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 "object-store.h"
17 #include "wildmatch.h"
21 #include "ewah/ewok.h"
22 #include "fsmonitor.h"
23 #include "submodule-config.h"
26 * Tells read_directory_recursive how a file or directory should be treated.
27 * Values are ordered by significance, e.g. if a directory contains both
28 * excluded and untracked files, it is listed as untracked because
29 * path_untracked > path_excluded.
39 * Support data structure for our opendir/readdir/closedir wrappers
43 struct untracked_cache_dir
*untracked
;
50 struct untracked_cache_dir
*ucd
;
53 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
54 struct index_state
*istate
, const char *path
, int len
,
55 struct untracked_cache_dir
*untracked
,
56 int check_only
, int stop_at_first_file
, const struct pathspec
*pathspec
);
57 static int resolve_dtype(int dtype
, struct index_state
*istate
,
58 const char *path
, int len
);
59 struct dirent
*readdir_skip_dot_and_dotdot(DIR *dirp
)
63 while ((e
= readdir(dirp
)) != NULL
) {
64 if (!is_dot_or_dotdot(e
->d_name
))
70 int count_slashes(const char *s
)
79 int fspathcmp(const char *a
, const char *b
)
81 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
84 int fspatheq(const char *a
, const char *b
)
86 return !fspathcmp(a
, b
);
89 int fspathncmp(const char *a
, const char *b
, size_t count
)
91 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
94 unsigned int fspathhash(const char *str
)
96 return ignore_case
? strihash(str
) : strhash(str
);
99 int git_fnmatch(const struct pathspec_item
*item
,
100 const char *pattern
, const char *string
,
104 if (ps_strncmp(item
, pattern
, string
, prefix
))
109 if (item
->flags
& PATHSPEC_ONESTAR
) {
110 int pattern_len
= strlen(++pattern
);
111 int string_len
= strlen(string
);
112 return string_len
< pattern_len
||
113 ps_strcmp(item
, pattern
,
114 string
+ string_len
- pattern_len
);
116 if (item
->magic
& PATHSPEC_GLOB
)
117 return wildmatch(pattern
, string
,
119 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
121 /* wildmatch has not learned no FNM_PATHNAME mode yet */
122 return wildmatch(pattern
, string
,
123 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
126 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
127 const char *string
, int stringlen
,
131 struct strbuf pat_buf
= STRBUF_INIT
;
132 struct strbuf str_buf
= STRBUF_INIT
;
133 const char *use_pat
= pattern
;
134 const char *use_str
= string
;
136 if (pattern
[patternlen
]) {
137 strbuf_add(&pat_buf
, pattern
, patternlen
);
138 use_pat
= pat_buf
.buf
;
140 if (string
[stringlen
]) {
141 strbuf_add(&str_buf
, string
, stringlen
);
142 use_str
= str_buf
.buf
;
146 flags
|= WM_CASEFOLD
;
147 match_status
= wildmatch(use_pat
, use_str
, flags
);
149 strbuf_release(&pat_buf
);
150 strbuf_release(&str_buf
);
155 static size_t common_prefix_len(const struct pathspec
*pathspec
)
161 * ":(icase)path" is treated as a pathspec full of
162 * wildcard. In other words, only prefix is considered common
163 * prefix. If the pathspec is abc/foo abc/bar, running in
164 * subdir xyz, the common prefix is still xyz, not xyz/abc as
167 GUARD_PATHSPEC(pathspec
,
176 for (n
= 0; n
< pathspec
->nr
; n
++) {
177 size_t i
= 0, len
= 0, item_len
;
178 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
180 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
181 item_len
= pathspec
->items
[n
].prefix
;
183 item_len
= pathspec
->items
[n
].nowildcard_len
;
184 while (i
< item_len
&& (n
== 0 || i
< max
)) {
185 char c
= pathspec
->items
[n
].match
[i
];
186 if (c
!= pathspec
->items
[0].match
[i
])
192 if (n
== 0 || len
< max
) {
202 * Returns a copy of the longest leading path common among all
205 char *common_prefix(const struct pathspec
*pathspec
)
207 unsigned long len
= common_prefix_len(pathspec
);
209 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
212 int fill_directory(struct dir_struct
*dir
,
213 struct index_state
*istate
,
214 const struct pathspec
*pathspec
)
219 unsigned exclusive_flags
= DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
;
220 if ((dir
->flags
& exclusive_flags
) == exclusive_flags
)
221 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
224 * Calculate common prefix for the pathspec, and
225 * use that to optimize the directory walk
227 prefix_len
= common_prefix_len(pathspec
);
228 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
230 /* Read the directory and prune it */
231 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
236 int within_depth(const char *name
, int namelen
,
237 int depth
, int max_depth
)
239 const char *cp
= name
, *cpe
= name
+ namelen
;
245 if (depth
> max_depth
)
252 * Read the contents of the blob with the given OID into a buffer.
253 * Append a trailing LF to the end if the last line doesn't have one.
256 * -1 when the OID is invalid or unknown or does not refer to a blob.
257 * 0 when the blob is empty.
258 * 1 along with { data, size } of the (possibly augmented) buffer
261 * Optionally updates the given oid_stat with the given OID (when valid).
263 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
264 size_t *size_out
, char **data_out
)
266 enum object_type type
;
273 data
= read_object_file(oid
, &type
, &sz
);
274 if (!data
|| type
!= OBJ_BLOB
) {
280 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
281 oidcpy(&oid_stat
->oid
, oid
);
289 if (data
[sz
- 1] != '\n') {
290 data
= xrealloc(data
, st_add(sz
, 1));
294 *size_out
= xsize_t(sz
);
300 #define DO_MATCH_EXCLUDE (1<<0)
301 #define DO_MATCH_DIRECTORY (1<<1)
302 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
305 * Does the given pathspec match the given name? A match is found if
307 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
308 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
309 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
310 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
312 * Return value tells which case it was (1-4), or 0 when there is no match.
314 * It may be instructive to look at a small table of concrete examples
315 * to understand the differences between 1, 2, and 4:
318 * | a/b | a/b/ | a/b/c
319 * ------+-----------+-----------+------------
320 * a/b | EXACT | EXACT[1] | LEADING[2]
321 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
322 * a/b/c | RECURSIVE | RECURSIVE | EXACT
324 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
325 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
327 static int match_pathspec_item(struct index_state
*istate
,
328 const struct pathspec_item
*item
, int prefix
,
329 const char *name
, int namelen
, unsigned flags
)
331 /* name/namelen has prefix cut off by caller */
332 const char *match
= item
->match
+ prefix
;
333 int matchlen
= item
->len
- prefix
;
336 * The normal call pattern is:
337 * 1. prefix = common_prefix_len(ps);
338 * 2. prune something, or fill_directory
339 * 3. match_pathspec()
341 * 'prefix' at #1 may be shorter than the command's prefix and
342 * it's ok for #2 to match extra files. Those extras will be
345 * Suppose the pathspec is 'foo' and '../bar' running from
346 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
347 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
348 * user does not want XYZ/foo, only the "foo" part should be
349 * case-insensitive. We need to filter out XYZ/foo here. In
350 * other words, we do not trust the caller on comparing the
351 * prefix part when :(icase) is involved. We do exact
352 * comparison ourselves.
354 * Normally the caller (common_prefix_len() in fact) does
355 * _exact_ matching on name[-prefix+1..-1] and we do not need
356 * to check that part. Be defensive and check it anyway, in
357 * case common_prefix_len is changed, or a new caller is
358 * introduced that does not use common_prefix_len.
360 * If the penalty turns out too high when prefix is really
361 * long, maybe change it to
362 * strncmp(match, name, item->prefix - prefix)
364 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
365 strncmp(item
->match
, name
- prefix
, item
->prefix
))
368 if (item
->attr_match_nr
&&
369 !match_pathspec_attrs(istate
, name
, namelen
, item
))
372 /* If the match was just the prefix, we matched */
374 return MATCHED_RECURSIVELY
;
376 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
377 if (matchlen
== namelen
)
378 return MATCHED_EXACTLY
;
380 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
381 return MATCHED_RECURSIVELY
;
382 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
383 match
[matchlen
- 1] == '/' &&
384 namelen
== matchlen
- 1 &&
385 !ps_strncmp(item
, match
, name
, namelen
))
386 return MATCHED_EXACTLY
;
388 if (item
->nowildcard_len
< item
->len
&&
389 !git_fnmatch(item
, match
, name
,
390 item
->nowildcard_len
- prefix
))
391 return MATCHED_FNMATCH
;
393 /* Perform checks to see if "name" is a leading string of the pathspec */
394 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
395 !(flags
& DO_MATCH_EXCLUDE
)) {
396 /* name is a literal prefix of the pathspec */
397 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
398 if ((namelen
< matchlen
) &&
399 (match
[namelen
-offset
] == '/') &&
400 !ps_strncmp(item
, match
, name
, namelen
))
401 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
403 /* name doesn't match up to the first wild character */
404 if (item
->nowildcard_len
< item
->len
&&
405 ps_strncmp(item
, match
, name
,
406 item
->nowildcard_len
- prefix
))
410 * name has no wildcard, and it didn't match as a leading
411 * pathspec so return.
413 if (item
->nowildcard_len
== item
->len
)
417 * Here is where we would perform a wildmatch to check if
418 * "name" can be matched as a directory (or a prefix) against
419 * the pathspec. Since wildmatch doesn't have this capability
420 * at the present we have to punt and say that it is a match,
421 * potentially returning a false positive
422 * The submodules themselves will be able to perform more
423 * accurate matching to determine if the pathspec matches.
425 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
432 * do_match_pathspec() is meant to ONLY be called by
433 * match_pathspec_with_flags(); calling it directly risks pathspecs
434 * like ':!unwanted_path' being ignored.
436 * Given a name and a list of pathspecs, returns the nature of the
437 * closest (i.e. most specific) match of the name to any of the
440 * The caller typically calls this multiple times with the same
441 * pathspec and seen[] array but with different name/namelen
442 * (e.g. entries from the index) and is interested in seeing if and
443 * how each pathspec matches all the names it calls this function
444 * with. A mark is left in the seen[] array for each pathspec element
445 * indicating the closest type of match that element achieved, so if
446 * seen[n] remains zero after multiple invocations, that means the nth
447 * pathspec did not match any names, which could indicate that the
448 * user mistyped the nth pathspec.
450 static int do_match_pathspec(struct index_state
*istate
,
451 const struct pathspec
*ps
,
452 const char *name
, int namelen
,
453 int prefix
, char *seen
,
456 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
468 if (!ps
->recursive
||
469 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
471 return MATCHED_RECURSIVELY
;
473 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
474 return MATCHED_EXACTLY
;
482 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
485 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
486 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
489 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
492 * Make exclude patterns optional and never report
493 * "pathspec ':(exclude)foo' matches no files"
495 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
496 seen
[i
] = MATCHED_FNMATCH
;
497 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
500 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
501 ps
->max_depth
!= -1 &&
502 how
&& how
!= MATCHED_FNMATCH
) {
503 int len
= ps
->items
[i
].len
;
504 if (name
[len
] == '/')
506 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
507 how
= MATCHED_EXACTLY
;
514 if (seen
&& seen
[i
] < how
)
521 static int match_pathspec_with_flags(struct index_state
*istate
,
522 const struct pathspec
*ps
,
523 const char *name
, int namelen
,
524 int prefix
, char *seen
, unsigned flags
)
526 int positive
, negative
;
527 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
528 prefix
, seen
, flags
);
529 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
531 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
533 flags
| DO_MATCH_EXCLUDE
);
534 return negative
? 0 : positive
;
537 int match_pathspec(struct index_state
*istate
,
538 const struct pathspec
*ps
,
539 const char *name
, int namelen
,
540 int prefix
, char *seen
, int is_dir
)
542 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
543 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
544 prefix
, seen
, flags
);
548 * Check if a submodule is a superset of the pathspec
550 int submodule_path_match(struct index_state
*istate
,
551 const struct pathspec
*ps
,
552 const char *submodule_name
,
555 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
556 strlen(submodule_name
),
559 DO_MATCH_LEADING_PATHSPEC
);
563 int report_path_error(const char *ps_matched
,
564 const struct pathspec
*pathspec
)
567 * Make sure all pathspec matched; otherwise it is an error.
570 for (num
= 0; num
< pathspec
->nr
; num
++) {
571 int other
, found_dup
;
576 * The caller might have fed identical pathspec
577 * twice. Do not barf on such a mistake.
578 * FIXME: parse_pathspec should have eliminated
579 * duplicate pathspec.
581 for (found_dup
= other
= 0;
582 !found_dup
&& other
< pathspec
->nr
;
584 if (other
== num
|| !ps_matched
[other
])
586 if (!strcmp(pathspec
->items
[other
].original
,
587 pathspec
->items
[num
].original
))
589 * Ok, we have a match already.
596 error(_("pathspec '%s' did not match any file(s) known to git"),
597 pathspec
->items
[num
].original
);
604 * Return the length of the "simple" part of a path match limiter.
606 int simple_length(const char *match
)
611 unsigned char c
= *match
++;
613 if (c
== '\0' || is_glob_special(c
))
618 int no_wildcard(const char *string
)
620 return string
[simple_length(string
)] == '\0';
623 void parse_path_pattern(const char **pattern
,
628 const char *p
= *pattern
;
633 *flags
|= PATTERN_FLAG_NEGATIVE
;
637 if (len
&& p
[len
- 1] == '/') {
639 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
641 for (i
= 0; i
< len
; i
++) {
646 *flags
|= PATTERN_FLAG_NODIR
;
647 *nowildcardlen
= simple_length(p
);
649 * we should have excluded the trailing slash from 'p' too,
650 * but that's one more allocation. Instead just make sure
651 * nowildcardlen does not exceed real patternlen
653 if (*nowildcardlen
> len
)
654 *nowildcardlen
= len
;
655 if (*p
== '*' && no_wildcard(p
+ 1))
656 *flags
|= PATTERN_FLAG_ENDSWITH
;
661 int pl_hashmap_cmp(const void *cmp_data UNUSED
,
662 const struct hashmap_entry
*a
,
663 const struct hashmap_entry
*b
,
664 const void *key UNUSED
)
666 const struct pattern_entry
*ee1
=
667 container_of(a
, struct pattern_entry
, ent
);
668 const struct pattern_entry
*ee2
=
669 container_of(b
, struct pattern_entry
, ent
);
671 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
675 return fspathncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
678 static char *dup_and_filter_pattern(const char *pattern
)
682 char *result
= xstrdup(pattern
);
688 /* skip escape characters (once) */
708 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
710 struct pattern_entry
*translated
;
713 const char *prev
, *cur
, *next
;
715 if (!pl
->use_cone_patterns
)
718 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
719 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
720 !strcmp(given
->pattern
, "/*")) {
725 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
730 if (given
->patternlen
< 2 ||
731 *given
->pattern
!= '/' ||
732 strstr(given
->pattern
, "**")) {
733 /* Not a cone pattern. */
734 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
738 if (!(given
->flags
& PATTERN_FLAG_MUSTBEDIR
) &&
739 strcmp(given
->pattern
, "/*")) {
740 /* Not a cone pattern. */
741 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
745 prev
= given
->pattern
;
746 cur
= given
->pattern
+ 1;
747 next
= given
->pattern
+ 2;
750 /* Watch for glob characters '*', '\', '[', '?' */
751 if (!is_glob_special(*cur
))
754 /* But only if *prev != '\\' */
758 /* But allow the initial '\' */
760 is_glob_special(*next
))
763 /* But a trailing '/' then '*' is fine */
769 /* Not a cone pattern. */
770 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
779 if (given
->patternlen
> 2 &&
780 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
781 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
782 /* Not a cone pattern. */
783 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
787 truncated
= dup_and_filter_pattern(given
->pattern
);
789 translated
= xmalloc(sizeof(struct pattern_entry
));
790 translated
->pattern
= truncated
;
791 translated
->patternlen
= given
->patternlen
- 2;
792 hashmap_entry_init(&translated
->ent
,
793 fspathhash(translated
->pattern
));
795 if (!hashmap_get_entry(&pl
->recursive_hashmap
,
796 translated
, ent
, NULL
)) {
797 /* We did not see the "parent" included */
798 warning(_("unrecognized negative pattern: '%s'"),
805 hashmap_add(&pl
->parent_hashmap
, &translated
->ent
);
806 hashmap_remove(&pl
->recursive_hashmap
, &translated
->ent
, &data
);
811 if (given
->flags
& PATTERN_FLAG_NEGATIVE
) {
812 warning(_("unrecognized negative pattern: '%s'"),
817 translated
= xmalloc(sizeof(struct pattern_entry
));
819 translated
->pattern
= dup_and_filter_pattern(given
->pattern
);
820 translated
->patternlen
= given
->patternlen
;
821 hashmap_entry_init(&translated
->ent
,
822 fspathhash(translated
->pattern
));
824 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
826 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
827 /* we already included this at the parent level */
828 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
836 warning(_("disabling cone pattern matching"));
837 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
838 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
839 pl
->use_cone_patterns
= 0;
842 static int hashmap_contains_path(struct hashmap
*map
,
843 struct strbuf
*pattern
)
845 struct pattern_entry p
;
847 /* Check straight mapping */
848 p
.pattern
= pattern
->buf
;
849 p
.patternlen
= pattern
->len
;
850 hashmap_entry_init(&p
.ent
, fspathhash(p
.pattern
));
851 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
854 int hashmap_contains_parent(struct hashmap
*map
,
856 struct strbuf
*buffer
)
860 strbuf_setlen(buffer
, 0);
863 strbuf_addch(buffer
, '/');
865 strbuf_addstr(buffer
, path
);
867 slash_pos
= strrchr(buffer
->buf
, '/');
869 while (slash_pos
> buffer
->buf
) {
870 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
872 if (hashmap_contains_path(map
, buffer
))
875 slash_pos
= strrchr(buffer
->buf
, '/');
881 void add_pattern(const char *string
, const char *base
,
882 int baselen
, struct pattern_list
*pl
, int srcpos
)
884 struct path_pattern
*pattern
;
889 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
890 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
891 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
893 pattern
= xmalloc(sizeof(*pattern
));
894 pattern
->pattern
= string
;
896 pattern
->patternlen
= patternlen
;
897 pattern
->nowildcardlen
= nowildcardlen
;
898 pattern
->base
= base
;
899 pattern
->baselen
= baselen
;
900 pattern
->flags
= flags
;
901 pattern
->srcpos
= srcpos
;
902 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
903 pl
->patterns
[pl
->nr
++] = pattern
;
906 add_pattern_to_hashsets(pl
, pattern
);
909 static int read_skip_worktree_file_from_index(struct index_state
*istate
,
911 size_t *size_out
, char **data_out
,
912 struct oid_stat
*oid_stat
)
917 pos
= index_name_pos(istate
, path
, len
);
920 if (!ce_skip_worktree(istate
->cache
[pos
]))
923 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
927 * Frees memory within pl which was allocated for exclude patterns and
928 * the file buffer. Does not free pl itself.
930 void clear_pattern_list(struct pattern_list
*pl
)
934 for (i
= 0; i
< pl
->nr
; i
++)
935 free(pl
->patterns
[i
]);
938 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
939 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
941 memset(pl
, 0, sizeof(*pl
));
944 static void trim_trailing_spaces(char *buf
)
946 char *p
, *last_space
= NULL
;
948 for (p
= buf
; *p
; p
++)
968 * Given a subdirectory name and "dir" of the current directory,
969 * search the subdir in "dir" and return it, or create a new one if it
970 * does not exist in "dir".
972 * If "name" has the trailing slash, it'll be excluded in the search.
974 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
975 struct untracked_cache_dir
*dir
,
976 const char *name
, int len
)
979 struct untracked_cache_dir
*d
;
982 if (len
&& name
[len
- 1] == '/')
986 while (last
> first
) {
987 int cmp
, next
= first
+ ((last
- first
) >> 1);
989 cmp
= strncmp(name
, d
->name
, len
);
990 if (!cmp
&& strlen(d
->name
) > len
)
1002 FLEX_ALLOC_MEM(d
, name
, name
, len
);
1004 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
1005 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
1006 dir
->dirs_nr
- first
);
1008 dir
->dirs
[first
] = d
;
1012 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
1016 dir
->untracked_nr
= 0;
1017 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1018 do_invalidate_gitignore(dir
->dirs
[i
]);
1021 static void invalidate_gitignore(struct untracked_cache
*uc
,
1022 struct untracked_cache_dir
*dir
)
1024 uc
->gitignore_invalidated
++;
1025 do_invalidate_gitignore(dir
);
1028 static void invalidate_directory(struct untracked_cache
*uc
,
1029 struct untracked_cache_dir
*dir
)
1034 * Invalidation increment here is just roughly correct. If
1035 * untracked_nr or any of dirs[].recurse is non-zero, we
1036 * should increment dir_invalidated too. But that's more
1040 uc
->dir_invalidated
++;
1043 dir
->untracked_nr
= 0;
1044 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1045 dir
->dirs
[i
]->recurse
= 0;
1048 static int add_patterns_from_buffer(char *buf
, size_t size
,
1049 const char *base
, int baselen
,
1050 struct pattern_list
*pl
);
1052 /* Flags for add_patterns() */
1053 #define PATTERN_NOFOLLOW (1<<0)
1056 * Given a file with name "fname", read it (either from disk, or from
1057 * an index if 'istate' is non-null), parse it and store the
1058 * exclude rules in "pl".
1060 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1061 * stat data from disk (only valid if add_patterns returns zero). If
1062 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1064 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1065 struct pattern_list
*pl
, struct index_state
*istate
,
1066 unsigned flags
, struct oid_stat
*oid_stat
)
1074 if (flags
& PATTERN_NOFOLLOW
)
1075 fd
= open_nofollow(fname
, O_RDONLY
);
1077 fd
= open(fname
, O_RDONLY
);
1079 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1081 warn_on_fopen_errors(fname
);
1086 r
= read_skip_worktree_file_from_index(istate
, fname
,
1092 size
= xsize_t(st
.st_size
);
1095 fill_stat_data(&oid_stat
->stat
, &st
);
1096 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1097 oid_stat
->valid
= 1;
1102 buf
= xmallocz(size
);
1103 if (read_in_full(fd
, buf
, size
) != size
) {
1112 if (oid_stat
->valid
&&
1113 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1114 ; /* no content change, oid_stat->oid still good */
1116 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1117 !ce_stage(istate
->cache
[pos
]) &&
1118 ce_uptodate(istate
->cache
[pos
]) &&
1119 !would_convert_to_git(istate
, fname
))
1120 oidcpy(&oid_stat
->oid
,
1121 &istate
->cache
[pos
]->oid
);
1123 hash_object_file(the_hash_algo
, buf
, size
,
1124 OBJ_BLOB
, &oid_stat
->oid
);
1125 fill_stat_data(&oid_stat
->stat
, &st
);
1126 oid_stat
->valid
= 1;
1130 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1134 static int add_patterns_from_buffer(char *buf
, size_t size
,
1135 const char *base
, int baselen
,
1136 struct pattern_list
*pl
)
1141 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1142 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1146 if (skip_utf8_bom(&buf
, size
))
1147 size
-= buf
- pl
->filebuf
;
1151 for (i
= 0; i
< size
; i
++) {
1152 if (buf
[i
] == '\n') {
1153 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1154 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1155 trim_trailing_spaces(entry
);
1156 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1159 entry
= buf
+ i
+ 1;
1165 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1166 int baselen
, struct pattern_list
*pl
,
1167 struct index_state
*istate
,
1170 return add_patterns(fname
, base
, baselen
, pl
, istate
, flags
, NULL
);
1173 int add_patterns_from_blob_to_list(
1174 struct object_id
*oid
,
1175 const char *base
, int baselen
,
1176 struct pattern_list
*pl
)
1182 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1186 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1190 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1191 int group_type
, const char *src
)
1193 struct pattern_list
*pl
;
1194 struct exclude_list_group
*group
;
1196 group
= &dir
->internal
.exclude_list_group
[group_type
];
1197 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1198 pl
= &group
->pl
[group
->nr
++];
1199 memset(pl
, 0, sizeof(*pl
));
1205 * Used to set up core.excludesfile and .git/info/exclude lists.
1207 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1208 struct oid_stat
*oid_stat
)
1210 struct pattern_list
*pl
;
1212 * catch setup_standard_excludes() that's called before
1213 * dir->untracked is assigned. That function behaves
1214 * differently when dir->untracked is non-NULL.
1216 if (!dir
->untracked
)
1217 dir
->internal
.unmanaged_exclude_files
++;
1218 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1219 if (add_patterns(fname
, "", 0, pl
, NULL
, 0, oid_stat
) < 0)
1220 die(_("cannot use %s as an exclude file"), fname
);
1223 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1225 dir
->internal
.unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1226 add_patterns_from_file_1(dir
, fname
, NULL
);
1229 int match_basename(const char *basename
, int basenamelen
,
1230 const char *pattern
, int prefix
, int patternlen
,
1233 if (prefix
== patternlen
) {
1234 if (patternlen
== basenamelen
&&
1235 !fspathncmp(pattern
, basename
, basenamelen
))
1237 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1238 /* "*literal" matching against "fooliteral" */
1239 if (patternlen
- 1 <= basenamelen
&&
1240 !fspathncmp(pattern
+ 1,
1241 basename
+ basenamelen
- (patternlen
- 1),
1245 if (fnmatch_icase_mem(pattern
, patternlen
,
1246 basename
, basenamelen
,
1253 int match_pathname(const char *pathname
, int pathlen
,
1254 const char *base
, int baselen
,
1255 const char *pattern
, int prefix
, int patternlen
)
1261 * match with FNM_PATHNAME; the pattern has base implicitly
1264 if (*pattern
== '/') {
1271 * baselen does not count the trailing slash. base[] may or
1272 * may not end with a trailing slash though.
1274 if (pathlen
< baselen
+ 1 ||
1275 (baselen
&& pathname
[baselen
] != '/') ||
1276 fspathncmp(pathname
, base
, baselen
))
1279 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1280 name
= pathname
+ pathlen
- namelen
;
1284 * if the non-wildcard part is longer than the
1285 * remaining pathname, surely it cannot match.
1287 if (prefix
> namelen
)
1290 if (fspathncmp(pattern
, name
, prefix
))
1293 patternlen
-= prefix
;
1298 * If the whole pattern did not have a wildcard,
1299 * then our prefix match is all we need; we
1300 * do not need to call fnmatch at all.
1302 if (!patternlen
&& !namelen
)
1306 return fnmatch_icase_mem(pattern
, patternlen
,
1312 * Scan the given exclude list in reverse to see whether pathname
1313 * should be ignored. The first match (i.e. the last on the list), if
1314 * any, determines the fate. Returns the exclude_list element which
1315 * matched, or NULL for undecided.
1317 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1319 const char *basename
,
1321 struct pattern_list
*pl
,
1322 struct index_state
*istate
)
1324 struct path_pattern
*res
= NULL
; /* undecided */
1328 return NULL
; /* undefined */
1330 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1331 struct path_pattern
*pattern
= pl
->patterns
[i
];
1332 const char *exclude
= pattern
->pattern
;
1333 int prefix
= pattern
->nowildcardlen
;
1335 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1336 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1337 if (*dtype
!= DT_DIR
)
1341 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1342 if (match_basename(basename
,
1343 pathlen
- (basename
- pathname
),
1344 exclude
, prefix
, pattern
->patternlen
,
1352 assert(pattern
->baselen
== 0 ||
1353 pattern
->base
[pattern
->baselen
- 1] == '/');
1354 if (match_pathname(pathname
, pathlen
,
1356 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1357 exclude
, prefix
, pattern
->patternlen
)) {
1366 * Scan the list of patterns to determine if the ordered list
1367 * of patterns matches on 'pathname'.
1369 * Return 1 for a match, 0 for not matched and -1 for undecided.
1371 enum pattern_match_result
path_matches_pattern_list(
1372 const char *pathname
, int pathlen
,
1373 const char *basename
, int *dtype
,
1374 struct pattern_list
*pl
,
1375 struct index_state
*istate
)
1377 struct path_pattern
*pattern
;
1378 struct strbuf parent_pathname
= STRBUF_INIT
;
1379 int result
= NOT_MATCHED
;
1382 if (!pl
->use_cone_patterns
) {
1383 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1386 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1398 strbuf_addch(&parent_pathname
, '/');
1399 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1402 * Directory entries are matched if and only if a file
1403 * contained immediately within them is matched. For the
1404 * case of a directory entry, modify the path to create
1405 * a fake filename within this directory, allowing us to
1406 * use the file-base matching logic in an equivalent way.
1408 if (parent_pathname
.len
> 0 &&
1409 parent_pathname
.buf
[parent_pathname
.len
- 1] == '/') {
1410 slash_pos
= parent_pathname
.len
- 1;
1411 strbuf_add(&parent_pathname
, "-", 1);
1413 const char *slash_ptr
= strrchr(parent_pathname
.buf
, '/');
1414 slash_pos
= slash_ptr
? slash_ptr
- parent_pathname
.buf
: 0;
1417 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1418 &parent_pathname
)) {
1419 result
= MATCHED_RECURSIVE
;
1424 /* include every file in root */
1429 strbuf_setlen(&parent_pathname
, slash_pos
);
1431 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1436 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1439 result
= MATCHED_RECURSIVE
;
1442 strbuf_release(&parent_pathname
);
1446 int init_sparse_checkout_patterns(struct index_state
*istate
)
1448 if (!core_apply_sparse_checkout
)
1450 if (istate
->sparse_checkout_patterns
)
1453 CALLOC_ARRAY(istate
->sparse_checkout_patterns
, 1);
1455 if (get_sparse_checkout_patterns(istate
->sparse_checkout_patterns
) < 0) {
1456 FREE_AND_NULL(istate
->sparse_checkout_patterns
);
1463 static int path_in_sparse_checkout_1(const char *path
,
1464 struct index_state
*istate
,
1465 int require_cone_mode
)
1468 enum pattern_match_result match
= UNDECIDED
;
1469 const char *end
, *slash
;
1472 * We default to accepting a path if the path is empty, there are no
1473 * patterns, or the patterns are of the wrong type.
1476 init_sparse_checkout_patterns(istate
) ||
1477 (require_cone_mode
&&
1478 !istate
->sparse_checkout_patterns
->use_cone_patterns
))
1482 * If UNDECIDED, use the match from the parent dir (recursively), or
1483 * fall back to NOT_MATCHED at the topmost level. Note that cone mode
1484 * never returns UNDECIDED, so we will execute only one iteration in
1487 for (end
= path
+ strlen(path
);
1488 end
> path
&& match
== UNDECIDED
;
1491 for (slash
= end
- 1; slash
> path
&& *slash
!= '/'; slash
--)
1494 match
= path_matches_pattern_list(path
, end
- path
,
1495 slash
> path
? slash
+ 1 : path
, &dtype
,
1496 istate
->sparse_checkout_patterns
, istate
);
1498 /* We are going to match the parent dir now */
1504 int path_in_sparse_checkout(const char *path
,
1505 struct index_state
*istate
)
1507 return path_in_sparse_checkout_1(path
, istate
, 0);
1510 int path_in_cone_mode_sparse_checkout(const char *path
,
1511 struct index_state
*istate
)
1513 return path_in_sparse_checkout_1(path
, istate
, 1);
1516 static struct path_pattern
*last_matching_pattern_from_lists(
1517 struct dir_struct
*dir
, struct index_state
*istate
,
1518 const char *pathname
, int pathlen
,
1519 const char *basename
, int *dtype_p
)
1522 struct exclude_list_group
*group
;
1523 struct path_pattern
*pattern
;
1524 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1525 group
= &dir
->internal
.exclude_list_group
[i
];
1526 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1527 pattern
= last_matching_pattern_from_list(
1528 pathname
, pathlen
, basename
, dtype_p
,
1529 &group
->pl
[j
], istate
);
1538 * Loads the per-directory exclude list for the substring of base
1539 * which has a char length of baselen.
1541 static void prep_exclude(struct dir_struct
*dir
,
1542 struct index_state
*istate
,
1543 const char *base
, int baselen
)
1545 struct exclude_list_group
*group
;
1546 struct pattern_list
*pl
;
1547 struct exclude_stack
*stk
= NULL
;
1548 struct untracked_cache_dir
*untracked
;
1551 group
= &dir
->internal
.exclude_list_group
[EXC_DIRS
];
1554 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1555 * which originate from directories not in the prefix of the
1556 * path being checked.
1558 while ((stk
= dir
->internal
.exclude_stack
) != NULL
) {
1559 if (stk
->baselen
<= baselen
&&
1560 !strncmp(dir
->internal
.basebuf
.buf
, base
, stk
->baselen
))
1562 pl
= &group
->pl
[dir
->internal
.exclude_stack
->exclude_ix
];
1563 dir
->internal
.exclude_stack
= stk
->prev
;
1564 dir
->internal
.pattern
= NULL
;
1565 free((char *)pl
->src
); /* see strbuf_detach() below */
1566 clear_pattern_list(pl
);
1571 /* Skip traversing into sub directories if the parent is excluded */
1572 if (dir
->internal
.pattern
)
1576 * Lazy initialization. All call sites currently just
1577 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1578 * them seems lots of work for little benefit.
1580 if (!dir
->internal
.basebuf
.buf
)
1581 strbuf_init(&dir
->internal
.basebuf
, PATH_MAX
);
1583 /* Read from the parent directories and push them down. */
1584 current
= stk
? stk
->baselen
: -1;
1585 strbuf_setlen(&dir
->internal
.basebuf
, current
< 0 ? 0 : current
);
1587 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1591 while (current
< baselen
) {
1593 struct oid_stat oid_stat
;
1595 CALLOC_ARRAY(stk
, 1);
1600 cp
= strchr(base
+ current
+ 1, '/');
1602 die("oops in prep_exclude");
1605 lookup_untracked(dir
->untracked
,
1608 cp
- base
- current
);
1610 stk
->prev
= dir
->internal
.exclude_stack
;
1611 stk
->baselen
= cp
- base
;
1612 stk
->exclude_ix
= group
->nr
;
1613 stk
->ucd
= untracked
;
1614 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1615 strbuf_add(&dir
->internal
.basebuf
, base
+ current
, stk
->baselen
- current
);
1616 assert(stk
->baselen
== dir
->internal
.basebuf
.len
);
1618 /* Abort if the directory is excluded */
1621 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = 0;
1622 dir
->internal
.pattern
= last_matching_pattern_from_lists(dir
,
1624 dir
->internal
.basebuf
.buf
, stk
->baselen
- 1,
1625 dir
->internal
.basebuf
.buf
+ current
, &dt
);
1626 dir
->internal
.basebuf
.buf
[stk
->baselen
- 1] = '/';
1627 if (dir
->internal
.pattern
&&
1628 dir
->internal
.pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1629 dir
->internal
.pattern
= NULL
;
1630 if (dir
->internal
.pattern
) {
1631 dir
->internal
.exclude_stack
= stk
;
1636 /* Try to read per-directory file */
1637 oidclr(&oid_stat
.oid
);
1639 if (dir
->exclude_per_dir
&&
1641 * If we know that no files have been added in
1642 * this directory (i.e. valid_cached_dir() has
1643 * been executed and set untracked->valid) ..
1645 (!untracked
|| !untracked
->valid
||
1647 * .. and .gitignore does not exist before
1648 * (i.e. null exclude_oid). Then we can skip
1649 * loading .gitignore, which would result in
1652 !is_null_oid(&untracked
->exclude_oid
))) {
1654 * dir->internal.basebuf gets reused by the traversal,
1655 * but we need fname to remain unchanged to ensure the
1656 * src member of each struct path_pattern correctly
1657 * back-references its source file. Other invocations
1658 * of add_pattern_list provide stable strings, so we
1659 * strbuf_detach() and free() here in the caller.
1661 struct strbuf sb
= STRBUF_INIT
;
1662 strbuf_addbuf(&sb
, &dir
->internal
.basebuf
);
1663 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1664 pl
->src
= strbuf_detach(&sb
, NULL
);
1665 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1667 untracked
? &oid_stat
: NULL
);
1670 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1671 * will first be called in valid_cached_dir() then maybe many
1672 * times more in last_matching_pattern(). When the cache is
1673 * used, last_matching_pattern() will not be called and
1674 * reading .gitignore content will be a waste.
1676 * So when it's called by valid_cached_dir() and we can get
1677 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1678 * modified on work tree), we could delay reading the
1679 * .gitignore content until we absolutely need it in
1680 * last_matching_pattern(). Be careful about ignore rule
1681 * order, though, if you do that.
1684 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1685 invalidate_gitignore(dir
->untracked
, untracked
);
1686 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1688 dir
->internal
.exclude_stack
= stk
;
1689 current
= stk
->baselen
;
1691 strbuf_setlen(&dir
->internal
.basebuf
, baselen
);
1695 * Loads the exclude lists for the directory containing pathname, then
1696 * scans all exclude lists to determine whether pathname is excluded.
1697 * Returns the exclude_list element which matched, or NULL for
1700 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1701 struct index_state
*istate
,
1702 const char *pathname
,
1705 int pathlen
= strlen(pathname
);
1706 const char *basename
= strrchr(pathname
, '/');
1707 basename
= (basename
) ? basename
+1 : pathname
;
1709 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1711 if (dir
->internal
.pattern
)
1712 return dir
->internal
.pattern
;
1714 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1719 * Loads the exclude lists for the directory containing pathname, then
1720 * scans all exclude lists to determine whether pathname is excluded.
1721 * Returns 1 if true, otherwise 0.
1723 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1724 const char *pathname
, int *dtype_p
)
1726 struct path_pattern
*pattern
=
1727 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1729 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1733 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1735 struct dir_entry
*ent
;
1737 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1742 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1743 struct index_state
*istate
,
1744 const char *pathname
, int len
)
1746 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1749 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->internal
.alloc
);
1750 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1753 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1754 struct index_state
*istate
,
1755 const char *pathname
, int len
)
1757 if (!index_name_is_other(istate
, pathname
, len
))
1760 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->internal
.ignored_alloc
);
1761 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1765 index_nonexistent
= 0,
1771 * Do not use the alphabetically sorted index to look up
1772 * the directory name; instead, use the case insensitive
1775 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1776 const char *dirname
, int len
)
1778 struct cache_entry
*ce
;
1780 if (index_dir_exists(istate
, dirname
, len
))
1781 return index_directory
;
1783 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1784 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1785 return index_gitdir
;
1787 return index_nonexistent
;
1791 * The index sorts alphabetically by entry name, which
1792 * means that a gitlink sorts as '\0' at the end, while
1793 * a directory (which is defined not as an entry, but as
1794 * the files it contains) will sort with the '/' at the
1797 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1798 const char *dirname
, int len
)
1803 return directory_exists_in_index_icase(istate
, dirname
, len
);
1805 pos
= index_name_pos(istate
, dirname
, len
);
1808 while (pos
< istate
->cache_nr
) {
1809 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1810 unsigned char endchar
;
1812 if (strncmp(ce
->name
, dirname
, len
))
1814 endchar
= ce
->name
[len
];
1818 return index_directory
;
1819 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1820 return index_gitdir
;
1822 return index_nonexistent
;
1826 * When we find a directory when traversing the filesystem, we
1827 * have three distinct cases:
1830 * - see it as a directory
1833 * and which one we choose depends on a combination of existing
1834 * git index contents and the flags passed into the directory
1835 * traversal routine.
1837 * Case 1: If we *already* have entries in the index under that
1838 * directory name, we always recurse into the directory to see
1841 * Case 2: If we *already* have that directory name as a gitlink,
1842 * we always continue to see it as a gitlink, regardless of whether
1843 * there is an actual git directory there or not (it might not
1844 * be checked out as a subproject!)
1846 * Case 3: if we didn't have it in the index previously, we
1847 * have a few sub-cases:
1849 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1850 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1851 * also true, in which case we need to check if it contains any
1852 * untracked and / or ignored files.
1853 * (b) if it looks like a git directory and we don't have the
1854 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1855 * show it as a directory.
1856 * (c) otherwise, we recurse into it.
1858 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1859 struct index_state
*istate
,
1860 struct untracked_cache_dir
*untracked
,
1861 const char *dirname
, int len
, int baselen
, int excluded
,
1862 const struct pathspec
*pathspec
)
1865 * WARNING: From this function, you can return path_recurse or you
1866 * can call read_directory_recursive() (or neither), but
1867 * you CAN'T DO BOTH.
1869 enum path_treatment state
;
1870 int matches_how
= 0;
1871 int check_only
, stop_early
;
1872 int old_ignored_nr
, old_untracked_nr
;
1873 /* The "len-1" is to strip the final '/' */
1874 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1876 if (status
== index_directory
)
1877 return path_recurse
;
1878 if (status
== index_gitdir
)
1880 if (status
!= index_nonexistent
)
1881 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1884 * We don't want to descend into paths that don't match the necessary
1885 * patterns. Clearly, if we don't have a pathspec, then we can't check
1886 * for matching patterns. Also, if (excluded) then we know we matched
1887 * the exclusion patterns so as an optimization we can skip checking
1888 * for matching patterns.
1890 if (pathspec
&& !excluded
) {
1891 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1895 DO_MATCH_LEADING_PATHSPEC
);
1901 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1902 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1904 * Determine if `dirname` is a nested repo by confirming that:
1905 * 1) we are in a nonbare repository, and
1906 * 2) `dirname` is not an immediate parent of `the_repository->gitdir`,
1907 * which could occur if the git_dir or worktree location was
1908 * manually configured by the user; see t2205 testcases 1-3 for
1909 * examples where this matters
1912 struct strbuf sb
= STRBUF_INIT
;
1913 strbuf_addstr(&sb
, dirname
);
1914 nested_repo
= is_nonbare_repository_dir(&sb
);
1917 char *real_dirname
, *real_gitdir
;
1918 strbuf_addstr(&sb
, ".git");
1919 real_dirname
= real_pathdup(sb
.buf
, 1);
1920 real_gitdir
= real_pathdup(the_repository
->gitdir
, 1);
1922 nested_repo
= !!strcmp(real_dirname
, real_gitdir
);
1926 strbuf_release(&sb
);
1929 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1930 (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
))
1932 return excluded
? path_excluded
: path_untracked
;
1936 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1938 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1939 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1942 * This is an excluded directory and we are
1943 * showing ignored paths that match an exclude
1944 * pattern. (e.g. show directory as ignored
1945 * only if it matches an exclude pattern).
1946 * This path will either be 'path_excluded`
1947 * (if we are showing empty directories or if
1948 * the directory is not empty), or will be
1949 * 'path_none' (empty directory, and we are
1950 * not showing empty directories).
1952 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1953 return path_excluded
;
1955 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1956 untracked
, 1, 1, pathspec
) == path_excluded
)
1957 return path_excluded
;
1961 return path_recurse
;
1964 assert(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
);
1967 * If we have a pathspec which could match something _below_ this
1968 * directory (e.g. when checking 'subdir/' having a pathspec like
1969 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1972 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1973 return path_recurse
;
1975 /* Special cases for where this directory is excluded/ignored */
1978 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1979 * hiding empty directories, there is no need to
1980 * recurse into an ignored directory.
1982 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1983 return path_excluded
;
1986 * Even if we are hiding empty directories, we can still avoid
1987 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
1988 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1990 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1991 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
1992 return path_excluded
;
1996 * Other than the path_recurse case above, we only need to
1997 * recurse into untracked directories if any of the following
1999 * - DIR_SHOW_IGNORED (because then we need to determine if
2000 * there are ignored entries below)
2001 * - DIR_SHOW_IGNORED_TOO (same as above)
2002 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
2003 * the directory is empty)
2006 !(dir
->flags
& (DIR_SHOW_IGNORED
|
2007 DIR_SHOW_IGNORED_TOO
|
2008 DIR_HIDE_EMPTY_DIRECTORIES
))) {
2009 return path_untracked
;
2013 * Even if we don't want to know all the paths under an untracked or
2014 * ignored directory, we may still need to go into the directory to
2015 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
2016 * an empty directory should be path_none instead of path_excluded or
2019 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
2020 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
2023 * However, there's another optimization possible as a subset of
2024 * check_only, based on the cases we have to consider:
2025 * A) Directory matches no exclude patterns:
2026 * * Directory is empty => path_none
2027 * * Directory has an untracked file under it => path_untracked
2028 * * Directory has only ignored files under it => path_excluded
2029 * B) Directory matches an exclude pattern:
2030 * * Directory is empty => path_none
2031 * * Directory has an untracked file under it => path_excluded
2032 * * Directory has only ignored files under it => path_excluded
2033 * In case A, we can exit as soon as we've found an untracked
2034 * file but otherwise have to walk all files. In case B, though,
2035 * we can stop at the first file we find under the directory.
2037 stop_early
= check_only
&& excluded
;
2040 * If /every/ file within an untracked directory is ignored, then
2041 * we want to treat the directory as ignored (for e.g. status
2042 * --porcelain), without listing the individual ignored files
2043 * underneath. To do so, we'll save the current ignored_nr, and
2044 * pop all the ones added after it if it turns out the entire
2045 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
2046 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
2047 * untracked paths so will need to pop all those off the last
2048 * after we traverse.
2050 old_ignored_nr
= dir
->ignored_nr
;
2051 old_untracked_nr
= dir
->nr
;
2053 /* Actually recurse into dirname now, we'll fixup the state later. */
2054 untracked
= lookup_untracked(dir
->untracked
, untracked
,
2055 dirname
+ baselen
, len
- baselen
);
2056 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
2057 check_only
, stop_early
, pathspec
);
2059 /* There are a variety of reasons we may need to fixup the state... */
2060 if (state
== path_excluded
) {
2061 /* state == path_excluded implies all paths under
2062 * dirname were ignored...
2064 * if running e.g. `git status --porcelain --ignored=matching`,
2065 * then we want to see the subpaths that are ignored.
2067 * if running e.g. just `git status --porcelain`, then
2068 * we just want the directory itself to be listed as ignored
2069 * and not the individual paths underneath.
2071 int want_ignored_subpaths
=
2072 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2073 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
2075 if (want_ignored_subpaths
) {
2077 * with --ignored=matching, we want the subpaths
2078 * INSTEAD of the directory itself.
2083 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
2084 FREE_AND_NULL(dir
->ignored
[i
]);
2085 dir
->ignored_nr
= old_ignored_nr
;
2090 * We may need to ignore some of the untracked paths we found while
2091 * traversing subdirectories.
2093 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2094 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
2096 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
2097 FREE_AND_NULL(dir
->entries
[i
]);
2098 dir
->nr
= old_untracked_nr
;
2102 * If there is nothing under the current directory and we are not
2103 * hiding empty directories, then we need to report on the
2104 * untracked or ignored status of the directory itself.
2106 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2107 state
= excluded
? path_excluded
: path_untracked
;
2113 * This is an inexact early pruning of any recursive directory
2114 * reading - if the path cannot possibly be in the pathspec,
2115 * return true, and we'll skip it early.
2117 static int simplify_away(const char *path
, int pathlen
,
2118 const struct pathspec
*pathspec
)
2122 if (!pathspec
|| !pathspec
->nr
)
2125 GUARD_PATHSPEC(pathspec
,
2134 for (i
= 0; i
< pathspec
->nr
; i
++) {
2135 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2136 int len
= item
->nowildcard_len
;
2140 if (!ps_strncmp(item
, item
->match
, path
, len
))
2148 * This function tells us whether an excluded path matches a
2149 * list of "interesting" pathspecs. That is, whether a path matched
2150 * by any of the pathspecs could possibly be ignored by excluding
2151 * the specified path. This can happen if:
2153 * 1. the path is mentioned explicitly in the pathspec
2155 * 2. the path is a directory prefix of some element in the
2158 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2159 const struct pathspec
*pathspec
)
2163 if (!pathspec
|| !pathspec
->nr
)
2166 GUARD_PATHSPEC(pathspec
,
2174 for (i
= 0; i
< pathspec
->nr
; i
++) {
2175 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2176 int len
= item
->nowildcard_len
;
2178 if (len
== pathlen
&&
2179 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2181 if (len
> pathlen
&&
2182 item
->match
[pathlen
] == '/' &&
2183 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2189 static int get_index_dtype(struct index_state
*istate
,
2190 const char *path
, int len
)
2193 const struct cache_entry
*ce
;
2195 ce
= index_file_exists(istate
, path
, len
, 0);
2197 if (!ce_uptodate(ce
))
2199 if (S_ISGITLINK(ce
->ce_mode
))
2202 * Nobody actually cares about the
2203 * difference between DT_LNK and DT_REG
2208 /* Try to look it up as a directory */
2209 pos
= index_name_pos(istate
, path
, len
);
2213 while (pos
< istate
->cache_nr
) {
2214 ce
= istate
->cache
[pos
++];
2215 if (strncmp(ce
->name
, path
, len
))
2217 if (ce
->name
[len
] > '/')
2219 if (ce
->name
[len
] < '/')
2221 if (!ce_uptodate(ce
))
2222 break; /* continue? */
2228 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2229 const char *path
, int len
)
2233 if (dtype
!= DT_UNKNOWN
)
2235 dtype
= get_index_dtype(istate
, path
, len
);
2236 if (dtype
!= DT_UNKNOWN
)
2238 if (lstat(path
, &st
))
2240 if (S_ISREG(st
.st_mode
))
2242 if (S_ISDIR(st
.st_mode
))
2244 if (S_ISLNK(st
.st_mode
))
2249 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2250 struct cached_dir
*cdir
,
2251 struct index_state
*istate
,
2252 struct strbuf
*path
,
2254 const struct pathspec
*pathspec
)
2257 * WARNING: From this function, you can return path_recurse or you
2258 * can call read_directory_recursive() (or neither), but
2259 * you CAN'T DO BOTH.
2261 strbuf_setlen(path
, baselen
);
2263 strbuf_addstr(path
, cdir
->file
);
2264 return path_untracked
;
2266 strbuf_addstr(path
, cdir
->ucd
->name
);
2267 /* treat_one_path() does this before it calls treat_directory() */
2268 strbuf_complete(path
, '/');
2269 if (cdir
->ucd
->check_only
)
2271 * check_only is set as a result of treat_directory() getting
2272 * to its bottom. Verify again the same set of directories
2273 * with check_only set.
2275 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2276 cdir
->ucd
, 1, 0, pathspec
);
2278 * We get path_recurse in the first run when
2279 * directory_exists_in_index() returns index_nonexistent. We
2280 * are sure that new changes in the index does not impact the
2281 * outcome. Return now.
2283 return path_recurse
;
2286 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2287 struct untracked_cache_dir
*untracked
,
2288 struct cached_dir
*cdir
,
2289 struct index_state
*istate
,
2290 struct strbuf
*path
,
2292 const struct pathspec
*pathspec
)
2294 int has_path_in_index
, dtype
, excluded
;
2297 return treat_path_fast(dir
, cdir
, istate
, path
,
2299 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2301 strbuf_setlen(path
, baselen
);
2302 strbuf_addstr(path
, cdir
->d_name
);
2303 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2306 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2308 /* Always exclude indexed files */
2309 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2311 if (dtype
!= DT_DIR
&& has_path_in_index
)
2315 * When we are looking at a directory P in the working tree,
2316 * there are three cases:
2318 * (1) P exists in the index. Everything inside the directory P in
2319 * the working tree needs to go when P is checked out from the
2322 * (2) P does not exist in the index, but there is P/Q in the index.
2323 * We know P will stay a directory when we check out the contents
2324 * of the index, but we do not know yet if there is a directory
2325 * P/Q in the working tree to be killed, so we need to recurse.
2327 * (3) P does not exist in the index, and there is no P/Q in the index
2328 * to require P to be a directory, either. Only in this case, we
2329 * know that everything inside P will not be killed without
2332 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2333 (dtype
== DT_DIR
) &&
2334 !has_path_in_index
&&
2335 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2338 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2341 * Excluded? If we don't explicitly want to show
2342 * ignored files, ignore it
2344 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2345 return path_excluded
;
2352 * WARNING: Do not ignore/amend the return value from
2353 * treat_directory(), and especially do not change it to return
2354 * path_recurse as that can cause exponential slowdown.
2355 * Instead, modify treat_directory() to return the right value.
2357 strbuf_addch(path
, '/');
2358 return treat_directory(dir
, istate
, untracked
,
2359 path
->buf
, path
->len
,
2360 baselen
, excluded
, pathspec
);
2364 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2365 0 /* prefix */, NULL
/* seen */,
2369 return path_excluded
;
2370 return path_untracked
;
2374 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2378 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2379 dir
->untracked_alloc
);
2380 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2383 static int valid_cached_dir(struct dir_struct
*dir
,
2384 struct untracked_cache_dir
*untracked
,
2385 struct index_state
*istate
,
2386 struct strbuf
*path
,
2395 * With fsmonitor, we can trust the untracked cache's valid field.
2397 refresh_fsmonitor(istate
);
2398 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2399 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2400 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2403 if (!untracked
->valid
||
2404 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2405 fill_stat_data(&untracked
->stat_data
, &st
);
2410 if (untracked
->check_only
!= !!check_only
)
2414 * prep_exclude will be called eventually on this directory,
2415 * but it's called much later in last_matching_pattern(). We
2416 * need it now to determine the validity of the cache for this
2417 * path. The next calls will be nearly no-op, the way
2418 * prep_exclude() is designed.
2420 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2421 strbuf_addch(path
, '/');
2422 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2423 strbuf_setlen(path
, path
->len
- 1);
2425 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2427 /* hopefully prep_exclude() haven't invalidated this entry... */
2428 return untracked
->valid
;
2431 static int open_cached_dir(struct cached_dir
*cdir
,
2432 struct dir_struct
*dir
,
2433 struct untracked_cache_dir
*untracked
,
2434 struct index_state
*istate
,
2435 struct strbuf
*path
,
2440 memset(cdir
, 0, sizeof(*cdir
));
2441 cdir
->untracked
= untracked
;
2442 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2444 c_path
= path
->len
? path
->buf
: ".";
2445 cdir
->fdir
= opendir(c_path
);
2447 warning_errno(_("could not open directory '%s'"), c_path
);
2448 if (dir
->untracked
) {
2449 invalidate_directory(dir
->untracked
, untracked
);
2450 dir
->untracked
->dir_opened
++;
2457 static int read_cached_dir(struct cached_dir
*cdir
)
2462 de
= readdir_skip_dot_and_dotdot(cdir
->fdir
);
2464 cdir
->d_name
= NULL
;
2465 cdir
->d_type
= DT_UNKNOWN
;
2468 cdir
->d_name
= de
->d_name
;
2469 cdir
->d_type
= DTYPE(de
);
2472 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2473 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2483 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2484 struct untracked_cache_dir
*d
= cdir
->untracked
;
2485 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2491 static void close_cached_dir(struct cached_dir
*cdir
)
2494 closedir(cdir
->fdir
);
2496 * We have gone through this directory and found no untracked
2497 * entries. Mark it valid.
2499 if (cdir
->untracked
) {
2500 cdir
->untracked
->valid
= 1;
2501 cdir
->untracked
->recurse
= 1;
2505 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2506 struct untracked_cache_dir
*untracked
,
2507 struct cached_dir
*cdir
,
2508 struct index_state
*istate
,
2509 struct strbuf
*path
,
2511 const struct pathspec
*pathspec
,
2512 enum path_treatment state
)
2514 /* add the path to the appropriate result list */
2517 if (dir
->flags
& DIR_SHOW_IGNORED
)
2518 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2519 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2520 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2521 exclude_matches_pathspec(path
->buf
, path
->len
,
2523 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2526 case path_untracked
:
2527 if (dir
->flags
& DIR_SHOW_IGNORED
)
2529 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2531 add_untracked(untracked
, path
->buf
+ baselen
);
2540 * Read a directory tree. We currently ignore anything but
2541 * directories, regular files and symlinks. That's because git
2542 * doesn't handle them at all yet. Maybe that will change some
2545 * Also, we ignore the name ".git" (even if it is not a directory).
2546 * That likely will not change.
2548 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2549 * to signal that a file was found. This is the least significant value that
2550 * indicates that a file was encountered that does not depend on the order of
2551 * whether an untracked or excluded path was encountered first.
2553 * Returns the most significant path_treatment value encountered in the scan.
2554 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2555 * significant path_treatment value that will be returned.
2558 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2559 struct index_state
*istate
, const char *base
, int baselen
,
2560 struct untracked_cache_dir
*untracked
, int check_only
,
2561 int stop_at_first_file
, const struct pathspec
*pathspec
)
2564 * WARNING: Do NOT recurse unless path_recurse is returned from
2565 * treat_path(). Recursing on any other return value
2566 * can result in exponential slowdown.
2568 struct cached_dir cdir
;
2569 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2570 struct strbuf path
= STRBUF_INIT
;
2572 strbuf_add(&path
, base
, baselen
);
2574 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2576 dir
->internal
.visited_directories
++;
2579 untracked
->check_only
= !!check_only
;
2581 while (!read_cached_dir(&cdir
)) {
2582 /* check how the file or directory should be treated */
2583 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2585 dir
->internal
.visited_paths
++;
2587 if (state
> dir_state
)
2590 /* recurse into subdir if instructed by treat_path */
2591 if (state
== path_recurse
) {
2592 struct untracked_cache_dir
*ud
;
2593 ud
= lookup_untracked(dir
->untracked
,
2596 path
.len
- baselen
);
2598 read_directory_recursive(dir
, istate
, path
.buf
,
2600 check_only
, stop_at_first_file
, pathspec
);
2601 if (subdir_state
> dir_state
)
2602 dir_state
= subdir_state
;
2605 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2606 0 /* prefix */, NULL
,
2607 0 /* do NOT special case dirs */))
2612 if (stop_at_first_file
) {
2614 * If stopping at first file, then
2615 * signal that a file was found by
2616 * returning `path_excluded`. This is
2617 * to return a consistent value
2618 * regardless of whether an ignored or
2619 * excluded file happened to be
2622 * In current usage, the
2623 * `stop_at_first_file` is passed when
2624 * an ancestor directory has matched
2625 * an exclude pattern, so any found
2626 * files will be excluded.
2628 if (dir_state
>= path_excluded
) {
2629 dir_state
= path_excluded
;
2634 /* abort early if maximum state has been reached */
2635 if (dir_state
== path_untracked
) {
2637 add_untracked(untracked
, path
.buf
+ baselen
);
2640 /* skip the add_path_to_appropriate_result_list() */
2644 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2645 istate
, &path
, baselen
,
2648 close_cached_dir(&cdir
);
2650 strbuf_release(&path
);
2655 int cmp_dir_entry(const void *p1
, const void *p2
)
2657 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2658 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2660 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2663 /* check if *out lexically strictly contains *in */
2664 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2666 return (out
->len
< in
->len
) &&
2667 (out
->name
[out
->len
- 1] == '/') &&
2668 !memcmp(out
->name
, in
->name
, out
->len
);
2671 static int treat_leading_path(struct dir_struct
*dir
,
2672 struct index_state
*istate
,
2673 const char *path
, int len
,
2674 const struct pathspec
*pathspec
)
2676 struct strbuf sb
= STRBUF_INIT
;
2677 struct strbuf subdir
= STRBUF_INIT
;
2678 int prevlen
, baselen
;
2680 struct cached_dir cdir
;
2681 enum path_treatment state
= path_none
;
2684 * For each directory component of path, we are going to check whether
2685 * that path is relevant given the pathspec. For example, if path is
2687 * then we will ask treat_path() whether we should go into foo, then
2688 * whether we should go into bar, then whether baz is relevant.
2689 * Checking each is important because e.g. if path is
2691 * then we need to check .git to know we shouldn't traverse it.
2692 * If the return from treat_path() is:
2693 * * path_none, for any path, we return false.
2694 * * path_recurse, for all path components, we return true
2695 * * <anything else> for some intermediate component, we make sure
2696 * to add that path to the relevant list but return false
2697 * signifying that we shouldn't recurse into it.
2700 while (len
&& path
[len
- 1] == '/')
2705 memset(&cdir
, 0, sizeof(cdir
));
2706 cdir
.d_type
= DT_DIR
;
2710 prevlen
= baselen
+ !!baselen
;
2711 cp
= path
+ prevlen
;
2712 cp
= memchr(cp
, '/', path
+ len
- cp
);
2716 baselen
= cp
- path
;
2718 strbuf_add(&sb
, path
, baselen
);
2719 if (!is_directory(sb
.buf
))
2722 strbuf_add(&sb
, path
, prevlen
);
2723 strbuf_reset(&subdir
);
2724 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2725 cdir
.d_name
= subdir
.buf
;
2726 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2728 if (state
!= path_recurse
)
2729 break; /* do not recurse into it */
2731 break; /* finished checking */
2733 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2734 &sb
, baselen
, pathspec
,
2737 strbuf_release(&subdir
);
2738 strbuf_release(&sb
);
2739 return state
== path_recurse
;
2742 static const char *get_ident_string(void)
2744 static struct strbuf sb
= STRBUF_INIT
;
2749 if (uname(&uts
) < 0)
2750 die_errno(_("failed to get kernel name and information"));
2751 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2756 static int ident_in_untracked(const struct untracked_cache
*uc
)
2759 * Previous git versions may have saved many NUL separated
2760 * strings in the "ident" field, but it is insane to manage
2761 * many locations, so just take care of the first one.
2764 return !strcmp(uc
->ident
.buf
, get_ident_string());
2767 static void set_untracked_ident(struct untracked_cache
*uc
)
2769 strbuf_reset(&uc
->ident
);
2770 strbuf_addstr(&uc
->ident
, get_ident_string());
2773 * This strbuf used to contain a list of NUL separated
2774 * strings, so save NUL too for backward compatibility.
2776 strbuf_addch(&uc
->ident
, 0);
2779 static unsigned new_untracked_cache_flags(struct index_state
*istate
)
2781 struct repository
*repo
= istate
->repo
;
2785 * This logic is coordinated with the setting of these flags in
2786 * wt-status.c#wt_status_collect_untracked(), and the evaluation
2787 * of the config setting in commit.c#git_status_config()
2789 if (!repo_config_get_string(repo
, "status.showuntrackedfiles", &val
) &&
2790 !strcmp(val
, "all"))
2794 * The default, if "all" is not set, is "normal" - leading us here.
2795 * If the value is "none" then it really doesn't matter.
2797 return DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2800 static void new_untracked_cache(struct index_state
*istate
, int flags
)
2802 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2803 strbuf_init(&uc
->ident
, 100);
2804 uc
->exclude_per_dir
= ".gitignore";
2805 uc
->dir_flags
= flags
>= 0 ? flags
: new_untracked_cache_flags(istate
);
2806 set_untracked_ident(uc
);
2807 istate
->untracked
= uc
;
2808 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2811 void add_untracked_cache(struct index_state
*istate
)
2813 if (!istate
->untracked
) {
2814 new_untracked_cache(istate
, -1);
2816 if (!ident_in_untracked(istate
->untracked
)) {
2817 free_untracked_cache(istate
->untracked
);
2818 new_untracked_cache(istate
, -1);
2823 void remove_untracked_cache(struct index_state
*istate
)
2825 if (istate
->untracked
) {
2826 free_untracked_cache(istate
->untracked
);
2827 istate
->untracked
= NULL
;
2828 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2832 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2834 const struct pathspec
*pathspec
,
2835 struct index_state
*istate
)
2837 struct untracked_cache_dir
*root
;
2838 static int untracked_cache_disabled
= -1;
2840 if (!dir
->untracked
)
2842 if (untracked_cache_disabled
< 0)
2843 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2844 if (untracked_cache_disabled
)
2848 * We only support $GIT_DIR/info/exclude and core.excludesfile
2849 * as the global ignore rule files. Any other additions
2850 * (e.g. from command line) invalidate the cache. This
2851 * condition also catches running setup_standard_excludes()
2852 * before setting dir->untracked!
2854 if (dir
->internal
.unmanaged_exclude_files
)
2858 * Optimize for the main use case only: whole-tree git
2859 * status. More work involved in treat_leading_path() if we
2860 * use cache on just a subset of the worktree. pathspec
2861 * support could make the matter even worse.
2863 if (base_len
|| (pathspec
&& pathspec
->nr
))
2866 /* We don't support collecting ignore files */
2867 if (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2868 DIR_COLLECT_IGNORED
))
2872 * If we use .gitignore in the cache and now you change it to
2873 * .gitexclude, everything will go wrong.
2875 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2876 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2880 * EXC_CMDL is not considered in the cache. If people set it,
2883 if (dir
->internal
.exclude_list_group
[EXC_CMDL
].nr
)
2886 if (!ident_in_untracked(dir
->untracked
)) {
2887 warning(_("untracked cache is disabled on this system or location"));
2892 * If the untracked structure we received does not have the same flags
2893 * as requested in this run, we're going to need to either discard the
2894 * existing structure (and potentially later recreate), or bypass the
2895 * untracked cache mechanism for this run.
2897 if (dir
->flags
!= dir
->untracked
->dir_flags
) {
2899 * If the untracked structure we received does not have the same flags
2900 * as configured, then we need to reset / create a new "untracked"
2901 * structure to match the new config.
2903 * Keeping the saved and used untracked cache consistent with the
2904 * configuration provides an opportunity for frequent users of
2905 * "git status -uall" to leverage the untracked cache by aligning their
2906 * configuration - setting "status.showuntrackedfiles" to "all" or
2907 * "normal" as appropriate.
2909 * Previously using -uall (or setting "status.showuntrackedfiles" to
2910 * "all") was incompatible with untracked cache and *consistently*
2911 * caused surprisingly bad performance (with fscache and fsmonitor
2912 * enabled) on Windows.
2914 * IMPROVEMENT OPPORTUNITY: If we reworked the untracked cache storage
2915 * to not be as bound up with the desired output in a given run,
2916 * and instead iterated through and stored enough information to
2917 * correctly serve both "modes", then users could get peak performance
2918 * with or without '-uall' regardless of their
2919 * "status.showuntrackedfiles" config.
2921 if (dir
->untracked
->dir_flags
!= new_untracked_cache_flags(istate
)) {
2922 free_untracked_cache(istate
->untracked
);
2923 new_untracked_cache(istate
, dir
->flags
);
2924 dir
->untracked
= istate
->untracked
;
2928 * Current untracked cache data is consistent with config, but not
2929 * usable in this request/run; just bypass untracked cache.
2935 if (!dir
->untracked
->root
) {
2936 /* Untracked cache existed but is not initialized; fix that */
2937 FLEX_ALLOC_STR(dir
->untracked
->root
, name
, "");
2938 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2941 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2942 root
= dir
->untracked
->root
;
2943 if (!oideq(&dir
->internal
.ss_info_exclude
.oid
,
2944 &dir
->untracked
->ss_info_exclude
.oid
)) {
2945 invalidate_gitignore(dir
->untracked
, root
);
2946 dir
->untracked
->ss_info_exclude
= dir
->internal
.ss_info_exclude
;
2948 if (!oideq(&dir
->internal
.ss_excludes_file
.oid
,
2949 &dir
->untracked
->ss_excludes_file
.oid
)) {
2950 invalidate_gitignore(dir
->untracked
, root
);
2951 dir
->untracked
->ss_excludes_file
= dir
->internal
.ss_excludes_file
;
2954 /* Make sure this directory is not dropped out at saving phase */
2959 static void emit_traversal_statistics(struct dir_struct
*dir
,
2960 struct repository
*repo
,
2964 if (!trace2_is_enabled())
2968 trace2_data_string("read_directory", repo
, "path", "");
2970 struct strbuf tmp
= STRBUF_INIT
;
2971 strbuf_add(&tmp
, path
, path_len
);
2972 trace2_data_string("read_directory", repo
, "path", tmp
.buf
);
2973 strbuf_release(&tmp
);
2976 trace2_data_intmax("read_directory", repo
,
2977 "directories-visited", dir
->internal
.visited_directories
);
2978 trace2_data_intmax("read_directory", repo
,
2979 "paths-visited", dir
->internal
.visited_paths
);
2981 if (!dir
->untracked
)
2983 trace2_data_intmax("read_directory", repo
,
2984 "node-creation", dir
->untracked
->dir_created
);
2985 trace2_data_intmax("read_directory", repo
,
2986 "gitignore-invalidation",
2987 dir
->untracked
->gitignore_invalidated
);
2988 trace2_data_intmax("read_directory", repo
,
2989 "directory-invalidation",
2990 dir
->untracked
->dir_invalidated
);
2991 trace2_data_intmax("read_directory", repo
,
2992 "opendir", dir
->untracked
->dir_opened
);
2995 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
2996 const char *path
, int len
, const struct pathspec
*pathspec
)
2998 struct untracked_cache_dir
*untracked
;
3000 trace2_region_enter("dir", "read_directory", istate
->repo
);
3001 dir
->internal
.visited_paths
= 0;
3002 dir
->internal
.visited_directories
= 0;
3004 if (has_symlink_leading_path(path
, len
)) {
3005 trace2_region_leave("dir", "read_directory", istate
->repo
);
3009 untracked
= validate_untracked_cache(dir
, len
, pathspec
, istate
);
3012 * make sure untracked cache code path is disabled,
3013 * e.g. prep_exclude()
3015 dir
->untracked
= NULL
;
3016 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
3017 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
3018 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
3019 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
3021 emit_traversal_statistics(dir
, istate
->repo
, path
, len
);
3023 trace2_region_leave("dir", "read_directory", istate
->repo
);
3024 if (dir
->untracked
) {
3025 static int force_untracked_cache
= -1;
3027 if (force_untracked_cache
< 0)
3028 force_untracked_cache
=
3029 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", -1);
3030 if (force_untracked_cache
< 0)
3031 force_untracked_cache
= (istate
->repo
->settings
.core_untracked_cache
== UNTRACKED_CACHE_WRITE
);
3032 if (force_untracked_cache
&&
3033 dir
->untracked
== istate
->untracked
&&
3034 (dir
->untracked
->dir_opened
||
3035 dir
->untracked
->gitignore_invalidated
||
3036 dir
->untracked
->dir_invalidated
))
3037 istate
->cache_changed
|= UNTRACKED_CHANGED
;
3038 if (dir
->untracked
!= istate
->untracked
) {
3039 FREE_AND_NULL(dir
->untracked
);
3046 int file_exists(const char *f
)
3049 return lstat(f
, &sb
) == 0;
3052 int repo_file_exists(struct repository
*repo
, const char *path
)
3054 if (repo
!= the_repository
)
3055 BUG("do not know how to check file existence in arbitrary repo");
3057 return file_exists(path
);
3060 static int cmp_icase(char a
, char b
)
3065 return toupper(a
) - toupper(b
);
3070 * Given two normalized paths (a trailing slash is ok), if subdir is
3071 * outside dir, return -1. Otherwise return the offset in subdir that
3072 * can be used as relative path to dir.
3074 int dir_inside_of(const char *subdir
, const char *dir
)
3078 assert(dir
&& subdir
&& *dir
&& *subdir
);
3080 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
3086 /* hel[p]/me vs hel[l]/yeah */
3087 if (*dir
&& *subdir
)
3091 return !*dir
? offset
: -1; /* same dir */
3093 /* foo/[b]ar vs foo/[] */
3094 if (is_dir_sep(dir
[-1]))
3095 return is_dir_sep(subdir
[-1]) ? offset
: -1;
3097 /* foo[/]bar vs foo[] */
3098 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
3101 int is_inside_dir(const char *dir
)
3110 rc
= (dir_inside_of(cwd
, dir
) >= 0);
3115 int is_empty_dir(const char *path
)
3117 DIR *dir
= opendir(path
);
3124 e
= readdir_skip_dot_and_dotdot(dir
);
3132 char *git_url_basename(const char *repo
, int is_bundle
, int is_bare
)
3134 const char *end
= repo
+ strlen(repo
), *start
, *ptr
;
3141 start
= strstr(repo
, "://");
3148 * Skip authentication data. The stripping does happen
3149 * greedily, such that we strip up to the last '@' inside
3152 for (ptr
= start
; ptr
< end
&& !is_dir_sep(*ptr
); ptr
++) {
3158 * Strip trailing spaces, slashes and /.git
3160 while (start
< end
&& (is_dir_sep(end
[-1]) || isspace(end
[-1])))
3162 if (end
- start
> 5 && is_dir_sep(end
[-5]) &&
3163 !strncmp(end
- 4, ".git", 4)) {
3165 while (start
< end
&& is_dir_sep(end
[-1]))
3170 * It should not be possible to overflow `ptrdiff_t` by passing in an
3171 * insanely long URL, but GCC does not know that and will complain
3172 * without this check.
3174 if (end
- start
< 0)
3175 die(_("No directory name could be guessed.\n"
3176 "Please specify a directory on the command line"));
3179 * Strip trailing port number if we've got only a
3180 * hostname (that is, there is no dir separator but a
3181 * colon). This check is required such that we do not
3182 * strip URI's like '/foo/bar:2222.git', which should
3183 * result in a dir '2222' being guessed due to backwards
3186 if (memchr(start
, '/', end
- start
) == NULL
3187 && memchr(start
, ':', end
- start
) != NULL
) {
3189 while (start
< ptr
&& isdigit(ptr
[-1]) && ptr
[-1] != ':')
3191 if (start
< ptr
&& ptr
[-1] == ':')
3196 * Find last component. To remain backwards compatible we
3197 * also regard colons as path separators, such that
3198 * cloning a repository 'foo:bar.git' would result in a
3199 * directory 'bar' being guessed.
3202 while (start
< ptr
&& !is_dir_sep(ptr
[-1]) && ptr
[-1] != ':')
3207 * Strip .{bundle,git}.
3210 strip_suffix_mem(start
, &len
, is_bundle
? ".bundle" : ".git");
3212 if (!len
|| (len
== 1 && *start
== '/'))
3213 die(_("No directory name could be guessed.\n"
3214 "Please specify a directory on the command line"));
3217 dir
= xstrfmt("%.*s.git", (int)len
, start
);
3219 dir
= xstrndup(start
, len
);
3221 * Replace sequences of 'control' characters and whitespace
3222 * with one ascii space, remove leading and trailing spaces.
3226 int prev_space
= 1 /* strip leading whitespace */;
3227 for (end
= dir
; *end
; ++end
) {
3229 if ((unsigned char)ch
< '\x20')
3240 if (out
> dir
&& prev_space
)
3246 void strip_dir_trailing_slashes(char *dir
)
3248 char *end
= dir
+ strlen(dir
);
3250 while (dir
< end
- 1 && is_dir_sep(end
[-1]))
3255 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
3259 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
3260 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
3261 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
3262 int purge_original_cwd
= (flag
& REMOVE_DIR_PURGE_ORIGINAL_CWD
);
3263 struct object_id submodule_head
;
3265 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
3266 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
3267 /* Do not descend and nuke a nested git work tree. */
3273 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
3274 dir
= opendir(path
->buf
);
3276 if (errno
== ENOENT
)
3277 return keep_toplevel
? -1 : 0;
3278 else if (errno
== EACCES
&& !keep_toplevel
)
3280 * An empty dir could be removable even if it
3283 return rmdir(path
->buf
);
3287 strbuf_complete(path
, '/');
3290 while ((e
= readdir_skip_dot_and_dotdot(dir
)) != NULL
) {
3293 strbuf_setlen(path
, len
);
3294 strbuf_addstr(path
, e
->d_name
);
3295 if (lstat(path
->buf
, &st
)) {
3296 if (errno
== ENOENT
)
3298 * file disappeared, which is what we
3303 } else if (S_ISDIR(st
.st_mode
)) {
3304 if (!remove_dir_recurse(path
, flag
, &kept_down
))
3305 continue; /* happy */
3306 } else if (!only_empty
&&
3307 (!unlink(path
->buf
) || errno
== ENOENT
)) {
3308 continue; /* happy, too */
3311 /* path too long, stat fails, or non-directory still exists */
3317 strbuf_setlen(path
, original_len
);
3318 if (!ret
&& !keep_toplevel
&& !kept_down
) {
3319 if (!purge_original_cwd
&&
3320 startup_info
->original_cwd
&&
3321 !strcmp(startup_info
->original_cwd
, path
->buf
))
3322 ret
= -1; /* Do not remove current working directory */
3324 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
3327 * report the uplevel that it is not an error that we
3328 * did not rmdir() our directory.
3334 int remove_dir_recursively(struct strbuf
*path
, int flag
)
3336 return remove_dir_recurse(path
, flag
, NULL
);
3339 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
3341 void setup_standard_excludes(struct dir_struct
*dir
)
3343 dir
->exclude_per_dir
= ".gitignore";
3345 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3347 excludes_file
= xdg_config_home("ignore");
3348 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
3349 add_patterns_from_file_1(dir
, excludes_file
,
3350 dir
->untracked
? &dir
->internal
.ss_excludes_file
: NULL
);
3352 /* per repository user preference */
3353 if (startup_info
->have_repository
) {
3354 const char *path
= git_path_info_exclude();
3355 if (!access_or_warn(path
, R_OK
, 0))
3356 add_patterns_from_file_1(dir
, path
,
3357 dir
->untracked
? &dir
->internal
.ss_info_exclude
: NULL
);
3361 char *get_sparse_checkout_filename(void)
3363 return git_pathdup("info/sparse-checkout");
3366 int get_sparse_checkout_patterns(struct pattern_list
*pl
)
3369 char *sparse_filename
= get_sparse_checkout_filename();
3371 pl
->use_cone_patterns
= core_sparse_checkout_cone
;
3372 res
= add_patterns_from_file_to_list(sparse_filename
, "", 0, pl
, NULL
, 0);
3374 free(sparse_filename
);
3378 int remove_path(const char *name
)
3382 if (unlink(name
) && !is_missing_file_error(errno
))
3385 slash
= strrchr(name
, '/');
3387 char *dirs
= xstrdup(name
);
3388 slash
= dirs
+ (slash
- name
);
3391 if (startup_info
->original_cwd
&&
3392 !strcmp(startup_info
->original_cwd
, dirs
))
3394 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3401 * Frees memory within dir which was allocated, and resets fields for further
3402 * use. Does not free dir itself.
3404 void dir_clear(struct dir_struct
*dir
)
3407 struct exclude_list_group
*group
;
3408 struct pattern_list
*pl
;
3409 struct exclude_stack
*stk
;
3410 struct dir_struct
new = DIR_INIT
;
3412 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3413 group
= &dir
->internal
.exclude_list_group
[i
];
3414 for (j
= 0; j
< group
->nr
; j
++) {
3417 free((char *)pl
->src
);
3418 clear_pattern_list(pl
);
3423 for (i
= 0; i
< dir
->ignored_nr
; i
++)
3424 free(dir
->ignored
[i
]);
3425 for (i
= 0; i
< dir
->nr
; i
++)
3426 free(dir
->entries
[i
]);
3430 stk
= dir
->internal
.exclude_stack
;
3432 struct exclude_stack
*prev
= stk
->prev
;
3436 strbuf_release(&dir
->internal
.basebuf
);
3438 memcpy(dir
, &new, sizeof(*dir
));
3441 struct ondisk_untracked_cache
{
3442 struct stat_data info_exclude_stat
;
3443 struct stat_data excludes_file_stat
;
3447 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3450 int index
; /* number of written untracked_cache_dir */
3451 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3452 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3453 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3455 struct strbuf sb_stat
;
3456 struct strbuf sb_sha1
;
3459 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3461 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3462 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3463 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3464 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3465 to
->sd_dev
= htonl(from
->sd_dev
);
3466 to
->sd_ino
= htonl(from
->sd_ino
);
3467 to
->sd_uid
= htonl(from
->sd_uid
);
3468 to
->sd_gid
= htonl(from
->sd_gid
);
3469 to
->sd_size
= htonl(from
->sd_size
);
3472 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3473 struct write_data
*wd
)
3475 struct stat_data stat_data
;
3476 struct strbuf
*out
= &wd
->out
;
3477 unsigned char intbuf
[16];
3478 unsigned int intlen
, value
;
3479 int i
= wd
->index
++;
3482 * untracked_nr should be reset whenever valid is clear, but
3485 if (!untracked
->valid
) {
3486 untracked
->untracked_nr
= 0;
3487 untracked
->check_only
= 0;
3490 if (untracked
->check_only
)
3491 ewah_set(wd
->check_only
, i
);
3492 if (untracked
->valid
) {
3493 ewah_set(wd
->valid
, i
);
3494 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3495 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3497 if (!is_null_oid(&untracked
->exclude_oid
)) {
3498 ewah_set(wd
->sha1_valid
, i
);
3499 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3500 the_hash_algo
->rawsz
);
3503 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3504 strbuf_add(out
, intbuf
, intlen
);
3506 /* skip non-recurse directories */
3507 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3508 if (untracked
->dirs
[i
]->recurse
)
3510 intlen
= encode_varint(value
, intbuf
);
3511 strbuf_add(out
, intbuf
, intlen
);
3513 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3515 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3516 strbuf_add(out
, untracked
->untracked
[i
],
3517 strlen(untracked
->untracked
[i
]) + 1);
3519 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3520 if (untracked
->dirs
[i
]->recurse
)
3521 write_one_dir(untracked
->dirs
[i
], wd
);
3524 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3526 struct ondisk_untracked_cache
*ouc
;
3527 struct write_data wd
;
3528 unsigned char varbuf
[16];
3530 const unsigned hashsz
= the_hash_algo
->rawsz
;
3532 CALLOC_ARRAY(ouc
, 1);
3533 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3534 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3535 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3537 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3538 strbuf_add(out
, varbuf
, varint_len
);
3539 strbuf_addbuf(out
, &untracked
->ident
);
3541 strbuf_add(out
, ouc
, sizeof(*ouc
));
3542 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3543 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3544 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3547 if (!untracked
->root
) {
3548 varint_len
= encode_varint(0, varbuf
);
3549 strbuf_add(out
, varbuf
, varint_len
);
3554 wd
.check_only
= ewah_new();
3555 wd
.valid
= ewah_new();
3556 wd
.sha1_valid
= ewah_new();
3557 strbuf_init(&wd
.out
, 1024);
3558 strbuf_init(&wd
.sb_stat
, 1024);
3559 strbuf_init(&wd
.sb_sha1
, 1024);
3560 write_one_dir(untracked
->root
, &wd
);
3562 varint_len
= encode_varint(wd
.index
, varbuf
);
3563 strbuf_add(out
, varbuf
, varint_len
);
3564 strbuf_addbuf(out
, &wd
.out
);
3565 ewah_serialize_strbuf(wd
.valid
, out
);
3566 ewah_serialize_strbuf(wd
.check_only
, out
);
3567 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3568 strbuf_addbuf(out
, &wd
.sb_stat
);
3569 strbuf_addbuf(out
, &wd
.sb_sha1
);
3570 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3572 ewah_free(wd
.valid
);
3573 ewah_free(wd
.check_only
);
3574 ewah_free(wd
.sha1_valid
);
3575 strbuf_release(&wd
.out
);
3576 strbuf_release(&wd
.sb_stat
);
3577 strbuf_release(&wd
.sb_sha1
);
3580 static void free_untracked(struct untracked_cache_dir
*ucd
)
3585 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3586 free_untracked(ucd
->dirs
[i
]);
3587 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3588 free(ucd
->untracked
[i
]);
3589 free(ucd
->untracked
);
3594 void free_untracked_cache(struct untracked_cache
*uc
)
3599 free(uc
->exclude_per_dir_to_free
);
3600 strbuf_release(&uc
->ident
);
3601 free_untracked(uc
->root
);
3607 struct untracked_cache_dir
**ucd
;
3608 struct ewah_bitmap
*check_only
;
3609 struct ewah_bitmap
*valid
;
3610 struct ewah_bitmap
*sha1_valid
;
3611 const unsigned char *data
;
3612 const unsigned char *end
;
3615 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3617 memcpy(to
, data
, sizeof(*to
));
3618 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3619 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3620 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3621 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3622 to
->sd_dev
= ntohl(to
->sd_dev
);
3623 to
->sd_ino
= ntohl(to
->sd_ino
);
3624 to
->sd_uid
= ntohl(to
->sd_uid
);
3625 to
->sd_gid
= ntohl(to
->sd_gid
);
3626 to
->sd_size
= ntohl(to
->sd_size
);
3629 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3630 struct read_data
*rd
)
3632 struct untracked_cache_dir ud
, *untracked
;
3633 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3634 const unsigned char *eos
;
3638 memset(&ud
, 0, sizeof(ud
));
3640 value
= decode_varint(&data
);
3644 ud
.untracked_alloc
= value
;
3645 ud
.untracked_nr
= value
;
3646 if (ud
.untracked_nr
)
3647 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3649 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3652 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3654 eos
= memchr(data
, '\0', end
- data
);
3655 if (!eos
|| eos
== end
)
3658 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3659 memcpy(untracked
, &ud
, sizeof(ud
));
3660 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3663 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3664 eos
= memchr(data
, '\0', end
- data
);
3665 if (!eos
|| eos
== end
)
3667 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3671 rd
->ucd
[rd
->index
++] = untracked
;
3674 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3675 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3681 static void set_check_only(size_t pos
, void *cb
)
3683 struct read_data
*rd
= cb
;
3684 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3688 static void read_stat(size_t pos
, void *cb
)
3690 struct read_data
*rd
= cb
;
3691 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3692 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3693 rd
->data
= rd
->end
+ 1;
3696 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3697 rd
->data
+= sizeof(struct stat_data
);
3701 static void read_oid(size_t pos
, void *cb
)
3703 struct read_data
*rd
= cb
;
3704 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3705 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3706 rd
->data
= rd
->end
+ 1;
3709 oidread(&ud
->exclude_oid
, rd
->data
);
3710 rd
->data
+= the_hash_algo
->rawsz
;
3713 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3714 const unsigned char *sha1
)
3716 stat_data_from_disk(&oid_stat
->stat
, data
);
3717 oidread(&oid_stat
->oid
, sha1
);
3718 oid_stat
->valid
= 1;
3721 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3723 struct untracked_cache
*uc
;
3724 struct read_data rd
;
3725 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3729 const char *exclude_per_dir
;
3730 const unsigned hashsz
= the_hash_algo
->rawsz
;
3731 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3732 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3734 if (sz
<= 1 || end
[-1] != '\0')
3738 ident_len
= decode_varint(&next
);
3739 if (next
+ ident_len
> end
)
3741 ident
= (const char *)next
;
3744 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3747 CALLOC_ARRAY(uc
, 1);
3748 strbuf_init(&uc
->ident
, ident_len
);
3749 strbuf_add(&uc
->ident
, ident
, ident_len
);
3750 load_oid_stat(&uc
->ss_info_exclude
,
3751 next
+ ouc_offset(info_exclude_stat
),
3753 load_oid_stat(&uc
->ss_excludes_file
,
3754 next
+ ouc_offset(excludes_file_stat
),
3755 next
+ offset
+ hashsz
);
3756 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3757 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3758 uc
->exclude_per_dir
= uc
->exclude_per_dir_to_free
= xstrdup(exclude_per_dir
);
3759 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3760 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3764 len
= decode_varint(&next
);
3765 if (next
> end
|| len
== 0)
3768 rd
.valid
= ewah_new();
3769 rd
.check_only
= ewah_new();
3770 rd
.sha1_valid
= ewah_new();
3774 ALLOC_ARRAY(rd
.ucd
, len
);
3776 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3780 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3785 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3790 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3794 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3795 rd
.data
= next
+ len
;
3796 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3797 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3802 ewah_free(rd
.valid
);
3803 ewah_free(rd
.check_only
);
3804 ewah_free(rd
.sha1_valid
);
3807 free_untracked_cache(uc
);
3813 static void invalidate_one_directory(struct untracked_cache
*uc
,
3814 struct untracked_cache_dir
*ucd
)
3816 uc
->dir_invalidated
++;
3818 ucd
->untracked_nr
= 0;
3822 * Normally when an entry is added or removed from a directory,
3823 * invalidating that directory is enough. No need to touch its
3824 * ancestors. When a directory is shown as "foo/bar/" in git-status
3825 * however, deleting or adding an entry may have cascading effect.
3827 * Say the "foo/bar/file" has become untracked, we need to tell the
3828 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3829 * directory any more (because "bar" is managed by foo as an untracked
3832 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3833 * was the last untracked entry in the entire "foo", we should show
3834 * "foo/" instead. Which means we have to invalidate past "bar" up to
3837 * This function traverses all directories from root to leaf. If there
3838 * is a chance of one of the above cases happening, we invalidate back
3839 * to root. Otherwise we just invalidate the leaf. There may be a more
3840 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3841 * detect these cases and avoid unnecessary invalidation, for example,
3842 * checking for the untracked entry named "bar/" in "foo", but for now
3843 * stick to something safe and simple.
3845 static int invalidate_one_component(struct untracked_cache
*uc
,
3846 struct untracked_cache_dir
*dir
,
3847 const char *path
, int len
)
3849 const char *rest
= strchr(path
, '/');
3852 int component_len
= rest
- path
;
3853 struct untracked_cache_dir
*d
=
3854 lookup_untracked(uc
, dir
, path
, component_len
);
3856 invalidate_one_component(uc
, d
, rest
+ 1,
3857 len
- (component_len
+ 1));
3859 invalidate_one_directory(uc
, dir
);
3863 invalidate_one_directory(uc
, dir
);
3864 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3867 void untracked_cache_invalidate_path(struct index_state
*istate
,
3868 const char *path
, int safe_path
)
3870 if (!istate
->untracked
|| !istate
->untracked
->root
)
3872 if (!safe_path
&& !verify_path(path
, 0))
3874 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3875 path
, strlen(path
));
3878 void untracked_cache_remove_from_index(struct index_state
*istate
,
3881 untracked_cache_invalidate_path(istate
, path
, 1);
3884 void untracked_cache_add_to_index(struct index_state
*istate
,
3887 untracked_cache_invalidate_path(istate
, path
, 1);
3890 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3891 const char *sub_gitdir
)
3894 struct repository subrepo
;
3895 struct strbuf sub_wt
= STRBUF_INIT
;
3896 struct strbuf sub_gd
= STRBUF_INIT
;
3898 const struct submodule
*sub
;
3900 /* If the submodule has no working tree, we can ignore it. */
3901 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3904 if (repo_read_index(&subrepo
) < 0)
3905 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3907 /* TODO: audit for interaction with sparse-index. */
3908 ensure_full_index(subrepo
.index
);
3909 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3910 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3912 if (!S_ISGITLINK(ce
->ce_mode
))
3915 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3916 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3918 * Skip entries with the same name in different stages
3919 * to make sure an entry is returned only once.
3923 sub
= submodule_from_path(&subrepo
, null_oid(), ce
->name
);
3924 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3925 /* .gitmodules broken or inactive sub */
3928 strbuf_reset(&sub_wt
);
3929 strbuf_reset(&sub_gd
);
3930 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3931 submodule_name_to_gitdir(&sub_gd
, &subrepo
, sub
->name
);
3933 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3935 strbuf_release(&sub_wt
);
3936 strbuf_release(&sub_gd
);
3937 repo_clear(&subrepo
);
3940 void connect_work_tree_and_git_dir(const char *work_tree_
,
3941 const char *git_dir_
,
3942 int recurse_into_nested
)
3944 struct strbuf gitfile_sb
= STRBUF_INIT
;
3945 struct strbuf cfg_sb
= STRBUF_INIT
;
3946 struct strbuf rel_path
= STRBUF_INIT
;
3947 char *git_dir
, *work_tree
;
3949 /* Prepare .git file */
3950 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3951 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3952 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3954 /* Prepare config file */
3955 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3956 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3957 die(_("could not create directories for %s"), cfg_sb
.buf
);
3959 git_dir
= real_pathdup(git_dir_
, 1);
3960 work_tree
= real_pathdup(work_tree_
, 1);
3962 /* Write .git file */
3963 write_file(gitfile_sb
.buf
, "gitdir: %s",
3964 relative_path(git_dir
, work_tree
, &rel_path
));
3965 /* Update core.worktree setting */
3966 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3967 relative_path(work_tree
, git_dir
, &rel_path
));
3969 strbuf_release(&gitfile_sb
);
3970 strbuf_release(&cfg_sb
);
3971 strbuf_release(&rel_path
);
3973 if (recurse_into_nested
)
3974 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3981 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3983 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3985 if (rename(old_git_dir
, new_git_dir
) < 0)
3986 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3987 old_git_dir
, new_git_dir
);
3989 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);
3992 int path_match_flags(const char *const str
, const enum path_match_flags flags
)
3994 const char *p
= str
;
3996 if (flags
& PATH_MATCH_NATIVE
&&
3997 flags
& PATH_MATCH_XPLATFORM
)
3998 BUG("path_match_flags() must get one match kind, not multiple!");
3999 else if (!(flags
& PATH_MATCH_KINDS_MASK
))
4000 BUG("path_match_flags() must get at least one match kind!");
4002 if (flags
& PATH_MATCH_STARTS_WITH_DOT_SLASH
&&
4003 flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
)
4004 BUG("path_match_flags() must get one platform kind, not multiple!");
4005 else if (!(flags
& PATH_MATCH_PLATFORM_MASK
))
4006 BUG("path_match_flags() must get at least one platform kind!");
4010 if (flags
& PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
&&
4014 if (flags
& PATH_MATCH_NATIVE
)
4015 return is_dir_sep(*p
);
4016 else if (flags
& PATH_MATCH_XPLATFORM
)
4017 return is_xplatform_dir_sep(*p
);