git

1
/*
2
 * This handles recursive filename detection with exclude
3
 * files, index knowledge etc..
4
 *
5
 * Copyright (C) Linus Torvalds, 2005-2006
6
 *		 Junio Hamano, 2005-2006
7
 */
8

9
#define USE_THE_REPOSITORY_VARIABLE
10

11
#include "git-compat-util.h"
12
#include "abspath.h"
13
#include "config.h"
14
#include "convert.h"
15
#include "dir.h"
16
#include "environment.h"
17
#include "gettext.h"
18
#include "name-hash.h"
19
#include "object-file.h"
20
#include "object-store-ll.h"
21
#include "path.h"
22
#include "refs.h"
23
#include "wildmatch.h"
24
#include "pathspec.h"
25
#include "utf8.h"
26
#include "varint.h"
27
#include "ewah/ewok.h"
28
#include "fsmonitor-ll.h"
29
#include "read-cache-ll.h"
30
#include "setup.h"
31
#include "sparse-index.h"
32
#include "submodule-config.h"
33
#include "symlinks.h"
34
#include "trace2.h"
35
#include "tree.h"
36
#include "hex.h"
37

38
 /*
39
  * The maximum size of a pattern/exclude file. If the file exceeds this size
40
  * we will ignore it.
41
  */
42
#define PATTERN_MAX_FILE_SIZE (100 * 1024 * 1024)
43

44
/*
45
 * Tells read_directory_recursive how a file or directory should be treated.
46
 * Values are ordered by significance, e.g. if a directory contains both
47
 * excluded and untracked files, it is listed as untracked because
48
 * path_untracked > path_excluded.
49
 */
50
enum path_treatment {
51
	path_none = 0,
52
	path_recurse,
53
	path_excluded,
54
	path_untracked
55
};
56

57
/*
58
 * Support data structure for our opendir/readdir/closedir wrappers
59
 */
60
struct cached_dir {
61
	DIR *fdir;
62
	struct untracked_cache_dir *untracked;
63
	int nr_files;
64
	int nr_dirs;
65

66
	const char *d_name;
67
	int d_type;
68
	const char *file;
69
	struct untracked_cache_dir *ucd;
70
};
71

72
static enum path_treatment read_directory_recursive(struct dir_struct *dir,
73
	struct index_state *istate, const char *path, int len,
74
	struct untracked_cache_dir *untracked,
75
	int check_only, int stop_at_first_file, const struct pathspec *pathspec);
76
static int resolve_dtype(int dtype, struct index_state *istate,
77
			 const char *path, int len);
78
struct dirent *readdir_skip_dot_and_dotdot(DIR *dirp)
79
{
80
	struct dirent *e;
81

82
	while ((e = readdir(dirp)) != NULL) {
83
		if (!is_dot_or_dotdot(e->d_name))
84
			break;
85
	}
86
	return e;
87
}
88

89
int count_slashes(const char *s)
90
{
91
	int cnt = 0;
92
	while (*s)
93
		if (*s++ == '/')
94
			cnt++;
95
	return cnt;
96
}
97

98
int git_fspathcmp(const char *a, const char *b)
99
{
100
	return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
101
}
102

103
int fspatheq(const char *a, const char *b)
104
{
105
	return !fspathcmp(a, b);
106
}
107

108
int git_fspathncmp(const char *a, const char *b, size_t count)
109
{
110
	return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
111
}
112

113
int paths_collide(const char *a, const char *b)
114
{
115
	size_t len_a = strlen(a), len_b = strlen(b);
116

117
	if (len_a == len_b)
118
		return fspatheq(a, b);
119

120
	if (len_a < len_b)
121
		return is_dir_sep(b[len_a]) && !fspathncmp(a, b, len_a);
122
	return is_dir_sep(a[len_b]) && !fspathncmp(a, b, len_b);
123
}
124

125
unsigned int fspathhash(const char *str)
126
{
127
	return ignore_case ? strihash(str) : strhash(str);
128
}
129

130
int git_fnmatch(const struct pathspec_item *item,
131
		const char *pattern, const char *string,
132
		int prefix)
133
{
134
	if (prefix > 0) {
135
		if (ps_strncmp(item, pattern, string, prefix))
136
			return WM_NOMATCH;
137
		pattern += prefix;
138
		string += prefix;
139
	}
140
	if (item->flags & PATHSPEC_ONESTAR) {
141
		int pattern_len = strlen(++pattern);
142
		int string_len = strlen(string);
143
		return string_len < pattern_len ||
144
			ps_strcmp(item, pattern,
145
				  string + string_len - pattern_len);
146
	}
147
	if (item->magic & PATHSPEC_GLOB)
148
		return wildmatch(pattern, string,
149
				 WM_PATHNAME |
150
				 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0));
151
	else
152
		/* wildmatch has not learned no FNM_PATHNAME mode yet */
153
		return wildmatch(pattern, string,
154
				 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0);
155
}
156

157
static int fnmatch_icase_mem(const char *pattern, int patternlen,
158
			     const char *string, int stringlen,
159
			     int flags)
160
{
161
	int match_status;
162
	struct strbuf pat_buf = STRBUF_INIT;
163
	struct strbuf str_buf = STRBUF_INIT;
164
	const char *use_pat = pattern;
165
	const char *use_str = string;
166

167
	if (pattern[patternlen]) {
168
		strbuf_add(&pat_buf, pattern, patternlen);
169
		use_pat = pat_buf.buf;
170
	}
171
	if (string[stringlen]) {
172
		strbuf_add(&str_buf, string, stringlen);
173
		use_str = str_buf.buf;
174
	}
175

176
	if (ignore_case)
177
		flags |= WM_CASEFOLD;
178
	match_status = wildmatch(use_pat, use_str, flags);
179

180
	strbuf_release(&pat_buf);
181
	strbuf_release(&str_buf);
182

183
	return match_status;
184
}
185

186
static size_t common_prefix_len(const struct pathspec *pathspec)
187
{
188
	int n;
189
	size_t max = 0;
190

191
	/*
192
	 * ":(icase)path" is treated as a pathspec full of
193
	 * wildcard. In other words, only prefix is considered common
194
	 * prefix. If the pathspec is abc/foo abc/bar, running in
195
	 * subdir xyz, the common prefix is still xyz, not xyz/abc as
196
	 * in non-:(icase).
197
	 */
198
	GUARD_PATHSPEC(pathspec,
199
		       PATHSPEC_FROMTOP |
200
		       PATHSPEC_MAXDEPTH |
201
		       PATHSPEC_LITERAL |
202
		       PATHSPEC_GLOB |
203
		       PATHSPEC_ICASE |
204
		       PATHSPEC_EXCLUDE |
205
		       PATHSPEC_ATTR);
206

207
	for (n = 0; n < pathspec->nr; n++) {
208
		size_t i = 0, len = 0, item_len;
209
		if (pathspec->items[n].magic & PATHSPEC_EXCLUDE)
210
			continue;
211
		if (pathspec->items[n].magic & PATHSPEC_ICASE)
212
			item_len = pathspec->items[n].prefix;
213
		else
214
			item_len = pathspec->items[n].nowildcard_len;
215
		while (i < item_len && (n == 0 || i < max)) {
216
			char c = pathspec->items[n].match[i];
217
			if (c != pathspec->items[0].match[i])
218
				break;
219
			if (c == '/')
220
				len = i + 1;
221
			i++;
222
		}
223
		if (n == 0 || len < max) {
224
			max = len;
225
			if (!max)
226
				break;
227
		}
228
	}
229
	return max;
230
}
231

232
/*
233
 * Returns a copy of the longest leading path common among all
234
 * pathspecs.
235
 */
236
char *common_prefix(const struct pathspec *pathspec)
237
{
238
	unsigned long len = common_prefix_len(pathspec);
239

240
	return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
241
}
242

243
int fill_directory(struct dir_struct *dir,
244
		   struct index_state *istate,
245
		   const struct pathspec *pathspec)
246
{
247
	const char *prefix;
248
	size_t prefix_len;
249

250
	unsigned exclusive_flags = DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO;
251
	if ((dir->flags & exclusive_flags) == exclusive_flags)
252
		BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
253

254
	/*
255
	 * Calculate common prefix for the pathspec, and
256
	 * use that to optimize the directory walk
257
	 */
258
	prefix_len = common_prefix_len(pathspec);
259
	prefix = prefix_len ? pathspec->items[0].match : "";
260

261
	/* Read the directory and prune it */
262
	read_directory(dir, istate, prefix, prefix_len, pathspec);
263

264
	return prefix_len;
265
}
266

267
int within_depth(const char *name, int namelen,
268
			int depth, int max_depth)
269
{
270
	const char *cp = name, *cpe = name + namelen;
271

272
	while (cp < cpe) {
273
		if (*cp++ != '/')
274
			continue;
275
		depth++;
276
		if (depth > max_depth)
277
			return 0;
278
	}
279
	return 1;
280
}
281

282
/*
283
 * Read the contents of the blob with the given OID into a buffer.
284
 * Append a trailing LF to the end if the last line doesn't have one.
285
 *
286
 * Returns:
287
 *    -1 when the OID is invalid or unknown or does not refer to a blob.
288
 *     0 when the blob is empty.
289
 *     1 along with { data, size } of the (possibly augmented) buffer
290
 *       when successful.
291
 *
292
 * Optionally updates the given oid_stat with the given OID (when valid).
293
 */
294
static int do_read_blob(const struct object_id *oid, struct oid_stat *oid_stat,
295
			size_t *size_out, char **data_out)
296
{
297
	enum object_type type;
298
	unsigned long sz;
299
	char *data;
300

301
	*size_out = 0;
302
	*data_out = NULL;
303

304
	data = repo_read_object_file(the_repository, oid, &type, &sz);
305
	if (!data || type != OBJ_BLOB) {
306
		free(data);
307
		return -1;
308
	}
309

310
	if (oid_stat) {
311
		memset(&oid_stat->stat, 0, sizeof(oid_stat->stat));
312
		oidcpy(&oid_stat->oid, oid);
313
	}
314

315
	if (sz == 0) {
316
		free(data);
317
		return 0;
318
	}
319

320
	if (data[sz - 1] != '\n') {
321
		data = xrealloc(data, st_add(sz, 1));
322
		data[sz++] = '\n';
323
	}
324

325
	*size_out = xsize_t(sz);
326
	*data_out = data;
327

328
	return 1;
329
}
330

331
#define DO_MATCH_EXCLUDE   (1<<0)
332
#define DO_MATCH_DIRECTORY (1<<1)
333
#define DO_MATCH_LEADING_PATHSPEC (1<<2)
334

335
/*
336
 * Does the given pathspec match the given name?  A match is found if
337
 *
338
 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
339
 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
340
 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
341
 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
342
 *
343
 * Return value tells which case it was (1-4), or 0 when there is no match.
344
 *
345
 * It may be instructive to look at a small table of concrete examples
346
 * to understand the differences between 1, 2, and 4:
347
 *
348
 *                              Pathspecs
349
 *                |    a/b    |   a/b/    |   a/b/c
350
 *          ------+-----------+-----------+------------
351
 *          a/b   |  EXACT    |  EXACT[1] | LEADING[2]
352
 *  Names   a/b/  | RECURSIVE |   EXACT   | LEADING[2]
353
 *          a/b/c | RECURSIVE | RECURSIVE |   EXACT
354
 *
355
 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
356
 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
357
 */
358
static int match_pathspec_item(struct index_state *istate,
359
			       const struct pathspec_item *item, int prefix,
360
			       const char *name, int namelen, unsigned flags)
361
{
362
	/* name/namelen has prefix cut off by caller */
363
	const char *match = item->match + prefix;
364
	int matchlen = item->len - prefix;
365

366
	/*
367
	 * The normal call pattern is:
368
	 * 1. prefix = common_prefix_len(ps);
369
	 * 2. prune something, or fill_directory
370
	 * 3. match_pathspec()
371
	 *
372
	 * 'prefix' at #1 may be shorter than the command's prefix and
373
	 * it's ok for #2 to match extra files. Those extras will be
374
	 * trimmed at #3.
375
	 *
376
	 * Suppose the pathspec is 'foo' and '../bar' running from
377
	 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
378
	 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
379
	 * user does not want XYZ/foo, only the "foo" part should be
380
	 * case-insensitive. We need to filter out XYZ/foo here. In
381
	 * other words, we do not trust the caller on comparing the
382
	 * prefix part when :(icase) is involved. We do exact
383
	 * comparison ourselves.
384
	 *
385
	 * Normally the caller (common_prefix_len() in fact) does
386
	 * _exact_ matching on name[-prefix+1..-1] and we do not need
387
	 * to check that part. Be defensive and check it anyway, in
388
	 * case common_prefix_len is changed, or a new caller is
389
	 * introduced that does not use common_prefix_len.
390
	 *
391
	 * If the penalty turns out too high when prefix is really
392
	 * long, maybe change it to
393
	 * strncmp(match, name, item->prefix - prefix)
394
	 */
395
	if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
396
	    strncmp(item->match, name - prefix, item->prefix))
397
		return 0;
398

399
	if (item->attr_match_nr &&
400
	    !match_pathspec_attrs(istate, name - prefix, namelen + prefix, item))
401
		return 0;
402

403
	/* If the match was just the prefix, we matched */
404
	if (!*match)
405
		return MATCHED_RECURSIVELY;
406

407
	if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
408
		if (matchlen == namelen)
409
			return MATCHED_EXACTLY;
410

411
		if (match[matchlen-1] == '/' || name[matchlen] == '/')
412
			return MATCHED_RECURSIVELY;
413
	} else if ((flags & DO_MATCH_DIRECTORY) &&
414
		   match[matchlen - 1] == '/' &&
415
		   namelen == matchlen - 1 &&
416
		   !ps_strncmp(item, match, name, namelen))
417
		return MATCHED_EXACTLY;
418

419
	if (item->nowildcard_len < item->len &&
420
	    !git_fnmatch(item, match, name,
421
			 item->nowildcard_len - prefix))
422
		return MATCHED_FNMATCH;
423

424
	/* Perform checks to see if "name" is a leading string of the pathspec */
425
	if ( (flags & DO_MATCH_LEADING_PATHSPEC) &&
426
	    !(flags & DO_MATCH_EXCLUDE)) {
427
		/* name is a literal prefix of the pathspec */
428
		int offset = name[namelen-1] == '/' ? 1 : 0;
429
		if ((namelen < matchlen) &&
430
		    (match[namelen-offset] == '/') &&
431
		    !ps_strncmp(item, match, name, namelen))
432
			return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
433

434
		/* name doesn't match up to the first wild character */
435
		if (item->nowildcard_len < item->len &&
436
		    ps_strncmp(item, match, name,
437
			       item->nowildcard_len - prefix))
438
			return 0;
439

440
		/*
441
		 * name has no wildcard, and it didn't match as a leading
442
		 * pathspec so return.
443
		 */
444
		if (item->nowildcard_len == item->len)
445
			return 0;
446

447
		/*
448
		 * Here is where we would perform a wildmatch to check if
449
		 * "name" can be matched as a directory (or a prefix) against
450
		 * the pathspec.  Since wildmatch doesn't have this capability
451
		 * at the present we have to punt and say that it is a match,
452
		 * potentially returning a false positive
453
		 * The submodules themselves will be able to perform more
454
		 * accurate matching to determine if the pathspec matches.
455
		 */
456
		return MATCHED_RECURSIVELY_LEADING_PATHSPEC;
457
	}
458

459
	return 0;
460
}
461

462
/*
463
 * do_match_pathspec() is meant to ONLY be called by
464
 * match_pathspec_with_flags(); calling it directly risks pathspecs
465
 * like ':!unwanted_path' being ignored.
466
 *
467
 * Given a name and a list of pathspecs, returns the nature of the
468
 * closest (i.e. most specific) match of the name to any of the
469
 * pathspecs.
470
 *
471
 * The caller typically calls this multiple times with the same
472
 * pathspec and seen[] array but with different name/namelen
473
 * (e.g. entries from the index) and is interested in seeing if and
474
 * how each pathspec matches all the names it calls this function
475
 * with.  A mark is left in the seen[] array for each pathspec element
476
 * indicating the closest type of match that element achieved, so if
477
 * seen[n] remains zero after multiple invocations, that means the nth
478
 * pathspec did not match any names, which could indicate that the
479
 * user mistyped the nth pathspec.
480
 */
481
static int do_match_pathspec(struct index_state *istate,
482
			     const struct pathspec *ps,
483
			     const char *name, int namelen,
484
			     int prefix, char *seen,
485
			     unsigned flags)
486
{
487
	int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
488

489
	GUARD_PATHSPEC(ps,
490
		       PATHSPEC_FROMTOP |
491
		       PATHSPEC_MAXDEPTH |
492
		       PATHSPEC_LITERAL |
493
		       PATHSPEC_GLOB |
494
		       PATHSPEC_ICASE |
495
		       PATHSPEC_EXCLUDE |
496
		       PATHSPEC_ATTR);
497

498
	if (!ps->nr) {
499
		if (!ps->recursive ||
500
		    !(ps->magic & PATHSPEC_MAXDEPTH) ||
501
		    ps->max_depth == -1)
502
			return MATCHED_RECURSIVELY;
503

504
		if (within_depth(name, namelen, 0, ps->max_depth))
505
			return MATCHED_EXACTLY;
506
		else
507
			return 0;
508
	}
509

510
	name += prefix;
511
	namelen -= prefix;
512

513
	for (i = ps->nr - 1; i >= 0; i--) {
514
		int how;
515

516
		if ((!exclude &&   ps->items[i].magic & PATHSPEC_EXCLUDE) ||
517
		    ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
518
			continue;
519

520
		if (seen && seen[i] == MATCHED_EXACTLY)
521
			continue;
522
		/*
523
		 * Make exclude patterns optional and never report
524
		 * "pathspec ':(exclude)foo' matches no files"
525
		 */
526
		if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
527
			seen[i] = MATCHED_FNMATCH;
528
		how = match_pathspec_item(istate, ps->items+i, prefix, name,
529
					  namelen, flags);
530
		if (ps->recursive &&
531
		    (ps->magic & PATHSPEC_MAXDEPTH) &&
532
		    ps->max_depth != -1 &&
533
		    how && how != MATCHED_FNMATCH) {
534
			int len = ps->items[i].len;
535
			if (name[len] == '/')
536
				len++;
537
			if (within_depth(name+len, namelen-len, 0, ps->max_depth))
538
				how = MATCHED_EXACTLY;
539
			else
540
				how = 0;
541
		}
542
		if (how) {
543
			if (retval < how)
544
				retval = how;
545
			if (seen && seen[i] < how)
546
				seen[i] = how;
547
		}
548
	}
549
	return retval;
550
}
551

552
static int match_pathspec_with_flags(struct index_state *istate,
553
				     const struct pathspec *ps,
554
				     const char *name, int namelen,
555
				     int prefix, char *seen, unsigned flags)
556
{
557
	int positive, negative;
558
	positive = do_match_pathspec(istate, ps, name, namelen,
559
				     prefix, seen, flags);
560
	if (!(ps->magic & PATHSPEC_EXCLUDE) || !positive)
561
		return positive;
562
	negative = do_match_pathspec(istate, ps, name, namelen,
563
				     prefix, seen,
564
				     flags | DO_MATCH_EXCLUDE);
565
	return negative ? 0 : positive;
566
}
567

568
int match_pathspec(struct index_state *istate,
569
		   const struct pathspec *ps,
570
		   const char *name, int namelen,
571
		   int prefix, char *seen, int is_dir)
572
{
573
	unsigned flags = is_dir ? DO_MATCH_DIRECTORY : 0;
574
	return match_pathspec_with_flags(istate, ps, name, namelen,
575
					 prefix, seen, flags);
576
}
577

578
/**
579
 * Check if a submodule is a superset of the pathspec
580
 */
581
int submodule_path_match(struct index_state *istate,
582
			 const struct pathspec *ps,
583
			 const char *submodule_name,
584
			 char *seen)
585
{
586
	int matched = match_pathspec_with_flags(istate, ps, submodule_name,
587
						strlen(submodule_name),
588
						0, seen,
589
						DO_MATCH_DIRECTORY |
590
						DO_MATCH_LEADING_PATHSPEC);
591
	return matched;
592
}
593

594
int report_path_error(const char *ps_matched,
595
		      const struct pathspec *pathspec)
596
{
597
	/*
598
	 * Make sure all pathspec matched; otherwise it is an error.
599
	 */
600
	int num, errors = 0;
601
	for (num = 0; num < pathspec->nr; num++) {
602
		int other, found_dup;
603

604
		if (ps_matched[num])
605
			continue;
606
		/*
607
		 * The caller might have fed identical pathspec
608
		 * twice.  Do not barf on such a mistake.
609
		 * FIXME: parse_pathspec should have eliminated
610
		 * duplicate pathspec.
611
		 */
612
		for (found_dup = other = 0;
613
		     !found_dup && other < pathspec->nr;
614
		     other++) {
615
			if (other == num || !ps_matched[other])
616
				continue;
617
			if (!strcmp(pathspec->items[other].original,
618
				    pathspec->items[num].original))
619
				/*
620
				 * Ok, we have a match already.
621
				 */
622
				found_dup = 1;
623
		}
624
		if (found_dup)
625
			continue;
626

627
		error(_("pathspec '%s' did not match any file(s) known to git"),
628
		      pathspec->items[num].original);
629
		errors++;
630
	}
631
	return errors;
632
}
633

634
/*
635
 * Return the length of the "simple" part of a path match limiter.
636
 */
637
int simple_length(const char *match)
638
{
639
	int len = -1;
640

641
	for (;;) {
642
		unsigned char c = *match++;
643
		len++;
644
		if (c == '\0' || is_glob_special(c))
645
			return len;
646
	}
647
}
648

649
int no_wildcard(const char *string)
650
{
651
	return string[simple_length(string)] == '\0';
652
}
653

654
void parse_path_pattern(const char **pattern,
655
			   int *patternlen,
656
			   unsigned *flags,
657
			   int *nowildcardlen)
658
{
659
	const char *p = *pattern;
660
	size_t i, len;
661

662
	*flags = 0;
663
	if (*p == '!') {
664
		*flags |= PATTERN_FLAG_NEGATIVE;
665
		p++;
666
	}
667
	len = strlen(p);
668
	if (len && p[len - 1] == '/') {
669
		len--;
670
		*flags |= PATTERN_FLAG_MUSTBEDIR;
671
	}
672
	for (i = 0; i < len; i++) {
673
		if (p[i] == '/')
674
			break;
675
	}
676
	if (i == len)
677
		*flags |= PATTERN_FLAG_NODIR;
678
	*nowildcardlen = simple_length(p);
679
	/*
680
	 * we should have excluded the trailing slash from 'p' too,
681
	 * but that's one more allocation. Instead just make sure
682
	 * nowildcardlen does not exceed real patternlen
683
	 */
684
	if (*nowildcardlen > len)
685
		*nowildcardlen = len;
686
	if (*p == '*' && no_wildcard(p + 1))
687
		*flags |= PATTERN_FLAG_ENDSWITH;
688
	*pattern = p;
689
	*patternlen = len;
690
}
691

692
int pl_hashmap_cmp(const void *cmp_data UNUSED,
693
		   const struct hashmap_entry *a,
694
		   const struct hashmap_entry *b,
695
		   const void *key UNUSED)
696
{
697
	const struct pattern_entry *ee1 =
698
			container_of(a, struct pattern_entry, ent);
699
	const struct pattern_entry *ee2 =
700
			container_of(b, struct pattern_entry, ent);
701

702
	size_t min_len = ee1->patternlen <= ee2->patternlen
703
			 ? ee1->patternlen
704
			 : ee2->patternlen;
705

706
	return fspathncmp(ee1->pattern, ee2->pattern, min_len);
707
}
708

709
static char *dup_and_filter_pattern(const char *pattern)
710
{
711
	char *set, *read;
712
	size_t count  = 0;
713
	char *result = xstrdup(pattern);
714

715
	set = result;
716
	read = result;
717

718
	while (*read) {
719
		/* skip escape characters (once) */
720
		if (*read == '\\')
721
			read++;
722

723
		*set = *read;
724

725
		set++;
726
		read++;
727
		count++;
728
	}
729
	*set = 0;
730

731
	if (count > 2 &&
732
	    *(set - 1) == '*' &&
733
	    *(set - 2) == '/')
734
		*(set - 2) = 0;
735

736
	return result;
737
}
738

739
static void clear_pattern_entry_hashmap(struct hashmap *map)
740
{
741
	struct hashmap_iter iter;
742
	struct pattern_entry *entry;
743

744
	hashmap_for_each_entry(map, &iter, entry, ent) {
745
		free(entry->pattern);
746
	}
747
	hashmap_clear_and_free(map, struct pattern_entry, ent);
748
}
749

750
static void add_pattern_to_hashsets(struct pattern_list *pl, struct path_pattern *given)
751
{
752
	struct pattern_entry *translated;
753
	char *truncated;
754
	char *data = NULL;
755
	const char *prev, *cur, *next;
756

757
	if (!pl->use_cone_patterns)
758
		return;
759

760
	if (given->flags & PATTERN_FLAG_NEGATIVE &&
761
	    given->flags & PATTERN_FLAG_MUSTBEDIR &&
762
	    !strcmp(given->pattern, "/*")) {
763
		pl->full_cone = 0;
764
		return;
765
	}
766

767
	if (!given->flags && !strcmp(given->pattern, "/*")) {
768
		pl->full_cone = 1;
769
		return;
770
	}
771

772
	if (given->patternlen < 2 ||
773
	    *given->pattern != '/' ||
774
	    strstr(given->pattern, "**")) {
775
		/* Not a cone pattern. */
776
		warning(_("unrecognized pattern: '%s'"), given->pattern);
777
		goto clear_hashmaps;
778
	}
779

780
	if (!(given->flags & PATTERN_FLAG_MUSTBEDIR) &&
781
	    strcmp(given->pattern, "/*")) {
782
		/* Not a cone pattern. */
783
		warning(_("unrecognized pattern: '%s'"), given->pattern);
784
		goto clear_hashmaps;
785
	}
786

787
	prev = given->pattern;
788
	cur = given->pattern + 1;
789
	next = given->pattern + 2;
790

791
	while (*cur) {
792
		/* Watch for glob characters '*', '\', '[', '?' */
793
		if (!is_glob_special(*cur))
794
			goto increment;
795

796
		/* But only if *prev != '\\' */
797
		if (*prev == '\\')
798
			goto increment;
799

800
		/* But allow the initial '\' */
801
		if (*cur == '\\' &&
802
		    is_glob_special(*next))
803
			goto increment;
804

805
		/* But a trailing '/' then '*' is fine */
806
		if (*prev == '/' &&
807
		    *cur == '*' &&
808
		    *next == 0)
809
			goto increment;
810

811
		/* Not a cone pattern. */
812
		warning(_("unrecognized pattern: '%s'"), given->pattern);
813
		goto clear_hashmaps;
814

815
	increment:
816
		prev++;
817
		cur++;
818
		next++;
819
	}
820

821
	if (given->patternlen > 2 &&
822
	    !strcmp(given->pattern + given->patternlen - 2, "/*")) {
823
		struct pattern_entry *old;
824

825
		if (!(given->flags & PATTERN_FLAG_NEGATIVE)) {
826
			/* Not a cone pattern. */
827
			warning(_("unrecognized pattern: '%s'"), given->pattern);
828
			goto clear_hashmaps;
829
		}
830

831
		truncated = dup_and_filter_pattern(given->pattern);
832

833
		translated = xmalloc(sizeof(struct pattern_entry));
834
		translated->pattern = truncated;
835
		translated->patternlen = given->patternlen - 2;
836
		hashmap_entry_init(&translated->ent,
837
				   fspathhash(translated->pattern));
838

839
		if (!hashmap_get_entry(&pl->recursive_hashmap,
840
				       translated, ent, NULL)) {
841
			/* We did not see the "parent" included */
842
			warning(_("unrecognized negative pattern: '%s'"),
843
				given->pattern);
844
			free(truncated);
845
			free(translated);
846
			goto clear_hashmaps;
847
		}
848

849
		hashmap_add(&pl->parent_hashmap, &translated->ent);
850
		old = hashmap_remove_entry(&pl->recursive_hashmap, translated, ent, &data);
851
		if (old) {
852
			free(old->pattern);
853
			free(old);
854
		}
855
		free(data);
856
		return;
857
	}
858

859
	if (given->flags & PATTERN_FLAG_NEGATIVE) {
860
		warning(_("unrecognized negative pattern: '%s'"),
861
			given->pattern);
862
		goto clear_hashmaps;
863
	}
864

865
	translated = xmalloc(sizeof(struct pattern_entry));
866

867
	translated->pattern = dup_and_filter_pattern(given->pattern);
868
	translated->patternlen = given->patternlen;
869
	hashmap_entry_init(&translated->ent,
870
			   fspathhash(translated->pattern));
871

872
	hashmap_add(&pl->recursive_hashmap, &translated->ent);
873

874
	if (hashmap_get_entry(&pl->parent_hashmap, translated, ent, NULL)) {
875
		/* we already included this at the parent level */
876
		warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
877
			given->pattern);
878
		goto clear_hashmaps;
879
	}
880

881
	return;
882

883
clear_hashmaps:
884
	warning(_("disabling cone pattern matching"));
885
	clear_pattern_entry_hashmap(&pl->recursive_hashmap);
886
	clear_pattern_entry_hashmap(&pl->parent_hashmap);
887
	pl->use_cone_patterns = 0;
888
}
889

890
static int hashmap_contains_path(struct hashmap *map,
891
				 struct strbuf *pattern)
892
{
893
	struct pattern_entry p;
894

895
	/* Check straight mapping */
896
	p.pattern = pattern->buf;
897
	p.patternlen = pattern->len;
898
	hashmap_entry_init(&p.ent, fspathhash(p.pattern));
899
	return !!hashmap_get_entry(map, &p, ent, NULL);
900
}
901

902
int hashmap_contains_parent(struct hashmap *map,
903
			    const char *path,
904
			    struct strbuf *buffer)
905
{
906
	char *slash_pos;
907

908
	strbuf_setlen(buffer, 0);
909

910
	if (path[0] != '/')
911
		strbuf_addch(buffer, '/');
912

913
	strbuf_addstr(buffer, path);
914

915
	slash_pos = strrchr(buffer->buf, '/');
916

917
	while (slash_pos > buffer->buf) {
918
		strbuf_setlen(buffer, slash_pos - buffer->buf);
919

920
		if (hashmap_contains_path(map, buffer))
921
			return 1;
922

923
		slash_pos = strrchr(buffer->buf, '/');
924
	}
925

926
	return 0;
927
}
928

929
void add_pattern(const char *string, const char *base,
930
		 int baselen, struct pattern_list *pl, int srcpos)
931
{
932
	struct path_pattern *pattern;
933
	int patternlen;
934
	unsigned flags;
935
	int nowildcardlen;
936

937
	parse_path_pattern(&string, &patternlen, &flags, &nowildcardlen);
938
	FLEX_ALLOC_MEM(pattern, pattern, string, patternlen);
939
	pattern->patternlen = patternlen;
940
	pattern->nowildcardlen = nowildcardlen;
941
	pattern->base = base;
942
	pattern->baselen = baselen;
943
	pattern->flags = flags;
944
	pattern->srcpos = srcpos;
945
	ALLOC_GROW(pl->patterns, pl->nr + 1, pl->alloc);
946
	pl->patterns[pl->nr++] = pattern;
947
	pattern->pl = pl;
948

949
	add_pattern_to_hashsets(pl, pattern);
950
}
951

952
static int read_skip_worktree_file_from_index(struct index_state *istate,
953
					      const char *path,
954
					      size_t *size_out, char **data_out,
955
					      struct oid_stat *oid_stat)
956
{
957
	int pos, len;
958

959
	len = strlen(path);
960
	pos = index_name_pos(istate, path, len);
961
	if (pos < 0)
962
		return -1;
963
	if (!ce_skip_worktree(istate->cache[pos]))
964
		return -1;
965

966
	return do_read_blob(&istate->cache[pos]->oid, oid_stat, size_out, data_out);
967
}
968

969
/*
970
 * Frees memory within pl which was allocated for exclude patterns and
971
 * the file buffer.  Does not free pl itself.
972
 */
973
void clear_pattern_list(struct pattern_list *pl)
974
{
975
	int i;
976

977
	for (i = 0; i < pl->nr; i++)
978
		free(pl->patterns[i]);
979
	free(pl->patterns);
980
	clear_pattern_entry_hashmap(&pl->recursive_hashmap);
981
	clear_pattern_entry_hashmap(&pl->parent_hashmap);
982

983
	memset(pl, 0, sizeof(*pl));
984
}
985

986
static void trim_trailing_spaces(char *buf)
987
{
988
	char *p, *last_space = NULL;
989

990
	for (p = buf; *p; p++)
991
		switch (*p) {
992
		case ' ':
993
			if (!last_space)
994
				last_space = p;
995
			break;
996
		case '\\':
997
			p++;
998
			if (!*p)
999
				return;
1000
			/* fallthrough */
1001
		default:
1002
			last_space = NULL;
1003
		}
1004

1005
	if (last_space)
1006
		*last_space = '\0';
1007
}
1008

1009
/*
1010
 * Given a subdirectory name and "dir" of the current directory,
1011
 * search the subdir in "dir" and return it, or create a new one if it
1012
 * does not exist in "dir".
1013
 *
1014
 * If "name" has the trailing slash, it'll be excluded in the search.
1015
 */
1016
static struct untracked_cache_dir *lookup_untracked(struct untracked_cache *uc,
1017
						    struct untracked_cache_dir *dir,
1018
						    const char *name, int len)
1019
{
1020
	int first, last;
1021
	struct untracked_cache_dir *d;
1022
	if (!dir)
1023
		return NULL;
1024
	if (len && name[len - 1] == '/')
1025
		len--;
1026
	first = 0;
1027
	last = dir->dirs_nr;
1028
	while (last > first) {
1029
		int cmp, next = first + ((last - first) >> 1);
1030
		d = dir->dirs[next];
1031
		cmp = strncmp(name, d->name, len);
1032
		if (!cmp && strlen(d->name) > len)
1033
			cmp = -1;
1034
		if (!cmp)
1035
			return d;
1036
		if (cmp < 0) {
1037
			last = next;
1038
			continue;
1039
		}
1040
		first = next+1;
1041
	}
1042

1043
	uc->dir_created++;
1044
	FLEX_ALLOC_MEM(d, name, name, len);
1045

1046
	ALLOC_GROW(dir->dirs, dir->dirs_nr + 1, dir->dirs_alloc);
1047
	MOVE_ARRAY(dir->dirs + first + 1, dir->dirs + first,
1048
		   dir->dirs_nr - first);
1049
	dir->dirs_nr++;
1050
	dir->dirs[first] = d;
1051
	return d;
1052
}
1053

1054
static void do_invalidate_gitignore(struct untracked_cache_dir *dir)
1055
{
1056
	int i;
1057
	dir->valid = 0;
1058
	dir->untracked_nr = 0;
1059
	for (i = 0; i < dir->dirs_nr; i++)
1060
		do_invalidate_gitignore(dir->dirs[i]);
1061
}
1062

1063
static void invalidate_gitignore(struct untracked_cache *uc,
1064
				 struct untracked_cache_dir *dir)
1065
{
1066
	uc->gitignore_invalidated++;
1067
	do_invalidate_gitignore(dir);
1068
}
1069

1070
static void invalidate_directory(struct untracked_cache *uc,
1071
				 struct untracked_cache_dir *dir)
1072
{
1073
	int i;
1074

1075
	/*
1076
	 * Invalidation increment here is just roughly correct. If
1077
	 * untracked_nr or any of dirs[].recurse is non-zero, we
1078
	 * should increment dir_invalidated too. But that's more
1079
	 * expensive to do.
1080
	 */
1081
	if (dir->valid)
1082
		uc->dir_invalidated++;
1083

1084
	dir->valid = 0;
1085
	dir->untracked_nr = 0;
1086
	for (i = 0; i < dir->dirs_nr; i++)
1087
		dir->dirs[i]->recurse = 0;
1088
}
1089

1090
static int add_patterns_from_buffer(char *buf, size_t size,
1091
				    const char *base, int baselen,
1092
				    struct pattern_list *pl);
1093

1094
/* Flags for add_patterns() */
1095
#define PATTERN_NOFOLLOW (1<<0)
1096

1097
/*
1098
 * Given a file with name "fname", read it (either from disk, or from
1099
 * an index if 'istate' is non-null), parse it and store the
1100
 * exclude rules in "pl".
1101
 *
1102
 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1103
 * stat data from disk (only valid if add_patterns returns zero). If
1104
 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1105
 */
1106
static int add_patterns(const char *fname, const char *base, int baselen,
1107
			struct pattern_list *pl, struct index_state *istate,
1108
			unsigned flags, struct oid_stat *oid_stat)
1109
{
1110
	struct stat st;
1111
	int r;
1112
	int fd;
1113
	size_t size = 0;
1114
	char *buf;
1115

1116
	if (flags & PATTERN_NOFOLLOW)
1117
		fd = open_nofollow(fname, O_RDONLY);
1118
	else
1119
		fd = open(fname, O_RDONLY);
1120

1121
	if (fd < 0 || fstat(fd, &st) < 0) {
1122
		if (fd < 0)
1123
			warn_on_fopen_errors(fname);
1124
		else
1125
			close(fd);
1126
		if (!istate)
1127
			return -1;
1128
		r = read_skip_worktree_file_from_index(istate, fname,
1129
						       &size, &buf,
1130
						       oid_stat);
1131
		if (r != 1)
1132
			return r;
1133
	} else {
1134
		size = xsize_t(st.st_size);
1135
		if (size == 0) {
1136
			if (oid_stat) {
1137
				fill_stat_data(&oid_stat->stat, &st);
1138
				oidcpy(&oid_stat->oid, the_hash_algo->empty_blob);
1139
				oid_stat->valid = 1;
1140
			}
1141
			close(fd);
1142
			return 0;
1143
		}
1144
		buf = xmallocz(size);
1145
		if (read_in_full(fd, buf, size) != size) {
1146
			free(buf);
1147
			close(fd);
1148
			return -1;
1149
		}
1150
		buf[size++] = '\n';
1151
		close(fd);
1152
		if (oid_stat) {
1153
			int pos;
1154
			if (oid_stat->valid &&
1155
			    !match_stat_data_racy(istate, &oid_stat->stat, &st))
1156
				; /* no content change, oid_stat->oid still good */
1157
			else if (istate &&
1158
				 (pos = index_name_pos(istate, fname, strlen(fname))) >= 0 &&
1159
				 !ce_stage(istate->cache[pos]) &&
1160
				 ce_uptodate(istate->cache[pos]) &&
1161
				 !would_convert_to_git(istate, fname))
1162
				oidcpy(&oid_stat->oid,
1163
				       &istate->cache[pos]->oid);
1164
			else
1165
				hash_object_file(the_hash_algo, buf, size,
1166
						 OBJ_BLOB, &oid_stat->oid);
1167
			fill_stat_data(&oid_stat->stat, &st);
1168
			oid_stat->valid = 1;
1169
		}
1170
	}
1171

1172
	if (size > PATTERN_MAX_FILE_SIZE) {
1173
		warning("ignoring excessively large pattern file: %s", fname);
1174
		free(buf);
1175
		return -1;
1176
	}
1177

1178
	add_patterns_from_buffer(buf, size, base, baselen, pl);
1179
	free(buf);
1180
	return 0;
1181
}
1182

1183
static int add_patterns_from_buffer(char *buf, size_t size,
1184
				    const char *base, int baselen,
1185
				    struct pattern_list *pl)
1186
{
1187
	char *orig = buf;
1188
	int i, lineno = 1;
1189
	char *entry;
1190

1191
	hashmap_init(&pl->recursive_hashmap, pl_hashmap_cmp, NULL, 0);
1192
	hashmap_init(&pl->parent_hashmap, pl_hashmap_cmp, NULL, 0);
1193

1194
	if (skip_utf8_bom(&buf, size))
1195
		size -= buf - orig;
1196

1197
	entry = buf;
1198

1199
	for (i = 0; i < size; i++) {
1200
		if (buf[i] == '\n') {
1201
			if (entry != buf + i && entry[0] != '#') {
1202
				buf[i - (i && buf[i-1] == '\r')] = 0;
1203
				trim_trailing_spaces(entry);
1204
				add_pattern(entry, base, baselen, pl, lineno);
1205
			}
1206
			lineno++;
1207
			entry = buf + i + 1;
1208
		}
1209
	}
1210
	return 0;
1211
}
1212

1213
int add_patterns_from_file_to_list(const char *fname, const char *base,
1214
				   int baselen, struct pattern_list *pl,
1215
				   struct index_state *istate,
1216
				   unsigned flags)
1217
{
1218
	return add_patterns(fname, base, baselen, pl, istate, flags, NULL);
1219
}
1220

1221
int add_patterns_from_blob_to_list(
1222
	struct object_id *oid,
1223
	const char *base, int baselen,
1224
	struct pattern_list *pl)
1225
{
1226
	char *buf;
1227
	size_t size;
1228
	int r;
1229

1230
	r = do_read_blob(oid, NULL, &size, &buf);
1231
	if (r != 1)
1232
		return r;
1233

1234
	if (size > PATTERN_MAX_FILE_SIZE) {
1235
		warning("ignoring excessively large pattern blob: %s",
1236
			oid_to_hex(oid));
1237
		free(buf);
1238
		return -1;
1239
	}
1240

1241
	add_patterns_from_buffer(buf, size, base, baselen, pl);
1242
	free(buf);
1243
	return 0;
1244
}
1245

1246
struct pattern_list *add_pattern_list(struct dir_struct *dir,
1247
				      int group_type, const char *src)
1248
{
1249
	struct pattern_list *pl;
1250
	struct exclude_list_group *group;
1251

1252
	group = &dir->internal.exclude_list_group[group_type];
1253
	ALLOC_GROW(group->pl, group->nr + 1, group->alloc);
1254
	pl = &group->pl[group->nr++];
1255
	memset(pl, 0, sizeof(*pl));
1256
	pl->src = src;
1257
	return pl;
1258
}
1259

1260
/*
1261
 * Used to set up core.excludesfile and .git/info/exclude lists.
1262
 */
1263
static void add_patterns_from_file_1(struct dir_struct *dir, const char *fname,
1264
				     struct oid_stat *oid_stat)
1265
{
1266
	struct pattern_list *pl;
1267
	/*
1268
	 * catch setup_standard_excludes() that's called before
1269
	 * dir->untracked is assigned. That function behaves
1270
	 * differently when dir->untracked is non-NULL.
1271
	 */
1272
	if (!dir->untracked)
1273
		dir->internal.unmanaged_exclude_files++;
1274
	pl = add_pattern_list(dir, EXC_FILE, fname);
1275
	if (add_patterns(fname, "", 0, pl, NULL, 0, oid_stat) < 0)
1276
		die(_("cannot use %s as an exclude file"), fname);
1277
}
1278

1279
void add_patterns_from_file(struct dir_struct *dir, const char *fname)
1280
{
1281
	dir->internal.unmanaged_exclude_files++; /* see validate_untracked_cache() */
1282
	add_patterns_from_file_1(dir, fname, NULL);
1283
}
1284

1285
int match_basename(const char *basename, int basenamelen,
1286
		   const char *pattern, int prefix, int patternlen,
1287
		   unsigned flags)
1288
{
1289
	if (prefix == patternlen) {
1290
		if (patternlen == basenamelen &&
1291
		    !fspathncmp(pattern, basename, basenamelen))
1292
			return 1;
1293
	} else if (flags & PATTERN_FLAG_ENDSWITH) {
1294
		/* "*literal" matching against "fooliteral" */
1295
		if (patternlen - 1 <= basenamelen &&
1296
		    !fspathncmp(pattern + 1,
1297
				   basename + basenamelen - (patternlen - 1),
1298
				   patternlen - 1))
1299
			return 1;
1300
	} else {
1301
		if (fnmatch_icase_mem(pattern, patternlen,
1302
				      basename, basenamelen,
1303
				      0) == 0)
1304
			return 1;
1305
	}
1306
	return 0;
1307
}
1308

1309
int match_pathname(const char *pathname, int pathlen,
1310
		   const char *base, int baselen,
1311
		   const char *pattern, int prefix, int patternlen)
1312
{
1313
	const char *name;
1314
	int namelen;
1315

1316
	/*
1317
	 * match with FNM_PATHNAME; the pattern has base implicitly
1318
	 * in front of it.
1319
	 */
1320
	if (*pattern == '/') {
1321
		pattern++;
1322
		patternlen--;
1323
		prefix--;
1324
	}
1325

1326
	/*
1327
	 * baselen does not count the trailing slash. base[] may or
1328
	 * may not end with a trailing slash though.
1329
	 */
1330
	if (pathlen < baselen + 1 ||
1331
	    (baselen && pathname[baselen] != '/') ||
1332
	    fspathncmp(pathname, base, baselen))
1333
		return 0;
1334

1335
	namelen = baselen ? pathlen - baselen - 1 : pathlen;
1336
	name = pathname + pathlen - namelen;
1337

1338
	if (prefix) {
1339
		/*
1340
		 * if the non-wildcard part is longer than the
1341
		 * remaining pathname, surely it cannot match.
1342
		 */
1343
		if (prefix > namelen)
1344
			return 0;
1345

1346
		if (fspathncmp(pattern, name, prefix))
1347
			return 0;
1348
		pattern += prefix;
1349
		patternlen -= prefix;
1350
		name    += prefix;
1351
		namelen -= prefix;
1352

1353
		/*
1354
		 * If the whole pattern did not have a wildcard,
1355
		 * then our prefix match is all we need; we
1356
		 * do not need to call fnmatch at all.
1357
		 */
1358
		if (!patternlen && !namelen)
1359
			return 1;
1360
	}
1361

1362
	return fnmatch_icase_mem(pattern, patternlen,
1363
				 name, namelen,
1364
				 WM_PATHNAME) == 0;
1365
}
1366

1367
/*
1368
 * Scan the given exclude list in reverse to see whether pathname
1369
 * should be ignored.  The first match (i.e. the last on the list), if
1370
 * any, determines the fate.  Returns the exclude_list element which
1371
 * matched, or NULL for undecided.
1372
 */
1373
static struct path_pattern *last_matching_pattern_from_list(const char *pathname,
1374
						       int pathlen,
1375
						       const char *basename,
1376
						       int *dtype,
1377
						       struct pattern_list *pl,
1378
						       struct index_state *istate)
1379
{
1380
	struct path_pattern *res = NULL; /* undecided */
1381
	int i;
1382

1383
	if (!pl->nr)
1384
		return NULL;	/* undefined */
1385

1386
	for (i = pl->nr - 1; 0 <= i; i--) {
1387
		struct path_pattern *pattern = pl->patterns[i];
1388
		const char *exclude = pattern->pattern;
1389
		int prefix = pattern->nowildcardlen;
1390

1391
		if (pattern->flags & PATTERN_FLAG_MUSTBEDIR) {
1392
			*dtype = resolve_dtype(*dtype, istate, pathname, pathlen);
1393
			if (*dtype != DT_DIR)
1394
				continue;
1395
		}
1396

1397
		if (pattern->flags & PATTERN_FLAG_NODIR) {
1398
			if (match_basename(basename,
1399
					   pathlen - (basename - pathname),
1400
					   exclude, prefix, pattern->patternlen,
1401
					   pattern->flags)) {
1402
				res = pattern;
1403
				break;
1404
			}
1405
			continue;
1406
		}
1407

1408
		assert(pattern->baselen == 0 ||
1409
		       pattern->base[pattern->baselen - 1] == '/');
1410
		if (match_pathname(pathname, pathlen,
1411
				   pattern->base,
1412
				   pattern->baselen ? pattern->baselen - 1 : 0,
1413
				   exclude, prefix, pattern->patternlen)) {
1414
			res = pattern;
1415
			break;
1416
		}
1417
	}
1418
	return res;
1419
}
1420

1421
/*
1422
 * Scan the list of patterns to determine if the ordered list
1423
 * of patterns matches on 'pathname'.
1424
 *
1425
 * Return 1 for a match, 0 for not matched and -1 for undecided.
1426
 */
1427
enum pattern_match_result path_matches_pattern_list(
1428
				const char *pathname, int pathlen,
1429
				const char *basename, int *dtype,
1430
				struct pattern_list *pl,
1431
				struct index_state *istate)
1432
{
1433
	struct path_pattern *pattern;
1434
	struct strbuf parent_pathname = STRBUF_INIT;
1435
	int result = NOT_MATCHED;
1436
	size_t slash_pos;
1437

1438
	if (!pl->use_cone_patterns) {
1439
		pattern = last_matching_pattern_from_list(pathname, pathlen, basename,
1440
							dtype, pl, istate);
1441
		if (pattern) {
1442
			if (pattern->flags & PATTERN_FLAG_NEGATIVE)
1443
				return NOT_MATCHED;
1444
			else
1445
				return MATCHED;
1446
		}
1447

1448
		return UNDECIDED;
1449
	}
1450

1451
	if (pl->full_cone)
1452
		return MATCHED;
1453

1454
	strbuf_addch(&parent_pathname, '/');
1455
	strbuf_add(&parent_pathname, pathname, pathlen);
1456

1457
	/*
1458
	 * Directory entries are matched if and only if a file
1459
	 * contained immediately within them is matched. For the
1460
	 * case of a directory entry, modify the path to create
1461
	 * a fake filename within this directory, allowing us to
1462
	 * use the file-base matching logic in an equivalent way.
1463
	 */
1464
	if (parent_pathname.len > 0 &&
1465
	    parent_pathname.buf[parent_pathname.len - 1] == '/') {
1466
		slash_pos = parent_pathname.len - 1;
1467
		strbuf_add(&parent_pathname, "-", 1);
1468
	} else {
1469
		const char *slash_ptr = strrchr(parent_pathname.buf, '/');
1470
		slash_pos = slash_ptr ? slash_ptr - parent_pathname.buf : 0;
1471
	}
1472

1473
	if (hashmap_contains_path(&pl->recursive_hashmap,
1474
				  &parent_pathname)) {
1475
		result = MATCHED_RECURSIVE;
1476
		goto done;
1477
	}
1478

1479
	if (!slash_pos) {
1480
		/* include every file in root */
1481
		result = MATCHED;
1482
		goto done;
1483
	}
1484

1485
	strbuf_setlen(&parent_pathname, slash_pos);
1486

1487
	if (hashmap_contains_path(&pl->parent_hashmap, &parent_pathname)) {
1488
		result = MATCHED;
1489
		goto done;
1490
	}
1491

1492
	if (hashmap_contains_parent(&pl->recursive_hashmap,
1493
				    pathname,
1494
				    &parent_pathname))
1495
		result = MATCHED_RECURSIVE;
1496

1497
done:
1498
	strbuf_release(&parent_pathname);
1499
	return result;
1500
}
1501

1502
int init_sparse_checkout_patterns(struct index_state *istate)
1503
{
1504
	if (!core_apply_sparse_checkout)
1505
		return 1;
1506
	if (istate->sparse_checkout_patterns)
1507
		return 0;
1508

1509
	CALLOC_ARRAY(istate->sparse_checkout_patterns, 1);
1510

1511
	if (get_sparse_checkout_patterns(istate->sparse_checkout_patterns) < 0) {
1512
		FREE_AND_NULL(istate->sparse_checkout_patterns);
1513
		return -1;
1514
	}
1515

1516
	return 0;
1517
}
1518

1519
static int path_in_sparse_checkout_1(const char *path,
1520
				     struct index_state *istate,
1521
				     int require_cone_mode)
1522
{
1523
	int dtype = DT_REG;
1524
	enum pattern_match_result match = UNDECIDED;
1525
	const char *end, *slash;
1526

1527
	/*
1528
	 * We default to accepting a path if the path is empty, there are no
1529
	 * patterns, or the patterns are of the wrong type.
1530
	 */
1531
	if (!*path ||
1532
	    init_sparse_checkout_patterns(istate) ||
1533
	    (require_cone_mode &&
1534
	     !istate->sparse_checkout_patterns->use_cone_patterns))
1535
		return 1;
1536

1537
	/*
1538
	 * If UNDECIDED, use the match from the parent dir (recursively), or
1539
	 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1540
	 * never returns UNDECIDED, so we will execute only one iteration in
1541
	 * this case.
1542
	 */
1543
	for (end = path + strlen(path);
1544
	     end > path && match == UNDECIDED;
1545
	     end = slash) {
1546

1547
		for (slash = end - 1; slash > path && *slash != '/'; slash--)
1548
			; /* do nothing */
1549

1550
		match = path_matches_pattern_list(path, end - path,
1551
				slash > path ? slash + 1 : path, &dtype,
1552
				istate->sparse_checkout_patterns, istate);
1553

1554
		/* We are going to match the parent dir now */
1555
		dtype = DT_DIR;
1556
	}
1557
	return match > 0;
1558
}
1559

1560
int path_in_sparse_checkout(const char *path,
1561
			    struct index_state *istate)
1562
{
1563
	return path_in_sparse_checkout_1(path, istate, 0);
1564
}
1565

1566
int path_in_cone_mode_sparse_checkout(const char *path,
1567
				     struct index_state *istate)
1568
{
1569
	return path_in_sparse_checkout_1(path, istate, 1);
1570
}
1571

1572
static struct path_pattern *last_matching_pattern_from_lists(
1573
		struct dir_struct *dir, struct index_state *istate,
1574
		const char *pathname, int pathlen,
1575
		const char *basename, int *dtype_p)
1576
{
1577
	int i, j;
1578
	struct exclude_list_group *group;
1579
	struct path_pattern *pattern;
1580
	for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1581
		group = &dir->internal.exclude_list_group[i];
1582
		for (j = group->nr - 1; j >= 0; j--) {
1583
			pattern = last_matching_pattern_from_list(
1584
				pathname, pathlen, basename, dtype_p,
1585
				&group->pl[j], istate);
1586
			if (pattern)
1587
				return pattern;
1588
		}
1589
	}
1590
	return NULL;
1591
}
1592

1593
/*
1594
 * Loads the per-directory exclude list for the substring of base
1595
 * which has a char length of baselen.
1596
 */
1597
static void prep_exclude(struct dir_struct *dir,
1598
			 struct index_state *istate,
1599
			 const char *base, int baselen)
1600
{
1601
	struct exclude_list_group *group;
1602
	struct pattern_list *pl;
1603
	struct exclude_stack *stk = NULL;
1604
	struct untracked_cache_dir *untracked;
1605
	int current;
1606

1607
	group = &dir->internal.exclude_list_group[EXC_DIRS];
1608

1609
	/*
1610
	 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1611
	 * which originate from directories not in the prefix of the
1612
	 * path being checked.
1613
	 */
1614
	while ((stk = dir->internal.exclude_stack) != NULL) {
1615
		if (stk->baselen <= baselen &&
1616
		    !strncmp(dir->internal.basebuf.buf, base, stk->baselen))
1617
			break;
1618
		pl = &group->pl[dir->internal.exclude_stack->exclude_ix];
1619
		dir->internal.exclude_stack = stk->prev;
1620
		dir->internal.pattern = NULL;
1621
		free((char *)pl->src); /* see strbuf_detach() below */
1622
		clear_pattern_list(pl);
1623
		free(stk);
1624
		group->nr--;
1625
	}
1626

1627
	/* Skip traversing into sub directories if the parent is excluded */
1628
	if (dir->internal.pattern)
1629
		return;
1630

1631
	/*
1632
	 * Lazy initialization. All call sites currently just
1633
	 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1634
	 * them seems lots of work for little benefit.
1635
	 */
1636
	if (!dir->internal.basebuf.buf)
1637
		strbuf_init(&dir->internal.basebuf, PATH_MAX);
1638

1639
	/* Read from the parent directories and push them down. */
1640
	current = stk ? stk->baselen : -1;
1641
	strbuf_setlen(&dir->internal.basebuf, current < 0 ? 0 : current);
1642
	if (dir->untracked)
1643
		untracked = stk ? stk->ucd : dir->untracked->root;
1644
	else
1645
		untracked = NULL;
1646

1647
	while (current < baselen) {
1648
		const char *cp;
1649
		struct oid_stat oid_stat;
1650

1651
		CALLOC_ARRAY(stk, 1);
1652
		if (current < 0) {
1653
			cp = base;
1654
			current = 0;
1655
		} else {
1656
			cp = strchr(base + current + 1, '/');
1657
			if (!cp)
1658
				die("oops in prep_exclude");
1659
			cp++;
1660
			untracked =
1661
				lookup_untracked(dir->untracked,
1662
						 untracked,
1663
						 base + current,
1664
						 cp - base - current);
1665
		}
1666
		stk->prev = dir->internal.exclude_stack;
1667
		stk->baselen = cp - base;
1668
		stk->exclude_ix = group->nr;
1669
		stk->ucd = untracked;
1670
		pl = add_pattern_list(dir, EXC_DIRS, NULL);
1671
		strbuf_add(&dir->internal.basebuf, base + current, stk->baselen - current);
1672
		assert(stk->baselen == dir->internal.basebuf.len);
1673

1674
		/* Abort if the directory is excluded */
1675
		if (stk->baselen) {
1676
			int dt = DT_DIR;
1677
			dir->internal.basebuf.buf[stk->baselen - 1] = 0;
1678
			dir->internal.pattern = last_matching_pattern_from_lists(dir,
1679
									istate,
1680
				dir->internal.basebuf.buf, stk->baselen - 1,
1681
				dir->internal.basebuf.buf + current, &dt);
1682
			dir->internal.basebuf.buf[stk->baselen - 1] = '/';
1683
			if (dir->internal.pattern &&
1684
			    dir->internal.pattern->flags & PATTERN_FLAG_NEGATIVE)
1685
				dir->internal.pattern = NULL;
1686
			if (dir->internal.pattern) {
1687
				dir->internal.exclude_stack = stk;
1688
				return;
1689
			}
1690
		}
1691

1692
		/* Try to read per-directory file */
1693
		oidclr(&oid_stat.oid, the_repository->hash_algo);
1694
		oid_stat.valid = 0;
1695
		if (dir->exclude_per_dir &&
1696
		    /*
1697
		     * If we know that no files have been added in
1698
		     * this directory (i.e. valid_cached_dir() has
1699
		     * been executed and set untracked->valid) ..
1700
		     */
1701
		    (!untracked || !untracked->valid ||
1702
		     /*
1703
		      * .. and .gitignore does not exist before
1704
		      * (i.e. null exclude_oid). Then we can skip
1705
		      * loading .gitignore, which would result in
1706
		      * ENOENT anyway.
1707
		      */
1708
		     !is_null_oid(&untracked->exclude_oid))) {
1709
			/*
1710
			 * dir->internal.basebuf gets reused by the traversal,
1711
			 * but we need fname to remain unchanged to ensure the
1712
			 * src member of each struct path_pattern correctly
1713
			 * back-references its source file.  Other invocations
1714
			 * of add_pattern_list provide stable strings, so we
1715
			 * strbuf_detach() and free() here in the caller.
1716
			 */
1717
			struct strbuf sb = STRBUF_INIT;
1718
			strbuf_addbuf(&sb, &dir->internal.basebuf);
1719
			strbuf_addstr(&sb, dir->exclude_per_dir);
1720
			pl->src = strbuf_detach(&sb, NULL);
1721
			add_patterns(pl->src, pl->src, stk->baselen, pl, istate,
1722
				     PATTERN_NOFOLLOW,
1723
				     untracked ? &oid_stat : NULL);
1724
		}
1725
		/*
1726
		 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1727
		 * will first be called in valid_cached_dir() then maybe many
1728
		 * times more in last_matching_pattern(). When the cache is
1729
		 * used, last_matching_pattern() will not be called and
1730
		 * reading .gitignore content will be a waste.
1731
		 *
1732
		 * So when it's called by valid_cached_dir() and we can get
1733
		 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1734
		 * modified on work tree), we could delay reading the
1735
		 * .gitignore content until we absolutely need it in
1736
		 * last_matching_pattern(). Be careful about ignore rule
1737
		 * order, though, if you do that.
1738
		 */
1739
		if (untracked &&
1740
		    !oideq(&oid_stat.oid, &untracked->exclude_oid)) {
1741
			invalidate_gitignore(dir->untracked, untracked);
1742
			oidcpy(&untracked->exclude_oid, &oid_stat.oid);
1743
		}
1744
		dir->internal.exclude_stack = stk;
1745
		current = stk->baselen;
1746
	}
1747
	strbuf_setlen(&dir->internal.basebuf, baselen);
1748
}
1749

1750
/*
1751
 * Loads the exclude lists for the directory containing pathname, then
1752
 * scans all exclude lists to determine whether pathname is excluded.
1753
 * Returns the exclude_list element which matched, or NULL for
1754
 * undecided.
1755
 */
1756
struct path_pattern *last_matching_pattern(struct dir_struct *dir,
1757
				      struct index_state *istate,
1758
				      const char *pathname,
1759
				      int *dtype_p)
1760
{
1761
	int pathlen = strlen(pathname);
1762
	const char *basename = strrchr(pathname, '/');
1763
	basename = (basename) ? basename+1 : pathname;
1764

1765
	prep_exclude(dir, istate, pathname, basename-pathname);
1766

1767
	if (dir->internal.pattern)
1768
		return dir->internal.pattern;
1769

1770
	return last_matching_pattern_from_lists(dir, istate, pathname, pathlen,
1771
			basename, dtype_p);
1772
}
1773

1774
/*
1775
 * Loads the exclude lists for the directory containing pathname, then
1776
 * scans all exclude lists to determine whether pathname is excluded.
1777
 * Returns 1 if true, otherwise 0.
1778
 */
1779
int is_excluded(struct dir_struct *dir, struct index_state *istate,
1780
		const char *pathname, int *dtype_p)
1781
{
1782
	struct path_pattern *pattern =
1783
		last_matching_pattern(dir, istate, pathname, dtype_p);
1784
	if (pattern)
1785
		return pattern->flags & PATTERN_FLAG_NEGATIVE ? 0 : 1;
1786
	return 0;
1787
}
1788

1789
static struct dir_entry *dir_entry_new(const char *pathname, int len)
1790
{
1791
	struct dir_entry *ent;
1792

1793
	FLEX_ALLOC_MEM(ent, name, pathname, len);
1794
	ent->len = len;
1795
	return ent;
1796
}
1797

1798
static struct dir_entry *dir_add_name(struct dir_struct *dir,
1799
				      struct index_state *istate,
1800
				      const char *pathname, int len)
1801
{
1802
	if (index_file_exists(istate, pathname, len, ignore_case))
1803
		return NULL;
1804

1805
	ALLOC_GROW(dir->entries, dir->nr+1, dir->internal.alloc);
1806
	return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
1807
}
1808

1809
struct dir_entry *dir_add_ignored(struct dir_struct *dir,
1810
				  struct index_state *istate,
1811
				  const char *pathname, int len)
1812
{
1813
	if (!index_name_is_other(istate, pathname, len))
1814
		return NULL;
1815

1816
	ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->internal.ignored_alloc);
1817
	return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
1818
}
1819

1820
enum exist_status {
1821
	index_nonexistent = 0,
1822
	index_directory,
1823
	index_gitdir
1824
};
1825

1826
/*
1827
 * Do not use the alphabetically sorted index to look up
1828
 * the directory name; instead, use the case insensitive
1829
 * directory hash.
1830
 */
1831
static enum exist_status directory_exists_in_index_icase(struct index_state *istate,
1832
							 const char *dirname, int len)
1833
{
1834
	struct cache_entry *ce;
1835

1836
	if (index_dir_exists(istate, dirname, len))
1837
		return index_directory;
1838

1839
	ce = index_file_exists(istate, dirname, len, ignore_case);
1840
	if (ce && S_ISGITLINK(ce->ce_mode))
1841
		return index_gitdir;
1842

1843
	return index_nonexistent;
1844
}
1845

1846
/*
1847
 * The index sorts alphabetically by entry name, which
1848
 * means that a gitlink sorts as '\0' at the end, while
1849
 * a directory (which is defined not as an entry, but as
1850
 * the files it contains) will sort with the '/' at the
1851
 * end.
1852
 */
1853
static enum exist_status directory_exists_in_index(struct index_state *istate,
1854
						   const char *dirname, int len)
1855
{
1856
	int pos;
1857

1858
	if (ignore_case)
1859
		return directory_exists_in_index_icase(istate, dirname, len);
1860

1861
	pos = index_name_pos(istate, dirname, len);
1862
	if (pos < 0)
1863
		pos = -pos-1;
1864
	while (pos < istate->cache_nr) {
1865
		const struct cache_entry *ce = istate->cache[pos++];
1866
		unsigned char endchar;
1867

1868
		if (strncmp(ce->name, dirname, len))
1869
			break;
1870
		endchar = ce->name[len];
1871
		if (endchar > '/')
1872
			break;
1873
		if (endchar == '/')
1874
			return index_directory;
1875
		if (!endchar && S_ISGITLINK(ce->ce_mode))
1876
			return index_gitdir;
1877
	}
1878
	return index_nonexistent;
1879
}
1880

1881
/*
1882
 * When we find a directory when traversing the filesystem, we
1883
 * have three distinct cases:
1884
 *
1885
 *  - ignore it
1886
 *  - see it as a directory
1887
 *  - recurse into it
1888
 *
1889
 * and which one we choose depends on a combination of existing
1890
 * git index contents and the flags passed into the directory
1891
 * traversal routine.
1892
 *
1893
 * Case 1: If we *already* have entries in the index under that
1894
 * directory name, we always recurse into the directory to see
1895
 * all the files.
1896
 *
1897
 * Case 2: If we *already* have that directory name as a gitlink,
1898
 * we always continue to see it as a gitlink, regardless of whether
1899
 * there is an actual git directory there or not (it might not
1900
 * be checked out as a subproject!)
1901
 *
1902
 * Case 3: if we didn't have it in the index previously, we
1903
 * have a few sub-cases:
1904
 *
1905
 *  (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1906
 *      just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1907
 *      also true, in which case we need to check if it contains any
1908
 *      untracked and / or ignored files.
1909
 *  (b) if it looks like a git directory and we don't have the
1910
 *      DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1911
 *      show it as a directory.
1912
 *  (c) otherwise, we recurse into it.
1913
 */
1914
static enum path_treatment treat_directory(struct dir_struct *dir,
1915
	struct index_state *istate,
1916
	struct untracked_cache_dir *untracked,
1917
	const char *dirname, int len, int baselen, int excluded,
1918
	const struct pathspec *pathspec)
1919
{
1920
	/*
1921
	 * WARNING: From this function, you can return path_recurse or you
1922
	 *          can call read_directory_recursive() (or neither), but
1923
	 *          you CAN'T DO BOTH.
1924
	 */
1925
	enum path_treatment state;
1926
	int matches_how = 0;
1927
	int check_only, stop_early;
1928
	int old_ignored_nr, old_untracked_nr;
1929
	/* The "len-1" is to strip the final '/' */
1930
	enum exist_status status = directory_exists_in_index(istate, dirname, len-1);
1931

1932
	if (status == index_directory)
1933
		return path_recurse;
1934
	if (status == index_gitdir)
1935
		return path_none;
1936
	if (status != index_nonexistent)
1937
		BUG("Unhandled value for directory_exists_in_index: %d\n", status);
1938

1939
	/*
1940
	 * We don't want to descend into paths that don't match the necessary
1941
	 * patterns.  Clearly, if we don't have a pathspec, then we can't check
1942
	 * for matching patterns.  Also, if (excluded) then we know we matched
1943
	 * the exclusion patterns so as an optimization we can skip checking
1944
	 * for matching patterns.
1945
	 */
1946
	if (pathspec && !excluded) {
1947
		matches_how = match_pathspec_with_flags(istate, pathspec,
1948
							dirname, len,
1949
							0 /* prefix */,
1950
							NULL /* seen */,
1951
							DO_MATCH_LEADING_PATHSPEC);
1952
		if (!matches_how)
1953
			return path_none;
1954
	}
1955

1956

1957
	if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1958
		!(dir->flags & DIR_NO_GITLINKS)) {
1959
		/*
1960
		 * Determine if `dirname` is a nested repo by confirming that:
1961
		 * 1) we are in a nonbare repository, and
1962
		 * 2) `dirname` is not an immediate parent of `the_repository->gitdir`,
1963
		 *    which could occur if the git_dir or worktree location was
1964
		 *    manually configured by the user; see t2205 testcases 1-3 for
1965
		 *    examples where this matters
1966
		 */
1967
		int nested_repo;
1968
		struct strbuf sb = STRBUF_INIT;
1969
		strbuf_addstr(&sb, dirname);
1970
		nested_repo = is_nonbare_repository_dir(&sb);
1971

1972
		if (nested_repo) {
1973
			char *real_dirname, *real_gitdir;
1974
			strbuf_addstr(&sb, ".git");
1975
			real_dirname = real_pathdup(sb.buf, 1);
1976
			real_gitdir = real_pathdup(the_repository->gitdir, 1);
1977

1978
			nested_repo = !!strcmp(real_dirname, real_gitdir);
1979
			free(real_gitdir);
1980
			free(real_dirname);
1981
		}
1982
		strbuf_release(&sb);
1983

1984
		if (nested_repo) {
1985
			if ((dir->flags & DIR_SKIP_NESTED_GIT) ||
1986
				(matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC))
1987
				return path_none;
1988
			return excluded ? path_excluded : path_untracked;
1989
		}
1990
	}
1991

1992
	if (!(dir->flags & DIR_SHOW_OTHER_DIRECTORIES)) {
1993
		if (excluded &&
1994
		    (dir->flags & DIR_SHOW_IGNORED_TOO) &&
1995
		    (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING)) {
1996

1997
			/*
1998
			 * This is an excluded directory and we are
1999
			 * showing ignored paths that match an exclude
2000
			 * pattern.  (e.g. show directory as ignored
2001
			 * only if it matches an exclude pattern).
2002
			 * This path will either be 'path_excluded`
2003
			 * (if we are showing empty directories or if
2004
			 * the directory is not empty), or will be
2005
			 * 'path_none' (empty directory, and we are
2006
			 * not showing empty directories).
2007
			 */
2008
			if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
2009
				return path_excluded;
2010

2011
			if (read_directory_recursive(dir, istate, dirname, len,
2012
						     untracked, 1, 1, pathspec) == path_excluded)
2013
				return path_excluded;
2014

2015
			return path_none;
2016
		}
2017
		return path_recurse;
2018
	}
2019

2020
	assert(dir->flags & DIR_SHOW_OTHER_DIRECTORIES);
2021

2022
	/*
2023
	 * If we have a pathspec which could match something _below_ this
2024
	 * directory (e.g. when checking 'subdir/' having a pathspec like
2025
	 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
2026
	 * need to recurse.
2027
	 */
2028
	if (matches_how == MATCHED_RECURSIVELY_LEADING_PATHSPEC)
2029
		return path_recurse;
2030

2031
	/* Special cases for where this directory is excluded/ignored */
2032
	if (excluded) {
2033
		/*
2034
		 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
2035
		 * hiding empty directories, there is no need to
2036
		 * recurse into an ignored directory.
2037
		 */
2038
		if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
2039
			return path_excluded;
2040

2041
		/*
2042
		 * Even if we are hiding empty directories, we can still avoid
2043
		 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
2044
		 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
2045
		 */
2046
		if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2047
		    (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING))
2048
			return path_excluded;
2049
	}
2050

2051
	/*
2052
	 * Other than the path_recurse case above, we only need to
2053
	 * recurse into untracked directories if any of the following
2054
	 * bits is set:
2055
	 *   - DIR_SHOW_IGNORED (because then we need to determine if
2056
	 *                       there are ignored entries below)
2057
	 *   - DIR_SHOW_IGNORED_TOO (same as above)
2058
	 *   - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
2059
	 *                                 the directory is empty)
2060
	 */
2061
	if (!excluded &&
2062
	    !(dir->flags & (DIR_SHOW_IGNORED |
2063
			    DIR_SHOW_IGNORED_TOO |
2064
			    DIR_HIDE_EMPTY_DIRECTORIES))) {
2065
		return path_untracked;
2066
	}
2067

2068
	/*
2069
	 * Even if we don't want to know all the paths under an untracked or
2070
	 * ignored directory, we may still need to go into the directory to
2071
	 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2072
	 * an empty directory should be path_none instead of path_excluded or
2073
	 * path_untracked).
2074
	 */
2075
	check_only = ((dir->flags & DIR_HIDE_EMPTY_DIRECTORIES) &&
2076
		      !(dir->flags & DIR_SHOW_IGNORED_TOO));
2077

2078
	/*
2079
	 * However, there's another optimization possible as a subset of
2080
	 * check_only, based on the cases we have to consider:
2081
	 *   A) Directory matches no exclude patterns:
2082
	 *     * Directory is empty => path_none
2083
	 *     * Directory has an untracked file under it => path_untracked
2084
	 *     * Directory has only ignored files under it => path_excluded
2085
	 *   B) Directory matches an exclude pattern:
2086
	 *     * Directory is empty => path_none
2087
	 *     * Directory has an untracked file under it => path_excluded
2088
	 *     * Directory has only ignored files under it => path_excluded
2089
	 * In case A, we can exit as soon as we've found an untracked
2090
	 * file but otherwise have to walk all files.  In case B, though,
2091
	 * we can stop at the first file we find under the directory.
2092
	 */
2093
	stop_early = check_only && excluded;
2094

2095
	/*
2096
	 * If /every/ file within an untracked directory is ignored, then
2097
	 * we want to treat the directory as ignored (for e.g. status
2098
	 * --porcelain), without listing the individual ignored files
2099
	 * underneath.  To do so, we'll save the current ignored_nr, and
2100
	 * pop all the ones added after it if it turns out the entire
2101
	 * directory is ignored.  Also, when DIR_SHOW_IGNORED_TOO and
2102
	 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2103
	 * untracked paths so will need to pop all those off the last
2104
	 * after we traverse.
2105
	 */
2106
	old_ignored_nr = dir->ignored_nr;
2107
	old_untracked_nr = dir->nr;
2108

2109
	/* Actually recurse into dirname now, we'll fixup the state later. */
2110
	untracked = lookup_untracked(dir->untracked, untracked,
2111
				     dirname + baselen, len - baselen);
2112
	state = read_directory_recursive(dir, istate, dirname, len, untracked,
2113
					 check_only, stop_early, pathspec);
2114

2115
	/* There are a variety of reasons we may need to fixup the state... */
2116
	if (state == path_excluded) {
2117
		/* state == path_excluded implies all paths under
2118
		 * dirname were ignored...
2119
		 *
2120
		 * if running e.g. `git status --porcelain --ignored=matching`,
2121
		 * then we want to see the subpaths that are ignored.
2122
		 *
2123
		 * if running e.g. just `git status --porcelain`, then
2124
		 * we just want the directory itself to be listed as ignored
2125
		 * and not the individual paths underneath.
2126
		 */
2127
		int want_ignored_subpaths =
2128
			((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2129
			 (dir->flags & DIR_SHOW_IGNORED_TOO_MODE_MATCHING));
2130

2131
		if (want_ignored_subpaths) {
2132
			/*
2133
			 * with --ignored=matching, we want the subpaths
2134
			 * INSTEAD of the directory itself.
2135
			 */
2136
			state = path_none;
2137
		} else {
2138
			int i;
2139
			for (i = old_ignored_nr + 1; i<dir->ignored_nr; ++i)
2140
				FREE_AND_NULL(dir->ignored[i]);
2141
			dir->ignored_nr = old_ignored_nr;
2142
		}
2143
	}
2144

2145
	/*
2146
	 * We may need to ignore some of the untracked paths we found while
2147
	 * traversing subdirectories.
2148
	 */
2149
	if ((dir->flags & DIR_SHOW_IGNORED_TOO) &&
2150
	    !(dir->flags & DIR_KEEP_UNTRACKED_CONTENTS)) {
2151
		int i;
2152
		for (i = old_untracked_nr + 1; i<dir->nr; ++i)
2153
			FREE_AND_NULL(dir->entries[i]);
2154
		dir->nr = old_untracked_nr;
2155
	}
2156

2157
	/*
2158
	 * If there is nothing under the current directory and we are not
2159
	 * hiding empty directories, then we need to report on the
2160
	 * untracked or ignored status of the directory itself.
2161
	 */
2162
	if (state == path_none && !(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
2163
		state = excluded ? path_excluded : path_untracked;
2164

2165
	return state;
2166
}
2167

2168
/*
2169
 * This is an inexact early pruning of any recursive directory
2170
 * reading - if the path cannot possibly be in the pathspec,
2171
 * return true, and we'll skip it early.
2172
 */
2173
static int simplify_away(const char *path, int pathlen,
2174
			 const struct pathspec *pathspec)
2175
{
2176
	int i;
2177

2178
	if (!pathspec || !pathspec->nr)
2179
		return 0;
2180

2181
	GUARD_PATHSPEC(pathspec,
2182
		       PATHSPEC_FROMTOP |
2183
		       PATHSPEC_MAXDEPTH |
2184
		       PATHSPEC_LITERAL |
2185
		       PATHSPEC_GLOB |
2186
		       PATHSPEC_ICASE |
2187
		       PATHSPEC_EXCLUDE |
2188
		       PATHSPEC_ATTR);
2189

2190
	for (i = 0; i < pathspec->nr; i++) {
2191
		const struct pathspec_item *item = &pathspec->items[i];
2192
		int len = item->nowildcard_len;
2193

2194
		if (len > pathlen)
2195
			len = pathlen;
2196
		if (!ps_strncmp(item, item->match, path, len))
2197
			return 0;
2198
	}
2199

2200
	return 1;
2201
}
2202

2203
/*
2204
 * This function tells us whether an excluded path matches a
2205
 * list of "interesting" pathspecs. That is, whether a path matched
2206
 * by any of the pathspecs could possibly be ignored by excluding
2207
 * the specified path. This can happen if:
2208
 *
2209
 *   1. the path is mentioned explicitly in the pathspec
2210
 *
2211
 *   2. the path is a directory prefix of some element in the
2212
 *      pathspec
2213
 */
2214
static int exclude_matches_pathspec(const char *path, int pathlen,
2215
				    const struct pathspec *pathspec)
2216
{
2217
	int i;
2218

2219
	if (!pathspec || !pathspec->nr)
2220
		return 0;
2221

2222
	GUARD_PATHSPEC(pathspec,
2223
		       PATHSPEC_FROMTOP |
2224
		       PATHSPEC_MAXDEPTH |
2225
		       PATHSPEC_LITERAL |
2226
		       PATHSPEC_GLOB |
2227
		       PATHSPEC_ICASE |
2228
		       PATHSPEC_EXCLUDE |
2229
		       PATHSPEC_ATTR);
2230

2231
	for (i = 0; i < pathspec->nr; i++) {
2232
		const struct pathspec_item *item = &pathspec->items[i];
2233
		int len = item->nowildcard_len;
2234

2235
		if (len == pathlen &&
2236
		    !ps_strncmp(item, item->match, path, pathlen))
2237
			return 1;
2238
		if (len > pathlen &&
2239
		    item->match[pathlen] == '/' &&
2240
		    !ps_strncmp(item, item->match, path, pathlen))
2241
			return 1;
2242
	}
2243
	return 0;
2244
}
2245

2246
static int get_index_dtype(struct index_state *istate,
2247
			   const char *path, int len)
2248
{
2249
	int pos;
2250
	const struct cache_entry *ce;
2251

2252
	ce = index_file_exists(istate, path, len, 0);
2253
	if (ce) {
2254
		if (!ce_uptodate(ce))
2255
			return DT_UNKNOWN;
2256
		if (S_ISGITLINK(ce->ce_mode))
2257
			return DT_DIR;
2258
		/*
2259
		 * Nobody actually cares about the
2260
		 * difference between DT_LNK and DT_REG
2261
		 */
2262
		return DT_REG;
2263
	}
2264

2265
	/* Try to look it up as a directory */
2266
	pos = index_name_pos(istate, path, len);
2267
	if (pos >= 0)
2268
		return DT_UNKNOWN;
2269
	pos = -pos-1;
2270
	while (pos < istate->cache_nr) {
2271
		ce = istate->cache[pos++];
2272
		if (strncmp(ce->name, path, len))
2273
			break;
2274
		if (ce->name[len] > '/')
2275
			break;
2276
		if (ce->name[len] < '/')
2277
			continue;
2278
		if (!ce_uptodate(ce))
2279
			break;	/* continue? */
2280
		return DT_DIR;
2281
	}
2282
	return DT_UNKNOWN;
2283
}
2284

2285
unsigned char get_dtype(struct dirent *e, struct strbuf *path,
2286
			int follow_symlink)
2287
{
2288
	struct stat st;
2289
	unsigned char dtype = DTYPE(e);
2290
	size_t base_path_len;
2291

2292
	if (dtype != DT_UNKNOWN && !(follow_symlink && dtype == DT_LNK))
2293
		return dtype;
2294

2295
	/*
2296
	 * d_type unknown or unfollowed symlink, try to fall back on [l]stat
2297
	 * results. If [l]stat fails, explicitly set DT_UNKNOWN.
2298
	 */
2299
	base_path_len = path->len;
2300
	strbuf_addstr(path, e->d_name);
2301
	if ((follow_symlink && stat(path->buf, &st)) ||
2302
	    (!follow_symlink && lstat(path->buf, &st)))
2303
		goto cleanup;
2304

2305
	/* determine d_type from st_mode */
2306
	if (S_ISREG(st.st_mode))
2307
		dtype = DT_REG;
2308
	else if (S_ISDIR(st.st_mode))
2309
		dtype = DT_DIR;
2310
	else if (S_ISLNK(st.st_mode))
2311
		dtype = DT_LNK;
2312

2313
cleanup:
2314
	strbuf_setlen(path, base_path_len);
2315
	return dtype;
2316
}
2317

2318
static int resolve_dtype(int dtype, struct index_state *istate,
2319
			 const char *path, int len)
2320
{
2321
	struct stat st;
2322

2323
	if (dtype != DT_UNKNOWN)
2324
		return dtype;
2325
	dtype = get_index_dtype(istate, path, len);
2326
	if (dtype != DT_UNKNOWN)
2327
		return dtype;
2328
	if (lstat(path, &st))
2329
		return dtype;
2330
	if (S_ISREG(st.st_mode))
2331
		return DT_REG;
2332
	if (S_ISDIR(st.st_mode))
2333
		return DT_DIR;
2334
	if (S_ISLNK(st.st_mode))
2335
		return DT_LNK;
2336
	return dtype;
2337
}
2338

2339
static enum path_treatment treat_path_fast(struct dir_struct *dir,
2340
					   struct cached_dir *cdir,
2341
					   struct index_state *istate,
2342
					   struct strbuf *path,
2343
					   int baselen,
2344
					   const struct pathspec *pathspec)
2345
{
2346
	/*
2347
	 * WARNING: From this function, you can return path_recurse or you
2348
	 *          can call read_directory_recursive() (or neither), but
2349
	 *          you CAN'T DO BOTH.
2350
	 */
2351
	strbuf_setlen(path, baselen);
2352
	if (!cdir->ucd) {
2353
		strbuf_addstr(path, cdir->file);
2354
		return path_untracked;
2355
	}
2356
	strbuf_addstr(path, cdir->ucd->name);
2357
	/* treat_one_path() does this before it calls treat_directory() */
2358
	strbuf_complete(path, '/');
2359
	if (cdir->ucd->check_only)
2360
		/*
2361
		 * check_only is set as a result of treat_directory() getting
2362
		 * to its bottom. Verify again the same set of directories
2363
		 * with check_only set.
2364
		 */
2365
		return read_directory_recursive(dir, istate, path->buf, path->len,
2366
						cdir->ucd, 1, 0, pathspec);
2367
	/*
2368
	 * We get path_recurse in the first run when
2369
	 * directory_exists_in_index() returns index_nonexistent. We
2370
	 * are sure that new changes in the index does not impact the
2371
	 * outcome. Return now.
2372
	 */
2373
	return path_recurse;
2374
}
2375

2376
static enum path_treatment treat_path(struct dir_struct *dir,
2377
				      struct untracked_cache_dir *untracked,
2378
				      struct cached_dir *cdir,
2379
				      struct index_state *istate,
2380
				      struct strbuf *path,
2381
				      int baselen,
2382
				      const struct pathspec *pathspec)
2383
{
2384
	int has_path_in_index, dtype, excluded;
2385

2386
	if (!cdir->d_name)
2387
		return treat_path_fast(dir, cdir, istate, path,
2388
				       baselen, pathspec);
2389
	if (is_dot_or_dotdot(cdir->d_name) || !fspathcmp(cdir->d_name, ".git"))
2390
		return path_none;
2391
	strbuf_setlen(path, baselen);
2392
	strbuf_addstr(path, cdir->d_name);
2393
	if (simplify_away(path->buf, path->len, pathspec))
2394
		return path_none;
2395

2396
	dtype = resolve_dtype(cdir->d_type, istate, path->buf, path->len);
2397

2398
	/* Always exclude indexed files */
2399
	has_path_in_index = !!index_file_exists(istate, path->buf, path->len,
2400
						ignore_case);
2401
	if (dtype != DT_DIR && has_path_in_index)
2402
		return path_none;
2403

2404
	/*
2405
	 * When we are looking at a directory P in the working tree,
2406
	 * there are three cases:
2407
	 *
2408
	 * (1) P exists in the index.  Everything inside the directory P in
2409
	 * the working tree needs to go when P is checked out from the
2410
	 * index.
2411
	 *
2412
	 * (2) P does not exist in the index, but there is P/Q in the index.
2413
	 * We know P will stay a directory when we check out the contents
2414
	 * of the index, but we do not know yet if there is a directory
2415
	 * P/Q in the working tree to be killed, so we need to recurse.
2416
	 *
2417
	 * (3) P does not exist in the index, and there is no P/Q in the index
2418
	 * to require P to be a directory, either.  Only in this case, we
2419
	 * know that everything inside P will not be killed without
2420
	 * recursing.
2421
	 */
2422
	if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
2423
	    (dtype == DT_DIR) &&
2424
	    !has_path_in_index &&
2425
	    (directory_exists_in_index(istate, path->buf, path->len) == index_nonexistent))
2426
		return path_none;
2427

2428
	excluded = is_excluded(dir, istate, path->buf, &dtype);
2429

2430
	/*
2431
	 * Excluded? If we don't explicitly want to show
2432
	 * ignored files, ignore it
2433
	 */
2434
	if (excluded && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
2435
		return path_excluded;
2436

2437
	switch (dtype) {
2438
	default:
2439
		return path_none;
2440
	case DT_DIR:
2441
		/*
2442
		 * WARNING: Do not ignore/amend the return value from
2443
		 * treat_directory(), and especially do not change it to return
2444
		 * path_recurse as that can cause exponential slowdown.
2445
		 * Instead, modify treat_directory() to return the right value.
2446
		 */
2447
		strbuf_addch(path, '/');
2448
		return treat_directory(dir, istate, untracked,
2449
				       path->buf, path->len,
2450
				       baselen, excluded, pathspec);
2451
	case DT_REG:
2452
	case DT_LNK:
2453
		if (pathspec &&
2454
		    !match_pathspec(istate, pathspec, path->buf, path->len,
2455
				    0 /* prefix */, NULL /* seen */,
2456
				    0 /* is_dir */))
2457
			return path_none;
2458
		if (excluded)
2459
			return path_excluded;
2460
		return path_untracked;
2461
	}
2462
}
2463

2464
static void add_untracked(struct untracked_cache_dir *dir, const char *name)
2465
{
2466
	if (!dir)
2467
		return;
2468
	ALLOC_GROW(dir->untracked, dir->untracked_nr + 1,
2469
		   dir->untracked_alloc);
2470
	dir->untracked[dir->untracked_nr++] = xstrdup(name);
2471
}
2472

2473
static int valid_cached_dir(struct dir_struct *dir,
2474
			    struct untracked_cache_dir *untracked,
2475
			    struct index_state *istate,
2476
			    struct strbuf *path,
2477
			    int check_only)
2478
{
2479
	struct stat st;
2480

2481
	if (!untracked)
2482
		return 0;
2483

2484
	/*
2485
	 * With fsmonitor, we can trust the untracked cache's valid field.
2486
	 */
2487
	refresh_fsmonitor(istate);
2488
	if (!(dir->untracked->use_fsmonitor && untracked->valid)) {
2489
		if (lstat(path->len ? path->buf : ".", &st)) {
2490
			memset(&untracked->stat_data, 0, sizeof(untracked->stat_data));
2491
			return 0;
2492
		}
2493
		if (!untracked->valid ||
2494
			match_stat_data_racy(istate, &untracked->stat_data, &st)) {
2495
			fill_stat_data(&untracked->stat_data, &st);
2496
			return 0;
2497
		}
2498
	}
2499

2500
	if (untracked->check_only != !!check_only)
2501
		return 0;
2502

2503
	/*
2504
	 * prep_exclude will be called eventually on this directory,
2505
	 * but it's called much later in last_matching_pattern(). We
2506
	 * need it now to determine the validity of the cache for this
2507
	 * path. The next calls will be nearly no-op, the way
2508
	 * prep_exclude() is designed.
2509
	 */
2510
	if (path->len && path->buf[path->len - 1] != '/') {
2511
		strbuf_addch(path, '/');
2512
		prep_exclude(dir, istate, path->buf, path->len);
2513
		strbuf_setlen(path, path->len - 1);
2514
	} else
2515
		prep_exclude(dir, istate, path->buf, path->len);
2516

2517
	/* hopefully prep_exclude() haven't invalidated this entry... */
2518
	return untracked->valid;
2519
}
2520

2521
static int open_cached_dir(struct cached_dir *cdir,
2522
			   struct dir_struct *dir,
2523
			   struct untracked_cache_dir *untracked,
2524
			   struct index_state *istate,
2525
			   struct strbuf *path,
2526
			   int check_only)
2527
{
2528
	const char *c_path;
2529

2530
	memset(cdir, 0, sizeof(*cdir));
2531
	cdir->untracked = untracked;
2532
	if (valid_cached_dir(dir, untracked, istate, path, check_only))
2533
		return 0;
2534
	c_path = path->len ? path->buf : ".";
2535
	cdir->fdir = opendir(c_path);
2536
	if (!cdir->fdir)
2537
		warning_errno(_("could not open directory '%s'"), c_path);
2538
	if (dir->untracked) {
2539
		invalidate_directory(dir->untracked, untracked);
2540
		dir->untracked->dir_opened++;
2541
	}
2542
	if (!cdir->fdir)
2543
		return -1;
2544
	return 0;
2545
}
2546

2547
static int read_cached_dir(struct cached_dir *cdir)
2548
{
2549
	struct dirent *de;
2550

2551
	if (cdir->fdir) {
2552
		de = readdir_skip_dot_and_dotdot(cdir->fdir);
2553
		if (!de) {
2554
			cdir->d_name = NULL;
2555
			cdir->d_type = DT_UNKNOWN;
2556
			return -1;
2557
		}
2558
		cdir->d_name = de->d_name;
2559
		cdir->d_type = DTYPE(de);
2560
		return 0;
2561
	}
2562
	while (cdir->nr_dirs < cdir->untracked->dirs_nr) {
2563
		struct untracked_cache_dir *d = cdir->untracked->dirs[cdir->nr_dirs];
2564
		if (!d->recurse) {
2565
			cdir->nr_dirs++;
2566
			continue;
2567
		}
2568
		cdir->ucd = d;
2569
		cdir->nr_dirs++;
2570
		return 0;
2571
	}
2572
	cdir->ucd = NULL;
2573
	if (cdir->nr_files < cdir->untracked->untracked_nr) {
2574
		struct untracked_cache_dir *d = cdir->untracked;
2575
		cdir->file = d->untracked[cdir->nr_files++];
2576
		return 0;
2577
	}
2578
	return -1;
2579
}
2580

2581
static void close_cached_dir(struct cached_dir *cdir)
2582
{
2583
	if (cdir->fdir)
2584
		closedir(cdir->fdir);
2585
	/*
2586
	 * We have gone through this directory and found no untracked
2587
	 * entries. Mark it valid.
2588
	 */
2589
	if (cdir->untracked) {
2590
		cdir->untracked->valid = 1;
2591
		cdir->untracked->recurse = 1;
2592
	}
2593
}
2594

2595
static void add_path_to_appropriate_result_list(struct dir_struct *dir,
2596
	struct untracked_cache_dir *untracked,
2597
	struct cached_dir *cdir,
2598
	struct index_state *istate,
2599
	struct strbuf *path,
2600
	int baselen,
2601
	const struct pathspec *pathspec,
2602
	enum path_treatment state)
2603
{
2604
	/* add the path to the appropriate result list */
2605
	switch (state) {
2606
	case path_excluded:
2607
		if (dir->flags & DIR_SHOW_IGNORED)
2608
			dir_add_name(dir, istate, path->buf, path->len);
2609
		else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
2610
			((dir->flags & DIR_COLLECT_IGNORED) &&
2611
			exclude_matches_pathspec(path->buf, path->len,
2612
						 pathspec)))
2613
			dir_add_ignored(dir, istate, path->buf, path->len);
2614
		break;
2615

2616
	case path_untracked:
2617
		if (dir->flags & DIR_SHOW_IGNORED)
2618
			break;
2619
		dir_add_name(dir, istate, path->buf, path->len);
2620
		if (cdir->fdir)
2621
			add_untracked(untracked, path->buf + baselen);
2622
		break;
2623

2624
	default:
2625
		break;
2626
	}
2627
}
2628

2629
/*
2630
 * Read a directory tree. We currently ignore anything but
2631
 * directories, regular files and symlinks. That's because git
2632
 * doesn't handle them at all yet. Maybe that will change some
2633
 * day.
2634
 *
2635
 * Also, we ignore the name ".git" (even if it is not a directory).
2636
 * That likely will not change.
2637
 *
2638
 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2639
 * to signal that a file was found. This is the least significant value that
2640
 * indicates that a file was encountered that does not depend on the order of
2641
 * whether an untracked or excluded path was encountered first.
2642
 *
2643
 * Returns the most significant path_treatment value encountered in the scan.
2644
 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2645
 * significant path_treatment value that will be returned.
2646
 */
2647

2648
static enum path_treatment read_directory_recursive(struct dir_struct *dir,
2649
	struct index_state *istate, const char *base, int baselen,
2650
	struct untracked_cache_dir *untracked, int check_only,
2651
	int stop_at_first_file, const struct pathspec *pathspec)
2652
{
2653
	/*
2654
	 * WARNING: Do NOT recurse unless path_recurse is returned from
2655
	 *          treat_path().  Recursing on any other return value
2656
	 *          can result in exponential slowdown.
2657
	 */
2658
	struct cached_dir cdir;
2659
	enum path_treatment state, subdir_state, dir_state = path_none;
2660
	struct strbuf path = STRBUF_INIT;
2661

2662
	strbuf_add(&path, base, baselen);
2663

2664
	if (open_cached_dir(&cdir, dir, untracked, istate, &path, check_only))
2665
		goto out;
2666
	dir->internal.visited_directories++;
2667

2668
	if (untracked)
2669
		untracked->check_only = !!check_only;
2670

2671
	while (!read_cached_dir(&cdir)) {
2672
		/* check how the file or directory should be treated */
2673
		state = treat_path(dir, untracked, &cdir, istate, &path,
2674
				   baselen, pathspec);
2675
		dir->internal.visited_paths++;
2676

2677
		if (state > dir_state)
2678
			dir_state = state;
2679

2680
		/* recurse into subdir if instructed by treat_path */
2681
		if (state == path_recurse) {
2682
			struct untracked_cache_dir *ud;
2683
			ud = lookup_untracked(dir->untracked,
2684
					      untracked,
2685
					      path.buf + baselen,
2686
					      path.len - baselen);
2687
			subdir_state =
2688
				read_directory_recursive(dir, istate, path.buf,
2689
							 path.len, ud,
2690
							 check_only, stop_at_first_file, pathspec);
2691
			if (subdir_state > dir_state)
2692
				dir_state = subdir_state;
2693

2694
			if (pathspec &&
2695
			    !match_pathspec(istate, pathspec, path.buf, path.len,
2696
					    0 /* prefix */, NULL,
2697
					    0 /* do NOT special case dirs */))
2698
				state = path_none;
2699
		}
2700

2701
		if (check_only) {
2702
			if (stop_at_first_file) {
2703
				/*
2704
				 * If stopping at first file, then
2705
				 * signal that a file was found by
2706
				 * returning `path_excluded`. This is
2707
				 * to return a consistent value
2708
				 * regardless of whether an ignored or
2709
				 * excluded file happened to be
2710
				 * encountered 1st.
2711
				 *
2712
				 * In current usage, the
2713
				 * `stop_at_first_file` is passed when
2714
				 * an ancestor directory has matched
2715
				 * an exclude pattern, so any found
2716
				 * files will be excluded.
2717
				 */
2718
				if (dir_state >= path_excluded) {
2719
					dir_state = path_excluded;
2720
					break;
2721
				}
2722
			}
2723

2724
			/* abort early if maximum state has been reached */
2725
			if (dir_state == path_untracked) {
2726
				if (cdir.fdir)
2727
					add_untracked(untracked, path.buf + baselen);
2728
				break;
2729
			}
2730
			/* skip the add_path_to_appropriate_result_list() */
2731
			continue;
2732
		}
2733

2734
		add_path_to_appropriate_result_list(dir, untracked, &cdir,
2735
						    istate, &path, baselen,
2736
						    pathspec, state);
2737
	}
2738
	close_cached_dir(&cdir);
2739
 out:
2740
	strbuf_release(&path);
2741

2742
	return dir_state;
2743
}
2744

2745
int cmp_dir_entry(const void *p1, const void *p2)
2746
{
2747
	const struct dir_entry *e1 = *(const struct dir_entry **)p1;
2748
	const struct dir_entry *e2 = *(const struct dir_entry **)p2;
2749

2750
	return name_compare(e1->name, e1->len, e2->name, e2->len);
2751
}
2752

2753
/* check if *out lexically strictly contains *in */
2754
int check_dir_entry_contains(const struct dir_entry *out, const struct dir_entry *in)
2755
{
2756
	return (out->len < in->len) &&
2757
		(out->name[out->len - 1] == '/') &&
2758
		!memcmp(out->name, in->name, out->len);
2759
}
2760

2761
static int treat_leading_path(struct dir_struct *dir,
2762
			      struct index_state *istate,
2763
			      const char *path, int len,
2764
			      const struct pathspec *pathspec)
2765
{
2766
	struct strbuf sb = STRBUF_INIT;
2767
	struct strbuf subdir = STRBUF_INIT;
2768
	int prevlen, baselen;
2769
	const char *cp;
2770
	struct cached_dir cdir;
2771
	enum path_treatment state = path_none;
2772

2773
	/*
2774
	 * For each directory component of path, we are going to check whether
2775
	 * that path is relevant given the pathspec.  For example, if path is
2776
	 *    foo/bar/baz/
2777
	 * then we will ask treat_path() whether we should go into foo, then
2778
	 * whether we should go into bar, then whether baz is relevant.
2779
	 * Checking each is important because e.g. if path is
2780
	 *    .git/info/
2781
	 * then we need to check .git to know we shouldn't traverse it.
2782
	 * If the return from treat_path() is:
2783
	 *    * path_none, for any path, we return false.
2784
	 *    * path_recurse, for all path components, we return true
2785
	 *    * <anything else> for some intermediate component, we make sure
2786
	 *        to add that path to the relevant list but return false
2787
	 *        signifying that we shouldn't recurse into it.
2788
	 */
2789

2790
	while (len && path[len - 1] == '/')
2791
		len--;
2792
	if (!len)
2793
		return 1;
2794

2795
	memset(&cdir, 0, sizeof(cdir));
2796
	cdir.d_type = DT_DIR;
2797
	baselen = 0;
2798
	prevlen = 0;
2799
	while (1) {
2800
		prevlen = baselen + !!baselen;
2801
		cp = path + prevlen;
2802
		cp = memchr(cp, '/', path + len - cp);
2803
		if (!cp)
2804
			baselen = len;
2805
		else
2806
			baselen = cp - path;
2807
		strbuf_reset(&sb);
2808
		strbuf_add(&sb, path, baselen);
2809
		if (!is_directory(sb.buf))
2810
			break;
2811
		strbuf_reset(&sb);
2812
		strbuf_add(&sb, path, prevlen);
2813
		strbuf_reset(&subdir);
2814
		strbuf_add(&subdir, path+prevlen, baselen-prevlen);
2815
		cdir.d_name = subdir.buf;
2816
		state = treat_path(dir, NULL, &cdir, istate, &sb, prevlen, pathspec);
2817

2818
		if (state != path_recurse)
2819
			break; /* do not recurse into it */
2820
		if (len <= baselen)
2821
			break; /* finished checking */
2822
	}
2823
	add_path_to_appropriate_result_list(dir, NULL, &cdir, istate,
2824
					    &sb, baselen, pathspec,
2825
					    state);
2826

2827
	strbuf_release(&subdir);
2828
	strbuf_release(&sb);
2829
	return state == path_recurse;
2830
}
2831

2832
static const char *get_ident_string(void)
2833
{
2834
	static struct strbuf sb = STRBUF_INIT;
2835
	struct utsname uts;
2836

2837
	if (sb.len)
2838
		return sb.buf;
2839
	if (uname(&uts) < 0)
2840
		die_errno(_("failed to get kernel name and information"));
2841
	strbuf_addf(&sb, "Location %s, system %s", get_git_work_tree(),
2842
		    uts.sysname);
2843
	return sb.buf;
2844
}
2845

2846
static int ident_in_untracked(const struct untracked_cache *uc)
2847
{
2848
	/*
2849
	 * Previous git versions may have saved many NUL separated
2850
	 * strings in the "ident" field, but it is insane to manage
2851
	 * many locations, so just take care of the first one.
2852
	 */
2853

2854
	return !strcmp(uc->ident.buf, get_ident_string());
2855
}
2856

2857
static void set_untracked_ident(struct untracked_cache *uc)
2858
{
2859
	strbuf_reset(&uc->ident);
2860
	strbuf_addstr(&uc->ident, get_ident_string());
2861

2862
	/*
2863
	 * This strbuf used to contain a list of NUL separated
2864
	 * strings, so save NUL too for backward compatibility.
2865
	 */
2866
	strbuf_addch(&uc->ident, 0);
2867
}
2868

2869
static unsigned new_untracked_cache_flags(struct index_state *istate)
2870
{
2871
	struct repository *repo = istate->repo;
2872
	char *val;
2873

2874
	/*
2875
	 * This logic is coordinated with the setting of these flags in
2876
	 * wt-status.c#wt_status_collect_untracked(), and the evaluation
2877
	 * of the config setting in commit.c#git_status_config()
2878
	 */
2879
	if (!repo_config_get_string(repo, "status.showuntrackedfiles", &val) &&
2880
	    !strcmp(val, "all"))
2881
		return 0;
2882

2883
	/*
2884
	 * The default, if "all" is not set, is "normal" - leading us here.
2885
	 * If the value is "none" then it really doesn't matter.
2886
	 */
2887
	return DIR_SHOW_OTHER_DIRECTORIES | DIR_HIDE_EMPTY_DIRECTORIES;
2888
}
2889

2890
static void new_untracked_cache(struct index_state *istate, int flags)
2891
{
2892
	struct untracked_cache *uc = xcalloc(1, sizeof(*uc));
2893
	strbuf_init(&uc->ident, 100);
2894
	uc->exclude_per_dir = ".gitignore";
2895
	uc->dir_flags = flags >= 0 ? flags : new_untracked_cache_flags(istate);
2896
	set_untracked_ident(uc);
2897
	istate->untracked = uc;
2898
	istate->cache_changed |= UNTRACKED_CHANGED;
2899
}
2900

2901
void add_untracked_cache(struct index_state *istate)
2902
{
2903
	if (!istate->untracked) {
2904
		new_untracked_cache(istate, -1);
2905
	} else {
2906
		if (!ident_in_untracked(istate->untracked)) {
2907
			free_untracked_cache(istate->untracked);
2908
			new_untracked_cache(istate, -1);
2909
		}
2910
	}
2911
}
2912

2913
void remove_untracked_cache(struct index_state *istate)
2914
{
2915
	if (istate->untracked) {
2916
		free_untracked_cache(istate->untracked);
2917
		istate->untracked = NULL;
2918
		istate->cache_changed |= UNTRACKED_CHANGED;
2919
	}
2920
}
2921

2922
static struct untracked_cache_dir *validate_untracked_cache(struct dir_struct *dir,
2923
						      int base_len,
2924
						      const struct pathspec *pathspec,
2925
						      struct index_state *istate)
2926
{
2927
	struct untracked_cache_dir *root;
2928
	static int untracked_cache_disabled = -1;
2929

2930
	if (!dir->untracked)
2931
		return NULL;
2932
	if (untracked_cache_disabled < 0)
2933
		untracked_cache_disabled = git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2934
	if (untracked_cache_disabled)
2935
		return NULL;
2936

2937
	/*
2938
	 * We only support $GIT_DIR/info/exclude and core.excludesfile
2939
	 * as the global ignore rule files. Any other additions
2940
	 * (e.g. from command line) invalidate the cache. This
2941
	 * condition also catches running setup_standard_excludes()
2942
	 * before setting dir->untracked!
2943
	 */
2944
	if (dir->internal.unmanaged_exclude_files)
2945
		return NULL;
2946

2947
	/*
2948
	 * Optimize for the main use case only: whole-tree git
2949
	 * status. More work involved in treat_leading_path() if we
2950
	 * use cache on just a subset of the worktree. pathspec
2951
	 * support could make the matter even worse.
2952
	 */
2953
	if (base_len || (pathspec && pathspec->nr))
2954
		return NULL;
2955

2956
	/* We don't support collecting ignore files */
2957
	if (dir->flags & (DIR_SHOW_IGNORED | DIR_SHOW_IGNORED_TOO |
2958
			DIR_COLLECT_IGNORED))
2959
		return NULL;
2960

2961
	/*
2962
	 * If we use .gitignore in the cache and now you change it to
2963
	 * .gitexclude, everything will go wrong.
2964
	 */
2965
	if (dir->exclude_per_dir != dir->untracked->exclude_per_dir &&
2966
	    strcmp(dir->exclude_per_dir, dir->untracked->exclude_per_dir))
2967
		return NULL;
2968

2969
	/*
2970
	 * EXC_CMDL is not considered in the cache. If people set it,
2971
	 * skip the cache.
2972
	 */
2973
	if (dir->internal.exclude_list_group[EXC_CMDL].nr)
2974
		return NULL;
2975

2976
	if (!ident_in_untracked(dir->untracked)) {
2977
		warning(_("untracked cache is disabled on this system or location"));
2978
		return NULL;
2979
	}
2980

2981
	/*
2982
	 * If the untracked structure we received does not have the same flags
2983
	 * as requested in this run, we're going to need to either discard the
2984
	 * existing structure (and potentially later recreate), or bypass the
2985
	 * untracked cache mechanism for this run.
2986
	 */
2987
	if (dir->flags != dir->untracked->dir_flags) {
2988
		/*
2989
		 * If the untracked structure we received does not have the same flags
2990
		 * as configured, then we need to reset / create a new "untracked"
2991
		 * structure to match the new config.
2992
		 *
2993
		 * Keeping the saved and used untracked cache consistent with the
2994
		 * configuration provides an opportunity for frequent users of
2995
		 * "git status -uall" to leverage the untracked cache by aligning their
2996
		 * configuration - setting "status.showuntrackedfiles" to "all" or
2997
		 * "normal" as appropriate.
2998
		 *
2999
		 * Previously using -uall (or setting "status.showuntrackedfiles" to
3000
		 * "all") was incompatible with untracked cache and *consistently*
3001
		 * caused surprisingly bad performance (with fscache and fsmonitor
3002
		 * enabled) on Windows.
3003
		 *
3004
		 * IMPROVEMENT OPPORTUNITY: If we reworked the untracked cache storage
3005
		 * to not be as bound up with the desired output in a given run,
3006
		 * and instead iterated through and stored enough information to
3007
		 * correctly serve both "modes", then users could get peak performance
3008
		 * with or without '-uall' regardless of their
3009
		 * "status.showuntrackedfiles" config.
3010
		 */
3011
		if (dir->untracked->dir_flags != new_untracked_cache_flags(istate)) {
3012
			free_untracked_cache(istate->untracked);
3013
			new_untracked_cache(istate, dir->flags);
3014
			dir->untracked = istate->untracked;
3015
		}
3016
		else {
3017
			/*
3018
			 * Current untracked cache data is consistent with config, but not
3019
			 * usable in this request/run; just bypass untracked cache.
3020
			 */
3021
			return NULL;
3022
		}
3023
	}
3024

3025
	if (!dir->untracked->root) {
3026
		/* Untracked cache existed but is not initialized; fix that */
3027
		FLEX_ALLOC_STR(dir->untracked->root, name, "");
3028
		istate->cache_changed |= UNTRACKED_CHANGED;
3029
	}
3030

3031
	/* Validate $GIT_DIR/info/exclude and core.excludesfile */
3032
	root = dir->untracked->root;
3033
	if (!oideq(&dir->internal.ss_info_exclude.oid,
3034
		   &dir->untracked->ss_info_exclude.oid)) {
3035
		invalidate_gitignore(dir->untracked, root);
3036
		dir->untracked->ss_info_exclude = dir->internal.ss_info_exclude;
3037
	}
3038
	if (!oideq(&dir->internal.ss_excludes_file.oid,
3039
		   &dir->untracked->ss_excludes_file.oid)) {
3040
		invalidate_gitignore(dir->untracked, root);
3041
		dir->untracked->ss_excludes_file = dir->internal.ss_excludes_file;
3042
	}
3043

3044
	/* Make sure this directory is not dropped out at saving phase */
3045
	root->recurse = 1;
3046
	return root;
3047
}
3048

3049
static void emit_traversal_statistics(struct dir_struct *dir,
3050
				      struct repository *repo,
3051
				      const char *path,
3052
				      int path_len)
3053
{
3054
	if (!trace2_is_enabled())
3055
		return;
3056

3057
	if (!path_len) {
3058
		trace2_data_string("read_directory", repo, "path", "");
3059
	} else {
3060
		struct strbuf tmp = STRBUF_INIT;
3061
		strbuf_add(&tmp, path, path_len);
3062
		trace2_data_string("read_directory", repo, "path", tmp.buf);
3063
		strbuf_release(&tmp);
3064
	}
3065

3066
	trace2_data_intmax("read_directory", repo,
3067
			   "directories-visited", dir->internal.visited_directories);
3068
	trace2_data_intmax("read_directory", repo,
3069
			   "paths-visited", dir->internal.visited_paths);
3070

3071
	if (!dir->untracked)
3072
		return;
3073
	trace2_data_intmax("read_directory", repo,
3074
			   "node-creation", dir->untracked->dir_created);
3075
	trace2_data_intmax("read_directory", repo,
3076
			   "gitignore-invalidation",
3077
			   dir->untracked->gitignore_invalidated);
3078
	trace2_data_intmax("read_directory", repo,
3079
			   "directory-invalidation",
3080
			   dir->untracked->dir_invalidated);
3081
	trace2_data_intmax("read_directory", repo,
3082
			   "opendir", dir->untracked->dir_opened);
3083
}
3084

3085
int read_directory(struct dir_struct *dir, struct index_state *istate,
3086
		   const char *path, int len, const struct pathspec *pathspec)
3087
{
3088
	struct untracked_cache_dir *untracked;
3089

3090
	trace2_region_enter("dir", "read_directory", istate->repo);
3091
	dir->internal.visited_paths = 0;
3092
	dir->internal.visited_directories = 0;
3093

3094
	if (has_symlink_leading_path(path, len)) {
3095
		trace2_region_leave("dir", "read_directory", istate->repo);
3096
		return dir->nr;
3097
	}
3098

3099
	untracked = validate_untracked_cache(dir, len, pathspec, istate);
3100
	if (!untracked)
3101
		/*
3102
		 * make sure untracked cache code path is disabled,
3103
		 * e.g. prep_exclude()
3104
		 */
3105
		dir->untracked = NULL;
3106
	if (!len || treat_leading_path(dir, istate, path, len, pathspec))
3107
		read_directory_recursive(dir, istate, path, len, untracked, 0, 0, pathspec);
3108
	QSORT(dir->entries, dir->nr, cmp_dir_entry);
3109
	QSORT(dir->ignored, dir->ignored_nr, cmp_dir_entry);
3110

3111
	emit_traversal_statistics(dir, istate->repo, path, len);
3112

3113
	trace2_region_leave("dir", "read_directory", istate->repo);
3114
	if (dir->untracked) {
3115
		static int force_untracked_cache = -1;
3116

3117
		if (force_untracked_cache < 0)
3118
			force_untracked_cache =
3119
				git_env_bool("GIT_FORCE_UNTRACKED_CACHE", -1);
3120
		if (force_untracked_cache < 0)
3121
			force_untracked_cache = (istate->repo->settings.core_untracked_cache == UNTRACKED_CACHE_WRITE);
3122
		if (force_untracked_cache &&
3123
			dir->untracked == istate->untracked &&
3124
		    (dir->untracked->dir_opened ||
3125
		     dir->untracked->gitignore_invalidated ||
3126
		     dir->untracked->dir_invalidated))
3127
			istate->cache_changed |= UNTRACKED_CHANGED;
3128
		if (dir->untracked != istate->untracked) {
3129
			FREE_AND_NULL(dir->untracked);
3130
		}
3131
	}
3132

3133
	return dir->nr;
3134
}
3135

3136
int file_exists(const char *f)
3137
{
3138
	struct stat sb;
3139
	return lstat(f, &sb) == 0;
3140
}
3141

3142
int repo_file_exists(struct repository *repo, const char *path)
3143
{
3144
	if (repo != the_repository)
3145
		BUG("do not know how to check file existence in arbitrary repo");
3146

3147
	return file_exists(path);
3148
}
3149

3150
static int cmp_icase(char a, char b)
3151
{
3152
	if (a == b)
3153
		return 0;
3154
	if (ignore_case)
3155
		return toupper(a) - toupper(b);
3156
	return a - b;
3157
}
3158

3159
/*
3160
 * Given two normalized paths (a trailing slash is ok), if subdir is
3161
 * outside dir, return -1.  Otherwise return the offset in subdir that
3162
 * can be used as relative path to dir.
3163
 */
3164
int dir_inside_of(const char *subdir, const char *dir)
3165
{
3166
	int offset = 0;
3167

3168
	assert(dir && subdir && *dir && *subdir);
3169

3170
	while (*dir && *subdir && !cmp_icase(*dir, *subdir)) {
3171
		dir++;
3172
		subdir++;
3173
		offset++;
3174
	}
3175

3176
	/* hel[p]/me vs hel[l]/yeah */
3177
	if (*dir && *subdir)
3178
		return -1;
3179

3180
	if (!*subdir)
3181
		return !*dir ? offset : -1; /* same dir */
3182

3183
	/* foo/[b]ar vs foo/[] */
3184
	if (is_dir_sep(dir[-1]))
3185
		return is_dir_sep(subdir[-1]) ? offset : -1;
3186

3187
	/* foo[/]bar vs foo[] */
3188
	return is_dir_sep(*subdir) ? offset + 1 : -1;
3189
}
3190

3191
int is_inside_dir(const char *dir)
3192
{
3193
	char *cwd;
3194
	int rc;
3195

3196
	if (!dir)
3197
		return 0;
3198

3199
	cwd = xgetcwd();
3200
	rc = (dir_inside_of(cwd, dir) >= 0);
3201
	free(cwd);
3202
	return rc;
3203
}
3204

3205
int is_empty_dir(const char *path)
3206
{
3207
	DIR *dir = opendir(path);
3208
	struct dirent *e;
3209
	int ret = 1;
3210

3211
	if (!dir)
3212
		return 0;
3213

3214
	e = readdir_skip_dot_and_dotdot(dir);
3215
	if (e)
3216
		ret = 0;
3217

3218
	closedir(dir);
3219
	return ret;
3220
}
3221

3222
char *git_url_basename(const char *repo, int is_bundle, int is_bare)
3223
{
3224
	const char *end = repo + strlen(repo), *start, *ptr;
3225
	size_t len;
3226
	char *dir;
3227

3228
	/*
3229
	 * Skip scheme.
3230
	 */
3231
	start = strstr(repo, "://");
3232
	if (!start)
3233
		start = repo;
3234
	else
3235
		start += 3;
3236

3237
	/*
3238
	 * Skip authentication data. The stripping does happen
3239
	 * greedily, such that we strip up to the last '@' inside
3240
	 * the host part.
3241
	 */
3242
	for (ptr = start; ptr < end && !is_dir_sep(*ptr); ptr++) {
3243
		if (*ptr == '@')
3244
			start = ptr + 1;
3245
	}
3246

3247
	/*
3248
	 * Strip trailing spaces, slashes and /.git
3249
	 */
3250
	while (start < end && (is_dir_sep(end[-1]) || isspace(end[-1])))
3251
		end--;
3252
	if (end - start > 5 && is_dir_sep(end[-5]) &&
3253
	    !strncmp(end - 4, ".git", 4)) {
3254
		end -= 5;
3255
		while (start < end && is_dir_sep(end[-1]))
3256
			end--;
3257
	}
3258

3259
	/*
3260
	 * It should not be possible to overflow `ptrdiff_t` by passing in an
3261
	 * insanely long URL, but GCC does not know that and will complain
3262
	 * without this check.
3263
	 */
3264
	if (end - start < 0)
3265
		die(_("No directory name could be guessed.\n"
3266
		      "Please specify a directory on the command line"));
3267

3268
	/*
3269
	 * Strip trailing port number if we've got only a
3270
	 * hostname (that is, there is no dir separator but a
3271
	 * colon). This check is required such that we do not
3272
	 * strip URI's like '/foo/bar:2222.git', which should
3273
	 * result in a dir '2222' being guessed due to backwards
3274
	 * compatibility.
3275
	 */
3276
	if (memchr(start, '/', end - start) == NULL
3277
	    && memchr(start, ':', end - start) != NULL) {
3278
		ptr = end;
3279
		while (start < ptr && isdigit(ptr[-1]) && ptr[-1] != ':')
3280
			ptr--;
3281
		if (start < ptr && ptr[-1] == ':')
3282
			end = ptr - 1;
3283
	}
3284

3285
	/*
3286
	 * Find last component. To remain backwards compatible we
3287
	 * also regard colons as path separators, such that
3288
	 * cloning a repository 'foo:bar.git' would result in a
3289
	 * directory 'bar' being guessed.
3290
	 */
3291
	ptr = end;
3292
	while (start < ptr && !is_dir_sep(ptr[-1]) && ptr[-1] != ':')
3293
		ptr--;
3294
	start = ptr;
3295

3296
	/*
3297
	 * Strip .{bundle,git}.
3298
	 */
3299
	len = end - start;
3300
	strip_suffix_mem(start, &len, is_bundle ? ".bundle" : ".git");
3301

3302
	if (!len || (len == 1 && *start == '/'))
3303
		die(_("No directory name could be guessed.\n"
3304
		      "Please specify a directory on the command line"));
3305

3306
	if (is_bare)
3307
		dir = xstrfmt("%.*s.git", (int)len, start);
3308
	else
3309
		dir = xstrndup(start, len);
3310
	/*
3311
	 * Replace sequences of 'control' characters and whitespace
3312
	 * with one ascii space, remove leading and trailing spaces.
3313
	 */
3314
	if (*dir) {
3315
		char *out = dir;
3316
		int prev_space = 1 /* strip leading whitespace */;
3317
		for (end = dir; *end; ++end) {
3318
			char ch = *end;
3319
			if ((unsigned char)ch < '\x20')
3320
				ch = '\x20';
3321
			if (isspace(ch)) {
3322
				if (prev_space)
3323
					continue;
3324
				prev_space = 1;
3325
			} else
3326
				prev_space = 0;
3327
			*out++ = ch;
3328
		}
3329
		*out = '\0';
3330
		if (out > dir && prev_space)
3331
			out[-1] = '\0';
3332
	}
3333
	return dir;
3334
}
3335

3336
void strip_dir_trailing_slashes(char *dir)
3337
{
3338
	char *end = dir + strlen(dir);
3339

3340
	while (dir < end - 1 && is_dir_sep(end[-1]))
3341
		end--;
3342
	*end = '\0';
3343
}
3344

3345
static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
3346
{
3347
	DIR *dir;
3348
	struct dirent *e;
3349
	int ret = 0, original_len = path->len, len, kept_down = 0;
3350
	int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
3351
	int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
3352
	int purge_original_cwd = (flag & REMOVE_DIR_PURGE_ORIGINAL_CWD);
3353
	struct object_id submodule_head;
3354

3355
	if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
3356
	    !repo_resolve_gitlink_ref(the_repository, path->buf,
3357
				      "HEAD", &submodule_head)) {
3358
		/* Do not descend and nuke a nested git work tree. */
3359
		if (kept_up)
3360
			*kept_up = 1;
3361
		return 0;
3362
	}
3363

3364
	flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
3365
	dir = opendir(path->buf);
3366
	if (!dir) {
3367
		if (errno == ENOENT)
3368
			return keep_toplevel ? -1 : 0;
3369
		else if (errno == EACCES && !keep_toplevel)
3370
			/*
3371
			 * An empty dir could be removable even if it
3372
			 * is unreadable:
3373
			 */
3374
			return rmdir(path->buf);
3375
		else
3376
			return -1;
3377
	}
3378
	strbuf_complete(path, '/');
3379

3380
	len = path->len;
3381
	while ((e = readdir_skip_dot_and_dotdot(dir)) != NULL) {
3382
		struct stat st;
3383

3384
		strbuf_setlen(path, len);
3385
		strbuf_addstr(path, e->d_name);
3386
		if (lstat(path->buf, &st)) {
3387
			if (errno == ENOENT)
3388
				/*
3389
				 * file disappeared, which is what we
3390
				 * wanted anyway
3391
				 */
3392
				continue;
3393
			/* fall through */
3394
		} else if (S_ISDIR(st.st_mode)) {
3395
			if (!remove_dir_recurse(path, flag, &kept_down))
3396
				continue; /* happy */
3397
		} else if (!only_empty &&
3398
			   (!unlink(path->buf) || errno == ENOENT)) {
3399
			continue; /* happy, too */
3400
		}
3401

3402
		/* path too long, stat fails, or non-directory still exists */
3403
		ret = -1;
3404
		break;
3405
	}
3406
	closedir(dir);
3407

3408
	strbuf_setlen(path, original_len);
3409
	if (!ret && !keep_toplevel && !kept_down) {
3410
		if (!purge_original_cwd &&
3411
		    startup_info->original_cwd &&
3412
		    !strcmp(startup_info->original_cwd, path->buf))
3413
			ret = -1; /* Do not remove current working directory */
3414
		else
3415
			ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
3416
	} else if (kept_up)
3417
		/*
3418
		 * report the uplevel that it is not an error that we
3419
		 * did not rmdir() our directory.
3420
		 */
3421
		*kept_up = !ret;
3422
	return ret;
3423
}
3424

3425
int remove_dir_recursively(struct strbuf *path, int flag)
3426
{
3427
	return remove_dir_recurse(path, flag, NULL);
3428
}
3429

3430
static GIT_PATH_FUNC(git_path_info_exclude, "info/exclude")
3431

3432
void setup_standard_excludes(struct dir_struct *dir)
3433
{
3434
	dir->exclude_per_dir = ".gitignore";
3435

3436
	/* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3437
	if (!excludes_file)
3438
		excludes_file = xdg_config_home("ignore");
3439
	if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
3440
		add_patterns_from_file_1(dir, excludes_file,
3441
					 dir->untracked ? &dir->internal.ss_excludes_file : NULL);
3442

3443
	/* per repository user preference */
3444
	if (startup_info->have_repository) {
3445
		const char *path = git_path_info_exclude();
3446
		if (!access_or_warn(path, R_OK, 0))
3447
			add_patterns_from_file_1(dir, path,
3448
						 dir->untracked ? &dir->internal.ss_info_exclude : NULL);
3449
	}
3450
}
3451

3452
char *get_sparse_checkout_filename(void)
3453
{
3454
	return git_pathdup("info/sparse-checkout");
3455
}
3456

3457
int get_sparse_checkout_patterns(struct pattern_list *pl)
3458
{
3459
	int res;
3460
	char *sparse_filename = get_sparse_checkout_filename();
3461

3462
	pl->use_cone_patterns = core_sparse_checkout_cone;
3463
	res = add_patterns_from_file_to_list(sparse_filename, "", 0, pl, NULL, 0);
3464

3465
	free(sparse_filename);
3466
	return res;
3467
}
3468

3469
int remove_path(const char *name)
3470
{
3471
	char *slash;
3472

3473
	if (unlink(name) && !is_missing_file_error(errno))
3474
		return -1;
3475

3476
	slash = strrchr(name, '/');
3477
	if (slash) {
3478
		char *dirs = xstrdup(name);
3479
		slash = dirs + (slash - name);
3480
		do {
3481
			*slash = '\0';
3482
			if (startup_info->original_cwd &&
3483
			    !strcmp(startup_info->original_cwd, dirs))
3484
				break;
3485
		} while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
3486
		free(dirs);
3487
	}
3488
	return 0;
3489
}
3490

3491
/*
3492
 * Frees memory within dir which was allocated, and resets fields for further
3493
 * use.  Does not free dir itself.
3494
 */
3495
void dir_clear(struct dir_struct *dir)
3496
{
3497
	int i, j;
3498
	struct exclude_list_group *group;
3499
	struct pattern_list *pl;
3500
	struct exclude_stack *stk;
3501
	struct dir_struct new = DIR_INIT;
3502

3503
	for (i = EXC_CMDL; i <= EXC_FILE; i++) {
3504
		group = &dir->internal.exclude_list_group[i];
3505
		for (j = 0; j < group->nr; j++) {
3506
			pl = &group->pl[j];
3507
			if (i == EXC_DIRS)
3508
				free((char *)pl->src);
3509
			clear_pattern_list(pl);
3510
		}
3511
		free(group->pl);
3512
	}
3513

3514
	for (i = 0; i < dir->ignored_nr; i++)
3515
		free(dir->ignored[i]);
3516
	for (i = 0; i < dir->nr; i++)
3517
		free(dir->entries[i]);
3518
	free(dir->ignored);
3519
	free(dir->entries);
3520

3521
	stk = dir->internal.exclude_stack;
3522
	while (stk) {
3523
		struct exclude_stack *prev = stk->prev;
3524
		free(stk);
3525
		stk = prev;
3526
	}
3527
	strbuf_release(&dir->internal.basebuf);
3528

3529
	memcpy(dir, &new, sizeof(*dir));
3530
}
3531

3532
struct ondisk_untracked_cache {
3533
	struct stat_data info_exclude_stat;
3534
	struct stat_data excludes_file_stat;
3535
	uint32_t dir_flags;
3536
};
3537

3538
#define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3539

3540
struct write_data {
3541
	int index;	   /* number of written untracked_cache_dir */
3542
	struct ewah_bitmap *check_only; /* from untracked_cache_dir */
3543
	struct ewah_bitmap *valid;	/* from untracked_cache_dir */
3544
	struct ewah_bitmap *sha1_valid; /* set if exclude_sha1 is not null */
3545
	struct strbuf out;
3546
	struct strbuf sb_stat;
3547
	struct strbuf sb_sha1;
3548
};
3549

3550
static void stat_data_to_disk(struct stat_data *to, const struct stat_data *from)
3551
{
3552
	to->sd_ctime.sec  = htonl(from->sd_ctime.sec);
3553
	to->sd_ctime.nsec = htonl(from->sd_ctime.nsec);
3554
	to->sd_mtime.sec  = htonl(from->sd_mtime.sec);
3555
	to->sd_mtime.nsec = htonl(from->sd_mtime.nsec);
3556
	to->sd_dev	  = htonl(from->sd_dev);
3557
	to->sd_ino	  = htonl(from->sd_ino);
3558
	to->sd_uid	  = htonl(from->sd_uid);
3559
	to->sd_gid	  = htonl(from->sd_gid);
3560
	to->sd_size	  = htonl(from->sd_size);
3561
}
3562

3563
static void write_one_dir(struct untracked_cache_dir *untracked,
3564
			  struct write_data *wd)
3565
{
3566
	struct stat_data stat_data;
3567
	struct strbuf *out = &wd->out;
3568
	unsigned char intbuf[16];
3569
	unsigned int intlen, value;
3570
	int i = wd->index++;
3571

3572
	/*
3573
	 * untracked_nr should be reset whenever valid is clear, but
3574
	 * for safety..
3575
	 */
3576
	if (!untracked->valid) {
3577
		untracked->untracked_nr = 0;
3578
		untracked->check_only = 0;
3579
	}
3580

3581
	if (untracked->check_only)
3582
		ewah_set(wd->check_only, i);
3583
	if (untracked->valid) {
3584
		ewah_set(wd->valid, i);
3585
		stat_data_to_disk(&stat_data, &untracked->stat_data);
3586
		strbuf_add(&wd->sb_stat, &stat_data, sizeof(stat_data));
3587
	}
3588
	if (!is_null_oid(&untracked->exclude_oid)) {
3589
		ewah_set(wd->sha1_valid, i);
3590
		strbuf_add(&wd->sb_sha1, untracked->exclude_oid.hash,
3591
			   the_hash_algo->rawsz);
3592
	}
3593

3594
	intlen = encode_varint(untracked->untracked_nr, intbuf);
3595
	strbuf_add(out, intbuf, intlen);
3596

3597
	/* skip non-recurse directories */
3598
	for (i = 0, value = 0; i < untracked->dirs_nr; i++)
3599
		if (untracked->dirs[i]->recurse)
3600
			value++;
3601
	intlen = encode_varint(value, intbuf);
3602
	strbuf_add(out, intbuf, intlen);
3603

3604
	strbuf_add(out, untracked->name, strlen(untracked->name) + 1);
3605

3606
	for (i = 0; i < untracked->untracked_nr; i++)
3607
		strbuf_add(out, untracked->untracked[i],
3608
			   strlen(untracked->untracked[i]) + 1);
3609

3610
	for (i = 0; i < untracked->dirs_nr; i++)
3611
		if (untracked->dirs[i]->recurse)
3612
			write_one_dir(untracked->dirs[i], wd);
3613
}
3614

3615
void write_untracked_extension(struct strbuf *out, struct untracked_cache *untracked)
3616
{
3617
	struct ondisk_untracked_cache *ouc;
3618
	struct write_data wd;
3619
	unsigned char varbuf[16];
3620
	int varint_len;
3621
	const unsigned hashsz = the_hash_algo->rawsz;
3622

3623
	CALLOC_ARRAY(ouc, 1);
3624
	stat_data_to_disk(&ouc->info_exclude_stat, &untracked->ss_info_exclude.stat);
3625
	stat_data_to_disk(&ouc->excludes_file_stat, &untracked->ss_excludes_file.stat);
3626
	ouc->dir_flags = htonl(untracked->dir_flags);
3627

3628
	varint_len = encode_varint(untracked->ident.len, varbuf);
3629
	strbuf_add(out, varbuf, varint_len);
3630
	strbuf_addbuf(out, &untracked->ident);
3631

3632
	strbuf_add(out, ouc, sizeof(*ouc));
3633
	strbuf_add(out, untracked->ss_info_exclude.oid.hash, hashsz);
3634
	strbuf_add(out, untracked->ss_excludes_file.oid.hash, hashsz);
3635
	strbuf_add(out, untracked->exclude_per_dir, strlen(untracked->exclude_per_dir) + 1);
3636
	FREE_AND_NULL(ouc);
3637

3638
	if (!untracked->root) {
3639
		varint_len = encode_varint(0, varbuf);
3640
		strbuf_add(out, varbuf, varint_len);
3641
		return;
3642
	}
3643

3644
	wd.index      = 0;
3645
	wd.check_only = ewah_new();
3646
	wd.valid      = ewah_new();
3647
	wd.sha1_valid = ewah_new();
3648
	strbuf_init(&wd.out, 1024);
3649
	strbuf_init(&wd.sb_stat, 1024);
3650
	strbuf_init(&wd.sb_sha1, 1024);
3651
	write_one_dir(untracked->root, &wd);
3652

3653
	varint_len = encode_varint(wd.index, varbuf);
3654
	strbuf_add(out, varbuf, varint_len);
3655
	strbuf_addbuf(out, &wd.out);
3656
	ewah_serialize_strbuf(wd.valid, out);
3657
	ewah_serialize_strbuf(wd.check_only, out);
3658
	ewah_serialize_strbuf(wd.sha1_valid, out);
3659
	strbuf_addbuf(out, &wd.sb_stat);
3660
	strbuf_addbuf(out, &wd.sb_sha1);
3661
	strbuf_addch(out, '\0'); /* safe guard for string lists */
3662

3663
	ewah_free(wd.valid);
3664
	ewah_free(wd.check_only);
3665
	ewah_free(wd.sha1_valid);
3666
	strbuf_release(&wd.out);
3667
	strbuf_release(&wd.sb_stat);
3668
	strbuf_release(&wd.sb_sha1);
3669
}
3670

3671
static void free_untracked(struct untracked_cache_dir *ucd)
3672
{
3673
	int i;
3674
	if (!ucd)
3675
		return;
3676
	for (i = 0; i < ucd->dirs_nr; i++)
3677
		free_untracked(ucd->dirs[i]);
3678
	for (i = 0; i < ucd->untracked_nr; i++)
3679
		free(ucd->untracked[i]);
3680
	free(ucd->untracked);
3681
	free(ucd->dirs);
3682
	free(ucd);
3683
}
3684

3685
void free_untracked_cache(struct untracked_cache *uc)
3686
{
3687
	if (!uc)
3688
		return;
3689

3690
	free(uc->exclude_per_dir_to_free);
3691
	strbuf_release(&uc->ident);
3692
	free_untracked(uc->root);
3693
	free(uc);
3694
}
3695

3696
struct read_data {
3697
	int index;
3698
	struct untracked_cache_dir **ucd;
3699
	struct ewah_bitmap *check_only;
3700
	struct ewah_bitmap *valid;
3701
	struct ewah_bitmap *sha1_valid;
3702
	const unsigned char *data;
3703
	const unsigned char *end;
3704
};
3705

3706
static void stat_data_from_disk(struct stat_data *to, const unsigned char *data)
3707
{
3708
	memcpy(to, data, sizeof(*to));
3709
	to->sd_ctime.sec  = ntohl(to->sd_ctime.sec);
3710
	to->sd_ctime.nsec = ntohl(to->sd_ctime.nsec);
3711
	to->sd_mtime.sec  = ntohl(to->sd_mtime.sec);
3712
	to->sd_mtime.nsec = ntohl(to->sd_mtime.nsec);
3713
	to->sd_dev	  = ntohl(to->sd_dev);
3714
	to->sd_ino	  = ntohl(to->sd_ino);
3715
	to->sd_uid	  = ntohl(to->sd_uid);
3716
	to->sd_gid	  = ntohl(to->sd_gid);
3717
	to->sd_size	  = ntohl(to->sd_size);
3718
}
3719

3720
static int read_one_dir(struct untracked_cache_dir **untracked_,
3721
			struct read_data *rd)
3722
{
3723
	struct untracked_cache_dir ud, *untracked;
3724
	const unsigned char *data = rd->data, *end = rd->end;
3725
	const unsigned char *eos;
3726
	unsigned int value;
3727
	int i;
3728

3729
	memset(&ud, 0, sizeof(ud));
3730

3731
	value = decode_varint(&data);
3732
	if (data > end)
3733
		return -1;
3734
	ud.recurse	   = 1;
3735
	ud.untracked_alloc = value;
3736
	ud.untracked_nr	   = value;
3737
	if (ud.untracked_nr)
3738
		ALLOC_ARRAY(ud.untracked, ud.untracked_nr);
3739

3740
	ud.dirs_alloc = ud.dirs_nr = decode_varint(&data);
3741
	if (data > end)
3742
		return -1;
3743
	ALLOC_ARRAY(ud.dirs, ud.dirs_nr);
3744

3745
	eos = memchr(data, '\0', end - data);
3746
	if (!eos || eos == end)
3747
		return -1;
3748

3749
	*untracked_ = untracked = xmalloc(st_add3(sizeof(*untracked), eos - data, 1));
3750
	memcpy(untracked, &ud, sizeof(ud));
3751
	memcpy(untracked->name, data, eos - data + 1);
3752
	data = eos + 1;
3753

3754
	for (i = 0; i < untracked->untracked_nr; i++) {
3755
		eos = memchr(data, '\0', end - data);
3756
		if (!eos || eos == end)
3757
			return -1;
3758
		untracked->untracked[i] = xmemdupz(data, eos - data);
3759
		data = eos + 1;
3760
	}
3761

3762
	rd->ucd[rd->index++] = untracked;
3763
	rd->data = data;
3764

3765
	for (i = 0; i < untracked->dirs_nr; i++) {
3766
		if (read_one_dir(untracked->dirs + i, rd) < 0)
3767
			return -1;
3768
	}
3769
	return 0;
3770
}
3771

3772
static void set_check_only(size_t pos, void *cb)
3773
{
3774
	struct read_data *rd = cb;
3775
	struct untracked_cache_dir *ud = rd->ucd[pos];
3776
	ud->check_only = 1;
3777
}
3778

3779
static void read_stat(size_t pos, void *cb)
3780
{
3781
	struct read_data *rd = cb;
3782
	struct untracked_cache_dir *ud = rd->ucd[pos];
3783
	if (rd->data + sizeof(struct stat_data) > rd->end) {
3784
		rd->data = rd->end + 1;
3785
		return;
3786
	}
3787
	stat_data_from_disk(&ud->stat_data, rd->data);
3788
	rd->data += sizeof(struct stat_data);
3789
	ud->valid = 1;
3790
}
3791

3792
static void read_oid(size_t pos, void *cb)
3793
{
3794
	struct read_data *rd = cb;
3795
	struct untracked_cache_dir *ud = rd->ucd[pos];
3796
	if (rd->data + the_hash_algo->rawsz > rd->end) {
3797
		rd->data = rd->end + 1;
3798
		return;
3799
	}
3800
	oidread(&ud->exclude_oid, rd->data, the_repository->hash_algo);
3801
	rd->data += the_hash_algo->rawsz;
3802
}
3803

3804
static void load_oid_stat(struct oid_stat *oid_stat, const unsigned char *data,
3805
			  const unsigned char *sha1)
3806
{
3807
	stat_data_from_disk(&oid_stat->stat, data);
3808
	oidread(&oid_stat->oid, sha1, the_repository->hash_algo);
3809
	oid_stat->valid = 1;
3810
}
3811

3812
struct untracked_cache *read_untracked_extension(const void *data, unsigned long sz)
3813
{
3814
	struct untracked_cache *uc;
3815
	struct read_data rd;
3816
	const unsigned char *next = data, *end = (const unsigned char *)data + sz;
3817
	const char *ident;
3818
	int ident_len;
3819
	ssize_t len;
3820
	const char *exclude_per_dir;
3821
	const unsigned hashsz = the_hash_algo->rawsz;
3822
	const unsigned offset = sizeof(struct ondisk_untracked_cache);
3823
	const unsigned exclude_per_dir_offset = offset + 2 * hashsz;
3824

3825
	if (sz <= 1 || end[-1] != '\0')
3826
		return NULL;
3827
	end--;
3828

3829
	ident_len = decode_varint(&next);
3830
	if (next + ident_len > end)
3831
		return NULL;
3832
	ident = (const char *)next;
3833
	next += ident_len;
3834

3835
	if (next + exclude_per_dir_offset + 1 > end)
3836
		return NULL;
3837

3838
	CALLOC_ARRAY(uc, 1);
3839
	strbuf_init(&uc->ident, ident_len);
3840
	strbuf_add(&uc->ident, ident, ident_len);
3841
	load_oid_stat(&uc->ss_info_exclude,
3842
		      next + ouc_offset(info_exclude_stat),
3843
		      next + offset);
3844
	load_oid_stat(&uc->ss_excludes_file,
3845
		      next + ouc_offset(excludes_file_stat),
3846
		      next + offset + hashsz);
3847
	uc->dir_flags = get_be32(next + ouc_offset(dir_flags));
3848
	exclude_per_dir = (const char *)next + exclude_per_dir_offset;
3849
	uc->exclude_per_dir = uc->exclude_per_dir_to_free = xstrdup(exclude_per_dir);
3850
	/* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3851
	next += exclude_per_dir_offset + strlen(exclude_per_dir) + 1;
3852
	if (next >= end)
3853
		goto done2;
3854

3855
	len = decode_varint(&next);
3856
	if (next > end || len == 0)
3857
		goto done2;
3858

3859
	rd.valid      = ewah_new();
3860
	rd.check_only = ewah_new();
3861
	rd.sha1_valid = ewah_new();
3862
	rd.data	      = next;
3863
	rd.end	      = end;
3864
	rd.index      = 0;
3865
	ALLOC_ARRAY(rd.ucd, len);
3866

3867
	if (read_one_dir(&uc->root, &rd) || rd.index != len)
3868
		goto done;
3869

3870
	next = rd.data;
3871
	len = ewah_read_mmap(rd.valid, next, end - next);
3872
	if (len < 0)
3873
		goto done;
3874

3875
	next += len;
3876
	len = ewah_read_mmap(rd.check_only, next, end - next);
3877
	if (len < 0)
3878
		goto done;
3879

3880
	next += len;
3881
	len = ewah_read_mmap(rd.sha1_valid, next, end - next);
3882
	if (len < 0)
3883
		goto done;
3884

3885
	ewah_each_bit(rd.check_only, set_check_only, &rd);
3886
	rd.data = next + len;
3887
	ewah_each_bit(rd.valid, read_stat, &rd);
3888
	ewah_each_bit(rd.sha1_valid, read_oid, &rd);
3889
	next = rd.data;
3890

3891
done:
3892
	free(rd.ucd);
3893
	ewah_free(rd.valid);
3894
	ewah_free(rd.check_only);
3895
	ewah_free(rd.sha1_valid);
3896
done2:
3897
	if (next != end) {
3898
		free_untracked_cache(uc);
3899
		uc = NULL;
3900
	}
3901
	return uc;
3902
}
3903

3904
static void invalidate_one_directory(struct untracked_cache *uc,
3905
				     struct untracked_cache_dir *ucd)
3906
{
3907
	uc->dir_invalidated++;
3908
	ucd->valid = 0;
3909
	ucd->untracked_nr = 0;
3910
}
3911

3912
/*
3913
 * Normally when an entry is added or removed from a directory,
3914
 * invalidating that directory is enough. No need to touch its
3915
 * ancestors. When a directory is shown as "foo/bar/" in git-status
3916
 * however, deleting or adding an entry may have cascading effect.
3917
 *
3918
 * Say the "foo/bar/file" has become untracked, we need to tell the
3919
 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3920
 * directory any more (because "bar" is managed by foo as an untracked
3921
 * "file").
3922
 *
3923
 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3924
 * was the last untracked entry in the entire "foo", we should show
3925
 * "foo/" instead. Which means we have to invalidate past "bar" up to
3926
 * "foo".
3927
 *
3928
 * This function traverses all directories from root to leaf. If there
3929
 * is a chance of one of the above cases happening, we invalidate back
3930
 * to root. Otherwise we just invalidate the leaf. There may be a more
3931
 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3932
 * detect these cases and avoid unnecessary invalidation, for example,
3933
 * checking for the untracked entry named "bar/" in "foo", but for now
3934
 * stick to something safe and simple.
3935
 */
3936
static int invalidate_one_component(struct untracked_cache *uc,
3937
				    struct untracked_cache_dir *dir,
3938
				    const char *path, int len)
3939
{
3940
	const char *rest = strchr(path, '/');
3941

3942
	if (rest) {
3943
		int component_len = rest - path;
3944
		struct untracked_cache_dir *d =
3945
			lookup_untracked(uc, dir, path, component_len);
3946
		int ret =
3947
			invalidate_one_component(uc, d, rest + 1,
3948
						 len - (component_len + 1));
3949
		if (ret)
3950
			invalidate_one_directory(uc, dir);
3951
		return ret;
3952
	}
3953

3954
	invalidate_one_directory(uc, dir);
3955
	return uc->dir_flags & DIR_SHOW_OTHER_DIRECTORIES;
3956
}
3957

3958
void untracked_cache_invalidate_path(struct index_state *istate,
3959
				     const char *path, int safe_path)
3960
{
3961
	if (!istate->untracked || !istate->untracked->root)
3962
		return;
3963
	if (!safe_path && !verify_path(path, 0))
3964
		return;
3965
	invalidate_one_component(istate->untracked, istate->untracked->root,
3966
				 path, strlen(path));
3967
}
3968

3969
void untracked_cache_invalidate_trimmed_path(struct index_state *istate,
3970
					     const char *path,
3971
					     int safe_path)
3972
{
3973
	size_t len = strlen(path);
3974

3975
	if (!len)
3976
		BUG("untracked_cache_invalidate_trimmed_path given zero length path");
3977

3978
	if (path[len - 1] != '/') {
3979
		untracked_cache_invalidate_path(istate, path, safe_path);
3980
	} else {
3981
		struct strbuf tmp = STRBUF_INIT;
3982

3983
		strbuf_add(&tmp, path, len - 1);
3984
		untracked_cache_invalidate_path(istate, tmp.buf, safe_path);
3985
		strbuf_release(&tmp);
3986
	}
3987
}
3988

3989
void untracked_cache_remove_from_index(struct index_state *istate,
3990
				       const char *path)
3991
{
3992
	untracked_cache_invalidate_path(istate, path, 1);
3993
}
3994

3995
void untracked_cache_add_to_index(struct index_state *istate,
3996
				  const char *path)
3997
{
3998
	untracked_cache_invalidate_path(istate, path, 1);
3999
}
4000

4001
static void connect_wt_gitdir_in_nested(const char *sub_worktree,
4002
					const char *sub_gitdir)
4003
{
4004
	int i;
4005
	struct repository subrepo;
4006
	struct strbuf sub_wt = STRBUF_INIT;
4007
	struct strbuf sub_gd = STRBUF_INIT;
4008

4009
	const struct submodule *sub;
4010

4011
	/* If the submodule has no working tree, we can ignore it. */
4012
	if (repo_init(&subrepo, sub_gitdir, sub_worktree))
4013
		return;
4014

4015
	if (repo_read_index(&subrepo) < 0)
4016
		die(_("index file corrupt in repo %s"), subrepo.gitdir);
4017

4018
	/* TODO: audit for interaction with sparse-index. */
4019
	ensure_full_index(subrepo.index);
4020
	for (i = 0; i < subrepo.index->cache_nr; i++) {
4021
		const struct cache_entry *ce = subrepo.index->cache[i];
4022

4023
		if (!S_ISGITLINK(ce->ce_mode))
4024
			continue;
4025

4026
		while (i + 1 < subrepo.index->cache_nr &&
4027
		       !strcmp(ce->name, subrepo.index->cache[i + 1]->name))
4028
			/*
4029
			 * Skip entries with the same name in different stages
4030
			 * to make sure an entry is returned only once.
4031
			 */
4032
			i++;
4033

4034
		sub = submodule_from_path(&subrepo, null_oid(), ce->name);
4035
		if (!sub || !is_submodule_active(&subrepo, ce->name))
4036
			/* .gitmodules broken or inactive sub */
4037
			continue;
4038

4039
		strbuf_reset(&sub_wt);
4040
		strbuf_reset(&sub_gd);
4041
		strbuf_addf(&sub_wt, "%s/%s", sub_worktree, sub->path);
4042
		submodule_name_to_gitdir(&sub_gd, &subrepo, sub->name);
4043

4044
		connect_work_tree_and_git_dir(sub_wt.buf, sub_gd.buf, 1);
4045
	}
4046
	strbuf_release(&sub_wt);
4047
	strbuf_release(&sub_gd);
4048
	repo_clear(&subrepo);
4049
}
4050

4051
void connect_work_tree_and_git_dir(const char *work_tree_,
4052
				   const char *git_dir_,
4053
				   int recurse_into_nested)
4054
{
4055
	struct strbuf gitfile_sb = STRBUF_INIT;
4056
	struct strbuf cfg_sb = STRBUF_INIT;
4057
	struct strbuf rel_path = STRBUF_INIT;
4058
	char *git_dir, *work_tree;
4059

4060
	/* Prepare .git file */
4061
	strbuf_addf(&gitfile_sb, "%s/.git", work_tree_);
4062
	if (safe_create_leading_directories_const(gitfile_sb.buf))
4063
		die(_("could not create directories for %s"), gitfile_sb.buf);
4064

4065
	/* Prepare config file */
4066
	strbuf_addf(&cfg_sb, "%s/config", git_dir_);
4067
	if (safe_create_leading_directories_const(cfg_sb.buf))
4068
		die(_("could not create directories for %s"), cfg_sb.buf);
4069

4070
	git_dir = real_pathdup(git_dir_, 1);
4071
	work_tree = real_pathdup(work_tree_, 1);
4072

4073
	/* Write .git file */
4074
	write_file(gitfile_sb.buf, "gitdir: %s",
4075
		   relative_path(git_dir, work_tree, &rel_path));
4076
	/* Update core.worktree setting */
4077
	git_config_set_in_file(cfg_sb.buf, "core.worktree",
4078
			       relative_path(work_tree, git_dir, &rel_path));
4079

4080
	strbuf_release(&gitfile_sb);
4081
	strbuf_release(&cfg_sb);
4082
	strbuf_release(&rel_path);
4083

4084
	if (recurse_into_nested)
4085
		connect_wt_gitdir_in_nested(work_tree, git_dir);
4086

4087
	free(work_tree);
4088
	free(git_dir);
4089
}
4090

4091
/*
4092
 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
4093
 */
4094
void relocate_gitdir(const char *path, const char *old_git_dir, const char *new_git_dir)
4095
{
4096
	if (rename(old_git_dir, new_git_dir) < 0)
4097
		die_errno(_("could not migrate git directory from '%s' to '%s'"),
4098
			old_git_dir, new_git_dir);
4099

4100
	connect_work_tree_and_git_dir(path, new_git_dir, 0);
4101
}
4102

4103
int path_match_flags(const char *const str, const enum path_match_flags flags)
4104
{
4105
	const char *p = str;
4106

4107
	if (flags & PATH_MATCH_NATIVE &&
4108
	    flags & PATH_MATCH_XPLATFORM)
4109
		BUG("path_match_flags() must get one match kind, not multiple!");
4110
	else if (!(flags & PATH_MATCH_KINDS_MASK))
4111
		BUG("path_match_flags() must get at least one match kind!");
4112

4113
	if (flags & PATH_MATCH_STARTS_WITH_DOT_SLASH &&
4114
	    flags & PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH)
4115
		BUG("path_match_flags() must get one platform kind, not multiple!");
4116
	else if (!(flags & PATH_MATCH_PLATFORM_MASK))
4117
		BUG("path_match_flags() must get at least one platform kind!");
4118

4119
	if (*p++ != '.')
4120
		return 0;
4121
	if (flags & PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH &&
4122
	    *p++ != '.')
4123
		return 0;
4124

4125
	if (flags & PATH_MATCH_NATIVE)
4126
		return is_dir_sep(*p);
4127
	else if (flags & PATH_MATCH_XPLATFORM)
4128
		return is_xplatform_dir_sep(*p);
4129
	BUG("unreachable");
4130
}
4131