2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
8 #include "git-compat-util.h"
14 #include "environment.h"
16 #include "name-hash.h"
17 #include "object-file.h"
18 #include "object-store.h"
21 #include "wildmatch.h"
25 #include "ewah/ewok.h"
26 #include "fsmonitor.h"
28 #include "submodule-config.h"
35 * Tells read_directory_recursive how a file or directory should be treated.
36 * Values are ordered by significance, e.g. if a directory contains both
37 * excluded and untracked files, it is listed as untracked because
38 * path_untracked > path_excluded.
48 * Support data structure for our opendir/readdir/closedir wrappers
52 struct untracked_cache_dir
*untracked
;
59 struct untracked_cache_dir
*ucd
;
62 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
63 struct index_state
*istate
, const char *path
, int len
,
64 struct untracked_cache_dir
*untracked
,
65 int check_only
, int stop_at_first_file
, const struct pathspec
*pathspec
);
66 static int resolve_dtype(int dtype
, struct index_state
*istate
,
67 const char *path
, int len
);
68 struct dirent
*readdir_skip_dot_and_dotdot(DIR *dirp
)
72 while ((e
= readdir(dirp
)) != NULL
) {
73 if (!is_dot_or_dotdot(e
->d_name
))
79 int count_slashes(const char *s
)
88 int fspathcmp(const char *a
, const char *b
)
90 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
93 int fspatheq(const char *a
, const char *b
)
95 return !fspathcmp(a
, b
);
98 int fspathncmp(const char *a
, const char *b
, size_t count
)
100 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
103 unsigned int fspathhash(const char *str
)
105 return ignore_case
? strihash(str
) : strhash(str
);
108 int git_fnmatch(const struct pathspec_item
*item
,
109 const char *pattern
, const char *string
,
113 if (ps_strncmp(item
, pattern
, string
, prefix
))
118 if (item
->flags
& PATHSPEC_ONESTAR
) {
119 int pattern_len
= strlen(++pattern
);
120 int string_len
= strlen(string
);
121 return string_len
< pattern_len
||
122 ps_strcmp(item
, pattern
,
123 string
+ string_len
- pattern_len
);
125 if (item
->magic
& PATHSPEC_GLOB
)
126 return wildmatch(pattern
, string
,
128 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
130 /* wildmatch has not learned no FNM_PATHNAME mode yet */
131 return wildmatch(pattern
, string
,
132 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
135 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
136 const char *string
, int stringlen
,
140 struct strbuf pat_buf
= STRBUF_INIT
;
141 struct strbuf str_buf
= STRBUF_INIT
;
142 const char *use_pat
= pattern
;
143 const char *use_str
= string
;
145 if (pattern
[patternlen
]) {
146 strbuf_add(&pat_buf
, pattern
, patternlen
);
147 use_pat
= pat_buf
.buf
;
149 if (string
[stringlen
]) {
150 strbuf_add(&str_buf
, string
, stringlen
);
151 use_str
= str_buf
.buf
;
155 flags
|= WM_CASEFOLD
;
156 match_status
= wildmatch(use_pat
, use_str
, flags
);
158 strbuf_release(&pat_buf
);
159 strbuf_release(&str_buf
);
164 static size_t common_prefix_len(const struct pathspec
*pathspec
)
170 * ":(icase)path" is treated as a pathspec full of
171 * wildcard. In other words, only prefix is considered common
172 * prefix. If the pathspec is abc/foo abc/bar, running in
173 * subdir xyz, the common prefix is still xyz, not xyz/abc as
176 GUARD_PATHSPEC(pathspec
,
185 for (n
= 0; n
< pathspec
->nr
; n
++) {
186 size_t i
= 0, len
= 0, item_len
;
187 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
189 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
190 item_len
= pathspec
->items
[n
].prefix
;
192 item_len
= pathspec
->items
[n
].nowildcard_len
;
193 while (i
< item_len
&& (n
== 0 || i
< max
)) {
194 char c
= pathspec
->items
[n
].match
[i
];
195 if (c
!= pathspec
->items
[0].match
[i
])
201 if (n
== 0 || len
< max
) {
211 * Returns a copy of the longest leading path common among all
214 char *common_prefix(const struct pathspec
*pathspec
)
216 unsigned long len
= common_prefix_len(pathspec
);
218 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
221 int fill_directory(struct dir_struct
*dir
,
222 struct index_state
*istate
,
223 const struct pathspec
*pathspec
)
228 unsigned exclusive_flags
= DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
;
229 if ((dir
->flags
& exclusive_flags
) == exclusive_flags
)
230 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
233 * Calculate common prefix for the pathspec, and
234 * use that to optimize the directory walk
236 prefix_len
= common_prefix_len(pathspec
);
237 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
239 /* Read the directory and prune it */
240 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
245 int within_depth(const char *name
, int namelen
,
246 int depth
, int max_depth
)
248 const char *cp
= name
, *cpe
= name
+ namelen
;
254 if (depth
> max_depth
)
261 * Read the contents of the blob with the given OID into a buffer.
262 * Append a trailing LF to the end if the last line doesn't have one.
265 * -1 when the OID is invalid or unknown or does not refer to a blob.
266 * 0 when the blob is empty.
267 * 1 along with { data, size } of the (possibly augmented) buffer
270 * Optionally updates the given oid_stat with the given OID (when valid).
272 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
273 size_t *size_out
, char **data_out
)
275 enum object_type type
;
282 data
= repo_read_object_file(the_repository
, oid
, &type
, &sz
);
283 if (!data
|| type
!= OBJ_BLOB
) {
289 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
290 oidcpy(&oid_stat
->oid
, oid
);
298 if (data
[sz
- 1] != '\n') {
299 data
= xrealloc(data
, st_add(sz
, 1));
303 *size_out
= xsize_t(sz
);
309 #define DO_MATCH_EXCLUDE (1<<0)
310 #define DO_MATCH_DIRECTORY (1<<1)
311 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
314 * Does the given pathspec match the given name? A match is found if
316 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
317 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
318 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
319 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
321 * Return value tells which case it was (1-4), or 0 when there is no match.
323 * It may be instructive to look at a small table of concrete examples
324 * to understand the differences between 1, 2, and 4:
327 * | a/b | a/b/ | a/b/c
328 * ------+-----------+-----------+------------
329 * a/b | EXACT | EXACT[1] | LEADING[2]
330 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
331 * a/b/c | RECURSIVE | RECURSIVE | EXACT
333 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
334 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
336 static int match_pathspec_item(struct index_state
*istate
,
337 const struct pathspec_item
*item
, int prefix
,
338 const char *name
, int namelen
, unsigned flags
)
340 /* name/namelen has prefix cut off by caller */
341 const char *match
= item
->match
+ prefix
;
342 int matchlen
= item
->len
- prefix
;
345 * The normal call pattern is:
346 * 1. prefix = common_prefix_len(ps);
347 * 2. prune something, or fill_directory
348 * 3. match_pathspec()
350 * 'prefix' at #1 may be shorter than the command's prefix and
351 * it's ok for #2 to match extra files. Those extras will be
354 * Suppose the pathspec is 'foo' and '../bar' running from
355 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
356 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
357 * user does not want XYZ/foo, only the "foo" part should be
358 * case-insensitive. We need to filter out XYZ/foo here. In
359 * other words, we do not trust the caller on comparing the
360 * prefix part when :(icase) is involved. We do exact
361 * comparison ourselves.
363 * Normally the caller (common_prefix_len() in fact) does
364 * _exact_ matching on name[-prefix+1..-1] and we do not need
365 * to check that part. Be defensive and check it anyway, in
366 * case common_prefix_len is changed, or a new caller is
367 * introduced that does not use common_prefix_len.
369 * If the penalty turns out too high when prefix is really
370 * long, maybe change it to
371 * strncmp(match, name, item->prefix - prefix)
373 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
374 strncmp(item
->match
, name
- prefix
, item
->prefix
))
377 if (item
->attr_match_nr
&&
378 !match_pathspec_attrs(istate
, name
, namelen
, item
))
381 /* If the match was just the prefix, we matched */
383 return MATCHED_RECURSIVELY
;
385 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
386 if (matchlen
== namelen
)
387 return MATCHED_EXACTLY
;
389 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
390 return MATCHED_RECURSIVELY
;
391 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
392 match
[matchlen
- 1] == '/' &&
393 namelen
== matchlen
- 1 &&
394 !ps_strncmp(item
, match
, name
, namelen
))
395 return MATCHED_EXACTLY
;
397 if (item
->nowildcard_len
< item
->len
&&
398 !git_fnmatch(item
, match
, name
,
399 item
->nowildcard_len
- prefix
))
400 return MATCHED_FNMATCH
;
402 /* Perform checks to see if "name" is a leading string of the pathspec */
403 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
404 !(flags
& DO_MATCH_EXCLUDE
)) {
405 /* name is a literal prefix of the pathspec */
406 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
407 if ((namelen
< matchlen
) &&
408 (match
[namelen
-offset
] == '/') &&
409 !ps_strncmp(item
, match
, name
, namelen
))
410 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
412 /* name doesn't match up to the first wild character */
413 if (item
->nowildcard_len
< item
->len
&&
414 ps_strncmp(item
, match
, name
,
415 item
->nowildcard_len
- prefix
))
419 * name has no wildcard, and it didn't match as a leading
420 * pathspec so return.
422 if (item
->nowildcard_len
== item
->len
)
426 * Here is where we would perform a wildmatch to check if
427 * "name" can be matched as a directory (or a prefix) against
428 * the pathspec. Since wildmatch doesn't have this capability
429 * at the present we have to punt and say that it is a match,
430 * potentially returning a false positive
431 * The submodules themselves will be able to perform more
432 * accurate matching to determine if the pathspec matches.
434 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
441 * do_match_pathspec() is meant to ONLY be called by
442 * match_pathspec_with_flags(); calling it directly risks pathspecs
443 * like ':!unwanted_path' being ignored.
445 * Given a name and a list of pathspecs, returns the nature of the
446 * closest (i.e. most specific) match of the name to any of the
449 * The caller typically calls this multiple times with the same
450 * pathspec and seen[] array but with different name/namelen
451 * (e.g. entries from the index) and is interested in seeing if and
452 * how each pathspec matches all the names it calls this function
453 * with. A mark is left in the seen[] array for each pathspec element
454 * indicating the closest type of match that element achieved, so if
455 * seen[n] remains zero after multiple invocations, that means the nth
456 * pathspec did not match any names, which could indicate that the
457 * user mistyped the nth pathspec.
459 static int do_match_pathspec(struct index_state
*istate
,
460 const struct pathspec
*ps
,
461 const char *name
, int namelen
,
462 int prefix
, char *seen
,
465 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
477 if (!ps
->recursive
||
478 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
480 return MATCHED_RECURSIVELY
;
482 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
483 return MATCHED_EXACTLY
;
491 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
494 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
495 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
498 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
501 * Make exclude patterns optional and never report
502 * "pathspec ':(exclude)foo' matches no files"
504 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
505 seen
[i
] = MATCHED_FNMATCH
;
506 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
509 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
510 ps
->max_depth
!= -1 &&
511 how
&& how
!= MATCHED_FNMATCH
) {
512 int len
= ps
->items
[i
].len
;
513 if (name
[len
] == '/')
515 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
516 how
= MATCHED_EXACTLY
;
523 if (seen
&& seen
[i
] < how
)
530 static int match_pathspec_with_flags(struct index_state
*istate
,
531 const struct pathspec
*ps
,
532 const char *name
, int namelen
,
533 int prefix
, char *seen
, unsigned flags
)
535 int positive
, negative
;
536 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
537 prefix
, seen
, flags
);
538 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
540 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
542 flags
| DO_MATCH_EXCLUDE
);
543 return negative
? 0 : positive
;
546 int match_pathspec(struct index_state
*istate
,
547 const struct pathspec
*ps
,
548 const char *name
, int namelen
,
549 int prefix
, char *seen
, int is_dir
)
551 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
552 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
553 prefix
, seen
, flags
);
557 * Check if a submodule is a superset of the pathspec
559 int submodule_path_match(struct index_state
*istate
,
560 const struct pathspec
*ps
,
561 const char *submodule_name
,
564 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
565 strlen(submodule_name
),
568 DO_MATCH_LEADING_PATHSPEC
);
572 int report_path_error(const char *ps_matched
,
573 const struct pathspec
*pathspec
)
576 * Make sure all pathspec matched; otherwise it is an error.
579 for (num
= 0; num
< pathspec
->nr
; num
++) {
580 int other
, found_dup
;
585 * The caller might have fed identical pathspec
586 * twice. Do not barf on such a mistake.
587 * FIXME: parse_pathspec should have eliminated
588 * duplicate pathspec.
590 for (found_dup
= other
= 0;
591 !found_dup
&& other
< pathspec
->nr
;
593 if (other
== num
|| !ps_matched
[other
])
595 if (!strcmp(pathspec
->items
[other
].original
,
596 pathspec
->items
[num
].original
))
598 * Ok, we have a match already.
605 error(_("pathspec '%s' did not match any file(s) known to git"),
606 pathspec
->items
[num
].original
);
613 * Return the length of the "simple" part of a path match limiter.
615 int simple_length(const char *match
)
620 unsigned char c
= *match
++;
622 if (c
== '\0' || is_glob_special(c
))
627 int no_wildcard(const char *string
)
629 return string
[simple_length(string
)] == '\0';
632 void parse_path_pattern(const char **pattern
,
637 const char *p
= *pattern
;
642 *flags
|= PATTERN_FLAG_NEGATIVE
;
646 if (len
&& p
[len
- 1] == '/') {
648 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
650 for (i
= 0; i
< len
; i
++) {
655 *flags
|= PATTERN_FLAG_NODIR
;
656 *nowildcardlen
= simple_length(p
);
658 * we should have excluded the trailing slash from 'p' too,
659 * but that's one more allocation. Instead just make sure
660 * nowildcardlen does not exceed real patternlen
662 if (*nowildcardlen
> len
)
663 *nowildcardlen
= len
;
664 if (*p
== '*' && no_wildcard(p
+ 1))
665 *flags
|= PATTERN_FLAG_ENDSWITH
;
670 int pl_hashmap_cmp(const void *cmp_data UNUSED
,
671 const struct hashmap_entry
*a
,
672 const struct hashmap_entry
*b
,
673 const void *key UNUSED
)
675 const struct pattern_entry
*ee1
=
676 container_of(a
, struct pattern_entry
, ent
);
677 const struct pattern_entry
*ee2
=
678 container_of(b
, struct pattern_entry
, ent
);
680 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
684 return fspathncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
687 static char *dup_and_filter_pattern(const char *pattern
)
691 char *result
= xstrdup(pattern
);
697 /* skip escape characters (once) */
717 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
719 struct pattern_entry
*translated
;
722 const char *prev
, *cur
, *next
;
724 if (!pl
->use_cone_patterns
)
727 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
728 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
729 !strcmp(given
->pattern
, "/*")) {
734 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
739 if (given
->patternlen
< 2 ||
740 *given
->pattern
!= '/' ||
741 strstr(given
->pattern
, "**")) {
742 /* Not a cone pattern. */
743 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
747 if (!(given
->flags
& PATTERN_FLAG_MUSTBEDIR
) &&
748 strcmp(given
->pattern
, "/*")) {
749 /* Not a cone pattern. */
750 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
754 prev
= given
->pattern
;
755 cur
= given
->pattern
+ 1;
756 next
= given
->pattern
+ 2;
759 /* Watch for glob characters '*', '\', '[', '?' */
760 if (!is_glob_special(*cur
))
763 /* But only if *prev != '\\' */
767 /* But allow the initial '\' */
769 is_glob_special(*next
))
772 /* But a trailing '/' then '*' is fine */
778 /* Not a cone pattern. */
779 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
788 if (given
->patternlen
> 2 &&
789 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
790 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
791 /* Not a cone pattern. */
792 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
796 truncated
= dup_and_filter_pattern(given
->pattern
);
798 translated
= xmalloc(sizeof(struct pattern_entry
));
799 translated
->pattern
= truncated
;
800 translated
->patternlen
= given
->patternlen
- 2;
801 hashmap_entry_init(&translated
->ent
,
802 fspathhash(translated
->pattern
));
804 if (!hashmap_get_entry(&pl
->recursive_hashmap
,
805 translated
, ent
, NULL
)) {
806 /* We did not see the "parent" included */
807 warning(_("unrecognized negative pattern: '%s'"),
814 hashmap_add(&pl
->parent_hashmap
, &translated
->ent
);
815 hashmap_remove(&pl
->recursive_hashmap
, &translated
->ent
, &data
);
820 if (given
->flags
& PATTERN_FLAG_NEGATIVE
) {
821 warning(_("unrecognized negative pattern: '%s'"),
826 translated
= xmalloc(sizeof(struct pattern_entry
));
828 translated
->pattern
= dup_and_filter_pattern(given
->pattern
);
829 translated
->patternlen
= given
->patternlen
;
830 hashmap_entry_init(&translated
->ent
,
831 fspathhash(translated
->pattern
));
833 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
835 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
836 /* we already included this at the parent level */
837 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
845 warning(_("disabling cone pattern matching"));
846 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
847 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
848 pl
->use_cone_patterns
= 0;
851 static int hashmap_contains_path(struct hashmap
*map
,
852 struct strbuf
*pattern
)
854 struct pattern_entry p
;
856 /* Check straight mapping */
857 p
.pattern
= pattern
->buf
;
858 p
.patternlen
= pattern
->len
;
859 hashmap_entry_init(&p
.ent
, fspathhash(p
.pattern
));
860 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
863 int hashmap_contains_parent(struct hashmap
*map
,
865 struct strbuf
*buffer
)
869 strbuf_setlen(buffer
, 0);
872 strbuf_addch(buffer
, '/');
874 strbuf_addstr(buffer
, path
);
876 slash_pos
= strrchr(buffer
->buf
, '/');
878 while (slash_pos
> buffer
->buf
) {
879 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
881 if (hashmap_contains_path(map
, buffer
))
884 slash_pos
= strrchr(buffer
->buf
, '/');
890 void add_pattern(const char *string
, const char *base
,
891 int baselen
, struct pattern_list
*pl
, int srcpos
)
893 struct path_pattern
*pattern
;
898 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
899 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
900 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
902 pattern
= xmalloc(sizeof(*pattern
));
903 pattern
->pattern
= string
;
905 pattern
->patternlen
= patternlen
;
906 pattern
->nowildcardlen
= nowildcardlen
;
907 pattern
->base
= base
;
908 pattern
->baselen
= baselen
;
909 pattern
->flags
= flags
;
910 pattern
->srcpos
= srcpos
;
911 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
912 pl
->patterns
[pl
->nr
++] = pattern
;
915 add_pattern_to_hashsets(pl
, pattern
);
918 static int read_skip_worktree_file_from_index(struct index_state
*istate
,
920 size_t *size_out
, char **data_out
,
921 struct oid_stat
*oid_stat
)
926 pos
= index_name_pos(istate
, path
, len
);
929 if (!ce_skip_worktree(istate
->cache
[pos
]))
932 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
936 * Frees memory within pl which was allocated for exclude patterns and
937 * the file buffer. Does not free pl itself.
939 void clear_pattern_list(struct pattern_list
*pl
)
943 for (i
= 0; i
< pl
->nr
; i
++)
944 free(pl
->patterns
[i
]);
947 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
948 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
950 memset(pl
, 0, sizeof(*pl
));
953 static void trim_trailing_spaces(char *buf
)
955 char *p
, *last_space
= NULL
;
957 for (p
= buf
; *p
; p
++)
977 * Given a subdirectory name and "dir" of the current directory,
978 * search the subdir in "dir" and return it, or create a new one if it
979 * does not exist in "dir".
981 * If "name" has the trailing slash, it'll be excluded in the search.
983 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
984 struct untracked_cache_dir
*dir
,
985 const char *name
, int len
)
988 struct untracked_cache_dir
*d
;
991 if (len
&& name
[len
- 1] == '/')
995 while (last
> first
) {
996 int cmp
, next
= first
+ ((last
- first
) >> 1);
998 cmp
= strncmp(name
, d
->name
, len
);
999 if (!cmp
&& strlen(d
->name
) > len
)
1011 FLEX_ALLOC_MEM(d
, name
, name
, len
);
1013 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
1014 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
1015 dir
->dirs_nr
- first
);
1017 dir
->dirs
[first
] = d
;
1021 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
1025 dir
->untracked_nr
= 0;
1026 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1027 do_invalidate_gitignore(dir
->dirs
[i
]);
1030 static void invalidate_gitignore(struct untracked_cache
*uc
,
1031 struct untracked_cache_dir
*dir
)
1033 uc
->gitignore_invalidated
++;
1034 do_invalidate_gitignore(dir
);
1037 static void invalidate_directory(struct untracked_cache
*uc
,
1038 struct untracked_cache_dir
*dir
)
1043 * Invalidation increment here is just roughly correct. If
1044 * untracked_nr or any of dirs[].recurse is non-zero, we
1045 * should increment dir_invalidated too. But that's more
1049 uc
->dir_invalidated
++;
1052 dir
->untracked_nr
= 0;
1053 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1054 dir
->dirs
[i
]->recurse
= 0;
1057 static int add_patterns_from_buffer(char *buf
, size_t size
,
1058 const char *base
, int baselen
,
1059 struct pattern_list
*pl
);
1061 /* Flags for add_patterns() */
1062 #define PATTERN_NOFOLLOW (1<<0)
1065 * Given a file with name "fname", read it (either from disk, or from
1066 * an index if 'istate' is non-null), parse it and store the
1067 * exclude rules in "pl".
1069 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1070 * stat data from disk (only valid if add_patterns returns zero). If
1071 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1073 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1074 struct pattern_list
*pl
, struct index_state
*istate
,
1075 unsigned flags
, struct oid_stat
*oid_stat
)
1083 if (flags
& PATTERN_NOFOLLOW
)
1084 fd
= open_nofollow(fname
, O_RDONLY
);
1086 fd
= open(fname
, O_RDONLY
);
1088 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1090 warn_on_fopen_errors(fname
);
1095 r
= read_skip_worktree_file_from_index(istate
, fname
,
1101 size
= xsize_t(st
.st_size
);
1104 fill_stat_data(&oid_stat
->stat
, &st
);
1105 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1106 oid_stat
->valid
= 1;
1111 buf
= xmallocz(size
);
1112 if (read_in_full(fd
, buf
, size
) != size
) {
1121 if (oid_stat
->valid
&&
1122 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1123 ; /* no content change, oid_stat->oid still good */
1125 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1126 !ce_stage(istate
->cache
[pos
]) &&
1127 ce_uptodate(istate
->cache
[pos
]) &&
1128 !would_convert_to_git(istate
, fname
))
1129 oidcpy(&oid_stat
->oid
,
1130 &istate
->cache
[pos
]->oid
);
1132 hash_object_file(the_hash_algo
, buf
, size
,
1133 OBJ_BLOB
, &oid_stat
->oid
);
1134 fill_stat_data(&oid_stat
->stat
, &st
);
1135 oid_stat
->valid
= 1;
1139 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1143 static int add_patterns_from_buffer(char *buf
, size_t size
,
1144 const char *base
, int baselen
,
1145 struct pattern_list
*pl
)
1150 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1151 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1155 if (skip_utf8_bom(&buf
, size
))
1156 size
-= buf
- pl
->filebuf
;
1160 for (i
= 0; i
< size
; i
++) {
1161 if (buf
[i
] == '\n') {
1162 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1163 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1164 trim_trailing_spaces(entry
);
1165 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1168 entry
= buf
+ i
+ 1;
1174 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1175 int baselen
, struct pattern_list
*pl
,
1176 struct index_state
*istate
,
1179 return add_patterns(fname
, base
, baselen
, pl
, istate
, flags
, NULL
);
1182 int add_patterns_from_blob_to_list(
1183 struct object_id
*oid
,
1184 const char *base
, int baselen
,
1185 struct pattern_list
*pl
)
1191 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1195 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1199 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1200 int group_type
, const char *src
)
1202 struct pattern_list
*pl
;
1203 struct exclude_list_group
*group
;
1205 group
= &dir
->internal
.exclude_list_group
[group_type
];
1206 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1207 pl
= &group
->pl
[group
->nr
++];
1208 memset(pl
, 0, sizeof(*pl
));
1214 * Used to set up core.excludesfile and .git/info/exclude lists.
1216 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1217 struct oid_stat
*oid_stat
)
1219 struct pattern_list
*pl
;
1221 * catch setup_standard_excludes() that's called before
1222 * dir->untracked is assigned. That function behaves
1223 * differently when dir->untracked is non-NULL.
1225 if (!dir
->untracked
)
1226 dir
->internal
.unmanaged_exclude_files
++;
1227 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1228 if (add_patterns(fname
, "", 0, pl
, NULL
, 0, oid_stat
) < 0)
1229 die(_("cannot use %s as an exclude file"), fname
);
1232 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1234 dir
->internal
.unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1235 add_patterns_from_file_1(dir
, fname
, NULL
);
1238 int match_basename(const char *basename
, int basenamelen
,
1239 const char *pattern
, int prefix
, int patternlen
,
1242 if (prefix
== patternlen
) {
1243 if (patternlen
== basenamelen
&&
1244 !fspathncmp(pattern
, basename
, basenamelen
))
1246 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1247 /* "*literal" matching against "fooliteral" */
1248 if (patternlen
- 1 <= basenamelen
&&
1249 !fspathncmp(pattern
+ 1,
1250 basename
+ basenamelen
- (patternlen
- 1),
1254 if (fnmatch_icase_mem(pattern
, patternlen
,
1255 basename
, basenamelen
,
1262 int match_pathname(const char *pathname
, int pathlen
,
1263 const char *base
, int baselen
,
1264 const char *pattern
, int prefix
, int patternlen
)
1270 * match with FNM_PATHNAME; the pattern has base implicitly
1273 if (*pattern
== '/') {
1280 * baselen does not count the trailing slash. base[] may or
1281 * may not end with a trailing slash though.
1283 if (pathlen
< baselen
+ 1 ||
1284 (baselen
&& pathname
[baselen
] != '/') ||
1285 fspathncmp(pathname
, base
, baselen
))
1288 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1289 name
= pathname
+ pathlen
- namelen
;
1293 * if the non-wildcard part is longer than the
1294 * remaining pathname, surely it cannot match.
1296 if (prefix
> namelen
)
1299 if (fspathncmp(pattern
, name
, prefix
))
1302 patternlen
-= prefix
;
1307 * If the whole pattern did not have a wildcard,
1308 * then our prefix match is all we need; we
1309 * do not need to call fnmatch at all.
1311 if (!patternlen
&& !namelen
)
1315 return fnmatch_icase_mem(pattern
, patternlen
,
1321 * Scan the given exclude list in reverse to see whether pathname
1322 * should be ignored. The first match (i.e. the last on the list), if
1323 * any, determines the fate. Returns the exclude_list element which
1324 * matched, or NULL for undecided.
1326 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1328 const char *basename
,
1330 struct pattern_list
*pl
,
1331 struct index_state
*istate
)
1333 struct path_pattern
*res
= NULL
; /* undecided */
1337 return NULL
; /* undefined */
1339 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1340 struct path_pattern
*pattern
= pl
->patterns
[i
];
1341 const char *exclude
= pattern
->pattern
;
1342 int prefix
= pattern
->nowildcardlen
;
1344 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1345 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1346 if (*dtype
!= DT_DIR
)
1350 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1351 if (match_basename(basename
,
1352 pathlen
- (basename
- pathname
),
1353 exclude
, prefix
, pattern
->patternlen
,
1361 assert(pattern
->baselen
== 0 ||
1362 pattern
->base
[pattern
->baselen
- 1] == '/');
1363 if (match_pathname(pathname
, pathlen
,
1365 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1366 exclude
, prefix
, pattern
->patternlen
)) {
1375 * Scan the list of patterns to determine if the ordered list
1376 * of patterns matches on 'pathname'.
1378 * Return 1 for a match, 0 for not matched and -1 for undecided.
1380 enum pattern_match_result
path_matches_pattern_list(
1381 const char *pathname
, int pathlen
,
1382 const char *basename
, int *dtype
,
1383 struct pattern_list
*pl
,
1384 struct index_state
*istate
)
1386 struct path_pattern
*pattern
;
1387 struct strbuf parent_pathname
= STRBUF_INIT
;
1388 int result
= NOT_MATCHED
;
1391 if (!pl
->use_cone_patterns
) {
1392 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1395 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1407 strbuf_addch(&parent_pathname
, '/');
1408 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1411 * Directory entries are matched if and only if a file
1412 * contained immediately within them is matched. For the
1413 * case of a directory entry, modify the path to create
1414 * a fake filename within this directory, allowing us to
1415 * use the file-base matching logic in an equivalent way.
1417 if (parent_pathname
.len
> 0 &&
1418 parent_pathname
.buf
[parent_pathname
.len
- 1] == '/') {
1419 slash_pos
= parent_pathname
.len
- 1;
1420 strbuf_add(&parent_pathname
, "-", 1);
1422 const char *slash_ptr
= strrchr(parent_pathname
.buf
, '/');
1423 slash_pos
= slash_ptr
? slash_ptr
- parent_pathname
.buf
: 0;
1426 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1427 &parent_pathname
)) {
1428 result
= MATCHED_RECURSIVE
;
1433 /* include every file in root */
1438 strbuf_setlen(&parent_pathname
, slash_pos
);
1440 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1445 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1448 result
= MATCHED_RECURSIVE
;
1451 strbuf_release(&parent_pathname
);
1455 int init_sparse_checkout_patterns(struct index_state
*istate
)
1457 if (!core_apply_sparse_checkout
)
1459 if (istate
->sparse_checkout_patterns
)
1462 CALLOC_ARRAY(istate
->sparse_checkout_patterns
, 1);
1464 if (get_sparse_checkout_patterns(istate
->sparse_checkout_patterns
) < 0) {
1465 FREE_AND_NULL(istate
->sparse_checkout_patterns
);
1472 static int path_in_sparse_checkout_1(const char *path
,
1473 struct index_state
*istate
,
1474 int require_cone_mode
)
1477 enum pattern_match_result match
= UNDECIDED
;
1478 const char *end
, *slash
;
1481 * We default to accepting a path if the path is empty, there are no
1482 * patterns, or the patterns are of the wrong type.
1485 init_sparse_checkout_patterns(istate
) ||
1486 (require_cone_mode
&&
1487 !istate
->sparse_checkout_patterns
->use_cone_patterns
))
1491 * If UNDECIDED, use the match from the parent dir (recursively), or
1492 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1493 * never returns UNDECIDED, so we will execute only one iteration in
1496 for (end
= path
+ strlen(path
);
1497 end
> path
&& match
== UNDECIDED
;
1500 for (slash
= end
- 1; slash
> path
&& *slash
!= '/'; slash
--)
1503 match
= path_matches_pattern_list(path
, end
- path
,
1504 slash
> path
? slash
+ 1 : path
, &dtype
,
1505 istate
->sparse_checkout_patterns
, istate
);
1507 /* We are going to match the parent dir now */
1513 int path_in_sparse_checkout(const char *path
,
1514 struct index_state
*istate
)
1516 return path_in_sparse_checkout_1(path
, istate
, 0);
1519 int path_in_cone_mode_sparse_checkout(const char *path
,
1520 struct index_state
*istate
)
1522 return path_in_sparse_checkout_1(path
, istate
, 1);
1525 static struct path_pattern
*last_matching_pattern_from_lists(
1526 struct dir_struct
*dir
, struct index_state
*istate
,
1527 const char *pathname
, int pathlen
,
1528 const char *basename
, int *dtype_p
)
1531 struct exclude_list_group
*group
;
1532 struct path_pattern
*pattern
;
1533 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1534 group
= &dir
->internal
.exclude_list_group
[i
];
1535 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1536 pattern
= last_matching_pattern_from_list(
1537 pathname
, pathlen
, basename
, dtype_p
,
1538 &group
->pl
[j
], istate
);
1547 * Loads the per-directory exclude list for the substring of base
1548 * which has a char length of baselen.
1550 static void prep_exclude(struct dir_struct
*dir
,
1551 struct index_state
*istate
,
1552 const char *base
, int baselen
)
1554 struct exclude_list_group
*group
;
1555 struct pattern_list
*pl
;
1556 struct exclude_stack
*stk
= NULL
;
1557 struct untracked_cache_dir
*untracked
;
1560 group
= &dir
->internal
.exclude_list_group
[EXC_DIRS
];
1563 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1564 * which originate from directories not in the prefix of the
1565 * path being checked.
1567 while ((stk
= dir
->internal
.exclude_stack
) != NULL
) {
1568 if (stk
->baselen
<= baselen
&&
1569 !strncmp(dir
->internal
.basebuf
.buf
, base
, stk
->baselen
))
1571 pl
= &group
->pl
[dir
->internal
.exclude_stack
->exclude_ix
];
1572 dir
->internal
.exclude_stack
= stk
->prev
;
1573 dir
->internal
.pattern
= NULL
;
1574 free((char *)pl
->src
); /* see strbuf_detach() below */
1575 clear_pattern_list(pl
);
1580 /* Skip traversing into sub directories if the parent is excluded */
1581 if (dir
->internal
.pattern
)
1585 * Lazy initialization. All call sites currently just
1586 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1587 * them seems lots of work for little benefit.
1589 if (!dir
->internal
.basebuf
.buf
)
1590 strbuf_init(&dir
->internal
.basebuf
, PATH_MAX
);
1592 /* Read from the parent directories and push them down. */
1593 current
= stk
? stk
->baselen
: -1;
1594 strbuf_setlen(&dir
->internal
.basebuf
, current
< 0 ? 0 : current
);
1596 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1600 while (current
< baselen
) {
1602 struct oid_stat oid_stat
;
1604 CALLOC_ARRAY(stk
, 1);
1609 cp
= strchr(base
+ current
+ 1, '/');
1611 die("oops in prep_exclude");
1614 lookup_untracked(dir
->untracked
,
1617 cp
- base
- current
);
1619 stk
->prev
= dir
->internal
.exclude_stack
;
1620 stk
->baselen
= cp
- base
;
1621 stk
->exclude_ix
= group
->nr
;
1622 stk
->ucd
= untracked
;
1623 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1624 strbuf_add(&dir
->internal
.basebuf
, base
+ current
, stk
->baselen
- current
);
1625 assert(stk
->baselen
== dir
->internal
.basebuf
.len
);
1627 /* Abort if the directory is excluded */
1630 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = 0;
1631 dir
->internal
.pattern
= last_matching_pattern_from_lists(dir
,
1633 dir
->internal
.basebuf
.buf
, stk
->baselen
- 1,
1634 dir
->internal
.basebuf
.buf
+ current
, &dt
);
1635 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = '/';
1636 if (dir
->internal
.pattern
&&
1637 dir
->internal
.pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1638 dir
->internal
.pattern
= NULL
;
1639 if (dir
->internal
.pattern
) {
1640 dir
->internal
.exclude_stack
= stk
;
1645 /* Try to read per-directory file */
1646 oidclr(&oid_stat
.oid
);
1648 if (dir
->exclude_per_dir
&&
1650 * If we know that no files have been added in
1651 * this directory (i.e. valid_cached_dir() has
1652 * been executed and set untracked->valid) ..
1654 (!untracked
|| !untracked
->valid
||
1656 * .. and .gitignore does not exist before
1657 * (i.e. null exclude_oid). Then we can skip
1658 * loading .gitignore, which would result in
1661 !is_null_oid(&untracked
->exclude_oid
))) {
1663 * dir->internal.basebuf gets reused by the traversal,
1664 * but we need fname to remain unchanged to ensure the
1665 * src member of each struct path_pattern correctly
1666 * back-references its source file. Other invocations
1667 * of add_pattern_list provide stable strings, so we
1668 * strbuf_detach() and free() here in the caller.
1670 struct strbuf sb
= STRBUF_INIT
;
1671 strbuf_addbuf(&sb
, &dir
->internal
.basebuf
);
1672 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1673 pl
->src
= strbuf_detach(&sb
, NULL
);
1674 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1676 untracked
? &oid_stat
: NULL
);
1679 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1680 * will first be called in valid_cached_dir() then maybe many
1681 * times more in last_matching_pattern(). When the cache is
1682 * used, last_matching_pattern() will not be called and
1683 * reading .gitignore content will be a waste.
1685 * So when it's called by valid_cached_dir() and we can get
1686 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1687 * modified on work tree), we could delay reading the
1688 * .gitignore content until we absolutely need it in
1689 * last_matching_pattern(). Be careful about ignore rule
1690 * order, though, if you do that.
1693 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1694 invalidate_gitignore(dir
->untracked
, untracked
);
1695 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1697 dir
->internal
.exclude_stack
= stk
;
1698 current
= stk
->baselen
;
1700 strbuf_setlen(&dir
->internal
.basebuf
, baselen
);
1704 * Loads the exclude lists for the directory containing pathname, then
1705 * scans all exclude lists to determine whether pathname is excluded.
1706 * Returns the exclude_list element which matched, or NULL for
1709 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1710 struct index_state
*istate
,
1711 const char *pathname
,
1714 int pathlen
= strlen(pathname
);
1715 const char *basename
= strrchr(pathname
, '/');
1716 basename
= (basename
) ? basename
+1 : pathname
;
1718 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1720 if (dir
->internal
.pattern
)
1721 return dir
->internal
.pattern
;
1723 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1728 * Loads the exclude lists for the directory containing pathname, then
1729 * scans all exclude lists to determine whether pathname is excluded.
1730 * Returns 1 if true, otherwise 0.
1732 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1733 const char *pathname
, int *dtype_p
)
1735 struct path_pattern
*pattern
=
1736 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1738 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1742 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1744 struct dir_entry
*ent
;
1746 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1751 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1752 struct index_state
*istate
,
1753 const char *pathname
, int len
)
1755 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1758 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->internal
.alloc
);
1759 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1762 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1763 struct index_state
*istate
,
1764 const char *pathname
, int len
)
1766 if (!index_name_is_other(istate
, pathname
, len
))
1769 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->internal
.ignored_alloc
);
1770 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1774 index_nonexistent
= 0,
1780 * Do not use the alphabetically sorted index to look up
1781 * the directory name; instead, use the case insensitive
1784 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1785 const char *dirname
, int len
)
1787 struct cache_entry
*ce
;
1789 if (index_dir_exists(istate
, dirname
, len
))
1790 return index_directory
;
1792 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1793 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1794 return index_gitdir
;
1796 return index_nonexistent
;
1800 * The index sorts alphabetically by entry name, which
1801 * means that a gitlink sorts as '\0' at the end, while
1802 * a directory (which is defined not as an entry, but as
1803 * the files it contains) will sort with the '/' at the
1806 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1807 const char *dirname
, int len
)
1812 return directory_exists_in_index_icase(istate
, dirname
, len
);
1814 pos
= index_name_pos(istate
, dirname
, len
);
1817 while (pos
< istate
->cache_nr
) {
1818 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1819 unsigned char endchar
;
1821 if (strncmp(ce
->name
, dirname
, len
))
1823 endchar
= ce
->name
[len
];
1827 return index_directory
;
1828 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1829 return index_gitdir
;
1831 return index_nonexistent
;
1835 * When we find a directory when traversing the filesystem, we
1836 * have three distinct cases:
1839 * - see it as a directory
1842 * and which one we choose depends on a combination of existing
1843 * git index contents and the flags passed into the directory
1844 * traversal routine.
1846 * Case 1: If we *already* have entries in the index under that
1847 * directory name, we always recurse into the directory to see
1850 * Case 2: If we *already* have that directory name as a gitlink,
1851 * we always continue to see it as a gitlink, regardless of whether
1852 * there is an actual git directory there or not (it might not
1853 * be checked out as a subproject!)
1855 * Case 3: if we didn't have it in the index previously, we
1856 * have a few sub-cases:
1858 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1859 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1860 * also true, in which case we need to check if it contains any
1861 * untracked and / or ignored files.
1862 * (b) if it looks like a git directory and we don't have the
1863 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1864 * show it as a directory.
1865 * (c) otherwise, we recurse into it.
1867 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1868 struct index_state
*istate
,
1869 struct untracked_cache_dir
*untracked
,
1870 const char *dirname
, int len
, int baselen
, int excluded
,
1871 const struct pathspec
*pathspec
)
1874 * WARNING: From this function, you can return path_recurse or you
1875 * can call read_directory_recursive() (or neither), but
1876 * you CAN'T DO BOTH.
1878 enum path_treatment state
;
1879 int matches_how
= 0;
1880 int check_only
, stop_early
;
1881 int old_ignored_nr
, old_untracked_nr
;
1882 /* The "len-1" is to strip the final '/' */
1883 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1885 if (status
== index_directory
)
1886 return path_recurse
;
1887 if (status
== index_gitdir
)
1889 if (status
!= index_nonexistent
)
1890 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1893 * We don't want to descend into paths that don't match the necessary
1894 * patterns. Clearly, if we don't have a pathspec, then we can't check
1895 * for matching patterns. Also, if (excluded) then we know we matched
1896 * the exclusion patterns so as an optimization we can skip checking
1897 * for matching patterns.
1899 if (pathspec
&& !excluded
) {
1900 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1904 DO_MATCH_LEADING_PATHSPEC
);
1910 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1911 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1913 * Determine if `dirname` is a nested repo by confirming that:
1914 * 1) we are in a nonbare repository, and
1915 * 2) `dirname` is not an immediate parent of `the_repository->gitdir`,
1916 * which could occur if the git_dir or worktree location was
1917 * manually configured by the user; see t2205 testcases 1-3 for
1918 * examples where this matters
1921 struct strbuf sb
= STRBUF_INIT
;
1922 strbuf_addstr(&sb
, dirname
);
1923 nested_repo
= is_nonbare_repository_dir(&sb
);
1926 char *real_dirname
, *real_gitdir
;
1927 strbuf_addstr(&sb
, ".git");
1928 real_dirname
= real_pathdup(sb
.buf
, 1);
1929 real_gitdir
= real_pathdup(the_repository
->gitdir
, 1);
1931 nested_repo
= !!strcmp(real_dirname
, real_gitdir
);
1935 strbuf_release(&sb
);
1938 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1939 (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
))
1941 return excluded
? path_excluded
: path_untracked
;
1945 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1947 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1948 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1951 * This is an excluded directory and we are
1952 * showing ignored paths that match an exclude
1953 * pattern. (e.g. show directory as ignored
1954 * only if it matches an exclude pattern).
1955 * This path will either be 'path_excluded`
1956 * (if we are showing empty directories or if
1957 * the directory is not empty), or will be
1958 * 'path_none' (empty directory, and we are
1959 * not showing empty directories).
1961 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1962 return path_excluded
;
1964 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1965 untracked
, 1, 1, pathspec
) == path_excluded
)
1966 return path_excluded
;
1970 return path_recurse
;
1973 assert(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
);
1976 * If we have a pathspec which could match something _below_ this
1977 * directory (e.g. when checking 'subdir/' having a pathspec like
1978 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1981 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1982 return path_recurse
;
1984 /* Special cases for where this directory is excluded/ignored */
1987 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1988 * hiding empty directories, there is no need to
1989 * recurse into an ignored directory.
1991 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1992 return path_excluded
;
1995 * Even if we are hiding empty directories, we can still avoid
1996 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
1997 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1999 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2000 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
2001 return path_excluded
;
2005 * Other than the path_recurse case above, we only need to
2006 * recurse into untracked directories if any of the following
2008 * - DIR_SHOW_IGNORED (because then we need to determine if
2009 * there are ignored entries below)
2010 * - DIR_SHOW_IGNORED_TOO (same as above)
2011 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
2012 * the directory is empty)
2015 !(dir
->flags
& (DIR_SHOW_IGNORED
|
2016 DIR_SHOW_IGNORED_TOO
|
2017 DIR_HIDE_EMPTY_DIRECTORIES
))) {
2018 return path_untracked
;
2022 * Even if we don't want to know all the paths under an untracked or
2023 * ignored directory, we may still need to go into the directory to
2024 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2025 * an empty directory should be path_none instead of path_excluded or
2028 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
2029 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
2032 * However, there's another optimization possible as a subset of
2033 * check_only, based on the cases we have to consider:
2034 * A) Directory matches no exclude patterns:
2035 * * Directory is empty => path_none
2036 * * Directory has an untracked file under it => path_untracked
2037 * * Directory has only ignored files under it => path_excluded
2038 * B) Directory matches an exclude pattern:
2039 * * Directory is empty => path_none
2040 * * Directory has an untracked file under it => path_excluded
2041 * * Directory has only ignored files under it => path_excluded
2042 * In case A, we can exit as soon as we've found an untracked
2043 * file but otherwise have to walk all files. In case B, though,
2044 * we can stop at the first file we find under the directory.
2046 stop_early
= check_only
&& excluded
;
2049 * If /every/ file within an untracked directory is ignored, then
2050 * we want to treat the directory as ignored (for e.g. status
2051 * --porcelain), without listing the individual ignored files
2052 * underneath. To do so, we'll save the current ignored_nr, and
2053 * pop all the ones added after it if it turns out the entire
2054 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2055 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2056 * untracked paths so will need to pop all those off the last
2057 * after we traverse.
2059 old_ignored_nr
= dir
->ignored_nr
;
2060 old_untracked_nr
= dir
->nr
;
2062 /* Actually recurse into dirname now, we'll fixup the state later. */
2063 untracked
= lookup_untracked(dir
->untracked
, untracked
,
2064 dirname
+ baselen
, len
- baselen
);
2065 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
2066 check_only
, stop_early
, pathspec
);
2068 /* There are a variety of reasons we may need to fixup the state... */
2069 if (state
== path_excluded
) {
2070 /* state == path_excluded implies all paths under
2071 * dirname were ignored...
2073 * if running e.g. `git status --porcelain --ignored=matching`,
2074 * then we want to see the subpaths that are ignored.
2076 * if running e.g. just `git status --porcelain`, then
2077 * we just want the directory itself to be listed as ignored
2078 * and not the individual paths underneath.
2080 int want_ignored_subpaths
=
2081 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2082 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
2084 if (want_ignored_subpaths
) {
2086 * with --ignored=matching, we want the subpaths
2087 * INSTEAD of the directory itself.
2092 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
2093 FREE_AND_NULL(dir
->ignored
[i
]);
2094 dir
->ignored_nr
= old_ignored_nr
;
2099 * We may need to ignore some of the untracked paths we found while
2100 * traversing subdirectories.
2102 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2103 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
2105 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
2106 FREE_AND_NULL(dir
->entries
[i
]);
2107 dir
->nr
= old_untracked_nr
;
2111 * If there is nothing under the current directory and we are not
2112 * hiding empty directories, then we need to report on the
2113 * untracked or ignored status of the directory itself.
2115 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2116 state
= excluded
? path_excluded
: path_untracked
;
2122 * This is an inexact early pruning of any recursive directory
2123 * reading - if the path cannot possibly be in the pathspec,
2124 * return true, and we'll skip it early.
2126 static int simplify_away(const char *path
, int pathlen
,
2127 const struct pathspec
*pathspec
)
2131 if (!pathspec
|| !pathspec
->nr
)
2134 GUARD_PATHSPEC(pathspec
,
2143 for (i
= 0; i
< pathspec
->nr
; i
++) {
2144 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2145 int len
= item
->nowildcard_len
;
2149 if (!ps_strncmp(item
, item
->match
, path
, len
))
2157 * This function tells us whether an excluded path matches a
2158 * list of "interesting" pathspecs. That is, whether a path matched
2159 * by any of the pathspecs could possibly be ignored by excluding
2160 * the specified path. This can happen if:
2162 * 1. the path is mentioned explicitly in the pathspec
2164 * 2. the path is a directory prefix of some element in the
2167 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2168 const struct pathspec
*pathspec
)
2172 if (!pathspec
|| !pathspec
->nr
)
2175 GUARD_PATHSPEC(pathspec
,
2183 for (i
= 0; i
< pathspec
->nr
; i
++) {
2184 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2185 int len
= item
->nowildcard_len
;
2187 if (len
== pathlen
&&
2188 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2190 if (len
> pathlen
&&
2191 item
->match
[pathlen
] == '/' &&
2192 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2198 static int get_index_dtype(struct index_state
*istate
,
2199 const char *path
, int len
)
2202 const struct cache_entry
*ce
;
2204 ce
= index_file_exists(istate
, path
, len
, 0);
2206 if (!ce_uptodate(ce
))
2208 if (S_ISGITLINK(ce
->ce_mode
))
2211 * Nobody actually cares about the
2212 * difference between DT_LNK and DT_REG
2217 /* Try to look it up as a directory */
2218 pos
= index_name_pos(istate
, path
, len
);
2222 while (pos
< istate
->cache_nr
) {
2223 ce
= istate
->cache
[pos
++];
2224 if (strncmp(ce
->name
, path
, len
))
2226 if (ce
->name
[len
] > '/')
2228 if (ce
->name
[len
] < '/')
2230 if (!ce_uptodate(ce
))
2231 break; /* continue? */
2237 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2238 const char *path
, int len
)
2242 if (dtype
!= DT_UNKNOWN
)
2244 dtype
= get_index_dtype(istate
, path
, len
);
2245 if (dtype
!= DT_UNKNOWN
)
2247 if (lstat(path
, &st
))
2249 if (S_ISREG(st
.st_mode
))
2251 if (S_ISDIR(st
.st_mode
))
2253 if (S_ISLNK(st
.st_mode
))
2258 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2259 struct cached_dir
*cdir
,
2260 struct index_state
*istate
,
2261 struct strbuf
*path
,
2263 const struct pathspec
*pathspec
)
2266 * WARNING: From this function, you can return path_recurse or you
2267 * can call read_directory_recursive() (or neither), but
2268 * you CAN'T DO BOTH.
2270 strbuf_setlen(path
, baselen
);
2272 strbuf_addstr(path
, cdir
->file
);
2273 return path_untracked
;
2275 strbuf_addstr(path
, cdir
->ucd
->name
);
2276 /* treat_one_path() does this before it calls treat_directory() */
2277 strbuf_complete(path
, '/');
2278 if (cdir
->ucd
->check_only
)
2280 * check_only is set as a result of treat_directory() getting
2281 * to its bottom. Verify again the same set of directories
2282 * with check_only set.
2284 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2285 cdir
->ucd
, 1, 0, pathspec
);
2287 * We get path_recurse in the first run when
2288 * directory_exists_in_index() returns index_nonexistent. We
2289 * are sure that new changes in the index does not impact the
2290 * outcome. Return now.
2292 return path_recurse
;
2295 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2296 struct untracked_cache_dir
*untracked
,
2297 struct cached_dir
*cdir
,
2298 struct index_state
*istate
,
2299 struct strbuf
*path
,
2301 const struct pathspec
*pathspec
)
2303 int has_path_in_index
, dtype
, excluded
;
2306 return treat_path_fast(dir
, cdir
, istate
, path
,
2308 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2310 strbuf_setlen(path
, baselen
);
2311 strbuf_addstr(path
, cdir
->d_name
);
2312 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2315 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2317 /* Always exclude indexed files */
2318 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2320 if (dtype
!= DT_DIR
&& has_path_in_index
)
2324 * When we are looking at a directory P in the working tree,
2325 * there are three cases:
2327 * (1) P exists in the index. Everything inside the directory P in
2328 * the working tree needs to go when P is checked out from the
2331 * (2) P does not exist in the index, but there is P/Q in the index.
2332 * We know P will stay a directory when we check out the contents
2333 * of the index, but we do not know yet if there is a directory
2334 * P/Q in the working tree to be killed, so we need to recurse.
2336 * (3) P does not exist in the index, and there is no P/Q in the index
2337 * to require P to be a directory, either. Only in this case, we
2338 * know that everything inside P will not be killed without
2341 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2342 (dtype
== DT_DIR
) &&
2343 !has_path_in_index
&&
2344 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2347 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2350 * Excluded? If we don't explicitly want to show
2351 * ignored files, ignore it
2353 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2354 return path_excluded
;
2361 * WARNING: Do not ignore/amend the return value from
2362 * treat_directory(), and especially do not change it to return
2363 * path_recurse as that can cause exponential slowdown.
2364 * Instead, modify treat_directory() to return the right value.
2366 strbuf_addch(path
, '/');
2367 return treat_directory(dir
, istate
, untracked
,
2368 path
->buf
, path
->len
,
2369 baselen
, excluded
, pathspec
);
2373 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2374 0 /* prefix */, NULL
/* seen */,
2378 return path_excluded
;
2379 return path_untracked
;
2383 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2387 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2388 dir
->untracked_alloc
);
2389 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2392 static int valid_cached_dir(struct dir_struct
*dir
,
2393 struct untracked_cache_dir
*untracked
,
2394 struct index_state
*istate
,
2395 struct strbuf
*path
,
2404 * With fsmonitor, we can trust the untracked cache's valid field.
2406 refresh_fsmonitor(istate
);
2407 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2408 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2409 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2412 if (!untracked
->valid
||
2413 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2414 fill_stat_data(&untracked
->stat_data
, &st
);
2419 if (untracked
->check_only
!= !!check_only
)
2423 * prep_exclude will be called eventually on this directory,
2424 * but it's called much later in last_matching_pattern(). We
2425 * need it now to determine the validity of the cache for this
2426 * path. The next calls will be nearly no-op, the way
2427 * prep_exclude() is designed.
2429 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2430 strbuf_addch(path
, '/');
2431 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2432 strbuf_setlen(path
, path
->len
- 1);
2434 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2436 /* hopefully prep_exclude() haven't invalidated this entry... */
2437 return untracked
->valid
;
2440 static int open_cached_dir(struct cached_dir
*cdir
,
2441 struct dir_struct
*dir
,
2442 struct untracked_cache_dir
*untracked
,
2443 struct index_state
*istate
,
2444 struct strbuf
*path
,
2449 memset(cdir
, 0, sizeof(*cdir
));
2450 cdir
->untracked
= untracked
;
2451 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2453 c_path
= path
->len
? path
->buf
: ".";
2454 cdir
->fdir
= opendir(c_path
);
2456 warning_errno(_("could not open directory '%s'"), c_path
);
2457 if (dir
->untracked
) {
2458 invalidate_directory(dir
->untracked
, untracked
);
2459 dir
->untracked
->dir_opened
++;
2466 static int read_cached_dir(struct cached_dir
*cdir
)
2471 de
= readdir_skip_dot_and_dotdot(cdir
->fdir
);
2473 cdir
->d_name
= NULL
;
2474 cdir
->d_type
= DT_UNKNOWN
;
2477 cdir
->d_name
= de
->d_name
;
2478 cdir
->d_type
= DTYPE(de
);
2481 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2482 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2492 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2493 struct untracked_cache_dir
*d
= cdir
->untracked
;
2494 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2500 static void close_cached_dir(struct cached_dir
*cdir
)
2503 closedir(cdir
->fdir
);
2505 * We have gone through this directory and found no untracked
2506 * entries. Mark it valid.
2508 if (cdir
->untracked
) {
2509 cdir
->untracked
->valid
= 1;
2510 cdir
->untracked
->recurse
= 1;
2514 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2515 struct untracked_cache_dir
*untracked
,
2516 struct cached_dir
*cdir
,
2517 struct index_state
*istate
,
2518 struct strbuf
*path
,
2520 const struct pathspec
*pathspec
,
2521 enum path_treatment state
)
2523 /* add the path to the appropriate result list */
2526 if (dir
->flags
& DIR_SHOW_IGNORED
)
2527 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2528 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2529 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2530 exclude_matches_pathspec(path
->buf
, path
->len
,
2532 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2535 case path_untracked
:
2536 if (dir
->flags
& DIR_SHOW_IGNORED
)
2538 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2540 add_untracked(untracked
, path
->buf
+ baselen
);
2549 * Read a directory tree. We currently ignore anything but
2550 * directories, regular files and symlinks. That's because git
2551 * doesn't handle them at all yet. Maybe that will change some
2554 * Also, we ignore the name ".git" (even if it is not a directory).
2555 * That likely will not change.
2557 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2558 * to signal that a file was found. This is the least significant value that
2559 * indicates that a file was encountered that does not depend on the order of
2560 * whether an untracked or excluded path was encountered first.
2562 * Returns the most significant path_treatment value encountered in the scan.
2563 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2564 * significant path_treatment value that will be returned.
2567 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2568 struct index_state
*istate
, const char *base
, int baselen
,
2569 struct untracked_cache_dir
*untracked
, int check_only
,
2570 int stop_at_first_file
, const struct pathspec
*pathspec
)
2573 * WARNING: Do NOT recurse unless path_recurse is returned from
2574 * treat_path(). Recursing on any other return value
2575 * can result in exponential slowdown.
2577 struct cached_dir cdir
;
2578 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2579 struct strbuf path
= STRBUF_INIT
;
2581 strbuf_add(&path
, base
, baselen
);
2583 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2585 dir
->internal
.visited_directories
++;
2588 untracked
->check_only
= !!check_only
;
2590 while (!read_cached_dir(&cdir
)) {
2591 /* check how the file or directory should be treated */
2592 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2594 dir
->internal
.visited_paths
++;
2596 if (state
> dir_state
)
2599 /* recurse into subdir if instructed by treat_path */
2600 if (state
== path_recurse
) {
2601 struct untracked_cache_dir
*ud
;
2602 ud
= lookup_untracked(dir
->untracked
,
2605 path
.len
- baselen
);
2607 read_directory_recursive(dir
, istate
, path
.buf
,
2609 check_only
, stop_at_first_file
, pathspec
);
2610 if (subdir_state
> dir_state
)
2611 dir_state
= subdir_state
;
2614 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2615 0 /* prefix */, NULL
,
2616 0 /* do NOT special case dirs */))
2621 if (stop_at_first_file
) {
2623 * If stopping at first file, then
2624 * signal that a file was found by
2625 * returning `path_excluded`. This is
2626 * to return a consistent value
2627 * regardless of whether an ignored or
2628 * excluded file happened to be
2631 * In current usage, the
2632 * `stop_at_first_file` is passed when
2633 * an ancestor directory has matched
2634 * an exclude pattern, so any found
2635 * files will be excluded.
2637 if (dir_state
>= path_excluded
) {
2638 dir_state
= path_excluded
;
2643 /* abort early if maximum state has been reached */
2644 if (dir_state
== path_untracked
) {
2646 add_untracked(untracked
, path
.buf
+ baselen
);
2649 /* skip the add_path_to_appropriate_result_list() */
2653 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2654 istate
, &path
, baselen
,
2657 close_cached_dir(&cdir
);
2659 strbuf_release(&path
);
2664 int cmp_dir_entry(const void *p1
, const void *p2
)
2666 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2667 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2669 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2672 /* check if *out lexically strictly contains *in */
2673 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2675 return (out
->len
< in
->len
) &&
2676 (out
->name
[out
->len
- 1] == '/') &&
2677 !memcmp(out
->name
, in
->name
, out
->len
);
2680 static int treat_leading_path(struct dir_struct
*dir
,
2681 struct index_state
*istate
,
2682 const char *path
, int len
,
2683 const struct pathspec
*pathspec
)
2685 struct strbuf sb
= STRBUF_INIT
;
2686 struct strbuf subdir
= STRBUF_INIT
;
2687 int prevlen
, baselen
;
2689 struct cached_dir cdir
;
2690 enum path_treatment state
= path_none
;
2693 * For each directory component of path, we are going to check whether
2694 * that path is relevant given the pathspec. For example, if path is
2696 * then we will ask treat_path() whether we should go into foo, then
2697 * whether we should go into bar, then whether baz is relevant.
2698 * Checking each is important because e.g. if path is
2700 * then we need to check .git to know we shouldn't traverse it.
2701 * If the return from treat_path() is:
2702 * * path_none, for any path, we return false.
2703 * * path_recurse, for all path components, we return true
2704 * * <anything else> for some intermediate component, we make sure
2705 * to add that path to the relevant list but return false
2706 * signifying that we shouldn't recurse into it.
2709 while (len
&& path
[len
- 1] == '/')
2714 memset(&cdir
, 0, sizeof(cdir
));
2715 cdir
.d_type
= DT_DIR
;
2719 prevlen
= baselen
+ !!baselen
;
2720 cp
= path
+ prevlen
;
2721 cp
= memchr(cp
, '/', path
+ len
- cp
);
2725 baselen
= cp
- path
;
2727 strbuf_add(&sb
, path
, baselen
);
2728 if (!is_directory(sb
.buf
))
2731 strbuf_add(&sb
, path
, prevlen
);
2732 strbuf_reset(&subdir
);
2733 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2734 cdir
.d_name
= subdir
.buf
;
2735 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2737 if (state
!= path_recurse
)
2738 break; /* do not recurse into it */
2740 break; /* finished checking */
2742 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2743 &sb
, baselen
, pathspec
,
2746 strbuf_release(&subdir
);
2747 strbuf_release(&sb
);
2748 return state
== path_recurse
;
2751 static const char *get_ident_string(void)
2753 static struct strbuf sb
= STRBUF_INIT
;
2758 if (uname(&uts
) < 0)
2759 die_errno(_("failed to get kernel name and information"));
2760 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2765 static int ident_in_untracked(const struct untracked_cache
*uc
)
2768 * Previous git versions may have saved many NUL separated
2769 * strings in the "ident" field, but it is insane to manage
2770 * many locations, so just take care of the first one.
2773 return !strcmp(uc
->ident
.buf
, get_ident_string());
2776 static void set_untracked_ident(struct untracked_cache
*uc
)
2778 strbuf_reset(&uc
->ident
);
2779 strbuf_addstr(&uc
->ident
, get_ident_string());
2782 * This strbuf used to contain a list of NUL separated
2783 * strings, so save NUL too for backward compatibility.
2785 strbuf_addch(&uc
->ident
, 0);
2788 static unsigned new_untracked_cache_flags(struct index_state
*istate
)
2790 struct repository
*repo
= istate
->repo
;
2794 * This logic is coordinated with the setting of these flags in
2795 * wt-status.c#wt_status_collect_untracked(), and the evaluation
2796 * of the config setting in commit.c#git_status_config()
2798 if (!repo_config_get_string(repo
, "status.showuntrackedfiles", &val
) &&
2799 !strcmp(val
, "all"))
2803 * The default, if "all" is not set, is "normal" - leading us here.
2804 * If the value is "none" then it really doesn't matter.
2806 return DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2809 static void new_untracked_cache(struct index_state
*istate
, int flags
)
2811 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2812 strbuf_init(&uc
->ident
, 100);
2813 uc
->exclude_per_dir
= ".gitignore";
2814 uc
->dir_flags
= flags
>= 0 ? flags
: new_untracked_cache_flags(istate
);
2815 set_untracked_ident(uc
);
2816 istate
->untracked
= uc
;
2817 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2820 void add_untracked_cache(struct index_state
*istate
)
2822 if (!istate
->untracked
) {
2823 new_untracked_cache(istate
, -1);
2825 if (!ident_in_untracked(istate
->untracked
)) {
2826 free_untracked_cache(istate
->untracked
);
2827 new_untracked_cache(istate
, -1);
2832 void remove_untracked_cache(struct index_state
*istate
)
2834 if (istate
->untracked
) {
2835 free_untracked_cache(istate
->untracked
);
2836 istate
->untracked
= NULL
;
2837 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2841 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2843 const struct pathspec
*pathspec
,
2844 struct index_state
*istate
)
2846 struct untracked_cache_dir
*root
;
2847 static int untracked_cache_disabled
= -1;
2849 if (!dir
->untracked
)
2851 if (untracked_cache_disabled
< 0)
2852 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2853 if (untracked_cache_disabled
)
2857 * We only support $GIT_DIR/info/exclude and core.excludesfile
2858 * as the global ignore rule files. Any other additions
2859 * (e.g. from command line) invalidate the cache. This
2860 * condition also catches running setup_standard_excludes()
2861 * before setting dir->untracked!
2863 if (dir
->internal
.unmanaged_exclude_files
)
2867 * Optimize for the main use case only: whole-tree git
2868 * status. More work involved in treat_leading_path() if we
2869 * use cache on just a subset of the worktree. pathspec
2870 * support could make the matter even worse.
2872 if (base_len
|| (pathspec
&& pathspec
->nr
))
2875 /* We don't support collecting ignore files */
2876 if (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2877 DIR_COLLECT_IGNORED
))
2881 * If we use .gitignore in the cache and now you change it to
2882 * .gitexclude, everything will go wrong.
2884 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2885 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2889 * EXC_CMDL is not considered in the cache. If people set it,
2892 if (dir
->internal
.exclude_list_group
[EXC_CMDL
].nr
)
2895 if (!ident_in_untracked(dir
->untracked
)) {
2896 warning(_("untracked cache is disabled on this system or location"));
2901 * If the untracked structure we received does not have the same flags
2902 * as requested in this run, we're going to need to either discard the
2903 * existing structure (and potentially later recreate), or bypass the
2904 * untracked cache mechanism for this run.
2906 if (dir
->flags
!= dir
->untracked
->dir_flags
) {
2908 * If the untracked structure we received does not have the same flags
2909 * as configured, then we need to reset / create a new "untracked"
2910 * structure to match the new config.
2912 * Keeping the saved and used untracked cache consistent with the
2913 * configuration provides an opportunity for frequent users of
2914 * "git status -uall" to leverage the untracked cache by aligning their
2915 * configuration - setting "status.showuntrackedfiles" to "all" or
2916 * "normal" as appropriate.
2918 * Previously using -uall (or setting "status.showuntrackedfiles" to
2919 * "all") was incompatible with untracked cache and *consistently*
2920 * caused surprisingly bad performance (with fscache and fsmonitor
2921 * enabled) on Windows.
2923 * IMPROVEMENT OPPORTUNITY: If we reworked the untracked cache storage
2924 * to not be as bound up with the desired output in a given run,
2925 * and instead iterated through and stored enough information to
2926 * correctly serve both "modes", then users could get peak performance
2927 * with or without '-uall' regardless of their
2928 * "status.showuntrackedfiles" config.
2930 if (dir
->untracked
->dir_flags
!= new_untracked_cache_flags(istate
)) {
2931 free_untracked_cache(istate
->untracked
);
2932 new_untracked_cache(istate
, dir
->flags
);
2933 dir
->untracked
= istate
->untracked
;
2937 * Current untracked cache data is consistent with config, but not
2938 * usable in this request/run; just bypass untracked cache.
2944 if (!dir
->untracked
->root
) {
2945 /* Untracked cache existed but is not initialized; fix that */
2946 FLEX_ALLOC_STR(dir
->untracked
->root
, name
, "");
2947 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2950 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2951 root
= dir
->untracked
->root
;
2952 if (!oideq(&dir
->internal
.ss_info_exclude
.oid
,
2953 &dir
->untracked
->ss_info_exclude
.oid
)) {
2954 invalidate_gitignore(dir
->untracked
, root
);
2955 dir
->untracked
->ss_info_exclude
= dir
->internal
.ss_info_exclude
;
2957 if (!oideq(&dir
->internal
.ss_excludes_file
.oid
,
2958 &dir
->untracked
->ss_excludes_file
.oid
)) {
2959 invalidate_gitignore(dir
->untracked
, root
);
2960 dir
->untracked
->ss_excludes_file
= dir
->internal
.ss_excludes_file
;
2963 /* Make sure this directory is not dropped out at saving phase */
2968 static void emit_traversal_statistics(struct dir_struct
*dir
,
2969 struct repository
*repo
,
2973 if (!trace2_is_enabled())
2977 trace2_data_string("read_directory", repo
, "path", "");
2979 struct strbuf tmp
= STRBUF_INIT
;
2980 strbuf_add(&tmp
, path
, path_len
);
2981 trace2_data_string("read_directory", repo
, "path", tmp
.buf
);
2982 strbuf_release(&tmp
);
2985 trace2_data_intmax("read_directory", repo
,
2986 "directories-visited", dir
->internal
.visited_directories
);
2987 trace2_data_intmax("read_directory", repo
,
2988 "paths-visited", dir
->internal
.visited_paths
);
2990 if (!dir
->untracked
)
2992 trace2_data_intmax("read_directory", repo
,
2993 "node-creation", dir
->untracked
->dir_created
);
2994 trace2_data_intmax("read_directory", repo
,
2995 "gitignore-invalidation",
2996 dir
->untracked
->gitignore_invalidated
);
2997 trace2_data_intmax("read_directory", repo
,
2998 "directory-invalidation",
2999 dir
->untracked
->dir_invalidated
);
3000 trace2_data_intmax("read_directory", repo
,
3001 "opendir", dir
->untracked
->dir_opened
);
3004 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
3005 const char *path
, int len
, const struct pathspec
*pathspec
)
3007 struct untracked_cache_dir
*untracked
;
3009 trace2_region_enter("dir", "read_directory", istate
->repo
);
3010 dir
->internal
.visited_paths
= 0;
3011 dir
->internal
.visited_directories
= 0;
3013 if (has_symlink_leading_path(path
, len
)) {
3014 trace2_region_leave("dir", "read_directory", istate
->repo
);
3018 untracked
= validate_untracked_cache(dir
, len
, pathspec
, istate
);
3021 * make sure untracked cache code path is disabled,
3022 * e.g. prep_exclude()
3024 dir
->untracked
= NULL
;
3025 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
3026 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
3027 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
3028 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
3030 emit_traversal_statistics(dir
, istate
->repo
, path
, len
);
3032 trace2_region_leave("dir", "read_directory", istate
->repo
);
3033 if (dir
->untracked
) {
3034 static int force_untracked_cache
= -1;
3036 if (force_untracked_cache
< 0)
3037 force_untracked_cache
=
3038 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", -1);
3039 if (force_untracked_cache
< 0)
3040 force_untracked_cache
= (istate
->repo
->settings
.core_untracked_cache
== UNTRACKED_CACHE_WRITE
);
3041 if (force_untracked_cache
&&
3042 dir
->untracked
== istate
->untracked
&&
3043 (dir
->untracked
->dir_opened
||
3044 dir
->untracked
->gitignore_invalidated
||
3045 dir
->untracked
->dir_invalidated
))
3046 istate
->cache_changed
|= UNTRACKED_CHANGED
;
3047 if (dir
->untracked
!= istate
->untracked
) {
3048 FREE_AND_NULL(dir
->untracked
);
3055 int file_exists(const char *f
)
3058 return lstat(f
, &sb
) == 0;
3061 int repo_file_exists(struct repository
*repo
, const char *path
)
3063 if (repo
!= the_repository
)
3064 BUG("do not know how to check file existence in arbitrary repo");
3066 return file_exists(path
);
3069 static int cmp_icase(char a
, char b
)
3074 return toupper(a
) - toupper(b
);
3079 * Given two normalized paths (a trailing slash is ok), if subdir is
3080 * outside dir, return -1. Otherwise return the offset in subdir that
3081 * can be used as relative path to dir.
3083 int dir_inside_of(const char *subdir
, const char *dir
)
3087 assert(dir
&& subdir
&& *dir
&& *subdir
);
3089 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
3095 /* hel[p]/me vs hel[l]/yeah */
3096 if (*dir
&& *subdir
)
3100 return !*dir
? offset
: -1; /* same dir */
3102 /* foo/[b]ar vs foo/[] */
3103 if (is_dir_sep(dir
[-1]))
3104 return is_dir_sep(subdir
[-1]) ? offset
: -1;
3106 /* foo[/]bar vs foo[] */
3107 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
3110 int is_inside_dir(const char *dir
)
3119 rc
= (dir_inside_of(cwd
, dir
) >= 0);
3124 int is_empty_dir(const char *path
)
3126 DIR *dir
= opendir(path
);
3133 e
= readdir_skip_dot_and_dotdot(dir
);
3141 char *git_url_basename(const char *repo
, int is_bundle
, int is_bare
)
3143 const char *end
= repo
+ strlen(repo
), *start
, *ptr
;
3150 start
= strstr(repo
, "://");
3157 * Skip authentication data. The stripping does happen
3158 * greedily, such that we strip up to the last '@' inside
3161 for (ptr
= start
; ptr
< end
&& !is_dir_sep(*ptr
); ptr
++) {
3167 * Strip trailing spaces, slashes and /.git
3169 while (start
< end
&& (is_dir_sep(end
[-1]) || isspace(end
[-1])))
3171 if (end
- start
> 5 && is_dir_sep(end
[-5]) &&
3172 !strncmp(end
- 4, ".git", 4)) {
3174 while (start
< end
&& is_dir_sep(end
[-1]))
3179 * It should not be possible to overflow `ptrdiff_t` by passing in an
3180 * insanely long URL, but GCC does not know that and will complain
3181 * without this check.
3183 if (end
- start
< 0)
3184 die(_("No directory name could be guessed.\n"
3185 "Please specify a directory on the command line"));
3188 * Strip trailing port number if we've got only a
3189 * hostname (that is, there is no dir separator but a
3190 * colon). This check is required such that we do not
3191 * strip URI's like '/foo/bar:2222.git', which should
3192 * result in a dir '2222' being guessed due to backwards
3195 if (memchr(start
, '/', end
- start
) == NULL
3196 && memchr(start
, ':', end
- start
) != NULL
) {
3198 while (start
< ptr
&& isdigit(ptr
[-1]) && ptr
[-1] != ':')
3200 if (start
< ptr
&& ptr
[-1] == ':')
3205 * Find last component. To remain backwards compatible we
3206 * also regard colons as path separators, such that
3207 * cloning a repository 'foo:bar.git' would result in a
3208 * directory 'bar' being guessed.
3211 while (start
< ptr
&& !is_dir_sep(ptr
[-1]) && ptr
[-1] != ':')
3216 * Strip .{bundle,git}.
3219 strip_suffix_mem(start
, &len
, is_bundle
? ".bundle" : ".git");
3221 if (!len
|| (len
== 1 && *start
== '/'))
3222 die(_("No directory name could be guessed.\n"
3223 "Please specify a directory on the command line"));
3226 dir
= xstrfmt("%.*s.git", (int)len
, start
);
3228 dir
= xstrndup(start
, len
);
3230 * Replace sequences of 'control' characters and whitespace
3231 * with one ascii space, remove leading and trailing spaces.
3235 int prev_space
= 1 /* strip leading whitespace */;
3236 for (end
= dir
; *end
; ++end
) {
3238 if ((unsigned char)ch
< '\x20')
3249 if (out
> dir
&& prev_space
)
3255 void strip_dir_trailing_slashes(char *dir
)
3257 char *end
= dir
+ strlen(dir
);
3259 while (dir
< end
- 1 && is_dir_sep(end
[-1]))
3264 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
3268 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
3269 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
3270 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
3271 int purge_original_cwd
= (flag
& REMOVE_DIR_PURGE_ORIGINAL_CWD
);
3272 struct object_id submodule_head
;
3274 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
3275 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
3276 /* Do not descend and nuke a nested git work tree. */
3282 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
3283 dir
= opendir(path
->buf
);
3285 if (errno
== ENOENT
)
3286 return keep_toplevel
? -1 : 0;
3287 else if (errno
== EACCES
&& !keep_toplevel
)
3289 * An empty dir could be removable even if it
3292 return rmdir(path
->buf
);
3296 strbuf_complete(path
, '/');
3299 while ((e
= readdir_skip_dot_and_dotdot(dir
)) != NULL
) {
3302 strbuf_setlen(path
, len
);
3303 strbuf_addstr(path
, e
->d_name
);
3304 if (lstat(path
->buf
, &st
)) {
3305 if (errno
== ENOENT
)
3307 * file disappeared, which is what we
3312 } else if (S_ISDIR(st
.st_mode
)) {
3313 if (!remove_dir_recurse(path
, flag
, &kept_down
))
3314 continue; /* happy */
3315 } else if (!only_empty
&&
3316 (!unlink(path
->buf
) || errno
== ENOENT
)) {
3317 continue; /* happy, too */
3320 /* path too long, stat fails, or non-directory still exists */
3326 strbuf_setlen(path
, original_len
);
3327 if (!ret
&& !keep_toplevel
&& !kept_down
) {
3328 if (!purge_original_cwd
&&
3329 startup_info
->original_cwd
&&
3330 !strcmp(startup_info
->original_cwd
, path
->buf
))
3331 ret
= -1; /* Do not remove current working directory */
3333 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
3336 * report the uplevel that it is not an error that we
3337 * did not rmdir() our directory.
3343 int remove_dir_recursively(struct strbuf
*path
, int flag
)
3345 return remove_dir_recurse(path
, flag
, NULL
);
3348 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
3350 void setup_standard_excludes(struct dir_struct
*dir
)
3352 dir
->exclude_per_dir
= ".gitignore";
3354 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3356 excludes_file
= xdg_config_home("ignore");
3357 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
3358 add_patterns_from_file_1(dir
, excludes_file
,
3359 dir
->untracked
? &dir
->internal
.ss_excludes_file
: NULL
);
3361 /* per repository user preference */
3362 if (startup_info
->have_repository
) {
3363 const char *path
= git_path_info_exclude();
3364 if (!access_or_warn(path
, R_OK
, 0))
3365 add_patterns_from_file_1(dir
, path
,
3366 dir
->untracked
? &dir
->internal
.ss_info_exclude
: NULL
);
3370 char *get_sparse_checkout_filename(void)
3372 return git_pathdup("info/sparse-checkout");
3375 int get_sparse_checkout_patterns(struct pattern_list
*pl
)
3378 char *sparse_filename
= get_sparse_checkout_filename();
3380 pl
->use_cone_patterns
= core_sparse_checkout_cone
;
3381 res
= add_patterns_from_file_to_list(sparse_filename
, "", 0, pl
, NULL
, 0);
3383 free(sparse_filename
);
3387 int remove_path(const char *name
)
3391 if (unlink(name
) && !is_missing_file_error(errno
))
3394 slash
= strrchr(name
, '/');
3396 char *dirs
= xstrdup(name
);
3397 slash
= dirs
+ (slash
- name
);
3400 if (startup_info
->original_cwd
&&
3401 !strcmp(startup_info
->original_cwd
, dirs
))
3403 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3410 * Frees memory within dir which was allocated, and resets fields for further
3411 * use. Does not free dir itself.
3413 void dir_clear(struct dir_struct
*dir
)
3416 struct exclude_list_group
*group
;
3417 struct pattern_list
*pl
;
3418 struct exclude_stack
*stk
;
3419 struct dir_struct
new = DIR_INIT
;
3421 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3422 group
= &dir
->internal
.exclude_list_group
[i
];
3423 for (j
= 0; j
< group
->nr
; j
++) {
3426 free((char *)pl
->src
);
3427 clear_pattern_list(pl
);
3432 for (i
= 0; i
< dir
->ignored_nr
; i
++)
3433 free(dir
->ignored
[i
]);
3434 for (i
= 0; i
< dir
->nr
; i
++)
3435 free(dir
->entries
[i
]);
3439 stk
= dir
->internal
.exclude_stack
;
3441 struct exclude_stack
*prev
= stk
->prev
;
3445 strbuf_release(&dir
->internal
.basebuf
);
3447 memcpy(dir
, &new, sizeof(*dir
));
3450 struct ondisk_untracked_cache
{
3451 struct stat_data info_exclude_stat
;
3452 struct stat_data excludes_file_stat
;
3456 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3459 int index
; /* number of written untracked_cache_dir */
3460 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3461 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3462 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3464 struct strbuf sb_stat
;
3465 struct strbuf sb_sha1
;
3468 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3470 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3471 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3472 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3473 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3474 to
->sd_dev
= htonl(from
->sd_dev
);
3475 to
->sd_ino
= htonl(from
->sd_ino
);
3476 to
->sd_uid
= htonl(from
->sd_uid
);
3477 to
->sd_gid
= htonl(from
->sd_gid
);
3478 to
->sd_size
= htonl(from
->sd_size
);
3481 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3482 struct write_data
*wd
)
3484 struct stat_data stat_data
;
3485 struct strbuf
*out
= &wd
->out
;
3486 unsigned char intbuf
[16];
3487 unsigned int intlen
, value
;
3488 int i
= wd
->index
++;
3491 * untracked_nr should be reset whenever valid is clear, but
3494 if (!untracked
->valid
) {
3495 untracked
->untracked_nr
= 0;
3496 untracked
->check_only
= 0;
3499 if (untracked
->check_only
)
3500 ewah_set(wd
->check_only
, i
);
3501 if (untracked
->valid
) {
3502 ewah_set(wd
->valid
, i
);
3503 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3504 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3506 if (!is_null_oid(&untracked
->exclude_oid
)) {
3507 ewah_set(wd
->sha1_valid
, i
);
3508 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3509 the_hash_algo
->rawsz
);
3512 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3513 strbuf_add(out
, intbuf
, intlen
);
3515 /* skip non-recurse directories */
3516 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3517 if (untracked
->dirs
[i
]->recurse
)
3519 intlen
= encode_varint(value
, intbuf
);
3520 strbuf_add(out
, intbuf
, intlen
);
3522 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3524 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3525 strbuf_add(out
, untracked
->untracked
[i
],
3526 strlen(untracked
->untracked
[i
]) + 1);
3528 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3529 if (untracked
->dirs
[i
]->recurse
)
3530 write_one_dir(untracked
->dirs
[i
], wd
);
3533 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3535 struct ondisk_untracked_cache
*ouc
;
3536 struct write_data wd
;
3537 unsigned char varbuf
[16];
3539 const unsigned hashsz
= the_hash_algo
->rawsz
;
3541 CALLOC_ARRAY(ouc
, 1);
3542 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3543 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3544 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3546 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3547 strbuf_add(out
, varbuf
, varint_len
);
3548 strbuf_addbuf(out
, &untracked
->ident
);
3550 strbuf_add(out
, ouc
, sizeof(*ouc
));
3551 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3552 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3553 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3556 if (!untracked
->root
) {
3557 varint_len
= encode_varint(0, varbuf
);
3558 strbuf_add(out
, varbuf
, varint_len
);
3563 wd
.check_only
= ewah_new();
3564 wd
.valid
= ewah_new();
3565 wd
.sha1_valid
= ewah_new();
3566 strbuf_init(&wd
.out
, 1024);
3567 strbuf_init(&wd
.sb_stat
, 1024);
3568 strbuf_init(&wd
.sb_sha1
, 1024);
3569 write_one_dir(untracked
->root
, &wd
);
3571 varint_len
= encode_varint(wd
.index
, varbuf
);
3572 strbuf_add(out
, varbuf
, varint_len
);
3573 strbuf_addbuf(out
, &wd
.out
);
3574 ewah_serialize_strbuf(wd
.valid
, out
);
3575 ewah_serialize_strbuf(wd
.check_only
, out
);
3576 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3577 strbuf_addbuf(out
, &wd
.sb_stat
);
3578 strbuf_addbuf(out
, &wd
.sb_sha1
);
3579 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3581 ewah_free(wd
.valid
);
3582 ewah_free(wd
.check_only
);
3583 ewah_free(wd
.sha1_valid
);
3584 strbuf_release(&wd
.out
);
3585 strbuf_release(&wd
.sb_stat
);
3586 strbuf_release(&wd
.sb_sha1
);
3589 static void free_untracked(struct untracked_cache_dir
*ucd
)
3594 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3595 free_untracked(ucd
->dirs
[i
]);
3596 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3597 free(ucd
->untracked
[i
]);
3598 free(ucd
->untracked
);
3603 void free_untracked_cache(struct untracked_cache
*uc
)
3608 free(uc
->exclude_per_dir_to_free
);
3609 strbuf_release(&uc
->ident
);
3610 free_untracked(uc
->root
);
3616 struct untracked_cache_dir
**ucd
;
3617 struct ewah_bitmap
*check_only
;
3618 struct ewah_bitmap
*valid
;
3619 struct ewah_bitmap
*sha1_valid
;
3620 const unsigned char *data
;
3621 const unsigned char *end
;
3624 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3626 memcpy(to
, data
, sizeof(*to
));
3627 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3628 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3629 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3630 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3631 to
->sd_dev
= ntohl(to
->sd_dev
);
3632 to
->sd_ino
= ntohl(to
->sd_ino
);
3633 to
->sd_uid
= ntohl(to
->sd_uid
);
3634 to
->sd_gid
= ntohl(to
->sd_gid
);
3635 to
->sd_size
= ntohl(to
->sd_size
);
3638 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3639 struct read_data
*rd
)
3641 struct untracked_cache_dir ud
, *untracked
;
3642 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3643 const unsigned char *eos
;
3647 memset(&ud
, 0, sizeof(ud
));
3649 value
= decode_varint(&data
);
3653 ud
.untracked_alloc
= value
;
3654 ud
.untracked_nr
= value
;
3655 if (ud
.untracked_nr
)
3656 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3658 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3661 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3663 eos
= memchr(data
, '\0', end
- data
);
3664 if (!eos
|| eos
== end
)
3667 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3668 memcpy(untracked
, &ud
, sizeof(ud
));
3669 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3672 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3673 eos
= memchr(data
, '\0', end
- data
);
3674 if (!eos
|| eos
== end
)
3676 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3680 rd
->ucd
[rd
->index
++] = untracked
;
3683 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3684 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3690 static void set_check_only(size_t pos
, void *cb
)
3692 struct read_data
*rd
= cb
;
3693 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3697 static void read_stat(size_t pos
, void *cb
)
3699 struct read_data
*rd
= cb
;
3700 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3701 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3702 rd
->data
= rd
->end
+ 1;
3705 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3706 rd
->data
+= sizeof(struct stat_data
);
3710 static void read_oid(size_t pos
, void *cb
)
3712 struct read_data
*rd
= cb
;
3713 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3714 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3715 rd
->data
= rd
->end
+ 1;
3718 oidread(&ud
->exclude_oid
, rd
->data
);
3719 rd
->data
+= the_hash_algo
->rawsz
;
3722 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3723 const unsigned char *sha1
)
3725 stat_data_from_disk(&oid_stat
->stat
, data
);
3726 oidread(&oid_stat
->oid
, sha1
);
3727 oid_stat
->valid
= 1;
3730 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3732 struct untracked_cache
*uc
;
3733 struct read_data rd
;
3734 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3738 const char *exclude_per_dir
;
3739 const unsigned hashsz
= the_hash_algo
->rawsz
;
3740 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3741 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3743 if (sz
<= 1 || end
[-1] != '\0')
3747 ident_len
= decode_varint(&next
);
3748 if (next
+ ident_len
> end
)
3750 ident
= (const char *)next
;
3753 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3756 CALLOC_ARRAY(uc
, 1);
3757 strbuf_init(&uc
->ident
, ident_len
);
3758 strbuf_add(&uc
->ident
, ident
, ident_len
);
3759 load_oid_stat(&uc
->ss_info_exclude
,
3760 next
+ ouc_offset(info_exclude_stat
),
3762 load_oid_stat(&uc
->ss_excludes_file
,
3763 next
+ ouc_offset(excludes_file_stat
),
3764 next
+ offset
+ hashsz
);
3765 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3766 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3767 uc
->exclude_per_dir
= uc
->exclude_per_dir_to_free
= xstrdup(exclude_per_dir
);
3768 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3769 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3773 len
= decode_varint(&next
);
3774 if (next
> end
|| len
== 0)
3777 rd
.valid
= ewah_new();
3778 rd
.check_only
= ewah_new();
3779 rd
.sha1_valid
= ewah_new();
3783 ALLOC_ARRAY(rd
.ucd
, len
);
3785 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3789 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3794 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3799 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3803 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3804 rd
.data
= next
+ len
;
3805 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3806 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3811 ewah_free(rd
.valid
);
3812 ewah_free(rd
.check_only
);
3813 ewah_free(rd
.sha1_valid
);
3816 free_untracked_cache(uc
);
3822 static void invalidate_one_directory(struct untracked_cache
*uc
,
3823 struct untracked_cache_dir
*ucd
)
3825 uc
->dir_invalidated
++;
3827 ucd
->untracked_nr
= 0;
3831 * Normally when an entry is added or removed from a directory,
3832 * invalidating that directory is enough. No need to touch its
3833 * ancestors. When a directory is shown as "foo/bar/" in git-status
3834 * however, deleting or adding an entry may have cascading effect.
3836 * Say the "foo/bar/file" has become untracked, we need to tell the
3837 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3838 * directory any more (because "bar" is managed by foo as an untracked
3841 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3842 * was the last untracked entry in the entire "foo", we should show
3843 * "foo/" instead. Which means we have to invalidate past "bar" up to
3846 * This function traverses all directories from root to leaf. If there
3847 * is a chance of one of the above cases happening, we invalidate back
3848 * to root. Otherwise we just invalidate the leaf. There may be a more
3849 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3850 * detect these cases and avoid unnecessary invalidation, for example,
3851 * checking for the untracked entry named "bar/" in "foo", but for now
3852 * stick to something safe and simple.
3854 static int invalidate_one_component(struct untracked_cache
*uc
,
3855 struct untracked_cache_dir
*dir
,
3856 const char *path
, int len
)
3858 const char *rest
= strchr(path
, '/');
3861 int component_len
= rest
- path
;
3862 struct untracked_cache_dir
*d
=
3863 lookup_untracked(uc
, dir
, path
, component_len
);
3865 invalidate_one_component(uc
, d
, rest
+ 1,
3866 len
- (component_len
+ 1));
3868 invalidate_one_directory(uc
, dir
);
3872 invalidate_one_directory(uc
, dir
);
3873 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3876 void untracked_cache_invalidate_path(struct index_state
*istate
,
3877 const char *path
, int safe_path
)
3879 if (!istate
->untracked
|| !istate
->untracked
->root
)
3881 if (!safe_path
&& !verify_path(path
, 0))
3883 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3884 path
, strlen(path
));
3887 void untracked_cache_remove_from_index(struct index_state
*istate
,
3890 untracked_cache_invalidate_path(istate
, path
, 1);
3893 void untracked_cache_add_to_index(struct index_state
*istate
,
3896 untracked_cache_invalidate_path(istate
, path
, 1);
3899 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3900 const char *sub_gitdir
)
3903 struct repository subrepo
;
3904 struct strbuf sub_wt
= STRBUF_INIT
;
3905 struct strbuf sub_gd
= STRBUF_INIT
;
3907 const struct submodule
*sub
;
3909 /* If the submodule has no working tree, we can ignore it. */
3910 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3913 if (repo_read_index(&subrepo
) < 0)
3914 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3916 /* TODO: audit for interaction with sparse-index. */
3917 ensure_full_index(subrepo
.index
);
3918 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3919 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3921 if (!S_ISGITLINK(ce
->ce_mode
))
3924 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3925 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3927 * Skip entries with the same name in different stages
3928 * to make sure an entry is returned only once.
3932 sub
= submodule_from_path(&subrepo
, null_oid(), ce
->name
);
3933 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3934 /* .gitmodules broken or inactive sub */
3937 strbuf_reset(&sub_wt
);
3938 strbuf_reset(&sub_gd
);
3939 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3940 submodule_name_to_gitdir(&sub_gd
, &subrepo
, sub
->name
);
3942 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3944 strbuf_release(&sub_wt
);
3945 strbuf_release(&sub_gd
);
3946 repo_clear(&subrepo
);
3949 void connect_work_tree_and_git_dir(const char *work_tree_
,
3950 const char *git_dir_
,
3951 int recurse_into_nested
)
3953 struct strbuf gitfile_sb
= STRBUF_INIT
;
3954 struct strbuf cfg_sb
= STRBUF_INIT
;
3955 struct strbuf rel_path
= STRBUF_INIT
;
3956 char *git_dir
, *work_tree
;
3958 /* Prepare .git file */
3959 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3960 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3961 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3963 /* Prepare config file */
3964 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3965 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3966 die(_("could not create directories for %s"), cfg_sb
.buf
);
3968 git_dir
= real_pathdup(git_dir_
, 1);
3969 work_tree
= real_pathdup(work_tree_
, 1);
3971 /* Write .git file */
3972 write_file(gitfile_sb
.buf
, "gitdir: %s",
3973 relative_path(git_dir
, work_tree
, &rel_path
));
3974 /* Update core.worktree setting */
3975 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3976 relative_path(work_tree
, git_dir
, &rel_path
));
3978 strbuf_release(&gitfile_sb
);
3979 strbuf_release(&cfg_sb
);
3980 strbuf_release(&rel_path
);
3982 if (recurse_into_nested
)
3983 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3990 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3992 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3994 if (rename(old_git_dir
, new_git_dir
) < 0)
3995 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3996 old_git_dir
, new_git_dir
);
3998 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);
4001 int path_match_flags(const char *const str
, const enum path_match_flags flags
)
4003 const char *p
= str
;
4005 if (flags
& PATH_MATCH_NATIVE
&&
4006 flags
& PATH_MATCH_XPLATFORM
)
4007 BUG("path_match_flags() must get one match kind, not multiple!");
4008 else if (!(flags
& PATH_MATCH_KINDS_MASK
))
4009 BUG("path_match_flags() must get at least one match kind!");
4011 if (flags
& PATH_MATCH_STARTS_WITH_DOT_SLASH
&&
4012 flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
)
4013 BUG("path_match_flags() must get one platform kind, not multiple!");
4014 else if (!(flags
& PATH_MATCH_PLATFORM_MASK
))
4015 BUG("path_match_flags() must get at least one platform kind!");
4019 if (flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
&&
4023 if (flags
& PATH_MATCH_NATIVE
)
4024 return is_dir_sep(*p
);
4025 else if (flags
& PATH_MATCH_XPLATFORM
)
4026 return is_xplatform_dir_sep(*p
);