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"
13 #include "environment.h"
15 #include "name-hash.h"
16 #include "object-file.h"
17 #include "object-store-ll.h"
21 #include "wildmatch.h"
25 #include "ewah/ewok.h"
26 #include "fsmonitor-ll.h"
27 #include "read-cache-ll.h"
29 #include "sparse-index.h"
30 #include "submodule-config.h"
36 * Tells read_directory_recursive how a file or directory should be treated.
37 * Values are ordered by significance, e.g. if a directory contains both
38 * excluded and untracked files, it is listed as untracked because
39 * path_untracked > path_excluded.
49 * Support data structure for our opendir/readdir/closedir wrappers
53 struct untracked_cache_dir
*untracked
;
60 struct untracked_cache_dir
*ucd
;
63 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
64 struct index_state
*istate
, const char *path
, int len
,
65 struct untracked_cache_dir
*untracked
,
66 int check_only
, int stop_at_first_file
, const struct pathspec
*pathspec
);
67 static int resolve_dtype(int dtype
, struct index_state
*istate
,
68 const char *path
, int len
);
69 struct dirent
*readdir_skip_dot_and_dotdot(DIR *dirp
)
73 while ((e
= readdir(dirp
)) != NULL
) {
74 if (!is_dot_or_dotdot(e
->d_name
))
80 int count_slashes(const char *s
)
89 int fspathcmp(const char *a
, const char *b
)
91 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
94 int fspatheq(const char *a
, const char *b
)
96 return !fspathcmp(a
, b
);
99 int fspathncmp(const char *a
, const char *b
, size_t count
)
101 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
104 int paths_collide(const char *a
, const char *b
)
106 size_t len_a
= strlen(a
), len_b
= strlen(b
);
109 return fspatheq(a
, b
);
112 return is_dir_sep(b
[len_a
]) && !fspathncmp(a
, b
, len_a
);
113 return is_dir_sep(a
[len_b
]) && !fspathncmp(a
, b
, len_b
);
116 unsigned int fspathhash(const char *str
)
118 return ignore_case
? strihash(str
) : strhash(str
);
121 int git_fnmatch(const struct pathspec_item
*item
,
122 const char *pattern
, const char *string
,
126 if (ps_strncmp(item
, pattern
, string
, prefix
))
131 if (item
->flags
& PATHSPEC_ONESTAR
) {
132 int pattern_len
= strlen(++pattern
);
133 int string_len
= strlen(string
);
134 return string_len
< pattern_len
||
135 ps_strcmp(item
, pattern
,
136 string
+ string_len
- pattern_len
);
138 if (item
->magic
& PATHSPEC_GLOB
)
139 return wildmatch(pattern
, string
,
141 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
143 /* wildmatch has not learned no FNM_PATHNAME mode yet */
144 return wildmatch(pattern
, string
,
145 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
148 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
149 const char *string
, int stringlen
,
153 struct strbuf pat_buf
= STRBUF_INIT
;
154 struct strbuf str_buf
= STRBUF_INIT
;
155 const char *use_pat
= pattern
;
156 const char *use_str
= string
;
158 if (pattern
[patternlen
]) {
159 strbuf_add(&pat_buf
, pattern
, patternlen
);
160 use_pat
= pat_buf
.buf
;
162 if (string
[stringlen
]) {
163 strbuf_add(&str_buf
, string
, stringlen
);
164 use_str
= str_buf
.buf
;
168 flags
|= WM_CASEFOLD
;
169 match_status
= wildmatch(use_pat
, use_str
, flags
);
171 strbuf_release(&pat_buf
);
172 strbuf_release(&str_buf
);
177 static size_t common_prefix_len(const struct pathspec
*pathspec
)
183 * ":(icase)path" is treated as a pathspec full of
184 * wildcard. In other words, only prefix is considered common
185 * prefix. If the pathspec is abc/foo abc/bar, running in
186 * subdir xyz, the common prefix is still xyz, not xyz/abc as
189 GUARD_PATHSPEC(pathspec
,
198 for (n
= 0; n
< pathspec
->nr
; n
++) {
199 size_t i
= 0, len
= 0, item_len
;
200 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
202 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
203 item_len
= pathspec
->items
[n
].prefix
;
205 item_len
= pathspec
->items
[n
].nowildcard_len
;
206 while (i
< item_len
&& (n
== 0 || i
< max
)) {
207 char c
= pathspec
->items
[n
].match
[i
];
208 if (c
!= pathspec
->items
[0].match
[i
])
214 if (n
== 0 || len
< max
) {
224 * Returns a copy of the longest leading path common among all
227 char *common_prefix(const struct pathspec
*pathspec
)
229 unsigned long len
= common_prefix_len(pathspec
);
231 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
234 int fill_directory(struct dir_struct
*dir
,
235 struct index_state
*istate
,
236 const struct pathspec
*pathspec
)
241 unsigned exclusive_flags
= DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
;
242 if ((dir
->flags
& exclusive_flags
) == exclusive_flags
)
243 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
246 * Calculate common prefix for the pathspec, and
247 * use that to optimize the directory walk
249 prefix_len
= common_prefix_len(pathspec
);
250 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
252 /* Read the directory and prune it */
253 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
258 int within_depth(const char *name
, int namelen
,
259 int depth
, int max_depth
)
261 const char *cp
= name
, *cpe
= name
+ namelen
;
267 if (depth
> max_depth
)
274 * Read the contents of the blob with the given OID into a buffer.
275 * Append a trailing LF to the end if the last line doesn't have one.
278 * -1 when the OID is invalid or unknown or does not refer to a blob.
279 * 0 when the blob is empty.
280 * 1 along with { data, size } of the (possibly augmented) buffer
283 * Optionally updates the given oid_stat with the given OID (when valid).
285 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
286 size_t *size_out
, char **data_out
)
288 enum object_type type
;
295 data
= repo_read_object_file(the_repository
, oid
, &type
, &sz
);
296 if (!data
|| type
!= OBJ_BLOB
) {
302 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
303 oidcpy(&oid_stat
->oid
, oid
);
311 if (data
[sz
- 1] != '\n') {
312 data
= xrealloc(data
, st_add(sz
, 1));
316 *size_out
= xsize_t(sz
);
322 #define DO_MATCH_EXCLUDE (1<<0)
323 #define DO_MATCH_DIRECTORY (1<<1)
324 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
327 * Does the given pathspec match the given name? A match is found if
329 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
330 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
331 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
332 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
334 * Return value tells which case it was (1-4), or 0 when there is no match.
336 * It may be instructive to look at a small table of concrete examples
337 * to understand the differences between 1, 2, and 4:
340 * | a/b | a/b/ | a/b/c
341 * ------+-----------+-----------+------------
342 * a/b | EXACT | EXACT[1] | LEADING[2]
343 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
344 * a/b/c | RECURSIVE | RECURSIVE | EXACT
346 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
347 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
349 static int match_pathspec_item(struct index_state
*istate
,
350 const struct pathspec_item
*item
, int prefix
,
351 const char *name
, int namelen
, unsigned flags
)
353 /* name/namelen has prefix cut off by caller */
354 const char *match
= item
->match
+ prefix
;
355 int matchlen
= item
->len
- prefix
;
358 * The normal call pattern is:
359 * 1. prefix = common_prefix_len(ps);
360 * 2. prune something, or fill_directory
361 * 3. match_pathspec()
363 * 'prefix' at #1 may be shorter than the command's prefix and
364 * it's ok for #2 to match extra files. Those extras will be
367 * Suppose the pathspec is 'foo' and '../bar' running from
368 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
369 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
370 * user does not want XYZ/foo, only the "foo" part should be
371 * case-insensitive. We need to filter out XYZ/foo here. In
372 * other words, we do not trust the caller on comparing the
373 * prefix part when :(icase) is involved. We do exact
374 * comparison ourselves.
376 * Normally the caller (common_prefix_len() in fact) does
377 * _exact_ matching on name[-prefix+1..-1] and we do not need
378 * to check that part. Be defensive and check it anyway, in
379 * case common_prefix_len is changed, or a new caller is
380 * introduced that does not use common_prefix_len.
382 * If the penalty turns out too high when prefix is really
383 * long, maybe change it to
384 * strncmp(match, name, item->prefix - prefix)
386 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
387 strncmp(item
->match
, name
- prefix
, item
->prefix
))
390 if (item
->attr_match_nr
&&
391 !match_pathspec_attrs(istate
, name
- prefix
, namelen
+ prefix
, item
))
394 /* If the match was just the prefix, we matched */
396 return MATCHED_RECURSIVELY
;
398 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
399 if (matchlen
== namelen
)
400 return MATCHED_EXACTLY
;
402 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
403 return MATCHED_RECURSIVELY
;
404 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
405 match
[matchlen
- 1] == '/' &&
406 namelen
== matchlen
- 1 &&
407 !ps_strncmp(item
, match
, name
, namelen
))
408 return MATCHED_EXACTLY
;
410 if (item
->nowildcard_len
< item
->len
&&
411 !git_fnmatch(item
, match
, name
,
412 item
->nowildcard_len
- prefix
))
413 return MATCHED_FNMATCH
;
415 /* Perform checks to see if "name" is a leading string of the pathspec */
416 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
417 !(flags
& DO_MATCH_EXCLUDE
)) {
418 /* name is a literal prefix of the pathspec */
419 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
420 if ((namelen
< matchlen
) &&
421 (match
[namelen
-offset
] == '/') &&
422 !ps_strncmp(item
, match
, name
, namelen
))
423 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
425 /* name doesn't match up to the first wild character */
426 if (item
->nowildcard_len
< item
->len
&&
427 ps_strncmp(item
, match
, name
,
428 item
->nowildcard_len
- prefix
))
432 * name has no wildcard, and it didn't match as a leading
433 * pathspec so return.
435 if (item
->nowildcard_len
== item
->len
)
439 * Here is where we would perform a wildmatch to check if
440 * "name" can be matched as a directory (or a prefix) against
441 * the pathspec. Since wildmatch doesn't have this capability
442 * at the present we have to punt and say that it is a match,
443 * potentially returning a false positive
444 * The submodules themselves will be able to perform more
445 * accurate matching to determine if the pathspec matches.
447 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
454 * do_match_pathspec() is meant to ONLY be called by
455 * match_pathspec_with_flags(); calling it directly risks pathspecs
456 * like ':!unwanted_path' being ignored.
458 * Given a name and a list of pathspecs, returns the nature of the
459 * closest (i.e. most specific) match of the name to any of the
462 * The caller typically calls this multiple times with the same
463 * pathspec and seen[] array but with different name/namelen
464 * (e.g. entries from the index) and is interested in seeing if and
465 * how each pathspec matches all the names it calls this function
466 * with. A mark is left in the seen[] array for each pathspec element
467 * indicating the closest type of match that element achieved, so if
468 * seen[n] remains zero after multiple invocations, that means the nth
469 * pathspec did not match any names, which could indicate that the
470 * user mistyped the nth pathspec.
472 static int do_match_pathspec(struct index_state
*istate
,
473 const struct pathspec
*ps
,
474 const char *name
, int namelen
,
475 int prefix
, char *seen
,
478 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
490 if (!ps
->recursive
||
491 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
493 return MATCHED_RECURSIVELY
;
495 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
496 return MATCHED_EXACTLY
;
504 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
507 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
508 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
511 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
514 * Make exclude patterns optional and never report
515 * "pathspec ':(exclude)foo' matches no files"
517 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
518 seen
[i
] = MATCHED_FNMATCH
;
519 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
522 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
523 ps
->max_depth
!= -1 &&
524 how
&& how
!= MATCHED_FNMATCH
) {
525 int len
= ps
->items
[i
].len
;
526 if (name
[len
] == '/')
528 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
529 how
= MATCHED_EXACTLY
;
536 if (seen
&& seen
[i
] < how
)
543 static int match_pathspec_with_flags(struct index_state
*istate
,
544 const struct pathspec
*ps
,
545 const char *name
, int namelen
,
546 int prefix
, char *seen
, unsigned flags
)
548 int positive
, negative
;
549 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
550 prefix
, seen
, flags
);
551 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
553 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
555 flags
| DO_MATCH_EXCLUDE
);
556 return negative
? 0 : positive
;
559 int match_pathspec(struct index_state
*istate
,
560 const struct pathspec
*ps
,
561 const char *name
, int namelen
,
562 int prefix
, char *seen
, int is_dir
)
564 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
565 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
566 prefix
, seen
, flags
);
570 * Check if a submodule is a superset of the pathspec
572 int submodule_path_match(struct index_state
*istate
,
573 const struct pathspec
*ps
,
574 const char *submodule_name
,
577 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
578 strlen(submodule_name
),
581 DO_MATCH_LEADING_PATHSPEC
);
585 int report_path_error(const char *ps_matched
,
586 const struct pathspec
*pathspec
)
589 * Make sure all pathspec matched; otherwise it is an error.
592 for (num
= 0; num
< pathspec
->nr
; num
++) {
593 int other
, found_dup
;
598 * The caller might have fed identical pathspec
599 * twice. Do not barf on such a mistake.
600 * FIXME: parse_pathspec should have eliminated
601 * duplicate pathspec.
603 for (found_dup
= other
= 0;
604 !found_dup
&& other
< pathspec
->nr
;
606 if (other
== num
|| !ps_matched
[other
])
608 if (!strcmp(pathspec
->items
[other
].original
,
609 pathspec
->items
[num
].original
))
611 * Ok, we have a match already.
618 error(_("pathspec '%s' did not match any file(s) known to git"),
619 pathspec
->items
[num
].original
);
626 * Return the length of the "simple" part of a path match limiter.
628 int simple_length(const char *match
)
633 unsigned char c
= *match
++;
635 if (c
== '\0' || is_glob_special(c
))
640 int no_wildcard(const char *string
)
642 return string
[simple_length(string
)] == '\0';
645 void parse_path_pattern(const char **pattern
,
650 const char *p
= *pattern
;
655 *flags
|= PATTERN_FLAG_NEGATIVE
;
659 if (len
&& p
[len
- 1] == '/') {
661 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
663 for (i
= 0; i
< len
; i
++) {
668 *flags
|= PATTERN_FLAG_NODIR
;
669 *nowildcardlen
= simple_length(p
);
671 * we should have excluded the trailing slash from 'p' too,
672 * but that's one more allocation. Instead just make sure
673 * nowildcardlen does not exceed real patternlen
675 if (*nowildcardlen
> len
)
676 *nowildcardlen
= len
;
677 if (*p
== '*' && no_wildcard(p
+ 1))
678 *flags
|= PATTERN_FLAG_ENDSWITH
;
683 int pl_hashmap_cmp(const void *cmp_data UNUSED
,
684 const struct hashmap_entry
*a
,
685 const struct hashmap_entry
*b
,
686 const void *key UNUSED
)
688 const struct pattern_entry
*ee1
=
689 container_of(a
, struct pattern_entry
, ent
);
690 const struct pattern_entry
*ee2
=
691 container_of(b
, struct pattern_entry
, ent
);
693 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
697 return fspathncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
700 static char *dup_and_filter_pattern(const char *pattern
)
704 char *result
= xstrdup(pattern
);
710 /* skip escape characters (once) */
730 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
732 struct pattern_entry
*translated
;
735 const char *prev
, *cur
, *next
;
737 if (!pl
->use_cone_patterns
)
740 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
741 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
742 !strcmp(given
->pattern
, "/*")) {
747 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
752 if (given
->patternlen
< 2 ||
753 *given
->pattern
!= '/' ||
754 strstr(given
->pattern
, "**")) {
755 /* Not a cone pattern. */
756 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
760 if (!(given
->flags
& PATTERN_FLAG_MUSTBEDIR
) &&
761 strcmp(given
->pattern
, "/*")) {
762 /* Not a cone pattern. */
763 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
767 prev
= given
->pattern
;
768 cur
= given
->pattern
+ 1;
769 next
= given
->pattern
+ 2;
772 /* Watch for glob characters '*', '\', '[', '?' */
773 if (!is_glob_special(*cur
))
776 /* But only if *prev != '\\' */
780 /* But allow the initial '\' */
782 is_glob_special(*next
))
785 /* But a trailing '/' then '*' is fine */
791 /* Not a cone pattern. */
792 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
801 if (given
->patternlen
> 2 &&
802 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
803 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
804 /* Not a cone pattern. */
805 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
809 truncated
= dup_and_filter_pattern(given
->pattern
);
811 translated
= xmalloc(sizeof(struct pattern_entry
));
812 translated
->pattern
= truncated
;
813 translated
->patternlen
= given
->patternlen
- 2;
814 hashmap_entry_init(&translated
->ent
,
815 fspathhash(translated
->pattern
));
817 if (!hashmap_get_entry(&pl
->recursive_hashmap
,
818 translated
, ent
, NULL
)) {
819 /* We did not see the "parent" included */
820 warning(_("unrecognized negative pattern: '%s'"),
827 hashmap_add(&pl
->parent_hashmap
, &translated
->ent
);
828 hashmap_remove(&pl
->recursive_hashmap
, &translated
->ent
, &data
);
833 if (given
->flags
& PATTERN_FLAG_NEGATIVE
) {
834 warning(_("unrecognized negative pattern: '%s'"),
839 translated
= xmalloc(sizeof(struct pattern_entry
));
841 translated
->pattern
= dup_and_filter_pattern(given
->pattern
);
842 translated
->patternlen
= given
->patternlen
;
843 hashmap_entry_init(&translated
->ent
,
844 fspathhash(translated
->pattern
));
846 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
848 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
849 /* we already included this at the parent level */
850 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
858 warning(_("disabling cone pattern matching"));
859 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
860 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
861 pl
->use_cone_patterns
= 0;
864 static int hashmap_contains_path(struct hashmap
*map
,
865 struct strbuf
*pattern
)
867 struct pattern_entry p
;
869 /* Check straight mapping */
870 p
.pattern
= pattern
->buf
;
871 p
.patternlen
= pattern
->len
;
872 hashmap_entry_init(&p
.ent
, fspathhash(p
.pattern
));
873 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
876 int hashmap_contains_parent(struct hashmap
*map
,
878 struct strbuf
*buffer
)
882 strbuf_setlen(buffer
, 0);
885 strbuf_addch(buffer
, '/');
887 strbuf_addstr(buffer
, path
);
889 slash_pos
= strrchr(buffer
->buf
, '/');
891 while (slash_pos
> buffer
->buf
) {
892 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
894 if (hashmap_contains_path(map
, buffer
))
897 slash_pos
= strrchr(buffer
->buf
, '/');
903 void add_pattern(const char *string
, const char *base
,
904 int baselen
, struct pattern_list
*pl
, int srcpos
)
906 struct path_pattern
*pattern
;
911 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
912 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
913 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
915 pattern
= xmalloc(sizeof(*pattern
));
916 pattern
->pattern
= string
;
918 pattern
->patternlen
= patternlen
;
919 pattern
->nowildcardlen
= nowildcardlen
;
920 pattern
->base
= base
;
921 pattern
->baselen
= baselen
;
922 pattern
->flags
= flags
;
923 pattern
->srcpos
= srcpos
;
924 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
925 pl
->patterns
[pl
->nr
++] = pattern
;
928 add_pattern_to_hashsets(pl
, pattern
);
931 static int read_skip_worktree_file_from_index(struct index_state
*istate
,
933 size_t *size_out
, char **data_out
,
934 struct oid_stat
*oid_stat
)
939 pos
= index_name_pos(istate
, path
, len
);
942 if (!ce_skip_worktree(istate
->cache
[pos
]))
945 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
949 * Frees memory within pl which was allocated for exclude patterns and
950 * the file buffer. Does not free pl itself.
952 void clear_pattern_list(struct pattern_list
*pl
)
956 for (i
= 0; i
< pl
->nr
; i
++)
957 free(pl
->patterns
[i
]);
960 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
961 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
963 memset(pl
, 0, sizeof(*pl
));
966 static void trim_trailing_spaces(char *buf
)
968 char *p
, *last_space
= NULL
;
970 for (p
= buf
; *p
; p
++)
990 * Given a subdirectory name and "dir" of the current directory,
991 * search the subdir in "dir" and return it, or create a new one if it
992 * does not exist in "dir".
994 * If "name" has the trailing slash, it'll be excluded in the search.
996 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
997 struct untracked_cache_dir
*dir
,
998 const char *name
, int len
)
1001 struct untracked_cache_dir
*d
;
1004 if (len
&& name
[len
- 1] == '/')
1007 last
= dir
->dirs_nr
;
1008 while (last
> first
) {
1009 int cmp
, next
= first
+ ((last
- first
) >> 1);
1010 d
= dir
->dirs
[next
];
1011 cmp
= strncmp(name
, d
->name
, len
);
1012 if (!cmp
&& strlen(d
->name
) > len
)
1024 FLEX_ALLOC_MEM(d
, name
, name
, len
);
1026 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
1027 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
1028 dir
->dirs_nr
- first
);
1030 dir
->dirs
[first
] = d
;
1034 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
1038 dir
->untracked_nr
= 0;
1039 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1040 do_invalidate_gitignore(dir
->dirs
[i
]);
1043 static void invalidate_gitignore(struct untracked_cache
*uc
,
1044 struct untracked_cache_dir
*dir
)
1046 uc
->gitignore_invalidated
++;
1047 do_invalidate_gitignore(dir
);
1050 static void invalidate_directory(struct untracked_cache
*uc
,
1051 struct untracked_cache_dir
*dir
)
1056 * Invalidation increment here is just roughly correct. If
1057 * untracked_nr or any of dirs[].recurse is non-zero, we
1058 * should increment dir_invalidated too. But that's more
1062 uc
->dir_invalidated
++;
1065 dir
->untracked_nr
= 0;
1066 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1067 dir
->dirs
[i
]->recurse
= 0;
1070 static int add_patterns_from_buffer(char *buf
, size_t size
,
1071 const char *base
, int baselen
,
1072 struct pattern_list
*pl
);
1074 /* Flags for add_patterns() */
1075 #define PATTERN_NOFOLLOW (1<<0)
1078 * Given a file with name "fname", read it (either from disk, or from
1079 * an index if 'istate' is non-null), parse it and store the
1080 * exclude rules in "pl".
1082 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1083 * stat data from disk (only valid if add_patterns returns zero). If
1084 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1086 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1087 struct pattern_list
*pl
, struct index_state
*istate
,
1088 unsigned flags
, struct oid_stat
*oid_stat
)
1096 if (flags
& PATTERN_NOFOLLOW
)
1097 fd
= open_nofollow(fname
, O_RDONLY
);
1099 fd
= open(fname
, O_RDONLY
);
1101 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1103 warn_on_fopen_errors(fname
);
1108 r
= read_skip_worktree_file_from_index(istate
, fname
,
1114 size
= xsize_t(st
.st_size
);
1117 fill_stat_data(&oid_stat
->stat
, &st
);
1118 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1119 oid_stat
->valid
= 1;
1124 buf
= xmallocz(size
);
1125 if (read_in_full(fd
, buf
, size
) != size
) {
1134 if (oid_stat
->valid
&&
1135 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1136 ; /* no content change, oid_stat->oid still good */
1138 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1139 !ce_stage(istate
->cache
[pos
]) &&
1140 ce_uptodate(istate
->cache
[pos
]) &&
1141 !would_convert_to_git(istate
, fname
))
1142 oidcpy(&oid_stat
->oid
,
1143 &istate
->cache
[pos
]->oid
);
1145 hash_object_file(the_hash_algo
, buf
, size
,
1146 OBJ_BLOB
, &oid_stat
->oid
);
1147 fill_stat_data(&oid_stat
->stat
, &st
);
1148 oid_stat
->valid
= 1;
1152 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1156 static int add_patterns_from_buffer(char *buf
, size_t size
,
1157 const char *base
, int baselen
,
1158 struct pattern_list
*pl
)
1163 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1164 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1168 if (skip_utf8_bom(&buf
, size
))
1169 size
-= buf
- pl
->filebuf
;
1173 for (i
= 0; i
< size
; i
++) {
1174 if (buf
[i
] == '\n') {
1175 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1176 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1177 trim_trailing_spaces(entry
);
1178 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1181 entry
= buf
+ i
+ 1;
1187 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1188 int baselen
, struct pattern_list
*pl
,
1189 struct index_state
*istate
,
1192 return add_patterns(fname
, base
, baselen
, pl
, istate
, flags
, NULL
);
1195 int add_patterns_from_blob_to_list(
1196 struct object_id
*oid
,
1197 const char *base
, int baselen
,
1198 struct pattern_list
*pl
)
1204 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1208 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1212 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1213 int group_type
, const char *src
)
1215 struct pattern_list
*pl
;
1216 struct exclude_list_group
*group
;
1218 group
= &dir
->internal
.exclude_list_group
[group_type
];
1219 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1220 pl
= &group
->pl
[group
->nr
++];
1221 memset(pl
, 0, sizeof(*pl
));
1227 * Used to set up core.excludesfile and .git/info/exclude lists.
1229 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1230 struct oid_stat
*oid_stat
)
1232 struct pattern_list
*pl
;
1234 * catch setup_standard_excludes() that's called before
1235 * dir->untracked is assigned. That function behaves
1236 * differently when dir->untracked is non-NULL.
1238 if (!dir
->untracked
)
1239 dir
->internal
.unmanaged_exclude_files
++;
1240 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1241 if (add_patterns(fname
, "", 0, pl
, NULL
, 0, oid_stat
) < 0)
1242 die(_("cannot use %s as an exclude file"), fname
);
1245 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1247 dir
->internal
.unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1248 add_patterns_from_file_1(dir
, fname
, NULL
);
1251 int match_basename(const char *basename
, int basenamelen
,
1252 const char *pattern
, int prefix
, int patternlen
,
1255 if (prefix
== patternlen
) {
1256 if (patternlen
== basenamelen
&&
1257 !fspathncmp(pattern
, basename
, basenamelen
))
1259 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1260 /* "*literal" matching against "fooliteral" */
1261 if (patternlen
- 1 <= basenamelen
&&
1262 !fspathncmp(pattern
+ 1,
1263 basename
+ basenamelen
- (patternlen
- 1),
1267 if (fnmatch_icase_mem(pattern
, patternlen
,
1268 basename
, basenamelen
,
1275 int match_pathname(const char *pathname
, int pathlen
,
1276 const char *base
, int baselen
,
1277 const char *pattern
, int prefix
, int patternlen
)
1283 * match with FNM_PATHNAME; the pattern has base implicitly
1286 if (*pattern
== '/') {
1293 * baselen does not count the trailing slash. base[] may or
1294 * may not end with a trailing slash though.
1296 if (pathlen
< baselen
+ 1 ||
1297 (baselen
&& pathname
[baselen
] != '/') ||
1298 fspathncmp(pathname
, base
, baselen
))
1301 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1302 name
= pathname
+ pathlen
- namelen
;
1306 * if the non-wildcard part is longer than the
1307 * remaining pathname, surely it cannot match.
1309 if (prefix
> namelen
)
1312 if (fspathncmp(pattern
, name
, prefix
))
1315 patternlen
-= prefix
;
1320 * If the whole pattern did not have a wildcard,
1321 * then our prefix match is all we need; we
1322 * do not need to call fnmatch at all.
1324 if (!patternlen
&& !namelen
)
1328 return fnmatch_icase_mem(pattern
, patternlen
,
1334 * Scan the given exclude list in reverse to see whether pathname
1335 * should be ignored. The first match (i.e. the last on the list), if
1336 * any, determines the fate. Returns the exclude_list element which
1337 * matched, or NULL for undecided.
1339 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1341 const char *basename
,
1343 struct pattern_list
*pl
,
1344 struct index_state
*istate
)
1346 struct path_pattern
*res
= NULL
; /* undecided */
1350 return NULL
; /* undefined */
1352 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1353 struct path_pattern
*pattern
= pl
->patterns
[i
];
1354 const char *exclude
= pattern
->pattern
;
1355 int prefix
= pattern
->nowildcardlen
;
1357 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1358 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1359 if (*dtype
!= DT_DIR
)
1363 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1364 if (match_basename(basename
,
1365 pathlen
- (basename
- pathname
),
1366 exclude
, prefix
, pattern
->patternlen
,
1374 assert(pattern
->baselen
== 0 ||
1375 pattern
->base
[pattern
->baselen
- 1] == '/');
1376 if (match_pathname(pathname
, pathlen
,
1378 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1379 exclude
, prefix
, pattern
->patternlen
)) {
1388 * Scan the list of patterns to determine if the ordered list
1389 * of patterns matches on 'pathname'.
1391 * Return 1 for a match, 0 for not matched and -1 for undecided.
1393 enum pattern_match_result
path_matches_pattern_list(
1394 const char *pathname
, int pathlen
,
1395 const char *basename
, int *dtype
,
1396 struct pattern_list
*pl
,
1397 struct index_state
*istate
)
1399 struct path_pattern
*pattern
;
1400 struct strbuf parent_pathname
= STRBUF_INIT
;
1401 int result
= NOT_MATCHED
;
1404 if (!pl
->use_cone_patterns
) {
1405 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1408 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1420 strbuf_addch(&parent_pathname
, '/');
1421 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1424 * Directory entries are matched if and only if a file
1425 * contained immediately within them is matched. For the
1426 * case of a directory entry, modify the path to create
1427 * a fake filename within this directory, allowing us to
1428 * use the file-base matching logic in an equivalent way.
1430 if (parent_pathname
.len
> 0 &&
1431 parent_pathname
.buf
[parent_pathname
.len
- 1] == '/') {
1432 slash_pos
= parent_pathname
.len
- 1;
1433 strbuf_add(&parent_pathname
, "-", 1);
1435 const char *slash_ptr
= strrchr(parent_pathname
.buf
, '/');
1436 slash_pos
= slash_ptr
? slash_ptr
- parent_pathname
.buf
: 0;
1439 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1440 &parent_pathname
)) {
1441 result
= MATCHED_RECURSIVE
;
1446 /* include every file in root */
1451 strbuf_setlen(&parent_pathname
, slash_pos
);
1453 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1458 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1461 result
= MATCHED_RECURSIVE
;
1464 strbuf_release(&parent_pathname
);
1468 int init_sparse_checkout_patterns(struct index_state
*istate
)
1470 if (!core_apply_sparse_checkout
)
1472 if (istate
->sparse_checkout_patterns
)
1475 CALLOC_ARRAY(istate
->sparse_checkout_patterns
, 1);
1477 if (get_sparse_checkout_patterns(istate
->sparse_checkout_patterns
) < 0) {
1478 FREE_AND_NULL(istate
->sparse_checkout_patterns
);
1485 static int path_in_sparse_checkout_1(const char *path
,
1486 struct index_state
*istate
,
1487 int require_cone_mode
)
1490 enum pattern_match_result match
= UNDECIDED
;
1491 const char *end
, *slash
;
1494 * We default to accepting a path if the path is empty, there are no
1495 * patterns, or the patterns are of the wrong type.
1498 init_sparse_checkout_patterns(istate
) ||
1499 (require_cone_mode
&&
1500 !istate
->sparse_checkout_patterns
->use_cone_patterns
))
1504 * If UNDECIDED, use the match from the parent dir (recursively), or
1505 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1506 * never returns UNDECIDED, so we will execute only one iteration in
1509 for (end
= path
+ strlen(path
);
1510 end
> path
&& match
== UNDECIDED
;
1513 for (slash
= end
- 1; slash
> path
&& *slash
!= '/'; slash
--)
1516 match
= path_matches_pattern_list(path
, end
- path
,
1517 slash
> path
? slash
+ 1 : path
, &dtype
,
1518 istate
->sparse_checkout_patterns
, istate
);
1520 /* We are going to match the parent dir now */
1526 int path_in_sparse_checkout(const char *path
,
1527 struct index_state
*istate
)
1529 return path_in_sparse_checkout_1(path
, istate
, 0);
1532 int path_in_cone_mode_sparse_checkout(const char *path
,
1533 struct index_state
*istate
)
1535 return path_in_sparse_checkout_1(path
, istate
, 1);
1538 static struct path_pattern
*last_matching_pattern_from_lists(
1539 struct dir_struct
*dir
, struct index_state
*istate
,
1540 const char *pathname
, int pathlen
,
1541 const char *basename
, int *dtype_p
)
1544 struct exclude_list_group
*group
;
1545 struct path_pattern
*pattern
;
1546 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1547 group
= &dir
->internal
.exclude_list_group
[i
];
1548 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1549 pattern
= last_matching_pattern_from_list(
1550 pathname
, pathlen
, basename
, dtype_p
,
1551 &group
->pl
[j
], istate
);
1560 * Loads the per-directory exclude list for the substring of base
1561 * which has a char length of baselen.
1563 static void prep_exclude(struct dir_struct
*dir
,
1564 struct index_state
*istate
,
1565 const char *base
, int baselen
)
1567 struct exclude_list_group
*group
;
1568 struct pattern_list
*pl
;
1569 struct exclude_stack
*stk
= NULL
;
1570 struct untracked_cache_dir
*untracked
;
1573 group
= &dir
->internal
.exclude_list_group
[EXC_DIRS
];
1576 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1577 * which originate from directories not in the prefix of the
1578 * path being checked.
1580 while ((stk
= dir
->internal
.exclude_stack
) != NULL
) {
1581 if (stk
->baselen
<= baselen
&&
1582 !strncmp(dir
->internal
.basebuf
.buf
, base
, stk
->baselen
))
1584 pl
= &group
->pl
[dir
->internal
.exclude_stack
->exclude_ix
];
1585 dir
->internal
.exclude_stack
= stk
->prev
;
1586 dir
->internal
.pattern
= NULL
;
1587 free((char *)pl
->src
); /* see strbuf_detach() below */
1588 clear_pattern_list(pl
);
1593 /* Skip traversing into sub directories if the parent is excluded */
1594 if (dir
->internal
.pattern
)
1598 * Lazy initialization. All call sites currently just
1599 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1600 * them seems lots of work for little benefit.
1602 if (!dir
->internal
.basebuf
.buf
)
1603 strbuf_init(&dir
->internal
.basebuf
, PATH_MAX
);
1605 /* Read from the parent directories and push them down. */
1606 current
= stk
? stk
->baselen
: -1;
1607 strbuf_setlen(&dir
->internal
.basebuf
, current
< 0 ? 0 : current
);
1609 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1613 while (current
< baselen
) {
1615 struct oid_stat oid_stat
;
1617 CALLOC_ARRAY(stk
, 1);
1622 cp
= strchr(base
+ current
+ 1, '/');
1624 die("oops in prep_exclude");
1627 lookup_untracked(dir
->untracked
,
1630 cp
- base
- current
);
1632 stk
->prev
= dir
->internal
.exclude_stack
;
1633 stk
->baselen
= cp
- base
;
1634 stk
->exclude_ix
= group
->nr
;
1635 stk
->ucd
= untracked
;
1636 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1637 strbuf_add(&dir
->internal
.basebuf
, base
+ current
, stk
->baselen
- current
);
1638 assert(stk
->baselen
== dir
->internal
.basebuf
.len
);
1640 /* Abort if the directory is excluded */
1643 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = 0;
1644 dir
->internal
.pattern
= last_matching_pattern_from_lists(dir
,
1646 dir
->internal
.basebuf
.buf
, stk
->baselen
- 1,
1647 dir
->internal
.basebuf
.buf
+ current
, &dt
);
1648 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = '/';
1649 if (dir
->internal
.pattern
&&
1650 dir
->internal
.pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1651 dir
->internal
.pattern
= NULL
;
1652 if (dir
->internal
.pattern
) {
1653 dir
->internal
.exclude_stack
= stk
;
1658 /* Try to read per-directory file */
1659 oidclr(&oid_stat
.oid
);
1661 if (dir
->exclude_per_dir
&&
1663 * If we know that no files have been added in
1664 * this directory (i.e. valid_cached_dir() has
1665 * been executed and set untracked->valid) ..
1667 (!untracked
|| !untracked
->valid
||
1669 * .. and .gitignore does not exist before
1670 * (i.e. null exclude_oid). Then we can skip
1671 * loading .gitignore, which would result in
1674 !is_null_oid(&untracked
->exclude_oid
))) {
1676 * dir->internal.basebuf gets reused by the traversal,
1677 * but we need fname to remain unchanged to ensure the
1678 * src member of each struct path_pattern correctly
1679 * back-references its source file. Other invocations
1680 * of add_pattern_list provide stable strings, so we
1681 * strbuf_detach() and free() here in the caller.
1683 struct strbuf sb
= STRBUF_INIT
;
1684 strbuf_addbuf(&sb
, &dir
->internal
.basebuf
);
1685 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1686 pl
->src
= strbuf_detach(&sb
, NULL
);
1687 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1689 untracked
? &oid_stat
: NULL
);
1692 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1693 * will first be called in valid_cached_dir() then maybe many
1694 * times more in last_matching_pattern(). When the cache is
1695 * used, last_matching_pattern() will not be called and
1696 * reading .gitignore content will be a waste.
1698 * So when it's called by valid_cached_dir() and we can get
1699 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1700 * modified on work tree), we could delay reading the
1701 * .gitignore content until we absolutely need it in
1702 * last_matching_pattern(). Be careful about ignore rule
1703 * order, though, if you do that.
1706 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1707 invalidate_gitignore(dir
->untracked
, untracked
);
1708 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1710 dir
->internal
.exclude_stack
= stk
;
1711 current
= stk
->baselen
;
1713 strbuf_setlen(&dir
->internal
.basebuf
, baselen
);
1717 * Loads the exclude lists for the directory containing pathname, then
1718 * scans all exclude lists to determine whether pathname is excluded.
1719 * Returns the exclude_list element which matched, or NULL for
1722 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1723 struct index_state
*istate
,
1724 const char *pathname
,
1727 int pathlen
= strlen(pathname
);
1728 const char *basename
= strrchr(pathname
, '/');
1729 basename
= (basename
) ? basename
+1 : pathname
;
1731 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1733 if (dir
->internal
.pattern
)
1734 return dir
->internal
.pattern
;
1736 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1741 * Loads the exclude lists for the directory containing pathname, then
1742 * scans all exclude lists to determine whether pathname is excluded.
1743 * Returns 1 if true, otherwise 0.
1745 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1746 const char *pathname
, int *dtype_p
)
1748 struct path_pattern
*pattern
=
1749 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1751 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1755 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1757 struct dir_entry
*ent
;
1759 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1764 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1765 struct index_state
*istate
,
1766 const char *pathname
, int len
)
1768 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1771 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->internal
.alloc
);
1772 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1775 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1776 struct index_state
*istate
,
1777 const char *pathname
, int len
)
1779 if (!index_name_is_other(istate
, pathname
, len
))
1782 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->internal
.ignored_alloc
);
1783 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1787 index_nonexistent
= 0,
1793 * Do not use the alphabetically sorted index to look up
1794 * the directory name; instead, use the case insensitive
1797 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1798 const char *dirname
, int len
)
1800 struct cache_entry
*ce
;
1802 if (index_dir_exists(istate
, dirname
, len
))
1803 return index_directory
;
1805 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1806 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1807 return index_gitdir
;
1809 return index_nonexistent
;
1813 * The index sorts alphabetically by entry name, which
1814 * means that a gitlink sorts as '\0' at the end, while
1815 * a directory (which is defined not as an entry, but as
1816 * the files it contains) will sort with the '/' at the
1819 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1820 const char *dirname
, int len
)
1825 return directory_exists_in_index_icase(istate
, dirname
, len
);
1827 pos
= index_name_pos(istate
, dirname
, len
);
1830 while (pos
< istate
->cache_nr
) {
1831 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1832 unsigned char endchar
;
1834 if (strncmp(ce
->name
, dirname
, len
))
1836 endchar
= ce
->name
[len
];
1840 return index_directory
;
1841 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1842 return index_gitdir
;
1844 return index_nonexistent
;
1848 * When we find a directory when traversing the filesystem, we
1849 * have three distinct cases:
1852 * - see it as a directory
1855 * and which one we choose depends on a combination of existing
1856 * git index contents and the flags passed into the directory
1857 * traversal routine.
1859 * Case 1: If we *already* have entries in the index under that
1860 * directory name, we always recurse into the directory to see
1863 * Case 2: If we *already* have that directory name as a gitlink,
1864 * we always continue to see it as a gitlink, regardless of whether
1865 * there is an actual git directory there or not (it might not
1866 * be checked out as a subproject!)
1868 * Case 3: if we didn't have it in the index previously, we
1869 * have a few sub-cases:
1871 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1872 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1873 * also true, in which case we need to check if it contains any
1874 * untracked and / or ignored files.
1875 * (b) if it looks like a git directory and we don't have the
1876 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1877 * show it as a directory.
1878 * (c) otherwise, we recurse into it.
1880 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1881 struct index_state
*istate
,
1882 struct untracked_cache_dir
*untracked
,
1883 const char *dirname
, int len
, int baselen
, int excluded
,
1884 const struct pathspec
*pathspec
)
1887 * WARNING: From this function, you can return path_recurse or you
1888 * can call read_directory_recursive() (or neither), but
1889 * you CAN'T DO BOTH.
1891 enum path_treatment state
;
1892 int matches_how
= 0;
1893 int check_only
, stop_early
;
1894 int old_ignored_nr
, old_untracked_nr
;
1895 /* The "len-1" is to strip the final '/' */
1896 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1898 if (status
== index_directory
)
1899 return path_recurse
;
1900 if (status
== index_gitdir
)
1902 if (status
!= index_nonexistent
)
1903 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1906 * We don't want to descend into paths that don't match the necessary
1907 * patterns. Clearly, if we don't have a pathspec, then we can't check
1908 * for matching patterns. Also, if (excluded) then we know we matched
1909 * the exclusion patterns so as an optimization we can skip checking
1910 * for matching patterns.
1912 if (pathspec
&& !excluded
) {
1913 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1917 DO_MATCH_LEADING_PATHSPEC
);
1923 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1924 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1926 * Determine if `dirname` is a nested repo by confirming that:
1927 * 1) we are in a nonbare repository, and
1928 * 2) `dirname` is not an immediate parent of `the_repository->gitdir`,
1929 * which could occur if the git_dir or worktree location was
1930 * manually configured by the user; see t2205 testcases 1-3 for
1931 * examples where this matters
1934 struct strbuf sb
= STRBUF_INIT
;
1935 strbuf_addstr(&sb
, dirname
);
1936 nested_repo
= is_nonbare_repository_dir(&sb
);
1939 char *real_dirname
, *real_gitdir
;
1940 strbuf_addstr(&sb
, ".git");
1941 real_dirname
= real_pathdup(sb
.buf
, 1);
1942 real_gitdir
= real_pathdup(the_repository
->gitdir
, 1);
1944 nested_repo
= !!strcmp(real_dirname
, real_gitdir
);
1948 strbuf_release(&sb
);
1951 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1952 (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
))
1954 return excluded
? path_excluded
: path_untracked
;
1958 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1960 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1961 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1964 * This is an excluded directory and we are
1965 * showing ignored paths that match an exclude
1966 * pattern. (e.g. show directory as ignored
1967 * only if it matches an exclude pattern).
1968 * This path will either be 'path_excluded`
1969 * (if we are showing empty directories or if
1970 * the directory is not empty), or will be
1971 * 'path_none' (empty directory, and we are
1972 * not showing empty directories).
1974 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1975 return path_excluded
;
1977 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1978 untracked
, 1, 1, pathspec
) == path_excluded
)
1979 return path_excluded
;
1983 return path_recurse
;
1986 assert(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
);
1989 * If we have a pathspec which could match something _below_ this
1990 * directory (e.g. when checking 'subdir/' having a pathspec like
1991 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1994 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1995 return path_recurse
;
1997 /* Special cases for where this directory is excluded/ignored */
2000 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
2001 * hiding empty directories, there is no need to
2002 * recurse into an ignored directory.
2004 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2005 return path_excluded
;
2008 * Even if we are hiding empty directories, we can still avoid
2009 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
2010 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
2012 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2013 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
2014 return path_excluded
;
2018 * Other than the path_recurse case above, we only need to
2019 * recurse into untracked directories if any of the following
2021 * - DIR_SHOW_IGNORED (because then we need to determine if
2022 * there are ignored entries below)
2023 * - DIR_SHOW_IGNORED_TOO (same as above)
2024 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
2025 * the directory is empty)
2028 !(dir
->flags
& (DIR_SHOW_IGNORED
|
2029 DIR_SHOW_IGNORED_TOO
|
2030 DIR_HIDE_EMPTY_DIRECTORIES
))) {
2031 return path_untracked
;
2035 * Even if we don't want to know all the paths under an untracked or
2036 * ignored directory, we may still need to go into the directory to
2037 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2038 * an empty directory should be path_none instead of path_excluded or
2041 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
2042 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
2045 * However, there's another optimization possible as a subset of
2046 * check_only, based on the cases we have to consider:
2047 * A) Directory matches no exclude patterns:
2048 * * Directory is empty => path_none
2049 * * Directory has an untracked file under it => path_untracked
2050 * * Directory has only ignored files under it => path_excluded
2051 * B) Directory matches an exclude pattern:
2052 * * Directory is empty => path_none
2053 * * Directory has an untracked file under it => path_excluded
2054 * * Directory has only ignored files under it => path_excluded
2055 * In case A, we can exit as soon as we've found an untracked
2056 * file but otherwise have to walk all files. In case B, though,
2057 * we can stop at the first file we find under the directory.
2059 stop_early
= check_only
&& excluded
;
2062 * If /every/ file within an untracked directory is ignored, then
2063 * we want to treat the directory as ignored (for e.g. status
2064 * --porcelain), without listing the individual ignored files
2065 * underneath. To do so, we'll save the current ignored_nr, and
2066 * pop all the ones added after it if it turns out the entire
2067 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2068 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2069 * untracked paths so will need to pop all those off the last
2070 * after we traverse.
2072 old_ignored_nr
= dir
->ignored_nr
;
2073 old_untracked_nr
= dir
->nr
;
2075 /* Actually recurse into dirname now, we'll fixup the state later. */
2076 untracked
= lookup_untracked(dir
->untracked
, untracked
,
2077 dirname
+ baselen
, len
- baselen
);
2078 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
2079 check_only
, stop_early
, pathspec
);
2081 /* There are a variety of reasons we may need to fixup the state... */
2082 if (state
== path_excluded
) {
2083 /* state == path_excluded implies all paths under
2084 * dirname were ignored...
2086 * if running e.g. `git status --porcelain --ignored=matching`,
2087 * then we want to see the subpaths that are ignored.
2089 * if running e.g. just `git status --porcelain`, then
2090 * we just want the directory itself to be listed as ignored
2091 * and not the individual paths underneath.
2093 int want_ignored_subpaths
=
2094 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2095 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
2097 if (want_ignored_subpaths
) {
2099 * with --ignored=matching, we want the subpaths
2100 * INSTEAD of the directory itself.
2105 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
2106 FREE_AND_NULL(dir
->ignored
[i
]);
2107 dir
->ignored_nr
= old_ignored_nr
;
2112 * We may need to ignore some of the untracked paths we found while
2113 * traversing subdirectories.
2115 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2116 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
2118 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
2119 FREE_AND_NULL(dir
->entries
[i
]);
2120 dir
->nr
= old_untracked_nr
;
2124 * If there is nothing under the current directory and we are not
2125 * hiding empty directories, then we need to report on the
2126 * untracked or ignored status of the directory itself.
2128 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2129 state
= excluded
? path_excluded
: path_untracked
;
2135 * This is an inexact early pruning of any recursive directory
2136 * reading - if the path cannot possibly be in the pathspec,
2137 * return true, and we'll skip it early.
2139 static int simplify_away(const char *path
, int pathlen
,
2140 const struct pathspec
*pathspec
)
2144 if (!pathspec
|| !pathspec
->nr
)
2147 GUARD_PATHSPEC(pathspec
,
2156 for (i
= 0; i
< pathspec
->nr
; i
++) {
2157 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2158 int len
= item
->nowildcard_len
;
2162 if (!ps_strncmp(item
, item
->match
, path
, len
))
2170 * This function tells us whether an excluded path matches a
2171 * list of "interesting" pathspecs. That is, whether a path matched
2172 * by any of the pathspecs could possibly be ignored by excluding
2173 * the specified path. This can happen if:
2175 * 1. the path is mentioned explicitly in the pathspec
2177 * 2. the path is a directory prefix of some element in the
2180 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2181 const struct pathspec
*pathspec
)
2185 if (!pathspec
|| !pathspec
->nr
)
2188 GUARD_PATHSPEC(pathspec
,
2196 for (i
= 0; i
< pathspec
->nr
; i
++) {
2197 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2198 int len
= item
->nowildcard_len
;
2200 if (len
== pathlen
&&
2201 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2203 if (len
> pathlen
&&
2204 item
->match
[pathlen
] == '/' &&
2205 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2211 static int get_index_dtype(struct index_state
*istate
,
2212 const char *path
, int len
)
2215 const struct cache_entry
*ce
;
2217 ce
= index_file_exists(istate
, path
, len
, 0);
2219 if (!ce_uptodate(ce
))
2221 if (S_ISGITLINK(ce
->ce_mode
))
2224 * Nobody actually cares about the
2225 * difference between DT_LNK and DT_REG
2230 /* Try to look it up as a directory */
2231 pos
= index_name_pos(istate
, path
, len
);
2235 while (pos
< istate
->cache_nr
) {
2236 ce
= istate
->cache
[pos
++];
2237 if (strncmp(ce
->name
, path
, len
))
2239 if (ce
->name
[len
] > '/')
2241 if (ce
->name
[len
] < '/')
2243 if (!ce_uptodate(ce
))
2244 break; /* continue? */
2250 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2251 const char *path
, int len
)
2255 if (dtype
!= DT_UNKNOWN
)
2257 dtype
= get_index_dtype(istate
, path
, len
);
2258 if (dtype
!= DT_UNKNOWN
)
2260 if (lstat(path
, &st
))
2262 if (S_ISREG(st
.st_mode
))
2264 if (S_ISDIR(st
.st_mode
))
2266 if (S_ISLNK(st
.st_mode
))
2271 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2272 struct cached_dir
*cdir
,
2273 struct index_state
*istate
,
2274 struct strbuf
*path
,
2276 const struct pathspec
*pathspec
)
2279 * WARNING: From this function, you can return path_recurse or you
2280 * can call read_directory_recursive() (or neither), but
2281 * you CAN'T DO BOTH.
2283 strbuf_setlen(path
, baselen
);
2285 strbuf_addstr(path
, cdir
->file
);
2286 return path_untracked
;
2288 strbuf_addstr(path
, cdir
->ucd
->name
);
2289 /* treat_one_path() does this before it calls treat_directory() */
2290 strbuf_complete(path
, '/');
2291 if (cdir
->ucd
->check_only
)
2293 * check_only is set as a result of treat_directory() getting
2294 * to its bottom. Verify again the same set of directories
2295 * with check_only set.
2297 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2298 cdir
->ucd
, 1, 0, pathspec
);
2300 * We get path_recurse in the first run when
2301 * directory_exists_in_index() returns index_nonexistent. We
2302 * are sure that new changes in the index does not impact the
2303 * outcome. Return now.
2305 return path_recurse
;
2308 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2309 struct untracked_cache_dir
*untracked
,
2310 struct cached_dir
*cdir
,
2311 struct index_state
*istate
,
2312 struct strbuf
*path
,
2314 const struct pathspec
*pathspec
)
2316 int has_path_in_index
, dtype
, excluded
;
2319 return treat_path_fast(dir
, cdir
, istate
, path
,
2321 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2323 strbuf_setlen(path
, baselen
);
2324 strbuf_addstr(path
, cdir
->d_name
);
2325 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2328 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2330 /* Always exclude indexed files */
2331 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2333 if (dtype
!= DT_DIR
&& has_path_in_index
)
2337 * When we are looking at a directory P in the working tree,
2338 * there are three cases:
2340 * (1) P exists in the index. Everything inside the directory P in
2341 * the working tree needs to go when P is checked out from the
2344 * (2) P does not exist in the index, but there is P/Q in the index.
2345 * We know P will stay a directory when we check out the contents
2346 * of the index, but we do not know yet if there is a directory
2347 * P/Q in the working tree to be killed, so we need to recurse.
2349 * (3) P does not exist in the index, and there is no P/Q in the index
2350 * to require P to be a directory, either. Only in this case, we
2351 * know that everything inside P will not be killed without
2354 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2355 (dtype
== DT_DIR
) &&
2356 !has_path_in_index
&&
2357 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2360 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2363 * Excluded? If we don't explicitly want to show
2364 * ignored files, ignore it
2366 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2367 return path_excluded
;
2374 * WARNING: Do not ignore/amend the return value from
2375 * treat_directory(), and especially do not change it to return
2376 * path_recurse as that can cause exponential slowdown.
2377 * Instead, modify treat_directory() to return the right value.
2379 strbuf_addch(path
, '/');
2380 return treat_directory(dir
, istate
, untracked
,
2381 path
->buf
, path
->len
,
2382 baselen
, excluded
, pathspec
);
2386 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2387 0 /* prefix */, NULL
/* seen */,
2391 return path_excluded
;
2392 return path_untracked
;
2396 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2400 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2401 dir
->untracked_alloc
);
2402 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2405 static int valid_cached_dir(struct dir_struct
*dir
,
2406 struct untracked_cache_dir
*untracked
,
2407 struct index_state
*istate
,
2408 struct strbuf
*path
,
2417 * With fsmonitor, we can trust the untracked cache's valid field.
2419 refresh_fsmonitor(istate
);
2420 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2421 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2422 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2425 if (!untracked
->valid
||
2426 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2427 fill_stat_data(&untracked
->stat_data
, &st
);
2432 if (untracked
->check_only
!= !!check_only
)
2436 * prep_exclude will be called eventually on this directory,
2437 * but it's called much later in last_matching_pattern(). We
2438 * need it now to determine the validity of the cache for this
2439 * path. The next calls will be nearly no-op, the way
2440 * prep_exclude() is designed.
2442 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2443 strbuf_addch(path
, '/');
2444 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2445 strbuf_setlen(path
, path
->len
- 1);
2447 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2449 /* hopefully prep_exclude() haven't invalidated this entry... */
2450 return untracked
->valid
;
2453 static int open_cached_dir(struct cached_dir
*cdir
,
2454 struct dir_struct
*dir
,
2455 struct untracked_cache_dir
*untracked
,
2456 struct index_state
*istate
,
2457 struct strbuf
*path
,
2462 memset(cdir
, 0, sizeof(*cdir
));
2463 cdir
->untracked
= untracked
;
2464 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2466 c_path
= path
->len
? path
->buf
: ".";
2467 cdir
->fdir
= opendir(c_path
);
2469 warning_errno(_("could not open directory '%s'"), c_path
);
2470 if (dir
->untracked
) {
2471 invalidate_directory(dir
->untracked
, untracked
);
2472 dir
->untracked
->dir_opened
++;
2479 static int read_cached_dir(struct cached_dir
*cdir
)
2484 de
= readdir_skip_dot_and_dotdot(cdir
->fdir
);
2486 cdir
->d_name
= NULL
;
2487 cdir
->d_type
= DT_UNKNOWN
;
2490 cdir
->d_name
= de
->d_name
;
2491 cdir
->d_type
= DTYPE(de
);
2494 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2495 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2505 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2506 struct untracked_cache_dir
*d
= cdir
->untracked
;
2507 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2513 static void close_cached_dir(struct cached_dir
*cdir
)
2516 closedir(cdir
->fdir
);
2518 * We have gone through this directory and found no untracked
2519 * entries. Mark it valid.
2521 if (cdir
->untracked
) {
2522 cdir
->untracked
->valid
= 1;
2523 cdir
->untracked
->recurse
= 1;
2527 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2528 struct untracked_cache_dir
*untracked
,
2529 struct cached_dir
*cdir
,
2530 struct index_state
*istate
,
2531 struct strbuf
*path
,
2533 const struct pathspec
*pathspec
,
2534 enum path_treatment state
)
2536 /* add the path to the appropriate result list */
2539 if (dir
->flags
& DIR_SHOW_IGNORED
)
2540 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2541 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2542 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2543 exclude_matches_pathspec(path
->buf
, path
->len
,
2545 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2548 case path_untracked
:
2549 if (dir
->flags
& DIR_SHOW_IGNORED
)
2551 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2553 add_untracked(untracked
, path
->buf
+ baselen
);
2562 * Read a directory tree. We currently ignore anything but
2563 * directories, regular files and symlinks. That's because git
2564 * doesn't handle them at all yet. Maybe that will change some
2567 * Also, we ignore the name ".git" (even if it is not a directory).
2568 * That likely will not change.
2570 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2571 * to signal that a file was found. This is the least significant value that
2572 * indicates that a file was encountered that does not depend on the order of
2573 * whether an untracked or excluded path was encountered first.
2575 * Returns the most significant path_treatment value encountered in the scan.
2576 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2577 * significant path_treatment value that will be returned.
2580 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2581 struct index_state
*istate
, const char *base
, int baselen
,
2582 struct untracked_cache_dir
*untracked
, int check_only
,
2583 int stop_at_first_file
, const struct pathspec
*pathspec
)
2586 * WARNING: Do NOT recurse unless path_recurse is returned from
2587 * treat_path(). Recursing on any other return value
2588 * can result in exponential slowdown.
2590 struct cached_dir cdir
;
2591 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2592 struct strbuf path
= STRBUF_INIT
;
2594 strbuf_add(&path
, base
, baselen
);
2596 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2598 dir
->internal
.visited_directories
++;
2601 untracked
->check_only
= !!check_only
;
2603 while (!read_cached_dir(&cdir
)) {
2604 /* check how the file or directory should be treated */
2605 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2607 dir
->internal
.visited_paths
++;
2609 if (state
> dir_state
)
2612 /* recurse into subdir if instructed by treat_path */
2613 if (state
== path_recurse
) {
2614 struct untracked_cache_dir
*ud
;
2615 ud
= lookup_untracked(dir
->untracked
,
2618 path
.len
- baselen
);
2620 read_directory_recursive(dir
, istate
, path
.buf
,
2622 check_only
, stop_at_first_file
, pathspec
);
2623 if (subdir_state
> dir_state
)
2624 dir_state
= subdir_state
;
2627 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2628 0 /* prefix */, NULL
,
2629 0 /* do NOT special case dirs */))
2634 if (stop_at_first_file
) {
2636 * If stopping at first file, then
2637 * signal that a file was found by
2638 * returning `path_excluded`. This is
2639 * to return a consistent value
2640 * regardless of whether an ignored or
2641 * excluded file happened to be
2644 * In current usage, the
2645 * `stop_at_first_file` is passed when
2646 * an ancestor directory has matched
2647 * an exclude pattern, so any found
2648 * files will be excluded.
2650 if (dir_state
>= path_excluded
) {
2651 dir_state
= path_excluded
;
2656 /* abort early if maximum state has been reached */
2657 if (dir_state
== path_untracked
) {
2659 add_untracked(untracked
, path
.buf
+ baselen
);
2662 /* skip the add_path_to_appropriate_result_list() */
2666 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2667 istate
, &path
, baselen
,
2670 close_cached_dir(&cdir
);
2672 strbuf_release(&path
);
2677 int cmp_dir_entry(const void *p1
, const void *p2
)
2679 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2680 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2682 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2685 /* check if *out lexically strictly contains *in */
2686 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2688 return (out
->len
< in
->len
) &&
2689 (out
->name
[out
->len
- 1] == '/') &&
2690 !memcmp(out
->name
, in
->name
, out
->len
);
2693 static int treat_leading_path(struct dir_struct
*dir
,
2694 struct index_state
*istate
,
2695 const char *path
, int len
,
2696 const struct pathspec
*pathspec
)
2698 struct strbuf sb
= STRBUF_INIT
;
2699 struct strbuf subdir
= STRBUF_INIT
;
2700 int prevlen
, baselen
;
2702 struct cached_dir cdir
;
2703 enum path_treatment state
= path_none
;
2706 * For each directory component of path, we are going to check whether
2707 * that path is relevant given the pathspec. For example, if path is
2709 * then we will ask treat_path() whether we should go into foo, then
2710 * whether we should go into bar, then whether baz is relevant.
2711 * Checking each is important because e.g. if path is
2713 * then we need to check .git to know we shouldn't traverse it.
2714 * If the return from treat_path() is:
2715 * * path_none, for any path, we return false.
2716 * * path_recurse, for all path components, we return true
2717 * * <anything else> for some intermediate component, we make sure
2718 * to add that path to the relevant list but return false
2719 * signifying that we shouldn't recurse into it.
2722 while (len
&& path
[len
- 1] == '/')
2727 memset(&cdir
, 0, sizeof(cdir
));
2728 cdir
.d_type
= DT_DIR
;
2732 prevlen
= baselen
+ !!baselen
;
2733 cp
= path
+ prevlen
;
2734 cp
= memchr(cp
, '/', path
+ len
- cp
);
2738 baselen
= cp
- path
;
2740 strbuf_add(&sb
, path
, baselen
);
2741 if (!is_directory(sb
.buf
))
2744 strbuf_add(&sb
, path
, prevlen
);
2745 strbuf_reset(&subdir
);
2746 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2747 cdir
.d_name
= subdir
.buf
;
2748 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2750 if (state
!= path_recurse
)
2751 break; /* do not recurse into it */
2753 break; /* finished checking */
2755 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2756 &sb
, baselen
, pathspec
,
2759 strbuf_release(&subdir
);
2760 strbuf_release(&sb
);
2761 return state
== path_recurse
;
2764 static const char *get_ident_string(void)
2766 static struct strbuf sb
= STRBUF_INIT
;
2771 if (uname(&uts
) < 0)
2772 die_errno(_("failed to get kernel name and information"));
2773 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2778 static int ident_in_untracked(const struct untracked_cache
*uc
)
2781 * Previous git versions may have saved many NUL separated
2782 * strings in the "ident" field, but it is insane to manage
2783 * many locations, so just take care of the first one.
2786 return !strcmp(uc
->ident
.buf
, get_ident_string());
2789 static void set_untracked_ident(struct untracked_cache
*uc
)
2791 strbuf_reset(&uc
->ident
);
2792 strbuf_addstr(&uc
->ident
, get_ident_string());
2795 * This strbuf used to contain a list of NUL separated
2796 * strings, so save NUL too for backward compatibility.
2798 strbuf_addch(&uc
->ident
, 0);
2801 static unsigned new_untracked_cache_flags(struct index_state
*istate
)
2803 struct repository
*repo
= istate
->repo
;
2807 * This logic is coordinated with the setting of these flags in
2808 * wt-status.c#wt_status_collect_untracked(), and the evaluation
2809 * of the config setting in commit.c#git_status_config()
2811 if (!repo_config_get_string(repo
, "status.showuntrackedfiles", &val
) &&
2812 !strcmp(val
, "all"))
2816 * The default, if "all" is not set, is "normal" - leading us here.
2817 * If the value is "none" then it really doesn't matter.
2819 return DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2822 static void new_untracked_cache(struct index_state
*istate
, int flags
)
2824 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2825 strbuf_init(&uc
->ident
, 100);
2826 uc
->exclude_per_dir
= ".gitignore";
2827 uc
->dir_flags
= flags
>= 0 ? flags
: new_untracked_cache_flags(istate
);
2828 set_untracked_ident(uc
);
2829 istate
->untracked
= uc
;
2830 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2833 void add_untracked_cache(struct index_state
*istate
)
2835 if (!istate
->untracked
) {
2836 new_untracked_cache(istate
, -1);
2838 if (!ident_in_untracked(istate
->untracked
)) {
2839 free_untracked_cache(istate
->untracked
);
2840 new_untracked_cache(istate
, -1);
2845 void remove_untracked_cache(struct index_state
*istate
)
2847 if (istate
->untracked
) {
2848 free_untracked_cache(istate
->untracked
);
2849 istate
->untracked
= NULL
;
2850 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2854 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2856 const struct pathspec
*pathspec
,
2857 struct index_state
*istate
)
2859 struct untracked_cache_dir
*root
;
2860 static int untracked_cache_disabled
= -1;
2862 if (!dir
->untracked
)
2864 if (untracked_cache_disabled
< 0)
2865 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2866 if (untracked_cache_disabled
)
2870 * We only support $GIT_DIR/info/exclude and core.excludesfile
2871 * as the global ignore rule files. Any other additions
2872 * (e.g. from command line) invalidate the cache. This
2873 * condition also catches running setup_standard_excludes()
2874 * before setting dir->untracked!
2876 if (dir
->internal
.unmanaged_exclude_files
)
2880 * Optimize for the main use case only: whole-tree git
2881 * status. More work involved in treat_leading_path() if we
2882 * use cache on just a subset of the worktree. pathspec
2883 * support could make the matter even worse.
2885 if (base_len
|| (pathspec
&& pathspec
->nr
))
2888 /* We don't support collecting ignore files */
2889 if (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2890 DIR_COLLECT_IGNORED
))
2894 * If we use .gitignore in the cache and now you change it to
2895 * .gitexclude, everything will go wrong.
2897 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2898 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2902 * EXC_CMDL is not considered in the cache. If people set it,
2905 if (dir
->internal
.exclude_list_group
[EXC_CMDL
].nr
)
2908 if (!ident_in_untracked(dir
->untracked
)) {
2909 warning(_("untracked cache is disabled on this system or location"));
2914 * If the untracked structure we received does not have the same flags
2915 * as requested in this run, we're going to need to either discard the
2916 * existing structure (and potentially later recreate), or bypass the
2917 * untracked cache mechanism for this run.
2919 if (dir
->flags
!= dir
->untracked
->dir_flags
) {
2921 * If the untracked structure we received does not have the same flags
2922 * as configured, then we need to reset / create a new "untracked"
2923 * structure to match the new config.
2925 * Keeping the saved and used untracked cache consistent with the
2926 * configuration provides an opportunity for frequent users of
2927 * "git status -uall" to leverage the untracked cache by aligning their
2928 * configuration - setting "status.showuntrackedfiles" to "all" or
2929 * "normal" as appropriate.
2931 * Previously using -uall (or setting "status.showuntrackedfiles" to
2932 * "all") was incompatible with untracked cache and *consistently*
2933 * caused surprisingly bad performance (with fscache and fsmonitor
2934 * enabled) on Windows.
2936 * IMPROVEMENT OPPORTUNITY: If we reworked the untracked cache storage
2937 * to not be as bound up with the desired output in a given run,
2938 * and instead iterated through and stored enough information to
2939 * correctly serve both "modes", then users could get peak performance
2940 * with or without '-uall' regardless of their
2941 * "status.showuntrackedfiles" config.
2943 if (dir
->untracked
->dir_flags
!= new_untracked_cache_flags(istate
)) {
2944 free_untracked_cache(istate
->untracked
);
2945 new_untracked_cache(istate
, dir
->flags
);
2946 dir
->untracked
= istate
->untracked
;
2950 * Current untracked cache data is consistent with config, but not
2951 * usable in this request/run; just bypass untracked cache.
2957 if (!dir
->untracked
->root
) {
2958 /* Untracked cache existed but is not initialized; fix that */
2959 FLEX_ALLOC_STR(dir
->untracked
->root
, name
, "");
2960 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2963 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2964 root
= dir
->untracked
->root
;
2965 if (!oideq(&dir
->internal
.ss_info_exclude
.oid
,
2966 &dir
->untracked
->ss_info_exclude
.oid
)) {
2967 invalidate_gitignore(dir
->untracked
, root
);
2968 dir
->untracked
->ss_info_exclude
= dir
->internal
.ss_info_exclude
;
2970 if (!oideq(&dir
->internal
.ss_excludes_file
.oid
,
2971 &dir
->untracked
->ss_excludes_file
.oid
)) {
2972 invalidate_gitignore(dir
->untracked
, root
);
2973 dir
->untracked
->ss_excludes_file
= dir
->internal
.ss_excludes_file
;
2976 /* Make sure this directory is not dropped out at saving phase */
2981 static void emit_traversal_statistics(struct dir_struct
*dir
,
2982 struct repository
*repo
,
2986 if (!trace2_is_enabled())
2990 trace2_data_string("read_directory", repo
, "path", "");
2992 struct strbuf tmp
= STRBUF_INIT
;
2993 strbuf_add(&tmp
, path
, path_len
);
2994 trace2_data_string("read_directory", repo
, "path", tmp
.buf
);
2995 strbuf_release(&tmp
);
2998 trace2_data_intmax("read_directory", repo
,
2999 "directories-visited", dir
->internal
.visited_directories
);
3000 trace2_data_intmax("read_directory", repo
,
3001 "paths-visited", dir
->internal
.visited_paths
);
3003 if (!dir
->untracked
)
3005 trace2_data_intmax("read_directory", repo
,
3006 "node-creation", dir
->untracked
->dir_created
);
3007 trace2_data_intmax("read_directory", repo
,
3008 "gitignore-invalidation",
3009 dir
->untracked
->gitignore_invalidated
);
3010 trace2_data_intmax("read_directory", repo
,
3011 "directory-invalidation",
3012 dir
->untracked
->dir_invalidated
);
3013 trace2_data_intmax("read_directory", repo
,
3014 "opendir", dir
->untracked
->dir_opened
);
3017 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
3018 const char *path
, int len
, const struct pathspec
*pathspec
)
3020 struct untracked_cache_dir
*untracked
;
3022 trace2_region_enter("dir", "read_directory", istate
->repo
);
3023 dir
->internal
.visited_paths
= 0;
3024 dir
->internal
.visited_directories
= 0;
3026 if (has_symlink_leading_path(path
, len
)) {
3027 trace2_region_leave("dir", "read_directory", istate
->repo
);
3031 untracked
= validate_untracked_cache(dir
, len
, pathspec
, istate
);
3034 * make sure untracked cache code path is disabled,
3035 * e.g. prep_exclude()
3037 dir
->untracked
= NULL
;
3038 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
3039 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
3040 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
3041 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
3043 emit_traversal_statistics(dir
, istate
->repo
, path
, len
);
3045 trace2_region_leave("dir", "read_directory", istate
->repo
);
3046 if (dir
->untracked
) {
3047 static int force_untracked_cache
= -1;
3049 if (force_untracked_cache
< 0)
3050 force_untracked_cache
=
3051 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", -1);
3052 if (force_untracked_cache
< 0)
3053 force_untracked_cache
= (istate
->repo
->settings
.core_untracked_cache
== UNTRACKED_CACHE_WRITE
);
3054 if (force_untracked_cache
&&
3055 dir
->untracked
== istate
->untracked
&&
3056 (dir
->untracked
->dir_opened
||
3057 dir
->untracked
->gitignore_invalidated
||
3058 dir
->untracked
->dir_invalidated
))
3059 istate
->cache_changed
|= UNTRACKED_CHANGED
;
3060 if (dir
->untracked
!= istate
->untracked
) {
3061 FREE_AND_NULL(dir
->untracked
);
3068 int file_exists(const char *f
)
3071 return lstat(f
, &sb
) == 0;
3074 int repo_file_exists(struct repository
*repo
, const char *path
)
3076 if (repo
!= the_repository
)
3077 BUG("do not know how to check file existence in arbitrary repo");
3079 return file_exists(path
);
3082 static int cmp_icase(char a
, char b
)
3087 return toupper(a
) - toupper(b
);
3092 * Given two normalized paths (a trailing slash is ok), if subdir is
3093 * outside dir, return -1. Otherwise return the offset in subdir that
3094 * can be used as relative path to dir.
3096 int dir_inside_of(const char *subdir
, const char *dir
)
3100 assert(dir
&& subdir
&& *dir
&& *subdir
);
3102 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
3108 /* hel[p]/me vs hel[l]/yeah */
3109 if (*dir
&& *subdir
)
3113 return !*dir
? offset
: -1; /* same dir */
3115 /* foo/[b]ar vs foo/[] */
3116 if (is_dir_sep(dir
[-1]))
3117 return is_dir_sep(subdir
[-1]) ? offset
: -1;
3119 /* foo[/]bar vs foo[] */
3120 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
3123 int is_inside_dir(const char *dir
)
3132 rc
= (dir_inside_of(cwd
, dir
) >= 0);
3137 int is_empty_dir(const char *path
)
3139 DIR *dir
= opendir(path
);
3146 e
= readdir_skip_dot_and_dotdot(dir
);
3154 char *git_url_basename(const char *repo
, int is_bundle
, int is_bare
)
3156 const char *end
= repo
+ strlen(repo
), *start
, *ptr
;
3163 start
= strstr(repo
, "://");
3170 * Skip authentication data. The stripping does happen
3171 * greedily, such that we strip up to the last '@' inside
3174 for (ptr
= start
; ptr
< end
&& !is_dir_sep(*ptr
); ptr
++) {
3180 * Strip trailing spaces, slashes and /.git
3182 while (start
< end
&& (is_dir_sep(end
[-1]) || isspace(end
[-1])))
3184 if (end
- start
> 5 && is_dir_sep(end
[-5]) &&
3185 !strncmp(end
- 4, ".git", 4)) {
3187 while (start
< end
&& is_dir_sep(end
[-1]))
3192 * It should not be possible to overflow `ptrdiff_t` by passing in an
3193 * insanely long URL, but GCC does not know that and will complain
3194 * without this check.
3196 if (end
- start
< 0)
3197 die(_("No directory name could be guessed.\n"
3198 "Please specify a directory on the command line"));
3201 * Strip trailing port number if we've got only a
3202 * hostname (that is, there is no dir separator but a
3203 * colon). This check is required such that we do not
3204 * strip URI's like '/foo/bar:2222.git', which should
3205 * result in a dir '2222' being guessed due to backwards
3208 if (memchr(start
, '/', end
- start
) == NULL
3209 && memchr(start
, ':', end
- start
) != NULL
) {
3211 while (start
< ptr
&& isdigit(ptr
[-1]) && ptr
[-1] != ':')
3213 if (start
< ptr
&& ptr
[-1] == ':')
3218 * Find last component. To remain backwards compatible we
3219 * also regard colons as path separators, such that
3220 * cloning a repository 'foo:bar.git' would result in a
3221 * directory 'bar' being guessed.
3224 while (start
< ptr
&& !is_dir_sep(ptr
[-1]) && ptr
[-1] != ':')
3229 * Strip .{bundle,git}.
3232 strip_suffix_mem(start
, &len
, is_bundle
? ".bundle" : ".git");
3234 if (!len
|| (len
== 1 && *start
== '/'))
3235 die(_("No directory name could be guessed.\n"
3236 "Please specify a directory on the command line"));
3239 dir
= xstrfmt("%.*s.git", (int)len
, start
);
3241 dir
= xstrndup(start
, len
);
3243 * Replace sequences of 'control' characters and whitespace
3244 * with one ascii space, remove leading and trailing spaces.
3248 int prev_space
= 1 /* strip leading whitespace */;
3249 for (end
= dir
; *end
; ++end
) {
3251 if ((unsigned char)ch
< '\x20')
3262 if (out
> dir
&& prev_space
)
3268 void strip_dir_trailing_slashes(char *dir
)
3270 char *end
= dir
+ strlen(dir
);
3272 while (dir
< end
- 1 && is_dir_sep(end
[-1]))
3277 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
3281 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
3282 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
3283 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
3284 int purge_original_cwd
= (flag
& REMOVE_DIR_PURGE_ORIGINAL_CWD
);
3285 struct object_id submodule_head
;
3287 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
3288 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
3289 /* Do not descend and nuke a nested git work tree. */
3295 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
3296 dir
= opendir(path
->buf
);
3298 if (errno
== ENOENT
)
3299 return keep_toplevel
? -1 : 0;
3300 else if (errno
== EACCES
&& !keep_toplevel
)
3302 * An empty dir could be removable even if it
3305 return rmdir(path
->buf
);
3309 strbuf_complete(path
, '/');
3312 while ((e
= readdir_skip_dot_and_dotdot(dir
)) != NULL
) {
3315 strbuf_setlen(path
, len
);
3316 strbuf_addstr(path
, e
->d_name
);
3317 if (lstat(path
->buf
, &st
)) {
3318 if (errno
== ENOENT
)
3320 * file disappeared, which is what we
3325 } else if (S_ISDIR(st
.st_mode
)) {
3326 if (!remove_dir_recurse(path
, flag
, &kept_down
))
3327 continue; /* happy */
3328 } else if (!only_empty
&&
3329 (!unlink(path
->buf
) || errno
== ENOENT
)) {
3330 continue; /* happy, too */
3333 /* path too long, stat fails, or non-directory still exists */
3339 strbuf_setlen(path
, original_len
);
3340 if (!ret
&& !keep_toplevel
&& !kept_down
) {
3341 if (!purge_original_cwd
&&
3342 startup_info
->original_cwd
&&
3343 !strcmp(startup_info
->original_cwd
, path
->buf
))
3344 ret
= -1; /* Do not remove current working directory */
3346 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
3349 * report the uplevel that it is not an error that we
3350 * did not rmdir() our directory.
3356 int remove_dir_recursively(struct strbuf
*path
, int flag
)
3358 return remove_dir_recurse(path
, flag
, NULL
);
3361 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
3363 void setup_standard_excludes(struct dir_struct
*dir
)
3365 dir
->exclude_per_dir
= ".gitignore";
3367 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3369 excludes_file
= xdg_config_home("ignore");
3370 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
3371 add_patterns_from_file_1(dir
, excludes_file
,
3372 dir
->untracked
? &dir
->internal
.ss_excludes_file
: NULL
);
3374 /* per repository user preference */
3375 if (startup_info
->have_repository
) {
3376 const char *path
= git_path_info_exclude();
3377 if (!access_or_warn(path
, R_OK
, 0))
3378 add_patterns_from_file_1(dir
, path
,
3379 dir
->untracked
? &dir
->internal
.ss_info_exclude
: NULL
);
3383 char *get_sparse_checkout_filename(void)
3385 return git_pathdup("info/sparse-checkout");
3388 int get_sparse_checkout_patterns(struct pattern_list
*pl
)
3391 char *sparse_filename
= get_sparse_checkout_filename();
3393 pl
->use_cone_patterns
= core_sparse_checkout_cone
;
3394 res
= add_patterns_from_file_to_list(sparse_filename
, "", 0, pl
, NULL
, 0);
3396 free(sparse_filename
);
3400 int remove_path(const char *name
)
3404 if (unlink(name
) && !is_missing_file_error(errno
))
3407 slash
= strrchr(name
, '/');
3409 char *dirs
= xstrdup(name
);
3410 slash
= dirs
+ (slash
- name
);
3413 if (startup_info
->original_cwd
&&
3414 !strcmp(startup_info
->original_cwd
, dirs
))
3416 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3423 * Frees memory within dir which was allocated, and resets fields for further
3424 * use. Does not free dir itself.
3426 void dir_clear(struct dir_struct
*dir
)
3429 struct exclude_list_group
*group
;
3430 struct pattern_list
*pl
;
3431 struct exclude_stack
*stk
;
3432 struct dir_struct
new = DIR_INIT
;
3434 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3435 group
= &dir
->internal
.exclude_list_group
[i
];
3436 for (j
= 0; j
< group
->nr
; j
++) {
3439 free((char *)pl
->src
);
3440 clear_pattern_list(pl
);
3445 for (i
= 0; i
< dir
->ignored_nr
; i
++)
3446 free(dir
->ignored
[i
]);
3447 for (i
= 0; i
< dir
->nr
; i
++)
3448 free(dir
->entries
[i
]);
3452 stk
= dir
->internal
.exclude_stack
;
3454 struct exclude_stack
*prev
= stk
->prev
;
3458 strbuf_release(&dir
->internal
.basebuf
);
3460 memcpy(dir
, &new, sizeof(*dir
));
3463 struct ondisk_untracked_cache
{
3464 struct stat_data info_exclude_stat
;
3465 struct stat_data excludes_file_stat
;
3469 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3472 int index
; /* number of written untracked_cache_dir */
3473 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3474 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3475 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3477 struct strbuf sb_stat
;
3478 struct strbuf sb_sha1
;
3481 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3483 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3484 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3485 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3486 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3487 to
->sd_dev
= htonl(from
->sd_dev
);
3488 to
->sd_ino
= htonl(from
->sd_ino
);
3489 to
->sd_uid
= htonl(from
->sd_uid
);
3490 to
->sd_gid
= htonl(from
->sd_gid
);
3491 to
->sd_size
= htonl(from
->sd_size
);
3494 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3495 struct write_data
*wd
)
3497 struct stat_data stat_data
;
3498 struct strbuf
*out
= &wd
->out
;
3499 unsigned char intbuf
[16];
3500 unsigned int intlen
, value
;
3501 int i
= wd
->index
++;
3504 * untracked_nr should be reset whenever valid is clear, but
3507 if (!untracked
->valid
) {
3508 untracked
->untracked_nr
= 0;
3509 untracked
->check_only
= 0;
3512 if (untracked
->check_only
)
3513 ewah_set(wd
->check_only
, i
);
3514 if (untracked
->valid
) {
3515 ewah_set(wd
->valid
, i
);
3516 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3517 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3519 if (!is_null_oid(&untracked
->exclude_oid
)) {
3520 ewah_set(wd
->sha1_valid
, i
);
3521 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3522 the_hash_algo
->rawsz
);
3525 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3526 strbuf_add(out
, intbuf
, intlen
);
3528 /* skip non-recurse directories */
3529 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3530 if (untracked
->dirs
[i
]->recurse
)
3532 intlen
= encode_varint(value
, intbuf
);
3533 strbuf_add(out
, intbuf
, intlen
);
3535 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3537 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3538 strbuf_add(out
, untracked
->untracked
[i
],
3539 strlen(untracked
->untracked
[i
]) + 1);
3541 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3542 if (untracked
->dirs
[i
]->recurse
)
3543 write_one_dir(untracked
->dirs
[i
], wd
);
3546 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3548 struct ondisk_untracked_cache
*ouc
;
3549 struct write_data wd
;
3550 unsigned char varbuf
[16];
3552 const unsigned hashsz
= the_hash_algo
->rawsz
;
3554 CALLOC_ARRAY(ouc
, 1);
3555 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3556 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3557 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3559 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3560 strbuf_add(out
, varbuf
, varint_len
);
3561 strbuf_addbuf(out
, &untracked
->ident
);
3563 strbuf_add(out
, ouc
, sizeof(*ouc
));
3564 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3565 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3566 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3569 if (!untracked
->root
) {
3570 varint_len
= encode_varint(0, varbuf
);
3571 strbuf_add(out
, varbuf
, varint_len
);
3576 wd
.check_only
= ewah_new();
3577 wd
.valid
= ewah_new();
3578 wd
.sha1_valid
= ewah_new();
3579 strbuf_init(&wd
.out
, 1024);
3580 strbuf_init(&wd
.sb_stat
, 1024);
3581 strbuf_init(&wd
.sb_sha1
, 1024);
3582 write_one_dir(untracked
->root
, &wd
);
3584 varint_len
= encode_varint(wd
.index
, varbuf
);
3585 strbuf_add(out
, varbuf
, varint_len
);
3586 strbuf_addbuf(out
, &wd
.out
);
3587 ewah_serialize_strbuf(wd
.valid
, out
);
3588 ewah_serialize_strbuf(wd
.check_only
, out
);
3589 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3590 strbuf_addbuf(out
, &wd
.sb_stat
);
3591 strbuf_addbuf(out
, &wd
.sb_sha1
);
3592 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3594 ewah_free(wd
.valid
);
3595 ewah_free(wd
.check_only
);
3596 ewah_free(wd
.sha1_valid
);
3597 strbuf_release(&wd
.out
);
3598 strbuf_release(&wd
.sb_stat
);
3599 strbuf_release(&wd
.sb_sha1
);
3602 static void free_untracked(struct untracked_cache_dir
*ucd
)
3607 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3608 free_untracked(ucd
->dirs
[i
]);
3609 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3610 free(ucd
->untracked
[i
]);
3611 free(ucd
->untracked
);
3616 void free_untracked_cache(struct untracked_cache
*uc
)
3621 free(uc
->exclude_per_dir_to_free
);
3622 strbuf_release(&uc
->ident
);
3623 free_untracked(uc
->root
);
3629 struct untracked_cache_dir
**ucd
;
3630 struct ewah_bitmap
*check_only
;
3631 struct ewah_bitmap
*valid
;
3632 struct ewah_bitmap
*sha1_valid
;
3633 const unsigned char *data
;
3634 const unsigned char *end
;
3637 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3639 memcpy(to
, data
, sizeof(*to
));
3640 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3641 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3642 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3643 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3644 to
->sd_dev
= ntohl(to
->sd_dev
);
3645 to
->sd_ino
= ntohl(to
->sd_ino
);
3646 to
->sd_uid
= ntohl(to
->sd_uid
);
3647 to
->sd_gid
= ntohl(to
->sd_gid
);
3648 to
->sd_size
= ntohl(to
->sd_size
);
3651 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3652 struct read_data
*rd
)
3654 struct untracked_cache_dir ud
, *untracked
;
3655 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3656 const unsigned char *eos
;
3660 memset(&ud
, 0, sizeof(ud
));
3662 value
= decode_varint(&data
);
3666 ud
.untracked_alloc
= value
;
3667 ud
.untracked_nr
= value
;
3668 if (ud
.untracked_nr
)
3669 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3671 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3674 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3676 eos
= memchr(data
, '\0', end
- data
);
3677 if (!eos
|| eos
== end
)
3680 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3681 memcpy(untracked
, &ud
, sizeof(ud
));
3682 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3685 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3686 eos
= memchr(data
, '\0', end
- data
);
3687 if (!eos
|| eos
== end
)
3689 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3693 rd
->ucd
[rd
->index
++] = untracked
;
3696 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3697 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3703 static void set_check_only(size_t pos
, void *cb
)
3705 struct read_data
*rd
= cb
;
3706 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3710 static void read_stat(size_t pos
, void *cb
)
3712 struct read_data
*rd
= cb
;
3713 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3714 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3715 rd
->data
= rd
->end
+ 1;
3718 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3719 rd
->data
+= sizeof(struct stat_data
);
3723 static void read_oid(size_t pos
, void *cb
)
3725 struct read_data
*rd
= cb
;
3726 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3727 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3728 rd
->data
= rd
->end
+ 1;
3731 oidread(&ud
->exclude_oid
, rd
->data
);
3732 rd
->data
+= the_hash_algo
->rawsz
;
3735 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3736 const unsigned char *sha1
)
3738 stat_data_from_disk(&oid_stat
->stat
, data
);
3739 oidread(&oid_stat
->oid
, sha1
);
3740 oid_stat
->valid
= 1;
3743 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3745 struct untracked_cache
*uc
;
3746 struct read_data rd
;
3747 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3751 const char *exclude_per_dir
;
3752 const unsigned hashsz
= the_hash_algo
->rawsz
;
3753 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3754 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3756 if (sz
<= 1 || end
[-1] != '\0')
3760 ident_len
= decode_varint(&next
);
3761 if (next
+ ident_len
> end
)
3763 ident
= (const char *)next
;
3766 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3769 CALLOC_ARRAY(uc
, 1);
3770 strbuf_init(&uc
->ident
, ident_len
);
3771 strbuf_add(&uc
->ident
, ident
, ident_len
);
3772 load_oid_stat(&uc
->ss_info_exclude
,
3773 next
+ ouc_offset(info_exclude_stat
),
3775 load_oid_stat(&uc
->ss_excludes_file
,
3776 next
+ ouc_offset(excludes_file_stat
),
3777 next
+ offset
+ hashsz
);
3778 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3779 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3780 uc
->exclude_per_dir
= uc
->exclude_per_dir_to_free
= xstrdup(exclude_per_dir
);
3781 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3782 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3786 len
= decode_varint(&next
);
3787 if (next
> end
|| len
== 0)
3790 rd
.valid
= ewah_new();
3791 rd
.check_only
= ewah_new();
3792 rd
.sha1_valid
= ewah_new();
3796 ALLOC_ARRAY(rd
.ucd
, len
);
3798 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3802 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3807 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3812 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3816 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3817 rd
.data
= next
+ len
;
3818 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3819 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3824 ewah_free(rd
.valid
);
3825 ewah_free(rd
.check_only
);
3826 ewah_free(rd
.sha1_valid
);
3829 free_untracked_cache(uc
);
3835 static void invalidate_one_directory(struct untracked_cache
*uc
,
3836 struct untracked_cache_dir
*ucd
)
3838 uc
->dir_invalidated
++;
3840 ucd
->untracked_nr
= 0;
3844 * Normally when an entry is added or removed from a directory,
3845 * invalidating that directory is enough. No need to touch its
3846 * ancestors. When a directory is shown as "foo/bar/" in git-status
3847 * however, deleting or adding an entry may have cascading effect.
3849 * Say the "foo/bar/file" has become untracked, we need to tell the
3850 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3851 * directory any more (because "bar" is managed by foo as an untracked
3854 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3855 * was the last untracked entry in the entire "foo", we should show
3856 * "foo/" instead. Which means we have to invalidate past "bar" up to
3859 * This function traverses all directories from root to leaf. If there
3860 * is a chance of one of the above cases happening, we invalidate back
3861 * to root. Otherwise we just invalidate the leaf. There may be a more
3862 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3863 * detect these cases and avoid unnecessary invalidation, for example,
3864 * checking for the untracked entry named "bar/" in "foo", but for now
3865 * stick to something safe and simple.
3867 static int invalidate_one_component(struct untracked_cache
*uc
,
3868 struct untracked_cache_dir
*dir
,
3869 const char *path
, int len
)
3871 const char *rest
= strchr(path
, '/');
3874 int component_len
= rest
- path
;
3875 struct untracked_cache_dir
*d
=
3876 lookup_untracked(uc
, dir
, path
, component_len
);
3878 invalidate_one_component(uc
, d
, rest
+ 1,
3879 len
- (component_len
+ 1));
3881 invalidate_one_directory(uc
, dir
);
3885 invalidate_one_directory(uc
, dir
);
3886 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3889 void untracked_cache_invalidate_path(struct index_state
*istate
,
3890 const char *path
, int safe_path
)
3892 if (!istate
->untracked
|| !istate
->untracked
->root
)
3894 if (!safe_path
&& !verify_path(path
, 0))
3896 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3897 path
, strlen(path
));
3900 void untracked_cache_remove_from_index(struct index_state
*istate
,
3903 untracked_cache_invalidate_path(istate
, path
, 1);
3906 void untracked_cache_add_to_index(struct index_state
*istate
,
3909 untracked_cache_invalidate_path(istate
, path
, 1);
3912 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3913 const char *sub_gitdir
)
3916 struct repository subrepo
;
3917 struct strbuf sub_wt
= STRBUF_INIT
;
3918 struct strbuf sub_gd
= STRBUF_INIT
;
3920 const struct submodule
*sub
;
3922 /* If the submodule has no working tree, we can ignore it. */
3923 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3926 if (repo_read_index(&subrepo
) < 0)
3927 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3929 /* TODO: audit for interaction with sparse-index. */
3930 ensure_full_index(subrepo
.index
);
3931 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3932 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3934 if (!S_ISGITLINK(ce
->ce_mode
))
3937 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3938 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3940 * Skip entries with the same name in different stages
3941 * to make sure an entry is returned only once.
3945 sub
= submodule_from_path(&subrepo
, null_oid(), ce
->name
);
3946 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3947 /* .gitmodules broken or inactive sub */
3950 strbuf_reset(&sub_wt
);
3951 strbuf_reset(&sub_gd
);
3952 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3953 submodule_name_to_gitdir(&sub_gd
, &subrepo
, sub
->name
);
3955 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3957 strbuf_release(&sub_wt
);
3958 strbuf_release(&sub_gd
);
3959 repo_clear(&subrepo
);
3962 void connect_work_tree_and_git_dir(const char *work_tree_
,
3963 const char *git_dir_
,
3964 int recurse_into_nested
)
3966 struct strbuf gitfile_sb
= STRBUF_INIT
;
3967 struct strbuf cfg_sb
= STRBUF_INIT
;
3968 struct strbuf rel_path
= STRBUF_INIT
;
3969 char *git_dir
, *work_tree
;
3971 /* Prepare .git file */
3972 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3973 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3974 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3976 /* Prepare config file */
3977 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3978 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3979 die(_("could not create directories for %s"), cfg_sb
.buf
);
3981 git_dir
= real_pathdup(git_dir_
, 1);
3982 work_tree
= real_pathdup(work_tree_
, 1);
3984 /* Write .git file */
3985 write_file(gitfile_sb
.buf
, "gitdir: %s",
3986 relative_path(git_dir
, work_tree
, &rel_path
));
3987 /* Update core.worktree setting */
3988 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3989 relative_path(work_tree
, git_dir
, &rel_path
));
3991 strbuf_release(&gitfile_sb
);
3992 strbuf_release(&cfg_sb
);
3993 strbuf_release(&rel_path
);
3995 if (recurse_into_nested
)
3996 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
4003 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
4005 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
4007 if (rename(old_git_dir
, new_git_dir
) < 0)
4008 die_errno(_("could not migrate git directory from '%s' to '%s'"),
4009 old_git_dir
, new_git_dir
);
4011 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);
4014 int path_match_flags(const char *const str
, const enum path_match_flags flags
)
4016 const char *p
= str
;
4018 if (flags
& PATH_MATCH_NATIVE
&&
4019 flags
& PATH_MATCH_XPLATFORM
)
4020 BUG("path_match_flags() must get one match kind, not multiple!");
4021 else if (!(flags
& PATH_MATCH_KINDS_MASK
))
4022 BUG("path_match_flags() must get at least one match kind!");
4024 if (flags
& PATH_MATCH_STARTS_WITH_DOT_SLASH
&&
4025 flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
)
4026 BUG("path_match_flags() must get one platform kind, not multiple!");
4027 else if (!(flags
& PATH_MATCH_PLATFORM_MASK
))
4028 BUG("path_match_flags() must get at least one platform kind!");
4032 if (flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
&&
4036 if (flags
& PATH_MATCH_NATIVE
)
4037 return is_dir_sep(*p
);
4038 else if (flags
& PATH_MATCH_XPLATFORM
)
4039 return is_xplatform_dir_sep(*p
);