2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
11 #include "object-store.h"
14 #include "wildmatch.h"
18 #include "ewah/ewok.h"
19 #include "fsmonitor.h"
20 #include "submodule-config.h"
23 * Tells read_directory_recursive how a file or directory should be treated.
24 * Values are ordered by significance, e.g. if a directory contains both
25 * excluded and untracked files, it is listed as untracked because
26 * path_untracked > path_excluded.
36 * Support data structure for our opendir/readdir/closedir wrappers
40 struct untracked_cache_dir
*untracked
;
47 struct untracked_cache_dir
*ucd
;
50 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
51 struct index_state
*istate
, const char *path
, int len
,
52 struct untracked_cache_dir
*untracked
,
53 int check_only
, int stop_at_first_file
, const struct pathspec
*pathspec
);
54 static int resolve_dtype(int dtype
, struct index_state
*istate
,
55 const char *path
, int len
);
57 int count_slashes(const char *s
)
66 int fspathcmp(const char *a
, const char *b
)
68 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
71 int fspathncmp(const char *a
, const char *b
, size_t count
)
73 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
76 int git_fnmatch(const struct pathspec_item
*item
,
77 const char *pattern
, const char *string
,
81 if (ps_strncmp(item
, pattern
, string
, prefix
))
86 if (item
->flags
& PATHSPEC_ONESTAR
) {
87 int pattern_len
= strlen(++pattern
);
88 int string_len
= strlen(string
);
89 return string_len
< pattern_len
||
90 ps_strcmp(item
, pattern
,
91 string
+ string_len
- pattern_len
);
93 if (item
->magic
& PATHSPEC_GLOB
)
94 return wildmatch(pattern
, string
,
96 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
98 /* wildmatch has not learned no FNM_PATHNAME mode yet */
99 return wildmatch(pattern
, string
,
100 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
103 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
104 const char *string
, int stringlen
,
108 struct strbuf pat_buf
= STRBUF_INIT
;
109 struct strbuf str_buf
= STRBUF_INIT
;
110 const char *use_pat
= pattern
;
111 const char *use_str
= string
;
113 if (pattern
[patternlen
]) {
114 strbuf_add(&pat_buf
, pattern
, patternlen
);
115 use_pat
= pat_buf
.buf
;
117 if (string
[stringlen
]) {
118 strbuf_add(&str_buf
, string
, stringlen
);
119 use_str
= str_buf
.buf
;
123 flags
|= WM_CASEFOLD
;
124 match_status
= wildmatch(use_pat
, use_str
, flags
);
126 strbuf_release(&pat_buf
);
127 strbuf_release(&str_buf
);
132 static size_t common_prefix_len(const struct pathspec
*pathspec
)
138 * ":(icase)path" is treated as a pathspec full of
139 * wildcard. In other words, only prefix is considered common
140 * prefix. If the pathspec is abc/foo abc/bar, running in
141 * subdir xyz, the common prefix is still xyz, not xyz/abc as
144 GUARD_PATHSPEC(pathspec
,
153 for (n
= 0; n
< pathspec
->nr
; n
++) {
154 size_t i
= 0, len
= 0, item_len
;
155 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
157 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
158 item_len
= pathspec
->items
[n
].prefix
;
160 item_len
= pathspec
->items
[n
].nowildcard_len
;
161 while (i
< item_len
&& (n
== 0 || i
< max
)) {
162 char c
= pathspec
->items
[n
].match
[i
];
163 if (c
!= pathspec
->items
[0].match
[i
])
169 if (n
== 0 || len
< max
) {
179 * Returns a copy of the longest leading path common among all
182 char *common_prefix(const struct pathspec
*pathspec
)
184 unsigned long len
= common_prefix_len(pathspec
);
186 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
189 int fill_directory(struct dir_struct
*dir
,
190 struct index_state
*istate
,
191 const struct pathspec
*pathspec
)
197 * Calculate common prefix for the pathspec, and
198 * use that to optimize the directory walk
200 prefix_len
= common_prefix_len(pathspec
);
201 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
203 /* Read the directory and prune it */
204 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
209 int within_depth(const char *name
, int namelen
,
210 int depth
, int max_depth
)
212 const char *cp
= name
, *cpe
= name
+ namelen
;
218 if (depth
> max_depth
)
225 * Read the contents of the blob with the given OID into a buffer.
226 * Append a trailing LF to the end if the last line doesn't have one.
229 * -1 when the OID is invalid or unknown or does not refer to a blob.
230 * 0 when the blob is empty.
231 * 1 along with { data, size } of the (possibly augmented) buffer
234 * Optionally updates the given oid_stat with the given OID (when valid).
236 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
237 size_t *size_out
, char **data_out
)
239 enum object_type type
;
246 data
= read_object_file(oid
, &type
, &sz
);
247 if (!data
|| type
!= OBJ_BLOB
) {
253 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
254 oidcpy(&oid_stat
->oid
, oid
);
262 if (data
[sz
- 1] != '\n') {
263 data
= xrealloc(data
, st_add(sz
, 1));
267 *size_out
= xsize_t(sz
);
273 #define DO_MATCH_EXCLUDE (1<<0)
274 #define DO_MATCH_DIRECTORY (1<<1)
275 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
278 * Does the given pathspec match the given name? A match is found if
280 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
281 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
282 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
283 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
285 * Return value tells which case it was (1-4), or 0 when there is no match.
287 * It may be instructive to look at a small table of concrete examples
288 * to understand the differences between 1, 2, and 4:
291 * | a/b | a/b/ | a/b/c
292 * ------+-----------+-----------+------------
293 * a/b | EXACT | EXACT[1] | LEADING[2]
294 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
295 * a/b/c | RECURSIVE | RECURSIVE | EXACT
297 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
298 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
300 static int match_pathspec_item(const struct index_state
*istate
,
301 const struct pathspec_item
*item
, int prefix
,
302 const char *name
, int namelen
, unsigned flags
)
304 /* name/namelen has prefix cut off by caller */
305 const char *match
= item
->match
+ prefix
;
306 int matchlen
= item
->len
- prefix
;
309 * The normal call pattern is:
310 * 1. prefix = common_prefix_len(ps);
311 * 2. prune something, or fill_directory
312 * 3. match_pathspec()
314 * 'prefix' at #1 may be shorter than the command's prefix and
315 * it's ok for #2 to match extra files. Those extras will be
318 * Suppose the pathspec is 'foo' and '../bar' running from
319 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
320 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
321 * user does not want XYZ/foo, only the "foo" part should be
322 * case-insensitive. We need to filter out XYZ/foo here. In
323 * other words, we do not trust the caller on comparing the
324 * prefix part when :(icase) is involved. We do exact
325 * comparison ourselves.
327 * Normally the caller (common_prefix_len() in fact) does
328 * _exact_ matching on name[-prefix+1..-1] and we do not need
329 * to check that part. Be defensive and check it anyway, in
330 * case common_prefix_len is changed, or a new caller is
331 * introduced that does not use common_prefix_len.
333 * If the penalty turns out too high when prefix is really
334 * long, maybe change it to
335 * strncmp(match, name, item->prefix - prefix)
337 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
338 strncmp(item
->match
, name
- prefix
, item
->prefix
))
341 if (item
->attr_match_nr
&&
342 !match_pathspec_attrs(istate
, name
, namelen
, item
))
345 /* If the match was just the prefix, we matched */
347 return MATCHED_RECURSIVELY
;
349 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
350 if (matchlen
== namelen
)
351 return MATCHED_EXACTLY
;
353 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
354 return MATCHED_RECURSIVELY
;
355 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
356 match
[matchlen
- 1] == '/' &&
357 namelen
== matchlen
- 1 &&
358 !ps_strncmp(item
, match
, name
, namelen
))
359 return MATCHED_EXACTLY
;
361 if (item
->nowildcard_len
< item
->len
&&
362 !git_fnmatch(item
, match
, name
,
363 item
->nowildcard_len
- prefix
))
364 return MATCHED_FNMATCH
;
366 /* Perform checks to see if "name" is a leading string of the pathspec */
367 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
368 !(flags
& DO_MATCH_EXCLUDE
)) {
369 /* name is a literal prefix of the pathspec */
370 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
371 if ((namelen
< matchlen
) &&
372 (match
[namelen
-offset
] == '/') &&
373 !ps_strncmp(item
, match
, name
, namelen
))
374 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
376 /* name doesn't match up to the first wild character */
377 if (item
->nowildcard_len
< item
->len
&&
378 ps_strncmp(item
, match
, name
,
379 item
->nowildcard_len
- prefix
))
383 * name has no wildcard, and it didn't match as a leading
384 * pathspec so return.
386 if (item
->nowildcard_len
== item
->len
)
390 * Here is where we would perform a wildmatch to check if
391 * "name" can be matched as a directory (or a prefix) against
392 * the pathspec. Since wildmatch doesn't have this capability
393 * at the present we have to punt and say that it is a match,
394 * potentially returning a false positive
395 * The submodules themselves will be able to perform more
396 * accurate matching to determine if the pathspec matches.
398 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
405 * do_match_pathspec() is meant to ONLY be called by
406 * match_pathspec_with_flags(); calling it directly risks pathspecs
407 * like ':!unwanted_path' being ignored.
409 * Given a name and a list of pathspecs, returns the nature of the
410 * closest (i.e. most specific) match of the name to any of the
413 * The caller typically calls this multiple times with the same
414 * pathspec and seen[] array but with different name/namelen
415 * (e.g. entries from the index) and is interested in seeing if and
416 * how each pathspec matches all the names it calls this function
417 * with. A mark is left in the seen[] array for each pathspec element
418 * indicating the closest type of match that element achieved, so if
419 * seen[n] remains zero after multiple invocations, that means the nth
420 * pathspec did not match any names, which could indicate that the
421 * user mistyped the nth pathspec.
423 static int do_match_pathspec(const struct index_state
*istate
,
424 const struct pathspec
*ps
,
425 const char *name
, int namelen
,
426 int prefix
, char *seen
,
429 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
441 if (!ps
->recursive
||
442 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
444 return MATCHED_RECURSIVELY
;
446 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
447 return MATCHED_EXACTLY
;
455 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
458 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
459 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
462 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
465 * Make exclude patterns optional and never report
466 * "pathspec ':(exclude)foo' matches no files"
468 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
469 seen
[i
] = MATCHED_FNMATCH
;
470 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
473 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
474 ps
->max_depth
!= -1 &&
475 how
&& how
!= MATCHED_FNMATCH
) {
476 int len
= ps
->items
[i
].len
;
477 if (name
[len
] == '/')
479 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
480 how
= MATCHED_EXACTLY
;
487 if (seen
&& seen
[i
] < how
)
494 static int match_pathspec_with_flags(const struct index_state
*istate
,
495 const struct pathspec
*ps
,
496 const char *name
, int namelen
,
497 int prefix
, char *seen
, unsigned flags
)
499 int positive
, negative
;
500 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
501 prefix
, seen
, flags
);
502 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
504 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
506 flags
| DO_MATCH_EXCLUDE
);
507 return negative
? 0 : positive
;
510 int match_pathspec(const struct index_state
*istate
,
511 const struct pathspec
*ps
,
512 const char *name
, int namelen
,
513 int prefix
, char *seen
, int is_dir
)
515 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
516 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
517 prefix
, seen
, flags
);
521 * Check if a submodule is a superset of the pathspec
523 int submodule_path_match(const struct index_state
*istate
,
524 const struct pathspec
*ps
,
525 const char *submodule_name
,
528 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
529 strlen(submodule_name
),
532 DO_MATCH_LEADING_PATHSPEC
);
536 int report_path_error(const char *ps_matched
,
537 const struct pathspec
*pathspec
)
540 * Make sure all pathspec matched; otherwise it is an error.
543 for (num
= 0; num
< pathspec
->nr
; num
++) {
544 int other
, found_dup
;
549 * The caller might have fed identical pathspec
550 * twice. Do not barf on such a mistake.
551 * FIXME: parse_pathspec should have eliminated
552 * duplicate pathspec.
554 for (found_dup
= other
= 0;
555 !found_dup
&& other
< pathspec
->nr
;
557 if (other
== num
|| !ps_matched
[other
])
559 if (!strcmp(pathspec
->items
[other
].original
,
560 pathspec
->items
[num
].original
))
562 * Ok, we have a match already.
569 error(_("pathspec '%s' did not match any file(s) known to git"),
570 pathspec
->items
[num
].original
);
577 * Return the length of the "simple" part of a path match limiter.
579 int simple_length(const char *match
)
584 unsigned char c
= *match
++;
586 if (c
== '\0' || is_glob_special(c
))
591 int no_wildcard(const char *string
)
593 return string
[simple_length(string
)] == '\0';
596 void parse_path_pattern(const char **pattern
,
601 const char *p
= *pattern
;
606 *flags
|= PATTERN_FLAG_NEGATIVE
;
610 if (len
&& p
[len
- 1] == '/') {
612 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
614 for (i
= 0; i
< len
; i
++) {
619 *flags
|= PATTERN_FLAG_NODIR
;
620 *nowildcardlen
= simple_length(p
);
622 * we should have excluded the trailing slash from 'p' too,
623 * but that's one more allocation. Instead just make sure
624 * nowildcardlen does not exceed real patternlen
626 if (*nowildcardlen
> len
)
627 *nowildcardlen
= len
;
628 if (*p
== '*' && no_wildcard(p
+ 1))
629 *flags
|= PATTERN_FLAG_ENDSWITH
;
634 int pl_hashmap_cmp(const void *unused_cmp_data
,
635 const struct hashmap_entry
*a
,
636 const struct hashmap_entry
*b
,
639 const struct pattern_entry
*ee1
=
640 container_of(a
, struct pattern_entry
, ent
);
641 const struct pattern_entry
*ee2
=
642 container_of(b
, struct pattern_entry
, ent
);
644 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
649 return strncasecmp(ee1
->pattern
, ee2
->pattern
, min_len
);
650 return strncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
653 static char *dup_and_filter_pattern(const char *pattern
)
657 char *result
= xstrdup(pattern
);
663 /* skip escape characters (once) */
683 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
685 struct pattern_entry
*translated
;
688 const char *prev
, *cur
, *next
;
690 if (!pl
->use_cone_patterns
)
693 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
694 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
695 !strcmp(given
->pattern
, "/*")) {
700 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
705 if (given
->patternlen
< 2 ||
706 *given
->pattern
== '*' ||
707 strstr(given
->pattern
, "**")) {
708 /* Not a cone pattern. */
709 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
713 prev
= given
->pattern
;
714 cur
= given
->pattern
+ 1;
715 next
= given
->pattern
+ 2;
718 /* Watch for glob characters '*', '\', '[', '?' */
719 if (!is_glob_special(*cur
))
722 /* But only if *prev != '\\' */
726 /* But allow the initial '\' */
728 is_glob_special(*next
))
731 /* But a trailing '/' then '*' is fine */
737 /* Not a cone pattern. */
738 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
747 if (given
->patternlen
> 2 &&
748 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
749 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
750 /* Not a cone pattern. */
751 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
755 truncated
= dup_and_filter_pattern(given
->pattern
);
757 translated
= xmalloc(sizeof(struct pattern_entry
));
758 translated
->pattern
= truncated
;
759 translated
->patternlen
= given
->patternlen
- 2;
760 hashmap_entry_init(&translated
->ent
,
762 strihash(translated
->pattern
) :
763 strhash(translated
->pattern
));
765 if (!hashmap_get_entry(&pl
->recursive_hashmap
,
766 translated
, ent
, NULL
)) {
767 /* We did not see the "parent" included */
768 warning(_("unrecognized negative pattern: '%s'"),
775 hashmap_add(&pl
->parent_hashmap
, &translated
->ent
);
776 hashmap_remove(&pl
->recursive_hashmap
, &translated
->ent
, &data
);
781 if (given
->flags
& PATTERN_FLAG_NEGATIVE
) {
782 warning(_("unrecognized negative pattern: '%s'"),
787 translated
= xmalloc(sizeof(struct pattern_entry
));
789 translated
->pattern
= dup_and_filter_pattern(given
->pattern
);
790 translated
->patternlen
= given
->patternlen
;
791 hashmap_entry_init(&translated
->ent
,
793 strihash(translated
->pattern
) :
794 strhash(translated
->pattern
));
796 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
798 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
799 /* we already included this at the parent level */
800 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
802 hashmap_remove(&pl
->parent_hashmap
, &translated
->ent
, &data
);
810 warning(_("disabling cone pattern matching"));
811 hashmap_free_entries(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
812 hashmap_free_entries(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
813 pl
->use_cone_patterns
= 0;
816 static int hashmap_contains_path(struct hashmap
*map
,
817 struct strbuf
*pattern
)
819 struct pattern_entry p
;
821 /* Check straight mapping */
822 p
.pattern
= pattern
->buf
;
823 p
.patternlen
= pattern
->len
;
824 hashmap_entry_init(&p
.ent
,
826 strihash(p
.pattern
) :
828 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
831 int hashmap_contains_parent(struct hashmap
*map
,
833 struct strbuf
*buffer
)
837 strbuf_setlen(buffer
, 0);
840 strbuf_addch(buffer
, '/');
842 strbuf_addstr(buffer
, path
);
844 slash_pos
= strrchr(buffer
->buf
, '/');
846 while (slash_pos
> buffer
->buf
) {
847 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
849 if (hashmap_contains_path(map
, buffer
))
852 slash_pos
= strrchr(buffer
->buf
, '/');
858 void add_pattern(const char *string
, const char *base
,
859 int baselen
, struct pattern_list
*pl
, int srcpos
)
861 struct path_pattern
*pattern
;
866 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
867 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
868 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
870 pattern
= xmalloc(sizeof(*pattern
));
871 pattern
->pattern
= string
;
873 pattern
->patternlen
= patternlen
;
874 pattern
->nowildcardlen
= nowildcardlen
;
875 pattern
->base
= base
;
876 pattern
->baselen
= baselen
;
877 pattern
->flags
= flags
;
878 pattern
->srcpos
= srcpos
;
879 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
880 pl
->patterns
[pl
->nr
++] = pattern
;
883 add_pattern_to_hashsets(pl
, pattern
);
886 static int read_skip_worktree_file_from_index(const struct index_state
*istate
,
888 size_t *size_out
, char **data_out
,
889 struct oid_stat
*oid_stat
)
894 pos
= index_name_pos(istate
, path
, len
);
897 if (!ce_skip_worktree(istate
->cache
[pos
]))
900 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
904 * Frees memory within pl which was allocated for exclude patterns and
905 * the file buffer. Does not free pl itself.
907 void clear_pattern_list(struct pattern_list
*pl
)
911 for (i
= 0; i
< pl
->nr
; i
++)
912 free(pl
->patterns
[i
]);
916 memset(pl
, 0, sizeof(*pl
));
919 static void trim_trailing_spaces(char *buf
)
921 char *p
, *last_space
= NULL
;
923 for (p
= buf
; *p
; p
++)
943 * Given a subdirectory name and "dir" of the current directory,
944 * search the subdir in "dir" and return it, or create a new one if it
945 * does not exist in "dir".
947 * If "name" has the trailing slash, it'll be excluded in the search.
949 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
950 struct untracked_cache_dir
*dir
,
951 const char *name
, int len
)
954 struct untracked_cache_dir
*d
;
957 if (len
&& name
[len
- 1] == '/')
961 while (last
> first
) {
962 int cmp
, next
= first
+ ((last
- first
) >> 1);
964 cmp
= strncmp(name
, d
->name
, len
);
965 if (!cmp
&& strlen(d
->name
) > len
)
977 FLEX_ALLOC_MEM(d
, name
, name
, len
);
979 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
980 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
981 dir
->dirs_nr
- first
);
983 dir
->dirs
[first
] = d
;
987 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
991 dir
->untracked_nr
= 0;
992 for (i
= 0; i
< dir
->dirs_nr
; i
++)
993 do_invalidate_gitignore(dir
->dirs
[i
]);
996 static void invalidate_gitignore(struct untracked_cache
*uc
,
997 struct untracked_cache_dir
*dir
)
999 uc
->gitignore_invalidated
++;
1000 do_invalidate_gitignore(dir
);
1003 static void invalidate_directory(struct untracked_cache
*uc
,
1004 struct untracked_cache_dir
*dir
)
1009 * Invalidation increment here is just roughly correct. If
1010 * untracked_nr or any of dirs[].recurse is non-zero, we
1011 * should increment dir_invalidated too. But that's more
1015 uc
->dir_invalidated
++;
1018 dir
->untracked_nr
= 0;
1019 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1020 dir
->dirs
[i
]->recurse
= 0;
1023 static int add_patterns_from_buffer(char *buf
, size_t size
,
1024 const char *base
, int baselen
,
1025 struct pattern_list
*pl
);
1028 * Given a file with name "fname", read it (either from disk, or from
1029 * an index if 'istate' is non-null), parse it and store the
1030 * exclude rules in "pl".
1032 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
1033 * stat data from disk (only valid if add_patterns returns zero). If
1034 * ss_valid is non-zero, "ss" must contain good value as input.
1036 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1037 struct pattern_list
*pl
, struct index_state
*istate
,
1038 struct oid_stat
*oid_stat
)
1046 fd
= open(fname
, O_RDONLY
);
1047 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1049 warn_on_fopen_errors(fname
);
1054 r
= read_skip_worktree_file_from_index(istate
, fname
,
1060 size
= xsize_t(st
.st_size
);
1063 fill_stat_data(&oid_stat
->stat
, &st
);
1064 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1065 oid_stat
->valid
= 1;
1070 buf
= xmallocz(size
);
1071 if (read_in_full(fd
, buf
, size
) != size
) {
1080 if (oid_stat
->valid
&&
1081 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1082 ; /* no content change, ss->sha1 still good */
1084 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1085 !ce_stage(istate
->cache
[pos
]) &&
1086 ce_uptodate(istate
->cache
[pos
]) &&
1087 !would_convert_to_git(istate
, fname
))
1088 oidcpy(&oid_stat
->oid
,
1089 &istate
->cache
[pos
]->oid
);
1091 hash_object_file(the_hash_algo
, buf
, size
,
1092 "blob", &oid_stat
->oid
);
1093 fill_stat_data(&oid_stat
->stat
, &st
);
1094 oid_stat
->valid
= 1;
1098 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1102 static int add_patterns_from_buffer(char *buf
, size_t size
,
1103 const char *base
, int baselen
,
1104 struct pattern_list
*pl
)
1109 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1110 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1114 if (skip_utf8_bom(&buf
, size
))
1115 size
-= buf
- pl
->filebuf
;
1119 for (i
= 0; i
< size
; i
++) {
1120 if (buf
[i
] == '\n') {
1121 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1122 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1123 trim_trailing_spaces(entry
);
1124 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1127 entry
= buf
+ i
+ 1;
1133 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1134 int baselen
, struct pattern_list
*pl
,
1135 struct index_state
*istate
)
1137 return add_patterns(fname
, base
, baselen
, pl
, istate
, NULL
);
1140 int add_patterns_from_blob_to_list(
1141 struct object_id
*oid
,
1142 const char *base
, int baselen
,
1143 struct pattern_list
*pl
)
1149 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1153 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1157 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1158 int group_type
, const char *src
)
1160 struct pattern_list
*pl
;
1161 struct exclude_list_group
*group
;
1163 group
= &dir
->exclude_list_group
[group_type
];
1164 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1165 pl
= &group
->pl
[group
->nr
++];
1166 memset(pl
, 0, sizeof(*pl
));
1172 * Used to set up core.excludesfile and .git/info/exclude lists.
1174 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1175 struct oid_stat
*oid_stat
)
1177 struct pattern_list
*pl
;
1179 * catch setup_standard_excludes() that's called before
1180 * dir->untracked is assigned. That function behaves
1181 * differently when dir->untracked is non-NULL.
1183 if (!dir
->untracked
)
1184 dir
->unmanaged_exclude_files
++;
1185 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1186 if (add_patterns(fname
, "", 0, pl
, NULL
, oid_stat
) < 0)
1187 die(_("cannot use %s as an exclude file"), fname
);
1190 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1192 dir
->unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1193 add_patterns_from_file_1(dir
, fname
, NULL
);
1196 int match_basename(const char *basename
, int basenamelen
,
1197 const char *pattern
, int prefix
, int patternlen
,
1200 if (prefix
== patternlen
) {
1201 if (patternlen
== basenamelen
&&
1202 !fspathncmp(pattern
, basename
, basenamelen
))
1204 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1205 /* "*literal" matching against "fooliteral" */
1206 if (patternlen
- 1 <= basenamelen
&&
1207 !fspathncmp(pattern
+ 1,
1208 basename
+ basenamelen
- (patternlen
- 1),
1212 if (fnmatch_icase_mem(pattern
, patternlen
,
1213 basename
, basenamelen
,
1220 int match_pathname(const char *pathname
, int pathlen
,
1221 const char *base
, int baselen
,
1222 const char *pattern
, int prefix
, int patternlen
,
1229 * match with FNM_PATHNAME; the pattern has base implicitly
1232 if (*pattern
== '/') {
1239 * baselen does not count the trailing slash. base[] may or
1240 * may not end with a trailing slash though.
1242 if (pathlen
< baselen
+ 1 ||
1243 (baselen
&& pathname
[baselen
] != '/') ||
1244 fspathncmp(pathname
, base
, baselen
))
1247 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1248 name
= pathname
+ pathlen
- namelen
;
1252 * if the non-wildcard part is longer than the
1253 * remaining pathname, surely it cannot match.
1255 if (prefix
> namelen
)
1258 if (fspathncmp(pattern
, name
, prefix
))
1261 patternlen
-= prefix
;
1266 * If the whole pattern did not have a wildcard,
1267 * then our prefix match is all we need; we
1268 * do not need to call fnmatch at all.
1270 if (!patternlen
&& !namelen
)
1274 return fnmatch_icase_mem(pattern
, patternlen
,
1280 * Scan the given exclude list in reverse to see whether pathname
1281 * should be ignored. The first match (i.e. the last on the list), if
1282 * any, determines the fate. Returns the exclude_list element which
1283 * matched, or NULL for undecided.
1285 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1287 const char *basename
,
1289 struct pattern_list
*pl
,
1290 struct index_state
*istate
)
1292 struct path_pattern
*res
= NULL
; /* undecided */
1296 return NULL
; /* undefined */
1298 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1299 struct path_pattern
*pattern
= pl
->patterns
[i
];
1300 const char *exclude
= pattern
->pattern
;
1301 int prefix
= pattern
->nowildcardlen
;
1303 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1304 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1305 if (*dtype
!= DT_DIR
)
1309 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1310 if (match_basename(basename
,
1311 pathlen
- (basename
- pathname
),
1312 exclude
, prefix
, pattern
->patternlen
,
1320 assert(pattern
->baselen
== 0 ||
1321 pattern
->base
[pattern
->baselen
- 1] == '/');
1322 if (match_pathname(pathname
, pathlen
,
1324 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1325 exclude
, prefix
, pattern
->patternlen
,
1335 * Scan the list of patterns to determine if the ordered list
1336 * of patterns matches on 'pathname'.
1338 * Return 1 for a match, 0 for not matched and -1 for undecided.
1340 enum pattern_match_result
path_matches_pattern_list(
1341 const char *pathname
, int pathlen
,
1342 const char *basename
, int *dtype
,
1343 struct pattern_list
*pl
,
1344 struct index_state
*istate
)
1346 struct path_pattern
*pattern
;
1347 struct strbuf parent_pathname
= STRBUF_INIT
;
1348 int result
= NOT_MATCHED
;
1349 const char *slash_pos
;
1351 if (!pl
->use_cone_patterns
) {
1352 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1355 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1367 strbuf_addch(&parent_pathname
, '/');
1368 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1370 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1371 &parent_pathname
)) {
1372 result
= MATCHED_RECURSIVE
;
1376 slash_pos
= strrchr(parent_pathname
.buf
, '/');
1378 if (slash_pos
== parent_pathname
.buf
) {
1379 /* include every file in root */
1384 strbuf_setlen(&parent_pathname
, slash_pos
- parent_pathname
.buf
);
1386 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1391 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1394 result
= MATCHED_RECURSIVE
;
1397 strbuf_release(&parent_pathname
);
1401 static struct path_pattern
*last_matching_pattern_from_lists(
1402 struct dir_struct
*dir
, struct index_state
*istate
,
1403 const char *pathname
, int pathlen
,
1404 const char *basename
, int *dtype_p
)
1407 struct exclude_list_group
*group
;
1408 struct path_pattern
*pattern
;
1409 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1410 group
= &dir
->exclude_list_group
[i
];
1411 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1412 pattern
= last_matching_pattern_from_list(
1413 pathname
, pathlen
, basename
, dtype_p
,
1414 &group
->pl
[j
], istate
);
1423 * Loads the per-directory exclude list for the substring of base
1424 * which has a char length of baselen.
1426 static void prep_exclude(struct dir_struct
*dir
,
1427 struct index_state
*istate
,
1428 const char *base
, int baselen
)
1430 struct exclude_list_group
*group
;
1431 struct pattern_list
*pl
;
1432 struct exclude_stack
*stk
= NULL
;
1433 struct untracked_cache_dir
*untracked
;
1436 group
= &dir
->exclude_list_group
[EXC_DIRS
];
1439 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1440 * which originate from directories not in the prefix of the
1441 * path being checked.
1443 while ((stk
= dir
->exclude_stack
) != NULL
) {
1444 if (stk
->baselen
<= baselen
&&
1445 !strncmp(dir
->basebuf
.buf
, base
, stk
->baselen
))
1447 pl
= &group
->pl
[dir
->exclude_stack
->exclude_ix
];
1448 dir
->exclude_stack
= stk
->prev
;
1449 dir
->pattern
= NULL
;
1450 free((char *)pl
->src
); /* see strbuf_detach() below */
1451 clear_pattern_list(pl
);
1456 /* Skip traversing into sub directories if the parent is excluded */
1461 * Lazy initialization. All call sites currently just
1462 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1463 * them seems lots of work for little benefit.
1465 if (!dir
->basebuf
.buf
)
1466 strbuf_init(&dir
->basebuf
, PATH_MAX
);
1468 /* Read from the parent directories and push them down. */
1469 current
= stk
? stk
->baselen
: -1;
1470 strbuf_setlen(&dir
->basebuf
, current
< 0 ? 0 : current
);
1472 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1476 while (current
< baselen
) {
1478 struct oid_stat oid_stat
;
1480 stk
= xcalloc(1, sizeof(*stk
));
1485 cp
= strchr(base
+ current
+ 1, '/');
1487 die("oops in prep_exclude");
1490 lookup_untracked(dir
->untracked
, untracked
,
1492 cp
- base
- current
);
1494 stk
->prev
= dir
->exclude_stack
;
1495 stk
->baselen
= cp
- base
;
1496 stk
->exclude_ix
= group
->nr
;
1497 stk
->ucd
= untracked
;
1498 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1499 strbuf_add(&dir
->basebuf
, base
+ current
, stk
->baselen
- current
);
1500 assert(stk
->baselen
== dir
->basebuf
.len
);
1502 /* Abort if the directory is excluded */
1505 dir
->basebuf
.buf
[stk
->baselen
- 1] = 0;
1506 dir
->pattern
= last_matching_pattern_from_lists(dir
,
1508 dir
->basebuf
.buf
, stk
->baselen
- 1,
1509 dir
->basebuf
.buf
+ current
, &dt
);
1510 dir
->basebuf
.buf
[stk
->baselen
- 1] = '/';
1512 dir
->pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1513 dir
->pattern
= NULL
;
1515 dir
->exclude_stack
= stk
;
1520 /* Try to read per-directory file */
1521 oidclr(&oid_stat
.oid
);
1523 if (dir
->exclude_per_dir
&&
1525 * If we know that no files have been added in
1526 * this directory (i.e. valid_cached_dir() has
1527 * been executed and set untracked->valid) ..
1529 (!untracked
|| !untracked
->valid
||
1531 * .. and .gitignore does not exist before
1532 * (i.e. null exclude_oid). Then we can skip
1533 * loading .gitignore, which would result in
1536 !is_null_oid(&untracked
->exclude_oid
))) {
1538 * dir->basebuf gets reused by the traversal, but we
1539 * need fname to remain unchanged to ensure the src
1540 * member of each struct path_pattern correctly
1541 * back-references its source file. Other invocations
1542 * of add_pattern_list provide stable strings, so we
1543 * strbuf_detach() and free() here in the caller.
1545 struct strbuf sb
= STRBUF_INIT
;
1546 strbuf_addbuf(&sb
, &dir
->basebuf
);
1547 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1548 pl
->src
= strbuf_detach(&sb
, NULL
);
1549 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1550 untracked
? &oid_stat
: NULL
);
1553 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1554 * will first be called in valid_cached_dir() then maybe many
1555 * times more in last_matching_pattern(). When the cache is
1556 * used, last_matching_pattern() will not be called and
1557 * reading .gitignore content will be a waste.
1559 * So when it's called by valid_cached_dir() and we can get
1560 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1561 * modified on work tree), we could delay reading the
1562 * .gitignore content until we absolutely need it in
1563 * last_matching_pattern(). Be careful about ignore rule
1564 * order, though, if you do that.
1567 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1568 invalidate_gitignore(dir
->untracked
, untracked
);
1569 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1571 dir
->exclude_stack
= stk
;
1572 current
= stk
->baselen
;
1574 strbuf_setlen(&dir
->basebuf
, baselen
);
1578 * Loads the exclude lists for the directory containing pathname, then
1579 * scans all exclude lists to determine whether pathname is excluded.
1580 * Returns the exclude_list element which matched, or NULL for
1583 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1584 struct index_state
*istate
,
1585 const char *pathname
,
1588 int pathlen
= strlen(pathname
);
1589 const char *basename
= strrchr(pathname
, '/');
1590 basename
= (basename
) ? basename
+1 : pathname
;
1592 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1595 return dir
->pattern
;
1597 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1602 * Loads the exclude lists for the directory containing pathname, then
1603 * scans all exclude lists to determine whether pathname is excluded.
1604 * Returns 1 if true, otherwise 0.
1606 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1607 const char *pathname
, int *dtype_p
)
1609 struct path_pattern
*pattern
=
1610 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1612 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1616 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1618 struct dir_entry
*ent
;
1620 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1625 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1626 struct index_state
*istate
,
1627 const char *pathname
, int len
)
1629 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1632 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
1633 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1636 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1637 struct index_state
*istate
,
1638 const char *pathname
, int len
)
1640 if (!index_name_is_other(istate
, pathname
, len
))
1643 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
1644 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1648 index_nonexistent
= 0,
1654 * Do not use the alphabetically sorted index to look up
1655 * the directory name; instead, use the case insensitive
1658 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1659 const char *dirname
, int len
)
1661 struct cache_entry
*ce
;
1663 if (index_dir_exists(istate
, dirname
, len
))
1664 return index_directory
;
1666 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1667 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1668 return index_gitdir
;
1670 return index_nonexistent
;
1674 * The index sorts alphabetically by entry name, which
1675 * means that a gitlink sorts as '\0' at the end, while
1676 * a directory (which is defined not as an entry, but as
1677 * the files it contains) will sort with the '/' at the
1680 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1681 const char *dirname
, int len
)
1686 return directory_exists_in_index_icase(istate
, dirname
, len
);
1688 pos
= index_name_pos(istate
, dirname
, len
);
1691 while (pos
< istate
->cache_nr
) {
1692 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1693 unsigned char endchar
;
1695 if (strncmp(ce
->name
, dirname
, len
))
1697 endchar
= ce
->name
[len
];
1701 return index_directory
;
1702 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1703 return index_gitdir
;
1705 return index_nonexistent
;
1709 * When we find a directory when traversing the filesystem, we
1710 * have three distinct cases:
1713 * - see it as a directory
1716 * and which one we choose depends on a combination of existing
1717 * git index contents and the flags passed into the directory
1718 * traversal routine.
1720 * Case 1: If we *already* have entries in the index under that
1721 * directory name, we always recurse into the directory to see
1724 * Case 2: If we *already* have that directory name as a gitlink,
1725 * we always continue to see it as a gitlink, regardless of whether
1726 * there is an actual git directory there or not (it might not
1727 * be checked out as a subproject!)
1729 * Case 3: if we didn't have it in the index previously, we
1730 * have a few sub-cases:
1732 * (a) if "show_other_directories" is true, we show it as
1733 * just a directory, unless "hide_empty_directories" is
1734 * also true, in which case we need to check if it contains any
1735 * untracked and / or ignored files.
1736 * (b) if it looks like a git directory, and we don't have
1737 * 'no_gitlinks' set we treat it as a gitlink, and show it
1739 * (c) otherwise, we recurse into it.
1741 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1742 struct index_state
*istate
,
1743 struct untracked_cache_dir
*untracked
,
1744 const char *dirname
, int len
, int baselen
, int excluded
,
1745 const struct pathspec
*pathspec
)
1748 * WARNING: From this function, you can return path_recurse or you
1749 * can call read_directory_recursive() (or neither), but
1750 * you CAN'T DO BOTH.
1752 enum path_treatment state
;
1753 int matches_how
= 0;
1754 int nested_repo
= 0, check_only
, stop_early
;
1755 int old_ignored_nr
, old_untracked_nr
;
1756 /* The "len-1" is to strip the final '/' */
1757 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1759 if (status
== index_directory
)
1760 return path_recurse
;
1761 if (status
== index_gitdir
)
1763 if (status
!= index_nonexistent
)
1764 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1767 * We don't want to descend into paths that don't match the necessary
1768 * patterns. Clearly, if we don't have a pathspec, then we can't check
1769 * for matching patterns. Also, if (excluded) then we know we matched
1770 * the exclusion patterns so as an optimization we can skip checking
1771 * for matching patterns.
1773 if (pathspec
&& !excluded
) {
1774 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1778 DO_MATCH_LEADING_PATHSPEC
);
1784 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1785 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1786 struct strbuf sb
= STRBUF_INIT
;
1787 strbuf_addstr(&sb
, dirname
);
1788 nested_repo
= is_nonbare_repository_dir(&sb
);
1789 strbuf_release(&sb
);
1792 return ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ? path_none
:
1793 (excluded
? path_excluded
: path_untracked
));
1795 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1797 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1798 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1801 * This is an excluded directory and we are
1802 * showing ignored paths that match an exclude
1803 * pattern. (e.g. show directory as ignored
1804 * only if it matches an exclude pattern).
1805 * This path will either be 'path_excluded`
1806 * (if we are showing empty directories or if
1807 * the directory is not empty), or will be
1808 * 'path_none' (empty directory, and we are
1809 * not showing empty directories).
1811 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1812 return path_excluded
;
1814 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1815 untracked
, 1, 1, pathspec
) == path_excluded
)
1816 return path_excluded
;
1820 return path_recurse
;
1823 /* This is the "show_other_directories" case */
1826 * If we have a pathspec which could match something _below_ this
1827 * directory (e.g. when checking 'subdir/' having a pathspec like
1828 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1831 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1832 return path_recurse
;
1835 * Other than the path_recurse case immediately above, we only need
1836 * to recurse into untracked/ignored directories if either of the
1837 * following bits is set:
1838 * - DIR_SHOW_IGNORED_TOO (because then we need to determine if
1839 * there are ignored directories below)
1840 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
1841 * the directory is empty)
1843 if (!(dir
->flags
& (DIR_SHOW_IGNORED_TOO
| DIR_HIDE_EMPTY_DIRECTORIES
)))
1844 return excluded
? path_excluded
: path_untracked
;
1847 * ...and even if DIR_SHOW_IGNORED_TOO is set, we can still avoid
1848 * recursing into ignored directories if the path is excluded and
1849 * DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1852 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1853 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
1854 return path_excluded
;
1857 * If we have we don't want to know the all the paths under an
1858 * untracked or ignored directory, we still need to go into the
1859 * directory to determine if it is empty (because an empty directory
1860 * should be path_none instead of path_excluded or path_untracked).
1862 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
1863 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
1866 * However, there's another optimization possible as a subset of
1867 * check_only, based on the cases we have to consider:
1868 * A) Directory matches no exclude patterns:
1869 * * Directory is empty => path_none
1870 * * Directory has an untracked file under it => path_untracked
1871 * * Directory has only ignored files under it => path_excluded
1872 * B) Directory matches an exclude pattern:
1873 * * Directory is empty => path_none
1874 * * Directory has an untracked file under it => path_excluded
1875 * * Directory has only ignored files under it => path_excluded
1876 * In case A, we can exit as soon as we've found an untracked
1877 * file but otherwise have to walk all files. In case B, though,
1878 * we can stop at the first file we find under the directory.
1880 stop_early
= check_only
&& excluded
;
1883 * If /every/ file within an untracked directory is ignored, then
1884 * we want to treat the directory as ignored (for e.g. status
1885 * --porcelain), without listing the individual ignored files
1886 * underneath. To do so, we'll save the current ignored_nr, and
1887 * pop all the ones added after it if it turns out the entire
1888 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
1889 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
1890 * untracked paths so will need to pop all those off the last
1891 * after we traverse.
1893 old_ignored_nr
= dir
->ignored_nr
;
1894 old_untracked_nr
= dir
->nr
;
1896 /* Actually recurse into dirname now, we'll fixup the state later. */
1897 untracked
= lookup_untracked(dir
->untracked
, untracked
,
1898 dirname
+ baselen
, len
- baselen
);
1899 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
1900 check_only
, stop_early
, pathspec
);
1902 /* There are a variety of reasons we may need to fixup the state... */
1903 if (state
== path_excluded
) {
1904 /* state == path_excluded implies all paths under
1905 * dirname were ignored...
1907 * if running e.g. `git status --porcelain --ignored=matching`,
1908 * then we want to see the subpaths that are ignored.
1910 * if running e.g. just `git status --porcelain`, then
1911 * we just want the directory itself to be listed as ignored
1912 * and not the individual paths underneath.
1914 int want_ignored_subpaths
=
1915 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1916 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
1918 if (want_ignored_subpaths
) {
1920 * with --ignored=matching, we want the subpaths
1921 * INSTEAD of the directory itself.
1926 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
1927 FREE_AND_NULL(dir
->ignored
[i
]);
1928 dir
->ignored_nr
= old_ignored_nr
;
1933 * We may need to ignore some of the untracked paths we found while
1934 * traversing subdirectories.
1936 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1937 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
1939 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
1940 FREE_AND_NULL(dir
->entries
[i
]);
1941 dir
->nr
= old_untracked_nr
;
1945 * If there is nothing under the current directory and we are not
1946 * hiding empty directories, then we need to report on the
1947 * untracked or ignored status of the directory itself.
1949 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1950 state
= excluded
? path_excluded
: path_untracked
;
1956 * This is an inexact early pruning of any recursive directory
1957 * reading - if the path cannot possibly be in the pathspec,
1958 * return true, and we'll skip it early.
1960 static int simplify_away(const char *path
, int pathlen
,
1961 const struct pathspec
*pathspec
)
1965 if (!pathspec
|| !pathspec
->nr
)
1968 GUARD_PATHSPEC(pathspec
,
1977 for (i
= 0; i
< pathspec
->nr
; i
++) {
1978 const struct pathspec_item
*item
= &pathspec
->items
[i
];
1979 int len
= item
->nowildcard_len
;
1983 if (!ps_strncmp(item
, item
->match
, path
, len
))
1991 * This function tells us whether an excluded path matches a
1992 * list of "interesting" pathspecs. That is, whether a path matched
1993 * by any of the pathspecs could possibly be ignored by excluding
1994 * the specified path. This can happen if:
1996 * 1. the path is mentioned explicitly in the pathspec
1998 * 2. the path is a directory prefix of some element in the
2001 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2002 const struct pathspec
*pathspec
)
2006 if (!pathspec
|| !pathspec
->nr
)
2009 GUARD_PATHSPEC(pathspec
,
2017 for (i
= 0; i
< pathspec
->nr
; i
++) {
2018 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2019 int len
= item
->nowildcard_len
;
2021 if (len
== pathlen
&&
2022 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2024 if (len
> pathlen
&&
2025 item
->match
[pathlen
] == '/' &&
2026 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2032 static int get_index_dtype(struct index_state
*istate
,
2033 const char *path
, int len
)
2036 const struct cache_entry
*ce
;
2038 ce
= index_file_exists(istate
, path
, len
, 0);
2040 if (!ce_uptodate(ce
))
2042 if (S_ISGITLINK(ce
->ce_mode
))
2045 * Nobody actually cares about the
2046 * difference between DT_LNK and DT_REG
2051 /* Try to look it up as a directory */
2052 pos
= index_name_pos(istate
, path
, len
);
2056 while (pos
< istate
->cache_nr
) {
2057 ce
= istate
->cache
[pos
++];
2058 if (strncmp(ce
->name
, path
, len
))
2060 if (ce
->name
[len
] > '/')
2062 if (ce
->name
[len
] < '/')
2064 if (!ce_uptodate(ce
))
2065 break; /* continue? */
2071 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2072 const char *path
, int len
)
2076 if (dtype
!= DT_UNKNOWN
)
2078 dtype
= get_index_dtype(istate
, path
, len
);
2079 if (dtype
!= DT_UNKNOWN
)
2081 if (lstat(path
, &st
))
2083 if (S_ISREG(st
.st_mode
))
2085 if (S_ISDIR(st
.st_mode
))
2087 if (S_ISLNK(st
.st_mode
))
2092 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2093 struct untracked_cache_dir
*untracked
,
2094 struct cached_dir
*cdir
,
2095 struct index_state
*istate
,
2096 struct strbuf
*path
,
2098 const struct pathspec
*pathspec
)
2101 * WARNING: From this function, you can return path_recurse or you
2102 * can call read_directory_recursive() (or neither), but
2103 * you CAN'T DO BOTH.
2105 strbuf_setlen(path
, baselen
);
2107 strbuf_addstr(path
, cdir
->file
);
2108 return path_untracked
;
2110 strbuf_addstr(path
, cdir
->ucd
->name
);
2111 /* treat_one_path() does this before it calls treat_directory() */
2112 strbuf_complete(path
, '/');
2113 if (cdir
->ucd
->check_only
)
2115 * check_only is set as a result of treat_directory() getting
2116 * to its bottom. Verify again the same set of directories
2117 * with check_only set.
2119 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2120 cdir
->ucd
, 1, 0, pathspec
);
2122 * We get path_recurse in the first run when
2123 * directory_exists_in_index() returns index_nonexistent. We
2124 * are sure that new changes in the index does not impact the
2125 * outcome. Return now.
2127 return path_recurse
;
2130 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2131 struct untracked_cache_dir
*untracked
,
2132 struct cached_dir
*cdir
,
2133 struct index_state
*istate
,
2134 struct strbuf
*path
,
2136 const struct pathspec
*pathspec
)
2138 int has_path_in_index
, dtype
, excluded
;
2141 return treat_path_fast(dir
, untracked
, cdir
, istate
, path
,
2143 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2145 strbuf_setlen(path
, baselen
);
2146 strbuf_addstr(path
, cdir
->d_name
);
2147 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2150 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2152 /* Always exclude indexed files */
2153 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2155 if (dtype
!= DT_DIR
&& has_path_in_index
)
2159 * When we are looking at a directory P in the working tree,
2160 * there are three cases:
2162 * (1) P exists in the index. Everything inside the directory P in
2163 * the working tree needs to go when P is checked out from the
2166 * (2) P does not exist in the index, but there is P/Q in the index.
2167 * We know P will stay a directory when we check out the contents
2168 * of the index, but we do not know yet if there is a directory
2169 * P/Q in the working tree to be killed, so we need to recurse.
2171 * (3) P does not exist in the index, and there is no P/Q in the index
2172 * to require P to be a directory, either. Only in this case, we
2173 * know that everything inside P will not be killed without
2176 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2177 (dtype
== DT_DIR
) &&
2178 !has_path_in_index
&&
2179 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2182 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2185 * Excluded? If we don't explicitly want to show
2186 * ignored files, ignore it
2188 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2189 return path_excluded
;
2196 * WARNING: Do not ignore/amend the return value from
2197 * treat_directory(), and especially do not change it to return
2198 * path_recurse as that can cause exponential slowdown.
2199 * Instead, modify treat_directory() to return the right value.
2201 strbuf_addch(path
, '/');
2202 return treat_directory(dir
, istate
, untracked
,
2203 path
->buf
, path
->len
,
2204 baselen
, excluded
, pathspec
);
2208 return path_excluded
;
2210 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2211 0 /* prefix */, NULL
/* seen */,
2214 return path_untracked
;
2218 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2222 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2223 dir
->untracked_alloc
);
2224 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2227 static int valid_cached_dir(struct dir_struct
*dir
,
2228 struct untracked_cache_dir
*untracked
,
2229 struct index_state
*istate
,
2230 struct strbuf
*path
,
2239 * With fsmonitor, we can trust the untracked cache's valid field.
2241 refresh_fsmonitor(istate
);
2242 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2243 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2244 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2247 if (!untracked
->valid
||
2248 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2249 fill_stat_data(&untracked
->stat_data
, &st
);
2254 if (untracked
->check_only
!= !!check_only
)
2258 * prep_exclude will be called eventually on this directory,
2259 * but it's called much later in last_matching_pattern(). We
2260 * need it now to determine the validity of the cache for this
2261 * path. The next calls will be nearly no-op, the way
2262 * prep_exclude() is designed.
2264 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2265 strbuf_addch(path
, '/');
2266 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2267 strbuf_setlen(path
, path
->len
- 1);
2269 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2271 /* hopefully prep_exclude() haven't invalidated this entry... */
2272 return untracked
->valid
;
2275 static int open_cached_dir(struct cached_dir
*cdir
,
2276 struct dir_struct
*dir
,
2277 struct untracked_cache_dir
*untracked
,
2278 struct index_state
*istate
,
2279 struct strbuf
*path
,
2284 memset(cdir
, 0, sizeof(*cdir
));
2285 cdir
->untracked
= untracked
;
2286 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2288 c_path
= path
->len
? path
->buf
: ".";
2289 cdir
->fdir
= opendir(c_path
);
2291 warning_errno(_("could not open directory '%s'"), c_path
);
2292 if (dir
->untracked
) {
2293 invalidate_directory(dir
->untracked
, untracked
);
2294 dir
->untracked
->dir_opened
++;
2301 static int read_cached_dir(struct cached_dir
*cdir
)
2306 de
= readdir(cdir
->fdir
);
2308 cdir
->d_name
= NULL
;
2309 cdir
->d_type
= DT_UNKNOWN
;
2312 cdir
->d_name
= de
->d_name
;
2313 cdir
->d_type
= DTYPE(de
);
2316 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2317 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2327 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2328 struct untracked_cache_dir
*d
= cdir
->untracked
;
2329 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2335 static void close_cached_dir(struct cached_dir
*cdir
)
2338 closedir(cdir
->fdir
);
2340 * We have gone through this directory and found no untracked
2341 * entries. Mark it valid.
2343 if (cdir
->untracked
) {
2344 cdir
->untracked
->valid
= 1;
2345 cdir
->untracked
->recurse
= 1;
2349 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2350 struct untracked_cache_dir
*untracked
,
2351 struct cached_dir
*cdir
,
2352 struct index_state
*istate
,
2353 struct strbuf
*path
,
2355 const struct pathspec
*pathspec
,
2356 enum path_treatment state
)
2358 /* add the path to the appropriate result list */
2361 if (dir
->flags
& DIR_SHOW_IGNORED
)
2362 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2363 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2364 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2365 exclude_matches_pathspec(path
->buf
, path
->len
,
2367 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2370 case path_untracked
:
2371 if (dir
->flags
& DIR_SHOW_IGNORED
)
2373 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2375 add_untracked(untracked
, path
->buf
+ baselen
);
2384 * Read a directory tree. We currently ignore anything but
2385 * directories, regular files and symlinks. That's because git
2386 * doesn't handle them at all yet. Maybe that will change some
2389 * Also, we ignore the name ".git" (even if it is not a directory).
2390 * That likely will not change.
2392 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2393 * to signal that a file was found. This is the least significant value that
2394 * indicates that a file was encountered that does not depend on the order of
2395 * whether an untracked or excluded path was encountered first.
2397 * Returns the most significant path_treatment value encountered in the scan.
2398 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2399 * significant path_treatment value that will be returned.
2402 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2403 struct index_state
*istate
, const char *base
, int baselen
,
2404 struct untracked_cache_dir
*untracked
, int check_only
,
2405 int stop_at_first_file
, const struct pathspec
*pathspec
)
2408 * WARNING: Do NOT recurse unless path_recurse is returned from
2409 * treat_path(). Recursing on any other return value
2410 * can result in exponential slowdown.
2412 struct cached_dir cdir
;
2413 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2414 struct strbuf path
= STRBUF_INIT
;
2416 strbuf_add(&path
, base
, baselen
);
2418 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2422 untracked
->check_only
= !!check_only
;
2424 while (!read_cached_dir(&cdir
)) {
2425 /* check how the file or directory should be treated */
2426 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2429 if (state
> dir_state
)
2432 /* recurse into subdir if instructed by treat_path */
2433 if (state
== path_recurse
) {
2434 struct untracked_cache_dir
*ud
;
2435 ud
= lookup_untracked(dir
->untracked
, untracked
,
2437 path
.len
- baselen
);
2439 read_directory_recursive(dir
, istate
, path
.buf
,
2441 check_only
, stop_at_first_file
, pathspec
);
2442 if (subdir_state
> dir_state
)
2443 dir_state
= subdir_state
;
2446 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2447 0 /* prefix */, NULL
,
2448 0 /* do NOT special case dirs */))
2453 if (stop_at_first_file
) {
2455 * If stopping at first file, then
2456 * signal that a file was found by
2457 * returning `path_excluded`. This is
2458 * to return a consistent value
2459 * regardless of whether an ignored or
2460 * excluded file happened to be
2463 * In current usage, the
2464 * `stop_at_first_file` is passed when
2465 * an ancestor directory has matched
2466 * an exclude pattern, so any found
2467 * files will be excluded.
2469 if (dir_state
>= path_excluded
) {
2470 dir_state
= path_excluded
;
2475 /* abort early if maximum state has been reached */
2476 if (dir_state
== path_untracked
) {
2478 add_untracked(untracked
, path
.buf
+ baselen
);
2481 /* skip the add_path_to_appropriate_result_list() */
2485 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2486 istate
, &path
, baselen
,
2489 close_cached_dir(&cdir
);
2491 strbuf_release(&path
);
2496 int cmp_dir_entry(const void *p1
, const void *p2
)
2498 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2499 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2501 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2504 /* check if *out lexically strictly contains *in */
2505 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2507 return (out
->len
< in
->len
) &&
2508 (out
->name
[out
->len
- 1] == '/') &&
2509 !memcmp(out
->name
, in
->name
, out
->len
);
2512 static int treat_leading_path(struct dir_struct
*dir
,
2513 struct index_state
*istate
,
2514 const char *path
, int len
,
2515 const struct pathspec
*pathspec
)
2517 struct strbuf sb
= STRBUF_INIT
;
2518 struct strbuf subdir
= STRBUF_INIT
;
2519 int prevlen
, baselen
;
2521 struct cached_dir cdir
;
2522 enum path_treatment state
= path_none
;
2525 * For each directory component of path, we are going to check whether
2526 * that path is relevant given the pathspec. For example, if path is
2528 * then we will ask treat_path() whether we should go into foo, then
2529 * whether we should go into bar, then whether baz is relevant.
2530 * Checking each is important because e.g. if path is
2532 * then we need to check .git to know we shouldn't traverse it.
2533 * If the return from treat_path() is:
2534 * * path_none, for any path, we return false.
2535 * * path_recurse, for all path components, we return true
2536 * * <anything else> for some intermediate component, we make sure
2537 * to add that path to the relevant list but return false
2538 * signifying that we shouldn't recurse into it.
2541 while (len
&& path
[len
- 1] == '/')
2546 memset(&cdir
, 0, sizeof(cdir
));
2547 cdir
.d_type
= DT_DIR
;
2551 prevlen
= baselen
+ !!baselen
;
2552 cp
= path
+ prevlen
;
2553 cp
= memchr(cp
, '/', path
+ len
- cp
);
2557 baselen
= cp
- path
;
2559 strbuf_add(&sb
, path
, baselen
);
2560 if (!is_directory(sb
.buf
))
2563 strbuf_add(&sb
, path
, prevlen
);
2564 strbuf_reset(&subdir
);
2565 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2566 cdir
.d_name
= subdir
.buf
;
2567 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2569 if (state
!= path_recurse
)
2570 break; /* do not recurse into it */
2572 break; /* finished checking */
2574 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2575 &sb
, baselen
, pathspec
,
2578 strbuf_release(&subdir
);
2579 strbuf_release(&sb
);
2580 return state
== path_recurse
;
2583 static const char *get_ident_string(void)
2585 static struct strbuf sb
= STRBUF_INIT
;
2590 if (uname(&uts
) < 0)
2591 die_errno(_("failed to get kernel name and information"));
2592 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2597 static int ident_in_untracked(const struct untracked_cache
*uc
)
2600 * Previous git versions may have saved many NUL separated
2601 * strings in the "ident" field, but it is insane to manage
2602 * many locations, so just take care of the first one.
2605 return !strcmp(uc
->ident
.buf
, get_ident_string());
2608 static void set_untracked_ident(struct untracked_cache
*uc
)
2610 strbuf_reset(&uc
->ident
);
2611 strbuf_addstr(&uc
->ident
, get_ident_string());
2614 * This strbuf used to contain a list of NUL separated
2615 * strings, so save NUL too for backward compatibility.
2617 strbuf_addch(&uc
->ident
, 0);
2620 static void new_untracked_cache(struct index_state
*istate
)
2622 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2623 strbuf_init(&uc
->ident
, 100);
2624 uc
->exclude_per_dir
= ".gitignore";
2625 /* should be the same flags used by git-status */
2626 uc
->dir_flags
= DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2627 set_untracked_ident(uc
);
2628 istate
->untracked
= uc
;
2629 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2632 void add_untracked_cache(struct index_state
*istate
)
2634 if (!istate
->untracked
) {
2635 new_untracked_cache(istate
);
2637 if (!ident_in_untracked(istate
->untracked
)) {
2638 free_untracked_cache(istate
->untracked
);
2639 new_untracked_cache(istate
);
2644 void remove_untracked_cache(struct index_state
*istate
)
2646 if (istate
->untracked
) {
2647 free_untracked_cache(istate
->untracked
);
2648 istate
->untracked
= NULL
;
2649 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2653 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2655 const struct pathspec
*pathspec
)
2657 struct untracked_cache_dir
*root
;
2658 static int untracked_cache_disabled
= -1;
2660 if (!dir
->untracked
)
2662 if (untracked_cache_disabled
< 0)
2663 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2664 if (untracked_cache_disabled
)
2668 * We only support $GIT_DIR/info/exclude and core.excludesfile
2669 * as the global ignore rule files. Any other additions
2670 * (e.g. from command line) invalidate the cache. This
2671 * condition also catches running setup_standard_excludes()
2672 * before setting dir->untracked!
2674 if (dir
->unmanaged_exclude_files
)
2678 * Optimize for the main use case only: whole-tree git
2679 * status. More work involved in treat_leading_path() if we
2680 * use cache on just a subset of the worktree. pathspec
2681 * support could make the matter even worse.
2683 if (base_len
|| (pathspec
&& pathspec
->nr
))
2686 /* Different set of flags may produce different results */
2687 if (dir
->flags
!= dir
->untracked
->dir_flags
||
2689 * See treat_directory(), case index_nonexistent. Without
2690 * this flag, we may need to also cache .git file content
2691 * for the resolve_gitlink_ref() call, which we don't.
2693 !(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
) ||
2694 /* We don't support collecting ignore files */
2695 (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2696 DIR_COLLECT_IGNORED
)))
2700 * If we use .gitignore in the cache and now you change it to
2701 * .gitexclude, everything will go wrong.
2703 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2704 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2708 * EXC_CMDL is not considered in the cache. If people set it,
2711 if (dir
->exclude_list_group
[EXC_CMDL
].nr
)
2714 if (!ident_in_untracked(dir
->untracked
)) {
2715 warning(_("untracked cache is disabled on this system or location"));
2719 if (!dir
->untracked
->root
) {
2720 const int len
= sizeof(*dir
->untracked
->root
);
2721 dir
->untracked
->root
= xmalloc(len
);
2722 memset(dir
->untracked
->root
, 0, len
);
2725 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2726 root
= dir
->untracked
->root
;
2727 if (!oideq(&dir
->ss_info_exclude
.oid
,
2728 &dir
->untracked
->ss_info_exclude
.oid
)) {
2729 invalidate_gitignore(dir
->untracked
, root
);
2730 dir
->untracked
->ss_info_exclude
= dir
->ss_info_exclude
;
2732 if (!oideq(&dir
->ss_excludes_file
.oid
,
2733 &dir
->untracked
->ss_excludes_file
.oid
)) {
2734 invalidate_gitignore(dir
->untracked
, root
);
2735 dir
->untracked
->ss_excludes_file
= dir
->ss_excludes_file
;
2738 /* Make sure this directory is not dropped out at saving phase */
2743 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
2744 const char *path
, int len
, const struct pathspec
*pathspec
)
2746 struct untracked_cache_dir
*untracked
;
2748 trace_performance_enter();
2750 if (has_symlink_leading_path(path
, len
)) {
2751 trace_performance_leave("read directory %.*s", len
, path
);
2755 untracked
= validate_untracked_cache(dir
, len
, pathspec
);
2758 * make sure untracked cache code path is disabled,
2759 * e.g. prep_exclude()
2761 dir
->untracked
= NULL
;
2762 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
2763 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
2764 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
2765 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
2767 trace_performance_leave("read directory %.*s", len
, path
);
2768 if (dir
->untracked
) {
2769 static int force_untracked_cache
= -1;
2770 static struct trace_key trace_untracked_stats
= TRACE_KEY_INIT(UNTRACKED_STATS
);
2772 if (force_untracked_cache
< 0)
2773 force_untracked_cache
=
2774 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2775 trace_printf_key(&trace_untracked_stats
,
2776 "node creation: %u\n"
2777 "gitignore invalidation: %u\n"
2778 "directory invalidation: %u\n"
2780 dir
->untracked
->dir_created
,
2781 dir
->untracked
->gitignore_invalidated
,
2782 dir
->untracked
->dir_invalidated
,
2783 dir
->untracked
->dir_opened
);
2784 if (force_untracked_cache
&&
2785 dir
->untracked
== istate
->untracked
&&
2786 (dir
->untracked
->dir_opened
||
2787 dir
->untracked
->gitignore_invalidated
||
2788 dir
->untracked
->dir_invalidated
))
2789 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2790 if (dir
->untracked
!= istate
->untracked
) {
2791 FREE_AND_NULL(dir
->untracked
);
2797 int file_exists(const char *f
)
2800 return lstat(f
, &sb
) == 0;
2803 int repo_file_exists(struct repository
*repo
, const char *path
)
2805 if (repo
!= the_repository
)
2806 BUG("do not know how to check file existence in arbitrary repo");
2808 return file_exists(path
);
2811 static int cmp_icase(char a
, char b
)
2816 return toupper(a
) - toupper(b
);
2821 * Given two normalized paths (a trailing slash is ok), if subdir is
2822 * outside dir, return -1. Otherwise return the offset in subdir that
2823 * can be used as relative path to dir.
2825 int dir_inside_of(const char *subdir
, const char *dir
)
2829 assert(dir
&& subdir
&& *dir
&& *subdir
);
2831 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
2837 /* hel[p]/me vs hel[l]/yeah */
2838 if (*dir
&& *subdir
)
2842 return !*dir
? offset
: -1; /* same dir */
2844 /* foo/[b]ar vs foo/[] */
2845 if (is_dir_sep(dir
[-1]))
2846 return is_dir_sep(subdir
[-1]) ? offset
: -1;
2848 /* foo[/]bar vs foo[] */
2849 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
2852 int is_inside_dir(const char *dir
)
2861 rc
= (dir_inside_of(cwd
, dir
) >= 0);
2866 int is_empty_dir(const char *path
)
2868 DIR *dir
= opendir(path
);
2875 while ((e
= readdir(dir
)) != NULL
)
2876 if (!is_dot_or_dotdot(e
->d_name
)) {
2885 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
2889 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
2890 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
2891 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
2892 struct object_id submodule_head
;
2894 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
2895 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
2896 /* Do not descend and nuke a nested git work tree. */
2902 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
2903 dir
= opendir(path
->buf
);
2905 if (errno
== ENOENT
)
2906 return keep_toplevel
? -1 : 0;
2907 else if (errno
== EACCES
&& !keep_toplevel
)
2909 * An empty dir could be removable even if it
2912 return rmdir(path
->buf
);
2916 strbuf_complete(path
, '/');
2919 while ((e
= readdir(dir
)) != NULL
) {
2921 if (is_dot_or_dotdot(e
->d_name
))
2924 strbuf_setlen(path
, len
);
2925 strbuf_addstr(path
, e
->d_name
);
2926 if (lstat(path
->buf
, &st
)) {
2927 if (errno
== ENOENT
)
2929 * file disappeared, which is what we
2934 } else if (S_ISDIR(st
.st_mode
)) {
2935 if (!remove_dir_recurse(path
, flag
, &kept_down
))
2936 continue; /* happy */
2937 } else if (!only_empty
&&
2938 (!unlink(path
->buf
) || errno
== ENOENT
)) {
2939 continue; /* happy, too */
2942 /* path too long, stat fails, or non-directory still exists */
2948 strbuf_setlen(path
, original_len
);
2949 if (!ret
&& !keep_toplevel
&& !kept_down
)
2950 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
2953 * report the uplevel that it is not an error that we
2954 * did not rmdir() our directory.
2960 int remove_dir_recursively(struct strbuf
*path
, int flag
)
2962 return remove_dir_recurse(path
, flag
, NULL
);
2965 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
2967 void setup_standard_excludes(struct dir_struct
*dir
)
2969 dir
->exclude_per_dir
= ".gitignore";
2971 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2973 excludes_file
= xdg_config_home("ignore");
2974 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
2975 add_patterns_from_file_1(dir
, excludes_file
,
2976 dir
->untracked
? &dir
->ss_excludes_file
: NULL
);
2978 /* per repository user preference */
2979 if (startup_info
->have_repository
) {
2980 const char *path
= git_path_info_exclude();
2981 if (!access_or_warn(path
, R_OK
, 0))
2982 add_patterns_from_file_1(dir
, path
,
2983 dir
->untracked
? &dir
->ss_info_exclude
: NULL
);
2987 int remove_path(const char *name
)
2991 if (unlink(name
) && !is_missing_file_error(errno
))
2994 slash
= strrchr(name
, '/');
2996 char *dirs
= xstrdup(name
);
2997 slash
= dirs
+ (slash
- name
);
3000 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3007 * Frees memory within dir which was allocated for exclude lists and
3008 * the exclude_stack. Does not free dir itself.
3010 void clear_directory(struct dir_struct
*dir
)
3013 struct exclude_list_group
*group
;
3014 struct pattern_list
*pl
;
3015 struct exclude_stack
*stk
;
3017 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3018 group
= &dir
->exclude_list_group
[i
];
3019 for (j
= 0; j
< group
->nr
; j
++) {
3022 free((char *)pl
->src
);
3023 clear_pattern_list(pl
);
3028 stk
= dir
->exclude_stack
;
3030 struct exclude_stack
*prev
= stk
->prev
;
3034 strbuf_release(&dir
->basebuf
);
3037 struct ondisk_untracked_cache
{
3038 struct stat_data info_exclude_stat
;
3039 struct stat_data excludes_file_stat
;
3043 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3046 int index
; /* number of written untracked_cache_dir */
3047 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3048 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3049 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3051 struct strbuf sb_stat
;
3052 struct strbuf sb_sha1
;
3055 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3057 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3058 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3059 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3060 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3061 to
->sd_dev
= htonl(from
->sd_dev
);
3062 to
->sd_ino
= htonl(from
->sd_ino
);
3063 to
->sd_uid
= htonl(from
->sd_uid
);
3064 to
->sd_gid
= htonl(from
->sd_gid
);
3065 to
->sd_size
= htonl(from
->sd_size
);
3068 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3069 struct write_data
*wd
)
3071 struct stat_data stat_data
;
3072 struct strbuf
*out
= &wd
->out
;
3073 unsigned char intbuf
[16];
3074 unsigned int intlen
, value
;
3075 int i
= wd
->index
++;
3078 * untracked_nr should be reset whenever valid is clear, but
3081 if (!untracked
->valid
) {
3082 untracked
->untracked_nr
= 0;
3083 untracked
->check_only
= 0;
3086 if (untracked
->check_only
)
3087 ewah_set(wd
->check_only
, i
);
3088 if (untracked
->valid
) {
3089 ewah_set(wd
->valid
, i
);
3090 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3091 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3093 if (!is_null_oid(&untracked
->exclude_oid
)) {
3094 ewah_set(wd
->sha1_valid
, i
);
3095 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3096 the_hash_algo
->rawsz
);
3099 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3100 strbuf_add(out
, intbuf
, intlen
);
3102 /* skip non-recurse directories */
3103 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3104 if (untracked
->dirs
[i
]->recurse
)
3106 intlen
= encode_varint(value
, intbuf
);
3107 strbuf_add(out
, intbuf
, intlen
);
3109 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3111 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3112 strbuf_add(out
, untracked
->untracked
[i
],
3113 strlen(untracked
->untracked
[i
]) + 1);
3115 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3116 if (untracked
->dirs
[i
]->recurse
)
3117 write_one_dir(untracked
->dirs
[i
], wd
);
3120 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3122 struct ondisk_untracked_cache
*ouc
;
3123 struct write_data wd
;
3124 unsigned char varbuf
[16];
3126 const unsigned hashsz
= the_hash_algo
->rawsz
;
3128 ouc
= xcalloc(1, sizeof(*ouc
));
3129 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3130 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3131 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3133 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3134 strbuf_add(out
, varbuf
, varint_len
);
3135 strbuf_addbuf(out
, &untracked
->ident
);
3137 strbuf_add(out
, ouc
, sizeof(*ouc
));
3138 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3139 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3140 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3143 if (!untracked
->root
) {
3144 varint_len
= encode_varint(0, varbuf
);
3145 strbuf_add(out
, varbuf
, varint_len
);
3150 wd
.check_only
= ewah_new();
3151 wd
.valid
= ewah_new();
3152 wd
.sha1_valid
= ewah_new();
3153 strbuf_init(&wd
.out
, 1024);
3154 strbuf_init(&wd
.sb_stat
, 1024);
3155 strbuf_init(&wd
.sb_sha1
, 1024);
3156 write_one_dir(untracked
->root
, &wd
);
3158 varint_len
= encode_varint(wd
.index
, varbuf
);
3159 strbuf_add(out
, varbuf
, varint_len
);
3160 strbuf_addbuf(out
, &wd
.out
);
3161 ewah_serialize_strbuf(wd
.valid
, out
);
3162 ewah_serialize_strbuf(wd
.check_only
, out
);
3163 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3164 strbuf_addbuf(out
, &wd
.sb_stat
);
3165 strbuf_addbuf(out
, &wd
.sb_sha1
);
3166 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3168 ewah_free(wd
.valid
);
3169 ewah_free(wd
.check_only
);
3170 ewah_free(wd
.sha1_valid
);
3171 strbuf_release(&wd
.out
);
3172 strbuf_release(&wd
.sb_stat
);
3173 strbuf_release(&wd
.sb_sha1
);
3176 static void free_untracked(struct untracked_cache_dir
*ucd
)
3181 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3182 free_untracked(ucd
->dirs
[i
]);
3183 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3184 free(ucd
->untracked
[i
]);
3185 free(ucd
->untracked
);
3190 void free_untracked_cache(struct untracked_cache
*uc
)
3193 free_untracked(uc
->root
);
3199 struct untracked_cache_dir
**ucd
;
3200 struct ewah_bitmap
*check_only
;
3201 struct ewah_bitmap
*valid
;
3202 struct ewah_bitmap
*sha1_valid
;
3203 const unsigned char *data
;
3204 const unsigned char *end
;
3207 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3209 memcpy(to
, data
, sizeof(*to
));
3210 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3211 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3212 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3213 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3214 to
->sd_dev
= ntohl(to
->sd_dev
);
3215 to
->sd_ino
= ntohl(to
->sd_ino
);
3216 to
->sd_uid
= ntohl(to
->sd_uid
);
3217 to
->sd_gid
= ntohl(to
->sd_gid
);
3218 to
->sd_size
= ntohl(to
->sd_size
);
3221 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3222 struct read_data
*rd
)
3224 struct untracked_cache_dir ud
, *untracked
;
3225 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3226 const unsigned char *eos
;
3230 memset(&ud
, 0, sizeof(ud
));
3232 value
= decode_varint(&data
);
3236 ud
.untracked_alloc
= value
;
3237 ud
.untracked_nr
= value
;
3238 if (ud
.untracked_nr
)
3239 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3241 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3244 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3246 eos
= memchr(data
, '\0', end
- data
);
3247 if (!eos
|| eos
== end
)
3250 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3251 memcpy(untracked
, &ud
, sizeof(ud
));
3252 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3255 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3256 eos
= memchr(data
, '\0', end
- data
);
3257 if (!eos
|| eos
== end
)
3259 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3263 rd
->ucd
[rd
->index
++] = untracked
;
3266 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3267 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3273 static void set_check_only(size_t pos
, void *cb
)
3275 struct read_data
*rd
= cb
;
3276 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3280 static void read_stat(size_t pos
, void *cb
)
3282 struct read_data
*rd
= cb
;
3283 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3284 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3285 rd
->data
= rd
->end
+ 1;
3288 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3289 rd
->data
+= sizeof(struct stat_data
);
3293 static void read_oid(size_t pos
, void *cb
)
3295 struct read_data
*rd
= cb
;
3296 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3297 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3298 rd
->data
= rd
->end
+ 1;
3301 hashcpy(ud
->exclude_oid
.hash
, rd
->data
);
3302 rd
->data
+= the_hash_algo
->rawsz
;
3305 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3306 const unsigned char *sha1
)
3308 stat_data_from_disk(&oid_stat
->stat
, data
);
3309 hashcpy(oid_stat
->oid
.hash
, sha1
);
3310 oid_stat
->valid
= 1;
3313 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3315 struct untracked_cache
*uc
;
3316 struct read_data rd
;
3317 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3321 const char *exclude_per_dir
;
3322 const unsigned hashsz
= the_hash_algo
->rawsz
;
3323 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3324 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3326 if (sz
<= 1 || end
[-1] != '\0')
3330 ident_len
= decode_varint(&next
);
3331 if (next
+ ident_len
> end
)
3333 ident
= (const char *)next
;
3336 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3339 uc
= xcalloc(1, sizeof(*uc
));
3340 strbuf_init(&uc
->ident
, ident_len
);
3341 strbuf_add(&uc
->ident
, ident
, ident_len
);
3342 load_oid_stat(&uc
->ss_info_exclude
,
3343 next
+ ouc_offset(info_exclude_stat
),
3345 load_oid_stat(&uc
->ss_excludes_file
,
3346 next
+ ouc_offset(excludes_file_stat
),
3347 next
+ offset
+ hashsz
);
3348 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3349 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3350 uc
->exclude_per_dir
= xstrdup(exclude_per_dir
);
3351 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3352 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3356 len
= decode_varint(&next
);
3357 if (next
> end
|| len
== 0)
3360 rd
.valid
= ewah_new();
3361 rd
.check_only
= ewah_new();
3362 rd
.sha1_valid
= ewah_new();
3366 ALLOC_ARRAY(rd
.ucd
, len
);
3368 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3372 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3377 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3382 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3386 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3387 rd
.data
= next
+ len
;
3388 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3389 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3394 ewah_free(rd
.valid
);
3395 ewah_free(rd
.check_only
);
3396 ewah_free(rd
.sha1_valid
);
3399 free_untracked_cache(uc
);
3405 static void invalidate_one_directory(struct untracked_cache
*uc
,
3406 struct untracked_cache_dir
*ucd
)
3408 uc
->dir_invalidated
++;
3410 ucd
->untracked_nr
= 0;
3414 * Normally when an entry is added or removed from a directory,
3415 * invalidating that directory is enough. No need to touch its
3416 * ancestors. When a directory is shown as "foo/bar/" in git-status
3417 * however, deleting or adding an entry may have cascading effect.
3419 * Say the "foo/bar/file" has become untracked, we need to tell the
3420 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3421 * directory any more (because "bar" is managed by foo as an untracked
3424 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3425 * was the last untracked entry in the entire "foo", we should show
3426 * "foo/" instead. Which means we have to invalidate past "bar" up to
3429 * This function traverses all directories from root to leaf. If there
3430 * is a chance of one of the above cases happening, we invalidate back
3431 * to root. Otherwise we just invalidate the leaf. There may be a more
3432 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3433 * detect these cases and avoid unnecessary invalidation, for example,
3434 * checking for the untracked entry named "bar/" in "foo", but for now
3435 * stick to something safe and simple.
3437 static int invalidate_one_component(struct untracked_cache
*uc
,
3438 struct untracked_cache_dir
*dir
,
3439 const char *path
, int len
)
3441 const char *rest
= strchr(path
, '/');
3444 int component_len
= rest
- path
;
3445 struct untracked_cache_dir
*d
=
3446 lookup_untracked(uc
, dir
, path
, component_len
);
3448 invalidate_one_component(uc
, d
, rest
+ 1,
3449 len
- (component_len
+ 1));
3451 invalidate_one_directory(uc
, dir
);
3455 invalidate_one_directory(uc
, dir
);
3456 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3459 void untracked_cache_invalidate_path(struct index_state
*istate
,
3460 const char *path
, int safe_path
)
3462 if (!istate
->untracked
|| !istate
->untracked
->root
)
3464 if (!safe_path
&& !verify_path(path
, 0))
3466 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3467 path
, strlen(path
));
3470 void untracked_cache_remove_from_index(struct index_state
*istate
,
3473 untracked_cache_invalidate_path(istate
, path
, 1);
3476 void untracked_cache_add_to_index(struct index_state
*istate
,
3479 untracked_cache_invalidate_path(istate
, path
, 1);
3482 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3483 const char *sub_gitdir
)
3486 struct repository subrepo
;
3487 struct strbuf sub_wt
= STRBUF_INIT
;
3488 struct strbuf sub_gd
= STRBUF_INIT
;
3490 const struct submodule
*sub
;
3492 /* If the submodule has no working tree, we can ignore it. */
3493 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3496 if (repo_read_index(&subrepo
) < 0)
3497 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3499 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3500 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3502 if (!S_ISGITLINK(ce
->ce_mode
))
3505 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3506 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3508 * Skip entries with the same name in different stages
3509 * to make sure an entry is returned only once.
3513 sub
= submodule_from_path(&subrepo
, &null_oid
, ce
->name
);
3514 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3515 /* .gitmodules broken or inactive sub */
3518 strbuf_reset(&sub_wt
);
3519 strbuf_reset(&sub_gd
);
3520 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3521 strbuf_addf(&sub_gd
, "%s/modules/%s", sub_gitdir
, sub
->name
);
3523 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3525 strbuf_release(&sub_wt
);
3526 strbuf_release(&sub_gd
);
3527 repo_clear(&subrepo
);
3530 void connect_work_tree_and_git_dir(const char *work_tree_
,
3531 const char *git_dir_
,
3532 int recurse_into_nested
)
3534 struct strbuf gitfile_sb
= STRBUF_INIT
;
3535 struct strbuf cfg_sb
= STRBUF_INIT
;
3536 struct strbuf rel_path
= STRBUF_INIT
;
3537 char *git_dir
, *work_tree
;
3539 /* Prepare .git file */
3540 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3541 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3542 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3544 /* Prepare config file */
3545 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3546 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3547 die(_("could not create directories for %s"), cfg_sb
.buf
);
3549 git_dir
= real_pathdup(git_dir_
, 1);
3550 work_tree
= real_pathdup(work_tree_
, 1);
3552 /* Write .git file */
3553 write_file(gitfile_sb
.buf
, "gitdir: %s",
3554 relative_path(git_dir
, work_tree
, &rel_path
));
3555 /* Update core.worktree setting */
3556 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3557 relative_path(work_tree
, git_dir
, &rel_path
));
3559 strbuf_release(&gitfile_sb
);
3560 strbuf_release(&cfg_sb
);
3561 strbuf_release(&rel_path
);
3563 if (recurse_into_nested
)
3564 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3571 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3573 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3575 if (rename(old_git_dir
, new_git_dir
) < 0)
3576 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3577 old_git_dir
, new_git_dir
);
3579 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);