get_be32(&from->sd_dev);
2429 to->sd_ino = get_be32(&from->sd_ino);
2430 to->sd_uid = get_be32(&from->sd_uid);
2431 to->sd_gid = get_be32(&from->sd_gid);
2432 to->sd_size = get_be32(&from->sd_size);
2433}
2434
2435static int read_one_dir(struct untracked_cache_dir **untracked_,
2436 struct read_data *rd)
2437{
2438 struct untracked_cache_dir ud, *untracked;
2439 const unsigned char *next, *data = rd->data, *end = rd->end;
2440 unsigned int value;
2441 int i, len;
2442
2443 memset(&ud, 0, sizeof(ud));
2444
2445 next = data;
2446 value = decode_varint(&next);
2447 if (next > end)
2448 return -1;
2449 ud.recurse = 1;
2450 ud.untracked_alloc = value;
2451 ud.untracked_nr = value;
2452 if (ud.untracked_nr)
2453 ud.untracked = xmalloc(sizeof(*ud.untracked) * ud.untracked_nr);
2454 data = next;
2455
2456 next = data;
2457 ud.dirs_alloc = ud.dirs_nr = decode_varint(&next);
2458 if (next > end)
2459 return -1;
2460 ud.dirs = xmalloc(sizeof(*ud.dirs) * ud.dirs_nr);
2461 data = next;
2462
2463 len = strlen((const char *)data);
2464 next = data + len + 1;
2465 if (next > rd->end)
2466 return -1;
2467 *untracked_ = untracked = xmalloc(sizeof(*untracked) + len);
2468 memcpy(untracked, &ud, sizeof(ud));
2469 memcpy(untracked->name, data, len + 1);
2470 data = next;
2471
2472 for (i = 0; i < untracked->untracked_nr; i++) {
2473 len = strlen((const char *)data);
2474 next = data + len + 1;
2475 if (next > rd->end)
2476 return -1;
2477 untracked->untracked[i] = xstrdup((const char*)data);
2478 data = next;
2479 }
2480
2481 rd->ucd[rd->index++] = untracked;
2482 rd->data = data;
2483
2484 for (i = 0; i < untracked->dirs_nr; i++) {
2485 len = read_one_dir(untracked->dirs + i, rd);
2486 if (len < 0)
2487 return -1;
2488 }
2489 return 0;
2490}
2491
2492static void set_check_only(size_t pos, void *cb)
2493{
2494 struct read_data *rd = cb;
2495 struct untracked_cache_dir *ud = rd->ucd[pos];
2496 ud->check_only = 1;
2497}
2498
2499static void read_stat(size_t pos, void *cb)
2500{
2501 struct read_data *rd = cb;
2502 struct untracked_cache_dir *ud = rd->ucd[pos];
2503 if (rd->data + sizeof(struct stat_data) > rd->end) {
2504 rd->data = rd->end + 1;
2505 return;
2506 }
2507 stat_data_from_disk(&ud->stat_data, (struct stat_data *)rd->data);
2508 rd->data += sizeof(struct stat_data);
2509 ud->valid = 1;
2510}
2511
2512static void read_sha1(size_t pos, void *cb)
2513{
2514 struct read_data *rd = cb;
2515 struct untracked_cache_dir *ud = rd->ucd[pos];
2516 if (rd->data + 20 > rd->end) {
2517 rd->data = rd->end + 1;
2518 return;
2519 }
2520 hashcpy(ud->exclude_sha1, rd->data);
2521 rd->data += 20;
2522}
2523
2524static void load_sha1_stat(struct sha1_stat *sha1_stat,
2525 const struct stat_data *stat,
2526 const unsigned char *sha1)
2527{
2528 stat_data_from_disk(&sha1_stat->stat, stat);
2529 hashcpy(sha1_stat->sha1, sha1);
2530 sha1_stat->valid = 1;
2531}
2532
2533struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
2534{
2535 const struct ondisk_untracked_cache *ouc;
2536 struct untracked_cache *uc;
2537 struct read_data rd;
2538 const unsigned char *next = data, *end = (const unsigned char *)data + sz;
1e8fef60
NTND
2539 const char *ident;
2540 int ident_len, len;
f9e6c649
NTND
2541
2542 if (sz <= 1 || end[-1] != '\0')
2543 return NULL;
2544 end--;
2545
1e8fef60
NTND
2546 ident_len = decode_varint(&next);
2547 if (next + ident_len > end)
2548 return NULL;
2549 ident = (const char *)next;
2550 next += ident_len;
2551
f9e6c649
NTND
2552 ouc = (const struct ondisk_untracked_cache *)next;
2553 if (next + ouc_size(0) > end)
2554 return NULL;
2555
2556 uc = xcalloc(1, sizeof(*uc));
1e8fef60
NTND
2557 strbuf_init(&uc->ident, ident_len);
2558 strbuf_add(&uc->ident, ident, ident_len);
f9e6c649
NTND
2559 load_sha1_stat(&uc->ss_info_exclude, &ouc->info_exclude_stat,
2560 ouc->info_exclude_sha1);
2561 load_sha1_stat(&uc->ss_excludes_file, &ouc->excludes_file_stat,
2562 ouc->excludes_file_sha1);
2563 uc->dir_flags = get_be32(&ouc->dir_flags);
2564 uc->exclude_per_dir = xstrdup(ouc->exclude_per_dir);
2565 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2566 next += ouc_size(strlen(ouc->exclude_per_dir));
2567 if (next >= end)
2568 goto done2;
2569
2570 len = decode_varint(&next);
2571 if (next > end || len == 0)
2572 goto done2;
2573
2574 rd.valid = ewah_new();
2575 rd.check_only = ewah_new();
2576 rd.sha1_valid = ewah_new();
2577 rd.data = next;
2578 rd.end = end;
2579 rd.index = 0;
2580 rd.ucd = xmalloc(sizeof(*rd.ucd) * len);
2581
2582 if (read_one_dir(&uc->root, &rd) || rd.index != len)
2583 goto done;
2584
2585 next = rd.data;
2586 len = ewah_read_mmap(rd.valid, next, end - next);
2587 if (len < 0)
2588 goto done;
2589
2590 next += len;
2591 len = ewah_read_mmap(rd.check_only, next, end - next);
2592 if (len < 0)
2593 goto done;
2594
2595 next += len;
2596 len = ewah_read_mmap(rd.sha1_valid, next, end - next);
2597 if (len < 0)
2598 goto done;
2599
2600 ewah_each_bit(rd.check_only, set_check_only, &rd);
2601 rd.data = next + len;
2602 ewah_each_bit(rd.valid, read_stat, &rd);
2603 ewah_each_bit(rd.sha1_valid, read_sha1, &rd);
2604 next = rd.data;
2605
2606done:
2607 free(rd.ucd);
2608 ewah_free(rd.valid);
2609 ewah_free(rd.check_only);
2610 ewah_free(rd.sha1_valid);
2611done2:
2612 if (next != end) {
2613 free_untracked_cache(uc);
2614 uc = NULL;
2615 }
2616 return uc;
2617}
e931371a 2618
73f9145f
NTND
2619static void invalidate_one_directory(struct untracked_cache *uc,
2620 struct untracked_cache_dir *ucd)
2621{
2622 uc->dir_invalidated++;
2623 ucd->valid = 0;
2624 ucd->untracked_nr = 0;
2625}
2626
2627/*
2628 * Normally when an entry is added or removed from a directory,
2629 * invalidating that directory is enough. No need to touch its
2630 * ancestors. When a directory is shown as "foo/bar/" in git-status
2631 * however, deleting or adding an entry may have cascading effect.
2632 *
2633 * Say the "foo/bar/file" has become untracked, we need to tell the
2634 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2635 * directory any more (because "bar" is managed by foo as an untracked
2636 * "file").
2637 *
2638 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2639 * was the last untracked entry in the entire "foo", we should show
2640 * "foo/" instead. Which means we have to invalidate past "bar" up to
2641 * "foo".
2642 *
2643 * This function traverses all directories from root to leaf. If there
2644 * is a chance of one of the above cases happening, we invalidate back
2645 * to root. Otherwise we just invalidate the leaf. There may be a more
2646 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2647 * detect these cases and avoid unnecessary invalidation, for example,
2648 * checking for the untracked entry named "bar/" in "foo", but for now
2649 * stick to something safe and simple.
2650 */
2651static int invalidate_one_component(struct untracked_cache *uc,
2652 struct untracked_cache_dir *dir,
2653 const char *path, int len)
2654{
2655 const char *rest = strchr(path, '/');
2656
2657 if (rest) {
2658 int component_len = rest - path;
2659 struct untracked_cache_dir *d =
2660 lookup_untracked(uc, dir, path, component_len);
2661 int ret =
2662 invalidate_one_component(uc, d, rest + 1,
2663 len - (component_len + 1));
2664 if (ret)
2665 invalidate_one_directory(uc, dir);
2666 return ret;
2667 }
2668
2669 invalidate_one_directory(uc, dir);
2670 return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
2671}
2672
e931371a
NTND
2673void untracked_cache_invalidate_path(struct index_state *istate,
2674 const char *path)
2675{
e931371a
NTND
2676 if (!istate->untracked || !istate->untracked->root)
2677 return;
73f9145f
NTND
2678 invalidate_one_component(istate->untracked, istate->untracked->root,
2679 path, strlen(path));
e931371a
NTND
2680}
2681
2682void untracked_cache_remove_from_index(struct index_state *istate,
2683 const char *path)
2684{
2685 untracked_cache_invalidate_path(istate, path);
2686}
2687
2688void untracked_cache_add_to_index(struct index_state *istate,
2689 const char *path)
2690{
2691 untracked_cache_invalidate_path(istate, path);
2692}