2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
8 #include "git-compat-util.h"
14 #include "environment.h"
16 #include "name-hash.h"
17 #include "object-file.h"
18 #include "object-store-ll.h"
22 #include "wildmatch.h"
26 #include "ewah/ewok.h"
27 #include "fsmonitor-ll.h"
28 #include "read-cache-ll.h"
30 #include "sparse-index.h"
31 #include "submodule-config.h"
38 * Tells read_directory_recursive how a file or directory should be treated.
39 * Values are ordered by significance, e.g. if a directory contains both
40 * excluded and untracked files, it is listed as untracked because
41 * path_untracked > path_excluded.
51 * Support data structure for our opendir/readdir/closedir wrappers
55 struct untracked_cache_dir
*untracked
;
62 struct untracked_cache_dir
*ucd
;
65 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
66 struct index_state
*istate
, const char *path
, int len
,
67 struct untracked_cache_dir
*untracked
,
68 int check_only
, int stop_at_first_file
, const struct pathspec
*pathspec
);
69 static int resolve_dtype(int dtype
, struct index_state
*istate
,
70 const char *path
, int len
);
71 struct dirent
*readdir_skip_dot_and_dotdot(DIR *dirp
)
75 while ((e
= readdir(dirp
)) != NULL
) {
76 if (!is_dot_or_dotdot(e
->d_name
))
82 int count_slashes(const char *s
)
91 int fspathcmp(const char *a
, const char *b
)
93 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
96 int fspatheq(const char *a
, const char *b
)
98 return !fspathcmp(a
, b
);
101 int fspathncmp(const char *a
, const char *b
, size_t count
)
103 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
106 unsigned int fspathhash(const char *str
)
108 return ignore_case
? strihash(str
) : strhash(str
);
111 int git_fnmatch(const struct pathspec_item
*item
,
112 const char *pattern
, const char *string
,
116 if (ps_strncmp(item
, pattern
, string
, prefix
))
121 if (item
->flags
& PATHSPEC_ONESTAR
) {
122 int pattern_len
= strlen(++pattern
);
123 int string_len
= strlen(string
);
124 return string_len
< pattern_len
||
125 ps_strcmp(item
, pattern
,
126 string
+ string_len
- pattern_len
);
128 if (item
->magic
& PATHSPEC_GLOB
)
129 return wildmatch(pattern
, string
,
131 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
133 /* wildmatch has not learned no FNM_PATHNAME mode yet */
134 return wildmatch(pattern
, string
,
135 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
138 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
139 const char *string
, int stringlen
,
143 struct strbuf pat_buf
= STRBUF_INIT
;
144 struct strbuf str_buf
= STRBUF_INIT
;
145 const char *use_pat
= pattern
;
146 const char *use_str
= string
;
148 if (pattern
[patternlen
]) {
149 strbuf_add(&pat_buf
, pattern
, patternlen
);
150 use_pat
= pat_buf
.buf
;
152 if (string
[stringlen
]) {
153 strbuf_add(&str_buf
, string
, stringlen
);
154 use_str
= str_buf
.buf
;
158 flags
|= WM_CASEFOLD
;
159 match_status
= wildmatch(use_pat
, use_str
, flags
);
161 strbuf_release(&pat_buf
);
162 strbuf_release(&str_buf
);
167 static size_t common_prefix_len(const struct pathspec
*pathspec
)
173 * ":(icase)path" is treated as a pathspec full of
174 * wildcard. In other words, only prefix is considered common
175 * prefix. If the pathspec is abc/foo abc/bar, running in
176 * subdir xyz, the common prefix is still xyz, not xyz/abc as
179 GUARD_PATHSPEC(pathspec
,
188 for (n
= 0; n
< pathspec
->nr
; n
++) {
189 size_t i
= 0, len
= 0, item_len
;
190 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
192 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
193 item_len
= pathspec
->items
[n
].prefix
;
195 item_len
= pathspec
->items
[n
].nowildcard_len
;
196 while (i
< item_len
&& (n
== 0 || i
< max
)) {
197 char c
= pathspec
->items
[n
].match
[i
];
198 if (c
!= pathspec
->items
[0].match
[i
])
204 if (n
== 0 || len
< max
) {
214 * Returns a copy of the longest leading path common among all
217 char *common_prefix(const struct pathspec
*pathspec
)
219 unsigned long len
= common_prefix_len(pathspec
);
221 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
224 int fill_directory(struct dir_struct
*dir
,
225 struct index_state
*istate
,
226 const struct pathspec
*pathspec
)
231 unsigned exclusive_flags
= DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
;
232 if ((dir
->flags
& exclusive_flags
) == exclusive_flags
)
233 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
236 * Calculate common prefix for the pathspec, and
237 * use that to optimize the directory walk
239 prefix_len
= common_prefix_len(pathspec
);
240 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
242 /* Read the directory and prune it */
243 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
248 int within_depth(const char *name
, int namelen
,
249 int depth
, int max_depth
)
251 const char *cp
= name
, *cpe
= name
+ namelen
;
257 if (depth
> max_depth
)
264 * Read the contents of the blob with the given OID into a buffer.
265 * Append a trailing LF to the end if the last line doesn't have one.
268 * -1 when the OID is invalid or unknown or does not refer to a blob.
269 * 0 when the blob is empty.
270 * 1 along with { data, size } of the (possibly augmented) buffer
273 * Optionally updates the given oid_stat with the given OID (when valid).
275 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
276 size_t *size_out
, char **data_out
)
278 enum object_type type
;
285 data
= repo_read_object_file(the_repository
, oid
, &type
, &sz
);
286 if (!data
|| type
!= OBJ_BLOB
) {
292 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
293 oidcpy(&oid_stat
->oid
, oid
);
301 if (data
[sz
- 1] != '\n') {
302 data
= xrealloc(data
, st_add(sz
, 1));
306 *size_out
= xsize_t(sz
);
312 #define DO_MATCH_EXCLUDE (1<<0)
313 #define DO_MATCH_DIRECTORY (1<<1)
314 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
317 * Does the given pathspec match the given name? A match is found if
319 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
320 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
321 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
322 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
324 * Return value tells which case it was (1-4), or 0 when there is no match.
326 * It may be instructive to look at a small table of concrete examples
327 * to understand the differences between 1, 2, and 4:
330 * | a/b | a/b/ | a/b/c
331 * ------+-----------+-----------+------------
332 * a/b | EXACT | EXACT[1] | LEADING[2]
333 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
334 * a/b/c | RECURSIVE | RECURSIVE | EXACT
336 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
337 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
339 static int match_pathspec_item(struct index_state
*istate
,
340 const struct pathspec_item
*item
, int prefix
,
341 const char *name
, int namelen
, unsigned flags
)
343 /* name/namelen has prefix cut off by caller */
344 const char *match
= item
->match
+ prefix
;
345 int matchlen
= item
->len
- prefix
;
348 * The normal call pattern is:
349 * 1. prefix = common_prefix_len(ps);
350 * 2. prune something, or fill_directory
351 * 3. match_pathspec()
353 * 'prefix' at #1 may be shorter than the command's prefix and
354 * it's ok for #2 to match extra files. Those extras will be
357 * Suppose the pathspec is 'foo' and '../bar' running from
358 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
359 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
360 * user does not want XYZ/foo, only the "foo" part should be
361 * case-insensitive. We need to filter out XYZ/foo here. In
362 * other words, we do not trust the caller on comparing the
363 * prefix part when :(icase) is involved. We do exact
364 * comparison ourselves.
366 * Normally the caller (common_prefix_len() in fact) does
367 * _exact_ matching on name[-prefix+1..-1] and we do not need
368 * to check that part. Be defensive and check it anyway, in
369 * case common_prefix_len is changed, or a new caller is
370 * introduced that does not use common_prefix_len.
372 * If the penalty turns out too high when prefix is really
373 * long, maybe change it to
374 * strncmp(match, name, item->prefix - prefix)
376 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
377 strncmp(item
->match
, name
- prefix
, item
->prefix
))
380 if (item
->attr_match_nr
&&
381 !match_pathspec_attrs(istate
, name
, namelen
, item
))
384 /* If the match was just the prefix, we matched */
386 return MATCHED_RECURSIVELY
;
388 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
389 if (matchlen
== namelen
)
390 return MATCHED_EXACTLY
;
392 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
393 return MATCHED_RECURSIVELY
;
394 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
395 match
[matchlen
- 1] == '/' &&
396 namelen
== matchlen
- 1 &&
397 !ps_strncmp(item
, match
, name
, namelen
))
398 return MATCHED_EXACTLY
;
400 if (item
->nowildcard_len
< item
->len
&&
401 !git_fnmatch(item
, match
, name
,
402 item
->nowildcard_len
- prefix
))
403 return MATCHED_FNMATCH
;
405 /* Perform checks to see if "name" is a leading string of the pathspec */
406 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
407 !(flags
& DO_MATCH_EXCLUDE
)) {
408 /* name is a literal prefix of the pathspec */
409 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
410 if ((namelen
< matchlen
) &&
411 (match
[namelen
-offset
] == '/') &&
412 !ps_strncmp(item
, match
, name
, namelen
))
413 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
415 /* name doesn't match up to the first wild character */
416 if (item
->nowildcard_len
< item
->len
&&
417 ps_strncmp(item
, match
, name
,
418 item
->nowildcard_len
- prefix
))
422 * name has no wildcard, and it didn't match as a leading
423 * pathspec so return.
425 if (item
->nowildcard_len
== item
->len
)
429 * Here is where we would perform a wildmatch to check if
430 * "name" can be matched as a directory (or a prefix) against
431 * the pathspec. Since wildmatch doesn't have this capability
432 * at the present we have to punt and say that it is a match,
433 * potentially returning a false positive
434 * The submodules themselves will be able to perform more
435 * accurate matching to determine if the pathspec matches.
437 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
444 * do_match_pathspec() is meant to ONLY be called by
445 * match_pathspec_with_flags(); calling it directly risks pathspecs
446 * like ':!unwanted_path' being ignored.
448 * Given a name and a list of pathspecs, returns the nature of the
449 * closest (i.e. most specific) match of the name to any of the
452 * The caller typically calls this multiple times with the same
453 * pathspec and seen[] array but with different name/namelen
454 * (e.g. entries from the index) and is interested in seeing if and
455 * how each pathspec matches all the names it calls this function
456 * with. A mark is left in the seen[] array for each pathspec element
457 * indicating the closest type of match that element achieved, so if
458 * seen[n] remains zero after multiple invocations, that means the nth
459 * pathspec did not match any names, which could indicate that the
460 * user mistyped the nth pathspec.
462 static int do_match_pathspec(struct index_state
*istate
,
463 const struct pathspec
*ps
,
464 const char *name
, int namelen
,
465 int prefix
, char *seen
,
468 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
480 if (!ps
->recursive
||
481 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
483 return MATCHED_RECURSIVELY
;
485 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
486 return MATCHED_EXACTLY
;
494 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
497 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
498 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
501 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
504 * Make exclude patterns optional and never report
505 * "pathspec ':(exclude)foo' matches no files"
507 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
508 seen
[i
] = MATCHED_FNMATCH
;
509 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
512 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
513 ps
->max_depth
!= -1 &&
514 how
&& how
!= MATCHED_FNMATCH
) {
515 int len
= ps
->items
[i
].len
;
516 if (name
[len
] == '/')
518 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
519 how
= MATCHED_EXACTLY
;
526 if (seen
&& seen
[i
] < how
)
533 static int match_pathspec_with_flags(struct index_state
*istate
,
534 const struct pathspec
*ps
,
535 const char *name
, int namelen
,
536 int prefix
, char *seen
, unsigned flags
)
538 int positive
, negative
;
539 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
540 prefix
, seen
, flags
);
541 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
543 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
545 flags
| DO_MATCH_EXCLUDE
);
546 return negative
? 0 : positive
;
549 int match_pathspec(struct index_state
*istate
,
550 const struct pathspec
*ps
,
551 const char *name
, int namelen
,
552 int prefix
, char *seen
, int is_dir
)
554 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
555 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
556 prefix
, seen
, flags
);
560 * Check if a submodule is a superset of the pathspec
562 int submodule_path_match(struct index_state
*istate
,
563 const struct pathspec
*ps
,
564 const char *submodule_name
,
567 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
568 strlen(submodule_name
),
571 DO_MATCH_LEADING_PATHSPEC
);
575 int report_path_error(const char *ps_matched
,
576 const struct pathspec
*pathspec
)
579 * Make sure all pathspec matched; otherwise it is an error.
582 for (num
= 0; num
< pathspec
->nr
; num
++) {
583 int other
, found_dup
;
588 * The caller might have fed identical pathspec
589 * twice. Do not barf on such a mistake.
590 * FIXME: parse_pathspec should have eliminated
591 * duplicate pathspec.
593 for (found_dup
= other
= 0;
594 !found_dup
&& other
< pathspec
->nr
;
596 if (other
== num
|| !ps_matched
[other
])
598 if (!strcmp(pathspec
->items
[other
].original
,
599 pathspec
->items
[num
].original
))
601 * Ok, we have a match already.
608 error(_("pathspec '%s' did not match any file(s) known to git"),
609 pathspec
->items
[num
].original
);
616 * Return the length of the "simple" part of a path match limiter.
618 int simple_length(const char *match
)
623 unsigned char c
= *match
++;
625 if (c
== '\0' || is_glob_special(c
))
630 int no_wildcard(const char *string
)
632 return string
[simple_length(string
)] == '\0';
635 void parse_path_pattern(const char **pattern
,
640 const char *p
= *pattern
;
645 *flags
|= PATTERN_FLAG_NEGATIVE
;
649 if (len
&& p
[len
- 1] == '/') {
651 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
653 for (i
= 0; i
< len
; i
++) {
658 *flags
|= PATTERN_FLAG_NODIR
;
659 *nowildcardlen
= simple_length(p
);
661 * we should have excluded the trailing slash from 'p' too,
662 * but that's one more allocation. Instead just make sure
663 * nowildcardlen does not exceed real patternlen
665 if (*nowildcardlen
> len
)
666 *nowildcardlen
= len
;
667 if (*p
== '*' && no_wildcard(p
+ 1))
668 *flags
|= PATTERN_FLAG_ENDSWITH
;
673 int pl_hashmap_cmp(const void *cmp_data UNUSED
,
674 const struct hashmap_entry
*a
,
675 const struct hashmap_entry
*b
,
676 const void *key UNUSED
)
678 const struct pattern_entry
*ee1
=
679 container_of(a
, struct pattern_entry
, ent
);
680 const struct pattern_entry
*ee2
=
681 container_of(b
, struct pattern_entry
, ent
);
683 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
687 return fspathncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
690 static char *dup_and_filter_pattern(const char *pattern
)
694 char *result
= xstrdup(pattern
);
700 /* skip escape characters (once) */
720 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
722 struct pattern_entry
*translated
;
725 const char *prev
, *cur
, *next
;
727 if (!pl
->use_cone_patterns
)
730 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
731 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
732 !strcmp(given
->pattern
, "/*")) {
737 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
742 if (given
->patternlen
< 2 ||
743 *given
->pattern
!= '/' ||
744 strstr(given
->pattern
, "**")) {
745 /* Not a cone pattern. */
746 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
750 if (!(given
->flags
& PATTERN_FLAG_MUSTBEDIR
) &&
751 strcmp(given
->pattern
, "/*")) {
752 /* Not a cone pattern. */
753 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
757 prev
= given
->pattern
;
758 cur
= given
->pattern
+ 1;
759 next
= given
->pattern
+ 2;
762 /* Watch for glob characters '*', '\', '[', '?' */
763 if (!is_glob_special(*cur
))
766 /* But only if *prev != '\\' */
770 /* But allow the initial '\' */
772 is_glob_special(*next
))
775 /* But a trailing '/' then '*' is fine */
781 /* Not a cone pattern. */
782 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
791 if (given
->patternlen
> 2 &&
792 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
793 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
794 /* Not a cone pattern. */
795 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
799 truncated
= dup_and_filter_pattern(given
->pattern
);
801 translated
= xmalloc(sizeof(struct pattern_entry
));
802 translated
->pattern
= truncated
;
803 translated
->patternlen
= given
->patternlen
- 2;
804 hashmap_entry_init(&translated
->ent
,
805 fspathhash(translated
->pattern
));
807 if (!hashmap_get_entry(&pl
->recursive_hashmap
,
808 translated
, ent
, NULL
)) {
809 /* We did not see the "parent" included */
810 warning(_("unrecognized negative pattern: '%s'"),
817 hashmap_add(&pl
->parent_hashmap
, &translated
->ent
);
818 hashmap_remove(&pl
->recursive_hashmap
, &translated
->ent
, &data
);
823 if (given
->flags
& PATTERN_FLAG_NEGATIVE
) {
824 warning(_("unrecognized negative pattern: '%s'"),
829 translated
= xmalloc(sizeof(struct pattern_entry
));
831 translated
->pattern
= dup_and_filter_pattern(given
->pattern
);
832 translated
->patternlen
= given
->patternlen
;
833 hashmap_entry_init(&translated
->ent
,
834 fspathhash(translated
->pattern
));
836 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
838 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
839 /* we already included this at the parent level */
840 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
848 warning(_("disabling cone pattern matching"));
849 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
850 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
851 pl
->use_cone_patterns
= 0;
854 static int hashmap_contains_path(struct hashmap
*map
,
855 struct strbuf
*pattern
)
857 struct pattern_entry p
;
859 /* Check straight mapping */
860 p
.pattern
= pattern
->buf
;
861 p
.patternlen
= pattern
->len
;
862 hashmap_entry_init(&p
.ent
, fspathhash(p
.pattern
));
863 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
866 int hashmap_contains_parent(struct hashmap
*map
,
868 struct strbuf
*buffer
)
872 strbuf_setlen(buffer
, 0);
875 strbuf_addch(buffer
, '/');
877 strbuf_addstr(buffer
, path
);
879 slash_pos
= strrchr(buffer
->buf
, '/');
881 while (slash_pos
> buffer
->buf
) {
882 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
884 if (hashmap_contains_path(map
, buffer
))
887 slash_pos
= strrchr(buffer
->buf
, '/');
893 void add_pattern(const char *string
, const char *base
,
894 int baselen
, struct pattern_list
*pl
, int srcpos
)
896 struct path_pattern
*pattern
;
901 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
902 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
903 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
905 pattern
= xmalloc(sizeof(*pattern
));
906 pattern
->pattern
= string
;
908 pattern
->patternlen
= patternlen
;
909 pattern
->nowildcardlen
= nowildcardlen
;
910 pattern
->base
= base
;
911 pattern
->baselen
= baselen
;
912 pattern
->flags
= flags
;
913 pattern
->srcpos
= srcpos
;
914 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
915 pl
->patterns
[pl
->nr
++] = pattern
;
918 add_pattern_to_hashsets(pl
, pattern
);
921 static int read_skip_worktree_file_from_index(struct index_state
*istate
,
923 size_t *size_out
, char **data_out
,
924 struct oid_stat
*oid_stat
)
929 pos
= index_name_pos(istate
, path
, len
);
932 if (!ce_skip_worktree(istate
->cache
[pos
]))
935 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
939 * Frees memory within pl which was allocated for exclude patterns and
940 * the file buffer. Does not free pl itself.
942 void clear_pattern_list(struct pattern_list
*pl
)
946 for (i
= 0; i
< pl
->nr
; i
++)
947 free(pl
->patterns
[i
]);
950 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
951 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
953 memset(pl
, 0, sizeof(*pl
));
956 static void trim_trailing_spaces(char *buf
)
958 char *p
, *last_space
= NULL
;
960 for (p
= buf
; *p
; p
++)
980 * Given a subdirectory name and "dir" of the current directory,
981 * search the subdir in "dir" and return it, or create a new one if it
982 * does not exist in "dir".
984 * If "name" has the trailing slash, it'll be excluded in the search.
986 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
987 struct untracked_cache_dir
*dir
,
988 const char *name
, int len
)
991 struct untracked_cache_dir
*d
;
994 if (len
&& name
[len
- 1] == '/')
998 while (last
> first
) {
999 int cmp
, next
= first
+ ((last
- first
) >> 1);
1000 d
= dir
->dirs
[next
];
1001 cmp
= strncmp(name
, d
->name
, len
);
1002 if (!cmp
&& strlen(d
->name
) > len
)
1014 FLEX_ALLOC_MEM(d
, name
, name
, len
);
1016 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
1017 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
1018 dir
->dirs_nr
- first
);
1020 dir
->dirs
[first
] = d
;
1024 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
1028 dir
->untracked_nr
= 0;
1029 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1030 do_invalidate_gitignore(dir
->dirs
[i
]);
1033 static void invalidate_gitignore(struct untracked_cache
*uc
,
1034 struct untracked_cache_dir
*dir
)
1036 uc
->gitignore_invalidated
++;
1037 do_invalidate_gitignore(dir
);
1040 static void invalidate_directory(struct untracked_cache
*uc
,
1041 struct untracked_cache_dir
*dir
)
1046 * Invalidation increment here is just roughly correct. If
1047 * untracked_nr or any of dirs[].recurse is non-zero, we
1048 * should increment dir_invalidated too. But that's more
1052 uc
->dir_invalidated
++;
1055 dir
->untracked_nr
= 0;
1056 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1057 dir
->dirs
[i
]->recurse
= 0;
1060 static int add_patterns_from_buffer(char *buf
, size_t size
,
1061 const char *base
, int baselen
,
1062 struct pattern_list
*pl
);
1064 /* Flags for add_patterns() */
1065 #define PATTERN_NOFOLLOW (1<<0)
1068 * Given a file with name "fname", read it (either from disk, or from
1069 * an index if 'istate' is non-null), parse it and store the
1070 * exclude rules in "pl".
1072 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1073 * stat data from disk (only valid if add_patterns returns zero). If
1074 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1076 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1077 struct pattern_list
*pl
, struct index_state
*istate
,
1078 unsigned flags
, struct oid_stat
*oid_stat
)
1086 if (flags
& PATTERN_NOFOLLOW
)
1087 fd
= open_nofollow(fname
, O_RDONLY
);
1089 fd
= open(fname
, O_RDONLY
);
1091 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1093 warn_on_fopen_errors(fname
);
1098 r
= read_skip_worktree_file_from_index(istate
, fname
,
1104 size
= xsize_t(st
.st_size
);
1107 fill_stat_data(&oid_stat
->stat
, &st
);
1108 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1109 oid_stat
->valid
= 1;
1114 buf
= xmallocz(size
);
1115 if (read_in_full(fd
, buf
, size
) != size
) {
1124 if (oid_stat
->valid
&&
1125 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1126 ; /* no content change, oid_stat->oid still good */
1128 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1129 !ce_stage(istate
->cache
[pos
]) &&
1130 ce_uptodate(istate
->cache
[pos
]) &&
1131 !would_convert_to_git(istate
, fname
))
1132 oidcpy(&oid_stat
->oid
,
1133 &istate
->cache
[pos
]->oid
);
1135 hash_object_file(the_hash_algo
, buf
, size
,
1136 OBJ_BLOB
, &oid_stat
->oid
);
1137 fill_stat_data(&oid_stat
->stat
, &st
);
1138 oid_stat
->valid
= 1;
1142 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1146 static int add_patterns_from_buffer(char *buf
, size_t size
,
1147 const char *base
, int baselen
,
1148 struct pattern_list
*pl
)
1153 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1154 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1158 if (skip_utf8_bom(&buf
, size
))
1159 size
-= buf
- pl
->filebuf
;
1163 for (i
= 0; i
< size
; i
++) {
1164 if (buf
[i
] == '\n') {
1165 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1166 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1167 trim_trailing_spaces(entry
);
1168 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1171 entry
= buf
+ i
+ 1;
1177 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1178 int baselen
, struct pattern_list
*pl
,
1179 struct index_state
*istate
,
1182 return add_patterns(fname
, base
, baselen
, pl
, istate
, flags
, NULL
);
1185 int add_patterns_from_blob_to_list(
1186 struct object_id
*oid
,
1187 const char *base
, int baselen
,
1188 struct pattern_list
*pl
)
1194 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1198 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1202 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1203 int group_type
, const char *src
)
1205 struct pattern_list
*pl
;
1206 struct exclude_list_group
*group
;
1208 group
= &dir
->internal
.exclude_list_group
[group_type
];
1209 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1210 pl
= &group
->pl
[group
->nr
++];
1211 memset(pl
, 0, sizeof(*pl
));
1217 * Used to set up core.excludesfile and .git/info/exclude lists.
1219 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1220 struct oid_stat
*oid_stat
)
1222 struct pattern_list
*pl
;
1224 * catch setup_standard_excludes() that's called before
1225 * dir->untracked is assigned. That function behaves
1226 * differently when dir->untracked is non-NULL.
1228 if (!dir
->untracked
)
1229 dir
->internal
.unmanaged_exclude_files
++;
1230 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1231 if (add_patterns(fname
, "", 0, pl
, NULL
, 0, oid_stat
) < 0)
1232 die(_("cannot use %s as an exclude file"), fname
);
1235 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1237 dir
->internal
.unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1238 add_patterns_from_file_1(dir
, fname
, NULL
);
1241 int match_basename(const char *basename
, int basenamelen
,
1242 const char *pattern
, int prefix
, int patternlen
,
1245 if (prefix
== patternlen
) {
1246 if (patternlen
== basenamelen
&&
1247 !fspathncmp(pattern
, basename
, basenamelen
))
1249 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1250 /* "*literal" matching against "fooliteral" */
1251 if (patternlen
- 1 <= basenamelen
&&
1252 !fspathncmp(pattern
+ 1,
1253 basename
+ basenamelen
- (patternlen
- 1),
1257 if (fnmatch_icase_mem(pattern
, patternlen
,
1258 basename
, basenamelen
,
1265 int match_pathname(const char *pathname
, int pathlen
,
1266 const char *base
, int baselen
,
1267 const char *pattern
, int prefix
, int patternlen
)
1273 * match with FNM_PATHNAME; the pattern has base implicitly
1276 if (*pattern
== '/') {
1283 * baselen does not count the trailing slash. base[] may or
1284 * may not end with a trailing slash though.
1286 if (pathlen
< baselen
+ 1 ||
1287 (baselen
&& pathname
[baselen
] != '/') ||
1288 fspathncmp(pathname
, base
, baselen
))
1291 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1292 name
= pathname
+ pathlen
- namelen
;
1296 * if the non-wildcard part is longer than the
1297 * remaining pathname, surely it cannot match.
1299 if (prefix
> namelen
)
1302 if (fspathncmp(pattern
, name
, prefix
))
1305 patternlen
-= prefix
;
1310 * If the whole pattern did not have a wildcard,
1311 * then our prefix match is all we need; we
1312 * do not need to call fnmatch at all.
1314 if (!patternlen
&& !namelen
)
1318 return fnmatch_icase_mem(pattern
, patternlen
,
1324 * Scan the given exclude list in reverse to see whether pathname
1325 * should be ignored. The first match (i.e. the last on the list), if
1326 * any, determines the fate. Returns the exclude_list element which
1327 * matched, or NULL for undecided.
1329 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1331 const char *basename
,
1333 struct pattern_list
*pl
,
1334 struct index_state
*istate
)
1336 struct path_pattern
*res
= NULL
; /* undecided */
1340 return NULL
; /* undefined */
1342 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1343 struct path_pattern
*pattern
= pl
->patterns
[i
];
1344 const char *exclude
= pattern
->pattern
;
1345 int prefix
= pattern
->nowildcardlen
;
1347 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1348 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1349 if (*dtype
!= DT_DIR
)
1353 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1354 if (match_basename(basename
,
1355 pathlen
- (basename
- pathname
),
1356 exclude
, prefix
, pattern
->patternlen
,
1364 assert(pattern
->baselen
== 0 ||
1365 pattern
->base
[pattern
->baselen
- 1] == '/');
1366 if (match_pathname(pathname
, pathlen
,
1368 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1369 exclude
, prefix
, pattern
->patternlen
)) {
1378 * Scan the list of patterns to determine if the ordered list
1379 * of patterns matches on 'pathname'.
1381 * Return 1 for a match, 0 for not matched and -1 for undecided.
1383 enum pattern_match_result
path_matches_pattern_list(
1384 const char *pathname
, int pathlen
,
1385 const char *basename
, int *dtype
,
1386 struct pattern_list
*pl
,
1387 struct index_state
*istate
)
1389 struct path_pattern
*pattern
;
1390 struct strbuf parent_pathname
= STRBUF_INIT
;
1391 int result
= NOT_MATCHED
;
1394 if (!pl
->use_cone_patterns
) {
1395 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1398 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1410 strbuf_addch(&parent_pathname
, '/');
1411 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1414 * Directory entries are matched if and only if a file
1415 * contained immediately within them is matched. For the
1416 * case of a directory entry, modify the path to create
1417 * a fake filename within this directory, allowing us to
1418 * use the file-base matching logic in an equivalent way.
1420 if (parent_pathname
.len
> 0 &&
1421 parent_pathname
.buf
[parent_pathname
.len
- 1] == '/') {
1422 slash_pos
= parent_pathname
.len
- 1;
1423 strbuf_add(&parent_pathname
, "-", 1);
1425 const char *slash_ptr
= strrchr(parent_pathname
.buf
, '/');
1426 slash_pos
= slash_ptr
? slash_ptr
- parent_pathname
.buf
: 0;
1429 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1430 &parent_pathname
)) {
1431 result
= MATCHED_RECURSIVE
;
1436 /* include every file in root */
1441 strbuf_setlen(&parent_pathname
, slash_pos
);
1443 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1448 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1451 result
= MATCHED_RECURSIVE
;
1454 strbuf_release(&parent_pathname
);
1458 int init_sparse_checkout_patterns(struct index_state
*istate
)
1460 if (!core_apply_sparse_checkout
)
1462 if (istate
->sparse_checkout_patterns
)
1465 CALLOC_ARRAY(istate
->sparse_checkout_patterns
, 1);
1467 if (get_sparse_checkout_patterns(istate
->sparse_checkout_patterns
) < 0) {
1468 FREE_AND_NULL(istate
->sparse_checkout_patterns
);
1475 static int path_in_sparse_checkout_1(const char *path
,
1476 struct index_state
*istate
,
1477 int require_cone_mode
)
1480 enum pattern_match_result match
= UNDECIDED
;
1481 const char *end
, *slash
;
1484 * We default to accepting a path if the path is empty, there are no
1485 * patterns, or the patterns are of the wrong type.
1488 init_sparse_checkout_patterns(istate
) ||
1489 (require_cone_mode
&&
1490 !istate
->sparse_checkout_patterns
->use_cone_patterns
))
1494 * If UNDECIDED, use the match from the parent dir (recursively), or
1495 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1496 * never returns UNDECIDED, so we will execute only one iteration in
1499 for (end
= path
+ strlen(path
);
1500 end
> path
&& match
== UNDECIDED
;
1503 for (slash
= end
- 1; slash
> path
&& *slash
!= '/'; slash
--)
1506 match
= path_matches_pattern_list(path
, end
- path
,
1507 slash
> path
? slash
+ 1 : path
, &dtype
,
1508 istate
->sparse_checkout_patterns
, istate
);
1510 /* We are going to match the parent dir now */
1516 int path_in_sparse_checkout(const char *path
,
1517 struct index_state
*istate
)
1519 return path_in_sparse_checkout_1(path
, istate
, 0);
1522 int path_in_cone_mode_sparse_checkout(const char *path
,
1523 struct index_state
*istate
)
1525 return path_in_sparse_checkout_1(path
, istate
, 1);
1528 static struct path_pattern
*last_matching_pattern_from_lists(
1529 struct dir_struct
*dir
, struct index_state
*istate
,
1530 const char *pathname
, int pathlen
,
1531 const char *basename
, int *dtype_p
)
1534 struct exclude_list_group
*group
;
1535 struct path_pattern
*pattern
;
1536 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1537 group
= &dir
->internal
.exclude_list_group
[i
];
1538 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1539 pattern
= last_matching_pattern_from_list(
1540 pathname
, pathlen
, basename
, dtype_p
,
1541 &group
->pl
[j
], istate
);
1550 * Loads the per-directory exclude list for the substring of base
1551 * which has a char length of baselen.
1553 static void prep_exclude(struct dir_struct
*dir
,
1554 struct index_state
*istate
,
1555 const char *base
, int baselen
)
1557 struct exclude_list_group
*group
;
1558 struct pattern_list
*pl
;
1559 struct exclude_stack
*stk
= NULL
;
1560 struct untracked_cache_dir
*untracked
;
1563 group
= &dir
->internal
.exclude_list_group
[EXC_DIRS
];
1566 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1567 * which originate from directories not in the prefix of the
1568 * path being checked.
1570 while ((stk
= dir
->internal
.exclude_stack
) != NULL
) {
1571 if (stk
->baselen
<= baselen
&&
1572 !strncmp(dir
->internal
.basebuf
.buf
, base
, stk
->baselen
))
1574 pl
= &group
->pl
[dir
->internal
.exclude_stack
->exclude_ix
];
1575 dir
->internal
.exclude_stack
= stk
->prev
;
1576 dir
->internal
.pattern
= NULL
;
1577 free((char *)pl
->src
); /* see strbuf_detach() below */
1578 clear_pattern_list(pl
);
1583 /* Skip traversing into sub directories if the parent is excluded */
1584 if (dir
->internal
.pattern
)
1588 * Lazy initialization. All call sites currently just
1589 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1590 * them seems lots of work for little benefit.
1592 if (!dir
->internal
.basebuf
.buf
)
1593 strbuf_init(&dir
->internal
.basebuf
, PATH_MAX
);
1595 /* Read from the parent directories and push them down. */
1596 current
= stk
? stk
->baselen
: -1;
1597 strbuf_setlen(&dir
->internal
.basebuf
, current
< 0 ? 0 : current
);
1599 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1603 while (current
< baselen
) {
1605 struct oid_stat oid_stat
;
1607 CALLOC_ARRAY(stk
, 1);
1612 cp
= strchr(base
+ current
+ 1, '/');
1614 die("oops in prep_exclude");
1617 lookup_untracked(dir
->untracked
,
1620 cp
- base
- current
);
1622 stk
->prev
= dir
->internal
.exclude_stack
;
1623 stk
->baselen
= cp
- base
;
1624 stk
->exclude_ix
= group
->nr
;
1625 stk
->ucd
= untracked
;
1626 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1627 strbuf_add(&dir
->internal
.basebuf
, base
+ current
, stk
->baselen
- current
);
1628 assert(stk
->baselen
== dir
->internal
.basebuf
.len
);
1630 /* Abort if the directory is excluded */
1633 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = 0;
1634 dir
->internal
.pattern
= last_matching_pattern_from_lists(dir
,
1636 dir
->internal
.basebuf
.buf
, stk
->baselen
- 1,
1637 dir
->internal
.basebuf
.buf
+ current
, &dt
);
1638 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = '/';
1639 if (dir
->internal
.pattern
&&
1640 dir
->internal
.pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1641 dir
->internal
.pattern
= NULL
;
1642 if (dir
->internal
.pattern
) {
1643 dir
->internal
.exclude_stack
= stk
;
1648 /* Try to read per-directory file */
1649 oidclr(&oid_stat
.oid
);
1651 if (dir
->exclude_per_dir
&&
1653 * If we know that no files have been added in
1654 * this directory (i.e. valid_cached_dir() has
1655 * been executed and set untracked->valid) ..
1657 (!untracked
|| !untracked
->valid
||
1659 * .. and .gitignore does not exist before
1660 * (i.e. null exclude_oid). Then we can skip
1661 * loading .gitignore, which would result in
1664 !is_null_oid(&untracked
->exclude_oid
))) {
1666 * dir->internal.basebuf gets reused by the traversal,
1667 * but we need fname to remain unchanged to ensure the
1668 * src member of each struct path_pattern correctly
1669 * back-references its source file. Other invocations
1670 * of add_pattern_list provide stable strings, so we
1671 * strbuf_detach() and free() here in the caller.
1673 struct strbuf sb
= STRBUF_INIT
;
1674 strbuf_addbuf(&sb
, &dir
->internal
.basebuf
);
1675 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1676 pl
->src
= strbuf_detach(&sb
, NULL
);
1677 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1679 untracked
? &oid_stat
: NULL
);
1682 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1683 * will first be called in valid_cached_dir() then maybe many
1684 * times more in last_matching_pattern(). When the cache is
1685 * used, last_matching_pattern() will not be called and
1686 * reading .gitignore content will be a waste.
1688 * So when it's called by valid_cached_dir() and we can get
1689 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1690 * modified on work tree), we could delay reading the
1691 * .gitignore content until we absolutely need it in
1692 * last_matching_pattern(). Be careful about ignore rule
1693 * order, though, if you do that.
1696 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1697 invalidate_gitignore(dir
->untracked
, untracked
);
1698 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1700 dir
->internal
.exclude_stack
= stk
;
1701 current
= stk
->baselen
;
1703 strbuf_setlen(&dir
->internal
.basebuf
, baselen
);
1707 * Loads the exclude lists for the directory containing pathname, then
1708 * scans all exclude lists to determine whether pathname is excluded.
1709 * Returns the exclude_list element which matched, or NULL for
1712 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1713 struct index_state
*istate
,
1714 const char *pathname
,
1717 int pathlen
= strlen(pathname
);
1718 const char *basename
= strrchr(pathname
, '/');
1719 basename
= (basename
) ? basename
+1 : pathname
;
1721 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1723 if (dir
->internal
.pattern
)
1724 return dir
->internal
.pattern
;
1726 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1731 * Loads the exclude lists for the directory containing pathname, then
1732 * scans all exclude lists to determine whether pathname is excluded.
1733 * Returns 1 if true, otherwise 0.
1735 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1736 const char *pathname
, int *dtype_p
)
1738 struct path_pattern
*pattern
=
1739 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1741 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1745 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1747 struct dir_entry
*ent
;
1749 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1754 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1755 struct index_state
*istate
,
1756 const char *pathname
, int len
)
1758 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1761 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->internal
.alloc
);
1762 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1765 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1766 struct index_state
*istate
,
1767 const char *pathname
, int len
)
1769 if (!index_name_is_other(istate
, pathname
, len
))
1772 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->internal
.ignored_alloc
);
1773 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1777 index_nonexistent
= 0,
1783 * Do not use the alphabetically sorted index to look up
1784 * the directory name; instead, use the case insensitive
1787 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1788 const char *dirname
, int len
)
1790 struct cache_entry
*ce
;
1792 if (index_dir_exists(istate
, dirname
, len
))
1793 return index_directory
;
1795 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1796 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1797 return index_gitdir
;
1799 return index_nonexistent
;
1803 * The index sorts alphabetically by entry name, which
1804 * means that a gitlink sorts as '\0' at the end, while
1805 * a directory (which is defined not as an entry, but as
1806 * the files it contains) will sort with the '/' at the
1809 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1810 const char *dirname
, int len
)
1815 return directory_exists_in_index_icase(istate
, dirname
, len
);
1817 pos
= index_name_pos(istate
, dirname
, len
);
1820 while (pos
< istate
->cache_nr
) {
1821 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1822 unsigned char endchar
;
1824 if (strncmp(ce
->name
, dirname
, len
))
1826 endchar
= ce
->name
[len
];
1830 return index_directory
;
1831 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1832 return index_gitdir
;
1834 return index_nonexistent
;
1838 * When we find a directory when traversing the filesystem, we
1839 * have three distinct cases:
1842 * - see it as a directory
1845 * and which one we choose depends on a combination of existing
1846 * git index contents and the flags passed into the directory
1847 * traversal routine.
1849 * Case 1: If we *already* have entries in the index under that
1850 * directory name, we always recurse into the directory to see
1853 * Case 2: If we *already* have that directory name as a gitlink,
1854 * we always continue to see it as a gitlink, regardless of whether
1855 * there is an actual git directory there or not (it might not
1856 * be checked out as a subproject!)
1858 * Case 3: if we didn't have it in the index previously, we
1859 * have a few sub-cases:
1861 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1862 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1863 * also true, in which case we need to check if it contains any
1864 * untracked and / or ignored files.
1865 * (b) if it looks like a git directory and we don't have the
1866 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1867 * show it as a directory.
1868 * (c) otherwise, we recurse into it.
1870 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1871 struct index_state
*istate
,
1872 struct untracked_cache_dir
*untracked
,
1873 const char *dirname
, int len
, int baselen
, int excluded
,
1874 const struct pathspec
*pathspec
)
1877 * WARNING: From this function, you can return path_recurse or you
1878 * can call read_directory_recursive() (or neither), but
1879 * you CAN'T DO BOTH.
1881 enum path_treatment state
;
1882 int matches_how
= 0;
1883 int check_only
, stop_early
;
1884 int old_ignored_nr
, old_untracked_nr
;
1885 /* The "len-1" is to strip the final '/' */
1886 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1888 if (status
== index_directory
)
1889 return path_recurse
;
1890 if (status
== index_gitdir
)
1892 if (status
!= index_nonexistent
)
1893 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1896 * We don't want to descend into paths that don't match the necessary
1897 * patterns. Clearly, if we don't have a pathspec, then we can't check
1898 * for matching patterns. Also, if (excluded) then we know we matched
1899 * the exclusion patterns so as an optimization we can skip checking
1900 * for matching patterns.
1902 if (pathspec
&& !excluded
) {
1903 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1907 DO_MATCH_LEADING_PATHSPEC
);
1913 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1914 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1916 * Determine if `dirname` is a nested repo by confirming that:
1917 * 1) we are in a nonbare repository, and
1918 * 2) `dirname` is not an immediate parent of `the_repository->gitdir`,
1919 * which could occur if the git_dir or worktree location was
1920 * manually configured by the user; see t2205 testcases 1-3 for
1921 * examples where this matters
1924 struct strbuf sb
= STRBUF_INIT
;
1925 strbuf_addstr(&sb
, dirname
);
1926 nested_repo
= is_nonbare_repository_dir(&sb
);
1929 char *real_dirname
, *real_gitdir
;
1930 strbuf_addstr(&sb
, ".git");
1931 real_dirname
= real_pathdup(sb
.buf
, 1);
1932 real_gitdir
= real_pathdup(the_repository
->gitdir
, 1);
1934 nested_repo
= !!strcmp(real_dirname
, real_gitdir
);
1938 strbuf_release(&sb
);
1941 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1942 (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
))
1944 return excluded
? path_excluded
: path_untracked
;
1948 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1950 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1951 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1954 * This is an excluded directory and we are
1955 * showing ignored paths that match an exclude
1956 * pattern. (e.g. show directory as ignored
1957 * only if it matches an exclude pattern).
1958 * This path will either be 'path_excluded`
1959 * (if we are showing empty directories or if
1960 * the directory is not empty), or will be
1961 * 'path_none' (empty directory, and we are
1962 * not showing empty directories).
1964 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1965 return path_excluded
;
1967 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1968 untracked
, 1, 1, pathspec
) == path_excluded
)
1969 return path_excluded
;
1973 return path_recurse
;
1976 assert(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
);
1979 * If we have a pathspec which could match something _below_ this
1980 * directory (e.g. when checking 'subdir/' having a pathspec like
1981 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1984 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1985 return path_recurse
;
1987 /* Special cases for where this directory is excluded/ignored */
1990 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1991 * hiding empty directories, there is no need to
1992 * recurse into an ignored directory.
1994 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1995 return path_excluded
;
1998 * Even if we are hiding empty directories, we can still avoid
1999 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
2000 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
2002 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2003 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
2004 return path_excluded
;
2008 * Other than the path_recurse case above, we only need to
2009 * recurse into untracked directories if any of the following
2011 * - DIR_SHOW_IGNORED (because then we need to determine if
2012 * there are ignored entries below)
2013 * - DIR_SHOW_IGNORED_TOO (same as above)
2014 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
2015 * the directory is empty)
2018 !(dir
->flags
& (DIR_SHOW_IGNORED
|
2019 DIR_SHOW_IGNORED_TOO
|
2020 DIR_HIDE_EMPTY_DIRECTORIES
))) {
2021 return path_untracked
;
2025 * Even if we don't want to know all the paths under an untracked or
2026 * ignored directory, we may still need to go into the directory to
2027 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2028 * an empty directory should be path_none instead of path_excluded or
2031 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
2032 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
2035 * However, there's another optimization possible as a subset of
2036 * check_only, based on the cases we have to consider:
2037 * A) Directory matches no exclude patterns:
2038 * * Directory is empty => path_none
2039 * * Directory has an untracked file under it => path_untracked
2040 * * Directory has only ignored files under it => path_excluded
2041 * B) Directory matches an exclude pattern:
2042 * * Directory is empty => path_none
2043 * * Directory has an untracked file under it => path_excluded
2044 * * Directory has only ignored files under it => path_excluded
2045 * In case A, we can exit as soon as we've found an untracked
2046 * file but otherwise have to walk all files. In case B, though,
2047 * we can stop at the first file we find under the directory.
2049 stop_early
= check_only
&& excluded
;
2052 * If /every/ file within an untracked directory is ignored, then
2053 * we want to treat the directory as ignored (for e.g. status
2054 * --porcelain), without listing the individual ignored files
2055 * underneath. To do so, we'll save the current ignored_nr, and
2056 * pop all the ones added after it if it turns out the entire
2057 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2058 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2059 * untracked paths so will need to pop all those off the last
2060 * after we traverse.
2062 old_ignored_nr
= dir
->ignored_nr
;
2063 old_untracked_nr
= dir
->nr
;
2065 /* Actually recurse into dirname now, we'll fixup the state later. */
2066 untracked
= lookup_untracked(dir
->untracked
, untracked
,
2067 dirname
+ baselen
, len
- baselen
);
2068 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
2069 check_only
, stop_early
, pathspec
);
2071 /* There are a variety of reasons we may need to fixup the state... */
2072 if (state
== path_excluded
) {
2073 /* state == path_excluded implies all paths under
2074 * dirname were ignored...
2076 * if running e.g. `git status --porcelain --ignored=matching`,
2077 * then we want to see the subpaths that are ignored.
2079 * if running e.g. just `git status --porcelain`, then
2080 * we just want the directory itself to be listed as ignored
2081 * and not the individual paths underneath.
2083 int want_ignored_subpaths
=
2084 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2085 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
2087 if (want_ignored_subpaths
) {
2089 * with --ignored=matching, we want the subpaths
2090 * INSTEAD of the directory itself.
2095 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
2096 FREE_AND_NULL(dir
->ignored
[i
]);
2097 dir
->ignored_nr
= old_ignored_nr
;
2102 * We may need to ignore some of the untracked paths we found while
2103 * traversing subdirectories.
2105 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2106 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
2108 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
2109 FREE_AND_NULL(dir
->entries
[i
]);
2110 dir
->nr
= old_untracked_nr
;
2114 * If there is nothing under the current directory and we are not
2115 * hiding empty directories, then we need to report on the
2116 * untracked or ignored status of the directory itself.
2118 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2119 state
= excluded
? path_excluded
: path_untracked
;
2125 * This is an inexact early pruning of any recursive directory
2126 * reading - if the path cannot possibly be in the pathspec,
2127 * return true, and we'll skip it early.
2129 static int simplify_away(const char *path
, int pathlen
,
2130 const struct pathspec
*pathspec
)
2134 if (!pathspec
|| !pathspec
->nr
)
2137 GUARD_PATHSPEC(pathspec
,
2146 for (i
= 0; i
< pathspec
->nr
; i
++) {
2147 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2148 int len
= item
->nowildcard_len
;
2152 if (!ps_strncmp(item
, item
->match
, path
, len
))
2160 * This function tells us whether an excluded path matches a
2161 * list of "interesting" pathspecs. That is, whether a path matched
2162 * by any of the pathspecs could possibly be ignored by excluding
2163 * the specified path. This can happen if:
2165 * 1. the path is mentioned explicitly in the pathspec
2167 * 2. the path is a directory prefix of some element in the
2170 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2171 const struct pathspec
*pathspec
)
2175 if (!pathspec
|| !pathspec
->nr
)
2178 GUARD_PATHSPEC(pathspec
,
2186 for (i
= 0; i
< pathspec
->nr
; i
++) {
2187 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2188 int len
= item
->nowildcard_len
;
2190 if (len
== pathlen
&&
2191 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2193 if (len
> pathlen
&&
2194 item
->match
[pathlen
] == '/' &&
2195 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2201 static int get_index_dtype(struct index_state
*istate
,
2202 const char *path
, int len
)
2205 const struct cache_entry
*ce
;
2207 ce
= index_file_exists(istate
, path
, len
, 0);
2209 if (!ce_uptodate(ce
))
2211 if (S_ISGITLINK(ce
->ce_mode
))
2214 * Nobody actually cares about the
2215 * difference between DT_LNK and DT_REG
2220 /* Try to look it up as a directory */
2221 pos
= index_name_pos(istate
, path
, len
);
2225 while (pos
< istate
->cache_nr
) {
2226 ce
= istate
->cache
[pos
++];
2227 if (strncmp(ce
->name
, path
, len
))
2229 if (ce
->name
[len
] > '/')
2231 if (ce
->name
[len
] < '/')
2233 if (!ce_uptodate(ce
))
2234 break; /* continue? */
2240 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2241 const char *path
, int len
)
2245 if (dtype
!= DT_UNKNOWN
)
2247 dtype
= get_index_dtype(istate
, path
, len
);
2248 if (dtype
!= DT_UNKNOWN
)
2250 if (lstat(path
, &st
))
2252 if (S_ISREG(st
.st_mode
))
2254 if (S_ISDIR(st
.st_mode
))
2256 if (S_ISLNK(st
.st_mode
))
2261 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2262 struct cached_dir
*cdir
,
2263 struct index_state
*istate
,
2264 struct strbuf
*path
,
2266 const struct pathspec
*pathspec
)
2269 * WARNING: From this function, you can return path_recurse or you
2270 * can call read_directory_recursive() (or neither), but
2271 * you CAN'T DO BOTH.
2273 strbuf_setlen(path
, baselen
);
2275 strbuf_addstr(path
, cdir
->file
);
2276 return path_untracked
;
2278 strbuf_addstr(path
, cdir
->ucd
->name
);
2279 /* treat_one_path() does this before it calls treat_directory() */
2280 strbuf_complete(path
, '/');
2281 if (cdir
->ucd
->check_only
)
2283 * check_only is set as a result of treat_directory() getting
2284 * to its bottom. Verify again the same set of directories
2285 * with check_only set.
2287 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2288 cdir
->ucd
, 1, 0, pathspec
);
2290 * We get path_recurse in the first run when
2291 * directory_exists_in_index() returns index_nonexistent. We
2292 * are sure that new changes in the index does not impact the
2293 * outcome. Return now.
2295 return path_recurse
;
2298 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2299 struct untracked_cache_dir
*untracked
,
2300 struct cached_dir
*cdir
,
2301 struct index_state
*istate
,
2302 struct strbuf
*path
,
2304 const struct pathspec
*pathspec
)
2306 int has_path_in_index
, dtype
, excluded
;
2309 return treat_path_fast(dir
, cdir
, istate
, path
,
2311 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2313 strbuf_setlen(path
, baselen
);
2314 strbuf_addstr(path
, cdir
->d_name
);
2315 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2318 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2320 /* Always exclude indexed files */
2321 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2323 if (dtype
!= DT_DIR
&& has_path_in_index
)
2327 * When we are looking at a directory P in the working tree,
2328 * there are three cases:
2330 * (1) P exists in the index. Everything inside the directory P in
2331 * the working tree needs to go when P is checked out from the
2334 * (2) P does not exist in the index, but there is P/Q in the index.
2335 * We know P will stay a directory when we check out the contents
2336 * of the index, but we do not know yet if there is a directory
2337 * P/Q in the working tree to be killed, so we need to recurse.
2339 * (3) P does not exist in the index, and there is no P/Q in the index
2340 * to require P to be a directory, either. Only in this case, we
2341 * know that everything inside P will not be killed without
2344 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2345 (dtype
== DT_DIR
) &&
2346 !has_path_in_index
&&
2347 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2350 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2353 * Excluded? If we don't explicitly want to show
2354 * ignored files, ignore it
2356 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2357 return path_excluded
;
2364 * WARNING: Do not ignore/amend the return value from
2365 * treat_directory(), and especially do not change it to return
2366 * path_recurse as that can cause exponential slowdown.
2367 * Instead, modify treat_directory() to return the right value.
2369 strbuf_addch(path
, '/');
2370 return treat_directory(dir
, istate
, untracked
,
2371 path
->buf
, path
->len
,
2372 baselen
, excluded
, pathspec
);
2376 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2377 0 /* prefix */, NULL
/* seen */,
2381 return path_excluded
;
2382 return path_untracked
;
2386 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2390 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2391 dir
->untracked_alloc
);
2392 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2395 static int valid_cached_dir(struct dir_struct
*dir
,
2396 struct untracked_cache_dir
*untracked
,
2397 struct index_state
*istate
,
2398 struct strbuf
*path
,
2407 * With fsmonitor, we can trust the untracked cache's valid field.
2409 refresh_fsmonitor(istate
);
2410 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2411 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2412 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2415 if (!untracked
->valid
||
2416 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2417 fill_stat_data(&untracked
->stat_data
, &st
);
2422 if (untracked
->check_only
!= !!check_only
)
2426 * prep_exclude will be called eventually on this directory,
2427 * but it's called much later in last_matching_pattern(). We
2428 * need it now to determine the validity of the cache for this
2429 * path. The next calls will be nearly no-op, the way
2430 * prep_exclude() is designed.
2432 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2433 strbuf_addch(path
, '/');
2434 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2435 strbuf_setlen(path
, path
->len
- 1);
2437 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2439 /* hopefully prep_exclude() haven't invalidated this entry... */
2440 return untracked
->valid
;
2443 static int open_cached_dir(struct cached_dir
*cdir
,
2444 struct dir_struct
*dir
,
2445 struct untracked_cache_dir
*untracked
,
2446 struct index_state
*istate
,
2447 struct strbuf
*path
,
2452 memset(cdir
, 0, sizeof(*cdir
));
2453 cdir
->untracked
= untracked
;
2454 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2456 c_path
= path
->len
? path
->buf
: ".";
2457 cdir
->fdir
= opendir(c_path
);
2459 warning_errno(_("could not open directory '%s'"), c_path
);
2460 if (dir
->untracked
) {
2461 invalidate_directory(dir
->untracked
, untracked
);
2462 dir
->untracked
->dir_opened
++;
2469 static int read_cached_dir(struct cached_dir
*cdir
)
2474 de
= readdir_skip_dot_and_dotdot(cdir
->fdir
);
2476 cdir
->d_name
= NULL
;
2477 cdir
->d_type
= DT_UNKNOWN
;
2480 cdir
->d_name
= de
->d_name
;
2481 cdir
->d_type
= DTYPE(de
);
2484 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2485 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2495 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2496 struct untracked_cache_dir
*d
= cdir
->untracked
;
2497 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2503 static void close_cached_dir(struct cached_dir
*cdir
)
2506 closedir(cdir
->fdir
);
2508 * We have gone through this directory and found no untracked
2509 * entries. Mark it valid.
2511 if (cdir
->untracked
) {
2512 cdir
->untracked
->valid
= 1;
2513 cdir
->untracked
->recurse
= 1;
2517 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2518 struct untracked_cache_dir
*untracked
,
2519 struct cached_dir
*cdir
,
2520 struct index_state
*istate
,
2521 struct strbuf
*path
,
2523 const struct pathspec
*pathspec
,
2524 enum path_treatment state
)
2526 /* add the path to the appropriate result list */
2529 if (dir
->flags
& DIR_SHOW_IGNORED
)
2530 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2531 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2532 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2533 exclude_matches_pathspec(path
->buf
, path
->len
,
2535 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2538 case path_untracked
:
2539 if (dir
->flags
& DIR_SHOW_IGNORED
)
2541 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2543 add_untracked(untracked
, path
->buf
+ baselen
);
2552 * Read a directory tree. We currently ignore anything but
2553 * directories, regular files and symlinks. That's because git
2554 * doesn't handle them at all yet. Maybe that will change some
2557 * Also, we ignore the name ".git" (even if it is not a directory).
2558 * That likely will not change.
2560 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2561 * to signal that a file was found. This is the least significant value that
2562 * indicates that a file was encountered that does not depend on the order of
2563 * whether an untracked or excluded path was encountered first.
2565 * Returns the most significant path_treatment value encountered in the scan.
2566 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2567 * significant path_treatment value that will be returned.
2570 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2571 struct index_state
*istate
, const char *base
, int baselen
,
2572 struct untracked_cache_dir
*untracked
, int check_only
,
2573 int stop_at_first_file
, const struct pathspec
*pathspec
)
2576 * WARNING: Do NOT recurse unless path_recurse is returned from
2577 * treat_path(). Recursing on any other return value
2578 * can result in exponential slowdown.
2580 struct cached_dir cdir
;
2581 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2582 struct strbuf path
= STRBUF_INIT
;
2584 strbuf_add(&path
, base
, baselen
);
2586 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2588 dir
->internal
.visited_directories
++;
2591 untracked
->check_only
= !!check_only
;
2593 while (!read_cached_dir(&cdir
)) {
2594 /* check how the file or directory should be treated */
2595 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2597 dir
->internal
.visited_paths
++;
2599 if (state
> dir_state
)
2602 /* recurse into subdir if instructed by treat_path */
2603 if (state
== path_recurse
) {
2604 struct untracked_cache_dir
*ud
;
2605 ud
= lookup_untracked(dir
->untracked
,
2608 path
.len
- baselen
);
2610 read_directory_recursive(dir
, istate
, path
.buf
,
2612 check_only
, stop_at_first_file
, pathspec
);
2613 if (subdir_state
> dir_state
)
2614 dir_state
= subdir_state
;
2617 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2618 0 /* prefix */, NULL
,
2619 0 /* do NOT special case dirs */))
2624 if (stop_at_first_file
) {
2626 * If stopping at first file, then
2627 * signal that a file was found by
2628 * returning `path_excluded`. This is
2629 * to return a consistent value
2630 * regardless of whether an ignored or
2631 * excluded file happened to be
2634 * In current usage, the
2635 * `stop_at_first_file` is passed when
2636 * an ancestor directory has matched
2637 * an exclude pattern, so any found
2638 * files will be excluded.
2640 if (dir_state
>= path_excluded
) {
2641 dir_state
= path_excluded
;
2646 /* abort early if maximum state has been reached */
2647 if (dir_state
== path_untracked
) {
2649 add_untracked(untracked
, path
.buf
+ baselen
);
2652 /* skip the add_path_to_appropriate_result_list() */
2656 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2657 istate
, &path
, baselen
,
2660 close_cached_dir(&cdir
);
2662 strbuf_release(&path
);
2667 int cmp_dir_entry(const void *p1
, const void *p2
)
2669 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2670 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2672 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2675 /* check if *out lexically strictly contains *in */
2676 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2678 return (out
->len
< in
->len
) &&
2679 (out
->name
[out
->len
- 1] == '/') &&
2680 !memcmp(out
->name
, in
->name
, out
->len
);
2683 static int treat_leading_path(struct dir_struct
*dir
,
2684 struct index_state
*istate
,
2685 const char *path
, int len
,
2686 const struct pathspec
*pathspec
)
2688 struct strbuf sb
= STRBUF_INIT
;
2689 struct strbuf subdir
= STRBUF_INIT
;
2690 int prevlen
, baselen
;
2692 struct cached_dir cdir
;
2693 enum path_treatment state
= path_none
;
2696 * For each directory component of path, we are going to check whether
2697 * that path is relevant given the pathspec. For example, if path is
2699 * then we will ask treat_path() whether we should go into foo, then
2700 * whether we should go into bar, then whether baz is relevant.
2701 * Checking each is important because e.g. if path is
2703 * then we need to check .git to know we shouldn't traverse it.
2704 * If the return from treat_path() is:
2705 * * path_none, for any path, we return false.
2706 * * path_recurse, for all path components, we return true
2707 * * <anything else> for some intermediate component, we make sure
2708 * to add that path to the relevant list but return false
2709 * signifying that we shouldn't recurse into it.
2712 while (len
&& path
[len
- 1] == '/')
2717 memset(&cdir
, 0, sizeof(cdir
));
2718 cdir
.d_type
= DT_DIR
;
2722 prevlen
= baselen
+ !!baselen
;
2723 cp
= path
+ prevlen
;
2724 cp
= memchr(cp
, '/', path
+ len
- cp
);
2728 baselen
= cp
- path
;
2730 strbuf_add(&sb
, path
, baselen
);
2731 if (!is_directory(sb
.buf
))
2734 strbuf_add(&sb
, path
, prevlen
);
2735 strbuf_reset(&subdir
);
2736 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2737 cdir
.d_name
= subdir
.buf
;
2738 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2740 if (state
!= path_recurse
)
2741 break; /* do not recurse into it */
2743 break; /* finished checking */
2745 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2746 &sb
, baselen
, pathspec
,
2749 strbuf_release(&subdir
);
2750 strbuf_release(&sb
);
2751 return state
== path_recurse
;
2754 static const char *get_ident_string(void)
2756 static struct strbuf sb
= STRBUF_INIT
;
2761 if (uname(&uts
) < 0)
2762 die_errno(_("failed to get kernel name and information"));
2763 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2768 static int ident_in_untracked(const struct untracked_cache
*uc
)
2771 * Previous git versions may have saved many NUL separated
2772 * strings in the "ident" field, but it is insane to manage
2773 * many locations, so just take care of the first one.
2776 return !strcmp(uc
->ident
.buf
, get_ident_string());
2779 static void set_untracked_ident(struct untracked_cache
*uc
)
2781 strbuf_reset(&uc
->ident
);
2782 strbuf_addstr(&uc
->ident
, get_ident_string());
2785 * This strbuf used to contain a list of NUL separated
2786 * strings, so save NUL too for backward compatibility.
2788 strbuf_addch(&uc
->ident
, 0);
2791 static unsigned new_untracked_cache_flags(struct index_state
*istate
)
2793 struct repository
*repo
= istate
->repo
;
2797 * This logic is coordinated with the setting of these flags in
2798 * wt-status.c#wt_status_collect_untracked(), and the evaluation
2799 * of the config setting in commit.c#git_status_config()
2801 if (!repo_config_get_string(repo
, "status.showuntrackedfiles", &val
) &&
2802 !strcmp(val
, "all"))
2806 * The default, if "all" is not set, is "normal" - leading us here.
2807 * If the value is "none" then it really doesn't matter.
2809 return DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2812 static void new_untracked_cache(struct index_state
*istate
, int flags
)
2814 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2815 strbuf_init(&uc
->ident
, 100);
2816 uc
->exclude_per_dir
= ".gitignore";
2817 uc
->dir_flags
= flags
>= 0 ? flags
: new_untracked_cache_flags(istate
);
2818 set_untracked_ident(uc
);
2819 istate
->untracked
= uc
;
2820 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2823 void add_untracked_cache(struct index_state
*istate
)
2825 if (!istate
->untracked
) {
2826 new_untracked_cache(istate
, -1);
2828 if (!ident_in_untracked(istate
->untracked
)) {
2829 free_untracked_cache(istate
->untracked
);
2830 new_untracked_cache(istate
, -1);
2835 void remove_untracked_cache(struct index_state
*istate
)
2837 if (istate
->untracked
) {
2838 free_untracked_cache(istate
->untracked
);
2839 istate
->untracked
= NULL
;
2840 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2844 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2846 const struct pathspec
*pathspec
,
2847 struct index_state
*istate
)
2849 struct untracked_cache_dir
*root
;
2850 static int untracked_cache_disabled
= -1;
2852 if (!dir
->untracked
)
2854 if (untracked_cache_disabled
< 0)
2855 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2856 if (untracked_cache_disabled
)
2860 * We only support $GIT_DIR/info/exclude and core.excludesfile
2861 * as the global ignore rule files. Any other additions
2862 * (e.g. from command line) invalidate the cache. This
2863 * condition also catches running setup_standard_excludes()
2864 * before setting dir->untracked!
2866 if (dir
->internal
.unmanaged_exclude_files
)
2870 * Optimize for the main use case only: whole-tree git
2871 * status. More work involved in treat_leading_path() if we
2872 * use cache on just a subset of the worktree. pathspec
2873 * support could make the matter even worse.
2875 if (base_len
|| (pathspec
&& pathspec
->nr
))
2878 /* We don't support collecting ignore files */
2879 if (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2880 DIR_COLLECT_IGNORED
))
2884 * If we use .gitignore in the cache and now you change it to
2885 * .gitexclude, everything will go wrong.
2887 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2888 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2892 * EXC_CMDL is not considered in the cache. If people set it,
2895 if (dir
->internal
.exclude_list_group
[EXC_CMDL
].nr
)
2898 if (!ident_in_untracked(dir
->untracked
)) {
2899 warning(_("untracked cache is disabled on this system or location"));
2904 * If the untracked structure we received does not have the same flags
2905 * as requested in this run, we're going to need to either discard the
2906 * existing structure (and potentially later recreate), or bypass the
2907 * untracked cache mechanism for this run.
2909 if (dir
->flags
!= dir
->untracked
->dir_flags
) {
2911 * If the untracked structure we received does not have the same flags
2912 * as configured, then we need to reset / create a new "untracked"
2913 * structure to match the new config.
2915 * Keeping the saved and used untracked cache consistent with the
2916 * configuration provides an opportunity for frequent users of
2917 * "git status -uall" to leverage the untracked cache by aligning their
2918 * configuration - setting "status.showuntrackedfiles" to "all" or
2919 * "normal" as appropriate.
2921 * Previously using -uall (or setting "status.showuntrackedfiles" to
2922 * "all") was incompatible with untracked cache and *consistently*
2923 * caused surprisingly bad performance (with fscache and fsmonitor
2924 * enabled) on Windows.
2926 * IMPROVEMENT OPPORTUNITY: If we reworked the untracked cache storage
2927 * to not be as bound up with the desired output in a given run,
2928 * and instead iterated through and stored enough information to
2929 * correctly serve both "modes", then users could get peak performance
2930 * with or without '-uall' regardless of their
2931 * "status.showuntrackedfiles" config.
2933 if (dir
->untracked
->dir_flags
!= new_untracked_cache_flags(istate
)) {
2934 free_untracked_cache(istate
->untracked
);
2935 new_untracked_cache(istate
, dir
->flags
);
2936 dir
->untracked
= istate
->untracked
;
2940 * Current untracked cache data is consistent with config, but not
2941 * usable in this request/run; just bypass untracked cache.
2947 if (!dir
->untracked
->root
) {
2948 /* Untracked cache existed but is not initialized; fix that */
2949 FLEX_ALLOC_STR(dir
->untracked
->root
, name
, "");
2950 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2953 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2954 root
= dir
->untracked
->root
;
2955 if (!oideq(&dir
->internal
.ss_info_exclude
.oid
,
2956 &dir
->untracked
->ss_info_exclude
.oid
)) {
2957 invalidate_gitignore(dir
->untracked
, root
);
2958 dir
->untracked
->ss_info_exclude
= dir
->internal
.ss_info_exclude
;
2960 if (!oideq(&dir
->internal
.ss_excludes_file
.oid
,
2961 &dir
->untracked
->ss_excludes_file
.oid
)) {
2962 invalidate_gitignore(dir
->untracked
, root
);
2963 dir
->untracked
->ss_excludes_file
= dir
->internal
.ss_excludes_file
;
2966 /* Make sure this directory is not dropped out at saving phase */
2971 static void emit_traversal_statistics(struct dir_struct
*dir
,
2972 struct repository
*repo
,
2976 if (!trace2_is_enabled())
2980 trace2_data_string("read_directory", repo
, "path", "");
2982 struct strbuf tmp
= STRBUF_INIT
;
2983 strbuf_add(&tmp
, path
, path_len
);
2984 trace2_data_string("read_directory", repo
, "path", tmp
.buf
);
2985 strbuf_release(&tmp
);
2988 trace2_data_intmax("read_directory", repo
,
2989 "directories-visited", dir
->internal
.visited_directories
);
2990 trace2_data_intmax("read_directory", repo
,
2991 "paths-visited", dir
->internal
.visited_paths
);
2993 if (!dir
->untracked
)
2995 trace2_data_intmax("read_directory", repo
,
2996 "node-creation", dir
->untracked
->dir_created
);
2997 trace2_data_intmax("read_directory", repo
,
2998 "gitignore-invalidation",
2999 dir
->untracked
->gitignore_invalidated
);
3000 trace2_data_intmax("read_directory", repo
,
3001 "directory-invalidation",
3002 dir
->untracked
->dir_invalidated
);
3003 trace2_data_intmax("read_directory", repo
,
3004 "opendir", dir
->untracked
->dir_opened
);
3007 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
3008 const char *path
, int len
, const struct pathspec
*pathspec
)
3010 struct untracked_cache_dir
*untracked
;
3012 trace2_region_enter("dir", "read_directory", istate
->repo
);
3013 dir
->internal
.visited_paths
= 0;
3014 dir
->internal
.visited_directories
= 0;
3016 if (has_symlink_leading_path(path
, len
)) {
3017 trace2_region_leave("dir", "read_directory", istate
->repo
);
3021 untracked
= validate_untracked_cache(dir
, len
, pathspec
, istate
);
3024 * make sure untracked cache code path is disabled,
3025 * e.g. prep_exclude()
3027 dir
->untracked
= NULL
;
3028 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
3029 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
3030 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
3031 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
3033 emit_traversal_statistics(dir
, istate
->repo
, path
, len
);
3035 trace2_region_leave("dir", "read_directory", istate
->repo
);
3036 if (dir
->untracked
) {
3037 static int force_untracked_cache
= -1;
3039 if (force_untracked_cache
< 0)
3040 force_untracked_cache
=
3041 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", -1);
3042 if (force_untracked_cache
< 0)
3043 force_untracked_cache
= (istate
->repo
->settings
.core_untracked_cache
== UNTRACKED_CACHE_WRITE
);
3044 if (force_untracked_cache
&&
3045 dir
->untracked
== istate
->untracked
&&
3046 (dir
->untracked
->dir_opened
||
3047 dir
->untracked
->gitignore_invalidated
||
3048 dir
->untracked
->dir_invalidated
))
3049 istate
->cache_changed
|= UNTRACKED_CHANGED
;
3050 if (dir
->untracked
!= istate
->untracked
) {
3051 FREE_AND_NULL(dir
->untracked
);
3058 int file_exists(const char *f
)
3061 return lstat(f
, &sb
) == 0;
3064 int repo_file_exists(struct repository
*repo
, const char *path
)
3066 if (repo
!= the_repository
)
3067 BUG("do not know how to check file existence in arbitrary repo");
3069 return file_exists(path
);
3072 static int cmp_icase(char a
, char b
)
3077 return toupper(a
) - toupper(b
);
3082 * Given two normalized paths (a trailing slash is ok), if subdir is
3083 * outside dir, return -1. Otherwise return the offset in subdir that
3084 * can be used as relative path to dir.
3086 int dir_inside_of(const char *subdir
, const char *dir
)
3090 assert(dir
&& subdir
&& *dir
&& *subdir
);
3092 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
3098 /* hel[p]/me vs hel[l]/yeah */
3099 if (*dir
&& *subdir
)
3103 return !*dir
? offset
: -1; /* same dir */
3105 /* foo/[b]ar vs foo/[] */
3106 if (is_dir_sep(dir
[-1]))
3107 return is_dir_sep(subdir
[-1]) ? offset
: -1;
3109 /* foo[/]bar vs foo[] */
3110 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
3113 int is_inside_dir(const char *dir
)
3122 rc
= (dir_inside_of(cwd
, dir
) >= 0);
3127 int is_empty_dir(const char *path
)
3129 DIR *dir
= opendir(path
);
3136 e
= readdir_skip_dot_and_dotdot(dir
);
3144 char *git_url_basename(const char *repo
, int is_bundle
, int is_bare
)
3146 const char *end
= repo
+ strlen(repo
), *start
, *ptr
;
3153 start
= strstr(repo
, "://");
3160 * Skip authentication data. The stripping does happen
3161 * greedily, such that we strip up to the last '@' inside
3164 for (ptr
= start
; ptr
< end
&& !is_dir_sep(*ptr
); ptr
++) {
3170 * Strip trailing spaces, slashes and /.git
3172 while (start
< end
&& (is_dir_sep(end
[-1]) || isspace(end
[-1])))
3174 if (end
- start
> 5 && is_dir_sep(end
[-5]) &&
3175 !strncmp(end
- 4, ".git", 4)) {
3177 while (start
< end
&& is_dir_sep(end
[-1]))
3182 * It should not be possible to overflow `ptrdiff_t` by passing in an
3183 * insanely long URL, but GCC does not know that and will complain
3184 * without this check.
3186 if (end
- start
< 0)
3187 die(_("No directory name could be guessed.\n"
3188 "Please specify a directory on the command line"));
3191 * Strip trailing port number if we've got only a
3192 * hostname (that is, there is no dir separator but a
3193 * colon). This check is required such that we do not
3194 * strip URI's like '/foo/bar:2222.git', which should
3195 * result in a dir '2222' being guessed due to backwards
3198 if (memchr(start
, '/', end
- start
) == NULL
3199 && memchr(start
, ':', end
- start
) != NULL
) {
3201 while (start
< ptr
&& isdigit(ptr
[-1]) && ptr
[-1] != ':')
3203 if (start
< ptr
&& ptr
[-1] == ':')
3208 * Find last component. To remain backwards compatible we
3209 * also regard colons as path separators, such that
3210 * cloning a repository 'foo:bar.git' would result in a
3211 * directory 'bar' being guessed.
3214 while (start
< ptr
&& !is_dir_sep(ptr
[-1]) && ptr
[-1] != ':')
3219 * Strip .{bundle,git}.
3222 strip_suffix_mem(start
, &len
, is_bundle
? ".bundle" : ".git");
3224 if (!len
|| (len
== 1 && *start
== '/'))
3225 die(_("No directory name could be guessed.\n"
3226 "Please specify a directory on the command line"));
3229 dir
= xstrfmt("%.*s.git", (int)len
, start
);
3231 dir
= xstrndup(start
, len
);
3233 * Replace sequences of 'control' characters and whitespace
3234 * with one ascii space, remove leading and trailing spaces.
3238 int prev_space
= 1 /* strip leading whitespace */;
3239 for (end
= dir
; *end
; ++end
) {
3241 if ((unsigned char)ch
< '\x20')
3252 if (out
> dir
&& prev_space
)
3258 void strip_dir_trailing_slashes(char *dir
)
3260 char *end
= dir
+ strlen(dir
);
3262 while (dir
< end
- 1 && is_dir_sep(end
[-1]))
3267 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
3271 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
3272 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
3273 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
3274 int purge_original_cwd
= (flag
& REMOVE_DIR_PURGE_ORIGINAL_CWD
);
3275 struct object_id submodule_head
;
3277 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
3278 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
3279 /* Do not descend and nuke a nested git work tree. */
3285 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
3286 dir
= opendir(path
->buf
);
3288 if (errno
== ENOENT
)
3289 return keep_toplevel
? -1 : 0;
3290 else if (errno
== EACCES
&& !keep_toplevel
)
3292 * An empty dir could be removable even if it
3295 return rmdir(path
->buf
);
3299 strbuf_complete(path
, '/');
3302 while ((e
= readdir_skip_dot_and_dotdot(dir
)) != NULL
) {
3305 strbuf_setlen(path
, len
);
3306 strbuf_addstr(path
, e
->d_name
);
3307 if (lstat(path
->buf
, &st
)) {
3308 if (errno
== ENOENT
)
3310 * file disappeared, which is what we
3315 } else if (S_ISDIR(st
.st_mode
)) {
3316 if (!remove_dir_recurse(path
, flag
, &kept_down
))
3317 continue; /* happy */
3318 } else if (!only_empty
&&
3319 (!unlink(path
->buf
) || errno
== ENOENT
)) {
3320 continue; /* happy, too */
3323 /* path too long, stat fails, or non-directory still exists */
3329 strbuf_setlen(path
, original_len
);
3330 if (!ret
&& !keep_toplevel
&& !kept_down
) {
3331 if (!purge_original_cwd
&&
3332 startup_info
->original_cwd
&&
3333 !strcmp(startup_info
->original_cwd
, path
->buf
))
3334 ret
= -1; /* Do not remove current working directory */
3336 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
3339 * report the uplevel that it is not an error that we
3340 * did not rmdir() our directory.
3346 int remove_dir_recursively(struct strbuf
*path
, int flag
)
3348 return remove_dir_recurse(path
, flag
, NULL
);
3351 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
3353 void setup_standard_excludes(struct dir_struct
*dir
)
3355 dir
->exclude_per_dir
= ".gitignore";
3357 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3359 excludes_file
= xdg_config_home("ignore");
3360 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
3361 add_patterns_from_file_1(dir
, excludes_file
,
3362 dir
->untracked
? &dir
->internal
.ss_excludes_file
: NULL
);
3364 /* per repository user preference */
3365 if (startup_info
->have_repository
) {
3366 const char *path
= git_path_info_exclude();
3367 if (!access_or_warn(path
, R_OK
, 0))
3368 add_patterns_from_file_1(dir
, path
,
3369 dir
->untracked
? &dir
->internal
.ss_info_exclude
: NULL
);
3373 char *get_sparse_checkout_filename(void)
3375 return git_pathdup("info/sparse-checkout");
3378 int get_sparse_checkout_patterns(struct pattern_list
*pl
)
3381 char *sparse_filename
= get_sparse_checkout_filename();
3383 pl
->use_cone_patterns
= core_sparse_checkout_cone
;
3384 res
= add_patterns_from_file_to_list(sparse_filename
, "", 0, pl
, NULL
, 0);
3386 free(sparse_filename
);
3390 int remove_path(const char *name
)
3394 if (unlink(name
) && !is_missing_file_error(errno
))
3397 slash
= strrchr(name
, '/');
3399 char *dirs
= xstrdup(name
);
3400 slash
= dirs
+ (slash
- name
);
3403 if (startup_info
->original_cwd
&&
3404 !strcmp(startup_info
->original_cwd
, dirs
))
3406 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3413 * Frees memory within dir which was allocated, and resets fields for further
3414 * use. Does not free dir itself.
3416 void dir_clear(struct dir_struct
*dir
)
3419 struct exclude_list_group
*group
;
3420 struct pattern_list
*pl
;
3421 struct exclude_stack
*stk
;
3422 struct dir_struct
new = DIR_INIT
;
3424 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3425 group
= &dir
->internal
.exclude_list_group
[i
];
3426 for (j
= 0; j
< group
->nr
; j
++) {
3429 free((char *)pl
->src
);
3430 clear_pattern_list(pl
);
3435 for (i
= 0; i
< dir
->ignored_nr
; i
++)
3436 free(dir
->ignored
[i
]);
3437 for (i
= 0; i
< dir
->nr
; i
++)
3438 free(dir
->entries
[i
]);
3442 stk
= dir
->internal
.exclude_stack
;
3444 struct exclude_stack
*prev
= stk
->prev
;
3448 strbuf_release(&dir
->internal
.basebuf
);
3450 memcpy(dir
, &new, sizeof(*dir
));
3453 struct ondisk_untracked_cache
{
3454 struct stat_data info_exclude_stat
;
3455 struct stat_data excludes_file_stat
;
3459 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3462 int index
; /* number of written untracked_cache_dir */
3463 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3464 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3465 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3467 struct strbuf sb_stat
;
3468 struct strbuf sb_sha1
;
3471 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3473 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3474 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3475 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3476 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3477 to
->sd_dev
= htonl(from
->sd_dev
);
3478 to
->sd_ino
= htonl(from
->sd_ino
);
3479 to
->sd_uid
= htonl(from
->sd_uid
);
3480 to
->sd_gid
= htonl(from
->sd_gid
);
3481 to
->sd_size
= htonl(from
->sd_size
);
3484 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3485 struct write_data
*wd
)
3487 struct stat_data stat_data
;
3488 struct strbuf
*out
= &wd
->out
;
3489 unsigned char intbuf
[16];
3490 unsigned int intlen
, value
;
3491 int i
= wd
->index
++;
3494 * untracked_nr should be reset whenever valid is clear, but
3497 if (!untracked
->valid
) {
3498 untracked
->untracked_nr
= 0;
3499 untracked
->check_only
= 0;
3502 if (untracked
->check_only
)
3503 ewah_set(wd
->check_only
, i
);
3504 if (untracked
->valid
) {
3505 ewah_set(wd
->valid
, i
);
3506 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3507 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3509 if (!is_null_oid(&untracked
->exclude_oid
)) {
3510 ewah_set(wd
->sha1_valid
, i
);
3511 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3512 the_hash_algo
->rawsz
);
3515 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3516 strbuf_add(out
, intbuf
, intlen
);
3518 /* skip non-recurse directories */
3519 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3520 if (untracked
->dirs
[i
]->recurse
)
3522 intlen
= encode_varint(value
, intbuf
);
3523 strbuf_add(out
, intbuf
, intlen
);
3525 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3527 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3528 strbuf_add(out
, untracked
->untracked
[i
],
3529 strlen(untracked
->untracked
[i
]) + 1);
3531 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3532 if (untracked
->dirs
[i
]->recurse
)
3533 write_one_dir(untracked
->dirs
[i
], wd
);
3536 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3538 struct ondisk_untracked_cache
*ouc
;
3539 struct write_data wd
;
3540 unsigned char varbuf
[16];
3542 const unsigned hashsz
= the_hash_algo
->rawsz
;
3544 CALLOC_ARRAY(ouc
, 1);
3545 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3546 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3547 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3549 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3550 strbuf_add(out
, varbuf
, varint_len
);
3551 strbuf_addbuf(out
, &untracked
->ident
);
3553 strbuf_add(out
, ouc
, sizeof(*ouc
));
3554 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3555 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3556 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3559 if (!untracked
->root
) {
3560 varint_len
= encode_varint(0, varbuf
);
3561 strbuf_add(out
, varbuf
, varint_len
);
3566 wd
.check_only
= ewah_new();
3567 wd
.valid
= ewah_new();
3568 wd
.sha1_valid
= ewah_new();
3569 strbuf_init(&wd
.out
, 1024);
3570 strbuf_init(&wd
.sb_stat
, 1024);
3571 strbuf_init(&wd
.sb_sha1
, 1024);
3572 write_one_dir(untracked
->root
, &wd
);
3574 varint_len
= encode_varint(wd
.index
, varbuf
);
3575 strbuf_add(out
, varbuf
, varint_len
);
3576 strbuf_addbuf(out
, &wd
.out
);
3577 ewah_serialize_strbuf(wd
.valid
, out
);
3578 ewah_serialize_strbuf(wd
.check_only
, out
);
3579 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3580 strbuf_addbuf(out
, &wd
.sb_stat
);
3581 strbuf_addbuf(out
, &wd
.sb_sha1
);
3582 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3584 ewah_free(wd
.valid
);
3585 ewah_free(wd
.check_only
);
3586 ewah_free(wd
.sha1_valid
);
3587 strbuf_release(&wd
.out
);
3588 strbuf_release(&wd
.sb_stat
);
3589 strbuf_release(&wd
.sb_sha1
);
3592 static void free_untracked(struct untracked_cache_dir
*ucd
)
3597 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3598 free_untracked(ucd
->dirs
[i
]);
3599 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3600 free(ucd
->untracked
[i
]);
3601 free(ucd
->untracked
);
3606 void free_untracked_cache(struct untracked_cache
*uc
)
3611 free(uc
->exclude_per_dir_to_free
);
3612 strbuf_release(&uc
->ident
);
3613 free_untracked(uc
->root
);
3619 struct untracked_cache_dir
**ucd
;
3620 struct ewah_bitmap
*check_only
;
3621 struct ewah_bitmap
*valid
;
3622 struct ewah_bitmap
*sha1_valid
;
3623 const unsigned char *data
;
3624 const unsigned char *end
;
3627 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3629 memcpy(to
, data
, sizeof(*to
));
3630 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3631 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3632 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3633 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3634 to
->sd_dev
= ntohl(to
->sd_dev
);
3635 to
->sd_ino
= ntohl(to
->sd_ino
);
3636 to
->sd_uid
= ntohl(to
->sd_uid
);
3637 to
->sd_gid
= ntohl(to
->sd_gid
);
3638 to
->sd_size
= ntohl(to
->sd_size
);
3641 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3642 struct read_data
*rd
)
3644 struct untracked_cache_dir ud
, *untracked
;
3645 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3646 const unsigned char *eos
;
3650 memset(&ud
, 0, sizeof(ud
));
3652 value
= decode_varint(&data
);
3656 ud
.untracked_alloc
= value
;
3657 ud
.untracked_nr
= value
;
3658 if (ud
.untracked_nr
)
3659 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3661 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3664 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3666 eos
= memchr(data
, '\0', end
- data
);
3667 if (!eos
|| eos
== end
)
3670 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3671 memcpy(untracked
, &ud
, sizeof(ud
));
3672 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3675 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3676 eos
= memchr(data
, '\0', end
- data
);
3677 if (!eos
|| eos
== end
)
3679 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3683 rd
->ucd
[rd
->index
++] = untracked
;
3686 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3687 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3693 static void set_check_only(size_t pos
, void *cb
)
3695 struct read_data
*rd
= cb
;
3696 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3700 static void read_stat(size_t pos
, void *cb
)
3702 struct read_data
*rd
= cb
;
3703 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3704 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3705 rd
->data
= rd
->end
+ 1;
3708 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3709 rd
->data
+= sizeof(struct stat_data
);
3713 static void read_oid(size_t pos
, void *cb
)
3715 struct read_data
*rd
= cb
;
3716 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3717 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3718 rd
->data
= rd
->end
+ 1;
3721 oidread(&ud
->exclude_oid
, rd
->data
);
3722 rd
->data
+= the_hash_algo
->rawsz
;
3725 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3726 const unsigned char *sha1
)
3728 stat_data_from_disk(&oid_stat
->stat
, data
);
3729 oidread(&oid_stat
->oid
, sha1
);
3730 oid_stat
->valid
= 1;
3733 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3735 struct untracked_cache
*uc
;
3736 struct read_data rd
;
3737 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3741 const char *exclude_per_dir
;
3742 const unsigned hashsz
= the_hash_algo
->rawsz
;
3743 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3744 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3746 if (sz
<= 1 || end
[-1] != '\0')
3750 ident_len
= decode_varint(&next
);
3751 if (next
+ ident_len
> end
)
3753 ident
= (const char *)next
;
3756 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3759 CALLOC_ARRAY(uc
, 1);
3760 strbuf_init(&uc
->ident
, ident_len
);
3761 strbuf_add(&uc
->ident
, ident
, ident_len
);
3762 load_oid_stat(&uc
->ss_info_exclude
,
3763 next
+ ouc_offset(info_exclude_stat
),
3765 load_oid_stat(&uc
->ss_excludes_file
,
3766 next
+ ouc_offset(excludes_file_stat
),
3767 next
+ offset
+ hashsz
);
3768 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3769 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3770 uc
->exclude_per_dir
= uc
->exclude_per_dir_to_free
= xstrdup(exclude_per_dir
);
3771 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3772 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3776 len
= decode_varint(&next
);
3777 if (next
> end
|| len
== 0)
3780 rd
.valid
= ewah_new();
3781 rd
.check_only
= ewah_new();
3782 rd
.sha1_valid
= ewah_new();
3786 ALLOC_ARRAY(rd
.ucd
, len
);
3788 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3792 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3797 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3802 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3806 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3807 rd
.data
= next
+ len
;
3808 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3809 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3814 ewah_free(rd
.valid
);
3815 ewah_free(rd
.check_only
);
3816 ewah_free(rd
.sha1_valid
);
3819 free_untracked_cache(uc
);
3825 static void invalidate_one_directory(struct untracked_cache
*uc
,
3826 struct untracked_cache_dir
*ucd
)
3828 uc
->dir_invalidated
++;
3830 ucd
->untracked_nr
= 0;
3834 * Normally when an entry is added or removed from a directory,
3835 * invalidating that directory is enough. No need to touch its
3836 * ancestors. When a directory is shown as "foo/bar/" in git-status
3837 * however, deleting or adding an entry may have cascading effect.
3839 * Say the "foo/bar/file" has become untracked, we need to tell the
3840 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3841 * directory any more (because "bar" is managed by foo as an untracked
3844 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3845 * was the last untracked entry in the entire "foo", we should show
3846 * "foo/" instead. Which means we have to invalidate past "bar" up to
3849 * This function traverses all directories from root to leaf. If there
3850 * is a chance of one of the above cases happening, we invalidate back
3851 * to root. Otherwise we just invalidate the leaf. There may be a more
3852 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3853 * detect these cases and avoid unnecessary invalidation, for example,
3854 * checking for the untracked entry named "bar/" in "foo", but for now
3855 * stick to something safe and simple.
3857 static int invalidate_one_component(struct untracked_cache
*uc
,
3858 struct untracked_cache_dir
*dir
,
3859 const char *path
, int len
)
3861 const char *rest
= strchr(path
, '/');
3864 int component_len
= rest
- path
;
3865 struct untracked_cache_dir
*d
=
3866 lookup_untracked(uc
, dir
, path
, component_len
);
3868 invalidate_one_component(uc
, d
, rest
+ 1,
3869 len
- (component_len
+ 1));
3871 invalidate_one_directory(uc
, dir
);
3875 invalidate_one_directory(uc
, dir
);
3876 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3879 void untracked_cache_invalidate_path(struct index_state
*istate
,
3880 const char *path
, int safe_path
)
3882 if (!istate
->untracked
|| !istate
->untracked
->root
)
3884 if (!safe_path
&& !verify_path(path
, 0))
3886 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3887 path
, strlen(path
));
3890 void untracked_cache_remove_from_index(struct index_state
*istate
,
3893 untracked_cache_invalidate_path(istate
, path
, 1);
3896 void untracked_cache_add_to_index(struct index_state
*istate
,
3899 untracked_cache_invalidate_path(istate
, path
, 1);
3902 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3903 const char *sub_gitdir
)
3906 struct repository subrepo
;
3907 struct strbuf sub_wt
= STRBUF_INIT
;
3908 struct strbuf sub_gd
= STRBUF_INIT
;
3910 const struct submodule
*sub
;
3912 /* If the submodule has no working tree, we can ignore it. */
3913 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3916 if (repo_read_index(&subrepo
) < 0)
3917 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3919 /* TODO: audit for interaction with sparse-index. */
3920 ensure_full_index(subrepo
.index
);
3921 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3922 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3924 if (!S_ISGITLINK(ce
->ce_mode
))
3927 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3928 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3930 * Skip entries with the same name in different stages
3931 * to make sure an entry is returned only once.
3935 sub
= submodule_from_path(&subrepo
, null_oid(), ce
->name
);
3936 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3937 /* .gitmodules broken or inactive sub */
3940 strbuf_reset(&sub_wt
);
3941 strbuf_reset(&sub_gd
);
3942 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3943 submodule_name_to_gitdir(&sub_gd
, &subrepo
, sub
->name
);
3945 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3947 strbuf_release(&sub_wt
);
3948 strbuf_release(&sub_gd
);
3949 repo_clear(&subrepo
);
3952 void connect_work_tree_and_git_dir(const char *work_tree_
,
3953 const char *git_dir_
,
3954 int recurse_into_nested
)
3956 struct strbuf gitfile_sb
= STRBUF_INIT
;
3957 struct strbuf cfg_sb
= STRBUF_INIT
;
3958 struct strbuf rel_path
= STRBUF_INIT
;
3959 char *git_dir
, *work_tree
;
3961 /* Prepare .git file */
3962 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3963 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3964 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3966 /* Prepare config file */
3967 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3968 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3969 die(_("could not create directories for %s"), cfg_sb
.buf
);
3971 git_dir
= real_pathdup(git_dir_
, 1);
3972 work_tree
= real_pathdup(work_tree_
, 1);
3974 /* Write .git file */
3975 write_file(gitfile_sb
.buf
, "gitdir: %s",
3976 relative_path(git_dir
, work_tree
, &rel_path
));
3977 /* Update core.worktree setting */
3978 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3979 relative_path(work_tree
, git_dir
, &rel_path
));
3981 strbuf_release(&gitfile_sb
);
3982 strbuf_release(&cfg_sb
);
3983 strbuf_release(&rel_path
);
3985 if (recurse_into_nested
)
3986 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3993 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3995 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3997 if (rename(old_git_dir
, new_git_dir
) < 0)
3998 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3999 old_git_dir
, new_git_dir
);
4001 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);
4004 int path_match_flags(const char *const str
, const enum path_match_flags flags
)
4006 const char *p
= str
;
4008 if (flags
& PATH_MATCH_NATIVE
&&
4009 flags
& PATH_MATCH_XPLATFORM
)
4010 BUG("path_match_flags() must get one match kind, not multiple!");
4011 else if (!(flags
& PATH_MATCH_KINDS_MASK
))
4012 BUG("path_match_flags() must get at least one match kind!");
4014 if (flags
& PATH_MATCH_STARTS_WITH_DOT_SLASH
&&
4015 flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
)
4016 BUG("path_match_flags() must get one platform kind, not multiple!");
4017 else if (!(flags
& PATH_MATCH_PLATFORM_MASK
))
4018 BUG("path_match_flags() must get at least one platform kind!");
4022 if (flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
&&
4026 if (flags
& PATH_MATCH_NATIVE
)
4027 return is_dir_sep(*p
);
4028 else if (flags
& PATH_MATCH_XPLATFORM
)
4029 return is_xplatform_dir_sep(*p
);