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
11 #include "object-store.h"
14 #include "wildmatch.h"
18 #include "ewah/ewok.h"
19 #include "fsmonitor.h"
20 #include "submodule-config.h"
23 * Tells read_directory_recursive how a file or directory should be treated.
24 * Values are ordered by significance, e.g. if a directory contains both
25 * excluded and untracked files, it is listed as untracked because
26 * path_untracked > path_excluded.
36 * Support data structure for our opendir/readdir/closedir wrappers
40 struct untracked_cache_dir
*untracked
;
47 struct untracked_cache_dir
*ucd
;
50 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
51 struct index_state
*istate
, const char *path
, int len
,
52 struct untracked_cache_dir
*untracked
,
53 int check_only
, int stop_at_first_file
, const struct pathspec
*pathspec
);
54 static int resolve_dtype(int dtype
, struct index_state
*istate
,
55 const char *path
, int len
);
56 struct dirent
*readdir_skip_dot_and_dotdot(DIR *dirp
)
60 while ((e
= readdir(dirp
)) != NULL
) {
61 if (!is_dot_or_dotdot(e
->d_name
))
67 int count_slashes(const char *s
)
76 int fspathcmp(const char *a
, const char *b
)
78 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
81 int fspatheq(const char *a
, const char *b
)
83 return !fspathcmp(a
, b
);
86 int fspathncmp(const char *a
, const char *b
, size_t count
)
88 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
91 unsigned int fspathhash(const char *str
)
93 return ignore_case
? strihash(str
) : strhash(str
);
96 int git_fnmatch(const struct pathspec_item
*item
,
97 const char *pattern
, const char *string
,
101 if (ps_strncmp(item
, pattern
, string
, prefix
))
106 if (item
->flags
& PATHSPEC_ONESTAR
) {
107 int pattern_len
= strlen(++pattern
);
108 int string_len
= strlen(string
);
109 return string_len
< pattern_len
||
110 ps_strcmp(item
, pattern
,
111 string
+ string_len
- pattern_len
);
113 if (item
->magic
& PATHSPEC_GLOB
)
114 return wildmatch(pattern
, string
,
116 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
118 /* wildmatch has not learned no FNM_PATHNAME mode yet */
119 return wildmatch(pattern
, string
,
120 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
123 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
124 const char *string
, int stringlen
,
128 struct strbuf pat_buf
= STRBUF_INIT
;
129 struct strbuf str_buf
= STRBUF_INIT
;
130 const char *use_pat
= pattern
;
131 const char *use_str
= string
;
133 if (pattern
[patternlen
]) {
134 strbuf_add(&pat_buf
, pattern
, patternlen
);
135 use_pat
= pat_buf
.buf
;
137 if (string
[stringlen
]) {
138 strbuf_add(&str_buf
, string
, stringlen
);
139 use_str
= str_buf
.buf
;
143 flags
|= WM_CASEFOLD
;
144 match_status
= wildmatch(use_pat
, use_str
, flags
);
146 strbuf_release(&pat_buf
);
147 strbuf_release(&str_buf
);
152 static size_t common_prefix_len(const struct pathspec
*pathspec
)
158 * ":(icase)path" is treated as a pathspec full of
159 * wildcard. In other words, only prefix is considered common
160 * prefix. If the pathspec is abc/foo abc/bar, running in
161 * subdir xyz, the common prefix is still xyz, not xyz/abc as
164 GUARD_PATHSPEC(pathspec
,
173 for (n
= 0; n
< pathspec
->nr
; n
++) {
174 size_t i
= 0, len
= 0, item_len
;
175 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
177 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
178 item_len
= pathspec
->items
[n
].prefix
;
180 item_len
= pathspec
->items
[n
].nowildcard_len
;
181 while (i
< item_len
&& (n
== 0 || i
< max
)) {
182 char c
= pathspec
->items
[n
].match
[i
];
183 if (c
!= pathspec
->items
[0].match
[i
])
189 if (n
== 0 || len
< max
) {
199 * Returns a copy of the longest leading path common among all
202 char *common_prefix(const struct pathspec
*pathspec
)
204 unsigned long len
= common_prefix_len(pathspec
);
206 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
209 int fill_directory(struct dir_struct
*dir
,
210 struct index_state
*istate
,
211 const struct pathspec
*pathspec
)
216 unsigned exclusive_flags
= DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
;
217 if ((dir
->flags
& exclusive_flags
) == exclusive_flags
)
218 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
221 * Calculate common prefix for the pathspec, and
222 * use that to optimize the directory walk
224 prefix_len
= common_prefix_len(pathspec
);
225 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
227 /* Read the directory and prune it */
228 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
233 int within_depth(const char *name
, int namelen
,
234 int depth
, int max_depth
)
236 const char *cp
= name
, *cpe
= name
+ namelen
;
242 if (depth
> max_depth
)
249 * Read the contents of the blob with the given OID into a buffer.
250 * Append a trailing LF to the end if the last line doesn't have one.
253 * -1 when the OID is invalid or unknown or does not refer to a blob.
254 * 0 when the blob is empty.
255 * 1 along with { data, size } of the (possibly augmented) buffer
258 * Optionally updates the given oid_stat with the given OID (when valid).
260 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
261 size_t *size_out
, char **data_out
)
263 enum object_type type
;
270 data
= read_object_file(oid
, &type
, &sz
);
271 if (!data
|| type
!= OBJ_BLOB
) {
277 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
278 oidcpy(&oid_stat
->oid
, oid
);
286 if (data
[sz
- 1] != '\n') {
287 data
= xrealloc(data
, st_add(sz
, 1));
291 *size_out
= xsize_t(sz
);
297 #define DO_MATCH_EXCLUDE (1<<0)
298 #define DO_MATCH_DIRECTORY (1<<1)
299 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
302 * Does the given pathspec match the given name? A match is found if
304 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
305 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
306 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
307 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
309 * Return value tells which case it was (1-4), or 0 when there is no match.
311 * It may be instructive to look at a small table of concrete examples
312 * to understand the differences between 1, 2, and 4:
315 * | a/b | a/b/ | a/b/c
316 * ------+-----------+-----------+------------
317 * a/b | EXACT | EXACT[1] | LEADING[2]
318 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
319 * a/b/c | RECURSIVE | RECURSIVE | EXACT
321 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
322 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
324 static int match_pathspec_item(struct index_state
*istate
,
325 const struct pathspec_item
*item
, int prefix
,
326 const char *name
, int namelen
, unsigned flags
)
328 /* name/namelen has prefix cut off by caller */
329 const char *match
= item
->match
+ prefix
;
330 int matchlen
= item
->len
- prefix
;
333 * The normal call pattern is:
334 * 1. prefix = common_prefix_len(ps);
335 * 2. prune something, or fill_directory
336 * 3. match_pathspec()
338 * 'prefix' at #1 may be shorter than the command's prefix and
339 * it's ok for #2 to match extra files. Those extras will be
342 * Suppose the pathspec is 'foo' and '../bar' running from
343 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
344 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
345 * user does not want XYZ/foo, only the "foo" part should be
346 * case-insensitive. We need to filter out XYZ/foo here. In
347 * other words, we do not trust the caller on comparing the
348 * prefix part when :(icase) is involved. We do exact
349 * comparison ourselves.
351 * Normally the caller (common_prefix_len() in fact) does
352 * _exact_ matching on name[-prefix+1..-1] and we do not need
353 * to check that part. Be defensive and check it anyway, in
354 * case common_prefix_len is changed, or a new caller is
355 * introduced that does not use common_prefix_len.
357 * If the penalty turns out too high when prefix is really
358 * long, maybe change it to
359 * strncmp(match, name, item->prefix - prefix)
361 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
362 strncmp(item
->match
, name
- prefix
, item
->prefix
))
365 if (item
->attr_match_nr
&&
366 !match_pathspec_attrs(istate
, name
, namelen
, item
))
369 /* If the match was just the prefix, we matched */
371 return MATCHED_RECURSIVELY
;
373 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
374 if (matchlen
== namelen
)
375 return MATCHED_EXACTLY
;
377 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
378 return MATCHED_RECURSIVELY
;
379 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
380 match
[matchlen
- 1] == '/' &&
381 namelen
== matchlen
- 1 &&
382 !ps_strncmp(item
, match
, name
, namelen
))
383 return MATCHED_EXACTLY
;
385 if (item
->nowildcard_len
< item
->len
&&
386 !git_fnmatch(item
, match
, name
,
387 item
->nowildcard_len
- prefix
))
388 return MATCHED_FNMATCH
;
390 /* Perform checks to see if "name" is a leading string of the pathspec */
391 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
392 !(flags
& DO_MATCH_EXCLUDE
)) {
393 /* name is a literal prefix of the pathspec */
394 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
395 if ((namelen
< matchlen
) &&
396 (match
[namelen
-offset
] == '/') &&
397 !ps_strncmp(item
, match
, name
, namelen
))
398 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
400 /* name doesn't match up to the first wild character */
401 if (item
->nowildcard_len
< item
->len
&&
402 ps_strncmp(item
, match
, name
,
403 item
->nowildcard_len
- prefix
))
407 * name has no wildcard, and it didn't match as a leading
408 * pathspec so return.
410 if (item
->nowildcard_len
== item
->len
)
414 * Here is where we would perform a wildmatch to check if
415 * "name" can be matched as a directory (or a prefix) against
416 * the pathspec. Since wildmatch doesn't have this capability
417 * at the present we have to punt and say that it is a match,
418 * potentially returning a false positive
419 * The submodules themselves will be able to perform more
420 * accurate matching to determine if the pathspec matches.
422 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
429 * do_match_pathspec() is meant to ONLY be called by
430 * match_pathspec_with_flags(); calling it directly risks pathspecs
431 * like ':!unwanted_path' being ignored.
433 * Given a name and a list of pathspecs, returns the nature of the
434 * closest (i.e. most specific) match of the name to any of the
437 * The caller typically calls this multiple times with the same
438 * pathspec and seen[] array but with different name/namelen
439 * (e.g. entries from the index) and is interested in seeing if and
440 * how each pathspec matches all the names it calls this function
441 * with. A mark is left in the seen[] array for each pathspec element
442 * indicating the closest type of match that element achieved, so if
443 * seen[n] remains zero after multiple invocations, that means the nth
444 * pathspec did not match any names, which could indicate that the
445 * user mistyped the nth pathspec.
447 static int do_match_pathspec(struct index_state
*istate
,
448 const struct pathspec
*ps
,
449 const char *name
, int namelen
,
450 int prefix
, char *seen
,
453 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
465 if (!ps
->recursive
||
466 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
468 return MATCHED_RECURSIVELY
;
470 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
471 return MATCHED_EXACTLY
;
479 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
482 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
483 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
486 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
489 * Make exclude patterns optional and never report
490 * "pathspec ':(exclude)foo' matches no files"
492 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
493 seen
[i
] = MATCHED_FNMATCH
;
494 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
497 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
498 ps
->max_depth
!= -1 &&
499 how
&& how
!= MATCHED_FNMATCH
) {
500 int len
= ps
->items
[i
].len
;
501 if (name
[len
] == '/')
503 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
504 how
= MATCHED_EXACTLY
;
511 if (seen
&& seen
[i
] < how
)
518 static int match_pathspec_with_flags(struct index_state
*istate
,
519 const struct pathspec
*ps
,
520 const char *name
, int namelen
,
521 int prefix
, char *seen
, unsigned flags
)
523 int positive
, negative
;
524 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
525 prefix
, seen
, flags
);
526 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
528 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
530 flags
| DO_MATCH_EXCLUDE
);
531 return negative
? 0 : positive
;
534 int match_pathspec(struct index_state
*istate
,
535 const struct pathspec
*ps
,
536 const char *name
, int namelen
,
537 int prefix
, char *seen
, int is_dir
)
539 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
540 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
541 prefix
, seen
, flags
);
545 * Check if a submodule is a superset of the pathspec
547 int submodule_path_match(struct index_state
*istate
,
548 const struct pathspec
*ps
,
549 const char *submodule_name
,
552 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
553 strlen(submodule_name
),
556 DO_MATCH_LEADING_PATHSPEC
);
560 int report_path_error(const char *ps_matched
,
561 const struct pathspec
*pathspec
)
564 * Make sure all pathspec matched; otherwise it is an error.
567 for (num
= 0; num
< pathspec
->nr
; num
++) {
568 int other
, found_dup
;
573 * The caller might have fed identical pathspec
574 * twice. Do not barf on such a mistake.
575 * FIXME: parse_pathspec should have eliminated
576 * duplicate pathspec.
578 for (found_dup
= other
= 0;
579 !found_dup
&& other
< pathspec
->nr
;
581 if (other
== num
|| !ps_matched
[other
])
583 if (!strcmp(pathspec
->items
[other
].original
,
584 pathspec
->items
[num
].original
))
586 * Ok, we have a match already.
593 error(_("pathspec '%s' did not match any file(s) known to git"),
594 pathspec
->items
[num
].original
);
601 * Return the length of the "simple" part of a path match limiter.
603 int simple_length(const char *match
)
608 unsigned char c
= *match
++;
610 if (c
== '\0' || is_glob_special(c
))
615 int no_wildcard(const char *string
)
617 return string
[simple_length(string
)] == '\0';
620 void parse_path_pattern(const char **pattern
,
625 const char *p
= *pattern
;
630 *flags
|= PATTERN_FLAG_NEGATIVE
;
634 if (len
&& p
[len
- 1] == '/') {
636 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
638 for (i
= 0; i
< len
; i
++) {
643 *flags
|= PATTERN_FLAG_NODIR
;
644 *nowildcardlen
= simple_length(p
);
646 * we should have excluded the trailing slash from 'p' too,
647 * but that's one more allocation. Instead just make sure
648 * nowildcardlen does not exceed real patternlen
650 if (*nowildcardlen
> len
)
651 *nowildcardlen
= len
;
652 if (*p
== '*' && no_wildcard(p
+ 1))
653 *flags
|= PATTERN_FLAG_ENDSWITH
;
658 int pl_hashmap_cmp(const void *unused_cmp_data
,
659 const struct hashmap_entry
*a
,
660 const struct hashmap_entry
*b
,
663 const struct pattern_entry
*ee1
=
664 container_of(a
, struct pattern_entry
, ent
);
665 const struct pattern_entry
*ee2
=
666 container_of(b
, struct pattern_entry
, ent
);
668 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
673 return strncasecmp(ee1
->pattern
, ee2
->pattern
, min_len
);
674 return strncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
677 static char *dup_and_filter_pattern(const char *pattern
)
681 char *result
= xstrdup(pattern
);
687 /* skip escape characters (once) */
707 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
709 struct pattern_entry
*translated
;
712 const char *prev
, *cur
, *next
;
714 if (!pl
->use_cone_patterns
)
717 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
718 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
719 !strcmp(given
->pattern
, "/*")) {
724 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
729 if (given
->patternlen
< 2 ||
730 *given
->pattern
== '*' ||
731 strstr(given
->pattern
, "**")) {
732 /* Not a cone pattern. */
733 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
737 prev
= given
->pattern
;
738 cur
= given
->pattern
+ 1;
739 next
= given
->pattern
+ 2;
742 /* Watch for glob characters '*', '\', '[', '?' */
743 if (!is_glob_special(*cur
))
746 /* But only if *prev != '\\' */
750 /* But allow the initial '\' */
752 is_glob_special(*next
))
755 /* But a trailing '/' then '*' is fine */
761 /* Not a cone pattern. */
762 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
771 if (given
->patternlen
> 2 &&
772 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
773 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
774 /* Not a cone pattern. */
775 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
779 truncated
= dup_and_filter_pattern(given
->pattern
);
781 translated
= xmalloc(sizeof(struct pattern_entry
));
782 translated
->pattern
= truncated
;
783 translated
->patternlen
= given
->patternlen
- 2;
784 hashmap_entry_init(&translated
->ent
,
785 fspathhash(translated
->pattern
));
787 if (!hashmap_get_entry(&pl
->recursive_hashmap
,
788 translated
, ent
, NULL
)) {
789 /* We did not see the "parent" included */
790 warning(_("unrecognized negative pattern: '%s'"),
797 hashmap_add(&pl
->parent_hashmap
, &translated
->ent
);
798 hashmap_remove(&pl
->recursive_hashmap
, &translated
->ent
, &data
);
803 if (given
->flags
& PATTERN_FLAG_NEGATIVE
) {
804 warning(_("unrecognized negative pattern: '%s'"),
809 translated
= xmalloc(sizeof(struct pattern_entry
));
811 translated
->pattern
= dup_and_filter_pattern(given
->pattern
);
812 translated
->patternlen
= given
->patternlen
;
813 hashmap_entry_init(&translated
->ent
,
814 fspathhash(translated
->pattern
));
816 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
818 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
819 /* we already included this at the parent level */
820 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
822 hashmap_remove(&pl
->parent_hashmap
, &translated
->ent
, &data
);
830 warning(_("disabling cone pattern matching"));
831 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
832 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
833 pl
->use_cone_patterns
= 0;
836 static int hashmap_contains_path(struct hashmap
*map
,
837 struct strbuf
*pattern
)
839 struct pattern_entry p
;
841 /* Check straight mapping */
842 p
.pattern
= pattern
->buf
;
843 p
.patternlen
= pattern
->len
;
844 hashmap_entry_init(&p
.ent
, fspathhash(p
.pattern
));
845 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
848 int hashmap_contains_parent(struct hashmap
*map
,
850 struct strbuf
*buffer
)
854 strbuf_setlen(buffer
, 0);
857 strbuf_addch(buffer
, '/');
859 strbuf_addstr(buffer
, path
);
861 slash_pos
= strrchr(buffer
->buf
, '/');
863 while (slash_pos
> buffer
->buf
) {
864 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
866 if (hashmap_contains_path(map
, buffer
))
869 slash_pos
= strrchr(buffer
->buf
, '/');
875 void add_pattern(const char *string
, const char *base
,
876 int baselen
, struct pattern_list
*pl
, int srcpos
)
878 struct path_pattern
*pattern
;
883 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
884 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
885 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
887 pattern
= xmalloc(sizeof(*pattern
));
888 pattern
->pattern
= string
;
890 pattern
->patternlen
= patternlen
;
891 pattern
->nowildcardlen
= nowildcardlen
;
892 pattern
->base
= base
;
893 pattern
->baselen
= baselen
;
894 pattern
->flags
= flags
;
895 pattern
->srcpos
= srcpos
;
896 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
897 pl
->patterns
[pl
->nr
++] = pattern
;
900 add_pattern_to_hashsets(pl
, pattern
);
903 static int read_skip_worktree_file_from_index(struct index_state
*istate
,
905 size_t *size_out
, char **data_out
,
906 struct oid_stat
*oid_stat
)
911 pos
= index_name_pos(istate
, path
, len
);
914 if (!ce_skip_worktree(istate
->cache
[pos
]))
917 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
921 * Frees memory within pl which was allocated for exclude patterns and
922 * the file buffer. Does not free pl itself.
924 void clear_pattern_list(struct pattern_list
*pl
)
928 for (i
= 0; i
< pl
->nr
; i
++)
929 free(pl
->patterns
[i
]);
932 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
933 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
935 memset(pl
, 0, sizeof(*pl
));
938 static void trim_trailing_spaces(char *buf
)
940 char *p
, *last_space
= NULL
;
942 for (p
= buf
; *p
; p
++)
962 * Given a subdirectory name and "dir" of the current directory,
963 * search the subdir in "dir" and return it, or create a new one if it
964 * does not exist in "dir".
966 * If "name" has the trailing slash, it'll be excluded in the search.
968 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
969 struct untracked_cache_dir
*dir
,
970 const char *name
, int len
)
973 struct untracked_cache_dir
*d
;
976 if (len
&& name
[len
- 1] == '/')
980 while (last
> first
) {
981 int cmp
, next
= first
+ ((last
- first
) >> 1);
983 cmp
= strncmp(name
, d
->name
, len
);
984 if (!cmp
&& strlen(d
->name
) > len
)
996 FLEX_ALLOC_MEM(d
, name
, name
, len
);
998 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
999 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
1000 dir
->dirs_nr
- first
);
1002 dir
->dirs
[first
] = d
;
1006 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
1010 dir
->untracked_nr
= 0;
1011 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1012 do_invalidate_gitignore(dir
->dirs
[i
]);
1015 static void invalidate_gitignore(struct untracked_cache
*uc
,
1016 struct untracked_cache_dir
*dir
)
1018 uc
->gitignore_invalidated
++;
1019 do_invalidate_gitignore(dir
);
1022 static void invalidate_directory(struct untracked_cache
*uc
,
1023 struct untracked_cache_dir
*dir
)
1028 * Invalidation increment here is just roughly correct. If
1029 * untracked_nr or any of dirs[].recurse is non-zero, we
1030 * should increment dir_invalidated too. But that's more
1034 uc
->dir_invalidated
++;
1037 dir
->untracked_nr
= 0;
1038 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1039 dir
->dirs
[i
]->recurse
= 0;
1042 static int add_patterns_from_buffer(char *buf
, size_t size
,
1043 const char *base
, int baselen
,
1044 struct pattern_list
*pl
);
1046 /* Flags for add_patterns() */
1047 #define PATTERN_NOFOLLOW (1<<0)
1050 * Given a file with name "fname", read it (either from disk, or from
1051 * an index if 'istate' is non-null), parse it and store the
1052 * exclude rules in "pl".
1054 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1055 * stat data from disk (only valid if add_patterns returns zero). If
1056 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1058 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1059 struct pattern_list
*pl
, struct index_state
*istate
,
1060 unsigned flags
, struct oid_stat
*oid_stat
)
1068 if (flags
& PATTERN_NOFOLLOW
)
1069 fd
= open_nofollow(fname
, O_RDONLY
);
1071 fd
= open(fname
, O_RDONLY
);
1073 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1075 warn_on_fopen_errors(fname
);
1080 r
= read_skip_worktree_file_from_index(istate
, fname
,
1086 size
= xsize_t(st
.st_size
);
1089 fill_stat_data(&oid_stat
->stat
, &st
);
1090 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1091 oid_stat
->valid
= 1;
1096 buf
= xmallocz(size
);
1097 if (read_in_full(fd
, buf
, size
) != size
) {
1106 if (oid_stat
->valid
&&
1107 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1108 ; /* no content change, oid_stat->oid still good */
1110 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1111 !ce_stage(istate
->cache
[pos
]) &&
1112 ce_uptodate(istate
->cache
[pos
]) &&
1113 !would_convert_to_git(istate
, fname
))
1114 oidcpy(&oid_stat
->oid
,
1115 &istate
->cache
[pos
]->oid
);
1117 hash_object_file(the_hash_algo
, buf
, size
,
1118 "blob", &oid_stat
->oid
);
1119 fill_stat_data(&oid_stat
->stat
, &st
);
1120 oid_stat
->valid
= 1;
1124 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1128 static int add_patterns_from_buffer(char *buf
, size_t size
,
1129 const char *base
, int baselen
,
1130 struct pattern_list
*pl
)
1135 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1136 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1140 if (skip_utf8_bom(&buf
, size
))
1141 size
-= buf
- pl
->filebuf
;
1145 for (i
= 0; i
< size
; i
++) {
1146 if (buf
[i
] == '\n') {
1147 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1148 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1149 trim_trailing_spaces(entry
);
1150 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1153 entry
= buf
+ i
+ 1;
1159 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1160 int baselen
, struct pattern_list
*pl
,
1161 struct index_state
*istate
,
1164 return add_patterns(fname
, base
, baselen
, pl
, istate
, flags
, NULL
);
1167 int add_patterns_from_blob_to_list(
1168 struct object_id
*oid
,
1169 const char *base
, int baselen
,
1170 struct pattern_list
*pl
)
1176 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1180 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1184 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1185 int group_type
, const char *src
)
1187 struct pattern_list
*pl
;
1188 struct exclude_list_group
*group
;
1190 group
= &dir
->exclude_list_group
[group_type
];
1191 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1192 pl
= &group
->pl
[group
->nr
++];
1193 memset(pl
, 0, sizeof(*pl
));
1199 * Used to set up core.excludesfile and .git/info/exclude lists.
1201 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1202 struct oid_stat
*oid_stat
)
1204 struct pattern_list
*pl
;
1206 * catch setup_standard_excludes() that's called before
1207 * dir->untracked is assigned. That function behaves
1208 * differently when dir->untracked is non-NULL.
1210 if (!dir
->untracked
)
1211 dir
->unmanaged_exclude_files
++;
1212 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1213 if (add_patterns(fname
, "", 0, pl
, NULL
, 0, oid_stat
) < 0)
1214 die(_("cannot use %s as an exclude file"), fname
);
1217 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1219 dir
->unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1220 add_patterns_from_file_1(dir
, fname
, NULL
);
1223 int match_basename(const char *basename
, int basenamelen
,
1224 const char *pattern
, int prefix
, int patternlen
,
1227 if (prefix
== patternlen
) {
1228 if (patternlen
== basenamelen
&&
1229 !fspathncmp(pattern
, basename
, basenamelen
))
1231 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1232 /* "*literal" matching against "fooliteral" */
1233 if (patternlen
- 1 <= basenamelen
&&
1234 !fspathncmp(pattern
+ 1,
1235 basename
+ basenamelen
- (patternlen
- 1),
1239 if (fnmatch_icase_mem(pattern
, patternlen
,
1240 basename
, basenamelen
,
1247 int match_pathname(const char *pathname
, int pathlen
,
1248 const char *base
, int baselen
,
1249 const char *pattern
, int prefix
, int patternlen
,
1256 * match with FNM_PATHNAME; the pattern has base implicitly
1259 if (*pattern
== '/') {
1266 * baselen does not count the trailing slash. base[] may or
1267 * may not end with a trailing slash though.
1269 if (pathlen
< baselen
+ 1 ||
1270 (baselen
&& pathname
[baselen
] != '/') ||
1271 fspathncmp(pathname
, base
, baselen
))
1274 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1275 name
= pathname
+ pathlen
- namelen
;
1279 * if the non-wildcard part is longer than the
1280 * remaining pathname, surely it cannot match.
1282 if (prefix
> namelen
)
1285 if (fspathncmp(pattern
, name
, prefix
))
1288 patternlen
-= prefix
;
1293 * If the whole pattern did not have a wildcard,
1294 * then our prefix match is all we need; we
1295 * do not need to call fnmatch at all.
1297 if (!patternlen
&& !namelen
)
1301 return fnmatch_icase_mem(pattern
, patternlen
,
1306 static int path_matches_dir_pattern(const char *pathname
,
1308 struct strbuf
**path_parent
,
1310 struct path_pattern
*pattern
,
1311 struct index_state
*istate
)
1313 if (!*path_parent
) {
1315 CALLOC_ARRAY(*path_parent
, 1);
1316 strbuf_add(*path_parent
, pathname
, pathlen
);
1317 slash
= find_last_dir_sep((*path_parent
)->buf
);
1320 strbuf_setlen(*path_parent
, slash
- (*path_parent
)->buf
);
1322 strbuf_setlen(*path_parent
, 0);
1326 * If the parent directory matches the pattern, then we do not
1327 * need to check for dtype.
1329 if ((*path_parent
)->len
&&
1330 match_pathname((*path_parent
)->buf
, (*path_parent
)->len
,
1332 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1333 pattern
->pattern
, pattern
->nowildcardlen
,
1334 pattern
->patternlen
, pattern
->flags
))
1337 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1338 if (*dtype
!= DT_DIR
)
1345 * Scan the given exclude list in reverse to see whether pathname
1346 * should be ignored. The first match (i.e. the last on the list), if
1347 * any, determines the fate. Returns the exclude_list element which
1348 * matched, or NULL for undecided.
1350 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1352 const char *basename
,
1354 struct pattern_list
*pl
,
1355 struct index_state
*istate
)
1357 struct path_pattern
*res
= NULL
; /* undecided */
1359 struct strbuf
*path_parent
= NULL
;
1362 return NULL
; /* undefined */
1364 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1365 struct path_pattern
*pattern
= pl
->patterns
[i
];
1366 const char *exclude
= pattern
->pattern
;
1367 int prefix
= pattern
->nowildcardlen
;
1369 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
1370 !path_matches_dir_pattern(pathname
, pathlen
, &path_parent
,
1371 dtype
, pattern
, istate
))
1374 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1375 if (match_basename(basename
,
1376 pathlen
- (basename
- pathname
),
1377 exclude
, prefix
, pattern
->patternlen
,
1385 assert(pattern
->baselen
== 0 ||
1386 pattern
->base
[pattern
->baselen
- 1] == '/');
1387 if (match_pathname(pathname
, pathlen
,
1389 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1390 exclude
, prefix
, pattern
->patternlen
,
1398 strbuf_release(path_parent
);
1406 * Scan the list of patterns to determine if the ordered list
1407 * of patterns matches on 'pathname'.
1409 * Return 1 for a match, 0 for not matched and -1 for undecided.
1411 enum pattern_match_result
path_matches_pattern_list(
1412 const char *pathname
, int pathlen
,
1413 const char *basename
, int *dtype
,
1414 struct pattern_list
*pl
,
1415 struct index_state
*istate
)
1417 struct path_pattern
*pattern
;
1418 struct strbuf parent_pathname
= STRBUF_INIT
;
1419 int result
= NOT_MATCHED
;
1422 if (!pl
->use_cone_patterns
) {
1423 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1426 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1438 strbuf_addch(&parent_pathname
, '/');
1439 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1442 * Directory entries are matched if and only if a file
1443 * contained immediately within them is matched. For the
1444 * case of a directory entry, modify the path to create
1445 * a fake filename within this directory, allowing us to
1446 * use the file-base matching logic in an equivalent way.
1448 if (parent_pathname
.len
> 0 &&
1449 parent_pathname
.buf
[parent_pathname
.len
- 1] == '/') {
1450 slash_pos
= parent_pathname
.len
- 1;
1451 strbuf_add(&parent_pathname
, "-", 1);
1453 const char *slash_ptr
= strrchr(parent_pathname
.buf
, '/');
1454 slash_pos
= slash_ptr
? slash_ptr
- parent_pathname
.buf
: 0;
1457 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1458 &parent_pathname
)) {
1459 result
= MATCHED_RECURSIVE
;
1464 /* include every file in root */
1469 strbuf_setlen(&parent_pathname
, slash_pos
);
1471 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1476 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1479 result
= MATCHED_RECURSIVE
;
1482 strbuf_release(&parent_pathname
);
1486 int init_sparse_checkout_patterns(struct index_state
*istate
)
1488 if (!core_apply_sparse_checkout
)
1490 if (istate
->sparse_checkout_patterns
)
1493 CALLOC_ARRAY(istate
->sparse_checkout_patterns
, 1);
1495 if (get_sparse_checkout_patterns(istate
->sparse_checkout_patterns
) < 0) {
1496 FREE_AND_NULL(istate
->sparse_checkout_patterns
);
1503 static int path_in_sparse_checkout_1(const char *path
,
1504 struct index_state
*istate
,
1505 int require_cone_mode
)
1508 enum pattern_match_result match
= UNDECIDED
;
1509 const char *end
, *slash
;
1512 * We default to accepting a path if there are no patterns or
1513 * they are of the wrong type.
1515 if (init_sparse_checkout_patterns(istate
) ||
1516 (require_cone_mode
&&
1517 !istate
->sparse_checkout_patterns
->use_cone_patterns
))
1521 * If UNDECIDED, use the match from the parent dir (recursively), or
1522 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1523 * never returns UNDECIDED, so we will execute only one iteration in
1526 for (end
= path
+ strlen(path
);
1527 end
> path
&& match
== UNDECIDED
;
1530 for (slash
= end
- 1; slash
> path
&& *slash
!= '/'; slash
--)
1533 match
= path_matches_pattern_list(path
, end
- path
,
1534 slash
> path
? slash
+ 1 : path
, &dtype
,
1535 istate
->sparse_checkout_patterns
, istate
);
1537 /* We are going to match the parent dir now */
1543 int path_in_sparse_checkout(const char *path
,
1544 struct index_state
*istate
)
1546 return path_in_sparse_checkout_1(path
, istate
, 0);
1549 int path_in_cone_mode_sparse_checkout(const char *path
,
1550 struct index_state
*istate
)
1552 return path_in_sparse_checkout_1(path
, istate
, 1);
1555 static struct path_pattern
*last_matching_pattern_from_lists(
1556 struct dir_struct
*dir
, struct index_state
*istate
,
1557 const char *pathname
, int pathlen
,
1558 const char *basename
, int *dtype_p
)
1561 struct exclude_list_group
*group
;
1562 struct path_pattern
*pattern
;
1563 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1564 group
= &dir
->exclude_list_group
[i
];
1565 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1566 pattern
= last_matching_pattern_from_list(
1567 pathname
, pathlen
, basename
, dtype_p
,
1568 &group
->pl
[j
], istate
);
1577 * Loads the per-directory exclude list for the substring of base
1578 * which has a char length of baselen.
1580 static void prep_exclude(struct dir_struct
*dir
,
1581 struct index_state
*istate
,
1582 const char *base
, int baselen
)
1584 struct exclude_list_group
*group
;
1585 struct pattern_list
*pl
;
1586 struct exclude_stack
*stk
= NULL
;
1587 struct untracked_cache_dir
*untracked
;
1590 group
= &dir
->exclude_list_group
[EXC_DIRS
];
1593 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1594 * which originate from directories not in the prefix of the
1595 * path being checked.
1597 while ((stk
= dir
->exclude_stack
) != NULL
) {
1598 if (stk
->baselen
<= baselen
&&
1599 !strncmp(dir
->basebuf
.buf
, base
, stk
->baselen
))
1601 pl
= &group
->pl
[dir
->exclude_stack
->exclude_ix
];
1602 dir
->exclude_stack
= stk
->prev
;
1603 dir
->pattern
= NULL
;
1604 free((char *)pl
->src
); /* see strbuf_detach() below */
1605 clear_pattern_list(pl
);
1610 /* Skip traversing into sub directories if the parent is excluded */
1615 * Lazy initialization. All call sites currently just
1616 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1617 * them seems lots of work for little benefit.
1619 if (!dir
->basebuf
.buf
)
1620 strbuf_init(&dir
->basebuf
, PATH_MAX
);
1622 /* Read from the parent directories and push them down. */
1623 current
= stk
? stk
->baselen
: -1;
1624 strbuf_setlen(&dir
->basebuf
, current
< 0 ? 0 : current
);
1626 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1630 while (current
< baselen
) {
1632 struct oid_stat oid_stat
;
1634 CALLOC_ARRAY(stk
, 1);
1639 cp
= strchr(base
+ current
+ 1, '/');
1641 die("oops in prep_exclude");
1644 lookup_untracked(dir
->untracked
, untracked
,
1646 cp
- base
- current
);
1648 stk
->prev
= dir
->exclude_stack
;
1649 stk
->baselen
= cp
- base
;
1650 stk
->exclude_ix
= group
->nr
;
1651 stk
->ucd
= untracked
;
1652 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1653 strbuf_add(&dir
->basebuf
, base
+ current
, stk
->baselen
- current
);
1654 assert(stk
->baselen
== dir
->basebuf
.len
);
1656 /* Abort if the directory is excluded */
1659 dir
->basebuf
.buf
[stk
->baselen
- 1] = 0;
1660 dir
->pattern
= last_matching_pattern_from_lists(dir
,
1662 dir
->basebuf
.buf
, stk
->baselen
- 1,
1663 dir
->basebuf
.buf
+ current
, &dt
);
1664 dir
->basebuf
.buf
[stk
->baselen
- 1] = '/';
1666 dir
->pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1667 dir
->pattern
= NULL
;
1669 dir
->exclude_stack
= stk
;
1674 /* Try to read per-directory file */
1675 oidclr(&oid_stat
.oid
);
1677 if (dir
->exclude_per_dir
&&
1679 * If we know that no files have been added in
1680 * this directory (i.e. valid_cached_dir() has
1681 * been executed and set untracked->valid) ..
1683 (!untracked
|| !untracked
->valid
||
1685 * .. and .gitignore does not exist before
1686 * (i.e. null exclude_oid). Then we can skip
1687 * loading .gitignore, which would result in
1690 !is_null_oid(&untracked
->exclude_oid
))) {
1692 * dir->basebuf gets reused by the traversal, but we
1693 * need fname to remain unchanged to ensure the src
1694 * member of each struct path_pattern correctly
1695 * back-references its source file. Other invocations
1696 * of add_pattern_list provide stable strings, so we
1697 * strbuf_detach() and free() here in the caller.
1699 struct strbuf sb
= STRBUF_INIT
;
1700 strbuf_addbuf(&sb
, &dir
->basebuf
);
1701 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1702 pl
->src
= strbuf_detach(&sb
, NULL
);
1703 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1705 untracked
? &oid_stat
: NULL
);
1708 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1709 * will first be called in valid_cached_dir() then maybe many
1710 * times more in last_matching_pattern(). When the cache is
1711 * used, last_matching_pattern() will not be called and
1712 * reading .gitignore content will be a waste.
1714 * So when it's called by valid_cached_dir() and we can get
1715 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1716 * modified on work tree), we could delay reading the
1717 * .gitignore content until we absolutely need it in
1718 * last_matching_pattern(). Be careful about ignore rule
1719 * order, though, if you do that.
1722 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1723 invalidate_gitignore(dir
->untracked
, untracked
);
1724 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1726 dir
->exclude_stack
= stk
;
1727 current
= stk
->baselen
;
1729 strbuf_setlen(&dir
->basebuf
, baselen
);
1733 * Loads the exclude lists for the directory containing pathname, then
1734 * scans all exclude lists to determine whether pathname is excluded.
1735 * Returns the exclude_list element which matched, or NULL for
1738 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1739 struct index_state
*istate
,
1740 const char *pathname
,
1743 int pathlen
= strlen(pathname
);
1744 const char *basename
= strrchr(pathname
, '/');
1745 basename
= (basename
) ? basename
+1 : pathname
;
1747 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1750 return dir
->pattern
;
1752 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1757 * Loads the exclude lists for the directory containing pathname, then
1758 * scans all exclude lists to determine whether pathname is excluded.
1759 * Returns 1 if true, otherwise 0.
1761 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1762 const char *pathname
, int *dtype_p
)
1764 struct path_pattern
*pattern
=
1765 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1767 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1771 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1773 struct dir_entry
*ent
;
1775 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1780 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1781 struct index_state
*istate
,
1782 const char *pathname
, int len
)
1784 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1787 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
1788 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1791 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1792 struct index_state
*istate
,
1793 const char *pathname
, int len
)
1795 if (!index_name_is_other(istate
, pathname
, len
))
1798 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
1799 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1803 index_nonexistent
= 0,
1809 * Do not use the alphabetically sorted index to look up
1810 * the directory name; instead, use the case insensitive
1813 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1814 const char *dirname
, int len
)
1816 struct cache_entry
*ce
;
1818 if (index_dir_exists(istate
, dirname
, len
))
1819 return index_directory
;
1821 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1822 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1823 return index_gitdir
;
1825 return index_nonexistent
;
1829 * The index sorts alphabetically by entry name, which
1830 * means that a gitlink sorts as '\0' at the end, while
1831 * a directory (which is defined not as an entry, but as
1832 * the files it contains) will sort with the '/' at the
1835 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1836 const char *dirname
, int len
)
1841 return directory_exists_in_index_icase(istate
, dirname
, len
);
1843 pos
= index_name_pos(istate
, dirname
, len
);
1846 while (pos
< istate
->cache_nr
) {
1847 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1848 unsigned char endchar
;
1850 if (strncmp(ce
->name
, dirname
, len
))
1852 endchar
= ce
->name
[len
];
1856 return index_directory
;
1857 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1858 return index_gitdir
;
1860 return index_nonexistent
;
1864 * When we find a directory when traversing the filesystem, we
1865 * have three distinct cases:
1868 * - see it as a directory
1871 * and which one we choose depends on a combination of existing
1872 * git index contents and the flags passed into the directory
1873 * traversal routine.
1875 * Case 1: If we *already* have entries in the index under that
1876 * directory name, we always recurse into the directory to see
1879 * Case 2: If we *already* have that directory name as a gitlink,
1880 * we always continue to see it as a gitlink, regardless of whether
1881 * there is an actual git directory there or not (it might not
1882 * be checked out as a subproject!)
1884 * Case 3: if we didn't have it in the index previously, we
1885 * have a few sub-cases:
1887 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1888 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1889 * also true, in which case we need to check if it contains any
1890 * untracked and / or ignored files.
1891 * (b) if it looks like a git directory and we don't have the
1892 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1893 * show it as a directory.
1894 * (c) otherwise, we recurse into it.
1896 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1897 struct index_state
*istate
,
1898 struct untracked_cache_dir
*untracked
,
1899 const char *dirname
, int len
, int baselen
, int excluded
,
1900 const struct pathspec
*pathspec
)
1903 * WARNING: From this function, you can return path_recurse or you
1904 * can call read_directory_recursive() (or neither), but
1905 * you CAN'T DO BOTH.
1907 enum path_treatment state
;
1908 int matches_how
= 0;
1909 int nested_repo
= 0, check_only
, stop_early
;
1910 int old_ignored_nr
, old_untracked_nr
;
1911 /* The "len-1" is to strip the final '/' */
1912 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1914 if (status
== index_directory
)
1915 return path_recurse
;
1916 if (status
== index_gitdir
)
1918 if (status
!= index_nonexistent
)
1919 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1922 * We don't want to descend into paths that don't match the necessary
1923 * patterns. Clearly, if we don't have a pathspec, then we can't check
1924 * for matching patterns. Also, if (excluded) then we know we matched
1925 * the exclusion patterns so as an optimization we can skip checking
1926 * for matching patterns.
1928 if (pathspec
&& !excluded
) {
1929 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1933 DO_MATCH_LEADING_PATHSPEC
);
1939 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1940 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1941 struct strbuf sb
= STRBUF_INIT
;
1942 strbuf_addstr(&sb
, dirname
);
1943 nested_repo
= is_nonbare_repository_dir(&sb
);
1944 strbuf_release(&sb
);
1947 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1948 (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
))
1950 return excluded
? path_excluded
: path_untracked
;
1953 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1955 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1956 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1959 * This is an excluded directory and we are
1960 * showing ignored paths that match an exclude
1961 * pattern. (e.g. show directory as ignored
1962 * only if it matches an exclude pattern).
1963 * This path will either be 'path_excluded`
1964 * (if we are showing empty directories or if
1965 * the directory is not empty), or will be
1966 * 'path_none' (empty directory, and we are
1967 * not showing empty directories).
1969 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1970 return path_excluded
;
1972 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1973 untracked
, 1, 1, pathspec
) == path_excluded
)
1974 return path_excluded
;
1978 return path_recurse
;
1981 assert(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
);
1984 * If we have a pathspec which could match something _below_ this
1985 * directory (e.g. when checking 'subdir/' having a pathspec like
1986 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1989 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1990 return path_recurse
;
1992 /* Special cases for where this directory is excluded/ignored */
1995 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1996 * hiding empty directories, there is no need to
1997 * recurse into an ignored directory.
1999 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2000 return path_excluded
;
2003 * Even if we are hiding empty directories, we can still avoid
2004 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
2005 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
2007 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2008 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
2009 return path_excluded
;
2013 * Other than the path_recurse case above, we only need to
2014 * recurse into untracked directories if any of the following
2016 * - DIR_SHOW_IGNORED (because then we need to determine if
2017 * there are ignored entries below)
2018 * - DIR_SHOW_IGNORED_TOO (same as above)
2019 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
2020 * the directory is empty)
2023 !(dir
->flags
& (DIR_SHOW_IGNORED
|
2024 DIR_SHOW_IGNORED_TOO
|
2025 DIR_HIDE_EMPTY_DIRECTORIES
))) {
2026 return path_untracked
;
2030 * Even if we don't want to know all the paths under an untracked or
2031 * ignored directory, we may still need to go into the directory to
2032 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2033 * an empty directory should be path_none instead of path_excluded or
2036 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
2037 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
2040 * However, there's another optimization possible as a subset of
2041 * check_only, based on the cases we have to consider:
2042 * A) Directory matches no exclude patterns:
2043 * * Directory is empty => path_none
2044 * * Directory has an untracked file under it => path_untracked
2045 * * Directory has only ignored files under it => path_excluded
2046 * B) Directory matches an exclude pattern:
2047 * * Directory is empty => path_none
2048 * * Directory has an untracked file under it => path_excluded
2049 * * Directory has only ignored files under it => path_excluded
2050 * In case A, we can exit as soon as we've found an untracked
2051 * file but otherwise have to walk all files. In case B, though,
2052 * we can stop at the first file we find under the directory.
2054 stop_early
= check_only
&& excluded
;
2057 * If /every/ file within an untracked directory is ignored, then
2058 * we want to treat the directory as ignored (for e.g. status
2059 * --porcelain), without listing the individual ignored files
2060 * underneath. To do so, we'll save the current ignored_nr, and
2061 * pop all the ones added after it if it turns out the entire
2062 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2063 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2064 * untracked paths so will need to pop all those off the last
2065 * after we traverse.
2067 old_ignored_nr
= dir
->ignored_nr
;
2068 old_untracked_nr
= dir
->nr
;
2070 /* Actually recurse into dirname now, we'll fixup the state later. */
2071 untracked
= lookup_untracked(dir
->untracked
, untracked
,
2072 dirname
+ baselen
, len
- baselen
);
2073 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
2074 check_only
, stop_early
, pathspec
);
2076 /* There are a variety of reasons we may need to fixup the state... */
2077 if (state
== path_excluded
) {
2078 /* state == path_excluded implies all paths under
2079 * dirname were ignored...
2081 * if running e.g. `git status --porcelain --ignored=matching`,
2082 * then we want to see the subpaths that are ignored.
2084 * if running e.g. just `git status --porcelain`, then
2085 * we just want the directory itself to be listed as ignored
2086 * and not the individual paths underneath.
2088 int want_ignored_subpaths
=
2089 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2090 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
2092 if (want_ignored_subpaths
) {
2094 * with --ignored=matching, we want the subpaths
2095 * INSTEAD of the directory itself.
2100 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
2101 FREE_AND_NULL(dir
->ignored
[i
]);
2102 dir
->ignored_nr
= old_ignored_nr
;
2107 * We may need to ignore some of the untracked paths we found while
2108 * traversing subdirectories.
2110 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2111 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
2113 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
2114 FREE_AND_NULL(dir
->entries
[i
]);
2115 dir
->nr
= old_untracked_nr
;
2119 * If there is nothing under the current directory and we are not
2120 * hiding empty directories, then we need to report on the
2121 * untracked or ignored status of the directory itself.
2123 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2124 state
= excluded
? path_excluded
: path_untracked
;
2130 * This is an inexact early pruning of any recursive directory
2131 * reading - if the path cannot possibly be in the pathspec,
2132 * return true, and we'll skip it early.
2134 static int simplify_away(const char *path
, int pathlen
,
2135 const struct pathspec
*pathspec
)
2139 if (!pathspec
|| !pathspec
->nr
)
2142 GUARD_PATHSPEC(pathspec
,
2151 for (i
= 0; i
< pathspec
->nr
; i
++) {
2152 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2153 int len
= item
->nowildcard_len
;
2157 if (!ps_strncmp(item
, item
->match
, path
, len
))
2165 * This function tells us whether an excluded path matches a
2166 * list of "interesting" pathspecs. That is, whether a path matched
2167 * by any of the pathspecs could possibly be ignored by excluding
2168 * the specified path. This can happen if:
2170 * 1. the path is mentioned explicitly in the pathspec
2172 * 2. the path is a directory prefix of some element in the
2175 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2176 const struct pathspec
*pathspec
)
2180 if (!pathspec
|| !pathspec
->nr
)
2183 GUARD_PATHSPEC(pathspec
,
2191 for (i
= 0; i
< pathspec
->nr
; i
++) {
2192 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2193 int len
= item
->nowildcard_len
;
2195 if (len
== pathlen
&&
2196 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2198 if (len
> pathlen
&&
2199 item
->match
[pathlen
] == '/' &&
2200 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2206 static int get_index_dtype(struct index_state
*istate
,
2207 const char *path
, int len
)
2210 const struct cache_entry
*ce
;
2212 ce
= index_file_exists(istate
, path
, len
, 0);
2214 if (!ce_uptodate(ce
))
2216 if (S_ISGITLINK(ce
->ce_mode
))
2219 * Nobody actually cares about the
2220 * difference between DT_LNK and DT_REG
2225 /* Try to look it up as a directory */
2226 pos
= index_name_pos(istate
, path
, len
);
2230 while (pos
< istate
->cache_nr
) {
2231 ce
= istate
->cache
[pos
++];
2232 if (strncmp(ce
->name
, path
, len
))
2234 if (ce
->name
[len
] > '/')
2236 if (ce
->name
[len
] < '/')
2238 if (!ce_uptodate(ce
))
2239 break; /* continue? */
2245 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2246 const char *path
, int len
)
2250 if (dtype
!= DT_UNKNOWN
)
2252 dtype
= get_index_dtype(istate
, path
, len
);
2253 if (dtype
!= DT_UNKNOWN
)
2255 if (lstat(path
, &st
))
2257 if (S_ISREG(st
.st_mode
))
2259 if (S_ISDIR(st
.st_mode
))
2261 if (S_ISLNK(st
.st_mode
))
2266 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2267 struct cached_dir
*cdir
,
2268 struct index_state
*istate
,
2269 struct strbuf
*path
,
2271 const struct pathspec
*pathspec
)
2274 * WARNING: From this function, you can return path_recurse or you
2275 * can call read_directory_recursive() (or neither), but
2276 * you CAN'T DO BOTH.
2278 strbuf_setlen(path
, baselen
);
2280 strbuf_addstr(path
, cdir
->file
);
2281 return path_untracked
;
2283 strbuf_addstr(path
, cdir
->ucd
->name
);
2284 /* treat_one_path() does this before it calls treat_directory() */
2285 strbuf_complete(path
, '/');
2286 if (cdir
->ucd
->check_only
)
2288 * check_only is set as a result of treat_directory() getting
2289 * to its bottom. Verify again the same set of directories
2290 * with check_only set.
2292 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2293 cdir
->ucd
, 1, 0, pathspec
);
2295 * We get path_recurse in the first run when
2296 * directory_exists_in_index() returns index_nonexistent. We
2297 * are sure that new changes in the index does not impact the
2298 * outcome. Return now.
2300 return path_recurse
;
2303 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2304 struct untracked_cache_dir
*untracked
,
2305 struct cached_dir
*cdir
,
2306 struct index_state
*istate
,
2307 struct strbuf
*path
,
2309 const struct pathspec
*pathspec
)
2311 int has_path_in_index
, dtype
, excluded
;
2314 return treat_path_fast(dir
, cdir
, istate
, path
,
2316 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2318 strbuf_setlen(path
, baselen
);
2319 strbuf_addstr(path
, cdir
->d_name
);
2320 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2323 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2325 /* Always exclude indexed files */
2326 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2328 if (dtype
!= DT_DIR
&& has_path_in_index
)
2332 * When we are looking at a directory P in the working tree,
2333 * there are three cases:
2335 * (1) P exists in the index. Everything inside the directory P in
2336 * the working tree needs to go when P is checked out from the
2339 * (2) P does not exist in the index, but there is P/Q in the index.
2340 * We know P will stay a directory when we check out the contents
2341 * of the index, but we do not know yet if there is a directory
2342 * P/Q in the working tree to be killed, so we need to recurse.
2344 * (3) P does not exist in the index, and there is no P/Q in the index
2345 * to require P to be a directory, either. Only in this case, we
2346 * know that everything inside P will not be killed without
2349 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2350 (dtype
== DT_DIR
) &&
2351 !has_path_in_index
&&
2352 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2355 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2358 * Excluded? If we don't explicitly want to show
2359 * ignored files, ignore it
2361 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2362 return path_excluded
;
2369 * WARNING: Do not ignore/amend the return value from
2370 * treat_directory(), and especially do not change it to return
2371 * path_recurse as that can cause exponential slowdown.
2372 * Instead, modify treat_directory() to return the right value.
2374 strbuf_addch(path
, '/');
2375 return treat_directory(dir
, istate
, untracked
,
2376 path
->buf
, path
->len
,
2377 baselen
, excluded
, pathspec
);
2381 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2382 0 /* prefix */, NULL
/* seen */,
2386 return path_excluded
;
2387 return path_untracked
;
2391 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2395 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2396 dir
->untracked_alloc
);
2397 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2400 static int valid_cached_dir(struct dir_struct
*dir
,
2401 struct untracked_cache_dir
*untracked
,
2402 struct index_state
*istate
,
2403 struct strbuf
*path
,
2412 * With fsmonitor, we can trust the untracked cache's valid field.
2414 refresh_fsmonitor(istate
);
2415 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2416 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2417 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2420 if (!untracked
->valid
||
2421 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2422 fill_stat_data(&untracked
->stat_data
, &st
);
2427 if (untracked
->check_only
!= !!check_only
)
2431 * prep_exclude will be called eventually on this directory,
2432 * but it's called much later in last_matching_pattern(). We
2433 * need it now to determine the validity of the cache for this
2434 * path. The next calls will be nearly no-op, the way
2435 * prep_exclude() is designed.
2437 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2438 strbuf_addch(path
, '/');
2439 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2440 strbuf_setlen(path
, path
->len
- 1);
2442 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2444 /* hopefully prep_exclude() haven't invalidated this entry... */
2445 return untracked
->valid
;
2448 static int open_cached_dir(struct cached_dir
*cdir
,
2449 struct dir_struct
*dir
,
2450 struct untracked_cache_dir
*untracked
,
2451 struct index_state
*istate
,
2452 struct strbuf
*path
,
2457 memset(cdir
, 0, sizeof(*cdir
));
2458 cdir
->untracked
= untracked
;
2459 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2461 c_path
= path
->len
? path
->buf
: ".";
2462 cdir
->fdir
= opendir(c_path
);
2464 warning_errno(_("could not open directory '%s'"), c_path
);
2465 if (dir
->untracked
) {
2466 invalidate_directory(dir
->untracked
, untracked
);
2467 dir
->untracked
->dir_opened
++;
2474 static int read_cached_dir(struct cached_dir
*cdir
)
2479 de
= readdir_skip_dot_and_dotdot(cdir
->fdir
);
2481 cdir
->d_name
= NULL
;
2482 cdir
->d_type
= DT_UNKNOWN
;
2485 cdir
->d_name
= de
->d_name
;
2486 cdir
->d_type
= DTYPE(de
);
2489 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2490 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2500 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2501 struct untracked_cache_dir
*d
= cdir
->untracked
;
2502 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2508 static void close_cached_dir(struct cached_dir
*cdir
)
2511 closedir(cdir
->fdir
);
2513 * We have gone through this directory and found no untracked
2514 * entries. Mark it valid.
2516 if (cdir
->untracked
) {
2517 cdir
->untracked
->valid
= 1;
2518 cdir
->untracked
->recurse
= 1;
2522 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2523 struct untracked_cache_dir
*untracked
,
2524 struct cached_dir
*cdir
,
2525 struct index_state
*istate
,
2526 struct strbuf
*path
,
2528 const struct pathspec
*pathspec
,
2529 enum path_treatment state
)
2531 /* add the path to the appropriate result list */
2534 if (dir
->flags
& DIR_SHOW_IGNORED
)
2535 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2536 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2537 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2538 exclude_matches_pathspec(path
->buf
, path
->len
,
2540 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2543 case path_untracked
:
2544 if (dir
->flags
& DIR_SHOW_IGNORED
)
2546 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2548 add_untracked(untracked
, path
->buf
+ baselen
);
2557 * Read a directory tree. We currently ignore anything but
2558 * directories, regular files and symlinks. That's because git
2559 * doesn't handle them at all yet. Maybe that will change some
2562 * Also, we ignore the name ".git" (even if it is not a directory).
2563 * That likely will not change.
2565 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2566 * to signal that a file was found. This is the least significant value that
2567 * indicates that a file was encountered that does not depend on the order of
2568 * whether an untracked or excluded path was encountered first.
2570 * Returns the most significant path_treatment value encountered in the scan.
2571 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2572 * significant path_treatment value that will be returned.
2575 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2576 struct index_state
*istate
, const char *base
, int baselen
,
2577 struct untracked_cache_dir
*untracked
, int check_only
,
2578 int stop_at_first_file
, const struct pathspec
*pathspec
)
2581 * WARNING: Do NOT recurse unless path_recurse is returned from
2582 * treat_path(). Recursing on any other return value
2583 * can result in exponential slowdown.
2585 struct cached_dir cdir
;
2586 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2587 struct strbuf path
= STRBUF_INIT
;
2589 strbuf_add(&path
, base
, baselen
);
2591 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2593 dir
->visited_directories
++;
2596 untracked
->check_only
= !!check_only
;
2598 while (!read_cached_dir(&cdir
)) {
2599 /* check how the file or directory should be treated */
2600 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2602 dir
->visited_paths
++;
2604 if (state
> dir_state
)
2607 /* recurse into subdir if instructed by treat_path */
2608 if (state
== path_recurse
) {
2609 struct untracked_cache_dir
*ud
;
2610 ud
= lookup_untracked(dir
->untracked
, untracked
,
2612 path
.len
- baselen
);
2614 read_directory_recursive(dir
, istate
, path
.buf
,
2616 check_only
, stop_at_first_file
, pathspec
);
2617 if (subdir_state
> dir_state
)
2618 dir_state
= subdir_state
;
2621 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2622 0 /* prefix */, NULL
,
2623 0 /* do NOT special case dirs */))
2628 if (stop_at_first_file
) {
2630 * If stopping at first file, then
2631 * signal that a file was found by
2632 * returning `path_excluded`. This is
2633 * to return a consistent value
2634 * regardless of whether an ignored or
2635 * excluded file happened to be
2638 * In current usage, the
2639 * `stop_at_first_file` is passed when
2640 * an ancestor directory has matched
2641 * an exclude pattern, so any found
2642 * files will be excluded.
2644 if (dir_state
>= path_excluded
) {
2645 dir_state
= path_excluded
;
2650 /* abort early if maximum state has been reached */
2651 if (dir_state
== path_untracked
) {
2653 add_untracked(untracked
, path
.buf
+ baselen
);
2656 /* skip the add_path_to_appropriate_result_list() */
2660 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2661 istate
, &path
, baselen
,
2664 close_cached_dir(&cdir
);
2666 strbuf_release(&path
);
2671 int cmp_dir_entry(const void *p1
, const void *p2
)
2673 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2674 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2676 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2679 /* check if *out lexically strictly contains *in */
2680 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2682 return (out
->len
< in
->len
) &&
2683 (out
->name
[out
->len
- 1] == '/') &&
2684 !memcmp(out
->name
, in
->name
, out
->len
);
2687 static int treat_leading_path(struct dir_struct
*dir
,
2688 struct index_state
*istate
,
2689 const char *path
, int len
,
2690 const struct pathspec
*pathspec
)
2692 struct strbuf sb
= STRBUF_INIT
;
2693 struct strbuf subdir
= STRBUF_INIT
;
2694 int prevlen
, baselen
;
2696 struct cached_dir cdir
;
2697 enum path_treatment state
= path_none
;
2700 * For each directory component of path, we are going to check whether
2701 * that path is relevant given the pathspec. For example, if path is
2703 * then we will ask treat_path() whether we should go into foo, then
2704 * whether we should go into bar, then whether baz is relevant.
2705 * Checking each is important because e.g. if path is
2707 * then we need to check .git to know we shouldn't traverse it.
2708 * If the return from treat_path() is:
2709 * * path_none, for any path, we return false.
2710 * * path_recurse, for all path components, we return true
2711 * * <anything else> for some intermediate component, we make sure
2712 * to add that path to the relevant list but return false
2713 * signifying that we shouldn't recurse into it.
2716 while (len
&& path
[len
- 1] == '/')
2721 memset(&cdir
, 0, sizeof(cdir
));
2722 cdir
.d_type
= DT_DIR
;
2726 prevlen
= baselen
+ !!baselen
;
2727 cp
= path
+ prevlen
;
2728 cp
= memchr(cp
, '/', path
+ len
- cp
);
2732 baselen
= cp
- path
;
2734 strbuf_add(&sb
, path
, baselen
);
2735 if (!is_directory(sb
.buf
))
2738 strbuf_add(&sb
, path
, prevlen
);
2739 strbuf_reset(&subdir
);
2740 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2741 cdir
.d_name
= subdir
.buf
;
2742 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2744 if (state
!= path_recurse
)
2745 break; /* do not recurse into it */
2747 break; /* finished checking */
2749 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2750 &sb
, baselen
, pathspec
,
2753 strbuf_release(&subdir
);
2754 strbuf_release(&sb
);
2755 return state
== path_recurse
;
2758 static const char *get_ident_string(void)
2760 static struct strbuf sb
= STRBUF_INIT
;
2765 if (uname(&uts
) < 0)
2766 die_errno(_("failed to get kernel name and information"));
2767 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2772 static int ident_in_untracked(const struct untracked_cache
*uc
)
2775 * Previous git versions may have saved many NUL separated
2776 * strings in the "ident" field, but it is insane to manage
2777 * many locations, so just take care of the first one.
2780 return !strcmp(uc
->ident
.buf
, get_ident_string());
2783 static void set_untracked_ident(struct untracked_cache
*uc
)
2785 strbuf_reset(&uc
->ident
);
2786 strbuf_addstr(&uc
->ident
, get_ident_string());
2789 * This strbuf used to contain a list of NUL separated
2790 * strings, so save NUL too for backward compatibility.
2792 strbuf_addch(&uc
->ident
, 0);
2795 static void new_untracked_cache(struct index_state
*istate
)
2797 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2798 strbuf_init(&uc
->ident
, 100);
2799 uc
->exclude_per_dir
= ".gitignore";
2800 /* should be the same flags used by git-status */
2801 uc
->dir_flags
= DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2802 set_untracked_ident(uc
);
2803 istate
->untracked
= uc
;
2804 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2807 void add_untracked_cache(struct index_state
*istate
)
2809 if (!istate
->untracked
) {
2810 new_untracked_cache(istate
);
2812 if (!ident_in_untracked(istate
->untracked
)) {
2813 free_untracked_cache(istate
->untracked
);
2814 new_untracked_cache(istate
);
2819 void remove_untracked_cache(struct index_state
*istate
)
2821 if (istate
->untracked
) {
2822 free_untracked_cache(istate
->untracked
);
2823 istate
->untracked
= NULL
;
2824 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2828 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2830 const struct pathspec
*pathspec
)
2832 struct untracked_cache_dir
*root
;
2833 static int untracked_cache_disabled
= -1;
2835 if (!dir
->untracked
)
2837 if (untracked_cache_disabled
< 0)
2838 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2839 if (untracked_cache_disabled
)
2843 * We only support $GIT_DIR/info/exclude and core.excludesfile
2844 * as the global ignore rule files. Any other additions
2845 * (e.g. from command line) invalidate the cache. This
2846 * condition also catches running setup_standard_excludes()
2847 * before setting dir->untracked!
2849 if (dir
->unmanaged_exclude_files
)
2853 * Optimize for the main use case only: whole-tree git
2854 * status. More work involved in treat_leading_path() if we
2855 * use cache on just a subset of the worktree. pathspec
2856 * support could make the matter even worse.
2858 if (base_len
|| (pathspec
&& pathspec
->nr
))
2861 /* Different set of flags may produce different results */
2862 if (dir
->flags
!= dir
->untracked
->dir_flags
||
2864 * See treat_directory(), case index_nonexistent. Without
2865 * this flag, we may need to also cache .git file content
2866 * for the resolve_gitlink_ref() call, which we don't.
2868 !(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
) ||
2869 /* We don't support collecting ignore files */
2870 (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2871 DIR_COLLECT_IGNORED
)))
2875 * If we use .gitignore in the cache and now you change it to
2876 * .gitexclude, everything will go wrong.
2878 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2879 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2883 * EXC_CMDL is not considered in the cache. If people set it,
2886 if (dir
->exclude_list_group
[EXC_CMDL
].nr
)
2889 if (!ident_in_untracked(dir
->untracked
)) {
2890 warning(_("untracked cache is disabled on this system or location"));
2894 if (!dir
->untracked
->root
)
2895 FLEX_ALLOC_STR(dir
->untracked
->root
, name
, "");
2897 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2898 root
= dir
->untracked
->root
;
2899 if (!oideq(&dir
->ss_info_exclude
.oid
,
2900 &dir
->untracked
->ss_info_exclude
.oid
)) {
2901 invalidate_gitignore(dir
->untracked
, root
);
2902 dir
->untracked
->ss_info_exclude
= dir
->ss_info_exclude
;
2904 if (!oideq(&dir
->ss_excludes_file
.oid
,
2905 &dir
->untracked
->ss_excludes_file
.oid
)) {
2906 invalidate_gitignore(dir
->untracked
, root
);
2907 dir
->untracked
->ss_excludes_file
= dir
->ss_excludes_file
;
2910 /* Make sure this directory is not dropped out at saving phase */
2915 static void emit_traversal_statistics(struct dir_struct
*dir
,
2916 struct repository
*repo
,
2920 if (!trace2_is_enabled())
2924 trace2_data_string("read_directory", repo
, "path", "");
2926 struct strbuf tmp
= STRBUF_INIT
;
2927 strbuf_add(&tmp
, path
, path_len
);
2928 trace2_data_string("read_directory", repo
, "path", tmp
.buf
);
2929 strbuf_release(&tmp
);
2932 trace2_data_intmax("read_directory", repo
,
2933 "directories-visited", dir
->visited_directories
);
2934 trace2_data_intmax("read_directory", repo
,
2935 "paths-visited", dir
->visited_paths
);
2937 if (!dir
->untracked
)
2939 trace2_data_intmax("read_directory", repo
,
2940 "node-creation", dir
->untracked
->dir_created
);
2941 trace2_data_intmax("read_directory", repo
,
2942 "gitignore-invalidation",
2943 dir
->untracked
->gitignore_invalidated
);
2944 trace2_data_intmax("read_directory", repo
,
2945 "directory-invalidation",
2946 dir
->untracked
->dir_invalidated
);
2947 trace2_data_intmax("read_directory", repo
,
2948 "opendir", dir
->untracked
->dir_opened
);
2951 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
2952 const char *path
, int len
, const struct pathspec
*pathspec
)
2954 struct untracked_cache_dir
*untracked
;
2956 trace2_region_enter("dir", "read_directory", istate
->repo
);
2957 dir
->visited_paths
= 0;
2958 dir
->visited_directories
= 0;
2960 if (has_symlink_leading_path(path
, len
)) {
2961 trace2_region_leave("dir", "read_directory", istate
->repo
);
2965 untracked
= validate_untracked_cache(dir
, len
, pathspec
);
2968 * make sure untracked cache code path is disabled,
2969 * e.g. prep_exclude()
2971 dir
->untracked
= NULL
;
2972 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
2973 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
2974 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
2975 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
2977 emit_traversal_statistics(dir
, istate
->repo
, path
, len
);
2979 trace2_region_leave("dir", "read_directory", istate
->repo
);
2980 if (dir
->untracked
) {
2981 static int force_untracked_cache
= -1;
2983 if (force_untracked_cache
< 0)
2984 force_untracked_cache
=
2985 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2986 if (force_untracked_cache
&&
2987 dir
->untracked
== istate
->untracked
&&
2988 (dir
->untracked
->dir_opened
||
2989 dir
->untracked
->gitignore_invalidated
||
2990 dir
->untracked
->dir_invalidated
))
2991 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2992 if (dir
->untracked
!= istate
->untracked
) {
2993 FREE_AND_NULL(dir
->untracked
);
3000 int file_exists(const char *f
)
3003 return lstat(f
, &sb
) == 0;
3006 int repo_file_exists(struct repository
*repo
, const char *path
)
3008 if (repo
!= the_repository
)
3009 BUG("do not know how to check file existence in arbitrary repo");
3011 return file_exists(path
);
3014 static int cmp_icase(char a
, char b
)
3019 return toupper(a
) - toupper(b
);
3024 * Given two normalized paths (a trailing slash is ok), if subdir is
3025 * outside dir, return -1. Otherwise return the offset in subdir that
3026 * can be used as relative path to dir.
3028 int dir_inside_of(const char *subdir
, const char *dir
)
3032 assert(dir
&& subdir
&& *dir
&& *subdir
);
3034 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
3040 /* hel[p]/me vs hel[l]/yeah */
3041 if (*dir
&& *subdir
)
3045 return !*dir
? offset
: -1; /* same dir */
3047 /* foo/[b]ar vs foo/[] */
3048 if (is_dir_sep(dir
[-1]))
3049 return is_dir_sep(subdir
[-1]) ? offset
: -1;
3051 /* foo[/]bar vs foo[] */
3052 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
3055 int is_inside_dir(const char *dir
)
3064 rc
= (dir_inside_of(cwd
, dir
) >= 0);
3069 int is_empty_dir(const char *path
)
3071 DIR *dir
= opendir(path
);
3078 e
= readdir_skip_dot_and_dotdot(dir
);
3086 char *git_url_basename(const char *repo
, int is_bundle
, int is_bare
)
3088 const char *end
= repo
+ strlen(repo
), *start
, *ptr
;
3095 start
= strstr(repo
, "://");
3102 * Skip authentication data. The stripping does happen
3103 * greedily, such that we strip up to the last '@' inside
3106 for (ptr
= start
; ptr
< end
&& !is_dir_sep(*ptr
); ptr
++) {
3112 * Strip trailing spaces, slashes and /.git
3114 while (start
< end
&& (is_dir_sep(end
[-1]) || isspace(end
[-1])))
3116 if (end
- start
> 5 && is_dir_sep(end
[-5]) &&
3117 !strncmp(end
- 4, ".git", 4)) {
3119 while (start
< end
&& is_dir_sep(end
[-1]))
3124 * Strip trailing port number if we've got only a
3125 * hostname (that is, there is no dir separator but a
3126 * colon). This check is required such that we do not
3127 * strip URI's like '/foo/bar:2222.git', which should
3128 * result in a dir '2222' being guessed due to backwards
3131 if (memchr(start
, '/', end
- start
) == NULL
3132 && memchr(start
, ':', end
- start
) != NULL
) {
3134 while (start
< ptr
&& isdigit(ptr
[-1]) && ptr
[-1] != ':')
3136 if (start
< ptr
&& ptr
[-1] == ':')
3141 * Find last component. To remain backwards compatible we
3142 * also regard colons as path separators, such that
3143 * cloning a repository 'foo:bar.git' would result in a
3144 * directory 'bar' being guessed.
3147 while (start
< ptr
&& !is_dir_sep(ptr
[-1]) && ptr
[-1] != ':')
3152 * Strip .{bundle,git}.
3155 strip_suffix_mem(start
, &len
, is_bundle
? ".bundle" : ".git");
3157 if (!len
|| (len
== 1 && *start
== '/'))
3158 die(_("No directory name could be guessed.\n"
3159 "Please specify a directory on the command line"));
3162 dir
= xstrfmt("%.*s.git", (int)len
, start
);
3164 dir
= xstrndup(start
, len
);
3166 * Replace sequences of 'control' characters and whitespace
3167 * with one ascii space, remove leading and trailing spaces.
3171 int prev_space
= 1 /* strip leading whitespace */;
3172 for (end
= dir
; *end
; ++end
) {
3174 if ((unsigned char)ch
< '\x20')
3185 if (out
> dir
&& prev_space
)
3191 void strip_dir_trailing_slashes(char *dir
)
3193 char *end
= dir
+ strlen(dir
);
3195 while (dir
< end
- 1 && is_dir_sep(end
[-1]))
3200 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
3204 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
3205 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
3206 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
3207 struct object_id submodule_head
;
3209 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
3210 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
3211 /* Do not descend and nuke a nested git work tree. */
3217 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
3218 dir
= opendir(path
->buf
);
3220 if (errno
== ENOENT
)
3221 return keep_toplevel
? -1 : 0;
3222 else if (errno
== EACCES
&& !keep_toplevel
)
3224 * An empty dir could be removable even if it
3227 return rmdir(path
->buf
);
3231 strbuf_complete(path
, '/');
3234 while ((e
= readdir_skip_dot_and_dotdot(dir
)) != NULL
) {
3237 strbuf_setlen(path
, len
);
3238 strbuf_addstr(path
, e
->d_name
);
3239 if (lstat(path
->buf
, &st
)) {
3240 if (errno
== ENOENT
)
3242 * file disappeared, which is what we
3247 } else if (S_ISDIR(st
.st_mode
)) {
3248 if (!remove_dir_recurse(path
, flag
, &kept_down
))
3249 continue; /* happy */
3250 } else if (!only_empty
&&
3251 (!unlink(path
->buf
) || errno
== ENOENT
)) {
3252 continue; /* happy, too */
3255 /* path too long, stat fails, or non-directory still exists */
3261 strbuf_setlen(path
, original_len
);
3262 if (!ret
&& !keep_toplevel
&& !kept_down
)
3263 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
3266 * report the uplevel that it is not an error that we
3267 * did not rmdir() our directory.
3273 int remove_dir_recursively(struct strbuf
*path
, int flag
)
3275 return remove_dir_recurse(path
, flag
, NULL
);
3278 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
3280 void setup_standard_excludes(struct dir_struct
*dir
)
3282 dir
->exclude_per_dir
= ".gitignore";
3284 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3286 excludes_file
= xdg_config_home("ignore");
3287 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
3288 add_patterns_from_file_1(dir
, excludes_file
,
3289 dir
->untracked
? &dir
->ss_excludes_file
: NULL
);
3291 /* per repository user preference */
3292 if (startup_info
->have_repository
) {
3293 const char *path
= git_path_info_exclude();
3294 if (!access_or_warn(path
, R_OK
, 0))
3295 add_patterns_from_file_1(dir
, path
,
3296 dir
->untracked
? &dir
->ss_info_exclude
: NULL
);
3300 char *get_sparse_checkout_filename(void)
3302 return git_pathdup("info/sparse-checkout");
3305 int get_sparse_checkout_patterns(struct pattern_list
*pl
)
3308 char *sparse_filename
= get_sparse_checkout_filename();
3310 pl
->use_cone_patterns
= core_sparse_checkout_cone
;
3311 res
= add_patterns_from_file_to_list(sparse_filename
, "", 0, pl
, NULL
, 0);
3313 free(sparse_filename
);
3317 int remove_path(const char *name
)
3321 if (unlink(name
) && !is_missing_file_error(errno
))
3324 slash
= strrchr(name
, '/');
3326 char *dirs
= xstrdup(name
);
3327 slash
= dirs
+ (slash
- name
);
3330 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3337 * Frees memory within dir which was allocated, and resets fields for further
3338 * use. Does not free dir itself.
3340 void dir_clear(struct dir_struct
*dir
)
3343 struct exclude_list_group
*group
;
3344 struct pattern_list
*pl
;
3345 struct exclude_stack
*stk
;
3346 struct dir_struct
new = DIR_INIT
;
3348 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3349 group
= &dir
->exclude_list_group
[i
];
3350 for (j
= 0; j
< group
->nr
; j
++) {
3353 free((char *)pl
->src
);
3354 clear_pattern_list(pl
);
3359 for (i
= 0; i
< dir
->ignored_nr
; i
++)
3360 free(dir
->ignored
[i
]);
3361 for (i
= 0; i
< dir
->nr
; i
++)
3362 free(dir
->entries
[i
]);
3366 stk
= dir
->exclude_stack
;
3368 struct exclude_stack
*prev
= stk
->prev
;
3372 strbuf_release(&dir
->basebuf
);
3374 memcpy(dir
, &new, sizeof(*dir
));
3377 struct ondisk_untracked_cache
{
3378 struct stat_data info_exclude_stat
;
3379 struct stat_data excludes_file_stat
;
3383 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3386 int index
; /* number of written untracked_cache_dir */
3387 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3388 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3389 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3391 struct strbuf sb_stat
;
3392 struct strbuf sb_sha1
;
3395 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3397 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3398 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3399 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3400 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3401 to
->sd_dev
= htonl(from
->sd_dev
);
3402 to
->sd_ino
= htonl(from
->sd_ino
);
3403 to
->sd_uid
= htonl(from
->sd_uid
);
3404 to
->sd_gid
= htonl(from
->sd_gid
);
3405 to
->sd_size
= htonl(from
->sd_size
);
3408 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3409 struct write_data
*wd
)
3411 struct stat_data stat_data
;
3412 struct strbuf
*out
= &wd
->out
;
3413 unsigned char intbuf
[16];
3414 unsigned int intlen
, value
;
3415 int i
= wd
->index
++;
3418 * untracked_nr should be reset whenever valid is clear, but
3421 if (!untracked
->valid
) {
3422 untracked
->untracked_nr
= 0;
3423 untracked
->check_only
= 0;
3426 if (untracked
->check_only
)
3427 ewah_set(wd
->check_only
, i
);
3428 if (untracked
->valid
) {
3429 ewah_set(wd
->valid
, i
);
3430 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3431 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3433 if (!is_null_oid(&untracked
->exclude_oid
)) {
3434 ewah_set(wd
->sha1_valid
, i
);
3435 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3436 the_hash_algo
->rawsz
);
3439 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3440 strbuf_add(out
, intbuf
, intlen
);
3442 /* skip non-recurse directories */
3443 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3444 if (untracked
->dirs
[i
]->recurse
)
3446 intlen
= encode_varint(value
, intbuf
);
3447 strbuf_add(out
, intbuf
, intlen
);
3449 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3451 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3452 strbuf_add(out
, untracked
->untracked
[i
],
3453 strlen(untracked
->untracked
[i
]) + 1);
3455 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3456 if (untracked
->dirs
[i
]->recurse
)
3457 write_one_dir(untracked
->dirs
[i
], wd
);
3460 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3462 struct ondisk_untracked_cache
*ouc
;
3463 struct write_data wd
;
3464 unsigned char varbuf
[16];
3466 const unsigned hashsz
= the_hash_algo
->rawsz
;
3468 CALLOC_ARRAY(ouc
, 1);
3469 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3470 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3471 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3473 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3474 strbuf_add(out
, varbuf
, varint_len
);
3475 strbuf_addbuf(out
, &untracked
->ident
);
3477 strbuf_add(out
, ouc
, sizeof(*ouc
));
3478 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3479 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3480 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3483 if (!untracked
->root
) {
3484 varint_len
= encode_varint(0, varbuf
);
3485 strbuf_add(out
, varbuf
, varint_len
);
3490 wd
.check_only
= ewah_new();
3491 wd
.valid
= ewah_new();
3492 wd
.sha1_valid
= ewah_new();
3493 strbuf_init(&wd
.out
, 1024);
3494 strbuf_init(&wd
.sb_stat
, 1024);
3495 strbuf_init(&wd
.sb_sha1
, 1024);
3496 write_one_dir(untracked
->root
, &wd
);
3498 varint_len
= encode_varint(wd
.index
, varbuf
);
3499 strbuf_add(out
, varbuf
, varint_len
);
3500 strbuf_addbuf(out
, &wd
.out
);
3501 ewah_serialize_strbuf(wd
.valid
, out
);
3502 ewah_serialize_strbuf(wd
.check_only
, out
);
3503 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3504 strbuf_addbuf(out
, &wd
.sb_stat
);
3505 strbuf_addbuf(out
, &wd
.sb_sha1
);
3506 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3508 ewah_free(wd
.valid
);
3509 ewah_free(wd
.check_only
);
3510 ewah_free(wd
.sha1_valid
);
3511 strbuf_release(&wd
.out
);
3512 strbuf_release(&wd
.sb_stat
);
3513 strbuf_release(&wd
.sb_sha1
);
3516 static void free_untracked(struct untracked_cache_dir
*ucd
)
3521 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3522 free_untracked(ucd
->dirs
[i
]);
3523 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3524 free(ucd
->untracked
[i
]);
3525 free(ucd
->untracked
);
3530 void free_untracked_cache(struct untracked_cache
*uc
)
3533 free_untracked(uc
->root
);
3539 struct untracked_cache_dir
**ucd
;
3540 struct ewah_bitmap
*check_only
;
3541 struct ewah_bitmap
*valid
;
3542 struct ewah_bitmap
*sha1_valid
;
3543 const unsigned char *data
;
3544 const unsigned char *end
;
3547 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3549 memcpy(to
, data
, sizeof(*to
));
3550 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3551 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3552 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3553 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3554 to
->sd_dev
= ntohl(to
->sd_dev
);
3555 to
->sd_ino
= ntohl(to
->sd_ino
);
3556 to
->sd_uid
= ntohl(to
->sd_uid
);
3557 to
->sd_gid
= ntohl(to
->sd_gid
);
3558 to
->sd_size
= ntohl(to
->sd_size
);
3561 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3562 struct read_data
*rd
)
3564 struct untracked_cache_dir ud
, *untracked
;
3565 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3566 const unsigned char *eos
;
3570 memset(&ud
, 0, sizeof(ud
));
3572 value
= decode_varint(&data
);
3576 ud
.untracked_alloc
= value
;
3577 ud
.untracked_nr
= value
;
3578 if (ud
.untracked_nr
)
3579 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3581 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3584 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3586 eos
= memchr(data
, '\0', end
- data
);
3587 if (!eos
|| eos
== end
)
3590 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3591 memcpy(untracked
, &ud
, sizeof(ud
));
3592 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3595 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3596 eos
= memchr(data
, '\0', end
- data
);
3597 if (!eos
|| eos
== end
)
3599 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3603 rd
->ucd
[rd
->index
++] = untracked
;
3606 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3607 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3613 static void set_check_only(size_t pos
, void *cb
)
3615 struct read_data
*rd
= cb
;
3616 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3620 static void read_stat(size_t pos
, void *cb
)
3622 struct read_data
*rd
= cb
;
3623 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3624 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3625 rd
->data
= rd
->end
+ 1;
3628 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3629 rd
->data
+= sizeof(struct stat_data
);
3633 static void read_oid(size_t pos
, void *cb
)
3635 struct read_data
*rd
= cb
;
3636 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3637 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3638 rd
->data
= rd
->end
+ 1;
3641 oidread(&ud
->exclude_oid
, rd
->data
);
3642 rd
->data
+= the_hash_algo
->rawsz
;
3645 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3646 const unsigned char *sha1
)
3648 stat_data_from_disk(&oid_stat
->stat
, data
);
3649 oidread(&oid_stat
->oid
, sha1
);
3650 oid_stat
->valid
= 1;
3653 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3655 struct untracked_cache
*uc
;
3656 struct read_data rd
;
3657 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3661 const char *exclude_per_dir
;
3662 const unsigned hashsz
= the_hash_algo
->rawsz
;
3663 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3664 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3666 if (sz
<= 1 || end
[-1] != '\0')
3670 ident_len
= decode_varint(&next
);
3671 if (next
+ ident_len
> end
)
3673 ident
= (const char *)next
;
3676 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3679 CALLOC_ARRAY(uc
, 1);
3680 strbuf_init(&uc
->ident
, ident_len
);
3681 strbuf_add(&uc
->ident
, ident
, ident_len
);
3682 load_oid_stat(&uc
->ss_info_exclude
,
3683 next
+ ouc_offset(info_exclude_stat
),
3685 load_oid_stat(&uc
->ss_excludes_file
,
3686 next
+ ouc_offset(excludes_file_stat
),
3687 next
+ offset
+ hashsz
);
3688 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3689 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3690 uc
->exclude_per_dir
= xstrdup(exclude_per_dir
);
3691 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3692 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3696 len
= decode_varint(&next
);
3697 if (next
> end
|| len
== 0)
3700 rd
.valid
= ewah_new();
3701 rd
.check_only
= ewah_new();
3702 rd
.sha1_valid
= ewah_new();
3706 ALLOC_ARRAY(rd
.ucd
, len
);
3708 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3712 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3717 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3722 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3726 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3727 rd
.data
= next
+ len
;
3728 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3729 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3734 ewah_free(rd
.valid
);
3735 ewah_free(rd
.check_only
);
3736 ewah_free(rd
.sha1_valid
);
3739 free_untracked_cache(uc
);
3745 static void invalidate_one_directory(struct untracked_cache
*uc
,
3746 struct untracked_cache_dir
*ucd
)
3748 uc
->dir_invalidated
++;
3750 ucd
->untracked_nr
= 0;
3754 * Normally when an entry is added or removed from a directory,
3755 * invalidating that directory is enough. No need to touch its
3756 * ancestors. When a directory is shown as "foo/bar/" in git-status
3757 * however, deleting or adding an entry may have cascading effect.
3759 * Say the "foo/bar/file" has become untracked, we need to tell the
3760 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3761 * directory any more (because "bar" is managed by foo as an untracked
3764 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3765 * was the last untracked entry in the entire "foo", we should show
3766 * "foo/" instead. Which means we have to invalidate past "bar" up to
3769 * This function traverses all directories from root to leaf. If there
3770 * is a chance of one of the above cases happening, we invalidate back
3771 * to root. Otherwise we just invalidate the leaf. There may be a more
3772 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3773 * detect these cases and avoid unnecessary invalidation, for example,
3774 * checking for the untracked entry named "bar/" in "foo", but for now
3775 * stick to something safe and simple.
3777 static int invalidate_one_component(struct untracked_cache
*uc
,
3778 struct untracked_cache_dir
*dir
,
3779 const char *path
, int len
)
3781 const char *rest
= strchr(path
, '/');
3784 int component_len
= rest
- path
;
3785 struct untracked_cache_dir
*d
=
3786 lookup_untracked(uc
, dir
, path
, component_len
);
3788 invalidate_one_component(uc
, d
, rest
+ 1,
3789 len
- (component_len
+ 1));
3791 invalidate_one_directory(uc
, dir
);
3795 invalidate_one_directory(uc
, dir
);
3796 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3799 void untracked_cache_invalidate_path(struct index_state
*istate
,
3800 const char *path
, int safe_path
)
3802 if (!istate
->untracked
|| !istate
->untracked
->root
)
3804 if (!safe_path
&& !verify_path(path
, 0))
3806 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3807 path
, strlen(path
));
3810 void untracked_cache_remove_from_index(struct index_state
*istate
,
3813 untracked_cache_invalidate_path(istate
, path
, 1);
3816 void untracked_cache_add_to_index(struct index_state
*istate
,
3819 untracked_cache_invalidate_path(istate
, path
, 1);
3822 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3823 const char *sub_gitdir
)
3826 struct repository subrepo
;
3827 struct strbuf sub_wt
= STRBUF_INIT
;
3828 struct strbuf sub_gd
= STRBUF_INIT
;
3830 const struct submodule
*sub
;
3832 /* If the submodule has no working tree, we can ignore it. */
3833 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3836 if (repo_read_index(&subrepo
) < 0)
3837 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3839 /* TODO: audit for interaction with sparse-index. */
3840 ensure_full_index(subrepo
.index
);
3841 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3842 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3844 if (!S_ISGITLINK(ce
->ce_mode
))
3847 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3848 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3850 * Skip entries with the same name in different stages
3851 * to make sure an entry is returned only once.
3855 sub
= submodule_from_path(&subrepo
, null_oid(), ce
->name
);
3856 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3857 /* .gitmodules broken or inactive sub */
3860 strbuf_reset(&sub_wt
);
3861 strbuf_reset(&sub_gd
);
3862 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3863 submodule_name_to_gitdir(&sub_gd
, &subrepo
, sub
->name
);
3865 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3867 strbuf_release(&sub_wt
);
3868 strbuf_release(&sub_gd
);
3869 repo_clear(&subrepo
);
3872 void connect_work_tree_and_git_dir(const char *work_tree_
,
3873 const char *git_dir_
,
3874 int recurse_into_nested
)
3876 struct strbuf gitfile_sb
= STRBUF_INIT
;
3877 struct strbuf cfg_sb
= STRBUF_INIT
;
3878 struct strbuf rel_path
= STRBUF_INIT
;
3879 char *git_dir
, *work_tree
;
3881 /* Prepare .git file */
3882 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3883 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3884 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3886 /* Prepare config file */
3887 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3888 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3889 die(_("could not create directories for %s"), cfg_sb
.buf
);
3891 git_dir
= real_pathdup(git_dir_
, 1);
3892 work_tree
= real_pathdup(work_tree_
, 1);
3894 /* Write .git file */
3895 write_file(gitfile_sb
.buf
, "gitdir: %s",
3896 relative_path(git_dir
, work_tree
, &rel_path
));
3897 /* Update core.worktree setting */
3898 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3899 relative_path(work_tree
, git_dir
, &rel_path
));
3901 strbuf_release(&gitfile_sb
);
3902 strbuf_release(&cfg_sb
);
3903 strbuf_release(&rel_path
);
3905 if (recurse_into_nested
)
3906 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3913 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3915 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3917 if (rename(old_git_dir
, new_git_dir
) < 0)
3918 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3919 old_git_dir
, new_git_dir
);
3921 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);