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 "object-store.h"
18 #include "wildmatch.h"
22 #include "ewah/ewok.h"
23 #include "fsmonitor.h"
25 #include "submodule-config.h"
30 * Tells read_directory_recursive how a file or directory should be treated.
31 * Values are ordered by significance, e.g. if a directory contains both
32 * excluded and untracked files, it is listed as untracked because
33 * path_untracked > path_excluded.
43 * Support data structure for our opendir/readdir/closedir wrappers
47 struct untracked_cache_dir
*untracked
;
54 struct untracked_cache_dir
*ucd
;
57 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
58 struct index_state
*istate
, const char *path
, int len
,
59 struct untracked_cache_dir
*untracked
,
60 int check_only
, int stop_at_first_file
, const struct pathspec
*pathspec
);
61 static int resolve_dtype(int dtype
, struct index_state
*istate
,
62 const char *path
, int len
);
63 struct dirent
*readdir_skip_dot_and_dotdot(DIR *dirp
)
67 while ((e
= readdir(dirp
)) != NULL
) {
68 if (!is_dot_or_dotdot(e
->d_name
))
74 int count_slashes(const char *s
)
83 int fspathcmp(const char *a
, const char *b
)
85 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
88 int fspatheq(const char *a
, const char *b
)
90 return !fspathcmp(a
, b
);
93 int fspathncmp(const char *a
, const char *b
, size_t count
)
95 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
98 unsigned int fspathhash(const char *str
)
100 return ignore_case
? strihash(str
) : strhash(str
);
103 int git_fnmatch(const struct pathspec_item
*item
,
104 const char *pattern
, const char *string
,
108 if (ps_strncmp(item
, pattern
, string
, prefix
))
113 if (item
->flags
& PATHSPEC_ONESTAR
) {
114 int pattern_len
= strlen(++pattern
);
115 int string_len
= strlen(string
);
116 return string_len
< pattern_len
||
117 ps_strcmp(item
, pattern
,
118 string
+ string_len
- pattern_len
);
120 if (item
->magic
& PATHSPEC_GLOB
)
121 return wildmatch(pattern
, string
,
123 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
125 /* wildmatch has not learned no FNM_PATHNAME mode yet */
126 return wildmatch(pattern
, string
,
127 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
130 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
131 const char *string
, int stringlen
,
135 struct strbuf pat_buf
= STRBUF_INIT
;
136 struct strbuf str_buf
= STRBUF_INIT
;
137 const char *use_pat
= pattern
;
138 const char *use_str
= string
;
140 if (pattern
[patternlen
]) {
141 strbuf_add(&pat_buf
, pattern
, patternlen
);
142 use_pat
= pat_buf
.buf
;
144 if (string
[stringlen
]) {
145 strbuf_add(&str_buf
, string
, stringlen
);
146 use_str
= str_buf
.buf
;
150 flags
|= WM_CASEFOLD
;
151 match_status
= wildmatch(use_pat
, use_str
, flags
);
153 strbuf_release(&pat_buf
);
154 strbuf_release(&str_buf
);
159 static size_t common_prefix_len(const struct pathspec
*pathspec
)
165 * ":(icase)path" is treated as a pathspec full of
166 * wildcard. In other words, only prefix is considered common
167 * prefix. If the pathspec is abc/foo abc/bar, running in
168 * subdir xyz, the common prefix is still xyz, not xyz/abc as
171 GUARD_PATHSPEC(pathspec
,
180 for (n
= 0; n
< pathspec
->nr
; n
++) {
181 size_t i
= 0, len
= 0, item_len
;
182 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
184 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
185 item_len
= pathspec
->items
[n
].prefix
;
187 item_len
= pathspec
->items
[n
].nowildcard_len
;
188 while (i
< item_len
&& (n
== 0 || i
< max
)) {
189 char c
= pathspec
->items
[n
].match
[i
];
190 if (c
!= pathspec
->items
[0].match
[i
])
196 if (n
== 0 || len
< max
) {
206 * Returns a copy of the longest leading path common among all
209 char *common_prefix(const struct pathspec
*pathspec
)
211 unsigned long len
= common_prefix_len(pathspec
);
213 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
216 int fill_directory(struct dir_struct
*dir
,
217 struct index_state
*istate
,
218 const struct pathspec
*pathspec
)
223 unsigned exclusive_flags
= DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
;
224 if ((dir
->flags
& exclusive_flags
) == exclusive_flags
)
225 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
228 * Calculate common prefix for the pathspec, and
229 * use that to optimize the directory walk
231 prefix_len
= common_prefix_len(pathspec
);
232 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
234 /* Read the directory and prune it */
235 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
240 int within_depth(const char *name
, int namelen
,
241 int depth
, int max_depth
)
243 const char *cp
= name
, *cpe
= name
+ namelen
;
249 if (depth
> max_depth
)
256 * Read the contents of the blob with the given OID into a buffer.
257 * Append a trailing LF to the end if the last line doesn't have one.
260 * -1 when the OID is invalid or unknown or does not refer to a blob.
261 * 0 when the blob is empty.
262 * 1 along with { data, size } of the (possibly augmented) buffer
265 * Optionally updates the given oid_stat with the given OID (when valid).
267 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
268 size_t *size_out
, char **data_out
)
270 enum object_type type
;
277 data
= repo_read_object_file(the_repository
, oid
, &type
, &sz
);
278 if (!data
|| type
!= OBJ_BLOB
) {
284 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
285 oidcpy(&oid_stat
->oid
, oid
);
293 if (data
[sz
- 1] != '\n') {
294 data
= xrealloc(data
, st_add(sz
, 1));
298 *size_out
= xsize_t(sz
);
304 #define DO_MATCH_EXCLUDE (1<<0)
305 #define DO_MATCH_DIRECTORY (1<<1)
306 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
309 * Does the given pathspec match the given name? A match is found if
311 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
312 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
313 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
314 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
316 * Return value tells which case it was (1-4), or 0 when there is no match.
318 * It may be instructive to look at a small table of concrete examples
319 * to understand the differences between 1, 2, and 4:
322 * | a/b | a/b/ | a/b/c
323 * ------+-----------+-----------+------------
324 * a/b | EXACT | EXACT[1] | LEADING[2]
325 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
326 * a/b/c | RECURSIVE | RECURSIVE | EXACT
328 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
329 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
331 static int match_pathspec_item(struct index_state
*istate
,
332 const struct pathspec_item
*item
, int prefix
,
333 const char *name
, int namelen
, unsigned flags
)
335 /* name/namelen has prefix cut off by caller */
336 const char *match
= item
->match
+ prefix
;
337 int matchlen
= item
->len
- prefix
;
340 * The normal call pattern is:
341 * 1. prefix = common_prefix_len(ps);
342 * 2. prune something, or fill_directory
343 * 3. match_pathspec()
345 * 'prefix' at #1 may be shorter than the command's prefix and
346 * it's ok for #2 to match extra files. Those extras will be
349 * Suppose the pathspec is 'foo' and '../bar' running from
350 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
351 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
352 * user does not want XYZ/foo, only the "foo" part should be
353 * case-insensitive. We need to filter out XYZ/foo here. In
354 * other words, we do not trust the caller on comparing the
355 * prefix part when :(icase) is involved. We do exact
356 * comparison ourselves.
358 * Normally the caller (common_prefix_len() in fact) does
359 * _exact_ matching on name[-prefix+1..-1] and we do not need
360 * to check that part. Be defensive and check it anyway, in
361 * case common_prefix_len is changed, or a new caller is
362 * introduced that does not use common_prefix_len.
364 * If the penalty turns out too high when prefix is really
365 * long, maybe change it to
366 * strncmp(match, name, item->prefix - prefix)
368 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
369 strncmp(item
->match
, name
- prefix
, item
->prefix
))
372 if (item
->attr_match_nr
&&
373 !match_pathspec_attrs(istate
, name
, namelen
, item
))
376 /* If the match was just the prefix, we matched */
378 return MATCHED_RECURSIVELY
;
380 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
381 if (matchlen
== namelen
)
382 return MATCHED_EXACTLY
;
384 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
385 return MATCHED_RECURSIVELY
;
386 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
387 match
[matchlen
- 1] == '/' &&
388 namelen
== matchlen
- 1 &&
389 !ps_strncmp(item
, match
, name
, namelen
))
390 return MATCHED_EXACTLY
;
392 if (item
->nowildcard_len
< item
->len
&&
393 !git_fnmatch(item
, match
, name
,
394 item
->nowildcard_len
- prefix
))
395 return MATCHED_FNMATCH
;
397 /* Perform checks to see if "name" is a leading string of the pathspec */
398 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
399 !(flags
& DO_MATCH_EXCLUDE
)) {
400 /* name is a literal prefix of the pathspec */
401 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
402 if ((namelen
< matchlen
) &&
403 (match
[namelen
-offset
] == '/') &&
404 !ps_strncmp(item
, match
, name
, namelen
))
405 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
407 /* name doesn't match up to the first wild character */
408 if (item
->nowildcard_len
< item
->len
&&
409 ps_strncmp(item
, match
, name
,
410 item
->nowildcard_len
- prefix
))
414 * name has no wildcard, and it didn't match as a leading
415 * pathspec so return.
417 if (item
->nowildcard_len
== item
->len
)
421 * Here is where we would perform a wildmatch to check if
422 * "name" can be matched as a directory (or a prefix) against
423 * the pathspec. Since wildmatch doesn't have this capability
424 * at the present we have to punt and say that it is a match,
425 * potentially returning a false positive
426 * The submodules themselves will be able to perform more
427 * accurate matching to determine if the pathspec matches.
429 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
436 * do_match_pathspec() is meant to ONLY be called by
437 * match_pathspec_with_flags(); calling it directly risks pathspecs
438 * like ':!unwanted_path' being ignored.
440 * Given a name and a list of pathspecs, returns the nature of the
441 * closest (i.e. most specific) match of the name to any of the
444 * The caller typically calls this multiple times with the same
445 * pathspec and seen[] array but with different name/namelen
446 * (e.g. entries from the index) and is interested in seeing if and
447 * how each pathspec matches all the names it calls this function
448 * with. A mark is left in the seen[] array for each pathspec element
449 * indicating the closest type of match that element achieved, so if
450 * seen[n] remains zero after multiple invocations, that means the nth
451 * pathspec did not match any names, which could indicate that the
452 * user mistyped the nth pathspec.
454 static int do_match_pathspec(struct index_state
*istate
,
455 const struct pathspec
*ps
,
456 const char *name
, int namelen
,
457 int prefix
, char *seen
,
460 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
472 if (!ps
->recursive
||
473 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
475 return MATCHED_RECURSIVELY
;
477 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
478 return MATCHED_EXACTLY
;
486 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
489 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
490 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
493 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
496 * Make exclude patterns optional and never report
497 * "pathspec ':(exclude)foo' matches no files"
499 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
500 seen
[i
] = MATCHED_FNMATCH
;
501 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
504 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
505 ps
->max_depth
!= -1 &&
506 how
&& how
!= MATCHED_FNMATCH
) {
507 int len
= ps
->items
[i
].len
;
508 if (name
[len
] == '/')
510 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
511 how
= MATCHED_EXACTLY
;
518 if (seen
&& seen
[i
] < how
)
525 static int match_pathspec_with_flags(struct index_state
*istate
,
526 const struct pathspec
*ps
,
527 const char *name
, int namelen
,
528 int prefix
, char *seen
, unsigned flags
)
530 int positive
, negative
;
531 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
532 prefix
, seen
, flags
);
533 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
535 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
537 flags
| DO_MATCH_EXCLUDE
);
538 return negative
? 0 : positive
;
541 int match_pathspec(struct index_state
*istate
,
542 const struct pathspec
*ps
,
543 const char *name
, int namelen
,
544 int prefix
, char *seen
, int is_dir
)
546 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
547 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
548 prefix
, seen
, flags
);
552 * Check if a submodule is a superset of the pathspec
554 int submodule_path_match(struct index_state
*istate
,
555 const struct pathspec
*ps
,
556 const char *submodule_name
,
559 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
560 strlen(submodule_name
),
563 DO_MATCH_LEADING_PATHSPEC
);
567 int report_path_error(const char *ps_matched
,
568 const struct pathspec
*pathspec
)
571 * Make sure all pathspec matched; otherwise it is an error.
574 for (num
= 0; num
< pathspec
->nr
; num
++) {
575 int other
, found_dup
;
580 * The caller might have fed identical pathspec
581 * twice. Do not barf on such a mistake.
582 * FIXME: parse_pathspec should have eliminated
583 * duplicate pathspec.
585 for (found_dup
= other
= 0;
586 !found_dup
&& other
< pathspec
->nr
;
588 if (other
== num
|| !ps_matched
[other
])
590 if (!strcmp(pathspec
->items
[other
].original
,
591 pathspec
->items
[num
].original
))
593 * Ok, we have a match already.
600 error(_("pathspec '%s' did not match any file(s) known to git"),
601 pathspec
->items
[num
].original
);
608 * Return the length of the "simple" part of a path match limiter.
610 int simple_length(const char *match
)
615 unsigned char c
= *match
++;
617 if (c
== '\0' || is_glob_special(c
))
622 int no_wildcard(const char *string
)
624 return string
[simple_length(string
)] == '\0';
627 void parse_path_pattern(const char **pattern
,
632 const char *p
= *pattern
;
637 *flags
|= PATTERN_FLAG_NEGATIVE
;
641 if (len
&& p
[len
- 1] == '/') {
643 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
645 for (i
= 0; i
< len
; i
++) {
650 *flags
|= PATTERN_FLAG_NODIR
;
651 *nowildcardlen
= simple_length(p
);
653 * we should have excluded the trailing slash from 'p' too,
654 * but that's one more allocation. Instead just make sure
655 * nowildcardlen does not exceed real patternlen
657 if (*nowildcardlen
> len
)
658 *nowildcardlen
= len
;
659 if (*p
== '*' && no_wildcard(p
+ 1))
660 *flags
|= PATTERN_FLAG_ENDSWITH
;
665 int pl_hashmap_cmp(const void *cmp_data UNUSED
,
666 const struct hashmap_entry
*a
,
667 const struct hashmap_entry
*b
,
668 const void *key UNUSED
)
670 const struct pattern_entry
*ee1
=
671 container_of(a
, struct pattern_entry
, ent
);
672 const struct pattern_entry
*ee2
=
673 container_of(b
, struct pattern_entry
, ent
);
675 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
679 return fspathncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
682 static char *dup_and_filter_pattern(const char *pattern
)
686 char *result
= xstrdup(pattern
);
692 /* skip escape characters (once) */
712 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
714 struct pattern_entry
*translated
;
717 const char *prev
, *cur
, *next
;
719 if (!pl
->use_cone_patterns
)
722 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
723 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
724 !strcmp(given
->pattern
, "/*")) {
729 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
734 if (given
->patternlen
< 2 ||
735 *given
->pattern
!= '/' ||
736 strstr(given
->pattern
, "**")) {
737 /* Not a cone pattern. */
738 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
742 if (!(given
->flags
& PATTERN_FLAG_MUSTBEDIR
) &&
743 strcmp(given
->pattern
, "/*")) {
744 /* Not a cone pattern. */
745 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
749 prev
= given
->pattern
;
750 cur
= given
->pattern
+ 1;
751 next
= given
->pattern
+ 2;
754 /* Watch for glob characters '*', '\', '[', '?' */
755 if (!is_glob_special(*cur
))
758 /* But only if *prev != '\\' */
762 /* But allow the initial '\' */
764 is_glob_special(*next
))
767 /* But a trailing '/' then '*' is fine */
773 /* Not a cone pattern. */
774 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
783 if (given
->patternlen
> 2 &&
784 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
785 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
786 /* Not a cone pattern. */
787 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
791 truncated
= dup_and_filter_pattern(given
->pattern
);
793 translated
= xmalloc(sizeof(struct pattern_entry
));
794 translated
->pattern
= truncated
;
795 translated
->patternlen
= given
->patternlen
- 2;
796 hashmap_entry_init(&translated
->ent
,
797 fspathhash(translated
->pattern
));
799 if (!hashmap_get_entry(&pl
->recursive_hashmap
,
800 translated
, ent
, NULL
)) {
801 /* We did not see the "parent" included */
802 warning(_("unrecognized negative pattern: '%s'"),
809 hashmap_add(&pl
->parent_hashmap
, &translated
->ent
);
810 hashmap_remove(&pl
->recursive_hashmap
, &translated
->ent
, &data
);
815 if (given
->flags
& PATTERN_FLAG_NEGATIVE
) {
816 warning(_("unrecognized negative pattern: '%s'"),
821 translated
= xmalloc(sizeof(struct pattern_entry
));
823 translated
->pattern
= dup_and_filter_pattern(given
->pattern
);
824 translated
->patternlen
= given
->patternlen
;
825 hashmap_entry_init(&translated
->ent
,
826 fspathhash(translated
->pattern
));
828 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
830 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
831 /* we already included this at the parent level */
832 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
840 warning(_("disabling cone pattern matching"));
841 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
842 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
843 pl
->use_cone_patterns
= 0;
846 static int hashmap_contains_path(struct hashmap
*map
,
847 struct strbuf
*pattern
)
849 struct pattern_entry p
;
851 /* Check straight mapping */
852 p
.pattern
= pattern
->buf
;
853 p
.patternlen
= pattern
->len
;
854 hashmap_entry_init(&p
.ent
, fspathhash(p
.pattern
));
855 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
858 int hashmap_contains_parent(struct hashmap
*map
,
860 struct strbuf
*buffer
)
864 strbuf_setlen(buffer
, 0);
867 strbuf_addch(buffer
, '/');
869 strbuf_addstr(buffer
, path
);
871 slash_pos
= strrchr(buffer
->buf
, '/');
873 while (slash_pos
> buffer
->buf
) {
874 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
876 if (hashmap_contains_path(map
, buffer
))
879 slash_pos
= strrchr(buffer
->buf
, '/');
885 void add_pattern(const char *string
, const char *base
,
886 int baselen
, struct pattern_list
*pl
, int srcpos
)
888 struct path_pattern
*pattern
;
893 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
894 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
895 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
897 pattern
= xmalloc(sizeof(*pattern
));
898 pattern
->pattern
= string
;
900 pattern
->patternlen
= patternlen
;
901 pattern
->nowildcardlen
= nowildcardlen
;
902 pattern
->base
= base
;
903 pattern
->baselen
= baselen
;
904 pattern
->flags
= flags
;
905 pattern
->srcpos
= srcpos
;
906 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
907 pl
->patterns
[pl
->nr
++] = pattern
;
910 add_pattern_to_hashsets(pl
, pattern
);
913 static int read_skip_worktree_file_from_index(struct index_state
*istate
,
915 size_t *size_out
, char **data_out
,
916 struct oid_stat
*oid_stat
)
921 pos
= index_name_pos(istate
, path
, len
);
924 if (!ce_skip_worktree(istate
->cache
[pos
]))
927 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
931 * Frees memory within pl which was allocated for exclude patterns and
932 * the file buffer. Does not free pl itself.
934 void clear_pattern_list(struct pattern_list
*pl
)
938 for (i
= 0; i
< pl
->nr
; i
++)
939 free(pl
->patterns
[i
]);
942 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
943 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
945 memset(pl
, 0, sizeof(*pl
));
948 static void trim_trailing_spaces(char *buf
)
950 char *p
, *last_space
= NULL
;
952 for (p
= buf
; *p
; p
++)
972 * Given a subdirectory name and "dir" of the current directory,
973 * search the subdir in "dir" and return it, or create a new one if it
974 * does not exist in "dir".
976 * If "name" has the trailing slash, it'll be excluded in the search.
978 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
979 struct untracked_cache_dir
*dir
,
980 const char *name
, int len
)
983 struct untracked_cache_dir
*d
;
986 if (len
&& name
[len
- 1] == '/')
990 while (last
> first
) {
991 int cmp
, next
= first
+ ((last
- first
) >> 1);
993 cmp
= strncmp(name
, d
->name
, len
);
994 if (!cmp
&& strlen(d
->name
) > len
)
1006 FLEX_ALLOC_MEM(d
, name
, name
, len
);
1008 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
1009 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
1010 dir
->dirs_nr
- first
);
1012 dir
->dirs
[first
] = d
;
1016 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
1020 dir
->untracked_nr
= 0;
1021 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1022 do_invalidate_gitignore(dir
->dirs
[i
]);
1025 static void invalidate_gitignore(struct untracked_cache
*uc
,
1026 struct untracked_cache_dir
*dir
)
1028 uc
->gitignore_invalidated
++;
1029 do_invalidate_gitignore(dir
);
1032 static void invalidate_directory(struct untracked_cache
*uc
,
1033 struct untracked_cache_dir
*dir
)
1038 * Invalidation increment here is just roughly correct. If
1039 * untracked_nr or any of dirs[].recurse is non-zero, we
1040 * should increment dir_invalidated too. But that's more
1044 uc
->dir_invalidated
++;
1047 dir
->untracked_nr
= 0;
1048 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1049 dir
->dirs
[i
]->recurse
= 0;
1052 static int add_patterns_from_buffer(char *buf
, size_t size
,
1053 const char *base
, int baselen
,
1054 struct pattern_list
*pl
);
1056 /* Flags for add_patterns() */
1057 #define PATTERN_NOFOLLOW (1<<0)
1060 * Given a file with name "fname", read it (either from disk, or from
1061 * an index if 'istate' is non-null), parse it and store the
1062 * exclude rules in "pl".
1064 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1065 * stat data from disk (only valid if add_patterns returns zero). If
1066 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1068 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1069 struct pattern_list
*pl
, struct index_state
*istate
,
1070 unsigned flags
, struct oid_stat
*oid_stat
)
1078 if (flags
& PATTERN_NOFOLLOW
)
1079 fd
= open_nofollow(fname
, O_RDONLY
);
1081 fd
= open(fname
, O_RDONLY
);
1083 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1085 warn_on_fopen_errors(fname
);
1090 r
= read_skip_worktree_file_from_index(istate
, fname
,
1096 size
= xsize_t(st
.st_size
);
1099 fill_stat_data(&oid_stat
->stat
, &st
);
1100 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1101 oid_stat
->valid
= 1;
1106 buf
= xmallocz(size
);
1107 if (read_in_full(fd
, buf
, size
) != size
) {
1116 if (oid_stat
->valid
&&
1117 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1118 ; /* no content change, oid_stat->oid still good */
1120 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1121 !ce_stage(istate
->cache
[pos
]) &&
1122 ce_uptodate(istate
->cache
[pos
]) &&
1123 !would_convert_to_git(istate
, fname
))
1124 oidcpy(&oid_stat
->oid
,
1125 &istate
->cache
[pos
]->oid
);
1127 hash_object_file(the_hash_algo
, buf
, size
,
1128 OBJ_BLOB
, &oid_stat
->oid
);
1129 fill_stat_data(&oid_stat
->stat
, &st
);
1130 oid_stat
->valid
= 1;
1134 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1138 static int add_patterns_from_buffer(char *buf
, size_t size
,
1139 const char *base
, int baselen
,
1140 struct pattern_list
*pl
)
1145 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1146 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1150 if (skip_utf8_bom(&buf
, size
))
1151 size
-= buf
- pl
->filebuf
;
1155 for (i
= 0; i
< size
; i
++) {
1156 if (buf
[i
] == '\n') {
1157 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1158 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1159 trim_trailing_spaces(entry
);
1160 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1163 entry
= buf
+ i
+ 1;
1169 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1170 int baselen
, struct pattern_list
*pl
,
1171 struct index_state
*istate
,
1174 return add_patterns(fname
, base
, baselen
, pl
, istate
, flags
, NULL
);
1177 int add_patterns_from_blob_to_list(
1178 struct object_id
*oid
,
1179 const char *base
, int baselen
,
1180 struct pattern_list
*pl
)
1186 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1190 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1194 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1195 int group_type
, const char *src
)
1197 struct pattern_list
*pl
;
1198 struct exclude_list_group
*group
;
1200 group
= &dir
->internal
.exclude_list_group
[group_type
];
1201 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1202 pl
= &group
->pl
[group
->nr
++];
1203 memset(pl
, 0, sizeof(*pl
));
1209 * Used to set up core.excludesfile and .git/info/exclude lists.
1211 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1212 struct oid_stat
*oid_stat
)
1214 struct pattern_list
*pl
;
1216 * catch setup_standard_excludes() that's called before
1217 * dir->untracked is assigned. That function behaves
1218 * differently when dir->untracked is non-NULL.
1220 if (!dir
->untracked
)
1221 dir
->internal
.unmanaged_exclude_files
++;
1222 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1223 if (add_patterns(fname
, "", 0, pl
, NULL
, 0, oid_stat
) < 0)
1224 die(_("cannot use %s as an exclude file"), fname
);
1227 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1229 dir
->internal
.unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1230 add_patterns_from_file_1(dir
, fname
, NULL
);
1233 int match_basename(const char *basename
, int basenamelen
,
1234 const char *pattern
, int prefix
, int patternlen
,
1237 if (prefix
== patternlen
) {
1238 if (patternlen
== basenamelen
&&
1239 !fspathncmp(pattern
, basename
, basenamelen
))
1241 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1242 /* "*literal" matching against "fooliteral" */
1243 if (patternlen
- 1 <= basenamelen
&&
1244 !fspathncmp(pattern
+ 1,
1245 basename
+ basenamelen
- (patternlen
- 1),
1249 if (fnmatch_icase_mem(pattern
, patternlen
,
1250 basename
, basenamelen
,
1257 int match_pathname(const char *pathname
, int pathlen
,
1258 const char *base
, int baselen
,
1259 const char *pattern
, int prefix
, int patternlen
)
1265 * match with FNM_PATHNAME; the pattern has base implicitly
1268 if (*pattern
== '/') {
1275 * baselen does not count the trailing slash. base[] may or
1276 * may not end with a trailing slash though.
1278 if (pathlen
< baselen
+ 1 ||
1279 (baselen
&& pathname
[baselen
] != '/') ||
1280 fspathncmp(pathname
, base
, baselen
))
1283 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1284 name
= pathname
+ pathlen
- namelen
;
1288 * if the non-wildcard part is longer than the
1289 * remaining pathname, surely it cannot match.
1291 if (prefix
> namelen
)
1294 if (fspathncmp(pattern
, name
, prefix
))
1297 patternlen
-= prefix
;
1302 * If the whole pattern did not have a wildcard,
1303 * then our prefix match is all we need; we
1304 * do not need to call fnmatch at all.
1306 if (!patternlen
&& !namelen
)
1310 return fnmatch_icase_mem(pattern
, patternlen
,
1316 * Scan the given exclude list in reverse to see whether pathname
1317 * should be ignored. The first match (i.e. the last on the list), if
1318 * any, determines the fate. Returns the exclude_list element which
1319 * matched, or NULL for undecided.
1321 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1323 const char *basename
,
1325 struct pattern_list
*pl
,
1326 struct index_state
*istate
)
1328 struct path_pattern
*res
= NULL
; /* undecided */
1332 return NULL
; /* undefined */
1334 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1335 struct path_pattern
*pattern
= pl
->patterns
[i
];
1336 const char *exclude
= pattern
->pattern
;
1337 int prefix
= pattern
->nowildcardlen
;
1339 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1340 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1341 if (*dtype
!= DT_DIR
)
1345 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1346 if (match_basename(basename
,
1347 pathlen
- (basename
- pathname
),
1348 exclude
, prefix
, pattern
->patternlen
,
1356 assert(pattern
->baselen
== 0 ||
1357 pattern
->base
[pattern
->baselen
- 1] == '/');
1358 if (match_pathname(pathname
, pathlen
,
1360 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1361 exclude
, prefix
, pattern
->patternlen
)) {
1370 * Scan the list of patterns to determine if the ordered list
1371 * of patterns matches on 'pathname'.
1373 * Return 1 for a match, 0 for not matched and -1 for undecided.
1375 enum pattern_match_result
path_matches_pattern_list(
1376 const char *pathname
, int pathlen
,
1377 const char *basename
, int *dtype
,
1378 struct pattern_list
*pl
,
1379 struct index_state
*istate
)
1381 struct path_pattern
*pattern
;
1382 struct strbuf parent_pathname
= STRBUF_INIT
;
1383 int result
= NOT_MATCHED
;
1386 if (!pl
->use_cone_patterns
) {
1387 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1390 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1402 strbuf_addch(&parent_pathname
, '/');
1403 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1406 * Directory entries are matched if and only if a file
1407 * contained immediately within them is matched. For the
1408 * case of a directory entry, modify the path to create
1409 * a fake filename within this directory, allowing us to
1410 * use the file-base matching logic in an equivalent way.
1412 if (parent_pathname
.len
> 0 &&
1413 parent_pathname
.buf
[parent_pathname
.len
- 1] == '/') {
1414 slash_pos
= parent_pathname
.len
- 1;
1415 strbuf_add(&parent_pathname
, "-", 1);
1417 const char *slash_ptr
= strrchr(parent_pathname
.buf
, '/');
1418 slash_pos
= slash_ptr
? slash_ptr
- parent_pathname
.buf
: 0;
1421 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1422 &parent_pathname
)) {
1423 result
= MATCHED_RECURSIVE
;
1428 /* include every file in root */
1433 strbuf_setlen(&parent_pathname
, slash_pos
);
1435 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1440 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1443 result
= MATCHED_RECURSIVE
;
1446 strbuf_release(&parent_pathname
);
1450 int init_sparse_checkout_patterns(struct index_state
*istate
)
1452 if (!core_apply_sparse_checkout
)
1454 if (istate
->sparse_checkout_patterns
)
1457 CALLOC_ARRAY(istate
->sparse_checkout_patterns
, 1);
1459 if (get_sparse_checkout_patterns(istate
->sparse_checkout_patterns
) < 0) {
1460 FREE_AND_NULL(istate
->sparse_checkout_patterns
);
1467 static int path_in_sparse_checkout_1(const char *path
,
1468 struct index_state
*istate
,
1469 int require_cone_mode
)
1472 enum pattern_match_result match
= UNDECIDED
;
1473 const char *end
, *slash
;
1476 * We default to accepting a path if the path is empty, there are no
1477 * patterns, or the patterns are of the wrong type.
1480 init_sparse_checkout_patterns(istate
) ||
1481 (require_cone_mode
&&
1482 !istate
->sparse_checkout_patterns
->use_cone_patterns
))
1486 * If UNDECIDED, use the match from the parent dir (recursively), or
1487 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1488 * never returns UNDECIDED, so we will execute only one iteration in
1491 for (end
= path
+ strlen(path
);
1492 end
> path
&& match
== UNDECIDED
;
1495 for (slash
= end
- 1; slash
> path
&& *slash
!= '/'; slash
--)
1498 match
= path_matches_pattern_list(path
, end
- path
,
1499 slash
> path
? slash
+ 1 : path
, &dtype
,
1500 istate
->sparse_checkout_patterns
, istate
);
1502 /* We are going to match the parent dir now */
1508 int path_in_sparse_checkout(const char *path
,
1509 struct index_state
*istate
)
1511 return path_in_sparse_checkout_1(path
, istate
, 0);
1514 int path_in_cone_mode_sparse_checkout(const char *path
,
1515 struct index_state
*istate
)
1517 return path_in_sparse_checkout_1(path
, istate
, 1);
1520 static struct path_pattern
*last_matching_pattern_from_lists(
1521 struct dir_struct
*dir
, struct index_state
*istate
,
1522 const char *pathname
, int pathlen
,
1523 const char *basename
, int *dtype_p
)
1526 struct exclude_list_group
*group
;
1527 struct path_pattern
*pattern
;
1528 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1529 group
= &dir
->internal
.exclude_list_group
[i
];
1530 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1531 pattern
= last_matching_pattern_from_list(
1532 pathname
, pathlen
, basename
, dtype_p
,
1533 &group
->pl
[j
], istate
);
1542 * Loads the per-directory exclude list for the substring of base
1543 * which has a char length of baselen.
1545 static void prep_exclude(struct dir_struct
*dir
,
1546 struct index_state
*istate
,
1547 const char *base
, int baselen
)
1549 struct exclude_list_group
*group
;
1550 struct pattern_list
*pl
;
1551 struct exclude_stack
*stk
= NULL
;
1552 struct untracked_cache_dir
*untracked
;
1555 group
= &dir
->internal
.exclude_list_group
[EXC_DIRS
];
1558 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1559 * which originate from directories not in the prefix of the
1560 * path being checked.
1562 while ((stk
= dir
->internal
.exclude_stack
) != NULL
) {
1563 if (stk
->baselen
<= baselen
&&
1564 !strncmp(dir
->internal
.basebuf
.buf
, base
, stk
->baselen
))
1566 pl
= &group
->pl
[dir
->internal
.exclude_stack
->exclude_ix
];
1567 dir
->internal
.exclude_stack
= stk
->prev
;
1568 dir
->internal
.pattern
= NULL
;
1569 free((char *)pl
->src
); /* see strbuf_detach() below */
1570 clear_pattern_list(pl
);
1575 /* Skip traversing into sub directories if the parent is excluded */
1576 if (dir
->internal
.pattern
)
1580 * Lazy initialization. All call sites currently just
1581 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1582 * them seems lots of work for little benefit.
1584 if (!dir
->internal
.basebuf
.buf
)
1585 strbuf_init(&dir
->internal
.basebuf
, PATH_MAX
);
1587 /* Read from the parent directories and push them down. */
1588 current
= stk
? stk
->baselen
: -1;
1589 strbuf_setlen(&dir
->internal
.basebuf
, current
< 0 ? 0 : current
);
1591 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1595 while (current
< baselen
) {
1597 struct oid_stat oid_stat
;
1599 CALLOC_ARRAY(stk
, 1);
1604 cp
= strchr(base
+ current
+ 1, '/');
1606 die("oops in prep_exclude");
1609 lookup_untracked(dir
->untracked
,
1612 cp
- base
- current
);
1614 stk
->prev
= dir
->internal
.exclude_stack
;
1615 stk
->baselen
= cp
- base
;
1616 stk
->exclude_ix
= group
->nr
;
1617 stk
->ucd
= untracked
;
1618 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1619 strbuf_add(&dir
->internal
.basebuf
, base
+ current
, stk
->baselen
- current
);
1620 assert(stk
->baselen
== dir
->internal
.basebuf
.len
);
1622 /* Abort if the directory is excluded */
1625 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = 0;
1626 dir
->internal
.pattern
= last_matching_pattern_from_lists(dir
,
1628 dir
->internal
.basebuf
.buf
, stk
->baselen
- 1,
1629 dir
->internal
.basebuf
.buf
+ current
, &dt
);
1630 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = '/';
1631 if (dir
->internal
.pattern
&&
1632 dir
->internal
.pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1633 dir
->internal
.pattern
= NULL
;
1634 if (dir
->internal
.pattern
) {
1635 dir
->internal
.exclude_stack
= stk
;
1640 /* Try to read per-directory file */
1641 oidclr(&oid_stat
.oid
);
1643 if (dir
->exclude_per_dir
&&
1645 * If we know that no files have been added in
1646 * this directory (i.e. valid_cached_dir() has
1647 * been executed and set untracked->valid) ..
1649 (!untracked
|| !untracked
->valid
||
1651 * .. and .gitignore does not exist before
1652 * (i.e. null exclude_oid). Then we can skip
1653 * loading .gitignore, which would result in
1656 !is_null_oid(&untracked
->exclude_oid
))) {
1658 * dir->internal.basebuf gets reused by the traversal,
1659 * but we need fname to remain unchanged to ensure the
1660 * src member of each struct path_pattern correctly
1661 * back-references its source file. Other invocations
1662 * of add_pattern_list provide stable strings, so we
1663 * strbuf_detach() and free() here in the caller.
1665 struct strbuf sb
= STRBUF_INIT
;
1666 strbuf_addbuf(&sb
, &dir
->internal
.basebuf
);
1667 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1668 pl
->src
= strbuf_detach(&sb
, NULL
);
1669 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1671 untracked
? &oid_stat
: NULL
);
1674 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1675 * will first be called in valid_cached_dir() then maybe many
1676 * times more in last_matching_pattern(). When the cache is
1677 * used, last_matching_pattern() will not be called and
1678 * reading .gitignore content will be a waste.
1680 * So when it's called by valid_cached_dir() and we can get
1681 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1682 * modified on work tree), we could delay reading the
1683 * .gitignore content until we absolutely need it in
1684 * last_matching_pattern(). Be careful about ignore rule
1685 * order, though, if you do that.
1688 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1689 invalidate_gitignore(dir
->untracked
, untracked
);
1690 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1692 dir
->internal
.exclude_stack
= stk
;
1693 current
= stk
->baselen
;
1695 strbuf_setlen(&dir
->internal
.basebuf
, baselen
);
1699 * Loads the exclude lists for the directory containing pathname, then
1700 * scans all exclude lists to determine whether pathname is excluded.
1701 * Returns the exclude_list element which matched, or NULL for
1704 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1705 struct index_state
*istate
,
1706 const char *pathname
,
1709 int pathlen
= strlen(pathname
);
1710 const char *basename
= strrchr(pathname
, '/');
1711 basename
= (basename
) ? basename
+1 : pathname
;
1713 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1715 if (dir
->internal
.pattern
)
1716 return dir
->internal
.pattern
;
1718 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1723 * Loads the exclude lists for the directory containing pathname, then
1724 * scans all exclude lists to determine whether pathname is excluded.
1725 * Returns 1 if true, otherwise 0.
1727 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1728 const char *pathname
, int *dtype_p
)
1730 struct path_pattern
*pattern
=
1731 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1733 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1737 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1739 struct dir_entry
*ent
;
1741 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1746 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1747 struct index_state
*istate
,
1748 const char *pathname
, int len
)
1750 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1753 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->internal
.alloc
);
1754 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1757 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1758 struct index_state
*istate
,
1759 const char *pathname
, int len
)
1761 if (!index_name_is_other(istate
, pathname
, len
))
1764 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->internal
.ignored_alloc
);
1765 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1769 index_nonexistent
= 0,
1775 * Do not use the alphabetically sorted index to look up
1776 * the directory name; instead, use the case insensitive
1779 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1780 const char *dirname
, int len
)
1782 struct cache_entry
*ce
;
1784 if (index_dir_exists(istate
, dirname
, len
))
1785 return index_directory
;
1787 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1788 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1789 return index_gitdir
;
1791 return index_nonexistent
;
1795 * The index sorts alphabetically by entry name, which
1796 * means that a gitlink sorts as '\0' at the end, while
1797 * a directory (which is defined not as an entry, but as
1798 * the files it contains) will sort with the '/' at the
1801 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1802 const char *dirname
, int len
)
1807 return directory_exists_in_index_icase(istate
, dirname
, len
);
1809 pos
= index_name_pos(istate
, dirname
, len
);
1812 while (pos
< istate
->cache_nr
) {
1813 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1814 unsigned char endchar
;
1816 if (strncmp(ce
->name
, dirname
, len
))
1818 endchar
= ce
->name
[len
];
1822 return index_directory
;
1823 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1824 return index_gitdir
;
1826 return index_nonexistent
;
1830 * When we find a directory when traversing the filesystem, we
1831 * have three distinct cases:
1834 * - see it as a directory
1837 * and which one we choose depends on a combination of existing
1838 * git index contents and the flags passed into the directory
1839 * traversal routine.
1841 * Case 1: If we *already* have entries in the index under that
1842 * directory name, we always recurse into the directory to see
1845 * Case 2: If we *already* have that directory name as a gitlink,
1846 * we always continue to see it as a gitlink, regardless of whether
1847 * there is an actual git directory there or not (it might not
1848 * be checked out as a subproject!)
1850 * Case 3: if we didn't have it in the index previously, we
1851 * have a few sub-cases:
1853 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1854 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1855 * also true, in which case we need to check if it contains any
1856 * untracked and / or ignored files.
1857 * (b) if it looks like a git directory and we don't have the
1858 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1859 * show it as a directory.
1860 * (c) otherwise, we recurse into it.
1862 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1863 struct index_state
*istate
,
1864 struct untracked_cache_dir
*untracked
,
1865 const char *dirname
, int len
, int baselen
, int excluded
,
1866 const struct pathspec
*pathspec
)
1869 * WARNING: From this function, you can return path_recurse or you
1870 * can call read_directory_recursive() (or neither), but
1871 * you CAN'T DO BOTH.
1873 enum path_treatment state
;
1874 int matches_how
= 0;
1875 int check_only
, stop_early
;
1876 int old_ignored_nr
, old_untracked_nr
;
1877 /* The "len-1" is to strip the final '/' */
1878 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1880 if (status
== index_directory
)
1881 return path_recurse
;
1882 if (status
== index_gitdir
)
1884 if (status
!= index_nonexistent
)
1885 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1888 * We don't want to descend into paths that don't match the necessary
1889 * patterns. Clearly, if we don't have a pathspec, then we can't check
1890 * for matching patterns. Also, if (excluded) then we know we matched
1891 * the exclusion patterns so as an optimization we can skip checking
1892 * for matching patterns.
1894 if (pathspec
&& !excluded
) {
1895 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1899 DO_MATCH_LEADING_PATHSPEC
);
1905 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1906 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1908 * Determine if `dirname` is a nested repo by confirming that:
1909 * 1) we are in a nonbare repository, and
1910 * 2) `dirname` is not an immediate parent of `the_repository->gitdir`,
1911 * which could occur if the git_dir or worktree location was
1912 * manually configured by the user; see t2205 testcases 1-3 for
1913 * examples where this matters
1916 struct strbuf sb
= STRBUF_INIT
;
1917 strbuf_addstr(&sb
, dirname
);
1918 nested_repo
= is_nonbare_repository_dir(&sb
);
1921 char *real_dirname
, *real_gitdir
;
1922 strbuf_addstr(&sb
, ".git");
1923 real_dirname
= real_pathdup(sb
.buf
, 1);
1924 real_gitdir
= real_pathdup(the_repository
->gitdir
, 1);
1926 nested_repo
= !!strcmp(real_dirname
, real_gitdir
);
1930 strbuf_release(&sb
);
1933 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1934 (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
))
1936 return excluded
? path_excluded
: path_untracked
;
1940 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1942 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1943 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1946 * This is an excluded directory and we are
1947 * showing ignored paths that match an exclude
1948 * pattern. (e.g. show directory as ignored
1949 * only if it matches an exclude pattern).
1950 * This path will either be 'path_excluded`
1951 * (if we are showing empty directories or if
1952 * the directory is not empty), or will be
1953 * 'path_none' (empty directory, and we are
1954 * not showing empty directories).
1956 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1957 return path_excluded
;
1959 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1960 untracked
, 1, 1, pathspec
) == path_excluded
)
1961 return path_excluded
;
1965 return path_recurse
;
1968 assert(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
);
1971 * If we have a pathspec which could match something _below_ this
1972 * directory (e.g. when checking 'subdir/' having a pathspec like
1973 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1976 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1977 return path_recurse
;
1979 /* Special cases for where this directory is excluded/ignored */
1982 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1983 * hiding empty directories, there is no need to
1984 * recurse into an ignored directory.
1986 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1987 return path_excluded
;
1990 * Even if we are hiding empty directories, we can still avoid
1991 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
1992 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1994 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1995 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
1996 return path_excluded
;
2000 * Other than the path_recurse case above, we only need to
2001 * recurse into untracked directories if any of the following
2003 * - DIR_SHOW_IGNORED (because then we need to determine if
2004 * there are ignored entries below)
2005 * - DIR_SHOW_IGNORED_TOO (same as above)
2006 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
2007 * the directory is empty)
2010 !(dir
->flags
& (DIR_SHOW_IGNORED
|
2011 DIR_SHOW_IGNORED_TOO
|
2012 DIR_HIDE_EMPTY_DIRECTORIES
))) {
2013 return path_untracked
;
2017 * Even if we don't want to know all the paths under an untracked or
2018 * ignored directory, we may still need to go into the directory to
2019 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2020 * an empty directory should be path_none instead of path_excluded or
2023 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
2024 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
2027 * However, there's another optimization possible as a subset of
2028 * check_only, based on the cases we have to consider:
2029 * A) Directory matches no exclude patterns:
2030 * * Directory is empty => path_none
2031 * * Directory has an untracked file under it => path_untracked
2032 * * Directory has only ignored files under it => path_excluded
2033 * B) Directory matches an exclude pattern:
2034 * * Directory is empty => path_none
2035 * * Directory has an untracked file under it => path_excluded
2036 * * Directory has only ignored files under it => path_excluded
2037 * In case A, we can exit as soon as we've found an untracked
2038 * file but otherwise have to walk all files. In case B, though,
2039 * we can stop at the first file we find under the directory.
2041 stop_early
= check_only
&& excluded
;
2044 * If /every/ file within an untracked directory is ignored, then
2045 * we want to treat the directory as ignored (for e.g. status
2046 * --porcelain), without listing the individual ignored files
2047 * underneath. To do so, we'll save the current ignored_nr, and
2048 * pop all the ones added after it if it turns out the entire
2049 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2050 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2051 * untracked paths so will need to pop all those off the last
2052 * after we traverse.
2054 old_ignored_nr
= dir
->ignored_nr
;
2055 old_untracked_nr
= dir
->nr
;
2057 /* Actually recurse into dirname now, we'll fixup the state later. */
2058 untracked
= lookup_untracked(dir
->untracked
, untracked
,
2059 dirname
+ baselen
, len
- baselen
);
2060 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
2061 check_only
, stop_early
, pathspec
);
2063 /* There are a variety of reasons we may need to fixup the state... */
2064 if (state
== path_excluded
) {
2065 /* state == path_excluded implies all paths under
2066 * dirname were ignored...
2068 * if running e.g. `git status --porcelain --ignored=matching`,
2069 * then we want to see the subpaths that are ignored.
2071 * if running e.g. just `git status --porcelain`, then
2072 * we just want the directory itself to be listed as ignored
2073 * and not the individual paths underneath.
2075 int want_ignored_subpaths
=
2076 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2077 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
2079 if (want_ignored_subpaths
) {
2081 * with --ignored=matching, we want the subpaths
2082 * INSTEAD of the directory itself.
2087 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
2088 FREE_AND_NULL(dir
->ignored
[i
]);
2089 dir
->ignored_nr
= old_ignored_nr
;
2094 * We may need to ignore some of the untracked paths we found while
2095 * traversing subdirectories.
2097 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2098 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
2100 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
2101 FREE_AND_NULL(dir
->entries
[i
]);
2102 dir
->nr
= old_untracked_nr
;
2106 * If there is nothing under the current directory and we are not
2107 * hiding empty directories, then we need to report on the
2108 * untracked or ignored status of the directory itself.
2110 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2111 state
= excluded
? path_excluded
: path_untracked
;
2117 * This is an inexact early pruning of any recursive directory
2118 * reading - if the path cannot possibly be in the pathspec,
2119 * return true, and we'll skip it early.
2121 static int simplify_away(const char *path
, int pathlen
,
2122 const struct pathspec
*pathspec
)
2126 if (!pathspec
|| !pathspec
->nr
)
2129 GUARD_PATHSPEC(pathspec
,
2138 for (i
= 0; i
< pathspec
->nr
; i
++) {
2139 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2140 int len
= item
->nowildcard_len
;
2144 if (!ps_strncmp(item
, item
->match
, path
, len
))
2152 * This function tells us whether an excluded path matches a
2153 * list of "interesting" pathspecs. That is, whether a path matched
2154 * by any of the pathspecs could possibly be ignored by excluding
2155 * the specified path. This can happen if:
2157 * 1. the path is mentioned explicitly in the pathspec
2159 * 2. the path is a directory prefix of some element in the
2162 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2163 const struct pathspec
*pathspec
)
2167 if (!pathspec
|| !pathspec
->nr
)
2170 GUARD_PATHSPEC(pathspec
,
2178 for (i
= 0; i
< pathspec
->nr
; i
++) {
2179 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2180 int len
= item
->nowildcard_len
;
2182 if (len
== pathlen
&&
2183 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2185 if (len
> pathlen
&&
2186 item
->match
[pathlen
] == '/' &&
2187 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2193 static int get_index_dtype(struct index_state
*istate
,
2194 const char *path
, int len
)
2197 const struct cache_entry
*ce
;
2199 ce
= index_file_exists(istate
, path
, len
, 0);
2201 if (!ce_uptodate(ce
))
2203 if (S_ISGITLINK(ce
->ce_mode
))
2206 * Nobody actually cares about the
2207 * difference between DT_LNK and DT_REG
2212 /* Try to look it up as a directory */
2213 pos
= index_name_pos(istate
, path
, len
);
2217 while (pos
< istate
->cache_nr
) {
2218 ce
= istate
->cache
[pos
++];
2219 if (strncmp(ce
->name
, path
, len
))
2221 if (ce
->name
[len
] > '/')
2223 if (ce
->name
[len
] < '/')
2225 if (!ce_uptodate(ce
))
2226 break; /* continue? */
2232 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2233 const char *path
, int len
)
2237 if (dtype
!= DT_UNKNOWN
)
2239 dtype
= get_index_dtype(istate
, path
, len
);
2240 if (dtype
!= DT_UNKNOWN
)
2242 if (lstat(path
, &st
))
2244 if (S_ISREG(st
.st_mode
))
2246 if (S_ISDIR(st
.st_mode
))
2248 if (S_ISLNK(st
.st_mode
))
2253 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2254 struct cached_dir
*cdir
,
2255 struct index_state
*istate
,
2256 struct strbuf
*path
,
2258 const struct pathspec
*pathspec
)
2261 * WARNING: From this function, you can return path_recurse or you
2262 * can call read_directory_recursive() (or neither), but
2263 * you CAN'T DO BOTH.
2265 strbuf_setlen(path
, baselen
);
2267 strbuf_addstr(path
, cdir
->file
);
2268 return path_untracked
;
2270 strbuf_addstr(path
, cdir
->ucd
->name
);
2271 /* treat_one_path() does this before it calls treat_directory() */
2272 strbuf_complete(path
, '/');
2273 if (cdir
->ucd
->check_only
)
2275 * check_only is set as a result of treat_directory() getting
2276 * to its bottom. Verify again the same set of directories
2277 * with check_only set.
2279 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2280 cdir
->ucd
, 1, 0, pathspec
);
2282 * We get path_recurse in the first run when
2283 * directory_exists_in_index() returns index_nonexistent. We
2284 * are sure that new changes in the index does not impact the
2285 * outcome. Return now.
2287 return path_recurse
;
2290 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2291 struct untracked_cache_dir
*untracked
,
2292 struct cached_dir
*cdir
,
2293 struct index_state
*istate
,
2294 struct strbuf
*path
,
2296 const struct pathspec
*pathspec
)
2298 int has_path_in_index
, dtype
, excluded
;
2301 return treat_path_fast(dir
, cdir
, istate
, path
,
2303 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2305 strbuf_setlen(path
, baselen
);
2306 strbuf_addstr(path
, cdir
->d_name
);
2307 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2310 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2312 /* Always exclude indexed files */
2313 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2315 if (dtype
!= DT_DIR
&& has_path_in_index
)
2319 * When we are looking at a directory P in the working tree,
2320 * there are three cases:
2322 * (1) P exists in the index. Everything inside the directory P in
2323 * the working tree needs to go when P is checked out from the
2326 * (2) P does not exist in the index, but there is P/Q in the index.
2327 * We know P will stay a directory when we check out the contents
2328 * of the index, but we do not know yet if there is a directory
2329 * P/Q in the working tree to be killed, so we need to recurse.
2331 * (3) P does not exist in the index, and there is no P/Q in the index
2332 * to require P to be a directory, either. Only in this case, we
2333 * know that everything inside P will not be killed without
2336 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2337 (dtype
== DT_DIR
) &&
2338 !has_path_in_index
&&
2339 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2342 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2345 * Excluded? If we don't explicitly want to show
2346 * ignored files, ignore it
2348 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2349 return path_excluded
;
2356 * WARNING: Do not ignore/amend the return value from
2357 * treat_directory(), and especially do not change it to return
2358 * path_recurse as that can cause exponential slowdown.
2359 * Instead, modify treat_directory() to return the right value.
2361 strbuf_addch(path
, '/');
2362 return treat_directory(dir
, istate
, untracked
,
2363 path
->buf
, path
->len
,
2364 baselen
, excluded
, pathspec
);
2368 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2369 0 /* prefix */, NULL
/* seen */,
2373 return path_excluded
;
2374 return path_untracked
;
2378 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2382 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2383 dir
->untracked_alloc
);
2384 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2387 static int valid_cached_dir(struct dir_struct
*dir
,
2388 struct untracked_cache_dir
*untracked
,
2389 struct index_state
*istate
,
2390 struct strbuf
*path
,
2399 * With fsmonitor, we can trust the untracked cache's valid field.
2401 refresh_fsmonitor(istate
);
2402 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2403 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2404 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2407 if (!untracked
->valid
||
2408 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2409 fill_stat_data(&untracked
->stat_data
, &st
);
2414 if (untracked
->check_only
!= !!check_only
)
2418 * prep_exclude will be called eventually on this directory,
2419 * but it's called much later in last_matching_pattern(). We
2420 * need it now to determine the validity of the cache for this
2421 * path. The next calls will be nearly no-op, the way
2422 * prep_exclude() is designed.
2424 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2425 strbuf_addch(path
, '/');
2426 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2427 strbuf_setlen(path
, path
->len
- 1);
2429 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2431 /* hopefully prep_exclude() haven't invalidated this entry... */
2432 return untracked
->valid
;
2435 static int open_cached_dir(struct cached_dir
*cdir
,
2436 struct dir_struct
*dir
,
2437 struct untracked_cache_dir
*untracked
,
2438 struct index_state
*istate
,
2439 struct strbuf
*path
,
2444 memset(cdir
, 0, sizeof(*cdir
));
2445 cdir
->untracked
= untracked
;
2446 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2448 c_path
= path
->len
? path
->buf
: ".";
2449 cdir
->fdir
= opendir(c_path
);
2451 warning_errno(_("could not open directory '%s'"), c_path
);
2452 if (dir
->untracked
) {
2453 invalidate_directory(dir
->untracked
, untracked
);
2454 dir
->untracked
->dir_opened
++;
2461 static int read_cached_dir(struct cached_dir
*cdir
)
2466 de
= readdir_skip_dot_and_dotdot(cdir
->fdir
);
2468 cdir
->d_name
= NULL
;
2469 cdir
->d_type
= DT_UNKNOWN
;
2472 cdir
->d_name
= de
->d_name
;
2473 cdir
->d_type
= DTYPE(de
);
2476 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2477 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2487 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2488 struct untracked_cache_dir
*d
= cdir
->untracked
;
2489 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2495 static void close_cached_dir(struct cached_dir
*cdir
)
2498 closedir(cdir
->fdir
);
2500 * We have gone through this directory and found no untracked
2501 * entries. Mark it valid.
2503 if (cdir
->untracked
) {
2504 cdir
->untracked
->valid
= 1;
2505 cdir
->untracked
->recurse
= 1;
2509 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2510 struct untracked_cache_dir
*untracked
,
2511 struct cached_dir
*cdir
,
2512 struct index_state
*istate
,
2513 struct strbuf
*path
,
2515 const struct pathspec
*pathspec
,
2516 enum path_treatment state
)
2518 /* add the path to the appropriate result list */
2521 if (dir
->flags
& DIR_SHOW_IGNORED
)
2522 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2523 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2524 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2525 exclude_matches_pathspec(path
->buf
, path
->len
,
2527 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2530 case path_untracked
:
2531 if (dir
->flags
& DIR_SHOW_IGNORED
)
2533 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2535 add_untracked(untracked
, path
->buf
+ baselen
);
2544 * Read a directory tree. We currently ignore anything but
2545 * directories, regular files and symlinks. That's because git
2546 * doesn't handle them at all yet. Maybe that will change some
2549 * Also, we ignore the name ".git" (even if it is not a directory).
2550 * That likely will not change.
2552 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2553 * to signal that a file was found. This is the least significant value that
2554 * indicates that a file was encountered that does not depend on the order of
2555 * whether an untracked or excluded path was encountered first.
2557 * Returns the most significant path_treatment value encountered in the scan.
2558 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2559 * significant path_treatment value that will be returned.
2562 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2563 struct index_state
*istate
, const char *base
, int baselen
,
2564 struct untracked_cache_dir
*untracked
, int check_only
,
2565 int stop_at_first_file
, const struct pathspec
*pathspec
)
2568 * WARNING: Do NOT recurse unless path_recurse is returned from
2569 * treat_path(). Recursing on any other return value
2570 * can result in exponential slowdown.
2572 struct cached_dir cdir
;
2573 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2574 struct strbuf path
= STRBUF_INIT
;
2576 strbuf_add(&path
, base
, baselen
);
2578 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2580 dir
->internal
.visited_directories
++;
2583 untracked
->check_only
= !!check_only
;
2585 while (!read_cached_dir(&cdir
)) {
2586 /* check how the file or directory should be treated */
2587 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2589 dir
->internal
.visited_paths
++;
2591 if (state
> dir_state
)
2594 /* recurse into subdir if instructed by treat_path */
2595 if (state
== path_recurse
) {
2596 struct untracked_cache_dir
*ud
;
2597 ud
= lookup_untracked(dir
->untracked
,
2600 path
.len
- baselen
);
2602 read_directory_recursive(dir
, istate
, path
.buf
,
2604 check_only
, stop_at_first_file
, pathspec
);
2605 if (subdir_state
> dir_state
)
2606 dir_state
= subdir_state
;
2609 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2610 0 /* prefix */, NULL
,
2611 0 /* do NOT special case dirs */))
2616 if (stop_at_first_file
) {
2618 * If stopping at first file, then
2619 * signal that a file was found by
2620 * returning `path_excluded`. This is
2621 * to return a consistent value
2622 * regardless of whether an ignored or
2623 * excluded file happened to be
2626 * In current usage, the
2627 * `stop_at_first_file` is passed when
2628 * an ancestor directory has matched
2629 * an exclude pattern, so any found
2630 * files will be excluded.
2632 if (dir_state
>= path_excluded
) {
2633 dir_state
= path_excluded
;
2638 /* abort early if maximum state has been reached */
2639 if (dir_state
== path_untracked
) {
2641 add_untracked(untracked
, path
.buf
+ baselen
);
2644 /* skip the add_path_to_appropriate_result_list() */
2648 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2649 istate
, &path
, baselen
,
2652 close_cached_dir(&cdir
);
2654 strbuf_release(&path
);
2659 int cmp_dir_entry(const void *p1
, const void *p2
)
2661 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2662 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2664 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2667 /* check if *out lexically strictly contains *in */
2668 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2670 return (out
->len
< in
->len
) &&
2671 (out
->name
[out
->len
- 1] == '/') &&
2672 !memcmp(out
->name
, in
->name
, out
->len
);
2675 static int treat_leading_path(struct dir_struct
*dir
,
2676 struct index_state
*istate
,
2677 const char *path
, int len
,
2678 const struct pathspec
*pathspec
)
2680 struct strbuf sb
= STRBUF_INIT
;
2681 struct strbuf subdir
= STRBUF_INIT
;
2682 int prevlen
, baselen
;
2684 struct cached_dir cdir
;
2685 enum path_treatment state
= path_none
;
2688 * For each directory component of path, we are going to check whether
2689 * that path is relevant given the pathspec. For example, if path is
2691 * then we will ask treat_path() whether we should go into foo, then
2692 * whether we should go into bar, then whether baz is relevant.
2693 * Checking each is important because e.g. if path is
2695 * then we need to check .git to know we shouldn't traverse it.
2696 * If the return from treat_path() is:
2697 * * path_none, for any path, we return false.
2698 * * path_recurse, for all path components, we return true
2699 * * <anything else> for some intermediate component, we make sure
2700 * to add that path to the relevant list but return false
2701 * signifying that we shouldn't recurse into it.
2704 while (len
&& path
[len
- 1] == '/')
2709 memset(&cdir
, 0, sizeof(cdir
));
2710 cdir
.d_type
= DT_DIR
;
2714 prevlen
= baselen
+ !!baselen
;
2715 cp
= path
+ prevlen
;
2716 cp
= memchr(cp
, '/', path
+ len
- cp
);
2720 baselen
= cp
- path
;
2722 strbuf_add(&sb
, path
, baselen
);
2723 if (!is_directory(sb
.buf
))
2726 strbuf_add(&sb
, path
, prevlen
);
2727 strbuf_reset(&subdir
);
2728 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2729 cdir
.d_name
= subdir
.buf
;
2730 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2732 if (state
!= path_recurse
)
2733 break; /* do not recurse into it */
2735 break; /* finished checking */
2737 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2738 &sb
, baselen
, pathspec
,
2741 strbuf_release(&subdir
);
2742 strbuf_release(&sb
);
2743 return state
== path_recurse
;
2746 static const char *get_ident_string(void)
2748 static struct strbuf sb
= STRBUF_INIT
;
2753 if (uname(&uts
) < 0)
2754 die_errno(_("failed to get kernel name and information"));
2755 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2760 static int ident_in_untracked(const struct untracked_cache
*uc
)
2763 * Previous git versions may have saved many NUL separated
2764 * strings in the "ident" field, but it is insane to manage
2765 * many locations, so just take care of the first one.
2768 return !strcmp(uc
->ident
.buf
, get_ident_string());
2771 static void set_untracked_ident(struct untracked_cache
*uc
)
2773 strbuf_reset(&uc
->ident
);
2774 strbuf_addstr(&uc
->ident
, get_ident_string());
2777 * This strbuf used to contain a list of NUL separated
2778 * strings, so save NUL too for backward compatibility.
2780 strbuf_addch(&uc
->ident
, 0);
2783 static unsigned new_untracked_cache_flags(struct index_state
*istate
)
2785 struct repository
*repo
= istate
->repo
;
2789 * This logic is coordinated with the setting of these flags in
2790 * wt-status.c#wt_status_collect_untracked(), and the evaluation
2791 * of the config setting in commit.c#git_status_config()
2793 if (!repo_config_get_string(repo
, "status.showuntrackedfiles", &val
) &&
2794 !strcmp(val
, "all"))
2798 * The default, if "all" is not set, is "normal" - leading us here.
2799 * If the value is "none" then it really doesn't matter.
2801 return DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2804 static void new_untracked_cache(struct index_state
*istate
, int flags
)
2806 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2807 strbuf_init(&uc
->ident
, 100);
2808 uc
->exclude_per_dir
= ".gitignore";
2809 uc
->dir_flags
= flags
>= 0 ? flags
: new_untracked_cache_flags(istate
);
2810 set_untracked_ident(uc
);
2811 istate
->untracked
= uc
;
2812 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2815 void add_untracked_cache(struct index_state
*istate
)
2817 if (!istate
->untracked
) {
2818 new_untracked_cache(istate
, -1);
2820 if (!ident_in_untracked(istate
->untracked
)) {
2821 free_untracked_cache(istate
->untracked
);
2822 new_untracked_cache(istate
, -1);
2827 void remove_untracked_cache(struct index_state
*istate
)
2829 if (istate
->untracked
) {
2830 free_untracked_cache(istate
->untracked
);
2831 istate
->untracked
= NULL
;
2832 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2836 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2838 const struct pathspec
*pathspec
,
2839 struct index_state
*istate
)
2841 struct untracked_cache_dir
*root
;
2842 static int untracked_cache_disabled
= -1;
2844 if (!dir
->untracked
)
2846 if (untracked_cache_disabled
< 0)
2847 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2848 if (untracked_cache_disabled
)
2852 * We only support $GIT_DIR/info/exclude and core.excludesfile
2853 * as the global ignore rule files. Any other additions
2854 * (e.g. from command line) invalidate the cache. This
2855 * condition also catches running setup_standard_excludes()
2856 * before setting dir->untracked!
2858 if (dir
->internal
.unmanaged_exclude_files
)
2862 * Optimize for the main use case only: whole-tree git
2863 * status. More work involved in treat_leading_path() if we
2864 * use cache on just a subset of the worktree. pathspec
2865 * support could make the matter even worse.
2867 if (base_len
|| (pathspec
&& pathspec
->nr
))
2870 /* We don't support collecting ignore files */
2871 if (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2872 DIR_COLLECT_IGNORED
))
2876 * If we use .gitignore in the cache and now you change it to
2877 * .gitexclude, everything will go wrong.
2879 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2880 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2884 * EXC_CMDL is not considered in the cache. If people set it,
2887 if (dir
->internal
.exclude_list_group
[EXC_CMDL
].nr
)
2890 if (!ident_in_untracked(dir
->untracked
)) {
2891 warning(_("untracked cache is disabled on this system or location"));
2896 * If the untracked structure we received does not have the same flags
2897 * as requested in this run, we're going to need to either discard the
2898 * existing structure (and potentially later recreate), or bypass the
2899 * untracked cache mechanism for this run.
2901 if (dir
->flags
!= dir
->untracked
->dir_flags
) {
2903 * If the untracked structure we received does not have the same flags
2904 * as configured, then we need to reset / create a new "untracked"
2905 * structure to match the new config.
2907 * Keeping the saved and used untracked cache consistent with the
2908 * configuration provides an opportunity for frequent users of
2909 * "git status -uall" to leverage the untracked cache by aligning their
2910 * configuration - setting "status.showuntrackedfiles" to "all" or
2911 * "normal" as appropriate.
2913 * Previously using -uall (or setting "status.showuntrackedfiles" to
2914 * "all") was incompatible with untracked cache and *consistently*
2915 * caused surprisingly bad performance (with fscache and fsmonitor
2916 * enabled) on Windows.
2918 * IMPROVEMENT OPPORTUNITY: If we reworked the untracked cache storage
2919 * to not be as bound up with the desired output in a given run,
2920 * and instead iterated through and stored enough information to
2921 * correctly serve both "modes", then users could get peak performance
2922 * with or without '-uall' regardless of their
2923 * "status.showuntrackedfiles" config.
2925 if (dir
->untracked
->dir_flags
!= new_untracked_cache_flags(istate
)) {
2926 free_untracked_cache(istate
->untracked
);
2927 new_untracked_cache(istate
, dir
->flags
);
2928 dir
->untracked
= istate
->untracked
;
2932 * Current untracked cache data is consistent with config, but not
2933 * usable in this request/run; just bypass untracked cache.
2939 if (!dir
->untracked
->root
) {
2940 /* Untracked cache existed but is not initialized; fix that */
2941 FLEX_ALLOC_STR(dir
->untracked
->root
, name
, "");
2942 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2945 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2946 root
= dir
->untracked
->root
;
2947 if (!oideq(&dir
->internal
.ss_info_exclude
.oid
,
2948 &dir
->untracked
->ss_info_exclude
.oid
)) {
2949 invalidate_gitignore(dir
->untracked
, root
);
2950 dir
->untracked
->ss_info_exclude
= dir
->internal
.ss_info_exclude
;
2952 if (!oideq(&dir
->internal
.ss_excludes_file
.oid
,
2953 &dir
->untracked
->ss_excludes_file
.oid
)) {
2954 invalidate_gitignore(dir
->untracked
, root
);
2955 dir
->untracked
->ss_excludes_file
= dir
->internal
.ss_excludes_file
;
2958 /* Make sure this directory is not dropped out at saving phase */
2963 static void emit_traversal_statistics(struct dir_struct
*dir
,
2964 struct repository
*repo
,
2968 if (!trace2_is_enabled())
2972 trace2_data_string("read_directory", repo
, "path", "");
2974 struct strbuf tmp
= STRBUF_INIT
;
2975 strbuf_add(&tmp
, path
, path_len
);
2976 trace2_data_string("read_directory", repo
, "path", tmp
.buf
);
2977 strbuf_release(&tmp
);
2980 trace2_data_intmax("read_directory", repo
,
2981 "directories-visited", dir
->internal
.visited_directories
);
2982 trace2_data_intmax("read_directory", repo
,
2983 "paths-visited", dir
->internal
.visited_paths
);
2985 if (!dir
->untracked
)
2987 trace2_data_intmax("read_directory", repo
,
2988 "node-creation", dir
->untracked
->dir_created
);
2989 trace2_data_intmax("read_directory", repo
,
2990 "gitignore-invalidation",
2991 dir
->untracked
->gitignore_invalidated
);
2992 trace2_data_intmax("read_directory", repo
,
2993 "directory-invalidation",
2994 dir
->untracked
->dir_invalidated
);
2995 trace2_data_intmax("read_directory", repo
,
2996 "opendir", dir
->untracked
->dir_opened
);
2999 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
3000 const char *path
, int len
, const struct pathspec
*pathspec
)
3002 struct untracked_cache_dir
*untracked
;
3004 trace2_region_enter("dir", "read_directory", istate
->repo
);
3005 dir
->internal
.visited_paths
= 0;
3006 dir
->internal
.visited_directories
= 0;
3008 if (has_symlink_leading_path(path
, len
)) {
3009 trace2_region_leave("dir", "read_directory", istate
->repo
);
3013 untracked
= validate_untracked_cache(dir
, len
, pathspec
, istate
);
3016 * make sure untracked cache code path is disabled,
3017 * e.g. prep_exclude()
3019 dir
->untracked
= NULL
;
3020 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
3021 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
3022 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
3023 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
3025 emit_traversal_statistics(dir
, istate
->repo
, path
, len
);
3027 trace2_region_leave("dir", "read_directory", istate
->repo
);
3028 if (dir
->untracked
) {
3029 static int force_untracked_cache
= -1;
3031 if (force_untracked_cache
< 0)
3032 force_untracked_cache
=
3033 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", -1);
3034 if (force_untracked_cache
< 0)
3035 force_untracked_cache
= (istate
->repo
->settings
.core_untracked_cache
== UNTRACKED_CACHE_WRITE
);
3036 if (force_untracked_cache
&&
3037 dir
->untracked
== istate
->untracked
&&
3038 (dir
->untracked
->dir_opened
||
3039 dir
->untracked
->gitignore_invalidated
||
3040 dir
->untracked
->dir_invalidated
))
3041 istate
->cache_changed
|= UNTRACKED_CHANGED
;
3042 if (dir
->untracked
!= istate
->untracked
) {
3043 FREE_AND_NULL(dir
->untracked
);
3050 int file_exists(const char *f
)
3053 return lstat(f
, &sb
) == 0;
3056 int repo_file_exists(struct repository
*repo
, const char *path
)
3058 if (repo
!= the_repository
)
3059 BUG("do not know how to check file existence in arbitrary repo");
3061 return file_exists(path
);
3064 static int cmp_icase(char a
, char b
)
3069 return toupper(a
) - toupper(b
);
3074 * Given two normalized paths (a trailing slash is ok), if subdir is
3075 * outside dir, return -1. Otherwise return the offset in subdir that
3076 * can be used as relative path to dir.
3078 int dir_inside_of(const char *subdir
, const char *dir
)
3082 assert(dir
&& subdir
&& *dir
&& *subdir
);
3084 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
3090 /* hel[p]/me vs hel[l]/yeah */
3091 if (*dir
&& *subdir
)
3095 return !*dir
? offset
: -1; /* same dir */
3097 /* foo/[b]ar vs foo/[] */
3098 if (is_dir_sep(dir
[-1]))
3099 return is_dir_sep(subdir
[-1]) ? offset
: -1;
3101 /* foo[/]bar vs foo[] */
3102 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
3105 int is_inside_dir(const char *dir
)
3114 rc
= (dir_inside_of(cwd
, dir
) >= 0);
3119 int is_empty_dir(const char *path
)
3121 DIR *dir
= opendir(path
);
3128 e
= readdir_skip_dot_and_dotdot(dir
);
3136 char *git_url_basename(const char *repo
, int is_bundle
, int is_bare
)
3138 const char *end
= repo
+ strlen(repo
), *start
, *ptr
;
3145 start
= strstr(repo
, "://");
3152 * Skip authentication data. The stripping does happen
3153 * greedily, such that we strip up to the last '@' inside
3156 for (ptr
= start
; ptr
< end
&& !is_dir_sep(*ptr
); ptr
++) {
3162 * Strip trailing spaces, slashes and /.git
3164 while (start
< end
&& (is_dir_sep(end
[-1]) || isspace(end
[-1])))
3166 if (end
- start
> 5 && is_dir_sep(end
[-5]) &&
3167 !strncmp(end
- 4, ".git", 4)) {
3169 while (start
< end
&& is_dir_sep(end
[-1]))
3174 * It should not be possible to overflow `ptrdiff_t` by passing in an
3175 * insanely long URL, but GCC does not know that and will complain
3176 * without this check.
3178 if (end
- start
< 0)
3179 die(_("No directory name could be guessed.\n"
3180 "Please specify a directory on the command line"));
3183 * Strip trailing port number if we've got only a
3184 * hostname (that is, there is no dir separator but a
3185 * colon). This check is required such that we do not
3186 * strip URI's like '/foo/bar:2222.git', which should
3187 * result in a dir '2222' being guessed due to backwards
3190 if (memchr(start
, '/', end
- start
) == NULL
3191 && memchr(start
, ':', end
- start
) != NULL
) {
3193 while (start
< ptr
&& isdigit(ptr
[-1]) && ptr
[-1] != ':')
3195 if (start
< ptr
&& ptr
[-1] == ':')
3200 * Find last component. To remain backwards compatible we
3201 * also regard colons as path separators, such that
3202 * cloning a repository 'foo:bar.git' would result in a
3203 * directory 'bar' being guessed.
3206 while (start
< ptr
&& !is_dir_sep(ptr
[-1]) && ptr
[-1] != ':')
3211 * Strip .{bundle,git}.
3214 strip_suffix_mem(start
, &len
, is_bundle
? ".bundle" : ".git");
3216 if (!len
|| (len
== 1 && *start
== '/'))
3217 die(_("No directory name could be guessed.\n"
3218 "Please specify a directory on the command line"));
3221 dir
= xstrfmt("%.*s.git", (int)len
, start
);
3223 dir
= xstrndup(start
, len
);
3225 * Replace sequences of 'control' characters and whitespace
3226 * with one ascii space, remove leading and trailing spaces.
3230 int prev_space
= 1 /* strip leading whitespace */;
3231 for (end
= dir
; *end
; ++end
) {
3233 if ((unsigned char)ch
< '\x20')
3244 if (out
> dir
&& prev_space
)
3250 void strip_dir_trailing_slashes(char *dir
)
3252 char *end
= dir
+ strlen(dir
);
3254 while (dir
< end
- 1 && is_dir_sep(end
[-1]))
3259 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
3263 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
3264 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
3265 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
3266 int purge_original_cwd
= (flag
& REMOVE_DIR_PURGE_ORIGINAL_CWD
);
3267 struct object_id submodule_head
;
3269 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
3270 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
3271 /* Do not descend and nuke a nested git work tree. */
3277 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
3278 dir
= opendir(path
->buf
);
3280 if (errno
== ENOENT
)
3281 return keep_toplevel
? -1 : 0;
3282 else if (errno
== EACCES
&& !keep_toplevel
)
3284 * An empty dir could be removable even if it
3287 return rmdir(path
->buf
);
3291 strbuf_complete(path
, '/');
3294 while ((e
= readdir_skip_dot_and_dotdot(dir
)) != NULL
) {
3297 strbuf_setlen(path
, len
);
3298 strbuf_addstr(path
, e
->d_name
);
3299 if (lstat(path
->buf
, &st
)) {
3300 if (errno
== ENOENT
)
3302 * file disappeared, which is what we
3307 } else if (S_ISDIR(st
.st_mode
)) {
3308 if (!remove_dir_recurse(path
, flag
, &kept_down
))
3309 continue; /* happy */
3310 } else if (!only_empty
&&
3311 (!unlink(path
->buf
) || errno
== ENOENT
)) {
3312 continue; /* happy, too */
3315 /* path too long, stat fails, or non-directory still exists */
3321 strbuf_setlen(path
, original_len
);
3322 if (!ret
&& !keep_toplevel
&& !kept_down
) {
3323 if (!purge_original_cwd
&&
3324 startup_info
->original_cwd
&&
3325 !strcmp(startup_info
->original_cwd
, path
->buf
))
3326 ret
= -1; /* Do not remove current working directory */
3328 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
3331 * report the uplevel that it is not an error that we
3332 * did not rmdir() our directory.
3338 int remove_dir_recursively(struct strbuf
*path
, int flag
)
3340 return remove_dir_recurse(path
, flag
, NULL
);
3343 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
3345 void setup_standard_excludes(struct dir_struct
*dir
)
3347 dir
->exclude_per_dir
= ".gitignore";
3349 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3351 excludes_file
= xdg_config_home("ignore");
3352 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
3353 add_patterns_from_file_1(dir
, excludes_file
,
3354 dir
->untracked
? &dir
->internal
.ss_excludes_file
: NULL
);
3356 /* per repository user preference */
3357 if (startup_info
->have_repository
) {
3358 const char *path
= git_path_info_exclude();
3359 if (!access_or_warn(path
, R_OK
, 0))
3360 add_patterns_from_file_1(dir
, path
,
3361 dir
->untracked
? &dir
->internal
.ss_info_exclude
: NULL
);
3365 char *get_sparse_checkout_filename(void)
3367 return git_pathdup("info/sparse-checkout");
3370 int get_sparse_checkout_patterns(struct pattern_list
*pl
)
3373 char *sparse_filename
= get_sparse_checkout_filename();
3375 pl
->use_cone_patterns
= core_sparse_checkout_cone
;
3376 res
= add_patterns_from_file_to_list(sparse_filename
, "", 0, pl
, NULL
, 0);
3378 free(sparse_filename
);
3382 int remove_path(const char *name
)
3386 if (unlink(name
) && !is_missing_file_error(errno
))
3389 slash
= strrchr(name
, '/');
3391 char *dirs
= xstrdup(name
);
3392 slash
= dirs
+ (slash
- name
);
3395 if (startup_info
->original_cwd
&&
3396 !strcmp(startup_info
->original_cwd
, dirs
))
3398 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3405 * Frees memory within dir which was allocated, and resets fields for further
3406 * use. Does not free dir itself.
3408 void dir_clear(struct dir_struct
*dir
)
3411 struct exclude_list_group
*group
;
3412 struct pattern_list
*pl
;
3413 struct exclude_stack
*stk
;
3414 struct dir_struct
new = DIR_INIT
;
3416 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3417 group
= &dir
->internal
.exclude_list_group
[i
];
3418 for (j
= 0; j
< group
->nr
; j
++) {
3421 free((char *)pl
->src
);
3422 clear_pattern_list(pl
);
3427 for (i
= 0; i
< dir
->ignored_nr
; i
++)
3428 free(dir
->ignored
[i
]);
3429 for (i
= 0; i
< dir
->nr
; i
++)
3430 free(dir
->entries
[i
]);
3434 stk
= dir
->internal
.exclude_stack
;
3436 struct exclude_stack
*prev
= stk
->prev
;
3440 strbuf_release(&dir
->internal
.basebuf
);
3442 memcpy(dir
, &new, sizeof(*dir
));
3445 struct ondisk_untracked_cache
{
3446 struct stat_data info_exclude_stat
;
3447 struct stat_data excludes_file_stat
;
3451 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3454 int index
; /* number of written untracked_cache_dir */
3455 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3456 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3457 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3459 struct strbuf sb_stat
;
3460 struct strbuf sb_sha1
;
3463 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3465 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3466 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3467 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3468 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3469 to
->sd_dev
= htonl(from
->sd_dev
);
3470 to
->sd_ino
= htonl(from
->sd_ino
);
3471 to
->sd_uid
= htonl(from
->sd_uid
);
3472 to
->sd_gid
= htonl(from
->sd_gid
);
3473 to
->sd_size
= htonl(from
->sd_size
);
3476 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3477 struct write_data
*wd
)
3479 struct stat_data stat_data
;
3480 struct strbuf
*out
= &wd
->out
;
3481 unsigned char intbuf
[16];
3482 unsigned int intlen
, value
;
3483 int i
= wd
->index
++;
3486 * untracked_nr should be reset whenever valid is clear, but
3489 if (!untracked
->valid
) {
3490 untracked
->untracked_nr
= 0;
3491 untracked
->check_only
= 0;
3494 if (untracked
->check_only
)
3495 ewah_set(wd
->check_only
, i
);
3496 if (untracked
->valid
) {
3497 ewah_set(wd
->valid
, i
);
3498 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3499 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3501 if (!is_null_oid(&untracked
->exclude_oid
)) {
3502 ewah_set(wd
->sha1_valid
, i
);
3503 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3504 the_hash_algo
->rawsz
);
3507 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3508 strbuf_add(out
, intbuf
, intlen
);
3510 /* skip non-recurse directories */
3511 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3512 if (untracked
->dirs
[i
]->recurse
)
3514 intlen
= encode_varint(value
, intbuf
);
3515 strbuf_add(out
, intbuf
, intlen
);
3517 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3519 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3520 strbuf_add(out
, untracked
->untracked
[i
],
3521 strlen(untracked
->untracked
[i
]) + 1);
3523 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3524 if (untracked
->dirs
[i
]->recurse
)
3525 write_one_dir(untracked
->dirs
[i
], wd
);
3528 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3530 struct ondisk_untracked_cache
*ouc
;
3531 struct write_data wd
;
3532 unsigned char varbuf
[16];
3534 const unsigned hashsz
= the_hash_algo
->rawsz
;
3536 CALLOC_ARRAY(ouc
, 1);
3537 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3538 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3539 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3541 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3542 strbuf_add(out
, varbuf
, varint_len
);
3543 strbuf_addbuf(out
, &untracked
->ident
);
3545 strbuf_add(out
, ouc
, sizeof(*ouc
));
3546 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3547 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3548 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3551 if (!untracked
->root
) {
3552 varint_len
= encode_varint(0, varbuf
);
3553 strbuf_add(out
, varbuf
, varint_len
);
3558 wd
.check_only
= ewah_new();
3559 wd
.valid
= ewah_new();
3560 wd
.sha1_valid
= ewah_new();
3561 strbuf_init(&wd
.out
, 1024);
3562 strbuf_init(&wd
.sb_stat
, 1024);
3563 strbuf_init(&wd
.sb_sha1
, 1024);
3564 write_one_dir(untracked
->root
, &wd
);
3566 varint_len
= encode_varint(wd
.index
, varbuf
);
3567 strbuf_add(out
, varbuf
, varint_len
);
3568 strbuf_addbuf(out
, &wd
.out
);
3569 ewah_serialize_strbuf(wd
.valid
, out
);
3570 ewah_serialize_strbuf(wd
.check_only
, out
);
3571 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3572 strbuf_addbuf(out
, &wd
.sb_stat
);
3573 strbuf_addbuf(out
, &wd
.sb_sha1
);
3574 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3576 ewah_free(wd
.valid
);
3577 ewah_free(wd
.check_only
);
3578 ewah_free(wd
.sha1_valid
);
3579 strbuf_release(&wd
.out
);
3580 strbuf_release(&wd
.sb_stat
);
3581 strbuf_release(&wd
.sb_sha1
);
3584 static void free_untracked(struct untracked_cache_dir
*ucd
)
3589 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3590 free_untracked(ucd
->dirs
[i
]);
3591 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3592 free(ucd
->untracked
[i
]);
3593 free(ucd
->untracked
);
3598 void free_untracked_cache(struct untracked_cache
*uc
)
3603 free(uc
->exclude_per_dir_to_free
);
3604 strbuf_release(&uc
->ident
);
3605 free_untracked(uc
->root
);
3611 struct untracked_cache_dir
**ucd
;
3612 struct ewah_bitmap
*check_only
;
3613 struct ewah_bitmap
*valid
;
3614 struct ewah_bitmap
*sha1_valid
;
3615 const unsigned char *data
;
3616 const unsigned char *end
;
3619 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3621 memcpy(to
, data
, sizeof(*to
));
3622 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3623 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3624 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3625 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3626 to
->sd_dev
= ntohl(to
->sd_dev
);
3627 to
->sd_ino
= ntohl(to
->sd_ino
);
3628 to
->sd_uid
= ntohl(to
->sd_uid
);
3629 to
->sd_gid
= ntohl(to
->sd_gid
);
3630 to
->sd_size
= ntohl(to
->sd_size
);
3633 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3634 struct read_data
*rd
)
3636 struct untracked_cache_dir ud
, *untracked
;
3637 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3638 const unsigned char *eos
;
3642 memset(&ud
, 0, sizeof(ud
));
3644 value
= decode_varint(&data
);
3648 ud
.untracked_alloc
= value
;
3649 ud
.untracked_nr
= value
;
3650 if (ud
.untracked_nr
)
3651 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3653 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3656 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3658 eos
= memchr(data
, '\0', end
- data
);
3659 if (!eos
|| eos
== end
)
3662 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3663 memcpy(untracked
, &ud
, sizeof(ud
));
3664 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3667 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3668 eos
= memchr(data
, '\0', end
- data
);
3669 if (!eos
|| eos
== end
)
3671 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3675 rd
->ucd
[rd
->index
++] = untracked
;
3678 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3679 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3685 static void set_check_only(size_t pos
, void *cb
)
3687 struct read_data
*rd
= cb
;
3688 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3692 static void read_stat(size_t pos
, void *cb
)
3694 struct read_data
*rd
= cb
;
3695 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3696 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3697 rd
->data
= rd
->end
+ 1;
3700 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3701 rd
->data
+= sizeof(struct stat_data
);
3705 static void read_oid(size_t pos
, void *cb
)
3707 struct read_data
*rd
= cb
;
3708 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3709 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3710 rd
->data
= rd
->end
+ 1;
3713 oidread(&ud
->exclude_oid
, rd
->data
);
3714 rd
->data
+= the_hash_algo
->rawsz
;
3717 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3718 const unsigned char *sha1
)
3720 stat_data_from_disk(&oid_stat
->stat
, data
);
3721 oidread(&oid_stat
->oid
, sha1
);
3722 oid_stat
->valid
= 1;
3725 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3727 struct untracked_cache
*uc
;
3728 struct read_data rd
;
3729 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3733 const char *exclude_per_dir
;
3734 const unsigned hashsz
= the_hash_algo
->rawsz
;
3735 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3736 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3738 if (sz
<= 1 || end
[-1] != '\0')
3742 ident_len
= decode_varint(&next
);
3743 if (next
+ ident_len
> end
)
3745 ident
= (const char *)next
;
3748 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3751 CALLOC_ARRAY(uc
, 1);
3752 strbuf_init(&uc
->ident
, ident_len
);
3753 strbuf_add(&uc
->ident
, ident
, ident_len
);
3754 load_oid_stat(&uc
->ss_info_exclude
,
3755 next
+ ouc_offset(info_exclude_stat
),
3757 load_oid_stat(&uc
->ss_excludes_file
,
3758 next
+ ouc_offset(excludes_file_stat
),
3759 next
+ offset
+ hashsz
);
3760 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3761 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3762 uc
->exclude_per_dir
= uc
->exclude_per_dir_to_free
= xstrdup(exclude_per_dir
);
3763 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3764 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3768 len
= decode_varint(&next
);
3769 if (next
> end
|| len
== 0)
3772 rd
.valid
= ewah_new();
3773 rd
.check_only
= ewah_new();
3774 rd
.sha1_valid
= ewah_new();
3778 ALLOC_ARRAY(rd
.ucd
, len
);
3780 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3784 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3789 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3794 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3798 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3799 rd
.data
= next
+ len
;
3800 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3801 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3806 ewah_free(rd
.valid
);
3807 ewah_free(rd
.check_only
);
3808 ewah_free(rd
.sha1_valid
);
3811 free_untracked_cache(uc
);
3817 static void invalidate_one_directory(struct untracked_cache
*uc
,
3818 struct untracked_cache_dir
*ucd
)
3820 uc
->dir_invalidated
++;
3822 ucd
->untracked_nr
= 0;
3826 * Normally when an entry is added or removed from a directory,
3827 * invalidating that directory is enough. No need to touch its
3828 * ancestors. When a directory is shown as "foo/bar/" in git-status
3829 * however, deleting or adding an entry may have cascading effect.
3831 * Say the "foo/bar/file" has become untracked, we need to tell the
3832 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3833 * directory any more (because "bar" is managed by foo as an untracked
3836 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3837 * was the last untracked entry in the entire "foo", we should show
3838 * "foo/" instead. Which means we have to invalidate past "bar" up to
3841 * This function traverses all directories from root to leaf. If there
3842 * is a chance of one of the above cases happening, we invalidate back
3843 * to root. Otherwise we just invalidate the leaf. There may be a more
3844 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3845 * detect these cases and avoid unnecessary invalidation, for example,
3846 * checking for the untracked entry named "bar/" in "foo", but for now
3847 * stick to something safe and simple.
3849 static int invalidate_one_component(struct untracked_cache
*uc
,
3850 struct untracked_cache_dir
*dir
,
3851 const char *path
, int len
)
3853 const char *rest
= strchr(path
, '/');
3856 int component_len
= rest
- path
;
3857 struct untracked_cache_dir
*d
=
3858 lookup_untracked(uc
, dir
, path
, component_len
);
3860 invalidate_one_component(uc
, d
, rest
+ 1,
3861 len
- (component_len
+ 1));
3863 invalidate_one_directory(uc
, dir
);
3867 invalidate_one_directory(uc
, dir
);
3868 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3871 void untracked_cache_invalidate_path(struct index_state
*istate
,
3872 const char *path
, int safe_path
)
3874 if (!istate
->untracked
|| !istate
->untracked
->root
)
3876 if (!safe_path
&& !verify_path(path
, 0))
3878 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3879 path
, strlen(path
));
3882 void untracked_cache_remove_from_index(struct index_state
*istate
,
3885 untracked_cache_invalidate_path(istate
, path
, 1);
3888 void untracked_cache_add_to_index(struct index_state
*istate
,
3891 untracked_cache_invalidate_path(istate
, path
, 1);
3894 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3895 const char *sub_gitdir
)
3898 struct repository subrepo
;
3899 struct strbuf sub_wt
= STRBUF_INIT
;
3900 struct strbuf sub_gd
= STRBUF_INIT
;
3902 const struct submodule
*sub
;
3904 /* If the submodule has no working tree, we can ignore it. */
3905 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3908 if (repo_read_index(&subrepo
) < 0)
3909 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3911 /* TODO: audit for interaction with sparse-index. */
3912 ensure_full_index(subrepo
.index
);
3913 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3914 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3916 if (!S_ISGITLINK(ce
->ce_mode
))
3919 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3920 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3922 * Skip entries with the same name in different stages
3923 * to make sure an entry is returned only once.
3927 sub
= submodule_from_path(&subrepo
, null_oid(), ce
->name
);
3928 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3929 /* .gitmodules broken or inactive sub */
3932 strbuf_reset(&sub_wt
);
3933 strbuf_reset(&sub_gd
);
3934 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3935 submodule_name_to_gitdir(&sub_gd
, &subrepo
, sub
->name
);
3937 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3939 strbuf_release(&sub_wt
);
3940 strbuf_release(&sub_gd
);
3941 repo_clear(&subrepo
);
3944 void connect_work_tree_and_git_dir(const char *work_tree_
,
3945 const char *git_dir_
,
3946 int recurse_into_nested
)
3948 struct strbuf gitfile_sb
= STRBUF_INIT
;
3949 struct strbuf cfg_sb
= STRBUF_INIT
;
3950 struct strbuf rel_path
= STRBUF_INIT
;
3951 char *git_dir
, *work_tree
;
3953 /* Prepare .git file */
3954 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3955 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3956 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3958 /* Prepare config file */
3959 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3960 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3961 die(_("could not create directories for %s"), cfg_sb
.buf
);
3963 git_dir
= real_pathdup(git_dir_
, 1);
3964 work_tree
= real_pathdup(work_tree_
, 1);
3966 /* Write .git file */
3967 write_file(gitfile_sb
.buf
, "gitdir: %s",
3968 relative_path(git_dir
, work_tree
, &rel_path
));
3969 /* Update core.worktree setting */
3970 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3971 relative_path(work_tree
, git_dir
, &rel_path
));
3973 strbuf_release(&gitfile_sb
);
3974 strbuf_release(&cfg_sb
);
3975 strbuf_release(&rel_path
);
3977 if (recurse_into_nested
)
3978 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3985 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3987 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3989 if (rename(old_git_dir
, new_git_dir
) < 0)
3990 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3991 old_git_dir
, new_git_dir
);
3993 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);
3996 int path_match_flags(const char *const str
, const enum path_match_flags flags
)
3998 const char *p
= str
;
4000 if (flags
& PATH_MATCH_NATIVE
&&
4001 flags
& PATH_MATCH_XPLATFORM
)
4002 BUG("path_match_flags() must get one match kind, not multiple!");
4003 else if (!(flags
& PATH_MATCH_KINDS_MASK
))
4004 BUG("path_match_flags() must get at least one match kind!");
4006 if (flags
& PATH_MATCH_STARTS_WITH_DOT_SLASH
&&
4007 flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
)
4008 BUG("path_match_flags() must get one platform kind, not multiple!");
4009 else if (!(flags
& PATH_MATCH_PLATFORM_MASK
))
4010 BUG("path_match_flags() must get at least one platform kind!");
4014 if (flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
&&
4018 if (flags
& PATH_MATCH_NATIVE
)
4019 return is_dir_sep(*p
);
4020 else if (flags
& PATH_MATCH_XPLATFORM
)
4021 return is_xplatform_dir_sep(*p
);