2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
8 #include "git-compat-util.h"
14 #include "environment.h"
16 #include "object-store.h"
19 #include "wildmatch.h"
23 #include "ewah/ewok.h"
24 #include "fsmonitor.h"
26 #include "submodule-config.h"
31 * Tells read_directory_recursive how a file or directory should be treated.
32 * Values are ordered by significance, e.g. if a directory contains both
33 * excluded and untracked files, it is listed as untracked because
34 * path_untracked > path_excluded.
44 * Support data structure for our opendir/readdir/closedir wrappers
48 struct untracked_cache_dir
*untracked
;
55 struct untracked_cache_dir
*ucd
;
58 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
59 struct index_state
*istate
, const char *path
, int len
,
60 struct untracked_cache_dir
*untracked
,
61 int check_only
, int stop_at_first_file
, const struct pathspec
*pathspec
);
62 static int resolve_dtype(int dtype
, struct index_state
*istate
,
63 const char *path
, int len
);
64 struct dirent
*readdir_skip_dot_and_dotdot(DIR *dirp
)
68 while ((e
= readdir(dirp
)) != NULL
) {
69 if (!is_dot_or_dotdot(e
->d_name
))
75 int count_slashes(const char *s
)
84 int fspathcmp(const char *a
, const char *b
)
86 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
89 int fspatheq(const char *a
, const char *b
)
91 return !fspathcmp(a
, b
);
94 int fspathncmp(const char *a
, const char *b
, size_t count
)
96 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
99 unsigned int fspathhash(const char *str
)
101 return ignore_case
? strihash(str
) : strhash(str
);
104 int git_fnmatch(const struct pathspec_item
*item
,
105 const char *pattern
, const char *string
,
109 if (ps_strncmp(item
, pattern
, string
, prefix
))
114 if (item
->flags
& PATHSPEC_ONESTAR
) {
115 int pattern_len
= strlen(++pattern
);
116 int string_len
= strlen(string
);
117 return string_len
< pattern_len
||
118 ps_strcmp(item
, pattern
,
119 string
+ string_len
- pattern_len
);
121 if (item
->magic
& PATHSPEC_GLOB
)
122 return wildmatch(pattern
, string
,
124 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
126 /* wildmatch has not learned no FNM_PATHNAME mode yet */
127 return wildmatch(pattern
, string
,
128 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
131 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
132 const char *string
, int stringlen
,
136 struct strbuf pat_buf
= STRBUF_INIT
;
137 struct strbuf str_buf
= STRBUF_INIT
;
138 const char *use_pat
= pattern
;
139 const char *use_str
= string
;
141 if (pattern
[patternlen
]) {
142 strbuf_add(&pat_buf
, pattern
, patternlen
);
143 use_pat
= pat_buf
.buf
;
145 if (string
[stringlen
]) {
146 strbuf_add(&str_buf
, string
, stringlen
);
147 use_str
= str_buf
.buf
;
151 flags
|= WM_CASEFOLD
;
152 match_status
= wildmatch(use_pat
, use_str
, flags
);
154 strbuf_release(&pat_buf
);
155 strbuf_release(&str_buf
);
160 static size_t common_prefix_len(const struct pathspec
*pathspec
)
166 * ":(icase)path" is treated as a pathspec full of
167 * wildcard. In other words, only prefix is considered common
168 * prefix. If the pathspec is abc/foo abc/bar, running in
169 * subdir xyz, the common prefix is still xyz, not xyz/abc as
172 GUARD_PATHSPEC(pathspec
,
181 for (n
= 0; n
< pathspec
->nr
; n
++) {
182 size_t i
= 0, len
= 0, item_len
;
183 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
185 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
186 item_len
= pathspec
->items
[n
].prefix
;
188 item_len
= pathspec
->items
[n
].nowildcard_len
;
189 while (i
< item_len
&& (n
== 0 || i
< max
)) {
190 char c
= pathspec
->items
[n
].match
[i
];
191 if (c
!= pathspec
->items
[0].match
[i
])
197 if (n
== 0 || len
< max
) {
207 * Returns a copy of the longest leading path common among all
210 char *common_prefix(const struct pathspec
*pathspec
)
212 unsigned long len
= common_prefix_len(pathspec
);
214 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
217 int fill_directory(struct dir_struct
*dir
,
218 struct index_state
*istate
,
219 const struct pathspec
*pathspec
)
224 unsigned exclusive_flags
= DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
;
225 if ((dir
->flags
& exclusive_flags
) == exclusive_flags
)
226 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
229 * Calculate common prefix for the pathspec, and
230 * use that to optimize the directory walk
232 prefix_len
= common_prefix_len(pathspec
);
233 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
235 /* Read the directory and prune it */
236 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
241 int within_depth(const char *name
, int namelen
,
242 int depth
, int max_depth
)
244 const char *cp
= name
, *cpe
= name
+ namelen
;
250 if (depth
> max_depth
)
257 * Read the contents of the blob with the given OID into a buffer.
258 * Append a trailing LF to the end if the last line doesn't have one.
261 * -1 when the OID is invalid or unknown or does not refer to a blob.
262 * 0 when the blob is empty.
263 * 1 along with { data, size } of the (possibly augmented) buffer
266 * Optionally updates the given oid_stat with the given OID (when valid).
268 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
269 size_t *size_out
, char **data_out
)
271 enum object_type type
;
278 data
= repo_read_object_file(the_repository
, oid
, &type
, &sz
);
279 if (!data
|| type
!= OBJ_BLOB
) {
285 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
286 oidcpy(&oid_stat
->oid
, oid
);
294 if (data
[sz
- 1] != '\n') {
295 data
= xrealloc(data
, st_add(sz
, 1));
299 *size_out
= xsize_t(sz
);
305 #define DO_MATCH_EXCLUDE (1<<0)
306 #define DO_MATCH_DIRECTORY (1<<1)
307 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
310 * Does the given pathspec match the given name? A match is found if
312 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
313 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
314 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
315 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
317 * Return value tells which case it was (1-4), or 0 when there is no match.
319 * It may be instructive to look at a small table of concrete examples
320 * to understand the differences between 1, 2, and 4:
323 * | a/b | a/b/ | a/b/c
324 * ------+-----------+-----------+------------
325 * a/b | EXACT | EXACT[1] | LEADING[2]
326 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
327 * a/b/c | RECURSIVE | RECURSIVE | EXACT
329 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
330 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
332 static int match_pathspec_item(struct index_state
*istate
,
333 const struct pathspec_item
*item
, int prefix
,
334 const char *name
, int namelen
, unsigned flags
)
336 /* name/namelen has prefix cut off by caller */
337 const char *match
= item
->match
+ prefix
;
338 int matchlen
= item
->len
- prefix
;
341 * The normal call pattern is:
342 * 1. prefix = common_prefix_len(ps);
343 * 2. prune something, or fill_directory
344 * 3. match_pathspec()
346 * 'prefix' at #1 may be shorter than the command's prefix and
347 * it's ok for #2 to match extra files. Those extras will be
350 * Suppose the pathspec is 'foo' and '../bar' running from
351 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
352 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
353 * user does not want XYZ/foo, only the "foo" part should be
354 * case-insensitive. We need to filter out XYZ/foo here. In
355 * other words, we do not trust the caller on comparing the
356 * prefix part when :(icase) is involved. We do exact
357 * comparison ourselves.
359 * Normally the caller (common_prefix_len() in fact) does
360 * _exact_ matching on name[-prefix+1..-1] and we do not need
361 * to check that part. Be defensive and check it anyway, in
362 * case common_prefix_len is changed, or a new caller is
363 * introduced that does not use common_prefix_len.
365 * If the penalty turns out too high when prefix is really
366 * long, maybe change it to
367 * strncmp(match, name, item->prefix - prefix)
369 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
370 strncmp(item
->match
, name
- prefix
, item
->prefix
))
373 if (item
->attr_match_nr
&&
374 !match_pathspec_attrs(istate
, name
, namelen
, item
))
377 /* If the match was just the prefix, we matched */
379 return MATCHED_RECURSIVELY
;
381 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
382 if (matchlen
== namelen
)
383 return MATCHED_EXACTLY
;
385 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
386 return MATCHED_RECURSIVELY
;
387 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
388 match
[matchlen
- 1] == '/' &&
389 namelen
== matchlen
- 1 &&
390 !ps_strncmp(item
, match
, name
, namelen
))
391 return MATCHED_EXACTLY
;
393 if (item
->nowildcard_len
< item
->len
&&
394 !git_fnmatch(item
, match
, name
,
395 item
->nowildcard_len
- prefix
))
396 return MATCHED_FNMATCH
;
398 /* Perform checks to see if "name" is a leading string of the pathspec */
399 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
400 !(flags
& DO_MATCH_EXCLUDE
)) {
401 /* name is a literal prefix of the pathspec */
402 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
403 if ((namelen
< matchlen
) &&
404 (match
[namelen
-offset
] == '/') &&
405 !ps_strncmp(item
, match
, name
, namelen
))
406 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
408 /* name doesn't match up to the first wild character */
409 if (item
->nowildcard_len
< item
->len
&&
410 ps_strncmp(item
, match
, name
,
411 item
->nowildcard_len
- prefix
))
415 * name has no wildcard, and it didn't match as a leading
416 * pathspec so return.
418 if (item
->nowildcard_len
== item
->len
)
422 * Here is where we would perform a wildmatch to check if
423 * "name" can be matched as a directory (or a prefix) against
424 * the pathspec. Since wildmatch doesn't have this capability
425 * at the present we have to punt and say that it is a match,
426 * potentially returning a false positive
427 * The submodules themselves will be able to perform more
428 * accurate matching to determine if the pathspec matches.
430 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
437 * do_match_pathspec() is meant to ONLY be called by
438 * match_pathspec_with_flags(); calling it directly risks pathspecs
439 * like ':!unwanted_path' being ignored.
441 * Given a name and a list of pathspecs, returns the nature of the
442 * closest (i.e. most specific) match of the name to any of the
445 * The caller typically calls this multiple times with the same
446 * pathspec and seen[] array but with different name/namelen
447 * (e.g. entries from the index) and is interested in seeing if and
448 * how each pathspec matches all the names it calls this function
449 * with. A mark is left in the seen[] array for each pathspec element
450 * indicating the closest type of match that element achieved, so if
451 * seen[n] remains zero after multiple invocations, that means the nth
452 * pathspec did not match any names, which could indicate that the
453 * user mistyped the nth pathspec.
455 static int do_match_pathspec(struct index_state
*istate
,
456 const struct pathspec
*ps
,
457 const char *name
, int namelen
,
458 int prefix
, char *seen
,
461 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
473 if (!ps
->recursive
||
474 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
476 return MATCHED_RECURSIVELY
;
478 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
479 return MATCHED_EXACTLY
;
487 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
490 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
491 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
494 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
497 * Make exclude patterns optional and never report
498 * "pathspec ':(exclude)foo' matches no files"
500 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
501 seen
[i
] = MATCHED_FNMATCH
;
502 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
505 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
506 ps
->max_depth
!= -1 &&
507 how
&& how
!= MATCHED_FNMATCH
) {
508 int len
= ps
->items
[i
].len
;
509 if (name
[len
] == '/')
511 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
512 how
= MATCHED_EXACTLY
;
519 if (seen
&& seen
[i
] < how
)
526 static int match_pathspec_with_flags(struct index_state
*istate
,
527 const struct pathspec
*ps
,
528 const char *name
, int namelen
,
529 int prefix
, char *seen
, unsigned flags
)
531 int positive
, negative
;
532 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
533 prefix
, seen
, flags
);
534 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
536 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
538 flags
| DO_MATCH_EXCLUDE
);
539 return negative
? 0 : positive
;
542 int match_pathspec(struct index_state
*istate
,
543 const struct pathspec
*ps
,
544 const char *name
, int namelen
,
545 int prefix
, char *seen
, int is_dir
)
547 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
548 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
549 prefix
, seen
, flags
);
553 * Check if a submodule is a superset of the pathspec
555 int submodule_path_match(struct index_state
*istate
,
556 const struct pathspec
*ps
,
557 const char *submodule_name
,
560 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
561 strlen(submodule_name
),
564 DO_MATCH_LEADING_PATHSPEC
);
568 int report_path_error(const char *ps_matched
,
569 const struct pathspec
*pathspec
)
572 * Make sure all pathspec matched; otherwise it is an error.
575 for (num
= 0; num
< pathspec
->nr
; num
++) {
576 int other
, found_dup
;
581 * The caller might have fed identical pathspec
582 * twice. Do not barf on such a mistake.
583 * FIXME: parse_pathspec should have eliminated
584 * duplicate pathspec.
586 for (found_dup
= other
= 0;
587 !found_dup
&& other
< pathspec
->nr
;
589 if (other
== num
|| !ps_matched
[other
])
591 if (!strcmp(pathspec
->items
[other
].original
,
592 pathspec
->items
[num
].original
))
594 * Ok, we have a match already.
601 error(_("pathspec '%s' did not match any file(s) known to git"),
602 pathspec
->items
[num
].original
);
609 * Return the length of the "simple" part of a path match limiter.
611 int simple_length(const char *match
)
616 unsigned char c
= *match
++;
618 if (c
== '\0' || is_glob_special(c
))
623 int no_wildcard(const char *string
)
625 return string
[simple_length(string
)] == '\0';
628 void parse_path_pattern(const char **pattern
,
633 const char *p
= *pattern
;
638 *flags
|= PATTERN_FLAG_NEGATIVE
;
642 if (len
&& p
[len
- 1] == '/') {
644 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
646 for (i
= 0; i
< len
; i
++) {
651 *flags
|= PATTERN_FLAG_NODIR
;
652 *nowildcardlen
= simple_length(p
);
654 * we should have excluded the trailing slash from 'p' too,
655 * but that's one more allocation. Instead just make sure
656 * nowildcardlen does not exceed real patternlen
658 if (*nowildcardlen
> len
)
659 *nowildcardlen
= len
;
660 if (*p
== '*' && no_wildcard(p
+ 1))
661 *flags
|= PATTERN_FLAG_ENDSWITH
;
666 int pl_hashmap_cmp(const void *cmp_data UNUSED
,
667 const struct hashmap_entry
*a
,
668 const struct hashmap_entry
*b
,
669 const void *key UNUSED
)
671 const struct pattern_entry
*ee1
=
672 container_of(a
, struct pattern_entry
, ent
);
673 const struct pattern_entry
*ee2
=
674 container_of(b
, struct pattern_entry
, ent
);
676 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
680 return fspathncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
683 static char *dup_and_filter_pattern(const char *pattern
)
687 char *result
= xstrdup(pattern
);
693 /* skip escape characters (once) */
713 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
715 struct pattern_entry
*translated
;
718 const char *prev
, *cur
, *next
;
720 if (!pl
->use_cone_patterns
)
723 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
724 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
725 !strcmp(given
->pattern
, "/*")) {
730 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
735 if (given
->patternlen
< 2 ||
736 *given
->pattern
!= '/' ||
737 strstr(given
->pattern
, "**")) {
738 /* Not a cone pattern. */
739 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
743 if (!(given
->flags
& PATTERN_FLAG_MUSTBEDIR
) &&
744 strcmp(given
->pattern
, "/*")) {
745 /* Not a cone pattern. */
746 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
750 prev
= given
->pattern
;
751 cur
= given
->pattern
+ 1;
752 next
= given
->pattern
+ 2;
755 /* Watch for glob characters '*', '\', '[', '?' */
756 if (!is_glob_special(*cur
))
759 /* But only if *prev != '\\' */
763 /* But allow the initial '\' */
765 is_glob_special(*next
))
768 /* But a trailing '/' then '*' is fine */
774 /* Not a cone pattern. */
775 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
784 if (given
->patternlen
> 2 &&
785 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
786 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
787 /* Not a cone pattern. */
788 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
792 truncated
= dup_and_filter_pattern(given
->pattern
);
794 translated
= xmalloc(sizeof(struct pattern_entry
));
795 translated
->pattern
= truncated
;
796 translated
->patternlen
= given
->patternlen
- 2;
797 hashmap_entry_init(&translated
->ent
,
798 fspathhash(translated
->pattern
));
800 if (!hashmap_get_entry(&pl
->recursive_hashmap
,
801 translated
, ent
, NULL
)) {
802 /* We did not see the "parent" included */
803 warning(_("unrecognized negative pattern: '%s'"),
810 hashmap_add(&pl
->parent_hashmap
, &translated
->ent
);
811 hashmap_remove(&pl
->recursive_hashmap
, &translated
->ent
, &data
);
816 if (given
->flags
& PATTERN_FLAG_NEGATIVE
) {
817 warning(_("unrecognized negative pattern: '%s'"),
822 translated
= xmalloc(sizeof(struct pattern_entry
));
824 translated
->pattern
= dup_and_filter_pattern(given
->pattern
);
825 translated
->patternlen
= given
->patternlen
;
826 hashmap_entry_init(&translated
->ent
,
827 fspathhash(translated
->pattern
));
829 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
831 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
832 /* we already included this at the parent level */
833 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
841 warning(_("disabling cone pattern matching"));
842 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
843 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
844 pl
->use_cone_patterns
= 0;
847 static int hashmap_contains_path(struct hashmap
*map
,
848 struct strbuf
*pattern
)
850 struct pattern_entry p
;
852 /* Check straight mapping */
853 p
.pattern
= pattern
->buf
;
854 p
.patternlen
= pattern
->len
;
855 hashmap_entry_init(&p
.ent
, fspathhash(p
.pattern
));
856 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
859 int hashmap_contains_parent(struct hashmap
*map
,
861 struct strbuf
*buffer
)
865 strbuf_setlen(buffer
, 0);
868 strbuf_addch(buffer
, '/');
870 strbuf_addstr(buffer
, path
);
872 slash_pos
= strrchr(buffer
->buf
, '/');
874 while (slash_pos
> buffer
->buf
) {
875 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
877 if (hashmap_contains_path(map
, buffer
))
880 slash_pos
= strrchr(buffer
->buf
, '/');
886 void add_pattern(const char *string
, const char *base
,
887 int baselen
, struct pattern_list
*pl
, int srcpos
)
889 struct path_pattern
*pattern
;
894 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
895 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
896 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
898 pattern
= xmalloc(sizeof(*pattern
));
899 pattern
->pattern
= string
;
901 pattern
->patternlen
= patternlen
;
902 pattern
->nowildcardlen
= nowildcardlen
;
903 pattern
->base
= base
;
904 pattern
->baselen
= baselen
;
905 pattern
->flags
= flags
;
906 pattern
->srcpos
= srcpos
;
907 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
908 pl
->patterns
[pl
->nr
++] = pattern
;
911 add_pattern_to_hashsets(pl
, pattern
);
914 static int read_skip_worktree_file_from_index(struct index_state
*istate
,
916 size_t *size_out
, char **data_out
,
917 struct oid_stat
*oid_stat
)
922 pos
= index_name_pos(istate
, path
, len
);
925 if (!ce_skip_worktree(istate
->cache
[pos
]))
928 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
932 * Frees memory within pl which was allocated for exclude patterns and
933 * the file buffer. Does not free pl itself.
935 void clear_pattern_list(struct pattern_list
*pl
)
939 for (i
= 0; i
< pl
->nr
; i
++)
940 free(pl
->patterns
[i
]);
943 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
944 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
946 memset(pl
, 0, sizeof(*pl
));
949 static void trim_trailing_spaces(char *buf
)
951 char *p
, *last_space
= NULL
;
953 for (p
= buf
; *p
; p
++)
973 * Given a subdirectory name and "dir" of the current directory,
974 * search the subdir in "dir" and return it, or create a new one if it
975 * does not exist in "dir".
977 * If "name" has the trailing slash, it'll be excluded in the search.
979 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
980 struct untracked_cache_dir
*dir
,
981 const char *name
, int len
)
984 struct untracked_cache_dir
*d
;
987 if (len
&& name
[len
- 1] == '/')
991 while (last
> first
) {
992 int cmp
, next
= first
+ ((last
- first
) >> 1);
994 cmp
= strncmp(name
, d
->name
, len
);
995 if (!cmp
&& strlen(d
->name
) > len
)
1007 FLEX_ALLOC_MEM(d
, name
, name
, len
);
1009 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
1010 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
1011 dir
->dirs_nr
- first
);
1013 dir
->dirs
[first
] = d
;
1017 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
1021 dir
->untracked_nr
= 0;
1022 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1023 do_invalidate_gitignore(dir
->dirs
[i
]);
1026 static void invalidate_gitignore(struct untracked_cache
*uc
,
1027 struct untracked_cache_dir
*dir
)
1029 uc
->gitignore_invalidated
++;
1030 do_invalidate_gitignore(dir
);
1033 static void invalidate_directory(struct untracked_cache
*uc
,
1034 struct untracked_cache_dir
*dir
)
1039 * Invalidation increment here is just roughly correct. If
1040 * untracked_nr or any of dirs[].recurse is non-zero, we
1041 * should increment dir_invalidated too. But that's more
1045 uc
->dir_invalidated
++;
1048 dir
->untracked_nr
= 0;
1049 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1050 dir
->dirs
[i
]->recurse
= 0;
1053 static int add_patterns_from_buffer(char *buf
, size_t size
,
1054 const char *base
, int baselen
,
1055 struct pattern_list
*pl
);
1057 /* Flags for add_patterns() */
1058 #define PATTERN_NOFOLLOW (1<<0)
1061 * Given a file with name "fname", read it (either from disk, or from
1062 * an index if 'istate' is non-null), parse it and store the
1063 * exclude rules in "pl".
1065 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1066 * stat data from disk (only valid if add_patterns returns zero). If
1067 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1069 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1070 struct pattern_list
*pl
, struct index_state
*istate
,
1071 unsigned flags
, struct oid_stat
*oid_stat
)
1079 if (flags
& PATTERN_NOFOLLOW
)
1080 fd
= open_nofollow(fname
, O_RDONLY
);
1082 fd
= open(fname
, O_RDONLY
);
1084 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1086 warn_on_fopen_errors(fname
);
1091 r
= read_skip_worktree_file_from_index(istate
, fname
,
1097 size
= xsize_t(st
.st_size
);
1100 fill_stat_data(&oid_stat
->stat
, &st
);
1101 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1102 oid_stat
->valid
= 1;
1107 buf
= xmallocz(size
);
1108 if (read_in_full(fd
, buf
, size
) != size
) {
1117 if (oid_stat
->valid
&&
1118 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1119 ; /* no content change, oid_stat->oid still good */
1121 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1122 !ce_stage(istate
->cache
[pos
]) &&
1123 ce_uptodate(istate
->cache
[pos
]) &&
1124 !would_convert_to_git(istate
, fname
))
1125 oidcpy(&oid_stat
->oid
,
1126 &istate
->cache
[pos
]->oid
);
1128 hash_object_file(the_hash_algo
, buf
, size
,
1129 OBJ_BLOB
, &oid_stat
->oid
);
1130 fill_stat_data(&oid_stat
->stat
, &st
);
1131 oid_stat
->valid
= 1;
1135 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1139 static int add_patterns_from_buffer(char *buf
, size_t size
,
1140 const char *base
, int baselen
,
1141 struct pattern_list
*pl
)
1146 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1147 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1151 if (skip_utf8_bom(&buf
, size
))
1152 size
-= buf
- pl
->filebuf
;
1156 for (i
= 0; i
< size
; i
++) {
1157 if (buf
[i
] == '\n') {
1158 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1159 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1160 trim_trailing_spaces(entry
);
1161 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1164 entry
= buf
+ i
+ 1;
1170 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1171 int baselen
, struct pattern_list
*pl
,
1172 struct index_state
*istate
,
1175 return add_patterns(fname
, base
, baselen
, pl
, istate
, flags
, NULL
);
1178 int add_patterns_from_blob_to_list(
1179 struct object_id
*oid
,
1180 const char *base
, int baselen
,
1181 struct pattern_list
*pl
)
1187 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1191 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1195 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1196 int group_type
, const char *src
)
1198 struct pattern_list
*pl
;
1199 struct exclude_list_group
*group
;
1201 group
= &dir
->internal
.exclude_list_group
[group_type
];
1202 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1203 pl
= &group
->pl
[group
->nr
++];
1204 memset(pl
, 0, sizeof(*pl
));
1210 * Used to set up core.excludesfile and .git/info/exclude lists.
1212 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1213 struct oid_stat
*oid_stat
)
1215 struct pattern_list
*pl
;
1217 * catch setup_standard_excludes() that's called before
1218 * dir->untracked is assigned. That function behaves
1219 * differently when dir->untracked is non-NULL.
1221 if (!dir
->untracked
)
1222 dir
->internal
.unmanaged_exclude_files
++;
1223 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1224 if (add_patterns(fname
, "", 0, pl
, NULL
, 0, oid_stat
) < 0)
1225 die(_("cannot use %s as an exclude file"), fname
);
1228 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1230 dir
->internal
.unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1231 add_patterns_from_file_1(dir
, fname
, NULL
);
1234 int match_basename(const char *basename
, int basenamelen
,
1235 const char *pattern
, int prefix
, int patternlen
,
1238 if (prefix
== patternlen
) {
1239 if (patternlen
== basenamelen
&&
1240 !fspathncmp(pattern
, basename
, basenamelen
))
1242 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1243 /* "*literal" matching against "fooliteral" */
1244 if (patternlen
- 1 <= basenamelen
&&
1245 !fspathncmp(pattern
+ 1,
1246 basename
+ basenamelen
- (patternlen
- 1),
1250 if (fnmatch_icase_mem(pattern
, patternlen
,
1251 basename
, basenamelen
,
1258 int match_pathname(const char *pathname
, int pathlen
,
1259 const char *base
, int baselen
,
1260 const char *pattern
, int prefix
, int patternlen
)
1266 * match with FNM_PATHNAME; the pattern has base implicitly
1269 if (*pattern
== '/') {
1276 * baselen does not count the trailing slash. base[] may or
1277 * may not end with a trailing slash though.
1279 if (pathlen
< baselen
+ 1 ||
1280 (baselen
&& pathname
[baselen
] != '/') ||
1281 fspathncmp(pathname
, base
, baselen
))
1284 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1285 name
= pathname
+ pathlen
- namelen
;
1289 * if the non-wildcard part is longer than the
1290 * remaining pathname, surely it cannot match.
1292 if (prefix
> namelen
)
1295 if (fspathncmp(pattern
, name
, prefix
))
1298 patternlen
-= prefix
;
1303 * If the whole pattern did not have a wildcard,
1304 * then our prefix match is all we need; we
1305 * do not need to call fnmatch at all.
1307 if (!patternlen
&& !namelen
)
1311 return fnmatch_icase_mem(pattern
, patternlen
,
1317 * Scan the given exclude list in reverse to see whether pathname
1318 * should be ignored. The first match (i.e. the last on the list), if
1319 * any, determines the fate. Returns the exclude_list element which
1320 * matched, or NULL for undecided.
1322 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1324 const char *basename
,
1326 struct pattern_list
*pl
,
1327 struct index_state
*istate
)
1329 struct path_pattern
*res
= NULL
; /* undecided */
1333 return NULL
; /* undefined */
1335 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1336 struct path_pattern
*pattern
= pl
->patterns
[i
];
1337 const char *exclude
= pattern
->pattern
;
1338 int prefix
= pattern
->nowildcardlen
;
1340 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1341 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1342 if (*dtype
!= DT_DIR
)
1346 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1347 if (match_basename(basename
,
1348 pathlen
- (basename
- pathname
),
1349 exclude
, prefix
, pattern
->patternlen
,
1357 assert(pattern
->baselen
== 0 ||
1358 pattern
->base
[pattern
->baselen
- 1] == '/');
1359 if (match_pathname(pathname
, pathlen
,
1361 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1362 exclude
, prefix
, pattern
->patternlen
)) {
1371 * Scan the list of patterns to determine if the ordered list
1372 * of patterns matches on 'pathname'.
1374 * Return 1 for a match, 0 for not matched and -1 for undecided.
1376 enum pattern_match_result
path_matches_pattern_list(
1377 const char *pathname
, int pathlen
,
1378 const char *basename
, int *dtype
,
1379 struct pattern_list
*pl
,
1380 struct index_state
*istate
)
1382 struct path_pattern
*pattern
;
1383 struct strbuf parent_pathname
= STRBUF_INIT
;
1384 int result
= NOT_MATCHED
;
1387 if (!pl
->use_cone_patterns
) {
1388 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1391 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1403 strbuf_addch(&parent_pathname
, '/');
1404 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1407 * Directory entries are matched if and only if a file
1408 * contained immediately within them is matched. For the
1409 * case of a directory entry, modify the path to create
1410 * a fake filename within this directory, allowing us to
1411 * use the file-base matching logic in an equivalent way.
1413 if (parent_pathname
.len
> 0 &&
1414 parent_pathname
.buf
[parent_pathname
.len
- 1] == '/') {
1415 slash_pos
= parent_pathname
.len
- 1;
1416 strbuf_add(&parent_pathname
, "-", 1);
1418 const char *slash_ptr
= strrchr(parent_pathname
.buf
, '/');
1419 slash_pos
= slash_ptr
? slash_ptr
- parent_pathname
.buf
: 0;
1422 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1423 &parent_pathname
)) {
1424 result
= MATCHED_RECURSIVE
;
1429 /* include every file in root */
1434 strbuf_setlen(&parent_pathname
, slash_pos
);
1436 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1441 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1444 result
= MATCHED_RECURSIVE
;
1447 strbuf_release(&parent_pathname
);
1451 int init_sparse_checkout_patterns(struct index_state
*istate
)
1453 if (!core_apply_sparse_checkout
)
1455 if (istate
->sparse_checkout_patterns
)
1458 CALLOC_ARRAY(istate
->sparse_checkout_patterns
, 1);
1460 if (get_sparse_checkout_patterns(istate
->sparse_checkout_patterns
) < 0) {
1461 FREE_AND_NULL(istate
->sparse_checkout_patterns
);
1468 static int path_in_sparse_checkout_1(const char *path
,
1469 struct index_state
*istate
,
1470 int require_cone_mode
)
1473 enum pattern_match_result match
= UNDECIDED
;
1474 const char *end
, *slash
;
1477 * We default to accepting a path if the path is empty, there are no
1478 * patterns, or the patterns are of the wrong type.
1481 init_sparse_checkout_patterns(istate
) ||
1482 (require_cone_mode
&&
1483 !istate
->sparse_checkout_patterns
->use_cone_patterns
))
1487 * If UNDECIDED, use the match from the parent dir (recursively), or
1488 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1489 * never returns UNDECIDED, so we will execute only one iteration in
1492 for (end
= path
+ strlen(path
);
1493 end
> path
&& match
== UNDECIDED
;
1496 for (slash
= end
- 1; slash
> path
&& *slash
!= '/'; slash
--)
1499 match
= path_matches_pattern_list(path
, end
- path
,
1500 slash
> path
? slash
+ 1 : path
, &dtype
,
1501 istate
->sparse_checkout_patterns
, istate
);
1503 /* We are going to match the parent dir now */
1509 int path_in_sparse_checkout(const char *path
,
1510 struct index_state
*istate
)
1512 return path_in_sparse_checkout_1(path
, istate
, 0);
1515 int path_in_cone_mode_sparse_checkout(const char *path
,
1516 struct index_state
*istate
)
1518 return path_in_sparse_checkout_1(path
, istate
, 1);
1521 static struct path_pattern
*last_matching_pattern_from_lists(
1522 struct dir_struct
*dir
, struct index_state
*istate
,
1523 const char *pathname
, int pathlen
,
1524 const char *basename
, int *dtype_p
)
1527 struct exclude_list_group
*group
;
1528 struct path_pattern
*pattern
;
1529 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1530 group
= &dir
->internal
.exclude_list_group
[i
];
1531 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1532 pattern
= last_matching_pattern_from_list(
1533 pathname
, pathlen
, basename
, dtype_p
,
1534 &group
->pl
[j
], istate
);
1543 * Loads the per-directory exclude list for the substring of base
1544 * which has a char length of baselen.
1546 static void prep_exclude(struct dir_struct
*dir
,
1547 struct index_state
*istate
,
1548 const char *base
, int baselen
)
1550 struct exclude_list_group
*group
;
1551 struct pattern_list
*pl
;
1552 struct exclude_stack
*stk
= NULL
;
1553 struct untracked_cache_dir
*untracked
;
1556 group
= &dir
->internal
.exclude_list_group
[EXC_DIRS
];
1559 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1560 * which originate from directories not in the prefix of the
1561 * path being checked.
1563 while ((stk
= dir
->internal
.exclude_stack
) != NULL
) {
1564 if (stk
->baselen
<= baselen
&&
1565 !strncmp(dir
->internal
.basebuf
.buf
, base
, stk
->baselen
))
1567 pl
= &group
->pl
[dir
->internal
.exclude_stack
->exclude_ix
];
1568 dir
->internal
.exclude_stack
= stk
->prev
;
1569 dir
->internal
.pattern
= NULL
;
1570 free((char *)pl
->src
); /* see strbuf_detach() below */
1571 clear_pattern_list(pl
);
1576 /* Skip traversing into sub directories if the parent is excluded */
1577 if (dir
->internal
.pattern
)
1581 * Lazy initialization. All call sites currently just
1582 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1583 * them seems lots of work for little benefit.
1585 if (!dir
->internal
.basebuf
.buf
)
1586 strbuf_init(&dir
->internal
.basebuf
, PATH_MAX
);
1588 /* Read from the parent directories and push them down. */
1589 current
= stk
? stk
->baselen
: -1;
1590 strbuf_setlen(&dir
->internal
.basebuf
, current
< 0 ? 0 : current
);
1592 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1596 while (current
< baselen
) {
1598 struct oid_stat oid_stat
;
1600 CALLOC_ARRAY(stk
, 1);
1605 cp
= strchr(base
+ current
+ 1, '/');
1607 die("oops in prep_exclude");
1610 lookup_untracked(dir
->untracked
,
1613 cp
- base
- current
);
1615 stk
->prev
= dir
->internal
.exclude_stack
;
1616 stk
->baselen
= cp
- base
;
1617 stk
->exclude_ix
= group
->nr
;
1618 stk
->ucd
= untracked
;
1619 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1620 strbuf_add(&dir
->internal
.basebuf
, base
+ current
, stk
->baselen
- current
);
1621 assert(stk
->baselen
== dir
->internal
.basebuf
.len
);
1623 /* Abort if the directory is excluded */
1626 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = 0;
1627 dir
->internal
.pattern
= last_matching_pattern_from_lists(dir
,
1629 dir
->internal
.basebuf
.buf
, stk
->baselen
- 1,
1630 dir
->internal
.basebuf
.buf
+ current
, &dt
);
1631 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = '/';
1632 if (dir
->internal
.pattern
&&
1633 dir
->internal
.pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1634 dir
->internal
.pattern
= NULL
;
1635 if (dir
->internal
.pattern
) {
1636 dir
->internal
.exclude_stack
= stk
;
1641 /* Try to read per-directory file */
1642 oidclr(&oid_stat
.oid
);
1644 if (dir
->exclude_per_dir
&&
1646 * If we know that no files have been added in
1647 * this directory (i.e. valid_cached_dir() has
1648 * been executed and set untracked->valid) ..
1650 (!untracked
|| !untracked
->valid
||
1652 * .. and .gitignore does not exist before
1653 * (i.e. null exclude_oid). Then we can skip
1654 * loading .gitignore, which would result in
1657 !is_null_oid(&untracked
->exclude_oid
))) {
1659 * dir->internal.basebuf gets reused by the traversal,
1660 * but we need fname to remain unchanged to ensure the
1661 * src member of each struct path_pattern correctly
1662 * back-references its source file. Other invocations
1663 * of add_pattern_list provide stable strings, so we
1664 * strbuf_detach() and free() here in the caller.
1666 struct strbuf sb
= STRBUF_INIT
;
1667 strbuf_addbuf(&sb
, &dir
->internal
.basebuf
);
1668 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1669 pl
->src
= strbuf_detach(&sb
, NULL
);
1670 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1672 untracked
? &oid_stat
: NULL
);
1675 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1676 * will first be called in valid_cached_dir() then maybe many
1677 * times more in last_matching_pattern(). When the cache is
1678 * used, last_matching_pattern() will not be called and
1679 * reading .gitignore content will be a waste.
1681 * So when it's called by valid_cached_dir() and we can get
1682 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1683 * modified on work tree), we could delay reading the
1684 * .gitignore content until we absolutely need it in
1685 * last_matching_pattern(). Be careful about ignore rule
1686 * order, though, if you do that.
1689 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1690 invalidate_gitignore(dir
->untracked
, untracked
);
1691 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1693 dir
->internal
.exclude_stack
= stk
;
1694 current
= stk
->baselen
;
1696 strbuf_setlen(&dir
->internal
.basebuf
, baselen
);
1700 * Loads the exclude lists for the directory containing pathname, then
1701 * scans all exclude lists to determine whether pathname is excluded.
1702 * Returns the exclude_list element which matched, or NULL for
1705 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1706 struct index_state
*istate
,
1707 const char *pathname
,
1710 int pathlen
= strlen(pathname
);
1711 const char *basename
= strrchr(pathname
, '/');
1712 basename
= (basename
) ? basename
+1 : pathname
;
1714 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1716 if (dir
->internal
.pattern
)
1717 return dir
->internal
.pattern
;
1719 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1724 * Loads the exclude lists for the directory containing pathname, then
1725 * scans all exclude lists to determine whether pathname is excluded.
1726 * Returns 1 if true, otherwise 0.
1728 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1729 const char *pathname
, int *dtype_p
)
1731 struct path_pattern
*pattern
=
1732 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1734 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1738 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1740 struct dir_entry
*ent
;
1742 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1747 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1748 struct index_state
*istate
,
1749 const char *pathname
, int len
)
1751 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1754 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->internal
.alloc
);
1755 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1758 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1759 struct index_state
*istate
,
1760 const char *pathname
, int len
)
1762 if (!index_name_is_other(istate
, pathname
, len
))
1765 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->internal
.ignored_alloc
);
1766 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1770 index_nonexistent
= 0,
1776 * Do not use the alphabetically sorted index to look up
1777 * the directory name; instead, use the case insensitive
1780 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1781 const char *dirname
, int len
)
1783 struct cache_entry
*ce
;
1785 if (index_dir_exists(istate
, dirname
, len
))
1786 return index_directory
;
1788 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1789 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1790 return index_gitdir
;
1792 return index_nonexistent
;
1796 * The index sorts alphabetically by entry name, which
1797 * means that a gitlink sorts as '\0' at the end, while
1798 * a directory (which is defined not as an entry, but as
1799 * the files it contains) will sort with the '/' at the
1802 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1803 const char *dirname
, int len
)
1808 return directory_exists_in_index_icase(istate
, dirname
, len
);
1810 pos
= index_name_pos(istate
, dirname
, len
);
1813 while (pos
< istate
->cache_nr
) {
1814 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1815 unsigned char endchar
;
1817 if (strncmp(ce
->name
, dirname
, len
))
1819 endchar
= ce
->name
[len
];
1823 return index_directory
;
1824 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1825 return index_gitdir
;
1827 return index_nonexistent
;
1831 * When we find a directory when traversing the filesystem, we
1832 * have three distinct cases:
1835 * - see it as a directory
1838 * and which one we choose depends on a combination of existing
1839 * git index contents and the flags passed into the directory
1840 * traversal routine.
1842 * Case 1: If we *already* have entries in the index under that
1843 * directory name, we always recurse into the directory to see
1846 * Case 2: If we *already* have that directory name as a gitlink,
1847 * we always continue to see it as a gitlink, regardless of whether
1848 * there is an actual git directory there or not (it might not
1849 * be checked out as a subproject!)
1851 * Case 3: if we didn't have it in the index previously, we
1852 * have a few sub-cases:
1854 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1855 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1856 * also true, in which case we need to check if it contains any
1857 * untracked and / or ignored files.
1858 * (b) if it looks like a git directory and we don't have the
1859 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1860 * show it as a directory.
1861 * (c) otherwise, we recurse into it.
1863 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1864 struct index_state
*istate
,
1865 struct untracked_cache_dir
*untracked
,
1866 const char *dirname
, int len
, int baselen
, int excluded
,
1867 const struct pathspec
*pathspec
)
1870 * WARNING: From this function, you can return path_recurse or you
1871 * can call read_directory_recursive() (or neither), but
1872 * you CAN'T DO BOTH.
1874 enum path_treatment state
;
1875 int matches_how
= 0;
1876 int check_only
, stop_early
;
1877 int old_ignored_nr
, old_untracked_nr
;
1878 /* The "len-1" is to strip the final '/' */
1879 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1881 if (status
== index_directory
)
1882 return path_recurse
;
1883 if (status
== index_gitdir
)
1885 if (status
!= index_nonexistent
)
1886 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1889 * We don't want to descend into paths that don't match the necessary
1890 * patterns. Clearly, if we don't have a pathspec, then we can't check
1891 * for matching patterns. Also, if (excluded) then we know we matched
1892 * the exclusion patterns so as an optimization we can skip checking
1893 * for matching patterns.
1895 if (pathspec
&& !excluded
) {
1896 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1900 DO_MATCH_LEADING_PATHSPEC
);
1906 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1907 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1909 * Determine if `dirname` is a nested repo by confirming that:
1910 * 1) we are in a nonbare repository, and
1911 * 2) `dirname` is not an immediate parent of `the_repository->gitdir`,
1912 * which could occur if the git_dir or worktree location was
1913 * manually configured by the user; see t2205 testcases 1-3 for
1914 * examples where this matters
1917 struct strbuf sb
= STRBUF_INIT
;
1918 strbuf_addstr(&sb
, dirname
);
1919 nested_repo
= is_nonbare_repository_dir(&sb
);
1922 char *real_dirname
, *real_gitdir
;
1923 strbuf_addstr(&sb
, ".git");
1924 real_dirname
= real_pathdup(sb
.buf
, 1);
1925 real_gitdir
= real_pathdup(the_repository
->gitdir
, 1);
1927 nested_repo
= !!strcmp(real_dirname
, real_gitdir
);
1931 strbuf_release(&sb
);
1934 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1935 (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
))
1937 return excluded
? path_excluded
: path_untracked
;
1941 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1943 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1944 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1947 * This is an excluded directory and we are
1948 * showing ignored paths that match an exclude
1949 * pattern. (e.g. show directory as ignored
1950 * only if it matches an exclude pattern).
1951 * This path will either be 'path_excluded`
1952 * (if we are showing empty directories or if
1953 * the directory is not empty), or will be
1954 * 'path_none' (empty directory, and we are
1955 * not showing empty directories).
1957 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1958 return path_excluded
;
1960 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1961 untracked
, 1, 1, pathspec
) == path_excluded
)
1962 return path_excluded
;
1966 return path_recurse
;
1969 assert(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
);
1972 * If we have a pathspec which could match something _below_ this
1973 * directory (e.g. when checking 'subdir/' having a pathspec like
1974 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1977 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1978 return path_recurse
;
1980 /* Special cases for where this directory is excluded/ignored */
1983 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1984 * hiding empty directories, there is no need to
1985 * recurse into an ignored directory.
1987 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1988 return path_excluded
;
1991 * Even if we are hiding empty directories, we can still avoid
1992 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
1993 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1995 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1996 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
1997 return path_excluded
;
2001 * Other than the path_recurse case above, we only need to
2002 * recurse into untracked directories if any of the following
2004 * - DIR_SHOW_IGNORED (because then we need to determine if
2005 * there are ignored entries below)
2006 * - DIR_SHOW_IGNORED_TOO (same as above)
2007 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
2008 * the directory is empty)
2011 !(dir
->flags
& (DIR_SHOW_IGNORED
|
2012 DIR_SHOW_IGNORED_TOO
|
2013 DIR_HIDE_EMPTY_DIRECTORIES
))) {
2014 return path_untracked
;
2018 * Even if we don't want to know all the paths under an untracked or
2019 * ignored directory, we may still need to go into the directory to
2020 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2021 * an empty directory should be path_none instead of path_excluded or
2024 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
2025 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
2028 * However, there's another optimization possible as a subset of
2029 * check_only, based on the cases we have to consider:
2030 * A) Directory matches no exclude patterns:
2031 * * Directory is empty => path_none
2032 * * Directory has an untracked file under it => path_untracked
2033 * * Directory has only ignored files under it => path_excluded
2034 * B) Directory matches an exclude pattern:
2035 * * Directory is empty => path_none
2036 * * Directory has an untracked file under it => path_excluded
2037 * * Directory has only ignored files under it => path_excluded
2038 * In case A, we can exit as soon as we've found an untracked
2039 * file but otherwise have to walk all files. In case B, though,
2040 * we can stop at the first file we find under the directory.
2042 stop_early
= check_only
&& excluded
;
2045 * If /every/ file within an untracked directory is ignored, then
2046 * we want to treat the directory as ignored (for e.g. status
2047 * --porcelain), without listing the individual ignored files
2048 * underneath. To do so, we'll save the current ignored_nr, and
2049 * pop all the ones added after it if it turns out the entire
2050 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2051 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2052 * untracked paths so will need to pop all those off the last
2053 * after we traverse.
2055 old_ignored_nr
= dir
->ignored_nr
;
2056 old_untracked_nr
= dir
->nr
;
2058 /* Actually recurse into dirname now, we'll fixup the state later. */
2059 untracked
= lookup_untracked(dir
->untracked
, untracked
,
2060 dirname
+ baselen
, len
- baselen
);
2061 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
2062 check_only
, stop_early
, pathspec
);
2064 /* There are a variety of reasons we may need to fixup the state... */
2065 if (state
== path_excluded
) {
2066 /* state == path_excluded implies all paths under
2067 * dirname were ignored...
2069 * if running e.g. `git status --porcelain --ignored=matching`,
2070 * then we want to see the subpaths that are ignored.
2072 * if running e.g. just `git status --porcelain`, then
2073 * we just want the directory itself to be listed as ignored
2074 * and not the individual paths underneath.
2076 int want_ignored_subpaths
=
2077 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2078 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
2080 if (want_ignored_subpaths
) {
2082 * with --ignored=matching, we want the subpaths
2083 * INSTEAD of the directory itself.
2088 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
2089 FREE_AND_NULL(dir
->ignored
[i
]);
2090 dir
->ignored_nr
= old_ignored_nr
;
2095 * We may need to ignore some of the untracked paths we found while
2096 * traversing subdirectories.
2098 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2099 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
2101 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
2102 FREE_AND_NULL(dir
->entries
[i
]);
2103 dir
->nr
= old_untracked_nr
;
2107 * If there is nothing under the current directory and we are not
2108 * hiding empty directories, then we need to report on the
2109 * untracked or ignored status of the directory itself.
2111 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2112 state
= excluded
? path_excluded
: path_untracked
;
2118 * This is an inexact early pruning of any recursive directory
2119 * reading - if the path cannot possibly be in the pathspec,
2120 * return true, and we'll skip it early.
2122 static int simplify_away(const char *path
, int pathlen
,
2123 const struct pathspec
*pathspec
)
2127 if (!pathspec
|| !pathspec
->nr
)
2130 GUARD_PATHSPEC(pathspec
,
2139 for (i
= 0; i
< pathspec
->nr
; i
++) {
2140 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2141 int len
= item
->nowildcard_len
;
2145 if (!ps_strncmp(item
, item
->match
, path
, len
))
2153 * This function tells us whether an excluded path matches a
2154 * list of "interesting" pathspecs. That is, whether a path matched
2155 * by any of the pathspecs could possibly be ignored by excluding
2156 * the specified path. This can happen if:
2158 * 1. the path is mentioned explicitly in the pathspec
2160 * 2. the path is a directory prefix of some element in the
2163 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2164 const struct pathspec
*pathspec
)
2168 if (!pathspec
|| !pathspec
->nr
)
2171 GUARD_PATHSPEC(pathspec
,
2179 for (i
= 0; i
< pathspec
->nr
; i
++) {
2180 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2181 int len
= item
->nowildcard_len
;
2183 if (len
== pathlen
&&
2184 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2186 if (len
> pathlen
&&
2187 item
->match
[pathlen
] == '/' &&
2188 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2194 static int get_index_dtype(struct index_state
*istate
,
2195 const char *path
, int len
)
2198 const struct cache_entry
*ce
;
2200 ce
= index_file_exists(istate
, path
, len
, 0);
2202 if (!ce_uptodate(ce
))
2204 if (S_ISGITLINK(ce
->ce_mode
))
2207 * Nobody actually cares about the
2208 * difference between DT_LNK and DT_REG
2213 /* Try to look it up as a directory */
2214 pos
= index_name_pos(istate
, path
, len
);
2218 while (pos
< istate
->cache_nr
) {
2219 ce
= istate
->cache
[pos
++];
2220 if (strncmp(ce
->name
, path
, len
))
2222 if (ce
->name
[len
] > '/')
2224 if (ce
->name
[len
] < '/')
2226 if (!ce_uptodate(ce
))
2227 break; /* continue? */
2233 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2234 const char *path
, int len
)
2238 if (dtype
!= DT_UNKNOWN
)
2240 dtype
= get_index_dtype(istate
, path
, len
);
2241 if (dtype
!= DT_UNKNOWN
)
2243 if (lstat(path
, &st
))
2245 if (S_ISREG(st
.st_mode
))
2247 if (S_ISDIR(st
.st_mode
))
2249 if (S_ISLNK(st
.st_mode
))
2254 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2255 struct cached_dir
*cdir
,
2256 struct index_state
*istate
,
2257 struct strbuf
*path
,
2259 const struct pathspec
*pathspec
)
2262 * WARNING: From this function, you can return path_recurse or you
2263 * can call read_directory_recursive() (or neither), but
2264 * you CAN'T DO BOTH.
2266 strbuf_setlen(path
, baselen
);
2268 strbuf_addstr(path
, cdir
->file
);
2269 return path_untracked
;
2271 strbuf_addstr(path
, cdir
->ucd
->name
);
2272 /* treat_one_path() does this before it calls treat_directory() */
2273 strbuf_complete(path
, '/');
2274 if (cdir
->ucd
->check_only
)
2276 * check_only is set as a result of treat_directory() getting
2277 * to its bottom. Verify again the same set of directories
2278 * with check_only set.
2280 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2281 cdir
->ucd
, 1, 0, pathspec
);
2283 * We get path_recurse in the first run when
2284 * directory_exists_in_index() returns index_nonexistent. We
2285 * are sure that new changes in the index does not impact the
2286 * outcome. Return now.
2288 return path_recurse
;
2291 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2292 struct untracked_cache_dir
*untracked
,
2293 struct cached_dir
*cdir
,
2294 struct index_state
*istate
,
2295 struct strbuf
*path
,
2297 const struct pathspec
*pathspec
)
2299 int has_path_in_index
, dtype
, excluded
;
2302 return treat_path_fast(dir
, cdir
, istate
, path
,
2304 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2306 strbuf_setlen(path
, baselen
);
2307 strbuf_addstr(path
, cdir
->d_name
);
2308 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2311 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2313 /* Always exclude indexed files */
2314 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2316 if (dtype
!= DT_DIR
&& has_path_in_index
)
2320 * When we are looking at a directory P in the working tree,
2321 * there are three cases:
2323 * (1) P exists in the index. Everything inside the directory P in
2324 * the working tree needs to go when P is checked out from the
2327 * (2) P does not exist in the index, but there is P/Q in the index.
2328 * We know P will stay a directory when we check out the contents
2329 * of the index, but we do not know yet if there is a directory
2330 * P/Q in the working tree to be killed, so we need to recurse.
2332 * (3) P does not exist in the index, and there is no P/Q in the index
2333 * to require P to be a directory, either. Only in this case, we
2334 * know that everything inside P will not be killed without
2337 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2338 (dtype
== DT_DIR
) &&
2339 !has_path_in_index
&&
2340 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2343 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2346 * Excluded? If we don't explicitly want to show
2347 * ignored files, ignore it
2349 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2350 return path_excluded
;
2357 * WARNING: Do not ignore/amend the return value from
2358 * treat_directory(), and especially do not change it to return
2359 * path_recurse as that can cause exponential slowdown.
2360 * Instead, modify treat_directory() to return the right value.
2362 strbuf_addch(path
, '/');
2363 return treat_directory(dir
, istate
, untracked
,
2364 path
->buf
, path
->len
,
2365 baselen
, excluded
, pathspec
);
2369 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2370 0 /* prefix */, NULL
/* seen */,
2374 return path_excluded
;
2375 return path_untracked
;
2379 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2383 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2384 dir
->untracked_alloc
);
2385 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2388 static int valid_cached_dir(struct dir_struct
*dir
,
2389 struct untracked_cache_dir
*untracked
,
2390 struct index_state
*istate
,
2391 struct strbuf
*path
,
2400 * With fsmonitor, we can trust the untracked cache's valid field.
2402 refresh_fsmonitor(istate
);
2403 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2404 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2405 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2408 if (!untracked
->valid
||
2409 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2410 fill_stat_data(&untracked
->stat_data
, &st
);
2415 if (untracked
->check_only
!= !!check_only
)
2419 * prep_exclude will be called eventually on this directory,
2420 * but it's called much later in last_matching_pattern(). We
2421 * need it now to determine the validity of the cache for this
2422 * path. The next calls will be nearly no-op, the way
2423 * prep_exclude() is designed.
2425 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2426 strbuf_addch(path
, '/');
2427 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2428 strbuf_setlen(path
, path
->len
- 1);
2430 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2432 /* hopefully prep_exclude() haven't invalidated this entry... */
2433 return untracked
->valid
;
2436 static int open_cached_dir(struct cached_dir
*cdir
,
2437 struct dir_struct
*dir
,
2438 struct untracked_cache_dir
*untracked
,
2439 struct index_state
*istate
,
2440 struct strbuf
*path
,
2445 memset(cdir
, 0, sizeof(*cdir
));
2446 cdir
->untracked
= untracked
;
2447 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2449 c_path
= path
->len
? path
->buf
: ".";
2450 cdir
->fdir
= opendir(c_path
);
2452 warning_errno(_("could not open directory '%s'"), c_path
);
2453 if (dir
->untracked
) {
2454 invalidate_directory(dir
->untracked
, untracked
);
2455 dir
->untracked
->dir_opened
++;
2462 static int read_cached_dir(struct cached_dir
*cdir
)
2467 de
= readdir_skip_dot_and_dotdot(cdir
->fdir
);
2469 cdir
->d_name
= NULL
;
2470 cdir
->d_type
= DT_UNKNOWN
;
2473 cdir
->d_name
= de
->d_name
;
2474 cdir
->d_type
= DTYPE(de
);
2477 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2478 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2488 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2489 struct untracked_cache_dir
*d
= cdir
->untracked
;
2490 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2496 static void close_cached_dir(struct cached_dir
*cdir
)
2499 closedir(cdir
->fdir
);
2501 * We have gone through this directory and found no untracked
2502 * entries. Mark it valid.
2504 if (cdir
->untracked
) {
2505 cdir
->untracked
->valid
= 1;
2506 cdir
->untracked
->recurse
= 1;
2510 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2511 struct untracked_cache_dir
*untracked
,
2512 struct cached_dir
*cdir
,
2513 struct index_state
*istate
,
2514 struct strbuf
*path
,
2516 const struct pathspec
*pathspec
,
2517 enum path_treatment state
)
2519 /* add the path to the appropriate result list */
2522 if (dir
->flags
& DIR_SHOW_IGNORED
)
2523 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2524 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2525 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2526 exclude_matches_pathspec(path
->buf
, path
->len
,
2528 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2531 case path_untracked
:
2532 if (dir
->flags
& DIR_SHOW_IGNORED
)
2534 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2536 add_untracked(untracked
, path
->buf
+ baselen
);
2545 * Read a directory tree. We currently ignore anything but
2546 * directories, regular files and symlinks. That's because git
2547 * doesn't handle them at all yet. Maybe that will change some
2550 * Also, we ignore the name ".git" (even if it is not a directory).
2551 * That likely will not change.
2553 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2554 * to signal that a file was found. This is the least significant value that
2555 * indicates that a file was encountered that does not depend on the order of
2556 * whether an untracked or excluded path was encountered first.
2558 * Returns the most significant path_treatment value encountered in the scan.
2559 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2560 * significant path_treatment value that will be returned.
2563 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2564 struct index_state
*istate
, const char *base
, int baselen
,
2565 struct untracked_cache_dir
*untracked
, int check_only
,
2566 int stop_at_first_file
, const struct pathspec
*pathspec
)
2569 * WARNING: Do NOT recurse unless path_recurse is returned from
2570 * treat_path(). Recursing on any other return value
2571 * can result in exponential slowdown.
2573 struct cached_dir cdir
;
2574 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2575 struct strbuf path
= STRBUF_INIT
;
2577 strbuf_add(&path
, base
, baselen
);
2579 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2581 dir
->internal
.visited_directories
++;
2584 untracked
->check_only
= !!check_only
;
2586 while (!read_cached_dir(&cdir
)) {
2587 /* check how the file or directory should be treated */
2588 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2590 dir
->internal
.visited_paths
++;
2592 if (state
> dir_state
)
2595 /* recurse into subdir if instructed by treat_path */
2596 if (state
== path_recurse
) {
2597 struct untracked_cache_dir
*ud
;
2598 ud
= lookup_untracked(dir
->untracked
,
2601 path
.len
- baselen
);
2603 read_directory_recursive(dir
, istate
, path
.buf
,
2605 check_only
, stop_at_first_file
, pathspec
);
2606 if (subdir_state
> dir_state
)
2607 dir_state
= subdir_state
;
2610 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2611 0 /* prefix */, NULL
,
2612 0 /* do NOT special case dirs */))
2617 if (stop_at_first_file
) {
2619 * If stopping at first file, then
2620 * signal that a file was found by
2621 * returning `path_excluded`. This is
2622 * to return a consistent value
2623 * regardless of whether an ignored or
2624 * excluded file happened to be
2627 * In current usage, the
2628 * `stop_at_first_file` is passed when
2629 * an ancestor directory has matched
2630 * an exclude pattern, so any found
2631 * files will be excluded.
2633 if (dir_state
>= path_excluded
) {
2634 dir_state
= path_excluded
;
2639 /* abort early if maximum state has been reached */
2640 if (dir_state
== path_untracked
) {
2642 add_untracked(untracked
, path
.buf
+ baselen
);
2645 /* skip the add_path_to_appropriate_result_list() */
2649 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2650 istate
, &path
, baselen
,
2653 close_cached_dir(&cdir
);
2655 strbuf_release(&path
);
2660 int cmp_dir_entry(const void *p1
, const void *p2
)
2662 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2663 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2665 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2668 /* check if *out lexically strictly contains *in */
2669 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2671 return (out
->len
< in
->len
) &&
2672 (out
->name
[out
->len
- 1] == '/') &&
2673 !memcmp(out
->name
, in
->name
, out
->len
);
2676 static int treat_leading_path(struct dir_struct
*dir
,
2677 struct index_state
*istate
,
2678 const char *path
, int len
,
2679 const struct pathspec
*pathspec
)
2681 struct strbuf sb
= STRBUF_INIT
;
2682 struct strbuf subdir
= STRBUF_INIT
;
2683 int prevlen
, baselen
;
2685 struct cached_dir cdir
;
2686 enum path_treatment state
= path_none
;
2689 * For each directory component of path, we are going to check whether
2690 * that path is relevant given the pathspec. For example, if path is
2692 * then we will ask treat_path() whether we should go into foo, then
2693 * whether we should go into bar, then whether baz is relevant.
2694 * Checking each is important because e.g. if path is
2696 * then we need to check .git to know we shouldn't traverse it.
2697 * If the return from treat_path() is:
2698 * * path_none, for any path, we return false.
2699 * * path_recurse, for all path components, we return true
2700 * * <anything else> for some intermediate component, we make sure
2701 * to add that path to the relevant list but return false
2702 * signifying that we shouldn't recurse into it.
2705 while (len
&& path
[len
- 1] == '/')
2710 memset(&cdir
, 0, sizeof(cdir
));
2711 cdir
.d_type
= DT_DIR
;
2715 prevlen
= baselen
+ !!baselen
;
2716 cp
= path
+ prevlen
;
2717 cp
= memchr(cp
, '/', path
+ len
- cp
);
2721 baselen
= cp
- path
;
2723 strbuf_add(&sb
, path
, baselen
);
2724 if (!is_directory(sb
.buf
))
2727 strbuf_add(&sb
, path
, prevlen
);
2728 strbuf_reset(&subdir
);
2729 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2730 cdir
.d_name
= subdir
.buf
;
2731 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2733 if (state
!= path_recurse
)
2734 break; /* do not recurse into it */
2736 break; /* finished checking */
2738 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2739 &sb
, baselen
, pathspec
,
2742 strbuf_release(&subdir
);
2743 strbuf_release(&sb
);
2744 return state
== path_recurse
;
2747 static const char *get_ident_string(void)
2749 static struct strbuf sb
= STRBUF_INIT
;
2754 if (uname(&uts
) < 0)
2755 die_errno(_("failed to get kernel name and information"));
2756 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2761 static int ident_in_untracked(const struct untracked_cache
*uc
)
2764 * Previous git versions may have saved many NUL separated
2765 * strings in the "ident" field, but it is insane to manage
2766 * many locations, so just take care of the first one.
2769 return !strcmp(uc
->ident
.buf
, get_ident_string());
2772 static void set_untracked_ident(struct untracked_cache
*uc
)
2774 strbuf_reset(&uc
->ident
);
2775 strbuf_addstr(&uc
->ident
, get_ident_string());
2778 * This strbuf used to contain a list of NUL separated
2779 * strings, so save NUL too for backward compatibility.
2781 strbuf_addch(&uc
->ident
, 0);
2784 static unsigned new_untracked_cache_flags(struct index_state
*istate
)
2786 struct repository
*repo
= istate
->repo
;
2790 * This logic is coordinated with the setting of these flags in
2791 * wt-status.c#wt_status_collect_untracked(), and the evaluation
2792 * of the config setting in commit.c#git_status_config()
2794 if (!repo_config_get_string(repo
, "status.showuntrackedfiles", &val
) &&
2795 !strcmp(val
, "all"))
2799 * The default, if "all" is not set, is "normal" - leading us here.
2800 * If the value is "none" then it really doesn't matter.
2802 return DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2805 static void new_untracked_cache(struct index_state
*istate
, int flags
)
2807 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2808 strbuf_init(&uc
->ident
, 100);
2809 uc
->exclude_per_dir
= ".gitignore";
2810 uc
->dir_flags
= flags
>= 0 ? flags
: new_untracked_cache_flags(istate
);
2811 set_untracked_ident(uc
);
2812 istate
->untracked
= uc
;
2813 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2816 void add_untracked_cache(struct index_state
*istate
)
2818 if (!istate
->untracked
) {
2819 new_untracked_cache(istate
, -1);
2821 if (!ident_in_untracked(istate
->untracked
)) {
2822 free_untracked_cache(istate
->untracked
);
2823 new_untracked_cache(istate
, -1);
2828 void remove_untracked_cache(struct index_state
*istate
)
2830 if (istate
->untracked
) {
2831 free_untracked_cache(istate
->untracked
);
2832 istate
->untracked
= NULL
;
2833 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2837 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2839 const struct pathspec
*pathspec
,
2840 struct index_state
*istate
)
2842 struct untracked_cache_dir
*root
;
2843 static int untracked_cache_disabled
= -1;
2845 if (!dir
->untracked
)
2847 if (untracked_cache_disabled
< 0)
2848 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2849 if (untracked_cache_disabled
)
2853 * We only support $GIT_DIR/info/exclude and core.excludesfile
2854 * as the global ignore rule files. Any other additions
2855 * (e.g. from command line) invalidate the cache. This
2856 * condition also catches running setup_standard_excludes()
2857 * before setting dir->untracked!
2859 if (dir
->internal
.unmanaged_exclude_files
)
2863 * Optimize for the main use case only: whole-tree git
2864 * status. More work involved in treat_leading_path() if we
2865 * use cache on just a subset of the worktree. pathspec
2866 * support could make the matter even worse.
2868 if (base_len
|| (pathspec
&& pathspec
->nr
))
2871 /* We don't support collecting ignore files */
2872 if (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2873 DIR_COLLECT_IGNORED
))
2877 * If we use .gitignore in the cache and now you change it to
2878 * .gitexclude, everything will go wrong.
2880 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2881 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2885 * EXC_CMDL is not considered in the cache. If people set it,
2888 if (dir
->internal
.exclude_list_group
[EXC_CMDL
].nr
)
2891 if (!ident_in_untracked(dir
->untracked
)) {
2892 warning(_("untracked cache is disabled on this system or location"));
2897 * If the untracked structure we received does not have the same flags
2898 * as requested in this run, we're going to need to either discard the
2899 * existing structure (and potentially later recreate), or bypass the
2900 * untracked cache mechanism for this run.
2902 if (dir
->flags
!= dir
->untracked
->dir_flags
) {
2904 * If the untracked structure we received does not have the same flags
2905 * as configured, then we need to reset / create a new "untracked"
2906 * structure to match the new config.
2908 * Keeping the saved and used untracked cache consistent with the
2909 * configuration provides an opportunity for frequent users of
2910 * "git status -uall" to leverage the untracked cache by aligning their
2911 * configuration - setting "status.showuntrackedfiles" to "all" or
2912 * "normal" as appropriate.
2914 * Previously using -uall (or setting "status.showuntrackedfiles" to
2915 * "all") was incompatible with untracked cache and *consistently*
2916 * caused surprisingly bad performance (with fscache and fsmonitor
2917 * enabled) on Windows.
2919 * IMPROVEMENT OPPORTUNITY: If we reworked the untracked cache storage
2920 * to not be as bound up with the desired output in a given run,
2921 * and instead iterated through and stored enough information to
2922 * correctly serve both "modes", then users could get peak performance
2923 * with or without '-uall' regardless of their
2924 * "status.showuntrackedfiles" config.
2926 if (dir
->untracked
->dir_flags
!= new_untracked_cache_flags(istate
)) {
2927 free_untracked_cache(istate
->untracked
);
2928 new_untracked_cache(istate
, dir
->flags
);
2929 dir
->untracked
= istate
->untracked
;
2933 * Current untracked cache data is consistent with config, but not
2934 * usable in this request/run; just bypass untracked cache.
2940 if (!dir
->untracked
->root
) {
2941 /* Untracked cache existed but is not initialized; fix that */
2942 FLEX_ALLOC_STR(dir
->untracked
->root
, name
, "");
2943 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2946 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2947 root
= dir
->untracked
->root
;
2948 if (!oideq(&dir
->internal
.ss_info_exclude
.oid
,
2949 &dir
->untracked
->ss_info_exclude
.oid
)) {
2950 invalidate_gitignore(dir
->untracked
, root
);
2951 dir
->untracked
->ss_info_exclude
= dir
->internal
.ss_info_exclude
;
2953 if (!oideq(&dir
->internal
.ss_excludes_file
.oid
,
2954 &dir
->untracked
->ss_excludes_file
.oid
)) {
2955 invalidate_gitignore(dir
->untracked
, root
);
2956 dir
->untracked
->ss_excludes_file
= dir
->internal
.ss_excludes_file
;
2959 /* Make sure this directory is not dropped out at saving phase */
2964 static void emit_traversal_statistics(struct dir_struct
*dir
,
2965 struct repository
*repo
,
2969 if (!trace2_is_enabled())
2973 trace2_data_string("read_directory", repo
, "path", "");
2975 struct strbuf tmp
= STRBUF_INIT
;
2976 strbuf_add(&tmp
, path
, path_len
);
2977 trace2_data_string("read_directory", repo
, "path", tmp
.buf
);
2978 strbuf_release(&tmp
);
2981 trace2_data_intmax("read_directory", repo
,
2982 "directories-visited", dir
->internal
.visited_directories
);
2983 trace2_data_intmax("read_directory", repo
,
2984 "paths-visited", dir
->internal
.visited_paths
);
2986 if (!dir
->untracked
)
2988 trace2_data_intmax("read_directory", repo
,
2989 "node-creation", dir
->untracked
->dir_created
);
2990 trace2_data_intmax("read_directory", repo
,
2991 "gitignore-invalidation",
2992 dir
->untracked
->gitignore_invalidated
);
2993 trace2_data_intmax("read_directory", repo
,
2994 "directory-invalidation",
2995 dir
->untracked
->dir_invalidated
);
2996 trace2_data_intmax("read_directory", repo
,
2997 "opendir", dir
->untracked
->dir_opened
);
3000 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
3001 const char *path
, int len
, const struct pathspec
*pathspec
)
3003 struct untracked_cache_dir
*untracked
;
3005 trace2_region_enter("dir", "read_directory", istate
->repo
);
3006 dir
->internal
.visited_paths
= 0;
3007 dir
->internal
.visited_directories
= 0;
3009 if (has_symlink_leading_path(path
, len
)) {
3010 trace2_region_leave("dir", "read_directory", istate
->repo
);
3014 untracked
= validate_untracked_cache(dir
, len
, pathspec
, istate
);
3017 * make sure untracked cache code path is disabled,
3018 * e.g. prep_exclude()
3020 dir
->untracked
= NULL
;
3021 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
3022 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
3023 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
3024 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
3026 emit_traversal_statistics(dir
, istate
->repo
, path
, len
);
3028 trace2_region_leave("dir", "read_directory", istate
->repo
);
3029 if (dir
->untracked
) {
3030 static int force_untracked_cache
= -1;
3032 if (force_untracked_cache
< 0)
3033 force_untracked_cache
=
3034 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", -1);
3035 if (force_untracked_cache
< 0)
3036 force_untracked_cache
= (istate
->repo
->settings
.core_untracked_cache
== UNTRACKED_CACHE_WRITE
);
3037 if (force_untracked_cache
&&
3038 dir
->untracked
== istate
->untracked
&&
3039 (dir
->untracked
->dir_opened
||
3040 dir
->untracked
->gitignore_invalidated
||
3041 dir
->untracked
->dir_invalidated
))
3042 istate
->cache_changed
|= UNTRACKED_CHANGED
;
3043 if (dir
->untracked
!= istate
->untracked
) {
3044 FREE_AND_NULL(dir
->untracked
);
3051 int file_exists(const char *f
)
3054 return lstat(f
, &sb
) == 0;
3057 int repo_file_exists(struct repository
*repo
, const char *path
)
3059 if (repo
!= the_repository
)
3060 BUG("do not know how to check file existence in arbitrary repo");
3062 return file_exists(path
);
3065 static int cmp_icase(char a
, char b
)
3070 return toupper(a
) - toupper(b
);
3075 * Given two normalized paths (a trailing slash is ok), if subdir is
3076 * outside dir, return -1. Otherwise return the offset in subdir that
3077 * can be used as relative path to dir.
3079 int dir_inside_of(const char *subdir
, const char *dir
)
3083 assert(dir
&& subdir
&& *dir
&& *subdir
);
3085 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
3091 /* hel[p]/me vs hel[l]/yeah */
3092 if (*dir
&& *subdir
)
3096 return !*dir
? offset
: -1; /* same dir */
3098 /* foo/[b]ar vs foo/[] */
3099 if (is_dir_sep(dir
[-1]))
3100 return is_dir_sep(subdir
[-1]) ? offset
: -1;
3102 /* foo[/]bar vs foo[] */
3103 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
3106 int is_inside_dir(const char *dir
)
3115 rc
= (dir_inside_of(cwd
, dir
) >= 0);
3120 int is_empty_dir(const char *path
)
3122 DIR *dir
= opendir(path
);
3129 e
= readdir_skip_dot_and_dotdot(dir
);
3137 char *git_url_basename(const char *repo
, int is_bundle
, int is_bare
)
3139 const char *end
= repo
+ strlen(repo
), *start
, *ptr
;
3146 start
= strstr(repo
, "://");
3153 * Skip authentication data. The stripping does happen
3154 * greedily, such that we strip up to the last '@' inside
3157 for (ptr
= start
; ptr
< end
&& !is_dir_sep(*ptr
); ptr
++) {
3163 * Strip trailing spaces, slashes and /.git
3165 while (start
< end
&& (is_dir_sep(end
[-1]) || isspace(end
[-1])))
3167 if (end
- start
> 5 && is_dir_sep(end
[-5]) &&
3168 !strncmp(end
- 4, ".git", 4)) {
3170 while (start
< end
&& is_dir_sep(end
[-1]))
3175 * It should not be possible to overflow `ptrdiff_t` by passing in an
3176 * insanely long URL, but GCC does not know that and will complain
3177 * without this check.
3179 if (end
- start
< 0)
3180 die(_("No directory name could be guessed.\n"
3181 "Please specify a directory on the command line"));
3184 * Strip trailing port number if we've got only a
3185 * hostname (that is, there is no dir separator but a
3186 * colon). This check is required such that we do not
3187 * strip URI's like '/foo/bar:2222.git', which should
3188 * result in a dir '2222' being guessed due to backwards
3191 if (memchr(start
, '/', end
- start
) == NULL
3192 && memchr(start
, ':', end
- start
) != NULL
) {
3194 while (start
< ptr
&& isdigit(ptr
[-1]) && ptr
[-1] != ':')
3196 if (start
< ptr
&& ptr
[-1] == ':')
3201 * Find last component. To remain backwards compatible we
3202 * also regard colons as path separators, such that
3203 * cloning a repository 'foo:bar.git' would result in a
3204 * directory 'bar' being guessed.
3207 while (start
< ptr
&& !is_dir_sep(ptr
[-1]) && ptr
[-1] != ':')
3212 * Strip .{bundle,git}.
3215 strip_suffix_mem(start
, &len
, is_bundle
? ".bundle" : ".git");
3217 if (!len
|| (len
== 1 && *start
== '/'))
3218 die(_("No directory name could be guessed.\n"
3219 "Please specify a directory on the command line"));
3222 dir
= xstrfmt("%.*s.git", (int)len
, start
);
3224 dir
= xstrndup(start
, len
);
3226 * Replace sequences of 'control' characters and whitespace
3227 * with one ascii space, remove leading and trailing spaces.
3231 int prev_space
= 1 /* strip leading whitespace */;
3232 for (end
= dir
; *end
; ++end
) {
3234 if ((unsigned char)ch
< '\x20')
3245 if (out
> dir
&& prev_space
)
3251 void strip_dir_trailing_slashes(char *dir
)
3253 char *end
= dir
+ strlen(dir
);
3255 while (dir
< end
- 1 && is_dir_sep(end
[-1]))
3260 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
3264 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
3265 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
3266 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
3267 int purge_original_cwd
= (flag
& REMOVE_DIR_PURGE_ORIGINAL_CWD
);
3268 struct object_id submodule_head
;
3270 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
3271 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
3272 /* Do not descend and nuke a nested git work tree. */
3278 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
3279 dir
= opendir(path
->buf
);
3281 if (errno
== ENOENT
)
3282 return keep_toplevel
? -1 : 0;
3283 else if (errno
== EACCES
&& !keep_toplevel
)
3285 * An empty dir could be removable even if it
3288 return rmdir(path
->buf
);
3292 strbuf_complete(path
, '/');
3295 while ((e
= readdir_skip_dot_and_dotdot(dir
)) != NULL
) {
3298 strbuf_setlen(path
, len
);
3299 strbuf_addstr(path
, e
->d_name
);
3300 if (lstat(path
->buf
, &st
)) {
3301 if (errno
== ENOENT
)
3303 * file disappeared, which is what we
3308 } else if (S_ISDIR(st
.st_mode
)) {
3309 if (!remove_dir_recurse(path
, flag
, &kept_down
))
3310 continue; /* happy */
3311 } else if (!only_empty
&&
3312 (!unlink(path
->buf
) || errno
== ENOENT
)) {
3313 continue; /* happy, too */
3316 /* path too long, stat fails, or non-directory still exists */
3322 strbuf_setlen(path
, original_len
);
3323 if (!ret
&& !keep_toplevel
&& !kept_down
) {
3324 if (!purge_original_cwd
&&
3325 startup_info
->original_cwd
&&
3326 !strcmp(startup_info
->original_cwd
, path
->buf
))
3327 ret
= -1; /* Do not remove current working directory */
3329 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
3332 * report the uplevel that it is not an error that we
3333 * did not rmdir() our directory.
3339 int remove_dir_recursively(struct strbuf
*path
, int flag
)
3341 return remove_dir_recurse(path
, flag
, NULL
);
3344 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
3346 void setup_standard_excludes(struct dir_struct
*dir
)
3348 dir
->exclude_per_dir
= ".gitignore";
3350 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3352 excludes_file
= xdg_config_home("ignore");
3353 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
3354 add_patterns_from_file_1(dir
, excludes_file
,
3355 dir
->untracked
? &dir
->internal
.ss_excludes_file
: NULL
);
3357 /* per repository user preference */
3358 if (startup_info
->have_repository
) {
3359 const char *path
= git_path_info_exclude();
3360 if (!access_or_warn(path
, R_OK
, 0))
3361 add_patterns_from_file_1(dir
, path
,
3362 dir
->untracked
? &dir
->internal
.ss_info_exclude
: NULL
);
3366 char *get_sparse_checkout_filename(void)
3368 return git_pathdup("info/sparse-checkout");
3371 int get_sparse_checkout_patterns(struct pattern_list
*pl
)
3374 char *sparse_filename
= get_sparse_checkout_filename();
3376 pl
->use_cone_patterns
= core_sparse_checkout_cone
;
3377 res
= add_patterns_from_file_to_list(sparse_filename
, "", 0, pl
, NULL
, 0);
3379 free(sparse_filename
);
3383 int remove_path(const char *name
)
3387 if (unlink(name
) && !is_missing_file_error(errno
))
3390 slash
= strrchr(name
, '/');
3392 char *dirs
= xstrdup(name
);
3393 slash
= dirs
+ (slash
- name
);
3396 if (startup_info
->original_cwd
&&
3397 !strcmp(startup_info
->original_cwd
, dirs
))
3399 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3406 * Frees memory within dir which was allocated, and resets fields for further
3407 * use. Does not free dir itself.
3409 void dir_clear(struct dir_struct
*dir
)
3412 struct exclude_list_group
*group
;
3413 struct pattern_list
*pl
;
3414 struct exclude_stack
*stk
;
3415 struct dir_struct
new = DIR_INIT
;
3417 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3418 group
= &dir
->internal
.exclude_list_group
[i
];
3419 for (j
= 0; j
< group
->nr
; j
++) {
3422 free((char *)pl
->src
);
3423 clear_pattern_list(pl
);
3428 for (i
= 0; i
< dir
->ignored_nr
; i
++)
3429 free(dir
->ignored
[i
]);
3430 for (i
= 0; i
< dir
->nr
; i
++)
3431 free(dir
->entries
[i
]);
3435 stk
= dir
->internal
.exclude_stack
;
3437 struct exclude_stack
*prev
= stk
->prev
;
3441 strbuf_release(&dir
->internal
.basebuf
);
3443 memcpy(dir
, &new, sizeof(*dir
));
3446 struct ondisk_untracked_cache
{
3447 struct stat_data info_exclude_stat
;
3448 struct stat_data excludes_file_stat
;
3452 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3455 int index
; /* number of written untracked_cache_dir */
3456 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3457 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3458 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3460 struct strbuf sb_stat
;
3461 struct strbuf sb_sha1
;
3464 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3466 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3467 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3468 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3469 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3470 to
->sd_dev
= htonl(from
->sd_dev
);
3471 to
->sd_ino
= htonl(from
->sd_ino
);
3472 to
->sd_uid
= htonl(from
->sd_uid
);
3473 to
->sd_gid
= htonl(from
->sd_gid
);
3474 to
->sd_size
= htonl(from
->sd_size
);
3477 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3478 struct write_data
*wd
)
3480 struct stat_data stat_data
;
3481 struct strbuf
*out
= &wd
->out
;
3482 unsigned char intbuf
[16];
3483 unsigned int intlen
, value
;
3484 int i
= wd
->index
++;
3487 * untracked_nr should be reset whenever valid is clear, but
3490 if (!untracked
->valid
) {
3491 untracked
->untracked_nr
= 0;
3492 untracked
->check_only
= 0;
3495 if (untracked
->check_only
)
3496 ewah_set(wd
->check_only
, i
);
3497 if (untracked
->valid
) {
3498 ewah_set(wd
->valid
, i
);
3499 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3500 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3502 if (!is_null_oid(&untracked
->exclude_oid
)) {
3503 ewah_set(wd
->sha1_valid
, i
);
3504 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3505 the_hash_algo
->rawsz
);
3508 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3509 strbuf_add(out
, intbuf
, intlen
);
3511 /* skip non-recurse directories */
3512 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3513 if (untracked
->dirs
[i
]->recurse
)
3515 intlen
= encode_varint(value
, intbuf
);
3516 strbuf_add(out
, intbuf
, intlen
);
3518 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3520 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3521 strbuf_add(out
, untracked
->untracked
[i
],
3522 strlen(untracked
->untracked
[i
]) + 1);
3524 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3525 if (untracked
->dirs
[i
]->recurse
)
3526 write_one_dir(untracked
->dirs
[i
], wd
);
3529 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3531 struct ondisk_untracked_cache
*ouc
;
3532 struct write_data wd
;
3533 unsigned char varbuf
[16];
3535 const unsigned hashsz
= the_hash_algo
->rawsz
;
3537 CALLOC_ARRAY(ouc
, 1);
3538 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3539 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3540 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3542 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3543 strbuf_add(out
, varbuf
, varint_len
);
3544 strbuf_addbuf(out
, &untracked
->ident
);
3546 strbuf_add(out
, ouc
, sizeof(*ouc
));
3547 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3548 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3549 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3552 if (!untracked
->root
) {
3553 varint_len
= encode_varint(0, varbuf
);
3554 strbuf_add(out
, varbuf
, varint_len
);
3559 wd
.check_only
= ewah_new();
3560 wd
.valid
= ewah_new();
3561 wd
.sha1_valid
= ewah_new();
3562 strbuf_init(&wd
.out
, 1024);
3563 strbuf_init(&wd
.sb_stat
, 1024);
3564 strbuf_init(&wd
.sb_sha1
, 1024);
3565 write_one_dir(untracked
->root
, &wd
);
3567 varint_len
= encode_varint(wd
.index
, varbuf
);
3568 strbuf_add(out
, varbuf
, varint_len
);
3569 strbuf_addbuf(out
, &wd
.out
);
3570 ewah_serialize_strbuf(wd
.valid
, out
);
3571 ewah_serialize_strbuf(wd
.check_only
, out
);
3572 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3573 strbuf_addbuf(out
, &wd
.sb_stat
);
3574 strbuf_addbuf(out
, &wd
.sb_sha1
);
3575 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3577 ewah_free(wd
.valid
);
3578 ewah_free(wd
.check_only
);
3579 ewah_free(wd
.sha1_valid
);
3580 strbuf_release(&wd
.out
);
3581 strbuf_release(&wd
.sb_stat
);
3582 strbuf_release(&wd
.sb_sha1
);
3585 static void free_untracked(struct untracked_cache_dir
*ucd
)
3590 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3591 free_untracked(ucd
->dirs
[i
]);
3592 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3593 free(ucd
->untracked
[i
]);
3594 free(ucd
->untracked
);
3599 void free_untracked_cache(struct untracked_cache
*uc
)
3604 free(uc
->exclude_per_dir_to_free
);
3605 strbuf_release(&uc
->ident
);
3606 free_untracked(uc
->root
);
3612 struct untracked_cache_dir
**ucd
;
3613 struct ewah_bitmap
*check_only
;
3614 struct ewah_bitmap
*valid
;
3615 struct ewah_bitmap
*sha1_valid
;
3616 const unsigned char *data
;
3617 const unsigned char *end
;
3620 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3622 memcpy(to
, data
, sizeof(*to
));
3623 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3624 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3625 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3626 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3627 to
->sd_dev
= ntohl(to
->sd_dev
);
3628 to
->sd_ino
= ntohl(to
->sd_ino
);
3629 to
->sd_uid
= ntohl(to
->sd_uid
);
3630 to
->sd_gid
= ntohl(to
->sd_gid
);
3631 to
->sd_size
= ntohl(to
->sd_size
);
3634 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3635 struct read_data
*rd
)
3637 struct untracked_cache_dir ud
, *untracked
;
3638 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3639 const unsigned char *eos
;
3643 memset(&ud
, 0, sizeof(ud
));
3645 value
= decode_varint(&data
);
3649 ud
.untracked_alloc
= value
;
3650 ud
.untracked_nr
= value
;
3651 if (ud
.untracked_nr
)
3652 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3654 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3657 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3659 eos
= memchr(data
, '\0', end
- data
);
3660 if (!eos
|| eos
== end
)
3663 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3664 memcpy(untracked
, &ud
, sizeof(ud
));
3665 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3668 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3669 eos
= memchr(data
, '\0', end
- data
);
3670 if (!eos
|| eos
== end
)
3672 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3676 rd
->ucd
[rd
->index
++] = untracked
;
3679 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3680 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3686 static void set_check_only(size_t pos
, void *cb
)
3688 struct read_data
*rd
= cb
;
3689 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3693 static void read_stat(size_t pos
, void *cb
)
3695 struct read_data
*rd
= cb
;
3696 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3697 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3698 rd
->data
= rd
->end
+ 1;
3701 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3702 rd
->data
+= sizeof(struct stat_data
);
3706 static void read_oid(size_t pos
, void *cb
)
3708 struct read_data
*rd
= cb
;
3709 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3710 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3711 rd
->data
= rd
->end
+ 1;
3714 oidread(&ud
->exclude_oid
, rd
->data
);
3715 rd
->data
+= the_hash_algo
->rawsz
;
3718 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3719 const unsigned char *sha1
)
3721 stat_data_from_disk(&oid_stat
->stat
, data
);
3722 oidread(&oid_stat
->oid
, sha1
);
3723 oid_stat
->valid
= 1;
3726 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3728 struct untracked_cache
*uc
;
3729 struct read_data rd
;
3730 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3734 const char *exclude_per_dir
;
3735 const unsigned hashsz
= the_hash_algo
->rawsz
;
3736 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3737 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3739 if (sz
<= 1 || end
[-1] != '\0')
3743 ident_len
= decode_varint(&next
);
3744 if (next
+ ident_len
> end
)
3746 ident
= (const char *)next
;
3749 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3752 CALLOC_ARRAY(uc
, 1);
3753 strbuf_init(&uc
->ident
, ident_len
);
3754 strbuf_add(&uc
->ident
, ident
, ident_len
);
3755 load_oid_stat(&uc
->ss_info_exclude
,
3756 next
+ ouc_offset(info_exclude_stat
),
3758 load_oid_stat(&uc
->ss_excludes_file
,
3759 next
+ ouc_offset(excludes_file_stat
),
3760 next
+ offset
+ hashsz
);
3761 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3762 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3763 uc
->exclude_per_dir
= uc
->exclude_per_dir_to_free
= xstrdup(exclude_per_dir
);
3764 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3765 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3769 len
= decode_varint(&next
);
3770 if (next
> end
|| len
== 0)
3773 rd
.valid
= ewah_new();
3774 rd
.check_only
= ewah_new();
3775 rd
.sha1_valid
= ewah_new();
3779 ALLOC_ARRAY(rd
.ucd
, len
);
3781 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3785 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3790 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3795 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3799 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3800 rd
.data
= next
+ len
;
3801 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3802 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3807 ewah_free(rd
.valid
);
3808 ewah_free(rd
.check_only
);
3809 ewah_free(rd
.sha1_valid
);
3812 free_untracked_cache(uc
);
3818 static void invalidate_one_directory(struct untracked_cache
*uc
,
3819 struct untracked_cache_dir
*ucd
)
3821 uc
->dir_invalidated
++;
3823 ucd
->untracked_nr
= 0;
3827 * Normally when an entry is added or removed from a directory,
3828 * invalidating that directory is enough. No need to touch its
3829 * ancestors. When a directory is shown as "foo/bar/" in git-status
3830 * however, deleting or adding an entry may have cascading effect.
3832 * Say the "foo/bar/file" has become untracked, we need to tell the
3833 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3834 * directory any more (because "bar" is managed by foo as an untracked
3837 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3838 * was the last untracked entry in the entire "foo", we should show
3839 * "foo/" instead. Which means we have to invalidate past "bar" up to
3842 * This function traverses all directories from root to leaf. If there
3843 * is a chance of one of the above cases happening, we invalidate back
3844 * to root. Otherwise we just invalidate the leaf. There may be a more
3845 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3846 * detect these cases and avoid unnecessary invalidation, for example,
3847 * checking for the untracked entry named "bar/" in "foo", but for now
3848 * stick to something safe and simple.
3850 static int invalidate_one_component(struct untracked_cache
*uc
,
3851 struct untracked_cache_dir
*dir
,
3852 const char *path
, int len
)
3854 const char *rest
= strchr(path
, '/');
3857 int component_len
= rest
- path
;
3858 struct untracked_cache_dir
*d
=
3859 lookup_untracked(uc
, dir
, path
, component_len
);
3861 invalidate_one_component(uc
, d
, rest
+ 1,
3862 len
- (component_len
+ 1));
3864 invalidate_one_directory(uc
, dir
);
3868 invalidate_one_directory(uc
, dir
);
3869 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3872 void untracked_cache_invalidate_path(struct index_state
*istate
,
3873 const char *path
, int safe_path
)
3875 if (!istate
->untracked
|| !istate
->untracked
->root
)
3877 if (!safe_path
&& !verify_path(path
, 0))
3879 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3880 path
, strlen(path
));
3883 void untracked_cache_remove_from_index(struct index_state
*istate
,
3886 untracked_cache_invalidate_path(istate
, path
, 1);
3889 void untracked_cache_add_to_index(struct index_state
*istate
,
3892 untracked_cache_invalidate_path(istate
, path
, 1);
3895 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3896 const char *sub_gitdir
)
3899 struct repository subrepo
;
3900 struct strbuf sub_wt
= STRBUF_INIT
;
3901 struct strbuf sub_gd
= STRBUF_INIT
;
3903 const struct submodule
*sub
;
3905 /* If the submodule has no working tree, we can ignore it. */
3906 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3909 if (repo_read_index(&subrepo
) < 0)
3910 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3912 /* TODO: audit for interaction with sparse-index. */
3913 ensure_full_index(subrepo
.index
);
3914 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3915 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3917 if (!S_ISGITLINK(ce
->ce_mode
))
3920 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3921 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3923 * Skip entries with the same name in different stages
3924 * to make sure an entry is returned only once.
3928 sub
= submodule_from_path(&subrepo
, null_oid(), ce
->name
);
3929 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3930 /* .gitmodules broken or inactive sub */
3933 strbuf_reset(&sub_wt
);
3934 strbuf_reset(&sub_gd
);
3935 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3936 submodule_name_to_gitdir(&sub_gd
, &subrepo
, sub
->name
);
3938 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3940 strbuf_release(&sub_wt
);
3941 strbuf_release(&sub_gd
);
3942 repo_clear(&subrepo
);
3945 void connect_work_tree_and_git_dir(const char *work_tree_
,
3946 const char *git_dir_
,
3947 int recurse_into_nested
)
3949 struct strbuf gitfile_sb
= STRBUF_INIT
;
3950 struct strbuf cfg_sb
= STRBUF_INIT
;
3951 struct strbuf rel_path
= STRBUF_INIT
;
3952 char *git_dir
, *work_tree
;
3954 /* Prepare .git file */
3955 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3956 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3957 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3959 /* Prepare config file */
3960 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3961 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3962 die(_("could not create directories for %s"), cfg_sb
.buf
);
3964 git_dir
= real_pathdup(git_dir_
, 1);
3965 work_tree
= real_pathdup(work_tree_
, 1);
3967 /* Write .git file */
3968 write_file(gitfile_sb
.buf
, "gitdir: %s",
3969 relative_path(git_dir
, work_tree
, &rel_path
));
3970 /* Update core.worktree setting */
3971 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3972 relative_path(work_tree
, git_dir
, &rel_path
));
3974 strbuf_release(&gitfile_sb
);
3975 strbuf_release(&cfg_sb
);
3976 strbuf_release(&rel_path
);
3978 if (recurse_into_nested
)
3979 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3986 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3988 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3990 if (rename(old_git_dir
, new_git_dir
) < 0)
3991 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3992 old_git_dir
, new_git_dir
);
3994 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);
3997 int path_match_flags(const char *const str
, const enum path_match_flags flags
)
3999 const char *p
= str
;
4001 if (flags
& PATH_MATCH_NATIVE
&&
4002 flags
& PATH_MATCH_XPLATFORM
)
4003 BUG("path_match_flags() must get one match kind, not multiple!");
4004 else if (!(flags
& PATH_MATCH_KINDS_MASK
))
4005 BUG("path_match_flags() must get at least one match kind!");
4007 if (flags
& PATH_MATCH_STARTS_WITH_DOT_SLASH
&&
4008 flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
)
4009 BUG("path_match_flags() must get one platform kind, not multiple!");
4010 else if (!(flags
& PATH_MATCH_PLATFORM_MASK
))
4011 BUG("path_match_flags() must get at least one platform kind!");
4015 if (flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
&&
4019 if (flags
& PATH_MATCH_NATIVE
)
4020 return is_dir_sep(*p
);
4021 else if (flags
& PATH_MATCH_XPLATFORM
)
4022 return is_xplatform_dir_sep(*p
);